Reliability of judgments of stuttering-related variables: The effect of language familiarity

Abstract

Previous studies demonstrate mixed results and some methodological limitations regarding judges’ ability to reliably assess stuttering-related variables in an unfamiliar language. The present study examined intra- and inter-rater reliability for percent syllables stuttered (%SS), stuttering severity (SEV), syllables per minute (SPM), and speech naturalness (NAT) when English-speaking judges viewed speech samples in English and in a language with which they had no or minimal familiarity (Spanish). Over two time periods, 21 judges viewed eight videos of four bilingual persons who stutter. Data were analyzed for relative and absolute intra- and inter-rater reliability as well as for an effect of language on time period differences. Intra- and inter-rater relative reliability were good or excellent for all measures in both languages, with the exception of inter-rater relative reliability for NAT in both languages and %SS in Spanish. Intra-rater absolute reliability was acceptable in both languages for NAT and SEV and unacceptable in both for SPM and %SS. Inter-rater absolute reliability in both languages was unacceptable for all measures, even with judges with the same training. There was a clinically significant effect of language on %SS scores, but, despite a statistically significant effect of language for SPM and SEV, the differences were not clinically significant. Results indicate that reliability across and within languages varies by measure and is impacted by intra- vs. inter-rater reliability, relative vs. absolute reliability, and language familiarity. Modifications in training may be able to address some of the limitations found, particularly with regard to SPM and NAT.