Juror perceptions of false confessions versus witness recantations

Abstract

In false confessions, someone confesses to a crime but then later retracts that confession, whereas in witness recantations, an eyewitness testifies but then later revokes that testimony. The revocations are conceptually similar, but they differ in the author of the revocation – the defendant versus a third party. The current study examines differences in juror perceptions of the legitimacy of false confessions versus witness recantations, and also takes contextual influences (coercion and crime severity) into account. False confessions were found to be judged more harshly than witness recantations, but, surprisingly, levels of coercion and the severity of the crime did not influence decision-making in the manner that the original hypotheses predicted. Implications for the application of this research are discussed.

Approximately 30% of DNA exoneration cases have included false confessions (The Innocence Project, 2015), meaning that conclusively innocent people do sometimes confess to crimes and then later try to retract those false accounts. Similarly, witness recantations can occur in wrongful conviction cases, with eyewitnesses changing or revoking their statements after the original trial. In data from the National Registry of Exonerations, approximately 23% of all identified exonerations involved a witness recantation (Gross & Gross, 2013). One main difference exists between these two types of changing accounts: in false confessions, the defendant revokes his or her statement, and in witness recantations a third party revokes his or her statement. Research has not yet examined the influence of the person who is revoking prior testimony on jurors’ judgments.

False confessions

Statements of confession have a major influence on jury decision-making (Kassin, 2008b). It makes intuitive sense that a person who admits guilt would indeed be guilty, and in fact confessions have more impact on case outcomes than other persuasive forms of evidence (Kassin, 2008a). Confessions are not in the self-interest of the defendant, so a confession is highly persuasive of guilt (Kassin, 2008a, 2012).

There are several reasons, both situational and dispositional, why an individual might falsely confess to a crime (Kassin, 2008b), but confessions are nevertheless hard to ignore (Henkel, Coffman, & Dailey, 2008) and can lead to assumptions of guilt even when there is evidence to the contrary (Leo & Liu, 2009). Demonstrating this in a mock-jury study, Kassin (2008b) presented participants with one of three versions of a murder trial – a low-pressure confession condition, a high-pressure confession condition, and a no-confession control condition – and found that jurors may not always be able to discount a confession even after they learn that it was coerced. The confession in the high-pressure condition was indicated as having been involuntary, and the participants claimed that it did not influence their decisions; nevertheless, the conviction rate is higher in the high-pressure condition than in the no-confession condition. Thus, the coerced confession did influence their decisions, even though the participants were specifically instructed to disregard it and subsequently claimed that they had followed these instructions. Evidently, it is not always possible for jurors to disregard a false confession, even when they believe that they have done so.

Why are confessions so powerful? The perceptions of others routinely involve a fundamental attribution error in which people disproportionately emphasize dispositional factors instead of contextual, external factors or situations (Jones, 1990). For instance, many people do not fully disregard false confessions because they believe that they personally would never confess to doing something they did not do (Henkel et al., 2008); believing (perhaps reasonably) that they would never confess to a crime they did not commit, they evaluate others using the same assumption (Kassin, 2012), focusing on others’ dispositions while ignoring the circumstances. Further, even when the coercive nature of interrogation tactics is recognized, this knowledge may not influence people's decision-making (Blandón-Gitlin, Sperry, & Leo, 2011). In a survey study, participants recognized that interrogation methods might be both coercive and able to elicit true confessions, but they denied that these techniques could elicit false confessions (Leo & Liu, 2009). Participants in a mock-jury study agreed that a confession elicited due to police threats and false promises is involuntary, but they rendered a guilty verdict on the defendant anyway (Kassin, 1997). Even when people are aware of the circumstances, they may find it difficult to reject or forget an initial admission of guilt. Indeed, there may be fundamental differences between personal admissions of guilt and other testimony that establishes guilt which is provided by a third party.

Witness recantations

Another type of revocation occurs in some trials: through an affidavit in the appeals process, a witness may retract testimony offered under oath in the original trial. A witness's correction of inaccurate testimony is admissible evidence if the original statement was not a deliberate falsification (Salzman, 1977); otherwise, it will likely be judged to be perjury. Witness recantations may be very influential in new trials, but literature on the topic is minimal.

