Search

SUMMARYIncreases in the scale and complexity of behavioral data pose an increasing challenge for data analysis. A common strategy involves replacing entire behaviors with small numbers of handpicked, domain-specific features, but this approach suffers from several crucial limitations. For example, handpicked features may miss important dimensions of variability, and correlations among them complicate statistical testing. Here, by contrast, we apply the variational autoencoder (VAE), an unsupervised learning method, to learn features directly from data and quantify the vocal behavior of two model species: the laboratory mouse and the zebra finch. The VAE converges on a parsimonious representation that outperforms handpicked features on a variety of common analysis tasks, enables the measurement of moment-by-moment vocal variability on the timescale of tens of milliseconds in the zebra finch, provides strong evidence that mouse ultrasonic vocalizations do not cluster as is commonly believed, and captures the similarity of tutor and pupil birdsong with qualitatively higher fidelity than previous approaches. In all, we demonstrate the utility of modern unsupervised learning approaches to the quantification of complex and high-dimensional vocal behavior.

While functional magnetic resonance imaging (fMRI) remains one of the most widespread and important methods in basic and clinical neuroscience, the data it produces—time series of brain volumes—continue to pose daunting analysis challenges. The current standard (“mass univariate”) approach involves constructing a matrix of task regressors, fitting a separate general linear model at each volume pixel (“voxel”), computing test statistics for each model, and correcting for false positives post hoc using bootstrap or other resampling methods. Despite its simplicity, this approach has enjoyed great success over the last two decades due to: 1) its ability to produce effect maps highlighting brain regions whose activity significantly correlates with a given variable of interest; and 2) its modeling of experimental effects as separable and thus easily interpretable. However, this approach suffers from several well-known drawbacks, namely: inaccurate assumptions of linearity and noise Gaussianity; a limited ability to capture individual effects and variability; and difficulties in performing proper statistical testing secondary to independently fitting voxels. In this work, we adopt a different approach, modeling entire volumes directly in a manner that increases model flexibility while preserving interpretability. Specifically, we use a generalized additive model (GAM) in which the effects of each regressor remain separable, the product of a spatial map produced by a variational autoencoder and a (potentially nonlinear) gain modeled by a covariate-specific Gaussian Process. The result is a model that yields group-level effect maps comparable or superior to the ones obtained with standard fMRI analysis software while also producing single-subject effect maps capturing individual differences. This suggests that generative models with a decomposable structure might offer a more flexible alternative for the analysis of task-based fMRI data.

Musical and athletic skills are learned and maintained through intensive practice to enable precise and reliable performance for an audience. Consequently, understanding such complex behaviours requires insight into how the brain functions during both practice and performance. Male zebra finches learn to produce courtship songs that are more varied when alone and more stereotyped in the presence of females1. These differences are thought to reflect song practice and performance, respectively2,3, providing a useful system in which to explore how neurons encode and regulate motor variability in these two states. Here we show that calcium signals in ensembles of spiny neurons (SNs) in the basal ganglia are highly variable relative to their cortical afferents during song practice. By contrast, SN calcium signals are strongly suppressed during female-directed performance, and optogenetically suppressing SNs during practice strongly reduces vocal variability. Unsupervised learning methods4,5 show that specific SN activity patterns map onto distinct song practice variants. Finally, we establish that noradrenergic signalling reduces vocal variability by directly suppressing SN activity. Thus, SN ensembles encode and drive vocal exploration during practice, and the noradrenergic suppression of SN activity promotes stereotyped and precise song performance for an audience. In male zebra finches, song practice and courtship song performance are associated with distinct patterns of neural activity in the basal ganglia, resulting in reduced vocal variability during performance.

Animals vocalize only in certain behavioral contexts, but the circuits and synapses through which forebrain neurons trigger or suppress vocalization remain unknown. Here we used transsynaptic tracing to identify two populations of inhibitory neurons that lie upstream of neurons in the periaqueductal gray that gate the production of ultrasonic vocalizations in mice (i.e., PAG-USV neurons). Activating PAG-projecting neurons in the preoptic hypothalamus (POAPAG neurons) elicited USV production in the absence of social cues. In contrast, activating PAG-projecting neurons in the central-medial boundary zone of the amygdala (AmgC/M-PAG neurons) transiently suppressed USV production without disrupting non-vocal social behavior. Optogenetics-assisted circuit mapping in brain slices revealed that POAPAG neurons directly inhibit PAG interneurons, which in turn inhibit PAG-USV neurons, whereas AmgC/M-PAG neurons directly inhibit PAG-USV neurons. These experiments identify two major forebrain inputs to the PAG that trigger and suppress vocalization, respectively, while also establishing the synaptic mechanisms through which these neurons exert opposing behavioral effects.

Bowling et al. (1) report musical consonance judgements of various musical chords and a model designed to assess their “vocal similarity,” claiming the fit between model and data provides evidence that the perceived consonance of tone combinations is explained in part by their “relative similarity to voiced speech sounds.” I believe, however, this claim is unjustified. The Bowling et al. (1) report states that consonance judgements are tested for “every possible chromatic dyad, triad, and tetrad within a single octave,” yet because a just intonation tuning system is employed rather than the standard equal-temperament system, the actual number of possible chords is more … [↵][1]1Email: jackgoffinet{at}gmail.com. [1]: #xref-corresp-1-1