REVIEW
JOSÉ LUIS BERMÚDEZ
Oxford & New York: Oxford University Press, 2003, cloth £25.00 / US$35.00
ISBN: 0 19 515969 1
Peter Carruthers
Thinking without Words is about the extent to which infants and non-human animals are capable of thought. As one might expect from an author whose previous book presupposed that there can be no conceptual thinking without language (The Paradox of Self-Consciousness, MIT 1998), the approach taken here is an extremely cautious one. Bermúdez argues that some animals are capable of something like thought, and that such animals are at best capable of a kind of proto-rationality significantly unlike human rationality. (Even the sort of intentional content that animal thoughts are said to have individuated via a form of success semantics is claimed to differ from the kind of content that human thoughts possess.)
Roughly the first half of the book is devoted to convincing us that there really are such phenomena as non-linguistic forms of thinking, construed realistically to involve discrete causally-effective structured content-bearing states. Bermúdez shows how attributions of such thoughts to animals and infants are warranted by the evidence, works out in some detail what such attributions might require, and replies to a range of possible objections to the idea. This half of the book is for the most part successful, and deserves to be widely read, especially within Oxford-influenced / Wittgenstein-influenced circles, where it is taken for granted that thought has some sort of important intrinsic connection with natural language.
The second half of the book is devoted to tracing the limits of languageless thought. This is much less convincing. Bermúdez, like a number of others who write on this topic, assumes that genuine thought requires instrumental rationality. And he likewise assumes that behaviors that dont manifest this sort of causal thinking will either be innate fixed action schemata or the product of some sort of associative conditioning. Both assumptions are deeply flawed, however, as I shall now briefly try to show. (For further elaboration, see my article On being simple minded, American Philosophical Quarterly, 2004; see also the references contained therein for many of the facts that follow.)
The basic point is that Bermúdez ignores the cognitive processes that are responsible for spatial navigation (at least in this context he does he actually does consider navigation data when discussing how one might distinguish between different modes of presentation amongst non-linguistic thoughts). These navigational processes are widely distributed throughout the animal kingdom from ants and bees through to birds and mammals and involve a number of distinct kinds of non-associative learning mechanism. Many animals use the sun, or the polarization of the suns light in the sky, to calculate a compass orientation. This means that they have to be capable of learning the solar ephemeris the arc that the sun makes through the sky at a given latitude and time of year in order that they should compute the direction of South, given the suns position in the sky and the time of day.
Many animals use dead reckoning, integrating measures of distance traveled in each direction with the angle of each of their turns in order to calculate the distance and direction of home. The Tunisian desert ant, for example, can travel as much as a hundred meters across a featureless desert, zigzag back and forth in a highly complex path in search of food, crossing terrain never before traveled, and then turn and head in a straight line back to its nest when it finds a food item, being accurate to within centimeters. Then when it returns to that same cache of food, it doesnt need to traverse its original route, but heads straight for it.
This is plainly an example of learning, not any sort of fixed action pattern the ant makes many unique forays, and has to figure out where it is afresh on each occasion. And it plainly isnt the result of any sort of associative conditioning. Moreover, the learning mechanism needed to do dead reckoning would obviously need to be different from the mechanism that learns the solar ephemeris and uses that information to figure out the direction of South (although the former mechanism presupposes the latter one the Tunisian ant, too, uses the sun to calculate directions of travel, which are taken as inputs to the dead reckoning process).
Many other animals (including all birds and mammals, and perhaps even bees) can construct a cognitive map of their environment, on which the relative positions of a range of salient landmarks are represented. (Note that the representational vehicles of mental maps need not be at all map-like. They might consist of sentences specifying, in respect of each landmark, the distance and direction of some of the other landmarks, together with an inference mechanism that can compute novel spatial relationships from such information.) When navigating to a target on the map such as a food cache, or home base the animal does so by identifying one or more landmarks, and calculating from that the direction in which it needs to travel.
