We had an interesting experience recently that made very real the problem with eye-witness testimony, and shows why I rarely rely on it as a prosecutor unless there's some very good corroborating evidence.
One weekday evening, my son and three friends were in a nearby park (he and the two girls were eleven, and another boy was eight). They encountered two boys who identified themselves as 8th graders at a nearby middle school. Unfortunately, these boys began throwing rocks at the younger kids, and threatening to kill and rape them.
My son is a tough lad but had no desire to stand his ground, so took off running. He fell and did this:
We reported the incident to the police, who can't do very much without an identification of those boys. So on Friday we rendezvous-ed at the elementary school to look at a yearbook from the middle school, to see if the kids could recognize the two boys. Actually, the kids and one parent, who'd gone to look for her daughter in the park and encountered them.
So, four kids and one adult.
They took turns leafing through the yearbook, so as not to influence each other. The photos weren't divided up by year, but by alphabet, so they had to look through several hundred pictures of sixth, seventh, and eighth graders to find the stone-throwers.
How hard could it be?
I'll tell you: very.
By the end of an hour, we had eleven suspects. The identifications varied from fifty percent to eighty percent certainty, but the interesting thing to me (apart from the number eleven) was that we had zero crossover. In other words, none of the five people perusing the book had chosen the same kid.
Pretty scary, eh?