Verbal Reasoning in Your Second Language: Strategies for Non-Native Speakers

There is a specific kind of panic that sets in around question fourteen. You have spent too long on a passage about comitology, the clock says eleven minutes remaining, and you realise you have read the same sentence three times without absorbing it. Your English is good — demonstrably good; you work in it, you studied in it, EPSO required C1 and you are well above that. And yet here you are, losing.

This is the experience most EPSO candidates have, because most EPSO candidates sit verbal reasoning in a language they did not grow up speaking. It is worth understanding exactly why — because the reason is not the one most people assume, and the fix is not “improve your English.”

First, the rules as they actually are

Before strategy, facts. A great deal of outdated and simply incorrect information circulates about EPSO language rules, including on sites that sell preparation material. Under the AD5 2026 competition notice (EPSO/AD/427/26, published in the Official Journal on 5 February 2026), the language regime works as follows:

• You must have thorough knowledge — minimum C1 — of one of the 24 official EU languages. This is Language 1.
• You must have satisfactory knowledge — minimum B2 — of a different language, chosen from the remaining 23. This is Language 2.
• The minimum levels apply to each linguistic ability declared: listening, reading, spoken interaction, spoken production, and writing.

And critically, the language each test is taken in:

Two corrections to widespread misinformation. First, Language 2 is not restricted to English, French or German in this competition — the notice states plainly it is chosen among the remaining 23 official EU languages. Second, all three reasoning tests are taken in Language 1, not just verbal. Some sources claim numerical and abstract are sat in Language 2; the notice lists “Reasoning tests — Language 1” as a single row. This matters for your strategy, and we will come back to it.

Why competent speakers underperform

The EPSO verbal reasoning test is not a language test. EPSO does not describe it as one; it assesses your ability to think logically and understand verbal information. But it is delivered through language, under time pressure — and that combination produces a specific failure mode.

When you read in a second language, even at a high level, you perform a small amount of additional cognitive work per sentence. Not much — marginally slower on low-frequency vocabulary, a fraction less automatic on a nested subordinate clause. In ordinary reading this is invisible; you have time, you reread, nobody notices. Under a timer, that small tax compounds. The AD5 verbal test gives you 20 questions in 35 minutes — 105 seconds per question, including reading the passage. There is no slack. The extra processing load does not just slow you down; it consumes working memory you need for the reasoning part. You end up understanding the passage but with insufficient capacity left to hold its logical structure in your head while you evaluate four options.

The five trap patterns

EPSO verbal questions present a passage — often on an EU policy or institutional topic — followed by a question with four options, exactly one supported by the text. The wrong options are constructed, and they exploit a small number of recurring patterns. Learn to recognise them and you convert a reading problem into a pattern-matching problem, which is far cheaper cognitively.

1. The absolute qualifier

The passage says something measured; the distractor says something absolute. “Generally” is not “all.” “Most” is not “every.” “May” is not “will.” “Contributes to” is not “causes.” For non-native speakers this trap is disproportionately effective, and the reason is instructive: quantifiers and modals are exactly the words your brain deprioritises in L2 — short, unstressed, functional rather than semantically rich. You skim past them to reach the content words. The trap is built precisely where your attention is thinnest.

Countermeasure: train yourself to physically notice quantifiers and modal verbs. When you practise, flag every all, some, most, always, never, may, must, can, should, generally, typically, in principle. Make it mechanical; within a few weeks it becomes automatic.

2. The scope shift

The passage establishes something about a limited domain; the distractor quietly widens or narrows it — along who (financial sector → all sectors), when (during the transition → permanently), where (in the euro area → across the EU), or what (reporting obligations → all obligations). EU texts are unusually rich in scope qualifiers, because EU law is unusually precise about jurisdiction. A phrase like “in respect of institutions established in the euro area” is doing real legal work. In your first language you register that clause as load-bearing; in your second, there is a tendency to compress it into background texture.

