Paired t-tests conducted separately for valence, arousal, and surprise ratings indicated that all six experimental conditions (see Table 1B) differed significantly from each other (p≤.05 in all tests), except valence and arousal ratings for unexpected and very unexpected in the expressive condition, and surprise ratings for unexpected and expected in the non-expressive condition.
Thus it was expected that neither ERAN nor N5 would be influenced by aspects giving rise to emotion, and that they would thus not differ between the non-expressive and the expressive condition.
The data from the expressive condition show that the neural mechanisms underlying the generation of both ERAN and N5 are also involved in the processing of real naturalistic music.
In addition to the larger N5 elicited by expressive compared to the non-expressive chords (see above), the expressive chords also elicited an increased negativity over central electrodes in the time window from 80 to 120 ms (this effect is presumably an increased N100 due to the fact that the critical chords were usually played more loudly in the expressive condition, see also Discussion).
Finally, the heart rate calculated for entire excerpts (reflecting the vagal tone during listening to the excerpts) was identical for expressive and non-expressive excerpts, rendering it unlikely that the task of detecting the timbre deviants was more engaging, or more attention-demanding, in the expressive condition.
Second, the behavioural data of the timbre detection task indicate that this task was comparably easy in both expressive and non-expressive conditions.
Figure 3A shows the skin conductance responses (SCRs) to expected, unexpected (original), and very unexpected chords, averaged across all subjects (and across both expressive and non-expressive conditions).
Note that purely physical differences between expressive and non-expressive conditions cannot account for this effect, because we compared the difference in SCRs between expected and (very) unexpected harmonies, separately for the expressive, and for the non-expressive music.
Although the ANOVAs did not indicate an interaction between factors chord and expression, we also inspected the single subject data sets for differences in SCR effects between the expressive and the non-expressive conditions (to exclude that the large variance typical for SCR data rendered the results of the ANOVA spurious).
Importantly, the ERAN did not differ between the expressive and the non-expressive condition.
