The Neurobiology of Addiction: Reward, Learning and Neuroadaptation
Drugs and Society
Learning objectives for this lesson:
- Describe the mesocorticolimbic dopamine system and summarize the evidence that drugs used non-medically raise dopamine in the nucleus accumbens through different molecular routes.
- Explain reward prediction error and calculate how a prediction is updated for a natural reward and for a drug that releases dopamine directly.
- Evaluate the hypothesis that drugs distort reinforcement learning, weighing evidence from dopamine recordings and from a blocking experiment.
- Distinguish wanting from liking and explain the incentive sensitization theory of addiction and the evidence for it.
- Explain cue-triggered craving and the incubation of craving, using evidence from studies of animals and people, and evaluate incentive sensitization theory in light of this evidence.
- Describe Koob and Volkow’s three-stage cycle of addiction and explain the shift from positive to negative reinforcement.
- Distinguish tolerance, physical dependence and addiction, and compare withdrawal syndromes across drug classes.
- Explain the iRISA model of prefrontal dysfunction and the cautions needed when interpreting brain images.
- Describe gambling disorder and the reward mechanisms it shares with drug addiction, use Canadian data and guidelines to analyze gambling risk, and apply criteria that guard against overpathologizing gaming and other everyday behaviour.
This course was developed by Dr. Kiffer G. Card.
Reward Circuits and Dopamine as a Teaching Signal
Introduction
In 1954, two psychologists at McGill University in Montreal, James Olds and Peter Milner, implanted fine electrodes in the brains of rats and let each animal press a lever that sent a tiny current through the electrode. When the electrodes sat near the septum and the nucleus accumbens, the rats pressed the lever again and again, in later experiments up to 2,000 times an hour (as described by Kringelbach & Berridge, 2010). Many observers concluded that Olds and Milner had found a “pleasure centre”. Later research linked the circuit to motivation and to learning about rewards, and this section and the next describe that account of what it does.
Lesson 3 introduced receptors, agonists and tolerance, and Lessons 4 to 7 applied them to each class of drugs. This section describes the brain’s system for learning about rewards, the role of dopamine in it and what happens to learning when a drug produces dopamine’s teaching signal directly. Lesson 9 examines why some people develop addiction and others do not, and HSCI 402 develops diagnosis and treatment.
Learning Objectives
- 1.1 Describe the mesocorticolimbic dopamine system, naming the ventral tegmental area, the nucleus accumbens and the prefrontal cortex and the part each plays in reward.
- 1.2 Summarize the evidence from Di Chiara and Imperato (1988) that drugs people use non-medically raise dopamine in the nucleus accumbens.
- 1.3 Explain how opioids, cannabis, nicotine, benzodiazepines, cocaine and amphetamine raise dopamine through different molecular targets, and state the limits of this shared dopamine action as an explanation of addiction.
- 2.1 Define reward prediction error and describe how dopamine neurons respond to unexpected, predicted and omitted rewards.
- 2.2 Calculate a prediction error and a learning update for a natural reward.
- 2.3 Calculate a prediction error and a learning update for a drug that releases dopamine directly, using a simplified form of Redish’s model.
- 3.1 Evaluate the hypothesis that drugs distort reinforcement learning by producing a teaching signal that learning cannot cancel, weighing evidence from dopamine recordings and from a blocking experiment.
1.1 The Brain’s Reward System
Every animal has to learn which foods, places and actions lead to good outcomes. Psychologists call anything that increases the behaviour that produces it a reinforcer, such as food for a hungry animal, and the circuits that use reinforcers to shape behaviour are often called the reward system. This part describes the main pathway, and the two parts that follow set out the evidence that drugs act on it and the routes by which they do so.
Nerve cells, or neurons, signal to one another by releasing chemical messengers called neurotransmitters, which bind to receptors on the receiving cell, as Lesson 3 described. Dopamine is one of these neurotransmitters. A small population of midbrain neurons makes dopamine and sends long fibres, called projections, to distant parts of the brain, so a few cells can influence many regions.
The mesocorticolimbic dopamine system originates in the ventral tegmental area (VTA) of the midbrain and projects most notably to the nucleus accumbens and the prefrontal cortex (Lüscher & Ungless, 2006). Its name combines the midbrain (meso), the cortex (cortico) and the limbic structures involved in emotion and motivation. The branch to the nucleus accumbens is called the mesolimbic pathway and the branch to the prefrontal cortex the mesocortical pathway. The nucleus accumbens lies in the lower, or ventral, part of the striatum, a large structure deep in the forebrain; the upper, or dorsal, striatum becomes important as drug use turns into habit (Berridge & Robinson, 2016; Koob & Volkow, 2016). Figure 8.1 places these structures in a side view of the brain, and the flip cards after it summarize the role of each.
3-D Model 8.1 shows the same pathways inside a standard adult brain, adds a third, the nigrostriatal pathway from the substantia nigra to the dorsal striatum, and previews the brain systems of the three-stage cycle of addiction that Section 3 describes.
1.2 Evidence That Drugs Act on This System
The idea that very different drugs share an action on dopamine was tested in the 1980s with microdialysis, in which a thin probe samples the fluid around neurons in a living animal so that the amount of neurotransmitter outside the cells can be measured. Case 8.1 describes a 1988 study that compared several drugs.
Di Chiara and Imperato measured dopamine outside the cells in the nucleus accumbens and in the dorsal caudate nucleus, a motor area of the dorsal striatum, of freely moving rats. Opiates, ethanol, nicotine, amphetamine and cocaine all increased dopamine in both areas, “but especially in the accumbens”. Drugs that animals avoid, such as kappa opioid receptor agonists, reduced dopamine, and three drugs that the authors classed as not used non-medically, the antidepressant imipramine, atropine and the antihistamine diphenhydramine, did not change it (Di Chiara & Imperato, 1988).
The authors concluded that stimulating limbic dopamine transmission “might be a fundamental property” of such drugs. Their title uses the older phrase “drugs abused by humans”; this course says drugs used non-medically.
Lüscher and Ungless (2006) later called this increase in mesocorticolimbic dopamine “a defining commonality” of addictive drugs. The comparison drugs matter too: imipramine and diphenhydramine act on the brain yet left dopamine unchanged, which suggests that the dopamine effect tracks a drug’s tendency to be taken for its own sake more closely than psychoactivity in general.
1.3 Different Drugs, Different Routes to Dopamine
Drugs reach the same outcome through very different molecular targets. Lüscher and Ungless (2006) grouped them into three classes, and Table 8.1 summarizes that classification with the lessons in this course where each drug appears.
Table 8.1. Three routes by which drugs raise dopamine in the mesocorticolimbic system, following the classification of Lüscher and Ungless (2006).
| Group | Examples | Main action that raises dopamine | Lessons |
|---|---|---|---|
| Drugs acting on G protein-coupled receptors | Opioids, THC, GHB | They inhibit the GABA neurons in the VTA that restrain dopamine neurons, through mu-opioid receptors for opioids and CB1 receptors for THC. | 4, 5, 7 |
| Drugs acting on ion channels | Nicotine, alcohol, benzodiazepines | Nicotine opens nicotinic receptors on dopamine and GABA neurons and their inputs; benzodiazepines may inhibit GABA neurons more than dopamine neurons; how alcohol does it “remains unclear”. | 4, 6 |
| Drugs acting on monoamine transporters | Cocaine, amphetamine, MDMA | Cocaine blocks the dopamine transporter, so dopamine stays outside the cell longer; amphetamine makes the transporter run in reverse. | 6, 7 |
The first group works by disinhibition, which resembles releasing a brake: GABA neurons normally slow nearby dopamine neurons, and a drug that quiets the GABA neurons lets the dopamine neurons fire more (Lüscher & Ungless, 2006). Lesson 5 showed that the mu-opioid receptor quiets the neurons that carry it; in the VTA those quieted neurons are the brake. The third group acts on dopamine terminals in the nucleus accumbens, so cocaine raises dopamine there even though dopamine neurons in the VTA fire more slowly after cocaine. The box below sets out what this shared action does not explain.
The limits of the dopamine story
Case 8.1 came from rats, and human imaging studies give a more varied picture. Reviewing the human imaging evidence, Nutt, Lingford-Hughes, Erritzoe and Stokes (2015) found strong evidence that stimulants raise dopamine in the human striatum and some evidence for alcohol, but “little evidence, if any” for cannabis and opioids. Dopamine release and receptor availability were reduced in people with stimulant or alcohol dependence but not in people with opioid, nicotine or cannabis dependence, and the authors noted that the dopamine view of addiction has not led to new treatments. In animals, mice that cannot make dopamine still learn to prefer a place where they received morphine, and Lüscher and Ungless (2006) reported that most experts in the field agreed that some aspects of reward, such as euphoria, do not depend on dopamine. Dopamine is therefore one part of the explanation, strongest for stimulants.
The reward system has now been located, and drugs raise dopamine in it by several routes. The next part turns to what dopamine does there.
2.1 Dopamine as a Teaching Signal
If dopamine simply signalled pleasure, every reward would release it in the same way every time. Recordings from single dopamine neurons in monkeys during the 1990s suggested that the neurons respond to surprise about rewards. This part explains that finding, and the next gives a simple way to calculate it.
2.1.1 Learning from Surprise
Learning depends on the gap between what we expect and what we get. Wolfram Schultz (2016) describes pressing a button at random on a drink machine in Japan whose labels he could not read and receiving exactly the blackcurrant juice he wanted. The outcome was better than his one-in-six expectation, and next time he pressed the same button.
This gap is called the reward prediction error: the reward received minus the reward predicted. It is positive when an outcome is better than expected, zero when it matches the prediction and negative when it is worse. Learning models such as the Rescorla-Wagner model use it to update expectations, so the error shrinks as the world becomes predictable (Schultz, 2016).
2.1.2 What Dopamine Neurons Do
Schultz, Dayan and Montague (1997) combined recordings from monkey dopamine neurons with a mathematical theory of learning and proposed that the neurons signal errors in the prediction of rewards. Figure 8.2 shows the three situations behind this proposal.
Schultz (2016) reports that most midbrain dopamine neurons in humans, monkeys and rodents are activated by more reward than predicted, stay at baseline for a fully predicted reward and are depressed when the reward is smaller than predicted or absent. The response also moves in time: neurons that first fire to a new reward stop firing once it is predictable and fire instead to the cue that predicts it (Koob & Volkow, 2016), as in Figure 8.2(b). Experiments in rodents that switch dopamine neurons on with light at precise moments suggest that the signal causes learning: an artificial burst at the time of reward can rescue learning that would otherwise fail (Keiflin & Janak, 2015). Gershman and colleagues (2024) describe the prediction error account as the most influential account of these brief dopamine signals, and judge that in its original form it is probably too simple to explain all of their properties.
2.2 A Simple Model of Learning by Prediction Error
The pattern in Figure 8.2 can be written as two equations, a simplified form of the temporal difference models used by Schultz and colleagues (1997) and Redish (2004) with timing within each trial left out. Rewards are measured in hypothetical units. Equation 8.1 defines the prediction error, and Equation 8.2 updates the prediction.
Here δ (delta) is the prediction error, r is the reward received, V is the value currently predicted and α (alpha) is a learning rate between 0 and 1. A learning rate of 0.3 means that each trial closes 30 percent of the gap between prediction and outcome.
Worked Example 8.1 (hypothetical numbers). A cue is followed by a reward of r = 10 units, the starting prediction is V = 0 and α = 0.3. On trial 1, δ = 10 − 0 = 10 and the new value is 0 + 0.3 × 10 = 3. On trial 2, δ = 10 − 3 = 7 and the value becomes 5.1. On trial 3, δ = 4.9 and the value becomes 6.57. By trial 10 the error is about 0.40 and the value about 9.72, so the dopamine burst at the time of reward has nearly disappeared, as in Figure 8.2(b). If the reward is then omitted (r = 0) when the value is 10, the error is −10, the dip in Figure 8.2(c), and the value falls to 10 + 0.3 × (−10) = 7. Calculator 8.1, after Worked Example 8.2, checks each step.
