Effects of personality, interrogation techniques and plausibility in an experimental false confession paradigm

Authors


Correspondence should be addressed to Jessica R. Klaver, Department of Psychology, Simon Fraser University, 8888 University Drive, Burnaby, BC, V5A 1S6, Canada (e-mail: jklavera@sfu.ca).

Abstract

Purpose. The goal of the present study was to investigate the effects of personality variables, interrogation techniques and the plausibility level of an alleged transgression on the experimental elicitation of false confessions.

Methods. Two hundred and nineteen undergraduate students assessed on measures of compliance, self-esteem, locus of control and interrogative suggestibility participated in the Kassin and Kiechel (1996) paradigm. Experimental manipulations included minimization and maximization interrogation techniques and high and low plausibility of the alleged typing mistake to examine rates of false confession and internalization.

Results. The overall false confession and internalization rates across all conditions were 43 and 10%, respectively. An increased likelihood of false confession behaviour was associated with higher Shift scores on the Gudjonsson Suggestibility Scale, the use of minimization interrogation techniques and an increase in the plausibility of the allegation. Females were more likely to falsely confess than males in the high plausibility condition, whereas Caucasian and Asian participants were equally likely to falsely confess. Personality variables, such as compliance, most influenced the behaviour of males and Asians.

Conclusions. The results of this study offer insight into false confession behaviour, suggesting that individuals who have a tendency to change their responses in the face of negative feedback may be more prone to false confession behaviour. The findings also serve to highlight the dangers of using minimization interrogation techniques and elucidate the limited generalizability of the paradigm to situations in which the alleged transgression is less plausible.

A confession has traditionally been viewed as the most influential type of evidence in criminal proceedings (McCormick, 1972; Wigmore, 1970). In simulated juror studies, confession evidence has demonstrated a stronger impact on verdicts than eyewitness testimony or character evidence (Kassin & Neumann, 1997). Further, the mere presence of a confession can increase the rate of guilty verdicts regardless of knowledge that the confession is involuntary and instructions to disregard it when making verdict decisions (Kassin & Sukel, 1997).

In real-life criminal settings, most suspects who confess during police interrogations are guilty and most confessions are corroborated (Leo & Ofshe, 1998). However, numerous cases of wrongful conviction resulting from false confessions have been documented (Bedau & Radelet, 1987; Borchard, 1932; Leo & Ofshe, 1998; Radelet, Bedau, & Putnam, 1992; Rattner, 1988). The prevalence of false confessions is unknown, but growing evidence points to an alarming rate of occurrence, with false confessions now recognized as one of the leading sources of erroneous convictions of innocent individuals (Connors, Lundregan, Miller, & McEwan, 1996; Drizin & Leo, 2004; Scheck, Neufeld, & Dwyer, 2000). Given the powerful sway of confession evidence and hence the perilous influence of false confessions, it is crucial to examine factors that may be involved in producing false confessions. This study aimed to investigate individual and situational factors that may contribute to the experimental elicitation of false confessions.

Individual variables related to false confessions

False confessions generally arise in the context of complex social interactions, resulting from a combination of the individual psychological makeup of the suspect and situational factors related to the police interrogation setting (Kassin & Gudjonsson, 2004). Several demographic, cognitive, personality, mental health and physiological factors have been identified theoretically or empirically as relevant to the generation of false confessions.

False confessions have been linked to younger age (e.g. adolescence, Brandon & Davies, 1973; Goldstein, Condie, Kalbeitzer, Osman, & Geier, 2003) low intelligence or mental retardation (Brodsky & Bennett, 2005; Clare & Gudjonsson, 1993, 1995; Fulero & Everington, 2004; Gudjonsson & MacKeith, 1994; Perske, 2000; Redlich, 2004), mental disturbance (Irving, 1980; Irving & McKenzie, 1989; Redlich & Appelbaum, 2004) and physiological factors, such as sleep deprivation (Blagrove, 1996) and alcohol or drug intoxication and withdrawal (Davison & Gossop, 1996; Gudjonsson, Hannesdottir, Peteursson, & Bjnornsson, 2002; Pearse, Gudjonsson, Clare, & Rutter, 1998; Santtila, Alkiora, Magnus, & Neimi, 1999; Sigurdsson & Gudjonsson, 1995, 1996a, 2001).

