Goal: use another observation to check a bounded statement. Proposed opportunities: AC9SFH01, AC9SFI01–AC9SFI04.
Prepare: same actual sample and target, log, marked view; educator-approved repeat. Choose one question arising from earlier raw rows, not an imposed expected outcome.
- Read the first claim · 3 min. “I saw ___ through this sample from here.” Ask whether a later view might match, differ or be unclear.
- Predict a repeat · 4 min. Record the child's prior words. State what will be kept in the same place. A different actual light condition must be recorded, not hidden.
- Make the second view · 7 min. Recheck the uncovered baseline, then repeat safely. Keep the second raw observation even if it disagrees; pause/change property or use an attributed report when access is unsuitable. No unavailable baseline is evidence that the cover hid the target.
- Compare records · 5 min. Child indicates same/different/unclear and one possible setup change. A suggested explanation is a question to check, not proved cause.
- Revise the helper message · 4 min. Include both dates/views when needed. “Both views showed ___” or “The views differed; we need another check” are useful outcomes.
- Keep and next · 2 min. Keep prior words, two rows and interruptions. Model-only access gets record-reasoning evidence, not direct experiment evidence.