Knowing 95% of the words on a page sounds like an A. In a real text, it's closer to a failing grade. At that level you hit an unknown word every twenty or so, which is often enough to keep you guessing instead of reading. The number that actually feels comfortable is 98%, and the gap between those two figures is where a lot of language learning quietly stalls.
That gap is the heart of second-language vocabulary acquisition, and the research on how words move from "seen once" to "use it without thinking" is specific enough to explain exactly what's happening to you. Word knowledge is not a light switch. It's a dial, with a long stretch of positions between "never met it" and "it comes out clean in a sentence," and most learners idle at partial knowledge for ages without ever crossing into real retention.
The good news is that the science is clear on what "partial" means, why it sticks, and what conditions actually push a word over the line. There are a handful of these conditions. Most learners are only using one or two of them.
Once you know the coverage thresholds and encounter counts the research points to, you'll have practical tips for learning vocabulary in a way that actually sticks.
Table of Contents
- What It Means
- The Coverage Threshold
- How Many Encounters?
- Practical Tips
- Tools for Vocabulary Acquisition
What Vocabulary Acquisition Actually Means
Vocabulary acquisition is the process by which a learner moves from first encountering a word to being able to use it reliably in speech and writing. It is not a single event but a gradual accumulation of knowledge across three dimensions — form (how the word looks and sounds), meaning (what it refers to and implies), and use (the grammar patterns, collocations, and register it belongs to). Research consistently finds that most words require 10–20 meaningful encounters across different contexts before they become automatic and fully usable, which is why the conditions and spacing of those encounters matter far more than most learners assume.
What Does It Mean to "Know" a Word?
Most learners treat vocabulary as binary: either you know a word or you don't. Paul Nation's vocabulary knowledge framework breaks this down into three dimensions that each need to be learned: form (how the word looks and sounds), meaning (what it refers to, what it implies, what words it associates with), and use (the grammar patterns it fits, the register it belongs to, the words it collocates with).
Full knowledge requires all three. The reason a word "disappears" when you need it is usually that you have the meaning but not the use — or you have the form and meaning but not the collocations that make it come out naturally. You've learned part of the word. The rest hasn't been practiced enough to be automatic.
The other major gap is between recognition and production. Your recognition vocabulary (words you understand when you see or hear them) will always be larger than your production vocabulary (words you can call up on demand). Recognition is picking a face out of a police lineup: the answer is standing right there, you just have to point. Production is sketching that same face from memory on a blank page. One is multiple choice. The other is an essay. Recognition only asks your brain to match a form to a meaning; production asks it to retrieve the form, the meaning, and the correct usage pattern at once, under real-time pressure. The surprising part is how wide the gap runs: studies commonly find productive vocabulary at somewhere between half and four-fifths of receptive vocabulary, and it widens for rarer words. Plenty of words just sit in the "I know that one" zone and refuse to come when called.
What Is a Word Family, and Why Does It Matter?
When researchers count vocabulary, they typically count word families, not individual word forms. A word family is a base word plus its inflected and closely related derived forms. The family around decide includes decides, decided, deciding, decision, decisive, indecisive: roughly one family. Learn the base and the relatives mostly come along for the ride, the way learning one good knot teaches you three knots that are basically the same wrist motion. Nation's (2001) framework uses word families as the counting unit for exactly that reason.
This matters practically: a vocabulary size of "5,000 words" means something completely different depending on whether someone is counting word forms or word families. The research thresholds covered below use word families. If you're tracking your vocabulary size with a tool that counts forms, your number will look larger than what the research is describing.
Incidental vs. Intentional Vocabulary Learning
There are two routes to vocabulary. Intentional learning is what most people think of: flashcards, word lists, deliberate study. Incidental learning is picking up words from context while reading or listening, without explicitly trying to memorize them.
