Scientific Understanding of Consciousness
Consciousness as an Emergent Property of Thalamocortical Activity

Gestures in Bird Song


Researchers must use animal models to explore of the neural circuitry of (FAPs, gestures) involved in the production of behavioral movements. Studies of the neural gestures birds use in the production of bird song can have application in how humans use FAPs to produce movements of vocal speech, riding a bicycle, playing a violin, etc.



Nature 495, 59–64 (07 March 2013)

Elemental gesture dynamics are encoded by song premotor cortical neurons

Department of Organismal Biology and Anatomy, University of Chicago, 1027 East 57th Street, Chicago, Ilinois 60637, USA

Ana Amador & Daniel Margoliash

Department of Physics, FCEN, University of Buenos Aires, Intendente Guiraldes 2160, Pabellon 1, Buenos Aires 1428, Argentina

Yonatan Sanz Perl & Gabriel B. Mindlin

Department of Physics, FCEN, University of Buenos Aires, Intendente Guiraldes 2160, Pabellon 1, Buenos Aires 1428, Argentina.

Ana Amador


Quantitative biomechanical models can identify control parameters that are used during movements, and movement parameters that are encoded by premotor neurons. We fit a mathematical dynamical systems model including subsyringeal pressure, syringeal biomechanics and upper-vocal-tract filtering to the songs of zebra finches. This reduces the dimensionality of singing dynamics, described as trajectories (motor ‘gestures’) in a space of syringeal pressure and tension. Here we assess model performance by characterizing the auditory response ‘replay’ of song premotor HVC neurons to the presentation of song variants in sleeping birds, and by examining HVC activity in singing birds. HVC projection neurons were excited and interneurons were suppressed within a few milliseconds of the extreme time points of the gesture trajectories. Thus, the HVC precisely encodes vocal motor output through activity at the times of extreme points of movement trajectories. We propose that the sequential activity of HVC neurons is used as a ‘forward’ model, representing the sequence of gestures in song to make predictions on expected behaviour and evaluate feedback.

In zebra finches, there is a notable state-dependent neuronal replay phenomenon associated with song learning, so that the strongest and most selective auditory responses are recorded in sleeping birds. We used the responses of HVC neurons in sleeping adult zebra finches to evaluate the responses to BOS and artifical BOS variants, and then tested emerging hypotheses by recording from singing birds.

The avian vocal organ is a nonlinear device that is capable of generating complex sounds even when driven by simple instructions. We extended a low-dimensional model of the avian syrinx and vocal tract that can capture a variety of acoustic features such as the precise relationship between fundamental frequency and spectral content of zebra finch song. A two-dimensional set of equations describes the labial dynamics. Flow fluctuations are fed into a vocal tract, generating an input sound (Pin(t)). The tract filters the sound and is characterized as a trachea, modelled by a tube, which connects to the oro-oesophageal cavity (OEC), modelled here as a Helmholtz resonator. The output of the model is a time trace representing the uttered sound (Pout(t)).

Using this model, we created synthetic versions of the songs that our test birds sang.

Song was described by the sequence of these pressure–tension trajectories, which we call gestures, with gesture onsets and offsets defined as discontinuities in either the pressure or tension functions. Gestures include movements that do not result in phonation, such as pressure patterns associated with mini-breaths between syllables19, but our recordings were limited here to airborne sounds. In a sample of 8modelled songs, there were 13±4 gestures per motif (largest basic unit of song, a repeated sequence of syllables). The distribution of gesture durations (mode = 22.5±2.5ms, range 4–142ms) was non-Gaussian, with 33% of the gestures lasting less than 30ms, and it had a long tail corresponding to slowly varying sounds, such as constant-frequency harmonic stacks.

This simple model captured the essential features of sound production in a framework of labial tension and subsyringeal pressure over which birds have direct motor control.

We have described song organization based on gestures, using the dynamical systems modelling framework to replace analysis of songs based on spectrographs. These features of motor systems organization may be represented in other systems and for other behaviors. Our data support Sherrington’s long-standing hypothesis that the motor cortex is a synthetic organ, representing segments of whole movements. In humans, the production of speech and the performance of athletes and musicians are an exceptional example of highly precise learned skilled behaviour that could have similar mechanisms to those described here. The development of corresponding models for human speech production should help to provide insight into speech and language pathologies in which sequential behaviour is disrupted.

