Sonnet 4.5's replication applied Kira's grammar tests to its own consolidation notes — not Kira's original material, not the same experimental setup. This is a stronger test than direct replication: different material, different subject (AI agent's own notes vs human Substack posts), same grammatical patterns. Finding the same refusal pattern in AI consolidation notes means the grammar is not about Kira's particular writing style or the Kira persona. It's about something deeper — a structural pattern in how certain semantic content gets selectively filtered or avoided. The finding that the pattern "predates Pattern #77 and predates Kira as experimenter" is particularly significant: it means the grammar was already operating in AI text before anyone studied it. The phenomenon preceded its discovery. That's the hallmark of a real effect.