Forest canopies are often stratified by species; little is known about the depth distribution of tree roots by species because they are not readily identified. We used diagnostic characteristics of wood anatomy to distinguish roots by species, and checked the validity of our separation using genetic techniques. We also learned to recognize roots by their gross morphology, which many researchers have thought was not possible. We applied these methods to test for differences in the rooting depth of sugar maple (Acer saccharum Marsh.), American beech (Fagus grandifolia Ehrh.), and yellow birch (Betula allegheniensis Britton) in two northern hardwood sites. We also identified viburnum (Viburnum alnifolium) and ash (Fraxinus americana). We studied the distribution of roots with depth by fitting curves of the form Y = 1 – β^d describing cumulative root fraction (Y) as a function of depth (d). There was no significant difference across species in the decline of root mass with depth (β); similarly, the proportion of species mass was also indistinguishable by depth. There was a significant difference in the distribution of roots by size class, with fine roots more concentrated near the soil surface. The two sites differed significantly in rooting depth, with roots at the Hubbard Brook site distributed more deeply than at the Bartlett site. The method we used requires somewhat more time than the already time-consuming process of picking roots from soil and sorting them by size class, but it is less expensive than genetic characterization. If the goal were simply to identify the distribution of roots by species, without measuring biomass, it would be possible to use genetic techniques alone, which might be more efficient than root picking.