Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Reward, Pleasure, and Incentive Salience

Abstract. Reward is a multicomponent psychological phenomenon involving hedonic impact, motivational incentive salience, and associative learning. This fundamental research monograph synthesizes the experimental history of reinforcement and taste reactivity, the dissociation of reward 'liking' from cue-triggered 'wanting,' opioid-sensitive hedonic hotspots, dopamine-dependent motivation and prediction, Pavlovian-to-instrumental transfer, outcome devaluation, dynamic state-dependent valuation, computational models, and translational research on food and addiction. It evaluates primary animal studies, human behavioral paradigms, electrophysiology, causal pharmacology, computational models, and recent methodological reviews while specifying the limits of inference across species and settings. The synthesis identifies where neural interventions establish causal effects on separately measured reward components, where alternative cognitive explanations remain plausible, and how future studies can improve validity, replication, and clinical translation. The review is a critical narrative synthesis of published evidence, without original experiments or claims of independently validated new theory.

Keywords: reward; pleasure; hedonic impact; liking; wanting; incentive salience; dopamine; opioid hedonic hotspots; nucleus accumbens; ventral pallidum; Pavlovian-to-instrumental transfer; outcome devaluation; reward prediction error; cue-triggered motivation; incentive sensitization; craving; reward learning; Artificial Era

Reward as a Multicomponent Psychological Phenomenon

Reward is an event or outcome that changes an organism's behavior through reinforcement, acquires positive affective significance, or comes to be pursued under specific conditions. These functions often co-occur, but they require different observations. A sweet solution may evoke positive facial reactions, support a learned action, attract approach to a predictive cue, and later be described as enjoyable. None of those indicators alone measures the whole phenomenon. The contemporary science of reward accordingly separates hedonic impact or 'liking,' incentive salience or cue-triggered 'wanting,' and associative learning about events and their consequences. These terms designate experimentally identifiable operations rather than interchangeable everyday descriptions. The distinction is grounded in dissociations between affective reactions and motivated performance, including dopamine manipulations that alter pursuit while preserving positive taste reactions Berridge & Robinson (1998) Berridge et al. (2009) Wyvell & Berridge (2000).

Pleasure denotes positive hedonic impact produced during an experience; in humans, it can be investigated through subjective ratings, reports, and convergent behavioral or physiological indices. The experimentally narrower concept of core 'liking' refers to positive affective reactions that may be measured without attributing conscious pleasure to a nonverbal organism. Wanting in incentive-salience theory denotes the motivational pull conferred on a reward or its predictor at the moment it is encountered. It is different from a stated preference, a deliberate intention, or an estimate of future happiness. Learned predictions and outcome associations allow cues to signal possibilities, while physiological and neural states determine how vigorously those possibilities are pursued. Pleasure, prediction and pursuit are interacting components whose relative contributions vary across settings and individuals Berridge (2012) Pool et al. (2016) Kringelbach & Berridge (2009).

The central research questions are why some learned cues become powerfully attractive, how positive hedonic reactions are generated and measured, what dopamine and opioid circuits contribute to each process, how bodily state transforms incentive value, and when craving persists despite diminished enjoyment. An adequate explanation must connect causal intervention, behavioral operationalization, neural recording and human self-report without assuming they are equivalent. This monograph is a critical narrative synthesis of original experiments, formal models, methodological analyses and peer-reviewed reviews, not a registered systematic review or a meta-analysis. It treats generalized claims about clinical disorders and artificial systems as hypotheses requiring their own tests Pool et al. (2016) Robinson & Berridge (2025) Berridge & Kringelbach (2015).

Conceptual Boundaries: Liking, Wanting, Learning, and Choice

An organism can learn that a tone predicts food yet decline to act on that knowledge when satiated; conversely, a cue can sharply increase an instrumental response without changing the previously learned stimulus–outcome association. Such patterns show why reward prediction, current incentive value and action execution should be separately operationalized. In a Pavlovian contingency, a stimulus predicts an outcome without depending on the organism's response. In an instrumental contingency, an action changes the probability of a consequence. Pavlovian-to-instrumental transfer tests the influence of learned predictive stimuli on independently trained actions, creating a controlled bridge between associative knowledge and motivational expression Holland (2004) Wyvell & Berridge (2000).

Hedonic impact is assessed close to consumption or stimulus experience. Instrumental choice depends on anticipated consequences, effort, delay, available alternatives and action capacity. A consumption measure may therefore change when the animal is hungry, when movement becomes easier, or when competing behavior disappears, even though positive taste reactivity remains unchanged. A self-reported wanting scale may combine desire, expectation, imagined pleasure and social meaning. The system of constructs must be interpreted with attention to measurement context. Pool and colleagues' methodological review of human and animal wanting/liking tasks describes variability in operational definitions and identifies expected pleasantness as a source of confounding in human tasks Pool et al. (2016) Grill & Norgren (1978).

Motivational salience in this dossier specifically means reward-oriented incentive significance, not the broader perceptual prominence of any unusual or aversive event. Attentional capture, approach, willingness to work and conditioned reinforcement are useful observations, but none automatically establishes the full incentive-salience mechanism. Similarly, explicit happiness ratings extend beyond momentary hedonic impact into appraisal, memory and comparison. Scientific language becomes more precise when reports of pleasure, microstructural affective reactions, learned outcome expectations and work expenditure are assigned to their corresponding empirical measurements. This precision is necessary when reward science is applied to depression, addiction and problems of self-regulation Berridge et al. (2009) Berridge & Kringelbach (2015) Pool et al. (2016).

Historical Foundations of Experimental Reward Research

Mid-twentieth-century research distinguished reinforcement from reflexive or consummatory behavior by manipulating consequences of action. Olds and Milner's 1954 experiments demonstrated that electrical stimulation of certain brain regions could reinforce responding in rats, establishing a tractable route to biological reward research. Intracranial stimulation made reinforcement experimentally controllable but stimulated multiple fibers and cellular populations. Its result did not independently demonstrate a specific neurotransmitter's role in pleasure, much less a direct measure of a private experience. The distinction between a reinforcing intervention and the affect it produces became a central interpretive issue that later motivational neuroscience had to resolve Olds & Milner (1954) Berridge & Robinson (1998).

Grill and Norgren introduced the taste-reactivity method in 1978 to measure organized facial and bodily responses elicited by intraoral tastes without requiring spontaneous drinking. Sweet and bitter solutions produce differentiable ingestive and aversive reaction patterns. This innovation made it possible to study positive and negative affective responses apart from sustained intake. Subsequent research on equal-preference tastes showed that preference and reaction patterns could express separable dimensions of palatability. The method supplies an observable index of hedonic reactions in animals, provided researchers avoid assuming that an outward motor pattern constitutes complete proof of conscious experience Grill & Norgren (1978) Berridge & Grill (1984).

