Elizabeth W. Freeman,1,2* Greg Guanano,2 Deborah Olson,3 Mike Keele,4 and Janine L. Brown1

1. Department of Reproductive Sciences, Conservation & Research Center, National 1
Zoological Park, Smithsonian Institution, Front Royal, Virginia
2. Department of Environmental Science and Policy, George Mason University, Fairfax, 2
3. Department of Conservation and Science, Indianapolis Zoo, Indianapolis, Indiana 3
4. Oregon Zoo, Portland, Oregon

Nearly one-third of reproductive age African elephants in North America that are
hormonally monitored fail to exhibit estrous cycle activity, which exacerbates the
nonsustainability of the captive population. Three surveys were distributed to
facilities housing female African elephants to determine how social and
environmental variables contribute to cyclicity problems. Forty-six facilities
returned all three surveys providing information on 90% of the SSP population
and 106 elephants (64 cycling, 27 non cycling and 15 undetermined). Logistic
analyses found that some physiological and social history variables were related
to ovarian acyclicity. Females more likely to be acyclic had a larger body mass
index and had resided longer at a facility with the same herd mates. Results
suggest that controlling the weight of an elephant might be a first step to helping
mitigate estrous cycle problems. Data further show that transferring females
among facilities has no major impact on ovarian activity. Last, social status
appears to impact cyclicity status; at 19 of 21 facilities that housed both cycling
and non cycling elephants, the dominant female was acyclic. Further studies on
how social and environmental dynamics affect hormone levels in free-living,

Grant sponsors: Smithsonian Scholarly Studies Program and Women’s Committee; Robison Family
Foundation; Shared Earth Foundation; Friends of the National Zoo; International Elephant Foundation
ÃCorrespondence to: Elizabeth W. Freeman, Department of Reproductive Sciences, Conservation &
Research Center, National Zoological Park, Smithsonian Institution, Front Royal, VA 22630.
E-mail: [email protected]
Received 5 June 2006; Revised 14 August 2007; Accepted 13 February 2008
DOI 10.1002/zoo.20187
Published online 16 June 2008 in Wiley InterScience (www.interscience.wiley.com).
cycling elephants are needed to determine whether acyclicity is strictly a captivity-
related phenomenon. Zoo Biol 28:1–15, 2009.

Keywords: African elephant; Loxodonta africana; ovarian acyclicity


Although enhanced breeding efforts have led to increased pregnancy and birth
rates among captive African elephants (Loxodonta africana), the North American
population is still not self-sustaining

[Olson and Wiese, 2000]. With restrictions and
political concerns associated with importation, zoos must enhance breeding success
of elephants to maintain the captive population. Reproductive rates of captive
elephants are low owing in part to a lack of available bulls for breeding and
occasional beha vioral incompatibilities between the sexes. However, the main
concern is the high rates of ovarian cyclicity problems that eliminate over one-third
of females from the breeding pool [Brown, 2000; Brown et al., 2004b; Olson and
Wiese, 2000].
It is not known what causes acyclicity or whether it is exclusively a captivity-
mediated phenomenon. Noninvasive fecal hormone monitoring of free-ranging
elephants has recently provided evidence that prolonged periods of anestrous are
related to reduced availability of quality browse [Wittemyer et al., 2007a]. However,
an analogous effect of altered nutritional status is not likely to occur in most zoo
settings. In fact, it is highly unlikely that any single factor is responsible for the high
rates of ovarian acyclicity observed among captive individuals [Brown et al., 2004a],
but rather both physiological and psychological conditions probably are involved.
Husbandry or management practices are not the sole contributing factors to
acyclicity; otherwise, all females at a given facility would have the same ovarian
status. Only 4% of North American facilities exclusively house noncycling females,
whereas 52% house both cycling and noncycling cows [Brown et al., 2004b]. Thus,
captivity-related factors probably mediate acyclicity only indirectly [Brown et al.,
2004a]. We propose that ovarian activity is related to social rank, phy sical conditions
and/or climate [Schulte et al., 2000], because both social [Schulte, 2000] and
environmental factors differ between captive and free-living elephants.
Failure of ex situ populations to reproduce at levels comparable with the wild
can be caused by a lack of species-appropriate socio-environmental conditions[Lindburg and Fitch-Synder, 1994]. Thus, captive managers need detailed know-
ledge of the life history traits and social organization of each species to ensure
successful propagation in captivity [Wielebnowski, 1998]. Systematic, cross-
institutional evaluations of the captive environment that integrate behavior and
husbandry research can identify causes of poor breeding success and contribute to
the creation of self-sustaining ex situ populations [Kleiman, 1994; Lindburg and
Fitch-Synder, 1994; Mellen, 1994; Wielebnowski, 1998]. Data can be collected in
person; however, written surveys are commonly used because of logistical and cost
considerations associated with direct interviews [Carlstead et al., 1999b; Mellen,
1994; Wielebnowski, 1998, 1999]. Surveys rely on the familiarity of humans with the
animals in their care, which enables them to filter and integrate individual traits
based on a variety of situations over long periods of time [Line, 1987]. This intimate
knowledge of individual animals relative to their cohorts ca n provide insight into
variables that contribute to overall health and reproduction [Carlstead et al., 1999b;
Wielebnowski, 1999].
For this study, three surveys were distributed to North American facilities
housing female African elephants to determine what social traits and environmental
conditions, if any, a re shared by noncycling females. Additional life history
information was gathered from the North American African Elephant Studbook[Olson, 2003]. Data on environmental parameters, including longitude, latitude,
elevation, average annual precipitation and temperature of each facility, were
acquired from the National Oceanic and Atmospheric Administration (NOAA).
Survey, studbook and environmental data were analyzed to produce a mathematical
model of variables associated with ovarian acyclicity. The goal was to enhance our
understanding of the mechanisms that control estrous cycle activity so that
mitigating strategies can be developed to create a self-sustaining captive population
of African elephants.


