Most research on differential item functioning (DIF) focuses on methods for detection rather than on understanding why DIF might occur. This study was designed to investigate whether two alternative approaches to parsing items based on structural characteristics related to particular cognitive strategies could be used to help explain gender DIF. The two coding schemes, one based on Multicontext Theory (Ibarra, 2001) and the second based on a cognitive structure analysis described by Gallagher (1998), were used to identify features of mathematics items that might specifically help or hinder correct responses from male or female participants. Previous research on reading test items suggested that Multicontext Theory might be useful in predicting gender DIF. In Study 1, a coding scheme was developed based on Multicontext Theory. In Study 2, predictions based on the Multicontext coding scheme were compared with predictions based on an application of Gallagher's method. The results suggested that the method based on Multicontext Theory was more effective than Gallagher's method, but both approaches demonstrated some utility in predicting gender DIF.