Data provided by Carlson, Khoo, Yaure, and Schneider (1990) show that with practice, Ss solve computerized versions of logic gates more quickly, with fewer requests for help, and more cheaply (in simulated costs). Ss reported that they changed strategies as they became more familiar with the task, began working backwards from the goal, and used larger perceptual units of analysis. Although these results provide an interesting account of the development of problem-solving expertise, they cannot be used to differentiate between single-workspace or distributed processing models of working memory.