Everyone knows how a police lineup works — it’s an obligatory scene in every TV Crime drama. The suspects, feigning a casual innocence, blinking in the glare of the lights, face forward then shuffle around to turn to the side. The nervous witness (usually an elderly woman with glasses, anxiously clutching her handbag) squints at the suspects’ faces, trying hard to identify the man she saw from her bedroom window fleeing the scene (”It was so dark, I can’t be sure…”).

The traditional approach to police lineups makes a kind of gritty, noir-ish sense. It seems, instinctively, that it would be easier for a witness to recognize a perp if they have the opportunity to directly compare his looks to 5 other guys who look similar. On the surface it feels like a good safeguard. The presence of the other guys (the “fillers”) forces the witness to focus on the smaller details that distinguish one person’s appearance from another’s. It would seem to improve accuracy.

Apparently, as reported in the NY Times, a new sequential approach to conducting lineups, where the witness views one suspect at a time as opposed to all together, is being tried in some jurisdictions and a recent study compared the efficacy of the two models. The Illinois study, which looked at 700 real-world lineups and their outcomes, concluded that witnesses in group lineups are 33% more likely to accurately identify a perpetrator and 66% less likely to ID the wrong person. In other words, the study proves that it’s easier to accurately identify someone when you can compare their appearance with others’.

This got me thinking about usability testing (yeah, I’m weird that way). We are often engaged by clients to test a redesign of an existing site or function. In most cases, we recruit non-users (i.e. folks who have never used the current interface) to test the new design on the assumption that their feedback will be a more accurate predictor of the general market reaction. These test participants are usually not shown the existing design — they are only asked to comment on the new design on its own merits. Is it possible that the participant feedback would be richer and more specific (i.e. useful) if participants were allowed to compare the old and new versions?

There are many good reasons for the “traditional” usability testing approach, but the Police Lineups study raises an interesting question about how people parse distinctions and arrive at conclusions.

This entry was posted on Friday, April 21st, 2006 at 3:06 pm and is filed under General, Usability Testing, Finds.