Low-dimensional dynamical model for the diversity of pressure patterns used in canary song

Automated annotation of birdsong with a neural network that segments spectrograms

Songbirds provide a powerful model system for studying sensory-motor learning. However, many analyses of birdsong require time-consuming, manual annotation of its elements, called syllables. Automated methods for annotation have been proposed, but these methods assume that audio can be cleanly segmented into syllables, or they require carefully tuning multiple statistical models. Here we present TweetyNet: a single neural network model that learns how to segment spectrograms of birdsong into annotated syllables. We show that TweetyNet mitigates limitations of methods that rely on segmented audio. We also show that TweetyNet performs well across multiple individuals from two species of songbirds, Bengalese finches and canaries. Lastly, we demonstrate that using TweetyNet we can accurately annotate very large datasets containing multiple days of song, and that these predicted annotations replicate key findings from behavioral studies. In addition, we provide open-source software to assist other researchers, and a large dataset of annotated canary song that can serve as a benchmark. We conclude that TweetyNet makes it possible to address a wide range of new questions about birdsong.

Birds breathe at an aerodynamic resonance

We present a dynamical model for the avian respiratory system and report the measurement of its variables in normal breathing canaries (Serinus canaria). Fitting the parameters of the model, we are able to show that the birds in our study breathe at an aerodynamic resonance of their respiratory system. For different respiratory regimes, such as singing, where rapid respiratory gestures are used, the nonlinearities of the model lead to a shift in its resonances toward higher frequency values.

Dynamical model for the neural activity of singing Serinus canaria

Vocal production in songbirds is a key topic regarding the motor control of a complex, learned behavior. Birdsong is the result of the interaction between the activity of an intricate set of neural nuclei specifically dedicated to song production and learning (known as the "song system"), the respiratory system and the vocal organ. These systems interact and give rise to precise biomechanical motor gestures which result in song production. Telencephalic neural nuclei play a key role in the production of motor commands that drive the periphery, and while several attempts have been made to understand their coding strategy, difficulties arise when trying to understand neural activity in the frame of the song system as a whole. In this work, we report neural additive models embedded in an architecture compatible with the song system to provide a tool to reduce the dimensionality of the problem by considering the global activity of the units in each neural nucleus. This model is capable of generating outputs compatible with measurements of air sac pressure during song production in canaries (Serinus canaria). In this work, we show that the activity in a telencephalic nucleus required by the model to reproduce the observed respiratory gestures is compatible with electrophysiological recordings of single neuron activity in freely behaving animals.

Song Feature Specific Analysis of Isolate Song Reveals Interspecific Variation in Learned Components

Studies of avian vocal development without exposure to conspecific song have been conducted in many passerine species, and the resultant isolate song is often interpreted to represent an expression of the genetic code for conspecific song. There is wide recognition that vocal learning exists in oscine songbirds, but vocal learning has only been thoroughly investigated in a few model species, resulting in a narrow view of birdsong learning. By extracting acoustic signals from published spectrograms, we have reexamined the findings of isolate studies with a universally applicable semi-automated quantitative analysis regimen. When song features were analyzed in light of three different production aspects (respiratory, syringeal, and central programming of sequence), all three show marked interspecific variability in how close isolate song features are to normal. This implies that song learning mechanisms are more variable than is commonly recognized. Our results suggest that the interspecific variation shows no readily observable pattern reflecting phylogeny, which has implications for understanding the mechanisms behind the evolution of avian vocal communication. We emphasize that song learning in passerines provides an excellent opportunity to investigate the evolution of a complex, plastic trait from a phylogenetic perspective.

Nonlinear dynamics in the study of birdsong

Birdsong, a rich and complex behavior, is a stellar model to understand a variety of biological problems, from motor control to learning. It also enables us to study how behavior emerges when a nervous system, a biomechanical device and the environment interact. In this review, I will show that many questions in the field can benefit from the approach of nonlinear dynamics, and how birdsong can inspire new directions for research in dynamics.

From perception to action in songbird production: dynamics of a whole loop

Birdsong emerges when a set of highly interconnected brain areas manage to generate a complex output. This consists of precise respiratory rhythms as well as motor instructions to control the vocal organ configuration. In this way, during birdsong production, dedicated cortical areas interact with life-supporting ones in the brainstem, such as the respiratory nuclei. We discuss an integrative view of this interaction together with a widely accepted "top-down" representation of the song system. We also show that a description of this neural network in terms of dynamical systems allows to explore songbird production and processing by generating testable predictions.

