Should I read papers in English or translate them to my first language?

Read in English with in-language glosses, and use translation only for sentences that stay confusing after glossing. Your field's vocabulary, conventions and career all live in English; every paper read in translation is vocabulary you didn't acquire.

How many unknown words are too many?

Research on extensive reading puts comfortable comprehension at roughly 95–98% known words. Below that, lookup friction dominates: switch to easier material in the same field (survey papers, textbook chapters) rather than pushing through with a translator.

Won't instant lookups stop me from learning vocabulary?

Evidence points the other way: the damage comes from abandoning texts you can't finish, not from quick answers. A 2–3 second gloss lets you finish far more papers, and volume of real reading is what builds vocabulary — especially when your lookup history doubles as a review list.

What about field-specific jargon that isn't in any dictionary?

Jargon is exactly where in-context explanation beats both dictionaries and translation. A tool that reads the surrounding paragraph can tell you what the term does in *this* paper's argument — which is the answer you actually need.

Reading papers in a second language: a 3-step workflow

For ESL researchers: gloss in the paper's language first, translate only as a fallback, and turn lookup history into vocabulary — a 3-step workflow.

Key takeaways

Gloss in the paper's own language first; translate whole sentences only as a fallback — comprehension and vocabulary both grow faster
Comfortable reading needs roughly 95–98% known words; below that, pick easier texts in the same field, don't power through
Unknown words in papers split into real vocabulary (worth learning) and field jargon (worth a deep-dive) — treat them differently
Your lookup history is a personalized review list; 10 minutes a week on it beats any wordlist app
Volume wins: a 2–3 second gloss keeps you reading, and finished papers are what build reading speed

The short answer

If you're reading papers in a second language, the workflow that builds both comprehension and vocabulary is: gloss in the paper's language first, translate only as a fallback, and review your lookup history weekly. The goal is not to make any single paper easy — it's to make the fiftieth paper faster than the first.

This post covers the second-language angle; the general strategy is in the complete guide to reading papers with unfamiliar terms.

Why second-language reading is a different problem

Native readers stall on terms; second-language readers stall on terms plus vocabulary plus syntax — often in the same sentence. That changes the math in three ways:

Lookup volume is 2–3× higher. A native reader might gloss 10 terms per paper; you might hit 20–30 unknown items. At 30–60 seconds per browser-tab lookup, that's the entire reading session gone.
The 95–98% threshold is real. Extensive-reading research consistently finds you need to know roughly 95–98% of the words on a page for comfortable, unassisted comprehension. Below that, you're not reading — you're decoding.
Translation is a trap at this volume. Translating whole paragraphs gets you through today's paper and leaves you exactly as unprepared for next week's. The field's vocabulary lives in English (or whatever the paper's language is), and every translated page is vocabulary you didn't acquire.

The 3-step workflow

Gloss in the paper's language first. When a word or term stops you, get a short explanation in English (or the paper's language), not a translation. An explanation that reads the surrounding sentence resolves both senses of "I don't know this": unknown vocabulary and unknown jargon. PopGloss's language-learning style is built for exactly this — it explains the term simply, in context, in 1–3 sentences, without leaving the PDF.
Translate only as a fallback. If a sentence is still confusing after its terms are glossed — tangled syntax, dense subordinate clauses — then translate that one sentence. You've now spent translation where it helps (structure) instead of where it hurts (vocabulary you needed to learn).
Review your lookup history weekly. Ten minutes, once a week, over the terms you actually looked up. This list beats any generic wordlist app because every item on it is a word a real paper in your field made you need. PopGloss stores lookup history locally on your Mac (searchable, never uploaded), so the list builds itself.

Split unknowns into two piles

Not every unknown deserves the same treatment:

Type	Examples	Treatment
Real vocabulary	"parsimonious", "confound", "prune"	Gloss in place, add to weekly review — these recur across all papers
Field jargon	"heteroscedasticity", "ablation", "epistasis"	Gloss now; if it recurs 3+ times, it earns a deep-dive (a textbook section, a survey)
One-off noise	Author names, dataset names, venue acronyms	Skip entirely — they don't affect the argument

The piles matter because they decay differently. Real vocabulary pays off across every paper you'll ever read. Jargon pays off inside your subfield. One-off noise never pays off — skipping it is a skill, not a shortcut.

What progress actually looks like

Weeks 1–4: 20–30 glosses per paper; heavy reliance on the fallback translation for methods sections
Weeks 5–10: glosses drop to 10–15; fallback needed mostly for long syntactic tangles
After a semester: 5–10 glosses per paper, mostly new jargon; the vocabulary pile has mostly become permanent

The mechanism is simple: field English is a finite vocabulary, and papers reuse it relentlessly. Every gloss you review is a word you'll never look up again. The readers who get fast are not the ones with the best memories — they're the ones whose lookup friction was low enough that they finished enough papers for the repetition to work.

Try it yourself: Download PopGloss — 10 free credits included, never expire.