The loud scratch: a newly identified gesture of Sumatran orangutan mothers in the wild

Abstract

The communicative function of primates' self-directed behaviours like scratching has gained increasing attention in recent years, but their intentional use is still debated. Here, we addressed this issue by exploring the communicative function of ‘loud scratches’ in wild Sumatran orangutans. Building on previous studies in chimpanzees, we examined the prediction that audio-visual loud scratches are used communicatively in mother–infant travel coordination. Specifically, we examined whether individual, social and scratch features affected the use of pre-move scratches, markers of intentional signal use and approach responses. We analysed a total of 1457 scratching bouts, produced by 17 individuals (including four mothers and their dependent offspring) observed during 305 h of focal follows. Overall, we found that scratching bouts preceded departure mainly when these were produced by mothers and showed features of exaggeration. If the scratching individual was a mother, associates were more likely to be visually attentive during pre-move scratches than in other contexts. Approach or follow responses to scratches by individuals in association were predicted by context, the relationship with the scratcher (i.e. offspring) and the associate's attentional state. We conclude that orangutan mothers use loud scratches as communicative strategies to coordinate joint travel with their infants.

1. Introduction

Self-directed behaviours in animals, including scratching, self-grooming and face-touching, are behaviours that seem not to fit the current context and were traditionally seen as by-products of physiological and psychological arousal [1]. In recent years, however, the communicative function of these behaviours, particularly scratching behaviour in primates, has received increasing consideration (e.g. [1–4]). Through learned or evolved associations between scratching and subsequent behaviour, an audience can gain valuable information about the scratcher's motivational state. Such sensitivity enables individuals to better navigate the social environment by mediating their interactions with others [1,2]. However, the extent to which this behaviour is used in goal-directed ways is still unresolved.

Great apes, our closest living relates, demonstrate the richest and most flexible gestural repertoires for social communication. Self-scratching behaviours in chimpanzees (Pan troglodytes) have been discussed repeatedly as serving a communicative function in various social contexts [3,5–7]. A recent systematic field study on mother–infant joint travel in two communities each of wild chimpanzees and bonobos reported that most mothers used the ‘loud scratch’ as an intentional gesture to initiate a carry—that is, these behaviours were accompanied by audience checking, goal persistence and/or sensitivity to attentional state [8,9]. Whether other primate species use self-scratching or other self-directed behaviours in their natural environments to achieve similar communicative goals has not yet been investigated.

Compared to the African great apes, the non-vocal communication of the arboreal orangutans (Pongo spp.) has to date been little studied in natural environments (but see [10,11–13]). Orangutans split from other hominids about 9–13 Ma [14], and their socio-ecology deviates substantially from that of the African apes: due to their semi-solitary nature, long juvenile periods and inter-birth intervals [15,16], it is presumed that mothers disproportionately affect their infants' behavioural development [12]. Dense vegetation in the canopy means that there are few opportunities for the direct lines-of-sight needed for visual communication. Nonetheless, in captivity, studies have demonstrated that orangutans possess a rich repertoire of tactile and visual gestures [17,18] and a facial mobility comparable to the African great apes [19]. In an experimental study on intentional communication, Cartmill & Byrne [20] have shown that orangutans adopt multimodal tactics to achieve communicative goals based on comprehension, by repeating signals if they are partially understood and switching sensory modalities if completely misunderstood.

The need to communicate of course also exists in wild orangutans, particularly among mother–offspring pairs. While infants emit vocalizations (e.g. cries and whimpers) when distressed, mothers and adults generally vocalize very infrequently (apart from the males’ long call; [21]). Resolution of conflicts between mothers and infants does not necessarily require vocalizations (which might also attract unwanted associates, such as males attempting forced matings or predators, such as tigers [22]). Mother–infant coordination is less relevant when the infant is still constantly clinging to the mother's body (i.e. during the first year of age; [23]), and once the infant is old enough to travel through the canopy independently (i.e. after 4–5 years of age; [24]). However, coordinating joint travel via ‘assistive bridges’ is still relevant for mothers of older infants that cannot yet cross tree gaps independently. Loud scratches, perceived audio-visually, are a prime candidate for circumventing the problem of line-of-sight access in the canopy, enabling efficient travel initiation and coordination among orangutan mother–infant dyads. Human observers can hear loud scratches at greater than or equal to 15 m in the noisy rain forest [25]. Although loud scratches have previously been described as part of chimpanzee and bonobo gestural repertoires used to coordinate mother–infant travel [5,8,9], no systematic study had yet addressed to what extent self-scratching behaviours exhibit features of signals and can thus be distinguished from regular self-maintenance scratches. A common problem in gestural research is that the form of the communicative act can resemble a regular action, and behaviours are then classified as gestures only if they meet the criterion of intentional use [26]. Therefore, we urgently need more studies assessing the communicative function of ‘mechanically effective’ behaviours such as those usually used for self-maintenance, by examining effects of context, form (e.g. intensity and exaggeration), intentional use and response of associates.

