Collaborative Research: Understanding and Predicting Eyewitness Identification Errors: Studies Using a Unique Set of Materials from Actual Lineups

在美国，法医DNA测试已经释放了数百名无辜的人，其中大约75%的定罪涉及错误的目击者指认。很明显，需要确定可靠的方法来区分准确和错误的识别。来自实验室模拟实验的理论和数据表明，准确和错误的识别在可测量的方面存在差异（例如，识别速度、置信度、证人言语）。实验室研究还表明，列队偏见（例如，无辜的嫌疑人脱颖而出或管理员无意中引导证人）可能会导致错误的识别。然而，实验室模拟参与者知道犯罪不是真实的，并且无法捕捉到实际证人潜在的重要动态（例如，恐惧，正义动机，对后果的关注）。研究人员有来自实际案件的独特材料，涉及855组照片，每组照片中只有一个人是嫌疑人。其他队员都是无辜的替补队员。填充标识构成错误标识，占所有标识的34%。研究人员提供的软件录下了证人的口供、指认速度和其他相关信息。大多数队列是双盲的；因此，阵容管理员不知道哪一张是嫌疑人的照片，因此无法引导证人。共进行了四项研究。研究1检验了阵容偏见是否能预测错误的识别。研究2测试证人的语言表达是否能预测错误。研究3测试的是，与识别嫌疑人相比，错误是否更慢、更不确定。研究4检验了双盲队列程序是否证明队列管理人员较少转向。数据分析包括层次、混合模型、对数线性分析或方差分析。这项工作可以帮助将实验室模拟结果转化为实际案例，从而帮助司法系统在错误和准确的识别之间进行分类。研究结果将与科学家、警察和检察官广泛分享。小型本科学院的本科生，包括第一代和有色人种，将在这个项目中接受研究训练。

Forensic DNA testing has freed hundreds of innocent people in the U.S. and approximately 75% of these convictions involved mistaken eyewitness identification. The need to identify reliable ways to sort between accurate and mistaken identifications is readily apparent. Theory and data from laboratory-simulation experiments indicate that accurate and mistaken identifications differ in measurable ways (e.g., identification speed, confidence, witness verbalizations). Lab research also shows that lineup biases (e.g., innocent suspect stands out or administrator unintentionally steers the witness) can inflate mistaken identifications. Lab simulation participants, however, know that the crime is not real and fail to capture potentially important dynamics of actual witnesses (e.g., fear, justice motivations, concern for consequences). The researchers have unique materials from actual cases involving 855 photo-lineups where only one person in each lineup was a suspect. The other lineup members were known-innocent fillers. Filler identifications, which constitute mistaken identifications, comprise 34% of all identifications. Software provided by the researchers made audio recordings of witnesses' verbalizations, speed of identifications, and other relevant information for each lineup. A majority of the lineups were administered double-blind; thus, lineup administrators were unaware of which was the suspect's photo and hence could not steer the witness. Four studies were conducted. Study 1 tests whether lineup bias predicts mistaken identifications. Study 2 tests whether witness verbalizations predict mistakes. Study 3 tests whether mistakes are slower and made with less certainty than suspect identifications. Study 4 tests whether double-blind lineup procedures evidence less steering by the lineup administrators. Data analyses involve hierarchical, mixed model log-linear analyses or ANOVAs. This work can help translate lab simulation findings to actual cases and thereby help the justice system sort between mistaken and accurate identifications. Results will be shared broadly with scientists, police and prosecutors. Undergraduate students, including first-generation and those of color at a small undergraduate college, will receive research training in this project.