Courts are often skeptical of recantation claims, and they consider a motion for a new trial only if standards are met under Larrison v. United States (1928), in which the court decided the following:

A new trial should be granted [based on a recanting witness] when a) the court is reasonably well satisfied that the testimony given by a material witness is false, b) that without it the jury might have reached a different conclusion, c) that the party seeking the new trial was taken by surprise when the false testimony was given and was unable to meet it or did not know of its falsity until after the trial. (quoted in Armbrust, 2008, p. 81)

Considerable skepticism is expressed toward recanted testimony, and courts often reject recantations as grounds for winning exonerations (Heder & Goldsmith, 2012). This reluctance to give credence to recantations stems from two sources. Firstly, untrustworthiness may seem to be in play because the witness has proven to be unreliable; either the original story was false or the basis of the new statement is false. Secondly, the principle of finality is influential: the law likes to close cases (Heder & Goldsmith, 2012). If there is doubt in the original trial then there can be doubt in the subsequent trials, and a line has to be drawn somewhere. Judges rarely admit recantation testimony; it is often viewed with suspicion and is the basis for new trials only in extraordinary cases (Repka, 1986). However, little is known about what happens when new testimony is admitted and a witness's revocation of a statement is considered by jury members. Several contextual factors may influence their judgments, which are examined in the current study.

Contextual factors

Crime severity

The severity of the crime that has been committed may influence jury decision-making. It is no surprise that as crimes increase in seriousness, jurors tend to assign more severe penalties, which is commonly referred to as just desserts (Carlsmith, Darley, & Robinson, 2002). However, this assumes a level of certainty of the defendant's guilt. When a recanted confession exists, there is less certainty, which may change how crime severity is processed during jury decision-making. For example, a severity–leniency hypothesis states that as the negative consequences of conviction increase, jurors become less willing to risk convicting someone who is innocent – so they require more evidence of guilt to convict on more serious charges (Devine, 2012; Freedman, Krismer, MacDonald, & Cunningham, 1994). In the case of statement revocations according to the severity–leniency hypothesis, if the stakes are high and a defendant revokes a confession, a juror might be less likely to convict if there is a lack of other evidence of guilt. On the other hand, false confessions might be a unique situation that interferes with the severity–leniency hypothesis. From this perspective, a juror may also be less likely to believe the defendant's revocation because the consequences are severe enough that an initial false confession would seem unreasonable. People trust confessions the way in which they trust other behaviors that counter self-interest (Kassin, 2008a), and if the stakes are high, they might believe confessions to an even greater extent due to this contradiction of self-interest.

Coercion

Coercion from external forces should also reasonably influence a juror's perception of statement revocations. In a survey of attitudes and beliefs about false confessions, it was found that coercion is thought to be commonplace in interrogations and is also judged to be influential (Henkel et al., 2008). Physical torture, denying a suspect food and/or water, and lying to a suspect about physical evidence are thought to be impermissible tactics. Therefore, jurors might find the revocation of a confession to be more plausible when the confession was elicited in high-coercion situations. A similar pattern would also seem likely in a witness recantation scenario; the more pressure and coercion witnesses faced when they provided their original testimony, the more likely a juror might be to consider a later revocation to be credible. Nevertheless, a survey of potential jurors found that the participants believed false confessions to be counter-intuitive and unlikely, even in response to coercive interrogation techniques (Leo & Liu, 2009). Along the same lines, another classic study on coercion showed that even though mock-jury participants recognized coercive tactics, they still convicted at a high rate (Kassin, 2008b). Coercion may or may not be a factor that shapes perceptions of revoked testimony.

The current study

The current study seeks to identify differences in juror perceptions of the legitimacy of false confession and witness recantation claims. Because false confessions and witness recantations share structural similarities – in both cases, key actors change their prior accounts of events pertinent to a trial – any difference in jury perception would seem to stem from the fact that a defendant versus a third party made the revoked statement, along with the motivations and consequences implied for both. The effect of crime severity on juror perception is also factored into the study, due to the likelihood that the revocation statement of someone faced with graver consequences would be judged more harshly than someone with less to lose. The intensity of the interrogations is also manipulated in order to test whether or not jurors take coercion into account when judging the believability of false confessions or witness recantations. It was expected that: 1) the participants would judge false confessions to be less legitimate than witness recantations, specifically that they would find false confessors guiltier and thus less believable, 2) high levels of coercion would lead the participants to make more favorable decisions on behalf of defendants, specifically that it would reduce guilt ratings, and 3) the participants would view false confessors as guiltier for a high-severity crimes than low-severity crimes.

