Wanting, Liking and Learning

A familiar cue appears. Perhaps it is a notification, a snack left in view or the next episode beginning automatically.

Your attention shifts. Anticipation rises. You act.

Yet when you obtain the outcome, it is less enjoyable than expected. The impulse to pursue it was clear, but the pleasure itself was weaker.

Several processes could have contributed. The cue may have gained significance through repetition. Familiar conditions may have made the response easier to initiate. The outcome may still serve a deliberate goal. Convenience may have made acting easier than choosing an alternative.

The behaviour alone cannot reveal which explanation is correct. What it does reveal is why an important distinction is needed: pursuit and pleasure do not always move together, and learning can influence both without being identical to either.

Wanting, liking and learning are functional distinctions within interacting systems—not three separate inner forces competing for control.

They often align. But one cannot be used as proof of the others.

Reward is not one thing

Across research traditions, an outcome may be called rewarding because it is approached, chosen, pleasurable or reinforcing. Researchers may also use reward language for predicted outcomes, learning signals or cues that acquire motivational significance.

These criteria identify different properties.

An outcome can be pleasurable without being worth substantial effort. A consequence can reinforce behaviour without being consciously enjoyed. A cue can motivate pursuit before the associated outcome is obtained. Something can be judged valuable without generating strong immediate attraction.

Reward is therefore a useful umbrella term only when the relevant function or measurement is clear.

Without that clarification, reward explanations can become circular.

Suppose a person repeatedly performs a behaviour. The repetition is treated as proof that the outcome is rewarding. Its rewarding nature is then used to explain why the behaviour repeats.

Nothing new has been identified. Repetition may reflect pleasure, but it may also reflect learned cues, habit, a deliberate goal, social expectations, convenience or limited alternatives.

This is why reward is not interchangeable with pleasure, reinforcement or value. These processes interact, but they answer different questions.

Wanting: what draws and energises pursuit

In everyday language, wanting can include conscious desire, preference, intention or a deliberate goal.

In the specialised reward literature, technical "wanting" has a narrower meaning. It refers to incentive salience: motivational attraction attributed to an outcome or to cues associated with it.

A cue can become:

• attention-grabbing;
• motivationally significant;
• capable of increasing approach or effort;
• more influential under particular conditions.

A notification sound can pull attention toward a device. The smell of food can increase readiness to approach. A familiar location can evoke anticipation before the expected outcome appears.

This cue-triggered attraction is related to salience, but motivational significance is not identical to attention alone. A cue can stand out without being pursued, while an attractive outcome may guide behaviour without dominating conscious attention.

Technical wanting is also not identical to consciously saying, "I want this."

A cue-triggered response may arise before deliberate evaluation. It can conflict with a longer-term goal or reflective judgment. Someone may feel drawn toward an outcome while deciding not to pursue it. Conversely, a person can deliberately pursue something important without strong immediate attraction.

Cues alter probabilities and priorities. They do not compel action in every encounter.

Wanting is also state-sensitive. Hunger can increase the motivational influence of food cues. Satiety can reduce it. Stress may alter how strongly a learned cue energises effort.

In one human experiment, an acute stress manipulation increased effort in the presence of a cue associated with a sweet odour without increasing reported pleasure from that odour. The finding was bounded to one task, outcome and operational measure, but it demonstrates that cue-triggered pursuit and pleasure reports need not change together (Pool et al., 2015).

Animal research provides stronger experimental separation. In one salt-depletion study, a previously unattractive salt-associated cue gained motivational influence when rats entered a novel physiological state—even before the salty outcome had been re-experienced in that state. The inference concerns operational measures of incentive motivation and hedonic reaction, not human-like conscious desire. Its value is showing how current state can recompute the motivational influence of a learned cue (Tindell et al., 2009).

Wanting is therefore not a fixed property stored inside a cue or outcome. It emerges through an interaction among learning, current state and context.

Nor can it be read directly from one behaviour. Approach and effort provide evidence, but they can also reflect goals, obligations, habits or constraints. Technical wanting is inferred from converging measures rather than observed as one internal quantity.

