In 1988, Byrne and Whiten coined the phrase Machiavellian intelligence to portray primate intelligence as geared primarily to the sorts of conniving we ascribe to Machiavelli — deceit, cunning, and other manipulative, self-serving tactics — in short, to navigating the social, not the physical world. As they point out in this second edition (Chapter 1), intelligence honed for sociality may employ strategies for gaining social advantage beyond self-serving social manoeuvres, such as exploiting others' expertise. Exploiting expertise has many faces, for many reasons. It can aim at varied targets – expertise itself, such as knowledge or skills, or the products of expertise, especially resources such as food. There are two sides to the story of exploitation, as there are with most social stories – the exploiter's and the exploited's. Exploiters come in many guises, from learners and partners to bullies and thieves. So do the exploited, from co-operative, supportive teachers, sharing partners and tolerant mothers, to unwilling, niggardly hoarders or even neutral, naive dupes; and their responses can affect the nature and success of the ploys exploiters use. All these sides of the story suggest a number of broad factors behind primates' efforts at exploiting others (see Table 7.1). Exploitation may also vary within the primates because of differences in intellectual capacity between primate species, notably between monkeys and great apes. This chapter explores some of the ways in which non-human primates exploit one another's expertise and how various tactical roles come into play when they do.

Exploiting others’ expertise

Exploiting others’ knowledge entails the social transfer of knowledge and skills.

In a ‘protected threat’, a baboon induces a dominant member of its group to attack a third one. The baboon appeases the dominant member whom it uses as a tool to threaten the target and manoeuvres to prevent the target from doing the same (Kummer, 1988). This ‘social tool use’ is mastered by baboons at puberty, whereas chimpanzees are adult before they learn to use a stone as a tool for cracking hard nuts (Boesch & Boesch, 1984). Primates appear to manipulate social objects with more ease and sophistication than physical tools.

Observations such as these have suggested that primate intelligence is designed primarily for the social rather than the physical and have led to the Machiavellian intelligence hypothesis (Whiten & Byrne, 1988a) or social intelligence hypothesis (Kummer et al., 1997). The term Machiavellian intelligence emphasises the besting of rivals for personal gain over co-operation, whereas the term social intelligence (which is the more general term) is neutral on the balance between exploitation and co operation.

The social intelligence hypothesis is both stimulating and vague. It is stimulating because it reminds us that whenever psychologists study intelligence and learning in humans or animals, it is almost invariably about inanimate objects: symbols, sticks and bananas. It is vague because the nature of the intelligence it invokes is largely unclear, and as a consequence, the mechanisms of social intelligence have not yet been specified. This combination of exciting and imprecise should be an alarming signal. The social intelligence hypothesis has the seductive power of a political party with no precise programme that allows everyone unhappy with the established system to project his or her own values on it.

Introduction

In 1988, one section of Machiavellian Intelligence asked ‘are primates mind-readers?’. There seemed an obvious logic for posing the question. One of the most powerful ways to succeed in a complex social world is to read the very minds of one's companions, and get one step ahead in whatever competitive or co-operative games are at stake. We know this most clearly from our own human Machiavellianism, but we also know that the complexities of primate social life suggest niches for the ability and we know that primates have advanced social cognition: thus, the distribution of mindreading mechanisms in the primate order clearly begs investigation. Unfortunately, despite the years that had elapsed since Premack & Woodruff (1978) first tried experimentally to answer the question, ‘Does the chimpanzee have a theory of mind?’ (see Chapter 1), relevant empirical studies were still few in 1988 and mostly restricted to our own species: those on non-human primates could be counted on the fingers of one hand.

The situation has changed dramatically since 1988. Growth in the study of mindreading has blossomed, perhaps more than any other subject dealt with in the predecessor to this volume. To be sure, this has principally focused on the development of the capacity in humans. Developmental psychologists were the first to pick up on the potential of Premack & Woodruff's ideas, and more specifically on suggestions in the peer commentary on the article, that testing whether an individual can discriminate another's false belief would be the most convincing way to demonstrate a true reading of ‘mind’.

