Trowbridge Foundation: Current Article in Full

Suggestibility and Confessions

By Brett C. Trowbridge, Ph.D., J.D.

This article supported by The Trowbridge Foundation

Originally published in the American Journal of Forensic Psychology, Vol 21, Issue 1, 2003

Introduction: The Problem of False Confessions

Police are taught that obtaining confessions is one of their most crucial objectives, and around 80% of criminal cases are solved by a confession. Confessions are considered to be compelling by jurors, who are more likely to convict on the basis of a confession than anything else, including eyewitness identification. We know that there are false confessions, because many instances of false confessions have been documented. Rattner cites an authority who estimates that five percent of people convicted were actually innocent. In a sample of 205 cases, coerced confessions accounted for 8.4 %.

Kassin and Wrightsman have described three psychologically distinct types of false confession: “voluntary”, “coerced-compliant” and “coerced-internalized” types. “Voluntary” false confessions are offered by people without any external pressure from the police. These individuals simply report themselves to the authorities, claiming they have committed a crime, often a notorious one they have read about in the newspapers or seen on television. Sometimes voluntary false confessors are unable to distinguish between fact and fantasy. Sometimes they are trying to protect the real culprit. “Coerced-compliant” false confessions result from the pressures of coercion during the interrogation process. The accused gives in to the pressure of the interrogation for an instrumental gain, such as terminating the interview or being allowed to go home. The suspect confesses despite knowing that he did not commit the crime, doing so because the perceived immediate gains outweigh the uncertain long-term consequences. In a “coerced-internalized” confession, on the other hand, the person being interrogated does gradually accept the suggestion of his guilt, temporarily (or sometimes permanently) believing he did commit the crime even though he has no memory of actually doing so. Because of subtle manipulation during the interrogation the suspects begin to distrust their own beliefs. Coerced-complaint false confessors are likely to retract their confession as soon as the immediate stress of the interrogation is over, but coerced-internalized false confessors only retract after they again become convinced they are truly innocent of the crime. Innocent people may remain permanently convinced that they have committed a crime. Coerced-internalized false confessors tend to be trusting of people in authority, to lack self-confidence, and to have high levels of suggestibility .

Kassin has concluded that our criminal justice system does not grant adequate protection to suspects in interrogation situations, as the police routinely use deception and trickery, which may at times cause innocent people to confess to crimes they did not commit. Standard textbooks for police detectives describe for the police the most effective ways of using trickery and deceit to obtain confessions, but provide little to help police determine whether the confessions they have obtained that way are true or false confessions. Under U.S. law, deliberate lying by interrogators in order to obtain confessions from suspects is permitted.

One experiment demonstrated how such trickery by interrogators can lead to large numbers of false confessions even among high functioning people. Subjects who were involved in what they thought was a reaction time experiment, were accused of damaging a computer by pressing the wrong key, a key that they had been told not to touch. All were truly innocent, but the computer screen went blank, and they were accused of having pressed the forbidden key. Some subjects had been very rushed in their task, while others had been allowed to type at a slow and relaxed pace. A confederate said (in one condition) that she had seen the subject press the forbidden key, or (in another condition) she had not seen the subject do so. The experimenters then asked the subjects to sign a false confession. The experimenters also unobtrusively recorded the way the subjects privately described what had happened soon afterward, by placing another confederate in the waiting room to ask the subject “what happened”. The subject’s reply was recorded verbatim, and scored for whether he had internalized guilt for what happened. Finally, after the sessions appeared to be over the experimenter re-appeared, took the subjects back into the lab, and asked them to reconstruct how they had hit the wrong key. This procedure was designed to determine whether subjects would confabulate details to the fit the allegation.

Overall, 69% of the 75 subjects signed the confession, 28% exhibited internalization, and nine percent confabulated details to support their false beliefs. Only 35% of the subjects in the slow-paced group with no witness signed the confession and none showed internalization or confabulation. However, in the fast-paced group with a witness stating she had seen the wrong key being pressed, 100% signed a confession, 65% came to believe they were guilty (in reality, they were not), and 35% confabulated details to support their false beliefs. This study provides strong support for the notion that the presentation of false incriminating evidence, an interrogation ploy routinely used by police, can induce people to internalize blame for outcomes they did not produce. These effects were exhibited by college students who are intelligent (mean score on the SAT over 1300), self-assured, and under minimal stress compared with crime suspects held in custody, often in isolation. In other words, even very bright college students can be “suggestible”.

