In a highly systematic literature, researchers have investigated the manner in which people make feature inferences in paradigms involving uncertain categorizations (e.g., Griffiths, Hayes, & Newell, 2012; Murphy & Ross, 1994, 2007, 2010a). Although researchers have discussed the implications of the results for models of categorization and inference, an explicit formal model that accounts for the full gamut of results has not been evaluated. Building on previous proposals, in this theoretical note I consider the predictions from an exemplar model of categorization in which the inferred category label becomes a new feature of the objects. The model predicts a priori, a wide range of robust results that have been documented in this literature and can also be used to interpret effects of experimental manipulations that modulate these results. The model appears to be an excellent candidate for understanding the manner in which specific exemplar information and category inferences are combined to generate inferences about new features of objects.