The examples given in earlier sections of this chapter have described how neuroimaging data can be used within a formal application of the MTMM approach. However, the general approach toward looking for convergence and discrepancies across methods and traits can be applied as an evaluative strategy, even in situations where it is not possible to use the same methods in the same subjects. In such a situation one cannot produce a covariance matrix across methods, but one can nevertheless use an emphasis on convergence and divergence for evaluating hypotheses.
Sarter, Cacioppo, Berntson, and colleagues (Cacioppo et al., 2003; Sarter, Berntson, & Cacioppo, 1996) have articulated the importance of understanding the type of information that functional neuroimaging studies provide relative to other types of neuroscientific data. Specifically, most neuroimaging studies provide information on the probability that a given brain area activates as a function of a cognitive process (i.e., the experimenter performs a task aimed at inducing a specific cognitive process and determines whether the task leads to activity in a specific brain region). In contrast, such studies do not typically provide information on the probability that a given cognitive process arises as a function of activation of a specific brain region (although researchers frequently make the erroneous interpretation that the results provide this information). Such a conclusion would only be true if there is a one-to-one correspondence between the brain region's activity and the cognitive process, and we rarely possess evidence for such a one-to-one correspondence. Sarter et al. argue that to fully understand the bidirectional relationship between brain activity and cognitive processes, one needs to integrate other types of paradigms (such as lesion or electrical stimulation data) that allow direct manipulation of brain regions and thus provide information on the probability of a cognitive process given activity (or lack of activity) within a specific brain region.
The preceding analysis parallels a classic distinction in the neurobehavioral field between brain areas that are activated in a task and brain areas that are necessary for performance of the task. Taken alone, neuroimaging typically only addresses the question of what is activated and fails to address whether that activation is necessary. In contrast, neuropsychological studies of patients address what is necessary, but not what is activated. Thus, to answer the question of what is both engaged and necessary in a task, one needs to use both methods. The greatest clarity arises when both methods converge to show that an area is both necessary for and engaged by a task involving a given psychological process, but is not necessary or engaged by tasks that do not require that psychological process.
Considered in this light, it also becomes necessary to expand the MTMM approach to include data from other species. Specifically, most techniques that allow us to look at the causative effects of manipulating brain regions can only ethically be carried out in nonhuman populations. These animal studies typically proceed on the assumption that (a) there are "homologous" brain regions across species, (b) these regions perform the same tasks, and (c) the regions perform the tasks in the same way. However, despite many features that are conserved across species, even a cursory study of neuroanatomy reveals substantial interspecies differences. Given these potential cross-species differences, we need evidence of convergence and divergence across methods used in different species. It thus may prove useful to take a multitrait-multi-method-multispecies approach to evaluating brain-behavior relationships. In summary, the core logic articulated by Campbell and Fiske provides an extremely useful overall strategy for placing neuroimaging research within the larger field of psychology and neuroscience, even in situations where formal MTMM analyses are not feasible.
Was this article helpful?