How Reciprocal Teaching Works | Predict, Question, Clarify, Summarise

eduKateSG Learning Node Series · 0125

A student can read every word on a page and still fail to notice that meaning has quietly disappeared.

The eyes continue. The sentences are pronounced correctly. The paragraph ends. If asked what happened, the learner gives back fragments, guesses from one familiar word, or discovers only then that the text stopped making sense three lines ago.

Reciprocal teaching was built for this problem. Instead of treating comprehension as something invisible that either happens or does not, it turns four useful reading moves into a conversation: predict, question, clarify and summarise. At first the teacher models the moves. Then teacher and learners share them. Eventually students lead more of the work themselves.

Reciprocal teaching works by making the hidden work of comprehension public long enough for learners to take it over.

The 50-Second Read

Reciprocal teaching is associated with Annemarie Palincsar and Ann Brown’s landmark work on reading comprehension.
The classic four activities are predicting, questioning, clarifying and summarising.
The method is not four worksheets. It is a structured dialogue around a text.
The teacher initially models expert comprehension and gradually transfers leadership to students.
Predicting activates a model of what may come next; questioning identifies important information; clarifying repairs breakdowns; summarising compresses meaning.
The four moves work together as a monitoring system: read, test understanding, detect trouble, repair and update.
The original research reported gains in comprehension and transfer, but later classroom studies show that implementation quality matters greatly.
Students need enough background knowledge and vocabulary to reason about the text; strategy labels cannot replace knowledge.
The dialogue should become less scripted as learners gain control.
The destination is independent comprehension monitoring, not permanent dependence on four named prompts.

Canonical Owner Boundary

This Learning Node owns the reciprocal-teaching method in which teacher and learners take turns leading comprehension dialogue through predicting, questioning, clarifying and summarising. How Dialogic Teaching Works owns classroom dialogue as a wider pedagogy. How Questioning Works in Teaching owns questioning more generally. Existing English and comprehension owners retain their task-specific reading skills. This page owns the orchestration of four comprehension-monitoring moves and the gradual handover of dialogue leadership.

1. The Original Problem: Weak Readers Often Do Not Know When Meaning Has Broken

Strong readers do more than decode. They notice when a sentence conflicts with an earlier idea, when a pronoun has no clear referent, when a technical term blocks understanding, when the main claim changes, or when they have reached the bottom of a page without a usable model of what the text says.

Weaker comprehenders may continue reading without launching a repair.

Palincsar and Brown’s reciprocal-teaching work made those monitoring and repair behaviours explicit. Instead of telling students simply to “understand better,” it gave them recurring cognitive jobs they could practise in dialogue.

2. Predicting Builds a Forward Model

Prediction is not fortune-telling.

The learner uses the title, headings, prior sentences, genre, evidence and background knowledge to anticipate what the text is likely to do next.

The important part arrives after reading continues. Was the prediction supported? Partly supported? Wrong? What evidence changed the model?

This makes prediction a comprehension check. A wildly inaccurate prediction can reveal that the learner misunderstood the current state of the text. A revised prediction shows that the reader is updating rather than merely marching forward.

3. Questioning Identifies What Matters

Students often ask weak questions because they do not yet distinguish central information from decorative detail.

“What colour was the bag?” may be answerable but irrelevant to the passage’s main idea. A stronger question might ask why a character changed position, what caused an outcome, which evidence supports a claim, or how two concepts differ.

Generating questions therefore becomes a test of importance. To ask a useful question, the learner must decide what is worth knowing.

4. Clarifying Is the Repair Engine

Clarifying begins with an admission: something is not clear.

That sounds trivial, but it is a major metacognitive act. Learners must detect the break before they can repair it.

The problem may be a word, sentence, reference, causal link, assumed fact, diagram or entire paragraph. Repair options include rereading, reading ahead, decomposing a sentence, checking a definition, connecting to a diagram, asking another reader, or revising the interpretation of an earlier passage.

“Clarify” is not one strategy. It is the launch command for a family of repair strategies.

5. Summarising Compresses Without Destroying Structure

A summary is not the first sentence copied more briefly.

The learner must decide what is central, what can be omitted, how ideas relate and what wording preserves the structure without reproducing every detail.

That makes summarising a demanding comprehension test. If the student cannot produce a coherent compressed version, the internal model may still be fragmented.

Good summaries also feed the next cycle. The reader begins the next paragraph with a compact representation of what has already been established.