On this account, dopamine neurons therefore signal the difference between what was expected and what arrived, so a natural reward stops producing a burst once it is predicted. The next part asks whether drugs obey the same rule.
2.3 A Teaching Signal That Learning Cannot Cancel
Drugs that raise dopamine directly may escape the rule that teaching stops once a reward is predicted, because the dopamine they release does not depend on what the brain expected. This part presents that hypothesis and a simplified model of it for calculation; the part that follows weighs the evidence for and against it.
Drugs reach the brain through the blood, and Schultz (2016) argues that their effects “mimic a positive dopamine reward prediction error, as they are not compared against a prediction”. Redish (2004) built this idea into a computational model by adding a “noncompensable drug-induced dopamine increase” to a temporal difference learning model, which then “over-selects actions leading to drug receipt”. Because the drug’s teaching signal is set by its pharmacology, in the model the value of actions that lead to the drug keeps increasing without limit and eventually exceeds the value of natural rewards (Keiflin & Janak, 2015). Redish himself judged it “highly unlikely” that values in a real brain grow without bound, because biological compensation would limit them. Equation 8.3 is a simplified, single-step version of the drug prediction error in Redish’s model, written in the notation of Equations 8.1 and 8.2; it leaves out the discounting of future rewards and the value of the next state (Redish, 2004). It applies on trials when the drug is delivered; on trials without the drug, the ordinary prediction error of Equation 8.1 applies, which is why a drug-paired cue can still lose value.
Here D is the extra dopamine signal from the drug, in hypothetical units. While the prediction is low, the drug adds its boost to an ordinary error; once the prediction catches up, an ordinary error would fall to zero, but the drug’s error stays at D.
Worked Example 8.2 (hypothetical numbers). Keep r = 10 and α = 0.3 and add a drug boost of D = 4. On trial 1, δdrug is the larger of 10 − 0 + 4 = 14 and 4, so it is 14, and the value becomes 4.2. On trial 2 the error is 9.8 and the value 7.14; on trial 3 the error is 6.86 and the value 9.20; on trial 4 the error is 4.80 and the value 10.64. From trial 5 the ordinary part of the error is negative, so the error stays at 4 and the value rises by 0.3 × 4 = 1.2 every trial. By trial 30 the drug-paired cue has a value of about 41.8, more than four times the value of 10 at which the natural reward settles. Use Calculator 8.1 to repeat any step of Worked Examples 8.1 and 8.2.
3.1 How Drugs Distort Reinforcement Learning
If the model is right, drugs distort reinforcement learning: actions that lead to the drug keep gaining value long after a natural reward would have stopped teaching. This part shows how that distortion unfolds over many trials and then weighs the evidence for and against it.
3.1.1 Watching the Two Signals over Many Trials
The difference between a natural reward and a drug appears only over many repetitions. Interactive 8.1 repeats Equations 8.1 to 8.3 trial after trial and plots either the prediction error at the time of reward, which corresponds to the dopamine burst, or the learned value of the cue. Set the drug boost to zero and the two lines become identical.
Prediction error simulator: a natural reward compared with a drug that releases dopamine directly
3.1.2 Evidence for and Against the Model
Redish’s model makes two testable predictions: the dopamine response to a drug should not shrink as the drug becomes expected, and learning effects that depend on prediction errors should fail with drugs. In the studies described here, the first found support and the second did not.
Keiflin and Janak (2015) review recordings in rats that pressed a lever for food or for intravenous cocaine. The dopamine burst to food progressively decreased as the reward became expected, while cocaine produced a delayed, long-lasting increase in dopamine transients that did not appear to depend on expectation and persisted after the task was well learned, as the model predicts.
The second prediction concerns blocking: if one cue already fully predicts a reward, a second cue added later is learned poorly, because the reward produces no prediction error. If drugs always produce an error, blocking should fail with drugs. Panlilio, Thorndike and Schindler (2007) paired a tone with each cocaine injection in rats and then added a light; the tone blocked learning about the light, so their results “fail to support a potentially powerful explanation of addiction”. The model also treats every dose as identical, while repeated drug use changes the circuits that compute prediction errors, and Keiflin and Janak (2015) concluded that the role of dopamine prediction errors in addiction “remains an intriguing open question”, and research on it continues. Try it 8.1 asks you to explore the model before weighing this evidence.
In Interactive 8.1, keep the defaults (reward size 10, learning rate 0.3, drug boost 4, 30 trials). (1) Record the prediction error on trial 30 for both learners. (2) Switch to “Learned value of the cue” and record both values on trial 30. (3) Choose an omission on trial 15 and describe the natural reward’s error on that trial and its value afterwards. (4) Explain in two sentences why even a small boost eventually produces a large difference in value. (5) State what the blocking experiment implies about this prediction.
Suggested answers: (1) About 0 for the natural reward and 4 for the drug. (2) About 10.0 and 41.8. (3) The error is about −10 and the value falls to about 7, then recovers. (4) In the model, the drug error never falls below the boost, so the value grows by the learning rate times the boost on every trial without limit; a larger boost only speeds the growth. (5) Because blocking still occurred with cocaine, the drug did not keep producing an error in that experiment, so the model is at best incomplete.
Summary
This section traced the mesocorticolimbic dopamine system from the ventral tegmental area to the nucleus accumbens and prefrontal cortex and showed that drugs raise dopamine there by three molecular routes. In the most influential account, dopamine neurons signal reward prediction errors, and Redish’s model proposes that drugs distort learning with an error that prediction cannot cancel; recordings with cocaine support part of the model and the blocking experiment does not. Kringelbach and Berridge (2010) report the suggestion that the electrodes of Olds and Milner produced a form of “wanting” for more stimulation, with little evidence of pleasure. Section 2 explains why wanting and pleasure can come apart.
Key Takeaways
- 1.1 The mesocorticolimbic dopamine system originates in the ventral tegmental area of the midbrain and projects to the nucleus accumbens (the mesolimbic pathway) and the prefrontal cortex (the mesocortical pathway). The nucleus accumbens lies in the ventral striatum, the main target where drugs raise dopamine, while the dorsal striatum becomes important as drug use turns into habit and the prefrontal cortex supports planning and stopping actions.
- 1.2 Using microdialysis in freely moving rats, Di Chiara and Imperato (1988) found that opiates, ethanol, nicotine, amphetamine and cocaine raised dopamine, especially in the nucleus accumbens, while imipramine, atropine and diphenhydramine did not. Because those comparison drugs are psychoactive, the dopamine effect appears to track a drug’s tendency to be taken for its own sake more closely than psychoactivity in general.
- 1.3 Drugs reach dopamine by three routes: opioids and THC quiet the GABA neurons that restrain dopamine neurons (disinhibition), nicotine, alcohol and benzodiazepines act on ion channels, and cocaine and amphetamine act on the dopamine transporter. A 2015 review of human imaging found strong evidence that stimulants raise striatal dopamine and little evidence that opioids and cannabis do, so dopamine is one part of the explanation of addiction, strongest for stimulants.
- 2.1 Reward prediction error is the reward received minus the reward predicted. In the most influential account, dopamine neurons fire for a better-than-expected outcome, stay at baseline for a fully predicted reward and dip below baseline when a predicted reward is omitted, and with learning their response moves from the reward to the cue that predicts it, a signal that teaches the brain what to repeat.
- 2.2 In the simple model, the prediction error is δ = r − V and the new value is V + α × δ. With a reward of 10 units and a learning rate of 0.3, the value rises from 0 to 3, 5.1 and 6.57 and reaches about 9.72 by trial 10, so the error and the dopamine burst at the time of reward nearly disappear; omitting the reward then gives an error of −10.
- 2.3 For a drug that releases dopamine directly, the simplified prediction error is the larger of (r − V + D) and D, so it never falls below the drug’s boost. With a boost of 4, the value of the drug-paired cue keeps rising by 1.2 units a trial once the prediction has caught up and reaches about 41.8 after 30 trials, more than four times the value at which the natural reward settles.
- 3.1 Redish’s model predicts that drug-induced dopamine responses do not shrink as the drug becomes expected and that blocking should fail with drugs. Cocaine produced long-lasting dopamine transients that persisted after the task was learned, which supports the first prediction, but a tone already paired with cocaine still blocked learning about a light, so the model is at best incomplete and the role of prediction errors in addiction remains an open question.
Reflection
A friend says: “Dopamine is the brain’s pleasure chemical, so drugs are addictive simply because they feel good.” Write a response of 150 to 200 words that uses three pieces of evidence from Section 1: (a) Di Chiara and Imperato’s 1988 finding that opiates, ethanol, nicotine, amphetamine and cocaine raised dopamine, especially in the nucleus accumbens, while imipramine, atropine and diphenhydramine did not; (b) the finding that dopamine neurons fire for an unexpected reward, stay at baseline for a fully predicted reward and dip when a predicted reward is omitted; and (c) Worked Examples 8.1 and 8.2, in which a natural reward of 10 units settles at a value of about 10 while a drug with a dopamine boost of 4 reaches a value of about 41.8 after 30 trials. Say what your friend gets right, what the evidence adds and one limit of the evidence.
Minimum 20 characters required.
Question 1: In the microdialysis study by Di Chiara and Imperato (1988), which result best supports the idea that raising limbic dopamine is a shared property of drugs people use non-medically?
Question 2: A dopamine neuron in a trained monkey fires to a light that predicts juice. On one trial the light appears but no juice follows. What does the prediction error account predict at the moment the juice was expected?
Question 3: Using Equations 8.1 and 8.2 with a reward r = 10, a predicted value V = 5 and a learning rate of 0.2, what are the prediction error and the new value for a natural reward?
Question 4: Which finding most directly challenges Redish’s (2004) proposal that drugs always produce a prediction error that learning cannot cancel?
Incentive Sensitization: Wanting, Liking and Cue-Triggered Craving
Introduction
By the 1980s most researchers accepted that dopamine produces the pleasure of rewards. Kent Berridge and Terry Robinson set out to add one more piece of evidence: they depleted nearly all the dopamine in rats’ brains and expected the animals’ facial reactions of pleasure to sweet tastes to fade. The reactions were “completely normal”, but the rats no longer sought or ate food (Berridge & Robinson, 2016). Berridge, Venier and Robinson (1989, as described by Berridge & Robinson, 2016) proposed that dopamine is needed for wanting a reward and that its pleasure, or liking, depends on other systems.
This section explains the difference between wanting and liking, sets out the incentive sensitization theory that Robinson and Berridge (1993) built on it and examines evidence that drug cues can trigger craving long after drug use has stopped. Section 1 showed that dopamine neurons respond to cues that predict rewards; this section explains how those cues acquire their pull.
Learning Objectives
- 4.1 Distinguish “wanting” (incentive salience) from “liking” (the pleasure of a reward) and describe the experiments in rats that separated them.
- 4.2 State the four main claims of Robinson and Berridge’s incentive sensitization theory of addiction.
- 4.3 Contrast sensitization with tolerance, explain what it means for a brain system to be hyper-reactive to drug cues and describe what affects who becomes sensitized.
- 5.1 Describe the evidence that drug cues trigger craving and brain responses in people, including the study by Childress and colleagues (1999).
- 5.2 Explain the incubation of craving and summarize the evidence for it in rats and in people who smoke.
- 5.3 Evaluate the strengths and limits of incentive sensitization theory as an account of addiction.
4.1 Wanting and Liking
In everyday speech, wanting something and liking it seem to be one feeling. Berridge and Robinson argue that the brain treats them as separate processes with separate circuits. This part defines them, describes how they are measured and compares them.
