Cited by 17Subtitles and Audiovisual MediaEnglish Language Learning and TeachingSecond Language Acquisition and Learning

The DOI matches its Crossref record; 31 of the 31 references it cites were found in a registry or an index; none of the 21 sources its claims cite is checked for support yet.

The effects of captions on L2 learners’ comprehension of vlogs

Open access
Aldukhayel, Dukhayel·2021·Language Learning & Technology

Abstract

This study investigated the effects of captions on the listening comprehension of vlogs. A total of 96 EFL learners watched three vlogs under one of three conditions: L2 captions, L1 captions, and no captions. Each group included low-, mid-, and high-level proficiency learners. The vlogs differed in the pictorial support of the audio, with Vlog 1 being highly supported, Vlog 2 being partially supported, and Vlog 3 being slightly supported by pictorial images. After each vlog, the participants took a multiple-choice test measuring their comprehension of details. Afterwards, participants completed a questionnaire about their perception of captions. The findings suggest that the availability of captions may not necessarily lead to better listening comprehension because students, particularly lower proficiency learners, were unable to simultaneously process the multiple modalities (images, audio, and captions) due to their limited capacities of working memory and cognitive load. High-proficiency learners achieved better comprehension than low- and mid-proficiency learners and achieved their best comprehension with L2 captions. A significant increase in comprehension of vlogs caused by high pictorial support was detected, with the inverse relationship also being true. Analysis of the questionnaire indicated that participants consider L2 captions useful. For both L2 and L1 captions, students think that their listening comprehension would decrease without captions. When considering vlogs for L2 listening, language proficiency and pictorial support are better indicators of levels of comprehension. Captions might be beneficial when learners’ proficiency level is high. When visual images are highly supportive for the audio, better comprehension of vlogs is likely.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

This study found that L2 and L1 captions did not significantly improve Arab EFL learners’ listening comprehension of vlogs compared with no captions. Proficiency level and pictorial support were stronger predictors, with high-proficiency learners benefiting most from L2 captions and comprehension declining as pictorial support decreased.

Key findings

  • Caption condition produced no significant differences in comprehension across the three vlogs: MANOVA Wilks’ Λ = 0.98, F(6, 182) = .32, p = .93, multivariate η² = .010.
  • Proficiency level significantly affected comprehension: Wilks’ Λ = 0.27, F(6, 182) = 28.32, p < .001, η² = .48; high-proficiency learners scored higher than mid-proficiency learners, who scored higher than low-proficiency learners. For high-proficiency learners, L2 captions produced significantly higher scores than L1 captions or no captions (F(2, 18) = 18.036, p < .001), with no significant difference between L1 captions and no captions.
  • Pictorial support significantly affected comprehension: F(1.86, 176.80) = 81.03, p < .001, R² = .46, η² = .46. Comprehension was highest for Vlog 1 (M = 7.82, SD = 2.02), lower for Vlog 2 (M = 5.77, SD = 2.33), and lowest for Vlog 3 (M = 5.16, SD = 1.88).
  • Questionnaire results showed the L2 captions group rated L2 captions as relatively useful (average scores higher than 3.80 on a 5-point scale); both caption groups agreed their comprehension would be negatively affected without captions, and the no-captions group agreed comprehension would be better with L2 captions.

Methods

  • Ninety-six Arabic-speaking male university EFL learners (Mage = 17.38 years, SD = .64) were assigned by classroom to L2 captions (n = 35), L1 captions (n = 32), or no captions (n = 29). Proficiency was measured with the Vocabulary Levels Test at the 2,000-, 3,000-, and 5,000-word levels and categorized as low, mid, or high.
  • Participants watched three YouTube vlogs twice under their assigned caption condition; the vlogs differed in pictorial support (Vlog 1 high, Vlog 2 partial, Vlog 3 slight). After each vlog, they completed a 10-item multiple-choice test measuring comprehension of details, followed by a questionnaire on caption perceptions.
  • Analyses included MANOVA for caption and proficiency effects, follow-up ANOVAs and Tukey HSD post hoc tests, and a repeated measures ANOVA with Greenhouse-Geisser correction for pictorial support.

