Mock jurors’ awareness of age-related changes in memory and cognitive capacity

Abstract

While age-related changes in memory have been well documented, findings about jurors’ perceptions of older witnesses are conflicting. We investigated the effect of victim age (25 vs. 75 years old) and crime severity (victim injured vs. not injured) on mock jurors’ decisions in a robbery trial. Jury-eligible participants (120 women; 84 men) read a mock trial summary and delivered their verdicts online. Mock jurors believed the young victim more than the older victim when the crime was severe, while no age differences emerged for the less severe crime. Whereas previous research demonstrated that juror characteristics were generally associated with culpability, we demonstrated that with case-specific information, these general views became less important. In all, mock jurors were aware of age-related decline in memory provided by eyewitnesses only to a limited extent. Accordingly, in trials involving older witnesses, jurors will benefit from educative information about age-related memory changes.

In the course of the last century life expectancy increased. An implication for criminal justice is that older people are more and more at risk of becoming victims of crimes. Crime statistics show that the victimization rate of older people in proportion to young people varies by crime. For instance, older people are at a higher risk of becoming victims of robbery or handbag snatching than victims of homicide (James & Graycar, 2000; Walker, 2018). Robbery is one of the cases that is most likely to be conducted by strangers against older people, along with burglary and homicide (Walker, 2018). The focus in this study was whether mock jurors were aware of the fact that identification of an unfamiliar person is less reliable in older age than in young age. Thus we chose a crime that typically involves strangers. In a jury trial it is for jurors to decide, based on the presented evidence, as to whether the defendant is guilty or not. Typically, criminal (and other jury) trials contain contradictory accounts of events (Bornstein & Greene, 2011). In cases in which the only witness is the victim, jurors are faced with the question of whether the victim was willing and able to tell the truth. In other words, they have to decide whether the victim provided an honest, truthful (i.e. credible in the sense of lack of deception) and accurate, reliable account (i.e. credible in the sense of lack of error; Sporer, 2008). These two constructs are not exclusive and need to be considered simultaneously when assessing witness credibility. This complexity is also mirrored in jury instructions, as in their decision the jury has the discretion to accept or reject the whole or a part of the evidence (Judicial Commission of New South Wales, 2016; Queensland Courts, 2016).

According to the conceptual model of juror decision-making (Devine & Caughlin, 2014), case characteristics (e.g. victim characteristics, crime type), juror characteristics (e.g. gender, experiences, legal trust) and defendant characteristics (e.g. race, criminal history) are associated with inferences made about the case, and ultimately with judgments of guilt. In all, jurors are ‘active information processors’ (Bornstein & Greene, 2011, p. 65) whose decision reflect not only evidence and legal guidelines (i.e. legal factors) but also juror stereotypes and attitudes (i.e. extra-legal factors), which in turn may affect case perceptions. The present study investigated the extent to which victim age and crime severity (operationalized as a sprained ankle vs. no injury) as case characteristics and legal trust as a juror characteristic contribute to juror decisions. Defendant characteristics were held constant in this study.

Victim age as a case characteristic

Stereotypes and attitudes are held not only about defendants but also about victims and witnesses. Here, we investigate perceptions of older people in a trial. Considering age-related reductions in the efficiency of the sensory system (Schieber, 2006) and age-related changes of brain structure (Coleman & O’Hanlon, 2008), older people have more difficulties in processing and remembering newly learned information. This applies to memory for faces (see the meta-analysis by Martschuk & Sporer, 2018), eyewitness recall of events (see the overview by Aizpura, Migueles, & Garcia-Bajos, 2014) and metacognitive judgments (e.g. Dodson & Krueger, 2006; Martschuk, Sporer, & Sauerland, 2019). For instance, despite the age-related decline in identification performance, expressed confidence that the identification was correct remained relatively unchanged across the lifespan (Martschuk et al., 2019). A meta-analysis of studies of stereotypes about older people showed that older people were perceived as less competent than young people, but the effect decreased when more detailed information about the target was provided than age alone (M. E. Kite, Stockdale, Whitley, & Johnson, 2005).

Studies that investigated perceptions of older witnesses (as opposed to older people in general) revealed conflicting patterns. While some studies found that older witnesses were perceived more positively than young or middle-aged adults (e.g. Allison, Brimacombe, Hunter, & Kadlec, 2006; Kwong See, Hoffman, & Wood, 2001; Ross, Dunning, Toglia, & Ceci, 1990), other studies showed the opposite effect (e.g. Neal, Christiansen, Bornstein, & Robicheaux, 2012). Factors, such as questioning type, speech style or demographic background were associated with perceived witness credibility (Brimacombe, Jung, Garrioch, & Allison, 2003; Nunez, McCoy, Clark, & Shaw, 1999). Further factors were the content or narrative style of a given testimony. For instance, a study investigated the quantity and quality of eyewitness statements regarding a staged crime video under direct and cross-examination of young (18–25 years), middle-aged (30–44 years) and older adults (65–85 years) and their consequences on perceived credibility (Brimacombe, Quinton, Nance, & Garrioch, 1997). Older adults were indeed less accurate under direct and cross-examination and provided fewer details of the offender than young and middle-aged adults. Consequently, when another group of participants evaluated these videotaped testimonies, they perceived older people as less credible (perceived description accuracy, perceived confidence and perceived competence) but also as more honest than younger adults.

Altogether, these findings reflect the complex construct of witness credibility, which consists of ‘witness competence, confidence, convincingness, accuracy, honesty, witness’s observation of the event, memory for the event, witness’s level of cognitive functioning, and witness suggestibility’ (Pittman, Toglia, Leone, Mueller-Johnson, 2014, p. 368), mirroring both elements of credibility: honesty and accuracy (Sporer, 2008). It includes the individual’s credibility (i.e. believability, often based on witness demeanour and attributed character) and evidentiary credibility (i.e. plausibility of the facts provided; Andrewartha, 2014).

Studies that investigated the impact of age on mock jurors’ perceptions included mock trials for the following crimes: aggravated robbery (Nunez et al., 1999), nonviolent theft (Brimacombe et al., 2003; Brimacombe et al., 1997), possession of cocaine (Ross et al., 1990), robbery-murder trial (Neal et al., 2012) and shoplifting (Kwong See et al., 2001). None of the studies manipulated type or severity of crime. The present study tested both types of credibility as a function of witness age (young vs. older adult) and crime severity (injury vs. no injury) in a robbery case scenario. Furthermore, this study considered the extent to which mock jurors’ beliefs about memory changes in old age were associated with their decisions. Due to the generally known age-related memory decline and the decreased identification performance associated with it, we hypothesized that mock jurors will be less likely to believe the testimony of a 75-year-old victim than a 25-year-old victim (Hypothesis 1a). Thus, mock jurors will perceive the prosecution case weaker (Hypothesis 1b) and the defence case stronger when the victim was 75 years old (Hypothesis 1c).

Crime severity as a case characteristic

Conflicting arguments exist as to whether and how crime severity influences juror decisions. According to the severity–leniency hypothesis (Kerr, 1978), jurors are less willing to convict when the crime is more severe. This hypothesis predicts that in order to avoid the risk of convicting an innocent person, jurors are more reluctant to convict in more serious cases (which would result in more severe sentences) and thus require more evidence of guilt. Studies that investigated the influence of charge severity supported the severity–leniency hypothesis inasmuch as jurors were more likely to convict the defendant when the charge was more severe (McComas & Noll, 1974, as cited in Kerr, 1978). This study, however, did not control for strength of evidence between the charges.

