69% of Innocent People in Experiment Signed False Confession

By Hans Sherrer

Justice:Denied magazine, Issue 27, Winter 2005, page 18

Sixty-nine percent of the 79 participants in a 1996 experiment at Williams College signed a confession after being falsely accused of causing a computer program to crash with the loss of data. The experimenters, Saul Kassin and Katherine Kiechel, wrote about their findings in The Social Psychology of False Confessions (see, Psychological Science, V. 7. N. 3, 125-8, May 1996).

The 40 males and 39 females that participated were told they were involved in a typing experiment. They were instructed to type the letters on a computer keyboard spoken by a person. However they were warned not to press the “ALT” key, since it would cause the computer program to crash and data to be lost.

The participants were not told that the object of the experiment was to find their response to being falsely accused and interrogated about their supposedly negligent act of pressing the “ALT” key while typing the letters.

The study was intended to provide information about four areas of false confessions about which little information is known:
 To measure the effect of a participant’s certainty of not pressing the “ALT” key on their likelihood of falsely confessing,the letters to be typed were read slowly (43 letters per minute) to half the participants by a tester and 50% faster (67 letters per minute) to the other half. The idea being that with more time to press the “right” key, the participants in the former group might be more certain of their innocence of pressing the “ALT” key, and thus have a “lower vulnerability” to falsely confessing than the latter group, who with less time between keystrokes might have a “higher vulnerability” to falsely confessing. The experiment found there was a correlation between certainty of one’s innocence and signing a false confession.

To measure the effect of confronting a suspect with false evidence of his or her guilt, half the participants were told an onlooker (a confederate in the experiment) was an eyewitness to their pressing of the “ALT” key, while the other half were not presented with independent evidence that they had pressed the key. The idea being that the former group might be less certain of their innocence than the latter group. The experiment found there was a correlation between being confronted with false evidence of one’s guilt and signing a false confession.

Since interrogators are trained to increase a suspect’s vulnerability by lying to a suspect that evidence ties him or her to the crime (e.g., an eyewitness, fingerprints, DNA, etc. that doesn’t actually exist), the experiments finding about the rate of false confessions when an “eyewitness” confirms that the “ALT” key was pressed is important for understanding what happens in the real world. The chart shows the studies results when an innocent person is confronted with the evidence of a false eyewitness to their alleged negligent act.

The table shows that when questioned about pressing the “ALT” key allegedly witnessed by another person, 94% of the innocent participants signed what was in fact a false confession. Perhaps more remarkably, 54% of those people believed they had pressed the “ALT” key when they hadn’t, and 20% confabulated facts explaining their non-existent pressing of the “ALT” key.

The meat of the experiment began when the computer screen suddenly went blank as a participant was typing a letter. The tester then examined the keyboard and the computer. After verifying that data was lost, the tester asked the participant: “Did you hit the “ALT” key?” All the participants initially denied pressing the “ALT” key (Mimicking the reflex response: ‘No Mommy! I didn’t take the cookies out of the cookie jar!’). However, whether the answer was affirmative or negative, the tester than wrote out a confession that the participant was asked to sign, “I hit the “ALT” key and caused the program to crash. Data was lost.” (126) Each participant was told a consequence of signing the confession would be a phone call from an investigator. If the participant refused to sign, they were asked a second time. Eventually, a high percentage signed the confession.

In addition to signing a confession, some participants made a voluntary statement. Among them, “I hit the wrong button and ruined the program.”; “I hit a button I wasn’t supposed to.”; and, “I hit it with the side of my hand right after you called out the ‘A’”. (126-7)

After each participant had either signed the confession or refused to do so, they were told that they hadn’t pressed the “ALT” or caused the loss of any data. According to Kassin and Kiechel, they mostly “reacted with a combination of relief (that they had not ruined the experiment), amazement (that their perceptions of their own behavior had been so completely manipulated), and a sense of satisfaction (at having played a meaningful role in an important study).” (127)

The experiment’s finding that there is a high likelihood an innocent person can be induced to falsely confess is highlighted by the finding that 35% of the participants falsely confessed who had a low vulnerability and no eyewitness claimed to have seen him or her press the “ALT” key. Its further finding that an innocent suspect is almost twice as likely to falsely confess when an interrogator lies about fake incriminatory evidence, suggests that techniques increases the unreliability of a confession to such a degree that it should be barred in real life.

Although critics correctly claim the experiment didn’t mimic the conditions of a police interrogation – that fact makes its findings all the more compelling. The participants were not subject to the overbearing pressure of being interrogated about a crime of which they knew nothing by the police in a hostile environment. They were all intelligent (avg. SAT over 1300), self-assured college students voluntarily participating in an activity and subjected to a grilling that they could have walked away from at any time. Furthermore, Kassin and Kiechel point out that the internalization of guilt and the fabrication of explanations by a significant percentage of the participants for their non-existent negligent action, “is not seriously compromised by the laboratory paradigm that was used.” In other words, it reflects what people do in the real world.

The experiment has serious implications for considering that a person’s claim of having falsely confessed has much more likelihood of validity than the incredulity that might intuitively be ascribed to such a claim. The importance of taking a false confession claim seriously is underscored by what was reported in a subsequent article that Kassin co-authored, Coerced Confessions and the Jury:

“In the studies reported in this article, mock jurors did not sufficiently discount a defendant’s confession in reaching a verdict – even when they saw the confession as coerced, even when the judge ruled the confession inadmissible, and even when participants said it did not influence their decision-making. The mere presence of a confession was thus sufficient to turn acquittal into conviction, irrespective of the contexts in which it was elicited and presented. (42)

…the presence of any confession powerfully increased the conviction rate – even when it was seen as coerced, even when it was ruled inadmissible, and even when participants claimed that it did not affect their verdicts.” (44) (Coerced Confessions and the Jury: An Experimental Test of the “Harmless Error” Rule, Saul M. Kassin and Holly Sukel, Law and Human Behavior, Vol. 21, No. 1, 1997, 27-46.)

So once made, the negative consequences of a false (or suspect) confession cannot be undone by anything less than dismissal of the charges. Since that is a rarity, the integrity of the criminal prosecution system is grievously undermined by the prevalence of false confessions, the techniques used to obtain them, and the deficient ability of police, prosecutors, judges and jurors to detect a real confession from a false one, or to discount it as evidence when it is known to be false.