These findings confirm earlier work showing the effects of misinformation on memory for observed events. Although these studies demonstrate that false confessions can easily be elicited, mock jury research tells us that juries are prone to believe confessions that are obtained using such deceptive methods. Apparently jurors find it hard to believe that anyone would confess to a crime they did not commit.

Interrogative Suggestibility

Given the fact that many people can be induced to make false confessions, and that juries will convict people who confess during police interrogation after being told untruths about the situation, it becomes important to try to determine which people are more “suggestible”, i.e. prone to be tricked into believing something true or saying that something is true when it is not when confronted in an interrogation context with false evidence suggesting that it is true, and or by interpersonal pressure to change an answer that has been given. Suggestibility in interrogation situations should be distinguished from hypnotic suggestibility, or “hynotizability”.

Binet first introduced the concept of interrogative suggestibility around the turn of the century. His procedure for measuring interrogative suggestibility involved asking leading questions concerning pictures subjects had been previously shown. In 1938 Stern demonstrated that a leading questions procedure distorted responses when they are phrased in such a way as to suggest the desired response, whether correct or incorrect. Gudjonsson and Clark have developed a theoretical model for explaining suggestibility in police interrogation situations. According to the model, interrogative suggestibility is dependent upon the coping strategies that subjects can generate and implement when dealing with uncertainty and expectations in the interrogative situation. Interrogative suggestibility is thought to be a trait, although suggestibility can be enhanced or diminished by demand characteristics of a situation, or by whether the person being questioned adopts a suggestible or resistant response set. Gudjonsson is convinced that stable individual differences in interrogative suggestibility can be measured reliably. According to the model a suggestible person who is subjected to leading questions and negative feedback may have his memory of the event in question distorted, particularly if the interviewee does not know for certain the right answers to the questions.

Gudjonsson Suggestibility Scales

Gisili Gudjonsson was initially a detective with the Reykjavik Criminal Investigation Department, and as such became aware that false confessions can occur. When he obtained a post as a psychologist at the Institute of Psychiatry at the University of London, he began being asked to evaluate both handicapped victims of alleged crimes and defendants who had retracted confessions made during police interviewing. It became evident to him that what was actually being asked for was some measure of the individuals’ levels of suggestibility. After informally asking leading questions to see if they would change their reports, and after informally putting some pressure on them to see if they would change their reports, Gudjonsson realized it would make scientific sense to develop a standardized psychological test for measuring “interrogative suggestibility”. This test would also provide a conceptual framework for assessing the reliability of evidence using psychological procedures, whether that evidence would come from the defendant himself in the form of a confession, or from a witness. This conceptual framework offers the potential to relate to both legal and psychological constructs.

Because there appeared to be no suitable instrument for quantitatively measuring this type of suggestibility, Gudjonsson constructed a suggestibility scale to assess an individual’s response to “leading questions” and to “negative feedback” instructions (being told your answers are incorrect, and to be more accurate), referred to as ‘yield’ and ‘shift’ respectively. The Gudjonsson Suggestibility Scale (GSS) is intended to measure individual differences in the tendency to ‘yield’ to leading questions and to ‘shift’ elements of one’s previous answers in response to criticism or interpersonal pressure.

Subjects taking the Gudjonsson Suggestibility Scale (GSS-1) are told: “I want you to listen to a short story. Listen carefully because when I finish I want to you to tell me everything you remember.” An audiotape of a story is then played, after which the subject is directed, “Now tell me everything you remember about the story”. The story is a narrative paragraph describing a fictitious robbery.

Each idea recalled from the story counts as one point, for a maximum of 40 points. Next, the subject is asked to answer twenty questions about the story, 15 of which are leading questions. These 15 questions are designed to find out if the subject “yields” to suggestive questions. Each leading question that is answered in the affirmative (i.e. with a false alternative reply) scores one point, for a maximum of 15 points on the yield scale. After answering all 20 questions the subjects are “firmly” told: “You have made a number of errors. It is therefore necessary to go through the questions once more, and this time try to be more accurate.”

The same questions are repeated, with each distinct change in the nature of the reply to any of the suggestive questions counting towards the total shift score. The total Suggestibility Scale is the sum of the yield and shift scores.