Another study showed support for the severity–leniency hypothesis only when evidence was more ambiguous in a more severe crime – that is, when the conduct was the same but more evidence was considered necessary to prove guilt in a first-degree murder versus second-degree murder versus manslaughter case (Freedman, Krismer, MacDonald, & Cunningham, 1994, Experiment 1). However, when Freedman et al. (1994, Experiments 2 to 4) controlled for the strength of evidence – that is, when proof was provided that the crime was committed – and mock jurors had to decide whether the defendant was the perpetrator, the association between crime severity (robbery vs. robbery and assault vs. robbery and aggravated assault) and verdict disappeared. In other words, mock jurors were as likely to convict when the crime was robbery as they were when the robbery was followed by assault or aggravated assault, when evidence strength was the same.

A different theoretical perspective postulates that evidence of a serious offence may evoke emotional arousal (Bright & Goodman-Delahunty, 2004). As a consequence, jurors’ perception of the defendant’s guilt is elevated by their negative emotional state. Referring to the potential prejudicial impact of gruesome evidence, the Australian Law Reform Commission (1985) cautioned that ‘the fact-finder may use the evidence to make a decision on an improper, perhaps emotional, basis, . . . logically unconnected with the issues in the case’ (pp. 351–352). Put another way, jurors may engage in a biased interpretation of emotionally laden information (e.g. verbal gruesome evidence), which in turn may increase conviction rates (Bright & Goodman-Delahunty, 2004). A recent meta-analysis confirmed that the presence of visual (photographic) gruesome evidence increased conviction rates in other studies (Grady, Reiser, Garcia, Koeu, & Scurich, 2018).

The same principle applies to the effect of crime severity and outcome severity, such that jurors confronted with evidence of a violent crime experience strong negative emotions. Accordingly, jurors may lower their threshold to convict or accord undue weight to the presented evidence when faced with the prospect of erroneously acquitting a violent perpetrator who then would return to the community (Devine, 2012).

Some evidence for the influence of crime severity on mock jurors’ emotions and verdicts was gathered in a study that compared responses to life-threatening conduct versus less threatening conduct (Goodman-Delahunty, Martschuk, & Ockenden, 2016). Mock jurors expressed more negative emotions and were four times more likely to convict the defendant when the crime was more severe. Notably, the effect disappeared when the defendant was additionally charged with a terrorist crime, presumably because the charge severity outshone or outweighed the conduct severity when terrorism was mentioned. The complex interplay of crime type (e.g. shoplifting, rape, murder, child abuse), type of evidence (none, witness, physical evidence, DNA evidence) and criminal history (none, prior conviction) was also investigated by Pearson et al. (2018). After controlling for evidence type, crime severity was associated with more guilty verdicts and higher punishment ratings.

A similar pattern to those observed in jurors’ judgments of guilt emerged in studies on sentencing decisions. Mock jurors were more punitive (i.e. awarded longer sentences) when the victim reported more severe injuries than when they reported mild injuries in cases of burglary and robbery (Nadler & Rose, 2003). Mock judges were also more punitive following unintended consequences of the defendant’s actions that involved the killing versus the injury of a third party (Sporer & Goodman-Delahunty, 2009).

Taken together, the foregoing research supports the theoretical notion that mock jurors are less conservative in their decisions and hence more likely to convict when the crime involved is more emotionally arousing and/or considered more severe. Accordingly, we hypothesized that mock jurors will be more likely to believe the complainant when she was injured during the robbery than when she was not (Hypothesis 2a). This should be reflected in the perceptions of a stronger prosecution case (Hypothesis 2b) and a weaker defence case (Hypothesis 2c). Finally, one might also predict that age of victim and crime severity exert an interactive influence – namely, older victims with a more severe injury will be perceived to be least credible and hence the prosecution’s case considered less strong (Hypothesis 3).

Juror attitudes as juror characteristics

In addition to case characteristics, the present study investigated attitudes as juror characteristics in legal decisions. The association between juror characteristics, juror attitudes and their decisions has been the focus of research since at least the 1950s (e.g. Adorno, Levinson, Frenkel-Brunswik, & Sanford, 1950). While some researchers investigated personality traits, others assessed legally relevant attitudes (for an overview, see Lecci & Myers, 2008). Studies repeatedly showed that juror attitudes have an important role in their legal decisions (see the meta-analysis by Devine & Caughlin, 2014), particularly when trial evidence is ambiguous.

As Sherif and Sherif (1969) posited long ago in the context of social perception, the more ambiguous a situation (i.e. the stimulus to be judged) becomes, the greater is the contribution of internal factors (e.g. an individual’s past experiences, social attitudes and temporary expectations) on decisions (see also Sporer & Goodman-Delahunty, 2009). In the context of criminal trials, the more conflicting or equivocal the evidence, the stronger is the contribution of juror characteristics and attitudes on the assessment of the presented evidence, and consequently on verdict. Considering that the prosecution case and the defence case are typically contradicting (Bornstein & Greene, 2011), jurors’ internal factors are likely to contribute to the evaluation of the presented evidence, at least to some extent. This was shown by a study that tested the predictive value of mock jurors’ pre-trial attitudes on pre- and post-deliberation verdicts in a simulated armed robbery trial (Lecci & Myers, 2009). Analyses revealed that mock jurors’ confidence in the justice system and conviction proneness were associated with their verdicts.

Based on this reasoning we derived the following additional hypotheses: mock jurors’ attitudes towards the justice system and towards ageing will be associated with the perceived likelihood that the identification of the perpetrator was correct (Hypothesis 4a), the probability of guilt (Hypothesis 4b) and the guilty verdict (Hypothesis 4c). Furthermore, we hypothesized that, after controlling for mock jurors’ attitudes, perceived victim credibility and perceptions of the prosecution and defence case would be associated with the perceived likelihood that the identification was correct (Hypothesis 5a), the probability of guilt (Hypothesis 5b) and the guilty verdict (Hypothesis 5c).

Method

Participants

After screening, 205 participants (120 women; 84 men; 1 unknown), aged between 18 and 85 years (M = 31.24, SD = 10.95) were included in the study. Seven participants were excluded from the study, because (a) they took less than one minute to read the trial scenario (n = 4), (b) they did not complete the study (n = 2), or (c) responses indicated misunderstanding of questions (n = 1). One participant did not provide demographic information but completed the experimental tasks and was, therefore, not excluded from the study. About half of the participants were U.K. citizens (52.7%), 31.7% were U.S. citizens, 12.7% Canadians, 4.9% Australians and 2.0% New Zealand citizens; 4% were dual citizens. Most participants were White (79.2%), 11.4% were Asian, 3.5% were Hispanic or Latino, 3.0% were Black, and 3.0% were mixed. More than half of the participants had a university degree (Bachelor’s Degree 39.7%, Master’s Degree 9.8%, Doctorate 3.9%), 21.6% had a Diploma or a trade certificate, and 25.0% had completed high school or less. One fifth of the participants reported that they had been called to jury duty (21.6), and 6.9% had served on a jury trial.

