Evaluating social housing strategies in zoo-housed aye-ayes (Daubentonia madagascariensis): behavioural considerations for welfare and management

Introduction

Primate social systems are often associated with the benefits they provide to individuals or group members within specific ecological contexts (Box, 2012). Understanding variation in these systems requires distinguishing between their core components - social organisation, social structure, and mating system - which are shaped by different selection pressures and do not necessarily co-vary (Kappeler and van Schaik, 2002; Kappeler, 2019). These systems are also flexible, shifting in response to ecological pressures, resource distribution, or environmental change, demonstrating behavioural flexibility.

Despite extensive research on primates, effort remains unevenly distributed across taxa, with many underrepresented species belonging to predominantly nocturnal groups (Chen et al., 2023). However, there is recognition that nocturnal primates may exhibit greater social complexity than previously assumed (Bearder, 1999; Geerah et al., 2019), including the aye-aye (Goodman, 2022), where complexity encompasses not only group size but also the nature and patterning of social relationships (Kappeler, 2019). Advances in field technologies, including radio telemetry and acoustic monitoring, alongside the expansion of captive breeding programmes, have improved the study cryptic and nocturnal species (Hending, 2016; Göbel and Wilms, 2025). For example, acoustic monitoring has revealed ultrasonic vocalisations in the Javan slow loris (Nycticebus javanicus) (Geerah et al., 2019), highlighting social communication systems that were previously overlooked.

Aye-ayes are the sole extant lemur of the Family Daubentoniidae (Sterling and McCreless, 2007). It has traditionally been classified as a solitary species, where as defined by Kappeler and van Schaik (2002), solitary social organisation refers to the absence of synchronised activity and foraging with conspecifics, rather than the absence of social relationships altogether. Ancrenaz et al. (1994) reported that aye-ayes nest solitarily, and this field observation has been used to classify the aye-aye’s primary social organisation as solitary in comparative analyses of primate social evolution, though occasional male-female social units have also been identified in the same population (Ancrenaz et al., 1994; Olivier et al., 2024). Classifications of social systems based on partial accounts of the diel cycle may underestimate the full complexity of a species’ social organisation. For example, grey mouse lemurs (Microcebus murinus), classified as solitary foragers, exhibit daytime sleeping associations that vary considerably by sex and season (Engelhardt et al., 2025), illustrating how social variation can be obscured when observations are restricted to the active period alone. More broadly, Makuya and Schradin (2024) further highlight that the costs and benefits of solitary living in mammals exist on a continuum, and that distinguishing between truly solitary species and solitary foragers that maintain social bonds, such as during inactivity, requires observations across the full diel cycle. For aye-ayes, resting period behaviour remains poorly documented. Individuals were not observed sharing nest sites on Nosy Mangabe, yet Sterling (1993) noted that adult males had been observed sharing nests on mainland Madagascar, suggesting further potential for sociality.

Behavioural field studies indicate that aye-ayes express considerable variable in social interactions (Sterling, 1993). While affiliative body contact is largely absent and individuals often forage alone, Sterling (1993) documented repeated foraging and travel associations between an adult male and female on eight occasions across 11 months outside of breeding periods, alongside male-male and male-female foraging units of up to three individuals, though inter-individual encounters were described as infrequent overall across 496 observation hours. Interaction patterns in males ranged from avoidance or absence of interaction (Sterling, 1993; Randimbiharinirina et al., 2018), chasing, dominance displays, and foraging in tandem in repeated non-breeding associations (Sterling and Richard, 1995), illustrating that aye-aye social structure defined as the patterning and quality of interactions amongst individuals (Kappeler and van Schaik, 2002), may be more variable than a strictly solitary social classification implies. These observations raise important questions regarding the functions of such associations, including whether they reflect underlying social motivations or are driven by ecological factors such as resource distribution. However, the absence of more precise quantitative data on the frequency of social interactions in wild aye-ayes limits interpretation and reflects the broader scarcity of detailed behavioural data for this species and highlights the need for further systematic study of their social behaviour.

Male-female social interactions include nose touching, mutual licking, allogrooming, and social play (Andriamasimanana, 1994); notably, a male-female pair on a semi-wild island site spent an average of 26.4% of observation time in proximity to one another, with association periods ranging from 5 minutes to over an hour (Andriamasimanana, 1994). Overlapping home ranges between males and females (Randimbiharinirina et al., 2018; Sterling, 1993), and amongst males (Sterling, 1993; Ancrenaz et al., 1994), may therefore naturally create opportunities for interactions to take place. Similarly, in captivity, allogrooming post-copulation has occurred, though allogrooming expression varied between dyads - occurring between an unrelated male and female but not between a male and his dam (Winn, 1994a), tentatively suggesting affiliative behavioural differences between related and unrelated individuals. As with Sterling (1993), neither study reports precise proportions of social versus solitary units, reflecting the lack of quantitative on relative frequency of social versus solitary behaviour for this species across both wild and captive contexts. Despite this, aye-ayes are typically managed solitarily in captivity outside of weaning and breeding contexts, and systematic evaluations of social housing are limited.

