Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/24268390

Behavioral Monitoring in Zoos and Aquariums: A Tool for Guiding Husbandry


and Directing Research

Article  in  Zoo Biology · January 2009


DOI: 10.1002/zoo.20207 · Source: PubMed

CITATIONS READS

60 2,679

3 authors:

Jason Watters Susan W Margulis


San Francisco Zoological Society Canisius College
41 PUBLICATIONS   1,327 CITATIONS    43 PUBLICATIONS   943 CITATIONS   

SEE PROFILE SEE PROFILE

Sylvia Atsalis

40 PUBLICATIONS   869 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Chemical Communication and Musth in Captive Asian Elephants, Elephas maximus View project

Female Gorilla & Chimpanzee Reproductive Aging View project

All content following this page was uploaded by Sylvia Atsalis on 03 November 2019.

The user has requested enhancement of the downloaded file.


Zoo Biology 28:35–48 (2009)

TECHNICAL REPORT

Behavioral Monitoring in Zoos and


Aquariums: A Tool for Guiding
Husbandry and Directing Research
Jason V. Watters,1 Susan W. Margulis,2,3 and Sylvia Atsalis2
1
Chicago Zoological Society, Brookfield Zoo, Brookfield, Illinois
2
Lincoln Park Zoo, Chicago, Illinois
3
Committee on Evolutionary Biology, University of Chicago, Chicago, Illinois

Behavioral monitoring is the scientific collection of animal behavior data to


understand normal patterns of behavior and changes in these patterns. This tool
is underutilized in the zoo industry although it can be an effective indicator of
many potential problems that compromise zoo animal well-being. We suggest
that a behavioral monitoring program should be a core component of a
zoological institution’s care program. We detail the benefits of such a program
and describe its components. We provide guidelines for implementing such a
program and make recommendations that will help institutions to employ
behavioral monitoring programs with reasonable expense. We argue that the
benefits of such a program, primarily increased detection of rising or potential
problems, far outweigh the minor costs of implementation. Zoo Biol 28:35–48,
2009. r 2008 Wiley-Liss, Inc.

Keywords: stereotypy; time budget; science based husbandry; behavioral monitoring

INTRODUCTION
Scientific monitoring is a basic tool widely applied in many disciplines. Overall,
the goal of systematic scientific monitoring is to detect change and understand its
causes. As behavioral biologists who work within zoological settings, we are
Grant sponsors: The Colonel Stanley R. McNeil Foundation; The Women’s Board of the Chicago
Zoological Society; IMLS; Grant number: ]IC-03-07-0113-07.
Correspondence to: Jason Watters, Chicago Zoological Society, Brookfield Zoo, 3300 Golf Road,
Brookfield, IL 60513. E-mail: jason.watters@czs.org
Received 9 January 2008; Revised 13 May 2008; Accepted 13 May 2008
DOI 10.1002/zoo.20207
Published online 17 September 2008 in Wiley InterScience (www.interscience.wiley.com).

r 2008 Wiley-Liss, Inc.


36 Watters et al.

especially interested in the significance of behavioral monitoring for animals in our


care. Behavioral monitoring is essential in establishing a sound understanding of
baseline behaviors, which serve as control states against which changes, should they
occur, are compared. In this article, we identify and discuss the benefits of routinely
implemented behavioral monitoring for the health and well-being of animals in zoos
and aquariums.
Behavioral monitoring is a demonstrated technique that can improve the
welfare of animals in zoos [Kleiman, 1992]. Surveys conducted by Finlay and
Maple [1986], and repeated by Stoinski et al. [1998] confirm that since the 1980s the
vast majority of zoo-based research involves behavior. More recent initiatives
to institute formal protocols for assessing behavior, such as the Methods for
Behavioral Assessment program, have met with some success [Carlstead and
Kleiman, 1998; Carlstead, 2000; Shepherdson and Carlstead, 2001]. Yet, despite its
efficacy, the application of formal data collection as a tool to monitor behavior
continues to be the exception rather than the norm, even as zoological institutions
intensify the use of more expensive and less direct forms of monitoring.
Periodic check-ups and complete physicals are now routine in institutions
accredited by the Association of Zoos and Aquariums (AZA) and similar
organizations. These checks monitor weight and other health indicators with an
eye toward achieving and safeguarding optimal body condition [e.g. Schwitzer
and Kaumanns, 2001]. Endocrine analyses are also useful for diagnosing or
even preventing certain health conditions. Because endocrine techniques are
expensive, they are most often applied to answer specific questions on reproductive
cycling [e.g. Atsalis and Margulis, 2006], or to gauge stress responses to individual
events [e.g. Owen et al., 2004]. Measuring food intake in order to ensure proper
nutritional state is another type of routine monitoring performed by many zoos and
aquariums [e.g. Zinner, 1999]. Overall, these institutions place a premium on the
well-being of the animals in their care, with different types of health-related
monitoring performed to varying degrees depending upon institutional culture and
financial ease.
Results from monitoring animals in our care allow the establishment of
normal, or at least common, levels of key health indices and expressions
of behavior. Therefore, the use of monitoring techniques is not necessarily
hypothesis-driven, nor is implementation specific to addressing known
problems. Baseline information gathered through recurrent monitoring
facilitates the detection of health-related issues, allowing us to predict when
reproductive efforts are likely to be most successful and to understand if
animals are nutritionally fit. Indeed, in the absence of quantified baseline
data, we cannot hope to understand whether variations are occurring outside the
norm—knowledge that is especially important if a change is signaling a potential
problem.
Overall, we believe that behavioral monitoring plays a significant role in
providing information on many important indicators of animal well-being. Behavior
can reveal whether animals are in the appropriate social environment as well as call
attention to health, diet, hormonal or husbandry-related problems [Crockett, 1996].
As physical, social or psychological problems are often associated with subtle
changes in behavior, inspection of behavioral monitoring data has the power to
inform many areas of husbandry.

