Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
1. What did recursion recurse? In their recent overview of current language-evolution research, Hauser, Chomsky and Fitch (2002) suggest that researchers should begin by seeking for some type of recursive ability (perhaps developed for navigation, number quantification or social relationships) among other species. They admit that the discovery of such an ability would merely “open the door to another suite of puzzles”, but regard as plausible the suggestion that a previously impenetrable module–for navigation, say–“may have become penetrable and domain-general” as a result of either “particular selective pressures” or as a “by-product” of “neural reorganization” (Hauser et al. 2002, 1578), thereby giving rise to language. Although Hauser et. al. do not use these terms, what they hypothesize could only be some form of what has been variously termed “pre-adaptation”, “exaptation”, or “change of function”, with a Gouldian twist in the second, “by-product” possibility (Gould and Lewontin 1979). Accordingly, we have to take into account the nature of such processes. Although these are quite widely attested–buoyancy chambers into lungs, cooling panels into wings, and so on–they normally involve new functions being found for old body parts, rather than for old computational mechanisms. I know of no case where an abstract computational mechanism devoted to one function has been exapted to serve some other function (granted, this may be merely my ignorance–in which case I hope for speedy enlightenment–or because no-one so far has looked for such developments). However, if we assume that this is possible, the circumstances attending such exaptation must necessarily differ from those where a physical organ is involved. In the latter case, that organ merely needs to modify its existing form for the new function to take hold. But any abstract capacity requires a separate and distinct domain over which it can range. For example, recursion would have to have something to recurse with. Recursion could not have applied to the units of any animal communication system (ACS) that preceded language, because ACS units are complete and self-contained. Two units in sequence mean simply the sum of the unit’s two meanings: a food call followed by an alarm call, for instance, simply means “There’s food here” plus “There’s a predator coming.” Essential to any recursive process is the capacity to embed one unit within another (for instance, to convert–as some early generative grammars did lierally–“A dog bit me” and “The dog ran away” into “The dog that bit me ran away”). Thus for recursion to have been able to operate at all, there must first of all have existed particulate units–units standing for smaller chunks of meaning than ACS units–that could be arranged and re-arranged in a variety of ways: in other words, words (or perhaps manual signs–the distinction is of no importance here). The approaches both of Hauser and his colleagues (Hauser et al., 2002; Fitch, Hauser and Chomsky 2005) and of their critics (Pinker and Jackendoff 2005) take a similar perspective. To both, language evolution is viewed as an abstract problem, not a dynamic process that occurred at a specific time or times, in a specific location or locations, as merely a part (albeit arguably the most significant part) of the overall evolution of the human species. Their focus is on what attributes of language are unique to humans and what are shared with other species. No attempt is made to determine the chronological sequence in which attributes developed, nor the manner in which, whether unique to humans or shared with other species, they were integrated with one another to form the well-articulated whole we now know as language.. While an approach of this type has obvious advantages, it needs to be supplemented by one that looks more closely at the dynamics and pragmatics of the evolutionary process. By no means all aspects of this process depend on our correctly interpreting sketchy and often equivocal archaeological evidence. Some can be determined by simple logic. The temporal precedence of particulate symbols over recursion (regardless of whether or not this was exapted from some prior function) is just one of such aspects. Logical considerations enable us to determine much more than this about the materials to which recursion would eventually apply. 2. Symbolism, predication and displacement. For there to exist units to which recursive operations might apply, two other novel properties–properties not found in ACSs, with a tiny handful of exceptions that I shall deal with shortly–would have had to emerge: symbolization and predication. Why , precisely, can ACS units not be concatenated? Not, for sure, because serial juxtaposition lies beyond the powers of non-humans. Animals that have been taught quasi-linguistic systems have without exception concatenated the units of those systems to form crude propositions, apparently with little if any formal instruction in proposition-making. Basic to such proposition-making is predication: a process that links two or more systemic units in a relationship where one (or more) of such units tells us something about another unit. Without some such semantic linkage, what would be the point of concatenating two or more items, and what could they possibly mean if you did concatenate them? A sequence like “Here’s some food of type X”, could not, of its very nature, tell us anything about “Look out, there’s a leopard!”–or vice versa. Thus not merely particulate units, but the predicative use of such units, must have preceded the exaptation of recursion. Such particulate units must also have been symbolic. ACS units are, almost without exception, indexical: they point to an existing state of affairs (“I am angry”, “I want to mate with you”, “Get out of my territory!” or whatever) or an object located within the sensory field (“Here’s food of type X”, ‘A predator’s coming”, or whatever). There is no way in which these could have been used to describe past or future states of affairs, or objects accessible only through the memory store. But once you have particulate units describing objects or events (or, to be more precise, describing concepts of such objects and events), in isolation from one another and from any particular instantiation of those objects and events, it becomes possible to achieve displacement: the capacity to refer to any object or event, irrespective of when or where it occurred or is/was/will be located. It is, perhaps, possible to imagine that recursion preceded