Why Some People Develop Addiction and Others Do Not
Drugs and Society
Learning objectives for this lesson:
- Explain how family, twin and adoption studies estimate the genetic contribution to substance use disorders, and calculate heritability with Falconer’s formula.
- Describe how variants of the ADH1B and ALDH2 genes protect against alcohol use disorder, and explain what polygenic scores can and cannot tell us about an individual’s risk.
- Describe how adolescent brain development and age at first use are associated with later substance use disorder, and calculate a relative risk and a risk difference.
- Distinguish a causal explanation of an association from a common liability explanation, using evidence from twin studies.
- Describe how childhood self-control and the personality dimensions measured by the Substance Use Risk Profile Scale are associated with later substance use, and explain three ways in which mental illness and substance use disorders may come to co-occur.
- Evaluate the evidence that childhood adversity and social rank shape vulnerability to addiction, including studies of social dominance in monkeys.
- Assess the Rat Park experiments at Simon Fraser University, the criticisms of their design and their replication record, and explain what animal models of the environment can and cannot tell us about human addiction.
- Explain the evidence for natural recovery and the clinician’s illusion, compare moral, disease, choice, learning and biopsychosocial accounts of addiction, and apply them to the brain disease model debate and to a Canadian law.
This course was developed by Dr. Kiffer G. Card.
Genetic Epidemiology: Families, Twins, Adoptees and Genes
Introduction
Lesson 8 described how drugs act on the brain’s reward and learning systems, and how repeated use produces tolerance, withdrawal and cue-triggered craving. Yet most people who drink alcohol, use cannabis or try other drugs never develop a substance use disorder. Volkow, Koob and McLellan (2016) put the point directly: “Only a minority of people who use drugs ultimately become addicted.” They attributed the difference to varying vulnerability to genetic, environmental and developmental factors, which this lesson examines in turn.
This first section starts from the observation that addiction runs in families. It explains how studies of families, twins and adoptees estimate genetic influence, works through the arithmetic of a twin study, examines two genes with well-understood effects on drinking, and closes with polygenic scores. Lesson 1 defined substance use disorder with the criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5). In this lesson, as in everyday usage, the word addiction refers to the same condition.
Learning Objectives
- 1.1 Explain how family studies show that substance use disorders aggregate in families, and why they cannot separate shared genes from shared homes.
- 1.2 Calculate heritability, shared environment and unique environment from twin correlations with Falconer’s formula, interpret the results, and state the assumptions on which the twin design rests.
- 1.3 Explain how adoption studies separate genetic transmission from the rearing environment, and compare the main assumption or limit of family, twin and adoption designs.
- 1.4 Describe the evidence that substance use disorders are moderately to highly heritable, including the meta-analysis by Verhulst, Neale and Kendler (2015).
- 2.1 Explain how variants of the ADH1B and ALDH2 genes protect against alcohol use disorder, and how changing drinking norms altered that protection in Japan.
- 2.2 Evaluate what polygenic scores can and cannot tell us about one person’s risk of addiction.
1.1 Why Addiction Runs in Families
A family study begins with a set of people who have a disorder, called probands, and compares their relatives with the relatives of people who do not have the disorder. If the disorder is more common among the relatives of probands, it is said to aggregate in families.
Kathleen Merikangas and colleagues (1998) carried out a controlled family study of this kind. They interviewed 231 probands with dependence on opioids, cocaine, cannabis or alcohol, 61 control probands, and 1,267 adult first-degree relatives (parents, siblings and children). Relatives of probands with drug disorders had an eight-fold increased risk of drug disorders themselves, across opioids, cocaine, cannabis and alcohol. The authors concluded that a family history of drug disorder is one of the most potent risk factors for developing one, and they found that relatives tended to share the proband’s predominant drug (Merikangas et al., 1998).
A family study cannot say why a disorder runs in families. Parents and children share genes, but they also share a home, a neighbourhood, a household income, attitudes toward drinking and access to substances. Any of these could explain the clustering. Separating the two sources of resemblance requires families in which genetic and environmental relatedness come apart, which is what twin and adoption studies provide.
1.2 The Logic of Twin Studies
Twins offer a natural comparison. Monozygotic (identical) twins develop from a single fertilized egg and share essentially all of their DNA. Dizygotic (fraternal) twins develop from two eggs and, like any pair of full siblings, share on average about half of the genetic variants that differ between people (Mayhew & Meyre, 2017). Both kinds of twins are usually raised together, so if identical twins resemble each other more closely than fraternal twins, the extra resemblance is attributed to their greater genetic similarity.
1.2.1 Three Sources of Variation
Twin researchers divide the variation in a trait into three parts. The additive genetic part (written A, and estimated as heritability, h²) is the share of variation linked to genetic differences. The shared environment (C, or c²) covers influences that make members of a family alike, such as household income or parents’ drinking. The unique environment (E, or e²) covers experiences that make family members different, such as different friends, different jobs or a serious injury, together with measurement error (Mayhew & Meyre, 2017). The three parts sum to 1, or 100 percent of the variation. Figure 9.1 shows how the two kinds of twin pairs share these sources.
1.2.2 Falconer’s Formula
The geneticist D. S. Falconer turned this reasoning into a simple formula, published in his 1960 textbook on quantitative genetics and still used to teach twin studies (Mayhew & Meyre, 2017). Researchers first measure how similar twins are on the trait, expressed as a correlation (r), a number from 0 (no resemblance) to 1 (perfect resemblance). Under the twin model, the correlation in identical pairs equals h² + c², and the correlation in fraternal pairs equals half of h² plus c². Rearranging these two statements gives Equation 9.1.
Worked Example 9.1. Suppose a hypothetical twin study measures liability to alcohol use disorder and finds a correlation of 0.62 in identical pairs and 0.38 in fraternal pairs. Heritability is h² = 2 × (0.62 − 0.38) = 2 × 0.24 = 0.48. The shared environment is c² = (2 × 0.38) − 0.62 = 0.76 − 0.62 = 0.14. The unique environment is e² = 1 − 0.62 = 0.38. The three parts sum to 0.48 + 0.14 + 0.38 = 1.00. About 48 percent of the variation in this hypothetical population is linked to genetic differences, 14 percent to environments that make family members alike and 38 percent to experiences that make them different. Use Calculator 9.1 to load the correlations implied by Verhulst and colleagues’ (2015) estimates, then try a fraternal correlation more than half, and much less than half, of the identical one.
The calculator shows two warning signs. When the fraternal correlation is more than half of the identical correlation, the shared environment is positive, and it grows as the fraternal correlation approaches the identical one. When it is less than half, Falconer’s formula produces a negative shared environment, which is impossible; this pattern suggests that genes act in non-additive ways (for example, through combinations of variants) or that an assumption of the model has failed.
1.2.3 The Assumptions Behind the Numbers
A much-debated assumption is the equal environments assumption: that identical and fraternal twins experience equally similar environments. If parents, teachers and friends treat identical twins more alike than fraternal twins, some environmental resemblance will be counted as genetic, and heritability will be overestimated. The model also assumes that the unique environment affects both kinds of twins equally and that siblings share exactly half of their varying genes, which is true only on average (Mayhew & Meyre, 2017). That figure of one half also assumes that parents pair up at random with respect to the trait. When partners resemble each other in liability to alcohol use disorder, a pattern called assortative mating, fraternal twins share more than half of the relevant genes, which inflates the estimate of the shared environment and correspondingly lowers the estimate of heritability (Verhulst et al., 2015). The box below sets out what a heritability estimate can and cannot tell us, since the word is often misread.
What a heritability estimate means
It describes a population. Heritability is the proportion of variation in a trait across a group of people that is linked to genetic differences among them. Every component of the twin model is specific to the population studied, so a change in environment changes the estimate (Mayhew & Meyre, 2017).
It does not divide an individual. A heritability of 0.50 does not mean that half of one person’s alcohol use disorder comes from genes and half from experience. For any individual, genes and experience act together throughout life.
It does not imply that a condition is fixed. Goldman, Oroszi and Ducci (2005) called the heritability of addiction paradoxical, because addiction requires drug use, a choice itself shaped by genes and environment. Case 9.1 shows a heritable risk changing with the environment.
Twin studies therefore estimate genetic influence under stated assumptions. Adoption studies test the same question with a different set of assumptions, which is why researchers place most confidence in conclusions that both designs support.
1.3 Adoption Studies
Adoption separates the two routes by which parents pass on risk. A child adopted early in life receives genes from biological parents and an upbringing from adoptive parents. If adoptees resemble their biological relatives, genes are implicated; if they resemble their adoptive relatives, the rearing environment is implicated.
One large adoption study of alcohol use disorder used eight Swedish national registers. Kendler and colleagues (2015) followed 18,115 adoptees born between 1950 and 1993 and identified alcohol use disorder from medical, legal and pharmacy records. Alcohol use disorder in adoptees was predicted by alcohol use disorder in their biological parents (odds ratio 1.46) and biological siblings (odds ratio 1.94), and also in their adoptive parents (odds ratio 1.40). An odds ratio compares the odds of an outcome in two groups; 1.46 means odds about one and a half times as high. The authors concluded that parent-to-child transmission of alcohol use disorder results from both genetic and environmental factors (Kendler et al., 2015).
Adoption studies have their own limits. Children adopted later in childhood have already shared an environment with their biological family. Biological parents, adoptive parents and adoptees may also differ from the general population, which limits how far the results generalize (Mayhew & Meyre, 2017). Register studies also capture only cases that came to the attention of hospitals, courts or pharmacies. Table 9.1 compares the three designs.
Table 9.1. Family, twin and adoption studies compare different kinds of relatives, separate genetic and environmental influences to different degrees and rest on different assumptions.
| Design | Who is compared | What it can separate | Main assumption or limit | Example |
|---|---|---|---|---|
| Family study | Relatives of people with and without the disorder | Whether the disorder clusters in families | Cannot separate shared genes from shared homes | Merikangas et al. (1998): eight-fold increased risk of drug disorders in relatives |
| Twin study | Identical and fraternal twin pairs raised together | Additive genetic, shared environmental and unique environmental variation | Equal environments assumption | Verhulst et al. (2015): meta-analysis of 12 twin studies |
| Adoption study | Adoptees with their biological and adoptive relatives | Genetic transmission and rearing environment | Late placement and unrepresentative families | Kendler et al. (2015): 18,115 Swedish adoptees |
1.4 How Heritable Are Substance Use Disorders?
With many studies of each design now published, meta-analysis can combine them. Verhulst, Neale and Kendler (2015) identified 12 twin studies and 5 adoption studies of alcohol use disorder and reanalyzed the data together. The best estimate of heritability was 0.49 (95 percent confidence interval 0.43 to 0.53), and the shared environment accounted for 0.10 (0.03 to 0.16) of the variation. Estimates did not differ by sex, design or assessment method, and the authors concluded that alcohol use disorder is approximately 50 percent heritable, with a modest contribution from family environments.
Other substances show similar or higher figures. A review by Deak and Johnson (2021) describes the heritability of substance use disorders as about 50 percent, and, reviewing large twin studies, Ducci and Goldman (2012) reported heritabilities for addictions ranging from 0.39 for hallucinogens to 0.72 for cocaine, and noted that heritability is usually higher for addiction than for substance use itself. Genes therefore appear to matter more for whether use becomes a disorder than for whether a person tries a substance.
Is there one genetic liability to addiction, or a separate one for each drug? Kendler, Myers and Prescott (2007) studied 4,865 twins from the Virginia Adult Twin Study and found that two genetic factors fit the data well: one loading mainly on cannabis and cocaine (illegal drugs) and one loading on alcohol, caffeine and nicotine (legal drugs). The two factors were highly correlated (+0.82), and nicotine and caffeine also had large substance-specific genetic effects. Together with the drug specificity found by Merikangas and colleagues (1998), these results suggest a broad shared liability to substance use disorders with some additional risk specific to each class.
