So far we have limited our discussion to (a) how the field of human memory has evolved because of the integration of multiple behavioral measurements, and (b) how the models that serve that function should be evaluated. Here I briefly confront the question of how to integrate behavioral measures with the types of data provided by research in cognitive neuroscience. Let me warn the excitable reader that I offer no good answers to this question. I am not alone in that regard, but I do offer a few suggestions that might help guide future advances on this front.
In particular, advances in medical imaging have brought to the forefront questions about the integration of physiological data into cognitive theorizing. The issues themselves are quite old, in fact; researchers have used the electroencephalogram (EEG) and galvanic skin response (GSR) to address cognitive-like issues for about a century (Berger, 1929; Fere, 1888; Tarchanoff, 1890). The advances alluded to refer primarily to measures that allow greater spatial precision in viewing the morphological structure of the brain, as well as the transient electrical, chemical, and hemodynamic events that occur during brain function. These techniques— both the new and the old—allow the construction of spatial and temporal maps of activity during the performance of different cognitive tasks. One tack to integrating cognition and neuroscience is a primarily exploratory approach. Using cognitive theory to compare tasks that differ in a single putative cognitive component, either parametrically or otherwise, allows the inspired cognitive neuroscientist to compare maps of brain activity and postulate a brain region or regions that are related to the manipulated cognitive component.
Hidden within this approach is the notion that the brain is likely to have divvied up cognitive functions in the same manner as experimental psychologists have. I fear that we have not had that kind of insight, but the approach is valuable nonetheless, for it allows for the evolution of cognitive neuroscience into a second, more mature phase of theoretical development. Using a hypothesis-testing approach, specific neural signatures known to accompany cognitive events are sought in paradigms in which there is theoretical debate about the contribution of those cognitive components to the behavior in question. For example, changes in blood flow are apparent in areas in Broca's area 17 during mental imagery (Le Bihan et al., 1993). In addition, "small" mental images elicit greater activation in posterior visual cortex, corresponding to foveal input, whereas
"large" mental images elicited greater activation in anterior visual cortex, an area that represents input from the periphery of the eye (Kosslyn et al., 1993). In each of these cases, the researchers used established knowledge about brain function—in this case, that regions of occipital cortex code visual input from the eye—to address the question of whether visual imagery is spatial or prepositional in format (Finke, 1980). The evidence revealed that imagery engaged visual areas of the brain and is thus likely spatial in representation. Other recent research has used this approach to address whether people learned an association between visual and auditory stimuli by examining blood flow in visual cortex following presentation of an auditory stimulus that had previously been paired with a visual stimulus (Mcintosh, Cabeza, & Lobaugh, 1998). Many other examples exist in the domains of perception, attention, memory, and language.
As results from exploratory cognitive neuroscience increase the number (and validity) of known relationships between neural signatures and cognitive components, the more scientists interested in cognitive phenomena will be able to exploit that knowledge for the purpose of furthering cognitive theory. The back-and-forth between exploratory and hypothesis-testing approaches illustrates one way by which to integrate measures from the two domains. But it is worth noting that the distinction between brain-based and behavioral measures is at least partly artificial. If we measure a button press or a verbal output from a subject, we consider that measure behavioral. Yet at multiple physiological levels, events occur during that press or vocalization that are unique to that output. Muscular events in the arm or larynx, as well as neuronal events in motor cortex, control those very actions that we measure behaviorally. Other neural events combine to derive that pattern of efferent control given the input from sensory organs. No matter what the task, a continuum of events guides the physical input (in the form of light or sound waves, for example) into physical representations in the brain into physical output (in the form of muscular contractions). Whether we measure those behavioral endpoints or the physical events that precede and determine them— inside or outside the brain—the logic for the combination of multiple measurements remains the same.
The endpoints of this continuum will always be critical measures, however, no matter how precise our measurements of the intervening processes become. Just as it would be impossible to draw any meaningful conclusions about psychology without knowing anything about the physical stimulation to the subject, it is also quite difficult to do so without actually examining behavior. Many behavioral measurements carry with them an inherent dimension of performance quality that other intervening measures do not. If a manipulation enhances the speed or accuracy with which subjects perform a task, we are licensed to attribute to that manipulation an interpretation of quality—that it improves learning, or problem-solving speed, or attentional focus, for example. There is nothing inherently "better" from a cognitive perspective about more blood flow to a particular brain region, greater skin conductance, or higher levels of chemical uptake, even though such effects may well accompany behavioral effects that do allow such an interpretation.
On the other hand, experimental tasks often suffer from a failure to approximate real-world circumstances that elucidate the contribution of the cognitive capacity under study. In part, this may be because of the contrived nature of the chosen behavioral measure. Researchers interested in language comprehension, for example, often measure the rapidity with which subjects can identify probe stimuli as words or nonwords as an index of the degree to which previously read sentences or heard utterances (related to those words) have been comprehended. Clearly, this artificial task makes the laboratory study of language comprehension quite unlike naturalistic language comprehension. Cognitive neuroscience methods provide an opportunity to reduce the reliance on such tasks by allowing measurements in the absence of an overt behavioral task. For any given experimental situation, the choice between behavioral and brain-based measures involves trade-offs, and as the astute reader might suspect, the combination of multiple types of measures across and within single studies often proves the most fruitful approach.
Was this article helpful?