There is an experiment described in the article. A group of volunteers take part in a test. In the test, they are made to believe that they saw a laptop being stolen. They are then asked to identify who stole it, from a line-up consisting of “suspects.” All the “suspects” are “innocent.” The volunteers record their judgement and mark their level of confidence. They are called back a couple of days later to help further with the investigation. Now some of them are told that all the suspects denied stealing, and the rest are told that one of them confessed. The volunteers who were told that one of the suspects had confessed showed a remarkable shift in their judgement. The fact that they were told of the confession was enough to convince a lot of them that the guy was really guilty, and their own earlier judgement about his innocence was wrong!!
Experiments such as this show how tricky things can get in a real court, because the eyewitnesses can be “tainted by confessions.”