The puzzle of egalitarian society

The common ancestor of humans and the African great apes was a nonmonogamous ape, one that lived in a closed social network and exhibited hostile relations between groups with stalking and killing of conspecifics by males. Wrangham (1987) deduced these ancient traits conservatively by tallying behaviours that humans share with all three African apes. However, the presence or absence of social dominance hierarchy was left in abeyance — even though hierarchies are all too apparent in chimpanzees, gorillas, bonobos, and Neolithic and modern humans. The problem, presumably, was that human foragers, being egalitarian, have been taken to exist without any significant hierarchy; this erroneous assumption has seriously confused the assessment of our own political nature.

More recently, Knauft (1991) has likened our political evolution to a ‘U-shaped curve’. He does assume our common ancestor lived hierarchically, but points out that this approach to social life disappeared for a long span of evolutionary time until it returned with chiefdoms, civilisation and modern nations. Focusing on the ‘simple-foragers’ in whose bands evolved the genes we now carry, Knauft characterises them as largely lacking hierarchy in the form of dominance relations or stratification among the males, and as exhibiting little or no leadership and very low levels of inter-group violence.

Recent work of my own (Boehm, 1993) has questioned the first of these assumptions. I have suggested that among foragers hierarchical behaviour did not actually disappear but rather it assumed a radically different form.

Consequences of behaviour and meta-learning, and social intelligence

Don't be too clever in order to be smart

Our title alludes to Clever and Smart, comic figures created by Ibañez, which contrast in the behaviour they use to reach goals. Whereas Clever shows much refinement, Smart acts without many detours, and succeeds as often as Clever. This result would be surprising to common sense or to analysts of social interaction, Machiavellian intelligence and cognitive competence (Handel, 1982; Hinde, 1983; Anderson, 1985). Indeed, the analysis of the cognitive prerequisites of social interaction would normally lead to the conclusion that sophisticatedly planned and performed behaviour is the means to achieve social goals. Yet there is a caveat. The point we want to make is that the consequences of behaviour, not the degree of underlying cognitive complexity, determine social success. Straightforward action—reaction behaviour such as reciprocity (an eye for an eye) may be as efficacious as subtle diplomacy. For example, reciprocity may be well suited to stop overt physical aggression such as a child's temper tantrum. Watzlawick et al. (1967) tell illuminating stories of unsuccessful communication resulting from either endless recursive mindreading or ignorance of quite simple interaction rules (a man who needs a hammer imagines that his neighbour may be unwilling to lend one, and after a lot of thinking on the neighbour's possible motives, knocks angrily at the other's door shouting that he would never accept even a donated hammer). Watzlawick et al. call on interactants to communicate on their communication rules (to meta-communicate) to resolve problems, and like Dennett (1983), think that we can manage only few embedded propositions.

Introduction

In the pain of his passion, Cyrano de Bergerac fed Christian a suite of elegantly crafted lines designed to capture Roxanne's heart. Cyrano's act was clearly deceptive. Christian knew that he was being deceptive, but did not know the basis for Cyrano's apparent act of generosity. And then there is Roxanne, who believes that the handsome Christian is a poetic spirit able to weave verse that strikes at the heart. She is, of course, deceived. Here then, in one triadic interaction, we have Cyrano who knowingly deceives and does so on the basis of his knowledge of what Christian and Roxanne believe and desire. We have Christian who deceives by both withholding information about his lack of eloquence and actively falsifies information by making Roxanne believe that what emerges from his lips are true inspirations from the heart. And Roxanne is our gullible recipient, swept off her feet by Christian's ersatz performance.

Humans perform such complex mental acrobatics all the time, at least those humans who are over the age of about 4 years old and have all of their species-typical neural faculties intact. Is it even reasonable to contemplate the possibility that non-human animals are similarly endowed, to lie, cheat and conceal valuable information from other group members? In this chapter, I wish to accomplish at least three things along the way to answering this question. Firstly, I will discuss several conceptual issues that are relevant to thinking about the origins and subsequent evolution of not only deceptive behaviour, but a mental capacity for deception.

