Review: “The Disappointment Dilemma: Short‐ and Long‐Term Learning From Negative Outcomes”

ABSTRACT

In this review, we synthesize key findings from the 2025 Spanish Society for Neuroscience (SENC) symposium “Understanding Habenulae Function in Emotional Behavior” and propose that the brain possesses a “disappointment circuit” that complements the classic dopaminergic reward system. The “disappointment dilemma” in financial investing—avoiding risk to prevent short‐term disappointment, only to later experience greater long‐term disappointment when safe choices underperform—illustrates how immediate and delayed negative outcomes can be traded off over time. We argue that this dilemma parallels partially specialized roles of the lateral habenula (LHb), which signals immediate disappointments as acute negative prediction errors, and the medial habenula (MHb), which shapes longer‐lasting negative expectations or mood through the accumulation of disappointing experiences over time. We propose that LHb outputs implement fast, trial‐by‐trial adjustments in coping strategies via RMTg‐ and VTA‐projecting pathways, whereas MHb–IPN circuits gradually reshape threat valuation, withdrawal states, and mood across repeated disappointing or stressful experiences. We discuss how LHb and MHb circuits contribute to short‐ and long‐term disappointment by integrating information on stress resilience, aversive learning, threat and fear responses, drug‐related behaviors, and mood adaptation from specific cell types and receptors that modulate fear, addiction, and motivation to circuit‐level mechanisms underlying aversive learning and emotional responses.

The habenula routes negative‐outcome signals to VTA, RMTg, and IPN to shape aversive learning and motivation. The lateral habenula encodes acute negative prediction errors to rapidly adapt behavior to threat, fear, and avoidance. The medial habenula integrates repeated stress, drug, and aversive experiences into long‐term negative mood and expectations. Together, they form a “disappointment circuit” balancing short‐ and long‐term learning from negative outcomes, complementing dopaminergic reward circuits.

Abbreviations

Acc

nucleus accumbens

BDNF

brain‐derived neurotrophic factor

CB1

cannabinoid receptor type 1

CRF

corticotropin releasing factor

ECS

endocannabinoid system

EPSP

excitatory postsynaptic current

FC

fear conditioning

GP

globus pallidus

IA

inhibitory avoidance

IPN

interpeduncular nucleus

LDTg

laterodorsal tegmental nucleus

LHb

lateral habenula

MHb

medial habenula

MOR

Mu opioid receptors

nAChR

nicotine acetylcholine receptors

RMTg

rostromedial tegmental nucleus

RPE

reward prediction error

S

septum

SN

substantia nigra

Sst

somatostatin

VGLUT

vesicular glutamate transporter

VTA

ventral tegmental area

1. Introduction

It was not until 2007 that researchers discovered that, rather than recruiting classic reward centers, the brain engages a different hub—the habenula—when outcomes are worse than expected, with habenula neurons firing robustly when rewards are omitted or when unexpected negative events occur (Matsumoto and Hikosaka 2007). In contrast, the classic reward centers were discovered much earlier, beginning with 1950s brain‐stimulation experiments showing that animals would repeatedly self‐stimulate specific brain regions to obtain pleasurable outcomes (Olds and Milner 1954). Subsequent pharmacological and anatomical studies in the 1960s–1970s established dopamine pathways—particularly the mesolimbic system—as central to reward processing and led to the formal concept of a “dopaminergic reward system” (Figure 1).

FIGURE 1.

In the reward/antireward circuitry, dopaminergic pathways originating from the ventral tegmental area (VTA) and substantia nigra (SN) signal reward and better‐than‐expected outcomes to the striatum and cortex, driving motivation and reinforcing learning. The lateral (LHb) and medial (MHb) habenulea integrate information about negative outcomes and disappointments and, via the fasciculus retroflexus (f. retroflexus), send directly or indirectly potent inhibitory signals to dopamine neurons and modulate serotonin pathways. This circuit thereby dynamically tunes whether dopamine activity is suppressed or maintained in response to ongoing outcomes. Acc = nucleus accumbens in ventral striatum, GP = globus pallidus, IPN = interpeduncular nucleus, RMTg = rostromedial tegmental nucleus, S = septum, Striatum (C) = caudatum in dorsomedial striatum, striatum (P) = putamen in dorsal striatum. For reviews of habenular circuitry, see Ciscato et al. (2025), Ables et al. (2023), Boulos et al. (2017), Proulx et al. (2014), and Loonen (2023).

The habenula is an evolutionarily ancient epithalamic hub that emerged in vertebrates more than 360 million years ago and relays information from diverse forebrain regions to midbrain monoaminergic centers. Its circuitry, including major afferent and efferent pathways, has been extensively reviewed (Ables et al. 2023; Boulos et al. 2017; Mathis and Kenny 2019; Molas, DeGroot, et al. 2017; Proulx et al. 2014; Ciscato et al. 2025). Despite its small size, it contains densely packed neurons and exceptionally high levels of neuromodulatory receptors, positioning it as a key node for encoding negative prediction errors and regulating coping strategies across stress, pain, and addiction paradigms. Within this architecture, the medial habenula–interpeduncular nucleus (MHb–IPN) pathway forms a parallel anti‐reward axis to the lateral habenula–rostromedial tegmental nucleus/ventral tegmental area (LHb–RMTg/VTA) routes using distinct molecular receptors to shape aversion, withdrawal, and drug intake (Mathis and Kenny 2019; Ables et al. 2023; Ciscato et al. 2025).

In the context of the 2025 Spanish Society for Neuroscience (SENC) symposium, this review proposes that the dopaminergic reward system and the habenula “disappointment system” constitute complementary circuits for encoding positive rewards and negative prediction errors over different timescales. It further argues that the brain is equipped with a dedicated disappointment circuit, consisting of two parallel pathways—the LHb and MHb circuits that detect acute versus sustained aversive signaling and work together to process negative outcomes and integrate them into behavioral adaptation and mood regulation. To illustrate these mechanisms, we present key findings from the SENC symposium, “Understanding Habenulae Function in Emotional Behavior.” The symposium presented some recent advances on the circuits and receptor mechanisms of the MHb and LHb in learned and innate fear, as well as drug‐related behaviors. Recent reviews, including work in EJN by Ciscato et al. 2025 and others on habenular receptor pharmacology and MHb–IPN nicotinic mechanisms, have provided in‐depth coverage of the molecular and synaptic organization of this circuit. In contrast, this article focuses on the broader concept of a “disappointment circuit” and jointly considers MHb and LHb outputs across threat processing, aversive learning, addiction, and mood‐related behavior.

We first summarize symposium contributions on how molecularly defined receptors, including nicotinic acetylcholine receptors (nAChRs), CB1 cannabinoid receptors, mu opioid receptors, and the habenula‐enriched GPCRs GPR151 and GPR139, organize computations in MHb–IPN synapses and mediate nicotine‐ and opioid‐related negative affect and withdrawal. We then review presentations showing how MHb–IPN circuits link peripheral immune and metabolic signals, endocannabinoid modulation, and adaptive threat processing to the accumulation of aversive memories and long‐term negative states. A final set of talks, together with additional recent work, focused on how segregated LHb outputs to RMTg and VTA gate passive versus active coping, regulate stress susceptibility, and implement fast adjustments in behavioral strategies after disappointing outcomes and how LHb‐generated aversive signals regulate the formation, persistence, and retrieval of fear memories in inhibitory avoidance and Pavlovian conditioning paradigms. Building on these presentations, we propose that MHb and LHb together form an integrated “disappointment system” encoding worse‐than‐expected outcomes across timescales and point to specific cells, receptors, and projections as potential circuit‐selective therapeutic targets (Table 1).

TABLE 1.

Complementary roles of reward and disappointment circuits.

Function/domain	Dopaminergic pathway (reward axis)	Habenula pathway (disappointment axis)
Outcome evaluation	Encodes better‐than‐expected outcomes and reward value	Encodes worse‐than‐expected outcomes and reward omission
Learning/motivation	Supports reward learning and reinforcement of positive cues and actions increasing pursuit of valued goals	Supports aversive learning and avoidance of cues predicting negative events; suppresses motivation under repeated failure
Mood/addiction	Supports positive affect and enjoyment of rewards; mediates drug reward, craving and compulsive seeking	Drives despair, anhedonia and anxiety; encodes withdrawal and punishment‐like aspects of drug use

2. Molecular Modulators of the Habenula: From Classical Receptors to Orphan GPCRs

The MHb–IPN pathway contains exceptionally high levels of nAChRs, especially receptors containing the α3, α5, and β4 subunits. These receptors drive nicotine‐evoked firing and transmitter release (Boulter et al. 1990; Quick et al. 1999; Shih et al. 2014) and make the habenula a major nodal point for nicotine dependence. Genetic and gain‐/loss‐of‐function studies of α3, α5, and β4 nAChR subunits show that modest changes in their expression shift the balance between nicotine intake and aversion, with increasing β4 strengthening nicotine‐evoked drive and aversion and loss or mutation of α5 weakening aversion and promoting compulsive use (Fowler et al. 2011; Frahm et al. 2011). This was initially unexpected because α4β2 and α7 receptors are far more ubiquitously expressed in the brain (Changeux et al. 1998), yet human genetic studies instead implicated the CHRNA5–CHRNA3–CHRNB4 locus in heavy smoking and nicotine dependence (Saccone et al. 2007), together defining the MHb–IPN circuit as a key determinant of the aversive limit on nicotine consumption (Antolin‐Fontes et al. 2015). In addition, MHb neurons express μ‐opioid receptors (MOR) (Gardon et al. 2014), CB1 cannabinoid receptors (CB1R) (Soria‐Gómez et al. 2015), and GPR139, an orphan GPCR recently proposed to act as a non‐canonical anti‐opioid receptor (Wang et al. 2019; Stoveken et al. 2020; Li et al. 2025). Together, these receptors allow opioids and cannabinoids and GPR139‐dependent signaling to strongly modulate glutamatergic and cholinergic output and thereby influence anxiety, negative affect, aversive learning, and relapse (Boulos et al. 2017; Antolin‐Fontes et al. 2020; Bailly et al. 2023; Chan and Ogawa 2025). Functionally, opioid and cannabinoid receptors couple to Gαi/o proteins to inhibit adenylyl cyclase (Liu et al. 2024), reduce presynaptic calcium influx (Wilson et al. 2001), and open potassium channels (Reeves et al. 2022), suppressing transmitter release and hyperpolarizing MHb/IPN neurons, whereas nAChRs are ligand‐gated cation channels that depolarize MHb cells and facilitate transmitter release into the IPN. Recent work shows that opioids and GABA_B receptors can paradoxically potentiate MHb–IPN glutamatergic and cholinergic transmission in a state‐ and input‐specific manner (Zhang et al. 2016; Koppensteiner et al. 2024; Singhal et al. 2025; Chittajallu et al. 2025), so that the same terminals co‐expressing nAChRs, MOR, CB1R, and GABA_B receptors become a shared presynaptic substrate where nicotine, opioids, cannabinoids, and GABAergic signals interact to shape computations related to disappointment, withdrawal, and relapse risk.

2.1. Pacemaking, Emergency Brakes, and IPN Microcircuits

Electrophysiological recordings reveal that MHb neurons behave as autonomous pacemakers that increase firing at low nicotine doses (Görlich et al. 2013) but, at higher concentrations, enter a long‐lasting refractory state driven by calcium‐activated channels and depolarization block, effectively engaging an intrinsic “emergency brake” (Kawai et al. 2025). This brake silences MHb output, prevents excitotoxicity, and transiently interrupts MHb → IPN transmission, providing a cellular explanation for behavioral avoidance at high nicotine doses. Cholinergic “habenular” neurons co‐release glutamate and acetylcholine (Ren et al. 2011), with acetylcholine enhancing vesicular glutamate loading and release, thereby boosting fast excitatory drive from MHb terminals to the IPN (Frahm et al. 2015). On the postsynaptic side, specific IPN neuron populations expressing α5‐containing nAChRs (Hsu et al. 2013), somatostatin and nitric oxide synthase (Ables et al. 2017), CRF receptors (Zhao‐Shea et al. 2015), and GABAergic markers (Zhao‐Shea et al. 2013) provide additional brakes and gating mechanisms that tune nicotine reward, anxiety, and coping behaviors during withdrawal (Molas, DeGroot, et al. 2017; Klenowski et al. 2022). Together, these data establish the MHb–IPN circuit as a hub where glutamatergic, cholinergic, GABAergic, peptidergic, and stress‐related signals are integrated into motivational output and negative affect across dependence and stress states.

2.2. GPR151: Orphan GPCR in MHb and LHb

Within this molecularly complex setting, the orphan GPCR GPR151 has emerged as a habenula‐enriched Gαi/o‐coupled receptor that integrates nicotine, opioid, and cannabinoid signals (Antolin‐Fontes et al. 2019, 2020). Anatomical and TRAP‐seq studies identify GPR151 as one of the most highly enriched GPCRs in cholinergic MHb neurons and their projections to the IPN and show that it is co‐expressed with α3β4 nAChRs, MOR, and CB1R on VGLUT‐positive presynaptic terminals (Ignatov et al. 2004; Broms et al. 2015; Antolin‐Fontes et al. 2020). GPR151 is also robustly expressed in lateral habenula neurons and their axons across species (Broms et al. 2015), reinforcing its designation as a habenula‐selective receptor (Kobayashi et al. 2013; DePasquale et al. 2025). At the cellular level, GPR151 activation couples to Gαi/o proteins to reduce cAMP, whereas loss of Gpr151 decreases miniature and spontaneous EPSC frequency, lowers synaptic fidelity during high‐frequency trains, and alters the readily releasable vesicle pool without major changes in nAChR or opioid/cannabinoid receptor expression levels (Antolin‐Fontes et al. 2019, 2020). In vivo, GPR151 deletion blunts neuronal and behavioral responses to nicotine and opioids, increases nicotine consumption, and reduces social reward, indicating that this receptor amplifies drug‐evoked signaling within MHb–IPN synapses while simultaneously constraining intake (Allain et al. 2022; Antolin‐Fontes et al. 2019, 2020).

