7.3 Spectro-temporal receptive fields

A1 neurons do not respond only to pure tones. The most informative way to describe a single A1 neuron is to characterize its spectro-temporal receptive field (STRF) — the function that says, for each frequency and each time-before-spike, how much that frequency at that lag contributes to the neuron’s firing probability.

An STRF can be measured directly. Play a long sequence of random spectrotemporal stimulus (a “spectrotemporal noise” or a “dynamic ripple”), record from a neuron, and compute the spike-triggered average of the stimulus over a window of, say, ±100 ms. The result is a 2-D function — frequency on one axis, time on the other — with positive (excitatory) and negative (inhibitory) regions.

Most A1 STRFs are band-pass in frequency (excitatory at the CF, inhibitory at nearby frequencies — sometimes called side-band inhibition) and have characteristic temporal structure: a quick excitatory peak around 10–30 ms before the spike, often followed or flanked by inhibitory regions that confer sensitivity to temporal modulations and spectral motion (frequency sweeps).

The interactive below renders a parametric STRF.

The pedagogical point is that A1 cells are richer than their CFs alone suggest. A neuron with the STRF you have just designed will respond preferentially to a particular stimulus class — say, an upward frequency sweep at a particular rate, or a sustained tone with amplitude modulation at a particular frequency — and weakly or not at all to other stimuli of equal energy. The cortex’s representation of sound is no longer just a tonotopic snapshot; it is a multidimensional feature space in which each cell encodes a slightly different kind of acoustic event.