When Student Data Contradict the Expected Result
An eighth-grade physical science class investigates which cup material best slows heat loss. Groups pour 200 mL of water at 60 °C into foam, paper, and bare metal cups, record temperature every 3 minutes for 15 minutes, and write a claim-evidence-reasoning conclusion. Group 4 reports that the paper cup lost only 9 °C while the foam cup lost 18 °C, contradicting both the class prediction and the textbook. Their lab log shows the foam cup lid was never seated and the thermometer rested on the bottom of the paper cup during readings. Group 4 concludes: 'Foam is the worst insulator because our data prove it.' Two other groups copy that conclusion into their own reports. Reports are due Friday, the teacher has one class period before then, and the district expects students to analyze sources of error rather than discard unexpected results.
1. Identify what is wrong with Group 4's investigation, decide whether the data should be kept or thrown out, and specify a redesigned procedure.
Read the worked answer
Two variables changed along with the cup material. An unseated lid on the foam cup opens a path for convective heat loss that the other cups do not have. A thermometer resting on the bottom of the paper cup reads conducted heat from the table and the cup base rather than the water column, biasing that reading in the direction the group observed. Because the treatment and the measurement conditions differ across conditions, the comparison is confounded, and the 9 °C versus 18 °C gap cannot be attributed to material.
Keep the data. Discarding them teaches students that inconvenient results disappear, which is the opposite of the standard being assessed. Instead, classify the data as 'not usable as evidence about materials, fully usable as evidence about experimental design.'
Redesign: same starting temperature within 1 °C, same volume, lids seated on all cups, thermometer suspended at mid-depth in every cup, identical room placement, and two trials per material with the average reported. Students should then name the one variable deliberately changed and list the variables held constant. If time is short, run only foam and paper with the corrected protocol, since the metal cup mainly serves as a contrast case.
2. Rewrite Group 4's claim as a defensible claim-evidence-reasoning statement, and explain how you would address the two groups that copied the conclusion.
Read the worked answer
Defensible rewrite: 'Our data cannot support a claim that foam is a poor insulator, because the foam cup had no lid and the paper cup thermometer touched the bottom, so the two cups were not tested under the same conditions. If lids and thermometer placement are controlled, we predict foam will lose less heat than paper because the air trapped in foam reduces conduction and convection.' This keeps the observation, states the limitation, and makes a testable prediction rather than a false certainty.
For the copying groups, treat it as a norms problem, not a cheating problem. Ask each group to show its own raw readings and ask which cup was its control. Neither group can answer, which makes the point without a lecture. Then set a class rule: a conclusion may cite another group's data only if you explain why that group's procedure was valid. Have the two groups write one sentence comparing their own numbers with Group 4's and noting whether their conditions matched. Ten minutes at the start of the period is enough; the remaining time goes to the redesign and rewritten conclusions.
3. Write a single short warm-up item for the start of that period that tests whether students can separate an observation from an inference and an evidence claim from a design flaw.
Read the worked answer
Item: 'A student writes: The sand cooled fastest (observation) because sand holds less heat (inference). The sand container was also the only one placed near the open window (design note). Label each part, then answer: which part would you need to change before comparing sand with soil, and why?'
The item works because it forces three separate judgments in about four minutes. Labeling the observation requires noticing what was directly measured. Labeling the inference requires noticing that 'holds less heat' is an explanation, not a reading. The design question requires recognizing that placement is a second difference between conditions, which is precisely the error Group 4 made with the lid.
Score it as three separate points rather than one, so the results tell you which misunderstanding dominates. If most students mislabel the inference as an observation, spend the period on the language of claim and evidence. If most catch the labels but miss the window, spend the period on controlled variables and go straight to the redesign. Either route keeps Group 4's data in the conversation as a teaching case rather than as a failure to be erased.