2.3. Circuit‐Specific Therapeutic Opportunities

Because GPR151 expression is largely confined to habenular projection neurons, with additional expression in dorsal root ganglia and limited spinal populations (Holmes et al. 2017; Jiang et al. 2018; Xia et al. 2021; DePasquale et al. 2025), it represents an interesting cell‐type‐selective target for circuit‐based interventions that may carry fewer systemic side effects than drugs targeting broader nicotinic, opioid, or cannabinoid systems. A multidisciplinary collaboration between the Ibanez‐Tallon laboratory and partners at NCATS, Scripps, and the Icahn School of Medicine at Mount Sinai is using stable GPR151 expression systems to support high‐throughput screening for brain‐penetrant modulators, with the aim of testing whether agonists or inverse agonists can enhance aversion, reduce craving, or attenuate withdrawal‐related negative affect for nicotine and opioids and possibly influence neuropathic pain. A major open question is the identity of the endogenous ligand, if any, and how GPR151 signaling is regulated by chronic exposure to nicotine, opioids, cannabinoids, and stress, as well as how GPR151‐positive MHb/LHb outputs interact with broader midbrain and brainstem networks that control active versus passive coping. Addressing these issues will be essential for translating this habenula‐specific GPCR, and related molecular targets in the MHb–IPN axis, into viable therapeutic strategies for nicotine and opioid addiction. Related orphan receptors such as GPR139, which is enriched in MHb and striatum and implicated in the modulation of mu‐opioid receptor signaling and aversive states, together with GPR151, illustrate how habenula‐enriched orphan GPCRs may offer complementary circuit‐selective entry points to regulate opioid‐related disappointment and withdrawal without broadly suppressing opioid signaling.

3. The Medial Habenula: A Missing Link in the Periphery‐to‐Brain Control of Emotions

Anatomically, the MHb is critically positioned as an interface between brain regions coding stress signals and areas regulating emotional behavior. Moreover, the MHb is a subventricular structure that harbors specialized cells and vasculature with enhanced transport function, potentially sensitive to known peripheral modulators of behavioural outputs. The connection between the immune system and the central nervous system had remained unknown for centuries, but the discoveries in the field (e.g., the presence of CNS lymphatic vessels) (Louveau et al. 2015) challenge current theories of brain–body interactions. Accumulating evidence indicates that the immune system participates in the development of neurological diseases (Solomon 1987; Togo et al. 2002). Interestingly, the anatomical presence of immune‐related molecules in the habenular complex confers a special sensitivity to immunological insults (Le Foll and French 2018), but their functional impact remains to be dissected. Furthermore, MHb expresses a wide variety of food‐ and metabolic‐related signals, including leptin receptors and various peptides (Yang et al. 2018). Moreover, a type of ependymal cells resembling hypothalamic third‐ventricular tanycytes (Cupédo and Weerd 1985) has been described in the MHb. Hypothalamic tanycytes play a crucial role in the control of blood‐hypothalamus exchanges of peripheral signaling factors (Langlet et al. 2013) (e.g., glucose and leptin). Importantly, such transport is dynamically set according to the energy status. Therefore, molecules sensing the body energy balance could target the MHb, potentially affecting emotional regulation. Thus, the MHb is a legitimate candidate to serve as a functional hub linking peripheral and central signals that regulate emotional responses. However, the cell‐type‐specific control of emotional responses in the medial habenular circuits is largely unknown. We recently started shedding light on this, showing that the presynaptic control of neurotransmitter release in habenular circuits regulates the expression of aversive memories (Soria‐Gómez et al. 2015).

3.1. Cannabinoid Modulation of Habenular Circuits

In the brain, a primary modulatory system, the endocannabinoid system (ECS), participates in the control of neurotransmitter release and has a direct impact on animal behavior (Busquets‐Garcia et al. 2015). The ECS comprises endogenous ligands (anandamide and 2‐arachidonoylglycerol [2‐AG]), cannabinoid receptors CB1 and CB2, and enzymes controlling endocannabinoid synthesis, release, and degradation. This system is characterized as a retrograde feedback mechanism (i.e., inhibition of neurotransmitter release) and is known to be involved in cognitive processes (Robledo‐Menendez et al. 2022) and mood regulation (Loomba and Patel 2025). The CB1 receptor is widely expressed in different brain regions (including habenular circuits), cell types (e.g., neurons, astrocytes, and microglia) (Busquets‐Garcia et al. 2018), and subcellular compartments (e.g., mitochondria) (Benard et al. 2012). Thus, the CB1 receptor is an optimal tool for studying the complexity of the brain and for identifying novel molecular mechanisms linking brain cell‐type specificity to behavioural outputs. In particular, the CB1 in habenular circuits regulates the expression of aversive memories. Briefly, we demonstrated that mice carrying a CB1 deletion in MHb cells exhibit reduced freezing behaviour. Such a phenotype is associated with a specific regulation of cholinergic transmission in the IPN (Soria‐Gómez et al. 2015). This study highlighted the importance of the fine regulation of habenular circuits and their impact on aversive responses. Interestingly, the multi‐diverse cellular milieu that constitutes the MHb suggests the existence of more complex interactions and molecular mechanisms involved in emotional regulation under cannabinoid influence. Currently, the Soria‐Gomez lab aims at studying the role of cannabinoid signaling in glial cells of the medial habenula in the control of innate and learned emotional behavior in males and females.