Juvenile aye-ayes experience prolonged weaning, extending to approximately 17 months of age (Rakotondrazandry et al., 2021a), during which food sharing and close maternal associations are common. Juveniles use food transfer to obtain hard-to-process food, such as coconuts or seeds (Ravelomandrato et al., 2026), via maternal provisioning, opportunistic theft, or leftover remains from adult feeding bouts, potentially including those of other adults (Andriamasimanana, 1994). This occurs until juveniles have developed specialised foraging skills and enlarged jaw muscles to access their full dietary range independently. Sexual maturity is typically reached at around 2.5 years (Winn, 1994a), although earlier maturation as early as 21 months old has been reported in captivity (Species360, 2026). Kinship strongly influences social interactions across vertebrates, including primates (Belisle and Chapais, 2001; Ren et al., 2018). However, beyond early developmental stages, the role of kinship in structuring adult social relationships and wider social organisation in aye-ayes remains unclear and warrants further investigation.

Notably, captive aye-ayes have been reported to exhibit abnormal behaviours, including stereotypies (Audra, 2016; Fuller et al., 2016; Göbel and Wilms, 2025) and self-directed behaviour such as excessive scratching and subsequent hair loss (Toyne, 2022). The underlying causes remain unclear, however, auto-grooming behaviour has been associated with stress in other species (Kruk et al., 1998). While such behaviours may function as coping mechanisms, they may also be deemed abnormal in frequency, duration, or intensity, and could impose welfare risks if associated with injury or chronic stress. Zoo environments may introduce stressors, such as spatial restriction, limited environmental complexity, or artificial social structures (Price and Stoinski, 2007), which have been linked to stereotypic behaviour in some species (Vaz et al., 2017). Although large-scale environmental modifications may be limited by practical and institutional constraints, social housing represents a potential strategy for increasing behavioural opportunities in captivity. This may be particularly relevant given that wild aye-ayes occupy substantially larger home ranges than those practically available in captivity (Andriamasimanana, 1994; see Methods). Moreover, the predominantly solitary management of aye-ayes in captivity contrasts with the variable social structure observed in some wild populations and may be associated with the expression of abnormal behaviours in this species.

Evidence from other primates with dispersed social systems suggests social housing can facilitate affiliative behaviours, including food sharing in orangutans (Pongo abelii) (Kopp and Liebal, 2016), and allogrooming and sleep-site sharing in pygmy lorises (Xanthonycticebus spp.) (Alejandro et al., 2021). Although social housing has been proposed for aye-ayes (Price and Cowen, 2016), empirical evidence remains scarce, with existing reports largely anecdotal or mixed (Carroll and Haring, 1994; Toyne, 2022).

Aims, hypotheses, and predictions

Given the documented variability in wild aye-aye social interactions and reports of abnormal behaviour in captive aye-ayes, this study aimed to assess whether managed social housing affects social and non-social behaviours in captive aye-ayes. Due to the welfare implications of stereotypic and self-directed behaviour, monitoring abnormal behaviour as well as maintenance of essential behaviours such as feeding and foraging was considered a key welfare outcome of this study.

Hypothesis 1: social behaviour and kin-structured group composition

Prediction 1a: Social behaviours will be significantly greater than zero levels observed in solitary housing, confirming increased expression under social contexts.

Expected magnitude: Social behaviour < 20 % of scan samples, consistent with the species’ classification as a largely solitary forager but evidencing latent social flexibility.

Prediction 1b: The proportion of social behaviour will increase with group size - highest in full social housing (three individuals), intermediate in dam-offspring dyads, and lowest in sire-offspring dyads, reflecting greater interaction opportunities and greater maternal tolerance.

Hypothesis 2: non-social behavioural time budgets

Prediction 2a: Relative to solitary housing, social and dyadic housing will show modest reductions in locomotion, exploration, and maintenance as activity is reallocated towards social interaction.

Prediction 2b: Feeding and foraging levels will remain stable, and abnormal behaviour will not increase, indicating no adverse short-term welfare effect.

