General Discussion

The main goal of our study was to extend tonal priming obtained for tonic versus subdominant for chords and melodies (Bigand & Pineau, 1997; Bigand & al., 2003; Marmel et al., In Press) to a subtler contrast in tonal stability: mediant versus leading tone. This choice of tonal degrees also answered to the criticism that expectations for the tonic at the end of a musical phrase might reflect expectations for tonal closure. Our study did not include the tonic as target and still reports priming. The facilitated processing for mediants over leading tones shows that fine-grained tonal knowledge can be assessed implicitly in listeners with limited or no formal musical training. The control of sensory expectations driven by tone repetition, melodic contour, or intervals, together with the removal of the confounded expectation for the tonic as providing closure, pleads for the interpretation of expectations involved in musical priming as reflecting cognitive representations of tonal stability. Participants seem to be able to extract tonal centers and process tonal stabilities with very little cues (melodies in Experiment 1b and 2 differed only by one note). This confirms that nonmusician listeners have elaborated tonal knowledge (Bigand & Poulin-Charronnat, 2006).

A computational model of tonal knowledge has been proposed by Bharucha (1987) and elaborated by Tillmann, Bharucha, and Bigand (2000). The MUSACT model is a three-layer network of interconnected tone, chord, and key units. The knowledge of Western tonal regularities is not stored explicitly, but emerges from activation reverberating between units. The tones of a musical piece activate the corresponding tone units, and activation spreads to related chord and key units. Activation then reverberates from key units to chord and tone units, and the reverse, until equilibrium is reached. MUSACT incorporates a memory decay: activation of the units depend on the activation due to previous events, weighted according to recency. For one event e, the total activation of a unit represents: 1) the sum of the bottom-up activation caused directly by event e, 2) the indirect activation received from other units in response to event e (i.e., the phasic activation spreading in the system), and 3) the decayed activation caused by previous events. MUSACT has been simulating numerous data on music perception, notably priming data with chord sequences. When prime contexts were presented to the model and the activation pattern for chord units representing the targets were read out, higher activations were reported for tonic than for subdominant chords. This was interpreted as stronger expectation for the tonic chord to come next, thus predicting faster response times (Tillmann et al., 2000). More recently, it has been shown that MUSACT also simulates the priming effect for tonic over subdominant target tones in melodies (Marmel et al., Submitted). We tested MUSACT with our present melodies and the fine tonal manipulation (mediants versus leading tones). Related and less-related melodies were presented to the model without the target tones, and activations of tone units representing the targets were read out and analyzed by two-sided paired t-tests. To focus on top-down activations, the simulations used only the reverberating phasic activations for the tone units (and not the initial input stimulation, see Tillmann et al., 2000). Activation levels were higher for tone units representing related targets than for those representing less-related targets (Experiment 1a: t(11) = 2.06, p = .06; Experiment 1b: t(10) = 4.58, p < .01; Experiment 2: t(11) = 2.98, p < .05). Thus, the activation pattern of tone units provides a good fit with our priming data and strengthens MUSACT as a model of listeners’ tonal knowledge and its influence on music processing.

The priming effects observed in our study were smaller than the effects reported for comparisons including the tonic degree. Effect sizes were calculated with Cohen’s d, which allows distinguishing between small effects (.20 < d < .50), medium effects (.50 < d < .80) and large effects (d > .80) (Cohen, 1988).The effect sizes calculated for the more continuous timbre in Experiment 1 and for the in-tune targets in Experiment 2 were in the small effect range (d = .34 and .39 for Experiment 1a and 1b respectively, d = .40 for Experiment 2). In comparison, the priming effect for tonic versus subdominant tones was a medium effect (d = .66) for the more continuous target timbre in a timbre discrimination task (Marmel et al., submitted), and it was a large effect (d = .97) for the in-tune condition of the intonation task (Marmel et al., In Press). These stronger effect sizes point to the cognitive importance of the tonic, which has been considered as a cognitive reference point in Western tonal music (Francès, 1958; Krumhansl, 1990). The special role of the tonic has been shown by harmonic priming studies comparing tonal contexts against baseline contexts (without a tonal center): A benefit of processing was observed only for the tonic, whereas a processing cost was observed for the subdominant, and neither benefit nor cost were observed for the dominant (Tillmann, Janata, Birk, & Bharucha, In Press).

In our view, the small effect sizes found in the present study need to be considered in light of the weakness of the acoustical manipulation (melodies in the two tonal conditions were almost identical) and in light of the weakness of the tonal manipulation (mediant versus leading tone). Observing an effect of this tonal manipulation on tone processing suggests the psychological relevance of the tonal expectancies (Prentice & Miller, 1992). Facilitated processing for mediants over leading tones was observed even in the less-experienced listeners whereas the difference in the completion judgments (see pretest of Experiment 1) was supported only by the moderately-experienced listeners. This confirms the interest of implicit paradigms for music cognition research: following Schellenberg et al. (2005) and Tillmann et al. (2007), we argue that implicit paradigms allow showing processes that would not be observed otherwise.