Friday, March 30, 2007

Weapon Focus Under Attack?

Perhaps the only puzzling part of the LeGrand opinion (PDF) was the section dealing with weapon focus -- that is, the phenomenon in which the presence of a weapon during the commission of a crime negatively affects the eyewitness's ability to later identify the perpetrator. Despite finding that the trial court should have allowed the defense expert to testify on several areas of eyewitness research, the LeGrand court "agree[d with the trial judge] that there was insufficient evidence [of the reliability of weapon focus] to confirm that the principles expounded by defendant’s expert witness are generally accepted by the relevant scientific community."

Is this a signal that the New York Court of Appeals did not believe sufficient scientific consensus on the reliability of weapon focus currently exists? Or was it simply a statement, unique to record in the LeGrand case, that the defense there did not present sufficient evidence there to meet its foundational burden under United States v. Frye?

I don't think the Court could reasonably have meant the former -- i.e., that there is currently not enough expert consensus about the reliability of weapon focus. Or at least, if that's what the Court meant, it is demonstrably wrong. The most recent (2001) Kassin survey of experts in the field showed that 87% of surveyed experts believe the research on weapon focus is now sufficiently reliable to form the basis of testimony. This percentage increased 30% over the 1989 study, in which only 57% of experts found the weapon focus phenomenon sufficiently documented. Kassin, et al, On the General Acceptance of Eyewitness Testimony Research: A New Survey of Experts, American Psychologist at 405 (May 2001).

Why the big increase from 1989 to 2001? The main reason was the publication, in 1992, of Dr. Nancy Steblay's meta-analytic review of the weapon focus effect. Nancy M. Steblay, A Meta-Analytic Review of the Weapon Focus Effect, 16 Law and Human Behavior 413-424 (1992). That review conducted a statistical analysis of the data from the 12 studies that had been done to that date on weapon focus (accounting for 19 individual tests of the hypothesis), to determine whether those results, as a whole, showed a weapon focus effect. The study demonstrated that a statistically significant weapon focus effect does in fact exist, and that the effect was "more pronounced in research scenarios that appear as real life to the subject." The study also showed an increase in the effect when the viewed object was clearly a threatening object, like a gun, even if the entire incident was being viewed on video.

To be sure, in the 19 tests reviewed, only 6 showed a statistically significant weapon focus effect (while the other 13 did not see an effect). But a meta-analysis does not simply count up the studies on the issue and stop there; if it did, a study of 5 witnesses would have equal weight with a study of 500 witnesses. Rather, the meta-analysis (and I'm simplifying here, in part because I'm not a statistician) reviews the entirety of all the witnesses and figures out whether any effect is statistically significant. See, e.g., R. Rosenthal, Combining results of independent studies, Psychological Bulletin, 85, 185-193 (1978); R. Rosenthal, Meta-analytic procedures for social research (1984). The analysis also controls for the possibility that one lab might be skewing the results so that if only one big study showed the phenomenon, that alone would not signify a weapon focus effect.

Conducted in this way, the meta-analysis showed a statistically significant weapon focus effect. It also showed that some of the studies in which the effect was reduced or not seen either didn't involve a weapon at all (some involved, for example, a person carrying a bloody meat cleaver but not using it as a weapon) and several others did not involve any sort of crime. Moreover, one of the experiments in which a small effect was seen involved a comparison between two groups in which the culprit had a weapon -- in one the weapon was used and in the other it was concealed; in such a scenario it is easy to see why any weapon focus effect would seem reduced because the control group itself may also have experienced a weapon focus effect as witnesses may well have been focused on the concealed weapon.

Following the meta-analysis, and several other experiments in the 1990's confirming these findings, experts moved from relatively skeptical to substantial consensus on the question of weapon focus. That is where experts are today, which is why LeGrand discussion cannot apply to the phenomenon generally, but can only reasonably be read to mean that the trial court there had not erred in excluding testimony because the defense had failed to demonstrate consensus on that record.

Presumably, when fully confronted with the facts on weapon focus, New York courts will do what others, like the Third Circuit, have done: Let the jury take the science on weapon focus into account when deciding the reliability of an identification.

No comments: