Abstracts
SESSION I: Orbital Prefrontal Cortex
Functional Subdivisions within Macaque Orbital Prefrontal Cortex
Elisabeth A. Murray, PhD, National Institute of Mental Health, NIH
Damage to the orbital prefrontal cortex (PFo) results in marked changes in goal-directed behavior and affect. On the basis of patterns of anatomical connectivity and cytoarchitecture, two distinct subregions have been identified within the PFo: 1) a region corresponding to Walker's areas 11 and 13, and 2) a region corresponding to Walker's area 14. Although it has been proposed that the functions of the subregions may differ, evidence is limited. To explore the contribution of the PFo subregions to reward-guided behavior, we studied rhesus monkeys (Macaca mulatta) with excitotoxic lesions targeting either areas 11/13 or area 14. The performance of the two groups was compared to that of a group of unoperated controls on a series of reward-based tasks that have been shown to be sensitive to lesions of the PFo as a whole (Walker's areas 11, 13 and 14). The extent of the lesions was confirmed by T2-weighted MRI. We used three tasks: reinforcer devaluation, object reversal learning, and extinction of responses to a previously rewarded object. Monkeys with lesions of area 11/13 were impaired on reinforcer devaluation but not extinction. Monkeys with lesions of area 14 showed the converse result, namely, impairment in extinction but not reinforcer devaluation. Neither operated group was impaired on object reversal learning. These results reveal a double dissociation of function within PFo. In addition, the findings, which show selective changes in goal-directed behavior in the absence of changes in object reversal learning, contrast with previous findings from macaques with complete PFo lesions.
Orbitofrontal Cortex Contributions to Tracking Value in an Uncertain World
Mark E. Walton, DPhil, University of Oxford
While the orbitofrontal cortex (OFC) has for decades been implicated as being a critical structure underlying aspects of learning and flexible decision making, its precise function has remained elusive. One reason for this may be that primate OFC is not a uniform structure, but instead can be divided into lateral (lOFC) and medial sectors (mOFC) on the basis of their distinct patterns of anatomical connections. To date, however, there has been little evidence as to what the behavioral consequences of this anatomical division might be. Moreover, the vast majority of previous studies to investigate OFC function have utilized tasks with limited options to choose between and where any changes in reinforcement have directly signaled a necessity to switch to an alternative course of action. By contrast, the world that we and other animals inhabit allows for multiple alternative courses of action and is often uncertain as well as changeable. I will present data from a series of studies comparing the effects of lOFC and mOFC lesions in macaques on learning and decision making during performance on 3-option, stochastic "bandit" tasks that mimic some of the problems faced by animals in the natural world. These show that lOFC plays a critical role in appropriate credit assignment, such that damage to this region renders animals unable to determine causal relationships between choices and their contingent outcomes. By contrast, mOFC appears essential to make appropriate value comparisons and decisions that are unaffected by irrelevant and potentially distracting alternatives.
Path Coding by Orbitofrontal Cortex Neurons
Matthew L. Shapiro, PhD, Mount Sinai School of Medicine
Memory guides adaptive behavior: the past informs the present so that we can anticipate the outcome of goal-directed responses. To be useful, memory retrieval must be selective, directed by the salient features of situations, and flexible to adapt to changing internal goals, environmental opportunities, and potential actions. The neural mechanisms of selective memory retrieval are largely unknown. We hypothesize that the orbital frontal cortex (OFC) and hippocampus interact, integrating reward history with episodes, and provide key mechanisms for selective memory retrieval of goal-related representations.
We trained rats to find food in plus maze tasks that required both the OFC and the hippocampus, and recorded unit activity during stable performance, reversal learning, and strategy switching. OFC firing distinguished different rewarded paths, journeys from a start arm to a goal arm. Path-selective OFC cells fired differently during overlapping journeys that led to different goals or from different starts, resembling journey-dependent coding by hippocampal neurons. Activity of individual cells and the population correlated with performance as rats learned newly rewarded outcomes. Local field potentials (LFPs) recorded simultaneously in the OFC and the hippocampus oscillated coherently in the theta band (5-12 Hz) during stable performance. LFP coherence diminished when rats adapted to altered reward contingencies and followed different paths. Thus, OFC neurons appear to participate in a distributed network including the hippocampus that associates spatial paths, recent memory, and integrated reward history.
