How do we select appropriate actions to reach a desired goal?

How do we implement control to retrieve information from memory?

Cognitive control involves choosing from among a set of actions or representations in order to achieve a certain goal or outcome. We use cognitive control to carry out a variety of tasks, from making a cup of coffee to remembering where the house keys are. The frontal lobes broadly support cognitive control; regions within the frontal lobe guide action selection, so that we add the coffee grounds before turning on the coffee maker. Regions within the frontal lobe also guide memory retrieval, so that we can still search for our keys even when limited external cues are available.

Our lab is interested in how the organization of the frontal lobe influences goal directed behavior, as well as dissociations and interactions between the frontal and medial temporal lobe regions in their contribution to memory function. In our research, we use a variety of methods to investigate these questions, including functional magnetic resonance imaging (fMRI) as well as behavioral testing of healthy adults and patient populations.

The Badre lab research is supported by the Alfred P. Sloan Foundation, The National Institute of Health and Brown University.

Current Projects

Exception-handling in the learning and generalization of rules


The flexibility of human behavior relies on our ability to deploy a variety of previously learnt, abstract rules in novel situations. However, rules often fit a novel situation imperfectly. Our ability to deal with exceptions to rules, therefore, is crucial, both for maintaining abstract rules, and enabling their use in different tasks. In this study, we examine exception learning strategies and how they impact rule learning. Behavioral experiments provide evidence for a dissociation between rule learning and exception learning, suggesting separate mechanisms. Moreover, different subjects are able to deploy different task rules to the same task by leveraging their ability to separately learn about exceptions. Further experiments will examine how exceptions impact rule learning and generalization.
Project Lead: Apoorva Bhandari

Hierarchical Control of Task Sequences


Frontal neocortex is thought to support our highest intellectual abilities, including our ability to plan and enact a sequence of tasks toward a desired goal. In everyday life, such task sequences are abstract in that they do not require consistent movement sequences and are often assembled “on the fly”. Yet, remarkably little is known about the necessity of frontal sub-regions for such control. We asked participants to repeat short, simple sequences of four simple tasks during fMRI scanning. We have found that rostrolateral prefrontal cortex (RLPFC) activation ramped over position in the sequence, and reset at the initiation of each new sequence. To establish the necessity of RLPFC in this task and specify its functional role, a second group of participants performed the sequential task while undergoing transcranial magnetic stimulation (TMS) of the RLPFC or a control region. RLPFC stimulation increasingly disrupted task performance as each sequence progressed. These data establish RLPFC as necessary for uncertainty resolution during sequence-level control. Ongoing studies are currently investigating the specific role of uncertainty in the performance of sequential tasks and how sequences of tasks are learned.
Project Lead: Theresa Desrochers

Striatal Contributions to Retrieval


Recently, there has been renewed focus on the relationship between the striatum and declarative memory. However, the striatal contribution to declarative memory retrieval remains unclear. We have recently proposed three hypotheses for striatum’s role in retrieval: (1) striatum modulates the re-encoding of retrieved items in accord with their expected utility (adaptive encoding), (2) striatum selectively admits information into working memory that is expected to increase the likelihood of successful retrieval (adaptive gating), and (3) striatum enacts adjustments in cognitive control based on the outcome of retrieval (reinforcement learning). Our ongoing work begins to test these hypotheses. In one project, we administer a recognition memory paradigm with probabilistic false positive feedback. Striatally-mediated reinforcement learning should drive a shift in decision criteria in this task. Our results suggest that people can adapt their criteria if external feedback sufficiently violates their expectations. In another ongoing project, we are testing whether striatally-mediated reinforcement learning could support changes in other aspects of memory, including the kinds of memory strategies that subjects adopt. An integrative theme across these studies is the use of functional neuroimaging in combination with an array of modeling techniques, including signal-detection theory, drift-diffusion models, and reinforcement-learning models.
Project Lead: Jason Scimeca

Learning the dynamic structure of a task


Learning how to perform a typical cognitive task often involves figuring out its 'rules' - how stimuli and actions are related to outcomes. The mechanisms underlying this process are well studied. But, learning a task also involves adapting internal cognitive processes to the task's 'dynamic structure' - i.e. the specific timing and order of events relevant to the task. This study investigates this latter process and how it interacts with rule learning. In behavioral experiments, we find evidence that subjects learn a task's dynamic structure in the initial trials, and they transfer this knowledge between tasks. Subjects showed both positive (same task structure) and negative (different task structure) transfer, independent of the rules of the task. fMRI experiments in the pipeline will investigate the neurobiological mechanisms underlying this process.
Project Lead: Apoorva Bhandari

The Cost of Cognitive Effort


Would you rather perform a long series of multiplications by hand or using a calculator? Despite individual differences, people often choose the latter when given the option, even if they are equally capable of performing both tasks. 'Demand avoidance' is argued to stem from the disutility of cognitive effort, the cost of which is integrated into the benefit of performing a task. In order to determine the nature of this cost, we aim to specify the cost function for cognitive effort and explore its origins in the brain. Preliminary behavioral findings showed that the subjective value of a task decreases with increasing cognitive effort. Reward processing areas of the brain, such as ACC and striatum might be tracking this cost associated with mental labor.
Project Lead: Ceyda Sayali

