Item

Abstract

Recent work has shown that in auditory cortex acoustic stimuli are potentially encoded by
different neural codes, each operating on different temporal scales. For example, the millisecond-
precise timing of individual neuron’s action potentials has been implicated similarly
as firing rate modulations on slower scales or the timing of spikes to ongoing oscillatory
background activity [1]. Here we asked whether the temporal precision of these putative
neural codes is fixed and inherent to the system, or whether their temporal precision is determined
by the acoustic stimulus.
Stimulus information in different codes was compared during stimulation with naturalistic
sounds and sequences of random tones. The natural sounds had a typical autocorrelation of
around 20–30 ms (computed from the envelope of individual frequency bands), while random
tones had a much shorter autocorrelation time (around 10 ms). Neural activity was
recorded using multiple electrodes in primary and secondary auditory cortex of macaque
monkeys passively listening to these stimuli. Mutual information between stimulus and neural
activity was characterized using previously established approaches [2,3].
We found that the precise time scale of each code depends on the acoustic stimulus. For binary
spike words (spike timing), the temporal precision required to decode maximal information
was higher during stimulation with random tones (average 7 ms) than with natural
sounds (average 12 ms). In addition, the degree to which field potentials were stimulus
locked (‘entrained’) varied between sound types: during stimulation with random tones entrainment
was stronger and extended to much higher frequencies (up to 60Hz) than during
stimulation with natural sounds (about 30 Hz).
These results extend previous finding in the visual thalamus and demonstrate that the temporal
precision of sensory neurons responses in auditory cortex depends on the temporal
structure of the stimulus. In particular, stimuli with shorter correlation times, hence faster intrinsic
time scales, induce responses that vary on shorter time scales. This implies that the
relevant time scales of neural codes are not fixed, but are dynamically adapted to, or reflect
the environment.