Stimulus material

Trial scenario

The trial was a written summary of a robbery trial (ca. 800 words), which included general instructions, opening of the prosecution and the defence case, evidence-in-chief and cross-examination of the victim and the investigating officer as the two prosecution witnesses. The 25- or 75-year-old victim alleged that the defendant robbed her in the parking garage by (a) grabbing her handbag; or (b) violently taking her handbag and pushing her over so she fell and sprained her ankle. She identified him in a fair photo line-up procedure at the police station. The investigating officer reported that all the investigation led to the defendant – for instance, that the police received reports that fit the description of the defendant. The defence argued that the allegations were unsubstantiated and that the victim had made a false identification. The jury instructions included information that burden of proof of guilt of the accused is placed on the Crown who must prove beyond reasonable doubt the guilt of the accused. The participants were informed that the defendant did not testify following his lawyer’s advice. All case facts were held constant, except for victim age and crime severity. The study materials are available from the first author upon request.

Questionnaire

The questionnaire included items in the named order. First, the questionnaire measured guilt both as a binary variable (guilty vs. not guilty) and as the probability of guilt (on an 11-point rating scale from 0–100%), the likelihood of a correct identification (11-point rating scale from 0–100%), and the strength of the prosecution and the defence case (7-point rating scale; 1 = strongly disagree; 7 = strongly agree).

Next, perceived victim credibility was assessed using the Eyewitness Credibility Scale with the subscales Victim Credibility (11 items; e.g. the victim accurately remembered the event; Cronbach’s α = .89) and perceived Victim Lenient Response Bias (3 items; e.g. the victim wanted to have someone to blame; Cronbach’s α = .73). Participants indicated their agreement on a 7-point rating scale from 1 = strongly disagree to 7 = strongly agree. Higher values on the subscale Victim Credibility indicate that participants believe the victim more, while higher values on the subscale Victim Lenient Response Bias indicate that participants perceived that the victim was more lenient – that is, prone to making a positive identification regardless of target presence. The inter-correlation between Victim Credibility and Victim Lenient Response Bias was r = −.24, p < .05. The results of the principal component analysis are in Appendix A.

Further, participants’ attitudes towards the justice system were assessed using the Pretrial Juror Attitude Questionnaire (PJAQ; Lecci & Myers, 2008, 2009) subscales System Confidence (6 items, e.g. a suspect who runs from police, probably committed a crime; Cronbach’s α = .70¹), Conviction Proneness (5 items; e.g. criminals should be caught and convicted by ‘any means necessary’; Cronbach’s α = .63), and Social Justice (4 items; e.g. wealthy individuals are almost never convicted of their crimes; Cronbach’s α = .66). Higher values indicate that participants have more positive attitudes, while lower values indicate more scepticism towards the justice system. System Confidence and Conviction Proneness were highly correlated with each other r = .58, p < .001, while they were not correlated with Social Justice. System Confidence and Conviction Proneness were combined (hereafter referred to as the Pre-Trial Juror Bias) to one composite measure yielding a higher degree of reliability (Cronbach’s α = .78) than either subscale separately.

Finally, participants’ beliefs about memory changes in old age were assessed using the subscale Stereotype of Capable–Incapable² of the Ageism Scale (10 items; e.g. with old age, people are inclined to become forgetful; Cronbach’s α = .85; Braithwaite, Lynd-Stevenson, & Pigram, 1993). For each item, participants indicated their agreement on a 7-point rating scale from 1 = strongly disagree to 7 = strongly agree. Higher values indicate that participants have more negative attitudes towards older people and their memory. Item order was randomized within each scale to control for order effects.

Research design

The study was a randomized 2× 2 between-participants design with victim age (25 years vs. 75 years) and crime severity (no injury vs. sprained ankle) as the manipulated variables. The four groups were (a) 25-year-old victim, no injury (n = 51); (b) 25-year-old victim, sprained ankle (n = 54); (c) 75-year-old victim, no injury (n = 54); and (d) 75-year-old victim, sprained ankle (n = 47).

Procedure

The study was approved by the Local Ethics Committee at the Justus-Liebig-University of Giessen (approval number: 2017-0009). Participants were recruited via Prolific, an online crowdsourcing platform (www.prolific.ac), and participated online.

Participants were randomly assigned to one of the experimental groups. They read the trial scenario and answered a series of questions about the trial and the victim. Further, they indicated their agreement to attitudinal items and provided demographic information. Participants could not re-read the trial scenario or re-do questions. They were then debriefed and reimbursed for their time (£2 = approximately US $2.60). Participation was about 10–15 min (M = 11.55, SD = 4.70).

Pilot study

A pilot study with 40 participants revealed that the manipulation of victim age was successful. When the victim was young, participants were more likely to convict the defendant (57.1% guilty) and perceive that he committed the crime (M = 74.3%, SD = 23.64) than when the victim was older: 19.2% guilty, χ²(1, 40) = 5.96, p = .015, φ = −.386, odds ratio (OR) = 5.6; probability of guilt: M = 51.2%, SD = 24.39, F(1, 40) = 7.05, p = .012, g_u = 0.94. Further, participants were more likely to agree that a young victim made a correct identification (M = 75.7%, SD = 21.74) than her older counterpart (M = 59.6%, SD = 23.06), F(1, 40) = 4.43, p = .042, g_u = 0.70. Crime severity was not associated with any of the dependent measures (p > .05).

Perceived victim credibility consisted of 18 items. Analyses revealed the presence of multicollinearity for several items (r > .80) and a very low corrected item–total correlation (CITC) for one item (CITC = .08). Redundant items were then removed or rephrased, and the item with low CITC was removed from the questionnaire. The final questionnaire consisted of 14 items.

Results

In the following, we first report results on verdict. Next, we present the results of 2× 2 between-participants analyses of variance (ANOVAs) of victim age and crime severity on perceived victim credibility and on mock jurors’ perceptions of the case strength for the prosecution and for the defence as dependent variables. In addition, we report a repeated measures t test to assess whether the prosecution or the defence case was stronger. Next, we report two hierarchical multiple regression analyses to assess the predictive value of mock jurors’ demographics, general attitudes, perceived victim credibility and case strength for the prosecution and the defence on (a) the perceived likelihood that the identification was correct, and (b) the probability of guilt.

The reported effect sizes are the corrected standardized mean difference Hedges g_u, Pearson correlations r and point biserial correlations r_pb (Lipsey & Wilson, 2001). In addition, semi-partial correlations (sr) are reported with multiple regression analyses to assess the unique contribution of each variable to the regression model (Tabachnick & Fidell, 2013). Effect sizes are interpreted as small for Pearson correlations and point biserial correlations of .10 and standardized mean differences of .20; moderate for Pearson correlations of .30 and point biserial correlations of .24 and a standardized mean difference of .50; and large for Pearson correlations of .50 and point biserial correlations of .37 and a standardized mean difference of .80 (Cohen, 1988).

