Witnesses and victims often choose not to report crimes to police. Previous research suggests that decisions to report and participate in the criminal justice process are the product of a complex, multi-determined confluence of cognitive, social, and cultural factors. The focus of this dissertation was to examine the relationship between attitudes regarding the legitimacy of police, expressed willingness to report crimes, and actual crime reporting. Study 1, conducted with 309 undergraduate students, was designed to assess whether the relationship between perceptions of legitimacy and expressed willingness to report varied depending on the specific crime conditions. The results showed that when presented with broad, hypothetical questions that do not specify the crime conditions (i.e., the typical method used in willingness studies), individuals imagine their own crime scenarios. Willingness to report varied depending on the specific conditions imagined and based on explicitly defined features of the crime. The legitimacy-willingness relationship was robust, regardless of whether respondents imagined the crime features or they were explicitly provided with them. Study 2 assessed attitudes about the legitimacy of police and crime reporting behavior among 75 undergraduates who had either been a witness or victim of a serious crime within the previous two years. Attitudes about the legitimacy of the police were strongly associated with respondents' expressed willingness to report crimes to the police, but were not associated with actual crime reporting behavior. Expressed willingness was strongly associated with actual crime reporting. The results underscore the importance of social factors associated with actual reporting. Specifically, victims and witnesses were less likely to report crimes when: (a) they were committed by perpetrators they knew; (b) they perceived the crime as not serious; (c) they did not fear retaliation; (d) they feared being labeled a "snitch;" (e) they did not talk to a parent or co-witness after the crime occurred; and, (f) they did not perceive crime reporting as a moral duty. Implications for process-based theories of control, police practice and procedure, and future research are discussed.

The dominant theory of decision-making in eyewitness identification, based on a distinction between absolute and relative judgments, assumes that relative judgments (identifying the best match relative to the other lineup members) increases identification errors (Wells, 1984). This distinction also assumes that comparisons among lineup members underlying relative judgments increases errors, as evidenced by the sequential lineup advantage (Lindsay & Wells, 1985). Sequential lineups preclude comparisons among lineup members by presenting lineup members separately. Recently, and in contrast to the dominant view, receiver operating characteristic (ROC) studies found a simultaneous accuracy advantage (Mickes, Flowe, & Wixted, 2012) that may be due to underlying comparative processes (Wixted & Mickes, 2014). This dissertation focuses on the effect of comparative strategies on eyewitness identification accuracy. Clark, Erickson, and Breneman (2011) simulated identification response probabilities for different decision rules and foil selection methods using the computational model, WITNESS (Clark, 2003). These simulations showed a disadvantage for relative judgment strategies for lineups with foils matched to the appearance of the suspect. Experiment 1 provides an empirical test of the WITNESS model's predictions. Participants viewed lineups that differed in suspect guilt or innocence and foil selection method. The innocent suspect either appeared in different-foils lineups, the foil selection method used by police, or same-foils lineups, a method used in eyewitness identification research that holds foils constant across lineup type. Additionally, lineup instructions either directed witnesses to use an absolute or relative judgment decision rule. ROC curves indicated a small advantage for relative decision rules for different-foils and same-foils lineups. Contrary to the dominant view, the results suggest that relative judgments (and comparisons among lineup members) may improve accuracy. Experiment 2 investigated comparative processes independently of decision rule. Participants viewed lineups that differed in arrangement. Similar foils surrounded the suspect in the high-similarity neighbor condition, designed to facilitate comparative processes, and dissimilar foils surrounded the suspect in the low-similarity neighbor condition. Results showed an accuracy advantage for high-similarity neighbors based on ROC analyses. Experiment 3 investigated the degree to which comparisons among lineup members underlie the simultaneous accuracy advantage. Witnesses viewed lineups differing in lineup presentation (simultaneous or sequential), and neighboring foil similarity. Results showed a larger accuracy advantage of simultaneous high-similarity neighbor lineups. High-similarity neighbors may facilitate comparative strategies in simultaneous lineups but not sequential lineups, corroborating recent reports of a simultaneous advantage and contrary to the widespread "sequential superiority" view.

