Alternative stories and the decision to prosecute: an applied approach against confirmation bias in criminal prosecution

ABSTRACT

In many legal systems, preliminary proceedings and the prosecutor’s charging decision predict the outcome of the trial, as reflected in high conviction rates. This charging decision is mostly based on the final police report, which is primarily aimed to substantiate the given suspicion and may lead to a preponderance of incriminating evidence. We investigated if a written statement by the defense (an alternative story) can balance out this information disparity and can alter the charging decision. Building on classical debiasing methods, the Story Model and narrative persuasion, we hypothesized that an alternative story reduces the likelihood of being charged, because the police report becomes less unique. Participants of three experiments (criminal-law students, N = 684) received either only the police report, or the police report plus an alternative story (vs. a statement merely claiming innocence) and were asked to make charging decisions. The presence of prior conviction evidence (PCE) was varied. Results show that 1) an alternative story reduces (and PCE increases) the likelihood of being charged, 2) a statement merely claiming innocence is less but still effective, and 3) PCE effects were small and inconsistent. Underlying mechanisms, implications for criminal procedural law, and applicability to adversarial systems are discussed.

KEYWORDS:

• Prosecutor
• alternative story
• charging decision
• narrative persuasion
• confirmation bias

In criminal procedure, after the prosecutor has charged a defendant, a court of law decides the case in a public and oral trial. According to the Austrian and German codes of criminal procedure the trial thus forms the center of the criminal prosecution. In reality, however, preliminary proceedings and the final decision of the prosecutor's office are strong predictors of the verdict: If a suspect is charged with a criminal offense, the trial will most likely end with a conviction. According to the Security Report of 2017, 76.9% of the criminal trials in Austria ended with convictions (Bundeministerium für Justiz, 2017). In Germany, the conviction rate, at 82%, is even higher (Statistisches Bundesamt, 2017). Furthermore, recent research done in Belgium and the Netherlands imply that judges often try to corroborate the prosecutors’ decisions by actively looking for confirming evidence (Tersago et al., 2020) or that judges focus on excluding alternative scenarios rather than finding the scenario that best explains the evidence (Maegherman, 2021). Also, Swedish prosecutors displayed a guilt-confirming mindset once they have charged a suspect (Lidén et al., 2019). Since the structure of the law of criminal procedure triggers psychological mechanisms in prosecutors and judges, thereby shaping the decisions of both, the initial preference for the incriminating story may not be compensated for in the trial (e.g. Englich et al., 2006; Rassin, 2020; Schmittat & Englich, 2016). Collectively, this stresses the importance of balanced and well-founded decisions of prosecutors.

However, the problem that arises here is that this important and directive decision might be based on incomplete or biased information in the police report, because investigator bias or tunnel vision are known to influence the investigative process in both adversarial and inquisitorial legal systems and may lead to a preponderance of incriminating evidence at this stage of criminal prosecution (Ask & Granhag, 2005; Findley & Scott, 2006). These psychological processes could be even more pronounced in the presence of context information such as a criminal record or an untrustworthy looking defendant (Rassin, 2020). The potentially biased information the prosecutor’s office receives may then jeopardize the objectivity of the prosecutor in his or her charging decision.¹

Additionally, suspects have little opportunity to tell their story, because their rights to defend themselves, especially by introducing their own perspective, are only rudimentary during the preliminary proceedings, according to procedural law. Independent and written statements by suspects usually do not enter the preliminary proceedings, resulting in an inherent information imbalance.

The present research represents a first step to examine if this information imbalance caused by potentially biased police reports as well as the suspects’ rudimentary rights to defend themselves, could be minimized by introducing the perspective of the suspect, thus, by introducing an alternative story. This might also reduce possible confirmation biases at this and later stages. Based on the debiasing method to consider-the-opposite (e.g. Hirt & Markman, 1995) as well as research on the Story Model (Pennington & Hastie, 1986; 1992) and narrative persuasion (e.g. Green & Brock, 2000), the current studies investigated if the presentation of a written alternative story can change the prosecutors’ decision to charge the accused (or the subsequent verdict by criminal judges). A clear understanding of the cognitive processes and boundary conditions of the balancing influence of an alternative story is essential to implement an effective structural change in criminal-legal procedure.

Legal background

According to prevailing opinion, the preliminary proceedings only have a preparatory function and do not influence the verdict. Therefore, no harm is done if the defendant’s right to a fair trial within the meaning of Article 6 of the European Convention of Human Rights is not fully developed prior to the trial. Suspects are merely interrogated on the police’s terms with little or no supervision of legal counsel (§164, para.2, Austrian Code of Criminal Procedure; §§163, 163a para.3, subpara 2 in connection with §168c, German Code of Criminal Procedure) and, of course, have the option to use their right to remain silent (§7, para.2, Austrian Code of Criminal Procedure; and §136, para. 1, German Code of Criminal Procedure). However, if suspects renounce their right to silence when they tell their point of view in a criminal case, they risk that they create evidence against themselves. Therefore, adversarial processes in the preliminary proceedings require legal integration.

Furthermore, unlike in the Anglo-American criminal proceedings, Austrian and German prosecutors are bound to be objective, that is, they must investigate both incriminating and exonerating evidence. Hence, prosecutors do not act as a party, but rather are obligated to accumulate all relevant information and to base their decisions solely on the information available. The information they receive from the police, however, could be incomplete or biased due to tunnel vision in the police investigation (e.g. Findley & Scott, 2006). This could make the already difficult task to think of possible alternative explanations for the evidence even more challenging. Especially if the prosecutor is already convinced by the final police file, the task of having to think of alternative explanations can backfire and bolster previous beliefs, as research on the devil’s advocate indicates (Nemeth et al., 2001). A solution could be to offer the suspects the possibility to submit their own alternative explanation to the prosecutors in writing.

Alternative stories as debiasing method

Legal storytelling and (biased) story construction

The importance of stories for legal decision-making was demonstrated by Pennington and Hastie’s research (1986, 1988, 1992, 1993). They showed that decision makers (jurors) automatically construct stories out of the presented trial evidence. According to Pennington and Hastie’s Story Model, the story with the highest coverage (best fit between evidence and story), coherence (most consistent, complete and plausible/realistic story) and uniqueness (only this story can explain the evidence) is chosen amongst all possible stories. The chosen story then influences the verdict (Pennington & Hastie, 1992). Overall, the importance of legal storytelling in the courtroom is a widely accepted persuasion tool within the legal community (e.g. Mazzocco & Green, 2011).

Since prosecutors are confronted with the police file first, the first stories that they construct are likely of incriminating nature. This construction process could be facilitated by irrelevant context information, for example, knowing that the accused has committed a similar crime might unjustly promote the incriminating story. Previous studies have suggested that prior conviction evidence (PCE) increases perceived chance of guilt as well as conviction rates mediated by judgments of the defendant’s criminal propensity (e.g. Greene & Dodge, 1995; for a meta-analysis of the influences of various characteristics on trial outcomes see Devine & Caughlin, 2014). An analysis of German conviction statistics show that over 60% of the defendants who were convicted in 2013 did have a prior record. Half of them even had five or more prior convictions. However, it remains unclear if the bias already happens during preliminary proceedings, thus, if suspects with a prior record already get charged more often, or if the prior record increases the likelihood of being convicted again (Schlotthauser & Yundina, 2016). Either way, PCE (or other irrelevant context information such as ethnicity or socioeconomic background) probably facilitates the incriminating story construction process.

The structure of an alternative story

Arguing against an already formed story is difficult (Ask & Granhag, 2005; Luus & Wells, 1994; O’Brien, 2009; Pyszczynski & Wrightsman, 1981 Rassin, 2010). Through coherence shifting, a cognitive process that continuously evaluates and reevaluates information in a confirmatory fashion (Holyoak & Simon, 1999; Simon et al., 2004), the incriminating story may become very coherent and therefore hard to argue against. Therefore, an effective alternative story needs to overcome the disadvantages of coming second.