Verdict

The vast majority of mock jurors acquitted the defendant (81.5% guilty, 18.5% not guilty). The conviction rate was irrespective of age of the victim, χ²(1, 205) = 1.79, p = .181, φ = −.093, OR = 1.63, 95% confidence interval, CI [0.79, 3.34], or crime severity, χ²(1, 205) = 1.55, p = .213, φ = .087, OR = 1.57, 95% CI [0.77, 3.20]. Table 1 shows the proportions of guilty verdicts by victim age and crime severity.

Table 1.

Victim credibility, case strength, probability of guilt and verdict as a function of victim age and crime severity.

	Young victim M (SD)	Older victim M (SD)	Effect size gu [95% CI]
Victim credibility	4.63 (1.02)	4.47 (1.00)	0.16 [−0.12, 0.43]
No injury	4.49 (1.04)	4.62 (0.91)	−0.13 [−0.51, 0.25]
Injury	4.76 (0.98)	4.29 (1.07)	0.45[0.06,0.85]
Victim lenient response bias	3.99 (1.45)	4.15 (1.29)	−0.11 [−0.38, 0.16]
No injury	4.20 (1.51)	4.30 (1.29)	−0.08 [−0.46, 0.31]
Injury	3.80 (1.37)	3.96 (1.28)	−0.13 [−0.51, 0.27]
Prosecution case strength	3.12 (1.70)	2.80 (1.63)	0.19 [−0.08, 0.46]
No injury	2.78 (1.54)	3.13 (1.61)	−0.22 [−0.60, 0.16]
Injury	3.43 (1.79)	2.43 (1.58)	0.59[0.19,0.98]
Defence case strength	4.52 (1.67)	4.95 (1.64)	−0.26 [−0.53, 0.02]
No injury	4.92 (1.44)	5.02 (1.42)	−0.07 [−0.45, 0.31]
Injury	4.13 (1.80)	4.87 (1.87)	−0.40 [−0.79, −0.01]
Likelihood of a correct ID	60.39 (20.19)	56.44 (19.58)	0.20 [−0.08, 0.47]
No injury	57.65 (21.22)	58.15 (19.04)	−0.03 [−0.41, 0.36]
Injury	63.02 (18.97)	57.47 (20.20)	0.28 [−0.11, 0.67]
Probability of guilt	55.39 (21.49)	51.68 (20.15)	0.18 [−0.10, 0.45]
No injury	51.76 (22.42)	51.48 (21.14)	0.01 [−0.37, 0.39]
Injury	58.87 (20.16)	51.91 (19.18)	0.35 [−0.04, 0.74]
Guilty verdict (%)	22.1	14.9	OR = 1.63 [0.79, 3.34]
No injury (%)	17.6	13.0	OR = 1.44 [0.49, 4.20]
Injury (%)	26.4	17.0	OR = 1.75 [0.66, 4.64]

Note: Bold values indicate a significant age effect. CI = confidence interval; OR = odds ratio.

Correlational analyses revealed that with increased Pre-Trial Juror Bias mock jurors were more likely to convict the defendant, r_pb = .18 (see Appendix B). Further, verdict was positively correlated with perceived Victim Credibility, r_pb = .31, and negatively correlated with perceived Victim Lenient Response Bias, r_pb = −.23. Finally, verdict was positively correlated with the perceived likelihood that the identification was correct, r_pb = .46, the probability of guilt, r_pb = .50 and the prosecution case strength, r_pb = .52. The cell sizes were too small for guilty verdicts (n < 10) to conduct further inferential statistical analyses.

Perceived victim credibility

Perceived victim credibility was measured using two factors, Victim Credibility (M = 4.55, SD = 1.01) and Victim Lenient Response Bias (M = 4.07, SD = 1.37). Table 1 shows the means and standard deviations of credibility measures, strength of the prosecution and the defence case and the perceived probability of guilt by victim age and injury presence.

Two-way ANOVAs showed no main effects of victim age and crime severity on perceived Victim Credibility, F(1, 201) = 1.45, p = .231, g_u= 0.20, 95% CI [−0.08, 0.47], and F(1, 201) = 0.05, p = .832, g_u= −0.05, 95% CI [−0.33, 0.22], respectively, nor on perceived Victim Lenient Response Bias, F(1, 201) = 0.51, p = .477, g_u= 0.18, 95% CI [−0.10, 0.45], and F(1, 201) = 3.70, p = .056, g_u= 0.19, 95% CI [−0.46, 0.08], respectively. The interaction between Victim Age and Crime Severity was significant for perceived Victim Credibility, F(1, 201) = 4.52, p = .035, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .022, but not for perceived Victim Lenient Response Bias, F(1, 201) = 0.02, p = .877, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .000. Post hoc analyses using a Tukey test showed no age effect in absence of injury, g_u= −0.13, 95% CI [−0.51, 0.25]. By contrast, when the victim reported she was injured during the robbery, mock jurors believed the 25-year-old victim more than the 75-year-old victim, g_u= 0.45, 95% CI [0.06, 0.85]. Figure 1 shows the interaction effect between victim age and crime severity on perceived victim credibility.

Figure 1.

Perceived witness credibility as a function of victim age and crime severity. Error bars are 95% confidence intervals.

Prosecution versus defence case strength

Overall, mock jurors perceived the defence case stronger, M = 4.73, SD = 1.67, than the prosecution case, M = 2.96, SD = 1.67, t(204) = 9.19, p < .001. Further, the defence and the prosecution case strength were negatively correlated with each other, r = −.37, p < .001. In other words, mock jurors were more likely to agree that the defence created reasonable doubt about the accused’s guilt than they were to agree that the prosecution case was strong.

A two-way ANOVA showed no main effects of victim age, F(1, 201) = 2.09, p = .149, g_u= 0.19, 95% CI [−0.08, 0.46], and crime severity, F(1, 201) = 0.01, p = .906, g_u= 0.00, 95% CI [−0.27, 0.27], on prosecution case strength. However, as predicted by Hypothesis 3, the interaction between Victim Age and Crime Severity was significant, F(1, 201) = 8.72, p = .004, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .042, as shown in Figure 2(a). Post hoc analyses using a Tukey test showed that mock jurors perceived the prosecution case stronger when the injured victim was 25 years old than when the victim was 75 years old, g_u= 0.58, 95% CI [0.19, 0.98]. No age effect was found in the absence of injury, g_u= −0.22, 95% CI [−0.60, 0.16]. Further, for the young victim mock jurors perceived the prosecution case stronger in the presence than in the absence of an injury, g_u= 0.39, 95% CI [0.00, 0.77]. The effect was reversed for the older victim, g_u= −0.43, 95% CI [−0.83, −0.04].

Figure 2.

Perceived case strength for the prosecution (Panel A) and the defence (Panel B). Error bars are 95% confidence intervals.

A two-way ANOVA for defence case strength showed a main effect of crime severity, F(1, 201) = 4.16, p = .043, g_u= 0.30, 95% CI [0.02, 0.57]. Mock jurors perceived the defence case stronger when the victim was not injured during the robbery. The effect of victim age, F(1, 201) = 3.33, p = .069, g_u= −0.26, 95% CI [−0.53, 0.02], and the interaction between Victim Age and Crime Severity were not significant, F(1, 201) = 1.97, p = .162, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .010. Mock jurors’ perceptions of the defence case strength are shown in Figure 2(b).

