African savanna elephants (Loxodonta africana) and Asian elephants (Elephas maximus) are a focus of welfare research in zoos due to their high intelligence, complex social structure, and sheer size. Due to these challenges, some argue that zoos are inherently incapable of providing appropriate care for elephants, while others believe that zoos can fulfill the needs of these species with improved husbandry. There is a general consensus from both within and outside of zoos, however, that zoos must improve their elephant programs or cease exhibiting these animals altogether. Now more than ever, applied research on zoo elephant welfare is needed to provide context for this debate. Researchers are interested in how far zoo elephants walk due to the potential health and welfare benefits of walking in these highly mobile species. Zoo researchers recently adopted GPS technology to study elephant walking, and preliminary evidence suggests that African elephants in large zoo exhibits walk distances that correspond with wild elephants under non-extreme conditions. However, data are limited from Asian elephants and from elephants in more typically-sized exhibits. In Chapter Two, I discuss important methodological considerations of utilizing GPS in a zoo environment, including an introduction to the technology, sources of error and mitigation, methods to improve GPS performance, and possible effects of GPS device attachment on animal behavior. This review shows GPS performance is adequate for tracking zoo elephant walking when proper methodological techniques are applied, and should serve as a useful reference for zoo researchers considering using GPS. In Chapter Three, I used GPS anklets to measure outdoor daily walking distance in 56 adult female African (n = 33) and Asian (n = 23) elephants housed in 30 zoos. I collected 259 days of data and found that elephants walked an average of 5.34 km/day with no significant difference between species. Multivariate regression models predicted that elephants with more dynamic feeding regimens (more diverse feeding types and frequencies; unscheduled feeding times) will walk more. Distance walked was also predicted to be higher in elephants that spend time in a greater number of different social groups. Distance walked was predicted to decline with age. Finally, I found a significant negative correlation between distance walked and nighttime space experience. The results of the analysis suggest that zoos that want to increase walking in their elephants need not rely solely on larger exhibits, but can increase walking by adding quality and complexity to exhibits. However, my results failed to establish a definitive link between walking distance and other validated measures of elephant welfare. Thus, the direct health and welfare benefits of walking in zoo elephants remain unresolved. Resting behaviors are an essential component of animal welfare, but have received little attention in zoological research. In Chapter Four, I used accelerometers in anklets to complete the first large-scale multi-species investigation of zoo elephant recumbence. I collected 344 days of data from 72 adult female African (n = 44) and Asian (n = 28) elephants at 40 zoos. I found that African elephants are recumbent an average of 2.14 hours/day, which is significantly less than Asian elephants at 3.22 hours/day. Multivariate regression models predicted that African elephant recumbence increases when they experience more space at night, and Asian elephant recumbence increases when they spend time housed alone. Both species showed a similar response to substrate, such that African elephants spending time on all-hard substrates are predicted to be recumbent less, while Asian elephants spending time on all-soft substrates are predicted to be recumbent more. The discovery that occasional non-recumbence is a common behavior in zoo elephants also introduces a new area of research that may have important animal welfare consequences. Finally, this study established that zoos should continue their efforts to replace hard substrate with soft substrate in order to provide zoo elephants with environments that facilitate recumbence. Overall, this work assessed walking and recumbence in zoo elephants, which will allow zoos to gauge the prevalence of these behaviors in their elephants as compared to the sub-population studied here. A variety of factors that are associated with these behaviors were also identified. With this information, zoos can prioritize modifications to their facilities and animal management programs to create an environment that encourages zoo elephants to express walking and recumbence behavior, should they choose to do so. This work is one component of the Elephant Welfare Project, the largest zoo animal welfare project ever undertaken, and is unprecedented in both scope and scale. The project was funded by the Institute of Museum and Library Services (IMLS), an independent, U.S., federal, grant-making agency that supports libraries, museums, and zoos. At the time of this writing, the first manuscripts from this project are being submitted to academic journals. These papers will describe the prevalence and distribution of a variety of elephant behaviors and welfare indicators, serve as a benchmark for future elephant welfare studies, and aid in decision making with regard to best practices in elephant management.

