Thursday, May 22, 2008

Thinking as a structured process

For some reason I was reminded of a classic and challenging article by Karl Lashley, "The Problem of Serial Order in Behavior" (1951). As I recall, the article was a pivotal contribution to new and productive thinking in what became "cognitive psychology." And it was one of the central components on Noam Chomsky's earliest attacks on radical behaviorism as a paradigm for psychology. Chomsky took the article to provide scientific evidence for the view that there is more to mental processes than a simple concatenation of stimulus-response mechanisms. Most basically, the article laid a basis for attributing substantial degrees of mental structure to ordinary human performances. (See this link for a nice reflection by George Houghton and Tom Hartley on Lashley's arguments on this subject.)

I'm sure there is more to the article than this, but what I remember most clearly is Lashley's study of "serial" behaviors such as typing, where there is some antecedent credibility in the idea that each discrete step is a probabilistic result of the previous step. This would assume that the performance is a Markov process -- one in which each step depends only on the prior one.

Lashley analyzes typing behavior and asks, what can we learn from analysis of errors? And simply, he argues that there are patterns of errors that demonstrate that the typist has a representation of the full series of actions as he/she proceeds through the performance. An example would be the incidence of transposition errors between characters in different parts of the word. If we find that there are numerous examples like this: "PRAZOC" instead of "PROZAC" --then this seems to imply that the typist has the whole sequence of finger movements in mind in the early stage of the performance. And this is inconsistent with the idea that the performance is a Markov chain of linked pairs.

Noam Chomsky soon linked that concept to the idea of a sentence being the manifestation of a syntactic structure as a real, underlying mental representation. (He presented this idea in a famous review of Skinner's book, Verbal Behavior; here is a snippet from Kenneth MacCorquodale describing Chomsky's argument.) Here the idea is that a language-user's comprehension and production of a sentence like "Fairbanks is the desolate capital of Montana" (a sentence, by the way, that has probably never been previously written or uttered) involves a cognitive representation of a number of abstract structures that are not serial or linear. Rather, the competent user arrives at an abstract representation of the whole sentence including its syntax, semantics, and phonology. And the point can be expanded: we might imagine that many kinds of complex performance -- dance, fencing, piano or cello performance -- involve higher-level planning, representations of abstract structures or grammars, and execution of the performance based on the abstract plans and representations.

The other alternative is that the full complex performance is generated by an initial setting and a set of rules deriving the next move based on the current move.

Now let's apply this perspective to thinking. The Markovian perspective would hold that an apparently complexly structured piece of reasoning or creating is actually a serial process in which one step entails its successor. And the more extreme forms of behaviorism evidently entailed that thinking is "nothing but" a series of learned responses to a current set of stimuli -- so the Markov-process interpretation seems to be entailed by the behaviorist premise.

There are pieces of human cognitive performance that probably do have something like this structure -- a meandering conversation or a stream-of-consciousness monologue, for example. But in general, we seem to be confronted with numerous instances of reasoning and creating that do not have this property. Instead, many instances seem to reflect an underlying strategic structure, in which elements that are introduced early in the production are crucial for later turns in the argument or creation. Why does Rawls introduce the idea of reflective equilibrium early in the book? Because he envisions a stage in the argument where this construct will be a vital part of the argument. Why does Monet sketch out the dark shadows around the windows of the Cathedral early in his painting work? Because he has a vision of the eventual gestalt of the finished painting. And, critically, we can speculate that these conceptions are cognitive realities in the mind of the philosopher or artist.

In each instance we seem to have a clear example of a thinker whose complex reasoning and creation is not solely serial. Instead, each is an example of an abstract representation of the whole guiding the particular steps of execution. And this suggests something very much like the Chomsky vision of syntactic representation of the whole guiding the building of the composition.

No comments: