Machine learning and statistical classification of birdsong link vocal acoustic features with phylogeny.

Birdsong is a longstanding model system for studying evolution and biodiversity. Here, we collected and analyzed high quality song recordings from seven species in the family Estrildidae. We measured the acoustic features of syllables and then used dimensionality reduction and machine learning classifiers to identify features that accurately assigned syllables to species. Species differences were captured by the first 3 principal components, corresponding to basic frequency, power distribution, and spectrotemporal features. We then identified the measured features underlying classification accuracy. We found that fundamental frequency, mean frequency, spectral flatness, and syllable duration were the most informative features for species identification. Next, we tested whether specific acoustic features of species' songs predicted phylogenetic distance. We found significant phylogenetic signal in syllable frequency features, but not in power distribution or spectrotemporal features. Results suggest that frequency features are more constrained by species' genetics than are other features, and are the best signal features for identifying species from song recordings. The absence of phylogenetic signal in power distribution and spectrotemporal features suggests that these song features are labile, reflecting learning processes and individual recognition.

Auditory sensitivity and vocal acoustics in five species of estrildid songbirds.

Comparative studies of acoustic communication in clades with diverse signal features provide a powerful framework for testing relationships between perception and behaviour. We measured auditory sensitivity in five species of estrildid songbirds with acoustically distinct songs and tested whether differences aligned with species differences in song frequency content. Species were chosen based on phylogeny and differences in song acoustics. Behavioural audiograms were obtained using operant training and testing. Adult audiograms were compared across species and between sexes within a species. Juvenile and adult audiograms were compared in one species. The audiograms of adults reared by their own species and those reared and tutored by another species were compared in one species. Results showed that audiograms were similar across species and similar to previous reports of songbird auditory sensitivity. Species differed in the highest frequency detected and the frequency of peak sensitivity. While hearing frequency range was not correlated with song frequency bandwidth, the frequency of peak sensitivity was highly corelated with the frequency of peak energy in song. Sensitivity did not differ based on sex, age or tutoring experience. Our findings suggest that adaptations in songbird auditory sensitivity are largely constrained by shared peripheral and central encoding mechanisms, with species-specific perception appearing only at peak sensitivity.

Auditory Selectivity for Spectral Contrast in Cortical Neurons and Behavior.

Vocal communication relies on the ability of listeners to identify, process, and respond to vocal sounds produced by others in complex environments. To accurately recognize these signals, animals' auditory systems must robustly represent acoustic features that distinguish vocal sounds from other environmental sounds. Vocalizations typically have spectral structure; power regularly fluctuates along the frequency axis, creating spectral contrast. Spectral contrast is closely related to harmonicity, which refers to spectral power peaks occurring at integer multiples of a fundamental frequency. Although both spectral contrast and harmonicity typify natural sounds, they may differ in salience for communication behavior and engage distinct neural mechanisms. Therefore, it is important to understand which of these properties of vocal sounds underlie the neural processing and perception of vocalizations.Here, we test the importance of vocalization-typical spectral features in behavioral recognition and neural processing of vocal sounds, using male zebra finches. We show that behavioral responses to natural and synthesized vocalizations rely on the presence of discrete frequency components, but not on harmonic ratios between frequencies. We identify a specific population of neurons in primary auditory cortex that are sensitive to the spectral resolution of vocal sounds. We find that behavioral and neural response selectivity is explained by sensitivity to spectral contrast rather than harmonicity. This selectivity emerges within the cortex; it is absent in the thalamorecipient region and present in the deep output region. Further, deep-region neurons that are contrast-sensitive show distinct temporal responses and selectivity for modulation density compared with unselective neurons.SIGNIFICANCE STATEMENT Auditory coding and perception are critical for vocal communication. Auditory neurons must encode acoustic features that distinguish vocalizations from other sounds in the environment and generate percepts that direct behavior. The acoustic features that drive neural and behavioral selectivity for vocal sounds are unknown, however. Here, we show that vocal response behavior scales with stimulus spectral contrast but not with harmonicity, in songbirds. We identify a distinct population of auditory cortex neurons in which response selectivity parallels behavioral selectivity. This neural response selectivity is explained by sensitivity to spectral contrast rather than to harmonicity. Our findings inform the understanding of how the auditory system encodes socially-relevant signals via detection of an acoustic feature that is ubiquitous in vocalizations.

