A Note on Memory

“Memories……like the corners of my distributed neural architecture…” -Barbara Streisand, Memories (Eliminativist Version).

Issues concerning the nature of memory, much like those concerning the nature of humour and of the paranormal, are underdescribed in the literature. Yes, philosophers have at one time or other recognised that memory carries with it a set of philosophical questions, and much interesting work has been done. But not as much as should have been and it is curious that it has not commanded more attention, nor achieved more centrality in areas such as the philosophy of mind. Continuity theories of the self, for example,  often follow Locke to make memory a necessary condition of Smith being the same person at aged 12 as at age 30 and at age 70. Locke noticed that in order to solve the problem of the persistence of persons over time it seems insufficient to posit the existence of some substance in which experiences inhere since those same experiences could be swapped between many such substances and our intuition would remain that these experiences were of the “same” person. Memory, he argued, provides a psychological criterion of persistence which is less injurious to this intuition. Needless to say there are problems with this, not least that the ascription of memory might presuppose that the question of person-persistence is already settled.  And whilst it might be the case that memory can secure our intuition that I am the same person as I was last week, is that the same claim as that I am identical to that person. Or does this not matter?

What I want to do here is to say a couple of things about the nature of memory and then to raise what might or might not be an interesting question for reductive physicalists i.e. those philsophers of mind who would want to argue that there are no features of memory which are anything more than complex features of brain function or brain chemistry.

“Memory” is not, of course, a unitary phenomenon and we can at the outset distinguish dispositional from (what I will call) the phenomenal character of memory.  By dispositional memory I mean the wealth of acquired practices we display from the time we get up until the time we go to bed: dressing, brushing our teeth, eating, driving, playing chess, riding a bike, telling the time etc. All of these involve learned cognitive competences which do not necessarily disclose themselves to consciousness. When I am playing my friend at chess I am demonstrating a memory of the rules of chess without necessarily rehearsing those rules to myself as I consider my next move (it is interesting that the more competent the player the less such rules impinge -grandmasters tend to see in positions patterns which do not smoothly transpose onto the rules that are constitutive of the game). What can we say about dispositional memory from the point of view of philosophy? Not necessarily that much. Philosophers like to distinguish the hard from the soft problem of consciousness in recognition of the fact that there are certain features of consciousness that might smoothly be integrated into the developing brain and behavioural sciences (the smooth features) and certain other features (the hard features) that will resist such integration, no matter how complete our understanding of brain function, architecture and chemistry turns out to be. Qualia (or sense data or contents presented experientially) might present a hard problem for the cognitive sciences; dispositional memory, on the other hand, might be their bread and butter – we are all of us to some degree a bit Sphexish.

I retain, as it happens, a memory of my first kiss – with a young lady called Suzanne, in a park in Oxford, on a damp Wednesday afternoon just before Christmas 1983. This memory seems not to be an instance of dispositional memory as it presents itself to me as a mental episode, with a certain character, and this character is in no way exhausted by whatever dispositional properties the memory might have. This character is phenomenal in that it feels a certain way (in this case wistful and pleasant and not a little sad). We can notice straight away that there is an interesting relation between phenomenal memory and the beliefs which accompany it, and which are used to describe it. Suppose the kiss took place not on a Wednesday but on a Friday. Would this make the memory a “false” one? Surely not. It seems as if the memory can remain veridical even if some, perhaps the majority, of the attendant beliefs are revealed by my diary to be inaccurate. Could it remain so if all the beliefs are false?  If it turns out that my first kiss was with Jane, in High Wycombe, in May 1984 we would surely want to say that my memory of Suzanne is a false one. Would realising this affect its phenomenal character? I suspect that it would. So we might want to say that there is a connection between an episode of phenomenal memory and our beliefs about that memory but that the connection is not one that can be easily articulated. To put it another way: phenomenal memory seems to be truth functional in a way that beliefs are not and yet the truth functionality of the latter somehow and to some degree feeds back into that of the former.

Reductive physicalists, of course, will want to assimilate phenomenal memory to the cognitive sciences. On this view my memory of Suzanne is no more than some cluster of features of my brain architecture and its content will turn out to be no more than a distributed property of certain neural activities occurring in parallel (or some such story). I wonder if the following feature of phenomenal memory might give them pause. Take two events, A and B. A occurred last year and was a major event in my life. B happened yesterday and was trivial. I remember both of these events and both these memories have what I have called a phenomenal character. As well as remembering A  (MemA) and remembering B (MemB)  I also have a mental episode C in which it is disclosed to me that A preceded B and that B succeeded A. In other words, there is something about the conjunction MemA and MemB which gives a phenomenal disclosure of the temporal ordering of the two episodes. C is a sort of second-order mental episode whose character is an internal relation given by MemA plus MemB. The question then is: (1) is my analysis of the phenomenology of temporal orderings of memory events plausible; and (2) if so does it present a particular problem for the reductive physicalist? My own answers to (1) and therefore to (2) tend to differ with the time of day, I’m happy to admit. So any thoughts welcome….