Contrasting Reward Signals in Orbitofrontal Cortex and Anterior Cingulate Cortex
Jonathan D. Wallis, PhD, Helen Wills Neuroscience Institute, University of California at Berkeley
Over the last decade, converging evidence has shown the critical importance of orbitofrontal cortex (OFC) for valuation and decision-making. At the same time, neurons in other brain areas, notably anterior cingulate cortex (ACC), have been found to encode much of the same information as OFC neurons. A current challenge is to determine how these two populations are functionally distinct. I will discuss two lines of research that aimed to meet this challenge.
In our first experiment, we examined whether the OFC and ACC differed with respect to the type of information that they were evaluating. Based on differences in the anatomical connections of the two areas, we hypothesized that ACC encoded value information as it related to internally-driven actions, while OFC encoded value information as it related to external stimuli in our environment. We found no evidence that this was the case: both areas were equally capable of encoding value information for actions or stimuli.
Our second experiment examined whether the value signal itself differed between the areas. Two related value signals are the prediction (the value of what one expects to receive) and the prediction error (the difference between what one expects and what one actually receives). These two signals are typically highly correlated. However, we were able to show that OFC encoded predictions not prediction errors, while the reverse was true for ACC. These results mark an important step in distinguishing different value signals in the frontal cortex.
The Involvement of Orbitofrontal Cortex in Goal-Directed Action
Bernard Balleine, PhD, Brain & Mind Research Institute, University of Sydney
We have previously reported that, whereas prelimbic prefrontal cortex mediates the effects of changes in experienced reward, the orbitofrontal cortex is more important for the influence of predicted reward particularly on choice between different goal-directed actions. More recently, we have found evidence implicating the lateral and ventral orbital regions in control by predicted reward based on: (1) the loss of sensitivity to select degradation of the pavlovian CS-US contingency after: (a) OFC lesions and (b) after meth-amphetamine exposure resulting in a reduction in c-fos labeling in OFC; and (2) evidence of functional connections between these regions of OFC and an area of the medial shell of the nucleus accumbens mediating outcome-specific Pavlovian-instrumental transfer. This region of the shell is critical for the influence of predicted reward on choice and takes glutamatergic afferents from the OFC. Delta-opioid receptors are likely present on the terminals of these projections presynaptic to dopamine D1, rather than D2, receptor-expressing MSNs. Thus we have found that: (1) delta-, not mu-, receptor knockout mice have a deficit in the specific transfer effects mediated by predicted reward values; (2) this deficit is replicated by infusion of naltrindole, and not CTAP, into the shell; (3) is also replicated by D1, not D2 antagonist infusions in the shell as well as by (4) contralateral infusions of naltrindole and a D1 antagonist. Together these data implicate the OFC in the circuitry mediating the effect of specific reward predictions on choice.
Orbitofrontal Contributions to Value-Driven Learning and Choice: Evidence from Humans with Focal Frontal Lobe Damage
Lesley K. Fellows, MD, CM, DPhil, Montreal Neurological Institute, McGill University
Orbitofrontal cortex (OFC) seems to play an important role in stimulus-value learning, and in value-driven decision making. Recent findings in animal models have begun to specify the nature of OFC contributions to learning from feedback in more detail. In parallel, neuroeconomics work has shown that OFC and ventromedial PFC encode the subjective value of options during choice. Here, we present evidence from studies in humans with focal frontal damage, aiming to establish which aspects of value-based learning and choice rely critically on OFC, and to probe whether these processes are dissociable. We examined stimulus-value learning from probabilistic feedback in a large sample of patients with frontal lobe damage. Learning in this dynamic stimulus-reinforcement environment was disrupted selectively by OFC damage, and not by damage to other frontal regions. A second study examined a selected group of patients with either OFC or dorsomedial PFC damage, comparing performance on the same stimulus-value learning task, and on an analogous action-value task. OFC damage disrupted the ability to sustain the correct choice of stimulus, but not of action, after positive feedback, while dorsomedial damage led to a similar deficit in action-value learning, while sparing stimulus-value learning. A final study tested whether OFC damage disrupted value-maximizing choices, using a paradigm developed in experimental economics to detect and quantify violations of utility theory. The findings were consistent with a critical role for OFC in value-maximizing choice.
