Predict – Observe – Explain
Process & connectPredict – Observe – Explain is a science and math structure in which students first write down what will happen and why, and only then watch. The board shows the setup of an experiment, a video or an image, with one question: "What happens if …?" Everyone predicts alone, partners compare, the class's predictions go up on the board, and then everyone watches in silence. Next, each student holds up ✓ or ✗, partners work out what in their reasoning did not hold, and a drawn pair explains it to the class. A demonstration stops being a show and becomes a test of what the class was thinking.
One demonstration · ± 17 min
Put it on the boardOne screen: the question, a link to a video, picture or slides if you have one, and one sentence about what is set up. Or let the AI suggest five demonstrations on your topic. Only you see it; the class sees it only when you press Start.
This exact screen goes on your board. The timer runs per step, "Watch. In silence." waits for you, and you stay in control: pause, add time or move on to the next step. Type the question yourself, or with a free account the AI suggests five demonstrations on your topic.
What the class sees on the board
7 steps · one demonstration, ± 11 min
For you: Only check that everyone writes something. Don't signal whether it's right, not even with your face.
Group size
alone, then in pairs · the whole class
Duration
± 17 min
incl. briefing and look-back
Level
grades 4–12, built for middle school
Materials
mini whiteboard or paper, ✓/✗ voting cards, A/B letter cards; a video or a live demo
Format
alone → pair → class
Purpose
testing a prediction and explaining it
Check one thing before the first run: can the class reason its way to the outcome without already knowing it? A good demonstration has an outcome you can predict with some prior knowledge, and that still surprises part of the class. If everyone knows the answer, there is nothing to explain. If nobody has anything to go on, predicting turns into guessing. So pick an experiment on material the class half knows.
The flow at a glance
± 17 min totalThe minutes above are the defaults, not a script: whoever prepares the structure sets the time per step. Step 2 has one clock for two turns: halfway, the screen passes the turn to B by itself. Step 4 has no clock: an experiment takes as long as it takes, and you move on when it is done. Step 7 grows with the number of explanations you want to hear out loud, three by default.
Step-by-step instructions
7 steps
Predict
1.5 minalone, in silenceThe board shows one question, "What happens if …?", and what is ready: the set-up of a demonstration, or the first frame of a video. Everyone writes on paper or a mini whiteboard what will happen, and why. Anyone unsure still picks one answer.
Compare with your partner
1.5 minin pairsA says what A thinks will happen and why; B listens. Halfway, the screen gives the turn to B: do you think the same? What is on paper stays there.
Predictions on the board
2 minthe whole classThe class names the predictions there are, and you type each different one into its own tile on the board, in the class's words. You confirm nothing: who is right, the class will see in a moment.
Watch. In silence.
1 minthe demonstration, in silenceYou show the demonstration: the video, the picture or the slides on the board, or you do it live. Everyone watches, nobody says a word until it is over. This step has no timer: it lasts as long as the demonstration, and you move on.
Did it come true?
1 minthe whole classOn three, everyone holds up the ✓ or ✗ card for their own prediction, all at once. Then you tap ✓ or ✗ on each tile of the board. A ✗ is not a mistake: it is where the explanation starts.
Explain together
2 minin pairsWhy did it happen this way? The pair works out an explanation, including for the prediction that did not come true: what did we think at first, and what do we think now? Both must be able to tell it.
Explain to the class
3 minthe whole classThe screen draws a pair, and they give their explanation; the class adds to it or offers another. You draw again, three times by default, and a pair that has spoken does not come up twice. Only after the last explanation do you add what is still missing.
No video or picture? Leave the link field empty: you demonstrate it live, and the sentence about what is set up goes on the board.
Which learning strategy does the class train?
Activate prior knowledge
Students first pull up what they already know, before anything new is added. New material attaches to what is already there, so bringing it up first makes the new stick. Source: Wijze lessen (Surma et al., 2019), building block ‘activate prior knowledge’.
