Eyewitness testimony generally doesn't affect the commission of a crime, but it can have profound effects on the investigation of a crime and subsequent adjudication. The potential problem with eyewitness testimony begins immediately upon observation of an event by the witness. Contrary to media representations of memory that suggest memory is a kind of camcorder in people's heads, human beings never simply observe; they interpret what they observe through past experience, beliefs and enculturation. People focus on particular parts of an experience based on what those aspects of an event mean to them.
Loftus and other researchers found that retelling an event actually alters the memory of it. In the process of retelling, certain aspects of the event are emphasized, and this emphasis in turn begins a process of distortion. Moreover, witnesses retell an event to someone, and witnesses tailor the retelling to the particular audience. They will tell the story to a spouse, friend and policeman in different ways, based on what they believe is important to the particular audience.
Witnesses become psychologically attached to a version of events, even if that version is inaccurate. In tests using simulated police lineups, once a witness had identified the wrong person, the witness identified the same person more confidently each subsequent time. This is not the same as falsifying. The witnesses themselves were convinced of their own accuracy of memory and were being as truthful as they knew how. The problem is not one of character but of the plasticity of memory itself.
Memory is suggestible. That means that the introduction of suggestions by third parties can actually affect memory. Loftus' original research in the 1970s was in response to several egregious cases of implanted memory, where "therapists" has actually assisted clients in "recovering repressed memories" of childhood in which clients believed they had been raped by parents and even been involved in satanic rituals that involved eating babies. Experiments showed that suggestions by a questioner can alter witness memory.
Cultural biases and stress also have been shown to have effects on memory. People often have greater difficulty accurately identifying individuals of another ethnicity. During the commission of crimes with weapons, many witnesses become hyper-focused on the weapon itself instead of the perpetrator. Fleeting glimpses, as opposed to sustained observations, are even more prone to bias and third-party suggestions.