MedUni Exam

Studying medicine in a second language: what to translate and what to leave alone

Studying medicine in a second language is rarely a fluency problem. It is two vocabularies for the same fact, and only one of them is printed on the exam paper. The rule that survives contact with the paper is this: read in the language of the source, test yourself in the language you will be graded in, and leave the source sentence alone. Translating your notes feels like work, and it rewrites the terms an examiner builds distractors from.

Who ends up studying in two languages

A student on an English track in Poland or the Czech Republic reads English textbooks all year, then answers a national exam written in the local language. A German student works through an English review book and takes the Physikum in German. Someone who moved abroad to study in English goes home to a final in their own language.

Reading, recall, and the sentence you cite when you argue with a question are three different jobs, and they do not all belong in the same language. Most students collapse them into one by translating everything into the exam language in the first weeks of the course, which is the decision that costs the most and the one made earliest.

Where translation costs points: drug names, eponyms, anatomy

Drug names are the clearest case. The same molecule is acetaminophen in one country and paracetamol in another, epinephrine on one paper and adrenaline on the next, and a list of five options can carry both forms. If your notes hold only the word your textbook used, you hesitate, and hesitation on a timed paper costs you a question you knew.

Eponyms split the same way. Some curricula keep the surname, others teach the descriptive name. A stem built on the descriptive form does not look like the disease you memorized under a person's name.

Anatomy adds a third problem. A course may keep the Latin term unchanged, use the local adaptation, or use the English form, and a translated note lands on whichever one the translator preferred. Units do the same: a reference range you copied in mg/dL is not the range on the paper if your faculty works in mmol/L.

None of that is careless translation. Ordinary translation smooths over exactly the distinctions an examiner builds distractors from. Translate your own recall. Never translate the source sentence.

Test in the language of the exam, quote in the language of the page

That habit has a mechanical version on this site. The question, its options, the explanation and the grading come out in one language: the one detected in your document, or the one you pick on the generate form. The quoted sentence does not follow. It is copied from your page exactly as printed, even when that is not the language of the question.

The reason is not style. Before a question reaches you, the server looks for its quote in the stored text of the page the question cited, and drops the question when the quote is not there. A translated quote is not there. A second model from a different family then answers from the quotes alone, without the key, and disagreement drops the item. Both checks are described on how it works, and neither one reads your language.

For a bilingual student that split is the useful one. You rehearse retrieval in the language of the exam, which is the skill being graded, while the evidence stays in the language of the book you actually read.

Practice questions in your own language: what each tier covers

The language picker lists 7,048 languages. Rather than claim one quality number for all of them, the language page sorts them into three tiers and says what each tier includes.

  • Tier 1, 28 languages: questions, grading, the post exam debrief and the full item writing filter, a hand written pattern list that throws out all of the above options and vague which of the following is true stems.
  • Tier 2, 11 languages: questions, grading and the debrief, with the language of your document detected for you, but no phrase level filter.
  • Tier 3, 7,009 languages: you pick the language by hand on the generate form.

What does not change between tiers is the part that matters here. The quote check, the blind second answer and the near duplicate index run in every language. The tier tells you how much of the lazy item filter is written in yours, not whether the evidence is checked.

The German and French pages were the first; fourteen exams now have a page in the language they are sat in, listed on the exams page.

Practicing on an English chapter for an exam in Polish, German or Spanish

Here is a run you can copy. Take a 40 slide cardiology deck in English and a module exam in German nine days away. A PDF chapter works the same way and cites page numbers instead of slide numbers.

  1. Upload the .pptx itself, not photographs of the slides. Each slide is stored as one slide, so every accepted question cites a slide number you can open. Speaker notes and text inside images are not read, which the slide deck page sets out.
  2. On the generate form the detected language reads English. Change the output language to German. Detection happens when the file is read, before any allowance is spent.
  3. Ask for single best answer questions on the definition slides and clinical vignettes on the management slides.
  4. Take twenty of them cold, in German. Written answers are graded in the language of the question.
  5. For each one you miss, open the slide it cites and read the English sentence printed under the question.
  6. Write down the pair and nothing else: the German term from the question, the English term from the quote.

