An important developmental milestone in children is joint attention/engagement, a dyadic process in which preverbal individuals begin to respond to, and initiate, nonverbal bids of communication via the use of gaze, gesture, and vocalizations (Adamson, 1996). For example, around 4–6 months of age, typically developing children begin to respond by orienting to the communicative actions and bids by humans (usually parents and caretakers), such as following gaze cues and/or pointing responses (Tomasello, 1995; Adamson, 1996), otherwise known as receptive joint attention (RJA). Between 9 and 12 months of age, typically developing children not only respond to communicative cues, but also begin initiating them, including directing communication toward social partners, synchronizing their behavior with these partners, and responding to the communicative behaviors of others (Tomasello, 1995; Adamson, 1996; Carpenter et al., 1998). Most early communication is nonverbal (e.g., facial expressions, gaze, gestures). Typically developing children usually learn that pointing and gaze direction refer to an external referent by 12–15 months of age, and they point at and direct their attention to the referent (i.e., initiation of joint attention behavior, IJA; Carpenter et al., 1998). Many have suggested that, given the universal expression of RJA and IJA by human infants in a variety of cultures, comprehension and initiation of joint attention behaviors are the foundation for subsequent language and speech development. For example, in developmentally-typical children, the rate of language development is predicted by the age of onset of joint attention behaviors (Bates et al., 1975, 1987; Carpenter et al., 1998; Nichols et al., 2005; Whalen et al., 2006; Mundy et al., 2007). The importance of joint attention for typical socio-communicative development has been clearly established in clinical studies of children at risk for or with a diagnosis of Autism Spectrum Disorder (ASD). There is significant overlap between social and communication impairment, particularly for pre-verbal children with ASD (Landa et al., 2007). Young children at risk for ASD or diagnosed with ASD display impaired IJA and RJA abilities (Dawson et al., 2002; Presmanes et al., 2007; Ibanez et al., 2012). The neurobiological mechanisms underlying ASD are still unknown and likely involve multiple systems (Cody et al., 2002; Lotspeich et al., 2004; Herbert et al., 2005; DiCicco-Bloom et al., 2006; Hadjikhani et al., 2006; Hazlett et al., 2006; Mosconi et al., 2006; Girgis et al., 2007; Minshew and Williams, 2007; Amaral et al., 2008; Ha et al., 2015; Mundy, 2017; Pan et al., 2021); however, the potential consequences of atypical communicative and social behavior on language development and the brain are likely significant.

Some researchers have made a distinction in joint attention between imperative and declarative communication, particularly with regard to pointing (Liszkowski et al., 2004; Tomasello, 2008). Imperative pointing, or initiating behavioral requests, reflects an individual’s request for an object or action from another social agent and is therefore instrumental in function. For example, a child pointing at a cookie as a request for another person to give them the cookie. In contrast, declarative pointing is allegedly motivated by the desire to simply share information with another individual with no explicit expectation of a response. For example, a child pointing to an airplane in the sky in order to share that information with another person. Similarly, with respect to comprehension, the distinction has been made between responding to a behavioral request or to a declarative signal. Declarative communication, and particularly pointing, likely reflects the initiation of shared joint attention, whereas initiating and responding to behavioral requests reflects different underlying cognitive, neurological, and motivational processes (Baron-Cohen, 1995; Mundy et al., 2007; Cochet and Vauclair, 2010a; Cochet and Vauclair, 2010b; Leavens, 2012; Mundy, 2017). Moreover, and critically, it has been hypothesized that declarative, rather than imperative pointing, is the more important and critical cognitive and linguistic dimension of joint attention in predicting the subsequent development of language and speech (Iverson and Goldin-Meadow, 2005; Tomasello, 2008; Liszkowski et al., 2009), as well as distinguishing children at risk for ASD (Camaioni et al., 2004), though empirical support for this hypothesis is tentative at best (Ibanez et al., 2012). Indeed, direct studies that have longitudinally assessed both imperative and declarative signaling as predictors of subsequent language development in the same subjects are somewhat limited in the literature and the findings are not entirely consistent (Colonessi et al., 2010).