The 1990s brought a decisive conceptual reorganization. Robinson and Berridge's 1993 incentive-sensitization theory addressed why drug-seeking could become persistent even when consumption was no longer especially enjoyed. Berridge and Robinson's 1998 review integrated evidence that severe dopamine depletion could leave certain affective taste reactions intact while modifying motivated behavior. In parallel, Schultz, Dayan and Montague connected dopamine-neuron responses to predictive discrepancies in reward learning. These programs specified different candidate functions for overlapping mesolimbic systems: motivational attraction, teaching signals and hedonic impact. Their scientific value lies in testable experimental contrasts, not in the assertion that one mechanism explains every reward-related response Robinson & Berridge (1993) Berridge & Robinson (1998) Schultz et al. (1997).

From Learned Signals to Incentive Objects

A predictive cue can become an object of attraction in its own right. Its perceptual features come to carry a learned relation to reward, and its presentation may redirect attention, trigger approach or energize an action trained separately. The motivational effect is temporally local: the same sound or image can have different influence before and after a meal, following a pharmacological manipulation, or after new experience changes expectation. Incentive-salience theory formalizes this distinction by treating cue-triggered motivation as a function of prior association and current state rather than as a fixed historical reward value Berridge (2012) Zhang et al. (2009).

The term incentive salience is especially useful where cues elicit behavior in the absence of immediate outcome delivery. Wyvell and Berridge trained rats on sucrose-reinforced instrumental responding and separately established a sucrose-predictive Pavlovian cue. During test sessions without response-contingent reinforcement, intra-accumbens amphetamine selectively amplified cue-evoked responding while failing to increase positive sucrose taste reactions. Because the test withheld new reinforcement, the observed enhancement could be distinguished from a simple increase in reinforcement received during testing. The experiment offers a strong intervention-based dissociation, although its interpretation remains bounded by one species, cue arrangement, outcome and neural manipulation Wyvell & Berridge (2000).

Repeated exposure to amphetamine can also produce sensitization of subsequent cue-triggered incentive motivation. Wyvell and Berridge reported enhanced responses to sucrose-associated cues following prior drug exposure, extending the motivational hypothesis beyond an acute injection. Later experimental work traced motivational amplification into activity patterns of ventral pallidal neurons. These observations point to distributed incentive processing and a distinction between momentary dopamine-related state and historical learning. They also caution against concluding that neural sensitization is sufficient for any particular human addictive disorder: vulnerability, access, context, social conditions and learning histories remain important explanatory variables Wyvell & Berridge (2001) Tindell et al. (2005) Robinson & Berridge (2025).

Taste Reactivity and the Experimental Identification of Hedonic Impact

An adequate hedonic measure must reduce dependence on locomotion, effort and acquisition rate. The taste-reactivity preparation introduced by Grill and Norgren delivers a controlled solution directly into the mouth and records temporally structured orofacial responses. Positive patterns to sucrose and aversive responses to quinine can be scored without requiring an animal to travel, press a lever or select between bottles. The design is important because the amount consumed in a free-feeding assay mixes palatability with hunger, accessibility, effort and motor function. Taste reactivity is particularly suited to within-animal comparisons across pharmacological or deprivation conditions, provided coding is blinded and general motor impairment is assessed Grill & Norgren (1978) Berridge & Grill (1984).

Positive taste reactivity is a validated behavioral proxy for hedonic impact under specified species and task conditions, yet its scope remains narrower than a first-person report of pleasure. Different drugs can influence swallowing, facial tone or state-dependent stereotypy; experimental investigators must determine whether the response changes reflect an affective alteration or a change in the expression apparatus. The strongest designs combine taste microstructure with consumption, cue pursuit, sensory discrimination and neural interventions. Parallel measurements demonstrate whether one manipulation affects only an index of enjoyment or several reward functions at once Berridge et al. (2009) Peciña & Berridge (2005).

Human pleasure research requires corresponding caution. Momentary liking ratings refer to introspected subjective experience, while retrospective preference reports may reflect memory and cultural categories. Brain activation during enjoyable activities is informative but is not itself a direct readout of pleasure. A study can support a dissociation when it independently measures consummatory pleasantness, instrumental vigor and anticipatory desire under controlled changes in reward availability. The systematic methodological review by Pool and colleagues underscores that human wanting and liking measures have sometimes been defined differently across experiments, complicating quantitative generalization Pool et al. (2016) Berridge & Kringelbach (2015) Kringelbach & Berridge (2009).

Localized Hedonic Hotspots and Distributed Pleasure Circuits

Experiments with mu-opioid receptor agonism mapped the sites at which neural stimulation increases positive reactions to sucrose. Peciña and Berridge found a spatially localized hedonic hotspot within the rostrodorsal medial shell of the rat nucleus accumbens. Positive taste reactions were amplified when stimulation occurred in that site, whereas increased eating could be produced across a larger territory. The contrast matters because enhanced ingestion is sometimes mistakenly treated as proof of enhanced pleasure. Local mapping establishes a causal contribution to a defined hedonic endpoint, not the sufficiency of a single microscopic location for all forms of human enjoyment Peciña & Berridge (2005) Berridge & Kringelbach (2015).

Ventral pallidum participates in a cooperating hedonic network. Smith and Berridge tested the functional interaction between accumbens and ventral pallidal hotspots, finding evidence that opioid stimulation in one node depended on coordinated activity with the other for enhanced positive sucrose taste reactions. At the same time, effects on food intake could follow a different pattern. The result encourages a circuit-level model in which localized sites have specific causal leverage within an interconnected system. The small territory supporting increased 'liking' should not be confused with the broader anatomical distribution of reward-related activity observed by imaging or electrophysiology Smith & Berridge (2007) Berridge & Kringelbach (2015).

The opioid system is a particularly informative mechanistic contrast to dopamine systems involved in approach. Wassum and colleagues described differentiated opioid circuit contributions to the palatability and desirability of rewarding outcomes. This reinforces the requirement to measure motivational and hedonic effects independently even when both depend on overlapping limbic anatomy. Human pleasure imaging implicates orbitofrontal, insular, ventral striatal and other areas across multiple reward modalities, but imaging associations alone cannot specify causal production of pleasure. Claims of an isolated 'pleasure center' replace a distributed, state-dependent and measurement-sensitive organization with an inaccurate localization Wassum et al. (2009) Kringelbach & Berridge (2009) Berridge & Kringelbach (2015).

Dopamine, Motivational Vigor, and the Limits of Pleasure Attribution

Dopamine is implicated in reward processing through several experimentally distinct operations: anticipatory signaling, cue-triggered pursuit, mobilization of work, aspects of action selection and learning-related updating. Its function changes with timescale, location, measurement and task demands. Berridge and Robinson assembled influential evidence against identifying dopamine with the generation of positive taste affect. Under substantial dopamine depletion, some positive and negative taste-reactivity patterns remained detectable despite impaired reward-seeking performance. The dissociation indicates that dopamine-dependent motivation and basic hedonic reactions can vary independently under controlled conditions; it does not imply that dopamine is irrelevant to all pleasant experiences or to every component of human affect Berridge & Robinson (1998) Berridge & Robinson (2016).