This test bears an obvious relationship to the police interrogation situation in which suspects are questioned, but are then asked leading questions and are given false information while being placed under interpersonal pressure. Thus, the scale has good “face validity”, and it is easy and quick to administer. The scale has the advantage of measuring the impact of both suggestive questions (yield) and interpersonal pressure (shift). The two types of suggestibility appear to be independent in as far as they are poorly correlated and load on separate factors.

A parallel form of the scale using a different story (GSS-2) has also been developed. The GSS-2 narrative contains a non-criminal story content, which describes a couple saving a boy from having a bicycle accident. Both forms have been shown to have high inter-rater reliability, and the correlation co-efficients for the suggestibility measures are high between the two forms. As in the case of the original scale, the “yield” and “shift” items loaded on separate factors on the new form (GSS-2). The two forms of the test have very similar mean and standard deviation scores, which means the two scales can be used interchangeably. The setting (whether prison or hospital) in which the subjects are tested does not appear to significantly influence their suggestibility scores on these scales.

A renowned forensic psychologist, Thomas Grisso, did an early evaluation of the utility of the GSS-1 In his 1986 book, Evaluating Competencies: Forensic Assessments and Instruments . He pointed out that “primary suggestibility” as defined by Eysenck is closely related to hynotizability, but is not closely related to Eysenck’s “secondary suggestibility” (often called gullibility) or interrogative suggestibility. Grisso states that Gudjonsson’s idea to measure “yielding” (response to suggestive questions) and “shifting” (response to negative feedback and interpersonal pressure) was developed from forensic consideration of potential sources of interrogative biasing effects in law enforcement activities.

Grisso suggested that the criminal content of the narrative used in the GGS-1 might interfere with the actual recollection of crime-related facts in individual cases, and it was apparently primarily because of this criticism that the GSS-2 was developed, which does not use criminal content in its narrative. Except for that criticism, Grisso had no critique of the conceptual basis of Gudjonsson’s theory and his use of the GSS-1 as an operational definition of interrogative suggestibility.

Grisso wrote:

The operational definitions of yielding and shifting, as evidence of degree of suggestibility, are quite logical, straightforward, and likely to be perceived by courts as manifesting good conceptual face validity. Overall, the conceptual and operational development of the GSS appears to have involved a very thoughtful process that provides a good foundation for its use and further development.

Grisso also reported that in his opinion the test was properly standardized, although he did suggest a tape-recording of the instructions might be used as a further standardization refinement. Grisso pointed out that the normative sample was obtained from citizens of Great Britain; it appears that even today there are no U.S. norms.

Grisso felt the GSS-1 had good construct validation, in that various predictions from Gudjonsson and Clark’s theoretical paper had been born-out by empirical research studies. These included findings that intelligence and GSS-1 suggestibility scores have been shown to be negatively correlated and significantly related, findings that total suggestibility on the GSS-1 is significantly and positively correlated with Eysenck’s Neuroticism and Eysenck’s social desirability index, and other similar findings, many of which will be reported later in this paper. Grisso concluded:

Much of the research by Gudjonsson and associates used deduction of hypothesis based on theoretical relations between the concept of suggestibility and other personality constructs, then used the GSS as an operational definition of suggestibility to test these hypothesis. This has produced an impressive network of results lending considerable construct validity to the GSS as an index of suggestibility.

Grisso did complain that few studies had been done up until that time (1986) which examined the ability of various GSS scores to predict or postdict the actual degree of suggestibility at the time of police questioning. One study had found that defendants who had consistently denied their involvement in alleged crimes obtained significantly lower GSS-1 scores than did defendants who had first confessed to police and then had retracted their confessions. This study showed that GSS-1 scores had the potential to discriminate between defendants who had responded in various ways to actual interrogation events. This study was criticized by Sharrock , who believed the differences in IQ between the two groups could have accounted for the differences in suggestibility scores. In response, Gudjonsson extended and replicated the original study with a larger number of subjects, and this time the two groups were matched with respect to age, sex, intelligence and memory capacity. Highly significant differences between their GSS-1 scores were recorded between alleged false confessors and resistors even after their intelligence and memory capacity had been controlled for. Thus, intelligence may be an important factor in differentiating between alleged false confessors and resisters, but suggestibility is also important.

