Chapter 5

ONTOGONY OF COMMUNICATION

Gordon M. Burghardt

How does an individual animal arrive at the condition in which he or she initiates and responds appropriately to communicatory signals? How does the animal seemingly "understand" what is meant and "know" what to do? These questions juxtapose ontogeny and communication, two central and controversial concerns in the modern study of behavior. Since a theoretical or empirical synthesis is not possible at the present time, I will limit myself to discussing some selected ideas and research on ontogeny and communication and to introducing some of the methods and concepts employed in their study, while attempting to provide a framework for organizing our knowledge and asking questions. The reader is encouraged to consult other statements that present overlapping treatments with different emphases (e.g., Beer, 1971; Candland, 1971; Hebb, 1972; Hess, 1973; Hinde, 1970, 1971b; Kuo, 1967; Lehrman and Rosenblatt, 1971; Schneirla, 1965).

The specific emphasis will be placed on ethology because it provides the most comprehensive and adaptable approach for a holistic treatment of communication, ontogeny, and other important questions in animal behavior (Tinbergen, 1963), while at the same time incorporating data from other fields, including psychology, physiology, and ecology (e.g., Hess, 1962; Hinde, 1970). Ethologists have devoted most of their efforts to species-characteristic behaviors and the signals involved in communication.

Ontogeny

According to the Oxford English Dictionary, ontogeny is "the origin and development of the individual being" and "the history or science of the development of the individual being," including, but going beyond, embryology. In other words, it is synonymous with "development," as applied to organisms. In using the word, it is clear that no presuppositions are made about the factors that influence the development of the organism and, consequently, its behavior. To study ontogeny means to investigate the complex antecedents and consequences of the interaction of genetic, structural, physiological, and environmental events from fertilization until death.

All social behavior involves communication and takes place over time; any temporal process can be viewed as involving some developmental aspect. It would simplify our problem immensely if we could limit ontogenetic processes to changes that occur over a relatively long (in the life of the individual) time span. Yet imprinting can occur in a few minutes; a single exposure to tainted food can lead to long-term aversions by rats, and a few seconds' exposure to light can synchronize metamorphosis in insects. Fleeting events can have long-lasting effects and be important parts of normal ontogeny. Potentially all processes (not just the obvious ones) that result in behavior change can be seen as relevant to ontogeny.

Most ethologists select birth or hatching as the starting point for ontogenetic studies. A case can be made that this is appropriate when dealing with communication since it is largely a postnatal phenomenon. However, for answering certain questions it is essential to study embryos, and stimulating prenatal studies dealing with communication are being performed (e.g., Gottlieb, 1971a; Hess, 1973; Impekoven and Gold, 1973). Serious technical problems have hindered rapid advances in this area and limited most prenatal studies on communication to auditory signals in birds.

Communication

Definitions of communication usually exclude phenomena that should be encompassed or include phenomena (such as most predator-prey interactions) that are best excluded. It is easy, then, to reach the position that no definition can really differentiate communication from noncommunication or that "We all know what communication means." If communicatory processes have evolved from noncommunicatory ones, then we should expect intergradations of all sorts and an absolute distinction will be unattainable. This position is unfortunate. Often examples at the borderline between accepted distinctions lead to clarifications enhancing understanding of unequivocal examples in either category (e.g., living and nonliving, plant and animal, classical and instrumental conditioning), and this seems true of communication also.

Communication is usually treated as synonymous with social behavior. If we accept "communication" as equivalent to "social behavior," then we do not need the term "communication." However, "communication" is used to refer not so much to the behavior itself as to its signal function, its information content, and its reception and interpretation by other organisms. Most examples of communication are, however, inferred from intraspecific social interactions and take wondrously varied forms (Otte, 1974).

Elsewhere I have reviewed definitions of communication and rejected most of them. A general, but useful, characterization of communication is that it occurs when one organism emits a stimulus that, when responded to by another organism, confers some advantage (or the statistical probability of it) to the signaler or its group (Burghardt, 1970b). While the emphasis of this functional definition is based on the signaler, it does not deny that "signalers" and "receivers" can have mutually evolved and interfaced communicational systems, nor does it preclude the possibility that disparate mechanisms may be involved.

This view of communication seems to isolate the relevant aspects of communication (e.g., eliminating predatory encounters yet keeping other interspecific ones) more clearly than some other recent formulations (Hinde, 1974; Wilson, 1975). Understanding the diversity of communication systems in animals entails a close analysis of signal characteristics, habitat, and adaptive value.

Questions about Ontogeny

Assuming the scientist has developed adequate techniques of observation, his first task is to discover which phenomena or events are related to which other phenomena and to ascertain the nature or form of these relationships. This first stage may be described as involving the discovery of the low order empirical laws holding among particular observed events. Playing important roles in the accomplishment of this task, of course, are the procedures or operations of experimentation and measurement. [Spence, 1953:283]

What are the ontogenetic factors, processes, or antecedent conditions shaping and altering the diverse kinds of communication in animals differing in almost every aspect: life span, speed of development, sensory capacities, motor abilities, ecological requirements, and complexity of communication? Obviously the ontogenetic mechanisms will differ a great deal. Not so obvious is the fact that psychologists and biologists have too often opted for only one or a few mechanisms, usually mutually exclusive ones at that. This seems to be due to the hope that most differences are superficial, not "real." Perhaps the success of genetics and physiology in elucidating principles (e.g., in blood circulation) that apply to entire orders, classes, and even phyla keep fertilizing the hope of harvesting specific yet widely applicable explanations. In relation to behavior, this may be true someday, perhaps never—but definitely not now.

The antecedents to communicatory acts must be characterized and a major aim of this chapter is such a brief characterization. As a starting point, a distinction between the species' evolutionary heritage and the individual's environment and consequent experience is useful if it is carefully and thoughtfully applied. Many words have been used to summarize the various aspects of this supposed nature-nurture dichotomy: innate, genetic, instinctive, unlearned, endogenous, constitutional, phylogenetically adapted, maturational, reflexive, stereotyped, and hereditary on one side; learned, experience, stimulation, modification, environmental, variable, conditioned, acquired, plastic, and emergent on the other. Failure to recognize that the terms in either group are not synonymous with one another, and that these two sets do not refer to necessarily mutually exclusive events, led to loose and often needless arguments. Scientists are now agreed that the behavior of any organism is a combined outcome of its heredity, environment, and sometimes active participation. Where workers differ is on the relative emphasis given one or the other, the questions asked, and the methods and terminology employed (see Burghardt, 1973; Gottlieb, 1973; Hebb, 1972; Hinde, 1970; Lehrman, 1970; Lorenz, 1965; Thorpe, 1974; and Whalen, 1971 for some contrasting views). Here we will try to short-circuit some of these issues by providing a framework that does not contain, or even imply, an inherent bias for either nature or nurture. The lack of such a framework for posing questions has led to many impatient workers' verbally ostracizing the entire issue of the causal analysis of ontogeny from modern ethology. But this stance is unacceptable if we truly want to understand behavior and its origins, regardless of whether we accept Tinbergen's (1963) characterization of the four classes of problems in behavior (ontogeny plus evolution, function, and causation) or Schneirla's (1966:284) more extreme position that "Behavioral ontogenesis is the backbone of comparative psychology."

The living organism we see, smell, and hear is a phenotype that results from a given constitution (the genotype)developing in a given environment. Anatomical, physiological, and behavioral characteristics of organisms are all phenotypic characters. Now note that the first set of "nature" terms above ("innate" et al.) traditionally refer to properties derived from the genotype, while the second, or "nurture," set ("learning" et al.) refer to postfertilization environmental effects on the phenotype. It is apparent that the two sets of terms cannot be in opposition with a sort of inverse relationship to each other. Obviously all genotypes need an environment and the environment can only act on phenotypes containing a given genetic constitution. But we cannot stop, as some would, with an "everything affects everything" view. It may, but not equally, and this must be explained.

