Parallel texts: help or crutch?

What counts as a parallel text?

A parallel text sets a passage in the language you are learning beside a translation in your own, either in two columns or on facing pages. It sits at one end of a range of ways to bring your first language into reading:

Most of the research is about the last two rows, because glosses are easy to control in an experiment. Studies of full parallel texts are rare, as the authors of the main one point out: bilingual books are everywhere, but very little research has looked at what they do for learners.

Do parallel texts help you learn words?

The most direct test is Zhang and Webb's 2019 study of 82 Chinese learners of English at high-beginner level. They read the same text in one of four ways: English only, English with glosses, English followed by the Chinese version, or Chinese followed by English. Twelve target words were tested before, straight after and a week later.

Both bilingual groups beat the English-only group on the delayed test. The order in which they read the two versions made no significant difference, and bilingual reading did about as well as glossed reading.

The larger body of research on glosses points the same way. Yanagisawa, Webb and Uchihara's 2020 meta-analysis pooled 42 studies with 3,802 learners. Readers with glosses learned 45.3% of the target words on the immediate test and 33.4% on the delayed test, against 26.6% and 19.8% for readers without. Glosses in the reader's first language led to more learning than glosses in the target language, and that held for beginners, intermediates and advanced learners alike.

Part of the reason is that readers left alone skip words. In Hulstijn, Hollander and Greidanus's 1996 study, advanced Dutch learners reading a French story with a dictionary looked up only 12–15% of the target words, and those without glosses simply ignored many of them. The marginal-gloss group did at least twice as well as the dictionary group.

When does a translation become a crutch?

The worry is reasonable: if the meaning is sitting right there in your own language, why work at the other one? Zhang and Webb set out to test exactly that, expecting that readers given the whole content in their first language might read the English less carefully. Their results didn't bear it out for vocabulary, but they tested only whether learners could translate the words, not whether they could use them.

The clearest warning comes from video rather than books. In Wang and Pellicer-Sánchez's 2022 eye-tracking study, 112 Chinese learners of English watched a programme with subtitles in both languages. They spent more time reading the Chinese translations than the English target words, and only the attention they gave to the English words predicted what they learned. Bilingual subtitles helped them recognise meanings, but did less than English captions for recognising the words' forms.

There is an older argument too. Hulstijn and his colleagues note earlier evidence that a meaning you infer yourself tends to be retained better than one you are handed, though inferring on your own often goes wrong. And a 2026 eye-tracking preprint, not yet peer reviewed, found that whether the translation sat in a second column or on the facing page made no difference to vocabulary scores, although the side-by-side layout drew more glances back and forth.

So the evidence is thin and mostly favourable. The crutch risk is real but behavioural: it depends on where your eyes go, not on the format.

How do you read a parallel text without leaning on it?

1. Read a paragraph in the new language first, covering the translation, and decide what you think happened.
2. Check against the translation, but only for the sentences you got stuck on.
3. Reread the original without looking across. This is the step most people skip, and it is where your attention returns to the language you're learning.
4. Wean yourself off it. When you check the translation less than once a paragraph, try a text with word-level help instead.

Here is the opening of a free Italian A1 chapter set out as a parallel text. Cover the right-hand column and see how far you get:

Then the last line of the paragraph, which the whole series turns on:

Contos doesn't print a translation beside the story. It sits one step along the range: the text is written at your level so that most of it needs no help, and you tap a word to see its meaning in your own language, which is the kind of first-language gloss the meta-analysis found most useful. If you like parallel texts, the habit to bring with you is the same either way: read first, check second, reread.