Applied Animal Behaviour Science 110 (2008) 377–385

www.elsevier.com/locate/applanim

Group differences in captive gorillas’ reaction

to large crowds
Christopher W. Kuhar *
Disney’s Animal Kingdom, P.O. Box 10000, Lake Buena Vista, FL 32830-1000, United States
Accepted 10 April 2007
Available online 18 May 2007

Abstract
The impact of visitor crowd size on captive animal behaviour can be difficult to interpret in many
institutions due to the lack of variation in crowd size over short periods of time. In attempts to compare
greater variation in crowd size, animal behaviour is often compared over conditions that create additional
confounds, such as day of week or season. Fluctuations in attendance over the holiday season at Disney’s
Animal Kingdom1 Theme Park provided an opportunity to examine the impact of variation in crowd size
on gorilla behaviour without the confounds found in many other studies. Ten western lowland gorillas
(Gorilla g. gorilla) in a bachelor group (n = 4) and a family group (n = 6) were observed over a period of
8 weeks in late 2005. Observations were classified into Large Crowd and Small Crowd days and a
repeated-measures ANOVA procedure was conducted to determine the effects of crowd size and social
group on gorilla behaviour patterns. Few overall differences were found in behaviour during the Large
Crowd and Small Crowd conditions. However, similar to previous research, gorillas were less visible
during the Large Crowd condition (F 1,8 = 14.15, P = 0.01). There was also an interaction of crowd size
and group (F 1,8 = 7.58, P = 0.01), indicating the bachelor group of gorillas was more aggressive during
the Large Crowd condition, whereas the family group showed no increase in aggression with large
crowds. These results indicate the importance of providing complex enclosures with visual barriers to
allow animals to move away from large crowds if they choose. Future research should focus on the
individual characteristics of animals and enclosures that may contribute to behavioural differences in
visitor–animal interaction research, as well as the proximate cues associated with behaviour change when
exposed to large crowds.
# 2007 Elsevier B.V. All rights reserved.

Keywords: Gorilla; Human–animal relationships; Bachelor groups; Agonistic behaviour; Social behaviour; Zoo animals

* Tel.: +1 407 938 3205; fax: +1 407 938 1982.

E-mail address: Chris.Kuhar@Disney.com.

0168-1591/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.applanim.2007.04.011
378 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385

