Jerry Coyne FUD, part II

In my first post on Coyne's two part critique of Andrews and Thomson's paper on depression, I argued that Coyne's hand-wringing about the paper was either paternalistic, naive, or, most likely, just a means to sow FUD about evolutionary psychology and evolutionary psychiatry. In this post and the next I discuss the critique itself.

It is obviously legitimate to critique A&T but this requires reading the entire paper and knowing something about the topic. Coyne's critique is sloppy at best. It shows he didn't read large chunks of the paper and is ignorant of well-known findings in depression research. Coyne also fails to recognize basic elements of his own field. This post will focus on points of strong disagreement between myself and Coyne, and the next will focus on points of partial agreement.

My gloss of Andrews and Thomson's analytical rumination hypothesis (ARH)

Before I continue, I'll provide my gloss of the Andrews and Thomson argument, which is essentially a refinement of the psychic pain hypothesis for depression first put forward by evolutionary biologists Richard Alexander and Randy Thornhill, and then taken up by the psychiatrist Randy Nesse and others. The psychic pain hypothesis proposes that psychic pain, such as depression, is analogous to physical pain. Physical pain is undoubtedly an adaptation whose function is to detect and avoid physical injuries. If you break your ankle, the pain informs you that you have seriously injured your ankle, it inhibits you from walking, thereby preventing further damage to the ankle, and it helps teach you not to repeat whatever you might have done to break your ankle in the first place.

According to the psychic pain hypothesis, depression has a similar function. If your valued spouse leaves you, the psychic pain informs you that something very bad has happened (as measured in the currency of fitness), it inhibits you from increasing the social damage (e.g., stop being a workaholic and maybe your spouse will return), and it helps teach you not to repeat whatever you might have done to cause your spouse to leave (e.g., if your spouse returns or you get another spouse, spend more time with him or her and less time at work).

In the Andrews and Thomson update of the psychic pain hypothesis, complex life problems, such as abandonment by a spouse, trigger depression, which then causes the depressed person to ruminate intensely about this problem, thereby increasing the chances that he or she will find some solution, such as a way to convince the spouse to return or to obtain another spouse.

Although Coyne doesn't mention it, the evidence is overwhelming that depressed folks ruminate a lot, they ruminate about their problems, and they say they are trying to understand and solve their problems (Nolen-Hoeksema et al. 2008 and references therein). What is unresolved is whether such rumination actually helps the depressed solve their problems, or instead impedes problem solving, thereby prolonging depression.

Andrews and Thomson make the case that rumination is actually helpful, contrary to the prevailing view that such rumination impairs cognitive performance and is therefore dysfunctional and harmful.

A key finding of A&T is that no study has ever examined whether depressive rumination improves one's ability to solve the problem that triggered the depression in the first place. Instead, the evidence for the prevailing view comes from studies in which depressed individuals solve cognitive tasks unrelated to the life problem they see as having triggered their depressions.

Imagine being depressed over a marital breakup, and then asked to generate random numbers in a lab. It's not too surprising that performance on this task will be "impaired" relative to folks who haven't recently suffered a marital breakup because, as A&T argue, the depressed folks are rightly ruminating about the marital breakup instead of focusing on random number generation.

In my opinion, the Andrews and Thomson hypothesis will eventually be viewed as simple common sense. Now, though, it's a radical revision of much current thinking on depressive rumination.

In the following discussion, I take Andrews and Thomson's Psych Review article as the authoritative source of their argument rather than their Scientific American article, because Scientific American heavily edits articles in order to shorten and simplify them for a popular audience, and this editing can distort the actual argument. I recently wrote a short piece for Scientific American, and even after I approved the final version my (heavily edited) wording was changed in a way I didn't like.

What, now, of Coyne's criticisms?

How to identify adaptations

Coyne questions the method by which Andrews and Thomson aim to demonstrate adaptation:

The authors claim that “such coordination [of features of depressive rumination] makes it very unlikely that depressive rumination is a by-product of biological processes or is attributable to chance. Just as the highly structured and complex design of the vertebrate eye must have been constructed by selection and not by chance, it is difficult to see how chance biological processes could have generated such coordination. It suggests that depression evolved by natural selection, probably because depression helped people analyze and solve the problems about which they were ruminating.”

Do I really need to debunk this logic? It’s not a kind of logic that I’m familiar with as an evolutionist. Any disease or malady, psychological or otherwise, involves a coordinated group of symptoms. Schizophrenia also involves a coordinated group of symptoms that often includes catatonia, hearing voices, disordered thinking, and changes in neurotransmitter quantity. Does that make it an adaptation? I haven’t seen anybody claim that, despite the fact that schizophrenia is also found in nearly all cultures.

