B.1.2.1.1. Salience effect time-course

One strong assumption behind the hypothesis that visual search could follow display saliences was that salience effects could last long. The present experiments addressed this issue by assessing how the saliences of multiple items in a display determined the respective processing times of these items. It was well demonstrated that the most salient item of a display could gain attentional priority (Yantis & Jonides, 1984; Theeuwes, 1992). However, how the saliences of the other, less salient, items influenced their processing was rarely studied, and theories about the follow-up steps of the attentional focus remain largely speculative. This matter might be informative about the processes underlying salience.

In the eye movement control domain, some evidence suggested that salience determined gaze orienting, at least partly, in natural scene free-viewing conditions, for a rather long duration (Parkhurst, Law, & Niebur, 2002). Einhäuser, Rutishauser, and Koch (2008) also showed that salience could influence eye movements in similar conditions, but that top-down influences could nevertheless completely override these salience effects (see also Foulsham & Underwood, 2007).

In the domain of covert visual attention, Donk and van Zoest (2008) recently presented contradictory experiments suggesting that salience effects were short-lived, even when relevant for the task at hand. These results questioned the important assumption that salience effects could endure rather long, at least when they are not harmful for the task (Itti & Koch, 2000; Michael et al., 2006; Theeuwes, 1993; Wolfe, 1994). Donk and van Zoest (2008) asked participants to perform salience comparison tasks. The display always contained two salient items of different saliences. Participants had to make a speeded saccade toward the most salient item (Exp.1), or a manual response about the side of the most salient item (Exp.2). The authors only observed salience effects in the first quintiles of the saccadic RT distribution, and for very short display presentation durations for manual responses. They concluded that visual salience could only be transiently represented. Unfortunately, these experiments might be criticized. There are two intermingled problems. First, the task relied on some kind of introspection. Second, the comparison task intrinsically involve to counteract salience.

On the one hand, participants had to make a direct judgement on the process that was studied, that is, on salience. This methodology should be used cautiously, since one do not know how their evaluation would influence the signal of salience, that was postulated to be exogenous. The instructions explicitly involved to compare the saliences of the two concurrent salient items (participants always had to select the most salient among two salient items). It is impossible to compare two stimuli without processing them both to some extent. And it is impossible to process the two items without counteracting salience, since one of them is more salient than the other --and should thus attract attention at the depend of the other.

The methodology of these authors, somewhat reminiscent of introspection, could be useful, but it could not inform unambiguously about the time-course of salience effects, as long as salience is considered as an exogenous signal. To avoid these effects, a more objective task should have been used, in which the incentive were strictly controlled. This was the case in the present experiments, which showed that salience effects could last long under some conditions. The real strategy of participants, and the mechanisms involved, should always be carefully evaluated, especially when a subjective judgement was used to estimate the process of interest.

Endogenous processes could dramatically influence salience effects (e.g. Einhäuser, Rutishauser, & Koch, 2008; Foulsham & Underwood, 2007; Lamy et al., 2004, Michael et al., 2006; Müller & Krummenacher, 2006). Most of the visual selective attention models considered that relevance and salience could have interactive effects (e.g. Fecteau & Munoz, 2006; Michael et al., 2006; Müller, Heller, & Ziegler, 1995; Navalapakkam & Itti, 2005; Wolfe, 1994; but see Theeuwes, 1993, 2010a, for an opposite opinion). The results of Donk and van Zoest (2008) might thus be linked to some strategical influences counteracting salience effects, thus shortening their life expectancy.