Professional Documents
Culture Documents
Ebook - Aversive Training Methods and Papers Ver1
Ebook - Aversive Training Methods and Papers Ver1
1 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Index
THE DEFINITION 4
USE IN APPLIED BEHAVIOR ANALYSIS (ABA) 4
WHAT BELONGS TO THE AVERSIVE METHODS? 6
SENSE BEHIND PUNISHMENT? 7
WHAT PROBLEMS CAN ARISE FROM THIS? 7
FROM AGAINST EACH OTHER 9
TO WITH EACH OTHER!
HOW DID THIS HAPPEN? 12
WHAT IS THE ALTERNATIVE?
DOG TRAINING METHODS AFFECT ATTACHMENT TO THE 12
OWNER
DOES OWNER PERSONALITY AFFECT DOG TRAINING 17
METHODS?
USING AVERSIVES 23
HOW TO RECOGNISE AVERSIVE TRAINING METHODS 30
THE EFFECT AVERSIVE TECHNIQUES CAN HAVE ON 33
ON YOUR DOG
WHY AVERSIVE TRAINING METHODS 34
DON’T WORK IN BEHAVIOR MODIFICATION
HOW NOT TO TRAIN A DOG: WHY AVERSIVES & PUNISHMENT 37
ARE BAD IDEAS
WHAT IS AN “AVERSIVE” AND HOW DOES IT 39
RELATE TO DOG TRAINING?
POSITIVE REINFORCEMENT & NEGATIVE 41
PUNISHMENT: THE PREFERRED DOG TRAINING
OPTIONS
POSITIVE PUNISHMENT: 43
A DOG TRAINING APPROACH TO AVOID
AVERSIVE DOG TRAINING TOOLS: 45
PRONG COLLARS, CHOKE CHAINS & E-COLLARS
WHY SHOULD AVERSIVES BE AVOIDED IN DOG TRAINING? 46
WHY SOME TRAINERS STILL USE E-COLLARS & OTHER 72
AVERSIVE TOOLS
BSAVA STATEMENT REGARDING AVERSIVE TRAINING 79
METHODS
ARE PRONGED COLLARS HARMFUL TO MY DOG? 81
WHAT ARE THE DANGERS OF USING CHOKE AND PRONG 85
COLLARS?
WHAT’S WRONG WITH THE PRONG? 87
1 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
2 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
AVERSIVE
TRAINING
METHODS:
NO, THANKS!
Aversive training methods are based on
punishment and have psychological and physical
consequences for your pet.
3 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
THE DEFINITION
In psychology, aversives are unpleasant stimuli that
induce changes in behavior via negative reinforcement or
positive punishment. By applying an aversive immediately
before or after a behavior the likelihood of the target
behavior occurring in the future is reduced. Aversives can
vary from being slightly unpleasant or irritating to
physically, psychologically and/or emotionally damaging.
4 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
5 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Moxon rope
Halti (in case of improper use)
Yelling
…and anything that frightens or pains your animal.
8 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
10 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
11 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
13 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
These results show that there was a secure base effect for
the dogs that had been trained with reward-based methods.
Comparing these results to previous research on
attachment in dogs, the scientists write, “When dogs had
been trained with reward-based methods, they were more
likely to play when the owner was in the room compared
to when the stranger was in the room. The dogs in the
reward-based group spent more time exploring the room in
the presence of the owner compared to when the stranger
was there. These dogs also engaged in more greeting
behaviours when the owner came back into the room.”
“Together with our results, this suggests it is not the
reward-based training in itself that generates a secure
attachment, but rather the aversive-based training that
may be related to the absence of a secure-base effect.”
There were no differences between the two groups of dogs
in terms of the two other aspects of attachment, separation
distress or maintaining contact with the owner.
14 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
It’s worth noting that there were some order effects in this
study, and some of the significant results only applied to
one of the orders. This is likely because in one order the
dog had one, long separation from the owner, compared to
the other design which had three, shorter separations.
34 dogs took part in the study. Half had attended one of
three reward-based training schools, and half had attended
one of three aversive-based dog training schools, for a
period of at least 2 months.
It makes sense that dogs trained with aversive methods are
more likely to have an insecure attachment, because the
same has been found with children when positive
punishment is used to discipline them.
Although more research is needed, this is a well-designed
study. The data is correlational (and so does not prove
causation), but the authors say it is difficult to do
randomized controlled trials in this area because of ethical
concerns with assigning people’s dogs to aversive training
methods when they are associated with stress.
The authors also point out that different kinds of people
might choose different dog training methods and training
schools, and say that more research on how and why
people make these choices would be welcome. Other
research has found that men with depression are more
likely to use aversive dog training methods.
This study provides more evidence that reward-based
methods are the best way to train dogs, and that aversive
methods may have a detrimental effect on the dog’s
15 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
16 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
17 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
18 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Other studies have also found men are more likely to use
aversive training methods than women, but the link with
moderate depression and dog behavior is surprising. It’s
also hard to interpret. The scientists write,
19 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
20 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
21 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
References
22 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
USING AVERSIVES
If you ask a group of dog trainers to define what an
aversive means, you will get many different answers as
well as some spirited discussion. Regarding behavior
modification techniques, an aversive is ‘the avoidance of
a thing, situation, or behavior that is achieved by using
an unpleasant or punishing stimulus.’ Adding or taking
something away from the dog, situation or environment,
dissuades the dog from repeating an undesired behavior,
but because ‘bad’ behavior means different things to
different people, there is much disagreement on when,
how and if an aversive should be applied.
24 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
25 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
may bark and lunge to try and keep the child farther
away. This reaction to children can then generalize to
other situations. The dog now begins to bark and lunge
at children who visit the home or who he sees out on
walks.
In the Eight Rules for Punishment, police dog trainer
Steve White, gives guidelines for the effective use of
positive punishment, not because he condones it, but
because he shows how difficult it is to use positive
punishment effectively. He explains how temporary
cessation of ‘bad’ behavior is extremely reinforcing to
the person who administers the aversive, at least for that
moment. Unfortunately, people often fail to pay
attention to whether the punishment actually reduces
future occurrences. If the behavior still happens just as
often after punishment is used, the punishment is not
working, no matter how good it makes the person feel at
that moment.
27 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
person who uses a choke chain will tell you that making
your dog wear something around her face is just as
unpleasant. While some dogs do well on headcollars,
others find them very uncomfortable. For this reason I
have stopped using them in favor of the harness, which
seems to be the most effective piece of equipment when
it comes to a dog’s comfort and safety.
If you are unsure about what aversive your dog finds
unpleasant, how do void using them? One of the most
effective ways to stop unwanted behaviors is to teach
your dog alternative ones. If a dog is doing something
that I deem to be unsafe or inappropriate, I will give
teach her new coping/life skills to deal with the
situation. This gives clients the freedom to teach their
dogs’ new skills rather than punishing old ones. For
example, I teach a jumping dog to run and get a toy to
show off to a new person that has just walked into their
home. This gives the dog something to do and diverts
excitable energy onto something that makes everyone
happy. The dog has the toy, the client does not have to
tell their dog off for jumping and the guest is relieved
not to be jumped on.
We can argue with each other about what training
methods and techniques are aversive, but at the end of
the day our dogs don’t lie and will usually tell us if they
find something unpleasant, even if their language is
disregarded or misunderstood. The learner decides what
is rewarding and what is aversive, and it is up to
handlers to use kindness as well as common sense, and
28 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
29 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
31 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
32 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
33 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
This does not mean the behaviour has gone away, and the
fact is, it is likely to return, often worse than it was to start
with. However, in some cases the behaviour doesn’t
return, but instead the dog psychologically suffers because
the initial stress that created the behaviour is still present, it
will just be expressed in a different way.
34 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
36 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Key Takeaways
“Aversive” dog training tools and techniques are those that
use pain or intimidation to achieve the desired result.
Common examples include prong collars, shock collars,
and “alpha” or “dominance-based” training philosophies.
There are a number of reasons to avoid using aversive
training methods. For starters, they’re simply not as
effective as positive-based approaches are, but they’re also
simply disrespectful to your dog, who you ostensibly love.
A subset of modern dog trainers still employ aversive tools
and techniques, but (thankfully) science is helping to turn
this tide. The empirical data on the subject of aversive
37 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
38 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
39 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
40 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
42 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
POSITIVE PUNISHMENT:
A DOG TRAINING APPROACH TO AVOID
43 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
rarely, and only when all other options have been explored
consistently and thoroughly for quite some time.
By contrast — and despite being referred to as “force-
free” or “positive-based” training — many modern dog
trainers use both positive reinforcement and negative
punishment. But these types of trainers will never use
positive punishment – or at least will use it very rarely, and
only when all other options have been explored
consistently and thoroughly for quite some time.
For the purposes of this article, From this point forward,
when I refer to “punishment,” I specifically mean
positive punishment, not negative punishment.
44 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
45 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
46 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
48 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
49 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
50 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
52 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
53 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Your dog learns that you will take him out for a walk
after making your morning coffee, and he ends up
becoming excited and waiting at the door when he
smells you brewing coffee — regardless of the hour.
55 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
56 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
As the bell rang, the dogs would get shocked. There were
no behaviors or actions the dogs could perform to stop the
shocks.
57 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
58 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
59 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Some dogs will completely shut down under this stress and
act more like robots than dogs.
60 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
In fact, I’m not all that convinced that dogs are capable of
respect. At least not the way we think about it.
61 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
62 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
64 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
65 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
66 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
67 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
68 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
That last one is where some owners may struggle the most.
“Why should I have to provide my do with motivation?”
you might ask. “Shouldn’t dogs follow our instruction out
of obedience?”
The idea that dogs are built to obey a human’s whim is a
bit of a pipe dream.
70 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
71 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
72 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
73 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
74 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
75 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
77 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
78 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
79 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
80 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
81 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
82 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
83 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
References
[1] Rooney NJ & Cowan S (2011) Training methods and owner-dog interactions. Links with
dog behaviour and learning ability. Applied Animal Behaviour Science 132: 169-177.
[2] Overall KL (2007): “Editorial: Considerations for Shock and “Training” Collars:
Concerns from and for the Working Dog Community” in Journal of Veterinary Behaviour:
Clinical App Res 2: 103-107.
[3] Blackwell EJ et al (2008) The relationship between training methods and the
occurrence of behaviour problems, as reported by owners, in a population of domestic
dogs. Journal of Veterinary Behavior 3:207-217.
[4] Lefebvre D et al (2007) The quality of the relationship between handler and military
dogs influences efficiency and welfare of dogs. Applied Animal Behaviour Science 104(1–
2): 49–60.
84 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
The most humane and safest option for walking a dog who
tends to want to pull is a front-leash attachment harness,
such as the Sense-ation. When dogs lunge or pull while
wearing the Sense-ation harness, the front leash attachment
85 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
86 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
every dog will like – you just have to try things out to find
what motivates your dog. Baby food? Tripe? Meatballs?
String cheese? Or, some dogs will work for a ball or a
game of tug!
90 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
93 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
94 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
13) Last but not least, aversive based training can inhibit
dogs from offering new behaviors as it may happen with
dogs who become tentative in picking up objects. If a
puppy is consistently punished for picking up objects with
his mouth such as the remote control or shoes, the puppy
95 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
"But I have been using these methods for years, they have
worked for me, why should I ditch them?" This is often a
self-defense mechanism due to fear of something new, it's
the (normal) resistance associated with the hesitancy of
embracing an unexplored world.
96 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
97 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
Scientific Studies
and References
Numerous scientific studies have shown the
harmful effects of aversive training methods and
equipment. Below you’ll find some of the most
important documents to date.
98 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
99 - The Evidence
AVERSIVE TRAINING METHODS: NO, THANKS!
If you have an aggressive dog, you may have heard about the purely positive vs
balanced dog training debate. In this video, I break down why purely positive dog
training methods cannot fix dog aggression. And neither can balanced dog training.
Watch this video for helpful dog aggression and aggressive dog training tips and
tricks.
Do you have a dog dealing with anxiety? Is your dog reactive to dogs, people, or
sounds? Maybe your dog is dealing with aggression. Then you might be the right fit
for my signature program, The Recovering Rover Program. In the RRP, I coach you
on how to treat your dog’s mental health. To find out more, go to
https://www.getacalmdog.com/
Join my Email list for even MORE free training tips, promos about the RRP, and to be
in the Dog Liaison community. https://mailchi.mp/8fc2973c78eb/join-... Follow me on
Instagram @dog_liaison https://www.instagram.com/dog_liaison/ Like me on
Facebook https://www.facebook.com/dogliaison/ Make sure to subscribe and hit the
notification bell on Youtube! Disclaimers: Jenna Romano is not a veterinarian.
Always consult a veterinarian before giving your dog any food. Use of this information
is strictly at your own risk. Jenna Romano will not assume any liability for direct or
indirect losses or damages that may result from the use of information contained in
this video including but not limited to economic loss, injury, illness or death to you or
your dog. This video is not instructional. This video is for entertainment purposes
only.
a r t i c l e i n f o a b s t r a c t
Article history: The methods by which owners train their pet dogs range widely, with some exclusively
Accepted 2 March 2011 using rewards, and others using a combination, or only punishment-based methods. This
Available online 21 April 2011
paper examines links between the way in which owners reported to have trained their dogs
and observations of the dogs’ subsequent behaviour. It also explores associations between
Keywords: behaviour of owner and dog when tested in their own home. A total of 53 owners were
Domestic dog
surveyed about their preferred methods for training each of seven common tasks, and were
Training methods
each filmed interacting with their dog in a series of standardised scenarios. Dogs owned by
Obedience
Learning ability subjects who reported using a higher proportion of punishment were less likely to interact
Punishment with a stranger, and those dogs whose owners favoured physical punishment tended to be
Reward less playful. However, dogs whose owners reported using more rewards tended to perform
better in a novel training task. Ability at this novel task was also higher in dogs belonging to
owners who were seen to be more playful and who employed a patient approach to training.
This study shows clear links between a dog’s current behaviour and its owner’s reported
training history as well as the owner’s present behaviour. High levels of punishment may
thus have adverse effects upon a dog’s behaviour whilst reward based training may improve
a dog’s subsequent ability to learn.
© 2011 Elsevier B.V. All rights reserved.
0168-1591/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.applanim.2011.03.007
170 N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177
ishing rather than rewarding behaviour in their handlers son and owner, response to being ignored, obedience to
or owners. But the results do suggest that frequent use basic commands, dog–owner play and response to owner
of punishment-based training does not result in superior training a novel task.
training performance. Within dog–owner partnerships, interactions are
In addition to obedience and training performance, numerous and varied with structured training repre-
numerous other aspects of the dog’s behaviour and of senting only a minority of the daily interaction time.
the dog–owner relationship, have been linked to choice Therefore, it is probable that many subtle aspects of the
of training methods. Punishment-based training has been owner’s behaviour affect the dog’s response to training,
shown to cause stress (Schalke et al., 2007), suffering as well as the training approach employed. For example
(Beerda et al., 1997), fearfulness (Schilder and van der previous studies have found links between the way in
Borg, 2004; Blackwell and Casey, 2006), to be associated which owners play and dimensions of the dog–owner rela-
with higher levels of reported behavioural problems (Hiby tionship (Rooney and Bradshaw, 2003). Therefore, we have
et al., 2004), specifically aggression to dogs (Haverbeke measured variations in the owners’ behaviour as well as
et al., 2008; Arhant et al., 2010) and to people (Arhant their training techniques and investigated whether these
et al., 2010), increased excitability (Arhant et al., 2010) correlate to aspects of the dogs’ behaviour, specifically its
and distraction (Haverbeke et al., 2008). A survey by obedience score and its response to learning a new task.
Blackwell et al. (2008) saw lower levels of several spe- Our study aims to investigate the following hypotheses:
cific behavioural problems in those dogs trained without
punishment. (a) Dogs reported to be trained using different methods
The majority of these previous studies of training meth- exhibit measurable differences in behaviour in a stan-
ods have relied upon the owners’ reports of their dog’s dardised test.
behaviour. Owners’ reports of certain aspects of their dog’s (b) Measurable differences in owner behaviour correlate
behaviour can be unreliable (e.g. Rooney, 1999; Stephen to differences in dog behaviour; specifically obedience
and Ledger, 2007). Therefore to further investigate the and performance at a novel task.
postulated links, we have carried out a study comparing
training methods and dog behaviour, via direct observa-
2. Methods
tion.
We predict that in a home environment the type
2.1. Subjects
of training methods employed may affect the way in
which a dog behaves in numerous situations. For example
A total of 53 volunteer dog owners (35 women and
punishment-based training methods may lead to general
18 men), ranging in age from 12 to 73 years (median = 46
anxiety (Blackwell et al., 2008) which may affect a dog’s
years) were recruited through veterinary practices and
social behaviour towards both its owner and towards other
personal communication in Glasgow and Bristol. All par-
people, or its response to denial of contact or attention
ticipants were the main owners, who had the majority
from its owner. Training history may also affect the way
of responsibility for the care of the dog and all had
a dog plays with its owner, since play behaviour is usu-
owned their dog for at least three months. If participants
ally only performed when conditions are good (e.g. Jensen
owned more than one dog, they randomly selected one
et al., 1998) and an animal is relaxed (Berman, 1980) and
to take part. There were 19 male and 34 female dogs,
aspects of play likely reflect the quality of the dog–owner
ranging from five months to 14 years of age (median = 4
relationship (Rooney and Bradshaw, 2003). The way in
years). A total of 16 were crossbreds and 37 were pure-
which a dog was trained in the past may also affect its
breds, and when categorised according to the UK Kennel
future aptitude and motivation to learn (e.g. Marshall-
Club categories (see ‘Discover Dogs’ at http://www.the-
Pescini et al., 2008) and therefore its performance at novel
kennel-club.org.uk); there were 24 gundogs, six pastoral
training tasks, as well as its general obedience. To our
(traditionally herding breeds), five terriers, one hound and
knowledge, these ideas have never been investigated in pet
one utility breed.
dogs.
Using a modified version of the questionnaire developed
by Hiby et al. (2004) and a sample of established dog–owner 2.2. Protocol
partnerships, we examined owners’ approaches to train-
ing. We quantified the proportion of tasks for which the To avoid the effects of a novel environment, which may
owner reported using reward, punishment and whether vary with the dog’s individual level of socialisation, the
they used physical punishment. We then recorded the experimenter (Sarah Cowan) visited each owner and dog
behaviour of the dog and the owner in each of five stan- in their own home. All the data was collected during a sin-
dardised test scenarios and compared this to the training gle visit, when owner, dog and experimenter were alone
methods advocated by their owner. The behaviour test in an enclosed room; the owner’s living or sitting room
used in this study was a modified version of that used by in which two chairs were placed at least 2 m apart. Vis-
Rooney and Bradshaw (2002, 2003) to quantify aspects of its lasted between 19.3 min and 39.4 min (mean = 27.2 min)
the dog–owner relationship, but here shortened to only and were filmed throughout. If other family members or
examine the key interactions hypothesised to be impor- dogs were at home, they were requested not to enter the
tant with the addition of a novel training task. The scenarios room. Owners were asked to remove any toys from the
used were: relaxed social behaviour towards a novel per- room prior to the experimenter’s arrival.
N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177 171
Upon arrival, the experimenter placed a video camera and a bag of 20 g of small, hard treats (Coachies, Com-
on a tripod (maximising the floor area recordable), com- pany of Animals), and could use either, neither or both of
menced filming, and then sat on a chair opposite to the these rewards. They were given 5 min to attempt to train
owner. She explained the protocol, before commencing the dog to perform the task. The experimenter stressed
the behaviour test, which was composed of five separate that there was no expectation that the dog would per-
subtests, always conducted in the same order (Sections form perfectly by the end of the time, and the purpose
2.2.1–2.2.5). Owners were instructed before each subtest of the exercise was to investigate the variety of methods
was undertaken and these instructions were repeated if that people use to train dogs, and the different ways dogs
required. learn.
2.2.1. Subtest A: relaxed social behaviour 2.2.6. Reported Training Methods (RTM) interview
This commenced when the experimenter first sat down. The interview was carried out at the same time as
If the dog approached the experimenter, it was briefly pet- Subtest A, and was a modified version of the written ques-
ted. The owner was instructed not to encourage the dog to tionnaire used by Hiby et al. (2004). Owners were asked
approach but to act naturally if it approached of its own 11 questions which included; four general questions (dog’s
accord. The experimenter then conducted the Reported name, dog’s age, where the dog was obtained from and
Training Method interview (see Section 2.2.6). Owners how long they had owned the dog) and then how they
were variable in the time they took to complete this inter- had trained their dog in each of seven common training
view, ranging from 2.8 to 9.6 min (mean = 5.5 min), so to scenarios:
standardise habituation effects, only the first 2 min of this
subtest was analysed. (a) toilet training,
(b) to sit on command,
2.2.2. Subtest B: ignoring (c) to recall,
The owner was asked to ignore the dog and the exper- (d) to leave objects,
imenter did the same. The owner was given two written (e) to walk to heel,
questionnaires to complete, one examining their attitudes (f) when the dog had stolen objects,
towards dogs and the second investigating their responses (g) when the dog had chewed things.
to pictures of dog behaviour. These were later analysed
to explore the effects of attitudes on owner behaviour, In the case of dogs that were already trained to per-
but the results are reported elsewhere and do not form form one or more of the behaviours before the owner
part of this paper. This subtest lasted between 5.2 and obtained them, owners were asked how they would choose
15.2 min (mean = 10.6 min) so for standardisation, only the to train a dog for that specific task. Owners were asked to
first 5 min were analysed. highlight the main method used and their responses were
transcribed by the experimenter.
2.2.3. Subtest C: obedience
The owner was instructed to ask the dog to “Sit”, to “Lie 2.3. Behavioural analysis and variable generation
Down” and then to “Stay”. The owner then took five steps
backwards and remained facing the dog until it rose or for 2.3.1. Owner behaviour
a maximum of 30 s. This component was repeated once. It From the video recorded behaviour test, eleven vari-
was stressed that the command “Stay” must only be given ables were measured which described the owners’
once and it was not important whether the dog obeyed or behaviour throughout three subtests (Table 1). Two of
not. If the dog was still lying down after 30 s it was called the initial variables were rejected: the frequency of con-
to the owner. tact (subtest A) was observed to be determined mainly
by the dog so was included as a dog behaviour variable
2.2.4. Subtest D: non-object play (Table 2), and the proportion of positive vocalisations (sub-
The owner was asked to play with their dog without any test E) was highly correlated to the proportion of negative
toys for 2 min. Owners were encouraged to play as they vocalisations (Rho = −0.77, P < 0.001), with neutral vocali-
would normally, starting and ending the session in their sations being very rare. The remaining nine variables each
usual way. showed considerable variation across the sample. They
were inspected for normality and three were transformed
2.2.5. Subtest E: novel training task to improve data spread.
We designed a training task which was unique to all
dog–owner partnerships. The owner was presented with 2.3.2. Dog behaviour
a 29 cm (width) × 25 cm (depth) × 14 cm (height) wooden A total of 12 variables were measured to describe the
block in which there were mounted two 14 cm high dogs’ behaviour during the five subtests. Two variables
wooden spoons. One of the spoons; the target, had a black were deemed too difficult to score consistently between
cross marked on it in ink, whereas the other spoon had breeds (tail and ear positions) and one was found to be
an open circle. The owner was asked to start to train the highly correlated to another variable (Involvement in play
dog to respond to the command “touch it” by touching (scale 1–5) which correlated to duration of play, Spear-
the correct spoon (the one with a cross) and not the non- man Rank Correlation: Rho = 0.77; P < 0.001), so these were
target spoon, with its nose. Owners were offered a ball eliminated leaving nine variables. The spread of each of
172 N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177
Table 1
Variables describing the owner’s behaviour during three of the subtests plus transformations necessary to improve data spread.
A. Relaxed Command Number of times owner calls, or orders dog Frequency Scale: 0–2,
social frequency 0 = 0, 1 = <5 and
behaviour 2 = ≥5
Total play Number of different categories of play signal Frequency
signal types handler uses to initiate and maintain play
(Rooney et al., 2001)
D. Non-object Rough-and- Handler plays with dog any game involving 0/1
play tumble wrestling and high levels of contact between
the players (Rooney and Bradshaw, 2003)
Owner Subjective rating of the extent to which the Scale: 1–5:
involvement owner was involved or committed to playing 1 = least
(Rooney and Bradshaw, 2003) committed–5 = most
committed
Total rewards Number of times owner delivers either ball or Frequency
treat to the dog
Patience Subjective rating of level of tolerance exhibited Scale 1–5:
by the owner over the 5 min session 1 = least
patient–5 = most
patient
E. Novel Positive Proportion of times owner touches the dog in Proportion Scale: 0 = 0,
training task contacts an affectionate way (petting, patting, stroking) 1 = <50, 2 = ≥50
as a proportion of the total number of physical and 3 = 100%
contacts made
Negative Proportion of times owner touches the dog in a Proportion Scale 0 = 0,
contacts potentially aversive way (shoving, hitting, 1 = <50, 2 = ≥50
tapping, moving or restraining) as a proportion and 3 = 100%
of the total number of physical contacts made
Negative Proportion of the total verbalisations which Proportion
vocalisations were intended to rebuke or command the dog
these variables was inspected and those which were not 2.4. Statistical analysis
normally distributed were transformed to improve data
spread, either by conversion to an ordinal scale or to 0/1 All statistical analyses were carried out using SPSS
categories (Table 2). 12.0 for Windows (© SPSS Inc.). Many of the variables col-
lected were not normally distributed so nonparametric
2.3.3. Reported Training Methods (RTM) tests were employed throughout. When testing the rela-
Owners’ responses to each question in the RTM inter- tionship between two ordinal variables Spearman’s Rank
view were classified as reward-based, miscellaneous, or Correlation Tests were used. Mann–Whitney U-tests were
punishment-based. We used the categorisation system of used to test the relationship between an ordinal and binary
Hiby et al. (2004), as a basis although there were also some variable whilst Chi-Square Tests were used when testing
answers given in this study, not previously listed by Hiby two binary variables.
et al. (2004; Table 3). For each owner, the total number of
times they mentioned using each method type was calcu- 2.4.1. Demographics
lated, and since some owners mentioned more than one We compared the male and female respondents for
method for each training scenario, the proportion of each their reported training methods, and for each of the owner
method was calculated (proportion reward-based, pro- behaviour variables. Similarly male and female dogs were
portion punishment-based and proportion miscellaneous). compared for each of the dog behaviour variables.
Proportion miscellaneous has little biological significance
and so was not analysed any further. Proportion reward- 2.4.2. Hypothesis testing
based and punishment-based showed only a moderate 2.4.2.1. Comparing training history and current dog
correlation to one another (Spearman Rank Correlation: behaviour. We explored the associations between both
Rho = −0.293), so both were retained for future analy- the proportion of reported training methods variables
sis. Some owners reported using methods which involved which were reward-based, and the proportion which were
physical punishment, and we hypothesised that these may punishment-based and each of the nine variables describ-
have different consequences to verbal punishment or aver- ing dog behaviour. We also compared the behaviour
sive stimuli and that could be avoided by the dog (e.g. bad of those dogs which were every trained using physical
tasting substances used to prevent chewing of furniture). punishment to those which were not, for each behavioural
Therefore we recorded an additional categorical variable variable in turn.
which described whether each owner ever used any of
the techniques described as “physical punishment” (see 2.4.2.2. Comparing owner behaviour and dog training per-
Table 3). formance. Next we explored links between each of the nine
N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177 173
Table 2
Variables describing the dog’s behaviour during each of the five subtests plus transformations necessary to improve data spread.
D. Non-object Play duration Time out of 2 min session for which Duration: s Scale: 0 = 0,
play dog is interacting with handler in a 1 = <60 s,
playful manner (for definition of 2 = ≥60 s,
play see Rooney, 1999) 3 = 120 s;
Interactivity Subjective rating of the degree of Scale: 1–5 (5 = most
reciprocity between the players reciprocal–1 = least reciprocal;
during the 2 min session Rooney and Bradshaw, 2003)
E. Novel Dog’s ability Rating of the extent to which the Scale: 0 = no interest in spoon,
training task dog had learnt the task as 1 = interest in correct spoon,
demonstrated by response to 2 = occasionally touching correct
handler’s commands during the spoon, 3 = usually touching correct
final 30 s spoon
All subtests Submissive Dog ever displays any of the Frequency 0/1
behaviour following behaviour(s): Lip licking,
cringing, rolling over, gaze
aversion, submissive grin, tail
tucked, ears back, licks owner or
lifts paw
variables describing owner behaviour in the behaviour test, behaviour than were males (Chi-squared test, 2 = 6.058,
and the dog’s C Total obedience score and E Dog’s ability in P = 0.02); no male dogs showed submissive behaviours,
the novel training task. whilst nine out of the 34 (26%) female dogs were recorded
to exhibit one or more submissive behaviour.
3. Results
3.2. Reported training methods
Table 4 shows the spread of the owner and dog
behaviour variables in the sample. None of the 53 owners reported using exclusively
reward-based methods or entirely punishment; all used
3.1. Demographics a combination. The proportion of tasks which own-
ers trained using reward-based methods ranged from
None of the owner behaviour variables, or the reported 0 to 0.64 (median = 0.22; 25th percentile = 0.09; 75th
training methods varied with gender of owner. When com- percentile = 0.44) whilst those using punishment-based
paring dog behaviour variables, one significant difference methods ranged from 0 to 0.6 (median = 0.25; 25th per-
was found; females were more likely to show submissive centile = 0.013; 75th percentile = 0.35). A total of 38% of
174 N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177
Table 3
Examples of categorisation of reported training techniques into four method types.
Punishment-based
Just classified as punishment-based Denial of social contact (e.g. put in cage, sent to bed)
Vocal punishment (e.g. tell off)
Removal of pleasurable stimuli (e.g. do not feed, take toy out of mouth)
Aversive (bad tasting food, throwing tin pot on ground)
Table 4
Distribution of variables describing owner and dog behaviour in sample of 53 dog–owner partnerships.
the subjects reported using reward-based training for P = 0.007) and the dog’s ability at the novel training task
more tasks than they used punishment, whilst for 49% (Rho = −0.278, N = 53, P = 0.044). The proportion of positive
of owners, punishment-based methods use outweighed training methods reported by the owner was positively
reward-based. In all, 34% of owners listed a “physical pun- correlated to the dog’s ability at the novel training task
ishment” method for at least one of the seven tasks. (Rho = 0.308, N = 53, P = 0.025). Owners who reported using
physical punishment owned dogs that were less interac-
3.3. Links between reported training methods and dog tive during play (Mann–Whitney U-test, U = 187.5, N = 53,
behaviour P = 0.015, medians = 3.25 vs. 4), and were less likely to con-
tact and interact with the experimenter (U = 196, N = 51,
When comparing the three variables describing train- P = 0.03, medians = 8 vs. 1.5) when compared to dogs whose
ing history and the dog’s behaviour in the test scenario, owners never used these methods.
five significant associations were seen. The proportion
of punishment-based training methods reported by the 3.4. Links between general owner behaviour and dog
owner was negatively associated with experimenter inter- training performance
action level during the relaxed social behaviour subtest
(Spearman’s Rank Correlation Test, Rho = −0.371, N = 51 (2 When testing the relationship between owner
missing data points due to dog out of view on video), behaviour and dog performance at basic obedience
N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177 175
tasks and a novel training task, four significant relation- they are sufficiently relaxed (Berman, 1980). Since play was
ships were found. The dog’s Total Obedience Score for less interactive in partnerships which use some physical
basic commands was positively associated with both the punishment, this may suggest that this method of train-
owner’s Patience when training the novel training task ing is associated with a reduced quality of dog–human
(Rho = 0.317, N = 53, P = 0.021), and to Owner Involvement relationship. This correlation cannot be used as proof that
during play (Rho = 0.316, N = 53, P = 0.021). The Dog’s punishment causes a reduction in interactivity, it may
Ability at the novel training task was positively associated equally be that owners who are less interactive or recipro-
with both the Total Rewards delivered (Rho = 0.468, N = 53, cal in their play with their dog, are also more likely to favour
P < 0.001), and owner’s Patience when training the novel a domineering or punishment-based approach to training.
training task (Rho = 0.300, N = 53, P = 0.029). But the association between play quality and dog–owner
behaviour supports links previously seen (Rooney and
4. Discussion Bradshaw, 2003).
Reported obedience, has previously been shown to be
4.1. Does training history affect current dog behaviour? higher in those dogs trained using more positive rewards
(Hiby et al., 2004) and more consistent training methods
This study is, to our knowledge, the first to examine (Arhant et al., 2010) and performance is higher in work-
associations between empirical measures of dog behaviour ing dogs trained using less aversive stimuli (Haverbeke
and their owner’s approach to training, although a recent et al., 2008). However, in this study, we did not detect a
study has reported similar observations of military dog systematic difference in recorded obedience scores when
handlers and their dogs (Haverbeke et al., 2008). We have dogs with different training histories performed three stan-
found several significant links between the types of meth- dard tasks. This may be because the tasks chosen; sit, lie
ods owners report to have used in the past and their and stay, only measure the dogs’ response to specific com-
dog’s current human-directed behaviour. This is a relatively mands, which could be very context specific, and not fully
small sample size and with multiple statistical testing, representative of a dog’s overall obedience. Past research
one needs to beware that one or more of the significant has suggested that the effects of negative and positive
associations may be due to chance. The subjective rat- training methods are specific to the context in which they
ings for obedience, involvement and interactivity in play, are delivered (Blackwell et al., 2008) and hence it may be
were recorded by the experimenter who had also inter- necessary to examine the training methods used for the
viewed the owners, and hence there remains a risk of precise tasks for which obedience is scored. However, we
unconscious bias. However several of the links support did detect a significant difference in the dogs’ ability at
findings from larger owner-reported data and hence are a novel task; owners who reported training more tasks
likely meaningful, so below we discuss all results signifi- using reward-based methods in the past were more suc-
cant at P < 0.05. cessful when training their dog to perform a new task, in
Those owners who chose to use punishment-based this case to touch a specific wooden spoon. This suggests
methods to train more tasks, tended to have dogs which that a past history of reward-based training increases a
showed lower levels of interaction with a novel tester dog–owner partnerships’ success in future training; pos-
within their home. The dogs reported to be trained using sibly by increasing the dog’s motivation and aptitude to
physical punishment, overall were less interactive when learn, because it learns to anticipate rewards. Dogs trained
playing with their owners as well as less likely to approach using more rewards may be less stressed, and more relaxed
a new person. In contrast, owners who favoured reward- in a training scenario than those trained using inconsis-
based methods for more tasks, and those who used fewer tent or punishment-based methods as has previously been
punishment-based methods, had dogs which scored higher shown in working dogs (Schilder and van der Borg, 2004).
for ability when trained to perform a novel task. Since high levels of stress can reduce an animals’ ability
The association between punishment-based prior train- to learn (e.g. Mendl, 1999), this could explain this asso-
ing and lack of interaction with the novel tester, may be ciation. Once again this result is a correlation, and it is
because punishment can lead a dog to become anxious or impossible to attribute cause and effect. It could equally
fearful towards people, as has previously been seen in dogs be that those dogs which show poor learning ability were
trained using electric collars (Schilder and van der Borg, less likely to instigate a reward-driven training approach
2004), and hence it may be less confident to approach a in their owners, that owners resort to punishment after
novel person within its home. It could also be that pun- reward-based methods have failed, or that owners who
ishment based training leads to an insecure attachment reported reward-based training also behaved subtly differ-
which manifests as “clingy” behaviour towards the owner ently in the training task and thereby enhanced learning.
(see Scott and Fuller, 1965, p. 145) which may explain why Owners who favour this approach may even have selected
those dogs trained using these methods were significantly dogs of different temperaments at the outset, or be gen-
less interactive with a novel person. erally keener or more experienced dog trainers. However,
There was a significant difference in play behaviour even if these alternative explanations are true, this study,
between owner-dog partnerships which reported using shows that a reduction of rewards, or an increase in punish-
physical punishment and those which did not; the use of ment does not lead to enhanced obedience or future ability
these methods was associated with lower levels of inter- to learn, and hence these results support those of Hiby et al.
activity during non-object play. Animals only play when (2004) that higher levels of reward-based methods appear
their basic needs have been met (Jensen et al., 1998) and to be beneficial.
176 N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177
4.2. Do measurable differences in owner behaviour not rule out the possibility that owners predisposed to
correlate to differences in dog behaviour? train in different ways originally selected dogs with dif-
fering temperaments and hence receptiveness to differing
When examining correlations between owner training techniques. Further studies examining behaviour
behaviour, and the dog’s obedience and learning behaviour within non-established dog–human pairing would help to
in a standardised test, several significant associations were elucidate such subtle effects.
seen. Within individual subtests, there were clear asso-
ciations between the behaviour of each of the partners.
For example, owners who scored higher for patience in
4.4. What affects ability to learn?
the training task tended to have dogs which scored higher
for ability. Similarly there was a correlation between
Since the behaviour of neither party was controlled, and
the number of rewards delivered and the dog’s ability
they are inevitably inter-related, one must be careful in
at the task, which would be expected since rewards are
drawing discrete conclusions about either dog or owner
more likely to be delivered following correct responses.
from this study. Significant findings are best interpreted as
However, there were also several interesting correlations
indicative of differences between dog–owner relationships
when comparing between subtests. Those owners who
rather than attributed to be cause and effect.
were more involved in play tended to have dogs which
The performance of the dog–owner partnerships
performed better in the novel training task, and which
when presented with a novel training task was sig-
were more obedient to basic commands. This supports
nificantly related to a number of measured variables.
for the idea of playfulness being indicative of a healthy
Those partnerships which performed the best, tended
relationship (Rooney and Bradshaw, 2003), since owners
to be those with a reported history of training many
who were more playful seemed also to have trained their
tasks using reward-based methods, whilst owners who
dogs better. Playful owners may also use play as a reward
showed high levels of involvement when engaging in
and thus motivate their dogs to learn quicker since play
non-object play and those who showed a high level
has been demonstrated to be an efficient reward for the
of patience and a high reward frequency when train-
training of many tasks (e.g. Rooney et al., 2004). When
ing the task also did well. This supports the popular
interpreting this and the other results however, one must
idea that the best way to build a dog–owner relation-
be aware that due to demographics, the majority of the
ship which encourages effective learning is to adopt a
dogs in this sample were gundogs, whose responses may
reward-based, playful yet patient approach to training
not be typical of all breeds. Since some gundogs breeds are
(e.g. Bailey, 2008).
particularly playful (e.g. Labrador Retrievers; Rooney and
Bradshaw, 2004), such associations may be less important
in a sample of different demographic. It would therefore
be advisable to repeat this study with a sample of different 5. Conclusion
breed types.
Despite a growing trend to use reward-based train-
4.3. Novel training task ing methods, there continue to be a large number of
punishment-based methods in common usage. This study
The training task developed in this study was an effec- empirically compares the behaviour of dogs trained using a
tive way of examining a dog and owners’ response to a variety of training methods. Different reported methods of
novel task, and subtle aspects of their communication dur- training were associated with differing levels of obedience
ing training. There was considerable diversity both in the and behavioural problems in past studies (Hiby et al., 2004;
way the owners behaved during the task and in the subse- Blackwell et al., 2008) and to variations in playfulness, lev-
quent ability of their dogs to learn. The owners’ behaviour els of interaction with a novel person, and ability to learn a
in the novel training task showed links to their dogs’ ability new task, in the current study.
to learn that task, and to its general obedience in previously Although this study uses a relatively small sample of
learnt tasks. The dog’s behaviour in the task was further dogs, similar to past research all the significant relation-
linked to their owners’ involvement in play in an earlier ships seen support the benefits of reward-based training
subtest. methods (Hiby et al., 2004; Blackwell et al., 2008). Whilst
Aspects of owner–dog communication during the train- the proportion of reward-based methods showed signifi-
ing task such as the number of affectionate physical cant associations with benefits such as enhanced ability in
contacts or the frequency of rebuking vocalisations did a new training task; the proportion of punishment-based
not appear to significantly affect the dogs’ ability at the methods only showed associations with potential detri-
task being trained, whilst the owner’s patience and the ments such as reduced interactivity during play and lower
frequency with which they rewarded the dog did corre- levels of interaction with new people. Thus, we conclude
late with the dog’s performance. This suggests that subtle that, in support of the questionnaire-based findings of Hiby
aspects of the owner’s communication when interacting et al. (2004), this observational study suggests that, for dog
may affect their dog’s performance at a novel task, and owners, the use of reward-based training appears to be
in a test scenario these may be better indicators of the the most beneficial for the dog’s welfare, since it is linked
owners’ natural interaction style as they are possibly less to enhanced learning and a balanced healthy dog–owner
affected by the owner being watched or filmed. One can- relationship.
N.J. Rooney, S. Cowan / Applied Animal Behaviour Science 132 (2011) 169–177 177
Acknowledgements Jensen, M.B., Vesergaard, K.S., Krohn, C.C., 1998. Play behaviour in dairy
calves kept in pens the effect of social contact and space allowance.
Appl. Anim. Behav. Sci. 56, 97–108.
We would like to thank all 53 kind dog owners, Marscark, E.D., Baenninger, R., 2002. Modification of instinctive herding
without whom this study would not have been pos- dog behaviour using reinforcement and punishment. Anthrozoos 15,
sible. We are also grateful to John Bradshaw for his 51–68.
Marshall-Pescini, S., Valsecchi, P., Petak, I., Previde, E.P., 2008. Does
help with statistical analysis and for his comments on training make you smarter? The effects of training on dogs’ perfor-
the manuscript, to Andrew Haig-Ferguson for his help mance (Canis familiaris) in a problem solving task. Behav. Process. 78,
in developing the methods. We thank Emily Blackwell 449–454.
Mendl, M., 1999. Performing under pressure: stress and cognitive func-
for great ideas about novel training tasks, and Sam tion. Appl. Anim. Behav. Sci. 65, 221–244.
Gaines and Emily for their helpful feedback on the Rooney, N.J., 1999. Play behaviour of the domestic dog, Canis familiaris and
manuscript. its effects on the dog–human relationship. PhD Thesis. University of
Southampton.
Rooney, N.J., Bradshaw, J.W.S., 2002. An experimental study of the effects
References of play upon the dog–human relationship. Appl. Anim. Behav. Sci. 75,
161–176.
Arhant, C., Bubna-Littitz, H., Bartels, A., Futschik, A., Troxler, J., 2010. Rooney, N.J., Bradshaw, J.W.S., 2003. The effects of play upon dominance
Behaviour of smaller and larger dogs: effects of training methods, and attachment dimensions of the dog–owner relationship. J Appl.
inconsistency of owner behaviour and level of engagement in activi- Anim. Welf. Sci. 6, 67–94.
ties with the dog. Appl. Anim. Behav. Sci. 123, 131–142. Rooney, N.J., Bradshaw, J.W.S., 2004. Breed and sex differences in
Bailey, G., 2008. The Perfect Puppy. Hamlyn Publisher. the behavioural attributes of specialist search dogs—a question-
Beerda, B., Schilder, M.B.H., van Hooff, J.A.R.A.M., de Vries, H.W., 1997. naire survey of trainers and handlers. Appl. Anim. Behav. Sci. 86,
Manifestations of chronic and acute stress in dogs. Appl. Anim. Behav. 123–135.
Sci. 52, 307–319. Rooney, N.J., Bradshaw, J.W.S., Robinson, I., 2001. Do dogs respond to play
Berman, M., 1980. Early agonistic experience and rank acquisition among signals given by humans? Anim. Behav. 61, 715–722.
free ranging infant rhesus monkeys. J. Primatol. 1, 153–170. Rooney, N.J., Bradshaw, J.W.S., Almney, H., 2004. Attributes of specialist
Blackwell, E.J., Casey, R.A., 2006. The Use of Shock Collars and their search dogs—a questionnaire survey of UK dog handlers and trainers.
Impact on the Welfare of Dogs: A Review of the Current Literature. J. Forensic Sci. 49, 300–306.
Report to the RSPCA, http://www.rspca.org.uk/servlet/Satellite? Schalke, E., Stichnoth, J., Ott, S., Jones-Baade, R., 2007. Clinical signs caused
blobcol=urlblob&blobheader=application%2Fpdf&blobkey=id& by the use of electric training collars on dogs in everyday life situation.
blobtable=RSPCABlob&blobwhere=1138718966544&ssbinary=true Appl. Anim. Behav. Sci. 105, 369–380.
(accessed on 18/02/10). Schilder, M.B.H., van der Borg, J.A.M., 2004. Training dogs with help of
Blackwell, E.J., Twells, C., Seawright, A., Casey, R.A., 2008. The relationship the shock collar: short and long term behavioural effects. Appl. Anim.
between training methods and the occurrence of behavior problems, Behav. Sci. 85, 319–334.
as reported by owners, in a population of domestic dogs. J. Vet. Behav. Scott, J.P., Fuller, J.L., 1965. Genetics and Social Behavior of the Dog. Uni-
3, 201–217. versity of Chicago Press, p. 145.
Haverbeke, A., Laporte, B., Depiereux, E., Giffroy, J.-M., Diederich, Stephen, J., Ledger, R., 2007. Relinquishing dog owners’ ability to pre-
C., 2008. Training methods of military dog handlers and their dict behavioural problems in shelter dogs post adoption. Appl. Anim.
effects on the team’s performances. Appl. Anim. Behav. Sci. 113, Behav. Sci. 197, 88–99.
63–69. The Kennel Club, 2006. Kennel Club Breed Standards, http://www.
Herron, M.E., Shofer, F.S., Reisner, I.R., 2009. Survey of the use and out- thekennelclub.org.uk/item/210 (accessed 18/02/10).
come of confrontational and non-confrontational training methods in Tortora, D.F., 1982. Understanding electronic dog training part 1. Canine
client-owned dogs showing undesired behaviors. Appl. Anim. Behav. Pract. 9, 17–22.
Sci. 117, 47–54. Yeon, S.C., Erb, H.N., Houpt, K.A., 1999. A retrospective study of canine
Hiby, E.F., Rooney, N.J., Bradshaw, J.W.S., 2004. Dog training methods: their house soiling: diagnosis and treatment. J. Am. Anim. Hosp. Assoc. 35
use, effectiveness and interaction with behaviour and welfare. Anim. (2), 101–106.
Welf. 13 (1), 63–69.
RESEARCH ARTICLE
1 Center for Canine Behavior Studies, Salisbury, CT, United States of America, 2 Martingale Consulting,
Media, PA, United States of America, 3 School of Veterinary Medicine, University of Pennsylvania,
Philadelphia, PA, United States of America
a1111111111 * serpell@vet.upenn.edu
a1111111111
a1111111111
a1111111111
a1111111111 Abstract
Behavioral problems are a major source of poor welfare and premature mortality in compan-
ion dogs. Previous studies have demonstrated associations between owners’ personality
and psychological status and the prevalence and/or severity of their dogs’ behavior prob-
OPEN ACCESS lems. However, the mechanisms responsible for these associations are currently unknown.
Citation: Dodman NH, Brown DC, Serpell JA Other studies have detected links between the tendency of dogs to display behavior prob-
(2018) Associations between owner personality lems and their owners’ use of aversive or confrontational training methods. This raises the
and psychological status and the prevalence of
possibility that the effects of owner personality and psychological status on dog behavior are
canine behavior problems. PLoS ONE 13(2):
e0192846. https://doi.org/10.1371/journal. mediated via their influence on the owner’s choice of training methods. We investigated this
pone.0192846 hypothesis in a self-selected, convenience sample of 1564 current dog owners using an
Editor: I. Anna S. Olsson, Universidade do Porto online battery of questionnaires designed to measure, respectively, owner personality,
Instituto de Biologia Molecular e Celular, depression, emotion regulation, use of aversive/confrontational training methods, and
PORTUGAL
owner-reported dog behavior. Multivariate linear and logistic regression analyses identified
Received: August 8, 2017 modest, positive associations between owners’ use of aversive/confrontational training
Accepted: January 31, 2018 methods and the prevalence/severity of the following dog behavior problems: owner-
directed aggression, stranger-directed aggression, separation problems, chasing, persis-
Published: February 14, 2018
tent barking, and house-soiling (urination and defecation when left alone). The regression
Copyright: © 2018 Dodman et al. This is an open
models also detected modest associations between owners’ low scores on four of the ‘Big
access article distributed under the terms of the
Creative Commons Attribution License, which Five’ personality dimensions (Agreeableness, Emotional Stability, Extraversion & Conscien-
permits unrestricted use, distribution, and tiousness) and their dogs’ tendency to display higher rates of owner-directed aggression,
reproduction in any medium, provided the original
stranger-directed fear, and/or urination when left alone. The study found only weak evidence
author and source are credited.
to support the hypothesis that these relationships between owner personality and dog
Data Availability Statement: The data file for this
behavior were mediated via the owners’ use of punitive training methods, but it did detect a
project is available on OPENICPSR at: http://doi.
org/10.3886/E101214V1. more than five-fold increase in the use of aversive/confrontational training techniques
among men with moderate depression. Further research is needed to clarify the causal rela-
Funding: The study was funded by The Simon
Foundation, Inc., a private 501(c)3 of Mark And tionship between owner personality and psychological status and the behavioral problems
Linda Greenberg to NHD; http://www. of companion dogs.
thesimonfoundation.org/. The funder provided
support in the form of research grants to Tufts
University and the University of Pennsylvania but
did not have any additional role in the study design,
Sample recruitment
Dog owners who participated in the study were recruited through The Simon Foundation,
Inc.’s Center for Canine Behavior Studies (CCBS) that operates as a 501(c)3 non-profit canine
behavior research center. To participate in the study, prospective participants were required to
have an active CCBS membership. Registration at CCBS required a valid email address, which
was authenticated through a time-sensitive confirmation email. Once registered at CCBS, pro-
spective participants were asked to register each dog they wished to include in the Study. Dog
registration asked (but did not require) information about the dog, including age, sex, steriliza-
tion status, breed information (pure, designer/hybrid, or unknown mixed), and acquisition
source. Each submitted dog registration was paired with a uniquely generated CCBS identifica-
tion code. This CCBS dog data collected during registration was included with the REDCap
data for analysis.
The study protocol was reviewed and approved by the Institutional Review Boards of both
Tufts University and the University of Pennsylvania. All study participants were required to
read the approved online Consent Form and to indicate their understanding of, and agreement
to, the terms of the study by checking a box before being able to gain access to the study ques-
tionnaires hosted on REDCap. Children under 18 years of age were excluded from participa-
tion in the study. The complete survey is available for viewing at: https://collaborate.tuftsctsi.
org/redcap/surveys/?s=ECBqHsA6yd&part_id=UvMPj5aelmfYySKikJrgq45OrT1QZ2jKoVqx
RrC2DIR2rrbB1sHhmDVDmMkcFtNTf3gWkJPNh
At the end of the survey, the following statement was posted for the benefit of participants
who may have felt sad, depressed, or suicidal after having completed the survey:
“Thank you for participating in this study. If in the course of participating in this study you
have become more aware of any personal emotional distress, we encourage you to seek help
within your local community. To find a community based Mental Health Treatment Center,
go to http://findtreatment.samhsa.gov/MHTreatmentLocator/faces/quickSearch.jspx. If you
experience suicidal ideation, we strongly urge you to reach out to the International Association
for Suicide Prevention, http://www.iasp.info/resources/Crisis_Centres”.
Statistical analysis
Descriptive statistics were calculated. Non-normally distributed variables were expressed as
medians and ranges, and categorical variables were expressed as frequencies and percentages.
Internal reliability of the Attitude to Training questionnaire was evaluated using factor analysis
with subsequent varimax rotation to confirm a single factor. Inter-item correlation matrices
were evaluated to detect negative correlations and to identify items that might have consis-
tently weak correlations with other items or the scale.
Depending on the skewness of the scores and model specifications, hierarchical linear or
logistic regression analyses were performed to assess the relationship between the owner psy-
chological variables (TIPI, BDI and ERQ) and the outcome variables (ATT and mini C-BARQ
scores) when the effects of potential confounding or intervening background variables (e.g.
dog age, sex, neuter status; owner gender, prior dog ownership, and number of dogs in the
household) were accounted for. Interactions among the main effects were also investigated.
Univariate analysis was performed and variables associated at a significance level of p < 0.20
were evaluated in each multivariable model. Background variables associated with a >15%
change in coefficients were retained in the models as confounders. All models were evaluated
for specification error, and variables were retained in each multivariable model when the value
of p for that variable was < 0.05. All analyses were performed using STATA 13 (StataCorp LP,
College Station, TX 77845 USA).
Table 1. All statistically significant associations between owner psychological variables, attitudes to training, and dog mini C-BARQ scores.
C-BARQ subscalesb C-BARQ miscellaneous items
Owner Psychological EXC ODA SDA DDA FDA SDF DDF NSF TS SEP ATT TRAIN CHASE ENERGY URIN DEF BARK
Variables ALONE ALONE ATT
ATT Scorea + + + + + + + + + + + + +
TIPI Agreeableness – – – – _
TIPI Conscientiousness – – – – – – – _
TIPI Emotional – – – – – – – – – – – – _
Stability
TIPI Extraversion – – – +
TIPI Openness – – – +
BDI Mild Depression
BDI Moderate – + + + +
Depression
BDI Severe Depression
ERQ Cognitive – –
Reappraisal
ERQ Expressive
Suppression
Key to C-BARQ variables: EXC = Excitability, ODA = Owner-directed aggression, SDA = Stranger-directed aggression, DDA = Dog -directed aggression,
FDA = Familiar dog aggression, SDF = Stranger-directed fear, DDF = Dog-directed fear, NSF = Nonsocial fear, TS = Touch sensitivity, SEP = Separation problems,
ATT = Attachment/attention-seeking, TRAIN = Trainability, CHASE = Chasing, URIN ALONE = Urination when left alone, DEF ALONE = Defecation when left
alone, BARK = Persistent barking.
+ As the psychological variable score increases, the ATT and C-BARQ subscale or item score increases
– As the psychological variable score increases, the ATT and C-BARQ subscale or item score decreases
a
A higher ATT score signifies use of more forceful/confrontational training methods
b
Higher mini C-BARQ subscale and item scores signify less favorable canine behavior
https://doi.org/10.1371/journal.pone.0192846.t001
• The median C-BARQ separation problems score was 0.7 (range 0 to 4). Controlling for other
variables in the model, the odds of dogs displaying separation problems is 1.78 for every
increase of 1 in the ATT score (95% CI 1.33, 2.38; p < 0.001).
• The median C-BARQ chasing score was 2.5 (range 0 to 4). Controlling for other variables in
the model, the baseline chasing score was 2.15. For every increase of 1 in the ATT score,
there is a 0.41 increase in the chasing score when the owners score low on conscientiousness
(95% CI 0.28, 0.54; p<0.001).
• The median C-BARQ energy score was 2 (range 0 to 4). Controlling for other variables in the
model, the baseline energy score was 1.29. For dog’s younger than 5 years old, energy scores
increased by 0.50 for every increase of 1 in the ATT score (95% CI 0.40, 0.60; p < 0.001).
• The median C-BARQ urinates when left alone score was 0 (range 0 to 4). Controlling for
other variables in the model, the baseline score was 0.26. For every increase of 1 in the ATT
score, dogs belonging to men with moderate depression have a 0.44 increase in their urinates
when left alone score (95%CI 0.28, 0.60; p<0.001). Dogs of owners who score higher on emo-
tional stability and agreeableness have a 0.15 lower urinates when left alone score (95% CI
0.06, 0.25; p = 0.002), compared to owners without those traits.
• The median C-BARQ ‘defecates when left alone’ score was 0 (range 0 to 4). Controlling for
the other variables in the model, the baseline score was 0.22. For every 1 increase in the ATT
score, dogs of men with moderate depression have a 0.46 higher ‘defecates when left alone’
score (95%CI 0.31, 0.60; p<0.001), compared to dogs of women without depression.
• The median C-BARQ ‘barks persistently’ score was 1 (range 0 to 4). Controlling for other var-
iables in the model, the baseline score was 1.02. For every 1 increase in the owner’s ATT
score, the dogs ‘persistent barking’ score increases by 0.39 (95%CI 0.26, 0.51; p<0.001).
• The median ATT score was 1.3 (range 1 to 7). Controlling for other variables in the model,
the baseline attitude to training score was 1.68. Men with moderate depression have a 5.31
greater ATT score (95% CI 4.34, 6.28; p<0.001), compared to women without depression.
As in a number of previous studies [22–27], the present findings confirmed a positive asso-
ciation between owners’ reported use of aversive or coercive training methods (ATT scores)
and the prevalence and severity of their dogs’ behavior problems as measured by the mini
C-BARQ (Table 1). Based on the results of the regression models, the main associations were
with owner-directed aggression, stranger-directed aggression, separation problems, chasing
(among dogs belonging to owners who score low on the TIPI conscientiousness scale), persis-
tent barking, urination when left alone, and defecation when left alone (in dogs belonging to
men with self-reported moderate depression).
The only mini C-BARQ temperament domain in which there was no association with coer-
cive training was fear/anxiety—stranger-directed fear, dog-directed fear, nonsocial fear and
touch-sensitivity showed no significant relationships with ATT scores (see Table 1). This find-
ing is somewhat at odds with the results of previous studies that found higher rates of anxiety/
fearfulness in the dogs of owners/handlers who employed aversive, punishment-based training
[22, 23]. Such inconsistencies highlight the difficulty of determining direction of causation in
cross-sectional studies of this type [36]. When positive associations between fear/anxiety and
confrontational training methods are found, it tends to be assumed that the dogs’ fear/anxiety
is induced by the aversive training techniques. Conversely, when the association is with
behavior problems other than fear/anxiety, as in the current study, a more parsimonious inter-
pretation might be that the dogs’ behavior problems are triggering their owners’ more con-
frontational training styles—except in the case of fearfulness when confrontation would
presumably be viewed by owners as counter-productive.
Also in common with previous findings [8–10], the results of this study detected a large
number of significant relationships between owner personality traits and the presence and
severity of their dogs’ behavior problems (Table 1). Most of these associations, however,
accounted for less than 10% of the variance in mini C-BARQ scores. According to the regres-
sion models, dogs of owners who rated themselves low on emotional stability displayed higher
rates of owner-directed aggression, stranger-directed fear and urination when left alone. Dogs
belonging to owners who scored low on conscientiousness showed higher rates of stranger-
directed fear. Those of owners who scored themselves as low for extraversion also displayed sig-
nificantly higher rates of stranger-directed fear, and those of owners who scored low on agree-
ableness showed higher rates of urination when left alone. Most of these associations make
reasonable sense. For example, owners who score low on emotional stability see themselves as
‘anxious, easily upset’ vs. ‘calm, emotionally stable’ [30], so the association with their dogs’
fear/anxiety-related behaviors, such as fear of strangers and separation-related urination, is
not unexpected. These dogs may be responding directly to the owners’ anxiety or indirectly as
a consequence of inadequate socialization. Like so-called ‘helicopter parents’[37], more anx-
ious and neurotic dog owners may be overprotective of their pets, thereby limiting their ability
to socialize or familiarize themselves with novel social and nonsocial situations and stimuli.
Previous studies have also found that ‘neurotic’ emotionally unstable dog owners are more
likely to report having dogs with owner-directed aggression problems [8–10]. This may be an
indication of these dogs’ greater willingness to assert themselves aggressively in competitive
interactions with their more anxious owners. Higher rates of stranger-directed fear among the
dogs of owners with low conscientiousness and extraversion also seem logical. Less extraverted
owners see themselves as ‘reserved, quiet’ vs. ‘extroverted, enthusiastic’ [30], and dogs belong-
ing to such owners are therefore likely to be less well socialized than those of more outgoing
owners. People with low conscientiousness scores see themselves as ‘disorganized, careless’ vs.
‘dependable, self-disciplined’ [30]. Like most animals, dogs tend find unpredictable environ-
ments stressful [38], and it is possible that the more pronounced fear of strangers displayed by
these dogs in some way reflects the stress of living with their more disorganized owners. How-
ever, why this effect should manifest itself only in relation to fear of strangers is not easily
explained. It is also unclear why the dogs of the less agreeable owners should show higher rates
of urination when left alone. Such owners view themselves as ‘critical, quarrelsome’ vs. ‘warm,
sympathetic’ [30] and it may be that they trigger the kinds of anxious attachments in their
dogs that could lead to separation-related house-soiling.
Contrary to expectations, the results did not provide strong support for our original
hypothesis that the link between owner personality and dog behavior problems is mediated or
effected predominantly via training methods. While the analysis did detect significant associa-
tions between certain personality types—namely agreeableness, conscientiousness and emo-
tional stability—and the use of punitive/confrontational training methods (Table 1), most of
these relationships accounted for very little of the variance in the behavior regression models,
suggesting that any influence of an owner’s personality on his or her dog’s behavior occurs pre-
dominantly through mechanisms other than training. Further research will be needed to eluci-
date what these other mechanisms might be.
The relationship between owners’ Beck Depression Inventory (BDI) scores and both ATT
and mini C-BARQ outcomes was particularly interesting, given the greater than five-fold
increase in the reported use of punitive training methods among men with moderate depres-
sion (compared with women without depression). This same group of men also reported sig-
nificantly higher rates of familiar dog aggression and house soiling (urination and defecation
when left alone) in their dogs. While it is unclear whether the dogs’ behavior problems are a
symptom or a cause of these male owners’ heavy-handed approach to training, the striking
gender difference is not unexpected. Recent studies suggest that men and women experience
different symptoms of depression, and that men tend to report higher rates of anger attacks/
aggression, substance abuse, and risk taking [39]. Depressed men are therefore more likely to
respond to their dogs’ behavior problems aggressively and punitively, and this suggests that
such individuals may be less than ideal candidates for dog ownership or adoption.
The two subscales of the Emotion Regulation Questionnaire (ERQ), cognitive reappraisal
and expressive suppression, were surprisingly unassociated with either ATT scores or mini
C-BARQ outcomes. Dogs belonging to owners who scored low on cognitive reappraisal tended
to have significantly higher excitability and energy scores on the C-BARQ (Table 1), but these
associations explained very little of the variance in the regression models.
Due to the cross-sectional nature of the current study it is impossible to determine causa-
tion in these relationships between owner personality/psychological status and dog behavior
problems. However, given that human (or owner) personality traits are generally considered
to be relatively stable and consistent over an individual’s lifetime [12, 13], it seems intuitively
more likely that owner personality contributes (either directly or indirectly) to the develop-
ment of the dog behavioral outcomes rather than vice versa. Whether the same can be said of
owner mood disorders is uncertain, though at least one previous longitudinal study detected
an apparent causal association between the onset of owners’ depressive symptoms and the
later occurrence of their dogs’ behavior problems [11].
A further limitation is that the current findings are based on indirect, owner-reported
behavioral assessments of their dogs’ behavior. While the C-BARQ has been validated in mul-
tiple studies, and is designed to reduce subjectivity by focusing on dogs’ responses to a range
of specific situations and stimuli in the recent past, it is impossible to entirely eliminate subjec-
tive biases in these sorts of questionnaire evaluations. Consequently, it could be argued that
the current findings reflect consistent biases in how people of different personality types evalu-
ate their dogs on the C-BARQ rather than any real differences in the dogs’ behavior. Without
independent behavioral observations of dogs belonging to people of different personality
types, this issue cannot be resolved, though it seems implausible that different personality fac-
tors would all result in the same systematic biases in owners’ responses to a canine behavioral
assessment survey.
Conclusions
The results of the present study confirmed a positive link between owners’ reported use of
aversive or coercive training and the prevalence and severity of their dogs’ behavior problems,
and also identified a strong association between male owner depression and the tendency to
use aversive or punitive training methods. While it is still unclear whether the owners’ use of
confrontational training is a symptom or a cause of the dogs’ behavior problems, the absence
of any associations with anxiety/fear-related behavior suggests that the former is more likely.
The study also detected significant associations between four of the ‘Big Five’ owner personal-
ity traits and the prevalence of some canine behavior problems, but found little evidence to
support the hypothesis that style of training mediates these effects. Further research will be
needed to clarify the mechanisms underlying these associations. Overall, the findings illustrate
the independent contributions of both human and dog psychological variables to the mainte-
nance of harmonious human-dog relationships, and have implications for the behavior and
welfare of both companion and working dogs, and the impact of dogs on the health and well-
being of their owners.
Supporting information
S1 File. Attitude to Training questionnaire.
(DOCX)
Acknowledgments
The authors gratefully acknowledge the technological support provided by Tufts University
(REDCap data capture program), Ian Dinwoodie, BSChE, and Small Fish Technologies, LLC.
Author Contributions
Conceptualization: Nicholas H. Dodman, James A. Serpell.
Data curation: Nicholas H. Dodman.
Formal analysis: Dorothy C. Brown, James A. Serpell.
Funding acquisition: Nicholas H. Dodman.
Investigation: Nicholas H. Dodman, Dorothy C. Brown, James A. Serpell.
Methodology: Nicholas H. Dodman, Dorothy C. Brown, James A. Serpell.
Project administration: Nicholas H. Dodman, James A. Serpell.
References
1. ASPCA. 2017. Shelter intake and surrender. https://www.aspca.org/animal-homelessness/shelter-
intake-and-surrender/pet-statistics (accessed Jul 20, 2017).
2. Arkow PS, Dow S. The ties that do not bind: A study of the human-animal bonds that fail. In: Anderson
RK, Hart BL, Hart LA, editors. The Pet Connection: Its Influence on Our Health and Quality of Life. Cen-
ter to Study the Human-Animal Bond, Minneapolis, MN. 1984. pp. 348–354.
3. Herron ME, Lord LK, Hill LN, Reisner IR. Effects of preadoption counseling for owners on house-training
success among dogs acquired from shelters. JAVMA 2007; 231: 558–562. https://doi.org/10.2460/
javma.231.4.558 PMID: 17696854
4. Miller DM, Stats SR, Partlo BS, Rada K. Factors Associated with the decision to surrender a pet to an
animal shelter. JAVMA 1996; 209: 738–742. PMID: 8756871
5. Patronek GJ, Glickman LT, Beck AM, McCabe GP, Ecker C. Risk factors for relinquishment of dogs to
an animal shelter. JAVMA 1996; 209: 572–581. PMID: 8755975
6. Salman MD, New JG, Scarlett JM, Kass PH, Ruch-Gallie R, Hetts S. Human and animal factors
related to the relinquishment of dogs and cats in 12 selected animal shelters in the United States.
J Appl Anim Welf Sci 1998; 1: 207–226. https://doi.org/10.1207/s15327604jaws0103_2 PMID:
16363966
7. Salman MD, Ruche-Gallie R, Kogan L, New JC, Kass PH, Scarlett JM. Behavioral reasons for relin-
quishment of dogs and cats to 12 shelters. J Appl Anim Welf Sci 2000; 3: 93–106.
8. O’Farrell V. Effects of owner personality and attitudes on dog behavior. In Serpell JA, editor. The
Domestic Dog: Its Evolution, Behaviour and Interactions with People, Cambridge: Cambridge Univer-
sity Press. 1995. pp. 153–158.
9. Podberscek AL, Serpell JA. Aggressive behaviour in English cocker spaniels and the personality of
their owners. Vet Rec 1997; 141: 73–76–227.
10. Dodman NH, Patronek G J, Dodman VJ, Zelin ML, Cottam N. Comparison of personality inventories of
owners of dogs with and without behaviour problems. Int. J Appl Res Vet Med 2004; 2(1): 55–61.
11. Hunt MG, Otto CM, Serpell JA, Alvarez J. Interactions between handler well-being and canine health
and behavior in search and rescue teams. Anthrozoös 2012; 25: 323–335.
12. American Psychological Association. 2017. Personality. http://www.apa.org/topics/personality/
(accessed Jul 20, 2017).
13. Pervin LA. Personality: Theory, Assessment, and Research. New York: John Wiley & Sons. 1970.
14. Kis A, Turcsán B, Miklósi Á, Gácsi M. The effect of owner’s personality on the behavior of owner-dog
dyads. Interact Stud 2012; 13: 373–385.
15. Cimarelli G, Turcsán B, Bánlaki Z, Range F, Virányi Z. Dog owners’ interaction styles: Their compo-
nents and associations with reactions of pet dogs to a social threat. Front Psychol 2016; 7: 1–14.
https://doi.org/10.3389/fpsyg.2016.00001
16. Clarke GI, Boyer WN. The effects of dog obedience training and behavioral counseling upon the
human-canine relationship. Appl Anim Behav Sci 1993; 37: 147–159.
17. Haverbeke et al. Training methods of military dog handlers and their effect on the team’s performances.
Appl Anim Behav Sci 2008; 113: 110–122.
18. Hoummady S, Péron F, Grandjean D, Cléro D, Bernard B, Titeux E, Desquilbet L, Gilbert C. Relation-
ship between personality of human-dog dyads and performances in working tasks. Appl Anim Behav
Sci 2016; 177: 42–51.
19. Lefebvre D, Diederich C, Delcourt M, Giffroy J-M. The quality of the relation between handler and mili-
tary dogs influences efficiency and welfare of dogs. Appl Anim Behav Sci 2007; 104: 49–60.
20. Rooney NJ, Cowan S. Training methods and owner-dog interactions: Links with dog behaviour and
learning ability. Appl Anim Behav Sci 2011; 132: 169–177.