To improve the starting position of an alternative story, it could be beneficial to falsify the incriminating story first. Three lines of research point towards this: First, Wilson and Brekke’s classic mental correction model (1994) proposes that in order to be open for alternatives, one needs to become aware of errors to see the need to change something. Second, research on narrative persuasion suggests that story readers may not have sufficient cognitive resources available to find independently counterarguments against the presented story themselves due to perspective-taking efforts (Green & Brock, 2000). Thus, it might be necessary to do that for them. Third, Pyszczynski and Wrightsman (1981) have demonstrated that the impact of the defense’s opening (i.e. the alternative story) largely depends on the extensiveness of the prosecution’s opening (i.e. the incriminating story). The defense’s opening was only effective when the prosecution did not provide a story or when the story was weak. This implies that if a story is not weak by default it could be weakened by discrediting it. Similarly, McKenzie et al. (2002) suggested there is a reference point to which a story is compared to (called minimum acceptable strength; MAS). The stronger the first story, the higher the MAS of the second. Thus, if the prosecution presented a strong story, the reference point shifts (the defense’s MAS increases) and thus the defense’s story needs to be even stronger. If the defense’s story falls short of the MAS, this can have a reverse impact on confidence: people are even more confident that the defendant is guilty. Consequently, to introduce an alternative story successfully, falsification of the first story should precede.

After falsification, the subsequently presented alternative version of the incriminating story needs to be convincing. Yet, previous studies suggest that alternative stories have little effect to change decisions. Rassin (2010), based on Ask and Granhag’s research (2005), investigated how the suggestion of a possible alternative perpetrator (who remained quite abstract) influences evidence evaluation and guilt perception in police officers, district attorneys and judges. Results showed that introducing a possible alternative perpetrator neither significantly impacted the evidence evaluation, nor the conviction rate, but the alternative did lead to a lower perceived chance of guilt. Tenney and colleagues studied the effectiveness of ‘the other dude did it’ strategy by explicitly suggesting at trial that one of the witnesses could have had a motive as well (Tenney et al., 2009). Results showed that presenting a concrete alternative perpetrator does reduce the likelihood of guilt. Thus, merely pointing out the abstract possibility of an alternative explanation may not be successful in changing people’s mind, but a detailed alternative story of the suspect’s point of view might be able to reduce perceived guilt of the defendant.

Consequently, to introduce an alternative story successfully, falsification of the first story should precede, followed by an elaborated alternative explanation of the available evidence. This should increase the effectiveness of an alternative story.

The present studies

This research sets out to test if an alternative story, which is given to the prosecutor along with the police file, can reduce the likelihood of being charged, thereby stopping the confirmatory processing of the incriminating story that is suggested by the police report. To our knowledge, this is the first set of empirical studies on this question. Although the emphasis was on the charging decision, all studies included an item for the verdict as well to investigate potential effects of the alternative story at this stage of criminal prosecution. Based on research on debiasing methods (Hirt & Markman, 1995), the Story Model (Pennington & Hastie, 1986; 1992) and narrative persuasion (e.g. Green & Brock, 2000), our first and main hypothesis is:

H1: A main effect of an alternative story will emerge such that participants will perceive suspects to be less guilty, evaluate the likelihood of conviction to be lower, charge suspects less often and acquit them more frequently when an alternative story is presented after the finale police report, compared to judgments and decisions solely based on the final police report (Studies 1–3).

Additionally, we investigate the effects of context information on guilt perception and decisions. We chose prior convictions evidence (PCE) as one example of context information that is legally not supposed to have an impact on charging decisions or verdicts (Greene & Dodge, 1995; Oswald, 2009; Schlotthauser & Yundina, 2016). Yet, context information (such as PCE) could facilitate the construction process of and the subsequent belief in incriminating stories (Studies 1 and 2). An alternative story could counteract this effect because the incriminating story (and PCE is part of it) is falsified and the alternative explanation shifts the focus away from PCE. Therefore, we hypothesized:

H2: A main effect of PCE will emerge, such that participants will perceive suspects to be more guilty, evaluate the likelihood of conviction to be higher, charge suspects more often and convict them more frequently when the police report included prior convictions compared to no information about prior convictions (Studies 1 and 2).

H3: An (ordinal) interaction effect of alternative story and PCE is expected, such as that the alternative story will negate the effects of PCE: Whereas PCE increases guilt and conviction likelihood as well as the percentage of charging decisions and convictions, an alternative story balances out the effects of PCE (Studies 1 and 2).

Furthermore, it is of great practical importance to investigate if an alternative story in fact needs to be in a story-like format (including falsification and alternative explanation), or if any type of statement that somewhat defends the suspect is enough to level out the information imbalance and to change the prosecutor’s mind. Again, based on considering-the-opposite (Hirt & Markman, 1995), Story Model (Pennington & Hastie, 1986; 1992) and narrative persuasion (Green & Brock, 2000) our last hypothesis is:

H4: A defensive statement that repeatedly claims the suspects’ innocence but neither argues against the incriminating story nor presents an alternative scenario, does not reduce guilt likelihood, conviction likelihood, charging decisions or conviction rate (Studies 2 and 3).

Finally, we explored underlying mechanisms (Studies 2 and 3): If the alternative story can change the perception of the first story, the reference point to which the alternative story is compared to is lowered (McKenzie et al., 2002). This is important, because in Austria and Germany the defense rarely has the opportunity to present new and exonerating evidence that would make its story extremely strong (also because of limited resources and possibilities to investigate themselves). Based on the Story Model and Tenney et al.’s findings (2009), we expected that after participants have read the alternative story, the police report is perceived as less unique. An alternative story could also increase the motivation to rethink the presented evidence and to make a fair and well-founded decision. Therefore, items based on the Story Model’s certainty principles as well as motivation and attention were added.

All studies used a vignette-based approach. The materials were designed in close collaboration with criminal-law experts and extensively pretested in order to present a case with an initial charging likelihood that is higher than 50%. A relatively high initial likelihood that the suspect is charged is essential, because if the case summary by itself is not incriminating enough, the alternative story can do little to lower the likelihood. To increase ecological validity, the case vignette used in Studies 2 and 3 was much more detailed than the one used in Study 1. All studies were approved by the ethical board of the German Psychology Association.

The studies were conducted with large law-student samples from a University in Austria. This approach allows us to test psychological mechanisms extensively with adequately large samples, before replicating the effect with an expert sample (prosecutors) in follow-up studies. Besides, research has demonstrated that influences on experts’ decision making might be highly similar to those on lay people’s decisions (Englich et al., 2006; Miller, 2019; Schmittat & Englich, 2016), supporting the proposed approach.

Study 1

To test hypotheses 1–3, we composed a short vignette about an automobile-radio theft (based on a real case). Exploratory items measuring cognitive mechanisms were added to investigate the underlying effects of the alternative story.

Method

Participants and design

Two hundred sixteen current and former law-students (118 women, 87 men, 18 without specification, M_Age = 30.19, SD = 8.25, ranging from 18 to 68 years) from an Austrian University participated in this study. All participants took a criminal-law class with one of the authors in the last ten years and were contacted via E-Mail. Participation was completely voluntary and neither monetary rewards nor class-credits were provided in exchange. Fifty-six participants indicated that they had finished their studies by the time they were contacted, 146 were still studying, and 14 left this item unanswered. Participants were randomly assigned to one of four conditions of a 2 (Defense: alternative story vs no-defensive statement) x 2 (PCE: yes vs. no) between-subjects design. A sensitivity power analysis using G*Power revealed that the collected sample size for Study 1 allowed to detect a main effect of the alternative story of Cohen’s d ≥ 0.38 and an interaction of η² = .03 with 80% power and α = 05 (Faul et al., 2007).

Materials and procedure

Case

Participants were presented with a short case summary, framed as a summary of the final police report that is sent to the prosecutor at the end of the investigation. The case is an adapted version of a real case, in which ‘Alexander’ is the main suspect. Alexander supposedly broke into a number of cars in a parking garage and stole the cars’ radios. One of the car owners called the police to the parking garage when he had noticed the theft. While canvassing the area, the police officers met Alexander, who was busy putting bags into the trunk of a car. It turns out that Alexander’s car was also broken into and that someone had tried to hot wire it. Most importantly, the officers found four car radios underneath the hood of Alexander’s car – between the engine and the hood. The parking attendant further implicated Alexander by stating that he saw Alexander returning to the parking garage at least 45 min prior to the call to the police. In total, six cars were broken into in that time period. There was no video surveillance at the time and fingerprints remained inconclusive. Alexander had an entrance ticket to the parking garage three hours prior to the call to the police. Alexander was arrested for theft. He refuses to make a statement (see supplementary materials for the exact case description and the alternative story).

PCE

In the PCE condition, participants additionally received the information that Alexander has multiple prior convictions for the theft of car radios. We chose the exact same offense as prior conviction (i.e. theft of car radios) in order to have a greater probability of an effect (Oswald, 2009; Wissler & Saks, 1985). Pretests confirmed this. Information on Alexander’s prior conviction was included in the case summary.