Zoo Biology
Behavioral Monitoring Programs in Zoos 37

Despite its value, behavioral monitoring is often under-utilized. Interpretation


of behavior is often viewed as subjective and nonquantitative and behavioral
monitoring itself is thought to be time-intensive. Furthermore, there is the
assumption that animal keepers already spend a great deal of their time watching
the animals in their care, and therefore that they would be intimately aware of
behavioral change.
In fact, animal keepers do not necessarily work with particular individuals or
groups of animals over long enough periods to assess changes across varying time
frames (e.g. annual changes). Thus, continuity even in basic, informal observations
may be lacking. The manner used to record observations may also vary considerably.
In addition, well-meaning observers often use inadequate, nonstandardized
techniques to record covariates, such as husbandry conditions and crowd levels.
Haphazard collection of covariate data makes it difficult to determine the causes of
behavioral change. Without systematic recording of behaviors and their covariates,
reliable patterns do not emerge from the data, and the potential for subjective
interpretations of specific behaviors is increased.
As is true for any scientific endeavor, animal behavior monitoring constitutes
an objective enterprise, one that seeks to evaluate the causes of behavior at
proximate and ultimate levels. In addition, behavioral monitoring is important in its
ability to detect and register rare or anecdotal events. Scientifically speaking,
‘‘anecdotal’’ is often synonymous with unreliable. In reality, anecdotal events can be
isolated but valuable indicators of change and well-being. The importance of
anecdotal observations may go unrecognized without regular behavioral monitoring,
which officially registers these rare events, allowing their interpretation to occur
within the broader behavioral context.
Overall, there are many advantages to implementing a behavioral monitoring
program. With a sound behavioral monitoring program in place, the presence of
atypical behavior in zoo animals is easier to recognize [Mallapur, 2005], enrichment
programs can be more easily evaluated [Boinski et al., 1999] and problems can be
more quickly identified and possibly rectified before developing into serious or
expensive concerns. In summary, systematic monitoring of animals in the captive
environment can lead to improvements in their psychological and physical health.
For these reasons, we argue that a regular schedule of behavioral monitoring,
applied broadly, should constitute the cornerstone of any institutional monitoring
program. Below, we describe techniques for the implementation of a behavioral
monitoring program, and demonstrate that it can be performed cost effectively and
with minimal daily time investment.

DEFINING BEHAVIORAL MONITORING


Behavioral monitoring is a means of objectively determining what animals are
doing and what they have been doing, a means of keeping track of animal activity on
a consistent basis. Observers obtain objectivity and reliability by adhering to a
specific set of strict rules for data collection. The rules are easy to follow and trained
observers have little difficulty adhering to them. A fundamental rule is that all
observers understand and agree upon the definitions and manifestations of specific
behaviors. Another rule is that observers must agree on how to collect data, or the
sampling schedule (described in more detail below). We address the need for

Zoo Biology
38 Watters et al.

collective and close adherence to specific rules because otherwise the behavioral
monitoring program will suffer from lack of objectivity, and will lose efficacy.
Without objectivity, it will be impossible to ascribe consistent patterns to the data
collected and the program will lack credibility. Anyone who understands and
adheres to the rules can perform valid and reliable behavioral monitoring.
A successful behavioral monitoring program integrates behavioral data and
information from a number of other sources, including but not limited to husbandry,
dietary and enrichment practices, as well as weather and zoo attendance numbers. In
so doing, we record the circumstances under which behaviors occur. As with any
behavioral observations, this collection of the ‘‘covariates’’ of behavior should be
objective and quantitative. Covariates are evaluated as correlates or determinants of
behavior. A behavioral monitoring program that lacks information on covariates
can only inform us of behavioral change, making it difficult to establish the cause of
the change or even the context within which behavior has changed. The timing of a
behavior and the circumstances surrounding its manifestation provide basic
information toward understanding its occurrence and developing relevant hypoth-
eses to identify causation. As such, the call for a behavioral monitoring program at
any institution should be accompanied by equal emphasis on standardized record
keeping of many other variables.
It is important to recognize that behavioral monitoring is not a subjective
approach to the interpretation of behavior. The behavioral monitor does not ascribe
function or intent to the actions of animals. The observer merely collects data on the
frequency and occurrence of behavior. Subjective analyses of behavior are not a
component of a behavioral monitoring program. Although subjective approaches
have been used successfully to describe animal behavior [e.g. Carlstead et al., 1999],
analysis of these data is always performed in an objective fashion. These techniques
tend to be time intensive and typically they focus on specific questions regarding
animal temperaments. However, a well-designed behavioral monitoring program
based on objective means of data collection can readily inform researchers
investigating animal temperaments.

