# Lesson 9: Why Some People Develop Addiction and Others Do Not

*Companion-podcast transcript • Sarah & Kiffer*  
*Office Hours episode to listen to after working through the lesson*

---

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. Heritability is often heard as a statement about one person's DNA, and Lesson nine, on why some people develop addiction and others don't, gives good reasons to be careful with that reading.

**Sarah:** Question one. What does a heritability of about fifty percent tell us about addiction, and what can change it?

**Kiffer:** Start with what kind of number it is. Heritability is a fraction. The top is the variation in liability that's linked to genetic differences, where liability means a person's underlying tendency toward the disorder. The bottom is all of the variation: genetic, shared family environment, and each person's unique experiences. Change either part and the answer changes, and the environment sits in the bottom part.

**Sarah:** Let's put numbers on it. Picture all of the variation in liability to alcohol use disorder in a hypothetical population as one hundred units. Genetic differences account for fifty units, the shared family environment for ten, and the unique environment for forty. That's close to the 2015 meta-analysis estimate from Verhulst, Neale and Kendler. The heritability is fifty out of one hundred, or 0.50.

**Kiffer:** Now suppose, hypothetically, that a society evens out the differences between households, so the shared family part falls to zero, and evens out enough differences between individual lives that the unique part falls from forty units to twenty-five. The genetic differences stay at fifty units. Here's the question for everyone listening. What's the heritability now? Pause the audio if you'd like more time. Take a few seconds.

*(Pause)*

**Sarah:** The total is now fifty plus zero plus twenty-five, which is seventy-five units. Genes still account for fifty of them, so the heritability is fifty divided by seventy-five, or about 0.67. It went from one half to two thirds.

**Kiffer:** And nobody's genes changed. The example also shows what the fraction can't tell you: whether more people or fewer developed the disorder after the change. Heritability describes how differences are spread among people and says nothing about the average level.

**Sarah:** And the Japanese data show that a gene's effect can depend on the setting. The protective ALDH2 variant itself didn't change between 1979 and 1992, yet among Japanese people diagnosed with alcoholism, the share carrying it rose from 2.5 percent to thirteen percent, most likely because drinking became more common.

**Kiffer:** Right, that's the gene-environment interplay from the lesson. Our arithmetic assumed that genes have the same effect in every setting, and twin studies of drinking suggest they may not. In a study of more than six thousand Australian twins published in 2009, genetic influences on alcohol dependence symptoms were considerably larger among people who first drank before age thirteen. Among people who started later, most of the variation fell in the unique environment.

**Sarah:** So the same kind of genetic risk showed up more in one group than another.

**Kiffer:** That's how the authors read it: early drinking may let genetic vulnerability be expressed. Finnish twin studies published in 2001 found a community version. Genetic influences on drinking at ages sixteen to eighteen were larger in cities and in communities where many people moved in and out, and the shared family environment mattered more in rural communities. The researchers suggested that stable communities, with more monitoring of young people, leave less room for genetic tendencies to show.

**Sarah:** And did that hold up?

**Kiffer:** Partly. A 2009 study of younger Finnish twins found a similar pattern for behaviour problems in twelve-year-old girls, and none for drinking at fourteen. So I'd call the community finding promising and unsettled. Either way, heritability describes one population in one setting, and changing the setting can push it up or down.

**Sarah:** Question two. People who first drink at fourteen or younger are about four times as likely to report later dependence as people who start at twenty or older. Is early drinking a cause, or a marker of risk that was already there?

**Kiffer:** The way to test that is to compare people who share most of their family liability and differ in when they started. Identical twins are the strongest version of that comparison.

**Sarah:** I want to try one myself, with made-up numbers. A hypothetical study finds one hundred pairs of identical twins in which one twin first drank at fourteen or younger and the co-twin, meaning the other twin in the pair, first drank at twenty or older. In thirty pairs both twins later developed alcohol dependence. In fifty-two pairs neither did. In eleven pairs only the early-drinking twin did, and in seven pairs only the later-drinking twin did. What's the relative risk for early drinking?

**Kiffer:** Before you start, let's put the same question to everyone listening. What's the relative risk for early drinking? Take a few seconds.

