The link between suggestibility, compliance, and false confessions: A review using experimental and field studies

Expert witnesses and scholarssometimesdisagree on whether suggestibility and compliance are related to people’s tendency to falsely confess. Hence, the principal aim of this review was to amass the available evidence on the link between suggestibility and complianceand false confessions. We reviewed experimental data in which false confessions were experimentally evoked and suggestibilityand compliancewere measured. Furthermore, we reviewed field data ofpotential false confessions and their relationship with suggestibility and compliance. These diverse databases converge to the same conclusion. We unequivocally found that high levels of suggestibility (andto a lesser extentcompliance)were associated with an increased vulnerability to falsely confess. Suggestibility measurements might be informative for expert witnesses who must evaluate the false confession potential in legal cases.


| INTRODUCTION
Are certain types of people more likely to confess to a crime that they did not commit? The issue of false confessions has attracted wide scientific and legal attention in the past decades. One reason for this is that false confessions are a prominent source of wrongful convictions (Kassin, 2017). Although some scholars have asserted that there are important individual differences that might impact people's willingness to confess to a crime that they did not commit (e.g., Gudjonsson, 2010Gudjonsson, , 2018, other scholars have argued that individual differences do not play a significant role in the susceptibility to form false confessions (Israëls, 2011;Mergaerts, 2019;Rassin & Israëls, 2014).
This disagreement can also be found when scholars are appointed as expert witnesses to testify on the potential false confessions, as happened, for example, in the highly publicized Dutch case of Kim V. In the summer of 2006, police were called by a mother who reported that someone had hurt her 2-year-old daughter and her 6-month-old son. When the police arrived at the scene, both of the children were found to have been stabbed. The mother, Kim V., claimed that a man had come to her apartment and had murdered her children. The police noticed bloodstains on her clothes and arrested her for the murder of her children. In custody, the young mother kept denying her involvement in the murder of her children. After several long police interrogations, she confessed to stabbing her children.
Shortly after, she recanted her confession, claimed her innocence and stated that she falsely confessed to the murder of her children in order to attend their cremation service. This case posed a difficult task for investigators: Did this young mother commit filicide or did she falsely confess to the murder of her children (De Ruiter & Kaser-Boyd, 2015)?
In 2007, the court acquitted the defendant in this case. However, the prosecution appealed the case and the appellate court appointed three expert witnesses (two cognitive and one legal psychologist) to investigate the defendant's statements obtained during the police interrogation (De Ruiter & Kaser-Boyd, 2015). When confronted with cases like these, the question arises whether expert witnesses need to be informed about possible mental health problems and personality traits (e.g., suggestibility) of a suspect so as to be able to determine whether risk factor for the formation of false confessions are present in the case at hand. In the Kim V. case, the expert witnesses solely received the tapes of the interrogation. Only one of the three expert witnesses (a legal psychologist) requested additional material (including mental health reports of the suspect). The more fundamental issue here is whether there is an established relationship between certain individual differences such as suggestibility and compliance and false confessions. In the Kim V. case, expert witnesses did not agree on whether it would be relevant to measure Kim V.'s suggestibility.
In the current review, we have assembled different lines of work on the relation between suggestibility, compliance and false confessions.
Specifically, we will present a quantitative analysis on experimental and field studies on these individual differences and false confessions.

| FALSE CONFESSIONS
Individuals confessing to a crime they have not committed is counterintuitive and defies human common sense. Yet, reported wrongful conviction cases around the world have revealed that false confessions do occur on a scale that was previously thought to be impossible (Huff & Killias, 2008;Kassin et al., 2010). In the US, two institutionsthe Innocent Project and the National Registry of Exonerationskeep track of miscarriages of justice cases and their causes. Based on their databases, between 12% and 27% of wrongful convictions involve a false confession as the leading cause of a legal miscarriage of justice (Innocence Project, 2020; The National Registry of Exonerations, 2020). This is not surprising as confession evidence is potent, especially in the eyes of a jury. One study found that up to 73% of recanted false confession cases ended up with a guilty verdict, even when contradictory evidence was available (Scheck et al., 2000). Given that false confessions and false pleas to guilt are counterintuitive, yet potent antecedents of miscarriages of justice, some authors have concluded that the body of knowledge on false confessions is, in fact, outside of the common knowledge of juries and professional judges and that expert testimony on false confessions should be admissible (Chojnacki et al., 2008). Should such testimony involve dispositional factors?
The prominent taxonomy of Kassin and Wrightsman (1985; see also Kassin & Gudjonsson, 2004) differentiates between three types of false confessions. The first type are voluntary false confessions, which emerge without external pressure during a police interrogation. Reasons for confessing voluntarily are, for example, to protect someone else or to gain fame. The second type are coerced-compliant false confessions, which occur due to police pressure and coercive interrogation techniques. In this category, confessors still know that they are innocent but often just want to get out of the uncomfortable interrogation situation.
The third type consists of coerced-internalized confessions that occur when innocent suspects are starting to wrongly believe in their guilt due to police pressure and the suggestive nature of the interrogation.