On the boardWhat do you already know? · First you pull up what you already know. New things stick better to something familiar.
Also trainable with this structure
- Answer first, then check · Write first, look second: ✓ or ✗ next to your own prediction.(language: vocabulary · text subjects)
- Apply it to something new · Before the formula: the concept has to predict something in a new situation.(maths & science)
- Ask why it is so · Explaining is the third step: why did what you saw happen?(natural sciences · text subjects)
Why does this work?
A demonstration is one of the strongest moments in a science class, and one of the easiest to waste. The class watches, someone says "whoa", and a week later nobody remembers what it showed. Predict – Observe – Explain puts three things around the demo that turn watching into learning.
In 2004, Catherine Crouch, Eric Mazur and colleagues looked at what physics students take away from lecture demonstrations. Students who only watched understood the principle no better than students who never saw the demonstration. Students who predicted the outcome first understood it clearly better. Garvin Brod and colleagues (2018) showed why that can work: surprise appeared only when someone had made a prediction beforehand, and the stronger the surprise, the more was learned. A wrong prediction you wrote down yourself is remembered. An outcome you only watched is entertainment. That is why everyone writes something in step 1, and not just what will happen but why.
Every class has someone who calls out what is happening at the first movement. At that moment the rest hear the answer instead of seeing it, and nobody checks it against their own prediction anymore. That is why "Watch. In silence." is a step of its own, not a remark in passing. Everyone watches with their own prediction still in mind, and answers "did it come true?" only afterwards, with a card in the air.
After the demonstration it is tempting to explain it yourself. Then the class hears the right answer, but the old reasoning stays where it was. In step 6 the pair works out why it happened this way, and above all what was wrong in the prediction that did not come true. That question brings out the idea that pointed the wrong way. The pair that explains in step 7 is drawn by the screen. Because nobody knows whose turn it is, every pair has an explanation ready, not just the hands that always go up.
The demonstration: a link or a live demo
The content is yours: one prediction question and one demonstration. The question works best when it starts with "What happens if …?" and names exactly what changes: "What will the scale read once the sugar has dissolved in the water?" beats "What happens to the sugar?"
A link. In the setup, paste a link to a video, an image or slides. While the class predicts, the board shows the setup but not the outcome: the video only starts when you press play in step 4. Pick a video that shows the setup calmly first, and check in advance that its title does not give the outcome away.
A live demo. If you run the experiment yourself, leave the link field empty and describe in one sentence what is set up: "Two cups of water, one with a lot of salt. An ice cube goes into each." That sentence stays on the board while the class predicts. Set the demo up where everyone can see it, and try it once before class.
Ready-made. For physical science, chemistry, life science and math there are preparations with a question and a demonstration ready to go. You will find them below, with the examples. Open one, see how it runs, then swap in your own question and link.
When it fits, what to watch for
When it fits
- When you expect a stubborn misconception. Heavy things fall faster, plants get their food from the soil, dissolved sugar no longer weighs anything. A student who writes the misconception down and then sees it fail lets go of it more easily than one who only hears the right answer.
- At the start of a new unit. The predictions on the board show at a glance what the class already thinks, before you have explained anything. That is where your explanation starts, and the class now wants to know how it really works.
- With a demo you would show anyway. Every demonstration you already planned gets stronger with a prediction in front of it. It costs a few extra minutes, and the whole class is involved instead of one student helping at the front.
- When you want to hear how they reason. Walk around during step 2 and listen to the reasons behind the predictions. A right answer with a wrong reason is something you hear here, and never see on a graded test.
- Outside science, too. A graph you only reveal after the prediction, a poll result, the outcome of a historical event: wherever the class can reason about something before it is revealed, the structure works. Only the demonstration changes.
What to watch for
- When the outcome is well known: change the question. An experiment that is in the textbook, or that the class saw last year, will not surprise anyone. Turn the question a quarter turn: what happens if you change one thing?
