Referential communication is defined as the interaction between signaler and receiver using referential signals that denote information about objects or events in the external environment. In humans, this typically corresponds to a mental representation in both parties and has been highlighted as an important element in the evolution of human language (Fitch 2010). In nonhuman animals (thereafter referred to as animals), the term of referential communication is used in a different way, as evidence for the relevant psychological states of mental representations is lacking, though the behavioral outcome appears to be comparable (Marler et al. 1992). To emphasize that while animal signals may function referential, but do not underlie the cognitive processes as in human communication, these signals have been called functionally referential (Macedonia and Evans 1993).
The prevailing text book example for functionally referential signals in animals are the different alarm call types used by vervet monkeys (Chlorocebus aethiops) that were described to show a distinct relationship to the appearance of specific types of predator types (Struhsaker 1967). Vervet monkeys produce structurally different alarm call types to terrestrial predators such as leopards than to eagles in the air or snakes. Playbacks in the natural habitat of the monkeys were used to demonstrate that these different alarm call types elicit different predator-specific escape responses as if the real predator was approaching (Seyfarth et al. 1980a, b). The playback of a terrestrial alarm call results in receivers running into the tree, while an aerial alarm call drives them either into the next bush or the center of the tree if already sitting up, and the snake alarm call induces them to stand up bipedally and monitor the ground next to them. These findings of the production of predator-specific alarm calls have been confirmed in several other species, including nonprimate mammal and bird species (for recent reviews on mammals, see Townsend and Manser 2013, on birds Gill and Bierema 2013; Smith 2017). These studies all indicate that the structural variation between these alarm call types seem meaningful, as they are emitted in highly specific contexts to a certain predator type and enable the receivers to show the appropriate predator-specific escape response.
Highly context-specific signaling referring to external events have also been described in the context of animals encountering food (Clay et al. 2012). The first functionally referential signal described is the honeybee waggle dance (von Frisch 1974), which refers to the direction and in some species also the distance to the foraging patch. Chimpanzees (Pan troglodytes) (Slocombe and Zuberbühler 2006) and bonobos (Pan paniscus), as well as rhesus macaques (Macaca mulatta) (Hauser and Marler 1993), capuchin monkeys (Cebus capucinus) (Gros-Louis 2002) produce calls when finding food that are acoustically distinct depending on food quality or desirability. The description of the specific context showed that the likelihood to produce one food call type and not another in apes also varied between groups (Clay et al. 2012), and in capuchin monkeys, it depends very much on which individuals are close by as potential receivers (Gros-Luis 2002). Food calls are also described for ravens (Corvus corax), where they increased in their number of “haa” calls depending on their food preferences (Bugnyar et al. 2001). However, whether the receivers relate the variation in vocalizations to the signaler’s perceived food quality has not been experimentally shown, although based on natural observations, these calls recruit other ravens to the location of the calling individual (Heinrich and Marzluff 1991).
Defining Functional Referentiality
The model put forward by Macedonia and Evans (1993) defining functionally referential communication requires a high production specificity and a high perception specificity of the different alarm call types within a species. Call types can only be regarded as functionally referential, if two specific conditions are met: the call type must be produced in close association to a specific external event or object and the receivers must respond to the call type as if the stimuli was present. However, this does not imply what aspect of the external stimuli is responsible to induce the specific call type. For example, studies on referential alarm calls suggest that in some species, call types are associated with specific predator species (Gunnision’s prairie dogs (Cynomys gunnisoni) Slobodchikoff and Placer 2006), while in others they seem to refer to behavior of the predator (Siberian jay (Griesser 2008), and in most cases, it is not clear whether it is the spatial location (air versus ground) or aerial versus terrestrial approach (fast versus slow) the call refers to (e.g., meerkats, Suricata suricatta) (Manser 2001).
Recent studies provide evidence for the combination of meaningful, highly specific call types put into a sequence, whereby the combinations of the calls carry the meaning of the single call types. Thereby different types of sequences have been described depending on the combination of the two call types (recent review Collier et al. 2014). A first type is the use of an affixation system by Campbell monkeys (Cercopithecus campbelli campbelli) (Ouattara et al. 2009). They add an affix to both of their predator-type specific calls to produce new calls with a different meaning. The terrestrial alarm call with the additional suffix is given to any disturbance, while the aerial alarm call with the affix is given to any disturbance in the canopy. It seems the addition of the affix modifies the meaning of the predator type specific calls into a less specific disturbance call in the according physical space indicated by the specificity of the predator type. The second type of meaningful sequences is the use of two predator-type specific calls, terrestrial and aerial calls, in combination, as described for putty-nosed monkeys (Cercopithecus nictitans) (Arnold and Zuberbühler 2006). The combination of the calls elicit movement of the calls, rather than the predator-type specific response given to each of the call types themselves. Different to the Campbell monkey combinations, here the meaning of the sequence seems not to derive from the meaning of the components. Recent studies on call sequences in animals have described the combinations of two meaningful call types; however, these are typically not based on functionally referential calls that seem to refer to an external event or object. Many call combinations have been described in the context of mobbing predators to recruit other group members to the caller (Suzuki et al. 2016), or in social context where calls are related to the behavior or the general context the caller experiences (Collier et al. 2014).
