Item

Abstract

Human and monkey neuroimaging and monkey electrophysiological studies suggest that neurons
in the anterior superior-temporal lobe are selective
for species-specific vocalizations. To better understand the basis of this selectivity, we studied
the coding properties of these neurons using
extracellular recordings in the awake macaque.
We used a paradigm based on a previous
macaque fMRI study to localize with electrophysiological recordings a voice-sensitive region in the anterior superior-temporal plane that prefers species-specific vocalizations over other complex
sound categories. This revealed a cluster
of vocalization-preferring sites about 5mm anterior
to the tonotopically organized field RT. To
evaluate the neurons’ sensitivity to different vocal
components, we used a set of 12 species-specific
vocalizations and several acoustical manipulations
of these calls. These controls involved, 1)
preserved spectrum (PS) versions of the calls, 2)
preserved envelope (PE) versions, i.e., pink noise
shaped with the Hilbert extracted call envelope,
and 3) preserved spectrum and envelope (PSE)
versions, which combine the first-order spectral
and temporal characteristics of the calls, i.e., their
extracted frequency spectrum shaped with their
envelope (see Figure 1).
Comparing the responses to original calls and the
controls, only 29 of the units significantly preferred
one of these four categories, suggesting
that the responses of many neurons are robust
to our spectro-temporal manipulations. Of the
selective units, the majority (60) favored the
preserved spectrum sounds (PS; Fig. 1), indicating
that these neurons are more sensitive to
spectral than temporal components. Yet, a linear
response classifier, inferring the identity of a vocalization
from a neuron’s single trial responses,
better decoded the original calls than the controls
in the population of neurons. In addition, we found
that the neurons are more selective for and more
sparsely encode the original calls than the acoustical
controls. Noteworthy, in comparison to previous
reports from the auditory core, belt, parabelt
and insular regions, the neurons in the anterior
superior-temporal plane were considerably more
selective to individual vocalizations (Fig. 1D).
We then tested whether these neurons encode
acoustical, phonetic, properties of calls or their
presumed functional meaning (semantics). More
units discriminated between acoustically different sounds belonging to a similar semantic category
(e.g., coo vs. grunt) than those that were acoustically
similar and from different semantic categories
(e.g., grunt vs. pant threat). These results
suggest that neuronal responses at this stage of
the auditory processing hierarchy are governed
by the acoustics of the calls.
In conclusion, evaluating single neuron responses
to the features of species-specific vocalizations
is clarifying the function of the voice-sensitive
regions of the primate brain. Although many of
the neurons did not prefer any specific call type,
they were selective for and could well decode
the species-specific vocalizations and their responses
revealed some preference for the spectral
features of the calls. Our findings suggest that
these neurons encode the acoustical features of
species-specific vocalization, such as the spectrum
of formant frequencies, which can provide
caller species and identity information perhaps
independently of a vocalization’s content.