BENEFITS OF BEHAVIORAL MONITORING


Overall, behavioral monitoring offers many benefits that support optimal
animal care. The fundamental purpose of behavioral monitoring is to provide a
background level of behavioral information to help identify patterns and track
changes. Although it is important not to confuse behavioral monitoring with
hypothesis-guided inquiry, an improved understanding of animal behavior will help
to develop working hypotheses when needed. To illustrate, veterinarians are not
testing a hypothesis when they request regular physical exams; rather, they are using
this information to monitor basic health and well-being, to detect changes that may
signal a possible problem, and to collect baseline information that can incrementally
provide improved indices of what is normal, and reveal deviations from that
standard. Thus, health monitoring parallels behavioral monitoring.
Another benefit that can arise from well-considered behavioral monitoring is
the ability to compare the behavior of animals at many institutions. Although the
benefits of behavioral monitoring accrue without large sample sizes, observing many
animals can provide insights into why some individuals fare better than others. This

Zoo Biology
Behavioral Monitoring Programs in Zoos 39

advantage becomes available only if observers at different institutions collect


comparable data. Using a common set of definitions for the behaviors being
recorded and developing a common understanding of how behaviors are manifested
facilitates the collection of comparable data across institutions.
Behavioral monitoring enhances insights gained through monitoring of
physiological measures. For example, although long-term ongoing hormonal
monitoring is not a common practice, glucocorticoid hormone analysis is regularly
used to monitor stress levels. One study monitored stress indices as a response to
ambient noise over a 4-year period in captive giant pandas [Owen et al., 2004].
Louder noises caused pandas to increase locomotion, scratching and vocalizations,
and the animals also exhibited high levels of urinary glucocorticoids. Only by
combining hormonal and behavioral records was the degree of susceptibility of the
animals to increased noise revealed. Significantly, females were especially sensitive
during estrus and lactation, highlighting the importance of monitoring in avoiding
grave effects on reproductive output.
Behavioral monitoring may be used to evaluate the quantity and variety of
enrichment needed to achieve psychological well-being of animals [e.g. Boinski et al.,
1999]. Data from behavioral monitoring programs can be used to demonstrate
animal behavior to management bodies, or other interested agencies or individuals.
Perhaps the most common benefit of behavioral monitoring is in its use to vet the
opinions of animal husbandry staff by quantitatively demonstrating change or stasis
in animal behavior. This particular use of behavioral data appears to garner
behavioral monitoring programs the most institutional support.

THE ROLE OF COVARIATES


When developing a behavioral monitoring program, one should consider
which covariates are likely to be important and measurable. For example, as sound
levels are shown to affect the behavior of giant pandas [Owen et al., 2004; Powell
et al., 2006] and likely affect the behavior of other species as well [Morgan and
Tromborg, 2007], it is reasonable to include the ambient noise level as a covariate
that potentially influences animal behavior. However, measuring sound levels
accurately requires specialized and costly equipment. Thus, institutions interested in
correlating the behavior of animals with ambient noise must determine the best way
to represent noise in their data set. Visitor attendance numbers can serve as an
appropriate proxy for sound levels, although this relationship should be validated
with borrowed or rented sound measuring equipment before deployment. Obviously,
a correlation between attendance and noise does not necessarily imply that visitors
cause noted behavioral changes. Additional potential covariates of behavior may
arise from weather conditions, diet and husbandry changes, the keeper staff on duty,
animal social groupings, and many other possibilities.
When establishing the covariates for data collection, researchers should also
consider the means by which they will manage animal behavioral and covariate data
for analysis. The program administrator should facilitate the analysis of behavioral
change by making it easy to compare behavior at times when different values of
covariates are measured. To this end, a single database that houses both behavioral
and covariate data is a simple solution. The database can be queried for analyses that
examine specific times of interest or changes in husbandry practices or other factors.

Zoo Biology
40 Watters et al.

A final consideration when determining which covariates to gather is that factors


that do not vary are not useful in analyses. To perform useful analyses, one must be
able to compare alternative states of specific factors to determine whether changes in
these states relate to changes in behavior.