Increasing human populations have resulted in the extensive conversion of natural forests and range lands into agricultural lands, resulting in an expansion of the interface between people and elephants across the elephant range countries of Asia and Africa. This interface describes the nature of two-way interactions between people and elephants, which can be positive and reverential or hostile and negative. Elephant crop-raiding, one of the most negative interactions for people at the interface, is not only the result of decreased food resources and space, but has also been attributed to a preference for cultivated crops and to damage caused during elephant movements between habitats. The aim of this thesis was an attempt to understand the use of coffee agroforestry areas by elephant populations in a South Indian district, Kodagu, and to assess the risks to elephants and people of coffee plantations. Geographically, located at a significant position in the Western Ghats, Kodagu district is a part of one of the largest wild Asian elephant ranges harbouring India’s largest elephant population. Kodagu has a unique topography and coffee agroforestry system in considered as the boon for conservation. This thesis is the first long term (one year) study on the elephant populations using coffee estates in Kodagu. Crop-raiding events across Kodagu and their intensity of occurrence were determined from the Forest Department compensation records. Virjapet taluk was one of the three administrative units of Kodagu with frequent incidences of crop-raiding, including elephant mortality and human deaths. High rates of crop-damage in Virajpet included both coffee and paddy rice produce as the land is conducive for the cultivation of both. To understand the use of coffee estates by elephants, coffee estates in Virjapet were directly and indirectly monitored for the presence of elephants using dung sampling (N=202), camera trapping, video and photo documentation, as well as sightings (N=408) and reports by local workers, in order to identify the individuals or groups of elephants frequenting these coffee estates. Lone male and all male groups used coffee estates most frequently and family herds ranging in group size from 2 to 10 were present mainly during the peak season of coffee ripening (post monsoon). Presence of large numbers of elephants, especially with large female groups, was associated with crop-damage during the months of December-January. As seasonal movements of elephants in Kodagu districts are still not known, it is unclear why the number of elephants in coffee estates post-monsoon increases when food availability should also be higher in forests. These large coffee estates were used as refuge areas by elephants during the day by all individuals and groups, and feeding on estates occurred during the night to early morning hours. Dung analysis and observations suggested that coffee estates were attractive for elephants due to the constant availability of water (for irrigation), green fodder, and cultivated fruit trees, especially jackfruit. Coffee plants were damaged both due to consumption (47% of dung samples in this study) and accidental damage during elephant movements within the estates. Although the dung sampling could not confirm whether coffee had become a novel food resource, the presence of large number of elephants during the coffee ripening season suggested that the potential for coffee berries to be added regularly to the diet in the future, with potential consequences for coffee invasion of native forests through dung seed dispersal. People working on large coffee estates were accustomed to the presence of elephants and were generally knowledgeable of the areas that elephants frequented, thus avoiding fatal encounters. However, safety of farmers and other people working on the estates remains a major concern, especially for large coffee estates owners. The constant interaction between elephants and people also led to more negative perceptions of elephants, and reduced the tolerance of elephants in the area. The unique topography of Kodagu as a mosaic of forests and farms challenges the number of possible mitigation methods to prevent negative encounters between people and elephant. The elephants of Kodagu may have adapted behaviourally to the presence of people, but long-term monitoring of the elephant population is important to understand their ecological and social adaptations to the various costs and benefits of using this agroforestry landscape. Suggestions for management of the elephant-human interface and mitigation of negative attitudes and actions were made, through a model that incorporates a multiple stakeholder (including elephants) action plan.

Elephants' complex societies, well-developed communication systems, evolutionary history and close working relationship with humans make them an important species for studies of cognition but research on elephant cognition is sparse. In this thesis I aim to illuminate the social cognition involved in the interpretation and production of visual signals by African elephants (Loxodonta africana). My results are intended to contribute to the cross-species literature on social cognition and help to elucidate wild elephant social behaviour. I studied captive elephants, housed at an elephant-back safari company in Victoria Falls, and wild elephants in Hwange National Park, both in Zimbabwe. Wild elephants display a vast array of postures, actions and signals. I found that elephants recognise visual attentiveness in others when they signal silently, producing more signals when their audience can see them, and using the body and face orientation of an audience to judge their attention. When responding to typically human visual signals, elephants immediately responded correctly to deictic gestures, including variants of pointing that they were unlikely to have already experienced. These results indicate elephants' astonishing sensitivity to even subtle social cues. I found no indication that elephants reason about mental states such as false beliefs, or rationality; however, limitations of the experimental design meant I was unlikely to find such an ability even if it is present in elephants. Furthermore, I discovered that elephants have a form of referential indication in their natural communication in the wild. Elephants match their direction of attention with a type of trunk action produced by a group member. Attending to human-like signals, and interpreting them as communicative is an advantage for any animal working with humans and that ability might explain the choice of species that are ancestors of today's domestic animals.

