Sumatran orangutan mothers differ in the extent and trajectory of their expression of maternal behaviour

Abstract

Mothers play a crucial role in the early development and survival of mammalian offspring, and differences in maternal care may affect offspring’s development. Whereas previous research has primarily focused on biological and socioecological factors to understand population-level variation in maternal behaviour, the individual as a source of variation remains understudied. We investigated between-individual variation in the average expression of, and plasticity in, six maternal behaviours in Sumatran orangutans (Pongo abelii), using 15 years of behavioural data. We found that mothers differed substantially in the average expression of four maternal behaviours, even after controlling for socioecological conditions, biological state characteristics and the offspring’s influence on these behaviours. Furthermore, not controlling for these confounding effects exaggerated or masked between-individual variation. Mothers also substantially differed in how they adjusted three of the maternal behaviours during offspring development, meaning that mothers differed in behavioural plasticity. Our results suggest that Sumatran orangutan mothers are constrained in the average expression of maternal behaviours and their plastic responses, potentially resulting in consistent differences among mothers, otherwise called maternal personality. Our findings highlight that there is biologically meaningful variation around the population mean in maternal behaviour and present novel opportunities to study evolutionary processes that shape maternal behaviour.

1. Introduction

In mammals, mothers are crucial for their offspring’s healthy development, survival into adulthood and eventual successful reproduction [1,2]. Mothers achieve this by providing protection, social learning opportunities and nutritional, coalitionary and reproductive support to their offspring [3–7]. Ultimately, maternal behaviour is an adaptive trait affecting the fitness of both the mother and the offspring [5].

Maternal behaviour towards offspring is particularly varied in long-lived species—such as primates [4]—in which mothers support their offspring through a lengthy developmental period, and the mother–offspring relationship often extends far beyond weaning [5,7]. In several primate species, the extent of expression of a maternal behaviour does not always completely overlap among all the individuals in a population, i.e. there may be behavioural variation among mothers [8] (henceforth, between-individual variation in the average expression of a maternal behaviour). Although such between-individual variation among mothers can arise from the differences in mother–offspring characteristics and their current socioecological conditions, previous research suggests that there may be fundamental differences among mothers, which may remain consistent across different conditions, including across their different offspring and different social and ecological environments [9,10].

Consistent differences among females in maternal behaviour can emerge through differences in social and environmental conditions that they faced during their early developmental period or they can be intrinsic [11,12]. In humans, consistent behavioural differences among individuals are often termed 'personality', and this term is increasingly applied to non-human species as well [13–15]. Although maternal behaviours are rarely included in the traits used to assign personality, they are interesting candidates to investigate consistent behavioural differences. Differences among mothers may lead to differences in the quantity and quality of offspring-directed maternal behaviours. In human and non-human primates, the quantity and quality of offspring-directed maternal behaviours may influence offspring’s pace of development, exploration tendencies, cognitive abilities, learning, skill acquisition, social behaviour and, ultimately, survival [16–20], (but see [21]). Therefore, beyond the immediate presence of the mother [1,2], the nature of her investment plays an important role in the successful rearing of her offspring.

Consistent differences between individuals in maternal behaviour over time do not mean that mothers will not exhibit short-term variation in their behaviour within their limits. Indeed, maternal behaviour is usually flexible in nature, in that individuals alter the expression of their behaviour in response to changes in their ecological and social environment [8]. Maternal behavioural expression is therefore seldom constant, even within one offspring’s developmental period. Flexibility in the expression of maternal behaviour is expected to be especially pronounced in long-lived, large-brained (relative to body size) species, as they are more likely to face fluctuating socioecological conditions during their lifetime and have a greater cognitive potential to exhibit behavioural flexibility [22–24]. Long-lived, large-brained species were thus hypothesized to achieve adaptation to prevailing conditions through being behaviourally flexible (otherwise called behavioural plasticity) [22]. Primate mothers, for example, flexibly vary the expression of maternal behaviours in response to experienced ecological conditions such as food availability or predation risk [25,26] and social conditions such as social unit size, dominance structure and presence of kin or a male in the social unit [27–29]. Primate mothers also vary their maternal behaviours as a function of their biological and social characteristics such as body condition, parity, age and dominance status [30,31], and of their offspring’s characteristics—most importantly, offspring age [30–34]. In other words, often there is within-individual variation or behavioural plasticity in maternal behaviour when measured over the ranges of different predictors.

Behavioural plasticity in maternal behaviour is hypothesized to positively affect female lifetime reproductive success [3]. Two recent studies found that not all primate mothers exhibit similar degrees of plasticity in maternal behaviours, such as offspring-directed communication, time spent in proximity to their offspring and retrieving, restraining or leaving their offspring [21,35]. Although behavioural plasticity and between-individual variation in maternal behavioural plasticity remain largely understudied in primates, a study on a non-primate species found that females with a higher phenotypic behavioural plasticity also had a higher reproductive success [36].