DEVELOPING AN ETHOGRAM
An ethogram is a list of behaviors that a particular species can exhibit. For a
given behavioral study, the ethogram used may be exhaustive, sometimes including
very subtle behaviors that occur at low frequencies. Alternatively, the ethogram can
be streamlined, including only primary behaviors that comprise the majority of an
animal’s time budget. The use of an exhaustive or a streamlined ethogram will
depend on the data of interest to the behavioral monitor.
For most behavioral monitoring programs, the measures of interest are the
animal’s time budget and the frequency of occurrence of selected behavioral events.
The time budget is the itemization of how an animal spends its day, and surprisingly
few behaviors need be recorded to develop a useful time budget. Bar graphs usually
represent time budgets, with the proportion of time that an animal spends exhibiting
a given behavior plotted alongside the proportion of time spent in other behaviors
(Fig. 1). Not surprisingly, an animal’s time budget varies seasonally and even over
the course of a single day. Behavioral monitors look for meaningful changes in
animal time budgets to focus their inquiry into potential problems. Useful time
budgets can develop with the use of either an exhaustive or a streamlined ethogram.
Under most circumstances, a streamlined ethogram will be sufficient to provide
data on basic activity patterns for behavioral monitoring. Use of an exhaustive
ethogram may be most appropriate when we know little about the natural history of
a species. In this case, observers develop the ethogram through careful behavioral
observation. Another possibility is that managers are interested in detailed time
budgets. For example, we may wish to know not only the proportion of time that an
animal is active, but also how much of that time is spent walking vs. running.
Exhaustive ethograms are used by many institutions to monitor the degree to which

70
animal A
60
animal B
50

40
% time

30

20

10

0
inactive locomote forage social aggression other
behavior

Fig. 1. A sample time budget representing two individuals (animal A and animal B). Note
that even with a very simple ethogram (in this example only six behaviors), clear patterns are
evident and differences between individuals apparent.

Zoo Biology
Behavioral Monitoring Programs in Zoos 41

aggression occurs in social groups. In this case, an ethogram detailed in terms of the
various types of aggression (i.e. threat, chase, bite, etc.) is very helpful when trying to
quantify the relative threat of injury. Notably, an ethogram can be expanded in one
functional context, such as aggression, and streamlined in another, such as
locomotion.
Streamlining an ethogram is useful for a number of reasons. Monitors may be
specifically interested in a particular type of behavior. For example, there is often an
interest in documenting animal visibility and activity in exhibits. In this case, the
ethogram may contain only these behaviors: visible and active, visible and not active,
not visible. However, with this streamlined focus of observation, behavioral
monitoring may not detect subtle changes in animal time budgets, offering little in
the way of preventative power, if heading off potential health and social problems is
a goal. Nevertheless, visibility monitoring is a valuable enterprise for some
institutions. Moreover, a streamlined ethogram is useful for maintaining strong
inter-observer reliability. When behaviors are compacted into primary functional
groups or ‘‘higher-order behaviors,’’ such as locomotion, social behavior, etc.,
observers can classify behavior more efficiently than when they are tasked with more
detailed behavioral classifications.
The level of detail included in an ethogram is a critical issue that requires
careful attention if the ethogram is to be widely accepted and applicable. As
mentioned above, while a surprisingly small number of behaviors are generally
necessary to account for most of an animal’s time, there may be particular situations
in which greater detail is required. Consequently, many researchers and animal
managers develop their own ethograms that often result in the inability to share and
compare information. One strategy for resolving this challenge is to structure an
ethogram such that the level of detail can be determined on a case-by-case basis,
while at the same time adhering to a consistent framework. A hierarchical ethogram
facilitates this process (Fig. 2). For example, an ethogram may contain a relatively
small set of higher-order behaviors, each of which can be broken down to a finer
level of detail.

High Order
Behavior Activity

Functional
Forage Locomote
Behavior

Detailed Crawl Walk Brachiate


Behavior

Fig. 2. An example that graphically depicts a hierarchical ethogram. At the top is the high-
order behavior marked as activity. This behavior is a very general one. Several functional
behavior categories can comprise the high-order behavior. In this example, the functional
behaviors that comprise activity are forage and locomote. An additional level of detail can be
added to the functional behaviors. No detailed behaviors were added for forage, but locomote
is broken down to include crawl, walk, and brachiate. While forage and brachiate are both
types of activity, they are dissimilar at the functional level.

Zoo Biology
42 Watters et al.

In theory, even detailed behaviors can be further subdivided to focus for


instance on specific body postures or limb positions. In all situations though,
detailed behaviors can be collapsed upwards to the higher-order categories. In this
way, even data collected at different levels of detail, for different purposes, or that
focus on different behavioral issues, are comparable at the highest-order level. To
share data across institutions then, securing agreement on terminology and
definition for these higher-order behavioral categories becomes pivotal. Efforts to
do so have met with success [Atsalis et al., 2005]. Guidelines for shareable behavioral
databases have been developed [Martins, 2004].
The key benefit of developing a taxon-specific ethogram that is agreed upon by
the zoological community is the enhanced ability to share information. In the
absence of such agreement, animal managers seeking to understand the behavior of
the animals in their care are, by definition, limited to small sample sizes, anecdotes,
and case studies.
Efforts to address the issues of small sample size have resulted in tremendous
growth in the area of multi-institutional studies. As the extent and scope of multi-
institutional studies expands, the use of common ethograms and terms becomes
more prominent. Multi-institutional studies are conducted by single researchers
visiting many institutions, or by training staff at different locations to collect data
[Mellen, 1991; Powell et al., 2008; Margulis and Pruett-Jones, 2008]. To develop a
behavioral monitoring program, the latter is a necessity. Moreover, whereas most
multi-institution studies are hypothesis-driven, the application of behavioral
monitoring at the multi-institution level does not require specific hypotheses. If
zoos use a common ethogram structure and apply this framework for routine
behavioral monitoring, the opportunity to use the information for research and
management purposes increases.