Single-neuron recordings established a different dimension of dopamine function. Schultz and colleagues observed response patterns consistent with reward prediction errors in specified primate midbrain populations, leading to influential temporal-difference interpretations. Reward prediction error is a computational discrepancy between obtained and expected outcomes; incentive salience describes the immediate motivational effect of an outcome or its cue. The two concepts can cooperate in a behaving organism because learning updates cue associations and motivational states change their current behavioral force. Neither an error-related firing pattern nor an increase in dopamine release is sufficient by itself to identify felt pleasure Schultz et al. (1997) Berridge (2012).

Mohebi and colleagues measured dopamine-related dynamics across neural signals and task phases, identifying dissociable patterns relevant to learning and motivation. Their observations support treating dopamine cell-body activity and local dopamine availability as related but nonidentical measurements. The same broad transmitter system can contribute to learning during one task epoch and to ongoing pursuit during another, with local dynamics shaped by circuit context. When a study reports dopamine activation, scientific interpretation therefore requires specifying recording site, temporal resolution, manipulation, behavioral variable and causal contrast. A simple 'dopamine equals happiness' formula cannot accommodate these experimentally demonstrated dissociations Mohebi et al. (2019) Berridge & Kringelbach (2015).

The Ventral Pallidum, Accumbens, and Dynamic Reward Coding

The ventral pallidum receives limbic information and participates in the expression of learned motivational significance. Tindell and collaborators recorded ventral pallidal neurons during a sequence of reward-predicting conditioned stimuli. Prior sensitization and acute amphetamine exposure altered neural responses associated with incentive motivation. These recordings supply a bridge from a behavioral account of cue pursuit to circuit activity, but they remain correlational at the level of individual recorded units; causal claims require interventions that manipulate the relevant projections or populations while preserving behavioral controls Tindell et al. (2005) Wyvell & Berridge (2001).

Not every cue acquires a constant appetitive sign. Tindell and colleagues' later work on dynamic computation demonstrates how motivational consequences can shift with changing physiological states and learned outcome relations. The finding challenges an account in which a conditioned stimulus simply carries a permanent scalar value copied from past reinforcement. Neural and behavioral coding of valence can depend on the state of the organism and the network. Berridge's analysis of affective modules and modes extends this lesson: anatomical nodes may have context-dependent affective functions that must be mapped through the conditions under which they are engaged Tindell et al. (2009) Berridge (2019).

A comprehensive interpretation distinguishes necessary participation, sufficient local intervention, correlated activation and distributed network computation. Opioid infusion that increases positive taste reactions in a localized region provides a different causal inference from a neuron whose firing predicts an upcoming reward. Ventral pallidal recordings can reveal state-dependent motivational coding without showing the neural origin of a private feeling. Experiments combining lesions, agonists, recordings, outcome devaluation and multiple behavioral endpoints narrow the interpretation. This methodological pluralism is indispensable because reward-related structures are reused for anticipation, movement, valuation, affect and learning Smith & Berridge (2007) Tindell et al. (2005) Mohebi et al. (2019).

Pavlovian-to-Instrumental Transfer as an Experimental Paradigm

Pavlovian-to-instrumental transfer, or PIT, tests whether a cue trained to predict an outcome alters independently learned action. In a common arrangement, one phase associates a neutral stimulus with a reward, another establishes an instrumental response for the reward, and a transfer test measures behavior when the conditioned stimulus is presented without ongoing reinforcement. General PIT refers to broad invigoration across reward-seeking actions, while outcome-specific PIT refers to selective changes in the action associated with the cue's predicted outcome. These different effects engage motivational and representational processes that require separate controls for motor activation, attention and expectancy Holland (2004) Badioli et al. (2024).

The interpretive strength of PIT depends on independently establishing both forms of learning and testing transfer outside new reward delivery. Wyvell and Berridge's amphetamine study showed a selectively enhanced Pavlovian cue effect on responding, while the animal's positive reactions to sucrose were not similarly amplified. The comparison joins motivational causation with a hedonic assay, making a stronger reward dissociation than either assay alone. However, behavior called PIT can also reflect goal-directed planning or learned stimulus-response associations; the term names a task effect that can arise through more than one underlying mechanism Wyvell & Berridge (2000) Seabrooke et al. (2017).

A preregistered meta-analysis of human PIT studies published by Badioli and colleagues in 2024 examined factors influencing the strength and reliability of transfer effects and emphasized methodological heterogeneity. This later evidence matters for translational inference. Human PIT tasks differ in reward type, instruction, awareness, conditioning, extinction, response affordances and whether experimenters assess general or specific transfer. An effect observed in rats under one pharmacological manipulation should not be declared a universal human biomarker of craving. Comparability requires matching operational definitions, measuring the learning phases and preregistering the hypothesized transfer contrast Badioli et al. (2024) Pool et al. (2016).

Outcome Devaluation and the Structure of Motivated Action

Outcome devaluation probes whether behavior depends on an updated representation of an outcome's desirability. After an action or cue has been learned, the outcome can be rendered less valuable through sensory-specific satiety, aversion learning or another defined manipulation. Investigators then assess whether pursuit declines and whether Pavlovian cues retain their ability to affect responding. If an action changes when the outcome is devalued, this supports sensitivity to current outcome value under the testing conditions; if it persists, several mechanisms remain possible, including stimulus-response habit, motivational transfer or incomplete learning about the new state Holland (2004) Allman et al. (2010).

Holland's 2004 experiments showed that relations between transfer and reinforcer devaluation depend on the training arrangement, including whether tasks involve a single or multiple reinforcers. Allman and collaborators subsequently studied reinforcer-selective PIT in humans following devaluation. These findings demonstrate that a PIT score cannot be interpreted independently of its action–outcome structure. A single outcome may allow generalized vigor, whereas multiple learned outcomes permit tests of identity-specific bias. A cue can change which action is favored without increasing all reward-seeking, and devaluation may alter intentional choice through a mechanism distinct from cue-triggered activation Holland (2004) Allman et al. (2010).

Seabrooke and colleagues examined human outcome-specific PIT using devaluation manipulations and reported evidence for a goal-directed component. Their findings challenge any interpretation that all PIT behavior is an automatic incentive reflex. A participant may retrieve the outcome signaled by a cue, infer its current usefulness and adjust choice strategically. The general research problem is to specify when conscious knowledge, inferred identity, state-dependent vigor and affectively charged cue attraction make independent contributions. Experiments should compare model predictions through orthogonal manipulations, rather than equating the presence of transfer with a unique underlying psychological process Seabrooke et al. (2017) Berridge (2012).

Physiological State, Hunger, Satiety, and Inference

Motivational state is an active causal input into reward seeking. Food restriction, satiety and sensory-specific changes in appetite may modify the intensity of cue-triggered pursuit without changing the historical fact that a cue predicts a given food. An incentive-salience account models this as an interaction between learned cue associations and the organism's current physiological state. Zhang and colleagues developed a computational model that represents dynamic amplification of learned reward cues through state variables, allowing incentive motivation to shift quickly when conditions change. The model is an explanatory formalization whose parameters must be independently connected to behavioral or neural observations Zhang et al. (2009) Berridge (2012).