- Without prior knowledge: give them a foothold first. If nobody can reason about it, the prediction becomes a guess, and explaining a guess gains little. A short question up front helps: what do we already know about why some things float and others sink?
- With a demo that sometimes fails: try it in advance. An experiment that goes differently in front of the class is not a disaster, but it is a different lesson. Run it once beforehand, or use a video of the same experiment.
- With students still learning English: use the language-learner variation. A prediction with a reason is a lot of language. With Language learners, students predict with a drawing or one word, start with "I think: …", and the steps are shorter.
- Short on time: use the short variation. As a lesson opener, Short fits: shorter steps, three predictions on the board and two explanations out loud. The order of the steps stays the same, because the order is what does the work.
One structure, several functions
Examples by context
- What happens if I drop a heavy ball and a light ball of the same size at the same time?
- What will the scale read once the sugar has completely dissolved?
- What happened to a strip of potato left in salt water for half an hour, and to one from plain water?
- What happens to the area of a square when its sides become twice as long?
Variations
By age
Common mistakes
Frequently asked questions
What exactly is the Predict – Observe – Explain structure?
A structure in which students predict the outcome of a demonstration before they see it, and then explain what happened. Seven steps: Predict (everyone alone, in silence, with a reason); Compare with your partner (first A talks, then B); Predictions on the board (you type the different predictions as tiles); Watch. In silence. (the demonstration, everyone watches without talking); Did it come true? (everyone holds up ✓ or ✗, you mark each tile ✓ or ✗); Explain together (the pair works out why it happened this way and what was wrong in the prediction); Explain to the class (a drawn pair explains). In science education research the approach is known as Predict-Observe-Explain, or POE.
How long does Predict – Observe – Explain take?
With the default settings the whole structure takes ± 17 minutes, briefing and look-back included.
Whoever prepares the structure sets the time for each step. Predicting and comparing are short; in step 2 the screen passes the turn from A to B halfway. Step 4 has no clock: the demonstration takes as long as it takes, and you click on. Step 7 grows with the number of explanations out loud, three by default. The Short variation shortens every step and uses three tiles and two explanations.
Why does the prediction have to be on paper, not just in their heads?
Because a prediction in your head quietly shifts afterwards. A student who wrote nothing down easily thinks "I knew that" after the demonstration, and then there is nothing to explain. A written prediction stays put: the student sees in black and white that they thought something else, and that collision is what does the work. That is why in step 5 everyone holds up ✓ or ✗ for their own prediction, not for what the pair worked out later.
Why does the class watch in silence?
Because otherwise the answer is heard instead of seen. The student who calls out first decides for the whole class what the outcome looked like, and the rest no longer compare it with their own prediction. "Watch. In silence." gives everyone their own moment with the outcome. The conversation comes afterwards, in step 6, and by then everyone has something to say.
What if everyone predicted it right?
Then the class knew more than you thought, and that is an outcome too. Step 6 then turns to the reasons: were they right for the right reason? Often they were not. If the reasons held up as well, keep the explanations out loud short and ask a harder follow-up: what happens if you change one thing?
How many predictions go on the board?
Four by default, and you can choose from two to six. Type only predictions that really differ: ten times "the heavy one first" is one tile. Start with the most common prediction and end with the most surprising one. For a question with two possible outcomes, two or three tiles are enough; for an open question, five or six tiles show the whole spread.
Where do you find a good demonstration?
For physical science, chemistry, life science and math, ready-made preparations sit with the examples on this page. Beyond that, in your own lab manual or demo collection, or in a short video of an experiment you cannot run in class. Looking for ideas? Ask Microsoft Copilot for "demonstrations about [topic] for students aged [age] with an outcome many students predict wrong", and pick one you can check yourself. The test is always the same: the class can reason its way to the outcome, but cannot know it for sure.
What if the experiment fails?
