A study was conducted that compared students’ solutions to word problems that differed by the degree to which they were grounded in realistic settings. In a repeated measures design, fourth-graders (N=115) solved three types of equal sharing problems: grounded problems, containing familiar objects, such as brownies; semi-grounded problems with less familiar objects, such as lengths of rope measured in centimeters; and idealized problems, containing non-words (e.g., “porams”). No difference was observed between grounded and idealized problems, but both problem types yielded superior performance relative to semi-grounded problems. Prior knowledge of fractions and multiplicative structures was associated with greater performance overall. One implication is that only partial familiarity with problem contexts may activate irrelevant schemas and hinder problem-solving performance.