Open a Spanish novel at random and count the words. Somewhere around eight in ten of them will come from the thousand most common words in the language. Not eight in ten of the interesting words — eight in ten of all of them, because que, de, ser, hacer and poder turn up on every page for the rest of your life.
This is the single most exploitable fact in language learning, and also the most widely misunderstood. It is true that a small vocabulary buys a lot of coverage. It is not true that 80% coverage means you understand 80% of what you read. Both halves matter.
The shape of the curve
The clearest numbers come from Paul Nation, whose 2006 study measured how much of a text you cover as your vocabulary grows. Counting word families — hablar, hablo, hablando as one item — against written English:
| Word families known | Coverage of written text | Coverage of speech |
|---|---|---|
| 1,000 | ~80% | — |
| 2,000 | ~88% | ~89% |
| 3,000 | — | ~96% |
| 4,000 | ~95% | — |
| 6,000–7,000 | — | ~98% |
| 8,000–9,000 | ~98% | — |
Nation, "How Large a Vocabulary is Needed for Reading and Listening?" (2006). Figures for the higher bands include proper nouns, and a dash means Nation did not report that band for that mode. Spanish behaves the same way; the curve is a property of how languages are used, not of English.
Look at what the curve does. The first thousand words buy you 80 points of coverage. The eighth thousand buys you about one. In terms of return per word learned, the earliest vocabulary is worth something like fifty times the latest.
The trap hiding in those numbers
Here is where people get burned. 80% coverage sounds like comfortable reading. It is not. At 80%, one word in five is unknown — that's roughly two unknown words per line, or forty on a page. Nobody reads like that. You grind, you look things up, you lose the thread of the sentence while you're decoding it, and you quit.
Nation's conclusion is that unassisted comprehension needs around 98% coverage, which for written text means a vocabulary in the range of 8,000 to 9,000 word families. Spoken language leans harder on common words, so speech gets there at roughly 6,000 to 7,000.
So both of these are true at once:
- The first 1,000 words are worth more than the next 7,000 combined, measured in coverage per word.
- 1,000 words is nowhere near enough to read comfortably. It's enough to start reading with support, which is a different and still very valuable thing.
A thousand words is a starting line, not a finish line
At 1,000 words you can follow the gist of simple conversation, read graded stories without stopping every sentence, and — this is the part that compounds — start learning new words from context instead of from a dictionary. That last one is the real prize. Below about 95% coverage, guessing from context mostly fails, because the surrounding words you'd guess from are themselves unknown. Above it, the language starts teaching itself.
Textbooks are surprisingly bad at this
You might assume a Spanish course would front-load the highest-frequency words. Mark Davies, who built the corpus behind the standard Spanish frequency dictionary, checked six widely used first- and second-year textbooks against it. The results are not flattering.
Of the top 1,000 Spanish words, the textbooks collectively covered 82%. Of the top 2,000, only 66%. Meanwhile a large share of the vocabulary they did teach sat outside the top 2,000 entirely.
The pattern in what gets skipped is the revealing part. Textbooks are organized by topic — the restaurant chapter, the airport chapter, the family chapter — which is great for concrete nouns and terrible for everything else. Davies found dozens of very common abstract words missing: momento, situación, ocurrir, importante. Words with no chapter to belong to. You end up able to name eleven vegetables and unable to say "it depends on the situation."
Frequency order isn't a complete plan
It would be dishonest to stop there, because pure frequency order has real problems and anyone who has tried it knows them.
The top of the list is the hardest part of it
The most frequent words in any language are function words — que, se, lo, por, para — and they are close to meaningless in isolation. "Se = oneself" is not knowledge. These words are learned by meeting them in hundreds of sentences, not by drilling a translation.
Frequency doesn't match your life
Corpus frequency is an average across newspapers, novels and transcripts. If you're learning Spanish for a job in a hospital, or because your partner's family is Colombian, a few hundred words matter far more to you than their rank suggests.
Rank differences get meaningless fast
The gap between word #12 and word #400 is enormous. The gap between #2,400 and #2,900 is noise. Past the first couple of thousand, frequency is a rough guide and interest should take over.
Frequency for the core, curiosity for the rest
Take the first 1,000–2,000 words in something close to frequency order, always inside sentences so the function words have something to attach to. Then switch: read and watch things you actually like, and let what you meet decide what you learn next. The frequency list gets you to the point where input works. Input takes you the rest of the way.
Three practical consequences
- Don't learn a word you haven't met and won't meet. If it isn't in the common core and it hasn't appeared in something you were reading, it can wait.
- Count coverage, not cards. "I know 1,400 words" is only useful as a proxy for "I can read this without drowning." Pick a text you want to read and check how many words per page stop you. Under two or three is the zone where reading becomes learning.
- Learn the core in sentences. The highest-frequency words are grammatical glue. A card that says lo → it teaches almost nothing; a card built on a sentence you understand teaches the word and its behaviour at once.
The core vocabulary first, in context, in order
Lernlang's decks are built straight from frequency data — you work through the words that actually carry the language rather than a chapter's worth of vegetables. Every word arrives in a sentence with audio, reviews are scheduled around when you're likely to forget, and level- matched stories recycle what you've learned so the coverage turns into reading rather than a number in an app.
The short version
The frequency curve is the closest thing language learning has to a cheat code, and it is strongest at exactly the moment you need it most — the beginning, when nothing makes sense and quitting is easiest. A thousand words, chosen well and met in real sentences, is a few months of honest work and it changes what you can do with the language.
Just don't mistake the on-ramp for the road. Eighty percent coverage is where reading becomes possible, not where it becomes pleasant. The way to 98% is not another frequency list — it's several hundred hours of reading and listening to things you'd want to read and listen to anyway.