ANIMAL BEHAVIOUR, 2002, 63, 1153–1162

doi:10.1006/anbe.2002.3017, available online at http://www.idealibrary.com on

Cannibalistic behaviour spread by social learning

SYLVIE CLOUTIER*, RUTH C. NEWBERRY*, KRISTEN HONDA* & J. RICHARD ALLDREDGE†

*Center for the Study of Animal Well-being, Department of Veterinary and Comparative Anatomy, Pharmacology
and Physiology, Washington State University
†Program in Statistics, Washington State University

(Received 2 April 2001; initial acceptance 7 June 2001;

final acceptance 29 November 2001; MS. number: A9030)

We hypothesized that social learning is involved in the spread of cannibalism in domestic fowl Gallus
gallus domesticus. To investigate this hypothesis without harming birds, we used an inanimate chicken
model as our cannibalism stimulus. We randomly assigned flocks of 12 White Leghorn pullets to one of
two treatments: (1) flocks with two trained demonstrators (N=9) and (2) control flocks (N=8). Demon-
strators were trained to pierce a membrane covering a dish of chicken blood and consume the blood. To
assess the effect of access to the cannibalism stimulus during demonstrations, we randomly assigned
observer pairs to one of two observer treatments: (1) observe stimulus through a wire mesh partition and
(2) observe stimulus within the same enclosure. We conducted five 10-min demonstration sessions, each
followed by a 10-min test of each observer pair in the absence of demonstrators, over a period of 15 days
when the birds were 41–55 days of age, and two further tests at 63–64 and 91–92 days of age. Pairs that
observed demonstrators piercing a membrane and consuming blood were more likely to perform this task
when tested than control pairs. Learning of the task was enhanced by direct access to the cannibalism
stimulus rather than observing it through a wire mesh partition. Blood consumption during tests was
increased by direct access to the cannibalism stimulus during demonstration sessions. The birds made
bigger holes in the membrane when tested after observing trained demonstrators and after having direct
access to the stimulus. Our results provide the first experimental evidence that social learning can
contribute to the spread of cannibalistic behaviour in domestic fowl. We suggest that stimulus
enhancement and observational conditioning were the social-learning mechanisms involved.
 2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.