4. Adaptive Threat Processing via MHb–IPN Circuits

Dysregulation of the MHb–IPN pathway has been implicated in several psychiatric disorders, including anxiety, depression, and substance use disorders. Early studies associated heightened MHb–IPN signaling with the affective components of nicotine withdrawal (Pang et al. 2016; Casserly et al. 2020; Klenowski et al. 2022; Ciscato et al. 2025). Subsequent investigations implicated this axis in anxiety (DeGroot et al. 2020; Molas et al. 2023) and the expression of associative fear memories (Soria‐Gómez et al. 2015; Zhang et al. 2016).

4.1. The IPN as a Central Hub for Processing the Negative Value of Threat Stimuli

Understanding how the MHb–IPN pathway processes stimuli of negative value to regulate affective states, particularly in the context of disappointment, requires rethinking this conserved circuit within the framework of its ancestral function. Across evolution, organisms have developed defensive strategies to detect and avoid predators and threats. Although these defensive responses are hardwired, long‐term survival relies on adaptive, behavioral flexibility that emerges from learning based on the negative outcomes of prior experiences. Despite the evolutionary significance of these processes, the contribution of the MHb–IPN circuit to innate defensive responses elicited by ethologically relevant threats has remained largely unexplored. To address this gap, we employed a visual looming‐threat paradigm in which an overhead expanding dark disk mimicking an approaching aerial predator evokes innate defensive behaviors. Importantly, although the looming is perceived as a threat, it is never followed by an aversive outcome. Recent studies from our group (Williams et al. 2025) and others (Mederos et al. 2025) demonstrate a reduction in defensive strategies with repeated looming sessions. Using fiber photometry, we observed robust engagement of IPN GABAergic neurons upon initial looming presentations, supporting their role in encoding aversive stimuli (Klenowski et al. 2022). However, IPN recruitment progressively decreased as animals adapted their defensive behaviors over days, consistent with these neurons controlling experience‐dependent responses to novel stimuli (Molas, Zhao‐Shea, et al. 2017; Tapper and Molas 2020; Molas et al. 2024). Optogenetic inhibition of IPN GABAergic neurons attenuated freezing and sheltering behaviors, thereby reducing the perceived threat value and promoting risk‐taking behaviors that facilitated adaptation. In contrast, photoactivation of these neurons during looming presentations prevented behavioral adaptation to threat (Williams et al. 2025). Notably, activation of IPN neurons alone was insufficient to elicit defensive responses, reinforcing the idea that the MHb–IPN pathway does not directly drive defensive motor outputs but instead amplifies aversive processing associated with potentially harmful situations (Liang et al. 2024).

4.2. The IPN–LDTg Circuit in Adaptive Defensive Learning

The IPN is reciprocally connected with several brain regions involved in motivation and affective behaviors, including the dorsal raphe nucleus, the laterodorsal tegmental nucleus (LDTg), and the locus coeruleus (Bueno et al. 2019; Quina et al. 2017). The LDTg provides excitatory input to VTA dopaminergic neurons, regulating their burst firing and dopamine release (Steidl et al. 2017). The IPN–LDTg circuit represents a key mechanism by which the MHb–IPN axis modulates mesolimbic dopamine transmission and motivational states (Wolfman et al. 2018; Monical and McGehee 2025). Through viral tracing and fiber photometry, we discovered that the IPN–LDTg circuit was strongly activated during initial visual threat presentations and adapted with repeated exposures. Selective optogenetic silencing of this circuit disrupted behavioral adaptation to visual looming stimuli, highlighting its role in updating threat value and guiding adaptive defensive strategies based on prior experience.

4.3. IPN Neuronal Populations Fine‐Tune Threat Responses

One of the most remarkable anatomical features of both the MHb and IPN is their exceptionally high neuronal density confined within relatively small, subdivided regions. This compact organization has important functional implications for local computation, signal integration, and precise input–output connectivity, although these remain poorly understood. Among the diverse IPN cell populations, somatostatin (Sst)‐expressing neurons have emerged as a functionally distinct subgroup (Luo et al. 2025). Sst is an inhibitory neuropeptide that acts at MHb–IPN synapses as a retrograde feedback signal to regulate overall IPN excitability (Ables et al. 2017). We found that IPN Sst neurons were activated by looming visual threats and by safety signals when animals entered a shelter. Genetic ablation of this population reduced sheltering and threat‐related avoidance behaviors.