Defense

After reading the case summary, participants in the alternative story condition received an additional statement. This statement was framed as a rebuttal from the defense attorney of Alexander, sent to the prosecution two weeks after the arrest (at that point Alexander was no longer in custody). In the statement the defense attorney describes the transpired events from Alexander’s point of view: Alexander was in the city for a business meeting and borrowed his girlfriend’s car to get there. After his meeting, he went back to the car, put his briefcase into the car and went shopping. He was putting the bags into the trunk of the car when the police approached him. His car was also broken into and the car radio was missing. However, until that moment he had not taken a closer look at the front of his car, which is why he had not noticed it. The defense attorney continues to argue how implausible it is that her client is guilty. He had both car registration papers and the key with him and would not have needed to break into it. It is also inexplicable why the car radio from Alexander’s girlfriend was not among the ones the police had discovered. According to Alexander, somebody else must have broken into the cars and needed a get-away-car afterwards. After failing to hot wire Alexander’s car, the person in question removed that car’s radio as well and fled together with the other more valuable radio. The other radios, which that person could not carry, were left behind.

Dependent variables

Guilt-likelihood ‘How likely do you think it is that Alexander committed the crime?’ and conviction-likelihood ‘How likely is a conviction?’ were measured with two Likert-type scales ranging from 0 = not at all likely, to 7 = very likely. Furthermore, participants were asked to reach a charging decision (If you were the prosecutor, how would you decide to proceed with the presented case? Charge vs. dismissal due to lack of suspicion²), as well as to reach a verdict (If you were the judge, would you acquit Alexander? Yes vs. no³). Additionally, four explorative variables were added to investigate the underlying cognitive processes such as attention (How carefully did you process the case material?), importance (How important is it for you to reach a fair decision in this case?), motivation (How high was your motivation to consider the accused’s side?), and plausibility (not in the sense of probability; how plausible is the police’s story?); all measured on Likert-type scales (0 = not at all, 7 = very much).

Procedure

Participants completed the study online. They first read the summary of the case. Half of the participants received a version with the information about Alexander’s prior convictions, the other half did not get any information about a prior record. Afterwards, participants in the alternative-story condition received the defense attorney’s statement. Participants in the no-defense statement condition proceeded straight to the dependent variables. The questionnaire ended with demographic variables. Participants had the opportunity to request information about the study, which was sent to them after the data collection had been completed.

Results

Likelihoods

Two 2 × 2 ANOVAs with Defense and PCE as between-subject variables and guilt- and conviction-likelihoods as dependent variables were conducted. Results revealed a large main effect for Defense and a non-significant trend for PCE on guilt likelihood (see Table 1 for means and standard deviations): Guilt- likelihood was higher in the conditions with no-defensive statement (M = 5.71, SD = 1.68) compared to conditions with an alternative story (M = 3.74, SD = 1.92), F(1,212) = 65.34, p < . 001, d = −1.09, 95% CI [0.80; 1.37]. Guilt-likelihood was also slightly higher with PCE (M = 4.91, SD = 2.05) than without PCE, M = 4.45, SD = 2.03 F(1, 212) = 3.78, p = .053, d =−.22, 95% CI [−0,04; 0.49], but failed to reach significance. The same pattern was revealed for conviction likelihood: in the no-defensive statement conviction likelihood was higher (M = 4.95, SD = 1.97) compared to the conditions that included an alternative story (M = 3.69, SD = 1.95), F(1,212) = 23.18, p < .001, d = 0.64, 95% CI [.37, .92]. Conviction-likelihood was also higher with PCE (M = 4.59, SD = 2.14) compared to the no PCE conditions (M = 3.97, SD = 1.92), F(1, 212) = 5.63, p = .019, d = 0.30, 95% CI [004; .57]. None of the interactions reached significance (all ps > .60).

Table 1. Means and Standard Deviations for guilt likelihood (1–7), conviction likelihood (1–7), as well as charging rate and conviction rate in % of Study 1.

	No-defensive Statement	Alternative Story	Total
PCE	M(SD) / %	M(SD) / %	M(SD) / %
perceived guilt
No PCE	5.52 (1.72)	3.43 (1.78)	4.45 (2.03)
PCE	5.89 (1.63)	4.02 (2.00)	4.91 (2.01)
Total	5.71 (1.68)	3.74 (1.92)	4.69 (2.05)
conviction likelihood
No PCE	4.68 (1.80)	3.30 (1.79)	3.97 (1.92)
PCE	5.20 (2.10)	4.03 (2.03)	4.59 (3.14)
Total	4.95 (1.97)	3.69 (1.95)	4.30 (2.05)
Charging rate
No PCE	56.00%	30.20%	42.70%
PCE	85.20%	45.80%	64.60%
Total	71.20%	38.40%	54.20%
Conviction rate
No PCE	36.00%	13.20%	24.80%
PCE	59.30%	20.30%	39.30%
Total	48.50%	17.30%	32.40%

Decisions

Logistic regressions (Enter Method) were performed to examine whether Defense and PCE and the interaction Defense x PCE predict the charging decision and the verdict. The categorical variables were dummy coded using the following codes, for charging decision: 0 = charged, 1 = not charged, for verdict: 0 = convict, 1 = acquit, for PCE: 0 = no PCE (reference), 1 = PCE, and for Defense: 0 = no defense (reference), 1 = alternative story. Logistic regression analysis indicated that the charging decision was predicted by Defense, β = 1.079, Wald χ² (1) = 6.83, p = .009, OR = 2.93, 95% CI [1.31, 6.61], as well as by PCE, β = −1.508, Wald χ² (1) = 9.98, p = .002, OR = .22, 95% CI [.09,4.56]. The odds that Alexander was not charged was 2.93 times higher with an alternative story and .22 lower with prior convictions. The interaction PCE x Defense did not predict the charging decision, β = .840, Wald χ² (1) = 1.83, p = .176, OR = 2.31, 95% CI [.69, 7.82].

Additional regression analysis revealed that Defense also significantly predicted the verdict, β = 1.317, Wald χ² (1) = 6.86, p = .009, OR = 3.73, 95% CI [1.39, 10.00], as did PCE, β = −.918, Wald χ² (1) = 5.13, p = .024, OR = .40, 95% CI [.18, .88]. Again, the odds that Alexander was acquitted was 3.73 times higher with an alternative story and .40 times lower with the presence of PCE. The interaction PCE x Defense was not a significant predictor, β = .401, Wald χ² (1) = .37, p = .543, OR = 1.49, 95% CI [.41, 5.44].

Furthermore, only 42,6% of the participants, who decided to charge Alexander, also convicted him. Therefore, to explore if Defense could have a delayed effect on the verdict, further analysis was conducted: Chi-Square Test of Independence revealed an effect of Defense, χ² (1, N = 115) = 10.07, p = .002, ω = .30: Of those participants who wanted the charge Alexander, but who did not receive an alternative story, 68.5% wanted to convict him, whereas only 38.1% wanted to convict him when they did receive an alternative story.

Cognitive processes

The three items measuring attention, importance of fair decision and motivation to take the suspect’s view into account were collapsed into one variable (Cronbach’s α = .73). A 2 × 2 ANOVA with this cumulated variable as dependent variable and Defense and PCE as independent variables revealed no significant main effects and no interaction (all Fs < 2.4, all ps> .1). The item about perceived plausibility of the police story was also analyzed with a 2 × 2 ANOVA. Results revealed a significant main effect for Defense, F (1, 207) = 12.78, p <.001, d = −0.50, 95% CI [−.78; −.23]. Participants perceived the police’s story to be more plausible without a defense (M = 5.50, SD = 1.67) than with the alternative story (M = 4.62; SD = 1.84).

Expertise/Practitioners

Since about one third of the participants had completed their law degree by the time they participated in the study, we investigated if PCE and Defense had a different effect on students compared to law practitioners. Results indicate that a law degree seems to have no effect on the likelihood ratings (guilt likelihood: F (1, 202) = 1.57, p > .2; conviction likelihood: F (1, 202) = 2.77, p = .098) and was not a significant predictor for charging decision or verdict when it was added to the logistic regression model (p >.8)

Discussion

This first study indicates that presenting an alternative story along with the final police report influences participant’s rating and charging decisions. Not only did participants think that the suspect is less likely to have committed the crime, but participants rated the conviction likelihood as lower, as well. Furthermore, Defense was a highly significant predictor for both charging decision and verdict: The odds that’s Alexander was not charged and acquitted was significantly higher when an alternative story was presented along with the police report. Hypothesis 1 was therefore supported in Study 1.