21. Topál J, Miklósi Á, Csányi V. Dog-human relationship affects problem solving behavior in the dog.
Anthrozoös 1997; 10: 214–224.
22. Arhant C, Bubna-Littitz H, Bartels A, Futschik A, Troxler J. behavior of smaller and larger dogs: Effects
of training methods, inconsistency of owner behavior and levels of engagement in activities with dog.
Appl Anim Behav Sci 2010; 123: 131–142.
23. Blackwell EJ, Twells C, Seawright A, Casey RA. The relationship between training methods and the
occurrence of behavior problems, as reported by owners, in a population of domestic dogs. J Vet Behav
2008; 3: 207–217.
24. Casey RA, Loftus B, Bolster C, Richards GJ, Blackwell EJ. Human-directed aggression in domestic
dogs (Canis familiaris): Occurrence in different contexts and risk factors. Appl Anim Behav Sci 2014;
152: 52–63.
25. Herron ME, Shofer FS, Reisner IR. Survey of the use and outcome of confrontational and non-confron-
tational training methods in client-owned dogs showing undesired behaviors. Appl Anim Behav Sci
2009; 117: 47–54.
26. Hiby E, Rooney NJ, Bradshaw JWS. Dog training methods: Their use, effectiveness and interactions
with behavior and welfare. Anim Welf 2004; 13: 63–69.
27. Hsu Y, Sun L. Factors associated with aggressive responses in pet dogs. Appl Anim Behav Sci 2010;
123: 108–123.
28. Deldalle S, Gaunet F. Effects of two training methods on stress-related behaviours of the dog (Canis
familiaris) and on the dog-owner relationship. J Vet Behav 2014; 9: 58–65.
29. Horváth Z, Dóka A, Miklósi A. Affiliative and disciplinary behavior of human handlers during play with
their dog affects cortisol concentrations in opposite directions. Hormones and Behav 2008; 54: 107–
114.
30. Gosling SD, Rentfrow PJ, Swann WB. A very brief measure of the Big Five personality domains. J Res
Pers 2003; 37: 504–528.
31. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for measuring depression. Arch
Gen Psychiat 1961; 4: 561–571. PMID: 13688369
32. Gross JJ, John OP. Individual differences in two emotion regulation processes: Implications for affect,
relationships and well-being. J Pers Soc Psych 2003; 85(2): 348–362.
33. Duffy DL, Kruger K, Serpell JA. Evaluation of a behavioral assessment tool for dogs relinquished to
shelters. Prevent Vet Med 2014; 117: 601–609.
34. Hsu Y, Serpell JA. Development and validation of a questionnaire for measuring behavior and tempera-
ment traits in pet dogs. JAVMA 2003; 223, 1293–1300. PMID: 14621216
35. Duffy DL, Serpell JA. Predictive validity of a method for evaluating temperament in young guide and ser-
vice dogs. Appl Anim Behav Sci 2012; 138: 99–109.
36. Fernandes JG, Olsson IAS, Vieira de Castro AC. Do aversive training methods actually compromise
dog welfare? A literature review. Appl Anim Behav Sci 2017; http://dx.doi.org/doi:10.1016/j.applanim.
2017.07.001
37. Segrin C, Woszidlo A, Givertz M, Montgomery N. Parent and child traits associated with overparenting.
J Soc Clin Psychol 2013; 32(6): 569–595.
38. Broom DM, Johnson KG. Stress and Animal Welfare. London: Chapman & Hall. 1993.
39. Martin LA, Neighbors HW, Griffith DM. The experience of symptoms of depression in men vs. women:
Analysis of the National Comorbidity Survey Replication. JAMA Psychiat 2013; 70(10): 1100–1106.
Does t raining make you smart er? T he effect s of t raining on dogs’ performance (Canis familia…
Pier Accorsi, Irena Pet ak
T he qualit y of t he relat ion bet ween handler and milit ary dogs influences efficiency and welfare of dogs
Claire Diederich
Abstract
Historically, pet dogs were trained using mainly negative reinforcement or punishment, but positive reinforcement using rewards has
recently become more popular. The methods used may have different impacts on the dogs' welfare. We distributed a questionnaire
to 364 dog owners in order to examine the relative effectiveness of different training methods and their effects upon a pet dog's
behaviour. When asked how they trained their dog on seven basic tasks, 66% reported using vocal punishment, 12% used physical
punishment, 60% praise (social reward), 51% food rewards and 11% play. The owner's ratings for their dog's obedience during eight
tasks correlated positively with the number of tasks which they trained using rewards (P < 0.01), but not using punishment (P = 0.5).
When asked whether their dog exhibited any of 16 common problematic behaviours, the number of problems reported by the
owners correlated with the number of tasks for which their dog was trained using punishment (P < 0.001), but not using rewards
(P = 0.17). Exhibition of problematic behaviours may be indicative of compromised welfare, because such behaviours can be caused
by or result in a state of anxiety and may lead to a dog being relinquished or abandoned. Because punishment was associated
with an increased incidence of problematic behaviours, we conclude that it may represent a welfare concern without concurrent
benefits in obedience. We suggest that positive training methods may be more useful to the pet-owning community.
Keywords: animal welfare, domestic dog, humananimal interaction, obedience, problematic behaviour, training methods
Introduction desired behaviours. These methods have been adopted as
Domestic dogs (Canis familiaris) are kept in large numbers part of the training regime of many working dog organisa-
by humans worldwide, and their welfare is worth serious tions, including the Guide Dogs for the Blind Association
ethical and scientific consideration. The majority of previous (Johnston 1995) and agencies training dogs to detect illegal
research in this area has been in laboratories and re-homing substances (Adams & Johnson 1994).
centres (eg Hubrecht 1993, 1995; Hennessey et al 1998; In this paper we survey the training methods used by the
Wells & Hepper 2000), but most domestic dogs are kept as general pet owning community in the UK. We also examine
pets within homes. The manner in which these dogs are the relative effectiveness of these methods for training a
trained may be of particular importance to their welfare. variety of specific tasks. This is important because positive
Most pet dogs receive at least rudimentary training. For training methods are likely to be adopted only if they are at
instance, dogs are trained to eliminate in an appropriate least as successful as their negative counterparts. We
place, to walk to heel, and to obey a range of commands hypothesise that specific training methods will be associated
including sitting and lying down. The methods by which with the performance of behaviours deemed problematic by
these tasks are trained vary greatly. owners (Overall 1997). The manifestation of problematic
Traditional training techniques have used mainly aversive behaviours may be important for dog welfare for several
stimuli, either in the form of positive punishment (applica- reasons: first, they are often caused by, or result in, states of
tion of an aversive stimulus in response to an undesirable anxiety; and second, problematic behaviours have been
behaviour) or negative reinforcement (removal of an aver- implicated in the weakening of the petowner relationship
sive stimulus leading to an increase in the performance of a (Mugford 1981), which may result either in the relinquish-
desirable behaviour) (Lieberman 1999). The use of aversive ment of dogs (Serpell 1996) or in euthanasia.
stimuli in training may have negative welfare implications: Previous studies of the relationship between training methods
it is thought to cause suffering (Beerda et al 1997), possibly and problematic behaviours have yielded apparently con-
poses health risks (through increased levels of physiological flicting results. Podberscek and Serpell (1997) and Voith et
stress), and has been found to be related to aggression al (1992) found no relationship between obedience training
towards other dogs (Roll & Unshelm 1997). and problematic behaviours, while Clark and Boyer (1993)
In contrast, other training methods utilise positive reinforce- and Jagoe and Serpell (1996) found a decrease in these
ment through the presentation of rewards in response to behaviours following obedience training. This discrepancy
Universities Federation for Animal Welfare Science in the Service of Animal Welfare
may be because these studies compared formal (under the then asked which methods they had used to train their dog
supervision of a dog trainer) versus informal training and to perform four tasks: to come when called, to sit on com-
took no account of the specific methods used (eg punishment mand, to leave or give up an object on command, and to
and reward). The aim of the current study is to document the walk to heel.
use of training methods by the pet-owning community and Obedience
investigate how these methods interact with both obedience
and problematic behaviours. Owners were asked to rate their dogs obedience on a scale
of 15 (5 being the most obedient) for each of the seven
Methods tasks. They were also asked to give an eighth obedience
score (15) for overall obedience. All eight obedience
Participants and distribution
scores were summed to give a summed obedience score.
Three main sites were used in Hampshire: Southampton
Problematic behaviours
Common (50°55´36´´N, 1°24´39´´W), Southampton Sports
Centre (50°56´28´´N, 1°25´14´´W), and Deers Leap Respondents were presented with a list of 16 common prob-
(50°52´53´´N, 1°30´45´´W). Four main sites were used in lematic behaviours (Barlow 2003) and asked to indicate
Cambridgeshire: Gog Magog Down (52°9´22´´N, whether their dog had shown each of them in the past,
0°10´25´´E), Cherry Hinton Estate Park (52°11´13´´N, currently or never. These 16 problems were reduced to 13
0°9´59´´E), Midsummer Common (52°12´36´´N, 0°7´46´´E), during analysis (Table 1).
and Jesus Green (52°12´43´´N, 0°7´25´´E). These sites were
Statistical analysis
chosen for their high concentration of dog walkers and
because they represented a cross-section of the types of All statistical analyses were carried out using SPSS 11 for
areas open for dog walking, ie both rural and urban. All sites Windows (SPSS Inc, 19892001).
were visited at various times of the day during daylight Demographics
hours, on both weekends and weekdays. All dog walkers The data from Hampshire and Cambridgeshire were com-
were approached by the experimenter and asked if they pared using the χ2 test for categorical data and the Mann-
would mind participating in a questionnaire survey about Whitney U test (Siegel & Castellan 1988) for ordinal data.
their dog. If amenable, they were given a questionnaire with
Training methods and obedience
a stamped addressed envelope.
Questionnaires with incomplete training sections, or in
In addition, 60 questionnaires were distributed to local vet-
which the respondents stated that they had not trained the
erinary surgeries and pet stores in each of the two counties
dog themselves, were excluded from analysis. We analysed
(Hampshire and Cambridgeshire) for voluntary pick-up.
the success of each commonly used method for training the
This ensured that not only people who walked their dogs
seven tasks. For each task in turn, we examined each training
received the questionnaire.
method mentioned by at least 10% of respondents. We used
To avoid reports of puppy behaviour, only people who Mann-Whitney U tests to compare the obedience scores
owned a dog over one year of age were recruited. If respon- given by those respondents who had used that training
dents owned more than one dog, they were asked to answer method with the obedience scores given by those respon-
with regard to the youngest dog (over one year), as it was dents who had not used it.
assumed that the training techniques used for the younger
Each reported training method was then categorised as
dog would be remembered better.
reward-based, punishment-based, or miscellaneous (ie not
Design obviously rewarding nor punishing) (see Table 2). The
The questionnaire contained 26 questions, 13 simple/multiple- authors and also two clinical animal behaviourists agreed
choice and 13 open-ended, covering the following topics. the categorisation. For each respondent, we then added up
Demographics the total number of times that they had reported use of each
method type (reward, punishment and miscellaneous) to
Respondents were asked their gender and age and that of generate three new variables: reward frequency, punishment
their dog. They also gave information on the breed, sexual frequency, and miscellaneous method frequency. These
status (entire or neutered), source and age at acquisition of variables were tested for correlation with the summed obe-
the dog, and how many dogs they owned both presently and dience scores using Spearmans Rank correlation tests
in the past. (Siegel & Castellan 1988).
Training methods Finally, we categorised the respondents according to their
Respondents were asked open-ended questions about the general training strategy. Each respondent was categorised
training methods that they had used to train seven common as using reward only, punishment only, a combination of
tasks. They were first asked about three specific training reward and punishment, or miscellaneous methods only.
situations: how they toilet trained their dog, how they react- The summed obedience scores for the four groups were
ed to their dog chewing household objects, and how they compared using the Kruskal-Wallis test (Siegel & Castellan
reacted if their dog stole food or other objects. They were 1988).
Problematic behaviour Never displayed Displayed in the past Currently being displayed
Barking at people 15.8 8.2 75.7
Aggression towards people 50.2 16.8 33.0
Barking at dogs 34.4 12.9 52.7
Nipping at dogs 74.8 7.7 13.8
Growling at dogs 40.1 17.5 42.4
Fear in a few situations 22.0 17.8 60.2
Fear in many situations 46.6 19.8 33.6
Excitement in a few situations 9.6 8.3 82.1
Excitement in many situations 37.1 15.1 47.8
Separation-related behaviours 46.3 38.0 15.6
Inappropriate mounting 71.3 14.7 14.1
Repetitive behaviours 80.1 10.1 9.8
Eating non-foodstuffs 57.6 25.9 16.5
dogs previously (median 2 versus 1; U = 11676.5, punishment, and then those using punishment only.
P = 0.05). Since no further differences were found, we con- Respondents using miscellaneous methods only reported
sidered the two populations to be sufficiently similar to the lowest obedience scores.
combine them for further analysis. Obedience and training for each task
Training methods When the obedience scores for each individual task were
Twelve training techniques were described (by >10% of the compared between respondents who had and who had not
respondents) in response to the seven training tasks. When used each of the common methods, three significant results
methods were categorised into the three types (Table 2), the were found:
type of training method was seen to vary depending on the (1) Obedience to leave or give up an object was significant-
task being trained. Punishment was commonly reported ly greater for those dogs that had been trained using play as
when training dogs not to chew household objects or steal; a reward, in comparison with those dogs that had not been
rewards were more common when training dogs to sit or trained using this method (U = 1977.5, P < 0.01).
come to call (Table 3).
(2) Obedience to walk to heel was significantly greater for
Overall 20.2% of respondents used reward-based methods those dogs that had been trained using praise as a reward, in
only, 9.8% used punishment-based methods only, 60.4% comparison with those dogs that had not been trained using
used a combination of reward and punishment, and 9.6% this method (U = 1612, P < 0.05).
used miscellaneous methods only or mentioned no methods
(3) Obedience for not chewing household objects was
at all.
greater in those dogs that had received an alternative object
Obedience to chew in response to their chewing behaviour, in compar-
The median summed obedience score was 33 (first quar- ison with those dogs that had not received an alternative
tile = 29, third quartile = 36). The scores ranged from 19 to (U = 3181.5, P < 0.05).
the maximum possible score of 40. No significant associations were found between the obedi-
Obedience and overall use of training methods ence scores for any of the other tasks and any specific train-
The reward frequency correlated positively with the ing technique.
summed obedience score (Rho = 0.26, P < 0.01). No signif-
Problematic behaviours
icant correlation was found between obedience and the fre-
quency of either punishment or miscellaneous methods. Of the respondents, 97.2% mentioned at least one of the 13
problematic behaviours. The most prevalent problems were
The summed obedience scores for the four categories of
showing excitement in a few situations, barking at people,
respondents (those that used reward only, punishment only,
and showing fear in a few situations (Table 1).
a combination of reward and punishment, or miscellaneous
methods only) differed significantly (Kruskal-Wallis; Problematic behaviours and use of training methods
χ2 = 8.152, P < 0.05). Highest obedience scores were given The punishment frequency correlated positively with the
by respondents using reward-based methods only; this was number of current problematic behaviours (Rho = 0.17,
followed by those using a combination of reward and P < 0.01). In contrast, no significant correlations were
found between the frequency of either reward or miscella- (Christiansen et al 2001; Marschark & Baenninger 2002).
neous methods and the incidence of problematic behaviours. For example, many owners used punishment-based training
The number of current problematic behaviours reported by to teach dogs not to chew or steal objects. However, this
the four categories of respondents (those that used reward survey suggests that for everyday training, punishment is
only, punishment only, a combination of reward and punish- not the most effective method. Furthermore, for certain tasks,
ment, or miscellaneous methods only) differed significantly reward-based methods are significantly more successful.
(Kruskal-Wallis; χ2 = 8.993, P < 0.05). The greatest number When we compared dogs obedience at seven basic tasks,
of current problematic behaviours was reported by respon- for four of these tasks (toilet training, stopping stealing
dents using punishment only, or a combination of both pun- objects, coming to call and sitting on command) we discov-
ishment and reward (both medians = 5); the lowest number ered no significant difference between dogs trained using
was reported by respondents using reward only or miscella- one specific method in comparison with another. However,
neous methods only (both medians = 4). for three tasks (leaving or giving up an object on command,
When the respondents using the four different training walking to heel and avoiding chewing household objects),
strategies were compared for the occurrence of each of the the use of specific methods was associated with significantly
13 problematic behaviours, three significant associations higher obedience scores. Although the most effective tech-
were found: nique varied according to the specific training task, for none
(1) Owners reporting reward-based methods only had the of the tasks was a punishment-based method most effective.
lowest percentage of dogs showing current over-excitement Even for chewing and stealing objects, where punishment is
(χ2 = 11.6, P < 0.01). very commonly used, those owners who used it did not
report greater obedience. Thus, examination of the individ-
(2) Owners using punishment, either alone or in combina-
ual tasks provides no support for the value of punishment.
tion with reward, had the highest percentage of dogs
exhibiting separation-related problems either currently or in Further evidence supporting the use of reward-based
the past (χ2 = 29.2, P < 0.001). methods was found when examining overall obedience
(3) Owners who reported using only miscellaneous meth- scores. These scores correlated significantly with the
ods, or no methods at all, reported the lowest percentage of number of times the owners reported using reward-based
dogs eating non-foodstuffs either currently or in the past training methods, but were unrelated to their reports of
(χ2 = 6.8, P < 0.01). either punishment-based or miscellaneous methods.
Furthermore, dogs trained exclusively using reward-based
Discussion methods were reported to be significantly more obedient
This survey has shown that reward-based training is used than those trained using either punishment or a combination
extensively within the dog-owning community, with over of reward and punishment.
three-quarters of respondents reporting using some form of These results suggest that there is a link between the use of
reward. However, it is unusual for owners to base all of their reward-based methods and obedience in pet dogs. However,
training on rewards, and the vast majority use a training this is a correlation and although the frequent use of rewards
regime that combines reward and punishment. Thus, pun- may lead to increased obedience, it is also possible that
ishment (verbal or physical) continues to feature prominent- when dogs show an initial high level of obedience, their
ly in the training of pet dogs. owner is more inclined to use reward-based training
Although there is an increasing concern that certain forms methods. However, regardless of which is cause and which
of punishment can cause suffering (Beerda et al 1997), is effect, it is clear that use by the general dog-owning com-
there remains a general belief that, for many canine tasks, munity of punishment-based methods, as compared to reward-
punishment is the most effective training technique based methods, does not result in a more obedient dog.
This is the final published version of the article (version of record). It first appeared online via Nature Research at
https://doi.org/10.1038/s41598-021-97743-0 . Please refer to any applicable terms of use of the publisher.
This document is made available in accordance with publisher policies. Please cite only the
published version using the reference above. Full terms of use are available:
http://www.bristol.ac.uk/red/research-policy/pure/user-guides/ebr-terms/
www.nature.com/scientificreports
Domestic dogs are trained using a range of different methods, broadly categorised as reward
based (positive reinforcement/negative punishment) and aversive based (positive punishment/
negative reinforcement). Previous research has suggested associations between use of positive
punishment-based techniques and undesired behaviours, but there is little research investigating
the relative welfare consequences of these different approaches. This study used a judgement bias
task to compare the underlying mood state of dogs whose owners reported using two or more
positive punishment/negative reinforcement based techniques, with those trained using only
positive reinforcement/negative punishment in a matched pair study design. Dogs were trained to
discriminate between rewarded and unrewarded locations equidistant from a start box, and mean
latencies recorded. Their subsequent latency to intermediate ‘ambiguous’ locations was recorded
as an indication of whether these were perceived as likely to contain food or not. Dogs trained using
aversive methods were slower to all ambiguous locations. This difference was significant for latency
to the middle (Wilcoxon Z = − 2.380, P = 0.017), and near positive (Wilcoxon Z = − 2.447, P = 0.014)
locations, suggesting that dogs trained using coercive methods may have a more negative mood
state, and hence that there are welfare implications of training dogs using such methods.
In contrast to commercially reared livestock, the welfare implications of different husbandry approaches for
domestic dogs have received relatively little research. There is a common perception that because dogs live in
domestic settings with their owners they have a ‘good life’. However, life experiences for dogs are potentially very
variable because they largely depend on the circumstances, knowledge, attitudes and lifestyle of their owners1–3.
One aspect in which owners vary considerably is in the choice of training methods used to teach new behaviours,
or to alter undesired behaviours, in their dogs. Approaches to dog training vary from the use of rewards (such
as attention, praise, play or food) when dogs show a desired response, to the use of aversive interventions when
they display unwanted behaviour4,5.
The different approaches are often described by dog training practitioners using terms developed by B.F. Skin-
ner in the 1930’s6. These terms describe the increase (reinforcement) or decrease (punishment) in the likelihood
of a specific target behaviour given the (positive) application or (negative) withdrawal of different stimuli. ‘Posi-
tive punishment’ is defined as the probability of a behaviour occurring later being decreased on the application of
a stimulus; ‘negative reinforcement’ is defined as the probability of a behaviour occurring later being increased on
withdrawal of a stimulus; ‘positive reinforcement’, is defined as the probability of a behaviour occurring later being
increased with the application of a stimulus; and ‘negative punishment’ is defined as the probability of a behav-
iour being decreased with the removal of a stimulus. However, positive punishment and negative reinforcement
inevitably occur together depending on the focal behaviour described by the these definitions: as one behaviour
within a context increases, another will be decreased. For example, the action of spraying a dog with water may
both reduce (positively punish) jumping up, but also increase (negatively reinforce) standing on all four paws
when the spraying is stopped. Similarly, a dog could be positively reinforced with attention for sitting to greet
people (behaviour increasing with the intervention), and negatively punished for jumping up by withdrawal of
attention if the dog does not sit7. Because these two categories occur together within contexts, we have combined
training techniques used by owners into ‘reward based’ (both positive reinforcement and negative punishment)
and ‘aversion based’ (both positive punishment and negative reinforcement).
Many veterinary, clinical behaviour, dog training and animal welfare organisations promote the use of reward
based training methods for dogs. Arguments for this recommendation are listed below:
1
Dogs Trust, Canine Behaviour and Research, Clarissa Baldwin House, 17, Wakley Street, London, UK. 2Gornate
Olona, Italy. 3Vets for Pets, 350 Southchurch Drive, Clifton, Nottingham, UK. 4Bristol Veterinary School, University
of Bristol, Bristol, UK. *email: Rachel.Casey@dogstrust.org.uk
Vol.:(0123456789)
www.nature.com/scientificreports/
1. That many undesired behaviours arise because dogs are anxious or fearful, and the implementation of
more aversive methods may enhance negative emotional states, resulting in further problems including
aggression8–10.
2. That dogs may associate the application of an aversive stimulus with unintended and coincidental e vents9.
For example, an electric stimulus used to deter a dog from running away may be applied when the dog is
approaching a child, with the risk of the aversive event becoming associated with children.
3. That the use of aversive methods may inhibit behaviours in that context in which the punishment is applied,
but not alter underlying emotional state, potentially leading to the subsequent return of the problem behav-
iour or alternative responses on exposure to the precipitating cue in other contexts.
4. That the use of aversive methods can create confusion and frustration in dogs because punishment of
an undesired behaviour alone does not enable the dog to understand what is required as the appropriate
response to a cue.
5. That aversive methods can risk causing physical injury or pain to dogs9.
Previous studies have suggested associations between the use of more aversive training methods with
unwanted behaviours4,5,8, and an increased occurrence of aggressive behaviour where such techniques are
used1,8,11. However, the causality of these relationships is unclear: whilst it may be that use of approaches involv-
ing aversive stimuli increase the risk of fear and aggression responses, it is also possible that owners whose dogs
show such behaviours are more likely to resort to these types of techniques.
There are also some indications that the use of more aversive methods of training have direct impacts on the
welfare of dogs. For example, German Shepherds trained with electronic (“shock”) collars were shown to display
more behavioural signs of fear towards their handlers than those trained with other methods12, and dogs showed
increases in cortisol when handlers tended to use discipline during interactions as compared to engaging in more
affiliative play13. However, there are often difficulties in interpreting behavioural and physiological indicators of
welfare, particularly when single indicators are used. Measures can be influenced by other factors and may not
necessarily be associated with emotional states of particular valence14.
A relatively novel approach to welfare assessment is the measurement of ‘cognitive bias’15. Cognitive biases
are observed where affective state influences human or animal responses to environmental stimuli, through dif-
ferences in how past events are remembered (‘memory bias’), ambiguous or future events are evaluated (‘judge-
ment bias’), or attention is focused (‘attention bias’)15. Moods are defined as relatively enduring affective states
which are detached from immediate triggering s timuli16,17. However, mood states are thought to be influenced
by cumulative experience, such that individuals which experience repeated aversive events, or fail to achieve
expected rewards, will have a more negative mood state than those successfully achieving rewards and avoid-
ing threats17,18. These changes in mood state with environmental circumstances are considered adaptive as they
enable individuals to adjust responses to ambiguity and expectations of future events according to previous
achievement (or not) of rewards, and avoidance (or not) of aversive stimuli16,17,19. For example, individuals
experiencing repeated aversive stimuli and hence in a negative mood are predicted to show a more cautious,
negative interpretation of future or ambiguous events. The judgement bias test developed by H arding20 aims to
evaluate whether animals have positive or negative expectations in ambiguous situations and hence to use these
judgements as proxy indicators of, respectively, positive or negative affect or m ood16,21.
Previous animal studies have confirmed that measures of judgement bias change with manipulations of the
environment likely to influence mood states, such as the removal or addition of e nrichment22–26, and judgement
bias methods developed for non-human a nimals20,27 have been successfully adapted for d ogs18. Given that mood
is hypothesized to be influenced by experience of reward and aversion, this approach is appropriate to evaluate
differences in affective state between dogs trained using rewarding or aversive stimuli. Previous research has
investigated differences in cognitive indicators of affect and welfare between groups of dogs attending training
classes where different approaches to training were u sed28. Dogs attending classes advocating mainly aversive
training methods were found to be more ‘pessimistic’ in a cognitive bias task than dogs from training classes
adopting a predominantly reward-based approach. The aim of this project was to further examine this relation-
ship by using a matched pair study design to test the hypothesis that dogs trained using owner reported training
techniques based on aversion (i.e. using positive punishment/negative r einforcement7) are more likely to show a
relatively ‘pessimistic’ judgement bias as compared to a breed, age and sex matched control population in which
these techniques had not been used.
Results
A spatial judgement bias task was used18,27 in which dogs were trained that a bowl would contain reward (food) on
trials in which it was placed at a standardized (positive) location on the left side of a test arena, and no reward on
trials in which it was placed at a standardized (negative) location on the right side of the arena (reward-location
combinations remained the same for each dog and were counter-balanced across dogs). Once dogs had learnt
this spatial discrimination and were running faster to the bowl on trials in which it was on the rewarded side,
trials in which the bowl was placed in one of three ambiguous locations (e.g. half-way between the left and right
locations) were presented. Latency to move to the bowl in these trials was used as an indicator of whether the
dog anticipated food to be present (short latency—positive (‘optimistic’) interpretation of ambiguity inferred to
reflect relatively positive affect) or absent (longer latency—negative (‘pessimistic’) interpretation of ambiguity
inferred to indicate a relatively negative affective state). 113 tested dogs completed the cognitive bias task. Of
these 13 were removed from the dataset because suitable matches were not identified for testing, leaving 100
subjects for analysis: 50 trained with two or more aversive methods (AT group), and 50 who had not experienced
these training methods according to owner report (RT group). A breakdown of the breeds, age categories, sex
Vol:.(1234567890)
www.nature.com/scientificreports/
Figure 1. Mean latency to positive and negative locations and unadjusted latencies to probe trials, showing 95%
confidence intervals, during the test phase for all dogs.
and neuter status for these dogs are show in Table 1. The mean latencies to reach goal locations in the test phase
for all dogs are shown in Fig. 1.
No significant effect of rewarded bowl location (left or right) was found for latencies to probe goal locations,
suggesting that side biases were unlikely to have influenced results. A final trial in which the bowl was presented
in the negative location but contained food was used to investigate whether dogs might be using olfactory cues
in the task. In fact, latency to the bowl location on this trial tended to be even slower than mean latencies to this
location during the test phase (Wilcoxon Z = − 2.721, P = 0.07) indicating that responses were unlikely to have
been influenced by olfactory cues.
Vol.:(0123456789)
www.nature.com/scientificreports/
Figure 2. Mean latency for each intermediate ‘probe’ goal location, divided by training group.
Dogs in the aversion training (AT) group and the reward training (RT) group did not differ with respect to
the number of trials to reach criterion (Wilcoxon, Z = − 0.368, P = 0.713) suggesting that the groups did not vary
in their ability to learn the spatial discrimination task. There was also no significant difference between groups
in the mean latency to the negative location or the mean latency to the positive location during the test phase
(Wilcoxon Z = − 1.105, P = 0.836 and Z = − 1.105, P = 0.269 respectively).
Dogs trained using no aversive methods (reward training; RT) had a lower latency (i.e. ran faster) to all probe
goal locations than those trained using two or more aversive methods (aversion training group: AT; Fig. 2). This
difference was significant for latency to the middle probe location (Wilcoxon Z = − 2.380, P = 0.017), and for
latency to the near positive probe location (Wilcoxon Z = − 2.447, P = 0.014).
Discussion
This study used a matched pair study design to compare dogs whose owners reported using at least two training
methods involving the application of an aversive stimulus with those trained with rewards only. Because mood
states are thought to be influenced by cumulative experience16,17, we hypothesized that dogs who experienced
aversive techniques during training would have a more negative mood state than those whose training was
focused on rewarding desired behaviours. Measurement of latency to reach ambiguous bowl locations in the
judgement bias task supported this hypothesis, with dogs in the group trained with aversive methods being more
‘pessimistic’ than those trained without these techniques, and moving significantly more slowly to the bowl when
on trials when it was in the central ambiguous location or that nearest the trained rewarded location.
This ‘pessimistic’ bias indicates that dogs which have been exposed to more aversive training methods have a
decreased expectation of reward than those which have not experienced these methods. This may reflect a nega-
tively biased processing of ambiguous information, as is found in depressed p eople29,30 rats exposed to chronic
31
psychosocial stress , and in a variety of other judgement bias studies involving animals in putative negative
affective states25,26. The results support the findings of Vieira de Castro et al.28, suggesting that the use of aversive
dog training methods may induce longer-term negative mood states and hence poorer welfare compared to
training methods that do not involve aversion.
It is important to note, however, that the relationship identified is correlational rather than causal. It is pos-
sible that rather than aversive training methods generating a negative affective state which leads to pessimistic
decision-making under ambiguity, dogs may show stable individual differences in the way in which they interpret
situations32,33 that influence their owners’ use of aversion based methods of training. For example, dogs with a
more pessimistic cognitive bias may be more likely to develop fear associated behaviour problems which own-
ers find difficult to manage, resulting in greater use of punitive techniques. Further research is needed to fully
understand the causal direction of the observed relationship between experience of reward and aversion and
cognitive bias.
Some caution is required in interpreting these findings as groups were defined according to owner reported
use of training methods. Although this was checked on the telephone on booking appointments, and verbally
with owner at time of the visit, there was no adjustment for how often the target methods were used, nor how
long ago they were last used –factors which may have made the experience of aversive training methods variable
in the AT group. However, the risk of classifying dogs in the AT group who had only experienced such methods
minimally was reduced by having an inclusion criterion of owners using at least two of the listed methods. Such
owners were assumed to be more likely to have an overall ethos of using this type of training method. There is
also potential for variation in how these methods were applied to dogs. For example, application of a punisher
poorly contingent with the behaviour of the dog is likely to be more stressful for the animal than where it is
Vol:.(1234567890)
www.nature.com/scientificreports/
Training method
Bark activated citronella collar
Remote activated citronella collar
Pet corrector
Physical punishment (e.g. smacking or shaking)
Remote activated electronic collar
Bark activated electronic collar
Water pistol
Check or choke chain
‘Rattle can’ or other sound based ‘distraction’ method
Table 2. Dogs were included in the aversive training group if owners reported using two or more of these
methods, and included in the reward training group if the owner reported never having used any.
appropriately applied12. Nevertheless, our findings appear to replicate those of Vieira de Castro et al.28, where
categorization of groups was conducted based on attendance to different training classes, with methods of train-
ing further categorized through video analysis of sessions.
The other limitation was that the inclusion criteria for the reward trained group was that owners had not used
the specific methods listed in Table 2. These owners may, however, have used other aversive methods that were
not specified. In addition, dogs in both groups may have experienced different training methods in the past (e.g.
if they had been rehomed). Finally, owners in both groups reported using reward based methods to train their
dogs (e.g. praise, play and food treats) as it is difficult to find owners who do not use any such methods with
their dog. This means that the groups were defined according to whether owners also reported using aversive
methods—a comparison between those using only positive reward to those using both positive reward based
and positive punishment based methods. It is therefore possible that group differences may be influenced by the
relative consistency of methods used by owners, as also suggested by Hiby5.
It is possible that variables such as breed, gender, neuter status and age may influence cognitive bias. Matching
pairs of dogs by these criteria in this study enabled us to conclude with greater confidence that methods used
in training rather than demographic variables were associated with mood state, and hence welfare, in dogs. The
strength of this study design adds weight to the previous findings of Vieira de Castro et al.28. Further research
to investigate the extent to which cognitive biases do or do not vary with factors such as breed, individual
personality, learning or stress coping style of dogs or owner based factors such as attitude or lifestyle would be
valuable, allowing future studies to better control for any confounding factors and also to be less constrained in
case recruitment for those characteristics found to be unrelated to judgement biases.
This study may have some general implications for animal training methods, including in other species. For
example, in line with our findings, training animals to perform desired responses using only positive reinforce-
ment for procedures in laboratory and zoo situations has been shown to have considerable benefits for welfare34,35.
Individual variation in sensitivity to rewarding and aversive stimuli may influence the appropriateness of different
training approaches. For example, just as human patients with clinical depression have been shown to have a
hypersensitive response to p unishment30, more ‘pessimistically’ biased dogs may be more sensitive to application
of punishment, and perhaps require less punitive methods in training than more ‘optimistic’ or less punishment
sensitive individuals. However, given that depressed people also have a reduced sensitivity to reward36, it may
be that reward-based training alone is not sufficient to alter behaviour in animals with a more ‘pessimistic’ bias,
although Beevers et al.37 suggest that punishment sensitivity is a more important component of depression than
reward hyposensitivity. Further research is required to assess the relevance of differences in sensitivity to reward
and punishment on dog behaviour, learning and welfare. Inferences from this study are limited by the fact that
the cognitive bias test used did not, for ethical reasons, involve any aversive stimulus (just lack of reward).
In conclusion, the results of this study have highlighted further the importance of understanding potential
welfare implications associated with training methods used in domestic dogs. The study replicates and strength-
ens the important findings of Vieira de Castro et al.28 through use of a matched pair study design to reduce the
impact of confounding variables and using an increased sample size. We demonstrated that dogs appear to show
a more pessimistic cognitive bias where trainers report using two or more aversive training methods, and this
has important implications for the kind of advice given to dog owners by veterinary surgeons, breeders, behav-
iourists or trainers, particularly for new owners or those acquiring a puppy. Although alternative interpretations
should be considered as discussed above, because of the potential welfare implications of using two or more
punishment-based techniques identified in this paper, we believe that professionals should advise owners to use
the least aversive methods available to them to modify their dog’s behaviour.
Materials and methods
Subject recruitment and matching. Dog owners living in the south-west of the U.K. were recruited to
take part in the study using a number of approaches: those who had indicated willingness to take part in the
further research as part of previous questionnaire s tudies38,39; through veterinary practices, dog training clubs,
forums, breeder lists and breed societies. Owners responding to recruitment calls were contacted initially by
telephone or e-mail, and subjects were selected to take part in the study if they fulfilled the following inclusion
criteria:
Vol.:(0123456789)
www.nature.com/scientificreports/
• Owners were willing for researchers to visit their house on a single occasion for a period of up to 3 h.
• Owners were willing to complete a questionnaire and sign an informed consent form.
• Owners had an area on their property of sufficient size to set out a 3 m × 4 m unobstructed arena for conduct-
ing the judgement bias task.
• Dogs were between 6 months and 12 years of age.
• Dogs were healthy, sound, and able to eat normal commercial dog food.
• Entire females were not in season at the time of testing.
• Dogs were not on any medication likely to influence their behavior.
• Dogs did not show human-directed aggression, for safety during testing.
• Dogs had either been regularly trained using two or more of the methods listed in Table 2 (Aversion-based
training (AT) group), or had been trained with none of the methods listed in Table 2 (Reward-based training
(RT) group) (see Table 2).
Only one dog from each household was selected for inclusion in the study. Initially all recruited dogs fulfilling
these criteria were tested, but as recruitment progressed, dogs were additionally selected based on matching pairs
of cases between the two groups. Criteria by which dogs were pair matched between groups were: specific breed;
age category (6 months to < 18 months; 18 to < 60 months; over 60 months); sex and neuter status (male entire;
male neutered; female entire; female neutered). Appointments were made with recruited owners by telephone,
and each was sent information about the study plus a participation consent form prior to visits. They were also
asked to complete a short questionnaire to verify their previous responses regarding training methods used with
the focal dog. All owners gave informed consent for the use of their dog and data in the study. Each owner and
dog was visited by a pair of researchers on a single occasion.
Judgement bias test. Each dog was tested for judgement bias during the visit, using methodology
described fully in Mendl et al.18, and summarized here. A 4 m × 3 m area was measured and set up as shown in
Fig. 3. Masking tape was used to indicate the positions for placement of food bowls during training and testing.
The dog was positioned at the start location, behind an opaque barrier, before each training and test trial. During
the training phase, dogs were taught to discriminate between a single food bowl placed either on the right or left
side of the arena through rewarding approach to one location with a small amount of food and not the other. The
rewarded (Positive; P) and unrewarded (Negative; N) sides were randomly allocated to left or right between dogs
but stayed consistent within subjects. Cues other than spatial location were controlled for by making the same
preparatory noises in ‘baiting’ each bowl and adding but removing food from the negative side to reduce any
potential difference in olfactory cues. The time taken for dogs to reach each goal from the starting position was
measured in seconds. Dogs were given up to 30 s to reach the target—where they did not reach the goal location,
they were given a maximum latency score of 30 s. After each trial, dogs were returned behind the screen whilst
the bowl was baited for the next trial. Positive and Negative trials were presented in a pseudo-random order,
such that each side was not presented more than twice consecutively, to reduce the risk of side biases developing.
The criterion for moving to the test phase was all latencies to P less than all latencies to N for six consecu-
tive trials. For the testing phase, dogs were presented with three unrewarded ‘probe’ trials with bowls placed in
ambiguous locations, spaced at equal quartiles between the positive and negative locations (near positive (NP),
middle (M) and near negative (NN); Fig. 3). The probe locations were presented in in the following order: M, NP,
NN, NP, NN, M, NN, M, NP (each location was presented first, second or third in each block of three test trials).
The nine test trials were each separated by four training trials. These training trials were also pseudo-randomly
ordered, ensuring two negative and two positive trials in each block to avoid extinction of the response. The
latency to each probe and training trial was recorded. Greater latencies in probe trials were interpreted as the
dog judging the bowl to be less likely to contain food (and hence a more ‘pessimistic’ cognitive bias), and faster
speeds a greater anticipation of food (and hence a more ‘optimistic’ cognitive bias).
At the end of testing, dogs were presented with a bowl containing food at the negative goal location and
the latency recorded. This was to eliminate the possibility that dogs were responding in trials based on direct
interpretation of olfactory cues, rather than expectation of reward. Methods were carried out in accordance with
relevant guidelines and regulations, and the methodology was approved by the local University of Bristol Ethical
Review Board (UIN UB/08/003) for both human and dog participants.
Vol:.(1234567890)
www.nature.com/scientificreports/
Statistical analysis. Any tested subjects without pair-matched controls were removed from the dataset.
Mean latencies to reach the P and N locations presented during the test phase (N = 32 presentations) were calcu-
lated, as were mean latencies to each of the three probe trials (NP, M and NN; 3 presentations of each). Latencies
to probe goal locations, number of trials to reach criterion, and latency to reach the final rewarded presentation
at the negative goal location were all tested for normality. None were normally distributed even after transforma-
tion hence non-parametric tests were used for all analyses.
Latencies to each probe were compared between dogs where the positive goal was located on the left and
those where it was on the right, to check for side biases, using a Mann–Whitney U test. The latency to the final
rewarded bowl in the negative location was compared with the mean latency to the negative goal during the
test phase, to check for the potential influence of olfactory cues on responding, using a Wilcoxon signed rank
test. The number of trials to reach criterion for testing was compared between AT and RT groups to test for dif-
ferences in ability to learn the spatial discrimination task with a Wilcoxon matched pairs test. The AT and RT
groups were also compared for mean latency to each probe goal location using Wilcoxon matched pairs tests.
All analyses were carried out using IBM SPSS Statistics v19. An open access datafile is available at: https://doi.
org/10.5061/dryad.312s0.
Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding
author on reasonable request.
References
1. Arhant, C., Bubna-Littitz, H., Bartels, A., Futschik, A. & Troxler, J. Behaviour of smaller and larger dogs: Effects of training methods,
inconsistency of owner behaviour and level of engagement in activities with the dog. Appl. Anim. Behav. Sci. 123, 131–142 (2010).
2. Dotson, M. J. & Hyatt, E. M. Understanding dog-human companionship. J. Bus. Res. 61, 457–466 (2008).
3. Bland, I. M., Guthrie-Jones, A., Taylor, R. D. & Hill, J. Dog obesity: Veterinary practices’ and owners’ opinions on cause and
management. Prev. Vet. Med. 94, 310–315 (2010).
4. Blackwell, E. J., Twells, C., Seawright, A. & Casey, R. A. The relationship between training methods and the occurrence of behavior
problems, as reported by owners, in a population of domestic dogs. J. Vet. Behav. 3, 207–217 (2008).
5. Hiby, E. F., Rooney, N. J. & Bradshaw, J. W. S. Dog training methods: Their use, effectiveness and interaction with behaviour and
welfare. Anim. Welf. 13, 63–69 (2004).
6. Skinner, B. F. The Behavior of Organisms: An Experimental Analysis (Appleton-Century, 1938).
7. Blackwell, E. J., Bolster, C., Richards, G., Loftus, B. & Casey, R. A. The use of electronic collars for training domestic dogs: Estimated
prevalence, reasons and risk factors for use, and owner perceived success as compared to other training methods. BMC Vet. Res.
8, 93. https://doi.org/10.1186/1746-6148-8-93 (2012).
8. Herron, M., Shofer, F. & Reisner, I. Survey of the use and outcome of confrontational and non-confrontational training methods
in client-owned dogs showing undesired behaviors. Appl. Anim. Behav. Sci. 117, 47–54 (2009).
9. Polsky, R. H. Electronic shock collars: Are they worth the risks?. J. Am. Anim. Hosp. Assoc. 30, 463–468 (1994).
10. Reisner, I. R., Shofer, F. S. & Nance, M. L. Behavioral assessment of child-directed canine aggression. Inj. Prev. 13, 348–351 (2007).
11. O’Sullivan, E. N., Jones, B. R., O’Sullivan, K. & Hanlon, A. J. The management and behavioural history of 100 dogs reported for
biting a person. Appl. Anim. Behav. Sci. 114, 149–158 (2008).
12. Schilder, M. Training dogs with help of the shock collar: Short and long term behavioural effects. Appl. Anim. Behav. Sci. 85,
319–334 (2004).
13. Horvath, Z., Doka, A. & Miklosi, A. Affiliative and disciplinary behavior of human handlers during play with their dog affects
cortisol concentrations in opposite directions. Horm. Behav. 54, 107–114 (2008).
14. Mason, G. & Mendl, M. Why is there no simple way of measuring animal welfare?. Anim. Welf. 2, 301–319 (1993).
15. Paul, E. S., Harding, E. J. & Mendl, M. Measuring emotional processes in animals: The utility of a cognitive approach. Neurosci.
Biobehav. Rev. 29, 469–491 (2005).
16. Mendl, M., Burman, O. H. P. & Paul, E. S. An integrative and functional framework for the study of animal emotion and mood.
Proc. R. Soc. B 277, 2895–2904 (2010).
17. Nettle, D. & Bateson, M. The evolutionary origins of mood and its disorders. Curr. Biol. 22, 712–721 (2012).
18. Mendl, M. et al. Dogs showing separation-related behaviour exhibit a “pessimistic” cognitive bias. Curr. Biol. 20, R839–R840
(2010).
19. Mendl, M. & Paul, E. S. Animal affect and decision-making. Neurosci. Biobehav. Rev. 112, 144–163. https://d oi.o
rg/1 0.1 016/j.n
eubi
orev.2020.01.025h (2020).
20. Harding, E. J., Paul, E. S. & Mendl, M. Animal behavior - cognitive bias and affective state. Nature 427, 312–312 (2004).
21. Mendl, M., Burman, O. H. P., Parker, R. M. A. & Paul, E. S. Cognitive bias as an indicator of animal emotion and welfare: Emerging
evidence and underlying mechanisms. Appl. Anim. Behav. Sci. 118, 161–181 (2009).
22. Brydges, N. M., Leach, M., Nicol, K., Wright, R. & Bateson, M. Environmental enrichment induces optimistic cognitive bias in
rats. Anim. Behav. 81, 169–175 (2011).
23. MacLeod, C., Rutherford, E., Campbell, L., Ebsworthy, G. & Holker, L. Selective attention and emotional vulnerability: Assessing
the causal basis of their association through the experimental manipulation of attentional bias. J. Abnorm. Psychol. 111, 107–123
(2002).
24. Richter, S. H. et al. A glass full of optimism: Enrichment effects on cognitive bias in a rat model of depression. Cogn. Affect. Behav.
Neurosci. 12, 527–542 (2012).
25. Neville, V. et al. Pharmacological manipulations of judgement bias: A systematic review and meta-analysis. Neurosci. Biobehav.
Rev. 108, 269–286. https://doi.org/10.1016/j.neubiorev.2019.11.008 (2020).
26. Lagisz, M. et al. Optimism, pessimism and judgement bias in animals: A systematic review and meta-analysis. Neurosci. Biobehav.
Rev. 118, 3–17. https://doi.org/10.1016/j.neubiorev.2020.07.012 (2020).
27. Burman, O. H. P., Parker, R., Paul, E. S. & Mendl, M. A spatial judgement task to determine background emotional state in labora-
tory rats, Rattus norvegicus. Anim. Behav. 76, 801–809 (2008).
28. de Castro, A. C. V. et al. Does training method matter? Evidence for the negative impact of aversive-based methods on companion
dog welfare. PLoS ONE 15(12), e0225023. https://doi.org/10.1371/journal.pone.0225023 (2020).
Vol.:(0123456789)
www.nature.com/scientificreports/
29. Gotlib, I. H. & Joormann, J. Cognition and depression: Current status and future directions. Annu. Rev. Clin. Psychol. 6, 285–312
(2010).
30. Eshel, N. & Roiser, J. P. Reward and punishment processing in depression. Biol. Psychiat. 68, 118–124 (2010).
31. Papciak, J., Popik, P., Fuchs, E. & Rygula, R. Chronic psychosocial stress makes rats more “pessimistic” in the ambiguous-cue
interpretation paradigm. Behav. Brain. Res. 256, 305–310 (2013).
32. Rygula, R., Papciak, J. & Popik, P. Trait pessimism predicts vulnerability to stress-induced anhedonia in rats. Neuropsychophar-
macology 38, 2188–2196 (2013).
33. Asher, L., Friel, M., Griffin, K. & Collins, L. M. Mood and personality interact to determine cognitive biases in pigs. Biol. Lett. 12,
20160402 (2016).
34. Laule, G. & Whittaker, M. Enhancing nonhuman primate care and welfare through the use of positive reinforcement training. J.
Appl. Anim. Welf. Sci. 10, 31–38 (2007).
35. Laule, G. Positive reinforcement training for laboratory animals. UFAW Handb. Care Manag. Lab. 8, 206–218 (2010).
36. Gradin, V. B. et al. Expected value and prediction error abnormalities in depression and schizophrenia. Brain 134, 1751–1764
(2011).
37. Beevers, C. G. et al. Influence of depression symptoms on history-independent reward and punishment processing. Psychiatry
Res. 207, 53–60 (2013).
38. Blackwell, E. J., Bradshaw, J. W. S. & Casey, R. A. Fear responses to noises in domestic dogs: Prevalence, risk factors and co-
occurrence with other fear related behaviour. Appl. Anim. Behav. Sci. 145, 15–25 (2013).
39. Casey, R. A., Loftus, B., Bolster, C., Richards, G. & Blackwell, E. J. Inter-dog aggression in a UK owner survey: Prevalence, co-
occurrence in different contexts and risk factors. Vet. Record. 172, 5 (2013).
Acknowledgements
We thank Bethany Loftus, Gemma Richards and Christine Bolster for assistance with data collection and data
entry. We also particularly thank all of the dog owners who assisted us by taking part in the study.
Author contributions
M.N.-O. and S.C. visited owners, ran tests, collected and entered data; R.C. conceived the project, supervised
data collection, collated and analysed data and drafted the paper; M.M. and E.B. supervised data collection,
contributed to study design and paper writing/editing. All authors reviewed the manuscript.
Competing interests
SC was funded with a Wellcome Trust vacation scholarship, MN-O was funded by a grant from the Anthrozool-
ogy Institute. We thank the Wellcome Trust and Anthrozoology Institute for assistance with research costs. EB
was funded by Dogs Trust at the time of the study, RC was at the University of Bristol at the time of the study but
is now employed at Dogs Trust. We thank Dogs Trust for their contribution to authors’ time on this project. MM
is employed by the University of Bristol. No authors have a conflict of interest regarding authorship of this paper.
Additional information
Correspondence and requests for materials should be addressed to R.A.C.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Vol:.(1234567890)
Do aversive-based training
methods actually
compromise dog welfare?: A
literature review
Gulherme Fernandes J, I Anna S Olsson, Vieira de
Castro A C
Originally published in Applied Animal Behaviour Science (2017), 196, 1-12
ABSTRACT
The methods by which dogs are trained vary between methods involving mainly negative
reinforcement and positive punishment (aversive-based methods) and methods based essentially in
positive reinforcement and negative punishment (reward-based methods). However, the use of
aversive-based methods is highly controversial. While some people defend their merits, others are
concerned with their potential negative effect on dog welfare. To date, some studies have been
performed aiming to assess the effects of aversive- and reward-based methods on the welfare and
behaviour of dogs. In the present paper we perform a comprehensive review of those studies with
the aim of characterizing the state of the art of scientific knowledge of the topic. Generally, the
published studies suggest that the use of aversive-based methods is correlated with indicators of
compromised welfare in dogs, namely stress‐related behaviours during training, elevated cortisol
levels and problematic behaviours such as fear and aggression. However, there are a number of
limitations that prevent any strong conclusion from being drawn. First, a considerable proportion of
the studies relied upon surveys rather than on objective measures. Second, they focused on sub-
populations of police and laboratory dogs and, thus, only represent a small portion of dogs
undergoing training. Finally, the empirical studies have concentrated mainly on the effects of
shock-collar training, which is only one of several tools used in aversive-based training, and, in some
studies, the description of the training methodologies lacks details. Here we present a description
of the published studies, discuss their limitations, debate other aspects that, in parallel with the
nature of the training methods, may affect dog welfare, and point to future directions for research
on the topic.
INTRODUCTION
Since domestication, dogs have been bred and selected for a variety of functions to support
humans in their activities. Among many others, dogs are used for herding and guarding of livestock,
for hunting, for search and rescue of people, for drug detection or as guide-dogs for blind people. In
order to develop performance in the activities they are assigned to, these dogs are subjected to
some sort of training (Coppinger and Coppinger, 2001). Nowadays, although dogs are mainly
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
adopted for their companionship rather than for working support, they also require some sort of
training (Reid, 2007).
For instance, dogs are trained for household rules (e.g., toilet training, not jumping to the
couch/bed, not chewing furniture), for general obedience (e.g., to walk on leash without pulling, to
come when called) or for more complex behaviours for hobby purposes (e.g., agility).
During training, the behaviour of dogs is modified through different learning processes. The most
commonly used is operant conditioning, through which the probability of occurrence of a given
behaviour is increased or decreased by arranging its consequences (Skinner, 1938). Depending on
whether the consequence of the behaviour is the addition or removal of a stimulus and on whether
the probability of occurrence of that same behaviour increases or decreases, operant conditioning is
divided in four quadrants: (a) positive reinforcement (R+), where a behaviour results in a (pleasant)
stimulus and the probability of its occurrence increases; (b) negative reinforcement (R−), where a
behaviour removes an (unpleasant) stimulus and the probability of its occurrence increases; (c)
positive punishment (P+), where a behaviour results in an (unpleasant) stimulus and the probability
of its occurrence decreases; and (d) negative punishment (P-) where a behaviour removes a
(pleasant) stimulus and the probability of its occurrence decreases. In general terms, in operant
conditioning, animals learn to perform specific behaviours because these result in the avoidance of
unpleasant stimuli and/or in the achievement of pleasant stimuli.
The way in which dogs are trained ranges widely from methods involving mainly negative
reinforcement and positive punishment (aversive-based techniques) to methods based essentially
in positive reinforcement and negative punishment (reward-based techniques). In dog training,
pleasant stimuli normally comprise vocal praise, stroking, food, interactive play and social contact,
whereas unpleasant stimuli involve vocal and physical reprimands or inflicted pain or discomfort
through tools specially designed for that goal, as, for example, shock and choke collars.
Learning through negative reinforcement and positive punishment has an important adaptive
value, in that these mechanisms help animals to avoid and escape from dangerous or harmful
situations. However, the use of aversive-based techniques in dog training has been mainly backed
by the traditional view that dogs, like wolves, are pack animals, whose social organization
encompasses a linear hierarchy, and that their behaviour is driven by a desire to be “dominant” or
the “alpha” in the pack. This view extends to the dog-human relationship, in the sense that dogs are
believed to view humans as a member of their pack. It has been widespread in the popular view as
well as in the literature that, in order to prevent disobedience and aggression, humans must be
hierarchically dominant over dogs and that the way to exert such dominance is to train them using
confrontational and coercive methods (e.g., Landsberg et al., 2003). However, the last decade or so
saw the emergence of a heated discussion on the validity and relevance of the dominance model
regarding dog and wolf social behaviour and also regarding the human-dog relationship. Whereas
the merits of the dominance model are still defended by some authors (e.g., Schilder et al., 2014), it
has been extensively questioned by others (e.g., Bradshaw et al., 2016, Coppinger and Coppinger,
2001, Yin, 2009). Importantly, the idea that dogs and even wolves form linear hierarchies has been
challenged, as well as the notions that dogs view humans as members of their packs and that
humans should adopt an “alpha dog” role (e.g., Bradshaw et al., 2016, Coppinger and Coppinger,
2001, Yin, 2009). As a consequence, the use of gentler techniques to train dogs, centered in the use
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
of positive reinforcement, has been gaining ground (e.g., Yin, 2009) and, in parallel, the use of
aversive‐based techniques is becoming more and more controversial.
Besides the criticism of the validity of the dominance model, aversive‐based techniques have also
been questioned for their potential negative effects on dog welfare. Several animal welfare,
behaviour, training, canine and veterinary organisations all over the world have launched public
statements discouraging the use of aversive-based methods on dog training (e.g., American
Veterinary Society of Animal Behavior, 2007; Welfare in Dog Training, n.d.). Some aversive-based
tools have indeed been legally banned in some countries. For example, shocks collars are not
allowed in Austria, Denmark, Finland, Germany and Wales (Companion Animal Welfare Council,
2012). The major arguments are that aversive techniques can cause physical damage (e.g., the use
of choke collars may cause injuries to the trachea or increase intraocular pressure) and lead to a
number of undesirable behavioural consequences. There is some support for such consequences in
early research into the undesirable side effects of using aversive methods for changing (human and
non-human animal) behaviour. For example, punishment was shown to lead to negative emotional
responses such as fear and anxiety and, consequently, to disturbances in learning and performance.
Additionally, it was found that punishment can lead to the general suppression of all behaviours,
including behaviours that can be of interest. A third major disadvantage is the fact that punishment
can lead to aggressive responses either towards the person applying the aversive stimulus or
whomever appears to be around (e.g., Azrin and Holz, 1966, Mazur, 2006). In the most extreme
case, exposure to unpredictable and uncontrollable aversive stimuli can lead to a long-term,
debilitating, depressive-like state in both human and non-human animals, referred to as learned
helplessness (Maier and Seligman, 1976). On the other hand, advocates of aversive‐based
methods assert that they are the most effective means to correct certain behaviours, such as poor
recall or predatory behaviour and that, by giving dogs more freedom to explore the environment in
safety, they indeed improve quality of life and welfare of dogs (e.g., Electronic Collar Manufacturers
Association, n.d.; Gellman, 2012).
The claims of the opponents of aversive-based dog training methods appear then to have some
support from these early data on the potential undesirable side effects of aversive methods. A
recent review paper also argues that aversive-based methods compromise the mental and physical
health of dogs (Ziv, 2017). Additionally, because in the view of author, this same literature shows
that reward-based methods appear to be more effective than aversive‐based ones, he defends
that the implementation of the latter should be avoided, and that some practices should even be
made illegal. On the other hand, the Companion Animal Welfare Council, in a review of the
arguments in favour and against the use of electronic pulse training aids in companion animal
training concluded that there is a lack of scientific evidence on the matter (Companion Animal
Welfare Council, 2012). In order to help solving the current controversy over the use of these
training methods and to draw policy decisions on the matter, solid scientific evidence is needed.
This evidence needs to come from research informed by expertise in animal behaviour and welfare
and providing statistically valid results.
The goal of the present paper is to perform a comprehensive review of the scientific literature on
the effects of different training methods in the welfare and behaviour of dogs. For the stronger level
of evidence that statistically significant results provide, in the present paper only such findings are
reported and discussed. With this review, we intend to provide the scientific community, the dog
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
training and behaviour professionals and the policy makers with a comprehensive picture of the
present scientific knowledge of the topic.
METHODS
An advanced search was conducted on the ISI Web of Science® database with the query TS =
“dog*” AND “train*” AND (“behavio$r*” OR “stress” OR “perform*” OR “health”). Results were
refined to include original research articles, reviews, case reports, and reports written in English and
Portuguese, published in journals of veterinary sciences, zoology, behavioural sciences, psychology,
and anthropology. From the 913 references resulting from the search and after a triage (Fig. 1), we
selected the articles addressing the effects of training methods on the welfare and behavioural
problems of dogs. A case report was excluded from the sample, because it reports an isolated
incident, which clearly limits the strength of the conclusions that can be drawn. Another study was
excluded, because it does not allow disentangling the effect of the training method from the effect
of a socialization programme, which was applied together with reward-based training. During the
course of the analysis, four additional relevant articles that did not result from our search were
found and included, ending with a sample of 14 articles.
Advanced search
on ISI Web of
Science®
32 abstracts further
analysed
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
Figure 1. Triage process. The first triage step involved reading each of the 913 abstracts and
excluding all papers that were not related to dog training and behaviour. The second triage step
excluded all papers that were not related to the effects of dog training methods on dog welfare and
behavioural problems, one case study, and one empirical study, and included four papers related to
effects of dog training methods on dog welfare that did not result from the initial web search.
Consequently, 14 articles were reviewed.
Nine of the reviewed papers report studies in which behaviour and welfare parameters were directly
measured by the researchers. These include both studies where data were collected in and around
the training situation and studies of dogs which had previously been trained with different
methods. The different studies are presented in detail in Table 1.
Different kinds of collars are used in aversive-based dog training. One is the electronic training
collar (also known as shock collar or e‐collar), which consists of a collar-mounted device capable of
delivering an electric shock to the neck (that can vary in intensity), causing discomfort and pain.
Another one, the pinch collar, is a metal collar with prongs in its inner face, and it is used to
implement P+ or R− by applying pressure on the neck through the prongs, which causes discomfort
and/or pain. A third one is the lemon spray collar, which is used to apply P+ in the form of a spray of
lemon juice to the dog’s face.
The behaviour of two groups of police dogs, previously trained for protection work either with
shock collars or without this device, was analysed during free walking on leash, obedience work and
protection work, in a study conducted by Schilder and van der Borg (2004). Dogs previously trained
with shock collars exhibited more stress-related behaviours than the control group, both within and
outside the training context, as well as in training (obedience and protection work) and non-training
activities (free walking). In another study, Christiansen et al. (2001) evaluated the effects of
exposure to electric shocks during training of hunting dogs to not attack sheep. Dogs underwent a
test to evaluate their reaction to sudden encounters with different stimuli, in order to assess fear
and anxiety. The test was performed twice: immediately before and one year after training. The
authors found no general effect of the use of shock collars on fear and anxiety. To investigate the
stress levels resulting from different ways of using the shock collar, Schalke et al. (2007) studied
three groups of dogs. In the Aversion group, dogs received a shock if they touched a dummy prey,
in the Here group dogs received a shock if not obeying a previously trained recall command and in
the Random group the electric shock was delivered arbitrarily. The results showed significant
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
differences in the cortisol levels of the three groups, with the Random group displaying the highest
levels, followed by the Here group and with the lowest cortisol levels for the Aversion group. The
authors related these differences to the differences in predictability and controllability in how
shocks were administered in the three situations, with shocks being most predictable and
controllable for the dogs in the Aversion group and completely unpredictable and uncontrollable for
dogs in the Random group. They stated that the lowest cortisol levels shown by dogs trained for
aversion to prey could be explained by the predictability (i.e., the electric pulse was delivered every
time the animal touched the prey) and controllability of the electric pulse (i.e., by avoiding touching
the prey, dogs avoided the electric shock). In turn, the highest cortisol levels shown by dogs to
whom electric shocks were delivered randomly could be explained by the fact that they could not
predict nor control the stimulus. Finally, the intermediate levels of cortisol shown by dogs trained to
respond to a recall cue were probably due to the fact that these dogs could predict but not control
the electric shocks. As hypothesized by the authors, because for this group the recall cue had been
previously trained but not in conjunction with the prey, dogs were able to predict the shocks
because they associated punishment with non-compliance to return to the handler, but they were
not able to control their initial reaction of chasing the prey. In one study, police dogs were trained
with pinch and shock collars through P+ to maintain a “heel” position with distractions. If the dogs
failed to maintain the position, P+ was applied. The results showed that dogs vocalized more often
in response to the shock collar than to the pinch collar. There was no difference in cortisol levels
between the two groups (Salgirli et al., 2012). Steiss et al. (2007) analysed the effect on
physiological parameters of stress of using shock and lemon spray collars for reducing barking in
dogs relinquished to shelters by their owners. Dogs were divided in three groups: shock collar
group, lemon spray collar group, and control group (half of dogs wearing an inactivated shock
collar, and the other half wearing an inactivated lemon spray collar). The results showed no
significant differences in cortisol levels and in ACTH levels between the three groups.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
Breed(s),
Number
Authors gender and Origin Methods/Treatment Dependent variable(s) Summary of results
of animals
age
Christiansen et Norwegian Not clear if N=114 Sheep confrontation test: Behaviour during the path test (dog Difference between year 2 and year 1 in
al. (2001) elkhound companion or walked along a 100-m path where it was performance (distance or time) in the path
(grey) working Dog wearing a shock collar released in exposed to sudden encounters with: rag, test:
(hunting) dogs a fenced area with a sheep flock unfamiliar human, bundle of cans thrown
English upon stones, single tied sheep): Dogs exposed to electric shocks showed a higher
setter If dog came within a distance of 1-2
m from the sheep, an electric shock increase in object discovery distance than dogs not
- Object contact latency exposed to electric shocks (p=0.003)
Hare hunting was applied
dog (Finnish - Object discovery latency
stoever, Dogs exposed to electric shocks maintained the
Hamilton - Human contact latency human contact latency, which was significantly
Some dogs were exposed to electric
stoever, shocks, other were not (depending on - Human discovery latency different from dogs not exposed to electric shocks,
Dunker, whether they came within 1-2 m from which showed an increase in the latency
Drever) the sheep or not) - Noise recovery time (p<0.001)
Gender and - Degree of interest in lone sheep
No significant differences were found between
age not groups for the remaining measures
- Reaction latency towards sheep
reported
Cooper et al. Various Companion N=63 Recall training in the presence of Behaviour during training (44 behaviours Time spent:
(2014) breeds livestock and other dogs: - e.g. panting, tail position and paw lift) In a tense behavioural state
n=21 A (24.6%) > C (3.96%) (p=0.007)
29 M (8; 9; Group A – Training with shock collars Salivary cortisol (before, during and after Yawning
12); performed by professional trainers training) A (0.9%) > C (0.19%) (p<0.01)
referred by the Electronic Collar Sniffing and interacting with the environment
34 F (13; 12; Manufacturers Association Urinary cortisol (periods of sampling not
9) reported) A (12.1%) < C (22.1%) (p<0.01)
Group B - Training without shock B (14.3%) < C (22.1%) (p=0.01)
Average age: collars performed by professional
46 months No significant differences were found between
trainers referred by the Electronic
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
Collar Manufacturers Association groups for other behaviours
Group C - Training without shock Salivary cortisol for all sampling periods (before,
collars performed by professional during and after training):
trainers from the Association of Pet C > B (p=0.001)
Dog Trainers, UK (focus on reward-
based training) No significant differences were found between
group A and B or C
Urinary cortisol: no significant differences were
found
Deldalle and Various Companion N=50 Obedience training with: Behaviour during training (6 behaviours Percentage of dogs displaying the behaviour
Gaunet (2014) breeds related to stress - e.g. mouth licking and during:
School A: Group A (School A) – R+ yawning, posture, gaze toward the owner,
School A: n=24 and body and gaze avoidance) Sit command
11 M; 13 F Group B (School B) – R- Mouth licking
1 - 7 years School B: R+ (8%) < R- (38%) (p=0.019)
old; average n=26 Yawning
of 2.88 years R+ (0%) < R- (23%) (p=0.023)
old Low posture
R+ (8%) <R- (46%) (p=0.0001)
School B: Gazing at owner
16 M; 10F 8 R+ (88%) >R- (38%) (p<0.0001)
months - 6 At least 1 out of 6 stress-related behaviours
years old; (mouth licking, yawning, scratching, sniffing,
average of shivering, whining):
2.41 years R+ (8%) < R-(65%) (p<0.0001)
old
Walking on leash
Gazing at owner
R+ (63%) >R- (4%) (p<0.0001)
No significant differences were found between
groups for other behaviours
Haverbeke et Belgian Military N=33 Dogs trained for obedience and Behaviour (body posture during training Body posture
al. (2008) shepherd protection with: (from very low - scored with -3, to high - Aversive stimuli (-0.22±0.19) < Appetitive stimuli
scored with 2); mouth-licking, tongue out, (0.49±0.09; p<0.01)
German Appetitive stimuli (R+) yawning, lifting front paw, replacement
shepherd behaviour (shacking and replacement No significant differences were found between
Aversive stimuli (P+ and R-) groups for other behaviours
26 M; 7F sniffing), jumping, and opening and
All dogs were trained with a mixture
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
3.06 ± 0.21 of appetitive and aversive stimuli. closing mouth)
years old
Salgirli et al. Belgian Police N=42 Dogs trained to “heel” (walking in Stress-related behaviours during training Percentage of dogs showing:
(2012) Malinois parallel next to the handler) with (reactions of ear, tail, and joint parts and Vocalizations
distractions using: vocalizations) Pinch collar < shock collar (p<0.0001)
33 M; 9 F
Shock collar (P+) Salivary cortisol Only descriptive analysis is available regarding the
3 - 10 years (before and after training) behaviour of dogs trained with the quitting signal
old Pinch collar (P+) (absolute and relative values- difference (only four dogs learned the behaviour)
Quitting signal (P-) between maximum and resting cortisol
values) No significant differences were found between
(Same dogs exposed to the three groups for other stress-related behaviours
treatments)
Relative cortisol values
Quitting signal > pinch collar (p = 0.0294)
During training, a decoy with a
protection sleeve and a whip in his
hand tried to provoke dogs to leave
the position.