Heritability estimates tell us that genes matter without naming any gene. The next part of the section turns to two genes whose effects on drinking are understood in detail.
2.1 Genes That Protect: ADH1B and ALDH2
Lesson 4 traced how the liver breaks down alcohol in two steps. The enzyme alcohol dehydrogenase (ADH) converts ethanol into acetaldehyde, a toxic compound, and aldehyde dehydrogenase (ALDH) converts acetaldehyde into acetate. Some people carry gene variants that change the speed of these steps, and a review by Edenberg and McClintick (2018) describes functional variants in two of these genes, ADH1B and ALDH2, as by far the strongest genetic contributors to alcohol use disorder.
Edenberg (2007) summarized their effects. Certain variants of the ADH1B and ADH1C genes encode particularly active enzymes, so ethanol is converted to acetaldehyde more rapidly. A variant of the ALDH2 gene, called ALDH2*2, encodes an essentially inactive enzyme, so acetaldehyde accumulates. Both variants are protective against alcohol use disorder. Figure 9.2 shows where each variant acts.
The ALDH2*2 variant is common in East Asia. Brooks and colleagues (2009) estimated that about 36 percent of East Asians (Japanese, Chinese and Korean people) show the alcohol flushing response, which includes facial flushing, nausea and a rapid heartbeat after drinking, and that at least 540 million people worldwide, about 8 percent of the world’s population, have deficient ALDH2 activity. People who inherit two copies of the variant have no detectable enzyme activity and are unable to drink significant amounts of alcohol. People with one copy have more than a hundred-fold reduction in activity and a milder response (Brooks et al., 2009). A meta-analysis of studies in Asian populations found that carrying one copy of either ALDH2*2 or ADH1B*2 was protective against alcohol dependence, that the two protective effects were independent, and that a second copy did not add significant protection (Luczak, Glatt & Wall, 2006).
A protective gene works by making alcohol unpleasant, so its effect depends on how much people drink and why. Higuchi and colleagues (1994) genotyped Japanese people diagnosed with alcoholism, the clinical term then in use for alcohol dependence. As Edenberg (2007) summarizes their study, the fraction of these patients carrying the ALDH2*2 variant rose from 2.5 percent in 1979 to 13 percent in 1992. Edenberg suggests that the likely cause of this decline in protection was a sociological change toward more drinking in Japan.
The same variant carries a second lesson. Brooks and colleagues (2009) describe how some people with one copy develop tolerance to the flushing response and, partly through social influences, become habitual heavy drinkers. Because acetaldehyde is carcinogenic, these drinkers face a much higher risk of cancer of the esophagus: case-control studies in Japan and Taiwan report odds ratios of 3.7 to 18.1 after adjustment for alcohol consumption. A variant that protects most carriers from alcohol use disorder becomes a serious risk for the carriers who drink heavily.
Case 9.1 is a clear example of gene-environment interplay: the effect of a gene on behaviour depends on the setting in which the gene is carried. Heritability estimates made in Japan in 1979 and in 1992 would not have been the same, even though the gene itself did not change.
2.2 Polygenic Scores and Their Limits
Variants with effects as large as ALDH2*2 are rare. Deak and Johnson (2021) describe substance use disorders as highly polygenic: most genetic influence comes from many variants across the genome, the vast majority with a very small effect. Genome-wide association studies (GWAS) scan the genome in very large samples for variants that are slightly more common in people with a disorder. A polygenic score adds up the small effects of many such variants to give each person a single number for genetic liability.
How well do these scores predict addiction? Barr and colleagues (2020) tested polygenic scores for alcohol use disorder in a population-based Finnish twin study and in a high-risk family study in the United States. Taken together, the scores explained about 2.5 to 3.5 percent of the variance in alcohol use disorder. Adding them to age, sex and ancestry only slightly improved classification, and the authors judged their usefulness for identifying people at increased risk to be “modest, at best.” Scores built from larger genome-wide studies explain more: Dick and colleagues (2026) report that recent polygenic scores for psychiatric and substance use disorders account for between 5 and 10 percent of the variance in outcomes. The gap between twin-based heritability of about 50 percent and the much smaller share explained by measured variants is called missing heritability (Mayhew & Meyre, 2017). The tabs below set out what a score can add and where its limits lie.
Polygenic scores confirm that liability to addiction is spread across very many genes. In the study by Barr and colleagues (2020), people in the top 5 percent of the score distribution reported more criteria for alcohol and nicotine use disorders and also for cannabis and opioid use disorders, which fits the twin evidence of a broad shared liability. Ducci and Goldman (2012) noted that genetic findings support the case for a biological contribution to addictive behaviour. Dick and colleagues (2026) also found that, judged by the criteria a United States genomics research network used to select scores for clinical study, current scores for substance use disorders are comparable to scores already being introduced into care for conditions such as breast cancer and heart disease.
A score that explains 3 percent of the variance leaves 97 percent unexplained, and even a score that explains 10 percent leaves most of the variation unexplained, so many people with high scores never develop a disorder and many with low scores do. A score can place a person above or below average in genetic liability, but it cannot say whether that person will develop a disorder. In 2012, Ducci and Goldman (2012) concluded that known genetic predictors accounted for a small portion of the genetic variance and had not been integrated into clinical care. As of 2026, reviews describe the clinical usefulness of polygenic scores for substance use disorders, meaning whether using them improves people’s health, as still being studied (Kember et al., 2024; Dick et al., 2026), and research in this area is moving quickly.
Most genome-wide studies have been carried out in people of European ancestry. Martin and colleagues (2019) showed that the polygenic scores available in 2019 were several times more accurate in people of European ancestry than in people of other ancestries, and warned that using them clinically would give greater benefit to populations already well served, widening health disparities. Dick and colleagues (2026) still list this underrepresentation, and the weaker prediction that follows from it, among the issues to address so that clinical use does not widen health inequities.
A news story reports a twin study under the headline “Addiction is 50 percent genetic: half of your risk is written in your DNA.” Using the box on what a heritability estimate means (1.2.3 The Assumptions Behind the Numbers), Case 9.1 and the polygenic score evidence, write three sentences: one stating what the 50 percent figure describes, one explaining why it does not describe any single reader, and one explaining what a genetic test can and cannot tell a reader about their own risk. Then open the item below to compare.
The 50 percent figure estimates heritability, the share of variation in alcohol use disorder across a particular population that is linked to genetic differences (Verhulst et al., 2015). It does not split one person’s risk into halves, because genes and experience act together in each life, and the estimate changes when drinking norms change, as the Japanese ALDH2 data show. A polygenic score can show whether a reader’s genetic liability is above or below average, and a 2026 review found current scores for substance use disorders comparable to scores being introduced into care for other conditions (Dick et al., 2026). It still cannot tell a reader whether they will develop a disorder: such scores explain a small share of the variance (about 2.5 to 3.5 percent in Barr et al., 2020), work less well outside European ancestry (Martin et al., 2019), and, as of 2026, whether using them improves care is still being studied.
Summary
Substance use disorders cluster strongly in families. A meta-analysis of twin and adoption studies, two designs with different assumptions, estimated that roughly half of the variation in alcohol use disorder is linked to genetic differences (Verhulst et al., 2015), and a review put heritability for other addictions at about 0.39 to 0.72 (Ducci & Goldman, 2012). Much of the genetic liability is shared across substances. Variants of ADH1B and ALDH2, described as by far the strongest genetic contributors to alcohol use disorder (Edenberg & McClintick, 2018), protect by making alcohol unpleasant, and their effect weakened in Japan as drinking became more common. Polygenic scores confirm that liability is spread over many genes. They explain a small share of the variation and cannot say whether a particular person will develop a disorder; current scores compare well with scores introduced into care for other conditions, but as of 2026 whether they improve care is still being studied (Dick et al., 2026). Genes shape vulnerability through the way a person responds to substances and settings over a lifetime. Section 2 follows that lifetime through childhood and adolescence.
Key Takeaways
- 1.1 A family study compares the relatives of people with a disorder (probands) with the relatives of people without it. Merikangas and colleagues (1998) found an eight-fold increased risk of drug disorders among relatives of probands with drug disorders, and relatives tended to share the proband’s predominant drug. Because relatives share both genes and homes, a family study shows that a disorder aggregates in families but cannot say why.
- 1.2 Identical twins share essentially all of their DNA and fraternal twins about half of their varying genes, while both usually share a family environment. Falconer’s formula estimates heritability as h² = 2 × (rMZ − rDZ), the shared environment as 2rDZ − rMZ and the unique environment as 1 − rMZ, so correlations of 0.62 and 0.38 give 0.48, 0.14 and 0.38. These estimates rest on the equal environments assumption and on random mating, and heritability describes variation across a population; it does not divide one person’s risk into genetic and environmental parts.
- 1.3 Adoption separates the genes received from biological parents from the upbringing provided by adoptive parents. In 18,115 Swedish adoptees, alcohol use disorder was predicted by the disorder in biological parents (odds ratio 1.46) and siblings (1.94) and also in adoptive parents (1.40), so transmission is both genetic and environmental (Kendler et al., 2015). Late placement, unrepresentative families and register-based case finding limit adoption studies, just as the equal environments assumption limits twin studies and shared homes limit family studies.
- 1.4 A meta-analysis of 12 twin and 5 adoption studies estimated the heritability of alcohol use disorder at 0.49 and the shared environment at 0.10 (Verhulst et al., 2015). In a 2012 review, heritabilities for other addictions ranged from about 0.39 for hallucinogens to 0.72 for cocaine, and heritability was usually higher for addiction than for use itself. Twin data point to a broad shared liability across substances with some additional risk specific to each class (Kendler et al., 2007).
- 2.1 The ADH1B*2 variant speeds the conversion of ethanol to acetaldehyde, and ALDH2*2 nearly stops its conversion to acetate, so both raise acetaldehyde and cause flushing, nausea and a rapid heartbeat that make drinking unpleasant. Among Japanese people diagnosed with alcoholism, the share carrying ALDH2*2 rose from 2.5 percent in 1979 to 13 percent in 1992 as drinking increased, an example of gene-environment interplay. Carriers who drink heavily face a much higher risk of cancer of the esophagus.
- 2.2 Polygenic scores add up the small effects of many common variants and confirm that liability is spread across very many genes. In a 2020 study, scores explained only about 2.5 to 3.5 percent of the variance in alcohol use disorder and added little to age and sex (Barr et al., 2020); scores reviewed in 2026 account for 5 to 10 percent (Dick et al., 2026), still leaving a large gap called missing heritability. The scores available in 2019 were also several times less accurate outside European ancestry, so clinical use could widen health disparities (Martin et al., 2019).
Reflection
A hypothetical friend tells you: “My father has alcohol use disorder. I read that it is 50 percent genetic, so I have a 50 percent chance of developing it, and there is nothing I can do about it.” Using the following evidence from Section 1, write a reply of 150 to 200 words that explains what your friend has right and what your friend has wrong: (1) Verhulst, Neale and Kendler (2015) estimated the heritability of alcohol use disorder at 0.49 and the shared environment at 0.10; (2) in a Swedish adoption study, alcohol use disorder in adoptees was predicted by the disorder in biological parents (odds ratio 1.46) and in adoptive parents (odds ratio 1.40); (3) among Japanese people diagnosed with alcoholism, the share carrying the protective ALDH2*2 variant rose from 2.5 percent in 1979 to 13 percent in 1992 as drinking became more common; (4) in a 2020 study, polygenic scores explained about 2.5 to 3.5 percent of the variance in alcohol use disorder, and a 2026 review describes their usefulness in care as still being studied.