Determination of Ovarian Status

At least a year of current progestagen data was used to categorize the ovarian
status of each elephant in the study. Serum progestagens were analyzed using a solid-
phase progesterone radioimmunoassay (Siemens Medical Solutions Diagnostics,
Inc., Los Angeles, CA) [Brown et al., 2004a]. Females were considered cyclic if they
exhibited normal estrous cycle ranges with an overall length of 14–16 weeks, a luteal
phase length of 8–12 weeks and a follicular phase length of 4–6 weeks [Plotka et al.,
1988; Hodges, 1998; Brown, 2000]. Some elephants exhibited irregular cycles with
periodic short luteal phases (o8 weeks in duration) or prolonged nonluteal phases
(46 weeks in duration). However, because these elephants demonstrated luteal
activity, they were categorized as ‘‘cycling.’’ Elephants that maintained continuous
baseline progestagen concentrations (o0.1 ng/mL), i.e. no luteal activity, were
categorized as ‘‘noncycling’’ [Brown et al., 2004b].

Surveys and Data Collection

Three surveys were distributed to facilities with adult, female African elephants
listed in the African Elephant Studbook [Olson, 2003]. Keepers with at least 6
months experience working directly with the elephants were asked to complete a
temperament survey on each female to determine whether noncycling females across
facilities share common temperament traits. Respondents were asked to complete
questions that ranked each elephant on a social scale (Table 2). The elephant
manager or SSP liaison at each facility was also asked to complete health and socio-
environmental surveys [Freeman, 2005]. The health survey was designed to
determine how the general health of the elephant (i.e. body condition and medical
history) related to ovarian status. The socio-environmental survey was designed
to investigate the relationship between captivity-related variables and ovarian
activity. It was composed of questions related to (1) facility design; (2) elephant
management; (3) changes within the herd (births, deaths and transfers) in the last 5
years; (4) number and experience of keeper staff; and (5) makeup of the elephant diet[Freeman, 2005].
For each elephant, age, number of herdmates, transfers between facilities and
the length of time spent at the current facility were obtained from the Studbook[Olson, 2003]. Location and climate data, including longitude, latitude and elevation
were downloaded for each facility from the NOAA National Climatic Data Centre
(NCDC) website http://www.ncdc.noaa.gov/oa/ncdc.html. The 30-year (1971–2001)
annual means for temperature and precipitation for each location were acquired
from the NCDC [National Oceanic and Atmospheric Administration, 2001]. We
used the 30-year means for temperature and precipitation because they re ect the
trends for each region, where most elephants have spent their adult lifetime, rather
than focusing on a specific year(s).

