Fig. 7 | Nature Communications

Fig. 7

From: Illusory sound texture reveals multi-second statistical completion in auditory scene analysis

Fig. 7

Experiment 6—statistical integration of illusory texture. a Step experiment paradigm. At the start of each trial block, participants were familiarized with a reference texture (synthesized from the statistics of a real-world texture recording) and a mean texture (synthesized from the average statistics of 50 real-world textures). On each trial participants heard two stimuli and judged which was most similar to the reference, basing their judgments on the endpoint of each stimulus. The first stimulus contained a subtle “step” in statistics, the direction and temporal position of which varied from trial to trial. The second stimulus (“morph”) had constant statistics in a given trial, but varied between the reference and mean across trials. b Main conditions and interpretation. Previous experiments suggested that humans estimate texture statistics using a temporal integration window of several seconds (symbolized in gray). A step occurring within this window should bias judgments away from the endpoint. In the critical condition (third panel), the middle of the step stimulus was replaced with 2 s of Gaussian masking noise. This was expected to elicit an illusory texture during the presumptive integration window that could bias judgments away from the endpoint if the illusory texture replicated the effects of an actual texture in the relevant part of the auditory system. The last condition contained a 200 ms silent gap before the noise. This was a control condition to confirm that eliminating the percept of illusory texture would eliminate any bias observed in the third condition. c Trial blocking. The experiment was divided into blocks of 10 trials related to a particular reference texture. The step condition and direction (towards or away from the reference) varied from trial to trial (randomly ordered over the course of the experiment subject to the blocking constraint). The 10 morph positions were presented once per block in random order. d Results of Experiment 6. Shaded regions show SEM of individual data points and curves plot logistic function fits. e Plot of the bias produced by the step in each condition, quantified as the difference between points of subjective equality for the upward and downward step conditions. Error bars show SEM, obtained by bootstrap (10,000 samples)

Back to article page