The role of state uncertainty in the dynamics of dopamine

Kappa opioid receptors diminish spontaneous dopamine signals in awake mice through multiple mechanisms

The role of the dynorphin/kappa opioid receptor (KOR) system in dopamine (DA) regulation has been extensively investigated. KOR activation reduces extracellular DA concentrations, but the exact mechanism(s) through which this is accomplished are not fully elucidated. To explore KOR influences on real-time DA fluctuations, we used the photosensor dLight1.2 with fiber photometry in the nucleus accumbens (NAc) core of freely moving male and female C57BL/6J mice. First, we established that the rise and fall of spontaneously arising DA signals were due to DA release and reuptake, respectively. Next, mice were systemically administered the KOR agonist U50,488H in the presence or absence of the KOR antagonist aticaprant. U50,488H reduced both the amplitude and width of spontaneous signals in both sexes. Further, the slope of the correlation between amplitude and width was increased, indicating that DA uptake rates were increased. U50,488H also reduced the frequency of occurrence of signals in males and females. The effects of KOR activation were stronger in males, while effects of KOR antagonism were stronger in females. Overall, KORs exerted significant inhibitory control over spontaneous DA signaling, acting through at least three mechanisms - inhibiting DA release, promoting DA transporter-mediated uptake, and reducing the frequency of signals.

Multi-timescale reinforcement learning in the brain

To thrive in complex environments, animals and artificial agents must learn to act adaptively to maximize fitness and rewards. Such adaptive behaviour can be learned through reinforcement learning1, a class of algorithms that has been successful at training artificial agents2-5 and at characterizing the firing of dopaminergic neurons in the midbrain6-8. In classical reinforcement learning, agents discount future rewards exponentially according to a single timescale, known as the discount factor. Here we explore the presence of multiple timescales in biological reinforcement learning. We first show that reinforcement agents learning at a multitude of timescales possess distinct computational benefits. Next, we report that dopaminergic neurons in mice performing two behavioural tasks encode reward prediction error with a diversity of discount time constants. Our model explains the heterogeneity of temporal discounting in both cue-evoked transient responses and slower timescale fluctuations known as dopamine ramps. Crucially, the measured discount factor of individual neurons is correlated across the two tasks, suggesting that it is a cell-specific property. Together, our results provide a new paradigm for understanding functional heterogeneity in dopaminergic neurons and a mechanistic basis for the empirical observation that humans and animals use non-exponential discounts in many situations9-12, and open new avenues for the design of more-efficient reinforcement learning algorithms.

Quantitative dynamics of neural uncertainty in sensory processing and decision-making during discriminative learning

Uncertainty is crucial in sensory processing, necessitating further quantitative research on its neural representation in the sensory cortex. Here, to address this need, we used a deep learning approach to quantify uncertainties in neural activity from the forelimb area of the primary somatosensory cortex (fS1) during a vibration frequency discrimination task, introducing a transformer model designed to decode neural data not consistently tracked over time. Our model shows that the neural representation of fS1 encodes uncertainties not only from vibratory stimuli but also from decision-making processes, emphasizing its crucial role across various biological contexts. We confirmed that uncertainty decreases as learning progresses and increases with interruptions in learning. In line with predictions from previous studies, we also observed that uncertainty is high at psychometric thresholds. Furthermore, high uncertainty correlates with incorrect decisions, and we have identified dynamics in uncertainty between previous and current trials. Such findings underscore the evolving role of fS1 in assessing uncertainty for the brain's downstream areas as learning progresses.

The devilish details affecting TDRL models in dopamine research

Over recent decades, temporal difference reinforcement learning (TDRL) models have successfully explained much dopamine (DA) activity. This success has invited heightened scrutiny of late, with many studies challenging the validity of TDRL models of DA function. Yet, when evaluating the validity of these models, the devil is truly in the details. TDRL is a broad class of algorithms sharing core ideas but differing greatly in implementation and predictions. Thus, it is important to identify the defining aspects of the TDRL framework being tested and to use state spaces and model architectures that capture the known complexity of the behavioral representations and neural systems involved. Here, we discuss several examples that illustrate the importance of these considerations.

Striatal dopamine represents valence on dynamic regional scales

Adaptive decision making relies on dynamic updating of learned associations where environmental cues come to predict valenced stimuli, such as food or threat. Cue-guided behavior depends on a network of brain systems, including dopaminergic projections to the striatum. Critically, it remains unclear how dopamine signaling across the striatum encodes multi-valent, dynamic learning contexts, where positive and negative associations must be rapidly disambiguated. To understand this, we employed a Pavlovian discrimination paradigm, where cues predicting food or threat were intermingled during conditioning sessions, and their meaning was serially reversed across training. We found that male and female rats readily distinguished these cues and updated their behavior rapidly upon valence reversal. Using fiber photometry, we recorded dopamine signaling in three major striatal subregions - the dorsolateral striatum (DLS), the nucleus accumbens (NAc) core, and the nucleus accumbens medial shell - finding that valence was represented uniquely across all three regions, indicative of local signals biased for value and salience. Further, ambiguity introduced by cue reversals reshaped striatal dopamine on different timelines: nucleus accumbens signals updated more readily than those in the DLS. Together, these results indicate that striatal dopamine flexibly encodes stimulus valence according to region-specific rules, and these signals are dynamically modulated by changing contingencies in the resolution of ambiguity about the meaning of environmental cues.Significance Statement Adaptive decision making relies on updating learned associations to disambiguate predictions of reward or threat. This cue-guided behavior depends on striatal dopamine, but it remains unclear how dopamine signaling encodes multi-valent, dynamic learning contexts. Here, we employed a paradigm where cues predicting positive and negative outcomes were intermingled, and their meaning was serially reversed across time. We recorded dopamine signaling, finding heterogeneous patterns of valence encoding across striatal subregions, and cue reversal reshaped subregional signals on different timelines. Our results suggest that dopamine flexibly encodes dynamic learning contexts to resolve ambiguity about the meaning of environmental cues.

Fiber photometry analysis of spontaneous dopamine signals: The z-scored data are not the data

Fluorescent sensors have revolutionized the measurement of molecules in the brain, and the dLight dopamine sensor has been used extensively to examine reward- and cue-evoked dopamine release, but only recently has the field turned its attention to spontaneous release events. Analysis of spontaneous events typically requires evaluation of hundreds of events over minutes to hours, and the most common method of analysis, z-scoring, was not designed for this purpose. Here, we compare the accuracy and reliability of three different analysis methods to identify pharmacologically induced changes in dopamine release and uptake in freely moving C57BL/6J mice. The D1-like receptor antagonist SCH23390 was used to prevent dLight sensors from interacting with dopamine in the extracellular space, while cocaine was used to inhibit uptake and raclopride to increase release of dopamine in the nucleus accumbens. We examined peak-to-peak frequency, peak amplitude, and width, the time spent above an established cutoff. The three methods were 1) the widely-used "Z-Score Method", which automatically smooths baseline drift and normalizes recordings using signal-to-noise ratios, 2) a "Manual Method", in which local baselines were adjusted manually and individual cutoffs were determined for each subject, and 3) the "Prominence Method" that combines z-scoring with prominence assessment to tag individual peaks, then returns to the preprocessed data for kinetic analysis. First, SCH23390 drastically reduced the number of signals detected as expected, but only when the Manual Method was used. Z-scoring failed to identify any changes, due to its amplification of noise when signals were diminished. Cocaine increased signal width as expected using the Manual and Prominence Methods, but not the Z-Score Method. Finally, raclopride-induced increases in amplitude were correctly identified by the Manual and Prominence Methods. The Z-Score Method failed to identify any of the changes in dopamine release and uptake kinetics. Thus, analysis of spontaneous dopamine signals requires assessment of the %ΔF/F values, ideally using the Manual Method, and the use of z-scoring is not appropriate.

Navigating the unknown: Leveraging self-information and diversity in partially observable environments

Reinforcement learning algorithms often struggle to learn in partially observable environments, where different states of the environment may appear identical. However, not all partially observable environments pose the same level of difficulty for learning. This work introduces the concept of dissonance distance, a metric that can estimate the difficulty of learning in such environments. We demonstrate that self-information, such as internal oscillations or memory of previous actions, can increase the dissonance distance and make learning easier in partially observable environments. Additionally, sensory occlusion may occur after learning was completed, leading to a lack of sufficient information and catastrophic failure. To address this, we propose a spatially layered architecture (SLA) inspired by the brain, which trains multiple policies in parallel for the same task. SLA can change the amount of external information processed at each timestep, providing an adaptive approach to handle the changing information in the environment state-space. We evaluate the effectiveness of our SLA method showing learnability and robustness against realistic noise and occlusion in sensory inputs for the partially observable Continuous Mountain Car environment. We hypothesize that multi-policy approaches like SLA might explain the complex dopamine dynamics in the brain that cannot be explained with the state of the art scalar Temporal Difference error.

Valence-dependent dopaminergic modulation during reversal learning in Parkinson's disease: A neurocomputational approach

Reinforcement learning, crucial for behavior in dynamic environments, is driven by rewards and punishments, modulated by dopamine (DA) changes. This study explores the dopaminergic system's influence on learning, particularly in Parkinson's disease (PD), where medication leads to impaired adaptability. Highlighting the role of tonic DA in signaling the valence of actions, this research investigates how DA affects response vigor and decision-making in PD. DA not only influences reward and punishment learning but also indicates the cognitive effort level and risk propensity in actions, which are essential for understanding and managing PD symptoms. In this work, we adapt our existing neurocomputational model of basal ganglia (BG) to simulate two reversal learning tasks proposed by Cools et al. We first optimized a Hebb rule for both probabilistic and deterministic reversal learning, conducted a sensitivity analysis (SA) on parameters related to DA effect, and compared performances between three groups: PD-ON, PD-OFF, and control subjects. In our deterministic task simulation, we explored switch error rates after unexpected task switches and found a U-shaped relationship between tonic DA levels and switch error frequency. Through SA, we classify these three groups. Then, assuming that the valence of the stimulus affects the tonic levels of DA, we were able to reproduce the results by Cools et al. As for the probabilistic task simulation, our results are in line with clinical data, showing similar trends with PD-ON, characterized by higher tonic DA levels that are correlated with increased difficulty in both acquisition and reversal tasks. Our study proposes a new hypothesis: valence, signaled by tonic DA levels, influences learning in PD, confirming the uncorrelation between phasic and tonic DA changes. This hypothesis challenges existing paradigms and opens new avenues for understanding cognitive processes in PD, particularly in reversal learning tasks.

