INTRODUCTION

A major focus in the study of bird song over the past three decades has been on the involvement of learning during development. At a basic level, two models of learning mechanisms have been proposed: instructive and selective (Jerne 1967; Changeux et al. 1984). In an instructive model, environmental stimulation adds information not already present in the behavioral repertoire. When a young bird memorizes a novel song, it is instructed. In contrast, in a selective model, learning consists of the selection of behavior(s) from a pre-existing repertoire as a function of experience. At the time of stimulation, the animal already possesses the potential or ability to perform the behavior. Therefore, the test to distinguish between the two models is to present a novel stimulus and to record whether it is learned.

Research on song learning has been guided largely by an instructive model of learning, embodied in the sensorimotor model first proposed in Konishi's (1965) study of song development in the white-crowned sparrow (Zonotrichia leucophrys). The sensorimotor model includes two stages: a sensory (instructive) phase in which songs are memorized, and a sensorimotor phase in which the bird compares its own song, via auditory feedback, to the memory trace acquired earlier.

One consequence of song learning is the formation of geographic “dialects” in which males at one location sing similar songs that differ from those of the same species at other locations. If vocal plasticity in birds is mediated solely by an instructive mechanism, then song matching dialects arise when males breed in the same area where they acquired their song(s).

INTRODUCTION

Experiments on song development have a long history dating from those performed by the Baron von Pernau with chaffinches (Fringilla coelebs), published in 1768 (Thielcke 1988). The Baron documented regional song dialects, showed that learning was restricted to a time window or “sensitive phase,” and demonstrated the bird's preference for learning songs of conspecifics over those of allospecifics or “stimulus filtering.” With the advent of instrumentation that allowed capture and analyses of sound, i.e., magnetic tape-recording and sound spectrographic technology, Thorpe (1958) was able to test these conclusions objectively. He played tape-recorded conspecific and allospecific songs to hand-raised, naive chaffinches reared in acoustic isolation. His experiments confirmed the existence of sensitive phases and stimulus filtering mechanisms during the song-learning process.

Thorpe's study set a standard for avian song ontogeny protocol: hand-raised naive experimental or pupil birds are isolated in sound-proof chambers and exposed to taperecorded vocalizations of model or tutor birds. An investigator is then able to: (a) control the number of songs played to the experimental birds to determine the minimum number of songs required to effect learning (Petrinovich 1985; Hultsch & Todt 1992; Peters et al. 1992; Hultsch 1993); (b) demonstrate that birds may recognize conspecific song by sound alone (Konishi 1985); and (c) test the effect of sound degradation on the choice of tutors by pupils (Morton et al. 1986).

That social factors could influence the choice of a song tutor by a pupil was first appreciated by Nicolai (1959). His work suggested that tape-tutoring experiments, although valuable, may in some cases test only what pupils can do under the conditions imposed by the investigator and not the pupil's actual or potential capabilities in nature.

INTRODUCTION

The study of vocal development in nonhuman animals has focused primarily on vocal production. However, vocal development involves not only production, but development of appropriate usage and appropriate responses to the calls of others. The relative neglect of usage and responses has led to a distorted view of vocal development as emphasized by Seyfarth & Cheney (Chapter 13) for monkeys and West et al. (Chapter 4) for birds. The argument is often made that nonhuman mammals differ fundamentally in vocal development from both birds and human beings but, as we have argued previously (Snowdon & Elowson 1992) and Seyfarth & Cheney (Chapter 13) argue, primates are similar to humans and birds if all three components of vocal development – production, usage and response – are evaluated.

In this chapter we argue that when functionally similar vocalizations are chosen from species that have similar social organization, then similar developmental processes will be found regardless of the taxon studied. We examine the idea that nonhuman primate vocal structures are fixed and review evidence that vocal plasticity may be quite common. We then provide three examples of phenomena from our research with marmosets and tamarins that illustrate how plasticity in pygmy marmoset trill vocalizations can be influenced by changes in social companions, how infant babbling in pygmy marmosets is a social interaction, and how the structure and usage of food-associated calls is acquired by cotton-top tamarins, and how social environments can inhibit the expression of these calls.

