How downplaying or exaggerating crime severity in a confession affects perceived guilt

Abstract

This study investigates how judgments of guilt are influenced by factual errors in confessions that either amplify or downplay the severity of the crime. Participants read a confession statement and police report in which either the confession was consistent with the police report, the suspect admitted to a worse crime or the suspect admitted to a lesser crime. Mediation analyses showed that, compared to consistent confessions, both types of directional errors reduced judgments of guilt. Inconsistencies that made the suspect look better – but not those that made the suspect look worse –also increased judgments of guilt via a direct effect. Confessions that contain errors that appear to exaggerate the severity of the crime prompt no higher judgments of suspect guilt; however, errors in confessions that are perceived to downplay the severity of the crime can prompt an increased perception of suspect guilt compared to a consistent confession.

False confessions have resulted in a surprising number of wrongful convictions, with the percentage of false confessions implicated in DNA exoneration cases estimated at anywhere between 16% (Garrett, 2011) and 25% (Innocence Project, n.d.). These numbers stand in contrast to the commonly held belief that innocent people will not confess unless mentally ill or tortured (Leo & Ofshe, 1998). Contributing to wrongful conviction rates is the difficulty that police investigators and jurors have in discovering and discounting false confessions, as there is a general inclination to simply trust that a confession is true. Drizin and Leo (2004) found that 30 out of 37 proven false confessors (81%) who decided to go to trial were found guilty by the jury, even after pleading not guilty. This highlights not only the overall acceptance of confessions, but the unparalleled power that confessions hold in courts (Kassin & Neumann, 1997), even when retracted (Kassin & Sukel, 1997; Leo & Ofshe, 1997).

Jurors are not always best equipped with the knowledge or skills needed to effectively apply court instructions, evidence and legal arguments to their decisions, and might instead use heuristics that are not reliable in the unique circumstances of a court case (Bornstein & Greene, 2011). One reason why people find confessions so convincing is the strength of the heuristic which assumes that as there is no advantage to confessing, the confession must be true. However, there is conflicting evidence in the literature about whether or not this heuristic is robust in the face of inconsistent confession evidence. Although inconsistencies in eyewitness testimony have been shown to reduce the credibility of the testimony and the likelihood of conviction (e.g. Berman et al., 1995), inconsistencies in confessions are often perceived as having minimal impact on jurors’ verdicts (Malloy & Lamb, 2010). However, some laboratory studies have found that inconsistencies can lead participants to reject a confession, resulting in a lower conviction rate (Henderson & Levett, 2016; Holt & Palmer, 2020; Jones et al., 2019; Palmer et al., 2016; Woestehoff & Meissner, 2016). Such conflicting findings in the context of confessions may be due to different presentations of inconsistencies being interpreted differently by jurors, such as one suspect giving multiple contradictory statements versus a suspect giving a single confession that contradicts the facts of the crime (e.g. Holt & Palmer, 2020; Jones et al., 2019). There is also the possibility that jurors might process confession inconsistencies differently according to the direction of the inconsistency – that is, whether the inconsistency appears to increase or decrease the severity of the crime.

Directional errors

The direction of confession inconsistencies can be important if we think about how false confessions come about. Innocent suspects sometimes confess to a crime without adequate consideration of the consequences. This occurs because the suspect believes that confessing is the only way to stop a stressful interrogation and begin the process of clearing their name (Kassin, 2005, 2012; Leo & Drizin, 2010). In order to produce a confession believable enough to satisfy the interrogator, the innocent suspect creates a story using the information they have at hand. This might include details that they have learned during the interrogation process, seen in the media or deduced using logic or common sense. However, the innocent suspect is at a disadvantage when creating an accurate confession as they do not have first-hand knowledge of the crime. In their ignorance of the specifics of the crime, the suspect might inadvertently understate or overstate key details, such as admitting to shooting the victim too few or too many times compared to the actual wounds inflicted. Although research has been conducted on the effects of confession inconsistencies on perception of suspect guilt, the effect of directional errors that might reasonably be expected of an innocent suspect have not to our knowledge been investigated.

