Cooperative vocal control in marmoset monkeys via vocal feedback

Abstract

Humans adjust speech amplitude as a function of distance from a listener; we do so in a manner that would compensate for such distance. This ability is presumed to be the product of high-level sociocognitive skills. Nonhuman primates are thought to lack such socially related flexibility in vocal production. Using predictions from a simple arousal-based model whereby vocal feedback from a conspecific modulates the drive to produce a vocalization, we tested whether another primate exhibits this type of cooperative vocal control. We conducted a playback experiment with marmoset monkeys and simulated "far-away" and "nearby" conspecifics using contact calls that differed in sound intensity. We found that marmoset monkeys increased the amplitude of their contact calls and produced such calls with shorter response latencies toward more distant conspecifics. The same was not true in response to changing levels of background noise. To account for how simulated conspecific distance can change both the amplitude and timing of vocal responses, we developed a model that incorporates dynamic interactions between the auditory system and limbic "drive" systems. Overall, our data show that, like humans, marmoset monkeys cooperatively control the acoustics of their vocalizations according to changes in listener distance, increasing the likelihood that a conspecific will hear their call. However, we propose that such cooperative vocal control is a system property that does not necessitate any particularly advanced sociocognitive skill. At least in marmosets, this vocal control can be parsimoniously explained by the regulation of arousal states across two interacting individuals via vocal feedback.

Keywords: arousal; auditory cortex; cingulate; primate vocalizations.

Fig. 1.

A simple model of cooperative vocal-amplitude control. Each marmoset's call dynamics is represented by 3 components: audition, drive, and vocalization. All the interactions are excitatory, except the interaction from audition to the drive, which is inhibitory. Top: hypothetical neural dynamics for the response to nearby (high-amplitude) calls. Bottom: hypothetical neural dynamics for the responses to far-away (low-amplitude) calls.

Fig. 2.

Experimental setup. A: dimension of the experiment room and position of the subjects. The speaker was located in the opposite side of the room. B: approximate distance between a source and the listener, simulated by the 2 different sound-stimulus amplitudes. The high-amplitude sound simulated a distance of ∼5 m from a source producing sound of 90 dB at 0.1 m. The low-amplitude sound simulated a distance of ∼60 m from the same source. C: power spectrum of the stimulus sounds recorded by the microphone at 0.76 m from the subjects. The nearby/high-amplitude (blue line) and far-away/low-amplitude (green line) stimuli are salient compared with the background noise (room + masking noise). The gray region indicates the fundamental frequency range of the typical phee call (6–10 kHz).

Fig. 3.

Responses to nearby and far-away stimuli. A: exemplar of spectrogram and amplitude of responses to the nearby and far-away stimuli; x-axis, time in seconds; y-axis, frequency in kilohertz. The response call to nearby stimulus (blue) had smaller amplitude than the response call to far-away stimulus (green). B: amplitude of the response calls to the nearby (blue circles) and far-away (green circles) stimuli; y-axis, average amplitude in decibels. Different sessions are indicated by lines with different colors joining 2 dots. Sessions from the same subject have the same color. The gray rectangles indicate the 95% confidence intervals. The horizontal black lines indicate the mean amplitude values. C: response time to loud and soft stimuli; y-axis, average response time in seconds. The color convention is as in B.

Fig. 4.

Relationship between amplitude and timing of the responses to nearby and far-away stimuli. A: discriminant analysis for the responses to nearby and far away; x-axis, amplitude in decibels; y-axis, response time in seconds. The responses to nearby stimuli (blue circles) were concentrated on lower-amplitude/longer-response-time region, and the responses to far-away stimuli (green circles) concentrated on high-amplitude/shorter-response-time region. The different line colors indicate different animals. The magenta line is the linear discrimination line separating the responses to nearby from far-away stimuli. B: correlation between the magnitude change in amplitude and response time; x-axis, magnitude change in decibels; y-axis, magnitude change of response time in seconds. The circles indicate different sessions, and the different colors indicate different animals. The orange plus sign indicates the outlier data point. The linear regression line (black) was obtained by applying a robust regression to the data set excluding the outlier.

Fig. 5.

Responses to nearby and far-away stimuli measured by other acoustic parameters. A: response rate to nearby and far-away stimuli; y-axis, response rate = number of responses divided by number of calls produced during baseline. The color conventions are the same as in Fig. 3B. B: number of syllables of the response calls for nearby and far-away stimuli; y-axis, average number of syllables of the response calls. The color conventions are the same as in A. C: call duration for nearby and far-away stimuli; y-axis, average call duration in seconds. Color conventions are the same as in A. D: harmonic attenuation of response calls to nearby and far-away stimuli; y-axis, harmonic attenuation = 10*log(amplitude of f0) − 10*log(amplitude of first harmonic). Color conventions are the same as in A.

Fig. 6.

Vocal production in noisy background. A: exemplar of spectrogram and amplitude (red line) of calls produced during 40-dB and 60-dB background noise; x-axis, time in seconds; y-axis left, frequency in kilohertz; y-axis right, amplitude in decibels. B: Amplitude of calls produced during 60-dB (blue circles), 50-dB (orange circles), and 40-dB (green circles) background noise; x-axis, noise level in decibels; y-axis, amplitude in decibels. The gray rectangles indicate the 95% confidence intervals, and the horizontal black lines indicate the mean relative amplitude values.

Fig. 7.

An elaborated model of marmoset vocal interactions as a function of vocal amplitude. The dynamics of activities in drive, auditory, and motor nodes are simulated. In the x-axis, time is in seconds; in the y-axis, neural activity is represented in arbitrary units. The blue line represents the activity in the drive node, and the red line represents the activity in auditory node. For visualization purposes, we multiplied the activity in auditory by a factor of 2,000. The threshold is indicated by the dashed line. The gray rectangles represent the motor node output (a vocalization). Their height represents the amplitude. The pink regions represent the heard conspecific calls.