When I was involved with the IAM, we had illustrated talks from police crash investigators, where they would bring along slides of accident scenes and ask us to speculate about how the crashes happened. They would then talk us through what the investigations revealed to have actually happened. The results were often very surprising indeed.
One lesson that stuck with me very clearly: eye witness testimony, while superficially compelling, is often wildly inaccurate. One of them talked about two experiments the police carried out as a result of being puzzled by the phenomenon of how ordinary members of the public could apparently just make stuff up that they hadn't actually seen yet were convinced they had.
The simplest one was to drive the same car past the same witness at the same point on the same road three times. The witness was asked to estimate the speed of the car. The experiment was run many times with a huge range of members of the public as witnesses. Typical speed estimates for the three runs were 25mph, 35mph and 50mph.
In fact, the speed of the car was exactly 30mph all three times. The difference was the car was in 4th gear for the first run, 3rd for the second and 2nd for the final one. People heard very different engine notes and extrapolated the speed from that, despite the fact that they had a clear view of the car as it passed them.
The second experiment was more elaborate. The Met has a mock town centre in its training centre at Hendon (used for riot training, etc). Members of the public were recruited to take part in what they thought was a questionnaire. They were taken in small groups from one building, along this mock street to another building. En-route, a woman pushed a pram onto a zebra crossing. A car clipped the pram then drove to the T-junction and turned right.
The public thought this was a real crash, and were taken immediately into separate rooms to give statements. Despite the fact that they were interviewed literally within three minutes of the incident, their statements were wildly variable. The colour of the car was black or red or white or silver. The car had one or two or three or four occupants. The driver was young or old, and was white, black or asian. The car hit the pram or the mother. The baby sometimes went flying out of the pram (which was in fact empty). The car was doing 40, 50, 60 or 70mph (it was doing 25mph). The car turned left, right or went straight on at the bottom of the road (the latter impossible as it was a T-junction).
Since then, I don't take much notice of press reports of witness statements.