The distinctive thing about this test is the integrated task. You read a short passage, then hear a lecture that qualifies or contradicts it, and then speak or write about the relationship between the two. Nothing else on the list of common English exams does this, and it is why a candidate with strong English can still do badly.
What it is measuring at that point is not English. It is whether you can hold a written argument in mind while a spoken one runs over the top of it, take notes on both, and pair them. That is an academic skill that happens to be tested in English, and it can be practised separately.
Eight facts about the format
- scored out of 120, with thirty points per section across reading, listening, speaking and writing.
- 80 to 100 is the usual requirement for university entry, varying widely by institution and course.
- it is taken on a computer throughout, including the writing, so typing speed is a real factor.
- the speaking is recorded, not conversational. You talk to a microphone against a timer, which many candidates find harder than a person.
- integrated tasks combine skills: read a passage, hear a lecture, then speak or write about both. This is the format's distinctive feature.
- note taking is allowed and is essential, because the lectures are long and played once.
- it is heavily North American in accent and in academic context, which is worth knowing if your listening practice has been British.
- the listening is played once, like IELTS, and the lectures run several minutes each.
The shape of it
Four sections, Reading, Listening, Speaking and Writing, each scored from zero to thirty, for a total out of a hundred and twenty. It is delivered on a computer, so the Speaking section is recorded rather than conducted with a person, and in 2023 the test was shortened to under two hours, which changed the number of tasks rather than what they ask.
The recorded speaking is the second thing that surprises candidates. You speak into a microphone, to a timer, in a room with other people doing the same, and there is nobody to react to what you say. Preparation that consisted of conversation practice does not transfer directly, because the difficulty here is producing a structured answer to a clock with no interlocutor.
The pairing is the whole task
In an integrated task, the passage makes several points and the lecture addresses each of them in order. The answer that scores is the one that pairs them explicitly: this is what the reading claimed, this is what the lecturer said about that claim, and this is the relationship between the two.
Candidates lose marks by summarising each source in turn and never joining them. Both summaries can be accurate and the answer still fails the task, because the task was the relationship and not the contents.
The mechanical version of that is worth carrying in: note the passage's points as a numbered list before the lecture starts, then attach what the lecture says to each number. The structure of the answer is then already on the page.
Where it stops: preparing English rather than note-taking
The stall is spending the preparation on vocabulary and grammar when the loss is happening in the notes. A candidate who reads well and listens well can still arrive at the end of a lecture with a page that does not support an answer, because writing the notes consumed the attention the listening needed.
Nothing in a score report distinguishes the two causes. A low integrated speaking score looks like a speaking problem, so the candidate practises speaking, and the next attempt goes the same way. The diagnosis is available in a minute: look at the notes you took and ask whether the answer could have been built from them by somebody who had not been in the room.
Practising on lectures that were not written for a test
MIT OpenCourseWare publishes complete university courses free and without an account: full lecture recordings, in academic English, at real speed, with the readings attached. That combination is exactly the material this test is built out of, and it is available in unlimited quantity.
It replaced waiting until you were at a university before you ever heard a university lecture, which is the position most candidates prepare from. The limitation is that real lectures are longer, less structured and less obliging than the test's, with no signposting written in to help a marker: they are harder than what you will sit, which is the useful direction for practice to be wrong in.
What it looks like when it goes wrong
The recognisable failure is an integrated answer that describes the reading accurately, describes the lecture accurately, and never says how they relate. It reads as complete to the person who wrote it and it has not done the task.
The second is running out of time in the recorded speaking because the first fifteen seconds were spent deciding how to begin. With no person opposite, nothing rescues a slow start, and the timer does not care that the rest of the answer would have been good.
The note-taking underneath all of this, and the rule about what is worth writing and what can be rebuilt afterwards, is catching the point while the speaker keeps going, and the academic vocabulary the lectures run on is the vocabulary that carries essays, papers and lectures.
The open question is whether a recorded and partly machine-marked speaking section suits you at all, because the same performance is scored differently by a person, and that difference is a reason to choose one test over another rather than a thing to complain about: what each test scores, and how to prepare for yours.