In particular, understanding the effect of understating the crime (inadvertently or otherwise) is important, as some police interrogation techniques may increase the possibility of a suspect confessing to an understated version of a crime. The commonly used Reid technique (Inbau et al., 2013) advocates the use of minimization in the police interrogation process, wherein the interrogator attempts to elicit a confession by downplaying the severity or consequences of the crime and providing the suspect with face-saving excuses for their involvement. By lessening the suspect’s perception of the severity of the crime, minimization could inadvertently encourage false narratives with inconsistencies that indicate a desire for minimal culpability. The use of such minimization techniques has been shown to increase the likelihood of both true and false confessions alike (Russano et al., 2005), providing an important reason for investigating the effect of directional errors on how jurors perceive the suspect’s level of guilt when processing confession evidence.

The aim of the present study is to investigate how factual errors in confessions that either amplify or downplay the severity of the crime influence judgements of guilt.

Attribution theory

Kelley’s (1973) attribution theory holds that the inferences an observer draws from another’s behavior depends on attributions about the motive underlying that behavior. An important implication of this principle is that a particular behavior can lead to very different inferences depending on the attributions made by the observer. An example of this principle, provided by Reeder et al. (2004), involves an observer who sees a student helping a professor. Without additional context, the observer may attribute this behavior to the student’s helpful nature. However, if the observer is aware of an ulterior motive (e.g. the professor is evaluating the student’s scholarship application), the helping behavior may be attributed to the pursuit of this ulterior motive rather than the student having a helpful nature. In this case, the perceived presence of an ulterior motive leads to discounting; that is, the helping behavior is discounted when making inferences about the disposition of the student. In contrast, some contextual information can lead to augmentation, whereby an observed behavior has a stronger effect on dispositional attributions. For example, if a student helps a professor without the professor’s knowledge, this might enhance an observer’s attributions about the student having a helpful nature; if the professor is unaware of being helped by the student, this reduces the scope for any ulterior motive to explain the behavior. Thus, the same behavior (a student helping a professor) can lead to very different inferences about the underlying cause of the behavior, depending on the attributions made by the observer.

Although the principles of discounting and augmentation were developed in the context of making dispositional attributions, we can consider how they might help to explain how jurors interpret directional inconsistencies in a confession. We will assume that jurors are motivated to explain inconsistencies in confessions and will therefore seek explanations for them (Palmer et al., 2016). Furthermore, we can expect that the inferences drawn from inconsistencies in confession evidence will vary systematically depending on the direction of the inconsistency. For example, if the evidence shows that the suspect has confessed to a crime that is less severe than the facts reveal (e.g. admitting to firing one shot when there is evidence that three were fired), the juror might reason that the suspect is lying in an attempt to look less guilty and attract a lesser penalty. If so, then according to the discounting principle, inconsistencies that imply a less severe version of the crime might not translate into reduced perceptions of guilt (relative to a confession without inconsistencies). However, if the evidence shows that the suspect has confessed to a worse version of the crime (e.g. firing ten shots when only three were fired), the juror might reasonably wonder what the suspect has to gain by lying in such a manner. As there is no perceived benefit in confessing to a crime of greater magnitude, the juror might suspect that the confession is the result of an unknown motive unrelated to guilt (e.g. perhaps the defendant is innocent and was pressured to confess). In turn, inconsistencies that imply a more severe version of the crime will lead to reduced perceptions of guilt (relative to a confession without inconsistencies). In this way, attribution theory provides a framework for investigating how jurors not only rationalize the presence of errors in a confession but also how they investigate the types of motivations that might underpin the directionality of those errors.¹

Although no prior studies have directly compared inconsistencies that imply a more severe versus less severe crime, two published studies have included manipulations of inconsistencies with some directional properties. Palmer et al. (2016) manipulated inconsistencies by including three details in a confession statement that differed from the verified facts in the police case file. Two of these details were non-directional (e.g. the time when the crime had occurred) but the third showed that the suspect had confessed to a crime that was more severe than what had actually occurred: the confession stated that the victim had been shot five times, whereas the police file stated that the victim had been shot once. In the two experiments that included this manipulation, the participants were found to be less likely to convict when reading the confession that was inconsistent with the case file compared to the one that was consistent.

