Using Variability in Masker Level to Study the Decision Process in Forward-masked Intensity Discrimination

The decision process in a forward-masked intensity discrimination task was studied by introducing within-trial variability in masker level. In a 2IFC paradigm, the level of the masker presented in interval 1 and interval 2, respectively, was sampled independently from a normal distribution in each trial. Mean and standard deviation of the distribution were varied. Standard level was constant; the level increment was fixed in each block. Correlational analyses revealed different response strategies depending on masker level. With mean masker level equal to standard level, listeners tended to select the interval with the higher masker level, behaving like an energy detector. For mean masker level larger than standard level, three of the four listeners showed a negative correlation between the masker level in a given interval and the probability of responding that the increment had been presented in this very interval. This indicates a strategy of forming a contrast between masker loudness and target tone loudness and voting for the interval in which their difference was smaller. The weight assigned to masker level was larger for the intermediate mean masker level and increased with masker variability. Non-simultaneous masking produces a rather complex pattern of effects on intensity resolution. Difference limens for a midlevel standard are strongly elevated by an intense forward masker (80-100 dB SPL), relative to the jnd in quiet. However, the effect of the masker on jnd’s for standards presented at low and high levels, respectively, is rather small (e.g., Zeng, Turner, & Relkin, 1991), resulting in the midlevel hump in intensity discrimination. The midlevel hump is also found for backward maskers and contralaterally presented maskers (e.g., Plack, Carlyon, & Viemeister, 1995), which precludes mechanisms in the auditory periphery as the origin of the effect. In experiments in which the masker-standard level difference was varied while keeping the standard level constant, the jnd elevation caused by a forward masker was larger for intermediate than for large masker-standard level differences (middifference hump; Oberfeld, 2003, in press). These observations are evidence for the similarity model proposed by Oberfeld (2003, 2005), which assumes that the masker degrades or biases the memory representations of the target tones (Plack & Viemeister, 1992; Carlyon & Beveridge, 1993), and that the perceptual similarity of masker and standard is crucial for the effect. Maskers strongly differing from the standard in at least one dimension (e.g., loudness, duration) are assumed to have only a relatively small effect on the memory representations and thus on intensity resolution, so that the jnd elevation is a non-monotonic function of the masker-standard level difference. The model is compatible with the reduced midlevel humps observed if a masker differing from the standard in duration or spectral content is presented (Schlauch, Lanthier, & Neve, 1997). The mid-level hump can be viewed as a special case of the mid-difference hump because standard level and the masker-standard level difference are correlated if masker level is fixed at, e.g., 90-dB SPL. The present study for the first time examined not only the effects forward masking on intensity difference limens (“molar psychophysics”, Green, 1964), but also assessed the decision process by introducing within-trial variability in masker level and analyzing the trial-by-trial data (“molecular psychophysics”, Green, 1964). For the 2IFC task used, it was assumed that listeners integrate the level of masker and target tone in both of the two observation intervals, and base their decisions on the overall level in each interval. An equivalent description of the expected decision process is that a listener behaves as an energy detector, comparing the output level of a temporal window for the first interval and the output level of a temporal window for the second interval, and voting for the interval where the output level was larger (Plack & Oxenham, 1998). Concerning intensity discrimination in quiet, Jesteadt, Schairer, and Neff (2005) analyzed data from an experiment in which external variability was added by randomly varying pedestal level in each of the two observation intervals. The relation between the level of the tone in the interval containing the standard only and the interval containing the standard-plus-increment, respectively, and performance was compatible with the pattern an energy detector would produce. The above assumptions result in the hypothesis that the binary response (“Increment in interval 1” or “Increment in interval 2”) be correlated with the within-trial difference in masker level. For example, listeners should tend to respond “Increment in interval 2” if the masker presented in the second interval is higher in level than the masker in interval 1. In line with the predictions of the similarity model, a second hypothesis was that the effect of the masker levels presented in a given trial on the response be smaller in conditions where the masker was found to cause only a small deterioration in performance. Method Stimuli and Apparatus The standard and the masker were 30-ms, 1-kHz pure tones. A 2I, 2AFC procedure was used. In one of the two observation intervals (randomly selected), an increment was added in-phase to the standard. Standard level was 25 dB SPL. Listeners were tested in quiet and with a forward masker presented in both intervals. In each trial, the sound pressure level of the masker presented in interval 1 and interval 2, respectively, was sampled independently from a normal distribution. Mean masker level was 25, 55, or 85 dB SPL. The standard deviation (SD) was 0 dB (fixed masker level), 2 dB, or 6 dB. Masker level was limited to a range of ± 2.5 SDs. The silent interval between masker offset and standard onset was 100 ms. The interval between the offset of the first target tone and the onset of the second target tone was 650 ms. All stimuli were generated digitally, played back via one channel of an RME ADI/S D/A converter (fS = 44.1 kHz, 24-bit resolution), attenuated (TDT PA5), buffered (TDT HB7), and presented to the right ear via Sennheiser HDA 200 headphones.