The Level of Specificity on the Production and Perception Side
A recurrent pattern described is that typically the alarm calls to aerial predators are more specific to a narrow range of predators, and therefore more functionally referential, than alarm calls to terrestrial predators. For example, in lemurs (Eulemur fulvus), a distinct aerial call type refers specifically to hawks while terrestrial alarm calls given in response to fossa, dog, and baboon calls are also given to other disturbances (Fitchel and Kappeler 2002). A similar pattern is found in blue monkeys (Cercopithecus mitis) who emit pyows to an even broader array of stimuli including dog, snakes, motorcycle, civet, and baboons (Fuller 2014). There are at least three possible explanations for this recurring pattern. First, aerial alarm calls are typically of high urgency and require immediate responses. Receivers cannot afford the time to attend to any contextual information before responding. In contrast, terrestrial predators typically approach at slower speeds which gives receivers more opportunities to gather additional contextual information and then respond. Second, animals may categorize the threat differently than us humans, and therefore we perceive the stimuli they respond to on a variation of terrestrial stimuli as broad, versus raptors to approach as a very clear defined narrow category. None of the studies so far published on functionally referential calls has thoroughly addressed this question, as it is a huge effort to determine how a species categorizes its environment (Townsend and Manser 2013; Zuberbühler and Neuman 2017).
We also find variation in the perception specificity of the receiver’s response to the functional referential signals. Recent studies on different vervet monkey populations have shown that the response of receivers to the originally described functionally referential call types elicit qualitatively different responses for the same predator type call. Ducheminsky et al. (2014) showed in response to playbacks of alarm calls that receivers often looked first towards the speaker and either did not show escape behavior or delayed. This variation in receiver response across populations could indicate a lack of perception specificity, or it may reflect the fact that we find variation in the recipients depending on their own situation. If they are close to shelter, they may not have to run immediately but can take additional cues into account for their appropriate response. Diana monkeys respond to guinea fowl alarm calls with leopard alarm calls themselves, as if a leopard were present, but they remain silent when they seem to have a situation indicating that the guinea fowls calls were given to chimpanzee or humans, potentially hunting for the monkeys (Zuberbühler 2000). This has also been demonstrated in dwarf mongoose, where receivers take their own situation into account whether to respond and how fast and intense (Kern et al. 2017).
Limitations of the Theoretical Framework
This variation in production and perception has recently brought up some criticism on the concept of functional referentiality and its current definition (Wheeler and Fischer 2012). The basic argument is that any call type, whether highly context specific to the external environment of the signaler or referring to the signaler’s behavioral state, conveys some referential information to the receiver. From this argument, the observed signals should be seen along a continuum rather than be divided into discrete categories, functionally referential versus other call types, in the past usually referred to as motivational signals expressing the behavioral context of the caller. Yet, the theoretical framework has hold up and seems to be helpful in stimulating new research (Scarantino and Clay 2015). Also, besides the specific information to the external event, it is likely any vocalization shows some variation relating to traits of the caller, such as indexical information or the intrinsic (motivational) state of the caller (Manser et al. 2002). One could then argue we should go along a multidimensional characterization of the signal structure, where referentiality could be identified for each of these different types of traits. For example, the individual signature conveying unique information on the identity of the caller, the expression of the motivational state reflecting the behavior of the caller, etc.… (Manser 2016). The obvious difference here is that it is not referring to a stimuli in the external environment of the caller.
