How guilty and innocent suspects perceive the police and themselves: suspect interviews in Germany
Abstract
Purpose
Suspects are central participants of a police interview and can provide crucial information on the interview interactions with the interviewers. This study examined how the way suspects perceive interviews relates to (a) their reported status of being guilty or innocent and (b) their decision to confess or deny.
Methods
A total of 250 convicted offenders completed a two-part questionnaire on their perceptions during the most recent suspect interview in which they had confessed to or denied a crime they had committed (Part 1) or not committed (Part 2).
Results
Participants reported a total of 334 police interviews – 223 for which they reported being guilty and 111 for which they reported being innocent. An exploratory factor analysis showed that three latent factors described how they viewed the interviewers and themselves: Respectful-Open Behaviors (non-accusatorial interviewer behaviour, and no pressure in suspects), Confession-Oriented Tactics by the interviewer (minimization and maximization tactics), and Suspects’ Psychological Distress (insecurity, fear, and lack of self-confidence). Suspects perceived less Psychological Distress and less Respectful-Open Behaviors in reported innocent (vs. guilty) interview situations. In reported guilty interview situations, confessions were associated positively with Respectful-Open Behaviors and Suspects’ Psychological Distress, whereas denials were associated positively with Confession-Oriented Tactics. Innocent interview situations showed a positive correlation between confessions and Suspects’ Psychological Distress.
Conclusions
In this study, suspects deliver an important message to the police and the legal system: The findings substantiate the benefits of an open-minded interviewing approach, and fail to support a confession-oriented interrogation approach.
Interviewing1 suspects is a central element of crime investigations. For suspects, a police interview is their first opportunity within a legal process to describe how they view the accusations. Their statements can be viewed as the outcome of interactions between at least one interviewer and one suspect (e.g. Moston et al., 1992). Most psycholegal literature focusing on the interview interaction has analysed police interviewers’ reports (e.g. Kassin et al., 2007), conducted experimental studies (e.g. May et al., 2017), examined videotapes of interviews (e.g. Surmon-Böhr et al., 2020), or developed theoretical approaches (e.g. Kelly, Miller, Redlich, & Kleinman, 2013). Only a few studies have examined the suspects’ perspective on interviewing (e.g. Cleary & Bull, 2019; Holmberg & Christianson, 2002; Kebbell, Alison, Hurren, & Mazerolle, 2010). However, none of these considered the reported guilt or innocence status of the suspects. Therefore, the present study aims to extend this line of research by examining whether suspects’ perceptions of the interview interactions are associated with (a) reported guilty and innocent interview situations and (b) confessions and denials.
Suspects’ statements and interview approaches
In many countries, suspects are legally entitled to decide whether to make a statement or remain silent (e.g. § 136 German code of criminal procedure). Under German law, if suspects decide to remain silent, interviewers have to end the interview immediately and are not permitted to ask further questions (i.e. the interview interaction is over). If suspects decide to make a statement, they can make either a confession or a denial. Moston et al. (1992) assumed that suspects’ initial statement within an interview is influenced by (1) background characteristics of the suspects and the offence (e.g. suspects’ age and criminal history), (2) contextual characteristics of the case (e.g. strength of evidence), and (3) the interviewers’ opening strategy. The suspects’ initial response may prompt the interviewers to change their interviewing style; and this, in turn, may influence the suspects’ subsequent decisions and, eventually, their final statement either confessing to or denying the accusation. Bull and Soukara (2010) analysed interview records in which confessions did not occur at the outset. In line with Moston et al.’s (1992) model, they found indications that suspects’ decisions on whether or not to confess might be influenced by the interviewers’ previously used interview tactics. This suggests that interview tactics are particularly relevant when suspects have not already decided to make a confession at the outset of the interview and initially deny the accusations.
In general, researchers have identified a range of tactics and techniques used by interviewers in suspect interviews. For example, in their taxonomy Kelly et al. (2013) sorted 71 tactics into six interviewing domains: rapport and relationship building; context manipulation; emotion provocation; confrontation and competition; collaboration; and presentation of evidence (see also Cabell, Moody, & Yang, 2020, for an effect taxonomy). Commonly, researchers have differentiated between two broad, contrary approaches: Information-gathering approaches aim to collect accurate and reliable information by, for example, developing rapport; explaining the conditions, purpose, and process of the interview; using open-ended questions; and generally treating suspects in a fair, respectful, empathic, and open-minded manner (e.g. Milne & Bull, 2000). In contrast, accusatorial approaches aim to break down suspects and force confessions by implementing confrontational, manipulative, persuasive, and suggestive tactics based on a guilt-presumptive mindset (e.g. Kassin et al., 2010). There are two main components of the accusatorial approach: first, so-called minimization tactics aim to give suspects a false sense of security and to trivialize the crime and its consequences by offering sympathy, face-saving excuses, and moral justifications. Second, maximization tactics aim to intimidate the suspects by exaggerating the strength of evidence and the seriousness of the offence (e.g. Gudjonsson, 2003; Kassin, 2015; Kassin & McNall, 1991).