Countermeasure: for each option, ask one question — does this claim cover exactly the same set of things as the passage? Not “is it true.” Same set, or different set.

3. “Not stated” versus “false”

This is the single largest source of lost marks, and it is a logic error rather than a language error — which is why strong speakers fall for it just as often. The task is to identify what can be logically inferred from the passage. Not what is true in the world; not what is plausible; not what you happen to know from reading the Treaties.

Candidates with genuine EU knowledge are at particular risk, because their knowledge fills gaps the passage left open. If you have been studying EU institutions intensively — and if you are preparing for this exam, you have — you must actively suppress that knowledge during verbal reasoning. The passage is the only universe that exists.

Countermeasure: before selecting an option, point to the specific sentence(s) that support it. If you cannot point to text, it is not the answer, no matter how obviously true it seems.

4. The reversed relation

The passage establishes a direction; the distractor flips it. A led to B is not B led to A. X reports to Y is not Y reports to X. The Council amended the Commission's proposal is not the Commission amended the Council's proposal. This one has a language-specific edge: passive constructions and prepositions of sequence — following, prior to, subsequent to, pending, upon — are handled less automatically in L2, and English legal-administrative register uses them constantly.

Countermeasure: when a passage describes a relationship between two things, note the direction with an arrow in your head before you look at the options. Then check each option against your arrow.

5. The unsupported causal link

The passage reports two facts; the distractor connects them causally. Correlation and temporal coincidence are not causation, and the passage did not claim it. Watch for options that introduce because, as a result of, due to, therefore, consequently, which led to where the passage merely used and or a full stop.

Choosing your Language 1

Given that all three reasoning tests are sat in Language 1, this decision deserves more thought than most candidates give it. The instinct many have is to pick English because it feels professional. For the testing phase of the AD5 2026 competition that is simply not required, and choosing on prestige rather than performance costs real marks.

The question to ask is narrow: in which language do I read fastest under pressure without loss of precision? Speed alone is not enough — you can read quickly and shallowly. Precision alone is not enough — you can be accurate and slow. And “under pressure” is not the same as reading comfortably on a Sunday afternoon. For most candidates the honest answer is their mother tongue, and for verbal reasoning that is usually correct: reading in L1 costs nothing and leaves your full working memory available for inference.

There are genuine counter-considerations. Practice-material availability: EPSO sample tests have been in transition and are strongest in English and French, so a less-served language may mean practising in one language and testing in another. Translation artefacts: items are authored in one language and translated; the notice allows you to request review within three calendar days if a specific MCQ error affected your answer — but complaints “merely pointing out alleged issues of translation, without specifying the problem” are not taken into account, so you must describe the actual error. Your professional reading language: if you have read policy documents in English for a decade and rarely read formal registers in your mother tongue, your fastest language for this genre may not be your first. Test it empirically — take timed sets in each candidate language and compare accuracy and completion. Whatever you decide, decide before the deadline: you cannot change your application afterwards.

A timing method that survives contact with the test

The budget is 105 seconds per question. Here is how to spend it.

• Read the question first, then the passage. The single highest-value habit change available. Knowing what you are looking for turns reading from comprehension into search — far cheaper.
• Do not read the options before the passage. Options are designed to be plausible; reading four plausible statements before the source primes you to see them in the text. Question first, passage second, options third.
• Set a hard ceiling and enforce it. At 100 seconds, choose your best remaining option and move. There is no negative marking in the notice, so an educated guess strictly dominates a blank. Answering all 20 at 65% beats answering 15 at 80%.
• Do a first pass and a second pass. Resolve everything you can confidently on the first pass; flag the rest and return with your remaining time. This prevents one hard question at position seven eating four minutes so you never see questions eighteen to twenty.
• Practise with the timer from day one. Untimed practice builds the habit of rereading — exactly what you cannot afford. Fifteen minutes today? Do a timed set of eight, not an untimed set of twenty.

