Sensitivity to temporal structure facilitates perceptual analysis of complex auditory scenes

Abstract

The notion that sensitivity to the statistical structure of the environment is pivotal to perception has recently garnered considerable attention. Here we investigated this issue in the context of hearing. Building on previous work (Sohoglu and Chait, 2016a; elife), stimuli were artificial 'soundscapes' populated by multiple (up to 14) simultaneous streams ('auditory objects') comprised of tone-pip sequences, each with a distinct frequency and pattern of amplitude modulation. Sequences were either temporally regular or random. We show that listeners' ability to detect abrupt appearance or disappearance of a stream is facilitated when scene streams were characterized by a temporally regular fluctuation pattern. The regularity of the changing stream as well as that of the background (non-changing) streams contribute independently to this effect. Remarkably, listeners benefit from regularity even when they are not consciously aware of it. These findings establish that perception of complex acoustic scenes relies on the availability of detailed representations of the regularities automatically extracted from multiple concurrent streams.

Keywords: Auditory scene analysis; Change deafness; Change detection; Predictive coding; Temporal regularity; Time perception.

Fig. 1

Example of the three variants (‘Change appear’, ‘Change disappear’, and ‘No change’) of a scene with 8 streams. Changing streams are indicated with arrows. Regular (REG) scenes are on the left, random (RAND) scenes on the right. The scenes are matched spectrally, with only temporal structure differing between REG and RAND scenes. The plots represent ‘auditory’ spectrograms, generated with a filter bank of 1/ERB wide channels equally spaced on a scale of ERB-rate. Channels are smoothed to obtain a temporal resolution similar to the Equivalent Rectangular Duration (Moore and Glasberg, 1983).

Fig. 2

Results of Experiment 1. Error bars are 1 standard error (SE). In all measures (d’ and response time) performance is significantly reduced in RAND relative to REG scenes. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

Fig. 3

Experiment 1B. [top] Schematic diagram of the loudspeaker array. [bottom] Results of Experiment 1b. Error bars are 1 standard error (SE). In all measures (d’ and response time) performance is significantly reduced in RAND relative to REG scenes.

Fig. 4

Experiments 2a and 2b. [A] Schematic representations of the regular patterns used. Scene streams in Experiment 1 (REG1) contained a fixed inter-tone-interval (T1) that was randomly chosen for each stream in each trial. Those in Experiment 2a (REG2) contained two different, regularly repeating, inter-tone-intervals (T1 and T2). T1 and T2 were randomly chosen for each stream in each trial. REG patterns in Experiment 2b (REG3) contained three different, regularly repeating, inter-tone-intervals (T1, T2 and T3). These were randomly chosen for each stream in each trial. [B] Results of Experiment 2a (Left) and Experiment 2b (Right) expressed in terms of d’ scores (top) and reaction times (bottom). Error bars are 1 standard error (SE). As in Experiment 1, performance is significantly increased in REG relative to RAND scenes.

Fig. 5

Results of Experiment 3. Appearance changes (CA; red colors) are on the left and disappearance changes (CD; blue colors) are on the right. REG context conditions are in darker colors; RAND context conditions are in lighter colors. REG changing-stream conditions are plotted with solid lines; RAND chaing-stream conditions are plotted with dashed lines. Error bars are 1 standard error (SE). The regularity of the changing stream as well as that of the context(non-changing) streams both contribute independently to the advantage of regularity (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 6

Experiment 4. [A] An Example of REG and RAND context scenes (left and right, respectively) with 4 streams. ‘Foil’ scenes (bottom) contain all REG or all RAND streams; ‘Target’ scenes (top) contain an odd stream – regular among random or vice versa, indicated with arrows. The plots represent ‘auditory’ spectrograms, generated with a filter bank of 1/ERB wide channels equally spaced on a scale of ERB-rate. Channels are smoothed to obtain a temporal resolution similar to the Equivalent Rectangular Duration. [B] Results of Experiment 4. The REG context condition is plotted with a solid line, the RAND context condition is plotted with a dashed line. The results demonstrate that it is consistently easier to detect a random stream among regular streams (REG context) than vice versa.