The problems with DNA evidence and contamination have been in the newsof late with the collapse of a rape trial. The defendant had seemingly been identified by a ‘database trawl’ and it came back to him with a 1 in a billion match. The Prosecution threw their hands in after it transpired that the sample from the rape victim may have been mixed up with an unrelated sample.
As some of you may have had the misfortune to know, talk of statistics and law (and DNA evidence in particular) can make me froth at the mouth at times. When I get stuck in to Bayes Theorem I take on the appearance of the crazy man shouting at the pigeons.
There is a general perception that DNA is infallible, or pretty much so (and not just among those people who watch CSI). Also, the way that evidence is presented in Court – a ‘1 in a billion match’ sounds pretty certain and anything that undermines it really just tinkering (which is presumably why it’s done like that). DNA is the only evidence that is done in this way (currently), other forensic evidence is presented on a scale of how much support the scientist finds for a match rather than a specific mathematical formula.
Examples of possible errors and difficulties include:
- Lab or other contamination (conscious)
- Lab contamination (unconscious)
- Mislabelled samples
- Mixtures of different DNA
- Separation of samples
- Difficult statistical analysis with partial matches
- Validity of underling database
I could talk for a long time about the use of statistics in Court (and will do at some point here). But, looking at the question of contamination – all labs are susceptible to error. It’s a fact of life and any scientist worth his or her salt will acknowledge that. A couple of years ago there was a spectacular example of this inGermany where a woman who worked in the factory making the swabs accidently contaminated them with her own DNA. The surprising thing is not that contamination has happened (it’s happened before and will happen again), but that there are not more examples of it.
The potential risk of contamination is recognised by the CPS, but again there is no attempt to quantify the risk, or to warn juries of it. Is it really likely that someone will be wrongly convicted though? What is the potential risk? Well, 2 of the first 200 people exonerated by the ‘Innocence Project’ by DNA testing were convicted of their offences originally on the basis of faulty DNA; that doesn’t inspire confidence.
Being concerned about contamination is not paranoid. All the errors that can be made have been (they are well documented here). There has not been sufficient research to say how common contamination errors are, but the general estimates are between 1 in 100 and 1 in 1,000. The documents linked above set out cases where people have been cleared due to the fact that, for example, they were in prison at the time of the crime, or were aged 4. Closer to home, one of the ‘ Omagh bombers’ was originally identified as a 14 year old English boy (where there was nothing to suggest that he had been toIreland). Given his age, location and lack of any connection with the Real IRA lead to it being categorised as the error it was.
Those people were fortunate in some ways. Due to the circumstances, they were out of the line of fire of police and prosecutors, or they had lawyers who fought on in the face of ‘irrefutable’ scientific evidence . Others may not be so lucky. It is pretty certain that there will be people out there who were convicted on a ‘1 in a billion’ match and whose protestations of innocence are being wrongly ignored.
What can be done about it? It is certainly possible to quantify the probability of this error mathematically (although I’m not aware of any case inEnglandandWaleswhere this has been attempted) through the use of Bayes Theorem (don’t get me started). This would rely on further research being conducted on the rates of contamination which isn’t easy.
Prosecutors wouldn’t like it however as it adds a layer of uncertainty (whether it was quantified or not) and they much prefer the crisp clarity of ‘1 in a billion’ as that seems nigh on impossible for an accused to argue with (even though it doesn’t mean what most people think it does).
The problems with the Court of Appeal’s analysis of statistics, and the general innumeracy of the population is a much wider concern which I may return to when
It is worth remembering the case of Lashley [25/02/2000] where the Court of Appeal concluded that no-one should be convicted solely on the basis of a DNA match. This is an ‘orphan’ case, one that’s not been referred to, but is still valid.