Both work, and the question of which one matters more has occupied the field for decades. James Coady and Thomas Huckin built an entire Cambridge Applied Linguistics book around that exact tension, and the field has mostly landed where they did: you need both, in different proportions at different stages. The research says intentional learning is more efficient for high-frequency words early on — you can load them quickly and get the form-meaning link established. Incidental learning becomes more powerful once you have enough coverage to actually infer meaning from context, and it's almost the only realistic route to the kind of broad vocabulary that produces natural, fluent language. The problem: incidental learning requires specific conditions to work, and most learners hit those conditions less often than they assume.
The Coverage Threshold: The Number Most Learners Haven't Heard
How Much Vocabulary Do You Need to Understand a Text?
The threshold finding is probably the most practically important result in vocabulary research, and it's higher than most learners expect. Hu and Nation (2000) ran one of the field's foundational pieces of empirical research, testing reading comprehension at four coverage levels: 80%, 90%, 95%, and 100% of known words. The result was a near-linear relationship between coverage and comprehension. At 80% coverage (one unknown word in five), comprehension was poor. At 95%, most learners still didn't reach what the researchers called "good comprehension." Around 98% is where independent reading becomes genuinely comfortable for most learners.
Picture the page itself. At 80% coverage, one word in five is blacked out like a redacted government file, and you're trying to reconstruct the secret from what's left. At 95%, the redactions thin out but never quite vanish; you trip on one every couple of sentences. At 98%, the marker is basically gone and the page just reads.
This single finding explains a lot. If you can comprehend the general drift of native content but can't really learn from it or enjoy it, the coverage gap is the likely cause. You're not making a comprehension effort. You're making a survival effort.
A 2023 replication by Kremmel et al., conducted with L2 learners in a non-WEIRD context, could not fully reproduce the original results and urged some caution about applying specific coverage cutoffs too rigidly. The clearest effect that did surface was that higher coverage tracked with better comprehension on the harder, expository text, where the 98% version was understood better than the 95% version. So the broad direction holds even if the exact thresholds are less clean than a single table suggests: getting from 90% to 98% coverage is not cosmetic. It changes whether text is a comprehension task or a genuine reading experience.
Here's the coverage picture across levels:
| Coverage Level | Unknown Words | Comprehension | Incidental Learning |
|---|---|---|---|
| 80% | 1 in 5 | Poor | Very limited |
| 90% | 1 in 10 | Degraded | Unreliable |
| 95% | 1 in 20 | Functional | Possible, inconsistent |
| 98% | 1 in 50 | Comfortable | Effective |
How Many Word Families Do You Need?
Nation (2006) mapped these thresholds to vocabulary size targets:
| Vocabulary Size | Coverage Type | Coverage Level |
|---|---|---|
| 1,000 word families | Informal spoken | ~78–84% |
| 3,000 word families | Everyday spoken | ~95% |
| 3,000 word families | General written text | ~90% |
| 6,000–7,000 word families | Spoken text | ~98% |
| 8,000–9,000 word families | General written text | ~98% |
The 3,000-family milestone is the most actionable target for intermediate learners. Getting there takes you to 95% of everyday spoken language and changes what's possible with listening-based input. The jump from 3,000 to 8,000 is longer and slower, but that's the range where reading becomes genuinely independent. Most learners underestimate how much vocabulary sustained reading actually requires.
Why Reading Below the Threshold Teaches You So Little
Here's the feedback loop that trips most learners: you hear that immersion works, so you start watching native TV or reading news articles. You push through, you look up some words, but you don't feel like you're actually learning the language. You are, a little. But the acquisition rate is much lower than it feels like it should be.
The reason is that incidental vocabulary learning from context depends on the surrounding words being known. Working out an unknown word from context is like solving a crossword: you fill the blank using the letters the crossing answers hand you. When most of those crossing answers are blank too, you're not deducing anything. You're just guessing, and your brain quietly stops bothering. The denser the unknown words, the less the context can do its job. The flip side is encouraging: Laufer (2020) found that learners at around 90% coverage could successfully infer roughly half of the unknown words, which lifted their effective coverage toward the 95% range often treated as the minimum for comprehension. Inferencing and coverage feed each other, which is exactly why the conditions around the unknown word matter so much.
