Donna Anthony spent eight years in jail, convicted of murdering her infant children within a year of each other. She blamed it on 'cot death' but paediatrician Professor Sir Roy Meadow testified that the chance of that happening twice in one family was so negligible as to make it unbelievable.

Now Donna has been freed on appeal, after other high-profile cases saw Sir Roy's evidence overturned when the courts re-visited them. New studies have pointed to a possible genetic link in such deaths, making repeated occurrences with a family less unlikely.

His 'evidence' was that the mothers suffered from a condition he called Munchausen's Syndrome By Proxy, in which parents harmed their children to bring attention to themselves. It rather calls to mind the Cleveland child-abuse cases of 1987, where two consultant paediatricians took case upon case as supporting fashionable new tests of widespread abuse. Again, the justice system overturned it all -- though after much heartache caused to parents who had their children taken from them by the authorities.

Juries may need some help to evaluate 'scientific' evidence, especially where social science is presented as if it were as objective as physical science. We loosely accept science as 'fact' -- but the reality is that science is a process of sifting through different theories. Experts promote their own pet theories, and hold on to them possessively, even after they have been well exploded. In social science, where interpretation of behaviour is often a factor, this can be intensified.

The philosopher Sir Karl Popper quoted a conversation he had as a young man with the great psychologist Adler, in which the latter identified an inferiority complex from childhood in a patient he had not even seen. "How can you be so sure?" asked Popper. "I know it from my thousand-fold experience," replied the great man. "Which I suppose," said Popper, "is now a thousand-and-one-fold."

Before we start removing children from their parents and sending people to jail, we ought to be assured that we are not dealing with case number one-thousand-and-one in the self-justification campaign of some supposed expert.