Neural-vocal phase coupling reveals structured timing in birdsong production

Abstract

Understanding how neural population activity is temporally coordinated with behavior remains a central challenge in neuroscience. Songbirds provide a powerful model system for addressing this question because learned vocal production requires precise coordination among neural dynamics, temporally structured motor output, and auditory feedback. However, quantifying neural-vocal interactions is challenging because both neural and acoustic signals are rhythmic, noisy, and highly nonstationary. Here, we investigate neural-vocal coordination during spontaneous canary singing using simultaneous recordings of neural population activity in a forebrain region of the song system and vocal behavior. Using a phase-resolved cross-correlation framework combined with surrogate-based statistical validation, we quantify neural-vocal interactions in short and highly variable song segments. Our analysis reveals that neural-vocal interactions are organized into distinct temporal regimes comprising positive, near-zero, and negative lags, consistent with neural activity preceding, accompanying, or following vocal output. The coexistence of these regimes is consistent with the integrative role of the recorded region, which receives auditory input, contributes to premotor control, and participates in the neural circuitry supporting song learning and the ongoing maintenance of adult song. We further find that correlated and anticorrelated interactions coexist throughout singing, with anticorrelated interactions consistently concentrated around near-zero lags. These anticorrelations identify periods in which decreases in neural population activity are closely aligned with sound production, revealing biologically relevant information that is obscured by analyses performed over complete song renditions. Together, these results uncover a robust temporal structure linking neural population activity to vocal behavior and provide a broadly applicable framework for extracting transient neural-behavioral interactions from complex biological signals. Author summaryModern neuroscience can simultaneously record the activity of large neural populations, yet extracting meaningful relationships between neural activity and behavior remains challenging because natural behaviors are highly variable. Songbirds provide a unique opportunity to study this problem because their learned vocalizations share key features with human speech while remaining experimentally accessible. Here, we analyzed simultaneous recordings of neural population activity and vocal behavior in freely singing canaries. Instead of averaging neural activity across entire songs, we examined brief time windows and combined local correlation analysis with statistical tests to identify reliable interactions between the brain and behavior. We found that these interactions switch among several preferred timing patterns: neural activity can precede vocal output, occur at nearly the same time, or follow acoustic events. This diversity is consistent with the combined roles of the recorded brain region. Our work provides an intuitive and broadly applicable framework for uncovering transient neural-behavioral interactions that remain hidden by traditional time-averaged approaches. Because it requires only simultaneous recordings of two time series and makes minimal assumptions about their dynamics, the framework can be applied to many biological systems involving complex temporal signals.