Marking effects in Pavlovian trace conditioning.

Abstract

In four experiments we investigated pigeons' acquisition of a successive discrimination with a trace autoshaping procedure. The conditioned stimuli were 5-s presentations of colored key lights, one of which was followed by food after a 5-s delay. In Experiment 1, which used spatially denned cues, we found that acquisition of differential responding to the reinforced cue was facilitated when a brief flash of light immediately followed both reinforced and nonreinforced cues. Experiment 2 found a similar enhancement by the added light flash in a purely visual discrimination. Experiment 3 found that the flash facilitated learning only when presented immediately after the discriminative cues, and not when it occurred immediately before the cues or at the time of reinforcement. A fourth experiment found this facilitation effect only when the flash and reinforcement occurred on the same trial. These results are interpreted in terms of marking: The flash enhanced learning because it triggered a backward scan through recent memory to search for possible predictors. Such a search would give extra processing to the preceding cues and increase the likelihood that they would be remembered and associated with the delayed food reinforcement. With possible sources of differential secondary reinforcement eliminated, a delay of as little as 5 s in following a response with reward or nonreward has been shown to severely impair a rat's ability to solve a visual discrimination task (Grice, 1948). The implication of this result is that two events must be temporally contiguous if they are to be associated one with the other. Recent research, however, indicates that lack of temporal contiguity between an animal's response and the consequences that follow from that response does not necessarily prevent the animal from being able to associate these two events. For example, rats have proved capable of learning the correct path through a T-maze to obtain food despite experiencing delays of up to 2 min between choosing one of two alleys and subsequently obtaining reward or nonreward in the goal box (Lieberman, Mclntosh, & Thomas, 1979; Thomas, Lieberman, Mclntosh, & Ronaldson, 1983). The critical feature enabling rats to form associations over a delay in these maze experiments was the occurrence of some brief, salient event (handling, white noise, a light) immediately after the rat had chosen one of the two alleys. Because the salient event followed both correct and incorrect choices, differential secondary reinforcement cannot be invoked as an explanation of the long-delay learning observed in these experiments. These results are interesting partly because they cannot be readily explained in terms of already known processes. The results may also necessitate a revision of current views of information pro