Suggested Reading

Adrian Cussins: The Connectionist Construction of Concepts (in The Philosophy of Artificial Intelligence (ed Boden 1990).

Mary Warnock: Memory (1987)

Wittgenstein: Philosophical Grammar (1974) (ed Rhees)

Suggested Listening

The Beautiful South: Song for Whoever.

Leave a comment ?


  1. Yes, your analysis of the phenomenology of temporal ordering of memory events is plausible. I could describe the way my mind works in the same terms.

    I’m not the person best suited to
    represent reductive physicalists, so I’ll let them answer your second question.

  2. Andy – If I was an A.I. entity I’d have a precise time stamp (T) on a MemA that happened in late 1983 (Ta) and a MemB in mid-84 (Tb).

    If MemA is your “Suzanne Mem” and MemC is the “trivial Mem” you mention separately that took place more recently I’d have to guess that over the next couple of decades your MemA will have continued to persist as some form of qualia while MemC would have probably faded to near nothingness most likely because of its dispositional or “routine” nature. In that time a “Jane Mem” (MemB) could pervade the “Suzanne Mem” and you may become a “bit fuzzy” about who was your first kiss – if and only if the two events were similar enough to allow that mixing – both Suzanne and Jane were blonds, say – a MemA-B if you will at Ta-b (“my first kiss,” you’ll say, “She was a blond girl I met around Christmas during my [whatever] year at Oxford… or was it around Easter? Just before a holiday time anyway, sigh… What was her name again?”)

    Meanwhile A.I. me with a similar Mem string would know MemA was at Ta, B at Tb and that C took place X:Y:Z hours:mins:secs ago at Tc. I’d probably have some form of ordering function to assign a value weight to each Mem but there’d be no mixing of discrete elements. Of course, as a fully functioning A.I. entity I’d imagine that I could be subject to a latency sub-routine to simulate the fuzziness of recollection from time well past and then I’d yield a similar dropped MemC and mixed MemA-B.

    Then you’d be walking across a park or a central-commons one particular day, a random event could take place – a gust of wind through the trees or against your cheek say: and the *full and true* MemA would flood back “as if it was yesterday.” A.I. me would have trouble “simulating” that, I think, seeing as I’d have to store a massive collection of possible random events, potential associations and outcomes in order to trigger those re-ordered thoughts – and that’s not to mention the various discriminators I’d need to have on-board to stop spurious linkages that would drive me towards reordering *all* my memories continuously.

    Seems to me what you describe as the “phenomenology of temporal orderings of memory events” via some form of an associative relationship of qualia is how memory works: it “makes sense” if you will. People do get mixed up in their memories and often they do – as if by magic; regain clarity: and I do think the R-P’s have a problem.

  3. Like folks above, I think that (1) seems perfectly plausible.

    I’m just not clear to me why it’s a threat to R-P. It’s only a threat if the reductionist has an unrealistic theory of mind, which imagines that the brain is like a digital computer (much like in Dana’s example). Like Dana said, that kind of theorist would be in deep trouble! But (I suppose) you can be a reductionist while still endorsing connectionist-sympathetic models of the mind. And these models could potentially engage in processing that resembles our flummoxed phenomenology.

    I might have missed the point, though. Could you say more?

  4. Thanks for replies so far guys. I’ll post some clarificatory stuff/responses soon as my child care duties are discharged….

  5. I cannot see how your first kiss with Jane entails that the memory of a so called first kiss with Suzanne is false. If it is false then you have been living in a fool’s paradise all this time which seems most unlikely. I would guess that you have undoubtedly kissed Suzanne at some time so good was it you called called it a first kiss. I understand the point you are making though, you have a fond memory of kissing Suzanne which remains veridical even though it is contradicted by dates. Outside of philosophical musings on this point we have to understand that memory plays us tricks it is not always reliable and we also have to consider the environment and frame of mind encompassing us when memories were laid down. Stressful conditions can make a difference so far as accuracy and inaccuracy are concerned .
    All this however does not seem to come close to the essential question here which if I Understand it correctly, and at this moment I am doubting, is that an event called C occurred such that it made a temporal distinction between two other events called A and B such that such that A is remembered as occurring before B. Examining this from a neurological viewpoint the situation is vastly more complicated and I am not sure if a thoroughgoing explanation would be forthcoming at this time. You are claiming and I cannot understand why, that there are no neural correlates for c and thus a complete reductive explanation for mental processes is not possible. I am far from sure I have understood this, which is nevertheless very interesting, and consequently is is only with reluctance and the fear of appearing stupid (should be used to that by now) that I have decided to comment. It seems to me that something must have happened such that temporality or before and after, is laid down and if it is not neural then where is it located and what is it?

