Dynamic-range compression affects the lateral position of sounds
Core lateralization results for the high-pass (left) and full-bandwidth (right) conditions, for sounds presented at a simulated azimuth of ±60°. Boxes range from the median of the reported leftmost extents of the auditory event to the median of the rightmost extents across all pooled responses (referred to the right hemisphere). The horizontal axes cover the range from the center of the head to a fully lateralized position. The dashed vertical lines are indicative of the average position suggested by the steady-state compressed interaural level difference in the high-frequency channel. Also shown are the dummy-trial results (averaged across all stimuli) for unprocessed sounds presented from simulated source azimuths between 0° and 45°.
Distributions of midpoint (MP) and span (SP) across participants in the high-pass (left) and full-bandwidth (right) conditions. Midpoint and span values have been converted back to the original response scale for plotting such that values between zero and one cover the range from the center of the head to a fully lateralized position. Central marks represent the median, boxes indicate the interquartile range (IQR), and whiskers extend to the most extreme data points still within 1.5 IQR of the box edges. Data points lying outside this range are plotted as small open circles. Asterisks indicate a significant difference compared to the unprocessed condition.
(Color online) Overall response option percentages in the high-pass (left) and full-bandwidth (right) conditions. In each set of three bars, the upper bar represents the unprocessed condition (UN), the middle bar the dynamic compression condition (DC), and the lower bar the static ILD bias condition (SB). Individual bars are divided into three parts representing the relative frequency of use of each response option: (1) “single, stationary image” (leftmost part); (2) “moving/gradually broadening image” (middle part); and (3) “split image” (rightmost part). The percentages within each bar were derived from a total of 220 trials.
Lateralization responses in the high-pass condition averaged across the SLOW ONSET NB, SAM NB, and SPEECH stimuli. Results in the dynamic compression (DC) condition are plotted separately for trials in which a “single, stationary image” was reported and trials in which a “moving/gradually broadening image” or “split image” was reported. The percentage of trials in each category is shown. Overall results for the unprocessed and static ILD bias conditions are included for comparison. Boxes extend from the median leftmost extent to the median rightmost extent calculated across all relevant trials (referred to the right hemisphere).
Normalized histograms of the short-term interaural level difference (ILD) in the high-frequency channel. For each stimulus, three histograms are plotted, representing ILDs in the unprocessed (upper panel), dynamic compression (middle panel), and static ILD bias (lower panel) conditions. The average ILD weighted by the signal power in each analysis window is shown (ILDw.a.). For the dynamic compression condition, the “proportion of partially compressed ILDs” (ILDp.p.c.) is also given, equal to the proportion of ILDs exceeding the 99th percentile of all ILDs in the static ILD bias condition (indicated by the dashed vertical lines).
Median span (average across participants) in the dynamic compression, high-pass condition plotted against the “proportion of partially compressed ILDs” (ILDp.p.c.) for each stimulus. The positive correlation indicates that the more compression caused ILDs to change dynamically during presentation of the stimulus, the greater was the overall lateral extent of the auditory event.
Details of the stimuli used in the experiment.
Significant effects (p < 0.05) of dynamic compression or a static ILD bias on midpoint and span. Median values across participants are given for the two processing conditions being compared in each case (along with the difference between these). All values have been converted back to the original response scale for consistency with Fig. 2. Exact p values are also given (Wilcoxon signed-rank tests, two-tailed, N = 11).
Article metrics loading...
Full text loading...