MULTI-INSTITUTION STUDIES
When developing a multi-institutional behavioral monitoring program, it is
inevitable that numerous variables, including housing conditions, will influence
animal behavior. The plethora of variables that may influence the expression of
behavior can easily become a hindrance to understanding the causal relationships
that underlie behavior. The difficulty arises when one tries to understand a multi-
variate data set with statistics not designed to handle these types of data. We argue
that given the proper statistical analyses and appropriate data collection techniques,
the inevitable excess of covariates can be an advantage for researchers seeking to
understand the responses of animals in different environments.
An example of a behavioral monitoring study that produced useful results by
comparing animal behavior across institutions comes from Mallapur [2005]. In one
study, the behavior of lion-tailed macaques was monitored at 12 institutions. The
analysis considered animal rearing history and zoo enclosure design to elucidate
factors that are associated with the expression of stereotypic behavior. Although
individual institutions vary in terms of the environments available to animals, there
is likely to be overlap of variable types across many institutions. In the above
example, Mallapur was able to classify macaque enclosure types into three categories
that additionally varied at three levels of complexity. This simplification was
informative in identifying potential causation of behavioral differences.

Zoo Biology
Behavioral Monitoring Programs in Zoos 43

Even with multi-institutional studies there may continue to be sample size


issues as only a few animals are likely to be housed at any one institution, usually
under identical conditions, so they may not be statistically independent. Also
multi-institutional studies often involve a larger number of independent variables
or covariates, and properly analyzing these requires a larger sample size. However,
as variables are likely to overlap across institutions—for example, zoos with
exhibit space larger than X and those with exhibit space smaller than X, or zoos that
house their animals indoors at night and those that do not, etc.—many issues
associated with small sample size will be eased [Kuhar, 2006; Margulis and Westhus,
2007].

HOW TO COLLECT BEHAVIORAL MONITORING DATA


The most commonly used methodologies for behavioral data collection are
continuous and instantaneous (or scan) sampling [Altmann, 1974]. The former
requires recording all behaviors of a single individual, noting time of onset of each
behavior. The latter collects data at pre-set intervals. In a standard research
paradigm, instantaneous sampling would require collection of a ‘‘point’’ of data at
specific, pre-determined intervals—for example, every 30 sec, or every 5 min.
Allocating substantial blocks of time to data collection is not a fundamental
requirement of either methodology. In fact, for the purposes of behavioral
monitoring, regular data collection is more important than collecting data for
single large blocks of time. Data collected regularly, even for as little as 10 min at a
time, can soon produce a respectable picture of how subjects spend their time. Even
if observers are unable to spend a continuous block of time watching animals, they
can collect valuable data with several quickly obtained scan samples spaced
throughout the day. Results obtained in this way provide valuable information
about time budgets [Margulis and Westhus, 2007].
At most institutions, animal care staff [Platt et al., 2002; Margulis and Weber,
2003] or researchers tasked with observing a broad spectrum of species, act as
observers. Thus, our vision of behavioral monitoring operates on the assumption
that observer time is extremely limited and in most situations we suggest utilizing an
instantaneous method [Margulis et al., 2008]. Although at times it may be possible to
identify a dedicated block of time (even a short one) for behavioral observations, in
reality a greater degree of flexibility may be required if animal care staff are to collect
data. Thus, an animal keeper that walks by an animal’s enclosure 8–10 times daily
can easily record behavioral data that soon begin to describe realistic time budgets.
Paper-and-pencil data collection may be appropriate for short-term investiga-
tions but this tool becomes less convenient for long-term efforts because the time
required for data entry is often prohibitive. Additionally, summarizing and
interpreting raw data may not fall within the purview of all observers. Hence, tools
for streamlining the data collection and interpretation process are necessary if
behavioral monitoring programs are to gain wide acceptance and use. Although
there are a number of commercially available data collection systems [Margulis,
2006], these are often costly or require substantial customization for full
effectiveness. Recognizing that the absence of inexpensive, automated tools posed
an obstacle to expanding the implementation of behavioral monitoring, in 2000 at
Brookfield Zoo the development of a data collection tool designed specifically for

Zoo Biology
44 Watters et al.

