Graham Macdonald and David Papineau, eds.: Teleosemantics.
Toronto and New York: Oxford University
US$87.72 (cloth ISBN-13: 978-0-19-927026-2); Cdn$45.95/US$35.00
(paper ISBN-13: 978-0-19-927027-9).
Teleosemantics is one of the more promising strategies for providing a naturalistic account of mental representation, primarily because it accounts for misrepresentation, something naturalistic theories have always foundered on. Suppose that, as indicator semantics has it, the meaning of a mental state is the property that tokens of the state carry information about--that 'dog' means dog because tokens of the former are caused only by instances of the latter. The problem is that 'dog' tokens are sometimes triggered by non-dogs, as when a large cat is glimpsed in bad light. How, then, can we avoid the conclusion that 'dog' means not dog, but dog or cat. How, that is, can we account for the fact that 'dog' continues to mean dog even when it's misapplied to cats? This is known as the disjunction problem, and it's a doozy.
Teleosemantics solves the disjunction problem by appealing to functions. If the meaning of a mental representation is the property it has the function of indicating rather than the property it actually indicates, misrepresentation becomes a kind of malfunctioning. Just as sperm cells malfunction when they fail to fertilize an ovum, so do 'dog' tokens when triggered by cats. What's left is only to provide a naturalistic analysis of function, which teleosemanticists do by appealing to natural selection. The function of sperm is to fertilize ova because that's what sperm cells were selected for. Similarly, the function of 'dog' is to indicate dogs because that's what it was selected for.
All of this is helpfully explained by Macdonald and Papineau in their introduction to this collection of essays written by leading figures in teleosemanitics. The volume begins with an essay by Kim Sterelny on the evolution of language--an issue of obvious interest to the teleosemanticist. Sterelny challenges the widespread view that linguistic competence is the product of a specialized 'language module', which is innately determined and encapsulated from other parts of the mind. His best objection appeals to what he calls 'the invasion problem' (30). Since modules are insensitive to environmental stimuli, the languages to which they give rise would function like fixed conceptual schemes; distinct modules would therefore give rise to incommensurable schemes. Assuming that linguistic competence is selected for facilitating communication, lone variants would be selected against, making it difficult to see how mutations to modules could gain a foothold in an existing population. Thus, defenders of modularity have a difficult time explaining how language modules could have evolved, which is particularly ironic given that modularists typically appeal to evolutionary considerations in defense of their thesis.
In Chapter 2, Peter Godfrey-Smith notes the recent loss of faith in naturalistic theories of content and proposes that the time is right for a reassessment of the representational theory of mind. Teleosemantics is a species of representationalism, so any objections to the latter apply with equal force to the former. However, Godfrey-Smith's strongest challenge is directed not so much at representationalism per se, but at versions that 'explain intelligence by giving the mind access to something with the same structure as its target' (54). The problem: 'If the mind's problem is dealing with things that exhibit (structure) S, how does it help to put something with S inside the head? The mind still has to detect and respond to S, just as it did when S was outside' (54). At best, this objection gives us reason to reject resemblance theories, but, as we've seen, it's possible for teleosemanticists to defend biologized versions of indicator semantics. Since indication is not resemblance, Godfrey-Smith's objection misses its intended target.
With the next two essays, the focus turns to teleosemantics proper. In Chapter 3, Fred Dretske attempts to ameliorate the difficulty teleosemantics has with self-knowledge. If the content of our thoughts is determined by their selection history, then knowing what we're thinking would require empirical knowledge of the (often) remote past. Thus teleosemantics appears to make introspection impossible. Dretske valiantly attempts to make all of this seem less absurd, but the consolation he offers (introspection cannot reveal that we are thinking, but it can reveal what we are thinking) would probably be rejected by anyone with passing knowledge of Descartes.
In Chapter 4, Frank Jackson shows that teleosemantics yields similar paradoxes concerning our knowledge of other minds. Since we don't know the selection histories of other peoples' mental states any better than our own, it follows that most of us know nothing about the minds of others and that nobody had such knowledge prior to Darwin. Again, this is a bitter pill to swallow, and Jackson doesn't make it any easier. He defends this epistemological objection against various replies, and does so convincingly.