In the late 1960s a series of developments in linguistics, developmental psycholinguistics and animal communication led to a convergent model of vocal development in human and nonhuman species. As presented by Lenneberg (1967) and Marler (1970), the development of language and of bird song required exposure to species-specific codes during a sensitive period of development, after which subsequent learning was extremely limited. The amount of input required could be quite small, and this input could be effective regardless of social interactions. Both birds and humans needed intact hearing and an extensive time for practice (babbling for human infants, subsong and plastic song for birds) to acquire adult competence in vocal production. Subsequent to this practice crystallization occurred, and further changes in vocal structure were rare.

This paradigm has led to extremely productive research over the past 25 years, not only in the study of vocal development but also in the understanding of the neurological controls of vocal production. However, as researchers interested in the ontogeny of primate and avian vocal communication, we have become increasingly aware of the need to consider some modifications to this paradigm. Some forms of social stimulation can extend sensitive periods for song learning in birds, and songs and calls in some species can be modified throughout life, often in response to changes in social stimuli. Parrots, dolphins and great apes with exceptional training acquired codes with some similarities to human language. Yet at the same time there was little evidence of vocal plasticity in nonhuman primates, suggesting a gap in continuity of developmental processes in the evolution from birds to humans.

INTRODUCTION

The first studies of language acquisition by the child described mainly the developmental stages of this specific human ability From Piaget (1923) to Brown (1973), authors were interested mostly in the different formal aspects of the acquisition: for example, the age of onset, total amount of language at any age, mean length of utterance, and emergence of grammar. These studies considered the abilities of each child as representative of the general linguistic abilities of the human species at this ontogenetic stage.

More recently, new trends have appeared, where language is studied in a more pragmatic way: it is considered as a means, at each developmental stage, for a child to elicit real communicative interactions. Thus, while admitting that the general stages of language development are alike in any child (Locke & Snow, Chapter 14) such an approach to the development of communication implies integrating various aspects that are usually considered separately by different researchers.

On the one hand, to consider the emerging linguistic skill as part of the larger phenomenon of communication implies integrating the analysis of the linguistic competence of a child at a given stage with that of previous stages, in particular with babbling. It implies also the integration of other communicative behaviors: for example, approaches, emotional addresses, and object exchanges. It can be hypothesized that, when a child is communicating, it is both acquiring the human language and developing its personal communicative style with human beings.

INTRODUCTION

Vocal learning involves the ability to modify and acquire new signals in an organism's vocal repertoire through the use of auditory information and feedback. Humans and many avian species, particularly songbirds, have demonstrated similarities and analogous patterns in the vocal acquisition of their respective repertoires. These similarities include the importance of auditory input, feedback, and social influences on vocal structure and acquisition, and stages of developmental overproduction, selective attrition, and vocal babbling/subsong (for reviews, see Kroodsma 1982; Pepperberg & Neapolitan 1988; Locke 1990, 1993a,b). Finding such parallels in phylogenetically distinct species is striking and suggests a convergence in strategies of vocal learning.

Evidence for vocal learning in other species is rare. Studies of vocal learning in nonhuman primates have suggested that learning plays a role in vocal development of contextual use and comprehension (Seyfarth et al. 1980; Cheney & Seyfarth 1982; Seyfarth 1986; Hauser 1988; Gouzoules & Gouzoules 1989) but clear evidence for the learning of vocal repertoires by nonhuman primates has been slow to emerge. However, recent results of studies of nonhuman primates (Elowson & Snowdon 1994; Snowdon et al., Chapter 12; Mitani & Brandt 1994) and birds (Brown & Farabaugh, Chapter 7) suggest greater vocal plasticity than was previously described and point to an importance of social factors on vocal structure and acoustic variability of calls. Therefore, to more clearly elucidate the phenomenon of vocal learning it is important to make a distinction between vocal learning (the ability to acquire new elements in one's vocal repertoire) and vocal plasticity (the ability to modify signal structure due to social or environmental conditions).