Henderson and Levett (2016) had participants read a confession containing errors that downplayed the severity of the crime. The suspect confessed to stabbing the victim 1 to 2 times, and the police case report stated that the victim was stabbed either 1 to 2 times (consistent condition) or 38 times (inconsistent condition). In two experiments using this manipulation, the manipulation of inconsistency had a marginal or null effect on the verdict. In a third experiment, the direction of the inconsistency was counterbalanced such that some participants read a confession that exaggerated the severity of the crime and others read a confession that downplayed the severity of the crime. However, comparing inconsistencies of different directions was not the focus of that study, and these two conditions were collapsed together when assessing the effects of the inconsistency manipulation. With the two directional conditions combined, the inconsistency was found to reduce the number of guilty verdicts.

Together, these results align with the rationale outlined above: when confessions contain inconsistencies that exaggerate the severity of the crime, guilty verdicts are reduced (as in Palmer et al., 2016), but when confessions contain inconsistencies that downplay the severity of the crime, the effects on verdicts are smaller and sometimes null (as in Henderson & Levett, 2016, Experiments 1 and 2). However, these studies did not conduct comparisons between inconsistencies of different directions, which is the focus of the present research.

Hypotheses

In line with Palmer et al. (2016), we hypothesized that confessions containing inconsistencies would be rated as less consistent than confessions that were consistent (i.e. we expected the jurors to notice the inconsistencies). We further hypothesized that juror perception of lower confession consistency would influence judgments of guilt dependent on how the jurors attributed the reasons for the errors in the confession – that is, if the confession errors increased the severity of the crime and acted to make the suspect look worse then we expected greater inconsistency to be associated with lower judgments of guilt. According to attribution theory, because there is no obvious ulterior motive for the worse inconsistencies, we expected some jurors to interpret them as an indicator that the suspect might not have committed the crime (‘Why would he make himself look even worse? Maybe he doesn’t know what happened because he didn’t commit the crime.’). However, if the inconsistencies decreased the crime severity and acted to make the suspect seem better, we predicted that while the presence of inconsistencies would reduce perceptions of the consistency of the confession, this reduction would not necessarily translate to lower judgments of guilt. In the case of a better confession, there is a clear ulterior motive for the inconsistencies wherein the suspect is self-servingly downplaying the severity of the crime (‘He’s making errors on purpose to make himself look better – of course he did it’).

Method

Participants

A total of 94 participants (67 female, 26 male, 1 undisclosed) aged 18 to 63 years (M = 26.49, SD = 9.21) volunteered from undergraduate psychology classes and from the greater University of Tasmania community. The majority were enrolled at university full time (99%). Participants were awarded with partial course credit or AU$10 remuneration. The majority of participants spoke English as their main language at home (83%). Two additional participants were excluded from the data set for misunderstanding the instructions, resulting in nonsensical answers such as believing that the police report had been written by the suspect. Their data were not analyzed. Participants who provided partial answers (e.g. a judgment of guilt but not a rating of perceived consistency) were retained in the data set, resulting in some analyses with a total n < 94.

Procedure

Supervised testing took place in a laboratory setting in small groups (n = 2–3) with participants giving informed consent. Participants acted as individual jurors and no deliberation or discussion took place. Random allocation was used to place participants in one of three confession conditions (worse, consistent, better) in a between-groups design. After reading two pieces of written evidence (a typed confession statement and a police report), participants answered the pen and paper questions in order, with instructions not to read ahead or change their answers once given. Completion time for the task was approximately 20 minutes.

Stimulus materials

Participants were given two pieces of evidence from a fictional case: a confession statement and a police report. The confession was presented as a typed statement in which the suspect admitted to committing armed robbery at a service station and discharging a firearm at the service station attendant. The typed police report gave detailed information based on the service station’s CCTV footage, including the number of shots fired and confirmation that the safe had been completely emptied in the robbery. The report confirmed that the safe’s contents had been counted and verified by two staff members prior to the robbery taking place. This was to ensure that participants did not dismiss any incorrect statement about the amount of money that was stolen as being due to an administrative error or unrelated theft by a staff member. The materials were made to appear as though they were copies of evidence from a real case.

