Re: Skinner

From: Harnad, Stevan (
Date: Mon Nov 20 1995 - 20:06:48 GMT

> From: "Harrison, Richard" <>
> Date: Tue, 14 Nov 1995 16:36:57 GMT
rh> I don't think having a theoretical framework to fit experimental results
rh> is necessarily the evil Skinner thinks it is. For example, changing
rh> from 'rate of response' to 'probability of response' involves a change
rh> from fact to theory (or at least interpretation) and this seems pretty
rh> innocuous.

Rate vs. probability here is a pretty trivial issue (and depends more on
views about foundations of probability theory, far from Skinner's domain
of expertise). The real issue is the status of theory in science in
general. Psychology is no different from other areas of scientific
inquiry: We want to understand the observables. Understanding comes from
a causal theory. Constructing such a theory usually calls for unobservables:
hypotheses about things whose existence or causal role is not yet
established. So we use them to guide our further investigations. We test
these hypotheses, check whether there is evidence that they are correct,
and in the process, we get a guided tour of the world. Otherwise it's
just an empty descriptive catalogue, leading nowhere, explaining

In the case of behaviour, what needs to be explained is not the trivial
historical path of reinforcement that shaped your behaviour, but what it
is that gave you the capacity to successfully tread that path: It's the
causal basis of behavioural capacity (unobservable, hence requiring
hypotheses to ascertain) that if Psychology's burden, and behavioural
analysis never shouldered it.

rh> It seems his point is that theory is wasting our time when we could be
rh> amassing data for a future time when we'll have enough data to
rh> understand what its all about. However, doesn't theory help us form
rh> hypothesis that we can go and test in the lab? Also, what will 'enough'
rh> data look like so we know when we can tie it all up in a theory of
rh> behaviour?

Yes, behaviourists always believed (though one wonders where they got
the idea, since it is not true in any scientific domain) that data are
self-explanatory: Just make and report your observations, and the facts
will speak for themselves. Tell that to Newton and the apple...

Yes, hypotheses guide research, suggesting where to look next, and what
to look for. But more important, they are what we really want to know:
For a system of correct, confirmed hypotheses will be the causal
explanation of the world.

rh> Skinner's analogy between behavioural science being to neuroscience
rh> what early genetic science was to the later study of genes is certainly
rh> appealing. However, if we maintain that there is a mind-body problem
rh> and the phenomenon of experience is not necessarily going to disappear
rh> with advancing neuroscience (even at our cerebroscope extreme) then
rh> there seems to be a fundamental flaw in the analogy. That is, there is
rh> no equivalent of experience in the study of genetics and genes. As
rh> scientists we aspire to explain everything (well eventually...) so the
rh> study of human behaviour should involve the study of cognition as well
rh> as behaviour from an external analysis and the study of neuroscience.

It's not just that experience makes psychology fundamentally different
from, say, genetics. It's that neuroscience is either investigating the
wrong thing (if it's investigating vegetative aspects of brain function)
or it is in the same position as behaviourism (if it's investigating the
brain's behavioural capacity). Neural data don't explain themselves; nor
do they explain behavioural data. And behavioural data don't explain
themselves either.

What is needed is a causal theory of how the brain (or any device)
generates behavioral capacity of the kind we observe in ourselves and
other species (or even in simpler systems that can do SOME of the
things organisms can do). What is needed, in other words, is the
underlying MECHANISM of behaviour. Skinner thinks this will amount to
something like the relation between molecular genetics and Mendelian
(behaviourist?) genetics. The truth is that both brain science and
behaviouristic psychology are waiting for a THEORY of that mechanism,
so the one will know what kind of physical implementation to look for,
and the other will know what generates the behaviour whose reinforcement
history it so closely catalogues.

rh> Also its a little disappointing that Skinner does not accept that
rh> behaviourism can be part of a wider psychology. And whilst he does have
rh> some valid criticisms about cognitive scientists it would appear that
rh> behaviourism, neuroscience and cognitive science could be integral
rh> parts of a comprehensive science of human behaviour.

I would be inclined to say that the behavioural data are already in; no
point cataloguing more reinforcement contingencies. The hand-writing is
on the wall. Now get to the real work of explaining its causal basis.
This causal basis is certainly subserved by the brain, but that doesn't
mean you will discover it by peeking and poking at the brain. Theory
(in the form of models that generate behaviour) must lead the way, as
everywhere else in science, but especially in reverse engineering, when
the data are already in.

In this quest, theoretical modeling is likely to play a more critical
role than brain or behavioural data-gathering.

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