Michigan Pioneer and Historical Society collections corpus

Summary

The project holds a complete text-derivative working corpus of the 40-volume Michigan Historical Collections run and its two contemporary indexes. The series is a broad Michigan historical compilation, not a Beaver Island archive, but it contains material relevant to the project’s early-state, treaty-era, Mackinac, Strangite, and historiographical research paths.

The corpus is retained as a searchable source layer and lead generator. It does not make every reprinted document primary in its own right, and a hit must be checked against its printed page image and underlying source before it supports a KB claim.

Initial screen and limits

The reproducible lexical screen counts “Beaver Island” and “Beaver Islands” across whitespace-expanded OCR, rather than requiring one literal space; Internet Archive OCR often inserts line breaks inside a printed phrase. It is a discovery screen, not a transcription or a count of historical events. Any hit can be a later reprint, editorial summary, unrelated namesake, or an OCR error, and must be checked on the printed page before use.

The screen also checks “Isles du Castor,” but a zero result for that or any other form is not evidence that the series contains no relevant material: OCR can fail, historical terminology varies, and relevant records may be indexed under people, missions, treaties, Mackinac, or other regional terms.

The screen’s per-volume counts are uneven. Most volumes return zero or a handful of hits; a few stand out as denser candidate material. Volume 32 (1902) returns 103 “Beaver Island”/“Beaver Islands” hits and 4 “Isles du Castor” hits, by far the highest of any volume in the run. Volume 18 (1891) returns 18 hits, Volume 39 (1915) returns 14 plus 1 “Isles du Castor” hit, and Volumes 9 (1886) and 26 (1894) each return 6. The two contemporary indexes, covering volumes 1–15 and 16–30, return 5 and 4 hits respectively. These are discovery counts from the held manifest, not a ranking of historical importance, and every volume — including the many at zero — can still hold relevant material indexed under a name other than the island’s.

The two indexes remain especially useful for an issue- and page-directed review. They should be used to narrow candidate passages, not treated as substitutes for reading the cited volume or for resolving whether a reprinted text is accurate.

Individually catalogued volumes

The 40-volume text corpus is a discovery layer; it is distinct from the smaller set of volumes the project has catalogued as their own citable articles. Ten of the forty volumes currently have an individual KB record: Volume 1 (1874), Volume 9 (1886), Volume 12 (1887), Volume 21 (1892), Volume 23 (1893), Volume 26 (1894), Volume 28 (1897), Volume 32 (1902), Volume 38 (1912), and Volume 39 (1915). Those ten are the volumes cited by name elsewhere in this knowledge base; the remaining thirty are held only as the raw text corpus described above, and a claim should not be attributed to “the Michigan Historical Collections” generally when only one of the thirty uncatalogued volumes has actually been read.

Added 2 August 2026. Recent changes