ABSTRACT
Models of the memory control of continuous visual search are discussed at three levels of analysis. At the sample level there are two rival views concerning the nature of the memory representation that enters the comparison between stimulus information and memory information: memory control of search can be exerted by a memory representation of the target(s) or of the nontarget(s) (target control versus context control). At the trial level two principal modes of temporal organization of subsequent sample analyses are distinguished: a one-process structure consisting of a chain of equivalent sample analyses and a two-process structure with an additional check at the end of the trial (scan-alone versus scan-and-check). At the level of training sessions, a model with three stages of practice is proposed. According to the model the search is under exclusive target control in Stage I and under exclusive context control in Stage III (scan-alone mode in both cases). Stage II mediates the transition from Stages I to III by means of a scan-and-check mode of processing, with the scan under context control and the check under target control. Evidence from a training experiment suggests that, within the range of task complexity employed, Stage I processing is restricted to the very first trials of the training period. Even the transition from Stage lI to III can only be observed when the context is rather complex. It is concluded that search can be conducted under context control even without much practice.