Minimum 20 characters required.
Question 1: A hypothetical twin study finds a correlation of 0.70 for liability to alcohol use disorder in identical twin pairs and 0.40 in fraternal twin pairs. What heritability does Falconer’s formula give?
Question 2: Verhulst, Neale and Kendler (2015) estimated the heritability of alcohol use disorder at 0.49. Which statement interprets this estimate correctly?
Question 3: How does the ALDH2*2 variant protect its carriers against alcohol use disorder?
Question 4: What did Barr and colleagues (2020) find when they tested polygenic scores for alcohol use disorder?
Development and Temperament: Adolescence, Early Use, Self-Control and Mental Health
Introduction
Section 1 showed that genes shape vulnerability to addiction, mostly through many small influences that unfold over a lifetime. This section follows that lifetime. Volkow, Koob and McLellan (2016) list the factors that increase vulnerability to addiction as family history, early exposure to drug use (“adolescence is among the periods of greatest vulnerability”), high-risk environments and certain mental illnesses. Section 1 covered family history, and Section 3 takes up environments. This section examines the developmental and psychological factors: the maturing adolescent brain, the age at which use begins, self-control in childhood, personality and co-occurring mental illness.
A theme runs through all five. Each factor is associated with later substance use disorder, and for each the harder question is whether it causes the disorder or marks a person who was already at higher risk for other reasons. Working through that question is the main skill this section builds.
Learning Objectives
- 3.1 Describe how the brain’s reward and control systems mature at different rates during adolescence, and explain how this pattern is linked to risk taking.
- 3.2 Describe the association between age at first drink and later alcohol dependence, and calculate a relative risk and a risk difference from survey data.
- 4.1 Explain why the association between early drinking and later dependence may not be causal, distinguishing a causal explanation from a common liability explanation with evidence from twin studies.
- 5.1 Describe the evidence from the Dunedin cohort that childhood self-control predicts adult substance dependence.
- 5.2 Describe the four personality dimensions measured by the Substance Use Risk Profile Scale.
- 5.3 Explain three ways in which mental illness and substance use disorders may come to co-occur, using survey data from the United States and Canada.
3.1 The Adolescent Brain
Adolescence is the period when most people first try alcohol and other drugs, and it is also a period of rapid brain development. Brain imaging studies that follow the same young people over many years show that different regions mature on different schedules.
Gogtay and colleagues (2004) scanned 13 healthy children with magnetic resonance imaging every two years for 8 to 10 years, covering ages 4 to 21. Their time-lapse maps of the cortex showed that higher-order association areas mature only after the lower-order sensory and visual areas whose functions they integrate, and that evolutionarily older brain areas mature earlier than newer ones. The prefrontal cortex, whose role in self-regulation Lesson 8 described, is one of these higher-order association areas.
Two influential reviews built on findings like these to explain why risk taking rises in adolescence. Laurence Steinberg (2008) proposed that risk taking increases between childhood and adolescence because changes around puberty in the brain’s socio-emotional system increase reward seeking, especially in the presence of peers, driven mainly by a remodelling of the brain’s dopamine system. Risk taking declines between adolescence and adulthood as the brain’s cognitive control system matures through adolescence and young adulthood, improving the capacity for self-regulation. Because the two systems develop on different timetables, he argued, mid-adolescence is a time of heightened vulnerability to risky behaviour. Casey, Jones and Hare (2008) reached a similar conclusion from imaging and animal studies: adolescents show heightened responsiveness to incentives and social and emotional contexts at a time when impulse control is still relatively immature. Figure 9.3 sketches this dual systems model.
The model describes average trends. Casey and colleagues (2008) noted that the imbalance may be greater in adolescents prone to strong emotional reactions, which is one way that temperament, discussed below, interacts with development. Later reviews have tested the model against newer data. Shulman and colleagues (2016) found the predicted average trends broadly supported, with reward sensitivity peaking in the late teens and cognitive control improving into the early twenties, while acknowledging that the model is an oversimplification. Meisel and colleagues (2019) argued that its central prediction, that a larger gap between the two systems leads to more risk taking, had not been rigorously tested. Research on the model is still evolving. The next question is whether starting substance use during this window itself raises the risk of a later disorder.
3.2 Age at First Use
Studies of age at first use ask whether people who start using a substance early are more likely to develop a disorder later. One example comes from the 1992 National Longitudinal Alcohol Epidemiologic Survey in the United States. Grant and Dawson (1997) analyzed 27,616 current and former drinkers and found that the rate of lifetime alcohol dependence declined from more than 40 percent among those who started drinking at age 14 or younger to roughly 10 percent among those who started at age 20 or older. After adjustment for other factors, the odds of dependence fell by 14 percent for each additional year of age at first drink. The survey used the DSM-IV categories of alcohol abuse and alcohol dependence, which Lesson 1 explained were later combined into alcohol use disorder.
Two simple measures compare the risk of an outcome in two groups. The relative risk divides the risk in one group by the risk in the other, and says how many times as likely the outcome is. The risk difference subtracts one risk from the other, and says how many extra cases occur per 100 people. Equation 9.2 gives both.
Worked Example 9.2. Using the approximate figures reported by Grant and Dawson (1997), take the risk of lifetime alcohol dependence as 40 percent for people who first drank at 14 or younger and 10 percent for people who first drank at 20 or older. The relative risk is 40 ÷ 10 = 4.0, so early starters were about four times as likely to report dependence. The risk difference is 40 − 10 = 30 percentage points, or about 30 extra cases for every 100 early starters. Because the abstract reports “more than 40 percent” and “roughly 10 percent”, these results are approximations. Calculator 9.2 lets you enter other pairs of risks and compare the two measures.
4.1 Does Early Drinking Cause Later Dependence?
A relative risk of 4 invites a policy conclusion: delay the first drink and dependence will fall. That conclusion holds only if early drinking causes later dependence. The box below explains why an association like this one can arise without a causal link.
Association, confounding and common liability
Two things can be associated because one causes the other, or because a third factor causes both. A third factor that produces an association between an exposure and an outcome is called a confounder. In addiction research, a common candidate is common liability: a set of genetic and family influences that leads some young people both to start drinking early and, separately, to develop a disorder later. If common liability explains the association, delaying the first drink would leave the underlying risk unchanged. Studies that compare twins or siblings, who share much of this liability, can test the idea.
Prescott and Kendler (1999) carried out exactly this test with 8,746 adult twins from a population-based registry. They found a strong association between early drinking and alcohol dependence, replicating earlier work. Comparisons within twin pairs, however, suggested that all of the association was due to familial sources, probably both genetic and shared environmental. The authors concluded that the association between age at first drink and alcohol dependence is noncausal and that “attempts to prevent the development of AD [alcohol dependence] by delaying drinking onset are unlikely to be successful.” Figure 9.4 contrasts the two explanations.
Later twin studies have qualified this conclusion. In the Minnesota Twin Family Study, Irons, Iacono and McGue (2015) assessed 1,512 twins at about ages 11, 14 and 24. Twins who drank in early adolescence had higher risk than their co-twins who had not for several of the adult outcomes studied, including an index of adult alcohol use, which points to some effect of early drinking itself. Agrawal and colleagues (2009), studying 6,257 adult twins in Australia, found that genetic influences on alcohol dependence symptoms were considerably larger among people who first drank before age 13, which suggests that early drinking may allow genetic vulnerability to be expressed. For alcohol dependence, these twin studies suggest that much of the association reflects earlier vulnerability, while a smaller causal contribution of early drinking remains possible. A systematic review of prospective cohort studies by Maimaris and McCambridge (2014) found only five eligible studies, reported that rigorous control for confounding markedly weakened or removed the associations, and concluded that there is no strong evidence that starting to drink earlier leads to adult alcohol problems. The authors called for more research, and the question remains open.
Early use can also carry other harms. Lesson 1 noted that Canada’s Lower-Risk Cannabis Use Guidelines advise delaying cannabis use until after adolescence; Lesson 7 examines the evidence behind that advice, and HSCI 402 develops prevention programs. The following parts of the section look at two markers that can be measured well before any substance use begins.
5.1 Childhood Self-Control
If vulnerability is visible before adolescence, it should be possible to detect it in childhood. A meta-analysis of 150 studies by Robson and colleagues (2020) found that better self-regulation in childhood was associated with less smoking and illicit drug use in early adolescence and with less alcohol and substance abuse in adulthood. One example is a cohort of about 1,000 children in Dunedin, New Zealand, followed from birth to age 32 by Moffitt and colleagues (2011). Childhood self-control, the capacity to regulate attention, emotion and behaviour, predicted adult physical health, substance dependence, personal finances and criminal offending. The relationship followed a gradient: outcomes worsened step by step from the highest to the lowest levels of childhood self-control.
Two features make the study persuasive. The effects of self-control could be separated from children’s intelligence and social class, and from mistakes they made as adolescents. In a second cohort of 500 British sibling pairs (non-identical twins) followed to age 12, the sibling with lower self-control at age 5 was more likely to have started smoking, to perform poorly in school and to behave antisocially, despite the shared family background (Moffitt et al., 2011; King’s College London, 2011). This comparison addresses part of the confounding problem raised earlier in the section (4.1 Does Early Drinking Cause Later Dependence?), although the sibling outcomes were measured in early adolescence, before substance dependence could be assessed. The authors suggested that interventions addressing self-control might reduce a wide range of costs to society. Because risk rose across the whole range of self-control, their suggestion concerns whole populations of children.
5.2 Personality Risk Profiles
Self-control is one dimension of temperament. Researchers have also asked whether distinct personality profiles lead to substance use by different routes. The Substance Use Risk Profile Scale (SURPS), developed by Patricia Conrod and colleagues, measures four personality dimensions: hopelessness, anxiety sensitivity, impulsivity and sensation seeking. The model behind the scale proposes that each dimension relates to particular patterns of substance use through a different kind of reinforcement, the learning process described in Lesson 8 (Woicik et al., 2009). In studies of undergraduate and high school students, the scale showed a stable four-factor structure and good reliability, and it identified younger students with problem drinking (Woicik et al., 2009). The flip cards below describe each dimension.
The profiles have been used to design prevention. In the Adventure trial, run in London secondary schools, Conrod and colleagues (2013) trained school staff to deliver brief interventions to 1,210 students with one of the four high-risk profiles, with 1,433 lower-risk students also followed; the students’ mean age was 13.7 years. Schools were randomly allocated to the program or to usual drug education. Over two years, the program reduced drinking rates, binge drinking and the growth of problem drinking among high-risk students, with some smaller benefits for low-risk students in the same schools. At the time of the trial report, Conrod was based at the Sainte-Justine university hospital research centre in Montréal. HSCI 402 develops school prevention programs and their evaluation.
5.3 Co-occurring Mental Illness
Substance use disorders and other mental disorders often occur in the same person, a pattern called co-occurring disorders or comorbidity. A meta-analysis of 22 population surveys by Lai and colleagues (2015) found strong associations between substance use disorders and mood and anxiety disorders, strongest between illicit drug use disorders and major depression (pooled odds ratio 3.80). An earlier example is the Epidemiologic Catchment Area study in the United States, which interviewed 20,291 people in the community and in institutions. Regier and colleagues (1990) reported that people with a mental disorder had 2.7 times the odds of also having an addictive disorder, with a lifetime prevalence of about 29 percent. Among people with an alcohol disorder, 37 percent had another mental disorder, and among people with other drug disorders, 53 percent did (odds ratio 4.5). Comorbidity was higher still among people in treatment settings and highest in prisons.
