This is a wikipedia article on the subject:
Is there any validity to this? I find my skeptic meter going on high alert but am willing to be educated.
This method is a kind of lie detector which has been proposed as early as the 80’s (see this wiki article on P 300)
There is some valid theoretical background in here, (yes, you can find signals in the brain which correlates with wether or not you have seen something already), but as always, the problem is how efficient the method is. Without a minimal degree of accuracy and reliability, the method is simply useless
This article from 1988 by Farewell describes a brain-machine interface method based on P300, used to spell characters on a computer with results (0.2 bits/sec) which are quite lower than what one would obtain now (rather around 1-2 bits/sec). Therefore, although there are indeed some old results published on the use of P300, there is nothing recent and nothing else than classical brain machine interface.
The wikipedia article mentions somewhere a 99% accuracy figure at the lab and in the field. That sound ridiculous (btw, I do research in neuroscience, near to the field of brain-machine interfaces, so I am familiar with the kind of figures you can expect in the field), and is not published.
Ultimately, this method is developped by a certain Dr Farewell. A little search on pubmed shows that he does have some article, but either on earlier works (before 1993) or on journals which are not in the neuroscience field (2001). Not-peer-reviewed scientific claims are simply worthless (some publications mentioned on the wiki page are simply posters abstracts and don’t count).
So, while there may be some background in this method, there is no evidence that it has enough efficiency and reliability to be usefull. Just as the classical lie detectors, in fact.
I don’t know if I’d really classify it as a lie detector. It seems to just determine if something is recognized. So it’s only useful for detecting lies that have to do with recognizing things. And whether something is recognized can be useful in ways that don’t have to do with lying per se.
I’m wondering how it managed to get approved for law enforcement?
I also wonder what qualifies as recognition. There’s obviously different levels of recognition for all stimuli…