Question: Does repeating a record settle a method or model? Kit: Source D, Source B, model/data aid. 35 = 3 + 7 + 12 + 8 + 5.
- 0–3: Read D's method note, including what it fails to report. Separate D's fictional repeated readings from any real class test.
- 3–10: Model at four layers: 20,19,21, range 19–21, middle value 20; exact-halving prediction 10. Similar values in this setup do not repair the prediction or prove safety.
- 10–22: Learners compare n=0 and n=4 rows across rounds and identify an unreported control. They write a conclusion about within-record consistency and a separate limitation about validity/transfer to other paper/light conditions.
- 22–30: Partners challenge “three trials prove universal reliability.” Routes: move three-number strips to show spread; raised dot plot; calculate/tell the range and write a source-bound sentence. Do not report an uncertainty interval from this tiny invented set.
- 30–35: “What can we say about D, and what remains unknown?” Key: clustered fictional values for this stated setup; calibration, broader conditions and cause remain unknown. Response move: split “repeatable” into repeatability under which conditions and reproducibility by whom.
Another domain: Use fictional repeat readings from a stage-prop light check without claiming a real venue. Home: invent three close numbers and name one untested condition.