False confessions have also been associated with personality variables, such as antisocial personality characteristics (Gudjonsson, Sigurdsson, Bragason, Einarsson, & Valdimarsdottir, 2004; Gudjonsson, Sigurdsson, & Einarsson, 2004; Gudjonsson, Sigurdsson, Finnbogadottir, & Jakobsdottir Smari, 2006; Sigurdsson & Gudjonsson, 1996b, 1997, 2001), anxiety (Gudjonsson, 1999a, 1999b), depression (Gudjonsson, Sigurdsson, Asgeirsdottir, & Sigfusdottir, 2006), compliance (Gudjonsson, 1990, 1991, 1999a, 1999b), suggestibility (Gudjonsson, 1990, 1991, 1992, 1993; Trowbridge, 2003) and low self-esteem (Gudjonsson, 1999b; Gudjonsson, Sigurdsson, Bragason, et al., 2004). One type of suggestibility that has been specifically linked to false confession behaviour is interrogative suggestibility, which refers to the ‘extent to which, within a closed social interaction, people come to accept messages communicated during formal questioning, as the result of which their subsequent behavioural response is affected’ (Gudjonsson & Clark, 1986, p. 84).

Although many individual variables have been linked to false confession behaviour, there are several others that have yet to be examined. For example, locus of control (Rotter, 1966) may be an important factor influencing false confession behaviour. Specifically, individuals with an external locus of control, who believe that powerful others, fate, or chance primarily determine events, may be at differential risk for false confession in comparison with those with an internal locus of control, who believe that events result primarily from their own behaviour and actions. Further, false confession behaviour has yet to be comprehensively investigated across different gender or ethnic groups.

Influence of the police interrogation setting

Although many individual factors are important in terms of increasing vulnerability to falsely confessing, situational variables also exert a large influence on the likelihood of false confessions. One such contextual factor is the police interrogation setting, during which the primary goal of investigators is to obtain a confession from a criminal suspect. In recent decades, interrogation strategies have become less physical and more psychological in nature (Leo, 2004; Skolnick & Leo, 1992). This change is reflected in contemporary police interrogation training manuals, such as the popular Criminal Interrogation and Confessions (Inbau, Reid, Buckley, & Jayne, 2001), which offers a variety of recommendations for obtaining confessions.

Police interrogation manuals justify deceitful tactics as necessary tools for criminal interrogation (Underwager & Wakefield, 1992; Wakefield & Underwager, 1998). Such techniques are widely used by police investigators in North America and have generally been considered legally admissible (Frazier v. Cupp, 1969; R. v. Oickle, 2000). Observational studies of police interrogations have revealed that psychologically manipulative tactics are commonly used in modern-day police practice (Leo, 2004). Common interrogation strategies include undermining the suspect's confidence in a denial of guilt, appealing to the importance of cooperation, offering moral justifications or psychological excuses, confronting suspects with false evidence of their guilt, using praise and flattery, appealing to the expertise or authority of the detective, appealing to the suspect's conscience and minimizing the moral seriousness of the offense.

Kassin and colleagues (Kassin, 1997; Kassin & McNall, 1991) summarized the interrogation strategies recommended by Inbau as falling into two general categories: maximization and minimization. Maximization strategies intimidate the suspect by using tactics such as making false claims about evidence and exaggerating the seriousness of the charges. Minimization strategies give the suspect a false sense of security using sympathy, flattery, offering legal or moral face-saving excuses for actions, conceptualizing actions as accidental, blaming the victim and underplaying the seriousness of the charges. Research has demonstrated that minimization techniques lead individuals to believe that they will receive leniency for confessing, even when it is not explicitly promised (Kassin & McNall, 1991).

Inbau et al. (2001) argue that their interrogation strategies rarely lead to false confessions because the police typically do not interrogate innocent suspects. Although the possibility of false confessions is recognized, it is presumed that their potential is minimized because police can distinguish truths from lies at high rates of accuracy. However, there appears to be little empirical evidence to support their claim that interrogators are able to distinguish guilty and innocent suspects (Kassin, 2005; Meissner & Kassin, 2002; Vrij, 2000). Because interrogation techniques used to obtain confessions from guilty suspects are similar to those that produce false confessions in some innocent suspects (Drizin & Colgan, 2004; Gudjonsson & MacKeith, 1988; Ofshe, 1989; see also Henkel & Coffman, 2004), innocent individuals subjected to powerful and psychologically manipulative interrogation techniques are at risk for false confession.

Experimental investigation of false confession behaviour

Research on false confessions has primarily comprised analyses of cases involving disputed confessions or known false confessions Gudjonsson, 1999a, 1999b). Considerable ethical constraints preclude the manipulation of variables in real-world settings to study false confessions. However, a paradigm has been developed to investigate the phenomenon experimentally (Kassin & Kiechel, 1996). Participants are required to type letters on a computer keyboard at a fast rate, but warned that pressing the ALT key on the keyboard will cause the computer to malfunction and the data to be lost. After 1 minute, an experimenter inconspicuously reboots the computer, feigns distress and accuses the participant of pressing the forbidden key by asking, ‘Did you hit the ALT key?’