THE SOCIAL FUNCTION OF LANGUAGE

This chapter analyzes alternative types of conversational action used to build social organization among girls and boys in an African-American working class neighborhood in Philadelphia. Participants work together to generate distinctive definitions of the situation appropriate to the task at hand, and the same individuals articulate talk and gender differently as they move from one activity to another. Making use of the same language system, children select alternative ways of putting these forms to use, constructing a range of diverse activites and social arrangements that can highlight either affiliation or competition.

From the perspective of ethology, Cullen (1972, p. 101) has argued that “all social life in animals depends on the coordination of interactions between them.” To achieve collaborative activity humans need to display to one another culturally meaningful behavior – articulating for their recipients what they are up to and how they expect others to respond. Sociologist Georg Simmel (1950, pp. 21–2) has stated that “if society is concerned as interaction among individuals, the description of the forms of this interaction is the task of the science of society in its strictest and most essential sense.” In that language provides the tool through which humans coordinate their behavior, then what is required for an adequate understanding of social organization is close attention to talk itself.

INTRODUCTION

In the mid-1960s Klaus Immelmann began a series of experiments with domesticated zebra finches (Taeniopygia guttata) that led to seminal contributions to two related fields, song learning and sexual imprinting. Immelmann (1969) manipulated the auditory and social experiences of young males in their first 100 days of life and found that those denied any contact with singing males failed to sing the normal zebra finch song at adulthood. He concluded that song in this species, like that of most songbirds, must be learned. When he isolated young from foster parents (Bengalese finches (Lonchura striata var. domesticaj) at different ages he found that the sensitive phase for song acquisition began as early as 25 days of age, about a week after fledging, and ended around 80 days of age, around the onset of sexual maturity. Furthermore, young males did not learn from just any singing adult, but preferred to copy from the male with whom they formed a personal bond. In most instances this was the father or foster father and the bond was based primarily on the provisioning relationship, the most basic filial bond. Immelmann (1969) hypothesized that wild zebra finches would be likely to learn the songs of their fathers, and an early end to the sensitive phase was necessary in order to prevent learning from heterospecific estrildines.

The experimental possibilities raised by Immelmann's intriguing study of song eventually stimulated a steady series of follow-up experiments by other researchers, in particular, by P. J. B. Slater and coworkers, who used song learning in domesticated zebra finches as a model for teasing apart the subtle interactions involved in the development of behaviour (for a review, see Slater et al. 1988).

INTRODUCTION

Vocal learning is a very widespread characteristic of songbirds, and a large variety of these learning processes has been described over the last decades. Learning can lead to different types of variation and results in song sharing that can be geographically localized and is then considered as “dialects.” The distribution of these variations can be limited to a few birds and/or cover large areas. Experimental studies have given precise information about the mechanisms involved, in particular in terms of “timing.” However, naturalistic validations may be necessary for us to fully understand the functional significance of vocal learning (see discussion of the sensitive periods, in the literature). There is a need for integrative studies and it is important to consider communication in its own context, which is that of social interaction (for language as a social act, see Goodwin 1990). A social organization needs an adapted communicative system, and comparative studies can give us hints about the evolutionary bases of vocal learning. Comparison can be made between species or phylogenetic groups but also between populations of a same species.

Likely candidates to help us to understand these processes are highly social animals that can adapt to different social environments. Starlings clearly correspond to this definition and here I examine through experimental and naturalistic studies the possible relation between song acquisition, song sharing and social organization.

The European starling (Sturnus vulgaris) occupies the largest geographic range of all the species of Sturnus (Feare 1984). It is present in Asia and in Europe from Scandinavia to Spain and Italy, and even in North Africa. It has also been introduced successfully in North America, New Zealand, and South Africa.

