B.1.3.2. Results

The number of values excluded averaged 3 % of trials per participant (including outliers and errors). An ANOVA was performed on mean correct RTs, with target size (small, medium, large) and homogeneity condition (homogeneous, heterogeneous) as within-subject factors (see Figure 1). The target size effect was significant (F(2,34) = 10.72, MSE: 39,917, p < .001), but not the homogeneity effect (F(1,17) = 0.45, MSE: 446, p > .51). The interaction between both factors was significant (F(2,34) = 9.25, MSE: 26,709, p < .001), indicating a stronger effect of the target size in the heterogeneous than in the homogeneous condition.

Planned comparisons were then computed with one-tailed Student t-tests, with alpha thresholds adjusted according to the Bonferroni-type procedure proposed by Larzerele and Mulaik (1977), to maintain the family-wise error inferior to alpha (α = .05). In the heterogeneous condition, RTs were significantly shorter for the large target (1015 ms) than for the medium target (1076 ms, t(17) = 3.04; t'_crit = 2.03; p < .005), and were shorter for this latter than for the small target (1135 ms, t(17) = 1.95; t'_crit = 2.69; p < .034). In the homogeneous condition, no difference in RTs should occur. In order to increase the statistical power, we conducted uncorrected one-tailed t-tests (rather than post-hoc). Even these liberal tests revealed no difference, neither between large (1071 ms) and medium targets (1060 ms, t(17) = 0.56, p > .29), nor between medium and small targets (1083 ms, t(17) = 0.97, p > .17), nor even between large and small targets (t(17) = 1, p > .16).

Figure 3: Response times to the visual search task, in Exp. 1 (left panel) and Exp. 2 (right panel), for the two display-size conditions (3 or 6 items), for the two WM load conditions (single task, dual task, and with or without a salient distractor. Bars represent 1 S.-E.