Saturday, October 30, 2004

The First (Sort Of) Request: Memory and Epistemology

In a comment, Clark Goble writes:

I've always wondered about the role of memory in epistemology. It seems that whenever we talk about rational reasons they are always dependent upon our memory. Even in a math equation, we assume we recall the early steps correctly. Memory is what allows us to transcend the temporal limits of reasoning. Yet, so far as I'm aware, it never gets brought up in epistemological articles. (Or at least I've missed most of them) The role of false memories seems to render a lot of epistemology quite problematic and at a minimum suggests that knowing we know doesn't follow naturally from knowing. While I may be completely wrong, I have this gut feeling that it ought to lead one to externalist accounts as well.

Now I'm no expert in epistemology, and I don't read a lot of the contemporary epistemology literature (I find it kind of boring, to be honest), but as I'm the foremost authority on epistemology among the authors of this blog, I will take it upon myself to post on it anyway. I do know of a few discussions of memory in the context of epistemology. One is online, here. It takes an internalist position, but there is a lengthy discussion of both externalist and internalist solutions to the fallibility of memory. Audi has a chapter on memory in his Epistemology: A Contemporary Introduction to the Theory of Knowledge, and I think a few other epistemology textbooks have chapters on memory as well.

Otherwise, memory has been a hot topic in philosophy for a long time. Aristotle discussed it, as did Hume (Section III here and Section IV and V here) and the other British empiricists. In the 20th century, Russell talked about memory often, including some great passages in The Analysis of Mind; the logical positivists talked about it fairly frequently (Ayer has a lengthy discussion of memory, and statements about the past in particular, in The Problem of Knowledge); and C. I. Lewis has an excellent chapter in An Analysis of Knowledge and Valuation in which he argues that the fallibility of memory demands a coherentist account of knowledge. Here is what writes about the implications of the potential for memory errors:

When the whole range of empirical beliefs is taken into account, all of them more or less dependent on memorial knowledge, we find that those which are most credible can be assured by their mutual support, or as we shall put it, by their congruence.

Of course, phenomenologists talk about memory a lot, as well. I'm sure Clark knows of Bergson's Matter and Memory, and (inspired by Bergson) Merleau-Ponty spends a lot of time talking about memory (as does Deleuze, at least when talking about Bergson). There was also a special issue of Philosophy and Phenomenological Research in the early 80s (82 or 83) that contained several excellent papers on memory. One of the papers in that issue is a brief review of the philosophy of memory over the last few millenia.

The philosopher who probably paid more attention to memory than any other in the last half century was Norman Malcolm. Much of his work on memory (e.g., in Memory and Mind ) involves a Wittgensteinian critique of representationalist, or trace theories of memory. In particular, he argues that the existence of "enduring representations" doesn't fit with our experience of direct access to memory, and creates the need for homunculi. He also talks about "false memories," though he doesn't call them that, and argues that memory is like knowledge, and therefore can't be mistaken. Any case that we might call "false memory" doesn't involve memory at all, or if it does, it involves a correct memory of a past impression which is what was actually mistaken. Here's an example Malcolm gives in "Memory and the Past" (which was in The Monist in 1963, vol 47):

Someone could point at a man in plain sight and say, "I met him last week." The even the refers to is meeting-that-man-last week. His memory is wrong, let us suppose, because he and the man pointed at were in different parts of the world last week. His erroneous memory does not pressupose some correct memory of the event referred to, for it did not take place. Still, his memory might be partly correct, for it might be that he remembered meeting this man but is wrong about when it happened. Or it could be that he had never met this particular man, but had met one who could easily be mistaken for him. Correct memory would here be mixed in with incorrect memory. Another possibility would be that previously he had dreamt of meeting this man, or had hallucinated it, or had formed in some other way an erroneous impression of having met him. But if his present belief was based on a previous false impression, then the present belief would not involve an error of memory: the error would be in the original impression.

I disagree with Malcolm's anti-trace arguments, in part because I think that the reconstructionist view of memory which is popular in cognitive psychology now does away with the problems he discusses. Under the reconstructionist view, which I share, memories do involve enduring representations, but remembering involves reconstructing those representations on-line, and this reconstruction is influenced by the structure of the representation as well as the context in which the remembering takes place. In this case, no homunculus is needed, as the present context is doing the work that the homunculus would have to do in older representationalist theories.

On the subject of false memories, Malcolm makes an interesting point, though what research shows is actually happening is that the memory mistakes the context of the original impression. The original impression is often not mistaken, however. For instance, during a therapy session involving suggestive visualization techniques, the visualizations may or may not be interpreted as products of the imagination or as memories of past events, but later memories of those visualizations may mistakenly involve the belief that they are memories of actual events, rather than a visualization exercise.


Moving on, I do think Clark is correct that the reconstructive nature of memory, as well as its fallibility, can pose problems for theories of knowledge which require that we have evidence to justify a belief. In most cases, that evidence will not be present at the time that we have a belief, except in the form of memories. I think epistemologists recognize this, though I'm not sure they're aware of much of the memory research of the last 30 years. For instance, research on the remember-know distinction (and similar distinctions, e.g., remember, know, and feel) has shown that people have qualitatively different types of memory experiences, and these types relate to accuracy -- when people feel they know something happened in the past, they tend to be correct more often than if they remember or feel it happened in the past. Also, research on the effect of background knowledge on memories for particular events has shown over and over again that background knowledge tends to intrude on memories of related events, causing us to falsely remember as being part of the event information in background knowledge that was not in fact present in the event. I suspect that these sorts of findings have implications for epistemological issues, but I'm not exactly sure what those implications are.

Here's another interesting finding from memory research that might have epistemological implications. In a classic experiment, Pichert and Anderson1 asked participants to read a story in which a house was described. The participants were told to read the story from one of two perspectives, either a potential home buyer or a burglar. After a delay, participants were asked to recall as much as they could about the story. During this first recall session, participants recalled significantly more information about the house that was relevant to their perspective (e.g., the potential home buyer might remember defects in the house, and burglars might remember information about the entrances and exits) than information that was relevant to the other perspective, but not theirs. After the first recall session, participants were told to think about the story again, but this time, from the other perspective (potential home buyers were now told to be burglars, and vice versa). Then, without reading the story again, they were told to recall as much as they could about the story again. During this second recall, participants were able to recall information about the house that was relevant to their new perspective, but which they had not recalled before. This result shows two things: 1.) The information that was irrelevant to their original perspective (schema) was encoded and 2.) This information was not accessible unless a relevant perspective (schema) was activated.

Now back to epistemology. Imagine a person has a belief, P, and has encoded the information, S, that would justify this belief, but because she has not activated the schema that would allow her to access S, she is not aware of this information. If asked to justify P, she could not access S, and therefore would not be able to do so. Does her belief constitute knowledge, even though she is (consciously) unaware of its justification residing in her memory? If it doesn't constitute knowledge, would it suddenly do so if she were to activate the relevant schema/perspective, and gain access to S?

There you have it, my response to my first request. If that wasn't rambling enough to dissuade anyone from ever asking me about epistemology again, I don't know what would be.

1 Pichert, J. W., & Anderson, R. C. (1977). Taking different perspectives on a story. Journal of Educational Psychology, 69, 309-315.