The next step, then, is to posit that any character (behavior or structure) can be differentially affected by given ranges of environmental (developmental) variation. That is, different phenotypes can result from the same genotype. Similarly, different genotypes can result in phenotypes that are not readily discriminable from one another. Many of the issues in ontogeny revolve around the nature of the limits set by different genotypes on phenotypes. Today a rather "tight" genotype-phenotype correlation would lead to labeling a behavior "genetically fixed," "environmentally stable," or "innate," and a "loose" correlation would lead to the label of "environmentally labile," "learned," or "acquired." But where to draw the line in a continuous gradation is a worry of many people (e.g., Marler, 1973). This issue could be handled if determining the nature of the genotypic limits was straightforward. But not only do methods of accomplishing this vary, but also researchers often take different levels or parameters of behavior as their units of analysis and then generalize to a broader functional unit. For example, Hess (1973) and many classical ethologists view imprinting in ducks and geese as a highly specialized form of instinctive behavior, bearing little resemblance to "learning" as studied by animal psychologists. Others stress the modifiability of imprinting and the operational similarity of some object acquisition aspects of imprinting to traditional learning paradigms (e.g., Hinde, 1970). We can see the origin of a labeling conflict in this example. If one is forced to label imprinting as "innate" or "acquired," it is obvious who will put it where, albeit grudgingly today.

Mayr (1974) attempts to solve the problem of the genotype-phenotype link by classifying behavior as due to either "closed genetic programs," in which "nothing can be inserted" through experience, and "open genetic programs," which allow for "additional input during the life span of the owner." While this terminology has some value, being derived from molecular biology and information theory, it is sometimes not useful at the level Mayr applies it. The open programs encompass too much, and too large a functional unit is used. Thus he calls imprinting an "open" system, a label that tends to make us lose sight of species differences and the initial visual and auditory preferences that put constraints on even some of the most "open" aspects of imprinting involving species recognition.

As another example, consider food recognition in newborn snakes (Burghardt, 1970a). Snakes of several colubrid genera (e.g., Thamnophis, Storeria) are born with the ability to recognize and attack chemical stimuli from specific types of animals, such as fish, earthworms, leeches, and slugs, which represent species- characteristic prey. Such responses to prey cues are stereotyped, show evolutionarily understandable differences and similarities across species, are difficult to habituate, are seemingly impervious to maternal prenatal experience (Burghardt, 1971), and remain intact over prolonged postnatal stimulus and response deprivation. These findings more than meet the criteria for a closed program advanced by Mayr. Yet within several feedings on actual prey, the young snakes increase their relative preference for it, while a single "bad taste" or illness associated with prey can dramatically decrease the food's palatability (Burghardt, et al., 1973). Even brief exposure to prey odors without ingestion can alter responses to prey! Is food recognition in snakes innate or learned? Genetically fixed or environmentally labile? Controlled by a closed or open genetic program? We seem to be in a quandary.

The resolution may be simple. Let us formalize what many workers are actually doing empirically, but discussing loosely. We need to inquire as to which aspects of a behavior (e.g., topography, patterning, stimulus control, duration) are determined by which genetic or environmental influences. The relevant phenotype is a parameter of the behavior, not necessarily a functional or topographic unit, regardless of the level at which it is defined. We must work to determine the evolutionary antecedents of the behavior repertoire of an individual and their relation to the historical development of phenotypic appearance. We can apply this approach to the question of whether differences in phenotypes are associated with differences in genotype or environment (Whalen, 1971). It is also compatible with asking whether the behavioral "informa-tion" contained in the organism (usually its nervous system) originated in the genome (genetic program or blueprint) or was acquired through experience (Lorenz, 1965; Mayr, 1974; Thorpe, 1974). Both populational and individual approaches (Burghardt, 1973) can thus be accommodated.

This "parameter of behavior" approach largely eliminates the need for arbitrary cutoff points on the continuum between "tight" and "loose" genotype-phenotype correlations. It also helps us realize that there is no such thing as "learned behavior," "innate behavior," and so forth (Verplanck, 1955). These are merely shorthand terms to refer to the origin of the aspect of the behavior in which we are interested. By this view, initial recognition of chemical cues from species-characteristic prey in snakes seems innate. Nothing is implied about the modifiability of such recognition after birth. On the other hand, we must still retain the distinction between inherited and environmentally based factors shaping a behavior, which careful writers on behavior have not been able to avoid, regardless of their verbal inventiveness and clever rhetoric. Since the time of Darwin we have known that phenotypic variation based on some inherited processes is necessary for adaptiveness and evolution.

The use of the term "parameter" here extends and makes more explicit the distinction Lorenz (1965) wanted to make between "character," on the one hand, and "organ" and "behavior pattern," on the other. However, he did not define his usage with sufficient clarity to avoid confusion. The "polythetic" approach advocated by Jensen (1970) and derived from numerical taxonomy is also compatible with this view of using operational approaches to confront, rather than avoid, critical empirical issues.

Species, Operations, and Outcomes

The development and use of instruments and experimental designs that provide for the isolation, control, and systematic variation of the factors in the situation under observation are an obvious requirement for the discovery of laws, especially when the situation is one in which a large number of factors are operating. [Spence, 1953:283]

Our formulation does not necessitate that extremes in the tightness or looseness of genotype-phenotype correlations exist, nor does it stress their relative prevalence; but it does indicate what to look for and how to interpret our findings. Only if the evidence shows that virtually all parameters of a response are "tight" or "loose" can we generalize to the entire functional unit. To illustrate these two points let us look at some experimental examples in a most well-studied area, auditory communication.

Crickets (Gryllus, Acheta) produce sounds through use of the forewing. This process, termed "stridulation," appears only in males. One of the sounds they produce is a species-characteristic call to attract females. Numerous studies have established the following ontogenetically important findings (references in Alexander, 1968, unless otherwise noted; see also Otte, this volume):

1. While external stimuli can trigger singing bouts, they can also occur without any known external clue.

2. Different cricket species may have different but similar calls. Such differences seem to be based largely on the number of separate sound pulses per trill, the number of trills per phrase, and the intervals between them. Females do not get confused and readily recognize their own species call.

3. Hybrid males are generally intermediate between the parent species in call parameters. Yet through backcrossing Bentley (1971) was able to show that some aspects of the call could be correlated to the presence or absence of a specific sex-linked chromosome.

4. Exposing various stages of the developing cricket to foreign sounds, including calls from related species, does not affect the calls of the male or the discrimination and preference for conspecific calls by females.

5. Male crickets deafened either before hearing their calls or as experienced adults still stridulated normally.

6. Severing the wings or otherwise altering proprioceptive feedback had no effect on male calls.

Such results indicate that in crickets the singing of males and the preference of females are stable in spite of a wide variety of environmental manipulations. If there are environmental events that would alter the singing, they have not yet been found. We have here a case of a very tight genotype-phenotype correlation, a nearly perfect example of Mayr's (1974) closed genetic program. Since virtually every aspect studied seems resistant to environmental modification, it is valid to refer to auditory courtship communication in crickets as genetically fixed, environmentally stable, or innate, but only as a shorthand summary of the specific experimental data. One might, however, make reasonable inferences about results to be expected from experiments using related species or alternative manipulations.