Schalke et al. Beagle Laboratory N=14 Dogs trained with shock collar (P+): Physiological measures taken in the Heart rate: no differences between the 3 groups
(2007) experimental environment 4 weeks after
9 M; 5 F Group A: aversion to prey (shocks the end of the training sessions: Salivary cortisol
delivered when dog touched the prey)
1.5 – 2 years Heart rate - R > H > A (p<0.001)
old Group H: recall on cue (shocks
delivered when dogs disobeyed the Salivary cortisol
recall cue during hunting)
Group R: shocks delivered arbitrarily
Schilder and Belgian Police N=46 Dogs previously trained for protection Behaviour (ears, tail and body positions, Frequency of behaviours during:
van der Borg Malinois work: and other 31 behaviours (e.g. panting,
(2004) crosses With shock yawning and crouch)) during: Free walking
collar Group A: With shock collar Low ear position
Pure bred (n=16) (Treatment group) - Free walking on leash (in a park and in A > B (p=0.006)
Belgian the training grounds)
Malinois Without Group B: Without shock collar Obedience work
shock (Control group) - Obedience work (in a park and in the Low ear position
German collar training grounds) A>B (p=0.041)
Tongue flicking
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
Shepherd (n=15) - Protection work (in the training A>B (p=0.006)
grounds) Paw lifting
Rottweiler A>B (p=0.009)
Gender and Protection work
age not Low ear position
reported A>B (p=0.004)
Paw lifting
A>B (p=0.007)
Walking with flexed limbs
A>B (p<0.05)
No significant differences between groups were
found for other behaviours
Steiss et al. Various Shelter dogs N=21 Dogs trained not to bark in the Physiological measures taken (post- No significant differences were found between
(2007) breeds (relinquished presence of barking stimulus (an training) during 3 consecutive days in two groups for cortisol or ACTH levels
by their unfamiliar dog on leash) with: different weeks:
11 M; 10F owners)
Shock collar (P+): device activated Plasma cortisol
10 - 64 when both sound and vibration of
months old; barking were detected ACTH
M=20
months; Lemon spray collar (P+): device
S.D.=14 activated when vibration of barking
months was detected
Another group of dogs used either
turned off shock or lemon spray
collars (control group)
Rooney and Various Companion N=52 Dogs previously trained for basic Behaviour during: Dogs trained mainly with physical punishment:
Cowen (2011) breeds education (e.g., toilet training) and Were less interactive during play (p=0.015)
obedience (e.g., walk to heel) with: Non-object play Contacted and interacted less with the
18 M; 34 F Owners played with their dog without any experimenter (p=0.007)
Punishment-based object
5 months to - No significant correlations were found between
14 years old; Reward-based methods Relaxed social test other tests and type of training method
median=4 An experimenter was sat down on a chair
years old (Same dogs could be exposed to both in the presence of the owners. If the dog
treatments) approached the experimenter, it was
petted.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
Table 1 - Description of the subjects, methodology and results of the observational and experimental studies that assessed the effect of different training methods in the behaviour and
welfare of dogs.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
The effects of aversive-based versus reward-based methods
More recently, attention has been increasing at directly comparing the effects of aversive‐based
techniques and reward-based techniques on dog welfare.
One study addressed the immediate reaction to aversive and appetitive stimuli. Haverbeke et al.
(2008) studied pairs of military dogs and handlers, where the handlers used a mix of aversive
(mainly pulling on the leash and hanging by the collar) and appetitive stimuli (mainly stroking and
patting) during training. During performance of obedience and protection work exercises, the dogs
exhibited a significantly lower body posture after aversive stimuli were presented than after the
presentation of appetitive stimuli, suggesting that the former generated more stress than the
latter.
Four studies compared the reaction of dogs systematically trained with aversive-based methods
with dogs systematically trained with reward-based methods. In the study by Salgirli et al. (2012)
reported in the previous section, the authors also compared the effects of aversive-based training
methods (shock and pinch collars) with the effects of a reward-based method (a quitting signal).
The quitting signal consisted of a conditioned signal for the withdrawal of a reward as a
consequence of an undesirable behaviour (P−). The authors report that the quitting signal resulted
in significantly higher cortisol levels than the pinch and the shock collar. However, this result is
difficult to interpret, given that only four dogs learned the quitting signal. For this same reason the
authors excluded the dogs that did not learn the quitting signal from the behavioural analysis,
making it impossible to make a statistical comparison between the groups. It is not clear why and
how they performed statistical analysis for the cortisol levels. Cooper et al. (2014) conducted an
experimental study with companion dogs trained with aversive‐based or reward-based methods.
The dogs were allocated to three groups which were all trained by professional trainers: training by
certified shock collar trainers, with and without shock collars, and training by certified reward-based
trainers. The results showed that dogs trained with shock collars spent significantly more time
exhibiting stress-related behaviours than dogs trained using reward-based methods. Deldalle and
Gaunet (2014) studied the behaviour of companion dogs as they were being trained by their owners
at two different training schools, one using R+ methods and the other using R−. The authors found
that dogs trained with R− showed more behaviours related with stress and also gazed less often to
owners than dogs trained with R+. Finally, Rooney and Cowan (2011) assessed the influence of the
training methods and owner behaviour on the behaviour of dogs by studying dog-owner pairs in
standard situations. They found that dogs belonging to owners reporting the use of more physical
punishment for basic education and obedience training were less interactive during play, and
contacted and interacted less with the experimenter during a relaxed social test.
Another set of studies surveyed the relationship between the behaviour and attitudes of owners in
educating and training and the behaviour of their dogs, mainly the occurrence of problematic
behaviours such as aggression, fear and separation-related problems. In these studies, the
researchers did not observe dog behaviour directly but instead based their data on owner reports.
The different studies are presented in detail in Table 2.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
Breed(s), Number of
Authors Origin Treatment Dependent variable(s) Summary of results
gender and age responses
Arhant et al. Various breeds Companion 1276 out of Dogs trained with: Problematic behaviours: Correlations between
(2010) 5000 problematic behaviours and
47% M; 53% F Punishment (e.g. leash jerk; scruff Aggression
training techniques:
shake/alpha roll)
0.25 to 17.5 Excitability
Higher aggression and
years old; M Reward-based responses to unwanted
Anxiety excitability scores with higher
= 6.3 ± 3.9 years behaviour (e.g. comfort with
frequency of punishment
old petting/speaking; distract with Fear
food/play) Higher anxiety and fear scores
with higher frequency of
Rewards (e.g. praise verbally; food)
punishment for small dogs
(dogs weighing less than 20 Kg)
Higher aggression and
excitability with higher
frequency of reward-based
methods to unwanted
behaviour
Lower aggression, excitability,
anxiety and fear scores with
higher frequency of rewards
No other significant
correlations between type of
technique and dog behaviour
were found
Blackwell et al. Various breeds Companion 192 out of 250 Dogs trained with: Problematic behaviours (36, Correlations between
(2008) e.g. fear; aggression) problematic behaviours and
52% M; 48% F R+
training techniques:
1 to 15 years old; R-
Lowest attention-seeking, fear
median = 5 years
P+ and aggression scores with the
old use of R+ alone
Combination of two or three
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
techniques Highest attention-seeking
scores with the use of both R+
and R-
Highest fear score with a
combination of all training
techniques
Highest aggression score with a
combination of P+ and R+
Highest aggression and fear
scores with P+ together with
any other method
No other significant
correlations between type of
technique and problematic
behaviour were found
Casey et al. (2014) Various breeds Companion 3897 out of Dogs trained with: Aggression towards people or The use of R- or P+ was
14566 hide and run away from correlated with an increased
48% M; 51% F R+ and/or P-
people (5, e.g. aggression risk of aggression towards
6 months to 17 R- and/or P+ towards familiar people; hide members of family/household
years old; M = 4 or run away from unfamiliar and towards unfamiliar people
years old people) outside of the house
No other significant
correlations between type of
technique and problematic
behaviour were found
Christiansen et al. Norwegian Not clear if 112 Sheep confrontation test: Dog behaviour: Statistical analysis of the data
(2001) elkhound (grey) companion were not reported
Dog wearing a shock collar released in Fear of gunshots
or working
English setter a fenced area with a sheep flock Descriptive data reported:
(hunting) Fear of unfamiliar people and
2 dogs not exposed to electric
Hare hunting dog dogs If dog came within a distance of 1-2 m dogs
shocks showed increased fear
(Finnish stoever, from the sheep, an electric shock was
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
Hamilton applied Aggression towards people to unfamiliar people, whereas 1
stoever, Dunker, and dogs dog exposed to electric shocks
Drever) presented this behaviour
During the test, 13 dogs received
Gender and age 1 dog not exposed to electric
electric shocks and 99 did not Questionnaire conducted one
not reported shocks showed increased fear
year after the sheep
to unfamiliar dogs, whereas
confrontation test none of the dogs exposed to
electric shocks presented this
behaviour
2 dogs not exposed to electric
shocks showed increased
aggression towards dogs,
whereas none of the dogs
exposed to electric shocks
presented this behaviour
Data regarding fear of gunshots
and aggression towards people
were not reported
Herron et al. Various breeds Companion 140 out of 217 Dogs trained with: Undesired behaviours (5, e.g. Owners whose dogs were
(2009) aggression towards people; trained with aversive-based
65% M; 35% F Aversive-based methods with direct
separation anxiety, and other methods reported their dogs
confrontation (e.g. roll dog onto back less common behaviours, e.g. responded with aggression
0.25 to 14 years and hold down; hold dog down on side,
old; M = 4.1 ± 2.8 aggression to cats and
legs extended) Few owners reported
years old barking)
aggression as a response to
Aversive-based methods with indirect
reward-based training
confrontation (e.g. make abrupt sound
to interrupt or correct undesirable The lowest percentage of dogs
behaviour; forcibly expose dog to showing over-excitement was
stimulus that frightens it - i.e., tile trained only with reward-based
floors, noise, people) methods
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
Reward-based methods (e.g. give food No other significant
as reward for desirable behaviour; use correlations between type of
food to trade for item in dog’s mouth) training method and undesired
behaviour were found
(Same dogs could be exposed to
different treatments)
Hiby et al. (2004) Various breeds Companion 326 out of 600 Dogs trained with: Problematic behaviours (13, The frequency of aversive-
e.g. aggression towards based methods used was
54.9% M; 45.1% Punishment-based method
people; barking at dogs; positively correlated with the
F
Reward-based methods separation-related behaviors) number of problematic
1 to 15 years old; behaviors
M = 61±40 Both punishment-based and
reward-based methods The lowest percentage of dogs
months old
showing over-excitement was
trained only with reward-based
methods
No other significant
correlations between type of
training method and
problematic behaviour were
found
Table 2 - Description of the subjects, methodology and results of the survey studies that assessed the relationship between different training methods and problematic or undesired
behaviours in dogs.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de Investigação e Inovação em Saúde in the University of Porto Open Repository For Open
Access to more of our publications, please visit http://repositorio-aberto.up.pt/
In a context of behavioural appointments Herron et al. (2009) found that many dog owners who
used aversive-based methods reported that their dogs responded with aggression. Conversely, few
owners reported aggression as a response to reward-based methods. Similarly, data collected from
questionnaires to dog owners showed a correlation between the use of positive punishment or
negative reinforcement and an increased risk of aggression towards members of family/household
and towards unfamiliar people outside of the house (Casey et al., 2014). Furthermore, Arhant et al.
(2010) found an association between a higher frequency of punishments and increased aggression
and excitability scores in dogs. A similar correlation was also found between punishment frequency
and anxiety and fear scores, but only for small dogs (less than 20 kg). Contrarily, a higher frequency
of rewards was correlated with lower aggression, excitability, anxiety and fear scores. However,
certain types of reward-based methods, namely reward-based responses to unwanted behaviour,
such as comforting dogs with petting or speaking and distracting with food or play, were also
correlated with increased aggression and excitability. Collecting data through a questionnaire to
dog owners, Hiby et al. (2004) found a positive correlation between the frequency of aversive‐
based methods and the number of problematic behaviours reported by the owners. Also based on
data from a questionnaire to dog owners, Blackwell et al. (2008) correlated scores of various
problematic behaviours reported by the owners with different training techniques that owners
reported to use (P+, R− and R+; alone or in a combination of two or three). Generally, it was found
that the number of undesirable behaviours, namely attention-seeking behaviours, fear behaviours
and aggressive behaviours, was positively correlated with the use of aversive-based methods. In
contrast, in the aforementioned study by Christiansen et al. (2001), there was no increase in fear or
aggression towards people or other dogs related to the exposure of electric shocks, as reported by
owners. Although there are a few contradictory results among the reported studies, in general it
was found that the frequency of problematic behaviours correlated positively with the use of
aversive-based training methods.
DISCUSSION
In the present paper, we present a comprehensive review of the scientific literature on the effects of
different training methods on dog welfare and behavioural problems. Generally, the existing
research papers on the topic suggest a correlation between the use of aversive-based training
methods and indicators of compromised welfare and behavioural problems in dogs, but the
evidence at present is not as clear as some advocates in the contemporary dog training discussion
claim.
In three of the studies comparing behaviour in dogs systematically trained with either aversive-
based or reward-based methods, there was an association between aversive-based training
methods and increased stress-related behaviours and/or reduced interactions with humans (Cooper
et al., 2014, Deldalle and Gaunet, 2014, Rooney and Cowan, 2011). The results reported by Salgirli
et al. (2012) are conflicting in that dogs presented higher cortisol levels when trained with the
reward‐based method (quitting signal) as compared to when trained with shock or pinch collars.
However, as reported above, it is not clear which dogs the authors included in this comparison,
since only four were able to learn the quitting signal. If the authors included the cortisol data from
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
all dogs, it is likely that the higher levels shown with the application of the quitting signal were due
to uncertainty and frustration resulting from dogs not understanding what was required from them,
and that this was more adverse than the stress resulting from the use of a shock or pinch collar. In
the studies investigating specifically the effect of shock collars, the existing literature is
inconsistent. The two studies where training with shock collars was directly compared with training
without these devices, Cooper et al. (2014) and Schilder and van der Borg (2004) found more stress-
related behaviours with shock collar training, whereas the other two studies that compared welfare
parameters between dogs exposed to electric shocks and dogs that were not exposed to electric
shocks reported no differences (Christiansen et al., 2001; Steiss et al., 1997). In addition to the
conflicting results, the widely disparate experimental designs of these studies make it speculative to
draw general conclusions. There are also indications that the training method affects dog‐human
relations beyond the training situation itself. Schilder and van der Borg (2004) found increased
stress-related behaviours also outside the training situation, which they suggest implies an
association between the presence of the handler and aversive events. Rooney and Cowan (2011)
found that dogs belonging to owners reporting the use of more physical punishment in training
interacted less with the owner during play, and contacted and interacted less with the unfamiliar
experimenter.
Regarding the relationship between training method and problematic behaviours, the evidence is
contradictory, yet stronger in the direction of a positive correlation between the use of aversive-
based methods and the appearance of problematic behaviours in dogs. From the reviewed studies,
four indicate an association between the use of aversive-based methods and problematic
behaviours (Blackwell et al., 2008, Casey et al., 2014, Herron et al., 2009, Hiby et al., 2004), one
shows no correlation (Christiansen et al., 2001) and another one shows positive correlations
between both the use of aversive-based methods and certain reward-based methods and
problematic behaviours (Arhant et al., 2010). As several of the authors suggest, a potential
explanation for the correlations found between the use of aversive ‐based methods and
problematic behaviours in dogs is that anxiety and conflict resulting from the use of such methods
might lead to the behavioural problems. However, as discussed further below, with this study
design it is not possible to draw conclusions about cause and effect from the correlational
association. The association between reward-based methods and aggression and excitability found
by Arhant et al. (2010) is more difficult to interpret. Nevertheless, and as was advanced by the
authors, it is possible that these reported actions were responses of the owners to the dog’s
aggression or excitability and not the other way around. In support of this view is the fact that such
correlation was found only for reward-based responses to unwanted behaviour.
In summary, together, the results published so far as regards training methods and dog welfare
seem to suggest that aversive‐based training might negatively influence dog welfare and dog-
human interactions. However, and apart from the existence of some contradictory results, there are
a number of important limitations in the existing literature, which prevent strong conclusions from
being drawn. First, a considerable part of the studies, namely those which used surveys as the
methodology for collecting data, relied upon reports of owners rather than on objective measures.
Data collected through surveys reveals, at its best, the perceptions of owners. We cannot exclude
the possibility that these reports do not match reality, both regarding the dog training techniques
that were used and the behaviour of dogs. But perhaps the most important limitation for drawing
conclusions about the effect of training method on welfare is that these studies are correlational.
Indeed, their aim was to investigate possible correlations between problematic behaviours and the
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
training techniques adopted or the attitudes of owners (Arhant et al., 2010; Blackwell et al., 2008,
Casey et al., 2014, Christiansen et al., 2001, Herron et al., 2009, Hiby et al., 2004). Yet, whether dogs
started exhibiting behavioural problems after starting being trained with aversive ‐ based
techniques, or whether owners adhered to this type of training techniques, or showed a more
confrontational approach because dogs had already displayed a problematic behaviour cannot be
revealed with this type of methodology, as was recognized by some authors (Blackwell et al., 2008,
Hiby et al., 2004).
Secondly, most of the empirical studies focused on sub-populations of working and laboratory
dogs, whose training regimes might not recapitulate those of companion dogs regarding frequency,
intensity, duration, exigency and type of behaviours trained (Haverbeke et al., 2008, Salgirli et al.,
2012, Schalke et al., 2007, Schilder and van der Borg, 2004). Additionally, the daily routines, living
quarters and amount of contact with humans are probably also different from those of the typical
companion dog. Whereas the greater potential for controlled and standardized conditions in
working and laboratory dog populations is valuable, companion dogs make up a much larger
proportion of the dogs undergoing training.
A third set of limitations has to do with the training methods. Firstly, the empirical studies
conducted so far have concentrated on the effects of shock-collar training (Christiansen et al., 2001,
Cooper et al., 2014, Salgirli et al., 2012, Schalke et al., 2007; Schilder and van der Borg, 2004; Steiss
et al., 2007), which comprises only a small part of the existing aversive-based techniques. Secondly,
in some studies, the description of the training methodologies lacks details regarding the tools and
reinforcement or punishment strategies that were used, which makes it difficult to draw
conclusions on some of the findings (Haverbeke et al., 2008, Cooper et al., 2014, Rooney and
Cowan, 2011, Schilder and van der Borg, 2004).
Our conclusions differ from those drawn by Ziv (2017), in a recent review paper that encompassed
more or less the same literature as the present paper. In our view, this difference can be explained
by how inclusion criteria were defined and results interpreted. In the present paper we have only
included studies with enough research subjects to allow statistical comparisons, thus excluding case
reports. One case report that came up in our search and that was covered by Ziv (2017) is that of a
dog with brain damage which was probably caused by strangulation during a training session
(Grohmann et al., 2013). The strangulation was a consequence of the use of an aversive-based
training technique called helicoptering or hanging, in which dogs are lifted from the ground and
swung by the collar. This case, although unlikely to be representative of the majority of the
aversive-based training techniques, highlights the danger of this particular technique. Furthermore,
as we have shown, the results of the relatively few studies that exist do not allow strong conclusions
to be drawn, especially not when considering the methodological limitations of the literature.
Whereas some of these limitations are noted by Ziv (2017), neither those nor contradictory results
seem to have been considered in formulating the conclusions. The importance of solid review
papers in which only studies fulfilling certain quality criteria are included, and in which the results
are reviewed in a systematic and unbiased manner is now widely recognized in medicine (Cochrane,
n.d.) and are of course equally important for other fields of research informing practice. The articles
reviewed in the present paper also shed some light on other aspects pertaining to training which
may influence the behaviour and welfare of dogs both negatively and positively. Inconsistency and
poor timing in the application of the reinforcers and/or punishers generate unpredictability and
uncontrollability of stimulus delivery and may therefore lead to stress and, consequently, have a
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
negative influence on dog welfare (e.g. Dess et al., 1983, Destrez et al., 2013, Weiss, 1972).
Additionally, a lower effectiveness of the training method may generate more unpredictability and
uncontrollability, and, consequently, an increase in stress. The existing literature is however
inconsistent, in that three studies suggest a higher efficacy of reward-based methods over aversive-
based methods (Blackwell et al., 2012, Haverbeke et al., 2010, Hiby et al., 2004) whereas another
points in the opposite direction (Salgirli et al., 2012), and a fifth shows no differences between the
two types of methods (Cooper et al., 2014). Over and above not compromising dog welfare, reward-
based methods may also influence dog welfare and dog-human interactions positively. Through
classical conditioning, a positive association may develop between people that are present during
training and the appetitive stimuli used, such as food, praise or play. In support of this view, some
studies showed that dogs and horses trained with reward-based methods presented less stress-
related behaviours and interacted more with familiar and unfamiliar people than those trained with
aversive‐based methods (Sankey et al., 2010, Schilder and van der Borg, 2004).
Based on this review we conclude that although currently there is limited scientific evidence of the
effect of training method on dog welfare, the existing literature indicates that, at least at some
level, aversive-based methods generate stress in dogs. However, further studies are needed to draw
strong conclusions on the topic. In particular, empirical and experimental studies are needed. These
should take into account the entire range of training tools and techniques used in aversive- and
reward-based training methods, in order to appropriately represent the effects of the two
methodologies. Furthermore, to get a comprehensive understanding of the effects of the different
training methods on dogs in general, more studies are needed on companion dogs and they should
include different breeds. Finally, because the effectiveness of the training methods and the
consistency and timing of the stimuli delivery may also influence welfare, future studies should also
take such aspects in account. Although the development of experimental studies on this topic may
raise ethical issues as a consequence of the stress, pain and discomfort that could be intentionally
inflicted to the research animals undergoing aversive-based training, empirical studies can be run in
real-life situations that overcome these same ethical issues. Several dog training schools exist that
work with aversive-based methods, where owners enrol in classes with their dogs. In this context,
both trainers and owners comply with the use of such methods. Hence, studying dogs that are
being trained at these schools is a good way of obtaining the data that is currently lacking without
raising the concern of intentionally allocating random animals to situations that could cause them
pain, stress and/or discomfort.
Finally, in addition to the effects on welfare, the efficacy of training methods is also relevant to
consider in the choice of training method and, regardless of what science will have to say about the
effects of different training methods on dog welfare, it is important to note that the choice of the
training method should not be based only on its effects in animal welfare. Dog training is a purpose-
built tool and, hence, its efficacy should also be considered in the equation. At present, there is also
a lack of scientific evidence on the efficacy of different training methods and it would be relevant to
combine this aspect with research on the effects of different training methods on dog welfare.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
ACKNOWLEDGMENTS
Ana Catarina Vieira de Castro was funded by Fundação para a Ciência e a Tecnologia with an
individual post-doc grant (SFRH/BPD/111509/2015).
The work undergoing the present review paper did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
REFERENCES
1. American Veterinary Society of Animal Behavior , 2007. AVSAB Position Statement The Use of
Punishment for Behavior Modification in Animals.
http://www.vetmed.ucdavis.edu/vmth/local_resources/pdfs/behavior_pdfs/AVSAB_Punishme
nt_Statements.pdf (accessed 17.02.02)
2. Arhant, C., Bubna-Littitz, H., Bartles, A., Futschik, A., troxler, J. 2010. Behaviour of smaller and
larger dogs: effects of training methods, inconsistency of owner behaviour and level of
engagement in activities with the dog. Appl. Anim. Behav. Sci. 123, 131-142.
doi:http://dx.doi.org/10.1016/j.applanim.2010.01.003
3. Azrin, N.H., Holz, W.C., Hake, D. F., 1963. Fixed-ratio punishment. J Exp Anal Behav. 6(2), 141-
148. doi:http://dx.doi.org/10.1901/jeab.1963.6-141
4. Azrin, N.H., Hake, D. F., Holz, W.C., Hutchison, R. R., 1965. Motivational aspects of escape
from punishment. J Exp Anal Behav. 8(1), 31-44. doi:http://dx.doi.org/10.1901/jeab.1965.8-31
5. Azrin, N.H., Holz, W.C., 1966. Punishment, in: Honig, W.K. (Ed.), Operant Behavior: Areas of
Research and Application. Upper Saddle River. NJ: Prentice Hall.
6. Blackwell, E.J., Twells, C., Seawright, A., Casey, R.A., 2008. The relationship between training
methods and the occurrence of behavior problems, as reported by owners, in a population of
domestic dogs. J. Vet. Behav. Clin. Appl. Res. 3, 207–217. doi:
http://dx.doi.org/10.1016/j.jveb.2007.10.008
7. Blackwell, E.J., Bolster, C., Richards, G., Loftus, B. A., Casey, R.A., 2012. The use of electronic
collars for training domestic dogs: estimated prevalence, reasons and risk factors for use, and
owner perceived. BMC Veterinary Research. 8(1), 93. doi:http://dx.doi.org/10.1186/1746-6148-
8-93
8. Bradshaw, J.W.S., Blackwell, E. J., Casey, R.A., 2016. Dominance in domestic dogs - A response
to Schilder et al. (2014). J. Vet. Behav. Clin. Appl. Res. 11, 102–108. doi:
http://dx.doi.org/10.1016/j.jveb.2015.11.008
9. Casey, R. A., Loftus, B., Bolster, C., Richards, G. J., Blackwell, E. J. 2014. Human direct
aggression in domestic dogs (Canis familiaris): Occurrence in different context and risk factors.
Appl. Anim. Behav. Sci. 152, 52-63. doi:http://dx.doi.org/10.1016/j.applanim.2013.12.003
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
10. Christiansen, F. O., Bakken, M., Braastad, B. O. 2001. Behavioural changes and aversive
conditioning in hunting dogs by the second-year confrontation with domestic sheep. Appl.
Anim. Behav. Sci. 72, 131-143. doi: http://dx.doi.org/10.1016/S0168-1591(00)00203-3
12. Cooper, J.J., Cracknell, N., Hardiman, J., Wright, H., Mills, D., 2014. The Welfare Consequences
and Efficacy of Training Pet Dogs with Remote Electronic Training Collars in Comparison to
Reward Based Training. PLoS One 9, e102722.
doi:http://dx.doi.org/10.1371/journal.pone.0102722
13. Companion Animal Welfare Council, 2012. The Use of Electric Pulse Training Aids (EPTAs) in
Companion Animals. http://eprints.lincoln.ac.uk/14640/1/CAWC%20ecollar%20report.pdf
(accessed 17.02.22)
14. Coppinger, R., Coppinger, L., 2001. Dogs: A Startling New Understanding of Canine Origin,
Behaviour & Evolution. New York, NY.
15. Deldalle, S., Gaunet, F., 2014. Effects of 2 training methods on stress-related behaviors of the
dog (Canis familiaris) and on the dog-owner relantionship. J. Vet. Behav. Clin. Appl. Res. 9, 58-
65. doi:http://dx.doi.org/10.1016/j.jveb.2013.11.004
16. Dess, N.K., Linwick, D., Patterson, J., Overmier, J.B., Levine, S., 1983. Immediate and proactive
effects of controllability and predictability on plasma cortisol responses to shocks in dogs.
Behav. Neurosci. 97, 1005–1016.
17. Destrez, A., Deiss, V., Leterrier, C., Boivin, X., Boissy, A., 2013. Long-term exposure to
unpredictable and uncontrollable aversive events alters fearfulness in sheep. Animal. 7, 476–
484. doi:http://dx.doi.org/10.1017/S1751731112001796
20. Grohmann, K., Dickomeit, M. J., Schmidt, M. J., Kramer, M. 2013. Severe brain damage after
punitive training technique with a choke chain collar in a German shepherd dog. J. Vet. Behav.
8, 180-184. doi:https://doi.org/10.1016/j.jveb.2013.01.002
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
21. Haverbeke, A., Laporte, B., Depiereux, E., Giffroy, J.M., Diederich, C., 2008. Training methods
of military dog handlers and their effects on the team’s performances. Appl. Anim. Behav. Sci.
113, 110–122. doi:http://dx.doi.org/10.1016/j.applanim.2007.11.010
22. Haverbeke, A., Messaoudi, F., Depiereux, E., Stevens, M., Giffroy, J.-M., Diederich, C., 2010.
Efficiency of working dogs undergoing a new Human Familiarization and Training Program. J.
Vet. Behav. Clin. Appl. Res. 5, 112–119. doi:http://dx.doi.org/10.1016/j.jveb.2009.08.008
23. Herron, M.E., Shofer, F.S., Reisner, I.R., 2009. Survey of the use and outcome of
confrontational and non-confrontational training methods in client-owned dogs showing
undesired behaviors. Appl. Anim. Behav. Sci. 117, 47–54.
doi:http://dx.doi.org/10.1016/j.applanim.2008.12.011
24. Hiby, E.F., Rooney, N.J., Bradshaw, J.W.S., 2004. Dog training methods: their use,
effectiveness and interaction with behaviour and welfare. Anim. Welf. 13, 63–69.
25. Landsberg, G., Hunthausen, W., Ackerman, L., 2003. Handbook of behavior problems in the
dog and cat. W.B. Saunders, Philadelphia, PA., pp. 385–426
26. Maier, S. F., Seligman, M. E. P., 1976. Learned helplessness: Theory and evidence. Journal of
Experimental Psychology: General, 105, 3-46.
27. Mazur, J., 2006. Learning and Behavior (6th ed). Upper Saddle River, N.J.: Pearson/Prentice
Hall.
28. Reid, P., 2007. Learning in Dogs, in: Jensen, P. (Ed.), The Behavioural Biology of Dogs. CABI,
pp. 120–144.
29. Rooney, N.J., Cowan, S., 2011. Training methods and owner–dog interactions: Links with dog
behaviour and learning ability. Appl. Anim. Behav. Sci. 132, 169–177.
doi:http://dx.doi.org/10.1016/j.applanim.2011.03.007
30. Salgirli, Y., Schalke, E., Hackbarth, H., 2012. Comparison of learning effects and stress
between 3 different training methods (electronic training collar, pinch collar and quitting
signal) in Belgian Malinois Police Dogs. Rev. Méd. Vét. 163, 530–535.
doi:http://dx.doi.org/10.1016/j.jveb.2009.05.014
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
31. Sankey, C., Richard-Yris, M.-A., Henry, S., Fureix, C., Nassur, F., Hausberger, M., 2010.
Reinforcement as a mediator of the perception of humans by horses (Equus caballus). Anim.
Cogn. 13, 753–764. doi:http://dx.doi.org/10.1007/s10071-010-0326-9
32. Schalke, E., Stichnoth, J., Ott, S., Jones-Baade, R., 2007. Clinical signs caused by the use of
electric training collars on dogs in everyday life situations. Appl. Anim. Behav. Sci. 105, 369–
380. doi:http://dx.doi.org/10.1016/j.applanim.2006.11.002
33. Schilder, M.B.H., van der Borg, J.A.M., 2004. Training dogs with help of the shock collar: short
and long term behavioural effects. Appl. Anim. Behav. Sci. 85, 319–334.
doi:http://dx.doi.org/10.1016/j.applanim.2003.10.004
34. Schilder, M.B.H., Vinke, C.M., van der Borg, J.A.M., 2014. Dominance in domestic dogs
revisited: Useful habit and useful construct? J. Vet. Behav. Clin. Appl. Res. 9, 184–191.
doi:http://dx.doi.org/10.1016/j.jveb.2014.04.005
35. Skinner, B.F., 1938. The behavior of organisms: an experimental analysis. Appleton-Century,
Oxford, England.
36. Steiss, J. E., Schaffer, C., Ahmad H. A., Voith, V. L. 2007. Evaluation of plasma cortisol levels
and behavior in dogs wearing bark control collars. Appl. Anim. Behav. Sci. 106, 96-106. doi:
http://dx.doi.org/10.1016/j.applanim.2006.06.018
37. Ulrich, R.E., Azrin, N.H., 1962. Reflexive fighting in response to aversive stimulation. J. Exp.
Anal. Behav. 511–520. doi:http://dx.doi.org/10.1901/jeab.1962.5-511
38. Ulrich, R.E., Wolff, P.C., Azrin, N.H., 1964. Shock as an elicitor of intra- and inter-species
fighting behaviour. Anim. Behav. 12, 14–15. doi:http://dx.doi.org/10.1016/0003-3472(64)90095-
8
39. Weiss, J.M., 1972. Psychological factors in stress and disease. Sci. Am. 226, 104–113.
40. Welfare in Dog Training, n.d. What are the Implications of Using Training Techniques Which
Induce Fear or Pain in Dogs? http://www.dogwelfarecampaign.org/implications-of-
punishment.php (accessed 17.02.02)
41. Yin, S., 2009. Low Stress Handling, Restraint and Behaviour Modification of Dogs and Cats.
Cattle Dog Publishing, Davis, CA.
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
42. Ziv, G. 2017. The effects of using aversive training methods in dogs- A review. J. Vet. Behav.
Clin. Appl. Res. 19, 50-60 doi:http://dx.doi.org/10.1016/j.jveb.2017.02.004
A01/00
Version: Postprint (identical content as published paper) This is a self-archived document from i3S – Instituto de
Investigação e Inovação em Saúde in the University of Porto Open Repository For Open Access to more of our
publications, please visit http://repositorio-aberto.up.pt/
NEUROBIOLOGY OF LEARNING AND MEMORY 67, 57–63 (1997)
ARTICLE NO. NL963747
†Center for the Neurobiology of Learning and Memory, ‡Department of Psychobiology and §Department of Pharmacology,
University of California, Irvine, Irvine, California 92717; and *Departamento de Neurociencias, Instituto de Fisiologia
Celular Urtiversidad Nacional Autónoma de México, Mexico City, 04510 Mexico
Introini-Collison, Nagahara, Cahill, Brioni, & were aimed at either the IC (AP Å /2; ML { 4.5;
Castellano, 1990; McGaugh, Cahill, Parent, Mesches, DV Å 05.0; from dura, nose bar Å 03.3 mm; n Å
Coleman-Mesches, & Salinas, 1995). Further, there is 15), or AM (AP Å 02.3; ML { 4.5; DV Å 08 from
evidence that the AM and the IC are functionally and dura, nose bar Å 03.3 mm; n Å 26; Paxinos & Wat-
reciprocally interconnected (Escobar, Fernández, son, 1986). The injection needle tips (which pro-
Guevara-Aguilar, & Bermúdez-Rattoni, 1989; Lasi- truded 1 mm from the tip of the cannulae) reached
ter & Glanzman, 1985). A direct projection between the injection targets. Behavioral procedures began
the AM and the IC via the internal capsule has been 1 week after surgery and injections of NMDA were
described (Norgren, & Wolf 1975; Kiefer, 1985). It has given to induce lesions 1 week after the behavioral
been suggested that cortical projections to the AM, training was completed.
including those from the IC, convey processed cogni-
tive information which is then integrated with emo- Apparatus and Behavioral Procedure
tional and motivational processes (Pascoe & Kapp,
The animals were trained in a trough-shaped
1987). Microinjections of tetrodotoxin, a reversible ac-
alley (91 cm long, 20 cm wide at the top, 6.4 cm
tivity-dependent sodium-channel blocker, into the IC
wide at the floor, 15 cm deep) that was divided
induce retrograde and anterograde memory impair-
into two compartments by a guillotine door that
ment in inhibitory avoidance and spatial learning
retracted into the floor. A safe compartment (31
tasks (Bermúdez-Rattoni et al., 1991). Moreover, re-
cm long) illuminated by a tensor lamp was sepa-
cent findings indicate that N-methyl-D-aspartate
rated from the darkened compartment (61 cm long)
(NMDA)-induced AM lesions impair inhibitory avoid-
where animals received the shock. In Experiment
ance retention in animals given extensive training
1 all animals received 10 escape training trials. In
prior to induction of the lesions (Parent, Tomaz, &
Experiment 2, different groups of rats received 1,
McGaugh, 1992; Parent, Avila, & McGaugh, 1995;
10, or 20 training trials. In each trial the rat was
Parent, Quirarte, Cahill, & McGaugh, 1995). However,
placed in the dark compartment facing away from
the effects of posttraining IC lesions in rats given dif-
the door leading to the illuminated compartment
ferent amounts of training prior to induction of the
and when the rat turned toward the door, the door
lesions are not yet known. To address this issue the
was opened and a timer was started. A footshock
first experiment assessed the effects on retention of
was administered (0.75 mA) until the rat escaped
NMDA-induced lesions of either the AM or the IC in
into the illuminated compartment. The rat was re-
rats given 10 escape-training trials 1 week prior to
tained in the illuminated compartment with the
induction of the lesions. The second experiment exam-
door closed for 60 s. Rats given only 1 training trial
ined the effects on retention of IC NMDA lesions in-
were then returned to their home cages. Rats given
duced 1 week after 1, 10, or 20 escape training trials.
10 or 20 trials were removed from the lighted com-
partment and placed back into the dark compart-
METHODS ment where the same procedure was followed for
the remaining trials.
Subjects
Male Sprague–Dawley rats (260–320 g) from Lesions
Charles River Laboratories (Wilmington, MA) were
One week after the escape training, the cannu-
used. They were individually housed and main-
lated animals were randomly assigned to one of two
tained in a 12-h light/dark cycle (light on at 7:00
groups, and bilateral NMDA-induced lesions (AM-
A.M.) with food and water freely available. The ani-
Lx; n Å 16; IC-Lx; n Å 7) were induced by injecting
mals were acclimated to the laboratory conditions
the excitotoxin (Sigma Chemical Co., St. Louis, MO).
for approximately 1 week before surgical procedures
The NMDA (10 mg/1 ml) was dissolved in 10 mM
were initiated.
phosphate-buffered saline, pH 7.0–7.4, back-filled
into a 30-gauge needle that was attached via poly-
Surgical Procedure
ethylene tubing (PE20) to a 10-ml syringe (Hamilton
The rats were implanted bilaterally with 15-mm Co., Reno, NV), and injected at a rate of 2.0 ml/min
23-gauge stainless steel cannulae under anesthesia for 4 min with the use of an automated Sage syringe
(a mixture of 100 mg/ml ketamine hydrochloride and pump (Orion Res., Inc., Boston, MA). The needle was
20 mg/ml xylazine, 87:13, 1.2 ml/kg i.p. and 0.4 mg/ kept in place for 5 min after the injection to max-
kg atropine i.p. as a preanesthetic) using standard imize diffusion of the NMDA into the IC or the AM.
stereotaxic procedures. The tips of the cannulae Sham control animals (AM-Sh, n Å 10; IC-Sh, n Å
IMPAIRED RETENTION BY AMYGDALA AND INSULAR CORTEX LESIONS 59
8) were given bilateral microinjections of buffer solu-
tion following the same protocols as used for the
NMDA microinjections.
Statistics
One week after the microinjections all the animals
were tested for retention. On the retention test, each
rat was placed in the illuminated compartment and
the latency to enter the dark compartment (maxi-
mum of 300 s) was recorded. Step-through retention
latencies were analyzed using the nonparametric
Kruskall–Wallis and two-tailed Mann–Whitney U
test and are expressed as medians (interquartile
range). The number of crossings between compart-
ments during the 300-s testing period was analyzed
by one-way ANOVAs and post-hoc pairwise compari-
sons using the Fisher test and was expressed as the
mean ({SEM).
Histology
After completion of the experiments, the animals FIG. 1. Schematic representation of the smallest (dark-
were perfused with 4% paraformaldehyde, 1% glu- shaded) and largest (light-shaded) lesions of the insular cortex
taraldehyde in phosphate buffer (pH 7.4); the brains (left) and the amygdaloid complex (right). The numbers beside
were excised and stained with cresyl violet and ex- each section indicate the distance anterior or posterior to bregma.
amined under a microscope to determine cannula Atlas plates are adapted from ‘‘The Rat Sterotaxic Coordinates,’’
by Paxinos & Watson, 1986.
placements.
RESULTS
latencies of the AM and IC lesioned groups were
The histological findings revealed that injections significantly shorter than those of the sham-oper-
of NMDA produced a loss of neurons and extensive ated groups (p õ .01 for both comparisons), sug-
gliosis in both the IC and the AM. The cannulae tips gesting that NMDA-induced lesions of the IC and
for the IC groups were located in the border of the AM induced comparable disruptive effects on inhibi-
granular and agranular IC and in some cases located tory avoidance retention. Figure 2B shows the num-
in the dorsal claustrum (Fig. 1). The data from two ber of crossings during the test trial. A one-way AN-
animals with unclear or unilateral lesions were dis- OVA on the number of crossings between the two
carded. The AM lesions were generally located in compartments revealed a significant difference be-
the central nucleus and extensive lesions reached tween groups (F(3,40) Å 5.95; p õ .01). The two
the basolateral and lateral nucleus (Fig. 1). The sham-lesioned groups did not differ in the number
data from four animals with unilateral or unclear of crossings. The number of crossings made by the
lesions were discarded. The histological findings are animals in the ‘‘IC-lesioned’’ and ‘‘AM-lesioned’’
similar to those of previous studies using NMDA or groups was significantly greater than that of ani-
ibotenic acid-induced lesions (Bermúdez-Rattoni & mals in the sham-IC lesioned (p õ .01) and sham
McGaugh, 1991; Jellestad & Garcia-Cabrera, 1986). AM-lesioned groups (p õ .05).
These findings indicate that both AM and IC le-
Experiment 1
sions had significant disruptive effects on the long-
Figure 2A shows the inhibitory avoidance reten- term retention of escape training. Recent findings
tion latencies 1 week after the animals received indicate that the disruptive effects of posttraining
NMDA-induced lesions in either the AM or the IC. AM lesions on retention are attenuated by additional
Statistical analysis of the retention latencies re- training (i.e., 20 training trials) prior to induction of
vealed significant differences among groups (H(3,41) the lesions (Parent et al., 1992). Thus, Experiment
Å 21.31; p õ .01). The step-through latencies for 2 was conducted to determine whether the retention-
both sham groups were very similar. The retention impairing effects of postraining IC lesions depend
60 BERMÚDEZ-RATTONI ET AL.
Experiment 2
The surgical and behavioral procedures and appa-
ratus were the same as those used in Experiment 1
except that the animals received only NMDA or
sham IC lesions. The animals received 1, 10, or 20
escape training trials as described under Methods.
One week after the escape training the animals were
assigned to one of two sets of groups: IC-lesioned
(Lx-1, n Å 10; Lx-10, n Å 9; Lx-20, n Å 10) or sham- FIG. 3. The effects of IC NMDA-induced lesions and degree
lesioned groups (Sh-1, n Å 10; Sh-10, n Å 11; Sh-20, of training in an escape task. (A) The mean latency (SEM) to
n Å 9). As described under Methods, the lesions were enter the dark compartment B. The number of crossings (SEM)
between compartments on the retention test. **p õ .01 vs sham
induced by infusing NMDA into the IC through im- control with the same number of training trials; ## p õ .01 vs
planted cannulae. Retention was tested 1 week later sham or NMDA 1-trial groups; #p õ .05 vs sham or NMDA 1-
and although the retention testing period was 600 trial groups; `p õ .01 vs sham 10-trial groups.
IMPAIRED RETENTION BY AMYGDALA AND INSULAR CORTEX LESIONS 61
between the two compartments (Fig. 3B) revealed sions of the IC or AM disrupted retention even after
significant differences among groups (F(5,52) Å 5.01, 10 or 20 trials of training. The NMDA-induced le-
p õ .01). The IC-lesioned and sham-lesioned animals sions of the IC significantly affected both the reten-
given only one training trial did not differ in the tion latency and the reduction in the number of
number of crossings between the two compartments. crossings when the animals received 10 escape train-
The sham groups given 10 or 20 training trials were ing trials. The lesioned group given 20 training trials
significantly different from those given only one did not have a significantly increased number of
training trial (p õ .01). For groups given 10 training crossings between the two compartments compared
trials, the number of crossings made by the IC-le- to sham controls. The lack of differences could be
sioned group was significantly greater than that due to the increased number of crossings by the
made by sham controls (p õ .01). The IC-lesioned sham group given 20 training trials, although the
and sham groups given 20 training trials did not number of crossings was not significantly different
differ significantly in the number of crossings. There from those of the control group given 10 training
were, however, significant differences between the trials. These findings are similar to those found with
lesioned group given 20 training trials and the sham AM lesions; that is, the retention performance was
group given 10 training trials (p õ .01). affected by the number of trials given prior to induc-
The findings of the present experiment indicate tion of amygdala lesions (Parent, West, & McGaugh
that one single training trial for inhibitory avoidance 1995). In this study, animals given 20 training trials
was not sufficient to produce significant retention in an avoidance task prior to the induction of excito-
when the animals were tested 2 weeks after train- toxic lesions of the IC were not as impaired as ani-
ing. These results were confirmed by two different mals given only 10 training trials prior to the induc-
measures, the retention latency and the number of tion of the lesions (see Fig. 3). These results indicate
crossings between the two compartments during the that the amount of training influenced the retention
retention test. The training with 10 or 20 training performances of both control and lesioned groups.
trials increased retention latencies and reduced the Parent and co-workers, in a series of experiments,
number of crossings when the sham animals were addressed the hypothesis that if the amygdala was
tested 2 weeks after training. The NMDA-induced the locus of permanent memory for aversively moti-
lesions groups trained with 10 or 20 training trials vated learning, then posttraining amygdala lesions
showed significant disruptive effects on both reten- should have disruptive effects regardless of the de-
tion latencies and number of crossings, although the gree of preoperative training (Parent & McGaugh,
effects of the NMDA-induced lesions were slightly 1994). The results indicate that the effects of amyg-
attenuated in the group given 20 training trials, as dala posttraining lesions on retention depend on the
demonstrated by the reduced number of crossings. number of training trials given before the lesion is
These results indicate that NMDA-induced lesions induced (Parent et al., 1995; Parent, West, &
of the IC significantly impair retention in an inhibi- McGaugh, 1995). Moreover, these authors report
tory avoidance task learned 2 weeks before the re- that the level of the unconditioned stimulus applied,
tention test. i.e., the shock intensity, also affects retention. That
is, when either the number of training trials or the
DISCUSSION shock intensity was increased, lesions of the amyg-
dala had decreased effects on retention. When the
The results of the present experiments support the same animals were subsequently retrained in a mul-
view that the IC and AM are involved in retention of tiple-trial inhibitory avoidance task after the lesions
an inhibitory avoidance task. In the first experiment, were induced, animals given 10 or 20 training trials
both AM and IC lesions have similar effects on reten- learned the task significantly faster than those given
tion; that is, even after 10 training trials, NMDA- fewer initial training trials. Such findings suggest
induced lesions produced severe impairment in the that the amygdala is involved in acquisition but is
recall of the task. The results of Experiment 2 sug- not the permanent site of storage for aversive learn-
gest that neither sham nor lesioned animals that ing. Increasing either the number of training trials
were trained with only 1 trial learned the task. The or the shock intensity leads to a better retention
shock intensity was low and therefore 1 training performance. In the present experiment, one single
trial was not sufficient to produce good acquisition training trial was not sufficient to produce good re-
of the task. Sham animals given 10 or 20 training tention, but when 10 or 20 training trials were given,
trials demonstrated good retention performance 2 the animals demonstrated good retention in two
weeks after training. In contrast, NMDA-induced le- measures: step-through latency and the number of
62 BERMÚDEZ-RATTONI ET AL.
crossings. There is increased variance in the step- IC after overtraining animals in an inhibitory avoid-
through latencies as well as in the number of cross- ance task.
ings in animals given 20 training trials. These data
might indicate that some nonspecific effects begin to REFERENCES
produce interference with learning after 10 training
trials using electric shocks. The posttraining IC le- Bermúdez-Rattoni, F., Fernández, J., Sánchez, M., Aguilar-Ro-
blero, R., & Drucker-Colin, R. (1987). Fetal brain transplants
sions had disruptive effects even in animals with 10
induce recuperation of taste aversion learning. Brain Re-
or 20 training trials but there was a slight tendency search, 416, 147–150.
to attenuate the disruptive effects with an increased Bermúdez-Rattoni, F., Introini-Collison, I. B., & McGaugh, J. L.
number of training trials. The findings with AM le- (1991). Reversible inactivation of the insular cortex by tetro-
sions by Parent et al. (1995) and the data presented dotoxin produces retrograde and anterograde amnesia for in-
here suggest that extensive training induces hibitory avoidance and spatial learning. Proceedings of the
changes in other brain areas that participate in the National Academy of Sciences USA, 88, 5379–5382.
long-lasting storage of aversively motivated memory Bermúdez-Rattoni, F., & McGaugh, J. L. (1991). Insular cortex
and amygdala lesions differentially affect acquisition of in-
(Prado-Alcalá, 1995). hibitory avoidance and conditioned taste aversion. Brain Re-
We have examined whether the IC is a site of per- search, 549, 165–170.
manent storage for conditioned taste aversion or is Bermúdez-Rattoni, F., Sánchez, M. A., & Prado-Alcalá, R. A.
only a site that modulates the storage of information (1989). Learning of external and visceral cue consequences
elsewhere in the brain, using fetal brain grafts as a may be subserved by different neouroanatomical substrates.
‘‘reversible long-lasting lesion’’ model. It has been In T. Archer & N. Lars-Goran (Eds.), Aversion avoidance and
anxiety (pp. 121–138). Mahwah, NJ: Lawrence Erlbaum As-
demonstrated that homotopical, but not occipital, sociates.
cortical grafts induce recovery of learning abilities, Bermúdez-Rattoni, F., Ormsby, Ch. E., Escobar, M. L., & Hernán-
including inhibitory avoidance, in IC-lesioned rats dez-Echeagaray. (1995). The role of the insular cortex in the
(Bermúdez-Rattoni, Fernández, Sánchez, Aguilar- acquisition and long lasting for aversively motivated behav-
Roblero, & Drucker Colı́n, 1987; Fernández-Ruı́z, ior. In J. L. McGaugh, F. Bermúdez-Rattoni, and R. A. Prado
Escobar, Pina, Dı́az Cintra, Cintra-McGlone, & Ber- Alcalá (Eds.), Plasticity in the central nervous system: Learn-
ing and Memory (pp. 67–82) Mahwah, NJ: Lawrence Erl-
múdez-Rattoni, 1991; Escobar et al., 1989; Escobar, baum Associates.
Russell, Booth, & Bermúdez-Rattoni, 1994). In a re-
Dunn, L. T., & Everitt, B. J. (1988). Double dissociations of the
cent series of experiments, we have demonstrated effects of amygdala and insular cortex lesions on conditioned
that both homotopical and occipital cortical grafts taste aversion, passive avoidance, and neophobia in the rat
restore a conditioned taste aversion even 70 days using the excitotoxic ibotenic acid. Behavioral Neuroscience,
after training. However, only those animals that re- 102, 3–22.
ceived homotopical, but not occipital, grafts recov- Escobar, M. L., Fernández, J., Guevara-Aguilar, R., & Bermúdez-
Rattoni, F. (1989). Fetal brain grafts induce recovery of learn-
ered the ability to acquire a taste aversion, using a
ing deficits and connectivity in rats with gustatory neocortex
different taste stimulus when retrained (Bermúdez- lesion. Brain Research, 478, 368–374.
Rattoni, Ormsby, Escobar, & Hernández-Echea- Escobar, M. L., Russell, R. W., Booth, R. A., & Bermúdez-Rattoni,
garay, 1995; Ormsby, Pina, & Bermúdez-Rattoni, F. (1994). Accelerating behavioral recovery after cortical le-
1991). These results are consistent with the notion sions. I. Homotopic implants plus NGF. Behavioral and Neu-
that the IC is involved in the learning and retention ral Biology, 61, 73–80.
of different tasks but is probably not a site of long- Fernández-Ruiz, J., Escobar, M. L., Piña, A. L., DıB az-Cintra, S.,
term storage. Cintra-McGlone, F. L., & Bermúdez-Rattoni, F. (1991). Time-
dependent recovery of taste aversion learning by fetal brain
Together, the present results and those of Parent transplants in gustatory neocortex-lesioned rats. Behavioral
and co-workers indicate that the IC and AM are im- and Neural Biology, 55, 179–193.
portant structures for the acquisition and consol- Jellestad, F. K., & GarcıB a-Cabrera, J. G. (1986). Exploration and
idation of inhibitory avoidance and for retrieval. Al- avoidance learning after ibotenic acid and radio frequency
though the participation of the amygdala is less im- lesions in the rat amygdala. Behavioral and Neural Biology,
portant for retrieval, other experiments are neces- 46, 196–215.
sary to determine fully the role of the insular cortex Kiefer, S. W. (1985). Neural mediation of conditioned food aver-
sions. Annals of the New York Academy of Sciences, 443, 100–
in aversively motivated learning and memory. Fi-
109.
nally, it is possible that functional loops are estab-
Krushel, L. A., & van der Kooy, D. V. D. (1988). Visceral cortex:
lished between the IC and AM and other brain struc- Integration of the mucosal with limbic information in the rat
tures responsible for long-term storage of aversive agranular insular cortex. Journal of Comparative Neurology,
information. We are currently evaluating this possi- 270, 34–54.
bility by producing lesions of both the AM and the Kornhuber, A. W. (1995). Unimanual motor learning impaired by
IMPAIRED RETENTION BY AMYGDALA AND INSULAR CORTEX LESIONS 63
frontomedial and insular lesions in man. Journal of Neurol- tention of inhibitory avoidance training. Brain Research, 661,
ogy, 242, 568–578. 97–103.
Lasiter, P. S., & Glanzman, D. I. (1985). Cortical substrates of Parent, M. B., Quirarte, G. L., Cahill, L., & McGaugh, J. L. (1995).
taste aversion learning: Involvement of the dorsalateral Spared retention of inhibitory avoidance learning following
amygdaloid nuclei and temporal neocortex in taste aversion posttraining amygdala lesions. Behavioral Neuroscience, 109,
learning. Behavioral Neuroscience, 99, 257–276. 803–807.
McGaugh, J. L., Cahill, L., Parent, M. B., Mesches, M. H., Cole- Parent, M. B., Tomaz, C., & McGaugh, J. L. (1992). Increased
man-Mesches, K., & Salinas, J. A. (1995). Involvement of the training in an aversively motivated task attenuates the mem-
amygdala in the regulation of memory storage. In J. L. ory-impairing effects of posttraining N-Methyl-D-aspartate-
McGaugh, F., Bermúdez-Rattoni, and R. A. Prado Alcalá induced amygdala lesions. Behavioral Neuroscience, 106,
(Eds.), Plasticity in the central nervous system: Learning and 791–799.
memory, Mahwah NJ: Lawrence Erlbaum Associates. Parent, M. B., West, M., & McGaugh, J. L. (1995). Memory of
McGaugh, J. L., Introini-Collison, I. B., Nagahara, A. H., Cahill, rats with amygdala lesions induced 30 days after footshock-
L., Brioni, J. D., & Castellano, C. (1990). Involvement of the motivated escape training reflects degree of original training.
amygdaloid complex in neuromodulatory influences on mem- Behavioral Neuroscience, 108, 1080–1087.
ory storage. Neuroscience and Behavioral Review, 14, 425– Paxinos, G., & Watson, C. (1986). The rat brain in stereotaxic
431. coordinates, San Diego: Academic Press.
Norgren, R., & Wolf, G. (1975). Projections of thalamic gustatory Prado-Alcalá, R. A. (1995). Serial and parallel processing during
and lingual areas in the rat. Brain Research, 92, 123–129. memory consolidation. In J. L. McGaugh, F. Bermúdez-Rat-
toni, & R. A., Prado Alcalá (Eds.), Plasticity in the Central
Ormsby, C. E., Piña, A. L., & Bermúdez-Rattoni, F. (1991). Long-
Nervous System: Learning and Memory. Mahwah, NJ: Law-
term retrograde amnesia of inhibitory avoidance and condi- rence Erlbaum Associates.
tioned taste aversion learning tasks by insular cortex lesions.
Saper, C. B. (1988). Convergence of autonomic and limbic projec-
Society for Neuroscience Abstracts, 17, 1045.
tions in the insular cortex of the rat. Journal of Comparative
Pascoe, I. P., & Kapp, B. S. (1987). Responses of amygdaloid cen- Neurology, 2, 163–173.
tral nucleus neurons to stimulation of the insular cortex in
Scott, L. (1995). A positron emission tomographic study of simple
awake rabbits. Neuroscience, 21, 471–485.
phobic symptom provocation. Archives of General Psychiatry,
Parent, M. B., Avila, E., & McGaugh, J. L. (1995). Footshock facil- 52, 20–28.
itates the expression of aversively motivated memory in rats Van-der-Kooy, D. L., Koda, L. Y., McGinty, C. R., Gerfen,
given posttraining amygdala basolateral complex lesions. C. R., & Bloom, F. E. (1984). The organization of projections
Brain Research, 676, 235–244. from the cortex, amygdala and hypothalamus to the nu-
Parent, M. B., & McGaugh, J. L. (1994). Posttraining lesions of cleus of the solitary tract in rat. Journal of Comparative
lidocaine into the amygdala basolateral complex impair re- Neurology, 224, 1 – 24.
PLOS ONE
RESEARCH ARTICLE
Funding: The current research study was especially if used in high proportions, compromise the welfare of companion dogs both
supported by FCT - Fundação Portuguesa para a within and outside the training context.
Ciência e Tecnologia (Fellowship SFRH/BPD/
111509/2015) and UFAW – Universities Federation
for Animal Welfare (Grant 14-16/17), with grants
awarded to ACVC. SP was supported by PIPOL -
Regione Friuli Venezia Giulia. The funders had no
role in study design, data collection and analysis, 1. Introduction
decision to publish or preparation of the
To fulfil their increasingly important role as companion animals, dogs need to be trained to
manuscript. FCT - Fundação Portuguesa para a
Ciência e Tecnologia: https://www.fct.pt/index.
behave in a manner appropriate for human households. This includes, for example, learning
phtml.pt UFAW – Universities Federation for to eliminate outdoors or walk calmly on a leash [1,2]. Dog behavioral problems are the most
Animal Welfare: https://www.ufaw.org.uk/. frequently cited reason for rehoming or relinquishment of dogs to shelters and for euthanasia
Competing interests: The authors have declared
[2], which suggests that such training is often missing or unsuccessful.
that no competing interests exist. Dog training most often involves the use of operant conditioning principles, and dog train-
ing methods can be classified according to the principles they implement: aversive-based
methods use mainly positive punishment and negative reinforcement and reward-based meth-
ods rely on positive reinforcement and negative punishment [3]. Within a given training
method, several factors may influence how dogs react, such as the characteristics of the behav-
ior under training and the timing of reinforcement/punishment [4]. However, the use of aver-
sive-based training methods per se is surrounded by a heated debate, as studies have linked
them to compromised dog welfare [5–10]. Some aversive-based tools, such as shock collars,
have indeed been legally banned in some countries [11]. However, a recent literature review by
[3] concluded that, because of important limitations, existing studies on the topic do not pro-
vide adequate data for drawing firm conclusions. Specifically, the authors reported that a con-
siderable proportion of the studies relied upon surveys rather than on objective measures of
both training methods and welfare; that they focused on sub-populations of police and labora-
tory dogs which only represent a small portion of dogs undergoing training; and, finally, that
the empirical studies have concentrated mainly on the effects of shock-collar training, which is
only one of several tools used in aversive-based training. In summary, limited scientific evi-
dence exists on the effects of the entire range of dog training techniques on companion dog
welfare.
Furthermore, previous empirical studies have focused on the effects of training methods on
dog welfare within the training context. Behavioral and physiological indicators of welfare,
such as the frequency of stress-related behaviors and the concentration of salivary cortisol,
have been collected in and around the training situation (e.g., [9,12]; see also [3]). However,
the welfare impact of training methods beyond the training scenario has not yet been exam-
ined. To our knowledge, only one study evaluated the effects of training on welfare outside the
training context. Christiansen et al (2001) [13] found no effect of shock collar training on dog
fear or anxiety, but this was based on dog owner reports of behavior and temperament tests
rather than on objective and animal-based welfare indicators. Importantly, a suitable assess-
ment of the effects of training methods on dog welfare should comprise an evaluation of their
effects both during and beyond the training scenario.
The affective states of animals are influenced by both immediate rewarding or punishing
experiences (effects on shorter-term states), and by the cumulative experience of rewarding or
punishing experiences (effects on longer-term states) [14]. Hence, due to the repeated expo-
sure to aversive stimuli, training with aversive-based methods is expected to also affect dogs’
affective states in a longer-term, transitioning to outside the training context. One way to
assess affective states is through the cognitive bias paradigm (e.g., [15]). The cognitive bias task
has been validated as an effective tool to evaluate the affective states of non-human animals
and has been extensively used with several species, including dogs [16–18]. The rationale
behind the paradigm is based on theoretical and empirical findings that an individual’s under-
lying affective state biases its decision-making and, specifically, that individuals experiencing
negative affective states make more ‘pessimistic’ judgements about ambiguous stimuli than
individuals experiencing more positive affective states [14,15,17].
Therefore, the aim of the present study was to perform a comprehensive evaluation of the
effects of different training methods on the welfare of companion dogs both within and out-
side the training context. By performing an objective assessment of training methods (through
the direct observation of training sessions) and by using objective measures of welfare (behav-
ioral and physiological data to assess effects during training, and a cognitive bias task to assess
effects outside training), we assessed the effects of reward-based and aversive-based methods
on companion dog welfare. We hypothesized that dogs trained using aversive-based methods
would display higher levels of stress during training, as determined by behavioral and physio-
logical indicators of stress during training sessions, and more ‘pessimistic’ judgments of
ambiguous stimuli during a cognitive bias task performed outside the training context, as com-
pared to dogs trained using reward-based methods. We used a quasi-experimental approach
in which dog-owner dyads were recruited to participate through the training school at which
they were enrolled. As treatment could not be randomized, data on potential confounders was
collected to be included in the analysis of treatment effects.
Understanding the effects of training methods on companion dog welfare has important
consequences for both dogs and humans. Both determining and applying those training meth-
ods that are less stressful for dogs is a key factor to ensure adequate dog welfare and to capital-
ize on the human benefits derived from interactions with dogs [19,20].
2.3. Subjects
The head trainer of each training school was asked to indicate at least fourteen dogs fitting our
inclusion criteria (described below), and we then approached the owners to ask if they were
willing to participate. The information about the study given to the owners was the same that
was given to the head trainers of the schools. The inclusion criteria for the dogs were: 1) to
have attended the training school for less than two months, in order to mitigate familiarization
to training methods, and 2) to be free of certain behavioral problems (e.g., aggression, fearful-
ness and separation anxiety, as determined by the owner and the first author), in order to pre-
vent any confounding stress.
Table 1. Definition of the intended operant conditioning procedures used to classify the dog training schools
according to their training methods.
Procedure Definition
Positive punishment Any (presumably) unpleasant stimulus that was applied to the dog after the exhibition of an
undesirable behavior. These included leash jerks (with either choke or pinch collars), shock
delivery through e-collars, slapping the dog, yelling at the dog and leaning towards the dog
in a threatening way.
Negative Any (presumably) unpleasant stimulus that was applied to the dog and that was stopped
reinforcement only after the dog exhibited the desired behavior. These included pulling the collar upward
and releasing the pressure only when the dog sat, pulling the collar downward and releasing
the pressure only when the dog laid down, and hanging the dog by the choke collar until he
or she calmed down.
Positive Any (presumably) pleasant stimulus that was applied to the dog after the exhibition of a
reinforcement desirable behavior. These included food treats, playing tug-of-war, verbal praise, and
petting the dog.
Negative Any (presumably) pleasant stimulus that was removed after the exhibition of an undesirable
punishment behavior. These involved turning the back to the dog as soon as they jumped or started to
mouth, and stopping to walk if the dog was pulling on the leash.
https://doi.org/10.1371/journal.pone.0225023.t001
Over the course of the study, which was conducted between October 2016 and March 2019,
the owners of 122 companion dogs agreed to participate. However, 30 dog owners dropped
out of the training schools before any meaningful data could be collected. Specifically, these
subjects dropped out before meeting our requirement that at least two training sessions were
video recorded and that the owner completed a written questionnaire. Consequently, our final
sample comprised 92 subjects, 28 from Group Aversive (Schools A and D: n = 14), 22 from
Group Mixed (School C: n = 8, School F: n = 14), and 42 from Group Reward (School B and
G: 15 dogs, School E: 12 dogs).
As for subjects’ demographics, the average age was 11.9 (SEM = ±1.0) months, 54% were
male and 35% were neutered/spayed. Thirty-four percent were mixed-breed dogs and the
remaining 66% belonged to a FCI-recognized breed group: 18% belonged to Group 1: Sheep-
dogs and Cattledogs (except Swiss Cattledogs), 13% to Group 2: Pinscher and Schnauzer–
Molossoid and Swiss Mountain and Cattledogs, 5% to Group 3: Terriers, 4% to Group 6: Scent
hounds and related breeds, 2% to Group 7: Pointing dogs, 20% to Group 8: Retrievers–Flush-
ing Dogs–Water Dogs, and 3% to Group 9: Companion and Toy Dogs.
2.4.2. Phase 2—Evaluating welfare outside the training context. After finishing data
collection for Phase 1, dogs participated in Phase 2, which consisted of a spatial cognitive bias
task. The end of Phase 1 did not correspond to the conclusion of the training programs for the
dogs, as this would result in different dogs being exposed to substantially different amounts of
training before being assessed for cognitive bias. Instead, for standardization purposes, we
ensured that 1) dogs had attended the training school for at least one month prior to Phase 2
and that 2) the cognitive bias task was conducted within one month of completing Phase 1.
Due to limited owner availability, 13 subjects either dropped out or did not meet the criteria
for Phase 2, resulting in 79 (24 from Group Aversive, 20 from Group Mixed and 35 from
Group Reward) of the original 92 dogs participating in Phase 2. The cognitive bias tasks were
scheduled according to owners’ availability, both on weekdays and Saturdays.
The test was conducted in an indoor room (7.7 x 3 meters) within a research building at the
Abel Salazar Biomedical Sciences Institute (ICBAS), University of Porto in Portugal. All dogs
were unfamiliar with the room prior to testing. Two experimenters conducted the test while
the dog’s owner(s) sat in a chair in a corner area of the room (see Fig 1). Dog owners were
asked not to look into the dog’s eyes or to speak to the dog during the test, unless the experi-
menters instructed otherwise. The entire test took place over one meeting for each dog. The
room was cleaned with water and liquid detergent at the end of each test.
2.4.2.1. Familiarization period. Prior to the start of the cognitive bias task, the dogs were
given the opportunity to familiarize with the test room and the researchers. This consisted of a
10-min period during which the dog was allowed to freely explore the room and engage with
the researchers and the owner(s).
2.4.2.2. Training phase. The methodology of Phase 2 was based on [18]. During the training
phase, dogs were trained to discriminate between a ‘positive’ (P) location of a food bowl,
which always contained a food reward, and a ‘negative’ (N) location, which never contained a
food reward. At the start of each trial, the dog was held by one trained experimenter—hereafter
the ‘handler’, behind a barrier (2 x 2 m, see Fig 1), while a second trained experimenter—here-
after the ‘timer’, baited (or did not bait, depending on the type of trial) the bowl with a piece of
sausage (approximately 1.25 g for smaller dogs and 2.5 g for larger dogs). To ensure that the
dog, the owner and the ‘handler’ were blind to whether or not the bowl contained food during
each trial, the bowl was baited out of their sight, on the opposite side of the barrier. Addition-
ally, the food reward was rubbed onto the food bowl before every trial to prevent the influence
of olfactory cues. The height of the food bowl was such that the dog could not visually judge
the presence or absence of food from the start position.
After baiting (or not baiting) the bowl, the ‘timer’ placed it at one of the two training loca-
tions. The ‘timer’ then determined the start of the trial, by verbally signaling to the ‘handler’,
upon which the ‘handler’ led the dog to the start position and released him. The ‘handler’
always led the dog to the start position on her left side. Because we found that dogs had some
difficulty noticing the bowl at the end of the room during pilot tests, the ‘handler’ walked
towards the bowl and pointed it out to the dog during the first four trials. For the remaining
trials, the ‘handler’ simply walked the dog to the start position and released him. After the dog
reached the food bowl and (when applicable) ate the reward, the ‘handler’ collected him and
led him behind the barrier to start the next trial. The latency to reach the bowl, defined as the
time elapsed between release at the start position and the dog putting its head in line with the
edge of the bowl, was recorded for each trial by the ‘timer’ using a stopwatch.
The position of the ‘positive’ and ‘negative’ locations was counterbalanced across subjects
and training schools, such that for half of the dogs from each training school, the ‘positive’
location was on the right hand side as they faced the test area, and for the other half it was on
the left. Initially, each dog received two consecutive ‘positive’ trials (bowl placed in the ‘posi-
tive’ location) followed by two ‘negative’ trials (bowl placed in the ‘negative’ location). Subse-
quently, ‘positive’ and ‘negative’ trials were presented in a pseudorandom order, with no more
than two trials of the same type being presented consecutively.
All dogs received a minimum of 15 training trials to learn the discrimination between bowl
locations. Dogs were considered to have learned an association between bowl location and
food (the learning criterion) when, after a minimum of 15 trials, the longest latency to reach
the ‘positive’ location was shorter than any of the latencies to reach the ‘negative’ location for
the preceding three ‘positive’ trials and the preceding three ‘negative’ trials. Each trial lasted a
maximum of 20 seconds. If the dog did not reach the bowl by that time, the trial automatically
ended and a latency of 20 seconds was recorded.
All but two dogs were able to complete the training phase. For the two dogs that failed to
complete training, one did not show any interest in the food reward and the other was food-
motivated but could not focus on the task. These two dogs belonged to Group Mixed. There-
fore, the total number of subjects completing Phase 2 in Group Mixed was 18.
2.4.2.3. Test phase. Testing began once the learning criterion was achieved. Test trials were
identical to training trials except that the bowl (empty) was placed at one of three ambiguous
locations equally spaced along an arc 4 m from the dog’s start position, between the ‘positive’
and ‘negative’ locations. The three test locations were: ‘near-positive’ (NP: one third of the way
along the arc from the ‘positive’ location), ‘middle’ (M: half way along the arc), ‘near-negative’
(NN: one third of the way along the arc from the ‘negative’ location). Three test trials were pre-
sented at each test location (nine test trials in total) in the following order for all dogs: M, NP,
NN, NP, NN, M, NN, M, NP (each location was presented first, second or third in each block
of three test trials). Each test trial was separated from the next one by two training trials identi-
cal to those conducted in the training phase (one ‘positive’ and one ‘negative’ trials presented
in a random order), in order to maintain the associations between the ‘positive’ and ‘negative’
locations and the presence or absence of food, respectively. Thus, the test phase included a fur-
ther sixteen training trials interspersed in blocks of two between the nine test trials.
