Discussion

Experiment 2 replicated Experiment 1 with pure tones instead of piano tones. The goal of this experiment was to confirm the cognitive interpretation of the priming effect of Experiment 1, while ruling out the alternative sensory explanation raised by the simulations with piano tones. In Experiment 2, human listeners showed a tonal priming effect with pure tones despite the failure of the sensory model to predict a difference between related and less-related targets. This finding thus provides strong evidence for cognitive processes based on tonal knowledge in non-expert listeners. Even when sensory components including tone repetition, melodic contour, intervals, and spectral information (i.e., overtones, virtual pitches) are controlled in the material, the melodic stimuli are still informative enough to generate tonal expectations at a cognitive level, resulting in facilitated processing for tonally related events.

This finding does not imply that sensory components do not have any influence in music processing (beyond tonal knowledge activation). To get a first indication of the respective importance of cognitive and sensory components in tonal priming, we compared the priming effects (for correct response times) between Experiment 1 and 2 in two ways: 1) by calculating the effect sizes (i.e., Cohen’s d) of the priming effects in both experiments, and 2) with a 2x2x2 ANOVA with Tonal Relatedness and Target Timbre as within-participant factors and Experiment as a between-participants factor. Cohen’s d is the difference between group means (here, related and less-related conditions) expressed in standard deviation units. The observed size of the tonal priming effect (collapsed across Target Timbres) was d = .59 in Experiment 1, which is categorized by Cohen as a medium effect, and d = .38 in Experiment 2, which is categorized as a small effect (small, medium and large effects are respectively defined by d > .20, d > .50, and d > .80; Cohen, 1988). The stronger effect size in Experiment 1 suggests that sensory components contribute to stronger tonal expectations, even if these expectations are still elicited without them (Experiment 2). However, no interaction between Tonal Relatedness and Experiment was observed in the ANOVA (p = .55). In addition, this analysis revealed a main effect of Experiment, F(1,54) = 11.58; MSE = 96036; p < .01, with longer response times for Experiment 2. This finding together with fewer correct responses and larger intra-participant standard deviations for correct response times in Experiment 2⁸ suggest that the difference in effect size might rather reflect the greater difficulty of the timbre discrimination task in Experiment 2 than stronger tonal expectations in Experiment 1.The increased task difficulty, which is likely to be explained by participants’ lack of familiarity with pure tones, renders thus difficult the interpretation of directly comparing the effect sizes. The reduced sensory components as well as the greater task difficulty might account for the smaller effect size in Experiment 2. Future studies should thus aim for priming tasks with equal difficulty in order to test whether adding sensory components to the cognitive components changes the effect size of the priming effect.⁹ However, the main goal of our study was more modest than assessing the respective contributions of sensory and cognitive components: we aimed to show whether cognitive expectations were strong and stable enough across nonmusician listeners to elicit significant priming effects when sensory expectations were controlled, which they did.