In Germany, the legal framework stipulates that interviewers have to apply an information-gathering approach (German code of criminal procedure). However, the ways police interviewers are trained differs in each of Germany's federal states, and there is no systematic review of which approaches are actually implemented in the field (see also Volbert & Baker, 2015).
In a meta-analysis of field studies, Meissner et al. (2014) found that both the information-gathering and the accusatorial approaches increased the likelihood of a confession compared to control conditions. However, in their meta-analysis of experimental studies, information-gathering approaches led to a higher ratio of true to false confessions than accusatorial approaches. This shows that there may well be a need to distinguish between guilty and innocent suspects when examining the effectiveness of suspect interviews. In experimental studies, it may be possible to control for the ground truth reliably. Nonetheless, field studies can also deliver an approximation, although this has to be treated with a degree of caution. For example, some research on the frequencies and reasons of false confessions is based on the suspects' reported status of innocence (e.g. Gudjonsson, 2010). In the current study, we asked former suspects via self-report whether or not they had committed the accused offences. This approach allowed us to examine whether the perception of different interview approaches was associated with confessions or denials in interview situations in which the suspects reported being guilty or innocent.
Suspects’ perceptions during the interview
Gudjonsson’s (2003) cognitive behavioural model assumes that confessions are the outcome of a specific constellation of social (e.g. isolation from friends or family, police pressure), emotional (e.g. distress, anxiety, fear, feelings of guilt or shame), cognitive (e.g. suspects’ thoughts, assumptions, and perceived strategies), situational (e.g. the circumstances of the arrest and familiarity with police procedures), and physiological antecedents (e.g. heightened arousal) along with their corresponding immediate (e.g. arousal reduction) and long-term consequences (e.g. return of arousal to base level). When examining emotional states and cognitive processes from the perspective of 83 Swedish suspects, Holmberg and Christianson (2002) found that four factors underlay their perceptions of the interview interaction: Suspects described the interviewers as being (1) more or less dominant (i.e. aggressive, brusque, and impatient) and (2) more or less humane (i.e. trying to cope with the suspects, showing an empathic, positive attitude towards them); and they described themselves as being (3) more or less respectful and respected (i.e. being friendly and obliging, and feeling acknowledged as a human being) and as being (4) more or less anxious (i.e. suffering from fear, stress, and sleeplessness). In line with this, Cleary and Bull (2018) found that 418 inmates in the United States most strongly endorsed the factor Humanity/Integrity followed by the factors Rapport and Sympathy/Perspective-Taking, but disaffirmed the factor Dominance/Control when asked how the police should interview suspects.
Turning to the statements made by suspects, Holmberg and Christianson (2002) found that confessors perceived their interviewers as being more humane, whereas deniers perceived their interviewers as being more dominant. Similarly, 43 Australian inmates who had confessed during their suspect interviews perceived their interviewers as being more ethical (i.e. taking an open-minded approach) and less dominant, and they felt treated more humanely in comparison with those who had denied the accusations (Kebbell et al., 2010). Furthermore, a study with 100 Canadian inmates found that the likelihood of a confession was higher when the interviewing style was more humane (e.g. respectful, patient or calm; Snook, Brooks, & Bull, 2015). Overall, these self-report studies indicate that confessions are associated more with a humane, ethical interviewing style, whereas denials are associated more with a dominant interviewing style.
However, these field studies did not include their self-reported guilt or innocence status. This is important, because in their meta-analysis of experimental studies, Houston, Meissner, and Evans (2014) found that innocent and guilty suspects differed in their cognitive processes and their emotional states. Specifically, both true and false confessions were associated with suspects’ considering the consequences of confessing or not confessing. False confessions were also associated with the suspects’ perceptions of external pressure from the interviewers and the interview context, whereas true confessions were associated with the suspects’ emotional reactions to the interview (i.e. experiences of stress, worry, and anxiety), and their perceptions of the evidence and guilt. However, this experiment-based meta-analysis focused solely on confessions and did not take the alternative statement behaviour of denying into account.
The present study
Suspect interviews are central elements of crime investigations in which the suspects are vital participants. Surveying suspects about their police interviews offers a way to understand their perceptions, deliberations, and decisions. Therefore, we designed a questionnaire to map how convicted offenders perceived their most recent suspect interviews. We focused only on cases in which suspects had waived their rights to remain silent (i.e. had made a statement). This is because according to German law, suspects are asked whether they want to exercise their right to remain silent at the beginning of the interview. If they decide to remain silent, the police are not allowed to ask further questions, and there is almost no interview interaction. Therefore, suspects who make statements include relatively longer and more extensive interactions between suspects and interviewers. Furthermore, we asked the suspects whether they had finally confessed or denied and differentiated between guilty and innocent interview situations on the basis of suspects’ reports. That is, suspects were asked about their perceptions of their most recent guilty and their most recent innocent interview, in which they had made a statement. These interviews did not have to concern the offences the suspects were most recently convicted for. Instead, they could refer to criminal investigations the suspects were involved previously. Overall, our main objective was to examine suspects’ perceptions of their police interviews by considering their reported guilt or innocence status and their decisions to confess or deny. Another aim was to map the practice of suspect interviewing in Germany from the suspects’ point of view.