To end the test phase, a final trial was conducted by placing an empty bowl in the ‘positive’
location to determine whether dogs ran to the empty bowl as quickly as they did to the baited
bowl. This was meant to establish that the dogs were not relying on olfactory or visual cues
during the test. During the entire test, each trial was kept as similar as possible in terms of
preparation time and activity, and dogs were handled in the same way throughout the test.
Due to circumstances beyond our control, namely people speaking loudly and other dogs
barking in the building during some of the tests, some subjects were clearly distracted and dis-
engaged from the task during some trials. Whenever this happened, no latency was recorded
for that trial. The experimenters waited for the dog to resettle and moved to the following trial.
2.5. Questionnaire
All owners were asked to complete a brief written questionnaire regarding dog demographics
and background, and owner demographics and experience with dogs and dog training. The
questionnaire was based on [10].
2.6.1.2. Cortisol analysis. Two dogs (one from School B and one from School E, both from
Group Reward) did not cooperate with the saliva collection procedure and, as such, no saliva
samples were extracted from them. For the remaining 90 dogs, only 23 dog owners (seven
from Group Aversive, five from Group Mixed and 11 from Group Reward) were able to appro-
priately collect six saliva samples. The samples from these subjects were selected for analysis.
An additional 40 dog owners (14 from Group Aversive, 11 from Group Mixed and 15 from
Group Reward) were able to properly collect at least four saliva samples. From these 40 dogs,
eight were randomly selected to have their samples analyzed (one from Group Aversive, three
from Group Mixed and four from Group Reward). In total, 8 dogs from Group Aversive, 8
dogs from Group Mixed and 15 dogs from Group Reward had their samples selected for analy-
sis (Schools A, C, D, E and F: n = 4; School B: n = 5; School G: n = 6). These samples were sent
to the Faculty of Sport Sciences and Physical Education of the University of Coimbra, Coim-
bra, Portugal, where they were assayed for cortisol concentration using standard ELISA kits
(Salimetrics1).
In order to investigate potential changes in salivary cortisol concentration as a result of
training methods, for each dog the baseline sample values (BL) and the post-training sample
values (PT) were averaged, and the difference between the post-training and the baseline aver-
ages computed (hereafter the post-training increase in cortisol concentration).
2.6.2. Phase 2 –Evaluating welfare outside the training context. For each dog, we calcu-
lated the average latency to reach the food bowl during each of the three types of test trials
(NP, M, NN) as well as the average latency to reach the ‘positive’ and ‘negative’ training loca-
tions during the test phase.
Seventy-three dogs completed the cognitive bias task. From these 73 subjects, 14 disengaged
from the task during some trials due to noise outside the test room. Thirteen disengaged for
one test trial (Group Aversive: five dogs at location NP, and three dogs at location NN; Group
Reward: one dog at location M, three dogs at location NP, and one dog at location NN), and
one (Group Reward) for three test trials (one at each test location). For these dogs, the average
latencies to the test locations were calculated from the remaining test trials. Of the remaining
four dogs, one (from School G) completed the first seven test trials (at locations M, NP, NN,
NP, NN, M, NN), two (one from School A and one from School E) completed the first five test
trials (at locations M, NP, NN, NP, NN), and one (from School G) completed the first three
test trials (at locations M, NP, NN); then they stopped cooperating with the task. Their average
latencies to the test locations were calculated from these trials.
• Kruskal-Wallis tests were used to evaluate the effects of group (Aversive, Mixed and Reward)
on the post-training increase in cortisol concentration, on the baseline and the post-training
levels of cortisol, as well as on the number of training classes attended by the dogs before
Phase 2.
• A Wilcoxon signed-rank test was used to compare, in the cognitive bias task, the latency to
reach the P location during test trials and during the final trial, when the bowl contained no
food, to verify whether dogs were relying on olfactory or visual cues to discriminate between
bowl locations.
Finally, a one-way ANOVA for independent samples was used to compare the number of
trials needed to reach the learning criterion in the cognitive bias task among groups (Aversive,
Mixed, Reward). All the statistical tests were two-tailed and the level of significance was set at
α = 0.05. When multiple comparisons were performed, a Bonferroni correction was applied.
Specifically, corrected p-values were used for the post-hoc pairwise comparisons performed
for the post-training increase in cortisol concentration and for the number of trials to criterion
in the cognitive bias task. The effect sizes for all the reported results were calculated as Cohen’s
d. The entire dataset is available in S2 Appendix.
3. Results
3.1. Questionnaire
3.1.1 Dog demographics and background. Concerning dog demographics, the three
groups did not differ in sex and neuter status ratios, but they differed with regards to age
(F = 13.9, p = 0.013) and FCI breed group (F = 25.3, p = 0.008). As for dog background, the
groups differed only in the age of separation from the mother (F = 20.8, p = 0.001, see Table 4).
3.1.2. Owner demographics, experience with dogs and dog training. Regarding owner
demographics, the three groups did not differ in owner age and family household size; how-
ever, they differed in owner gender (F = 8.4, p = 0.013) and in whether the household included
children (F = 6.2, p = 0.044). Regarding owner experience with dogs and dog training, the
groups did not differ in whether owners had attended training classes with a previous dog nor
in whether they had had other dog(s) in the past, but they differed in the information owners
used to choose the dog training school (F = 19.9, p = 0.005, see Table 4).
Table 4. Variables obtained from the questionnaire (dog demographics and background, and owner demographics and experience with dogs and dog training).
Fisher’s exact tests were used to compare the three Groups (Aversive, Mixed, Reward).
Group
Variable Statistical results Aversive (n) Mixed (n) Reward (n)
Dog demographics
Breed (FCI) F = 25.3, p = 0.008�
Mixed breed 4 7 20
Sheepdogs and Cattledogs 7 6 4
Pinscher and Schnauzer 7 4 1
Terriers 1 2 2
Dachshunds - - -
Spitz and primitive types - - -
Scent hounds and related breeds 1 - 3
Pointing dogs 2 - -
Retrievers, Flushing Dogs and Water Dogs 5 2 11
Companion and Toy Dogs 1 1 1
Age F = 13.9, p = 0.013�
younger than 6 months 1 - 7
6–11 months 10 14 25
1–3 years 15 8 9
4–7 years 2 - -
older than 7 years 0 - 0
Sex F = 2.7, ns
Neuter status F = 0.1, ns
Dog background
Dog origin F = 16.8, ns
Age of dog when separated from the mother F = 20.8, p = 0.001�
less than 1 month 1 - -
1–1.5 months (inclusive) 1 1 9
1.5–2 months (inclusive) 13 5 11
2–2.5 months (inclusive) 4 7 6
2.5–3 months (inclusive) 5 2 2
3–4 months (inclusive) - 4 -
4–5 months (inclusive) 1 - -
more than 5 months - - -
don’t know 3 3 14
Other animals at home F = 0.2, ns
Age of dog when adopted F = 20.7, ns
Age of dog when first taken out F = 17.0, ns
Frequency of walks during the first 5 months of life F = 22.7, ns
Owner demographics
Gender F = 8.5, p = 0.013�
Male 15 13 11
Female 13 9 31
Children F = 6.2, p = 0.044�
Yes 15 5 12
No 13 17 30
Age F = 5.3, ns
Family household size F = 5.2, ns
(Continued )
Table 4. (Continued)
Group
Variable Statistical results Aversive (n) Mixed (n) Reward (n)
Owner experience with dogs and dog training
Searched for information before choosing current school F = 19.9, p = 0.005�
Yes, by visiting dog training schools 3 1 1
Yes, on the different training methods 3 5 13
No, already knew the present school 6 3 2
No, followed third party recommendation 8 12 24
No, the present school was the closest 8 1 2
Other - - -
Owned other dog F = 4.8, ns
Training classes with other dog F = 0.2, ns
�
significant differences at α = 0.05.
https://doi.org/10.1371/journal.pone.0225023.t004
vs. Group Reward: 4.11±0.37; Z = 7.3, p<0.001, d = 1.99) compared to Group Reward. There
was also a tendency for move away to be affected by group (X22,265 = 5.3, p = 0.073). Overall,
when group affected stress-related behaviors, these were performed less frequently in Group
Reward than in both Group Aversive and Group Mixed. No effect of group was found for
scratch, yelp, whine and paw lift. None of the stress-related behaviors was affected by training
session. The average frequencies of stress-related behaviors by group are depicted in Fig 2.
Presence of children in the household was a significant confounder which increased the fre-
quency of body turn (Z = -2.4, p = 0.018), but decreased the frequency of body shake (Z = 2.4,
p = 0.016). Additionally, as dog age increased, the frequency of body turn (Z = -2.6, p = 0.011)
and yawn (Z = -2.8, p = 0.006) decreased.
There were not enough occurrences of salivating and lying on side/back to perform nega-
tive binomial regressions. Salivation frequency was (M±SEM) 0.29±0.14, 0.03±0.03, and 0.02
±0.02 in Groups Aversive, Mixed, and Reward, respectively. In a similar pattern, the frequency
of lying on side/back was 0.99±0.09 and 0.02±0.02 in Groups Aversive and Mixed, respectively,
Fig 2. Number of occurrences of each stress-related behavior averaged across the three training sessions for Group
Aversive (white bars), Group Mixed (grey bars) and Group Reward (black bars). Vertical bars show the SEM. � stands for
statistically significant differences for least square means at α = 0.05.
https://doi.org/10.1371/journal.pone.0225023.g002
Fig 3. Average number of scans in the different behavioral states in training sessions 1 (S1), 2 (S2) and 3 (S3) for
Group Aversive (left), Group Mixed (middle) and Group Reward (right).
https://doi.org/10.1371/journal.pone.0225023.g003
and no occurrences were observed in Group Reward. Fear-related elimination was never dis-
played during this study. Model details are available in S3 Appendix.
3.2.1.2. Overall behavioral state. The frequency of scans in which dogs were observed in an
excited state was lower in Group Aversive compared to both Group Mixed (Z = 6.2, p<0.001,
d = 1.62) and Group Reward (Z = 9.0, p<0.001, d = 2.63). Dogs from Group Mixed were also
observed less frequently (Z = -4.2, p<0.001, d = 1.31) in an excited state than those of the
Group Reward (Fig 3). Similarly, dogs were observed in a relaxed state less frequently in Group
Aversive than in Group Mixed (Z = -2.5, p = 0.033, d = 0.66) and Group Reward (Z = -2.8,
p = 0.017, d = 0.78), but no differences were observed between Group Mixed and Group
Reward (Z = -0.1, p = 0.999). On the other hand, dogs from Group Aversive were observed in
tense and low states more frequently than those from Group Mixed (Z = 5.9, p<0.001, d = 1.85
for tense; Z = 3.7, p<0.001, d = 1.07 for low) and Group Reward (Z = 14.6, p<0.001, d = 2.96
for tense; Z = 3.9, p<0.001, d = 0.81 for low). Dogs were also tense more frequently in Group
Mixed compared to Group Reward (Z = 7.6, p<0.001, d = 1.72 see Fig 3). Dogs were observed
more often in a low state when children were present in the household (Z = -2.6, p = 0.011).
Training session tended to affect the occurrence of the behavioral state relaxed (X22,265 =
5.1, p = 0.077) and significantly affected the behavioral state excited (X22,265 = 10.3, p = 0.006),
in that dogs were observed more frequently in an excited state in the first, S1, than in the sec-
ond, S2, (Z = 2.7, p = 0.019) and in the last, S3, (Z = 3.3, p = 0.003) training sessions.
3.2.1.3. Panting. The frequency of scans in which dogs were observed panting in Group
Aversive was higher than in Group Reward (Z = 4.6, p<0.001, d = 1.02). Panting frequency
was also observed to be higher in Group Mixed compared to Group Reward (Z = 2.5,
p = 0.042, d = 0.59, Fig 4). Training session did not affect the frequency of panting.
3.2.2. Physiological data. Baseline cortisol concentrations did not differ among groups
[Group Aversive: 0.15±0.02 vs. Group Mixed: 0.14±0.02 vs. Group Reward: 0.13±0.02 μg/dL;
H(2) = 1.689, p = 0.430], but differences among groups were found for post-training levels
[Group Aversive: 0.26±0.05 vs. Group Mixed: 0.23±0.05 vs. Group Reward: 0.13±0.02 μg/dL;
H(2) = 8.634, p = 0.013]. As a result, there was an effect of group in the average post-training
increase in cortisol concentration [H(2) = 9.852, p = 0.007]. Specifically, as depicted in Fig 5,
Fig 4. Number of scans panting averaged across training sessions for Group Aversive (left), Group Mixed
(middle) and Group Reward (right). Vertical bars show the SEM. � stands for statistically significant differences for
least square means at α = 0.05.
https://doi.org/10.1371/journal.pone.0225023.g004
the average post-training increase in cortisol was higher in Group Aversive than in Group
Reward (U = 107.0, p = 0.003, d = 0.13), but no differences were found between Group Mixed
and Group Reward (U = 90.0, p = 0.183) nor between Group Mixed and Group Aversive
(U = 39.5, p = 0.826).
Fig 5. Average post-training increase in cortisol concentration (PT-BL) for Group Aversive (left), Group Mixed
(middle) and Group Reward (right). Vertical bars show the SEM. � stands for statistically significant differences for
the averages at α = 0.05.
https://doi.org/10.1371/journal.pone.0225023.g005
Fig 6. Average latency to reach the food bowl as a function of bowl location: P—‘positive’, NP–‘near positive’,
M–‘middle’, NN–‘near negative’, N–‘negative’, for Group Aversive (white circles), Group Mixed (grey circles) and
Group Reward (black circles). Vertical bars show the SEM.
https://doi.org/10.1371/journal.pone.0225023.g006
3.3.1. Training phase. Dogs took (M±SEM) 27.14±0.85 trials to reach the learning crite-
rion. Dogs from Group Aversive, Group Mixed and Group Reward took, respectively, 28.71
±1.35, 29.61±1.79 and 24.80±1.26 trials. The differences among groups were statistically signif-
icant [F(2,74) = 3.5, p = 0.037], with dogs from Group Mixed showing a tendency to require
more trials to reach criterion than Group Reward [t(51) = -2.2, p = 0.090], but no differences
being found between Group Aversive and Group Reward [t(57) = -2.1, p = 0.124] nor between
Group Aversive and Group Mixed [t(40) = 0.4, p = 0.968].
3.3.2. Test phase. Fig 6 depicts the average latencies to reach the two training locations
(P, N) and the three test locations (NP, M, NN) for Group Aversive, Group Mixed and Group
Reward. Group and bowl location affected the latency for the dogs to reach the bowl, but there
was no significant group� bowl location interaction. Dogs of Group Aversive took longer to
reach all bowl locations (t = 2.6, p = 0.032, d = 3.99) compared to dogs from Group Reward,
but no differences were found between Group Mixed and Group Reward (t = 2.0, p = 0.153),
as well as between Group Aversive and Group Mixed (t = 0.4, p = 0.999).
Lastly, an analysis comparing the average latency to reach the P location during test trials
and the latency to reach this same location during the final trial (when the bowl contained no
food) revealed no significant differences (T = 1295.5, p = 0.328), confirming that the dogs were
not relying on olfactory or visual cues to discriminate between bowl locations.
4. Discussion
This is the first empirical study to systematically investigate the effects of different training
methods on the welfare of companion dogs within and outside the training context. We objec-
tively classified training methods, extended the study of aversive-based methods to other tech-
niques and tools besides shock collars, and used objective and validated measures for the
assessment of dog welfare within the training context (behavioral and physiological stress
responses during training) and outside the training context (cognitive bias task). Since it
became evident during data collection that the recruited dog training schools that employed
aversive-based methods did so to a substantially different extent, for analysis the participating
schools were divided into three groups: Group Aversive, composed by two schools that used
over 75% of intended aversive-based methods, Group Mixed, composed by two schools that
used less than 40% of intended aversive-based methods, and Group Reward, composed by
three schools that used no intended aversive stimuli. Overall, our results show that Group
Aversive and Group Mixed were in poorer welfare during training than Group Reward, and
that Group Aversive was also in poorer welfare than Group Reward outside the training con-
text. Additionally, although no differences between Groups Aversive and Mixed were found
outside the training context, Group Aversive displayed poorer welfare during training.
During the welfare assessment in the training sessions, dogs from Group Aversive were
observed more frequently in low behavioral states than dogs from Group Reward, and dogs
from both Group Aversive and Group Mixed were observed more frequently in tense behav-
ioral states and more frequently panting than dogs from Group Reward. Dogs from Group
Aversive were also observed more frequently in tense and low behavioral states than dogs from
Group Mixed. Tense and low body postures reflect states of distress and fear in dogs (e.g.,
[25]), while panting has been associated with acute stress in dogs (e.g., [9,26]). Additionally,
overall, dogs from Group Aversive displayed stress-related behaviors more frequently than
dogs from both Group Mixed and Group Reward, and dogs from Group Mixed displayed
stress-related behaviors more frequently than dogs from Group Reward. In previous studies,
high levels of lip licking and yawning behaviors have been consistently associated with acute
stress in dogs (e.g., [10,27]). Importantly, lip licking has been associated with stressful social
situations [27]. This most likely explains the large magnitude of this behavior observed in
Group Mixed and the even larger magnitude in Group Aversive, as aversive-based training
methods comprise social and physical confrontation with the dog. The display of avoidance
behaviors such as body turn, move away, crouch and lying on side/back, specifically in
response to training techniques, highlights the aversive nature of the training sessions at the
schools employing aversive-based methods. Notably, lying on side/back was only displayed in
Groups Aversive and Mixed (and mostly in School A, which employed the highest proportion
of intended aversive-based training methods). Finally, no differences were found among
groups for scratch, paw lift, whine and yelp. Previous studies on dog training methods have
also failed to identify significant differences in scratch and paw lift [9,10], suggesting that these
may not be reliable indicators of stress, at least in the context of training. It is possible that
scratching and paw lift behaviors are also associated with excitement and arousal rather than
just distress. In turn, whining has also been associated with attention seeking and/or food beg-
ging behavior in dogs [28], and as such, is most likely also not a reliable indicator of distress.
Finally, yelping may be interpreted as a response to pain [27]. However, besides the fact that
no differences were found between groups, this behavior occurred very rarely in the present
study. Combined with the observed differences in other stress-related behaviors, this seems to
suggest that the aversive-based methods used in the present study caused emotional rather
than physical distress. Hence, the present study shows a strong association between the use of
aversive-based training methods and an increased frequency of stress behaviors during com-
panion dog training. However, our results also show that the proportion of aversive-based
methods used also matters, with lower proportions of aversive stimuli resulting in lower fre-
quencies of stress behaviors exhibited by the dogs. These results strengthen and extend the
findings of previous studies on companion dogs, which suggested a positive correlation
between the use of both shock collars [9] and other aversive techniques [10] with stress behav-
iors in the context of dog training.
An effect of training session was found for the behavioral state excited, with dogs being
more frequently excited in Session 1 than in Sessions 2 and 3. This result is most likely a conse-
quence of dogs’ familiarization with the training context. The tendency of the relaxed behav-
ioral state to increase with training session possibly reflects the reduction in excitement.
With regards to physiological measures of stress, the average post-training increase in corti-
sol concentration (PT-BL) was higher in Group Aversive than in Group Reward, whereas no
differences were found between Group Mixed and Group Reward nor between Group Aver-
sive and Group Mixed. Previous studies investigating cortisol levels in dogs in relation to train-
ing have yielded contradictory results. Schalke et al (2007) [29] found significant differences in
the cortisol levels of three groups of laboratory dogs trained using shock collars with different
degrees of shock predictability (the lower the predictability, the higher the cortisol levels).
However, studies comparing aversive- and reward-based training methods have found either
no significant differences or the opposite pattern: the effect on cortisol from shock collar and
lemon-spray bark collars did not differ from the control treatment [9,30], and a negative pun-
ishment training method (a quitting signal) resulted in higher levels of cortisol than the use of
a pinch collar (aversive-based technique) [31]. Hence, the present study is the first to report a
significant increase in cortisol levels in dogs trained with aversive-based methods as compared
to dogs trained with reward-based methods.
The average post-training increase in cortisol levels observed in the present study
(M = 0.11 μg/dL for Group Aversive and M = 0.08 μg/dL for Group Mixed) was lower than
those reported in other studies that found significant increases after dogs were exposed to aver-
sive stimuli (0.20–0.30 μg/dL in [29] and 0.30–0.40 μg/dL in [23]). One possible explanation
for this difference in magnitude may be related to the nature of the stimuli used in the different
studies. Whereas the reported elevations in cortisol in [29] and [23] appeared after the presen-
tation of non-social stimuli (shocks in [29], and shocks, sound blasts and a falling bag in [23]),
the stimuli used during training in the present study were mainly of a social nature (i.e.: leash
jerks, physical manipulation or yelling at the dog). Stimuli administered in a social context
may be more predictable or better anticipated and, therefore, generate less acute stress
responses [23]. In support of this view, [23] did not find elevations in cortisol after the presen-
tation of social stimuli (physical restraint and opening an umbrella).
When considering welfare outside the training context, we found that, in the cognitive bias
task, dogs from Group Aversive displayed higher latencies for all the stimuli than dogs from
Group Reward, with no differences being found between Groups Aversive and Mixed nor
between Groups Reward and Mixed. Although affect is hypothesized to exert a greater influ-
ence on decision-making under ambiguity (i.e., for the test stimuli: NN, M, NP) than under
certainty (i.e., for training stimuli: N, P), other studies in cognitive bias have also found differ-
ences for both test and training stimuli [e.g., 32–35, see 35 for a review]. This type of result,
with differences found for (at least one of) the training stimuli has also been interpreted as evi-
dence for differences in the valence of the affective states. The fact that differences can emerge
for both training and test stimuli has been proposed to result from the fact that choice in the
cognitive bias task depends on two different components of the decision-making process: per-
ceived probability and perceived valuation of rewards (and punishments). An individual may
be less likely to make a less ‘risky’ or more ‘pessimistic’ response if they consider the reward to
be less probable (or punisher more probable) and/or if they consider the reward to be less valu-
able (or the punisher more aversive) [35,36]. In summary, affective states may influence the
responses to both the training and the test stimuli in the cognitive bias task, although different
components of the decision-making process may be playing a role. Therefore, the most likely
explanation for the present findings is that dogs from Group Aversive considered the food
reward less probable (as indicated by the higher latencies to the test stimuli) and also showed a
higher valuation of reward loss relative to win (as indicated by the higher latencies to the train-
ing stimuli) [36]. Overall, these results indicate that dogs from Group Aversive were in a less
positive affective state than dogs from Group Reward. To our knowledge, the only other study
in dogs that addressed the welfare effects of training methods outside the training context was
performed by Christiansen et al (2001) [13], who studied the use of shock collars to train hunt-
ing dogs not to attack sheep. No general effects of the use of shock collars on dog fear and anxi-
ety were found one year after training took place. However, unlike the test used by
Christiansen et al (2001) [13], which was a modified version of a temperament test used by the
Norwegian Kennel Club, the cognitive bias approach used in the current study is a widely
established and well-validated method for evaluating animal welfare (e.g., [14–18]). Hence, to
our knowledge, this is the first study to reliably assess and report the effects of aversive- and
reward-based training methods in the affective states of dogs outside the training context.
Dogs from Group Reward showed a tendency to learn the cognitive bias task faster than
dogs from Group Mixed. Similar findings were observed previously by Rooney et al. (2011)
[37], who found a positive correlation between the reported use of reward-based training
methods and a dog’s ability to learn a novel task (touching a spoon with its nose). In another
study, Marshall-Pescini et al (2008) [38] found that dogs with high-level training experience
were more successful at opening a box to obtain food than dogs which had received either
none or only basic training. Although Marshall-Pescini et al (2008) [38] reported that all sub-
jects’ training included positive reinforcement methods, they did not specify whether positive
punishment and/or negative reinforcement were used in combination. Altogether, previous
research suggests that training using positive reinforcement may improve the learning ability
of dogs. It remains unclear why a difference was not observed between Group Aversive and
Group Reward in the present study. Still, it is important to mention that in all previous studies
cited above, animals were required to perform a given behavior in order to obtain a positive
reinforcer. Thus, it is unclear whether the same effect would stand if the dogs had to learn a
task whose goal was, for example, to perform a behavior to escape from an unpleasant situa-
tion. It may be the case that dogs trained with positive reinforcement develop a specific ‘learn-
ing set’ [39] for tasks involving positive reinforcement, but that dogs trained with aversive-
based methods perform better in tasks involving some sort of aversive stimuli. Further
research is needed to clarify the relationship between training methods and learning ability in
dogs.
Notably, we found that the higher the proportion of aversive stimuli used in training, the
greater the impact on the welfare of dogs (both within and outside the training context). This
result is in line with the findings of a previous survey study, which showed that a higher fre-
quency of punishment was correlated with higher anxiety and fear scores [8]. Still, in the pres-
ent study, welfare differences were found even when comparing Groups Reward and Mixed,
which used a lower proportion of intended aversive-based techniques as compared to Group
Aversive. Dogs from Group Mixed showed higher frequencies of stress-related behaviors, were
found more frequently in tense states and panted more frequently during training than dogs
from Group Reward. When comparing Group Mixed and Group Aversive, the latter showed a
higher frequency of stress-related behaviors and was more frequently found in tense and low
behavioral states during training. This seems to suggest that, although dogs trained in ‘low
aversive’ schools do not show as many indicators of poor welfare as those trained in ‘highly
aversive’ schools, their welfare may still be at stake.
Moreover, our results suggest that the proportion of aversive stimuli used in training plays
a greater role on dogs’ stress levels than the specific training tools used. As an example, one
school from Group Mixed used pinch and e-collars, whereas another school from Group Aver-
sive only used choke collars during training. Although the tools used by the former school may
be perceived as more aversive, the frequency of stress behaviors was higher in dogs being
trained at the latter school. The type of (intended) positive reinforcers also appears to be rele-
vant. All schools except the aforementioned school from Group Aversive used primarily food
treats as rewards, whereas the latter only used petting. Although this was not the school using
the highest proportion of aversive stimuli, it was the school whose dogs showed the highest fre-
quency of stress behaviors (data not shown). Previous research has shown that petting is a less
effective reward than food in training [40]. Having a highly valuable reward might thus be
important in reducing stress when aversive stimuli are used in training. The goal of the present
study was to test the overall effect on dog welfare of aversive- and reward-based methods as
they are used in the real world, but it may be interesting for future studies to focus on disentan-
gling the effects of the different types of stimuli used in training (as has been done with e-col-
lars) [e.g., 9,25].
Finally, some limitations of the present study must be considered. Firstly, because this was a
quasi-experimental rather than an experimental study, we cannot infer a true causal relation-
ship between training methods and dog welfare. To do so would require a randomized control
trial. However, conducting an experimental study where dogs are designedly subjected to aver-
sive-based methods would raise ethical concerns, as previous studies have already suggested an
association between the use of aversive-based methods and indicators of stress in dogs [3, but
see 9]. Because we did not randomly allocate dogs to the treatments (training methods), we
cannot discard the possibility that there are significant differences between dog-owner pairs
that led some owners to choose an aversive-based school and others to choose a reward-based
school. There were indeed differences among groups in owner gender, in whether or not the
household included children and in the information owners relied on for choosing the dog
training school. There were also differences among groups in dog age, FCI breed group and
age of separation from the mother. The study was not designed to evaluate the effect of these
factors and they were therefore treated as potential confounders in the statistical analysis, in
order to account for the possibility that they would affect our results. The effects of training
method reported in the study are robust to these confounders. We tested for dog age, presence
of children in the household and owner gender, factors which have been shown to potentially
affect dog stress and welfare [e.g., 41–44]. The presence of children in the family has been
found to be negatively associated with the owners’ perception of the relationship with their
dogs, in what is to our knowledge the only study addressing how this factor affects dog behav-
ior [43]. Most research into the relationship between dog age and stress indicators has been
conducted in senior dogs and consistently shows higher baseline cortisol and higher cortisol
responses to stressful stimuli in aged dogs [45,46]; however, our study did not include any
senior dog. Schöberl et al (2017) [44] found cortisol to decrease with increasing age of the dog
in adult dogs, whereas Henessy et al (1998) [42] found that the juveniles and adults had higher
plasma cortisol levels than puppies. Two of the potential confounders were not included in the
analysis because of insufficient reliable data: breed (34% mixed breeds, mainly unknown) and
age of separation from the mother (22% unknown). Breed differences in behavior are well
established [43] but the classification of breeds into groups has not been found to systemati-
cally correlate with behavioral similarities [e.g., 47], and the large percentage of mixed breed
dogs where the actual breeds were unknown further constrains a meaningful analysis of this
factor in our sample. Literature shows that both early [e.g., 48] and late [e.g., 49,50] separation
from the mother (before and after 8 weeks-old, respectively) can be associated with stress-
related behavioral problems in dogs. Whereas we do not know the animals’ stress levels before
the start of training, cortisol data shows no differences between training groups on non-train-
ing days.
Secondly, a volunteer bias cannot be excluded and hence any generalization of the present
results must take this in account. Finally, this study focused on welfare and did not compare
the efficacy of training methods. Presently, the scientific literature on the efficacy of the differ-
ent methodologies is scarce and inconsistent [3]. Whereas some studies suggest a higher effi-
cacy of reward methods [5,12,51–53], one points in the opposite direction [31] and three show
no differences between methods [9,54,55]. This limits the extent of evidence-based recommen-
dations. If reward-based methods are, as the current results show, better for dog welfare than
aversive-based methods, and also prove to be more or equally effective to aversive-based meth-
ods, there is no doubt that owners and dog professionals should use reward-based training
practices. If, on the other hand, aversive-based methods prove to be more effective, the recom-
mendation may be to use aversive stimuli as infrequently as possible during training, and use
them in combination with reward-based techniques. This applies not only to training in a for-
mal school setting but whenever owners use reinforcement or punishment in their interactions
with the dog.
5. Conclusions
Overall, our results show that companion dogs trained with aversive-based methods experi-
enced poorer welfare during training sessions than dogs trained with reward-based methods.
Additionally, dogs trained with higher proportions of aversive-based methods experienced
poorer welfare outside the training context than dogs trained with reward-based methods.
Moreover, whereas different proportions of aversive-based methods did not result in differ-
ences in dog welfare outside the training context among aversive-based schools, a higher pro-
portion of aversive-based methods resulted in poorer welfare during training. To our
knowledge, this is the first comprehensive and systematic study to evaluate and report the
effects of dog training methods on companion dog welfare. Critically, our study points to the
fact that the welfare of companion dogs trained with aversive-based methods is at risk, espe-
cially if these are used in high proportions.
Supporting information
S1 Appendix. Proportion (mean ± standard deviation) of intended aversive-based tech-
niques used during the six training sessions analyzed for each training school. For each
training session, the number of intended positive punishments and negative reinforcements
was divided by the total number of intended positive punishments, negative reinforcements,
positive reinforcements and negative punishments. Schools A and D were categorized as
Group Aversive, Schools C and F as Group Mixed and Schools B, E and G as Group Reward.
(DOCX)
S2 Appendix. Raw data underlying all the analyzes performed in the current research
paper.
(XLSX)
S3 Appendix. Negative binomial and generalized linear mixed model details. S3a Table.
Analysis of Generalized Estimating Equation for the stress-related behaviors analysis. S3b
Table. Analysis of Generalized Estimating Equation for the behavioral state analysis. S3c Table.
Analysis of Generalized Estimating Equation for the panting analysis. S3d Table. Solutions for
fixed effects from the generalized linear mixed model for the cognitive bias analysis.
(DOCX)
Acknowledgments
We are grateful, first and foremost, to all dogs and their owners who participated in this study;
without them this research would never have been possible. A very special acknowledgment to
the dog training schools and their trainers that opened their doors for our participant recruit-
ment and data collection.
We would also like to thank Joana Guilherme-Fernandes for the support provided in the
development of the setup for the cognitive bias task, for helping with data collection and espe-
cially for all the invaluable discussions during study planning and data interpretation. A special
acknowledgment also for Margarida Lencastre and Flávia Canastra, who also helped in data
collection. We also want to thank Igor M Lopes for the critical help provided with statistical
analysis and Jennifer Barrett for input given during data collection and analysis.
Finally, we would like to thank Dr. Carolyn Walsh and the anonymous reviewers for the
detailed input on our work, which helped to substantially improve its quality.
Author Contributions
Conceptualization: Ana Catarina Vieira de Castro, Liliana de Sousa, I. Anna S. Olsson.
Data curation: Ana Catarina Vieira de Castro.
Formal analysis: Ana Catarina Vieira de Castro, Danielle Fuchs, Gabriela Munhoz Morello,
Stefania Pastur.
Funding acquisition: Ana Catarina Vieira de Castro, I. Anna S. Olsson.
Investigation: Ana Catarina Vieira de Castro, Danielle Fuchs, Stefania Pastur.
Methodology: Ana Catarina Vieira de Castro.
Project administration: Ana Catarina Vieira de Castro, I. Anna S. Olsson.
Resources: Liliana de Sousa, I. Anna S. Olsson.
Supervision: Ana Catarina Vieira de Castro, Liliana de Sousa, I. Anna S. Olsson.
Validation: Ana Catarina Vieira de Castro, I. Anna S. Olsson.
Visualization: Ana Catarina Vieira de Castro.
Writing – original draft: Ana Catarina Vieira de Castro.
Writing – review & editing: Ana Catarina Vieira de Castro, Danielle Fuchs, Gabriela Munhoz
Morello, I. Anna S. Olsson.
References
1. Reid P. Learning in dogs. In: Jensen P, editor. The Behavioural Biology of Dogs. CABI; 2007. pp. 120–
144.
2. Reisner I. The learning dog: A discussion of training methods. In: Serpell J, editor. The Domestic Dog:
Its Evolution, Behavior and Interactions with People, 2nd Edition. Cambridge University Press;
2017. pp. 211–226.
3. Guilherme-Fernandes J, Olsson IAS, Vieira de Castro AC. Do aversive-based training methods actually
compromise dog welfare?: A literature review. Appl Anim Behav Sci. 2017; 196, 1–12.
4. Hineline PN, Rosales-Ruiz J. Behavior in relation to aversive events: Punishment and negative rein-
forcement. In Madden GJ, Dube WV, Hackenberg TD, Hanley G P, Lattal KA, editors. APA Handbook
of Behavior Analysis, Vol. 1. Methods and Principles. APA Handbooks in Psychology®; 2013. pp.
483–512. https://doi.org/10.1158/1078-0432.CCR-12-3764 PMID: 23596103
5. Hiby EF, Rooney NJ, Bradshaw JWS. Dog training methods: their use, effectiveness and interaction
with behaviour and welfare. Anim Welf. 2004; 13, 63–69.
6. Blackwell EJ, Twells C, Seawright A, Casey RA. The relationship between training methods and the
occurrence of behavior problems, as reported by owners, in a population of domestic dogs. J Vet
Behav. 2008; 3, 207–217.
7. Herron ME, Shofer FS, Reisner IR. Survey of the use and outcome of confrontational and non-confron-
tational training methods in client-owned dogs showing undesired behaviors. Appl Anim Behav Sci.
2009; 117, 47–54.
8. Arhant C, Bubna-Littitz H, Bartles A, Futschik A, Troxler J. Behaviour of smaller and larger dogs: effects
of training methods, inconsistency of owner behavior and level of engagement in activities with the dog.
Appl Anim Behav Sci. 2010; 123, 131–142.
9. Cooper JJ, Cracknell N, Hardiman J, Wright H, Mills D. The welfare consequences and efficacy of train-
ing pet dogs with remote electronic training collars in comparison to reward based training. PLoS One.
2014; 9, e102722. https://doi.org/10.1371/journal.pone.0102722 PMID: 25184218
10. Deldalle S, Gaunet F. Effects of 2 training methods on stress-related behaviors of the dog (Canis famil-
iaris) and on the dog-owner relationship. J Vet Behav. 2014; 9, 58–65.
11. Companion Animal Welfare Council, 2012. The Use of Electric Pulse Training Aids (EPTAs) in Compan-
ion Animals. Available from: http://eprints.lincoln.ac.uk/14640/1/CAWC%20ecollar%20report.pdf
(Accessed 12 August 2019).
12. Haverbeke A, Laporte B, Depiereux E, Giffroy JM, Diederich C. Training methods of military dog han-
dlers and their effects on the team’s performances. Appl Anim Behav Sci. 2008; 113, 110–122.
13. Christiansen FO, Bakken M, Braastad BO. Behavioural changes and aversive conditioning in hunting
dogs by the second-year confrontation with domestic sheep. Appl Anim Behav Sci. 2001; 72, 131–143.
https://doi.org/10.1016/s0168-1591(00)00203-3 PMID: 11278032
14. Mendl M, Paul ES. Animal affect and decision-making. Neurosci Biobehav Rev. 2020; 112, 144–163.
https://doi.org/10.1016/j.neubiorev.2020.01.025 PMID: 31991192
15. Harding EJ, Paul ES, Mendl M. Animal behaviour: cognitive bias and affective state. Nature. 2004; 427,
312–312. https://doi.org/10.1038/427312a PMID: 14737158
16. Boissy A, Erhard HW. How studying interaction between animal emotion, cognition, and personality can
contribute to improve farm animal welfare. In: Grandin T, Deesing MJ, editors. Genetics and the Behav-
ior of Domestic Animals, 2nd Edition. Academic Press; 2014. pp. 81–113.
17. Mendl M, Burman OHP, Parker RMA, Paul ES. Cognitive bias as an indicator of animal emotion and
welfare: emerging evidence and underlying mechanisms. Appl Anim Behav Sci. 2009; 118, 161–181.
18. Mendl M, Brooks J, Basse C, Burman O, Paul E, Blackwell E, et al. Dogs showing separation-related
behaviour exhibit a ‘pessimistic’ cognitive bias’. Curr Biol. 2010; 20, R839—R840. https://doi.org/10.
1016/j.cub.2010.08.030 PMID: 20937467
19. Barker SB, Wolen AR. The benefits of human-companion animal interaction: a review. J Vet Med Educ.
2008; 35(4), 487–495. https://doi.org/10.3138/jvme.35.4.487 PMID: 19228898
20. Crawford EK, Worsham NL, Swinehart ER. Benefits derived from companion animals, and the use of
the term “attachment”. Anthrozoos. 2006; 19 (2), 98–112.
21. Skinner BF. Science and Human Behavior. Macmillan; 1953.
22. Dreschel NA, Granger DA. Methods of collection for salivary cortisol measurement in dogs. Horm
Behav. 2009; 55(1), 163–188. https://doi.org/10.1016/j.yhbeh.2008.09.010 PMID: 18948108
23. Beerda B, Schilder MBH, van Hooff JARAM, de Vries HW, Mol JA. Behavioural, saliva cortisol and
heart rate responses to different types of stimuli in dogs. Appl Anim Behav Sci. 1998; 58, 365–381.
24. Kobelt A, Hemsworth PH, Barnett JL, Butler KL. Sources of sampling variation in saliva cortisol in dogs.
Res Vet Sci. 2003; 75, 157–161. https://doi.org/10.1016/s0034-5288(03)00080-8 PMID: 12893165
25. Beerda B, Schilder MBH, van Hooff JARAM, de Vries HW. Manifestations of chronic and acute stress in
dogs. Appl Anim Behav Sci. 1997; 52, 307–319.
26. Voith VL, Borchelt PL. Fears and phobias in companion animals. In: Voith VL, Borchelt PL, editors.
Readings in companion animal behaviour. Trenton, NJ: Veterinary Learning Systems; 1996. pp. 140–
152. https://doi.org/10.1002/eji.1830261009 PMID: 8898941
27. Schilder MBH, van der Borg JAM. Training dogs with help of the shock collar: short and long term beha-
vioural effects. Appl Anim Behav Sci. 2004; 85, 319–334.
28. Mills DS, Beral A, Lawson S. Attention seeking behavior in dogs–what owners love and loathe! J Vet
Behav: Clin Ap Res. 2020; 5, 60.
29. Schalke E, Stichnoth J, Ott S, Jones-Baade R. Clinical signs caused by the use of electric training col-
lars on dogs in everyday life situations. Appl Anim Behav Sci. 2007; 105, 369–380.
30. Steiss JE, Schaffer C, Ahmad HA, Voith VL. Evaluation of plasma cortisol levels and behavior in dogs
wearing bark control collars. Appl Anim Behav Sci. 2007; 106, 96–106.
31. Salgirli Y, Schalke E, Hackbarth H. Comparison of learning effects and stress between 3 different train-
ing methods (electronic training collar, pinch collar and quitting signal) in Belgian Malinois Police Dogs.
Rev Méd Vét. 2012; 163, 530–535.
32. Horváth M, Pichová K, Košťál L. The effects of housing conditions on judgement bias in Japanese quail.
Appl Anim Behav Sci. 2016; 185, 121–130.
33. Deakin A, Mendl M, Browne WJ, Paul ES, Hodge JJL. State-dependent judgement bias in Drosophila:
evidence for evolutionarily primitive affective processes. Biol Lett. 2018; 14: 20170779. https://doi.org/
10.1098/rsbl.2017.0779 PMID: 29491031
34. Zidar J, Campderrich I, Jansson E, Wichman A, Winber S, Keeling L, et al. Environmental complexity
buffers against stress-induced negative judgement bias in female chickens. Sci Rep. 2018; 8:5404.
https://doi.org/10.1038/s41598-018-23545-6 PMID: 29599444
35. Neville V, Nakagawa S, Zidar J, Paul ES, M Lagisz M, Bateson M, et al. Pharmacological manipulations
of judgement bias: A systematic review and meta-analysis. Neurosci Biobehav Rev. 2020; 108: 269–
286. https://doi.org/10.1016/j.neubiorev.2019.11.008 PMID: 31747552
36. Iigaya K, Jolivald A, Jitkrittum W, Gilchrist ID, Dayan P, Paul E, et al. Cognitive Bias in Ambiguity Judge-
ments: Using Computational Models to Dissect the Effects of Mild Mood Manipulation in Humans. PLoS
ONE. 2016; 11(11): e0165840. https://doi.org/10.1371/journal.pone.0165840 PMID: 27829041
37. Rooney NJ, Cowan S. Training methods and owner-dog interactions: links with dog behaviour and
learning ability. Appl Anim Behav Sci. 2011; 132, 169–177.
38. Marshall-Pescini S, Valsecchi P, Petak I, Accorsi PA, Prato-Previde E. Does training make you smar-
ter? The effects of training on dogs’ performance (Canis familiaris) in a problem solving task. Behav Pro-
cess. 2008; 78, 449–454.
39. Harlow HF. The formation of learning sets. Psychol Rev. 1949; 56, 51–65. https://doi.org/10.1037/
h0062474 PMID: 18124807
40. Feuerbacher EN, Wynne CD. Relative efficacy of human social interaction and food as reinforcers for
domestic dogs and hand-reared wolves. J Exp Anal Behav. 2012; 98(1): 105–29 https://doi.org/10.
1901/jeab.2012.98-105 PMID: 22851794.
41. Buttner AP, Thompson B, Strasser R, Santo J. Evidence for a synchronization of hormonal states
between humans and dogs during competition. Physiol Behav. 2015; 147, 54–62. https://doi.org/10.
1016/j.physbeh.2015.04.010 PMID: 25862521
42. Hennessy MB, Williams MT, Miller D, Douglas CW, Voith VL. Influence of male and female petters on
plasma cortisol and behaviour: can human interaction reduce the stress of dogs in a public animal shel-
ter? Appl Anim Behav Sci. 1998; 61(1), 63–77.
43. Meyer I, Forkman B. Dog and owner characteristics affecting the dog–owner relationship. J Vet Behav.
2014; 9(4), 143–150.
44. Schöberl I, Wedl M, Beetz A, Kotrschal K. Psychobiological Factors Affecting Cortisol Variability in
Human-Dog Dyads. PLoS One. 2017; 12(2): e0170707. https://doi.org/10.1371/journal.pone.0170707
PMID: 28178272
45. Piotti P, Szabó D, Bognár Z, Egerer A, Hulsbosch P, Carson RS, et al. Effect of age on discrimination
learning, reversal learning, and cognitive bias in family dogs. Learn Behav. 2018; 46(4): 537–553.
https://doi.org/10.3758/s13420-018-0357-7 PMID: 30251103
46. Reul JMHM, Rothuizen J, de Kloet ER. Age-related changes in the dog hypothalamic-pituitary-adreno-
cortical system: Neuroendocrine activity and corticosteroid receptors. J Steroid Biochem Mol Biol.
1991; 40(1–3): 63–69. https://doi.org/10.1016/0960-0760(91)90168-5 PMID: 1659883
47. Turcsán B, Kubinyi E, Miklósi A. Trainability and boldness traits differ between dog breed clusters
based on conventional breed categories and genetic relatedness. Appl Anim Behav Sci. 2011; 132:
61–70.
48. Tiira K, Lohi H. Reliability and validity of a questionnaire survey in canine anxiety research. Appl Anim
Behav Sci. 2014; 155: 82–92.
49. Jokinen O, Appleby D, Sandbacka-Saxén S, Appleby T, Valros A. Homing age influences the preva-
lence of aggressive and avoidance-related behaviour in adult dogs. Appl Anim Behav Sci. 2017; 195:
87–92.
50. Puurunen J, Hakanen E, Salonen MK, Mikkola S, Sulkama S, Araujo C, et al. Inadequate socialisation,
inactivity, and urban living environment are associated with social fearfulness in pet dogs. Sci Rep.
2020; 10(1): 1–10. https://doi.org/10.1038/s41598-019-56847-4 PMID: 31913322
51. China L, Mills DS, Cooper JJ. Efficacy of Dog Training With and Without Remote Electronic Collars vs.
a Focus on Positive Reinforcement. Front Vet Sci. 2020; 7, 508. https://doi.org/10.3389/fvets.2020.
00508 PMID: 32793652
52. Blackwell EJ, Bolster C, Richards G, Loftus BA, Casey RA. The use of electronic collars for training
domestic dogs: estimated prevalence, reasons and risk factors for use, and owner perceived success
as compared to other training methods. BMC Vet Res. 2012; 8, 93. https://doi.org/10.1186/1746-6148-
8-93 PMID: 22748195
53. Haverbeke A, Messaoudi F, Depiereux E, Stevens M, Giffroy JM, Diederich C. Efficiency of working
dogs undergoing a new Human Familiarization and Training Program. J Vet Behav. 2010; 5: 112–119.
54. Sankey C, Richard-Yris MA, Henry S, Fureix C, Nassur F, Hausberger M. Reinforcement as a mediator
of the perception of humans by horses (Equus caballus). Anim Cogn. 2010; 13: 753–764. https://doi.
org/10.1007/s10071-010-0326-9 PMID: 20490592
55. Visser EK, Van Dierendonck M, Ellis AD, Rijksen C, Van Reenen. A comparison of sympathetic and
conventional training methods on responses to initial horse training. Vet J. 2009; 181: 48–52. https://
doi.org/10.1016/j.tvjl.2009.03.009 PMID: 19375363
In this paper, we review basic and applied findings on punishment and discuss the im-
portance of conducting further research in this area. The characteristics of responding
during punishment and numerous factors that interact with basic processes are delineated
in conjunction with implications for the treatment of behavior disorders in clinical pop-
ulations. We conclude that further understanding of punishment processes is needed to
develop a highly systematic, effective technology of behavior change, including strategies
for improving the efficacy of less intrusive procedures and for successfully fading treat-
ment.
DESCRIPTORS: behavior disorders, functional analysis, punishment, treatment
Punishment is generally defined as an en- ment have been studied extensively in the
vironmental change contingent on behavior laboratory. Nevertheless, basic research on
that produces a decrease in responding over punishment has been declining rapidly de-
time (Michael, 1993). Numerous procedural spite substantial gaps in knowledge (Baron,
variations of punishment have been devel- 1991; Crosbie, 1998). The generality of basic
oped for clinical use. Results of research con- findings to clinical populations and problems
ducted over the past four decades have also is questionable (Hayes & McCurry,
shown that punishment is effective in reduc- 1990). Most studies evaluated the effects of
ing problem behavior in clinical popula- intense, unconditioned punishers (e.g., elec-
tions, and in some cases, may be an essential tric shock), and a number of important re-
component of treatment (see Kazdin, 2001, lations have not yet been replicated with hu-
and O’Brien, 1989, for reviews of this lit- mans or clinically relevant punishers.
erature). However, more knowledge is need- The purpose of this paper is to review ba-
ed about factors that may influence the ef- sic and applied research findings on punish-
fects of punishment on problem behavior. ment, identify gaps in the literature, and dis-
Few strategies have been identified for en- cuss the implications of these findings for the
hancing the effectiveness of less intrusive use of punishment in clinical settings. Basic
punishment procedures, for attenuating un- findings that contradict common assump-
desirable aspects of punishment, or for suc- tions about punishment effects found in text-
cessfully fading treatment with punishment. books and review papers and that help elu-
The direct and indirect effects of punish- cidate inconsistent results on punishment in
the applied literature are highlighted. The
We thank Don Baer, Alan Baron, Linda LeBlanc, main premise of this paper is that further un-
Tony Nevin, and the anonymous reviewers for their derstanding of punishment processes may
comments on earlier versions of this paper. lead to an improved technology of behavior
Reprints may be obtained by contacting Dorothea change. We extend the most recent review
C. Lerman, 236 Audubon Hall, Louisiana State Uni-
versity, Baton Rouge, Louisiana 70803 (e-mail: papers on punishment (Matson & Di-
dlerman@lsu.edu). Lorenzo, 1984; Van Houten, 1983) by (a)
431
432 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
providing a broader overview of the direct Knowledge about punishment also is im-
and indirect effects of punishment and fac- portant because common treatments that are
tors that influence basic processes (e.g., his- associated with other processes may in fact
tory), (b) identifying areas in need of further reduce problem behavior through the mech-
research from both basic and applied litera- anism of punishment. For example, proce-
tures, and (c) discussing recent research find- dures such as response blocking, guided
ings on punishment within the context of ad- compliance, and the application of protec-
vances in the functional analysis of behavior.1 tive equipment are often presumed to reduce
Some authors have suggested that addi- problem behavior by terminating the rein-
tional applied research on punishment is un- forcement contingency that maintains the
necessary in light of refinements to the func- response (i.e., through extinction; e.g., Reid,
tional analysis methodology and treatment Parsons, Phillips, & Green, 1993; Rincover,
with reinforcement (Donnellan & LaVigna, 1978). Research findings suggest that these
1990; Guess, Helmstetter, Turnbull, & procedural variations of extinction may
Knowlton, 1987). Results of numerous stud- function as punishment instead of, or in
ies conducted over the past 15 years have combination with, extinction (e.g., Lerman
shown that the function of problem behav- & Iwata, 1996b; Mazaleski, Iwata, Rodgers,
ior often can be determined and that this Vollmer, & Zarcone, 1994). Some authors
information can be used to develop treat- also have suggested that the contingent loss
ments based on extinction, reinforcement, of reinforcement associated with differential
and other processes such as establishing op- reinforcement procedures (e.g., differential
erations (e.g., Iwata, Pace, Dorsey, et al., reinforcement of other behavior [DRO])
1994). Nevertheless, punishment may be may constitute a form of punishment (e.g.,
critical to treatment success when the vari- Rolider & Van Houten, 1990). Thus, the
ables maintaining problem behavior cannot process of punishment may underlie a num-
be identified or controlled (for further dis- ber of popular function-based treatments.
cussion, see Axelrod, 1990; Iwata, Vollmer, Functional analysis methodology now
& Zarcone, 1990; Vollmer & Iwata, 1993). permits more precise investigations of im-
Punishment also may be preferable to rein- portant environment–behavior relations in
forcement-based treatments when problem the area of punishment. Basic findings in-
behavior must be suppressed rapidly to pre- dicate that various parameters of reinforce-
vent serious physical harm (Dura, 1991; see ment influence the direct and indirect effects
also Iwata et al.; Vollmer & Iwata). More of punishment and interact with nearly ev-
important, results of several studies indicate ery other factor that has been found to in-
that treatments derived from functional fluence responding during punishment (e.g.,
analyses (e.g., differential reinforcement of punishment schedule and magnitude). Such
alternative behavior [DRA]) may not always interactions have important clinical impli-
reduce behavior to clinically acceptable levels cations because punishment is more likely to
without a punishment component (e.g., be used when the response–reinforcer rela-
Grace, Kahng, & Fisher, 1994; Hagopian, tion cannot be terminated completely. In ap-
Fisher, Sullivan, Acquisto, & LeBlanc, 1998; plied studies, important reinforcement vari-
Wacker et al., 1990). ables identified via functional analysis could
be manipulated prior to and during punish-
1 Much of the applied research on functional anal-
ment, even when the behavior is maintained
ysis and treatment of behavior disorders has been con-
ducted with individuals diagnosed with developmental by nonsocial consequences (i.e., the behavior
disabilities. Thus, this review reflects this emphasis. is maintained in the absence of socially me-
PUNISHMENT 433
diated reinforcers, such as attention, tangible than 30 years ago with nonhumans (see
items, and escape from instructions; see, e.g., Azrin & Holz, 1966, for a review of this
Lerman & Iwata, 1996b). literature). Procedural variations of punish-
Important issues related to the ethics and ment examined in the laboratory have in-
acceptability of using punishment to treat cluded the delivery of stimuli, often called
problem behavior in individuals with devel- positive punishment, and the removal of stim-
opmental disabilities have been the subject of uli, often called negative punishment (i.e., re-
numerous articles over the past 30 years. An sponse cost, time-out from positive rein-
overview of these issues is beyond the scope forcement).2 The majority of studies, how-
of this paper but can be found in a variety ever, employed contingent electric shock. In
of sources (see Donnellan & LaVigna, 1990; the earliest basic studies, the effects of pun-
Emerson, 1992; Guess et al., 1987; Jacob- ishment were evaluated while the punished
Timm, 1996; Sidman, 1989; Van Houten et response was undergoing extinction (e.g.,
al., 1988). Suggestions for further applied Estes, 1944; Skinner, 1938; Thorndike,
studies on punishment are made throughout 1932). However, in most subsequent studies,
this paper, with the assumption that pertinent punishment contingencies were introduced
guidelines and cautions about the application with no change in the prevailing reinforce-
of punishment will accompany published re- ment schedule. This latter arrangement as-
search findings (see Alberto & Troutman, sured a certain level of responding by which
1999; Lovaas & Favell, 1987; Matson & to examine the effects of punishment inde-
DiLorenzo, 1984). pendently of those produced by extinction
Ultimately, the consumers of behavioral (Azrin & Holz, 1966). Complex interactions
technologies (e.g., clinicians, caregivers) will between reinforcement and punishment pro-
determine which treatments are used with cesses also could be evaluated. A number of
individuals with developmental disabilities authors suggested that this laboratory ar-
(see Iwata, 1988, for a cogent discussion of rangement may be more pertinent to appli-
this issue). These decisions are at least partly cation because punishment is most likely to
guided by information generated by the sci- be used when the reinforcer that maintains
entific community. Safe, acceptable, and problem behavior cannot be identified or
highly effective technologies of behavior controlled (Azrin & Holz, 1966; Dinsmoor,
change should be available to consumers 1952).
who request them, including procedures that In fact, the function of problem behavior
are based on punishment. was not determined prior to treatment in
Basic and applied research findings on most clinical studies on punishment. Pun-
clinically relevant factors that influence the ishment thus was superimposed on an un-
direct effects of punishment will be discussed known schedule of reinforcement that likely
in the first half of the paper. Other charac- took the form of extinction when the be-
teristics of punished responding, including havior was maintained by social consequenc-
maintenance, generalization, and side effects,
will be discussed in the second half. 2 Consistent with previous articles and chapters on
punishment, few distinctions will be drawn between
positive and negative punishment in this review. Cur-
FACTORS THAT INFLUENCE rent research findings suggest that the procedures are
THE DIRECT EFFECTS associated with similar direct and indirect effects on
OF PUNISHMENT responding. Nevertheless, the literature on positive
punishment far exceeds that on negative punishment.
Much of the basic research on the direct Basic processes may differ in important ways under
effects of punishment was conducted more these two forms of punishment.
434 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
es (Iwata, Pace, Cowdery, & Miltenberger, DiLorenzo, 1984, for reviews of this litera-
1994). That is, social consequences that may ture). Applied findings also indicate that the
have maintained responding during baseline effects of punishment are superior to those
(e.g., verbal reprimands, escape from in- obtained with less intrusive procedures
structions) often were removed with the in- alone, such as differential reinforcement
troduction of punishment. A substantial (e.g., Barrett, Matson, Shapiro, & Ollen-
portion of applied findings thus may have dick, 1981; Favell et al., 1982; Scotti, Evans,
little generality to contemporary treatment Meyer, & Walker, 1991). Although results
approaches because punishment is most like- of such comparisons are consistent with
ly to be used when problem behavior con- those obtained in the laboratory, the findings
tinues to produce reinforcement. Basic find- are difficult to interpret because numerous
ings suggest that reinforcement parameters parameters likely influence the effects of
can influence the effects of punishment in these behavior-reduction procedures. For ex-
important ways. ample, a dense schedule of differential rein-
forcement may reduce behavior more effec-
THE DIRECT EFFECTS tively than a mild punisher.
OF PUNISHMENT More important, the relative efficacy of
Basic research findings have shown that treatment with reinforcement versus punish-
response-contingent shock, noise, blasts of ment likely depends on a variety of factors
air, response cost, and time-out can produce (e.g., history; use of extinction; type,
a rapid decrease in the frequency of behavior amount, and schedule of the consequence).
and, in some cases, may lead to complete These complex interactions need to be eval-
response suppression in rats, pigeons, mon- uated to generate more definitive findings
keys, and humans (e.g., college students, about the suppressive effects of punishment
psychiatric patients; Azrin, 1960; Crosbie, relative to other procedures. Further research
Williams, Lattal, Anderson, & Brown, on strategies to improve the efficacy of pun-
1997). Several studies with humans and ishment would be more pragmatic over the
nonhumans also found that the initial re- long run than additional, complex compar-
ductive effects of punishment with shock or ative studies of reinforcement versus punish-
point loss occurred more rapidly, or to a ment.
greater extent, than those produced by ex- Several authors have suggested that treat-
tinction, satiation, and differential reinforce- ment with punishment is so effective be-
ment (e.g., Holz & Azrin, 1963; Johnson, cause punishment usually can compete suc-
McGlynn, & Topping, 1973; Rawson & cessfully with reinforcement contingencies
Leitenberg, 1973). that maintain problem behavior (e.g., Van
The potential benefit of using punish- Houten, 1983). Although punishment often
ment to treat intractable behavior problems was confounded inadvertently with extinc-
led to the development of numerous punish- tion in applied research, recent studies have
ment procedures for clinical use. Results of demonstrated that common punishment
research have shown that treatment with a procedures (e.g., time-out, brief manual re-
wide variety of punishers (e.g., verbal repri- straint) can be effective in the absence of
mands, restraint, water mist, lemon juice, extinction (Fisher, Piazza, Bowman, Hago-
shock, removal of reinforcing activities or pian, & Langdon, 1994; Lerman, Iwata,
conditioned reinforcers) can produce an im- Shore, & DeLeon, 1997; Thompson, Iwata,
mediate, substantial suppression in problem Conners, & Roscoe, 1999). Reductions in
behavior (see Kazdin, 2001, and Matson & behavior were obtained even after unsuc-
PUNISHMENT 435
cessful attempts to treat the behavior with much of the research on these factors has
less intrusive procedures (e.g., Fisher et al., been conducted in the basic laboratory, and
1993; Lindberg, Iwata, & Kahng, 1999). our knowledge of some important complex
Nevertheless, the generality of these find- relations is relatively incomplete (Baron,
ings may be limited because data on effective 1991).
treatments are more likely to be published
than those that show unsuccessful outcomes. History
Potentially important reinforcement param- Basic findings indicate that previous ex-
eters also were unspecified in these studies. posure to certain factors can alter responding
Results of pretreatment functional analyses during punishment, a phenomenon that is
indicated that the behavior was maintained especially relevant to the application of pun-
independent of social consequences but did ishment because clinical populations typical-
not isolate the precise reinforcer. Methods to ly have diverse learning histories. Results of
identify the type of nonsocial reinforcement numerous basic studies have shown that pri-
(often called automatic reinforcement) that is or experience with the punishing stimulus
functionally related to problem behavior either contingently or noncontingently can
have been examined in a number of studies decrease a behavior’s sensitivity to punish-
(e.g., Goh et al., 1995; Kennedy & Souza, ment (e.g., Capaldi, Sheffer, Viveiros, Da-
1995; Patel, Carr, Kim, Robles, & Eastridge, vidson, & Campbell, 1985; Halevy, Feldon,
2000; Piazza, Adelinis, Hanley, Goh, & De- & Weiner, 1987). For example, research
lia, 2000). Although further methodological findings with rats indicate that exposure to
refinements are needed, these strategies may intermittent punishment with shock decreas-
be useful for identifying and manipulating es the efficacy of continuous punishment
various reinforcement parameters (e.g., re- with shock, even when several days or weeks
inforcer schedule and magnitude) while lapse between intermittent and continuous
treating automatically reinforced problem punishment (Banks, 1967; Halevy et al.;
behavior with punishment (e.g., Lerman & Shemer & Feldon, 1984). Deur and Parke
Iwata, 1996b). As described in more detail (1970) replicated this effect with normally
below, results of studies employing these developing children and a loud buzzer as the
types of manipulations with behavior main- punishing stimulus.
tained by either social or nonsocial conse- Although adaptation to the punishing
quences could lead to a greater understand- stimulus may account for these findings (Ca-
ing of punishment processes and improved paldi et al., 1985), a similar relation has
treatments. been obtained with intermittent reinforce-
As discussed in the following sections, a ment. That is, rats and college students with
number of factors directly relevant to the de- a history of intermittent reinforcement also
velopment of an applied technology have showed less response suppression under ei-
been found to influence the direct effects of ther continuous or intermittent punishment
punishment. These factors include historical with shock than participants with a history
variables (e.g., prior experience with the of continuous reinforcement (e.g., Brown &
punishing stimulus or intermittent reinforce- Wagner, 1964; Estes, 1944; Halevy et al.,
ment); the use of conditioned punishers; re- 1987; Vogel-Sprott, 1967). Moreover, Eisen-
inforcement variables (e.g., schedule, avail- berger, Weier, Masterson, and Theis (1989)
ability of alternative sources of reinforce- found that resistance to shock punishment
ment); and punishment variables (e.g., mag- increased for one response (lever pressing) in
nitude, immediacy, schedule). However, rats even after a topographically different be-
436 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
havior (running) was exposed to intermittent 1966). Alternating among several effective
reinforcement. punishment procedures in lieu of using a
In clinical settings, an individual is likely single procedure is another potentially useful
to experience common punishers (e.g., ver- strategy for minimizing exposure to any sin-
bal reprimands, time-out) before these con- gle punisher (e.g., Charlop, Burgio, Iwata,
sequences are specifically arranged to treat a & Ivancic, 1988).
particular inappropriate behavior. Further-
more, exposure to intermittent schedules of Conditioned Stimuli
reinforcement and punishment is typical in Neutral stimuli that are paired with pun-
the natural environment. Consequences of- ishing stimuli eventually may acquire prop-
ten are delivered on intermittent schedules erties of the punishing stimuli. Results of ba-
because it is difficult for parents and teachers sic studies indicate that these conditioned
to reinforce or punish every occurrence of stimuli can function as punishers when de-
behavior. When initial attempts to treat a livered contingent on behavior in the ab-
behavior with intermittent punishment fail, sence of the primary, or unconditioned,
caregivers may switch to a continuous sched- stimulus (e.g., Hake & Azrin, 1965). Con-
ule of punishment in an attempt to improve ditioned punishers may be useful for increas-
the efficacy of the treatment. Basic findings ing both the efficacy and acceptability of
indicate that a history with intermittent punishment in clinical settings. Suppose that
punishment may complicate treatment suc- a relatively nonintrusive but ineffective con-
cess in these cases, such that more intense sequence (e.g., a brief verbal cue) was estab-
punishers will be required to suppress be- lished and maintained as a potent condi-
havior effectively (Halevy et al., 1987; tioned punisher via intermittent pairings
Shemer & Feldon, 1984). However, the rel- with a more restrictive, time-consuming in-
evance of these findings to the types of pun- tervention (e.g., overcorrection, time-out).
ishers that are more commonly used in clin- Application of the conditioned punisher
ical settings is unknown because nearly all would reduce the individual’s exposure to
basic studies in this area evaluated the effects the intrusive intervention and the degree of
of shock, and no applied studies have repli- effort required by caregivers to implement
cated and extended these findings to prob- treatment, factors that might circumvent
lem behavior. problems with program inconsistency, habit-
Nevertheless, basic findings in this area uation to the unconditioned punisher, and
suggest some important guidelines for clin- ethical issues associated with the use of re-
ical research and practice. First, it may be strictive procedures.
beneficial for caregivers to identify novel Various stimuli have been established as
punishers when designing treatments and to conditioned punishers in the basic labora-
avoid using common consequences, such as tory, including tones, lights, and low-voltage
verbal reprimands and time-out, in unsys- shock (e.g., Crowell, 1974; Davidson, 1970;
tematic or unplanned ways. Second, inter- Hake & Azrin, 1965). With a few excep-
mittent schedules of punishment should not tions (e.g., Trenholme & Baron, 1975), the
be implemented prior to continuous sched- unconditioned stimulus was electric shock
ules. Third, if adaptation to the punishing and the subjects were rats or pigeons. Con-
stimulus accounts for the decreased sensitiv- ditioned punishers were established via one
ity of behavior (Capaldi et al., 1985), brief of two primary methods. Under one meth-
hiatus from punishment may be useful, as od, the neutral stimulus was presented be-
described in more detail below (Rachlin, fore the onset of an inescapable stimulus
PUNISHMENT 437
(e.g., shock delivered independent of re- ulus (e.g., intensity or saliency), should be
sponding) and then was either removed with examined in further research. In addition,
the onset of the unconditioned stimulus conditioned stimuli typically were estab-
(e.g., Hake & Azrin, 1965; Mowrer & Sol- lished and maintained independent of re-
omon, 1954) or remained in the environ- sponding, a method that likely would invoke
ment while the unconditioned stimulus was ethical concerns if extended to clinical pop-
delivered periodically (e.g., Orme-Johnson ulations. Although it may be more accept-
& Yarczower, 1974). Results of several stud- able to pair conditioned and unconditioned
ies on this method indicated that more con- stimuli contingent on problem behavior, op-
ditioning occurred if the neutral stimulus portunities to condition the stimulus would
was presented prior to the onset of the un- be severely restricted if the unconditioned
conditioned stimulus rather than simulta- punisher suppressed problem behavior to
neously with or after its onset (e.g., Evans, low levels.
