Statistical learning and relative processing of tone sequences

Purpose: We investigated how the statistical learning of auditory sequences is reflected in neuromagnetic responses. Methods: Complex tones with fundamental (F0) frequencies in a five-tone equal temperament and seven different combinations of the first (F1) and second (F2) formant frequencies, as two distinct sets of five Japanese vowels, were generated by a formant synthesizer. The tones were ordered with the constraint that the probability of forthcoming tone was statistically defined (80% for one specific tone; 5% for the other four) by the latest two successive tones (second-order Markov chains) to form different types of tone sequences: F0-variable music-like tone sequences (F1, F2: constant) and F1, F2-variable language-like tone sequences. Each sequence consisted of 500 tones and successive 250 tones with a relative shift of fundamental and/or formant frequencies based on the same Markov transitional matrix. Neuromagnetic responses to music-like tone sequences in explicit and implicit learning conditions (experiment 1) and to music-like and language-like tone sequences in explicit learning conditions (experiment 2) were recorded from fourteen right-handed Japanese participants. Temporal profiles of the N1m responses to the tones with higher and lower transitional probabilities were compared. Results: The N1m responses to the tones with a higher transitional probability were significantly decreased compared with those to tones with a lower transitional probability in the latter half of the 500-tone sequence. Furthermore, this difference was retained even after fundamental and/or formant frequencies were relatively shifted (Figure 1, Figure 2). Discussion: The N1m responses may be useful as a marker for statistical learning, and such statistically acquired knowledge may also be utilized when spectral frequencies are relatively shifted