symbolism, but much harder to imagine what value recursion would have had if it hadd been hobbled within the here-andnow. Note, further, other characteristics of language that become available once symbolization and predication are in place. To the extent that ACSs are referential, their reference is indissolubly linked with their indexicality. If a leopard is present, a vervet monkey’s leopard alarm call can be referential, but if no leopard is present, it is merely a mistake or a tactical deception Unlike words, ACS units cannot be questioned (“Is that a leopard?” or simply “leopard?”) or negated (“No leopards here!”). Recursion is not a pre- requisite for such utterances: simple predication, which does not necessarily involve more than two units casually concatenated, suffices to produce them. And since, as suggested above, predication preceded recursion, we may conclude that questions and negations, essential ingredients of language as we know it, also preceded the emergence of recursion. Symbolization, predication and displacement thus form the foundation on which the complex structure of language subsequently arose. If these factors necessarily preceded recursion in the evolution of language, they might seem worthy of more attention than they have hitherto received (see Table 1). (Insert Table 1 about here) Their appearance must, at the very least, indicate the boundary line between ACSs and a communicative system (call it protolanguage, or whatever) as unique to human ancestors as language itself. Indeed, if they constitute a necessary as well as a sufficient condition for the emergence of true language, they may represent a more significant Rubicon between humans and nonhumans than recursion itself, unique to our species though that be. Such would be even more clearly the case if recursion turned out to be, rather than a preexisting, exapted faculty, an entirely new and emergent property for which the emergence of symbolism and predication formed an important selective pressure. However, such issues lie far beyond the scope of the present paper, which limits itself to possible origins for symbolization and predication. 3. Among the hymenoptera If we seek out precursors of predication and symbolization by the means Hauser at al. suggest–comparative studies–our search takes us in an unexpected direction. The only species where such possible precursors occur are found among the hymenoptera: ants and bees. It has long been claimed that bees have a “dance language” that is symbolic in that it can convey information about pollen and nectar that is at some distance from, and beyond the sensory range of, both the communicating and receiving bees (von Frisch 1967; Gould 1976; Dyer and Gould 1983; but see Wenner 2002 for a contradictory view of bee communication). Less well publicized, but perhaps even more impressive, are the communication systems of ants (Wilson 1962; Hangartner 1969; Moglich and Holldobler 1975; Holldobler 1978), which feature not only symbolization of non-present objects but even a precursor of predication. Why should this be the case? Why should these relatively primitive organisms have access to capacities that even the great apes lack? The answer can only be that, for bees and ants, the standard symbolless, predication-less ACS must be inadequate. Other species do not develop symbolization and predication under natural conditions because they can live their lives perfectly well without such things. The fact that at least some species can acquire symbolization and predication under training shows that motivation, rather than the necessary neural infrastructure, is what is lacking here. Under the appropriate motivation, it may be that an extremely wide range of species might develop at least rudimentary forms of the functions in question. Why would a standard ACS be inadequate for bees or ants? Both are extractive foragers. Both operate from a central point, foraging individually or in small groups. In the case of ants, a variety of different food sources are utilized, some involving objects too large for a single ant to deal with. Both for ants and bees, the food source may be transient and require numbers for its efficient exploitation. These factors make recruitment a necessary strategy. It is necessary to make very clear what is meant here by recruitment. It means inducing conspecifics to exploit food sources of otherwise unpredictable location that neither the sender of the signal nor its receiver can directly perceive by any sensory means. It does NOT mean inducing conspecifics to exploit a food source that the sender can directly perceive at the moment of signaling. For that, a simple food call or rallying cry would suffice. Recruitment in the sense intended here demands at least some type of symbolization, if not predication. Recruitment among ants takes several forms: tandem running, where a single ant presents a food sample to another ant and then leads it towards the food; light trail laying plus a waggle display, used to recruit groups; and the establishment of a semi-permanent chemical trail, sometimes varying in strength according to the quality of the food source, where the source is large and mass recruitment is required (Sudd & Franks 1987). Tandem running would seem to be the closest of these strategies to predication. “When a forager finds a plentiful food source it returns to the nest and regurgitates food to its nestmates. Then it raises its gaster and extrudes its sting bearing a droplet of liquid. This attracts nestmates to it in the nest. As soon as the first of these nestmates reaches the caller, the caller runs out of the nest and leads the nestmate to the food source” (Sudd & Franks 1987, p. 112, based on the description of Leptothorax species, Moglich et al. 1974). It may not seem overly fanciful to paraphrase this as, “Come, X food [is] this way”–a subject with a comment about it, in other words an act of predication. Obviously, a communication system composed of such elements falls far short of the simplest system that we would recognize as a language. First, the elements it can symbolize are sharply limited; second, messages in such a system can involve only the limited domain of foraging; third, since the meaningful behaviors involved are innately programmed, the system cannot freely add additional units, as any language can freely add words. However, such systems are capable of transferring factual information of a type that other ACSs cannot transfer, just as human language is. 4. Factors favoring recruitment The need for recruitment arises from the