This study also investigated the influence of the suspect’s prior record. PCE increased the likelihood that the suspect is convicted for the crime and slightly increased the guilt likelihood (yet this trend did not reach significance). PCE was also a significant predictor for charging decision and verdict. The odds that Alexander was not charged and acquitted were significantly lower when prior convictions were mentioned in the police report. H2 was therefore supported.

There were no interactions between PCE and alternative story on any of the dependent variables, which we had hypothesized (H3). On a descriptive level, PCE seem to increase both charging and conviction rates without the presence of an alternative story, yet with the alternative story this increase was less pronounced. Overall, the effects of PCE were small and may have even been magnified by being identical to the present offense (Wissler & Saks, 1985). With small (d = .3) to non-significant main effects of PCE, this study was underpowered to find an ordinal interaction between PCE and Defense (Perugini et al., 2018).

Whether participants had completed their law-degree or were still studying, had neither an effect nor was it a significant predictor on any of the dependent variables. Although future research should replicate the effect of the alternative story with a prosecutor sample, our results suggest that legal experts react to the presented case materials in a similar way as law-students do.

The exploratory variables measuring underlying cognitive mechanisms suggest that neither the presence of PCE nor the presentation of an alternative story altered participant’s motivation and attention. The alternative story did reduce the plausibility of the police’s story, suggesting that presenting an alternative story changes the perception of the first story.

Further exploratory analysis revealed that charging decision and verdict are not an identical representation of guilt. Only about half of the participants who wanted to charge Alexander would also convict him. Hence, this study demonstrated that the threshold to charge somebody seems to be lower and one does not need to be completely convinced of the suspect’s guilt. There was also a carry-over effect of the alternative story: Even if the alternative story had not stopped participants from charging the suspect, it did stop them from convicting him.

Study 1 does have a few limitations: The case material was rather short; in terms of layout and editing, it did not look like a real final police report and may have not been as detailed as many other real cases. Although legal psychological research is often done with similar short materials (e.g. Ask & Granhag, 2005; Rassin, 2010), this is not an ideal replica of reality. Therefore, and in order to show generalizability of the present findings, more elaborated material is required. Also, the effect needs to be replicated with another offense and with a different alternative story. Furthermore, other explanations for the reduced charging rate need to be ruled out, for example, increased engagement in case processing through the surplus of information added by the alternative story. Although no new evidence in the sense of new DNA, video or fingerprint analyses was added in the alternative story, the alternative story did add relevant information (suspect had the keys; own radio was not among those discovered). However, since suspects are not legally bound to tell the truth and neither suspect nor lawyer submitted any supporting proof, the added information has little probative value. It is only a claim. Still, participants might have still rated the information as important. It thus remains undetermined if this surplus of information drove the effect rather than the alternative story per se. Therefore, a control condition was added in Study 2 and a different case vignette was used that allowed the construction of an alternative story without introducing any new relevant information.

Finally, the question ‘would you acquit Alexander?’ may have been too biased towards acquittal. Participants might have thought that acquit is the default or right answer. This item should be rephrased into ‘If you were the judge in this case, how would you decide?’ An attention check was also added to Study 2.

Study 2

The aim of Study 2 was twofold. First, the limitations of Study 1 are addressed (i.e. different offence and different alternative story to increase generalization and replicability; control condition) For a control condition we chose a statement of innocence (i.e. repeatedly stating that the suspect did not do it and that the evidence is weak). It could be argued that any defensive statement that is sent to the prosecution in addition to the police file can reduce charging rate, because it is a legal document and might therefore unfold its own persuasive power, or because of increased engagement with the case simply by more reading. However, we hypothesized that only an alternative story can significantly reduce the likelihood to charge the suspect (H4), because narratives are particularly convincing (van Laer et al., 2014).

The second aim was to further explore the underlying mechanisms of the alternative story’s effect. The effect on plausibility in Study 1 provides initial evidence that the alternative story influences how the first story (police’s story) is perceived (i.e. less plausible). To get a more sophisticated assessment, we included more items that were aimed to measure the different features of a story, based on the Story Model’s certainty principles: coverage, coherence (which includes consistency, completeness and plausibility) and uniqueness (Pennington & Hastie, 1992). It is important to note here, that Pennington and Hastie defined plausibility as ‘the extent to which the story is consistent with knowledge of real or imagined events in the real world’ (Pennington & Hastie, 1992, P. 191) and did not mean ‘probable’ in the sense that it is likely to have happened that way, which would be more like the uniqueness principle. We expected that the alternative story should lower the uniqueness of the police’s story, because through the other explanation, the police’s story becomes less unique. Although Tenney et al. (2009) did not measure the certainty principles directly, their results support this hypothesis. Coverage and coherence, however, should remain the same, because the alternative story neither provides new evidence and therefore does not cover more information, nor does it render the police’s story as less consistent or implausible in the sense that it is unrealistic.

Method

Participants and design

Three hundred and fifteen current law students from an Austrian University participated in this study. Eleven participants failed the attention check item (‘In order to ensure data quality, please move the slider to the number nine’). They were excluded from analysis, resulting in a total of 304 participants (169 women, 129 men, 6 unknown, M_Age = 30.48, SD = 9.09, ranging from 18 to 60 years). All participants took a criminal-law class with one of the authors within the last two years and were contacted via E-Mail. Participation was completely voluntary and neither monetary rewards nor class-credits were provided in exchange. Participants were randomly assigned to one of six conditions of a 3 (Defense: no-defensive statement vs. statement of innocence vs. alternative story) x 2 (PCE: yes vs. no) between-subjects design. A sensitivity power analysis using G*Power revealed that the collected sample size for Study 2 allowed to detect main effects of d = .36 and interactions of η² = 0.03 with 80% power and α = 05 (Faul et al., 2007).

Materials and procedure

Case

As in Study 1, participants were presented with a short case summary (pre-tested). The case summary was about three pages long and was therefore much more elaborated than the case in Study 1. Furthermore, the layout and editing of the case file replicated a real final police report as much as possible (see Supplementary Online Materials). The case describes an armed robbery. The alleged perpetrator is ‘Ralf’ who is suspected to have threatened the victim ‘Olivia’ with a knife, and to have taken her purse including 150 Euros in cash as well as her iPhone. The robbery occurred in an underpass in the late evening. Both entryways of the underpass were equipped with a surveillance camera, yet, only one camera was working that night. The robbery occurred close to the entrance / exit with the broken camera. On the surveillance tape of the working camera, a man wearing a hooded sweatshirt enters the underpass around the time of the robbery. Olivia was already in the underpass and talking on her phone when she was attacked. Olivia gave a description of her attacker and is certain that the man on the video was indeed her attacker. One of the police officers identified the person on the tape as Ralf, because Ralf had been in contact with the police before (the robbery occurred in a small city, therefore one of the officers recognized Ralf). The knife was found in the underpass, but there were no usable fingerprints on it. Police officers did not find the purse or the iPhone in Ralf’s apartment. The iPhone could not be located electronically. Ralf was interrogated and said that he was indeed in the underpass, but that he left once he had walked through. He said that he saw a woman, but that she was on the phone and had her back turned towards him. He is unable to provide more information. Ralf was arrested for armed robbery.

PCE

In conditions with PCE participants additionally received the information that Ralf has two prior convictions, one for minor assault with a knife, and one for shoplifting. These convictions were chosen because they contain both elements of an armed robbery, signaling that Ralf is – in theory – capable of the crime that he is accused of.

Defense

Participants either received no-defensive statement at all, a statement of innocence, or an alternative story. The key differences between the statement of innocence and the alternative story are that the alternative story provides valid arguments against the police’s story (i.e. the police file) as well as providing an alternative version of what could have happened instead. The statement of innocence, on the other hand, simply states eloquently that the police’s argumentation is unbelievable, and that Ralf did not do it.