4.1.1 Two Processes with Two Meanings
Liking, which Berridge and Robinson write in quotation marks to mark it as a technical term, is the hedonic impact of a reward: the pleasure of consuming it. Wanting is incentive salience, a psychological process that transforms the perception of stimuli, “imbuing them with salience, making them attractive, ‘wanted’, incentive stimuli” (Robinson & Berridge, 1993). Incentive salience is what makes the smell of a bakery or a familiar logo grab attention and pull a person toward a reward.
Berridge and Robinson (2016) also separate incentive salience from cognitive wanting, the conscious desire for a goal that a person can explain and plan for. The two usually go together, but they can split: a person in recovery may sincerely want to stay abstinent and still feel a surge of incentive salience when a drug cue appears, which can lead to use against the person’s own goals.
4.1.2 Measuring Liking in Animals and in People
Pleasure is subjective, so researchers need indirect measures. A sweet taste elicits relaxed facial expressions and rhythmic tongue movements, and a bitter taste elicits gapes of disgust; these reactions are similar in human infants, apes, monkeys and rats, so they can measure liking across species (Berridge & Robinson, 2016).
With this measure, Berridge’s laboratory found in rats that liking is generated by a few hedonic hotspots that are tiny, chemically specific and easily disrupted; one in the nucleus accumbens is about one cubic millimetre in a rat and probably about one cubic centimetre in humans. Stimulating a hotspot with the brain’s own opioid or cannabinoid-like chemicals amplifies liking reactions. Wanting, by contrast, is produced by large, widespread systems that include mesolimbic dopamine, which may be one reason why intense desires are more common in life than intense pleasures (Berridge & Robinson, 2016).
Berridge and Robinson (2016) describe human studies that point the same way: suppressing dopamine in volunteers did not reduce their pleasure ratings of cocaine or amphetamine even when it reduced their desire to take more, and imaging studies found that dopamine changes tracked ratings of wanting more closely than ratings of liking. A systematic review by Pool and colleagues (2016) found, however, that human experiments on wanting and liking had produced contradictory results, so the human evidence is mixed. Table 8.2 brings these findings together.
Table 8.2. “Liking” and “wanting” compared, as described by Berridge and Robinson (2016).
| Feature | “Liking” (hedonic impact) | “Wanting” (incentive salience) |
|---|---|---|
| What it is | The pleasure of consuming a reward | The pull that makes a reward and its cues attractive |
| Measure in animals | Facial reactions to tastes | Effort to obtain a reward and approach to its cues |
| Brain systems | Small, fragile hedonic hotspots | Large systems that include mesolimbic dopamine |
| Main chemical signals | Opioid and cannabinoid-like signals | Dopamine, with glutamate and other signals |
| Effect of removing dopamine | Reactions to sugar remain normal | Motivation to seek food disappears |
| Effect of repeated drug use in susceptible people | Need not increase and may decrease | Can become sensitized, especially to drug cues |
The “pleasure centre” revisited
Section 1 opened with the rats of Olds and Milner (1954), which pressed a lever thousands of times for stimulation near the nucleus accumbens. Kringelbach and Berridge (2010) note that when electrodes were placed in the brains of human patients in the 1950s and 1960s, some patients pressed a stimulation lever compulsively, but their reports never made clear that the stimulation caused real pleasure. They report the suggestion of some researchers that the electrodes produced a form of wanting with little liking, which would make the experiments of Olds and Milner an early demonstration of the separation described in this section.
Evidence, most of it from rats, therefore suggests that wanting and liking depend on different systems and can be pulled apart. The next part explains how Robinson and Berridge used this separation to build a theory of addiction.
4.2 The Four Claims of Incentive Sensitization Theory
A theory of addiction must explain why people keep seeking a drug that may no longer give them much pleasure, and why craving returns long after withdrawal has ended. This part sets out the incentive sensitization theory and its four claims; the next explains sensitization and asks who becomes sensitized.
Robinson and Berridge (1993) asked why people with addiction crave drugs, why craving persists after long abstinence and whether wanting drugs is the same as liking them. Their answer rests on four claims, which the flip cards below summarize.
The fourth claim carries the theory’s most striking prediction: sensitized incentive salience “can produce addictive behavior (compulsive drug seeking and drug taking) even if the expectation of drug pleasure or the aversive properties of withdrawal are diminished and even in the face of strong disincentives, including the loss of reputation, job, home and family” (Robinson & Berridge, 1993). Figure 8.3 sketches this divergence.
4.3 Sensitization
4.3.1 What Sensitization Is
Sensitization is a progressive increase in a response to the same dose of a drug given repeatedly, often measured in animals as increased movement (Keiflin & Janak, 2015). It is the mirror image of tolerance, which Lesson 3 defined, and both can occur in one person for different effects of the same drug. Berridge and Robinson (2016) explain that compensatory changes, such as fewer receptors, produce tolerance to the drug’s high and withdrawal when use stops, while the same people show heightened brain responses to the cues that trigger their urges.
Sensitization develops especially when drugs are taken repeatedly at high doses spaced apart, as in weekend binges, and once induced it is very long lasting and possibly permanent (Berridge & Robinson, 2016). It changes glutamate neurons that project from the cortex to the nucleus accumbens as well as dopamine neurons, and the same review describes imaging studies in which relatively low oral doses of amphetamine sensitized mesolimbic dopamine release in people with no history of drug use.
A sensitized system is quiet most of the time and overreacts when drug cues and contexts appear, producing pulses of dopamine release and motivation that last seconds or minutes (Berridge & Robinson, 2016). This hyper-reactivity fits a common pattern in addiction: long periods of relative calm interrupted by sudden, intense urges in particular places or situations.
4.3.2 Who Becomes Sensitized
Incentive sensitization theory locates part of the difference between people in how readily the brain attaches incentive salience to cues. Robinson’s laboratory distinguishes sign-trackers, rats that approach and work for a cue that predicts food, such as a lever that appears, from goal-trackers, which go straight to where the food will arrive. Cocaine and opioid cues acquire greater incentive salience in sign-trackers, although both types eventually show addiction-like intake after weeks of intermittent access to cocaine, and Berridge and Robinson (2016) suggest that most individuals may become sensitized if they take drugs for long enough at high enough doses.
The intensity of wanting also depends on the person’s state when a cue appears. Stress, emotional excitement, appetites and intoxication heighten dopamine reactivity and amplify cue-triggered urges, which Berridge and Robinson (2016) suggest is one reason why many people find it hard to stop at one dose and why even happy life events can make a return to use more likely. Lesson 9 examines other sources of vulnerability.
Interactive 8.2 asks you to apply the distinctions in this section and Section 1 to real research findings. For each finding, decide whether it mainly concerns liking, wanting or learning by prediction error.
Wanting, liking or learning? Classify research findings
Sensitization is therefore long lasting, triggered by cues and states, and able to grow while pleasure fades. The next part turns to the human evidence on cues and craving.
5.1 Cue-Triggered Craving
People with addiction often describe urges set off by particular sights, places or feelings. Goldstein and Volkow (2011) list the stimuli that become drug cues: the objects used to take the drug, the people who supply it and the emotional states that drug use once relieved or triggered. This part explains how cues acquire their power.
5.1.1 From Conditioning to Craving
Lesson 3 introduced Pavlovian conditioning through Shepard Siegel’s work on conditioned tolerance. Cues can also acquire motivating power: Koob and Volkow (2016) define conditioned reinforcement as the process by which a neutral stimulus paired with a reinforcer, such as a drug, comes to strengthen behaviour and becomes a reinforcer in its own right. Section 1 showed the dopamine side of this process, the burst that moves to the cue, and incentive sensitization theory adds that in some people the cue’s pull becomes exaggerated. Case 8.2 describes one test in the human brain, from 1999.
Anna Rose Childress and colleagues at the University of Pennsylvania used positron emission tomography (PET), an imaging method that can measure regional blood flow, to study 14 men who had used cocaine and had completed withdrawal treatment and 6 comparison participants who had never used it. During a video of cocaine-related scenes, the men who had used cocaine reported craving and showed increased blood flow in the amygdala and anterior cingulate cortex, limbic regions, compared with a video unrelated to drugs. The comparison participants showed no such pattern (Childress et al., 1999).
The authors concluded that limbic activation is one component of cue-induced craving. The sample was small, and the participants who had used cocaine were all men, which limits generalization.
5.1.2 Dopamine, Cues and Craving
Later studies tested the dopamine prediction of the theory. Koob and Volkow (2016) summarize studies in which cues associated with cocaine craving increased dopamine release in the striatum, the amygdala and parts of the prefrontal cortex, and PET studies have found that “the greater the cue-induced dopamine release the greater the craving” (Leyton & Vezina, 2013, as quoted in Berridge & Robinson, 2016). Many other studies used functional magnetic resonance imaging (fMRI), which measures changes in blood oxygen and infers neural activity from them; Berridge and Robinson (2016) caution that fMRI does not measure dopamine directly. Cues therefore trigger craving and brain activation, and the next part asks how long that power lasts.
5.2 The Persistence of Drug Memories
Craving can return months or years after a person has stopped using a drug and long after withdrawal has ended. Berridge and Robinson (2016) argue that sensitized wanting “can persist for years”, even when the person does not want to use and does not expect the drug to be pleasant. This part examines a related effect that researchers have measured directly.
5.2.1 The Incubation of Craving
In 1986, Gawin and Kleber suggested from clinical observations that cue-induced craving in people who used cocaine increased over the first weeks of abstinence and stayed high for long periods (Pickens et al., 2011). Grimm, Hope, Wise and Shaham (2001) found a similar effect in rats: after the rats stopped self-administering cocaine, their lever pressing in response to cocaine-paired cues increased with time since their last dose (as described by Pickens et al., 2011). The effect became known as the incubation of craving, and later studies found it after self-administration of heroin, nicotine, methamphetamine, alcohol and even sugar solutions (Pickens et al., 2011). Case 8.3 describes a 2011 test in people.
Gillinder Bedi and colleagues paid 86 adults who smoked at least 10 cigarettes a day, and who were not seeking treatment, to stop smoking for 7, 14 or 35 days, with abstinence checked daily, and then measured their craving in response to smoking cues. On some measures, cue-induced craving was greater after 35 days than after 7 days, and on no measure did it decrease, even as background craving and withdrawal symptoms faded (Bedi et al., 2011).
The authors described this as initial evidence of incubation of craving in humans. A study of 56 inpatients who used methamphetamine found a similar contrast in timing: depressive and psychotic symptoms largely resolved within a week of abstinence, while craving lasted at least five weeks (Zorick et al., 2010).
The evidence from people is more limited than the evidence from animals. A systematic review by Bergeria and colleagues (2024) found that 31 of the 44 animal studies it included reported greater drug seeking after longer abstinence, while none of the 20 clinical studies of opioid craving it included found that craving rose with longer abstinence. Incubation is therefore supported mainly by animal studies, with some support from people who smoke, and research on whether it occurs in people is evolving.
5.2.2 What Persistence Means
The explanations of persistent craving in this section overlap, and the accordion below summarizes how each accounts for an urge that returns long after a person has stopped using.
Through Pavlovian learning, cues that regularly accompany drug use acquire the power to trigger responses and to reinforce behaviour in their own right (Koob & Volkow, 2016). Lesson 3 described the related phenomenon of conditioned tolerance.
In rats, cue-induced drug seeking increases with time since the last dose, and in people who smoke, cue-induced craving was greater after five weeks than after one week of abstinence on some measures (Pickens et al., 2011; Bedi et al., 2011). A 2024 systematic review found the effect mainly in animal studies, while clinical studies of opioid craving showed no such rise (Bergeria et al., 2024).
These findings help explain why Koob and Volkow (2016) describe addiction as a “chronically relapsing disorder”. A return to drug use months after withdrawal has ended is consistent with this research, and it says little about a person’s character or commitment. Many people also stop using without returning to it, and Lesson 9 examines natural recovery and the debate over describing addiction as a brain disease. HSCI 402 develops treatment and recovery support; the final part of this section assesses the theory as a whole.
