3.4 Reading Task Formats, Text Length & Item Density
Key Takeaways
- EF states that each test taker answers approximately 50 reading questions, that the number of reading prompts varies, and that texts run 1 to 5 paragraphs.
- A single text carries 6 questions on average, with a published range of 1 to 12, so a long passage is worth far more of your section than a short notice.
- EF's published task-design specifications use two reading task models: a single passage with multiple-choice or multiple-option questions plus completion items, and a fixed-format task pairing two short passages with six statements.
- Text length scales with CEFR level in EF's design specifications, from roughly 100 words at A1 to 450 to 600 words at C1, which is why the passage in front of you is itself a signal of how you were routed.
- The first reading prompt is set at intermediate difficulty, and the next prompt becomes harder or easier depending on how you handled it.
Reading Task Formats, Text Length & Item Density
Quick Summary: Most candidates prepare reading skills and ignore reading packaging. That is a mistake on an adaptive test, because the EF SET does not deliver 50 independent questions — it delivers a handful of texts, each carrying a cluster of items. EF publishes the shape of that packaging, and the numbers change how you should spend your 25 minutes.
The Published Numbers
Four facts come straight from EF's description of the reading section:
| Published fact | Value | Why it matters to you |
|---|---|---|
| Questions per test taker | Approximately 50 | Averaged flat across 1,500 seconds, that is about 30 seconds per question. |
| Number of prompts | Varies by session | The adaptive panel decides how many texts you see; there is no fixed count to pace against. |
| Questions per text | 6 on average, range 1 to 12 | A single text can be worth a fifth of the section — or almost nothing. |
| Text length | 1 to 5 paragraphs | Nothing on this test is a full-page academic article. |
EF also states that "individual tasks on the reading section are not timed, so test takers have as much time as they need to read each text and answer questions, but the entire reading section is limited to 25 minutes." The only clock that can hurt you is the section clock.
The Density Rule
Because item density varies from 1 to 12 questions per text, read the item count before you read the text. The economics are stark:
A 5-paragraph text carrying 10 questions
→ worth ~20% of the section → a full structural skim is justified
A 1-paragraph notice carrying 1 question
→ worth ~2% of the section → answer it in 35 seconds and move on
Investing four minutes in a text that carries one question is one of the most expensive mistakes available on this exam, and it is invisible unless you check the density first.
The Two Reading Task Models
EF's academic and technical development report specifies two task-model families for reading. Both survive in recognisable form in the live test.
1. Single Passage with Multiple-Choice or Multiple-Option Questions
One continuous text followed by a cluster of items. At higher levels EF's specifications add completion items — sentence or summary completions — alongside the multiple-choice set. This is the workhorse format and the one your two-phase skim protocol from Section 3.1 is built for.
2. Fixed-Format Paired Passages
Two short passages presented together, followed by a fixed set of six statements that must each be assigned to one of a small number of fixed choices — for example, Passage A, Passage B, both, or neither. This format tests comparison and contrast rather than sequential comprehension, and the strategy is different:
- Read both passages for stance and claim, not for detail.
- Build a two-column mental note: what does A assert, what does B assert.
- Take each statement to the columns rather than back to the passages.
- Watch for statements that are true of the world but asserted in only one of the two texts — that is the format's characteristic trap.
How Length and Item Count Scale With Level
EF's published design specifications show text length and question count rising together across the CEFR range. These figures describe the design grammar EF used to build its item families, and the pattern is what matters:
| Target level | Single-passage length | Items on that passage | Paired-passage length (each) |
|---|---|---|---|
| A1 | ~100 words | ~4 questions | ~50 words |
| A2 | ~150–200 words | ~5 questions | ~100 words |
| B1 | ~350 words | ~6 questions | ~150 words |
| B2 | ~450–500 words | ~6 questions plus a completion | ~150–200 words |
| C1 | ~450–600 words | ~8 questions plus 2 completions | ~150–200 words |
Read this table as a routing signal. If the third text you meet is a 500-word argumentative passage with eight items, the panel has routed you upward. If your texts stay at notice length with one or two items each, it has not. Section 7.3 explains what to do with that information; here the point is simply that the stimulus itself tells you where you are.
[!NOTE] Spelling conventions. EF specifies that reading passages use standard American or British spelling and punctuation, without regional variants. Do not treat organise/organize or colour/color as a clue to anything — both are in scope and neither signals a trap.
What the Questions Are Actually Testing
EF's test blueprint lists the constructs assessed by every reading task. All six appear on the certificate-level test, and each maps to a technique you have already met:
| Construct in EF's blueprint | What the item asks you to do | Where it is trained |
|---|---|---|
| Gist | Identify the overall sense of a text | Section 3.1 |
| Major ideas | Identify the controlling claim of a paragraph or section | Section 3.1 |
| Specific details | Retrieve a stated fact, figure or condition | Section 3.2 |
| Inference | Derive an unstated but entailed conclusion | Section 3.3 |
| Sensitivity to rhetorical organization | Recognise how the text is built — preface, concession, conclusion | Sections 2.3 and 3.1 |
| Sensitivity to cohesion | Track reference, substitution and connectives across sentences | Section 2.3 |
EF also documents the levers it uses to make an item harder. Knowing them tells you where the difficulty in front of you is coming from:
- Amount of text in a fixed time and increasing syntactic complexity
- Increasing breadth and complexity of vocabulary
- Movement along the concrete-to-abstract and familiar-to-unfamiliar topic continua
- Distance from a direct reference in the stimulus to the key — the correct answer stops being a paraphrase of one sentence
- Closeness of the distractors — wrong options converge on the right one
- Inference on rhetorical features (tone, attitude, structure), which EF identifies as the hardest category of all
[!TIP] Diagnose before you struggle. When an item feels hard, name the lever. If the distractors are close, slow down and contrast them word by word. If the key is far from any direct reference, stop hunting for a matching phrase and reason from the argument instead. The two failures require opposite responses, and treating every hard item the same way is why candidates lose time.
A Practical Order of Operations
- Check the density. How many items hang off this text?
- Match the budget. One or two items: answer in place. Six or more: run the two-phase skim from Section 3.1.
- Classify each item. Gist, detail, inference, cohesion or organization — the label tells you whether to look at the whole text or one sentence.
- Answer in sequence. Items on a passage generally track the text in order, so use the previous item's location as the starting point for the next.
- Commit and move. Individual tasks are untimed; the section is not.
EF publishes the item density of its reading texts. What does it state?
You reach a one-paragraph notice that carries a single comprehension question. What is the correct time allocation?
EF's design specifications describe a fixed-format reading task built from two short passages and six statements. What does this format primarily assess?
An item's four options all paraphrase the passage closely and differ only in a single qualifier. Which difficulty lever from EF's documentation is in play, and what is the right response?