Delimiting meaningful parataxa in dispersed dicot leaf floras is among paleobotany’s greatest methodological challenges. Neogene oaks present a particularly thorny problem because of their morphological variation. Nonetheless, biologically meaningful assessments of taxonomic richness are essential to a variety of studies including those focusing on diversification of clades and paleoclimate reconstruction. To develop a repeatable method for delimiting taxa from a mixed collection like those in the fossil record, we assembled leaves from vouchered herbarium specimens from three co-occurring and morphologically similar California oaks: Quercus agrifolia Nee, Q. chrysolepis Liebm., and Q. wislizenii A.DC. For each leaf, we took digital images, measured 11 quantitative variables, and calculated five derived variable. These variable have been used to diagnose parataxa in Neogene oaks. ANOVA also showed that length, width, length/width, and average secondary vein distance showed statistically significant difference among species. However, no single variable distinguished all three species. Discriminant analysis correctly assigned 82% of the specimens to species. This analysis shows that (1) a quantitative morphometric analysis can be useful in delimiting leaf paraspecies within oaks, and (2) that some portion of any fossil collection is likely to remain in the morphological overlap between parataxa. A morphometric approach will help refine diagnoses of fossil leaf parataxa to include quantitative and empirical ranges of variation. To demonstrate this approach, we performed the reverse analysis on a large sample of Middle Miocene oaks from the Mint Canyon Formation of southern California. Multivariate ordination suggests extensive overlap in morphology among the oaks in this collection. Only two taxa, rather than the three originally described, can be conservatively recognized from this collection.

Key words: leaf morphology, morphometrics, Neogene, Quercus