More specifically, the alternative story is submitted by Ralf’s defense attorney and is about 2/3 of a page long. The alternative story begins by describing that the police’s story cannot be true for a number of reasons, for example, because Ralf was only recorded entering the underpass, but there is no footage of the crime itself. The attorney points out that the camera on the other entrance did not work, therefore other people who entered or exited the underpass on that side remained unseen. Also, there were no fingerprints of Ralf on the knife – the attacker must have worn gloves. The alternative story continues by depicting the night of question from Ralf’s point of view. Ralf went to a concert and he would not have been able to enter the concert with a knife in his pocket (security checks at the entrance). On the way home to his wife and children, he went through the underpass. Since it was cold that night, Ralf put his hands into the pockets of his sweatshirt, as can be seen in the video. Had he been wearing gloves this would not have been necessary. The attorney adds that it is much more logical that another person entered the underpass from the other entranceway (the one with no working camera) after Ralf had already left. That person also left the underpass that same way, since it was not only the quickest way out of the underpass, but that way the attacker remained undetected.

In the statement of innocence, participants are informed that Ralf’s attorney submitted a statement to the prosecution (2/3 of a page). The statement begins by simply repeating information that is already known from the police report in order to create a statement that is similar in length to the alternative story, but without any new interpretations. It is added that Ralf went to a concert and was on his way home. In the underpass, Ralf noticed a woman who had her back turned towards him and who was on the phone. He left the underpass and had no idea that the woman was attacked until the police showed up at his apartment. The statement continues by repeatedly expressing that Ralf cannot be the attacker, because the body of evidence is poor. The circumstantial evidence does not imply Ralf’s guilt. Furthermore, anybody going through an underpass would be a potential attacker according to the police’s argumentation. A hooded sweatshirt is also a typical piece of clothing and does not only fit Ralf’s description. Furthermore, Ralf did not have a motive to commit such a crime. The attorney adds that it could even be speculated that the crime did not take place at all. The victim is not credible and it is possible that she made up the whole story. In conclusion, Ralf was just at the wrong place at the wrong time (see supplementary materials for the exact case description, the alternative story, and the statement of innocence).

Main dependent variables

Guilt-likelihood and conviction-likelihood were measured with two Likert-type scales ranging from 1 = not at all likely, to 9 = very likely. Furthermore, participants were asked if they wanted to charge Ralf (If you were the prosecutor, how would you decide to proceed? Charge vs. dismissal due to lack of suspicion), as well as to reach a verdict (If you were the judge, which verdict would you choose? Acquittal vs. conviction).

Certainty principles

Participants were asked to indicate their agreement with 18 statements (scale: 0 = strongly disagree, 7 = strongly agree). Items for coverage (3 items) and coherence (3 items for each subscale: consistency, completeness and plausibility), were based on Yale’s narrative believability scale (2013) and translated into German, for example an item for coverage was ‘There were lots of ‘holes’ in the final police report’ and an item for coherence was ‘It was easy to understand the final police report’. We also constructed three items to measure the uniqueness of the presented story, e.g. ‘The description of the crime in the case file is the only reasonable explanation for the available facts’, and also added three more general items that were aimed to measure participants’ overall perception of the police’s story as well, for example ‘I think it is highly likely that perpetrator and victim acted the way it was described in the case file (see Supplementary Online Materials for a detailed description of all items).

Procedure

The procedure was identical to Study 1. After the verdict, the certainty principle items followed. The questionnaire ended with demographic variables. Participants had the opportunity to request information about the study, which was sent to them after the data collection had been completed.

Results

Likelihoods

Two 3 × 2 ANOVAs with Defense and PCE as between-subjects variables and guilt- and conviction-likelihoods as dependent variables revealed a main effect for Defense on both guilt- (F(2,297) = 17.45, p < . 001, η_p²= .11, 90% CI [.05, .16]) and conviction-likelihood, F(2,298) = 17.30, p < . 001, η_p²= .10, 90% CI [.05, .16], see Table 2 for all Means and Standard Deviations. As predicted, post-hoc tests with Bonferroni correction show that guilt-likelihood was significantly higher with no-defensive statement (M = 6.50, SD = 1.52) compared to statement of innocence (M = 5.67, SD = 1.98, p = .002, d = .47, 95% CI [.19, .74]) and compared to the alternative story (M = 5.00, SD = 1.85, p < .001, d = .89, 95% CI [.59, 1.18]). Additionally, guilt-likelihood was significantly lower in the alternative story condition than in the statement of innocence condition (p = .022, d = −.35, 95% CI [−.07, −.63]). Similarly, post-hoc tests with Bonferroni correction show that conviction-likelihood was higher with no-defensive statement (M = 5.98, SD = 2.09) compared to both statement of innocence (M = 4.33, SD = 2.12), p < .001, d = .74, 95% CI [.45, 1.03] and alternative story (M = 4.41, SD = 2.13), p < .001, d = .70, 95% CI [.41, .99], but there was no significant difference between alternative story and statement of innocence (p = 1.0).

Table 2. Means and Standard Deviations for guilt likelihood (1–9), conviction likelihood (1–9), as well as charging rate and conviction rate in % of Study 2.

	No-defensive Statement	Statement of Innocence	Alternative Story	Total
PCE	M(SD) / %	M(SD) / %	M(SD) / %	M(SD) / %
guilt likelihood
No PCE	6.70 (1.63)	5.96 (1.97)	4.91 (2.08)	5.88 (2.02)
PCE	6.32 (1.41)	5.48 (1.97)	5.08 (1.66)	5.62 (1.78)
Total	6.50 (1.52)	5.68 (1.97)	5.00 (1.85)	5.74 (1.89)
conviction likelihood
No PCE	5.87 (2.20)	4.78 (2.02)	4.00 (1.90)	4.91 (2.18)
PCE	5.91 (2.00)	3.98 (2.14)	4.74 (2.25)	4.83 (2.27)
Total	5.89 (2.09)	4.33 (2.12)	4.41 (2.13)	4.87 (2.22)
charging rate
No PCE	80.90%	58.70%	51.20%	64.00%
PCE	75.90%	55.70%	53.80%	61.00%
Total	78.20%	57.00%	52.60%	62.70%
conviction rate
No PCE	47.80%	28.30%	18.60%	31.90%
PCE	49.90%	29.50%	26.90%	34.80%
Total	48.50%	29.00%	23.20%	33.40%

Analyses revealed no main effects of PCE on guilt likelihood (F (1, 297) = 1.24, p = .266, η_p²= .004, 90% CI [0, 0]) or on conviction likelihood (F (1, 297) < .01, p > .9, η_p²= .0, 90% CI [0, 0]). Furthermore, a significant interaction became apparent between Defense and PCE on conviction-likelihood (see Table 2 for means and standard deviations), F(2,298) = 3.35, p = .037, η_p²= .02, 90% CI [0, .05]. Simple effect analyses show that with no-defensive statement, there was no difference between no-PCE and PCE conditions on conviction-likelihood, F(1,298) < .01, p >.9. η_p²= .00, 90% CI [0, 0]. With a statement of innocence, conviction-likelihood was slightly higher with no-PCE than with PCE, but this trend failed to reach significance, F (1,298) = 3.81, p = .052, η_p²= .01, 90% CI [0, .04]. The opposite trend was revealed with the alternative story. Here, conviction-likelihood was slightly lower with no-PCE than with PCE, but again this trend failed to reach significance, F(1,298) = 2.92, p = .088, η_p²= .01, 90% CI [0, .04].

Decisions

Logistic regressions (Enter Method) were performed to examine whether Defense and PCE and the interaction Defense x PCE predict the charging decision and the verdict. The categorical variables were dummy coded using the following codes, for charging decision: 0 = charged, 1 = not charged; for verdict: 0 = convict, 1 = acquit, for PCE: 0 = no PCE (reference), 1 = PCE, and for Defense: 0 = no defense (reference), 1 = statement of innocence, 2 = alternative story. Results indicated that Defense was a significant predictor, Wald χ² (2) = 8.89, p = .012. Compared to the no-defensive statement, the odds that the suspect is not charged is 2.97 times higher with a statement of innocence, β = 1.089, Wald χ² (1) = 5.22, p = .022, OR = 2.97, 95% CI [1.17, 7.56] and 4.03 times higher with an alternative story, β = 1.394, Wald χ² (1) = 8.43, p = .004, OR = 4.03, 95% CI [1.57, 10.33]. On a descriptive level, 78.20% of the participants in the no-defensive statement condition, 57.00% in the statement of innocence and 52.60% in the alternative story chose to charge the suspect Ralf.