4.4. Assessing Environmental Risk and Adapting Defensive Behaviors to Guide Future Actions

These findings position the MHb–IPN circuit as a central integrative hub for processing aversive threat stimuli in the environment and guiding adaptive behavioral responses. Rather than directly driving defensive actions, the IPN appears to integrate threat salience, internal aversive states, and experience‐based prediction about environmental risks. Through its projections to the LDTg and potential influence on dopaminergic circuits, the IPN contributes to mood regulation and the encoding of negative motivational states. Critically, this circuit updates the perceived value of environmental threats, thereby enabling appropriate risk assessment, experience‐dependent learning, and adaptive decision‐making. Future studies should aim to resolve the precise circuit architecture by which MHb and extra‐habenular inputs converge onto distinct IPN cell types and to map the functional outputs of these populations. In addition, elucidating the plasticity mechanisms underlying threat adaptation will be essential for understanding how the MHb–IPN axis supports both acute and long‐term regulation of aversive memories, threat adaptive behavior, and how failures in this plasticity succumb to pathological accumulation of disappointment.

These findings suggest that MHb–IPN pathways do not simply encode acute aversive events; instead, they update the perceived threat value across repeated exposures, allowing cumulative disappointing or threatening experiences to bias future risk assessment and mood.

5. Functional Specialization of LHb Outputs: Gating Active vs. Passive Coping

The LHb is a phylogenetically conserved structure that links limbic–basal ganglia circuits to midbrain monoaminergic systems (Michel et al. 2024; Proulx et al. 2014). Although the LHb is central to encoding aversive outcomes, recent evidence suggests its outputs are functionally segregated to manage the “disappointment dilemma” through distinct strategies: One pathway processes the negative value of failure, whereas the other organizes the active behavior required to prevent it.

5.1. The LHb–RMTg Pathway: Encoding Negative Value and Disengagement

The projection to the RMTg functions as a critical governor of motivational value. As characterized in primates and rodents (Matsumoto and Hikosaka 2007; Jhou et al. 2009; Hong et al. 2011), this pathway conveys negative reward prediction error (RPE) signals. Phasic excitation of LHb neurons occurs specifically when an outcome is worse than predicted. This signaling implements a teaching rule that discourages actions associated with suboptimal outcomes, consistent with reinforcement‐learning models in which negative RPEs guide trial‐by‐trial behavioral adjustment. Crucially, this pathway does not simply suppress general movement; rather, it selectively gates the motivation to exert effort (Proulx et al. 2018). Using fiber photometry in freely behaving rats, we demonstrated that LHb → RMTg terminal activity increased specifically during transitions into immobility in the forced swim test, the moment the animal ceases its escape attempts but did not correlate with ongoing locomotion above the immobility threshold, effectively encoding the behavioral decision to disengage. Optogenetic activation of this pathway decreased the motivation to struggle by increasing the frequency and duration of immobile bouts, without affecting the animal's capacity to move vigorously. Conversely, inhibiting it restored the drive to escape. Notably, this motivational role extends beyond aversive contexts: Driving LHb → RMTg activity also reduced effort exerted to obtain a reward in an appetitive ratio task, without altering the hedonic value of the reward itself. Together, these findings confirm that the LHb → RMTg axis encodes a domain‐general “disappointment” signal governing the willingness to exert effort, biasing behavior toward passive coping and energy conservation when continued effort appears futile.

5.2. The LHb–VTA Pathway: Encoding Salience and Active Defense

In contrast to the passive disengagement mediated by the RMTg pathway, the LHb projection to the VTA drives the “go” signal for defensive action. Using an intersectional viral strategy to selectively target VTA‐projecting LHb neurons, we monitored calcium activity across behavioral states indicating that this pathway is essential for negative associative learning and the organization of defensive strategies (Ihidoype et al. 2025). VTA‐projecting LHb neurons were excited by aversive outcomes and predictive during active avoidance acquisition, activity that diminished as conditioned avoidance became established, consistent with a role in encoding predictive aversive value rather than sustained threat representation. Notably, these neurons also showed increased calcium transients specifically at movement onset in the tail suspension test (TST), with signals larger than those observed during unconstrained locomotion in the open field, indicating that their recruitment is context and effort dependent rather than a general motor signal.

Selectively silencing LHb → VTA synaptic transmission impaired avoidance learning, prolonged escape latency, and reduced both the persistence and vigor of active escape responses in the TST, without affecting baseline locomotion, indicating that this pathway is required specifically for organizing coping under aversive conditions, not for movement per se. Anatomically, LHb terminals innervate both dopaminergic (TH⁺) and non‐dopaminergic (TH⁻) VTA neurons, with learning‐dependent, population‐specific synaptic plasticity: Enhanced drive onto TH⁺ neurons emerged early during initial cue–outcome learning, whereas delayed plasticity at TH⁻ synapses appeared once avoidance was well consolidated, suggesting that dopaminergic populations support the initial formation of aversive associations, whereas non‐dopaminergic populations are subsequently recruited to sustain organized defensive responding. Notably, activity at VTA‐projecting LHb neurons also decreases as avoidance behavior becomes established, consistent with the view that this pathway tracks the aversive predictive value of a stimulus, a signal that is rapidly recalibrated when the threat can be successfully avoided.

