Content area
Abstract
This study evaluated the effectiveness of two types of categorical ratings used as supplements to the baseline speech identification task. The ratings of: (1) labeling-response confidence; and (2) speech quality ratings each were solicited, in separate protocols, as follow-ups to the consonant labeling responses. The stimuli were an 8-member continuum of "DAE" to "TAE", developed via digital adjustment of a spoken syllable. This stimulus continuum was known to contain some generally ambiguous and some similarly-labeled items. Thirty-two normal-hearing adults supplied first, the consonant identification label, then the rating category. Consonant labeling latencies also were collected. Group- and individual-level analyses of the data were performed. Participant background characteristics and listeners' procedure preference feedback were obtained.
Consonant labeling functions did not vary with the inclusion of the ratings solicitations, though the analysis of labeling latencies reflected an apparent difference in the process of labeling in the presence versus absence of the ratings. The confidence and the quality ratings were reliable and reasonably consistent. Labeling latencies, while more informative than the labeling results alone, generally were not as sensitive a supplemental measure as were either the confidence or the quality ratings. While with the selected combination of stimuli and listeners, the quality ratings appeared to be more sensitive, listener preference for the confidence rating procedure over the quality rating procedure was overwhelming. The solicitation of the ratings seemed to promote an earnest engagement of the listener in the perception evaluation process, as evidenced by the response latency data as well as by listener responses to the exit survey items. Professional experience or training in speech and hearing may be an advantage for utilizing the quality ratings to distinguish among similarly-labeled speech stimuli. Considerable inter-listener differences with respect to ratings use were found among this normal-hearing listener group. The use of individual and group-level analyses provided the opportunity to explore the possible influence of more-idiosyncratic, personality-driven, or strategic rating response behaviors.