Assessment of psychological constructs has traditionally focused on behaviors that are either directly observable by a researcher or can be reported by the examinee. Neuroimaging can supplement these methods of assessment by providing information at a neural level. Although, one might be tempted to view this at a causal level (i.e., the brain activity causes the behavior, or the behavior causes the brain activity), it need not be viewed as such. Rather, neuroimaging data can be viewed as just another indicator or correlate of a psychological process or trait. However, neuroimaging data are qualitatively different from most other types of measures in psychological research in that the brain's response can be measured without requiring the subject to make a behavioral response or use introspection. Thus response may be measured uncontaminated by requirements to self-monitor or control a motor act (both of which may add method variance in psychological studies).
Imagine, for instance, the assessment of a personality trait. A number of investigators have found neural correlates of personality either in terms of resting data or the degree of activation during stimulation (see Canli, Sivers, Whitfield, Gotlib, & Gabrieli, 2002; Gusnard et al., 2003; Zald et al., 2004). By combining neuroimaging data with other self-report, observer rating, or experimental performance measures, we may increase accuracy in assessment. In such a paradigm, levels of regional brain activity would be predicted to converge with self-report and objective ratings of the trait of interest, but not other traits.
The MTMM approach can similarly be applied to the assessment of a psychological process. Imagine you are testing subliminal processing of visual stimuli using a tachistoscopic method. The presence of subliminal processing is traditionally tested by having subjects "guess" about stimulus features in the absence of an explicit awareness of having seen the stimulus. Performance significantly above chance provides evidence for subliminal processing. Now, if we simultaneously scan subjects with fMRI and see BOLD responses that are temporally linked to the presentation of the stimuli, we could use the fMRI data as a second source of evidence that subliminal processing occurred. Because neither measure is likely to be 100% sensitive or selective, the combination of the two types of data may dramatically increase predictive power.
A critical problem must be resolved before including functional imaging data in a MTMM matrix. Specifically, the precise relationship between activations and behavioral performance cannot always be predicted in advance. In some cases, higher activations may reflect greater performance or ability level. However, in some cases, subjects with lower ability may have to activate a region more to perform a task at an equivalent level to a more skilled person. This issue has been particularly salient in the psychiatric imaging literature, where researchers attempt to draw conclusions about the relationship between functional activations and the neural substrates of psychiatric conditions. This is essentially an empirical question. Once we understand the nature of performance-activation relationships, it becomes reasonable to consider the neuroimaging data in a MTMM matrix.
Unfortunately, because of the expense of collecting neuroimaging data, it seems unlikely that neuroimaging data will be routinely used as a component in MTMM matrices. However, its utility may be appraised in terms of a cost-benefit analysis. In situations where the neuroimaging data has significantly greater sensitivity or selectivity than other forms of data, then the benefit of its inclusion may outweigh the costs. Plus, with the advent of data sharing through the fMRI data center (http://www.fmridc.org), which is a public repository of peer-reviewed published fMRI data, researchers can potentially pool data from dozens of studies that fit key cells in MTMM matrices.
Was this article helpful?