These findings align with primate and rodent data demonstrating that a subset of VTA dopamine neurons, distinct from those inhibited by the LHb → RMTg circuit, is activated by aversive or alerting stimuli, encoding motivational salience independent of valence (Brischoux et al. 2009; Bromberg‐Martin et al. 2010; Lammel et al. 2011). Crucially, salience here refers to the attribution of attentional and motivational weight to a stimulus regardless of its hedonic sign, a property distinct from valence per se. Thus, rather than simply broadcasting a disappointment signal, the LHb → VTA pathway energizes the active avoidance behaviors that, when successful, allow the organism to escape the disappointing outcome entirely.

Taken together, LHb → RMTg and LHb → VTA outputs support rapid evaluation of whether continued effort is worth exerting after a disappointing outcome, implementing short‐timescale updates in coping strategy.

5.3. A Functional Switch for Coping: Dynamic Calibration of Behavior

This dual‐output architecture suggests the LHb functions not merely as a static “disappointment” center but as a dynamic logic gate that calibrates behavior on a rapid timescale to optimize survival in uncertain environments, with the LHb → RMTg signaling when continued effort is futile, and the LHb → VTA signaling when active defense remains the optimal strategy. We propose that the LHb integrates prediction and controllability to select the appropriate coping strategy. When a threat is predictive and controllable, the LHb → VTA axis is engaged to motivate active defense. When the outcome is worse than predicted or effort is futile, the LHb → RMTg axis dominates, enforcing inhibition and withdrawal (Bromberg‐Martin et al. 2010). Through the recurrent engagement of these complementary pathways, the LHb does not simply inhibit motivation, but rather orchestrates the acquisition and consolidation of adaptive avoidance behaviors. The “Disappointment Dilemma” becomes pathological when this balance is lost. In mice susceptible to chronic social defeat stress, we observed distinct, pathway‐specific synaptic adaptations (Hernandez Silva et al. 2024). Susceptible mice displayed synaptic potentiation at LHb → RMTg terminals (amplifying the “give up” signal) and synaptic depression at LHb → VTA synapses (muting the “active defense” signal). This effectively locks the switch: The brain becomes hypersensitive to disappointment while simultaneously losing the drive to actively escape it, providing a circuit‐level mechanism for the transition from adaptive learning to depressive hopelessness.

6. Fear and the LHb

During threat exposure, the LHb evaluates predictability and controllability and engages RMTg or VTA outputs to promote the most efficient coping strategy. In addition, LHb‐generated information influences the formation of the associated memory. To analyze how the LHb regulates threat/fear memories, two paradigms can be used: step‐down inhibitory avoidance (IA), in which animals learn not to step down from a platform onto an electrified grid floor, and classical Pavlovian fear conditioning (FC).

In the IA paradigm, inactivation of the LHb during training disrupted the long‐term persistence of fear memory (Tomaiuolo et al. 2014). This was evidenced by a conserved avoidance behavior 24 h after training but a significant amnesia 1 week later. The temporal stability of IA memory depends on two critical windows: The first occurs during training, when shock intensity determines memory persistence; the second occurs 12 h after acquisition, when dopamine‐induced brain‐derived neurotrophic factor (BDNF) synthesis in the hippocampus is required for long‐term stability (Bekinschtein et al. 2007; Rossato et al. 2009). Notably, the weak amnesia induced by LHb inactivation during strong training was reversed by hippocampal infusion of BDNF or D1 agonists 12 h after (Tomaiuolo et al. 2014). In contrast, LHb inactivation did not affect memory when training was weak. Thus, LHb inactivation selectively transformed a strong IA memory into a weak one with reduced temporal stability.

In the FC paradigm, inactivation of the LHb during training drastically reduced both contextual and cued fear conditioning (Sachella et al. 2022). Nevertheless, when the cue was presented in the training context, a conserved fear response was observed. A similar effect was observed with sustained optogenetic activation of the LHb, indicating that memory impairment could be induced by disrupting endogenous LHb activity, regardless of the sign of the change (Sachella et al. 2022). As in the IA paradigm, the temporal persistence of this residual combined fear memory was also impaired. Therefore, LHb inactivation does not prevent the formation of associative fear memories but increases the degree of similarity between training and testing conditions required for effective retrieval.

A role for the LHb in fear acquisition has also been shown by other authors (Durieux et al. 2020). Notably, some of them are coincident with ours in showing that the LHb modifies FC aspects related to the interaction between hippocampally encoded information and non‐hippocampal FC components, such as overshadowing or memory extinction retrieval (Velazquez‐Hernandez and Sotres‐Bayon 2021), and also in showing that continuous activation of the LHb disrupts normal FC (Levinstein et al. 2022).