The results of this study can be summarized as follows: First, when considered as a collective group, chimpanzees scored significantly higher than chance on the OC task; however, when evaluating individual performance based on binomial z-scores, there was no significant difference in the number of OC+ (passing) and OC− (failing). Second, OC task performance was not significantly heritable nor were there any significant effects of sex, rearing history, or colony. Lastly, significant differences in gray matter covariation were found between chimpanzees that passed or failed the OC task.

Many have reported that chimpanzees fail tasks that require responding to, or the production of, declarative signals, particularly for the OC task (Mulcahy and Hedge, 2012; Clark et al., 2019). That is to say, chimpanzees, as a species, are allegedly poor at or incapable of comprehending or producing declarative communicative signals. Consistent with the meta-analysis reported by Clark et al. (2019), the results reported here challenge this claim; however, this inference depends, in part, on the interpretation of the group compared to individual results. Notably, consistent with nearly every other report in the literature on OC task performance (Clark et al., 2019), though the average performance of the chimpanzees was significantly better than chance based on the one sample t-test, there was no statistical difference between the number of OC+ and OC− chimpanzees. In our view, the dissociation in interpretation of performance based on the group and individual levels of analysis is not a trivial issue.

There are numerous studies in the literature on animal social cognition that utilize group averages as a basis for inferring either (Adamson, 1996) species differences in performance or (Tomasello, 1995) group or species level capacities for a given psychological trait (Karg et al., 2015; Joly et al., 2017). As argued elsewhere (Hopkins and Sherwood, 2022), one problem in using group average data to evaluate species-specific capacities or species differences in performance is that, within a given sample of subjects, it is feasible that the group as a whole can perform significantly better than chance, while no individual may perform the task significantly better than chance. As an example, 8 chimpanzees could be administered 24 trials on the OC task and perform at 50%, 52%, 53%, 55%, 55%, 58%, 62%, and 67% correct. A one-sample t-test on these data reveals that the chimpanzees, as a group, perform statistically better than chance t(7) = 3.265, p = 0.014 while not a single individual ape performed better than chance based on their binomial z-scores. This is particularly true in studies where the subjects are administered very few trials, which can often be the case in experiments of cognition in human and nonhuman primates (Buttelmann et al., 2007; Herrmann et al., 2007, 2010; Russell et al., 2011; Schmitt et al., 2011; Karg et al., 2015; Rosati et al., 2016; Krupenye and Hare, 2018; Bettle and Rosati, 2019). In short, we believe that, in addition to group averages, there is also a need to report individual performance data in studies of animal cognition (including humans; Herrmann et al., 2006). This would increase the rigor of these studies and likely improve their repeatability and replication across laboratories and settings.

The findings reported here also failed to find any significant sex or rearing history effects on OC task performance. Further, OC task performance was not significantly heritable. These findings differ from previous reports on OC task performance among apes in two ways. First, regarding rearing history, Lyn et al. (2010) and Call and Tomasello (1994) reported that human enculturated apes perform significantly better than age-sex matched controls who were not human enculturated. Recall that a portion of our chimpanzees were raised by humans in a nursery setting for the first 3 years of life which some might suggests would constitute human enculturation. However, the human enculturated apes tested by Benga (2005)) and Call and Tomasello (1994) were individuals with extensive, life-long rearing histories, including training on alternative communication systems with humans which differs substantially from the NR chimpanzees in this report. Therefore, it appears that rearing of chimpanzees by humans in a nursery setting does not provide the level of experience or stimulation to enhance their OC task performance over apes raised by conspecifics. Second, performance on the OC task has been reported to be significantly heritable in canines (largely retriever breeds; Bray et al., 2021), a finding that differs from our results in chimpanzees. Because canines have been domesticated and the subjects were purebred animals, one interpretation of this difference may be that the selective breeding for traits strongly expressed in retriever canines facilitated the processing of human communicative cues in some canine species. This explanation is certainly possible but performance on the OC task varies substantially among different breeds of canines, as well as between pet dogs compared to individuals living in shelters or other environments (Udell et al., 2010; Clark et al., 2019). Thus, it is possible that selective breeding for certain traits manifest in retriever canine breeds may account for the reported heritability but, in the absence of data from non-retriever breeds, it is not possible to attribute these results to a broader genetic effect attributable to domestication.

