In 1989 a female jogger was beaten senseless, raped and left for dead in New York City's Central Park. Her skull had multiple fractures, her eye socket was crushed, and she lost three quarters of her blood. She survived, but she cannot remember anything about the incident. Within 48 hours of the attack, solely on the basis of confessions obtained by police, five African- and Hispanic-American boys, 14 to 16 years old, were arrested. The crime scene had shown a horrific act but carried no physical traces at all of the defendants. Yet it was easy to understand why detectives, under the glare of a national media spotlight, aggressively interrogated the teenagers, at least some of whom were "wilding" in the park that night.
Four of the confessions were videotaped and later presented at trial. The tapes were compelling, with each of the defendants de scribing in vivid -- though, in many ways, erroneous--detail how the jogger was attacked and what role he had played. One boy reenacted the way he pulled off her running pants. Another said he felt pressured by the others to participate in his "first rape"; he expressed remorse and promised that it would not happen again. After their arrest, the youths recanted these confessions, because they had believed that making a confession would have enabled them to go home. Regardless of the denials, the tapes collectively persuaded police, prosecutors, two trial juries, a city and a nation; the teenagers were convicted and sentenced to prison.
Thirteen years later Matias Reyes, who was in jail for three rapes and a murder committed after the jogger attack, stepped forward of his own initiative. He volunteered that he was the Central Park assailant and that he had acted alone. The Manhattan district attorney's office questioned Reyes and discovered that he had accurate, privileged and independently corroborated knowledge of the crime and crime scene. DNA testing further revealed that the semen samples recovered from the victim--which had conclusively excluded the boys as donors--belonged to Reyes. (Prosecutors had argued at trial that just because police did not capture all the alleged perpetrators did not mean they did not get some of them.) In December 2002 the five teenagers convictions were vacated.
Despite its notoriety, the case illustrates a phenomenon that is not new or unique. The pages of legal history reveal many tragic miscarriages of justice involving innocent men and women who were prosecuted, wrongfully convicted, and sentenced to prison or to death. Opinions differ on prevalence rates, but it is clear that a disturbing number of cases have involved defendants who were convicted based only on false confessions that, at least in retrospect, could not have been true. Indeed, as in the case of the Central Park incident, disputed false confessions have convicted some people notwithstanding physical evidence to the contrary. As a result of technological advances in forensic DNA typing--which enables the review of past cases in which blood, hair, semen, skin, saliva or other biological material has been preserved--many new, high-profile wrongful convictions have surfaced in recent years, up to 157 in the U.S. alone at the time of this writing. Typically 20 to 25 percent of DNA exonerations had false confessions in evidence.
Why would an innocent person confess to a crime? A scan of the scientific literature reveals how a complex set of psychological factors comes into play. First, techniques commonly used by investigators during interviews make them prone to see deceit in suspects, a perception that tends to bias the outcome of the questioning. When the accused waive their constitutional rights to silence and to counsel during questioning by the police, they may also unwittingly lose procedural safeguards and put themselves at greater risk of making a false confession. Other contributors include a given person's tendencies toward compliance or suggestibility in the face of two common interrogation tactics--the presentation of false incriminating evidence and the impression that giving a confession might bring leniency. In short, sometimes people confess because it seems like the only way out of a terrible situation.
More troubling, confession evidence is inherently prejudicial, influencing juries even when they are shown evidence of coercion and even when there is no corroboration. Ultimately, we believe, society should discuss the urgent need to reform practices that contribute to false confessions and to require mandatory videotaping of all interviews and interrogations.
Discerning the Truth
A 2004 conference on police interviewing attended by the two of us illustrates the problem of bias during questioning. Joseph Buckley--president of John E. Reid and Associates (which has trained tens of thousands of law-enforcement professionals) and co-author of the manual Criminal Interrogation and Confessions (Aspen Publishers, 2001)--presented the influential Reid technique of interviewing and interrogation. Afterward, an audience member asked if the persuasive methods did not at times cause innocent people to confess. Buckley replied that they did not interrogate innocent people.
To understand the basis of this remark, it is important to know that the highly confrontational, accusatory process of interrogation is preceded by an information-gathering interview intended to determine whether the suspect is guilty or innocent. Sometimes this initial judgment is reasonably based on witnesses, informants or other extrinsic evidence. At other times, however, such judgments may be based on nothing more than a hunch, a clinical impression that investigators form during a preinterrogation interview.