  6. Lots of good stuff in all those replies.

    OK a couple of points.

    First off, a word about AI (Dana got me thinking here). It’s been a while since I dipped into this (yes I know it shows), but is it still the case that the connectionist paradigm is in the ascendancy? I’ve always argued (following Smolensky (1987?)) that the proper treatment of connectionism supports eliminative, rather than reductive, physicalism. In other words that connectionist AI success would describe highly distributed mental contents and that our current understanding of intentional and phenomenal states would not be smoothly reducible to these. Now eliminative physicalists do not have a problem, normally, with qualia based objections when those objections take the form of pointing out that some features of our mental life seem recalcitrant and resistant to the reductionist urge. They simply junk them. A completed neuroscience seems inconsistent with our common sense view of consciousness? No problem (they argue) just ditch the common sense view. We might want to call this the Gil Grissom view: go with the science, wherever it leads. So (and slightly pace what I take Benjamin’s view to be)the sort of mental contents I refer to in the OP might or might not have neural correlates and there is a strain of physicalism which remains unabashed either way.

    Now to characterise the mental state C a little more sharply: it is a mental episode, with a certain phenomenal character, which occurs at a certain time,t, and in addition discloses a certain temporal property which is independent of t (call this TP). Now you can have a pretty good neural story regarding C and you can possibly account for t within that story but can you account for TP?

    An obvious move here would be to suggest that I have done nothing more than redescribe the qualia/knowledge argument stuff. Sometimes that move seems to me to be a strong one. But I can’t shake the feeling that the property TP makes this a problem of a slightly different order. At the risk of sounding Kantian, I’m tempted to say that TP structures and makes possible coherent experience in a way that the more normal qualia examples might not.

    Catch me in a certain mood and I go along with the polite scepticism of Don and Benjamin; catch me in another and I can’t help but feel there’s more to it than that!

  7. From what I understand, the current consensus is what we might call the computational-representational theory of mind. The underlying gambit is that connectionism is not at odds with symbolicism, but complementary to it. (Granted, I am biased — I am at Waterloo, which is a CRUM school.)

    How mental states map onto brains states is a separate question, and I don’t know what the general opinion is. Eliminativism suffered a blow when Churchland abandoned the position — as theories go, I’m not sure why it’s worth the time of day. Functionalism hasn’t been defeated, but it’s not as immensely popular as it was.

    So at first, I had suspected that token supervenience (with or without anomalous monism) was the most convincing doctrine. But Paul Thagard at Waterloo endorses a kind of non-reductionist identity theory. (Full disclosure: I mention him here because a) he’s my supervisor, b) he’s enormously well cited, c) he has strong and clear opinions, d) he’s a cool guy.) Thagard rejects folk psychology, but insists that we have to preserve the idea of mental representations. In his view, our best explanations involve interacting systems at multiple levels, and reductionism just tends to be a poor explanatory strategy (when it comes to mental representations in general). Still, there’s no entrenched defence of qualia involved there, at least in the sense that he’s not wedded to the preservation of beliefs or desires in our vocabulary.

    The idea is that mental representations do actually have (distributed) neural correlates. As he puts it: a representation in the brain is a population of neurons whose firing patterns encode information by virtue of having acquired regular responses to particular kinds of input. And it’s not a just-so story: in addition to fMRI and EEG studies, we’re just starting to develop mathematical models that predict patterns of neural-like firing!

    So do we have an explanation of TP as a mental representation? It’ll depend on the details. One hiccup is the encoding of time, which is hard to do with representations alone — we might need help from something like a dynamic systems theory, which is non-representational. But folks around here (e.g. Chris Eliasmith) have been arguing is that such approaches will supplement representational theories instead of junking them.

    Does that help? How do you respond to that account, intuitively? (And please tell me if I’ve misunderstood.)

  8. Benjamin – I know this sounds perverse but I’ll have a think and then let you know how I respond intuitively :-) I’m familiar with some of that connectionist work but have (or used to have) my own reasons for rejecting the complementarity with classicism.

    Qualia-based objections to physicalism are not quite the same as the “problem” of intentional states such as “belief” abd I shouldn’t have conflated the two.