SESSION II: Orbitofrontal Cortex and Emotional Memory
Interactions between the Hippocampus and Orbitofrontal Cortex in Memory for Overlapping Sequences of Nonspatial Events
Seth J Ramus, PhD, Bowdoin College
It has been proposed that long-term declarative memories are ultimately stored through interactions between the hippocampal memory system and the neocortical association areas that initially processed the to-be-stored information. One association neocortex, the orbitofrontal cortex (OF) is strongly and reciprocally connected with the hippocampal memory system and plays an important role in odor recognition memory in rats. We have previously reported the firing patterns of neurons in the OF rats performing a passive, 8-odor-sequence memory task. Most interesting were neurons that fired selectively in anticipation of specific odors - i.e., the cells demonstrated odor-selective firing prior to the rat inserting his nose in the odor port. We found that hippocampal lesions abolished the anticipatory firing in OF, suggesting that these anticipatory responses (memory) were in fact dependent on the hippocampus (Ramus, et al., 2007). Consistent with these findings, we have also found complex-spike neurons in the hippocampus that fired selectively prior to the presentation of odors within their sequential context (Ginther et al. 2011). Together, these studies suggest that the OF and hippocampus interact during the storage of long-term memories for sequences of odors, and that this anticipatory firing in the hippocampus might represent specific temporal locations within the odor sequence. These findings are also consistent with recent human neuroimaging studies that have revealed increased functional connectivity between the hippocampus and OF during the retrieval of overlapping sequences of images (Sherrill et al. 2010).
Different Time Courses for Learning-Related Changes in Amygdala and Orbitofrontal Cortex
C. Daniel Salzman, MD, PhD, Columbia University
The orbitofrontal cortex (OFC) and amygdala are thought to participate in reversal learning, a process that occurs when cue-outcome associations are switched. However, current theories disagree on whether OFC directs reversal learning in the amygdala. Here, we show that during reversal of cues' associations with rewarding and aversive outcomes, neurons that respond preferentially to stimuli predicting aversive events update more quickly in amygdala than OFC; meanwhile, OFC neurons that prefer reward-predicting stimuli update more quickly than those in the amygdala. After learning, in contrast, OFC consistently encodes impending reinforcement more rapidly than amygdala. Finally, analysis of local field potentials (LFPs) reveals a disproportionate influence of OFC on amygdala that emerges after learning. Current theories of OFC-amygdala function do not account for these data, which implies that reversal learning is supported by complex interactions between neural circuits spanning the amygdala and OFC, rather than directed by any single structure.
Prefrontal Cortex Contributions to the Control of Fear
Elizabeth A. Phelps, PhD, New York University
Research in non-human animals suggests the infralimbic cortex plays a critical role in the recollection of extinction learning in classical fear conditioning. It appears the infralimbic cortex enables the expression of extinction memory by inhibiting the expression of the conditioned fear memory represented in the amygdala. In humans, the analogue of this region is proposed to lie in the ventromedial prefrontal cortex (vmPFC). In this talk, I will discuss the role of the vmPFC in the control of conditioned fear in humans. Specifically, we compare BOLD responses when conditioned fears are diminished through 4 techniques: extinction, emotion regulation, reversal and reconsolidation. For extinction, emotion regulation and reversal, we find evidence consistent the involvement of the vmPFC in inhibiting the amygdala, with the magnitude of the BOLD response increasing with greater inhibition in reversal. In contrast, when conditioned fear is diminished by interfering with fear memory reconsolidation, presumably altering the amygdala-dependent representation of the fear memory rather than inhibiting it, we fail to see involvement of the vmPFC.