Directional errors manipulation

To test the directional effect of inconsistencies, key information in the confession statement was altered to create three versions: one that matched the facts of the crime in the police report (consistent), one in which the suspect admitted to a greater crime than that outlined in the police report (making him look worse) and one in which the suspect admitted to a lesser crime than that outlined in the police report (making him look better). The confession errors related to key facts of the crime (the amount of money stolen and the number of shots fired) to ensure that they were salient enough to attract notice and warrant the participants questioning the truthfulness of the confession. The police report stated that three shots were fired and AU$2100 was stolen. However, in the better condition, the confession stated that one shot was fired and several hundred dollars were stolen, whereas in the worse condition, the confession stated that around 10 shots were fired and AU$10,000 was stolen.

Measures

After reading the confession statement and police report, participants were asked to give a dichotomous (not guilty, guilty) verdict and rate their confidence in their verdict on a scale where 1 = not confident at all and 10 = completely confident. Previous research has found that the majority of people see a confession as an honest admission of guilt and will render a guilty verdict accordingly (Drizin & Leo, 2004). Therefore, dichotomous verdicts may not prove a subtle enough measure of the possible variation in juror belief in suspect guilt. To reveal more nuanced perceptions of guilt, we utilized a common practice in juror research (see Appleby & Kassin, 2016; Leippe et al., 2004; Sauer et al., 2017; Sommers & Kassin, 2001; Tenney et al., 2007), creating a continuous variable of judgment of guilt by combining the dichotomous verdict and participants’ confidence in that verdict. The resulting scalar variable measures judgment of guilt on a scale from −9.5 (complete confidence in a not guilty verdict) to 9.5 (complete confidence in a guilty verdict). Participants who score close to zero on this scalar variable are those who have little strength of belief that their verdict is correct. Following the verdict and confidence questions, the participants were asked to rate the consistency of the facts of the crime between the two pieces of evidence on a scale where 1 = not consistent at all and 10 = completely consistent.

To prime the participants to think more closely about the consistency of the evidence, they were asked to list why they thought differences (if any) between the pieces of evidence might have occurred. Two raters coded a sample of verbatim participant responses into those which indicated a belief that the suspect’s errors were a ploy to make himself look better, reduce perception of the severity of the crime or lessen the consequences of the crime. These were coded as downplaying, with responses that did not meet this criterion coded as no downplaying. Full agreement between the two raters was achieved.

Results

Judgment of guilt

Table 1 shows the means, standard deviations and confidence intervals for judgments of guilt and perception of consistency ratings for each of the conditions. There is no significant overall difference between the judgment of guilt scores for the consistent, better and worse conditions, with the mean scores for all three groups indicating that participants were reasonably confident that the suspect was guilty.² The manipulation check of consistency indicates that participants were able to perceive that the two inconsistent conditions (better and worse) contained confession inconsistencies and that the consistent condition did not, F(2, 90) = 56.10, p < .001.

Table 1.

Mean judgment of guilt and perception of confession consistency scores.

	n	M (SD)	95% CI
Judgment of guilt
Consistent condition	31	6.95 (4.50)	[5.30, 8.60]
Worse condition	32	5.27 (5.77)	[3.19, 7.35]
Better condition	30	5.80 (5.47)	[3.76, 7.84]
Perception of consistency
Consistent condition	32	8.98 (0.97)	[8.63, 9.34]
Worse condition	31	4.77 (1.95)	[4.06, 5.49]
Better condition	30	5.27 (2.08)	[4.50, 6.04]

Notes. CI = confidence intervals. Condition ns differ across measures because a few participants provided a judgment of guilt or consistency but not both.

Mediation analyses were conducted to test whether or not the effects of the directional manipulations on judgments of guilt were mediated by perceived consistency (from the consistency manipulation check). Perceived consistency is considered an important mediator, as the actual presence of inconsistencies might be unrelated to whether participants consciously noticed them not. A similar concept was explored in Palmer et al. (2016), wherein the degree of perceived consistency (as measured by the number of inconsistencies reported) mediated the effect of their manipulation on juror verdicts. To test the directional inconsistencies in the present study, two separate mediation analyses were conducted: consisten versus better and consistent versus worse.

