Honing, H. (2005). Music Cognition: Theory Testing and Model Selection. Proceedings of the Cognitive Science Society Conference, Stresa: University of Turin.

Abstract

While the most common way of evaluating a computational model is by showing a good fit with the empirical data, recently the literature on theory testing and model selection criticizes the assumption that this is actually strong evidence for a model. The paper presents a case study from music cognition (modeling the ritardandi in music performance) and compares two families of computational models (kinematic and perceptual) using three different model selection criteria: goodness-of-fit, model's simplicity, and the amount of surprise in the predictions. While both models fit the empirical data equally well, in the light of what accounts as strong evidence for a model, i.e. making precise (constrained), non-smooth,and relatively surprising predictions, the perception-based model is preferred over the kinematic model, however simpler and natural the latter model might seem.

Full paper (pdf)