WVitness trees from the land survey records have proved useful as indices of the character of the original vegetation and have served as a primary basis (Sears 1925; Kenoyer 1930, 1940; Blewett and Potzger 1950; Rohr and Potzger 1950; Potzger and Potzger 1950) or as a confirmatory or supplementary source of information (Lutz 1930; Crowl 1937; Dobbins 1937; Shanks 1937, 1938; Gordon 1940; Fassett 1944; Cottam 1949; Norris 1949) in a number of vegetation surveys in Ohio, Michigan, Indiana, Pennsylvania, New York, and WN;isconsin. The present study is based on the land survey records for Shelby County, Ohio, a typical sample area in the heart of the beech-maple forest region of western Ohio (Transeau and Sampson 1935; Braun 1950). Its objectives are: (1) species composition of the original forest based on the witness trees as a representative sample of the vegetation; (2) forest composition differences with differences in soil type and relief; (3) changes in species composition in the secondary forest; (4) analysis of the degree of association between the species used as witness trees in the original land surveys, based on the random pairs of trees used as witnesses at section and quarter-section corners; an(l (5) inferences as to habitat range of major species and their role in the forests. The original land surveys of Shelby County were made by nine surveyors during the years 1798-1832. The typical congressional land survey procedure was used, the land being divided by a grid into sections one mile square. Survey points were established at section corners an(l at intervening half-mile intervals (quarter-section corners) on alternate survey lines. Thus, each mile-square section had six survey points touching it, but the over-all average was two points per square mile, involving four witness trees per square mile. WVith the addition of survey corners along the river and county lines, 2030 trees were listed for this sample area of approximately 413 square miles. These witness tree records were transcribed onto topographic maps for certified copies of the survey records in the County Courthouse, Sidney, Ohio. Several assumptions underlie the analyses and conclusions of this paper: (1) that the species names in the survey records are consistent and identifiable; (2) that the samples are large enough to have validity; (3) that the witness trees are random pairs; and (4) that species can be used as habitat indices. The first two of these assumptions have commonly been discussed in papers which have justified the use of witness trees as a basis of vegetation inferences (S e a r s 1925; 1 Contribution from the Botanical Laboratory, The University of Tennessee, N. Ser. 146. Review of the manuscript and suggestions concerning soil interpretation and terminology by W. S. Ligon, Principal Soil Scientist, Division of Soil Survey, U. S. Dept. of Agriculture, are gratefully acknowledged.