The frequency curve, and what it buys
Word frequency in natural language is extremely lopsided. A small number of function words and very common verbs and nouns account for an outsized share of everything spoken, and after the first few thousand items each additional word appears less and less often. The exact figures depend on the language and the corpus, and anyone quoting a precise percentage should be treated with mild suspicion, but the shape is consistent and it has a direct practical consequence.
| Vocabulary reached | Roughly what it feels like |
|---|---|
| First few hundred words | Survival. You can ask for things, understand direction and price, and follow the shape of a slow sentence about you. |
| Around 1,000 | Simple conversation about familiar topics, with gaps. Most sentences you hear contain at least one word you know is carrying the meaning. |
| 2,000 to 3,000 | Ordinary everyday speech becomes mostly followable. Native material aimed at learners is comfortable. Unfiltered television is still hard. |
| 5,000 or so | Novels and news are readable with a dictionary at hand rather than in your lap. Conversation is fluent within familiar domains. |
| 8,000 to 10,000+ | Unfiltered native material across topics. This range is where progress becomes slow and unglamorous, because each new word appears rarely. |
The reason this matters is that most beginner courses are organized by topic — colors, animals, the classroom, at the airport — rather than by frequency, so learners spend early hours on words that appear once in a thousand utterances while missing high-frequency verbs and connectives that appear constantly. Learning the first thousand words in frequency order is the single highest-leverage decision in early language study, and it is available for free for every major language.
Two qualifications worth keeping. Frequency lists count word forms in one corpus, and a written corpus differs from a spoken one; a list built from newspapers will overweight words you rarely say. And frequency does not capture usefulness for your life — if you are learning for a specific job or a specific family, some low-frequency words matter enormously to you. Use frequency for the base and your own needs for the top.
Input
The most consistently supported idea in second-language research is that large volumes of understandable material drive acquisition. The operative word is understandable: input at a level where you follow most of it and infer the rest does work, and input where you catch nothing does very little. Sitting with a podcast in a language you do not speak, hoping something soaks in, is one of the most popular and least productive habits in the field.
What "mostly understandable" means in practice is that you are getting the gist without translation, and being stopped by roughly one word in twenty rather than one in three. Below that ratio the material is a decoding exercise and you should find easier material, which is not a defeat — graded readers, children's content, and learner-aimed audio exist precisely because volume at the right level beats struggle at the wrong one.
Two kinds of listening do different jobs. Extensive listening is volume at a comfortable level, ideally daily, for building the ear and absorbing patterns. Intensive listening is short — a couple of minutes — worked over repeatedly until you can account for every word, with a transcript at the end. Most learners do far too little of the second, which is the one that fixes specific gaps.
Output, and why waiting is a mistake
Input alone tends to produce people who understand a great deal and freeze when asked to speak. Producing language is a different skill with its own retrieval demands, and it improves mainly by being done, badly, repeatedly, with correction. The common plan of studying quietly for a year before attempting to speak reliably produces a year of accumulated avoidance.
What makes output productive rather than merely stressful is feedback. Speaking to nobody consolidates your errors as efficiently as it consolidates the correct forms; the value comes from someone or something telling you when what you produced was not what you meant. Writing has the same property with more thinking time, and a paragraph a day corrected by a competent reader is a genuinely strong practice for the money it costs, which can be nothing.
Shadowing — repeating audio a beat behind the speaker, matching rhythm and intonation — is a middle case. It builds articulation and prosody, which are otherwise slow to develop, and it does not build the ability to construct a sentence under pressure. Treat it as pronunciation training rather than speaking practice.
Grammar, in proportion
Two camps overstate their case. One says grammar study is unnecessary because children acquire without it; the other builds courses that are almost entirely grammar tables. The defensible middle is that explicit grammar is efficient for adults as an explanation of patterns you have already met in input, and inefficient as a substitute for meeting them.
The order that works for most adult learners is: encounter a structure in input several times, notice you cannot produce it, read the rule, then look for it in the next thing you read or hear. Ten minutes of grammar reference to resolve a pattern you have been tripping over is worth more than an hour of drills on a structure you have never encountered in the wild.
Some structures repay early explicit study because they are pervasive and unguessable — the case system in a language that has one, verb aspect where your language does not distinguish it, tone where it changes meaning. Get those explained early; leave the fine points of subjunctive usage until you can hear them.
