Buffer storage modality and identification time in tachistoscopic recognition.

Abstract

According to Sperling's (1967) model of short-term memory briefly presented masked stimuli are rapidly read into a non-visual Recognition Buffer (the RB model). An alternative interpretation of the data is that the stimulus information is coded into a non-iconic Visual Buffer where it is held while a much slower recognition process takes place (the VB model). The high frequency of errors in experiments with sequentially presented stimuli appears to refute the possibility that recognition is as rapid as suggested by the RB model. However these data may be attributed to variations in effective stimulus duration and stimulus quality rather than to slow recognition time. In an experiment to control for these effects, normal, laterally inverted and spaced digits were presented in a rapid sequence (1–10 items/s) with intervening pattern masks to keep the stimulus/mask interval constant. The recall data showed that order errors increased with rate of presentation but that item errors remained invariant. At the fastest rates of presentation there were fewer order errors for spaced than for coincident digits. It was argued that the results, as a whole, were more consistent with the VB than the RB model and that there is no evidence for identification times as fast as 10–40 ms/item.