Efficient filtering of irrelevant global and local information when target level and location are random

Summary Using compound letter stimuli (Navon, 1977), Paquet and Merikle (1988) found that subjects are very good at ignoring distractor letters shown in an irrelevant location when the distractors are also shown at the level of the structure opposite to that of the target. However, the procedure used by Paquet and Merikle (1988) might have biassed the results toward showing relatively poor non-target processing in this instance. The target level of structure was a blocked variable, and so subjects could ignore the other level of structure completely. In the present experiment, the target level was either blocked, requiring focussed attention to one level of structure, or else it varied randomly from trial to trial. This latter condition should require attention to be allocated to both levels of structure. Results were the same in both cases, namely efficient filtering of nontarget information when it was not at the target level of structure. The results are related to proposals (Duncan, 1981; Van der Heijden, 1981) that semantic analysis of visual information is an unlimited-capacity process.