behavioral monitoring was initiated. The Colonel Stanley R. McNeil Foundation’s


EthoTrak Observation System (EthoTrak) uses an inexpensive hand-held computer
for data collection, and incorporates a user interface that provides immediate
graphical feedback and summary statistics on the data collected. Use of the system
requires no specific expertise in data analysis, statistics, or database maintenance.
The implementation of behavioral monitoring using EthoTrak was gradually
expanded at Brookfield Zoo over a period of several years during which time,
behavioral research staff trained animal care staff in use of the system. Following
development of the program the zoo invested in a small number of PDA’s. Ethotrak
staff provided ongoing technological support in order to train animal care staff in its
use. Once staff were comfortable with the technology, formal inter-observer
reliability tests were conducted to ensure consistency. This stage generally took
several months, after which expansion to all keeper staff, and in some cases volunteer
observers as well, was instituted. The flexibility inherent in the tool allows animal
keepers to fit observations in as a part of their daily routines.
Developing a successful behavioral monitoring program with the right tools is
relatively easy, yet there is no replacement for well-trained observers who follow closely
the agreed-upon rules of observation. Without question, the greatest expense involved
with a behavioral monitoring program is the time involved in observer training and
maintaining inter-observer reliability. Observers record behavior at the prescribed
interval and are not disappointed when a ‘‘more interesting’’ behavior occurs just
before or just after the time when behavior is to be recorded. Observers who record
behaviors at times when they should not be recorded create erroneous data sets that
result in skewed time budgets where interesting or rare behaviors are represented at
proportions much higher than normal. Observers who influence the behavior of the
animals they are watching also create biased data sets. In some circumstances it may be
easier to videotape behavior or to watch it on a closed circuit television to avoid
observer effect on the data. Although a reliable observer does not need to be a species
expert, observers do need to follow the rules and strive for inter-observer agreement.
Consequently, it is critical that all observers in a given monitoring program observe
animals in a similar fashion. A goal of 90% inter-observer agreement is easily obtained
once observers are properly trained, and guidelines for measuring inter-observer
agreement are available [Lehner, 1996]. The importance of maintaining a high degree
of inter-observer reliability cannot be overstated. Monitoring programs should
maintain observer reliability with regular checks.

RESULTS FROM BEHAVIORAL MONITORING


We have found that enabling observers to see the fruits of their labor becomes
a positive feedback loop: seeing the quantitative patterns that develop when
behavioral data accumulate encourages observers to gather more data. Simple time
budget and behavioral frequency graphs are easy to generate, and in fact are
generated automatically by data collection software. These figures stimulate the
interest of data collectors and managers alike to continue data collection and to
consider whether husbandry changes are necessary. Individuals managing behavioral
monitoring programs should strive to generate monthly or quarterly reports that
graphically depict animal behavior. These reports should be distributed to curators
and animal care staff alike.

Zoo Biology
Behavioral Monitoring Programs in Zoos 45

Those managing a behavioral monitoring program should also be aware of


statistical approaches by which they can analyze the data to confirm or refute shifts
from a known baseline. Valid statistical approaches for small sample sizes include
randomization tests [Todman and Dugard, 2001], nonparametric analyses [Siegel
and Castellan, 1988] and other tests [for an overview, see Kuhar, 2006].
Randomization tests are perhaps most suited to many analyses as they do not
attempt to extrapolate from samples to populations. Instead, they identify
differences in the sample set [Edgington and Onghena, 2007]. Statistics users should
take care to understand the assumptions and power of any test they intend to
employ. In many cases, statistical analyses will not be necessary in order to note
changes in behavior. Comparisons of time budget graphs and average values and
standard deviations will often suffice to point managers to areas needed for deeper
investigation. Those working with data should pay attention to the number of
observations that comprise the data set of interest as it is easy to suggest that there
has been a change in the composition of the behavioral repertoire—even if there has
not—when comparing unequal or small numbers of observations.
Behavioral monitoring can help us to understand individual animals in the
context of the rest of the captive population, but in many cases institutions have
access only to their own data and are able to generate reports based on a limited
number of subjects. Yet, even a restricted data set collected through behavioral
monitoring can be a valuable tool when it includes baseline information. The
importance of baseline data became evident during a 6-year monitoring program of
female tigers [Miller and Kuhar, 2008]. From regular comparisons of current to
baseline data, the program showed that aggressive behaviors increased the longer the
group remained together. Because female tigers do not normally live in large groups,
the monitoring program gave insight into behavior when information was lacking
from other sources, allowing keepers to be flexible in their animal management
practice.
Eliminating or reducing stereotypic behavior is essential in our quest to achieve
optimal animal care and health. Behavioral monitoring can be instrumental in
assessing the value of particular enrichment regimens on animal well-being. In one
study, the presence of a manipulable object, woodchip bedding, was shown to
significantly reduce undesirable behaviors in captive chimpanzees [Brent, 1992]. The
ability to evaluate the outcome of specific behavioral management techniques
showcases the value of pursuing baseline behavioral data through a behavior
monitoring program. Changes in feeding strategies and environmental and social
manipulations are interventions whose efficacy can be quantitatively assessed only in
comparison with pre-treatment–baseline behavior. Thus, behavioral monitoring is
an essential component to behavioral management approaches [Laule, 1993].