Coyne pulled the above quote from near the end of A&T's article, but this is how A&T introduce their method (p. 622):

When a trait has features that proficiently promote a specific effect, that will go a long way to demonstrating that the effect is an evolved function of the trait because it is highly unlikely that chance processes could be completely responsible for the trait’s features (Andrews et al., 2002a; Thornhill, 1990; G. C. Williams, 1966). An exploration of a trait’s features in relation to its effects for the purpose of evaluating whether it has an evolved function is often called a design analysis, which is the form of argument used in this article.

A&T are not claiming that coordination of depressive symptoms per se is evidence of adaptation. Rather, the evidence of adaptation comprises the coordination of depression-induced neurophysiological changes for the specific purpose of effectively analyzing the life problem that triggered the depression. This type of analysis is standard stuff in evolutionary biology. Here, for example, is Darwin on the coordination of woodpecker traits for the purpose of catching insects:

Naturalists continually refer to external conditions, such as climate, food, etc., as the only possible cause of variation. In one very limited sense, as we shall hereafter see, this may be true; but it is preposterous to attribute to mere external conditions, the structure, for instance, of the woodpecker, with its feet, tail, beak, and tongue, so admirably adapted to catch insects under the bark of trees. (On the origin of species, Introduction.)

A&T don't bother citing Darwin on this point, but they do cite George C. Williams, a giant of modern evolutionary biology. Here is Williams, in the cited reference, explaining the relationship between organism design and natural selection:

If...we were asked how the apple acquired it various properties, and why it has these properties instead of others, we would need the theory of natural selection, at least by implication. Only thus could we explain why the apple has a waterproof wax on the outside, and not elsewhere, or why it contains dormant embryos and not something else. We would find an impressive list of structural details and processes of the apple can be understood as elements of a design for an efficient role in the propagation of the tree from which it came. We attribute the origin and perfection of this design to a long period of selection for effectiveness in this particular role. (Adaptation and Natural Selection, p. 6.)

Note that the question I'm addressing now is not whether A&T provide a convincing case that depression shows evidence of design for analytical rumination (I think they do, but I'll examine that later), nor is the question whether design analysis is the best way to identify adaptations. Rather, the question is whether A&T employ a logic that would be familiar to an evolutionary biologist. The answer is clearly yes.

It's not obvious to me why Coyne claims he is unfamiliar with this logic. Is he really unfamiliar with a basic approach to identifying adaptations, as laid out by both Darwin and G.C. Williams? Is he being deliberately misleading? Or is he writing about an article that he didn't carefully read? On current evidence, it is impossible to tell which of these is true.

Do negative life events cause depression?

It is critical for the ARH that complex life problems cause depression. Coyne questions the evidence for that:

Is depression caused by difficult social problems? On this crucial point the authors give virtually no evidence. [NOTE: It's since been called to my attention that there is work suggesting some association between depression and  difficult life situations. Again, however, there is a cause-and-effect problem here: people with a tendency to be depressed may more often get themselves into difficult life situations.]  People who have experience depression often say that its onset is often mysterious, not associated with an identifiable problem (see, for example, some of the commenters on yesterday’s post). But the authors wave that difficulty away:  “Many people may be reluctant to disclose the reasons for their depression because the problem is embarrassing, reputationally damaging or otherwise insensitive, which is often why depressive episodes may appear to be endogenous. . .”  This is almost Freudian in its hauteur.  Here the authors presume that there is a social reason for depression rather than treating it as a hypothesis.

The authors do note that “interpersonal conflict is commonly associated with depression. . “.  One example is that “in married couples, the risk for major depression is about 40 times greater if the couple is unhappily married.”  Well, you can see the problems with this: what is the cause, and what the consequence?  Perhaps the social problems result from undiagnosed depression. It’s easy to see that being married to someone with incipient or undiagnosed depression could cause unhappiness.  Depressed people are often hard to get along with.

In reality, mental health researchers have known for years that severe negative life events are strongly correlated with depression and probably cause it, at least in part. The causal role of negative life events in depression is a well-supported, if not yet proven, hypothesis. As A&T say in their paper:

Depression is commonly thought to be caused by severe problems or stressors, often of a social nature (Brown & Harris, 1978; Hammen, 1992; Kendler, Karkowski, & Prescott, 1999).

And as for the haughty suggestion that some depressed people might be embarrassed to disclose the reasons for their depression, A&T cite Leff et al. (1970) on this point (but Coyne omitted this cite when he pulled the quote from A&T).

It was not A&T's job to bring Jerry Coyne up to speed on the last several decades of depression research, including all the evidence -- well known to depression researchers -- supporting a causal role for negative life events. That's why they cited three other papers in support of this assertion (and by the way, the evidence that Coyne claims was called to his attention is the Kendler et al. study cited by A&T).

