Detectability index measures of binaural masking level difference across populations of inferior colliculus neurons

Abstract

In everyday life we continually need to detect signals against a background of interfering noise (the "cocktail party effect"): a task that is much easier to accomplish using two ears. The binaural masking level difference (BMLD) measures the ability of listeners to use a difference in binaural attributes to segregate sound sources and thus improve their discriminability against interfering noises. By computing the detectability of tones from rate-versus-level functions in the presence of a suprathreshold noise, we previously demonstrated that individual low-frequency delay-sensitive neurons in the inferior colliculus are able to show BMLDs. Here we consider the responses of a population of such neurons when the noise level is held constant (as conventionally in psychophysical paradigms). We have sampled the responses of 121 units in the inferior colliculi of five guinea pigs to identical noise and 500 Hz tones at both ears (NoSo) and to identical noise but with the 500 Hz tone at one ear inverted (NoSpi). The result suggests that the neurons subserving detection of So tones in No (identical noise at the two ears) noise are those neurons with best frequencies (BFs) close to 500 Hz that respond to So tones with an increase in their discharge rate from that attributable to the noise. The detection of the inverted (Spi) signal is also attributable to neurons with BFs close to 500 Hz. However, among these neurons, the presence of the Spi tone was indicated by an increased discharge rate in some neurons and by a decreased discharge rate in others.

Fig. 1.

A, Histogram of the BF distribution of 121 units. B, Threshold levels for binaurally presented BF signals and No noise. C, Difference between No level and noise threshold plotted against the BF of the units.D, Histogram showing the level of the No noise with respect to noise threshold for the 121 units.

Fig. 2.

Complete profile of responses of a single inferior colliculus neuron. A, Noise delay function.B, Interaural phase difference histogram for 500 Hz tones measured using binaural beats (Kuwada et al., 1979).C, Masked rate-versus-level function for NoSo.D, Masked rate-versus-level function for NoSπ. The binaural configurations are linked to the delay functions inA and B by arrows.

Fig. 3.

Use of the standard separation analysis to determine the masked thresholds for two inferior colliculus neurons (A, B). Rate-versus-level functions are shown for So (filled circles) and Sπ (open squares) 500 Hz tones in the presence of noise in the same phase (No) at the two ears. The discharge rates over the first 10 dB of the rate-versus-level function were used to estimate the mean and SD of the discharge rates attributable to the noise alone. The standard separation values are plotted for each 1 dB step in tone level in (C, D). The criterion of a D value of 1.0 was used to determine the minimum sound level at which a response to the tone (either an increase or a decrease in the discharge rate) could be detected (i.e., the masked threshold). The masked thresholds are shown by the arrows in C andD, and these are replotted in the insetsof A and B.

Fig. 4.

Masked thresholds for So and Sπ 500 Hz tones in No noise as a function of the neuron BF. A, Masked thresholds across a population of low-frequency inferior colliculus neurons pooled from five guinea pigs for 500 Hz tones in So (filled circles) and Sπ (open squares) configurations in the presence of a fixed level (23 dB SPL/ $\sqrt{\text{Hz}}$) of No noise. The dashed lines indicate the region from 300–800 Hz centered around the tone frequency of 500 Hz, which is marked by an arrow. B, The average of all of the masked thresholds across this population is plotted for So (filled circles) and for Sπ (open squares). The difference between these values is significant at the p < 0.001 level by Student’st test. C, The average of the masked thresholds from the frequency range 300–800 Hz, indicated by thedashed lines in A, is plotted for So (filled circles) and for Sπ (open squares). The difference between these values is significant at the p < 0.001 level by Student’s ttest.

Fig. 5.

Proportions of neurons showing increases or decreases in discharge rate at masked threshold and their average masked thresholds. A, Histogram of the number of units in our population that yielded positive (type P) or negative (type N)D values at the masked threshold in response to So (filled bars) or Sπ (open bars) 500 Hz tones. B, Averages of the masked thresholds for So (filled circles) and Sπ (open squares) tones computed separately for units yielding positive and negative D values, as indicated by thearrows from the histogram bars to the corresponding masked thresholds.

Fig. 6.

Population profiles of detectability. Here we have taken two arbitrary signal-to-noise ratios, which are likely to provide an indication of the activity in the IC when signals are close to the behavioral threshold for the guinea pig. The standard separationD is plotted against BFs for 121 units for NoSo and NoSπ for within-channel S/N ratios of 0 (A,B) and −5 dB (C, D) (see Materials and Methods for details).

Fig. 7.

Masked thresholds for NoSo and NoSπ plotted against the NBD for those units in Figure 6 with significant changes in their discharge rate. Thresholds for NoSo and NoSπ for a single unit are indicated by the same number.Dashed line, S/N ratio = 0 dB; dotted line, S/N ratio = −5 dB.