5.3 Weighing the Theory
Robinson and Berridge proposed incentive sensitization theory in 1993, and in 2016 they reviewed the evidence and judged that its key tenets still seemed well supported (Berridge & Robinson, 2016). The tabs below set out its strengths, limits and open questions.
The theory explains why craving can be intense when pleasure is modest, why urges follow cues and states and why risk persists after withdrawal ends. Its predictions have been tested in animals and people. It also connects to evidence beyond drugs: Berridge and Robinson (2016) report that some patients with Parkinson’s disease who are treated with medications that directly activate dopamine receptors develop intense compulsive motivations, including for gambling, which usually fade when the medications are stopped. Section 4 examines a 2010 study of how common such disorders are.
Much of the evidence comes from rats, human experiments on wanting and liking have produced contradictory results (Pool et al., 2016), and many human studies used fMRI, which does not measure dopamine directly. Some studies find blunted dopamine responses when people with addiction take a drug; Berridge and Robinson (2016) reply that these concern the drug itself, while responses to drug cues are heightened. If most people would become sensitized with enough exposure, other factors must explain who reaches that point (Lesson 9).
Researchers debate how incentive sensitization fits with the prediction error account in Section 1 and with the stress and self-control systems in Section 3. Koob and Volkow (2016) combine exaggerated incentive salience with habits, negative emotional states and weakened executive control in a three-stage model, which Section 3 presents. Whether sensitization can occur without drugs is a question for Section 4.
A hypothetical person who stopped using cocaine three months ago, and who wants to remain abstinent, says the drug “stopped being much fun” long before he quit, and that walking past the street where he used to buy it now produces a stronger urge than in his first week without it. In 100 to 150 words, explain (1) how liking could fall while wanting stayed high, (2) why the urge is tied to a place and (3) why it might be stronger at three months than at one week, naming a study for each point.
Suggested answer: Repeated use can sensitize the dopamine-based wanting system while liking, which depends on small opioid and cannabinoid hotspots, stays level or declines (Robinson & Berridge, 1993). The street is a drug cue that has become a reinforcer in its own right, and such cues trigger limbic activation and craving in people who have used cocaine (Childress et al., 1999); a sensitized system overreacts to them. The stronger urge at three months fits the incubation of craving, seen in rats after cocaine (Grimm et al., 2001) and in people who stopped smoking (Bedi et al., 2011).
Summary
This section separated wanting from liking: dopamine-depleted rats still like sugar but no longer seek it, and in rats liking depends on small hotspots and wanting on large dopamine-based systems, while human findings are mixed. Incentive sensitization theory proposes that repeated use makes the wanting system hyper-reactive to drug cues in susceptible people, so craving can grow while pleasure fades, and studies of the incubation of craving, mostly in animals, suggest that cue-induced urges can grow during abstinence. Section 3 adds the other half of the picture: what happens when the drug wears off and how the brain systems that stop actions change.
Key Takeaways
- 4.1 Liking is the hedonic impact of a reward, generated in rats by small, fragile hotspots that respond to opioid and cannabinoid-like signals; wanting, or incentive salience, makes a reward and its cues attractive and depends on large systems that include mesolimbic dopamine. Rats depleted of dopamine still showed normal liking reactions to sugar but no longer sought food, and suppressing dopamine in volunteers reduced their desire for stimulants without lowering their pleasure ratings, so the two processes can come apart, although other human experiments on wanting and liking have produced contradictory results.
- 4.2 Robinson and Berridge (1993) claim that addictive drugs share an action on mesocorticolimbic dopamine, that this system attributes incentive salience to associated events, that repeated use sensitizes it in some individuals and that sensitized wanting can occur independently of liking and withdrawal. Drug seeking can therefore persist when the expected pleasure has faded and in the face of strong disincentives.
- 4.3 Sensitization is a progressive increase in the response to the same dose given repeatedly, the mirror image of tolerance; it develops especially with high doses spaced apart and can be very long lasting. A sensitized system is quiet most of the time and overreacts to drug cues and contexts. Sign-trackers attribute more incentive salience to cues, most individuals may become sensitized with enough exposure, and stress, excitement and intoxication amplify cue-triggered urges.
- 5.1 Through conditioned reinforcement, cues paired with a drug become reinforcers in their own right. Childress and colleagues (1999) found that a cocaine video triggered craving and limbic activation in men who had used cocaine but not in comparison participants, and PET studies have found that greater cue-induced dopamine release goes with greater craving, although fMRI does not measure dopamine directly.
- 5.2 In the incubation of craving, cue-induced drug seeking in rats increases with time since the last dose, and in people who smoke, cue-induced craving was greater after 35 days than after 7 days of abstinence on some measures (Bedi et al., 2011). Craving can therefore return long after withdrawal has ended, which helps explain why addiction is described as a chronically relapsing disorder without saying anything about a person’s character.
- 5.3 Incentive sensitization theory explains intense craving with modest pleasure, urges that follow cues and states, and risk that persists after withdrawal, and it connects to the compulsive motivations that some people develop when treated with dopamine medications for Parkinson’s disease. Its limits include heavy reliance on rats and on fMRI, blunted dopamine responses to drugs themselves and the open question of who becomes sensitized and how the theory fits with prediction errors and stress systems.
Reflection
A hypothetical friend stopped smoking five weeks ago. Her withdrawal symptoms have faded, but she says the urge she feels when she sees a lighter or steps outside on a work break is stronger now than in her first week, and she worries that something is wrong with her. In 150 to 200 words, explain what may be happening using (a) the difference between wanting and liking, (b) conditioned reinforcement and incentive sensitization, and (c) the study by Bedi and colleagues (2011), in which 86 adults who smoked were paid to abstain for 7, 14 or 35 days and cue-induced craving was greater at 35 days than at 7 days on some measures and did not decrease on any measure. End with one implication of this evidence for the coming months.
Minimum 20 characters required.
Question 1: Berridge and colleagues removed nearly all the dopamine from rats’ brains. Which result led them to separate wanting from liking?
Question 2: Which statement best describes sensitization as used in incentive sensitization theory?
Question 3: In the study by Bedi and colleagues (2011), people who smoked were paid to abstain for 7, 14 or 35 days. What did the researchers find?
Question 4: Why do Berridge and Robinson (2016) caution that many brain imaging studies of cue reactivity do not measure dopamine directly?
The Addiction Cycle: Reinforcement, Dependence and Prefrontal Control
Content note. This section discusses dangerous withdrawal and, briefly, overdose after a loss of tolerance. If you would like support, the 9-8-8 Suicide Crisis Helpline is available by call or text at 9-8-8, the National Overdose Response Service can be reached at 1-888-688-6677, and the Hope for Wellness Help Line for Indigenous people can be reached at 1-855-242-3310.
Introduction
Koob and Volkow (2016) define drug addiction as “a chronically relapsing disorder, characterised by compulsion to seek and take the drug, loss of control in limiting intake, and emergence of a negative emotional state (eg, dysphoria, anxiety, irritability) when access to the drug is prevented”; in diagnostic language it falls under substance use disorder (Lesson 1). Sections 1 and 2 explained compulsion through learning and wanting. This section explains the negative emotional state that appears when the drug is gone and the changes in self-control that make limiting intake difficult.
It moves from reinforcement to a three-stage model of addiction, separates tolerance, physical dependence and addiction, compares withdrawal across drug classes and ends with the prefrontal cortex. Koob and Volkow note that “only some substance users make this transition”; Lesson 9 asks why.
Learning Objectives
- 6.1 Define positive and negative reinforcement and explain how the motivation for drug use can shift from one to the other.
- 6.2 Describe Koob and Volkow’s three-stage cycle of addiction, naming the main brain regions and chemical changes in each stage.
- 6.3 Distinguish within-system and between-system neuroadaptations, with an example of each.
- 7.1 Distinguish tolerance, physical dependence and addiction, and explain why DSM-5 does not count tolerance and withdrawal from supervised medical use toward a diagnosis.
- 7.2 Compare withdrawal syndromes across drug classes in their features, timing and medical risks.
- 8.1 Explain the iRISA model of prefrontal dysfunction.
- 8.2 Evaluate what brain imaging can and cannot show about self-control in addiction.
6.1 Positive and Negative Reinforcement
Psychologists distinguish two ways in which a consequence can strengthen a behaviour, and the distinction explains how the reasons for drug use can change over time. This part defines both and describes the shift between them.
6.1.1 Two Kinds of Reinforcement
Positive reinforcement strengthens a behaviour by adding something pleasant, as when praise makes a child more likely to repeat a helpful act. Negative reinforcement strengthens a behaviour by removing something unpleasant, as when the relief of a headache makes a person more likely to take the same pain reliever next time. Negative reinforcement is different from punishment, which weakens a behaviour; both kinds of reinforcement make a behaviour more likely.
Drugs can reinforce in both ways. Early use is often driven by positive effects such as euphoria; with repeated use, relief from withdrawal and negative emotion becomes a reason in its own right. Benowitz (2010) describes how, in people who smoke, nicotinic receptors recover during overnight abstinence, craving and withdrawal begin and the first cigarette relieves them (Lesson 6). Case 8.4 shows the same shift with opioids.
Weiss and colleagues (2014) asked 653 people seeking treatment for dependence on prescription opioids, in a multi-site treatment trial in the United States, why they had started and why they continued to use. Among participants with chronic pain who had begun prescription opioids mainly for pain relief, 56.5 percent said that their main reason for current use was to avoid withdrawal. Kosten and Baxter (2019) cite this study in a review, quote people who describe using only so as not to feel sick and report another study in which withdrawal was the reason most often given for moving from prescription opioids to heroin. They conclude that avoiding withdrawal “often becomes the most powerful force driving continued use”.
6.1.2 From Impulsive to Compulsive Use
Koob and Volkow (2016) connect this shift to a change in the character of use. They define impulsivity as “a predisposition toward rapid, unplanned reactions to internal and external stimuli without regard for the negative consequences” and compulsivity as “perseverative, repetitive actions that are excessive and inappropriate”. Impulsive acts often bring gratification, while compulsions often reduce tension or anxiety. In their account, use moves from impulsive to compulsive as its drive shifts from positive to negative reinforcement, although both can coexist at every stage.
The idea that a pleasant drug effect is followed by an opposing unpleasant state has a long history. Solomon and Corbit (1974) proposed an opponent-process theory of motivation, and Koob and Volkow (2016) build on the proposal that “opponent processes that are a general feature of biological systems also act to limit reward”, a concept they call anti-reward. Figure 8.4 sketches how the balance between a drug’s effect and the opposing process can change with repeated use.
Relief from an unpleasant state can therefore sustain drug use once the pleasant effects have faded. The next part places this shift in a model of the whole addiction cycle.
6.2 The Three-Stage Cycle of Addiction
Koob and Volkow (2016) describe addiction as “a three-stage, recurring cycle” that “worsens over time” and involves changes in the brain’s reward, stress and executive function systems. They present the stages as a heuristic framework, a simplified map for organizing research, drawn from the psychology of self-regulation failure, psychiatry and brain imaging. Figure 8.5 shows the cycle, and the tabs that follow describe each stage.
This stage involves the rewarding effects of drugs, the development of incentive salience and the formation of drug-seeking habits, through changes in dopamine and opioid peptides in the basal ganglia, the group of deep structures that includes the striatum. Koob and Volkow describe PET studies in which intoxicating doses of alcohol and drugs released dopamine and opioid peptides in the ventral striatum, with fast, steep release of dopamine associated with the subjective “high”; the box on the limits of the dopamine story in Section 1 noted that the human evidence for dopamine release is strongest for stimulants. With repetition, activity in the ventral striatum recruits loops that engage the dorsal striatum, a shift linked to habits and to compulsive responding (Koob & Volkow, 2016). Sections 1 and 2 described the learning and wanting processes of this stage.