Method

Participants

The participants (n = 237) were recruited from Amazon's Mechanical Turk (mTurk) website in order to pursue a community-based sample. The survey took approximately 30 to 45 minutes to complete and involved reading five vignettes, answering questions, and responding to a demographics survey. The participants received $0.25 for completing the online survey. While this is a lengthy survey with minimal compensation, the length of the survey was advertised on the website and the compensation level offered is typical for the rates on mTurk, and was approved by the institutional review board overseeing the study. A total of 23 participants were excluded because they incorrectly answered at least two of the four total experimental manipulation checks. A threshold of at least two incorrect responses to the four manipulation checks was chosen as the exclusion criterion, as this type of responding would be similar to chance and not indicative of careful responding. Regardless, this rather arbitrary cut-off was not found to impact the results of the study. Of those whose data were retained for further inspection, 45% were female and 56% were male, with ages ranging from 20 to 72 years (M = 35, SD = 11). A total of 52% of the participants are White and non-Hispanic, 14% are Hispanic, 4% are Black and non-Hispanic, and 30% identified as ‘other’, with 1 participant declining to answer the question. Politically, 29% were liberal, 37% were slightly liberal, 21% were slightly conservative, and 13% were conservative, with 1 participant declining to answer the question.

Materials

The vignettes manipulated the independent variables of statement revocation type, crime severity, and coercion. In a between-subjects manipulation, participants considered scenarios in which, during an appeal of a prior conviction, a defendant sought to retract a confession or an eyewitness sought to change his testimony. In addition, in within-subjects manipulations, each participant responded to four different vignettes that varied the nature of the crime and the nature of the contested evidence. The past offense was one of either high (first-degree murder) or low (property theft) crime severity, and, cross-cutting this manipulation, the key evidence – a confession or a damning eyewitness account – had been obtained using a higher or lower degree of coercion: the interrogation had lasted 10 hours and involved isolation, deprivation, and threats about consequences from the police, or it had lasted only 2 hours and had not involved any pressure from the police. Each vignette offered details about (a) the crime, (b) the defendant (or eyewitness), (c) the confession (or eyewitness testimony), (d) the revocation statement (claiming the confession or eyewitness statement was false), and (e) the reasoning behind the revocation (see Appendix A for the false confession group and Appendix B for the witness recantation group). It then described the revocation statement and reasons behind the revocation during an appeal. These vignettes were created by the authors and are not based on real cases.

A fifth vignette provided an offset control group with a scenario wherein a defendant confessed to stealing property under conditions of lower coercion and did not try to retract his confession (and, for the witness recantation group, the control vignette presented an eyewitness who did not retract his or her statement). This provided an opportunity to determine how any retracted statement differed from a vignette in which no retraction was made at all. Thus, there were five total vignettes for each between-subjects group. Questions followed each vignette. Participants were asked to decide on a verdict for each defendant (guilty beyond a reasonable doubt or not guilty) and offer the following three ratings: (a) the extent to which they found the defendant guilty (where 1 = not at all guilty and 10 = completely guilty), (b) their confidence in the accuracy with which they offered their verdict (where 1 = not at all confident and 10 = completely confident), and (c) the believability of the defendant (where 1 = not at all believable and 10 = completely believable). They were also asked to decide how many years of imprisonment the defendant should face, within a range of 0 to 100 years. Although ratings of defendant guilt and believability are most directly related to the hypotheses, confidence in accuracy of the verdict and sentencing are included as dependent variables in order to obtain more nuanced detail in terms of the participants’ responses and provide other ways to indirectly measure the concept of guilt. A manipulation check – a multiple-choice question about a specific detail from the case (e.g. the duration of the interrogation, the name of the defendant, the name of the eyewitness) – was also used to test participants’ attention and understanding for each vignette.^1.

