TOEFL Speaking Test: 7 Proven ESL Drills
Seven of TOEFL speaking’s eleven scored items are Listen and Repeat sentences, and together they carry 35 of the 55 points behind the band score — more weight than the interview task most lesson plans still treat as the main event. ETS swapped the old four-task speaking section for this two-task format on tests taken from January 21, 2026 onward, and dropped the planning time that used to buy students fifteen or twenty seconds to organize an answer. A student who used to build a mental outline before speaking now has to start the moment the prompt ends. That shift changes what works in class: rehearsed templates score worse than fast, clean repetition and short answers delivered without hesitation.

What the TOEFL Speaking Test Actually Scores
Each Listen and Repeat sentence gets a response window of eight to twelve seconds, and ETS’s AI model checks three things: whether the words match, whether the delivery stays smooth, and whether stress and rhythm sound natural rather than word-by-word. A student can swap one minor word and still score well if the sentence comes out fluent and quickly. A student who nails every word but drags out the sentence with pauses scores worse. That trade-off surprises most teachers the first time they see it, because it rewards speed and rhythm over textbook accuracy.
Take an Interview works differently. Four questions arrive on a single theme — a campus routine, a personal preference, a hypothetical plan — and each gets a 45-second response with no interviewer feedback in between. Grammar and coherence still count here, but they’re graded on whether the idea holds together across 45 seconds of continuous speech, not on whether every clause is textbook-perfect. A response with one dropped article and a clear throughline beats a grammatically flawless answer that trails off halfway through.
This is a genuinely different animal from the IELTS speaking test, which still puts a live examiner across the table and gives students a full minute to prepare notes before the two-minute cue card task. TOEFL’s version is faster, colder, and entirely machine-graded, which means the coaching has to be too.
Drill 1: The Progressive Shadowing Ladder
Start every Listen and Repeat unit with eight-word sentences and build toward twenty-word ones over three or four sessions, matching the way the real test sentences grow in length and clause complexity. Play a recording once, give students the actual response window — set a timer for ten seconds, not sixty — and have them repeat without looking at a transcript. The point isn’t memorization; it’s training the ear to catch a full sentence on a single pass, which is the actual skill ETS is testing.
Most classes skip straight to full-length practice sentences and wonder why students freeze on the real test. Build the ladder instead. A student who can shadow a clean eight-word sentence cold usually needs only two or three weeks of step-ups to handle the twenty-word versions without panicking.

Drill 2: Record, Score, Repeat
Have students record every Listen and Repeat attempt on a phone and grade their own playback against four criteria before you ever touch it: did the words match, did any filler sound (um, uh) slip in, did the sentence land inside the response window, and did it come out smooth or choppy. Self-grading forces students to hear their own hesitation the way the AI scorer does, instead of relying on a teacher’s ear for feedback that arrives a day later.
This drill also surfaces a pattern fast: students who sound confident often score themselves generously, while quieter students undersell clean repetitions. Compare three or four self-scores against your own assessment in the first week to calibrate the class, then let the self-check run on its own after that.
Drill 3: Sentence-Stress Mapping
English is stress-timed, not syllable-timed, which means content words — nouns, verbs, adjectives — carry the stress while function words like “the,” “a,” and “is” get swallowed in natural speech. Hand students a Listen and Repeat-style sentence, have them mark which words they think a native speaker would stress, then play the audio and check their guesses. Students consistently over-stress function words on the first attempt, which is exactly the habit that makes repetitions sound robotic to the AI scorer.
Run this drill with sentences pulled from the same topic bank as your shadowing ladder so the two drills reinforce each other instead of competing for class time.