The interaction is especially informative where a novel need emerges. If a learned cue gains motivational power after a state change, researchers must ask whether the animal already possessed relevant knowledge about the specific outcome, newly learned its value, or generalized from familiar physiological experiences. Immediate changes in performance are evidence for rapid modulation, while demonstrated absence of learning requires controls over prior exposure and history. The conceptual boundary between incentive-state modulation and inferential value updating can be difficult to identify from choices alone. A stronger experiment measures moment-by-moment responses, confirms learned cue identity and uses conditions that pit model-derived predictions against each other Tindell et al. (2009) Zhang et al. (2009).

Physiological modulation does not entail that all desires are homeostatic. Monetary rewards, novelty, social approval, sexual stimuli and artistic experiences recruit different representational and cultural resources. Their influence can fluctuate with mood, expectations, fatigue, loss and interpersonal context. The generalizable claim is that reward pursuit emerges from the coupling of prior information and present organismic state, while the exact biological and social variables differ by reward domain. Theoretical extension beyond controlled food and drug paradigms requires construct-equivalent measurements and tests of external validity Berridge (2012) Berridge & Kringelbach (2015) Pool et al. (2016).

Computational Accounts of Reward Learning and Incentive Value

Reinforcement-learning models represent how expected value is updated from experience. A standard prediction-error term compares actual reward with an expected value and uses the discrepancy to revise a learned estimate. This is a model of acquisition and prediction, not a complete specification of hedonic experience or immediate motivational force. In incentive-salience models, a previously acquired reward prediction is combined with current physiological or pharmacological variables to generate a cue's present attractiveness. The distinction yields empirical tests: manipulated state can change cue pursuit without requiring a new episode of reinforcement, while prediction error can alter future association strengths even if immediate pursuit is weak Schultz et al. (1997) Zhang et al. (2009).

Zhang and colleagues' 2009 model illustrates how formally related functions can be assigned separate computational roles. A stored association informs the identity or historical relevance of a cue, whereas a dynamic state transformation changes the magnitude and possibly the sign of incentive motivation. Such a formulation permits researchers to generate quantitative predictions about changes in approach, effort and choice under hunger, sensitization or satiety. Model fit, however, is conditional on task design and identifiability: multiple models may explain the same observed button presses. Discriminating models demands out-of-sample prediction, state manipulations and independent affective measures Zhang et al. (2009) Berridge (2012).

Reward-learning theory and incentive-motivation theory are complementary where experiments genuinely separate them. Dopamine responses associated with prediction errors can participate in teaching during conditioning, while local dopamine dynamics may track ongoing motivation during action. Mohebi and colleagues' work on dissociable dopamine dynamics offers physiological evidence that a single scalar label for dopamine function obscures distinct temporal computations. Any strong computational synthesis should specify whether it predicts learning rate, cue attraction, action vigor, subjective valuation or consummatory pleasure, and should report which of those outcomes has been measured Mohebi et al. (2019) Schultz et al. (1997).

Measurement Across Species and Reward Modalities

Cross-species measurement is possible when procedures preserve the causal structure of the construct. A rat's positive orofacial reaction to sucrose, a person's contemporaneous pleasure rating, and a person's choice to exert effort for a reward are related observations, but they are not measurement-identical. To compare them, researchers need convergent indicators, measurement-invariance tests where feasible, and explicit reasoning about how one behavior indexes the proposed psychological operation. Translating licking or approach into a claim about human pleasure without an intermediate validation chain risks attributing unobserved subjective content to a motor response Grill & Norgren (1978) Pool et al. (2016).

Experimental reward modalities also differ. Food and fluid reinforcement depend on deprivation state, oral sensation and sensory-specific satiety. Money involves symbolic exchange, anticipated future opportunities and social learning. Social rewards can depend on relationship history and interpersonal meaning; pharmacological rewards can affect the neural processes used to measure them. General reward tasks often exchange these outcomes for a common numerical value, but equal money amounts or equal food calories do not guarantee equal hedonic intensity, incentive appeal or preference across participants. Task validity must include state, familiarity, opportunity and the temporal interval between prediction, action and experience Berridge & Kringelbach (2015) Kringelbach & Berridge (2009).

The most informative multimethod protocol examines learned contingencies, reports of expected and actual pleasure, time-resolved approach, response vigor, costs paid and neural activity under factorial manipulations. It should include tests for motor capacity, attention, reward identity and shifts in general arousal. Researchers should avoid interpreting a null difference between pleasure ratings as proof of identical feelings when the rating scale is insensitive or the sample is small. Similarly, statistically significant activation in a reward-related area may reflect prediction, novelty, learning, salience or motor preparation. Predefined measurements and transparent inferential boundaries determine whether a claimed separation is scientifically meaningful Pool et al. (2016) Badioli et al. (2024) Mohebi et al. (2019).

Incentive Sensitization and the Persistence of Craving

The incentive-sensitization theory of addiction proposes that repeated drug exposure can produce lasting adaptations of mesolimbic motivational systems in vulnerable organisms. These adaptations increase the incentive salience attributed to drug-associated cues, making those cues potent triggers of pursuit even when the drug produces less enjoyment than earlier. Robinson and Berridge framed the proposal in 1993 and reviewed its subsequent evidence in 2016 and 2025. The theory therefore concerns a selective strengthening of one dimension of reward rather than an overall increase in reward value. It must be evaluated against observed cue reactivity, behavioral persistence, sensitization procedures and competing mechanisms of addiction Robinson & Berridge (1993) Berridge & Robinson (2016) Robinson & Berridge (2025).

Animal experiments demonstrate that psychostimulant-related sensitization can enhance conditioned incentive effects for natural rewards as well as drug-related contexts. These findings make the mechanism more general than a direct drug–pleasure relation. Yet addiction in humans involves complex interactions among reinforcement, habit, withdrawal, emotional regulation, accessibility, social structure and individual vulnerabilities. Clinical craving reports are not identical to the theoretical 'wanting' construct: a subjective urge can contain expectations, affective distress, interpretations and beliefs. The degree to which each component contributes varies across people and episodes. A valid clinical translation requires longitudinal evidence and intervention studies that isolate predictive effects beyond established clinical measures Wyvell & Berridge (2001) Robinson & Berridge (2025).

The 2025 reassessment by Robinson and Berridge compares incentive sensitization with habit, opponent-process and impaired-control accounts and addresses questions about craving, compulsion, attentional capture and uncertainty. The research significance of this comparison is methodological. Persistent drug use may follow pursuit of an attractive cue, relief from withdrawal, a well-trained action sequence or a perceived lack of alternatives. No single behavior reveals which process operated. Scientific models become more useful when they produce mutually discriminating predictions about exposure history, cue specificity, state transitions, devaluation sensitivity and long-term outcomes Robinson & Berridge (2025) Holland (2004).

