How many Spanish words do you need?
Fewer than you have been told, and the honest answer depends entirely on which book — which is the thing the usual answer never says. «Five thousand words to be fluent» is a number with no text under it. Here is one with a text under it.
The book is «Cuentos de la selva», Horacio Quiroga's eight jungle tales of 1918: 17,546 running words drawn from only 1,516 distinct dictionary words between them. Every figure below is counted on that text when this page is asked for.
The answer, at four depths
| Dictionary words | Of every word printed | What that feels like |
|---|---|---|
| 35 | half | the grammar and the glue — de, que, no, un, ser. You can parse a sentence and still not know what happened in it. |
| 237 | 80% | one word in five is a stranger. Not reading — decoding. |
| 781 | 95% | one word in twenty. Readable with a dictionary at your elbow, and slow. |
| 1,166 | 98% | one word in fifty — the level Paul Nation's work associates with reading unassisted, and the level at which you stop noticing you are working. |
So: 781 words for 95% of a real Spanish book, and 1,166 for 98%. Not five thousand. And the whole book — every word Quiroga uses across all eight tales — is 1,516.
Why coverage is the wrong number to plan around
Because the missing words are scattered through the sentences rather than piled up at the end. At 781 words you know 95% of what is printed — and of this shelf's 1,298 sentences, only 924 have nothing unknown left in them at all. At 1,166 words it is 1,156. Both figures are counted on the same text as the table above.
That is the gap nobody's word-count answer accounts for, and it is why «learn 3,000 words first» is bad advice however good the list is. You do not cross a line into reading; you read with a growing number of sentences that come free.
The part no word list can fix
594 of the 1,516 words on this shelf — 39% of them — appear exactly once. A word met once is a word you will not learn from this text, and no course prepares you for it, because the whole point of a frequency list is that these are at the bottom of it. Looking each one up is what actually kills the reading.
Which is the argument for a glossed text rather than a longer word list. In the reading room every one of those 1,516 words carries its dictionary form and meaning where it stands, and the English sits under the sentence until you ask for it — so the 39% cost you nothing and the 781 arrive while you read.
Start with the shortest one
«Las medias de los flamencos» is 1401 words, free and entire, with no account. The word lists are ordered by how many whole sentences each word finishes rather than by raw frequency, and the reading order prices every book on the shelf in words you do not yet have.