Randall C. O'Reilly - US grants

This project aims to futher our understanding of the neural representations that underlie working memory. Working memory refers to the active maintenance of information in the service of complex cognitive tasks such as problem solving and planning. The team of investigator with study how the unique demands placed on the working memory system shape its representations during learning and development, and how these representations affect the use of working memory by the cognitive system as a whole. Their primary source of insight into this process will come from a computational analysis, which will be used to integrate and explore relevant findings from neurobiology and developmental and adult cognition. The primary hypothesis is that to maintain information in an active state over delays and in the face of interference, working memory representations should be discrete. A discrete representation admits to only a finite set of possible states, rather than representing continuous states. For example, the integers from 1 to 100 form a discrete set, in contrast to the real numbers in this range. Discreteness imparts a measure of robustness to the representation because small amounts of "noise" can be overcome by interpreting an observed state as the nearest discrete stat. From the central property of discreteness, a number of other properties follow. For example, discrete representations should be more categorical, more easily verbalize, better for perceiving or performing a sequence of steps, and more accessible to awareness. All these properties have component of the proposed research is to explore the idea that they all follow from the more basic property of dicreteness. The initial goal of the project will be establish through experimental studies the validity of the hypothesis that working memory representations are indeed discrete. This will be done by exploring a key behavioral consequence of this hypothesis: working memory representations should be more categorical than other representations. This predicition will be investigated with a set of existing empirical paradigms that have elucidated variables that affect the working memory system, including developmental age, delay, dual-task demands, and brain damage. Working memory plays a central role in most accounts of complex human behavior, because working memory is required in any task that involves multiple stos ir a temporally extended focus of attention. This kinds of tasks are important and pervasive throughout society, including: economic, political, and military planning; air-traffic control; and scientific research, to name just a few. It is essential to understand the nature of the representations in the working memory system and to understand how people learn to use working memory in the service of complex behavior. This research will advance our knowledge in this important area, and may provide insight into techniques for rehabilitation of working memory following brain injury and techniques for assisting the development of working memory in children.

[unreadable] DESCRIPTION (provided by applicant): [unreadable] The overarching goal of the proposed research is to understand how the brain performs cognitive control. By cognitive control, we mean the ability of the cognitive system to flexibly control its own behavior in response to task demands or other contingencies, favoring the processing of task-relevant information over other sources of competing information, and mediating task-relevant behavior over habitual or otherwise prepotent responses. There is virtually universal agreement that the prefrontal cortex (PFC) plays a critical role in cognitive control. However, exactly what it does and how it does it, in terms of concrete neural mechanisms, remain considerably more controversial questions. Explicit computational models that incorporate biological mechanisms can provide a powerful means of testing theoretical ideas about the biological basis of cognitive control. However, existing models have each only simulated one or a few phenomena, leaving open questions about their general applicability across the entire domain of cognitive control. We propose to address this limitation by applying a single computational model that includes critical features of the underlying neurobiology (including the basal ganglia (BG) and its interactions with the PFC) to a wide range of benchmark behavioral and neural phenomena. This model includes powerful learning mechanisms that should produce intelligent, controlled behavior without relying on the kind of homunculus that often lies behind theories of cognitive control. Specific Aim 1: Modeling Behavioral and Neural Data. We test the sufficiency of our framework by evaluating whether a single instantiation of the model, trained on a single large corpus of tasks, can simulate a wide range of benchmark data in cognitive control, and we use this model to make a number of testable predictions. Specific Aim 2: Nature and Learning of PFC/BG Representations. We address the fundamental question: how can people quickly adapt to performing novel cognitive tasks, when it takes monkeys months of highly-focused training to learn a single new task? We hypothesize that people develop an extensive repertoire of basic cognitive operations throughout the long developmental period into adulthood, and can rapidly and flexibly combine them to solve novel tasks. Demonstrating this principle in an explicit computational model will have important implications for understanding human intelligence, education, and development. [unreadable] [unreadable] [unreadable] [unreadable] [unreadable] [unreadable]

The computational modeling core is designed to support the development of biologically plausible computational models across the range of projects proposed in the center. Various existing models have been developed for the experimental tasks employed across all of the proposed projects. This includes models of cognitive control (Project 1), learning (Project 2), emotion (Project 3), dompainergic infulences on executive funciton (Project 4) and child development (Project 5). Because these models are biologically inspired they consider both neurobiological and psychological constraints on processing. The development of computational models provides an important way of synthesizing experimental results in a mechanistic fashion, and generating predictions for further experimental studies. These models will be used in parallel with empirical investigations so that they can inform each other. Because of these characteristics, the computational core explicitly addresses a center goal of linking levels of analysis (computational, psychological, and neurobiological) in our understanding of executive function. The personnel associated with the computational core will provide both the faculty and post-doctoral expertise with regards to the creation, implementation and evaulation of comptuational models for all projects within the center. In addition, it will provide the computational infrastructure for those investigations as well as training materials and training opportunity for center members. Computational models of executive function can provide insights into psychiatric disorders in numerous ways. For example, models can be lesioned or altered in ways that are impossible to do experimentally for either ethical or practical reasons. As such, they can provide insights into the mechanisms that support disordered thoughts or emotional processes. Furthermore, they may provide insights into mechanisms or interventions that might be effective in overcoming such disorders. This is especially true of the models proposed for this center, which are explicitly linked to the neuroanatomy and functioning of brain systems as well as the psychological processes that support executive function.

The overarching goal of the proposed research is to understand the specific roles of different brain areas in the learning of executive function, by empirically testing and fur ther developing a biologically-based computational model of the prefrontal cor tex and associated subcor tical systems (including the basal ganglia and midbrain dopaminergic nuclei). We focus on two specific issues: (a) the learning of abstract, rule-like representations in prefrontal cor tical areas, which suppor t flexible behavior by enabling better generalization to novel circumstances;and (b) the mechanisms of feedback-driven learning, which shape the adaptive modulation of prefrontal executive function representations by the basal ganglia, according to our model. Several predictions from this computational modeling framework have already been successfully tested in diverse populations, including Parkinson's patients and people with ADHD, both of which are thought to involve disorders of the dopaminergic system as it affects the basal ganglia and prefrontal cor tex. Thus, this model has impor tant implications for understanding the neural basis of executive function, both in neurologically intact and disordered populations. We propose to test the following hypotheses: [unreadable] Specific Aim 2.1: Factors Affecting Learning of Representations in Prefrontal Cortex: First, we test in both young adults and children a set of predictions from our computational model. For example, blocked training should facilitate the development of abstract, rule-like representations, which in turn suppor t better generalization to novel task contexts. Second, because the degree of abstraction learning depends on the duration of active maintenance in our model, different regions of PFC may be organized according to relative degree of abstraction, and corresponding maintenance duration. We explore this idea in the model. [unreadable] Specific Aim 2.2: Factors Affecting Feedback Learning. We provide a more direct test of dopaminergic mediation of event-related potential signals responsive to feedback information (ERN) by administering dopamine D2 receptor agonists/antagonists, which should have dissociable effects on these signals according to our model. We also test whether mood induction can shift the balance of feedback responsiveness, as measured by ERP's. Finally, we attempt to disentangle multiple factors influencing learning of executive function tasks by using coordinated executive function and negative feedback learning studies in children.