Statistics and Misidentifications - The weeks Detail

[Outsider]

Michele Triplett:

Some of the conversations in this thread are interesting but I don’t see the relevance (such as if a subject was wearing gloves). I’ve been involved in cases where the subjects wore gloves but still left a print (without removing his gloves). This may seem unlikely to those outside our discipline but once shown how this works it’s easy to understand.

I am intrigued by this. Are we talking about the thin latex gloves used in laboratories or normal gloves that keep hands warm. How does it work?

[Daktari]

Shirley McKie can tell you all about latex gloves.
The first time she was done for leaving her fingerprints where they should not be was the 'Dead Baby' case. She claimed she only handled the bag wearing gloves when she became production officer in the case.
She claimed on the notorious Frontline Scotland programme, hosted by the Shelly Joffre, daughter of former SNP MSP Kay Ullrich, that Strathclyde Police had later circulated a memo because of this. James MacKay dismisses this as fanciful thinking. Maybe Iain McKie just forgot to tell MacKay what to say on this one.

[Outsider]

I think latex is a special case. It is non-porous with a very smooth surface. In offset printing machines latex is used to transfer the oily ink from the plate to the paper with very fine precision. It is not so difficult to imagine that a thin latex glove can transfer or impress a fingerprint or palm print in some way.

But what about normal gloves? I would have thought that the likelihood of depositing a fingerprint while wearing thick gloves is much lower than when gloves are not worn, even if it is theoretically possible. I was wondering if this is the sort of case where a low “prior probability” should be ringing alarm bells somewhere in the fingerprint department.

[Taggart]

Steve,

I hope the following link will help you understand a little more clearly

http://shirleymckie.myfastforum.org/_qu ... out15.html

[Outsider]

Hi Taggart,

In the linked message you say:

This does not happen often because the required conditions of handling with pressure and the rubbing against a fatty surface do not often occur.

And this is with latex gloves. I am interested in the cases that Michele Triplett cited earlier in the thread. Are they all latex examples? If not, how does the fingerprint transfer?

The reason I am interested in the gloves is because this is the sort of case where I would assume that the prior probability (prior to the fingerprint analysis) of the criminal leaving an identifiable print at the crime scene is very low. So this is exactly the sort of case were I think alarm bells should be ringing somewhere in the fingerprint department.

[Taggart]

Steve,

Thanks for your reply, however I cannot claim to be the author of the article I gave the link to. I found the article extremely interesting as I thought you would too.
Hopefully someone else might be able to answer your questions.

thanks

[adroitcaledonian]

L J Steele, you made a pertinent and salient point regarding juries:

"The prosecution's job (and the defense's) is to explain technical evidence at a level the jury can understand. This is difficult as, at least in the US, a juror can't raise his hand and ask a question if he's unclear about a concept. Are you suggesting that fingerprint evidence is too complex to be explained to a typical juror? Or doubting the ability of a juror to weigh conflicting testimony?"

Here is the suggestion of a judge in a prominent Scottish Fingerprint trial:

"She does not have to prove anything, as I’ve said, she remains innocent until that presumption is rebutted.
But the fact remains, that she has told you she is innocent and, if you believe that, that’s the end of the matter and you need take the matter no further. If you are satisfied to that effect simply conclude that she is an honest and straightforward witness and you can come back to me in 5 minutes’ time and tell me that; you needn’t take the matter any further, and you needn’t even start to consider the question of
fingerprints."

[Outsider]

Adroit, I don’t know what message you are replying to. I would like to keep this thread on the topic of statistics and misidentifications.

I am wondering if the explanation that a glove has transferred or impressed a fingerprint is used in some cases because of the assumed infallibility of the fingerprint ID. Maybe the possibility of misidentification is not pursued vigorously enough when the prosecution case has to include an unlikely event (low “prior probability” in the Bayes equation).

Here’s a thought, has anybody re-checked the McKie – “Baby in the Bag” identification.

I will be away for a couple of days so I won’t be able to respond.

[Daktari]

As I understand it the mark left by McKie was of her palm print. It was on a plastic bag, not a moist or fatty substance. She has never denied it was her mark.
Although it has been claimed since that this led to a change in police procedure James MacKay, who investigated the whole matter, says it had nothing to do with McKie.
It goes some way, however, in explaining McKie's denials in the Marion Ross case. Had she admitted being in the house she may have been sacked and lost her pension. Imagine her having to tell Iain that!

[Outsider]

Does anybody know if any research has been done into fingerprint impression or transfer through gloves? The reason I am interested is because if wearing gloves makes it, say, hundreds of times less likely that an identifiable print will be deposited, then Bayes Theorem predicts that the probability of misidentification will be hundreds of times greater (compared with a non-glove case, the likelihood of a good ID of the criminal will be hundreds of times lower but the likelihood of a mis-ID is the same). So investigating glove cases would be fertile ground for finding misidentifications.

The “brass rubbing” explanation is plausible but that would result in a negative image (the valleys between the ridges would leave the fatty deposit). Latex as an ideal medium to hold a fingerprint then transfer it to a surface but that, of course, would leave a mirror image on the surface.

