The idea of distributed mental representations is old in cognitive science. I see you had a look at some of those PDP models from the 80's and 90's? In those, a piece of knowledge is not represented in any one "unit" in the model, it is represented in patterns of interconnections in the whole model. The retrieval of that knowledge is also not represented in a single unit, but by a sequence of events that engages many units and their interconnections. (of course, these are models of cognition, not of brain function, they're not supposed to be neurally plausible, but the distributed knowledge idea is as old as that).
Similarly, recent models of working memory propose that the current content of awareness are represented by resonating activation throughout a network of interconnected units representing relevant concepts and other elements. These elements may be localised within very different brain regions. In fact, both prefrontal cortex and some region of posterior cortex are minimally required.
Turning to the brain in earnest now, once you start thinking in network terms, it seems decreasing likely that any experience would be "localised" to any one region. We might have thought like this once, but it was a long time ago now, and it didn't get us very far. I don't see that one-to-one localisation is required at all to make something "part of the world of physical science".
Dear Woolie,
It is not so much that I had a look at a few papers. I was copied in to an email yesterday that said:
Below, you will find a Call for Participation for a Frontiers in Psychology, Section Cognition, research topic on "Representation in the Brain", hosted by Asim Roy, Leonid Perlovsky, Juyang Weng,
Jonathan C. W. Edwards, and Tarek Besold.
I actually edit books on this topic! The subtitle is likely to be something like 'the role of individual neurons in distributed processing'.
We are all agreed that knowledge is represented as a widely distributed pattern of shifts in connection weightings. Even Leibniz seems to have deduced that, having looked at cells through Leeuwenhoek's microscope and argued from first principles. Retrieval of knowledge will require signals to be sent through the network. But to make any use of an output from that retrieval process that specifies some image or concept or name or whatever it has to arrive somewhere. I know my six times table whether I am asleep, listening to Tristan or gardening. But if someone asks me what is four times six, the answer 24 has to arrive somewhere in the brain in order for me to 'think of 24' and for my speech apparatus to be fed the commands to report it.
In a standard connectionist model, as I understand it, the output will consist of a pattern of firings of some units amongst a bank of final layer integrator units, or possibly a temporal sequence from a single final unit, although this requires a routine that I do not think connectionist systems usually make use of. If integrator units (neurons) in the brain were serial in operation then the second model will do - and since it involves a single locus it fits with what I am suggesting. But neurons do not work like that and the slowness would not compete with predators using convergent-divergent systems and since the rest of the connectionist net is convergent-divergent anyway it makes no real sense even in the model. But in the first model the output pattern consists of a set of events that have no causal interrelation as they stand. To make use of them all as a pattern they all have to relate causally to at least one downstream event - they have to converge on some individual integrator unit. So we are back to what I am suggesting - that an entire output pattern, such as 24, has to be the input to at least one individual integrator unit.
And of course this requires just the same architecture as the connectionist rows beforehand. Having one unit receiving 24, or 12, or 18 does not provide a large enough repertoire of response so you want lots of units all receiving 24 through the divergent component of the network architecture - connectionist again. But we have to posit that, at least in this row, there are inputs to individual units that represent the entire concept or idea or image being experienced. Note that I am
in no way suggesting that the experience of 'thinking 24 at that moment' occurs in only
one place. That was Descartes's mistake, and also Leibniz's. Ironically, William James makes the same mistake having realised he did not need to. The experience will be massively multiple - as in the rows of a connectionist net. The solution to the paradox is that
experiencing is distributed but
an individual instance of experience is local. (This is basically the topic of the book Asim Roy and I are producing.)
The stuff about resonating circuits is, I think, complete guff. It arises from Wolf Singer and Walter Freeman's work and von der Malsburg's suggestion about binding by synchrony from way back. It has become fashionable, with Tononi, Edelman, Dehaene and whoever piling in. I know Walter and we argue about this, but it is incompatible with physics and in fact it is incompatible with the mechanism used by von der Malsburg to justify it. Synchrony can only have a causal effect locally. It is Imperial tailoring and no more. The whole wild goose chase dates back to the paper Horace Barlow wrote in Perception in 1972, which Horace now admits sent people off on the wrong tack completely.
So we are agreed that experience is unlikely to be localised to one region but physics does require
each instance of experience to be local if it is to be a real physical event with specific causal powers. The reason why previous thinking did not get far is that so few people have considered the multiplicity issue. Hyperaspistes did in 1641, the philosopher Elizabeth Anscombe did sometime around 1970, Semir Zeki and Horace Barlow have glanced off it and now it is back on the agenda. The problem is the intuitive resistance to the idea that there is more than one 'I' in one's brain. But if one can get over that hurdle then it becomes possible to turn the sort of models that Mark Edwards likes into something both coherent and consistent with physical science.
