Pitches that Wire Together Fire Together: Scale Degree Associations Across Time Predict Melodic Expectations

Area of Science:

• Cognitive Science
• Music Cognition
• Computational Musicology

Background:

• Music listening relies on generating expectations.
• Previous research focused on transition probabilities and n-grams for statistical learning in melodies.
• Other statistical learning mechanisms, like temporal associations, may be underexplored.

Purpose of the Study:

• To investigate the role of temporal associations between scale degrees in melodic expectation.
• To compare the predictive power of expectation networks against established models (Markov models, music theory models).
• To assess how generalized scale degree associations influence melodic continuation predictions.

Main Methods:

• Utilized expectation networks to model temporal associations between scale degrees.
• Combined learned associations to predict melodic continuations.
• Tested predictions against listener responses in a musical cloze task, comparing with IDyOM and Temperley's model.

Main Results:

• Expectation networks, informed by adjacent and nonadjacent note relationships, significantly predicted melodic expectations.
• These networks explained unique variance in listener predictions beyond n-gram models.
• Coefficient estimates were highest for expectation networks, suggesting stronger predictive power.

Conclusions:

• Generalized scale degree associations are crucial for listeners' melodic predictions.
• Expectation networks offer a valuable computational approach to understanding musical expectation.
• Broader statistical learning processes beyond simple sequential patterns are vital for modeling music cognition.