Can You Learn Vocabulary Just by Listening?

Yes, but slowly. In the best-known study, learners who only listened to a story could pick the right meaning for about 29 percent of its new words right afterward, could write the meaning of 2 percent, and three months later recalled none. Listening teaches words when you also see them written, hear them often, and review them on purpose.

What happens when you only hear a new word

Brown, Waring and Donkaewbua (2008) ran the cleanest version of this experiment. Thirty-five Japanese university students went through three graded readers, one book per mode: reading silently, reading while listening to the audio, or listening only. Inside each book, 28 real words had been replaced with invented spellings so nobody could know them in advance. Seven of the 28 appeared 15 to 20 times in the story, seven appeared 10 to 13 times, seven appeared 7 to 9 times, and seven appeared only two or three times.

Then came the tests. A multiple-choice test, where you pick the meaning from options, and a translation test, where you write the meaning with no help. Immediately, one week later, and three months later.

Mode	Picked the right meaning, immediately	Wrote the meaning, immediately	Wrote the meaning after 3 months
Reading only	45%	16%	about 1 word in 28
Reading while listening	48%	15%	about 1 word in 28
Listening only	29%	2%	0 in 28

Read that bottom row again. These were stories where the target words came around as many as twenty times. Listening alone produced something at the recognition level, almost nothing at the recall level, and nothing at all that survived a season. The authors’ own summary was blunt: new words could be learned in all three modes, but most words were not learned.

I want to be fair to listening here. The reading rows are not exactly a triumph either. One word in 28 after three months is a slow way to build a vocabulary in any mode. But the listening row is in a different league of slow.

Why the ear is the hardest place to meet a word

Nothing about this is a verdict on your ears. Listening is a hard place to meet a new word for reasons that have to do with speech itself.

Speech does not come with spaces. On the page a new word arrives with a boundary on each side. In speech you have to decide where it starts and stops while the next three words are already arriving. A reader gets a second look for free. A listener gets one pass at the speaker’s speed, and by the time you have decided the strange thing you heard was a word and not two, the sentence has moved on.

Then there is English spelling. In a language where writing tracks sound closely, hearing a word tells you how to write it, so a listener loses nothing a reader has. English is not that language. Hear “colonel” for the first time and you will not go looking under c. This is such a common dead end that we wrote a separate guide to looking up a word you only heard.

The third reason is the strange one. Listening is forgiving in a way that works against you. In 2013, van Zeeland and Schmitt tested how much of a spoken text listeners needed to know to follow it, at 100, 98, 95 and 90 percent coverage. Most of their native and non-native listeners coped adequately at 90 percent, and at 95 percent comprehension was relatively good with much less variation. That is a far lower bar than the 98 percent usually quoted for reading, and it means roughly 2,000 to 3,000 word families get you through everyday listening where reading at 98 percent needs 6,000 to 7,000.

Good news for comprehension. Bad news for learning. An unknown word in every ten or twenty slides past without derailing the story, which is exactly why it also slides past without being learned. Your brain has no reason to stop on a word it did not need.

Reading pulls ahead, and then the gap closes

Vidal (2011) ran the comparison at a larger scale. Two hundred and thirty university students either read three academic texts, watched three lectures, or did nothing but take the tests. She then split the results by proficiency, four levels of it.

Reading won at every level. The readers’ gains ran from 19 to 38 percent depending on proficiency, the listeners’ from 7 to 28 percent. But the interesting part is the shape of the gap, which shrank as proficiency went up. A month later the readers still retained more than the listeners at every level except the top one, where there was no significant difference between the two.

So the rule is not “reading beats listening.” It is closer to “the better you already are at listening, the more listening gives you.” That matches a 2020 finding by Zhang and Graham, who tested 137 Chinese learners of English and found that listening proficiency predicted how much vocabulary they picked up from audio better than the size of their existing vocabulary did. Ears that keep up have attention to spare for the new word. Ears that are just holding on do not.

The study where listening won

Now the counterexample, because there is one and it is recent. Aliyar, Siyanova-Chanturia and Skalicky (2026) had 88 learners of Italian either read part of an authentic Italian novel, listen to its audiobook, or serve as controls. They tested form recognition, meaning recall and meaning recognition, immediately and three weeks later. Both groups learned words and kept them at three weeks. And the audiobook group gained substantially more than the readers.

