Do captions help students learning English? What a 2013 review of studies found, and where it stops

Do captions help students who are learning English, or do they give them one more thing to read while they are already working hard to listen? It's a fair worry. A student still learning English has to split fast speech into separate words, and a caption bar adds a second line that keeps moving whether the student has caught up or not.

Language researchers have studied captions on video since the 1980s. In 2013, three of them published a meta-analysis of that work, and a newer review calls theirs the first to establish that captions help with listening and vocabulary. This post walks through what the 2013 review found and where it stops helping a teacher in front of a live class.

What a meta-analysis is

A meta-analysis is a study of studies. Researchers collect the experiments that ask the same question, convert each result to a common scale, and combine them, so one small or unusual study has less pull on the answer. That scale is usually an effect size: how far apart two groups finished, measured in standard deviations, which is the typical spread of scores around a group's average. An effect size of 0.5 puts one group's average half a standard deviation above the other's.

What the 2013 review found

Maribel Montero Perez, Wim Van Den Noortgate and Piet Desmet gathered 18 studies of captioned video, meaning on-screen text in the same language as the audio, such as English text on an English video. The team ran one analysis on listening comprehension, using 15 of the studies, and another on vocabulary, using 10. In a 2026 review, Dru Sutton and Stuart Webb say the comparison group in those studies watched the same video with no on-screen text, so the 2013 review can't tell you how subtitles in a student's home language compare.

In both analyses, learners who watched with captions did better than those who watched without, and the authors call both effects large. The newer review reports the vocabulary effect size as 0.87, which leaves the average captioned viewer close to nine-tenths of a standard deviation ahead of the average viewer without captions.

Two details matter for teachers. The type of test changed how big the comprehension effect looked, so the benefit depended partly on how researchers measured understanding. And the learners' level made no significant difference to either result, although the newer review counts only one to three studies behind each proficiency level in that vocabulary analysis, too few to settle the question.

Who was in the studies

Most of these experiments follow the same design: learners watch a recorded video, some with captions and some without, then take a test on what they understood or which words they picked up. The newer review I come back to below shows who those learners tend to be: of the 48 experiments in its school-level comparison, 32 ran with university students, 10 with secondary school students and 6 with primary school students. The average reported age was about 18, and in 43 of its 49 experiments the learners were studying English.

That limits what the research can tell you about a live Grade 8 lesson, for two reasons.

• A Grade 8 student is 13 or 14, several years younger than most learners in these studies.
• Both reviews pooled studies of recorded video, so neither says anything direct about live captions of a teacher's own speech.

A third difference is mine to claim, because no study here tests it: a recorded clip usually shows pictures that match the words, and my lessons carry fewer pictures and change pace as questions come up.

What a second study adds

In 2010, Paula Winke, Susan Gass and Tetyana Sydorenko tested 150 students at a large university in the US Midwest, in their second or fourth year of Spanish or Russian, or their second year of Arabic or Chinese. Each student watched three animal documentaries of three to five minutes, dubbed into the language they were studying, and saw each one twice, most groups with captions on one viewing. Two extra groups of second-year Spanish students had captions both times or not at all.

1. Among second-year Spanish students, the 9 who had captions both times outscored the 8 who never had them on both vocabulary tests and on the comprehension test. Groups that small give only a rough answer.
2. Across all four languages, students who had captions on the first viewing beat those who had them on the second on a vocabulary test that played the new words aloud. The order made no significant difference on the written vocabulary test or the comprehension test.

The interviews answer part of the worry I started with: twenty-six students talked to the researchers afterwards, and they said captions helped them see where one word ended and the next began, check what they thought they had heard, and decide what to listen for on the second viewing. Some also called captions a crutch, and one fourth-year Spanish student said that sometimes you "only read and not listen."

A newer review

Satsuki Kurokawa, Aung Myo Hein and Takumi Uchihara published the newer meta-analysis in Language Learning, online in December 2024 and in print in 2025. They kept only studies of incidental vocabulary learning, which means nobody told the students about a word test or pointed them to the words.

Captions still helped, with an effect size of 0.56, which the authors call medium. They give two reasons for the drop from 0.87. Only four of their 49 experiments had also been in the 2013 review, and the 2013 review counted studies that told learners to watch for the words, which this one left out. In the 18 experiments that reported enough detail, groups with captions learned about 14 percent of the target words and groups without them 11 percent.

Several findings matter when you choose a video.

• Intermediate learners gained the most, and beginners and advanced learners gained much less. That gap fell short of statistical significance, and the authors still warn against overestimating what captions do for students with the least English.
• Secondary students, the group closest in age to Grade 8, showed a small benefit, the same size as university students. Primary students showed a much larger one, though nine of those twelve results came from intermediate-level children, so the authors put the gap down to language level rather than school level.
• Lecture-style videos made for native speakers showed almost no benefit. Six studies used them, three built on TED Talks, and the authors blame the thin supply of pictures and the vocabulary those videos carry. Of all the video types, that one looks most like a teacher talking.

Only eight studies tested students again after a delay, so the review can't say how long the words last.

Try this tomorrow

Turn captions on when you show a video, and before you press play, teach the three words students need to follow it. Say each word, write it on the board, and give a short meaning in plain English, with a picture or a home-language equivalent when that helps. Point to them again just before the clip starts.

The number three is my own choice, and no study picks it, though one classroom study backs teaching the words first. In a 2019 study, Geòrgia Pujadas and Carmen Muñoz followed four Grade 8 classes at a state school near Barcelona for a school year. The students, aged 13 and 14, were beginner to low-intermediate learners of English, and they watched 24 episodes of Fresh Off the Boat, two classes with English captions and two with Spanish subtitles. One captions class and one subtitles class opened each session with short activities on that episode's five target words, such as matching exercises and crosswords, and those two classes learned more of the target words than the two that skipped the task, with captions or with subtitles. Each condition ran in only one class, and the activities themselves taught the words, so the study can't say how much of that learning came from the episodes.

My own reason for the habit, which no study here tests: a student who met a word ten minutes ago has a better chance of catching it when it flashes past in a caption.

I build MAMAKO, which puts live English captions of a teacher's speech on the classroom screen. Because the research above comes from recorded video, I would still teach the three words first.