The Debated Mind

Evolutionary Psychology versus Ethnography

Harvey Whitehouse

• Publisher: Routledge
• Year: 2020
• ISBN:
  9781000180862
• Edition: 1
• Topic: Social Sciences
• Subtopic: Anthropology
• Index: Social Sciences

In a further development of the nature-nurture debate, this collection of articles questions how the human mind influences the content and organization of culture. In the study of mental activity, can the effects of evolution and history be teased apart? Evolutionary psychologists argue that cultural transmission is constrained by our genetic inheritance. Few social and cultural anthropologists have found this argument to be relevant to their work and many would doubt its validity. This book uniquely pitches the arguments for innatism against ethnographic perspectives that call into question the theoretical foundations of orthodox evolutionary biology and cognitive science. Ultimately the aim of the debate is to create an original set of mutually compatible theories that will open up new areas for interdisciplinary research.

PART ONE

1
Mental Modularity and Cultural Diversity¹

Dan Sperber

In The Modularity of Mind, published in 1983, Jerry Fodor attacked the then dominant view according to which there are no important discontinuities between perceptual processes and conceptual processes. Information flows freely, 'up' and 'down', between these two kinds of processes, and beliefs inform perception as much as they are informed by it. Against this view, Fodor argued that perceptual processes (and also linguistic decoding) are carried out by specialized, rather rigid mechanisms. Each of these 'modules' has its own proprietary data base, and does not draw on information produced by conceptual processes.

Although this was probably not intended and has not been much noticed, 'modularity of mind' was a paradoxical title, for, according to Fodor, modularity is to be found only at the periphery of the mind, in its input systems.² In its centre and bulk, Fodor's mind is decidedly nonmodular. Conceptual processes - that is, thought proper - are presented as a big holistic lump lacking joints at which to carve. Controversies have focused on the thesis that perceptual and linguistic decoding processes are modular, much more than on the alleged nonmodularity of thought.

In this chapter, I have two aims. My first aim is to defend the view that thought processes might be modular too (what Fodor [1987: 27] calls 'modularity theory gone mad' - oh well!). Let me however echo Fodor and say that, 'when I speak of a cognitive system as modular, I shall.. . always mean "to some interesting extent'" (Fodor, 1983: 37). My second aim is to articulate a modular view of human thought with the naturalistic view of human culture that I have been developing under the label 'epidemiology of representations' (Sperber, 1985b). These aims are closely related: Cultural diversity has always been taken to show how plastic the human mind is, whereas the modularity of thought thesis seems to deny that plasticity. I want to show how, contrary to the received view, organisms endowed with truly modular minds might engender truly diverse cultures.

Two commonsense arguments against the modularity of thought

Abstractly and roughly at least, the distinction between perceptual and conceptual processes is clear: Perceptual processes have, as input, information provided by sensory receptors and, as output, a conceptual representation categorizing the object perceived. Conceptual processes have conceptual representations both as input and as output. Thus seeing a cloud and thinking 'here is a cloud' is a perceptual process. Inferring from this perception 'it might rain' is a conceptual process.

The rough idea of modularity is also clear: A cognitive module is a genetically specified computational device in the mind/brain (henceforth: The mind) that works pretty much on its own on inputs pertaining to some specific cognitive domain and provided by other parts of the nervous systems (e.g., sensory receptors or other modules). Given such notions, the view that perceptual processes might be modular is indeed quite plausible, as argued by Fodor. On the other hand, there are two main commonsense arguments (and several more technical ones) that lead one to expect conceptual thought processes not to be modular.

The first commonsense argument against the modularity of thought has to do with integration of information. The conceptual level is the level at which information from different input modules, each presumably linked to some sensory modality, gets integrated into a modality-independent medium: a dog can be seen, heard, smelled, touched, and talked about. The percepts are different, the concept is the same. As Fodor points out,

the general form of the argument goes back at least to Aristotle: The representations that input systems deliver have to interface somewhere, and the computational mechanisms that effect the interface must ipso facto have access to information from more than one cognitive domain (Fodor, 1983: 101-2).

The second commonsense argument against the modularity of thought has to do with cultural diversity and novelty. An adult human's conceptual processes range over an indefinite variety of domains, including party politics, baseball history, motorcycle maintenance, Zen Buddhism, French cuisine, Italian opera, chess playing, stamp collecting, and Fodor's chosen example, modern science. The appearance of many of these domains in human cognition is very recent and not relevantly correlated with changes in the human genome. Many of these domains vary dramatically in content from one culture to another, or are not found at all in many cultures. In such conditions, it would be absurd to assume that there is an ad hoc genetically specified preparedness for these culturally developed conceptual domains.