In another study Gudjonsson compared the GSS-1 scores of 76 alleged false confessors, 38 forensic patients who had not retracted their confessions, and 15 criminal suspects or defendants who had been able to resist confessing during police interrogation in spite of evidence which subsequently resulted in their convictions. The three groups differed highly significantly after IQ and memory on the GSS-1 were controlled for by an analysis of covariance, with the false confessors significantly more suggestible than the forensic patients, who were themselves significantly more suggestible than the resistors. Indeed, in that study the differences between the three groups were more marked with regard to suggestibility than intelligence, suggesting suggestibility may be a better indicator of how people cope with police interrogation than intellectual functioning.

One study suggests that the GSS-1 is able to some degree to predict the accuracy of witnesses’ testimony during police interviewing. 45 subjects, 30 or whom were intellectually handicapped, were involved in a staged scenario involving the taking away of flower pots; at the time they were not aware they would later be answering questions about the incident. Police officers then video-recorded questioning them about the flower pot incident, and their interviews were scored for accuracy. Gudjonsson reanalyzed the data to compute the correlation between suggestibility and accuracy. Suggestibility correlated negatively with the number of items of accurate information and positively with the amount of erroneous information given.

Another approach to validity is to determine if GSS scores correlate with independent ratings of suggestibility. In one study the GSS-1 was administered to 31 delinquent boys, ages 11-16 years, who had been independently rated by two teachers on suggestibility. The teachers’ behavioral ratings of suggestibility correlated highly significantly with the GSS-1 shift score. Further information about these studies and others that demonstrate the validity of the GSS-1 and GSS-2 is available in the test manual. The GSS-1 has survived a Daubert challenge in a least one appellate Federal case.

In the manual Gudjonsson suggests the GSS-1 and GSS-2 may be appropriate for children as young as six-years-old. A similar test for children ages 4 to 10, the Bonn Test of Statement Suggestibility (BTSS) is under development. However, more large scale studies are needed to determine whether the BTSS is really useful as a forensic tool. Research with the GSS-1 has been done with children as young as six years of age.

What Suspects Would be Most Suggestible?

As we have seen, even highly intelligent college students can be induced to give false confessions when given false information, so even highly intelligent individuals are somewhat “suggestible”. However, all of the research shows that people with low IQ scores are even more suggestible (as measured by the Gudjonsson Suggestibility Scales) than those with higher IQ’s. Clare and Gudjonsson studied people with IQ’s from 57-75, and found them to be much more susceptible to leading questions than normals. They concluded that those with limited intellectual ability are potentially vulnerable to giving erroneous testimony during interrogations, as intelligence and suggestibility are negatively correlated.

Some people have a tendency to answer questions posed in a “yes” or “no” format with the answer “yes” much more frequently than with the answer “no”. This phenomenon has been labeled “acquiescence”, and it refers to the tendency to answer questions affirmatively regardless of content. Acquiescence is especially common among mentally deficient and mentally retarded individuals; research shows that acquiescence is negatively correlated with intelligence, so special precautions against acquiescence need to be employed when they are being interviewed.
Not surprisingly, individuals who are acquiescent also tend to be suggestible as measured by the GSS-1, i.e. suggestibility and acquiescence are correlated.

Suggestibility is also negatively correlated with age, i.e. on average children are more suggestible than adults. The fact that children are highly suggestible when being questioned is well known and well accepted. Numerous studies have been done showing that when children witness events their memories of those events can be greatly influenced by leading questions. Age differences in suggestibility among children are discussed by Ceci and Bruck.
One authority has concluded, “For now, the clear message sent by all of the reported research, some dating back to the turn of the century, is that younger children display a heightened susceptibility to suggestion, even inadvertent suggestion, by a neutral “adult”.

Although few very young children are ever exposed to police interrogation (in many states including Washington a child cannot be charged with a crime until age eight), adolescents frequently are, as they comprise one of the largest groups of accused lawbreakers. Adolescent delinquent boys do about as well as adults in resisting leading questions (yield on the GSS-1), but are more suggestible than adults to negative feedback (shift on the GSS-1) suggesting that delinquent adolescents may be particularly responsive to interpersonal pressure during interrogations. Serious juvenile offenders also have much higher “shift” scores on the GSS-1 than adult offenders, although their “yield” scores are about the same as those of adult offenders. This study provides further confirmation that adolescents are especially vulnerable to interrogative pressure as measured by the tendency to change their previous answers following negative feedback.