Vocabulary that recurs in EU texts

EPSO passages draw heavily on EU institutional and policy subject matter. The vocabulary is finite and repetitive; learning it removes a large share of the lookup cost from your reading.

Legislative and procedural. Ordinary legislative procedure, delegated act, implementing act, comitology, first reading, trilogue, Conciliation Committee, qualified majority, unanimity, subsidiarity, proportionality, transposition, infringement procedure, reasoned opinion, derogation, opt-out, recital, article, annex, entry into force, applicable from.

Institutional. Commission, Council of the European Union (not the European Council, and not the Council of Europe — three different bodies), European Parliament, Court of Justice of the European Union, General Court, European Court of Auditors, European Central Bank, European External Action Service, Committee of the Regions, European Economic and Social Committee, agency, joint undertaking, Directorate-General, College of Commissioners, rapporteur, presidency.

Budgetary and financial. Multiannual Financial Framework, own resources, commitment appropriations, payment appropriations, discharge, cohesion policy, structural funds, co-financing, headings, mobilisation, contingency margin.

Hedging and qualifying language — the vocabulary that carries the traps. Notwithstanding, without prejudice to, subject to, insofar as, in principle, as a rule, where appropriate, mutatis mutandis, inter alia, pursuant to, in accordance with, by way of derogation, save as otherwise provided. “Without prejudice to X” means X remains unaffected — not that X is unimportant. “Subject to Y” means conditional on Y. “In principle” signals that exceptions exist. These phrases do precise logical work and appear constantly in both passages and options.

Register-mismatch pairs — where administrative English differs from general English. Shall (obligation, not future tense) · may (permission or possibility) · provide for (establish in law) · address (a decision addressed to a Member State) · communication (a formal Commission document) · opinion (a formal instrument) · measure / instrument (a legal act) · party (to an agreement).

A three-week practice structure

Week 1 — untimed accuracy. Work through questions with no clock. For every one you get wrong, write one sentence naming which trap caught you. The act of naming the error is what makes it recognisable next time.

Week 2 — introduce the clock. Sets of ten, timed at 105 seconds each. Expect accuracy to drop; that drop is the information you are after — the size of your time-pressure penalty, and whether your problem is comprehension speed or inference reliability.

Week 3 — full simulation. Complete 20-question sets in 35 minutes, one sitting, no interruptions. Because the AD5 tests are proctored remotely, practise in the environment you will test in — your own machine, your own desk, webcam on. Throughout, keep a running list of vocabulary you had to slow down for; that list is worth more than any generic one because it reflects your actual gaps.

The strategic point about weighting

One final thing, because it should shape how you allocate preparation time across the whole competition. In the AD5 2026 competition, verbal reasoning does not merely gate your progression — it is the single most heavily weighted test in the entire competition. Under the notice it carries 40% of the preliminary combined score and 35% of the final combined score. No other test weighs more: EU knowledge and digital skills each carry 30% preliminary and 25% final, and the EUFTE carries 15% of the final score only.

By contrast, the numerical and abstract reasoning tests are not included in either combined score at all. You need a combined 10/20 across them to proceed, and beyond that threshold additional marks earn you nothing. Numerical and abstract are pass/fail hurdles; verbal reasoning is where your ranking is built. For a candidate sitting the test in a second language, that is either bad news or a considerable opportunity — depending entirely on whether you treat verbal reasoning as the priority it actually is.

If you are still weighing which route into the institutions to target, our guide to AD vs AST vs CAST career paths maps every option, with the official salary figures.

This article reflects the rules in Notice of Open Competition EPSO/AD/427/26, published in the Official Journal of the European Union C/2026/711 on 5 February 2026. Test structures, weightings and language regimes differ between competitions and change over time. Always verify against the notice governing the competition you are sitting — where any summary and the official notice diverge, the notice and its Annex I “General rules” prevail.

Frequently asked questions