Azaria's mother, Lindy, and father, Michael, were convicted by the Australian criminal justice system, Lindy of murder and Michael of being an accessory after the fact - in effect found guilty of having helped Lindy conceal the death of the child and her involvement in killing it, after he discovered her involvement. The trial was apparently exemplary; long after a royal commission found himself unable to criticise the Crown Prosecutor, Ian Barker, QC, for overstating the evidence he had, or of concealing material pointing to their innocence. A full bench of the Federal Court had carefully considered the Crown case and refused to interfere with the verdict, even though the trial judge was privately expressing amazement at the verdict. The High Court refused to interfere with the decision, if on new and somewhat priggish grounds, that justice did not enter the matter if no error of law could be demonstrated. Beyond the trials had been two intensive inquests, and, between them, a forensic investigation, with top scientists from three continents, brought on by NT detectives smarting under criticism of the quality of their work by the first coroner, a former detective.

The case against them was circumstantial but impressive. The forensic evidence was particularly persuasive. The jury was convinced beyond any reasonable doubt that Lindy had sat in the front seat of her car with a pair of scissors, stabbed the child, held it until it died, then concealed it, probably in the boot of the car while setting off a hue and cry about a dingo in the family tent at an Ayers Rock camping ground.

No one has ever suggested that a police or forensic witness gave consciously false evidence. All were honest and decent people, even if some were not as clever and experienced as they thought they were. Lindy Chamberlain was not ''verballed,'' nor was evidence implicating her ''planted''. It is important to remember that the law - judges and lawyers - failed Lindy as badly as the police and scientists.

The essential trouble was that both the investigating police, and the principal scientific witnesses were victims of what some called theory-dependence. They believed - in some cases just ''knew'' in their gut - that the Chamberlains were guilty and the dingo story cock and bull. They looked for evidence that tended to confirm their theories, and tended to ignore or explain away evidence which did not.

The biggest shocker, as it turned out, was a forensic biologist, Joy Kuhl, who, testing the Chamberlain car thought she had detected foetal haemoglobin in stains on the front wall of the front passenger side. If Azaria had been murdered in the front of the car, these ''bloodstains'' were just where one might have expected them to be. On her evidence, as well as an assertion by a scientist that tears on the baby's jumper were scissor cuts, not bite marks, Lindy was convicted.

What Joy Kuhl ignored was that the chemical test she had used to detect foetal haemoglobin levels also detected the presence of iron. There was no blood on the car's panels, but lots of paint, dust and ore. Painstaking examination of Kuhl's findings found a repeated tendency to overconfidently report evidence seeming to implicate Lindy, and a tendency to discount material which did not. She had ceased to be an objective expert, and became an advocate for a theory we now know positively to be quite wrong.

Could it happen today? Of course. Some might think that major advances in forensic science - and in the professionalism of police investigators and forensic scientists would now prevent such mistakes. Some might think modern science - say of DNA - is fairly settled, with jurors able to be told, for example, that the chance of a person's having the same DNA by coincidence is one in, say, 80 billion.

What jurors are, however, only rarely told is that the sloppiness of many murder investigations is such that the risk of cross contamination of evidence samples taken from scenes of crime may be as high, on average, as one in 50, that is, a very real possibility.

When such ''evidence'' leads investigators away from an initially open mind to a strong theory of a particular person's guilt, the blindness can be almost comical, if it was not for the disastrous consequences. We have in recent times seen a number of Western Australian convictions overturned, not on the basis of the emergence of significant doubts, but the discovery of evidence which proved the defendant innocent. In some cases, there was confessional material - or at least words spoken which police interpreted as, and urged upon jurors, as admissions. Often, as with Chamberlain, and significant miscarriages in Queensland, WA and NSW, the cases had been repeatedly before the courts, convictions being rubber stamped by judges seemingly unable to imagine that police, or experts, could get it all wrong.

A new technical book, Crucial errors in murder investigations, by Ted Duhs of Bond University, shows how investigators can make honest, but fatal, mistakes. His cases are investigations mucked up by Queensland detectives, although there is cross reference to famous disasters elsewhere, including to the Chamberlain case. (Professor Barry Boettcher, who exposed a good many of the fundamental scientific errors of the Chamberlain case, has written a foreword.)

Most stem from the human error of theory-dependence - focusing on evidence that supports a theory already formed, ignoring evidence that doesn't, making connections that send investigations down certain paths, holding on to theories even after they have been falsified. Duhs shows in case after case how this caused vital clues to be ignored, and investigations marshalled to prove things which were wrong. Triumphant detectives are rarely given to self-doubt.

It should not be thought that concerns about miscarriages of justice come only from an anti-police brigade. Here in Australia a number of forensic scientists have expressed concerns about the quality of evidence sometimes going before courts, the use of bad procedures, failures of chains of custody, incapacity to replicate results, the use of samples too small to guarantee results, sloppy systems of avoiding contamination and cross contamination, the use of impure chemicals, and the use of equipment which had not been validated or which were known for inaccurate results. As problematic, in some jurisdictions has been an association so close between forensic scientist and investigating police, as well as a pressure of overwork, that presents real risks of undermining the detachment that experts pretend. The scientific method, of course, is ever anxious to test a theory, not least with evidence that challenges it, rather than focusing only whittling facts until they fit the theory. The Chamberlain case is a warning about legal lynchings, and a reminder never to think a case is closed.

Jack Waterford is Editor-at-large.

Crucial errors in murder investigations, by Ted Duhs, Bond University Press, 212pp $39.95.