The risk of error at this stage is clear, as in the 1986 Florida case involving Tom Sawyer, whom investigators accused of sexual assault and murder and interrogated for 16 hours, extracting a confession. His statement was later suppressed by the judge, and the charges were dropped. Sawyer had become a prime suspect because his face flushed and he appeared embarrassed during an initial interview, a reaction interpreted as a sign of deception. Investigators did not know that Sawyer was a recovering alcoholic with a social anxiety disorder that caused him to sweat profusely and blush in evaluative social situations. Many of the characteristics associated with acting "guilty" are also signs of a person under high stress.
Separating truths from lies is tricky. In fact, most experiments have shown that people perform at no better than chance levels and that training programs produce, at best, small and inconsistent improvements compared with naive control groups. In general, professional lie catchers, such as police detectives, psychiatrists, customs inspectors and polygraph examiners, exhibit accuracy rates in the 45 to 60 percent range, with a mean of 54 percent.
Even with those statistics, trained investigators believe they are more accurate in determining guilt or innocence. In 2002 Christian Meissner of Florida International University and one of us (Kassin) conducted a meta-analysis to examine their performance. Across studies, investigators and educated participants, relative to naive controls, exhibited a proclivity to judge targets as deceptive--and to do so with confidence [see table above]. Expressing a particularly cynical but telling point of view, one detective is quoted as saying in a 1996 article by Richard A. Leo of the University of California at Irvine, "You can tell if a suspect is lying by whether he is moving his lips."
Protections Averted
Miranda may not yield the protective effect for which it was designed for two reasons. First, a number of suspects--because of their youth, level of intelligence, lack of education or mental health status--do not have the capacity to understand and apply the rights they are given. Second, police use methods of presentation that elicit waivers. After observing live and videotaped police interrogations, Leo found that roughly four out of five suspects waive their rights and submit to questioning. He also observed that individuals who have no prior felony record are more likely to waive their rights than are those with a history of criminal justice "experience." In a 2004 study by one of us (Kassin) and Rebecca Norwick of Harvard University, subjects guilty or innocent of a mock crime (stealing $100) were confronted by a neutral, sympathetic, or hostile "Detective McCarthy" who asked if they would waive their rights and talk. Only 36 percent of guilty subjects agreed, but 81 percent of innocents waived these rights, saying later they had nothing to hide or fear [see chart above].
Interrogation Tactics
In the past, American police routinely practiced "third degree" methods of custodial interrogation--inflicting physical or mental pain and suffering to extract confessions and other types of information from crime suspects. Such tactics have mostly faded into the annals of criminal justice history, but modern police interrogations remain powerful enough to elicit confessions. At the most general level, it is clear that the two-step approach employed by Reid-trained investigators and others--in which an interview generates a judgment of truth or deception, which in turn determines whether or not to proceed to interrogation--is inherently biased.
For innocents who are initially misjudged, one would hope that interrogators would remain open-minded and reevaluate their beliefs over the course of questioning. A warehouse of psychology research suggests, however, that once people form a belief, they selectively seek, collect and interpret new data in ways that verify their opinion. This distorting cognitive confirmation bias makes such personal convictions resistant to change, even in the face of contradictory evidence. It also contributes to the errors committed by forensic examiners whose judgments of handwriting samples, bite marks, tire marks, ballistics, fingerprints and other "scientific" observations are often corrupted by a priori expectations, a problem uncovered in many DNA exoneration cases.
In one instance in 2002, Bruce Godschalk was exonerated of two rape convictions after 15 years in prison when laboratories for both the state and the defendant found from his DNA that he was not the rapist. Yet the district attorney whose office had convicted Godschalk--even though Godschalk disavowed his initial confession--argued that the DNA tests were flawed and refused at first to release him from prison. When the district attorney was asked what foundation he had for his decision, he asserted, "I have no scientific basis. I know because I trust my detective and his tape-recorded confession. Therefore, the results must be flawed until someone proves to me otherwise."