If the participant denies the allegation, general interrogation statements are used to induce the participant to sign a written confession. Shortly thereafter, the participant encounters a confederate upon returning to the waiting room. The dependent variables of interest in this design are false confession and internalization. A false confession is operationalized as whether or not the participant signs the written statement admitting guilt and internalization is coded as whether or not the participant accepts personal responsibility for pressing the ALT key in response to the private inquiry by the confederate.

This computer paradigm has been remarkably successful in eliciting false confessions and internalization in adolescents and undergraduate students and is thus promising in terms of exploring factors influencing false confession behaviour. In the original study (Kassin & Kiechel, 1996), 69% of participants signed the false confession and 28% internalized responsibility. Further, in a condition in which participants were instructed to type at a faster speed and were presented with a false witness, 100% of participants falsely confessed to pressing the ALT key.

Researchers have assessed personality variables and introduced various manipulations into the paradigm in order to investigate factors potentially related to false confession behaviour. Forrest and colleagues (Forrest, Wadkins, & Larson, 2006; Forrest, Wadkins, & Miller, 2002) have investigated the roles of gender, stress and personality variables in the paradigm. Females were more likely to falsely confess and internalize responsibility than males and males had higher internalization rates in stressful conditions compared with the conditions without stress. False confession was related to an increased susceptibility to leading questions on a measure of interrogative suggestibility and internalization was related to the personality variables of external locus of control, anxiety and authoritarianism.

Horselenberg and colleagues (Horselenberg, Merckelbach, & Josephs, 2003; Horselenberg et al., 2006) found high false confession rates in the paradigm when false evidence was presented by an authority figure in addition to an immediate negative consequence of monetary loss. Further, fantasy proneness scores distinguished false confessors from nonconfessors. Candel, Merckelbach, Loyen, and Reyskens (2005) used the paradigm with a sample of young children, finding that, of the 36% of children who falsely confessed, 89% also internalized responsibility.

Redlich and Goodman (2003) examined the impact of age, interrogative suggestibility and the presentation of false evidence in the paradigm. Younger participants (aged 12–13 and 15–16) were more likely to sign the false confession than college students, with the 15- and 16-year olds being especially influenced by the presentation of false evidence. Additionally, participants who had a tendency to succumb to leading questions were more likely to sign the false confession.

Russano, Meissner, Narchet, and Kassin (2005) designed a novel paradigm to investigate both true and false confession behaviour. Interrogation techniques and the offer of a ‘deal’ were manipulated in conditions where cheating did or did not take place on a problem-solving task. Implying leniency by using minimization interrogation techniques increased the likelihood of false confession in comparison with a condition without interrogation. Further, participants presented with a deal explicitly offering leniency in terms of consequences were more likely to falsely confess than those who were not offered the deal.

The present study

The goal of the present study was to further examine factors that may be related to the experimental elicitation of false confessions using the Kassin and Kiechel (1996) paradigm. First, although there is an accumulating body of theoretical and clinical evidence supporting the potential impact of personality variables on confession behaviour (Gudjonsson, 2003), researchers have yet to consistently demonstrate these effects empirically using this paradigm (e.g. Horselenberg et al., 2003). In this study, the effects of compliance, interrogative suggestibility, self-esteem and locus of control on false confession and internalization rates were examined.

Second, there is little theoretical or empirical consensus on the effects of gender and ethnicity on false confession and internalization behaviour (Forrest et al., 2002; Gudjonsson & Sigurdsson, 1994; Gudjonsson, Sigurdsson, Bragason, et al., 2004). This study aimed to investigate differences in false confession and internalization rates in males vs. females and across ethnicities to determine whether there were gender- or ethnicity-specific differences in scores on personality measures related to false confession and internalization behaviour.

Third, it is unclear which types of interrogation techniques may differentially impact false confession and internalization behaviour. To date, only a mix of general interrogation statements has been used to pressure participants to sign the written confession in the paradigm. This study used interrogation statements adapted from Forrest et al. (2002) to compare the effects of minimization and maximization interrogation techniques, as described by Kassin (1997).

Finally, the role of plausibility in the paradigm is unknown. The original experimental design calls for the computer to crash immediately following instructions to press a key adjacent to the forbidden ALT key. Thus, pressing the ALT key is a highly plausible, unintentional and momentary transgression likely to be admitted and internalized by many. In this study, the plausibility of the alleged act was manipulated by adding a condition in which participants were accused of pressing the ESC key, one markedly farther from the typing area. In this way, the generalizability of the paradigm to circumstances involving less plausible allegations was examined.

