The present study explores the relationship between temporal fluency and second language (L2) learners' self-assessed and externally assessed proficiency level across the target languages French, Spanish, and Portuguese. Temporal fluency is operationalized as articulation rate (the speed of speech), the mean length of utterance, and the silent pause frequency. Participants (n = 283) in this study are native English speakers enrolled in upper-division language courses at Brigham Young University (BYU). Each participant completed both a self-assessment (the Language Ability Self-Evaluation Resource or LASER) as well as the Computerized Oral Proficiency Interview (OPIc) to receive an external proficiency assessment. Temporal fluency measures were automatically extracted from audio recordings used for the self-assessment (LASER) using a Praat script developed by De Jong & Wempe (2009). The results from this study find a strong, significant relationship between all temporal fluency measures and both self-assessed and OPIc-assessed proficiency. This relationship remains strong when comparing across L2 groups for all three temporal fluency measures and OPIc proficiency rating. For self-assessed proficiency, only some temporal fluency measures are found to be significantly related to self-assessed proficiency across each target language group, with articulation rate found to be the most consistently significant temporal fluency measure across all three target language groups. Together, these findings suggest that as speakers speak faster, produce longer utterances, and pause less frequently, they also tend to increase in proficiency, both according to their own self-assessment as well as within a formal oral proficiency assessment (OPIc).

This study investigates the interplay between temporal fluency measures, self-assessment, and language proficiency scores in novice- to intermediate- level language learners of Spanish and French. Analyzing data from 163 participants, the research employs both traditional linear regression and advanced XGBoost machine learning models. Findings demonstrate a moderate positive correlation between self-assessment and Oral Proficiency Interview by Computer (OPIc) scores, underscoring the dependable self-awareness of learners. Notably, XGBoost performs as well as linear regression in predicting OPIc scores and has more potential, underlining the efficacy of advanced methodologies. The study identifies Mean Length of Utterance (MLU) as a crucial predictor, highlighting specific temporal fluency measures' significance in determining proficiency. These findings contribute to language assessment practices, advocating for the integration of machine learning for enhanced precision in predicting language proficiency and informing tailored instructional approaches.