Vocal Learning in Mammals

Vincent M. Janik; Peter J.B. Slater    School of Biological and Medical Sciences University of ST. Andrews ST. Andrews, Fife, United Kingdom

I INTRODUCTION

In this chapter we survey the occurrence of vocal learning in mammals and discuss possible reasons it has evolved. But first it is important to be clear about what we mean by “vocal learning.” The term has been used to describe the influence of learning on a variety of different aspects of vocal communication. Learning can affect the generation of sounds, their usage, and their comprehension. While modifications in sound generation as a result of experience can be described as learning sounds, those in comprehension and usage are rather different phenomena, which are perhaps better described as learning about sounds. Vocal learning, as we discuss it here, refers only to learning sounds, that is, to instances where the vocalizations themselves are modified in form as a result of experience with those of other individuals. Learning that affects usage and comprehension of sounds will be referred to as contextual learning as opposed to vocal learning. Contextual learning in relation to vocal communication is relatively common among mammals. The list of animals in which the utterance of a vocal signal has been brought under conditional control, i.e., which have learned to change the context in which they are using sounds, comprises rats (Lal, 1967), guinea pigs (Burnstein and Wolff, 1967), dogs (Salzinger and Waller, 1962), cats (Molliver, 1963), sea lions (Schusterman and Feinstein, 1965), primates (Myers, Horel, and Pennypacker, 1965; Randolph and Brooks, 1967; Wilson, 1975; Aitken and Wilson, 1979), and dolphins (Lilly, 1965). Other forms of contextual learning in vocal communication involve learning to recognize particular sounds, or learning to react to sounds differently as a result of experience. These are important ways in which learning may influence vocal communication, and the behavior associated with it, but they are not examples of vocal learning in the strict sense in which we use the term here.

Some types of modification through learning are likely to be easier to achieve than others because different sound parameters are controlled by different mechanisms. Overall duration and amplitude of a sound can be altered by simple modifications of exhalation alone. A longer exhalation phase or higher air pressure while producing the sound is all that is required to cause such changes. For learning to occur here, only the activity of respiratory muscles need be modified by experience. Such changes in duration or amplitude do not affect the overall form of a call. On the other hand, the detailed matching of a sound pattern requires much more complicated alterations. The fundamental frequency of a sound can be altered only if the activity of muscles controlling the vocal apparatus itself can be modified by experience. In tonal signals this is required to achieve change in the frequency contour of a call. Of course, other more complex processes can be involved in sound production. Rapid amplitude modulation, for example, can cause additional frequency bands in a call. Seemingly subtle changes that require a high degree of coordination between respiratory, laryngeal, and articulatory muscles, like those leading to differences in voice-onset time, can also be involved. Where possible we look at modifications in duration and amplitude and those in frequency parameters separately, and point to these different levels of motor control as they occur in the vocal learning of different groups.

Vocal learning has been described only in birds and mammals, and even among these the evidence is patchy. It has been found in all songbirds (Oscines) studied to date (Kroodsma and Baylis, 1982), but appears not to occur in the closely related Suboscines (Kroodsma, 1984, 1989). Convincing evidence comes from only two of the twenty or so other orders of birds: the hummingbirds (Apodiformes) (Baptista and Schuchmann, 1990; Gaunt, Baptista, Sanchez, and Hernandez, 1994) and the parrots (Psittaciformes) (Todt, 1975; Pepperberg, 1981). The three groups showing vocal learning are only distantly related to each other, suggesting that it has evolved among birds on at least three separate occasions. In mammals, the importance of vocal learning in our own species contrasts remarkably with the scarcity of evidence elsewhere. Part of the reason for this may be the lack of relevant studies. Absence of evidence for vocal learning in a particular species is certainly not evidence for its absence.

In our review of the literature on vocal learning in mammals we attempt to determine the extent to which it occurs in species other than our own, and whether it is widespread or patchily distributed as in birds. This survey enables us to compare and contrast birds and mammals, and to consider the possible functional significance of vocal learning. It may perhaps also shed some light on why it occurs in humans. But first we discuss the methods that have been used in various studies and the extent to which these can give unequivocal evidence for or against vocal learning.

II EVIDENCE FOR VOCAL LEARNING

There are several pitfalls in trying to show vocal learning in a particular species. A fundamental problem is the question of whether a particular call was in an animal’s repertoire before it was first noted. Apparent changes in the call repertoire of an animal often relate to other changes in its environment. Examples are alterations in social context, because of the introduction of foreign animals or a change in status, changes in the habitat that alter its sound transmission characteristics, or seasonal events that influence the diurnal behavior patterns of the animal. If a new call arises at the time of such an event, it could be because of a change in the frequency of occurrence of calls that were already present in the repertoire rather than vocal learning.

But even a truly new call could arise for different reasons. Maturational processes or improvisation could be responsible, rather than copying from other individuals. Maturational processes lead to changes in vocal tract morphology that can influence sound characteristics. Thus, simple observations of changes in the call repertoire during ontogeny are difficult to interpret. Vocal learning may or may not be involved.

Improvisation is another process that leads to the production of new calls. Various different mechanisms can be used to achieve improvisation, and vocal learning, again, may or may not be involved. One possibility is the production of completely new sounds through random sound generation. This form of improvisation would be an interesting case of vocal flexibility. According to our definition it does not involve vocal learning, however, since experience is not required. As we see in our survey, completely random sound production has so far never been the only possible explanation for an observed change in call structure. But there are other forms of improvisation that do involve learning. If an animal produces a completely new call that avoids overlap with calls of other individuals, experience might be used to achieve this avoidance. This would be a case of vocal learning according to our definition. A more restricted form of improvisation might involve a recombination of given subunits of a call. If these units can be produced on their own, this form of improvisation represents a special case of contextual learning. It is simply a matter of calls that are already present in an individual’s repertoire being produced in a new context. Finally, an animal could learn different parts of other individuals’ calls and combine them to form a new call. This sort of improvisation would be a clear case of vocal learning.

Vocal learning is obviously difficult to investigate if changes in calls do not result in matching those of other individuals or model sounds. The most clinching evidence for vocal learning comes from experiments in which animals are trained to imitate sounds that have not been in their repertoire before. If animals are able to copy sounds that are very different from sounds in their natural repertoire, like human speech or computergenerated sounds that were designed to be different, vocal learning has clearly been demonstrated. In animals that are not capable of imitating such sounds, vocal learning within the natural repertoire can nevertheless be shown by rearing experiments, if infants that were raised with different acoustic stimuli are found to match the sounds they heard in detail. It is unlikely that selective reinforcement of randomly produced sounds could result in detailed matching of sounds produced by other animals.

Such rearing experiments have commonly been carried out on birds, but few have been attempted on mammals. There are probably two main reasons for this. First, many of the mammalian species involved, such as whales and dolphins, are difficult to keep in the highly controlled acoustic environments necessary for such studies. Second, the species involved are highly sociable and subjecting individuals to experimental treatments involving deprivation likely to lead to suffering is not easy to justify. Because of these difficulties, for many mammals the evidence for or against vocal learning is more circumstantial.

The main source of such less direct evidence comes from geographic variation in vocal signals. Where neighboring animals, or those in a social group, share sounds that differ from more distant individuals or those in...