Effect of speech dialect on speech naturalness ratings: a systematic replication of Martin, Haroldson, and Triden (1984).

This study investigated the effect of speech dialect on listeners' speech naturalness ratings by systematically replicating Martin, Haroldson, and Triden's (1984) study using three groups of speaker samples. Two groups consisted of speakers with General American dialect--one with persons who stutter and the other with persons who do not stutter. The third group also consisted of speakers who do not stutter but who spoke non-General American dialect. The results showed that speech naturalness ratings distinguished among the three speaker groups. The variables that appeared to influence speech naturalness ratings were type of dialect, speech fluency, and speaking rate, though they differed across speaker groups. The findings also suggested that strength of speech dialect may be a scaleable dimension that judges can rate with acceptable levels of reliability. Dialect may also be an important factor that needs to be incorporated or controlled within systems designed to train speech naturalness ratings. It may also be an important factor in determining the extent to which stuttering treatment produces natural sounding speech.