This is not a reason to avoid authentic content. It's a reason to match content to your actual vocabulary level, so that the incidental learning engine can run properly. Content calibrated to your level is not "easier content." It's content where the conditions for acquisition are actually met.
How Many Encounters Does a Word Actually Need?
The Repetition Requirement
The research consensus on this is consistent and slightly uncomfortable: a word needs around 10–20 meaningful encounters before it's solidly retained (Nation, 2001; Webb, 2007). Webb (2007) tested 1, 3, 7, and 10 encounters and found that knowledge gains kept climbing with each step up, with sizable gains by 10 encounters and clear signals that even more repetitions help: enough that fuller knowledge across spelling, meaning, and grammatical function lands at the higher end of the 10–20 range.
"Meaningful encounter" is the important qualifier. Think about how you learn a person's name. Someone tells you once at a party and it's gone by the time you've crossed the room. Bump into that same person five times over a month, in the hallway, at lunch, on the train, and you couldn't forget the name if you tried. Words work the same way. Seeing one in a vocabulary list twenty times keeps it in short-term memory, idling. Meeting it in five different sentences across five different sessions builds the distributed, contextual knowledge that actually sticks.
This is also why vocabulary size goals need to be tempered by realistic encounter rates. A word that appears rarely in the texts you're reading will take years to reach 10 meaningful encounters through incidental reading alone. For low-frequency vocabulary, intentional study is the faster route.
Does the Type of Encounter Matter?
Substantially. Encounters where you have to do more cognitive work with the word produce better retention. Inferring a word's meaning from context is more valuable than reading its definition. Using a word to produce output — writing it in a sentence, answering a question that requires it — builds more durable knowledge than passive recognition. Emotionally or narratively significant contexts also tend to produce better retention, which is one reason vocabulary encountered in stories you're genuinely engaged with tends to stick differently than vocabulary encountered in dry practice exercises.
Spacing matters as much as count. Ten encounters spread across three weeks produces much stronger retention than ten encounters in one session — this is the spaced repetition principle applied specifically to vocabulary. The spacing forces retrieval each time, and retrieval practice is one of the most durable retention mechanisms in the cognitive science literature.
What About Cognates and Loanwords?
If your first language is English and you're learning Spanish, French, Portuguese, or Italian, you have a genuine early advantage: transparent cognates (words that share form and meaning across languages) can give you functional access to 30–40% of high-frequency vocabulary without formal study. These words can be treated as near-known for coverage purposes. This is not a trivial advantage — it's a meaningful head start on the coverage thresholds above.
The flip side is false cognates, words that look like they should mean the same thing in both languages but don't. These feel like coverage but mislead you, and the confusion they cause can be sticky. Understanding how your first language shapes (and sometimes skews) your target-language acquisition is the subject of language transfer, and it's worth understanding explicitly rather than leaving to chance.
Practical Tips to Learn Vocabulary That Actually Work
None of this is abstract pedagogy. Every tip below maps directly to one of the conditions above, translated into something you can actually do this week.
Learn High-Frequency Words First
The research case for front-loading high-frequency word families is strong. The most frequent 1,000–3,000 families have the highest payoff per word learned because they appear constantly in both reading and listening, which means each word gets to its 10–20 encounter threshold much faster than rare vocabulary would. Learning rare vocabulary early, before you have solid coverage of the common words, is counterproductive: the rare words are slow to reinforce because you don't encounter them often, and the coverage they add is small.
The practical implication: if you're at a beginner or early intermediate level, a dedicated high-frequency vocabulary study program is one of the highest-return investments you can make. Get to 3,000 solid word families. The reading and listening you can do at 95% spoken coverage reflects a genuinely different level of proficiency than what's possible at 84%.
Know Whether You're at the Acquisition Threshold
A simple rule of thumb from the research: if more than one word in twenty is unknown, you're below the threshold where incidental learning from context works reliably. You're reading to extract meaning, not to acquire vocabulary in any efficient sense. This calibration question is the whole engine behind extensive reading: stay near that easy line and vocabulary grows almost as a side effect of the pages you turn; drop well below it and you tip into reading pain, where every sentence is decoding rather than acquisition.