The Representational Capacity of the Human Prefrontal Cortex


The PFC supports complex behaviors by providing flexible task representations. It has been proposed that neurons in the PFC that show mixed selective responses combine as a population to produce distributed representations of all combinations of task-relevant dimensions. This high-dimensional capacity theoretically allows any combinatorial mixture of features to be read out in support of flexible cognitive control. However, it is unknown whether high-dimensional capacity is unique to PFC or is, alternatively, a fundamental and general computational characteristic of association cortices. Critically, it is necessary to test whether the high-dimensional capacity observed in non-human primate PFC is conserved in humans, capable of the most complex of behaviors without extensive training. Thus, this study aimed to estimate the dimensional capacity of human PFC and other cortical areas in the human brain, using a similar pattern classification approach as has been implemented in the non-human primate. Human participants were scanned with high-resolution fMRI while performing two-item sequential memory tasks, adapted from studies in monkeys. We found that many combinations of task aspects (cues/task-types) were indeed decodable in distributed voxels within PFC, implicating high-dimensional capacity and consistent with non-human primate PFC. However, high representational capacity was observed in additional neocortical areas as well, particularly within occipital cortex. By contrast, other regions, such as motor and parietal, showed relatively low dimensional capacity, encoding fewer task representations. We consider these results with regard to the functional organization and computational nature of the neocortex – both in PFC and in the brain more broadly.
Project Lead: Patti Shih

Development of Rule-Guided Behavior and Hierarchical Cognitive Control


Previous work indicates that developmental change in rule-guided behavior in the transition from childhood to adolescence derives from improvements in the capacity to manage higher levels of rule (policy) abstraction but not in the ability to manage an increasing number of competing options at a given level of abstraction. We suggest that these findings can be understood within the framework of gating mechanisms in working memory (WM), that control the updating of relevant contextual information in order to bias the selection of appropriate behavioral rules. Recent neurocomputational models of corticostriatal circuits in WM and learning distinguish two gating functions: an “input gate” for updating task-relevant information into WM and an “output gate” for selecting which of the currently maintained representations exerts a top-down influence on attention and behavior. In this project, we test whether developmental improvements in abstract rule use result from change in input gating, change in output gating, or both. Our results suggest that higher output gating demands are associated with greater behavioral cost in children compared to adolescents, while there was less evidence for age differences in input gating. Despite their apparent difficulties with output gating, children were less likely to proactively prepare a response but rather adopted a more costly output gating strategy even when there was no requirement to do so.
Project Lead: Kerstin Unger

Transfer of Hierarchical Task Structure


When faced with a novel situation, we often draw on experiences from analogous scenarios to respond appropriately. In this study, we show that people likely accomplish this through the extraction of a tasks' core features and the application of this simplifying abstraction to a new task, despite surface level differences. Behaviorally, participants show positive transfer across tasks sharing similar underlying structure in their stimulus-response mappings. A comparison to several control groups whose first tasks varied in their relation to the second tasks shows that the performance increase is driven by facilitation, in the case of analogous situations, rather than inhibition, in the case of non-analogous situations. The transfer of abstract structure may be responsible for this facilitation.
Project Lead: Chris Gagne

Striatal Correlates of Working Memory Utility


WM resources should be reallocated when the utility of a maintained item drops, but prior work indicates such reallocation is sluggish. Here we show that the historical utility of maintained items predicts both this sluggishness and activity in ventral striatum (VS). One possibility is that VS activity encodes the utility of these maintained items, and may delay or even “veto” their removal.
Project Lead: Chris Chatham

Hierarchical Gating of the Output From Working Memory


Although information in working memory (WM) supports control over attention and behavior, not all information in WM will be relevant for behavior at any one moment. For this reason many models predict that the influence of working memory – that is, its output – is controlled by a gate. In contrast to gate-like modulation of the input to WM (“input gating”), such output gating has received relatively little attention. In this project, we tested whether output gating could act as a striatal site of hierarchical control, as suggested by one prior model. We found that higher-order frontal regions were differentially sensitive to demands on selective output gating, as opposed to input gating. This recruitment predicted individual differences in output gating efficiency. Overlapping regions of frontal cortex were also more strongly coupled with striatum during output gating. This coupling also predicted output gating efficiency. Together, these results suggest that output gating may be striatally-mediated and act as a key interface for hierarchical control.
Project Lead: Chris Chatham

Multiple Pathways To Retrieval


Ventrolateral prefrontal cortex (VLPFC) contributes to the control of retrieval however, little is known about the functional paths by which VLPFC interacts with memory processing regions. Functional magnetic resonance imaging (fMRI) studies have shown that the need for controlled retrieval, a process that assists in memory access, is associated with activation in anterior VLPFC (pars opercularis [BA47]). We demonstrated that the MRI signal in anterior VLPFC, temporal pole, anterior parahippocampal cortex and hippocampus was inversely correlated with source availability at retrieval suggesting their involvement in controlled retrieval. Additional functional connectivity analyses indicate that aVLPFC, temporal pole and hippocampus form a functional network. This ventral retrieval pathway is separable from neighboring functional networks associated with different VLPFC subregions. Further analysis of these data suggest mid VLPFC contributes to post-retrieval control rather than controlled retrieval and functionally couples with a separable fronto-parietal network.
Project Lead: Jennifer Barredo

Hemodynamic Correlates of Theta Oscillations During Retrieval


Evidence from fMRI has consistently located a widespread network of frontal, parietal, and temporal lobe regions during episodic retrieval. But how do these distributed regions of the brain dynamically interact in support of control of episodic retrieval? We recently proposed that theta oscillations represent interactions between brain systems for the cognitive control of episodic retrieval (Nyhus & Curran, 2010). In order to identify networks related to theta oscillations during episodic retrieval we used simultaneous EEG/fMRI. Combined EEG/fMRI results showed that theta power was correlated with the fMRI BOLD response in VLPFC, parietal, and temporal cortex. These results suggest that a fronto-parietal network interacts at theta frequency during episodic retrieval.
Project Lead: Erika Nyhus