Anhedonia, Depression, and the Diversity of Reward Disturbances

Clinical language often uses anhedonia as if it were a single reduction in pleasure. Reward research instead motivates a more differentiated assessment of anticipatory desire, willingness to expend effort, consummatory enjoyment, reinforcement learning and the value attributed to future outcomes. An individual may report diminished enjoyment, reduced motivation, impaired concentration or lower reward expectancy, and different mechanisms may contribute to those reports. These observations can be related but should not be presumed equivalent. A precise research protocol determines which function was measured, which treatment or disease state was present, and whether observed behavior was limited by fatigue, medication, physical ability or context Berridge & Kringelbach (2015) Pool et al. (2016).

Hedonic hotspot studies reveal why motivational impairment is an insufficient measure of reduced pleasure. Some neural interventions decrease instrumental pursuit while leaving positive taste responses comparatively intact; others amplify positive reactions within specific opioid-sensitive circuits. This evidence yields a principled distinction for clinical hypothesis generation, while the original causal experiments largely concern nonhuman animals. Human imaging and behavioral tasks extend the framework, but they do not provide a diagnostic assay in which a single reward-related brain response establishes clinical anhedonia. Symptom definitions, diagnostic classification and therapeutic conclusions require their own validated clinical evidence Berridge & Robinson (1998) Smith & Berridge (2007) Berridge & Kringelbach (2015).

Reward-function decomposition has translational value because a treatment directed at effort allocation may affect daily functioning without necessarily changing the immediate pleasantness of consummation. Equally, a change in self-reported pleasure may coexist with minimal behavioral change if opportunities for reward remain restricted. These are research implications, not prescriptions. The practical goal is to design studies measuring multiple endpoints and to test whether improvements in one reward component predict meaningful change in symptoms and life participation. The relevant clinical evidence should be evaluated by randomized design, adequate follow-up, safety reporting and replication before being translated into recommendations Pool et al. (2016) Berridge & Kringelbach (2015).

Food Rewards, Appetite, and Cue-Driven Consumption

Food is the principal experimental model for disentangling reward functions because palatability, hunger, cue associations and intake can be manipulated separately. Taste-reactivity tests investigate hedonic consequences of oral stimulation, whereas PIT and approach paradigms test effects of learned predictors on action. The contrast is decisive when food intake increases after an intervention that leaves positive taste reactions unchanged. Such a result demonstrates stronger consumption or pursuit under the observed conditions and calls for further experiments to determine whether the effect reflects current incentive salience, relief of deprivation, changed perception or altered action capacity Wyvell & Berridge (2000) Peciña & Berridge (2005).

Natural feeding episodes combine sensory pleasure, metabolic regulation, expectation, learned context and the availability of alternatives. A food cue can become conspicuous because of prior learning, while the organism's hunger state can change the strength of the reaction. Sensory-specific satiety creates an opportunity to vary outcome value without deleting knowledge of which stimulus predicts it. The scientific challenge is to infer whether a cue has acquired attractive motivational properties or whether behavior is explained by explicit valuation and selection. Outcome-devaluation procedures, conditioned reinforcement, attentional measures and independent liking assays help discriminate these explanations Holland (2004) Zhang et al. (2009).

Claims about highly processed foods or sugar as universally 'addictive' require additional evidence beyond the existence of reward circuits shared with pharmacological reinforcement. Overlap in neural systems does not show identical exposure effects, withdrawal profiles, diagnostic status or treatment requirements. Incentive salience can help formulate hypotheses about cue-responsive eating, yet food behavior is also shaped by cost, access, culture, hunger and social interaction. Responsible scientific interpretation identifies the task, organism, endpoint and inferential distance from everyday eating instead of converting a laboratory reward effect into a general diagnosis Berridge & Robinson (2016) Robinson & Berridge (2025).

Attentional Capture, Approach, and Agency

Reward-associated cues can acquire attentional priority and induce approach, providing observable consequences of motivational significance. Yet salience as attentional capture and incentive salience as reward-oriented pursuit are not coextensive constructs. A novel, threatening or highly visible stimulus may draw attention without being wanted, while a remembered reward can guide goal-directed action without reflexive approach to its physical cue. A complete experimental interpretation compares attention, response selection, conditioned reinforcement, and current reward value. Such separations are important in tasks where rapid eye movements or reaction times are used to infer internal motivation Robinson & Berridge (2025) Pool et al. (2016).

Agency in reward seeking includes the capacity to select, inhibit or revise actions in relation to context and valued outcomes. A sudden incentive pulse may bias choice, but deliberate reasoning can also influence what a participant does after encountering the cue. Human PIT experiments show why these processes must be investigated together: goal-directed contributions can appear alongside learned Pavlovian influence. Therefore cue-triggered motivation is best viewed as one contributor to action organization, whose behavioral expression is affected by instructions, uncertainty, action feasibility and competing goals Seabrooke et al. (2017) Allman et al. (2010).

A theoretically decisive design creates circumstances in which an automatic motivational bias predicts one response and an informed representation of the outcome predicts another. Investigators can manipulate reward identity, value, action cost and attention, then examine which variable governs behavior across conditions. Trial-level modeling may detect changing weights on cue influence and outcome-specific planning; questionnaire ratings should be interpreted as a further measurement channel. The objective is causal decomposition of observed behavior, not a binary declaration that a participant is either controlled by cues or acting intentionally Holland (2004) Seabrooke et al. (2017) Badioli et al. (2024).

Interindividual, Developmental, and Contextual Variation

The strength of cue-triggered reward pursuit varies with learning history, physiological sensitivity, stress exposure, opportunity and experience. Vulnerability is central to incentive-sensitization theory because repeated exposure does not produce the same motivational adaptation in every organism. A human study that correlates a PIT score with self-reported impulsivity or craving must control for measurement reliability, reward familiarity, training success and group differences in sensory or motor performance. Correlation can motivate a hypothesis about individual susceptibility, but it cannot identify the direction or mechanism of causation Robinson & Berridge (1993) Robinson & Berridge (2025) Badioli et al. (2024).

Developmental generalization requires particular caution. Children's reward environments, motor abilities, language for expressing pleasure and capacity for explicit outcome reasoning change across age. Older adults may differ in sensory discrimination, health, economic context or task familiarity. Cultural norms influence which rewards are meaningful and whether wanting can be freely acknowledged in a questionnaire. These variables make a one-size-fits-all incentive index scientifically fragile. Cross-group claims require equivalence of task contingencies, observable endpoints and measurement scales before inferential differences are attributed to motivational circuits Pool et al. (2016) Berridge & Kringelbach (2015).

Context also includes artificial feedback environments. Recommendation interfaces, reward schedules and interactive systems can repeatedly expose users to predictive cues. Reward theory offers experimentally testable hypotheses about attention, return behavior, expectancy and cue-induced desire in these environments; it does not warrant inferring subjective pleasure from click frequency. Longitudinal human–artificial interaction studies should compare prediction, momentary liking, willingness to pursue and actual well-being using separate measures. This is a cross-disciplinary research agenda under the Artificial Era horizon, rather than an established empirical claim about the psychology or phenomenal experience of artificial systems Zhang et al. (2009) Pool et al. (2016).