6. The Four Moves Form a Control Loop

Reciprocal teaching is strongest when the four activities stop feeling like separate classroom tricks.

Predict → read → question → detect uncertainty → clarify → summarise → update prediction → continue.

The loop has the same shape as other intelligent systems: form an expectation, receive evidence, compare, repair and update.

That is why reciprocal teaching belongs inside a wider learning library. It is a reading method, but beneath it is a general model of monitored cognition.

7. The Teacher Models the Invisible Work

Novices cannot imitate thinking they cannot see.

At the beginning, the teacher can think aloud:

“The heading says the city expanded after the railway arrived, so I predict the next paragraph will explain how transport changed trade. This sentence about population surprises me, so my prediction may be too narrow. The word ‘agglomeration’ is blocking me; the surrounding sentence suggests it means activities clustering together. Let me summarise: the railway did not just move people—it changed where firms and workers concentrated.”

The content matters, but so does the sequence of mental moves. Students hear what expert monitoring sounds like before they are asked to perform it independently.

8. Reciprocal Means the Leadership Moves

If the teacher always predicts, always asks the questions, always explains difficult words and always provides the summary, the method has lost its defining mechanism.

Leadership shifts.

A student becomes discussion leader for one passage. Another asks the key question. Another challenges the summary. The teacher coaches, corrects, prompts and protects disciplinary accuracy while reducing control where learners can carry it.

The handover is the learning design.

9. Scaffolding Should Fade, Not Fossilise

Many classrooms turn strategies into permanent worksheets with four boxes labelled Predict, Question, Clarify and Summarise.

The boxes can help initially. They externalise the routine.

But if advanced readers cannot approach a page without filling the four boxes, the scaffold has become architecture.

Fade prompts when the behaviours become internal. The goal is not ritual completion. The goal is a reader who notices which move is needed and uses it without waiting for a label.

10. The Original 1984 Research

Palincsar and Brown’s 1984 study, Reciprocal Teaching of Comprehension-Fostering and Comprehension-Monitoring Activities, reported two instructional studies with seventh-grade poor comprehenders. The four activities were summarising, questioning, clarifying and predicting, and teacher and students took turns leading dialogue around the text.

The reported results included improvements in the quality of questions and summaries, gains on comprehension measures, maintenance and transfer to novel tasks. The study became influential partly because it linked strategy instruction with modelling, dialogue and gradual transfer of responsibility rather than teaching isolated tricks.

Historical influence should not be confused with a guarantee that any four-strategy activity will reproduce the original results. Implementation is part of the treatment.

11. Later Evidence Warns Against Treating the Method as Plug-and-Play

A 2018 classroom study of low-achieving adolescents, Effects of reciprocal teaching on reading comprehension of low-achieving adolescents, found no overall main effect after a year of implementation, while effects were moderated by the quality of strategy instruction. The researchers noted that high-quality modelling and guidance in small groups were difficult to implement.

This is an important corrective. Reciprocal teaching is not four vocabulary words printed on a poster. Teachers must model comprehension processes, diagnose student dialogue, prompt without taking over and gradually fade support.

A 2021 systematic review covering 18 studies likewise concluded that reciprocal teaching often produced significant gains while emphasising scaffolding, comprehension monitoring and the learning environment as key features.

12. Knowledge Still Matters

A student cannot strategy their way around missing domain knowledge indefinitely.

If a science text assumes knowledge of cells, membranes and concentration, predicting and questioning will not manufacture those concepts from nothing. If a history passage refers to institutions and events the learner has never encountered, the comprehension burden rises.

Reciprocal teaching helps readers use and monitor knowledge. It does not abolish the need to build knowledge.

13. Vocabulary Can Be the First Weak Link

Clarifying often exposes lexical problems.

A student may think the whole paragraph is confusing when one technical term is carrying the relationship. Another learner may know every word individually but misunderstand a phrase such as “by contrast,” “subject to,” or “net effect.”

Teach students to locate the smallest unit of confusion. Is the problem one word, one relationship, one sentence or the whole model?

Repair becomes faster when the break is localised.

14. Predicting Can Become Empty Guessing

“I predict the next paragraph will be interesting.”

That is not useful prediction.

Require evidence: “Because the previous paragraph introduced two competing explanations, I predict the next section will compare them.”

The prediction should be constrained enough that later evidence can support, refine or reject it.