Hypothesis 3: effects of social housing duration

Prediction 3a: During three weeks of consecutive group housing, locomotion, exploration, and maintenance will stabilise or decline as individuals habituate.

Prediction 3b: Feeding and foraging will remain constant, and abnormal behaviour will not increase, reflecting neutral or positive welfare outcomes.

Additional expectation: Behavioural trajectories may differ between individuals due to variation in age, sex, and social tolerance.

Age class and sex were included as fixed effects to control for potential variation in behaviour. Although the small sample (one adult male, one adult female, one juvenile male) prevents formal age x sex interaction testing, the juvenile male was anticipated to display higher social play and lower maintenance behaviours than adults, consistent with developmental patterns. In line with Prediction 1a, affiliative social interactions between the adult and juvenile males were expected to occur less frequently than interactions involving the adult female, reflecting expected differences between paternal and maternal social relationships in aye-ayes. Because age and sex are partially confounded, these effects are interpreted cautiously as indicative rather than generalisable.

Materials and methods

Study site and subjects

The study was conducted at Bristol Zoo (UK) within the former nocturnal zone Twilight World. Three captive-born aye-ayes were observed: an adult male, an adult female, and their juvenile male offspring (Table 1). All individuals were familiar with one another due to prior short-term social housing periods of up to ~40 days outside of the female’s oestrus. At the onset of the study, the juvenile was 16 months old, no longer nursing, and had previously experienced temporary separation from the dam, including a ~50 day period housed with the adult male. During the study, animal care staff confirmed that the adult female was pregnant (Weeks 1–6 of study).

Table 1. Study subjection information.

The enclosure complex comprised two adjacent indoor areas (combined dimensions: 12.3 × 4.6 x 6 m) (Figure 1), connected via internal doors to allow flexible social management. Both forest-styled enclosures were of equal sizes and contained horizontal and vertical climbing structures and two elevated nest boxes. A reversed light cycle simulated nocturnal conditions (dim “moonlight” lighting during visitor hours; bright lighting during resting periods).

Housing conditions and experimental design

Three housing conditions were tested: (1) ‘Solitary’ (adult only; juvenile not housed alone due to age-related welfare considerations), (2) ‘Social’ (all three individuals housed together), and (3) pair housing, comprising three dyadic combinations: ‘Social with offspring’, ‘Social with dam’, and ‘Social with sire’ (adult-juvenile combinations, as detailed in Table 2). Housing conditions were applied for one-week blocks. During weeks 7–9, continuous social housing was maintained to assess short-term effects of housing duration. A precautionary split condition was implemented in Week 4, with daytime male-juvenile pairing (4B), and overnight reunion of dam-juvenile pairing (4A). During solitary housing, individuals were separated into adjacent enclosures, preventing tactile contact but allowing auditory and olfactory communication. During social group housing, both enclosures were accessible, effectively doubling the available space relative to solitary or pair housing; this spatial difference is acknowledged as a potential confound. All individuals received a standardised daily diet consisting of a gruel-based mixture of egg and pelleted feed in the morning, four whole vegetables and mixed nuts at midday, and live insect feed in the afternoon. Whilst the specific vegetables offered varied on a daily basis, the feeding schedule was consistent at the weekly level, with each vegetable item appearing in the diet a minimum of three times per week. This feeding regime remained consistent across all housing conditions.

Table 2. Housing conditions applied throughout the study period.

Behavioural data collection

Individual behaviour was recorded in each housing condition and used as the measured variable to compare between the housing conditions. Behavioural observations were conducted five days per week over a nine-week period (November 2021 to May 2022). Observation sessions lasted 60- minutes and were conducted during visitor hours (10:00–16.30). Sessions were scheduled at varying times within this window to reduce temporal bias, rather than at a fixed time each day. The specific start time of each session was recorded but the time of day was not included as a variable in the analysis, which represents a limitation acknowledged in the discussion.

Instantaneous scan sampling was conducted at one-minute intervals, yielding up to 60 scans per session. At each scan point, the behaviour of each visible individual was recorded separately, serving as the focal subject in turn, and allowing individual-level behavioural data across all housing conditions. This approach enabled differentiation of individual behaviour whilst maintaining the efficiency of scan sampling across a small, identifiable group (Altmann, 1974). Focal sampling was used when collecting data when the adult female or adult male were under solitary conditions. State behaviours were recorded at each scan point, whilst event behaviours (e.g. approach, avoid) were recorded if occurring at the scan point. Behaviours were classified according to a predefined ethogram (Table 3) and grouped into the following categories: social, locomotion, maintenance, feeding/foraging, exploration abnormal, and unrecorded. The abnormal behaviour category included stereotypic and self-directed behaviours (e.g. pacing, hair plucking, excessive scratching) identified by animal care staff as occurring at elevated frequency or intensity.