Nonlinear resonances and multi-stability in simple neural circuits

This article describes a numerical procedure designed to tune the parameters of periodically driven dynamical systems to a state in which they exhibit rich dynamical behavior. This is achieved by maximizing the diversity of subharmonic solutions available to the system within a range of the parameters that define the driving. The procedure is applied to a problem of interest in computational neuroscience: a circuit composed of two interacting populations of neurons under external periodic forcing. Depending on the parameters that define the circuit, such as the weights of the connections between the populations, the response of the circuit to the driving can be strikingly rich and diverse. The procedure is employed to find circuits that, when driven by external input, exhibit multiple stable patterns of periodic activity organized in complex tuning diagrams and signatures of low dimensional chaos.

An integrated model for motor control of song in Serinus canaria

Birdsong is a learned motor behavior controlled by an interconnected structure of neural nuclei. This pathway is bilaterally organized, with anatomically indistinguishable structures in each brain hemisphere. In this work, we present a computational model whose variables are the average activities of different neural nuclei of the song system of oscine birds. Two of the variables are linked to the air sac pressure and the tension of the labia during canary song production. We show that these time dependent gestures are capable of driving a model of the vocal organ to synthesize realistic canary like songs.

A circular model for song motor control in Serinus canaria

Song production in songbirds is controlled by a network of nuclei distributed across several brain regions, which drives respiratory and vocal motor systems to generate sound. We built a model for birdsong production, whose variables are the average activities of different neural populations within these nuclei of the song system. We focus on the predictions of respiratory patterns of song, because these can be easily measured and therefore provide a validation for the model. We test the hypothesis that it is possible to construct a model in which (1) the activity of an expiratory related (ER) neural population fits the observed pressure patterns used by canaries during singing, and (2) a higher forebrain neural population, HVC, is sparsely active, simultaneously with significant motor instances of the pressure patterns. We show that in order to achieve these two requirements, the ER neural population needs to receive two inputs: a direct one, and its copy after being processed by other areas of the song system. The model is capable of reproducing the measured respiratory patterns and makes specific predictions on the timing of HVC activity during their production. These results suggest that vocal production is controlled by a circular network rather than by a simple top-down architecture.

Parameter estimation, nonlinearity, and Occam's razor

Nonlinear systems are capable of displaying complex behavior even if this is the result of a small number of interacting time scales. A widely studied case is when complex dynamics emerges out of a nonlinear system being forced by a simple harmonic function. In order to identify if a recorded time series is the result of a nonlinear system responding to a simpler forcing, we develop a discrete nonlinear transformation for time series based on synchronization techniques. This allows a parameter estimation procedure which simultaneously searches for a good fit of the recorded data, and small complexity of a fluctuating driving parameter. We illustrate this procedure using data from respiratory patterns during birdsong production.

Motor control of sound frequency in birdsong involves the interaction between air sac pressure and labial tension

Frequency modulation is a salient acoustic feature of birdsong. Its control is usually attributed to the activity of syringeal muscles, which affect the tension of the labia responsible for sound production. We use experimental and theoretical tools to test the hypothesis that for birds producing tonal sounds such as domestic canaries (Serinus canaria), frequency modulation is determined by both the syringeal tension and the air sac pressure. For different models, we describe the structure of the isofrequency curves, which are sets of parameters leading to sounds presenting the same fundamental frequencies. We show how their shapes determine the relative roles of syringeal tension and air sac pressure in frequency modulation. Finally, we report experiments that allow us to unveil the features of the isofrequency curves.

Evidence and control of bifurcations in a respiratory system

We studied the pressure patterns used by domestic canaries in the production of birdsong. Acoustically different sound elements ("syllables") were generated by qualitatively different pressure gestures. We found that some ubiquitous transitions between syllables can be interpreted as bifurcations of a low dimensional dynamical system. We interpreted these results as evidence supporting a model in which different timescales interact nonlinearly.

Temperature induced syllable breaking unveils nonlinearly interacting timescales in birdsong motor pathway

The nature of telencephalic control over premotor and motor circuits is debated. Hypotheses range from complete usurping of downstream circuitry to highly interactive mechanisms of control. We show theoretically and experimentally, that telencephalic song motor control in canaries is consistent with a highly interactive strategy. As predicted from a theoretical model of respiratory control, mild cooling of a forebrain nucleus (HVC) led to song stretching, but further cooling caused progressive restructuring of song, consistent with the hypothesis that respiratory gestures are subharmonic responses to a timescale present in the output of HVC. This interaction between a life-sustaining motor function (respiration) and telencephalic song motor control suggests a more general mechanism of how nonlinear integration of evolutionarily new brain structures into existing circuitry gives rise to diverse, new behavior.