The presumption of guilt also influences the way police conduct interrogations, perhaps leading them to adopt an aggressive and confrontational questioning style. Demonstrating that interrogators can condition the behavior of suspects through an automatic process of social mimicry, Lucy Akehurst and Aldert Vrij of the University of Portsmouth in England found in 1999 that increased gestures and physical activity among police officers triggered movement among interviewees--fidgeting behavior that is then seen by others as suspicious.
It is important to scrutinize the specific practices of social influence that get people to confess. Proponents of the Reid technique advise interrogators to conduct the questioning in a small, barely furnished, soundproof room. The purpose is to isolate the suspect, increasing his or her anxiety and desire to escape. To further heighten discomfort, the interrogator may seat the suspect in a hard, armless, straight-backed chair; keep light switches, thermostats and other control devices out of reach; and encroach on the suspect's personal space over the course of interrogation.
Against this physical backdrop, the Reid operational nine-step process begins when an interrogator confronts the suspect with unwavering assertions of guilt (1); develops "themes" that psychologically justify or excuse the crime (2); interrupts all efforts at denial and defense (3); overcomes the suspect's factual, moral and emotional objections (4); ensures that the passive suspect does not withdraw (5); shows sympathy and understanding and urges the suspect to cooperate (6); offers a face-saving alternative construal of the alleged guilty act (7); gets the suspect to recount the details of his or her crime (8); and converts the latter statement into a full written or oral confession (9). Conceptually, this system is designed to get suspects to incriminate themselves by increasing the anxiety associated with denial, plunging the suspect into a state of despair and then minimizing the perceived consequences of confession.
Rates of confession vary in different countries, indicating the underlying role that institutional and cultural influences play. For example, suspects detained for questioning in the U.S. confess at a rate around 42 percent, whereas in England the figure is closer to 60 percent. In Japan, where few restraints are placed on police interrogations and where social norms favor confession as a response to the shame brought by transgression, more than 90 percent of suspects confess.
In so-called self-report studies, researchers ask why people confessed. In 1991 one of us (Gudjonsson) and Hannes Petursson of University Hospital in Reykjavik, Iceland, published the first work in this area carried out on Icelandic prison inmates, which was replicated in Northern Ireland and in a larger Icelandic prison population with an extended version of a 54-item self-report instrument, the Gudjonsson Confession Questionnaire.
Although most suspects confess for a combination of reasons, the most critical is their belief about the strength of the evidence against them. That is why the tactic of presenting false evidence--as when police lie to suspects about an eyewitness that does not exist; fingerprints, hair or blood that has not been found; or lie detector tests they did not really fail--can lead innocent people to confess. In a 1996 laboratory experiment that illustrates the point, Kassin and Katherine L. Kiechel of Williams College falsely accused college students of crashing a desktop computer by hitting a key that they were told was off-limits. When a fellow student who was present said she had witnessed the students hit the forbidden key, the number induced to sign a confession increased by 45 percent. Also increased were the numbers who internalized a belief in their own guilt and fabricated false memories to support that belief.
False Confessions
In 2004 Steven A. Drizin of Northwestern University School of Law and Leo analyzed 125 cases of proved false confessions in the U.S. from between 1971 and 2002, the largest sample ever studied. Approximately two thirds were exonerated before the trial, and the rest came after conviction. Ninety-three percent of the false confessors were men. Overall, 81 percent occurred in murder cases, followed by rape (8 percent) and arson (3 percent). The most common bases for exoneration were that the real perpetrator was identified (74 percent) and that new scientific evidence was discovered (46 percent). The sample was disproportionately represented by persons who were young (63 percent were younger than 25; 32 percent were under 18), mentally retarded (22 percent) and diagnosed with mental illness (10 percent). Astonishingly, 30 percent of the cases contained more than one false confession to the same crime, as in the Central Park jogger case, typically indicating that one false confession was used to get others.
Recognizing that people confess in different ways and for different reasons, psychologists categorize false confessions into three groups:
Voluntary false confessions. When aviator Charles Lindbergh's baby was kidnapped in 1932, some 200 people stepped forward to confess. In the 1980s Henry Lee Lucas falsely admitted to hundreds of unsolved murders, making him the most prolific serial confessor in history. People might voluntarily give a false confession for reasons including a pathological desire for notoriety; a conscious or unconscious need to expiate feelings of guilt over prior transgressions; an inability to distinguish fact from fantasy; and a desire to aid and protect the real criminal.