Surprisingly, mammalian vocalizations have been less well studied. However, a few examples can be given (see Salzinger, 1973):

1. Squirrel monkey (Saimiri sciureus) infants raised by surgically muted mothers and isolated from all species-characteristic sounds evidenced normal calls. Further, an infant deafened five days after birth also showed the normal vocal repertoire (Winter et al., 1973).

2. Salzinger and others have shown that barking in dogs, meowing in cats, and calls in monkeys can be brought under operant control; that is, the rate of vocalization can be modified by reinforcement (reward) and conditioned to a stimulus (such as a light) that is present during reinforcement. The form of the responses has not been shown to vary, but both the rate and the stimulus control are influenced by events occurring throughout the life of the individual.

3. Sea lions (Zalophus californianus) were conditioned to produce a "click" sound, never given previously in captivity, in the presence of a circular target (Schusterman and Feinstein, 1965). The sound was even used as a means of determining auditory thresholds (Schusterman et al., 1972).

4. Great apes that have been hand-reared from infancy have been taught to speak a few words of human speech in the proper context.

Thus, in some mammals we find a tight genotype-phenotype correlation in the form and pattern of species-characteristic sounds, but less so than with crickets, especially in relation to frequency and stimulus control.

Let us now turn to birds. References for the following summary of findings on birds can be found in Marler and Mundinger (1971) and Marler (1973):

1. Bird vocalizations can be divided into the longer "songs," occurring in territorial and mating behavior, especially by males, and the shorter "calls," used for alarm and flight. The calls are minimally modified by social isolation or cross-fostering with other species with different repertoires. This is true of all or most calls of eastern meadowlarks (Sternella magna), white-throats (Sylvia communis), song sparrows (Melospiza melodia), chaffinches (Fringilla coelebs), ring doves (Streptopelia risoria), and domestic chickens.

2. Rearing birds in complete auditory isolation led to some species' developing normally (chickens) and some abnormally (robins [Turdus migratorius] and grosbeaks [Pheucticus melanocephalus]: Konishi, 1965).

3. Doves and chickens deafened a day or two after birth still vocalized normally at maturity. This did not hold for robins and grosbeaks (Konishi, 1965).

4. Ducklings devocalized in the egg and hence unable to hear themselves still have normal preferences for maternal calls (Gottlieb, 1971a).

5. Arizona juncos (Junco phaeonotus) reared from four to five days of age in individual isolation develop a much simpler song than is normal. Hearing the adult conspecific song, however, was not necessary for developing the normal song, as young males reared with an age-mate developed songs that were indistinguishable from those of wild adults. In the chaffinch such joint experience only enhanced song complexity, while still leaving it abnormal.

6. The chaffinch will not develop the species-characteristic song unless it hears the adult conspecific song. But if exposed to a call slightly different from the normal song only a small amount of imitation will occur. Further, if the bird is exposed to both the normal and an alien song, it will selectively learn just the conspecific song. When the chaffinch is old enough to sing itself, at 300 days of age, its song is no longer capable of being changed by such experience and is virtually unmodifiable.

7. In the white-crowned sparrow (Zonotrichia leucophrys) brief exposure to the species-characteristic song for a three-week period (the "sensitive period") between the tenth and fiftieth day of life was all that was needed for normal song at the later appropriate time. Later experience had little effect (Marler, 1973). Exposure to only an alien call led to songs resembling those of social isolates.

8. Immelmann (1969), who deafened estrildid finches after the sensitive period, found that male zebra finches (Taeniopygia guttata castanotis) foster-reared by society finches (Lonchura striata) would later sing the foster parent song, even if conspecific calls were given in the area during the sensitive period. Hence the parental bond can be of importance in focusing attention on what is to be learned.

9. Minor, but consistent, population differences in the songs of white-crowned sparrows are due to exposure to those variants during the sensitive period.

10. Some birds (parrots, mynahs) readily imitate human words, calls of otheranimals, and even whistles as adults—sounds they presumably never heard earlier. In the case of whistling, the addition of food reward had no effect on the speed or quality of imitation (Foss, 1964).

11. The frequency of sounds normally part of the animal's repertoire can be increased or decreased through standard conditioning techniques. Chirping in domestic chicks is a good example (Lane, 1961).

In birds, in contrast to crickets and nonhuman mammals, a large variety of modifications can be made in some species, but generally the acquisition and modification of species-characteristic sounds are possible only under special circumstances or antecedent conditions. Bird vocalizations demonstrate the wide variety of experimental operations that may or may not affect the outcome. This shows us that a simple correlation of a behavioral phenotype with genotype or environment, while necessary, is not sufficient to understand ontogeny. It does not distinguish for us the myriad ways in which diverse geno-typic and environmental variables interact, and it fails to give us more than guideposts to the elucidation of how behavior arises in the development of the individual.

A Model of Life History

The genetic constitution of an organism remains constant throughout life, sort of an indelible serial number affixed at the factory (somatic mutations excepted). But from the moment egg and sperm unite, a complex interaction begins between the zygote's genotype and the diverse aspects of its external and internal environment. A dynamic system has begun, which will not end until death. But the types of processes that occur and their rates differ across species and individuals. Observation of animals developing, either prenatally or after birth, leads to descriptions of characteristic modes and levels of anatomical, neurological, and behavioral development. Cross-sectional (comparison of animals at different ages) and longitudinal (long-term studies on individual animals) studies are necessary to establish adequate descriptions and correlations. More loosely, references are made to stages or plateaus of development such as are found in embryological studies (e.g., blastula, type I motility) and in behavior studies (infant, juvenile). While the criteria used for distinguishing stages of behavioral ontogeny are often arbitrary, inconsistent, and not at all sharp, scientists often find them useful (e.g., Hinde, 1971a). The remarkable incorporation of Piagetian thought into all areas of child development seems primarily due to Piaget's codifying a plethora of stages and substages (Etienne, 1973). These useful, but often crude, classifications must remain only descriptive until they are tied to specific relationships.¹ The occurrence of discrete events is also important; hatching, birth, molting, pupating, and metamorphosis. In addition, many behaviors appear suddenly and seemingly "full blown," as is true of many fixed (modal) action patterns and complexes of seasonal behaviors like courtship or producing and caring for offspring. Prenatal studies are more often neurological or physiological, but that does not eliminate their relevance to later communicatory phenomena. A comprehensive but concise treatment of the well-studied prenatal behavior of the chicken (Oppenheim, 1974) is most useful in demonstrating the unification of descriptive and experimental approaches.

A vivid and accurate portrayal of ontogeny can be elaborated from Waddington's (1956) epigenetic landscape, in which a marble rolling down a hill with ridges and valleys represents the zygote from the moment of fertilization (the top of the hill) to the end of life. The width, depth, and steepness of the terrain represent the speed and plasticity of development, and the almost level areas represent stages. Further, the mass, size, and shape of the "marble" determine its interactions with the environmental topography. Using this model, we can appreciate that the narrower the valleys and the higher the ridges the more difficult it is for the marble to shift from one development pattern (valley) to another.² The ontogeny of animal vocalizations in different species, we have seen, incorporates many such widely varying "terrains."

Now let us consider the life cycle of an individual in relation to studying communication. From the descriptive information about the communicatory behavior of the adult of the species, we should be able to find the moment when a given aspect of behavior is first shown in its communicatory context or is synthesized out of already extant elements. We then either focus our attention on this first occurrence or move in two directions—earlier (which usually means studying different individuals of the same species) or later (which often involves longitudinal observations). If we look earlier, we aim at uncovering those antecedent conditions that led to the later functional behavior (conditions I shall call "prerequisites"). If we stay at the moment the behavior first occurred and the specific internal and external stimuli involved, we are studying the "corequisites." Much of the ethological sign stimulus and "Innate Releasing Mechanism" work is done at this point of first occurrence. If we look at what happens, or can happen, after the behavior is in the animal's repertoire, we are dealing with change or modification. Here we will deal not with the immediate causation or corequisite question but only with points earlier and later than the first occurrence. However, the corequisites can be viewed as being closely related to the prerequisites.