A national Canadian estimate came from the 2002 Canadian Community Health Survey on Mental Health and Well-being, with 36,984 respondents. Rush and colleagues (2008) found a 12-month population prevalence of co-occurring substance use and other mental disorders of 1.7 percent. Other mental disorders were more common among people with illicit drug problems than among those with alcohol problems, and more common among people with dependence than among those with less severe problems. Rates were higher in British Columbia and lower in Quebec. The authors noted that these Canadian rates sat at the lower end of the international range, possibly because the survey excluded several disorders known to co-occur often with substance use disorders. The accordion below sets out three explanations for co-occurrence that researchers continue to test.
The same genes, childhood experiences and environments may raise the risk of both kinds of disorder. Kendler and colleagues’ (2015) Swedish adoption study found that genetic risk for alcohol use disorder reflected both a specific liability and a liability shared with other externalizing disorders, and identified a subgroup of adoptees with alcohol use disorder who had high rates of psychiatric illness. Section 3 examines childhood adversity, which raises the risk of many conditions at once.
People may use substances to manage symptoms such as anxiety, low mood or trauma-related distress. Lesson 10 develops this self-medication hypothesis, and the SURPS dimensions of hopelessness and anxiety sensitivity in 5.2 Personality Risk Profiles describe traits that may lead to use for relief.
Substance use and withdrawal can produce or worsen psychiatric symptoms. Lesson 6 described stimulant psychosis and Lesson 7 the association between daily high-potency cannabis use and psychosis. HSCI 402 develops the treatment of concurrent disorders, and HSCI 214 develops mental illness in depth.
The three explanations are compatible, and the balance among them probably differs across disorders and people. Table 9.2 summarizes the developmental factors in this section.
Table 9.2. Developmental and psychological factors associated with later substance use disorder, with the main evidence and the main caution for each.
| Factor | Key evidence | Finding | Main caution |
|---|---|---|---|
| Adolescent brain development | Gogtay et al. (2004); Steinberg (2008); Casey et al. (2008) | Reward sensitivity peaks before cognitive control matures | Describes average trends; individuals vary, and its central prediction needs stronger tests (Meisel et al., 2019) |
| Early first drink | Grant & Dawson (1997) | Lifetime dependence above 40 percent for a first drink at 14 or younger, about 10 percent at 20 or older | Twin studies trace much of the association to familial factors (Prescott & Kendler, 1999; Irons et al., 2015), and a review found no strong evidence of a causal effect (Maimaris & McCambridge, 2014) |
| Low childhood self-control | Robson et al. (2020); Moffitt et al. (2011) | Better self-regulation linked to less later substance use; a gradient of adult dependence in one cohort | Observational studies; mechanisms are not established |
| Personality risk profiles | Woicik et al. (2009); Conrod et al. (2013) | Four dimensions linked to different patterns of use; targeted program reduced drinking | Trial outcomes measured over two years in one city |
| Co-occurring mental illness | Lai et al. (2015); Regier et al. (1990); Rush et al. (2008) | Comorbidity is common, especially with drug disorders and in treatment settings | Direction of influence differs across people |
Consider this claim: “People who start drinking at 14 or younger are four times as likely to develop alcohol dependence as people who start at 20 or older, so delaying everyone’s first drink to age 20 would cut alcohol dependence by three quarters.” Use Calculator 9.2, the box on association, confounding and common liability in 4.1 Does Early Drinking Cause Later Dependence? and the twin study by Prescott and Kendler (1999) to write a short response that identifies the assumption the claim makes and explains what the twin evidence implies about it. Then open the item below to compare.
The relative risk of about 4 (40 percent compared with 10 percent) is a real association, but the claim assumes that early drinking causes later dependence, so that changing the age of the first drink would change the risk. Prescott and Kendler (1999) found that analyses within twin pairs attributed all of the association to familial sources, both genetic and shared environmental. Their conclusion was that the association is noncausal and that delaying drinking onset is unlikely to prevent dependence. Later co-twin evidence suggests that early drinking may add some risk of its own (Irons et al., 2015), but because much of the association reflects familial liability, any reduction would be far smaller than three quarters. Even a fully causal effect would cut dependence by about three quarters only among early starters, from about 40 to about 10 percent (a fall of 30 of 40 percentage points); people who already start at 20 or older would see no change, so the fall across the whole population would be smaller still. The claim therefore does not follow from the survey data, even if delaying use is advised for other reasons.
Summary
In the dual systems model, adolescence brings a period in which sensitivity to reward and peers rises before the brain’s control systems mature; reviews find these average trends broadly supported while still debating the model, and most substance use begins in this period. In one national survey, people who started drinking early were about four times as likely to report later dependence, but twin studies trace much of this association to familial liability, and whether early drinking also causes dependence remains uncertain. Lower childhood self-regulation is linked to later substance use, and in one cohort it predicted adult substance dependence along a gradient, personality dimensions such as sensation seeking and anxiety sensitivity mark different routes into use, and mental illness and substance use disorders frequently co-occur for several reasons. Many of these individual differences are themselves shaped by experience. Section 3 examines that experience directly: childhood adversity, social rank and the physical and social environment, including the Rat Park experiments carried out at Simon Fraser University.
Key Takeaways
- 3.1 Imaging studies that follow the same young people show that higher-order association areas, including the prefrontal cortex, mature late (Gogtay et al., 2004). In the dual systems model, reward sensitivity rises sharply around puberty, especially in the presence of peers, while cognitive control matures slowly through young adulthood, so mid-adolescence is a window of heightened vulnerability (Steinberg, 2008; Casey et al., 2008). The model describes average trends, individuals vary, and later reviews find the trends broadly supported while calling for stronger tests of its central prediction.
- 3.2 In the 1992 National Longitudinal Alcohol Epidemiologic Survey, lifetime alcohol dependence was more than 40 percent among people who first drank at 14 or younger and roughly 10 percent among those who first drank at 20 or older, and the odds fell by 14 percent for each later year of onset (Grant & Dawson, 1997). The relative risk is 40 ÷ 10 = 4.0 and the risk difference is 30 percentage points, or about 30 extra cases per 100 early starters. The same relative risk can correspond to very different numbers of extra cases.
- 4.1 An association can arise because one factor causes the other or because a third factor, a confounder, causes both. Within twin pairs, Prescott and Kendler (1999) suggested that all of the association between early drinking and dependence was due to familial sources, which supports common liability, although whether early drinking also causes dependence remains uncertain. Later twin studies suggest that early drinking may add a smaller risk of its own or allow genetic vulnerability to be expressed (Irons et al., 2015; Agrawal et al., 2009).
- 5.1 In the Dunedin cohort of about 1,000 children followed to age 32, lower childhood self-control predicted adult substance dependence, poorer health, finances and offending along a gradient, independently of intelligence and social class (Moffitt et al., 2011). In 500 British sibling pairs, the sibling with lower self-control at age 5 was more likely to have started smoking by age 12, which addresses part of the confounding problem. Because risk rose across the whole range, the authors’ suggestion of self-control interventions concerns whole populations of children.
- 5.2 The Substance Use Risk Profile Scale measures hopelessness, anxiety sensitivity, impulsivity and sensation seeking, each linked to particular patterns of use through a different kind of reinforcement (Woicik et al., 2009). In the Adventure trial in London schools, brief interventions targeted to these profiles reduced drinking, binge drinking and the growth of problem drinking among high-risk students over two years (Conrod et al., 2013).
- 5.3 In the Epidemiologic Catchment Area study, people with a mental disorder had 2.7 times the odds of an addictive disorder, and 37 percent of people with an alcohol disorder and 53 percent with other drug disorders had another mental disorder, with higher rates in treatment settings and prisons (Regier et al., 1990). A Canadian estimate put the 12-month prevalence of co-occurring disorders at 1.7 percent (Rush et al., 2008). Shared risk factors, substance use as a response to symptoms, and substance use causing or worsening symptoms are compatible explanations whose balance differs across disorders and people.
Reflection
A hypothetical school board is choosing between two programs. Program A is a campaign to delay the first drink, justified by Grant and Dawson’s (1997) finding that lifetime alcohol dependence was more than 40 percent among people who first drank at 14 or younger and roughly 10 percent among those who first drank at 20 or older. Program B aims to strengthen self-control in all elementary school children, justified by Moffitt et al.’s (2011) finding, in about 1,000 children of the Dunedin cohort in New Zealand followed to age 32, that childhood self-control predicted adult substance dependence along a gradient independently of intelligence and social class; in a second cohort of 500 British sibling pairs, the sibling with lower self-control at age 5 was more likely to have started smoking and to behave antisocially by age 12. Prescott and Kendler (1999) found, in 8,746 adult twins, that the association between early drinking and alcohol dependence was due to familial sources, while Irons, Iacono and McGue (2015) found, in 1,512 Minnesota twins, that twins who drank in early adolescence had higher risk than their co-twins for several adult outcomes. In 150 to 250 words, calculate the relative risk behind Program A, explain what each program assumes about causation, and say which evidence supports or challenges each assumption.
Minimum 20 characters required.
Question 1: Grant and Dawson (1997) reported lifetime alcohol dependence of about 40 percent among people who first drank at 14 or younger and about 10 percent among those who first drank at 20 or older. What is the relative risk?
Question 2: What did Prescott and Kendler (1999) conclude from their study of 8,746 adult twins?
Question 3: According to the dual systems model described by Steinberg (2008) and Casey et al. (2008), why is mid-adolescence a period of heightened vulnerability?
Question 4: Which four dimensions does the Substance Use Risk Profile Scale measure?
Adversity, Social Rank and Environment: From Childhood Adversity to Rat Park
Content note. This section discusses childhood abuse, neglect and exposure to violence in the home. 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
Sections 1 and 2 looked for the sources of vulnerability inside the person: genes, brain development, temperament and mental health. This section turns to the circumstances around the person. Volkow, Koob and McLellan (2016) describe high-risk environments as “socially stressful environments with poor familial and social supports and restricted behavioral alternatives,” together with environments where drugs are easy to obtain and attitudes toward drug taking are permissive. Each part of that description can be tested.
The section begins with childhood adversity in human populations, then turns to two lines of animal research that manipulated the environment directly: studies of social rank in monkeys and the Rat Park experiments carried out at Simon Fraser University in the late 1970s. Rat Park is widely discussed, and the section examines both its findings and its contested replication record. HSCI 214 develops adverse childhood experiences and the biology of the stress response, and Lesson 10 develops the social determinants of substance use; this section concentrates on what the research shows about vulnerability to addiction.
Learning Objectives
- 6.1 Describe the evidence from the Adverse Childhood Experiences study and Canadian survey data that links childhood adversity to later substance use problems, and state its main limitations.
- 6.2 Explain how social rank changed dopamine receptors and cocaine self-administration in male and female monkeys.
- 7.1 Describe the design and findings of the Rat Park experiments at Simon Fraser University.
- 7.2 Describe the main criticisms of the Rat Park design.
- 7.3 Evaluate the replication record of Rat Park alongside more recent animal studies of social reward.
- 7.4 Explain what animal models of the environment can and cannot tell us about human addiction.
6.1 Childhood Adversity
Much of the human evidence on early environment comes from studies of adverse childhood experiences (ACEs): abuse, neglect and serious problems in the household, such as violence between parents. Researchers count how many categories of adversity a person experienced in childhood to give an ACE score.
