When the usual destination is occupied.

Learning from the result

An actual occupied-destination refusal can be linked to a later contact-confirmed corrective demonstration. The correction comes from what was observed, not from treating a teacher-proposed coordinate as a learned fact.

Right-of transfer errors in the recorded offset and distractor layouts were 8.68 mm and 4.75 mm. Matched ordinary right placement and an earlier left correction were retained. A fresh-process comparison took place about 58 minutes after teaching.

What the controls do—and do not—say

Read-only planning with correction disabled chose occupied endpoints. That comparison is a planning ablation, not a physical off rollout. This is narrow clearance adaptation across known object pairings and a few layouts. It does not establish general semantic repair or open-ended English.

This checkpoint recorded 19 native events and 790 visual pairs. Those counts are scoped records—including refusals—not scores for intelligence. Later posts retain this checkpoint and report the subsequent totals separately.