Chris Mortensen & Gerard O'Brien (1993) Representation and Causal Asymmetry. Psycoloquy: 4(19) Fodor Representation (5)

Volume: 4 (next, prev) Issue: 19 (next, prev) Article: 5 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 4(19): Representation and Causal Asymmetry

Commentary on Wallis on Fodor-Representation

Chris Mortensen & Gerard O'Brien
Department of Philosophy
The University of Adelaide
Adelaide, SA 5001, Australia


The considerations Wallis brings to bear do not undermine the substantive aspects of Fodor's account as such, but merely suggest that it is in need of more complicated modification.


color vision, Fodor, mind/body problem, perception, representation, semantics, sensory transduction, verificationism
1. One popular approach to the naturalization of intentionality is to equate the semantic properties of mental representations with the information they carry, where the information they carry is unpacked in terms of reliable causal covariance. Information-based approaches to semantics, however, have a notorious problem in accounting for cases of misrepresentation. If mental tokens represent their reliable causes, then as long as such tokens are reliably caused they cannot be false. Yet such false tokens are reliably caused all the time.

2. Jerry Fodor (1987, 1990) has addressed the problem of misrepresentation in informational semantics. He argues that an examination of the counterfactual causal relations that surround the tokening of mental states reveals a special kind of nomic dependency that privileges certain of their causes over the rest. In particular, there is a crucial asymmetry, he argues, between (what can be described as) the `true' and `false' causes of a mental state token, since the false causes of such a token are nomically dependent on the true causes but not conversely. This asymmetry is demonstrated by the fact that breaking the true-cause-to-mental-token link breaks the false link, but breaking the false link does not break the true. Thus, Fodor's suggestion is that while representation cannot be simply reduced to information, it can be reduced to information-plus-asymmetric-dependence.

3. Wallis (1993) argues that Fodor's account of representation gives implausible results when applied to cases of misrepresentation in transduction, feature detection and object recognition. In response, we argue that the considerations Wallis brings to bear do not undermine the substantive aspects of Fodor's account as such, but merely suggest that it is in need of more complicated modification.

4. Wallis's argumentative strategy is to begin with an interpretation of the activity of neural states (cells or cell assemblies) inspired by current information-processing theories in that domain. He then seeks to demonstrate that Fodor's theory of representation is unable to sustain this favoured interpretation because circumstances can be contrived such that the interpretation fails to meet the criterion of asymmetric dependence. We maintain, however, that what Wallis highlights in these cases is that Fodor's condition of asymmetric dependence is trivially false as it stands. There will always be a way of breaking the false-cause-to-mental-token link which breaks the true-cause-to- mental-token link. Apart from the utterly trivial ways (either destroying the organism or breaking the false link by means of breaking the true), one can, because true and false causal chains converge at some point if they produce mental tokens of the same representational type, break the false link anywhere after that point. If the first point of convergence is the mental token itself, break the false link by removing the capacity of the organism to have the mental token. Similarly, there are always trivial ways to break the true link that do not break the false one. Since true and false links diverge in their causes at some point, screen the organism from true causes earlier than that point. If the first point of convergence is reception/transduction at or near the skin surface, modify the external environment so that no true stimuli get through, but false stimuli do.

5. Wallis's objections to Fodor in cases of transduction and object recognition fall into these categories, we claim. Considering light transducers (rods) which are also susceptible to heat, Wallis notes that modifying the transducers (rhodopsin) to prevent false heat stimuli from producing characteristic effects is just as apt to prevent true photon stimuli doing the same as vice versa. He also notes the symmetry between environmental modifications to screen true and false stimuli respectively: each can be achieved independently of the other. Considering colour receptors (cones): Wallis notes that colour-response curves overlap, which permits phantom/false phenomena, but also ensures that there is no physically possible way of changing the transduction properties relative to one colour without doing the same for another colour. Again, environmental modifications obviously will show the same symmetries. Considering object recognition, Wallis notes that breaking the barn-facsimile to barn-representation connection (by breaking the internal causal chain at the rectangle-triangle-red-above to barn- representation connection) also breaks the barn to barn-representation connection. And externally, breaking the barn to barn-representation connection by screening the organism from real barns does not inevitably break the barn-facsimile to barn-representation link.

