Fig. 2.
Audio-visual congruency enhances visual motion representations. Single trial linear discriminant analysis was used to extract EEG activations sensitive to the direction of visual motion (left in panels A,B) and to single trial choice (right). A) Discriminant performance (Az: area under the receiver operator characteristic) in data aligned to response. Time epochs with significant performance are indicated in red (at least p<0.01). B) Upper panels: Projection of single trial discriminant components, Y, extracted at time points of interest in the motion or choice discriminants (M1; C1). For a definition of the component activation see Eq. (1). These are shown separately for congruent and incongruent trials, normalized for effects of visual coherence and only for correct trials (see main text for results pertaining to all trials). Lower panels: Statistical contrast for a congruency effect (significant for the visual motion component M1; -0.34 s to -0.25 s; p<10-5; not in the choice component, C1). C) Single subject ramp onset times differed significantly between congruent and incongruent trials for the motion component M1 (p=0.02), but not for the motion component M2 or the choice component C1. D) Statistical contrast for a congruency effect in the discriminant component M1 when aligned to stimulus onset (significant from 0.31 to 0.37 s; p=0.001). E) Single trial modelling of choice revealed a significant influence of visual motion direction (p<10−5) and of the discriminant component (black; -0.31 s to -0.23 s, p<0.001). Including alpha power into the model furthermore revealed a significant interaction of alpha power with the discriminant component (gray; -0.24 s to -0.20 s, p<0.001). Lines and shaded regions indicate means and standard errors across participants (n=18). Boxplots indicate medians and quartiles. Δ: Congruent – incongruent.