1962; Mowrer & Aiken, 1954). The use of conditioned punishers in treat-
Under the other method, the neutral ing problem behavior has been reported in
stimulus was established as a discriminative surprisingly few applied studies. More im-
stimulus for punishment. That is, the pres- portant, no applied studies have focused ex-
ence of the neutral stimulus was correlated clusively on methods to develop and main-
with delivery of the unconditioned stimulus tain stimuli as conditioned punishers in clin-
contingent on responding (e.g., Davidson, ical settings. Lovaas and Simmons (1969)
1970). Although the discriminative stimulus paired the word ‘‘no’’ with shock contingent
then was shown to suppress responding on severe self-injury with 1 participant. The
when delivered contingent on behavior, re- brief verbal reprimand then was presented
sults of Orme-Johnson and Yarczower for self-injury in the absence of shock during
(1974) indicated that stimuli established as a limited number of sessions, and results
discriminative stimuli were much less effec- suggested that the stimulus had acquired the
tive as conditioned punishers than stimuli suppressive properties of the original punish-
established via the former method. Regard- er. In a more thorough evaluation, Dorsey,
less of the conditioning method, research Iwata, Ong, and McSween (1980) paired the
findings have shown that the effects of con- word ‘‘no’’ with contingent water mist for 2
ditioned punishers on behavior are tempo- participants who engaged in self-injury. Re-
rary unless the conditioned stimulus and the sults showed that contingent presentation of
unconditioned punisher continue to be the verbal stimulus maintained low levels of
paired in some manner (Davidson, 1970; self-injury when water mist was withdrawn
Hake & Azrin, 1965). from the original treatment setting. Further-
A few basic studies have evaluated factors more, the suppressive effects of the verbal
that appear to influence the conditioning reprimand generalized to a setting that had
process, such as the magnitude of the un- not been previously associated with the wa-
conditioned stimulus (Mowrer & Solomon, ter mist procedure, as well as to other ther-
1954) and the duration of the conditioned apists who had never delivered the water
stimulus (Hake & Azrin, 1965). However, mist. Finally, Dixon, Helsel, Rojahn, Cipol-
other clinically relevant parameters of con- lone, and Lubetsky (1989) paired a mild,
ditioning, such as the number of pairings less effective punisher (visual screen) with a
between the conditioned and unconditioned more effective punisher (the odor of am-
stimuli, the type of unconditioned punisher monia) while treating aggression and disrup-
used, and characteristics of the neutral stim- tion exhibited by a young boy with devel-
438 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
opmental disabilities. Problem behavior re- clinicians could determine the maximum
mained suppressed for a short time when the number of times that the conditioned stim-
visual screen was used alone. ulus could be presented before the condi-
Although results of these three studies in- tioning effect begins to be extinguished. The
dicated that conditioned punishers were es- conditioned and unconditioned stimulus
tablished successfully for a clinical problem, then could be paired prior to that number
the efficacy of treatment was evaluated across on a regular basis.
a limited number of brief sessions. Basic
findings have shown that the effects of con- Reinforcement Schedule
ditioned punishers on behavior are tempo- Basic findings indicate that the character-
rary unless the conditioned stimulus and the istics of responding during punishment may
unconditioned punisher continue to be depend on the reinforcement schedule that
paired in some manner (Davidson, 1970; maintains the behavior, a factor that is es-
Hake & Azrin, 1965). In addition, details pecially relevant to application because prob-
necessary to replicate the conditioning pro- lem behavior is likely to be maintained by
cedure (e.g., method of pairing, total num- some form of reinforcement during treat-
ber of pairings, rules for determining when ment. Behavior may be concurrently ex-
to test the conditioned effect) were not de- posed to schedules of reinforcement and
lineated. The generality of these findings punishment when caregivers do not com-
and those obtained in the basic laboratory pletely withhold social consequences during
also may be limited by the use of relatively treatment or when the behavior is main-
intrusive unconditioned punishers (i.e., tained by automatic reinforcement. Results
shock, water mist, ammonia). of basic studies generally showed that the
Thus, current knowledge about condi- amount of response suppression under pun-
tioned punishment is fairly incomplete, and ishment was negatively related to the density
prescriptions for the application of condi- of the reinforcement schedule, with extinc-
tioned punishers should await further re- tion producing the greatest decrease in re-
search. The efficacy of pairing various types sponding (Azrin & Holz, 1966). Various pa-
of auditory, tactile, and visual stimuli should rameters of punishment (e.g., schedule, in-
be evaluated with more common and so- tensity) also appear to interact with the re-
cially acceptable forms of punishment (e.g., lation between response suppression and
time-out). The number of pairings necessary reinforcement density (Bouzas, 1978; Brad-
to produce conditioning and factors that shaw, Szabadi, & Bevan, 1977, 1978). For
might alter the outcome of conditioning example, Bradshaw and his colleagues found
(e.g., intensity of the conditioned stimulus) that the negative relation between reinforce-
could be evaluated by periodically testing the ment density and responding was much
suppressive effects of the paired stimulus in more pronounced when human subjects
the absence of the unconditioned punisher. were exposed to a variable-ratio (VR) pun-
The durability of conditioning could be de- ishment schedule of monetary loss than to a
termined by presenting the conditioned variable-interval (VI) punishment schedule.
stimulus without the unconditioned punish- Intermittent reinforcement schedules ex-
er until the effects on responding dissipate. amined in the laboratory have included
This strategy also may be useful when de- fixed-interval (FI), fixed-ratio (FR), VI, and
veloping a schedule for pairing the condi- VR schedules. Although these reinforcement
tioned and unconditioned stimuli to main- schedules have been found to interact dif-
tain conditioning over time. For example, ferentially with the effects of punishment,
PUNISHMENT 439
this interaction has not been well studied ment may be difficult to modify (but see
and likely depends on various factors, such Lerman & Iwata, 1996b, for one approach),
as the reinforcement density, punishment results of this research may lead to useful
schedule, and amount of reinforcement lost guidelines for designing effective yet practi-
due to a reduction in responding (e.g., Pow- cal treatments when caregivers will be un-
ell, 1970; Scobie & Kaufman, 1969; see also likely or unable to withhold social reinforce-
Baron, 1991, for further discussion). The ment for problem behavior. Current knowl-
various ways in which these schedules influ- edge indicates that reinforcement for prob-
ence punishment effects are relevant to an lem behavior should be withheld or
applied technology because social contingen- diminished if possible. Thus, for example,
cies for problem behavior often approximate when caregivers cannot withhold reinforce-
these laboratory arrangements in the natural ment completely during punishment (e.g.,
environment (e.g., Lalli & Goh, 1993; Voll- ignore all instances of self-injury), the den-
mer, Borrero, Wright, Van Camp, & Lalli, sity and magnitude of reinforcement for
2001). Such complex interactions between problem behavior should be reduced and
reinforcement and punishment schedules punishment should be delivered on a con-
also are likely responsible for some inconsis- tinuous schedule (see further discussion be-
tent findings reported in both the basic and low).
applied literatures on punishment (see fur-
ther discussion below). Additional basic re- Availability of Alternative Reinforcement
search in this area is needed to clarify these Most textbooks and literature reviews on
relations. application highlight the benefits of combin-
Knowledge about basic processes and pre- ing punishment with some type of reinforce-
scriptions for best practices when using pun- ment procedure, such as DRA (e.g., Cooper,
ishment in clinical settings will be incom- Heron, & Heward, 1987; Matson & Di-
plete without further evaluation of potential Lorenzo, 1984). Basic studies with rats, pi-
interactions between reinforcement sched- geons, and psychiatric patients have shown
ules and parameters of punishment. Never- that the suppressive effects of contingent
theless, no applied studies have examined shock, noise, or time-out were enhanced
the effects of reinforcement schedule or den- when reinforcement could be obtained in
sity on the outcome of treatment with pun- some manner other than, or in addition to,
ishment. Further research should determine engaging in the punished response (e.g.,
if reducing the density of the reinforcement Boe, 1964; Herman & Azrin, 1964; Holz,
schedule operating in the natural environ- Azrin, & Ayllon, 1963; Rawson & Leiten-
ment would substantially enhance the effi- berg, 1973). Although these findings suggest
cacy of commonly used punishment proce- that increasing the density of alternative re-
dures. If so, strategies are needed to thin the inforcement might enhance the efficacy of
schedule of reinforcement for problem be- mild punishers in clinical settings, few basic
havior during treatment with punishment. studies have evaluated clinically relevant
The parameters under which reinforcement punishers or the parameters under which re-
schedule is and is not an important factor inforcement may provide optimal benefits
when treating problem behavior with pun- during punishment.
ishment also should be evaluated. Potentially In a study with pigeons reported by Azrin
relevant parameters include the schedule, and Holz (1966), for example, an FR 25
type, and intensity of the punisher. schedule of alternative reinforcement did not
Although nonsocial sources of reinforce- increase sensitivity to punishment when a re-
440 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
sponse maintained by the same reinforce- arranged for both response options also sug-
ment schedule was punished with low-volt- gest that the amount of suppression pro-
age shock (less than 50 V). It is possible, duced by punishment for a given behavior
however, that a denser schedule of alterna- can be influenced by contingencies that op-
tive reinforcement would have enhanced the erate on other behavior, including the rela-
efficacy of this mild punisher. Results of a tive schedule, delay, and magnitude of re-
study by Fantino (1973) indicated that the inforcement and punishment (e.g., Deluty,
beneficial effects of alternative reinforcement 1976, 1978; Farley, 1980).
were compromised when the reinforcement Despite obvious clinical implications,
rate provided by a concurrent VI schedule only one applied study has evaluated the re-
prior to punishment could not be obtained lation between punishment effects and the
via exclusive responding on the unpunished availability of alternative reinforcement.
alternative. Punishment parameters such as Thompson et al. (1999) examined the sep-
schedule and delay and various reinforce- arate and combined effects of punishment
ment parameters also likely modify the ef- and reinforcement on self-injury after results
fects of alternative reinforcement. Thus, ba- of a functional analysis indicated that the
sic findings suggest that various factors (e.g., behavior was maintained by automatic re-
type or intensity of the punisher, density of inforcement. Reinforcement was arranged
available reinforcement prior to punish- for an alternative behavior (toy manipula-
ment) must be considered when combining tion) by giving the participants access to pre-
reinforcement with punishment in clinical ferred toys (thereby establishing automati-
settings. cally reinforced toy play) or by delivering
However, other commonly used reinforce- food contingent on toy manipulation. Re-
ment arrangements, such as noncontingent sults for the 4 participants indicated that al-
reinforcement (NCR) and differential rein- ternative reinforcement enhanced the effi-
forcement of low response rates (DRL), have cacy of relatively mild punishers (e.g., brief
not been evaluated in the context of a con- manual restraint). Furthermore, reinforce-
current punishment contingency in the lab- ment alone was fairly ineffective for all par-
oratory. More important, basic findings on ticipants, and punishment alone was ineffec-
alternative reinforcement may have limited tive for 1 participant. As noted by the au-
generality to application because clinically thors, however, the combined treatment was
relevant factors (e.g., response topography; differentially confounded with an additional
reinforcement quality, schedule, delay, and contingency (i.e., time-out from positive re-
magnitude) usually were held constant inforcement—access to programmed rein-
across available response options. Treatment forcement was withheld during punishment
with punishment and differential reinforce- delivery) that may constrain the generality
ment typically will incorporate different re- of the findings.
sponse, reinforcement, and punishment pa- Further research is needed on clinical
rameters across the targeted behaviors, es- strategies to enhance the efficacy of mild
pecially when problem behavior is main- punishers through the use of DRA, DRO,
tained by unknown or uncontrolled sources DRL, and NCR procedures. Thus far, basic
of reinforcement. In such cases, alternative findings in this area suggest a number of ten-
reinforcement may neither suppress the pun- tative prescriptions for application. Natural-
ished response nor increase adaptive behav- istic reinforcement schedules for targeted
ior. Results of basic studies in which sched- problem and alternative behavior should be
ules of reinforcement and punishment were considered first when developing treatments
PUNISHMENT 441
that combine reinforcement and punish- tion) should be used to treat behavior prob-
ment. The type, schedule, and magnitude of lems, and strategies that increase the effec-
reinforcement maintaining problem behav- tiveness of mild punishers should be incor-
ior should be identified, so that a larger porated into treatment. Basic studies on
amount of the same reinforcer could be pro- magnitude have shown that response sup-
vided independent of undesirable respond- pression is positively related to the intensity
ing or contingent on alternative behavior and duration of shock with rats, pigeons,
(Fantino, 1973). When the functional rein- monkeys, and college students (Church,
forcer cannot be identified or delivered by 1969; Deluty, 1978; Scobie & Kaufman,
others, selecting reinforcers that compete 1969), the duration of time-out with normal
with or substitute for maintaining reinforc- humans (Kaufman & Baron, 1968; N. B.
ers may be critical to effective treatment Miller & Zimmerman, 1966), and the num-
(e.g., Shore, Iwata, DeLeon, Kahng, & ber of points lost as part of response cost
Smith, 1997). with normal humans (Weiner, 1964). In
As many of the available sources of rein- fact, recovery during punishment and fol-
forcement as possible also should be deter- lowing the termination of the punishment
mined prior to treatment, so that steps can contingency was most likely to occur with
be taken to ensure that the total amount of mild punishers, such as a bar slap (Skinner,
obtainable reinforcement can be sustained or 1938) and low-voltage shock (Hake, Azrin,
exceeded despite a reduction in the punished & Oxford, 1967). Furthermore, punishment
behavior. To this end, differential reinforce- with high-intensity shock or forceful air
ment procedures should target simple free- blasts was found to be ineffective with rats
operant responses or adaptive behavior that if the intensity of the punisher was initially
is already in the individual’s repertoire. Pro- low and then gradually increased over time
cedures such as NCR, DRO, and DRL, (Cohen, 1968; N. E. Miller, 1960; Terris &
which do not require an alternative response Barnes, 1969).
for reinforcement delivery, may be preferable On the basis of these findings, numerous
during the initial stages of treatment with authors have recommended using moderate
punishment to insure a sufficient density of or high-intensity punishers to treat problem
available reinforcement. In fact, multiple re- behavior and cautioned against increasing
sponses, reinforcers, and reinforcement con- the intensity of punishment gradually over
tingencies (e.g., DRO plus NCR) should be time (e.g., Cooper et al., 1987; Martin &
incorporated into treatment such that pun- Pear, 1996; O’Brien, 1989). These guide-
ishment for a restricted number of responses lines may be difficult to reconcile with eth-
is implemented within the context of a rich- ical mandates to identify the least restrictive
ly reinforcing environment. procedure that is effective. Moreover, close
examination of basic findings in this area in-
Punishment Magnitude dicates that the relation between responding
The relation between the effects of pun- and punishment magnitude is more complex
ishment and the magnitude or amount of than frequently assumed. As a result, strat-
punishment delivered for responding is es- egies that are based on recommendations de-
pecially germane to the efficacy and accept- lineated in applied textbooks (e.g., Cooper
ability of punishment in clinical settings. et al.; Martin & Pear) and literature reviews
From an ethical and practical standpoint, may not influence behavior as expected.
the least amount of punishment that is ef- As previously noted, the basic relation be-
fective (i.e., lowest intensity, shortest dura- tween response suppression and punishment
442 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
magnitude may be influenced by the avail- son, & Manning, 1981). For example, Cole
ability of alternative reinforcement (e.g., et al. found that treatment with overcorrec-
Holz et al., 1963) and the type of reinforce- tion produced similar decreases in stereotyp-
ment schedule that maintains behavior (e.g., ic behavior, regardless of whether the inter-
Powell, 1970; Scobie & Kaufman, 1969). vention lasted 30 s, 2 min, or 8 min. The
Other variables (e.g., immediacy; Cohen, effects of overcorrection, however, may have
1968) also may alter the relation between been confounded with those of extinction
punishment magnitude and responding. and verbal reprimands. Results of studies on
Such complex interactions may be respon- the duration of time-out have shown a pos-
sible for some contradictory findings on itive relation (e.g., Burchard & Barrera,
punishment magnitude that have been re- 1972; Hobbs, Forehand, & Murray, 1978),
ported in the applied literature. The extent a negative relation (e.g., Kendall, Nay, & Jef-
to which basic findings are directly compa- fers, 1975), and no relation (e.g., White,
rable to applied findings also is limited be- Nielsen, & Johnson, 1972) between dura-
cause most basic studies examined the mag- tion length and treatment effects. These
nitude of shock, whereas applied studies in findings are difficult to interpret because the
this area have focused on more clinically ac- function of problem behavior was not iden-
ceptable forms of punishment (e.g., overcor- tified (thus, time-out may have been contra-
rection, time-out). indicated for some participants), and se-
In one of the few applied studies to eval- quence effects may have confounded the re-
uate the magnitude of electric shock, D. E. sults (Matson & DiLorenzo, 1984).
Williams, Kirkpatrick-Sanchez, and Iwata Further research on the relation between
(1993) compared the efficacy of treatment punishment magnitude and response to
for self-injury under two shock-intensity lev- treatment, as well as on factors that can alter
els (3.5 mA vs. 18.5 mA). Results were con- this relation (e.g., reinforcement schedule,
sistent with those obtained in basic studies. punishment delay), may be useful for rec-
The high-intensity shock produced larger, onciling inconsistent findings in the litera-
more immediate decreases in behavior than ture and for developing more comprehensive
the low-intensity shock. The generality of prescriptions for application. The common
this finding, however, is somewhat limited assumption that a larger magnitude of a giv-
because punishment was combined with ex- en punisher will be more effective than a
tinction. In addition, sequence effects could smaller magnitude is not strongly supported
have influenced the outcome because the in the current literature, with the exception
participant was exposed to the lower inten- of findings on contingent shock. Magnitude
sity shock prior to the higher intensity, and should be manipulated with a variety of
a reversal to the low-intensity condition was punishers and in a variety of ways that have
not implemented. not been examined in basic research. For ex-
Results of studies examining the relation ample, the amount of reinforcement avail-
between punishment magnitude and treat- able during ‘‘time-in’’ is another potentially
ment efficacy using other types of punishers important dimension of magnitude when
(e.g., physical restraint, unpleasant smells, treatment with time-out is implemented
time-out) have been inconsistent and often (e.g., Solnick, Rincover, & Peterson, 1977).
appeared to be confounded with other var- Strategies to enhance response suppression
iables (e.g., Altman, Haavik, & Cook, 1978; and maintenance under less effective values
Cole, Montgomery, Wilson, & Milan, 2000; of punishment magnitude also should be ex-
Marholin & Townsend, 1978; Singh, Daw- plored. Basic findings indicate that smaller
PUNISHMENT 443
magnitudes of punishment may be more ef- 180 s. Results of subsequent treatment anal-
fective if the punisher is delivered immedi- yses indicated that the procedure associated
ately following the behavior (e.g., Cohen, with the highest levels of negative vocaliza-
1968) and if reinforcement is available for tions, avoidance, or escape responses was the
an alternative response (e.g., Holz et al., most effective punisher for problem behav-
1963). Basic studies with shock also indicate ior. However, results of the initial assessment
that a less intense punisher may be effective, did not differentiate among the various du-
at least temporarily, if a high intensity level ration lengths for any participant, possibly
is decreased gradually over time (e.g., Azrin, because the procedures were alternated rap-
1960; Cohen, 1968; Hake et al., 1967). An idly across a limited number of trials. The
approach that involves periodically inter- utility of such assessments should be evalu-
spersing less intense punishers with more in- ated in further studies. For example, a sep-
tense punishers may be useful for maintain- arate assessment of punishment magnitude,
ing treatment effects while the magnitude of similar in design to that conducted by Fisher
punishment is gradually reduced. As dis- et al., might be useful after an initial assess-
cussed above, further research on condi- ment has identified a potent punisher.
tioned punishers and treatments combining Until further applied research on magni-
reinforcement with punishment also may tude is conducted, practitioners should select
lead to methods for increasing the effective- magnitudes that have been shown to be safe
ness of mild punishers. and effective in clinical studies, as long as
Finally, further studies should evaluate the magnitude is considered acceptable and
methods for identifying the most appropri- practical by those who will be implementing
ate magnitude of a given punishment pro- treatment. Punishment should be combined
cedure prior to treatment implementation in with some type of reinforcement procedure,
the natural environment. The typical trial- and the punisher should be delivered as im-
and-error approach to punishment selection mediately as possible following occurrences
is inefficient and may be counterproductive of problem behavior (see further discussion
if an individual receives prolonged exposure below). If the punisher fails to suppress be-
to ineffective procedures (e.g., N. E. Miller, havior over time, alternative procedures
1960; Terris & Barnes, 1969). Efficient probably should be considered instead of in-
strategies for identifying the least restrictive, creasing the magnitude of the punisher un-
effective treatment are surprisingly absent der the presumption that this strategy will
from the applied literature. In two studies improve the efficacy of treatment.
conducted by Fisher and colleagues (Fisher,
Piazza, Bowman, Hagopian, & Langdon, Immediacy of the Punisher
1994; Fisher, Piazza, Bowman, Kurtz, et al., Consequences for problem behavior are
1994), the potential suppressive effects of frequently delayed in the natural environ-
various procedures (e.g., time-out, facial ment. Caregivers and teachers often are un-
screen, contingent demands) were rapidly as- able to monitor behavior closely or to deliver
sessed by exposing participants to the puta- lengthy punishers (e.g., 15-min contingent
tive punishers while negative vocalizations work) immediately following instances of
(e.g., yelling, crying) and avoidance or es- problem behavior (Azrin & Powers, 1975).
cape responses (e.g., dropping to the floor) Punishment also may be delayed when the
were measured. Each punishment procedure individual actively resists application of the
was delivered noncontingently across five programmed consequences by struggling
different durations, ranging from 15 s to with the punishing agent or running away.
444 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
In some cases, problem behavior occurs pri- Surprisingly few applied studies have eval-
marily in the absence of the punishing agent, uated the efficacy of delayed punishment or
necessarily delaying programmed conse- strategies to improve treatment effects when
quences until the behavior is detected consequences do not occur contiguous to
(Grace, Thompson, & Fisher, 1996; Van the behavior. In one of the few studies to
Houten & Rolider, 1988). compare immediate and delayed punish-
For these reasons, research on delayed ment, Abramowitz and O’Leary (1990)
punishment is especially pertinent. Labora- found that immediate verbal reprimands
tory findings with rats indicate that the were much more effective in decreasing off-
lengthier the delay between the occurrence task behavior in school children than were
of the response and delivery of contingent reprimands that were delayed by 2 min.
shock, the smaller the amount of response These results are somewhat difficult to in-
suppression under punishment (e.g., Baron, terpret, however, because delayed repri-
Kaufman, & Fazzini, 1969; Camp, Ray- mands were delivered only if off-task behav-
mond, & Church, 1967). Even brief delays ior had occurred continuously for 2 min,
of 10 s or 20 s have been found to seriously during which time the students had varied
compromise the effects of contingent shock opportunities to interact with other students
with rats and college students (e.g., Banks & and non-task-related objects. Thus, the ef-
Vogel-Sprott, 1965; Goodall, 1984) and of fects of punishment delay were not separated
reinforcement loss with college students from those of reinforcement schedule and
(Trenholme & Baron, 1975). punishment schedule.
Stimuli that might bridge the interval be- Results of just two studies have delineated
tween a response and its consequence have conditions under which delayed punishment
been notably absent from laboratory ar- may produce effective outcomes. Rolider
rangements involving delayed punishment. and Van Houten (1985) and Van Houten
Results of at least one study suggest that fac- and Rolider (1988) demonstrated the effi-
tors such as the presence of a conditioned cacy of delayed punishment using various
punisher and delivery of instructions can al- mediated consequences with children with
ter the efficacy of delayed punishment. In emotional and developmental disabilities.
Trenholme and Baron (1975), delays of 10 One form of mediation involved playing au-
s, 20 s, and 40 s were equally effective with diotape recordings of the child’s disruptive
college students when a brief noise that was behavior that were collected earlier in the
paired with reinforcement loss also occurred day. The punishing consequence (physical
immediately following the behavior. A sub- restraint, verbal reprimands) then was deliv-
sequent experiment showed that delayed ered. In some cases, the tape recorder was
punishment was just as effective as imme- clearly visible to the child while the record-
diate punishment when the participants re- ings were being collected, and a verbal ex-
ceived instructions about the delay. The gen- planation of its role in the delivery of de-
erality of these findings to clinical popula- layed punishment was provided. These fac-
tions, such as individuals with developmen- tors may have served to bridge the temporal
tal disabilities, has not been determined. In gap between inappropriate behavior and its
addition, no basic studies have evaluated the consequence (e.g., by functioning as dis-
effects of numerous other potentially impor- criminative stimuli for punishment; Tren-
tant factors on delayed punishment (e.g., holme & Baron, 1975). However, for 1 par-
history, reinforcement schedule, availability ticipant, neither instructions about delayed
of alternative reinforcement). punishment nor an immediate consequence
PUNISHMENT 445
designed to signal that punishment was (e.g., Linscheid, Iwata, Ricketts, Williams,
forthcoming (a mark placed on the child’s & Griffin, 1990).
hand) was as effective as the audiotape pro- However, the timing of punishment in re-
cedure. lation to reinforcement delivery also should
In Van Houten and Rolider (1988), care- be considered, because some basic studies
givers physically guided 2 participants to en- have found that immediate punishment was
gage in the problem behavior (aggression or less effective than delayed punishment if the
theft) after occurrences of the behavior were immediate punisher preceded reinforcement
detected or reported. Caregivers then deliv- delivery but the delayed punisher followed
ered the punishing consequence (physical re- it (e.g., Epstein, 1984; Rodriguez & Logan,
straint) immediately following the guided re- 1980). It is conceivable that diligent caregiv-
sponse. Although treatment was effective, ers may respond to problem behavior by first
the efficacy of delayed punishment without delivering the prescribed punisher (e.g., con-
the guided response component was not ex- tingent work, time-out), followed (inadver-
amined. The length of the delay and care- tently) by the maintaining social reinforcer
givers’ immediate response to the problem (e.g., access to materials). In a similar man-
behavior also were not specified. ner, automated punishment may be deliv-
Current knowledge indicates that the ered immediately prior to social or nonsocial
mild punishers typically used in clinical set- consequences for problem behavior. Results
tings will be ineffective unless the conse- of other basic studies, in which the avail-
ability of reinforcement for one response was
quence immediately follows problem behav-
perfectly correlated with the delivery of mild
ior. Thus, further research is needed on fac-
punishment for an immediately preceding
tors that might enhance treatment effects
response, indicated that pairing punishment
under delayed punishment, especially pro-
and reinforcement in this manner estab-
cedures or stimuli that would bridge the
lished the mild punisher as a conditioned
temporal gap between a response and its positive reinforcer (e.g., Murray & Nevin,
consequence. Research on the utility of de- 1967; D. R. Williams & Barry, 1966). Thus,
livering conditioned punishers, instructions, the timing of punishment and reinforcement
and other types of stimuli associated with in the natural environment should be eval-
delayed consequences is needed with clinical uated carefully as part of treatment.
populations.
Until further applied research is conduct- Schedule of Punishment
ed, teachers and caregivers should be con- The effects of punishment schedules on
cerned with selecting punishers that can be responding have important implications for
readily delivered as soon as the behavior oc- the efficacy and acceptability of treatment
curs. Consequences that do not require the with punishment. Intermittently delivered
close proximity of the caregiver (i.e., stimuli consequences that successfully reduce prob-
that can be delivered or removed from a dis- lem behavior are easier to use, less time con-
tance) and technology to increase the prac- suming, and less intrusive than consequences
ticality of immediate punishment may be es- that must follow each occurrence of behav-
pecially useful in this regard. Electronic de- ior. Results of basic research with pigeons
vices that detect occurrences of problem be- and rats suggest that punishment with shock
havior and either alert caregivers or deliver or time-out will not produce acceptable re-
consequences automatically might circum- sults unless the punisher follows nearly every
vent the problems of delayed punishment occurrence of the behavior in situations in
446 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
which no alternative is available or when the FI; Azrin, 1956; Camp, Raymond, &
density of reinforcement is not reduced (Ap- Church, 1966), although the nature of this
pel, 1968; Azrin, Holz, & Hake, 1963; De- relation is complex (e.g., Arbuckle & Lattal,
luty, 1976; Farley, 1980; Thomas, 1968). 1992) and appears to be influenced by the
Although a number of applied studies schedule of reinforcement that maintains the
have examined the efficacy of intermittent behavior (e.g., Bradshaw et al., 1977, 1978;
punishment for treating problem behavior, Powell, 1970; Scobie & Kaufman, 1969).
results have been inconsistent, and the con- In the only applied study that examined
ditions under which intermittent punish- the interaction between type or intensity of
ment might be effective remain unclear. In punishment and intermittent punishment
some studies, for example, intermittent pun- schedules, Cipani, Brendlinger, McDowell,
ishment schedules were associated with a so- and Usher (1991) found that a VR 4 sched-
cially significant reduction in behavior, par- ule of punishment with contingent applica-
ticularly if responding was already sup- tion of lemon juice was just as effective as a
pressed to low levels via continuous punish- continuous schedule in reducing a child’s
ment (e.g., Clark, Rowbury, Baer, & Baer, stereotypic behavior. A ‘‘manual guidance
1973; Rollings & Baumeister, 1981; Ro- overcorrection’’ procedure (i.e., physically
manczyk, 1977). Conversely, even dense in- guiding the child’s arms over the head and
termittent punishment schedules were inef- to the sides 10 times) also was effective when
fective for some individuals in other studies the procedure was delivered under a contin-
(e.g., Calhoun & Matherne, 1975; Lerman uous schedule. Unlike the lemon juice, how-
et al., 1997). Basic findings on factors that ever, a VR 4 schedule with the overcorrec-
interact with the effects of punishment tion procedure did not produce clinically
schedules (e.g., reinforcement schedule) may significant reductions in behavior.
explain why clinical applications have pro- Finally, interactions between schedules of
duced inconsistent results. reinforcement and punishment may account
First, punishment appeared to be con- for the idiosyncratic effects of intermittent
founded with extinction and other potential punishment on problem behavior. For ex-
punishers (e.g., verbal reprimands) in ap- ample, Lerman et al. (1997) treated 5 par-
plied studies that showed significant treat- ticipants’ self-injurious behavior (SIB) with
ment effects under thin punishment sched- a continuous schedule of punishment after
ules (e.g., Barton, Brulle, & Repp, 1987; results of a functional analysis indicated that
Clark et al., 1973; Romanczyk, 1977). Sec- the behavior was maintained by automatic
ond, important parameters of punishment reinforcement. Initial application of inter-
(i.e., type, intensity, and schedule) varied mittent punishment (FI 2 min or FI 5 min)
considerably among these studies. Basic was ineffective for 4 of the 5 participants.
findings indicate that these variables alter the The continuous punishment schedule then
relation between intermittent punishment was successfully thinned to FI 5 min for 2
and response suppression. Increasing the in- of these participants. In contrast, continuous
tensity of a punisher, for example, can either punishment was necessary to suppress SIB
enhance or degrade the efficacy of intermit- for the other 2 participants, despite repeated
tent schedules, depending on other factors attempts to thin the schedule. Although the
(e.g., Appel, 1968; Lande, 1981). Certain function of SIB had been identified prior to
punishment schedules (e.g., VI) also have treatment, important parameters of the
been associated with greater decreases in re- maintaining reinforcers were unknown (e.g.,
sponding than other schedules (e.g., FR or schedule, density, magnitude). These param-
PUNISHMENT 447
eters, which likely varied across participants, that is considered problematic only because
may have been responsible for the inconsis- it occurs at high rates or in bursts. Further
tent success of the schedule-thinning proce- understanding of these schedules also is im-
dure. portant because they may commonly operate
Further research on interactions between in the natural environment. For example,
punishment schedule (e.g., VR vs. VI) and caregivers may be more likely to deliver pun-
other potentially important parameters of ishment when problem behavior occurs in-
punishment and reinforcement is needed to frequently (i.e., is characterized by long in-
clarify the conditions under which intermit- terresponse times) than when behavior oc-
tent punishment would and would not be curs at high rates (see Arbuckle & Lattal,
effective. Few studies have directly evaluated 1992, for a discussion of this issue). Such an
strategies to systematically thin punishment arrangement could compromise the efficacy
schedules or to utilize highly variable (and of treatment by increasing the frequency of
thus unpredictable) schedules. Combining a short interresponse times.
thin schedule of punishment with a rich Current knowledge about punishment
schedule of conditioned punishment is an- schedules suggests that parents and teachers
other potential approach for increasing the should punish each occurrence of problem
efficacy of intermittent punishment. De- behavior unless the behavior is simulta-
pending on the nature of the conditioned neously exposed to extinction. Until further
and unconditioned punishers, this arrange- research is conducted, clinicians should be
ment may be more practical than using a extremely cautious when attempting to thin
rich schedule of unconditioned punishment the punishment schedule, utilize DRH
alone. A dense schedule of alternative rein- schedules, or evaluate other strategies for im-
forcement also may promote the efficacy of proving the effects of intermittent punish-
intermittent punishment. ment (e.g., employing variable schedules or
Other types of punishment schedules ex- conditioned punishers). A continuous sched-
amined in the basic laboratory, such as the ule of punishment always should be imple-
differential punishment of high (DPH) or
mented initially, and intermittent schedules
low (DPL) response rates, also may be useful
should be considered only if the continuous
in clinical settings. These schedules do not
schedule remains effective in suppressing
specify a direct contingency between the de-
problem behavior to low levels over a con-
livery of the punisher and the occurrence of
siderable amount of time.
a response. For example, under DPH or
DPL, punishment is delivered contingent on
the pause length that immediately preceded FACTORS RELATED TO
a response (i.e., the selective punishment of MAINTENANCE, GENERALIZATION,
certain lengths of interresponse times). Re- AND INDIRECT EFFECTS
sults of basic studies on DPH and DPL
schedules showed that overall responding in- A much smaller proportion of basic and
creased when relatively long interresponse applied studies on punishment have evalu-
times (DPL) were punished and decreased ated the long-term maintenance, generaliza-
when short interresponse times (DPH) were tion, and side effects of punishment relative
punished (e.g., Galbicka & Branch, 1981; to those on direct effects. The extent to
Laurence, Hineline, & Bersh, 1994). DPH which punishment effects are maintained
schedules may be more beneficial than con- over time, transfer across settings and con-
tinuous punishment when treating behavior texts, and produce changes in other behavior
448 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
has significant implications for treatments to the numerous months (and sometimes
involving punishment. years) over which problem behavior requires
treatment. Moreover, few studies have ex-
Maintenance amined factors that may influence the du-
The durability of treatment with punish- rability of punishment effects.
ment is one of the most important consider- Although brief treatment evaluations are
ations for practitioners, teachers, and caregiv- predominant in the applied literature on
ers of individuals with behavior disorders. A punishment, an increasing number of stud-
number of authors, however, have suggested ies have examined the long-term efficacy of
that the clinical effects of punishment are rel- punishment over the past 10 years. Treat-
atively short-lived, even when the treatment ment effects have been examined for 1 to 60
remains unchanged over time (e.g., Parsons, months after punishment was initiated and
Hinson, & Sardo-Brown, 2001; Walker & continued with minor changes to the pro-
Shea, 1999). In basic studies with both hu- cedure (Duker & Seys, 1996; Ricketts,
mans and nonhumans, various punishers have Goza, & Matese, 1993; D. E. Williams,
been associated with continued response sup- Kirkpatrick-Sanchez, & Crocker, 1994), and
pression under punishment, including relative- after the original punishment component
ly intense levels of electric shock with pigeons was withdrawn (Arntzen & Werner, 1999;
and rats (Azrin, 1960; Crosbie et al., 1997), Foxx, Bittle, & Faw, 1989; Rolider, Wil-
point or monetary loss with normal humans liams, Cummings, & Van Houten, 1991).
(Crosbie et al.; Weiner, 1962), and time-out Results have shown varying success in main-
from positive reinforcement with squirrel taining the reduction in behavior, yet poten-
monkeys (McMillan, 1967). Response recov- tial reasons for the inconsistent outcomes
ery, however, has been associated with less in- have not yet been identified.
tense punishers, such as low-voltage shock For example, D. E. Williams et al. (1993)
with pigeons (Rachlin, 1966), a bar slap with observed a relapse in treatment with contin-
rats (Skinner, 1938), and noise with pigeons gent electric shock 6 months after punish-
(Holz & Azrin, 1962). Basic findings on the ment was initiated. Conversely, Linscheid,
maintenance of response suppression follow- Hartel, and Cooley (1993) found that con-
ing the termination of the punishment con- tingent electric shock continued to suppress
tingency also generally showed that response 2 individuals’ self-injurious behavior for 5
rates immediately returned to prepunishment years. Duker and Seys (1996) examined the
levels—sometimes even temporarily exceeding long-term efficacy of contingent shock with
baseline—unless intense punishers were used 12 individuals by obtaining information on
(e.g., high-voltage shock; Azrin, 1960). the degree of physical restraint each required
These results suggest that sufficiently in- from 2 to 47 months after the initiation of
tense punishers, including some commonly punishment. Results at follow-up suggested
used clinical procedures (e.g., time-out), that treatment remained effective for 7 par-
may produce lasting reductions in problem ticipants, including 1 individual who was
behavior as long as the punishment contin- evaluated at 36 months and another who
gency remains in effect. Nevertheless, basic was evaluated at 47 months.
findings may not be applicable to treatment Conclusions about applied findings on
outcomes in clinical settings because the maintenance are difficult to draw for a num-
time periods evaluated in the laboratory ber of reasons. First, the majority of studies
(e.g., 30-min to 60-min punishment sessions examined the long-term effectiveness of con-
across 10 to 20 days) may have little relation tingent electric shock, so results may not be
PUNISHMENT 449
applicable to other (or more mild) punishers clinical settings. One strategy that may de-
(Azrin & Holz, 1966). Second, the reinforc- crease the likelihood of habituation is the use
ing consequences of problem behavior were of hiatus from punishment. In several basic
not identified prior to treatment in most cas- studies with pigeons, response suppression
es. Long-term maintenance may have been under shock punishment was enhanced fol-
more likely to occur if the maintaining re- lowing brief time periods during which the
inforcer was withheld contingent on prob- subject was removed from the punishing sit-
lem behavior or readily available for engag- uation or exposed to reinforcement only
ing in more appropriate behavior (Estes, (e.g., Rachlin, 1966). Further research is
1944). Third, other factors potentially re- needed, however, because the beneficial ef-
sponsible for both successful and unsuccess- fect of this procedure was found to wane
ful cases of treatment maintenance may have across repeated punishment–hiatus cycles
varied widely across the studies (e.g., pun- (e.g., Orme-Johnson, 1967). Other strate-
ishment schedule, availability of reinforcers gies that may prevent or attenuate habitua-
that competed with or substituted for the tion, such as using intermittent, varied, or
maintaining reinforcer). In fact, components brief punishers (e.g., Charlop et al., 1988),
of the original intervention were modified should be evaluated in further studies.
over time in some studies (e.g., additional Research also is needed on strategies to
behavioral procedures or drugs were intro- maintain punishment effects while the inter-
duced; Duker & Seys, 1996), and it is dif- vention is systematically faded. Basic find-
ficult to determine which, if any, treatment ings with pigeons and monkeys have shown
modifications may have been responsible for that responding will remain suppressed un-
the outcomes. Furthermore, the lengthy der low-intensity shock if an initially intense
time period required to conduct these stud- shock is reduced very gradually (e.g., Hake
ies increased the likelihood that unplanned et al., 1967). Further applied research is
changes or other uncontrolled factors inter- needed to determine if treatment effects will
acted with the efficacy of the original treat- be maintained while the intensity or dura-
ment in either desirable or undesirable ways. tion of a punishment procedure is altered
Finally, the number of treatment relapse cas- very gradually or less intrusive procedures
es reported in the literature may not accu- are simultaneously introduced. For example,
rately reflect the prevalence of this problem it may be possible to reduce a 5-min time-
in applied settings because such cases are less out to a 1-min time-out over time. The use
likely to be submitted or accepted for pub- of conditioned punishers may enhance the
lication than successful cases of treatment likelihood of fading certain dimensions of
maintenance. intrusive punishers while treatment effects
Identifying factors or processes associated are maintained over the long run. Moreover,
with long-term maintenance is key to the basic studies have found that response re-
design of a systematic technology for pre- covery is more gradual following the with-
venting and remediating treatment relapse. drawal of intermittent shock punishment
Several authors have suggested that adapta- than following the removal of other punish-
tion, or habituation, to the punishing stim- ment schedules (e.g., Azrin et al., 1963;
ulus accounts for instances of recovery (i.e., Camp et al., 1966). Thus, strategies to in-
repeated exposure decreases the aversiveness crease the utility of conditioned punishers
of the punisher; Goodall, 1984). Moreover, and intermittent punishment for routine
adaptation is more likely to occur with mild clinical practice also may promote the long-
punishers, which are typically employed in term efficacy of punishment.
450 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
Several authors have suggested that com- consequences of problem behavior also
bining punishment with differential rein- should be minimized or withheld if possible.
forcement may increase the likelihood that Potential problems with habituation may
punishment can be faded successfully (e.g., be curtailed by limiting exposure to the pun-
Kazdin, 2001). Although this clinical strategy isher in various ways. For example, caregivers
has not been evaluated directly, one study could schedule brief vacations from punish-
found that differential reinforcement was ment on a regular basis (Rachlin, 1966) or
more effective in reducing problem behavior restrict the use of specific procedures to one
after a participant had been exposed to a pe- or two problem behaviors (e.g., those of
riod of punishment (contingent work) than greatest concern) instead of applying the
when differential reinforcement preceded same treatment for a variety of responses.
punishment (Fisher et al., 1993). Research Comprehensive punisher assessments also
findings on the indirect effects of punishment should be employed to identify clinically ac-
suggest that punishment may increase res- ceptable procedures that produce the great-
ponsivity to reinforcement (see below for fur- est reduction in behavior and, hence, would
ther discussion). Thus, punishment may en- lead to the least amount of exposure to the
hance the efficacy of reinforcement for estab- punisher (see Fisher, Piazza, Bowman, Ha-
lishing appropriate behavior that competes gopian, & Langdon, 1994). If the assess-
with or replaces inappropriate behavior, an ment identifies more than one effective form
outcome that in turn may increase the like- of punishment, caregivers could alternate
lihood that punishment can be withdrawn. among several procedures to minimize ex-
Until additional research on long-term posure to any single punisher.
maintenance is conducted, practitioners and Finally, practitioners and caregivers should
caregivers should not assume that punish- have a plan for dealing with treatment re-
ment will remain effective over the long run. lapse when it occurs during punishment or
Strategies for increasing the likelihood of following the removal of punishment. The
maintenance should be employed from the first step is to identify and rectify other fac-
outset of treatment. Although basic findings tors that may be responsible for treatment
suggest that relatively intense punishers may failure. Many instances of relapse likely are
be associated with successful long-term out- attributable to problems with treatment in-
comes, the use of analogous procedures to tegrity (D. E. Williams et al., 1993), espe-
treat problem behavior probably would raise cially when the punishment procedure is
ethical concerns for all but the most serious complex or time consuming (e.g., Foxx &
cases. Caregivers instead should focus on the Livesay, 1984). The next step is to reassess a
use of reinforcement to insure that alterna- wide range of stimuli and activities that may
tive behavior is at high strength in the rep- function as potent reinforcers for appropri-
ertoire of individuals who are exposed to ate behavior and to target additional, mul-
punishment. Systematic reinforcer assess- tiple responses that might compete with the
ments and functional analyses of problem punished behavior. Alternative forms of
behavior always should precede treatment punishment should be considered only after
implementation (Fisher et al., 1992). The ef- determining that the current punisher is in-
fects of punishment may last longer if ap- effective within the context of a richly rein-
propriate behavior is maintained by the same forcing environment. At this point, another
reinforcers that maintain problem behavior comprehensive punisher assessment should
or by reinforcers that are effective substitutes be conducted to identify other effective pun-
for maintaining reinforcers. The reinforcing ishers. Increasing the intensity or magnitude
PUNISHMENT 451
Thus, it is not surprising that punishment factors that are useful for promoting gener-
effects typically failed to generalize in the ab- alization under reinforcement, such as de-
sence of procedures designed to promote layed or intermittent contingencies, have
transfer (Stokes & Baer, 1977). Several au- been found to undermine the efficacy of
thors have suggested that techniques found punishment (Azrin et al., 1963; Goodall,
to enhance generalization under reinforce- 1984; Trenholme & Baron, 1975). The sup-
ment may be similarly effective under pun- pressive effects of naturalistic consequences
ishment (e.g., Matson & DiLorenzo, 1984; and component derivatives of complex in-
Miltenberger, 2001). A few applied studies terventions likely depend on the process of
on punishment have evaluated generalization conditioned punishment, an area that re-
strategies analogous to those used to promote quires further study. Finally, the develop-
reinforcement effects. For example, common ment of stimulus control under punishment
stimuli were introduced into treatment and has been evaluated in few basic or applied
generalization settings (e.g., a discriminative studies (see Rollings & Baumeister, 1981,
stimulus for punishment was presented in the for a notable exception). Successful general-
generalization setting; Birnbrauer, 1968), ization may hinge on the presence of stimuli
stimuli that might acquire discriminative that have acquired tight control over re-
control over the behavior were removed from sponding, such that few responses occur in
the treatment setting (e.g., the therapist was the absence of the punishment contingency.
hidden from view; Corte et al., 1971; Tate Relative to reinforced responding, it may be
& Baroff, 1966), and training was conducted difficult to establish control over punished
with multiple stimulus exemplars (e.g., sev- responding with stimuli that are not per-
eral different therapists delivered shock; Lo- fectly correlated with the delivery of punish-
vaas & Simmons, 1969). In nearly all cases, ment. Thus, further research in the areas of
however, these strategies were ineffective un- conditioned punishment, stimulus control,
less punishment was delivered in the gener- and intermittent or delayed punishment ap-
alization context. pears to be critical for developing a technol-
Other potential tactics drawn from the lit- ogy of generalization.
erature on reinforcement include pairing the The current literature indicates that pun-
punisher with naturalistic consequences ishment must be delivered consistently in all
(e.g., verbal reprimands), varying the stim- relevant contexts. Nevertheless, various gen-
ulus conditions during initial treatment with eralization strategies described by Stokes and
punishment, providing instruction on self- Baer (1977) may be useful for promoting
management, and using delayed or intermit- treatment generality when the procedure is
tent punishment (i.e., indiscriminable con- extended beyond the initial treatment set-
tingencies; Stokes & Baer, 1977; see also ting. For example, a variety of stimulus con-
O’Donnell & Crosbie, 1998, Experiments 3 ditions could be arranged in the initial treat-
and 4). Generalization also may be achieved ment setting (e.g., different caregivers and
by implementing a modified form of the peers could be present, diverse activities
treatment in generalization contexts (e.g., could be scheduled, physical features of the
delivering a smaller amount of the punisher environment could vary). Stimuli common
or a single component of a multicomponent to other settings and contexts in which pun-
treatment procedure). ishment will be applied could be introduced
Current knowledge about punishment, in the initial treatment setting before the in-
however, is insufficient to guide the appli- tervention is widely implemented. Treat-
cation of such strategies. For example, many ment generality also may be enhanced by en-
PUNISHMENT 453
suring that reinforcement is implemented cause the contingency itself provides an es-
consistently across settings, incorporating cape response (i.e., delivery of the punisher
certain aspects of self-management into can be avoided by refraining from the pun-
treatment (e.g., self-monitoring), and estab- ished behavior). In fact, results of several
lishing salient discriminative stimuli for pun- studies with rats showed that emotional re-
ishment in all settings and contexts (see sponses in the form of crouching and defe-
Stokes & Baer for further discussion of gen- cation were more pronounced and persistent
eralization procedures). when subjects were exposed to unavoidable
shock than to response-contingent stimula-
Indirect Effects of Punishment tion (Hearst, 1965; Hunt & Brady, 1955).
The effects of punishment on responses On the other hand, numerous basic stud-
that can occur concurrently with the pun- ies indicate that other forms of unpunished
ished behavior or in a different context as the behavior, including responses that occur in
punished behavior also have been studied in the absence of programmed consequences
basic and applied research. Among these side (e.g., species-specific behavior) and those
effects, collateral increases in aggression, es- that are maintained by experimenter-deliv-
cape behavior, and emotional reactions are ered reinforcement, may increase, decrease,
most commonly described in basic textbooks or remain unchanged during punishment.
and literature reviews (e.g., Azrin & Holz, Factors that determine whether unpunished
1966; Mazur, 1998) and by authors who rec- behavior will increase or decrease (called con-
ommend against using punishment in clinical trast and induction, respectively) have not
settings (e.g., LaVigna & Donnellan, 1986; been thoroughly studied. The function of
McGee, Menolascino, Hobbs, & Menousek, the behavior, schedule and intensity of the
1987; Parsons et al., 2001). punisher, and prior exposure to the punisher
Aggression (i.e., attacking nearby subjects, may be important (see Crosbie et al., 1997,
biting inanimate objects) in rats, pigeons, for a discussion).
and monkeys has been associated with non- In a series of studies, Dunham and col-
contingent delivery of unavoidable stimuli, leagues examined the effects of punishment
including shock and intense heat (e.g., on multiple responses in gerbils by deliver-
Hutchinson, 1977; Ulrich & Azrin, 1962). ing shock contingent on one response (e.g.,
Although this phenomenon is often called eating) while changes in alternative responses
punishment-elicited aggression, few studies (e.g., digging, grooming, running) were
have examined this side effect of punish- measured. Results indicated that the most
ment. Basic findings on the effects of ines- probable of the unpunished responses in-
capable, intense punishers probably have creased during punishment, whereas re-
limited generality to the application of pun- sponses that tended to follow the punished
ishment (see Linscheid & Meinhold, 1990, response decreased (Dunham, 1977, 1978;
for further discussion). Furthermore, elicited Dunham & Grantmyre, 1982). A subse-
aggression in monkeys and rats has been ob- quent study suggested that the function of
served to decrease when the subject could the unpunished behavior also may determine
exhibit a response (e.g., lever press) to escape these side effects. Baker, Woods, Tait, and
from the situation in which the stimulus was Gardiner (1986) found that when eating in
delivered (e.g., Azrin, Hutchinson, & Hake, gerbils was punished by shock or noise, dig-
1966). This finding suggests that elicited ag- ging increased even though running was the
gression may be less problematic during most probable response during baseline. The
punishment than is commonly assumed be- authors suggested that digging was a species-
454 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
specific response to food deprivation (i.e., Cummings, & Van Houten, 1991) and with
digging was in the same response class as decreases in unpunished inappropriate behav-
eating). Results of studies on behavior that ior, including aggression and crying (e.g., Bit-
is maintained in the absence of experiment- good, Crowe, Suarez, & Peters, 1980; Lovaas
er-arranged reinforcement may have some & Simmons, 1969; Linscheid et al., 1990;
relevance to problem behavior that produces Ricketts et al., 1993; Singh, Watson, & Win-
its own reinforcing consequences (e.g., sen- ton, 1986). On the other hand, decreases in
sory stimulation). appropriate behavior (e.g., toy play, speaking)
In other basic studies with humans and pi- and increases in unpunished problem behav-
geons, the same reinforcer was used to estab- ior (e.g., aggression, emotional reactions, ste-
lish and maintain two or more functionally reotypic behavior) also have been reported
equivalent responses (e.g., key pecking, lever (e.g., Bitgood et al., 1980; Duker & Seys,
pulling) under concurrent or multiple sched- 1996; Foxx & Azrin, 1973; Harris & Wol-
ules. Punishment in the form of shock, time- chick, 1979; Pendergrass, 1971; Singh, Man-
out, and point or monetary loss was then de- ning, & Angell, 1982; Thompson et al.,
livered for one member of the response class 1999). A number of authors have suggested
while the effects on the other responses were that desirable side effects are more likely to
observed. The most common finding for both occur than undesirable side effects during
humans and nonhumans was an increase in treatment with punishment (see Lundervold
unpunished behavior (i.e., contrast; Bennett & Bourland, 1988, and Matson & Taras,
& Cherek, 1990; Bradshaw, Szabadi, & Bev- 1989, for reviews). The prevalence of these
an, 1979; Brethower & Reynolds, 1962; Cros- side effects is unknown, however, because rel-
bie, 1991; Powell, 1971; Thomas, 1968). atively few studies have directly examined the
However, results were inconsistent both with- effects of punishment on unpunished behav-
in and across subjects, and the effects often ior in clinical settings. More important, data
were short-lived. The main findings of Dun- collection often was limited to one or two
ham and colleagues (i.e., an increase in the collateral behaviors (see Sisson, Hersen, &
most probable behavior and a decrease in be- Van Hasselt, 1993, for a notable exception).
havior that often followed the punished be- Positive side effects also may be reported
havior; Dunham, 1977, 1978; Dunham & more frequently than undesirable side effects
Grantmyre, 1982) were not replicated with in clinical studies on punishment, regardless
college students when up to 10 responses were of the actual prevalence.
reinforced simultaneously and one response Although relatively incomplete, basic
was exposed to contingent point loss (Crosbie, findings in this area suggest some possible
1990, 1991). explanations for the inconsistent outcomes
From a clinical standpoint, collateral in- obtained in applied studies. Punishment-
creases in appropriate behavior and collateral elicited aggression and emotional reactions
decreases in unpunished inappropriate behav- are less probable when exposure to the pun-
ior would be desirable. In fact, applied re- ishing stimulus can be reduced or avoided
search findings suggest that a variety of de- (e.g., Azrin et al., 1966; Hunt & Brady,
sirable and undesirable side effects can occur 1955). Thus, these undesirable side effects
within and across individuals. Punishment of may have been less likely to occur in clinical
problem behavior has been associated with studies when a relatively brief punisher sup-
increases in appropriate behavior, such as pressed behavior to low levels or when an
compliance and toy play (e.g., Koegel, Fire- alternative source of reinforcement was high-
stone, Kramme, & Dunlap, 1974; Rolider, ly effective in strengthening an incompatible
PUNISHMENT 455
response. Conversely, aggression and other was suppressed via satiation or response
inappropriate behavior may have increased blocking (e.g., Dunham & Grantmyre,
in some applied studies because the respons- 1982), indicating that such effects may be
es were in the same response class as the associated with response suppression per se.
punished behavior (Baker et al., 1984). Re- In fact, several undesirable side effects of
sults of several basic studies with rats and punishment (e.g., increases in aggression, es-
pigeons also indicate that the likelihood of cape, and emotional reactions) have been as-
contrast is positively related to the density of sociated with extinction procedures (Lerman
the punishment schedule (e.g., Deluty, & Iwata, 1996a) that often were confounded
1976; Thomas, 1968). Thus, increases in with punishment in applied research.
unpunished behavior may have been more Nevertheless, basic findings suggest a
probable when the punishment procedure number of potentially useful clinical strate-
was implemented with a high degree of in- gies. Punishment-elicited aggression or emo-
tegrity (i.e., consistently followed each oc- tional responses (e.g., crying) may be atten-
currence of the response). uated by implementing procedures that min-
Some basic findings suggest that context imize exposure to the punishing stimulus
may be an important factor in determining (e.g., using brief stimuli that produce near-
which responses increase or decrease during complete suppression of the behavior;
punishment. Bolles, Holtz, Dunn, and Hill strengthening competing responses by deliv-
(1980) found that induction in rats was ering potent reinforcers on rich schedules).
more likely to occur when the unpunished Comprehensive descriptive or functional
response was performed under the same analyses of appropriate and inappropriate
stimulus condition as the punished response behavior in an individual’s repertoire may be
(e.g., pushing vs. lifting the same lever) than useful for predicting the likelihood of un-
under different stimulus conditions (e.g., desirable collateral effects and for arranging
manipulating separate levers). Crosbie et al. conditions to increase desirable effects. In-
(1997) found that induction in both hu- suring that caregivers withhold reinforce-
mans and nonhumans was more likely to oc- ment for unpunished problem behavior
cur under mixed schedules during the initial might prevent collateral increases in behavior
exposure to punishment, whereas contrast that is in the same response class as the pun-
was more likely to occur under multiple ished behavior. Alternative strategies for
schedules. Together, these findings suggest managing undesirable contrast effects (e.g.,
that functionally equivalent unpunished re- arranging punishment for other inappropri-
sponses may have been likely to decrease in ate responses) may be necessary if extinction
applied studies if the behavior tended to be cannot be used.
exhibited in the same context as the pun- Identifying and punishing precursors to
ished behavior. dangerous behavior (e.g., mild forms of self-
Further research on factors that determine injury that consistently precede more severe
whether a particular unpunished behavior forms) may lead to collateral reductions in
will increase, decrease, or remain unchanged the severe behavior, augmenting the safety
during punishment is needed so that the di- and efficacy of treatment (e.g., Dunham,
rection and nature of side effects can be pre- 1977, 1978). Collateral reductions in func-
dicted and controlled. Further research also tionally equivalent appropriate behavior
is needed to determine if these side effects might be avoided by arranging reinforce-
are specific to punishment because similar ment for the behavior under stimulus con-
results have been obtained when a response ditions that are distinctly different from
456 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
those present when the punished behavior 1998; Wacker et al., 1990); punishment may
occurs (e.g., by modifying features of the en- underlie the effects of certain common func-
vironment or introducing reinforcement in tion-based treatments (e.g., Lerman & Iwa-
a new setting). To this end, caregivers should ta, 1996b; Mazaleski et al., 1994); and care-
ensure that alternative reinforcement is avail- givers continue to use punishment to reduce
able for multiple responses across a variety problem behavior in the natural environ-
of contexts and settings. ment (e.g., Peterson & Martens, 1995).
Finally, the potential for undesirable side Further understanding of punishment
effects that involve caregiver behavior (e.g., processes is needed to develop a systematic,
overusing punishment, dehumanizing the effective technology of behavior change. A
punished individual) should be given more review of the applied literature indicates that
attention in the applied literature. Although a wide range of punishment procedures can
these effects are frequently described in text- successfully treat severe behavior disorders in
books and reviews (e.g., Cooper et al., 1987; clinical populations. Variables that contrib-
Guess et al., 1987; Kazdin, 2001), few stud- ute to the findings reported in this research,
ies have directly evaluated changes in the however, have not been delineated. The ma-
punishing agent’s behavior. Research findings jority of studies focused on procedural vari-
that are relevant to this phenomenon have ations of punishment rather than on factors
not supported the hypothesis that using pun- that may influence the direct and indirect
ishment negatively affects caregiver behavior effects of punishment. A review of labora-
or attitudes toward the punished individual tory research on punishment suggests that
(e.g., Bihm, Sigelman, & Westbrook, 1997; numerous variables alter basic processes in
Goza, Ricketts, & Perkins, 1993; Harris, complex ways. Factors such as history, rein-
Handleman, Gill, & Fong, 1991; Propst & forcement schedule, various punishment pa-
Nagle, 1981). Bihm et al., for example, found rameters, and alternative sources of reinforce-
that college students’ attitudes toward a fic- ment may influence the immediate effects of
titious client (i.e., ratings of client compe- punishment and other clinically relevant out-
tence, adjustment, and learning potential) comes, such as long-term maintenance, gen-
were related to treatment success rather than eralization, and the emergence of side ef-
to the type of intervention used (i.e., rein- fects. Basic findings also contradict some
forcement vs. mild or intense punishment). commonly held assumptions about punish-
ment effects and provide possible explana-
tions for the inconsistent findings that have
CONCLUSIONS been reported in applied studies.
The use of punishment to treat problem Nevertheless, current knowledge about
behavior in clinical populations has re- basic processes is insufficient for translation
mained controversial for many years (see to application. The basic literature on some
Iwata, 1988; Johnston, 1991). Results of ba- important relations remains incomplete
sic and applied research indicate that current (Baron, 1991; Crosbie, 1998). More impor-
treatment approaches based on punishment tant, the extent to which findings with non-
have advantages (e.g., they are highly effec- humans and response-contingent electric
tive) and disadvantages (e.g., there are un- shock can be extrapolated to the treatment
predictable side effects). Nevertheless, pun- of behavior disorders in clinical populations
ishment is still sometimes needed to reduce may be substantially restricted. Although
destructive behavior to acceptable levels some basic studies have been conducted with
(e.g., Grace et al., 1994; Hagopian et al., human participants, the typical punisher
PUNISHMENT 457
(point or monetary loss) and population treatment with reinforcement alone (e.g.,
(normal adults) may limit the relevance of differential reinforcement) or with common
the findings for individuals with restricted punishment procedures (e.g., time-out, ver-
verbal skills (see Hayes & McCurry, 1990, bal reprimands) will be successful. For ex-
for further discussion). ample, time-out is more likely to be effective
Research on punishment only tentatively than contingent verbal reprimands if prob-
supports most prescriptions for application lem behavior is sensitive to attention (Iwata,
discussed in previous articles on punishment Pace, Dorsey, et al., 1994). Problem behav-
and in this updated review of the literature. ior that occurs at high levels across a wide
Basic findings also suggest that the process variety of conditions may be less responsive
of punishment may be more complex than to treatment with reinforcement than behav-
frequently assumed. As such, clinicians, par- ior that is differentially low under certain
ents, and teachers should be cautioned about conditions (e.g., the play or control condi-
the need for further research on potential tion of the functional analysis; see Paisey,
strategies for increasing the effectiveness of Whitney, & Hislop, 1990). Functional anal-
mild punishers, for attenuating undesirable ysis methodology also may be useful for
aspects of punishment, and for successfully identifying potent reinforcers to arrange as
fading treatment with punishment. part of treatment (e.g., for use in differential
The basic literature suggests a number of reinforcement or noncontingent reinforce-
avenues for further research. In particular, ment procedures) and for identifying func-
knowledge about conditioned, intermittent, tionally equivalent appropriate behavior to
and delayed punishment and the interactive strengthen as part of treatment.
effects of reinforcement and punishment may Finally, a greater willingness to publish
lead to technological advances that increase studies showing treatment failures and other
the effectiveness and acceptability of punish- undesirable outcomes during punishment
ment in clinical settings. Concurrent sched- may be helpful. For example, unsuccessful at-
ules of reinforcement and punishment are es- tempts to attenuate side effects, to increase
pecially germane to application because mul- the efficacy of delayed punishment, or to pro-
tiple contingencies typically operate in the mote long-term maintenance and generaliza-
natural environment. Treatment also may tion may guide further research for delineat-
have a greater likelihood of immediate, long- ing variables that are and are not relevant to
term, and generalized success when multiple clinical application. This knowledge ultimate-
sources of alternative reinforcement are ar- ly may lead to more effective, empirically
ranged within the context of punishment. sound recommendations for treatment.
The refinement of functional analysis
methodologies has led to greater understand-
ing of variables that can maintain problem REFERENCES
behavior and an emphasis on function-based Abramowitz, A. J., & O’Leary, S. G. (1990). Effec-
intervention. Yet, potentially important links tiveness of delayed punishment in an applied set-
between function and treatments based on ting. Behavior Therapy, 21, 231–239.
Alberto, P. A., & Troutman, A. C. (1999). Applied
punishment often may be overlooked. The behavior analysis for teachers (5th ed.). Upper Sad-
utility of functional analysis for guiding de- dle River, NJ: Prentice Hall.
cisions about potential treatments that in- Altman, K., Haavik, S., & Cook, J. W. (1978). Pun-
clude a punishment component should be ishment of self-injurious behavior in natural set-
tings using contingent aromatic ammonia. Behav-
emphasized in further research. Results of iour Research and Therapy, 16, 85–96.
functional analyses may indicate whether Appel, J. B. (1968). Fixed-interval punishment. Jour-
458 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
nal of the Experimental Analysis of Behavior, 11, havior of mentally retarded children. Applied Re-
803–808. search in Mental Retardation, 2, 247–256.
Arbuckle, J. L., & Lattal, K. A. (1992). Molecular Barton, L. E., Brulle, A. R., & Repp, A. C. (1987).
contingencies in schedules of intermittent punish- Effects of differential schedule of time-out to re-
ment. Journal of the Experimental Analysis of Be- duce maladaptive responding. Exceptional Chil-
havior, 58, 361–375. dren, 53, 351–356.
Arntzen, E., & Werner, S. B. (1999). Water mist Bennett, R. H., & Cherek, D. R. (1990). Punished
punishment for two classes of problem behaviour. and nonpunished responding in a multiple sched-
Scandinavian Journal of Behaviour Therapy, 28, ule in humans: A brief report. The Psychological
88–93. Record, 40, 187–196.
Axelrod, S. A. (1990). Myths that (mis)guide our Bihm, E. M., Sigelman, C. K., & Westbrook, J. P.
profession. In A. C. Repp & N. N. Singh (Eds.), (1997). Social implications of behavioral inter-
Perspectives on the use of nonaversive and aversive ventions for persons with mental retardation.
interventions for persons with developmental dis- American Journal on Mental Retardation, 101,
abilities (pp. 59–72). Sycamore, IL: Sycamore. 567–578.
Azrin, N. H. (1956). Some effects of two intermittent Birnbrauer, J. S. (1968). Generalization of punish-
schedules of immediate and non-immediate pun- ment effects: A case study. Journal of Applied Be-
ishment. Journal of Psychology, 42, 3–21. havior Analysis, 1, 201–211.
Azrin, N. H. (1960). Effects of punishment intensity Bitgood, S. C., Crowe, M. J., Suarez, Y., & Peters, R.
during variable-interval reinforcement. Journal of D. (1980). Imbolization: Effects and side effects
the Experimental Analysis of Behavior, 3, 123–142. on stereotyped behavior in children. Behavior
Azrin, N. H., & Holz, W. C. (1966). Punishment. Modification, 4, 187–208.
In W. K. Honig (Ed.), Operant behavior: Areas of Boe, E. E. (1964). Extinction as a function of inten-
research and application (pp. 380–447). New York: sity of punishment, amount of training, and re-
Appleton-Century-Crofts. inforcement of a competing response. Canadian
Azrin, N. H., Holz, W. C., & Hake, D. F. (1963). Journal of Psychology, 18, 328–342.
Fixed-ratio punishment. Journal of the Experimen- Bolles, R. C., Holtz, R., Dunn, T., & Hill, W. (1980).
tal Analysis of Behavior, 6, 141–148. Comparisons of stimulus learning and response
Azrin, N. H., Hutchinson, R. R., & Hake, D. F. learning in a punishment situation. Learning and
(1966). Extinction induced aggression. Journal of Motivation, 11, 78–96.
the Experimental Analysis of Behavior, 9, 191–204. Bouzas, A. (1978). The relative law of effect: Effects
Azrin, N. H., & Powers, M. A. (1975). Eliminating of shock intensity on response strength in multiple
classroom disturbances of emotionally disturbed schedules. Journal of the Experimental Analysis of
children by positive practice procedures. Behavior Behavior, 30, 307–314.
Therapy, 6, 525–534. Bradshaw, C. M., Szabadi, E., & Bevan, P. (1977).
Baker, A. G., Woods, W., Tait, R., & Gardiner, K. Effect of punishment on human variable-interval
(1986). Punishment suppression: Some effects on performance. Journal of the Experimental Analysis
alternative behaviour. The Quarterly Journal of Ex- of Behavior, 27, 275–279.
perimental Psychology: Comparative and Physiolog- Bradshaw, C. M., Szabadi, E., & Bevan, P. (1978).
ical Psychology, 38B, 191–215. Effect of variable-interval punishment on the be-
Banks, R. K. (1967). Intermittent punishment effect havior of humans in variable-interval schedules of
(IPE) sustained through changed stimulus condi- monetary reinforcement. Journal of the Experimen-
tions and through blocks of nonpunished trials. tal Analysis of Behavior, 29, 161–166.
Journal of Experimental Psychology, 73, 456–460. Bradshaw, C. M., Szabadi, E., & Bevan, P. (1979).
Banks, R. K., & Vogel-Sprott, M. (1965). Effect of The effect of punishment on free-operant choice
delayed punishment on an immediately rewarded behavior in humans. Journal of the Experimental
response in humans. Journal of Experimental Psy- Analysis of Behavior, 31, 71–81.
chology, 70, 357–359. Brethower, D. M., & Reynolds, G. S. (1962). A fa-
Baron, A. (1991). Avoidance and punishment. In I. cilitative effect of punishment on unpunished be-
H. Iversen & K. A. Lattal (Eds.), Experimental havior. Journal of the Experimental Analysis of Be-
analysis of behavior (Part 1, pp. 173–217). Am- havior, 5, 191–199.
sterdam: Elsevier Science. Brown, R. T., & Wagner, A. R. (1964). Resistance to
Baron, A., Kaufman, A., & Fazzini, D. (1969). Den- punishment and extinction following training
sity and delay of punishment of free-operant with shock or nonreinforcement. Journal of Ex-
avoidance. Journal of the Experimental Analysis of perimental Psychology, 68, 503–507.
Behavior, 12, 1029–1037. Burchard, J. D., & Barrera, F. (1972). An analysis of
Barrett, R. P., Matson, J. L., Shapiro, E. S., & Ollen- timeout and response cost in a programmed en-
dick, T. H. (1981). A comparison of punishment vironment. Journal of Applied Behavior Analysis, 5,
and DRO procedures for treating stereotypic be- 271–282.
PUNISHMENT 459
Calhoun, K. S., & Matherne, P. (1975). The effects humans. Journal of the Experimental Analysis of Be-
of varying schedules of time-out on aggressive be- havior, 68, 161–175.
havior of a retarded girl. Journal of Behavior Ther- Crowell, C. R. (1974). Conditioned aversive aspects
apy and Experimental Psychiatry, 6, 139–143. of electric shock. Learning and Motivation, 5,
Camp, D. S., Raymond, G. A., & Church, R. M. 209–220.
(1966). Response suppression as a function of the Davidson, R. S. (1970). Conditioned punishment
schedule of punishment. Psychonomic Science, 5, and conditioned negative reinforcement on a mul-
23–24. tiple schedule. Psychonomic Science, 20, 163–165.
Camp, D. S., Raymond, G. A., & Church, R. M. Deluty, M. Z. (1976). Choice and the rate of pun-
(1967). Temporal relationship between response ishment in concurrent schedules. Journal of the
and punishment. Journal of Experimental Psychol- Experimental Analysis of Behavior, 25, 75–80.
ogy, 74, 114–123. Deluty, M. Z. (1978). Self-control and impulsiveness
Capaldi, E. D., Sheffer, J. D., Viveiros, D. M., Da- involving aversive events. Journal of Experimental
vidson, T. L., & Campbell, D. H. (1985). Shock Psychology, 4, 250–266.
preexposure and the reduced effectiveness of Deur, J. L., & Parke, R. D. (1970). Effects of incon-
shock. Learning and Motivation, 16, 357–380. sistent punishment on aggression in children. De-
Charlop, M. H., Burgio, L. D., Iwata, B. A., & Ivan- velopmental Psychology, 2, 403–411.
cic, M. T. (1988). Stimulus variation as a means Dinsmoor, J. A. (1952). A discrimination based on
of enhancing punishment effects. Journal of Ap- punishment. Quarterly Journal of Experimental
plied Behavior Analysis, 21, 89–95. Psychology, 4, 27–45.