Introduction

Humphrey (1976) suggested that primate information processing skills, as displayed in laboratory learning tasks, exceed the demands of finding food, finding shelter and avoiding danger in the natural habitat, and he hypothesised that the apparent surplus capability evolved in response to the demands posed by life in complex social groups. Other chapters in this book attempt to spell out the empirical predictions of Humphrey's hypothesis, termed the ‘social intelligence’ or ‘Machiavellian intelligence’ hypothesis (Byrne & Whiten, 1988; Cheney & Seyfarth, 1992), and to evaluate the evidence from studies of social behaviour and brain size. In 1976 Humphrey had little information available on how primates find food in their natural habitat, and his statement about the relative simplicity of this task was mainly speculation. An alternative to Humphrey's viewpoint is that primates are capable of learning the relative positions and characteristics of a very large number of objects and topographical features in their natural habitat and that they use this stored information, in combination with current cues, to find food, to discriminate food from non-food objects, and to travel efficiently and safely. In this chapter, the questions of what information primates possess about the structure of their habitat and how such information might contribute to their survival and reproductive success are examined.

To form an evolutionary explanation for as complex a phenomenon as primate learning and memory capabilities requires that the capabilities of interest be well described and that a reasonable guess can be made about their biological value in the animals' evolutionarily relevant environment.

Introduction

The discovery of primate social complexity during the last 20 years stimulated a reinterpretation of the nature and evolution of primate intelligence. In this chapter we attempt to do three things. Firstly we present a short background highlighting some inherent difficulties with the current ‘social complexity/cognition’ model from which the Machiavellian Intelligence hypothesis derives. Next we explore the consequences of these problematic issues with data on sexual consorts in baboons. Finally we present another way to frame the social complexity/cognition link that we feel has the potential to more fully explain our consort data and to resolve some of the inherent ambiguities in the social complexity model of intelligence. In the process we are left to wonder whether Machiavellian intelligence is really ‘Machiavellian’.

The model

The intellectual events that culminated in the Machiavellian Intelligence hypothesis look slightly different from the description offered by Byrne and Whiten (Chapter 1) when seen from the perspective of primate field studies (Strum & Fedigan, 1997). This vantage point may help to explain why the Chance–Jolly–Kummer—Humphrey (Chance & Mead, 1953; Jolly, 1966; Kummer, 1967; Humphrey, 1976) hypotheses about ‘social intelligence’ did not actually begin to constitute a ‘domain’ of knowledge and research for nearly 20 years. Field data and shifts in theoretical orientations were crucial. Long-term studies of chimpanzees (see Goodall, 1986 and references therein) and baboons (Altmann, 1980; Ransom, 1981; Strum, 1981; Stein, 1984), in particular, documented an array of social relationships. These were initially treated as mere ‘social noise‘ resulting from many social animals living together (e.g. Ransom & Ransom, 1971; Ransom, 1981; Goodall 1986).

After a very slow germination in the more than 20 years leading up to 1988, the ‘Machiavellian intelligence hypothesis’ has subsequently been evoked as an explanatory theory in a wide range of contexts: neurophysiology (Brothers, 1990), social anthropology (Goody, 1995), medicine (Crow, 1993) and even news broadcasting (Venables, 1993), in addition to its impact on psychology and studies of primate evolution. All of a sudden, the idea that intelligence began in social manipulation, deceit and cunning co-operation seems to explain everything we had always puzzled about. This popularity may, of course, simply reflect its correctness. However, the vagueness of the theory may also have helped, allowing it to be ‘all things to all men’. The book that brought in the name did not even contain a single, clear definition of the Machiavellian intelligence hypothesis (Byrne & Whiten, 1988a)! This was not simply carelessness, but a reflection of the reality. In many ways, ‘Machiavellian intelligence’ is better seen, not as a precise theory, but as a banner for a cluster of hypotheses that have been under active investigation since before we coined the label.

All these hypotheses share one thing: the implication that possession of the cognitive capability we call ‘intelligence’ is linked with social living and the problems of complexity it can pose. In the mid-1980s, we thought we could discern a rise in the number of studies that acknowledged the potential explanatory power of the hypothesis. However, these were often rather disparate strands: the time, we felt, was ripe for an attempt to orchestrate them into what we hoped would be the beginnings of a more coherent and focused appraisal.