Liking: the pleasurable quality of the experience

Liking refers to hedonic impact: the pleasurable quality of the experience itself, not merely a favourable judgment about it.

It answers a different question from wanting.

Wanting concerns motivational attraction and pursuit. Liking concerns what the outcome feels like when encountered or consumed.

The two often align, but liking does not guarantee pursuit.

A person may enjoy an activity when invited without being motivated to organise it. They may like a particular food but decide it is not worth the effort, price or delay. Another goal may matter more.

Pleasure can contribute to future motivation without being sufficient to organise pursuit by itself.

Liking is also not a fixed property of an outcome. Pleasure varies with hunger, satiety, context, comparison, attention and the actual quality of the experience.

Several judgments must remain distinct:

• expected pleasure: how enjoyable an outcome is predicted to be;
• experienced pleasure: enjoyment during the experience;
• remembered pleasure: how the experience is later reconstructed;
• preference: which option is judged better than another.

Expected pleasure can contribute to an outcome's expected value, but the two are not identical. Expected value can also include probability, cost, effort, delay, goals and social consequences.

Research operationalises liking differently across species. Human studies commonly use pleasure or pleasantness reports. Animal studies may use species-typical positive hedonic reactions, especially to taste. Those reactions can support a liking-related inference, but they are not verbal reports of subjective experience.

Human food studies using pleasantness ratings and forced-choice measures have found partially different liking and wanting patterns across hunger and satiety. No measure is pure, but the findings support the conclusion that enjoyment and pursuit need not move together (Finlayson et al., 2007).

Consumption does not prove liking. A preference report is not identical to hedonic experience. Low pursuit does not establish low pleasure.

Learning: how experience changes acquired relations

Learning refers to changes in acquired relations, expectations or behavioural tendencies resulting from experience.

It is not one mechanism.

A person or animal may learn that:

• a cue predicts an outcome;
• an action produces an outcome;
• a context signals availability;
• a consequence changes later behaviour;
• a familiar cue makes a practised response easier to initiate.

These relationships are part of the broader learning and reinforcement architecture the Library establishes upstream.

Learning can influence wanting by giving cues predictive and motivational significance. It can influence expected pleasure by shaping what an outcome is believed to provide. It can alter later choice through action–outcome knowledge or reinforcement.

But learning is not identical to wanting, liking or behaviour.

One influential idea in reward learning is prediction error. In a particular learning model, a positive prediction error occurs when an obtained outcome is better—or larger in modelled value—than predicted. An omitted expected outcome can produce a negative error.

In several tasks, phasic responses in some midbrain dopamine neurons track differences between predicted and obtained outcomes in ways captured by reward-prediction-error models. This supports the architecture explained more fully in prediction error drives updating (Schultz, 1998).

Prediction error is not pleasure. It is not identical to conscious surprise in every research use, and it is not the whole learning process.

Reinforcement is similarly specific. A consequence is described as reinforcing when its contingent presentation increases or maintains the relevant behaviour under the tested conditions.

This functional effect does not establish:

• conscious enjoyment;
• present liking;
• current value;
• deliberate desire;
• universal future persistence.

A consequence can alter behaviour without pleasure being the identified mechanism.

When wanting, liking and learning align

In many ordinary situations, the three processes support one another.

A previous experience was pleasurable. Cues associated with it create anticipation. The person pursues the outcome. The new experience is enjoyable. It then provides information that may maintain or update later expectations and behaviour.

Reflective preference may agree with the same pattern. The person expects to enjoy the outcome, wants it, chooses it and likes it when it arrives.

The distinctions remain, but no obvious conflict exposes them.

Separability does not mean independence. Wanting, liking and learning usually operate within an interacting reward architecture. Their divergence matters because alignment is common.

When the processes diverge

Wanting can exceed liking

A familiar cue may still capture attention and energise pursuit even when the eventual pleasure has weakened.