Emergent tuning for learned vocalizations in auditory cortex.

Vocal learners use early social experience to develop auditory skills specialized for communication. However, it is unknown where in the auditory pathway neural responses become selective for vocalizations or how the underlying encoding mechanisms change with experience. We used a vocal tutoring manipulation in two species of songbird to reveal that tuning for conspecific song arises within the primary auditory cortical circuit. Neurons in the deep region of primary auditory cortex responded more to conspecific songs than to other species' songs and more to species-typical spectrotemporal modulations, but neurons in the intermediate (thalamorecipient) region did not. Moreover, birds that learned song from another species exhibited parallel shifts in selectivity and tuning toward the tutor species' songs in the deep but not the intermediate region. Our results locate a region in the auditory processing hierarchy where an experience-dependent coding mechanism aligns auditory responses with the output of a learned vocal motor behavior.

Sex differences and endocrine regulation of auditory-evoked, neural responses in African clawed frogs (Xenopus).

Mating depends on the accurate detection of signals that convey species identity and reproductive state. In African clawed frogs, Xenopus, this information is conveyed by vocal signals that differ in temporal patterns and spectral features between sexes and across species. We characterized spectral sensitivity using auditory-evoked potentials (AEPs), commonly known as the auditory brainstem response, in males and females of four Xenopus species. In female X. amieti, X. petersii, and X. laevis, peripheral auditory sensitivity to their species own dyad-two, species-specific dominant frequencies in the male advertisement call-is enhanced relative to males. Males were most sensitive to lower frequencies including those in the male-directed release calls. Frequency sensitivity was influenced by endocrine state; ovariectomized females had male-like auditory tuning while dihydrotestosterone-treated, ovariectomized females maintained female-like tuning. Thus, adult, female Xenopus demonstrate an endocrine-dependent sensitivity to the spectral features of conspecific male advertisement calls that could facilitate mating. Xenopus AEPs resemble those of other species in stimulus and level dependence, and in sensitivity to anesthetic (MS222). AEPs were correlated with body size and sex within some species. A frequency following response, probably encoded by the amphibian papilla, might facilitate dyad source localization via interaural time differences.

Coding principles of the canonical cortical microcircuit in the avian brain.

Mammalian neocortex is characterized by a layered architecture and a common or "canonical" microcircuit governing information flow among layers. This microcircuit is thought to underlie the computations required for complex behavior. Despite the absence of a six-layered cortex, birds are capable of complex cognition and behavior. In addition, the avian auditory pallium is composed of adjacent information-processing regions with genetically identified neuron types and projections among regions comparable with those found in the neocortex. Here, we show that the avian auditory pallium exhibits the same information-processing principles that define the canonical cortical microcircuit, long thought to have evolved only in mammals. These results suggest that the canonical cortical microcircuit evolved in a common ancestor of mammals and birds and provide a physiological explanation for the evolution of neural processes that give rise to complex behavior in the absence of cortical lamination.

Sparse and background-invariant coding of vocalizations in auditory scenes.

Vocal communicators such as humans and songbirds readily recognize individual vocalizations, even in distracting auditory environments. This perceptual ability is likely subserved by auditory neurons whose spiking responses to individual vocalizations are minimally affected by background sounds. However, auditory neurons that produce background-invariant responses to vocalizations in auditory scenes have not been found. Here, we describe a population of neurons in the zebra finch auditory cortex that represent vocalizations with a sparse code and that maintain their vocalization-like firing patterns in levels of background sound that permit behavioral recognition. These same neurons decrease or stop spiking in levels of background sound that preclude behavioral recognition. In contrast, upstream neurons represent vocalizations with dense and background-corrupted responses. We provide experimental evidence suggesting that sparse coding is mediated by feedforward suppression. Finally, we show through simulations that feedforward inhibition can transform a dense representation of vocalizations into a sparse and background-invariant representation.

Conserved mechanisms of vocalization coding in mammalian and songbird auditory midbrain.

