Understanding the neural mechanisms of cognitive control is essential to understanding the relation of human brain function and behavior. Cognitive control refers to processes supporting goal-directed behavior and cognition. A rich collection of executive functions are under the purview of cognitive control, including (but not limited to): decision making, problem solving, endogenous attention, mood regulation, motivation, working memory, and many forms of learning. In fact, many modern neuroscientists consider the term cognitive control to be synonymous with executive functioning. Importantly, cognitive control is also thought to be negatively impacted in many mental illnesses, which altogether makes it an imperative topic for a varied cross-section of researchers.
Prior research has identified large-scale neural systems collectively termed cognitive control networks. Of these, two prominent control networks include the frontoparietal network (FPN) and the cingulo-opercular network (CON). Brain regions of cognitive control networks are thought to enact control by large-scale network interactions that are flexible to the current goal or task context (Waskom et al., 2014). This framework is built upon the biased competition theory (Desimone et al., 1990, 1995), then later the guided activation theory (Miller and Cohen, 2001), and then more recently, the flexible hub theory (Cole et al., 2013). The flexible hub theory was developed by our lab and posits that well-connected hub regions exhibiting global connectivity changes (across tasks) are critical to adaptive task control.
Prior work has delineated the cognitive control related properties of FPN regions well. However, control properties of CON regions are less established. In our recent paper published in The Journal of Neuroscience, we examined the neural mechanisms of both the FPN and CON by quantifying each of their large-scale network interactions during a high-control-demand cognitive paradigm (as in Cole et al., 2010, 2013; Ito et al., 2017). We used fMRI data acquired during both resting and task states for 100 healthy adults. Estimates of functional connectivity were used to probe network interactions. This was based on the idea that these estimates capture interregional information flow in a functionally-relevant manner. To characterize the dynamic, functional properties of control networks, we applied network science tools (including a novel metric termed “network partition deviation”), decoding analyses, and developed a multidimensional functional cartography (or “mapping”; panel C in above Figure).
Expanding upon the flexible hub theory (and related theories mentioned above), we found FPN regions exhibited a “flexible coordinator” mechanism (panel A in above Figure). The “flexible” aspect of this characterization was observed in between-network connectivity patterns demonstrating high variability across task states. This suggests that FPN regions exhibit network interactions that are continuously adaptive to task goals. The “coordinator” aspect of this characterization was observed in FPN’s within-network configuration being maintained across task states. This internal stability could facilitate the binding of stimulus information with task-rule information during cognitive computations.
We found that the CON exhibited mechanisms of cognitive control that were complementary to the FPN: namely, acting as a “flexible switcher” (panel B in above Figure). Here, the “flexible” portion was demonstrated by CON regions exhibiting transient, task-relevant connectivity changes. These connectivity changes were observed as disbandments from their resting-state configuration, constituting the “switcher” portion of the CON’s characterization. This suggests that CON is capable of augmenting the processing of goal-relevant regions (or networks) by lending them processing resources. This proposed role for CON in controlled processing expands upon findings that suggest it specifies overall task-set modes (Dosenbach et al., 2007; Power and Petersen, 2013; Sadaghiani and D’Esposito, 2015).
Taken together, this account highlights the FPN and CON enacting cognitive control in a complementary manner. But what may be the benefit of the control networks exhibiting these specific properties? An idea central to the theories mentioned above is competition. Neural representations compete for resources: in any given moment (i.e., during a task), representations of stimuli, actions, and/or thoughts are all vying for the limited processing resources of the brain. It is thought that top-down signals from control network regions can shift this competition in favor of neural representations that are relevant to the task or goal at hand. This is the essence of cognitive control. To this end, FPN connectivity patterns appear important for flexibly coordinating top-down biases across task states. However, interference (from goal-irrelevant representations) is still possible. The flexible switching of CON connections can step in to lend resources to the goal-relevant systems and help them win competitions. Thus, the FPN and CON may be acting in-concert toward effective controlled processing. Moreover, this account allows for the possibility that the control networks are managing a tradeoff in computational demand.
Looking forward, we hope that this work is built upon by incorporating other task paradigms, clinical cohorts, and other applications targeting cognitive control. Methodological considerations can be taken into account, such as improving the estimation and interpretation of functional connectivity. The cartographic approach could be applied to other functions of interest (and accordingly, task paradigms with other outcomes), such as visual processing or attention. Moreover, recent findings across a variety of research questions point to a broad distinction between sensorimotor versus cognitive network mechanisms (e.g., Bassett et al., 2013). Our pipeline could be implemented as a tool for exploring these distinctions further.
Check out the full report here.