particular niches developed by ants and bees respectively: niches that have some obvious differences (bees forage aerially, ants terrestrially, for example) but have many more similarities than differences. By examining these similarities we can specify the characteristics of both species and niche that would select for an ACS with the essential feature of displacement, a central component of symbolism: that is, the conveying of objective information about things that are not physically present. With regard to the species, recruitment can only arise in a species that is social. In addition, it requires that such a species be co-operative. Species can, of course, be social without being co-operative; primates notoriously will go to some lengths to keep prized portions of food to themselves. There must, therefore, be some specific reason for cooperation to arise in a social species. One such reason is eusociality; if all the members of a group are siblings, inclusive fitness ensures that the good of all will be the good of each. But eusociality, if perhaps a sufficient cause of co-operation, may not be a necessary one. Another may be vulnerability to predation; a social species is more likely to be co-operative if it has come under heavy pressure from predators and co-operation between members can reduce this threat. Characteristics of niches that select for recruitment are as follows. Basic is that the means of subsistence should be extractive foraging. Carnivory and herbivory do not require recruitment. Carnivores can’t recruit for a food source because that source is mobile and by the time any recruit gets back to where it was, it probably won’t be there. Herbivores don’t need to recruit for a food source because (except in occasional severe droughts) there is usually enough in plain view to render communication unnecessary. Food sources may be widely scattered; if food is clumped and contiguous, there is correspondingly little need for information about it to be exchanged. Sources should also be unpredictable; if groups know where and when a particular food source will be present, the need to communicate about it is again weakened. If in addition to being scattered and unpredictable, sources are also transient, communication is still more strongly encouraged, while an intake of food that is diversified puts an adaptive premium on information about type and quality. Another factor favoring recruitment applies if the food is sizeable; if potential food is several sizes larger than the forager, recruitment becomes almost inescapable. Other necessary niche characteristics involve the means by which extractive foraging is carried out. The groups engaged in it should have a fission-fusion structure. Clearly, if a group forages as a unit, there is no need for members to communicate their finds to one another. The pattern of foraging is likelier to encourage recruitment if the foragers operate from a central place where information can be exchanged; central-place foraging is in turn made likelier if the species concerned habitually engages in provisioning relatively immobile young. Ants and bees both meet a substantial subset of the relevant criteria. Both are eusocial (hence co-operation is inbuilt, so to speak) and both engage in extractive foraging. Though food sources for bees are not diverse and the food items themselves not sizeable, the flowers that furnish them with nectar and pollen are widely scattered (often at considerable distances from the hive), transient (many blooms open only at certain hours, others last less than a day), and largely unpredictable (or at least, too numerous and irregular in their appearance for any organism with a bee-sized brain to keep track of all of them in memory, although single persistent patches may be returned to by bees even after overwintering, Winston 1987, p. 176). Ants eat a wide diversity of foods: “even those species which count as specialists for meat or vegetable food are not always consistent in this” (Dumpert 1978, p. 236). To the extent that they function as predators and scavengers (rather than fungus growers or aphid dairymen), their food sources are transient, unpredictable and scattered. They are also often sizeable; ants often capture larger animals which would be impossible for individuals to manage (Wilson 1958). Both ants and bees employ a fission-fusion strategy in foraging, except for some species such as army ants, which are dramatically unselective in their choice of diet.. All honey bees, and most ants (some are nomadic) forage from a central place (the nest or hive) to which they return with their booty to provision their larval young. It follows that for both ants and bees, recruitment plays a vital role in subsistence. It is possible that they could survive without it. It is certain that, if there were insect colonies that practiced recruitment and insect colonies that did not, those that practiced it would prosper at the expense of the others. Accordingly, recruitment strategies have proven highly adaptive and have been adopted by the majority of bee and ant species. As with predications, and in striking contrast to the signals in other ACSs, recruitment signals in ants can be shown to require concatenation. Holldobler (1971) demonstrated this experimentally in Camponotus socius, an ant that first uses shaking behavior to draw recruits and then follows a chemical trail of its own making to the food source. However, removal of this lead ant causes the recruits to abandon the trail. By mixing contents of the ant’s bladder and poison gland and ejecting this mixture, Holldobler caused leaderless ants to persist in following the trail, thus indicating that both the directional trail and continuing signals from the leader were required to insure recruitment. 