A Cognitive Information Theory of Music: A Computational Memetics Approach

This thesis offers an account of music cognition based on information theory and memetics. My research strategy is to split the memetic modelling into four layers: Data, Information, Psychology and Application. Multiple cognitive models are proposed for the Information and Psychology layers, and the MDL best-fit models with published human data are selected. Then, for the Psychology layer only, new experiments are conducted to validate the best-fit models.

In the information chapter, an information-theoretic model of musical memory is proposed, along with two competing models. The proposed model exhibited a better fit with human data than the competing models. Higher-level psychological theories are then built on top of this information layer. In the similarity chapter, I proposed three competing models of musical similarity, and conducted a new experiment to validate the best-fit model. In the fitness chapter, I again proposed three competing models of musical fitness, and conducted a new experiment to validate the best-fit model. In both cases, the correlations with human data are statistically significant.

All in all, my research has shown that the memetic strategy is sound, and the modelling results are encouraging. Implications of this research are discussed.