> From: "Willoughby Sarah" <SEW295@psy.soton.ac.uk>
> Date: Mon, 20 May 1996 11:32:06 GMT
>
> 73) What was wrong with Behaviourism?
>
> Firstly, there is no clear boundary for what is behaviour and what is
> merely the body functioning the way it should. It is generalised
> into, 'what an organism does'.
If one counts the OBSERVABLES available to a behaviourist (or to a
cognitivist, for that matter): one can observe what the whole body does
deliberately, or what it does involuntarily (as in reflexes), or what
parts of the body, inside and outside, "do" (e.g., twitches, blushing,
brain waves, neuron action potentials, neurochemical secretion). These
are all observable, but where does behaviour end and something else
(what?) begin?
> The main problem with Behaviourism is that it does not attempt to
> explain behaviour by formulating testable theories. Therefore it can
> not explain how the mind works because it refuses to look at internal
> processes.
Behaviourists do formulate testable theories, but not theories about
what is going on (unobservably) INSIDE the organism to generate its
behaviour: only theories about what the organism will do under this or
that external (observable) condition, for example, after a certain
history of rewards or punishments. No other branch of science denies
itself the right to hypothesise unobservables to explain and predict
observables.
> The Behaviourist idea is just to observe outward behaviours and
> predict from them alone. Their only explanation is that learning is
> a product of rewards and punishments but this still does not explain
> the internal processes that cause our reactions to such stimuli.
It not only fails to explain the internal processes that cause our
REACTIONS to such stimuli; it also fails to explain the internal
(structures and) processes that give us the capacity to LEARN from
rewards and punishments as we do. Behaviourists did not take on the
task of "reverse engineering" our behavioural capacities by explaining
their internal mechanisms (because those internal mechanisms are
unobservable -- or, alternatively, because they are the responsibility
of another field: brain science, not behavioral science, which is
concerned only with the analysis of behaviour). Instead, they
concentrated on "behavioural engineering," which was just the control of
behaviour.
This self-limitation arose in part from an over-reaction originating in
something that behaviourism was actually RIGHT about, namely, that ONE
kind of unobservable, the mental (i.e., what goes on in the mind, as
revealed by introspection), cannot be used either as data or as theory.
But in excluding the unobservable contents of the MIND, behaviourists
also excluded all other internal unobservables (including the real and
hypothetical structures and processes in the HEAD that generate our
behaviour).
> Another problem for behaviourists is trying to account for the
> differences in capacity such as intelligence and the ability to learn
> language. These sorts of questions can only be answered by a theory
> of what is going on inside which has, in part, been explained by the
> capabilities of artificial intelligence.
It's not just DIFFERENCES in capacity that behaviourists did not account
for, it's the capacities themselves. One important capacity among these
is our grammatical capacity: Our capacity to recognise and generate all
and only the grammatically correct sentences in a natural language. It
turns out that for this capacity you need a complicated internal
structure called Universal Grammar (UG), and it also turns out that UG
cannot be learnt by trial and error and reinforcement, because
hardly any trial and error goes on before children are already showing
that they have UG.
One can use the word (human) "intelligence" to refer to some or all
(human) behavioural capacities (e.g., recognising/generating
grammatically correct sentences, doing logical reasoning, recognising
patterns, remembering faces). Behavioural analysis never gave an
explanation of how we -- or anyone, or anything -- could have such
capacities. Artificial Intelligence (AI) was the first to actually
show how something -- anything -- could do the kinds of things we can
do. AI showed that computers can do it, through computation. That didn't
mean we do it that way; but until other theories of how we do it started
to appear, computation was the only theory there was; behaviourism had
none.
> Behaviourists fall down over the issue of language and perception
> which they claim is again, based on reward and punishment.
Perception depends in part on inborn sensory mechanisms (like the frog's
"bug-detector"). Behaviourism cannot explain these mechanisms.
Perceptual learning depends on mechanisms for learning patterns
(perhaps through neural nets); behaviourism had no explanation. Reward
and punishment may teach you the pattern, but what kind of internal
mechanism do you have to have to be able to LEARN the pattern from
those rewards and punishments? The same is true of language learning.
> This can
> not explain how humans learn universal grammar nor how children learn
> grammar when many adults speak ungrammatically.
This is a bit garbled. Adults speak ungrammatically in the sense of
saying "Who did he gave that to?" instead of "To whom did he give that?",
but those are just violations of local grammar; in fact, in the next
generation, they could be considered grammatically correct. They are not
violations of UG. A violation of UG would be to say "Who did he think
that left?" instead of "Who did he think left?" No one says that, child
or adult, and there are an infinite number of such things that no one
ever says, child or adult. Hence we did not learn not to say them; we
were born "knowing" not to say them. We were born with UG.
> Also, they have no
> explanation for why we are the only species capable of language .
Correct. In fact, they have no explanation for any species differences
in any behavioral capacities, because they have no explanation for any
behavioural capacities.
> Therefore, Behaviourism appears to be largely faulted by its' own
> restrictive warriors.
Yes, the limits of behaviourism were largely self-imposed; but it has
to be admitted that until AI came along, no one had any idea about the
internal mechanisms underlying humans' and animals' behavioral
capacities or anything like them. So it was not just the exclusion of
theory about internal mechanisms that limited behaviourism, it was the
lack of any good idea about what such mechanisms might be (and the
failure to realise that that "reverse engineering" needed to be done at
all).
This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:41 GMT