15. Questioning Can Become Trivia Production

A student who asks ten tiny factual questions may appear active while still missing the text’s central structure.

Teach question hierarchy.

What happened?
Why did it happen?
What evidence supports that explanation?
How does this idea connect with the previous section?
What would change if one condition changed?
What claim is the author trying to establish?

Different texts need different questions, but strong questions point toward load-bearing meaning.

16. Clarifying Can Become Teacher Rescue

A student says, “I don’t understand this sentence.” The teacher immediately explains it.

The misunderstanding disappears, but the learner may not have practised repair.

Try a ladder: Which part is unclear? What do you think it might mean? Does the next sentence help? Which word connects the two clauses? Can another student restate it? What evidence would reject that interpretation?

Provide the answer when needed, but preserve some of the repair work for the learner.

17. Summarising Can Become Deletion Instead of Compression

Weak summarising often means removing random sentences until the passage is shorter.

Strong summarising identifies the relationship that makes the details belong together.

Ask: What claim would make these three facts relevant to the same paragraph? Which detail could disappear without changing the point? What relation—cause, contrast, sequence, example, qualification—holds the information together?

A summary preserves the structure while compressing the surface.

18. Reciprocal Teaching in English Comprehension

With narrative or nonfiction passages, reciprocal teaching can make interpretation collaborative without reducing comprehension to opinion.

Predict how a conflict will develop. Ask which detail is central. Clarify an ambiguous pronoun or figurative phrase. Summarise the change in a character’s position. Then return to the text to check.

For examination reading, the dialogue can also distinguish reading from answering. Understanding the passage is one layer; matching the answer form to the question is another.

19. Reciprocal Teaching in Science

Scientific texts contain dense causal chains and technical vocabulary.

Prediction can target mechanism: if particle speed increases, what should happen next? Questioning can target evidence: which observation supports the model? Clarifying can separate familiar everyday meanings from technical meanings. Summarising can compress a process into inputs, mechanism and outputs.

The discussion should remain accountable to scientific accuracy. Reciprocal does not mean every explanation is equally valid.

20. Reciprocal Teaching in Mathematics

Mathematics is not usually the first domain associated with reciprocal teaching, but the logic can transfer to worked solutions and word problems.

Predict the next transformation. Ask why a step is legal. Clarify notation or a hidden assumption. Summarise the method as a sequence of structural decisions rather than copied operations.

The method is especially useful when students can follow a solution line by line but cannot reconstruct why the route works.

21. Reciprocal Teaching in History and Humanities

Humanities texts often demand source awareness, causation and argument tracking.

Predict the author’s next move. Question the evidence. Clarify whether a sentence states fact, interpretation or evaluation. Summarise not only what happened but the argument about why it mattered.

Students begin to read as participants in a disciplinary conversation rather than as collectors of sentences.

22. Group Size Changes the Dialogue

In a whole class, only a small number of students can lead at once. In a three- or four-person group, every learner can take a role quickly.

Small groups increase participation but also increase the teacher’s monitoring challenge. Misconceptions can circulate without correction. One strong reader can perform every strategy while others agree.

Use brief teacher sampling, written traces, rotating leadership and periodic whole-class synthesis to keep small-group dialogue connected to standards.

23. Three-Student Tutorials Create High-Resolution Reciprocal Reading

With three learners, the tutor can hear the reasoning behind every prediction and summary.

One learner predicts. Another challenges the evidence. The third clarifies a sentence. Roles rotate. The tutor notices whether the problem is vocabulary, inference, background knowledge or answer scope.

The small-group advantage is not merely more turns. It is diagnostic resolution.

24. The Teacher Should Eventually Hear Less Strategy Language

At first, students may need explicit labels: “Now predict.” “Now clarify.”

Later, mature readers may simply say, “That cannot be right because the previous paragraph says…” or “I think the author is moving toward…”

The named strategies have become natural reading operations.

This is success. Strategy instruction should disappear into competent performance.

25. Do Not Grade Every Reciprocal Move

If every prediction and question becomes a mark, students optimise for safe performance.

Prediction requires permission to be wrong for a good reason. Clarifying requires permission to admit confusion. Questioning requires experimentation.

Use low-stakes feedback while the method is being learned. Assess comprehension outcomes and quality of reasoning without making every monitoring act a public score.