How the review schedule fits
Vocabulary is the part of language learning that spacing was made for: discrete items, clear right and wrong, thousands of them. The scheduling principles in spaced repetition apply directly, with a few language-specific notes.
| Practice | Reason |
|---|---|
| Card the word in a sentence, not alone | Isolated pairs teach a translation, not a usage. The sentence carries the grammar, the register and the collocations. |
| Test in the direction you need | Recognizing a word when you see it and producing it when you need it are different abilities. If you want to speak, at least some cards must run from your language to the target. |
| Add audio for anything you will hear | A word you know only in writing is often unrecognizable in speech, particularly in languages where the orthography and pronunciation part ways. |
| Cap daily new items | Adding 60 a day for a week produces a review load that collapses the following month. Twenty is sustainable for most people; ten is sustainable for almost everyone. |
| Prune ruthlessly | Words you scraped from one article and will never meet again are a permanent tax on every session. |
For working out how many items a day a fixed list and a fixed deadline imply, word memorize planner does the division, and the spaced review planner lays out the review dates. The cards themselves can come from the flashcard maker.
Time, plateaus and honest expectations
How long it takes depends heavily on how far the target language is from yours — the difference between a closely related language and a distant one is a factor of several in study hours, and government language schools have published estimates that reflect this. A realistic frame for an adult studying an hour a day with real practice: recognizable progress in months, conversational competence in a familiar domain within a year or two for a close language, considerably longer for a distant one. Anyone promising fluency in weeks is selling something.
The intermediate plateau is real and it is a structural feature rather than a personal failing. Early progress is fast because the first thousand words transform your comprehension; the fourth thousand does much less per word because each item appears rarely. The plateau is what learning the long tail feels like from the inside. What gets people through it is narrowing — picking a domain you care about and going deep, so that the words you learn recur — and increasing the difficulty of your input, since comfortable material stops teaching once it is comfortable.
The habits that survive contact with a busy life are small and daily rather than large and weekly. Twenty minutes of review and thirty minutes of listening every day beats a four-hour Sunday, for the same reason spacing beats massing everywhere else. If you want the calendar built for you, the habit tracker maker and time block planner handle the scheduling; nothing handles the part where you have to talk to a person and be bad at it for a while, which is the step that separates the people who learn from the people who study.
Questions people ask
How many words do I need to have a conversation?
Word frequency is lopsided enough that a small set carries a lot of ordinary speech. Around a thousand words in frequency order gives you a footing in simple conversation about familiar topics, and two to three thousand makes everyday speech mostly followable. The important half of that answer is the phrase in frequency order: a thousand words chosen by frequency is worth far more than three thousand chosen by textbook topic, because the frequent ones keep appearing and the topic-list ones do not.
Does passive listening in the background help?
Very little if you understand almost none of it. The evidence supports input you can largely follow, where you are inferring the unknown parts from the known ones. Audio in a language you do not speak yet mostly trains you to tune it out. Once you are past the beginner stage it becomes worthwhile as extra exposure, but even then, twenty minutes of attentive listening to material at the right level does more than three hours of background noise.
When should I start speaking?
Earlier than feels comfortable. Understanding and producing are different skills, and the second improves mainly through being done badly and corrected. Learners who wait until they feel ready usually find that the waiting itself built the avoidance. What matters more than the timing is that some form of feedback exists — speaking with no correction consolidates your errors as reliably as your correct forms. Writing works too, with more thinking time, and a corrected paragraph a day is a serious practice.
Do I need to study grammar explicitly?
Some, in proportion, and best used as an explanation for patterns you have already run into. Adults learn efficiently from a rule that resolves something they have been tripping over, and inefficiently from drills on structures they have never encountered. A few features are worth front-loading because they are pervasive and impossible to guess — a case system, an aspect distinction your language lacks, tone that changes meaning. The finer points can wait until your ear can hear them.
Why did my progress stop after the first few months?
Because the first thousand words changed everything and the fourth thousand cannot. Early gains are large per word since the frequent items are everywhere; later gains are small per word since each new item appears rarely. The plateau is that arithmetic, not a personal failure. Two things move people through it: narrowing to a domain you care about so the vocabulary recurs and consolidates, and raising the difficulty of your input, since material you find comfortable has stopped teaching you anything.