When broken down into its components, the results of the consistent versus better conditions (n = 61) suggest that there were two different effects occurring (see Figure 1). First, there was a significant indirect effect in which the presence of inconsistencies reduced perceived confession consistency: (a), b = −3.71, 95% CI [−4.55, −2.89], t = −8.98, p < .001. In turn, as perception of confession consistency reduced, judgments of guilt similarly reduced: (b), b = 1.34 [0.61, 2.07], t = 3.67, p < .001. The significant indirect effect of inconsistencies on judgments of guilt (ab) confirms that the confession errors overall reduced the judgments of guilt by decreasing the perceived consistency, b = −4.99 [−8.42, −1.72]. This finding replicates the indirect effect of inconsistencies found in Palmer et al. (2016) and suggests that the greater the perceived inconsistency, the larger the reduction in guilt ratings.

Figure 1.

Mediation model testing the effect of inconsistencies (consistent versus better) on judgments of guilt, mediated by perceived consistency of the confession.

Note. *p ≤ .05; **p ≤ .01; ***p ≤ .001.

Second, and separate to the indirect effect, the direct effect of inconsistencies (c') operated in the opposite direction, b = 3.84 [0.26, 7.42], t = 2.15, p = .036, with inconsistent testimony increasing judgments of guilt. Opposing direct and indirect effects might seem counterintuitive, but these are not uncommon and are indicative of a mediation model in which two opposing mediational processes result in a relationship between the predictor and outcome variables that is effectively zero (MacKinnon et al., 2007). In the case of our study, the results suggest that two opposing mechanisms influenced judgments of guilt. One of them involves the indirect effect described above, whereby the presence of inconsistencies reduced judgments of guilt via differences in perceived consistency of the confession evidence. The other – opposing – mechanism is linked to the direct effect, whereby the presence of inconsistencies that made the suspect look better were associated with higher judgments of guilt. We return to this issue after reporting the mediation for the consistent versus worse conditions below.

The results of the consistent versus worse conditions (n = 62) reflect similar findings to the consistent versus better conditions for the indirect effect but not for the direct effect (see Figure 2). For the consistent versus worse conditions, the presence of inconsistencies significantly reduced perceived consistency: (a), b = −2.10 [−2.50, −1.71], t = −10.56, p < .001. Again, lower perceived consistency in turn resulted in lower judgments of guilt: (b), b = 1.06 [0.22, 1.90], t = 2.53, p = .014. The significant indirect effect of inconsistencies on judgments of guilt (ab), b = −2.23 [−4.23, −0.33], also aligns with the indirect effect of inconsistencies on verdicts found by Palmer et al. (2016), whereby confession inconsistencies reduced judgments of guilt. However, unlike in the consistent versus better conditions, there is no significant direct effect of the manipulation on judgments of guilt in the consistent versus worse conditions, b = 1.38 [−0.79, 3.56], t = 1.27, p = .209. Therefore, there is no evidence that the confession inconsistencies which exaggerated the severity of the crime affected judgments of guilt separate to the mechanism of inconsistencies reducing perception of confession consistency.

Figure 2.

Mediation model testing the effect of inconsistencies (consistent versus worse) on judgments of guilt, mediated by perceived consistency of the confession.

Note. *p ≤ .05; **p ≤ .01; ***p ≤ .001.

Downplaying of guilt

The comparison of the consistent versus better conditions yielded evidence for two opposing mechanisms, one linked to the indirect effect (whereby inconsistencies reduced judgments of guilt) and one linked to the direct effect (whereby inconsistencies increased judgments of guilt). We suggest that the most plausible interpretation of these results is that the inconsistencies which made the suspect look better caused participants to become suspicious that the suspect had deliberately made mistakes in an attempt to downplay the severity of the crime. This, in turn, led to higher guilt ratings, independent of perceived consistency of the confession.

