The causal structure of psychopathology and why it matters

Psychiatry is blessed or cursed with theoretical questions that never go away. They mostly take the form of dichotomous oppositions. Nature versus nurture is the most familiar1, but there is also mind versus brain, medication versus talk therapy, conscious versus unconscious, diseases versus problems in living, and categorical versus continuous. These chestnuts have been around so long and debated so many times that by now they are usually dismissed as either settled or irresolvable, yet to everyone’s frustration they refuse to go away. This essay proposes that they are tied together, and could eventually be resolved by consideration of the ways in which behavioral entities are recognized, defined and caused. The universe is organized hierarchically, with big things made out of little ones. Organisms – to which I will limit myself here – are no different. Atoms make molecules, molecules make chemical processes, which make cells, organs, and eventually individuals. Individuals go on to make families, groups and societies. As a hierarchically organized entity changes – that is as it behaves – the changes are manifest at all levels. The tiniest gears turn as a clock tells time, and the hands of the clock move along with motion in the gears. The whole thing churns together. Once again, organisms – now limiting myself further to people – are no different. An individual lapsing into a depressive episode is churning in the cells and genes and neurons, in a marriage and on a therapeutic couch, as part of a history that is personal, cultural and evolutionary. An observer of this multi-level developmental churning is faced with a vastly complex, chaotic system that does not come with ready-made edges to make sense of it. Human observers of human behavior have evolved the ability to create linguistic categories that impose edges on the chaos. Entities created by language regularize developmental chaos by simplifying it. By regularizing it, they make it possible to talk about behavior in systematic ways. Geneticists, neuroscientists, psychiatrists and sociologists create entities at different scales. The parable about the blind men and the elephant misses the point: it is not one scientist holding the trunk and another the tail; it is one with a microscope and another with an aerial camera. Any entity that is created by human observers depends on characteristics that exist at some scale. If we say that an individual is suffering from anaclitic depression, we are referring to edges that are imposed on behavior in personal relationships and in therapy. Observing that a person is a non-responder to selective serotonin reuptake inhibitors (SSRIs) finds edges in pharmacological metabolism and responses on the Beck Depression Inventory. Observing that a person is in a high-risk group based on a polygenic score for depression does not refer to behavior, but rather to the DNA that he/she has inherited. These different entities are to some extent independent of each other, but it is not the case that an entity specified at one scale does not exist at another: it is just less well defined. It becomes blurry2. People diagnosed by their analyst with anaclitic depression may have detectable differences in their SSRI responses or polygenic scores, but psychoanalysts cannot predict drug response, and one cannot identify anaclitic depression in the genes. Except that, sometimes, one can. It occasionally turns out that what appears to be an entity defined at one scale is actually a blurry version of another entity more sharply defined at a different, usually smaller, scale. Chorea is a term from descriptive neurology that refers to a variety of abnormal movements that occur for many different reasons, including phenomena such as hysterical contagion among groups of young people at a school. But in 1872 (i.e., before Mendel), G. Huntington3 noticed that a subset of choreiform movements appeared in families in a structured way, along with a variety of other devastating symptoms. The now eponymous chorea that Huntington called “hereditary” is a large-scale manifestation of a crisply defined entity on a genetic scale. This kind of cross-level reduction is powerful when it occurs, but it must be remembered that it is not inevitable. You could study bankruptcy for the rest of time and never discover a sub-syndrome that is crisply explained by its relation to genetic or neurological processes, despite the fact that bankrupt individuals certainly show blurry genetic and neurological differences from the financially solvent. Behavioral entities are composed of molecules, neurons and genes, but they are caused by other behavioral entities defined on the same scale. This is true even for Huntington’s disease. It is tempting to say that the choreiform movements of the patient with Huntington’s disease are caused by the HTT gene on chromosome 4, but doing so risks explaining the cause of an entity’s behavior with its own structure on a different scale. The turning of gears in a clock does not cause it to tell time: it is the clock telling time. Although it does not much matter in something as well-understood as Huntington’s disease, it is better to say that the cause of the disease is the inheritance of the HTT gene from an exogenous parent; the motor and cognitive symptoms that ensue are a larger-scale representation of the smaller-scale biological consequences of carrying the gene. The distinction between causal and compositional explanation is more important in entities that lack the well-understood structure of Huntington’s disease. Let’s say that an individual’s bankruptcy is caused by redlining practices in the local real estate market. The physical and behavioral entities that might make someone a target of redlining, such as socially defined race, economic class, and personal budget practices, all have blurry representations at smaller scales. So when redlining ⇒ bankruptcy, the whole blurry structure of the neurology of home ownership gets crossed with the equally blurry genetics of savings plans. That guarantees that the geneticist who conducts a genome-wide association study (GWAS) of bankruptcy will find something, but what is found will not be causal, and it will not explain anything. It will just be a blurry genetic representation of a process that can only be meaningfully analyzed at a larger scale. These examples illustrate why psychiatry’s old dichotomies are so persistent: they are not merely empirical puzzles waiting for better data, but reflections of the difficulty of drawing stable, meaningful boundaries around behavioral entities across multiple explanatory scales. Understanding the causal structure of complex behavioral entities requires a different skillset than the technical tools that are applied under the assumption that everything will eventually yield to biogenetic reduction. It requires philosophical discourse (this essay owes much to the philosopher of science W. Wimsatt4), advanced statistical methods (the clinical psychologist and philosopher P. Meehl devoted the second half of his career to the problem5), and (hardest of all), if recent history is any guide, attention to the null hypothesis. One must ask questions such as: “Are candidate gene studies of depression working, and what does it mean if they are not?“6. What it means is that the causal structure of depression may be more like the causal structure of bankruptcy than it is like Huntington’s disease. If that turns out to be the case, depression will fall more in the domain of clinical and social science as opposed to GWAS and brain imaging. The so-called biopsychosocial model is tautologically true in a hierarchical universe, but it obscures something important: entities are better defined, discussed and explained at some scales than they are at others. Determination of that optimal scale is prior to specification of causes. It must be understood what depression is before it can be understood what causes it. When we debate nature versus nurture, or mind versus brain, we are really debating how best to impose conceptual order on a system that resists tidy partitions. There may be no final resolution to these tensions, but there can be progress if we remain vigilant about the scales at which explanations operate, and maintain a distinction between genuine explanatory reduction, which is rare but powerful when it occurs, and simple compositional analysis, which is universally possible but generally uninformative.