Responses to dichotic tone-in-noise stimuli in the inferior colliculus

Abstract

Human listeners are more sensitive to tones embedded in diotic noise when the tones are out-of-phase at the two ears (N₀S_π) than when they are in-phase (N₀S₀). The difference between the tone-detection thresholds for these two conditions is referred to as the binaural masking level difference (BMLD) and reflects a benefit of binaural processing. Detection in the N₀S_π condition has been explained in modeling studies by changes in interaural correlation (IAC), but this model has only been directly tested physiologically for low frequencies. Here, the IAC-based hypothesis for binaural detection was examined across a wide range of frequencies and masker levels using recordings in the awake rabbit inferior colliculus (IC). IAC-based cues were strongly correlated with neural responses to N₀S_π stimuli. Additionally, average rate-based thresholds were calculated for both N₀S₀ and N₀S_π conditions. The rate-based neural BMLD at 500 Hz matched rabbit behavioral data, but the trend of neural BMLDs across frequency differed from that of humans.

Keywords: binaural cues; binaural detection; binaural masking level difference; interaural correlation; midbrain.

FIGURE 1

Example neural ITD responses (black solid curve) and fitted Gabor function (blue dashed curve) for peak-like (left) and trough-like (right) NDFs. Vertical dotted line indicates the best ITD. The neuron’s CF, Best or Worst ITD (d_BITD), and for cyclic ITD curves, the ITD tuning frequency (f_ITD) are described in the text.

FIGURE 2

Distribution of MTFs across CF (in one-octave bins) for the units presented in this study. Gray shades from light to dark indicate units with band-enhanced (BE), band-suppressed (BS), hybrid and all-pass (AP) MTF shapes. Two neurons with CF of 12.1k were included in the last bin for simplicity. Most MTF types were represented across the range of CFs, although hybrid MTFs were not observed at the lower CFs.

FIGURE 3

Responses of two example neurons (top and bottom row respectively). (A,E) MTF, response rates to amplitude-modulated noise; stars indicate modulation frequencies that had rates significantly different from the unmodulated condition. (B,F) ITD sensitivity, response rates vs. time delay in contralateral side (negative indicates ipsilateral side has delay). (C,D,G,H) responses to N₀S₀ and N₀S_π stimuli at different noise levels (different symbols) vs. SNR (from left to right); filled symbols indicate supra-thresholds. Errorbars indicate standard deviation. MTF shape and tone frequency for TIN stimuli (close to CF) are shown on the left. The example BE neuron had decreasing rate upon addition of a tone for both N₀S₀ and N₀S_π, while the example BS neuron had increasing rate for both conditions.

FIGURE 4

Responses of six example neurons (A–F). The left two columns show the neuron’s MTF and ITD sensitivity, respectively. The right three columns show the neuron’s response to N₀S₀ (blue circles) and N₀S_π (red squares) TIN stimuli at noise levels of 35, 55, and 75 dB SPL, respectively; filled symbols indicate supra-threshold responses. MTF shape and tone frequency of TIN stimuli (close to CF) are shown on the left.

FIGURE 5

Correlation between dynamic ranges of responses to N₀S_π and ITD (A), N₀S_π and ILD (B), and ILD and ITD (C) at 65 dB SPL (as indicated in titles). Correlation coefficients and p-values are shown at the top right of each panel; a star indicates that the correlation coefficient was significant after Bonferroni correction (p < 0.017). Neurons with CF below 1.5 kHz (low–CF) are shown with filled triangles, whereas neurons with CF above 1.5 kHz (high–CF) are shown with open circles. Solid gray lines indicate linear regressions.

FIGURE 6

Correlation between the rate difference elicited by addition of a dichotic tone (N₀S_π) at 0 dB SNR and the rate difference between responses to N₀ and N_u conditions. Correlation coefficients and p-values are shown; a star indicates that the correlation coefficient was significant after Bonferroni correction (p < 0.0014). Neurons with CF below 1.5 kHz (low–CF) are shown in filled triangles, whereas neurons with CF above 1.5 kHz (high–CF) are shown in open circles. Solid lines show linear regressions.

FIGURE 7

Rate-based threshold for N₀S₀ (A) and N₀S_π (B) conditions. Thresholds of most sensitive neurons across frequencies matched human behavioral data for the N₀S₀ condition, but had a trend different from human for the N₀S_π condition. Neural thresholds at 500 Hz matched rabbit behavioral data for both conditions.

FIGURE 8

Binaural masking level differences (BMLDs) calculated based on single-neuron thresholds for both N₀S₀ and N₀S_π conditions. Open triangles indicate that the direction of change in rate vs. SNR at threshold for the N₀S_π condition was the same as for the N₀S₀ condition, whereas filled triangles indicate opposite direction of change in rate at threshold for the N₀S₀ and N₀S_π conditions. Only neurons that had measurable thresholds in both N₀S₀ and N₀S_π conditions are shown here.

FIGURE 9

N₀S₀ (solid blue line) and N₀S_π thresholds (dashed red line) of the neural population across frequency. Individual neural thresholds at 79.1% correct for N₀S₀ (blue square) and N₀S_π (red diamond) conditions, for noise levels of 55–75 dB SPL are shown for all neurons with measurable thresholds above the lowest SNR tested. Symbols with a black star indicate that the threshold was lower than the lowest measured SNR. Human detection thresholds are from van de Par and Kohlrausch (1999) and shifted up by 4 dB for comparison with neural thresholds, which were computed using a higher criterion. Neural binaural masking level differences (BMLDs) had a different trend across frequency compare to human BMLDs.