1. Introduction

Along with the growing concern for animal welfare in zoological parks, over the past several
decades has come increasing interest in the impact of zoo visitors on the behaviour and welfare of
zoo animals. Despite the assumption that visitors may impact animal behaviour in some manner,
the research results are somewhat ambiguous. The impact of visitors has been alternatively
viewed as stressful, enriching, and of no consequence to zoo-housed animals (see Hosey, 2000
for review). Additionally, the presence of visitors has been argued to be more of a condition than a
variable and that some other specific factor (e.g. visitor activity or noise) is the proximate factor
affecting animal behaviour (Hosey, 2005).
Hosey (2000) has discussed the major hypotheses associated with the visitor effect. Despite
this commentary, two factors are under-explored in the visitor-effect literature. The first is
individual differences of the animals being studied. While some studies have indicated
differences in aggression based on age or sex (e.g. Mitchell et al., 1991), these are rarely factored
into summative discussions of the impact of zoo visitors on the behaviour of a species. Thus,
significant results, or the lack thereof, may be driven by the demographic makeup of the group, as
individual animals may be more or less susceptible to the effects of visitors.
The second factor that may impact studies of the visitor effect is research methodology.
Research on the visitor effect can be broadly divided into two classes: (1) instantaneous
evaluation and (2) daily evaluation. In instantaneous evaluation, both animal behaviour and
visitor characteristics (e.g. density and intensity) are evaluated throughout the day using an
interval sampling technique (Martin and Bateson, 1993). This method allows for the assessment
of current crowd size at an exhibit and provides a variable measure of crowd size throughout the
day. However, this method does not allow for analysis of cumulative effects. Using instantaneous
evaluation, an interval of low visitor density that was preceded by several hours of high visitor
density is weighted equally to an interval of low visitor density that has followed several hours
without a single visitor. Successive intervals are not independent of one another, and interval
evaluation methods do not account for this lack of independence.
Additionally, the instantaneous method is subject to a confound of visitor effect and visitor
attraction (Hosey, 2000). Large crowds may be attracted to aggressive behaviour displayed in an
exhibit, as opposed to the large crowds causing the behaviour, as many visitor-effect studies
assume. When instantaneous evaluation is used it can be difficult, if not impossible, to
completely separate the cause and effect relationship of crowd size and behaviour.
Alternatively, daily evaluation uses the number of individuals to enter the park or pass in front of
an exhibit throughout the day as the independent variable in the analysis. This methodology
controls for cumulative effects and can be used to avoid the visitor-effect/visitor-attraction
confound. However, the daily evaluation technique does not provide any information on specific
causes of behavioural change related to crowds or specific aspects of the animal–visitor interaction.
Using this methodology one cannot determine if it is, for example, visitor numbers, noise, activity,
or the increased chance of a single harassing visitor that causes behavioural change.
Given these advantages and limitations, daily evaluations are desirable to determine whether
crowd size is related to behavioural change and instantaneous evaluations provide more in-depth
information on the relationship between visitors and animals. Unfortunately, many zoos do not
have sufficient variation in daily attendance over a short period of time to adequately assess the
impact of crowd size. As a result, researchers are forced to compare weekdays with weekends or
summer versus winter attendance figures (e.g. Wells, 2005), thereby ignoring the potential
confounds of work schedule or weather/season.
 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385 379

The gorilla groups at Disney’s Animal Kingdom1 Theme Park provide an opportunity to
explore these issues. Disney’s Animal Kingdom1 Theme Park houses both a family group and a
bachelor group of gorillas. This provides the opportunity to examine differences in reactions to
crowds based on group characteristics. Second, at a theme park, attendance patterns fluctuate
greatly during the holiday season, providing an opportunity to assess the impact of daily variation in
attendance without the confounds of season or day of the week. Third, the exhibits are located near
the end of a one-way walking trail. Because visitors cannot see the gorillas when they enter the trail,
the number of visitors that pass the gorilla exhibit is not related to current gorilla behaviours. Thus,
behavioural change can be assumed to be caused by some aspect of the differences in crowd size
instead of different crowd sizes being attracted to different behaviour patterns.

2. Methods

2.1. Subjects

Subjects were 10 western lowland gorillas (Gorilla g. gorilla) housed in two groups at Disney’s Animal
Kingdom1 Theme Park: a bachelor group of four adult males (age: 24, 14, 14 and 12 years), and a family
group consisting of a single adult male (age: 24 years), two adult females (age: 34 and 22 years), two
juvenile males (age: 10 and 7 years), and a juvenile female (age: 6 years). Each group is housed in a large
(>0.4 ha) outdoor exhibit situated at the end of the Pangani Forest Trail. Each exhibit contains dense plant
material, rock outcroppings, and other visual barriers. These exhibits can be viewed from multiple locations
along an approximately 130 m section of the trail, including a 60 m section which separates and offers open
observation into both exhibits. The distance from visitors to the animals varies within exhibits as visitors
travel along the trail, but does not differ between the two exhibits.