Dopamine release plateau and outcome signals in dorsal striatum contrast with classic reinforcement learning formulations

We recorded dopamine release signals in centromedial and centrolateral sectors of the striatum as mice learned consecutive versions of visual cue-outcome conditioning tasks. Dopamine release responses differed for the centromedial and centrolateral sites. In neither sector could these be accounted for by classic reinforcement learning alone as classically applied to the activity of nigral dopamine-containing neurons. Medially, cue responses ranged from initial sharp peaks to modulated plateau responses; outcome (reward) responses during cue conditioning were minimal or, initially, negative. At centrolateral sites, by contrast, strong, transient dopamine release responses occurred at both cue and outcome. Prolonged, plateau release responses to cues emerged in both regions when discriminative behavioral responses became required. At most sites, we found no evidence for a transition from outcome signaling to cue signaling, a hallmark of temporal difference reinforcement learning as applied to midbrain dopaminergic neuronal activity. These findings delineate a reshaping of striatal dopamine release activity during learning and suggest that current views of reward prediction error encoding need review to accommodate distinct learning-related spatial and temporal patterns of striatal dopamine release in the dorsal striatum.

Mating proximity blinds threat perception

Romantic engagement can bias sensory perception. This 'love blindness' reflects a common behavioural principle across organisms: favouring pursuit of a coveted reward over potential risks1. In the case of animal courtship, such sensory biases may support reproductive success but can also expose individuals to danger, such as predation2,3. However, how neural networks balance the trade-off between risk and reward is unknown. Here we discover a dopamine-governed filter mechanism in male Drosophila that reduces threat perception as courtship progresses. We show that during early courtship stages, threat-activated visual neurons inhibit central courtship nodes via specific serotonergic neurons. This serotonergic inhibition prompts flies to abort courtship when they see imminent danger. However, as flies advance in the courtship process, the dopaminergic filter system reduces visual threat responses, shifting the balance from survival to mating. By recording neural activity from males as they approach mating, we demonstrate that progress in courtship is registered as dopaminergic activity levels ramping up. This dopamine signalling inhibits the visual threat detection pathway via Dop2R receptors, allowing male flies to focus on courtship when they are close to copulation. Thus, dopamine signalling biases sensory perception based on perceived goal proximity, to prioritize between competing behaviours.

Explaining dopamine through prediction errors and beyond

The most influential account of phasic dopamine holds that it reports reward prediction errors (RPEs). The RPE-based interpretation of dopamine signaling is, in its original form, probably too simple and fails to explain all the properties of phasic dopamine observed in behaving animals. This Perspective helps to resolve some of the conflicting interpretations of dopamine that currently exist in the literature. We focus on the following three empirical challenges to the RPE theory of dopamine: why does dopamine (1) ramp up as animals approach rewards, (2) respond to sensory and motor features and (3) influence action selection? We argue that the prediction error concept, once it has been suitably modified and generalized based on an analysis of each computational problem, answers each challenge. Nonetheless, there are a number of additional empirical findings that appear to demand fundamentally different theoretical explanations beyond encoding RPE. Therefore, looking forward, we discuss the prospects for a unifying theory that respects the diversity of dopamine signaling and function as well as the complex circuitry that both underlies and responds to dopaminergic transmission.

Frontal Norepinephrine Represents a Threat Prediction Error Under Uncertainty

BACKGROUND

To adapt to threats in the environment, animals must predict them and engage in defensive behavior. While the representation of a prediction error signal for reward has been linked to dopamine, a neuromodulatory prediction error for aversive learning has not been identified.

METHODS

We measured and manipulated norepinephrine release during threat learning using optogenetics and a novel fluorescent norepinephrine sensor.

RESULTS

We found that norepinephrine response to conditioned stimuli reflects aversive memory strength. When delays between auditory stimuli and footshock are introduced, norepinephrine acts as a prediction error signal. However, temporal difference prediction errors do not fully explain norepinephrine dynamics. To explain noradrenergic signaling, we used an updated reinforcement learning model with uncertainty about time and found that it explained norepinephrine dynamics across learning and variations in temporal and auditory task structure.

CONCLUSIONS

Norepinephrine thus combines cognitive and affective information into a predictive signal and links time with the anticipation of danger.

A feature-specific prediction error model explains dopaminergic heterogeneity

The hypothesis that midbrain dopamine (DA) neurons broadcast a reward prediction error (RPE) is among the great successes of computational neuroscience. However, recent results contradict a core aspect of this theory: specifically that the neurons convey a scalar, homogeneous signal. While the predominant family of extensions to the RPE model replicates the classic model in multiple parallel circuits, we argue that these models are ill suited to explain reports of heterogeneity in task variable encoding across DA neurons. Instead, we introduce a complementary 'feature-specific RPE' model, positing that individual ventral tegmental area DA neurons report RPEs for different aspects of an animal's moment-to-moment situation. Further, we show how our framework can be extended to explain patterns of heterogeneity in action responses reported among substantia nigra pars compacta DA neurons. This theory reconciles new observations of DA heterogeneity with classic ideas about RPE coding while also providing a new perspective of how the brain performs reinforcement learning in high-dimensional environments.