5. Niche construction. If natural selection by the environment, acting on variation within the gene pool of a species, was the sole force driving evolution, it would be hard to explain why humans are so different from the other great apes. The genetic difference involved here–less than 2% in some cases–is of an order that might be expected to produce differences in behavior no more radical than those between horses and donkeys, or lions and tigers. Recently, however, it has been suggested that there is another force driving evolution: the active construction by species of unique niches, which in turn impose novel selection pressures on that species, leading to a feedback process (Laland et al. 2000; Odling-Smee et al. 2003). Hitherto, evolution has been generally perceived as a one-way process in which a variable but largely autonomous environment selected from genetic variation in populations of living organisms. The role of those organisms was thus essentially a passive one. The environment, however, is itself to a large extent the creation of the organisms that inhabit it. Without photosynthesizing organisms there would be insufficient oxygen to support multicellular life; without earthworms and similar creatures there would be insufficient soil to support the majority of plant forms on which all mobile organisms ultimately subsist; without beavers, many wetlands that support a variety of fish, birds, amphibians and other life forms would not exist. In other words, evolution is a two- way not a one-way street. Environmental factors may select for genetic traits, but those factors themselves may result from prior activity by the species concerned; for instance, the physical shapes of beavers’ lips, teeth, oil-glands, eyelids, tails and feet were influenced if not wholly determined by the consequences of their initial “decision” to create an environment in which their lodges would not be threatened by the periodic shrinkages and dangerous overflows characteristic of running streams. On this view, organisms play a role in their own destinies that goes far beyond anything envisioned by the neo-Darwinian consensus. One significant advantage of niche construction theory is that, in contrast to previous biological theories, it does not merely account for human-ape differences, but directly predicts them. Other apes all inhabit roughly the same kind of niches and have remained in those niches, as far as we know, throughout the last seven million years. Human ancestors, in contrast, have inhabited at least four distinct niches. First came a foraging niche, characteristic of at least some species of australopithecines, in which tubers furnished a significant portion of the food supply. Then came a scavenging niche, in which a substantial percentage of nutrition was derived from the carcases of animals, including megafauna. There followed a hunting niche, in which scavenging and foraging were supplemented by the communal pursuit of living animals. Finally and quite recently we entered a farming niche based on agriculture and the domestication of animals. Since the niche forms the organism just as much as the organism forms the niche, differences in primate niches are the obvious places to look if we are trying to understand differences in primate adaptations–in particular, those adaptations that gave rise to language, 6. Homo in the scavenging niche. Around 2mya, world climates became cooler and drier, and savannas largely replaced the mosaic woodlands that had been the previous habitats of human ancestors (Reed 1997). Hominid patterns of behavior changed; day-ranges probably increased considerably (McHenry 1994). Scavenging already provided part of hominid diet, but it was probably low-end scavenging, conducted after other predators and scavengers had taken their share, and involving mostly harvesting of foods inaccessible to other species (Binford 1985). Bone marrow, for instance, could be obtained if primitive stone tools were used to crack open bones, but little meat would have been left by the time earlier and more powerful scavengers had done with the carcass. However, in a mosaic woodland habitat other food sources would have been readily available, including fruit and nuts and small animals such as red colobus monkeys, which chimpanzees catch readily without weapons (Stanford 1998). Although such a niche might have required ranges larger than those of contemporary apes, it would not necessarily have placed a premium on recruitment. However, when woodlands were replaced by grasslands, food sources other than meat (except for tubers, O’Connell et al. 1999) became much more rare. At this point, hominids engaged in counteractive niche construction (Odling-Smee et al., 2003: 46): a process in which a species counteracts an environmental change by relocating to a new environment or changing some aspect of its behavior. The new niche was a logical move for organisms that had already used stones to break open bones and (probably) digging sticks to excavate tubers. All they needed were the flakes struck off in the production of Oldowan tools. At least by the Plio-Pleistocene boundary hominids had moved their focus from brain and bone-marrow to active scavenging and butchery (Bunn & Kroll 1987 Blumenschine 1987). The efficacy of Oldowan tools was practically demonstrated by Kathy Schick and Ray Dezzani, who used them to butcher an elephant that had died of natural causes (Schick and Toth 1993, pp. 166ff). They “were amazed as a small lava flake sliced through the steel gray skin, about one inch thick, exposing enormous quantities of rich, red elephant meat.” Since “modern scavengers normally do not eat a dead elephant until it has decomposed for several days”–they can’t, their teeth cannot penetrate the skin until decay and the expansion of internal gases has split it open–“such carcasses may have provided occasional bonanzas for Early Stone Age hominids” (see also Blumenschine et al, 1994). In fact, the bonanzas may not have been so occasional. While it is true that bones of animals larger than 2000 kg are rarely found at catchment sites, by or soon after 2mya hominids had moved to territory scavenging rather than catchment scavenging (Larick and Ciochon 1996), making it likelier that large carcases would have been exploited in situ, and would accordingly remain widely scattered and rarely unearthed by researchers. It seems plausible that under prevailing conditions there were quite large numbers of elephants, rhinoceroses and hippopotami, as well as other megafauna now extinct. Such animals would have presented a formidable challenge to most predators; besides, why would a predator kill something knowing it would have to wait days for it to be ready to eat? For this reason, a high percentage of megafauna must have died natural deaths. This enables us to explain the phenomena of tooth-marks superimposed on cut-marks, which suggest that hominids were sometimes accessing carcases before carnivores (Monahan 1996). Some researchers who, based on good evidence that meat formed a substantial part of early diets, are reviving the “early-hunting” hypothesis (e.g. Dominguez-Rodrigues 2002) often write as if all animal deaths represented kills, and therefore that the cut-before-bite evidence must result from either hunting or confrontational scavenging (in which hominids are seen as driving competitors away from a kill