*(Pause)*

**Sarah:** Okay. Among the early-drinking twins, the ones with dependence are the thirty pairs where both had it plus the eleven where only the early twin had it. That's forty-one out of one hundred, or forty-one percent. The survey found about ten percent among people who started at twenty or older. Forty-one divided by ten is about four. So the twins back up the survey.

**Kiffer:** Which group did you divide by, and is that the comparison the design was built for?

**Sarah:** The survey's later starters. Oh. That throws away the whole design. The comparison group is the co-twins in these same pairs. Among the later-drinking twins, the ones with dependence are the thirty pairs where both had it plus the seven where only the later twin had it. That's thirty-seven percent.

**Kiffer:** So what's the relative risk inside the pairs?

**Sarah:** Forty-one divided by thirty-seven, which is about 1.1. The risk difference is four percentage points. So almost all of the fourfold gap disappears when the comparison is someone with the same genes and the same upbringing.

**Kiffer:** And look at that thirty-seven percent. In the survey, starting late went with a risk of about ten percent, yet more than a third of these late-starting twins developed dependence. That's common liability in numbers: much of the risk travels with the family. The numbers are invented, but the pattern resembles what Prescott and Kendler reported in 1999 from eight thousand seven hundred forty-six twins, where twin-pair analyses attributed all of the association to familial sources.

**Sarah:** And the small leftover in my example looks more like the Minnesota twin study published in 2015, where early-drinking twins had higher risk than their co-twins for several, though not all, adult outcomes.

**Kiffer:** Yes. And a co-twin design has a blind spot. It removes everything the twins share, and it can't remove something that happened to only one of them. If one twin fell in with older friends at thirteen, that could lead to early drinking and, separately, to later problems. So a difference between co-twins is evidence for a causal effect, though weaker than a randomized trial, which nobody could ethically run.

**Sarah:** The self-control research uses a similar comparison. In the Dunedin cohort in New Zealand, followed to age thirty-two, lower childhood self-control predicted adult substance dependence, and in the same 2011 report Moffitt and colleagues compared siblings in a second cohort, five hundred British pairs of non-identical twins.

**Kiffer:** They did, and it helps to be exact about what that covered. Those siblings were followed to age twelve. The sibling with lower self-control at age five was more likely to have started smoking, to do poorly in school and to behave antisocially. Dependence couldn't be measured that young. So the sibling comparison addresses part of the family confounding for those early outcomes, and the link to adult dependence rests on the Dunedin cohort, which compared children from different families after accounting for intelligence and social class.

**Sarah:** So what should we make of the four-times figure?

**Kiffer:** I'd call it a real association and a poor guide to what delaying the first drink would achieve on its own. Public health guidance can still recommend delaying use for other reasons, and HSCI 402 develops how prevention programs are tested. A claim that delaying the first drink would cut early starters' dependence by three quarters assumes the whole association is causal, and the twin evidence traces much or all of it to family liability.

**Sarah:** Question three. Rat Park was run here at Simon Fraser University in the late 1970s, and it's one of the most famous animal studies in addiction research. How much can a study like that carry?

**Kiffer:** Start with what it found. Rats living together in an enclosure of 8.8 square metres drank less morphine solution than rats housed alone in small cages, under particular conditions. The colony rats did drink some morphine. The popular version, in which social rats refused the drug, goes further than the data.

**Sarah:** And the design problems are about how much weight that difference can bear. The first study ended with two isolated females, six isolated males, seven colony females and twelve colony males. With two animals in a group, one unusual rat can move the group's average a long way.

**Kiffer:** There's a second problem, the number of comparisons. Morphine drinking was measured in three ways, across eight choice days and four groups, which leaves room for many comparisons. Researchers usually call a difference statistically significant when a difference that large would turn up by chance less than one time in twenty if housing made no difference. A difference that passes that test by chance alone is a false alarm.

**Sarah:** So here's a question for everyone. Suppose a hypothetical study makes twenty separate comparisons between isolated and colony rats, and housing truly makes no difference at all. About how many false alarms would you expect? Take a few seconds.

*(Pause)*

**Kiffer:** About one. Twenty comparisons, each with a one-in-twenty chance of a false alarm, gives twenty times one twentieth, which is one. If the comparisons were independent, meaning one result told you nothing about another, the chance of at least one false alarm would be nearly two in three. Each comparison stays quiet with a chance of 0.95, and 0.95 multiplied by itself twenty times is about 0.36, so the chance that at least one speaks up is about 0.64.