Positive and Negative Emotion Regulation in the Orbital and Lateral Prefrontal Cortex
A.C. Roberts, PhD, Institute of Behavioural and Clinical Neurosciences, University of Cambridge
Dysregulation in the orbitofrontal cortex (OFC) is associated with a variety of anxiety and mood disorders. Consistent with this, the BOLD signal within the human OFC and unit activity in rodent and primate OFC represent negative outcomes as well as predict such outcomes. Together these findings implicate the OFC in the regulation of negative emotion. Despite this wealth of evidence, only a handful of studies have investigated the effects of lesions to the OFC on negative emotion and these have produced conflicting results. For example, large lesions of the OFC in rhesus macaques, including areas 11,12,13 and 14, reduce overall levels of anxiety, whilst restricted lesions to ventral orbital and agranular insula regions of rat OFC either heighten fear reactivity and anxiety, or have no effect. Since there is increasing evidence for heterogeneity of function within OFC subdivisions, this paper will present recent findings comparing the effects of excitotoxic lesions of the medial (area 13) and lateral (area 12) regions in marmosets on the regulation of behavioural and cardiovascular arousal in conditioned fear and anxiety. These findings will be compared to our previous studies implicating medial OFC in the regulation of positive emotion. Since serotonin is a major target for drugs used to treat anxiety and mood disorders and there is a reciprocal relationship between the OFC and the serotonin-containing neurons in the dorsal raphe nucleus, this paper will also describe the effects of reductions in OFC serotonin on negative emotion regulation, including aspects of punishment processing and avoidance learning.
The Development of Decision Making and Emotional Memory in Monkeys with Neonatal Orbitofrontal Lesions
Jocelyne Bachevalier, PhD, Yerkes National Primate Research Center, Emory University
An increasing number of studies using a variety of experimental procedures in both animals and humans have demonstrated the significant contribution made by the orbital frontal cortex to the flexible monitoring of actions based on rewards processing. Yet, much remains to be discovered about the role played by this structure in the development of emotional responses and goal-directed behaviors, which are the prerequisites for the development of complex social behavior.
Three studies will be presented investigating behavioral and cognitive changes following neonatal damage of the orbital frontal cortex in infant rhesus monkeys. The first study demonstrates that neonatal orbital frontal lesions alter the modulation of fear and defensive responses towards threatening social stimuli, indicating poor modulation of social stimuli. These findings were confirmed by a second study showing that these same neonatal lesions disrupted choice selection predicted by affective signals but not by visual signals conveying reward contingency. However, the last study investigated how the orbital frontal cortex modulate fear reactivity and show that the neonatal orbital lesions did not impact fear learning or fear modulation by safety signals.
Discussion: These results are consistent with orbital frontal damage altering the complex and flexible monitoring of the reward values of emotional and social cues to select appropriate actions, even when the damage occurs in infancy. These functional alterations may in turn reduce general motivation to engage in social interactions. These experimental data shed some lights into the crucial role of this cortical area in developmental psychopathology.
SESSION III: Orbitofrontal Cortex, Value, and Decisions
Identity and Value Signaling Relevant to Learning in the Orbitofrontal Cortex and Ventral Striatum
Geoffrey Schoenbaum, MD, PhD, University of Maryland School of Medicine
Appetitive learning is thought to be driven by reward prediction errors that reflect the difference between the value of what we expect and what we actually receive. Yet learning also occurs when the identity of what we receive violates expectations, even if the "value" was just as expected. For example, if your employer starts paying you in bananas (or subprime mortgage backed securities if you work for AIG!) rather than cash, it will not matter how closely the number of bananas matches the value of your salary, you will notice the switch! And if you stick around, you will learn to expect bananas on payday. This type of learning, while critical to adaptive behavior, is not well accounted for by current theories of reinforcement learning, which predicate learning on changes in value, nor has its neural substrate been well explored. Here we employed unit recording and Pavlovian unblocking to assess the respective contributions of the orbitofrontal cortex and ventral striatum to learning driven by value- versus identity-based prediction errors. Consistent with popular Actor-Critic reinforcement learning theories, we found that only ventral striatum was necessary for learning in response to changes in reward value. By contrast, both regions, along with downstream regions in midbrain, were required for learning in response to changes in reward identity. These results, along with phasic activity in orbitofrontal and striatum neurons at the time of a reward shift, suggest that learning in response to changes in identity and value is mediated by partially - but not entirely - dissociable neural systems. Existing reinforcement learning models used to explain activity in these circuits should be modified to incorporate information about the identity of expected rewards. Supported by R01DA015718 to GS.
The Role of the Orbitofrontal Cortex in Simple Choice
Antonio Rangel, PhD, California Institute of Technology
A growing consensus in neuroscience suggests that the brain makes simple choices by assigning values to the options under considerations that are then compared to make a decision. This talk will describe recent human fMRI experiments showing how values are computed and compared in the orbitofrontal cortex at the time of choice, as well as the role that attention plays in such value signals.