These two commonsense arguments are so compelling that Fodor's more technical considerations (having to do with 'isotropy', illusions, rationality, etc.) look like mere nails in the coffin of a dead idea. My goal will be to shake the commonsense picture and to suggest that the challenge of articulating modularity, conceptual integration, and cultural diversity may be met and will turn out to be a source of psychological and anthropological insights.

Notice, to begin with, that both the informational integration argument and the cultural diversity argument are quite compatible with partial modularity at the conceptual level

True, it would be functionally self-defeating to reproduce at the conceptual level the same domain partition found at the perceptual level, and have a different conceptual module treat separately the output of each perceptual module. No integration whatsoever would take place, and the dog seen and the dog heard could never be one and the very same mastiff Goliath. But who says conceptual domains have to match perceptual domains? Why not envisage, at the conceptual level, a wholly different, more or less orthogonal domain partition, with each domain-specific conceptual mechanism getting its inputs from several input mechanisms? For instance, all the conceptual outputs of perceptual modules that contain the concept MASTIFF (and that are therefore capable of recognizing the presence of a mastiff) might be fed into a specialized module (say a domain-specific inferential device handling living-kind concepts) which takes care (inter alia) of Goliath qua mastiff. Similarly, all the conceptual outputs of input modules which contain the concept THREE might be fed into a specialized module which handles inference about numbers, and so forth. In this way, information from different input devices might get genuinely integrated, though not into a single, but into several conceptual systems.

Of course if you have, say, a prudential rule that tells you to run away when you encounter more than two bellicose dogs, you would not really be satisfied to be informed by the living-kinds module that the category BELLICOSE DOG is instantiated in your environment, and by the numerical module that there are more than two of something. Some further, at least partial, integration had better take place. It might even be argued - though that is by no means obvious - that a plausible model of human cognition should allow for full integration of all conceptual information at some level. Either way, partial or full integration might take place further up the line, among the outputs of conceptual rather than of perceptual modules. Conceptual integration is not incompatible with at least some conceptual modularity.

Similarly, the conceptual diversity argument implies that some conceptual domains (expertise in postage stamps for instance) could not be modular. It certainly does not imply that none of them could be. Thus, in spite of superficial variations, living-kind classification exhibits strong commonalities across cultures (see Berlin, 1978) in a manner that does suggest the presence of a domain-specific cognitive module (see Atran, 1987, 1990).

The thesis that some central thought processes might be modular gets support from a weal th of recent work (well illustrated in Hirschfeld and Gelman 1994) tending to show that many basic conceptual thought processes, found in every culture and in every fully developed human, are governed by domain-specific competences. For instance, it is argued that people's ordinary understanding of the movements of an inert solid object, of the appearance of an organism, or of the actions of a person are based on three distinct mental mechanisms, a naive physics, a naive biology, and a naive psychology (see for instance Atran 1987,1994; Carey, 1985; Keil 1989,1994; Leslie 1987,1988,1994; Spelke 1988). It is argued moreover that these mechanisms, at least in rudimentary form, are part of the equipment that makes acquisition of knowledge possible, rather than being acquired competences.

Accepting as a possibility some degree of modularity in conceptual systems is innocuous enough. Jerry Fodor himself recently considered favourably the view that 'intentional folk psychology is, essentially, an innate, modularized database' (Fodor, 1992: 284 - italics added) without suggesting that he was thereby departing from his former views on modularity. But what about the possibility of massive modularity at the conceptual level? Do the two commonsense arguments, integration and diversity, really rule it out?

Modularity and evolution

If modularity is a genuine natural phenomenon, an aspect of the organisation of the brain, then what it consists of is a matter of discovery, not stipulation. Fodor himself discusses a number of characteristic and diagnostic features of modularity. Modules, he argues, are 'domainspecific, innately specified, hardwired, autonomous' (1983: 36). Their operations are mandatory (p. 52) and fast (p. 61); they are' informationally encapsulated' (p. 64), that is, the only background information available to them is that found in their proprietary data base. Modules are 'associated with fixed neural architecture' (p. 98). Fodor discusses still other features that are not essential to the present discussion.