A species-specific view of song representation in a sensorimotor nucleus

Songbirds constitute a powerful model system for the investigation of how complex vocal communication sounds are represented and generated, offering a neural system in which the brain areas involved in auditory, motor and auditory-motor integration are well known. One brain area of considerable interest is the nucleus HVC. Neurons in the HVC respond vigorously to the presentation of the bird's own song and display song-related motor activity. In the present paper, we present a synthesis of neurophysiological studies performed in the HVC of one songbird species, the canary (Serinus canaria). These studies, by taking advantage of the singing behavior and song characteristics of the canary, have examined the neuronal representation of the bird's own song in the HVC. They suggest that breeding cues influence the degree of auditory selectivity of HVC neurons for the bird's own song over its time-reversed version, without affecting the contribution of spike timing to the information carried by these two song stimuli. Also, while HVC neurons are collectively more responsive to forward playback of the bird's own song than to its temporally or spectrally modified versions, some are more broadly tuned, with an auditory responsiveness that extends beyond the bird's own song. Lastly, because the HVC is also involved in song production, we discuss the peripheral control of song production, and suggest that interspecific variations in song production mechanisms could be exploited to improve our understanding of the functional role of the HVC in respiratory-vocal coordination.

Interaction between telencephalic signals and respiratory dynamics in songbirds

The mechanisms by which telencephalic areas affect motor activities are largely unknown. They could either take over motor control from downstream motor circuits or interact with the intrinsic dynamics of these circuits. Both models have been proposed for telencephalic control of respiration during learned vocal behavior in birds. The interactive model postulates that simple signals from the telencephalic song control areas are sufficient to drive the nonlinear respiratory network into producing complex temporal sequences. We tested this basic assumption by electrically stimulating telencephalic song control areas and analyzing the resulting respiratory patterns in zebra finches and in canaries. We found strong evidence for interaction between the rhythm of stimulation and the intrinsic respiratory rhythm, including naturally emerging subharmonic behavior and integration of lateralized telencephalic input. The evidence for clear interaction in our experimental paradigm suggests that telencephalic vocal control also uses a similar mechanism. Furthermore, species differences in the response of the respiratory system to stimulation show parallels to differences in the respiratory patterns of song, suggesting that the interactive production of respiratory rhythms is manifested in species-specific specialization of the involved circuitry.

The anatomy of onomatopoeia

Virtually every human faculty engage with imitation. One of the most natural and unexplored objects for the study of the mimetic elements in language is the onomatopoeia, as it implies an imitative-driven transformation of a sound of nature into a word. Notably, simple sounds are transformed into complex strings of vowels and consonants, making difficult to identify what is acoustically preserved in this operation. In this work we propose a definition for vocal imitation by which sounds are transformed into the speech elements that minimize their spectral difference within the constraints of the vocal system. In order to test this definition, we use a computational model that allows recovering anatomical features of the vocal system from experimental sound data. We explore the vocal configurations that best reproduce non-speech sounds, like striking blows on a door or the sharp sounds generated by pressing on light switches or computer mouse buttons. From the anatomical point of view, the configurations obtained are readily associated with co-articulated consonants, and we show perceptual evidence that these consonants are positively associated with the original sounds. Moreover, the pairs vowel-consonant that compose these co-articulations correspond to the most stable syllables found in the knock and click onomatopoeias across languages, suggesting a mechanism by which vocal imitation naturally embeds single sounds into more complex speech structures. Other mimetic forces received extensive attention by the scientific community, such as cross-modal associations between speech and visual categories. The present approach helps building a global view of the mimetic forces acting on language and opens a new venue for a quantitative study of word formation in terms of vocal imitation.

Average dynamics of a driven set of globally coupled excitable units

We investigate the behavior of the order parameter describing the collective dynamics of a large set of driven, globally coupled excitable units. We derive conditions on the parameters of the system that allow to bound the degree of synchrony of its solutions. We describe a regime where time dependent nonsynchronous dynamics occurs and, yet, the average activity displays low dimensional, temporally complex behavior.

Hormonal acceleration of song development illuminates motor control mechanism in canaries

In songbirds, the ontogeny of singing behavior shows strong parallels with human speech learning. As in humans, development of learned vocal behavior requires exposure to an acoustic model of species-typical vocalizations, and, subsequently, a sensorimotor practice period after which the vocalization is produced in a stereotyped manner. This requires mastering motor instructions driving the vocal organ and the respiratory system. Recently, it was shown that, in the case of canaries (Serinus canaria), the diverse syllables, constituting the song, are generated with air sac pressure patterns with characteristic shapes, remarkably, those belonging to a very specific mathematical family. Here, we treated juvenile canaries with testosterone at the onset of the sensorimotor practice period. This hormone exposure accelerated the development of song into stereotyped adultlike song. After 20 days of testosterone treatment, subsyringeal air sac pressure patterns of song resembled those produced by adults, while those of untreated control birds of the same age did not. Detailed temporal structure and modulation patterns emerged rapidly with testosterone treatment, and all previously identified categories of adult song were observed. This research shows that the known effect of testosterone on the neural circuits gives rise to the stereotyped categories of respiratory motor gestures. Extensive practice of these motor patterns during the sensorimotor phase is not required for their expression.