Christopher D. Green (1998) Idealisation is Fine; Opportunism is not. Psycoloquy: 9(37) Connectionist Explanation (30)

Volume: 9 (next, prev) Issue: 37 (next, prev) Article: 30 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 9(37): Idealisation is Fine; Opportunism is not

Reply to Opie on Connectionist-Explanation

Christopher D. Green
Department of Psychology
York University
Toronto, Ontario M3J 1P3


Opie (1998) seems to accept my worries about the ontological base of connectionism and resolves them by accepting my suggestion that neurology is the ground of connectionist cognitive science. He goes on to argue that scientists should be given some room to idealise the entities under study. One cannot agree. However, he seems to begs the question in suggesting that units and connections are the very things that units and connections model. There also appear to be some problems with the use of sources in support of his position.


artificial intelligence, cognition, computer modelling, connectionism, epistemology, explanation, methodology, neural nets, philosophy of science, theory.
1. There is relatively little to disagree with in Opie's (1998) commentary on Green (1998). He stresses the fact that scientists often idealise the entities they study when incorporating them into theories. Who would dispute this? He does, however, seem altogether too cavalier in his use of this observation. Idealisation can be a very tricky game. One person's "idealisation" is all too often another's unjustified assumption. That it worked out for Newton (for a while, anyway) has no bearing whatever on whether it will work out for connectionists. Moreover, an idealisation that blatantly conflicts with the known causal powers of the object of study (e.g., the indiscriminate use of "dual-action" nodes throughout) seems very much like an unjustified assumption MASQUERADING as an idealization in the hope that some solution will be found before the mask is torn away.

2. On the whole, however, Opie essentially seems to have accepted my main point, which was that connectionist modeling of cognitive processes is probably, in principle, research into brain function; that nodes and units will ultimately cash out as neurons or -- as already noted a number of times in this discussion, although many critics have tried to "straw-man" me by ignoring it -- some alternative form of neurological units (or it won't cash out at all). Indeed, Opie seems to hold to the neurological leanings of the position even more strongly than I: he claims that "the neural heritage is crucial, because if connectionism is not in the business of modelling the brain, then it is hard to take seriously as a source of COGNITIVE theory" (footnote 2). I myself would not go nearly this far. Traditional "folk psychology" certainly is not in the business of "modelling the brain." Instead, it seeks to model the inferential-cum-causal relations among propositional attitudes. but whatever problems it may have, it has not proven "hard to take seriously." Indeed the scorn regularly heaped upon it by connectionists shows just how serious a contender it is (or, was until very recently). (By way of comparison, connectionists do not spend much time criticizing, say, the Medieval ventricular theory of cognition.)

3. To pre-empt what is likely to be the immediate response of many to this point, it is important to note here that the fact that beliefs and desires are thought to be instantiated in the brain does NOT make folk psychology a brain-modeling theory. Consider that if it turned out that we had been entirely wrong about the brain, and found that beliefs and desires were instantiated instead in, say, the liver, folk psychology would need to change little in order to accommodate this discovery. In contrast, connectionism -- precisely because it is primarily, even if covertly, a theory of the neurological function underlying cognition -- would have to change very dramatically (probably by attempting to develop some other, as yet unknown, ontological basis for nodes and connections). In the absence of such an account, it is only plausible as long as it holds out the promise that its theoretical entities will ultimately be cashed out somehow in terms of brain activity.

4. Opie tries, as I have noted about so many other connectionists, to back away from this conclusion just as it looms largest: he writes, "PDP does not engage a microstructurally realistic model of the brain, but rather a model that captures certain high-level characteristics of neural networks (those described in terms of units, connections, connection weights, activation patterns, and so forth)" (para. 6). This is just question- begging. What makes Opie believe that there IS such a level of description about brain activity? The whole thrust of my target article was that we simply have no idea what units, connections, and the like might be OUTSIDE OF THE VERBAL CIRCLE OF THE THEORY (unless they be something neurological). Put bluntly, what are they about? Are they about beliefs and desires? Most certainly not, connectionists leap to tell us. Are they about neurons, then? Well, sort of, but not really. This will not do.

5. Despite what seems to be our basic agreement, I must take Opie to task a little for his use of sources. Invoking Imre Lakatos (1978) as an authority in the history of science is a questionable move, at best. Important as his philosophy of science might have been, he has, in the main, been dismissed as a dabbler by most serious historians of science. Gerald Holton (1978, p. 106), for instance, once described his historical work as "an historical parody that makes one's hair stand on end." Ian Hacking (1983), quoting the passage, mooted only that "many colleagues agree" (p. 122).

6. From so conservative a thinker as Lakatos, Opie rapidly swings to the opposite extreme of the spectrum, citing the work of Nancy Cartwright (1983). Opie is, of course, welcome to adopt Cartwright's radical views on science, but I seriously doubt that many connectionists are going to come along for the ride as he explains to them "How the findings (I dare not say "laws") of connectionist cognitive science lie" (if the allusion is obscure, see the title of Cartwright's book). I must also question the coherence of seeming to endorse simultaneously the so obviously CLASHING views of a Lakatos, on the one hand, and a Cartwright, on the other. To put things succinctly, Lakatos believed that there is, at root, a deductive basis for scientific progress. Cartwright (most emphatically) does not. Which does Opie endorse?

7. Finally, Opie's use of Dennett's work in support of his point strikes me as odd. It may well be that "realism is not an all or nothing affair" (Opie 1998, para. 4), but using the most controversial part ("the patterns are real") of a most controversial theory of mind (Dennett's) does not seem to be the stuff of authority. Indeed, even Dennett (1991, p. 27) concedes that his position "has been attacked from both sides of the ontological dichotomy" (p. 27). He says he finds this "amusing." I, by contrast, find it telling. Realism is, indeed, a very flexible doctrine. Hacking (1983) would appear to have presented a much more believable, much less opportunistic, account of its parameters.


Cartwright, N. (1983) How the Laws of Physics Lie. Oxford University Press.

Dennett, D.C. (1991) Real patterns. Journal of Philosophy 88:27-51

Green, C.D. (1998) Are connectionist models theories of cognition? PSYCOLOQUY 9(4)

Hacking, I. (1983) Representing and Intervening. Cambridge: Cambridge University Press.

Holton, G.J. (1978) The scientific imagination: Case studies. Cambridge: Cambridge University Press.

Lakatos, I. (1978) The methodology of scientific research programmes. J. Worrall & G. Currie (eds.). Cambridge University Press.

Opie, J. (1998). Connectionist modeling strategies: Comment on Green on connectionist-explanation. PSYCOLOQUY 9(30) psyc.98.9.30.connectionist-explanation.27.opie

Volume: 9 (next, prev) Issue: 37 (next, prev) Article: 30 (next prev first) Alternate versions: ASCII Summary