Data Analyses

A model was designed to determine the relationship between the survey
variables and the ovarian activity. Model variables (Tables 1 and 2) consisted of
factors that were significant in preliminary analyses [Freeman, 2005] as well as those
that are of particular interest to captive elephant managers (Elephant Taxon
Advisory Committee/Species Survival Plan). To determine changes in the herd
structure over the 5 years before our analyses, we created an index of herd changes
(‘‘changes’’) by adding the births and transfers in, and subtracting the deaths and
transfers out (Table 2). The number of respondents completing the temperament
surveys varied among facilities (range, 1–10). The inter-rater reliability on all
temperament survey questions was assessed for each elephant that had two or more
raters using Kendall’s coefficient of concordance (W), which tests for the degree of
association among multiple raters [Lehner, 1996]. Survey questions in which
concordance coefficients failed to reach statistical significance at a level of P<0.05 for inter-rater reliability on a given elephant were excluded from further analyses. The mean survey score among raters for each elephant was used in subsequent data analyses. Descriptive statistics were run on all the variables. Variables that were highly skewed or displayed extreme kurtosis were analyzed for outliers using stem and leaf plots; outliers were removed from further analyses. Principal components analysis (PCA) [Martin and Bateson, 1993] was used to reduce related variables (such as temperament and social history) to a few underlying factor scores that could be used for subsequent analyses in our model. Components with eigenvalues41 were retained for interpretation and labeled according to the variables that showed the highest loadings [Stevenson-Hinde et al., 1980; Gold and Maple, 1994; Wielebnowski, 1999]. The factor scores generated by the PCA were used in subsequent analyses. Spearman rank order correlation was used to determine the relationships among the model variables. To test for multicollinea rity we regressed each variable in turn against all other independent variables. Variables with an r2>0.80 and a variance in ation factor (VIF)>4.00 have a high degree of
collinearity and were excluded from the model. Once a list of independent variables
that were free of collinearity problems was identified, logistic regression was used to
determine the link between the captivity-related factors and ovarian acyclicity.
Logistic regression allows for the prediction of the dichotomous dependent variable
(i.e. noncycling or cycling) from one or more independent variables [Demaris, 1992].

TABLE 1. Definitions of questions and response options on questionnaires that were used to
assess the temperament of ex situ female African elephants in North American zoos

TABLE 2. Additional questionnaire, climate, and studbook data acquired to predict the ovarian
status of ex situ captive African elephants in North American facilities

Statistical analyses were computed using Sigma Stat (2004, v. 3.01.0, Sy stat
Software, Inc., San Jose, CA) and SigmaPlot (2001, v. 7.1, Systat Software, Inc., San
Jose, CA). Kolmogorov–Smirnov test was used to test for normality of the
underlying population and the Levene Median test was used to measure for equal
variances. All values were reported as mean±SEM and statistical significance was
assumed at a P<0.05.


Fifty-four facilities completed at least one of the three surveys. More
temperament than health or socio-environmental surveys were completed on each
female (Table 3). Three hundred and seventy-six temperament surveys were received
for 123 elephants, with an average of 3.1±0.2 surveys per female. Health surveys
were completed on 107 elephants at 45 facilities, whereas 49 socio-environmental
surveys were returned describing the captive environment of 113 females (Table 3).
All three surveys were returned from 46 facilities describing 106 females (64 cycling,

TABLE 3. Surveys received on North American female African elephants
Surveys received

TABLE 4. Major components of individual behavioral variation (temperament variables) and
studbook data (social history) of 123 captive African elephants at 54 North American facilities
obtained through principal components analysis

27 noncycling and 15 undetermined). Because ovarian status was not known for the
undetermined elephants, they were excluded from subsequent data analyses.
Tables 1 and 2 present the independent variables extracted from the surveys
and the studbook that were used in our analyses. PCA of the temperament survey
variables (Table 1) resulted in two components with eigenvalues41 and accounted
for 65.37% of variance in the data (Table 4). Temperament component I showed
high loadings for interactions with herdmates and was named ‘‘herd.’’ Temperament
component II showed high loadings for interactions with people and was named
‘‘human.’’ To determine the social history of each elephant at their current facility
and with their current herdmates, PCA was conducted on the variable moves,
together and years (Table 4), which resulted in one component with an eigenvalue41
and accounted for 63.08% of variance in the data.
Many of the model variables on the social dynamics, population demographics
and life history of elephants were significantly correlated (P<0.05; Table 5). For
example, age had a positive correlation to history and social status. History had a
negative correlation to the number of males and changes to the herd structure.
Females had a positive correlation to males and changes, but a negative correlation