Even though behavioural traits from an individual’s perspective can be viewed through the above two contrasting ideas of consistency and flexibility, most behaviours are neither rigid nor unlimitedly plastic, i.e. they are shaped by both the flexible and consistent nature of individuals [8]. Estimating between-individual variation in the average expression of a behaviour is often complicated by the fact that between-individual variation is confounded by the factors influencing within-individual variation (i.e. confounding effects of maternal behaviour). Therefore, only through the simultaneous investigation of between- and within-individual variation can we fully understand the role of individuals in the expression of maternal behaviour.

To understand the adaptive value of between- and within-individual variation, it is important to investigate these in wild animals. However, estimating between- and within-individual variation in maternal behaviour using field data is challenging, as data on wild individuals are rarely complete owing to the practical challenges associated with uniformly sampling the individuals through an extended period of offspring development, such as in primates. For example, individuals cannot always be found in the study area or only a limited number of individuals can be observed simultaneously owing to logistic constraints. This introduces sampling discrepancies among individuals, resulting in a significant risk of wrongly inferring the presence of between-individual variation in maternal behaviour when there is none (i.e. false-positive difference), or a failure to detect the presence of between-individual variation when it is actually present (i.e. false-negative similarity), as discussed further in [21]. These constraints raise the question of whether studies actually measure consistent, between-individual variation when they do not control for the confounding effects (i.e. offspring age, offspring sex, mother’s parity, sample size, etc., as discussed above).

Behavioural reaction norms [36–39] offer a way to reliably and simultaneously estimate between- and within-individual variation in maternal behaviour, especially when faced with sampling discrepancies and/or confounding effects. The approach entails controlling for confounding effects and estimating between- and within-individual behavioural variation—without isolating them—using a mixed model approach (electronic supplementary material, box S1), allowing one to estimate between-individual variation relative to within-individual variation. The approach has been successfully applied to several behavioural traits, including maternal behaviours, movement behaviour, calving date, egg laying date, chick provisioning behaviour and male territorial aggressive behaviour, in different species [36,39–44]. Studies on mammalian maternal behavioural variation using behavioural reaction norms are sparse (but see [21,35]). As long-term data on individually identified animals and large sample sizes are necessary to identify individual variation, such studies are challenging in species with slow life-histories and prolonged maternal care.

In this study, we aimed to shed light on the extent to which maternal behaviour in wild Sumatran orangutans (Pongo abelii) is shaped by individual variation by integrating both consistent individual differences and variation in individuals’ plasticity. We investigated between-individual variation in the average behavioural expression and between-individual variation in plasticity in six maternal behaviours while controlling for previously established predictors of, as well as the offspring’s potential influence on, maternal behaviour, using longitudinal and cross-sectional behavioural data. Previous studies on individual variation in maternal behaviour in primates did not control for all known confounding effects before quantifying individual variation [9,21,33,35]. Therefore, one of the aims of this study was to identify whether not controlling for known confounding effects (i.e. the biological characteristics of the mother and/or offspring, the prevailing socioecological conditions experienced by the mother–offspring pair, sample size and sampling scheme; see §2) would influence the estimation of between-individual variation in the average expression of maternal behaviour.

Sumatran orangutans are large-brained and long-lived arboreal apes, living for up to at least 50−60 years in the wild [45]. Sumatran orangutan females start reproducing from around 15 years of age [46]. With a duration of about 8 years, their inter-birth interval is the longest among all the non-human primates [46]. Sumatran orangutans are an excellent study system in which to investigate individual maternal behavioural variation because mothers provide care to each of their offspring for about 6−9 years without help from other individuals [47], through extensive caretaking behaviours, such as carrying, bridging, nest sharing and maintaining proximity to offspring, while serving as role models for the offspring’s acquisition of learning-intensive subsistence skills, such as foraging and nest building [48,49].

In line with the hypothesis that individuals may be constrained in their behavioural expression owing to their genetic and/or developmental background [11,12], a previous study on wild and captive orangutans found that orangutans show between-individual variation in communicative behaviours, including offspring-directed maternal communication and responsiveness [35]. Based on this result and the results from studies on other primates [21,27,33,50],

• —we predicted that Sumatran orangutan mothers would differ in their average expression of maternal behaviours even after controlling for known confounding effects. Since a mother may face differential demands from her different offspring, we predicted that a part—but not all—of the variation among mothers in their expression of maternal behaviour will be explained by the offspring’s identity.

In previous work on the study population [34], we found that offspring’s biological characteristics (i.e. offspring age, offspring sex) and experienced socioecological factors (i.e. food availability, association size and presence of a male) had significant effects on several maternal behaviours. We thus hypothesized that the estimation of between-individual variation in the average expression of maternal behaviour is influenced by whether these known confounding effects are controlled for in the analysis.

• —We predicted that not controlling for the confounding effects would either exaggerate or mask the estimated between-individual variation in the average expression of maternal behaviour and will increase the uncertainty of these estimates.

