In measurements of the frequency threshold curves (FTCs) of single cochlearnerve fibres in animal with acute pathological conditions of the cochlea deterioration in the FTC bandwidth is generally directly related to CF-threshold elevation (e. g. Evans, 1975). Our early studies of such changes in chronic pathological conditions (kanamycin poisoning), however, suggested that on occasion, deterioration in tuning bandwidth could occur without substantial change in threshold. In our own psychoacoustical investigations, and those of others, on patients with hearing loss of cochlear origin, substantial deviations from this direct relationship have also been found. In these latter studies, although impaired frequency resolution is correlated with absolute threshold elevation, most studies report correlation coefficients in the range 0.4 to 0.8 (e. g. Pick et al., 1977, Dreschler, 1980, Ritsma et al., 1980, and Tyler et al., 1982). It would seem that not all of this variance can be ascribed to poor experimental design. This belief is substantiated by two additional observations: a) Pick (1980a) reported evidence which suggested that poor frequency resolution can be dissociated from threshold in TTS, and b) more recently, clear physiological evidence has been obtained that FTC bandwidths in guinea pig can be impaired, without substantial elevation of threshold, after chronic or acute administration of ototoxic diuretics (Pratt and Comis, 1982, and Comis et al., 1981).

This chapter discusses about high-technology hearing aids to compensate for hearing impairment of cochlear origin. Patients having substantial hearing losses of cochlear origin have less than expected intelligibility at higher levels of amplification. The impairment of hearing, therefore, must be attributable to processes other than attenuation. A number of types of distortion of perception have been identified as associated with cochlear damage. The first distortion has been identified for many years and has been given the somewhat misleading name of recruitment. Recruitment is a distortion of perceived intensity, whereby an intense sound requires much less amplification than a quiet sound. A second type of distortion which occurs within the pathological cochlea is impairment of frequency resolution. In general, the greater the cochlear hearing loss, the greater is the deterioration in frequency resolution, A third type of distortion is of the temporal parameters of encoding of the sound. Patients with cochlear hearing loss are less able to hear a short silent period of a few milliseconds embedded in a sound than people with normal hearing. A fourth distortion also operates in the temporal domain and is probably closely related to recruitment. For normal ears, a short-duration sound needs to be of a higher amplitude than a longer-duration sound in order to be equally loud. For sounds with durations in the range of approximately 20-200 ms, sound energy appears to be perfectly integrated, so that the longer sound needs to have 3-dB less amplitude for each doubling of duration in order to be equally loud. However, in cochlear hearing loss, this figure is less than 3dB, the discrepancy becoming greater for greater hearing loss.

From gated tone thresholds measured in the presence of ’’comb-filtered’’ (sinusoidal energy spectrum) noise maskers, the shape of the ’’auditory filter’’ was estimated in five frequency regions, 0.5, 1, 2, 4, and 8 kHz, at a number of masker levels. The auditory filter shape became wider, asymmetrical, and shifted towards lower frequencies as the level increased, for filters with center frequencies in excess of 1 kHz. The assumptions inherent in the derivation of the filter functions are discussed, and it is concluded that the likely deviation from these assumptions (in particular, as a result of the generation of nonlinear distortion products) leads to only small errors in the shapes of the derived filter functions.

This paper examines analytically and by stimulation a method described by Møller, (Møller, A.R. (1970): Acta Physiol. Scand. 78, 299–314) using click-pairs to characterize the linear component of a system which contains nonlinearities. Møller has used the method to obtain the filter function of the response of neural units in the peripheral auditory system. We show that exact solutions for the filter functions are obtained with this method only if the system nonlinearity is quadratic. The method offers approximate solutions for systems containing other even-order nonlinearities, since the resulting filter functions are ‘contaminated’ by higher-order terms.

The suggestion that the frequency dependence of the slope of the cochlear fibre discharge rate vs. intensity function is evidence for a frequency-dependent basilar-membrane nonlinearity, is challenged by a demonstration that such frequency dependence of slope can be derived from a model of the cochlea which involves only linear filters and memoryless, frequency-independent nonlinearities. Some of the shortcomings of both models are briefly discussed. A third type of model, involving feedback, is tentatively proposed, which might more parsimoniously explain neurophysiological results.

The “paired-click” technique for characterizing the linear filtering properties exhibited by single neurons of the auditory system was examined. This method, as described by Møller [Acta Physiol. Scand. 78, 299–314 (1970)], makes use of an acoustic stimulus consisting of two clicks, of the same or opposite polarity, separated by time Δt;; the number of neural spikes evoked by the click pair is measured as a function of Δt. From this measurement, the autocorrelation function of the linear part of the system and its Fourier transform (the energy response of the system) may be determined. Using an analytical treatment of the technique supplemented with a computer simulation model, we show that only if the system contains an even-order nonlinearity does the method yield the characterizing autocorrelation function. Moreover, unless the nonlinearity is purely quadratic, the autocorrelation function is “contaminated” by higher-order terms. In addition, the effect of the “equivalent stimulus intensity” transformation and the amplitude of the click pair is examined.

A digital CMOS comb-filtered noise generator has been developed for use as a stimulus in auditory physiological and psycho-physical experiments. It incorporates two identical, 34-bit shift register pseudo-random noise generators whose outputs are summed and low-pass filtered. By introducing a delay between the generators, a noise power spectrum results which is co-sinusoidal on a linear frequency scale (i.e., "comb-filtered"). The generator may be operated either manually or automatically with a range of delays of 1 Ps or greater, selectable in 1 ps increments. The polarity of one of the shift registers may be changed to produce an inverted output spectrum. The peak-to-valley ratio of the spectrum is adjustable in discrete steps over the range from 0 to 35 dB. The output noise is low-pass riltered at 50 kHz, and has Gaussian amplitude characteristics to 3 a.

In order to provide a starting point for neurophysiologists who seek neurons in the cat sensitive to frequency transitions, the 170 frequency transitions present in 36 cat vocalizations were analyzed in terms of octave rate of change and duration. The results are presented as histograms and scattergrams.