Table 3. Ethogram of aye-aye behaviours (Modified from Roe, 2018).

In total, 49 observation hours were collected: 24 hours of social group housing, 25 hours of pair housing (15 hours of adult male-juvenile male housing; 10 hours of adult female-juvenile male housing) and 25 hours of solitary housing (10 hours adult male; 15 hours adult female). A single observation session (adult female, week 9) was excluded prior to analysis due to insufficient scan data (one scan point, representing <2% of the expected observations, precluding reliable proportion estimation). All other sessions were retained.

Data processing and statistical analysis

Behavioural data were summarised as proportions of total scan samples per individual per observation session for behavioural categories. Feeding and foraging behaviours were combined and analysed as one behavioural category due to their functional overlap. Unrecorded behaviours, recorded to reflect observation effort, were excluded from statistical analyses as they represented periods when individuals were not visible and did not yield behavioural data. Exploratory visualisations were generated in R version 4.4.3 (R Core Team, 2024) using ggplot2 to examine behavioural patterns across individuals and weeks prior to model fitting.

Although some dyadic housing conditions involved the same pair of individuals, categories were analysed from the perspective of the focal individual. For example, “social with offspring” reflects the adult focal individual housed with the juvenile, whilst “social with dam” and “social with sire” reflect the juvenile focal individual when housed with the mother or father, respectively. These categories represent distinct focal behavioural contexts rather than separate housing conditions, with the unit of analysis being the focal behavioural response.

To test Hypothesis 1, social behaviour was analysed using a generalised linear mixed model (GLMM) fitted in the glmmTMB package. Behavioural data were modelled as binomial response proportions using a two-column response matrix (cbind(social behaviour scans, total scans − social behaviour scans)), representing the number of scan samples in which social behaviour was observed relative to all scans per observation session. Housing condition (five levels: solitary [reference], social, social with dam, social with offspring, social with sire), age class, and sex were included as fixed effects.

To test Hypothesis 2, separate GLMMs were fitted in glmmTMB for each non-social behavioural category (locomotion, maintenance, exploration, feeding and foraging, and abnormal behaviour). Behavioural data were modelled as binomial response proportions using a two-column response matrix (cbind(successes, total scans − successes)), representing the number of scan samples in which a behaviour was observed relative to all scans per observation session to account for the bounded proportional scan data and the fixed sampling effort inherent in scan sampling protocols. Housing condition (five levels: solitary [reference], social, social with dam, social with offspring, and social with sire), age class, and sex were included as fixed effects.

To test Hypothesis 3, a subset of data from the three consecutive weeks of group housing (weeks 7–9) were analysed using GLMMs fitted in glmmTMB. Behavioural data were modelled as binomial proportions, expressed as the number of scans in which a behaviour was observed relative to the total number of scans per session using a two-column response matrix (cbind(successes, total scans − successes)). This approach directly accounted for the bounded nature of the data and the fixed upper limit imposed by scan sampling. Separate models were fitted for each behavioural category (locomotion, maintenance, exploration, feeding and foraging, and abnormal behaviour). Each model included week (categorical: 1-3), age class, and sex as fixed effects.

For all models, individual identity was included as a random intercept, with an additional observation-level random effect (one level per row) included to account for potential residual overdispersion. All models were fitted using a binomial error distribution with a logit link function. Model significance was assessed using likelihood ratio tests (LRTs) comparing full models to intercept-only null models, with model fit and dispersion diagnostic checks performed in the DHARMa package. For Hypothesis 1, to obtain all pairwise comparisons amongst housing conditions without inflating Type II error or reducing statistical power the same model was re-run with each housing condition set as the reference level in turn.

Results

H₁: Descriptive overview of social behaviour

Across social housing conditions, all individuals engaged in measurable social behaviours accounting for 16.25% of scans in the juvenile male, 14.02% in the adult female, and 12.37% in the adult male. Overall, proportions of social behaviours varied across conditions and generally increased with the number of social partners (Figure 2).

Affiliative behaviours including allogrooming, social play and social foraging were the most frequently observed in all individuals and across all social housing contexts relative to agonistic interactions (<0.35% total scans) (Figure 3). The adult male expressed similar proportions of most social behaviours across social and dyadic housing conditions, with the greatest variation observed in social mounting, scent marking, and allogrooming. Social play was only observed during social housing and, across all individuals, was most frequently expressed under this condition. Comparatively to both the adult and juvenile males, the adult female expressed the greatest variation in proportions of social behaviours across social and dyadic housing conditions, shifting between higher proportions of social play in group housing to higher proportions of social foraging in dyadic housing. This pattern is also visible in the juvenile male, who otherwise expressed similar proportions for most other social behaviours. Within the low proportions of agonistic interactions across all individuals and conditions, fighting was highest in dam-offspring dyadic housing.

