Trophic niche studies are essential for evaluating ecological interactions between and within species and their evolutionary implications. For example, fundamental aspects of a wide range of hypotheses concerning population divergence, evolution of sexual dimorphism, and adaptations to fluctuating environments rely on dietary evidence. The accuracy of different methodologies used to estimate trophic specializations is therefore a fundamental issue. Under the assumption that direct observations of gastric contents provide accurate information about dietary preferences, I examined how reliably diet is reflected in fecal pellets. I conducted two main comparative tests on living lizards. First, I fed individuals of 23 species with hard- and soft-bodied organisms, and compared fecal pellets. Second, I examined prey items from natural diets represented in feces and gastric contents (stomach-flushing) in a wild population of the lizard Liolaemus tenuis. My results reveal that fecal samples provide inaccurate estimates of lizard trophic preferences, mainly because soft-bodied organisms are destroyed by digestive processes. Even though soft-bodied prey may be essential dietary items (as inferred from gastric analysis), these organisms may be almost entirely absent from feces. I suggest that direct gastric analyses should be the preferred method for analyzing reptilian diets.