Eyewitness testimony is a compelling form of evidence, and mistaken identification is a factor in most wrongful convictions (Gross & Schafer, 2012). It is crucial to understand whether jurors can judge the accuracy of the eyewitness identification from the facts of the testimony, and subsequently factor that information into their verdicts. Some research suggests that jurors are good at this task, while other research disagrees. The current dissertation includes a meta-analysis of this conflicting literature, followed by an experimental test of juror understanding using video, audio-only, written transcript, and written summary testimony. The meta-analysis data suggested that using written testimony resulted in greater juror sensitivity to factors relating to potential eyewitness accuracy and greater application of that knowledge to verdicts, than was seen using videotaped testimony. Based on these data, it was hypothesized that reading the testimony allows jurors greater opportunity to discern these factors, remember them better, and apply them more efficiently to their verdicts, compared to seeing or hearing the witness provide the actual testimony. This prediction was, in part, supported by the experimental data. Whether the juror saw, heard, or read the testimony, they noted the key information relating to eyewitness accuracy. They also retained that information, and understood its relationship to the probable accuracy of the identification. However, across stimulus formats, they did not equally apply that information to their verdicts. This dissertation examines those differences and suggests potential explanations for the discrepancies.

The changed-trace and multiple-trace theories of interference are tested in a set of six experiments. The changed-trace hypothesis attributes interference to a rewriting of an initial memory trace. The multiple-trace hypothesis attributes interference to a competition between separate memory traces. Experiments 1 and 2 replicated the modified recognition test used by Chandler (1989, 1991) and provide support for the changed-trace hypothesis due to the strong evidence of retroactive interference, but lack of evidence for proactive interference. The rest of the experiments modify the basic paradigm by changing the type of stimuli (Experiments 3 and 4 introduce words as stimuli instead of images) and the number of presentations of stimuli (Experiments 5 and 6 increase the number of times the interfering stimuli are shown). These changes resulted in evidence for both proactive and retroactive interference. Proactive interference was found in the experiments that used a modified version of Chandler's methodology, supporting the multiple trace hypothesis. If a memory trace is changed, proactive interference will not occur. The lack of evidence for proactive interference in the original task could be due to a recency effect and a large difference in the strength of the original memory trace as compared to the strength of the memory trace for the interfering information.

Collaborative memory research has focused on the negative effects of group remembering, specifically emphasizing how collaborative memory can be worse than individual memory. Previous research has shown that collaboration can impair memory by limiting group output through retrieval disruption (Basden, Basden, Bryner & Thomas, 1997), by altering one's memory through socially induced forgetting (Coman, Manier, & Hirst, 2009), and through the social transmission of contagious errors (Roediger, Meade, & Bergman, 2000). The current dissertation includes three experiments designed to examine the effect that discussion has on subsequent individual memory reports. Experiment 1 systematically examines the transmission and acceptance of correct versus incorrect information within the specific context of a social contagion memory paradigm developed by Roediger et al. (2000). Experiment 2 examines the differential effects of social contagion when the overall amount of recalled information varies across cued and non-cued recall tasks (Tulving & Pearlstone, 1966). Experiment 3 examines how reconstructive memory processes may produce schema-consistent memory errors in recall, within the context of a social contagion memory paradigm using DRM stimuli (Deese, 1959; Roediger & McDermott, 1995). Consistent across all three experiments, pairs of participants recalled items from a set of stimulus materials, discussed their recall with each other, and then recalled the items again individually. The current research provides strong evidence that there is more to collaboration than just the transmission of errors. Overall, participants were exposed to more correct than incorrect information during discussion, even though exposure information was less accurate than each participants' initial recall. Participants incorporated more correct than incorrect exposure items into a later memory report, suggesting that people can distinguish correct information from incorrect information. However, there was little to no change in accuracy following discussion. Within each pair of participants who discussed, recall accuracy increased from initial to final recall for the initially less accurate participant and decreased for the initially more accurate participant. This suggests that during discussion there may be a redistribution of accuracy between participants.