Here, we examine whether loud scratches serve to coordinate mother–infant travel in a natural population of Sumatran orangutans, as in chimpanzees [5,9,27]. Specifically, we tested the following predictions. First, mothers with spatially independent offspring should scratch more before moving away. Second, orangutans should use scratching intentionally to alter the behaviour of conspecifics, and we therefore tested whether scratching bouts were accompanied by key markers of intentional communication: audience checking (i.e. visual orientation towards the associate), attentional state of associate (i.e. visual orientation towards the scratcher), and persistence to the goal (i.e. serial scratching bouts) [20,28–30]. Third, we predicted that we would see mainly responses in the form of approaches in spatially dependent infants to their mother's pre-move scratches.

2. Material and methods

The study was conducted on wild Sumatran orangutans, Pongo abelii, at the Suaq Balimbing study site (3°42′ N, 97°260′ E, Gunung Leuser National park, Aceh Selatan, Indonesia) [31,32]. The 17 focal subjects consisted of four mothers and their dependent offspring (i.e. four infants and three juveniles), three adult males and three adolescent females; see the electronic supplementary material, table S1 for detailed information on subjects and datasets.

Data collection for this study took place between September 2013 and February 2014. We used an integrated focal sampling approach [33] and recorded the behaviour before, during and after scratching bouts with a digital high-definition camera, along with observer comments (continuous recording; [34]). During 305 h of full-day (nest-to-nest) and partial-day observation, we recorded about 41 h of video footage on the focal individuals.

We coded a total of 1457 high-quality recordings of clearly visible scratching bouts, with ‘bout’ defined as a sequence of scratches produced without pause. Coded variables included scratcher's identity, number of scratches within a scratch bout, duration of scratching bout, individual versus serial scratch bouts, body parts deployed for scratches (e.g. left and/or right hand) and body parts being scratched (i.e. head, limbs, body, switches), scratch range (two levels: large = large scratch to body and/or limbs, small = small scratch to head or face), visual orientation of scratchers and associates (i.e. potential recipients), subsequent behaviour within 10 s of scratch bout (e.g. feeding, moving, resting, other), as well as presence, identity and responses of associates. Infants were only considered as potential recipients if they were not on the mother, because the communicative use of scratches and body orientation would be futile. For each scratching bout, we then coded whether it was accompanied by visual orientation of scratcher towards audience (audience checking), visual orientation of associate relative to scratcher (recipient's attentional state), and whether it was a serial scratch bout (persistence to the goal) [28,30,35–37].

To investigate the sources of variation in the production of (1) pre-move scratches (i.e. scratches that were followed by moving behaviour), (2) markers of intentional use ((a) visual orientation of scratcher, (b) visual orientation of audience, (c) serial scratch bout) and (3) response by associates, we used generalized linear mixed models (GLMMs; [38]) with a binomial error structure and logit link function. Depending on the model, we included age-sex class of scratcher, scratch duration, scratch rate, scratch range (1, 2, 3), association with conspecifics (1), pre-move scratching (2, 3), kin relationship and visual orientation of candidate audience (3) as our key test predictors and fixed effects (table 1). To test for interdependence of the effects, we initially included the interaction terms between age class and scratching variables (1, 2), as well as between age class and pre-move scratching (2), which were subsequently excluded from further analyses if these were not significant. For more details on coding procedure as well as model specification and implementation, see the electronic supplementary material.