**Sarah:** So one striking difference, found on one day with one measure, could easily be a false alarm.

**Kiffer:** It could. Comparisons made on the same rats aren't independent, so the exact figure would differ, but the direction holds. That's why a stronger design names its main comparison in advance. And the original authors were careful in their conclusions. In 1978 they wrote that sex and housing are important variables, and that theories ignoring them run the risk of oversimplification.

**Sarah:** Then there's the replication record, which is mixed. In 1989, Bozarth, Murray and Wise, at Concordia University in Montréal, found that isolated and group-housed rats both quickly learned to self-administer heroin and cocaine by injection. Isolated rats learned to take heroin earlier, but by the fifth week the groups took similar amounts, and cocaine intake didn't differ.

**Kiffer:** They concluded that isolation isn't necessary for these drugs to be reinforcing. In 1996, Petrie tried twice to repeat the 1979 study and couldn't, because the isolated rats drank much less morphine solution than in the original. Petrie suggested the supplier's rat strain might have changed.

**Sarah:** Which is itself a point about genes, in rats.

**Kiffer:** It is. The strongest recent support comes from a different design. In 2018, Venniro and colleagues at the United States National Institute on Drug Abuse let rats choose between a drug and a chance to interact with another rat. Social interaction prevented self-administration of methamphetamine and of heroin across both sexes, different doses and different housing conditions. The protection weakened when the social reward was delayed or punished.

**Sarah:** So Rat Park supports a narrower claim: social settings can change how much drug animals take, and isolation is one influence among several. Is it fair to keep telling the popular version because its message is a good one?

**Kiffer:** I don't think so, and Gage and Sumnall, who reread the studies in 2019, make a similar point. They judge the core message, that addiction is more complicated than a biological response to a drug, to be important, and they still urge people to read the study itself.

**Sarah:** Question four. If most people with a substance use disorder eventually recover, why does addiction look so chronic in clinics?

**Kiffer:** Let's start with the best-known study of natural recovery. Lee Robins and her colleagues evaluated United States Army enlisted men who served in Vietnam in 1970 and 1971. Thirty-four percent used heroin there, and twenty percent had symptoms of heroin dependence. In the first year back home, only one percent became addicted again, although ten percent tried heroin again.

**Sarah:** That's a huge drop. But coming home changed almost everything around them at once.

**Kiffer:** Right, which is why Hall and Weier, writing in 2017, read it as showing many of the influences on addiction beyond the drug's pharmacology: price, availability, how the drug is taken, social norms, education and life circumstances. It's also why the study says less about someone whose surroundings stay the same. The men who went on to inject at home tended to have used drugs and to have had behaviour problems before their military service.

**Sarah:** And population surveys show the pattern well beyond Vietnam. In an American national survey from 2001 and 2002, analyzed by Lopez-Quintero and colleagues, lifetime remission was estimated at about ninety-one percent for alcohol dependence. In Canada, Sobell, Cunningham and Sobell analyzed two telephone surveys of adults, a national survey from 1989 and an Ontario survey from 1993. In both, about seventy-eight percent of people who'd recovered from an alcohol problem for a year or more had done it without help or treatment.

**Kiffer:** With two cautions. Half of the people with alcohol dependence in that American survey took about fourteen years to remit, which is a long stretch of a life. And for illicit drug dependence, an analysis by Cunningham, in Toronto, published in 2000, found that a substantial share of people had used treatment before they recovered.

**Sarah:** Now the clinic. The lesson's example had eighty people with two-year episodes and twenty with twenty-year episodes, and the long cases made up about seventy-one percent of a clinic's patients on a typical day. But that assumed everyone with the disorder went to the clinic.

**Kiffer:** So let's change that assumption, hypothetically. Suppose one in ten of the people with short episodes comes to the clinic and stays for the whole episode, and one in two of the people with long episodes does, on the assumption that longer, harder episodes push people toward care. As in the lesson, a person-year is one person ill for one year. What share of the clinic's patients on a typical day have the long-lasting form? Take a few seconds.

*(Pause)*

**Sarah:** Okay. Short episodes: one in ten of eighty is eight people, each in the clinic for two years, so sixteen person-years. Long episodes: one in two of twenty is ten people, each there for twenty years, so two hundred person-years. Two hundred divided by two hundred sixteen is about 0.93. So about ninety-three percent of the clinic's patients have the long form.