Social learning refers to situations in which observation
of, or interaction with, another animal or its by-products
facilitates the acquisition of a new pattern of behaviour
(Galef 1988). Social learning allows behavioural inno-
vations to spread rapidly through a population, thereby
facilitating adaptation to different environments and
unpredictable conditions (Boyd & Richerson 1988). Social
transmission of foraging behaviour has been reported in a
wide range of animals (Palameta & Lefebvre 1985) includ-
ing junglefowl, Gallus gallus spadiceus, and their domesti-
cated relatives, G. gallus domesticus. Junglefowl show
feeding site preference and a shorter latency to feed if
they have previously observed a conspecific demonstrator
forage successfully from a marked dish (McQuoid & Galef
1992, 1993) and social transmission of key pecking for a
food reward has been reported in domestic laying hens
(Nicol & Pope 1992, 1993, 1994, 1999).
Correspondence: S. Cloutier, Department of Veterinary and Compara-
tive Anatomy, Pharmacology and Physiology, Washington State Uni-
versity, P.O. Box 646520, Pullman, WA 99164-6520, U.S.A. (email:
scloutie@vetmed.wsu.edu). J. R. Alldredge is at the Program in Statis-
tics, Washington State University, Pullman, WA 99164-3144, U.S.A.
1153
0003–3472/02/$35.00/0  2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.
Cannibalistic behaviour is widespread in the animal
kingdom, both in wild and captive populations (Elgar &
Crespi 1992). In poultry, cannibalism involves beak-
inflicted damage followed by the consumption of blood
and other tissues of conspecifics, either while they are still
alive or after death. Cannibalism constitutes a serious
welfare and economic problem when live birds are
injured and killed. Although variable between strains and
among flocks within strain, mortality due to cannibalism
above 30% has been reported in flocks of laying hens
with intact beaks (Craig & Muir 1996; Gunnarsson et al.
1999). It appears that bleeding resulting from feather
damage during bouts of feather pecking or from damag-
ing pecks to other body parts such as the cloaca or toes
can stimulate cannibalism (Hughes & Duncan 1972;
Cuthbertson 1978; Cloutier et al. 2000). Mechanisms
underlying the development of feather pecking and can-
nibalistic behaviour in poultry are poorly understood.
Recent evidence (Huber-Eicher & Wechsler 1997, 1998)
supports the hypothesis that feather pecking and canni-
balism represent foraging behaviour that becomes
redirected towards conspecifics in the absence of
1154 ANIMAL BEHAVIOUR, 63, 6
adequate alternative pecking stimuli within the captive
environment.
In poultry husbandry, it is considered important to
remove eggs and dead birds from the flock as soon as
possible so that the birds do not learn to eat them. It has
been suggested that deleterious behaviour may be copied,
leading to rapid proliferation of the behaviour within a
flock (Appleby et al. 1992; Nicol 1995). Among laying
hens, there is a significant trend for one death from
cannibalism to be followed by more in the same cage
(Allen & Perry 1975; Craig & Muir 1996) and adjacent
cages (Tablante et al. 2000). Within a cage, this finding
could be accounted for by the same bird attacking mul-
tiple victims. Cannibalism in adjacent cages in a limited
area within a poultry house could be due to individual
learning of the behaviour by birds exposed to specific
environmental conditions in that location, such as rela-
tively bright lighting. Alternatively, or additionally, can-
nibalistic behaviour may spread among birds through
social learning. Recently, Zeltner et al. (2000) have
reported evidence of a social influence on proliferation of
feather pecking in chickens. We were interested to deter-
mine whether social learning could be implicated in
outbreaks of cannibalism.
We hypothesized that social learning contributes to the
spread of cannibalistic behaviour in poultry and that
social learning is enhanced by direct access to victims
of cannibalism. We investigated these hypotheses in
chickens of a strain selected for high egg production
using an inanimate model (a dish of chicken blood
covered by a membrane) to which cannibalistic behav-
iour similar to that used in cannibalistic attacks on live
birds (damaging pecking followed by consumption of
chicken blood) could be directed. In previous exper-
iments, naïve chickens and pheasants, Phasianus colchi-
cus, avoided dishes containing conspecific blood (Jones &
Black 1979; Jones & Faure 1981, 1983), suggesting that
our model would not initially be recognized as a food
source. Thus, we considered our experimental construct
to be ecologically valid for investigating the acquisition
and diffusion of cannibalism within a population. We
predicted that observer chickens given the opportunity to
watch trained demonstrator chickens pecking open the
model and ingesting the blood would be more likely to
perform this behaviour subsequently, and to learn the
behaviour more quickly, than control chickens not given
this opportunity. We also predicted that observers able to
interact directly with the model and trained demonstra-
tors would be more likely to learn the behaviour, and
learn it sooner, than observers physically separated from
the model.
MATERIAL AND METHODS
Animals and Husbandry
We obtained female White Leghorn chicks from a
commercial hatchery and randomly divided them into 18
flocks of 12 individuals. We housed them in wire-floored
brooder units until 4 weeks of age. At this time, we
transferred them to enriched floor cages having a large
main compartment (9191100 cm high), where
the birds were housed, and a smaller compartment
(3091100 cm high) separated from the main com-
partment by a wire fence. Each compartment had a water
cup, and a food trough and wooden perch running across
its full length. The floor was covered with wood shavings.
Each flock was visually isolated from adjacent flocks by
solid walls. We provided food (standard pullet mash) and
water ad libitum. The duration of the photoperiod was
12 h. The ambient temperature was 29 C in the first week
and then gradually decreased to 21 C. The average light
intensity at the level of the feeder was 9420.7 lx in the
brooders and 726.8 lx in the floor cages.
During week 6, we adopted out some males that had
been incorrectly sexed at the hatchery and redistributed
females from one flock to the other flocks so that there
were 17 flocks of 12 birds. To facilitate individual identi-
fication, we marked each bird with plastic leg bands and
a coloured mark on her back. We visually monitored the
birds for health problems at least twice daily. Any sick or
injured bird was immediately separated from her flock-
mates and given veterinary care until fully recovered.
Although the birds had been vaccinated for Marek’s
disease, one bird developed symptoms and was eutha-
nized at 56 days. After the experiment, the birds were
adopted out to a private farm where they were kept under
free-range conditions.
Experimental Design
Within each flock, we selected two demonstrators (see
below). The remaining 10 birds were randomly divided
into five pairs of observers. We used pairs of demon-
strators and observers for the experiment because isolated
fowl may engage in prolonged periods of distress vocal-
ization and remain immobile (McQuoid & Galef 1992).
We randomly assigned flocks to one of two demonstrator
treatments: (1) flocks with naïve demonstrators (Control,
N=8) and (2) flocks with trained demonstrators (Trained,
N=9). In the Trained treatment, the demonstrators were
trained to pierce a membrane covering a dish of chicken
blood (see below) whereas, in the Control treatment, the
demonstrators received no such training. This condition
controlled for the probability that the task could be
learned individually. To assess the effect of access to the
dish of chicken blood during demonstrations, we ran-
domly assigned the observer pairs to one of two observer
treatments: (1) observe through a wire mesh partition
(O1) and (2) observe within the same compartment (O2).
Chicken Models
To avoid unnecessary harm, we used ‘chicken’ models
rather than live birds to test our hypotheses about social
transmission of cannibalistic behaviour. A chicken model
was a clean petri dish (9 cm diameter, Pyrex, Corning,
Corning, New York, U.S.A.) containing 32 g of chicken
blood covered by a smooth membrane of nontoxic para-
film (Parafilm M, American National Can, Menasha,
Wisconsin, U.S.A.). To prevent the dish from being
turned over and to contain any spillage, we attached the

