The Loftus and Palmer experiment was carried out by Elizabeth Loftus - she's made huge contributions in terms of things like false memories and how people RECONSTRUCT memories. In 1974, Loftus along with her colleague Palmer wanted to find out whether eye-witness testimony was genuine (they believed that often if not always, it wasn't genuine as people modified their memory of an incident based on past experience, social influences, their own expectations, etc). They made participants watch a video of a motor accident, after which they interrogated them as if they were being cross-examined in court. Some participants were asked ‘How fast were the cars going when they COLLIDED with each other?’, while other participants had other words such as bumped into, hit, contacted, smashed into substituted for ‘collided’.
Conclusion? It was found that the more impactful or more severe and intense the word in the question sounded like, the higher would be the eyewitnesses estimate of the speed. This was proven by the results as well which indicated that a high intensity crash implying word like "SMASHED" generated an estimated speed by the eyewitnesses of 40.5 m/ph; whereas a low intensity and less severe crash implying word generated an estimated speed of only 31.8 m/ph. This suggests that memories are never 100% reliable as even direct observation of an accident can have varying witness recounts (or in this case, estimates of speed) depending on HOW that question is asked to the eyewitness or the words used in the question which may alter or influence the eyewitnesses memory of what they actually saw.