Gallistel (1990) speaks about a demonstration used to teach students about the animals’ ability to learn rates of return.  During the demonstration, a rat was trained to run through a T-maze where the left arm was randomly baited 75% of the time, and the right arm baited 25% of the time.  If the rat did not choose the baited arm, the trial ended without it receiving any reinforcement.  A light bulb (not visible to the rat) indicated to the students which arm was baited.  The students were asked to guess which arm the rat would choose.  Since the rat is not reinforced when it chooses correctly:

“The rat learns to choose the higher rate of payoff on most of the trials, thereby nearly maximizing its rate of success.  Since there is no way to know in advance which feeder is armed, the strategy that maximizes success is always to choose the more frequently armed side.” (p. 352)

Thus, the rat consistently chooses the left arm of the maze in order to receive the maximum 75% of reinforcers.  The students, on the other hand:

“almost never chose the high payoff side exclusively.  In fact, as a group their percentage choice of that side was invariably within one or two points of 75 percent.  They matched the relative frequency of their choices to the relative frequencies with which the sides were armed.  They were greatly surprised to be shown when the demonstration was over that the rat’s behaviour was more intelligent than their own.” (p. 352)

Of course, if the rat were able to correct its behaviour and choose the baited arm after making an incorrect choice, it would learn to match its responses to the relative frequencies of reinforcement (as according the Herrnstein’s matching law).

            Gallistel raises an interesting question with this demonstration.  Specifically, with all our irrational thoughts, heuristics and biases, are we dumber than rats (at least in terms of probability learning)?  The answer seems to be 'no'.  However, when given equal opportunity (correction procedure) the rats are able match probabilities as we do.

            Animals are in fact very accurate in their abilities to learn and match probabilities.  This is not surprising since learning rates of return is important in all foraging situations.  Perhaps, however, this should serve as a reminder of humility to humans.  After all, "fifteen thousand years [of evolution], for what? We're still nothing but animals" (Cohen, E.).

Questions/Concerns/Links not working?  pls. contact egray@ualberta.ca