Search

Vocal communication in both songbirds and humans relies on categorical perception of smoothly varying acoustic spaces. Vocal perception can be biased by expectation and context, but the mechanisms of this bias are not well understood. We developed a behavioral task in which songbirds, European starlings, are trained to to classify smoothly varying song syllables in the context of predictive syllable sequences. We find that syllable-sequence predictability biases perceptual categorization following a Bayesian model of probabilistic information integration. We then recorded from populations of neurons in the auditory forebrain while birds actively categorized song syllables, observing large proportions of neurons that track the smoothly varying natural feature space of syllable categories. We observe that predictive information in the syllable sequences dynamically modulates sensory neural representations. These results support a Bayesian model of perception where predictive information acts to dynamically reallocate sensory neural resources, sharpening acuity (i.e. the likelihood) in high-probability regions of stimulus space.One-Sentence SummaryPredictive information in vocal sequences biases Bayesian categorical perception through rapid sensory reorganization.Graphical Abstract

To convey meaning, human language relies on hierarchically organized, long-range relationships spanning words, phrases, sentences, and discourse. The strength of the relationships between sequentially ordered elements of language (e.g., phonemes, characters, words) decays following a power law as a function of sequential distance. To understand the origins of these relationships, we examined long-range statistical structure in the speech of human children at multiple developmental time points, along with non-linguistic behaviors in humans and phylogenetically distant species. Here we show that adult-like power-law statistical dependencies precede the production of hierarchically-organized linguistic structures, and thus cannot be driven solely by these structures. Moreover, we show that similar long-range relationships occur in diverse non-linguistic behaviors across species. We propose that the hierarchical organization of human language evolved to exploit pre-existing long-range structure present in much larger classes of non-linguistic behavior, and that the cognitive capacity to model long-range hierarchical relationships preceded language evolution. We call this the Statistical Scaffolding Hypothesis for language evolution.1Significance StatementHuman language is uniquely characterized by semantically meaningful hierarchical organization, conveying information over long timescales. At the same time, many non-linguistic human and animal behaviors are also often characterized by richly hierarchical organization. Here, we compare the long-timescale statistical dependencies present in language to those present in non-linguistic human and animal behaviors as well as language production throughout childhood. We find adult-like, long-timescale relationships early in language development, before syntax or complex semantics emerge, and we find similar relationships in non-linguistic behaviors like cooking and even housefly movement. These parallels demonstrate that long-range statistical dependencies are not unique to language and suggest a possible evolutionary substrate for the long-range hierarchical structure present in human language.

The National Institute of Health (NIH) sets postdoctoral (postdoc) trainee stipend levels that many American institutions and investigators use as a basis for postdoc salaries. Although salary standards are held constant across universities, the cost of living in those universities' cities and towns vary widely. Across non-postdoc jobs, more expensive cities pay workers higher wages that scale with an increased cost of living. This work investigates the extent to which postdoc wages account for cost-of-living differences. More than 27,000 postdoc salaries across all US universities are analyzed alongside measures of regional differences in cost of living. We find that postdoc salaries do not account for cost-of-living differences, in contrast with the broader labor market in the same cities and towns. Despite a modest increase in income in high cost of living areas, real (cost of living adjusted) postdoc salaries differ by 29% ($15k 2021 USD) between the least and most expensive areas. Cities that produce greater numbers of tenure-track faculty relative to students such as Boston, New York, and San Francisco are among the most impacted by this pay disparity. The postdoc pay gap is growing and is well-positioned to incur a greater financial burden on economically disadvantaged groups and contribute to faculty hiring disparities in women and racial minorities.

1. Background: The manual detection, analysis and classification of animal vocalizations in acoustic recordings is laborious and requires expert knowledge. Hence, there is a need for objective, generalizable methods that detect underlying patterns in these data, categorize sounds into distinct groups and quantify similarities between them. Among all computational methods that have been proposed to accomplish this, neighbourhood-based dimensionality reduction of spectrograms to produce a latent space representation of calls stands out for its conceptual simplicity and effectiveness.2. Goal of the study/what was done: Using a dataset of manually annotated meerkat Suricata suricatta vocalizations, we demonstrate how this method can be used to obtain meaningful latent space representations that reflect the established taxonomy of call types. We analyse strengths and weaknesses of the proposed approach, give recommendations for its usage and show application examples, such as the classification of ambiguous calls and the detection of mislabelled calls.3. What this means: All analyses are accompanied by example code to help researchers realize the potential of this method for the study of animal vocalizations. published

This review systematizes the emerging literature for causal inference using deep neural networks under the potential outcomes framework. It provides an intuitive introduction on how deep learning can be used to estimate/predict heterogeneous treatment effects and extend causal inference to settings where confounding is non-linear, time varying, or encoded in text, networks, and images. To maximize accessibility, we also introduce prerequisite concepts from causal inference and deep learning. The survey differs from other treatments of deep learning and causal inference in its sharp focus on observational causal estimation, its extended exposition of key algorithms, and its detailed tutorials for implementing, training, and selecting among deep estimators in Tensorflow 2 available at github.com/kochbj/Deep-Learning-for-Causal-Inference.