Method

Participants

Participants were 219 undergraduate students at a western Canadian university. The modal age of participants was 18 years (range 18–45 years). Participants were 79% (N = 174) female and 21% (N = 45) male. The ethnicity of the sample varied, including 49% (N = 107) Caucasian, 36% (N = 80) Asian and 15% (N = 31) identified as another ethnicity. Thirty-four per cent (N = 74) of participants indicated English as a second language and the mean self-reported English fluency of these students was 8.3 out of 10.

Materials

Gudjonsson Suggestibility Scale (GSS; Gudjonsson, 1987, 1997)

The GSS is a memory task designed to measure interrogative suggestibility. Measures on the GSS include Memory Recall, Yield 1, Yield 2 and Shift. The GSS involves an experimenter reading aloud a short story and then asking the participant to recall everything he or she remembers from the story. The Memory Recall score is the number of story elements recalled correctly. Following the recall, the experimenter asks 20 questions related to the story, 15 of which are leading questions. Yield 1 is the number of leading questions answered in the affirmative by the participant.

After the 20 questions have been answered, the experimenter gives the participant ‘negative feedback’ by stating that the questions must be repeated because several responses were incorrect. The 20 questions are then repeated and Yield 2 is the number of affirmative responses to leading questions following the negative feedback. Total Yield 1 and Yield 2 scores range from 0 to 15. The Shift score is the number of distinct changes in the nature of any reply to a question following the administration of the negative feedback.

Total Shift scores range from 0 to 20. Mean GSS Recall, Yield 1, Yield 2 and Shift scores were 19.3 (SD = 5.6), 5.2 (SD = 3.3), 7.3 (SD = 4.0) and 4.5 (SD = 2.9), respectively. On the GSS Recall, females (M = 19.8, SD = 5.4) scored higher than males (M = 17.1, SD = 5.8), t (201) = 2.85, p = .005, and Caucasians (M = 20.6, SD = 5.6) scored higher than Asians (M = 17.4, SD = 5.1), t (170) = 3.86, p = <.001. On GSS Yield 1, Asians (M = 6.1, SD = 3.2) scored higher than Caucasians (M = 4.6, SD = 3.4), t (170) = 2.84, p = .005, indicating a greater tendency to acquiesce to suggestive questions.

Gudjonsson Compliance Scale (GCS; Gudjonsson, 1989, 1997)

The GCS is a 20-item self-report questionnaire measuring the tendency to conform to requests made by others, particularly authority, to please them or avoid conflict and confrontation. Each statement is endorsed as true or false and total GCS scores range from 0 to 20, with higher GCS scores indicating higher levels of compliance. The mean GCS score was 9.0 (SD = 3.8) and the internal consistency (Cronbach's α) of the measure was .77. On the GCS, Asians (M = 10.8, SD = 4.1) scored higher than Caucasians (M = 8.1, SD = 3.2), t (183) = 4.96, p <.001.

Rosenberg Self-Esteem Scale (Rosenberg, 1965)

The Rosenberg Self-Esteem Scale is a 10-item self-report scale measuring feelings of self-worth. Each item is scored on a 4-point Likert scale. Scores on this scale range from 0 to 30, with 30 indicating the highest level of self-esteem. The mean Self-Esteem Scale score was 22.6 (SD = 4.8) and the internal consistency of the measure was .86. Ratings of self-esteem were higher for males (M = 24.0, SD = 4.2) compared with females (M = 22.2, SD = 4.9), t (215) = 2.20, p = .029, and higher in Caucasians (M = 23.0, SD = 4.9) compared with Asians (M = 21.3, SD = 4.9), t (183) = 2.38, p = .018.

The Rotter Internal–External Locus of Control Scale (Rotter, 1966)

The Rotter Locus of Control Scale, presented to participants as an inventory of personal beliefs, is a 29-item self-report scale assessing the extent to which an individual feels in control of his or her life circumstances. Items on the scale are forced-choice, with one response reflecting a belief in internal control and the other a belief in external control. The total score, which ranges from 0 to 23, is the number of external choices endorsed. Thus, a high score on the scale indicates an external locus of control, whereas a low score indicates an internal locus of control. The mean Locus of Control Scale score was 11.2 (SD = 3.8) and the internal consistency of the measure was .72. Females (M = 11.5, SD = 3.8) scored higher than males (M = 9.8, SD = 3.7), t (215) = 2.72, p = .007, indicating a more external locus of control.

