In defense of a modular architecture for the number-processing system: reply to Campbell and Clark.

In several recent articles we have developed a model of the cognitive number-processing and calculation systems. Campbell and Clark (1988), commenting on one of these articles (McCloskey, Sokol, & Goodman, 1986), called into question our model's assumption of a modular functional architecture and a single form of internal numerical representation. Campbell and Clark proposed as an alternative a nonmodular encoding-complex view. In this reply we discuss the results offered by Campbell and Clark as evidence against our model, arguing that several of these results are in fact consistent with the model and that the remaining results, while raising significant issues, by no means justify abandonment of the modular framework and the constraints it imposes. We also point out that whereas our model provides specific, well-motivated interpretations for a substantial body of empirical findings, the encoding-complex view is so underspecified and unconstrained as to be vacuous.