There's a worrying problem with the way neuroscience made the headlines at the end of last month. Scientists at the University of California, Berkeley, had looked at the electrical signals generated in the brains of patients listening to speech. From those signals, they were able to infer some of the words being heard. Such work may one day help us listen to the thoughts of comatose or locked-in patients. So what's the problem? It will add to the unfounded faith in scientists' ability to read minds - a ticking time bomb for the legal system.
Research has shown that people are much more likely to believe a statement prefixed with the phrase "Brain scans indicate . . ." That has been true even when the statements were obviously scientifically flawed. It was also true for neuroscience students who should have known much better.
Things are made even worse with functional magnetic resonance imaging (fMRI) scans, pretty pictures that look like they contain much more data than they actually do. These differentiate the relative amounts of blood flow in the brain. The differences are generally displayed on "false-colour" pictures: like a Tube map, the colours are arbitrary and are used just for contrast; they are not an absolute measure of anything. But they look ever so significant.
That would explain why, in 2009, a defence team in the US tried to use fMRI scans to convince a jury to reduce the sentence of a convicted rapist and murderer. Brian Dugan had confessed to the 1983 abduction, rape and murder of a ten-year-old, Jeanine Nicarico. He was already serving a life sentence for two other murders and faced the death penalty for this crime. With the aid of Kent Kiehl, an expert on psychopathy at the University of New Mexico, the defence team showed the jury scientific evidence derived from fMRI scans. The evidence, according to the defence, showed Dugan to be a psychopath.
The jury eventually rejected the evidence and sentenced him to death. But the jurors' hesitation - they stayed out for five hours - provided a moral victory for the defence and gave grounds for appeal against the sentence.
It's worth pointing out to anyone who might find himself on a jury that whether you can infer anything solid from one person's fMRI scan is so far entirely unproven. All tests of fMRI have been carried out on groups of people; the indicative calibration maps of relative blood flow are an average over many individuals.
What's more, all fMRI calibration studies for brain-reading so far involve hugely artificial situations - students being paid to sit in scanners and tell banal lies, for example. Applying inferences from that kind of research to real-world situations is simply not justifiable. The technology can't distinguish between wishes or desires and what happened. If you imagine you have done something - whether or not you did it - the scan will see it: it cannot distinguish between real and false memories.
The biggest responsibility for the appliance of science here will lie not with jurors but with the neuroscientists who get involved in legal proceedings. They will face pressure to overstate the permissible conclusions.
Having carried out the fMRI scans on Dugan, Kiehl told the BBC that Dugan "struggles to try to understand why people even care about what he did". However, the defence attorney handling the appeal told the Chicago Tribune that Dugan had "expressed his sorrow and his remorse for what he had done".
The truth is out there but not everybody wants to expose it - and endorsing half-baked science won't help.
Michael Brooks's "Free Radicals: the Secret Anarchy of Science" is published by Profile Books (£12.99 paperback)