At the start of the procedure, the participants provided demographic information on their gender, age, United States (US) citizenship, ethnicity, political views, and the extent of their previous experience on a jury.

Procedure

Participants were recruited for the study through an advertisement on Amazon's mTurk (i.e. ‘Read crime scenarios and answer questions about them’). They consented online to participating in the study and then completed the study's materials in a survey hosted by Surveymonkey.com. Materials for the two statement-revocation conditions were posted on mTurk at different times, and measures were taken to ensure that no one who filled out a survey in the false-confession condition could also take the witness-recantation survey. A feature from Surveymonkey.com was also used to randomize the presentation of the five vignettes to each participant in order to control for order effects.

Results

A mixed-factor analysis was used to account for the examination of both within-subjects and between-subjects factors. The data were examined with a 2 (statement revocation: false confession vs witness recantation) × 2 (degree of coercion: lower vs higher) × 2 (crime severity: high vs low) multivariate analysis of variance (MANOVA) that assessed the effects of the manipulated variables. The statistically significant multivariate effects are presented in Table 1.

Table 1.

MANOVA effects.

		F (Wilks’ Λ)	p	η2p
Between-subjects	Revocation type	8.29	.00	.16
Within-subjects	Coercion	1.82	.13	.02
	Severity	5.55	.00	.09
	Coercion × Severity	5.12	.00	.09
	Coercion × Revocation type	0.65	.63	.01
	Severity × Revocation type	1.65	.17	.03
	Coercion × Severity × Revocation type	0.84	.50	.02

Note. Bold font indicates statistical significance. All degrees of freedom are (4, 257).

Statement revocation

Addressing the first hypothesis (i.e. participants would judge false confessions to be less legitimate than witness recantations), the MANOVA revealed significant main effects of the type of revocation statement (false confession or witness recantation) on participants’ perceptions of guilt. The defendant was more likely to receive a guilty verdict when he retracted his confession (60%) than when the eyewitness recanted his testimony (51%), which is analyzed using the dichotomous ‘guilty’ and ‘not guilty’ verdicts. The defendant was also rated as guiltier in the false-confession condition (M = 6.8, SD = 3.2) than in the witness-recantation condition (M = 5.2, SD = 3.2), F(1, 210) = 24.68, p < .001, η²_p = .09.

The participants also expressed more confidence in the accuracy of their decision-making in the false-confession condition (M = 6.8) than in the witness-recantation condition (M = 6.2), F(1, 265) = 70.08, p = .01, η²_p = .025. Thus, when the defendant withdrew his confession, participants felt more confident and found him guiltier than they did when the eyewitness recanted his statement. The type of revocation did not, however, have a significant effect on the believability of the defendant or the sentencing.

Although a comparison of the experimental vignettes to the control vignettes was not included in the original hypotheses, it was deemed to be important to demonstrate the influential nature of a retracted statement. The control vignettes determined how participants’ perceptions differed when the defendant (or eyewitness) did not retract his confession (or testimony). In the false-confession condition, the guilt ratings are significantly higher (M = 7.1, SD = 2.8, p < .05 by Dunnett's test) in the control vignette, demonstrating that the defendant's retraction of his confession in the other vignettes did have an impact on the participants’ perception of his guilt. The guilt ratings are also higher in the control vignette in the witness-recantation condition (M = 5.9, SD = 2.9, p < .05 by Dunnett's test) than they are when the eyewitness recanted his testimony, so the recantation is also shown to have benefited the defendant. Nevertheless, and importantly, when the defendant withdrew the confession that had convicted him, the participants found him guiltier than they did when the eyewitness recanted testimony that had been equally damning.

Contextual factors

Coercion and severity

In regard to the second hypothesis (i.e. high levels of coercion would lead participants to make more favorable decisions on behalf of defendants), the level of police coercion does not have a reliable main effect on any dependent variable in the MANOVA, F(1, 207) = 2.18, p < .07. It was expected that coercion would to lead to leniency in judgments of the defendants. It is difficult to ascertain if this null effect resulted from the true performance of this variable or from a weak manipulation.^2.