Method
Participants
Originally, 280 convicted offenders under current probation or parole were recruited for this study. However, 30 participants had to be excluded because they quitted while completing the questionnaire or gave contradictory answers. Hence, the final sample consisted of 250 participants: 211 (84.4%) were male, 37 (14.8%) were female, and two (0.8%) gave no information on their sex. Their mean age was 38.4 years (SD = 11.59) and ranged from 18 to 73 years (six missing values [2.4%]). The majority of participants had been born in Germany (n = 212 [84.8%]) and a minority abroad (n = 34 [13.6%]; four missing values [1.6%]). Educational levels varied between no school graduation (n = 34 [17.2%]), graduation after 9 years (n = 78 [31.2%]), 10 years (n = 89 [35.6%]), and 13 years (n = 37 [14.8%]; three missing values [1.2%]). Most participants reported earlier psychotherapeutic or psychiatric treatment (1–5 sessions: n = 48 [19.2%]; several months: n = 58 [23.6%]; several years: n = 48 [19.2%]; no treatment: n = 93 [37.2%]; three missing values [1.2%]).
The majority of participants (n = 190 [76.0%]) reported that they had been imprisoned during the course of their lives with a mean overall imprisonment length of 4.6 years (SD = 5.4) ranging from 1 month to 44 years (two missing values [0.8%]). The reported offences for which they were under current probation or parole were assault (n = 77 [30.8%]), theft (n = 70 [28.0%]), fraud (n = 59 [23.6%]), violation of the narcotics law (n = 58 [23.2%]), robbery (n = 36 [14.4%]), material damage (n = 27 [10.8%]), offences against personal freedom (n = 13 [5.2]%), homicide (n = 9 [3.6%]), traffic offences (n = 9 [3.6%]), sexual offences (n = 6 [2.4%]), arson (n = 6 [2.4%]), and other offences (n = 26 [10.4%]. A total of 92 participants [39.2%] reported more than one crime; 15 participants [6.0%] did not name a crime.
Procedure
We collected data in three judicial social service institutions in Berlin, Germany. Convicted offenders visit these regularly for appointments regarding their probation or parole. A research associate addressed them and informed them about the study, the voluntariness of their participation, and their anonymity. Participants were given the questionnaire, had it explained to them, and filled it out in the waiting room or in individual rooms. A member of the research team was present to answer questions and to collect the completed questionnaires. The questionnaire was read out loud to participants with reading or minor cognitive difficulties. Completing the questionnaire took approximately 25 min. All participants received 5 euros for their participation. The study was approved by the Berlin Justice Senate's criminological service.
Self-report questionnaire
We developed a self-report questionnaire based on prior research (Gudjonsson, 2003; Holmberg & Christianson, 2002; Houston et al., 2014; Kassin et al., 2010; Kebbell et al., 2010; Kelly et al., 2013; Snook et al., 2015) and adapted it to the German legal system. The questionnaire first gathered sociodemographic data (gender, age, place of birth, education); data on psychotherapeutic/psychiatric treatment; the offences for which they were under current probation or parole; and their criminal history (previous convictions, length of imprisonment during their lifetime). Next, participants were asked a set of questions about the last interview in which they had been interviewed when guilty and had waived their right to remain silent (Part 1) and then the same set of questions about the last interview in which they had been interviewed when innocent and made a statement (Part 2). That is, the reported interviews did not have to be related to the offences for which the participants were currently under probation or parole (for the current offence they possible remained silent). Participants were free to answer none, one, or both parts of the questionnaire (depending on whether they had ever experienced such interview situations). Each part started with questions about the general conditions of the interview (accused offences, age at the time of the interview). Then participants were asked to what extent they had confessed or denied the accusations on a 4-point scale ranging from 1 (confessed to all accusations) to 4 (denied all accusations). They also indicated to what extent an attorney was present during the interview on a 5-point scale ranging from 1 (does not apply at all) to 5 (applies very much).2 Subsequently, they were asked to rate the extent to which they agreed with a set of 32 items presented on 5-point scales ranging from 1 (disagree) to 5 (agree). These were 21 items on the interviewer's behaviour (e.g. ‘The police interviewer…’ ‘treated me respectfully’, ‘said that I could get a more lenient penalty for a confession’) and 11 items on the suspect's own behaviour and internal states (e.g. ‘I [the suspect] …’ ‘acted cooperatively’, ‘felt put under pressure’; see Supporting Information for all items, including M and SD values).