dish to a clean plastic tray (23 cm in diameter) using
Velcro. Whenever the chicken model was presented, a
clean sheet of brown paper was placed under it that
covered the whole floor area. Although the experiment
could have been conducted using whole dead chickens,
we were concerned that behaviour directed to carcasses
would be too easily generalized to live birds. We could
have provided chicken meat or internal organs in a
membrane-covered dish, as these are consumed during
cannibalistic attacks, but we chose to use blood as it
provides the first sign of injury in a cannibalistic attack.
Also, a bird pecking at a bleeding wound appears to
attract others to join in the attack. We obtained the blood
from chickens humanely slaughtered at a commercial
abattoir and desanguinated into a heparinized container
to prevent coagulation. The blood was stored at
70 C
until needed for the experiment.
Selection and Training of Demonstrators
Since high social status of demonstrators can enhance
social learning (Nicol & Pope 1994, 1999) but dominance
relationships in female chicks do not become fully estab-
lished before 9–10 weeks of age (Rushen 1984), we used a
‘worm running’ competitive test (Rogers & Astiningsih
1991) to select the demonstrators. We scored competition
among birds to pick up and run with a piece of wound
white pipe cleaner, 30 mm in length and approximately
3 mm in diameter, during 10-min tests at 8, 10 and 12
days of age. At each 5-s interval, we recorded the identity
of the bird carrying the pipe cleaner. Birds were ranked
according to their cumulative score over the three tests.
We selected the two highest-ranking birds in each flock to
be the demonstrators. Familiar demonstrators were used
because hens tend to avoid unfamiliar conspecifics when
possible (Nicol & Pope 1999).
We conducted demonstrator training during 13 train-
ing sessions between 21 and 40 days of age. Demon-
strators were food deprived before training sessions by
removing the food trough in the morning before the
lights came on. Once the lights were on, we moved each
pair of demonstrators to a separate, visually isolated cage
for a 30-min period. At the end of each training period,
we washed off any blood adhering to demonstrators and
returned them to their home cages. Food (apart from that
provided by the chicken model) was withheld from the
demonstrators assigned to the Trained treatment during
the training sessions. To habituate these demonstrators to
blood, we introduced a chicken model minus a mem-
brane and with some (familiar) food mixed into the
blood. In five subsequent training sessions, we reduced
the proportion of food to blood until the birds readily ate
the blood with no added food. On the seventh training
session, we presented a chicken model with a membrane
in which we had made two incisions. In subsequent
sessions, we reduced the number and size of the incisions
until the demonstrators were using their beaks to pierce
the membrane by themselves. Demonstrator training
continued until the birds reached a high level of ef-
ficiency (<60 s latency to pierce the membrane upon
presentation of the model in the 13th training session).
CLOUTIER ET AL.: CANNIBALISM AND SOCIAL LEARNING 1155
The demonstrators assigned to the Control treatment
were not presented with a chicken model during training
periods and were given access to food and water ad
libitum, as were the remaining birds in their home cages.
Trials
Trials were conducted in cages similar to the birds’
home cages and situated in the same room. We con-
ducted the first five trials over a 15-day period when the
birds were between 43 and 55 days of age. In these trials,
a demonstration session was followed by a test of each
observer pair in the absence of demonstrators. Trials 6
and 7 were conducted when the birds were 63–64 and
91–92 days of age, respectively. In these trials, we tested
the observer pairs without a preceding demonstration.
Because food deprivation of observers hinders social
learning (Nicol & Pope 1993), we did not deprive ob-
servers of food prior to demonstration sessions (trials
1–5). We did withhold food during trials 6 and 7, when
only tests for retention were given, by removing the food
trough before the lights came on in the morning.
For the O1 demonstration sessions, we placed the three
pairs of O1 observers from a flock together in the main
compartment of a cage. For flocks on the Trained treat-
ment, we placed the pair of trained demonstrators for
that flock in the adjacent smaller compartment of the
cage and introduced a chicken model for 10 min. The O1
observers were, therefore, able to observe the demon-
strators pierce the membrane and consume chicken blood
by watching them through the wire mesh partition. For
flocks on the Control treatment, we placed a model in the
small compartment without any demonstrators. The six
observers could look at the model through the wire mesh
partition but received no demonstration. For the O2
demonstration sessions, we placed the two pairs of O2
observers in the large compartment along with the pair of
demonstrators. We then introduced a chicken model for
10 min. The O2 observers on the Trained treatment were,
thus, able to observe the demonstrators cannibalize the
chicken blood at close range as well as having the oppor-
tunity to peck and pierce the model and to scrounge on
blood exposed by demonstrators or fellow observers. The
O2 observers on the Control treatment had the oppor-
tunity to peck and pierce a model and to scrounge on
blood exposed by one of the five other birds present.
After the 10-min demonstration sessions, we tested
each observer pair sequentially in a random predeter-
mined order that differed for every trial. Tests were
conducted in the small cage compartments but with
brown paper covering the wire mesh to prevent any birds
in the main compartments from observing the tests. A
chicken model was placed in the centre of the compart-
ment with the observers for 10 min and their responses
were recorded on videotape. We limited the duration of
exposure to the chicken model to minimize the chance
that any birds becoming contaminated with blood would
become targets of cannibalistic pecking. We washed off
any blood adhering to birds’ beaks, feet or feathers before
returning birds to their home cages.
1156 ANIMAL BEHAVIOUR, 63, 6

Table 1. Effects of demonstrator type (Trained versus Control) (D), observer type (O1 versus O2) (O), trial (T) and
their interactions on responses of pairs of observer chickens to a dish of blood covered by a membrane*
D O D*O T D*T O*T D*O*T

Peck membrane (proportion)

F 7.99 7.15 1.22 22.95 1.33 2.22 0.60
P 0.01 0.02 0.29 0.0001 0.24 0.04 0.73
df 1,15 1,15 1,15 6,180 6,180 6,180 6,180
Pierce membrane (proportion)
F 5.56 6.96 0.01 8.94 1.25 1.35 0.76
P 0.03 0.02 0.91 0.0001 0.28 0.24 0.60
df 1,15 1,15 1,15 6,180 6,180 6,180 6,180
Latency to peck membrane (s)
F 6.37 5.84 4.16 39.93 1.09 4.38 0.50
P 0.02 0.03 0.06 0.0001 0.37 0.0002 0.81
df 1,15 1,15 1,15 6,525 6,525 6,525 6,525
Latency to pierce membrane (s)
F 5.19 9.61 0.04 13.99 1.36 2.60 0.83
P 0.04 0.007 0.84 0.0001 0.23 0.02 0.55
df 1,15 1,15 1,15 6,525 6,525 6,525 6,525
Membrane score†
F 5.48 9.83 0.13 13.40 1.92 0.48 0.72
P 0.03 0.007 0.73 0.0001 0.08 0.82 0.64
df 1,15 1,15 1,15 6,515 6,515 6,515 6,515
Blood eaten (g)
F 2.64 6.34 0.66 20.86 1.35 2.52 0.44
P 0.13 0.02 0.43 0.0001 0.23 0.02 0.85
df 1,15 1,15 1,15 6,505 6,505 6,505 6,505

*Mixed model analysis of ranked data, with trial as a repeated measure.

†1=No damage, 2=indentations, 3=hole(s) less than 5 mm long, 4=hole(s) at least 5 mm long.