This stage involves chronic irritability, emotional pain, malaise, dysphoria, stress and a loss of motivation for natural rewards. In animals, every major drug produces raised reward thresholds during withdrawal, meaning that stimulation which was rewarding before now produces less reward, and these elevations precede and closely track increases in drug intake. Two kinds of change contribute: reduced function of the dopamine reward system and recruitment of brain stress chemicals, such as corticotropin-releasing factor (CRF) and dynorphin, in the extended amygdala. Human imaging studies show reduced sensitivity of reward circuits to natural rewards during this stage (Koob & Volkow, 2016).
This stage, often called craving, involves disrupted signals from the prefrontal cortex and the insula, a cortical region involved in awareness of the body’s internal state, to the basal ganglia and extended amygdala, including signals carried by the neurotransmitter glutamate. Koob and Volkow (2016) propose two opposing systems: a “Go” system that drives craving and engages habits through the basal ganglia, and a “Stop” system in the prefrontal cortex that weighs the value of choices and suppresses emotional responses. In addiction, the Stop system is weakened. The last two parts of this section return to the prefrontal cortex.
6.3 Within-System and Between-System Neuroadaptations
A neuroadaptation is a lasting change in the nervous system produced by repeated exposure to a drug. Koob and Volkow (2016) distinguish two kinds, and the distinction explains why withdrawal so often looks like the opposite of the drug’s effect.
In a within-system neuroadaptation, “the primary cellular response element to the drug” adapts to neutralize the drug’s effects, and the persistence of these opposing changes after the drug is gone produces withdrawal. In the human imaging studies that Koob and Volkow review, for example, the striatal dopamine response to amphetamine or methylphenidate was 50 percent lower in people who had stopped using and 80 percent lower in people actively using, compared with people without a substance use disorder (Koob & Volkow, 2016). Ashton (2002) described the same logic for benzodiazepines: stopping suddenly after long-term use can replace sleep with insomnia, relaxation with muscle tension, calm with anxiety and the drugs’ anticonvulsant effect with seizures.
In a between-system neuroadaptation, brain systems other than those producing the drug’s reward are recruited by chronic use. The stress system that uses CRF is disturbed by long-term use of every major drug, with raised CRF in the amygdala during acute withdrawal (Koob & Volkow, 2016), and in rats blocking CRF receptors prevents the anxiety and increased nicotine intake that follow nicotine abstinence (Benowitz, 2010). 3-D Model 8.2 lets you follow a within-system neuroadaptation at a dopamine synapse in the nucleus accumbens, from the bursts of dopamine described in Section 1 to the smaller dopamine release and the fewer available D2 receptors (one type of dopamine receptor) that imaging studies have found in people with stimulant or alcohol dependence (Nutt et al., 2015). The next part separates three terms that these adaptations often blur.
7.1 Tolerance, Physical Dependence and Addiction
The words tolerance, dependence and addiction are often used as if they meant the same thing. They describe different processes, and confusing them can lead to poor care and to stigma. This part defines each and explains how the diagnostic manual separates them.
7.1.1 Three Different Ideas
Tolerance, as Lesson 3 explained, is a reduced response to a drug after repeated use, so that a larger dose is needed for the same effect. Physical dependence is a state of adaptation in which stopping or sharply reducing a drug produces a withdrawal syndrome, a characteristic set of symptoms for that drug class. Both are predictable consequences of the neuroadaptations described above and can occur in anyone who takes certain drugs regularly, including as prescribed. Addiction, in contrast, is defined by compulsive seeking and taking, loss of control and continued use despite harm (Koob & Volkow, 2016). A person can be physically dependent without addiction, as many patients taking prescribed medications are, and a person can have an addiction with mild physical withdrawal, as the comparison of withdrawal syndromes below shows for stimulants and Section 4 shows for gambling.
7.1.2 How DSM-5 Handles Medical Use
The DSM-5 criteria for substance use disorder, which Lesson 1 introduced, include tolerance and withdrawal among eleven criteria, and a diagnosis requires two or more. The work group that wrote the criteria noted a problem: stimulants, opioids, sedatives and other drugs “can produce tolerance and withdrawal as normal physiological adaptations when used appropriately for supervised medical purposes”. A patient taking such a medication could reach the threshold of two criteria without any other sign of a disorder. The work group therefore stated that, under these conditions, tolerance and withdrawal without other criteria “do not indicate substance use disorders and should not be diagnosed as such” (Hasin et al., 2013). The same group reported a continuing tension in the field over terminology: some experts favour addiction as the general term and want dependence reserved for tolerance and withdrawal (Hasin et al., 2013). The box below explains why the distinction matters in practice.
Why the distinction matters
Treating physical dependence as addiction can stigmatize people who take prescribed medications exactly as directed and can lead to abrupt discontinuation, which Lesson 4 showed to be dangerous for benzodiazepines and alcohol. Kosten and Baxter (2019) note that most patients using opioid analgesics long term for pain not caused by cancer do not meet the criteria for opioid use disorder (although one study they cite found a lifetime prevalence of 35 percent in this group), and that the 2016 guideline of the United States Centers for Disease Control and Prevention, which they cite, recommended tapering such medications slowly. The reverse error also matters: mild withdrawal, as with stimulants or gambling, does not rule out addiction.
Interactive 8.3 asks you to apply these definitions to hypothetical patterns of use. Decide whether each pattern shows tolerance only, physical dependence without a substance use disorder, or a pattern consistent with addiction, then check the feedback.
Tolerance, physical dependence or addiction? Classify hypothetical patterns
Tolerance, physical dependence and addiction are therefore separate ideas. The next part compares withdrawal syndromes across drug classes, which differ widely in their features and their dangers.
7.2 Withdrawal Syndromes by Drug Class
Each drug class produces its own withdrawal syndrome, and the pattern often mirrors the drug’s acute effects: depressants produce withdrawal marked by overactivity of the nervous system, while stimulant withdrawal is marked by low mood and fatigue. Table 8.3 summarizes the main classes, with the lessons where each drug was introduced. HSCI 402 develops the clinical management of withdrawal.
Table 8.3. Withdrawal syndromes by drug class: common features, typical timing and notes on risk, from the sources named in each row.
| Drug class | Common features | Typical timing | Notes on risk |
|---|---|---|---|
| Alcohol (Lesson 4) | Tremor, sweating, rapid heart rate, insomnia, anxiety, headache; in more severe cases hallucinations, seizures and delirium tremens | Starts about 6 hours after the last drink; seizures 6 to 48 hours after; delirium tremens begins 48 to 72 hours after and can last up to 2 weeks | Can be life-threatening; with early detection and treatment, expected mortality is 1 percent or less (Jesse et al., 2017) |
| Benzodiazepines (Lesson 4) | Anxiety and panic, insomnia and nightmares, muscle tension, perceptual changes | Depends on the drug’s half-life (Lesson 4) | Rapid withdrawal, especially from high-potency drugs, can cause seizures; slow tapering minimizes symptoms (Ashton, 2002) |
| Opioids (Lesson 5) | Aches, cramps, nausea, vomiting, diarrhea, anxiety, insomnia, sweating, runny nose, dilated pupils, yawning, gooseflesh | Short-acting opioids: begins within 12 hours, peaks at 36 to 72 hours, tapers over 4 to 7 days; methadone: milder but may last 2 weeks or more | Severe distress drives continued use (Kosten & Baxter, 2019); after withdrawal, lost tolerance raises overdose risk (Lesson 5) |
| Methamphetamine (Lesson 6) | Depressive symptoms and, early in abstinence, psychotic symptoms; craving | In one inpatient study, depressive and psychotic symptoms largely resolved within a week, while craving lasted at least 5 weeks | Withdrawal symptoms other than craving were mild on average (Zorick et al., 2010) |
| Nicotine (Lesson 6) | Irritability, depressed mood, restlessness, anxiety and craving | Begins during overnight abstinence | Anxiety and stress are powerful incentives to smoke again (Benowitz, 2010) |
| Cannabis (Lesson 7) | Irritability, anxiety, sleep difficulty, decreased appetite, depressed mood, physical symptoms such as shakiness or sweating | Develops within about a week of stopping heavy use; often peaks between days 2 and 6; CB1 receptors recover within about 4 weeks | Mainly mood and behavioural symptoms of light to moderate intensity (Bonnet & Preuss, 2017) |
| Hallucinogens and inhalants (Lessons 4 and 7) | No withdrawal syndrome is recognized in DSM-5 | Not applicable | Evidence was judged insufficient to add one (Hasin et al., 2013) |
Table 8.3 shows that the medical danger of withdrawal does not track the strength of addiction. Withdrawal from alcohol and benzodiazepines can be life-threatening, which is why medical supervision and gradual tapering matter. Opioid withdrawal is intensely unpleasant, and its main danger lies in what follows: as tolerance falls during abstinence, a dose that a person once took regularly can cause an overdose, the risk that Lesson 5 described after release from prison. Stimulant and cannabis withdrawal are milder physically, yet methamphetamine craving outlasted the other symptoms by weeks in the study by Zorick and colleagues (2010), which fits the persistence of craving described in Section 2. The rest of this section turns from the body’s adaptations to the brain systems that control behaviour.
8.1 Prefrontal Control and the iRISA Model
The third stage of Koob and Volkow’s cycle involves the prefrontal cortex, the part of the brain most involved in planning, weighing consequences and stopping actions. Goldstein and Volkow (2011) reviewed a decade of imaging studies of the prefrontal cortex in addiction. This part describes their model, and the next sets out the cautions that apply to it.
Goldstein and Volkow (2011) proposed that disrupted prefrontal function produces a syndrome of impaired response inhibition and salience attribution (iRISA). The syndrome has three parts: excessive salience attributed to the drug and drug cues, reduced sensitivity to non-drug rewards and a reduced ability to inhibit harmful or disadvantageous behaviour. As a result, drug seeking becomes a main motivational drive at the expense of other activities. The first part corresponds to the incentive salience of Section 2; the second to the reduced reward sensitivity of the withdrawal stage; the third to the weakened Stop system.
The regions most often implicated are the orbitofrontal cortex, the anterior cingulate cortex and the dorsolateral prefrontal cortex. Goldstein and Volkow (2011) describe dorsal regions, such as the dorsolateral prefrontal cortex, as supporting higher-order control and ventral regions, such as the medial orbitofrontal cortex, as supporting more automatic, emotion-related processes. In their model, input from the dorsal regions normally suppresses automatic responses that could lead to drug taking, and during craving and withdrawal drug-related responses begin to override it. They describe the resulting cognitive problems as “mild, yet pervasive”, affecting attention, working memory, decision making and the ability to wait for larger rewards. They also report a mismatch between self-report and behaviour: people who used cocaine reported how engaged they were in a task, and their reports did not match their actual performance, which suggests reduced insight into their own motivation.
8.2 What Brain Images Can and Cannot Show
Goldstein and Volkow (2011) describe imaging studies that found reduced grey matter in the prefrontal cortex of people with addiction to alcohol, cocaine, methamphetamine, heroin and nicotine, with losses of up to 20 percent, and in people with alcohol addiction these reductions persisted from 6 to 9 months up to 6 years or more of abstinence. A meta-analysis of 60 imaging studies by Pando-Naude and colleagues (2021) found less grey matter, on average, among people who use substances, in clusters centred on the anterior cingulate cortex, the thalamus and the insula. Interpreting these findings requires care, and the box below sets out the main cautions.
Reading brain images with care
Most of these imaging studies compare people with addiction to people without it at one point in time, so they cannot show whether a difference came before drug use or after it, a limit that the authors of the 2021 meta-analysis note for the studies they pooled (Pando-Naude et al., 2021). Goldstein and Volkow (2011) note both possibilities: the persistence of prefrontal changes into long abstinence suggests pre-existing factors that might predispose people to addiction, while the absence of such changes in adolescents who use alcohol or cannabis suggests that they could be a dose-dependent consequence of use. Group averages also hide wide variation among individuals. A PET study described by Goldstein and Volkow (2011) suggests that the prefrontal cortex can be engaged, too: when people who used cocaine were told to inhibit craving while watching a cocaine video, activity in part of the orbitofrontal cortex fell. Research on what these brain differences mean continues. Lesson 9 takes up the wider debate over whether such findings justify calling addiction a brain disease.
