Two experiments were conducted to explore the switching of attention between perception and memory. In Experiment 1, college students performed a task that required them to cycle or switch attention between perceptual and memorial inputs. Switching times of 293 and 376 msec per switch were obtained on the basis of two formulas. In Experiment 2, the attentional load was manipulated by varying the number of perceptual and memorial inputs. Switching time increased as a function of list length, indicating that item load affects both the control processes that set attentional allocation policies and one's ability to perform memory and/or perceptual tasks. These results suggest that modularity, or encapsulation of item and control-process systems, does not hold. A model is presented that depicts the relation between item and control-process representation in rapidly alternating attention between perception and memory.Many common tasks, such as reading (Baddeley, 1982; Byrd & Gholson, 1984) and using a memorized shopping list, seem to involve a rapid cycling of attention or switching between information available in the environment and information stored in memory. The nature of switching between perception and memory is the concern of the present study.In the past, typical methods used to study attention switching involved either dichotic listening, in the classical tradition of Broadbent (1958), or the manipulation of expectancies as to the spatial location of critical events (Shulman, Remington, & McLean, 1978). In the dichotic listening task, each ear concurrently receives a different set of digits to be recalled. The pattern of recall does not typically follow the order of presentation, but is first by one ear and then the other, presumably because of the difficulty or time required to alternate attention between ears. In the expectancy manipulation approach, the probable location of a target is cued, and then the target is presented. In cases of agreement between cued and real target information, there is a savings in reaction time over a control condition with a noninformative cue. In cases of disagreement between cue and target location, there is a substantial increase in processing over a control condition. This latter difference could be interpreted as a switching time.However excellent the dichotic and expectancy methods may be for the study of certain aspects of attention, they may not be entirely appropriate for the study of rapid selfpaced switches between perception and memory. TheThe first author would like to acknowledge the support and use of facilities from the Department of Psychology and Social Relations, Harvard University, during his sabbatical leave. D. B. Burt is now at the