Selective Pressures on Evolving Functionally Referential Signals
The evolution of referential signals in animals seems to vary depending on the context, whereby in all cases the need to provide accurate information on the external stimuli to enable the receiver to make the appropriate decision how to respond seems the driving factor. This is very obvious in the situation of the alarm call types in vervet monkeys and some other species, where different escape responses to the different predator types in a three-dimensional space are of benefit (Macedonia and Evans 1993). However, e.g., in meerkats, this seems not the full explanation, as the ground squirrel living in the same habitat do not use predator type specific calls. Rather, the argument has been that social aspects to keep cohesion with the group may explain the need for functional referential calls (Furrer and Manser 2009). Even less clear are the selective pressures underlying food calls (Clay et al. 2012). While in alarm calls, the urgency aspect which does not allow to monitor for additional contextual cue due to time constraints seems logical, this does not apply on food calls. Here, rather social reasons, in particular the social structure of the group has been put forward as explanation, where finding high quality food and share it with group members may be linked to social rank and enhance the establishment of the reputation of the caller.
Nonvocal Functionally Referential Signals
Referential signals in animals other than vocalizations include visual signals, such as the bee dance or gestures described for primates, birds, and potentially also fish (Smith 2017). Most research has been done on the acoustic modality, likely due to the difficulties in identifying and manipulating visual signals, and convincingly test whether the criteria for production and perception specificity apply. In the gesture studies in primates, the production of single gestures and also combinations of them is highly specific to an external stimuli (Hobaiter and Byrne 2014). In birds, they point with their body and head towards the predator and the other group members apply the same pointing position. This behavior has been suggested to meet the criteria for referential gestures, as the signal is directed towards the eliciting referent and seems ineffective for other purposes (Pika and Bugynar 2011). Whether the body vibration signal of the fish to recruit an eal to join the cooperative hunt (Vail et al. 2013) also fits the criteria of a functional referential signal is something to discuss (Smith 2017). Similar to the vocal recruitment calls in the bird studies, it may just represent the social recruitment for hunting together from the signaler side, while for the receiver it may refer to prey. These would then only be object specific on the receiver side but not on the production side (Seyfarth and Cheney 2003).
Functionally Referential Signals and the Evolution of Human Language
An obvious approach, when trying to understand the evolution of human language, is the comparison to animal communication, and in particular identify and compare the underlying cognitive mechanisms. Zuberbühler and Neuman (2017) suggest that in humans at least three levels of information processes for the referential nature of human language need to be considered: (a) fulfill the criteria of functional referentiality with the production of distinct signals to specific external events or objects; (b) the signaler taking the nature and the composition of the audience into account; and (c) the ability of the signaler to take the knowledge of the audience into account. Over the last decades, empirical evidence in several animal species for the first aspect on the production of functionally referential signals has been convincingly accumulated, and in several circumstances, the signalers also seem to be aware of the nature of the audience and adjust their signaling accordingly. In contrast, only for chimpanzees, empirical evidence exists that animals are also able to take into account the knowledge state of the receiver (Crockford et al. 2012), while in human communication, all of the three cognitive processes seem to be present, often at the same time.
The study of referential signals in animals still seems to be a useful theoretical framework for the understanding of the evolution of animal communication and human language. It has stimulated much research in identifying the underlying cognitive mechanisms involved in animal communication. The critical evaluation of the theoretical framework of referentiality in animals has helped to rethink and potentially in the future also to extend it in aspects that may advance the understanding of information processes involved in communication. The application of linguistic concepts can help to formulate new hypotheses but also needs to be applied with caution and any conclusion need an objective evaluation of the existing empirical evidence in animal communication. To produce such evidence is challenging, as the according experiments in animals are difficult, in particular also to perform the correct control situations. We are still far away in understanding how animals categorize their environment, and therefore it is difficult to make unbiased interpretations of what information affects the signaler in the production of signals, and what information is of importance in the response of the receivers.
The University of Zurich has funded the writing of the article.
- Clay, Z., Smith, C. L., & Blumstein, D. T. (2012). Food-associated vocalizations in mammals and birds: what do these calls really mean ?: Animal Behaviour, 83(2), 323–330.Google Scholar
- Heinrich, B., & Marzluff, J. (1991). Do common ravens yell because they want to attract others? Behavioral Ecology and Sociobiology, 28, 13–21.Google Scholar
- Marler, P., Evans, C. S., & Hauser, M. (1992). Animal signals: Motivational, referential or both? In H. Papoušek, U. Jürgens, & M. Papoušek (Eds.), Nonverbal vocal communication: Comparative & developmental approaches (pp. 66–86). Cambridge, UK: Cambridge University Press.Google Scholar
- Struhsaker, T. T. (1967). Auditory communication among vervet monkeys (Cercopithecus aethiops). In S. Altmann (Ed.), Social communication among primates (pp. 281–324). Chicago: University of Chicago Press.Google Scholar
- Zuberbühler, K., & Neuman, C. (2017). Referential communication in nonhuman animals. APA handbook of comparative psychology.Google Scholar