INTRODUCTION

It might seem odd to suggest that the study of bird song needs a social agenda. Isn't it obvious that birds sing to attract or to repel one another? Isn't it now clear that birds need social experience to develop species-typical repertoires? Although the answer to these questions is “yes,” we are convinced that only the surface structure of social influence has been uncovered – the deep structure remains to be explored. The purpose of the chapter is to defend this position by examining some of our own efforts to study social influences. We begin with an account of some of the formative experiences that shaped the directions of our research. We follow with some of the historical themes that guided us and others studying songbirds. Then, we describe some of our most recent efforts to examine new themes relevant to avian communication, themes quite familiar to those studying primates. In particular, we focus on the difference between communicative form and communicative competence. We wish to draw a greater distinction between the processes involved in developing a potentially communicative signal and the processes involved in learning how to use those signals effectively (Seyfarth & Cheney 1986 and see Chapter 13; Snowdon et al., Chapter 12). We conclude that analyses restricted only to the structural nature of vocal signals are inadequate to capture the developmental processes leading to vocal communication. We must go beyond studies of songs and focus on the singers, listeners, and the contexts framing communication.

FLASHBULB MEMORIES: A BIOASSAY OF SONG, A BIOASSAY OF SINGERS

In May of 1973, we witnessed an event that led to a series of studies spanning the next two decades.

The capacity of a visual system to resolve fine spatial detail depends on several factors, some associated with the stimulus object, others with the eye, and still others with the central visual pathways. Sensitivity to spatial detail is commonly called acuity, but it is important to remember that this word is applied to several different performance measures. One reads, for example, of minimum separable acuity, grating acuity, vernier acuity, and stereoscopic acuity. The neural mechanisms that mediate these discriminative capacities are not necessarily all the same, despite the fact that the same word, “acuity,” is used for them. It should also be kept in mind that these measurements usually reflect performance at the extreme limit of some functional capacity, rather like determinations of the absolute threshold for detection of light. Thus, acuity measurements of all kinds can give misleading impressions of the sensory tasks in which the nervous system is routinely engaged.

Minimum Separable Acuity and Minimum Angle of Resolution

When two dots or short line segments are made to approach each other in the visual field, a separation is reached at which the subject reports the presence of only one object. The critical angular spacing of the stimuli when they are just resolved is called the minimum angle of resolution (MAR) (Figure 9-1), a measure analogous to the two-point discrimination threshold in somatic sensation. The MAR is affected by many factors, including the brightness of the stimuli, the state of retinal adaptation, and the position of the stimuli on the retina.

One of the main tasks of vision is to distinguish objects from their backgrounds, which at a minimum requires detection of spatial differences in patterns of retinal illumination. This capability is greatly enhanced if the organism can also discern differences in the wavelengths of the light forming the retinal image. These spectral or chromatic patterns can also be useful in the identification, as well as the detection, of important objects such as food sources or predators. The spectral composition of the retinal image of a given object depends on the spectral content of the incident light, the object's reflectance characteristics, and to some extent the differential absorption of certain wavelengths in the ocular media before light reaches the retinal photoreceptors. It is the photoreceptors that first dissect the retinal image into its chromatic components, and in the vertebrate retina this process depends on the cones.

Color vision generally requires the presence in the retina of two or more photopigments with different spectral sensitivities. As described previously, the photopigments of all animals are composed of an opsin, a large, membrane-spanning protein, and the chromophore, 11-cis retinaldehyde. When the chromophore absorbs a light quantum, it changes shape and activates the opsin, which then functions as a catalyst for further reactions in the photoreceptor (see Chapter 5). Retinaldehyde by itself is colorless, but when combined with the opsin its geometry is modified, and the combination becomes a photopigment that absorbs light in the visible range. Changes in one amino acid group at critical points in the opsin can significantly alter the spectral sensitivity of the opsin-chromophore combination.

All information about the retinal image that is directly available to the brain is transmitted by the axons of retinal ganglion cells. In higher mammals, and particularly primates, the projection to the cortex via the lateral geniculate nucleus of the thalamus appears to be essential for conscious visual perception. Retinal axons also reach the hypothalamus, superior colliculus, pretectum, and various other nuclei of the brain stem and diencephalon that subserve a variety of functions, such as reflex orientation to visual stimuli, stabilization of gaze, and control of pupil diameter. This chapter focuses primarily on the retino-geniculate component of the projection to the cerebral cortex, but several of the principles dealt with here are relevant to brain-stem projections as well.

