Reliability of individual differences in lipreading

Publisher Website
Google Scholar
Abstract
Evaluation of the benefits of sensory aids requires stimulus materials whose psychometric characteristics are known. Generalizability theory was applied to data from two experiments to estimate measurement error arising from different materials, different talkers, and practice. Stimulus materials consisted of CID sentences and CV nonsense syllables produced by a male and a female talker and stored on video laserdisk. In experiment I, 104 normal-hearing subjects lipread the CID sentences in a single test session. Results indicate that there are substantial individual differences in lipreading performance among subjects, but that there are also systematic differences among sentences and among talkers that must be taken into account in interpreting test performance. In experiment II, 15 normal-hearing subjects who participated in an intensive training protocol involving vibrotactile supplements to lipreading were given pre- and posttests on the CID sentences and on the CV syllables. Individual differences in performance on sentences were highly stable over the training period, suggesting relatively uniform improvements due to training. Reliability over time was lower for the CV syllables and correlations between the CV and sentence materials were weak both at the pretest and posttest. [Work supported by NIH.]

This publication has 0 references indexed in Scilit: