A cognitive architecture-based model of graph comprehension

I present a model of expert comprehension performance for 2×2 "interaction" graphs typically used to present data from two-way factorial research designs. Developed using the ACTR cognitive architecture, the model simulates the cognitive and perceptual operations involved in interpreting interaction graphs and provides a detailed characterisation of the information extracted from the diagram, the prior knowledge required to interpret interaction graphs, and the knowledge generated during the comprehension process. The model produces a scan path of attention fixations and a symbolic description of the interpretation which can be compared to human eye movement and verbal protocol data respectively, provides an account of the strategic processes that control comprehension, and makes explicit what underlies the differences between expert and novice performance.