| Experimental studies
False confession research has accumulated at a steady pace over the last three decades. Experimental studies have demonstrated that false confessions can relatively easily be induced in a laboratory setting. A major impetus for the experimental work was the by now classic ALT-key experiment (Kassin & Kiechel, 1996; see for a meta-analysis: Stewart et al., 2018) in which students were asked to type letters on a computer that were read to them (either slow or fast) by a confederate.
Participants were instructed to not hit the ALT key, but then were accused of hitting that key and got blamed for crashing the computer.
Overall, 69% signed a statement admitting to hitting the ALT key. Under certain conditions (fast pace and the confederate posing as a false witness who claimed to have seen the participant hit the key), all innocent participants signed a written confession. Even though this study demonstrates how easy it is to produce false confessions, it has been criticized for its lack of ecological validity. Specifically, critics argued that unlike false confessions in the Kassin and Kiechel paradigm, false confessions in real life will come with serious negative consequences (e.g., Horselenberg et al., 2003). With this critique in mind, several other experimental paradigms have been developed to study the prevalence of false confessions in a laboratory setting (e.g., Horselenberg et al., 2003;Perillo & Kassin, 2011;Russano et al., 2005).
For instance, Horselenberg et al. (2003) conducted a conceptual replication of the Kassin and Kiechel (1996) paradigm with several adjustments to increase its generalizability to real life situations. Additionally, their aim was to examine whether a relationship would exist between false confessions and individual differences such as compliance, suggestibility, fantasy-proneness, dissociation, and cognitive failures. In their study, participants were asked to type letters that were appearing on the computer screen and were instructed to not touch the SHIFT-key to avoid crashing the computer and losing all of the data. After a while, the computer did crash and participants were accused by the experimenter (who claimed to have seen that they touched the SHIFT-key) to be responsible for the computer crash. Different from the original ALT-key paradigm, participants were paid for their involvement in the experiment, and were threatened with losing 80% of their financial remuneration if they failed to comply with the instructions. Despite these changes, results were in line with the original study. Interestingly, false confessions did not appear to be related to individual differences in suggestibility and compliance. Similarly to Horselenberg et al. (2003), Klaver et al. (2008) used the Kassin and Kiechel (1996) paradigm to examine whether individual differences (e.g., suggestibility) would affect proneness to false confessions. In line with the previous study, Klaver and colleagues were not able to demonstrate a relationship between compliance and false confessions.