INTRODUCTION

In 1974, Ernst Mayr published a now classic paper on the distinction between innate and acquired characteristics. In this work, Mayr broke away from some of the more confining features of traditional accounts of innateness, and proposed a dimension along which behaviors might be expected to vary with respect to innateness. Mayr proposed a distinction between “closed” and “open” programs1 – a program that does not allow appreciable modifications during the lifespan of its owner is a “closed” program, while a program that does allow for the effects of additional input is “open.” Since it seems unlikely that any developmental program can be completely closed, Wimsatt (1986) has suggested that Mayr's notion of a closed program may be most fruitfully viewed as a relative one – “relative to the period of time of development under investigation, and the class of inputs being investigated, and probably also to the environment and the prior state of the developing phenotype” (Wimsatt 1986, p. 203). In Wimsatt's terms, a closed developmental program is one which is canalized with respect to the relevant inputs.

Mayr's classificatory schemes distinguished two types of behavior: A behavior is considered communicative if it is directed toward a recipient who is capable of responding with behavior of its own, and noncommunicative if it is directed toward a “recipient” that is passive and does not itself react (e.g., behaviors involved in selecting a habitat or seeking food).

INTRODUCTION

Striking parallels exist between the development of speech in human infants and the development of song in birds. Many sparrows, for example, learn their songs more readily during a sensitive period than at other times during development, require practice, and must hear themselves sing for normal song to develop (Baptista & Petrinovich 1986). These same features characterize both the earliest speech of human infants (see e.g., Ferguson et al. 1992) and second language learning, whether spoken or signed, among older individuals (Johnson & Newport 1989). Song production in zebra finches and canaries, like speech production in humans, is under lateralized neural control (Arnold & Bottjer 1985; Nottebohm 1991). Damage to any one of these areas, like damage to Broca's or Wernicke's area in (usually) the left temporal cortex of the human brain (for reviews, see Caplan 1987, 1992), produces highly specific deficits in the production or processing of communicative sounds.

As a result of these parallels in both behavior and neurobiology, studies of avian song development currently provide the best animal model for research on the mechanisms underlying speech development (Marler 1987). In contrast, while nonhuman primates are our closest living relatives and have often been used as animal models for the study of human social development (see e.g., Hinde 1984), their vocal communication is generally thought to provide no useful parallels with the development of human speech.

INTRODUCTION

In this chapter, we review work on the nature of vocal learning in human primates, comparing them en passant to nonhuman primates who share many of their capacities but are both less eager and less successful vocal learners. The basic question underlying this review is whether the precocious and prolific vocal learning of human primates can be explained by biological mechanisms that are specific to the language system or whether it relates to more general social capacities and to the particular social context of vocal learning in humans.

We know that young human primates are particularly good at vocal learning. One bit of evidence in support of this contention is that all national languages are spoken, even though extremely subtle articulatory and auditory discriminations are relied on to carry meaning in spoken languages. In addition, babbling and vocal play are early developmental activities universally observed in normally developing children (Locke 1992; Locke & Pearson 1992). Imitative vocal behavior is also universal in young children and common even in more mature language users. Furthermore, language learning, particularly word learning, by young children, is quite rapid and efficient.

Although there is much emphasis on children's preparedness for language learning, in fact children everywhere seem to enter the language system of conventional words use through the use of vocal forms that are more like adult forms in sound than in semantic or syntactic function. These early forms could be argued, though, to foreshadow a major function of oral language even in adulthood, namely to effect participation in social interaction rather than transmission of information.

INTRODUCTION

Vocal learning in birds has evolved independently in several different avian orders, but is common in only two groups, oscine songbirds and parrots. The diversity and complexity of vocal repertoire structure among the species in these two groups is enormous. However, much scientific attention in avian song learning has focused on a small group of songbirds, north temperate migrant species, in which song is restricted mainly to males and the occurrence of song and territoriality is seasonal. In addition to these birds there is also a vast number of laboratory studies on the development and neural control of male song in the zebra finch (Taenopygia guttata), an Australian species with an unusually compressed developmental period. This large body of research has resulted in general models of song and vocal learning drawn from only a small subset of the world's birds.

While these many studies have broadened our understanding of learned vocal communication, it is our contention that there is much to be gained by study of the vocal behavior of avian species with more complex social relationships. Such species exhibit long-term associations between well-acquainted individuals, and, for many, longterm associations are related to permanent residence in an area. Another important characteristic is the tendency to live in stable groups for at least part of the year. These groups could be, for example, a winter foraging flock of chickadees, a breeding colony of caciques, a foraging flock of cockatoos, or a permanently territorial pair of tropical wrens.

By consideration of common features of disparate groups that represent more fully the portion of the world's birds that learn vocally, we can approach a more truly universal model of vocal learning.

