³ÉÈËÍ·Ìõ

Tolman is the great systematic learning theorist whose thought seems closest to how we think of learning today.  He uses cognitive terminology to describe behavior on the part of his rats that looks to us today as a description of intelligent behavior.  But, be careful.  His position is a bit more ambiguous.  He thinks of himself as a behaviorist who believes that the best way to describe the behavior he is observing is that it is as if the rat was behaving intelligently, demonstrating a purpose, or executing a plan.  He describes himself as a purposive behaviorist.  He is proposing an approach to explanation that would seem acceptable to his fellow behaviorists.  Yet, at all other points he proceeds to explain behavior in what looks to us as in cognitive, goal directed terms.  He speaks of his rats as forming cognitive maps, and as developing expectations and hypotheses.  My own view is that the behaviorism part of purposive behaviorism was simply Tolman stating that he was going to be basing his theories on the observations of behavior.  The terms of his theories, though, certainly look cognitive.

Tolman was also a very creative experimentalist.  Although he didn’t invent the rat maze-running experiment, he was among the most creative maze builders, and who tied the nature of the mazes to his theoretical questions. Sometimes those mazes could be elaborate such as his sunburst maze or as simple such as the + maze.  Incidentally, both of these mazes were ways of testing the response-learning vs place-learning hypothesis.  Other behaviorist learning theories claim that what is being learned are specific behaviors (response learning), but Tolman argues that when mastering a maze the subject learns the spatial layout of the maze (place learning).  Furthermore, he is well known for introducing the idea that place learning is accomplished by having the subject form a cognitive map of the maze.  This is a far cry from the theoretical terminology used by Thorndike, Guthrie, Hull, and Spence.

As for Tolman’s theories.  He was always willing to change his views, but he certainly believed that you could not explain behavior simply by correlating observable stimuli with observable responses.  You need to posit theoretical entities that operated on the incoming environmental information in order to produce the observed behavior.  Tolman called these theoretical entities, intervening variables.  Now, we have seen that others like Guthrie, Hull, and Spence have hypothesized that there are internal stimuli (such as drive stimuli) that play a role in explaining behavior.  Tolman feels no need to couch these intervening variables in stimulus response terms.

Tolman’s most systematic account of the intervening variables he felt were needed was “Determiners of Behavior at a Choice Point” from 1938.  ‘Choice point’ here refers to the point in a maze in which a rat can either left or right.  The chart from this article is reproduced in our text and identifies an intervening (theoretical) variable corresponding to each independent (environmental variable).  Corresponding to the independent variable maintenance schedule (how long the subject has been deprived of food) is the intervening variable of demand (demand on the part of the subject for food, for example).  Corresponding to the independent variable goal (appropriate to the particular deprivation state) is the intervening variable appetite (for food as opposed to water, for example).  In all, his intervening variables are demand (being hungry, e.g.), appetite (for a specific goal), differentiation (discriminating between various stimuli in the maze relevant to reaching one’s goal), motor skill (what behaviors are required in this particular maze), hypothesis (expectations formed from previous runs through the maze), and bias (does the nature of the maze encourage the rat to turn right at this choice point?).  The above intervening variables are best understood as events happening internal to the organism between environmental changes and observable behaviors.  These six intervening variables each correspond to an observable environmental variable and, in principle, can be studied by manipulating that environmental variable while holding all the other environmental variables constant. 

Perhaps, Tolman’s most well-known result is demonstrating the distinction between learning vs performance.  This was demonstrated in his famous latent learning experiments.  He would have one group of rats wander around the maze for a number of days without being reinforced for wherever they ended up.  Then after a number of days they started getting reinforced upon reaching the goal box.  Those rats showed nearly immediate mastery of the maze (as opposed to a gradual mastery found by a group of rats that received consistent rewards).  Why did this group show sudden learning of the maze as opposed to gradual learning (predicted by most learning theories)?  Because they had been learning all along--- even when they weren’t being reinforced.  There was no sudden jump in learning.  Only a sudden jump in performance.  This was a finding that other learning theorists had to take seriously, and caused some of them (such as Spence) to drop the assumption that reinforcement was necessary for learning.  Reinforcement influences performance, not learning.