Classroom misconceptions

Why can a student give the correct answer yet still hold a mistaken underlying model?

A student writes the correct definition, gets full marks, and then, asked a week later why summer is warmer, says the Earth is closer to the Sun. Both performances are real. The temptation is to say they did not learn it properly, or forgot — but neither fits, since the marked answer was fluent and the wrong explanation comes with equal confidence.

So the assumption worth questioning is that a correct answer is evidence of a correct model. What if answering and explaining draw on different things?

Consider how the correct answer could be produced without the model. An exam question arrives with cues: this is the physics paper, this phrasing means the tilt question, and the tilt answer goes here. That is a rule keyed to a context, perfectly reliable inside it and connected to nothing else. Has the student learned anything false? No — something true, but local.

Meanwhile the intuitive model has been in place for years and was built from something more powerful than instruction: direct experience. Closer means hotter is a lifetime of fires and radiators; things stop unless you keep pushing is a lifetime of friction. These models earn credibility by working, constantly, in daily life. So why would a term of lessons displace them? The intuition has the better track record in the situations the student actually meets.

This has been measured, and the results are what made the field take it seriously. Conceptual inventories in mechanics found that traditional lecture courses moved students' intuitive answers about force and motion remarkably little, even while those students passed the exams. Later comparisons across many courses found classes built on interactive engagement roughly doubled the average gain on those inventories — an improvement, but note what it implies about the baseline.

Does the correct model eventually replace the wrong one? Apparently not simply. Work comparing experts with novices finds that people with advanced scientific training remain measurably slower to reject statements conflicting with intuition than statements that do not — suggesting the naive model is overridden rather than deleted. If that is right, teaching cannot aim at erasure.

There is also real disagreement about what the naive model even is: coherent alternative theories, or a loose kit of intuitive fragments assembled on the spot depending on the question. The two imply different teaching, and the argument is unresolved — a live debate, not a conclusion.

Two corrections. First, misconceptions are not stupidity or sloppiness; they are reasonable generalizations from real evidence, and often what makes everyday competence possible. Treating them as errors to be scolded away drives them underground rather than out. Second, more explanation is a weak remedy alone: if the intuitive model is never challenged where it visibly fails, a clearer statement of the correct one simply gets filed alongside it as school knowledge.

What helps follows from that diagnosis. The classic account names four conditions: the student must become dissatisfied with the existing conception, and the new one must be intelligible, plausible, and fruitful — visibly better at problems the old one could not touch. In practice that means prediction before demonstration, so the student commits and then watches the commitment fail; and assessment designed so the correct model and the misconception give different answers, since a question they both answer correctly is blind to this whole problem.

• Whether the naive model is ever really replaced or only reliably suppressed, and what that implies for expertise.
• How two-tier questions — answer plus reason — change what a test can detect.
• Why some misconceptions are near-universal across cultures while others are local to how a topic was taught.

Every term the collection defines is gathered in the glossary.