David Badre - US grants

DESCRIPTION (provided by applicant): PFC mediated cognitive control processes permit flexible behavior by biasing selection of goal-relevant responses. The research proposed here seeks to test two related hypotheses regarding the organization of RFC control processes. A functional magnetic resonance imaging (fMRI) study will identify regions of RFC processing different levels of representation en route to selecting a response. These levels are hypothesized to be individuated on the basis of the abstractness of the representation guiding selection, ranging from concrete features (e.g., red) in premotor cortex, to more abstract dimensions (e.g., color) in posterior lateral regions, to a general goal or context (e.g., color naming task) in frontal pole. Subsequently, a study of neurological patients with focal damage in RFC and a study of focal disruption by transcranial magnetic stimulation (TMS) will test the functional necessity and independence of these RFC regions. Furthermore, these data, along with functional connectivity analysis of the fMRI data, can test the hypothesis that this representational hierarchy is reflected in the relationship among RFC regions, instantiated in a flow of processing from superordinate to subordinate regions. Basic knowledge gained regarding the organization of RFC can provide important insights into disorders impacting the frontal lobes and toward the development of directed clinical assessments that index executive function.

DESCRIPTION (provided by applicant): The goal of this research program is to advance our understanding of the functional organization of frontal cortex and the ways in which interactions among frontal regions permit internal control of thought and action, an ability termed cognitive control. The proposed experiments test the hypothesis that the rostro-caudal axis of the frontal lobes comprises a processing hierarchy whereby more anterior regions support cognitive control involving progressively more abstract goals, plans, and action representations. First, a set of fMRI experiments, using factorial logic, seek to specify the dimensions of abstraction that define regional differences along the rostro-caudal axis of the frontal lobes. Several experiments have demonstrated a functional gradient along the rostro-caudal axis as cognitive control involves more abstract representations. However, these experiments have differed in their definition of abstraction, including domain generality, temporal abstraction, and rule complexity. Though these dimensions of abstraction are not mutually exclusive, they may also be independently varied under controlled circumstances. Thus, the first set of fMRI experiments take advantage of this independence to provide a more explicit definition of abstraction as it relates to frontal lobe function. Second, a hierarchical system requires that superordinate processors can modulate subordinate processors but not vice versa. Translated to a frontal hierarchy, this asymmetry predicts that control processing by a given region of the frontal cortex should directly influence regions posterior to but not anterior to its locus. Thus, a line fMRI experiments will use effective connectivity analysis to test whether (a) cognitive control processing in anterior frontal regions can modulate processing in posterior frontal regions, (b) whether cognitive control processing in posterior frontal regions can affect concurrent control processing by anterior frontal regions. Third, a set of experiments will investigate frontal lobe function during learning and adaptation of hierarchically structured rule sets. In general, hierarchies are useful structures for supporting learning of novel tasks because they can represent the same information at multiple levels of abstraction and so can provide access points for the application of prior knowledge, such as through analogy. Thus, a set of experiments test whether the putative frontal lobe hierarchy reflects such dynamics during learning of novel rule sets, the transfer of learning to new tasks, and the application of transformational rules to existing rule sets. Hence, across its aims, this program of research triangulates the hypothesis of a hierarchical organization of the frontal lobe by testing three different classes of predictions that each arises from an assumption of a hierarchical system. PUBLIC HEALTH RELEVANCE: A variety of disorders impact cognitive control and frontal function, including schizophrenia, attention-deficit hyperactivity disorder (ADHD), obsessive-compulsive disorder, major depression, autism, Alzheimer's disease, and Parkinson's disease. However, a lack of basic understanding of cognitive control and frontal function remains a persistent obstacle to directed assessment and rehabilitation of frontal lobe dysfunction (Royall et al., 2002). The proposed research seeks to fill these gaps in basic understanding of frontal lobe function, and so may form a basis for development of directed assessments that refine diagnosis and improve rehabilitation.