Limitations

  • The paper states that keyword captions were not explored; comprehension was measured only with multiple-choice detail tests rather than global comprehension or inferencing; only three vlogs were used; only male students participated; and the sample was Arabic-speaking EFL learners, so findings cannot be generalized to other L1 groups or non-English vlogs.
  • Evident from the text: the study used a controlled laboratory procedure in which learners could not change speed or replay vlogs, which may limit ecological validity for typical vlog viewing.
  • Evident from the text: some proficiency-by-treatment cells were small (e.g., high-proficiency learners totaled n = 21 across three groups), which may affect the stability of subgroup analyses.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.64152/10125/73439

Cites

31 of 31 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

26 Own result · 14 Own method · 12 Own interpretation · 6 Sources not found · 5 No full text of the sources · 4 Not checked yet · 2 No basis given

Verify all 1 references — free account
  • It may state two separate claims as one. 15
  • Probably not a substantial claim. 20
  • The quote only partly states this claim. 1
31 of 31
  • The paper relies on this without support.

    The questionnaire responses accurately reflect participants' perceptions of captions.

    “Participants completed a mini questionnaire consisting of a five-point Likert scale, including statements on the usefulness of L2 and L1 captions (e.g., ''My comprehension improved because of the English/Arabic captions'') and questions about their caption needs (e.g., ''My comprehension would be better with the Arabic/English captions'').”
  • The paper relies on this without support.

    The difficulty factors (accent, pronunciation, speaker speed) did not differ between Vlog 2 and Vlog 1.

    “Yet, while this might be true for Vlog 3, what needs to be emphasized is that the difficulty of those factors were not different between Vlog 2 and Vlog 1.”
2 of 2

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • ConsistentF(2, 93) = 0.248reported p = .78 · recomputed p = .781

    “F(2, 93) = .248,p = .78”

  • ConsistentF(6, 182) = 0.32reported p = .93 · recomputed p = .926

    “F(6, 182) = .32,p = .93”

  • ConsistentF(6, 182) = 28.32reported p < .001 · recomputed p < .001

    “F(6, 182) = 28.32,p < .001”

  • ConsistentF(2, 27) = 1.7reported p = .203 · recomputed p = .202

    “F(2, 27) = 1.70, p = .203”

  • ConsistentF(2, 42) = 3.207reported p = .051 · recomputed p = .051

    “F(2, 42) = 3.207, p = .051”

  • ConsistentF(2, 18) = 18.036reported p < .001 · recomputed p < .001

    “F(2, 18) = 18.036, p < .001”

  • ConsistentF(1.86, 176.8) = 81.03reported p < .001 · recomputed p < .001

    “F(1.86, 176.80) = 81.03,p < .001”

7 of 7

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 19 references

Research works printed but not linked: 1

  1. Werner, E. A. (2012). Rants, reactions, and other rhetorics: Genres of the YouTube vlog [Unpublished doctoral dissertation]. University of North Carolina at Chapel Hill.

Cited by

Sorted by citations

See all 17

Metadata

Authors
Aldukhayel, D.
Affiliations
Qassim University (SA)
Corresponding author
Aldukhayel, Dukhayel
Venue
Language Learning & Technology
Published
2021-06-01
Volume
25
Issue
2
Pages
178-191
ISSN
1094-3501 (ISSN-L)
OpenAlex
W3192002680
Cited by
17
References
19
Citations by year
2026: 3 · 2025: 2 · 2024: 4 · 2023: 6 · 2022: 2
FWCI
9.43
Type
journal article
Language
en
Open access
Locations
  • Language learning & technology (publishedVersion, cc-by-nc-nd)
  • ScholarSpace (University of Hawaii at Manoa) (submittedVersion)
  • Language learning & technology (cc-by-nc-nd)
Topics
  • Social Sciences › Arts and Humanities › Language and Linguistics › Subtitles and Audiovisual Media
  • Social Sciences › Education › English Language Learning and Teaching
  • Social Sciences › Psychology › Developmental and Educational Psychology › Second Language Acquisition and Learning
Keywords
captions, vlogs, listening comprehension, second language learners, EFL learners, pictorial support, working memory, cognitive load, language proficiency, multimodal processing, video-based language learning

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review