Questions on the Nature of memory, personality, etc.
robertmaertin at gmx.de
Wed May 19 11:03:11 EST 2004
Thanks for answers Glen & Lester
Okay. One can describe experimental psychology in many fashions. One
may for example use behaviourist terms.
But in contrast to 1950s psychology we can make assumptions about
underlying computations. And I'm not talking about abstract symbol
manipulating theories from cognitive psychology. I'm talking about
underlying data-processing structures (nets of neurons of varying
properties). One could for example ask :
"How is it achieved that a set of neurons in IT with their receptive
fields at the site of attention (located via behavioural tests and
eyetracking) is firing more vigorously than neighbouring neurons even
though neurons in V4 (the direct predecessor of this IT region) show
no difference in mean firing rate ?"
Pretty straight forward...
What do we do now ? Besides the fact that we are in need of new
terminology we are right inside the Black Box. We can now make
assumptions classical behaviourism does not offer the tools for. (Does
A very nice example combining both behaviourist experiments and "tha
new style" was the modeling of the cerebellar function on in timing of
conditioned stimuli. (Done in Rabbits, I think) Pavlovian learning was
done (Tone paired with airpuff on the rabbit's eye). The Interstimulus
interval was varied. The strength of the putative underlying learning
mechanism was calculated as a function of the ISI by measuring the
Now it was assumed that the cerebellum is a crucial structure for the
learning of the relative timing between tone and airpuff. A
computermodel of the cerebellum was designed including everything one
knew about cerebellar structure. Input to the cerebellum in during the
learning task was measured during extensive electrophysiological
sessions. This input was fed into the comutational model. The
artificial cerebellum produced output that would (regarding the
projections from cerebellar nuclei) result in an exact timing of the
The real and artificial cerebellum were compared using
electrophysiology and lesion studies (artificial destruction of parts
of the model vs. impairments after lesion).
And the funny thing was : months later the authors critizised their
own model and proposed even better implementations (even though their
results were pretty cool before).
So. Do behavioural terms suffice for such kinds of research ? (not
I know that I am not argueing on an abstract, conceptual level here...
but I hope these examples give you an idea what I am worrying about.
I think that we are past a point were we can take a term like "spatial
attention", get its definition and look for (or propose) underlying
mechanisms. Once such a theory has been created, it may be falsified,
extended, cursed or implemented.
"Cognition" and "Speech processing" are such HUGE terms. But if you
stick with smaller bits... we might just find something :)
BTW: I thought about O'Regan and Noe. If you assume that you don't
need any internal representations or symbol manipulating mechanisms
and that quality and function of a modality are determined by the
action of the agent in its environment... this appears to be not
"cognitive" at all. Right. Sounds a bit like functionaliusm without
ascription of beliefs
But what keeps me from saying: "Behaviour is all we need" is that I
think, given our means of neuroscientific methods (as well as
computational models) it is our duty to at least try to explain
Don't you agree ?
(If I'm misusing behavioural terms... notify me.)
thx for reading & regards
More information about the Neur-sci