The extent of behavioural plasticity is hypothesized to be positively correlated with brain size [22]. Accordingly, orangutans have been found to show considerable plasticity in maternal behaviours and maternal gestural communication [34,35]. Since behavioural plasticity is expected to balance the costs and benefits of maternal behaviour and given that mothers are unlikely to face similar pressures during their offspring’s developmental period, plastic responses may differ among Sumatran orangutan mothers. In fact, a recent study found that mothers differed in how they altered their responsiveness towards their offspring depending on the context of their interaction [35].

• —We predicted that Sumatran orangutan mothers will differ in their plasticity in maternal behaviours—i.e. mothers will differ in their maternal behavioural trajectories over offspring age. Since differences among offspring in their demands on the mother may vary across the developmental period, we predicted that offspring identity would contribute to maternal behavioural plasticity but not fully account for it.

To carry out this study, we focused on maternal behaviours related to proximity maintenance, skill acquisition and locomotory support [34] in 15 wild Sumatran orangutans at the Suaq Balimbing monitoring station in South Aceh, Indonesia.

2. Methods

(a). Study site and data collection

To study individual variation in maternal behaviour, we used long-term data on six maternal behaviours, namely initiation and termination of body contact and close proximity, carrying and feeding in close proximity (described in electronic supplementary material, table S1), that were collected on wild Sumatran orangutan mother–offspring pairs (Pongo abelii). The data were collected between 2007 and 2022 in the Suaq Balimbing research area, which is a coastal peat swamp forest in the Gunung Leuser National Park in South Aceh, Indonesia. These behaviours are commonly expressed by primate mothers and are used to measure the dynamics of the mother–offspring relationship in many primate and non-primate species [21,34,51–53].

Behavioural data were collected during focal follows, conducted continuously from morning nest to night nest when possible, based on an established protocol for orangutan data collection (https://www.ab.mpg.de/571325/standarddatacollectionrules_suaq_detailed_jan204.pdf). During the focal follows, scan sampling [54] was conducted every 2 minutes, during which the activity of the focal individual (i.e. a mother with a dependent offspring) and distances (measured in classes of 0 m (contact), >0–2, 2−5, 5−10 and 10−50 m) between the focal individual and all the association partners (of any age) were noted down. A party was defined as two or more individuals that were within 50 m of one another. When the distance class changed between two scans, the observers noted down the identity of the individual initiating the change. Further details about the study site and data collection are provided in previous publications [47,48].

(b). Sample size

We had behavioural data on 15 mothers, for nine of which we had data on one dependent offspring (between birth and 8 years of age), and for 6 others we had data on more than one offspring (three mothers with offspring of either sex; three mothers with either a male or a female offspring; electronic supplementary material, S2, table S1). Although data on multiple offspring per mother are ideal for detecting differences among mothers that are independent of offspring characteristics (e.g. offspring sex), such datasets are difficult to obtain in species with long inter-birth intervals. Including data on multiple offspring for some mothers and data on only a single offspring for others can conflate between-mother and between-offspring differences, which can make it harder to tease apart their effects. To avoid erroneously over- or underestimating differences among the mothers, we conducted two sets of analyses: in the first set of analyses, we included data on all of the mother–offspring pairs (i.e. data on some mothers with multiple offspring while others had only one offspring); in the second set of analyses, we retained data on only those mothers with multiple offspring (i.e. data on all mothers with multiple offspring). The latter set of analyses was conducted only for those behaviours for which we had data on multiple offspring per mother for at least five mothers to get reliable random effects estimates of mother and offspring identities.

We operationalized maternal behaviour as the proportion of scans in which the mother showed a behaviour, where the denominator controls for the opportunities available to the mother to show that behaviour (e.g. number of scans during which a mother moved was the denominator for the behaviour carrying; electronic supplementary material, S1, table S1). As a result of opportunistic sampling and as we only included follows during which the mother had the opportunity to show the analysed maternal behaviour at least once, the number of focal follows and total observation duration differed among the mothers and across behaviours (electronic supplementary material, S2, table S1). The number of focal follows per behaviour ranged between 622 and 868 (i.e. the number of days during which focal individuals were followed, and the behaviour could be assessed), which amounted to between 5853.7 and 7915.1 h of observation time (electronic supplementary material, S2, figure S1). The range of offspring age during which maternal behaviour could be recorded (i.e. sampling scheme) varied substantially among the offspring (electronic supplementary material, S2, figure S2).

(c). Statistical analysis

(i). Estimating between-individual variation in average maternal behavioural expression and plasticity

