Discussion Process Decomposition from Double Dissociation of Subprocesses

Between-Task Double Dissociation is defined as a particular pattern of effects of two factors (manipulations) F and G on performance measures MT1 and MT2 in two tasks implemented by different complex processes. F and G might be the amounts of damage in two brain regions. The factors must influence the measures selectively: MT1 influenced by F but invariant with respect to G; MT2 influenced by G but invariant with respect to F. Let “dd(MT1, MT2; F, G)” denote this set of two influence properties and two invariance properties. Investigation of such cases may be called task comparison; what are dissociated are the complex processes used to carry out the two tasks. I consider below a type of within-task double dissociation that can serve as evidence for the modularity of subprocesses of a single complex process, and is the basis of a method for process decomposition, an idea developed in detail in Sternberg (2001) with a dozen diverse applications; see also Sternberg (in press). Unlike task comparison, which is often used in a way that requires modularity to be assumed without test (Shallice, 1988, Ch. 11; Sternberg, 2001, Appendix A.1.), this method incorporates such a test. The inferential logic, summarized briefly below, depends on whether our measures of the subprocesses are “pure” or “composite”. The comments that follow the summaries are intended to clarify some aspects of the logic; several also apply to between – task double dissociation. Within-Task Double Dissociation: Pure Measures. Suppose we have two different measures MA and MB of performance in a single task. Examples include the sensitivity and criterion measures of signal detection theory (SDT) in a brightness discrimination experiment (McCarthy and Davison, 1984), and the amounts of fMRI activation in two brain regions during a number comparison task (Pinel et al., 2001). Suppose further that MA and MB are pure measures of two different parts, A and B, of the complex mental or neural process used to carry out the task. That is, MA = MA(A) (MA depends only on A) and MB = MB(B). We wish to ask whether A and B are separately modifiable, such that each can be changed independently of the other. Because we can observe only the measures (not the processes themselves), the evidence used to support separate modifiability is the finding of factors F and G for which dd(MA, MB; F, G) obtains, which also indicates (Sternberg, 2001, p. 149) that A and B are functionally distinct. Such evidence supports a theory with three components: (i) The task is accomplished by a complex process that contains two functionally distinct and separately modifiable parts (i.e., modules) A and B. (In SDT they would be the sensory and decision processes.) (ii) Module A is influenced by F