Regarding the SBM results, OC+ chimpanzees contributed more to the creation of two gray matter components that included a number of brain regions within the social brain network including the amygdala, hippocampus, insula and middle and posterior superior temporal gyrus (Adolphs, 2009; Kennedy and Adolphs, 2013; see Figure 1). These findings are consistent with the Mundy and Newell hypothesis regarding the brain regions implicated in the posterior attentional network and corroborate previous findings in chimpanzees of an association between gray matter volume in the superior temporal gyrus and a different measure of RJA. The findings are also consistent with hypotheses regarding the role of the middle and superior temporal gyrus and sulcus in the processing and interpretation of social stimuli (Redcay, 2008; Redcay et al., 2012). OC+ chimpanzees contributed less to component 3 compared to the OC− apes. Interestingly, the largest brain region within component 3 was the rostral and caudal anterior cingulate, a brain region hypothesized to be part of the anterior attention network and hypothesized to be linked to the initiation of imperative and declarative signals (Benga, 2005; Mundy and Newell, 2007). Previous studies have reported significant associations between gray matter volume within the anterior cingulate in chimpanzees(Hopkins and Taglialatela, 2012), suggesting that there is some dissociation in the brain regions associated with initiation of and responding to RJA cues.

The shared intentionality hypothesis proposes that increased selection for cooperative and prosocial behavior led to the emergence of the ability for shared intentions between interlocutors, and manifests in a variety of advanced, human-specific cognitive abilities, such as triadic joint attention, declarative communication, imitative learning and teaching (Tomasello, 1999). It has been correctly recognized that there is some evidence that apes raised by humans in complex socio-linguistic environments do engage in triadic joint attention and declarative pointing (Leavens and Bard, 2011). Indeed, previous research has shown that bonobos and chimpanzees that have been raised by humans in a socio-communicative linguistic environment, much like human infants, perform significantly better on the OC task than apes raised in standard captive settings by humans or conspecifics (Lyn et al., 2010). Similarly, among language-trained apes that communicate with humans via visual-graphic symbols, between 10% and 15% of their keyboard utterances have been characterized as declarative in function (Lyn et al., 2011). Though this percentage was much lower than data from 15-month-old human children, it does suggest that the differences in declarative pointing ability between humans and apes are not absolute, but rather quantitative and potentially malleable, as a function of the early social rearing experience of the subjects.

Importantly, as pointed out by Call (2009), evidence of declarative pointing in language-trained apes begs the question of why non-language-trained apes do not use these presumed innate abilities in the wild or captivity? We would offer two possible explanations. First, evidence from this study suggests that performance on the OC task is not heritable; thus, the assumption that this is an inherent, innate ability that is presumably under some genetic control is not supported in this study, and, indeed, we know of no studies that have explicitly demonstrated a genetic basis for joint attention in humans. It may be the case that social learning or other experiential factors explain the development of RJA in humans and that a similar explanation might elucidate the performance of the enculturated apes compared to standard mother- or nursery-reared individuals; i.e., non-language-trained apes likely experience less responsive caregiving environments than language-trained and other enculturated apes(Bard et al., 2014). That is to say, attempts to compare declarative communication abilities between human children and nonhuman primates is exceedingly difficult without controlling for the extent of reinforcement history and learning experiences that apply to each species’ experiences prior to testing.