There is one feature of modularity that is implied by Fodor's description, but that he does not mention or discuss. If, as Fodor argues, a module is innately specified, hardwired, and autonomous, then it follows that a cognitive module is an evolved mechanism with a distinct phylogenetic history. This is a characteristic, but hardly a diagnostic feature, because we know close to nothing about the actual evolution of cognitive modules. But I have been convinced by Leda Cosmides and John Tooby (see Cosmides 1989, Cosmides and Tooby 1987, 1994; Tooby and Cosmides 1989, 1992)³ that we know enough about evolution on the one hand and cognition on the other to come up with well-motivated (though, of course, tentative) assumptions as to when to expect modularity, what properties to expect of modules, and even what modules to expect. This section of the chapter owes much to their ideas.

Fodor himself does mention evolutionary considerations, but only in passing. He maintains that, phylogenetically, modular input systems should have preceded nonmodular central systems:

Cognitive evolution would thus have been in the direction of gradually freeing certain sorts of problem-solving systems from the constraints under which input analysers labour - hence of producing, as a relatively late achievement, the comparatively domain-free inferential capacities which apparently mediate the higher flights of cognition (Fodor 1983: 43).

Let us spell out some of the implications of Fodor's evolutionary suggestion. At an early stage of cognitive evolution we should find modular sensory input analysers directly connected to modular motor controllers. There is no level yet where information from several perceptual processes would be integrated by a conceptual process. Then there emerges a conceptual device, that is, an inferential device that is not itself directly linked to sensory receptors. This conceptual device accepts input from two or more perceptual devices, constructs new representations warranted by these inputs, and transmits information to motor-control mechanisms.

Initially, of course, this conceptual device is just another module: It is specialized, innately wired, fast, automatic, and so forth. Then, so the story should go, it grows and becomes less specialized - possibly it merges with other similar conceptual devices - to the point where it is a single big conceptual system, able to process all the outputs of all the perceptual modules, and able to manage all the conceptual information available to the organism. This true central system cannot, in performing a given cognitive task, activate all the data accessible to it, or exploit all of its many procedures. Automaticity and speed are no longer possible. Indeed, if the central system automatically did what it is capable of doing, this would trigger a computational explosion with no end in sight.

An evolutionary account of the emergence of a conceptual module in a mind that had known only perceptual processes is simple enough to imagine. Its demodularization would be much harder to explain.

A toy example might go like this: Organisms of a certain species, call them 'protorgs', are threatened by a danger of a certain kind. This danger (the approach of elephants that might trample the protorgs, as it might be) is signalled by the co-occurrence of a noise N and soil vibrations V. Protorgs have an acoustic perception module that detects instances of N and a vibration-perception module that detects instances of V. The detection either of N by one perceptual module or of V by the other activates an appropriate flight procedure. Fine, except that when N occurs alone, or when V occurs alone, it so happens that there is no danger. So protorgs end up with a lot of 'false positives', uselessly running away, and thus wasting energy and resources.

Some descendants of the protorgs, call them 'orgs', have evolved another mental device: A conceptual inference mechanism. The perceptual modules no longer directly activate the flight procedure. Rather their relevant outputs, that is, the identification of noise N and that of vibrations V, go to the new device. This conceptual mechanism acts essentially as an AND-gate: When, and only when both N and V have been perceptually identified, does the conceptual mechanism get into a state that can be said to represent the presence of danger, and it is this state that activates the appropriate flight procedure.

Orgs, so the story goes, competed successfully with protorgs for food resources, and that is why you won't find protorgs around. The orgs' conceptual mechanism, though not an input module, is nevertheless a clear case of a module: It is a domain-specific problem-solver; it is fast, informationally encapsulated, associated with fixed neural architecture, and so forth. Of course, it is a tiny module, but nothing stops us from imagining it becoming larger: Instead of accepting just two bits of information from two simple perceptual modules, the conceptual module could come to handle more from more sources, and to control more than a single motor procedure, but still be domain-specific, automatic, fast, and so on.

At this juncture, we have two diverging evolutionary scenarios on offer. According to the scenario suggested by Fodor, the conceptual module should evolve towards less domain-specificity, less informational encapsulation, less speed, and so on. In other words, it would become less and less modular, possibly merging with other demodularized devices, and ending up like the kin...

Table of contents

• Cover
• Half Title
• Dedication Page
• Title Page
• Copyright Page
• Contents
• List of Contributors
• Introduction
• PART ONE
• PART TWO
• Conclusion: Towards a Reconciliation
• Index