Prefrontal Cortex and Hybrid Learning in Competitive Games
Daeyeol Lee, PhD, Kavli Institute for Neuroscience, Yale University School of Medicine
Animals learn not only from actual outcomes of their actions but also from hypothetical outcomes that could have resulted from unchosen actions. To improve the animal's behavioral strategies based on hypothetical outcomes, they must be attributed correctly to specific unchosen actions. To investigate the neural mechanisms for learning simultaneously from actual and hypothetical outcomes, we trained rhesus monkeys in a virtual rock-paper-scissors task. We found that the animal's choices were systematically influenced not only by the actual outcomes of their previous actions, but also by the hypothetical outcomes from their unchosen actions. In addition, neurons in the dorsolateral (DLPFC) and orbital prefrontal (OFC) cortices encoded actual and hypothetical outcomes related to specific actions. In particular, DLPFC neurons tended to change their activity similarly according to the actual and hypothetical outcomes from the same action. By contrast, OFC neurons tend to modulate their activity related to actual and hypothetical outcomes independently of the animal's actions. Neural signals in DLPFC and OFC related to both actual and hypothetical outcomes emerged immediately after the outcomes from chosen and unchosen actions were revealed. These results suggest that the prefrontal cortex might be involved in updating the values of chosen and unchosen actions simultaneously, thereby improving the efficiency of reinforcement learning.
Reward Value and Risk
Wolfram Schultz, MD, PhD, University of Cambridge
Rewards can be viewed as probability distributions of reward values. The key parameters defining probability distributions are expected value, which is the mean expected after many trials, and standard deviation, which in economics is also referred to as 'risk'. This definition of risk differs from the common notion that considers risk as the probability of losing (which is negative value). Value and risk are fundamental variables for economic decision making.
Neurons in the orbitofrontal cortex of monkeys process both value and risk, but mostly in separate subpopulations. These responses increase monotonically with higher standard deviations of binary equiprobable distributions of reward magnitudes. Their latency is sufficiently short to allow them to participate in decision making before overt choices.
Human imaging studies reveal risk coding distinct from value coding in frontal cortex. The risk-related activations covary with degrees of risk aversion in individual participants. The terms 'risk aversion' and 'risk seeking' indicate that risk influences the subjective valuation of outcomes; this concept constitutes a basic tenet of economic utility theory. As direct neuronal correlate, risk enhances neuronal value responses in lateral frontal cortex of human risk seekers and reduces value responses in risk avoiders.
Subpopulations of orbitofrontal neurons show adaptation of reward-related responses to the value and risk of predicted reward probability distributions. Adaptive neuronal coding may explain such behavioral phenomena as contrast effect and reference dependent coding. Adaptive coding serves to use the limited neuronal coding range for efficient discrimination between the almost unlimited numbers and values of rewards.
Evaluating the Contribution of Medial Orbitofrontal Cortex in Action and Stimulus-Based Decision Making
John P. O'Doherty, DPhil, California Institute of Technology
The medial orbitofrontal cortex and adjacent medial prefrontal cortex has been implicated in value-based decision making in humans. However, the precise contributions of this region to the decision process has remained elusive. Previous studies have found a role for this region specifically in goal-directed behavior, wherein actions are selected based on the incentive value of associated outcomes (or goals). Moreover, activity in this region has been found to correlate with the expected value (or utility) of the to be chosen option. However, an open question is precisely what type of decision mechanism underpinssuch response patterns. One type of decision-making involves rendering decisions over which of several different actions to select in order to obtain a reward outcome. Another form of decision-making is choice performed over "goods" or "outcomes", without explicit consideration of the particular action needed to achieve those outcomes. In this talk I will present evidence from fMRI to suggest a role for medial OFC in both of these two variants of decision-making. Together these findings suggest a very general role for this region in the integration of outcome values with the stimuli and responses necessary for goal-directed decision-making.
