How Many Words You Learn From Reading: The Data

One question haunts everyone who tries to learn a language by reading: is this actually working? You read a chapter, meet a dozen unfamiliar words, and have no idea how many of them you will remember tomorrow. It feels productive, but it is hard to tell whether the words are sticking or just passing through. Fortunately, researchers have measured exactly this, and the numbers give you a realistic picture of what reading does for your vocabulary — and where it falls short.

A meta-analysis published in Language Teaching pooled 24 studies, 29 effect sizes, and 2,771 participants to quantify how much vocabulary learners pick up incidentally from meaning-focused input. The findings are encouraging and clarifying in equal measure: reading genuinely teaches words, the gains last, but the per-text rate is low enough that volume and a review layer both matter. Here is what the data says and how to act on it.

The headline numbers

The study reported large overall effects for incidental vocabulary learning — around g = 1.14 on immediate posttests and g = 0.93 on delayed posttests taken weeks later. Effect sizes that large are uncommon in education research, and the fact that they held up after a delay is the important part: this is not a flattering spike that fades by the weekend, but durable acquisition. The takeaway is simple and reassuring. Reading for meaning, with no deliberate word study at all, reliably builds vocabulary. The hours you spend reading at your level are not just comprehension practice; they are genuine vocabulary acquisition.

The proportion of words actually learned per exposure tells the other half of the story. Across the studies, learners acquired roughly 9 to 18 percent of the unknown target words on immediate tests and 6 to 17 percent on delayed tests. So from a single pass through a text, you might learn one unknown word in six to ten. That is the reality behind the encouraging effect sizes: each word is learned slowly, over multiple encounters, not captured the first time you see it. Understanding this prevents the most common form of discouragement — the feeling that you "should" remember every new word and are failing when you do not.

Why small per-word odds still add up

If you only learn one unknown word in eight from a given text, it is fair to ask how anyone ends up with a large vocabulary at all. The answer is sheer scale. Reading exposes you to thousands of words, and the words you fail to learn today are not gone — they wait for you in the next text, and the one after that. High-frequency words in particular appear relentlessly, so even a low per- encounter learning rate guarantees that common words get learned quickly simply because you meet them so often. The low odds only hurt for rare words, which by definition you encounter seldom.

This is the mathematical case for reading in volume. Doubling how much you read roughly doubles your encounters with every word, and more encounters mean more of those partial memories cross the threshold into solid knowledge. It is also why easy material beats hard material for building vocabulary: you can only read a lot when the text is comfortable, so ease drives volume and volume drives words. We make the full argument for reading large amounts of easy text in our guide to what extensive reading is, and the meta-analysis is the evidence underneath it.

Reading vs listening vs video

Because the meta-analysis compared input channels, it can tell you where to spend your time. Gains were broadly similar across reading (about 17 percent immediate, 15 percent delayed), listening (15 percent, 13 percent), and reading while listening (13 percent immediate, 17 percent delayed). In other words, text-based input is among the most efficient ways to pick up words, and pairing reading with audio holds its own — particularly on delayed tests, where reading-while-listening posted the strongest retention. This lines up with our deeper look at reading versus listening for vocabulary gains, which found the two channels are complementary rather than rivals.

Video was the clear laggard, with learners picking up only about 7 percent of words on immediate tests and 5 percent on delayed tests. The likely reason is attention: on-screen action, images, and plot pull cognitive resources away from the language itself, so fewer words get encoded. That does not make video worthless — it is excellent for listening practice, motivation, and cultural exposure — but if your specific goal is to grow vocabulary efficiently, reading is the better lever. If you do learn from video, treating the transcript as the primary text, as in our guide to learning a language from YouTube transcripts, helps pull the language back to the foreground.

What makes reading teach more words

The moderator analysis is where the practical advice gets specific. Proficiency mattered a lot: learners beyond the basic level gained far more than beginners (g = 1.40 versus 0.70), because once you can already follow a text, you have the spare capacity to absorb the new words in it. This is a virtuous cycle — the more you know, the faster you learn — and a reason to push through the difficult early stage where reading feels slow and unrewarding. It gets dramatically more efficient once you cross a few thousand words.

Text choice mattered just as much. Narrative texts produced much larger gains than dense expository writing (g = 1.43 versus 0.61), and material written for language learners beat material written for native speakers (g = 1.56 versus 0.71). The message is to read stories at your level and lean on learner-oriented material — graded readers, simplified content — especially early on, rather than forcing your way through authentic news or academic prose before you are ready. Finally, spaced exposure beat massed exposure (g = 1.51 versus 0.97): reading a little most days outperforms occasional marathon sessions, which is worth remembering when you plan your week.

Why you still need spaced repetition

The same data that validates reading also explains its limit. If you learn only a fraction of unknown words per exposure, then the words you need but rarely encounter — the ones that do not happen to recur in your next few texts — can take an extremely long time to stick, and some will keep slipping away no matter how much you read. Reading alone leaves these words to chance. That is precisely the gap spaced repetition is built to close, by deliberately resurfacing a chosen word on an expanding schedule so it gets reinforced just before you would have forgotten it.

The research even gestures at why this works: spaced exposure outperformed massed exposure in the meta-analysis, and spacing is exactly what an SRS automates and optimizes. The efficient workflow is not reading or flashcards but reading plus a thin layer of flashcards — mine the handful of genuinely high-value words that keep blocking you, drop them into a deck with the sentence they came from, and let the algorithm handle the timing. Our guide to building an SRS vocabulary deck from your reading walks through how to do this without turning reading into a chore. Reading supplies the words and their context in bulk; spaced repetition guarantees the important ones survive.