Is It OK to Use Translation When Reading in English?

In 2023, Eun Seon Chung of Hankyong National University in South Korea published an experiment on how learners use machine translation when reading English fiction (CALL-EJ, vol. 24, no. 3). The participants were 30 Korean university and graduate students, all with TOEIC scores of 785 or higher. They read three excerpts from English short stories, each 480 to 500 words long. One of them came from the Sherlock Holmes story “The Red-Headed League”, which also comes up in the post on reading Sherlock Holmes in the original.
Each participant read the three excerpts in three different ways. Original only. Original first, then an unedited machine translation into Korean. Machine translation first, then the original. Reading the translation was optional.
Many learners feel a little guilty when they open a translator in the middle of a book. This post looks at that experiment, and at research on word glosses and mental translation, to see how translation fits into reading.
Before or after the original
In Chung’s experiment, only 2 of the 30 participants never looked at the translation.
Scores on the comprehension questions did not differ significantly between the three conditions. On average, though, scores for every question type were lowest when people read the original alone.
The clear difference was in how people spent their time. When they read the original first, they spent a long time on it and less on the translation. When they read the translation first, the time on the original dropped sharply and the time on the translation went up. The original still got more time than the translation in both cases, but seeing the translation first made people lean on it more.
There was also a pattern, not statistically significant, in which people who spent longer on the original answered more questions correctly, and people who spent longer on the translation answered fewer.
From these results, Chung strongly recommends reading the original first and reading the translation afterwards, as a check. Chung also writes that raw machine translation helps with the gist and the overall context, but not so much with details, and that a deeper understanding needs a close reading of the original.
Errors in machine translation
Each translated excerpt in the experiment contained about 8 to 10 errors. None were grammatical. Most were errors of meaning: literal translations taken out of context, or awkward phrases that missed cultural knowledge.
Three of the five questions on each excerpt asked about details in the parts the translation got wrong. These were the hardest questions in every condition, including the one with no translation at all. Chung reads this as a sign that the difficulty came more from understanding and remembering small details than from being misled by the translation. Chung also notes that people who relied heavily on the translation seemed to accept its errors.
In the survey afterwards, most participants said the translation gave them a rough idea of the story and quick access to unfamiliar words and complex sentences. Some said it helped them break long sentences into chunks they could follow. The few who were unhappy with it named inaccurate and awkward translations.
The participants were advanced learners reading difficult literary texts. Chung points out that results may differ for lower-level learners, easier texts or other genres.
First-language glosses for words
Separate from translating whole passages, there is a large body of research on glosses: short notes that give the meaning of an unfamiliar word in a text.
Akifumi Yanagisawa, Stuart Webb and Takumi Uchihara combined data from 42 studies with 3,802 participants (Studies in Second Language Acquisition, 2020). With glosses, learners knew 45.3% of the target words on a test right after reading and 33.4% on a later test. Without glosses, the figures were 26.6% and 19.8%.
Glosses in the learners’ first language led to more word learning than glosses in the language they were learning. This difference did not change with proficiency, whether the learners were beginners, intermediate or advanced.
Results vary between studies, however. In a 2006 study by Makoto Yoshii (Language Learning & Technology, vol. 10, no. 3), first-language and second-language glosses on texts read in a multimedia environment showed no significant difference on either a test that asked for definitions or a recognition test. The study concludes that both kinds of gloss help vocabulary learning, but that long-term retention may differ between them.
Translating in your head
Even without an app, readers often turn the text into their own language in their heads. Richard Kern studied this “mental translation” by asking 51 intermediate-level students of French to think aloud while reading French texts (Studies in Second Language Acquisition, 1994).
Mental translation was less frequent among students with higher reading ability. At the same time, Kern argues that it can help readers build and keep meaning, because it lets them hold a stretch of foreign text that goes beyond what they can process in a more familiar form.
An order for using translation
Here is how the research above can shape the way you use a translator.
One option is the order Chung recommends: read in English first, then look at the translation to check. The translation works as an answer key.
Another is to keep reading the original closely after you have seen the translation. Chung writes that machine translation helps with the gist, but that without a careful reading of the original, learners can still be misled by its errors.
Which words to look up and when English-English definitions work better than translations have their own posts.
There is also the view that the question is not whether translation is cheating, but whether you are still giving the original its time.
Translating in ABC George
I make ABC George, a free iPhone and iPad app for reading Project Gutenberg books with a built-in dictionary. It has no ads and no in-app purchases.
Press and hold a word to bring up the selection menu, then tap “Translate”, and Google Translate opens for the selected text in a sheet inside the app. You can drag the handles to select a whole sentence or a paragraph. The dictionary sheet has a translate button too. Choose the target language from about 100 under “Translate into” in Settings; it starts as your device language. Translation needs a connection.
You can also check a word with “Look up”. The built-in dictionary shows a plain-English definition and the WordNet senses, and each word you look up is saved to your word book with the sentence you met it in. Getting to your first book is covered in Getting Started with ABC George.