Learning to express reward prediction error-like dopaminergic activity requires plastic representations of time

The dominant theoretical framework to account for reinforcement learning in the brain is temporal difference learning (TD) learning, whereby certain units signal reward prediction errors (RPE). The TD algorithm has been traditionally mapped onto the dopaminergic system, as firing properties of dopamine neurons can resemble RPEs. However, certain predictions of TD learning are inconsistent with experimental results, and previous implementations of the algorithm have made unscalable assumptions regarding stimulus-specific fixed temporal bases. We propose an alternate framework to describe dopamine signaling in the brain, FLEX (Flexibly Learned Errors in Expected Reward). In FLEX, dopamine release is similar, but not identical to RPE, leading to predictions that contrast to those of TD. While FLEX itself is a general theoretical framework, we describe a specific, biophysically plausible implementation, the results of which are consistent with a preponderance of both existing and reanalyzed experimental data.

Uncertainty of treatment efficacy moderates placebo effects on reinforcement learning

The placebo-reward hypothesis postulates that positive effects of treatment expectations on health (i.e., placebo effects) and reward processing share common neural underpinnings. Moreover, experiments in humans and animals indicate that reward uncertainty increases striatal dopamine, which is presumably involved in placebo responses and reward learning. Therefore, treatment uncertainty analogously to reward uncertainty may affect updating from rewards after placebo treatment. Here, we address whether different degrees of uncertainty regarding the efficacy of a sham treatment affect reward sensitivity. In an online between-subjects experiment with N = 141 participants, we systematically varied the provided efficacy instructions before participants first received a sham treatment that consisted of listening to binaural beats and then performed a probabilistic reinforcement learning task. We fitted a Q-learning model including two different learning rates for positive (gain) and negative (loss) reward prediction errors and an inverse gain parameter to behavioral decision data in the reinforcement learning task. Our results yielded an inverted-U-relationship between provided treatment efficacy probability and learning rates for gain, such that higher levels of treatment uncertainty, rather than of expected net efficacy, affect presumably dopamine-related reward learning. These findings support the placebo-reward hypothesis and suggest harnessing uncertainty in placebo treatment for recovering reward learning capabilities.

Valence ambiguity dynamically shapes striatal dopamine heterogeneity

Adaptive decision making relies on dynamic updating of learned associations where environmental cues come to predict positive and negatively valenced stimuli, such as food or threat. Flexible cue-guided behaviors depend on a network of brain systems, including dopamine signaling in the striatum, which is critical for learning and maintenance of conditioned behaviors. Critically, it remains unclear how dopamine signaling encodes multi-valent, dynamic learning contexts, where positive and negative associations must be rapidly disambiguated. To understand this, we employed a Pavlovian discrimination paradigm, where cues predicting positive and negative outcomes were intermingled during conditioning sessions, and their meaning was serially reversed across training. We found that rats readily distinguished these cues, and updated their behavior rapidly upon valence reversal. Using fiber photometry, we recorded dopamine signaling in three major striatal subregions -,the dorsolateral striatum (DLS), the nucleus accumbens core, and the nucleus accumbens medial shell - and found heterogeneous responses to positive and negative conditioned cues and their predicted outcomes. Valence ambiguity introduced by cue reversal reshaped striatal dopamine on different timelines: nucleus accumbens core and shell signals updated more readily than those in the DLS. Together, these results suggest that striatal dopamine flexibly encodes multi-valent learning contexts, and these signals are dynamically modulated by changing contingencies to resolve ambiguity about the meaning of environmental cues.

A dopamine mechanism for reward maximization

Individual survival and evolutionary selection require biological organisms to maximize reward. Economic choice theories define the necessary and sufficient conditions, and neuronal signals of decision variables provide mechanistic explanations. Reinforcement learning (RL) formalisms use predictions, actions, and policies to maximize reward. Midbrain dopamine neurons code reward prediction errors (RPE) of subjective reward value suitable for RL. Electrical and optogenetic self-stimulation experiments demonstrate that monkeys and rodents repeat behaviors that result in dopamine excitation. Dopamine excitations reflect positive RPEs that increase reward predictions via RL; against increasing predictions, obtaining similar dopamine RPE signals again requires better rewards than before. The positive RPEs drive predictions higher again and thus advance a recursive reward-RPE-prediction iteration toward better and better rewards. Agents also avoid dopamine inhibitions that lower reward prediction via RL, which allows smaller rewards than before to elicit positive dopamine RPE signals and resume the iteration toward better rewards. In this way, dopamine RPE signals serve a causal mechanism that attracts agents via RL to the best rewards. The mechanism improves daily life and benefits evolutionary selection but may also induce restlessness and greed.

Mesolimbic dopamine ramps reflect environmental timescales

Mesolimbic dopamine activity occasionally exhibits ramping dynamics, reigniting debate on theories of dopamine signaling. This debate is ongoing partly because the experimental conditions under which dopamine ramps emerge remain poorly understood. Here, we show that during Pavlovian and instrumental conditioning, mesolimbic dopamine ramps are only observed when the inter-trial interval is short relative to the trial period. These results constrain theories of dopamine signaling and identify a critical variable determining the emergence of dopamine ramps.

