B.1.3.3. Discussion

This first experiment confirmed that size superiority (i.e. being larger) induced genuine salience effects (Proulx & Egeth, 2006) and evidenced that these effects could last rather long. According to the pattern of results in the heterogeneous condition, one interpretation was that the attentional focus moved progressively from the most to the least salient items. The more salient (larger) was the target, the shorter the RTs; this was true for each salience variation. These processing priorities -of the large item over the medium one and of the medium item over the small one- were most likely due to salience. Indeed, several other possible explanations could easily be dismissed. First, the pattern of RTs did probably not reflect any strategy, since all three items had exactly the same relevance (probability of being the target), and since this was clearly explained to the participants. Thus, these latter had probably not adopted an attentional set, neither toward a singleton (since all items were actually singletons, strictly speaking), nor toward the size dimension (since it was not relevant at all). In addition, no physical or sensory property could account for the larger item RT benefit. Indeed, the size effect was significantly weaker in the homogeneous than in the heterogeneous condition, up to be virtually null. Even when search was made unnecessary (in the pre-test; see footnote 1), no difference emerged between RTs for large and small targets. Finally, all the salience effects observed for mean RTs were also observed for the first RT quintiles (see Figure 2, top panel and Tableau 1). One possible alternative account of the mean RT decrease with salience might have been that people attended the salient item more often at first place, without selecting the salient items before, or faster, than others. The analysis of the first RT quintiles showed that this interpretation was inaccurate and that size effects corresponded to genuine processing acceleration. Taken together, all these results evidenced that the size effects observed were indeed essentially salience effects.

The graded effects observed in the random conditions showed that the largest item was salient, while the smallest was not. Treisman and Gormican (1988) showed that maximal singletons (e.g. a large target among small distractors) could be processed more efficiently than minimal singletons (e.g. a small target among large distractors) in visual search. Consistently, Braun (1994) evidenced that a concurrent task disrupted more a visual search for minimal than for maximal singletons. The present results were in line with these conclusions, suggesting that only maximal singletons captured attention. Some authors showed that minimal singletons could be found efficiently in a visual search (e.g. Hodsoll and Humphreys, 2001; Nothdurft, 2006). However, Yantis and Egeth (1999) argued that search efficiency should not be equated with exogenous control of visual attention. Search efficiency and salience not only depend on target-distractors similarity (Duncan & Humphreys, 1989). The relative intensity of the feature inducing this salience seems to determine salience too. This point might only be evidenced, however, when endogenous influences were strictly controlled.

Nothdurft (2000) suggested that exogenous salience could be graded on a scale. This idea, consistent with the influence of intensity, was strengthened by the present results. Indeed, RTs differed between each salience level: Not only did the large target take less time to be processed than the medium one, but this latter also took less time than the small one.

Finally, observing these graded effects of salience also supported the main hypothesis of this article, that salience effects could last long. Indeed, the processing time difference between the small and medium targets occurred although these items were never the most salient of the display and were not processed at first place. This suggested that salience effects were still at play when the second most salient item came to be processed -- since it was processed faster than the least salient one. Furthermore, the salience effects between the large and medium items (and even between the medium and small ones) remained significant for the longest RTs. These results suggested that salience endured rather long and was not restricted to the first "attentional shift".

In summary, size superiority seemed to induce gradable salience, and hence to attract attention exogenously, even in absence of any attentional set. These salience effects appeared long-lived, in line with the classical view of salience-based attentional guidance of visual search (Itti & Koch, 2001; Koch & Ullman, 1985; Theeuwes, 1993). These data seemed at odd with those of Donk and van Zoest (2008), who found only transient salience effects in a salience comparison task. Nevertheless, we postulated that endogenous influences were at play in their experiments (see Introduction). Following the same reasoning, in the present experiments, cueing one size or an other could probably have strong influences on salience effects.