CONCLUSIONS
Identifying the circumstances under which particular behaviors occur allows us
to establish behavioral patterns. Collection of baseline behavioral data is an essential
step in attempts to resolve complicated behavioral issues. It is a key to information
gathering, accelerating the process of resolving behavioral and other time-intensive
issues [see Laule, 1993]. Behavioral monitoring is the essential tool to meet
husbandry and animal conservation goals. The tool is most effective when observers

Zoo Biology
46 Watters et al.

follow the rules of observation and it reaches its full potential when data are viewed
on a regular basis. If collected data suggest husbandry changes, further analyses may
indicate where to begin making modifications. Effective intervention using the data
from behavioral monitoring may be the best proof of a successful behavioral
monitoring program. These programs are relatively easy and inexpensive to launch,
though care should be taken to ensure that they are adequately supported in order to
maintain quality data collection and analysis over time. It is important to
appropriately design any monitoring program before it is implemented. Gathering
enough data to assist in answering questions that may arise is essential, but hyper-
collection of data can be a waste of resources.
An integrated monitoring system, one that incorporates behavior, health,
hormones, and a reasonable array of environmental variables, is the best way to
ensure animal welfare. Ideally, this kind of a system would include a composite
database with all of these components. With this system in place, the likelihood of
developing an early response to potential problems is increased. Workers who
participate in behavioral monitoring will also benefit from an ability to generate and
test reasoned hypotheses from the data collected via monitoring.
Finally, it is important to recognize that, specifically with zoo animals, it is
difficult to assume average behavior of a species. Considerable individual variation
in behavior means that members of the same species do not always act similarly.
There is much interest in the individual behavioral types of animals in zoo
collections [e.g. Gold and Maple, 1994; Wielebnowski, 1999] because individuals that
express alternative behavioral types do not respond in the same ways to the same
stimuli. Thus, individuals with alternative behavioral types are likely to require
individual-based management plans in order to achieve optimal well-being. By
deciding upon the appropriate behavioral covariates, we can gain a deeper
understanding of individual behavioral types through the data collected via
behavioral monitoring. Over time, we will know the relative consistency of
individual responses to different types of events, as well as their general tendencies
to behave in particular ways (i.e. whether they are consistently active, aggressive,
etc.). From these types of data, we can design management plans for individuals,
rather than focus on the average response of a species. One benefit of such an
enterprise is the ability to choose the most appropriate potential mates based on
behavioral as well as genetic compatibility. Another is establishing how social
group structure accounts for individual variation and the roles that individuals play
within the group (at a level much more intricate than simple considerations of
demographic features such as age and sex). Yet another benefit is the informed
ability to modify habitats to support as adequately as possible the individuals
residing in them.

ACKNOWLEDGMENTS
J. W.’s research and monitoring program is funded in part by The Colonel
Stanley R. McNeil Foundation, The Women’s Board of the Chicago Zoological
Society, and IMLS grant ]IC-03-07-0113-07. We would like to thank Todd Oakley
and Zaldy Apura for helping us to figure out reasonable ways to collect data
electronically. We are indebted to the late Allison Walsh, who initiated the original
EthoTrak programming, and dedicate this article to her memory.