We used a generalized linear mixed effect model (GLMM) to quantify between-individual variation in the average expression of a behaviour and between-individual variation in behavioural plasticity. In a recent study on the Suaq orangutan population, we found that certain offspring characteristics (offspring age (continuous), offspring sex (binary; male/female) and socioecological factors (prevailing food availability (henceforth, FAI; continuous), average association size during a follow (continuous) and male presence during a follow (binary; presence/absence)) were significant predictors of different maternal behaviours, but not of a mother’s parity [34]. Therefore, as detailed in §1, any variation among mothers in their maternal behaviour is likely to be at least partly caused by the variation introduced by these predictors. Hence, we modelled each maternal behaviour as a function of the respective known significant predictors of that behaviour. Further details about these predictors are available in [34; electronic supplementary material, S3, table S1]. The sampling scheme discrepancies in our study (electronic supplementary material, S2, figure S2) additionally underscore the importance of including offspring age as a predictor. We z-standardized the continuous predictors to a mean of zero and a standard deviation of one, and we dummy-coded and mean-centred offspring sex and male presence/absence to ease model convergence and interpretation [see footnote of electronic supplementary material, S4, table S1]. We included a random intercept of mother identity and a nested random intercept of offspring identity within mother identity because there were mothers with more than one offspring. The estimated standard deviation of random intercepts of mother identity gives us the extent of between-individual variation in the average expression of maternal behaviour. The estimated standard deviation of random intercepts of offspring identity gives us the extent of variation among offspring of the same mother. A non-zero estimate of the standard deviation of random intercepts of the mother and offspring identities would mean that offspring of the same mother differentially influence maternal behaviour and also that mothers behave in a consistent manner towards each of their offspring. A non-zero estimate for the mother but not for the offspring identity would mean that mothers behave in a consistent manner towards each of their offspring and that offspring of the same mother do not differentially influence maternal behaviour. Alternatively, a non-zero estimate for the offspring but not for the mother's identity would mean that offspring of the same mother differentially influence maternal behaviour and that mothers do not behave in a consistent manner with each of their offspring. We further included random slopes of offspring age within mother identity and offspring identity. The estimated standard deviation of random slopes within mother identity gives us the extent of between-individual variation in maternal behavioural plasticity. We could not calculate repeatability (calculated as variance among individuals/variance within individuals over time; [42,55]) to get population-level estimates of the fraction of maternal behavioural variation that is owing to between-individual variation, as models included complex random effect structure with multiple random slopes.

(ii). Modelling procedure

We fit the GLMMs explained above (henceforth, full model) using a Bayesian framework in R, v. 4.2.2 [56], using Stan, v. 2.32.5 [57], through the brms package [58], which retrieves posterior distributions of estimated parameters. We fit the models using a beta-binomial error distribution for initiation and termination of body contact and close proximity (events), and with a beta error distribution for carrying and feeding in close proximity (which are both states rather than events), both with a logit link function. We included the numbers of scans during which the mother showed a behaviour as the number of successes and the opportunities to show a behaviour (electronic supplementary material, S1, table S1) as the number of trials in the beta-binomial models. We operationalized the dependent variables as proportions in the beta models. To examine the goodness-of-fit of the models, we computed Bayesian R² using the bayes_R2 function.

(iii). Statistical support for random effects

In addition to each of the full models, we fit a model excluding the random intercepts of mother identity and offspring identity nested within mother identity and random slope of offspring age (henceforth, fixed effects model) and one including the random intercepts of mother identity and offspring identity nested within mother identity but excluding the random slope of offspring age (henceforth, random intercept model). We then compared these models using the leave-one-out cross-validation (LOO-CV) as implemented in the loo package (v. 2.7.0) [59] using the loo function. LOO-CV allowed us to examine whether the out-of-sample predictive performance of a model is improved by the addition of model terms (here, the random intercepts and slopes), which is analogous to the Akaike information criterion. We chose the more parsimonious model whose ΔELPD ± 2 × s.e. did not overlap zero to draw inferences about between-individual differences in maternal behaviour.

(iv). Influence of confounding effects

To understand whether and how not controlling for known confounding effects affect the estimation of between-individual variation in the average expression of maternal behaviour, we fit another model to each of the maternal behaviours without any fixed effects (i.e. not controlling for the known confounding effects) but with the random intercepts of mother identity and offspring identity nested within mother identity. We then compared the estimated between-individual variation in the average expression of maternal behaviour and the uncertainty associated with it between the full model and the above model. We assessed these results via visual inspection.

We report the population-level effects in the electronic supplementary material (electronic supplementary material, S4, table S1, figure S1).

3. Results

(a). Between-individual variation in the average expression of maternal behaviour

After controlling for mother–offspring characteristics, socioecological conditions and the offspring’s influence on maternal behaviour, we found significant and consistent (i.e. across one’s different offspring) between-individual variation among mothers in the expression of four of the maternal behaviours, namely contact termination, close proximity termination, carrying and feeding in close proximity (electronic supplementary material, S5, table S1); the posterior distributions of standard deviation of the random intercept of mother identity did not overlap zero in these behaviours (electronic supplementary material, S6, figure S1). Between-individual standard deviations for mothers ranged between 0.32 and 0.90 across these models (table 1). Offspring of the same mother contributed to, but did not fully explain, the variation in mothers’ expression of maternal behaviours (s.d. of the random intercept of offspring ranged between 0.47 and 0.71; table 1, figure 1, electronic supplementary material, S6, figure S1). Furthermore, mothers differed in their average expression of maternal behaviour across behaviours, and there was a significant, negative correlation between the random effect intercepts of mother identity in the carrying model and the model for feeding in close proximity (but not between any of the random effect intercepts of any of the other behaviours; electronic supplementary material, S7, table S1).

Table 1.

