To assess the degree to which forest litter reflects the source forest, three 1-ha plots of temperate deciduous forest were mapped and litter accumulating in these forests was sampled. Identity, position, and diameter of all stems 2 cm or larger diameter at breast height are known for each forest. Composition of the leaf litter is governed by two key factors: (1) abscised leaves are deposited primarily on the forest floor directly underneath the canopy that produced them, and (2) the leaf mass of a species is highly correlated with its stem cross-sectional area. These factors produce autochthonous litter samples that correspond closely in composition to the forest within a circle of canopy-height radius or less. Even relatively small litter samples (350 leaves) consistently contained all the common species in the local area. However, the rarer tree species were seldom recovered in the litter samples. Correlation coefficients for litter mass and basal area by species are typically over .80.These observations have three important implications for interpreting autochthonous compression-fossil assemblages. First, approximate relative abundances of locally dominant and subdominant forest taxa can be obtained from relatively small samples of autochthonous compression-fossil assemblages. Second, representation of rare forest species, even in large fossil samples, will be fortuitous. For this reason, complete species lists and consistent estimates of richness cannot be derived directly from most existing samples of autochthonous compression-fossil assemblages. Third, the strong tendency for leaves to fall beneath the canopy of the tree that sheds them suggests that properly sampled autochthonous fossil leaf assemblages may yield information on crown size of individual trees and the spatial distribution of individuals and species, aspects of vegetational structure that have been thought accessible only in well-preserved “fossil forests” with standing trunks.