More widespread and rigid neuronal representation of reward expectation underlies impulsive choices

Impulsive choices prioritize smaller, more immediate rewards over larger, delayed, or potentially uncertain rewards. Impulsive choices are a critical aspect of substance use disorders and maladaptive decision-making across the lifespan. Here, we sought to understand the neuronal underpinnings of expected reward and risk estimation on a trial-by-trial basis during impulsive choices. To do so, we acquired electrical recordings from the human brain while participants carried out a risky decision-making task designed to measure choice impulsivity. Behaviorally, we found a reward-accuracy tradeoff, whereby more impulsive choosers were more accurate at the task, opting for a more immediate reward while compromising overall task performance. We then examined how neuronal populations across frontal, temporal, and limbic brain regions parametrically encoded reinforcement learning model variables, namely reward and risk expectation and surprise, across trials. We found more widespread representations of reward value expectation and prediction error in more impulsive choosers, whereas less impulsive choosers preferentially represented risk expectation. A regional analysis of reward and risk encoding highlighted the anterior cingulate cortex for value expectation, the anterior insula for risk expectation and surprise, and distinct regional encoding between impulsivity groups. Beyond describing trial-by-trial population neuronal representations of reward and risk variables, these results suggest impaired inhibitory control and model-free learning underpinnings of impulsive choice. These findings shed light on neural processes underlying reinforced learning and decision-making in uncertain environments and how these processes may function in psychiatric disorders.

Local regulation of striatal dopamine: A diversity of circuit mechanisms for a diversity of behavioral functions?

Striatal dopamine governs a wide range of behavioral functions, yet local dopamine concentrations can be dissociated from somatic activity. Here, we discuss how dopamine's diverse roles in behavior may be driven by local circuit mechanisms shaping dopamine release. We first look at historical and recent work demonstrating that striatal circuits interact with dopaminergic terminals to either initiate the release of dopamine or modulate the release of dopamine initiated by spiking in midbrain dopamine neurons, with particular attention to GABAergic and cholinergic local circuit mechanisms. Then we discuss some of the first in vivo studies of acetylcholine-dopamine interactions in striatum and broadly discuss necessary future work in understanding the roles of midbrain versus striatal dopamine regulation.

Prediction error in dopamine neurons during associative learning

Dopamine neurons have long been thought to facilitate learning by broadcasting reward prediction error (RPE), a teaching signal used in machine learning, but more recent work has advanced alternative models of dopamine's computational role. Here, I revisit this critical issue and review new experimental evidences that tighten the link between dopamine activity and RPE. First, I introduce the recent observation of a gradual backward shift of dopamine activity that had eluded researchers for over a decade. I also discuss several other findings, such as dopamine ramping, that were initially interpreted to conflict but later found to be consistent with RPE. These findings improve our understanding of neural computation in dopamine neurons.

The neural correlates of continuous feedback processing

Feedback processing is commonly studied by analyzing the brain's response to discrete rather than continuous events. Such studies have led to the hypothesis that rapid phasic midbrain dopaminergic activity tracks reward prediction errors (RPEs), the effects of which are measurable at the scalp via electroencephalography (EEG). Although studies using continuous feedback are sparse, recent animal work suggests that moment-to-moment changes in reward are tracked by slowly ramping midbrain dopaminergic activity. Some have argued that these ramping signals index state values rather than RPEs. Our goal here was to develop an EEG measure of continuous feedback processing in humans, then test whether its behavior could be accounted for by the RPE hypothesis. Participants completed a stimulus-response learning task in which a continuous reward cue gradually increased or decreased over time. A regression-based unmixing approach revealed EEG activity with a topography and time course consistent with the stimulus-preceding negativity (SPN), a scalp potential previously linked to reward anticipation and tonic dopamine release. Importantly, this reward-related activity depended on outcome expectancy: as predicted by the RPE hypothesis, activity for expected reward cues was reduced compared to unexpected reward cues. These results demonstrate the possibility of using human scalp-recorded potentials to track continuous feedback processing, and test candidate hypotheses of this activity.

Multi-timescale reinforcement learning in the brain

To thrive in complex environments, animals and artificial agents must learn to act adaptively to maximize fitness and rewards. Such adaptive behavior can be learned through reinforcement learning1, a class of algorithms that has been successful at training artificial agents2-6 and at characterizing the firing of dopamine neurons in the midbrain7-9. In classical reinforcement learning, agents discount future rewards exponentially according to a single time scale, controlled by the discount factor. Here, we explore the presence of multiple timescales in biological reinforcement learning. We first show that reinforcement agents learning at a multitude of timescales possess distinct computational benefits. Next, we report that dopamine neurons in mice performing two behavioral tasks encode reward prediction error with a diversity of discount time constants. Our model explains the heterogeneity of temporal discounting in both cue-evoked transient responses and slower timescale fluctuations known as dopamine ramps. Crucially, the measured discount factor of individual neurons is correlated across the two tasks suggesting that it is a cell-specific property. Together, our results provide a new paradigm to understand functional heterogeneity in dopamine neurons, a mechanistic basis for the empirical observation that humans and animals use non-exponential discounts in many situations10-14, and open new avenues for the design of more efficient reinforcement learning algorithms.