That is one evening. It leaves you the German phrasing of twenty facts you already half knew in English, and a short list of the pairs that did not transfer. The list beats a translated deck because you produced it out of your own failures.

Medical terminology in two languages: a glossary built from misses

Two columns: the term in the language of the exam, the term in the language of the source. A third column for the quoted sentence if you have room. The discipline is that nothing goes in unless you got a question wrong.

A glossary copied out of a chapter word list is long, flat, and mostly made of things you already know. A glossary built from misses is short, and every line in it is a fact about you. After a few sessions the shape of it shows up: drug names, one or two organ systems, and the words your lecturer uses differently from your textbook. Terminology in two languages is not a general vocabulary problem but a problem in a specific handful of places, and your own misses locate them.

Review it by covering one column and producing the other, then swap sides. Reading both columns tests nothing.

When a lecturer mixes two languages in one deck

Mixed documents are normal in medicine. A German deck carries Latin anatomy and English trial acronyms. The generator is asked for one output language and told to copy the quoted sentence exactly as printed, so a Latin term inside a German chapter comes back the way the slide had it.

Detection needs prose. A page of tables, abbreviations and doses may not carry enough running text to tell two close languages apart, and an unclear result leaves the setting at Same as the document. Set the language by hand when your file looks like that.

One more case matters before you scan anything. Optical character recognition loses diacritics when it misreads a character, and a quote that no longer matches the stored page is dropped rather than shown to you. In Polish, Czech or Hungarian that shows up as fewer accepted questions from a bad scan, not as wrong ones. Scan flat, and read the first few, which is the same advice as on scanned material.

A week of bilingual review that does not double your workload

The mistake is doing everything twice. What you need is one reading and two retrieval languages, weighted toward the one on the paper.

  1. Days one to three: read the material in the language it was written in, once, and translate nothing.
  2. Days three to five: generate questions in the language of the exam from each chapter you have read, take them cold, and start the two column list from what you miss.
  3. Day six: cover the exam language column and produce the source term, then swap sides. Ten minutes.
  4. Day seven: generate a second set from the same chapters. New items are compared against everything already made from that document, so you get different questions instead of the same facts restated.
  5. The evening before: read the two column list and the questions you got wrong twice, and nothing else.

The cost of the second language is twenty minutes a day on the pairs, not a second pass through the book.

Language changes nothing about what you pay. The free plan is 300 verified questions a month with an account and 60 as a guest, in any language on the list, and pricing is the same whichever one you pick.

Questions people ask

Should I translate my lecture notes before I study them?

Translate your recall, not the page. A translated note teaches you the translator's word choice, while the distractors on the paper are built from your department's word choice. Read the original once, test yourself in the language of the exam, and write down only the terms that did not transfer.

Can I generate questions in Polish from an English PDF?

Yes. The output language is set on the generate form and does not have to match the document. The question, the options, the explanation and the grading come out in Polish, while each quoted sentence stays in the English of the page it was copied from, because the server checks that quote against that page.

Which language should I answer practice questions in?

The one the exam is written in. Retrieval is language specific, and producing a fact in English does not prove you can produce it in German under time pressure. Use the source language only for checking the quote after you get something wrong.

Do questions in my language cost more or work worse?

They cost the same. Quality does vary, because a model has seen far more English than most languages, so on a tier 3 language read the first few questions before you generate a hundred more. The quote check and the blind second answer run identically in every language.

What if my lecture notes mix two languages on the same page?

Nothing breaks. The generator is asked for one output language and told to copy quotes exactly as printed, so an English acronym or a Latin term inside a German page comes back as it was printed. That is also how the exam itself will show it.

Published , updated .