Table 1.

Effects on the proportion of scratching bouts (a) followed by moving behaviour, (b) associated with an attentive associate, and (c) followed by approach response by an associate. GLMMs were used with a binomial error structure and logit link function. Effects with p-values < 0.05 are depicted in italics.

	estimate	s.e.	${\chi }_1^2$	p-value
(1) pre-move scratch
intercept	−1.649	0.279	—	—
age-sex class [mother]	1.004	0.242	—	—
age-sex class [adult/m]	0.041	0.326	0.016	0.900
age-sex class [immature/f]	0.338	0.288	1.372	0.241
association [yes]	−0.548	0.197	7.828	0.005
scratch duration	0.41	0.118	—	—
scratch rate	−0.004	0.123	0.001	0.973
scratch range [large]	0.854	0.198	19.481	<0.001
mother × scratch duration	−0.398	0.192	4.516	0.034
(2) associate's attentional state
intercept	−5.193	1.219	—	—
pre-move scratch	0.343	0.891	—	—
age-sex class [mother]	5.093	1.369	—	—
age-sex class [adult/m]	−0.988	1.411	0.529	0.467
age-sex class [immature/f]	0.348	1.115	0.098	0.755
scratch duration	0.292	0.324	0.860	0.354
scratch rate	−0.276	0.246	0.622	0.430
scratch range [large]	−0.511	0.637	0.666	0.415
mother × pre-move scratch	3.552	2.486	5.140	0.023
(3) response
intercept	−9.259	2.127	—	—
pre-move scratch	2.365	0.82	9.559	0.002
relationship [offspring]	5.098	1.503	16.502	<0.001
associate's attentional state	2.846	0.791	11.413	0.001
scratch duration	−0.415	0.838	0.278	0.598
scratch rate	0.204	0.332	0.374	0.541
scratch range [large]	0.189	0.859	0.048	0.826

3. Results

First, we examined the sources of variation in the production of pre-move scratches. Overall, the full model fitted the data better than the null model, as the likelihood ratio test (LRT) revealed a significant effect of the key test predictors on pre-moving scratches (LRT: ${\chi }_8^2=52.173,$ p < 0.001, N = 1231). We found a significant interaction between the scratcher's age-sex class and scratch duration: while pre-move scratch bouts were significantly longer, they were more likely to be produced by mothers irrespective of duration (figure 1). Moreover, pre-move scratches were less likely if the scratcher was in association with (non-clinging) others, and more likely for large scratch range (table 1a).

Figure 1.

Effects of age-sex class and scratch bout duration on the production of pre-move scratches. Depicted are the raw proportions, separately for measured scratch bout durations (area of the dots corresponds to the sample size per measured duration; range = 1–259); the solid and dashed lines represent the fitted model and confidence intervals based on all other covariates centred to a mean of zero. (Online version in colour.)

For associates’ attentional state, we found that the full model fitted the data significantly better than the null model (LRT: ${\chi }_9^2=37.485,$ p < 0.001, N = 592). We found a significant interaction effect between age-sex class and post-scratch context (table 1b): potential recipients (i.e. offspring) visually oriented more towards mothers during the scratch bout if scratchers moved location after the scratching bout (figure 2). With regard to both scratcher's orientation and persistence, we found that the full model was not a significantly better fit as compared to the null model (LRT for scratcher's orientation towards associate: ${\chi }_9^2=10.858,$ p = 0.286, N = 715; serial scratch bout: ${\chi }_9^2=9.883,$ p = 0.360, N = 1181).

Figure 2.

Proportion of scratch bouts with associate visually oriented towards scratcher in relation to age-sex class and post-scratch behaviour (no. of individuals included: N_Mother = 4, N_other = 9). Indicated are individual means (small circles), median (horizontal lines), quartiles (boxes), and percentiles (2.5% and 97.5%, vertical lines).

With regard to associates' responses, the full model fitted the data significantly better than the null model (LRT: ${\chi }_6^2=32.012,$ p < 0.001, N = 440). We found that both relationship and attentional state had a significant effect on the outcome variable: responses were more common if the associate was the scratcher's offspring and visually oriented towards the scratcher (table 1c, see electronic supplementary material, figure S1). No other effects in the models were significant (table 1).

