DET Speaking Question Types

Speaking used to be the most mechanical part of the DET. Since July 2025 it is the most conversational: an on-screen character now asks you six to eight questions and adapts to your answers, and the two read-and-repeat tasks that used to pad the section are gone.

What the Speaking Subscore Measures

Speaking takes up to about 11 minutes of the sitting. Three of the four tasks are single recordings with one attempt each; the fourth is a live-feeling conversation that changes direction based on what you say.

Question type	Per test	Time	Adaptive	Scored as
Speak About the Photo	1	90 seconds	No	Speaking criteria: content, discourse coherence, fluency, grammar, lexis, pronunciation
Read, Then Speak	1	90 seconds	No	Speaking criteria
Interactive Speaking	6-8 questions	35 seconds per answer	Yes	Speaking criteria
Speaking Sample	1	3 minutes	No	Speaking criteria, and sent to institutions as video

There is no retake on any speaking task. Once you click Record Now, the answer you give is the answer that is graded. Practising the first ten seconds of an answer is therefore worth more than practising the whole of it.

The Speaking Tasks

Speak About the Photo

Describe a photograph out loud for up to ninety seconds. One recording, no second take.

Read, Then Speak

Read a prompt with bullet points, prepare for twenty seconds, then talk for a minute and a half.

Interactive Speaking

A spoken conversation with an on-screen character across two topics. Each question is asked once.

Speaking Sample

The longest speaking turn on the test, recorded on video and sent to the universities you apply to.

Read Aloud

Read a written sentence out loud and let the transcriber check it. Retired, still the cheapest pronunciation drill. Retired from the live test in July 2025.

What Changed in July 2025

Duolingo replaced two monologue tasks with one adaptive conversation. Read Aloud and Listen, Then Speak were removed from the live test, and Interactive Speaking was added in their place. If your preparation book still drills Read Aloud, it predates the change.

	Before July 2025	Now
Speaking tasks	Read Aloud, Listen Then Speak, Speak About the Photo, Read Then Speak, Speaking Sample	Speak About the Photo, Read Then Speak, Interactive Speaking, Speaking Sample
Adaptive speaking	Read Aloud only	Interactive Speaking, fully adaptive
What is rewarded	Clear pronunciation of written sentences	Answering an unpredictable question in 35 seconds, twice on two different topics
Minimum speaking times	Enforced on several tasks	Removed

Read Aloud is still the cheapest drill in existence for chunking and pronunciation, which is why it survives as practice on this site — see the Read Aloud page — but you will not meet it on a live test.

The Six Criteria

Speaking answers are graded on the four writing criteria plus two that only apply to speech.

Criterion	What is measured	What raises it
Content	Task achievement, style, development, effect on the listener.	Answer the actual question, then give one reason and one example. Do not describe what you are about to say.
Discourse coherence	Clarity, cohesion, progression of ideas.	Signpost out loud: "There are two reasons. First… The other thing is…"
Fluency	Speed, chunking, breakdowns and repairs.	Pause at clause boundaries, not mid-phrase. There is no ideal speed; there is an ideal rhythm.
Grammar	Grammatical range and accuracy.	One conditional, one relative clause, more than one tense. Range counts even when accuracy slips.
Lexis	Diversity, sophistication, word choice, word formation.	Precise verbs. If a word will not come, describe the thing rather than going silent.
Pronunciation	Intelligibility, individual sounds, word stress, sentence stress, intonation.	All standard accents are accepted. What is graded is whether you are easy to understand, and whether stress lands on the right syllable.

Accent is not penalised. Word stress is. baNAna is correct and banaNA is not, and stress errors are what the pronunciation score actually reacts to.

Handling Interactive Speaking

A Duolingo character asks six to eight questions, three or four on each of two unrelated topics. You get 35 seconds per answer, the timer starts the moment the question ends, and you hear each question exactly once. Your answer is transcribed and scored in real time, and it decides how hard the next question is.

Thirty-five seconds is roughly three to five sentences. Aim for a direct answer, a reason, and an example, then stop.
There is no preparation time and no replay. If you miss the question, say something relevant to the topic rather than nothing — silence scores lower than an imperfect answer.
The topic switch happens mid-set. Do not carry the previous topic’s content across.
Because it adapts, the questions getting harder is confirmation that you are doing well.
The questions use the same everyday and academic topics as the rest of the test. No specialist vocabulary is required.

Common Speaking Mistakes

Silence in the first three seconds. The recording starts when you click. Have a first sentence ready that works for any prompt.
Filling time with a preamble. "That is a very interesting question" costs you five of thirty-five seconds and earns nothing under any criterion.
Stopping early on the long tasks. Speak About the Photo gives you ninety seconds and the Speaking Sample gives three minutes. A short answer caps Content no matter how clean it is.
Going silent on a missing word. Describe it instead: "the thing you use to…". Fluency drops far more from a five-second gap than from an imprecise phrase.
Reciting a memorised answer. It breaks the test rules, and the graders are trained on exactly this pattern.

How to Prepare

Record 35-second answers to unfamiliar questions daily. Length control is a trainable skill and it is the one Interactive Speaking tests hardest.
Describe a photo out loud for ninety seconds every day without stopping. Set a timer and do not pause it.
Listen back to your own recordings. Count the silent gaps over two seconds; that number is your fluency score in miniature.
Drill word stress on the long academic words you actually use. Stress errors are the cheapest pronunciation points to win back.
Rehearse the three-minute Speaking Sample at full length. Most test takers have never spoken continuously for three minutes in any language.

Prompts and model answers are on the DET Speaking samples page, and the full band descriptors are on the DET Speaking scoring page.

Kirill O'Nayd, Duolingo English Test expert

Written by Kirill O'Nayd

English test expert and educator at TestSucceed, where he builds the Duolingo English Test practice tasks and mock tests and has spent years taking apart how computer-adaptive, AI-graded exams award their points. See all TestSucceed preparation materials.