Statistical analyses
We performed an exploratory factor analysis (EFA) to explore the latent structure of the 32 items representing suspects’ perceptions of the interviewer and themselves. We did not separate items on the suspect from those on the interviewer because of possible dyadic factors. Because each item occurred twice (once in the guilty and once in the innocent interview situation), we used exploratory structural equation modelling (ESEM; Marsh, Morin, Parker, & Kaur, 2014), while enforcing strict measurement invariance and allowing for correlated residuals across guilty and innocent interview situations. This ensured that the latent factors would represent the same constructs in both interview situations, enabling us to make meaningful comparisons between mean levels of factor scores (Liu et al., 2017). We determined the optimal number of factors per interview situation by comparing fit indices of ESEM models with increasing numbers of factors (Fabrigar, Wegener, MacCallum, & Strahan, 1999). We focused particularly on the comparative fit index (CFI; acceptable fit >.90, good fit >.95), the Tucker–Lewis index (TLI; acceptable fit >.90, good fit >.95), and the root-mean-square error of approximation (RMSEA; acceptable fit <.08, good fit <.05; Hopwood & Donnellan, 2010). To improve the interpretability of the final model, we applied oblique geomin rotation, thereby considering possible correlations between latent factors. Means and variances of the latent factors were fixed to 0 and 1, respectively, for the innocent interview situation and were estimated freely for the guilty interview situation. This allowed us to interpret factor means in the guilty interview situation directly as mean differences between interview situations. All models were estimated using robust (i.e. means- and variance-adjusted) weighted least squares (WLSMV) estimation based on the polychoric correlation matrix (Moshagen & Musch, 2014) to account for a non-normal distribution and the ordinal scaling of the items. We expanded the final ESEM model by including suspects’ confession or denial as correlates of the latent factors. In order to explore whether the presence of an attorney or the time delay between police interview and data collection could influence our analyses, we added both as further correlates of the latent factors. Prior to model estimation, we dichotomized ratings of confession or denial (1–4; values ≤ 2 indicating confession) and presence of an attorney (1–5; values ≥ 4 indicating presence). All analyses were performed with the statistical software Mplus 8.0 (Muthén & Muthén, 2017) or R 3.5 (R Core Team, 2017).
Results
Descriptive analyses
The 250 participants reported a total of 334 interview situations in which they had waived their right to remain silent: 223 in which they reported being guilty (66.8%) and 111 in which they reported being innocent (33.2%). Suspects’ indicated mean age at the time of the police interview was 30.5 years (SD = 11.5) for guilty interviews (one missing value [0.5%]) and 28.5 years (SD = 10.3) for innocent interviews (four missing values [4.4%]). Average time between interview and data collection was 7.4 years (SD = 7.0) for guilty interviews (seven missing values [3.1%]) and 9.7 years (SD = 8.5) for innocent interviews (six missing values [5.4%]). Suspects indicated the presence of an attorney in 34 guilty interviews (15.2%) and 11 innocent interviews (9.9%; no missing values). Table 1 shows the types of offences participants were suspected of separately for reported guilty and reported innocent interview situations.
| Suspected offence | Reported guilty (n = 222) | Reported innocent (n = 111) | ||
|---|---|---|---|---|
| n | % | n | % | |
| Theft | 80 | 35.9 | 31 | 27.9 |
| Assault | 58 | 26.0 | 29 | 26.1 |
| Violation of narcotics law | 55 | 24.7 | 16 | 14.4 |
| Material damage | 39 | 17.5 | 8 | 7.2 |
| Fraud | 40 | 17.9 | 14 | 12.6 |
| Robbery | 27 | 12.1 | 15 | 13.5 |
| Offences against personal freedom | 11 | 4.9 | 9 | 8.1 |
| Homicide | 11 | 4.9 | 5 | 4.5 |
| Sexual offences | 6 | 2.7 | 6 | 5.4 |
| Arson | 1 | 0.4 | 0 | 0 |
| Other offences | 5 | 2.2 | 5 | 4.5 |
Note
- Multiple answers possible. One missing answer in guilty interviews.
In the reported guilty interview situations, 158 suspects (70.9%) stated that they had fully or mainly confessed, whereas 62 suspects (27.8%) stated that they had fully or mainly denied the accusations (3 missing values [1.3%]). In the reported innocent situations, five suspects (4.5%) stated that they had fully or mainly confessed, whereas 106 suspects (95.5%) stated that they had denied fully or mainly their guilt (no missing values). Table 2 shows this distribution in more detail.
| Type of statement | Reported guilt (n = 220) | Reported innocence (n = 111) | ||
|---|---|---|---|---|
| n | % | n | % | |
| Fully confessed | 74 | 33.2 | 1 | 0.9 |
| Mainly confessed | 84 | 37.7 | 4 | 3.6 |
| Mainly denied | 32 | 14.3 | 17 | 15.3 |
| Fully denied | 30 | 13.5 | 89 | 80.2 |
Note
- Three missing answers in reported guilty interviews.