Procedure

Participants were invited to take part in a task investigating their personality, memory and typing skills. Informed consent was obtained from participants and all were treated in accordance with ethical standards set forth by the American Psychological Association. In small groups, participants completed several personality questionnaires, including the Rosenberg Self-Esteem Scale, Rotter Locus of Control Scale and the Gudjonnson Compliance Scale. They also completed a demographics questionnaire inquiring about age, gender, ethnicity, English fluency and educational background. As they finished the questionnaires, participants were led individually to a different room in order to complete the GSS and the typing task with a male Caucasian experimenter. The GSS was presented as a test of memory skills and read aloud by the experimenter. After completion of the GSS, participants were informed that the final activity of the session involved a typing test examining their ability to type letters quickly onto a computer keyboard.

The typing task used in this study was adapted from the paradigm developed by Kassin and Kiechel (1996). A metronome was used to ensure that letters were consistently read at a rate of 67 letters per minute. Two experimental manipulations were introduced into the design. First, the plausibility of the alleged act was varied. In the high plausibility condition, participants were warned not to press the ALT key because doing so would cause the program to crash and data to be lost. In this condition, a replication of Kassin and Kiechel, the reading of the letters aloud is designed so that the participant is instructed to press the Z key immediately preceding the crashing of the computer. Due to the close proximity of the Z and the ALT keys on the keyboard, pressing the ALT key is a highly plausible act during this moment of the task. In the low plausibility condition, participants were warned not to press the ESC key. The ESC key, located in the upper left of the keyboard, was considerably farther from the Z key, thus reducing the plausibility that it could be pressed during the course of the typing task.

Second, statements adapted from Forrest et al. (2002) reflecting two different types of interrogation strategies were used to induce the participant to sign the written statement falsely confessing to pressing the forbidden key. In the minimization condition, the experimenter offered statements designed to give the participant a false sense of security about the apparent transgression by normalizing the act, deeming it an accident and blaming the computer. The following statements were read, in order, until either the participant agreed to sign the confession or all the statements had been read: ‘Don't worry. It was just an accident. You didn't mean to hit the ALT/ESC key. Several participants so far have pressed the ALT/ESC key during this task. Are you sure you didn't press it?’ ‘This program seems not to be working lately. The ALT/ESC key is too sensitive, and registers even when it is barely pressed. Is that what happened?’

In the maximization condition, the experimenter used tactics designed to induce feelings of intimidation and exaggerate the seriousness of the alleged transgression. The experimenter read the following statements, in order, until either the participant agreed to sign the confession or both statements had been read: ‘We have run over 50 people in the past three weeks. The computer hasn't crashed any of those times. I know that the only time it crashes is when the ALT/ESC key is pressed. You must have pressed it, didn't you?’ ‘That file contained all the data collected so far in this study. There is no way to recover the data. It looks like the entire project may be delayed now. Why did you press the ALT/ESC key?’

In all conditions, participants were asked to sign a written confession stating, ‘I pressed the ALT key and caused the computer to crash. All data were lost.’ Regardless of whether the written confession was signed, the experimenter stated that, as a result of the computer problems, the experimental session would need to be rescheduled in order for the participant to be awarded course credit. The experimenter indicated that he needed to exit the room in order confer with the project supervisor. After the experimenter exited the room, a third experimenter, a female Caucasian confederate posing as a student in the area, entered the study room stating, ‘I heard a lot of noise. What happened in here?’ The response of the participant was recorded verbatim.

The two dependent measures recorded in this study were false confession and internalization, both coded dichotomously (present or absent). A false confession was evaluated according to whether or not the participant signed the written confession. Internalization was measured from responses to the inquiry made by the confederate. In order to reduce bias due to the confederate not being blind to the experimental conditions, the criterion for internalization was an unambiguous response by the participant clearly assuming full personal responsibility for pressing the forbidden key during the computer task. For example, if the participant answered in the affirmative (e.g. ‘I pressed the ALT/ESC key and the computer turned off’) the response was coded as internalization; however, if the participant indicated uncertainty (e.g. ‘I think so’, ‘I'm not sure’, ‘maybe’ or ‘he said I pressed the ALT/ESC key’), the response was not coded as internalization.

Immediately following the conversation with the confederate, each participant was thoroughly debriefed in person as to the nature of the experimental design and the necessity of the deception involved. Participants were reassured that no additional sessions would need to be rescheduled in order for course credit to be obtained. No participants reported any serious adverse effects resulting from participation in the study protocol. Most expressed relief that they had not in fact caused any damage to the computer. Following verbal debriefing, participants were asked to reconfirm their consent to the use of their data by signing their name to the statement, ‘I understand the true purpose of the study I have completed and consent to the use of my data.’ All participants agreed to this request. In order to maintain the integrity of the experimental design, participants were asked to refrain from discussing the details of the study with other potential participants until the estimated completion date of the study. None of the participants indicated during debriefing that they were aware in advance of the true purpose of the study. Upon completion of the session, participants were thanked and received course credit for their psychology class.