Students on the ontogeny of communication, then, look for the variables, prerequisites, or antecedent conditions that lead to the expression of species-characteristic communication in the individual at the proper time and situation (hereafter referred to as Question O for origins) or are capable of modifying, altering, or maintaining existing patterns of communication (hereafter referred to as Question M for modification ). This distinction between origins (first occurrence) and modification is important because processes involved in the developmental shaping of behavior may have little in common with those subsequently altering such behavior, but are often confounded in discussions. Studying the conditions that lead to the chaffinch's singing its species-characteristic song for the first time relates to Question O. Attempts to condition the frequency of chaffinch singing relate to Question M.

More on the Question of Origins: Prolegomenon to Precursors

The fact that a behavior occurs at a certain age does not mean we should no longer be concerned with its topography or function. First, it may not yet be in adult form. Longitudinal studies show that in jungle fowl (Gallus gallus) fighting develops out of hopping, which gradually, from one week of age, becomes directed toward other chicks and subsequently incorporates pecking and kicking (Kruijt, 1964). Similar processes of the phenotypic appearance of ritualized behavior have been shown in the facial expressions of canids (Fox, 1969). Second, even if the behavior is in its adult form and context, various refinements may be seen over time. Questions O and M, then, are specific to the age and response selected. Most studies that fall under Question M (modification) deal with changing the frequency, duration, or stimulus control of a response that is already in its adult form, as in the barking of dogs (Salzinger, 1973).

The answering of Question O (origins) has suffered most from confusion dealing with the nature of the antecedent conditions leading to the communicatory act. A modification and extension of Gottlieb's (1973) discussion of "precursors" in behavioral embryology is useful here in characterizing the problems of postnatal as well as prenatal origins (Burghardt, 1975). Recall the opening question of this chapter: "How does an individual animal arrive at the condition in which he or she initiates and responds appropriately to communicatory signals?"

There is a grouping of precursors that influence the pace of ontogeny temporally or quantitatively: that hasten, retard, or prevent the development in the individual of its species-characteristic patterns of behavior. Such retardation can take the extreme form of completely abolishing certain abilities ("atrophy"). For instance, raising an animal in the dark may irreversibly destroy pattern perception. If the species involved is one in which visual cues play an important role in mate selection, it is clear that the animal may show deficiencies or even be an abject failure in courtship and the communication involved therein. Exposure to patterned light is thus a "facilitative precursor" in this hypothetical example and perhaps a maintenance factor too. But such evidence does not tell us why the species comes to communicate in one way and not another. It is the "determinative precursors" that actually "force or channel neurobe-havioral development in one direction rather than another" (Gottlieb, 1973:6). They are the source of the "adaptive information" contained in the behavior (using the terminology of Lorenz and Thorpe). With facilitative precursors all we can say is that relatively nonspecific experience or stimulation of a certain kind is necessary (prerequisite) for certain abilities to develop and manifest themselves at the appropriate time. They are often likely to affect several behavior patterns. Note that in the courtship example both patterned light perception and courtship behavior are involved, and that a given event may have a determinative, facilitative, or maintenance effect depending upon the aspect of behavior being considered. Bateson (1976) discusses "specific" and "non-specific" determinants, both inherited and environmental, which may affect any given behavior pattern.

Gottlieb (1971a) demonstrated a somewhat specific prenatal facilitative precursor. He established that newly-hatched peking ducklings (Anas platyrhynchos) selectively approach the species-characteristic maternal call when simultaneously presented with the maternal call of related species. Even prior to hatching the ducklings selectively respond to the maternal call, as ingeniously measured by bill clapping and heart rate. But embryos that could hear the sounds made by other embryos showed a faster development than sound-isolated embryos. Nonetheless, postnatal behavior was unchanged in spite of the different history of occurrence. Ducklings auditorily isolated and devocalized in the egg, to prevent an embryo from even hearing itself, showed decreased discrimination after hatching; but they could still discriminate between mallard and wood duck maternal calls. However, at 48 hours discrimination between the mallard maternal call and the more similar chicken maternal call was seemingly abolished, although some improvement was noted later. Thus, we see that Gottlieb's analysis is an elegant example of what seems beyond quarrel: "Exposure to normally occurring stimulation helps to regulate the time of appearance of the perfected response, as well as the latency of the perfected response" (1971a: 134).

The danger of considering behavior a unitary phenotype is made evident by the concept of facilitative precursors.³ For example, a child's temper tantrum is a complex sequence of behavior that is all too perfect the first time it is performed. Hebb claims that to call the behavior instinctive or unlearned is wrong because "it is not possible without the learning required for the development of purposive behavior" (1953:43). He also claims that the dramatic onset of fear of strangers in a chimpanzee baby "is not learned: but it is definitely a product of learning, in part, for it does not occur until the chimpanzee has learned to recognize his caretakers" (ibid.). Such confusion and subsequent retreat to the "everything is everything" view are due to the error of considering a behavior as a unitary whole and not treating as separate questions which aspects (topography, rate of occurrence, etc.) are due to what. In the two examples above the topography of movements is not determined by the hypothesized prior learning, while the elicitation of the behavior definitely seems to be, that is, the latter is facilitated by a certain type of experience.

Facilitative precursors may either be necessary (but insufficient) for the manifestation of a behavior or, by their absence, cause deterioration of an already present (but yet undemonstrated) capability. The latter process occurs in cortical neurons that respond to specific types of visual cues in cats (see Blakemore, 1973). The indirect and subtle effects of many kinds of experience may thus influence behavioral development. Gross sensory and motor abilities, as well as certain experiences, may obviously be prerequisite for the manifestation of the complex and specific behavior and perception involved in communication.

Facilitative prerequisites based on sensory stimulation in the embryo seem much easier to demonstrate prenatally than determinative ones based on stimulation (Gottlieb, 1973). Prenatally, genes and hormones have been shown to be determinative. Postnatally, a variety of other processes may be at work (conditioning, imprinting) but often it is not easily shown that they are determinative. This contrasts with the expressed hopes of theorists such as Schneirla (1965) and Moltz (1965) that the environment is "actively implicated in determining the very structure and organization of each response system" (Moltz, 1965:44). In fact, most studies that have been presented as disproving traditional conceptions of innate behavior or instinct unfortunately do no more than prove the necessity of certain facilitative or maintenance precursors for normal behavioral ontogeny. Rarely do they treat the essential question of determinative precursors. Studies on Question M (modification) are even less related to determinative factors. Classical ethologists (e.g., Lorenz, 1965; Thorpe, 1974) and evolutionary theorists (e.g., Mayr, 1974) clearly view the determinative factors underlying communication as being either ultimately genetic or shaped strongly by evolved mechanisms. Too many authors limit nongenetic determinative precursors to "reinforcement learning" or a similar unitary process (e.g., Dawkins, 1968). A summarizing diagram sets forth the relationships developed here (Fig. 1).

Some Antecedents and Modifiers of Communication

In the early stages of development of knowledge in any field the scientist does not know all of the variables entering into a particular set of phenomena, and he must, therefore, hypothesize or hazard guesses as to what these unknown factors might be. [Spence, 1953:284]

Some Theoretical Issues

Fig. 1. An outline of the relations between genotype, phenotype, and the questions about antecedent conditions responsible for a given communicatory event. A typical but simplified ontogenetic sequence is incorporated into the diagram, which relates to the following generation. Mating feeds back to the genotype, and parental care to the phenotype.

