Texas Sharpshooter Fallacy

I first spotted that the assumption that McKie was lying looked like a case of the Texas Sharpshooter Fallacy. This is a common misunderstanding of probability illustrated by a gunslinger taking a quick shot at the side of a barn. He then paints target rings round the bullet hole and claims to be a sharpshooter.

 

In normal fingerprint work the police drawing a cordon round a crime scene is painting the target and they do this before firing the shot (the fingerprint analysis). If fingerprinting has a very small error rate it would be very unlikely (but not completely impossible) for a misidentification to occur within any set of only a few hundred latent fingerprints that have a connection to a crime.  But as long as the error rate is not zero some misidentifications will occur in the millions of identifications produced by the whole of fingerprinting activity over a period of years.  These misidentifications will turn up one day and if we assume that someone is lying when they deny depositing a fingerprint, and it is not connected with any known crime, then we are painting the target after firing the shot.

 

The core point about this fallacy is if something can happen either by chance or by an assignable cause, and you want to prove that an example of it didn’t just happen by chance, you cannot pick the case because it comes to your attention.  For example if you first notice that a village has an exceptionally high cluster of cancers and then you suggest that pollution from a nearby factory is the cause, you must find other proof that the factory is causing the cancers.  But if you start by having an idea that the type of pollution emitted from the factory causes cancers then you carry out a survey of the local village and find an exceptionally high rate of the disease, you can use the low probability that this cluster could have happened by chance as part of your proof that the factory is the cause. In fingerprinting most identifications are not denied so one that is denied attracts attention. A correct identification that is denied is caused by the criminal being in the crime scene and lying about it.  Many misidentifications will be random in nature, wrongly identifying someone by chance who cannot be linked to the crime under investigation.

 

A more common definition of the fallacy is that you cannot use the same information to both construct and test the same hypothesis.  In a normal criminal case the hypothesis is something like "the criminal has deposited a fingerprint at the crime scene and will deny it".  The hypothesis is constructed before the fingerprint work starts from the knowledge that a crime has occurred. The crime itself is not part of the hypothesis because we know it has happened.  In the McKie case, the hypothesis is "Shirley McKie entered Marion Ross' house in breach of police procedures and deposited a fingerprint". This is constructed entirely from the fingerprint match between McKie and Y7 and includes the "crime" of entering the murder scene.  It is unusual to have the crime as part of the hypothesis in fingerprinting. When David Asbury won his appeal the question arises, "If Asbury did not kill Marion Ross then who did"?. When Shirley McKie was found not guilty of perjury nobody asked the question "If it wasn't McKie who entered the house without permission then who did"? 

 

Back to Shirley McKie case – logical and statistical aspects

 

Back to Fingerprinting Innocentville