Measurements demonstrator type were random terms in the model. The

subjects were the observer pairs. We used the compound
Before and after each test, we weighed the chicken
symmetry covariance structure where possible but the
model so that we could determine the amount of blood
PS(POW) covariance structure provided a better fit to the
disappearing from the model during the test through
data for the membrane score and the proportion vari-
ingestion by the birds and any spillage, designated as
ables. Because the residuals were not normally distrib-
‘blood eaten’. After the test, we assigned a membrane
uted, we applied the mixed linear model to both
score, where 1=no damage, 2=indentations but no holes,
untransformed and ranked data. Similar results for the
3=one or more holes less than 5 mm long, and 4=one or
two analyses indicated reliability of the analysis on the
more holes at least 5 mm long. The following data were
ranked data (Zar 1996). Only results from the analyses of
collected from the videotapes of each observer pair by a
the ranked data are presented for these variables. To assess
human observer blind to the treatments: latency for the
rates of learning in each treatment combination, the
first bird of the pair to peck the membrane with or
cumulative proportion of pairs performing pecking and
without making a hole in it (peck latency) and latency for
piercing behaviour at each trial were summarized by
the first bird of the pair to make a hole in the membrane
linear (Proc GLM) and nonlinear (Proc NLIN) regression
as a result of pecking at it (pierce latency). If the behav-
models (SAS Institute 1999). We used SAS Proc NLMIXED
iour was not performed, a maximum latency of 600 s was
to construct confidence intervals to allow testing for
recorded. We also determined the proportion of pairs that
differences in learning rates. We excluded trials 6 and 7
performed pecking and piercing. We noted isolated inci-
from the models because these trials tested retention
dents in which a bird broke the membrane by scratching
rather than acquisition of the cannibalistic behaviour. To
at it or landing on it with her claws but these events did
determine whether the birds generalized from the model
not occur more than once within the same flock and were
to real birds, we used SAS Proc GLM to assess the effect of
too rare for statistical analysis.
demonstrator type (Trained or Control) on the frequency
of cannibalistic events occurring in the 5 weeks following
Statistical Analysis the demonstration trials (1–5).
We used the mixed linear model (Proc Mixed) of the
RESULTS
SAS Institute (1996), with trial (1–7) as a repeated
measure, to assess effects of demonstrator type (Trained Results of the mixed model analysis on the proportion of
or Control), observer type (O1 or O2), trial and their pairs, and latency, to peck and to pierce the membrane,
interactions on the measured variables. Flock within the quantity of blood eaten and the membrane scores are
demonstrator type, and observer type within flock and shown in Table 1. A higher proportion of observer pairs
 CLOUTIER ET AL.: CANNIBALISM AND SOCIAL LEARNING 1157

1.2 (a)
(a) 600
1.0

0.8
400
0.6

0.4
200
0.2
Proportion

0.0

Latency (s)
1 2 3 4 5 6 7
0
1 2 3 4 5 6 7
1.2
(b)
1.0 (b)
600
0.8

0.6
400
0.4

0.2
200
0.0
1 2 3 4 5 6 7
Trial
0
Figure 1. Mean proportion (±SE) of pairs of chickens in which one or 1 2 3 4 5 6 7
both chickens (a) pecked and (b) pierced a membrane covering a Trial
dish of blood (chicken model) during seven 10-min trials following
demonstration sessions in which they could see: two trained dem- Figure 2. Mean latency (±SE) (s) of the first bird of a pair of chickens
onstrators behind a wire partition piercing the membrane of a to (a) peck and (b) pierce a membrane covering a dish of blood
chicken model and consuming blood (TDO1: ); two trained (chicken model) during seven 10-min trials following one of four
demonstrators with a chicken model in their own cage (TDO2: ); a demonstration sessions (TDO1: ; TDO2: ; CO1: ; CO2: ; see
chicken model behind a wire partition without demonstrators (CO1: Fig. 1). The demonstrator, observer and trial main effects, and the
); or a chicken model in their own cage without trained demon- observer*trial interaction, were significant (mixed linear model on
strators (CO2: ). The demonstrator, observer and trial main effects the ranked data, P<0.05).
were significant for both peck and pierce the membrane, and the
observer*trial interaction was significant for peck the membrane
(mixed linear model on ranked data, P<0.05). 4

from the Trained, than the Control, treatment pecked

Membrane score

and pierced the membrane (Fig. 1). Observer pairs from 3

the Trained treatment pecked and pierced the membrane
with a shorter latency than Control pairs (Fig. 2) and had
higher membrane scores (Fig. 3). The quantity of blood
2
eaten did not differ between pairs from the Trained and
Control treatments (Fig. 4).
A higher proportion of observer pairs with direct access
to the chicken model during demonstrations (O2), rather 1
1 2 3 4 5 6 7
than contact through a wire mesh partition (O1), pecked
Trial
and pierced the membrane when subsequently tested
(Fig. 1). The O2 pairs had shorter latencies to peck and Figure 3. Mean membrane score (±SE) of pairs of observer chickens
pierce the membrane than the O1 pairs (Fig. 2). The O2 during seven 10-min trials following one of four demonstration
sessions (TDO1: ; TDO2: ; CO1: ; CO2: ; see Fig. 1). The
pairs also had higher membrane scores (Fig. 3), and
membrane score could range from 1 (no damage) to 4 (hole(s)
ingested more blood (Fig. 4), than the O1 pairs.
≥5 mm long). The demonstrator, observer and trial main effects
The proportion of pairs that pecked and pierced the were significant (mixed linear model on the ranked data, P<0.05).
membrane, the membrane score and the amount of
blood ingested all increased over time (trials 1–7),
whereas the latency to peck and pierce the membrane proportion of pairs pecking the membrane, latency to
decreased over time (Figs 1–4). There were significant peck and to pierce the membrane and the amount of
interactions between observer treatment and trial on the blood eaten (Table 1, O * T). There were no significant
1158 ANIMAL BEHAVIOUR, 63, 6

15 1.2
(a)
12
Blood eaten (g)