It is of course not ethically permissible to do experiments in which adolescents are exposed to harsh interrogation techniques in order to test whether this might cause them to falsely confess. One way to address this problem is to simply ask adolescents how they think they might react to various police interrogation techniques by giving them hypothetical situations. In one study 55 adolescent boys in a juvenile justice facility were told a story of police interrogating an adolescent falsely suspected of having stolen a watch, and having kicked a boy. Twenty-six police behaviors were added to the interrogation (e.g. police told the child if he confessed, they would allow him to leave; police told the child if he did not confess at that time, he would spend the rest of his life in prison; police told the child if he did not confess, his parents would be very disappointed in him.) These techniques were drawn from a popular United States police-training manual. Following the description of each of the 26 police behaviors, the adolescent gave his likelihood (on a one to six scale) of offering a false confession. 24% of the participants said they would definitely offer a false confession in at least one of the interrogation situations, and 42% said they leaned towards offering a false confession in at least one situation. Young age was the most important risk factor for self-reported likelihood of offering a false confession, with 13, 14 and 15 year-olds having much higher average false confession scores than 16,17 and 18 year-olds.

Clearly interrogation at a police station is stressful, although perhaps more stressful for some subjects than for others. In one study 161 suspects detained at two police stations were assessed prior to being questioned by the police. They were given the Gudjonsson Suggestibility Scale-2 (GSS-2), 3 sub-tests of an IQ test (WAIS-R), and the Spielbeiger State-Trait Anxiety Inventory (STAI), which consists of two 20 item self-reported rating scales for measuring State and Trait anxiety, respectively. State anxiety refers to a transitory feeling of tension or apprehension and is measured by the subjects describing the intensity of their feelings of distress and discomfort at a particular moment in time. Trait anxiety, on the other hand, refers to relatively stable individual differences in anxiety proneness, and is thought to be relatively independent of situational stress. Trait anxiety correlated highly with suggestibility, but there was a lesser correlation between state anxiety and suggestibility scores. Interestingly, Afro-Caribbean suspects were significantly more suggestible than the Caucasian suspects even after their GSS-2 memory and IQ scores were controlled for.

In another similar study the subjects were a group of anxiety disorder patients and normal adults. Suggestibility as measured by the GSS-2 correlated significantly with both state and trait anxiety as measured by the STAI, and anxious patients scored higher on the suggestibility factors than normal controls. Anxious patients also had higher “dissociation” and “absorption” scores than normal controls.

A related finding is that low self-esteem (as measured by Semantic Differential Scales) is also correlated with suggestibility as measured by the GSS-1. Other personality characteristics that have been shown to correlate with interrogative suggestibility as measured by the GSS-1 are perceptual defensiveness and belief in witchcraft. Research shows that there is a correlation between sleep deprivation and suggestibility. Suggestibility is negatively correlated with a suspect’s number of previous contacts with the justice system. The more convictions, the less suggestible the suspect, particularly in regard to “shift” scores .

Conclusion

All of the discussion above suggests strongly that the police interrogators should be very careful to avoid asking leading or suggestive questions when doing interrogations. This is important for all interrogations, but is especially important among groups shown to have high interrogative suggestibility, such as children and adolescents, mentally retarded and mentally deficient, highly acquiescent, and highly anxious suspects. One important approach with these types of suspects is to use the “cognitive interview”, a method based on findings and ideas established in controlled studies of memory which specifically avoids the use of suggestive or leading questions. Many experiments have compared the cognitive interview to standard police interview techniques, and all of them find that the cognitive interview yields gains in witness recall. Although many U.S. police departments do not audiotape or videotape their interrogations, tape-recording is now standard practice in England .

Many police, including all police forces in England and Wales, now receive training in the cognitive interview, and several features of the cognitive interview are included in the guidelines for collecting eyewitness evidence developed by the Justice Department.

Police should also realize that in many cases their work is not done once they have obtained a confession. After obtaining the confession the police still have an obligation to corroborate the confession. If the suspect truly committed the crime the police should be able to find aspects of the evidence which show that the suspect who confessed really did commit the crime.