Combined predictive value of attitudes, credibility and case strength

In response to Hypotheses 5a and 5b, hierarchical multiple regression analyses were conducted to assess the combined predictive value of mock jurors’ demographics and attitudes (Step 1), perceived victim credibility (Step 2) and perceptions of case strength (Step 3) on (a) the perceived likelihood of the identification to be correct, and (b) probability of guilt. Victim age and crime severity were excluded from the regression models, as they were not associated with the dependent measures. (ANOVAs revealed that victim age, crime severity and their interaction were not associated with the perceived likelihood of a correct identification, p > .10, or with probability of guilt, p > .10.) Correlations between mock jurors’ pre-trial attitudes, case evaluations and decisions are in Appendix B.

The subscale Stereotype of Capable–Incapable was negatively correlated with the perceived Victim Credibility, r(203) = −.27, p < .001, and positively correlated with the perceived Victim Lenient Response Bias, r(203) = .15, p = .014. Further, the subscale Stereotype of Capable–Incapable was negatively correlated with the likelihood that the identification was considered correct, r(203) = −.24, p < .001, the probability of guilt, r(203) = −.21, p = .001, and case strength for the prosecution, r(203) = −.24, p < .001. That is, an increased degree of beliefs that memory changes in older age was associated with decisions in favour of the defence.

Participants’ age was not correlated with Pre-Trial Juror Bias, r(202) = .07, p = .322, while it was negatively correlated with the subscale Stereotype of Capable–Incapable, r(202) = −.37, p < .001, indicating that ageism stereotypes decreased with increasing age.

Likelihood of a correct identification

Table 2 shows hierarchical regression results for the likelihood that the identification was considered correct. The model and model changes were significant at each step, indicating a significant contribution to the model when variables were added to the model.

Table 2. Hierarchical multiple regression for the likelihood of a correct identification.

Predictors	B	β	t	p	sr	R2
Step 1						.160
Participant gender	6.05	.15	2.15	.033	.14
Participant age	0.07	.04	0.55	.580	.04
Pre-trial juror bias	5.03	.25	3.69	<.001	.24
Beliefs about memory changes in old age	−4.71	−.21	−2.86	.005	−.19
Step 2						.382
Participant gender	4.07	.10	1.67	.097	.09
Participant age	−0.08	−.05	−0.73	.469	−.04
Pre-trial juror bias	3.82	.19	3.23	.001	.18
Beliefs about memory changes in old age	−2.38	−.10	−1.64	.102	−.09
Victim credibility	8.09	.41	6.51	<.001	.37
Victim lenient response bias	−3.07	−.21	−3.59	<.001	−.20
Step 3						.499
Participant gender	2.40	.06	1.08	.282	.06
Participant age	−0.02	−.01	−0.20	.846	−.01
Pre-trial juror bias	1.83	.09	1.64	.102	.08
Beliefs about memory changes in old age	−1.32	−.06	−1.00	.320	−.05
Victim credibility	4.02	.20	3.13	.002	.16
Victim lenient response bias	−2.21	−.15	−2.79	.006	−.14
Prosecution case strength	4.83	.40	5.98	<.001	.30
Defence case strength	−0.99	−.08	−1.49	.139	−.08
Regression constant	35.66		3.819	.002

Note: Model at Step 1: F(4, 199) = 9.45, p < .001; Model at Step 2: F(6, 197) = 20.33, p < .001; Model at Step 3: F(8, 195) = 24.25, p < .001. sr = semi-partial correlation.

At Step 1, participants’ gender but not age was associated with the likelihood of a correct identification. Particularly, women were more likely to perceive that the identification was correct than were men, β = .15, p = .033, sr = .14. Further, the greater Pre-Trial Juror Bias, β = .25, p < .001, sr = .24, and the lower the Stereotypes of Capable–Incapable, β = −.21, p < .001, sr = −.19, the higher was the perceived likelihood that the identification was correct. The model explained 16.0% of the variance, F(4, 199) = 9.45, p < .001.

When victim credibility was added to the model at Step 2, Pre-Trial Juror Bias was the only predictor from Step 1 to remain significant, β = .19, p = .001, sr = .18. Victim Credibility had the highest predictive value in the model, β = .41, p < .001, sr = .37. In other words, the higher the perceived Victim Credibility, the more likely were mock jurors to assume that the victim made a correct identification. Further, the higher the perceived Victim Lenient Response Bias, the less likely they were to agree that the identification was correct, β = −.21, p < .001, sr = −.20. The model at Step 2 was significant, F(6, 197) = 20.33, p < .001, and explained 38.2% of the variance.

The full model was also statistically significant, F(6, 198) = 32.42, p < .001, and explained 49.9% of the variance. Mock jurors’ perception of the prosecution case strength had the largest predictive value regarding the likelihood that the identification was correct, β = .40, p < .001, sr = .30, while perceived defence strength was not associated with the perceived likelihood of a correct identification, β = −08, p = .139, sr = −.08. Perceived Victim Credibility and the perceived Victim Lenient Response Bias remained significant at Step 3, as shown in Table 3.

Table 3.

Hierarchical multiple regression for probability of guilt.

Predictors	B	β	t	p	sr	R2
Step 1						.130
Participant gender	7.21	.17	2.41	.017	.16
Participant age	0.13	.07	0.97	.333	.06
Pre-trial juror bias	3.94	.19	2.72	.007	.18
Beliefs about memory changes in old age	−3.74	−.16	−2.14	.034	−.14
Step 2						.400
Participant gender	4.94	.12	1.97	.050	.11
Participant age	−0.03	−.02	−0.24	.808	−.01
Pre-trial juror bias	2.62	.12	2.15	.033	.12
Beliefs about memory changes in old age	−1.04	−.04	−0.07	.486	−.04
Victim credibility	8.64	.42	6.57	<.001	.37
Victim lenient response bias	−4.13	−.27	−4.70	<.001	−.26
Step 3						.538
Participant gender	3.05	.07	1.37	.173	.07
Participant age	0.04	.02	0.40	.689	.02
Pre-trial juror bias	0.35	.02	0.32	.753	.02
Beliefs about memory changes in old age	0.14	.01	0.10	.919	.01
Victim credibility	4.05	.20	3.15	.002	.15
Victim lenient response bias	−3.12	−.21	−2.94	<.001	−.19
Prosecution case strength	5.36	.43	6.63	<.001	.32
Defence case strength	−1.35	−.11	−2.02	.045	−.10
Regression constant	30.37		2.71	.007

Note: Model at Step 1: F(4, 199) = 7.41, p < .001; Model at Step 2: F(6, 197) = 21.88, p < .001; Model at Step 3: F(8, 195) = 28.39, p < .001.

Probability of guilt

The regression model change was significant at each step, indicating that each pair of variables added a significant contribution to the respective model. The hierarchical regression results for probability of guilt are presented in Table 3.

At Step 1, participant gender, Pre-Trial Juror Bias, and Stereotypes of Capable–Incapable were associated with the probability of guilt. Female participants and those with stronger Pre-Trial Juror Biases assigned higher probability of guilt, β = .17, p = .017, sr = .16, and β = .19, p = .007, sr = .18. Participants who were more likely to agree that memory changes with increasing age assigned lower probability of guilt, β = −.16, p = .034, sr = −.14. The model explained 13.0% of the variance in probability of guilt, F(4, 199) = 7.41, p < .001.