before any of them have had time to dismember it). While it is likely that relatively few deer-sized or smaller animals survive to die natural deaths, the same cannot be true for megafauna. Moreover, optimal foraging theory (Stephens & Krebs 1986), supported by a variety of field and/or experimental studies on organisms as diverse as gulls (Irons et al. 1986), stream insects (Velasco& Millan 1998) and white-tailed deer (Schmitz 1992), indicates that, other things being equal, any species will select the food that yields the highest calorific intake relative to the energy expended in obtaining it. Since for hominids this goal would be best represented by untouched megafauna carcases, optimal foraging theory suggests that they may have gone to some trouble to locate and exploit dead megafauna, even in preference to more widespread, more easily obtainable, but less nutritious alternatives (although they doubtless continued to exploit the more readily available brains and bone-marrow from stripped carcases, as well as vegetable foods, whenever fresh megafauna were unavailable). 7. Homo and recruitment conditions. The niche described in the preceding section was one that fulfilled most of the conditions favoring recruitment strategies described in Section 4.1 above. The hominids involved (probably h. erectus or h. ergaster, if indeed these are distinct species) were a social species whose niche would surely have made them co-operate more than most primates. A major factor would have been the risk of predation by savanna-dwelling carnivores. Forest apes, though not immune from predation, seem very seldom to serve as prey (except to humans), witness the contrast between the specialized alarm calls of vervets and the complete absence even of a generalized alarm call among chimpanzees or bonobos. Relatively few predators inhabit forested areas or climb trees, and apes can climb faster than leopards. In contrast, predators of several species, some now extinct, ranged the savannas and impacted heavily on hominds (Lewis 1997); even today, in areas where major predators are found, modern humans frequently fall victim to them (Treves & Naughton-Treves 1999). Two consequences follow from this state of affairs. The need for trust and co-operation, to be sure that there would always be somebody watching your back, must have been considerably raised among hominids. The tendency towards Machiavellian strategies, intrigues and one-upmanship must have been correspondingly lowered. Moreover, larger day ranges and the need to be constantly on the lookout for predators would between them have substantially lowered both the available time and the sense of security necessary for prolonged and intensive socializing. Next, consider foraging patterns. These are uncertain, due in part to ongoing controversy over whether home bases existed in the PlioPleistocene (Isaacs 1978; Binford 1981; Rose and Marshall 1996; Cavallo 1997). If there were home bases then by definition there was central-place foraging (there may also have been provisioning of the young, but, as with home bases, it cannot be determined at what stage of hominid evolution this began). But is the converse necessarily true? To determine this we should look at the behavior of other grounddwelling primates, such as hamadryas baboons (Kummer 1968). These have no fixed base; however, they alternate between several “safe places” (usually rocks or cliffs near a river bed) where they overnight, shifting periodically from one to another. The net result, however, is not significantly different from central-place foraging, Typically they start from one such place and return to it, either remaining there for the following night or traveling in a group to the next safe place. Given the risk of nocturnal predation, their inability to climb as quickly as other primates, and the extreme scarcity of trees to climb, it seems almost certain that h. habilis (and probably also early h. erectus) employed a similar strategy. That they also employed, in their foraging, a fission-fusion strategy also seems more than a reasonable conjecture. This strategy, in some form or other, is common to advanced primates generally, whether bonobos or baboons. We can ignore, for present purposes, the kind of fissionfusion that relates to longer time periods, in which overall group membership changes as some members drift off by themselves or to other groups while members of other groups wander in. What concerns us here is fission-fusion within the day-range as a strategy to maximize resource extraction. If a range has a high density of resources, such maximization requires only a minimal separation of subgroups; subgroups seldom need to travel outside eye- or earshot of one another, so that the need for communication between subgroups (and a fortiori the need for recruitment to exploit resources) is minimal. However, if a range has a low density of resources (as was probably true, in a majority of cases, around 2 mya and later) it becomes inefficient for subgroups to remain in close contact, since foraging on a narrow front would result in no-one getting enough to eat. If, however, the main group split into four subgroups, and if each selected a different quadrant with reference to the central place, four times the area could be covered, and (assuming a uniform resource distribution in all quadrants) each individual would get four times as much to eat. The foregoing is not a consideration for baboons, whose diet consists almost entirely of leaves (96% at the end of the dry season) or seeds and flowers (87% in the long rains; data from Kummer 1968, Table XIV), and who accordingly do not need to cover a wide area for even groups of a hundred or more to be adequately fed. However, to some extent for australopithecines and perhaps early h. habilis in a mosaic woodland setting, and to a much greater extent for later homo in a grassland setting, the foraging logic would have become inescapable. On the one hand, the scarcity of good night refuges plus the need for numbers to protect against a mass attack by nocturnal predators of the hyena class would have pushed towards large nighttime group sizes. On the other, the thin and dispersed state of food sources would have pushed towards small daytime group sizes. The optimal solution would have been for nighttime groups of, say, ~50 to split into several smaller groups for daytime foraging, then regroup at the original point of dispersal. Finally, consider to what extent properties of the food supply satisfied conditions for recruitment. The diet as a whole was necessarily