Causal Inference, Replication, and Evidence Hierarchies

Reward science draws on evidence of different causal strength. A neural recording may show that a signal predicts a cue or behavioral response; a randomized pharmacological perturbation may establish that changing a receptor system alters a specified endpoint; a lesion may demonstrate necessity under particular conditions; and a computational model may identify parameters that predict behavior without locating their cellular implementation. Each result answers a different question. For example, a local mu-opioid manipulation increasing positive facial responses supports a causal claim about that intervention and that hedonic assay. It does not by itself establish a general conscious-pleasure mechanism in humans. Strong synthesis preserves these distinctions when moving across experimental scales Peciña & Berridge (2005) Smith & Berridge (2007) Mohebi et al. (2019).

Replication requires a specified target claim. Reproducing a PIT difference under identical conditioning tests reliability within a paradigm, while extending the result to different cues, rewards or species tests generality. In human research, methodological diversity has sometimes made comparison difficult, especially where studies use different transfer phases or classify wanting and liking with divergent instruments. The reviews by Pool and colleagues and Badioli and colleagues demonstrate the value of explicit task definitions, preregistration and cross-design comparison. A positive pooled association, when reported, must still be interpreted in relation to between-study heterogeneity and design-specific biases Pool et al. (2016) Badioli et al. (2024).

Publication bias, selective endpoint reporting and underpowered interactions can distort a literature even when individual findings are technically correct. Experiments claiming an interaction between liking and wanting should report the reliability of both measures, prespecified contrasts and uncertainty intervals. Particularly informative tests contrast the predicted effects of one manipulation on at least two independent endpoints rather than relying on separate significant and nonsignificant tests. A null finding on one task is informative only to the extent that the design had sensitivity to detect a theoretically meaningful effect. Open protocols and accessible data can strengthen independent evaluation without changing the meaning of the underlying construct Pool et al. (2016) Badioli et al. (2024).

Theoretical Disputes and Alternative Explanations

One debate concerns whether incentive salience is required as a separate state-dependent motivational construct or whether flexible, model-based valuation can explain cue-driven behavior. Formal incentive models capture rapid changes in the attraction exerted by a learned cue; cognitive explanations emphasize outcome identity, expectations and planned action. The explanatory challenge is empirical: both mechanisms can produce increased responding in common tasks. Outcome-devaluation studies and human goal-directed PIT findings show that transfer may incorporate deliberative representations. Their presence does not invalidate incentive motivation, but it requires specific experiments to identify when a proposed incentive-state variable adds predictive power Zhang et al. (2009) Seabrooke et al. (2017) Holland (2004).

A second debate addresses dopamine's functional heterogeneity. The teaching-signal account correctly explains key characteristics of reward prediction errors in sampled neurons and tasks; incentive-salience research demonstrates modulation of pursuit without corresponding changes in the hedonic reaction measure. These statements need not compete for exclusive ownership of the transmitter system. Dopamine projections, cell bodies, terminals and time-dependent release dynamics can express different relations to learning and action. The stronger scientific question asks which causal manipulation, neural recording and behavioral endpoint support each proposed computation in a defined circuit Schultz et al. (1997) Berridge (2012) Mohebi et al. (2019).

A third dispute concerns how reward-related adaptations contribute to addiction. Incentive sensitization, habit, negative reinforcement, allostatic responses and control dysfunction offer partially overlapping accounts of persistent harmful pursuit. A clinical episode may contain more than one such process, and the relative contribution may change with withdrawal, treatment and environmental opportunity. Robinson and Berridge's 2025 review explicitly considers competitors and critiques, illustrating a research program in which testable alternatives remain central. Future designs should seek measurements that discriminate the proposed causes of persistence and establish whether the mechanism predicts clinical outcomes above simpler indicators Robinson & Berridge (2025) Berridge & Robinson (2016).

Reward Uncertainty, Predictability, and Cue Strength

Predictability changes both the informational role of reward cues and the motivational impact they can produce. An unexpected reward can elicit a neural response corresponding to a positive prediction error; after learning, the earliest predictive event may instead evoke the response. The temporal shift documented in classic dopamine studies reflects acquired information about expected outcomes. Yet a cue that has become informative is not necessarily maximally attractive in every setting: the motivational force of that prediction also depends on state, learning history, the available response and the possibility of action. A careful design therefore varies expected probability and access to the outcome while measuring learning and pursuit separately. This makes it possible to assess whether changes in choice reflect better predictions, stronger cue attraction, greater uncertainty or a change in opportunity costs Schultz et al. (1997) Berridge (2012) Zhang et al. (2009).

Uncertainty can affect motivation through several routes. A partially reliable cue offers less precise information about whether reward will arrive, while the absence of a definite prediction may sustain attention or exploration. A variable schedule can also change the temporal distribution of opportunities to act and the informational value of new observations. These alternatives can produce similar aggregate approach scores. The incentive-sensitization literature treats uncertainty as a contemporary problem in understanding cue-driven seeking, including cases where motivational attraction remains strong despite uncertain reinforcement. This claim should be tested with factorial manipulations of reward probability, magnitude, temporal predictability and energetic cost. Behavioral analyses need to establish that the animal or participant actually learned the intended contingencies before attributing uncertainty-related performance differences to incentive salience Robinson & Berridge (2025) Schultz et al. (1997).

The computational distinction is instructive. A prediction error concerns a discrepancy at the moment of an outcome or state transition; incentive salience concerns the present motivational magnetism of a learned cue. An uncertain cue may evoke a powerful response because of state-sensitive incentive processes even when its predicted expected value is modest. Alternatively, an organism may pursue the cue because it expects information that will improve future choices. Experiments can separate these proposals by changing the action's informational consequences without changing its immediate reward consequences, by assessing pleasure independently, and by fitting models to prospective held-out trials. No single neural signal establishes both computations at once; their joint role must be demonstrated through converging evidence across conditions Zhang et al. (2009) Mohebi et al. (2019) Pool et al. (2016).

Pleasure in Time: Anticipation, Consumption, and Remembered Value

Reward unfolds across time. A person may anticipate an enjoyable experience, approach a cue announcing it, encounter an outcome that produces momentary pleasure, and later recall the experience when deciding whether to seek it again. Each phase has different available measurements. Anticipation can be studied using prospective ratings, willingness to wait, choice and effort; consummation can be studied through reports collected during the experience or through validated affective reactions; remembrance requires a later judgment under conditions in which memory and reconstruction contribute to the answer. A long delay between reward consumption and liking ratings increases the role of recall. Studies that call all these measurements pleasure risk hiding phase-specific changes, especially when future choice depends more on expectation than on the immediate experience under investigation Kringelbach & Berridge (2009) Berridge & Kringelbach (2015) Pool et al. (2016).

The distinction has direct consequences for interpretation of anhedonic symptoms and craving. A person may expect little enjoyment and therefore decline an activity, while still experiencing positive affect when it occurs. Another may anticipate a reward with intense desire but report muted pleasure on receipt. Both patterns are compatible with the broader architecture of liking, wanting and learning, yet each demands its own evidence. Human studies can combine repeated experience sampling with contemporaneous behavioral tasks, separating predictions before reward from ratings during and after delivery. The validity problem is to identify the temporal referent of every measure: an anticipated feeling, an observed reaction, a remembered evaluation or the effort someone is prepared to expend. These indicators may covary within a person without being interchangeable Berridge & Kringelbach (2015) Pool et al. (2016) Berridge et al. (2009).