    And yes – your reply is very helpful.

  9. Of course. And that’s not at all perverse! Thinking then talking — that’s good philosophy.

  10. Ben:-
    I did not know Churchland had given up on Eliminativism when was that and which Churchland was it?
    You say our best explanations involve interacting systems at multiple levels, and reductionism just tends to be a poor explanatory strategy (when it comes to mental representations in general). So presumably the interaction systems at multiple levels have some sort of existence somewhere (and this does seem to be admitted with mention of the populations of neurons firing in patterns) there cannot be interactions in the absence of matter. Surely the problem we speak of here is taking place in our heads and that is where we need to look, ultimately. I am not saying that concepts like Connectionism and Functionalism are useless they may well assist in due course to establishing a Physicalist explanation. I am quite prepared to bow to greater knowledge when Profession Thagard says that Reductionism has poor explanatory strategy, but I cannot help my doubts in that connection just now. When I was introduced to Functionalism it seemed obvious that this was never going to explain at bottom what is actually going on. Connectionism the same. I have refreshed my mind on these concepts today, for me no easy job, and |I come back to the feeling that yet again I am standing before a motor car knowing nothing much about it trying to work out exactly how it does what it does. I might evolve some interesting theories in connection thereto but If I would only open the bonnet and start exploring some definite and ultimate explanation may well be forthcoming.
    It was for this reason rightly or wrongly, I gave up on computational-representational theories of mind opened the bonnet of the skull, and turned to exploring the Brain and Neuroscience, far more interesting too.
    Does this not all come down to the hard problem of consciousness wherein the irritation of nervous tissue brings about something as remarkable as conscious states. Huxley in 1866 likened this to being just as unaccountable as the appearance of Djin when Aladdin rubbed his lamp. So we may find Neural correlates for all sorts of human activity but how that would translate into what we call mental states experience, consciousness, seems at present without solution. Maybe Colin McGinn is right we are currently, just not up to it.

    None of this throws any light on the encoding of time but I am still not clear why this is a problem. We all know what before and after is, our brains must be doing it somehow, it must be done neurologically, how does it adversely affect a reductionist approach. if one is satisfied that a purely neurological explanation is thoroughgoing one. Where does explanation end?

  11. Don, it’s Patricia Churchland — I don’t know about Paul. But Patricia now talks about how our terms “co-evolve” as we gain better insight into how causal mechanisms relate to the appearances. Co-evolution is a very different explanatory strategy than elimination.

    Yep, Paul Thagard’s view is that the interacting systems are thoroughly physicalistic. Mental representations are part of higher-order physical systems. He says that these make for good explanations. And, as you mentioned, functionalism doesn’t care very much about the bottom-level specifics; so Thagard says, so much the worse for functionalism.

    On the other hand, connectionism has undergone some development. Eliasmith and others have found ways of using vectors to mimic the patterns of spiking neurons. I think it’s fair to say that they’re only in the beginning stages of understanding how the mind works at that level, but to an outsider like myself it sounds pretty promising. So my impression is that, so long as connectionism is regarded as complementary to computationism instead of being a competitor, and so long as the cogsci community is keeping a careful eye on whether their theories track any developments in neuroscience, then the explanations are only bound to get richer.

    I’ve never asked Paul (T) what he thinks about the hard problem, but I can predict with some certainty that his reaction would be “there is no hard problem”. Or, as Simon Blackburn put it in the newest TPM: “The only hard problem of consciousness is convincing people that there is no hard problem of consciousness.”

    But that’s not the same as saying that all explanations end at neuroscience, since Paul (T) wants to keep serious talk about multiple levels. To say that every explanation ends at neuroscience is just the same as saying you’re a reductionist. And I see little point in debates about reductionism. Convincing people out of reductionism is like deconverting them from a religion. They’ll let go of it only when they’re ready.

    Like you, I’m unclear about what the problems are when it comes to encoding time, but as I said earlier it’s not my field so I wouldn’t be surprised if I’m missing a nuance.

  12. I don’t think the problem is about the encoding of time. Rather it has to do with accounting for the phenomenal character of certain states which disclose a temporal character. The two things are different.

  13. Andy – I like Ben’s turn of phrase around “flummoxed phenomenology” (@ 12/15/8:57) and on re-reading it a memory I’ll call “reading about Strange Loops” (Mem-SL) has just occurred to me. I remember seeing that title on the clearance table of my favourite discount bookstore and being interested, I bought it (Mem-bSL). I know that I did that well in the past (two years ago, or was it one or three?) I simply can’t recall. (Next I’m home I’ll peek at the title page where I normally jot down my name and year I got it – just in case it walks away from me… They do that, you know.)