In contrast, crime severity was found to be more influential, addressing the third hypothesis (i.e. participants would view false confessors as guiltier for high-severity crimes than low-severity crimes). The MANOVA revealed a main effect of crime severity on the participants’ sentencing decisions, F(1, 210) = 13.66, p < .001, η²_p = .08; defendants charged with crimes of higher severity received longer sentences (M = 18.9, SD = 30.0) than those charged with crimes of lower severity (M = 14.0, SD = 26.9). Additionally, there is a significant main effect of crime severity on accuracy ratings such that the participants felt more confident in their judgments if the severity of the crime was lower (M = 7.0) rather than higher (M = 6.0), F(1, 260) = 4.30, p = .04, η²_p = .02. Defendant believability, however, was not found to be statistically significant.

A statistically significant interaction of coercion and crime severity was obtained in the MANOVA, F(1, 210) = 72.14, p < .001, η²_p = .12; the means are presented in Table 2. When the crime severity was high, the participants found defendants to be less guilty when the degree of coercion was also high than when it was low, F(1, 234) = 26.85, p < .001, d = 0.28, suggesting some sensitivity to the influence of coercion in high-stakes situations. On the other hand, when the crime severity was low, defendants were judged to be guiltier when the degree of coercion was high, F(1, 231) = 9.36, p = .002, d = 0.14. This finding is surprising and may suggest that the participants felt that if the degree of coercion is high in a low-stakes situation then the police must have had good reasons for using it. When coercion was not used to extract a confession, defendants were perceived as guiltier for crimes of higher, rather than lower, severity, F(1, 234) = 24.15, p < .001, d = 0.28. But once again, when the degree of coercion was high in response to lesser crimes, defendants were judged to be guiltier than the case of more severe crimes, F(1, 231) = 7.63, p = .006, d −0.14.

Table 2.

Interactions of coercion and severity on ratings of guilt, M (SD).

		Severity
		Low	High
Coercion	Low	5.70ac (3.00)	6.50ad (2.81)
	High	6.10bc (2.79)	5.70bd (2.96)

Note. Means with the same single-letter subscript differ by at least p < .05.

Discussion

The goal of this study is to assess the differences in juror perceptions of court cases in which a defendant falsely confesses to a crime versus a third-party witness who recants a statement. These seem to be conceptually similar scenarios, with someone revoking an important piece of evidence, but it was hypothesized that jurors would judge differently based on the identity of the individual revoking the statement.

The first hypothesis that the participants would judge false confessors more harshly than recanting witnesses is confirmed. The participants rated false confessors as guiltier than the defendants in the cases where the witness recanted, and the dichotomous verdicts also show higher percentages of guilty verdicts in the false-confession condition versus the witness-recantation condition. The participants also felt more confident in rating the guilt of false confessors than when the eyewitness statement was revoked, suggesting a certain confidence in these harsher judgments. Additionally, there is a bigger difference between the ratings of guilt for the control group (the scenario where no statement is retracted) for the false-confession condition than for the witness-recantation condition. This supports the notion that there is something fundamentally different about false confessions; they are judged more harshly than other forms of revoked statement. The fundamental attribution error is one likely explanation as to why there is a difference in participant perceptions between these two groups; it is possible that internal characteristics were overestimated as the explanation for the criminal behavior, rather than contextual information, which is supported due to the fact that contextual factors did not significantly influence participant decision-making in these cases. The disbelief that a person would confess to something that he or she did not do, and the mock juror’s own disbelief that he or she would personally ever give a false confession, are likely reasons for this discrepancy in perception. As expected, the confession held great weight in judging guiltiness, despite the other manipulated contextual circumstances.

The second hypothesis was that the presence of police coercion would cause the participants to view a defendant more favorably because of an acknowledgement of the contextual factor's influence. Surprisingly, this is not the case, and the hypothesis is not supported. It seems reasonable to imagine that differences in the degree of police pressure placed on a defendant to confess to a crime or on an eyewitness to make a statement would affect how that confession or statement is perceived by a juror, but the manipulation of the degree of coercion in this study did not have a profound effect on the participants’ judgments. Unfortunately, due to the nature of the manipulation checks used during this study, it is difficult to ascertain whether the minimal effect of the degree of coercion was due to the true performance of the variable or to a weak manipulation. Nonetheless, these results support the conclusion that even though people may recognize coercion, they often find defendants guilty anyway (Kassin, 1997).

