言語的符号化が標的音声の話者同定に与える影響

抄録

This study explored the effects of the similarity of verbalized information between target and distractor voices on speaker identification. Participants listened to the target voice, and either described it using one of three catego ries (“adjective-selection,” “adjective-rating,” or “free-description”) or did not describe it (Control). They then rat ed “the degree to which the voice sounded like the target person” for three types of voice stimuli: the target voice, a high-similarity voice (verbalized information was similar to the target voice), or a low-similarity voice (verbal ized information was dissimilar to the target voice). Results showed that “the degree to which the voice sounded like the target person” in the free-description category was significantly higher than in the control group in the high-similarity voice condition. These results suggest that verbalization leads participants to misidentify the high-similarity voice as the target person. Therefore, a verbal overshadowing effect in previous studies may have occurred because verbalization caused participants to perceive the high-similarity voice as the target person as much as the target voice, leading them to choose the high-similarity voice.