The ubiquity of social vocalizations among animals provides the opportunity to identify conserved mechanisms of auditory processing that subserve communication. Identifying auditory coding properties that are shared across vocal communicators will provide insight into how human auditory processing leads to speech perception. Here, we compare auditory response properties and neural coding of social vocalizations in auditory midbrain neurons of mammalian and avian vocal communicators. The auditory midbrain is a nexus of auditory processing because it receives and integrates information from multiple parallel pathways and provides the ascending auditory input to the thalamus. The auditory midbrain is also the first region in the ascending auditory system where neurons show complex tuning properties that are correlated with the acoustics of social vocalizations. Single unit studies in mice, bats and zebra finches reveal shared principles of auditory coding including tonotopy, excitatory and inhibitory interactions that shape responses to vocal signals, nonlinear response properties that are important for auditory coding of social vocalizations and modulation tuning. Additionally, single neuron responses in the mouse and songbird midbrain are reliable, selective for specific syllables, and rely on spike timing for neural discrimination of distinct vocalizations. We propose that future research on auditory coding of vocalizations in mouse and songbird midbrain neurons adopt similar experimental and analytical approaches so that conserved principles of vocalization coding may be distinguished from those that are specialized for each species. This article is part of a Special Issue entitled "Communication Sounds and the Brain: New Directions and Perspectives".

Experience dependence of neural responses to different classes of male songs in the primary auditory forebrain of female songbirds.

There is both extensive species-specificity and critical experience-dependence in the recognition of own species songs in many songbird species. For example, female zebra finches Taeniopygia guttata raised by their parents show behavioral preferences for the songs of the father over unfamiliar conspecific males and for unfamiliar songs of conspecifics over heterospecifics. Behavioral discrimination between different species' songs is also displayed by females raised without exposure to any male songs but it is diminished in females raised by heterospecific foster parents. We tested whether neural responses in the female auditory forebrain paralleled each of these known behavioral patterns in song-class discrimination. We analyzed spike rates, above background levels, recorded from single units in the L2a subregion of the field L complex of female zebra finches. In subjects raised by genetic parents, spike rates were similar to songs of fathers and unfamiliar male zebra finches, and higher to unfamiliar conspecific over unfamiliar heterospecific songs. In females raised in isolation from male songs, we also found higher spike rates to unfamiliar conspecific over heterospecific songs. In females raised by heterospecific foster parents, spike rates were similar in response to songs of the foster father and unfamiliar males of the foster species, similar between unfamiliar songs of conspecifics and the heterospecific foster species, and higher to unfamiliar songs of the foster species over a third finch species. Thus, in parallel to the experience-dependence of females' behaviors in response to different male song classes, differences in social experiences can also alter neural response patterns to male song classes in the auditory forebrain of female zebra finches.

Early experience shapes vocal neural coding and perception in songbirds.

Songbirds, like humans, are highly accomplished vocal learners. The many parallels between speech and birdsong and conserved features of mammalian and avian auditory systems have led to the emergence of the songbird as a model system for studying the perceptual mechanisms of vocal communication. Laboratory research on songbirds allows the careful control of early life experience and high-resolution analysis of brain function during vocal learning, production, and perception. Here, I review what songbird studies have revealed about the role of early experience in the development of vocal behavior, auditory perception, and the processing of learned vocalizations by auditory neurons. The findings of these studies suggest general principles for how exposure to vocalizations during development and into adulthood influences the perception of learned vocal signals.

A behavioral framework to guide research on central auditory development and plasticity.

The auditory CNS is influenced profoundly by sounds heard during development. Auditory deprivation and augmented sound exposure can each perturb the maturation of neural computations as well as their underlying synaptic properties. However, we have learned little about the emergence of perceptual skills in these same model systems, and especially how perception is influenced by early acoustic experience. Here, we argue that developmental studies must take greater advantage of behavioral benchmarks. We discuss quantitative measures of perceptual development and suggest how they can play a much larger role in guiding experimental design. Most importantly, including behavioral measures will allow us to establish empirical connections among environment, neural development, and perception.

Extra-classical tuning predicts stimulus-dependent receptive fields in auditory neurons.

