A recent study suggests that perfect agreement among witnesses in a criminal case is so improbable that it should be considered an indicator that something is wrong. In the study published in the Proceedings of the Royal Society A, social scientists used statistical analysis to prove that the probability of perfect agreement among witnesses is close to zero.
The proponents of that study suggested that complete agreement between a large number of witnesses is so unlikely that it suggests some unreliable factor is is influencing the witnesses. In the study, the scientists determined that witness testimony identifying a suspect is fairly reliable up when up to three witnesses identify the suspect, but with every witness over three who corroborates the identification the "the confidence that . . . the right person is being identified" erodes.
Social scientists have long understood that eye-witness accounts are particularly vulnerable to influence from external influences. For this reason, law enforcement should follow very strict procedures when eliciting statements and identifications from eye witnesses. If law enforcement doesn't take appropriate precautions to ensure that an eye-witness identification is not tainted by some outside influence, that evidence may be inadmissible in a criminal prosecution.