Results

False confession and internalization

All but a small number of participants (N = 4) initially denied the allegation that they had pressed the forbidden key; the data for the individuals who did not deny and thus were not subjected to interrogation was removed from subsequent analyses. The overall false confession and internalization rates across all conditions were 43 and 10%, respectively. In the high plausibility condition in which participants were accused of pressing the ALT key, a replication of Kassin and Kiechel (1996), 59% of participants signed the false confession and 16% internalized responsibility. In the low plausibility condition, 13% of participants signed the false confession and none internalized responsibility. Not surprisingly, false confession and internalization behaviour were positively correlated, φ = .39, p <.001. Also as expected, there were no cases in which an individual internalized but did not sign the false confession. Frequencies and percentages of false confession and internalization for all conditions are presented in Table 1.

Table 1.  Frequency and Percentage of False Confession and Internalization by Plausibility and Interrogation ConditionThumbnail image of

A logistic regression was used to determine the impact of gender, ethnicity, personality variables, plausibility level and interrogation technique on the likelihood of false confession (Table 2). The model was significant, χ2(10) = 66.10, p <.001, with 73% of participants correctly classified. An increased likelihood of false confession was significantly related to GSS Shift scores, the plausibility level of the alleged act and the type of interrogation technique used. Specifically, each 1-point increase in Shift score was associated with a 1.26 times higher likelihood of signing the false confession. Participants accused of pressing the ALT key in the high plausibility condition were 16.24 times more likely to sign the false confession when compared with those accused of pressing the ESC key in the low plausibility condition. Participants subjected to minimization interrogation techniques were 4.31 times more likely to sign the false confession than those subjected to maximization interrogation techniques.

Table 2.  Logistic Regression Results for Variables Predicting the Likelihood of False ConfessionThumbnail image of

Another logistic regression was used to determine the impact of gender, ethnicity, personality variables, plausibility level and interrogation technique on the likelihood of internalization. This model was non-significant with no variables influencing internalization rates.

Gender and ethnicity differences in false confession and internalization rates

To examine gender and ethnicity differences on behaviour in the paradigm, false confession and internalization rates were compared between males and females and between Caucasians and Asians for all conditions. Gender- and ethnicity-specific differences in scores on personality measures in relation to false confession and internalization behaviour were also examined in all conditions.

There were no overall gender differences in false confession or internalization rates. However, there was an overall trend suggesting a higher false confession rate in females (46%) when compared with males (31%), χ2(1) = 3.0, p = .08. Further, in the high plausibility condition in which participants were accused of pressing the ALT key, the false confession rate was higher in females (65%) than in males (39%), χ2(1) = 5.9, p = .02, rpb = .21. Males who signed the false confession scored higher on compliance (M = 10.3, SD = 4.9) when compared with those who did not sign the false confession (M = 7.4, SD = 3.6), t (37) = 2.1, p = .04. Males who internalized had higher Yield 1 scores (M = 8.8, SD = 4.4) when compared with those who did not internalize (M = 5.1, SD = 3.0), t (37) = 2.3, p = .03. Also, males who internalized had higher compliance scores (M = 13.3, SD = 6.1) than those who did not internalize (M = 7.8, SD = 3.6), t (37) = 2.7, p = .01.

There were no overall ethnicity differences in false confession or internalization rates between Caucasians and Asians, the two most highly represented ethnic groups. Caucasians who signed the false confession had higher Shift scores (M = 5.4, SD = 3.3) when compared with those who did not sign (M = 3.9, SD = 3.1), t (96) = 2.3, p = .03. Asians who signed the false confession had higher compliance scores (M = 5.4, SD = 3.3) than those who did not sign the false confession (M = 12.4, SD = 4.2), t (72) = 2.7, p = .01. Asians who signed the false confession also had a more external locus of control (M = 13.4, SD = 3.5) when compared with those who did not sign the false confession (M = 10.3, SD = 3.8), t (72) = 3.6, p = .001. Additionally, Asians who signed the false confession had lower self-esteem scores (M = 19.8, SD = 4.8) when compared with those who did not sign the false confession (M = 22.0, SD = 4.5), t (72) = 2.0, p = .04. Asians who internalized responsibility had higher compliance scores (M = 14.4, SD = 4.1) than those who did not internalize (M = 10.8, SD = 3.9), t (72) = 2.0, p = .05.

