Top-down contingent attentional capture during feed-forward visual processing.

Theeuwes (2010) summarizes an impressive number of studies demonstrating interference by irrelevant visual singletons in computer experiments with humans. In these studies, if participants search for a relevant singleton target, such as the single diamond among circles (i.e., a shape singleton), an irrelevant singleton distractor, such as the single red circle among the green stimuli (i.e., a color singleton), delays the correct response to the target (cf. Theeuwes, 1991, 1992). If this distractor – as was the case in the relevant experiments – does not predict the likely target position, it is said to be irrelevant. Theeuwes (2010) argues that the interference by the irrelevant distractor reflects stimulus-driven or exogenous capture of attention by singletons or feature contrasts, which he assumes to occur early during the feed-forward phase of visual processing. According to this explanation, a stimulus initially captures attention exogenously (i.e., in a stimulus-driven way) to its position, to the degree to which it is salient. Salience in turn is defined as the summed local contrast in terms of color, luminance, and orientation difference to the surrounding stimuli or the background (cf. Bergen & Julesz, 1983; Itti & Koch, 2001). Theeuwes (2010) assumes that this salience-driven capture of attention is fast and occurs within 150 ms since singleton onset, during the feed-forward phase of visual processing (cf. Lamme, 2003). In contrast to Theeuwes (2010), we think that top–down contingent capture is the rule and explains initial and fast attention capture effects in the first feed-forward phase of visual processing. During a later phase and under some conditions exogenous capture of attention possibly follows. At the same time, we propose that the evidence presented by Theeuwes fails to support exogenous orienting because it fails to exclude a top–down contingent capture explanation. We present our arguments in two sections. First, we review the evidence directly supporting our claim that salience capture is subject to top–down control. From this evidence we derive the exogenouscapture criterion that must be met by experiments for demonstrating exogenous spatial attention, and show that this criterion in notmet by the studies reviewed by Theeuwes. Second, we review studies showing attentional capture by stimuli during the feed-forward phase, and conclude that only top–down contingent capture but no