1. Two segregates can be readily distinguished in the virgin hemlock forest which occupies the flats and lower slopes in the Joyce Kilmer Memorial, a part of the Nantahala National Forest in western North Carolina. The first of these types, restricted to the flats and lowermost parts of the slopes along the streams, is characterized by a dense undergrowth of Rhododendron and a sparse herb layer. The other type, occurring higher up on the slopes, has a very open shrub stratum but a fairly dense and diversified herb layer. 2. Tsuga canadensis is the dominant tree species in both types, with Fagus grandifolia, Tilia americana, Betula lenta, Fraxinus pennsylvanica, and Halesia carolina as almost constant associates. Liriodendron tulipifera and Acer saccharum are more abundant in the Herb type, while Betula lutea is restricted to the Rhododendron type. The total tree density is lower, but the BA is greater in the latter type (60% more). Hemlock reproduction is about equivalent at the seedling stage, but saplings are considerably less numerous in the Rhododendron type. Age-diameter relationships in hemlock indicate that the species is able to stand suppression relatively well but not so long as some other climax conifers. 3. Herbs which are constants for both types include Mitchella repens and Viola rotundifolia, observed in every stand, and Tiarella cordifolia, Laportea canadensis, Trillium sessile, Dryopteris spinulosa var. intermedia, and Oxalis montana, with lower presence values. In spite of their general distribution, these are the important herbs of the Rhododendron type. The characteristic herbs of the Herb type are therefore those numerous additional species which appear. Among these, Carex plantaginea, Medeola virginiana, Adiantum pedatum, and Polygonatum pubescens are exclusive and constant. 4. The soils are very similar under both types. They are sandy loams with an A-C profile, good incorporation of organic matter, and good structure. The better growth of dominant components in the Rhododendron type is attributed to better soil-water relations, owing to topography and possibly to microclimatic differences. 5. A comparison is made with the two types (Rhododendron and Polycodium) of virgin hemlock forest found in Ravenel's Woods. Differences in species composition and in growth of hemlock lead to the conclusion that there are at least four types of hemlock cove forest, which are believed to be relic association-segregates of Braun's Tertiary Mesophytic Climax. The considerably greater BA of hemlock in Ravenel's Woods and the differences in associated species are attributed to differences in general climate (higher precipitation and lower temperature). 6. The sampling method used in this study is discussed. It consists of a strip random sampling in which each strip is in turn systematically subsampled at a 50% intensity.