Factor structure of suspects’ perceptions
Table 3 summarizes the goodness-of-fit statistics for the ESEM models with varying numbers of factors per interview situation. The three-factor model was chosen as the most adequate representation of the data, because all fit indices either met (RMSEA) or were close to meeting (CFI, TLI) minimal requirements for a good fit. Extracting further factors yielded only a minimal increase of fit (but led to estimation problems such as Heywood cases).
| Model* | Total number of latent variables | P | χ 2 | df | CFI | TLI | RMSEA [90% CI] |
|---|---|---|---|---|---|---|---|
| 1 factor | 2 | 195 | 3383.321 | 2077 | .839 | .844 | .050 [.047, .053] |
| 2 factor | 4 | 232 | 2830.318 | 2040 | .903 | .904 | .039 [.036, .043] |
| 3 factor | 6 | 271 | 2452.847 | 2001 | .944 | .944 | .030 [.026, .034] |
| 4 factor | 8 | 312 | 2337.268 | 1960 | .954 | .952 | .028 [.023, .032] |
| 5 factor | 10 | 355 | 2253.143 | 1917 | .959 | .956 | .022 [.022, .031] |
Note
- N = 250. The test statistics (χ2) of all models reached a level of significance of p < .001. P, number of parameters; CFI, comparative fit index; TLI, Tucker–Lewis index; RMSEA, root-mean-square error of approximation; CI, confidence interval.
- * Exploratory structural equation models based on polychoric correlations with increasing number of factors per interview situation. We enforced strict measurement invariance across interview situations by fixing item loadings and item thresholds, and allowed for correlated residuals across interview situations.
Table 4 shows item loadings for the three-factor model after geomin rotation. Factor 1, labelled Respectful-Open Behaviors, included items describing an open-minded and non-coercive behaviour by interviewers, and feelings of cooperation and no pressure in the suspects. Factor 2 was named Confession-Oriented Tactics and subsumed mainly interviewers’ tactics of maximization (e.g. the interviewer made it clear that the situation would become worse without a confession) and minimization (e.g. the interviewer said that the suspect would feel better if she or he were to confess). Factor 3 was labelled Suspects’ Psychological Distress and represented the majority of items on suspects’ emotions, cognitions, and psychophysiological aspects. A small number of cross-loadings (≥ |.4|) were observed between Factors 1 and 2 (four items) and between Factors 2 and 3 (two items). Two items (chummy treatment and substance influence) showed only marginal loadings and were not subsumed under any factor.
| Item | Factor loadings | Residual correlations | ||
|---|---|---|---|---|
| Factor 1: Respectful-Open Behaviors | Factor 2: Confession-Oriented Tactics | Factor 3: Suspects’ Psychological Distress | ||
| The interviewer… | ||||
| Treated me respectfully | .94 | −.03 | .02 | .86* |
| Took my point of view on the accusations seriously | .88 | .20 | .05 | .17 |
| Acted patiently | .87 | .001 | .10 | .31 |
| Seemed open-minded | .81 | .22 | .002 | .36* |
| Acted factually | .81 | .05 | −.04 | .21 |
| Acted in a friendly way | .81 | −.05 | .08 | −.06 |
| Acted aggressively | −.72 | .33 | .04 | .47* |
| Acted dismissively | −.71 | .12 | .03 | .23 |
| Honestly wanted to help me | .65 | .25 | −.10 | .15 |
| Wanted to provoke me | −.65 | .51 | −.06 | .16 |
| Wanted to intimidate me | −.63 | .52 | −.01 | .32 |
| Tried to convince me to make a confession | −.57 | .53 | .07 | .27 |
| Said that I would feel better if I confessed | −.21 | .80 | −.09 | .16 |
| Made clear that my situation would only get worse if I did not confess | −.31 | .78 | .01 | .48 |
| Pretended to be on my side | .01 | .74 | −.21 | .31* |
| Said that others in my situation would also have committed such a crime | −.15 | .71 | −.09 | .72*** |
| Said that I could get a more lenient penalty through a confession | −.22 | .68 | .05 | .61*** |
| Offered other advantages for a confession (e.g. phone calls, food, visitors) | −.46 | .62 | .01 | .18 |
| Made clear that she or he would not accept a denial | −.32 | .61 | .07 | .70*** |
| Wanted to trick me | −.49 | .60 | −.01 | .39* |
| Treated me in a chummy way | .2 | .36 | −.22 | .66*** |
| I (the suspect)… | ||||
| Felt put under pressure | −.52 | .19 | .29 | .30* |
| Felt relieved | .51 | .21 | −.12 | .11 |
| Acted cooperatively | .43 | −.11 | .19 | .54*** |
| Felt self-confident | .04 | .14 | −.73 | .18 |
| Felt relaxed | .21 | .24 | −.71 | .29* |
| Felt made insecure | −.24 | .01 | .71 | .39** |
| Was afraid | .12 | .32 | .71 | .60*** |
| Had concentration difficulties during the interview | .01 | .41 | .65 | .30 |
| Felt physically unwell during the interview (e.g. pain) | −.12 | .36 | .57 | .39** |
| Had sleep difficulties the days before the interview | .29 | .40 | .52 | .37* |
| Was under the influence of substances (alcohol/drugs/medicine/withdrawal symptoms) during the interview | −.05 | .35 | .36 | .91*** |
Note
- N = 250. Factor loadings >.4 printed in boldface. Factor loadings were standardized within the innocent interview situation. Residual correlations represent the rank-order stability of item residuals across the two interview situations.
- For residual correlations: *p < .05; **p < .01; ***p < .001.