An informative experiment also varies sensory identity, deprivation state and cue availability independently. Suppose participants can report expected sweetness before tasting a solution, evaluate its palatability during consumption and later choose whether to work for it. The investigator can compare three theoretically distinct outcomes without presuming that a preference reversal is identical to altered core liking. In animals, taste reactivity offers a complementary way to isolate the immediate hedonic phase, while learned instrumental responses expose decision processes influenced by prior outcomes. Such paired designs support strong claims about functional dissociation when interventions produce independently reliable effects on distinct measures. They also reveal where subjective pleasure remains a specifically human self-report construct and where behavioral analogies provide only a partial bridge across species Grill & Norgren (1978) Wyvell & Berridge (2000) Peciña & Berridge (2005) Pool et al. (2016).

Design Principles for the Next Generation of Reward Studies

A rigorous experiment should specify the distinct outcome classes it intends to measure: hedonic reaction, remembered prediction, attraction to a cue, instrumental vigor, cost sensitivity, choice, consumption or subjective experience. Baseline reward valuation and training must be established before any causal perturbation. If an intervention changes eating, the protocol should check oral motor performance, general arousal and current appetite before attributing the effect to pleasure. If it changes PIT, the analysis should establish that conditioned associations and instrumental training were acquired as intended and distinguish general from outcome-specific transfer Grill & Norgren (1978) Wyvell & Berridge (2000) Badioli et al. (2024).

Neural studies benefit from within-subject tests that dissociate reward prediction, release dynamics, movement and direct consumption. Imaging associations gain explanatory value when combined with behavioral measures sensitive to the construct of interest, while causal animal studies gain translational value when their outcomes can be mapped to validated human tasks. Formal models should generate held-out predictions under changes of reward identity, delay, devaluation or state, and competing models should be compared using criteria appropriate to the data. These designs create a bridge between conceptual claims and discriminating observations without reducing psychological reward to any one recording modality Mohebi et al. (2019) Zhang et al. (2009) Pool et al. (2016).

Longitudinal research is essential to distinguish transient incentive responses from lasting sensitization. Repeated measurement can show whether cues acquire motivational power, whether states modulate that power, and whether the dynamics predict meaningful behavioral outcomes. Human investigations should account for clinical heterogeneity, cultural context, prior learning and ethical restrictions on exposure. Replication across sites and transparent reporting are particularly valuable when translating animal manipulations into claims about depression, addiction, food behavior or digital environments. Conclusions should identify the evidence level and the remaining inferential gap, rather than masking uncertainty with general reward vocabulary Robinson & Berridge (2025) Badioli et al. (2024).

Reward and Psychological Organization in the Artificial Era

Artificial learning systems routinely employ variables called reward, preference or value, but those labels identify formal optimization targets established by a training or interaction protocol. A numerical reward assigned to a machine output has a demonstrable mathematical role in updating or selecting behavior. Whether an artificial system possesses affective experience is a separate scientific and philosophical question that cannot be inferred from the presence of a reward scalar or a preference model. The psychology of reward provides a vocabulary for examining functional analogies across biological and artificial agents, while preserving the evidential requirements specific to each system Schultz et al. (1997) Zhang et al. (2009).

Human–artificial environments are a practical site for reward research. Interfaces can present cues that forecast social response, information, recognition or personalized feedback. Researchers can test how predictive reliability, immediacy, variable scheduling and learned associations influence users' engagement, effort and felt satisfaction. A responsible experiment measures engagement and subjective pleasure separately and assesses how incentives interact with agency, preferences and longer-term well-being. Animal findings about mesolimbic dopamine offer mechanistic hypotheses for human follow-up, yet direct application to interface design requires human behavioral evidence and careful ethical evaluation Pool et al. (2016) Robinson & Berridge (2025).

The broader Artificial Era raises a durable comparative question: which principles of reward organization are general properties of adaptive control, and which depend on biological embodiment, affective systems or social development? A testable program could compare learned predictive representations, state-sensitive goal prioritization, behavioral persistence and counterfactual adjustment across biological and artificial systems. Results would establish functional similarities or differences under explicit operational criteria. They would not automatically settle questions of artificial pleasure, phenomenal consciousness or personal identity. The research horizon is productive precisely because it specifies what must be measured before an ontological conclusion is proposed Berridge (2012) Berridge & Kringelbach (2015).

Conclusion

The cumulative evidence supports a structured account of reward in which hedonic impact, incentive salience and associative learning are partially dissociable functions. Taste-reactivity assays reveal affective responses that cannot be inferred from consumption alone. Localized opioid-sensitive hedonic hotspots make a causal contribution to positive taste reactions within interacting limbic circuits. Dopamine-related interventions and recordings identify important roles in incentive motivation and reward prediction, with effects that depend on neural site, timescale and experimental conditions. Pavlovian-to-instrumental transfer connects learned predictors to action, while devaluation experiments reveal how cue influence can interact with current value and explicit outcome representations Grill & Norgren (1978) Peciña & Berridge (2005) Wyvell & Berridge (2000) Mohebi et al. (2019).

The most significant theoretical achievement is an explanatory separation among liking an outcome, knowing what it predicts, and being motivated to pursue it. These functions can reinforce one another during ordinary adaptive behavior, and their divergence illuminates phenomena including cue-triggered craving, altered motivation and reward-seeking that exceeds actual enjoyment. The evidence is strongest where controlled manipulations produce divergent effects on independent endpoints. Its limits include variation across species, uncertainty about human subjective reports, heterogeneous task definitions and the fact that multiple psychological processes can produce similar observed choices Pool et al. (2016) Robinson & Berridge (2025) Badioli et al. (2024).

Future scholarship should integrate independent hedonic measurement, conditioned approach, effort, prediction updating, state change and temporally resolved circuit observations in designs that adjudicate between competing accounts. Clinical translation calls for longitudinal, ethically reviewed and replicated studies that link well-defined mechanisms to meaningful outcomes. Research on artificial systems can illuminate computational analogies while keeping questions about affect and experience distinct from formal reward optimization. Reward, pleasure and incentive salience consequently form a coherent research object at the intersection of emotion, motivation, learning and behavioral neuroscience, with identifiable empirical foundations and clear problems for continued investigation Berridge (2012) Berridge & Kringelbach (2015) Robinson & Berridge (2025).

