Effects of stimulus repetition on L2-accented English sentence recognition thresholds and psychometric function slopes

Characterizing second language (L2)-accented sentence recognition in noise using traditional methods requires large corpora of stimuli, which are not always available. Repeating sentences at ascending target-to-masker ratios (TMRs), instead of using novel sentences for each trial, reduces the number of sentences required for a given experiment. For native-accented speech, repeating sentences tends to elevate sentence recognition thresholds (SRTs) and flatten psychometric function slopes compared to using novel stimuli for each trial; poorer performance for repeated stimuli may be due to error perseveration, whereby incorrect recognition at low TMRs is retained as the TMR increases. The present study extends that work with native-accented speech to evaluate how using repeated versus novel sentences influences L2-accented speech recognition. Repeating L2-accented stimuli could be more prone to error preservation due to unfamiliar pronunciation or prosody. Twenty-two English monolingual adults with normal hearing were tested with sentences in three accent conditions: Midland (native), Korean, and Spanish. The masker was a speech-shaped noise. Tasks were blocked by the test method (novel or repeated stimuli), counterbalanced across participants, with accent conditions randomized within blocks. The effects of stimulus repetition and L2 accent on SRTs and slopes will be discussed.