Why you understand everything and can say nothing

The gap is real, measured, and not a personal failing

You follow a whole episode without subtitles. You read a comment thread and understand every line. Then somebody asks you a question in a café and nothing comes out. It is the single most demoralising experience in language learning because it feels like proof that the hours were fake.

They were not fake. They bought exactly what they were spent on. Recognising a word and producing a word are separate abilities, they are measured separately, and the difference between them is one of the most reliably reproduced findings in vocabulary research.

What the research actually found

Stuart Webb's 2008 study is the clean one, because it fixed the design problem that muddied earlier attempts: it used equivalent test formats for both directions, with different target words on each side, so the two scores were genuinely comparable rather than one being a rerun of the other.

Two findings came out, and the second is the one nobody repeats.

Words you recogniseWords you can produce

1. The most frequent wordseveryday, high-frequency vocabulary

   widenearly as wide

2. Mid-frequency wordsthe working vocabulary of an adult conversation

   widenoticeably narrower

3. Lower-frequency wordsthe specific ones you reach for and miss

   widemuch narrower

The first finding is the expected one: total receptive vocabulary was larger than productive vocabulary, and "the difference between receptive and productive knowledge increas[ed] as the frequency of the words decreased." The rarer the word, the more one-directional your knowledge of it.

The second finding is the encouraging one, and it changes what the problem is. When Webb scored the same responses for partial knowledge rather than full knowledge, "there was little difference among vocabulary sizes at each frequency band." You do not not know those words. You half-know them. What you are missing is not the word — it is fast, reliable access to it.

Where the delay actually happens

Saying something in a second language is a chain, and only one link in it gets practised by reading and listening.

1. 1

   You decide what you want to say

   This happens in your own language, fast, and it is never the bottleneck. This is why the feeling is specifically one of having something to say and no way to say it.

2. 2

   You retrieve the words — from meaning to form

   This is the direction almost nothing you do practises. Reading goes form to meaning. Listening goes form to meaning. Flashcards usually go form to meaning. Production is the other way round, and it is a different route.

3. 3

   You assemble them, under a clock

   Word order, particles, the verb ending. Doable on paper with ten seconds; a different task at conversational speed, where the window is under a second and the other person is looking at you.

4. 4

   You recognise what comes back

   The one link input practice genuinely builds — and it is the reason the whole thing is so confusing. This link is strong, which makes the failure of links 2 and 3 feel inexplicable.

What each activity actually builds

This is the table worth pinning up. None of these activities is bad. They are just not interchangeable, and most study plans contain four items from the top half and nothing from the bottom.

Activity	What it builds	What it does not build
Watching with subtitles	Recognition, listening speed, a feel for register and rhythm	Retrieval. You are never asked to produce anything.
Reading	Recognition, grammar intuition, the largest vocabulary gains per hour	Retrieval, and any speed constraint at all — you set the pace.
Flashcards, form → meaning	Recognition, fast and measurably	Production, unless the card is deliberately reversed.
Flashcards, meaning → form	Retrieval of single words, in the right directionThe cheapest fix on this table.	Assembly into sentences, and anything under social pressure.
Shadowing	Articulation, prosody, mouth mechanics, confidence in your own sound	Retrieval — the words are supplied to you. You are copying, not choosing.
Writing	Retrieval and assembly, genuinely both	Speed. You can stop, look things up and rewrite — which is exactly what conversation forbids.
Speaking to a person	Retrieval, assembly and speed together, which is the only place all three meet	Coverage. A conversation will not systematically cover a syllabus.

Why this feels worse than it is

Two things make the gap more painful than the underlying skill difference warrants.

The first is that your receptive ability sets your expectations. You know what a good sentence in this language sounds like — you have heard thousands of them. So when you produce a clumsy one, you can hear exactly how clumsy it is. A beginner who understands nothing is spared that comparison. You are not, and the better your listening gets, the sharper the contrast.

The second is that recognition is what gets tested. Reading comprehension, listening comprehension, multiple choice: the things that produce a score and a sense of progress are overwhelmingly receptive. It is entirely possible to pass increasingly difficult tests for years while the productive side moves very little, and to have no feedback at all telling you so.

What to do about it, concretely

Reverse your flashcards. If the card shows the Korean and you say the English, you are practising the direction that is already strong. Show the English, produce the Korean, out loud. It is the smallest change on this list and it targets link 2 directly.

Put a clock on it. The gap Webb found is partly a speed problem — partial knowledge that cannot be reached fast enough is functionally absent in conversation. Practice that allows unlimited time trains the wrong thing.

Write, then say the same thing. Writing gives you retrieval and assembly without the time pressure; saying it afterwards adds the pressure to material you have already built. It is the cheapest ramp onto the hard version.

And then have the conversation, because it is the only activity on that table where retrieval, assembly and speed happen at once, and the research on doing it alone is not encouraging. If the blocker is not knowing what to say, the first ten minutes is solvable.

Sources

1. Webb, S. (2008) Receptive and Productive Vocabulary Sizes of L2 Learners. Studies in Second Language Acquisition, 30(1). ERIC EJ784734.

   Used equivalent receptive and productive test formats with different target words on each side. Found total receptive vocabulary larger than productive, with the gap widening as word frequency decreased — and, when scored for partial knowledge, little difference between the two at any band.

   Checked 2026-09-27

2. Toyama, M., & Yamazaki, Y. (2021) Classroom Interventions and Foreign Language Anxiety: A Systematic Review With Narrative Approach. Frontiers in Psychology, 12, 614184.

   Cited here only for the point that part of this gap is affective rather than lexical, and for Liu et al. (2018), in which learners who did a production task individually finished with higher anxiety than they began with.

   Checked 2026-09-27