Then you have a genuine science question: why did it go differently than expected? Mark ✓ and ✗ against what actually happened, and let the pairs work out in step 6 what was different about the setup. If you still want to show the intended outcome, play a video of the same experiment afterwards. To prevent it: try the experiment once before class.
Why does the screen draw a pair instead of students raising hands?
Because otherwise the same pairs explain every time, and the rest never needs an explanation. With the draw, nobody knows who will speak for the class, so every pair prepares an explanation in step 6. You draw again for each explanation, three times by default.
How is this different from Stand Up & Sit Down and Four Corners?
All three have the class take a position before anyone talks. Stand Up & Sit Down and Four Corners are about opinions and positions: there is no outcome that settles who was right. In Predict – Observe – Explain the demonstration settles it, and the conversation afterwards is about why a prediction did not come true. Choose Predict – Observe – Explain when there is an answer you can show.
Where does this structure come from?
The approach comes from physics education research. Around 1980, Audrey Champagne, Leopold Klopfer and John Anderson at the University of Pittsburgh used a sequence of demonstrating, observing and explaining to find out how students thought about motion. In their book Probing Understanding (1992), Richard White and Richard Gunstone turned it into Predict-Observe-Explain: first a prediction with a reason, then the observation, then an explanation of the difference. Crouch, Mazur and colleagues (2004) showed that it is the predicting that makes a demonstration a learning tool. In the Netherlands, the ShowdeFysica demonstration books and TU Delft's open textbook Show the Physics describe how to use prediction with demonstrations. LudensFlow added the build from cooperative learning: alone first, then in pairs, then with the class, and a drawn pair that explains.
Thinking before you know: which structure fits?
| Structure | Purpose | Group | Duration | When it fits |
|---|---|---|---|---|
| Predict – Observe – Explain | test a prediction against a demonstration | alone, then pairs | ± 17 min | an outcome the class can reason about |
| Which approach? | recognise the type of problem, without calculating | alone, then pairs | ± 16 min | the class knows the approaches, not when to use them |
| Design thinking in miniature | walk the whole design cycle once | teams of 3–4 | ± 35 min | before a design project, on a small task |
| Stand Up & Sit Down | show a position: agree or disagree | whole class | ± 9 min | quick picture + movement, binary choice |
| Four Corners | choose between four positions | whole class | ± 13 min | nuance and a conversation per corner |
| Numbered minds | make sure together that everyone understands | 2–6 | ± 16 min | one right answer, everyone accountable |
| Snowball (1-2-4-All) | gather and organize ideas | 8–32 | ± 18 min | open question, everyone alone first |
Sources
- Richard White & Richard Gunstone, Probing Understanding, London: Falmer Press (1992): the origin of Predict-Observe-Explain as a way to find out how students think.
- Catherine Crouch, Adam Fagen, J. Paul Callan & Eric Mazur, Classroom demonstrations: Learning tools or entertainment?, American Journal of Physics 72(6) (2004), 835–838: only watching gave no better understanding than not seeing the demo; predicting first did.
- Matthew Kearney, Classroom use of multimedia-supported predict-observe-explain tasks in a social constructivist learning environment, Research in Science Education 34(4) (2004), 427–453: POE with video in pairs, and the conversation that predicting and reasoning set off.
- Garvin Brod, Marcus Hasselhorn & Silvia Bunge, When generating a prediction boosts learning: The element of surprise, Learning and Instruction 55 (2018), 22–31: predicting beforehand did more than judging afterwards, and surprise appeared only after a prediction.
- Freek Pols & Peter Dekkers, Show the Physics, TU Delft OPEN Publishing (2024), doi 10.59490/tb.101: 99 physics demonstrations from the Dutch ShowdeFysica series, with the teaching of prediction around a demonstration.
Compiled by Oksana Oliinyk · Last updated 23 October 2026
Was this page useful?
After your answer we ask three short questions. Half a minute, and it genuinely helps us.