Similarly, Defense was a significant predictor for verdict, Wald χ² (2) = 8.73, p = .013. Compared to the no-defensive statement condition, the odds were 2.33 higher that the suspect was acquitted with a statement of innocence, although this contrast failed to reach conventional levels of significance, β = .845, Wald χ² (1) = 3.67, p = .055, OR = 2.33, 95% CI [.98, 5.52]. The odds for an acquittal were also 2.63 times higher with an alternative story compared to the no-defensive condition, β = .967, Wald χ² (1) = 4.62 p = .032, OR = 2.63, 95% CI [1.09, 6.35]. On a descriptive level, 48.50% of the participants in the no-defensive statement condition, 29.00% in the statement of innocence condition and 23.20% in the alternative story condition chose to convict Ralf (see Table 2 for more percentages). PCE and the Defense x PCE interaction were not significant predictors (both Wald χ² < .5 and ps >.5).

Certainty principles

The 18 items that measured how the police report was perceived were analyzed via principle component analysis with Varimax rotation. Three factors with eigenvalues > 1.0 emerged, accounting for 62.3% of the data. The three factors correspond with the three certainty principles, although some items that were aimed to measure coherence measured uniqueness instead; Coherence (8 items, Cronbach’s α = .88), coverage (4 items, Cronbach’s α = .78), and uniqueness (6 items, Cronbach’s α = .88).

Multiple 2 (PCE) x 3 (Defense) ANOVAs revealed only a significant main effect of Defense on the scale uniqueness, F(2, 299) = 5.46, p = .005, η_p² = .04, 90% CI [0.01, .07], all other Fs < 1, all ps > .5. Uniqueness, which represents the belief that only the police’s story can explain the existing information of the case, was higher when no-defensive statement was presented (M = 3.95, SD = 1.34) compared to when the statement of innocence (M = 3.75, SD = 1.25) or the alternative story (M = 3.35, SD = 1.31) were presented. Post-Hoc analysis with Bonferroni correction revealed that only the difference between no-defensive statement and alternative story was significant (p = .004, d = .45, 95% CI [.17, 74]). There was no significant main effect of PCE and no significant interaction (all ps > .3). To investigate if the certainty principles were differently affected by the Defense, a repeated measures ANOVA with the certainty principles coherence, coverage and uniqueness as repeated factor and Defense as between-subject factor was conducted. Analysis revealed a non-significant interaction between the certainty principles and the Defense F < 1.4, p > .2.

Discussion

Using different stimulus materials, Study 2 further supports the hypothesis that presenting an alternative story compared to no-defensive statement (H1) not only reduces the perception that the suspect is guilty of the accused crime, but the suspect is also charged and convicted less often. However, the statement of innocence was not as ineffective as predicted (H4). A possible explanation for the unexpected effect of the statement of innocence is that, in hindsight, many elements of an alternative story are actually included in the statement of innocence as well, making it more of an alternative story than had been intended. For example, the statement of innocence points out that a black hooded sweatshirt is a very common piece of clothing and implies that the suspect is not the only one who fits the description. In combination with the information that the victim’s story is questionable – another hint that there is an alternative explanation – this might have pushed participants to decide in favor of the suspect.

Inconsistently with Study 1, PCE had almost no effect, thus, both H2 and H3 could not be supported. There might be a simple explanation for this: The suspect Ralf is recognized by the police officer, because he had been in contact with the police before. This information is identical in all conditions, independent of the PCE manipulation. Therefore, having been in contact with the police might be perceived to be just as bad as having been convicted. This would support the hypothesis that the mechanism behind prior convictions is the defendant’s underlying criminal propensity (Greene & Dodge, 1995). Besides, the given interaction between Defense and PCE on conviction likelihood is difficult to interpret. Since this was a small effect (η_p²= .02), it may be an incidental finding and should therefore be replicated.

The items measuring the perception of the police report (based on the certainty principles) suggests that uniqueness was perceived differently: After being presented with an alternative story, participants believed less that the police’s story was the only story that can explain the evidence. This is in line with the results of Study 1, where plausibility of the police’s story was reduced by the alternative story. Since the interaction between certainty principles and Defense was insignificant, it cannot be concluded that the manipulation affected the principles differently. The expectation that an alternative story might not need to provide a better explanation per se (in the sense that it explains more evidence [coverage], or that it is more logical [coherence]), but that it could be enough to give an impulse to rethink the first story as the only explanation (uniqueness), could not be supported.

Study 3

This last study was conducted with the purpose to get a closer look at the difference between alternative story and the statement of innocence. All of the information in the statement of innocence that hinted at an alternative explanation was removed in order to get a statement of innocence that only repeatedly stresses the innocence of the suspect. Additionally, based on our results in Study 2, a condensed version of the certainty principle scale was carefully construed.

Method

Participants and design

One hundred and seventy current law students from an Austrian University participated in this study. Two participants failed the attention check item (same as in Study 2) and four participants were already familiar with the case. They were excluded from analysis, resulting in a total of 164 participants (95 women, 63 men, 6 left this item unanswered, M_Age = 31.64, SD = 10.06, ranging from 19 to 65 years). All participants took a criminal-law class with one of the authors within the last two years and were contacted via E-Mail. Participation was completely voluntary and neither monetary rewards nor class-credits were provided in exchange. Participants were randomly assigned to one of three conditions (Defense: no-defensive statement vs. statement of innocence vs. alternative story). A sensitivity power analysis using G*Power revealed that the collected sample size for Study 1 allowed to detect an effect of Defense of η_p² ≥ 0.06 with 80% power and α = 05 (Faul et al., 2007).

Materials and procedure

Case

The same case was used as in Study 2. For this study, there was no manipulation of PCE. All participants received the case materials that included two prior convictions of the accused, one for minor assault with a knife and one for shoplifting (see supplementary materials for the exact case description, the alternative story, and the statement of innocence).

Defense

As in the previous study, participants either received no-defensive statement, a statement of innocence, or an alternative story. In the statement of innocence, we deleted all passages that contained an alternative explanation of the events. For example, the information that the suspect went to a concert that evening, or that he was on the way home to his wife and his kids when he went through the underpass was erased. Also, we removed that anybody who uses an underpass could possibly be guilty, that a black hooded sweatshirt is a typical piece of clothing in young adults that does not only fit the suspect, and that all evidence was only circumstantial evidence. Additionally, the part that questions the victim’s credibility was deleted. Subsequently, in order to have a statement that is about as long as the alternative story, two sentences with little legal meaning were also added: ‘the accusations do not fit the standards of modern criminalistics’ and ‘the accusations against my client are nothing more than a conglomeration of assumptions’. The alternative story was kept nearly identical, but two sentences about the client wearing gloves were moved from the last paragraph to the first paragraph.

Dependent variables and certainty principles

The same main dependent variables as in Study 2 were used. In response to Study 2, we reduced and adapted the certainty principles items to a total of nine: three items for coverage (one item was changed), three items for coherence (one item for plausibility, one for completeness, and one for consistency), and three items for uniqueness (identical to Study 2).

Procedure

The procedure was identical to the two previous studies.

Results and discussion

Likelihoods

Two ANOVAs with Defense as between-subjects variable and guilt- and conviction likelihoods as dependent variables revealed a main effect of Defense on guilt likelihood, F(2, 161) = 14.67, p < .001, η_p² = .15, 90% CI [.07, .23]. The suspect’s guilt was perceived to be higher when there was no-defensive statement (M = 6.47, SD = 1.44) compared to statement of innocence (M = 5.96, SD = 1.93) and alternative story (M = 4.72, SD = 1.75). Post-hoc analysis with Bonferroni adjustment indicates that only the differences between no-defensive statement and alternative story (p <.001, d = 1.06, 95% CI [.66, 1.51]), and between statement of innocence and alternative story are significant (p = .001, d = 0.65, 95% CI [.28, 1.06]) but not the difference between no-defensive statement and statement of innocence (p = .376, d = 0.3, 95% CI [−.69, .09]). This supports our hypothesis (H4) that it needs to be an alternative story to reduce perceived guilt. A statement claiming innocence is not as effective.

Furthermore, results indicate that the effect of Defense on conviction likelihood failed to reach significance, F(2,161) = 2.61, p = .077, η_p² = .03, 90% CI [0, .06]. On a descriptive level, participants did perceive it to be more likely that the suspect gets convicted in the no-defensive statement condition (M = 5.62, SD = 2.20) compared to the statement of innocence condition (M = 4.77, SD = 2.20) and the alternative story (M = 4.83, SD = 2.06).

