How to Read in a Foreign Language Without Knowing Every Word

Most people who give up on reading in a foreign language do not give up because the book was too hard. They give up because of what they did about the words they did not know. Somewhere on the second page they looked one up, and then another, and after forty minutes they had covered a page and a half and remembered none of it. The book goes back on the shelf, and the conclusion drawn is that the reader is not ready.

Usually the reader was ready. What was missing was a way of handling the unknown that does not destroy the reading while it works.

Why looking everything up fails

A dictionary lookup costs between fifteen and forty seconds, and that is the smaller half of the price. The larger half is that it empties the sentence out of your head. Working memory holds a clause in a foreign language with some effort; interrupt it and the effort has to start again. Ten lookups on a page do not slow your reading by ten times fifteen seconds — they prevent you from ever building the thing that reading is for, which is a continuous representation of what is happening.

There is a second cost, less obvious. A word met once in a dictionary is a word met in a list. A word met four times across two stories, in different sentences, is a word you are beginning to own. Vocabulary from reading is acquired in small increments across many encounters, and stopping to look up every unknown word trades a large number of shallow encounters for a small number of shallow encounters plus a definition you will have forgotten by the next chapter.

The aim is not to abolish the dictionary. It is to spend it where it buys something.

The one rule: finish the sentence

When you meet a word you do not know, keep going to the end of the sentence. Then ask a single question: do I still know what is going on?

If the answer is yes, read on. That is the whole procedure. The unknown word has done its job — you have met it once in a context you understood, which is exactly the kind of encounter that builds vocabulary. It will come back.

If the answer is no, the next question is whether the sentence matters. A description of a coat is usually not load-bearing; the sentence in which somebody agrees to a duel usually is. Skipping a paragraph of scenery costs nothing. Skipping the turn of the plot means the next five pages will not make sense, and you will pay for it with far more than one lookup.

Which words are worth stopping for

Three kinds of word repay the interruption, and almost nothing else does.

• Words that carry the plot. If you cannot tell whether the character accepted or refused, stop. Comprehension of the story is the thing that makes all the rest of the input comprehensible.
• Words you have now met three times. A word that keeps returning is frequent in this writer, this topic or this language. The third encounter is the moment when a definition will stick to something, because you already have three contexts waiting for it.
• Words that block a whole sentence. Not one unknown word among known ones — a word in a sentence you have otherwise reconstructed and which still refuses to resolve.

Everything else can wait. A useful discipline is to mark, not look up: a pencil dot in the margin, or a tap on a reader, and carry on. At the end of the chapter you will find that half the marked words were explained by the following pages and the other half are worth ten minutes.

Guessing from context, and when it stops working

Inferring a meaning from its surroundings is the mechanism the whole approach depends on, and it is worth knowing what it actually needs. It needs the surroundings to be known. You can guess the coachman flicked the reins and the horses moved off if the only unfamiliar word is reins. You cannot guess anything at all if three of the other words are also new, because the context you would be guessing from is not there.

This is the practical face of the coverage figures from vocabulary research: unassisted reading works when you know about 98 per cent of the running words — roughly one unknown in fifty, or two or three per page — and degrades quickly below 95 per cent. Above that threshold, guessing succeeds often enough to be worth doing. Below it, guessing is a euphemism for inventing, and inventing reliably teaches you the wrong meaning.

So if you find yourself guessing constantly and getting nowhere, the problem is not your tolerance for ambiguity. It is the text, and the fix is to change it rather than to try harder.

Two smaller cautions. Cognates are generous and occasionally treacherous, in proportion to how closely the languages are related — French actuellement does not mean actually. And a guess you have made once tends to stick, right or wrong, which is a good argument for confirming the words that matter rather than living with a private and slightly wrong lexicon.

What makes the unknown cheaper

Some of this is the reader’s skill; a good deal of it is a property of the material.

A gloss on the page. A footnote where the difficulty is costs perhaps two seconds and does not empty the sentence out of your head. That is a different order of interruption from a dictionary, and it is why the volumes in this series put the footnotes on the page rather than in an appendix at the back.

Rereading. The second pass over a story is nearly free and unreasonably productive: the plot is known, so all your attention goes to the language, and words you skipped the first time now resolve on their own.

Audio. Listening to a text you have read attaches a sound to the words you acquired silently, and often clarifies a sentence whose structure you had been reading wrongly. It also imposes a pace, which is a cure for the habit of stopping.

Knowing the story already. The oldest trick in extensive reading: a book you have read in your own language, or a story whose shape you can predict, supplies the context in advance. This is one reason classic short fiction is such good material — much of it you half know before you begin.

The languages complicate this differently

How expensive an unknown word is depends on what it takes to find it. In French or Spanish, a word on the page is usually close enough to its dictionary form that looking it up is mechanical, and the cost is mostly the interruption.

In Russian it is not. A noun on the page may be in one of six cases, a verb in an aspect pair you have to recognise before you can search, and the form in front of you may share no visible ending with the entry you need. For a learner below B2 this turns a fifteen-second lookup into a small research project, several times a page. Reading past the unknown is not a nicety in a heavily inflected language; it is the only way the activity remains possible at all, which is why glossed and graded texts pay off there earlier than anywhere else.

Tolerating ambiguity is a trained skill

Learners often describe their discomfort with unknown words as a personality trait: I need to understand everything. It behaves much more like a habit, and it responds to practice. Two exercises are enough for most people.

1. Read one story with the dictionary closed. Not permanently — once. Use only the footnotes, and accept that you will miss things. At the end, read a summary and see how much you actually had. The gap is almost always smaller than it felt, and that discovery does more than any amount of advice.
2. Time yourself for ten minutes. Read for ten minutes without stopping, then note where you got to. Repeat on subsequent days with the same book. Watching the distance grow makes the cost of interruptions concrete in a way that nothing else does.

This is the reading behaviour that comprehensible input asks for, and it is only half the picture: there is a place for slow, close, dictionary work too, provided it happens on its own terms. Reading widely and reading closely do different jobs.