2.2. Behavioural data collection

Data collection began during the week of November 6, 2005. Data were collected twice daily (one AM
and one PM observation), a minimum of 3 days per week to the week of January 1, 2006. Data were
collected on any day of the week except Sunday and Monday. This allowed for control of variations in
staffing and park operating hours. Morning observations were conducted at 09:30 h and 10:00 h for the
bachelor group and the family group, respectively. This allowed for each group to be in the exhibit for
approximately one hour prior to data collection. Afternoon observations were conducted at 15:00 h and
15:30 h for the bachelor and family groups, respectively. Afternoon observations were conducted approxi-
mately one hour prior to the animals coming into the holding building at night. All observations were
conducted from visitor areas along the Pangani Forest Trail.
Observation sessions were 30 min in duration. Behavioural data were collected using instantaneous
sampling on the group with a 2 min inter-sample interval. This methodology was chosen because the social
behaviours in these groups tend to be long-lasting and easily scored with a scan methodology and because
interval sampling allows for the mathematical control of time spent out of view. An all-occurrence
behavioural sampling technique, typically utilized to capture event behaviours, requires constant viewing of
animals in order to provide accurate counts across visibility conditions. However, interval sampling does not
require constant monitoring and allows control of poor visibility by conducting analyses on the proportion of
scans that an animal was visible. The ethogram of behaviours for this study (Table 1) was based on gorilla
research previously conducted at this institution (Stoinski et al., 2004a).

2.3. Crowd size

Crowd size was based on turnstile counts of visitors exiting the Pangani Forest Trail. These turnstiles
are an accurate indicator of the number of visitors that have passed the gorilla exhibits, as the gorilla
380 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385

Table 1
Ethogram listing behaviours and definitions used in an examination of the impact of large and small crowds on captive
gorilla behaviour
Behaviour Definition
Animal Not Visible (ANV) Entire animal cannot be seen by observer
Behaviour Not Visible (BNV) Animal is visible but is positioned such that the behaviour cannot be identified
Inactive (Inact) Animal is not actively performing any behaviours
Feed/Forage (Feed) Animal is engaged in search, preparation or ingestion of food
Autogroom (AGroom) Animals is picking, scratching, or otherwise manipulating some part of its
body with its hands, mouth, or feet
Abnormal (Abn) Animal is engaged in stereotypic behaviour (including pacing and rocking),
coprophagia, urophagia, hair plucking, or regurgitation and reingestion
Affiliative (Affil) Animal is engaged in social grooming, social play, socio-sexual behaviour,
non-aggressive contact or investigating another animal
Aggressive (Aggr) Animal is engaged in contact aggression, including biting and hitting,
or non-contact aggression, including stiff-legged stance, tight-lipped
face or chest-beat displays
Active Other (Oth) Animal is actively performing a behaviour other than those listed above

exhibit immediately precedes this exit and all visitors must pass in front of the gorilla exhibits to exit the
trail.

2.4. Data analysis

A total of 49 observation sessions were conducted on each group (25 AM sessions, 24 PM sessions). A
50th session was cancelled due to staffing issues. Turnstile counts on the Pangani Forest Trail from
observation days were ordered by magnitude and the 20 observations with the smallest turnstile counts were
categorized as Small Crowd (SC) observations, while the 20 largest turnstile counts were categorized as
Large Crowd (LC) observations. The remaining nine middle observations were discarded and not used in
any of the analyses. On average, crowds on Large Crowd days (9906  665) were twice as large as crowds
on Small Crowd days (4921  295).
To verify that SC and LC observations did not differ with environmental conditions, an independent
samples t-test was used to compare the temperatures recorded during LC days and SC days.
Behavioural data were summarized as the proportion of scans within each crowd size condition. Thus, a
single value for each behaviour was calculated for each animal in each crowd size condition (e.g. for each
behaviour, a single value for each animal in the Large Crowd condition and a single value for each animal in
the Small Crowd condition) in order to control for the lack of independence of multiple observations (Kuhar,
2006). Animal Not Visible was summarized as the proportion of total scans, while all other behaviours were
summarized as the proportion of visible scans (total scans minus Animal Not Visible scans) in each crowd
size condition. These values were used as the within-subjects factor in a repeated-measures ANOVA
analysis. Group (bachelor or family) was used as a between-subjects factor to prevent pseudoreplication
associated with pooling the two groups (Kuhar, 2006) and to determine whether the two groups reacted
differently to large crowds. All data analyses were conducted using SPSS1 version 13.0. An alpha value of
0.05 was used as the rejection criterion for all tests.
Prior to analysis the assumptions of repeated-measures analysis were tested. For all dependent variables,
Kolmogorov–Smirnov tests were used to verify univariate normality, while Box’s test was used to verify
equality of covariance matrices and Levene’s test was used to test for homogeneity of variance (Stevens,
2002). Based on these results Affiliative and Abnormal behaviour categories were log transformed prior to
analysis in order to meet the assumption of normality. Box’s test indicated a lack of equality of covariance
 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385 381

