A systems-neuroscience model of phasic dopamine

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.

Prediction error in models of adaptive behavior

Pavlovian conditioning is evident in every species in which it has been assessed, and there is a consensus about its interpretation across behavioral,1,2 brain,3,4,5,6 and computational analyses7,8,9,10,11: conditioned behavior reflects the formation of a directional associative link from the memory of one stimulus (e.g., a visual stimulus) to another (e.g., food), with learning stopping when there is no error between the prediction generated by the visual stimulus and what happens next (e.g., food). This consensus fails to anticipate the results that we report here. In our experiments with rats, we find that arranging predictive (visual stimulus→food) and nonpredictive (food→visual stimulus) relationships produces marked and sustained changes in conditioned behaviors when the visual stimulus is presented alone. Moreover, the type of relationship affects (1) the distribution of conditioned behaviors related to the properties of both food (called goal-tracking) and the visual stimulus (called sign-tracking) and (2) when in the visual stimulus, these two behaviors are evident. These results represent an impetus for a fundamental shift in how Pavlovian conditioning is interpreted: animals learn about the relationship between two stimuli irrespective of the order in which they are presented, but they exhibit this knowledge in different ways. This interpretation and our new results are captured by a recent model of Pavlovian conditioning,12,13 HeiDI, and both are consistent with the need for animals to represent the fact that the impact of a cause (e.g., the ingestion of nutrients or the bite of a predator) can be felt before or after the cause has been perceived.

Striatal μ-opioid receptor activation triggers direct-pathway GABAergic plasticity and induces negative affect

Withdrawal from chronic opioid use often causes hypodopaminergic states and negative affect, which may drive relapse. Direct-pathway medium spiny neurons (dMSNs) in the striatal patch compartment contain μ-opioid receptors (MORs). It remains unclear how chronic opioid exposure and withdrawal impact these MOR-expressing dMSNs and their outputs. Here, we report that MOR activation acutely suppressed GABAergic striatopallidal transmission in habenula-projecting globus pallidus neurons. Notably, withdrawal from repeated morphine or fentanyl administration potentiated this GABAergic transmission. Furthermore, intravenous fentanyl self-administration enhanced GABAergic striatonigral transmission and reduced midbrain dopaminergic activity. Fentanyl-activated striatal neurons mediated contextual memory retrieval required for conditioned place preference tests. Importantly, chemogenetic inhibition of striatal MOR+ neurons rescued fentanyl withdrawal-induced physical symptoms and anxiety-like behaviors. These data suggest that chronic opioid use triggers GABAergic striatopallidal and striatonigral plasticity to induce a hypodopaminergic state, which may promote negative emotions and relapse.

A gradual temporal shift of dopamine responses mirrors the progression of temporal difference error in machine learning

A large body of evidence has indicated that the phasic responses of midbrain dopamine neurons show a remarkable similarity to a type of teaching signal (temporal difference (TD) error) used in machine learning. However, previous studies failed to observe a key prediction of this algorithm: that when an agent associates a cue and a reward that are separated in time, the timing of dopamine signals should gradually move backward in time from the time of the reward to the time of the cue over multiple trials. Here we demonstrate that such a gradual shift occurs both at the level of dopaminergic cellular activity and dopamine release in the ventral striatum in mice. Our results establish a long-sought link between dopaminergic activity and the TD learning algorithm, providing fundamental insights into how the brain associates cues and rewards that are separated in time.

"Liking" as an early and editable draft of long-run affective value

Psychological and neural distinctions between the technical concepts of "liking" and "wanting" pose important problems for motivated choice for goods. Why could we "want" something that we do not "like," or "like" something but be unwilling to exert effort to acquire it? Here, we suggest a framework for answering these questions through the medium of reinforcement learning. We consider "liking" to provide immediate, but preliminary and ultimately cancellable, information about the true, long-run worth of a good. Such initial estimates, viewed through the lens of what is known as potential-based shaping, help solve the temporally complex learning problems faced by animals.

