Contribution
The authors’ summary, a quarter-century on, of the theory they proposed in 1993. It retells how the liking–wanting dissociation was found, says what incentive salience is psychologically, and defends sensitization against the rival view that addicts have too little dopamine. It then extends the theory, tentatively, to behavioural addictions.
Key insight
Addiction is a disorder of ‘wanting’, not of pleasure or withdrawal. “Addiction is not so much about satisfaction, pleasure, need or withdrawal, by this view, as it is about ‘wanting’” (Psychological Features of Incentive Salience). What sensitization changes is how strongly cues pull, and that pull is cut loose both from how much the outcome is enjoyed and from what the person cognitively wants.
Assumptions
- Two senses of wanting. ‘Wanting’ in quotation marks is “a particular form of desire – namely, mesolimbic incentive salience”. Wanting without quotation marks is “a cognitive desire with a declarative goal” (Psychological Features). The theory is about the first.
- ‘Liking’ is measured by taste reactivity. Orofacial reactions to sweet and bitter, homologous across infants, apes, monkeys and rats, are taken as the measure of hedonic impact (introduction).
- BOLD reactivity indicates dopamine. Most human evidence is fMRI, and “most reports of such hyper-reactivity used fMRI measures, which do not directly measure dopamine” (Sensitization in Human Addiction). The authors cite optogenetic and PET work to bridge the gap.
Key results
As reported by the review:
- The founding dissociation. Near-total dopamine depletion left rats’ ‘liking’ reactions to sugar normal while abolishing their motivation to eat. Stimulation of the same system raised ‘wanting’ without raising ‘liking’ (introduction).
- Incentive salience is cue-bound. It is “less connected to cognitive goals and more tightly linked to reward cues, making those cues attention-grabbing and attractive”. “The intensity of the triggered urge depends both on the cue’s reward association and on the current state of dopamine-related brain systems in an individual”. So stress, excitement, appetite or intoxication can amplify it (Psychological Features).
- It dissociates from cognitive desire and from report. It “can occur either in opposition to a cognitive desire or even unconsciously in absence of any cognitive desire”. Subliminal cues can raise motivation “even though the person remains unable to report any change in subjective feelings while motivation increases are revealed in their behavior” (after Figure 1).
- The same system makes dread. “fearful salience is generated by the same mesolimbic circuitry as incentive salience” (Motivational Salience in Desire Versus Dread).
- Sensitization. Repeated drug use, especially at high doses spaced apart, makes the system hyper-reactive. “Once induced, sensitization is very long lasting, and possibly even permanent.” “‘Liking’, by contrast, need not increase with sensitization, and may even decrease. Sensitized ‘wanting’ can persist for years, even if the person cognitively doesn’t want to take drugs, doesn’t expect the drugs to be very pleasant, and even long after withdrawal symptoms have subsided” (Addiction Distorts ‘Wanting’ Versus ‘Liking’).
- Reactive, and gated by context. “A sensitized dopamine system is not always hyper-active, but it is hyper-reactive to drug cues and contexts”. Testing outside drug contexts can hide sensitization entirely (Addiction Distorts; Disentangling Reports).
- Against the dopamine-deficit view. Blunted responses to drugs themselves are compensatory tolerance and fade within weeks. Cue hyper-reactivity does not fade: “if anything, sensitization grows for some time during abstinence”, the incubation of craving (Disentangling Reports).
- Vulnerability. “only about 30% of people who take cocaine actually go on to become long-term addicts”. Genes, sex hormones, stress, route and dose matter, as does a sign-tracking disposition to attribute salience to discrete cues (Individual vulnerability).
- Behavioural addictions and Parkinson’s. “neural sensitization of ‘wanting’ mechanisms may in some cases occur without drugs”, on “emerging” cue-reactivity evidence. Dopamine agonists give about 15% of Parkinson’s patients compulsive gambling, shopping, sex or internet use, “intense and idiosyncratically diverse ‘wants’”, which fade when the medication stops, and “such patients with intense motivations have virtually never been reported to experience intense pleasure from their medications or compulsions” (Other Behavioral Addictions?).
