Excellent! This has always been my guess about depressive realism, but I never did the careful lit review to see if it lined up.
A bit of overconfidence is, curiously, rational. That's if you're also burdened with some unfortunate cognitive limitations that prevent your brain from perfectly telling times when you should err on the side of optimism from times you shouldn't.
And here we all are trying to think with monkey brains.
I posit: high enough that you’re slightly overoptimistic about stuff you can’t control, so that for stuff where confidence itself makes the difference, you squeak in.
FWIW, see Notes on Optimism, Hope, and Trust for more on this hypothesis, including William James's speculations.
When subjects completed the task in the presence of an observer, depressed students perceived themselves to have more control than did nondepressed students.
So...depressed people should get out more? :)
great post. I found Bertrand Russell's "The Conquest of Happiness", especially the 2nd chapter on Byronic Unhappiness to be a well articulated argument against a commonly held belief that to truly be an aware or intelligent person either requires or results in depression.
Depressive realism is the idea that depressed people have more accurate beliefs than the general population. It’s a common factoid in “things I learned” lists, and often posited as a matter of settled science.
In this post, I’ll explore whether it’s true.
The depressive realism hypothesis was first studied by Lauren Alloy and Lyn Yvonne Abramson, in a paper called Judgment of contingency in depressed and nondepressed students: Sadder but wiser?. Undergraduates had access to a button, which might (or might not) influence whether a light came on. Depressed students - but not nondepressed students - were pretty good at guessing how much influence the button had over the light. Specifically, nondepressed students tended to think they had more control than they actually did.
In other words, the seminal study showed something a lot more specific than “depressed people are more accurate”. It showed that they were better at inferring their own degree of influence. Or to put a finer point on it, (mildly) depressed people suffer less from the illusion of control.
There are two lines of criticism I want to explore here, before moving on to other studies. First, is the study probably accurate? And second, if it is, does it imply what it seems to?
Methodology
To get the big one out of the way first: Alloy and Abramson’s sample only consisted of undergraduates. This is an extremely common approach for psychological studies, because the people doing the studies are professors and undergrads are close at hand. But undergraduates are not exactly representative of humanity writ large. Also, the entire field of psychology is on fire, so “business as usual” is a poor defense.[1]
Furthermore, the total number of undergraduates in the study was 288: 144 depressed, and 144 not depressed. Depression was diagnosed through the Beck Depression Inventory, which seems fine, but does mean we’re probably dealing with self-report, so mileage may vary there too.
Finally, the study was done in 1979. This is the same decade as the Stanford Prison Experiment, one of the most famous studies of all time and now on shaky enough ground that Vox calls it straight up fraud.
As far as seminal studies from the 1970s go, this one seems pretty good. There have been replications (we’ll get to those), and if there’d been fraud or misconduct it would be a pretty good scoop. But again, 288 undergraduates in the 1970s should probably not have too large an impact on your views about much of anything.
Implications
What if this study were unimpeachable, and replicated perfectly every time? Would we have to agree with its normative implications, and chalk one up for depression? Not necessarily!
In most life situations, when someone is trying to figure out how much influence they have, their feeling is one vector of that influence. Especially situations that actually matter. If someone believes they deserve a promotion, for example, they’ll probably argue more persuasively for it than someone merely going through the motions. A depressed person might say “well, if I were to approach that attractive person I’d just be rejected”, and even if their friends egg them on and they go for it, their very lack of confidence might end up proving them right.
As a toy model, imagine everyone has to determine a single, global locus of control. Believe that your entire destiny is always fully in your hands, and you’ll wager it all on the roulette wheel and believe you’ve got to win because you want it so much. Believe you’re totally powerless in your own life, and you’ll sit in the basement drinking gin and watching paint dry. The right choice is somewhere in the middle. But where?
I posit: high enough that you’re slightly overoptimistic about stuff you can’t control, so that for stuff where confidence itself makes the difference, you squeak in. So sure, a slightly depressed person might be a little more correct about whether a button turns on a light. But if in exchange they’re wrong about if they should bother learning a new skill, or taking a walk every morning, or drinking more water, well, that’s a bad trade.
(Also, one more time, it was just 288 undergraduates in 1979.)
There have been several replication attempts, expansions, and meta-analyses over the years. I’ll highlight a few. The headline: the very specific effect holds up, but it is indeed very specific.
Benassi and Mahler, 1985
In 1985, Victor Benassi and Heiki Mahler published the sassily named Contingency judgments by depressed college students: sadder but not always wiser. They replicated Alloy and Abramson’s findings, but found that variations upended their result. From their abstract:
When subjects completed the task in the presence of an observer, depressed students perceived themselves to have more control than did nondepressed students. In Experiment 2, the observer effects found in Experiment 1 were replicated, and we extended those results by showing that when response-independent outcomes occurred relatively infrequently, depressed and nondepressed subjects who completed the task in the presence of an observer did not reliably differ in their estimates of personal control.