Suspects’ perceptions of reported guilty versus innocent interview situations
Concerning to the (standardized) latent means, in self-reported guilty interview situations, suspects showed significantly higher values on Factor 1 (Respectful-Open Behaviors; M = 0.26, SE = 0.12, p < .05) and Factor 3 (Suspects’ Psychological Distress; M = 0.35, SE = 0.11, p = .001) compared to reported innocent interview situations (with means fixed to 0). The latent means of Factor 2 (Confession-Oriented Tactics) did not differ significantly between reported innocent and guilty interview situations (M = 0.13, SE = 0.11, p = .23; see Figure 1, Panel A-C for a comparison of the factor distributions between reported guilty and innocent interview situations).
Correlates of suspects’ perceptions in reported guilty versus innocent interview situations
Table 5 presents the correlations of latent factors and covariates within and between interview situations. Factor 1 (Respectful-Open Behaviors) and Factor 2 (Confession-Oriented Tactics) were negatively correlated within and between interview situations. All three factors showed significant autocorrelations (i.e. rank-order stability) across the reported guilty and innocent interview situations, with Factor 3 (Suspects’ Psychological Distress) showing the highest stability. Length of the time between interview and data collection did not significantly correlate with any of the factors within interview situations. The presence of an attorney was significantly negatively correlated with F3 (Suspects’ Psychological Distress) in innocent interview situations, while the remaining correlations within interviews did not reach statistical significance.
| Reported Innocence | Reported Guilt | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| F1 | F2 | F3 | Confession | Attorney | Passed Time | F1 | F2 | F3 | Confession | Attorney | Passed Time | |
| Reported innocence | ||||||||||||
| F1 | − | |||||||||||
| F2 | −.29*** | − | ||||||||||
| F3 | −.30* | .18 | − | |||||||||
| Confession | .15 | .44 | .31** | − | ||||||||
| Attorney | .25 | .32 | −.30* | .22 | − | |||||||
| Passed Time | −.11 | −.06 | −.03 | −.33 | −.46* | − | ||||||
| Reported Guilt | ||||||||||||
| F1 | .56*** | −.27* | −.15 | .44 | .04 | .23 | − | |||||
| F2 | −.30** | .65*** | .16 | .15 | .30 | 0 | −.37*** | − | ||||
| F3 | −.19 | .09 | .84*** | −.16 | −.37* | −.02 | −.06 | .12 | − | |||
| Confession | .09 | −.13 | −.07 | .38 | −.22 | −.15 | .40*** | −.19* | .22* | − | ||
| Attorney | .04 | .16 | −.15 | .46 | .42 | .20 | .17 | .18 | −.14 | .22 | − | |
| Passed Time | .23* | −.30 | −.13 | −.34 | .21 | .44*** | −.02 | .04 | .03 | −.04 | .17 | − |
Note
- N = 250. The correlations were estimated based on an expanded ESEM including six covariates (i.e. suspects’ confession or denial, presence of an attorney, and time since the interview, both for innocent and guilty interviews). Model fit was acceptable, with CFI = .945, TLI = .943, RMSEA = .027. F1 = Respectful-Open Behaviors. F2 = Confession-Oriented Tactics. F3 = Suspects’ Psychological Distress.
- * p < .05; **p < .01; ***p < .001.
Confessions (vs. denials) in reported guilty interview situations showed significant positive correlations with Factor 1 (Respectful-Open Behaviors) and Factor 3 (Suspects’ Psychological Distress) and a significant negative correlation with Factor 2 (Confession-Oriented Tactics). Reported innocent interview situations showed a significant positive correlation between confessions and Factor 3 (Suspects’ Psychological Distress). In reported innocent interview situations, correlations between confessions and Factor 1 (Respectful-Open Behaviors) and between confessions and Factor 2 (Confession-Oriented Tactics) did not attain statistical significance (for the standardized distributions with respect to the suspects’ confessions/denials, see Figure 1, Panel D-F).3
Discussion
This study provides insights into suspects’ perceptions of real police interviews and how their perceptions are associated with their reported innocence or guilt in the interview situations and their confessions or denials. An exploratory factor analysis revealed that a three-factor model most adequately represents individual differences in how suspects view interviewers and themselves. In line with previous research (e.g. Holmberg & Christianson, 2002), two of these factors describe mainly contrary approaches in police behaviour: Confession-Oriented Tactics relates to police behaviour based on minimization and maximization tactics; Respectful-Open Behaviors relates to an open-minded, respectful, friendly, patient, and factual police behaviour. In addition, three items capturing self-perceptions (absence of perceived pressure, feeling of relief, and cooperative behaviour) were associated with Factor 1. This points towards an aspect of positive reciprocity within the factor labelled Respectful-Open Behaviors. Overall, these two factors comprise the two basic descriptors of police questioning mentioned above: Confession-Oriented Tactics overlaps with the accusatorial approach involving a more coercive and psychological manipulative interviewing style in order to obtain confessions. In contrast, Respectful-Open Behaviors covers main elements of the information-gathering approach aiming to gather information through showing respect, non-coercive questioning, and an open mindset (see also Loftus, 2011; Shepherd, 2007 on interrogation vs. investigative interviewing with regard to this).
