TOEFL 2026 guide
The Listening section, task by task
Four task types, about 29 minutes, and one constraint that changes everything: you hear it a fixed number of times and then it is gone. Here is what each task rewards, and the specific listening habits that cost marks.
The shape of the section
Listening runs about 29 minutes and moves from very short items to longer ones. It opens with single lines you respond to, then gives you campus conversations, announcements, and finally academic talks. The audio gets longer as you go; the questions get more interpretive.
| Task | What you hear | What you do |
|---|---|---|
| 1 · Choose the Response | One line — usually a question | Pick the best reply from four |
| 2 · Conversation | Two speakers, campus setting | Two questions per conversation |
| 3 · Announcement | One speaker, practical information | Two questions per announcement |
| 4 · Academic Talk | A short lecture extract | Two questions per talk |
The constraint that matters most
You cannot go back. In reading you can re-read a sentence twenty times; in listening the sentence carrying the answer lasts about two seconds and then it is gone forever. Almost every listening mistake is really a failure to catch one clause — not a failure to understand English.
Task 1 is not a comprehension test
Choose the Response looks like the easy part and is where careful students quietly bleed marks. The reason is that the wrong options are engineered around what a struggling listener actually does: catch one word and pick the option containing it.
The single most useful habit is negative: learn what CANNOT be the answer, and eliminate before you interpret. Two rules do most of the work.
- A question beginning What, Which, Who, Where, When, Why or How cannot be answered Yes or No. If you catch the question word and nothing else, you can still delete every Yes/No option.
- An “X or Y” question cannot be answered Yes or No either — there is nothing for the Yes to attach to. Hearing the word “or” is enough, even if you missed both halves.
Beyond that, treat familiarity as a warning rather than a comfort. An option that repeats a word you definitely heard is the commonest trap in the format, because recognising a word feels like understanding a question.
Tasks 2–4: the questions are predictable
Conversations, announcements and lectures differ in content and barely at all in what they ask. Across all three you get some combination of: what is this mainly about, one specific detail, why did the speaker mention X, and what can be concluded. Knowing that in advance changes how you listen.
- Main idea — nearly always available in the first third. Speakers state their business early and qualify it later.
- Detail — attaches to numbers, dates, conditions and names. When you hear one, expect a question about it.
- Intention (“Why does the speaker mention…”) — answered by the sentence AFTER the mention, where the speaker spends the fact. Listen forward, not back.
- Inference — lives in the concession: “still argued”, “not straightforward”, “and it is not a small one”. The last clause of the last sentence is disproportionately often the answer.
Announcements telegraph their own test items
Phrases like “please note”, “this is the part people get wrong” and “the timetable itself is unchanged” exist because the speaker expects a misunderstanding. A sentence that says what is NOT changing is there because people assume it did — and that is exactly what gets asked.
Are you mishearing, or misreading the options?
These are two completely different problems with the same symptom, and most students never separate them. It is worth ten minutes to find out which one you have, because the fixes have nothing in common.
Take any recording with a transcript. Answer the questions first. Then read the transcript and answer them again. If your score jumps, the audio is your problem. If it barely moves, you are hearing fine and losing marks to the answer options — which is a reading problem wearing a listening costume.
- If the audio is the problem: transcribe. Take one minute of speech and write down every word, replaying as often as you need. It is slow, it is unpleasant, and nothing else works nearly as well.
- If the options are the problem: stop practising listening. Study the wrong answers on items you got right — the traps repeat, and they are the same named traps the Reading section uses.
- If both are shaky: start with the words. Comprehension built on half-heard sentences has nothing underneath it.
Note-taking, honestly
You are allowed to take notes, and most students take far too many. Writing competes with listening for the same attention, and a page of notes taken while missing the argument is a net loss.
- Write numbers, dates, times, prices and names. These are hard to hold and easy to test.
- Write the turn: “but”, “however”, “the problem is”. Whatever follows one of those is usually the point.
- Do not transcribe. If you are writing sentences, you have stopped listening.
- For lectures, one line per paragraph of speech is plenty — you are recording structure, not content.
Pacing and the last five minutes
Listening has a peculiar timing profile: the audio takes as long as it takes, and the only part you control is how long you spend choosing. That means the recoverable time is all in the short items, and hesitation there is expensive.
- Task 1 items should take seconds, not a minute. If you did not catch it, eliminate what you can and move.
- Never let an unanswered question follow you. There is no going back, so a blank is a guaranteed zero where a guess is a real chance.
- Read the question stems for a talk BEFORE the audio if the interface lets you. Knowing you need a reason changes what you hold on to.
Notes on this section
Shorter pieces, each about one thing — a single fact about the 2026 format, or a single pattern that costs people marks.
The task nobody has practised →New in 2026, and up to 19 of the 47 items.Hold the question word →Topic sticks; question words do not. What to retain, and why.