| Field studies
In contrast to these mixed findings, several field studies have shown a relationship between suggestibility and the likelihood to falsely confess (Gudjonsson, 1990(Gudjonsson, , 1991a(Gudjonsson, , 1991b(Gudjonsson, , 1992(Gudjonsson, , 2003(Gudjonsson, , 2010(Gudjonsson, , 2018Gudjonsson & MacKeith, 1990;Sigurdsson & Gudjonsson, 1996). In one study, Sigurdsson and Gudjonsson (1996) interviewed a large sample (N = 509) of Icelandic prisoners about false confessions. Additionally, all participants in this study underwent psychological testing with various instruments, the most relevant ones for this article being the Gudjonsson Suggestibility Scale (GSS; Gudjonsson, 1997) and the Gudjonsson Compliance Scale (GCS; Gudjonsson, 1997). The field study demonstrated that compliance was significantly related to reported false confessions overall whereas suggestibility was specifically related to internalized false confessions. In another study (Gudjonsson, 2010), 34 British cases between 1989 and 2009 were examined in which convictions had been overturned on appeal because of false confessions. What was found was that suggestibility and compliance were the vulnerabilities of most importance to the appeal. Of course, the limitation of field studies is that the ground truth is oftentimes unknown. Thus, it remains unclear whether the reported confessions of the prisoners are, in fact, false.
It is important to realize that false confessions are not merely the result of individual differences such as suggestibility and compliance.
There are a host of others factors that can contribute to innocent people falsely confessing to a crime. Specifically, research has suggested certain risk factors such as situational (e.g., style of police interrogation) and personal (e.g., age, IQ) risk factors can increase the risk of a false confession (e.g., Gudjonsson, 2018;Kassin & Gudjonsson, 2004;Leo, 2009;Stewart et al., 2018; see, for a metaanalytic review: Meissner et al., 2014). For example, case studies have also shown that specific vulnerable populations (those with mental health disorders and those of young age) have been overrepresented in proven false confession cases (Kassin et al., 2010; see also Blair, 2007). Furthermore, people who show memory distrust (i.e., doubts about one's own memory functioning) are more likely to make false confessions than people who do not have memory distrust (e.g., Gudjonsson, 2017;Van Bergen et al., 2008).
Recently, these different factors have been synthesized in the cumulative-disadvantage framework (Scherr et al., 2020). This framework stipulates different phases (e.g., precustodial interviews, custodial interrogations) containing several factors that might lead an innocent suspect to falsely confess leading to wrongful convictions.
For example, this framework stresses that manipulative police tactics such as suggestive interrogation techniques can lead to false confessions. Of importance for the current review is that this framework also mentions that at several phases (i.e., precustodial interviews, custodial interrogations, and guilty pleas and trial convictions), vulnerable suspects such as those who are highly suggestible and compliant are at risk to falsely confess to a crime.

| REALITY MONITORING, SUGGESTIBILITY, COMPLIANCE, AND FALSE CONFESSIONS
A promising way to look into the theoretical link between internalized false confessions and suggestibility and compliance is by resorting to the Reality Monitoring framework (Johnson & Raye, 1981). Reality monitoring refers to how people decide whether a memory originates from an internal source such as imagination or an external source such as perception. The basic tenet is that people are likely to attribute a memory to an external source when a memory representation contains a multitude of details such as visual and auditory details. However, memories are likely to be attributed to an internal source when a representation lacks, for example, detailedness and distinctiveness. Obviously, internal-external source discrimination is relevant to the field of false confessions, particularly internalized false confessions (see, for an example, Gudjonsson et al., 1999). Innocent suspects who are suggestively interrogated about a crime might start to imagine how such a crime could have been perpetrated by them (e.g., Gudjonsson, 2017;Henkel & Coffman, 2004;Kassin, 2017). Through repeated imagination exercises, they might attribute their mental representation to an external source while it actually originated from an internal source. Furthermore, internal-external source distrimination might become more difficult in people suffering from memory distrust which has been linked to internalized false confessions as well (see Gudjonsson, 2003Gudjonsson, , 2018Gudjonsson & MacKeith, 1982).
Certain individual characteristics might make the discrimination between internal and external sources difficult. An often-mentioned characteristic in this context is suggestibility. Suggestibility refers to people's tendency to acquiesce to external suggestion and subsequently incorporate this misleading information in their memory reports (Gudjonsson, 1997). The link between suggestibility and false confessions has a long tradition and the sentiment is that suggestible individuals are more easily influenced by external suggestions, which might make them prone to falsely confess. Gudjonsson (1997) developed the Gudjonsson Suggestibility Scale (GSS) as an instrument to measure individual levels of suggestibility. In the GSS, people have to read a short story and afterwards are asked to freely recall the story. Following this, they have to answer 20 questions of which 15 are misleading. They also receive feedback that some answers were incorrect. This way, the GSS measures several aspects of suggestibility. The Yield score (maximum: 15) refers to the extent to which people go along with the suggestive/ misleading questions. The Shift score (maximum 20) refers to people's tendency to change their answers when they receive feedback on their answers and how they give in to interrogative pressure. Total suggestibility (maximum 35) is calculated by summing up the Yield and Shift scores. Psychometric research has shown that the internal reliability of these different GSS parameters is generally acceptable (e.g., Gudjonsson et al., 2016;Merckelbach et al., 1998).
Another individual characteristic related to the risk of reality monitoring errors is compliance. Compliance refers to the extent to which people go along with leading questions even if they know the correct answer (Gudjonsson, 1989). The standard questionnaire to measure levels of compliance is the Gudjonsson Compliance Scale (GCS; Gudjonsson, 1997). This scale consists of 20 true/false statements measuring compliant behavior (e.g., "I give in easily to people when I am pressured"). The more people accept these compliantrelated statements, the higher is their total compliance score (maximum score = 20). Psychometric indices of the GCS, such as, Cronbach's alphas, are generally adequate (e.g., Ray & Jones, 2012).
As with suggestibility, when people are high on compliance, they may under some circumstances eventually start to have reality monitoring difficulties leading to internal-external confusions (see, for an example, Levy & Gudjonsson, 2006).
Collectively, the discussed individual characteristics (i.e., suggestibility, compliance) have been put forward to explain the formation of false confessions. In what follows, we will provide a (meta-analytic) review of experimental and field studies in which the link between some (or all) of these individual characteristics and false confessions have been examined.