A hypothetical man has drunk heavily every day for years. Drinking now makes him feel “normal” more than good, his hands shake in the morning until he drinks, he plans his day around drinking and he has failed to cut down several times. (1) Identify a feature of each stage of Figure 8.5 in his account. (2) Is his drinking now mainly positively or negatively reinforced? (3) Using Table 8.3, why could stopping suddenly without medical support be dangerous? (4) Name one iRISA function that his failed attempts might reflect, and one caution about that conclusion.
Suggested answer: (1) Daily heavy drinking (binge and intoxication), morning tremor relieved by drinking (withdrawal and negative affect) and planning around drinking (preoccupation and anticipation). (2) Mainly negatively reinforced, since drinking relieves withdrawal. (3) Alcohol withdrawal can bring seizures within 48 hours and delirium tremens after 48 to 72 hours (Jesse et al., 2017). (4) Impaired response inhibition; but imaging findings describe group averages and cannot show whether such a difference caused his drinking or resulted from it.
Summary
This section explained how drug use can shift from positive to negative reinforcement as opposing processes grow, and set out Koob and Volkow’s three-stage cycle and its brain systems. Within-system adaptations oppose the drug’s own action, and between-system adaptations recruit stress systems. Tolerance and physical dependence can occur with prescribed use and differ from addiction; withdrawal varies by class in features and danger; and the iRISA model links addiction to weakened prefrontal control, with cautions about what imaging can show. Section 4 asks whether similar mechanisms can produce addiction without any drug.
Key Takeaways
- 6.1 Positive reinforcement strengthens a behaviour by adding something pleasant and negative reinforcement by removing something unpleasant; both make a behaviour more likely, unlike punishment. With repeated use, relief from withdrawal can become the main reason to use, as for the 56.5 percent of treatment seekers with chronic pain in a 2014 study who had begun prescription opioids for pain and said they continued mainly to avoid withdrawal. In Koob and Volkow’s account, use moves from impulsive to compulsive as opponent processes grow and its drive shifts from positive to negative reinforcement.
- 6.2 Koob and Volkow describe a recurring cycle that worsens over time: binge and intoxication, linked to the basal ganglia (reward, incentive salience and habits); withdrawal and negative affect, linked to the extended amygdala (raised reward thresholds and stress chemicals such as CRF and dynorphin); and preoccupation and anticipation, linked to the prefrontal cortex and insula (craving and a weakened Stop system). They present the stages as a heuristic framework for organizing research.
- 6.3 In a within-system neuroadaptation, the drug’s own target adapts to oppose it, as when, in the imaging studies that Koob and Volkow review, striatal dopamine responses to stimulants were 50 percent lower in people who had stopped using and 80 percent lower in people actively using. In a between-system neuroadaptation, other systems are recruited, such as the CRF stress system, whose blockade in rats prevents the anxiety and increased nicotine intake that follow abstinence.
- 7.1 Tolerance is a reduced response to repeated doses, physical dependence is an adaptation revealed by a withdrawal syndrome on stopping, and addiction is compulsive seeking, loss of control and continued use despite harm. DSM-5 does not count tolerance and withdrawal from supervised medical use toward a diagnosis, because treating physical dependence as addiction can stigmatize patients and lead to dangerous abrupt discontinuation, while mild withdrawal does not rule out addiction.
- 7.2 Withdrawal syndromes often mirror a drug’s acute effects. Alcohol and benzodiazepine withdrawal can bring seizures and be life-threatening; opioid withdrawal begins within 12 hours and peaks at 36 to 72 hours, and lost tolerance afterwards raises overdose risk; stimulant and cannabis withdrawal are milder physically, while in one inpatient study methamphetamine craving outlasted other symptoms by weeks. The medical danger of withdrawal therefore does not track the strength of addiction.
- 8.1 Goldstein and Volkow’s iRISA model links disrupted prefrontal function to excessive salience for the drug and drug cues, reduced sensitivity to non-drug rewards and a reduced ability to inhibit harmful behaviour, so drug seeking becomes a main motivational drive. The orbitofrontal, anterior cingulate and dorsolateral prefrontal cortices are most often implicated, and the resulting cognitive problems are mild yet pervasive, with reduced insight into one’s own motivation.
- 8.2 Imaging studies report prefrontal grey matter reductions of up to 20 percent that can persist into long abstinence, but most compare groups at one point in time and cannot show whether a difference came before drug use or after it. Group averages hide wide individual variation, and the prefrontal cortex can still be engaged, as when people in a PET study who were told to inhibit craving showed lower orbitofrontal activity.
Reflection
Consider two hypothetical patients. Patient A took a prescribed opioid exactly as directed for six weeks after surgery; when it was stopped abruptly she had aches, yawning, gooseflesh and diarrhea, and she then tapered with her physician and stopped. Patient B uses methamphetamine in binges, has tried and failed to cut down several times, lost a job because of use and keeps using despite frightening paranoia; when he stops, he has mainly low mood and tiredness. A colleague says: “Patient A is the one with the addiction, because her withdrawal was more obvious.” In 150 to 200 words, respond using the definitions of tolerance, physical dependence and addiction; the DSM-5 work group’s decision that tolerance and withdrawal from supervised medical use do not, on their own, indicate a substance use disorder; the three-stage cycle; and the methamphetamine findings in Table 8.3 (in one inpatient study, depressive symptoms resolved within a week and craving lasted at least five weeks). Explain one harm that could follow from the colleague’s view.
Minimum 20 characters required.
Question 1: A person continues drinking mainly because it relieves morning tremor and anxiety. Which process does this illustrate?
Question 2: In Koob and Volkow’s (2016) three-stage cycle, which pairing of stage and brain system is correct?
Question 3: According to the DSM-5 work group (Hasin et al., 2013), how should tolerance and withdrawal be treated in a patient taking a prescribed opioid exactly as directed under medical supervision?
Question 4: Which withdrawal syndrome in Table 8.3 carries a risk of seizures and delirium tremens and can be life-threatening without treatment?
Gambling Disorder and Other Behavioural Addictions
Introduction
In 2018, nearly two-thirds (64.5 percent) of people in Canada aged 15 or older reported gambling in the past year, most often by buying a lottery or raffle ticket (Rotermann & Gilmour, 2022). For most, gambling causes no harm; for some it becomes a pattern with the features of addiction. In 2013 the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) placed gambling disorder in the chapter that had been reserved for substance-related disorders (Hasin et al., 2013).
This section asks whether the mechanisms of Sections 1 to 3 can produce addiction without a drug. It covers gambling disorder, shared reward mechanisms, online and sports betting in Canada, and gaming disorder. HSCI 402 develops prevention, treatment and regulation.
Learning Objectives
- 9.1 Explain why DSM-5 moved gambling disorder into the chapter on substance-related and addictive disorders, and list its criteria and threshold.
- 9.2 Describe the prevalence of gambling and of gambling problems in Canada and worldwide, using dated survey and review estimates.
- 9.3 Explain how gambling engages the reward mechanisms of Sections 1 and 2, including the evidence from near misses and from dopamine medications in Parkinson’s disease.
- 9.4 Describe the growth of Ontario’s regulated online gambling market since April 2022 using its published data.
- 9.5 Calculate the expected loss from a total amount bet and a house edge, and compare the result with Canada’s Lower-Risk Gambling Guidelines.
- 9.6 Describe gaming disorder as defined in ICD-11 and apply criteria that guard against overpathologizing everyday behaviour.
9.1 From “Pathological Gambling” to Gambling Disorder
Gambling means staking money or something of value on an uncertain outcome in the hope of winning more. This part traces how gambling problems came to be classified as an addiction and sets out the diagnostic criteria; the next summarizes how common gambling problems are.
Pathological gambling entered the DSM in 1980, and in the fourth edition (1994) it was listed among the impulse-control disorders. The DSM-5 work group recommended moving it to the substance chapter because it often co-occurs with substance use disorders and resembles them in symptoms, biological dysfunction, genetic liability and treatment approaches (Hasin et al., 2013). The chapter was renamed “Substance-Related and Addictive Disorders”. DSM-5 also renamed the condition gambling disorder because “the term pathological is pejorative and redundant”, removed the criterion of committing illegal acts to finance gambling, and lowered the threshold for diagnosis from five criteria to four of the remaining nine within 12 months (Hasin et al., 2013; Petry et al., 2014). Table 8.4 lists the nine criteria alongside the substance use disorder features they parallel.
Table 8.4. The nine criteria for gambling disorder in DSM-5 (four or more in 12 months), worded from the interview items in Challet-Bouju et al. (2016) and the DSM-5 changes in Petry et al. (2014), with the parallel feature of substance use disorders.
| Gambling disorder criterion (past 12 months) | Parallel in substance use disorders |
|---|---|
| Preoccupied with gambling: reliving past gambling, planning the next bet or thinking of ways to get money to gamble | Craving and preoccupation |
| Needs to gamble with increasing amounts of money to achieve the desired excitement | Tolerance |
| Repeated unsuccessful efforts to control, cut back or stop gambling | Unsuccessful efforts to cut down |
| Restless or irritable when trying to cut down or stop | Withdrawal |
| Gambles to escape problems or relieve low mood, guilt or anxiety | Negative reinforcement (Section 3) |
| After losing money, often returns another day to win it back (“chasing” losses) | No direct parallel |
| Lies to family, a therapist or others to conceal the extent of gambling | No direct parallel |
| Has jeopardized or lost a significant relationship, job, or educational or career opportunity | Continued use despite harm to roles and relationships |
| Relies on others to provide money to relieve a desperate financial situation caused by gambling | No direct parallel |
9.2 How Common Are Gambling Problems?
The Canadian estimates used here come from the 2018 Canadian Community Health Survey of 24,983 people in the ten provinces, which used the Canadian Problem Gambling Index (CPGI), a nine-item measure of problem gambling behaviour and its consequences. Over half of Canadians (51.8 percent) had bought a lottery or raffle ticket in the past year, 12.6 percent had used electronic gambling machines and 7.9 percent had bet on sports. Of those who gambled, 1.6 percent, about 304,400 people, scored 3 or more on the CPGI, a moderate-to-severe risk of gambling problems. Men were more likely to be at risk, and so were people in lower-income households: 53.8 percent of the lowest-income group gambled, compared with 71.5 percent of the highest, but 2.7 percent of the lowest-income group were at moderate-to-severe risk, compared with 1.1 percent (Rotermann & Gilmour, 2022). These data predate the legalization of single-event sports betting in 2021.
Worldwide, the Lancet Public Health Commission on gambling estimated in 2024 that about 5.5 percent of women and 11.9 percent of men experience “any risk gambling”, meaning at least occasional experience of at least one behavioural symptom or adverse consequence of gambling. Extrapolated to the world’s adults, this would be about 448.7 million people, of whom about 80 million experience gambling disorder or problematic gambling. The Commission also estimated that gambling disorder or problematic gambling could affect 15.8 percent of adults and 26.4 percent of adolescents who gamble on online casino or slot products, and 8.9 percent of adults who bet on sports (Wardle et al., 2024). With gambling disorder and its scale defined, the next part asks how gambling acts on the brain.
9.3 Shared Reward Mechanisms
If gambling disorder belongs with the substance use disorders, gambling should engage the reward, learning and control systems that drugs engage. This part reviews three kinds of evidence: the response of the reward system to gambling, the effects of dopamine medications and the special power of near misses.