At Step 2, perceived Victim Credibility and Victim Lenient Response Bias were both associated with the probability of guilt, F(6, 197) = 21.88, p < .001, explaining 40.0% of the variance. The more mock jurors believed the victim, the higher was the probability of guilt, β = .42, p < .001, sr = .37. Further, the more mock jurors believed that the victim had a more Lenient Response Bias, the lower was the perceived probability of guilt, β = −.27, p < .001, sr = −.26.

The full regression model with all eight predictors was significant, F(8, 195) = 28.39, p < .001, and explained 53.8% of the variance. The stronger the perceived prosecution case the higher was the perceived probability of guilt, β = .43, p < .001, sr = .32. Further, the stronger the perception that the defence created reasonable doubt the lower the probability of guilt, β = −.11, p = .045, sr = −.10. Notably, mock jurors’ gender, Pre-Trial Juror Biases and Stereotypes of Capable–Incapable did no longer have any predictive value of probability of guilt in the full regression model, p > .10, as shown in Table 3. Inspection of the semi-partial correlations indicated that mock jurors’ perception of prosecution strength had the greatest predictive value of probability of guilt, sr = .32, followed by perceived Victim Lenient Response Bias, sr = −.19, and Victim Credibility, sr = .15.

Discussion

Age-related decline in the sensory system (Schieber, 2006), cognitive functioning and memory (Coleman & O’Hanlon, 2008; Park et al., 2002) have been well documented. Yet studies showed conflicting results about how jurors perceived older people compared to young people in regards to their credibility and reliability as witnesses (e.g. Kwong See et al., 2001; Neal et al., 2012; Ross et al., 1990). The differences may be attributed to characteristics other than victim age alone, such as crime type or juror characteristics, as proposed by the conceptual model of juror decision-making that linked defendant, case and juror characteristics to their verdict (Devine & Caughlin, 2014). This model implies that both legal and extra-legal factors drive the decision process (Sporer & Goodman-Delahunty, 2009). The present study aimed to investigate the effect of manipulated victim age and crime severity as legal factors and juror characteristics as extra-legal factors on mock jurors’ decisions. First, we examined whether mock jurors were aware of age-related memory changes and whether they believed more the testimony of the young than the older victim. Second, we investigated whether crime severity influenced mock jurors’ decisions. Third, we investigated the extent to which mock juror characteristics were associated with their decisions.

Unlike many other studies that only measured verdict (Lundrigan, Dhami, & Mueller-Johnson, 2018), the present study investigated mock jurors’ decision process from different perspectives. Perceptions of the victim were assessed using questions about credibility in the sense of deception and in the sense of error (Sporer, 2008). Analyses showed that these two constructs were not independent of each other, as reflected in the subscale Victim Credibility, which included items about both honesty and correct memory of the victim. The subscale Victim Lenient Response Bias assessed the perceived victim’s proneness to choose anyone from the line-up, which was moderately correlated with the subscale Victim Credibility. Finally, to gain insight into different aspects of the decision process beyond the binary variable verdict, the present study included jurors’ perceptions of the likelihood that the victim correctly identified the offender and of the probability of guilt. Both measures showed a strong correlation with the binary verdict.

Mock jurors’ awareness of memory changes in old age

Mock jurors’ awareness of memory changes in old age was tested via their beliefs about cognitive capacity in older age and by manipulating victim age. Analyses showed that mock jurors expressed a moderate degree of agreement to memory changes in older age (i.e. mean values above the midpoint of the scale), when they were asked about their beliefs about cognitive capacity in old age. These findings suggest that participants tended to associate old age with some losses in cognitive capacity, confirming previous research findings (Braithwaite et al., 1993). This is discussed in more detail below.

The manipulation of victim age was successful only to some extent. Compared to a young victim, mock jurors neither perceived the older victim as more credible (cf. Allison et al., 2006; Ross et al., 1990) nor doubted their testimony more than that of the young victim (cf. Neal et al., 2012). Further, no age effect was found for the perceived Victim Lenient Response Bias, indicating that mock jurors were unaware that older people may have a more liberal response criterion than young people (see the meta-analysis by Martschuk & Sporer, 2018).

Notably, the overall victim credibility was relatively low – that is, around the mean of the scale. One of the reasons may be the nature of the prosecution case relying mainly on the testimony of the victim with no other eyewitnesses, which would support or discredit the victim’s statement. However, when crime severity was taken into account, an age effect was found for the severe crime condition such that mock jurors believed more the young than the older victim. This pattern was also reflected in the perceptions of the prosecution versus the defence case strength. When Figures 2(a) and 2(b) are juxtaposed it becomes clear that the prosecution case was much weaker than the defence case. In the absence of injury, the age of the victim did not affect case strength. However, compared to the young victim mock jurors perceived the prosecution case weaker and the defence case stronger in the severe case involving an older victim. Thus, although both victims delivered the same testimony, mock jurors seemed to be not as persuaded by the older prosecution witness as they were by the young prosecution witness. Still, these perceptions were not reflected in the perceived likelihood that the identification was correct, probability of guilt, or verdict.

Further, no age effect was found for perceptions of the strength of the prosecution, or the defence case, failing to support Hypotheses 1b and 1c. These unexpected findings may be explained by the mode of trial presentation, namely identical trial summaries without any visual cues of ageing or differences in speech styles consistent with ageing (Pittman et al., 2014). Some support for this explanation was provided in a series of experiments by Brimacombe et al. (1997). They showed that older witnesses provided weaker testimonies than young witnesses (i.e. less accurate, more frequent use of negative qualifiers, higher expressed hesitance). As a consequence, mock jurors perceived testimonies of older witnesses as less credible, irrespective of whether they knew the age of the witness or not. Thus, speech style influenced perceived credibility more than the actual witness age.

Influence of crime severity on mock jurors’ decisions

Crime severity in the form of victim injury was the second case characteristic manipulated in the present study. Although injury should not be considered when evaluating the strength of the evidence, if crime is committed in a violent manner, mock jurors interpret the evidence in a biased way by putting more weight to the emotionally laden information (Bright & Goodman-Delahunty, 2004). Thus, we hypothesized that mock jurors would believe the victim more (Hypothesis 2a) and perceive the prosecution case stronger when the crime was more severe (Hypothesis 2 b). The present findings confirmed this theoretical notion for the young victim only: mock jurors perceived the prosecution case of a more severe crime stronger than that of a less severe crime. However, the effect was reversed for the older victim such that mock jurors perceived the prosecution case weaker when the crime was severe than when it was less severe, confirming Hypothesis 3. These findings indicate that mock jurors might have seen the injury as a further proof that the crime happened for the young victim, while they had doubts about the case when an older victim was injured.

The present study further showed that crime severity influenced perceptions of the defence case: Mock jurors were more likely to agree that the defence created reasonable doubt of the defendant’s guilt in the less severe case, confirming Hypothesis 2c. Put another way, victim injury weakened the defence case – a finding that was previously shown in cases of sexual abuse and child sexual abuse (e.g. Blackwell & Seymour, 2014; Fitzgerald, 2006; Gray-Eurom, Seaberg, & Wears, 2002).

