Previous research has cast doubt on the efficacy of utilizing external representations as an aid to solving word problems. The present study replicates previous findings that concrete representations hinder college students' ability to solve probability word problems, and extends those findings to apply to a multimedia instructional context. Our concrete-representation group was instructed in how to use Venn diagrams to solve probability problems involving non-mutually exclusive events; our procedural group was instructed only in the formulation of equations. Results showed that the procedural group outperformed the concrete-representation group; additionally, there was a main effect for performance on transfer problems, favoring near-transfer problems over far-transfer problems. Contrary to previous findings, however, in the present study cognitive load did not appear to be a factor in the lower performance of the concrete-representation condition. (Contains 1 figure and 1 table.)