Ever had a go at translating something? It’s trickier than it looks. You might assume that a translated text is simpler than the original, but some new research suggests the story is far less straightforward.
The study
Researchers compared 500 native English texts (from the FLOB corpus) with 500 English texts translated from Chinese [the COCE, or Corpus of Chinese into English, which doesn’t seem to be publicly available online]. All texts were balanced across four genres: news, general prose, academic writing & fiction. To measure complexity, they used entropy, a concept from information theory that captures unpredictability in data. Here, they examined:
- Wordform entropy, or lexical complexity: How varied & evenly distributed are the words?)
- POS entropy, or syntactic complexity: How varied is the distribution of grammatical categories (or Parts of Speech)?
All texts were trimmed to 1,500 words to avoid length-related distortion. POS tagging was done with the Penn Treebank tagset [interestingly & tangentially, ‘to’ & ‘existential there’ are the only two words to have a category all to themselves], then entropy was calculated using Shannon’s formula [well above my paygrade].
The findings
Contrary to the long-standing simplification hypothesis, translated English was lexically more complex, not simpler:
Translated texts showed significantly higher wordform entropy, especially in news & fiction.
- No significant difference was found in POS entropy, meaning syntactic complexity remained similar across translated & non-translated texts.
- Genre played a strong role: fiction & news were the most complex lexically, academic prose the least.
- Translators (all L1 Chinese, working into English as L2) tended to use more unique word forms, not fewer.
This challenges classic work suggesting translated texts are lexically simpler (Laviosa 1998; Olohan 2004) & aligns more with newer perspectives on explicitation & source-language “shining through”. Chinese has relatively high lexical entropy, so its patterns may subtly imprint onto the English translations.
To visualise this: think of an original English news article using predictable high-frequency vocabulary. A translated version might choose broader or more varied wording to convey nuances—like opting for “precautions”, “protective measures” & “preventative steps” where an original might just repeat “measures”.
I’d also see this as a bit of a bonus for non-native speaker teachers of English, who are often unjustly accused of “watering down” the language. These findings offer useful counter-evidence: the L2 translators in the study actually produced more lexically complex English than native writers, suggesting that being a non-native professional can enrich rather than reduce linguistic variation.
Teacher Takeaways?
- Get translating! Have 2 short texts- 1 in English & 1 in the learners L1. Ask different groups of learners to translate them, swap them over & repeat. Finally, regroup learners to look for interesting differences between the input & the output.
- Vocabulary scaffolding matters. When using translated materials, pre-teach/focus on key lexical sets: translators often introduce a wider range of synonyms.
- Genre awareness helps learners read strategically. Students can expect denser vocabulary in news or fiction, regardless of translation status.
Do you ever include translation in class?
Leave a Reply