Although rarely attributed to Campbell and Fiske's pioneering descriptions of the MTMM approach, the core ideas inherent in the MTMM concept can be seen in the convergence approaches that are used in many neuroimaging studies. Convergence approaches have become increasingly popular in functional neuroimaging in response to one of the core problems in the field. Specifically, a multitude of stimulation tasks or procedures exist that can be used to engage a particular psychological construct or brain region, with each variation possessing slightly different properties. In other words, each stimulation paradigm comes with its own method variance, and not surprisingly, substantial inconsistencies emerge in the literature. To deal with this, many neuroimaging researchers have begun to use procedures to look at the convergence of responses across procedures. In its simplest form, this is accomplished with a simple logistic analysis in which each effect is transformed voxel by voxel into a binary representation of whether the voxel was activated above a certain threshold. These binary representations are then summed or multiplied across contrasts to produce a spatial map of areas activated in more than one condition.
A convergence approach also helps deal with the problem of pure insertion. As noted earlier, in a simple subtraction design it is impossible to determine if a change in brain activity relates to the inserted cognitive component or to changes in other components that arise as a consequence of the inserted component. However, by using multiple stimulation-control contrasts it becomes possible to more clearly parse the component in question from its effects on other task components. Imagine, for instance, a judgment task in a given sensory modality that is contrasted with a passive task in which the stimulus is presented but no judgment is made. It is difficult to know if changes in brain activity are related to the judgment itself or if the act of making the judgment caused modality-
specific changes in sensory processing because of increased attention to the stimulus rather than the act of making the judgment. Now, if we run similar experiments in other sensory modalities, we can analyze them to determine common vs. modality-specific activations. The areas that are active in all tasks can be considered modality independent processes and cannot be attributed to factors such as increased attention to a specific stimulus category. Thus, even if the assumption of pure insertion fails in a given task, it becomes possible to separate activations related to the component of interest (the judgment) from changes in other processes (modality-specific attention) that arise as a consequence of the task insertion. Of course, a delineation of the common activations may fail to detect sensory-specific processes that are directly related to the component in question. Nevertheless, the remaining modality insensitive, common regions of activation will be more clearly attributable to the component of interest.
Price and Friston (1997) referred to this approach of examining the commonalities between activations arising in different contrasts as "cognitive conjunction analysis." It is worth noting that when applied in neuroimaging, particularly among researchers using the popular SPM program (Department of Cognitive Neurology, London, UK), the conjunction refers to the presence of a main effect in the absence of differences in simple effects at a given voxel. The analysis is performed by taking the sum of all activations [(stimulationj -control^ + (stimulation - control2) . . . ] and eliminating voxels where there exist significant differences among the individual contrasts [(stimulationj - control j ) - (stimulation2 - control2) . . . ].
Was this article helpful?