A Comparative Study of Unassisted and Audio-Assisted Repeated Reading for Improving EFL Learners' Reading Speed

Reading fluency, particularly reading speed, is crucial for academic success, especially at the university level. Laberge and Samuel's automaticity theory (AT) posits that fluent readers decode words automatically, freeing cognitive resources for comprehension. Slow readers, however, struggle with decoding, hindering comprehension. Samuels (1979) proposed repeated reading (RR) – UARR and AARR – to improve reading rates. While extensive L1 research supports RR's effectiveness, L2 research is limited and inconsistent due to methodological heterogeneity. This study aims to directly compare UARR and AARR's effects on EFL learners' reading speed, addressing prior research limitations such as varying sample sizes, group matching inconsistencies, dissimilar pre/post-tests, limited vocabulary overlap, and the impact of comprehension questions. The study seeks to provide a more robust comparison of the two methods and offer practical implications for EFL instruction.

The literature review examines existing research on both UARR and AARR in EFL contexts. UARR studies, often following Anderson's (1983) five-step procedure, show generally positive results but lack consistency due to methodological variations. Studies employing UARR have explored factors such as text matching to student levels, number of sessions, and the inclusion of comprehension assessments, with results ranging from significant to negligible gains. Similarly, AARR research, often building on Taguchi's work, demonstrates variable results. Studies using AARR have explored a variety of parameters, such as the number of repetitions per passage and the inclusion of comprehension questions, again leading to mixed results. Inconsistencies arise from differing methodologies, including sample sizes, group matching, and the inclusion of comprehension questions, potentially influencing reading speed due to students slowing down to understand the text.

This study employed a quantitative experimental design with a pre-test and post-test design. Participants (50 Vietnamese EFL learners) were divided into two groups: UARRG and AARRG (25 students each) using a matched ability nonrandom sampling approach, matched based on their pre-test scores, ensuring similarity in reading speed. The pre-test involved the first reading of a single text, which was specifically selected to be slightly below each students' current reading level in accordance with the i-minus-1 theory. The UARRG underwent a five-day treatment comprising 60 repeated readings of three passages following Anderson's (1983) and Baker's (2015) procedures. The AARRG also received 60 repeated readings over 12 days, adapting Gorsuch and Taguchi's (2010) method. To ensure comparability, both groups completed the same number of repeated readings. Instructions for both methods were provided in the participants’ native language (Vietnamese) to minimize any language related errors in administration. The pre-test and post-test assessed reading speed in words per minute (wpm). The same material was used for pre-test and post-test to better measure reading speed. A two-tailed independent-samples t-test examined differences between the groups' reading speed pre- and post-treatment. The study controlled for confounding variables by using a single text to minimize vocabulary differences and matching participants across groups, ensuring similar demographics.

The study reported that both UARRG and AARRG showed significant improvements in reading speed post-treatment. Statistical analysis revealed significant gains in reading speed for both groups. However, there was no statistically significant difference between the improvement in reading speeds seen in the two groups. This suggests that both UARR and AARR are effective methods for enhancing reading speed in EFL learners, and that the additional complexities of AARR might not translate to greater gains compared to the more straightforward UARR.

The findings suggest that both UARR and AARR are effective strategies for improving EFL learners' reading speed. The lack of a significant difference between the two methods might indicate that the audio assistance in AARR does not provide a substantial advantage over UARR in this context. This challenges some previous research which highlighted the potentially detrimental impact of audio accompaniment, particularly for learners who require more pauses. The consistent gains with both methods highlight the importance of repeated exposure to text as a key component of developing reading fluency. The study's rigorous methodology, using matched groups and controlling for potential confounders, enhances the reliability of the results, providing stronger evidence than many previous studies. These findings have implications for EFL instructors, suggesting that both UARR and AARR can be effectively integrated into reading instruction, with the choice potentially guided by factors such as resource availability and learner preferences.

This study provides strong evidence supporting the effectiveness of both UARR and AARR in improving EFL learners' reading speed. The lack of significant difference between the methods highlights the value of repeated reading in general. Future research could investigate the interaction between learner characteristics (e.g., prior reading experience, learning styles) and the effectiveness of each method. Exploring different text types and lengths and integrating comprehension tasks are also avenues for future research to better understand the effectiveness of RR in developing more holistic reading fluency.

While the study controlled for several variables, some limitations remain. The sample size, while larger than many previous studies, might not represent the full diversity of EFL learners. The study focused solely on reading speed; future studies might consider incorporating comprehension measures to provide a more comprehensive assessment of fluency. The use of a single text, while minimizing vocabulary issues, also limits the generalizability of the findings to other text types.