Martin Seligman

The question arose out of a malfunction. In Richard Solomon’s laboratory in Pennsylvania, dogs in a shuttle box were meant to learn to avoid an announced electric shock by jumping over a low barrier into the safe half. Some of the animals did not: they lay down and let the shocks pass over them although the way out stood open. For learning theory this was a nuisance — behaviour that could not be reinforced because it did not occur. The doctoral student Martin Seligman and his colleague Steven Maier suspected that these animals had previously learned something else: that nothing helps.

The design with which they tested this in 1967 is a model piece of discrimination and has been called the triadic experiment ever since. Three groups: the first received shocks it could switch off by a response; the second got exactly the same shocks for exactly as long — yoked to the animal of the first group — but could change nothing about them; the third got none. The yoking is decisive: both groups received the same amount of pain, and differed solely in control. The next day all came into the shuttle box, in which escape was possible for every animal. The animals of the first and third groups jumped. The animals of the second mostly stayed lying and took the blows, although the barrier was low and the way out visible.

The reading that came of this was called learned helplessness and was bold: the animals had learned that their behaviour has no effect, and transferred this expectation to a new situation. Out of the animal finding came a model of depression — passivity, loss of drive, giving up — and the transfer to humans followed promptly: in experiments with unsolvable tasks or uncontrollable noise, students afterwards performed worse on solvable tasks. In one stroke the model explained why people in hopeless situations stop acting, and gave clinical psychology its first laboratory model of depression.

The first rebuild came from within. Because not all people became helpless under the same circumstances, Lyn Abramson, Seligman and John Teasdale extended the model in 1978 by attribution: what matters is how someone explains the uncontrollability to themselves — as lasting or temporary, as general or confined to this area, as their own failure or as circumstance. Out of this version came the pessimistic explanatory style, a questionnaire construct with a research line of its own — and, via the hopelessness theory of depression, a bridge to Beck’s cognitive model in the same folder.

The transfer to humans was tested in the 1970s to the same plan, only with a bearable stimulus. Instead of shocks there was loud noise: one group could switch it off by pressing a button, a second heard the same noise without influence, a third none at all. Afterwards all had to work at a hand box in which one movement ended the tone — and the participants of the second group more often left their hand lying. Further experiments showed that the effect transfers between domains: unsolvable reasoning tasks made people passive later with noise, and the other way round. With that the pattern was no longer a peculiarity of the animal laboratory.

To the file belongs also the objection that concerns not the methodology but the procedure itself. The dogs of the second group could not escape the pain; that was the purpose of the arrangement and therefore not avoidable by better planning. Experiments of this kind would no longer be approved in this form under today’s animal-protection rules, and Seligman himself later said he gave up working with dogs because he could no longer bear to watch. The archive records this because the sentences about learned helplessness are readily quoted in self-help books without their origin.

This file’s archival turn is the most radical in the whole holding, because it comes from the originators themselves. In 2016 Maier and Seligman published, under the title Learned helplessness at fifty, a reappraisal that reverses the original reading. Neuroscience shows: what follows uncontrollable stress is not learned passivity — passivity is the brainstem’s default response to prolonged stress. What is learned is the opposite: experienced control brings the prefrontal cortex to inhibit this basic response actively, and this inhibition is transferred to later situations. It is not helplessness that is learned but confidence. The theory’s name, the authors write drily, was in retrospect badly chosen.

The finding on which the reversal rests is localised in the brain. Sustained, uncontrollable stress excites a small serotonergic nucleus in the brainstem, and this excitation produces the passivity — immediately, without learning, without expectation. What control accomplishes happens one storey up: an area of the front of the brain registers that one’s own action ends the stimulus, and from there inhibits the nucleus. This inhibition is the actual learning, it is tied to the experience of control, and it is carried over to later, similar situations. The animals of the second group had thus learned nothing — they had merely learned nothing that could switch off the basic response.

For the archive this is a special case in two respects. The finding itself — animals and humans who have experienced uncontrollable stress behave more passively afterwards — is robust across fifty years and hundreds of studies; it is one of the best-reproduced behavioural patterns in psychology. What has been overturned is the explanation, and by methods that did not exist in 1967. And it was overturned by the pair of authors themselves, half a century later, in the same journal in which the theory had grown large. Whoever looks in this archive for an example of self-correction that works finds it here — beside Kahneman’s recantation of the priming chapters.

The applications that followed from the optimistic turn are the weakest part of the file. Out of explanatory style came a school programme for the prevention of depression, in course form, over weeks — and the reviews of it find small effects that largely disappear after six to twelve months and come out smaller in studies without the involvement of the programme’s developers. The same idea was sold from 2009 on a large scale to the US Army: a resilience programme for over a million soldiers, with a contract volume in the tens of millions, introduced without prior randomised testing. The criticism was pointed and came from within the field — a programme of this size, it said, should have been tested first and rolled out afterwards, and the later efficacy figures came from those involved themselves.

Best evidenced is the part that belongs to the neighbouring file in the same folder. Out of the attributional version came the hopelessness theory of depression, and that was tested the way a prediction must be tested: prospectively. A study laid out over years followed undergraduates who had been sorted in advance by explanatory style into a high-risk and a low-risk group, and found markedly more first depressive episodes in the high-risk group. That is the strongest evidence that the habit of thinking precedes the disorder rather than merely accompanying it — and at the same time the point at which Seligman’s line runs into Beck’s cognitive model.

What stands today is therefore in three parts. The behavioural pattern: replicated. The reading of 1967: revised by its originators. The clinical application: mixed — pessimistic explanatory style predicts depressive symptoms moderately, the interventions derived from it work modestly, and the positive psychology that followed on has lost considerable ground in the replication crisis: exercises such as gratitude letters or writing down three good things show small effects in meta-analyses that shrink further after correction for study quality and publication bias. The file’s starting point nonetheless remains valid, only with the sign reversed: whoever has experienced control bears strain better. That is better news than the one the theory became famous for.