Higher-Order Conditioning: What Is Learnt and How it Is Expressed

Pairing a neutral conditioned stimulus (CS) with a motivationally significant unconditioned stimulus (US) results in the CS coming to elicit conditioned responses (CRs). The widespread significance and translational value of Pavlovian conditioning are increased by the fact that pairing two neutral CSs (A and X) enables conditioning with X to affect behavior to A. There are two traditional informal accounts of such higher-order conditioning, which build on more formal associative analyses of Pavlovian conditioning. But, higher-order conditioning and Pavlovian conditioning have characteristics that are beyond these accounts: Notably, the two are influenced in different ways by the same experimental manipulations, and both generate conditioned responses that do not reflect the US per se. Here, we present a formal analysis that sought to address these characteristics.

Biological constraints on neural network models of cognitive function

Neural network models are potential tools for improving our understanding of complex brain functions. To address this goal, these models need to be neurobiologically realistic. However, although neural networks have advanced dramatically in recent years and even achieve human-like performance on complex perceptual and cognitive tasks, their similarity to aspects of brain anatomy and physiology is imperfect. Here, we discuss different types of neural models, including localist, auto-associative, hetero-associative, deep and whole-brain networks, and identify aspects under which their biological plausibility can be improved. These aspects range from the choice of model neurons and of mechanisms of synaptic plasticity and learning to implementation of inhibition and control, along with neuroanatomical properties including areal structure and local and long-range connectivity. We highlight recent advances in developing biologically grounded cognitive theories and in mechanistically explaining, on the basis of these brain-constrained neural models, hitherto unaddressed issues regarding the nature, localization and ontogenetic and phylogenetic development of higher brain functions. In closing, we point to possible future clinical applications of brain-constrained modelling.

The Role of the Lateral Habenula in Inhibitory Learning from Reward Omission

The lateral habenula (LHb) is a phylogenetically primitive brain structure that plays a key role in learning to inhibit distinct responses to specific stimuli. This structure is activated by primary aversive stimuli, cues predicting an imminent aversive event, unexpected reward omissions, and cues associated with the omission of an expected reward. The most widely described physiological effect of LHb activation is acutely suppressing midbrain dopaminergic signaling. However, recent studies have identified multiple means by which the LHb promotes this effect as well as other mechanisms of action. These findings reveal the complex nature of LHb circuitry. The present paper reviews the role of this structure in learning from reward omission. We approach this topic from the perspective of computational models of behavioral change that account for inhibitory learning to frame key findings. Such findings are drawn from recent behavioral neuroscience studies that use novel brain imaging, stimulation, ablation, and reversible inactivation techniques. Further research and conceptual work are needed to clarify the nature of the mechanisms related to updating motivated behavior in which the LHb is involved. As yet, there is little understanding of whether such mechanisms are parallel or complementary to the well-known modulatory function of the more recently evolved prefrontal cortex.

The Neural Correlates of Cued Reward Omission

Compared to our understanding of positive prediction error signals occurring due to unexpected reward outcomes, less is known about the neural circuitry in humans that drives negative prediction errors during omission of expected rewards. While classical learning theories such as Rescorla-Wagner or temporal difference learning suggest that both types of prediction errors result from a simple subtraction, there has been recent evidence suggesting that different brain regions provide input to dopamine neurons which contributes to specific components of this prediction error computation. Here, we focus on the brain regions responding to negative prediction error signals, which has been well-established in animal studies to involve a distinct pathway through the lateral habenula. We examine the activity of this pathway in humans, using a conditioned inhibition paradigm with high-resolution functional MRI. First, participants learned to associate a sensory stimulus with reward delivery. Then, reward delivery was omitted whenever this stimulus was presented simultaneously with a different sensory stimulus, the conditioned inhibitor (CI). Both reward presentation and the reward-predictive cue activated midbrain dopamine regions, insula and orbitofrontal cortex. While we found significant activity at an uncorrected threshold for the CI in the habenula, consistent with our predictions, it did not survive correction for multiple comparisons and awaits further replication. Additionally, the pallidum and putamen regions of the basal ganglia showed modulations of activity for the inhibitor that did not survive the corrected threshold.