LEWISTON - Eyewitness accounts are always as reliable as people think, says a Bates College professor.

And, as with other evidence collected at a crime scene, it's important to preserve what the witness observed and lessen the likelihood of contamination, said Amy Bradfield Douglass, an associate professor of psychology. She spoke Thursday to about 50 people during an hour-long presentation at the Marsden Hartley Cultural Center at the Lewiston Public Library. She was featured as this month's speaker at the Great Falls Forum series.

Douglass, who teaches a course at the local college on psychology and law, said eyewitness identifications are often used in law enforcement and presented in court. They hold "very powerful" sway over juries. But they aren't as reliable as most people think, she said. They're often right; but they're also often wrong.

Of the 213 defendants exonerated using DNA testing by nationally-renowned The Innocence Project, 61 out of the first 70 included mistaken eyewitness identifications, Douglass said.

A 1993 report showed that more than 77,000 people became defendants each year after identification from lineups or photo spreads. Those procedures for identifying suspects often are flawed, she said.

When showing witnesses lineups of possible suspects, law enforcement officials should - but don't always - include in their instructions that the person they saw "may or may not be present."

When shown photos of possible suspects, witnesses should view individual pictures rather than group photos because they are inclined to compare the faces in a group and pick the one closest to the person they saw, even if that person's portrait isn't included in the spread, she said.

Using a PowerPoint presentation, Douglass asked her audience to pick the proper illustration of an American penny from a lineup. Few were successful, despite the fact that the penny is a common object.

A perspective drawing of a railroad track shows two railroad ties. Depending on where on the track a log is placed, the object appears to be smaller or larger than another the same size, she said.

In another experiment, a video created by Daniel Simons, an associate professor at the University of Illinois, shows a group of people passing basketballs among two subgroups divided into white shirts and black shirts. Viewers at Thursday's forum were told to focus on how many times the white-shirted ball passers handled the ball. During the sequence, someone dressed in a black gorilla suit entered the group, beat its chest, then exited.

Many in the audience failed to note the presence of the gorilla because they were so focused on the white shirts.

"It shows you how the mind is vulnerable to these pretty profound errors," she said during an interview after her presentation.

Interaction with the person representing law enforcement can influence a witness during suspect identification, often without the officer knowing it's happening, she said. In some cases, positive reinforcement given by an officer after a witness identifies a suspect can bolster the confidence that witness later has in his or her initial sighting.

Jurors will gauge the reliability of the witness on that person's level of confidence, regardless of whether the witness is right or wrong, Douglass said.