UMAP is a nonparametric graph-based dimensionality reduction algorithm using applied Riemannian geometry and algebraic topology to find low-dimensional embeddings of structured data. The UMAP algorithm consists of two steps: (1) computing a graphical representation of a data set (fuzzy simplicial complex) and (2) through stochastic gradient descent, optimizing a low-dimensional embedding of the graph. Here, we extend the second step of UMAP to a parametric optimization over neural network weights, learning a parametric relationship between data and embedding. We first demonstrate that parametric UMAP performs comparably to its nonparametric counterpart while conferring the benefit of a learned parametric mapping (e.g., fast online embeddings for new data). We then explore UMAP as a regularization, constraining the latent distribution of autoencoders, parametrically varying global structure preservation, and improving classifier accuracy for semisupervised learning by capturing structure in unlabeled data.

Animals produce vocalizations that range in complexity from a single repeated call to hundreds of unique vocal elements patterned in sequences unfolding over hours. Characterizing complex vocalizations can require considerable effort and a deep intuition about each species’ vocal behavior. Even with a great deal of experience, human characterizations of animal communication can be affected by human perceptual biases. We present a set of computational methods for projecting animal vocalizations into low dimensional latent representational spaces that are directly learned from the spectrograms of vocal signals. We apply these methods to diverse datasets from over 20 species, including humans, bats, songbirds, mice, cetaceans, and nonhuman primates. Latent projections uncover complex features of data in visually intuitive and quantifiable ways, enabling high-powered comparative analyses of vocal acoustics. We introduce methods for analyzing vocalizations as both discrete sequences and as continuous latent variables. Each method can be used to disentangle complex spectro-temporal structure and observe long-timescale organization in communication., Author summary Of the thousands of species that communicate vocally, the repertoires of only a tiny minority have been characterized or studied in detail. This is due, in large part, to traditional analysis methods that require a high level of expertise that is hard to develop and often species-specific. Here, we present a set of unsupervised methods to project animal vocalizations into latent feature spaces to quantitatively compare and develop visual intuitions about animal vocalizations. We demonstrate these methods across a series of analyses over 19 datasets of animal vocalizations from 29 different species, including songbirds, mice, monkeys, humans, and whales. We show how learned latent feature spaces untangle complex spectro-temporal structure, enable cross-species comparisons, and uncover high-level attributes of vocalizations such as stereotypy in vocal element clusters, population regiolects, coarticulation, and individual identity.

Human speech possesses a rich hierarchical structure that allows for meaning to be altered by words spaced far apart in time. Conversely, the sequential structure of nonhuman communication is thought to follow non-hierarchical Markovian dynamics operating over only short distances. Here, we show that human speech and birdsong share a similar sequential structure indicative of both hierarchical and Markovian organization. We analyze the sequential dynamics of song from multiple songbird species and speech from multiple languages by modeling the information content of signals as a function of the sequential distance between vocal elements. Across short sequence-distances, an exponential decay dominates the information in speech and birdsong, consistent with underlying Markovian processes. At longer sequence-distances, the decay in information follows a power law, consistent with underlying hierarchical processes. Thus, the sequential organization of acoustic elements in two learned vocal communication signals (speech and birdsong) shows functionally equivalent dynamics, governed by similar processes., By examining the organization of bird song and human speech, the authors show that the two types of communication signals have similar sequential structures, following both hierarchical and Markovian organization.

There is a rich tradition of building computational models in cognitive science, but modeling, theoretical, and experimental research are not as tightly integrated as they could be. In this paper, we show that computational techniques—even simple ones that are straightforward to use—can greatly facilitate designing, implementing, and analyzing experiments, and generally help lift research to a new level. We focus on the domain of artificial grammar learning, and we give five concrete examples in this domain for (a) formalizing and clarifying theories, (b) generating stimuli, (c) visualization, (d) model selection, and (e) exploring the hypothesis space., Zuidema et al. illustrate how empirical AGL studies can benefit from computational models and techniques. Computational models can help clarifying theories, and thus in delineating research questions, but also in facilitating experimental design, stimulus generation, and data analysis. The authors show, with a series of examples, how computational modeling can be integrated with empirical AGL approaches, and how model selection techniques can indicate the most likely model to explain experimental outcomes.