For sessions where vocabulary growth is the goal, a high unknown-word rate is a signal to drop to easier material rather than grind through.
Atlas Runa's Reader is calibrated to match content to your current vocabulary level, keeping you in the acquisition zone: enough known words that context inference works, enough unknown words that new vocabulary is actually entering. The Reader's unknown-word data is also a real-time diagnostic — it shows you where you are relative to the threshold, not just how many words you've saved.
Use Flashcards as a Supplement, Not a Substitute
Flashcards and spaced repetition systems are the most efficient form of direct vocabulary instruction, genuinely useful for the early stage of vocabulary acquisition: establishing the form-meaning link for new high-frequency words. They're also good for consolidating words you've encountered in reading but haven't retained. The cognitive science behind spaced repetition is robust, and tools that automate the review schedule are solving a real problem.
The trap is using flashcards as a substitute for reading, rather than a complement to it. Flashcard vocabulary is thin vocabulary: you know the translation, you might recognize the word, but you haven't built the collocational, grammatical, and contextual knowledge that makes the word actually usable. That knowledge comes from encounters in context, across varied sentences and situations, over time. Flashcards build the first encounter efficiently; reading and listening build the next fifteen.
Track Where You Are Across the Depth Dimension
Because word knowledge is a spectrum and not a binary, it's worth thinking about vocabulary in terms of depth, not just size. Having 5,000 word families at shallow knowledge (can recognize but not produce) is a different thing from having 3,000 word families at genuine depth (form, meaning, and use all solid). For communication, depth matters more. The learner with 3,000 deep words will outperform the learner with 5,000 shallow ones in almost every real use case.
This means the question to ask isn't just "how many words do I know?" but "how well do I know the words I've already met?" Atlas Runa's Mastery-Aware Word Tracking keeps that distinction visible: it monitors not just whether you've encountered a word, but how solidly you know it, so vocabulary recommendations reflect where you actually are, not just what you've seen.
A Few Moves That Make Practice Sessions More Effective
Not all vocabulary study time is equal. These in-session habits push encounters from shallow toward durable:
- Warm up with review before new input. A short spaced-repetition session before reading primes the words most at risk of being forgotten, which makes encounters with those words in the reading session more valuable.
- Reframe flashcard study as context recovery. When you see a word in isolation, try to recall a sentence you've seen it in, not just the translation. This retrieval practice builds contextual knowledge alongside form-meaning knowledge.
- Aim for diverse text types over more pages. Encountering a word in a news article, a story, and a conversation builds richer knowledge than encountering it ten times in the same genre.
- Flag words you recognize but can't produce. The recognition-to-production gap is real and measurable. Words you recognize passively benefit from active retrieval practice — writing sentences with them, using them in output.
- Space your encounters deliberately. If you're learning a batch of new words, don't front-load all the review into one session. Distribute it across three or four sessions over the next week or two. The spacing is part of what builds retention.
The best tool for vocabulary acquisition
Vocabulary acquisition was never about grinding harder at memorization. Words move from "sort of know it" to "can use it" when a few specific conditions are met, and the right tool holds all of them in place at once:
| What makes a word stick | How Atlas Runa delivers it |
|---|---|
| High-frequency words first | frequency-aware reading and word lists, so the words that matter most come first |
| Enough coverage for context to work | reading pitched to your level, where enough is familiar for context to fill the gaps |
| Spaced encounters over time | words you've met returning at the spacing that makes them stick |
| Doing something with the word, not just nodding | one-tap lookup and save, then reuse in later reading, writing, and speaking |
| Seeing words in fresh contexts | reading, writing, listening, speaking, and targeted drills, all driven by data on what you know and what you still need to see |
Atlas Runa holds all of those vocabulary-acquisition conditions in place at once, so it stops being a separate chore and becomes a side effect of reading things you actually want to read. You don't manage any of it with Atlas Runa; you just read at a level where reading is a pleasure, and the words take root. For the science underneath, see second language acquisition, comprehensible input, and the intermediate plateau.