I do not think this overturns the English results. It explains them. Italian spelling is close to a transcript of Italian sound, so a listener who catches a word can write it, and the second-look advantage of print mostly disappears. An audiobook also arrives with a performer’s phrasing and pauses, which does some of the segmenting for you. The authors themselves call for the same design to be run in other languages, and until it is, the safe reading is this: listening is not inherently a weak way to learn words. It is weak where sound and spelling diverge, and English is the poster child for that.

The mode that beats both: read it while you hear it

Go back to the table. Reading while listening scored 48 percent to reading’s 45 on the immediate recognition test, a difference the authors report as not statistically significant. Against listening alone, though, it was not close. Give the ear a page to follow and the listening problem largely goes away.

Webb and Chang (2015) tested the same idea over a whole course rather than a single book. Sixty-one Taiwanese learners of English either worked through ten graded readers, reading while listening to the audio, or followed a conventional coursebook. Across 100 target words, the extensive reading group’s relative gains were 44.06 percent right after the program and 36.66 percent three months later. Relative gains means the share of the words they did not know at the start that they knew at the end, so that is a lot of vocabulary surviving a lot of time, from ten short books.

The mechanism is not mysterious. The audio gives you the boundaries and the sound. The text gives you the spelling and the second look. You meet the word in both forms at once, which is the same thing a good dictionary entry does when it puts a headword, its IPA and an audio button on one line. If you take only one thing from this article, take this: the transcript is the part of the method that makes the audio work. Using it is not cheating.

Words you meet once are words you meet by accident

There is a second problem hiding in the Brown table. It is about count, not mode.

In that study the words that appeared more often were more likely to be learned and more resistant to decay. That is the expected result, and it is the reason a story where words repeat is better teaching material than a news bulletin where they do not. But the more sobering finding came from Webb and Chang. Across their ten books, the relationship between how often a word occurred and whether it was learned was non-significant. Frequency, they wrote, was perhaps one of many factors. In other words, you cannot rely on a text to repeat a word enough times, and even when it does, that alone does not settle whether the word sticks. We have a whole post on how many encounters it takes to learn a word, and the honest answer is more than most audio will give you.

Think about what a podcast episode actually provides. Most of the words you do not know appear in it once. Once, through the ear only, at speed, in a language where you cannot spell what you heard. Every condition from the previous sections lines up against that word. If listening is going to build your vocabulary, the repeats have to come from somewhere else.

What a little explanation does

One more study, because it changes what “listening practice” should mean. Zhang and Graham (2020), the same 137 learners from above, listened to material with the new words explained aloud in one of three styles, or with no explanation at all. No explanation was the least helpful condition. The most helpful, for learners at every level, was the style that explicitly drew attention to the word’s form, not just its meaning.

That is a quiet argument against pure exposure. The raw stream of speech is the weakest form of listening for vocabulary. A moment that names the word, shows you what it looks like, and tells you what it means is what turns a thing you heard into a thing you know. Learner podcasts that pause and explain are built on this. So is doing it yourself with a transcript and a dictionary, which has the advantage that it works with anything you actually want to listen to.

How to make listening count

Here is the routine the research points to. None of it is exotic.

• Choose audio that comes with a transcript, and use the transcript at least once. Reading while listening is the mode that wins, so build your listening around material where it is possible.
• Pick a level where you can follow without straining. Vidal’s and Zhang and Graham’s results both say the same thing: comprehension you already have is what frees the attention that learning a new word needs.
• Stop on a new word. Not every one, but when a word matters, pause, and write down what you heard even if the spelling is a guess.
• Find the real word and look at its IPA next to its spelling. This is the step that repairs the English problem. The sound you caught and the letters you will meet in print get tied together on the spot. Say it once, out loud.
• Put it into spaced repetition. The story was never going to repeat that word 15 times for you. A scheduler will bring it back tomorrow, then in a few days, then in a few weeks, which is the repetition curve the incidental studies say the word needs.
• Listen to the same episode again a day or two later. A second pass, now that you know the words, is worth more to those words than a new episode is.

What you are doing with that list is converting listening from an incidental mode into a deliberate one, one word at a time. The input still comes from the audio you enjoy. The learning comes from what you do in the ten seconds after you hear something new.

Where Vocaby fits

Vocaby is built to be that ten seconds. Every one of its 29,000+ English words comes with IPA, audio, a plain definition and real example sentences, so each entry is a small reading-while-listening moment: the sound and the spelling on the same card. The review side runs on FSRS, a spaced repetition scheduler that estimates when a word is about to fade and brings it back just before, which is the repeat count no podcast can promise. Browse the collection at our word lists.

Keep listening. Just do not ask your ears to do the whole job alone.

Learn English vocabulary with Vocaby on the App Store