Dual credit assignment processes underlie dopamine signals in a complex spatial environment

Animals frequently make decisions based on expectations of future reward ("values"). Values are updated by ongoing experience: places and choices that result in reward are assigned greater value. Yet, the specific algorithms used by the brain for such credit assignment remain unclear. We monitored accumbens dopamine as rats foraged for rewards in a complex, changing environment. We observed brief dopamine pulses both at reward receipt (scaling with prediction error) and at novel path opportunities. Dopamine also ramped up as rats ran toward reward ports, in proportion to the value at each location. By examining the evolution of these dopamine place-value signals, we found evidence for two distinct update processes: progressive propagation of value along taken paths, as in temporal difference learning, and inference of value throughout the maze, using internal models. Our results demonstrate that within rich, naturalistic environments dopamine conveys place values that are updated via multiple, complementary learning algorithms.

Striatal Dopamine Signals and Reward Learning

We are constantly bombarded by sensory information and constantly making decisions on how to act. In order to optimally adapt behavior, we must judge which sequences of sensory inputs and actions lead to successful outcomes in specific circumstances. Neuronal circuits of the basal ganglia have been strongly implicated in action selection, as well as the learning and execution of goal-directed behaviors, with accumulating evidence supporting the hypothesis that midbrain dopamine neurons might encode a reward signal useful for learning. Here, we review evidence suggesting that midbrain dopaminergic neurons signal reward prediction error, driving synaptic plasticity in the striatum underlying learning. We focus on phasic increases in action potential firing of midbrain dopamine neurons in response to unexpected rewards. These dopamine neurons prominently innervate the dorsal and ventral striatum. In the striatum, the released dopamine binds to dopamine receptors, where it regulates the plasticity of glutamatergic synapses. The increase of striatal dopamine accompanying an unexpected reward activates dopamine type 1 receptors (D1Rs) initiating a signaling cascade that promotes long-term potentiation of recently active glutamatergic input onto striatonigral neurons. Sensorimotor-evoked glutamatergic input, which is active immediately before reward delivery will thus be strengthened onto neurons in the striatum expressing D1Rs. In turn, these neurons cause disinhibition of brainstem motor centers and disinhibition of the motor thalamus, thus promoting motor output to reinforce rewarded stimulus-action outcomes. Although many details of the hypothesis need further investigation, altogether, it seems likely that dopamine signals in the striatum might underlie important aspects of goal-directed reward-based learning.

Learning to Express Reward Prediction Error-like Dopaminergic Activity Requires Plastic Representations of Time

The dominant theoretical framework to account for reinforcement learning in the brain is temporal difference (TD) reinforcement learning. The TD framework predicts that some neuronal elements should represent the reward prediction error (RPE), which means they signal the difference between the expected future rewards and the actual rewards. The prominence of the TD theory arises from the observation that firing properties of dopaminergic neurons in the ventral tegmental area appear similar to those of RPE model-neurons in TD learning. Previous implementations of TD learning assume a fixed temporal basis for each stimulus that might eventually predict a reward. Here we show that such a fixed temporal basis is implausible and that certain predictions of TD learning are inconsistent with experiments. We propose instead an alternative theoretical framework, coined FLEX (Flexibly Learned Errors in Expected Reward). In FLEX, feature specific representations of time are learned, allowing for neural representations of stimuli to adjust their timing and relation to rewards in an online manner. In FLEX dopamine acts as an instructive signal which helps build temporal models of the environment. FLEX is a general theoretical framework that has many possible biophysical implementations. In order to show that FLEX is a feasible approach, we present a specific biophysically plausible model which implements the principles of FLEX. We show that this implementation can account for various reinforcement learning paradigms, and that its results and predictions are consistent with a preponderance of both existing and reanalyzed experimental data.

Emergence of belief-like representations through reinforcement learning

To behave adaptively, animals must learn to predict future reward, or value. To do this, animals are thought to learn reward predictions using reinforcement learning. However, in contrast to classical models, animals must learn to estimate value using only incomplete state information. Previous work suggests that animals estimate value in partially observable tasks by first forming "beliefs"-optimal Bayesian estimates of the hidden states in the task. Although this is one way to solve the problem of partial observability, it is not the only way, nor is it the most computationally scalable solution in complex, real-world environments. Here we show that a recurrent neural network (RNN) can learn to estimate value directly from observations, generating reward prediction errors that resemble those observed experimentally, without any explicit objective of estimating beliefs. We integrate statistical, functional, and dynamical systems perspectives on beliefs to show that the RNN's learned representation encodes belief information, but only when the RNN's capacity is sufficiently large. These results illustrate how animals can estimate value in tasks without explicitly estimating beliefs, yielding a representation useful for systems with limited capacity.

Impulse control disorder in Parkinson's disease is associated with abnormal frontal value signalling

Dopaminergic medication is well established to boost reward- versus punishment-based learning in Parkinson's disease. However, there is tremendous variability in dopaminergic medication effects across different individuals, with some patients exhibiting much greater cognitive sensitivity to medication than others. We aimed to unravel the mechanisms underlying this individual variability in a large heterogeneous sample of early-stage patients with Parkinson's disease as a function of comorbid neuropsychiatric symptomatology, in particular impulse control disorders and depression. One hundred and ninety-nine patients with Parkinson's disease (138 ON medication and 61 OFF medication) and 59 healthy controls were scanned with functional MRI while they performed an established probabilistic instrumental learning task. Reinforcement learning model-based analyses revealed medication group differences in learning from gains versus losses, but only in patients with impulse control disorders. Furthermore, expected-value related brain signalling in the ventromedial prefrontal cortex was increased in patients with impulse control disorders ON medication compared with those OFF medication, while striatal reward prediction error signalling remained unaltered. These data substantiate the hypothesis that dopamine's effects on reinforcement learning in Parkinson's disease vary with individual differences in comorbid impulse control disorder and suggest they reflect deficient computation of value in medial frontal cortex, rather than deficient reward prediction error signalling in striatum. See Michael Browning (https://doi.org/10.1093/brain/awad248) for a scientific commentary on this article.

