Condition

CORRECTED IN V3.2: this query asked an affliction question and answered it with demons. Of the 718 MENTIONS_ENTITY edges reaching a Condition, 314 reached a THREAT and 88 a PATHOGEN_OR_CAUSE, so rakshas, sorcery, curses, worms and poison were all ranking as diseases. The 8 THREAT and 2 PATHOGEN_OR_CAUSE conditions are now excluded here by `condition_kind` and remain reachable through passages_protecting_against and condition_neighbourhood, both of which stay deliberately broad. Naming an affliction is weaker than treating it. CORRECTED IN V3.1: this caveat read 'There is deliberately no separate fever entity: takman- forms sit in YAKSMA-DISEASE'. That is no longer true -- VG:CONCEPT:TAKMAN-FEVER exists and carries 33 mentions -- so the caveat was denying the existence of an entity the same database returns. The reason it originally gave still matters, and is now a live hazard rather than a settled policy: the takman- paradigm is split across two entities, so a query filtering on either alone undercounts fever. Read TAKMAN-FEVER and YAKSMA-DISEASE together, and treat the split as unmeasured until someone reconciles the paradigm.

`passage_count` counts DISTINCT passages reaching the object through the lexical, conceptual and ritual predicates this surface counts, and is the same figure the detail view reports. It is NULL, never 0, where no passage reaches it: the mention layer's measured recall is 0.8857, so a null is a lower bound and not an assertion that the corpus is silent.