Towards a Cognitive Model of Melodic Similarity

In recent years the interest in melodic similarity has mushroomed mainly due to the increased importance of music information retrieval (MIR). A great number of similarity models and algorithms have been developed, but little or no attention has been paid to cognitive or perceptual aspects to the issue at hand. Questions, about the relevant parameters and the appropriate implementation are under-researched as are experimental data. This paper focuses on the pitch aspect of melodic similarity, scrutinising the term pitch replacing it by a less ambivalent and overused term, which we will call meloton. Based on the term meloton the paper suggests to approach the issue of ‘melotonic’ similarity from a transformational angle, where transformations are executed as reflections and translations. ‘Melotonic’ similarity then is seen as related to the transformation process in form of a transpositional and interval vector. Finally, melotonic similarity as portrait in a psychological context emerges as a multi-facet phenomenon requiring the development of flexible models.