The manipulation of crime severity significantly influenced decision-making such that defendants received longer sentences for high-severity crimes such as murder. This supports the third hypothesis – that jurors would view false confessors as guiltier in the case of high-severity crimes compared to low-severity crimes. Also, the degree of coercion did interact with crime severity. When the crime severity was high, the participants rated defendants as less guilty when the degree of coercion was also high, suggesting that the participants were more sensitive to the influence of coercion in high-stakes situations. On the other hand, when the crime severity was low, defendants were found to be guiltier when the degree of coercion was high, which is a surprising finding; participants may have presumed that if a high degree of coercion was used for a low-stakes crime then the police officers must have had good reasons for using coercion. This interaction of crime severity and coercion supports the severity–leniency hypothesis (Devine, 2012; Freedman et al., 1994). In general, the defendants were perceived to be guiltier for the high-severity crimes; however, when a high degree of coercion was also used, the defendants were perceived to be less guilty than when a low degree of coercion was used. This may be because the introduction of coercion led to doubts in the participants’ minds, thus meaning that they required more evidence of guilt for serious charges. However, this study also shows that the opposite may also be true: when a high degree of coercion was used for a low-severity crime, the defendants were perceived to be more guilty than when a low degree of coercion was used. Perhaps it was the case that the participants were sensitive to the detrimental impact of coercion when their responsibility in making judgments was greater (i.e. for high-severity crimes), but less so when their judgments would have less of an impactful outcome (e.g. for low-severity crimes).

It is important to note that this study has certain limitations. First, several participants were excluded due to manipulation-check errors. The literature has generally found that mTurk users are less likely to pay attention to experimental materials than traditional samples (Goodman, Cryder, & Cheema, 2012). However, mTurk is considered to provide a reliable and valid group of participants in terms of responses that represent the wider community compared to students on a university campus. Although the participants are paid a small amount of compensation for their time, this does not appear to affect the quality of the data that are obtained (Buhrmester, Kwang, & Gosling, 2011). These users tend to produce reliable results that are consistent with standard decision-making biases, making them a worthwhile source for data collection (Goodman et al., 2012). A second limitation is the artificiality of the study's materials, which are not necessarily representative of actual courtroom procedures. Jurors do not actually rate guilt levels or have the opportunity to rate their confidence in the accuracy of their decision-making; they simply provide a verdict, and sometimes recommend sentences. Perhaps more importantly, reading a brief scenario on paper is not the same as experiencing a trial first-hand. The materials are brief and not as impactful as real trial procedures would usually be. However, this study's methodology provides an important first look at attitudes toward false confessions versus witness recantations. Future research should attempt to acquire information about juror judgments in real courtroom settings with real jurors in real cases in order to provide additional ecological validity.

No previous studies have compared juror perceptions of false confessions and witness recantations. Although the two concepts are similar, they have not been compared in this way. Thus, this is an important study for several reasons. First, it shows that false confessions are viewed as a fundamentally different and unique type of revoked statement. Even when people recognize that false confessions exist, they are judged more harshly than other types of retracted testimony. Second, few studies have ever examined witness recantations from a psychological perspective. Witness recantations are involved in 72% of the first 325 DNA exoneration cases, making false eyewitness testimony the most common cause of wrongful convictions in these cases (Innocence Project, 2015). As work that uncovers miscarriages of justice becomes more common and more exonerations occur, witness recantations become an important problem to be solved. It will be valuable to know how jury members perceive statement revocations from eyewitnesses because such knowledge may be important evidence for many future exonerations.

This study takes a first step in examining jurors’ perceptions of different causes of wrongful convictions, but there is much more work to be done. Future research should clarify the nature of harsh opinions about false confessions. This study assumes an explanation for these opinions via the fundamental attribution error, but qualitative opinions regarding the reasoning behind making these decisions could be beneficial. Research should also begin to focus on the psychological processes of recanting witnesses themselves. Witness recantations are common in many cases of wrongful conviction, but the origins of such revoked testimony remain uncertain. Finally, it will be useful to examine how general versus particular beliefs and attitudes related to a specific case can affect juror decision-making. These future directions are becoming increasingly important as the prevalence of wrongful convictions becomes more and more acknowledged and the innocence movement gains momentum and builds public awareness of its existence.

