If you were not able to watch PBS Nova’s presentation of “Forensics on Trial” last night, you can stream it here.
To sum up my overall reaction, I was quite disappointed. They failed to educate the audience about the actual underlying scientific shortcomings of forensics, and for a good portion of the show, I felt like I was watching an episode of CSI in which the investigator sits down at a computer and some magical software application solves the crime.
They did say that there needs to be more “science” in “forensic science”, but did not address what or how. They did say that most all forensics is subject to an examiner’s interpretation of the evidence, and that it can be wrong. Beyond that, I found little value in the program. My impression is that it was clearly produced more to entertain than to educate.
The program addressed three areas of current forensics – fingerprints, blood spatter, and bite marks. Let’s look at some specifics.
Fingerprints. The Brandon Mayfield case, which was presented, is the CLASSIC example of latent print identification gone haywire. They did not discuss the issues with computer matching of partial prints, quality of original print (briefly), or all the research that’s been done to show that examiner competence and bias plays a huge role. The new “columnar thin film” technology they featured is quite interesting, but latent matching still comes down to the quality of the original latent print, and the capabilities of the examiner.
Blood Spatter. Blood spatter evidence, like all forensics is not useless. It has a role to play. It can determine things like direction of blow or injury, possibly suggest type of weapon, determine location of attack, or whether a body has been moved. But it can’t do much more than that. The “experiments” that were shown were the typical kind of demonstrations that blood spatter experts pass off as science. They are totally lacking in the control of independent variables in the process – like blood viscosity, and they use surrogates of questionable value. Bashing a dead pig’s head with a baseball bat might give you some information about what happens when you bash a dead pig’s head, but to suggest this is a substitute for a human head is doubtful. The bone, tissue, and fluid structures are very different, and dead things have no blood pressure or muscle tone, and have lost many of the fluids that were present during life.
Bite Marks. The case that was used to discuss bite marks is very interesting. One logically supportable use for bite mark evidence is to exclude a suspect. In this case, the suspect should have been excluded by the examiner, but his interpretation of the evidence caused him to make a false identification, which resulted in a wrongful conviction. Bite mark identification is fraught with issues, and even veteran forensic odontologists are beginning to disavow the discipline. This was not mentioned.
The “futures” stuff they talked about were really quite interesting, and will certainly have some future value, but I question their broad range utility. The “virtual autopsy” that was shown does have the advantage of being able to detect trapped gasses, and also provides the ability to go back at any time to view the body “as scanned”. However, today’s MRI resolution is about 1 mm, which would not be sufficient for very fine features which may require microscopic examination, and the virtual autopsy cannot present tissue color, texture, resiliency, or odor. The laser-generated 3D crime scene reconstruction will also have value in being able to go back at any time to review the scene “as scanned”. But as for the staged example they showed of catching a suspect in a lie, any human observer standing in the same spot could have observed the very same problem with his story.
Most importantly, what the program did not address was the non-scientific underpinnings of forensics – the fact that these disciplines evolved from logically flawed inductive reasoing. ”I’ve seen 100 roses, and they’re all red. Therefore, all roses are red.” Or, “I’ve never seen that before, so this must be unique.” The basic, fundamental question that was not asked is, “Show me the data from which I can compute a statistically valid probability of occurence.” If you ask a fingerprint examiner what the statistical probability is of two prints matching, he can’t tell you. Likewise for bite marks, or blood spatter, or hair & fiber, or tool marks, or ballistics. Why can’t they tell you? Because the data doesn’t exist. Therein lies the fundamental problem with the integrity and dependability of forensics.