Faking in the Autobiographical Implicit Association Test

The autobiographical Implicit AssociationTest (aIAT) was recently introduced in this journal as a new and promising lie-detection tool. The initial report found 91% accuracy in determining which of two autobiographical events was true. It was suggested that the aIAT, unlike other lie-detection tests, is resistant to faking. We investigated whether participants can strategically alter their performance on the aIAT. Experiment 1 showed that participants guilty of a mock theft were able to obtain an innocent test outcome. Two additional experiments showed that guilty participants can fake the aIAT without prior experience with the aIATand when a response deadline is imposed. The aIAT is subject to the same shortcomings as other lie-detection tests. In its war on terror, the U.S. government now uses handheld polygraphs for rapid screening of suspects of terrorism (e.g., ‘‘Are you a member of the Taliban?’’). This illustrates the great need that exists among law-enforcement agencies for lie-detection tools that are easy to apply. The enthusiastic use of the pocket polygraph stands in sharp contrast to the highly critical evaluation of this method by the prominent National Research Council (2003), which concluded that errors frequently occur and that successful faking in polygraph tests is possible. An important recommendation was to develop new lie-detection methods. The autobiographical Implicit Association Test (aIAT) is a very simple new lie-detection tool. The aIAT is based on reaction times and requires only a standard computer. The aIAT can be used to assess which of two autobiographical events is true. In an IAT used for criminal investigations, for example, sentences related to four categories are used: true statements unrelated to the crime (e.g., ‘‘I’m in front of a computer’’), false statements unrelated to the crime (e.g., ‘‘I’m in the city library’’), confession statements confirming that the participant committed the crime (e.g., ‘‘I stole the CD-ROM containing exam questions’’), and denial statements denying that the participant committed the crime (e.g., ‘‘I did not steal the CD-ROM’’). Sentences are presented one by one, and participants are required to give a speeded response depending on the task (see Table 1). The aIAT consists of two tasks. In the confession-true task, confession and true statements are mapped to one key, and denial and false statements are mapped to the second key. In the denial-true task, the assignments are reversed (denial and true statements are assigned to one key, confession and false statements assigned to the other key). Sartori, Agosta, Zogmaister, Ferrara, and Castiello, et al. (2008) argued that guilty participants should be faster in the confession-true task than in the denial-true task, whereas the reverse should be the case for innocent participants. In a series of six experiments, Sartori et al. found that the aIAT has an extremely high accuracy in determining which of two autobiographical events is true (overall 91%). The authors concluded that ‘‘the aIAT is an accurate method to detect concealed knowledge and outperforms currently available lie-detection techniques’’ (p. 780). This conclusion may be premature because it is unknown whether the aIAT is susceptible to faking. In Experiment 1, we investigated whether participants can strategically alter their performance on the aIAT by providing them with a single instruction sheet on how to beat the aIAT. Subsequent experiments examined the conditions under which effects on the aIAT can be faked.