H₁: Effects of housing condition on social behaviour

Housing condition significantly influenced the proportion of scans in which social behaviour was observed (Table 4). Relative to solitary housing, all social housing conditions were associated with significantly higher levels of social behaviour (all p < 0.001), with the strongest effect observed in the standard social group condition. Sex also significantly influenced social behaviour, with males exhibiting higher levels of social behaviour than females. In contrast, age did not significantly affect social behaviour, with no difference detected between juveniles and adults. A likelihood ratio test confirmed that inclusion of fixed effects significantly improved model fit compared to a null model (χ² = 82.15, df = 6, p < 0.001), indicating that housing condition and sex explained meaningful variation in social behaviour.

Table 4. Binomial GLMM results for social behaviour (Reference level = solitary; logit link; binomial error structure with observation-level random effect).

To further explore differences amongst social housing configurations, pairwise comparisons were derived by re-running the same model with each condition as the reference level in turn (Table 5). These results revealed that all social housing conditions differed significantly from solitary housing, confirming a robust increase in social behaviour under group living conditions. However, no significant differences were detected amongst social housing configurations themselves (all p > 0.06). This suggests that whilst social housing consistently promoted social behaviour relative to solitary housing, the specific composition of the social group did not significantly influence the level of social behaviour observed. These findings, taken together alongside the overview of descriptive social behaviour, indicate that H₁a is supported through significant social behaviour expression (social housing > solitary), whilst H₁b is only partially supported (the group condition was not statistically higher than dyads), and they are consistent with the prediction that social behaviour would remain below 20% of total activity budgets.

Table 5. Pairwise comparisons of housing conditions (binomial GLMM, logit link; no correction applied).

H₂: Effects of housing condition on non-social behavioural time budgets

Behavioural time budgets differed across housing conditions, although effects were behaviour-specific rather than uniform across the activity budget (Table 6). LRTs comparing full models to reduced models retaining age and sex indicated that inclusion of housing condition significantly improved model fit for maintenance (χ² = 28.70, df = 4, p < 0.001), exploration (χ² = 34.06, df = 4, p < 0.001), and abnormal behaviour (χ² = 25.99, df = 4, p < 0.001), but not for locomotion (χ² = 7.16, df = 4, p = 0.13) or feeding and foraging (χ² = 5.00, df = 4, p = 0.29).

Table 6. GLMM results for the effects of housing condition, age, and sex on behavioural time budgets (binomial GLMMs, logit link).

Behavioural time budgets showed clear, behaviour-specific responses to housing condition relative to solitary baseline, rather than a uniform shift across the activity budget (Table 6). Locomotion was significantly higher in social housing compared to solitary housing, whilst no differences were detected between solitary and dyadic social configurations. Maintenance behaviour showed a significant reduction in social housing relative to solitary, indicating reduced maintenance time in social contexts, whereas other social configurations did not differ significantly from solitary. Exploration was significantly elevated in the social-with-sire condition compared to solitary housing, suggesting context-dependent increases in exploratory behaviour depending on social composition.

Feeding and foraging behaviour did not differ significantly between solitary and any social housing condition, indicating that feeding time allocation remained comparable to solitary baseline levels across housing contexts. Abnormal behaviour showed the most pronounced effects of housing condition, with significant increases observed in both social-with-offspring and solitary contexts relative to standard social housing, indicating heightened abnormal behaviour under these conditions. No significant differences were observed for other social configurations (Figures 4A–E). Age and sex also contributed significantly to behavioural variation; however, given the small sample size and limited representation of each category across individuals, these effects are best interpreted as reflecting individual-level differences rather than independent population-level effects.

Model diagnostics indicated acceptable fit across all models. There was no evidence of substantial overdispersion for locomotion (p = 0.464), exploration (p = 0.552), feeding and foraging (p = 0.288), or abnormal behaviour (p = 0.512), whilst maintenance showed no significant deviation from expected dispersion under the binomial model (p = 0.784). Residual uniformity tests indicated adequate model fit across all behavioural categories (all p > 0.05), suggesting that the binomial GLMMs appropriately captured variation in behavioural proportions across housing conditions.