**Kiffer:** That's compared with twenty percent of everyone who ever had the disorder. The duration bias that Cohen and Cohen named the clinician's illusion in 1984 and the question of who seeks care stack on top of each other. A clinician there would see that almost every patient has been ill for years and would reasonably conclude that the condition is chronic.

**Sarah:** And they'd be right about their own patients, which is what makes it an illusion and an accurate picture at the same time. Okay. Now the part where we disagree, which is the brain disease model.

**Kiffer:** You're the one defending it, so go ahead and make the case.

**Sarah:** I'd defend the label more than you do. People who reach treatment are often the people for whom chronic and relapsing is accurate. In 2000, McLellan and colleagues found relapse and adherence to treatment in drug dependence similar to type 2 diabetes, high blood pressure and asthma. The label also ties addiction to health care. A 2021 review by an international group, including researchers in Hamilton and at the Centre for Addiction and Mental Health in Toronto, argued that denying addiction is a brain disease contributes to reduced access to care, and that remission doesn't negate the model, because the brain is also where the capacity for change comes from.

**Kiffer:** I agree with the biology. Both sides accept that addiction has a biological basis, and even Heyman, who argues for a choice account, writes that it must. Where I part ways is the general claim that addiction is chronic and relapsing. The population surveys don't support it as a description of everyone who meets the criteria. I'd describe addiction as a condition with a real biology that most people recover from, often slowly, and I'd keep chronic and relapsing for the severe, long-lasting cases that clinics see most.

**Sarah:** But a disease label has real effects. In American surveys, the share of the public endorsing treatment by a psychiatrist for alcohol dependence rose from sixty-one percent in 1996 to seventy-nine percent in 2006, as neurobiological explanations became more popular.

**Kiffer:** That's the strongest point for the label. The same surveys found that perceived danger and social distance, meaning how much people want to keep away from someone, didn't fall, and where a neurobiological view was linked to stigma at all, it increased community rejection. And a 2013 meta-analysis of twenty-eight experiments on psychological problems in general found that biological explanations reduced blame and increased pessimism about recovery. For a condition most people recover from, that pessimism is a real cost.

**Sarah:** Pessimism measured in a short experiment is a long way from someone being turned away from care, though. I'd still want a person with a severe disorder to walk into a clinic that treats addiction like any other illness.

**Kiffer:** That's fair, and it's a trade-off the evidence we have can't settle. The wording matters in law too. Alberta's Compassionate Intervention Act received royal assent in May 2025, and as of May 2026 the government's page on it opened with a sentence that begins, "Addiction is an illness that if left untreated will worsen." That fits many people in treatment. As a general claim, the population surveys complicate it. Whether the law's approach helps people is a separate question, and HSCI 402 takes up involuntary treatment and its evidence.

**Sarah:** So where does that leave you?

**Kiffer:** Where I land is that the disease label fits the severe cases better than it fits the many shorter episodes in the population, and I'd want any description of addiction to say which group it means.

**Sarah:** I'm more comfortable with the label than Kiffer is, but we agree on the practical rule. When someone tells you addiction is chronic, or that most people grow out of it, ask which group the evidence describes: people in treatment, or everyone who has ever met the criteria.

**Kiffer:** Let's pull it together with three things to take away.

**Sarah:** First, heritability is a share of variation in one population at one time. Change the setting and the number can rise or fall while nobody's genes change, and it never splits one person's risk into parts.

**Kiffer:** Second, an association like early drinking and later dependence needs a comparison that holds family liability fixed. Co-twin and sibling designs do that, and they still leave room for causes that one twin experienced and the other didn't.

**Sarah:** And third, the picture of addiction depends on where you look. Clinics see long, severe cases, and populations show that most people recover, often without treatment and often slowly. Before you repeat the popular version of a famous study like Rat Park or the Vietnam research, read what it found.

**Kiffer:** If you'd like more practice, rework the clinic problem from this episode with one in five of the people with short episodes coming to the clinic, and see how far the clinic's picture still sits from the population's.

**Sarah:** Next time, it's Lesson ten, Why People Use Substances and Why Risks Differ, where motives for use, the social ecological model and the social determinants of harm come in.

**Kiffer:** Take care, everyone.

**Sarah:** See you in Lesson ten.