Darwin was troubled by elaborate male traits observed in many species that are seemingly maladaptive for survival, the peacock’s tail being the most iconic of all. He wrote "The sight of a feather in a peacock’s tail, whenever I gaze at it, makes me sick" because it challenged his theory of evolution by natural selection for adaptive traits. The extreme length of the tail may render a peacock more vulnerable to predation and therefore maladaptive for survival. To account for the evolution of apparently maladaptive traits he proposed the theory of sexual selection, wherein, traits that directly enhance mating success may be selected for, either as weapons in male-male competition for mates or as ornaments preferred by females. Male and female elephants in the proboscidean evolutionary radiation have had tusks and show extreme exaggeration in size and form. However, tusk in the Asian elephant (Elephas maximus) is sexually dimorphic as it is expressed only in the males, hinting at a possibility that opposing selection (sexual selection advantage to males and natural selection disadvantage to females) may have been the processes behind this pattern of tusk expression. Intriguingly, tuskless males (male dimorphism with respect to tusk) also occur at fairly high frequencies in some Asian elephant populations (∼50% in norteastern India and ∼95% in Sri Lanka). Theory states that dimorphic males can also occur in a population in stable frequencies as a consequence of sexual selection. I explored sexual selection on elephant tusks as possible mechanism leading to the observed patterns of tusk dimorphism in the elephants. All elephant populations on earth have been harvested for ivory, therefore, artificial selection (selective poaching of tusked elephants for ivory) is another possible cause of tusk dimorphism. I developed mathematical models of population genetics, population dynamics and conducted field observations of mating behavior of Asian elephant in Kaziranga National Park, Assam to understand the evolution of tusk dimorphism in elephants. Darwin’s sexual selection theory was controversial when proposed in 1871 and continues to remain so in 2014. In the introduction of my thesis I have discussed Darwin’s two classical mechanisms of sexual selection, namely, male-male combats for mates and female mate choice based on male traits. The latter was viewed with considerable skepticism by his con-temporary Alfred Russell Wallace and more recently deemed "fundamentally flawed" by Joan Roughgarden. Therefore, I have also discussed the arguments against female mate choice for male traits found in literature. I have reviewed current knowledge about sexual selection for sexually dimorphic male traits of body size and musth, in the African and Asian elephant and state why I have hypothesized that tusks may also be under sexual selection. Sexually selected traits are expected to be genetically determined, therefore, I explored mathematically (Chapter 1) the genetic basis of evolution of sexual dimorphism. Fisher proposed that sexually selected male display traits originate in both the sexes but are suppressed in the females by modifier genes, when the trait becomes deleterious to females. Thus, sexually antagonistic selection on a trait and sex-specific gene expression can lead to the evolution of sexual dimorphism. Tusk is sexually monomorphic in the probocideans that are ancestral to both the African (Loxodonta africana) and Asian elephant (Elephas maximus). Tusk continues to remain monomorphic in the African elephant but has become sexually dimorphic in the Asian elephant. Tusk, therefore could be a sexually selected male trait that evolved according to the Fisherian model. Intriguingly, tuskless males occur at very high frequencies in some Asian elephant populations. The tusked and tuskless male morphs could be alternate male mating strategies, occurring at evolutionarily stable frequencies. Alternatively, the observed male tusk dimorphism, could be a consequence of artificial selection against tusked individuals, due to selective harvest of tusked males. Furthermore, male African elephants are more intensely poached for ivory than female elephants. Yet the frequency of tuskless individuals has increased more rapidly among females than in males. In essence, sexual dimorphism could be evolving among such poached populations. Is such rapid, contemporary evolution of sexual dimorphism, possible through the Fisherian modifier gene mechanism? A 2-loci genetic model (with X-linked trait gene and an autosomal modifier gene) (Rice 1984), a slight variant of the model (with X-linked modifier gene, and an autosomal trait gene) and an entirely autosomal model, were analyzed for the rate of evolution of sexual dimorphism, under different selection pressures for tusk possession. Negative frequency dependent selection was introduced into the model of tusk evolution in accordance with Gadgil’s model for the evolution of male dimorphism as consequence of sexual selection (Gadgil 1972). In two of the 2-loci models (in which tusk gene in autosomal), tusklessness evolved much more rapidly in females than in males, under equal negative selection pressures. The models predict several combinations of time-lines and negative selection pressures for effecting a particular change in the frequency of tusklessness. Model predictions were com-pared with observed changes in the frequency of female tusklessness, in one South Ugandan, African elephant