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

• Published in: Proceedings of meetings on acoustics Vol. 56; no. 1
• Main Authors: Cowan, Tiana Marie, Leibold, Lori, Buss, Emily
• Format: Conference Proceeding
• Published: 18.05.2025
• ISSN: 1939-800X
• DOI: 10.1121/2.0002088

Traditional methods for estimating the masked speech reception threshold (SRT) for open-set sentence recognition entails presenting novel sentences at a family of different target-to-masker ratios (TMRs). Obtaining an accurate estimate of SRT using these methods often requires a large number of unique stimuli (e.g., >30 per estimate). An alternative approach was recently proposed which repeats each sentence at ascending TMRs, reducing the number of unique stimuli required to estimate SRT. Existing data obtained with speech produced in the talker’s first (L1)-language indicate similar results for these two methods, with a slightly (0.5 dB) higher SRTs for repeated than novel stimuli. The present study was designed to see whether this result would generalize to speech produced by second language (L2) talkers. Stimuli were English sentences produced with a Midland English, Spanish, or Korean accent. In this dataset, repeating sentences led to significantly higher SRTs and shallower psychometric function slopes than the traditional procedures using novel stimuli for each trial, although both of these effects were small in magnitude. These findings demonstrate that repeating stimuli at ascending TMRs is a practical threshold estimation procedure, especially when materials are limited.