Figure 4. Effects of congruency on test sound ERPs. Congruent (blue ) and incongruent (red ) test sound evoked grand-averaged ERPs, as well as their difference wave (congruent – incongruent, in black ) are plotted at the Fz and the Pz electrodes. A series of topographical plots showing the difference between congruent and incongruent sounds across time and sensor space illustrates the spread of the significant positive cluster across time. Sensors exhibiting significant differences across conditions are highlighted in black.

Correlations between neural and behavioural effects of agency

Differences in %Correct between the agent and observer condition exhibited a great variability across individuals (M = 9.7%, SD = 7.1%, range = 0-26%). Thus, active learning was more efficient than learning from observation for some, but not all participants. Aiming to uncover the neural underpinnings of active learning benefits, we tested for possible correlations between the benefit that a participant would have of agent over observer acquisition for their memory of the movement-sound associations, and the degree to which agency modulated auditory ERPs during early stages of acquisition, when the active learning benefits were maximal.
We found a significant correlation between agency effects on N1 amplitude of acquisition sounds at Fz in the early learning stage and agency effects in memory performance [r(21) = 0.6, p = .002]. The stronger the N1 agency effect that a participant experienced during early acquisition, the stronger the memory advantage for the agent vs. the observer conditions (figure 5).
In order to visualize this finding, we decided to split the participants into two groups, using the median of the difference between agent and observer %Correct as a cutoff point (M = 7.8%).
Participants who had stronger behavioural agency effects than the median were considered “active learners” (N = 11), and participants that fell below that median of behavioural effects were considered “indifferent learners” (N = 12).
We compared the overall performance (agent + observer condition) of active and indifferent learners. Active learners performed overall better in the behavioural task (M = 78%, SD = 13%) as compared to indifferent learners (M = 68%, SD = 11%). A two-sample t-test showed that this difference is statistically significant [t(21) = 3.844, p < .001, d = 0.80].