DESCRIPTION (provided by applicant): Development is charactarized by increasing flexibility in thought and action, enabeling the generalization of rules from item-specific instances to abstract classes of rules that support flexible behavior and can be generalized to novel situations. We capitalized on an established paradigm that has been effective in charactarizing this skill in adults in behavioral and fMRI experiments, and also in patients with frontal damage. This work has shown that increasing complexity in rule-guided action recruites the frontal cortex along a rostro- caudal axis. We hypothesize here that the functional development of the frontal cortex, from middle childhood to adulthood, and associated behavioral performance will follow the same pattern. Specifically, we predict that mature behavior in simple rule tasks will be evident earlier in development than more complex tasks and that this will be associated with functional effiency at the level of the frontal cortex in caudal prior to rostral regions. Using th extant literature, we predict that this functional development is a reflection of increases in the stability of cortical coding of rule-relevant representations. We test this hypothesis using multi-voxel pattern classification algorithms designed to decode the stability of rule-relevant representations across development. Moreover, we test the prediction that representational stability is dependent on experience. Portions of the frontal cortex have an especially protracted developmental course and have been shown to be susceptible to both positive and negative environmental experiences. Rather than use age as a proxy for experience, we turn to socioeconomic status (SES), which has been shown to be a predictor of cognitive control development. Availability of items, interactions, and experiences in high relative to low SES families may prove relevant for experience-based changes in representational stability at the level of the frontal cortex. Moreover, variability among experiences is necessarily more relevant for the development of abstract rather than concrete rule representations, indicating that the impact of SES differences may be evident specific to the development of abstract rule structures. We will examine the specific contribution of SES, as a proxy for enriching experiences, to the development of rule-guided action in both behavioral and neuroimaging experiments. RELEVANCE Item-specific concreteness and perseveration are ubiquitous in developmental disorders including Autism, ADHD, and OCD. We propose to characterize development of rule use at varying levels of abstraction, allowing for novel specificity in understanding the deveopmental course of these skills. Uncovering a specific relationship with experience-based factors (SES) in this development has the added potential of defining variables that confer risk specific to cognitive flexibility and will inform social and educational policy moving forward. PUBLIC HEALTH RELEVANCE: We focus this work on the development of cognitive control and rule-guided behaviour during the transition from childhood to adolescence and adulthood. This is a period marked by extensive change in the child's environment, including changes in peer relations and increased independence and academic demand. Our aim is to expose the developmental interactions among components of rule-guided behaviour, their functional neural substrates, and the nature and stability of rule-relevant representations coded by the frontal cortex across this developmental period.

PROJECT SUMMARY/ABSTRACT The concept of ?representation? is a central pillar in the interpretive framework of neuroscience. Characterizing neural representations is thus a central concern in every domain of cognitive neuroscience. Though considerable attention has been paid to the ?content? of neural representations, as in what information in encoded, far fewer studies have considered the ?format? of neural representations or how this information is organized. Yet, this latter property is of paramount importance. The format in which a population of neurons represents information constrains how accessible that information is to downstream circuits, and so necessarily shapes circuit and network level computation, as well as behavior. However, our views about the format of neural representations in humans are largely informed by single neuron electrophysiology carried out in other species. A critical property of a neural representation?s format is its dimensionality, which controls a trade-off between the flexibility of a representation and its efficiency. The dimensionality of neural task representations in human prefrontal cortex is believed to be critical to cognitive control, or our ability to flexibly guide our cognition in a goal-directed manner. Current methods for estimating the dimensionality of neural representations rely on analyzing activity patterns over multiple voxels and are hampered in regions of the brain like the prefrontal cortex which have a different underlying spatial micro-structure of neuronal populations. A different approach for estimating dimensionality based on measuring repetition suppression effects has been proposed but never implemented. In the absence of a reliable method for estimating dimensionality, important questions about human neural representations remain unaddressed, and the state of neural representations currently plays little part in our thinking about neurological or psychiatric disorders. Across its aims, the proposed research program seeks to develop and validate a new repetition suppression based non-invasive method for building whole-brain, voxel-resolution maps of representational dimensionality in the human brain. This method will be employed to test important hypotheses about human task representations. In particular, based on prior work in macaques, we hypothesize that neural populations in the prefrontal cortex serve to integrate different sources of information and represent them in a high-dimensional format that make this information useful and accessible for downstream neurons. In addition, we hypothesize that the maintenance of a high-dimensional representation is critical for behavioral performance. We test these hypotheses in an experiment that applies the repetition suppression based method for estimating dimensionality with fMRI. More broadly, this method will enable systematic, quantitative studies of representational format across several domains of cognitive neuroscience, in both healthy people at different developmental stages, and patients with brain disease or disorder.

