The Effects of Time Expansion on English as a Second Language Individuals

John S. Novak, Daniel Bunn, Robert V. KenyonInterspeech 2019

TL;DR

This paper investigates whether giving English as a Second Language (ESL) learners real-time control over the playback speed of audio materials improves their listening comprehension. While learners prefer slower playback and actively use this control, the study found no significant improvement in their comprehension scores. This suggests a potential disconnect between perceived usefulness and actual performance enhancement.

What Problem Does This Paper Address?

Acquiring a second language heavily relies on listening comprehension, which can be particularly challenging for ESL learners. A common strategy to aid comprehension is "Clear Speech," where speakers naturally slow down and modify their speech. However, the effectiveness of these modifications, especially when applied artificially, is not fully understood. Previous research has shown mixed results on whether slowing down speech improves comprehension for non-native speakers. This paper addresses this by exploring whether fine-grained, user-controlled time expansion of audio can personalize the listening experience and, crucially, whether this personalization translates into measurable comprehension gains for ESL learners.

Methods

The researchers developed custom software that allowed participants to control the playback speed of audio clips using a slider interface. This "phase vocoder" technique allowed for time expansion (slowing down) without altering the pitch of the audio. The expansion range was from 1.0 to 2.5 times the original length (effectively slowing playback to 40% to 100% of original speed).

The experiment involved ESL participants listening to audio tracks (fictitious lectures and conversations) from official TOEFL iBT tests. For some trials ("treatment trials"), participants could adjust the playback speed, while for others ("control trials"), the speed was fixed. After each audio clip, participants completed a multiple-choice quiz. The software recorded playback speed adjustments, quiz scores, and other user data. A Likert scale survey and a free-form interview were also conducted to gather subjective feedback.

Datasets

The study utilized audio tracks and quizzes from the Official TOEFL iBT Tests, Vol. 1. Nine audio tracks were selected, comprising five fictitious lectures and four fictitious conversations. These were converted to WAV audio format, and their corresponding quizzes were transcribed. No new datasets were introduced.

Key Insights and Ideas

  • Learner Preference vs. Performance: The most striking insight is the divergence between what ESL learners prefer and what demonstrably improves their comprehension. Learners actively sought slower playback, indicating a perceived benefit, yet their objective test scores did not improve.
  • Personalization is Used, But Not Always Effective: The fine-grained control allowed for personalization, and a significant majority of participants used this feature. However, this personalized experience didn't lead to better comprehension outcomes.
  • Individual Differences Matter (and are Complex): The study hints at individual differences in how ESL learners benefit from time expansion, with a modest correlation observed between TOEFL scores and the degree of expansion used (lower scores tended to use more expansion). This suggests that while some personalization exists, it's not a simple one-size-fits-all solution.
  • Subjective Utility Doesn't Guarantee Objective Gains: Participants generally found audio expansion useful and the interface easy to use, but this positive subjective experience did not translate into improved quiz scores. This highlights the challenge of designing interventions based solely on user feedback.

Key Results

  • High Usage of Time Expansion: 88.5% of subjects used audio expansion in at least one of their three treatment trials, and 65.5% used it in all three. The average modal expansion across all treatment trials was 1.11 (11% added playback time).
  • No Significant Comprehension Improvement: A Kruskal-Wallis non-parametric test found no statistically significant difference between control and treatment quiz scores (p > 0.05). The mean control score was 0.70 (SD=0.19), and the mean treatment score was 0.65 (SD=0.15).
  • Modest Correlation with TOEFL Scores: A linear regression showed a mild reduction in audio expansion use as TOEFL scores increased (R²=0.15). A 10-point increase in TOEFL score was associated with a 2.4% reduction in playback time.
  • High Subjective Utility: 81% of subjects agreed or strongly agreed that audio expansion was useful for listening skill in general, and 61% found it useful for the specific audio files. The interface was rated as easy to use by 88%.
  • Recurring Subject Concerns: Interview data revealed that participants struggled with the length of passages, vocabulary difficulties (which expansion didn't solve), and found lectures more challenging than conversations.

Limitations and Future Work

Limitations: * The study used a linear time expansion method, which may not fully replicate natural "Clear Speech" modifications. * The sample size, while reasonable for an experimental study, might limit the generalizability of findings. * The TOEFL scores were self-reported, which could introduce some variability.

Future Work: The authors suggest that future research could explore:

  • Investigating the physiological cognitive load associated with audio expansion to see if it reveals a decreased load despite no comprehension gain.
  • Exploring more sophisticated audio modification techniques that better mimic natural Clear Speech.
  • Further examining the individual differences that might predict who benefits from time expansion.
  • Investigating the impact of vocabulary difficulty, which appeared to be a significant barrier for participants.

Summary updated

BibTeX
@inproceedings{novak2019effects,
  author = {John S. Novak and Daniel Bunn and Robert V. Kenyon},
  title = {{The Effects of Time Expansion on English as a Second Language Individuals}},
  booktitle = {Interspeech},
  year = {2019},
}
Figure