According to this, approximately 77,000 cases a year in the US rely solely on eye witness testimony.
Of course we would not want to disregard eye witness testimony otherwise, criminals would be able to run rampant as long as they know to destroy the evidence. It isn't like we have CCTV every where in the US. I think it is far more likely you would punish one innocent to catch 1000 criminals.
I read of an informal experiment done once at what I think was a psychologist's conference. The way the story went, a presentation was being held, full room of a sitting audience, all of them educated on matters of observations and perceptional bias.At a glance I can't seem to find any indication of the statistic when it comes to crimes with a single eye witness, so that probably cuts the cases down a bit, at least I'd assume that some of those 77,000 would involve multiple witnesses.
Also, this
"But the empirical facts about the validity of EW testimony, especially EW identification, are very disturbing. Aside from those few well-publicized cases of mistaken identity offering clear evidence that sometimes real witnesses make grave errors and innocentpeople are either imprisoned or executed (Wall, 1965), the outcomes of nearly all controlled psychological investigations employing realistic simulated crimes or staged events strongly suggest that identification errors in the real world may be relatively frequent , rather than rare . "
Is disconcerting.
A man came running through screaming, and he was followed by anothe man, waving a gun. They ran through the room, coming in from different entrances, and out, through different exits. Naturally, a lot of panic, before the speaker told everyone to calm down, and explained the experiment: everybody had a pad and pencil under their seats to record what they saw, and would they please write it out while it was fresh.
On review of these witness testimonies, it was found that not one got all the details right: the entrances and exits of the victim and aggressor, their clothes, skin color, facial descriptions. Most people only correctly reported a few attributes correctly each. It would have been a terrible mess trying to identify the correct details with only the collected testimony.
If anything, our legal systems place too much emphasis on eyewitness testimony as it is. It's not reliable, and it's prone to both prosecution and defense "massaging" the details in their "testimony prep". I am inclined against believing in any conviction based exclusively on eyewitness testimony, with no corroborating physical evidence.