Estimated between-individual variation in the average expression of a behaviour and between-individual variation in behavioural plasticity. Means and standard deviations of the posterior distribution of the parameters, their 95% credible intervals (CI), along with effective sample size (ESS) for the mother and offspring random effects are shown for the four maternal behaviours in which there was significant between-individual variation among mothers. MotherID : OffspringID denotes nested random effect. *Offspring age was z-transformed. Mean and s.d. of the raw variables are provided in electronic supplementary material, S4, table S1.

random effect	estimate	estimated error	95% CI	rhat	bulk ESS	tail ESS
contact termination; random intercept model; n = 719 focal follows
~motherID (number of levels: 11)
s.d.(Intercept)	0.90	0.45	[0.13, 1.95]	1.00	588	547
~motherID : offspringID (number of levels: 14)
s.d.(intercept)	0.57	0.33	[0.14, 1.37]	1.01	618	1153

close proximity termination; full model; n = 716 focal follows
~motherID (number of levels: 12)
s.d.(intercept)	0.32	0.23	[0.01, 0.87]	1.00	693	1300
s.d.(offspring age)	0.33	0.23	[0.02, 0.86]	1.00	705	1394
Cor(intercept,offspring age)	0.11	0.57	[−0.93, 0.96]	1.00	1292	1732
~motherID : offspringID (number of levels: 16)
s.d.(intercept)	0.47	0.17	[0.23, 0.88]	1.00	1315	1932
s.d.(offspring age)	0.37	0.17	[0.14, 0.79]	1.00	1157	1661
cor(intercept,offspring age)	−0.52	0.39	[−0.99, 0.38]	1.00	969	2007

carrying; full model; n = 830 focal follows
~motherID (number of levels: 13)
s.d.(intercept)	0.66	0.37	[0.05, 1.46]	1.00	512	966
s.d.(offspring age)	0.69	0.38	[0.04, 1.50]	1.01	589	741
cor(intercept,offspring age)	0.46	0.46	[−0.73, 0.98]	1.00	666	1246
~motherID : offspringID (number of levels: 20)
s.d.(intercept)	0.71	0.32	[0.26, 1.43]	1.00	380	1438
s.d.(offspring age)	0.72	0.25	[0.34, 1.34]	1.00	596	1343
cor(intercept,offspring age)	−0.25	0.51	[−0.98, 0.68]	1.00	371	986

feeding in close proximity; full model; n = 868 focal follows
~motherID (number of levels: 15)
s.d.(intercept)	0.60	0.34	[0.04, 1.31]	1.00	510	969
s.d.(offspring age)	0.16	0.15	[0.01, 0.54]	1.00	1400	1950
s.d.(offspring age2)	0.17	0.13	[0.01, 0.50]	1.00	1364	1712
Cor(intercept,offspring age)	0.09	0.49	[−0.84, 0.89]	1.00	2951	2306
Cor(intercept,offspring age2)	−0.21	0.50	[−0.94, 0.82]	1.00	2454	2155
Cor(offspring age,offspring age2)	0.05	0.50	[−0.87, 0.89]	1.00	2484	2442
~motherID : offspringID (number of levels: 22)
s.d.(intercept)	0.59	0.27	[0.12, 1.16]	1.00	608	1400
s.d.(offspring age)	0.17	0.15	[0, 0.57]	1.00	985	1595
s.d.(offspring age2)	0.38	0.12	[0.17, 0.66]	1.00	1077	1274
Cor(intercept,offspring age)	0.14	0.49	[−0.82, 0.91]	1.00	2070	2143
Cor(intercept,offspring age2)	−0.02	0.40	[−0.68, 0.79]	1.00	684	866
Cor(offspring age,Oofspring age2)	0.17	0.49	[−0.82, 0.92]	1.01	464	764

Figure 1.

Between-individual variation in the average expression of four maternal behaviours after controlling for all the known confounding effects. Posterior distributions (on the logit scale) are shown for (A) body contact termination (random intercept model), (B) close proximity termination (full model), (C) carrying (full model) and (D) feeding in close proximity (full model) for the individual mothers and offspring relative to the estimated population-level average. Black dots show the posterior mean for each individual. Positive estimates indicate higher-than-average behavioural expression, while negative estimates indicate lower-than-average behavioural expression. Random intercepts of the mothers were calculated based on their behaviour with 1−3 offspring. Maternal behavioural variation did not purely result from the variation among offspring, as mothers showed consistent, significant behavioural differences regardless of whether their offspring behaved similarly. For example, in close proximity termination, Sarabi’s offspring Simba and Sazu were both below the population average, while in carrying, Friska’s offspring Fredy and Frankie were below and above the average, respectively.

For two other behaviours, contact initiation and close proximity initiation, the models’ predictive performance did not improve with the addition of random effects of mother identity (electronic supplementary material, S5, table S1), suggesting that mothers did not significantly differ from one another in these behaviours.

We found the largest estimated between-individual variation among mothers for the behaviours contact termination and carrying (table 1, figure 1A,C), while close proximity termination and feeding in close proximity had lower estimated between-individual variation among mothers (table 1, figure 1B,D). Furthermore, the estimated average expression of a behaviour for each mother varied across the behaviours (figure 1A–D).