| THE CURRENT REVIEW
The paramount aim of the current paper is to provide a review of whether suggestibility and compliance might drive the formation of false confessions. To investigate this, we will first focus on studies in which false confessions were experimentally induced and in which suggestibility and compliance were measured as well. Second, we will discuss several field studies where detainees claimed to have falsely confessed to a crime and in which individual differences were measured. The reason for examining this link by using multiple sources (i.e., experimental and field studies) is because the combined findings from these separate sources will provide us with an overarching and more comprehensive picture of this link. some studies point to a link between false confessions and individual characteristics (e.g., Gudjonsson, 2010Gudjonsson, , 2018Gudjonsson, , 2013Sigurdsson & Gudjonsson, 1996) while others found that "[t]here was no evidence that individual differences modulate participants' susceptibility to false confessions" (Horselenberg et al., 2003, p. 1). A review of false confessions and suggestibility and compliance could clarify the reasons for these inconsistencies. Third, the phenomenon of false confessions has also been related to the phenomenon of false memories (Ost et al., 2001). In both phenomena, people report an event that was not experienced in reality. The category of internalized false confessions comes particularly close to false memory creation as here, people truly believe (and sometimes remember) events that they did not experience themselves. Why this commonality is relevant to stress is because recent studies have shown that there are no individual differences that are strongly related to false memory propensity (Patihis et al., 2018).
Extrapolating from this finding, one might argue that false confessions too are unrelated to specific individual characteristics. Taken together, these issues demand a critical review of suggestibility and compliance and their possible link to false confessions.

| EXPERIMENTAL STUDIES
Several paradigms have been used to experimentally evoke false confessions in the lab. One of the most studied methods is the alreadydiscussed ALT-key paradigm. The cheating paradigm is another method to experimentally induce false confessions (Russano et al., 2005). Although the cheating paradigm is very potent in inducing false confessions (i.e., false confessions rates up to 43%), individual difference factors have, to the best of our knowledge, not been explored within the context of this paradigm.
Experimental research on suggestibility, compliance, and susceptibility to falsely confess has exclusively relied on the ALT-key paradigm. We conducted a search of all papers using the ALT-key paradigm in which individual differences were measured. On November 2 to November 6, 2020, PsychInfo, Web of Science, and ProQuest were used to identify relevant experimental articles.
ProQuest was used to potentially identify unpublished work and account for possible publication bias. We used as search terms in the title and/or text "false confession" AND "individual difference," "false confession" AND "compliance," "false confession" AND "suggestibility," and "false confession" AND "personality." For a more detailed overview of our search strategy, see https://osf.io/rmtu4/.
Based on these search terms, 229 results were identified for PsychInfo, 226 for Web of Science, 618 for ProQuest. We also compared our search results with a meta-analysis on the prevalence of experimentallyinduced false confessions (Stewart et al., 2018). Based on this meta-analysis, we found one additional study in which false confessions were experimentally evoked and in which suggestibility was measured (i.e., Newring & Donohue, 2008). However, in that study, no information was given about suggestibility scores between participants who did and did not falsely confess. So, in total, we identified 1110 results (229 + 226 + 618 + 1). Inclusion criteria were studies in which false confessions were experimentally induced in adults and in which individual differences (i.e., suggestibility, compliance) were measured. Exclusion criteria were studies in which false confessions were not experimentally induced and in which children participated (Candel et al., 2005). We identified six studies that met this criterion (see Table 1). Not all information was available in the identified studies (e.g., standard deviation of age) and hence, we contacted the authors to receive the relevant information. Of these studies (k = 5), none of the authors could provide us with the missing data. 1 Importantly, in the ALT-key paradigm, when participants falsely confess, they are asked to sign a document stating that they crashed the computer. However, some participants also internalized the confession and indicated that they believe that they truly crashed the computer. This internalization happens when a confederate asks subjects what happened and when subjects then provide statements that they hit the key, they have internalized the false confession (Kassin & Kiechel, 1996). We examined all papers and compared different scores on the various individual differences questionnaires between participants who falsely confessed and those who did not. Specifically, we compared these scores separately for participants who signed (or not) and internalized the false confession (or not). To accomplish this, we calculated an effect size (Cohen's d) as a measure of the mean difference (Cohen, 1988) using the Practical Meta-analysis Effect Size Calculator (https://campbellcollaboration.org/escalc/html/ EffectSizeCalculator-Home.php). We used JASP (version 0.12.2) and used the meta-analysis module to obtain the mean effect size across studies. We applied the Hunter-Schmidt method for our analyses (Hunter & Schmidt, 2000). The data underlying the current review can be found on https://osf.io/uw5fv/

