Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.04.29.068163v1?rss=1 Authors: Loui, P. Abstract: Knowledge of language and music depends upon the ability to perceive relationships between sounds in order to form a stable mental representation of grammatical structure. Although abundant evidence exists for the learning of statistical structure (specifically event probability structure) from units of sounds, little research has explored what constitutes a distinct unit of sound. Here we show that spectral content is an important cue within sound units for acquiring a stable mental representation of sound structure. Tone sequences in a novel scale were constructed and presented to participants on three different timbres with spectral cues that were either congruent with the scale structure, incongruent with the scale structure, or did not contain spectral cues (neutral). Participants completed probe tone ratings before and after a half-hour period of exposure to melodies in the artificial grammar, using timbres that were either congruent, incongruent, or neutral. Learning outcome was assessed by correlating pre/post-exposure ratings and the statistical structure of tones within the exposure period. Participants did best at learning with congruent timbres, and worst with incongruent timbres. Results show that spectral content is a useful cue for statistical learning, and suggest that the frequency structure of sounds might be acquired through exposure to spectral distribution in natural sounds. Copy rights belong to original authors. Visit the link for more info