6.1. Two‐Fear Memories, One Common Role for the LHb

Under threatening conditions, aversion‐related information generated by the LHb determines the strength of the associated memory. This is evidenced by reduced temporal stability of IA and FC learning and by more stringent conditions for effective fear retrieval in FC (Sachella et al. 2022; Tomaiuolo et al. 2014). Notably, a similar idea has recently been proposed by Zichó et al. (2025), who described increased fear generalization through disinhibition of the LHb during training. By playing this role, the LHb would be in a key position to be targeted to treat psychiatric diseases related to fear dysregulation, such as panic attacks or phobias.

7. General Conclusion

Both habenulae participate in a wide range of different aspects of threat processing through specialized input and output pathways (Table 2). Notably, these two structures do not share primary inputs or outputs; their remarkable functional overlap suggests a coordinated framework of action. Indeed, recent work highlights the medial and lateral habenula as two structures with higher transcriptional responses to fear‐conditioned stimuli. We propose that, together, MHb and LHb constitute a disappointment pathway specialized in encoding worse‐than‐expected scenarios across a wide range of environmental conditions and temporal scales in a cell‐type‐specific manner. Although their behavioral domains overlap, LHb operates closer to action selection and value updating within limbic–basal ganglia–dopamine loops, calibrating rapid changes in coping strategy. In contrast, the MHb–IPN axis is biased toward integrating interoceptive, immune, metabolic, and drug‐related signals into more slowly evolving aversive states and mood, providing a substrate for the long‐term accumulation of disappointment.

TABLE 2.

Summary of key findings and open questions.

Habenular circuit	Behavioral domain	Key points from this review	Open questions
MHb → IPN	Nicotine aversion, withdrawal, adaptive threat processing, mood	nAChR subunit balance (α3, α5, α4) sets nicotine intake versus aversion; CB1, MOR,GPR139, and GPR151 modulate glutamatergic/cholinergic output transmission and opioid and cannabinoid sensitivity; IPN GABAergic and Sst neurons tune adaptive threat responses and novelty/exploration	How distinct MHb/IPN cell types and microcircuits differentially contribute to acute versus long‐term disappointment; how peripheral, metabolic, and immune signals acting on MHb reshape long‐term mood and disappointment accumulation; whether manipulating MHb‐enriched GPCRs (e.g., GPR151 and GPR139) can selectively adjust aversive limits on drug intake without broad side effects
LHb → RMTg	Negative RPE, passive coping	Encodes worse‐than‐expected outcomes, gates disengagement in forced swim and tail suspension, and can bias the system toward “give‐up” strategies when potentiated in chronic stress	How plasticity at LHb–RMTg synapses determines the transition from adaptive disengagement to depressive‐like hopelessness; whether specific molecular targets can normalize this output without blocking learning and how individual differences in this pathway confer vulnerability or resilience to chronic stress
LHb → VTA	Salience, active defense	Supports aversive learning and active avoidance; required to maintain escape responses under stress and likely drives salience‐encoding dopamine populations	How LHb–VTA dynamics encode controllability and predictability; how aberrant weakening of this pathway contributes to anergia and loss of motivated avoidance; and whether boosting salience‐encoding dopamine populations can restore active coping without increasing maladaptive reward seeking

Author Contributions

Christophe D. Proulx: conceptualization, investigation, writing – review and editing. Edgar Soria‐Gómez: conceptualization, investigation, funding acquisition, writing – review and editing. Ines Ibañez‐Tallon: conceptualization, writing – review and editing, investigation, writing – original draft. Joaquin Piriz: conceptualization, investigation, funding acquisition, writing – review and editing. Susanna Molas: conceptualization, investigation, writing – review and editing.

Funding

This work was supported by the Fundación Tatiana Pérez de Guzmán el Bueno, the Kellen Women's Entrepreneurship, the Brain and Behavior Research Foundation Young Investigator Award (30616), National Institute of Mental Health (MH129040), the Canadian Institutes of Health Research (PJT169117), Natural Science and Engineering Research Council of Canada (RGPIN‐2025‐04515), Fonds de Recherche en Santé du Québec, Basque Government (PIBA_2024_1_0057), Achucarro Basque Center for Neuroscience, EU COFUND (H2020‐MSCA‐COFUND‐2020‐101034228‐WOLFRAM2), International Brain Research Organization, and the Ministerio de Ciencia, Innovación y Universidades (PID2024 160818NB‐I00, PID2021‐125763NB‐I00).

Conflicts of Interest

The authors declare no conflicts of interest.

Ibañez‐Tallon, I. , Molas S., Proulx C. D., Piriz J., and Soria‐Gómez E.. 2026. “Review: “The Disappointment Dilemma: Short‐ and Long‐Term Learning From Negative Outcomes”.” European Journal of Neuroscience 63, no. 9: e70522. 10.1111/ejn.70522.

Data Availability Statement

No new datasets were generated or analyzed in this review. All data supporting the conclusions of this article are derived from previously published studies cited throughout the manuscript.