Hold until first response/capture; do not score a single preferred creative question or require all adult caveats in a child's speech. Exact source detail/reasoning, not reading, motor, volume, timed finish, record length or arbitrary repeat-count criterion. A partial truthful response informs next teaching. Neutral exact read-aloud, scribing/sign/AAC support attributed; later short components permitted. Two paper cases do not establish actual SFI02 investigation, full SFU03 composition, full achievement or a term grade.
Check AJ · the first contact report
Fixed source: ROOF-CHECK-AJ-33, original saved text/diagram, fictional notebook; creator/date/purpose in source. Batch J paper roof, low cardboard supports, broad very light card K/middle placement. Prior prediction keeps gap. Before reported gap discernible; after reported middle touches mat, no support/card shift. Hence reported result differs from prediction; earlier available baseline supports reported first-contact sequence only in this fictional case. No actual observer/test supplied; no universal all-paper strength, real shelter/weather/child sensory claim.
Open criterion: a relevant safe next question/check with local preparation/access, not a preferred fold or fixed number of repeats. It might ask to observe another comparable try, another approved paper form/source or a directly accessible unloaded property. Recognise report cannot settle every paper roof. A plan is not an actual trial.
Possible response after capture: “The notebook reports a gap before, contact afterwards; that differs from its prediction. I would like another safe try with the same checked setup to see what happens.” Another useful question is welcome; repeat is not compulsory.
Next teaching: if prior/after confused, use new unrelated practice source to place BEFORE words before action and compare later. If contact inferred without baseline, inspect actual unloaded setup only if available, otherwise attributed report/model. If all-paper claim, compare specific conditions with one unknown. Preserve first response.
Check AK · two different gaps
Fixed source: ROOF-CHECK-AK-34, same original creator/date, fictional reports. Batch S, same supports/light card M/middle placement reported; flat row narrow gap retains after gap; creased row different wide gap reports after contact. Both before gaps discernible, no slip reported. Form AND support gap differ, so rows cannot isolate form as cause or prove creasing always strengthens every roof. Neither prior prediction recorded; cannot decide a match/mismatch from invented after-only prediction. Schematics not actual measurements/material verification.
Open criterion: relevant safe child question and useful condition to check/keep. Examples include same verified gap/source/load/placement when comparing forms, a locally approved repeat, or a different accessible unloaded property. Accept uncertainty/one short question; no Year 1 fair-test checklist or exact fold pattern. Explicitly note a mismatch if actual matching cannot be established.
Possible response after capture: “The flat row keeps a gap and the creased row touches, but the gaps differ as well as the form. No prediction was recorded. I wonder what these forms would do with the same checked gap and card.” This is a proposal, not an experiment or single correct answer.
Next teaching: if child says crease changed material, revisit verified paper identity versus form using unrelated source/actual object. If cause overclaimed, notice the second setup change without demanding complex variables vocabulary. If missing prediction treated as wrong, distinguish NOT RECORDED from a prior prediction that differed. Use later actual accessible observation only if available/approved.
Bounded judgement
SFU03 needs actual object composition/material identity plus observable property description for the relevant partial claim; a drawing/creased form alone does not establish it. SFH01 questions/observations can support a particular actual example; fictional observer isn't human professional evidence. SFI01 prior prediction must precede genuinely new action; source prediction reasoning separate. SFI02 needs actual safe sensory access, not report/model converted to child's vision. SFI03/SFI04 can sample bounded supplied-record reasoning; actual pattern and own prediction comparison need corresponding real rows/records. SFI05 actual explanation can be to educator/willing person; private rehearsal not actual sharing. Use a body of actual dated samples and local professional judgement, no all-code/whole-year claims from AJ/AK.