Parallel Processing of the Retinal Image and Classes of Ganglion Cells

Previous chapters have shown how the retinal circuitry establishes one channel to signal increments and another to signal decrements of illumination. These on and off channels encode complementary versions of the distribution of light on the retina and transmit them in parallel to the central nervous system. Also, as discussed earlier, other features of the retinal image, such as movement direction, can be signaled by specialized ganglion cells. Thus, the retinal image is not transmitted in raw form to the brain, but is analyzed in different ways by different ganglion cells, which then communicate their “Views” of the image along separate channels. Different animals segregate different aspects of the retinal image as part of this strategy of parallel processing.

Major Anatomic Features of the Eye

Figure 2.1 illustrates schematically the major components of the human eye, which resembles that of most other primates. The sclera is a tough outer coat that is fibrous in humans but contains bone or cartilage in some other vertebrate species. The cornea is continuous with the sclera and provides the first element of the refracting media that bend the light to form an image on the retina. The lens lies behind the iris and in front of the vitreous humor, which fills the greater part of the globe. Aqueous humor fills the posterior chamber (the space between the lens and iris) and the anterior chamber (the space between the iris and the cornea). The posterior and anterior chambers are continuous through the pupil, the aperture formed by the iris.

The general features of the retina, the multilayered neural structure lining the back of the eyeball, can be visualized in the living eye with an ophthalmoscope or special camera (Figure 2.2). Axons leave the retina through the optic disc or optic papilla and enter the optic nerve to reach the brain. At the posterior pole of the eye, the retina thins to form the fovea, an area specialized for high-acuity vision. The visual axis is an imaginary line from the fovea through the center of the pupil (Figure 2.1). Behind the retina is the pigment epithelium, which is separated from the sclera by the vascular choroid.

This chapter provides an overview of the projections from the retina to the brain in vertebrates and reviews the key terms used in describing the pathway. The major components of the pathway and their functions are examined in greater detail in subsequent chapters.

The Visual Fields

The central projections of the two eyes map the visible world onto the brain. To understand this process, it is important to know how the visual field of each eye is described and how the projections from the two eyes are combined in the central pathways. The retina of each eye is conventionally divided into nasal and temporal parts, on the basis of proximity to the nose or temporal bone, respectively. Similarly, the visual field of each eye is divided into nasal and temporal parts, and because of the inversion of the retinal image by the eye's optics, the nasal visual field is imaged on the temporal retina, and the temporal field on the nasal retina. Figure 4.1 schematizes the projections of the visual fields in an animal whose eyes are located at the sides of its head. In such lateral-eyed animals, the axons from one retina generally cross completely in the optic chiasm, so that the input from that eye is directed at the contralateral hemisphere of the brain.

Figure 4.2 illustrates diagrammatically the monocular visual fields as they would appear to a frontal-eyed human observer, left eye (oculus sinister, O.S.) on the left, right eye (oculus dexter, O.D.) on the right.

When the eyes face the front, the central part of the visual field is imaged on both retinas (see Figure 4.4). This bestows certain advantages for depth perception, but also creates a formidable problem for the brain: how to ensure that the two retinal images are transformed to yield a unified perception of the part of the visual field seen by both eyes. Failure to achieve this, which sometimes happens in pathological conditions of the nervous system, results in diplopia, the perception that there are two objects when there really is only one. Diplopia can be demonstrated by pushing gently on the skin at the side of one eye to misalign the two visual axes.

Binocular Single Vision

The encoding of the two retinal images of a single object to yield a unique perception results in perceptual fusion of the two images. In discussing fusion, it is important to distinguish between it and two other phenomena, fixation and focus. If the visual axis of one eye is directed at an object so that the image is positioned on the fovea, the eye is said to fixate the object. It is possible to deliberately place an image outside the fovea, but the term “fixation” is generally used to mean foveal fixation. The fixated object will be in focus only if its distance from the eye and the power of the eye's optics permit the formation of a crisp retinal image.