matrices for Abnormal behaviour, suggesting caution in interpreting those results. Sphericity is assumed in
the output because of single pairwise comparisons in the repeated-measures condition (Stevens, 2002).

3. Results

Data from the 10 days with largest crowds and 10 days with smallest crowds were used in the
analyses. Each day included one AM and one PM observation. Twenty 30 min observations in
each crowd size condition (40 total observations) were analysed, resulting in 10 h of data in each
condition (20 h of total observations). There was no difference in mean temperature between LC
days (18.3  4.1 8C) and SC days (20.1  3.2 8C) (t19 = 1.13, P = 0.27). With the exception of
2 days, there was no rain and the sky was sunny and clear. Both of the days when it rained, the rain
was light and it occurred on SC days.
A repeated-measures analysis revealed no group by crowd size interaction for all behaviour
categories (P > 0.05) with the exception of Aggression. There was an interaction between group
and crowd size for Aggressive behaviours (F 1,8 = 7.58, P = 0.03). In the LC condition, the
bachelor group engaged in more Aggressive behaviours, whereas the family group did not
(Fig. 1).
When group was examined using repeated-measures ANOVA, there were few behavioural
differences between the groups (P > 0.05) (Fig. 2). Animal Not Visible was scored more
frequently in the family group (F 1,8 = 6.09, P = 0.04) and individuals in the family group were
less likely to be engaged in Active Other (F 1,8 = 16.18, P < 0.01). The bachelor group was more
likely to engage in Abnormal behaviour (F 1,8 = 5.30, P = 0.05), but this significant result was
driven by a single animal who engaged in regurgitation and reingestion in approximately 4% of
scans.

Fig. 1. This figure represents the means and standard errors of the occurrence of aggressive behaviours in a bachelor and a
family group of gorillas when exposed to small crowds of visitors and large crowds of visitors.
382 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385

Fig. 2. Means and standard errors of occurrences of behaviours calculated across crowd size conditions for a family group
and a bachelor group of gorillas. *P < 0.05 (ANV, Animal Not Visible; BNV, Behaviour Not Visible; Inact, inactive;
ActOth, Active Other; Feed, feed/forage; Agroom, autogroom; Abn, Abnormal behaviour; Aggr, Aggressive behaviours;
Affil, Affiliative behaviours).

Fig. 3. Means and standard errors of occurrences of behaviours summarized across both gorilla groups during small
visitor crowd and large visitor crowd conditions. *P < 0.05 (ANV, Animal Not Visible; BNV, Behaviour Not Visible;
Inact, inactive; ActOth, Active Other; Feed, feed/forage; Agroom, autogroom; Abn, Abnormal behaviour; Aggr,
Aggressive behaviours; Affil, Affiliative behaviours).
 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385 383

While there were no impacts of crowd size for most behaviours (P > 0.05), when large crowds
were present, the Animal Not Visible category was scored more often (F 1,8 = 14.15, P = 0.01)
(Fig. 3).

