When facts are uncertain (height, weight, distance, sequenced events, identity of perpetrators, green lights or red) people are incredibly suggestible. That's why negotiators learn how to "frame" offers in such a way as to make them seem better than they are -- we'll take less than a million rather than more than $750,000 for instance. Now the social scientists have demonstrated not only that our estimates of facts before our eyes are subject to framing, but our recollection of events is as well.
Thanks to @JuryVox in my twitter network (follow him now!) I'm able to bring you some of the most recent research in the field in the article Memory Manipulated After the Event from PsyBlog:
[A]lthough Loftus and Palmer's (1974) study was quite simple, its implications were profound. In the first of two experiments, 45 participants watched a film of a car accident. Nine of these were then asked this specific question: "About how fast were the cars going when they hit each other?". The other four groups of nine were asked an almost identical question but with one important difference. Instead of the word hit, the words 'smashed', 'collided', bumped, and 'contacted' were used.
The participants who were asked the question using the word 'smashed' as opposed to 'contacted', estimated the cars were travelling, on average, almost 10mph faster. The other words were fanned out in between.
So, what is this telling us? Probably, that because people are not good at gauging the speed of a car, the cue comes from the experimenter. If the experimenter asks the question using the word 'smashed', the participant assumes it was going faster than if they say 'contacted'.
It's in the follow-up study that things get interesting. The same experiment was repeated roughly as before, but with 150 participants. This time, however, participants filled out a questionnaire about the crash and were then asked to return in a week. As before, the question about the speed of the crash was varied between groups. Some read 'hit', some 'smashed' and so on.
One week later participants returned and were asked to fill out a questionnaire about the accident in which was hidden a crucial question: "Did you see any broken glass?". As broken glass is indicative of a more serious accident, so greater speed, Loftus and Palmer expected the group in which the word 'smashed' had been used would be more likely to indicate there was broken glass. This was exactly what they found.