- Disease, qualified. A sensitized person “who faces a series of hundreds of intense temptations may eventually be expected to fail”, so the “brain disease” label is “a reasonably fair description” (Is incentive-sensitization a brain disease?).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | ’Wanting’ and ‘liking’ have dissociable brain mechanisms; dopamine mediates the first only | strong (rat), moderate (human) | lesion and stimulation studies; human dopamine-suppression and imaging studies, reviewed |
| C2 | Incentive salience is cue-triggered and can oppose cognitive desire or run without report | moderate | relapse phenomenology; subliminal-cue studies, cited only |
| C3 | Drugs sensitize the ‘wanting’ system lastingly, and ‘liking’ need not rise | strong (rat), moderate (human) | animal sensitization work; human PET and fMRI cue reactivity |
| C4 | Sensitization is reactive, context-gated, and grows in abstinence | moderate | context-gating and incubation studies, mostly animal |
| C5 | Blunted drug responses in addicts are compensatory tolerance, not the core of addiction | moderate | argued from timescales; partly interpretive |
| C6 | Behavioural addictions may involve sensitization without drugs | weak | ”emerging” cue-reactivity evidence; Parkinson’s dopamine dysregulation as proof of principle |
| C7 | Addiction may fairly be called a brain disease | argued | probabilistic argument from repeated strong temptation |
Concepts
- incentive salience (‘wanting’): motivational magnetism attributed to cues and representations of reward, mesolimbic and dopamine-related.
- cognitive wanting: desire with a declarative goal, which usually goes with ‘wanting’ and can come apart from it.
- incentive sensitization: lasting hyper-reactivity of the ‘wanting’ system to drug cues and contexts.
- incubation of craving: relapse vulnerability that rises over the first month or so of abstinence.
- fearful salience: the negatively valenced mode of the same mesolimbic system.
- sign-tracker / goal-tracker: an animal that approaches the cue that predicts reward, against one that goes to where the reward will arrive.
- dopamine dysregulation syndrome: compulsions produced by dopamine-agonist treatment in Parkinson’s disease.
Connections
The same programme as Berridge and Kringelbach’s review (LIT-1459, NOTE-1110, THEORY-893), which states the dissociation and the ‘liking’ hotspots in more detail. This paper says more about sensitization and addiction. Dayan and Berridge (LIT-584, NOTE-463) give ‘wanting’ a model-based Pavlovian side. Ainslie’s précis (LIT-586, NOTE-443) takes Berridge and Robinson’s “wanted” but not “liked” behaviours as instances of its itch-like cycle. Daw, Niv and Dayan (LIT-580) supply the arbitration view of controllers that QUESTION-266 borrows for measuring potential. Flayelle et al. (LIT-1465) cite incentive sensitization as one of two model-free routes by which technology design promotes uncontrolled use, and propose that unpredictable non-drug rewards can cause it.
Bearing on the record
QUESTION-267: a mechanism for proxy-keyed potential in persons, which meets the term’s kind clause no better than ‘liking’ does.
-
The marks are all there, with a mechanism for each. On TERM-076, potential is priority, persistence and resumption (THEORY-055).
- Priority: sensitized ‘wanting’ wins against “a genuine cognitive desire to abstain”.
- Persistence: it “can persist for years”.
- Resumption: cues reinstate it, and in incubation it grows while interrupted.
The mechanism is a re-keying of strength. How hard a ranking pulls is set by “the cue’s reward association and … the current state of dopamine-related brain systems”, not by what the outcome does for the person, and repeated stimulation inflates it while ‘liking’ and expectation fall. That is a ranking that once tracked the organism’s good and has come apart from it, the question’s personal case, with a named mechanism.
