5.4 Rate-place coding above the phase-lock limit

What about frequencies above 5 kHz? Phase locking is gone. The only information a fiber can convey is its average firing rate, modulated by the input. This is rate coding, and it is what carries the high-frequency content of speech — the fricatives, the sibilants, the consonant releases — that we saw clustered in the basal half of the cochlea in 5.4.

Rate coding by itself is a noisy code. A fiber’s spontaneous firing follows approximately a Poisson process, so the variance of the spike count equals its mean — and rate estimates require many spikes to be reliable. To extract a high-frequency signal from rate coding alone, the brain has to integrate across multiple fibers: the population of fibers with CFs near the stimulus frequency, taken together, has many more samples than any one fiber. Population rate-place coding — knowing which fibers are firing more than baseline — is what tells the brain that a particular high frequency is present.

Below 1.5 kHz, by contrast, the brain has both place coding and phase locking available. Most theories of low-frequency pitch perception take advantage of phase locking, because phase information is sub-millisecond accurate while place information is broad. This dual encoding is part of why low-frequency pitch perception is so much more accurate than high-frequency pitch perception: you can perceive a 100 Hz pitch to within a few cents, but a 7 kHz pitch is much vaguer.