Deborah Orr: Is it fair to put mothers in jail on the basis of glib formulas and untested theories?

There were feverish predictions of a deluge of miscarriages of justice. The scale of legal failure was flagged up as being unprecedented in the history of the British system. Since first Sally Clark and then Angela Cannings had their convictions overturned, the final results of a review of 297 other baby-death cases have been awaited avidly.

Yesterday the attorney general, Lord Goldsmith, delivered those last findings. Only three cases among 88 convictions for shaken baby syndrome would be referred to the Criminal Cases Review Commission. For all those campaigners convinced that the diagnosis of shaken baby syndrome is made too widely, this news will be greeted with scepticism and bitterness.

Already it is being suggested that the review is flawed, with no pathologist asked to help with the interpretation of complex medical reports, and only the clinical evidence presented in court examined. If these accusations are true, then there is cause for concern. It is argued by some observers that the paediatrician Sir Roy Meadow and his colleague David Southall have been scapegoated by caring professions unwilling to examine their wider assumptions. These findings will do little to dispel such worries.

It's fair to say, I think, that these is fairly widespread confusion about what has happened in these cases, and exactly what contribution was made by Meadow and Southall. Meadow came up with the theory of Munchausen's syndrome by proxy, which offered an explanation of why parents seemingly without motive harmed their children (to get attention for themselves). Southall gained graphic evidence of such a syndrome when he covertly filmed parents whose babies were in hospital actually conducting the behaviour outlined by Meadow.

The two men offered their theories and findings to an audience eager to accept them. In his defence of Sir Roy Meadow (who was recently struck off by the General Medical Council), Dr Michael Fitzpatrick points to the Cleveland child sex abuse scandal, the earnest debate about the prevalence of satanic abuse in Britain, and the explosion of claims of "false memory syndrome", as other indicators of a widespread willingness to believe that the most awful crimes were hidden in the family unit. This view, once considered as a subversive attack on the sanctity of the nuclear family, is now mainstream enough to be crushingly banal, as viewers of Desperate Housewives can clearly see for themselves.

The work of the two men, then, contributed to and benefited from a widespread climate in which it was increasingly being assumed that parents, far from being the most passionate advocates for their children on the planet, were pretty likely to be doing them harm. Meadow's partial subscription to this idea was turned into a glib formula in the case of multiple cot deaths, when he began telling the courts that one cot death in a family was a tragedy, two was grounds for suspicion and three, unless proved otherwise, was murder. In fact, as her grandmother testified in the Trupti Patel case, the real tragedy is that multiple cot death can run in families.

This, it seems, is what occurred in the Sally Clark, Angela Cannings and Donna Anthony cases, among others, with the opinions of Meadow treated as expert evidence when in fact they were far from being such. His "formula" was the "smoking gun" that convicted the women. But in shaken baby syndrome, there is no simple "smoking gun". Instead there is merely a symptomatic tick list that contends that if certain injuries (mainly subdural haematomas and retinal haemorrages) are present, then shaking has occurred. Since, obviously, no babies have been shaken under laboratory conditions in order to prove that these assumptions are sound, there is no benchmark by which such assertions can be measured.

Rather, as campaigners point out, the presence of these injuries, combined with the wider assumption that parents commonly abuse their children, tends to create a climate in which over-diagnosis is almost certain. The worry is that without a straightforward alternative explanation - as was provided by Trupti Patel in response to Meadows' assertion about multiple sudden infant deaths - shaken baby diagnoses are standing simply because there is no other explanation.

After the Angela Cannings case, the judge wisely ruled that expert medical witnesses should stick to offering facts rather than opinions. This was clearly a refutation of Meadows' "law" about more than one SID in a family being suspicious. But the awful truth is that unless there has been a witness to the shaking, or a credible admission of shaking (parents are often told that they will be more likely to see their children again if they admit to abuse), then all diagnoses of shaken baby syndrome are just that - an expert opinion. The opinion is based on medical observation going back to the 1940s, it is true. But opinion it remains.

Anyway, beyond these worries about allowing an unproven cluster of symptoms to dictate almost irrefutably the existence of a very particular instance of physical abuse, there are other worries that have not been addressed at all by Goldsmith's review. In one of the three cases that have been selected for reassessment, the carer has been convicted of murder. I find it amazing that this man was not only convicted on such specious evidence, but was convicted not of manslaughter (bad enough) but of murder - with mandatory life.

Even if one accepts that these symptoms are cast-iron evidence that shaking occurred (which they aren't), then how can they be evidence too that shaking occurred with the intention of causing death? This latter, in combination with all we know about the stresses and strains on the carers for small babies, is a highly enthusiastic adoption of the worst-case scenario, just as so many of the other "epidemics" of child abuse we saw in the last century proved to be.

The ghastly truth is that this supposition can only be attributed to the general climate of fear and loathing around child abuse, and the assumption that families - the more respectable the more sinister - almost exist to hide abuse. Here is a comment on the subject from Lucy Thorpe, an NSPCC policy advisor: "Nobody wants to see parents prosecuted or convicted for a crime they did not commit, or guilty parents getting away with murder." Surely, this issue cannot be as black-and-white as that?

Poor Donna Anthony, when she was convicted of killing her babies, was described in the media as "evil". Now, it turns out, she was not. Perhaps if society was less keen to pillory and condemn those adjudged to be bad parents, and more keen to understand and assist them, then the injuries babies can sustain when their parents lose control would not be such a matter for supposition and guesswork.

d.orr@independent.co.uk