By contrast, in the remove condition, a two-way ANOVA revealed a highly significant underadditive interaction between cue validity and onset presence, F (1, 23) = 18.89, p < .001, η p 2 = .45, indicating that the effect of cue validity was smaller when the onset was present (123 ms) than when it was absent (143 ms).
The interaction between cue validity and onset presence only approached significance in the “classic” situation in which all stimuli remained on screen, F (1, 37) = 3.33, p = .076, η p 2 = .08, but attained significance in the remove condition, when all items but the target disappeared after a short period, F (1, 37) = 8.27, p = .006, η p 2 = .18.
The remove condition also led to a slight, 2 % reduction of three-saccade trajectories when the eyes were initially captured by the onset, but this difference failed to reach significance, t (23) = 0.859, p = .377 (as did all other paired comparisons between corresponding saccade proportions in the remove and not-remove conditions).
In view of this, finding the underadditive interaction approaching significance in the not-remove condition of Experiment 2 is not very remarkable.