Previous research has shown that students construct equations for word problems in which many of the terms have no referents. Experiment 1 attempted to eliminate some of these errors by providing instruction on canceling units. The failure of this method was attributed to the cognitive overload (Sweller, 2003) imposed by adding units to the equations and to the mechanical nature of the intervention (P. W. Thompson, 1994). Experiment 2 used contrasting pairs instruction (D. L. Schwartz & Bransford, 1998) that required students to determine which of 2 mathematical expressions referred to a real-world quantity. This brief practice at the beginning of the session significantly increased the number of correct equations relative to a control group that practiced constructing equations. The contrasting cases instruction can help prepare students for additional instruction, such as the iconic manipulations used in the Animation Tutor[TM] (Reed & Hoffman, 2006; see also Reed, 2005) modules.