Earlier learning may continue to give the cue motivational significance. Stress or deprivation may amplify attraction. Familiar conditions may make the practised response easier to initiate. Expected pleasure may not yet reflect recent experience.

Animal experiments have shown particularly clear dissociations. Dopamine-related manipulations have increased measured cue-triggered incentive motivation without increasing positive hedonic reactions to the same outcome (Wyvell and Berridge, 2000). Similar separation has been reported in hyperdopaminergic mice pursuing sweet rewards (Peciña et al., 2003).

These are animal operationalisations, not direct reports of human experience.

Human studies involving food, alcohol and stress also provide evidence that wanting-related measures and pleasure reports can diverge under some conditions. But pursuit is never a pure measure of technical wanting, and ordinary divergence is not diagnostic.

Addiction research examines stronger and more persistent forms of wanting–liking separation. Strong wanting, repeated pursuit or reduced enjoyment does not establish pathology by itself.

Liking can exceed wanting

The reverse is also possible.

An outcome can be pleasurable without motivating substantial pursuit. A person may enjoy an experience when it becomes available but rarely plan for it. The effort, cost or delay may outweigh its expected pleasure. A competing goal may matter more.

Pleasure can shape later motivation without determining it.

Low effort does not prove low liking, just as high effort does not prove high pleasure.

Knowledge and behaviour can update at different rates

A person may consciously learn that an outcome has changed while old cues or familiar responses retain influence.

They may know that a notification is unlikely to contain anything important yet still orient toward it. They may know that a familiar route has changed and briefly begin the old turn.

The reverse is also possible: behaviour can change before a person can clearly explain why. Knowledge and action do not have one universal update order.

Persistence after an outcome changes can reflect:

• cue-triggered attraction;
• habitual control;
• difficulty retrieving updated value;
• contextual differences;
• incomplete learning;
• task demands.

A momentary slip does not establish a durable habit.

Human habit research is method-sensitive. Some studies find behaviour that becomes less responsive to changes in outcome value. Other preregistered experiments have struggled to produce clear habit induction using particular procedures (Molinero et al., 2025).

Ordinary repetition therefore cannot prove habit. The fuller boundary between repeated action and habitual control is set out in what makes behaviour habitual.

Repetition does not prove enjoyment

Repeated behaviour may reflect:

• reinforcement history;
• habitual control;
• cue exposure;
• deliberate goals;
• social demands;
• convenience;
• avoidance;
• obligation;
• limited alternatives.

Outcome-insensitive persistence can support a habit inference under appropriate tests. Repetition by itself cannot.

Nor does repetition uniquely identify wanting or liking. The behaviour may continue because it serves a different goal or because the environment makes alternatives costly.

This is also why reinforcement must not be confused with pleasure—a distinction that becomes especially important where avoidance becomes self-reinforcing.

Anticipated pleasure can differ from experience

Pursuit begins before an outcome is obtained, so action often depends partly on prediction.

Those predictions can be inaccurate.

A person may overestimate how satisfying an outcome will be and pursue it more strongly than the eventual experience appears to justify. They may also underestimate an experience and enjoy it more than expected.

Affective-forecasting research has found both patterns. The direction and size of the error vary with the activity, population, context and information available (Moore et al., 2019).

Expectation can influence motivation even when it does not match later pleasure. New experience can update expectation, but updating may be incomplete when feedback is inconsistent or memory emphasises particular moments.

Expected pleasure is therefore neither false pleasure nor experienced liking. It is a prediction about a future hedonic response.

It is one influence on value and action. Delay introduces further changes in valuation, treated separately in the Library, but delay is not the only reason prediction and experience diverge.

Behaviour does not reveal one hidden process

The same action can emerge from different combinations of wanting, liking, learning, goals and constraints.

Approach can reflect cue-triggered attraction, a deliberate goal, social instruction or habit.

Effort can reflect wanting, obligation, identity, expected future value or the absence of easier alternatives.

Consumption can involve hunger, availability, routine, social participation or enjoyment.

Repetition can reflect reinforcement, habit, environmental design or a continuing goal.

Behaviour remains informative, but it requires interpretation alongside other evidence.