Dube and colleagues (2003) examined illicit drug use among 8,613 adults who attended a primary care clinic in California and completed a survey on 10 categories of childhood adversity. Each category of adverse experience increased the likelihood of starting illicit drug use by age 14 two- to four-fold. The ACE score showed a strong dose-response relationship, meaning that risk rose step by step with each additional category, for starting drug use at every age, for drug use problems, for addiction and for injection drug use. Compared with people who reported no ACEs, people with five or more were 7 to 10 times as likely to report illicit drug use problems, addiction to illicit drugs and injection drug use. The relationship held in each of four birth cohorts dating back to 1900, which led the authors to argue that the effect of adversity persisted through changes in drug availability and attitudes over the century.
Canadian data show a similar pattern. Afifi and colleagues (2014) analyzed the 2012 Canadian Community Health Survey on Mental Health, with 23,395 respondents aged 18 and older in the ten provinces. Thirty-two percent reported at least one of three types of child abuse: physical abuse, sexual abuse or exposure to intimate partner violence. Every type was associated with every mental condition studied, including alcohol abuse or dependence and drug abuse or dependence (the older DSM-IV categories that the survey used), after adjustment for sociodemographic factors (adjusted odds ratios from 1.4 to 7.9), and the odds rose with the number of abuse types experienced. All three types remained associated with drug abuse or dependence after full adjustment.
A meta-analysis has combined studies of this kind. Hughes and colleagues (2017) pooled 37 studies with 253,719 adult participants, leaving out high-risk and clinical samples, and found that people with at least four adverse childhood experiences had more than seven times the odds of problematic drug use of people with none, one of the strongest associations among the 23 health outcomes examined.
Both studies relied on adults recalling their childhoods, so some experiences may be forgotten or reported differently by people with current problems. Both measured childhood adversity and adult outcomes at the same time, which limits conclusions about cause. The confounding problem from Section 2 also applies: a parent with a substance use disorder may both create adversity in the home and pass on genetic liability, so some of the association could reflect shared genes. The consistency of the dose-response relationship across studies and countries, and across birth cohorts, makes a causal contribution of adversity plausible, but human surveys cannot isolate it. That is one reason researchers turned to animal studies, in which the environment can be assigned by the experimenter.
6.2 Social Rank, Dopamine and Cocaine in Monkeys
In an animal study, the researcher decides which environment each animal lives in, so differences that emerge afterward can be attributed to the environment. Studies of monkeys are especially informative because macaques live in social groups with stable hierarchies of dominant and subordinate animals.
Morgan and colleagues (2002) used positron emission tomography (PET), a brain imaging method that uses a radioactive tracer, to measure the availability of dopamine D2 receptors in 20 male cynomolgus macaques. Lesson 8 described dopamine’s role as a teaching signal in the reward system; D2 receptors are one of the receptor types through which dopamine acts. While the monkeys lived alone, their D2 receptor availability did not differ. After they were moved into social groups and hierarchies formed, D2 availability rose in the monkeys that became dominant and did not change in subordinate monkeys. The researchers then allowed the monkeys to self-administer cocaine. Cocaine functioned as a reinforcer, meaning the monkeys worked to obtain it, in subordinate monkeys but not in dominant ones. The authors concluded that changes in an organism’s environment can produce profound biological changes, including changes in vulnerability to cocaine.
The same research group later repeated the design with 16 female monkeys (Nader et al., 2012). D2-type receptor availability again rose in females that became dominant, and the availability of the dopamine transporter fell in subordinate females. The behavioural result was reversed: dominant females acquired cocaine self-administration at lower doses than subordinate females. The authors concluded that the social environment profoundly affects the dopamine system in both sexes, with different functional consequences for females and males. Figure 9.5 summarizes the two studies.
The monkey studies support the central claim of this section: the social environment can change the brain in ways that alter vulnerability to a drug. They also warn against a simple story in which low status always means more drug taking. The same environmental change had different behavioural consequences depending on sex, and the samples were small, so the findings are best treated as suggestive. A widely discussed environmental experiment in addiction research, conducted with rats in Burnaby, raises similar questions.
7.1 Rat Park
By the 1970s, drug self-administration by rats and monkeys housed in laboratory cages was a well-established research model (Gage & Sumnall, 2019). Bruce K. Alexander and his colleagues at Simon Fraser University asked how much the isolation of caged animals contributed to their drug taking. Case 9.2 describes what they built.
Alexander’s team built a colony enclosure of 8.8 square metres, covered the floor with cedar sawdust, and added open boxes, tubes and a climbing pole, so that rats could live together in a setting closer to their natural social life. They compared rats in this “Rat Park” with rats housed alone in standard laboratory cages measuring 18 by 25 by 18 centimetres, with no view of other animals (Gage & Sumnall, 2019).
In the first study, rats were given morphine solution as their only fluid for 57 days and then offered a choice between morphine and water. On choice days, and during a later period of abstinence, the isolated rats drank significantly more morphine solution than the social rats, and females drank more than males (Alexander, Coambs & Hadaway, 1978). In a second study with rats never before exposed to morphine, isolated rats offered increasingly sweet morphine-sucrose solutions drank significantly more; in the phase of greatest consumption, isolated females drank five times and isolated males sixteen times as much morphine (by body weight) as their colony counterparts (Hadaway et al., 1979). In a third study, rats were raised alone or in the colony and then half were switched; rats living in the colony at the time of testing drank less morphine solution than isolated rats, regardless of where they had been raised (Alexander et al., 1981).
The original authors drew measured conclusions. In the 1978 paper they wrote that “sex and housing are important variables” and that theories ignoring them “run the risk of oversimplification” (quoted in Gage & Sumnall, 2019). The 1981 paper suggested that colony rats avoid morphine because it interferes with complex, species-specific behaviour such as the social life of the colony. Gage and Sumnall note that the original papers did not draw conclusions about human addiction. Alexander later developed those implications in an adaptive model of addiction and in his 2008 book The Globalisation of Addiction, which Lesson 10 discusses.
7.2 Criticisms of the Rat Park Design
Writing in the journal Addiction, Suzanne Gage and Harry Sumnall (2019) reviewed the three studies four decades later. They identified several features of the design that limit confidence in the results. The accordion below sets them out.
Isolated rats’ fluid intake was measured by weighing their bottles, with a crude adjustment for leaks and evaporation. Colony rats drank from a tube large enough for one rat, where a camera identified the animal and counters recorded each drop. When groups are measured differently, a systematic error in one method can create a difference that is not really there. Gage and Sumnall note that water consumption did not follow the same pattern, which offers some protection against this bias.
The first study began with 32 rats: 4 isolated females, 6 isolated males, 10 social females and 12 social males. By its end only 2 isolated females, 6 isolated males, 7 social females and 12 social males remained, 27 rats in all, and morphine consumption was highest in the group with only two animals. Gage and Sumnall's text refers to four deaths, while their summary table records five (two females in each housing condition during the period of forced morphine drinking and one social female during abstinence), which matches the 27 survivors. Small groups produce unstable estimates.
Morphine consumption was measured in three ways, across eight choice days and four groups. With so many comparisons, some differences will appear by chance, and differences emerged only under specific conditions, such as particular concentrations of morphine and sugar.
Males and females lived together in Rat Park, so pregnancy, nursing and competition for resources may have affected females’ behaviour and drinking, and the papers do not explain how offspring were handled. Isolated rats faced none of these influences.
Gage and Sumnall also corrected a common misreading: the rats in Rat Park did drink some morphine. The finding was a difference in amount between housing conditions, which is a weaker claim than the popular version in which social rats refused the drug.
7.3 The Replication Record
A finding gains credibility when independent researchers obtain it again. Gage and Sumnall (2019) describe independent attempts to repeat and extend the Rat Park findings as producing mixed results. An earlier study by Hill and Powell (1976) had reared rats for 100 days in enriched or impoverished conditions and offered morphine or cocaine in drinking water; morphine intake did not differ, and enriched rats drank significantly more cocaine. Bozarth, Murray and Wise (1989) tested intravenous self-administration and found that group-housed and isolated rats both quickly learned to self-administer heroin and cocaine. Isolated rats learned to take heroin earlier, but the two groups took similar amounts by the fifth week, and cocaine intake did not differ. The authors concluded that isolation is not a necessary condition for heroin or cocaine to be reinforcing. Gage and Sumnall (2019) summarize this study differently, describing isolated rats that quickly learned to self-administer cocaine and socially housed rats that did not consistently do so; the result given here is the one Bozarth and colleagues report in their own abstract, in which the two groups learned cocaine self-administration equally quickly and took similar amounts. Petrie (1996) tried twice to repeat the 1979 study and could not replicate it, because the isolated rats drank much less morphine solution than in the original; Petrie suggested that the supplier’s rat strain had changed.
More recent work has revived the social question with stronger designs. Venniro and colleagues (2018), at the United States National Institute on Drug Abuse, trained rats to choose between a drug and the opportunity to interact with another rat. Across both sexes, two drugs (methamphetamine and heroin) at different doses, and different housing conditions, the opportunity for social interaction prevented drug self-administration. The protection weakened when the social reward was delayed or punished. The authors argued that neuroscience-based addiction research needs to incorporate social factors. Research on social reward and drug use is still evolving. Use Interactive 9.1 to weigh these studies against the claim that a richer social environment reduces drug taking.
Weigh the Rat Park evidence
Why Rat Park became famous
Gage and Sumnall (2019) point out that popular talks about addiction cite Rat Park, that a comic and an animation about it have had millions of views, and that organizations campaigning for drug policy reform cite it alongside Lee Robins’s studies of Vietnam veterans (Section 4). In their assessment, the popular interpretation that Rat Park provided a new understanding of addiction is overstated: it tested one manipulation (housing) on one behaviour (oral morphine intake), and most rodent self-administration studies, as of 2019, still housed animals individually. They also judge its core message, that addiction is more complicated than a biological response to a drug, to be important, and they suggest it has endured because advocating policy change requires a good story with at least some basis in evidence. They urge readers to read the study itself with a critical eye, and Hall and Weier (2017) make a similar recommendation about Robins’s work: seminal studies need to be read in the original.
7.4 From Animal Models to Human Environments
Animal studies allow researchers to assign environments, but human environments differ from cages and enclosures in many ways. Gage and Sumnall (2019) note that in human life the supply, availability and opportunity to use substances are major influences that no experimenter controls, and that “environment” includes poor access to services and social exclusion. Hall and Weier (2017) make a similar point about Lee Robins’s research on soldiers who used heroin in Vietnam, which Section 4 examines: it shows how price, availability, social norms, education and life circumstances shape addiction beyond the pharmacology of the drug. Table 9.3 compares the lines of evidence in this section.
Table 9.3. Evidence that environments shape vulnerability to addiction comes from different designs, each with its own strengths and limits.
| Line of evidence | Design | Main finding | Main limit |
|---|---|---|---|
| Adverse childhood experiences | Adult surveys recalling childhood (Dube et al., 2003; Afifi et al., 2014) and a meta-analysis (Hughes et al., 2017) | Dose-response relationship between adversity and drug problems; more than seven times the odds of problematic drug use with four or more experiences | Recall and confounding by family liability |
| Social rank in monkeys | Experimental housing with brain imaging (Morgan et al., 2002; Nader et al., 2012) | Rank changed dopamine receptors and cocaine reinforcement | Small samples; direction differed by sex |
| Rat Park | Colony compared with isolated housing (Alexander et al., 1978 to 1981) | Isolated rats drank more morphine solution | Small groups, unequal measurement, mixed replication |
| Social choice in rats | Choice between drug and social interaction (Venniro et al., 2018) | Social reward prevented drug self-administration | Laboratory choice differs from human social life |
Suppose you could run Rat Park again. Using the four criticisms in the accordion in 7.2 Criticisms of the Rat Park Design, describe three changes to the design that would make its results more convincing, and say which criticism each change addresses. Then open the item below to compare.
