Re: Skinner

From: Harnad, Stevan (harnad@cogsci.soton.ac.uk)
Date: Mon Nov 20 1995 - 18:58:41 GMT


> From: "Smith, Wendy" <WS93PY@psy.soton.ac.uk>
> Date: Tue, 14 Nov 1995 13:37:49 GMT
>
ws> I find behaviourism a very difficult topic. As I'm reading, I
ws> continually feel as if I'm missing the point. It seems to "explain"
ws> everything - and yet I don't feel any more enlightened at the end, so
ws> it didn't really explain anything to me! All the essential issues (or
ws> what appear to me to be essential - is this where I'm missing the
ws> point?) seem to be avoided eg "What is the current status of the
ws> theory?" (SH) "Radical behaviourism is anti-theory" (BFS) or "What is
ws> behaviour?" (SH) "There is no essence of behaviour" (BFS).

Could that sense you have that behavioural analysis claims to do it all
and yet seems to do very little perhaps arise from the following?

Behavioural analysis takes behavioural capacity for granted, and looks
only at how rewards and punishments shape it. But the real question is:
What gives us the CAPACITY to be shaped by rewards and punishments as we
are?

Example: If I put the problem of categorisation to Skinner (How do we
manage to categorise all the things we do, as we do?), his reply is that
it is shaped by the consequences of miscategorisation. True enough. But
why is it that people can be shaped by the consequences of
miscategorisation into the complex, world-sorting creatures we are,
with our dictionary-sized vocabulary of names for objects and events
and states of affairs, whereas other creatures cannot be?

A satisfying answer to this question, one that does not beg the
question, would tell us how one could give a system -- any system,
natural or artificial: this need not even be an explanation of OUR
behavioural capacity in the first instance; ANY explanation that will
actually deliver the behavioural goods will do -- how one could give a
system the capacity do what WE clearly have the capacity to do? It's no
more than asking how a plane can fly: There's no point replying that it
is shaped by the history of buttons the pilot pushes.

Ask yourself this question: If someone were trying to build a robot that
could do what we can do -- get about in the world, learn, categorise,
name, describe, respond to descriptions -- how much would it help him in
his project if Skinner told him that the robot's behaviour is shaped by
its consequences, by its history of rewards and punishments?

Note that the roboticist has not changed the subject; he is not talking
about mental states any more than the behaviourist is. He too is just
talking about behaviour. But he doesn't take the behaviour for GRANTED:
He must explain the capacity that generates it. That's exactly what
behaviourists never faced the problem of doing.

ws> Behaviourism appears to be putting more and more data and observations
ws> into a pot - but not combining them into a coherent whole. Skinner
ws> criticises the use of statistical control, rather than controlling via
ws> the experimental method. therefore lots of participants are used, when
ws> very interesting findings can be about individuals. But it strikes me
ws> that a science which can only offer explanations on an individual
ws> basis, and with a detailed history of the individual known, is not very
ws> useful. This doesn't seem to be an understanding of human behaviour,
ws> but a mere description of it.

The methodological issue of single-case generalisations vs. population
or repeated-test generalisations should be kept separate from the
question of whether a history of reinforcement is an explanation in
Psychology. Whether the history is an individual one or a statistical
one, it still doesn't explain behaviour in the sense I discussed above.
Imagine a physicist trying to explain collisions between objects in
terms of objects' "collision histories": We'd never arrive at Newton's
Laws of Motion that way. This would merely be "collisionism."

You are right, though, that this kind of historicism in Skinner is
related to his adherence to individual cases; it is all part, I think,
of a mistaken idea that scientific explanation amounts to a mere
description of observations and their regularities. Behaviourists, in an
over-reaction against the private and untestable theorising of
introspectionists, threw out the baby with the bathwater: Rejecting
theories about the unobservable things going on in the MIND, they also
rejected theorising about the unobservable things going on in the HEAD
(except inasmuch as they could be observed by brain science, but then it
was another subject, no longer psychology). Instead, they wanted only
to describe the regularities among observables. "Theories" were
unnecessary.

But theory is what makes science explanatory. We want to know the
underlying causes of observations: What makes the apple fall to the
ground, the earth rotate around the sun, the volcano erupt, the plane
fly, and -- the organism able to behave! We want to know the causal
basis of our behavioural capacities. Paradoxically, it is precisely the
science of behaviour that declined to analyse these, because they were
unobservable. Cognitive science is no more nor less than Behaviourism
finally granted the permission to theorise about the internal causes of
behaviour.

ws> Also, Skinner acknowledges that there is neural activity; and that
ws> there is overt behaviour (eg lever pressing). Presumably there is a
ws> link between the two (perhaps this is where cognition fits in?)
ws> However, this seems to be considered "unnecessary" or "not the province
ws> of behaviourism". Skinner says "They [learning theories] consist of
ws> references to 'mental' events, as in saying that an organism learns to
ws> behave in a certain way because it finds 'something pleasant' or
ws> because 'it expects something to happen'". However, this is not the
ws> cognitive psychology that we practice, is it?

That's right. Skinner is just conflating mentalistic explanation with
ANY form of internalist explanation. One can agree that MENTAL causes
are not explanatory, but that still leaves a lot that might be going on
inside the head to generate our behaviour.

Skinner's strategy of relegating other kinds of internal explanation to
"brain science," hence to another subject, was also wrong-headed: It all
depends what you mean by brain science. There are many things a brain
can do: it can generate action potentials, release neurochemicals, alter
neuronal connectivity, etc. That's brain science. But it can also run,
jump, see, hear, categorise, manipulate, talk, etc. That's behaviour.
Brain scientists typically don't get very deeply involved in the brain's
behavioural capacities (with the exception of neuropsychologists, to
whom I will return); they're more focused on the brain's "vegetative"
functions. So it is behavioural scientists who must take on the task of
explaining the brain's behavioural capacity.

(Neuropsychologists like Luria and his contemporary successors have
become increasingly cognitive, to the point where it is not always clear
when they are talking about the real brain and when about a hypothetical
model brain -- that is fine, in fact, modeling is one of the best ways
to do "reverse engineering," when you are trying to figure out how a
system built by nature works.)

Cognitive science includes behavioural and cognitive brain science.
What makes anything cognitive is that it explains in terms of internal
structures and processes. Some have tried to be more restrictive,
saying it's only certain KINDS of internal structures and processes
(e.g., computational or "representational" ones) that count as
cognitive. We'll be going into this more later in the seminar...



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:56 GMT