The "Baby Doe" Investigation Showcases Forensic Science at Its Purest

The story of "Baby Doe"--the little girl whose body was found along a Boston shoreline in June--has captured the nation's attention. So far, we know virtually nothing about what happened. We don't know who she was, what happened to her, or who was responsible.

Forensic analysts from multiple disciplines are working on the Baby Doe case in an effort to provide leads as to what may have happened. One technician has generated an image of what Baby Doe's face probably looked like, others have assessed her body for signs of trauma, and still others are doing toxicological tests for the presence of drugs or poison. And although the forensic field as a whole has come under heavy criticism lately, the type of work being done at this stage of the Baby Doe case is likely to be uncontroversial and universally appreciated. The reasons for this lack of controversy may provide some guidance on how to improve forensic science as a whole.

One key reason why people will assume the best from these analysts is that their work is being done at a point in the investigation at which there is no apparent reason to bias the results in any direction. No suspect has been identified, and indeed it's not yet clear that Baby Doe's death itself was criminal (there are certainly strong indications of that, but there have been cases in which children's bodies have been improperly disposed of following non-homicidal deaths). So there's no reason for the analysis to be anything other than a true "fact-finding" mission, directed solely towards finding out what actually happened.

What about the more typical case, in which the forensic work takes place after a suspect has been identified? In these cases, for example, an analyst might assess whether a shoe print from the crime scene "matches" a shoe found at the suspect's house. This, obviously, creates an entirely new set of issues.

Although the specific framework varies between jurisdictions, forensic analysts are often employed by the government and work closely with police officers and prosecutors. When an investigator brings a forensic question to an analyst, there's obviously a risk that the analyst will tailor his conclusions to fit the case theory. This risk becomes more problematic the more subjective the particular field is. DNA analysis, for example, is essentially objective, with little opportunity for the analyst to "nudge" the ultimate result one way or the other. A DNA profile from a crime scene either will or will not match that from a suspect. There may be error from contamination, or even outright fraud if the investigator or analyst is so inclined, but there's virtually no subjectivity involved.

Other disciplines, however, are much more subjective. For example, in the shoe print technique mentioned above, there's no real set of objective criteria for analysts to follow in determining how much of an impression is necessary for a comparison, or how many and what types of features must be considered before finding a "match." (I should note that the use of the term "match" is often inappropriate in the forensic setting; it's used here for simplicity and to avoid wading into the terminology controversy.)

With that degree of flexibility, it can be easy for an analyst--whether hired by the prosecution or the defense--to reach the "right" conclusion. This problem isn't limited to deliberate fraud, though of course that does exist. Given the subjectivity of these methods and the way human beings make decisions, even scrupulously honest analysts can be affected, even if only subconsciously, by knowing what the results "should" be.

There may be ways to make forensic science across the board more similar to the way it's being used in the Baby Doe case. For example, I've always thought that forensic analysis is best done "blind." In the shoe print example, the analyst should be presented with the impression from the scene and the suspect's shoe, along with any additional technical information (such as soil composition) necessary to the analysis. But nothing else--not, for example, information that the suspect has allegedly confessed, or that there was a bloodstain on his glove. (I've had a good faith disagreement on this issue with a leading expert in one of these fields. He believes that that type of additional information can be useful to provide context for the analyst's work; I think it's precisely that type of "context" that can lead to biased results.)

Of course, blind testing would only do so much; an analyst could still assume, from what's being submitted, that the police have someone in mind and that a positive forensic result would support their theory. To counter the resulting bias in favor of finding a "match," I've also suggested that analysts be provided, on a frequent but unpredictable basis, with "dummy" requests. So, for example, about ten percent of the time an analyst could be given a shoe print impression along with a shoe that, although similar, didn't actually make that impression. If the analyst tends to err by finding "matches" in these situations, that would call the quality of his work into question--and the resulting statistics could be made available to both prosecutors and defense attorneys in the analyst's cases. Additionally, knowing that any given request might be a "dummy" might cause analysts to exercise more caution before claiming to find matches in doubtful cases.

These and other steps might bring courtroom forensic work more closely in line with the more neutral work being done in the Baby Doe investigation. In the meantime, all of our thoughts and support will continue to be with those analysts as they search for answers in that tragic case.