As a preliminary test of this explanation, we conducted additional exploratory analyses. In order to explore the attributions that participants made about the directional confession inconsistencies, they were asked to note if they found any inconsistencies between the two pieces of evidence and explain why they thought those inconsistencies may have occurred. Responses were coded in terms of whether or not the participants believed that the confession inconsistencies were due to the suspect deliberately attempting to make his actions seem less severe. Examples of responses coded as downplaying include ‘cash amount differed – this was probably due to the perpetrator deliberately trying to downplay his actions’ and ‘the person who robbed the place is trying to make the crime a lot less serious than it actually was.’

Chi square analysis revealed a significant difference between the consistent and better conditions on the number of participants who gave downplaying as a reason for the confession inconsistencies, χ²(1) = 11.23, p < .001. A total of 9 of the 30 participants in the better condition responded that the confession inconsistencies were due to the suspect deliberately trying to make his actions seem less severe. No participants in the consistent or worse conditions believed that the suspect was deliberately trying to downplay the crime. The responses in the worse condition were too diverse to be coded into clear categories for the purpose of further analysis. They generally reflect a broad perception that the suspect was an inexperienced criminal who was not capable of estimating large sums of money or remaining calm enough to count the number of shots he had fired. Examples of responses include ‘if this was his first offence, $2,000 may look like $10,000 to him’ and ‘this might be due to the robber being in a hurry to leave and didn’t count or was illiterate’.

A t-test was used to analyze differences in judgments of guilt between those in the better condition who specifically mentioned the suspect deliberately downplaying the crime and those in the same condition who did not mention this possibility. The mean judgment of guilt is significantly higher for those who believed the suspect to be deliberately attempting to downplay the crime (M = 8.28) compared to those who did not mention downplaying as a reason for the confession inconsistencies (M = 4.74), t (22.68) = −2.52, p = .019, d = 0.67, difference between means = 3.54 [0.63, 6.45].

These results are consistent with the idea that participants attributed confession inconsistencies that made a suspect appear better (but not worse) to the suspect having the ulterior motive of trying to reduce the perceived consequences of their crime. The lack of a similar direct effect (opposing the indirect effect) in the consistent versus worse conditions adds further weight to this evaluation of the results, in that there is no apparent ulterior motive for deliberately making oneself look worse whereas there is a plausible ulterior motive for wanting to downplay one’s involvement in a crime.

Discussion

The results show that the participants were clearly able to see when inconsistencies were present in the confessions. However, the direction of the confession inconsistencies affected juror decisions in a way that we had not hypothesized. We predicted that the two types of confession inconsistencies (better and worse) would both affect perceptions of confession consistency but that they would differ in their relationship between perceived consistency and judgments of guilt. We hypothesized that greater perceived inconsistency would translate to lower judgments of guilt for worse confessions but not for better confessions. Instead, the results show that both types of inconsistencies were well detected by participants and translated to lower judgments of guilt for both conditions. The effect of confession inconsistencies is more robust than expected, with participants noticing and acting on inconsistent evidence, regardless of the direction of the inconsistency. This finding is consistent with Woestehoff and Meissner (2016), who suggest that juror beliefs and understandings might be changing in a way that gives promise of a more thoughtful and knowledgeable juror than previously thought.

In all conditions, the effect of the consistency manipulation on judgments of guilt was mediated by perception of consistency. This indicates that the effect of inconsistencies is reliant on the participant actually noticing the inconsistencies and then perceiving them to be inconsistent. The important role that perception of consistency might play in affecting judgments of guilt can be related to theoretical reasoning about how jurors might use information in their decision-making process. Theories of decision-making imply that a juror’s internal narrative of the crime adapts to accommodate information that has not specifically been provided, such as motivation (Kelley, 1973; Pennington & Hastie, 1986), and this information may act to alter their judgments of guilt. In the present study, motivation for the confession can be extended to encompass the suspect’s reasons for giving a confession that is partly inconsistent with the facts of the crime.

Crucially, there is a difference in the direct effect of inconsistencies on judgments of guilt between the better and worse conditions. In the better condition (but not the worse) there is a direct effect indicating that inconsistencies acted to increase judgments of guilt through a mechanism unrelated to the degree of perceived consistency. This suggests that, regardless of the extent to which the confession was perceived as inconsistent, inconsistencies that were seen as an attempt to deliberately downplay the severity of the crime triggered higher judgments of guilt.

The backlash effect of the ‘better’ confession

While mediation analysis showed that the inconsistencies overall acted to reduce judgments of guilt, when the suspect confessed to a lesser version of the crime there was a backlash in which judgments of guilt increased. The apparent mechanism behind the backlash is as follows: if the participant believed that the inconsistencies were an attempt by the suspect to deliberately downplay the severity of the crime, the suspect would be given a higher judgment of guilt than if the participant did not believe that the suspect was lying for their own gain. Verbatim responses about why there were inconsistencies between the suspect’s confession and the police report frequently indicated a perception that the suspect was an experienced criminal who had deliberately lied about the crime in order to reduce his sentence. Participants in the better condition who specifically mentioned that the inconsistencies between the confession and police report could be explained by the suspect deliberately downplaying the crime severity judged the suspect as guiltier than those who did not mention downplaying as a possible explanation for the inconsistencies.

This backlash suggests that although errors in a confession can act to suppress automatic judgment heuristics in jurors (where a confession is analogous to guilt), this suppression is less likely if the inconsistencies make the suspect look like they are admitting to a lesser version of the crime. When a suspect appears to be denying the severity of the crime, thus reducing their admitted culpability, the denial acts to increase the belief that the confession is an admission of guilt, with the denial of responsibility for the full crime rejected on the basis of a strong ulterior motive to deceive. This finding highlights an additional reason why the use of minimization techniques in police interrogations are problematic; not only does minimization increase the risk of an understated false confession by suggesting to the suspect that the crime is not so terrible after all, but the resulting false confession can then inflate jurors’ perceptions of that suspect’s guilt later on if the confession details a lesser version of the crime.

Limitations

A common criticism of lab-based juror studies is that court cases are, by their nature, infinitely more variable and complex than can be replicated in a laboratory setting. In the present study there is no attempt to claim that the methodology captures an entirely realistic jury experience or that the findings can be mapped directly onto any single case involving a false confession. For example, the participants in Kassin and Kiechel’s (1996) experiment were coerced into falsely confessing to crashing a computer program in a laboratory. While those participants are clearly not in the same position as a person being wrongfully accused of murder in a police interrogation, the findings of that experiment allowed greater understanding of the situational pressures that can increase the likelihood of an innocent person falsely confessing. Similarly, our work aims to illustrate the situations in which people are able to consider the inconsistencies in confessions, thereby inspiring new thinking about how jurors might be processing evidence that is of varying quality. The present study also helps to explain how people might process directional errors when they come across them in confession evidence, and how that might then affect their perception of a suspect’s guilt. Ongoing research into the sources of variance in juror decision-making allows researchers to make advances in understanding how and why jurors sometimes make wrong decisions based on poor-quality evidence. The hope is that such research can eventually be applied to the ongoing improvement and evolution of legal processes.

Summary

This study has applied principles of attribution theory to an important applied issue, that of the ways in which jurors process confession evidence. The ways in which a juror might make attributions about the errors in a suspect’s confession is important when considering why some false confessions are accepted as truthful by juries while others are dismissed. The present study shows that any suspect who has confessed to the crime will be overwhelmingly judged as guilty, but if they admit to a lesser version of the crime then some jurors will be even more certain that they are guilty. As a jury cannot know if a confession is false, any accidental error that reduces the severity of the crime could make jurors even more certain that the suspect has committed the crime due to the error being attributed to a perceived ulterior motive (the suspect attempting to downplay their actions), even though the error itself should make jurors question the veracity of the confession.

Ethical standards

Declaration of conflicts of interest

Glenys A. Holt has declared no conflicts of interest

Matthew A. Palmer 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.

Funding Statement

This work was supported by the Community Law and Policing Research Group and the Institute for Social Change at the University of Tasmania, Australia; and by an Australian Government Research Training Program Scholarship.