Recognition of digitized CV syllables in multitalker babble.

Nonsense syllable recognition by 10 normal-hearing listeners was assessed in quiet and three levels of multitalker babble. Stimuli were 19 consonants paired with /a/, /i/, and /u/ in a CV format. Performance measures included total syllable recognition, consonant feature-class recognition, and consonant feature errors. Results of each analysis were strongly affected by vowel coarticulation and noise level. However, a distinctive pattern of performance in the babble was observed. Comparisons are made between this pattern and those obtained previously in other types of noise backgrounds. Implications for clinical evaluation and auditory training are discussed.