TABLE 6. Multiple logistic regression coefficients (1SE) for survey and studbook variables
that were related to ovarian acyclicity in captive African elephants

with management. Similarly, males had a negative correlation with management, but
a positive correlation with changes. Lastly, herd interactions had a positive
correlation with social status.
Many of the location and climate variables were significantly related to each
other (P<=0.05), as well as to the demographic and social variables (Table 5). Latitude
was negatively correlated to precipitation, females and males, and positively correlated
to elevation and management. Longitude had a negative correlation with precipitation
and a positive correlation with elevation. Elevation was negatively correlated to
temperature and precipitation. Lastly, temperature had a positive correlation with
precipitation and females, but a negative correlation with history and management.
Given the number of correlated variables within our model, we ran multiple
linear regressions of each variable against all of the other independent variables to
test for multicollinearity. Regression of the variable males had an r2=0.82. Males
also had a VIF44.00 when regressed as an independent variable against the
dependent variables of age, body mass index (BMI), herd, people, social rank,
latitude, longitude, elevation, temperature, precipitation, management and history
(everything except females and changes). Regression of precipitation had an
r2=0.82. Similar to males, precipitation had a VIF44.0 when it was regressed as
an independent variable against the dependent variables of age, BMI, females, herd,
people, social rank, latitude, elevation, temperature, precipitation, management,
changes and history (everything except longitude). Linear regression of all of the
other model variables had r2<0.80. We removed males and precipitation from the
model and reran the multiple linear regressions. All of the remaining variables had
r2<0.80 and VIF<4.0.
To determine what contributed the most to the occurrence of ovarian acyclicity
a multiple logistic regression was run on the remaining model variables (Table 6).
The model was significant (w2=5.75, d.f.=13, P<0.05) and had a relatively
strong fit (r2=0.47). History and BMI were the only significant variables in the
multiple logistic model (P<0.05) and they were both positively related to a female
being acyclic.


