spatial tasks. A second task variable might be responsible for variations in resistance to retroactive interference from delay-interval stimulus changes. In conventional delayed conditional discrimination tasks, the to-be- remembered stimuli are rather simple (a colored pecking key, for example). In spatial tasks, the functional, extramaze cues are multidimensional (e.g., a window, a relay rack, the experimenter). In the spatial task, then, the memory trace is likely to be richer (composed of more attributes) than in the delayed conditional discrimination task. The resulting redundancy might then provide some measure of protection against interference from delay- interval stimuli in the spatial task. Now the challenge is to construct a theory of working memory from which the effects of task variables can be deduced. If it is to cope with these and other facts (e.g., Beatty & Shavalia, 1980a; Shavaliaet al., 1981). such a theory is likely to be much more complex than the simplistic model of memory maintenance embodied in Equation 1.
Tasks like delayed matching-to-sample can be used to assess working memory in animals. Performance is accurate to the extent that the subject remembers a conditional stimulus (the sample) throughout a retention interval (delay). A common finding is that a change in delay-interval stimuli reduces matching accuracy (retroactive interference). One interpretation of retroactive interference 'earlier formulated holds that novel delay-interval stimuli capture some of the animal's limited attentional capacity thereby detracting from rehearsal of the sample memory and permitting trace decay. In its simple form, the rehearsal hypothesis fails to predict two sets of phenomena described in this chapter. First, with sufficient training, pigeons' delayed matching performance recovered from the initially detrimental effects of a change in delay-interval illumination. The rate of recovery was faster if the change occurred early than if it occurred late in the delay. Recovery from light interpolated early in the delay produced faster recovery from subsequent whole-delay illumination than did recovery from light interpolated late in the delay. Second, using a large T-maze, pigeons were found to perform delayed alternation well at long delays and to rely on extramaze cues. Memory for spatial locations exhibited in these experiments was resistant to potential sources of retroactive interference like changes in delay-interval illumination. At the present time, recovery from and resistance to retroactive interference remain problems for a theory of working memory.
The author's research described in this chapter was supported by grant RO1-MH3 1432 from the National Institute of Mental Health. I thank Becky Aaland, Curt Borchert, Laurel Knoell. Rebecca Nelson, Robin White, and particularly Kathryn Bengtson and Deborah Olson for their assistance with and contributions to many aspects of the work.