Zoo Biology
Behavioral Monitoring Programs in Zoos 47

REFERENCES
Altmann J. 1974. Observational study of behavior: Mallapur A. 2005. Managing primates in zoos:
sampling methods. Behaviour 49:227–265. lessons from animal behaviour. Curr Sci India
Atsalis, S, Margulis SW. 2006. Sexual and 89:1214–1219.
hormonal cycles in geriatric Gorilla gorilla Margulis S. 2006. Behavior meets technology: data
gorilla. Int J Primatol 27:1663–1687. collection tools and software. Workshop, Animal
Atsalis S, Kasnicka C, Margulis S, McGee J, Behavior Society Meeting, Snowbird, UT, Au-
Pruett-Jones, M. 2005. EthoTrak: lessons learned gust 12–16, 2006.
in an electronic behavioral monitoring program. Margulis SW, Pruett-Jones M. 2008. Integrating
American Zoo and Aquarium Association An- science and husbandry: less is more. In: Bettinger
nual Conference Proceedings: 5 p. T, Bielitzki J, editors. The well-being of animals
Boinski S, Swing SP, Gross TS, Davis JK. 1999. in zoo and aquarium sponsored research: putting
Environmental enrichment of brown capuchins best practices forward. Greenbelt, MD: Scientists
(Cebus apella): behavioral and plasma and fecal Center for Animal Welfare. p 25–38.
cortisol measures of effectiveness. Am J Primatol Margulis SW, Weber TM. 2003. Keepers and
48:49–68. behavioral research: fostering the connection.
Brent L. 1992. Woodchip bedding as enrichment Anim Keepers’ Forum 30:285–287.
for captive chimpanzees in an outdoor enclosure. Margulis SW, Westhus EJ. 2007. Evaluation of
Anim Welfare 1:161–170. different observational sampling regimes for use
Carlstead K. 2000. Methods of behavioral assess- in zoological parks. Appl Anim Behav Sci
ment: constructing behavioral profiles for zoo 110:363–376.
animals—incorporating behavioral information Margulis SW, Atsalis S, Watters J. 2008. Methods
into captive population management. AZA and for behavioral monitoring of zoo primates. Am J
Oregon Zoo: Behavioral Husbandry Advisory Primatol 70:22.
Group. 75 p. Martins E. 2004. EthoSource: storing, sharing, and
Carlstead K, Kleiman DG. 1998. Behavior profiles combining behavioral data. BioScience 54:886–887.
of individual animals: a new tool for multi- Mellen JD. 1991. Factors influencing reproductive
institutional studies. American Zoo and Aqua- success in small captive exotic felids (Felis spp.): a
rium Association Annual Conference Proceed- multiple regression analysis. Zoo Biol 10:95–110.
ings. p 39–43. Miller A, Kuhar CW. 2008. Long-term monitoring
Carlstead K, Mellen J, Kleiman DG. 1999. of social behavior in a grouping of six female
Black rhinoceros (Diceros bicornis) in U.S. tigers (Panthera tigris). Zoo Biol 27:89–99.
Zoos 1. Individual behavior profiles and their Morgan KN, Tromborg CT. 2007. Sources of stress
relationship to breeding success. Zoo Biol in captivity. Appl Anim Behav Sci 102:262–302.
18:17–34. Owen MA, Swaisgood RR, Czekala NM, Stein-
Crockett CM. 1996. Data collection in the zoo man K, Lindburg DG. 2004. Monitoring stress in
setting, emphasizing behavior. In: Kleiman DG, captive giant pandas (Ailuropoda melanoleuca):
Allen ME, Thompson KV, Lumpkin S, editors. behavioral and hormonal responses to ambient
Wild mammals in captivity. Chicago: University noises. Zoo Biol 23:147–164.
of Chicago Press. p 545–565. Platt, JA, Brown-Palsgrove M, Ross SR. 2002.Ot-
Edgington ES, Onghena P. 2007. Randomization ter enrichment and the benefits of keeper
tests, fourth edition. New York: Chapman & involvement in behavioral research. Anim Kee-
Hall/CRC. 345 p. pers’ Forum 29:457–460.
Finlay TW, Maple TL. 1986. A survey of research Powell DM, Carlstead K, Tarou LR, Brown JL,
in American zoos and aquariums. Zoo Biol Monfort SL. 2006. Effects of construction noise
5:261–268. on behavior and cortisol levels in a pair of
Gold KC, Maple TL. 1994. Personality assessment captive giant pandas (Ailuropoda melanoleuca).
in the gorilla and its utility as a management Zoo Biol 25:391–408.
tool. Zoo Biol 13:509–522. Powell, DM, Hong L, Carlstead K, Kleiman D,
Kleiman DG. 1992. Behavior research in Zhang H, Zhang G, Zhang Z, Yu J, Zhang J G,
zoos: past, present, and future. Zoo Biol Lu Y, Ng TS, Tang JCL, Snyder R.
11:301–312 2008. Relationships between temperament,
Kuhar CW. 2006. In the deep end: Pooling husbandry, management, and socio-sexual beha-
data and other statistical challenges of vior in captive male and female giant pandas
zoo and aquarium research. Zoo Biol 25: (Ailuropoda melanoleuca). Acta Zoo Sin 54:
339–352. 169–175.
Laule G. 1993. The use of behavioral management Schwitzer C, Kaumanns W. 2001. Body weights of
techniques to reduce or eliminate abnormal ruffed lemurs (Varecia variegata) in European
behavior. Anim Welfare Information Center zoos with reference to the problem of obesity.
Newslett 4:1–11. Zoo Biol 20:261–269.
Lehner PN. 1996. Handbook of ethological Shepherdson D, Carlstead K. 2001. New ap-
methods, 2nd edition. Cambridge, England: proaches to the evaluation of zoo animal well-
Cambridge University Press. 672 p. being using multiple institutions, assessment of

Zoo Biology
48 Watters et al.

behavior and temperament, and non-invasive Todman JB, Dugard P. 2001. Single-case and
physiological measures. American Zoo and small-n experimental designs: a practical guide to
Aquarium Association Annual Conference Pro- randomization tests. Mahwah, NJ: Lawrence
ceedings. p 131–136. Erlbaum Associates, Inc. 245 p.
Siegel S, Castellan NJ. 1988. Nonparametric Wielebnowski NC. 1999. Behavioral differences as
statistics for the behavioral sciences, predictors of breeding status in captive cheetahs.
2nd edition. New York: McGraw Hill. Zoo Biol 18:335–349.
399 p. Zinner D. 1999. Relationship between feeding time
Stoinski TS, Lukas KE, Maple TL. 1998. A survey and food intake in Hamadryas baboons (Papio
of research in North American zoos and aqua- hamadryas) and the value of feeding time as
riums. Zoo Biol 17:167–180. predictor of food intake. Zoo Biol 18:494–505.

Zoo Biology

View publication stats

You might also like