Second, the capacity for declarative communication may be present in chimpanzees and perhaps, many other primate species, but it manifests itself differently than the communicative behaviors of typically developing preverbal children. For instance, recent studies in wild chimpanzees have shown that they will produce alarm calls in the presence of a predator (harmful snake) more frequently when the chimpanzees accompanying the focal subjects are unaware or ignorant of the presence of the snake (Crockford et al., 2012; Schel et al., 2013). Chimpanzees are less likely to produce the alarm calls when the accompanying chimpanzees are aware of the presence of the snake. Unlike the production of alarm calls in response to predators in other nonhuman primates, such as the well-known case of vervet monkeys (Seyfarth et al., 1980; Seyfarth and Cheney, 1993; Seyfarth and Cheney, 2010), the chimpanzees appear to be “informing” other group members of the presence of the snake, rather than simply responding fearfully, to which other group members respond (Seyfarth and Cheney, 2010). By definition, this constitutes a form of declarative communication; however, this is a vocal signal, not a manual pointing response, which is the primary behavioral measure investigators have used to characterize declarative communication in other studies of apes and children. Similarly, in some studies with children, the subjects are presented with novel objects in the presence of a caretaker and the researchers record the amount of time the child spends showing the object or the number of alternations in gaze the child makes between the object and the caregiver. This type of communicative behavior, which primarily manifests in gaze behavior, is considered declarative in function by some (Mundy et al., 2007) and defined as triadic joint attention by others (Carpenter et al., 1995). In either case, the overt behaviors measured are not a manual pointing response, which is again, the primary basis by which declarative communication has been measured in nonhuman apes. As Bates and her colleagues observed (Bates et al., 1975) in their paper introducing what they called the protodeclarative aspects of infant preverbal communication, pointing is simply one of a number of such communicative acts, including exhibition of self, showing of objects to others, and giving objects to others. All of these declarative behaviors are well-described in the communicative interactions of chimpanzees (Plooij, 1978; Zamma, 2002; Silk et al., 2013). In our view, an arguably less narrow, species-centric approach to defining declarative communication may provide a more valuable conceptual and methodological framework for comparative tests of the shared intentionality hypothesis.

This study is not without limitations. First, though we statistically controlled for the age of the chimpanzees when they were tested on the OC task and when they were scanned, the collection of these two data sets was spread over multiple years and therefore were not necessarily obtained in close temporal proximity to one another. Indeed, though all but one chimpanzee was tested on the OC task after the collection of the sMRI scan, the average time between behavioral testing and scanning was 5.37 years with no significant difference between OC+ M = 5.27 SD = 4.00 and OC− M = 5.45 SD = 4.29, t(116) = −0.235 p = 0.814. In light of the fact that there are age-related changes in gray matter covariation among chimpanzees (Hopkins et al., 2019; Mulholland et al., 2021), ideally the behavioral and sMRI data would have been collected at approximately the same time. Unfortunately, this is not possible, as MRI scans can no longer be acquired from chimpanzees for research purposes. Second, most if not all of the chimpanzees in this study have been subjects in previous studies on joint attention and related tasks assessing social cognition (Whiten et al., 2007; Hopper et al., 2008; Taglialatela et al., 2012; Hopkins et al., 2014; Brosnan et al., 2015; Leavens et al., 2015; Hopkins et al., 2020). It is possible that testing experiences on these previous tasks may have benefitted or hindered performance on the OC task and we cannot rule this out as a source of some of the individual variation in performance.

In summary, consistent with previous reports, the overall results demonstrate that chimpanzees comprehend declarative communication signals as measured by the object choice task, though there is considerable individual variability in performance with a significant number of apes performing below chance. OC task performance was not significantly heritable nor were there any significant effects of sex, colony, or rearing history. OC task performance was associated with gray matter covariation in the middle and superior temporal cortex, and a number of other regions, a finding consistent with the hypothesized role of these specific brain areas in RJA, and more broadly, social cognition.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by the Animal Care and Use Committee of Emory University and MD Anderson Cancer Center.

WH designed the work, analyzed and interpreted the data, and wrote the initial drafts. MMu analyzed the data and wrote the initial drafts. SS, MMa, and SW acquired data. All authors contributed to the article and approved the submitted version.

This work was supported, in part, by NIH grants AG-067419, NS-073134, NS-042867, and NS-092988. Chimpanzee maintenance at the National Center for Chimpanzee Care was previously funded by NIH/NCRR U42-OD-011197.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

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.