First, measure fluid intake in the same way for both groups, for example with the same individually identified drinking stations in colony and cage housing, which removes the bias from unequal measurement. Second, use larger groups of each sex, so that a result does not depend on two animals, and house males and females in separate colonies so that pregnancy and nursing cannot affect drinking. Third, state in advance which outcome and which comparison will test the main hypothesis, so that a difference found in one of many comparisons is not reported as the main finding. Venniro and colleagues’ (2018) choice design shows another improvement: offering social interaction and the drug as direct alternatives tests whether social reward competes with the drug.
Summary
Childhood adversity shows a dose-response relationship with later drug problems in large American and Canadian samples, and a meta-analysis linked four or more adverse experiences to more than seven times the odds of problematic drug use (Hughes et al., 2017), although recall and shared family liability limit causal conclusions. Two small experiments in monkeys suggest that social rank changes the dopamine system and the reinforcing effect of cocaine, with opposite behavioural results in males and females. Rat Park, carried out at Simon Fraser University in the late 1970s, found that isolated rats drank more morphine solution than rats housed in a social colony; its small groups, unequal measurement and mixed replication record led Gage and Sumnall (2019) to judge its popular interpretation overstated, while a newer social choice study supports the idea that social reward can compete with drugs. Environments matter, and they act alongside the individual differences in Sections 1 and 2. Section 4 asks what happens over time: how often people recover from addiction, often without treatment, and what that implies for the debate over whether addiction is a brain disease.
Key Takeaways
- 6.1 Among 8,613 Californian adults, each category of childhood adversity raised the likelihood of starting illicit drug use by age 14 two- to four-fold, and people with five or more ACEs were 7 to 10 times as likely to report drug problems, addiction and injection drug use, a dose-response relationship that held across four birth cohorts (Dube et al., 2003). In Canada, 32 percent of adults reported child abuse, and every type was associated with drug abuse or dependence (Afifi et al., 2014). Recall, measurement at a single time and confounding by family liability limit causal conclusions from these surveys.
- 6.2 In 20 male macaques, D2 receptor availability rose in monkeys that became dominant, and cocaine reinforced subordinate but not dominant monkeys (Morgan et al., 2002). In 16 females, D2-type availability again rose in dominant animals and the dopamine transporter fell in subordinates, but dominant females acquired cocaine self-administration at lower doses (Nader et al., 2012). The social environment can change the brain in ways that alter vulnerability, but the consequences differed by sex and the samples were small.
- 7.1 Bruce Alexander’s team at Simon Fraser University compared rats living in an 8.8 square metre colony with rats housed alone in small laboratory cages. Across three studies in the late 1970s and 1981, isolated rats drank more morphine solution than colony rats, and rats living in the colony at the time of testing drank less regardless of where they had been raised. The original papers drew measured conclusions and did not make claims about human addiction.
- 7.2 Gage and Sumnall (2019) identified four problems: the two groups’ intake was measured by different methods, the groups were very small (only two isolated females remained in the first study), many comparisons were made, and males and females lived together in the colony. They also noted that colony rats did drink some morphine, so the finding was a difference in amount, not a refusal of the drug.
- 7.3 Gage and Sumnall describe independent studies as giving mixed results: Hill and Powell (1976) found no housing difference for morphine, Bozarth and colleagues (1989) found that isolated and group-housed rats both learned to self-administer heroin and cocaine, and Petrie (1996) could not replicate the 1979 study. Venniro and colleagues (2018) found that the opportunity for social interaction prevented drug self-administration across sexes, drugs and doses, which revives the social question with a stronger design. Gage and Sumnall judged Rat Park’s popular interpretation overstated, although its core message that addiction is more than a biological response to a drug remains important.
- 7.4 Animal studies let researchers assign environments, but human environments include supply, availability, opportunity, poor access to services and social exclusion, which no experimenter controls (Gage & Sumnall, 2019). Robins’s research on soldiers returning from Vietnam shows how price, availability, social norms, education and life circumstances shape addiction beyond pharmacology (Hall & Weier, 2017). Each line of evidence has its own strengths and limits, and together they show that environments matter alongside individual differences.
Reflection
A popular video states: “Rat Park proved that isolation causes addiction and that drugs themselves are harmless.” Using the following evidence from Section 3, write 150 to 250 words evaluating this claim: (1) in the first Rat Park study, isolated rats drank more morphine solution than colony rats, but by the end only 2 isolated females, 6 isolated males, 7 colony females and 12 colony males remained, and the two groups’ intake was measured by different methods; (2) colony rats drank some morphine; (3) Hill and Powell (1976) found no housing difference in morphine intake, Bozarth et al. (1989) found that isolated and group-housed rats both learned to self-administer heroin and cocaine, and Petrie (1996) could not replicate the 1979 study; (4) Venniro et al. (2018) found that the opportunity for social interaction prevented drug self-administration in rats; (5) Morgan et al. (2002) and Nader et al. (2012) found that social rank changed dopamine receptors in monkeys, with cocaine more reinforcing for subordinate males and for dominant females.
Minimum 20 characters required.
Question 1: In the Adverse Childhood Experiences study (Dube et al., 2003), how did people with five or more adverse childhood experiences compare with people who reported none?
Question 2: What did the monkey studies by Morgan et al. (2002) and Nader et al. (2012) show about social rank and cocaine?
Question 3: What was Gage and Sumnall’s (2019) main concern about how fluid intake was measured in Rat Park?
Question 4: Which statement best summarizes the replication record of Rat Park?
Natural Recovery and the Brain Disease Model Debate
Introduction
Sections 1 to 3 asked who develops addiction. This section asks what happens afterward, and the answer depends on where one looks. In treatment programs, addiction often appears as a long-lasting condition with frequent relapse. McLellan and colleagues (2000) compared drug dependence with type 2 diabetes, high blood pressure and asthma and found that rates of relapse and of adherence to treatment were similar across all four conditions. In surveys of the general population, by contrast, many people who have ever met the criteria for a substance use disorder no longer do, and many of them stopped without treatment.
The section first examines the evidence for remission and natural recovery, then explains why clinical and population samples tell different stories. It then sets out five accounts of what addiction is and the debate over one of them, the brain disease model. HSCI 402 develops treatment, recovery support and the evaluation of services; the aim here is to weigh what the evidence on recovery implies about the nature of addiction.
Learning Objectives
- 8.1 Describe the evidence from the study of Vietnam veterans and from population surveys that many people with substance use disorders eventually remit, often without treatment.
- 8.2 Explain maturing out and the clinician’s illusion, and why treatment samples and population samples give different pictures of addiction.
- 8.3 Compare moral, disease, choice, learning and biopsychosocial accounts of addiction and the evidence each draws on.
- 8.4 Evaluate the arguments for and against the brain disease model of addiction, including evidence on how biological explanations affect stigma.
- 8.5 Apply the accounts of addiction in this lesson to the way a Canadian law describes addiction.
8.1 Remission and Natural Recovery
Remission means that a person who once met the criteria for a substance use disorder no longer meets them. Natural recovery, also called unassisted or spontaneous recovery, means remission without formal treatment or mutual-help groups. A classic early study of natural recovery followed American soldiers returning from war, as Case 9.3 describes. A decade earlier, Winick (1962) had found that many people reported to the United States Federal Bureau of Narcotics as addicted to narcotics stopped appearing in its records in their thirties, and he called this process “maturing out of narcotic addiction”.
Between 1972 and 1974, the psychiatric epidemiologist Lee Robins and her colleagues evaluated United States Army enlisted men who had served in Vietnam in 1970 and 1971, and compared them with a matched group (Robins, 1993). As summarized by Hall and Weier (2017), 34 percent of the soldiers had used heroin in Vietnam and 20 percent had symptoms of heroin dependence while serving. In the first year after returning to the United States, only 1 percent became re-addicted to heroin, although 10 percent tried the drug again.
Most of the men who had used heroin in Vietnam had smoked or snorted it, and only a minority had injected it. After their return, most used heroin very occasionally or not at all. Predictors of injecting heroin after their return included opiate and other drug use before military service, behaviour that Robins classed as deviant before and during service, and associating with people who used drugs before service (Robins & Slobodyan, 2003). Robins (1993) noted that the low rates of re-addiction were “still not entirely incorporated into public and scientific views of heroin addiction.”
Hall and Weier (2017) read the Vietnam study as showing, in miniature, the factors beyond a drug’s pharmacology that shape addiction: price, availability, the way the drug is delivered, the availability of other substances, social norms, education and life circumstances. The study also shows the individual risk factors of Sections 1 and 2 at work, since the men who went on to inject heroin at home tended to have used drugs and shown behaviour that Robins classed as deviant before they enlisted.
8.1.1 Evidence from Population Surveys
Later surveys of whole populations found that many people remit, although estimates vary with who is studied and how. Lopez-Quintero and colleagues (2011) analyzed the 2001 to 2002 National Epidemiologic Survey on Alcohol and Related Conditions, a nationally representative survey of 43,093 adults in the United States, and estimated the lifetime probability of remission among people who had ever been dependent on nicotine, alcohol, cannabis or cocaine. Figure 9.6 shows their results.
The estimated lifetime probability of remission was 83.7 percent for nicotine, 90.6 percent for alcohol, 97.2 percent for cannabis and 99.2 percent for cocaine, and half of the cases had remitted about 26, 14, 6 and 5 years after the onset of dependence. Men, Black respondents and people with personality disorders or another substance use disorder had lower rates of remission for at least two substances (Lopez-Quintero et al., 2011). The survey could not explain the difference for Black respondents; Lessons 10 and 11 describe structural conditions, such as unequal access to care and unequal enforcement, that are candidate explanations. These are long-term probabilities: half of the people with alcohol dependence took about 14 years to remit, which is a long time in a person’s life.
Other estimates are lower. In a meta-analysis of 21 studies, Fleury and colleagues (2016) found that 35 to 54 percent of people with substance use disorders had achieved remission after a mean follow-up of 17 years, and concluded that for a substantial number of people these disorders are long-term rather than acute; studies with longer follow-up and lower retention of participants reported higher remission rates. The authors called for more longitudinal research.
8.1.2 With and Without Treatment
How many people recover without help? Two Canadian-led analyses give different answers for different substances. Sobell, Cunningham and Sobell (1996), at the Addiction Research Foundation in Toronto, analyzed two random telephone surveys of adults: Canada’s 1989 National Alcohol and Other Drugs Survey, with 11,634 respondents, and the 1993 Ontario Alcohol and Drug Opinion Survey, with 1,034 respondents (Dawson et al., 2005, identify the two surveys). In both, about 78 percent (77.5 and 77.7 percent) of people who had recovered from an alcohol problem for a year or more had done so without help or treatment, and 38 and 63 percent respectively reported drinking moderately after resolving their problem. A systematic review by Mellor, Lancaster and Ritter (2019) found that remission from alcohol problems without treatment is common, but that estimates range widely with how studies define alcohol problems, remission and treatment and how they draw their samples. Cunningham (2000), studying people who had recovered from dependence on illicit drugs such as cannabis, cocaine, hallucinogens, amphetamines or heroin, concluded that a substantial proportion had used addiction treatment services before their remission.
