Learned helplessness

Why does an animal that once could not escape stop trying long after escape has become easy?

An animal is exposed to something unpleasant it cannot avoid. Later the situation changes: a low barrier now separates it from safety, and stepping over would end the discomfort at once. It does not step over. It sits there taking it, in a world that would reward one small action.

The obvious reading is that it learned something in the first phase and is now applying it wrongly — it learned that nothing it does matters, and carried that lesson into a situation where things do matter. That reading gave the phenomenon its name and held for half a century. It is also, on the current evidence, the wrong way round. Working out why is more instructive than the original story.

First, be sure the phenomenon is about controllability rather than about the unpleasantness. That needs a particular design. Take two animals wired so they receive the same stimulation at the same moments, but one can terminate it by acting and the other simply receives whatever the first one gets. Identical exposure; only the contingency between action and outcome differs. Test both later where escape is easy. The one that had control escapes; the one that did not, largely does not.

That is a clean result and it is the foundation of everything after. Whatever is happening is caused by the relation between behaviour and outcome, not by the stimulation.

Now ask what the animal must have acquired. The original account said: an expectation that responding is futile, generalised beyond the situation that taught it. Reasonable — but notice it assumes the passivity is the learned part. Is there an alternative? Suppose instead that prolonged uncontrollable aversive stimulation simply produces passivity, directly, as an unlearned reaction. Then the yoked animal that could act is the one who acquired something — namely that it had control — and what it acquired was protection.

The two stories predict the same experiment and differ entirely in what carries the memory. How would you tell them apart? Look inside, at what changes and where.

What is found is that intense uncontrollable stress strongly activates a brainstem cluster of serotonin-producing cells, and that activating that cluster directly is enough to produce the passivity and the failure to escape — in an animal that had no exposure at all. Nothing was learned; the machinery for passivity is already there and simply gets driven. Meanwhile, when the animal does have control, a region of prefrontal cortex that detects the contingency becomes engaged and acts to inhibit that brainstem cluster. Silence the prefrontal region in an animal that has control and it behaves as though it had none. Engage it in an animal that has none and the passivity is prevented.

So the direction reverses. Passivity is the default response to prolonged aversive stimulation; what is learned is control, and learning it installs a standing inhibition that carries forward into later situations. The animal in the box has not learned helplessness — it has failed to learn the thing that would have protected it. The two authors of the original work published this reversal themselves, fifty years on.

Does the revised account make its own risky prediction? It makes a striking one. If control is what gets learned, then giving an animal experience of control beforehand should immunise it against later uncontrollable exposure, and teaching control afterwards should undo the passivity. Both are found. Under the old account, prior control has nothing to act on and the immunisation result would be a puzzle; the refuting observation for the revision would be a demonstration that passivity persists undiminished after successful control training, and that is not what happens.

This is the promised complement to Habituation versus sensitization, which noted that an inescapable stimulus sensitises where the same stimulus under one's own control habituates. The fixed point of difference is the same one: it is the contingency, not the stimulus, that the nervous system is tracking, and the yoked design is what isolates it in both cases.

Two clarifications, then the honest limits.

The revision does not overturn the experimental result, only its interpretation — the yoked finding stands exactly as reported. And "default" does not mean inevitable: passivity requires sustained, intense, uncontrollable exposure, and a great deal short of that produces nothing of the kind.

Now the limits, which matter because this is a finding whose fame far exceeds the range of its evidence.

The animal work is strong and the circuit account is built on direct intervention rather than correlation. The human extension is a different matter. Laboratory inductions of helplessness in people are typically small studies with mild manipulations, and the reformulation that added attributional style — whether a person explains failure as internal, stable and global — moved the theory into territory where measurement is by questionnaire and the effects are far less securely established. Treating the animal circuit as though it had been demonstrated in people is a leap the evidence does not license.

And the everyday use of the phrase runs further still. "Learned helplessness" is now applied to poverty, to abusive relationships, to disengaged employees — situations where the constraint may be entirely real rather than a mis-learned expectation, and where calling the response a learned deficit locates the problem in the person. The original finding is about acute uncontrollable stress in a controlled paradigm. I would not carry it further than that, and would not quote effect sizes for any of the human versions.

• How closely the animal circuit maps onto human depression, and why the attributional reformulation is much weaker evidence than the yoked experiments.
• Whether perceived control without actual control produces the same protection, and what that would imply about the detector.

Every term the collection defines is gathered in the glossary.