-
But the ordering is over cues. Incentive salience is “tightly linked to reward cues” and “less connected to cognitive goals”. TERM-076’s kind clause says that a ranking of world states, however potential, is not valence “until it ranks the system’s own states by their bearing on its maintenance”. CLAIM-433 used exactly that clause to set ‘liking’ aside, as “an ordering over stimuli”. On this paper’s account ‘wanting’ is one too. There are two ways out, and they cost the same. The kind clause can exclude sensitized ‘wanting’ too, and then the question’s answer for addiction is “not valence at all”. Or the clause can be read so that a ranking of cues counts as a ranking of the system’s prospective states (obtaining, consuming), and then ‘liking’ can be read the same way, as a ranking of consummatory states. That was the rescue CLAIM-438 left open. Either way, the asymmetry CLAIM-433 draws between the two on the kind clause does not survive this paper.
CLAIM-433: the pairing of strength with ‘wanting’ and sign with goal conduciveness. The paper supplies the case the claim’s third defeater asks for. Its recovering addict holds a cognitive goal, abstinence, and relapses on cue. Here incentive salience sets the strength, and the person’s declarative goal and maintenance make the outcome goal-obstructive. Sensitized ‘wanting’ persists even when the person “doesn’t expect the drugs to be very pleasant”. CLAIM-433 resolves the addict as “goal conduciveness positive (the goal is attained)”. That holds only if the goal is the drug: the sign read along the ordering that has prevailed, not along the person’s goals or maintenance. Read so, the claim stays consistent, but its “goal conduciveness” has become conduciveness to whatever prevails. That presupposes an answer to QUESTION-267, namely that proxy-keyed potential is valence with its sign read on the proxy. Read the component process model’s way instead, by the person’s own goals, the relapsing addict has strong potential and negative goal conduciveness, and taking the drug is obstructive. The pairing then predicts negative valence at the moment of taking. The paper does not report on that. It is an empirical question, and it is where the defeater should be tested.
Two things that support the account.
- Strength and sign are separable. “fearful salience is generated by the same mesolimbic circuitry as incentive salience”. The salience system supplies strength for both signs, which fits CLAIM-433’s separation of strength from sign, and CLAIM-440’s view that intensity is not itself signed.
- Potential without report. Subliminal cues raise motivation “even though the person remains unable to report any change in subjective feelings”. This is CLAIM-439’s category of unfelt valence, potential without the system’s model of it, found by the layer-1 against layer-3 dissociation that QUESTION-266 proposes. It is cited, not read here (Winkielman et al. 2005; Childress et al. 2008).
QUESTION-268: measuring potential. “not always hyper-active, but … hyper-reactive”, and testing outside drug contexts can hide sensitization. Potential of this kind is dispositional and gated by context. A conflict record taken where the cues are absent will read it as weak. The paper’s complaint about scanner studies is the measurement problem the question faces, and CLAIM-440 already flags the same dispositional character under the counterfactual eraser.
The organizational side, as analogy only. Dopamine dysregulation syndrome shows that the potential’s strength can be set from outside, with its targets “idiosyncratically diverse”, and that it fades when the drive stops. An incentive system that amplifies whatever metric is cued, as Wells Fargo’s did for cross-sell (CASE-151, CLAIM-432), is the analogue QUESTION-267 has in mind. Nothing in this paper is about organizations, and the record would need an argument that a firm’s incentive system plays the role dopamine does before this grounds anything there.
Not an instruction for machine-learning practice.
Limitations
- A review by the theory’s authors, of mostly their own programme. Rival views (dopamine deficit, habit accounts) are presented in order to be answered.
- The extension to behavioural addictions rests on “emerging” evidence and an unexplained mechanism (“through mechanisms not yet understood”).
- Human evidence is largely fMRI cue reactivity, an indirect index of dopamine.
- ‘Liking’ outside taste is not measured. The human ‘liking’ evidence is ratings.
Open questions
- At the moment of relapse, is the person’s goal-conduciveness appraisal positive or negative, if measured apart from choice? The answer decides CLAIM-433’s third defeater.
- Does sensitization by unpredictable non-drug rewards (Flayelle et al.’s proposal) occur in people, by a measure other than self-reported craving?