[end of paraphrase]



Nature 495, 56–57 (07 March 2013)

Neuroscience: The units of a song

Todd W. Troyer

Todd W. Troyer is in the Department of Biology, University of Texas at San Antonio, San Antonio, Texas 78249, USA.


What is the basic unit of speech? The word? The syllable? The phoneme? This question has been vexing speech and language researchers for decades, and similar questions have challenged those who study songbirds. Whereas behavioural evidence1 supports the grouping of songs into 100–250-millisecond vocalizations called syllables, neurophysiological data suggest that the premotor areas at high levels in the hierarchy of motor neurons in the brain act more like a clock, providing a continuous stream of activity on a 10-millisecond timescale. Researchers reconcile these data by providing evidence that the song code generated by motor neurons of zebra finches (Taeniopygia guttata) is indeed broken into discrete 'gestures', which are significantly shorter than song syllables.

The study has roots in two research programmes that started at opposite ends of the motor-coding problem. One group studied the highest levels of the motor system, in which sensory signals about a song's acoustics change the song motor program during learning. The researchers discovered that, for every rendition of the bird's song, individual neurons produce short bursts of activity with incredible regularity and precision. They also demonstrated a remarkable correspondence between the motor activity that was recorded when the bird was singing and the auditory activity that resulted from playing the bird's song back to it when it was asleep.

The other team investigated how sound is generated by the avian vocal organ, the syrinx. They developed a simplified biophysical model of the syrinx with two dynamic parameters: the pressure in the bird's air sac and the spring-like tension on a vibrating membrane controlled by the muscles surrounding the syrinx. Analysis of the model showed that small changes in pressure and tension can lead to output that is a passable imitation of the sounds produced by several species of songbird. This work also suggested that, to sing, birds may not need precise control over a large ensemble of muscles. Rather, two basic signals may suffice, as long as the signals are controlled in a temporally precise manner.

Combining their previous approaches, the two research programmes now come together. Amador et al. focus on a high-level cluster of neurons called the HVC, which is essential for singing, but — in terms of synaptic connections between neurons — is the most distant from the syrinx. They recorded the activity of individual HVC cells either while the birds sang or during playback of the bird's own song while it slept. They also tuned the syrinx model to reproduce each bird's song. By defining a vocal gesture as a period of time when both the pressure and tension parameters were either unchanged or strictly increasing or decreasing, they could divide the song into a sequence of distinct gestural units.

On aligning the neural and behavioural data, the authors found that activity bursts in HVC neurons occurred at specific time points in the song, namely at the boundaries between gestures. The results suggest that the gesture — which is longer than a burst but shorter than a syllable — might be the basic unit of song production.

This finding contrasts with the reigning view of the motor code for birdsong that was originally developed to account for the precise bursting activity of HVC neurons. Finding no clear relationship between burst timing and the division of song into syllable-base units, researchers proposed that the HVC acted more like a clock: bursting in one set of HVC neurons triggered a burst in the next set, forming a continuous set of 'ticks' throughout the song.

Although the clock and gesture hypotheses lead to different views of the motor code for song, it is entirely possible that whereas bursting activity in HVC neurons tends to align with gesture transitions, a sufficient number of HVC neurons is active throughout each gesture to sustain clock-like functionality. Because ruling out this variation on the clock hypothesis would require demonstrating a negative — that there are no HVC neurons active during gestures — the debate over the status of the two hypotheses will probably linger for some time.

Amador and colleagues' results also contain a deeper mystery, the resolution of which may yield insight into how a bird learns its song. The mystery stems from their observation that the average delay between an HVC burst and its associated gesture transition was near zero milliseconds. However, neural signals in the HVC must be relayed through several stages before they can alter the contraction of respiratory and syringeal muscles, a process estimated to take 20 milliseconds8. Thus, the bursts recorded during singing occur too late to actually cause gesture transitions. Similarly, the sound signal that arrives at the bird's ears has to traverse several synapses, causing an estimated delay of 15 milliseconds, before a sensory representation of it is registered in the HVC. This means that the bursts recorded during sleep, which align to sound with a zero-millisecond delay, occur too early to be caused by the auditory detection of a gesture transition.

Although we cannot yet expect definitive answers to the question of how high-level motor representations determine the control signals for song production, the syringeal- modelling approach pursued by Amador et al. provides both a method for breaking the song down into its basic units and evidence that HVC bursts are related to specific events in a bird's song. With a better understanding of the basic units, these results provide a foundation for understanding how birds learn to string these pieces back together to produce a whole song.

[end of paraphrase]



Return to — Fixed Action Patterns (FAPs)