population (∼2% to 10% in 5 to 9 generations) and male tusklessness (∼5% to 50% in 25 to 40 generations) in one north eastern Indian, Asian elephant population. The models predict strong selection pressures of 30% to 50% reduction in fitness, that can effect an 8% increase in tusklessness, in the African elephant population, within time-lines of 9 to 5 generations (∼225 to 125 years) respectively. For the male Asian elephants, natural selection against tusked males on an already sexually dimorphic population, must have been in operation and shifted the population to 5% male tusklessness. The models predict that artificial selection with 20 to 30% fitness cost to tusked males, operating for 40 to 25 generations (∼1000 to 600 years) respectively, can further shift the population from ∼5% to ∼50% tusklessness. Asian elephant populations may already have been in a transient phase of evolution, tending towards tusklessness, with recent artificial selection hastening the process. The two major pre-dictions from this modeling exercise are (1) artificial selection could have played a significant role in the evolution of male tusk dimorphism in the Asian elephant (2) a lack of or very mild current sexual selection on tusks in the male Asian elephant. Both these predictions may be empirically verified. Chapters 2 and 3 are attempts at empirical verification of prediction (1) and Chapters 4 and 5 of prediction (2). From historical references to elephant harvest in Assam, we do know that artificial selection has been in operation, but whether it has played a major role in causing male tusk dimorphism needs to be established. It may be possible to detect signatures of significant past harvest from current demographic structure of an elephant population. Sustained biased harvesting of a particular sex and or age class from an animal population alters the sex ratio and age structure (relative proportion of individuals in each age and sex class) of a population considerably (Sukumar 1989). It may be possible to back infer the harvest scenario by studying the deviation of current age and sex ratios from natural age and sex ratios. In Chapter 2, I explored models of population dynamics under different harvest regimes and its effect on age and sex ratios. I described a method to infer unknown harvest rates and numbers from age and sex ratios, namely, adult female to male ratio, male old-adult to young-adult ratio, and proportion of adult males in the population using Jensen’s(2000) 2-sex, density-dependent Leslie matrix model. The specific combination of male and female harvest rates and numbers can be deter-mined from the history of harvest and an estimate of population size. I validated this model with published data on age and sex ratios of one Asian and African elephant population with fairly reliable data on elephant harvest as well. In Chapter 3, I applied this model to the demographic data that I collected from a wild Asian elephant population in Kaziranga National Park, Assam, India (where more than 50% of the adult males are tuskless). Male polymorphism of sexually-selected male traits occur at stable frequencies in populations of several species. The different male morphs of the trait are hypothesized to be alternate male mating strategies with equal life time reproductive fitness. Male Asian elephants of Kaziranga National Park, Assam are dimorphic with respect to tusk possession: ∼50% of the males are tuskless (and are locally called makhnas). Makhnas could be trading tusk for either longevity, larger body size, testicular volume and or duration of musth as alternate mating strategies. On the other hand makhnas may have increased to a very high frequency primarily due to selective removal (captures for domestication and hunting for ivory) of tusked males from the population for centuries. The aim of Chapter 3 was to examine the role of artificial selection in the evolution of makhnas. Prolonged male-biased harvest(removal from the population) is bound to alter the demographic structure of the population and leave a signature of the intensity and type of harvest on the residual population structure. The Kaziranga elephant population was considered as representative of elephant populations of north east India; A harvest modeling approach (described and validated in Chapter 2) was used to infer unknown harvest of elephants from demographic parameters estimated by sampling this elephant population during 458 field days in the dry season months of 2008–2011. The Kaziranga elephant population appears to have been harvested approximately for the past 700 to 1000 years with adult tusked males being harvested at approximately twice the rate of adult tuskless males, adult females and their immature offspring of both the sexes. The currently observed high frequency of tuskless males in Kaziranga therefore, may be a consequence of sustained artificial selection against tusked males for several centuries. The previous two Chapters have only examined some mechanisms for the loss of tusks in elephants. I proceeded to examine the possibility of evolution of tusks through Darwin’s mechanisms of male-male competition for mates and female mate choice. Elephant tusks are cited as an example of a male trait that has evolved as a weapon in male-male combats. In Chapter 4 I examined the role of tusks in establishing dominance along with two other known male–male signals, namely, body size and musth (a temporary physiologically heightened