Overall, these results indicate that H₂a is partially supported (locomotion decreased and maintenance increased under solitary as predicted, with a context-specific rise in exploration rather than a general reduction) and H₂b is largely supported (feeding and foraging remained stable and abnormal behaviour did not increase under social housing relative to solitary).

H₃: Effects of social housing duration on non-social behavioural time budgets

Behavioural composition changed across the first three consecutive weeks of group housing (weeks 7–9 of the study), with effects varying across behavioural categories (Table 7). LRTs comparing full models to intercept-only null models indicated that inclusion of fixed effects significantly improved model fit for exploration (χ² = 12.08, df = 4, p = 0.017), feeding and foraging (χ² = 9.89, df = 4, p = 0.042), locomotion (χ² = 9.64, df = 4, p = 0.047) and abnormal behaviour (χ² = 24.18, df = 4, p < 0.001), but not for maintenance (χ² = 1.98, df = 4, p = 0.740).

Table 7. GLMM results for behavioural change across group housing duration.

Exploration increased significantly in week 2 relative to week 1 (Table 7), whereas no significant difference was detected between week 3 and week 1, although individual-level variation in trajectories was evident (Figure 5A). Feeding and foraging showed a significant increase in week 3 relative to week 1, with no effect observed in week 2 despite variation amongst individuals (Figure 5B). Abnormal behaviour increased significantly across the study period, with higher levels in both week 2 and week 3 compared to week 1. In contrast, locomotion and maintenance did not vary significantly across the three-week period, although a non-significant increasing trend was observed in the juvenile (Figure 5A).

Age and sex were significant predictors across several behaviour categories, but like results for hypothesis 2, these effects are interpreted cautiously due to the non-independence of age, sex, and individual identity within the small sample. Model diagnostics showed acceptable fit for all models. Specifically, there was no evidence of overdispersion based on DHARMa tests and no issues with residual uniformity (all uniformity p- values > 0.05), indicating that the models adequately captured the variability in the behavioural data.

Collectively, the duration analyses show H₃a is partially supported (locomotion and maintenance stabilised, but exploration increased rather than declined) and H₃b is partially supported (feeding and foraging was largely stable with an increase by week 3, whereas abnormal behaviour increased contrary to the prediction of no rise).

Discussion

This study investigated the effects of solitary, dyadic, and group housing on the behaviour of zoo-housed aye-ayes and provides preliminary evidence that social housing is behaviourally tolerated in this species.

H₁a predicted that social behaviour would remain a relatively small component of activity budgets (<20%). This prediction was supported with social behaviour accounting for 12.37–16.25% of total scans across individuals. This represents a modest proportion of overall activity budgets, broadly consistent with expectations for a largely solitary species whilst also suggesting latent social plasticity under managed conditions. H₁b predicted that social behaviour would be higher in social housing than solitary housing, with the highest levels in group housing. This prediction was partially supported; social behaviour was significantly higher in all social housing conditions relative to solitary housing, demonstrating that aye-ayes engage in social interactions when opportunities arise. However, contrary to the expectation that group housing would produce the highest levels, differences amongst social housing configurations were not statistically significant. This suggests that the presence of social partners may be more influential than the specific composition of the social group. Together, these findings align with evidence of a more nuanced social organisation in wild aye-ayes (Goodman, 2022), characterised by variable interactions (Sterling, 1993), and are consistent with tolerance of, and potential motivation for together, these findings align with a more nuanced social organisation in wild aye-ayes (Goodman, 2022), characterised by variable interactions (Sterling, 1993), and are consistent with a degree of social tolerance and potential motivation for proximity to conspecifics under specific ecological or developmental contexts. For example, descriptive trends show the juvenile male engaged in social play more frequently than the adults and allocated slightly less time to maintenance, consistent with developmental predictions, although these differences were not statistically significant given the limited sample size.

Additional descriptive trends of higher levels of social play in social housing across all individuals suggest the presence of specific social partners may influence social outcomes. This raises the possibility that the presence of the dam may have facilitated a more affiliative context, that was absent or replaced with a dominance-oriented environment in the sire-juvenile dyad alone, indicated by their highest levels of scent marking in this condition. While this interpretation remains speculative, it highlights a need for further research explicitly examining the role of social partner identity and kinship in shaping interaction patterns. Furthermore, the lack of significant difference in the juvenile’s social behaviour when housed with his sire compared to his dam is not clearly supported or explained by existing literature. While adult male proximity to infants has been reported in wild populations (Andriamasimanana, 1994), including a rare case of infanticide (Rakotondrazandry et al., 2021b), detailed information on sire-juvenile spatial relationships remains limited. Behavioural patterns between sire and offspring here therefore cannot be readily contextualised within a broader framework. Nevertheless, both this result and the general lack of significant effects across housing conditions, may be partly influenced by individuals’ prior familiarity with one another, which could have lessened the behavioural impact of altered housing conditions.

