An atlas is only as findable as its labels, and its labels are pictures. No machine reads them: they curve along boundaries, they are set in five sizes, and half of them are in a copperplate italic. So they were read by eye and typed out, sheet by sheet, and this is the result — the whole county, not a sample of it.
12,219 of the entries are a name inside a boundary and nothing else: the ordinary householders and farmers who make up the body of every sheet. 222 are estates and 53 are farms, named as such on the plate. 157 are houses with names but no estate around them. The remaining 2,197 are what the plate marks besides property — mills, churches, schools, cemeteries, stations, reservoirs, quarries, ice houses, ponds, brooks and mountains.
Names are given exactly as the plate sets them. Where the atlas contradicts itself between sheets it is left contradicting itself: Tanglewood is R. S. Dixey on plate 21 and R. C. Dixey on plate 18, and both are here; Great Barrington’s plate 33 has D. Malampy where plate 34 has D. Malumpy and plate 36 has D. Malumphy, for what is plainly one family. Misprints stay misprinted — plate 36 titles Risingdale Scalc 400 feet to an inch. Nothing is normalized against the directories, and nothing is corrected against what a name later became.
445 entries carry the note reading uncertain. That means the letters are on the sheet and were not resolved with confidence at any magnification — a surname under a contour line, a name split by a parcel boundary, an rn that may be an m. Where even a doubtful reading could not be made, the name was left out rather than invented, so this is a transcription with holes in it, not a complete census of the ink.