Modals (e.g., "can," "must") vary along two dimensions of meaning: "force" (i.e., possibility or necessity), and "flavor" (i.e., possibilities relative to knowledge [epistemic], goals [teleological], or rules [deontic] …). Comprehension studies show that children struggle with both force and flavor dimensions of modals. However, given the complex one-to-many mappings from forms to meanings, it is not clear what force or flavor children assign to the modals being tested. In this study, we use a sentence-repair task to test which modals 3- and 4-year-old children themselves prefer to produce in teleological (goal-oriented) and epistemic (knowledge-based) possibility and necessity contexts, and how these preferences differ from those of adults. Our results provide a first controlled look at which modals children use to express the major flavor and force dimensions of modal verb meanings. We shed new light on children's modal systems, and show that learners generally distinguish modal flavors but struggle distinguishing forces.