In contrast to previous research (e.g. Holmberg & Christianson, 2002), we found only one factor describing exclusively suspects’ self-perceptions. The factor Suspects’ Psychological Distress relates to negative emotions, cognitions, and physical aspects of the suspects (e.g. feeling insecure, fearful, and having concentration difficulties).
Earlier studies found that suspects’ confessions were associated more strongly with a humane (ethical) interviewing style, whereas denials were associated more strongly with a dominant interviewing style (e.g. Holmberg & Christianson, 2002; Kebbell et al., 2010). However, these studies did not differentiate between suspects’ reported status of guilt or innocence. In line with a self-report study of forensic patients in Germany (Volbert, May, Hausam, & Lau, 2019), most participants in the current study reported having been interviewed at least once when guilty (89.2%), although a considerable proportion also reported having been interviewed at least once when innocent (44.4%). This differentiation allows a more detailed examination of police interviews from the suspects’ perspective.
In reported innocent interview situations, suspects stated that they felt less Psychological Distress compared to guilty interview situations. This is in line with two psychological phenomena explaining innocent suspects’ decision-making processes (Kassin, 2005): First, in line with the belief in a just world (Lerner, 1980), innocent suspects may trust in the fairness of the justice and legal system and expect that their innocence will be believed if they ‘just tell it like it happened’. Second, based on the illusion of transparency (Gilovich, Savitsky, & Medvec, 1998), innocent suspects may believe that interviewers will be able to read their thoughts and emotions and hence will ‘see their innocence’ (see also Kassin et al., 2010). These considerations are in line with an experimental study in which innocent suspects showed less stress (especially physiological arousal) compared to guilty suspects (Guyll et al., 2013). The present study confirms this with data from real suspects reporting on their perceptions of their real police interviews. In general, Kassin (2005) viewed the rather relaxed approach of innocent suspects as an explanation for why they tend to waive their rights to remain silent – which, in turn, is a necessary antecedent for a false confession.
Regarding perceptions of police behaviour, suspects reported more Respectful-Open Behaviors in reported guilty (vs. innocent) interview situations. Assuming that interviewers predominantly conduct suspect interviews when they presume some degree of guilt (instead of being open-minded), their aim is more to gather incriminating information than exonerating information from suspects (e.g. Fahsing, 2016; Volbert & May, 2016). In the reported guilty interview situations, most suspects reported having confessed (70.9%) and hence showed the statement behaviour desired by the interviewers. In contrast, in the reported innocent interview situations, most suspects reported having denied the accusations (95.5%). Correspondingly, interviewers might have applied a more open-minded, respectful, and non-accusatorial interviewing style when questioning guilty suspects, because the suspects showed the statement behaviour interviewers were aiming to elicit. In line with this, confessions correlated positively with Respectful-Open Behaviors in the guilty interview situations.
In reported guilty interview situations, Confession-Oriented Tactics correlated positively with denials and negatively with confessions. This is in line with previous field studies showing that suspects who perceived more dominant police behaviour more often denied, whereas those who perceived more humanity more often confessed (Holmberg & Christianson, 2002; Kebbell et al., 2010; Snook et al., 2015). However, these studies did not consider the reported guilt or innocence status. Our findings indicate that this finding holds only for guilty interview situations. Nonetheless, further research is needed to examine this in more detail. Overall, our findings provide no indication that an accusatorial approach helps to obtain confessions. Instead, they are more in line with the idea that an information-gathering approach is beneficial when aiming to collect incriminating information from guilty suspects.
In the reported guilty interview situations, confessions were related to Suspects’ Psychological Distress. This is in line with Meissner et al.’s (2014) meta-analysis of experimental studies showing that true confessions were associated with suspects’ emotions. While the presence of an attorney did no not influence suspects’ perceptions and statements in general, it affected this finding. In reported guilty interviewing situations, the correlation between confessions and Suspects’ Psychological Distress did not remain significant when analysing only the cases in which no attorney was present (i.e. this relation was rather less robust). In the present study, confessions also related to Suspects’ Psychological Distress in reported innocent interview situations (but we found no correlation between false confessions and the other factors here). This contradicts Meissner et al.’s (2014) finding that false confessions were associated with external pressure. In a field study with persons suffering from serious mental illness, Redlich, Kulish, and Steadman (2011) also found that false confessors reported more external pressure and less internal pressure compared to true confessors. Nonetheless, only five suspects reported having given a false confession in the current study, and so, this finding must be interpreted with caution. Furthermore, we found no differences in Confession-Oriented Tactics between reported guilty and innocent interview situations. The experimental findings on this in the literature are mixed (e.g. Hill, Memon, & McGeorge, 2008; Kassin, Goldstein, & Savitsky, 2003; Narchet, Meissner, & Russano, 2011). Further research needs to examine these issues in more detail.
