The study of speech acoustics dates back to the late 1800s with the invention of the phonograph (Pickett, 1999). In the 1940s and ‘50s, the spectrographic display gained attention at Bell Telephone Laboratories and the Detroit Day School for the Deaf for its potential usefulness for computerized speech synthesis and as a feedback tool for teaching speech to deaf students (Kent & Read, 1992; Potter, 1945, as cited in Stewart, Larkin & Houde, 1980). Limitations in resolution and cost of the early SD technologies led Stewart, Larkin, and Houde (1976) to develop the first SD device, the Visual Speech Training Aid, dedicated to the habilitation of speech with speakers with hearing-impairments.
They demonstrated its effectiveness with a case study of a college-aged deaf student. Since then, numerous other researchers have investigated the effectiveness of using SDs as a speech therapy tool for a variety of clinical populations and of various ages.
While Stewart, Larkin, and Houde (1976) were able to demonstrate successful application of SDs with one individual, its usefulness would depend on how easily spectrograms could truly be read. In a study conducted by Maki, Gustafson, Conklin, and Humphrey-Whitehead (1981), 10 adults with hearing-impairments who were familiar with SDs were able to successfully “decode” speech spectrograms, identify various visual features of speech (i.e., voice onset time), and evaluate various speech error patterns. The authors went further to delineate those error types that were most commonly interpreted incorrectly by the group, findings that will contribute to our discussion about target selection and clinical instruction (see below). It should be noted that the participants had already undergone at least 30 hours of speech therapy with SDs prior to the initiation of the study. Greene, Pisoni, and Carell (1984), in contrast, demonstrated that students wholly unfamiliar to speech science and SDs were able to read spectrograms at high accuracy rates with relatively little formal training and were able to generalize their learning to different speakers and different tokens. These results directly opposed previous contentions from Liberman, Copper, Shankweiler, and Studdert-Kennedy (1968) that visual displays of the acoustic signal were impractical therapy aids for speakers with hearing-impairments due to the abstractness and complexity of the display. Read more.