The Anatomy of Consonance/Dissonance: Evaluating Acoustic and Cultural Predictors Across Multiple Datasets with Chords
Acoustic and musical components of consonance and dissonance perception have been recently identified. This study expands the range of predictors of consonance and dissonance by three analytical operations: In the first stage, we identify the underlying structure of acoustic and musical predictors within a large set of potential variables using an extensive dataset of chords. In Experiment 1, we evaluate the current model of consonance-dissonance by Harrison and Pearce (2020) based on an generalised linear mixed model analysis a subset of three previously published datasets. This operates also allows us to optimise the predictors in the model in several ways. We bring an additional category, sharpness to complement roughness, harmonicity, and familiarity, but we also propose and assess a number of new predictors representing harmonicity and familiarity that are superior to the past formulations of the models representing these categories. In Experiment 2, the current and the optimised model are evaluated with the aid of nine full datasets that provide empirical mean ratings of consonance and dissonance for a range of intervals and chords. The results within datasets show good prediction rates for the present model (R squared=0.64) and significant improvement for the optimal model (R squared=0.77). Similar patter of differences holds for the analysis across all datasets and there is a significant improvement in the predictive rate when the model has been optimised based on the Experiment 1 analysis. In particular, the new elements, tonal dissonance, familiarity as coded by a corpus-driven predictor, and sharpness are substantial additions to account for the dissonance ratings. The discussion draws attention to the role of harmonicity, which in this analysis is captured by a predictor reflecting the knowledge of Western idiom.