| Signed versus not-signed
We examined all experimental studies in which we could calculate Cohen's d for the difference between the means of the signed and not-signed falsely confessed participants regarding suggestibility and compliance (see Table 2). For suggestibility and compliance, we could calculate a mean Cohen's d using four studies. The attained effect sizes were of small to medium strength. Specifically, the highest effect sizes were found for suggestibility and compliance which could be seen as small-to-medium in range (see also Table 4). As can be seen, for suggestibility, our mean effect size across studies was 0.33 (95% CI 0.10-0.55; see Figure 1). Also, the heterogeneity (I 2 ) of our analysis was 34.11. For compliance, our mean effect size was 0.12 (95% CI −0.18 to 0.43; see Figure 2) and the heterogeneity of our analysis was I 2 = 21.68. Our heterogeneity analyses suggest that the inconsistency across studies is not large. Note that the effect of compliance on false confession was not significant.

| Internalized versus not-internalized
Few studies examined the link between suggestibility, compliance, and internalized false confessions (see Table 3). Because of the limited studies in this area, we will only report the mean effect sizes (see Table 4). Also, the results should be treated with much caution as we did not obtain convincing evidence for a link between internalized false confessions and suggestibility and compliance.
Accordingly, based on the available experimental work, we found evidence that participants who scored high on suggestibility and compliance were more likely to have a false confession than participants scoring lower on these traits (see also Table 4). However, although we found that the effect of suggestibility on false confession was significant, this pattern was not observed for compliance. Below, we will review field studies that have looked at individual differences and false confessions.

| FIELD STUDIES
We searched for field studies in which individuals (i.e., suspects, detainees) claimed to have falsely confessed and whose suggestibility and compliance levels were measured (see Table 5). We compared alleged false confessors with control groups by looking at Total GSS F I G U R E 1 Forest plot of experimental studies on suggestibility between participants who signed and did not sign their false confession F I G U R E 2 Forest plot experimental studies on compliance between participants who signed and did not sign their false confession scores and the total score of the GCS. As in the experimental studies, we calculated Cohen's d as an estimation of the effect size (Table 6) using the Practical Meta-analysis Effect Size Calculator (https:// campbellcollaboration.org/escalc/html/EffectSizeCalculator-Home. php). We made use of JASP (version 0.12.2) and its meta-analysis module to obtain the mean effect size across studies. We applied the Hunter-Schmidt method for our analyses (Hunter & Schmidt, 2000).
Concerning our search, on November 12, 2020, PsychInfo, Web of Science, and ProQuest were used to identify relevant field studies.
We used as search term in the title and/or text "false confessor" For a more detailed overview of our search strategy, see https://osf.io/ ctx7u/. Based on these search terms, 31 results were identified for PsychInfo, 49 for Web of Science, and 180 for ProQuest. In the reference section of one paper (Gudjonsson, 1991a(Gudjonsson, , 1991b, we identified an additional paper (Gudjonsson, 1984). So, in total, we found 261 (31 + 49 + 180 + 1). Inclusion criteria were the following. First, studies should have included individuals in forensic or prison settings claiming to have falsely confessed. Second, suggestion and/or compliance should have been measured. Based on these criteria, we identified five suitable field studies (see Table 5). 2 For suggestibility, we found a mean effect size of 1.09 (95% CI 0.75-1.43; see also  Sigurdsson and Gudjonsson (1996) Compliance False confessors: n = 62 (10.6(3.1)); no false confessors: n = 298 (9.4(3.4)) Note: NR = Not reported. Bold = same study, Italics = same study, underscript = same study, bold/italics = same study. a Deniers are subjects who continuously denied any involvement in the crime, resisters were defendants who were able to resist police interrogation, others refer to other forensic referrals, no false confessors refer to inmates not stating to have falsely confessed.