4. Discussion

The study of the impact of zoo visitors on animal behaviour is often complicated by
methodological challenges. While this study controlled many of the confounds in visitor–animal
interaction research, it is subject to many of the constantly varying factors and potential
confounds of the real world. However, this study was able to capitalize on attendance fluctuations
over a short time period to examine the impact of large crowds on two demographically different
gorilla groups. Although behavioural patterns of the two groups were largely similar, some
differences did emerge. Whereas members of the family group were scored as Animal Not
Visible more often, members of the bachelor group engaged in Active Other and Abnormal
behaviours more often. Behaviours in the Active Other category consisted primarily of moving
from one area of the exhibit to another, whereas the Abnormal behaviours were usually
regurgitation and reingestion. Overall there were relatively few differences in behaviours
between the two crowd size conditions, but during the large crowd condition, gorillas were scored
as Animal Not Visible more often. Additionally, the bachelor group of gorillas, but not the family
group, was Aggressive more often during the Large Crowd condition.
That animals in both groups were scored as Animal Not Visible more often during periods of
high attendance is important. The provision of choice opportunities for animals has been
discussed as an important component of animal welfare in zoo exhibits (Snowdon, 1989) and the
implementation of visual barriers and environmental complexity has been discussed as one way
to provide these opportunities for primates (Coe et al., 2001; Stoinski et al., 2004b). The gorilla
exhibits at Disney’s Animal Kingdom1 are large, complex, and densely vegetated. Within these
exhibits there are multiple locations where animals can hide, both from the visitors and from each
other. The decrease in the amount of time animals were visible during the Large Crowd condition
indicates these gorillas may be using these visual barriers to avoid the public.
Given that animals were scored as Animal Not Visible in approximately 30% of observations,
it is possible that behaviours were performed but not recorded. However, it should be noted that
the areas in which animals could get out of public view were fairly small. These small spaces
allowed only enough room for a single animal to occupy each one at a time. Given the small
spaces, affiliative behaviours, such as play and grooming, are unlikely to have been missed
because only a single animal could be out of view in that area at a time. Likewise, it is unlikely
that any aggressive behaviour occurred in these areas without the observers being able to see
animals enter or exit these areas in an aggressive posture. Thus, while it was possible for solitary
behaviours to be occurring out of view of the data collector, significant numbers of social
behaviours were unlikely to have been missed. This is significant because social behaviours,
particularly aggressive ones, are often of the greatest interest to gorilla managers (Stoinski et al.,
2004b; Kuhar et al., 2006).
In another assessment of the impact of crowds on gorilla behaviour, Wells (2005) found
increased activity, aggression, abnormal behaviour, and autogrooming when crowd sizes were
large. The current study also found an increase in aggression, but not in activity or autogrooming,
although the amount of time that animals spent out of view in the current study should be noted. It
may be that providing extensive opportunities for gorillas to avoid large crowds mediates some of
the potential negative impacts that visitors may have on these animals. However, the differences
384 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385

in results between the studies may also result from methodological differences or individual
animal differences.
The idea that individual differences may drive results is an important one. When research is
conducted on a small sample of individuals, it is possible for the behaviour of a small number of
animals to impact the results. Instances of individual differences have been shown in other
research on visitor effects. Vrancken et al. (1990) found a single female eastern lowland gorilla to
sit near the glass when visitors were present, whereas the other four animals in the group were
unaffected by the presence of visitors. Sellinger and Ha (2005) found individual differences in the
relationship between visitor density and intensity of pacing and aggression levels in a pair of
jaguar. Although Hosey (2000) discussed the potential for individual characteristics to greatly
impact the results of visitor-effect studies, little research has examined these factors in detail.
The results presented here suggest crowd size may have influenced aggression differently in
the two groups. While there was a four-fold increase in aggression within the bachelor group
during the LC condition, the family group did not exhibit an increase in aggression. Because the
gorilla groups in this study are housed in similar enclosures and were exposed to the same crowd
conditions, the most obvious difference between the two groups is their demographic makeup.
However, the current design and analysis does not allow for the determination of which factors,
whether individual or environmental, caused these differences. While these data were examined
by conducting analyses at the group level, the potential for individual differences to influence
research results on the impact of crowd size implies that future research should carefully consider
the demographic makeup of groups of animals and employ statistical techniques which examine
variation in individual behaviour patterns. Until more research is conducted on individual
differences or differences between groups, we cannot be certain what factors resulted in the few
behavioural differences reported here.
In the future, research using instantaneous methodologies should be developed to help
determine the proximal cues to behaviour change caused by zoo visitors. Additionally, this
research may increase the understanding of why there are individual or group differences in
responses to crowds. The implications of this research could then positively impact animal
welfare if the findings are incorporated into the designs of future animal enclosures.