As Ruth Millikan observes in Chapter 5, teleosemantics has trouble explaining contents that don't have biological utility. This is a problem because it's difficult to see how, for example, Rover's representation of a ball helps him propagate his genes. Millikan attempts to defuse this worry by arguing that instrumental learning and practical reasoning produce their own sorts of biological purposes, derived from but not reducible to the 'purposes' of genes. Consider operant conditioning. There is an obvious similarity between the psychological process of reinforcement and the biological process of selection. In each case, randomly generated variations (stimulus-response pairs in the former case, phenotypic modifications in the latter) are selected for based on their agreement with environmental variables. Moreover, since the mechanisms of conditioning were naturally selected for producing stimulus-response correlations, there's a sense in which the representation of a given stimulus has the derivative function of eliciting its conditioned response. Thus, it's not too much of a stretch to speak of conditioning producing acquired content just as selection produces innate content. Rover represents balls when retrieving because nature has selected the mechanisms of conditioning for producing stimulus-response pairs and because his conditioning regime reinforced retrieval-behavior only when caused by ball-stimuli. More sophisticated content can be explained by appealing to the mechanisms of practical reasoning and compositional semantics.
In Chapter 6, Dan Ryder applies some of these themes to the difficult problem of representing kinds. It would seem that in order for the representation of a kind ([H.sub.2]O) to exclude superficially similar kinds (XYZ) it would need to include something like an intention to refer to kinds. But, as Ryder observes, 'if concepts of particular kinds are difficult to account for ... the concept of kindhood seems even more difficult' (118). Ryder's solution is to describe a class of neuronal structures that have the function of indicating kinds and that, when properly calibrated, have the derivative function of indicating specific kinds, thus explaining how kinds can be represented without explicit representations of kindhood. Ryder's essay is an example of naturalistic philosophy at its best, expertly weaving together themes from teleosemantics and his own account of neuronal functioning.
Mohan Matthen (Chapter 7) and Karen Neander (Chapter 8) come down on opposite sides of the consumer semantics debate. Consumer semantics explains the content of a state in terms of the functionally appropriate responses to it. The problem is that for higher organisms there doesn't seem to be a single appropriate response to a given perceptual state. What, exactly, is the appropriate response to an apple or to the color blue? Matthen responds by locating appropriate responses within the sensing organism rather than in bodily action. The appropriate response to an apple is a certain change of epistemic state, which may be as simple as changing the weight on a synaptic connection. Since actions occur as a result of the interactions between epistemic states and other cognitive states, context-dependence is explained without sacrificing univocality.
Neander argues that the content attributed by consumer-oriented analyses of this sort is at odds with the content required by cognitive science. When we apply a consumer-oriented analysis to the frog's representation of flies, we get something like 'frog food', but when we consider the informational content, we get something closer to 'small, dark, moving object'. Neander argues that the latter analysis is better because only informational content explains behavior, and it's behavior that cognitive science aims to explain. Neander's essay is a welcome reminder that the goal is to provide not only naturalistic analyses, but analyses in the service of scientific explanation. Naturalists are supposed to be making the mind safe for science, so analyses that do not cohere with the sciences are otiose, whatever their naturalistic bona fides.
The volume concludes with Robert Cummins et al. (Chapter 9) arguing that teleosemantics cannot account for 'unexploited content'--'content a representation has, but which the system that harbors it is currently unable to exploit' (195)--and with Carolyn Price (Chapter 10) arguing that teleosemantics has the unique resources to account for the not-quite-fully-cognitive content of our emotional states.
Teleosemantics is probably not the best introduction to the subject, although the introductory essay by Macdonald and Papineau is quite good. Many of the essays presuppose familiarity with the literature, and the papers are not fully representative of the major issues confronting the project--the problem of functional indeterminacy, for example, gets short shrift. It is, however, an excellent peek into the current state of the art and required reading for anyone interested in how things now stand with the project of providing a naturalistic account of mental representation. The impression one is left with is that new problems are piling up faster than old ones can be solved.
|Printer friendly Cite/link Email Feedback|
|Publication:||Philosophy in Review|
|Date:||Oct 1, 2008|
|Previous Article:||Leonard Lawlor: This is Not Sufficient: An Essay on Animality and Human Nature in Derrida.|
|Next Article:||Iain Macdonald and Krzysztof Ziarek, eds.: Adorno and Heidegger: Philosophical Questions.|