Silicification is an important mode of fossil preservation but the extent to which silicified material represents an unbiased sampling of the total fossil assemblage within a given rock sample remains poorly quantified. Here, we use paired analyses of thin sections and acid-extracted silicified specimens from the same samples to examine the biases introduced during silicification of Lower Triassic Virgin Limestone carbonates preserved in the Muddy Mountains of southern Nevada. Bivalves dominate most thin sections in the point count data, but rarely silicify completely enough to be recognized in residue. Echinoderms and gastropods are less abundant in thin section but dominate the residues. The abundances of these groups in thin section and residue are only weakly correlated. These findings suggest that although silicification generally captures relative trends in proportional abundance of higher taxa among samples, the silicification process can be taxonomically biased. Given the biases that can occur during silicification, it should not be assumed that silicified collections present a pristine picture of taxonomic or paleoecologic composition. Petrographic analysis has the potential to illuminate the reliability of paleontological data based on silicified collections.