Here’s an lightly-edited excerpt from this week’s installment of my chapter-by-chapter analysis of William Lane Craig’s book, On Guard:

In a study published in 2003 [PDF], psychology researchers Gary Wells and Elizabeth Loftus gave an example of how eyewitness testimony can evolve over time. A young woman was sexually assaulted and her friend was murdered. The young woman, Sherry Gillaspey worked with a police artist to put together a composite sketch of the assailant, and based this sketch, a young man named Thomas Brewster became a “person of interest.”

What happened next is a textbook case of how eyewitness testimony can be “improved” over time.

Here’s the sequence of events, as documented by Wells and Loftus.

On December 19, 1984, Gillaspey was shown a photo lineup with Brewster’s photo in it. She could not make a positive identification. One day later, Gillaspey was shown a live lineup in which Brewster appeared. Again, Gillaspey could not make a positive identification… Nearly four years later, in August 1988, detectives again showed Gillaspey a photo lineup with Brewster’s picture in it. Once again she could not make a positive identification.

In 1995, 11 years after the murder, two new detectives were assigned to the case. These detectives brought photos and, after interviewing her with the photos, she signed a statement saying that Brewster was the killer. Six days later, she identified Brewster from a live lineup.

The report goes on to look at details of how the two new detectives, apparently believing or wanting Brewster to be the attacker, subtly guided Gillaspey into “remembering” Brewster as the perpetrator. Initially, the woman did not remember Brewster as being the man who assaulted her and murdered her friend, despite being an eyewitness to the whole thing. Under the influence of the two detectives, however, this “memory” appeared—more than a decade later! Nor is this a rare case. Wells and Loftus cite research documenting at least 80 innocent men in recent years who have been wrongly imprisoned—or executed—based on eyewitness testimony, sometimes involving multiple witnesses. Other research explains how this can happen, as Wells and Loftus summarize:

[D]ecades of research has shown that postevent information, particularly when it is misleading, can also alter recollections of other details about key events. A typical finding is that after receiving new information that is misleading in some way, people make errors when they report what they saw. The new, postevent information is often incorporated into the recollection, supplementing or altering it, sometimes in dramatic ways.

In short, eyewitness testimony is the category of evidence that is THE most likely to be influenced and/or modified in the presence of stress, peer pressure, the power of suggestion, and so on. This is why it is important to make a distinction, in historical research, between “independent” accounts that arise via collaboration in a context of shared religious fervor and perceived persecution, versus accounts that are truly independent (i.e. that arise without any collaboration, religious identification, shared goals, etc., between the parties). It is especially significant, then, that William Lane Craig consistently fails to make any such distinction in his own historical “research.”

Specific details are at Evangelical Realism.