B.3.6.4. Discussion

As expected, both increasing salience and providing information about a target's feature could facilitate attentional orienting toward this target. First, a target was processed faster when it was made salient by homogenizing the colour of the background items. This confirmed that the contextual salience of a target accelerated its attentional selection independently of this target's perceptual features (Beck & Kastner, 2005; Duncan and Humphreys, 1989; Nothdurft et al., 1999). These target salient trials reflected rather strictly exogenous control of attention

Second, cueing the target's colour also accelerated the RTs. This could not be attributed to any perceptual priming effects, since the cue was a word, and thus shared no feature with any item. Moreover, due to the heterogeneity of the background, the target was not salient in these trials. Thus, these cueing effects might reasonably be attributed to endogenous control of attention. This confirmed old results that showed that some items could be rapidly selected on the basis of some of their features (e.g. Von Wright, 1968).

Third, these two processes could have cumulative effects. Knowing a non-spatial feature of an already salient target further accelerated its attentional selection, even when salience and relevance concerned a different dimension. This contradicted the assumption that attention is only driven by salience in a first stage, during which the processes involved are cognitively "impenetrable" (Theeuwes, 1993, 2010a). This also suggested that relevance could not only act by modulating, or "weighting", the computations involved in salience (Müller et al., 1995; Wolfe, 1994; Navalapakkam & Itti, 2005), but that it could be considered as an independent signal combining with the salience signal. Indeed, in the different dimension condition, the cue did facilitate the attentional selection of a salient target, not by indicating (and maybe weighting) the dimension in which the target would be salient, as considered by these "dimension weighting" models, but by providing knowledge about a non-salient feature. This kind of effects is more consistent with the biased competition hypothesis.

From the present results, several conclusions could be addressed. First, feature-based endogenous and exogenous processes could act concurrently to bias competition in perceptual systems. They are not independent processes that only race to control attentional selection, but rather could genuinely integrate together (at least when they concerned the same perceptual dimension). This suggested a second conclusion: that these signals likely integrate within the perceptual system. These results strongly supported the integrated competition hypothesis. Indeed, they could be interpreted without referring to the salience map concept, casting some doubt on its usefulness and validity under some conditions.