reinforcement ( Shimp, 1982a). The procedure of the present experiment is much simpler than the symbolic matching-to-sample procedure and seems as convenient and easy to arrange as the autoshaping procedure. Thus, we now have three ways by means of which we can ask a nonverbal organism to provide self reports of its own temporally patterned behavior.
An important next stop would seem to be to ask what kind of integrative theory can assimilate the present dissociation. That is, why, as preference becomes more extreme, does the animal appear to know less and less about what it is doing. This result seems counter-intuitive. It would seem natural for an animal, as two response alternatives become more difficult for it to discriminate, to care less about the difference between them: for example, Baum ( 1974) has suggested that undermatching in concurrent performance is attributable to a failure to discriminate perfectly between response alternatives. Here, however, the opposite result was obtained: as the subject knew less and less about the two response alternatives, it cared more and more about the difference between them.
The present dissociation, and related dissociations between behavior and metabehavior in pigeons ( Shimp, 1983) may be related to various other dissociations, such as those obtained with split brain preparations and amnesiacs. Similarly it might be related to dissociations between perceptual-motor skills in general and verbal self reports of those skills ( Nisbett & Wilson, 1977; Jacoby & Dallas, 1981) and perhaps also to dissociations between various levels of intentionality as described by Dennett ( Dennett, 1983).
An experiment was conducted on the local organization of behavior in rats in order to examine the generality of previous work on pigeons' key pecking, Lever pressing by two rats was reinforced according to a concurrent schedule of reinforcement for shorter and longer classes of interresponse times. Shorter and longer reinforced classes were terminated by presses on left and right lovers, respectively, and were cued by illumination changes within the chamber: the two reinforced behavioral patterns differed not only in temporal duration but also in the spatial location of their terminal lever presses. The interresponse-time distributions showed reasonably good control by the durations of the reinforced classes. As in previous experiments with pigeons, preference between patterns depended not only on relative but also on absolute durations of the patterns: preference for the shorter pattern increased as the absolute durations of both were increased, even though the longer was always three times as long as the shorter. This result has proven to be theoretically diagnostic and apparently has considerable empirical generality. Finally, the procedure permitted a kind of self-description by a rat of its own performance: the temporal patterning of behavior may be viewed as a skill, and the relative frequency with which a given temporal pattern is terminated by a response to the correct lever may be viewed as the