These forms of navigation can serve to ground a simple kind of belief / desire psychology. For in each case, we have a range of structured belief states capable of interacting with a range of different goal-states (for example, to forage, or to return home) in such a way as to determine behaviors that may well be unique in the life of the animal. And moreover, they involve what seem to be genuine forms of decision-making as when, for example, an animal makes a choice between two potential foraging sites on the grounds of the distance that would need to be traveled from its current location.
Bermúdez insists, in contrast, that the minimal requirement of genuine decision-making is that the selection of a particular course of action from the contrast space of alternative possible courses of action should be made on consequence-sensitive grounds (p.124). By this criterion rats engage in genuine decision-making, since their behavior has been shown to be sensitive to the causal efficacy of their own actions; as do tool-using ravens, apes, and hominids. But vast swathes of animal behavior will not count as resulting from genuine decision-making, on this account.
Reflection on the navigation cases shows, however, that Bermúdezs criterion is unmotivated and unnecessarily restrictive. Why must we say that a bird choosing between two potential food caches that are represented in distinct locations on its mental map, for example, isnt taking a genuine decision, merely on the grounds that the causality of the birds own behavior isnt explicitly represented? For humans, too, in such circumstances wont explicitly represent the causality of their own actions, either. When I am thirsty, realize that there is cold beer in the fridge, and walk to the appropriate location, it seems unlikely that my decision is taken on consequence-sensitive grounds. Rather, I just put my goal together with information about the location of something that would satisfy that goal, and then implement an action-schema (walking to the fridge) that is guided in its execution by the location information. I dont need to think of walking as something that will cause the satisfaction of my goal. Likewise, then, with the bird: the fact that it doesnt explicitly represent the causal properties of flight need not prevent it from counting as having engaged in genuine decision-making.
When it comes to characterizing the sort of rationality of which non-linguistic creatures are capable, Bermúdez diagnoses two ways in which the rationality of animals falls short of our own hence only counting as a form of proto-rationality or proto-logic. One is that they dont engage in calculations of maximum expected utility (p.138). The other is that they dont engage in reflection on what outcomes would satisfy their desires, together with the likelihood that the different courses of action have of generating those outcomes (p.148). But as regards the first, there is now an extensive body of literature demonstrating that humans dont engage in calculations of maximum utility, either, but rather deploy a range of simpler decision-making heuristics. (See, e.g., G. Gigerenzer and P. Todd, Simple Heuristics that Make us Smart, OUP 1999.) And as regards the second, Bermúdez mistakes the basic form of the practical reasoning syllogism, which isnt (as he alleges) a meta-representational one.
One basic kind of practical reasoning syllogism can be represented thus: DES [P], BEL [if I do Q then P], so INTEND [I do Q]. Here the operators DES, BEL and INTEND are meant to designate the first-order attitudes of desire, belief and intention that figure in the episode of practical reasoning. They are not supposed to be second-order representations of these states figuring within the content of the agents reasoning process. The only contents that so figure are P, I do Q, and the conditional, if I do Q then P. While humans can engage in meta-representational forms of practical reasoning thinking, not just about items wanted and how to get them, but also about our own desires themselves and how reliable our beliefs are there is no reason to believe that this is required for practical reasoning as such, nor for thinking that all or most of our practical reasoning takes such a form.
In the final two chapters of the book Bermúdez argues that all higher-order, meta-representational, thinking depends upon language. But he moves illicitly from a claim that many of us might grant namely, that a certain sort of reflexive thinking involves language (viz., conscious thinking where our thoughts are themselves immediately available to further thought) to the claim that all higher-order thought involves language. From the fact that some of our conscious thinking has natural language vehicles, it doesnt follow that our basic capacity to attribute thoughts to others or to ourselves is similarly dependent upon language. And indeed, there is direct evidence from some cases of aphasia that a capacity for higher-order thought can survive the destruction of the language faculty.
In sum, the evidence actually shows (contra Bermúdez) that attitudes of belief and desire, together with the most fundamental forms of inference and practical reasoning, are well neigh ubiquitous throughout the animal kingdom. While there are, no doubt, some fancy forms of cognition that are unique to humans, and that depend upon language, neither thinking nor higher-order thinking is one of them, and neither is decision making.