Current format, sourced
The 2026 TOEFL® Listening section: four task types, two modules
The current TOEFL® test Listening section is two-stage adaptive: everyone takes a router stage of about 18 minutes, then is routed to one of two second-stage modules — a lower module of 7 minutes or an upper module of 11 minutes. Four task types cover daily-life and academic listening.
What the specification lists
ETS states — from the blueprint and Listening tables of the 2026 test specifications, accessed 17 July 2026.
| Listed field | Listed value |
|---|---|
| Task types | Listen and Choose a Response; Listen to a Conversation; Listen to an Announcement; Listen to an Academic Talk |
| Items | 47 |
| Maximum raw points | 35 |
| Estimated time | ~18 min router + a second-stage module of 7 min (lower) or 11 min (upper) |
- These are the published blueprint’s listed values, which ETS marks as subject to minor revision until the official launch.
- Listening item counts include non-scored items used to gather item-performance data.
- The two module lengths are discrete listed values — 7 minutes (lower) or 11 minutes (upper) — not a 7-to-11-minute range.
- Listening items are machine-scored selected responses, per the specification’s item-type notes. Machine-scored is not exclusive to Reading and Listening: the Writing section’s ten Build a Sentence items are machine-scored too.
What the lower/upper module split means for you
Reach120 interprets — our reading of the ETS materials, not ETS’s own advice.
You cannot fix your question count in advance
Router performance determines your second-stage module, so “how many Listening questions will I see?” has no single answer for an individual sitting. Prepare for the task types, not for a fixed count.
The modules are discrete, not a dial
The blueprint lists exactly two second-stage lengths — 7 minutes or 11 minutes — with no published in-between. A prep source describing a sliding “7–11 minute” range is imprecise about the design.
Module assignment is not a score report
ETS publishes no rule converting router performance or module assignment into a 1–6 band. Do not treat a practice router result as a predicted band.
Adaptive changes pacing, not preparation
The design changes how many items you meet, not which skills are measured. Covering all four task types prepares you for either module.
How to practice each task type
The four task types are the specification’s; the practice advice is Reach120’s.
- Listen and Choose a Response
- Short conversational turn; you pick the natural reply. Practice reacting to function — requesting, offering, declining, confirming. The correct response answers the function, not the surface words.
- Listen to a Conversation
- Two speakers in a daily-life situation. Practice tracking each speaker’s goal and what changes by the end: what was asked, what was agreed, what happens next.
- Listen to an Announcement
- One speaker, practical information. Practice extracting the actionable core — who is affected, what changes, when — and expect questions about purpose and consequence rather than decorative detail.
- Listen to an Academic Talk
- Short academic monologue. Practice the claim–example link: what point is made, and what does the example prove? Habits built on the retired long-lecture format overshoot these shorter stimuli.
One item, one decision
Reach120-original worked example, illustrating the Listen to an Announcement pattern. It is not an ETS item and reproduces no official material. In a real sitting this stimulus is audio; it is shown here as a transcript.
“Attention, library visitors: starting this Saturday, the ground-floor group-study rooms will be closed for carpet replacement for about two weeks. Group bookings made for those rooms will be moved automatically to the third-floor rooms — you don’t need to rebook. Individual quiet-study seating on the ground floor stays open as usual.”
What should students with existing group bookings do?
- (A)Rebook a room on the third floor
- (B)Nothing — their bookings move automatically
- (C)Use the ground-floor quiet-study seating instead
- (D)Wait two weeks and then rebook
The decision
Every wrong option echoes real words from the audio — “third floor,” “quiet-study seating,” “two weeks.” That is the trap pattern: distractors recycle surface vocabulary. The question asks about action required, and the announcement’s function sentence is “moved automatically — you don’t need to rebook.” Option (B) paraphrases that function without repeating its words; (A) and (D) contradict “automatically,” and (C) confuses individual seating with group bookings.
The transferable decision: answer from the speaker’s function — what changes, for whom, what action is required — never from word-matching. An option that repeats the audio’s exact words is more often bait than answer.
What a Listening result means
Listening items are selected responses, machine-scored against answer keys — so a practice result here is a deterministic raw count, and AI may explain a result but never rewrite it.
That raw count is not a band. Official score reports use the 1–6 half-band scale for tests taken on or after January 21, 2026, and that scale is the primary official one; a comparable 0–120 estimate is also shown during ETS’s transition through January 2028. ETS publishes no simple raw-to-band conversion, so any practice figure is a practice estimate for the sections actually practiced — never an official score and never a whole-test readiness number.
Common questions
- Is the 2026 TOEFL Listening section adaptive?
- Yes. The 2026 Reading and Listening sections use multistage adaptive designs. Speaking and Writing are linear.
- How long is the Listening section?
- The specification lists an ~18-minute router plus a second-stage module of 7 minutes (lower) or 11 minutes (upper) — two discrete listed lengths, not a range, and subject to ETS’s stated pre-launch revisions.
- How many Listening questions are there?
- The specification lists 47 Listening items worth 35 raw points, and notes the count includes non-scored items used to gather item-performance data. Your own sitting depends on which second-stage module the router assigns.
- Is Listening scored by AI?
- In our reading of the specification’s item-type notes, no — Listening items are machine-scored selected responses against answer keys, and AI scoring applies to constructed responses in Writing and Speaking. Machine-scored is not exclusive to Reading and Listening: the Writing section’s ten Build a Sentence items are machine-scored too.
- Do older TOEFL listening materials still work?
- Check any material against the four listed task types — Listen and Choose a Response, Listen to a Conversation, Listen to an Announcement, Listen to an Academic Talk. Material organized entirely around long lectures and long conversations was written for an earlier format; some listening skill transfers, but the task shapes do not.
What this page does not tell you
This restates the ETS blueprint’s listed Listening values, which ETS itself marks as subject to minor revision until official launch. Item counts include non-scored items. Module assignment rules beyond “router then lower or upper module” are not published, and nothing here converts raw counts to bands or predicts an individual’s score. The worked item is a Reach120-original illustration, not an official item, and its transcript stands in for audio.
Sourced from the ETS 2026 test specifications, accessed 17 July 2026. Next review 18 August 2026, or immediately on any ETS specification change.