9 0.8

0.4
3

0
1 2 3 4 5 6
Trial
Figure 4. Mean quantity (±SE) of blood eaten (g) by pairs of
observer chickens during seven 10-min trials following one of four
demonstration sessions TDO1: ; TDO2: ; CO1: ; CO2: ; see
Fig. 1). The quantity of blood eaten was not measured in trial 1. The
observer and trial main effects, and the observer * trial interaction
were significant (mixed linear model on the ranked data, P<0.05).
interactions between demonstrator and observer treat-
ments (Table 1, D * O) or between these two treatments
and trials (Table 1, D * O * T).
For all treatment combinations, the logistic model
(Y=A/1+eB
Kx where x is the age of the bird in days)
yielded the best fit to the cumulative proportion of pairs
that pecked the membrane as a function of time for all
conditions (Y=0.9984/1+e(21.0673–0.4679x); Y=1.0121/
1+e(25.4806–0.5661x); Y=1.0068/1+e(23.7920–0.5601x);
Y=1.0286/1+e(12.6117–0.3018x) for Control O1, Control
O2, Trained O1 and Trained O2 conditions, respectively)
(Fig. 5a). The location of the inflection point on the
X axis, calculated as x=B/K, indicates that learning to
peck the membrane occurred significantly (P<0.05)
more rapidly in the Trained O1 and Trained O2
conditions (B/K=42.5, 95% confidence interval,
upper=42.7, lower=42.3 and B/K=41.8, upper=42.4,
lower=41.2, respectively) than the Control O1 (B/K=45,
upper=45.3, lower=44.8) and Control O2 (B/K=45,
upper=45.3, lower=44.8) conditions. Polynomial regres-
sions best fit the cumulative proportion of Trained O1
pairs (Y=0.55048
0.031516x+0.0005407x2) and Con-
trol O1 pairs (Y=
16.4206+1.1018x
0.0245x2 +
0.00018x3) piercing the membrane over time (Fig. 5b).
The logistic model adequately summarized the cumula-
tive proportion of Trained O2 pairs (Y=0.8019/
1+e(16.8116–0.3831x)) and Control O2 pairs (Y=0.4108/
1+e(33.8674–0.7155x)) that pierced the membrane as a
function of time (Fig. 5b). The Trained O2 pairs learned to
pierce the membrane significantly (P<0.05) faster
(B/K=43.9, 95% confidence limits, upper=45.0,
lower=42.8) than the Control O2 pairs (B/K=47.3, 95%
confidence limits, upper=48.0, lower=46.7).
There was no significant effect of the demonstrator
treatment on the frequency of cannibalistic attacks
(N/group/day) occurring in the 5 weeks following the
demonstration trials (F1,16 =2.48, P=0.14), although the
frequency was numerically higher in groups with trained
demonstrators (XSE, Control: 0.070.029 and
Trained: 0.160.041).
Cumulative proportion
7
0.0
43 45 47 49 51 53 55
1.0
(b)
0.8
0.6
0.4
0.2
0.0
43 45 47 49 51 53 55
Age (days)
Figure 5. Best-fit curves for the cumulative proportion of pairs of
chickens that (a) pecked and (b) pierced a membrane covering a
dish of blood (chicken model) during their first five exposures to a
chicken model (between 43 and 55 days of age), each occurring
immediately after one of four demonstration sessions (TDO1: m;
TDO2: ; CO1: _; CO2: ; see Fig. 1).
DISCUSSION
Response to Trained Demonstrators
Across trials, observing trained demonstrators peck and
pierce a membrane to access and consume chicken blood
consistently increased the proportion of observer pairs
performing the same task, and the size of the holes that
they made in the membrane, when tested in the absence
of the demonstrators. In addition, the latency of the
observers to perform the task was consistently lower than
that of the control birds who did not observe trained
demonstrators. Our results are in agreement with those of
Nicol & Pope (1992) who showed that observation of a
trained demonstrator facilitates learning of a novel feed-
ing task in domestic fowl. There are previous reports of
spatial clustering of cannibalistic events in the same and
adjacent cages (Allen & Perry 1975; Craig & Muir 1996;
Tablante et al. 2000) and there has been speculation that
cannibalistic behaviour is socially transmitted (Appleby
et al. 1992; Tablante et al. 2000). Our results provide the
first experimental evidence in support of this hypothesis.
Social learning requires that the naïve observer retains
and reproduces the demonstrator behaviour in the
absence of the demonstrator (Nicol 1995). This condition