diversified. Setting aside leaves and grass (save in dire emergencies), hominids probably ate whatever they could get. We need not, for the moment at least, consider bones, which though scattered and unpredictable in their locations would have remained in place for long periods. The same is true of roots and tubers, which were not even unpredictable, given that hominids should have had good cognitive season-by-season maps of all the vegetable resources within their home range. We should focus, therefore, on the most nutritious and highly-prized food-source, the intact carcasses of large dead animals. These fulfilled all the conditions requiring recruitment. They were scattered, indeed relatively rare. Their location was unpredictable, except perhaps that in drought seasons they were somewhat likelier to be found around dried-up waterholes. They were highly transient, in that they offered a very narrow window of opportunity between the moment of death and the first rupture of skin. Post-rupture, they would have been highly dangerous, surrounded by hungry and impatient animals in considerable numbers. Indeed, prospective scavengers may have start arriving well ahead of the first rupture, so that the sooner hominids located the carcasses, the better. One thing going in their favor was that predators and scavengers alike are good cost-benefit analyzers. No animal is going to risk being lamed by a well-aimed stone when the feast is a prospect still several hours away (when the carcass is open and ready and you either grab your piece or go hungry, it’s a different story). But the most crucial condition was that the carcasses were sizeable–many times larger than any hominid. This was a situation no other primate ever had to contend with. The paradigm case for all other primates is that of an animal confronted by a rare and delicious fruit, probably not more than a pound or so in weight, whose possession of it is in no way threatened by any other species– only by other members of its own group. Under such circumstances, selfishness pays off, and the animal will employ any stratagem that allows it to eat undisturbed. Now consider an animal confronted by a rare and delicious carcass, weighing in excess of a ton, whose possession of it is threatened by a variety of very ferocious species. Selfishness is futile here, co-operation is the only way to go; animals that do not co-operate get nothing. Suppose a small sub-group of hominids, alerted by vultures or just lucky, stumbles on a huge carcass, still unruptured. Possibly other animals are already stalking around. The group could, of course, take its chance and start cutting. But to start cutting would have the same effect as a rupture, it would immediately trigger a feeding frenzy in the other scavengers, and the smells of blood and meat might draw more of them. The whole sub-group could be overrun in minutes and end up as dinner too. Besides, how could they, even if they drove off all other scavengers, eat all that meat by themselves? Some might have close relatives in other sub-groups; inclusive fitness would then also come into play. In other words, almost every factor in the situation would cause them to seek additional helpers.. We are now in a position to compare the niches of ants, bees, apes and early Homo with respect to the conditions favoring recruitment that obtain in all four niches, as shown in Table 2. (Insert Table 2 about here) The homo scavenging niche is much closer to the niches occupied by ants and bees than it is to the niches occupied by closely-related primate species. Similar niches yield similar pressures. If the selective pressures in the hominid niche favored recruitment, how exactly could recruitment be undertaken in a species many times larger and with vastly more brain power than the hymenoptera? 6.0 How language began. Having a brain the size of a coconut rather than the size of a pinhead has disadvantages as well as advantages. One disadvantage, in this context, is that you are then an individual rather than a cog in a machine; you have your own agenda, your own preferences, and definitely a will of your own. Your subgroup may have found some tempting source of food, but another subgroup might have made finds of its own (piles of bones, bees’ nests, termite mounds...). The state of mind might often have been: why should I go to your find, why don’t you come to mine? Optimal foraging required some means by which the finds of different sub-groups could be compared and evaluated. Recruitment, therefore, couldn’t happen unless the nature of the find could somehow be indicated. One of the advantages of the coconutsize brain is that it holds primary representations (Bickerton 1990) of a very broad range of organisms and entities, including all the species with which the individual habitually interacts. To express these, all that is needed is a layer of secondary representation: some unique labels that will signify the objects concerned and trigger associations with those objects in the minds of others. Much ink has been spent on whether the original form of language was signed or spoken: often it seems to be assumed that these are mutually exclusive options (Hewes 1973; MacNeilage 1998). But this is not necessarily the case. Ant “language” contains visual, chemical and tactile elements. It could well have been the case that the original form of language was equally mixed, involving sounds, manual signs, facial expressions, mimesis (Donald 1991) or pantomime (Arbib 2004): whatever worked best communicatively. Its original units need not all have been symbols: some of them, like those of ants and bees, could have been indexical or iconic. Symbolism would enter through the use of such units in the context of fission-fusion foraging, where it would be clear, precisely from this context, that objects not physically present were being referred to, In other words, displacement would have formed the wedge that introduced symbolism into the human communication system. Thus a group that had located the carcass of an elephant, rhinoceros, hippopotamus or other large mammal could transmit this imitation by imitating the sound made by the animal or by mimicking some aspect of its behavior, following this with arm-waving and a pointing gesture–a sequence that would roughly correspond to the ant’s “Come, X food [is] this way” (see example in Section 4) and that is, of course, a predication. Once this type of communication had become functional, it could be expanded to include other information relevant to deciding which food sources to pursue. An optimal decision would have to take into account not merely the energy budget that helps determine all animal food choices–is the input going to be worth the effort?