Thursday, November 03, 2011

Onus on state, local departments to improve eyewitness testimony

The issue of eyewitness identification has been much in the news this week, with the US Supreme Court hearing oral arguments on the subject yesterday and, here in Texas, Sam Houston State University (SHSU) coming out with its draft of a model policy for Texas police departments on how to conduct eyewitness IDs.

Draft TX eyewitness ID policy released
Readers will recall that Texas passed legislation this year mandating law enforcement agencies establish detailed written procedures for eyewitness ID, directing a police training center at SHSU to develop a draft model policy. (In the interest of full disclosure, your correspondent lobbied for the bill at the Legislature on behalf of the Innocence Project of Texas and participated in a working group convened by SHSU on the subject.)

At present, 88% of Texas law enforcement agencies have no written policies on how to conduct lineups. Typically practices differ not only from department to department, but from detective to detective. Only a handful of Texas departments - notably led by Dallas PD - have policies that actually represent best practices developed through the scientific research. Under the statute, departments must adopt their own policies - either using the SHSU model or one of their own choosing - by September of next year.

To their credit, SHSU does tell agencies to use sequential presentation (showing photos one by one) instead of group photo arrays. Though language in the policy which describes "Conducting the photo array" fails to mention sequential presentation, in an appendix the explicit instructions to witnesses say, "The photos will be shown to you one at a time and are not in any particular order." One hopes that provision won't change - and preferably will be strengthened and included in the main body of the policy - after the end of the public comment period.

SHSU advocating the sequential method is good news because, as Patricia Kilday-Hart reported in the Houston Chronicle on Monday, law enforcement pushed hard for a model policy allowing group presentation. However in the latest, most comprehensive study in a long list of research on the topic (described here), group photo arrays were found to generate 50% more errors - 18% picked "fillers" in group arrays compared to just 12% with sequential presentation. It would have been hard for SHSU to claim they were recommending "best practices" if they suggested a method producing half-again as many errors.

Still problematic, though, SHSU suggests that "Because witnesses typically view perpetrators of crimes in three dimensions, a live lineup offering a three dimensional view is preferable." But that view flies directly in the face of research on the topic. In practice, even non-witnesses are able to pick suspects out of live lineups with greater frequency than chance because they pick up facial and body language cues based on fear or psychological stress from the person accused. Photos provide a more neutral medium. Further, particularly in smaller agencies, it's a LOT easier to find appropriate fillers in a photo array than for a live lineup. (See this video describing how photo fillers are compiled electronically by police in the UK.)

Given that the statute told SHSU to base its model policy on "credible field, academic, or laboratory research on eyewitness memory," Grits cannot imagine why they would "prefer" a live lineup method that's more likely to bias results. SHSU's original draft included no such preference, so this provision clearly resulted from lobbying by law-enforcement interests, though I don't understand the reasoning behind it.

At this point, Grits views obstructionist efforts by the police chiefs association and others in law enforcement as strangely at odds with those groups' own self-interest. That's because the Texas Court of Criminal Appeals earlier this fall ruled that a trial court abused its discretion by not allowing expert testimony on the credibility of eyewitness procedures, meaning that if agencies don't follow best practices they'll later find themselves embarrassed in court. If they use a group photo array, for example, police can expect an expert witness to tell the jury the chance of error is 50% higher than if photos were shown one by one. Given that, why not just get it right on the front end?

(If you want to comment on the SHSU draft policy, the email address is

Little help expected from SCOTUS
Coincidentally, the US Supreme Court yesterday heard oral arguments on eyewitness procedures, though SCOTUSBlog reports that "the Court very likely will decide this case, Perry v. New Hampshire (10-8947), without overruling any of the eyewitness case precedents going back to 1957." Instead, a majority on the court appear to view this case as an opportunity to promote the "casting aside of the Due Process Clause as a barrier to unreliable criminal evidence," which is a disturbing and disappointing take on how that case should be decided. At The Crime Report this week, attorney and author James Doyle predicted just that outcome:
It’s true that the criminal justice system is absorbing the science of eyewitness evidence. But it is also true that different components of the system are reacting to the science in different ways and at different speeds.

The police are leading the way; judges are bringing up the rear. The real story that the Perry case may ultimately tell is that the courts are fighting a determined rearguard action aimed at holding off the full mobilization of the lessons of eyewitness science.
In Texas' case, Mr. Doyle perhaps overstates the extent to which "police are leading the way," but clearly he had his finger on the pulse of the Supreme Court. Preventing the "full mobilization of the lessons of eyewitness science" appears to be exactly where SCOTUS is headed.

That said, the question before them was pretty narrow. Noted SCOTUSBlog, "the only issue that the Court had agreed to review was whether the absence of any police manipulation of that witness was sufficient to have allowed the identification to be put before the jury." So it's possible this case may be decided in favor of the state without causing too much damage. But in an era when much common evidence in criminal trials has been called into question - from eyewitness IDs to flawed forensics - it will be unfortunate if SCOTUS undertakes a long-term project of reducing barriers to introducing unreliable evidence. (See the transcript from the case here [pdf]; Grits may have more to say on the subject once I've had a chance to read it in more detail.)

Bottom line: It's a good thing the state of Texas is working to bring science into the police station and the courtroom on this question, because the US Supreme Court seems highly unlikely to do so.

MORE: See coverage of the SCOTUS arguments from the Washington Post, the New York Times, the Los Angeles Times, and Slate. AND MORE: From the American Constitution Society and Amnesty International. The New York Times had a followup editorial arguing that "The Supreme Court should maintain the law’s focus on reliability — and require that courts keep out unreliable identification that is likely to prejudice a verdict."

See related Grits posts:


Adam Poole said...

Regarding sequential versus non-sequential lineups...

I fairly recently cross examined Steven M. Smith on this topic. He is very well published in this area and frankly impressed me very much. He testified that the current research does not show, (at least not clearly) that sequential lineups are less suggestive or otherwise less error prone than a non-sequential photo array. I think this is an area where there are conflicting studies.

For what its worth...

Gritsforbreakfast said...

Adam, most studies have shown sequential result to be superior, but the reason he said that is that in 2006 a study that most people agree has methodological problems (pdf) found higher rates of false IDs with blind sequential than with traditional non-blinded group arrays. That study was not conducted by memory researchers but the general counsel of the Chicago Police Department.

Over the last year, a large, four-department field study (pdf) by the American Juudicature Society that included Austin PD as one of the participants corrected for those methodological errors and found that "fillers" (i.e., non-suspects) were wrongly identified 12% of the time in sequential lineups compared to 18% in photo arrays presented as a group. Those were preliminary findings, but nobody really expects the peer-review process will change the raw data, and once its complete the dispute surrounding that Illinois study will be largely resolved, at least among those who aren't reflexively opposing change just to oppose it.

So what Mr. Smith said is accurate - that the question has been in dispute - but more among law-enforcement than memory researchers, and even that's changing as the Illinois study is refuted.

The issue is that in sequential suspects compare the photo to their own memory, whereas with group arrays they compare them to one another and pick the one that most similar, which is a different mnemonic process.

Gritsforbreakfast said...

Whoops, the last sentence should say "witnesses," not "suspects" compare the photo to their own memory with sequential lineups.