Church, R. M. (1969). Response suppression. In B. Dixon, M. J., Helsel, W. J., Rojahn, J., Cipollone, R.,
A. Campbell & R. M. Church (Eds.), Punishment & Lubetsky, M. J. (1989). Aversive conditioning
and aversive behavior (pp. 111–156). New York: of visual screening with aromatic ammonia for
Appleton-Century-Crofts. treating aggressive and disruptive behavior in a de-
Cipani, E., Brendlinger, J., McDowell, L., & Usher, velopmentally disabled child. Behavior Modifica-
S. (1991). Continuous vs. intermittent punish- tion, 13, 91–107.
ment: A case study. Journal of Developmental and Doke, L. A., & Epstein, L. H. (1975). Oral overcor-
Physical Disabilities, 3, 147–156. rection: Side effects and extended applications. Jour-
Clark, H. B., Rowbury, T., Baer, A. M., & Baer, D. nal of Experimental Child Psychology, 20, 496–511.
M. (1973). Timeout as a punishing stimulus in Donnellan, A. M., & LaVigna, G. W. (1990). Myths
continuous and intermittent schedules. Journal of about punishment. In A. C. Repp & N. N. Singh
Applied Behavior Analysis, 6, 443–455. (Eds.), Perspectives on the use of nonaversive and
Cohen, P. S. (1968). Punishment: The interactive ef- aversive interventions for persons with developmental
fects of delay and intensity of shock. Journal of the disabilities (pp. 33–57). Sycamore, IL: Sycamore.
Experimental Analysis of Behavior, 11, 789–799. Dorsey, M. F., Iwata, B. A., Ong, P., & McSween, T.
Cole, G. A., Montgomery, R. W., Wilson, K. M., & E. (1980). Treatment of self-injurious behavior
Milan, M. A. (2000). Parametric analysis of over- using a water mist: Initial response suppression
correction duration effects: Is longer really better and generalization. Journal of Applied Behavior
than shorter? Behavior Modification, 24, 359–378. Analysis, 13, 343–353.
Cooper, J. O., Heron, T. E., & Heward, W. L. Duker, P. C., & Seys, D. M. (1996). Long-term use
(1987). Applied behavior analysis. New York: Mac- of electrical aversion treatment with self-injurious
millan. behavior. Research in Developmental Disabilities,
Corte, H. E., Wolf, M. M., & Locke, B. J. (1971). 17, 293–301.
A comparison of procedures for eliminating self- Dunham, P. J. (1977). The nature of reinforcing
injurious behavior of retarded adolescents. Journal stimuli. In W. K. Honig & J. E. R. Staddon
of Applied Behavior Analysis, 4, 201–213. (Eds.), Handbook of operant behavior (pp. 98–
Crosbie, J. (1990). Some effects of response cost on 124). Englewood Cliffs, NJ: Prentice Hall.
a free-operant multiple-response repertoire with Dunham, P. J. (1978). Changes in unpunished re-
humans. The Psychological Record, 40, 517–539. sponding during response-contingent punish-
Crosbie, J. (1991). The effects of punishment on un- ment. Animal Learning & Behavior, 6, 174–180.
punished reinforced free-operant responses. Aus- Dunham, P. J., & Grantmyre, J. (1982). Changes in
tralian Journal of Psychology, 43, 1–5. a multiple-response repertoire during response-
Crosbie, J. (1998). Negative reinforcement and pun- contingent punishment and response restriction:
ishment. In K. A. Lattal & M. Perone (Eds.), Sequential relationships. Journal of the Experimen-
Handbook of research methods in human operant tal Analysis of Behavior, 37, 123–133.
behavior (pp. 163–189). New York: Plenum. Dura, J. R. (1991). Controlling extremely dangerous
Crosbie, J., Williams, A. M., Lattal, K. A., Anderson, aggressive outbursts when functional analysis fails.
M. M., & Brown, S. M. (1997). Schedule inter- Psychological Reports, 69, 451–459.
actions involving punishment with pigeons and Eisenberger, R., Weier, F., Masterson, F. A., & Theis,
460 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
L. Y. (1989). Fixed-ratio schedules increase gen- ment of aggression. American Journal on Mental
eralized self-control: Preference for large rewards Retardation, 94, 27–36.
despite high effort or punishment. Journal of Ex- Foxx, R. M., & Livesay, J. (1984). Maintenance of
perimental Psychology: Animal Behavior Processes, response suppression following overcorrection.
15, 383–392. Analysis and Intervention in Developmental Dis-
Emerson, E. (1992). Self-injurious behaviour: An abilities, 1, 65–79.
overview of recent trends in epidemiological and Galbicka, G., & Branch, M. N. (1981). Selective
behavioural research. Mental Handicap Research, punishment of interresponse times. Journal of the
5, 49–81. Experimental Analysis of Behavior, 35, 311–322.
Epstein, R. (1984). An effect of immediate reinforce- Goh, H., Iwata, B. A., Shore, B. A., DeLeon, I. G.,
ment and delayed punishment, with possible im- Lerman, D. C., Ulrich, S. M., et al. (1995). An
plications for self-control. Journal of Behavior Ther- analysis of the reinforcing properties of hand
apy and Experimental Psychiatry, 15, 291–298. mouthing. Journal of Applied Behavior Analysis,
Estes, W. K. (1944). An experimental study of punish- 28, 269–283.
ment. Psychological Monographs, 57 (Serial No. 3). Goodall, G. (1984). Learning due to the response-shock
Evans, W. O. (1962). Producing either positive or contingency in signaled punishment. Quarterly Jour-
negative tendencies to a stimulus associated with nal of Experimental Psychology, 36B, 259–279.
shock. Journal of the Experimental Analysis of Be- Goza, A. B., Ricketts, R. W., & Perkins, T. S. (1993).
havior, 5, 335–337. The social validity of an argument supporting a
Fantino, E. (1973). Aversive control. In J. A. Nevin ban on aversive procedures. Journal of Intellectual
& G. S. Reynolds (Eds.), The study of behavior: Disability Research, 37, 449–458.
Learning, motivation, emotion, and instinct (pp. Grace, N. C., Kahng, S. W., & Fisher, W. W. (1994).
239–279). Glenview, IL: Scott, Foresman. Balancing social acceptability with treatment ef-
Farley, J. (1980). Reinforcement and punishment ef- fectiveness of an intrusive procedure: A case re-
port. Journal of Applied Behavior Analysis, 27,
fects in concurrent schedules: A test of two mod-
171–172.
els. Journal of the Experimental Analysis of Behav-
Grace, N. C., Thompson, R. H., & Fisher, W. W.
ior, 33, 311–326.
(1996). The treatment of covert self-injury
Favell, J. E., Azrin, N. H., Baumeister, A. A., Carr, E.
through contingencies on response products. Jour-
G., Dorsey, M. F., Forehand, R., et al. (1982). nal of Applied Behavior Analysis, 29, 239–242.
The treatment of self-injurious behavior. Behavior Guess, D., Helmstetter, E., Turnbull, H. R., III, &
Therapy, 13, 529–554. Knowlton, S. (1987). Use of aversive procedures
Fisher, W. W., Piazza, C. C., Bowman, L. G., Hago- with persons who are disabled: An historical review
pian, L. P., & Langdon, N. A. (1994). Empirical- and critical analysis. Seattle: The Association for
ly-derived consequences: A data-based method for Persons with Severe Handicaps.
prescribing treatments for destructive behavior. Re- Guttman, N., & Kalish, H. I. (1956). Discrimina-
search in Developmental Disabilities, 15, 133–149. bility and stimulus generalization. Journal of Ex-
Fisher, W. W., Piazza, C. C., Bowman, L. G., Hago- perimental Psychology, 51, 79–88.
pian, L. P., Owens, J. C., & Slevin, I. (1992). A Hagopian, L. P., Fisher, W. W., Sullivan, M. T., Ac-
comparison of two approaches for identifying re- quisto, J., & LeBlanc, L. A. (1998). Effectiveness
inforcers for persons with severe and profound of functional communication training with and
disabilities. Journal of Applied Behavior Analysis, without extinction and punishment. Journal of
25, 491–498. Applied Behavior Analysis, 31, 211–235.
Fisher, W. W., Piazza, C. C., Bowman, L. G., Kurtz, Hake, D. F., & Azrin, N. H. (1965). Conditioned
P. F., Sherer, M. R., & Lachman, S. R. (1994). punishment. Journal of the Experimental Analysis
A preliminary evaluation of empirically derived of Behavior, 8, 279–293.
consequences for the treatment of pica. Journal of Hake, D. F., Azrin, N. H., & Oxford, R. (1967). The
Applied Behavior Analysis, 27, 447–457. effects of punishment intensity on squirrel mon-
Fisher, W. W., Piazza, C. C., Cataldo, M. F., Harrell, keys. Journal of the Experimental Analysis of Be-
R., Jefferson, G., & Conner, R. (1993). Func- havior, 10, 95–107.
tional communication training with and without Halevy, G., Feldon, J., & Weiner, I. (1987). Resis-
extinction and punishment. Journal of Applied Be- tance to extinction and punishment following
havior Analysis, 26, 23–36. training with shock and non-reinforcement: Fail-
Foxx, R. M., & Azrin, N. H. (1973). The elimination ure to obtain cross-tolerance. Quarterly Journal of
of autistic self-stimulatory behavior by overcorrec- Experimental Psychology, 39B, 147–160.
tion. Journal of Applied Behavior Analysis, 6, 1–14. Harris, S. L., Handleman, J. S., Gill, M. J., & Fong,
Foxx, R. M., Bittle, R. G., & Faw, G. D. (1989). A P. L. (1991). Does punishment hurt? The impact
maintenance strategy for discontinuing aversive of aversives on the clinician. Research in Develop-
procedures: A 32-month follow-up of the treat- mental Disabilities, 12, 17–24.
PUNISHMENT 461
Harris, S. L., & Wolchick, S. A. (1979). Suppression Iwata, B. A., Vollmer, T. R., & Zarcone, J. R. (1990).
of self-stimulation: Three alternative strategies. The experimental (functional) analysis of behavior
Journal of Applied Behavior Analysis, 12, 185–198. disorders: Methodology, applications, and limita-
Hayes, L. J., & McCurry, C. (1990). Moral and sci- tions. In A. C. Repp & N. N. Singh (Eds.), Per-
entific aspects of the punishment controversy. In spectives on the use of nonaversive and aversive in-
A. C. Repp & N. N. Singh (Eds.), Perspectives on terventions for persons with developmental disabili-
the use of nonaversive and aversive interventions for ties (pp. 301–330). Sycamore, IL: Sycamore.
persons with developmental disabilities (pp. 87– Jacob-Timm, S. (1996). Ethical and legal issues as-
101). Sycamore, IL: Sycamore. sociated with the use of aversives in the public
Hearst, E. (1965). Stress induced breakdown of an schools: The SIBIS controversy. School Psychology
appetitive discrimination. Journal of the Experi- Review, 25, 184–198.
mental Analysis of Behavior, 8, 135–146. Johnson, D. L., McGlynn, F. D., & Topping, J. S.
Herman, R. L., & Azrin, N. H. (1964). Punishment (1973). The relative efficiency of four response-
by noise in an alternative response situation. Jour- elimination techniques following variable-ratio re-
nal of the Experimental Analysis of Behavior, 7, inforcement training. The Psychological Record, 23,
185–188. 203–208.
Hobbs, S. A., Forehand, R., & Murray, R. G. (1978). Johnston, J. M. (1991). What can behavior analysis
Effects of various durations of timeout on the learn from the aversives controversy? The Behavior
non-compliant behavior of children. Behavior Analyst, 14, 187–196.
Therapy, 9, 652–656. Kaufman, A., & Baron, A. (1968). Suppression of
Hoffman, H. S., & Fleshler, M. (1965). Stimulus as- behavior by timeout punishment when suppres-
pects of aversive controls: The effects of response sion results in loss of positive reinforcement. Jour-
contingent shock. Journal of the Experimental nal of the Experimental Analysis of Behavior, 11,
Analysis of Behavior, 8, 89–96. 595–607.
Holz, W. C., & Azrin, N. H. (1962). Recovery dur- Kazdin, A. E. (2001). Behavior modification in applied
ing punishment by intense noise. Psychological Re- settings (6th ed.). Belmont, CA: Wadsworth/
ports, 11, 655–657. Thompson Learning.
Holz, W. C., & Azrin, N. H. (1963). A comparison Kendall, P. C., Nay, W. R., & Jeffers, J. (1975). Time-
of several procedures for eliminating behavior. out duration and contrast effects: A systematic
Journal of the Experimental Analysis of Behavior, 6, evaluation of a successive treatments design. Be-
399–406. havior Therapy, 6, 609–615.
Holz, W. C., Azrin, N. H., & Ayllon, T. (1963). Kennedy, C. H., & Souza, G. (1995). Functional
Elimination of behavior of mental patients by re- analysis and treatment of eye poking. Journal of
sponse-produced extinction. Journal of the Exper- Applied Behavior Analysis, 28, 27–37.
imental Analysis of Behavior, 6, 407–412. Koegel, R. L., Firestone, P. B., Kramme, K. W., &
Honig, W. K., & Slivka, R. M. (1964). Stimulus gen- Dunlap, G. (1974). Increasing spontaneous play
eralization of the effects of punishment. Journal of by suppressing self-stimulation in autistic children.
the Experimental Analysis of Behavior, 7, 21–25. Journal of Applied Behavior Analysis, 7, 521–528.
Hunt, H. F., & Brady, J. V. (1955). Some effects of Lalli, J. S., & Goh, H. (1993). Naturalistic observa-
punishment and intercurrent ‘‘anxiety’’ on a sim- tions in community settings. In R. Reichle & D.
ple operant. Journal of Comparative and Physiolog- P. Wacker (Eds.), Communicative alternatives to
ical Psychology, 48, 305–310. challenging behavior (pp. 11–39). Paul H. Brookes.
Hutchinson, R. R. (1977). By-products of aversive Lande, S. D. (1981). An interresponse time analysis
control. In W. K. Honig & J. E. R. Staddon of variable-ratio punishment. Journal of the Exper-
(Eds.), Handbook of operant behavior (pp. 415– imental Analysis of Behavior, 35, 55–67.
431). Englewood Cliffs, NJ: Prentice Hall. Laurence, M. T., Hineline, P. N., & Bersh, P. J.
Iwata, B. A. (1988). The development and adoption (1994). The puzzle of responding maintained by
of controversial default technologies. The Behavior response-contingent shock. Journal of the Experi-
Analyst, 11, 149–157. mental Analysis of Behavior, 61, 135–153.
Iwata, B. A., Pace, G. M., Cowdery, G. E., & Mil- LaVigna, G. W., & Donnellan, A. M. (1986). Alter-
tenberger, R. G. (1994). What makes extinction natives to punishment: Solving behavior problems
work: An analysis of procedural form and func- with non-aversive strategies. New York: Irvington.
tion. Journal of Applied Behavior Analysis, 27, Lerman, D. C., & Iwata, B. A. (1996a). Developing
131–144. a technology for the use of operant extinction in
Iwata, B. A., Pace, G. M., Dorsey, M. F., Zarcone, J. clinical settings: An examination of basic and ap-
R., Vollmer, T. R., Smith, R. G., et al. (1994). plied research. Journal of Applied Behavior Analysis,
The functions of self-injurious behavior: An ex- 29, 345–382.
perimental-epidemiological analysis. Journal of Ap- Lerman, D. C., & Iwata, B. A. (1996b). A method-
plied Behavior Analysis, 27, 215–240. ology for distinguishing between extinction and
462 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
punishment effects associated with response Sensory extinction or punishment effects? Journal
blocking. Journal of Applied Behavior Analysis, 29, of Applied Behavior Analysis, 27, 345–355.
231–233. Mazur, J. E. (1998). Learning and behavior (4th ed.).
Lerman, D. C., Iwata, B. A., Shore, B. A., & DeLeon, Upper Saddle River, NJ: Prentice Hall.
I. G. (1997). Effects of intermittent punishment McGee, J. J., Menolascino, F. J., Hobbs, D. C., &
on self-injurious behavior: An evaluation of sched- Menousek, P. E. (1987). Gentle teaching: A non-
ule thinning. Journal of Applied Behavior Analysis, aversive approach to helping persons with mental re-
30, 187–201. tardation. New York: Human Sciences Press.
Lindberg, J. S., Iwata, B. A., & Kahng, S. W. (1999). McMillan, D. E. (1967). A comparison of the pun-
On the relation between object manipulation and ishing effects of response-produced shock and re-
stereotypic self-injurious behavior. Journal of Ap- sponse-produced time out. Journal of the Experi-
plied Behavior Analysis, 32, 51–62. mental Analysis of Behavior, 10, 439–449.
Linscheid, T. R., Hartel, F., & Cooley, N. (1993). Are Michael, J. L. (1993). Concepts and principles of be-
aversive procedures durable? A five year follow-up havior analysis. Kalamazoo, MI: Society for the
of three individuals treated with contingent elec- Advancement of Behavior Analysis.
tric shock. Child and Adolescent Mental Health Miller, N. B., & Zimmerman, J. (1966). The effects
Care, 3, 67–76. of a pre-time-out stimulus on matching-to-sample
Linscheid, T. R., Iwata, B. A., Ricketts, R. W., Wil- of humans. Journal of the Experimental Analysis of
liams, D. E., & Griffin, J. C. (1990). Clinical Behavior, 9, 487–499.
evaluation of the self-injurious behavior inhibiting Miller, N. E. (1960). Learning resistance to pain and
system (SIBIS). Journal of Applied Behavior Anal- fear effects over learning, exposure, and rewarded
ysis, 23, 53–78. exposure in context. Journal of Experimental Psy-
Linscheid, T. R., & Meinhold, P. (1990). The con- chology, 60, 137–145.
troversy over aversives: Basic operant research and Miltenberger, R. G. (2001). Behavior modification:
Principles and procedures (2nd ed.). Belmont, CA:
side effects of punishment. In A. C. Repp & N.
Wadsworth.
N. Singh (Eds.), Perspectives on the use of nonav-
Mowrer, O. H., & Aiken, E. G. (1954). Contiguity
ersive and aversive interventions for persons with de-
vs. drive-reduction in conditioned fear: Temporal
velopmental disabilities (pp. 435–450). Sycamore,
variations in conditioned and unconditioned stim-
IL: Sycamore. ulus. American Journal of Psychology, 67, 26–38.
Lovaas, O. I., & Favell, J. E. (1987). Protection for Mowrer, O. H., & Solomon, L. N. (1954). Conti-
clients undergoing aversive/restrictive interven- guity vs. drive-reduction in conditioned fear: The
tions. Education and Treatment of Children, 10, proximity and abruptness of drive-reduction.
311–325. American Journal of Psychology, 67, 15–25.
Lovaas, O. I., & Simmons, J. Q. (1969). Manipula- Murray, M., & Nevin, J. A. (1967). Some effects of
tion of self-destruction in three retarded children. correlation between response-contingent shock
Journal of Applied Behavior Analysis, 2, 143–157. and reinforcement. Journal of the Experimental
Lundervold, D., & Bourland, G. (1988). Quantita- Analysis of Behavior, 10, 301–309.
tive analysis of treatment of aggression, self-injury, O’Brien, F. (1989). Punishment for people with de-
and property destruction. Behavior Modification, velopmental disabilities. In E. Cipani (Ed.), The
12, 590–617. treatment of severe behavior disorders: Behavior
Marholin, D., & Townsend, N. M. (1978). An ex- analysis approaches (pp. 37–58). Washington, DC:
perimental analysis of side effects and response American Association of Mental Retardation.
maintenance of a modified overcorrection proce- O’Donnell, J., & Crosbie, J. (1998). Punishment
dure. Behavior Therapy, 9, 383–390. generalization gradients with humans. The Psycho-
Martin, G., & Pear, J. (1996). Behavior modification: logical Record, 48, 211–232.
What it is and how to do it (5th ed.). Upper Saddle O’Donnell, J., Crosbie, J., Williams, D. C., & Saun-
River, NJ: Prentice Hall. ders, K. J. (2000). Stimulus control and gener-
Matson, J. L., & DiLorenzo, T. M. (1984). Punish- alization of point-loss punishment with humans.
ment and its alternatives: New perspectives for be- Journal of the Experimental Analysis of Behavior,
havior modification. New York: Springer. 73, 261–274.
Matson, J. L., & Taras, M. E. (1989). A 20 year Orme-Johnson, D. W. (1967). Response suppression
review of punishment and alternative methods to as a function of a vacation from punishment. Psy-
treat problem behaviors in developmentally de- chonomic Science, 8, 277–278.
layed persons. Research in Developmental Disabil- Orme-Johnson, D. W., & Yarczower, M. (1974).
ities, 10, 85–104. Conditioned suppression. Journal of the Experi-
Mazaleski, J. L., Iwata, B. A., Rodgers, T. A., Vollmer, mental Analysis of Behavior, 21, 57–74.
T. R., & Zarcone, J. R. (1994). Protective equip- Paisey, T. J. H., Whitney, R. B., & Hislop, P. M.
ment as treatment for stereotypic hand mouthing: (1990). Client characteristics and treatment selec-
PUNISHMENT 463
tion: Legitimate influences and misleading infer- opmentally disabled children. Journal of Abnormal
ences. In A. C. Repp & N. N. Singh (Eds.), Per- Psychology, 6, 299–310.
spectives on the use of nonaversive and aversive in- Risley, T. R. (1968). The effects and side effects of
terventions for persons with developmental disabili- punishing the autistic behaviors of a deviant child.
ties (pp. 175–197). Sycamore, IL: Sycamore. Journal of Applied Behavior Analysis, 1, 21–34.
Parsons, R. D., Hinson, S. L., & Sardo-Brown, D. Rodriguez, W. A., & Logan, F. A. (1980). Preference
(2001). Educational psychology: A practitioner-re- for punishment of the instrumental or the con-
searcher model of teaching. Belmont, CA: Wads- summatory response. Animal Learning & Behav-
worth/Thomson Learning. ior, 8, 116–119.
Patel, M. R., Carr, J. E., Kim, C., Robles, A., & Eas- Rolider, A., Cummings, A., & Van Houten, R.
tridge, D. (2000). Functional analysis of aberrant (1991). Side effects of therapeutic punishment on
behavior maintained by automatic reinforcement: academic performance and eye contact. Journal of
Assessments of specific sensory reinforcers. Re- Applied Behavior Analysis, 24, 763–773.
search in Developmental Disabilities, 21, 393–407. Rolider, A., & Van Houten, R. (1985). Suppressing
Pendergrass, V. E. (1971). Effects of length of time- tantrum behavior in public places through the use
out from positive reinforcement and schedule of of delayed punishment mediated by audio record-
application reinforcement and schedule of appli- ings. Behavior Therapy, 16, 181–194.
cation in suppression of aggressive behavior. The Rolider, A., & Van Houten, R. (1990). The role of
Psychological Record, 21, 75–80. reinforcement in reducing inappropriate behavior:
Peterson, F. M., & Martens, B. K. (1995). A com- Some myths and misconceptions. In A. C. Repp
parison of behavioral interventions reported in & N. N. Singh (Eds.), Perspectives on the use of
treatment studies and programs for adults with nonaversive and aversive interventions for persons
developmental disabilities. Research in Develop- with developmental disabilities (pp. 119–127). Syc-
mental Disabilities, 16, 27–41. amore, IL: Sycamore.
Piazza, C. C., Adelinis, J. D., Hanley, G. P., Goh, H., Rolider, A., Williams, L., Cummings, A., & Van Hou-
& Delia, M. D. (2000). An evaluation of the ten, R. (1991). The use of a brief movement re-
effects of matched stimuli on behaviors main- striction procedure to eliminate severe inappro-
tained by automatic reinforcement. Journal of Ap- priate behavior. Journal of Behavior Therapy and
plied Behavior Analysis, 33, 13–27. Experimental Psychiatry, 22, 23–30.
Powell, R. W. (1970). The effect of punishment Rollings, J. P., & Baumeister, A. A. (1981). Stimulus
shock intensity upon responding under multiple control of stereotypic responding: Effects on target
schedules. Journal of the Experimental Analysis of and collateral behavior. American Journal of Men-
Behavior, 14, 201–211. tal Deficiency, 86, 67–77.
Powell, R. W. (1971). Some effects of punishment Rollings, J. P., Baumeister, A. A., & Baumeister, A. A.
shock intensity upon discriminative responding. (1977). The use of overcorrection procedures to
Journal of the Experimental Analysis of Behavior, eliminate the stereotyped behaviors of retarded in-
15, 109–116. dividuals: An analysis of collateral behaviors and
Propst, L. B., & Nagle, R. J. (1981). Effects of la- generalization of suppressive effects. Behavior
beling and a child’s reaction to punishment on Modification, 1, 29–46.
subsequent disciplinary practices of adults and Romanczyk, R. G. (1977). Intermittent punishment
peers. American Journal of Mental Deficiency, 86, of self-stimulation: Effectiveness during applica-
287–294. tion and extinction. Journal of Consulting and
Rachlin, H. (1966). Recovery of responses during Clinical Psychology, 45, 53–60.
mild punishment. Journal of the Experimental Scobie, S. R., & Kaufman, A. (1969). Intermittent
Analysis of Behavior, 9, 251–263. punishment of human responding maintained by
Rawson, R. A., & Leitenberg, H. (1973). Reinforced intermittent reinforcement. Journal of the Experi-
alternative behavior during punishment and ex- mental Analysis of Behavior, 12, 137–147.
tinction with rats. Journal of Comparative and Scotti, J. R., Evans, I. M., Meyer, L. H., & Walker,
Physiological Psychology, 85, 593–600. P. (1991). A meta-analysis of intervention re-
Reid, D. J., Parsons, M. B., Phillips, J. F., & Green, search with problem behavior: Treatment validity
C. W. (1993). Reduction of self-injurious hand and standards of practice. American Journal on
mouthing using response blocking. Journal of Ap- Mental Retardation, 96, 233–256.
plied Behavior Analysis, 26, 139–140. Shemer, A., & Feldon, J. (1984). Long-term partial re-
Ricketts, R. W., Goza, A. B., & Matese, M. (1993). inforcement extinction effect and long-term partial
A 4-year follow-up of treatment of self-injury. punishment effect in a one-trial-a-day paradigm.
Journal of Behavior Therapy and Experimental Psy- Bulletin of the Psychonomic Society, 22, 221–224.
chiatry, 24, 57–62. Shore, B. A., Iwata, B. A., DeLeon, I. G., Kahng, S.,
Rincover, A. (1978). Sensory extinction: A procedure & Smith, R. G. (1997). An analysis of reinforcer
for eliminating self-stimulatory behavior in devel- substitutability using object manipulation and
464 DOROTHEA C. LERMAN and CHRISTINA M. VORNDRAN
self-injury as competing responses. Journal of Ap- on human behavior (pp. 13–44). New York: Aca-
plied Behavior Analysis, 30, 21–40. demic Press.
Sidman, M. (1989). Coercion and its fallout. Boston: Van Houten, R., Axelrod, S., Bailey, J. S., Favell, J.
Authors Cooperative. E., Foxx, R. N., Iwata, B. A., et al. (1988). The
Singh, N. N., Dawson, M. J., & Manning, P. J. right to effective behavioral treatment. The Behav-
(1981). The effects of physical restraint on self- ior Analyst, 11, 111–114.
injurious behaviour. Journal of Mental Deficiency Van Houten, R., & Rolider, A. (1988). Recreating
Research, 25, 207–216. the scene: An effective way to provide delayed
Singh, N. N., Manning, P. J., & Angell, M. J. (1982). punishment for inappropriate motor behavior.
Effects of oral hygiene punishment procedure on Journal of Applied Behavior Analysis, 21, 187–192.
chronic rumination and collateral behaviors in Vogel-Sprott, M. (1967). Partial-reward training for
monozygous twins. Journal of Applied Behavior resistance to punishment and to subsequent ex-
Analysis, 15, 309–314. tinction. Journal of Experimental Psychology, 75,
Singh, N. N., Watson, J. E., & Winton, A. S. (1986). 138–140.
Treating self-injury: Water mist spray versus facial Vollmer, T. R., Borrero, J. C., Wright, C. S., Van
screening or forced arm exercise. Journal of Applied Camp, C., & Lalli, J. S. (2001). Identifying pos-
Behavior Analysis, 19, 403–410. sible contingencies during descriptive analyses of
Sisson, L. A., Hersen, M., & Van Hasselt, V. B. severe behavior disorders. Journal of Applied Be-
(1993). Improving the performance of youth with havior Analysis, 34, 269–287.
dual sensory impairment: Analyses and social val- Vollmer, T. R., & Iwata, B. A. (1993). Implications
idation of procedures to reduce maladaptive re- of a functional analysis technology for the use of
sponding in vocational and leisure settings. Be- restrictive behavioral interventions. Child and Ad-
havior Therapy, 24, 553–571. olescent Mental Health Care, 3, 95–113.
Skinner, B. F. (1938). The behavior of organisms: An Wacker, D. P., Steege, M. W., Northup, J., Sasso, G.
experimental analysis. Acton, MA: Copley. M., Berg, W., Reimers, T., et al. (1990). A com-
Solnick, J. V., Rincover, A., & Peterson, C. R. (1977). ponent analysis of functional communication
Some determinants of the reinforcing and punish- training across three topographies of severe behav-
ing effects of timeout. Journal of Applied Behavior ior problems. Journal of Applied Behavior Analysis,
Analysis, 10, 415–424. 23, 417–429.
Stokes, T. F., & Baer, D. M. (1977). An implicit Walker, J. E., & Shea, T. M. (1999). Behavior man-
technology of generalization. Journal of Applied agement: A practical approach for educators (7th
Behavior Analysis, 10, 349–367. ed.). Upper Saddle River, NJ: Prentice Hall.
Tate, B. G., & Baroff, G. S. (1966). Aversive control Weiner, H. (1962). Some effects of response cost
of self-injurious behavior in a psychotic boy. Be- upon human operant behavior. Journal of the Ex-
haviour Research and Therapy, 4, 281–287. perimental Analysis of Behavior, 5, 201–208.
Terris, W., & Barnes, M. (1969). Learned resistance Weiner, H. (1964). Response cost and fixed-ratio per-
to punishment and subsequent responsiveness to formance. Journal of the Experimental Analysis of
the same and novel punishers. Psychonomic Sci- Behavior, 7, 79–81.
ence, 15, 49–50. White, G. D., Nielsen, G., & Johnson, S. M. (1972).
Thomas, J. R. (1968). Fixed-ratio punishment by Timeout duration and the suppression of deviant
timeout of concurrent variable-interval behavior. behavior in children. Journal of Applied Behavior
Journal of the Experimental Analysis of Behavior, Analysis, 5, 111–120.
11, 609–616. Williams, D. E., Kirkpatrick-Sanchez, S., & Crocker,
Thompson, R. H., Iwata, B. A., Conners, J., & Roscoe, W. T. (1994). A long-term follow-up of treat-
E. M. (1999). Effects of reinforcement for alter- ment for severe self-injury. Research in Develop-
native behavior during punishment of self-injury. mental Disabilities, 15, 487–501.
Journal of Applied Behavior Analysis, 32, 317–328. Williams, D. E., Kirkpatrick-Sanchez, S., & Iwata, B.
Thorndike, E. L. (1932). Fundamentals of learning. A. (1993). A comparison of shock intensity in
New York: Teachers College. the treatment of longstanding and severe self-in-
Trenholme, I. A., & Baron, A. (1975). Immediate and jurious behavior. Research in Developmental Dis-
delayed punishment of human behavior by loss of abilities, 14, 207–219.
reinforcement. Learning and Motivation, 6, 62–79. Williams, D. R., & Barry, H., III. (1966). Counter
Ulrich, R. E., & Azrin, N. H. (1962). Reflexive fight- conditioning in an operant conflict situation.
ing in response to aversive stimulation. Journal of Journal of Comparative and Physiological Psycholo-
the Experimental Analysis of Behavior, 5, 511–520. gy, 61, 154–156.
Van Houten, R. (1983). Punishment: From the ani- Received September 1, 2001
mal laboratory to the applied setting. In S. Axel- Final acceptance August 12, 2002
rod & J. Apsche (Eds.), The effects of punishment Action Editor, Joseph Spradlin
UC Davis
UC Davis Previously Published Works
Title
Owner attachment and problem behaviors related to relinquishment and training techniques of
dogs.
Permalink
https://escholarship.org/uc/item/9f79z91b
Journal
Journal of applied animal welfare science : JAAWS, 16(2)
ISSN
1088-8705
Authors
Kwan, Jennifer Y
Bain, Melissa J
Publication Date
2013
DOI
10.1080/10888705.2013.768923
Peer reviewed
To cite this article: Jennifer Y. Kwan & Melissa J. Bain (2013) Owner Attachment and Problem
Behaviors Related to Relinquishment and Training Techniques of Dogs, Journal of Applied Animal
Welfare Science, 16:2, 168-183, DOI: 10.1080/10888705.2013.768923
168
ATTACHMENT AND BEHAVIOR 169
have problem behaviors (Diesel, Brodbelt, & Pfeiffer, 2009; Salman et al., 2000;
Segurson, Serpell, & Hart, 2005; Sigler, 1991). The number of dogs euthanized
annually in American shelters due to behavioral problems is speculated to
potentially be in the millions (National Council on Pet Population Study and
Policy, 1997). Behavior, therefore, has a considerable impact on the welfare of
canine companions and on the human-nonhuman animal bond.
In addition to problematic behaviors, or perhaps as a sequela of them, a weak-
ened attachment is likely related to relinquishment. Attachment is a complex
phenomenon that can be described as a combination of time and activities spent
with the pet (companion animal), interest and emotional closeness toward the
pet, knowledge about the pet and his or her care, and behavioral responses to the
pet (Melson, 1990). Prior studies have shown that there is a correlation between
strength of owner (companion animal guardian) attachment and relinquishment
to shelters (Kobelt, Hemsworth, Barnett, & Coleman, 2003; Serpell, 1996), but
none of the studies used validated survey tools. One study (Segurson et al.,
2005) that used a validated survey tool demonstrated that dogs relinquished to
shelters were more likely to have problem behaviors than those in the general
population. However, neither owner attachment nor the training methods used
were evaluated. There is also evidence that breeds differ in their behavior (Duffy,
Hsu, & Serpell, 2008; Turcsán, Kubinyi, & Miklósi, 2011). As with most studies
of companion animals with problem behaviors, it is the owner’s perception that a
behavior is a problem, and this is often due to unmet expectations and potential
owner-dog mismatch.
Researchers have demonstrated a correlation between training and techniques
utilized in training and the incidence of problematic behaviors. They also found
that obedience training was correlated with a decrease in unwanted behaviors
(Alexander, Friend, & Haug, 2011; Clark & Boyer, 1993; Jagoe & Serpell,
1996), increased adoptability from a rescue shelter (Hays, 2004), and retention
of dogs in a household (Duxbury, Jackson, Line, & Anderson, 2003). There
is a wide variety of training methods and tools used to train dogs, from pos-
itive reinforcement-based methods (utilizing such rewards as food and praise)
to positive punishment-based aversive methods (utilizing such punishments as
giving collar corrections, pinning a dog to the ground, and yelling). Tools often
associated with the delivery of positive punishment include choke/slip collars,
prong/pinch collars, and shock collars. Dogs who had worn a head collar, most
often used by people who utilize positive reinforcement-based techniques, were
more likely to be retained in their homes than dogs who did not wear one of
these collars (Duxbury et al., 2003).
Positive reinforcement methods have been shown to be equally or more ef-
fective than punishment-based methods for teaching desired behaviors and elim-
inating undesired behaviors, and they result in better trained dogs overall (Hiby,
Rooney, & Bradshaw, 2004; Rockwood & Bain, 2007). In one study, working
170 KWAN AND BAIN
METHODS
Selection
The study was approved by the Institutional Review Board on Human Subjects
Study. Two groups of dog owners were solicited during shelter hours of operation
during July and August of 2010. A power analysis was not performed a priori.
ATTACHMENT AND BEHAVIOR 171
Relinquishing owners were solicited from three animal shelters: Sacramento So-
ciety for the Prevention of Cruelty to Animals (SPCA), the Sacramento County
Animal Care and Regulation (County), and the City of Sacramento Animal Care
and Control (City), which were all located in Sacramento, California. Continuing
owners were solicited from owners at vaccination clinics held at these three
shelters. Vaccination clinics are places in which owners can have their pets
preventatively vaccinated but with limited veterinary examination. The reason for
soliciting continuing owners from these vaccination clinics was the potential that
those seeking medical care at these facilities would most likely relinquish their
dogs to that specific shelter; however, we could not state this for certain. There
is no evidence that owners who take their dogs to vaccination clinics are less
attached to their dogs compared with those who take their dogs to a traditional
veterinary clinic (Helms & Bain, 2009). The SPCA had regularly scheduled
vaccination clinics during most hours of operation, whereas the County and
City shelters had once monthly mobile vaccination clinics during specific hours.
Pilot Test
While developing the survey tools, owners at the University of California School
of Veterinary Medicine (UCD-SVM) Veterinary Medical Teaching Hospital were
asked to fill out the continuing owner survey and make comments regarding the
clarity of questions and length of time it took to fill out the survey. We evaluated
their answers to open-ended questions in order to best design the closed-answer
questions. The same was done for relinquishing owners at City and County. The
surveys were subsequently rewritten on the basis of responses and comments
received.
Survey
The surveys were offered in English and Spanish. Surveys from relinquishing
dog owners were completed by one investigator (J. K.) via in-person interviews
or handed directly to owners to fill out at the SPCA and County shelters. At the
City shelter, a paid shelter staff employee handed the survey to relinquishing
dog owners so that they could fill it out. Surveys were distributed during hours
of operation in which the public was allowed to relinquish dogs. Surveys from
continuing dog owners were completed by the same investigator via in-person
interviews or handed directly to owners to fill out at the shelters. Surveys dis-
tributed at the SPCA were filled out during normal hours of regularly scheduled
vaccination clinics. At the City and County shelters, data were collected during
two monthly clinics.
All survey participants were notified that no identifying information would be
collected and that shelter staff would not see the responses of their survey, thus
172 KWAN AND BAIN
helping to allay concerns of the owners that the surveys would affect the care
and treatment of their dogs. It was conveyed to them that the care or treatment of
their dogs was not dependent on whether they took part in the study. The survey
was labeled with the UCD-SVM logo, and contact information was clearly from
the university. The investigator wore a badge that identified her as a UCD-SVM
student. The surveys were directly placed into a provided envelope that was
immediately sealed. Sampling from both groups was via convenience sampling
in which some owners were not asked to take part in the survey to ensure the
investigator’s safety from potentially aggressive dogs or from owners who were
overtly emotional during relinquishment. Exclusion criteria included dogs less
than 4 months of age to decrease the chance of having a dog enter the study as
an unplaced puppy from a whelped litter. A dog was also excluded if he or she
had been owned for less than 3 months to allow us to focus on the relationship
between training tools, overall behavioral satisfaction, and owner attachment
(New et al., 2000). Dogs who were relinquished specifically for euthanasia for
medical reasons were also excluded to avoid additional emotional trauma to
these owners.
Owners were asked closed- and open-ended questions in the following cat-
egories via a questionnaire (available from M. B. upon request): demographic
information of the owner and dog, training methods and tools, frequency in
which his or her dog engaged in problematic behaviors, satisfaction with the
behavior of his or her dog as well as with the methods and tools used for
training, and owner attachment to his or her dog via the validated Lexington
Attachment to Pets Scale (LAPS; Helms & Bain, 2009; Johnson et al., 1992).
Relinquishing owners were asked closed- and open-ended questions regarding
their reasons for relinquishing their dogs to a shelter, including unemployment,
foreclosures, and other financial considerations, with scores ranging from 0 to
5 for each choice (0 being no influence, 5 being a very strong influence). They
were also asked closed- and open-ended questions regarding behavioral reasons
for relinquishment, with scores ranging from 0 to 5 for the following behaviors
(0 being no influence, 5 being a very strong influence): feces/urine in the home,
destruction in the home, escaping the home/yard, excessive barking, fearfulness,
barking/growling at people, barking/growling at dogs, nipping/biting people,
nipping/biting dogs, excessive excitement, and leash pulling on walks. For the
two questions, owners also had a choice to fill in “other” and rate that choice.
Statistical Analysis
Statistical analyses were performed via computerized statistical software (STATA
9, Statacorp, College Station, TX). Ordinal data, including the mean responses
from LAPS, satisfaction with the behavior of the dog, some demographic data,
and types of training methods utilized were evaluated using a Mann-Whitney
ATTACHMENT AND BEHAVIOR 173
rank sum test. Chi-square and Fisher’s exact tests were utilized to evaluate
categorical data. All tests were two-tailed, and significance was set at p < .05.
RESULTS
TABLE 1
Canine Demographic Information
Location
City shelter 10 (8%) 5 (6%) 7 (14%)
County shelter 18 (14%) 4 (6%) 14 (28%)
SPCA 99 (78%) 70 (88%) 29 (58%)
Sex
Male intact 28 (22%) 22 (28%) 6 (12%)
Male castrated 47 (36%) 24 (30%) 23 (47%)
Female intact 20 (15%) 13 (16%) 7 (14%)
Female spayed 32 (25%) 20 (25%) 12 (25%)
Unknown 2 (2%) 1 (1%) 1 (2%)
Weight
<30 lbs 56 (44%) 28 (35%) 28 (57%)
>30 lbs 70 (54%) 49 (61%) 21 (43%)
Unknown 3 (2%) 3 (4%) 0
Age (years)
Mean z* D 2.1, p D .04 3.66 (SE D 0.3) 4.14 (SE D 0.418) 2.85 (SE D 0.367)
Mixed vs. Purebred
Mixed breed 81 (63%) 53 (66%) 28 (57%)
Purebred 45 (35%) 24 (30%) 21 (43%)
Unknown 3 (2%) 3 (4%) 0
Attended Training Classes? n D 122 n D 75 n D 47
Yes 25 (20%) 14 (19%) 11 (23%)
No 97 (40%) 61 (81%) 36 (77%)
Length of Ownership (years)
Mean 2.93 (SE D 0.298) 3.29 (SE D 0.42) 2.34 (SE D 0.37)
Laps
Mean z* D 2.85, p D .006 2.98 (SE D 0.048) 2.88 (SE D 0.068) 3.14 (SE D 0.052)
Note. Pound equivalents: <13.6 kg and >13.6 kg. LAPS D Lexington Attachment to Pets Scale;
SPCA D Society for the Prevention of Cruelty to Animals.
* D significant differences between relinquished and owned dogs.
TABLE 2
Owner Demographic Information
Location n D 129 n D 80 n D 49
City shelter 12 (9%) 5 (6%) 7 (14%)
County shelter 18 (14%) 4 (6%) 14 (28%)
SPCA 99 (77%) 70 (88%) 29 (58%)
Age n D 123 n D 76 n D 47
19–24 15 (12%) 6 (8%) 9 (19%)
25–40 45 (37%) 30 (39%) 15 (32%)
41–60 50 (41%) 29 (39%) 21 (45%)
>60 13 (10%) 11 (14%) 2 (4%)
Gender n D 128 n D 79 n D 48
Male 50 (39%) 36 (46%) 14 (29%)
Female 78 (61%) 43 (54%) 34 (71%)
Marital Status n D 115 n D 71 n D 44
Never married 22 (19%) 9 (13%) 13 (30%)
Married/Domestic partnership 63 (55%) 41 (58%) 22 (50%)
Separated 8 (7%) 5 (7%) 3 (7%)
Divorced 18 (16%) 13 (18%) 5 (11%)
Widowed 4 (3%) 3 (4%) 1 (2%)
Number of Children n D 121 n D 76 n D 45
Mean (˙SE) 1.14 (SE D 0.11) 1.28 (SE D 0.14) 0.91 (SE D 0.176)
Median 1 (range 0–5) 1 (range 0–5) 0 (range 0–4)
Ethnicity n D 129 n D 76 n D 53
Of Hispanic ethnicity 18 (14%) 10 (13%) 8 (30%)
Not of Hispanic ethnicity 103 (86%) 66 (87%) 37 (70%)
Primary Race n D 122 n D 76 n D 46
White 89 (72%) 59 (78%) 30 (65%)
Asian 22 (18%) 10 (13%) 11 (24%)
Black 9 (7%) 5 (7%) 4 ((9%)
Hawaiian/Pacific Islander 2 (2%) 1 (1%) 1 (2%)
American Indian/Alaskan Native 1 (1%) 1 (1%) 0
Annual Income n D 94 n D 57 n D 37
<$10,000 11 (11%) 9 (16%) 2 (5%)
$10,000–29,999 16 (17%) 12 (21%) 4 (11%)
$30,000–49,999 16 (17%) 8 (14%) 8 (22%)
$50,000–69,999 10 (11%) 7 (12%) 3 (8%)
$70,000–89,999 21 (22%) 12 (21%) 9 (24%)
$90,000–109,999 8 (9%) 4 (7%) 4 (11%)
$110,000 or greater 12 (13%) 5 (9%) 7 (19%)
(continued )
176 KWAN AND BAIN
TABLE 2
(Continued)
Note. GED D general education development; SPCA D Society for the Prevention of Cruelty
to Animals.
TABLE 3
Breeds of Dogs as Reported by Owners
Chihuahua
Purebred 8 5 3
Mixed 14 8 6
Pit bull-type
Purebred 12 10 2
Mixed 7 5 2
Labrador retriever
Purebred 6 4 2
Mixed 4 4 0
German shepherd
Purebred 2 2 0
Mixed 7 6 1
Golden retriever
Purebred 4 0 4
Mixed 2 2 0
Note. All other breeds, purebred and owner-selected mixes of those breeds, totaled 5 or less.
If the owners stated that their dogs were mixed breeds, the breeds in the table were the first ones
written by the owners.
ATTACHMENT AND BEHAVIOR 177
very satisfied with their dogs’ behavior (2 [1, n D 75] D 5.59, p D .018), and
they were no more likely to state that behavior problems (specifically aggression)
were reasons for relinquishment than the owners of castrated dogs. Taking into
consideration all owners (relinquishing and continuing), those who used collars
that are primarily used via punishment-based methods (choke and prong collars)
reported less satisfaction with their dogs’ overall behavior and leash-walking
behavior, respectively (z D 2.26, p D .024; z D 2.2, p D .03).
Taking into consideration all owners, owners of Pit bull-type or Chihuahua
dogs were no more likely to state that they were less than very satisfied with
their dogs’ behavior compared with all other breeds combined, respectively (z D
0.85, p D .39; z D 1.4, p D .16).
When asked to assess behavioral problems as a reason for dog relinquishment
on a scale from 0 to 5 (0 being no influence, 5 being a very high influence), for
those for whom we received complete data from these questions (80 of 84), 52
of 80 relinquishing owners (65%) selected scores from 1 to 5, indicating some
problem behavior had some role in their decision to relinquish their dogs. When
assessing stronger influences of these behavior problems with scores from 3 to
5, 39 of 80 owners (49%) indicated that at least one problematic behavior was
a relatively strong influence on their decision to relinquish.
With regard to aggression (barking/growling and nipping/biting at people
and/or dogs), 38 of the relinquishing owners (48%) selected scores from 1 to
5, indicating that aggression had some role in their decision to relinquish their
dogs. When assessing a stronger influence of aggression with scores from 3 to
5, 23 owners (29%) indicated that aggression was a relatively strong influence
on their decision to relinquish. Reasons for relinquishment are summarized in
Table 4.
Regarding other pets in the household, there were no differences in the mean
number of cats or dogs in the households of those who relinquished their dogs
compared with those who did not (cats: z D 0.32, p D .75; dogs: z D 0.147,
p D .88), and there were no differences regarding whether they had a cat or dog
in the household (cats: 2 [1, n D129] D 0.2, p D .6; dogs: 2 [1, n D129] D 1.4,
p D .23). In evaluating those relinquishing owners in which moving had strongly
influenced their reason for relinquishment (n D 27), 48% of them still had at
least one other dog remaining in their household, which was not significantly
different from those who stated that moving had not strongly influence their
decision (2 [1, n D80] D 0.1, p D .75).
We also determined that finances did not strongly influence reasons for
relinquishment, as 42% of relinquishing owners had at least one other dog
remaining in their household, which was not significantly different from those
who had stated that finances did not strongly influence their decision (2 [1,
n D79] D 0.03, p D .86). Data on income reported by owners can be found in
Table 5.
178 KWAN AND BAIN
TABLE 4
Reasons for Relinquishment (n D 80)
Somewhat
Not a Reason of a Reason Strong Reason
Reason (Score D 0) (Scores D 1–5) (Scores D 3–5) M (SE) Median
DISCUSSION
In relation to our first hypothesis, the results of our study show that owners
relinquishing their dogs were significantly less attached to the dogs. Additionally,
they rated their dogs’ behavior as less than perfect compared with continuing
dog owners. We acknowledge that few dogs in any setting are perfectly behaved
in all aspects. Although this result may seem intuitive, our results imply that
TABLE 5
Owner Income Levels
<$10,000 7 2
$10,000–29,999 6 6
$30,000–49,999 4 4
$50,000–69,999 3 4
$70,000–89,999 5 7
$90,000–109,999 0 4
$110,000 or greater 1 4
Note. Owners indicated if their decision to relinquish their dogs was influenced by veterinary
costs, food and other animal costs, unemployment or other cuts in income by responding to the
question, “How much did each of the following influence your decision to relinquish this dog to the
shelter?” (n D 57).
ATTACHMENT AND BEHAVIOR 179
there was a break in the owner-pet bond or there was not a significant bond to
begin with for these relinquished dogs compared with those of continuing dog
owners. However, we cannot neglect the high likelihood that most, if not all, of
these owners struggled with the difficult decision to relinquish their dogs and
they might not have had adequate resources to help them during the time that
they owned their dogs (DiGiacomo, Arluke, & Patronek, 1998).
Our second hypothesis, that owners of relinquished dogs were less likely
to have attended training classes compared with continuing owners, was not
upheld. One study failed to show a relationship between problem behaviors and
attendance at obedience training classes (Voith, Wright, & Danneman, 1992).
However, these data may have been skewed by two separate factors: the low
number of owners who took their dogs to training classes and the fact that
owners might not have taken the specific dogs in the study to training classes
due to training and experience through the ownership of another dog, which was
information we did not solicit (Shore, Burdsal, & Douglas, 2008).
Our third hypothesis, that relinquishing owners were more likely to have used
punishment-based techniques/tools than continuing owners, was not upheld. Al-
though there was not a difference in the frequency of use with punishment-based
collars in either the relinquishing or continuing groups, the use of punishment-
based methods of training, including use of choke and prong collars, was
associated with greater dissatisfaction with their dogs’ leash-walking behavior,
which is consistent with another study (Hiby et al., 2004).
Whether owners decide to use these types of tools because of the behavior
of their dogs or whether the use of these types of tools is at least in part a cause
of problematic behavior remains to be determined. If we encourage owners to
select dogs based on compatible behaviors, provide an appropriate environment,
and train using positive reinforcement methods, we may be able to enhance the
human-animal bond and decrease chances of relinquishment to shelters.
Our study partly supported another study in demonstrating a correlation
between owner-reported inappropriate behaviors and relinquishment of dogs to
shelters (Salman et al., 2000; Salman et al., 1998; Scarlett, Salman, New, & Kass,
1999). It is important that we continue to realize the serious effects that problem
behaviors have on the human-animal bond. We did not collect information on
the behavior of the dog who remained in the house, so we cannot state whether
the relinquished dog was more ill-behaved compared with the remaining dog(s).
Financial concerns ranked near the top, which was likely the result of the
serious financial recession that had overtaken the United States for the prior few
years, affecting job stability and income. Compared with the rest of the country
during the time of data collection, the Sacramento area was greatly affected by
high unemployment rate (12.5% during the summer of 2010), furloughs of state
government employees, and a high number of home foreclosures (U.S. Depart-
ment of Labor, n.d.). Although no systematic study has been done regarding
180 KWAN AND BAIN
the specific effects of the current financial crisis on relinquishment of dogs and
cats to shelters, a review (Nowicki, 2011) summarized the many news reports
and statements by shelter managers regarding the increase of relinquished pets
to shelters. One such statement was from the director of the SPCA, who stated
that there was a 20% increase in relinquishments, identifying foreclosures as the
main reason for this increase (Waters, 2008).
There was a study conducted that evaluated the effects of the financial crisis as
part of the reason for the sharp increase in the number of unwanted horses in the
United States (Holcomb, Stull, & Kass, 2010). However, owners who selected
finances or moving as reasons for relinquishing their dogs were as likely to
have another dog remaining in their household as those who did not select those
reasons for relinquishment. This underscores the likelihood that other reasons,
such as behavioral problems, were at least contributing to the owners’ reasons
to relinquish their dogs.
We recognize that convenience sampling is not the most appropriate manner
in which to get complete data. A power analysis was not performed a priori
as solicitation for surveys was limited by the researcher’s time availability.
Some owners of very aggressive-appearing dogs were not interviewed because
the interviewer would have had to hold the dog, often Pit bull-type dogs,
which could have been a safety issue at vaccination clinics. We are aware that
this could have affected our result of Pit bull-type dogs being more likely to
be in the relinquished group. However, Pit bull-type dogs are reported to be
overrepresented in the population of shelter dogs (Brettman, 2008; DogsBite.org,
2009; K. Hurley, personal communication, October 2, 2012). Our result that
owners of Pit bull-type dogs were no more likely to state that they were less
than very satisfied with their dogs’ behavior could have been affected by avoiding
interviewing owners holding aggressive dogs.
It would have enhanced the study if owners who requested medical euthanasia
for their dogs had been surveyed so that we could fully assess whether there
were other reasons for relinquishment aside from medical issues. However, we
decided against this so as to not interfere with the owners’ grief.
We understand that these data cannot be representative of all populations of
dogs in the country or even in Sacramento. We attempted to control between the
two groups by soliciting owners attending vaccination clinics at the respective
shelters in an assumption that if these owners were to relinquish their dogs,
they would return to that particular shelter. However, we cannot predict that this
would be the case.
We interviewed two owners who were non-English speakers. In 2010, the
percentage of residents in Sacramento County born outside of the United States
was 20% (Social Explorer, n.d.). Certainly a lot of them speak fluent English,
but approximately 10% of Sacramento-area residents born outside of the United
States self-reported that they “speak English less than very well” (CensusScope,
ATTACHMENT AND BEHAVIOR 181
n.d.). This could have affected our results in relation to reasons for relinquish-
ment, owner attachment, and behavioral data. Additionally, owners may be less
than truthful when filling out a survey during a potentially emotional time
(Segurson et al., 2005). However, by stating in the surveys that confidentiality
would be maintained and placing them in closed envelopes at some shelters, we
attempted to avoid this issue as much as possible.
To get a more wide-ranging pool of people to fill out the survey, it would be
ideal for shelter employees to distribute the survey to every owner relinquishing
his or her dog for any reason and to every owner attending vaccination clinics.
Without an interviewer present to collect the data regularly, we run the risk of
decreased owner compliance and an increased chance of surveys being filled out
incompletely, which we determined during the pilot testing phase.
As with all studies that evaluate correlation, we recognize that there are mul-
tiple factors that play a role in a companion animal’s behavior and the decision
to relinquish a pet. We chose to evaluate specific parameters but acknowledge
that there are many other factors that play a role in these issues.
CONCLUSION
Within the limitations of this study, these findings show that the attachment to
their dogs was less for relinquishing owners than for continuing owners. It must
be stated that there are multiple factors that play a role in their lower attachment.
One factor that was elucidated in this study was that owners relinquishing their
dogs more frequently stated that they were less than very satisfied with their
dogs’ behavior and that the behaviors of their dogs affected their decision
to relinquish them. As a society we should take a multifactorial approach to
help keep dogs in their homes, which includes helping owners solve problem
behaviors of their dogs with tools that are utilized in humane training.
ACKNOWLEDGMENTS
REFERENCES
Alexander, M. B., Friend, T., & Haug, L. (2011). Obedience training effects on search dog perfor-
mance. Applied Animal Behaviour Science, 132, 152–159.
Brettman, A. (2008). Pit bulls overpopulate shelters, but adoptions are elusive. Retrieved from http://
www.oregonlive.com/environment/index.ssf/2008/05/pit_bulls_overpopulate_shelter.html
CensusScope. (n.d.). Ability to speak English among those speaking a language other than English,
2000. Retrieved from http://www.censusscope.org/us/m6920/chart_language.html
Clark, G. I., & Boyer, W. N. (1993). The effects of dog obedience training and behavioural
counselling upon the human-canine relationship. Applied Animal Behaviour Science, 37, 147–
159.
Diesel, G., Brodbelt, D., & Pfeiffer, D. U. (2009). Characteristics of relinquished dogs and their
owners at 14 rehoming centers in the United Kingdom. Journal of Applied Animal Welfare Science,
13, 15–30.
DiGiacomo, N., Arluke, A., & Patronek, G. (1998). Surrendering pets to shelters: The relinquisher’s
perspective. Anthrozoös: A Multidisciplinary Journal of the Interactions of People & Animals,
11(1), 41–51.
DogsBite.org (2009). 2009 U.S. shelter data: Pit bulls account for 58% of dogs euthanized. Retrieved
from http://blog.dogsbite.org/2009/08/2009-us-shelter-data-shows-that-pit.html
Duffy, D. L., Hsu, Y., & Serpell, J. A. (2008). Breed differences in canine aggression. Applied
Animal Behaviour Science, 114, 441–460.
Duxbury, M. M., Jackson, J. A., Line, S. W., & Anderson, R. K. (2003). Evaluation of association
between retention in the home and attendance at puppy socialization classes. Journal of the
American Veterinary Medical Association, 223(1), 61–66.
Haverbeke, A., Laporte, B., Depiereux, E., Giffroy, J. M., & Diederich, C. (2008). Training methods
of military dog handlers and their effects on the team’s performances. Applied Animal Behaviour
Science, 113, 110–122.
Hays, L. D. (2004). Effects of a standardized obedience program on approachability and problem be-
haviors in dogs from rescue shelters (Unpublished doctoral dissertation, Texas A&M University).
Retrieved from http://handle.tamu.edu/1969.1/1261
Helms, T. D., & Bain, M. J. (2009). Evaluation of owner attachment to dogs on the basis of whether
owners are legally considered guardians of their pets. Journal of the American Veterinary Medical
Association, 234, 896–900.
Herron, M. E., Shofer, F. S., & Reisner, I. R. (2008). Retrospective evaluation of the effects of
diazepam in dogs with anxiety-related behavior problems. Journal of the American Veterinary
Medical Association, 233, 1420–1424.
Hiby, E. F., Rooney, N. J., & Bradshaw, J. W. S. (2004). Dog training methods: Their use,
effectiveness and interaction with behaviour and welfare. Animal Welfare, 13, 63–69.
Holcomb, K. E., Stull, C. L., & Kass, P. H. (2010). Unwanted horses: The role of nonprofit equine
rescue and sanctuary organizations. Journal of Animal Science, 88, 4142–4150.
Jagoe, A., & Serpell, J. (1996). Owner characteristics and interactions and the prevalence of canine
behaviour problems. Applied Animal Behaviour Science, 47, 31–42.
Johnson, T. P., Garrity, T. F., & Stallones, L. (1992). Psychometric evaluation of the Lexington
Attachment to Pets Scale (LAPS). Anthrozoös: A Multidisciplinary Journal of the Interactions of
People & Animals, 5, 160–175.
Kobelt, A. J., Hemsworth, P. H., Barnett, J. L., & Coleman, G. J. (2003). A survey of dog ownership
in suburban Australia: Conditions and behaviour problems. Applied Animal Behaviour Science,
82, 137–148.
Melson, G. F. (1990). Studying children’s attachment to their pets: A conceptual and methodological
review. Anthrozoös: A Multidisciplinary Journal of the Interactions of People and Animals, 4, 91–
99.
ATTACHMENT AND BEHAVIOR 183
National Council on Pet Population Study and Policy. (1997). Shelter statistics survey. Retrieved
from www.petpopulation.org/statsurvey.html
New, J. C., Salman, M. D., King, M., Scarlett, J. M., Kass, P. H., & Hutchison, J. M. (2000).
Characteristics of shelter-relinquished animals and their owners compared with animals and their
owners in U.S. pet-owning households. Journal of Applied Animal Welfare Science 3, 179–201.
Nowicki, S. A. (2011). Give me shelter: The foreclosure crisis and its effect on America’s Animals.
Stanford Journal of Animal Law & Policy, 4, 97–121.
Reisner, I. R., Houpt, K. A., & Shofer, F. S. (2005). National survey of owner-directed aggression in
English springer spaniels. Journal of the American Veterinary Medical Association, 227, 1594–
1603.
Rockwood, C., & Bain, M. (2007, July). Canine training methods: Association with obedience, prob-
lem behaviors, and dog-owner relationships. In J. Ciribassi & D. Horwitz (Chairs), ACVB/AVSAB
scientific paper and poster session (p. 93). The Washington, DC: ACVB/AVSAB.
Salman, M. D., Hutchison, J., Ruch-Gallie, R., Kogan, L., New, J. C., Kass, P. H., & Scarlett, J. M.
(2000). Behavioral reasons for relinquishment of dogs and cats to 12 shelters. Journal of Applied
Animal Welfare Science, 3, 93–106.
Salman, M. D., New, J. G., Scarlett, J. M., Kass, P. H., Ruche-Gallie, R., & Hetts, S. (1998). Human
and animal factors related to relinquishment of dogs and cats in 12 selected animal shelters in
the United States. Journal of Applied Animal Welfare Science, 1, 207–226.
Scarlett, J. M., Salman, M. D., New, J. G., & Kass, P. H. (1999). Reasons for relinquishment
of companion animals in U.S. animal shelters: Selected health and personal issues. Journal of
Applied Animal Welfare Science, 2, 41–57.
Schilder, M. B. H., & van der Borg, J. A. M. (2004). Training dogs with help of the shock collar:
Short and long term behavioural effects. Applied Animal Behaviour Science, 85, 319–334.
Segurson, S. A., Serpell, J. A., & Hart, B. L. (2005). Evaluation of a behavioral assessment
questionnaire for use in the characterization of behavioral problems of dogs relinquished to animal
shelters. Journal of the American Veterinary Medical Association, 227, 1755–1761.
Serpell, J. A. (1996). Evidence for an association between pet behavior and owner attachment levels.
Applied Animal Behaviour Science, 47, 49–60.
Shore, E. R., Burdsal, C., & Douglas, K.(2008). Pet owners’ views of pet behavior problems and
willingness to consult experts for assistance. Journal of Applied Animal Welfare Science, 11,
63–73.
Sigler, L. (1991). Pet behavioral problems present opportunities for practitioners. AAHA Trends, 4,
44–45.
Social Explorer. (n.d.). Place of birth for the foreign-born population. Retrieved from http://www.
socialexplorer.com/pub/reportdata/home.aspx
Turcsán, B., Kubinyi, E., & Miklósi, A. (2011). Trainability and boldness traits differ between dog
breed clusters based on conventional breed categories and genetic relatedness. Applied Animal
Behaviour Science, 132, 61–70.
U.S. Census Bureau. (2010). Overview of race and Hispanic origin: 2010. Retrieved from http://
www.census.gov/prod/cen2010/briefs/c2010br-02.pdf
U.S. Department of Labor. (n.d.). Sacramento; Arden-Arcade; Roseville, CA metropolitan area
statistics. Retrieved from www.bls.gov/news.release/archives/metro_09012010.pdf
Voith, V. L., Wright, J. C., & Danneman, P. J. (1992). Is there a relationship between canine
behavior problems and spoiling activities, anthropomorphism, and obedience training? Applied
Animal Behaviour Science, 34, 263–272.
Waters, J. (2008). Foreclosing on Fido: Financially overwhelmed owners abandon homes, with
pets still inside. Retrieved from http://www.marketwatch.com/story/fido-foreclosed-overwhelmed-
owners-abandon-homes-pets-and-all
The Pet Professional Guild
Position Statement on the Use of Choke and Prong Collars
It is the position of the Pet Professional Guild that and long-term psychological damage to animals.
effective animal training procedures lay the
Common problems which can result from the use of
foundation for an animal’s healthy socialization and
choke and prong collars:
training and helps prevent behavior problems. The
general pet-owning public should be educated by Physical problems
organizations and associations to ensure pet animals
live in nurturing and stable environments to better While precise information on the potential damage
prevent behavior problems and help ensure the caused by the use of choke and prong collars is still being
overall well-being of the animal. Consistent with this collected, there are many cases of dogs suffering soft
effort, it is the position of the Pet Professional Guild tissue damage, eye problems, strangulation (leading to
(PPG) that the use of collars and leads that are death), tracheal/esophageal damage and neurological
intended to apply constriction, pressure, pain or problems resulting from the use of choke/prong collars.
force around a dog’s neck (such as choke chains and Training Fallout
prong collars) should be avoided.
Choke chains and prong collars are designed to
Though data demonstrating the exact damage that administer negative reinforcement and positive
can be potentially caused by using choke and prong punishment. Training
collars is incomplete, experience has shown that soft techniques based in these
tissue injuries are common and, as is the case with two learning theory
any harsh training method, the damage to the quadrants are prone to side
animal-human relationship results. Studies and the effects. As an example, a
experience of the PPG’s membership finds that dog wearing a choke or
training and behavior problems are consistently and prong collar that fearfully
effectively solved without the use of choke or prong barks and lunges at another dog would then be choked
collars with the alternative methods reinforcing the or pain inflicted by the prong collar. The pain and
animal-human bond. Evidence indicates that rather choking then adds to the negative association the dog
than speeding the learning process, harsh training wearing the collar has with other dogs. This is the polar
methods actually slow the training process, add to opposite of what an ideal training protocol is designed to
the animal’s stress and can result in both short-term accomplish.
use of flat buckle collars, head halters, harnesses and Anderson DR, Davis EB. Sensitivities of ocular tissues to acute
other types of control equipment that are safer for the pressure-induced ischemia. Arch Ophthalmol 1975;94:267-274.
animal.
Piette S, Liebman JM, Ishikawa H, et al. Acute conformational
changes in the optic nerve head with rapid intraocular pressure
Sources
elevation: implications for LASIK surgery. Ophthalmic Surg Lasers
Pauli, A., et al, Effects of the Application of Neck Pressure by a Imaging 2003;34:334-341.
Collar or Harness on Intraocular Pressure in Dogs. J Am Anim
Hamor RE, Gerding PA, Ramsey DT, et al. Evaluation of short-term
Hosp Assoc 2006;42:207-211.
increased intraocular pressure on flash- and pattern-generated
Gum GG, Gelatt KN, Ofri R. Physiology of the eye. In: Gelatt KN, electroretinograms of dogs. Am J Vet Res 2000;61:1087-1091.
ed. Veterinary Ophthalmology. 3rd ed. Philadelphia: Lippincott
Siliprandi R, Bucci MG, Canella R, et al. Flash and pattern
Williams & Wilkins, 1999:165-167.
electroretinograms during and after acute intraocular pressure
Slatter D. Fundamentals of Veterinary Ophthalmology. 3rd ed. elevation in cats. Invest Ophthalmol Vis Sci 1988;29:558-565.
Philadelphia: WB Saunders, 2001.
© Copyright 2012-2017 Pet Professional Guild. All
Schuman JS, Massicotte EC, Connolly S, et al. Increased rights reserved. If quoting any part of this article,
intraocular pressure and visual field defects in high resistance please respect our copyright and attribute it to the Pet
wind instrument players. Ophthalmology 2000;107:127-133.
Professional Guild (2012-2017) and include a link back
Strubbe DT, Gelatt KN. Ophthalmic examination and diagnostic to the original article on the PPG website.
procedures. In: Gelatt KN, ed. Veterinary Ophthalmology. 3rd ed.
https://www.petprofessionalguild.com/chokeandprong
Philadelphia: Lippincott Williams & Wilkins, 1999:427-466.
collarpositionstatement
Teng C, Gurses-Ozden R, Liebmann JM, et al. Effect of a tight
necktie on intraocular pressure. Br J Ophthal 2003;87:946-952.