6. The story is more complicated for the case of edge detection. Edge detectors are made of cells which work by decreasing response to angles further from the given angle, or fade-out over a range rather than a sharp spike at a single angle; but an edge detector might also be identified as an assemblage big enough to have only the sharp spike response. Wallis proposes that the spread of response in the former case eliminates any possibility of asymmetry of Fodor's kind. Now what is to count as false/phantom behaviour here? If the true angle is identified as the peak response, a definite A degrees, then any response of that module to a different angle counts as phantom. Here it would seem that it would, at least in principle, be easier to find some way to tune out a given phantom response than to find some way of tuning out peak response without interfering with a given phantom response (a promissory note, admittedly). If the true angle is rather a preferred central range, then only responses from stimuli outside that range count as phantom; but then there is also the problem of why one should identify the latter as abnormal functioning. If, alternatively, an edge detector is big enough to have a sharp response, then it has to be shown that phantom phenomena do occur. Thus, Wallis's argument around these points would seem to deserve an account of precisely what abnormal response he is claiming symmetry for.

7. But if Fodor's view is trivially false, one should consider the possibility that it was merely expressed carelessly. Two obvious modifications present themselves. First, to avoid the objections which turn on linkage breaking in the external environment one simply rules them out: only breakings of causal chains within the skin are to be considered. After all, causal chains outside the skin threaten asymmetry too easily, because it is too easy to break the true link by appropriate screening without breaking the false link. Second, to avoid the objections which turn on linkage breaking beyond the point of commonality in the causal chains, one interprets Fodor as intending the asymmetry to be as follows: the true link cannot be broken without breaking the false link, but there are ways of breaking the false link which do not break the true link. Then, the existence of ways of breaking both is an insufficient objection.

8. Needless to say, one must show that there are the appropriate asymmetries. But at least for transduction and feature detection, this is not so implausible. Interference with the mechanism in `normal' operation of the sensory mechanism ought to throw the whole system off, while arguably there are always some ways to get rid of false stimuli without obstructing normal operations.

9. The modifications proposed here do not overcome one of Wallis's objections to Fodor's account in the case of object recognition: In the language of possibilia, in any nearby possible world in which barn- facsimiles fail to cause the tokening of barn-representations, real barns would also fail to do so. Consequently, the counterfactuals fail to elicit any asymmetry at this level. This problem with Fodor's account has been highlighted by a number of authors (e.g. Godfrey-Smith, 1989; Sterelny, 1990, Chp.6; Jones, Mulaire and Stich, 1991). Fodor's (1990) response is to go beyond pure informational semantics to seek the necessary asymmetry in the causal history of mental state tokens. That is, representation reduces to asymmetric nomic-dependency-plus-actual-casual-history. Barn-facsimile-caused barn-representations are false because they are historically asymmetrically dependent on barn-caused barn-representations. This provides the ontological depth that cannot be accounted for by causal laws alone.

10. Wallis is dismissive of this "line of retreat," but we find his reasons unconvincing. It is surely unreasonable to require a perceptual mechanism to display asymmetric behaviour with respect to a facsimile barn. An object-recognizer ought to be positively fooled by such cases. So if there is any asymmetric condition to separate the normal operation of a barn recognizer from a barn-representation produced by a facsimile barn, it will be causal-historical asymmetry. Moreover, much of Wallis's disquiet at this point seems to be based on a misunderstanding of the explanatory role of mental content in computational cognitive science. He seems to suggest that any theory of mental representation that implicates the causal history of mental states is incompatible with a computational theory of object recognition (e.g. Biederman, 1987). Now this incompatibility would only exist if the content of a representational state played a direct causal role in mental processes. But, according to computational cognitive science at least, mental processes do not have access to the semantic properties of the mental states over which they are defined. Hence the role of mental content is indirect, but quite compatible with a theory of representation that defers to causal history.


Biederman, I. 1987, Recognition by Components: A Theory of Image Understanding, Psychological Review 94, 115-147.

Fodor, J.A. 1987, Psychosemantics, Cambridge, Ma, MIT Press.

Fodor, J.A. 1990, A Theory of Content II: The Theory. In A Theory of Content and Other Essays, Chapter 4, Cambridge, Ma, MIT Press.

Godfrey-Smith, P. 1989, Misinformation, Canadian Journal of Philosophy, 19, 533-550.

Jones, T., Mulaire, E. and Stich, S. 1991, Starving off Catastrophe: a Critical Notice of Jerry Fodor's Psychosemantics, Mind and Language, 6, 58-82.

Sterelny, K. 1990, The Representational Theory of Mind, Oxford, Blackwell

Wallis, C. (1992) Asymmetric Dependence and Mental Representation. PSYCOLOQUY 3(70) fodor-representation.1.

Volume: 4 (next, prev) Issue: 19 (next, prev) Article: 5 (next prev first) Alternate versions: ASCII Summary