So basically, if somebody’s watching, now it’s depressed students who are overconfident, and also (I think) if you make the button useless only occasionally, depressed people are no better than average at figuring the situation out.
I’ll chalk this up as a weak replication, which also suggests the original effect is pretty specific.
Dobson and Franche, 1989
In 1989, the extremely Canadian-named duo of Keith Dobson and Renée-Louise Franche released A conceptual and empirical review of the depressive realism hypothesis. To quote their abstract:
Positive evidence for the existence of depressive realism was found although the strength of that finding diminished as the ecological validity of studies increased.
I wasn’t able to get the full text of the study, but seems like basically a successful replication, with the caveat that in more natural settings, the effect looks less powerful.
Dobson and Pusch, 1995
Also by Keith Dobson, but now with Dennis Pusch, A test of the depressive realism hypothesis in clinically depressed subjects. Basically, they redid the original test from 1979, but digitally, and for clinically depressed (rather than just self-reporting undergraduate) patients. The sample size was only 45 women total (15 depressed, 15 formerly depressed, and 15 never depressed), so take with a hefty grain of salt, but, they claim, “The results provided limited support for the hypotheses.”
So a replication, but still in the very limited scope of the original study, with a small sample size, and “limited” support.
Jain, Bearden, and Filipowicz, 2013
If you’re wondering how depressive realism generalizes to situations that have nothing to do with the (variably) depressed person, look no further. This study, Depression and forecast accuracy: Evidence from the 2010 FIFA World Cup, saw if depressed people made savvier bets on a major sporting event.
If you’d like to place a bet yourself, now’s your chance! What do you think they found?
Across two different waves of predictions and with multiple measures and components of prediction accuracy, we find that depressed forecasters were less accurate.
I’d guess the causality is a mess here, but basically buy this study’s finding that no, being depressed doesn’t improve forecasting ability about random world events. Interestingly, too:
The poorer accuracy amongst the more depressed forecasters was primarily driven by a neglect of base rate probabilities: the depressed participants assigned probabilities that departed from the base rates more substantially, particularly for low base rate events.
So, depressed people tended to think more unusual stuff would happen, while non-depressed people played it safe. It wouldn’t surprise me if this is sports specific, or something. But in any case, the original study (and any follow ups I know of) had no opinion on the forecasting abilities of depressed people, and this study says depression makes you worse at it, so we definitely don’t have evidence for depressive realism on impersonal questions of fact.
Moore and Fresco, 2012
To cap things off, we have Depressive realism: a meta-analytic review. This one covers a lot of ground, so I’ll quote the abstract in full:
The current investigation represents the first meta-analysis of the depressive realism literature. A search of this literature revealed 75 relevant studies representing 7305 participants from across the US and Canada, as well as from England, Spain, and Israel. Results generally indicated a small overall depressive realism effect (Cohen's d=-.07). Overall, however, both dysphoric/depressed individuals (d=.14) and nondysphoric/nondepressed individuals evidenced a substantial positive bias (d=.29), with this bias being larger in nondysphoric/nondepressed individuals. Examination of potential moderator variables indicated that studies lacking an objective standard of reality (d=-.15 versus -.03, for studies possessing such a standard) and that utilize self-report measures to measure symptoms of depression (d=.16 versus -.04, for studies which utilize structured interviews) were more likely to find depressive realism effects. Methodological paradigm was also found to influence whether results consistent with depressive realism were found (d's ranged from -.09 to .14).
Basically: the effect size is very small, and the more rigorous the study, the smaller it tends to be.[2] I do think this meta-analysis bears out the original depressive realism study’s hypothesis: very impressive for a splashy study from the 1970s! But again, we’re talking a teensy effect under specific circumstances.
Sorry, not really. Depressive realism is a big deal, so there are way more studies that branch off in various directions. We haven’t touched on FMRI studies, studies specific to social judgments (whether depressed people judge social situations more accurately by being less self-serving), or studies of people rating their own performance on some task. I’m sure all these sub-ideas have their own lineages and debates. This post has mostly just traced the mainline of the concept, from its seminal study to a more recent meta-analysis.
That being said, as a nonexpert I’m confident enough to make the following claims:
This isn’t medical advice, but depression is bad. Try to avoid it, and don’t think having it would make you less wrong about the world.
This is a linkpost from my Substack. I don't crosspost very often, and if you enjoyed this post, you may like my posts there, too.
As in, replications of studies from top psychology journals only work somewhere in the ballpark of 30% of the time.
Specifically, studies without a rigorous definition of the relevant ground truth found a larger effect (clearly easier to fudge), and studies that relied on self-report for depression status found a larger effect.