was met in our experiment. The pairs of observers per-
formed the task efficiently even in trial 7, at least 36 days
after their last demonstration (in trial 5), indicating that
the birds had retained the task in their long-term
memory.
Even though pairs that had watched trained demon-
strators were faster at pecking and piercing the mem-
brane, there was no significant difference in the quantity
of blood eaten by these pairs and the control pairs. Given
that the birds were provided with a nutritionally com-
plete ration ad libitum, they were probably not motivated
to consume large amounts of blood within the brief time
frame of the tests. Other factors that may have influenced
the amount of blood consumed include aggressive
defence of the blood source by one member of the
observer pair, exploration of the test arena and
fear reactions resulting from being placed in the test
enclosure.
Response to Access to the Chicken Model
The ability to make close contact with the chicken
model during demonstration sessions (O2) increased the
proportion of pairs pecking and piercing the membrane
during subsequent tests compared with the condition in
which the observers were separated from the chicken
model by a wire mesh partition (O1). The opportunity to
interact directly with the chicken model also decreased
the latency to peck and pierce the membrane, increased
the size of holes made in the membrane and increased the
quantity of blood consumed during subsequent testing.
Thus, direct access to the chicken model facilitated acqui-
sition of cannibalistic behaviour towards the model.
There were no significant interactions between exposure
to trained demonstrators and access to the chicken
model. These two factors, thus, had additive effects, with
the greatest learning response occurring in the condition
in which observers had close contact with both the
chicken model and trained demonstrators.
Nicol & Pope (1994) observed a decrease in acquisition
of a key peck response by hens when seven observers were
in physical contact with a single demonstrator and the
apparatus, possibly because only three out of the eight
hens could stand side by side in the response box where
the key was located and, thus, the key was easily defend-
able by one bird. In our study, six birds were together
during observation sessions and the chicken model was
not easily defendable by a single bird. Although we did
not quantify agonistic interactions during observation
sessions, there was no indication that the birds were
competing for access to the chicken model at this time.
Rather, the presence of one or more birds at the model
appeared to increase the confidence of other birds
to approach the model. Contrary to the findings of
Giraldeau & Lefebvre (1987), scrounging does not appear
to have impaired learning in our study since a greater
proportion of pairs from the O2 condition learned the
task than in O1 condition where scrounging was not
possible.
Lefebvre & Giraldeau (1994) reported that the rate of
learning increased with an increased ratio of demon-
CLOUTIER ET AL.: CANNIBALISM AND SOCIAL LEARNING 1159
strators to bystanders. In our O2 condition, once
observers had learned the task, they, too, could be dem-
onstrators for remaining observers, thus increasing the
ratio of demonstrators to observers and, potentially,
increasing the rate at which the task was learned by the
remaining observers. By contrast, in the O1 condition,
observer pairs that had learned the task could not act as
demonstrators for the remaining observer pairs. Another
factor that may have contributed to the difference in
task acquisition between the O1 and O2 conditions was
the opportunity for the birds in the O2 condition to
coordinate their behaviour with demonstrators.
Despite physical separation, pairs observing trained
demonstrators were consistently more likely to peck and
pierce the model than pairs observing the chicken
model alone. Research on chicken vision (Dawkins &
Woodington 1997) suggested that observers would see
the chicken model and demonstrators most effectively if
located within 30 cm of them. The observers in the O1
condition were within that range when standing by the
wire partition. It was apparent during the trials that,
when trained demonstrators were present, O1 observers
stood by the wire partition and watched the demon-
strators. However, they did not appear to pay much
attention to the chicken model when it was presented
alone. Although not investigated in our study, auditory
(McQuoid & Galef 1993) and olfactory (Galef 1996)
contact may have contributed to acquisition of the task.
We minimized the possibility that olfactory cues persisted
beyond each test by washing the chicken model compo-
nents between each use, applying a fresh membrane each
time, changing the paper on the floor and washing off
any blood adhering to birds before returning them to
their home group. Jones & Black (1979) and Jones &
Faure (1981, 1983) suggested that olfaction might play a
role in initial aversion to blood. Our results indicate that
any initial aversion to blood was overcome, especially
when birds had the opportunity to investigate blood in
the presence of flockmates actively consuming blood
from the same source.
Rate of Learning
In all treatments, the proportion of pairs acquiring the
task increased over time, indicating that the cannibalistic
behaviour directed at the chicken model was learned.
Many of the control pairs learned to perform the task
even though they were not exposed to trained demon-
strators. This finding indicates that individual learning
occurred in the absence of social learning. As observer
pairs learned the task, they made bigger holes in the
membrane to access the blood, suggesting that, as the
birds learned about blood as a food source, they became
more motivated and more efficient in accessing the
blood. Apparently the task was easy enough for the
inexperienced birds to learn independently. In future
experiments, a stronger membrane could be used to
increase the difficulty of the task.
Acquisition of the task was enhanced by exposure to
trained demonstrators, indicating that, in addition
to individual learning, social learning was occurring.
1160 ANIMAL BEHAVIOUR, 63, 6