Finally, the case of a more severe crime may evoke an emotional reaction, which in turn would result in more guilty verdicts. A number of studies demonstrated that crime severity was associated with more guilty verdicts (e.g. Bright & Goodman-Delahunty, 2004; Goodman-Delahunty et al., 2016; Pearson et al., 2018) and more punitive sentencing decisions (Nadler & Rose, 2003; Sporer & Goodman-Delahunty, 2009). The present results revealed very low conviction rates (despite a perceived probability of guilt above 50%), independent of crime severity. These findings were in line with a study that used a similar trial scenario, namely a robbery trial with varying degrees of violence against the victim (Freedman et al., 1994). Taken together, more research is needed to further investigate whether and how crime severity influences perceptions of a victim, particularly of an older age, and ultimately verdict. An intriguing, related question may also be to what extent the testimony of older versus younger witnesses may be perceived as consistent.

Juror attitudes and their decisions

The extra-legal factors investigated in the present study were juror characteristics (i.e. age, gender), their trust in the legal system and perceptions of older people. Analyses revealed that participant gender but not age were associated with the perceived likelihood of a correct identification and probability of guilt. These findings were in line with previous research showing that female mock jurors were more likely to render guilt judgments than men (see the meta-analysis by Devine & Caughlin, 2014).

Further, the present findings showed that, without any case-specific information, mock jurors’ evaluation of the evidence (i.e. the likelihood that the identification is correct) and their ratings of the probability of guilt were associated with their Pre-Trial Juror Bias (i.e. System Confidence and Conviction Proneness) and with their beliefs about ageing (Stereotype of Capable–Incapable). Specifically, the more confident mock jurors were in the legal system and the more prone they were to convict the more likely they were to decide in favour of the prosecution. These findings support previous research findings that jurors’ trust in the legal system was associated with guilt judgments (Devine & Caughlin, 2014; Lecci & Myers, 2009).

Finally, mock jurors holding stronger beliefs about memory changes in old age were more sceptical of the victim’s testimony regardless of victim age, identification accuracy and probability of guilt. In all, they were more sceptical of the prosecution case and less likely to render guilt judgments. These findings also indicated that mock jurors holding stronger beliefs about memory changes in old age were in general more sceptical of memory performance of the victim. A large body of research showed that juror characteristics are only moderately correlated with their decisions (see review by Devine, 2012) and explained only about 10–15% of the variance in their decision, while the presented evidence had a stronger impact on juror decisions (Saks & Spellman, 2016). For instance, a study that tested juror decisions across multiple types of trials showed that perceived case strength was associated with higher probability of guilt (Pearson et al., 2018). The present results showed the same pattern: Mock juror characteristics explained together 13% of the variance in probability of guilt and 16% of the variance in the likelihood of a correct identification. Once case-specific facts were considered, influence of juror characteristics and general views became less important. Instead, mock jurors’ perceptions of the victim and their evaluation of the prosecution case were much more predictive of the perceived likelihood that the identification was correct and the probability of guilt than juror characteristics, as assumed in Hypotheses 5a and 5b. These findings lend further support to Sherif and Sherif’s (1969) Gestalt theoretical postulate that internal factors affect judgments only when the stimulus configuration is rather ambiguous (see also Sporer & Goodman-Delahunty, 2009).

Unfortunately, the cell sizes for guilty verdicts were too small to conduct logistic regression analyses to test Hypothesis 5c that after controlling for juror characteristics perceptions of the prosecution and the defence case would be associated with verdict. Nevertheless, correlational analyses confirmed that perceptions of the victim and of the prosecution case strength were associated with verdict. A reason for the low conviction rate may be that neither witnesses other than the victim nor forensic evidence were presented supporting the prosecution case – factors that influence perceptions of case strength and ultimately guilty judgments, as shown in a study by Pearson et al. (2018). Thus, mock jurors applied the principle of in dubio pro reo in this case and acquitted the defendant.

Limitations

The following methodological issues must be considered when drawing conclusions from the findings of the present study. Our manipulation of crime severity was rather weak. A stronger manipulation of severity than victim injury alone may lead to stronger effects. The mode of trial presentation was a written (simulated) trial summary as opposed to a live evidence presentation at court. Mock jurors were aware that their decision did not have any consequences for the defendant or the community. This may have limited the extent to which mock juror decisions resemble those in real cases. Further, mock jurors made their decision individually and not in groups, which limits the ecological validity of decisions due to the lack of mutual influence of other jury members and discussions that arise during jury deliberations. Notably, a previous study showed similarities in the decision processes between individual and group decisions in the control group (i.e. in absence of educative information provided by an expert; Goodman-Delahunty & Martschuk, 2017).

Finally, participants were jury-eligible citizens and not actual jurors. Compared to empanelled jurors, participants in the present study were more educated (i.e. a higher proportion of participants held a university degree), and the proportion of female participants was higher (cf. Goodman-Delahunty et al., 2007; Trimboli, 2008). However, explicit attempts were made to include potential jurors; for instance, only participants from countries in which jury trials are held were eligible. Moreover, more than 20% of participants reported that they had been previously called to jury duty, and one third of them reported they had served in a jury trial.

Implications for research and practice

The fact that victim age did not influence mock jurors’ decisions in the present study indicated that mock jurors were unaware of memory changes in old age. Notably, the findings that mock jurors were sceptical of memory in older age, and were also more sceptical of victim testimony in general, suggested that mock jurors applied their knowledge about memory (or lack thereof) when making decisions. Following this, in cases in which memory of an older witness might be an issue, mock jurors would benefit from educative information about a memory decline in older age. Similar proposals involving witness and victim memory have already been made (e.g. Conway, 2012; Doyle, 2012), and professional psychological associations (e.g. The British Psychological Society Research Board, 2010; Wells et al., 2020) prepared empirical guidelines to inform legal decision-makers about different aspects of memory. In jury trials, educative information can be provided in the form of judicial instructions or expert evidence. However, more research is needed to assess which mode of presentation is more effective to inform decision makers about memory.

Further research is needed to test whether the present findings hold following a simulated video trial and group deliberations. Analyses of jury deliberations will assist in assessing the reasoning process and the extent to which, for instance, victim age is discussed as part of evidence.

Conclusion

In summation, our findings demonstrated the importance of including moderators, such as age, when evaluating juror decisions. The theoretical notion that mock jurors evaluated the case more in favour of the prosecution when the crime was more severe was only confirmed for the young victim, while the effect was reversed for the older victim. Further, the data showed that mock jurors were only moderately aware of age-related changes in memory. While mock juror characteristics were associated with their decisions to some extent, case-specific facts were much more predictive of the evaluation of evidence and judgments of the defendant’s guilt.

Appendix A.

Component structure of the perceived victim credibility scale

All credibility items elicited the full range of possible responses on the 7-point rating scale, indicating a wide range of perceptions of victim credibility. Table A1 presents item means, standard deviations and inter-item correlations.

Table A1.

Items used to measure perceived victim credibility and victim lenient response bias items, means, standard deviations and inter-item correlations.

