The focus of this project is on the human capacity for cognitive control. Cognitive control can be defined as the mechanisms responsible for the pursuit of intentional (e.g., goal-directed) behavior in the face of competing alternatives. This capacity lies at the heart of our ability to pursue long-term interests, both individual and social. The flexibility and sophistication with which we can deploy cognitive control is one of the primary characteristics that distinguishes us from other species: It is fundamental to a variety of higher level mental faculties that are characteristically human, such as planning, problem solving and social interaction. It also relies on brain structures, most notably the prefrontal cortex, that are the most expanded relative to other species. The capacity for cognitive control is engaged in every walk of life, from the regulation of emotions and impulsive behaviors required in daily tasks (such as parenting, dieting and savings), to the decision making processes underlying ethical behavior (e.g., moral reasoning, forgiveness, and promise-making). Given the centrality of cognitive control to social behavior, understanding how it operates — and fails to operate — is also of central importance to all spheres of policy making, from educational to economic and diplomatic.

Our approach to studying cognitive control draws on two scientific traditions that have converged over the past two decades: psychological and neuroscientific. On the one hand, cognitive control is characteristically psychological: Its properties and effects are most coherently described at the behavioral level. On the other hand, great progress has been made in identifying neural mechanisms that are critical to its function. Our work integrates behavioral with neuroscientific experimentation and model building, working toward a deeper understanding of how cognitive control operates, and how it arises from the functioning of the brain. Such an understanding may point the way to new interventions for strengthening cognitive control in healthy individuals, and restoring it where it is impaired, as in neuropsychiatric disorders.

While a considerable body of previous work has addressed the neural mechanisms governing cognitive control, this work has been limited in three ways: 1) It has focused on simple forms of control that are convenient to study in the laboratory, but do not approximate the complexity of real-world human behavior. 2) Theoretical efforts have focused on descriptive rather than normative models. 3) Experimental work has been restricted by the use of methods originally designed to address functions other than cognitive control. Our project aims to surmount these limitations in three corresponding ways:

1. Higher-level forms of cognitive control. Existing work has focused on how control guides attention or action once goals or task rules have been specified. It has addressed the question: How does the brain decide what information is relevant and how to respond, given a pre-specified goal or task? However, everyday life involves decisions not just about attention and actions, but also about what goals to pursue and what tasks need to be accomplished to pursue them. These higher-level decisions do not necessarily lead to immediate responses, but rather modify internal states (goals and task sets) that can affect future behavior. A focus on this higher-level decision problem distinguishes this project from other work.

2. Normative approach to theory. Most theories of how cognitive control operates are descriptive: They seek to describe underlying mechanisms, either by drawing upon observations at the neural level, or constructing psychological models that explain behavioral observations. This contrasts with a normative approach to theory, which starts by formally specifying the problem faced by a given system, and then asking how it can be solved optimally. A major focus of our work is to apply a normative approach to the questions that are central to higher-level forms of cognitive control: How do we decide what goals are best to pursue now, and when it is best to defer them until later and explore other potentially more valuable tasks or goals now? How can memory be used optimally to store and retrieve task rules or goals when they need to be pursued at a later time? How are goals and task rules learned and organized in the brain to optimize the ability to make these decisions? And, finally, where cognitive control is found to fall short of optimal in human performance, does this reflect fixed constraints or opportunities for improvement?

3. Development of methods customized to the study of cognitive control. One of the most important recent advances in human brain imaging has been the development of methods for multivariate pattern analysis (MVPA). These methods use sophisticated machine learning classifier algorithms, applied to brain imaging data, to decode what people are thinking at a particular point in time. MVPA has been used successfully to study perception, memory, and language. To date, however, MVPA has not been applied aggressively to the study of cognitive control. In this project, we aim to develop the use of MVPA in three new ways: 1) model-based MVPA methods, that will exploit the theoretical models we propose to develop to improve the ability of MVPA to identify and track control representations in the brain; 2) the use of MVPA to analyze patterns of correlations among different brains areas (“functional connectivity analysis”), in order to test hypotheses about the dynamics of cognitive control, and the influence that control mechanisms have on parts of the brain responsible for task execution; and 3) real-time MVPA, to provide individuals being scanned with real-time feedback about the control-state of their brain, to test whether this can be used to improve their capacity for cognitive control.

Our project is composed of research in five Focus Areas, each of which is directed by two of the principal investigators.

Focus Areas 1 through 4 each address specific forms of higher-level control.  They draw heavily on formal theoretical modeling, including optimality analysis and neural network simulation, as well as empirical studies. The four specific topics include:

(1) Self-control over impulsive behavior – Botvinick & Cohen

(2) The execution of complex, hierarchically-organized behavior involving goals and subgoals – Botvinick & Niv

(3) Prospective memory and planning – Norman & Cohen

(4) Balancing goal-directed behavior (exploitation) with discovery and learning (exploration) – Cohen & Niv

Focus Area 5 involves developing innovative new MVPA-based brain imaging methods for studying cognitive control – Cohen, Norman & Turk-Browne