C4grasses form the foundation of warm-climate grasslands and savannas and provide important food crops such as corn, but their Neogene rise to dominance is still not fully understood. Carbon isotope ratios of tooth enamel, soil carbonate, carbonate cements, and plant lipids indicate a late Miocene-Pliocene (8–2 Ma) transition from C3vegetation to dominantly C4grasses at many sites around the world. However, these isotopic proxies cannot identify whether the C4grasses replaced woody vegetation (trees and shrubs) or C3grasses. Here we propose a method for reconstructing the carbon isotope ratio of Neogene grasses using the carbon isotope ratio of organic matter trapped in plant silica bodies (phytoliths). Although a wide range of plants produce phytoliths, we hypothesize that in grass-dominated ecosystems the majority of phytoliths will be derived from grasses, and will yield a grass carbon isotope signature. Phytolith extracts can be contaminated by non-phytolith silica (e.g., volcanic ash). To test the feasibility of the method given these potential problems, we examined sample purity (phytolith versus non-phytolith silica), abundance of grass versus non-grass phytoliths, and carbon isotope ratios of phytolith extracts from late Miocene-Pliocene paleosols of the central Great Plains. Isotope results from the purest samples are compared with phytolith assemblage analysis of these same extracts. The dual record spans the interval of focus (ca. 12–2 Ma), allowing us, for the first time, to investigate how isotopic shifts correlate with floral change.We found that many samples contained high abundances of non-biogenic silica; therefore, only a small subset of “pure” samples (>50% of phytoliths by volume) with good preservation were considered to provide reliable carbon isotope ratios. All phytolith assemblages contained high proportions (on average 85%) of grass phytoliths, supporting our hypothesis for grass-dominated communities. Therefore, the carbon isotope ratio of pure, well-preserved samples that are dominated by grass biosilica is considered a reliable measure of the proportion of C3and C4grasses in the Neogene.The carbon isotope ratios of the pure fossil phytolith samples indicate a transition from predominantly C3grasses to mixed C3-C4grasses by 5.5 Ma and then a shift to more than 80% C4grasses by 3–2 Ma. With the exception of the Pliocene sample, these isotopic data are broadly concordant with phytolith assemblages that show a general increase in C4grasses in the late Miocene. However, phytolith assemblage analysis indicates lower relative abundance of C4grasses in overall vegetation than do the carbon isotopes from the same phytolith assemblages. The discrepancy may relate to either (1) incomplete identification of (C4) PACMAD phytoliths, (2) higher production of non-diagnostic phytoliths in C4grasses compared to C3grasses, or (3) biases in the isotope record toward grasses rather than overall vegetation. The impact of potential incomplete characterization of (C4) PACMAD phytoliths on assemblage estimates of proportion of C4, though important, cannot reconcile discrepancies between the methods. We explore hypothesis (2) by analyzing a previously published data set of silica content in grasses and a small data set of modern grass leaf assemblage composition using analysis of variance, independent contrasts, and sign tests. These tests suggest that C4grasses do not have more silica than C3grasses; there is also no difference with regard to production of non-diagnostic phytoliths. Thus, it is most likely that the discrepancy between phytolith assemblages and isotope ratios is a consequence of hypothesis (3), that the isotope signature is influenced by the contribution of non-diagnostic grass phytoliths, whereas the assemblage composition is not. Assemblage-based estimates of % C4within grasses, rather than overall vegetation, are in considerably better agreement with the isotope-based estimates. These results support the idea that, in grass-dominated assemblages, the phytolith carbon isotope method predominantly records shifts in dominant photosynthetic pathways among grasses, whereas phytolith assemblage analysis detects changes in overall vegetation. Carbon isotope ratios of fossil phytoliths in conjunction with phytolith assemblage analysis suggest that the late Neogene expansion of C4grasses was largely at the expense of C3grasses rather than C3shrubs/trees. Stable isotopic analysis of phytoliths can therefore provide unique information about grass community changes during the Neogene, as well as help test how grass phytolith morphology relates to photosynthetic pathway.