It must be emphasized that the terms "factor," "precursor," or "variable" as applied to processes affecting communication are crude and oversimplified. Factors are often sets of heterogeneous mechanisms that are only superficially similar, while differences between the various factors may themselves be superficial. Here we will discuss some of these antecedent conditions. In actuality, any given act of communication can be affected by several of them. Most can conceivably act as facilitative precursors, determinative precursors, maintenance factors, or modifiers. Following is a catalogue of what seem to be the most debated and studied types of precursor and modifier.

GENOTYPE

Genetic differences are often associated with differences in communicatory behavior and hence survival and evolution, as inthe cricket examples. Manning (1971) and Gould (1974) give excellent reviews of genetic mechanisms affecting behavior, with many examples from the social realm. But it must be remembered that any given aspect of adult behavior or structure is usually due to the action of several to many hundreds of different genes (alleles). This contrasts with the one-or two-gene models derived from simple Mendelian heredity, with clear-cut dominant and recessive factors. Similarly, a change in one or a few genes may affect several characters in the adult (pleiotropy). An especially pertinent case familiar to cat fanciers is the Siamese cat, in which a recessive gene yields individuals showing both a specific pigmentation and a distinctive vocalization with equal conspicuousness. Recent studies (Benzer, 1973) with Drosophila demonstrate that precise localization of behavior patterns to specific mutations can be attained.

Genotype is obviously prerequisite to the organism and hence to all communication, and in this sense Ginsburg (1958:404) is correct in stating that "all aspects of an organism may be thought of as 100 percent genetic but not 100 percent determined," genetically determined that is. But genotype is often a determinative factor in, for example, courtship communication, which can differ greatly between closely related species (e.g., crickets, ranid frogs). Such "isolating mechanisms" prevent similar species, with overlapping ranges, from mating successfully. However, most genetic studies necessarily focus on the often much less dramatic differences seen within or between populations of the same species. An exception is the discovery that different strains of mice refract identical experiences in opposite ways, becoming more or less aggressive, while still other strains are not affected at all (Ginsburg, 1966). This example also illustrates that genetic factors are prerequisite to and influence the modifiability of behavior ("learning")—a standard, if only newly rediscovered, point in many recent texts and symposia. But how genotype affects modifiability of any behavior, least of all communication, is too rarely studied.

DEPRIVATION OR AUGMENTATION

As an alternative to genetic experiments we may find that animals reared in complete social isolation (Kasper Hauser on deprivation experiments) show the normal repertoires of signaling behaviors and can respond properly to signals from conspecifics (Lorenz, 1965). These studies show that whatever events in the ontogeny of the individual may be necessary for the origin of these behaviors, they do not include commerce with other members of the species. Of course, abnormal rearing conditions are highly destructive where such normal experience is needed (facilitative or determinative) for critical environmental input. Rhesus monkeys are frequently cited as cases in which social deprivation can have an adverse effect on later social behavior (Harlow and Harlow, 1965), but one should remember the magnitude of the abnormal conditions necessary to produce such an effect. Even in rhesus monkeys, however, isolation-reared individuals recognize visual signals, such as threat (Sackett, 1966). When communication is unimpaired in spite of postnatal social deprivation, we can state that the determinative factor for information contained in the display is probably ultimately tied to the genotype and that its manifestation (phenotype) is impervious to certain kinds of experience. A distinction must be made between the performance of the display and the response to the display. These may be differentially affected by experience and deprivation. The commonly used term "expression" and the less-common companion term "impression" also get at the distinction (Leyhausen, 1973).

Prenatal facilitative factorsmay be involved but, as we have seen, evidencefor experiential prenatal determinative factorsis sparse indeed. As the bird examples show, thegenetic variable can be manifested in variousways, ranging from a predisposition to learn the call exposed to in the nest (finch), through a predisposition to learn only the species-characteristic call (chaffinch), or to a mainly genotypic determinative factor that seems virtually impervious to all kinds of deprivation or augmented experience as well (chicken).

A final example will underline the importance of distinguishing facilitative and determinative precursors. In the oft-quoted controversy on nest building in naive virgin rats (see review in Burghardt, 1973), Riess (1954) argued, in effect, that prior manipulatory experience was the source of origin of the movements involved, as his rats reared without the opportunity to manipulate objects did not build nests. Eibl-Eibesfeldt's (1961) elegant demolishing of Reiss's position by taking into account the home environment only succeeded in bringing out another facilitative factor. Even if Eibl-Eibesfeldt's conditions had not led to nest building, we would still not know how the nest-building movements of rats originate ontogenetically.

EMBRYOLOGICAL FACTORS

Caspari (1971) cites the importance of differentiation. All the varied cells and structures of the adult organism begin from a fertilized egg with one combined set (from both parents) of genetic instructions. Yet while early cells seem to look the same, all the adult cells, regardless of their differences, still contain the same genes and therefore the same information. As Caspari (1971:3) puts it: "How do cells which contain the same genetic information assume a number of different biochemical and morphological states?"

In addition to differentiation, we also need to be concerned with "morphogenesis"—how and why differentiation occurs among groups of cells at different times and locations. Why does one sensory system develop before another (Gottlieb, 1971b)? How does a leg form differently from an arm? Although we will not dwell on such issues here, it is interesting to note that Caspari claims that the biological principles underlying differentiation and morphogenesis are sufficient to explain the development of behavior (see also McClearn and DeFries, 1973). Since most scientists consider the essence of communication to be behavior, the implication concerning the ontogeny of communication is obvious. But such an oversimplistic approach may lead to an ignoring of the special complexities and diversities found in the behavior of interacting organisms.

MATURATION

The question of origins often revolves around the issue of maturation. Animals and their behavior patterns change with age, and an older individual often evidences behavior not seen in the embryo or the neonate (newly born or hatched animal). But controversial issues involve the intervening steps between DNA and behavior: the influences, sources of information and variation, and reciprocal interactions that are responsible for an observed change in behavior. The phenotype changes, but the genotype does not. Is behavior that is seen in older organisms but not in the neonate, such as courtship, due only to maturational processes of the nervous system, physiology, and structure involving growth, proliferation, migration, and differentiation (Gottlieb, 1973)? If so, then the genotype-phenotype correlation is high given a minimal environment for survival. This phenomenon, called "maturation," is often posited as a factor on the basis of deprivation experiments. Calling in male crickets is as good an example as one could want. But while maturation is undoubtedly an important antecedent process in behavior, it is clearly not the only one even incrickets, and analyses then become snagged on conceptions of "experience" and "learning." The"closed genetic program" of Mayr (1974) closely resembles maturation. Because of the biological value of rapid and unambiguous species recognition, he concluded that communication behavior is much more closed than behavior patterns involving food or habitat, which involve ecological conditions that are muchless stable and predictable for the species than are the appearance and behavior of conspecifics.

FUNCTION

While the function-structure issue can relate to broad evolutionary questions—such as whether a distinctive wing patch arose before or after the behavior of flashing that area of the wing (e.g., as shown by mockingbirds), its specific current referent is in ontogeny. During maturation, what facilitative or determinative role for the later mature system is played by the functioning of the immature system? Does structure lead function, or can function influence structure even while maturation is taking place? In our formulation, "mature system" refers to the communicatory act (Fig. 1), but most of the current lively debate on this issue is played in the embryological theater. There the issue deals with the effects of use or disuse of muscles or sensory stimulation on neurobehavioral development (see papers in Gottlieb, ed., 1973) and is the homology of the "nonspecific experience" issue in postnatal organisms.