–but also the choice between high gain/high risk and low gain/low risk. Should the group take a chance of losing some of its members for a highprotein food that would last days, or content itself with the patch of tubers that some of its members were already engaged in harvesting? A number of factors might influence the choice: distance to carcass, hours of light available, length of time the carcass had been “on the market”, number and species of scavengers already at the site (if any), patch size of the tubers, and so on. To be able to transmit any of this information would make for a better choice. Once the breakthrough had been made, units would progress from iconicity to true symbolism by a process well attested for both spoken and signed languages. Units of the former suffer constant phonological attritions and mutations: “laboratory” goes to “labratry”, “forehead” to “forred”, while changes such as that in the first syllable of “breakfast” sever the word from its original meaning of “breaking a fast”. Thus even the auditory imitation of an elephant’s trumpeting would in time become shortened and stylized into a stereotypic form that might give the illusion of arbitrariness. Similar process apply in signed languages. Klima and Bellugi (1979, pp. 67-83) give a telling example: the American Sign Language sign for “sweetheart” originated as the pressure of both hands on the region of the heart, but changed gradually over time into an abstract gesture in open space before the center of the body. According to some writers (e.g. Mithen 1997; Jackendoff 2002) initial attempts at language would have been limited to some narrow function–if language developed to aid foraging, it would have been restricted to foraging contexts. Granted that ape and bee “languages” are limited in this way, but we are now talking about organisms with brains several orders of magnitude larger than hymenopteran brains, and I see no possible justification for supposing that those brains would be subject to similar restrictions. The brains of possibly all organisms above a particular brain/body ratio contain a wide range of primary representations. (Herrnstein 1979; Bickerton 1990). Once it has been discovered how to voluntarily label just one of these, it becomes potentially possible to voluntarily label all: there are no privileged lines of access between those areas of the brain that store concepts and the various motor areas, and no internal barriers that would block access to other conceptual representations. If what we are talking about is something with (at least some of) the defining properties of language, rather than those of a PCS, the units (equivalent to words or signs in a modern language) would not have been innate and genetically determined, but added voluntarily, invented on the fly and used until they were generally understood or replaced by some equivalent that was better understood. A number of different attempts to express the concept “elephant” might have had to be made before one of them, or some amalgam of more than one, got selected as the elephant. And since such a unit (as distinct from an alarm call) would be expected in the absence rather than the presence of its referent, the progress towards true symbolism (hence use of the term in a variety of contexts) might have been slow, but could hardly have been avoided in the long run.. The foregoing scenario would work well in a context of central-place, fission-fusion foraging. But it does not crucially depend on such a scenario. A paper that argues against both the central-place model and the centrality of meat-eating (O’Connell et al. 1999) contains a striking vignette of recruitment along similar lines: “Neither would transport of parts to ‘central places’ be indicated...; individuals or groups may simply have called attention to any carcass they encountered or acquired, just as do modern hunters...If the carcass had not yet been taken, the crowd so drawn could have done so, then consumed it on or near the spot, again just as modern hunters sometimes do” (O’Connell et al. 1999, p. 478, emphasis added). In order to “call attention” to such a carcass, and recruit “the crowd ” that was needed to exploit the situation, some means of going beyond the prior ACS into the realm of displacement had to be found. Once found, that means could be, and was, exploited in a potentially infinite number of ways. A test of the approach proposed here would be to examine the niches of other species in order to determine whetherthey need to practice recruitment, and if so, how they set about it. The prediction is that, given a sufficient subset of the conditions for recruitment described above, any species should develop some precursors of predication and symbolism. Unfortunately, surprisingly few species seem to require recruitment inj the sense used here. Animals that forage co-operatively may signal the presence of food to conspecifics, but only when the source lies within their sensory field, and a standardized call, in isolation, is adequate for this purpose. The only case known to me, outside those discussed above, is that of ravens (Heinrich 1989). Young ravens who have not yet mated or staked out a territory compete for prey carcasses with well-established pairs. To do so their must recruit other birds, and this seems to take place in the nightly roosting areas. However, the means by which this is done remain unclear, and the time-lag between discovery of and access to prey is much longer (a minimum of a whole night) than in the case of bee or ant recruitment, and probably longer than in the case of human ancestors. More research is needed here. But the hypothesis is clear enough: if and only if recruitment is required will a species move beyond the limitations of a standard ACS. If this is so, then recruitment for scanvenging megafauna carcasses represents the only way in which human ancestors could have broken through into language. REFERENCES Arbib, M.A. (2004) From monkey-like action recognition to human language: an evolutionary framework for neurolinguistics. Behavioral and Brain Sciences . Bickerton, D. (1981) Roots of language. Ann Arbor: Karoma. Binford, L.S. (1981) Bones: Ancient men and modern myths. New York: Academic Press. _____ (1985) Human ancestors: changing views of their behavior. Journal of Anthropological Archaeology, 4:292-327. Blumenschine, R.J. (1987) Characteristics of an early hominid scavenging