Subjective reports answer different questions. A desire report captures conscious experience or judgment; it does not reveal every motivational influence. A pleasure rating provides evidence about reported enjoyment; it does not determine how much effort the person will later invest.

Neural measures provide another class of evidence. Activity associated with cues, learning or outcome receipt can constrain explanations. It does not turn a psychological construct into a directly visible biological object.

No single measure provides complete access to wanting, liking or learning.

Dopamine is not a pleasure chemical

Dopamine is often described as the chemical that makes experiences pleasurable.

That is too simple.

Animal research has repeatedly shown that dopamine-related manipulations can increase cue-triggered pursuit, effort or consumption without corresponding increases in measured positive hedonic reactions. Authoritative reviews therefore distinguish dopamine's contribution to incentive motivation from the neural processes more directly associated with hedonic impact (Berridge and Kringelbach, 2015).

But replacing "dopamine is pleasure" with "dopamine is wanting" would create another error.

Dopamine contributes to several functions involving:

• learning;
• cue responsiveness;
• motivation;
• effort;
• action;
• movement;
• anticipation.

Its central role in movement also demonstrates why it cannot be reduced to reward alone.

Different circuits, receptors and timescales matter. In some tasks, phasic dopamine responses align with reward-prediction-error models. Other dopamine-related effects concern behavioural activation or motivational influence.

Human pharmacological research has also found that L-DOPA can alter expected pleasure for imagined future events. That concerns anticipation, not direct evidence that dopamine created pleasure during consumption (Sharot et al., 2009).

The correct conclusion is not that dopamine is unrelated to pleasure-related behaviour. It influences learning, anticipation and pursuit around potentially pleasurable outcomes.

Contribution is not identity.

Dopamine is not pleasure itself. It is not wanting itself. It is not learning itself. No single neurotransmitter explains reward-related behaviour.

Context changes the relationship

Human wanting is shaped by more than bodily appetite and conditioned sensory cues.

People pursue symbolic outcomes:

• achievement;
• recognition;
• belonging;
• security;
• identity;
• moral commitments;
• imagined futures.

A cue may become motivationally significant because of what it represents socially. Deliberate goals can organise action across time, inhibit immediate pursuit or motivate behaviour without strong current attraction—an architecture the Library develops separately where goals organise behaviour.

Valuation adds another distinction. A person may judge an outcome valuable without feeling strongly drawn toward it. They may experience cue-triggered attraction while judging pursuit unwise. They may perform something unenjoyable because it supports a meaningful future outcome.

Valuation, wanting and liking can therefore disagree.

Environments shape which cues are visible, frequent and easy to act upon. Availability reduces effort. Defaults make repetition easier. Social proof changes expectations. Designed reminders repeatedly return particular outcomes to attention.

Frequent pursuit in such an environment does not prove greater pleasure. Environmental influence changes probabilities and costs; it does not determine every action.

These distinctions do not divide the person into literal inner agents. Wanting, liking and learning are functional descriptions within a wider architecture containing goals, beliefs, bodily states, social meaning and environmental structure—one reason motivation is a family of processes rather than a single force.

Pursuit, pleasure and learning tell different parts of the story

Return to the familiar cue.

It attracts attention. A response begins. The eventual outcome is less pleasurable than expected.

The action alone cannot reveal why. Learned significance may have increased attraction. Familiar conditions may have eased initiation. Expected pleasure may have remained high. Convenience or another goal may have supported the behaviour.

What we can say is that pursuit does not prove pleasure.

The reverse matters too. Pleasure does not guarantee pursuit. Learned influence can remain after judgment changes, while reflective knowledge can still redirect behaviour.

Wanting, liking and learning can diverge without becoming independent systems.

Reward-related behaviour emerges from interacting motivational, hedonic and learned processes. Human action also includes valuation, deliberate goals, symbolic meaning and real environmental constraints.

To understand why something attracts, pleases or persists, we must identify which process is being measured rather than assume that one behaviour—or one neurotransmitter—explains them all.