9.3.1 Uncertain Rewards and the Dopamine System
Schultz (2016) notes that less is known about gambling than about drugs, but that gambling, “with its strong sensory stimulation and prospect of large gains”, activates dopamine-rich areas in humans and dopamine neurons in animals. The prediction error model of Section 1 suggests why uncertainty matters: a fully predicted reward produces no prediction error, while outcomes that vary from play to play keep producing errors. Koob and Volkow (2016) report that changes similar to those in drug addiction, including reward deficits, stress sensitization, impaired inhibitory control and reduced striatal dopamine D2 receptors, have been observed in behavioural addictions, and they describe the imaging data as preliminary. This is despite the fact that “the house always wins”: the expected value of gambling is negative (Clark et al., 2009). The box below describes evidence from dopamine medications.
Dopamine medications and gambling in Parkinson’s disease
Parkinson’s disease is treated with medications that replace dopamine or activate dopamine receptors directly (dopamine agonists). Weintraub and colleagues (2010) studied 3,090 patients treated at 46 movement disorder centres in the United States and Canada. An impulse control disorder was identified in 13.6 percent, including problem or pathological gambling in 5.0 percent. Impulse control disorders were more common in patients taking a dopamine agonist than in those not taking one (17.1 percent compared with 6.9 percent; adjusted odds ratio 2.72). Because the study was cross-sectional, it shows an association; other factors, including younger age, smoking and a family history of gambling problems, were also associated with these disorders. Grall-Bronnec and colleagues (2018) reviewed 90 studies and concluded that dopamine agonist use is associated with a higher risk of impulse control disorders, under the combined influence of several other factors.
9.3.2 Near Misses
Slot machines and many online games display losses that look almost like wins, such as two matching symbols with the third just above the payline. Case 8.5 describes a 2009 study that examined near misses in the brain, and Figure 8.6 shows its three kinds of outcome.
Luke Clark and colleagues at the University of Cambridge built a two-reel slot machine task. Wins (worth 50 pence) occurred on one-sixth of trials, near misses, in which the reel stopped one position from the payline, on two-sixths, and full misses on three-sixths, frequencies they describe as comparable to real machines. In 40 volunteers who did not gamble regularly, near misses were rated less pleasant than full misses but increased the desire to continue playing, an effect seen only on trials where the player had chosen the symbol. In a second group of 15 volunteers scanned with fMRI, near misses activated the ventral striatum and anterior insula, regions that also responded to real wins, and insula activity to near misses correlated with a questionnaire measure of gambling-related thinking (Clark et al., 2009).
The near-miss finding connects the sections of this lesson. A near miss is a loss, so it brings no money, yet it recruits the circuitry that responds to wins. It also resembles the wanting and liking dissociation in Section 2: pleasure fell while the urge to continue rose. Clark and colleagues (2009) concluded that near misses “invigorate gambling through the anomalous recruitment of reward circuitry”. Later reviews give a mixed picture: a systematic review by Barton and colleagues (2017) found that near misses motivate continued play and are associated with activity in reward-processing areas, while a review by Pisklak and colleagues (2020) found inconsistent evidence that near misses make people gamble for longer. Gambling can therefore engage the reward system; the next part turns to the products through which most Canadians now meet it.
9.4 Online and Sports Betting in Canada
Gambling in Canada changed rapidly after 2021. The Safe and Regulated Sports Betting Act legalized single-event sports betting in Canada in 2021, and Statistics Canada links it to the launch of Ontario’s regulated online gambling (igaming) market in April 2022 (Rotermann & Gilmour, 2022). This part uses Ontario’s published market data, and the next introduces the arithmetic of the house edge.
Ontario’s open, regulated igaming market launched on 4 April 2022 (Alcohol and Gaming Commission of Ontario [AGCO], 2023). iGaming Ontario publishes monthly data on cash wagers, the total of real-money bets placed (bonus bets excluded); on non-adjusted gross gaming revenue (NAGGR), which is cash wagers minus player winnings; and on active player accounts, which are not unique players, because one person can hold accounts with several operators (iGaming Ontario, 2026). The figures cover the operators that have agreements with iGaming Ontario and exclude the Ontario Lottery and Gaming Corporation’s own online site and betting on horse races. In August 2026, Ontarians placed $9,740 million in cash wagers, operators kept $396.3 million in NAGGR and there were 1,208,000 active player accounts; casino games, which include slots and table games, made up 90 percent of wagers. Summing the monthly figures for April 2025 to March 2026 gives about $103.3 billion in cash wagers and $4.27 billion in NAGGR. Interactive 8.4 lets you explore the monthly series.
Explorer: Ontario’s regulated online gambling market, April 2022 to August 2026
The data show steep growth: cash wagers rose from $1,079 million in the partial first month to more than $9 billion a month by late 2025. The hold, the share of wagers that operators keep, stayed near 4 percent from April 2022 to August 2026 (4.07 percent in August 2026), which means that on average about 96 cents of every dollar bet in the market was paid back to players as winnings. In 2023, after the first year of the market, the AGCO announced that from 28 February 2024 operators could no longer use active or retired athletes in igaming advertising except to promote responsible gambling, and it restricted celebrities and influencers likely to appeal to minors, because it judged that such advertising could harm people under the legal gambling age (AGCO, 2023).
9.5 The House Edge and Expected Loss
Every game in a regulated market is designed so that, over many bets, the operator keeps a share of the money staked. This share is the house edge; its complement, the share paid back as winnings, is called the return to player. Equation 8.4 gives the expected loss over a period.
The expected loss is an average over many bets; any one person in any one month may win or lose much more. Worked Example 8.3 (hypothetical numbers). A person with a household income before tax of $60,000 a year ($5,000 a month) places bets totalling $1,500 in a month on games with an average house edge of 6 percent. The expected loss is L = $1,500 × 0.06 = $90, which is 1.8 percent of monthly household income, while the total amount bet is 30 percent of monthly income. At the Ontario market’s August 2026 hold of 4.07 percent, the same betting would have an expected loss of about $61. The box on Canada’s Lower-Risk Gambling Guidelines below explains which of these two percentages Canada’s gambling guidelines refer to, and Calculator 8.2 lets you vary each input.
Canada’s Lower-Risk Gambling Guidelines
On 15 September 2021, the Canadian Centre on Substance Use and Addiction (CCSA) released the Lower-Risk Gambling Guidelines, developed by a scientific working group co-chaired by Matthew Young and David Hodgins from data on more than 60,000 people who gamble in eight countries. The guidelines set three limits that are meant to be followed together: gamble no more than 1 percent of household income before tax per month, gamble no more than 4 days per month and avoid regularly gambling at more than 2 types of games (CCSA, 2021). The report that set the limits measured gambling spending from people’s reports of the money they lost on gambling in a month, net of winnings, and of the share of income they spent on it (Young et al., 2021), so the 1 percent limit is best compared with money lost. As of November 2026, the guidelines website words the limit as “Don’t bet more than 1% of your household income before tax per month”, which a reader could take to mean the total amount bet. Calculator 8.2 therefore compares the expected loss with the limit and shows the total amount bet alongside it for comparison.
A regulated online market can therefore grow quickly, as in Ontario, and the house edge turns large volumes of betting into predictable losses. The next part asks how far the idea of behavioural addiction should extend.
9.6 Gaming Disorder and the Limits of Behavioural Addiction
If gambling can become an addiction, other rewarding behaviours, such as video gaming, shopping or exercise, might too. This part describes gaming disorder and the arguments of researchers who warn against turning everyday behaviour into disorders.
9.6.1 Gaming Disorder
The DSM-5 work group reviewed more than 200 publications on internet gaming addiction, mostly case reports or case series of young men in Asia, and found no standard criteria and limited data on prevalence, course or brain function. It placed internet gaming disorder in Section 3 of DSM-5, among conditions needing further study (Hasin et al., 2013). The World Health Organization went further. Its eleventh revision of the International Classification of Diseases (ICD-11), released in 2018, added gaming disorder to the section on addictive disorders (WHO Regional Office for Africa, 2018), and ICD-11, adopted by the World Health Assembly in 2019, came into effect on 1 January 2022 (World Health Organization, 2019). ICD-11 defines gaming disorder by impaired control over gaming, increasing priority given to gaming over other activities, and continuation or escalation despite negative consequences, severe enough to cause significant impairment in personal, family, social, educational or occupational functioning and normally evident for at least 12 months. In 2020, WHO noted that studies suggest the disorder “affects only a small proportion of people” who play digital or video games (World Health Organization, 2020).
9.6.2 The Risk of Overpathologizing
Some researchers argue that the concept of behavioural addiction has been stretched too far. Billieux and colleagues (2015) observed that recent publications suggested “nearly all daily life activities might lead to a genuine addiction”, and they criticized research that starts from an anecdote and applies checklists of substance addiction symptoms to activities such as work, exercise and even Argentine tango. Kardefelt-Winther and colleagues (2017) note that symptoms such as preoccupation appear with most engaging activities and do not by themselves show functional impairment. They proposed an operational definition of behavioural addiction as “a repeated behavior leading to significant harm or distress” that “is not reduced by the person and persists over a significant period of time”, with harm that is “of a functionally impairing nature”. The flip cards below give their four exclusion criteria.
These criteria do not deny that gambling disorder or gaming disorder can be serious. They ask researchers and clinicians to show harm and persistence before applying an addiction label, and to consider other explanations first. Research on behavioural addictions, including how best to define them, is still evolving. Try it 8.4 asks you to apply them.
Using the definition and exclusion criteria of Kardefelt-Winther and colleagues (2017), decide whether each hypothetical pattern fits a behavioural addiction, and name the criterion that guides your answer. (a) A student plays an online game for six hours a day during a three-week break and then returns to a normal routine with no problems. (b) A competitive dancer trains 25 hours a week and has had several injuries. (c) A man who lost his job and is being treated for depression has started gaming most of the day. (d) A woman has gambled online most nights for 18 months, has borrowed money to cover losses, has failed to cut down and has separated from her partner over the debts.
Suggested answers: (a) No: intense involvement without impairment (criterion 3). (b) No: a willful choice in high-level activity (criterion 2). (c) Probably not as a separate addiction: the gaming may be better explained by depression or be a coping strategy (criteria 1 and 4), although it deserves clinical attention. (d) Yes: repeated behaviour, significant harm and persistence despite attempts to reduce it; she also meets several DSM-5 criteria for gambling disorder in Table 8.4.
Ontario opened a regulated online gambling market on 4 April 2022. In August 2026, iGaming Ontario reported $9,740 million in cash wagers, $396.3 million in NAGGR and 1,208,000 active player accounts, with casino games making up 90 percent of wagers (iGaming Ontario, 2026). From 28 February 2024, the AGCO banned the use of athletes in igaming advertising except to promote responsible gambling (AGCO, 2023). Use the whole lesson to complete four tasks. (1) Calculate the market’s hold for August 2026 with Equation 8.4 rearranged (h = NAGGR ÷ wagers), explain what it means for the average result of $100 bet, and explain why 1,208,000 accounts is not the number of people who gambled. (2) Using prediction error (Section 1) and incentive salience (Section 2), explain why fast online slot play with unpredictable wins and frequent near misses might sustain play even when losses mount. (3) Explain the AGCO’s athlete ban in terms of cue-triggered wanting, and name the evidence you would need to judge whether it reduced harm among young people. (4) Using Statistics Canada’s 2018 estimate (made before single-event sports betting and Ontario’s online market) that 1.6 percent of people who gambled were at moderate-to-severe risk, and Kardefelt-Winther’s criteria, explain why heavy use of the market alone does not show that most players have a disorder.
Summary
This section described evidence that gambling disorder shares criteria, some brain responses and treatment approaches with the substance use disorders. In 2018, before single-event sports betting and Ontario’s online market, 1.6 percent of Canadians who gambled were at moderate-to-severe risk, more often in lower-income households. Gambling may engage the reward system through uncertain outcomes and near misses, although a 2020 review found inconsistent evidence that near misses lengthen play. Ontario’s online market grew rapidly after 2022 with a hold near 4 percent, and criteria for behavioural addiction guard against labelling ordinary passions as disorders. The reflection and final assessment that follow bring all four sections together.
