Method, and the traps this record sets

The build's standing source discipline, stated where a checker will find it

How sources were treated: first-hand reads govern; a claim is printed at the strength of the carrier that actually holds it, and a negative is printed only with the search that produced it. Official series outrank press, press outranks aggregators, and anything an aggregator alone carries — a founding year, a grant, a name — is treated as unverified until a real carrier holds it (one such claim, a Carnegie grant for a South Bend branch, is dissected in the libraries fragment as a worked example). The traps are named so they stop catching people: the other Roselands — Nebraska, New Jersey, Chicago, Louisiana (whose town site sits at the .org address a searcher will try first), and Rosedale in Parke County — have contributed more false Roseland facts than any other source class, and every one found was discarded whole rather than corrected into place; the name twins — two Ullery cemeteries in two townships, two Harrisons Crockett, the Ulery/Ullery spelling seam — are printed as pairs and never merged; the census seam — a tabulated count and its corrected count coexisting in the same reference article (conflict C1) — is the standing warning that a figure without its instrument is not a fact. How to check us: every article in this build names its carriers in its head-line; the registers of conflicts and gaps in the passes' notes state what would settle each open point, and the pull list above states what no screen can settle at all.