(b). Influence of confounding effects

When we compared the estimated between-mother variation in the average expression of maternal behaviour obtained from models in which we did and did not control for the known confounding effects, for three of the analysed maternal behaviours we found that not controlling for the confounding effects resulted in a higher estimated between-mother variation than controlling for these effects (figure 2). In contrast, for three other behaviours, not controlling for the confounding effects reduced the estimated between-individual variation. This difference was most pronounced for body contact initiation and termination, where the models that did not control for the known confounding effects (electronic supplementary material, S3, table S1) estimated the between-individual variation to be more than twice as high as the models that did control for the confounding effects (figure 2). The differences between the estimated between-individual variations were moderate for close proximity termination and feeding in close proximity and negligible for close proximity initiation and carrying. Additionally, for three of the behaviours (i.e. body contact initiation, termination and close proximity termination), the 95% credible intervals associated with the estimates were wider when the confounding effects were not controlled than when they were controlled for in the analysis (figure 2). Conversely, for the rest of the behaviours (i.e. carrying, close proximity initiation and feeding in close proximity), the 95% credible intervals associated with the estimates were wider when the confounding effects were controlled than when they were not controlled for in the analysis (figure 2).

Figure 2.

Comparison of the estimated differences among mothers in maternal behaviour. Estimated between-individual variation in the average expression of maternal behaviour (i.e. the estimated standard deviation of mother identity intercepts) is shown for each of the six maternal behaviours when the respective confounding effects were (full model) and were not controlled for in the analysis. The vertical lines represent 95% CI associated with the estimates. BCI, body contact initiation; BCT, body contact termination; CAR, carrying; CPI, close proximity initiation; CPT, close proximity termination; FCP, feeding in close proximity.

(c). Between-individual variation in maternal behavioural plasticity

For three of the behaviours—close proximity termination, carrying and feeding in close proximity—there was evidence of significant and consistent (i.e. across their different offspring) between-individual variation among mothers (s.d.: 0.16−0.69; table 1) in behavioural plasticity over offspring age (electronic supplementary material, S5, table S1). Furthermore, offspring identity contributed to, but did not fully explain, the variation in maternal behavioural plasticity over offspring age (s.d.: 0.17−0.72). The largest between-individual variation in behavioural plasticity among mothers was observed for carrying (table 1, electronic supplementary material, S8, figure S1). The significant effect of random slope of offspring age within mother identity suggests that mothers varied in how they adjusted their behaviour as their offspring developed, leading to mother-specific trajectories of maternal behaviour over offspring age (figure 3).

Figure 3.

Consistent between-mother variation in behavioural plasticity. Fitted models and their credible intervals (95%, 80% and 50% in grey) are shown, as an example, for close proximity termination for (A) two offspring of the same mother and (B) and (C) two offspring of two unrelated mothers from the full model to visualize the distinct trajectories of maternal behaviour over offspring age. Plots without the datapoints are shown in the insets for clearer visualization of the trajectories.

When we repeated the analysis with only those mothers with more than one offspring for carrying and feeding in close proximity, we again found significant and consistent between-mother variation in the average expression of maternal behaviour, as well as its plasticity (electronic supplementary material, S9, tables S1 and S2, figures S1 and S2).

4. Discussion

Much research to date on primate maternal behaviour has focused on the effects of the mother's and offspring's biological characteristics, of social factors and of ecological conditions on maternal behaviour [30,34,60,61], but little attention has been paid to individuals as a source of variation. Here, we used behavioural reaction norms to investigate between-individual variation in the average expression of maternal behaviour and between-individual variation in behavioural plasticity during offspring development in Sumatran orangutans. We found that mothers substantially differed in their average expression of maternal behaviour even after controlling for known confounding effects. Furthermore, we found that not controlling for the confounding effects either exaggerated or masked estimated between-individual variation. We further found that mothers differed in how they modified their behaviour in response to their offspring’s increasing age—i.e. mothers showed individualistic trajectories of maternal behaviour during offspring development and, thus, displayed between-individual variation in behavioural plasticity. Importantly, we found that the identity of the offspring also contributed to variation among mothers in the average expression of maternal behaviour and plasticity in maternal behaviour over offspring age. However, maternal variation remained significant even after accounting for the offspring’s contribution, suggesting that mothers behave in a consistent manner across their different offspring [33].