Although social learning was not necessary for the expres-
sion of cannibalistic behaviour, it increased the rate of
diffusion of the behaviour. The learning model of Cavalli-
Sforza & Feldman (1981) predicts sigmoidal diffusion of a
socially transmitted behaviour. Sigmoidal models best fit
our data on pecking the membrane and on diffusion of
piercing behaviour in the O2 condition, whereas poly-
nomial models, as used by Lefebvre (1986), best fit our
data on diffusion of piercing behaviour in the O1 con-
dition. Our data are based on the first bird of an observer
pair to perform the task and are, thus, not affected by
interference and scrounging that could alter the shape of
the learning curve (Lefebvre & Giraldeau 1994).
Performance of the cannibalism task was still high
when the birds were tested in trial 6, at least 8 days after
their last demonstration (in trial 5), and in trial 7, at least
27 days after their most recent previous exposure to the
chicken model (in trial 6). Our results extend those of
McQuoid & Galef (1992) who showed that fowl were
attracted to a feeding site where they had seen con-
specifics feeding 48 h earlier. The enhanced performance
of the task in trials 6 and 7 relative to earlier trials, and the
increased amount of blood ingested, might have been
affected by overnight food deprivation of the birds prior
to these two trials, resulting in an elevated motivation to
consume blood.
Mechanisms of Social Learning
Three processes can be ruled out as possible mech-
anisms of social transmission of the cannibalism task in
our study. Contagion, in which the performance of a
more or less instinctive pattern of behaviour by one
individual acts as a releaser for the same behaviours in
others (Thorpe 1956), operates on an immediate basis.
Although it could have occurred between birds during
observation sessions, it could not have been responsible
for the timing of the first performance of the task by one
observer of a pair when tested in the absence of the
demonstrators. Local enhancement, in which a demon-
strator draws attention to a specific location or object
with which it is interacting (Thorpe 1956), cannot
explain our results because the observation and testing
sessions occurred in different locations and because dem-
onstrators were not present during tests (see Palameta &
Lefebvre 1985; Galef 1988; McQuoid & Galef 1992).
Social facilitation, in which the mere presence of
conspecifics influences the response of an individual
towards a stimulus (Zajonc 1965), was possible in our
experiment since the observers were tested in pairs.
However, it cannot explain differences in learning
between the control pairs and those exposed to trained
demonstrators.
Stimulus enhancement occurs when the observers’
attention is drawn not only towards a particular object
with which the demonstrator is interacting but also to
objects of the same type at other locations and to the
same object at other times (Spence 1937). Stimulus
enhancement would explain enhanced attention of
observers to a chicken model similar to that with which
trained demonstrators were previously observed interact-
ing, when presented to observers in the test enclosure.
It would also explain why the cannibalistic behaviour
did not generalize to real birds. Zeltner et al. (2000)
found that feather pecking increased in groups of
chickens to which feather-pecking chickens (demonstra-
tors) were added compared with groups where
nonfeather-pecking chickens were added. They sug-
gested that feather pecking could have been socially
transmitted through stimulus enhancement. However,
since they did not test their birds in the absence of the
demonstrators or at other locations or times, any effect
of stimulus enhancement on the transmission of feather
pecking cannot be separated from contagion and local
enhancement. The design of our study provides more
definitive evidence for stimulus enhancement in the
social transmission of a deleterious behaviour pattern in
domestic fowl.
Stimulus enhancement alone may not have been suf-
ficient for acquisition of the cannibalism task. Palameta
& Lefebvre (1985) found that observer pigeons were
sufficiently motivated to pierce through a paper lid on a
dish of food only when they saw demonstrators pecking
through the lid and then eating the seeds beneath. The
pairing of the stimulus with a reward may be of crucial
relevance for the observers (Akins & Zentall 1998). The
mechanism by which a demonstrator’s behaviour is
paired with a secondary reinforcer is termed obser-
vational conditioning (Zentall 1996; Fritz & Kotrschal
1999). Chickens have been found to pay attention to the
responses of others towards novel foods (Johnson et al.
1998). Thus, our observers may have learned about a
novel food source (blood) not just through stimulus
enhancement but also by attending to the consumption
of blood by the demonstrators. To investigate this possi-
bility in future research, it would be necessary to vary the
contingency between the demonstrators’ behaviour
towards the model and reinforcement with conspecific
tissue.
Conclusion
Our results suggest that we were successful in designing
an ethical construct for investigating cannibalistic behav-
iour that did not result in generalization to live birds.
Nevertheless, we feel that the results are applicable to
cannibalism in poultry flocks since the same type of
behaviour was involved (damaging pecks followed by
consumption of conspecific blood) although directed to a
different target. Our results indicate that cannibalistic
behaviour is a learned response that is enhanced by
observing other birds engaging in the behaviour. We
contend that stimulus enhancement and, possibly, obser-
vational conditioning are involved in social transmission
of this behaviour. Our results provide the first experimen-
tal evidence that social learning can contribute to the
spread of cannibalistic behaviour in domestic fowl. This
finding has important implications for preventing canni-
balistic behaviour in poultry. Furthermore, given that
cannibalistic behaviour is widespread in the animal king-
dom (Elgar & Crespi 1992), our results suggest that the
social transmission of cannibalistic behaviour could be