Our training program for Interactionist Cognitive Neuroscience (ICoN) seeks to provide student-focused, interdisciplinary training in computational cognitive neuroscience that integrates data from multiple scales and levels of analysis. Transformative gains in understanding the human brain and mental health require integration across multiple levels of analysis. Recent historic advances in genetics and cellular biology are paving the way for understanding fundamentals of neural function. At the other end of the spectrum, methods for imaging and stimulating human brains non-invasively have led to revolutionary advances in discovering the macro-scale organization supporting perception, motivation, and cognition. Now, a major effort at the `systems' level between these two scales is beginning to uncover the activity, connectivity, and computations of neural circuits. The advent of this systems-level progress holds the promise of linking core circuit computations to emergent human behavior and leading to detailed, transdiagnostic models of mental illness. However, as we recently argued (Badre, Frank and Moore, 2015 Neuron), fulfilling this promise requires making direct links between circuit-level computation and the emergent function of the human system. We believe that integrating systems- and human neuroscience in this way demands a systematic approach built on two key strategies. First, formal computational models must be used to provide principled links between levels of analysis; and, second, complementary methods must be applied, and in the ideal case parallel human and non-human studies conducted in coordination. Achieving these aims requires a new generation of scientists that can take full advantage of multiple techniques and data sources, and who are deeply versed in computational theory. Traditional neuroscience training relies on an apprenticeship model that limits students to a single lab and level of inquiry. Thus, a specialized training program is required to specifically equip neuroscientists for this `Interactionist' approach. ICoN will provide this training emphasizing the two tenets: I. Computation is key to translating between levels. Students must be rigorously quantitatively trained in formal theory. A close corollary is that they must be fluent in the advanced analysis methods necessary for cross-level integration (e.g., machine learning). II. Next-generation scholars must have expertise at multiple levels. Students must be trained to use and integrate multiple methods and data sources. Further, they must have the skills (and courage) to pursue ideas to their next most logical step, to be question driven and not technique limited. Students will be trained to conduct integrative research projects across domains such as human cognitive neuroscience, systems neuroscience, and computational neuroscience.

PROJECT SUMMARY Cognitive control allows us to flexibly guide our actions based on our goals. Central to most prominent theories of cognitive control is the control representation. For control to be successful, this representation is maintained in working memory by the prefrontal cortex (PFC) where it allows the same input to map to different responses depending on the context. Convergent evidence has found that the PFC encodes multiple task-relevant features of a task. However, little is known about the computational features of these control representations based on how they organize this information. This is a fundamental gap in our understanding. Here we focus on one such property, termed representational dimensionality. In technical terms, representational dimensionality refers to the number of axes needed to explain the variance in activity of a neural population across its inputs. Theoretical neuroscience has demonstrated that the dimensionality of a neural population determines a fundamental computational trade-off. A low dimensional representation will discard irrelevant information and form abstractions over its inputs. It is therefore suitable for generalization to new situations. A high dimensional representation encodes multiple mixtures of inputs into highly separable firing patterns without overlap. Understanding how generalizability and separability relate to cognitive control function promises gains on some of the most fundamental problems in control, including context-guided behavior, interference resolution, multitasking, and controlled-to-automatic behavior. The goal of this research program is to link the computational properties of high dimensional control representations to cognitive control function. Our overall hypothesis is that PFC forms high dimensional representations of task features which are needed in behavioral circumstances benefitting from separability. This hypothesis is motivated by theoretical neuroscience and foundational studies that have tested the dimensionality of PFC representations in animal models. However, no study in humans has studied high dimensional codes in PFC and no evidence in any species links dimensionality to cognitive control function. Through an NINDS R21 (NS108380), we have developed and refined two novel, complementary methods for estimating representational dimensionality from fMRI and EEG data. Using these approaches, we have found preliminary evidence that the dorsolateral PFC (DLPFC) forms a high dimensional code relative to other brain areas. We also find evidence from EEG that separability of high dimensional codes improves efficient, flexible behavior and may aid stable readout. Thus, we build on these initial observations to establish the nature, functional significance, and temporal dynamics of high dimensional control representations in the human brain.