Individuals displayed a wide range of social behaviours, with affiliative interactions such as allogrooming, social foraging, and social play predominating across all social conditions in comparison to agonistic behaviour, which remained rare (<0.35% of total scans). This balance between affiliative and agonistic interactions suggests that, when given the opportunity, these individuals indicate a degree of positive social motivation to engage with one another. Within the limitations of behavioural indicators as proxies for welfare state, the predominance of affiliative over agonistic interactions suggests that social engagement did not coincide with clear behavioural indicators of compromised welfare in this study, indicative of neutral or positive welfare state in line with contemporary affect-based welfare frameworks (Mellor, 2016) and multi-criteria interpretations of animal welfare (Fraser, 2008). This interpretation is further supported by experimental evidence in other nocturnal strepsirrhines, where social housing has been associated with increased affiliative behaviour alongside reductions in physiological stress indicators and stereotypic behaviour (Alejandro et al., 2023). This is broadly consistent with some field observations describing both tolerant associations and occasional dominance-related interactions between wild aye-aye individuals (Sterling, 1993; Sterling and Richard, 1995), and may further suggest that captive aye-ayes retain a level of social flexibility comparable to that documented in wild populations. However, comparisons of findings in this study with wild activity budgets remain difficult due to the scarcity of quantitative data on social time allocation in this species. Despite this the proportion of affiliative behaviours such as social grooming observed here appears broadly comparable to that reported for other socially dispersed nocturnal primates in captive settings (Alejandro et al., 2021). No co-sleeping was recorded between adult individuals; this aligns with reports of predominantly solitary nesting in wild aye-ayes (Ancrenaz et al., 1994) and supports the importance of providing a minimum one nest box per individual in captivity. The persistence of dam-juvenile associations, including social foraging, is consistent with dam-juvenile affiliative behaviours in captivity (Winn, 1994b), and likely reflects prolonged maternal tolerance during extended weaning periods documented in wild populations (Rakotondrazandry et al., 2021a).

H₂a predicted relative to solitary housing, social and dyadic housing will show modest reductions in locomotion, exploration, and maintenance as activity is reallocated towards social interaction. This predication is partially supported; maintenance behaviour increased under solitary housing, which may reflect a redistribution of solo activity in the absence of social opportunities. In contrast, locomotion reduced under solitary housing, consistent with the possibility that social housing may promote increased movement opportunities through social interaction. However, this effect may reflect the reduced space available under solitary conditions, representing an important spatial confound that should be considered when interpreting this finding. Exploration did not differ between solitary and most social conditions, but was higher in the sire-juvenile condition, which may reflect increased environmental investigation and further suggest a less affiliative pairing; however, this interpretation remains speculative.

Feeding and foraging behaviour remained stable and abnormal behaviour did not increase under social housing conditions relative to solitary largely supporting H₂b predictions. These findings suggest that short-term social housing did not disrupt these essential behaviour processes and provide no clear behavioural evidence of adverse welfare effects based on this proxy. On the contrary, social foraging observed across all individuals and social housing conditions may contribute to elevated overall time allocated to feeding and foraging behaviours. Moreover, increased abnormal behaviour under solitary housing supports concerns regarding the potential welfare costs of isolating individuals with demonstrated social motivation, and is consistent with observations of stereotypic and self-directed behaviours in captive aye-ayes under managed conditions (Fuller et al., 2016; Göbel and Wilms, 2025). However, it is important to note that these studies do not isolate solitary housing as a causal factor. Descriptive trends, including the increased abnormal behaviour observed in the social with offspring condition is more difficult to interpret, as this period coincided with the adult female’s pregnancy, which may have influenced behavioural expression independently of housing condition, representing a confounding factor. In addition, given that male home ranges are substantially larger than those of females (Sterling, 1993), encounter rates between males in the wild are likely especially lower than those experienced under captive social housing. Spatial limitations in captivity may therefore influence male-male tolerance differently than mixed-sex housing, offering a potential explanation for the increased abnormal behaviour under the social with offspring condition for the adult male. Nevertheless, behavioural indicators alone provide a limited assessment of welfare, and future studies would benefit from incorporating physiological and health-based indicators such as body condition, weight stability, and endocrine measures over time.