sexual state) in an Asian elephant population in northeastern India with equal proportions of tusked and tuskless males. I observed 116 agonistic interactions with clear dominance outcomes between adult (>15 years) males during 458 field days in the dry season months of 2008–2011. A generalized linear mixed-effects model was used to predict the probability of winning as a function of body size, tusk possession and musth status relative to the opponent. A hierarchy of the three male–male signals emerged from this analysis, with musth overriding body size and body size overriding tusk possession. In this elephant population tusk possession thus played a relatively minor role in male–male competition. An important implication of musth and body size being stronger determinants of dominance than tusk possession is that it could facilitate rapid evolution of tuskless males in the population under artificial selection against tusked individuals, which are poached for ivory. If not a weapon, tusks could be a male ornament that female elephants find attractive. I explored the interplay of the three male traits (body size, musth and tusk), male mating strategies and female mate choice in Chapter 5. In some species males obtain mating opportunities by harassment of females. Given the striking size difference between an adult male and female elephant, with males weighing at least 30% more than females, male coercion of females to mate is a possibility. A detailed study of the courtship behavior revealed that overt male harassment of females is rare and the ability of a male to mount and stay mounted on a female for copulation is under female control. Therefore female Asian elephants can exercise choice to mate but this is subtly different from exercising mate choice itself. Age-related male mating strategy (reported for the first time in the Asian elephant) exists in the Kaziranga elephant population and this strategy limits the ability of females to exercise choice. Young males (<25 years) predominantly show a sneak mating strategy. Middle-aged males (25–40 years), when in musth, mate–guarded oestrous females from sneakers and attempted mating but sometimes resorted to sneak mating when out of musth. Old males (> 40 years) attempted mating only during their musth phase and were seldom sneakers. Large/musth males received positive responses from estrous females towards courtship attempts significantly more often than did small/non–musth males. Tusked non–musth males attempted courtship significantly more often than did their tuskless peers, and had a higher probability of receiving positive responses than did tuskless males. A positive response, however, may not translate into mating because of mate–guarding by the dominant male. Females permitted large/musth males to stay mounted significantly longer than small/non-musth males. Musth and large body size may be signals of male fertility. Female mate choice in elephants thus seems primarily for traits that signal direct benefits of assurance of conception. Tusked males may attain sexual maturity faster than tuskless males. Therefore it is worth exploring if tusks function as signals of male fertility when males are young (15 to 25 years); this may be possible through hormonal and behavioral profiling of young tusked and tuskless males from 10 to 20 years of age. Overall all musth and body size appear to play a larger role in enhancing male mating success than tusks. Tusked males appear to have a weak sexual selection advantage (male-male domi-nance and female preference) over their tuskless peers, only in the young age class (15 to 25 years) in this population. Males in this age age class, seldom come into musth that would over-ride tusk as a signal of male dominance. Current sexual selection on tusks in this population, appeared to be insignificant and this may be verified through genetic analysis of paternity success. An important implication of musth and body size being stronger determinants of mating success than tusk possession is that, it could facilitate rapid evolution of tuskless males in the population under artificial selection against tusked individuals, even in a slow breeder such as the elephant. Musth may have evolved much later than tusks in elephants, therefore it is possible that tusks evolved under sexual selection before musth evolved. However, body size, in mammals in gen-eral appear to be under both natural and sexual selection. Gould has shown that the absurdly large and palmate antlers of the extinct Irish elk, scales allometrically with body size (Gould & Lewontin 1979). Phylogenetic studies of elephant evolutionary radiation indicate a general trends towards increase in body-size with size reduction and tendency towards dwarfism occurring only in island habitats (Palombo 2001). Tusk development, which is essentially tooth development may be closely linked to cranium development. Cranium development in turn may be linked to body size through allometric scaling laws. If so, any selection on body size is bound to act on tusk size. I propose that the evolution of elaborate tusks seen in elephants is primarily due to natural and or sexual selection acting on body size, and tusk just hitched a ride with body size. Tusks may be maintained in spite of tuskless males occurring in the population only because of a rather weak sexual selection advantage to tusk possession in contests in which males are symmetrical with respect to body size and musth status.

