B.1.4.2.1. Vincentization

Given the lack of significant interaction involving the nature of the cue, we merged RTs of both conditions (symbolic and perceptual) to compute vincentized RT distributions (Ratcliff, 1979). We then computed planned comparisons between conditions for the two first and two last quintiles, as in Experiment 1. Again, differences were considered significant when probabilities were below threshold (or closed to be, i.e. : p < .075) for two consecutive quintiles. Statistical results are presented on Tableau 1, and values on Figure 2 (middle panel). They grossly confirmed the mean RT results. In the random condition, differences between large and medium, and between medium and small items were significant for both shortest and longest RTs. In the large-relevant condition, these differences were also significant for the shortest RTs, as for mean RTs. On the other, only the large-medium comparison reached significance for the longest RTs. In the small-relevant condition, the only significant effects were between small and medium items for the shortest RTs. Finally, cueing the size also benefited to the relevant target, relative to the random condition, even for the shortest RT quintiles. Indeed, for the small target, the RTs were shortened by cueing in both first quintiles (71 ms, SE: 13; t(37) = 5.31; p < .001; 83 ms, SE: 15; t(37) = 5.64; p < .001). For the large target, the RTs were also shortened by cueing in both first quintiles (46 ms, SE: 10; t(37) = 4.67; p < .001; 73 ms, SE: 14; t(37) = 5.24 ; p < .001).