The role of clinical experience in listening for covert contrasts in children’s speech.

University of Minnesota M.A. thesis. June 2010. Major: Speech-Language Pathology. Advisor: Benjamin Munson, Ph.D. 1 computer file (PDF); vii, 61 pages, appendices A-B.Children acquire speech sounds gradually. This gradual acquisition is reflected in numerous aspects of speech-sound development, from an infant’s ability to distinguish between sounds that have slight variations to the production of sounds that are identifiably adult-like. Evidence of gradual acquisition is seen in acoustic studies of children's speech-sound production, many of which have shown that children develop contrasts in certain speech sounds gradually and produce intermediate stages as they progress from incorrect to correct productions. It has also been shown that adults can perceive these fine differences in young children’s speech. This study examined whether experienced speech-language pathologists perceive children's consonants differently from untrained listeners. The stimuli sets consisted of /t/-/k/ (88 tokens), /s/-/q/ (200 tokens), and /d/-/g/ (135 tokens). Forty-two participants (21 experienced speech-language clinicians and 21 non-clinician undergraduate students) heard consonant-vowel syllables truncated from words produced by children ages two through five. Listeners were asked to provide a rating of the beginning target sound using a visual-analog scale (VAS), which contained a double-headed arrow labeled with the target sound on each side. For example, one end was labeled “the ‘t’ sound” and the other end was labeled “the ‘k’ sound.” The rating involved clicking on the line at a location that represented the token’s proximity to an ideal /t/ or /k/. The participants’ click locations on the VAS line are strongly correlated with the acoustic parameters that differentiate between the endpoint categories for a variety of contrasts for the stimuli sets. Results indicated three main differences between the way clinicians and laypeople perceived the stimuli. First, clinicians were more willing to click closer to the ends of each scale, indicating that a token was closer to a perfect representation of the target sound; second, clinicians had higher intra-rater reliability than the naïve listeners; and third, clinicians showed a tighter relationship between the acoustic properties and the VAS ratings than laypeopl