| GENERAL DISCUSSION
The overarching aim of the present review was to examine the link between suggestibility, compliance, and false confessions. Specifically, by combining available data from different lines of research (experimental and field studies), we investigated whether certain individual differences might make people susceptible to falsely confess. Based on our review, we found consistent support that high levels of suggestibility are related to the predisposition to falsely confess. High levels of compliance were also related to false confessions in the field studies.
In the experimental data, a link between suggestibility, compliance, and false confession was examined. We specifically compared scores of adult participants who signed or did not sign a false confession and who internalized or did not internalize the false confession.
Although we found that participants who signed the confession had higher suggestibility and compliance scores than those who did not sign, this difference was only significant for suggestibility. This difference was medium in size. Such a pattern of findings was not detected when focusing on participants who internalized (or not) the false confession.
When we concentrated on the field data, we also found that suggestibility and compliance scores were elevated in people who claimed to have falsely confessed. Importantly, this effect was shown to be large in nature. Of course, a limitation of these field data is that ground truth is unknown and, hence, it is not certain whether the tested people falsely confessed to a crime. Nonetheless, together with the experimental data, the consistent story is that suggestibility (and to a lesser extent compliance) seems to be related to false confessions in that higher levels of suggestibility (and to a lesser extent compliance) go along with an increased susceptibility to false confessions.
Collectively, the present review provides a rather consistent picture in that especially high levels of suggestibility are risk factors for innocent people to falsely confess. This general finding fits well with the cumulative-disadvantage framework put forward by Scherr et al. (2020). This framework specifies the various processes that can make innocent people falsely confess to a crime. The main tenets of this framework are that during a case, there are several phases (e.g., precustodial interviews, postconvictions) that cumulatively increase the probability that innocent people will falsely confess. Of importance for the current review is that this framework specifically denotes that vulnerable suspects are at risk to form false confessions during custodial interrogations, but that they are also vulnerable to engage in other types of behavior that could potentially result in wrongful convictions (e.g., waiving interrogation rights). The current review shows that suggestibility (and to a lesser extent compliance) are likely vulnerability factors in this cumulative disadvantage framework.
From a theoretical perspective, high levels of suggestibility and compliance might lead to reality monitoring difficulties thereby making it difficult to withhold external suggestive pressure (Henkel & Coffman, 2004). That is, reality monitoring errors would likely make innocents suspects at risk to internalized false confessions while high levels compliance would make suspects susceptible to all types of false confessions (see Gudjonsson, 2018). However, if true, we would F I G U R E 3 Forest plot field studies on suggestibility between participants who signed and did not sign their false confession F I G U R E 4 Forest plot field studies on compliance between participants who signed and did not sign their false confession have also expected to find higher suggestibility and compliance scores in people who internalized the confession in the reviewed experimental studies. Our review was limited to test this proposal considering the fact that the data on internalized false confessions was limited.
This latter point concerning limited number of studies is a limitation of all included experimental studies. That is, our quantitative comparisons were based on a limited amount of data with a maximum of five studies (e.g., for suggestibility and compliance). Furthermore, many of these studies had small sample sizes (but see Klaver et al., 2008). So, in general, our findings concerning the link between suggestibility, compliance, and experimentally induced false confessions should be interpreted with caution. It is clear that future work could benefit from conducting more high-powered studies. While it is true that caution should be exerted when interpreting the experimental data, the findings concerning suggestibility align well with what was found in the field data. Importantly, sample sizes in the field studies were oftentimes much larger than those found in the experimental data, although the heterogeneity in the field studies was substantial. In addition, the limited ecological validity of experimental studies could be attenuating the real effects of suggestive techniques in the propensity to create police-induced false confessions produced during high stake interrogations.
