This is a bit afield of the primaries, but forensics expert Roger Koppl has a fascinating (if alarming) piece in Forbes this week arguing that forensic evidence isn't as foolproof as shows like CSI might suggest:
If only forensics were that reliable. Instead, to judge by the most comprehensive study on the reliability of forensic evidence to date, the error rate is more than 10% in five categories of analysis, including fiber, paint and body fluids. (Meaning: When the expert says specimen X matches source Y, there's a 10% probability he's wrong.) DNA and fingerprints are more reliable but still not foolproof. The 1995 study, in the Journal of Forensic Sciences, looked at proficiency tests labs take to see whether their work is sound.
More recent studies have also shown problems. Though a 2005 study in the Journal of Criminal Law & Criminology suggests a fingerprint false-positive rate a bit below 1%, a widely read 2006 experiment shows an alarming 4% false-positive rate.
Radley Balko offers up some further commentary. Among other points, it turns out that forensic experts can be biased simply by talking to prosecutors and police officers beforehand (which, as I understand it, they usually need to do). Koppl proposes a few fixes, including this: "Each jurisdiction should include several competing labs" to double-check each other's work. The extra cost would quickly pay for itself by reducing the number of false convictions. Balko, for his part, emphasizes the idea that the defense should have guaranteed access to its own forensic experts, which is admittedly a more radical step.
P.S. Ooh, can't believe I almost overlooked this: Jeffrey Toobin had a terrific piece in The New Yorker last May exploring the various complexities in forensic science. And, on a weirder note, Target of all places now runs its very own crime lab, which was originally intended to sniff out shoplifters, but now comes to the aid of local law enforcement in a wider variety of cases.