The victim	M	SD	1	2	3	4	5	6	7	8	9	10	11	12	13
(1) accurately remembered the event.	4.07	1.42
(2) was cautious when she selected somebody from the line-up.	3.62	1.68	.38
(3) wanted to choose someone from the line-up to help the police.	4.41	1.70	−.02	−.10
(4) wanted to have someone to blame.	4.40	1.72	−.20	−.07	.50
(5) would have identified anyone from the line-up for justice to be served.	3.40	1.67	−.25	−.14	.38	.55
(6) gave a consistent account of the event.	3.75	1.56	.64	.35	−.03	−.18	−.27
(7) was plausible.	5.17	1.37	.53	.18	.06	−.08	−.27	.55
(8) was confident when she identified the robber at the line-up procedure.	5.30	1.50	.35	.12	−.10	−.25	−.27	.35	.44
(9) was confident when she testified at trial.	4.54	1.39	.42	.15	−.15	−.21	−.21	.40	.42	.52
(10) was truthful.	5.31	1.36	.52	.26	.02	−.15	−.35	.44	.66	.35	.37
(11) was credible.	4.55	1.39	.61	.37	−.09	−.22	−.35	.55	.62	.43	.45	.62
(12) was unbiased.	3.75	1.60	.41	.21	−.03	−.15	−.17	.41	.34	.24	.25	.32	.36
(13) was honest.	5.19	1.36	.49	.26	.00	−.13	−.33	.52	.65	.43	.39	.78	.61	.34
(14) was trustworthy.	4.78	1.33	.60	.35	−.01	−.11	−.30	.55	.57	.36	.40	.67	.68	.41	.70

Note: N = 205. Higher mean scores denote greater agreement to the items. Possible scores: 1–7. Correlations r > .11 are significant (p < .05).

A principal components analysis (PCA) was conducted on 14 items that measured perceived witness credibility. Initial analyses revealed suitability of data for PCA: (a) inspection of the correlation matrix revealed that all items were between r(205) = .33 and r(205) = .78 (see Table A1), and, therefore, between the recommended values of r = .3 and r = .8 (Field, 2013); (b) the Kaiser–Myer–Olkin measure (KMO) was KMO = .89, and all KMO values for individual items were greater than KMO = .68 and above the recommended value of .6 (Kaiser, 1974; Kaiser & Rice, 1974); (c) the Bartlett’s Test of Sphericity reached statistical significance, χ²(91) = 1347.39, p < .001.

Multiple methods were used to assess the number of components. Inspection of the eigenvalues revealed three components that exceeded 1, and explained 41.8%, 13.0% and 7.6% of the variance, respectively. The scree plot revealed one point of reflection after the second factor, and a second point of reflection after the third factor. The Monte Carlo PCA for Parallel Analysis (Watkins, 2000) with randomly generated data matrix using 1000 replications supported the two-component structure.

Solutions for two and three components were examined using varimax (orthogonal) rotation of the component-loading matrix. The component loadings in Table A2 measured the relationship between the components and the items indicating the overlapping variance among components, which are interpreted as follows: λ ≥ .71, 50% overlapping variance; λ ≥ .63, 40% overlapping variance; λ ≥ .55, 30% overlapping variance; λ ≥ .45, 20% overlapping variance; λ ≥ .32, 10% overlapping variance (Comrey & Lee, 1992; as cited in Tabachnick & Fidell, 2013). This was the case for each item for both models.

Table A2.

Summary of principal component analysis of items for perceived victim credibility and perceived victim lenient response bias.

	2 Components		3 Components
The victim	1	2	1	2	3
was trustworthy.	0.83	−0.04	0.59	0.59	−0.01
was honest.	0.82	−0.05	0.74	0.40	−0.01
was credible.	0.81	−0.17	0.61	0.55	−0.14
was truthful.	0.81	−0.05	0.70	0.42	−0.01
was plausible.	0.80	0.01	0.77	0.33	0.06
accurately remembered the event.	0.76	−0.10	0.47	0.64	−0.08
gave a consistent account of the event.	0.73	−0.11	0.46	0.61	−0.09
was confident when she testified at trial.	0.56	−0.25	0.67	0.08	−0.21
was confident when she identified the robber at the line-up procedure.	0.53	−0.28	0.72	−0.03	−0.23
was unbiased.	0.52	−0.08	0.27	0.50	−0.07
was cautious when she selected somebody from the line-up.	0.42	−0.07	−0.11	0.81	−0.10
wanted to have someone to blame.	−0.11	0.85	−0.13	−0.06	0.84
wanted to choose someone from the line-up to help the police.	0.08	0.81	0.09	−0.04	0.81
would have identified anyone from the line-up for justice to be served.	−0.31	0.71	−0.30	−0.17	0.70
Eigenvalues	5.85	1.83	5.85	1.83	1.07
% of variance	41.80	13.04	41.80	13.04	7.65
Cronbach's α	.89	.73	.86	.78	.73

Note: N = 205. Loadings on the respective component appear in bold.

Inspection of the three-component solution revealed a complex structure. Six of 14 items showed at least 20% overlapping variance on two components. Additional analyses showed that Components 1 and 2 were highly correlated, r(205) = .71, p < .001. The correlation between Components 1 and 3 was r(205) = −.27, p < .001, and between Component 2s and 3 r(205) = −.23, p = .001.

By contrast, the two-component matrix revealed a simple structure, with both components containing a number of strong loadings and items loading substantially on only one of the components. The inter-correlation between Component 1 and Component 2 was r(205) = −.27, p < .05.

The two-component solution was preferred, because Components 1 and 2 in the three-component solution were statistically and semantically related, indicating that they measured the same construct. Component 1 contained 11 items that measured Victim Credibility. Component 2 contained three items that measured Victim Lenient Response Bias. The reliabilities were Cronbach’s α = .89, and Cronbach’s α = .73, respectively, and above the acceptable value of .7.

Appendix B.

Appendix B Correlations between independent and dependent variables.

	M	SD	Victim credibility	Victim lenient response bias	Prosecution case strength	Defence case strength	Likelihood of correct identification	Probability of guilt	Verdict
Victim age: 25 vs. 75 years	—	—	−.08	.06	−.09	.13	−.10	−.09	−.09
Crime severity: not injured vs. injured	—	—	−.01	−.14	−.00	−.15	.03	.10	.09
Gender: male (1) vs. female (2)	—	—	.21	−.10	.26	−.12	.26	.26	.12
Age	31.34	11.01	.32	−.04	.15	−.06	.17	.18	.06
Pre-trial juror bias	3.66	0.98	.15	−.02	.29	−.17	.26	.20	.17
Social justice	4.41	0.97	.01	.02	.03	.06	.06	.07	.09
Beliefs about memory changes in old age	4.58	0.87	−.27	.15	−.24	.06	−.24	−.21	−.06
M	—	—	4.55	4.07	4.73	2.96	58.44	53.56	—
SD	—	—	1.01	1.37	1.67	1.67	19.94	20.87	—

Note: Point biserial correlations r_pb are reported for victim age, crime severity, gender and verdict. All other correlations are Pearson's r. Absolute values of any r ≥ .12 are significant at p < .05, of r ≥ .16 are significant at p < .01, and of r ≥ .23 are significant at p < .001.

Ethical standards

Declaration of conflicts of interest

Natalie Martschuk has declared no conflicts of interest

Siegfried L. Sporer has declared no conflicts of interest

Ethical approval

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

Informed consent

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