Finally, we want to highlight two descriptive results: First, a considerable number of participants reported having been interviewed when innocent (44.4%). This shows that it is crucial to conduct interviews in an open-minded manner aiming to gather both incriminating and exonerating information (instead of focusing only on confessions). Second, in the reported guilty interview situations, most suspects confessed (70.9%; denials: 27.8%), whereas in the reported innocent interview situations, most suspects denied (95.5%; confessions: 4.5%). This indicates that according to the suspects, their final statements during the interviews were predominantly correct, and this is in line with previous self-reports by forensic patients in Germany (Volbert et al., 2019).
Limitations
This study is based on retrospective self-reports with the methodological limitations these entail (e.g. social desirability and possible discrepancies compared to the perceptions of interviewers or observers). Here, an important issue is the time between the interview and data collection, which may result in memory distortions. However, for this study we found no relation between the length of time delay and suspects’ perceptions (factors), statements (confession vs. denial), or reported status of being guilty or innocent. Second, this study includes no information to validate the participants’ reports about their status of being guilty or innocent. In general, this holds true when surveying inmates (e.g. Cleary & Bull, 2018) but also police investigators (e.g. Kassin et al., 2007), who are both central participants in the interview interaction. However, suspects are the ones who ultimately decide the outcome of the interview, and their perspectives provide crucial information on the interview. Third, the cross-sectional design does not permit causal interpretations. For example, respectful and open-minded interviewing may have triggered confessions, or, vice versa, suspects’ incriminating statements may have led to respectful behaviour. Indeed, due to cognitive distortions, confessors may retrospectively describe interviewers more positively while deniers may describe them more negatively. Such methodological limitations can be countered with various research approaches (e.g. experimental studies or analysing video records). Fourth, the current sample limits the representativeness of the results. Also, it is difficult to interpret some results, as most guilty suspects confessed and most innocents denied. Future studies should include more and a wider range of participants (e.g. persons from other German federal states, non-German speakers, and suspects with intellectual disabilities). Finally, there is a link between suspects’ perception of the evidence held against them and their statements (Granhag & Hartwig, 2015; May et al., 2017). This study did not explicitly measure suspects’ perception of the evidence held by the interviewer. In line with previous studies (e.g. Holmberg & Christianson, 2002), suspects’ perception of the evidence held by the interviewer was only included implicitly in several items (e.g. ‘the interviewer tried to convince me to make a confession’). Future research should examine explicitly how real suspects’ perceptions of the interviewer's knowledge influence their decisions to make incriminating or exonerating statements.
Conclusions
In January 2020, Germany introduced a law making it mandatory to record police suspect interviews in homicide cases and with vulnerable persons (§ 136 German code of criminal procedure). This may bring light into interview rooms and suspects’ interviews. Nonetheless, it is also important to capture the subjective point of view in both interviewers and suspects in order to understand their perceptions, considerations, and decisions. This will help to explain why police and suspects behave in specific ways, thereby providing important information for lecturers, practitioners, researchers, and policymakers in police and legal settings. The current study is the first systematic examination of interview interactions in Germany from the suspects’ point of view. In line with international research, it revealed three major findings: Suspects perceived less Psychological Distress in reported innocent situations than in reported guilty situations. This might be seen as an essential factor for false confessions, because it may explain innocent suspects’ (naive) decisions in interviews (e.g. to waive their right to remain silent; Kassin, 2005). In the reported guilty interview situations, confessions were associated with a more open-minded, respectful, and non-coercive approach by the interviewer, whereas denials were associated more with minimization and maximization tactics. Overall, this study delivered an important message to lecturers, practitioners, researchers, and policymakers in police and legal settings from the suspects themselves: it outlined the benefits of an information-gathering approach and provided no support for an accusatorial approach of questioning suspects.
Acknowledgements
We thank Miriam Bach, Jasmin Ghalib, Clara Goers, Caroline Huss, Pauline Lauda, Claudia Zimmermann, and Pierre Pantazidis for their help with data collection. We acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité – Universitätsmedizin Berlin. Open access funding enabled and organized by ProjektDEAL.
Open Research
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
- 1 Although the literature sometimes uses the terms interrogation and interview interchangeably (Alison, Alison, Noone, Elntib, & Christiansen, 2013), interrogation is preferred in the United States; interview, in Europe. Because German law indicates that police should conduct interviews (instead of interrogations; Eisenberg, 2017), this term is preferred in the following unless specifically stated otherwise.
- 2 This study was part of a broader collection of data on suspect interviewing in Germany. Alongside the topic of the present study, further questions addressed (a) the frequencies of and reasons for suspects’ confessions, denials, and remaining silent; (b) suspects’ planning of the interview; and (c) their understanding of their rights during the interview. These data will be analysed in different studies.
- 3 To check the robustness of the correlations between confessions (vs. denials) and factors of suspects’ perceptions, we estimated an ESEM after excluding the 45 interviews in which suspects indicated the presence of an attorney (N = 221). In this model, confessions were still significantly correlated with F3 (r = .26, p < .05) in innocent interview situations, and with F1 (r = .34, p < .001) and F2 (r = −.24, p < .01) in guilty interview situations. However, the correlation with F3 in guilty interview situations was no longer significant (r = .16, p = .12).
