Journal Article10.1017/s0142716423000346
Second language speech comprehensibility and acceptability in academic settings: Listener perceptions and speech stream influences
TL;DR: Comprehensible L2 speech is generally perceived as acceptable in academic settings. Speech stream characteristics have a significant impact on judgments of both comprehensibility and acceptability.
read more
Abstract: Abstract Ideally, comprehensible second language (L2) speech would be seen as acceptable speech. However, the association between these dimensions is underexplored. To investigate the relationship between comprehensibility and “academic acceptability,” defined here as how well a speaker could meet the demands of a given role in an academic setting, 204 university stakeholders judged L2 speech samples elicited from a standardized English test used for university admissions. Four tasks from 100 speakers were coded for 13 speech stream characteristics. Judgments for comprehensibility and acceptability correlated strongly (r = .93). Linear mixed-effects models, used to examine judgments across all tasks and separately for each task, indicated that while random intercepts (i.e., speaker ability, listener severity) explained a substantial amount of total variation (32–44%) in listener judgments compared to speech characteristic fixed effects (8–21%), fixed effects did account for variation in speaker random effects (reducing variation compared to intercept-only models by 50–90%). Despite some minimal differences across task types, the influence of speech characteristics across both judgments was mostly similar. While providing evidence that comprehensible speech can indeed be perceived as acceptable, this study also provides evidence that speakers demonstrate both consistent and less consistent performance, in reference to speech stream production, across performances.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Figures

Table 4. Intraclass correlations of speech judgments 
Table 3. Pearson’s correlations among comprehensibility and acceptability items (scores averaged across listeners) 
Table 7. Variance explained (R2) and speaker variance for task-specific comprehensibility and acceptability models 
Table 2. Intercoder agreement for hand-coded speech variables 
Figure 2. Speech judgment questions and interface. 
Figure 3. Correlations and distributions of study variables.
Citations
Transdisciplinary Intersections in Second Language Pronunciation Learning and Teaching
Okim Kang,Shelley Staples +1 more
TL;DR: This study explores transdisciplinary intersections in second language pronunciation learning and teaching, drawing from fields like corpus linguistics, pragmatics, and technology to inform advancements in L2 pronunciation instruction and assessment, particularly with AI and ASR.
References
•Book
Hierarchical Linear Models: Applications and Data Analysis Methods
James A. Calvin
- 03 Mar 1992
TL;DR: The Logic of Hierarchical Linear Models (LMLM) as discussed by the authors is a general framework for estimating and hypothesis testing for hierarchical linear models, and it has been used in many applications.
24.4K
Forming inferences about some intraclass correlation coefficients.
Kenneth O. McGraw,S. P. Wong +1 more
TL;DR: A review of the distinction between various forms of intraclass correlation coefficients (ICC) can be found in this article, followed by a discussion of the relationship between the two types of ICCs.
6.6K
Hierarchical Linear Models: Applications and Data Analysis Methods.
Harvey Goldstein,Anthony S. Bryk,Stephen W. Raudenbush +2 more
3.1K
Computing inter-rater reliability and its variance in the presence of high agreement
TL;DR: This paper explores the origin of these limitations, and introduces an alternative and more stable agreement coefficient referred to as the AC1 coefficient, and proposes new variance estimators for the multiple-rater generalized pi and AC1 statistics, whose validity does not depend upon the hypothesis of independence between raters.
1.6K