5.2 Frequency code
Although the evidence for the place theory of frequency coding is compelling, there is some question as to whether the tuning curves obtained from neurons in the auditory nerve provide a mechanism for frequency discrimination that is fine enough to account for behavioural data. People can detect remarkably small differences in frequency – in some cases as small as 3 Hz (for a 1000 Hz signal at moderate intensity). What accounts for this ability? As early as 1930, the American experimental psychologists Glen Wever and Charles Bray proposed that in response to a pure tone, the vibration of the basilar membrane matches the input frequency. They further suggested that the auditory receptors respond in such a way that the temporal pattern of basilar membrane vibration is reproduced in the firing pattern of the neuron. This could be achieved if auditory nerve fibres respond by firing one or more action potentials at the same time in every cycle of a pure tone. This is known as a phase-locked response, since the response appears locked to a certain point (e.g. the peak) in the stimulus (Figure 24a). By phase locking, the response pattern of the nerve fibre would accurately reflect the frequency of the sound wave. This is called the frequency code.

This idea is attractive and there is evidence that it occurs, but only at low frequencies. The reason for this is that neurons cannot fire much faster than about 1000 action potentials per second (they have an absolute refractory period of about 1 ms and cannot fire twice in succession at intervals of less than 1 ms). Therefore they cannot fire during each cycle of the stimulus for stimuli above 1000 Hz (1 kHz). This realisation led Wever and Bray to propose the operation of a volley principle illustrated in Figure 24b. In this figure, the frequency of the sound wave illustrated on top is too high for a single fibre to fire on every cycle. According to the volley principle each fibre only fires at a certain point in the cycle although it does not respond to each cycle. Each of the eight fibres illustrated is firing in phase; that is, if on any cycle a given auditory nerve fibre does fire, it does so at the same relative position within the cycle. If the responses of all fibres are then combined, as may happen further up in the auditory system, then information regarding signal frequency is preserved. The bottom trace in the figure shows the combined responses of all eight fibres; while none of the individual fibres reproduces the pattern of the wave, the combined response is sufficient to reproduce the frequency of the incoming signal. Using this principle, fibres can phase lock to signals with frequencies up to 5 kHz thereby enhancing the transmission of information about stimulus frequency. Above this level, the variability inherent in neural firing becomes too great for such fine patterns to be resolved, and the frequency is probably coded for solely by the place code.
The central nervous system therefore gains information about stimulus frequency in two ways. First there is the place code: the fibres are arranged in a tonotopic map such that position is related to characteristic frequency. Second, there is the frequency code: fibres fire at a rate reflecting the frequency of the stimulus. Below 50 Hz, it appears that frequency is encoded solely by the frequency code. Frequency coding is also of particular importance when the sound is loud enough to saturate the neural firing rate (Section 6). Fibres of many characteristic frequencies will respond to a loud signal because it will be above threshold even for fibres with characteristic frequencies that are different from the signal frequency (although they will respond less vigorously). However, frequency information will still be encoded in the temporal firing pattern of all stimulated fibres.