Surveys of the captive African elephant population identified several variables
that were related to the occurrence of ovarian acyclicity. Results demonstrate that
physical size and social history of captive African elephants have a relationship with
ovarian acyclicity. Our model suggests that females with a higher BMI and those
that have moved less frequently between facilities and have longer relationships with
cohorts are more likely to be acyclic.
BMI was the only health variable related to acyclicity, suggesting that some
noncycling elephants may be overweight. Obesity has been related to reproductive
dysfunction in humans [Pasquali et al., 2003], horses [Vick et al., 2006], bats [Chanda
et al., 2003], sheep [Christman et al., 2000] and rats [Marin-Bevins and Olster, 1999].
In humans, obese women are more likely to have irregular cycles or chronic
an ovulation and have increased risks of miscarriage, pre-term delivery and maternal
death [Pasquali et al., 2003]. The exact mechanism by which excess weight negatively
affects reproduction is not well understood [Grodstein et al., 1994; Pasquali et al.,
2003] and likely is species-specific. Obesity acquired through overeating can have a
direct effect on reproduction, or obesity may be related to a more complex metabolic
disorder that results in energy being stored in adipose tissue rather than being
available for other bodily functions, such as reproduction [Wade et al., 1996]. One
reason to suspect that obesity may not be because of simple overeating is that weight
loss has varied effects on reinitiating normal estrous cycle activity. In women, losing
excess weight can improve fertility and facilitate conception [Bray, 1997; Clark et al.,
1995; Norman et al., 2004; Pasquali et al., 2003]. However, in other species, feed
restriction does not alter estrous cycle activity [horses, Vick et al., 2006; Zucker rats,
Marin-Bevins and Olster, 1999]. These latter studies suggest that excess body weight
may not cause reproductive dysfunction per se, but rather may only be a
contributing factor [Wade et al., 1996; Marin-Bev ins and Olster, 1999; Vick et al.,
2006]. Because of species differences in the relationship between weight and
reproduction, interpretation of the finding of higher BMI associated with ovarian
acyclicity in elephants should be done cautiously. Regardless of whether acyclicity in
elephants is because of a direct or indirect effect of increased BMI, however, it makes
sense to control the weight of zoo elephants. It also will be necessary to determine
whether captive elephants become overweight because of an improperly balanced
diet [Hatt and Clauss, 2006], overeating or metabolic disorders that impact energy
storage of adipose tissue [Wade et al., 1996].
The significant, positive relationship between social history and ovarian
acyclicity suggests that females maintained for long-term at a facility with the same
herdmates are more likely to be acyclic. Sociality is an important factor for wild
female African elephants that live in strongly bonded family groups of related cows.
Each individual holds a social status based on her age, size and individual disposition[Dublin, 1983; Thouless, 1996; Archie et al., 2006] and the largest, eldest female in the
group is the matriarch and she is crucial to family survival [Douglas-Hamilton, 1972;
Poole and Moss, 1989; Archie et al., 2006]. Bull elephants leave their family group in
their teens (12–15 yr of age) and then live either alone or in small bachelor groups[Poole, 1987]. Because of the fission–fusion society, social groups are dynamic and
change daily and seasonally, even though the family unit remains stable [Poole and
Moss, 1989; Wittemyer et al., 2005]. Matriarchs are crucial to herd survival because
of their knowledge of natural resources and coordination of herd defense [Douglas-
Hamilton, 1972; Dublin, 1983; McComb et al., 2001; Poole and Moss, 1989].
In some aspects, captive elephant social groups resemble those in the wild.
Females are traditionally housed together and most males are remov ed from the
herd at a relatively young age [Schulte, 2000]. Many behaviors of captive elephants
also resemble those of their wild counterparts [Adams and Berg, 1980; Garai, 1992;
Schulte, 2000]. There are some major differences, however. Captive herds are
traditionally composed of unrelated females often of similar age; thus, there is no
age-ordered hierarchy of related individuals. Additionally, most captive individuals
have limited contact with new fema les, calves or males [Schulte, 2000], and most
females are nulliparous. Thus, the majority of elephants within captive herds have
not had the opportunity to acquire social knowledge based on the range of
behavioral interactions normally found in the wild, which may explain why creating
long-term social bonds by maintaining females together within small, captive groups
throughout their lifetime seems to increase the tendency for some females to become
acyclic. Although this finding suggests that keeping captive elephants together for a
long time may be detrimental to fecundity, we believe that social stability is vital to
captive elephant welfare. The challenge then is to design a management stra tegy that
will maximize breeding success among reproductive-aged cows without compromis-
ing herd stability.
Analyses of previous temperament surveys demonstrated that dominant
females tend to be acyclic [Freeman et al., 2004]. In this study, at all but two of
the 21 facilities that housed both cycling and noncycling elephants, the dominant
female was acyclic. Without long-term studies on captive elephants exposed to
different socio-env ironmental changes, it may be impossible to establish a cause and
effect relationship among these trends. Still, these data are the first to link social
history with a specific reproductive problem in elephants.
In ma ny mammalian species, dominant fema les often use social interactions to
suppress reproduction in subordinates [e.g. Packard et al., 1985; Faulkes et al., 1990;
Creel et al., 1992]. Whether reproductive suppression is important in free-ranging
African elephants is not clear. It has been suggested that dominant females use
aggressive behaviors to suppress subordinates when resources are limited [Sikes,