Appendix

Appendix A

Example of a high-severity, high-coercion, false-confession vignette: State v. Juarez

This case involves Felix Juarez, a 30-year-old Hispanic male, who was convicted of first-degree murder. During the trial, the prosecution contended that it was Felix Juarez who, at 2:20am on March 12, 2011, entered a liquor store on Maple Street in Joliet, Illinois, and, in the process of attempting to rob the liquor store, shot and killed 65-year-old Lewis Beem, who was a patron of that store. The robber instructed the clerk to put money in a bag, and when he saw Mr. Beem at the back of the store he shot him in the head. Mr Beem was a frequent customer at the liquor store. He was pronounced dead upon arrival at the Joliet Medical Center.

An eyewitness two blocks away saw someone running from the direction of the store and called the police. A few minutes later, Mr. Juarez was picked up by the police on East Maple Street, 11 blocks from where the liquor store was located. Mr. Juarez was walking down the street, and he fit the general description that came across the radio dispatch – a Hispanic male, in his early thirties, approximately 5'9'' tall, and weighing approximately 160 pounds. Mr. Juarez was initially surprised about being picked up for the crime, but after ten hours of being questioned at the police station, he confessed to the crime. At his trial, the jury found him guilty. His confession was the only evidence that clearly pointed to him as the robber, and without the confession he would not have been convicted.

Six months later, Mr. Juarez is appealing his conviction, asserting that the issue of his confession had not been properly handled. He claims his original confession was not true. He maintains that he is innocent. He claims that he confessed because he wanted to return home after a brutal ten-hour interrogation. Mr. Juarez said the police coerced him into confessing, yelling at him and intimidating him. He says the police threatened to get his family into trouble if he did not confess. He was also left in isolation for hours on end and was deprived of food, water, and sleep.

Appendix B

Example of a high-severity, high-coercion, witness-recantation vignette: State v. Juarez

This case involves Felix Juarez, a 30-year-old Hispanic male, who was convicted of first-degree murder. During the trial, the prosecution contended that it was Felix Juarez who, at 2:20am on March 12, 2011, entered a liquor store on Maple Street in Joliet, Illinois, and, in the process of attempting to rob the liquor store, shot and killed 65-year-old Lewis Beem, who was a patron of that store. The robber instructed the clerk to put money in a bag, and when he saw Mr. Beem in the back of the store he shot him in the head. Mr. Beem was a frequent customer at the liquor store. He was pronounced dead upon arrival at the Joliet Medical Center.

An eyewitness, Mr. Hill, standing outside the laundromat next to the liquor store, saw the robber for several long seconds as he exited the liquor store. A few minutes later, Mr. Juarez was picked up by the police on Maple Street, 12 blocks from where the liquor store was located. Mr. Juarez was walking down the street, and he fit the general description given by Mr. Hill – a Hispanic male, in his early thirties, approximately 5'6'' tall, and weighing approximately 150 pounds. After ten hours of being questioned at the police station, Mr. Hill named Mr. Juarez as the robber. Mr. Juarez was surprised about being picked up for the crime. At his trial, the jury found him guilty. The eyewitness testimony of Mr. Hill was the only evidence that clearly pointed to Mr. Juarez as the robber, and without this testimony he would not have been convicted.

Six months later, Mr. Hill is revoking his original eyewitness testimony, asserting that he identified Mr. Juarez because of intense pressure from the police to do so. Mr. Hill asserts that the police wanted him to identify Mr. Juarez as the robber, and threatened to get his family into trouble if he did not comply. Mr. Hill was left in isolation for several hours and was deprived of food, water, and sleep until he identified Mr. Juarez as the robber. He claims his original statement was not true. He maintains that Mr. Juarez is innocent.

Acknowledgements

This paper was previously presented at the Annual Meeting of the American Psychology-Law Society, Atlanta, GA, 10–12 March 2016.

Declaration of conflicts of interest

Paula Bernhard, M.A. has declared no conflicts of interest

Rowland Miller, Ph.D. has declared no conflicts of interest

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study