The influence of temporal regularities on the implicit learning of pitch structures

Abstract

Implicit learning is the acquisition of complex information without the intention to learn. The aim of this study was to investigate the influence of temporal regularities on the implicit learning of an artificial pitch grammar. According to the dynamic attending theory (DAT) external regularities can entrain internal oscillators that guide attention over time, inducing temporal expectations that influence perception of future events. In the present study, the presentation of the artificial pitch grammar in the exposure phase was temporally either regular or irregular for one of two participant groups. Based on the DAT, it was hypothesized that the regular temporal presentation would favour implicit learning of tone structures in comparison to the irregular temporal presentation. Results demonstrated learning of the artificial grammar for the group with the regular exposure phase and partial learning for the group with the irregular exposure phase. These findings suggest that the regular presentation helps listeners to develop perceptual expectations about the temporal occurrence of future tones and thus facilitates the learning of the artificial pitch grammar.

Keywords:

• Implicit learning
• Artificial grammar
• Tone sequences
• Rhythm
• Temporal expectations

We thank Simone Dalla Bella, Daniele Schön, and Emmanuel Bigand for discussions and Alexandra Corneyllie for help with analysing the sensory motor synchronization data. We also wish to thank anonymous reviewers for their attentive reading of the manuscript and their valuable comments.

The Auditory Cognition and Psychoacoustics Team is part of the LabEx CeLyA (“Centre Lyonnais d'Acoustique”, ANR-10-LABX-60).

This research was supported by a grant from EBRAMUS ITN (Europe BRAin and MUSic) [grant number 238157].

Notes

¹A foreperiod is a single time interval occurring between a warning signal and a target.

²Calculation of C-scores were as follows for each temporal pattern (of 10 or 12 tones) using potential clocks with periods of 220, 440, 660, 880, and 1100 ms. The best clock corresponded to the minimal C-score; this final score then reflects the degree of metrically of a given time pattern. Thus, for R-patterns, the most strongly induced clock for both 10-tone sequences and 12-tone sequences was always one with a period of 880 ms. This corresponds to a quadruple metre. However, for the IR patterns, no single clock emerged across the sequences: The period of the best clock for the different IR patterns varied from a period of 660 ms to a period of 880 ms. Povel's clock calculations suggest that R patterns may induce a mentally represented clock with a period of 880 ms. In contrast, the high C-scores for the IR patterns confirmed the intended irregular nature of the IR patterns, with various possible underlying clocks.

³Weighted DAT scores attribute a stronger accent to a tone that coincides with both 440- and 880-ms oscillation peaks than to a tone aligning with only the 440-ms cycle. Weighted scores are calculated with the following formula:DAT_weighted = [(2 × P) + (P′ – D)]⁄N_ev, (3)where P is a tone occurring in agreement with both periods of 440 and 880 ms. P′ is a tone aligned only with beats of the 440-ms periodic cycle; D is a deviant tone that coincides with neither 440- nor 880-ms oscillations peaks and thus disturbs the regularity of the temporal structure; N_ev is the total number of pattern tones.Results of weighted DAT score for these stimuli confirmed simple DAT scorings. For R patterns, the mean weighted DAT scores were 1.15 ± 0.08 for the 10-tone patterns and 1.38 ± 0.13 for the 12-tone patterns. For IR patterns, the mean weighted DAT scores were 0.46 ± 0.16 for the 10-tone patterns and 0.52 ± 0.15 for the 12-tone patterns. Again R patterns had higher weighted DAT scores indicating greater metric regularity than IR patterns [t(6) = 15.9, p < .001 for 10-tone patterns; t(6) = 19.72, p < 0.001 for 12-tone patterns (paired t-test].

⁴The Levene Test for equality of variance confirmed homogeneity of variance in both groups, all Fs < 1.70, all ps > .21.

⁵An additional analysis on log RTs (reaction times) confirmed the data pattern of this ANOVA, notably with its main effect of grammaticality [F(1, 30) = 8.79, MSE = 0.002, p =.006, ] and the three-way interaction between grammaticality, length, and group [F(1, 30) = 4.21, MSE = 0.001, p =.049, ].