1971; Dublin, 1983], but this has yet to be systematically tested. Physiological and
demographic analyses show that reproductive rates of African elephants decline with
age [Laws et al., 1970; Smuts, 1975; Moss, 2001; Freeman et al., 2008] and as they
reach matriarchal status (B. Archie, Amboseli, data unpublished). Density and
behavioral mechanisms can cause older females to reach reproductive senescence at
an earlier age in the wild [Laws, 1969; Laws et al., 1970]. It is not known whether the
decline in fertility in older wild elephants is because of social pressures or
reproductive senescence. If it is socially mediated, such an a daptation may explain
why so many older, dominant females are not cycling in captivity. Necropsies of an
elephant culled in Kruger Nationa l Park, South Africa, revealed that many females
over 50 years of age have inactive ovaries; i.e. no follicle, corpus luteum or corpus
albicantia [Freeman et al., 2008]. Although it seems likely that these older females
were no longer cycling, systematic studies of endocrine activity have not been
monitored in free-ranging matriarchs. If some older females do stop cycling once
they reach matriarchal status, it could be functionally similar to what occurs in the
North America popula tion, albeit at a younger age.
An examination of environmental conditions found no significant effect of
facility location or climate on ovarian cyclicity status. In the wild, seasonal and
annual uctuations in rainfall and the availability of quality browse affect African
elephant social behaviors, coordinate herd movements [Wittemyer, 2001; Wittemyer
et al., 2005] and in uence birth rates [Laws et al., 1970; Foley et al., 2001; Moss,
2001; Wittemyer, 2001; Wittemyer et al., 2007b]. Additionally, decreased seasonal
precipitation causes a decrease in the availability of browse and causes reductions in
progestagen concentrations in pregnant cows [Foley et al., 2001; Wittemyer et al.,
2007a]. It is not expected that seasonal uctuations in precipitation similarly affect
captive elephants because they are provided regular allocations of food and water
year round [Schulte, 2000].
In recent years, questions have been raised about whether elephants should
be kept in zoos in northern latitudes because of harsh climates. To date, there is
little scientific evidence to support this belief. In fact, wild African elephants are
known to be highly adaptable to a wide range of temperatures, altitudes, latitudes
and terrains, and can withstand extreme conditions such as drought or heat
waves for considerable lengths of time [Sikes, 1971]. One study of African
elephants at a zoo in Rhode Island found a higher incidence of temporary ovarian
inactivity during some winter months [Schulte et al., 2000]. The authors speculated
that it was because of increased time spent indoors and exposure to indoor
contaminants and/or pheromones. However, results of our multi-institutional study
found no relationship between ovarian acyclicity and facility latitude or mean annual
temperature, both of which are correlated with time spent indoors in the winter[Freeman, 2005].
Multivariate, multi-institutional studies of other exotic species have
demonstrated that social and management factors contribute to breeding success[Carlstead et al., 1999a,b; Mellen, 1994; Wielebnowski, 1999; Wielebnowski et al.,
2002]. In addition, knowledge acquired from these types of studies has been used to
change management practices [Wielebnowski et al., 2002]. One encouraging finding
from this study is that acyclicity does not appear to be a permanent condition in
captive African elephants [Hermes et al., 2004; Schulte et al., 2000]. There now are
several examples of noncycling females that resumed cycling, either after a transfer
or alteration in the herd dynamics (Brown, unpublished). Unfortunately, there also
are cases of cycling females that shut down reproductive cyclicity after similar
Results of this study suggest that there are a few changes captive managers
can make to return acyclic females to the breeding pool. Reducing the weight
of young, acyclic cows may be an important first step to reinitiating normal
ovarian cycle activity. Although there has been some reluctance to moving
individuals in the past because it might disrupt the captive unit, survey data suggest
that transferring elephants should not affect long-term cyclicity status.
One possibility might be to create herds that contain an older, nonreproductive-
aged female that acts as the noncycling matriarch, allowing younger females
to cycle normally. That could potentially maximize reproductive potential within
the herd without compromising social stability. For those facilities reluctant to
move animals, it might be efficacious to use hormone treatments to stimulate
normal ovarian activity, although none have been proven to be effective as yet[Brown et al., 2004a].


1. Multiple logistic analyses suggest that social history is related to ovarian
acyclicity in captive African elephants. Acyclic elephants are more likely to have
long-term social relationships with their captive herdmates and facilities. These
results suggest that transferring elephants between facilities should not negatively
affect cyclicity status.
2. The majority of acyclic females hold a dominant social status within the herd.
3. BMI was the only health-related factor associated with ovarian acyclicity.
Controlling weight of acyclic elephants may help reinitiate normal cyclicity.


We thank the 54 facilities that completed one or more of the elephant surveys.
This project would not have been possible without their willingness to participate. We
acknowledge the following individuals for providing advice about the design of this
study and/or reviews of the early versions of the surveys: Kyle Burks, Debra Flynn,
Marie Galloway, Kirsten Leong, Jill Mellen, Alessia Ortolani, Heidi Riddle, Carol
Rizkalla, Sean Royals and Bruce Upchurch. Funding for this study was generously
provided by the Smithsonian Scholarly Studies Program and Women’s Committee,
the Robison Family Foundation, Philip D. Reed, Jr., the Shared Earth Foundation,
Friends of the National Zoo and the International Elephant Foundation.