How to Choose the Right Texts for Comprehensible Input

Choosing the text is the decision that determines whether reading works for you at all. Everything else — how long you read, how you handle unknown words, what you review afterwards — operates inside the margin that the choice of book leaves you. A well-chosen text makes an ordinary reader look disciplined. A badly chosen one makes a disciplined reader look like someone who has no aptitude for languages.

The number that governs everything

Vocabulary research gives a usable threshold. To read on your own with good comprehension you need to know somewhere around 98 per cent of the running words on the page: about one unknown word in fifty, which on a normal page of prose is two or three. At about 95 per cent — one in twenty — comprehension is still possible but the work becomes visible, and support of some kind is needed to keep it pleasant. Below that, comprehension falls away quickly, and so does the ability to guess unknown words from context, because the context is made of words you also do not know.

Two things follow that are worth stating plainly. The first is that comprehensible is a much higher bar than eventually decipherable. The second is that this threshold is about coverage of running words, not of distinct words, so it is far less demanding than it sounds: the few hundred most frequent words of any language make up half of almost any text.

Testing a text in ninety seconds

You do not need software. Open the book at a page somewhere in the middle — not the first page, which authors write more carefully than the rest — and read it.

• Count the words you do not know. More than about five on a full page and you are below the level where unassisted reading is comfortable. The old classroom version of this is the five-finger test: a finger for each unknown word on one page, and if the hand closes, the book is too hard.
• Ask what happened. Close the book and say, in your own language, what the page was about. If you cannot, the unknown words were the wrong ones — they were carrying the meaning.
• Notice the pace. A page that takes three times as long as a page in your own language is fine. Ten times is not: at that speed you are decoding, and you will not sustain the habit for a month.

Do this with a page rather than a blurb. Cover copy is written by marketing departments in a register of its own and tells you nothing about the prose.

Level in words, roughly

The CEFR describes what a learner can do, not how many words they have, but the rough correspondences are useful when reading a publisher’s catalogue. A2 tends to imply something like 1,500 word families, B1 around 2,500 to 3,000, B2 in the region of 5,000, with wide variation between individuals and between languages. Unsimplified adult fiction generally asks for 8,000 to 9,000 word families for that 98 per cent coverage — which is why the gap between “I finished a B1 course” and “I can read a novel” is not a small step but several years of input.

The purpose of graded material is to occupy that gap with real reading instead of waiting.

The sources, and what each is good for

Graded readers

A graded reader is a text written or rewritten to stay inside a controlled vocabulary and a limited range of structures. The tradition goes back to Michael West’s work in the 1920s and is now a small industry, with series from most of the large educational publishers and a body of research — the extensive reading literature — behind it.

Their advantage is exactly the one that matters: the coverage figure is engineered rather than hoped for. Their weakness is that a controlled vocabulary can flatten a text into something no adult would read voluntarily, and a reader who is bored stops, which returns you to zero input. The distinction worth making when choosing is between a text that has been simplified — cut down, its plot summarised away — and one that has been rewritten, where the structure, the irony and the ending survive and only the language moves. The second is worth reading twice. The first is worth reading never.

Children’s books

The obvious choice and a frequent disappointment. Books for young native speakers assume a native five-year-old’s vocabulary, which is nothing like a foreign adult’s: they are full of animal names, nursery idiom, diminutives, invented words and dialect. They are also, for most adults, unreadable for reasons that have nothing to do with difficulty. Books for older children — nine to twelve — are often a better bet than picture books, and translations of books you already know are better still.

Texts you already know

A novel you have read in your own language, a folk tale, a film you can recite: prior knowledge of the plot supplies context in advance, which raises effective comprehension without lowering the language. This is the cheapest way to read something slightly too hard and get away with it.

Non-fiction on a subject you know

Domain knowledge does the same work. An engineer reading about engineering in Spanish knows what the sentence is likely to say before decoding it, and technical vocabulary is often transparent between European languages. The register is usually simpler than literary prose, too.

Original literature

The destination, not the vehicle. There is no virtue in arriving early: reading Flaubert with a dictionary at B1 produces about forty pages of misery and no acquisition. Nineteenth-century prose in particular is harder than its reputation — long periods, obsolete idiom, a wide lexicon — which is why the same stories work so well when the language is lowered and the structure kept.

Narrow reading

One of Krashen’s most practical suggestions has nothing to do with level. Read narrowly: several books by one author, or several texts on one subject, rather than a scattering across the language. Each writer has a limited lexicon and a set of habits, so the second book by an author is markedly easier than the first, and the fourth is easy. Vocabulary repeats, which is what makes it stick, and the effective level of the text drops while the printed level stays the same.

A collection of stories by one writer is a small version of the same effect. It is one of the reasons these volumes group a handful of authors rather than sampling twenty.

Interest is a technical requirement

It is tempting to treat enjoyment as a soft consideration next to the hard arithmetic of coverage. It is not. The quantity of input you get is the product of your reading rate and the number of hours you actually read, and the second factor is entirely governed by whether you want to turn the page. A book at a perfect level that bores you delivers no input at all, because it is not read.

The corollary is permission to abandon books. A text you are not enjoying at page thirty is not a debt. Put it down and take another; nothing in this method depends on finishing any particular volume.

When to move up

Two signals, and they should both be present. You are reading the current level at something close to your normal reading pace, and you are finishing pages without noticing the language — the sensation of having read rather than translated. If either is missing, another book at the same level costs you nothing and is doing more for you than a harder one would.

When you do move, move once and only in one dimension. A new level, a new author, a new period and a new genre at the same time is how people conclude they have plateaued. Change the level and keep the author, or change the author and keep the level. And expect a dip: the first fifty pages of a new level always feel worse than the last fifty of the old one, and that feeling is not evidence.

Once the text is right, the remaining question is what to do with it: read a great deal quickly, or a little slowly and thoroughly. Both, in a particular ratio — and here is the whole routine.