Contributions of Orbitofrontal Cortex and Lateral Prefrontal Cortex to Economic Choice and the Good-to-Action Transformation
Camillo Padoa-Schioppa, PhD, Washington University in St Louis
During economic choice, individuals assign values to the available options; a decision is then made by comparing values. Substantial work in recent years examined the neuronal encoding of subjective value, defined by the integration of multiple determinants (the commodity, its quantity, probability, physical effort, time delays, etc.). In a nutshell, current evidence shows that neurons in the orbitofrontal cortex (OFC) encode subjective values. In contrast, much less is known about the processes through which values are compared to make a decision. Traditionally, the leading hypothesis in neurophysiology has been that decisions unfold as processes of action selection ‒ a view variously referred to as "intentional framework", "standard model" or "action-based model". However, it is known that neurons in the OFC encode subjective values independently of the visuo-motor contingencies of choice (in goods space). This observation opens two important questions. First, in what space are values compared to make a decision (goods space or actions space)? Second, assuming that decisions are made in goods space, what are the neural mechanisms through which choice outcomes are transformed into suitable action plans? We examined these issues using a behavioral paradigm that dissociates in time economic choice and action planning. We recorded in monkeys from several frontal regions including OFC and lateral prefrontal cortex (LPFC). Our results indicate that decisions can be made independent of action planning. Furthermore, the activity of neurons in the LPFC reflects the initial stages of the good-to-action transformation.
SESSION IV: Orbitofrontal Cortex, Uncertainty, and Decisions
Olfactory Perceptual Decision-Making in Human Orbitofrontal Cortex
Jay A. Gottfried, MD, PhD, Northwestern University Feinberg School of Medicine
Recent studies suggest that accumulation of sensory evidence, with respect to time, is one plausible mechanism to disambiguate noisy inputs. However, current neuroscientific understanding of perceptual decisions is largely derived from animal models, and though human imaging studies have begun to examine how decisions are made in the wake of input ambiguity, the fundamental assumption that sensory evidence in the human brain accumulates over time has not been shown. We combined functional magnetic resonance imaging (fMRI) with computational modeling and temporal deconvolution techniques to assess within-trial accumulation of odor information in the human olfactory system. Subjects participated in a two-alternative categorization task in which they decided which of two odor perceptual qualities (lemon or clove) was dominant in a series of odorant mixtures ranging between 100% citral (lemon) and 100% eugenol (clove). The magnitude of neural activity in posterior piriform cortex (PPC) and orbitofrontal cortex (OFC) systematically varied with the rate of information accumulation toward a decision boundary. The response in PPC reached an early plateau, whereas the response in OFC increased gradually over the course of the trial and peaked at the time of decision. Our findings suggest that PPC represents the incoming ambiguity of olfactory signals while OFC integrates these signals to enable effective decisions.
Orbitofrontal Cortex Inactivation Impairs Confidence Reporting in Rats
Adam Kepecs, PhD, Cold Spring Harbor Laboratory
A fundamental component of decision-making under uncertainty is the ability to assign appropriate levels of confidence to each decision. Although our recent behavioral, electrophysiological and computational studies on orbitofrontal cortex (OFC) revealed that OFC activity correlates with decision confidence (Kepecs et al., 2008), this brain structure may be one of several nodes in a network subserving confidence-guided decisions. To explore the causal role of OFC in confidence judgments, we designed a novel behavioral paradigm in which rats' willingness to wait for a reward is proportional to their confidence about the correctness of their sensory decision. This waiting time measure provides a quantitative trial-by-trial estimate of rats' decision confidence.
To test the role of OFC in our confidence-reporting task, we bilaterally inactivated the lateral OFC with muscimol or infused it with a saline control solution. OFC inactivation produced a large and significant reduction in the dependence of reward waiting time on decision confidence without a change in the mean waiting time or in the choice accuracy of the sensory decision. These results provide evidence for the causal involvement of OFC in confidence estimation and establish a behavioral assay suitable for examining the neural mechanisms underlying this process.
Neural Dynamics and Ensemble Coding of Reward Expectancy and Probability in Rat Orbitofrontal Cortex
CMA Pennartz, PhD, Center for Neuroscience and Cognitive Science Center Amsterdam, University of Amsterdam
To study how orbitofrontal cortex (OFC) codes information affecting decision-making, it is important to gain access to its population activity. The impact of OFC on behavior depends on effects of its mass activity on target structures such as the striatum. We made ensemble recordings from rat orbitofrontal cortex during sensory discrimination tasks, monitoring activity of ~30-60 isolated neurons simultaneously. Trials consisted of odor sampling, a decision and locomotor phase, a waiting and an outcome phase. The entire task sequence appeared 'tesselated' by subsets of OFC neurons that fire specifically during one or two phases. When rats learned to associate different stimuli to distinct reward amounts, OFC ensembles conveyed significant information about expected reward magnitude. When animals learned to associate stimuli with a varying likelihood of receiving reward, OFC firing activity depended on reward probability, but not uncertainty.
