‘Your answers on this questionnaire suggest moderate to severe depression,’ I said gently. ‘Can I ask how long that’s been going on?’

My client, a middle-aged, widowed military veteran with chronic back pain, looked taken aback. ‘Depression?’ he repeated, in a voice of disbelief. ‘No, that can’t be right. I can’t be depressed.’

‘I’m sorry,’ I said, puzzled, ‘but what do you mean you can’t be depressed?’

‘Well, depression…’ he mumbled, ‘it means you’re unreliable. It means I might be a danger to myself or others. It means I can’t volunteer at the animal shelter anymore because I might hurt the animals.’

‘Whoa!’ I said quickly. ‘That is very much not what that means.’ I put down his intake questionnaire and gestured vaguely out the clinic window. ‘Do you know how many people out there will develop depression?’

The proportion of the population who have experienced a psychiatric disorder, such as ‘major depression’, at some point in their lives is what epidemiologists refer to as ‘lifetime prevalence’. This statistic has a range of important practical implications. For example, higher rates of a disorder might suggest that correspondingly greater resources should be allocated to support early diagnosis and treatment. Evidence that disorders are relatively common might also provide some comfort to those afflicted, helping to reduce stigma. Given that many common mental health problems can be brought on or exacerbated by structural problems in society, data showing a high prevalence of psychiatric disorders might even encourage us to take a closer look at the major sources of stress in modern life, such as income inequality, systemic racism or exposure to violence. Finally, evidence for high prevalence could challenge some of our assumptions about the nature of mental health – including, for example, that a life free from mental disorder is ‘the default’, and any life that falls short of this standard is an unfortunate aberration.

Given these implications, it might come as a surprise that prevalence statistics in psychiatry have only recently begun to approximate ‘settled science’. A key reason is that, for most of the 20th century, psychiatric disorders were too nebulous to be diagnosed reliably by lay interviewers in large, community surveys. One way of circumventing this problem was to look at the proportion of individuals diagnosed by trained professionals in treatment settings but, because most people with a diagnosable mental health problem don’t seek or receive treatment, the number of clients treated for mental health problems in healthcare systems will almost always be far lower than the number of people struggling with these issues in the general population.

Eventually, however, the diagnostic criteria and interviews used in psychiatry improved to the point that diagnoses could be made reliably by lay interviewers. These advancements laid the foundation for the National Comorbidity Survey (NCS). Conducted in the early 1990s, the NCS was the first large-scale study in the US to assess the prevalence of mental disorders in a nationally representative, community sample. Based on interviews with more than 8,000 adult participants, the survey revealed that close to one in three of them had met criteria for a psychiatric diagnosis in the past 12 months, and that nearly half had experienced a diagnosable psychiatric disorder at some point in their lives.

Mental health problems are a lot like physical health problems – common, often temporary and, mostly, an unavoidable consequence of ‘normal’ life

Early reactions to the NCS were mixed, with some interpreting the findings as evidence of a hidden ‘epidemic’ of mental illness and others arguing that the high rates of disorder in this and other studies simply provided more evidence that psychiatry was ‘medicalis[ing] normality’. Against this contentious backdrop, newer longitudinal studies that followed research participants for decades began drawing attention to a new statistical wrinkle: ‘recall failure’.

Recall failure is a catch-all term used in epidemiology to describe discrepancies between what a person reports at one time point versus another. For example, if you follow the same people for three decades and ask them every few years if they’ve ever had diabetes, research shows that the number of people who say ‘yes’ at some point during this 30-year period will generally be the same as the number who say ‘yes’ when you ask them again at the very end of the study. Thus, diabetes is a condition with minimal recall failure because people who have diabetes hardly ever forget it.

By contrast, longitudinal studies of psychiatric disorders find that these conditions register as more than twice as common when people are assessed for them repeatedly versus when they are assessed at a single time point and asked to look back over their lives – a sign of high recall failure. The reasons for the high rates of recall failure in psychiatry are likely complex and multifaceted, but chief among them are that most episodes of mental disorder are temporary, and people (especially those who never seek treatment) will often forget about or re-frame their mental health struggles once they come out on the other side feeling better.

More: Psyche