26. Cross-Domain Comparison: Debugging Software

A programmer runs code with an expectation, observes output, notices a mismatch, isolates the problem, repairs it and reruns.

Reciprocal reading has the same control shape. Prediction establishes expected meaning. Questions inspect the system. Clarification locates and repairs faults. Summarising records the current stable state.

Reading comprehension is not software, but the debugging comparison helps explain why monitoring matters as much as decoding.

27. Cross-Domain Comparison: Navigation

A navigator does not wait until the destination to discover they took the wrong road fifty kilometres ago.

They maintain position, compare against route, notice deviation and correct.

Reciprocal teaching gives readers checkpoints before the end-of-passage question reveals that comprehension drifted long ago.

28. Cross-Domain Comparison: Team Briefings

High-reliability teams often restate the current situation, anticipate what comes next, surface uncertainty and ask what remains unresolved.

The purpose is shared situational awareness.

Reciprocal dialogue creates shared textual awareness. Students reveal not only the answer but the state of their model of the text.

29. A Practical Reciprocal-Teaching Protocol

Choose a text with real comprehension load: not so easy that monitoring is unnecessary, not so difficult that missing knowledge dominates.
Preteach essential background knowledge: strategies need something to work with.
Model one full cycle: think aloud and show why each move is used.
Let students practise one move at a time: especially clarifying and summarising.
Combine the moves: connect them into a reading loop.
Rotate leadership: every learner should carry the dialogue.
Probe reasoning: ask what evidence supports predictions and summaries.
Coach repair: do not immediately rescue every confusion.
Protect accuracy: intervene when group dialogue stabilises a false model.
Fade prompts: remove role cards and boxes as the behaviours become internal.
Transfer: use different genres, subjects and reading conditions.
Check independence: ask whether the learner uses monitoring without the reciprocal-teaching routine.

30. Failure Mode: Four Roles, No Comprehension

Students mechanically take turns saying a prediction, question, clarification and summary.

Repair: reconnect every move to evidence in the text and to a genuine comprehension need. A strategy label that changes nothing is ceremony.

31. Failure Mode: Students Lack the Knowledge to Discuss the Text

The group keeps “clarifying” because the text assumes concepts nobody knows.

Repair: teach missing domain knowledge or select a more appropriate text. Strategy instruction cannot compensate indefinitely for a knowledge vacuum.

32. Failure Mode: The Strongest Reader Does Everything

One student produces every useful question while others copy the summary.

Repair: require independent thinking before discussion, rotate leadership and occasionally ask another student to challenge or restate the leader’s reasoning.

33. Failure Mode: The Teacher Never Fades

The dialogue remains teacher-led months later.

Repair: deliberately hand over one function at a time. If students can clarify vocabulary independently, stop doing it first. Preserve support only where evidence says it is still needed.

34. Failure Mode: The Scaffold Fades Too Early

Students are told to lead before they have seen strong modelling.

Repair: return to worked dialogue. Show what a high-quality question or summary sounds like and why. Fading should follow capability, not a calendar.

35. Failure Mode: Strategy Performance Replaces Reading

Students spend so much time filling strategy templates that the text fragments into administrative tasks.

Repair: use the smallest visible scaffold that still helps. The strategy exists to support reading, not become the main event.

36. The Missing-Node Scan

If students read fluently but cannot explain what they read, if they rarely notice when meaning breaks, if difficult words stop the entire passage, if summaries reproduce details without structure, or if comprehension problems are discovered only after the questions at the end, the missing node may be comprehension monitoring.

Reciprocal teaching is one disciplined way to build that monitoring. It gives the invisible reader four temporary handles: anticipate, interrogate, repair and compress.

37. The Return Path

Return to the student whose eyes reached the end of the page while meaning vanished halfway through.

The mature reader should not need a teacher standing beside every paragraph.

They need an internal conversation.

What do I expect? What matters here? What stopped making sense? What does this section amount to? Does the next evidence fit the model I built?

Reciprocal teaching makes that conversation social first so it can become private later.

Reciprocal teaching works when the dialogue eventually disappears into the reader: prediction becomes anticipation, questioning becomes discrimination, clarification becomes repair, and summarising becomes a stable model of meaning.

Research and Further Reading

eduKateSG Learning Node Series · 0125 · Previous: 0124 — How Dialogic Teaching Works.

How Reciprocal Teaching Works | Turn Reading Into Predicting, Questioning, Clarifying and Summarising