The receptive fields of many sensory neurons are sensitive to statistical differences among classes of complex stimuli. For example, excitatory spectral bandwidths of midbrain auditory neurons and the spatial extent of cortical visual neurons differ during the processing of natural stimuli compared to the processing of artificial stimuli. Experimentally characterizing neuronal nonlinearities that contribute to stimulus-dependent receptive fields is important for understanding how neurons respond to different stimulus classes in multiple sensory modalities. Here we show that in the zebra finch, many auditory midbrain neurons have extra-classical receptive fields, consisting of sideband excitation and sideband inhibition. We also show that the presence, degree, and asymmetry of stimulus-dependent receptive fields during the processing of complex sounds are predicted by the presence, valence, and asymmetry of extra-classical tuning. Neurons for which excitatory bandwidth expands during the processing of song have extra-classical excitation. Neurons for which frequency tuning is static and for which excitatory bandwidth contracts during the processing of song have extra-classical inhibition. Simulation experiments further demonstrate that stimulus-dependent receptive fields can arise from extra-classical tuning with a static spike threshold nonlinearity. These findings demonstrate that a common neuronal nonlinearity can account for the stimulus dependence of receptive fields estimated from the responses of auditory neurons to stimuli with natural and non-natural statistics.

Automated auditory recognition training and testing.

Laboratory training and testing of auditory recognition skills in animals is important for understanding animal communication systems that depend on auditory cues. Songbirds are commonly studied because of their exceptional ability to learn complex vocalizations. In recent years, mounting interest in the perceptual abilities of songbirds has increased the demand for laboratory behavioural training and testing paradigms. Here, we describe and demonstrate the success of a method for auditory discrimination experiments, including all the necessary hardware, training procedures and freely-available, versatile software. The system can run several behavioural training and testing paradigms, including operant (go-nogo, stimulus preference, and two-alternative forced choice) and classical conditioning tasks. The software and some hardware components can be used with any laboratory animal that learns and responds to sensory cues. The peripheral hardware and training procedures are designed for use with songbirds and auditory stimuli. Using the go-nogo paradigm of the training system, we show that adult zebra finches learn to recognize and correctly classify individual female calls and male songs. We also show that learning the task generalizes to new stimulus classes; birds that learned the task with calls subsequently learned to recognize songs faster than did birds that learned the task and songs at the same time.

Anesthetic state modulates excitability but not spectral tuning or neural discrimination in single auditory midbrain neurons.

The majority of sensory physiology experiments have used anesthesia to facilitate the recording of neural activity. Current techniques allow researchers to study sensory function in the context of varying behavioral states. To reconcile results across multiple behavioral and anesthetic states, it is important to consider how and to what extent anesthesia plays a role in shaping neural response properties. The role of anesthesia has been the subject of much debate, but the extent to which sensory coding properties are altered by anesthesia has yet to be fully defined. In this study we asked how urethane, an anesthetic commonly used for avian and mammalian sensory physiology, affects the coding of complex communication vocalizations (songs) and simple artificial stimuli in the songbird auditory midbrain. We measured spontaneous and song-driven spike rates, spectrotemporal receptive fields, and neural discriminability from responses to songs in single auditory midbrain neurons. In the same neurons, we recorded responses to pure tone stimuli ranging in frequency and intensity. Finally, we assessed the effect of urethane on population-level representations of birdsong. Results showed that intrinsic neural excitability is significantly depressed by urethane but that spectral tuning, single neuron discriminability, and population representations of song do not differ significantly between unanesthetized and anesthetized animals.

Coevolution in communication senders and receivers: vocal behavior and auditory processing in multiple songbird species.

Communication is a strong selective pressure on brain evolution because the exchange of information between individuals is crucial for fitness-related behaviors, such as mating. Given the importance of communication, the brains of signal senders and receivers are likely to be functionally coordinated. We study vocal behavior and auditory processing in multiple species of estrildid finches with the goal of understanding how species identity and early experience interact to shape the neural systems that subserve communication. Male finches learn to produce species-specific songs, and both sexes learn to recognize songs. Our studies indicate that closely related species exhibit different auditory coding properties in the midbrain and forebrain and that early life experience of vocalizations contributes to these differences. Moreover, birds that naturally sing tonal songs can learn broadband songs from heterospecific tutors, providing an opportunity to examine the interplay between species identity and early experience in the development of vocal behavior and auditory tuning.

Incorporating naturalistic correlation structure improves spectrogram reconstruction from neuronal activity in the songbird auditory midbrain.

