FEEDFORWARD PROCESSING OF PICTURES 1
Supplementary Materials
Additional analyses. A different way of analyzing yes-no data that is closely related to d’ compares the proportion of correct yes responses (hits) and incorrect yes responses (false alarms). It has the merit of showing the proportions of hits and false yeses (see Figure S1), rather than merging them in a single d’ measure, and it also includes a measure of the significance of the difference between hits and false yeses, equivalent to the significance of d’ from 0.0. In each experiment repeated-measures analyses of variance (ANOVAs) were carried out on the proportion of yes responses as a function of before/after group, target presence (hits) or absence (false yeses) and presentation duration (80, 53, 27, or 13 ms per picture).Planned paired ttests at each duration, separately for each group in each experiment, compared the proportions of hits and false yeses. Overall, the results in these analyses track very closely the d’ results reported in the main text.
FigureS1. Results of Experiments 1 and 2 in which participants detected a picture that matched a name given before or after the sequence of six (Experiment 1: left panel) or twelve (Experiment 2: right panel) images.Figures depict proportion yes responses to target (hits) and non-target trials (false alarms). Error bars depict the standard error of the means. This information is presented in the paper as d’ values.
With six pictures (left panel), there was a main effect of target presence: the proportion of hits (correct yeses on target-present trials, M = .69) was significantly higher, F(1, 30) = 525.73, p < .001, ηG2= .66, than the proportion of false yeses on target-absent trials (M = .25). Notably, there was no main effect of group (Before/After), F < 1, and only one interaction with group, the triple interaction with target presence and duration, F(3, 90) = 7.31, p < .001,ηG2 =.033. Having the target name presented before rather than after the sequence benefited detection substantially at 80 ms but not at all at 13 ms, with the other durations falling in between. There was also a main effect of duration, F(3, 90) = 4.21, p < .01,ηG2 =.037, and an interaction with target presence, F(3, 90) = 42.20, p < .001,ηG2 = .165. Detection (the difference between the proportion of hits and false yeses) improved as the duration increased from 13 to 80 ms. Separate paired t tests showed that the effect of target presence was significant at each duration in each group (p < .001).
The main results of Experiment 2 (right panel) were similar to those of Experiment 1. Overall, the proportion of hits (correct yeses on target-present trials, M = .63) was significantly higher, F(1, 30) = 362.27, p < .001, ηG2= .53, than the proportion of false yeses on target-absent trials (M = .26). As in Experiment 1, there was no main effect of group (Before vs. After), F < 1, but there was an interaction with target presence, F(1, 30) = 8.03, p < .01, ηG2= .024, with a lower false yes rate when the target was presented before the sequence. The main effect of duration was not significant, although there was an interaction with target presence, F(3,90) = 27.87, p < .001,ηG2 = .165: the longer the presentation duration, the lower the false yes rate and the higher the proportion of hits. No other effects were significant. Separate paired t tests showed that the difference between the proportion of hits and false yeses was significant at each duration in each group (p < .001, except p < .01 at 13 ms in the After group.