What are graded readers, and do they work?

How are graded readers levelled?

Publishers grade their books by headwords, the number of word families a level is allowed to use, chosen mostly by frequency. Rob Waring gives the example of a 1,200-headword reader, in which very few words fall outside the 1,200 most frequent families. Grammar is controlled as well, mainly by sentence complexity and which verb forms appear.

David Hill's 1997 survey for the Edinburgh Project on Extensive Reading (EPER) found that most series had four to six stages, with books running from about 1,000 words at the first stage to 20,000 or more at the top. To compare publishers, Hill put them on one eight-level scale.

There is a catch. Waring warns that publishers use different word-frequency lists, especially at the higher levels, so level 3 in one series isn't necessarily level 3 in another.

Do graded readers actually work?

Most of the evidence comes from studies of extensive reading, an approach that, as the Extensive Reading Foundation puts it, often makes use of graded readers. Jeon and Day's 2016 meta-analysis pooled 71 samples from 49 studies published between 1980 and 2014. It found a small to medium effect on reading comprehension, reading rate and vocabulary, with bigger effects for adults than for children and in foreign-language settings than in second-language ones.

Results that look at simplified books specifically are more mixed. Paul Nation's 1997 review describes learners whose writing improved after reading simplified texts, although one group's written vocabulary didn't grow, probably because the books sat far below what those learners already knew. Nation also measured how tightly a graded reader controls its language: in the one he analysed, 99% of the words came from the 2,000 most frequent plus names, much tighter than the native-speaker stories he compared it with.

Graded readers are not magic, either. Krashen describes a university class in Taiwan in which only 18.5% of students found their graded readers interesting. The readers still beat a traditional class on a cloze test, but a group taught vocabulary directly did better on less common words, which the graded readers rarely contained.

Hill's verdict, drawing on some 20 years of EPER data, was that graded readers are "a superb language learning resource" when used systematically over a long period.

Aren't graded readers too artificial?

It's the commonest objection. Bamford and Day's answer is that a text is authentic when it communicates with the audience it was written for, and learners are a real audience. They call the best books written for learners language learner literature, and argue it is the most suitable material for beginners and intermediate readers.

Quality varies, though. Hill criticised weak adaptations that read like summaries, rushing through plots and characters, and found that a series' own original stories could be livelier than its simplified classics. Most beginner and elementary titles are originals; most upper-level ones are adaptations.

Native-speaker books have a cost too. Waring describes learners who slog through a classic far above their level and never read in the language again.

How do you pick the right level?

Waring's rule of thumb: you should be able to read a page in about two minutes and understand all but two or three words, not counting names. He has learners find the level that feels right and then go down one.

Contos works like a graded reader in instalments. Each series is written at one CEFR level, and its chapters are short. Instead of headword lists, Contos moves you up a level based on how many words you know.