Decisions

As in Studies 1 and 2, logistic regressions analyses (Enter Method) were performed to examine whether Defense can predict the charging decision and the verdict. Results indicate that Defense predicted the charging decision, but just failed to reach conventional levels of significance, Wald χ² (2) = 5.96, p = .051. Since our hypothesis stated that the alternative story, but not the statement of innocence affects the decision, we continued with the analysis. As predicted, compared to the no-defensive condition, the statement of innocence did not predict the charging decision, β = .591, Wald χ² (1) = .40, p = .140, OR = 1.81, 95% CI [.82, 3.96], but the alternative story did, β = .987, Wald χ² (1) = 5.94, p = .015, OR = 2.68, 95% CI [1.21, 5.93]. The odds that the suspect is not charged, was 2.68 times higher with an alternative story, compared to the no-defensive condition. On a descriptive level, the charging rates was 69.8% in the no-defensive statement condition, 56.1% in the statement of innocence condition, and 46.3% in the alternative story condition.

Defense was also a significant predictor for the verdict, Wald χ² (2) = 6.23, p = .044. Compared to the no-defensive statement condition, the statement of innocence did not reduce or increase the odds of Ralf getting acquitted, β = .567, Wald χ² (1) = 1.86, p = .173, OR = 1 .76, 95% CI [.78, 3.99], but the odds that Ralf is acquitted was 3.15 higher with the alternative story compared to the no-defensive statement, β = 1.149, Wald χ² (1) = 6.11, p = .013, OR = 3.15, 95% CI [1.27, 7.853]. Again, on a descriptive level, 39.2% convicted in the no-defensive statement condition, 26.8% in the statement of innocence condition and 17% in the condition with an alternative story. Thus, whereas the statement of innocence seems to reduce the conviction rate as well, it was not a significant predictor, supporting H4.

Certainty principles

Reliability analyses revealed that the adapted items for coverage and coherence were suitable (Cronbach’s α = .88 and Cronbach’s α = .77, respectively). Uniqueness, the scale that remained identical to Study 2, showed unacceptable reliability (Cronbach’s α = .59). The item ‘a different description of the crime could not explain the available facts as well as the description of the final police report’ seems to be an unsuitable item for the scale – if deleted reliability improves to α = .74. Therefore, analysis was conducted without this item. ANOVA analyses revealed that Defense had no influence on any of the certainty principles (uniqueness: F(2, 158) = 1.77, p = .175, η_p² = .02, 90% CI [0, .06]; coverage: F(2, 158) = 1.48, p = .23, η_p² = .02, 90% CI [0, .06]; coherence: F(2, 158) = .36, p = .699, η_p² = .005, 90% CI [0, .03]). On a descriptive level, the means of the uniqueness scale were highly similar to those of Study 2 (no-defensive statement: M = 3.56, SD = 1.42; statement of innocence: M = 3.38, SD = 1.69; alternative story: M = 3.00, SD = 1.61), but N was smaller in this Study, suggesting that this study was underpowered to detect an effect.

General discussion

Up to now, the presentation of a written alternative story by the defense is not part of the preliminary proceedings in Austria and Germany. The police’s report only consists of a one-sided, incriminating story, which guards the investigative process. Prosecutorial decisions to file charges are therefore based on potentially one-sided and biased information. Focusing on this problematic structural imbalance in the informational basis of prosecutorial decision making, the presented studies provide first empirical evidence that an elaborated alternative story by the defense, presented in a written format, can change the ratings and decisions of law-students acting as prosecutors. Three studies employing two different case files presenting two distinct offenses show that an alternative story can reduce guilt likelihood, conviction likelihood, charging rate and conviction rate, supporting H1. Additionally, Study 1 suggests that law-experts (i.e. law-practitioners) respond to the alternative story in a similar manner as law-students do, however, future studies should replicate this effect with a prosecutor sample.

We also tested the hypothesis that under certain conditions the prosecutorial process is biased towards an assumption of guilt (H2). For instance, prejudice against the accused due to the accused’s ethnicity (e.g. Sommers & Ellsworth, 2001) or socio-economic status (Hagan & Parker, 1985) can bias judicial decision-making and might also apply to the context of prosecutorial decision-making. In all these cases, an alternative story might be especially important to balance out biased information processing. We investigated prior conviction evidence (PCE) as an example for disadvantageous information that may serve as an interpretation aid for the incriminating story. The results were mixed. Whereas Study 1 indicates that PCE increases guilt and conviction likelihoods as well as charging likelihood and convictions, Study 2 did not replicate the main effect of PCE. It seems that there are a number of boundary conditions for when and how PCE impact legal decisions: The prior convictions need to be similar to the present offense (Wissler & Saks, 1985), but also relevant (e.g. not an old juvenile record, Clary & Shaffer, 1980), and important for the story, especially when there is not much other evidence available. For instance, in Study 1, the prior conviction may have led to the conclusion that the accused has both knowledge and experience to break into cars, steal the radios, and hide the stolen goods beneath the hood. In Study 2, both convictions (shoplifting and assault with a knife) are elements of armed robbery, but they are not identical to the present offense. Also, the suspect in Study 2 was not unknown to the police, which may already convey criminal propensity, making the presence or absence of prior conviction irrelevant. Overall, the present research provides some support of the existence of the effect of PCE (H2), but the effect is small (see also, Oswald, 2009) and boundary conditions seem to determine the existence of the effect.

Due to the inconsistency and small effects of the PCE effect, H3 cannot be evaluated conclusively. Future studies should replace PCE for other biased information such as ethnicity or socio-economic status.

The last hypothesis (H4) was only partially supported. Whereas a statement of innocence was not as ineffective as we had predicted (Study 2), this could be explained by the unclear operationalization of the statement. Compared to Study 3, the statement of innocence in Study 2 unintentionally contained elements of an alternative story. When these elements were removed, the predicted effect became apparent. Overall, these results underline the importance of telling the suspect’s perspectives and suggest that even some elements could be beneficially for the suspect. Additionally, these results also point out that more research needs to be done to study the individual contribution of the specific characteristics of the alternative story (e.g. combination of falsification and presentation of alternative version).

Furthermore, the presented research explored if and how the alternative story changes the perception of the first story (the police’s story). It was expected that an alternative story makes the police report weaker. Therefore, plausibility of the police’s story (Study 1) and the Story Model’s certainty principles (Study 2 and 3) were measured (Pennington & Hastie, 1992). Results indicated that participants hardly changed their perception of the final police’s story after reading the alternative story. However, the plausibility of the police’s story was reduced by the alternative story (Study 1), and the police’s story was also perceived as less unique (Study 2, but not in Study 3). Although these effects were rather small, it might have been enough to tip the scales in favor of the suspect. Still, the used scales (partially adapted from Yale, 2013) need more extensive testing, due to fluctuating reliability (especially the uniqueness scale).

Recently, Willmott et al. (2018) developed a Juror Decision Scale that is also based on the certainty principles. They proposed 16 items, including items that measure all of the principles, in addition to believability of the two competing stories (complainant and defendant), as well as decision confidence. Analysis shows that the items can be collapsed into three variables: Decision confidence, complainant believability and defendant believability. A greater belief in the complainant’s story was associated with guilty verdicts, whereas a greater belief in the defendant’s story was associated with not-guilty verdict, providing support for Pennington and Hastie’s certainty principles (1992). Furthermore, Willmott and colleagues’ results also demonstrate that inferences and pre-existing attitudes influenced believability of the stories, implying that the persuasiveness of a story is complex and multifaceted. Overall, to which degree these principles determine the legal decisions in the end, or if a more generalized feeling of doubt or distrust in the investigative process, which is caused by the alternative story, could also be important factors, still needs to be determined.

The present research extends the rather limited experimental research on decision-making in the preliminary proceedings. To this date, psychological research on legal decision-making (in either inquisitorial or adversarial legal system) has mostly focused on jurors and juries (e.g. Blandón-Gitlin et al., 2011; Carlson & Russo, 2001; Daftary-Kapur et al., 2010; Ellison & Munro, 2009; Mazzella & Feingold, 1994; Sommers & Ellsworth, 2001) and occasionally on professional judges (Eisenberg, 1994; Englich, 2006; Englich et al., 2005; 2006; Miller, 2019; Rachlinski et al., 2007; Schmittat & Englich, 2016). However, very few empirical studies have focused on prosecutors’ decision-making. The majority of these studies are case analyses (e.g. Alderden & Ullman, 2012; Franklin, 2010; Holleran et al., 2010) or interviews with prosecutors (e.g. Darwinkel et al., 2014). Only a handful are experimental studies (e.g. Darwinkel et al., 2015; Lidén et al., 2019; Pezdek & O’Brien, 2014). Most of these studies investigate the influences of various case characteristics in sexual assault cases. Therefore, more research in this specific area of legal decision-making is needed and the current research contributes to this. Although future studies should replicate the effect of alternative stories with prosecutors, the present research uses a close alternative (law-students and law-practitioners).