SUMMARY RÉSUMÉ
of acute stress in animals [1, 11]. According to BEERDA [1], Two test instructors were present during the entire experiment.
non-invasive sampling methods must be used to assess stress The main responsibilities of the test instructors were the ob-
in animals since those methods have minimal influence on the servation and control of the test sessions. Additionally, two
results. He further emphasized the necessity of using a phy- experienced canine officers who were also professional dog
siological parameter concomitant with behavioural parameters trainers took part in the study as decoys. The main responsi-
in order to support the interpretation of the behavioural data bility of the decoys was provoking the dogs during the test
when assessing stress. Therefore salivary cortisol values and sessions in order that the dogs make a mistake. Another res-
direct behavioural reactions of the dogs were evaluated as ponsibility of the decoys during this study was the adminis-
stress parameters in the present study. Several researchers re- tration of the electronic training collar. During the sessions in
ported a number of different behavioural indicators of acute which the electronic training collar was tested, they held the
stress in dogs, which comprise lowering and arching of the receiver of the collar and gave the electric impulse whenever
body [1, 7, 18], flattening ears [1, 7, 18] and lowering the tail the dog made the mistake. Each of the helpers provoked one
and/or holding the tail tightly between the legs [1, 18] when group during the whole experiment. The aim of using the
the dogs are confronted with the aversive situation. In the present same person as decoy for all dogs in the same group was to
study, the aim of the test design was to determine acute stress minimize the variability arising from the provocation style
arising from the training methods. Thus, above mentioned be- and, also, to the decoy himself.
havioural elements were evaluated as acute stress parameters. Each training method used in this research required a proper
The aim of this study was to investigate whether any stress training aid. Dogtra 600 NCP/2® electronic training collar,
is caused by the use of specific conditioned signal, quitting Klickstachelhalsung® pinch collar, the standard normal collar
signal, and/or pinch collars and if they do so, whether the and the 5 m long leash were used as training aids.
stress produced in the process is comparable to the one with
electric training collars. Corrections made by pinch collar and
electronic training collar were considered as representatives EXPERIMENTAL DESIGN AND TESTS
of the positive punishment while correction made by the quitting
In the current literature, it was cited that there are diffe-
signal was considered as the application of the negative pu-
rences in response to aversive stimulus between individuals
nishment. In this context, positive punishment refers to an ap-
[8, 19]. LINDSAY [12] additionally suggested that in order
plication of an aversive stimulus as a consequence of an
to amplify the statistical results in studies in which the elec-
undesired behaviour and negative punishment connotes the tronic training collar was tested, “within subject design’’ must
prevention or withholding of the delivery of an appetitive sti- be applied since individual variables can cause incorrect data.
mulus in case that the undesired behaviour occurs [15]. We Therefore, in the present study ‘’within subject design’’ was
set out to investigate the direct behavioural reactions and sa- applied as experimental design for comparison of stress and
livary cortisol values of the dogs upon administration of above learning effects of the different training methods. Namely,
mentioned training methods. We were especially interested in each training method was tested on each dog on different days
finding out which method leads to less stress in dogs during during the research. In order to eliminate the effects of the ad-
training with high level of arousal. Furthermore, this study ministration orders of the training methods on the results, six
examined the learning effects of the electronic training collar, subgroups, A, B, C, D, E, F, were established, to each of them
the pinch collar and the quitting signal. Thus, the compatibility a different administration order of the training methods was
of the learning effect of ‘’negative punishment’’ method with applied (randomized cross-over design, Table 1).
the ‘’positive punishment’’ method, in a training with high
level of arousal and motivation were also be assessed.
Subgroups 1st test day 2nd test day 3rd test day
A Q E P
Material and Methods B Q P E
SUBJECTS AND TEST PERSONS C E Q P
D E P Q
Forty-two adult police dogs of both genders (33 males and
E P Q E
9 females) and varying ages (3-10 years old) of the breed Ma-
F P E Q
linois (Shorthaired Belgian Sheepdog) served as subjects for
Q: Quitting signal, P: Pinch Collar, E: Electronic Training Collar.
this study. All dogs in the study were official police service
dogs and recruited from two different police departments in
Germany. The decision to use only Belgian Malinois was em- TABLE I: Cross-over design of the study conducted in order to compare
ployed in an attempt to avoid the variability due to breed cha- 3 different learning methods (electronic training collar, pinch collar
racteristics. During the study, dogs participated the sessions and quitting signal) in Belgian Malinois police dogs.
with their own handlers.
All tests were conducted on open air training grounds which
were also used for routine police dog training. In order to obtain Prior to the main experiment, since 7 dogs had never been
the standardization in respect to the test area, each dog was trained with the electronic training collar before, an adaptation
tested on the same place where it started to be tested during phase, which lasted six weeks, was conducted for them. For
the entire experiment. the adaptation phase, the dogs carried the electronic training
collars during the normal daily training routine. The same pro-
cedure was applied for the rest of the dogs for a week since
they were already familiar to the electronic training collar.
This procedure was conducted in order to achieve habituation
of the dogs to the device again since the electronic training
collars are forbidden in Germany since 2006.
Quitting signal is a conditioned signal which evokes feeling
of frustration in dogs since it has a meaning of withdrawal of
the reward. The main principle of the quitting signal is to
condition a feeling of frustration, and thus, to abandon of a
distinctive behaviour towards a specific signal. In order to par-
ticipate in the experiment, the dog should withdraw itself from
the toy immediately after the first instruction of the signal.
Before the main experiment started, quitting signal training
procedure was completed and the signal was tested on each dog. FIGURE 1: Test situation for comparing 3 different learning methods (elec-
tronic training collar, pinch collar and quitting signal) in Belgian Ma-
The main experiment took three test days for each dog. The linois police dogs.
time interval between test days was one week. On each test
day, a different training method was tested on the dog. Maximal
three main test sessions were conducted for each dog per day
dog abandoned the undesired behaviour reliably after the cor-
to test the concerning training method. Learning effect of training
rection, the same test procedure was repeated after an hour in
methods was evaluated considering these three sessions. Re-
order to see whether the method had a learning effect. During
gardless of which method was tested, each dog carried three
the repetition of the test, the same procedure as in the first test
collars around its neck, which were standard, pinch and elec-
was carried out. The decoy did exactly the same provocation
tronic training collars, during the entire experiment in order
against the dog. If the dog did not repeat the same mistake,
to ensure the standardization among the training methods. The
the test session was terminated and it was noted that the method
dogs were brought to the training area with a leash on standard
had a learning effect. In case that the dog showed a reaction
collar and were kept on the leash throughout the entire expe-
against the decoy again, the test was repeated after an hour
riment.
for the last time. If the dog did not abandon the undesired be-
Before conducting the main test, two different sessions were haviour reliably after the first correction, the handler and the
performed with each dog, which were obedience and play ses- dog left the test area and no repetition session was conducted.
sions. During the obedience session, the dog did some obe-
dience exercises for eighty seconds. This session brings the
dog to a certain level of arousal, so that the optimal results MEASUREMENTS
could be achieved at the main test. After the obedience session,
a standardized play session, which lasted for forty seconds, Salivary cortisol measurements
was carried on. In the play session, the handler played freely Salivary cortisol concentration was measured as one of the
with his/her dog. The goal of performing the play session between stress parameters. The saliva samples were obtained from the
the obedience- and test sessions was to avoid misevaluation dogs with cotton buds manufactured by Salivettten® der Firma
of extra-stress caused by frustration. At the end of the two minutes, Sarstedt AG & C. Each dog handler took the saliva sample from
the dog and its handler came to the determined point at which his/her own dog. In order to stimulate the secretion of saliva,
they took up the heel position. After the dog and its owner citric acid (Amos Vital Vitamin C Pulver der Firma Amos
took up the heel position with their backs turned to the entry Vital GmbH) had been put into the dogs’ mouths before the
of the test area, the decoy with a protection sleeve and a whip
saliva samples were collected. Following samples were mea-
in his hand entered the test area. As soon as the handler took
sured during the research:
his position up at a distance of approximately 3 m from the
- Resting samples: Two samples were taken with one hour
dog and the handler, the owner gave the “Heel” command to
interval in a home environment.
his/her dog and started to walk by the decoy (figure 1). From
- Basal samples: The same procedure as in the main expe-
that moment on, the decoy tried to provoke the dog in order
riment without any intrusion of the decoy was conducted. The
that it made a mistake. As the dog made the mistake, the trai-
dogs were brought to the training ground. The decoy stood
ning method which would be tested was administered. As testing
still at a distance of approximately 3 m from the dog and its
the quitting signal, the handler used a 5 m long leash. The first
owner. However, the owner was not allowed to apply any pu-
reason of using 5 m long leash for the quitting signal was to
nishment to the dog if it made any mistake. The saliva samples
allow reaction time for the dog handler and also for the dog
were collected at 5, 10 and 15 minutes after the dogs leaving
and, thus, to be able to evaluate clearly whether the dog stopped
the training ground and the cortisol values were recorded as
due to the influence of the collar or to the signal. The second
basic values.
reason was to be sure that the dog could not reach the decoy
- Experimental samples: The saliva samples were collected
and, thus, it could not get rewarded by being able to catch him.
at 5, 10 and 15 minutes after the administration of the concer-
As mentioned before, the training method which would be ning training method. The investigation of the samples was
tested was administered as the dog made the mistake. If the carried out in the laboratory of pharmacology and toxicology
Behavioural observations
The entire experiment was filmed on DVDs using a video
camera. The recorded DVDs were reviewed later in order to
analyze the direct behavioural reactions of the dogs after the
administration of the training methods. The direct reactions
of the dogs were evaluated using an extensive ethogram. One-
zero sampling method was used in order to assess the direct
behavioural reaction of the dog upon the administration of the
above mentioned methods. FIGURE 2: Learning effects of the 3 different learning methods (electronic
training collar (ETC), pinch collar (PC) and quitting signal (QS)) in
Belgian Malinois police dogs.
STATISTICAL ANALYSIS
Data analysis was performed with SPSS 16.0 Inc. software. 64.3% of the dogs exhibited this typical reaction after corrected
Two significance levels were set at the levels 95% (P<0.05**) by the pinch collar, but the difference between the 2 methods
and 99% (P<0.01*). Kruskal-Wallis tests were used for the was not statistically significant. No statistically significant dif-
comparison of the learning effect of the training methods bet- ference was found in comparison for the tail reactions (lowering)
ween groups and subgroups. Learning effects between the between the electronic training collar and the pinch collar (t-test,
training methods as well as the saliva cortisol concentrations P = 0,165). Comparing the first joint reactions of the dogs to
were analyzed by paired-sample t-tests. Frequency analyses the pinch- and the electronic training-collar, it was found that
have also been used for the detection of direct behavioural ef- the correction applied by the pinch collar caused lower body
fects of training methods. posture than the one applied by the electronic training collar
(figure 3). Moreover, 4.8% of the dogs exhibited “extreme lo-
wering of body posture” as a reaction to the pinch collar, while
this reaction was observed in none of the dogs using the elec-
Results tronic training collar. Dogs elicited vocalizations more fre-
LEARNING EFFECTS quently as a reaction to the electronic training collar than to
the pinch collar (P < 0.0001) (figure 4). As mentioned above,
As shown in figure 2, with electronic training collar, the me- only 4 dogs out of 42 subjects abandoned the behaviour after
thod had learning effect on 39 of 42 dogs. In other words, at receiving the quitting signal during the first session. Therefore,
the end of the experiment the wrong behaviour was abandoned only the reactions of these 4 dogs were tested. Consequently,
in 39 out of 42 dogs. With pinch collar, learning effect was it was observed that 2 dogs showed “backward ear position”
obtained on 32 of 42 dogs, that is, 32 dogs abandoned the mis- and 1 dog showed “extreme lowering of body posture” together
behaviour at the end of pinch collar part of the experiment. with the “crouching” after getting the correction.
Four dogs could be tested for the learning effect of the quitting
signal since the other 38 dogs did not reliably quit the beha-
viour after the instruction of the signal. In fact, the signal had
learning effect on only 3 dogs out of 42 subjects. The learning
effect was maximal with electronic training collar, not signi-
ficantly compared to the pinch collar (P = 0.160) but signifi-
cantly compared to the quitting signal (P < 0.0001). In
parallel, the learning effects of the pinch collar were also si-
gnificantly higher than those obtained with the quitting signal
(P < 0.0001). Considering learning effect of the training me-
thods, no significant difference was found between subgroups.
total of 17 dogs showed maximal cortisol values after the ins- pinch collar, it is difficult to achieve a sensitive adjustment,
truction of the quitting signal, while 15 and 10 dogs exhibited as factors such as strength and motivation of the handler are
maximal hormone concentrations after administration of signal essential for the effectiveness of the mechanical training de-
with electronic training collar and with pinch collar, respectively vices [10, 12, 20]. In addition to that, the pinch collar had to
(figure 5). To have a better comparison, relative cortisol values be administered by the dog handlers themselves. That obligation
established as difference between maximum cortisol values might also cause different results since the ability of each dog
and resting cortisol values were compared. Both relative cortisol handler in timing and application of the method was different.
values after the administration of the pinch collar (P = 0.0004) An explanation for the low learning effect of quitting signal
and the electronic training collar (P = 0.0065) were signifi- is similar to that of pinch collar. Since the dog handler applied
cantly lower than the relative basal value. No statistically si- the method, the optimal timing could not be achieved for each
gnificant difference was found between the relative basal dog. Moreover, as using the conditioned quitting signal, feeling
cortisol value and the relative cortisol value after the instruc- of frustration upon application of the signal might not be as
tion of the quitting signal. However, no significant difference strong as the feeling arised during the provocation of the
was found between electronic training collar and pinch collar decoy in most of the dogs.
or quitting signal (P = 0.2006 and P = 0.1782, respectively)
One of the goals of the present study was to compare the
while the relative cortisol value after the application of quit-
direct behavioural reactions of the dogs to three different training
ting signal was significantly higher than after the application
methods. Considering the body posture and ear positions,
of pinch collar (P = 0.0294). As a result, the highest cortisol
pinch collars seem to induce more behavioural reactions, in
concentrations were measured after the instruction of the quitting
the form of distress, than the electronic training collar However,
signal (figure 5). In addition, considering the relative cortisol
the maximal cortisol values after the administration of elec-
values, no significant difference was found between the sub-
tronic training collar were higher, but not significantly, than
groups.
that after application of pinch collar. This contradiction between
physiological and behavioural measures support the hypothesis
that behavioural responses are not always concomitant to phy-
Discussion siological parameters in case of exposing to stress [5, 21]. Fur-
thermore, it might also support the claim that in a study
The results of the present study showed that the electronic
involving highly exciting training sessions such as police dog
training collar as well as the pinch collar was convenient as
training, the behavioural data is more reliable than the phy-
training aids to cease the unwanted behaviour. The electronic
siological measures such as cortisol concentrations and heart-
training collar was, however, found to be more effective in
rate frequency [17].
comparison to the pinch collar. By contrast, the desirable learning
effect could not be achieved in application of the quitting signal. Within the frame of the study it was detected that more
The reason of obtaining high learning effect with electronic vocal reactions were elicited by the electronic training collar
training collars might be due to the fact that the device fulfils than by the pinch collars. The explanation of more vocaliza-
the essential punishment criteria such as timing and sensitive tions with electronic training collar might be that these vocal
adjustment, which the learning theories require [1, 6, 14, 16, reactions are elicited as “startle reactions” [2] rather than
20]. In case an appropriate device is chosen, the electronic “pain-induced vocalizations”. Visibility of the punishment as
training collar gives possibility of making a sensitive adjust- in the application of the pinch collar can cause that the dog
ment for the level of punishment. In the present study, the dose reacts in different ways according to the cues given by the
of the stimulus was reliably adjusted for each dog by two pro- handler subconsciously, such as facial expression, body lan-
ficient dog trainers. These dog trainers were also responsible guage, holding style of the leash etc.. That is, the application
for administration of the electronic training collar. Thus, the of the pinch collar as well as of the quitting signal is directly
correct timing was obtained. On the other hand, while using linked to the handler since the correction made by the owner
FIGURE 4: Vocal reactions to the electronic training collar (ETC) and the FIGURE 5: Number of the dogs with maximal cortisol values per learning
pinch collar (PC) in Belgian Malinois police dogs. method (electronic training collar (ETC), pinch collar (PC) and quitting
signal (QS)).
is visible to the dog. Furthermore, the feeling of the leash on 2. - BROOM D.M., JOHNSON K.G.: Stress and animal welfare, BROOM
the pinch collar could be a signal for the dog as forthcoming D.M. and JOHNSON K.G (ed), 1st edition, Chapman and Hall Publi-
shing, TJ Press, Great Britain, 1993, pp.: 88.
punishment, whereas no signal could be perceived by the dog
3. - BSAVA’s POSITION STATEMENT: Electronic training devices: a be-
while testing the electronic training collar. All those associations havioural perspective. J. Small Anim. Pract., 2003, 44, 95-96.
the dog makes between the punishment and the owner are un- 4. - CHRİSTİANSEN F.O., BAKKEN M., BRAADSTADT B.O.: Beha-
desirable in training since the dog can show a submissive vioural changes and aversive conditioning in hunting dogs by the se-
communication with its owner if it perceives that the punish- cond year confrontation with domestic sheep. Appl. Anim. Behav. Sci.,
ment comes from him/her. On the other hand, as applying the 2001, 72, 131-143.
electronic training collar, if exact timing is obtained, the dog 5. - CRONİN G.M., HEMSWOTH P.H., BARNETT J.L., JONGMAN
E.C.J., NEWMAN E.A., Mc CAULEY I.: An anti-barking muzzle for
makes association only between the situation and the effect. dogs and its short-term effects on behaviour and saliva cortisol concen-
In fact, this might be another explanation of the high learning trations. Appl. Anim. Behav. Sci., 2003, 83, 215-226.
effect of the electronic training collar. 6. - DESS N.K., LINWICK D., PATTERSON J., OVERMEIER J.B.: Im-
mediate and proactive effects of controllability and predictability on
Considering the maximum cortisol concentrations, it was plasma cortisol response to shocks in dogs. Behav. Neurosci., 1983,
evaluated that most of the dogs exhibited maximal values after 97, 1005-1016.
the instruction of the quitting signal. It was further detected 7. - FEDDERSEN-PETERSEN D., OHL F.: Ausdrucksverhalten beim
that even the dogs on which the quitting signal had learning Hund: Mimik und Körpersprache, Kommunikation und Verstaendi-
effect exhibited clear stress related behaviours. This point is gung, BECK A. (ed), Franckh-Kosmos Verlags-GmbH, Stuttgart,
2008, pp.: 172.
noteworthy since this result raises an important question about
8. - JACQUES J., MYERS S.: Electronic training devices: a review of cur-
whether the psychic stressors such as frustration and uncer- rent literature. Anim. Behav. Consult.: Theor. Pract., 2007, 3, 22-39.
tainty produce more stress than the physical exposures in animals. 9. - KLEIN D: Telereizgeraete, Sachkunde zur Anwendung in der Hun-
The significant difference found between the quitting signal deausbildung, 1st edition, Books on demand Gmbh, Germany, 2003,
and the pinch collar supports this hypothesis, at least for the pp.: 38.
dogs of Malinois breed. Considering the relative cortisol values, 10. - LANDSBERG G., HUNTHAUSEN W., ACKERMANN L.: Hand-
the basal values were significantly higher than the values ob- book of behaviour problems of the dog and cat. LANDSBERG G.,
HUNTHAUSEN W. and ACKERMANN L. (eds), 2nd edition, Saun-
tained using the electronic training collar and the pinch collar. ders, Edinburgh, London, 2003, pp.: 64.
As mentioned above, no correction was applied on dogs even
11. - LAY D.C. Jr., FRIEND T.H., BOWERS C.L., GRISSOM K.K., JEN-
though they made a mistake before the collection of basic KINS O.C.: A comparative study of physiological and behavioural
value. This uncertainty might cause a high stress just as the study of freeze and hot-iron branding using dairy cows. J. Anim. Sci.,
feeling of frustration during the use of quitting signal. 1992, 70, 1121-1125.
12. - LINDSAY S.R.: Handbook of applied dog behaviour and training.
As a conclusion, in the present study, it was found that the Vol 3, Procedures and Protocols, LINDSAY S.R. (ed.), Iowa State
electronic training collar had higher learning effect and induced University Press, Ames, 2005, pp.: 557-634.
less stress to cease the unwanted behaviour in comparison to 13. - OVERALL K.L.: Why electric shock is not behaviour modification?.
the other training methods in a situation with high motivation. J. Vet. Behav. Clin. Appl. Res., 2007, 2, 1-4.
However, for achieving this result, it is essential to prove the 14. - POLSKY R.H.: Electronic shock collars: Are they worth the risks?
J. Am. Anim. Hosp. Assoc., 1994, 30, 463-468.
administrator’s practical and theoretical knowledge. Overall,
15. - REID P.: Learning in dogs. In: The behavioural biology of dogs, JEN-
the debates over training methods should include not only the SEN P. (ed.), CABI publishing, 2007, pp.: 120-144.
specific training aids but also their significance for animal
16. - SCHALKE E., STICHNOTH J., OTT S., JONES-BAADE R.: Cli-
welfare prospect should be covered. The qualifications of pro- nical signs caused by the use of electric training collars on dogs in
fessional dog trainers such as practical and theoretical know- everyday life situations. Appl. Anim. Behav. Sci., 2007, 105, 369-380.
ledge requirements that trainers must fulfil should also be 17. - SCHILDER M.B.H., VAN DER BORG J.A.M.: Training dogs with
considered in new discussions. Therefore, we strongly recom- help of the shock collar: short and long term behavioural effects.
mend to put more emphasize on qualification of the trainer Appl. Anim. Behav. Sci., 2004, 85, 319-334.
when assessing the effectiveness of training methods. We finally 18. - SCHWIZGEBEL D.: Zusammenhänge zwischen dem Verhalten des
deutschen Schäferhundes im Hinblick auf tiergerechte Ausbildung.
recommend that further studies should be performed to inves- In: Aktuelle Arbeiten zur artgemäßen Tierhaltung, 1982, KTBL-
tigate whether the same findings could be achieved with the Schrift 291, pp.: 138-148.
dogs of other breeds. 19. - SHEPPARD G., MILLS D.S.: The development of a psychometric
scale for the evaluation of the emotional predispositions of pet dogs.
Internat. J. Comp. Psychol., 2002, 15, 201-222.
20. - TORTORA D.F.: Understanding electronic dog-training. 2nd edition,
References Tri-Tronics Inc., Tucson, Arizona, 1982, pp.: 21-32.
1. - BEERDA B.: Stress and well-beings in dogs. Thesis (PhD), Universiteit 21. - VINCENT I.C., MICHELL A.R.: Comparison of cortisol concentra-
Utrecht, Netherlands, 1997, pp.: 29-44. tions in saliva and plasma of dogs. Res. Vet. Sci., 1992, 53, 342-345.
RESEARCH
KEYWORDS: Abstract In a questionnaire survey of dog owners, 88% of respondents’ dogs had received some form
dog behavior; of training. Training methods varied; 16% of owners said that they used only positive reinforcement,
training; 12% used a combination of positive reinforcement and negative reinforcement, 32% used a combina-
reinforcement; tion of positive reinforcement and positive punishment, and the remaining 40% used a combination of
punishment; all categories. Seventy-two percent of owners used some form of positive punishment. The mean num-
behavior problems; ber of potentially undesirable behaviors reported was 11.3 per dog. Attendance at formal training clas-
undesirable behaviors ses did not significantly affect the total number of potentially undesirable behaviors reported. However,
dogs that had attended puppy socialization classes were less likely to show an undesirable reaction to
dogs from outside the household, and owners who carried out informal training at home, but did not
attend any form of formal training class, were more likely to report some form of aggression in their
dog. The training method used by owners was also related to the total number of potentially undesir-
able behaviors shown by the dogs. When individual categories of potentially undesirable behavior were
investigated, the type of training method used was also significantly associated with attention-seeking
score, fear (avoidance) score, and aggression score. Other factors related to the overall number of
potentially undesirable behaviors included the age and origin of the dog.
Ó 2008 Published by Elsevier Inc.
undesirable behaviors (Voith et al., 1992; Jagoe and Serpell, only, or a combination of both. The potential negative
1996; O’Farrell, 1997; Hiby et al., 2004; Bennett and effects on behavior of using aversive training techniques
Rohlf, 2007) provides evidence as to which factors influ- have also been identified in other studies (Schilder et al.,
ence the development and persistence of undesirable behav- 2004).
iors and is therefore fundamental in improving rational Given that the majority of pet dogs receive at least some
prevention and treatment strategies. form of basic training to make them manageable compan-
It is widely accepted that the development of behavior in ions, and that there is some indication that the methods used
any species is influenced by both genetic and environmental may relate to the prevalence of undesirable behaviors, the
factors. Although genetic factors clearly predispose indi- aim of this study was to investigate further the relationship
vidual dogs to develop particular behavioral phenotypes between types of training methods, attendance at different
(Overall et al., 2006), environmental factors also have a categories of training classes, and the occurrence of poten-
profound effect (Appleby et al., 2002). The domestic dog tially undesirable behaviors in domestic dogs.
is clearly a highly social species capable of complex com-
munication with people (Hare and Tomasello, 2005) and
able to associate even subtle visual signals from their Materials and methods
owners with positive or negative outcomes (Rooney et al.,
2001; Cullinan et al., 2004). It is therefore perhaps not sur- A convenience sample of dog owners was recruited for
prising that differences in owners’ global approach to han- this survey. Dog owners out walking their dogs, or visiting
dling and training appears to have an influence on the a veterinary hospital in several different locations in the
occurrence of undesirable behaviors in their dogs (Hiby United Kingdom (Shropshire, Somerset, and Leicester-
et al., 2004). There have, however, been mixed findings shire) were asked to complete a standard questionnaire.
as to which specific aspects of ‘‘owner behavior’’ are im- The questionnaire obtained basic demographic data about
portant in terms of the development of behavior problems the dog and owner, information about the type of training
(Voith et al., 1992; O’Farrell, 1997; Jagoe and Serpell, conducted with the dog (e.g., training carried out infor-
1996; Atkinson et al., 2005). The relationship between pre- mally at home or attendance at training classes such as
vious dog ownership, or owners’ general knowledge about puppy socialization, agility, obedience, or show handling),
dogs, and occurrence of undesirable behaviors has also and the specific training techniques used (e.g., sound
varied between studies (Borchelt and Voith, 1982; Jagoe distraction, verbal praise, shutting away, telling off). In
and Serpell, 1996; Kobelt et al., 2003; Bennett and Rohlf, addition, the questionnaire asked owners the frequency
2007). with which their dog showed 36 behaviors that are
Some of the differences in reported findings may clearly commonly perceived to be undesirable, on a scale of 0
be accounted for by methodological differences between to 4, where 0 was ‘‘never’’ and 4 was ‘‘frequently.’’ The
studies. For example, there are often distinct differences in individual behaviors were described as objectively as
owner reports between questionnaires containing objective possible (e.g., ‘‘Does your dog not come back when
(e.g., ‘‘Does your dog withdraw when unfamiliar people called on a walk?’’ or ‘‘Does your dog chew or destroy
approach?’’) or more subjective (e.g., ‘‘Is your dog fearful items when you are out of the house?’’) to avoid differ-
of other dogs?’’) types of questions. However, these mixed ences in owner perception of more subjective concepts,
findings may also be indicative that more subtle aspects of such as ‘‘disobedience’’ or ‘‘anxiety.’’ Where one of these
owner behavior may interact with global effects of owner behaviors occurred in their dog, respondents were also
attitude or training. For example, the consistency, or asked to rate whether they perceived the behavior to be a
otherwise, of an owner’s behavior toward his or her dog problem.
within a single context is also related to the occurrence of A total of 250 questionnaires were given out, and 192
undesirable behaviors (Cullinan et al., 2004). In addition, completed questionnaires were returned. Sixty-seven per-
the effect of an ‘‘anthropomorphic attitude’’ is likely to cent of respondents were female. The age of the owner
be complex; such an attitude can lead to owners both under- ranged from under 20 years to over 60, with the majority of
and overestimating the cognitive abilities of their pets participants falling into the 41-60 category. Most of the
(Bradshaw and Casey, 2007). dogs in the sample lived in households with 2 adults (71%
Previous studies have reported an association between of households), and there were no children in 43% of the
reduced prevalence of undesirable behaviors in pet dogs households. The total number of dogs in the sample
and attendance at obedience training classes (Clark and households ranged from 1 to 5 dogs. Thirty-eight percent
Boyer, 1993; Jagoe and Serpell, 1996) or engagement of dogs lived in a house with 1 other dog, 6.8% lived with 2
with any form of training (Kobelt et al., 2003; Bennett other dogs, 2% lived with 3, and 2% lived with 4 other
and Rohlf, 2007). In addition, Hiby et al (2004) found dogs. Single-dog households made up the remaining 52%
significantly fewer behavior problems in dogs that had of the sample. Most respondents had owned dogs before
been trained using rewards only, as compared with dogs their current dog, with first-time dog owners making up just
that had been trained using some form of punishment 15% of the sample.
Blackwell et al Training methods and behavior problems in dogs 209
The dogs included in the survey ranged in age from 1 to of any underlying ‘‘temperament’’ or ‘‘behavioral strategy’’
15 years, with a median age of 5 years. The ratio of male to existing across contexts were created by summating the
female dogs in the survey was roughly equal, with 52% contexts in which 3 common behavioral responses (aggres-
males and 48% females. Fifty-eight percent of the males sion, avoidance, and attention seeking) were shown and
and 63% of the females were neutered. Purebred dogs then dividing by number of contexts measured for each.
dominated the sample population, with only 19% of Second, the presence or absence of any behavioral
respondents owning cross-breeds. All of the kennel club response within specific contexts (e.g., being left alone
breed groups were represented in the sample. Gun dogs by the owner) or toward specific targets (e.g., reaction to
were the largest group with 35% of the dogs being gun dog other dogs outside the household) was calculated. A ‘‘con-
breeds. The next largest group was the pastoral breeds trol score’’ was also developed from the mean number of
(18%), followed by terriers (12%), hounds (8%), toy breeds contexts in which dogs showed goal-directed behaviors
(3%), and breeds from the utility group (2%). that are generally inhibited through effective training
The majority of dogs (71%) were acquired by the current techniques (e.g., not coming back when called). The
owner within the first 3 months of the dogs’ lives. Dogs individual undesirable behaviors contributing to each
acquired by their current owner at ages over 3 months were category are listed in Table 2. Nonparametric tests were
obtained predominantly from either a private home or from a used for the analysis, as the data were not normally distrib-
re-homing center. Most dogs (72%) were acquired by the uted. Relationships between behavioral scores and contex-
owners from the person who bred the dog, 20% of owners got tual categories were compared with attendance at training
their dogs from re-homing centers, and 2% of owners bred the classes and the use of different training methods using
dogs themselves. Finally, 6% of owners acquired their dogs Spearman rank correlations, Mann-Whitney U tests, chi-
from other sources, such as pet shops, family, and friends. square tests, and Kruskal-Wallis tests.
shown in Table 3. When combined into categories, the mean Eighteen percent of owners had sought help for unde-
‘‘aggression score’’ across all dogs was 0.22, the mean ‘‘fear sirable behaviors. Of these, 32% had asked a vet for advice,
(avoidance) score’’ was 0.27, the mean ‘‘attention-seeking 9% a vet nurse, 26% an animal behaviorist, 47% a dog
score’’ was 0.39, and the mean ‘‘control’’ score was 0.43. trainer, and 9% a friend or relative. Help had also been
Sixty-five dogs (34%) showed a behavioral response in the sought from rescue centers, breeders, and books. Owners
context of being left alone (i.e., separation-related behav- were found to be most likely to seek advice from someone
ior); 15% (n 5 28) showed an undesirable behavioral re- if they were experiencing control problems with their dog,
sponse to other dogs within the household; 99 dogs (52%) aggression to unfamiliar dogs, aggression to unfamiliar
showed an undesirable response to dogs outside the house- people, and jumping up.
hold; 25 (13%) showed an undesirable response toward fam-
ily members; 153 (80%) showed an undesirable response to
unfamiliar people; and 13% of dogs (n 5 25) showed an un- Relationship between training classes attended
desirable response when told off. and occurrence of undesirable behaviors
When owners whose dog showed a potentially undesir-
able behavior were asked whether they considered the In this study, attendance at any type of training class did
behavior that their dog displayed to be a ‘‘problem,’’ 76 % not significantly affect the total number of undesirable
of owners considered the expression of at least 1 potentially behaviors shown by dogs. However, attendance at puppy
undesirable behavior by their dog to be problematic to socialization classes was associated with reduced reaction
them. The most common behaviors considered a problem to dogs from outside the household (c2 5 6.843, df 5 1,
by owners whose dogs showed them (Table 4) were differ- P , 0.01). An association was also found between owners
ent from those occurring most commonly in the population, doing informal training at home and not attending any
although there was a significant positive correlation training classes, with an increased aggression score
between owners rating an undesirable behavior as occurring (Mann-Whitney U, z 5 21.974, P , 0.05). Interestingly,
‘‘frequently’’ and considering it to be ‘‘problematic’’ the control problem score did not appear to be affected
(Spearman’s rho 5 0.462, n 5 192, P , 0.01). by attendance at any type of training class.
Table 4 Behaviors most commonly considered to be problems by owners whose dogs showed them
Aggression score
all types
positive reinforcement and punishment
positive and negative reinforcement
positive reinforcement only
all types
0 5 10 15 20 25 30 35
Percentage of dogs
% reacting to other dogs % reacting to unfamiliar people
Figure 3 Percentage of dogs showing an undesirable response to unfamiliar people and dogs outside the house in different categories of
training method used by owners.
Other influences on occurrence of undesirable P , 0.05) and control problems (Spearman’s rho
behaviors 20.293, P , 0.01). However, older dogs were more likely
to show an undesirable reaction to other dogs outside
Age of dog the home (Mann-Whitney U, Z 5 22.356, P , 0.05),
A significant, although weak, negative correlation was and unfamiliar people (Mann-Whitney U, Z 5 22.279,
found between dog age and total number of undesirable P , 0.05).
behaviors (Spearman’s rho 20.235, P 5 0.01). When
compared with the different categories of behavior, dog Breed of dog
age was found to be significantly negatively correlated No significant effect of breed type was found on total
with attention-seeking score (Spearman’s rho 20.389, number of undesirable behaviors, control problem score,
P , 0.01), but not fear (avoidance) score or aggression avoidance score, or attention-seeking score. The number of
score. Younger dogs were more likely to show separa- dogs in the utility (n 5 4) and toy (n 5 5) breed groups
tion-related behaviors (Mann-Whitney U, Z 5 22.691, were too low to allow for further analysis.
Avoidance score
Aggression score
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Mean score
Did not use punishment Used punishment
Figure 4 Effect of whether owners used any positive punishment methods at all on undesirable behavior scores in dogs.
214 Journal of Veterinary Behavior, Vol 3, No 5, September/October 2008
classes may provide increased opportunities for social people and to dogs from outside the household were
interaction, facilitating the development of appropriate significantly associated with training type. In both cases,
social communication with unfamiliar dogs. the use of positive reinforcement alone, or in combination
Most owners reported using some kind of positive with negative reinforcement, was significantly associated
reinforcement when interacting with their dogs, particularly with a lower reporting of undesirable response, compared
verbal praise and stroking. However, only 16% of owners to where owners used methods involving positive punish-
use purely positive reinforcements in their training, similar ment. The finding supports that of Roll and Unshelm
to that found by Hiby and colleagues (2004). When combi- (1997), who found a link between punishment-based train-
nations of training category used were compared with ing and dog–dog aggression. This result may suggest that
undesirable behaviors, the use of positive reinforcement the use of punishment techniques in this context increases
alone was significantly associated with a lower number of behaviors associated with anxiety. However, since this is
undesirable behaviors, as also found by Hiby and col- a correlational finding, this result may have also arisen be-
leagues (2004). These results, therefore, confirm the finding cause owners whose dogs show these behaviors are more
that the use of punishment-based techniques when training likely to resort to punishment.
dogs is associated with a higher number of undesirable Factors other than training methods were also found to
behaviors, although this is obviously not necessarily a influence undesirable behaviors. As has been found in other
causal link. It is possible that increased undesirable behav- studies (e.g., Bradshaw et al., 2002), younger dogs were
iors occur through dogs associating punishment either with more likely to show a higher number of undesirable behav-
the person carrying it out, or the context in which it occurs, iors. When examined in more detail, young dogs appeared
rather than (as is intended) their own behavioral response. to have higher attention-seeking scores and were more
This inadvertent learning can lead to fear or anxiety devel- likely to have control problems and show separation-related
oping in association with the handler or context. Alterna- behavior. It makes sense for young dogs to show more
tively, it may be simply that owners of dogs that show a attention-seeking behaviors and control problems, since
larger number of undesirable behaviors are more likely to behavioral inhibition is generally greater in adults than ju-
resort to using punishment. veniles. This is the case because the immediate behavioral
The use of training methods involving positive rein- response to emotions in young animals becomes increas-
forcement only were also significantly associated with a ingly inhibited by pathways from higher cortical centers
lower attention-seeking score, aggression score, and fear as the latter develop with age and experience (Panksepp,
(avoidance) score, but not control problem score. This 2003). An older population of dogs may also partially rep-
finding tends to support the hypothesis that punishment- resent a ‘‘residual population’’ of animals who have not
based training techniques are related to the development of been relinquished or euthanized for their behavior, and/or
undesirable behavioral traits, since it would be expected whose owners have learned to tolerate or live with their un-
that punishment in response to the occurrence of undesir- desirable behaviors. In contrast, an undesirable response to
able behaviors would apply to control problems as much as unfamiliar people and dogs outside the household was
other behavioral responses. The highest mean attention- higher in older dogs. Since these behaviors often become
seeking score was found in dogs whose owners use a reinforced over time, either because they are ‘‘successful’’
combination of positive reinforcement and negative rein- in resolving the situation for the individual, or because of
forcement. This finding is consistent with what might be the owner’s response, an increase with age is not unexpected.
expected, since the variable reinforcement of attention- There was no real effect of breed type on undesirable
seeking responses is likely to make them more persistent behavior shown. The gender of dogs was also unrelated to
over time. However, since the most prevalent technique measures of undesirable behavior, as found in other studies
used within the ‘‘negative reinforcement’’ category was (e.g., Bradshaw et al., 2002; Bennett and Rohlf, 2007).
‘‘shutting away,’’ it is also possible that owners resort to this Neuter status was also not significant in the study reported
technique in response to attention-seeking behaviors from here, in contrast to Bennett and Rohlf (2007), who found
their dogs. The highest aggression score was found in those that owners rated ‘‘anxiety/destructiveness’’ and ‘‘nervous-
dogs whose owners used a combination of positive rein- ness’’ differently for neutered and entire dogs. This differ-
forcement and positive punishment. This finding may ence may be accounted for by methodological
support the hypothesis that aggressive responses in dogs differences, as Bennett and Rohlf (2007) measured owner
can develop as a result of ‘‘conflict,’’ or anxiety about an perceptions rather than specific behaviors.
uncertain response to their behaviors from inconsistent The origin of dogs appeared to have a significant effect
owners (Cullinan et al., 2004). Alternatively, again, the on undesirable behaviors in this study, in contrast to Brad-
findings may reflect the types of responses that owners shaw and colleagues (2002). The number of undesirable
try in response to aggressive behaviors in their dogs. behaviors was lower in dogs from a breeder than those
When the contexts in which undesirable behaviors from rescue centers. This may be a reflection of the num-
occurred were compared with category of training method bers relinquished to rescue centers for behavior problems
used by owners, only an undesirable response to unfamiliar (Bailey, 1992; Miller et al., 1996; Patronek et al., 1996;
216 Journal of Veterinary Behavior, Vol 3, No 5, September/October 2008
Serpell, 1996; Salman et al., 1998, 2000; Scarlett et al., behaviors, training engagement and shared activities. Appl. Anim.
1999; New et al., 2000; Marston and Bennett, 2003; Shore Behav. Sci. 102, 65-84.
Blackwell, E., Casey, R.A., Bradshaw, J.W.S., 2006a. Controlled trial of
et al., 2003; Mondelli et al., 2004; Shore, 2005). In partic- behavioral therapy for separation-related disorders in dogs. Vet. Rec.
ular, the occurrence of separation-related behavior was 158, 551-554.
higher in dogs from a rescue center. This result supports find- Blackwell, E., Casey, R.A., Bradshaw, J.W.S., 2006b. The prevalence of
ings from some other studies (Guthrie, 1999), but not others noise fears and phobias in pet dogs in the UK. Proceedings of 49th
(Bradshaw et al., 2002), where the population included a Annual Congress of British Small Animal Veterinary Association,
Birmingham, UK, p. 490.
larger proportion of dogs obtained from re-homing centers. Borchelt, P.L., Voith, V.L., 1982. Classification of animal behavior prob-
In contrast to other studies (Jagoe and Serpell, 1996; lems. Vet. Clin. North Am. Small Anim. Pract. 12, 571-585.
Kobelt et al., 2003; Bennett and Rohlf, 2007), but in agree- Bradshaw, J.W.S., Casey, R.A., 2007. Anthropomorphism and anthropo-
ment with Borchelt and Voith (1982), no relationship was centrism as influences in the quality of life of companion animals.
found between first-time dog ownership and the occurrence Anim. Welf. 16(S), 149-155.
Bradshaw, J.W.S., McPherson, J.A., Casey, R.A., Larter, I.S., 2002. Aeti-
of potentially undesirable behaviors. This finding may be ology of separation-related behavior in domestic dogs. Vet. Rec.
related to the relatively small proportion (15%) of respon- 151, 43-46.
dents in this study who were first-time dog owners. Casey, R.A., Bradshaw, J.W.S., 2001. A comparison of referred feline clin-
The number of children in households had no significant ical behavior cases with general population prevalence data. Scientific
effect on total number of undesirable behaviors or any of Proceedings of the British Small Animal Veterinary Association
Congress, Birmingham, UK, p. 529.
the undesirable behavior scores, in contrast to Kobelt and Clark, G.I., Boyer, M.N., 1993. The effects of dog obedience training and
colleagues (1993), who reported a link between the pres- behavioral councelling upon the human-canine relationship. Appl.
ence of children in a household and the occurrence of Anim. Behav. Sci. 37, 147-159.
‘‘overexcitability’’ of dogs. The mean number of children in Cullinan, P., Blackwell, E. J., Casey, R. A., 2004. The relationships
the household was significantly lower, however, where dogs between owner consistency and ‘problem’ behaviors in dogs: a prelim-
inary study. Proceedings of 1st meeting of the European College of
showed separation-related behavior. This finding may be Veterinary Behavioral MedicinedCompanion Animals. Cremona,
owing to dogs being less dependent on their owner’s Italy. 22nd October, 2004.
attention in households where parents’ attention is directed Guthrie, A., 1999. Dogs behaving badlydcanine separation disorder re-
toward children rather than dogs, or it may be that dogs search. Vet. Pract. 31, 12-13.
living in households with children are less likely to be left Hare, B., Tomasello, M., 2005. Human-like social skills in dogs? Trends
Cogn. Sci. 9, 439-444.
alone. The increased food stealing with increasing numbers Hart, L.A., 1995. Dogs as human companions: a review of the relationship.
of children probably reflects the greater opportunity for In: Serpell, J.A. (Ed.), The Domestic Dog: Its Evolution, Behavior and
successfully getting food from children, or the increased Interactions with People. Cambridge University Press, Cambridge,
availability of discarded food in households that include UK, pp. 161-178.
children. Hiby, E.F., Rooney, N.J., Bradshaw, J.W.S., 2004. Dog training methods:
their use, effectiveness and interaction with behavior and welfare.
Anim. Welf. 13(1), 63-69.
Jagoe, A., Serpell, J., 1996. Owner characteristics and interactions and the
Conclusions prevalence of canine behavior problems. Appl. Anim. Behav. Sci. 47,
31-42.
In conclusion, it seems that there is an association between Kobelt, A.J., Hemsworth, P.H., Barnett, J.L., Coleman, G.J., 2003. A sur-
a lower number of potentially undesirable behaviors vey of dog ownership in suburban Australiadconditions and behavior
problems. Appl. Anim. Behav. Sci. 82, 137-148.
reported in dogs trained without the use of punishment-
Landsburg, G., Ruehl, W., 1997. Geriatric behavioral problems. Vet. Clin.
based techniques. However, attendance at formal training North Am. Small. Anim. Pract. 27, 537-1559.
classes appears to have little effect on the overall occur- Marston, L.C., Bennett, P.C., 2003. Reforging the bonddtowards success-
rence of undesirable behaviors in adult dogs. ful canine adoption. Appl. Anim. Behav. Sci. 83, 227-245.
Miller, D.D., Staats, S.R., Partlo, C., Rada, K., 1996. Factors associated
with the decision to surrender a pet to an animal shelter. J. Am. Vet.
Med. Assoc. 209, 738-742.
References Mondelli, F., Previde, E.P., Verga, M., Levi, D., Magistrelli, S.,
Valsecchi, P., 2004. The bond that never developed: adoption and re-
Appleby, D., Bradshaw, J.W.S., Casey, R.A., 2002. Relationship between linquishment of dogs in a rescue shelter. J. Appl. Anim. Behav. Sci. 7,
aggressive and avoidance behavior by dogs and their experience in 253-266.
the first six months of life. Vet. Rec. 150, 434-438. New, J.C., Salman, M.D., King, M., Scarlett, J.M., Kass, P.H.,
Atkinson, T., Casey, R. A., Bradshaw, J. W. S., 2005. Are the amount and Hutchinson, J.M., 2000. Characteristics of shelter-relinquished ani-
type of exercise taken, and proximity to owners at night, related to the mals and their owners compared with animals and their owners in
incidence of behavior problems in the domestic dog? Proceedings of U.S. pet-owning households. J. Appl. Anim. Welf. Sci. 3, 179-201.
the Companion Animal Behavior Therapy Study Group Study Day, O’Farrell, V., 1997. Owner attitudes and dog behavior problems. Appl.
Birmingham, 2005. Anim. Behav. Sci. 52, 205-213.
Bailey, G., 1992. Parting with a pet survey. Blue Cross Publication, Blue Overall, K.L., Hamilton, S.P., Chang, M.L., 2006. Understanding the ge-
Cross, Burford, Oxon, U.K. netic basis of canine anxiety: phenotyping dogs for behavioral, neuro-
Bennett, P.C., Rohlf, V.I., 2007. Owner-companion dog interactions: rela- chemical and genetic assessment. J. Vet. Behav. Clin. Appl. Res. 1,
tionships between demographic variables and potentially problematic 124-141.
Blackwell et al Training methods and behavior problems in dogs 217
Panksepp, J., 2003. At the interface of the affective, behavioral, and cog- Schilder, M.B.H., van der Borg, J.A.M., 2004. Training dogs with the help
nitive neurosciences: Decoding the emotional feelings of the brain. of the shock collar: short and long term behavioral effects. Appl.
Brain Cogn. 52, 4-14. Anim. Behav. Sci. 85, 3-4, 319-334.
Patronek, G.J., Glicklman, L.T., Beck, A.M., McCabe, G.P., Ecker, C., Serpell, J.A., 1996. Evidence for an association between pet behavior and
1996. Risk factors for relinquishment of dogs to an animal shelter. owner attachment levels. Appl. Anim. Behav. Sci. 47, 49-60.
J. Am. Vet. Med. Assoc. 209, 572-581. Shore, E.R., Petersen, C.L., Douglas, D.K., 2003. Moving as a reason for
Roll, A., Unshelm, J., 1997. Aggressive conflicts amongst dogs and factors pet relinquishment: a closer look. J. Appl. Anim. Welf. Sci. 6, 39-52.
affecting them. Appl. Anim. Behav. Sci. 52, 229-242. Shore, E.R., 2005. Returning a recently adopted companion animal:
Rooney, N.J., Bradshaw, J.W.S., Robinson, I.H., 2001. Do dogs respond to adopters’ reasons for and reactions to the failed adoption experience.
play signals given by humans? Anim. Behav. 61, 715-722. J. Appl. Anim. Welf. Sci. 8, 187-198.
Salman, M.D., New, J.C., Scarlett, J.M., Kass, P.H., 1998. Human and animal Voith, V.L., Wright, J.C., Danneman, P.J., 1992. Is there a relationship be-
factors related to the relinquishment of dogs and cats in selected animal tween canine behavior problems and spoiling activities, anthropomor-
shelters in the United States. J. Appl. Anim. Welf. Sci. 1, 207-226. phism, and obedience training? Appl. Anim. Behav. Sci. 34, 263-272.
Salman, M.D., Hutchinson, J.M., Ruch-Gallie, R., Kogan, L., New, J.C., Wells, D.L., Hepper, P.G., 2000. Prevalence of behavior problems reported
Kass, P.H., Scarlett, J.M., 2000. Behavioral reasons for relinquishment by owners of dogs purchased from an animal rescue shelter. Appl.
of dogs and cats to 12 shelters. J. Appl. Anim. Welf. Sci. 3, 93-106. Anim. Behav. Sci. 69, 55-65.
Scarlett, J.M., Salman, M.D., New, J.C., Kass, P.H., 1999. Reasons for re- Westbrook, C., Pinchbeck, G. L., Bradshaw, J. W. S., Dawson, S, Gaskell,
linquishment of companion animals in U.S. animal shelters: selected R. M., Christley, R. M., In press. Dog-human and dog-dog interactions
health and personal issues. J. Appl. Anim. Welf. Sci. 2, 41-57. in a UK pet dog community. Vet. Rec.
Journal of Veterinary Behavior 19 (2017) 50e60
Canine Review
a r t i c l e i n f o a b s t r a c t
Article history: The purpose of this study was to review a series of studies (N ¼ 17) regarding the effects of using various
Received 18 March 2016 methods when training dogs. The reviewed studies examined the differences between training methods
Received in revised form (e.g., methods based on positive reinforcement, positive punishment, escape/avoidance, et cetera) on a
13 December 2016
dog’s physiology, welfare, and behavior toward humans and other dogs. The reviewed studies included
Accepted 11 February 2017
Available online 22 February 2017
surveys, observational studies, and interventions. The results show that using aversive training methods
(e.g., positive punishment and negative reinforcement) can jeopardize both the physical and mental
health of dogs. In addition, although positive punishment can be effective, there is no evidence that it is
Keywords:
dog training
more effective than positive reinforcementebased training. In fact, there is some evidence that the
reinforcement opposite is true. A few methodological concerns arose from the reviewed studies. Among them are small
punishment sample sizes, missing data on effect size, possible bias when coding behavior in observational studies,
training methods and the need to publish case reports of bodily damage caused by aversive training methods. In
electronic collar conclusion, those working with or handling dogs should rely on positive reinforcement methods and
avoid using positive punishment and negative reinforcement as much as possible.
Ó 2017 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jveb.2017.02.004
1558-7878/Ó 2017 Elsevier Inc. All rights reserved.
G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60 51
Table
A summary of studies of the effects of aversive training methods on dogs (N ¼ 17, studies are presented by year of publication and name of first author)
Roll and Unshelm, 1997 Dog owners with a dog injured by Questionnaires completed by all Sex Only descriptive statistics reported.
another dog (n ¼ 151) and dog owners to compare aggressors and Breed Dogs that were trained by hitting or
owners with a dog that injured victims. Age “shaking” tended to have a history
another dog (n ¼ 55). Training methods of biting other dogs.
Degree of injury A higher % of victim dogs were
Location of fight owned by dog owners who shouted
Owners’ demographics and gave clear commands and
owners who believed that training
should be fun and that it was
advantageous to have a trained
dog.
A higher % of aggressor dogs were
owned by owners who believed
that a dog would be out of control
without training.
Polsky, 2000 Adult dogs (n ¼ 5). Descriptive data collected in the Dog sex, age, reproductive status, All 5 attacks carried out by adult
form of legal documents, animal breed, location of attack relative to intact males between the ages of 2-
control officers and police officers’ border of containment system, and 3 years.
reports. victim’s familiarity with dog Adult victims were familiar with
Case studies of aggression that was the dog, child victims were not.
elicited by electronic pet In 4 of 5 cases, the dog received a
containment systems were shock at the time of the attack.
examined. Not 1 of the dogs showed
threatening behavior before attack.
All attacks included repeated biting
of victim.
No gross warning signals were
given before biting.
Hiby et al., 2004 Dog owners (n ¼ 326) with dogs Questionnaires given to dog Demographics Using punishment positively
aged more than 1 year. owners as they were walking their Training methods correlated with problem behaviors.
dog or through veterinary clinics. Obedience level No correlation between use of
Questionnaires returned by mail. Problem behaviors reward-based training and
problem behaviors.
Number of behavior problems:
punishment only > combination of
punishment and reward > reward
only or miscellaneous methods.
Punishment-based training was
never the most effective for
achieving obedience goals.
Overall obedience was related to
reward-based but not to
punishment-based methods.
Schilder and van der Dogs that were trained for official Videotaping 107 shocks delivered Direct effect of shocks on body Direct reaction to shocks (e.g.,
Borg, 2004 certificate of police service (n ¼ 15). to 31 dogs and comparing body language. lowering body, high-pitched yelps,
Dogs that trained with (n ¼ 16) and language to control dogs that did Body language of dogs that barks, squeals, redirected
without (n ¼ 15) electric shocks. not receive shocks. received shocks in the past during aggression, avoidance) lasted a
Comparing behavior of dogs that various conditions in which no fraction of a second.
received shocks in the past to dogs shocks were given Long-term effects: comparison of
that did not. shocked (S) dogs and control dogs
(C).
During free walking, obedience,
and police work: lower ear
postures and stress-related
behaviors: S > C.
Differences were seen even when
walking in a park outside the
training grounds.
Stress and lower ear postures:
training > free walking.
Schalke et al., 2007 Laboratory-bred Beagles between Training to stop prey behavior. Heart rate Absolute and relative cortisol
the ages of 1.5-2 years (n ¼ 14). Three groups: (1) receive shock Salivary cortisol levels:
precisely at the moment they touch Group (1)dincrease by w22 and
a prey dummy, (2) receive shock 31%, respectively.
when failing to obey a recall during Group (2)dincrease by w114 and
hunting, (3) receive arbitrary, 160%, respectively.
unpredictable shocks. Group (3)dincrease by w336 and
328%, respectively.
Increase in maximal heart rate in
group (3) on days of training with
shock.
(continued on next page)
G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60 53
Table (continued )
Steiss et al., 2007 Dogs from a private no-kill shelter Dogs randomly assigned to 3 Barking Reduced barking in S and E groups
(n ¼ 21). groups: control (C), electronic Activity compared to C.
antibark collar (E), lemon spray Plasma cortisol No statistically significant changes
antibark collar (S). ACTH in plasma cortisol and ACTH.
Baseline measurements and However, large effect sizes of
measurements after collar was increased plasma cortisol and ACTH
activated. in S and E groups on day 1 of
wearing the active collar compared
to baseline.
Blackwell et al., 2008 Dog owners (n ¼ 192) with dogs Questionnaires given to dog Demographics Training methods: 16% Rþ, 12% Rþ
between the ages of 1-15 years. owners walking their dogs or in Training methods and R, 32% Rþ and Pþ, 40%
veterinary hospitals in the United Obedience level combination of all categories.
Kingdom. Problem behaviors 72% used some Pþ.
Rþ only: lowest score on attention
seeking behaviors, fear, and
aggression.
Rþ and R: highest score on
attention seeking behavior.
Rþ and Pþ: highest score on
aggression.
Reactivity to other dog and people:
lowest in Rþ only.
Fear and aggression problems
higher in dogs that received any
type of punishment.
Haverbeke et al., 2008 33 dog and handler teams of the Standardized evaluation of 8 Team performance Team performance: 66% success in
Belgian Defense. obedience exercises and 5 Handler’s behavior obedience and 39% success in
Dogs between the ages of 1-5 years protection work exercises to assess Dog behavior protection work.
and have been working between teams’ performance. Use of Rþ (57.12%) > use of
3 months and 3 years. Two evaluations with 20 days in aversive stimuli (R and Pþ)
between were performed. (21.88%).
Dogs divided into high- Use of aversive stimuli:
performance and low-performance protection work > obedience
groups. high-performance < low-
performance dogs
distracted dogs > slightly
distracted, not distracted
2nd evaluation > 1st evaluation.
Dogs showed lower posture after
aversive stimuli in the 2nd
compared to the 1st evaluation
Herron et al., 2009 Dog owners who scheduled an Survey regarding previous Frequency of intervention use % of dogs responding aggressively
appointment for behavior behavior interventions and Aggressive response due to to confrontational training
consultation in a veterinary outcomes sent by email, fax, or intervention. methods:
hospital (n ¼ 140). postal mail to owners. Effect of intervention on behavior “Alpha roll” (31%), forced release of
Survey included a list of 30 possible problem item in dog’s mouth (38%), hit or
interventions. kick dog (43%), grab jowls/scruff
Interventions categorized by (26%), “dominance down” (29%).
researchers to: aversive, indirect % of dogs responding aggressively
confrontation, reward training, and to indirect confrontational training
neutral. methods:
“stare down” (30%), water pistol/
spray bottle (20%), growl at dog
(41%), yelling “no” (15%).
Only between 0%-6% of dogs
responded aggressively to neutral
and reward-based methods.
Arhant et al., 2010 Survey randomly sent to 5000 dog Questionnaire with 237 short Dog behavior Using Pþ: small ¼ large dogs.
owners. 1,405 surveys returned, questions regarding demographics, Training techniques used Higher frequency of punishments
1,276 analyzed. and dog and owner behavior. related to higher aggression and
Comparison between small excitability scores in small and
(<20 kg) and large (>20 kg) dogs. large dogs.
Relationship between punishment
and aggression stronger in small
dogs.
Relationship between punishment
and fearfulness and anxiety found
only in small dogs.
Use of reward-based responses to
unwanted behavior related to
higher frequency of aggressive
behavior.
All correlations < 0.3
(continued on next page)
54 G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60
Table (continued )
Rooney and Cowan, Dog owners and their dogs (n ¼ Researcher visited and filmed the Owner-reported training methods None of the 54 owners reported
2011 53). behavior of each owner and dog at Dog behavior using reward-based or
their home while following specific Owner behavior punishment-based methods
instructions including ignoring the exclusively.
dog and performing obedience Proportion of punishment-based
exercises. methods negatively correlated
Owners also completed a with dog interaction with
questionnaire. experimenter.
Proportion of reward-based
training methods positively
correlated with dog performance at
a novel task.
Using physical punishment: dogs
less interactive during play and less
likely to interact with
experimenter (compared to not
using physical punishment at all).
Dog performance in novel task
positively correlated with total
rewards delivered and owner
patience.
Blackwell et al., 2012 14,566 questionnaires distributed Questionnaires regarding Training method Only 3.3% reported using
to dog owners, 27% returned (n ¼ demographics, choice of training Problem behavior E-collars, 1.4% bark E-collars,
3,897). method, and prevalence of Demographics and 0.9% electronic containment
undesired behaviors. Training success system.
Higher % of owners using
reward-based methods
reported success for recall/chasing
problems (w97%) compared to
E-collar use (w83%) or other
aversive methods (w94%).
Occurrence of undesired behaviors
did not differ between training
methods.
Salgirli et al., 2012 Adult police dogs (n ¼ 42). Setup of training exercise with Salivary cortisol Use of E-collar and pinch collar led
human decoys meant to distract Behavioral observations to learning to disregard the
the dog into breaking a heel. distraction while heeling.
Repeated-measures design. % of dogs with behavior reactions to
Counterbalanced order of 3 punishment: no statistically
aversive stimuli: (1) quitting signal, significant differences between
(2) pinch collar, (3) E-collar. One pinch (64.3%) and electronic
week between conditions for each (38.1%) for extreme backward ears.
dog. Extreme low body posture: 4.8% of
dogs for pinch, 0% E-collar.
Vocalization: w60% E-collar >
w23% pinch.
# of dogs with maximal salivary
cortisol values: 17 quitting signal,
15 E-collar, 10 pinch.
Highest cortisol concentration
after quitting signal.
Grohmann et al., 2013 1-year-old intact male German A case study of severe brain Description of punishment Dog suspended a few feet in
Shepherd. damage after punitive technique Description of symptoms the air by a choke collar for
with a choke collar. approximately 60 seconds.
Dog panicked and lost
consciousness. After a few
hours, dog became ataxic
on all 4 limbs and was circling to
left. Several neurological
symptoms.
MRI revealed legions which
led to a diagnosis of severe
cerebral edema due to
ischemia.
Owner chose to euthanize
the dog.
Casey et al., 2014 Same questionnaire as Blackwell Questionnaires regarding Risk factors for aggressive behavior Compared to Rþ and P, using Pþ
et al., 2012. demographics, choice of training and/or R were related to
14,566 questionnaires distributed method, and prevalence of increased risk of aggression toward
to dog owners, 27% returned (n ¼ undesired behaviors. members of the family and toward
3,897). unfamiliar people outside the
house (odds ratio 2.8 and 2.2,
respectively).
(continued on next page)
G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60 55
Table (continued )
Cooper et al., 2014 63 dogs (no differences between Examining effects of E-collar use on Behavioral and physiological Training success: no differences
groups in age, sex, and breed) dogs’ welfare. measures before, during, and after between groups in owners’
divided to 3 groups based on dog Three groups trained for 5 days (2 training satisfaction and perception of dogs’
characteristics and past training sessions per day) for recall in the improvement.
history. presence of distractions: Behavioral measures:
(1) E-collar use, (2) same trainers time spent in tense state: 1 > 2, 3
with no use of E-collar, (3) trainers Low tail carriage: 1, 2 > 3
who don’t advocate E-collar use Yawning: 1 > 3
and no use of E-collar. Vocalization increase with E-collar
intensity increase.
# of commands given: 1, 2 > 3
(twice as many)
Interaction with environment: 1, 2
<3
Salivary cortisol: 3 >1, 2
throughout study.
Post-training minus pretraining
cortisol: 1 ¼ 2 ¼ 3
Deldalle and Gaunet, Dogs in advanced training class Observation of owner and dog Performance While walking on leash: % of dogs
2014 from 2 schools: Rþ school (n ¼ 24), performance of a walk on leash and Dog behavior gazing at owner: Rþ (63%) > R
R school (n ¼ 26). a sit command in advanced group Owner behavior (4%).
classes. During the “sit” command: # of
dogs showing mouth licking and
yawning and showing at least 1 of 6
stress-related behaviors:
R > Rþ.
% of dogs gazing at owner: Rþ (88%)
> R (38%).
% of dogs with low body posture: R
(46%) > Rþ (8%).
ACTH, adrenocorticotropic hormone; E-collar, electronic collar; MRI, magnetic resonance imaging; Rþ, positive reinforcement; R, negative reinforcement; Pþ, positive
punishment.
training methods and behavior in small (<20 kg) and large dogs aggression scores. As the authors of this study suggested, it is
(>20 kg) (Arhant et al., 2010). As in previous surveys, positive re- possible that the inconsistency in training methods led to uncer-
lationships between the use of punishment and excitability, and tainty or anxiety in the dogs, which in turn led to aggressive
between punishment and aggression, were found in both small and behavior. It is also possible that this finding merely suggests that
large dogs. A relationship between punishment and fearfulness was owners tried a variety of training methods to modify aggressive
found in small dogs only. In addition, reward-based responses to behavior. As this is the only study that reported this finding, more
behavior problems (e.g., calming or distracting the dog) were research on this topic is warranted.
related to aggression as well. Correlation values were low to
moderate (usually <0.3). In addition, the cutoff weight for small and Studies with direct observations of dogs
large dogs (i.e., 20 kg) may not have been specific enough to reveal Researchers in 3 studies directly observed the training tech-
differences in dogs of different sizes. niques and the behavior of owners and their dogs (Deldalle and
In summary, it appears that the use of training methods that are Gaunet, 2014; Haverbeke et al., 2008; Rooney and Cowan, 2011).
based on positive punishment and negative reinforcement are In 1 study, one of the researchers visited the home of each of 53
related to higher incidences of behavior problems, aggression, and dogs and their owners and conducted several tests of behavior and
fear. Reward-based responses to behavior problems were related to obedience (Rooney and Cowan, 2011). All sessions were filmed, and
aggression only in 1 study (Arhant et al., 2010). It is important to a history of training methods used by owners was obtained. The
note, however, that the 5 reviewed studies (i.e., Arhant et al., 2010; dogs of owners who tended to use punishments showed less
Blackwell et al., 2008; Casey et al., 2014; Herron et al., 2009; Hiby interaction with the experimenter. In addition, dogs that were
et al., 2004) were based on the owners’ subjective answers to trained with physical punishment were less likely to approach a
questionnaires and are prone to methodological difficulties such as stranger and played less with their owners. It was also found that
recall bias and the misunderstanding of terms or questions that dogs that were mostly trained with reward-based methods scored
were presented in the questionnaires. In addition, causality cannot higher in their ability to learn a novel task.
be established from the data in these studies, as it is not known In another exploratory study (Deldalle and Gaunet, 2014), one of
whether the dogs’ aggression or the presence of behavior problems the researchers observed training classes at 2 dog training schools.
led to the use of aversive training methods, or if the use of aversive In one school, the dogs were trained using positive reinforcement,
training methods caused aggression and other behavior problems. and in the other school, they were trained with negative rein-
Hence, the results of such studies should be interpreted appropri- forcement. The researcher observed and recorded the dogs’
ately and serve as a basis for future studies that are based on direct behavior as they were walking on a leash and responding to the
observation or specific interventions, rather than as a basis for “sit” command from their owners. More dogs showed stress-related
causal inference. In future studies, researchers should also take into behaviors and low body postures in the negative reinforcement
account the dogs’ size as a covariate in their statistical analysis, as it group compared to the positive reinforcement group. In addition,
can have an effect on dog owners’ choice of training methods. fewer dogs gazed at their owners in the negative reinforcement
Finally, at least according to 1 study (Blackwell et al., 2008), group, compared to the positive reinforcement group (during
inconsistency in training methods was related to the highest walking on leash, 4% [1 of 26 dogs] vs. 63% [15 of 24 dogs],
56 G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60
respectively; during the “sit” command, 38% [10 of 26 dogs] vs. 88% injured by another dog (n ¼ 151) and dog owners with a dog that
[21 of 24 dogs], respectively). These results suggest that dogs’ had injured another dog (n ¼ 55) while they were in a veterinary
welfare may be threatened by the use of negative reinforcement. clinic. The results showed that dogs that were trained by hitting or
Moreover, the data on gazes toward the owner may suggest that the “shaking,” and dogs owned by individuals who believed that
relationship between owner and dog can be compromised when without training a dog will be out of control, tended to be on the
training with negative reinforcement. However, it is also possible aggressor side of the dog-to-dog aggression. In contrast, dogs of
that the dogs from the positive reinforcement school looked at their owners who believed that training should be fun and that it would
owners because they were waiting for a reward or that they were be advantageous to have a trained dog were found to be more often
reinforced more often for this behavior compared to the dogs from on the victim side. In addition, dogs of owners who shouted and
the negative reinforcement school. It is important to note that gave clear commands were also found more often on the victim
positive punishment and negative punishment were not evaluated side. The explanation for the last finding is not readily apparent. It is
in this study. Both the history of punishment and the acute pun- possible that the dogs that were on the receiving end of shouting
ishments delivered throughout the training sessions could have were more fearful in general and had been punished for previous
affected the results of this study. For example, when using positive displays of aggression.
reinforcement, the withdrawal of food can be considered negative
punishment. In contrast, in order for a stimulus to serve as a Electronic collars and electronic containment systems
negative reinforcement, one has to apply positive punishment first
so that there will be an aversive stimulus to remove. The use of electronic collars is highly controversial. Some
Finally, a third study carried out evaluations of the performance trainers suggest that such collars are effective in modifying
of 33 military dog-handler dyads when performing standardized behavior, whereas other trainers find them inhumane and avoid
obedience exercises and protection work (Haverbeke et al., 2008). using them. When possible, practitioners should base their choices
In general, the teams’ performance (i.e., score based on correct and of training methods on scientific data. Hence, data on the rela-
incorrect exercises) was relatively low, with w66% success in tionship between the use of electronic collars and dogs’ behavior
obedience exercises and w39% success in protection work exer- are described in this section.
cises. Dogs that received more aversive stimuli (either positive
punishment or negative reinforcementde.g., pulling on the leash, Observational studies and surveys
hanging dog by collar, verbal scolding, hitting) were more In 1 study (Polsky, 2000), descriptive data about dog aggression
distracted and showed poorer performance compared to dogs that that may have been elicited by electronic containment systems
received less-aversive stimuli. In addition, the dogs showed a lower were collected. Five cases in which dogs inflicted multiple unin-
posture after the infliction of aversive stimuli by their handlers. The hibited bites on humans in the presence of an active electronic
authors of this study suggested that the welfare of dogs that containment system were found. In all cases, the dog was an intact
received aversive stimuli during training was threatened, although male with no prior displays of aggression. During the attack, all
this could not be directly proven by the data. dogs failed to show gross warning signs (e.g., snarling, growling)
In summary, the reviewed studies suggest that aversive training before biting. In 4 of the 5 attacks, the dogs received a shock before
methods (e.g., positive punishment and negative reinforcement) the attack. These results should be read with caution, as the data
may negatively affect the behavior and welfare of dogs. Moreover, were collected from legal reports and the behavior was not directly
none of the studies showed any evidence that aversive training observed by trained professionals. Still, because all of the dogs had
methods are more effective than reward-based training. In fact, not shown serious aggression before the incident, it is plausible that
according to Haverbeke et al. (2008), the opposite appears to be the aggressive behavior was elicited by the shock, by the classically
true. Hence, those working with dogs are encouraged to rely on conditioned response to a threat tone that precedes the shock, or by
reward-based training methods. a response to other environmental stimuli that preceded the shock.
