The talking cure on trial 

The first challenge was philosophical. Karl Popper, who grew up in the Vienna of these arguments, used psychoanalysis as his standing example of a theory that explains too much. Whatever a patient did, the theory could accommodate it: a man who drowns a child and a man who dies saving one could both be described in terms of repression and sublimation. A claim that no observation can contradict, Popper argued, is not scientific, however true it might be. Freud's defenders answered that clinical evidence is real evidence, which is fair, and that psychoanalysis makes many specific claims, which is also fair.

The sharper challenge came from inside philosophy of science, from Adolf Grünbaum, who rejected Popper's blanket verdict and then examined Freud's actual arguments in detail. His conclusion was harder to dismiss: the clinical evidence Freud relied on cannot support his causal claims, because the analyst's suggestions, the patient's wish to please and the long process of joint story-building contaminate the data. What looks like recovered material may be produced by the treatment.

Then the outcome question. Hans Eysenck argued in 1952 that patients receiving psychotherapy improved no more than untreated patients did, a claim that was overstated and that provoked exactly the research programme it needed to. Decades of trials followed. The honest current picture: talking therapies do help many people, and modern psychodynamic therapy, which is Freud's clinical descendant rather than his theory, performs respectably in trials for depression and some anxiety and personality disorders. What the evidence does not support is that its benefits come from the mechanisms Freud described.

The famous cases have not aged well either. Historians examining the records have found follow-ups that were less successful than published, patients whose disagreement was recorded as resistance, and at least one, the woman Freud called Dora, whose situation as a teenager pressured by an adult family friend looks to modern eyes like a straightforward account of abuse that was interpreted as fantasy. Frank Cioffi and others argued that Freud sometimes knew his interpretive practice was circular and continued anyway.

Specific doctrines have fared badly under testing: dreams as disguised wish fulfilment, the Oedipus complex as a universal developmental stage, penis envy, the notion that repressed traumatic memories are reliably stored and can be excavated intact. Memory research by Elizabeth Loftus and others showed the opposite: suggestive questioning can create confident false memories, a finding with real casualties in the recovered-memory cases of the 1980s and 1990s.

And yet the residue is substantial. Unconscious processing is now mainstream cognitive science, though it looks less like a seething basement and more like ordinary automatic computation. Defence mechanisms, transference, ambivalence and the value of a trained listener survived into practice. Freud was wrong about the machinery and right that the interior life is worth taking seriously.

By the time the arguments got going in earnest, their subject was an old man with jaw cancer in a city that had just voted itself into the Reich.