As predicted, our analyses revealed substantial between-individual variation in the average expression of carrying, feeding in close proximity, body contact termination and close proximity termination among mothers when controlling for confounding effects. These findings suggest that differences in the expression of maternal behaviours among Sumatran orangutan mothers extend beyond the effects of immediate socioecological conditions. Differences among mothers in maternal behaviours, such as carrying, rejection, restraint and ventral contact have been previously found in Guinea baboons (Papio papio), rhesus macaques (Macaca mulatta) and vervet monkeys (Chlorocebus pygerythrus) [9,21,33], but these studies did not control for all likely confounding effects, leading to uncertainty in the extent of estimated variation. Furthermore, a recent study found individual differences in offspring-directed maternal communication in orangutans. However, this study included only one offspring per mother, making it impossible to tease apart the contribution of the mother versus the offspring to maternal behavioural variation [35]. Our results indicate that the differences in maternal behaviour are likely consistent in individual mothers across their different offspring, strengthening the notion that there is personality—specifically maternal personality—in non-human primates [11]. In species with large brains, developmental effects that lead to consistent individual differences are likely to occur [62]. This is because, as brain size increases relative to body size, developmental effects become more important since large-brained species are more sensitive to developmental inputs [63] and a large share of brain development happens after birth [64]. In Sumatran orangutans, including in our study population, body contact initiation and close proximity initiation with offspring are both rarely shown by mothers [34]. Accordingly, most of the data points for these behaviours in our dataset were zero, which makes it difficult to discern variation with the currently available analysis techniques.

In our study, the magnitude of variation among mothers was higher for body contact termination and carrying than that for close proximity termination and feeding in close proximity. In orangutans, carrying is a form of locomotory support [34,49] and close-range feeding can provide social learning opportunities for offspring [48], whereas body contact and proximity maintenance are often associated with protection/thermoregulation in primates [65]. Carrying is energy intensive [66] and close-range feeding may increase feeding competition and thus also have adverse energetic effects on the mothers. Individual differences in the expression of a maternal behaviour could be the consequence of differences in the maximum investment that individual mothers can show (based on their genetic foundation or irreversible effects of early developmental experience). A mother’s genetically or developmentally determined body condition may thus constrain her investment in energetically critical maternal behaviour. However, since mother–offspring interactions are always dyadic in nature, it is possible that the differences among mothers may be at least partly driven by differences among their offspring. As predicted, we found non-zero variation among offspring of the same mother, suggesting that differences among offspring in their characteristics and demands can lead to differential expression of maternal behaviour. However, offspring explained only part of the variation in maternal behaviour.

Mothers differed in their average expression across the different behaviours. In primates, variation in maternal behaviours usually falls along two dimensions, namely protection and rejection [27,60], meaning that a mother’s behaviour in one dimension does not predict her behaviour in another dimension. Our results showed that orangutan mothers exhibit distinct individual patterns of investment across the different maternal behaviours. This dissimilarity suggests that the maternal behaviours that we analysed may fall on different dimensions in Sumatran orangutans, like that seen in captive chimpanzees and bonobos and many wild primates [17,27,60]. However, mothers who showed higher than average carrying showed significantly less than average feeding in close proximity with their offspring. In Sumatran orangutans, carrying is at its peak for the first two years of life and steeply declines thereafter, and feeding in proximity peaks between 5 and 6 years of age and then declines [34]. This suggests that early life investment in offspring may lead to faster feeding skill acquisition, resulting in a decrease in feeding in close proximity between the mother and the offspring later in life.

As predicted, we found that failing to appropriately control for the known confounding effects while estimating between-individual variation in the average expression of maternal behaviours can result in exaggerated estimations of variation among individuals. When differences in the biological characteristics and the socioecological environment under which the different mother–offspring pairs were observed [34] cause differences among-individual behaviour, then by not deliberately adding these factors as predictors in their models, researchers risk wrongly assigning these differences to individual effects (i.e. creating false-positive between-individual differences) [21]. In addition, as predicted, our results show that one may also end up underestimating individual variation (i.e. creating false-negative similarity between individuals) when not controlling for confounding effects. Failure to control for the confounding effects not only affects the estimated variation between mothers but also the uncertainty associated with it. Credible intervals were wider when confounding effects were not controlled for than when they were controlled for in half of our models. As the values of the different predictors, such as offspring age, association size, etc., change over time and as females are responding to the changes in these variables [34], differences in the effects of these variables further add to the inherent differences among individuals, resulting in increased uncertainty around the estimates. In general, this result highlights the importance of controlling for the confounding effects in animal personality studies.

Although variation in behavioural plasticity is usually studied along an environmental or a social gradient [35–37,43], we investigated whether mothers differ in how they adjust their behaviour during offspring development. As predicted, our analyses revealed that offspring identity contributed to differences in behavioural plasticity among mothers. Second, our analyses revealed substantial variation in behavioural plasticity among mothers in response to offspring’s age in close proximity termination, carrying and feeding in close proximity, even after accounting for the offspring’s contribution. This means that mothers differed in how they adjusted their behaviour during offspring development. This may be a result of some mothers being more plastic than others. As with between-individual difference in the average expression of these behaviours, because body contact and close proximity initiation are rarely shown by mothers, it is difficult to discern between-individual variation in plasticity in these two behaviours. Our analyses did not detect substantial between-individual variation in plasticity in contact termination, suggesting that the range of behavioural expression for body contact termination is indeed similar across mothers, despite the significant between-individual variation in the average expression.