niche. Current Anthropology 28: 383-407. _____, Cavallo, J.A, & Capaldo, S.P. (1994). Competition for carcases and early hominid behavioral ecology. Journal of Human Evolution, 27:197214. Bunn, H.T. & Kroll, E,M, (1986) Systematic butchery by Plio-Pleistocene hominids at Olduvai Gorge, Tanzania. Current Anthropology 27: 431-52. Cavallo, J.A. (1997) A re-examination of Isaac’s central-place foraging hypothesis. Unpublished Ph.D. dissertation, Rutgers University.. Deacon, T. (1997) The symbolic species: The co-evolution of language and the human brain. New York: Norton. Dominguez-Rodrogues, M. (2002) Hunting and scavenging by early humans: The state of the debate. Journal of World Prehistory 16: 1-54. Donald, M. (1991) Origins of the modern mind. Cambridge, Mass.: Harvard University Press. Dumpert, K. (1978) The social biology of ants. Boston: Pitman. Dyer, F.C. & Gould, J.L. (1983) Honey bee navigation. American Scientist 71:587-97. Fitch, T.W., Hauser, M.D., & Chomsky, N. (2005) Evolution of the language faculty: clarifications and implications. Cognition. Frisch, K. von. (1967) Honeybees: do they use direction and distance information provided by their dancers? Science 158: 1072-76. Gould, J.L. (1976) The dance-language controversy. Quarterly Review of Biology 51:211-44. Hangartner, W. (1969) Trail-laying in the subterranean ant Acanthomyops interjectus. Journal of Insect Physiology 15:1-4. Hauser, M.D., Chomsky, N. & Fitch, W.T. (2002) The language faculty: What is it, who has it, and how did it evolve? Science 298.:1569-79. Heinrich, B. (1989) The raven in winter. New York: Random House. Herrnstein, R.J. (1979) Acquisition, generalization, and discrimination reversal of a natural concept. Journal of Experimental Psychology (Animal Behavior Processes. 5:116-29. Hewes, G.W. (1973) Primate communicaton and the gestural origins of language. Current Anthropology 14:5-24. Holldobler, B. (1971) Recruitment behavior in Camponotus socius. Zeitschrift fur Vergleichende Physiologie 75: 123-42. _____ (1978) Ethological aspects of chemical communication in ants. Advances in the Study of Behavior 8:75-115. Irons, D.B., Anthony, R.G. & Estes J.A. (1986) Foraging strategies of glaucouswinged gulls in a rocky intertidal community. Ecology 67: 1460-74. Isaac, G.L. (1978) Food sharing and human evolution: archaeological evidence from the Plio-Pleistocene of East Africa. Journal of Anthropological Research, 34: 311-25. Jackendoff, R. (2002) Foundations of language: brain, meaning, gramar, evolution. New York: Basic Books. Klima, E. & Bellugi, U. (1979) The signs of language. Cambridge, MA: Harvard University Press. Kummer, H. (1968) Social organization of hamadryas baboons. Basel, Switzerland: Karger. Laland, K.N., Odling-Smee, F.J. & Feldman, M.W. (2000) Niche construction, biological evolution and cultural change. Behavioral and Brain Sciences 23: 131-75. Larick, R. & Ciochon, R.L. (1996) The African emergence and early Asia dispersals of the genus Homo. American Scientist 84: 538-51.. Lewis, M.E. (1997) Carnivorean paleoguilds of Africa: implications for hominid food procurement strategies. Journal of Human Evolution 32:257-8 MacNeilage, P.F. (1998) The frame/content theory of the evolution of speech production. Behavioral and Brain Sciences 21:499-546. McHenry, H.M. (1994) Behavioral ecological implications of early hominid body size. Journal of Human Evolution 27.77-88. Mithen, S. (1997) The prehistory of the mind. London: Thames & Hudson. Moglich, M. & Holldobler, B. (1975) Communication and orientation during foraging and emigration in the ant Formica fusca, Journal of Comparative Physiology 101: 275-88. _____, Maschwitz, U. & Holldobler, B. (1974) Tandem calling: a new kind of signal in ant communication. Science 186: 1046-7. Monahan, C.M.(1996) New zooarchaeological data from Bed II, Olduvai Gorge, Tanzania: implications for hominid behavior in the early Pleistocene. Journal of Human Evolution 31:93-128. O’Connell, J.F., Hawkes, K. & Blurton-Jones, N.G. (1999). Granmothering and the evolution of Homo erectus. Journal of Human Evolution 36:461-85. Odling-Smee, F.J., K.N. Laland & M.W. Feldman. (2003) Niche Construction: the neglected process in evolution. Princeton: Princeton University Press. Pinker, S., & Jackendoff. R. (2005) The faculty of language: what’s special about it? Cognition xx.1-36 Reed, K.E. (1997) Early hominid evolution and ecological change through the African Plio-Pleistocene. Journal of Human Evolution 32:289-322. Rose. L. & Marshall, F. (1996) Meat eating, hominid sociality and home bases revisited. Current Anthropology 37: 307-38. Schick, K.D. & Toth, N. (1993) Making silent stones speak: Human evolution and the dawn of technology. New York: Simon and Schuster. Schmitz, O.J. (1992) Optimal diet selection by white-tailed deer: Balancing rproduction with starvation risk. Evolutionary Ecology 6: 125-41. . Stephens, D.W. & Krebs, J.R. (1986) Foraging theory. Princeton, NJ: Princeton University Press. Sudd, J.H., & Franks, N.R. (1987) The behavioral ecology of ants. New York: Chapman & Hall.. Treves, A, & Naughton-Treves, L. (1999) Risk and opportunity for humans coexisting with large carnivores. Journal of Human Evolution 36: 275-82. Velasco, J. & Millan, V.H. (1998) Feeding habits of two large insects from a desert stream: Abedus herberti (Hemiptera: Belostomatidae) and Thermonectus marmoratus (Coleoptera: Dytiscidae). Aquatic Insects 20: Wilson, E.O. (1958) Studies on the ant fauna of Melanesia. I. The tribe Leptogenyini. II. The tribes Amblyoponini and Platythyreini. Bulletin of the Museum of Comparative Zoology, Harvard, 118: 101-53. _____ (1962) Chemical communication in the fire ant Solenopsis saevissima. Animal Behavior 10:134-64. Winston, M.L. (1987) The biology of the honey bee. Cambridge, MA: Harvard University Press. Hauser, Chomsky & Fitch 2002 recursion Pinker & Jackendoff 2005 Fitch, Hauser & Chomsky 2005 19 68 47 symbolization 0 0 0 predication 0 0 0 displacement 0 0 0 Table 1 Frequencies of mention for four significant categories in three recent papers on language evolution Chimpanzees/ Homo Bonobos Predation risk Low High Central place No ??* foraging Fission-fusion Yes Yes social structure Provisioning No ??* Food sources: widely scattered No Yes transient Not often Often diversified Often Yes unpredictable Often Often sizeable No Sometimes Ants Bees High Yes High Yes Yes Yes Yes Yes Usually Often Yes Yes Sometimes Yes Usually No Yes No Table 2 Comparison of recruitment-favoring factors in the niches of primates and hymenoptera *Dates of origin of these functions are still unknown,