The calls of the wild

When I tell people that I study vocal communication in chimpanzees and that I work in a psychology department I am often greeted with surprised faces. I go on to explain thatI am interested in language and trying to trace the evolutionary history of this remarkable human capacity. Language is one of the most intricate and complex behaviours, and is one of the few that clearly distinguishes humans from the rest of the living world.

Finding an evolutionary explanation for the origins of human language is, however, an extremely difficult challenge. Spoken language has left no fossil remains and very few clues as to when and how it emerged. Recent genetic evidence suggests that our hominid ancestors did not have modern speech abilities until approximately 200,000 years ago, when the two key mutations in the FOXP2 gene are thought to have stabilised in the human population (Enard et al., 2002). In addition, archaeological records indicate that representational artefacts, which require symbolic thought, only date back 50,000 years (Deacon, 1997).

These strands of evidence suggests that humans had an exceedingly short time in which to evolve such a complex capacity as modern spoken language. One powerful premise is that many of the underlying cognitive capacities involved in language processing are much older than language itself, with their phylogenetic roots deep in the primate lineage (Hauser et al., 2002). Thus, a promising empirical approach to understanding the origins of human behaviour, including language, has been to examine the capacities of extant primates, whose phylogenetic relationships to modern humans are known (Fitch, 2000; McGrew, 1991).

This comparative approach is vital for identifying the elements of language that appear to have evolved gradually from a common primate ancestor and those that have no clear evolutionary path, which may be the 'novel' elements that caused human language to evolve into its current uniquely complex state (Hauser & Fitch, 2003). Chimpanzees are the closest living approximation to the common human–chimpanzee ancestor (McGrew, 1991), and as such are one of the most informative models for the comparative approach (Lieberman, 2000). Thus by studying vocal communication in chimpanzees I aim to identify prerequisites and elements of language that are shared between humans and chimpanzees, and those that are not.

The global aim of my PhD was to assess whether chimpanzees were capable of communicating about events and objects in the external world. This type of communication has been found in a number of monkey species – most famously in the vervet monkey. Seyfarth and colleagues in 1980 showed that these monkeys give different calls for different predators and when the alarm calls are played back in the absence of the predator, listeners respond as if they have seen the predator themselves. This led to the conclusion that these calls were functioning to refer to different types of predator, in a roughly analogous way to how humans label events with words.

Surprisingly, there was no evidence available that any of the great ape species was capable of such 'functionally referential' communication. This had caused some to doubt the significance of the monkey evidence for evolutionary accounts of language, claiming we may be looking at a case of convergent evolution, rather than a precursor of human abilities. I sought to address this issue by studying vocal communication in wild and captive chimpanzees.

I identified two situations where it may be advantageous for chimpanzees to communicate with others about external events: food discovery and agonistic interactions. I then systematically investigated grunts given in response to food and screams given during agonistic encounters. I conducted observational work with habituated individuals from the wild population of chimpanzees in the Budongo Forest in Uganda (Reynolds, 2005),  and captive individuals at Edinburgh Zoo.

Rough grunts

Chimpanzees often produce calls when encountering food and these calls are referred to as 'rough grunts'. I investigated whether rough grunts contained information about the nature of the discovered food source with the captive chimpanzees housed at Edinburgh Zoo.

First, we constructed a food-preference hierarchy by providing the chimpanzees with a choice of two food types and recording their responses. This way we were able to objectively rank nine common foods in order of preference. We grouped the nine food types into three different value groups (high, medium and low) and then recorded rough grunts produced upon discovering the different foods. We first found that chimpanzee rough grunts were acoustically different according to the value of food encountered (Slocombe & Zuberbühler, 2006). High-value food elicited long, high-pitched grunts which graded into the short, low-pitched, noisy grunts elicited by low-value food. Chimpanzees thus reliably produced grunts that reflected the value of the food they were eating.

We then examined whether chimpanzee calls could have a greater degree of referential specificity by acting as unique labels for specific food types. We needed to control for the effect their preference had on the acoustic structure of the rough grunts. Therefore we examined whether the grunts given to the three food types that made up each value class differed in acoustic structure. We found no evidence for this in low- and medium-value foods, but within the high-value category we found the grunts given to bread, mango and banana were acoustically distinct. We found that the grunts given to bread and bananas remained stable across different feeding events. This suggested they could function as labels for individual food types.

We then examined rough grunts from the wild chimpanzees of the Budongo Forest, given in response to three medium-high preference foods. The acoustic structure of these grunts was comparable to the grunts given by captive chimpanzees to high- or medium-preference foods, giving the captive results a degree of ecological validity. We were, however, unable to replicate the finding that different food types elicited acoustically distinct grunts. We believe the most parsimonious explanation for these results is that labelling of specific food types is likely to be a by-product of a preference driven system, within the special conditions of captivity. In captivity the quality, quantity and novelty of the different food types were relatively consistent, and food type may thus become the main determinant of food value in captive chimpanzees. (Type is just one of many factors influencing food value in the wild – size of tree, number of fruits, ripeness of fruit, seasonal availability are also important – and therefore it is unlikely we will see labelling of food types in this environment. We are currently analysing a larger dataset to test this hypothesis.) In summary, chimpanzees produce rough grunts that reference the value of a discovered food source, but under certain circumstances this system may result in specific foods eliciting distinct rough grunts (Slocombe & Zuberbühler, 2006).