Discussion

The goal of the present study was to evaluate the impact of various individual and situational factors on false confession and internalization behaviour in an experimental false confession paradigm. The results of this study further demonstrate the ability of the paradigm to successfully elicit false confession and internalization behaviour in an undergraduate sample. The overall false confession and internalization rates in the current study were 43 and 10%, respectively. In the high plausibility condition in which participants were accused of pressing the ALT key, a replication of Kassin and Kiechel (1996), the false confession and internalization rates were 59 and 16%, respectively.

The internalization rate in this condition was lower than the 28% found in the original study. This discrepancy may be attributable to a more stringent criteria used for coding internalization in the current study. Specifically, internalization was coded only for a clear acceptance of responsibility for pressing the forbidden key during the typing task. Any ambiguity in the response was coded as no internalization. This result may also be due to the varied ethnic composition of the sample in this study, as internalization frequencies were almost three times lower for Asian (N = 5) when compared with Caucasian participants (N = 14), although this difference was not statistically significant.

Interrogation techniques, plausibility level, and false confession behaviour

In contrast to past studies utilizing a mixed list of general interrogation statements to pressure participants into signing a written false confession, this study directly compared two distinct sets of interrogation techniques – minimization and maximization – modelled after those recommended and used in real-life criminal interrogation settings (Inbau et al., 2001; Kassin, 1997). Overall, false confession rates were highest when participants were pressured to admit to a highly plausible transgression while being subjected to minimization interrogation techniques. Compared with maximization statements, participants were over four times more likely to sign the false confession if minimization techniques were used to induce the confession. Thus, in a controlled laboratory setting, undergraduate students were more likely to sign false confessions admitting to pressing a forbidden computer key when interrogated by an experimenter who feigned sympathy and blamed external sources. This is consistent with recent research by Russano et al. (2005) who used a novel paradigm, described earlier, to demonstrate that minimization techniques reduced the diagnostic value of an elicited confession by increasing the rate of false confessions.

Despite important differences between real-life circumstances and the paradigm used in this study, the dangers of using minimization interrogation techniques deserve attention. The use of such techniques may be a powerful situational factor impacting the likelihood of false confession behaviour. Using minimization techniques, investigators take advantage of natural defence mechanisms, such as rationalization and projection, used by individuals to justify or minimize transgressions. A seemingly sincere and empathic investigator rationalizes the criminal act, projects blame onto others, minimizes the seriousness of the crime and frames a confession as an opportunity for the suspect to tell his or her story. Thus, suspects are offered a dignified way to ‘save face’ while admitting their involvement in a crime.

In the experimental false confession paradigm, participants are typically accused of causing the computer to crash by pressing the ALT key during the typing task, a highly plausible action. Thus, in order to examine the role of plausibility in the paradigm, a condition was added in which participants were accused of pressing the ESC key, one markedly further from the typing area of the keyboard. Compared with those accused of the highly plausible act of pressing the ALT key, participants accused of the less plausible act of pressing the ESC key were more than 16 times less likely to sign the false confession. Further, no participant accused of pressing the ESC key internalized responsibility for the alleged transgression.

This finding indicates a striking effect of plausibility in the paradigm and is consistent with the literature on event plausibility. For example, Pezdek and colleagues (Pezdek, Finger, & Hodge, 1997; Pezdek & Hodge, 1999) have demonstrated that events can be suggestively planted in memory to the degree that the suggested event is plausible. In both children and adults, implausible false childhood memories were recalled significantly less often than plausible false memories. Thus, the generalizability of the current paradigm may be limited to conditions in which the alleged act is highly plausible.

Impact of personality, gender and ethnicity on false confession and internalization

Personality, gender and ethnicity were investigated as individual factors potentially related to behaviour in the paradigm. Interestingly, most personality variables, including compliance, self-esteem, locus of control and various indices of interrogative suggestibility, were unrelated to false confession or internalization behaviour. This is in line with previous research indicating a lack of relationship between personality measures and behaviour in the paradigm (Horselenberg et al., 2003, 2006).

However, consistent with Redlich and Goodman (2003), participants with higher Shift scores were more likely to sign the false confession. Interestingly, there was some evidence that this finding was specific to Caucasian participants. Thus, individuals who have a tendency to change their responses in the face of negative feedback may be more prone to false confession behaviour. The emergence of a consistent relationship between Shift scores and false confession behaviour in the paradigm supports the utility of this score as an indicator of vulnerability to interrogation techniques involving the presentation of negative feedback following a denial of responsibility.

Given that personality variables have repeatedly been linked to false confession behaviour both theoretically and in case studies (Gudjonsson, 2003), it is interesting that the effect has been difficult to demonstrate experimentally in the paradigm. There are several possible reasons for the general finding of a lack of relationship between personality variables and false confession behaviour. First, it may be personality factors other than those that have been examined impact behaviour more strongly in the paradigm. Second, it is plausible that the powerful situational demands of an interrogation setting transcend any personality influences (see Kendrick, Neuberg, & Cialdini, 2002). Finally, this paradigm may not be as useful as other methods, such as case studies, to investigate personality influences on false confession behaviour.

