B.3.6.3.2. RT Distributions and race model

On the basis of these results, the hypothesis of a genuine integration between salience and relevance was tested for each similarity condition, through the race model inequality proposed by Miller (1982; Ullrich, Miller, & Schröter, 2007). When relevance and salience concerned a different dimension, no integration was evidenced at any point of the RT distribution (see Figure 11, panel B). Thus, no statistical test was necessary. On the contrary, when relevance and salience concerned the same dimension (see Figure 11, panel A), the Miller's inequality was falsified in all the 11 first percentiles (i.e. from 5 % to 55 %). Following the recommendations of Kiesel, Miller, and Ulrich (2007) to reduce type-I statistical errors, the inequality was only tested at the four percentiles 5 % 10 %, 15 %, 20 % and 25 %. The integration significantly violated the Miller's inequality at percentiles 10 % to 20 % (all ps < .03), and nearly significantly at percentiles 25 % (p < .065). Interestingly, the integration values differed significantly from the similar to the different condition dimensions (all ps < .05) for all percentiles.

Analyses were also carried out on raw data (see Data analysis, section , and Figure 11, panel C). For the different-dimension condition, the integration was positive on 19 time points between 233 to 257. The differences were never significant. For the same-dimension condition, the integration was positive from time points 202 to 284, and from 294 to 398. These values differed significantly from zero from the point 310 to the point 346 (all successive ps < .05, unilateral uncorrected t-tests; see Figure 11, panel C, red shaded bar).). Then, from point 347 to point 379, all ps but 2 were close to be significant (all ps < .10. Finally, these integration values were compared between the two conditions of similarity: they differed significantly from time points 310 to 475 (see Figure 11, panel C, blue shaded bar).

These results confirmed those obtained with mean RTs, suggesting a stronger interaction between salience and relevance when both processes concerned the same dimension. Moreover, only in this latter condition was evidenced a genuine integration, and the effect of perceptual similarity on this integration was significant.

Figure 11 (Next page) A.-B.: Cumulative probability distributions for the salience alone (red dotted line), relevance alone (blue dotted line) and salience and relevance combined (purple solid line) and the cumulative probability predicted by the race model (black trace), for the same dimension condition (panel A) and for the different dimension condition (panel B). C.: The difference in response probability between integration trials and race model predictions for the same dimension condition (red bold line; red thin line represent 1. S-E.) and the different dimension condition (blue bold line; blue thin line represent 1. S-E.) illustrate a genuine integration only for the former one (red shaded bar: 310 - 346 ms). The two integration values differed one from the other from time point 310 to 475 ms (blue shaded bar).