5. Conclusion

The data presented here showed few differences in the behaviour patterns of western lowland
gorillas when exposed to large and small crowds. However, these data emphasized the
importance of the provision of visual barriers and the opportunities for animals to avoid large
crowds if they choose. These data also showed the potential for group differences in responses to
crowd sizes. These findings further emphasize Hosey’s (2000) call for research into specific
characteristics of the visitor effect and suggest that individual differences should be examined in
future studies on the effects of zoo visitors.

Acknowledgements

I thank the following individuals for their assistance in data collection and data entry: Linda
Cory, Jon Wild, Oscar Roussett, Keelyn Walsh, Stephanie Joseph, and Kathleen Smith. I thank
Dr. Kyle Burks for his assistance in obtaining visitor count information. I acknowledge the
support of the Animal Husbandry and Education and Science teams at Disney’s Animal
Kingdom1 for their cooperation and support of this project.
 C.W. Kuhar / Applied Animal Behaviour Science 110 (2008) 377–385 385

References

Coe, J.C., Fulk, R., Brent, L., 2001. Chimpanzee facility design. In: Brent, L. (Ed.), The Care and Management of Captive
Chimpanzees. American Society of Primatologists, San Antonio, TX, pp. 39–81.
Hosey, G.R., 2000. Zoo animals and their human audiences: What is the visitor effect? Anim. Welfare 9, 343–357.
Hosey, G.R., 2005. How does the zoo environment affect the behaviour of captive primates? Appl. Anim. Behav. Sci. 90,
107–129.
Kuhar, C.W., 2006. In the deep end: pooling data and other statistical errors in zoo and aquarium research. Zoo Biol. 25,
339–352.
Kuhar, C.W., Stoinski, T.S., Lukas, K.E., Maple, T.L., 2006. Gorilla behaviour index revisited: age, housing, and
behaviour. Appl. Anim. Behav. Sci. 96, 315–326.
Martin, P., Bateson, P., 1993. Measuring Behaviour, second ed. Cambridge University Press, Cambridge, pp. 84–100.
Mitchell, G., Obradovich, S., Herring, F.H., Dowd, B., Tromborg, C., 1991. Threats to observers, keepers, visitors, and
others by zoo mangabeys (Cercocebus galeritus chrysogaster). Primates 32, 515–522.
Sellinger, R.L., Ha, J.C., 2005. The effects of visitor density and intensity on the behaviour of two captive jaguars
(Panthera onca). J. Appl. Anim. Welfare Sci. 8, 233–244.
Snowdon, C.T., 1989. The criteria for successful propagation of endangered primates. Zoo Biol. 8 (Suppl. 1), 149–161.
Stevens, J.P., 2002. Applied Multivariate Statistics for the Social Sciences, fourth ed. Lawrence Erlbaum Associates,
Mahway, NJ, pp. 268–269, 500–502.
Stoinski, T.S., Kuhar, C.W., Lukas, K.E., Maple, T.L., 2004a. Social dynamics of captive western lowland gorillas living
in all-male groups. Behaviour 141, 169–195.
Stoinski, T.S., Lukas, K.E., Kuhar, C.W., Maple, T.L., 2004b. Factors influencing the formation and maintenance of all-
male gorilla groups in captivity. Zoo Biol. 23, 189–203.
Vrancken, A., Van Elsacker, L., Verheyen, R.F., 1990. Preliminary study on the influence of the visiting public on the
spatial distribution in captive eastern lowland gorillas (Gorilla gorilla graueri Matchsie, 1914). Acta Zool. Pathol.
Ant. 81, 9–15.
Wells, D.L., 2005. A note on the influence of visitors on the behaviour and welfare of zoo-housed gorillas. Appl. Anim.
Behav. Sci. 93, 13–17.