We next considered the dynamics of information processing in OFC. The fine temporal structure of spike trains may greatly influence the way OFC outputs are processed by target areas and affect synaptic plasticity. A striking feature of OFC dynamics was the phase synchronization of neurons to Theta rhythm, which correlated strongly to reward expectancy. In contrast, phase-locking of spikes to Gamma oscillations occurred when activity of neurons coding action-outcome information was suppressed. Altogether, OFC coding of task elements can be considered a temporal ´scaffold´ that is modulated by outcome parameters. OFC phase synchronization to rhythmic mass activity is likely to affect neural mechanisms for integrating OFC output in target structures and behavior.
Prior and Likelihood Uncertainty are Differentially Represented in the Human Brain
Konrad Kording, PhD, Northwestern University
Background: Many experiments in motor control have shown that humans, when performing a movement (for example an arm-reaching task), can integrate information both from past history of movements (prior) and available sensory feedback (likelihood)[1]. Moreover, they seem to integrate prior and likelihood based on the respective uncertainties, in a way that is often close to the Bayesian optimal. This means that uncertainty in both the prior and the likelihood needs to be encoded somehow in the human brain[2]. However, how and where it is encoded is still largely unknown.
Objective: Here we wanted to know first if human behavior is also close to optimal in a more cognitive task and second, which brain areas are involved in representation and integration of prior and likelihood uncertainty, and whether these areas overlap.
Methods: A total of twenty-seven adult subjects participated in the behavioral portion of this study. Each subject performed a decision-making task which consisted of guessing the position of a hidden target on a screen. The position of the target was sampled from a 1-D Gaussian distribution (the prior) in which the mean was fixed and the variance was kept constant inside each block of trials, but changed between blocks. In every trial 5 dots were shown in the screen (the likelihood), whose x-position was drawn from another 1-D Gaussian distribution, in which the mean was the hidden target and the variance changed randomly between trials. The subjects then tried to guess the position of the hidden target and, after the choice was made, the real position of the target was shown. Fifteen of these subjects then performed the same task in an fMRI scanner.
Results: We found that people readily combined information from both the position of the likelihood dots as well as previous knowledge about the target distribution in a way similar to the predictions from Bayesian decision theory. Analysis of the fMRI results showed that higher prior uncertainty was correlated with stronger activations in the insula, caudate, amygdala and putamen and the lateral orbito-frontal cortex. Higher uncertainty in the likelihood, on the other side, activated areas in the visual cortex, which do not disappear after controlling for potential confounds such as foveal representation. This suggests that the uncertainty associated with visual sensory feedback may be preprocessed already in the visual cortex, while prior uncertainty might be represented more deeply in the brain.
Conclusions: Our results indicate that the human brain makes use of different pathways to represent and integrate uncertainty about prior and likelihood, and offer a potential neural mechanism for optimal Bayesian decision making.
Social Value Representation in Primate Orbitofrontal Cortex
Karli K. Watson, PhD, Center for Cognitive Neuroscience, Duke University
In order to successfully navigate their social environment, biological organisms must be able to ascribe appropriate reward value to the social stimuli around them. Human fMRI studies have demonstrated that orbital and medial prefrontal regions are typically activated in response to rewarding stimuli across a myriad of modalities, and electrophysiological recordings in animals have revealed that neurons in the orbitofrontal cortex (OFC) are modulated by the value of nutritive rewards and their predictive cues. Here we show that, in the context of a decision making task, single neurons in the macaque monkey OFC display firing rate changes in response to social image display as well as to the receipt of a fluid reward. A significant proportion of these neurons encode social image category, the size of the juice reward, or both. Moreover, firing rate is scaled by the relative value of the social images as well as by the size and value of nutritive rewards, and neuronal activity in posterior lateral OFC utilizes a common encoding scheme across these two different modalities of rewards. These data endorse the notion that the orbitofrontal cortex assesses social as well as gustatory value and plays a role in computing tradeoffs between various types of commodities.