Consistent with Forrest et al. (2002), we found a gender difference in false confession behaviour, with females falsely confessing more often than males in the high plausibility (ALT key) condition. This finding may be attributed to gender differences in coping strategies used in stressful situations (Gudjonsson & Sigurdsson, 2003). However, the finding should be interpreted with caution, as the sample was predominantly female, and past research has been equivocal in directly relating gender to false confession behaviour.

It is likely that the relationship between gender and false confession behaviour is complex. For example, the interaction between the gender of interrogator and the gender of the suspect may be an important predictor of behaviour. Furthermore, gender and personality variables could interact to produce certain types of behaviour in interrogation settings. Accordingly, in this sample, an interesting finding was that personality factors, such as compliance and a tendency to acquiesce to leading questions, were related to false confession and internalization behaviour in males only. Thus, although females were generally more susceptible to falsely confessing in the paradigm, personality factors played a larger role in the behaviour of males.

Importantly, this is the first study to demonstrate false confession and internalization behaviour in an Asian sample using this paradigm, suggesting that the utility of the paradigm generalizes across ethnic groups. There were no differences in false confession or internalization rates between Asian and Caucasian participants. However, personality factors, such as compliance, were more related to false confession and internalization behaviour for Asians than for Caucasians. Thus, unique personality factors may be related to confession behaviour for different ethnic groups. The importance of this finding is highlighted by the recent work of Chang (2004) who identified culture-specific persuasive questioning used to extract confessions in Chinese criminal cases, including techniques designed to capitalize on power differences and invoke shame in defendants.

Limitations and future research

A major limitation of the present study is the restricted ecological validity of the experimental design. First, the current paradigm is limited by its ability to represent real-life legal situations. For example, there may have been few perceived consequences for falsely confessing to the momentary and unintentional allegation of pressing the forbidden key. Second, although an attempt was made to operationalize minimization and maximization interrogation techniques within the paradigm, the variety of statements used may have introduced imprecision in comparing their effects. Third, the current methodology was used to compare minimization to maximization interrogation techniques, but was unable to directly assess the independent effects of each technique because there was no control condition without interrogation statements. Fourth, there is a possibility that the administration of the GSS before the typing task may have impacted the tendency to falsely confess, as participants may have been either compelled to behave consistently in these persuasive tasks or primed by the suggestibility task to act similarly in a later task designed to induce false beliefs.

Finally, for ethical reasons, experimental designs aimed at investigating false confession behaviour cannot cause undue stress to participants (Costanzo & Leo, 2007). For example, the conceptualization and measurement of internalization in the current protocol is less than ideal, as this construct typically implies change across a longer term occurring in situations different from the one in which compliance is obtained (Kelman, 1958, 1961). A delayed assessment of internalization would be more favourable, although the potential negative psychological ramifications of delayed debriefing may preclude such a design.

Despite these limitations, it is important to continue the investigation of these phenomena using experimental designs. Future research should explore other individual and situational factors that may be related to false confession behaviour. In terms of personality factors, variables of interest include anxiety and coping style, which have been theoretically linked to aspects of interrogative suggestibility (Forrester, McMahon, & Greenwood, 2001; Gudjonsson, 1988, 1995; Howard & Hong, 2002). False confession behaviour could also be examined in other groups using this paradigm, including children and older adults.

Alternate interrogation techniques or tactics adapted from interrogation training manuals could also be introduced as experimental manipulations in the false confession paradigm. For example, a ‘good cop, bad cop’ condition could be created in order to examine the combined effects of minimization and maximization techniques, which are typically used sequentially in interrogations. Further, minimization techniques could be deconstructed in order to examine which components are most influential in eliciting false confessions. The experimental study of confession behaviour would also benefit from further examination of factors impacting true confession behaviour, using newly developed paradigms (see Russano et al., 2005). A continued investigation of the factors that contribute to false confessions and confession behaviour in general will greatly inform our understanding of the phenomenon and aid in efforts to prevent the occurrence of false confessions and their liberty-depriving consequences.

Acknowledgements

This study is based on a paper presented at the 2003 meeting of the American Psychology-Law Society in Edinburgh, Scotland. This research was supported in part by an American Psychology-Law Society Grant-in-Aid to the third author and a Social Sciences and Humanities Research Council of Canada doctoral fellowship to the second author. We would like to thank Amanda Murphy and Sandi Holman for their assistance with data collection.

Ancillary