Key Takeaways
- 9.1 DSM-5 moved gambling disorder into the renamed chapter “Substance-Related and Addictive Disorders” because it often co-occurs with substance use disorders and resembles them in symptoms, biological dysfunction, genetic liability and treatment. It renamed the condition, removed the criterion of illegal acts and requires four or more of nine criteria within 12 months, criteria that parallel craving, tolerance, withdrawal, negative reinforcement and continued use despite harm.
- 9.2 In 2018, 64.5 percent of people in Canada aged 15 or older gambled in the past year, and 1.6 percent of those who gambled, about 304,400 people, were at moderate-to-severe risk on the CPGI, more often men and people in lower-income households; these data predate single-event sports betting. The Lancet Public Health Commission estimated in 2024 any risk gambling in 11.9 percent of men and 5.5 percent of women worldwide, with higher rates of disorder or problematic gambling among people who gamble on online casino or slot products.
- 9.3 Gambling can activate dopamine-rich areas, and on the prediction error model, outcomes that vary from play to play keep producing prediction errors even though the expected value of gambling is negative. Dopamine agonists in Parkinson’s disease are associated with impulse control disorders including gambling, and in a 2009 study near misses were rated less pleasant than full misses yet raised the desire to play and activated the ventral striatum and insula, a pattern that resembles the dissociation of wanting from liking.
- 9.4 Ontario’s regulated igaming market launched on 4 April 2022, and cash wagers rose from $1,079 million in the partial first month to more than $9 billion a month by late 2025, with $9,740 million in August 2026 and a hold near 4 percent. Active player accounts are not unique players, and from 28 February 2024 the AGCO banned athletes in igaming advertising except to promote responsible gambling.
- 9.5 Expected loss equals the total amount bet multiplied by the house edge (L = B × h), so $1,500 bet at a 6 percent edge gives an expected loss of $90, or 1.8 percent of a $5,000 monthly income. Canada’s Lower-Risk Gambling Guidelines set limits of 1 percent of household income before tax per month, 4 days per month and 2 types of games, and the income limit is best compared with money lost rather than with the total amount bet.
- 9.6 ICD-11 defines gaming disorder by impaired control, increasing priority given to gaming and continuation despite negative consequences, with significant impairment normally evident for at least 12 months, while DSM-5 lists internet gaming disorder as a condition for further study. Kardefelt-Winther and colleagues require repeated behaviour with significant, functionally impairing harm that persists, and they exclude behaviour better explained by another disorder, a willful choice, intense involvement without impairment or a coping strategy.
Reflection
A commentator says: “Online slots and video games both release dopamine, so they are addictions just like heroin, and anyone who plays every day has an addiction.” In 150 to 200 words, respond using: the DSM-5 criteria for gambling disorder (four or more of nine criteria in 12 months); the ICD-11 definition of gaming disorder (impaired control, increasing priority, continuation despite negative consequences, significant impairment, normally evident for at least 12 months); the near-miss study by Clark and colleagues (2009), in which near misses were rated less pleasant than full misses but increased the desire to play and activated reward regions that respond to wins; Statistics Canada’s 2018 estimate that 1.6 percent of Canadians who gambled were at moderate-to-severe risk of gambling problems; and the exclusion criteria of Kardefelt-Winther and colleagues (2017). Say what the commentator gets right and what the claim gets wrong.
Minimum 20 characters required.
Question 1: Which change did DSM-5 make to the diagnosis of gambling disorder?
Question 2: In the near-miss study by Clark and colleagues (2009), how did near misses compare with full misses?
Question 3: Using Equation 8.4, what is the expected monthly loss for someone who bets a total of $2,000 in a month on games with a house edge of 5 percent?
Question 4: Which pattern falls under one of the exclusion criteria for behavioural addiction proposed by Kardefelt-Winther and colleagues (2017)?
Final Assessment
Bringing It All Together
This lesson followed addiction from a single dopamine neuron to the design of an online casino. Dopamine neurons in the ventral tegmental area project to the nucleus accumbens and prefrontal cortex, and drugs that people use non-medically raise dopamine there by acting on receptors, ion channels or transporters. In the most influential account, dopamine neurons signal reward prediction errors that teach the brain what to repeat; a natural reward stops teaching once it is predicted, and Redish’s model proposes that drugs keep teaching because their dopamine signal is not cancelled by prediction, a proposal that some evidence supports and a blocking experiment does not.
In animal studies, dopamine is more closely tied to wanting than to liking, while human findings are mixed. Incentive sensitization theory holds that repeated drug use can make the wanting system hyper-reactive to drug cues while pleasure stays level or fades, and studies of cue-induced craving and its incubation, mostly in animals, suggest that urges can persist and even grow long after withdrawal ends. Koob and Volkow’s three-stage cycle adds a stress system that produces negative emotion when the drug wears off, a shift from positive to negative reinforcement and weakened prefrontal control. Tolerance and physical dependence are normal adaptations that can occur with prescribed use and are distinct from addiction.
Similar mechanisms may operate without a drug. Gambling disorder shares criteria and some reward responses with the substance use disorders; in a 2009 study, near misses increased the urge to play while reducing pleasure; and Ontario’s online market shows how quickly new products can spread. Careful criteria for behavioural addiction, which require persistent and impairing harm, help keep these insights from turning everyday enthusiasms into disorders.
Key Takeaways from this lesson
- The mesocorticolimbic dopamine system runs from the ventral tegmental area to the nucleus accumbens and prefrontal cortex, and drugs used non-medically raise dopamine there through G protein-coupled receptors, ion channels or monoamine transporters. A 2015 review of human imaging found strong evidence that stimulants raise striatal dopamine and little evidence that opioids and cannabis do, so dopamine is one part of the explanation of addiction.
- In the most influential account, dopamine neurons signal reward prediction errors: they fire for better-than-expected outcomes, stay at baseline for predicted ones and dip when a predicted reward is omitted, and the signal teaches. A natural reward stops teaching once its value reaches the reward, while a drug whose dopamine boost cannot be cancelled keeps adding value on every trial.
- Redish’s model proposes that drugs produce a teaching signal that prediction cannot cancel, so drug-paired cues gain value without limit; dopamine recordings with cocaine support part of the model and a blocking experiment does not.
- In rats, wanting, or incentive salience, depends on large dopamine-based systems, while liking depends on small opioid and cannabinoid hotspots, so the two can come apart; human findings are mixed. Incentive sensitization theory holds that repeated, intermittent drug use can make the wanting system hyper-reactive to drug cues in susceptible people while pleasure stays level or fades.
- Drug cues trigger craving and limbic activation, and cue-induced craving can grow during abstinence, most clearly in animal studies, as the incubation of craving in rats and in a study of people who smoke suggests. These findings support incentive sensitization theory, although much of the evidence comes from rats and from imaging that does not measure dopamine directly.
- Koob and Volkow’s three-stage cycle links binge and intoxication to the basal ganglia, withdrawal and negative affect to the extended amygdala, and preoccupation and anticipation to the prefrontal cortex and insula, and over repeated cycles drug use shifts from positive toward negative reinforcement.
- Tolerance and physical dependence can occur with prescribed use; addiction is defined by compulsion, loss of control and continued use despite harm, and the severity of withdrawal, which varies widely across drug classes, does not measure it.
- The iRISA model links addiction to excessive salience for drugs, reduced sensitivity to other rewards and weakened inhibitory control in the prefrontal cortex, but cross-sectional brain images cannot show whether such differences preceded drug use or followed it.
- Gambling disorder shares criteria and reward mechanisms with substance use disorders; in 2018, before single-event sports betting was legal, 1.6 percent of Canadians who gambled were at moderate-to-severe risk, and Ontario’s regulated online market has grown rapidly since April 2022. Criteria that require persistent, functionally impairing harm and exclude other explanations guard against labelling everyday activities, including gaming, as addictions.
Core Concepts Reviewed
Section 1: the mesocorticolimbic dopamine system, the ventral tegmental area and nucleus accumbens, microdialysis evidence, disinhibition and transporter actions, reward prediction error, the learning update and the drug prediction error model.
Section 2: liking and wanting, incentive salience, hedonic hotspots, incentive sensitization theory, sensitization and tolerance, sign-trackers, conditioned reinforcement, cue-induced craving and the incubation of craving.
Section 3: positive and negative reinforcement, impulsivity and compulsivity, opponent processes, the three-stage cycle, within-system and between-system neuroadaptations, tolerance, physical dependence, withdrawal syndromes and the iRISA model.
Section 4: gambling disorder and its criteria, Canadian and global prevalence, near misses, dopamine agonists in Parkinson’s disease, Ontario’s online market, the house edge and expected loss, the Lower-Risk Gambling Guidelines, gaming disorder and overpathologizing.
The final reflection asks you to bring the four sections together to explain one hypothetical person’s experience of craving after withdrawal.
Reflection
A hypothetical man used opioids from the unregulated supply daily for several years and has now been abstinent for two months. Withdrawal is long over, but he reports strong urges when he passes the neighbourhood where he used to buy drugs, especially on stressful days, and says he had stopped enjoying the drug long before he quit. In 200 to 250 words, explain his experience by integrating the four sections of this lesson: (1) how drug-induced dopamine and reward prediction error could have strengthened the value of drug-related cues; (2) why cue-triggered urges can persist or grow after withdrawal ends, and why stress can intensify them; (3) where his past and present experience fits in the three-stage cycle, including the shift from positive to negative reinforcement, the opioid withdrawal timing in Table 8.3 (begins within 12 hours, peaks at 36 to 72 hours, tapers over 4 to 7 days) and why his lost tolerance matters; and (4) which of these mechanisms gambling disorder shares and which it lacks.
Minimum 30 characters required.
Final Knowledge Assessment
Question 1: Which sequence best links Sections 1, 2 and 3 in explaining how drug cues come to drive behaviour?
Question 2: Why does Redish’s model predict that the value of a drug-paired cue keeps rising while the value of a natural reward levels off?
Question 3: A hypothetical person says cocaine is “no longer much fun”, yet his urges to use are stronger than ever. Which account explains this most directly?
Question 4: Which pair of findings shows that dopamine is more closely tied to wanting than to liking?
Question 5: How do tolerance and sensitization differ?
Question 6: In a 2014 study of people seeking treatment for prescription opioid dependence, cited by Kosten and Baxter (2019), 56.5 percent of those with chronic pain who began opioids for pain said that avoiding withdrawal was their main reason to continue. Which stage and form of reinforcement does this illustrate?
Question 7: Which statement about withdrawal and addiction is best supported by this lesson?
Question 8: Why is the period after opioid withdrawal a time of high overdose risk, linking Sections 2 and 3 with Lesson 5?
Question 9: What does the iRISA model add to the three-stage cycle?
Question 10: Goldstein and Volkow (2011) report prefrontal grey matter reductions that persist into long abstinence. Which conclusion is most appropriate?
Question 11: Which evidence best supports placing gambling disorder alongside the substance use disorders?
Question 12: In August 2026, Ontario’s regulated online market reported cash wagers of $9,740 million and NAGGR of $396.3 million. What was the hold, and what does it mean?
Question 13: A hypothetical person with a household income of $48,000 a year expects to lose $60 in a month of gambling. How does this compare with the money limit in the Lower-Risk Gambling Guidelines?
Question 14: How does ICD-11’s approach to gaming disorder compare with that of DSM-5?
Question 15: Which conclusion best integrates the lesson’s evidence on behavioural addiction?
Glossary: Key Terms, People & Frameworks
📚 Reference page, available throughout the lesson
Terms, frameworks and people introduced in this lesson on the neurobiology of addiction, with short definitions.