Similarity among responses is important, but it...

2010-04-14T21:54:25.638-05:00

Similarity among responses is important, but it's not clear cut, and it may be misleading to characterize it as necessarily physical. I'll try to explain:

The "dimensions" of a behavior sample are underdetermined, insofar as any given sample can be parametrically represented in many different ways (by many different sets of orthogonal features). That is not to say that the dimensions do not exist, only that it is in some sense arbitrary which ones are chosen to describe behavior.

Many sorts of dimensions can be chosen. Even something as straightforwardly physical as motion can be relative to different frames of reference.

The similarities among responses may also be more broadly structural, not just "physical." Responses can also be composed out of smaller pieces, as in sequences of actions (or, for that matter, words).

Not all features or feature-sets are equally useful for scientists or applied workers, though.

I believe the Skinnerian strategy - yielding perhaps the best return on analytic effort - would be to identify the responses by their orderly relations to controlling variables such as discriminative stimuli and rate of reinforcement.

This focus on the functional relationship is really what set Skinner apart from his predecessors in a "radical" way.

I think one quite often finds that the response classes defined
in such a way exhibit a coherence which is neither purely arbitrary (the only thing in common being the outcome) nor purely "physical" (the only thing in common being simpler features such as rate or speed or force). Nor are these coherences limited to distinctions which the animal cannot produce or discriminate.

Rather, many natural response classes cohere along lines which are structural and/or relational.

This is discouraging because any choice of dimensions may seem arbitrary, is not guaranteed to show the kind of beautiful order one can get out of (say) peck rates in pigeons, and (not unimportantly) may be hard to collect a lot of data on in a well-controlled and fully automated way.

However, EAB and behavioral science in general will not have a very strong grip on the response class until there are more workers exploring the richness of the response dimensions and digging out the underlying regularities in learning across different sets of response dimensions.

I see what you mean by the two have to be similar ...

2008-02-17T23:04:00.000-06:00

I see what you mean by the two have to be similar but not the same. In 129 the professor lectures from the left then lectures from the right he is still doing the same thing, lecturer but doing it from different angles which makes it different giving a different response.

Comments on Dead Man Test: What is a response class?

Similarity among responses is important, but it...

I see what you mean by the two have to be similar ...