Sample discrimination of frequency differences with irrelevant context–revisited

Abstract

This experiment replicated earlier work to provide additional data for post hoc analyses of perceptual weights for conditions with large effects of context on performance. The non-adaptive, 2AFC task was sample discrimination of frequency differences (SD-F) in quiet and with added context stimuli. Listeners were to select the interval in which a pair of target tones was drawn from the higher of two Gaussian frequency distributions (means of 2000 and 2150 Hz). Pairs of context stimuli were added at frequency regions above and below the targets, at distances of 600, 1000 and 1400 Hz. Context stimuli were fixed-frequency tones, noise bands, or random-frequency tones. All stimuli were 100 ms with 5-ms ramps, presented simultaneously. Conditions were tested with and without Gaussian level jitter (standard deviation=3 dB). Mean levels of target and context stimuli were equated or systematically varied across conditions. The results confirm earlier work, with little effect of fixed-frequency tones or noise-band context at any distance from the targets. Random-frequency context tones produced large detrimental effects even at remote distances, due predominantly to effects of the lower-frequency context tone. Level variation had little effect. Different approaches to calculating perceptual weights influences data interpretation. [Work supported by NIDCD.]