    You see – I go to this place almost every payday and it’s the kind of thing I do, if not routinely – often enough that it just doesn’t stand out, so perhaps Mem-bSL has become what I’ll call “flummoxed routinized” temporal event. Fact is *I know* that Mem-bSL proceeds Mem-SL: logic dictates that – but it also feels right, too: quite literally, picking it up and looking at it in the store, some of the pictures; feeling of it’s size and it’s weight while I was reading it propped up in bed; that sort of thing. Further, I distinctly remember moving the book from one shelf to another just last week (Mem-mSL) when I was reorganizing some of my books (by weight in accordance with bookshelf bendiness – they do that, you know!) but to tell you the exact date of Mem-bSL, or even the season in which I bought it right at this point in time is; quite simply: impossible. If pushed, I’d have to guess it was a couple of years ago, again: it kind of “feels right” – but I could be wrong.

    I’m think that’s somewhat along the lines you’re describing @ 12/20/6:15. I see what your saying, I think. To me it’s almost a psuedo-ordered, semi-sequential, but strangely contiguous time and while it “feels” ordered, it is not, as you point out to us: “encoded time”…

    Now here’s something interesting: on reading “flummoxed phenomenology” @ Ben above, would Mem-SL have been trigger had I not had a Mem-mSL? Would the connection of Mem-mSL and your distinction of “not encoded time” stood out for me without having Mem-SL?

    I peeked at Wiki and I don’t know where Douglas Hofstadter’s “Strange Loop” concept would fit in the discussion other to say it felt to me that he described some of same paradoxical phenomena that you point us to in your opening post.

  14. I think these interesting Strange Loop examples harken back to the point that Andy made earlier — about how our episodic memories can be belief-dependent. As Dana put it, memories don’t necessarily encode information about its time sequence. In her episode of Deja-Vu above, Dana infers that Mem-(bSL) occurred prior to (SL), because she can triangulate that information by way of an intermediate event (mSL).

    The lesson is: our mental representations operate in a holistic environment. They’re all part of the same network. And that’s exactly what the new-fangled connectionist would expect. So: physicalists rejoice!

    But suppose that Dana wasn’t able to infer when bSL happened. Then we have to ask (as Andy asked in the OP), in what sense is her experience of SL when she was ignorant of mSL identical to her experience when she had the benefit of mSL? In other words: does her loss of the ability to remember the time of the event mean that she’s actually remembering something else?

    (At this point, I think this is going to turn into confusing analytic writing unless we start using English phrases. So let’s call the flummoxed recollection of bSL “the weird memory”, and the recollection of bSL with the help of mSL “the cleared memory”.)

    The answer to that question is messy.

    There’s a relevant sense in which the weird memory is not the same as the cleared memory. Our experience of having a flummoxed phenomenology seems to exert a kind of warranted confusion. So at the very least, I think we’re entitled to say that the mental representations are different. The weird memory does not represent the same thing as the cleared memory. They are qualitatively distinct mental events, and ideally we will one day be able to predict distinct patterns of neural firing associated with each.

    But there’s also a relevant sense in which it is the same memory, because it’s about the same event. Whether or not we recall all the details is irrelevant, in that sense — the point is just that if the representations of the event had been faithful to what caused them, then there’d be no difference between the weird memory and the cleared memory. (Or — to be more precise — the weird memory would never have arisen in the first place.) And this is just the problem of intentionality, stated in a new form (which is what caused my initial puzzlement, and was why I thought I was missing something).

    Is intentionality a challenge to physicalism or to reductionist physicalism? Multi-level physicalists like Thagard say “no”. For Paul, the answer to the problem of intentionality is less about mental-states on their own, and more about the whole system that mental-states interact in. But then, it would seem that intentionality isn’t just a question of mental-states anymore — it’s about the states of the wider system. It’s not clear to me, though, how the reductionist gets to say that the two events are “the same”. And maybe that’s what Andy’s intuition was (which is only now dawning on me; derp).

  15. Buy north face
    north face coat After receiving my new boots the next day after I ordered them, I wore

    them on a 6-hour, strenuous hike with a 2,000 ft. elevation gain. These boots are totally fantastic! They performed impeccably on

    dirt, shale, glacier fields, boulders, streams, mud — you name it, plus waterproof to the max! I wore med. wt. hiking socks and used

    my orthotic inserts. No blisters, no rubbing, great protection against toe stubbing. Feet felt great the next day! I love these boots

    and would highly recommend them for hikers that love adventure and the high country buy

    north face !hjy

Leave a Comment

NOTE - You can use these HTML tags and attributes:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>