Fig. 34.1
Top-view schematic examples of the potential spatial percepts of an acoustic scene with a point source presented in front of the listener. Panel a shows a punctate percept accurately representing the direction and distance of the acoustic event. Panels b and c show the punctate source perceived at different radial and angular positions (i.e. a localisation bias). Panels d–f show intracranial images: a punctate image (d), a broad image (e) and dual images at the two ears (f); headphone presentation without applying proper ear and headphone equalisation would be expected to result in these types of images. Panels g–i show broadened distal images, from slightly broader (g) to stretching across the azimuth (h) to being perceived as all around the listener or fully diffuse (i). Images resembling these could be expected from presentation in a reverberant or multiple-source environment. Panel j shows an unstable spatial percept, with the source location moving, and panel k indicates an indescribable percept
To experimentally determine if there were differences in the percept of sound sources between NH and HI listeners, the current study employed two methods of visual analogy: a drawing task and an identification task. In order to reduce variability, emphasis was placed on the apparent width of the stimulus presented over headphones, not the stimulus location. In the drawing or visual-description task, listeners drew their representation of sound sources with different interaural coherences and simulated positions onto a pre-drawn schematic of a mannequin head. In the identification task, listeners chose the closest match to their perception of the source from a predefined set of 15 images representing hypothetically narrow to wide images presented to the left, centre and right.
2 Methods
A group of 21 HI adults matched for hearing loss with ages ranging from 48 to 77 years were recruited by post. Better ear pure-tone threshold averages ranged from 33- to 43-dB HL with asymmetries of 0–10-dB HL. Four NH listeners (2 female), aged 28–41 years with normal hearing based on pure-tone audiometric thresholds between 500 and 4,000 Hz less than 20-dB HL, were also recruited from employees and students of the Institute of Hearing Research. All listeners had participated in an auditory-source-width discrimination task prior to testing, giving them some familiarity with the concept of source width and the stimuli being used in the current tasks. 2
The stimuli were composed of 500-ms third-octave narrowband noises with octave-spaced centre frequencies from 500 to 4,000 Hz. Each narrowband component was generated using Plenge’s symmetric generator method (1972) to reduce variability within each band: two independent noises were independently attenuated, then added and subtracted, respectively, from each other in the left and right channel for the desired IC. In the drawing task, three IC values were tested: 0.6, 0.8 and 1. In the identification task, five IC values were tested: 0.6–1 in 0.1 increments. The broadband stimuli were presented at 75 dB(A) over circumaural headphones. To examine the effect of location and maintain interest throughout the experimental session, the stimuli were presented from three simulated positions (−30°, 0° and +30°) and monaurally left or right, randomly chosen on each trial. The ±30° positions were created by applying global ITDs and ILDs of 229 μs and 4.8 dB (being the average values across multiple HRTF databases) to the stimuli.
In the visual-description procedure, HI and NH listeners were required to sketch, using a touch screen, the perceived size of the sound they heard on each trial onto a 450-pixel square image of a mannequin head with an ear-to-ear distance of 360 pixels. Listeners’ responses were sometimes incomplete, necessitating a recursive sliding or boxcar average to complete the shapes. The only additional instruction was that, as the experiment was concerned with the size of the sound the listener heard, they were to project any sounds heard to the rear to the front for their response. Each listener was given fully coherent (IC = 1) stimuli with the five locations (L, −30°, 0°, +30° and R) as practice. All listeners then commenced with the experiment, sketching their percept of ten presentations of each of 11 combinations of IC and position (3 IC × 3 positions plus monaural stimuli). Stimuli were presented in random order.
A problem was evident in the visual-description responses for the ±30° stimuli: sketches were constrained by the image of the head. Listeners did not draw images beyond the two ears, so that the area and width were confounded by the centre of the response. As responses for ±30° stimuli were confounded by the visual anchor of the mannequin head, those stimuli were not included in further analysis of the measurement technique. Two of the older HI listeners did not draw shapes at all, only drawing dots to indicate positions (e.g. panel d in Fig. 34.1 above), and three older HI listeners placed left- and right-positioned stimuli in the opposite hemifield, indicating a possible momentary lapse in understanding the mirror-image aspect of the task. The results of these five listeners were not included in the analysis.
Immediately following the drawing task, HI and NH listeners completed the visual identification task. They were presented with stimuli with ICs of 0.6–1 and simulated positions of −30°, 0° and 30° (no monaural stimuli were used). After each stimulus presentation, a 5 × 3 matrix of images was displayed with 20-pixel high grey bars made of visual noise. The widths of the pre-drawn bars ranged across the five columns from 20 to 100 pixels in 20-pixel increments based on previously established near-linear relationships between IC and width (cf. Keet 1968). The positions of the bars across the three rows were at approximately −30°, 0° and +30°. The 15 conditions were presented in a randomised order for each block; after two blocks of practice, listeners completed four test blocks (i.e. a total of 90 trials).
3 Results and Discussion
For the drawing task, the raw results of the 16 HI listeners and four NH listeners are shown in Fig. 34.2; individual sketches were collated into density plots for the 0° stimuli with ICs of 0.6. 0.8 and 1 and the monaural stimuli. Three basic results are clear: (1) changes in IC did not produce noticeably different sized responses for HI listeners, though they did for NH listeners, (2) diotic (IC = 1) stimuli were drawn smaller and narrower by NH listeners relative to HI listeners, and (3) drawing methods varied greatly, especially among HI listeners.
