Informational masking of speech in children: auditory-visual integration

Abstract

The focus of this study was the release from informational masking that could be obtained in a speech task by viewing a video of the target talker. A closed-set speech recognition paradigm was used to measure informational masking in 23 children (ages 6-16 years) and 10 adults. An audio-only condition required attention to a monaural target speech message that was presented to the same ear with a time-synchronized distracter message. In an audiovisual condition, a synchronized video of the target talker was also presented to assess the release from informational masking that could be achieved by speechreading. Children required higher target/distracter ratios than adults to reach comparable performance levels in the audio-only condition, reflecting a greater extent of informational masking in these listeners. There was a monotonic age effect, such that even the children in the oldest age group (12-16.9 years) demonstrated performance somewhat poorer than adults. Older children and adults improved significantly in the audiovisual condition, producing a release from informational masking of 15 dB or more in some adult listeners. Audiovisual presentation produced no informational masking release for the youngest children. Across all ages, the benefit of a synchronized video was strongly associated with speechreading ability.

FIG. 1

Psychometric functions for individual listeners in the audio-only condition of the CRM speech recognition task. The four panels show data from listeners in the four different age groups. The different symbols in each panel represent different listeners. The mean psychometric function in each group is shown by the dashed line. The mean excludes data from the one “outlier” in the 6–9.9 years age group whose data are represented by open stars.

FIG. 2

Mean psychometric functions in the audio-only condition of the CRM task for listeners in each of the four age groups. Data from the current study (excluding that from the “outlier” shown in Fig. 1) are represented by open symbols and data from the previous study (Wightman and Kistler, 2005) by filled symbols. The error bars represent 95% confidence intervals for the mean.

FIG. 3

Individual psychometric functions from seven children tested with modulated noise as a distracter (audio-only) and retested with the speech distracter. The original data obtained with a speech distracter are labeled “Speech 1” and the retest data are labeled “Speech 2.” The bottom right panel shows all the individual modulated noise functions along with the data obtained in a similar condition by (Brungart, 2001b). Brungart's data are plotted as stars.

FIG. 4

Speechreading scores (video-only condition) for all listeners in the current study plotted as a function of age. The circled symbol represents the score for the young listener whose data were excluded from the mean computations shown in Figs. 1, 2, 5, and 6. The correlation listed in the inset is a Spearman r, computed with the data indicated by the circled symbol removed.

FIG. 5

Same as Fig. 1, except here the individual psychometric functions are from the audiovisual condition.

FIG. 6

Mean psychometric functions in the audio-only condition (open symbols) and the audiovisual condition (filled symbols) of the CRM task for listeners in each of the four age groups. Solid horizontal lines show mean performance in the video-only condition. The error bars and dashed lines represent 95% confidence intervals for the mean.

FIG. 7

A/V_Benefit scores for all listeners in the current study plotted as a function of age. The circled symbol represents the score for the young listener whose data were excluded from the mean computation shown in Figs. 1, 2, 5, and 6. The correlation listed in the inset is a Spearman r, computed with the data indicated by the circled symbol removed.

FIG. 8

Same as Fig. 5 except this figure includes only the data from the 7-years-old who performed better than the others in the same age group. This is the listener whose data were excluded from the mean computation shown in Figs. 1, 2, 5, and 6.