Frequency-based vocabulary: learn the right words first
Why learning words in order of how often they appear buys you the most comprehension for the least effort.
Frequency-based vocabulary means learning a language’s words in order of how often they occur, starting with the most common, so that your earliest effort buys the most comprehension. Because a small number of words does most of the work in any text, front-loading them is the quickest way to get reading.
Zipf’s law: a few words do most of the work
Word frequencies are wildly uneven. The linguist George Kingsley Zipf described the pattern in the 1930s and 1940s: a word’s frequency is roughly inversely proportional to its rank, so the most common word appears about twice as often as the second, three times as often as the third, and so on. In English the single word “the” accounts for something like 5 to 7 percent of all the words on a typical page. A handful of words are everywhere; the vast majority are rare. This lopsidedness, Zipf’s law, is exactly why frequency ordering is such high leverage: the first words you learn are the ones you will meet again and again.
Coverage is not comprehension
Here is the crucial caveat, and the reason frequency lists are oversold. You will often hear that the most common few hundred words cover around 80 percent of any text. As a statement about running words that is roughly true, but it deceives, because those top words are the grammatical skeleton (“the”, “a”, “of”, “is”, “with”) and the meaning of a text lives in the rarer content words that make up the rest. Knowing the words behind 80 percent of the tokens on a page is nothing like understanding 80 percent of what it says.
Coverage tells you how many words on the page you recognise. Comprehension is whether you understand what they are saying. They are not the same thing, and the gap between them is where real reading ability is built.
The research bears this out. Hu and Nation’s 2000 study set 98 percent known-word coverage as the benchmark for “unguided” comprehension, reading comfortably without looking things up, and Paul Nation’s later work estimates that reaching 98 percent coverage of a novel takes a vocabulary on the order of 8,000 to 9,000 word families. Webb and Macalister found in 2013 that even texts written for children can demand a vocabulary of around 10,000 words for that same 98 percent coverage. The exact numbers are contested (Kremmel and colleagues could not fully reproduce Hu and Nation’s results) but the direction is not: you need to know a high proportion of the words, on the order of 85 percent or more, to deeply understand a text without a dictionary. “Getting the gist” is a lower and more attainable bar for everyday reading, especially with a tool that makes lookups cheap. Lector’s methodology works through this in detail.
How to use frequency lists and decks
The practical move is to drill the most frequent few thousand words deliberately, then let reading take over. A frequency-ordered flashcard deck or cloze bank front-loads the words that unlock the most text, and pairing it with spaced repetition keeps them from slipping away. Once you can follow simple texts, sentence mining from your own reading takes over from the generic list, because your reading becomes a personalised frequency list of exactly the words your chosen material uses. For a ready-made starting point, the Afrikaans reference data ships a free, frequency-ordered word list and starter deck, and the other languages link hand-picked frequency decks.
Caveats: function words, domains, and word families
- Function words versus content words. The very top of any frequency list is function words that carry little meaning alone. They are worth knowing cold, but do not mistake mastering them for understanding; the content words below them are where comprehension is won.
- Frequency is domain-specific. A general list reflects average language. The words you need for medical texts, football commentary, or a particular novelist diverge quickly from that average, which is another reason to shift to mining your own material as soon as you can.
- Word families, not word forms. Coverage figures are usually counted in word families: a headword plus its inflected and derived forms (read, reads, reading, reader). One family can be many forms, so a “9,000 word” target is larger in raw forms than it sounds.
Frequency is where the reading-first method starts, not where it ends. Once you have the common core, your own reading takes the wheel.
Frequently asked
There is no single number, and "fluent" is a spectrum. As a rough guide, reading a novel comfortably without constant lookups takes roughly 8,000 to 9,000 word families (about 98 percent coverage); everyday conversation needs fewer. You can start reading useful material far sooner, with a few thousand common words plus good tools.
Roughly yes, if you count running words, because those few hundred words are so common they make up most of the tokens on a page. But coverage is not comprehension: they are mostly grammatical glue, and the meaning sits in the rarer content words. Treat 80 percent coverage as a foothold, not fluency.
Open data sources and language-learning tools publish them. Lector's reference-data pages provide free, frequency-ordered word lists and Anki decks per language; the Afrikaans page, for instance, includes a full frequency-sorted deck you can download and drill.