References

  • Allman, M. J., DeLeon, I. G., Cataldo, M. F., Holland, P. C., & Johnson, A. W. (2010). Learning processes affecting human decision making: An assessment of reinforcer-selective Pavlovian-to-instrumental transfer following reinforcer devaluation. Journal of Experimental Psychology: Animal Behavior Processes, 36(3), 402–408. https://doi.org/10.1037/a0017876
  • Badioli, M., Degni, L. A. E., Dalbagno, D., Danti, C., Starita, F., di Pellegrino, G., Benassi, M., & Garofalo, S. (2024). Unraveling the influence of Pavlovian cues on decision-making: A pre-registered meta-analysis on Pavlovian-to-instrumental transfer. Neuroscience & Biobehavioral Reviews, 164, Article 105829. https://doi.org/10.1016/j.neubiorev.2024.105829
  • Berridge, K. C., & Robinson, T. E. (1998). What is the role of dopamine in reward: Hedonic impact, reward learning, or incentive salience? Brain Research Reviews, 28(3), 309–369. https://doi.org/10.1016/S0165-0173(98)00019-8
  • Berridge, K. C., Robinson, T. E., & Aldridge, J. W. (2009). Dissecting components of reward: 'Liking', 'wanting', and learning. Current Opinion in Pharmacology, 9(1), 65–73. https://doi.org/10.1016/j.coph.2008.12.014
  • Berridge, K. C. (2012). From prediction error to incentive salience: Mesolimbic computation of reward motivation. European Journal of Neuroscience, 35(7), 1124–1143. https://doi.org/10.1111/j.1460-9568.2012.07990.x
  • Berridge, K. C., & Kringelbach, M. L. (2015). Pleasure systems in the brain. Neuron, 86(3), 646–664. https://doi.org/10.1016/j.neuron.2015.02.018
  • Berridge, K. C., & Robinson, T. E. (2016). Liking, wanting, and the incentive-sensitization theory of addiction. American Psychologist, 71(8), 670–679. https://doi.org/10.1037/amp0000059
  • Berridge, K. C. (2019). Affective valence in the brain: Modules or modes? Nature Reviews Neuroscience, 20, 225–234. https://doi.org/10.1038/s41583-019-0122-8
  • Berridge, K. C., & Grill, H. J. (1984). Isohedonic tastes support a two-dimensional hypothesis of palatability. Appetite, 5(3), 221–231. https://doi.org/10.1016/S0195-6663(84)80017-3
  • Grill, H. J., & Norgren, R. (1978). The taste reactivity test. I. Mimetic responses to gustatory stimuli in neurologically normal rats. Brain Research, 143(2), 263–279. https://doi.org/10.1016/0006-8993(78)90568-1
  • Holland, P. C. (2004). Relations between Pavlovian-instrumental transfer and reinforcer devaluation. Journal of Experimental Psychology: Animal Behavior Processes, 30(2), 104–117. https://doi.org/10.1037/0097-7403.30.2.104
  • Kringelbach, M. L., & Berridge, K. C. (2009). Towards a functional neuroanatomy of pleasure and happiness. Trends in Cognitive Sciences, 13(11), 479–487. https://doi.org/10.1016/j.tics.2009.08.006
  • Mohebi, A., Pettibone, J. R., Hamid, A. A., Wong, J.-M. T., Vinson, L. T., Patriarchi, T., Tian, L., Kennedy, R. T., & Berke, J. D. (2019). Dissociable dopamine dynamics for learning and motivation. Nature, 570, 65–70. https://doi.org/10.1038/s41586-019-1235-y
  • Olds, J., & Milner, P. (1954). Positive reinforcement produced by electrical stimulation of septal area and other regions of rat brain. Journal of Comparative and Physiological Psychology, 47(6), 419–427. https://doi.org/10.1037/h0058775
  • Peciña, S., & Berridge, K. C. (2005). Hedonic hot spot in nucleus accumbens shell: Where do mu-opioids cause increased hedonic impact of sweetness? Journal of Neuroscience, 25(50), 11777–11786. https://doi.org/10.1523/JNEUROSCI.2329-05.2005
  • Pool, E., Sennwald, V., Delplanque, S., Brosch, T., & Sander, D. (2016). Measuring wanting and liking from animals to humans: A systematic review. Neuroscience & Biobehavioral Reviews, 63, 124–142. https://doi.org/10.1016/j.neubiorev.2016.01.006
  • Robinson, T. E., & Berridge, K. C. (1993). The neural basis of drug craving: An incentive-sensitization theory of addiction. Brain Research Reviews, 18(3), 247–291. https://doi.org/10.1016/0165-0173(93)90013-P
  • Robinson, T. E., & Berridge, K. C. (2025). The incentive-sensitization theory of addiction 30 years on. Annual Review of Psychology, 76, 29–58. https://doi.org/10.1146/annurev-psych-011624-024031
  • Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593–1599. https://doi.org/10.1126/science.275.5306.1593
  • Seabrooke, T., Le Pelley, M. E., Hogarth, L., & Mitchell, C. J. (2017). Evidence of a goal-directed process in human Pavlovian-instrumental transfer. Journal of Experimental Psychology: Animal Learning and Cognition, 43(4), 377–387. https://doi.org/10.1037/xan0000147
  • Smith, K. S., & Berridge, K. C. (2007). Opioid limbic circuit for reward: Interaction between hedonic hotspots of nucleus accumbens and ventral pallidum. Journal of Neuroscience, 27(7), 1594–1605. https://doi.org/10.1523/JNEUROSCI.4205-06.2007
  • Tindell, A. J., Berridge, K. C., Zhang, J., Peciña, S., & Aldridge, J. W. (2005). Ventral pallidal neurons code incentive motivation: Amplification by mesolimbic sensitization and amphetamine. European Journal of Neuroscience, 22(10), 2617–2634. https://doi.org/10.1111/j.1460-9568.2005.04411.x
  • Tindell, A. J., Smith, K. S., Berridge, K. C., & Aldridge, J. W. (2009). Dynamic computation of incentive salience: 'Wanting' what was never 'liked'. Journal of Neuroscience, 29(39), 12220–12228. https://doi.org/10.1523/JNEUROSCI.2499-09.2009
  • Wassum, K. M., Ostlund, S. B., Maidment, N. T., & Balleine, B. W. (2009). Distinct opioid circuits determine the palatability and the desirability of rewarding events. Proceedings of the National Academy of Sciences, 106(30), 12512–12517. https://doi.org/10.1073/pnas.0905874106
  • Wyvell, C. L., & Berridge, K. C. (2000). Intra-accumbens amphetamine increases the conditioned incentive salience of sucrose reward: Enhancement of reward 'wanting' without enhanced 'liking' or response reinforcement. Journal of Neuroscience, 20(21), 8122–8130. https://doi.org/10.1523/JNEUROSCI.20-21-08122.2000
  • Wyvell, C. L., & Berridge, K. C. (2001). Incentive sensitization by previous amphetamine exposure: Increased cue-triggered 'wanting' for sucrose reward. Journal of Neuroscience, 21(19), 7831–7840. https://doi.org/10.1523/JNEUROSCI.21-19-07831.2001
  • Zhang, J., Berridge, K. C., Tindell, A. J., Smith, K. S., & Aldridge, J. W. (2009). A neural computational model of incentive salience. PLoS Computational Biology, 5(7), Article e1000437. https://doi.org/10.1371/journal.pcbi.1000437