A national survey in the United States suggests that both routes are common. Kelly and colleagues (2017) found that 9.1 percent of American adults reported having resolved a significant alcohol or other drug problem. Of these, 53.9 percent had used some form of assistance, most often mutual-help groups (45.1 percent), followed by formal treatment (27.6 percent) and recovery support services (21.8 percent). People with a lifetime diagnosis, involvement with a drug court or opioid problems were more likely to have used assistance, and people with cannabis problems less likely. Heyman (2013), reviewing this kind of evidence, concluded that most people who meet the criteria for addiction to illegal drugs stop by about age 30, usually without professional help, and that the correlates of quitting include legal concerns, economic pressures and the desire for respect, particularly from family members.
8.2 Maturing Out and the Clinician’s Illusion
Heavy drinking and alcohol dependence rise through adolescence and the early twenties and then decline sharply in the late twenties. This decline is called maturing out, and it is usually explained by role incompatibility: new roles such as marriage, parenthood and full-time work conflict with a heavy-drinking lifestyle (Vergés et al., 2012). Using longitudinal data from the same national survey as Lopez-Quintero, Vergés and colleagues (2012) found that the picture is more complicated. The age pattern also reflects fewer new cases and fewer recurrences at older ages, persistence of dependence tends to increase with age, and the effects of life transitions differ between younger and older adults and between men and women. Maturing out, in their words, is “only a small piece of the puzzle.”
If many people remit, why do clinicians and many researchers describe addiction as chronic and relapsing? The box below explains a sampling effect that helps reconcile the two pictures.
The clinician’s illusion
Cohen and Cohen (1984) pointed out that clinicians see the people who currently have a disorder, a prevalence sample, while population studies try to capture everyone who has ever had it. In a prevalence sample, the chance that a person appears is proportional to how long the disorder lasts, so long-lasting cases are overrepresented. They called this bias the clinician’s illusion and named alcoholism and opiate addiction among the conditions whose long-term outlook is disputed partly for this reason.
Worked Example 9.3 (hypothetical). Suppose that 100 people develop a disorder: 80 have episodes lasting 2 years and 20 have episodes lasting 20 years. Over time, the short episodes contribute 80 × 2 = 160 person-years of illness and the long episodes contribute 20 × 20 = 400 person-years. If everyone with the disorder attended a clinic and new cases arose at a steady rate, then on a typical day long-lasting cases would make up 400 ÷ (160 + 400) = 400 ÷ 560, or about 71 percent, of the clinic’s patients, although they make up only 20 percent of everyone who ever had the disorder.
Treatment samples also differ in severity. Regier and colleagues (1990), as Section 2 noted, found more co-occurring disorders among people in treatment settings, and Lopez-Quintero and colleagues (2011) found that co-occurring disorders lowered the chance of remission. Estimates therefore differ with the people studied, and both pictures are accurate for the populations they describe: addiction is often chronic among people who reach treatment, and often time-limited across the whole population. Any account of what addiction is must explain both, and the next part of the section compares five accounts.
8.3 Five Accounts of Addiction
An account of addiction is a general explanation of what addiction is and why it persists. Each account emphasizes different evidence and implies different responses. The tabs below introduce five accounts that recur in research and policy debates, and Table 9.4 compares them.
Lesson 2 described the moral model of the temperance era: excessive drinking as a freely chosen moral failing, to be met with warnings, religious correction and punishment. The model survives in everyday language about willpower and in policies that treat drug use mainly as wrongdoing. It does not account for the evidence in Sections 1 to 3 that vulnerability is heritable, shaped by development and adversity, and linked to measurable brain changes.
Alan Leshner (1997), then director of the United States National Institute on Drug Abuse, argued in Science that “drug addiction is a chronic, relapsing disease that results from the prolonged effects of drugs on the brain,” and that behavioural and social context are important parts of the disorder itself. McLellan and colleagues (2000) argued that heritability, personal choice and environmental factors are comparably involved in addiction and in diabetes, hypertension and asthma, and that addiction should be insured, treated and evaluated like other chronic illnesses. Volkow, Koob and McLellan (2016) concluded that neuroscience continues to support the brain disease model.
Gene Heyman (2013) argued that all behaviour, including voluntary action, has a biological basis, so the question is whether drug use in addiction is compulsive. Because most people stop by about age 30, usually without professional help and in response to legal, financial and family pressures, he concluded that “the correlates of quitting are the correlates of choice.” He described addiction as a disorder that follows established principles of choice, such as hyperbolic discounting, the tendency to prefer smaller immediate rewards over larger delayed ones.
Lesson 8 showed that drugs act on the brain’s systems for reward and learning. The neuroscientist Marc Lewis, writing from the University of Toronto, set out a learning account in Neuroethics (Lewis, 2017) and in a 2018 review in the New England Journal of Medicine titled “Brain change in addiction as learning, not disease”. In the 2017 article he argued that the brain changes found in addiction are similar to those generally observed when recurrent, highly motivated goal seeking produces deep habits, Pavlovian (cue) learning and reduced engagement of the prefrontal cortex. On this account, the brain changes described in Lesson 8 are real, addiction is learning that has been accelerated and entrenched, and recovery is a further stage of learning and development in which the brain reorganizes again (Lewis, 2017).
The psychiatrist George Engel (1977) argued in Science that the dominant biomedical model left no room for the social, psychological and behavioural dimensions of illness, and proposed a biopsychosocial model as a blueprint for research, teaching and care. Gage and Sumnall (2019) described biopsychosocial models as the ones that predominated in addiction research when they wrote. They combine genes and brains (Section 1 and Lesson 8), development and personality (Section 2), and environment (Section 3).
Table 9.4. Five accounts of addiction differ in their core claim, the evidence they emphasize, the responses they imply and the criticism most often made of them.
| Account | Core claim | Evidence emphasized | Response implied | Question it must answer |
|---|---|---|---|---|
| Moral | Excessive use is a freely chosen failing | Everyday experience of choosing | Correction and punishment | Why is vulnerability heritable and linked to adversity? |
| Disease | Addiction is a chronic, relapsing brain disease | Neuroimaging, animal studies, relapse in treatment samples | Long-term medical care | Why do many people remit, often without treatment? |
| Choice | Use responds to costs, rewards and alternatives | Natural recovery and its correlates | Change incentives and alternatives | Why do some people keep using despite severe costs? |
| Learning | Brain changes reflect intense, repeated learning | Reward learning and brain change (Lesson 8) | Support new learning and goals | What separates addiction from other deeply learned habits? |
| Biopsychosocial | Biological, psychological and social causes interact | Twin, developmental and environmental studies together | Responses at several levels | How do the levels combine in a particular person? |
The questions in the last column are ones each account must answer to be convincing, and 8.4 The Brain Disease Model Debate shows how the two sides of the main debate answer some of them. Use Interactive 9.2 to practise recognizing each account in statements drawn from the sources in this section.
Which account is speaking?
8.4 The Brain Disease Model Debate
The brain disease model has been prominent in research and public messaging since the late 1990s, and it has attracted sustained criticism. Both sides include leading addiction scientists, and Canadian researchers have contributed on both sides.
8.4.1 The Case Against
Hall, Carter and Forlini (2015) noted that the United States National Institute on Drug Abuse had advocated the brain disease model since 1997. They argued that the model is not supported by animal and neuroimaging evidence to the extent its advocates suggest, that it has not delivered more effective treatments, and that its effect on public policy has been modest. Its focus on disordered brain function in a minority of severely affected people, they argued, undermines the implementation of cost-effective population policies to discourage smoking and heavy drinking, and pursuing high-technology brain treatments is questionable when most people with addiction lack access to effective psychosocial and drug treatments. Heyman (2013) and Lewis (2017, 2018) add the arguments from natural recovery and from learning described above.
8.4.2 The Case For
Heilig and colleagues (2021), whose authors include researchers in Hamilton and at the Centre for Addiction and Mental Health in Toronto, listed the main criticisms: that the model is deterministic, fails to account for variation in remission and recovery, places too much emphasis on compulsion and has not identified a specific neural signature of addiction. They acknowledged that some of these criticisms have merit. They argued that its foundational premise, that addiction has a neurobiological basis, is sound; that apparently spontaneous remission does not negate it; and that compulsive behaviour can coexist with sensitivity to alternative rewards. They also argued that denying that addiction is a brain disease is harmful, because it contributes to reduced access to health care and treatment. They described the brain as the biological basis of both addiction and the capacity for change, and called for multidisciplinary research that integrates neuroscientific, behavioural, clinical and sociocultural perspectives.
8.4.3 Does a Disease Label Reduce Stigma?
Supporters of the model have argued that it may reduce the moral judgement attached to addiction (Volkow et al., 2016). The evidence on this point is mixed. A meta-analysis of 28 experiments on psychological problems in general found that biogenetic explanations reduced blame (Hedges’ g = −0.32) but increased pessimism about recovery (g = 0.26) and, less certainly, perceptions of dangerousness (g = 0.20), and did not typically change the desire for social distance (Kvaale, Haslam & Gottdiener, 2013). In the United States, the share of the public endorsing treatment by a psychiatrist for alcohol dependence rose from 61 percent in 1996 to 79 percent in 2006, as neurobiological explanations became more popular, but social distance and perceived danger did not decrease (Pescosolido et al., 2010). In that study, holding a neurobiological view increased support for treatment but was generally unrelated to stigma, and where it was related, it increased community rejection. Lesson 1 introduced the evidence on stigmatizing language, and HSCI 214 develops stigma theory.
Where the two sides agree
Heilig and colleagues (2021) describe the criticism of the brain disease model as “acerbic”, yet the two sides agree on several points. Both sides accept that addiction has a biological basis: Heyman (2013) writes that addiction “must” have a biology, and Heilig and colleagues (2021) build their defence on that premise. Both accept that remission is common and that environments matter. Hall and colleagues (2015) and Heilig and colleagues (2021) both worry about access to effective treatment. The disagreement concerns whether “disease” is the most accurate and useful description, and which responses follow from it.
8.5 Applying this Lesson
Accounts of addiction become concrete when governments write them into law. The task card below applies the accounts in 8.3 Five Accounts of Addiction to the way a 2025 Canadian law describes addiction.
Alberta introduced the Compassionate Intervention Act as Bill 53 in April 2025. It received royal assent on 15 May 2025, with certain sections taking effect at assent and the rest to take effect on proclamation (Government of Alberta, 2025). The Act allows family members, health care professionals, police and others to apply for an order requiring treatment for an adult whose substance use makes the person likely to cause substantial harm to themselves or others. As of May 2026, the government’s page on the Act stated that the first secure treatment beds would begin operating in 2027. The same page introduces the Act with this sentence: “Addiction is an illness that if left untreated will worsen and can eventually lead to a person losing their ability to make good decisions for their health and wellbeing” (Government of Alberta, 2026).
Write a short analysis (about 250 words) of that sentence that (1) identifies which of the five accounts in 8.3 Five Accounts of Addiction it draws on most clearly and which claims of the brain disease model it echoes, (2) explains how the choice account (Heyman, 2013) and the learning account (Lewis, 2017) would each describe the same person’s addiction, and (3) uses two findings from this lesson that are consistent with the sentence and two that complicate it, stating whether each comes from a treatment sample or a population sample (8.2 Maturing Out and the Clinician’s Illusion). Involuntary treatment itself, including its legal and ethical questions and the evidence on whether it improves outcomes, belongs to HSCI 402, which develops treatment and its evaluation.
Summary
Many people who develop a substance use disorder eventually remit, often without treatment, although estimates vary with who is studied and how: Vietnam veterans rarely returned to heroin addiction at home, one American survey projected lifetime remission from 84 percent for nicotine to 99 percent for cocaine, and a meta-analysis found 35 to 54 percent in remission after a mean follow-up of 17 years. Unassisted recovery is common for alcohol problems, while people with more severe problems and with opioid problems more often use assistance, and co-occurring disorders slow remission. The clinician’s illusion helps explain why treatment settings see a chronic, relapsing condition while populations show frequent recovery. Moral, disease, choice, learning and biopsychosocial accounts emphasize different parts of this evidence. Both sides of the brain disease debate accept that addiction has a biology; they disagree about whether the disease label is accurate and what it implies for stigma and policy. With the reflection and knowledge check below, you can test your grasp of this section before the final assessment draws the lesson together.
