A learner can understand every word in a meeting and still enter half a second late. By then, another colleague has started, the learner yields, and the contribution disappears. Turn timing deserves direct listening practice because English conversation asks a listener to prepare an answer before the current speaker has fully stopped.
This article is for a B1 office worker in Taipei who can answer clearly after a pause yet struggles to join a quick video call.
Conversation leaves very small gaps
Everyday talk contains less empty time than a language exercise. Stivers and colleagues measured question-and-answer sequences in ten languages. Their cross-linguistic PNAS study found a shared tendency to limit both silence and overlap. Average timing varied across languages, though the language means stayed within 250 milliseconds of the overall mean. The result describes informal conversation. It cannot supply a single correct gap for a Taiwanese workplace meeting, where hierarchy, connection quality and the number of participants also shape timing.
The small gap creates a processing problem. Speech planning takes time, so a listener who waits for acoustic silence before deciding what to say will often arrive late. A fluent turn begins with prediction: the listener estimates the likely action, the likely final words and the likely completion point while the speaker is still talking. Fast entry alone is a poor target. Interrupting at every possible completion point would make the learner harder to speak with. The useful skill is recognising a place where the sentence, action and intonation all make speaker change possible.
Words often carry the strongest clue
De Ruiter, Mitterer and Enfield tested how listeners projected endings of turns taken from natural Dutch conversations. In their Language experiment, removing pitch movement while preserving words and syntax barely changed projection accuracy. Scrambling the words while preserving the original intonation damaged it. Their participants relied heavily on lexicosyntactic information. That finding matters for lesson design. Worksheets often make the pitch line look more decisive than the unfinished clause underneath it. A falling pitch is easy to draw on a worksheet, yet it cannot tell the whole story. Compare these two starts:
If we move the launch to Thursday...
If we move the launch to Thursday, we can keep the same reviewer.
The first stretch may sound complete in isolation. The opening if projects another clause, so a listener should keep the floor open. Verbs, conjunctions and the action being performed help predict what remains. A question such as Could you send the revised file by four? also projects an answer before the final syllable arrives.
There is a boundary here. The 2006 experiment used native listeners and Dutch materials. It shows which cues supported projection in that task, rather than proving that English learners will weight the same cues in the same order.
Learners can use prosody too
Emilia Nottbeck's 2026 book reports two reaction-time experiments comparing English native speakers with learners at different developmental stages. The publisher's research summary says prosody helped L2 learners predict turn endings, with effects that differed by question type and proficiency. Grammar and pitch should therefore appear together in practice. I would use real two-person audio, not isolated sentences read from a list. Natural turns include hesitation, continuation and small repairs. Those apparent imperfections carry timing information.
Eighteen seconds can be enough.
Choose an 18-second exchange with two clear changes of speaker. On the first listen, the learner follows the meaning. On the second, pause 300 milliseconds before each change and ask the learner to tap when the first speaker could reasonably finish. Reveal the rest of the audio immediately. A tap too early exposes the cue that was missed: perhaps an unfinished because clause, a list with one item still pending, or a level pitch that continued into another phrase. This resembles the work in teaching pauses at idea boundaries, though the listening target is different. Phrase boundaries divide one speaker's message. Turn boundaries offer the floor to someone else.
Prepare content before the silence
A second pass should add response planning. Stop the clip two or three words before the expected ending and give the learner two tasks: predict the final idea, then begin an appropriate response within 700 milliseconds of the actual end. The answer can start with a modest launch such as Right, I can do that, About Thursday, or One concern is... The opening buys enough time to finish planning without filling the gap with a long uh. Measure only one thing at first: entry timing. Grammar feedback can wait until the learner has completed the response. Combining timing, accuracy, pronunciation and politeness in the same score turns a useful diagnostic into fog.
Video adds gaze and hand movement, but audio should come first for an online meeting learner. A frozen camera and screen sharing often remove visual cues. Once the audio task is stable, replay the same exchange with video and ask whether the speaker's gaze changes the prediction.
Keep overlap in the data
Clean textbook dialogue teaches an accidental rule: speaker B begins only after speaker A produces perfect silence. Natural conversation contains brief overlap. Some overlap shows successful anticipation. Some steals the turn. Use a four-row log after six attempts: early interruption, usable entry, late entry, and no entry. Record the cue beside each attempt. After a few clips, a pattern may appear. The learner might read grammar well and miss pitch, or hear final pitch and ignore a projected clause. That pattern gives the next lesson a target.
For tomorrow's practice, take one 18-second exchange from a recorded interview. Mark two possible completion points, listen again without the transcript, and start a reply before the waveform shows 700 milliseconds of silence.