Or, why Bourbon Barrel Ale leads to bad exam-studying and useless philosophizing.
So this model has arisen in two classes of mine, but I've felt uncomfortable considering memory in such terms. One of the professors really seemed to push it, however I wasn't about to interrupt his lecture with my concerns; furthermore, they weren't as well formulated at the time.
I think this model deserves an introduction first. The model is one used since the 1960s to describe memory and encoding of that memory, in a largely conceptual framework. It posits that memory is, initially, sensory datum (Sensory memory, SM). From there, if attention is directed toward that sensory datum, it is encoded into short term memory (STM), where it can last, according to some, between 15 seconds and a minute or so. Rehearsal keeps objects in STM, where it eventually falls into a long-term memory storehouse (LTM). According to the model, most memory from STM and LTM is semantic in form.
Maybe this is because I'm not immersed in memory research, but something about such a simple diagram throws me off. Primarily, I don't like modules that aren't neuroanatomically based. The division of labor seems more arbitrary that way, but that may be Biology-Major me speaking through.
Anyway, on to the more specific criticisms.
First, the 3-store model considers SM as primarily either iconic or echonic; visual or auditory. Certainly that is how most consciously direct memory works, if introspection can be given any credit at all. However, scent associations tend to be the strongest in evoking prior memories, and this model seems to not accommodate this result at all. I readily understand the unwillingness to allow for something that isn't auditory and visual; after all, it's hard to express linguistically something that doesn't fall into one of those two categories. However, are we to expect most memory to be of those categories rather than tactile, proprioceptive, or scent-or-taste based?
Most encoding to STM from sensory information, following the prior point, comes to the root of my problem. It is posited, by this model, that encoding is done on a semantic basis.
And this isn't the Bio-Major me speaking, this is the Philosophy-of-Mind me speaking when I say that I'm cautious that semantically-interpretable belief states are simply taken for granted at this point. I'm not a fan of Daniel Dennett's work on consciousness as a whole, but his critique as to the commonsense notion of a semantically-interpretable belief state is fairly devastating (as is Patricia Churchland and many others). What is the "belief state" of a frog flicking its tongue out? Is it that the black object moving is a fly? Or is it that the black object moving is likely to be food? Or is it just a reflex to a black object? Multiple ways to phrase a behavior semantically will result in a similar behavioral output, and in this, taking mental-semantics for granted is immediately suspect.
I really really want to sympathize with wanting to make semantics a basis for memory. Wittgenstein provided his critique with the Private Language arguments, and as such it seems unnatural for memory to have a basis in anything but a publically-expressible language. However, this can be escaped in an associationist framework if memory has a realization in a physically realizable network which does not see connection changes unless prompted by proper stimuli; in other words, the "meaning" (in the loose, not properly semantic sense) is set by physiological patterns rather than some arbitrary assigning of the individual using the memory (and, in this case, there is none, barring a homonculus regression).
So at this point I really hope that the literature I've found thus far on this model has been lacking. Otherwise, it will be held in suspect, in spite of the fact that it may appear on an exam tomorrow.