The validity of the structure-function dichotomy seems meaningful to people on either side of the issue. But is it? As soon as we move out of the embryo, we have ample evidence that environmental events can facilitate or determine aspects of developing (maturing) systems, as in imprinting. Rather than allude to an almost mystical "function" (structure can similarly be unwisely invoked) we should instead isolate and investigate the role of the various possible precursors. One would certainly expect to find a range of genotype-phenotype correlations. While the experimental evidence may be in doubt in a given case, use of the framework outlined here precludes broad function-structure debates on inappropriate levels.

HORMONES

Secretions of hormones can determine the course of behavior development, most conspicuously in sex-related behavior, but elsewhere as well. Consequently, hormones may be determinative precursors. Many experiments have shown that castration and/or supplemental injections of either male or female hormones can lead to major changes in sexually dimorphic behavior in many and diverse vertebrates (Young, 1965). Normally, phenotypic sex is virtually always the same as genetic sex, but experimental work shows that hormone manipulation is one of the few known cases where a genetic determinative precursor can be qualitatively overridden (not just suppressed). The hormonal organizing effect may be limited to a certain period in the animal's life.

SENSITIVE PERIODS

In a number of communicatory phenomena, including classical imprinting, song learning, and hormonal organization, a limited time span is most important in the establishment or modification of certain modes of response or perception. These may be measured in hours, as in filial imprinting in ducks; minutes, as in maternal imprinting in goats; or weeks or months, as in sexual imprinting or song learning. Sensitive periods differ in their "criticality" and other attributes. Hess (1973) documents many examples and presents a convenient tripartite classification that helps to order the diversity. The examples of bird song learning clearly show the importance of sensitive periods.

It is useful to recognize two broad functions of sensitive periods. In one, the animal acquires new information necessary or useful for survival (e.g., species-characteristic song). In the other, experience during the sensitive period is necessary for abilities already prepotent or maturationally dependent to be expressed, perfected, or maintained (e.g., sensory effects). These functions can be combined, but often one or the other predominates. In other words, sensitive periods can be involved as both determinative and facilitative precursors.

ASSOCIATION

Association or contiguity refers to the pairing, or almost simultaneous occurrence, in time and place of two events. Obviously many of the experiential precursors to communication involve association, as in imprinting and song learning. Psychologists, unfortunately, have sought to elucidate the mechanism of association by elaboration of a highly specialized, abstract, and behaviorally impoverished procedure. If one of the two events (e.g., food) elicits a certain behavior (e.g., salivation) and the second (such as sound) normally does not, the second may begin to act as an effective stimulus even in the absence of the first. Such is the paradigm of classical or respondent conditioning (Type I or Type S).

While classical conditioning has assumed great theoretical importance in psychology, it has been considered much too narrowly, especially in relation to communication, and has mainly been applied to the modification of existing noncommunicatory processes. An exception is the history of attempts over more than forty years to condition the vocalizations of rats when shocked with electricity, which is nicely reviewed by David and Hubbard (1973). Successful conditioning is related to the animal's mobility and the use of a variable inter-trial interval. Thus, far too often, the "principles" emerging from psychological research on conditioning are artifacts of the particular events chosen for juxtaposition and of the convenient species whose members are studied. What else can we say about the history of attempts to demonstrate "conditioning" in embryos by the rigid application of an arbitrary paradigm? We know now that classical conditioning cannot override biological constraints on what stimuli can be associated with what responses, and that close temporal contiguity of UCS (unconditioned stimulus) and CS (conditioned stimulus) is not always necessary (see Hinde and Stevenson-Hinde, 1973; Seligman and Hager, 1972). Principles based on prenatal association have been invoked in ingenious (but empirically unsupported) ways by Schneirla (1965) in his approach-withdrawal theory of behavioral ontogeny.

REINFORCEMENT AND PUNISHMENT

Reinforcement (reward) refers to the presentation or withdrawal of events (e.g., food or shock, respectively) that lead to an increased frequency or probability of a response. Punishment refers to the presentation to the animal of an event (usually shock) that decreases responding. The animal must first respond, then a consequence follows, which increases or decreases the probability of its performing the behavior again under the same or similar circumstances. Skinner (1966) has drawn a parallel between reinforcement and natural selection, the former directly altering behavior patterns, and the latter the genotype. As with the link between association and classical conditioning, reinforcement is too often discussed only in the context of instrumental or operant conditioning, thus implying a certain ubiquitous procedure for studying it. Another danger is that all changes in behavior can be attributed to the action of some reinforcer somewhere. The recent views of Premack (1965) are helping us to realize that reinforcers are, in fact, responses, andhe thus broadens the classical ethologists' view that the performance of fixed (modal) action patterns is reinforcing.

It is certainly true that rewards, such as food or water, and punishment, such as electric shock, can change the frequency and situation-specificity of behaviors, as the examples of chirping in chicks and barking in dogs suggest. Also, aggressive behavior in a male chicken can be eliminated quickly if he is shocked each time he attacks another chicken (Radlow, Hale, and Smith, 1958), while Thompson (1964) showed that the visual image of a rooster, in the form of a second animal or a mirror, would reinforce a key pecking response in an operant situation. Thompson (1966) also demonstrated parallels between ethological "releasers" and operant "rein-forcers" in that the former can act as the latter in Siamese fighting fish (Betta splendens). Melvin and Anson (1969) showed not only that male fighting fish would greatly increase their frequency of swimming through a narrow aperture when reinforced with a mirror to which they could display, but also that moderate electric shock punishment facilitated rather than suppressed the operant response. Such studies bring home the fact that the organism and the response need to be considered as important as the reinforcer.

But all such studies manipulate some aspects of the performance of the behavior and only deal with modification, usually with respect to frequency or stimulus control. They simply do not face the problem of the ontogenetic and phylogenetic origin or "roots" of the behavior (Question O). Some of the events that prove positively or negatively reinforcing to the individual prove to be both species-characteristic and response-specific.

There are two ways in which reinforcement may be involved as determinative precursors of communication. The traditional way that complex communication systems could be established, compatible with a strict application of reinforcement principles, is through "shaping." Here successive approximations to the desired final behavior are rewarded until what appears as qualitatively new responses and/or stimulus control are shown. While this view is elegant, its problems can easily be summarized: There is no direct evidence that this order of events occurs with respect to the normal ontogeny of communication in any species except, perhaps, in language in humans. A second and more recent approach to origins can be found in the concept of "autoshaping." Rats and pigeons exposed to stimuli and reinforcers noncontingently soon begin to respond to the stimulus in the usual "shaped" fashion (Hearst and Jenkins, 1974).

Self-reinforcement, however, may be an important factor in the ontogeny of communication. For instance, ducklings may quickly learn to depress a pedal to self-imprint in response to a flashing rotating light (Bateson and Reese, 1969), and song learning in some birds also involves the reward (or feedback) of hearing oneself sing.

Dimond (1970) has stayed close to traditional operant conditioning while emphasizing the action of reinforcement and other learning processes in animal social interactions and the answering of Question O. His premise is that "much of the social behavior must relate to the fact that one animal finds the presence of another animal rewarding. Within this system must surely lie the roots of gregariousness: aggregation, flock formation, and social grouping" (1970:118). He then suggests a framework involving reward systems of food, warmth, comfort, and shelter that interact with early experience to form the appropriate associations. Such processes may occur with mammals and birds, but it is as difficult to rule out "unlearned" social attractions as to prove them. In newborn snakes (Thamnophis, Storeria, Natrix) we now know that species-specific aggregations occur immediately after birth in the absence of variables involving food, warmth, or parental care (Burghardt, in press).