Similarly to H₂a, H₃a predicted modest reductions, or stabilisation, in locomotion, exploration, and maintenance behaviours following social housing overtime. This prediction was partially supported. Locomotion and maintenance behaviours remained relatively stable over the three-week period, indicating a degree of behavioural adjustment. However, exploration increased during the second week of social housing, suggesting ongoing behavioural adaptation rather than immediate stabilisation. H₃b was also partially supported with stable feeding and foraging was stable in week 2 relative to week 1, but showed a significant increase in week 3 relative to week 1, reinforcing the earlier finding that overall time allocated to feeding and foraging may increase under social housing. In contrast, abnormal behaviour increased across the study period, with higher levels observed in weeks 2 and 3 relative to week 1 - particularly by the adult female during week 3, which was not predicted and cannot be clearly explained by known husbandry or health changes. Unmeasured external factors such as visitor presence or noise may have contributed to this variation. Alternatively, this pattern may reflect individual responses to prolonged social exposure, although the small sample size precludes firm conclusions. This study found age and sex were significant predictors of several behavioural categories; however, these effects likely reflect individual-specific variation rather than population-level patterns, due to the non- independence of age, sex, and identity within a sample of three individuals. Individual responses to changing stimuli in captivity are not uncommon (Boyle et al., 2020), but further studies are needed to determine whether individual responses reflect genuine behavioural differences or whether differences are attributable to limited statistical power given the small sample. The persistence of abnormal behaviours across housing conditions more broadly confirms that factors beyond social structure, including environmental conditions, may influence behavioural welfare outcomes and highlight the need for evidence-based evaluations to environmental modifications.

Methodological factors should be considered when interpreting these findings. Instantaneous scan sampling may have underrepresented brief or infrequent behaviours, particularly social interactions if these occurred between scan points. Continuous event sampling with timed-interval scoring of state behaviours could more comprehensively capture both the frequency and qualitative aspects of social interactions. The small sample size represents a major constraint, limiting the generalisability of results, and the adult female’s pregnancy during part of the study introduces an additional confounding factor. Data collection was restricted to visitor hours (10:00–16:30); although observation times were varied to reduce temporal bias, behaviours occurring outside of this period were not captured, potentially underestimating behavioural patterns across the diel cycle. Future studies would benefit from expanding temporal coverage to determine whether behavioural changes stabilise following acclimatisation as well as including time of day as a covariate. Differences in enclosure access between housing conditions represent a spatial confound that may have influenced behavioural outcomes. Consequently, spatial availability, individual history, health status, breeding recommendations, and studbook guidance should inform social housing decisions, with separation facilities remaining essential. While continuous mixed-sex housing carries reproductive management considerations, short-term or managed social exposure outside of oestrus may provide a flexible approach to promoting affiliative behaviour whilst maintaining reproductive control.

Overall, this study provides novel baseline information on the social behaviour of captive aye-ayes under varying housing conditions. The results tentatively support the provision of short-term, managed social housing between familiar individuals, as affiliative behaviours were observed across all social housing conditions without clear behavioural evidence of adverse welfare effects. Extended group housing warrants further investigation to determine whether prolonged exposure is associated with sustained increases in abnormal behaviour or whether these reflect a transient adjustment period. Mixed-sex and adult male-juvenile housing both appeared tolerated in this study, with highest affiliative behaviours observed under group housing, although responses varied between individuals, potentially reflecting the influence of social partners or kinship relationships. Given the limited sample size, these findings should be interpreted cautiously, and multi-female housing was not assessed and cannot be recommended on the basis of the present data.

Importantly, these findings are broadly consistent with evidence from other nocturnal strepsirrhines, where social housing has been associated with increased affiliative behaviour and stable or improved behavioural welfare indicators when appropriate space and environmental complexity are provided. For example, pygmy lorises introduced into group enclosures showed increased social engagement, including co-nesting and affiliative interactions, suggesting greater social flexibility than traditionally assumed in solitary strepsirrhines (Alejandro et al., 2021). While species differences must be considered, such comparative evidence supports the interpretation that social tolerance in captive strepsirrhines may be context-dependent and facilitated by enriched and choice-based housing conditions. Future multi-institutional studies, alongside improved contemporary data on wild social behaviour, will be important for refining social housing recommendations and informing evidence-based behavioural welfare management of this species.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During the preparation of this work, the authors used ChatGPT to improve the language and readability of the paper. After using this tool/service, the authors reviewed and edited the content as needed and took full responsibility for the publication’s content.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Paul Rose, University of Exeter, United Kingdom

Reviewed by: Cristiano Azevedo, Universidade Federal de Ouro Preto, Brazil

Sacha C. Engelhardt, University of Göttingen, Germany

Zoe Newnham, Marwell Zoo, United Kingdom