Based on recent findings in cognitive neuroscience and psychology as well as computational models of working memory and reasoning, I argue that fluid intelligence (fluid reasoning) can amount to representing in the mind the key relation(s) for the task at hand. Effective representation of relations allows for enormous flexibility of thinking but depends on the validity and robustness of the dynamic patterns of argument–object (role–filler) bindings, which encode relations in the brain. Such a reconceptualization of the fluid intelligence construct allows for the simplification and purification of its models, tests, and potential brain mechanisms.