Another consideration is that healthy, critical students who participated in false confession experiments will, as a group, exhibit lowered suggestibility and compliance scores, which limits the detectability of the individual difference moderators of false confessions.
That this point has much credibility becomes clear when one compares mean GSS and GCS scores of false confessors in experimental studies and of defendants who claimed false confessions (e.g., see Tables 2 and 5). So, the lower effect sizes in the experimental studies might be due to a restriction of range and to tackle this issue, future research might attempt to include more diverse populations (e.g., elderly, children).
Our results can also be explained when we glance through the lens of the concept of memory distrust. Memory distrust refers to the phenomenon that individuals go along with external suggestions because they do not trust their own memory performance (e.g., Gudjonsson et al., 2014). Memory distrust has been implicated in people's willingness to falsely confess (Van Bergen et al., 2008).
Although there is research suggesting that suggestibility and memory distrust are related to each other (see for a review: Gudjonsson et al., 2013), the exact relationship between the two is unclear. For example, one possibility might be that suggestibility and compliance might be caused by an underlying trait such as memory distrust.
Our review started with the observation that scholars disagree on the relationship between individual differences (suggestibility and compliance) and false confessions. That is, some scholars argue that such a relationship does not exist (e.g., Israëls, 2011) while others propose that a meaningful relationship is present (e.g., Gudjonsson, 1991aGudjonsson, , 1991b.
Finding an answer to this issue is imperative considering the fact that expert witnesses might choose to test suspects on their levels of suggestibility and compliance in disputed false confession evidence.
Such information might become relevant in cases when there is also information that suggestive pressure was present during the interrogation. Based on this review, the answer to this issue is that a meaningful relationship between suggestibility and false confessions does seem to exist. However, it also has to be noted that the number of experimental studies examining the link between false confessions and individual differences was rather small, so any conclusions based on these data should be drawn with the utmost care. Furthermore, although the field data yielded large effect sizes, the main limitation with these data is that ground truth is unknown. Thus, it is not certain whether the tested population were innocent and formed false confessions. Nonetheless, if anything, different sources of data do seem to reveal that certain individual differences are risk factors to falsely confess to a crime.
The follow-up conundrum is to decide whether tests such as the GSS and the GCS should be applied by expert witnesses when they are asked for their opinion about whether a suspect falsely confessed.
Our review shows that apart from looking at other sources (e.g., information on how the interrogation was conducted), especially measuring suggestibility might be an important source of information in expert witness case work. This is important to emphasize considering the fact that many psychological tests used in the courtroom often lack sound psychometric properties, are not generally accepted by the scientific community, and are sometimes seldom tested for their scientific merits (Neal et al., 2019). The GSS however has been subjected to empirical scrutiny, has shown to possess acceptable levels of psychometrics and has been extensively examined in various populations (e.g., Merckelbach et al., 1998;Polczyk, 2005; but see also Gignac & Powell, 2009). If possible, expert witness case work on possible false confessions could benefit from measuring suggestibility (and perhaps to a lesser extent compliance), in combination with an analysis of how the confession was obtained, and as much collateral information to inform a global conclusion.
In sum, the current review focused on the link between suggestibility, compliance, and false confessions. Reviewing experimental and field studies, we can conclude that high levels of suggestibility (and to a lesser extent compliance) have been shown to elevate the risk for false confessions, to varying degrees according to the environment in which it is elicited. Although there are many other factors that might affect people's susceptibility to falsely confess (e.g., situational risk factors), suggestibility and compliance measures such as the GSS and the GCS should be considered when expert witnesses are tasked with a possible case of a false confession. Obtaining convergent evidence from various sources might assist triers of fact when rendering legal decisions on cases of possible false confessions.

ACKNOWLEDGMENT
We want to thank Hannah Clemens for her help in this project.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.