occurring in natural populations of animals when
environmental conditions and cognitive aptitude favour
social learning.
Acknowledgments
We thank C. J. Anderson, T. M. Donaldson, A. Sorensen
and R. Trudelle-Schwarz for technical assistance, and L.
Boyd, K. Feldner and B. A. Wheeler for animal care. This
material is based upon work supported by a grant from
the Cooperative State Research, Education and Extension
Service of the United States Department of Agriculture,
under Agreement No. 97-35204-4812, to R. C. Newberry
and C. M. Ulibarri. The research presented here was
evaluated and approved by the Washington State Univer-
sity’s Animal Care and Use Committee (Protocol No.
2674).
References
Akins, C. K. & Zentall, T. R. 1998. Imitation in Japanese quail: the
role of reinforcement of demonstrator responding. Psychonomic
Bulletin & Review, 5, 694–697.
Allen, J. & Perry, G. C. 1975. Feather pecking and cannibalism in a
caged layer flock. British Poultry Science, 16, 441–451.
Appleby, M. C., Hughes, B. O. & Elson, H. A. 1992. Poultry
Production Systems. Behaviour, Management and Welfare.
Wallingford, Oxon: CAB International.
Boyd, R. & Richerson, P. J. 1988. An evolutionary model of social
learning: the effects of spatial and temporal variation. In: Social
Learning: Psychological and Biological Perspectives (Ed. by T. R.
Zentall & B. G. Galef, Jr), pp. 29–48. Hillsdale, New Jersey:
L. Erlbaum.
Cavalli-Sforza, L. L. & Feldman, M. W. 1981. Cultural Transmission
and Evolution: a Quantitative Approach. Princeton, New Jersey:
Princeton University Press.
Cloutier, S., Newberry, R. C., Foster, C. T. & Girsberger, K. M.
2000. Does pecking at inanimate stimuli predict cannibalistic
behaviour in domestic fowl? Applied Animal Behaviour Science, 66,
119–133.
Craig, J. V. & Muir, W. M. 1996. Group selection for adaptation to
multiple-hen cages: beak-related mortality, feathering, and body
weight responses. Poultry Science, 75, 294–302.
Cuthbertson, G. J. 1978. The development of feather pecking and
cannibalism in chicks. Applied Animal Ethology, 4, 292.
Dawkins, M. S. & Woodington, A. 1997. Distance and the presen-
tation of visual stimuli to birds. Animal Behaviour, 54, 1019–1025.
Elgar, M. A. & Crespi, B. J. 1992. Ecology and evolution of
cannibalism. In: Cannibalism. Ecology and Evolution among Diverse
Taxa (Ed. by M. A. Elgar & B. J. Crespi), pp. 1–12. Oxford: Oxford
University Press.
Fritz, J. & Kotrschal, K. 1999. Social learning in common ravens,
Corvus corax. Animal Behaviour, 57, 785–793.
Galef, B. G., Jr. 1988. Imitation in animals: history definition and
interpretation of data from the psychological laboratory. In: Social
Learning: Psychological and Biological Perspectives (Ed. by T. R.
Zentall & B. G., Galef, Jr.), pp. 3–28. Hillsdale, New Jersey:
L. Erlbaum.
Galef, B. G., Jr. 1996. Social enhancement of food preferences in
Norway rats: a brief review. In: Social Learning in Animals. The
Roots of Culture (Ed. by C. M. Heyes & B. G. Galef, Jr.), pp. 49–64.
San Diego: Academic Press.
CLOUTIER ET AL.: CANNIBALISM AND SOCIAL LEARNING 1161
Giraldeau, L.-A. & Lefebvre, L. 1987. Scrounging prevents cul-
tural transmission of food-finding behaviour in pigeons. Animal
Behaviour, 35, 387–394.
Gunnarsson, S., Keeling, L. J. & Svedberg, J. 1999. Effects of
rearing factors on the prevalence of mislaid eggs, cloacal canni-
balism and feather pecking in commercial flocks of loose housed
laying hens. British Poultry Science, 40, 12–18.
Huber-Eicher, B. & Wechsler, B. 1997. Feather pecking in domestic
chicks: its relation to dustbathing and foraging. Animal Behaviour,
54, 757–768.
Huber-Eicher, B. & Wechsler, B. 1998. The effect of quality and
availability of foraging materials on feather pecking in laying hens
chicks. Animal Behaviour, 55, 861–873.
Hughes, B. O. & Duncan, I. J. H. 1972. The influence of strain and
environmental factors upon feather pecking and cannibalism in
fowls. British Poultry Science, 13, 525–547.
Johnson, A. N. B., Burne, T. H. J. & Rose, S. P. R. 1998. Observation
learning in day-old chicks using a one-trial passive avoidance
learning paradigm. Animal Behaviour, 56, 1347–1353.
Jones, R. B. & Black, A. J. 1979. Behavioral responses of the
domestic chick to blood. Behavioral and Neural Biology, 27, 319–
329.
Jones, R. B. & Faure, J.-M. 1981. The effect of age on the responses
of pheasants (Phasianus colchicus) to conspecific blood. Inter-
national Research Communications System Medical Science, 9, 498–
499.
Jones, R. B. & Faure, J.-M. 1983. A further analysis of the responses
of pheasants (Phasianus colchicus) to conspecific blood. Inter-
national Research Communications System Medical Science, 11,
955–956.
Lefebvre, L. 1986. Cultural diffusion of a novel food-finding behav-
iour in urban pigeons: an experimental field test. Ethology, 71,
295–304.
Lefebvre, L. & Giraldeau, L.-A. 1994. Cultural transmission in
pigeons is affected by the number of tutors and bystanders
present. Animal Behaviour, 47, 331–337.
McQuoid, L. M. & Galef, B. G., Jr. 1992. Social influences on
feeding site selection by Burmese fowl (Gallus gallus). Journal of
Comparative Psychology, 106, 137–141.
McQuoid, L. M. & Galef, B. G., Jr. 1993. Social stimuli influencing
feeding behaviour of Burmese fowl: a video analysis. Animal
Behaviour, 46, 13–22.
Nicol, C. J. 1995. The social transmission of information and
behaviour. Applied Animal Behaviour Science, 44, 79–98.
Nicol, C. J. & Pope, S. J. 1992. Effects of social learning on the
acquisition of discriminatory key-pecking in hens. Bulletin of
Psychonometric Sociology, 30, 293–296.
Nicol, C. J. & Pope, S. J. 1993. Food deprivation during
observation reduces social learning in hens. Animal Behaviour, 45,
193–196.
Nicol, C. J. & Pope, S. J. 1994. Social learning in small flocks of
laying hens. Animal Behaviour, 47, 1289–1296.
Nicol, C. J. & Pope, S. J. 1999. The effect of demonstrator social
status and prior foraging success on social learning in laying hens.
Animal Behaviour, 57, 163–171.
Palameta, B. & Lefebvre, L. 1985. The social transmission of a
food-finding technique in pigeons: what is learned? Animal Behav-
iour, 33, 892–896.
Rogers, L. J. & Astiningsih, K. 1991. Social hierarchies in very young
chicks. British Poultry Science, 32, 47–56.
Rushen, J. 1984. Frequencies of agonistic behaviours as measures of
aggression in chickens: a factor analysis. Applied Animal Behaviour
Science, 12, 167–176.
SAS Institute 1996. SAS/STAT Software, Changes and Enhancements
through Release 6.11. Cary, North Carolina: SAS Institute.
SAS Institute 1999. SAS/STAT User’s Guide. Version 8. Cary, North
Carolina: SAS Institute.
1162 ANIMAL BEHAVIOUR, 63, 6
Spence, K. W. 1937. Experimental studies of learning and higher
mental processes in infra-human primates. Psychological Bulletin,
34, 806–850.
Tablante, N. L., Vaillancourt, J.-P., Martin, S. W., Shoukri, M.
& Estevez, I. 2000. Spatial distribution of cannibalism mor-
talities in commercial laying hens. Poultry Science, 79, 705–
708.
Thorpe, W. H. 1956. Learning and Instinct in Animals. Cambridge,
Massachusetts: Harvard University Press.
Zajonc, R. B. 1965. Social facilitation. Science, 149, 269–274.
Zar, J. H. 1996. Biostatistical Analysis. 3rd edn. Upper Saddle River,
New Jersey: Prentice Hall.
Zeltner, E., Klein, T. & Huber-Eicher, B. 2000. Is there social
transmission of feather pecking in groups of laying hen chicks?
Animal Behaviour, 60, 211–216.
Zentall, T. R. 1996. An analysis of imitative learning in animals. In:
Social Learning in Animals. The Roots of Culture (Ed. by C. M. Heyes
& B. G. Galef, Jr.), pp. 221–244. San Diego: Academic Press.