In order for a call to be classified as a functionally referential signal, it is necessary to show that the listeners understand the meaning of the call and its environmental referent. We therefore conducted a playback study to examinea listener's response to rough grunt calls, to see if they meaningful refer to food value (Slocombe & Zuberbühler, 2005b).This study was also conducted at Edinburgh Zoo, where we first established two artificial trees in their enclosure. These 'trees' regularly dropped tubes, filled either with bread (high value) or apples (low value), into the study group's outdoor enclosure. To ensure the chimpanzees did not monopolise the preferred bread tree, we introduced the contingency that only one tree ever fruited successfully, with the other tree producing empty tubes.

Once the chimpanzees became accustomed to these rules we started the playback trials. Before any playback trial started we waited until all individuals were in the indoor enclosure. Whilst they were absent we dropped empty tubes from both trees to simulate a successful fruiting event by one of the trees. As soon as the first individual emerged from the inside enclosure we played back rough grunts, originally given to either bread or apples, from a speaker that was positioned equidistant from the two trees. We then filmed the focal animal's response to monitor its foraging strategy. We were only able to collect systematic data from one individual, a six year-old male, named Liberius, who was usually the first individual that came out into the outside enclosure.

In the first few trials Liberius consistently visited the 'correct tree' first, that is, the tree indicated by the rough grunts. However, soon thereafter he then adopted an idiosyncratic foraging strategy, during which he consistently searched his favourite bread tree first, regardless of the types of rough grunts used as playback stimuli. We therefore analysed subtler behavioural measures collected during all trials, to examine whether he was processing the rough grunts in a meaningful manner. We hypothesised that if Liberius had an expectation about where to find food based on the grunts he had heard, he should put different amounts of effort into searching the two locations. We found that when Liberius heard 'apple grunts' he spent longer searching and he tended to search more tubes under the correct apple tree than the incorrect bread tree. In contrast, when he heard 'bread grunts' he searched more tubes and tended to search for longer under the correct bread tree than the incorrect apple treeIn control trials when Liberius heard no grunts, he showed no such biases in search effort. We concluded that this chimpanzee understood that his group members' grunts referred to foods of different value and used this information to aid his own search for food. This was the first experimental evidence that any ape species used their vocalisations in a referential manner (Slocombe & Zuberbühler, 2005b).

Agonistic screams

Chimpanzees often scream when they are involved in agonistic encounters. In the Budongo Forest in Uganda, I recorded screams with sound equipment along with a detailed record of the accompanying context. I then examined whether chimpanzees were producing acoustically distinct call variants in different contexts.

I first assessed whether the social role of the caller in a fight was encoded in the acoustic structure of the screams they produced. We compared the acoustic structure of screams produced by 14 individuals, in the roles of both victim and aggressor, and found subtle but consistent acoustic differences (Slocombe & Zuberbühler, 2005a). Having found reliable differences in the production of these calls, we needed to test if listening individuals were able to extract meaningful social information about the role of the caller from these screams. Playback experiments designed to test this premise will be completed this summer.

I then went on to consider victim screams in greater detail. We found that victim screams varied in their acoustic structure as a function of the severity of he aggression they were receiving (Slocombe & Zuberbühler, 2007). Victims receiving severe aggression gave longer bouts of screams in which each call was longer in duration and higher in frequency than screams produced by victims of mild aggression. In collaboration with Simon Townsend, I have just completed playback experiments that indicate that listening individuals can meaningfully distinguish between mild and severe victim screams (Slocombe, Townsend & Zuberbühler, in prep). This suggests that individuals out of sight of a fight can infer the nature of the fight simply by listening to the screams.

Thus it appears that a number of levels of social information about the nature of a fight and the role of the caller are encoded in the screams of chimpanzees. As playbacks have confirmed that listening chimpanzees can extract this information about ongoing events, these calls can be classed as functionally referential calls.

We also encountered another extremely interesting phenomenon when analysing the victim screams. We found that victims receiving severe aggression were sensitive to the composition of the listening audience and they modified the acoustic structure of the screams accordingly. If there was an individual equal or higher ranking than the aggressor (an individual who could effectively challenge the aggressor) present in the vicinity, then the victim produced screams that were acoustically consistent with extremely severe aggression. This vocal exaggeration of the level of aggression only occurred when the chimpanzees most needed aid; when they encountered severe aggression, not mild aggression. It seems that victim screams thus function to recruit aid (Slocombe & Zuberbühler, 2007). This raises the possibility that chimpanzees are capable of intentional deception (exaggerating level of aggression to alter the high ranking chimpanzee's understanding of the event and thus increase the chances of receiving aid); however, we need to collect more data to exclude other more low-level explanations for this behaviour, before this possibility can be seriously considered.

Conclusions

My research into chimpanzee rough grunts succeeded in providing the first evidence of functionally referential communication in any ape species. My work on chimpanzee screams has identified this call type as having a number of potentially meaningful variants and therefore a promising candidate for also acting as a referential signal. In addition, we identified an interesting degree of control and flexibility in the scream producer as well as highlighting the complex social cognition of these animals.

In terms of the evolution of language debate, these findings address the previous chasm between monkey and human evidence for referential abilities. I am still actively researching this topic and I believe with more research effort the chimpanzee vocal system will shed more light on the elements of human language and its cognitive underpinnings that are phylogenetically old, and those that are unique to humans.