Drill 4: The Zero-Prep 45-Second Response
Since Take an Interview gives students no planning time, the single worst habit a class can carry into test day is pausing to organize before speaking. Fix it by running true zero-prep rounds: read a question aloud, start a 45-second timer the instant you finish the sentence, and require students to begin talking within two seconds. No “let me think,” no silent planning — if they’re not speaking, the clock is burning real response time.
Students resist this drill at first because it feels unfair compared to how they studied for other tests. Push through three sessions of it anyway. The resistance usually breaks once a student notices their own 45-second answers are getting longer and more organized without any extra prep time at all — the organization is happening in real time instead of beforehand, which is the actual skill being measured.
Drill 5: Four Openers for Four Question Types
Take an Interview questions fall into four recurring types, and giving students one reliable opening phrase per type removes the two or three seconds most students lose deciding how to start:
- Description questions (“Describe a place you visit often”) — open with “I usually…” or “One place I…”
- Opinion questions (“Do you agree that…”) — open with “I’d say…” or “My view is…”
- Prediction questions (“What do you think will happen if…”) — open with “I expect…” or “It’s likely that…”
- Abstraction questions (“What does X mean to you”) — open with “To me, X means…” or “I see X as…”
These are starters, not scripts — the goal is killing the opening hesitation, not memorizing a canned answer that an AI scorer flags for sounding rehearsed. Drill the four openers in isolation for one class period before attaching them to full 45-second responses.

Drill 6: Partner Interview Swaps
Pair students and give each one a stack of Take an Interview-style question cards pulled from a single theme. Partner A reads a question cold, Partner B answers in 45 seconds with no prep, then they swap roles. Because a classmate is asking instead of a recording, students have to actually listen and respond to a live voice — closer to the conversational pressure of the real interview task than solo practice with headphones ever gets.
Build in a thirty-second peer feedback round after each pair of answers: one thing that sounded natural, one filler word or long pause the partner caught. Students catch each other’s “um” habit faster than they catch their own, and the format keeps the whole exercise moving without eating a full class period.

Run this drill twice a week rather than daily. Students need a day between sessions to notice their own filler words shrinking, and running it every single class turns useful pressure into fatigue, which shows up as flatter, quieter answers instead of sharper ones.
Keep a simple tally for each student across sessions — filler count, average response length in seconds, one note on topic drift — rather than scoring every round. A running tally shows improvement over three weeks in a way a single 45-second clip never will, and it gives you something concrete to show a student who insists they haven’t gotten any better.
Drill 7: The Speech-to-Text Mirror Check
Have students speak a practice response directly into their phone’s voice typing feature and compare the transcribed text against what they meant to say. When the transcription comes out garbled or substitutes the wrong word, that’s usually a pronunciation or clarity issue the student can’t hear in their own voice but a scoring algorithm would catch the same way. It’s a blunt tool, but a free one, and it gives students an outside check between class sessions.

Fitting Seven Drills Into a 50-Minute Class
Nobody runs all seven drills in one session, and trying to is how a speaking unit collapses into a vocabulary review by week three. A workable rotation looks like: ten minutes of shadowing ladder as a warm-up, fifteen minutes split between stress mapping and record-score-repeat, fifteen minutes of zero-prep interview rounds or partner swaps, and five minutes of speech-to-text self-checks as homework setup. Rotate which two or three drills anchor each class so students see every drill at least twice a week without any single one going stale.
The broader context matters here too. TOEFL’s 2026 format overhaul touched all four sections, not just speaking, and a teacher coaching the Writing or Reading sections with pre-2026 habits is working from a test that no longer exists. Students weighing TOEFL against the Duolingo English Test should know the speaking demands are nothing alike either — DET’s speaking prompts still allow a short prep window, which makes TOEFL’s zero-prep format the harder adjustment for most learners.

Students who treat the TOEFL speaking test like a shorter version of the old format tend to plateau around the middle bands, not because their English is weak but because they’re still pausing to plan on a test that no longer gives them time to. The seven drills above exist to break that one habit. Run the shadowing ladder and zero-prep rounds consistently for three weeks before judging whether a student has actually improved — speaking gains on a timed, AI-scored test show up slower than on paper, and a single mock session won’t tell you much either way.
سرچینې
- TOEFL Speaking Section 2026: Format, Tasks, Timing & Tips — Test Ninjas — breakdown of the two-task format, scoring weights, and AI scoring criteria.
- Ten Awesome Tips for the Speaking Section of the TOEFL — TST Prep — source for shadowing, self-assessment checklist, and rapid-fire interview drill concepts.