Key Takeaways
- 8.1 Remission means no longer meeting the criteria for a disorder, and natural recovery means remission without treatment or mutual-help groups. Of soldiers who served in Vietnam, 34 percent used heroin and 20 percent had symptoms of dependence, yet only 1 percent became re-addicted in their first year home (Robins, 1993; Hall & Weier, 2017). In a national United States survey, the projected lifetime probability of remission was 83.7 percent for nicotine, 90.6 percent for alcohol, 97.2 percent for cannabis and 99.2 percent for cocaine (Lopez-Quintero et al., 2011), and about 78 percent of Canadians who had recovered from an alcohol problem did so without help (Sobell et al., 1996), although people with more severe or opioid problems more often used assistance (Kelly et al., 2017). A meta-analysis found lower figures, 35 to 54 percent in remission after a mean follow-up of 17 years (Fleury et al., 2016).
- 8.2 Maturing out describes the decline in heavy drinking and dependence in the late twenties, usually explained by role incompatibility, although Vergés and colleagues (2012) found it to be only a small piece of the puzzle. Clinicians see a prevalence sample in which long-lasting cases are overrepresented, the clinician’s illusion (Cohen & Cohen, 1984): in the worked example, cases that are 20 percent of everyone affected make up about 71 percent of a clinic’s patients. Treatment samples are also more severe, so addiction is often chronic among people who reach treatment and often time-limited across the whole population.
- 8.3 The moral account treats excessive use as a freely chosen failing; the disease account treats addiction as a chronic, relapsing brain disease (Leshner, 1997); the choice account argues that the correlates of quitting are the correlates of choice (Heyman, 2013); the learning account treats brain change as accelerated and entrenched learning (Lewis, 2017); and the biopsychosocial account combines biological, psychological and social causes (Engel, 1977). Each emphasizes different evidence, implies different responses and faces a question it must answer to be convincing.
- 8.4 Critics argue that the brain disease model outruns its evidence, has not delivered better treatments and can undermine population policies (Hall et al., 2015), while defenders argue that its neurobiological premise is sound and that remission does not negate it (Heilig et al., 2021). Biogenetic explanations reduce blame but increase pessimism about recovery (Kvaale et al., 2013), and growing neurobiological views raised support for treatment without reducing social distance (Pescosolido et al., 2010). Both sides accept that addiction has a biology; they disagree about whether the disease label is accurate and useful.
- 8.5 Alberta’s Compassionate Intervention Act, which received royal assent on 15 May 2025, is introduced with a sentence describing addiction as an illness that worsens if untreated and can erode a person’s ability to make decisions, which draws most clearly on the disease account. The choice and learning accounts would describe the same person differently, and findings from population samples on natural recovery complicate the sentence while findings from treatment samples fit it more closely. Asking which kind of sample a finding comes from is central to judging how a law or policy frames addiction.
Reflection
An addiction physician says, “In my experience, addiction is a chronic, relapsing illness.” A researcher replies, “Population surveys show that many people recover, often without treatment.” Using the following evidence from Section 4, write 150 to 250 words explaining how both can be right and what this implies for the debate over the brain disease model: (1) Lopez-Quintero et al. (2011) estimated lifetime remission of 83.7 percent for nicotine, 90.6 percent for alcohol, 97.2 percent for cannabis and 99.2 percent for cocaine dependence, with half remitting about 26, 14, 6 and 5 years after onset; (2) in Worked Example 9.3, 80 people with 2-year episodes and 20 people with 20-year episodes produce a clinic population in which about 71 percent have the long-lasting form; (3) co-occurring disorders were more common in treatment settings and lowered the chance of remission; (4) Heilig et al. (2021) argued that apparently spontaneous remission does not negate the brain disease model.
Minimum 20 characters required.
Question 1: What did Lopez-Quintero et al. (2011) estimate about remission from dependence in the United States?
Question 2: What is the clinician’s illusion described by Cohen and Cohen (1984)?
Question 3: Which argument did Heyman (2013) make for a choice account of addiction?
Question 4: What did the meta-analysis by Kvaale, Haslam and Gottdiener (2013) find about biogenetic explanations of psychological problems?
Final Assessment
Bringing It All Together
This lesson asked why only a minority of people who use substances develop addiction. Section 1 showed that substance use disorders run strongly in families and that a meta-analysis of twin and adoption studies estimated that about half of the variation in alcohol use disorder is linked to genes, with heritability for other addictions from about 0.39 to 0.72. Variants of ADH1B and ALDH2, which a 2018 review describes as by far the strongest genetic contributors to alcohol use disorder, protect by making alcohol unpleasant, and the weakening of that protection in Japan as drinking increased showed that genes act through behaviour and setting. Polygenic scores confirm that liability is spread across many genes; they explain a small share of the variance, and as of 2026 whether they improve care is still being studied.
Sections 2 and 3 followed vulnerability through development and environment. In the dual systems model, reward sensitivity rises in adolescence before cognitive control matures; early drinking is strongly associated with later dependence, although twin studies trace much of the association to familial liability and its causal role remains uncertain; and lower childhood self-control predicted adult dependence along a gradient in one cohort. Childhood adversity shows a dose-response relationship with later drug problems, social rank changed the dopamine system and cocaine reinforcement in two small monkey studies, and Rat Park at Simon Fraser University, read alongside its critics and replications, supports a narrower claim than its popular version.
Section 4 showed that many people remit, often without treatment, although estimates vary with the people studied and the method, while treatment settings see a chronic, relapsing condition because of the clinician’s illusion and greater severity. Moral, disease, choice, learning and biopsychosocial accounts draw on different parts of this evidence, and the brain disease debate concerns what the disease label implies more than whether addiction has a biology.
Key Takeaways from this lesson
- A meta-analysis of twin and adoption studies estimated that about half of the variation in alcohol use disorder is linked to genetic differences, with a modest contribution from the shared family environment. Heritability describes variation across a population at a particular time and place; it does not divide one person’s risk into genetic and environmental parts.
- The ADH1B*2 and ALDH2*2 variants protect against alcohol use disorder by raising acetaldehyde, and their protective effect depends on drinking norms. Polygenic scores explained about 2.5 to 3.5 percent of the variance in alcohol use disorder in a 2020 study and work less well outside European ancestry; a 2026 review found them comparable to scores introduced into care for other conditions, but whether they improve care is still being studied.
- The dual systems model describes reward sensitivity peaking in adolescence before cognitive control matures, which makes mid-adolescence a period of heightened vulnerability; reviews find the average trends broadly supported and still debate the model. In one survey, early first drinking was associated with about four times the risk of later dependence, a risk difference of about 30 percentage points.
- Twin studies trace much of the association between early drinking and later dependence to familial liability, so whether early drinking causes dependence remains uncertain.
- Childhood self-control, personality traits such as sensation seeking and anxiety sensitivity, and co-occurring mental illness all mark different routes to substance use disorder.
- Childhood adversity shows a dose-response relationship with later drug problems, and social rank changed the dopamine system and the reinforcing effect of cocaine in two small monkey studies, so environments can shape vulnerability.
- Rat Park supports the idea that social environments matter, while its small, unequally measured groups and mixed replications limit what it can show, and no animal model reproduces the human environments that shape addiction.
- Many people with lifetime dependence eventually remit, often without treatment, although estimates vary with the sample and method, while treatment settings overrepresent long-lasting cases. Moral, disease, choice, learning and biopsychosocial accounts of addiction emphasize different evidence, and both sides of the brain disease debate accept that addiction has a biological basis.
Core Concepts Reviewed
Section 1: family, twin and adoption designs, heritability and Falconer’s formula, the shared and unique environment, protective ADH1B and ALDH2 variants, gene-environment interplay and polygenic scores.
Section 2: the dual systems model of adolescent development, age at first use, relative risk and risk difference, confounding and common liability, childhood self-control, personality risk profiles and co-occurring disorders.
Section 3: adverse childhood experiences and dose-response relationships, social rank and dopamine D2 receptors in monkeys, Rat Park and its criticisms, replication and social choice experiments.
Section 4: remission and natural recovery, maturing out, the clinician’s illusion, and the moral, disease, choice, learning and biopsychosocial accounts of addiction.
Before the final knowledge assessment, bring the four sections together by answering the lesson’s central question in your own words.
Reflection
Write 200 to 250 words answering the lesson’s question, “Why do some people develop addiction and others do not?” Use at least one finding from each section: Section 1 (for example, heritability of alcohol use disorder of about 0.49 and the protective ALDH2*2 variant whose effect weakened in Japan as drinking increased); Section 2 (for example, a relative risk of about 4 for an early first drink, much of which twin studies trace to familial liability, or the gradient between childhood self-control and adult dependence); Section 3 (for example, a 7- to 10-fold higher risk of drug problems with five or more adverse childhood experiences, or social rank changing dopamine receptors in monkeys); and Section 4 (for example, projected lifetime remission of 84 to 99 percent in one American survey, or 35 to 54 percent remission after a mean of 17 years in a meta-analysis). End by stating which account of addiction (moral, disease, choice, learning or biopsychosocial) best fits the evidence you used and one limitation of that account.
Minimum 30 characters required.
Final Knowledge Assessment
Question 1: Which pair of findings best shows that both genes and the rearing environment contribute to alcohol use disorder?
Question 2: Among Japanese people diagnosed with alcoholism, the share carrying the protective ALDH2*2 variant rose from 2.5 percent in 1979 to 13 percent in 1992. What does this change show?
Question 3: A hypothetical twin study finds correlations of 0.60 in identical pairs and 0.45 in fraternal pairs. What does Falconer’s formula imply?
Question 4: A study finds that people who first used cannabis early are more likely to develop cannabis use disorder. Which design would best test whether common liability explains the association?
Question 5: Why did Moffitt et al. (2011) also study 500 sibling pairs in addition to their main birth cohort?
Question 6: Rush et al. (2008) found a 12-month prevalence of co-occurring substance use and other mental disorders of 1.7 percent in Canada in 2002. Why might this rate sit at the lower end of the international range?
Question 7: What do the monkey studies of social rank add to the evidence from surveys of adverse childhood experiences?
Question 8: Which statement about the Rat Park experiments at Simon Fraser University is accurate?
Question 9: How did the design of Venniro et al. (2018) differ from the original Rat Park studies?
Question 10: What did Lee Robins’s study of Army enlisted men who served in Vietnam in 1970 and 1971 show, as summarized by Hall and Weier (2017)?
Question 11: Which conclusion about treatment and recovery best fits the evidence in Section 4?
Question 12: Suppose that 90 people have a disorder for 1 year and 10 people have it for 30 years (hypothetical figures). If all attend a clinic and new cases arise at a steady rate, about what share of the clinic’s patients on a typical day have the long-lasting form?
Question 13: Which criticism did Hall, Carter and Forlini (2015) make of the brain disease model of addiction?
Question 14: How did Heilig et al. (2021) respond to the argument that frequent natural recovery contradicts the brain disease model?
Question 15: Alberta’s government describes addiction as “an illness that if left untreated will worsen.” Which finding from this lesson most directly complicates this description as a general claim?
Glossary: Key Terms, People & Frameworks
📚 Reference page, available throughout the lesson
Terms, frameworks and researchers introduced in this lesson, in the order of the sections where they first appear.