4. Discussion

We examined whether loud scratches in a population of wild Sumatran orangutans have a communicative function. We found that they are significantly tied to a particular context: mother–infant joint travel. Pre-moving scratch bouts were deployed more by mothers (adult females) than any other age-sex class, had a longer duration and a larger range than scratches in other contexts. We also found that adjustment to the attentional state of the associate (i.e. visually oriented to scratchers) was significantly more likely if scratching bouts were produced by orangutan mothers that were leaving a particular location, but found no evidence that audience checking or persistence in the form of serial repetitions of scratch bouts played a substantial role. Approach responses by associates were predicted by the relationship with the scratchers (‘dependent offspring’) and the associate's visual orientation, showing that loud scratches are at least sometimes used in goal-oriented ways. These are therefore not regular self-maintenance scratches, since they have signal features attached: they are produced in the context of moving, are longer and larger in range (i.e. ‘exaggerated’), mostly directed at an attentive audience and result in a response by dependent offspring.

As expected, loud scratches before moving were most commonly performed by mothers, who did so selectively in the presence of their offspring rather than conspecifics generally. This conclusion is in line with reports from three different populations (and two subspecies) of wild chimpanzees [5,9]. Interestingly, individuals of other age-sex classes do not seem to have the need to signal joint travel intentions when in associations with conspecifics. Thus, there is no carry-over effect from mother–infant context to adult relationships and no two-sided maintenance of association.

We have to consider that the intentionality markers developed for great apes were originally operationalized in captive chimpanzees, and might be too conservative and less applicable for arboreal orangutans living in visually obstructed habitat [26]. In these species, scratcher's visual orientation and serial use of scratch bouts might be poor indicators of intentional signal use. In the loud scratch, the auditory component of the signal complements the visual one, so audience checking is not necessary: if the aim is to indicate travel direction and onset, there might be no need to gaze at the audience. By contrast, considering the recipient's visual orientation is valid even for a signal that has a salient auditory component: to perceive it one still needs to orient in the direction of the auditory stimulus, perhaps to hear better or to acquire additional information. Orangutans generally use very few vocalizations in their dyadic communication, probably to avoid attracting the attention of other orangutans and predators. Loud scratches may thus serve as low urgency signals in the context of mother–offspring coordination.

In sum, the context-dependent and presumably intentional use of loud scratches by orangutan mothers and subsequent responses by infants suggests that pre-move scratches are used communicatively during travel initiation and coordination. Infant-directed scratches are different from scratches used in other contexts as they showed features of exaggeration and include markers of intentional use (i.e. audience adjustment as evinced by visual orientation towards the scratcher). Hence, we can rule out the possibility that loud scratches are merely self-directed behaviours indicating a behavioural change (mostly referred to as ‘cues’), with infants associating their mother's scratching with her subsequent moving behaviour. To better understand whether these gestures reflect evolved signals or culturally transmitted behaviours, it will be critical to investigate mother–offspring travel coordination in other orangutan populations and species. These findings will help to assess whether primates can invent meaningful ‘signals’ from cues that do not require any evolutionary changes in morphology.

Ethics

We were not required to complete an ethical assessment or to obtain ethical approval prior to conducting research. The research protocols were approved by the Ministry of Research and Rechnology (RISTEK; Research Permit no.: 152/SIP/FRP/SM/V/2012) and adhered to the legal requirements of Indonesia.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material.

Authors' contributions

M.F. designed the study, carried out the analyses and drafted the manuscript. K.L. conceived of the study, collected field data, participated in the analyses and commented on the manuscript. T.M.S. and C.S. helped to coordinate the project and commented on the manuscript. C.P.v.S. designed the study and helped to draft the paper. All authors gave final approval for publication and agree to be held accountable for the work performed therein.

Competing interests

We have no competing interests.

Funding

This study was funded by the A.H. Schultz Foundation and the University of Zurich. M.F. was supported by research fellowships of the Sponsorship Society of the German Primate Center and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) (grant no. FR 3986/1-1); and by the Forschungskredit Postdoc of the University of Zurich (grant no. FK-17-106).