The only thing a Google search has come up with is an index entry for Fingerprint Whorld - B. Comber, "Fingerprints through gloves," Fingerprint Whorld 23 (90), 125 (1997), but the article is not on-line.

[Outsider]

Since nobody could point to any research about fingerprints and gloves I did my own quick test.

I had two different types of thin “lab type” gloves. I pressed my gloved finger against the glass of a clear photographic filter (UV). I found that if I held the filter so that the surface was reflecting a lamp, with a dark object behind the filter, I could see the fingerprints very easily. The glass was cleaned with Eclipse fluid (methanol) between tests. I got best results by rubbing the gloved finger on the side of my nose before pressing it on to the glass. Pictures were hand held with a macro lens:

http://www.pbase.com/stevehorn/image/80449870.jpg

The image on the left was using a thin latex glove which had a slightly rough surface. No matter what I tried, I could not get any ridge detail to transfer to the glass with this glove.

The image on the right was with a thin vinyl glove which had a very smooth surface. This is not a negative impression or anything complex, it is just a straight print of the ridges.

I was expecting that it would be possible, but difficult, to deposit a print through a glove. I would then have made the point that a low “prior probability” of depositing prints while wearing gloves leads to a high “posterior probability” of misidentification after the fingerprint analysis - according to Bayes’ Theorem. That is not what I found. In my test it was all or nothing. One glove could not deposit a fingerprint no matter how hard I tried but another glove would transfer my fingerprint very easily.

[Pat A. Wertheim]

Steve --

Thanks for the photos! Graphic demonstration of how easily fingerprints can be transferred through smooth latex gloves.

[Outsider]

Hi Pat,

What is amazing is that the two gloves look almost exactly the same, they are just normal lab-style gloves. It is not obvious to the naked eye that the surface textures are different.

If they are of any use please feel free to copy the photos (you can save a copy by right-clicking the picture in the web browser).

Here is the smooth one “double gloved”:

http://www.pbase.com/stevehorn/image/80462381.jpg

[Outsider]

As, perhaps, a final thought, it has struck me recently that my statistical observation about the McKie case will seem strange from a fingerprint analysis point of view. From this perspective all misidentifications are the same, an error is made and an innocent person is damaged. For me it is the circumstances of the police allegation that makes it stand out. The police should never have to construct an entire case, including a hypothetical act of wrongdoing, from a fingerprint identification alone - with no independent concrete evidence of anything.

In a normal case, before the fingerprint work starts you have a real crime and you have a good chance (prior probability) that the criminal will have left fingerprints while committing the crime. This means that if an identification is made from a latent print in the crime scene and the person denies having deposited it, the overwhelming likelihood is that it the criminal was correctly identified - even if the the fingerprint department has not used 100% best practise so an error is possible. Take away the ID from the McKie case you have nothing, no crime, no suspicious behaviour, no reason to imagine that the fingerprints of a disobedient police officer might be found, nothing. Bayes’ Theorem says that that if there is a low prior probability of a criminal’s fingerprint being deposited there is a higher probability of misidentification. It is not that the low prior probability causes the ID to be wrong, it is that (unbiased) misidentifications incriminate at random, and in high prior probability situations they are swamped by the good IDs of the criminals.

I think that eyebrows should be raised whenever the police propose that a very unlikely event must have occurred for a fingerprint to have been deposited. For example a criminal depositing an identifiable fingerprint while wearing gloves (unless we know they are the type that impress fingerprints easily), or something very unusual must have happened and there is no independent evidence of it. Misidentifications are also very unlikely events. It is the ratio of the likelihoods of the two events which defines the probability that the identification is good or bad, not the very small chance of an identification being erroneous (chosen at random). In normal circumstances we KNOW that the criminal attended the crime scene, so depositing an identifiable fingerprint in that location would not be a unlikely event.

Accepting that this statistically-based line of thinking might be useful does not require abandoning the principle that a competent fingerprint identification has an error rate of zero (although I think that the fingerprint profession should encourage research into this). When a verified ID requires that a very rare event must have occurred for the fingerprint to have been deposited, this would be a good time to invite an independent out-of-department view of the identification and working practices.

For reference, here are my URLs

Main essay:
http://www.stevehornsc.pwp.blueyonder.co.uk/short.htm

Fingerprinting Innocentville:
http://www.stevehornsc.pwp.blueyonder.c ... tville.htm

Bayes’ Theorem as it applies to the McKie case:
http://www.stevehornsc.pwp.blueyonder.co.uk/bayes.htm

[Outsider]

I thought it would be useful to fully explore my thoughts about the McKie case from a logical, rather than statistical, point of view. To do this I created a simplified example of the Prosecutor’s Fallacy to expose its core logic. The two outwardly similar fingerprint examples which follow illustrate important logical differences.

If you fully understand my perspective and subsequently decide that fingerprinting has little to gain from this line of thinking, then that will be an informed decision. I will remain convinced, however, that a logical and statistical understanding of the McKie case would have led to an early intervention that would have limited the damage.

The message is quite long so I’ll post it in 4 sections.

If you want it all in one page, it is here

http://www.stevehornsc.pwp.blueyonder.co.uk/pf.htm