Solving simple arithmetic word problems is a major ability that children must acquire throughout the primary-grade mathematics curriculum. However, this skill is often challenging for them. For instance, "unknown referent problems" are more difficult to solve than "unknown compare problems." In unknown compare problems, the relational term (e.g., "more than") is consistent with the arithmetic operation (e.g., addition) required to find the solution, whereas in unknown referent problems, the relational term (e.g., "more than") is inconsistent with the arithmetic operation (i.e., subtraction). To determine whether solving unknown referent problems relies on the ability to inhibit the heuristic "add if more, subtract if less" strategy, we devised a negative priming (NP) paradigm for children, adolescents, and adults. For each test trial, an unknown referent problem served as a prime, in which the inhibition of the heuristic strategy was necessary and was immediately followed by an unknown compare problem that served as a probe, in which the heuristic strategy became the appropriate strategy to use. For each control trial, a neutral problem served as a prime and was followed by an unknown compare problem as a probe. All participants performed the unknown compare problems more slowly when they were preceded by unknown referent problems, which reflected typical NP effects. Our results confirm that success in completing unknown referent problems reveals not only participants' ability to grasp the underlying logic of the problem but also their ability to inhibit a misleading strategy. This study sheds light on the importance of inhibitory control in learning mathematics.