Last but not least, participants were asked if they would charge the suspect as well as asking for a verdict. Results of all studies indicated that these two decisions are not identical, because suspects were charged much more often than they were convicted. Therefore, a different benchmark and different decision-making strategies are used for the charging decision, implying that research that investigated how verdicts are made cannot simply be transferred to prosecutorial decision-making easily.

International relevance and transferability to the adversarial system

The possibility of a potentially one-sided or even biased final police report landing in the hands of the prosecution where it can influence charging decisions is not limited to the inquisitorial system; it is also a possibility in the adversarial system. Conceptually, inquisitorial and adversarial legal systems are fundamentally different (Herrmann, 1971): In inquisitorial legal systems, public prosecutors and the courts are responsible for determining the truth by means of objective and thorough clarification of the facts. The function of the defense is limited to correcting or adding to the narrative developed by prosecutor and judge. In contrast, in the adversarial system, it is assumed that the facts of the case are best determined if the two opposing parties, prosecution and defense, develop their own hypotheses, with contradictions or unsolved issues being resolved by means of a competition between the parties. This scenario, however, which represents the quintessence of adversariness, does not reach its highpoint until the trial, and 95% of cases charged do not proceed to trial (Hertel, 2010; Trüg, 2008). For this reason, there are remarkable parallels between the two systems as far as preliminary proceedings are concerned: In the adversarial system, a number of different authorities – magistrates, prosecutors, grand juries (Gerberding, 2005) – decide how to proceed with a case. At this preliminary stage, the important adversarial element is essentially missing – just as in the inquisitorial system. Suspects are not automatically entitled by law to introduce their own perspective (Schmid, 1993). Confrontation of different perspectives is not possible until, at the earliest, the plea-bargaining stage. However, introduction of the defense perspective at this point may be too late. Similar to developments in the inquisitorial system, by the time plea negotiations have begun, prosecutors in the adversarial system may have already made up their minds on the basis of the final police report.

Consequently, problems experienced in inquisitorial and adversarial systems during preliminary proceedings are similar. Given the ubiquity of cognitive biases in legal decision-making (e.g. Englich, 2009; Rachlinski et al., 2007) and given the numerous studies that have shown that these biases are largely independent of expertise in any kind of legal decision-making (see e.g. Miller, 2019), we can assume that our results can also be transferred to adversarial systems if charging decisions are made under similar conditions.

Limitations and future research

The created alternative story offers a different perspective and a new interpretation of the already presented facts including the suspect’s point of view. Generally, the alternative story implies that the evidence is not as incriminating as the summary of the preliminary investigations suggests. While the alternative stories did not introduce new evidence (in the sense of DNA or fingerprint analyses, surveillance videos or affidavits) they do include new information. For example, in Study 1 the suspect explained why he was in the parking garage, that the car radio of his girlfriend’s car was not among the ones that were found or that he was in possession of the car keys. Whereas introducing new information is probably inevitable, because telling a different perspective automatically increases the amount of available information, it needs to be discussed if the alternative story’s effects could also be explained by it.

We believe that the introduction of ‘new’ information in our alternative stories does not explain the effect: First of all, even though the information may be new and relevant, the probative value of it is low. Since suspects are not legally obliged to tell the truth and can lie, the included new information of the alternative story does not carry a lot of weight yet. It is only a claim and neither the suspect nor his lawyer delivered any proof. The participating law students should know this. Nevertheless, this knowledge might not have protected them from still using this information for their judgment (see also research on inadmissible evidence, Wistrich et al., 2005). To eliminate this problem, the alternative stories used in Studies 2 and 3 kept the new information to a bare minimum. The suspect’s story only introduces context information that tells a different perspective but that is based on the same evidence.⁴ The effect of the alternative story was still there.

Additionally, research on narrative persuasion indicates that argument-based persuasion attempts are not as convincing as stories (Adaval & Wyer, 1998; Chestek, 2010), implying that the surplus of information without the context of the story are likely less effective. This is supported by the results of Study 2’s statement of innocence, which indicate that already some context information (i.e. the suspect went to a concert first and then home to his family) paired with unspecific statements implying innocence, is better than no defense. This notwithstanding, future research should disentangle this more.

It could be argued that an alternative story is not necessary at all, because it already is the prosecutor’s task to think of alternative explanations in order to determine the truth (at least in Austria and Germany). From a psychological perspective, thinking of alternatives can be a potent strategy to reduce decision-making biases. For instance, ‘consider-the-opposite’ is an effective approach to reduce biases in a number of different domains (e.g. Hirt & Markman, 1995). Therefore, generating an exonerating alternative should consequently be a good strategy to balance out biased incriminating reports. Yet, there is a potential backfire-effect: What if the prosecutor cannot think of an alternative explanation? If it is too difficult to simulate an alternative story, this difficulty can strengthen the prosecutor’s initial belief in the incriminating story. Likewise, research on the effectiveness of playing the devil’s advocate, which is comparable to the consider-the-opposite strategy, has shown that this strategy may ultimately bolster the initial position, because, if the dissent is not genuine, it does not create original ideas (Nemeth et al., 2001). Consequently, self-generated alternative stories may not lead to the desired results of a balanced information evaluation but may backfire instead. This suggests that presenting the alternative story could be the safer strategy to convince the prosecutor of the exonerating story.

Practical implications and conclusion

The decision to charge is based on potentially incomplete or biased files and shapes the course of the following trial, as this decision – the first notification the court receives – may cause the judge(s)/ jury to cling to the hypotheses of the investigative authorities. Motions of the defendant that seek only to initiate additional investigation without a concrete factual basis are usually not successful. Compensation of this imbalance by cautious reasoning and attainment of the principle of in dubio pro reo is in fact hardly possible and may be legally inadmissible. This is true even if the court is aware of deficiencies in the preliminary proceedings; it can and may not compensate for reliability problems in the investigations with cautious reasoning. The drawback of doing so would be an increase in the acquittal rate and, indeed, an increase in unjustified acquittals (Velten 2016).

If further research replicates the present results, it should be discussed if the submission of a written alternative story during the preliminary proceedings needs to be institutionalized. This would imply a number of changes in the code of criminal procedure, for instance, extending the defense of the accused in a way that would make the submission of the written alternative story possible (similar to civil proceedings). The court should not draw any conclusions when the defendant chooses to remain silent apart from that.

Acknowledgements

Birte Englich, co-author of this article, suddenly passed away in September 2019. She greatly contributed to this project and was very involved in the first version of this manuscript. Her wit and insights will be missed tremendously.

Data availability statement

The data that support the findings of these studies are available from the corresponding author upon request.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Austrian and German prosecutors are bound to be objective. Objectivity could thus be jeopardized through biased information in inquisitorial legal systems.

2 Study 1 also included the decision option ‘dismissal under certain conditions’. Only 4.7% of the participants chose this option, because dismissal under certain conditions is as an extrajudicial guilt- decision: According to Austrian law, dismissal under certain conditions means that the suspect is not charged and therefore is not tried in front of a court of law, yet, there is still enough evidence that a conviction would be highly likely. The suspect agrees to take at least some responsibility for the crime, although a confession is not necessary. The suspect therefore agrees to the conditions and to accept the consequences (e.g., monetary fine, community service, out-of-court offence resolution). When the participants who chose this option were completely removed, analyses lead to similar results. Since participants of Study 1 clearly did not understand or did not know when to choose the possibility to dismiss under certain conditions this option was deleted for the following studies.

3 If participants acquitted Ralf, they were asked about the reason for this decision (open question). We expected that the alternative story elicits more specific reasons for acquittal, because specific reasons were provided in this defensive statement compared to no reasons in the no-defensive statement or generalized reasons in the statement of innocence. The provided reasons (the top four) were independent of the Defense condition (all ps > .5, for detailed information on the listed reasons see Supplementary Online Materials).

4 Both statement of innocence and alternative story started with the sentence that the suspect Ralf was no longer in custody. Whereas this could imply that Ralf is not as guilty, otherwise he would still be in custody, the Austrian system only keeps suspects in pretrial detention if the crime is severe, the risk of fleeing is high, or the risk of covering-up evidence is high. Whereas it should have been included in the police report that Ralf was released, this does not imply that he is not guilty.