Compliant false confessions. In these cases, the suspect confesses to achieve some end: to escape an aversive situation, to avoid an explicit or implied threat, or to gain a promised or implied reward. In Brown v. Mississippi in 1936, for example, three black tenant farmers admitted to murder after they were whipped with a steel-studded leather belt. And in the Central Park jogger case, each boy said he had confessed despite innocence because he was stressed and expected to go home if he cooperated.
Internalized false confessions. During interrogation, some suspects--particularly those who are young, tired, confused, suggestible and exposed to false information--come to believe that they committed the crime in question, even though they did not. In a classic case, 18-year-old Peter Reilly of Falls Village, Conn., returned home one night to find that his mother had been murdered. Reilly immediately called the police but was suspected of matricide. After gaining Reilly's trust, the police told him that he failed a lie detector test (which was not true), and which indicated that he was guilty even though he had no conscious memory of the event.
After hours of interrogation, the audiotape reveals that Reilly underwent a chilling transformation from denial to confusion, self-doubt, conversion ("Well, it really looks like I did it") and finally a full confession ("I remember slashing once at my mother's throat with a straight razor I used for model airplanes.... I also remember jumping on my mother's legs"). Two years later independent evidence revealed that Reilly could not have possibly committed the murder.
Trial jurors, like others in the criminal justice system who precede them, can be overly influenced by confessions. Archival analyses of actual cases containing confessions later proved false tell a disturbing tale. In these cases, the jury conviction rates ranged from 73 percent (as found by Richard Ofshe of the University of California at Berkeley and Leo in 1998) to 81 percent (as found by Drizin and Leo in 2004)--about the same as cases in which the defendants had made true confessions.
In light of such findings, the time is ripe for law-enforcement professionals, policymakers and the courts to reevaluate current methods of interrogation. Although more research is needed, certain practices clearly pose a risk to the innocent. One such factor concerns time in custody and interrogation. The 2004 study by Drizin and Leo found that in proved false confession cases, the interrogations lasted for an average of 16.3 hours. In the Central Park case, the five boys were in custody for 14 to 30 hours by the time they confessed. Following the Police and Criminal Evidence Act of 1986 (PACE) guidelines implemented in England and Wales, policy discussions should begin with a proposal for the imposition of time limits for detention and interrogation or at least flexible guidelines, as well as periodic breaks for rest and meals.
A second problem concerns the tactic of lying to suspects about the evidence. Research shows that people capitulate when they believe that the authorities have strong evidence against them. The practice of confronting suspects with real evidence, or even their own inconsistent statements, should increase the reliability of the confessions ultimately elicited. When police misrepresent the evidence, however, innocent suspects come to feel as trapped as the perpetrators--which increases the risk of false confession.
A third matter revolves around the use of minimization, as when police suggest to a suspect that the conduct in question was provoked, an accident or otherwise morally justified. Such tactics lead people to infer leniency in sentencing on confession, as if explicit promises had been made. In a study that is now in press, Melissa Russano of Roger Williams University and her colleagues found that such covert assurances can contribute to false confessions.
The Need for Reforms
To assess any given confession accurately, police, judges, lawyers and juries should have access to a videotaped record of the interrogation that produced it. In Great Britain, PACE mandated that all sessions be taped. In the U.S., four states--Minnesota, Alaska, Illinois and Maine--have mandatory videotaping, although the practice is often found elsewhere on a voluntary basis. Videotaping deters interrogators from using the most aggressive, psychologically coercive methods. It also will block frivolous defense claims of coercion where none existed. And it provides an objective and accurate record of all that transpired, avoiding disputes about how the confession came about.
A 1993 National Institute of Justice study revealed that many U.S. police departments already have videotaped interrogations--and the vast majority found the practice useful. More recently, in 2004, Thomas P. Sullivan of the law firm Jenner & Block interviewed officials from 238 police and sheriff's departments in 38 states who made such recordings voluntarily and found that they enthusiastically favored the practice, which increases accountability, provides an instant replay of the suspect's statement that reveals information initially overlooked and reduces the amount of time spent in court defending their interrogation conduct. As a counter to the most common criticisms, those interviewed found that videotaping is not costly and does not inhibit suspects from talking to police.
Such reforms are sorely needed. Only then can society trust the process of interrogation and the confessions that it produces--and help to promote justice for all.


