Popular Mechanics has a terrific cover story this month on the crumbling integrity of forensic science. Here's a taste:

The scientific method is instrumental to our understanding of the physical world. To scientists, the process is sacrosanct: Research your topic, generate a hypothesis, test the hypothesis, analyze your data and then publish the results for peer review. Forensic science, however, was not developed by scientists. It was created by cops—often guided by little more than common sense—looking for reliable ways to match patterns from clues with evidence tied to suspects. What research has been done understandably focuses on finding new techniques for putting criminals in jail. 

In the academic community the legal sciences get a comparative trickle of federal funding. In 2007, the National Institute of Justice awarded 21 grants for forensic research (excluding DNA) totaling $6.6 million; the National Institutes of Health awarded 37,275 grants totaling $15 billion. And without a wealth of statistically defensible research to back up their evidence, forensic examiners generally rely upon their own intuition and the experience of their colleagues. “You can’t take a few case studies and say, ‘Oh, it worked on these people; it must be reliable,’” says Karen Kafadar, an Indiana University statistics professor and a member of the NAS committee. “That is hardly a placebo-controlled, double-blind randomized trial.” 

The article includes a skeptical look at four common forensic specialties, including fingerprint analysis, ballistic evidence, trace evidence, and biological evidence, and explains how none are as certain as they're often portrayed in the courtroom.

For more on this, be sure to check out the paper Roger Koppl wrote for the Reason Foundation on how to introduce real scientific rigor to the forensic process, or the piece on the same topic that Koppl and I co-wrote for Slate.