A widely observed phenomenon in children’s word-extensions and generalizations is the characteristic-to-defining shift, whereby young children initially generalize words based on typical properties and gradually transition into generalizing words using abstract, logical information. In this paper, we propose a statistically principled model of conceptual development grounded in the trade-off between simplicity and fit to the data. We run our model based on informant-provided family trees and the real-life characteristic features of people on those trees. We demonstrate that the characteristic-to-defining shift does not necessarily depend on discrete change in representation or processes. Instead, the shift could fall out naturally from statistical inference over conceptual hypotheses. Our model finds that the shift occurs even when abstract logical relations are present from the outset of learning as long as characteristic features are informative but imperfect in their ability to capture the underlying concept to be learned—a property of our elicited features.