Toward characterizing the neural correlates of component processes of cognition

Introduction Human cognition is a challenging area of inquiry. Ironically, the same intricacies of the mind that allow us to examine it also frustrate our progress; getting our thinking devices to understand their own mechanisms of operation sometimes feels like chasing one’s shadow. The mind’s flexibility requires many concepts to describe its many functions: For example, in the domain of memory, we use different terms for “remembering” how to ride a bicycle and “remembering” the events of the day the training wheels came off (procedural vs. declarative memory [Cohen & Squire, 1980]), or for remembering the phone number we just looked up (working memory [Baddeley, 1992; Baddeley & Hitch, 1974]) and our phone number from childhood (long-term memory). We categorize memory by its informational content (e.g., episodic vs. semantic memory [Tulving, 1983]), by the types of processes we think are engaged (e.g., familiarity vs. recollection [Atkinson & Juola, 1974; Jacoby & Dallas, 1981; Mandler, 1980; Tulving, 1985]; shallow vs. deep encoding [Craik & Lockhart, 1972]; or perceptual vs. reflective processing [Johnson & Hirst, 1991]), or by the brain regions that are involved (e.g., the medial temporal lobe vs. the basal ganglia [Poldrack & Packard, 2003; Poldrack & Rodriguez, 2004]). Such broad categorizations of memory are not necessarily mutually exclusive; for instance, whether one is able to recollect a stimulus or merely recognize it as familiar may have something (but not everything) to do with whether it was initially encoded deeply or shallowly (see Yonelinas, 2002). In turn, such seemingly different subjective experiences as a feeling of familiarity or of more embellished recollection may involve partially overlapping brain structures. A closely related issue specific to process-oriented approaches is that key concepts may be complex and involve multiple subprocesses. For example, even simple working memory tasks require encoding, maintenance, updating, and selection processes. One such task sometimes used to operationalize the process of “working memory” is the N-back task, which minimally requires one not only to perceive the features of a stimulus, construct an internal representation of it, and add that representation to an existing queue of N previously presented stimuli, but also to compare the first and last representations to decide if they are the same, recall the appropriate action to take, make a button press or other overt response, and remove the oldest representation from the queue. Furthermore, some of these sub-processes could easily be shared with a number of other cognitive tasks which may or may not be considered “working memory” tasks per se. Also, as the complexity of a task grows, there is the increasing likelihood that different people will use different strategies (i.e., differing combinations or sequences of component processes) to perform the task (Johnson et al., 2005). Thus, for multiple reasons, the greater the complexity of a task or a proposed cognitive process, the more difficult it may be to characterize. At the same time, general concepts used to characterize mental activity during complex tasks, for example, “working memory,” “executive function,” and “cognitive control,” likely share some or many of the same underlying cognitive and neural components. While general constructs such as working memory, executive function, and cognitive control focus attention on important domains and help organize findings, researchers also recognize the importance of unpacking these complex ideas into constituent elements (e.g., structures or processes): for example, the work of Baddeley and colleagues in characterizing the phonological loop, visuo-spatial sketchpad, and central executive subcomponents of working memory (Baddeley, 1984, 1996; Baddeley, Lewis, & Vallar, 1984; Baddeley & Lieberman, 1980; Salame & Baddeley, 1982) or the work of Cohen, Carter and others in dissociating elements of cognitive control, particularly the role of the anterior cingulate cortex (ACC) in detecting conflict (Botvinick, Nystrom, Fissell, Carter, & Cohen, 1999; Carter et al., 1998; Kerns et al., 2004; MacDonald, Cohen, Stenger, & Carter, 2000). In our lab, we have found it useful to adopt a component-process approach, using a model that defines a set of basic “building blocks” of cognition that, when combined, could form the many more complex operations of which the mind is capable. Here, we first provide an overview of this model and then describe studies using neuroimaging to test and more completely characterize its component processes.