INTRODUCTION

Most studies of the effects of social interaction on the ontogeny of vocal communication in birds and primates concentrate on the normal course of development of species-specific codes: how birds learn conspecific song, how nonhuman primates develop their natural repertoire of calls, and how human infants develop language. The effects of social interaction, however, are probably even more important during exceptional learning (Pepperberg 1985): learning that is unlikely to occur in the normal course of events. Such learning, defined and described below, has been documented for a number of species, including humans. I have been particularly interested in examining how social interaction can influence a specific type of exceptional learning – the development of interspecies communication between humans and birds. My research on the effects of social interaction on the acquisition of a vocal, English-based code by grey parrots (Psittacus erithacus) clearly demonstrates how social and environmental input1 can engender learning that would not otherwise occur (e.g., Pepperberg 1990a). Interestingly, an analysis of research on ape language also demonstrates how social interaction may be a particularly effective means of teaching nonvocal human-based communication codes to nonhuman primates.

Although characterizing the effects of social and environmental influences on exceptional learning is not a simple task, my work has shown that a conceptual framework, social modelling theory, can be used (a) to characterize how social input influences learning and (b) to delineate the critical features of input necessary for exceptional learning.

INTRODUCTION

Marine mammals stand out among nonhuman mammals in their abilities to modify their vocalizations on the basis of auditory experience. While there is good evidence that terrestrial mammals learn to comprehend and use their calls correctly, there is much less evidence for modification of vocal production (Seyfarth & Cheney, Chapter 13). In contrast, vocal learning has evolved independently in at least two marine mammal taxa, the seals and cetaceans, and is widespread among the whales and dolphins. We concentrate our focus in this chapter on vocal learning and development in the bottlenose dolphin (Tursiops truncatus) because it is the marine mammal species in which vocal learning and imitation has been best studied.

Dolphins produce a variety of sounds. The two predominant sound types are clicks, which can be used for echolocation, and frequency-modulated whistles, which are used for social communication. In addition to whistles, dolphins produce short frequency upsweeps that have been called chirps (Caldwell & Caldwell 1970). The dolphin vocal repertoire also includes a variety of burst pulsed sounds and combinations of pulses and whistles.

Captive bottlenose dolphins of both sexes are highly skilled at imitating synthetic pulsed sounds and whistles (Caldwell & Caldwell 1972; Herman 1980). Once a dolphin learns to copy a sound, the novel sound can be incorporated into its vocal repertoire, and the dolphin can produce the sound even when it does not hear the model. Bottlenose dolphins may imitate sounds spontaneously within a few seconds after the first exposure (Herman 1980), or after only a few exposures (Reiss & McCowan 1993).

INTRODUCTION

Songbirds learn their songs by hearing others and then copying them, matching or improvising on the song theme (Slater 1989; Catchpole & Slater 1995). Their social behavior varies among species – they are migratory or resident, solitary or group-living, faithful partners to a single mate, polygynous or with no pair bond, and parental or nonparental in the care of their offspring (brood parasites lay in nests of other species, and their fosterers rear the young). All songbirds depend on parental care, and it has been suggested that this is the time when the young learn their songs. Later, when they are independent, the birds engage in a wider range of social interactions.

Field studies suggest that most songbirds learn their songs after the time of natal dispersal, when a bird moves from the site where it was reared to an area where it copies the song of a neighbor, rather than singing the song of its father: Bewick's wrens (Thryomanes bemickii), marsh wrens (Cistothorus palustris), saddlebacks (Philesturnus carunculatus), indigo buntings (Passerina cyanea), white-crowned sparrows (Zonotrichia leucophrys), and corn buntings (Emberiza calandra) (Kroodsma 1974; Verner 1976; Jenkins 1978; Payne et al. 1987; Baptista & Morton 1988; Petrinovich 1988; McGregor & Thompson 1988; McGregor et al. 1988). Song sparrows (Melospiza melodia) copy at least one song from three or four neighboring males when they settle on a territory, some time after the first four weeks of life (Nice 1943; Beecher et al. 1994). In two species males often copy their father (Darwin's finches (Geospiza fortis), zebra finches (Taeniopygia guttata); Millington & Price 1985; Gibbs 1990; Zann 1990), but not all individuals do this.