Clinical psychologists are increasingly being asked to evaluate cases in which confessions have been retracted. The Gudjonsson Suggestibility Scales will be helpful in assessing whether a given confession might have been “coerced” or “involuntary”. Defense attorneys who suspect that confessions were false confessions produced by police interrogation and suggestibility on the part of their client might consider hiring a psychologist to measure the client’s level of suggestibility using the Gudjonsson Suggestibility Scales. If the client tested to be highly suggestible an attack on the reliability of the confession could be considered, using the psychologist as an expert witness.

During an individual forensic assessment the suggestibility score needs to be compared with those of the appropriate normative group for the general population, because courts will need to know whether or not the obtained scores fall within or outside normal limits for the general population. For purposes of a forensic assessment, the scores obtained on any one psychological test should not be interpreted in isolation from other information. Obviously IQ data, achievement test data, and personality assessment are also important. Furthermore, third-party information including police reports and tapes of interrogation sessions should be reviewed, as well as medical records, school records, and records of prior mental health treatment. During the clinical interview the psychologist should take a detailed history, assess the defendant’s psychological functioning and obtain the defendant’s account of what happened during the interrogation process. An assessment of alcohol or drug abuse is relevant, particularly if the defendant was intoxicated at the time of the interrogation. Determining the relevance of the test scores to the individual case and interpreting them appropriately in the overall context of the case is the most difficult part of the forensic assessment.

Suggestibility is only relevant if the individual was actually asked leading questions or was pressured during the interrogation. Accordingly, even significant mental impairment will not automatically render a confession involuntary absent some showing of police misconduct. One U.S. Supreme Court case holds that unless the police somehow used “coercion” to take advantage of the suspects “vulnerability” the confession will be ruled valid since there was no police conduct “casually related to the confession”. In other words, there is no requirement that in order for a confession to be voluntary, it must be a product of a free will; coercive police activity is also a necessary predicate to a finding that a confession was not voluntary within the meaning of the due process clause of the Fourteenth Amendment. On the other hand, another U.S. Supreme Court case holds that it is important that courts take into account characteristics of the detainee which affect the extent to which his free will is likely to have been overborne. Thus, the psychologist must evaluate whether the behavior of the police in interaction with the suspect’s susceptibility could have led to an “involuntary” confession. Relevant issues would include whether the disability, if any, would have been apparent to a normal policeman, or whether the police were told about the disability before questioning. The expert may also wish to address the body of literature pertaining to the amount of psychological coercion resulting form various types of interrogation methods.

One obvious problem is that the suggestibility scores will likely be obtained weeks or months after the confession event. This can make the assessment difficult because the defendant may have changed in the interviewing time period. Another issue is whether the defendant might malinger his responses on the GSS-1 or GSS-2. Although Gudjonsson feels the tests are subtle enough that their true purposes (assessing “yield” in response to leading questions and “shift” in response to negative feedback) cannot be discerned, obviously as the use of the tests becomes more widespread, lawyers could begin to counsel their clients to appear more “suggestible”, and the tests have no built-in scale to measure malingering. Thus, administering a test specifically designed to detect malingering, such as the Validity Indicator Profile (VIP), the Test of Memory Malingering (TOMM), or the Victoria Symptom Validity Test (VSVT) might be appropriate.

Voluntary false confessions are most often made by schizophrenics, or by those who feel extraordinarily guilty because of psychotic depression. Coerced-compliant false confessions do not involve the suspect actually believing in his own guilt, and are thus mediated more by “compliance” than by “suggestibility” . The GSS-1 and GSS-2 are most relevant to cases of alleged coerced-internalized false confessions. However, the different types of false confessions should not be viewed as exclusive categories. For example, suspects may be partially persuaded by the police that they were involved in the crime but may nevertheless confess mainly to escape from an intolerable situation.

It is clear there is no “cut-off score” on the GSS-1 or GSS-2 which indicates a given confession was “involuntary” or “unreliable”, and psychologists should probably never provide testimony to the ultimate issue in the case by stating the defendant’s confession was “false”, since there is usually no way of determining the “ground truth” of a disputed confession. Similarly, a psychologist should not reach the issue of whether a given confession was “coerced” or “involuntary”. Instead, psychologists should point to reasons why the confession might have been involuntary, coerced and/or unreliable under the circumstances.