This evidence includes that negative life events typically precede depression, ruling out the possibility that depression caused the negative life event (this doesn't, rule out the possibility of a confound, however). The single most powerful "predictor" of depression is the death of a close relative. Is it really plausible that depression causes close relatives to die? Or that depression and death of close relatives are only linked by a confound -- some unknown factor that causes close relatives to die and causes depression, but with no causal connection between the death and the depression?

Kessler's (1997) review on the relationship between stressful events and depression found causality in both directions:

The evidence reviewed above clearly shows that inventories of stressful events predict subsequent depression. A smaller number of controlled comparative studies of people exposed to single major life events provide strong evidence that at least part of this association is due to events causing depression. It is also clear from other studies that this relationship can be reciprocal and that depression can elicit or exacerbate certain stressful events and difficulties. (p. 209)

The empirical evidence is more than strong enough to warrant theorizing about depression as if life stressors actually do, in part, cause depression. That's certainly the assumption made by many mainstream (non-evolutionary) theories of depression, such as the hugely influential diathesis-stress model (e.g., Metalsky et al. 1987 and references therein), which posits that life stressors precipitate depression in constitutionally vulnerable individuals (the diathesis part).

The causal role of life stressors in depression is also the assumption made by the many researchers exploring the neurophysiological effects of hypothalamic-pituitary-adrenal (HPA) axis activation on the brain and subsequent depression. This latter work also explores severe stressors occurring in childhood, which are thought to predispose the adult to depression (for review, see here). These folks see depression as the dysfunctional effect of the biochemical cascades caused by stress, of course, but in their model life stressors are a cause of depression nonetheless.

If A&T are wrong about the causal role of life problems in depression, they'll have a lot of company.

A final point: many of the stressors that are most closely associated with major depression, such as death of a close relative and relationship problems, are common and almost certainly not evolutionarily novel. Hence, it seems that the brain should have evolved to respond to them without suffering serious dysfunction.

The significance of depression's high prevalence

Coyne first argues, correctly, that just because a trait is common does not mean that trait is an adaptation:

The authors assert that the “high prevalence estimates” of depressive disorders, and their worldwide presence “suggest[s] that much of what is currently classified as depressive disorder represents normal psychological functioning.” This suggests nothing of the sort, any more than the frequency and ubiquity of toothaches suggests that these are part of normal dental functioning. The only way around this grotesque conclusion is the semantic tactic that anything that occurs in more than 15% of people is “normal” by definition.

In their Psych Review article, here is what Andrews and Thomson's actually say about the high prevalence of depression:

Such evidence [of the high prevalence and universality of depression] suggests that much of what is currently classified as depressive disorder represents normal psychological functioning (Horwitz & Wakefield, 2007).

Notice that Andrews and Thomson are citing Horwitz and Wakefield (2007) on this point, so Coyne really needs to engage with the book-length argument of the latter two scholars, not with Andrews and Thomson (I'm beginning to wonder if Coyne understands the purpose of citations). I haven't yet read Horwitz and Wakefield, so I don't know how they make their case, but I'm pretty surprised that Coyne, a well-regarded evolutionary biologist, couldn't easily see why, from an evolutionary perspective, depression differs profoundly from a toothache.

Toothaches are caused by irritated nerves, and these nerves become irritated by dental infections or by physical tooth damage (I'm simplifying for the sake of argument). From an evolutionary perspective, diseases caused by infections are expected to be common because there is an evolutionary arms race between pathogens and hosts. This arms race means that the human immune system will often fail to prevent infection. Hence, many people will suffer a dental infection at some point in their life.

Diseases caused by physical damage are also expected to be common. The tooth, as an adaptation, exhibits a number of tradeoffs between strength, size, and function. Sometimes teeth will be subject to physical forces that exceed their "engineering specifications" and they will break, causing a toothache.

But, so far as we know, most depressions are not caused by pathogens and they're not caused by physical brain injuries resulting from, e.g., blunt force trauma or toxin exposure. (People with traumatic brain injuries do appear to suffer higher rates of depression, but the vast majority of depressions are not caused by traumatic brain injury.)

As Coyne should know, several evolutionary theories of senescence predict that, absent infection or physical injury, dysfunctions of critical adaptations, such as the heart or brain, should be relatively low in young adults but increase with age. That's the pattern we see, for example, with heart attacks, which are rare among young adults but common among elderly adults. That's also the pattern we see with dementia, which is very rare before the age of 60 but increases exponentially for at least the next 25 years (Finch and Zelinsky 2005).

Andrews, Thomson, Horwitz, and Wakefield are right to wonder why the brain, unlike most other organs, is thought to suffer very high rates of a major dysfunction in young adults, in the form of depression, despite no evidence that most such dysfunctions are caused by infection or physical damage. None of this proves that depression is an adaptation. There is, after all, still debate about the various evolutionary theories of aging, among other things. But it sure should make mental health researchers, and Coyne, pause and think.