English speech recognition from phonemes, words, to sentences in quiet and noise for second language learners

The goals of this study were to investigate the recognition of English speech stimuli from phonetic, to word, and sentence level in quiet and long-term speech-shaped (LTSS) noise with and without temporal modulation for 60 Chinese college students and whether speech recognition across these speech materials was correlated to each other, e.g., whether students with better phonetic identification also showed better sentence recognition. Twelve isolated English vowels, twenty-one English consonants in the /aCa/ context, NU6 words, the coordinate response measure (CRM), and IEEE sentences served as speech stimuli. Overall, across the five English speech materials, Chinese students’ performance followed the order: CRM > consonant > NU6 > vowel > IEEE. In quiet, speech recognition scores were significantly correlated with each other, and particularly sentence recognition was significantly related to both vowel and consonant identification, indicating the importance of phonetic processing in sentence recognition in quiet. However, the significant correlation between sentence recognition and phonetic identification was not presented in noise with low signal-to-noise ratios, suggesting that listeners may rely more on top-down processing like linguistic cues in sentence recognition in challenging listening conditions.