Models of memory: Wittgenstein and cognitive science

The model of memory as a store, from which records can be retrieved, is taken for granted by many contemporary researchers. On this view, memories are stored by memory traces, which represent the original event and provide a causal link between that episode and one's ability to remember it. I argue that this seemingly plausible model leads to an unacceptable conception of the relationship between mind and brain, and that a nonrepresentational, connectionist, model offers a promising alternative . I also offer a new reading of Wittgenstein's paradoxical remarks about thought and brain processes : as a critique of the cognitivist thesis that information stored in the brain has a linguistic structure and a particular location. On this reading, Wittgenstein's criticism foreshadows some of the most promising contemporary work on connectionist models of neural functioning. Since the time of the ancient Greeks, our memory has been compared to writing on a wax tablet or a storehouse which contains many written records ; more recently, analogies have been drawn with information storage in a computer. But while a conventional digital computer will always retrieve an item from its memory in the same form, our recall of an event such as a family gathering long ago is often a much more fluid affair: a great deal depends on the context in which one tries to remember, and how much prompting goes on . Recalling a distant event is often a reconstructive activity, one in which there is no sharp dividing line between factual recall and piecing together the most plausible and relevant story (Neisser, 1982 ; Neisser & Winograd, 1988, chs 3-5) . Our memory for faces is another example that does not sit well with the storehouse model: most people are much better at recognizing an acquaintance than remembering his or her appearance, let alone describing that person on the basis of free recall . Advocates of the storehouse analogy treat these cases as anomalies which can be accomodated by postulating regulations which control access to the storehouse. But these are, I believe, telling instances of the limitations of that analogy . For memory is, after all, an ability to think and act in certain ways ; we do not have to assume that these actions must be explained in terms of memory traces, or that the words we utter are a "translation of something that was there before" (Wittgenstein, 1980a, p. 736) . [1] Storehouse theorists take these traces for granted, for they think