Computational modeling of rhythmic expectations: Perspectives, pitfalls, and prospects.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Rhythmic structure is fundamental to human communication, influencing speech and music perception.
• Computational models attempt to explain human rhythmic sequence processing, but their interrelations and evaluation methods are unclear.
• Existing models include entrainment, probabilistic, and timekeeper approaches, each with distinct strengths and limitations.

Purpose of the Study:

• To review and critically assess three major classes of computational models for rhythmic expectations: entrainment, probabilistic, and timekeeper models.
• To identify how these models relate to each other and how they can be effectively evaluated.
• To propose a path forward for the field through model comparison and integration.

Main Methods:

• Comparative analysis of entrainment, probabilistic, and timekeeper models based on explanatory power, behavioral scope, learning/enculturation, and integration of features like pitch.
• Identification of challenges in comparing diverse models, including signal variability and differing research aims.
• Formulation of practical recommendations for advancing the field.

Main Results:

• Entrainment, probabilistic, and timekeeper models capture different aspects of rhythmic expectations.
• Substantial differences exist in what each model class can explain regarding rhythmic perception and production.
• Current model evaluation and comparison methods are insufficient for comprehensive understanding.

Conclusions:

• Model comparison and integration are crucial for advancing the study of rhythmic expectations.
• Standardizing input/output signals, employing diverse evaluation metrics, and integrating models are key recommendations.
• Openly sharing code and data will facilitate collaborative progress in understanding the cognitive and neural basis of rhythm.