A few words about the potential pitfalls of human memory, communication, and perception might shed light on how the "myth" surrounding the Roswell event evolved. The following observations are strictly from the standpoint of common sense and real-world experience, not from some abstract or esoteric psychological theory.
Faulty memory is obviously one of the biggest detriments to accurately reconstructing past events, especially if a great deal of time has elapsed. While memory generally serves us well, it is unlike the playback of a digital tape -- it is far from precise or absolute. With the passage of time, images from different events can be inadvertently blended together. Also, mental images of things only imagined can be inadvertently blended or confused with memories of things actually observed or events actually experienced.
If memory were perfect, everyone of us would have consistently accomplished perfect scores in our comprehensive exams throughout high school and college. We would remember with perfect accuracy every name, face, scene, place, and event that we had ever heard, seen, or experienced. Obviously, human memory doesn't work this way. Interestingly, however, in the case of witness testimony, there often seems to be an unrealistic assumption that it does.
Like faulty memory, inaccurate communication can also be a detriment to accurately reconstructing past events. Miscommunication is definitely a factor with secondhand testimony. Perhaps, that is part of the reason secondhand testimony is not admissible in a court of law. There is a game sometimes played by children that amply illustrates the problem of miscommunication, and consequently also secondhand testimony. A group stands in a circle and a message is whispered from one person to the next. After being passed around a few times, the original message is usually completely unrecognizable. It doesn't take much imagination to realize how much the problem would be exacerbated if there was a gap of several years each time the message was passed on. Interestingly, in the Roswell case, much of the secondhand testimony, upon which so much importance is placed, comes from information passed on many years back.
Errors in perception are probably more frequent than most people realize. Perception is often influenced by expectation. We have probably all noticed this phenomenon on a micro scale when trying to proofread something. Similar to a form of computer enhancement, the brain tends to automatically comp nsate for missing letters or words which are supposed to be there. In other words, it tends to automatically "fill in the blanks."
The problem of expectation influencing perception is exacerbated when recalling past events. Memory seems to somehow have the ability to distort an observation retroactively to make it better match expectation. This phenomenon can be a real problem with eyewitness accounts to traffic accidents, and even more so to aviation accidents -- even accounts from experienced observers. For example, in August 1987, a Northwest Airlines MD-88 took off from Detroit Metro airport. The crew, however, had failed to lower the flaps to the takeoff position (15 degrees). Under most circumstances, this error would not have been a fatal mistake. Unfortunately, due to other contributing factors (short runway, hot day, fully loaded aircraft, etc.), they didn't make it. During the investigation of the accident, the crew of the aircraft in line for takeoff immediately behind the Northwest airplane testified that its flaps were definitely lowered to the takeoff position. Analysis of the wreckage and flight-data recorder, however, showed otherwise. The testimony of the crew of the second aircraft was wrong. What they remembered was what they thought they should have seen rather than what they actually saw.