Taphonomic processes may filter in a biased manner the tiny fraction of leaves preserved as fossils. A common perception is that large leaves are underrepresented; this is based both on intuition (large leaves are more likely to break apart) and some observations of extant vegetation. Characterizing leaf area correctly is critical for reconstructing climate and for studying evolutionary and biogeographic patterns. In extant dicotyledonous angiosperms, leaf area generally scales with the inverse of second-order vein density. This scaling offers the potential to test if fossil leaf fragments were derived from leaves that were larger than complete (or nearly complete) fossil leaves of the same species. Here we test vein scaling on 573 complete leaves from the latest Cretaceous Hell Creek Formation and earliest Paleocene Fort Union Formation in the Williston Basin of western North and South Dakota. We find a strong scaling similar to extant vegetation, with a somewhat shallower slope (1.67 vs. 2.04) and lower r² (0.64 vs. 0.80). We apply these two scalings to 41 species-site pairs from the Williston Basin that are each represented by complete (n = 355) and fragmented (n = 387) leaves. With both scalings, the reconstructed leaf areas of fragments are on average 10% larger (±36% 1σ) than their complete companions. This small but noisy signal means that the underrepresentation of large leaves, as captured by our study design, is probably not critical for most fossil applications. Comparing directly the reconstructed areas of complete and fragmented leaves appears reasonable, thus expanding the usefulness of fossil leaf fragments.