Dimond goes on to state, "Each animal learns to communicate with others. The suggestion arises that a sophisticated behavioral language is learnt from the earliest period of the neonate's existence . . . each animal acts to the other as an important communication source" (1970:118). Dimond accepts social contact as initially rewarding because it allows the animal to learn the communication system of its species; he implies that the social attractiveness is due to conditioning of primary biological needs (e.g., a secondary rein-forcer). In fact, little ontogenetic work has actually been done on the role of such "social learning." A much-lauded study that concluded that conditioning was a determinative factor in regurgitation feeding of squabs by parent ring doves (Lehrman, 1955) surprisingly omitted a crucial ontogenetic variable—age of squab (Klinghammer and Hess, 1964). Thus, while the methods and precision of experimental psychology will be increasingly useful in the study of communication, liberation from narrow, theoretically oriented approaches seems essential if learning psychologists are to make substantive contributions.

HABITUATION

Habituation is considered the simplest and most basic form of learning: unpunished learning not to respond. Consider the gaping response of a young bird to cues associated with the arrival at the nest of a parent bringing food. If one presents those cues (such as branch shaking) often enough, the gaping response will wane if food is not forthcoming (Prechtl, 1953). Such habituation is very common and certainly seems relevant to communicatory signals, such as "shaping" through disuse. Some responses, once habituated, stay that way while others recover quickly. Sensory and motor factors need to be controlled. Further, in some short term cases, the opposite to habituation, sensitization, occurs, in which previously ineffective stimuli become effective through generalization. Habituation can also be viewed as a broad category, which includes extinction of conditioned behavior.

As Denny and Ratner (1970) point out, it seems easier to habituate preliminary appetitive behaviors (as in using a certain food-searching behavior) than consummatory ones. Communicatory gestures and responses to them are clearly subject to habituation. For instance, threat displays of male Siamese fighting fish habituate to live male fish as well as to mirror images (Baenninger, 1966; Baenninger, Bergman, and Baenninger, 1969), although exposure to a mirror can also act as a reinforcer in the acquisition of a new response (Thompson, 1966). Unfortunately, there is little information on habituation in actual social interactions, although it is undoubtedly involved in changes in intensity of courtship behavior or maternal care occurring over time. But there are too many other changes superimposed on any habituation effect to tell a clear story. In the ontogeny of communication it likewise seems clear that habituation should play a role and it is similarly unestablished.

OBSERVATIONAL LEARNING AND IMITATION

This process and the following two are controversial, but only in the sense that the mechanisms involved may or may not be basically different from the foregoing three, which psychologists have generally considered most basic. They have been largely ignored by ethologists and psychologists, but are included here to point out areas where some new thinking and methods may offer insights about communication (see also Griffin, this volume and 1976).

When a naive animal observes another engaging in some activity and later performs the same behavior in the absence of the model, some acquisition process is involved. Mainardi and Pasquali (1968) have demonstrated this occurrence in mice. Traditionally it has been called imitation, with emphasis on problem solving: The subject observes a trained animal choosing the "right" stimulus or manipulating a lock correctly to get at food. Unfortunately, while this situation seems very common and obvious in people, the evidence in animals with arbitrary responses is controversial. One area of communication where it has been clearly established is in the vocal learning of songbirds discussed earlier. The fact that a white-crowned sparrow will show the song "dialect" it was exposed to during rearing leads us to conclude that certain "traditions" can be passed from one individual to another through social interaction (Wickler, 1965). This is especially true for responses to nonsocial objects such as food. Galef (1976) has shown that adult rats experiencing a sublethal dose of poison after eating highly palatable food can transmit an aversion for this food to offspring who have never experienced any illness because of eating it. Here communicatory processes between parents and young are important in modifying food preferences in early ontogeny, although classical imitation does not seem involved. Imitation of unnatural gestures used in sign language communication in chimpanzees has been demonstrated (see Fouts, volume 2), but the relevance to development of normal communication has yet to be demonstrated.

PERCEPTUAL LEARNING AND PRACTICE

Mere repeated exposure to an array of highly similar, but not identical, objects can lead to discrimination without the operation of traditional differential pairing or differential reinforcement. When students begin to watch a social group of animals, such as baboons, they want to mark them individually for easy recognition. But that type of aid soon becomes unnecessary as subtle differences in size, proportion, natural markings, scars, and behavior become clear. While reinforcement through information feedback is also involved, the case can be made that such reinforcement is not necessary (Hebb, 1972). Perceptual learning differs from the nonspecific facilitative precursors discussed earlier in that it does not refer to the experience necessary for the development of sensory abilities (e.g., acuity, pattern, odor, tones) but to the use of these abilities in making finer discriminations (cf. search image; Krebs, 1973). Practice is the motoric equivalent of perceptual learning.

Perceptual learning is of greater importance in socially stratified animals than is usually recognized and could be the basis for individual recognition. For instance, gull chicks quickly learn to recognize subtle differences in their own parents' calls. Indeed, exposure to the calls prenatally leads to discrimination (Impekoven and Gold, 1973). Clearly, an animal reared in a social group has ample opportunity for such perceptual learning. For example, as soon as a female northern elephant seal (Mirounga angustirostris) gives birth she emits a "warble" vocalization, which is answered by a distinctive sound by the pup. Such duets can occur throughout the nursing period (Le Boeuf et al., 1972). The phenomena of "selective attention" (Chance and Jolly, 1970) and "local enhancement" may facilitate perceptual learning and imitation by causing the animal to be attracted to important learning situations.

INSIGHT

A controversial legacy of the Gestalt psychologists, "insight" refers to the sudden apprehension of certain relationships that allows the animal to do something it previously struggled to accomplish rather unproductively. Chimps abruptly learn to stack up boxes to acquire food formerly out of reach (Köhler, 1927). Less dramatic are all the related instances of rapid learning after a period of poor responding. In fact, some psychologists argue that all learning takes place in discrete steps. Curiously, classical insight shares several attributes with instinct in that an integrated adaptive response is made without prior specific experience. But the behavior does seem purposeful because the animal uses it to gain an immediate goal or reward, while instinctive responses are performed for their own sake and only indirectly lead to adaptive results (e.g., ingestion of food, offspring).

The question is how much influence such innovative behavior has on communication and ontogeny. On one level it probably has very little effect. Even in mammals and birds it appears too variable, infrequent, and unstable to be a basis for the development of the ritualized communication involved in such fundamental and critical duties as courtship, maternal-offspring interaction, and defense. In developmental studies it is also difficult to separate insightful factors involved in the transition from one stage to another with that caused by "readiness" or "preparedness" due to maturation or prior experience, which may facilitate or even trigger such sudden progress.

PLAY

In recent years many investigators have turned to studying the rather poorly understood phenomena jointly termed play. Since it is generally found exclusively or primarily in infants and juveniles it is clearly ontogenetically relevant. We still do not know what play is, but we know it is important. But important for what? To develop communication gestures and skills, say some primate workers (Jolly, 1972; Mason, 1965), although several other functions of play have been mentioned in the literature, for example, to develop predatory skills. A stimulating symposium on play in mammals has recently been published (Amer. Zool. 1974), and some findings discussed there illustrate the problems in determining the relations between ontogeny and communication.

Animals reared in social isolation that, perforce, do not play, often show deficiencies in the patterning of later adult communicatory behaviors and in the "understanding" of the responses of others. Such evidence has been used to argue for the definitive importance of play in the later behavior of adults. Aside from the fact that social isolation deprives an animal of more than just play experience, it is not possible to state from the evidence whether play is a determinative or merely a facilitative factor in later communication. Play may exert primarily a refining influence in terms of response- and stimulus-specificity, patterning, and so forth. Evidence for the former would entail experiments such as this: Foster-rear an animal in a family of a related species with a different communicatory repertoire. If the foster animal later shows characteristics associated with the foster species, then to that extent play may be determinative. Note, however, the possible confusion with imitation or observation. To separate the two might entail prevention of physical interaction. Bekoff's retraction of earlier views on the necessity of play (facilitative or determinative) to an animal "in order to acquire species-typical social skills" (1974:337) is a recognition of the difficulties involved (see also Symons, 1974).

