All articles

Fade speaking prompts after a successful turn

30 October 2026

Choc Education


A sentence frame can rescue the first turn and quietly occupy the next twelve. Once a learner has used “One reason I prefer ___ is ___” successfully, leaving the whole frame on screen may turn later answers into slot-filling. Speaking support should shrink after evidence of a successful turn, with the teacher ready to restore one piece when performance breaks down. Listen first. The timing comes from the learner’s response, which gives a teacher fresher evidence than the week number or the order of slides in a plan.

Support has three moving parts

Educational research uses “scaffolding” so loosely that almost any helpful worksheet can acquire the label. Van de Pol, Volman and Beishuizen’s review of a decade of teacher-student research offers a stricter account. It identifies contingency, fading and transfer of responsibility as linked features. The teacher adjusts help to current performance, reduces it over time and hands more of the task to the learner. Those features make a useful test for language materials. A permanent box listing six sentence starters provides support. It becomes scaffolding only when someone checks whether the learner still needs all six and changes the help accordingly.

The review also exposes a limit. The underlying studies used varied definitions and research designs, and effectiveness evidence was mixed. “Fade the prompt” is a design principle that still needs judgement. It is no universal recipe for faster language acquisition.

Watch the turn before touching the slide

In 2025, a System study examined teacher fading in EFL classroom interaction. The data covered two teachers, 261 K-11 students in a northern Chinese senior high school and 180 minutes of video from four lessons. Through conversation analysis, the researchers traced moments when teachers reduced their contribution after learners produced correct short answers, partially correct answers or more complex responses that showed potential for independent work. Fading took several forms. A teacher could wait instead of completing the answer, narrow a question instead of supplying language, or let peers extend a point. The important unit was the unfolding exchange. Support changed after visible evidence in a learner’s turn.

This is close to what a speaking teacher can hear online. Suppose a learner receives this prompt for a product comparison:

I would choose ___ because ___. Compared with ___, it is ___, which matters when ___.

On attempt one, the full frame may let the learner express a comparison. On attempt two, the screen can show only “Compared with...”. If the learner completes the contrast and reason, attempt three can retain the product names and remove the language frame. If the sentence collapses after the first clause, restore the connective rather than restarting the whole explanation. That sequence is an example, not a finding from the System study. The study observed four lessons in one school and did not randomly assign learners to fading schedules. It can reveal how teachers time support in interaction; it cannot give a causal estimate of how many prompts to remove.

Fixed fading can leave too early

Research outside language classrooms adds a useful warning. Tawfik and colleagues compared sustained and faded question prompts in ill-structured argumentation tasks. Their measures did not all point in the same direction. Rubric scores showed no significant difference between conditions, while analysis of conceptual links favoured sustained support. The authors concluded that novices may need scaffolds for longer before working independently. This boundary matters for a thirteen-year-old preparing a two-minute school presentation. Removing every sentence starter in the second week merely because the plan says “independent practice” can expose the same unresolved language problem again. Independence is the goal, not the date.

Some support can remain while another part fades. Keep a content diagram and remove the model sentence. Keep four topic words and remove the word order. Let the learner ask for one prompt after 8 seconds. Support can move sideways before it disappears.

Use a four-level prompt ladder

Teachers need a quick system that works during live interaction. One ladder can serve across several speaking tasks:

  1. Show a complete model and substitute one detail.
  2. Show the sentence opening and the required connective.
  3. Show content cues only.
  4. Give the task with no visible language prompt.

After each turn, record the lightest level at which the learner completed the communicative job, using a task-specific definition that may require a position, one reason and room for the partner in a disagreement task, or the clear location of two objects that a partner must place in a picture task. Accuracy remains part of the record, though it should not control every move. A learner who conveys the comparison with one article error may be ready for a lighter prompt. A learner who produces a flawless memorised frame and cannot answer a follow-up has supplied weaker evidence of independent control.

This approach pairs well with making teacher talk interruptible. When learners can ask “Can I have the first word?” or “Can I see the connector again?”, restored support becomes a normal classroom move rather than a public declaration of failure.

Assess the amount of help

Conventional speaking notes often record only the final answer. Add one small field: prompt level. Two learners may produce the same sentence after very different amounts of help. The distinction changes the next lesson. For a class of four, use a grid with learner names down the side and task rounds across the top. Enter 4, 3, 2 or 1 for the level of help used, plus one short code such as C for content, L for lexis or S for sentence structure. A sequence of 4-3-2 across three rounds is evidence for reducing support. A sequence of 2-4-3 says the task or topic changed the demand. It does not describe a general lack of ability.

Do not turn the numbers into a grade. Prompt level depends on topic familiarity, partner, time pressure and the exact wording of the task. It is local diagnostic information.

End with an unsupported transfer

The cleanest check comes after a delay and a small change. Give the original frame during a discussion of remote work, fade it across repeated turns, then ask about four-day workweeks 17 minutes later with no frame visible. The learner now has to transfer the language to a related claim. If that final turn fails, the earlier support still did useful work. It showed the learner what a complete answer could contain. The next step is specific: restore one connector, rehearse another comparison and try the transfer again. The prompt ladder moves in both directions.

Tomorrow, remove one line only after the learner has used it successfully. Write down what remains when the line is gone.