Determining Optimal Talker Variability for Nonnative Speech Training: A Systematic Review and Bayesian Network...

Area of Science:

• Speech-language pathology
• Second language acquisition
• Educational psychology

Background:

• Talker variability in nonnative speech training is crucial for improving pronunciation and intelligibility.
• Previous research has yielded mixed results on the optimal level of talker variability.
• A systematic approach is needed to synthesize existing evidence.

Purpose of the Study:

• To determine the optimal level of talker variability in training to maximize second-language speech learning.
• To compare different talker variability training conditions using a meta-analysis.

Main Methods:

• Systematic search of studies on talker variability in nonnative speech training.
• Bayesian network meta-analysis of 32 studies with 998 participants.
• Ranking of six training conditions by effectiveness using surface under the cumulative ranking curve (SUCRA) values.

Main Results:

• Training with six talkers was most effective (SUCRA = 94%), but conditions with four or two talkers, depending on presentation format, also showed high effectiveness.
• Moderate talker variability, considering both number and presentation format, demonstrated significant improvements in second-language speech outcomes.
• The analysis revealed moderate between-study heterogeneity.

Conclusions:

• Optimizing talker variability in nonnative speech training requires balancing the number of talkers and their presentation format.
• A moderate level of talker variability is most effective for enhancing second-language speech training outcomes.
• This study provides evidence-based recommendations for designing effective speech training programs.