Thorndike [Powers]

[From Bruce Abbott (951102.1325 EST)]

Bill Powers (951101.1450 MST) --

    Bruce Abbott (951101.1535 EST)

What framework would Thorndyke have
fallen back upon if someone had claimed that consequences do not have
any particular effects on the actions of the organism that produces

He would have pointed to the data, which appeared to show that such claim
was false. Stripped of its "connectionist" language, all the law of effect
says is that certain actions that are reliably and immediately followed by a
"satisfying state of affairs" tend to be repeated. That is what his
observations seemed to show. Much later (early 1940s, I believe) Tolman
demonstrated that rats would use varied means to reach the same ends. I
believe that Thorndike was still around then; I wonder if he had any
comments on this demonstration?

    So not only didn't Thorndike attempt to infer what the cat was
    doing in the puzzle-box, he deliberately avoided doing so, because
    he wanted to avoid the trap of ascribing his own intellectual
    abilities to the cat.

Yes, this was the style of the time and in some quarters still is the
style. But the antidote was no better than the disease: ascribing to the
"situation" the causal properties which Thorndyke found subjectively

Avoiding this "trap" is a good idea only to the extent that is is really
a trap and not a route to valid insights. The problem is that in
avoiding attributing _all_ his own capabilities to the cat, Thorndyke
was asserting that the cat had _none_ of his own capabilities. You know,
either horses can count or they are just bundles of reflexes. By ruling
out intentional behavior, for example, Thorndyke was left with nothing
but external causation as an explanation, which to my mind is far worse
than giving too much credit to the cat's brain.

Thorndike was taking the very conservative view that complex processes
should not be assumed if a simple process will serve. That is, he believed
in parsimony. If he had felt that explaining the cat's behavior required a
more complex process than simple association, I'm sure he would have
attempted to determine what that process might be. However, his data
satisfied him that simple association was sufficient.

But that would have been all right if Thorndyke had laid out his own
framework in detail and shown how his interpretation of the cat's
behavior followed from that framework. Presumably, he would have had
experimental evidence to help persuade us that he was using the right
framework, so we would be willing to go along with its application to
these experiments.

What is easy to overlook is just how novel Thorndike's approach was at the
time. Prior to Thorndike, nobody -- nobody -- had thought to experimentally
investigate learning in animals. And when Ivan Pavlov, the Russian
physiologist, began his own investigations quite independently of
Thorndike's, the theory Pavlov formulated to explain his observations was
essentially the same: the development of new reflexes through a process of
association. These were not stupid people who lacked an understanding of
scientific method: Pavlov was awarded the Nobel prize in 1904 for his
experimental investigations of digestion. Their mistake was to choose the
most strongly supported ideas of their time as the basis for their own

    By saying this you make Thorndike sound silly, but it is not what
    Thorndike said or implied. It is the cat, not the behavior, that
    is confronted with a particular situation, and different situations
    tend to be associated with a particular mix of behaviors that
    appear and reappear with a given frequency, as the observer sees

Then why did you say that the behaviors were confronted with the
situation, and so on (as you did)?

Perhaps you could point out the passage. I don't recall saying anything of
the kind. Behaviors are just output. Here on campus, students sometimes
cut across the grass, thus taking a more direct route from Point A to Point
B. Most students will stay on the sidewalk, probably because they don't
want to kill the grass. However, after enough students have used the same
route across the grass, the grass begins to die there and a path appears.
The presence of a visible path encourages more students to take it (after
all, the damage is done), and this positive feedback loop keeps working,
resulting in the development of a wide, ugly, grassless path across the
lawn. The behavior (cutting across the lawn) changed conditions (loss of
grass along a path) thereby increasing the liklihood of the behavior (more
students cutting across the grass). This is like the process Thorndike
envisioned was responsible for the observed changes in the behavior of his
cats. With each escape from the box, the associative pathways connecting
the situation to a particular pattern of behavior were becoming more deeply
rutted. Like the students confronted with the problem of getting to the
entrance of the building visible across the grass, the cats were becoming
more and more likely to take the well-worn path (pushing against the pole)
to their destination (escape from the box).