This paper outlines a general approach for analyzing siliceous microfossils from plants (phytoliths) in geologic sediments and uses this method to reconstruct the evolution of grasslands in the central Great Plains during the late Tertiary. Phytolith assemblage analysis is widely employed for reconstructing vegetation types in Pleistocene–Holocene paleoecology and archaeology, but no standardized methodology is currently available. Published phytolith studies are often system-specific and tailored after specific modern analogues. As a result, they vary in extraction technique, classification scheme, and method of inference. For ancient phytolith assemblages whose modern analogues are not well understood, a more general analytic approach is necessary. This analytic approach includes studying all size fractions (2–250 mm), using a classification scheme based on the available literature and a comprehensive modern phytolith reference collection, and comparing relative frequencies through time of all diagnostic phytoliths to interpret changes in habitat structure. This general method was applied to late Eocene to early Miocene sediment samples from northwestern Nebraska to deduce vegetation history and the results were compared to those obtained using a more specific technique [Strömberg, Paleogeogr. Paleoclimatol. Paleoecol. 177 (2002) 59]. In both studies, phytoliths were extracted by means of heavy liquid separation from the following lithostratigraphic units: (1) the late Eocene Chadron Formation and the early Oligocene Brule Formation (White River Group); (2) the late Oligocene–early Miocene Monroe Creek Formation, the early Miocene Harrison Formation and Anderson Ranch Formation (Arikaree Group); and (3) the early Miocene Runningwater Formation and Box Butte Formation (Ogallala Group). The two techniques, general vs. specific, produced results that differ markedly. Data from the general approach indicate the presence of relatively closed habitats in late Eocene to early Oligocene, typified by the presence of phytoliths from bambusoid grasses, woody and herbaceous dicotyledons, and palms. The introduction and spread of modern, open-habitat grasses, such as pooids, arundinoids, and panicoids, occurred between the late Oligocene and early Miocene. This resulted in a savanna or woodland type habitat by the early Miocene. In contrast, analysis using the specific method suggested the presence of open habitats in northwestern Nebraska from the late Eocene onward. The difference in interpretation based on the two approaches is principally due to inclusion of phytoliths >50 mm and the use of an expanded reference collection in the current study (general analytic approach). These results point to the importance of methodological choices in phytolith analysis and emphasize the usefulness of phytoliths as paleoecological indicators in the fossil record.