An audio-visual test was developed for ascertaining ability inrecognizing misarticulations. Two films, each with a different set of 23 words representing 23 initial and 20 final sounds, were produced. Each test included 138 consonant sounds in phrases, 276 in isolated words, and 144 in trios of words. An empirical “correct” answer sheet was prepared by two experienced clinicians after repeated viewings of the films. Six children with misarticulations presented stimuli in the films.
One-hundred-fifty subjects included speech clinicians with five or more years of experience, graduating senior majors in speech pathology, and experienced teachers. Each subject took three tests: Film A (audio-visual), Film B (audio-visual), and the sound-track of Film A presented without motion pictures (audio only). Correctly identified misarticulations, falsely identified misarticulations, and total correct responses were tabulated. Both forms of the audio-visual test were considered valid since the clinicians and students were significantly better than the teachers in identifying misarticulations. Satisfactory reliability was also established, since Films A and B were not significantly different in the accurate identification of sounds in words.
Performance was significantly better for audio-visual representation than for audio only; and identification of misarticulation was best with isolated words and worst with phrases. Experienced teachers did not identify 12 of the sounds as accurately as experienced clinicians.
SpecTalk: Conforming IoT Implementations to Sensor Specifications