Differences in mothers’ physiological conditions can lead to differences in their ability or willingness to respond to their offspring’s needs, which may be a source of individual differences in plasticity in maternal behaviour. In line with this reasoning, we found the most pronounced between-individual variation in plasticity for the most energetically costly behaviour, namely carrying. Maternal condition was also found to influence between-individual differences in time spent in and out of contact with offspring and frequency of rejection in other primates [61]. Furthermore, as evidence suggests that orangutan mothers adjust their behaviour in response to their offspring’s age [34], variation in the development of their offspring’s locomotory skill levels may lead to differences in the mothers’ readiness to carry them. In other words, differences in offspring’s pace of development and, especially, variation in the magnitude of these differences can bring about individual differences in maternal behavioural plasticity. However, we controlled for offsping differences in all our models. In addition, differences in offspring health could drive differences in maternal behaviour. However, all the focal offspring in this study were in apparent good health, and none died during or after the study (consistent with published extremely low infant mortality [46]).

Overall, our results support that there is consistent between-individual behavioural variation in maternal behaviour in Sumatran orangutans. However, as with any field-based study, individual mothers differed in the conditions under which they were observed, which may affect both individual differences in the average expression and plasticity of the maternal behaviours. Even though we controlled for all previously established confounding effects [34], there may be other biological state characteristics or socioecological variables that affect variation in maternal behaviour that we currently cannot quantify or are unaware of. Furthermore, the complexity of these effects may exceed the way we quantify them. For example, we took into account the current presence or absence of males because of their known effects on maternal behaviours, which likely arise from the risk they may pose to immature individuals [34,67]. However, individual males likely differ in the risk they pose to the offspring [68] and thus also in the effects they have on maternal behaviour. Controlling for more accurately quantified prevailing social risk faced by offspring would thus be a better approach, although this is not feasible with the current dataset. Despite our study making use of around 6000 h of behavioural observation conducted over the span of 15 years, it is constrained by the limited number of focal individuals, the offspring’s lengthy developmental periods and long inter-birth intervals. Accordingly, sample size and sampling scheme substantially varied among the mothers, which inevitably limits the inferences possible with our study. For a more accurate quantification of between-individual differences in maternal behaviour, we need data on mother–offspring pairs spanning the entire range of offspring developmental periods and, ideally, data on multiple offspring for all the mothers.

Our study aimed to disentangle meaningful patterns from noise through partitioning observed variation in maternal behaviour. First, our results suggest that Sumatran orangutan maternal behaviour is shaped by individuality, apart from mother–offspring social and biological state characteristics and socioecological conditions [34]. Our findings thus add to previous research on maternal personality in primates and other species in showing that there are consistent differences among mothers in their behaviour towards offspring [9,21,27,33,40,52]. Second, our results signify the importance of expanding the research focus from average behaviour to individual behaviour to capture biologically meaningful variation around the population mean. Consistent between-individual variation in the average expression of behaviour and plasticity can have significant consequences and present novel opportunities for studies of the evolutionary processes that shape the behaviour.

For consistent, heritable variation to undergo natural selection, there must be fitness consequences associated with it. Investigating whether differences in maternal behaviour are associated with differences in the speed at which offspring reach developmental milestones and the offspring’s future reproductive success will shed light on the adaptive value of maternal behaviour for both the mother and the offspring. Our study also sets the stage to address whether daughters inherit maternal personality—socially and/or biologically—potentially giving rise to matrilineal differences in maternal behaviour [9,50]. Ultimately, studies on individual variation in maternal behaviour can help shed light on the causes, maintenance and consequences of such variation.

Ethics

All observers kept a minimal distance of 7 m away from the orangutans. Since the orangutans of the study population are arboreal, this distance was in fact substantially larger than 7 m for most of the time. This minimizes the effect of observer presence on the natural behaviour of the individuals and the risk human observers present to them. Data collection was purely observational and non-invasive. The research followed the recommendations of the Animals (Scientific Procedures) Act 1986, as published by the UK government, and the principles of Ethical Treatment of Non-Human Primates as stated by the American Society of Primatologists. Approval for this study was obtained from the Indonesian State Ministry for Research and Technology (RISTEK, to date called National Research and Innovation Agency of Indonesia, BRIN).

Data accessibility

Data and original code are deposited at Dryad data repository [69].

Supplementary material is available online [70].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

T.R: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, writing—original draft, writing—review and editing; R.M.: methodology, writing—review and editing; S.S.U.-A.: project administration, writing—review and editing; T.U.A.: data curation, writing—review and editing; M.A.v.N.: conceptualization, writing—review and editing; M.F.: conceptualization, writing—review and editing; P.-C.B.: methodology, writing—review and editing; C.S.: conceptualization, data curation, funding acquisition, methodology, project administration, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

We thank the Max Planck Institute of Animal Behavior, the University of Zurich, the A.H. Schultz Foundation, the Leakey Foundation (Primate Research Fund and project grant), the SUAQ Foundation, the Volkswagen Stiftung (Freigeist fellowship to CS) and the Stiftung für Mensch und Tier (Freiburg i.Br.) for financial support. Open access funding was provided by the Max Planck Institute of Animal Behaviour.