Enhanced locomotor behaviour is mediated by activation of tyrosine hydroxylase in the silkworm brain

Animal behaviour regulation is a complex process involving many factors, and the nervous system is an essential factor in this process. In many species, pathogens can alter host behaviour by affecting the host's nervous system. An interesting example is that the silkworm shows enhanced locomotor behaviour after being infected with the nucleopolyhedrosis virus. In this study, we analysed the transcriptome of the silkworm brain at different time points after infection and found that various genes related to behaviour regulation changed after infection. In-depth analysis showed that the tyrosine hydroxylase gene might be a key candidate gene, and the content of dopamine, its downstream metabolite, increased significantly in the brain of silkworms infected with the virus. After the injection of tyrosine hydroxylase inhibitor into the infected silkworm, the dopamine content in the silkworm brain decreased and the locomotor behaviour caused by the virus was blocked successfully. These results confirm that tyrosine hydroxylase is involved in regulating enhanced locomotor behaviour after virus infection in silkworms. Furthermore, the tyrosine hydroxylase gene was specifically overexpressed in the brain of the silkworm, and the transgenic silkworm was enhanced in locomotor behaviour and foraging behaviour. These results suggest that the tyrosine hydroxylase gene plays a vital role in regulating insect behaviour.

Dissociation and integration of outcome and state uncertainty signals in cognitive control

Signals related to uncertainty are frequently observed in regions of the cognitive control network, including anterior cingulate/medial prefrontal cortex (ACC/mPFC), dorsolateral prefrontal cortex (dlPFC), and anterior insular cortex. Uncertainty generally refers to conditions in which decision variables may assume multiple possible values and can arise at multiple points in the perception-action cycle, including sensory input, inferred states of the environment, and the consequences of actions. These sources of uncertainty are frequently correlated: noisy input can lead to unreliable estimates of the state of the environment, with consequential influences on action selection. Given this correlation amongst various sources of uncertainty, dissociating the neural structures underlying their estimation presents an ongoing issue: a region associated with uncertainty related to outcomes may estimate outcome uncertainty itself, or it may reflect a cascade effect of state uncertainty on outcome estimates. In this study, we derive signals of state and outcome uncertainty from mathematical models of risk and observe regions in the cognitive control network whose activity is best explained by signals related to state uncertainty (anterior insula), outcome uncertainty (dlPFC), as well as regions that appear to integrate the two (ACC/mPFC).

Emergence of belief-like representations through reinforcement learning

To behave adaptively, animals must learn to predict future reward, or value. To do this, animals are thought to learn reward predictions using reinforcement learning. However, in contrast to classical models, animals must learn to estimate value using only incomplete state information. Previous work suggests that animals estimate value in partially observable tasks by first forming "beliefs"-optimal Bayesian estimates of the hidden states in the task. Although this is one way to solve the problem of partial observability, it is not the only way, nor is it the most computationally scalable solution in complex, real-world environments. Here we show that a recurrent neural network (RNN) can learn to estimate value directly from observations, generating reward prediction errors that resemble those observed experimentally, without any explicit objective of estimating beliefs. We integrate statistical, functional, and dynamical systems perspectives on beliefs to show that the RNN's learned representation encodes belief information, but only when the RNN's capacity is sufficiently large. These results illustrate how animals can estimate value in tasks without explicitly estimating beliefs, yielding a representation useful for systems with limited capacity.

Author Summary

Natural environments are full of uncertainty. For example, just because my fridge had food in it yesterday does not mean it will have food today. Despite such uncertainty, animals can estimate which states and actions are the most valuable. Previous work suggests that animals estimate value using a brain area called the basal ganglia, using a process resembling a reinforcement learning algorithm called TD learning. However, traditional reinforcement learning algorithms cannot accurately estimate value in environments with state uncertainty (e.g., when my fridge's contents are unknown). One way around this problem is if agents form "beliefs," a probabilistic estimate of how likely each state is, given any observations so far. However, estimating beliefs is a demanding process that may not be possible for animals in more complex environments. Here we show that an artificial recurrent neural network (RNN) trained with TD learning can estimate value from observations, without explicitly estimating beliefs. The trained RNN's error signals resembled the neural activity of dopamine neurons measured during the same task. Importantly, the RNN's activity resembled beliefs, but only when the RNN had enough capacity. This work illustrates how animals could estimate value in uncertain environments without needing to first form beliefs, which may be useful in environments where computing the true beliefs is too costly.

