Psychoacoustic data for the study "Notched noise reveals differential improvement in the neural representation of sound"

Psychoacoustic data for Matlab including audiometric thresholds from the study "Notched noise reveals differential improvement in the neural representation of sound"

The study was approved by the Ethics committee of the University of Oldenburg.

Stimuli were sinusoidally amplitude-modulated (SAM) tones with a fixed modulation frequency fm = 128 Hz at full modulation depth of 100% and fc = 4000 Hz. 

The procedure was a two-interval, two-alternative forced-choice task (2I-2AFC). The task of the subject was to indicate whether the SAM tone in the second interval was perceived to the left or the right of that presented in the first interval. The stimuli had synchronous onset and offset gating in both ears, so that the SAM tones differed only in their ongoing interaural phase difference of the envelope (IPDENV). SAM tones had a carrier frequency fc of 4 kHz and were presented to the subject via Sennheiser HD-650 headphones at a SPL of 65 dB. Levels where calibrated using a sound-level meter and an Artificial Ear Type 4153 (Brüel & Kjær). Digital-analog conversion was carried out by ADI-s DAC FS (RME) with 32 bit and a 48-kHz sampling rate. Stimuli were generated digitally using the AFC-software package. The subjects were seated in a double-walled, sound-attenuating booth and responded by pressing a key on a standard computer keyboard. Visual feedback was provided after each trial. The 300-ms SAM tone duration included 20-ms cos2 rise-decay ramps. A 50-ms silent interval separated the two intervals. A next pair of intervals was presented 500 ms after the subject responded. The IPDENV of the stimuli presented in the two intervals were symmetrical around zero, so that in one of the two intervals, the right ear was leading, in the other the left ear led by the same IPDENV. Subjects could thus make their decision based on a ΔIPDENV (= 2 IPDENV used in the individual intervals) difference between the two intervals.

The experiment consisted of two phases. In the first phase (Figure 1), the threshold IPD-value was estimated for the baseline condition (with spectrally-flanking notched noise presented to both ears with no interaural correlation). An adaptive ‘four-down, one-up’ staircase procedure controlled the ΔIPDENV, meaning that the ΔIPDENV was decreased after four correct responses in a row and increased after each incorrect response, adapting towards a proportion correct of 0.84. This target rate of 0.84 provides a large result space for a deterioration before a floor is hit at chance level 0.5. Each adaptive track started at a ΔIPDENV of 0.4p radians. The maximum allowed ΔIPDENV value was 0.8p, because a decline of sensitivity is expected for higher values. With this upper limit for the adaptive tracking variable, a slightly lower upper bound of approximately 0.7p radians, is expected for the mean values. The initial step size was a factor of 2, which was reduced to 1.414 (21/2) and 1.189 (21/4) after the first and second ‘down-up-reversal’. An adaptive track was terminated after 10 reversals at the minimum step size. This sequence was repeated three times for each subject.

In the second part (Figure 2), the proportion correct for a fixed ΔIPDENV was measured in four different conditions: (1) in quiet, (2, 3) in spectrally-flanking notched noise at only the left or right headphone channel, and (4) in a control condition with the same interaurally uncorrelated notched noise as in the first part.

For the second, third, and fourth conditions, the spectrally-flanking noise was presented continuously, to minimize binaural interference. The ΔIPDENV was the geometric mean of the thresholds from the first part. The spectrally-flanking notched noise had a spectral level of -35 dB relative to the SAM tone. The notch was centered at 4 kHz and had a width of 800 Hz between 3600 and 4400 Hz (w = 20%). To avoid the use of low-frequency distortion products arising from nonlinear peripheral auditory processing, a continuous diotic noise, low passed at 1.3 kHz with a relative spectral level of g = -35 dB.Hz-1 was added in all conditions.