Our mental words are populated with concepts — rich representations of knowledge about things in the world (e.g., diamonds, pumpkins). In language, words are used to refer to these concepts (e.g., “diamond”, “pumpkin”) and to communicate with others. This is quite impressive given that a word does not activate the same information each time it is used: conceptual information is flexibly activated based on the context. For example, the phrases “raw diamond”, “baseball diamond”, and “diamond eyes” evoke different kinds of diamond information. This flexible concept use is not only exemplified in creative language, but in creative thought and natural language more generally. The goal of this thesis was to leverage methods in cognitive neuroscience, network science, and computational modeling to explore the kinds of conceptual structure that can support this flexible concept use. In the first study (Chapter 2) I capture the global structure of concepts in novel feature-based networks, and show that aspects of this network structure relate to text-based and empirical measures of flexible concept use. I subsequently narrow in on the local representations of conceptual features that relate to flexible concept use by observing what happens when concepts combine. In one fMRI study (Chapter 3) I show that feature uncertainty predicts the extent to which features (e.g., green, salty) are flexibly modulated in the brain during comprehension of adjective-noun combinations (e.g., “green pumpkin”, “salty cookie”). In follow-up studies (Chapter 4) I further reveal the relationship between feature uncertainty and flexible feature activations in combined concepts. In combinations that modify conceptual brightness (e.g., “dark diamond”, “light night”), an explicit behavioral measure of conceptual feature modulation is predicted by feature uncertainty as well as by a related predictive combinatorial Bayesian model. An associated fMRI study reveals that flexible feature modulation and feature uncertainty relate to responses in left inferior frontal gyrus (LIFG) and left anterior temporal lobe (LATL), suggesting roles for these regions in flexible concept activation. Taken together, this work reveals relationships between conceptual structure and flexible concept use in behavior and in the brain.