Forest elephant (Loxodonta africana cyclotis) sociality is relatively little-studied due to the difficulties of making direct observations in rainforests. In Central Africa elephants aggregate at large natural forest clearings known as bais, which have been postulated to offer social benefits in addition to nutritional resources. This thesis explores the role of these clearings as social arenas by examining bai use within three main themes; ecology, sociality and risk factors. Seasonal changes in elephant use of the Maya Nord bai (Republic of Congo) are described, along with the demography of the visiting population. Elephant visit rate was highly variable; the number of elephants using Maya Nord in an observation day ranged from 0 to 117 animals. This variability was unrelated to local resource availability and productivity suggesting that bai use occurs year round. Elephants in Odzala-Kokoua do not show high fidelity to a single clearing; 454 elephants were individually identified and re-sighted an average of 1.76 times (range 1-10) during the twelve month study period. Previous bai studies have yet to quantify how elephants associate with one another within the bai area. This study examines socio-spatial organisation and associate choice using two measures of association within the 0.23 km2 bai area; aggregations (all elephants present in the clearing) and parties (elephants spatially co-ordinated in activity and movement) and distinguishes these from parties that range together (i.e. arrive and leave together). Social network analyses (SocProg) were used to describe inter- and intra-sexual multi-level organisation in the bai environment, and to illustrate the non-random nature of elephant aggregations and parties. Bais were shown to function as social arenas; female elephants showed active choice of certain associates and active avoidance of others when creating parties, whereas males were less discriminatory. Parties formed in the clearing (mean size= 3.93, SE= 0.186) were larger than ranging parties (mean size= 2.71, SE= 0.084) and elephants stayed for 50% longer in the clearing when they associated with individuals from outside their ranging party. Inter- and intra-sexual relationships were maintained within the clearing, and these are suggested to offer elephants essential opportunities for social learning. The patterning and nature of the relationships observed at the Maya Nord clearing indicates that forest elephants use a fission-fusion social structure similar to that of savannah elephants (Loxodonta africana africana); relationships are significantly structured by age- and sex- and underpinned by individual identity. Old experienced females hold key roles for forest elephants, and male relationships are superimposed on the network of female associations. Odzala-Kokoua elephants use bais to maintain their social relationships despite being highly sensitive to the anthropogenic risks involved in using these open areas. The results of this study suggest that forest and savannah elephants lie on the same social continuum, balancing social “pulls” to aggregate against the ecological “pushes” that force groups to fission. Previous models of savannah elephant sociality construct levels of association and social complexity upwards from the basic mother-calf unit (e.g. Wittemyer & Getz 2007). My results suggest that it may be more appropriate to consider elephant sociality and associations as in dynamic equilibrium between social and ecological influences acting at all levels of grouping, and to explicitly test how these underlie the opportunity costs that elephants are willing to pay in order to maintain social groupings.

This study presents an investigation into the population dynamics and social structure of a small, closed elephant population. Specifically, it examined population growth rates for evidence of density-dependent regulation. It also quantified the association patterns of female elephants groups, and male elephants groups. Social structure was examined using Hamilton’s kinship theories of inclusive fitness, and age. Male-female patterns of association were also examined for inbreeding avoidance behaviours. The study population was located in Addo Elephant National Park, South Africa. Density-dependence was assessed using a long-term data set. Densities were considerably higher than estimated carrying capacities. Population growth rate was positively correlated with increasing density. No relationship between birth rate, the age of first calving or calf sex ratio and elephant density was detected but there was a positive relationship between birth rate and rainfall during conception year. Mortality rates, particularly for juveniles, were low, and mean inter-calf interval was 3.3 years. There is no evidence of density dependent regulation in this population. These findings indicate that density dependence should not be considered as an option in the control of elephant numbers in this Park, or where elephant resources are not seasonally limited. Examination of association patterns of the adult female component revealed that associations were not random at the population, family or individual scale. This is the second study on African elephants to confirm previous behavioural studies that predicted that preferred associates were close maternal relatives. This supports many studies showing that social species preferentially associate with their kin. The adult males in this population were found to have a well differentiated society with non-random associations. Generally, males were found to have weak associations with most other males and strong associations with only a few males. This association pattern was found to be persistent over the time frame of the study, as indicated by the time lag analysis. Males returned to their natal family, even when maternally related females were in oestrus. Oestrous females directed positive behaviours towards musth males. It appears that behavioural inbreeding avoidance mechanisms in this small, closed population are inhibited: musth status seems to override inbreeding avoidance. General principles from this case study were interpreted in terms of their applicability to other small, closed populations.