The sandwich effect: The role of attentional factors in serial recall.

Five experiments tested the prediction, from a simple chaining model, that interleaving irrelevant material will substantially disrupt immediate serial recall. Experiment 1 interpolated long or short words between items in an auditory digit span test. These two conditions disrupted recall to an equal but moderate extent. Experiment 2 presented mixed lists of digits and words, cuing one or the other before or after presentation. Precuing led to substantially better recall. Experiment 3 used articulatory suppression to rule out the hypothesis that recall was protected from the sandwich effect by subvocal rehearsal. Experiment 4 combined the sandwich effect with a concurrent task, finding clear effects of both but no interaction. Experiment 5 showed that the predictability of interpolated material did not influence recall. These results can be explained by adding an attentional preprocessor to standard chaining models. Immediate memory span, like speech itself, requires that order information be preserved. This has presented a major problem both to models of immediate serial recall and to models of language behavior. Although the range of models of serial order is wide, they tend to fall into two broad categories, one assuming that order is imposed by means of an external structure and one assuming a form of chaining whereby associative links are formed between successive items (see Lewandowsky & Murdock, 1989, for a more detailed discussion of existing models).