Detailed descriptive studies of the normal ontogeny of play are necessary prior to either speculation or gross experiments on the causal significance of play to later behavior. Some developmental studies of note involving monkeys, bears, canids, sea lions, ferrets, and meercats are presented in the above mentioned symposium.

The Interaction of Several Precursors

If. . . the system under study is a complex one involving a relatively large combination of interacting variables, the discovery of what the relevant factors are and the nature of the relations holding between them involves much more [theorizing]. . . . Theorizing at this stage, then, consists primarily in guesses as to what the unknown relevant factors might be and guesses as to the form the relations between them might take. [Spence, 1953:284]

The organism should not be viewed as an aggregation of precursors, but as a dynamic system with a history, a history that influences, to varying degrees, the individual's future commerce with the world. While environmental events in the ecological theater can have critical short-term individual as well as long-term evolutionary effects, internal control system properties involving feedback, maturation, self-reinforcement, and other self-correcting effects must not be ignored. There are three broad phenomena, touched on throughout this chapter, which explicitly involve several antecedents in a determinative fashion and bring out, to varying degrees, these self-adjusting mechanisms.

INSTINCT-CONDITIONING INTERCALATION

Early ethological theory held that much social communication of animals was based on releasers, fixed (modal) action patterns, and their reciprocal interactions. A morphological structure or behavior was the signal to a second individual, who responded with a behavior that either acted as a releaser or made prominent a structure (such as a colored feather), that was the releaser. A very useful way of viewing communication, this scheme made terminology clear and comparisons illuminating. The units of analysis were usually stereotyped, species-characteristic, and largely based on genetic differences and similarities. The signals, responses, and the code were prematurely viewed as innately determined and impervious to normal and even much abnormal experience (Lorenz, 1937).

The ethologists who formulated this theory in the 1930s and 1940s did not rule out learning as an important factor. The use of environmental factors, however, was clearly influenced by the conditioning approach then current in psychology. The ethologists knew that any sequence of behavior did not exist as a unit but as a series of apparently discrete parts. With the fixed (modal) action pattern (FAP) this is still controversial (Moltz, 1965; Barlow, this volume). But experience was often needed for optimal performance as indicated by the following evidence: (1) Deprived animals would often perform the behavioral units, but unrefined, or the arrangement would be out of sequence, as in nest building in virgin rats deprived of manipulatory experience (Eibl-Eibesfeldt, 1961). (2) Certain patterns would be entirely absent. This is especially true of orientation movements. For instance, social play is important for polecats (Eibl-Eibesfeldt, 1970) to direct the neck bite properly in mating and prédation. (3) The releasers for social behavior could be directed normally to a small subset of potentially effective stimuli through specific early experience (e.g., imprinting, which will be discussed separately). (4) Some behavior patterns dropped out with repeated stimulation (habituation, discussed above).

Lorenz (1935) initially formulated the idea of instinct-conditioning intercalation to handle seemingly "blank" spaces in behavior sequences that could be filled in through conditioning or imprinting. Today, of course, we know that experience can have subtle effects and do more than "fill in the blanks." Indeed, fixed (modal) action patterns and releasers operate by diverse mechanisms and vary greatly in their parameters. Studies tracing the development of specific behavior patterns are needed, and too few of these exist (e.g., Kruijt, 1964; Williams, 1972). But we do know that even specific aspects of a behavior may be ontogenetically shaped through the involvement of several precursors.

IMPRINTING

Imprinting is the process whereby an animal attaches behavior to a subclass of potential stimuli in a manner seemingly different from the normal conditioning paradigm. As originally defined it was restricted to maternal and sexual objects (classical imprinting), but phenomena possessing some similar operational characteristics also seem involved in food choice, habitat and nest site selection, song learning and other behaviors (Burghardt, 1973). Nonetheless, classical imprinting is not a unitary concept, and even in birds the degree and mechanism of imprinting varies enormously. Recent reviews of imprinting are available and should be contacted for a thorough introduction to this complex area, which has gathered more experimental attention than any other problem in ethology (e.g., Hess, 1973; Sluckin, 1965, 1972; Bateson, 1971; Immelmann, 1972).

In classical imprinting the experiential determinative effect concentrates entirely on the releaser or perceptual side of behavior. When ducks imprint, it is to the object of filial or sexual behavior, not to the way the motor movements are manifested, which seem largely due to maturation.

Imprinting is important to communication because the animal can respond with filial or courtship behavior to signals emanating from distinctly unnatural objects. On the other hand, recent work indicates that imprinting is often strongest to objects that possess or accentuate some characteristic found in the natural parent or mate. Imprinting, therefore, is a phenomenon that combines several precursors, including genotype, association, self-reinforcement, and sensitive periods.

TEMPLATES

The classical notion of the Innate Releasing Mechanism (IRM) arose from the idea of an internal schema that, when matched, released the "proper" response. Today we know that aspects of such releasers, and hence releasing mechanisms or schemata, vary from tight to loose genotype-phenotype correlations, from innate to largely acquired, from rigid to easily modifiable, or from genetically closed to genetically open (e.g., Hailman, 1967; Schleidt, 1962).

Vocalizations, while often signals, are nonetheless motor responses of the signaler. The findings on bird songs, summarized earlier, allow us to see that several ontogenetic factors, including genotype, sensitive periods, auditory feedback, and conditioning, can be involved. These findings led to the template concept, a way of accounting for diverse findings by a hypothetical mechanism. A template is

visualized as a mechanism or mechanisms residing at one or more locations in the auditory pathway which provide a model to which the bird can match feedback from its vocalizations. . . . In some species, such as the song sparrow, an individual raised in complete isolation from species members seems to possess a sufficiently well-specified template so that it can generate normal song, as long as the individual can hear itself. . . . However in other species, such as the white-crowned sparrow or the chaffinch, the initial specifications for the template are less complete. [Marler and Mundinger, 1971:428-29]

While the template model seems restricted to song learning, there is no reason why it cannot be much more widespread factor in behavioral development, as Marler (1973) himself postulates. However, it will be difficult to uncover such mechanisms in species that do not learn or perform such precise and relatively easily analyzed behaviors as songs. A major problem with the template idea is that it is the result of several experimental operations, each leading to somewhat different results depending on the species, age, sounds used, and so forth. The question is whether the template idea will actually allow us to make predictions or restrain us at the level of a summarizing concept that is tied closely to specific results.

1. The proliferation of developmental stages parallels that of "instincts" as carried out by McDougall's more enthusiastic followers (see Bernard, 1924). Doubtless it will lead to a corresponding, unwarranted discrediting of Piaget's constructive aspects in years to come.

2. W. S. Vcr plain k suggested to me the analogy of a snowball rolling down a snow covered landscape. rapidly increasing in size and weight Thus changes in the marble occur which influence subsequent interaction with the environment Taking this even further, large boulders and trees could play the obvious role of hazards.

3. Facilitative precursors as defined here are similar to the so-called nonspecific factors, used by some writers to discredit "instinctivists" by showing that in theory and practicesome manipulation of organism or environment can abolish or alter any and all "innate" responses. Use problems in distinguishing specific from nonspecific factors are almost insurmounlable, but may be of heuristic value (see also Bateson, 1976).