However, 2 limitations are noteworthy. First, although observa- In contrast to the previous study, a survey of 3,897 dog owners
tional studies provide more robust data than surveys, they still do not did not find any difference in undesirable behaviors among those
necessarily provide support for causation. Second, in the first study using various training techniques (Blackwell et al., 2012). Results
(Rooney and Cowan, 2011), only the researcher who visited the dogs showed that only a small proportion of dog owners used electronic
rated their behaviors. This researcher also interviewed the owners. collars (3.3%, N ¼ 133), electronic bark collars (1.4%, N ¼ 54), or
Hence, as the authors mentioned, there is a risk for unconscious bias. electronic containment systems (0.9%, N ¼ 36). It is possible that
Similarly, in the second study (Deldalle and Gaunet, 2014), only 1 these small numbers prevented a meaningful statistical comparison
researcher recorded the behaviors of dogs while visiting the between methods. As for training effectiveness, the reported
training centers. In the third study (Haverbeke et al., 2008), the training success for teaching a dog to come when called or to pre-
authors did not report whether more than 1 researcher rated vent the dog from chasing was higher in the reward-based group
the behaviors. In future studies, 2 researchersdpreferably blind to compared to the electronic collars group. However, this could be
the hypotheses or study groupsdshould code behaviors, and due to a number of confounding factors, such as the seriousness of
interrater reliability should be reported. the problem behavior and owners’ perception of success.
Finally, because only 2 training schools were compared by Finally, the effects of using electronic collars were examined in a
Deldalle and Gaunet (2014), individual variations could have led to study that directly observed 32 dogs that were trained as police
some of the reported differences. Hence, as the authors of this study service dogs or watchdogs (Schilder and van der Borg, 2004). The
suggested, future studies should include a larger sample of training researchers also observed the behavior of dogs that were shocked in
schools to compensate for these individual variations. past training with dogs that had never received shock before. Dogs
trained with electronic collars vocalized and presented body pos-
Dog-to-dog aggression tures associated with stress or fear (e.g., tongue flicking, lowering
ear positions, lowering of the body/tail) for a fraction of a second
Only 1 study examined the relationship between training after receiving an electric shock. Both during free walking and
methods and dog-to-dog aggression (Roll and Unshelm, 1997). during training sessions without shocks, dogs that were trained
Questionnaires were given to dog owners with a dog that was with shocks in the past showed more stress-related behaviors than
G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60 57
dogs that were not trained with shocks. In addition, shocked dogs were examined in a study of 42 police dogs (Salgirli et al., 2012). The
were more stressed than control dogs on the training grounds and dogs were required to heel while a person serving as a decoy tried
also in a park unrelated to training. The fact that stress-related to distract them and cause them to leave the handler’s side. In a
behaviors were seen outside the training grounds, but in the counterbalanced design, the dogs received either a pull on a pinch
presence of the handlers, suggests that dogs associate the possi- collar, a shock from an electronic collar, or a quitting signal that was
bility of getting shocked with the presence of their handlers. The conditioned to signify the withdrawal of a reward. A similar number
authors of this study concluded that using electronic shocks for of dogs learned to disregard the distraction with the use of the
training is not only unpleasant but is also painful and frightening for electronic collar (n ¼ 39) and the pinch collar (n ¼ 32), compared to
the dogs. Hence, it appears that even dogs that make it through only 3 dogs with the use of a quitting signal. A plausible explanation
demanding training programs suffer from the aversive training for these results is that the dogs receiving the quitting signal did not
methods. understand what was expected of them in this specific setting.
Indeed, the training of the quitting signal was done with a toy and
Interventional studies not with a provoking person. Expecting the dogs to generalize the
One study compared the application of shock to 3 groups of quitting signal with a toy to a different scenario seems unrealistic.
laboratory-bred Beagles (Schalke et al., 2007). The first group of Hence, it is not surprising that the quitting signal failed to elicit the
dogs received the shock precisely when they grabbed a prey required behavior.
dummy. This led to an increase of w22% and w31% in absolute and Although not statistically significant, 64.3% (N ¼ 27) of the dogs
relative salivary cortisol, respectively. The second group received showed an extreme backwards ear position after being punished
the shock if they failed to respond to a recall while hunting the prey with the pinch collar, compared to 38.1% (N ¼ 16) that were pun-
dummy (w114% and w160% increase in absolute and relative sali- ished using the electronic collar. In addition, approximately 43%
vary cortisol, respectively). The third group received arbitrary and (Nw18) of the dogs being trained with a pinch collar or an elec-
unpredictable shocks (w336% and w328% increase in absolute and tronic collar showed a lowering of their back, and approximately
relative salivary cortisol, respectively). The fact that the Beagles that 31% (Nw13) of the dogs crouched. Vocalization was seen in
were shocked unpredictably had extremely high cortisol levels is approximately 60% of the dogs with the electronic collar (Nw25)
not surprising, as they could not predict and had no control over the compared to approximately 23% of the dogs with the pinch collar
coming shocks. In the second group, it is possible that the elevated (Nw10) (exact values were not reported in the original study;
levels of cortisol were due to the fact that the recall was trained values extracted from figures). Finally, of the 4 dogs that responded
without a prey dummy but was tested with it. Hence, it was difficult to the quitting signal, 2 showed a backward ear position and one
for the dogs to control their first reaction to chasing the prey. In the showed an extreme lowering of body posture and crouching.
first group, the predictability of the shock could have led to the Unfortunately, cortisol levels are difficult to interpret in this
relatively small increase in cortisol levels, but another explanation study. This is because actual cortisol values and effect sizes were not
is possible. Although an increase in the concentration of cortisol can reported and because there are contradictions in the reporting of
represent an increase in stress (Dreschel and Granger, 2005), it can results. For example, the authors state that no significant differ-
also represent the physical activity level of the dog. Indeed, eleva- ences in relative cortisol levels were found between groups, but at
tion in cortisol concentration can occur as a result of both low- the same time, they report that the relative cortisol level after the
intensity and high-intensity exercise (Radosevich et al., 1989). quitting signal was significantly higher than after the use of a pinch
However, the dogs in this study (Radosevich et al., 1989) exercised collar (Salgirli et al., 2012, p. 534). Still, 17 dogs showed maximal
for 90 minutes on a treadmill, and although plasma cortisol grad- cortisol values with the quitting signal, 15 dogs showed maximal
ually increased with the duration of exercise, large elevations were cortisol values with the electronic collar, and 10 with the pinch
seen only after 15-30 minutes of exercise. Because the dogs in collar. Statistical significance was not reported for these values.
Schalke et al.’s (2007) study ran after prey for less than 2 minutes a The results of this study suggest that the use of positive pun-
day, and because plasma cortisol samples were taken 10 minutes ishment in the form of a pinch collar or an electronic collar can have
after the administration of the shock, it is unlikely that the short detrimental effects on dogs’ physical and mental welfare. In addi-
exercise contributed significantly to the elevation in cortisol levels. tion, using negative punishment without clear or consistent in-
A second study (Cooper et al., 2014) compared the behavior and structions of what is expected of the dogs can lead to fear and stress.
cortisol levels of dogs trained to come when called, in 3 training
groups: (1) using an electronic collar, (2) training without an Bark collars
electronic collar but by the same trainers from group 1, and (3) Bark collars are a different type of electronic collars. Unlike
training without an electronic collar by trainers who believe in regular electronic collars that are operated manually by the handler,
reward-based training. The study found no differences in training these collars are designed to automatically deliver a shock every
effectiveness between groups. The dogs that were trained with time a dog barks. One study (Steiss et al., 2007) compared the use of
electronic collars tended to spend more time in a tense state, car- electronic bark collars, lemon spray collars (i.e., instead of an
ried their tails lower, interacted less with the environment, and electronic shock, a spray of an unfavorable odor is sprayed during
yawned more compared to the dogs that were trained without the barking), and control (inactivated) collars. Dogs in both the elec-
electronic collars by reward-based trainers. No differences in uri- tronic and spray collars groups barked less than dogs in the control
nary cortisol levels were found between groups, but salivary group. Although no statistically significant elevation in plasma
cortisol levels were higher in the group trained by the reward- cortisol or in adrenocorticotropic hormone (ACTH) was reported in
based trainers compared to the 2 other groups throughout the any of the groups (perhaps due to the small sample sizes), calcu-
experiment. However, comparing the difference between post- lations of effect sizes portray a different picture. These calculations
training cortisol to pretraining cortisol revealed no differences be- revealed an increase in plasma cortisol between the first day of
tween the 3 groups, suggesting that these values were not due to wearing the activated collars and the acclimation stage to the col-
training methods. lars (Cohen’s d ¼ 1.3 and 1.7 for the electronic collar and the spray
Although the previous studies compared aversive and non- collar, respectively). Similar effect sizes were calculated for ACTH
aversive training methods, the effects of 3 aversive training (Cohen’s d ¼ 0.82 and 1.9 for the electronic collar and the spray
methods on learning as well as on possible unintended outcomes collar, respectively). These effect sizes are considered large and
58 G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60
represent significant elevations in plasma cortisol and ACTH as a studies reviewed here examined acute stress responses, but it is
result of wearing the activated collars. mostly chronic stress that can negatively affect physical health in
The fact that bark collars can in some cases reduce barking was the long term (Beerda et al., 1997). One study found a relationship
shown previously, in a study that compared the satisfaction of between dogs’ fear of strangersda possible powerful stressor, and a
owners when using either an antibark citronella collar (88.9% shortened lifespan (Dreschel, 2010). The chronic effect of stress due
satisfaction; N ¼ 8) or an antibark electronic collar (44.4% satis- to aversive training on dogs’ physical health should be a relevant
faction; N ¼ 4) in a sample of 9 dogs (Juarbe-Diaz and Houpt, 1995). topic for future research. More research is needed to clarify the
However, the more relevant question is not whether these bark relationship between dogs’ behavior and stress because behavioral
collars work, but rather whether there are other effective training responses can vary between individual dogs and between various
methods that can alleviate barking without the added stress that is stimuli and can be misinterpreted.
associated with them. It is also possible that using certain punitive techniques presents
In summary, except for 1 study (Blackwell et al., 2012), all of the more direct health risks. One study showed increased intraocular
observational and interventional studies reviewed suggest that the pressure in dogs while pulling against a collar (Pauli et al., 2006).
various types of electronic collars may pose risks to dogs’ welfare. Importantly, 1 extreme case report of the effects of a specific pu-
Indeed, Schilder and van der Borg (2004) showed the risks to dogs’ nitive technique on the physical health of a 1-year-old German
welfare even when the collars are operated by experienced trainers. Shepherd dog was found (Grohmann et al., 2013). The dog was
The fact that dogs associate shocks with the presence of handlers hung several feet in the air with a choke collar for approximately
(probably due to classical conditioning) is not surprising and is 60 seconds and subsequently lost consciousness. A few hours after
troubling. It could be argued that the association these dogs make this incident, the dog developed several neurological symptoms
between their handlers and the painful shocks can make them less (ataxia in all 4 limbs, circling to the left, disorientation). Magnetic
reliable in situations when faultless performance is most needed. resonance imaging revealed severe cerebral edema due to ischemia.
This putative association should be examined in future studies. Tragically, the owner chose to euthanize the dog. Although this is an
The smallest elevations in cortisol levels when using electronic extreme case in the published literature, the author suggested that
collars were reported when well-timed shocks were delivered in a the punitive technique of choking a dog while hanging it in the air is
controlled environment (Schalke et al., 2007). Although no data are not uncommon, and more cases such as this one may have gone
available on this topic, it is unlikely that dog owners would have the unreported. Veterinarians and trainers should be made aware that
necessary skills or experience to use such collars, nor would they hanging a dog in the air or “helicoptering” it (e.g., lifting up by the
operate the collars in a controlled environment. choke chain and circling the dog in the air) presents a severe threat
Finally, shock collars, even in the hands of the most experienced to dogs’ health and obviously should be avoided. Legislators would
trainers, can only provide information regarding what behavior not do well to make such practices illegal.
to perform. These devices do not give the dog a choice of an alter-
native behavior to perform. Hence, given the available data and to Discussion
avoid risking the dogs’ welfare, trainers should avoid using elec-
tronic collars when training dogs. The discussion is divided into 2 sections: (1) Methodological
concerns and (2) Implications for practitioners and researchers.
The effects of aversive training techniques on physical health
Methodological concerns
The psychological unintended outcomes of aversive training
methods have been described, but the effect of aversive training This section discusses 4 methodological concerns regarding the
methods on the physical health of dogs should also be examined. It reviewed literature.
appears that stress can be associated with aversive training
methods. Increased cortisol levels followed shocks from electronic Sample size and lack of the reporting of effect size
collars (Schalke et al., 2007) and were found in dogs with activated The reporting of effect sizes allows readers to assess the practical
electronic and citronella bark collars (Steiss et al., 2007). Training significance of group differences and is not related to sample sizes.
inconsistency and the use of electronic or pinch collars were related This information is of importance because the lack of statistical
to maximal cortisol levels (Salgirli et al., 2012). Importantly, Beerda significance does not necessarily mean that differences between
et al. (1998) reported that unanticipated stimuli such as short groups are not of consequence. For example, in 1 of the reviewed
electric shocks and sound blasts led to increased salivary cortisol in studies (Steiss et al., 2007), calculation of effect sizes revealed
dogs. Low body posture, body shaking, crouching, yawning, and relatively large differences in plasma cortisol, although those were
restlessness were also indicators of acute stress (Beerda et al., 1998). not statistically significant. The lack of statistical significance was
Finally, Dess et al. (1983) showed a marked elevation in mean probably due to the small sample sizes (i.e., 6-8 dogs in each of 3
cortisol with (258% increase) or without (400% increase) control groups). Statistical analyses in future studies should make sure to
over electronic shocks. Control over the situation was assessed by report effect sizes in addition to null-hypothesis testing.
allowing or preventing the dogs from pushing a lever to stop the
shock. In addition, elevations in mean cortisol were seen whether The reliance on surveys and observational data, and the ethics of
the dogs could have predicted (291% increase) or could not have randomized interventions
predicted (374% increase) the coming shock (Dess et al., 1983). Many of the reviewed studies were based on surveys and ob-
Predictability was introduced by presenting an auditory tone before servations. Although these studies are valuable, they do not allow
the shock or shocking the dogs without signaling it beforehand. researchers to assess causal relationships. Hence, performing ran-
Such stress can affect dogs’ physical health. Indeed, a recent domized controlled interventions in which dogs are randomly
review of the effects of stress on animals’ health suggests that stress assigned to punishment-based training groups versus reward-
is associated with various damaging changes to physical health in based training groups may be warranted. However, the results of
dogs, including suppression of the immune system, gastrointestinal this review, and the vast literature on punishment in general (e.g.,
problems (e.g., diarrhea, vomiting, decreased appetite), delayed Durrant and Ensom, 2012; Sidman, 2000), suggest that punishment
puberty, and decreased sperm quality (Mills et al., 2014). The comes with negative unintended outcomes that can be detrimental
G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60 59
to an animal. From an ethical perspective, researchers should be practitioner has good teaching skills and the required knowledge of
cautious before performing such interventions and should ensure behavior. It may be noted that such ethical hierarchies of inter-
that the dogs’ welfare is not threatened. Once these conditions are vention that begin with the least intrusive and end with the most
met, researchers can be encouraged to continue learning about the intrusive are practiced in children’s education as well (Carter &
effects of aversive training methods on dogs’ behavior and welfare Wheeler, 2005). As Friedman (2009) suggests, it may be wise to
within the realm of observational studies. Direct observations of borrow such guidelines from the field of applied behavior analysis,
larger sample sizes, with a robust methodology of coding the as both animals and humans who require behavior modification are
observed behaviors (e.g., using 2 observers who are blinded to the often vulnerable and frequently cannot protect themselves. If
hypotheses, reporting interrater reliability), will allow researchers aversive or intrusive methods are chosen, the competence of the
to provide enlightening data on this topic without endangering the handlers is critical in order that they may achieve the proper timing
dogs’ welfare. and consistency required to allow for quick learning and to avoid
abusing dogs and threatening their physical and mental well-being.
Reporting case studies on physical health risks due to aversive Handlers’ competence should be defined, regulated, and assessed
training methods by relevant regulating agencies based on the recommendations of
Only 1 study reported a severe case of brain damage, due to accredited and experienced animal behaviorists.
hanging a dog in the air with a choke chain as a punitive technique One could rightly suggest that more studies with better meth-
(Grohmann et al., 2013). Because the use of such aversive methods odologies concerning the effects of aversive training methods on
may still be prevalent, it is unlikely that this is the only case. In dogs’ welfare are needed to strengthen the evidence on this topic.
addition, many punitive techniques involve pressure on the neck, However, the data emerging from the current review, as well as
which may lead to dangerous increases in intraocular pressure (see available data on the negative unintended outcomes of aversive
Pauli et al., 2006). Veterinarians should be encouraged to publish training methods in other species, such as in humans (e.g., Durrant
case studies in which dogs were injured due to the use of aversive and Ensom, 2012; Sidman, 2000) and rats (e.g., Ulrich and Azrin,
training methods. Publication of this type of information will allow 1962), suggest that it is perhaps time to pursue a different focus
those who work with dogs to understand the negative physical and approach of research. This new line of research will examine
symptoms related to aversive training methods and perhaps reduce how humane, reward-based methods can be improved to facilitate
their occurrences. better communication between humans and dogs. In turn, such
outcomes will allow dogs to modulate their stress and, at the same
The operation of electronic collars by experienced trainers time, improve their ability to effectively understand and respond to
In all of the observational and interventional studies that re- the behavior displayed toward them.
ported the use of electronic collars, experienced trainers operated
the collars. The results of the studies suggest that even when Acknowledgments
experienced trainers operate these collars, the welfare of the dogs
could be compromised. However, most dogs are not owned by The author would like to thank Dinah Olswang and Daveena
professional trainers, and the effects of regular pet owners using Tauber for their editorial assistance. The idea for the study was
such collars on dogs’ welfare are not known. Indeed, it is likely that conceived by the author. The study was written solely by the author.
the threat to dogs’ welfare would be even greater in the hands of
unskilled dog owners, who might lack the timing and consistency Conflict of interest
needed for this type of training to be successful. In such cases, due
to the aversive nature of these devices and the likelihood of training The author has no conflicts of interest relevant to the content of
ineffectiveness, their use can be abusive. this review.
Despite the methodological concerns, it appears that aversive No ethical approval was required as this is a review of literature.
training methods have undesirable unintended outcomes and that
using them puts dogs’ welfare at risk. In addition, there is no evi-
References
dence to suggest that aversive training methods are more effective
than reward-based training methods. At least 3 studies in this re- Arhant, C., Bubna-Littitz, H., Bartels, A., Futschik, A., Troxler, J., 2010. Behaviour of
view suggest that the opposite might be truedin both pets and smaller and larger dogs: effects of training methods, inconsistency of owner
working dogs (Blackwell et al., 2012; Haverbeke et al., 2008; Hiby behaviour and level of engagement in activities with the dog. Appl. Anim.
Behav. Sci. 123, 131e142.
et al., 2004). Because this appears to be the case, it is recom- Beerda, B., Schilder, M.B.H., van Hooff, J.A., de Vries, H.W., 1997. Manifestations of
mended that the dog training community embrace reward-based chronic and acute stress in dogs. Appl. Anim. Behav. Sci. 52, 307e319.
training and avoid, as much as possible, training methods that Beerda, B., Schilder, M.B.H., van Hooff, J.A., de Vries, H.W., Mol, J.A., 1998. Behav-
ioural, saliva cortisol and heart rate responses to different types of stimuli in
include aversion. For this purpose, it is proposed that Friedman’s dogs. Appl. Anim. Behav. Sci. 58, 365e381.
(2009) hierarchy of intervention strategies may be a good tool for Blackwell, E.J., Bolster, C., Richards, G., Loftus, B.A., Casey, R.A., 2012. The use of
choosing the least intrusive, yet effective, behavior modification electronic collars for training domestic dogs: estimated prevalence, reasons and
risk factors for use, and owner perceived success as compared to other training
tools. Friedman (2009) lists 6 levels of intervention: (1) arranging methods. BMC Vet. Res. 8, 93.
distant antecedents (least intrusive); (2) arranging immediate an- Blackwell, E.J., Twells, C., Seawright, A., Casey, R.A., 2008. The relationship between
tecedents; (3) positive reinforcement; (4) differential reinforce- training methods and the occurrence of behavior problems, as reported by
owners, in a population of domestic dogs. J. Vet. Behav.: Clin. Appl. Res. 3, 207e
ment of alternative behavior; (5) negative punishment, negative
217.
reinforcement, extinction; and (6) positive punishment (most Carter, S.L., Wheeler, J.J., 2005. Considering the intrusiveness of interventions. Int. J.
intrusive). According to Friedman (2009), levels 1-4 are sufficient Spec. Educ. 20, 136e142.
for solving the vast majority of behavior problems in animals. Level Casey, R.A., Loftus, B.A., Bolster, C., Richards, G.J., Blackwell, E.J., 2014. Human
directed aggression in domestic dogs (Canis familiaris): occurrence in different
5 may occasionally and under certain conditions be the effective contexts and risk factors. Appl. Anim. Behav. Sci. 152, 52e63.
and ethical choice. Level 6 is rarely needed or suggested when the Chance, P., 2003. Learning and Behavior. Wadsworth Publishing, Belmont, CA.
60 G. Ziv / Journal of Veterinary Behavior 19 (2017) 50e60
Christiansen, F.O., Bakken, M., Braastad, B.O., 2001. Behavioural changes and aver- Mills, D., Karagiannis, C., Zulch, H., 2014. Stressdits effects on health and behavior:
sive conditioning in hunting dogs by the second-year confrontation with do- a guide for practitioners. Vet. Clin. North Am.: Small Anim. Pract. 44, 525e541.
mestic sheep. Appl. Anim. Behav. Sci. 72, 131e143. Overall, K.L., 2007. Why electric shock is not behavior modification. J. Vet. Behav.:
Cooper, J.J., Cracknell, N., Hardiman, J., Wright, H., Mills, D., 2014. The welfare Clin. Appl. Res. 2, 1e4.
consequences and efficacy of training pet dogs with remote electronic training Pauli, A.M., Bentley, E., Diehl, K.A., Miller, P.E., 2006. Effects of the application of
collars in comparison to reward based training. PLoS One 9, e102722. neck pressure by a collar or harness on intraocular pressure in dogs. J. Am.
Deldalle, S., Gaunet, F., 2014. Effects of 2 training methods on stress-related be- Anim. Hosp. Assoc. 42, 207e211.
haviors of the dog (Canis familiaris) and on the dogeowner relationship. J. Vet. Polsky, R., 2000. Can aggression in dogs be elicited through the use of electronic pet
Behav.: Clin. Appl. Res. 9, 58e65. containment systems? J. Appl. Anim. Welf. Sci. 3, 345e357.
Dess, N.K., Linwick, D., Patterson, J., Overmier, J.B., Levine, S., 1983. Immediate and Radosevich, P.M., Nash, J.A., Lacy, D.B., O’Donovan, C., Williams, P.E., Abumrad, N.N.,
proactive effects of controllability and predictability on plasma cortisol re- 1989. Effects of low-and high-intensity exercise on plasma and cerebrospinal
sponses to shocks in dogs. Behav. Neurosci. 97, 1005e1016. fluid levels of ir-b-endorphin, ACTH, cortisol, norepinephrine and glucose in the
Dreschel, N.A., 2010. The effects of fear and anxiety on health and lifespan in pet conscious dog. Brain Res. 498, 89e98.
dogs. Appl. Anim. Behav. Sci. 125, 157e162. Reid, P.J., 1996. Excel-erated Learning. James & Kenneth Publishers, Berkeley, CA.
Dreschel, N.A., Granger, D.A., 2005. Physiological and behavioral reactivity to stress Roll, A., Unshelm, J., 1997. Aggressive conflicts amongst dogs and factors affecting
in thunderstorm-phobic dogs and their caregivers. Appl. Anim. Behav. Sci. 95, them. Appl. Anim. Behav. Sci. 52, 229e242.
153e168. Rooney, N.J., Cowan, S., 2011. Training methods and owneredog interactions: links
Durrant, J., Ensom, R., 2012. Physical punishment of children: lessons from 20 years with dog behaviour and learning ability. Appl. Anim. Behav. Sci. 132, 169e177.
of research. Can. Med. Assoc. J. 184, 1373e1377. Salgirli, Y., Schalke, E., Boehm, I., Hackbarth, H., 2012. Comparison of learning effects
Friedman, S., 2009. What’s wrong with this picture? Effectiveness is not enough. and stress between 3 different training methods (electronic training collar,
J. Appl. Comp. Anim. Behav. 3, 41e45. pinch collar and quitting signal) in Belgian Malinois police dogs. Rev. Med. Vet.
Grohmann, K., Dickomeit, M.J., Schmidt, M.J., Kramer, M., 2013. Severe brain damage 163, 530e535.
after punitive training technique with a choke chain collar in a German shep- Schalke, E., Stichnoth, J., Ott, S., Jones-Baade, R., 2007. Clinical signs caused by the
herd dog. J. Vet. Behav.: Clin. Appl. Res. 8, 180e184. use of electric training collars on dogs in everyday life situations. Appl. Anim.
Haverbeke, A., Laporte, B., Depiereux, E., Giffroy, J.M., Diederich, C., 2008. Training Behav. Sci. 105, 369e380.
methods of military dog handlers and their effects on the team’s performances. Schilder, M.B.H., van der Borg, J.M., 2004. Training dogs with help of the shock collar:
Appl. Anim. Behav. Sci. 113, 110e122. short and long term behavioural effects. Appl. Anim. Behav. Sci. 85, 319e334.
Herron, M.E., Shofer, F.S., Reisner, I.R., 2009. Survey of the use and outcome of Sidman, M., 2000. Coercion and its Fallout. Authors Cooperative Inc, Boston, MA.
confrontational and non-confrontational training methods in client-owned Revised edition.
dogs showing undesired behaviors. Appl. Anim. Behav. Sci. 117, 47e54. Steiss, J.E., Schaffer, C., Ahmad, H.A., Voith, V.L., 2007. Evaluation of plasma cortisol
Hiby, E.F., Rooney, N.J., Bradshaw, J.W.S., 2004. Dog training methods: their use, levels and behavior in dogs wearing bark control collars. Appl. Anim. Behav. Sci.
effectiveness and interaction with behaviour and welfare. Anim. Welf. 13, 63e69. 106, 96e106.
Juarbe-Diaz, S.V., Houpt, K.A., 1995. Comparison of two antibarking collars for Ulrich, R.E., Azrin, N.H., 1962. Reflexive fighting in response to aversive stimulation.
treatment of nuisance barking. J. Am. Anim. Hosp. Assoc. 32, 231e235. J. Exp. Anal. Behav. 5, 511e520.
The Behavior Analyst 1991, 14, 187-196 No. 2 (Fall)
barely probed the experimental possibil- largely whatever each participant wants
ities that remain available. In reconcep- it to be, and the resulting collective def-
tualizing how we can best use the field's inition of the field should be reason for
resources to meet society's demands for concern.
behavioral services, it will help if we rec- Perhaps we should worry about where
ognize that the effectiveness of our ser- this might lead. There is a risk that the
vices ultimately depends on the scientific modal doctoral level individual working
quality and scope of our applied research in some applied areas who is generally
enterprise. These values should be well acknowledged by himself or herself and
represented in our training programs. by others to be a behavior analyst might
Aside from reflecting a distinction be- become someone who is fairly unfamiliar
tween applied research versus service de- with the philosophical underpinnings of
livery, we also need to develop a better this approach to the study and manage-
sense of the literature and what it says ment of behavior, who has only a super-
about the field's directions and needs. In ficial appreciation of the research meth-
other words, we need to take an organi- ods of behavior analysis, and who has a
zational interest in our literature by re- poor understanding of the basic research
lating its active and inactive areas to cur- literature from which applied procedures
rent issues and needs. For instance, the supposedly derive. Although there might
shortcomings of the applied literature re- be disagreement about the appropriate or
garding punishment and alternative ways necessary degree of expertise in these ar-
of decreasing behavior have always been eas, it may be argued that such individ-
available for anyone to see, but few have uals are not likely to have state-of-the-
been interested in looking at the litera- art skills in analyzing and ameliorating
ture in terms of disciplinary or even po- behavioral problems. In fact, it is not at
litical needs. Although each researcher all uncommon to find individuals work-
must always be free to pursue whatever ing in retardation settings who have grad-
questions he or she wishes, the field of uate degrees but little formal behavioral
behavior analysis-like other scientific training and who proclaim at least prac-
disciplines-must begin to take regular tical expertise in this technology. Al-
stock of its research directions and relate though we certainly want to encourage
them to other scientific and societal in- identification with the field of applied be-
terests. The motivation to look at liter- havior analysis, this situation presents a
atures in this way can be inculcated real problem for the identification of the
through graduate education. field.
Even though these general training
Professional Education problems may contribute somewhat in-
directly to the aversives controversy, their
The professionals who are seen as rep- effect is probably pervasive. The lack of
resenting the field of behavior analysis in agreement about a core doctoral curric-
the area ofdevelopmental disabilities may ulum in behavior analysis certainly con-
also have something to do with the av- tributes to the shortcomings in the ap-
ersives controversy. First, because the plied literature suggested above because
field has generally not addressed curnic- the literature is primarily influenced by
ular issues, it has no minimum or even the educational history of those who have
recommended standards for what con- contributed to it. The field's preference
stitutes proper (minimum) training in be- for demonstrating the applied power of
havior analysis at either Master's or doc- operant contingencies instead of analyz-
toral levels. As a result, it sometimes ing behavior and procedures so as to de-
seems as if behavior analysts (and, thus, velop more effective and reliable tech-
the field itself) are increasingly being de- niques is partly the result of how behavior
fined by self-nomination rather than by analysts were trained to think about be-
educational history. To some extent, what havior and behavioral technology.
it means to be a behavior analyst can be The field can begin addressing these
192 J. M. JOHNSTON
bilities so that we can produce an ade- reaucrats, and parties interested in de-
quate supply of properly trained person- velopmental disabilities about the nature
nel. of the field of behavior analysis and the
The field can address these problems technology that is being offered to serve
by becoming much more aware ofits ed- this population. Even though behavior
ucational capabilities, its annual training analysis has made respectable contribu-
output at different degree levels, and var- tions to retardation journals, its focus
ious other features of its training efforts tends to have been narrowly technique
and the characteristics of its personnel. oriented. There seems to have been less
It should then regularly consider these tendency to publish on a broader spec-
personnel matters in the context of the trum of topics that would help educate
field's desires for growth in different di- other professionals about the implica-
rections and in relation to the personnel tions of behavioral philosophy, methods,
needs in various employment markets. and literature for the care of develop-
In other words, we need to do what other mentally disabled persons. Perhaps be-
mature disciplines do; we must attempt cause behavior analysts are sometimes
to exert some control over the production unfamiliar with talking to non-academic
ofbehavior analysts. Such controls would communities, we may do an even poorer
necessarily be indirect at the individual job communicating with people who do
level, but they might focus on widely not usually read journals. As a result, al-
varying features, such as the level of de- though bureaucrats, lawyers, parents, and
gree earned, the general area of special- interested citizens may know a little bit
ization, and the type of employment about behavioral programming, it may
planned. sometimes be worse than knowing noth-
For instance, even though this kind of ing.
information has never been gathered, it This superficial familiarity would seem
seems clear that our ability to offer ser- to assure a fertile ground for misunder-
vices to the field of developmental dis- standings, suspicions, and receptiveness
abilities and the demand from this field to systematic programs of deceit. For in-
for our graduates mean that we need to stance, the number of individuals who
produce far more individuals who have have taken an active role in arguing
a certain type of training at the bachelor's against the use of procedures involving
level and many more individuals with a aversive consequences is not large
different type of training at the Master's (Schroeder, 1990), but their arguments
level. In addition, we may also need to (both reasoned and emotional) have been
produce more doctoral level individuals sufficiently attractive to a much larger
who have the interest and training to de- constituency to gamer considerable sup-
velop research careers in developmental port. Might there have been broader re-
disabilities in academic and research set- sistance to such political efforts ifbehav-
tings, although we must carefully evalu- ioral technology and its foundations had
ate the employment market for such in- been better understood by the full range
dividuals. Of course, we also need to turn of interested parties?
out more doctoral graduates than we are This kind ofcommunication challenge
presently producing who are interested must be met at national, state, and local
in and trained for careers in the service levels. As the national representative of
delivery sector of this field. Although we behavior analysis, ABA must reach out
do not yet know what proportions ofthese to retardation and other fields in more
types of training are needed, this is ex- effective ways than it has in the past. For
actly what the field's ongoing study of instance, although there has been interest
personnel issues must learn. on the part of the American Association
There is a third kind of training need for Mental Retardation for coordinating
not yet effectively addressed by the field. the scheduling of its national meeting with
That is, the field has not done a very good ABA's annual meeting so that members
job of educating other professionals, bu- of both organizations can conveniently
194 J. M. JOHNSTON
attend parts ofboth gatherings, it has been certainly evident in the aversives contro-
difficult to arrange logistically. Neverthe- versy. For instance, the field is finding it
less, this is exactly the kind of national difficult to respond in a coherent manner.
gesture that is called for. At state and Because behavior analysis is still strug-
local levels, the effort must usually be gling with basic issues such as how it
more personal, although there are many should be defined and how its national
opportunities for programmatic activi- association should function (Etzel, 1988),
ties as well. State-level behavior analysis it has belatedly reacted to the aversives
organizations can do a lot here through issues at a national level with ad hoc spe-
their annual meetings and other outreach cial interest groups (e.g., the Internation-
programs (see Johnston & Shook, 1987). al Association for the Right to Effective
Treatment, Inc.) and the efforts of a few
Politics individuals who happen to be especially
interested in the problem. The picture is
In a more explicitly political vein, we no better at the state level. Behavior anal-
have not done a very good job of inte- ysis seems to have few allies in the bu-
grating effectively with the larger group reaucratic power structure in most states,
of professionals who constitute the power and those who are in power have many
structure in developmental disabilities. available mechanisms for pursuing their
This is an academically heterogeneous agenda.
population of individuals who may share We need to teach our students to be
only a career in this field and includes fully involved in the professional com-
professors, full-time researchers, officials munities defined by their interests, and
in state and federal service agencies, di- these interests must be broadly defined.
rectors of facilities, leaders of profession- In the case ofdevelopmental disabilities,
al and special interest associations, and for example, we need to encourage stu-
so forth. dents to be active members of a variety
Although there are certainly some of developmental disability associations
laudatory exceptions, the field of behav- and special interest groups, to publish in
ior analysis is generally not powerfully a broad array of developmental disability
represented in a broad range of other journals, and to play the politics neces-
groups and organizations. Neither do we sary to work their way into the power
seem to have infiltrated ourselves very structure. We need to encourage students
well into state service delivery systems. to look at administrative careers as no
Just one behavior analyst in even a mod- less valuable to the field's long-term in-
erately powerful position in a state agen- terests than research or service delivery
cy office may sometimes be more valu- careers. And we need to teach students
able than a number of behavior analysts how to be effective in their dealings with
working in service facilities. We do not those holding other perspectives without
seem to be intimately involved in setting losing their own identity (Morse & Bruns,
policy at either academic or bureaucratic 1983).
levels. The field tends to react to the pol-
icy initiatives of others and usually in a
fairly uncoordinated way. In other words, PIG'S EARS AND SILK PURSES
behavior analysis is not as well repre-
sented in the academic, state, and federal In summary, the lessons that we can
power structure of developmental disa- learn from the aversives controversy seem
bilities as it must be to avoid inappro- to revolve around how new behavior an-
priately constraining regulations and to alysts are created. Graduate (and under-
achieve its long term goals. graduate) education has a pervasive in-
Although behavior analysis has some fluence on who we are as professionals
legitimate excuses-it is, after all, a rel- and, thus, on how the field evolves. Our
atively small discipline, only a portion of professional education not only gives us
which is directly involved with the area a technical repertoire, it largely deter-
of retardation-this general deficiency is mines the nature of our general interests,
LEARNING FROM THE AVERSIVES CONTROVERSY 195
which eventually become the interests of of the scientific literature regarding the
the field. If we tend to limit our profes- treatment of severe behavior disorders
sional involvement to behavioral organ- that will bear directly on these issues (Na-
izations and journals, for example, it is tional Institutes of Health, 1989). Of
because we were not trained to under- course, there have been dozens of sym-
stand and appreciate the benefits of in- posia and other discussions ofthese mat-
volvement with other professional spe- ters at various national meetings.
cialties. Similarly, if we tend to ignore Third, the controversy has already re-
the needs of the punishment literature, it sulted in increased research funding de-
is because we were not adequately trained signed to improve our knowledge of ways
to evaluate the needs of the literature or to manage serious problem behavior. A
to appreciate and practice certain styles center grant was awarded by NIDRR to
of research. Robert Homer, Edward Carr, Glen Dun-
Finally, even though the aversives con- lap, Robert Koegel, and Wayne Sailor,
troversy is disruptive and will lead to among others, to conduct this kind of
some legal and regulatory actions that re- research. This is certainly not the only
quire correction (see Sherman, 1991), it research effort that has been occasioned
will also have some benefits for the field by these issues.
of behavior analysis and its interests in It is not yet clear what the resolution
the area of developmental disabilities. of the aversives controversy will bring.
First, it will prompt various appraisals It is obvious that the issues this fight has
such as this with their suggestions for how raised will be with us in different forms
behavior analysis functions as a science, for some years, and we will see actions
as a technology, and as a profession. These from federal, state, and professional en-
reviews may lead to some changes along tities affecting the delivery of behavioral
the lines suggested here. technology that will involve laws, regu-
Second, a number of debates, sympo- lations, policies, and research (Sherman,
sia, and special events have already pro- 1991). Nevertheless, it is possible to be
vided us with the belated opportunity to optimistic about the effects of the aver-
educate colleagues and others about the sives controversy. At one of Florida's re-
issues. For instance, in the fall of 1988, tardation institutions in 1972, there was
a small group ofindividuals were brought a major episode centering around the use
together by the National Institute for of behavioral procedures. It was nega-
Disability and Rehabilitation Research tively covered by the state's newspapers
(NIDRR) to educate some of their top and led to the governor appointing a blue
officials on these matters (at which there ribbon advisory committee to make rec-
was an opportunity to explain basic con- ommendations about how behavioral
cepts to NIDRR staff, the director of the programming should be regulated. Out
institute, and the U.S. assistant attorney of that mess eventually evolved one of
general).2 The Cambridge center for Be- the strongest behavioral programming
havioral Studies also held a debate in service delivery systems in the country,
Boston in December of 1988 that has led which is now spreading from the retar-
to films and other educational materials.3 dation system into the mental health sys-
Finally, the National Institute of Child tem (Johnston & Shook, 1987). Can we
Health and Human Development held a use the aversives controversy to lead us
consensus conference in September of toward improvements in how the field
1989 that provided a formal assessment goes about delivering its technologies?
2 The meeting was
organized by Rob Homer, and REFERENCES
the participants included Mike Cataldo, Barbara
Etzel, Judy Favel, Doug Guess, Jim Johnston, Gail Brantner, J. P., & Doherty, M. A. (1984). A review
McGee, Steve Schroeder, and Marti Snell. of timeout: A conceptual and methodological
I
The participants included Bea Barrett, Gary analysis. In S. Axelrod & J. Apsche (Eds.), The
LaVigna, Jim Johnston, Tim Paisey, Tom Near- effects ofpunishment on human behavior (pp. 87-
ny, Bob Sherman, Jane Salzana, Marcia Smith, and 132). New York: Academic Press.
Travis Thompson. Etzel, B. C. (1988). Message from the president:
196 J. M. JOHNSTON
Is an intervention called for? TheABA Newsletter, Morse, L. A., & Bruns, B. J. (1983). Nurturing
11(1), 2-5. behavioral repertoires within a nonsupportive
Foxx, R. M., & Azrin, N. H. (1973). The elimi- environment. The Behavior Analyst, 6, 19-25.
nation of autistic self-stimulatory behavior by National Institutes of Health. (1989). Treatment
overcorrection. Journal of Applied Behavior of destructive behaviors in persons with develop-
Analysis, 6, 1-14. mental disabilities. Washington, DC: Office of
Foxx, R. M., & Bechtel, D. R. (1984). Overcor- Research Reporting, National Institute of Child
rection: A review and analysis. In S. Axelrod & Health and Human Development.
J. Apsche (Eds.), The effects of punishment on Salzberg, C. L., Favell, J. E., Greene, B. F., Hopkins,
human behavior (pp. 133-220). New York: Ac- B. L., & Schneider, H. C. (1989, May). Supply
ademic Press. and demand ofbehavior analysts: Are we attend-
Johnston, J. M. (1972). Punishment of human ing? Panel discussion conducted at the meeting
behavior. American Psychologist, 27, 1033-1054. of the Association for Behavior Analysis, Mil-
Johnston, J. M. (in press). A model for developing waukee, WI.
and evaluating behavioral technology. In R. Van Schover, L. R., & Newsome, C. D. (1976). Ov-
Houten & S. Axelrod (Eds.), Effective behavioral erselectivity, developmental level and overtrain-
treatment: Issues and implementation. New York: ing in autistic and normal children. Journal of
Plenum. Abnormal Child Psychology, 4, 289-298.
Johnston, J. M., & Shook, J. L. (1987). Devel- Schreibman, L., Koegel, R. L., & Craig, M. S.
oping behavior analysis at the state level. The (1977). Reducing stimulus overselectivity in au-
Behavior Analyst, 10, 199-233. tistic children. Journal ofAbnormal Child Psy-
LaVigna, G. W., & Donnellan, A. M. (1986). Al- chology, 5, 425-436.
ternatives to punishment: Solving behavior prob- Schroeder, S. (1990). A history of the aversives
lems with non-aversive strategies. New York: Ir- controversy (1960-1990): A functional analysis.
vington. Unpublished manuscript.
Lovaas, 0. I., Koegel, R. L., & Schreibman, L. Schroeder, S. R., & Schroeder, C. S. (1989). The
(1979). Stimulus overselectivity and autism: A role of the AAMR in the aversives controversy.
review of research. Psychological Bulletin, 86, Mental Retardation, 27(3), iii-v.
1236-1254. Sherman, R. A. (1991). Aversives, fundamental
McGee, J. J., Menolascino, F. J., Hobbs, D. C., & rights, and the courts. The Behavior Analyst, 14,
Menousek, P. E. (1987). Gentleteaching:A non- 197-206.
aversive approach to helping persons with mental
retardation. New York: Human Sciences Press.
Applied Animal Behaviour Science 85 (2004) 319–334
Abstract
Behavioural effects of the use of a shock collar during guard dog training of German shepherd
dogs were studied. Direct reactions of 32 dogs to 107 shocks showed reactions (lowering of body
posture, high pitched yelps, barks and squeals, avoidance, redirection aggression, tongue flicking)
that suggest stress or fear and pain. Most of these immediate reactions lasted only a fraction of a
second. The behaviour of 16 dogs that had received shocks in the recent past (S-dogs) was compared
with the behaviour of 15 control dogs that had received similar training but never had received shocks
(C-dogs) in order to investigate possible effects of a longer duration. Only training sessions were used
in which no shocks were delivered and the behaviour of the dogs (position of body, tail and ears,
and stress-, pain- and aggression-related behaviours) was recorded in a way that enabled comparison
between the groups. During free walking on the training grounds S-dogs showed a lower ear posture
and more stress-related behaviours than C-dogs. During obedience training and during manwork (i.e.
excercises with a would-be criminal) the same differences were found. Even a comparison between
the behaviour of C-dogs with that of S-dogs during free walking and obedience exercises in a park
showed similar differences. Differences between the two groups of dogs existed in spite of the fact
that C-dogs also were trained in a fairly harsh way. A comparison between the behaviour during free
walking with that during obedience exercises and manwork, showed that during training more stress
signals were shown and ear positions were lower. The conclusions, therefore are, that being trained
is stressful, that receiving shocks is a painful experience to dogs, and that the S-dogs evidently have
learned that the presence of their owner (or his commands) announces reception of shocks, even
outside of the normal training context. This suggests that the welfare of these shocked dogs is at stake,
at least in the presence of their owner.
© 2003 Elsevier B.V. All rights reserved.
∗ Corresponding author. Present address: Department of Animals & Society, Yalelaan 17, 3584 CL Utrecht, The
0168-1591/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.applanim.2003.10.004
320 M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334
1. Introduction
A large variety of training methods are currently being used in dog training. These run
from very “friendly” methods (e.g. clicker training) to the use of apparently harsh methods,
like beatings and use of electronic and other equipment, that could cause wounds, pain and
mental harm to the animals.
For many decades the electric collar was being used in the Netherlands, especially in
training of police and guard dogs and for hunting and rescue work. The electric collar
consists of a collar that includes a battery and electrodes, and a remote control, through
which the trainer can deliver shocks of various durations and intensities to the dog. Some
types of collar can be tuned finely to the sensitivity of the dog being trained; other types
possess only a limited possibility of adaptation. Some collars include a feature to warn the
dog by sounding a beep, before a shock is delivered. Shock duration may vary from 1/1000
of 1–30 s; shock intensity will vary with coat structure and humidity, but in general a current
of a few thousand volts is used. Shocks are used mostly as a punishment, although some
manufacturers promote the electric collar by stating that it should be used as a negative
reinforcer, e.g. using the alleviation of pain after a series of very mild shocks has been
delivered as a reward. Use of the shock collar has been promoted by Tortora (1982, 1984).
Previously, shock has been used in research into avoidance learning in dogs (Solomon and
Wynne, 1953; Church et al., 1966).
Although the pros and cons of the use of this type of collar have been evaluated by Polski
(1994), and some descriptive data on the behavioural reactions of dogs to shocks has been
provided by Feddersen-Petersen (1999), no systematic investigations regarding possible
long term effects of the use of the collar have been published. Only one study (Beerda
et al., 1997) showed behavioural and cortisol effects upon the reception of some shocks
in laboratory dogs, that suggests that shocks are unpleasant. However, the results of this
study cannot be applied to the actual dog-training situation for several reasons. First, the
laboratory dogs did not expect any punishment at all; therefore their reactions could well be
much stronger than those of dogs during training. Secondly, dogs in guard and police dog
training are of another breed and thirdly, they are very excited during training. This could
well influence pain sensitivity. Therefore, we set out to investigate the direct behavioural
reactions of dogs upon receiving a shock during training, with the aim of finding what
behavioural responses are elicited by the reception of a shock. We were interested especially
in finding occurrences of pain, fear, avoidance, pain-induced aggression and submission.
Secondly, we wanted to investigate what the long-term impact of shocks could be. Our
reasoning, regarding the negative impact shocks could have on the dog’s welfare, was the
following. Suppose, that reactions to a shock only would involve cessation of the unwanted
behaviour, then the impact on welfare could be assessed as negligible. Suppose, however,
that signals suggesting pain and fear were shown concomitantly. Then, the impact should
be considered as being larger. But, if the behavioural signals would fade after shocking
was terminated and the dogs would behave completely normal thereafter, then the welfare
implications should be evaluated as present but minimal. But if on the training grounds, and
when no training was occurring, one could distinguish dogs that had received shocks in the
past from dogs that had never received shocks, then the impact of shocks should be regarded
as much greater and an interpretation of these shocks as “traumatic experiences” is likely
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 321
to be accurate. In that case, the dogs should have associated getting shocked with being on
the training grounds. Evidence would be stronger if shocked dogs could be distinguished
from non-shocked controls outside of the training grounds. This would hold if the dogs had
learned to associate their handler, or being ordered around, with the reception of shocks.
In that case, shocking must indeed have been perceived as aversive by the dogs. In the
latter case, a long-term impact on welfare is quite likely. On this basis we investigated
the behaviour of shocked and non-shocked control dogs before and during training and
also during a walk in a park, with the aim to find out if there were indications, that once
shocked, dogs were indeed more fearful than non-shocked controls. We limited ourselves
to behavioural reactions and omitted cortisol or other measures of physiological nature,
because dogs during training are highly excited, which in our belief would render such
measures useless. Also, measurements would tend to disrupt training, which would make
cooperation by trainers and handlers impossible.
Table 1
Ear, tail and body position
Description
Ear positions
Pinnae maximally backwards
Pinnae backwards The pinnae are backwards for more than half, are upright of
buckled, they are in one line with the stop of the nose and are
not flat in the neck
Pinnae partly backwards The pinnae turned backwards halfway and upwards; opening
is completely visible from the side
Neutral ears Pinnae partly sidewards and completely upwards, ear
openings are partly visible from the side.
Pinnae partly high Position between neutral and high
Pinnae maximally forward Pinnae maximally forwards and turned towards another and
forwards. Inside of pinnae not visible from the side.
Tail positions
Tail very low Tail tugged between hind legs
Tail low Upper side of tail against back, tail forms an S.
Tail half low Tail lower than neutral
Tail is neutral Tail follows line of lower back of dog and appears not above
the line of the back
Tail half high Tail is held above the contour of the back
Tail high Tail in a maximally high position
Body positions while walking and standing
Normal sit Sit in a normal position: legs stretched and head held
Shoulders While sitting the shoulders are withdrawn
Walk normal Walks with straight, legs not flexed
Bent legs Walks with flexed hind legs
Completely flexed Walks with flexed fore and back legs
Crouch stalk Walks with strongly flexed fore and hind legs
groups of dogs. First, a “free” walk on the leash, lasting about 2 min. No orders were given
to the dogs; secondly, obedience work which included exercising the following commands:
sit and down in motion, heeling in slow, normal and fast walking speed with changes of
direction, and recall to the handler; thirdly, protection work. This protection work included
a number of exercises (search for criminal, hold and bark at criminal, escape and defence,
followed by attack by the criminal, and finally, transport back). Also, we filmed these 31
dogs, while being walked in a park. There, the handlers were asked to walk the dog in a
“free-walk on the leash”, and subsequently to perform the same series of obedience exercises
as done on the training grounds. This enabled us to make the following comparisons (see
Fig. 1):
The data were sampled in the following way: We used an extensive ethogram, including
separate ear, tail and body positions (Table 1) in addition to a number of behaviours (Table 2).
Some behaviours or postures could only be scored in specific contexts, as indicated in
Table 2.
It appeared, that the execution of the exercises we had filmed varied to a large degree.
In order to sample data from the two groups of dogs that could be compared, we singled
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 323
Table 2
Behaviours, scored in four different contexts
Behaviour Description Scored during
out instances that occurred in all exercises in all dogs, in order to score the behaviours
that occurred at those moments. During free walk it was sufficient to score tail, body
and ear positions just once and exactly 1 min after the beginning of the excersise. We
used a one–zero sampling method to sample the other behaviours of the ethogram. During
324 M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334
Fig. 1. The design of the study: the numbered arrows between conditions denote the comparisons that have been
investigated. Control: control dogs, shocked: shocked dogs.
obedience exercises (14 commands) we could sample the ear positions 19 times, and the
tail and body positions each seven times. Using one–zero sampling on 17 occasions again
scored other behaviours. During protection work, the execution turned out to be so variable,
that meaningful comparisons between both groups were possible only when using the data
on “transport back of the criminal”, which stands for the following sequence: the handler
orders the would be criminal to walk; the dog has to heel and follow the criminal together
with the handler for about 30 m.
Here we used one–zero sampling for postures and behaviours. The tail and ear postures
were scored 3 s after the first command; other behaviours were scored during 10 s after that
first command, using one–zero sampling. All these observations lasted about 7 min per dog.
To each of the six ear and tail positions a score was allotted: the highest position got a
−1, the lowest position got a −6. For each of the three contexts (free walk, obedience and
protection work), for each training situation (park and training grounds) and for each dog
an average ear, tail and body position was calculated.
Data for the two groups were compared using the Mann–Whitney U (MWU) test. When
different contexts for the same group of dogs were compared we used the Wilcoxon matched
pairs (WMP) test. In comparison, we left out behaviours that occurred in less than 50% of
the dogs. We did this to prevent data from being influenced by just a few dogs.
Statistical tests were two-tailed, except in those cases, where we formulated explicit
expectations: namely when comparing control with shocked dogs and when comparing
free walking with training. Nominal P-values using an improved Bonferroni method (cf
Hochberg, 1988), were also calculated, but are not shown here. When correction lead
to a P-value becoming >0.05, the P-value concerned has been marked with an asterisk
(*).
3. Results
For data from two separate observation sessions, made in shocked dogs that walked freely
and did obedience exercises on the training grounds, we compared the scores for the ear and
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 325
tail positions and frequencies of some five behaviours that occurred in more than 50% of
the dogs. Therefore, the number of dogs may vary per test. For this purpose we calculated
Spearman’s rank correlation coefficients.
During free walking, the ear positions for the two data sets were highly correlated:
r = 0.92, P < 0.000. Tail positions correlated highly also: r = 0.70, P = 0.003. Other
behaviours could not be tested due to a too low number of dogs performing these.
During obedience work, both the ear positions (N = 15, r = 0.90, P < 0.000), and
the tail positions (N = 15, r = 0.74, P = 0.003) were highly correlated. For most of
the remainder of the behaviours that could be tested, the same was true—tongue flicking:
N = 9, r = 0.69, P = 0.006; lick mouth: N = 6, r = 0.85, P < 0.000; yawning: N = 9,
r = 0.59, P = 0.03; and turn away: N = 15, r = 0.71, P = 0.005. The only exception
was the behaviour lift front paw: N = 8, r = 0.57, P = 0.34.
We concluded, that for the tail and ear positions and for most of the stress-related be-
haviours that we were able to test, that the results from two different data sets were suffi-
ciently similar and that our measuring methods therefore were reliable.
The most frequent wrongdoings by the dogs that elicited a shock from the trainer, were
the following: dog does not obey “let go” command: 34×; dog heels ahead of the handler:
33×; dogs bites the criminal at wrong moment 12×; dog reacts too late on command “heel”:
8×. In six more contexts the dogs received four or fewer shocks.
The frequencies of behavioural reactions are shown in Table 3.
Table 3
Direct reactions of 32 dogs to 107 shocks
Behaviour Frequency No. of dogs
3.3.1. Does the behaviour on the training grounds of shocked dogs differ from that of
control dogs?
Testing for possible differences between control and shocked dogs (N = 15 and 16,
respectively) on the training grounds (comparison 1, see Fig. 2a–e) rendered the following
results, using one-tailed MWU tests.
During free walking, shocked dogs had lower ear positions than controls (U = 51.5,
P = 0.006), but tail positions did not differ (U = 82.5, P = 0.14).
Regarding stress-related behaviours, only the occurrence of licking lips could be tested
using the Fisher exact probability test. Significantly more shocked dogs showed licking lips
than did control dogs (P = 0.005). The other behaviours occurred too rarely to be tested.
During obedience exercises, again, the shocked dogs had lower ear positions than did
the control dogs (U = 68, P = 0.041*), tail positions again did not differ (U = 110.5,
Fig. 2. (a–e) Significant differences in postures and behaviours between control dogs and shocked dogs on the
training grounds. Averages and S.E.M.’s are given. Av.: average; freq.: frequency; observ.: total observation time.
For (a) and (b), the more negative the number on the y-axis, the lower the position of the ears.
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 327
Fig. 2. (Continued ).
P = 0.211). We could test five stress-related behaviours for differences between the two
groups. Tongue flicking was shown more by shocked dogs (U = 51, P = 0.006); as was
lift front paw (U = 55, ∗ P = 009). The remaining three behaviours that could be tested,
lick lips (U = 114, P = 0.830), yawning (U = 118, P = 0.953), and turn away (U = 88,
P = 0.216) were shown about equally often by both groups.
During protection work, ear positions in shocked dogs were lower (U = 45, P = 0.004)
than in control dogs, but tail positions did not differ (U = 92.5, P = 0.412). Also, shocked
dogs showed more paw lifting (U = 40.5, P = 0.007). Furthermore, significantly more
shocked dogs walked with completely flexed limbs than control dogs (7 versus 2, Fisher
exact probability test, ∗ P < 0.05), but there were no differences regarding the numbers of
dogs in both groups that showed tongue flicking or licking lips. Average values and S.E.M.’s
are presented in Fig. 2a–e.
3.3.2. Does the behaviour of shocked dogs in a park differs from that of control dogs?
Comparison 2 (all MWU tests, one tailed, with N = 15 for shocked dogs and N = 10
for control dogs) rendered the following results.
While walking on the leash, shocked dogs showed lower ears more often (U = 34, P =
0.023) but not lower tail positions (U = 67.5, P = 0.680) than control dogs. Stress-related
behaviours were too rare to be tested. During obedience exercises, again, ear positions of
328 M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334
Fig. 3. (a–c) Significant differences in postures and behaviours between control dogs and shocked dogs in a park.
Averages and S.E.M.’s are given. Av.: average; freq.: frequency; observ.: total observation time. For (a) and (b),
the more negative the number on the y-axis, the lower the position of the ears.
shocked dogs were lower than for control dogs (U = 30, ∗ P = 0.012), but not so the tail
positions (U = 49.5, P = 0.160). Of five stress-related behaviours that could be tested,
only tongue flicking was shown more by shocked dogs (U = 12.5, P = 0.001), lick lips
(U = 54, P = 0.261), lift front leg (U = 54, P = 0.261), yawning (U = 70, P = 0.810)
and turn away (U = 49.5, P = 0.160) were shown equally often by both groups of dogs.
Average values and S.E.M.’s are presented in Fig. 3a–c.
3.4. Is being on the training grounds more stressful than being in the park?
3.4.1. Are control dogs more frightened on the training grounds than in the park?
Two-tailed WMP tests for N = 10 dogs, resulted in no significant differences for ear or
tail positions during free walking of control dogs (N = 1 (9 ties), T = 0, P = 0.317 in both
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 329
Fig. 4. Significant difference in behaviour between park and training grounds for control dogs. Average and S.E.M.
are given. Av.: average; freq.: frequency; observ.: total observation time.
situations). During obedience work, again, no differences between ear and tail positions in
park and training grounds could be detected. (N = 10, T = 13.5, P = 0.152 and N = 9,
T = 15.0, P = 0.370, respectively).
From the stress-related behaviours, five occurred frequently enough to be tested. Only
tongue flicking occurred more on the training grounds than in the park (N = 7, T = 2.5,
P = 0.003). The other tested behaviours, lick lips (N = 8, T = 16, P = 0.763), lift
front paw (N = 5, T = 6, P = 0.680), yawning (N = 1 (9 ties), T = 1.0, P = 0.655),
turn away (N = 8, T = 10.5, P = 0.271), were shown equally often in both conditions.
Average value and the S.E.M. of the only significant difference found are presented in
Fig. 4.
Fig. 5. (a–c) Significant differences in postures and behaviours between park and training grounds for shocked
dogs. Averages and S.E.M.’s are given. Av.: average; freq.: frequency, observ.: total observation time. For (a), the
more negative the number on the y-axis, the lower the position of the tails.
differences in tail positions when comparing walking and protection work (N = 10, T =
16, P = 0.256). Differences in the occurrence of stress-related behaviours could not be
tested.
For control dogs, trends were the same. Ear positions were lower during obedience than
during walking (N = 14, T = 4, P = 0.002). During protection work, ear positions also
were lower than during walking (N = 9, T = 0, P = 0.006). Tail positions also were lower
during obedience than during walking (N = 15, T = 0, P = 0.001), and during protection
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 331
work these were almost significantly lower than during free walking (N = 7, T = 3.5,
P = 0.058).
4. Discussion
The fact that several simultaneous comparisons have been made when answering a ques-
tion necessitates the use of a (improved) Bonferroni correction. On the other hand, we
measured the same behaviours and postures in different conditions, so that in fact a kind of
repeated measurement design was used. The risk of using a Bonferroni correction is, that
meaningful significant differences may disappear. This is illustrated by the disappearance of
the significant difference between shocked and control dogs concerning their ear position
during obedience exercises on the training grounds (Section 3.3.1). A similar difference
shows up in Section 3.3.2. This significance also would disappear after correction. The
fact that this repeated measurement would render identical significant results by chance is
remote. This is the reason why we showed uncorrected P-values.
In spite of the impossibility of controlling the events taking place in the park and on the
training grounds, a careful choice of opportunities to score behaviours and postures made it
possible to sample behavioural data that enabled us to compare the postures and behaviours
of shocked dogs with those of control dogs and thus to investigate a possible lasting impact
of being shocked. The differences that we found between shocked and control dogs were
small but consistent. From the facts that: (a) shocked dogs consistently showed a lower
ear posture than control dogs during free walking and obedience in park and on training
grounds, and that shocked dogs showed lower ear positions while involved in protection
work; (b) shocked dogs show tongue flicking more often during obedience work in park
and training field; (c) shocked dogs show more lifted front paw during obedience work, we
conclude that:
(1) shocked dogs are more stressed than control dogs on the training grounds;
(2) shocked dogs are also more stressed than control dogs in the park;
(3) shocked dogs connect their handlers with getting shocks;
(4) shocked dogs may also connect orders given by their handlers with getting shocked.
The second point illustrates that effects of the electric collar, at least when used in a harsh
way, may be visible outside the training area. The most likely factor here is the presence of
the handler. In spite of the fact that some 75% of handlers and trainers that were interviewed
by a student of ours are of the opinion, that the dogs do not relate the presence of the handler
with getting shocked, the dogs obviously do. Secondly, we have some evidence that getting
an order, which previously was immediately followed by a shock or shocks, had obtained
a negative connotation: for example one dog, shocked immediately after getting a “heel”
command, yelped after getting the next “heel” commands without being shocked.
All this means, that when in presence of the handler, the dog has learned to expect
something aversive. Obviously, the enormous rewards the dogs experience during training,
i.e. chasing down, catching a criminal and winning the sleeve, do not counter the negative
effects of getting shocked. This is in spite of the fact that handlers of non-shocked dogs
admitted that they use prong collars, and that their dogs experienced beatings and other
harsh punishments, such as kicks or choke collar corrections.
Afterwards we wished that we had done a control experiment; namely have the shocked
and control dogs walked by an unknown person. Our results largely confirm the findings of
Schwizgebel (1982) who showed that dogs from trainers that punished a lot, showed lower
postures, a flexed sitting position and a lifted paw and tongue flicking.
We also have shown that the training in itself is stressful for both groups of dogs as
compared to walking on the leash. The example given above is one of the many mistakes
that were made during protection dog training: the command was followed by a shock so
quickly that the dog was not able to prevent a shock. This leads to an unwanted conditioning:
the dog has learned that getting a command predicts getting a shock.
We hope a future comparison of German shepherd dogs trained in a more friendly way
will bear out that indeed a friendly training regime leads to less signals of stress. We have
not proved that the long-term welfare of the shocked dogs is hampered, but we have made
clear that it is under serious threat.
M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334 333
4.4. Why is there so much and such heavy punishment during police and guard dog
training?
First, this type of training is typically and traditionally work by and for men: it is mostly
men, that do these trainings and they have been doing it their way and successfully for
many years. Men mostly are harder on animals than women, men may be perceived as more
threatening than females (Wells and Hepper, 1999; Hennesy et al., 1997). Secondly, training
time is too short. Thirdly, prestige is an important factor: championships or high rankings
count heavily. All this promotes severe punishment in order to get quick results. Success
in training does not promote willingness to change the type of training. Fourth, the type of
dog used here are highly motivated, hard, temperamental and possess low biting thresholds.
They stem from special breeding lines and quickly become so excited, that mistakes are
easily made and commands not obeyed. This also promotes punishments. Also, the high
excitement may have analgesic effects, so that softer physical punishments do not have the
desired effect.
Fifth, during training excessive emphasis is being laid upon biting. The “let go” command,
that is often not obeyed and therefore elicits punishment, is trained much later than biting.
Last, these dogs are expensive to extremely expensive animals. Handlers and trainers
both want to get the maximum out of the animals.
We concluded that shocks received during training are not only unpleasant but also painful
and frightening. Furthermore, we found that shocked dogs are more stressful on the training
grounds than controls, but also in a park. This implies, that whenever the handler is around,
the dog seems to expect an aversive event to occur. A second unwanted association might
be that the dogs have learned to associate a specific command with getting a shock.
Apart from the acute pain and fear, these expectations may influence the dog’s well being
in the long term in a negative way. To counter misuse of the shock collar, it is proposed
to ban its use for “sports”, but save it for therapeutic applications, such as for suppressing
hunting and killing sheep. The effects we found occurred in spite of the fact that control
dogs also underwent fairly harsh training regimes.
Trainers and handlers should study learning theory far better and review the structure
of the training in order to teach the let go command in an earlier phase and to reduce the
number of mistakes. They should incorporate more rewards during exercises. Also, less
temperamental and less forceful dogs should be bred. This also would decrease the chance
that dogs make mistakes for which they could receive punishment.
Acknowledgements
We wish to thank all the officials, trainers and dog handlers who had the courage to
participate in the research. They did this in spite of the fact that the results could be critical
regarding their handling of the dogs. We also thank them for their friendly reception. We
especially wish to thank Debbie Reijnders, who gave us access to the rather closed world
334 M.B.H. Schilder, J.A.M. van der Borg / Applied Animal Behaviour Science 85 (2004) 319–334
of guard dog training. We also wish to thank the Bond tot Bescherming van Honden (Asso-
ciation for Dog Protection) who provided the funding. Finally, we thank Han de Vries for
statistical advice and critical reading of the manuscript.
References
Beerda, B., Schilder, M.B.H., van Hooff, J.A.R.A.M., de Vries, H.W., Mol, J., 1997. Behavioural, saliva cortisol
and heart rate responses to different types of stimuli in dogs. Appl. Anim. Behav. Sci. 58, 365–381.
Beerda, B., Schilder, M.B.H., van Hooff, J.A.R.A.M., de Vries, H.W., Mol, J., 1999. Chronic stress in dogs
subjected to social and spatial restriction. I. Behavioural responses. Physiol. Behav. 66 (2), 233–234.
Beerda, B., Schilder, M.B.H., Bernardina, W., van Hooff, J.A.R.A.M., de Vries, H.W., Mol, J., 2000. Behavioural
and hormonal indicators of enduring environmental stress in dogs. Anim. Welfare 9, 49–62.
Church, R.M., LoLordo, V., Overmier, J.B., Solomon, R.L., Turner, L.H., 1966. Cardiac responses to shock in
curarized dogs: effects of shock intensity and duration, warning signal, and prior experience with shock. J.
Comp. Physiol. 62 (1), 1–7.
Conzemius, M.G., Hill, C.M., Sammarco, J.L., Perkowski, S.Z., 1997. Correlation between subjective and objective
measures used to determine severity of postoperative pain in dogs. J. Am. Vet. Med. Ass. 210, 1619–1622.
Feddersen-Petersen, D.U., 1999. Elektrogeräte: Ihr Einsatz bei der Ausbildung ist tierschutzrelevant. Kleintier 1,
16–17.
Fox, M.W., 1974. Understanding your dog. Blond & Briggs, London.
Hellyer, P.W., 1999. Minimizing postoperative discomfort in dogs and cats. Vet. Med. March, 259–266.
Hennesy, M.B., Davis, H.N., Williams, M.T., Mellot, C., Douglas, C.W., 1997. Plasma cortisol levels of dogs at a
county animal shelter. Physiol. Behav. 62, 481–490.
Hochberg, Y., 1988. A sharper Bonferroni procedure for multiple tests of significance. Biometrica 75, 800–802.
Klein, D., 2000. Elektrogeräte: Grundlagen, Wirkungen und mögliche Gefahren im Hinblick auf die Anwendung
in der Hundeausbildung. Der Gebrauchshund 1, 38–48.
Light, G.S., Hardie, E.M., Young, M.S., Hellyer, P.W., Brownie, C., Hansen, B.D., 1993. Pain and anxiety behaviors
of dogs during intravenous catheterization after premedication with placebo, acepromazine or oxymorphone.
Appl. Anim. Behav. Sci. 37, 331–343.
Martin, P., Bateson, P., 1993. Measuring Behaviour. Cambridge University Press, Cambridge.
Noonan, G.J., Rand, J.S., Blackshaw, J.K., Priest, J., 1996. Behavioural observations of puppies undergoing tail
docking. Appl. Anim. Behav. Sci. 49, 335–342.
Polski, R.H., 1994. Electronic shock collars: are they worth the risks? J. Am. Anim. Hosp. Ass. (30) 463–468.
Polski, R.H., 1998. Shock collars and aggression in dogs. Anim. Behav. Consultant Newslett, April 1998.
Schwizgebel, D., 1982. Zusammenhänge zwischen dem Verhalten des deutschen Schäferhundes im Hinblick auf
tiergerechte Ausbildung. Aktuel. Arbeit. Artgemass. Tierh., pp. 138–148.
Solomon, R.L., Wynne, L.C., 1953. Traumatic avoidance learning: acquisition in normal dogs. Psychol. Monogr.,
Gen. Appl. 67 (4), 1–19.
Tortora, D.F., 1982. Understanding electronic dog training, Part I. Canine Pract. 9, 17–22.
Tortora, D.F., 1984. Safety training: the elimination of avoidance motivated aggression in dogs. Aust. Vet. Pract.
14 (2), 70–74.
Ulrich, R., 1966. Pain as a cause of aggression. Am. Zoologist 6, 643–662.
van Hooff, J.A.R.A.M., Wensing, J.B., 1987. Dominance and its behavioural measures in a captive wolf pack. In:
Frank, H. (Ed.), Man and Wolf, pp. 219–252, ISBN 90-6193-614-4. Dr. W. Junk Publishers, Dordrecht, The
Netherlands.
Wells, D.L., Hepper, P.G., 1999. Male and female dogs respond differently to men and women. Appl. Anim. Behav.
Sci. 61, 341–349.
AVERSIVE TRAINING METHODS: NO, THANKS!