Dual credit assignment processes underlie dopamine signals in a complex spatial environment

Dopamine in the nucleus accumbens helps motivate behavior based on expectations of future reward ("values"). These values need to be updated by experience: after receiving reward, the choices that led to reward should be assigned greater value. There are multiple theoretical proposals for how this credit assignment could be achieved, but the specific algorithms that generate updated dopamine signals remain uncertain. We monitored accumbens dopamine as freely behaving rats foraged for rewards in a complex, changing environment. We observed brief pulses of dopamine both when rats received reward (scaling with prediction error), and when they encountered novel path opportunities. Furthermore, dopamine ramped up as rats ran towards reward ports, in proportion to the value at each location. By examining the evolution of these dopamine place-value signals, we found evidence for two distinct update processes: progressive propagation along taken paths, as in temporal-difference learning, and inference of value throughout the maze, using internal models. Our results demonstrate that within rich, naturalistic environments dopamine conveys place values that are updated via multiple, complementary learning algorithms.

Opponent Learning with Different Representations in the Cortico-Basal Ganglia Circuits

The direct and indirect pathways of the basal ganglia (BG) have been suggested to learn mainly from positive and negative feedbacks, respectively. Since these pathways unevenly receive inputs from different cortical neuron types and/or regions, they may preferentially use different state/action representations. We explored whether such a combined use of different representations, coupled with different learning rates from positive and negative reward prediction errors (RPEs), has computational benefits. We modeled animal as an agent equipped with two learning systems, each of which adopted individual representation (IR) or successor representation (SR) of states. With varying the combination of IR or SR and also the learning rates from positive and negative RPEs in each system, we examined how the agent performed in a dynamic reward navigation task. We found that combination of SR-based system learning mainly from positive RPEs and IR-based system learning mainly from negative RPEs could achieve a good performance in the task, as compared with other combinations. In such a combination of appetitive SR-based and aversive IR-based systems, both systems show activities of comparable magnitudes with opposite signs, consistent with the suggested profiles of the two BG pathways. Moreover, the architecture of such a combination provides a novel coherent explanation for the functional significance and underlying mechanism of diverse findings about the cortico-BG circuits. These results suggest that particularly combining different representations with appetitive and aversive learning could be an effective learning strategy in certain dynamic environments, and it might actually be implemented in the cortico-BG circuits.

Mesolimbic dopamine adapts the rate of learning from action

Recent success in training artificial agents and robots derives from a combination of direct learning of behavioural policies and indirect learning through value functions^1-3. Policy learning and value learning use distinct algorithms that optimize behavioural performance and reward prediction, respectively. In animals, behavioural learning and the role of mesolimbic dopamine signalling have been extensively evaluated with respect to reward prediction⁴; however, so far there has been little consideration of how direct policy learning might inform our understanding⁵. Here we used a comprehensive dataset of orofacial and body movements to understand how behavioural policies evolved as naive, head-restrained mice learned a trace conditioning paradigm. Individual differences in initial dopaminergic reward responses correlated with the emergence of learned behavioural policy, but not the emergence of putative value encoding for a predictive cue. Likewise, physiologically calibrated manipulations of mesolimbic dopamine produced several effects inconsistent with value learning but predicted by a neural-network-based model that used dopamine signals to set an adaptive rate, not an error signal, for behavioural policy learning. This work provides strong evidence that phasic dopamine activity can regulate direct learning of behavioural policies, expanding the explanatory power of reinforcement learning models for animal learning⁶.

Dopamine mediates the bidirectional update of interval timing

The role of dopamine (DA) as a reward prediction error (RPE) signal in reinforcement learning (RL) tasks has been well-established over the past decades. Recent work has shown that the RPE interpretation can also account for the effects of DA on interval timing by controlling the speed of subjective time. According to this theory, the timing of the dopamine signal relative to reward delivery dictates whether subjective time speeds up or slows down: Early DA signals speed up subjective time and late signals slow it down. To test this bidirectional prediction, we reanalyzed measurements of dopaminergic neurons in the substantia nigra pars compacta of mice performing a self-timed movement task. Using the slope of ramping dopamine activity as a readout of subjective time speed, we found that trial-by-trial changes in the slope could be predicted from the timing of dopamine activity on the previous trial. This result provides a key piece of evidence supporting a unified computational theory of RL and interval timing. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

A unidirectional but not uniform striatal landscape of dopamine signaling for motivational stimuli

Dopamine signals in the striatum are critical for motivated behavior. However, their regional specificity and precise information content are actively debated. Dopaminergic projections to the striatum are topographically organized. Thus, we quantified dopamine release in response to motivational stimuli and associated predictive cues in six principal striatal regions of unrestrained, behaving rats. Absolute signal size and its modulation by stimulus value and by subjective state of the animal were interregionally heterogeneous on a medial to lateral gradient. In contrast, dopamine-concentration direction of change was homogeneous across all regions: appetitive stimuli increased and aversive stimuli decreased dopamine concentration. Although cues predictive of such motivational stimuli acquired the same influence over dopamine homogeneously across all regions, dopamine-mediated prediction-error signals were restricted to the ventromedial, limbic striatum. Together, our findings demonstrate a nuanced striatal landscape of unidirectional but not uniform dopamine signals, topographically encoding distinct aspects of motivational stimuli and their prediction.

Dopamine ramps for accurate value learning under uncertainty

Dopamine signals ramping towards reward timings have become widely reported, but their functions remain elusive. Through modeling analyses and experiments in mice, a recent study by Mikhael, Kim et al. shows that such signals represent reward prediction errors used for accurate value learning in conditions with uncertainty about upcoming state and its resolution by sensory feedback.