Birdsong is comprised of rich spectral and temporal organization, which might be used for vocal perception. To quantify how this structure could be used, we have reconstructed birdsong spectrograms by combining the spike trains of zebra finch auditory midbrain neurons with information about the correlations present in song. We calculated maximum a posteriori estimates of song spectrograms using a generalized linear model of neuronal responses and a series of prior distributions, each carrying different amounts of statistical information about zebra finch song. We found that spike trains from a population of mesencephalicus lateral dorsalis (MLd) neurons combined with an uncorrelated Gaussian prior can estimate the amplitude envelope of song spectrograms. The same set of responses can be combined with Gaussian priors that have correlations matched to those found across multiple zebra finch songs to yield song spectrograms similar to those presented to the animal. The fidelity of spectrogram reconstructions from MLd responses relies more heavily on prior knowledge of spectral correlations than temporal correlations. However, the best reconstructions combine MLd responses with both spectral and temporal correlations.

A generalized linear model for estimating spectrotemporal receptive fields from responses to natural sounds.

In the auditory system, the stimulus-response properties of single neurons are often described in terms of the spectrotemporal receptive field (STRF), a linear kernel relating the spectrogram of the sound stimulus to the instantaneous firing rate of the neuron. Several algorithms have been used to estimate STRFs from responses to natural stimuli; these algorithms differ in their functional models, cost functions, and regularization methods. Here, we characterize the stimulus-response function of auditory neurons using a generalized linear model (GLM). In this model, each cell's input is described by: 1) a stimulus filter (STRF); and 2) a post-spike filter, which captures dependencies on the neuron's spiking history. The output of the model is given by a series of spike trains rather than instantaneous firing rate, allowing the prediction of spike train responses to novel stimuli. We fit the model by maximum penalized likelihood to the spiking activity of zebra finch auditory midbrain neurons in response to conspecific vocalizations (songs) and modulation limited (ml) noise. We compare this model to normalized reverse correlation (NRC), the traditional method for STRF estimation, in terms of predictive power and the basic tuning properties of the estimated STRFs. We find that a GLM with a sparse prior predicts novel responses to both stimulus classes significantly better than NRC. Importantly, we find that STRFs from the two models derived from the same responses can differ substantially and that GLM STRFs are more consistent between stimulus classes than NRC STRFs. These results suggest that a GLM with a sparse prior provides a more accurate characterization of spectrotemporal tuning than does the NRC method when responses to complex sounds are studied in these neurons.

Automating the design of informative sequences of sensory stimuli.

Adaptive stimulus design methods can potentially improve the efficiency of sensory neurophysiology experiments significantly; however, designing optimal stimulus sequences in real time remains a serious technical challenge. Here we describe two approximate methods for generating informative stimulus sequences: the first approach provides a fast method for scoring the informativeness of a batch of specific potential stimulus sequences, while the second method attempts to compute an optimal stimulus distribution from which the experimenter may easily sample. We apply these methods to single-neuron spike train data recorded from the auditory midbrain of zebra finches, and demonstrate that the resulting stimulus sequences do in fact provide more information about neuronal tuning in a shorter amount of time than do more standard experimental designs.

Discrimination of communication vocalizations by single neurons and groups of neurons in the auditory midbrain.

Many social animals including songbirds use communication vocalizations for individual recognition. The perception of vocalizations depends on the encoding of complex sounds by neurons in the ascending auditory system, each of which is tuned to a particular subset of acoustic features. Here, we examined how well the responses of single auditory neurons could be used to discriminate among bird songs and we compared discriminability to spectrotemporal tuning. We then used biologically realistic models of pooled neural responses to test whether the responses of groups of neurons discriminated among songs better than the responses of single neurons and whether discrimination by groups of neurons was related to spectrotemporal tuning and trial-to-trial response variability. The responses of single auditory midbrain neurons could be used to discriminate among vocalizations with a wide range of abilities, ranging from chance to 100%. The ability to discriminate among songs using single neuron responses was not correlated with spectrotemporal tuning. Pooling the responses of pairs of neurons generally led to better discrimination than the average of the two inputs and the most discriminating input. Pooling the responses of three to five single neurons continued to improve neural discrimination. The increase in discriminability was largest for groups of neurons with similar spectrotemporal tuning. Further, we found that groups of neurons with correlated spike trains achieved the largest gains in discriminability. We simulated neurons with varying levels of temporal precision and measured the discriminability of responses from single simulated neurons and groups of simulated neurons. Simulated neurons with biologically observed levels of temporal precision benefited more from pooling correlated inputs than did neurons with highly precise or imprecise spike trains. These findings suggest that pooling correlated neural responses with the levels of precision observed in the auditory midbrain increases neural discrimination of complex vocalizations.