Author's personal copy Regulating accuracy on university tests with the plurality option q Philip A. Higham * School of Psychology, University of Southampton, Highfield, Southampton SO17 1BJ, UK article info Article history: Received 27 January 2012 Received in revised form 1 August 2012 Accepted 10 August 2012 Keywords: Accuracy regulation Metacognition Calibration Base-rate fallacy Underconfidence abstract A single experiment is reported in which introductory psychology students were administered a multiple-choice test on psychology with either 4 (n ¼ 78) or 5 alternatives (n ¼ 92) prior to any lectures being delivered. Two answers were generated for each question: a small answer consisting of their favorite alternative, and a large answer consisting of all alternatives except for their least favorite one. They also rated confidence of the accuracy of both answers and selected one for grading (plurality option). Replicating previous research, there was evidence of a confidence-accuracy (C-A) dissociation for selected answers. Specifically, accuracy was higher, but confidence was lower, for selected large answers compared to selected small answers. However, unlike previous research, the C-A dissociation was coupled with marked underconfidence for both types of selected answers. The results are discussed in terms of option fixation, the base-rate fallacy, response criteria, and alternative plausibility. Ó 2012 Elsevier Ltd. All rights reserved. 1. Introduction A number of recent studies have investigated how people metacognitively regulate the accuracy of their memory reports. One means to regulate accuracy is to use a report option that can be used to filter out unwanted, low-quality information by with- holding it (e.g., Higham, 2002; Koriat & Goldsmith, 1996; Roebers & Schneider, 2005). For example, because the stakes are high, eyewitnesses queried about the identity of a suspect may choose to withhold their answer unless they are very confident. A second means to regulating accuracy is a grain-size option, which can be used to reduce the amount of detail in memory reports (e.g., Ackerman & Goldsmith, 2008; Goldsmith, Koriat, & Pansky, 2005; Goldsmith, Koriat, & Weinberg-Eliezer, 2002; Weber & Brewer, 2008; Yaniv & Foster, 1995, 1997). For example, when asked about the timing of a robbery, an unsure eyewitness may report “some time between 9pm and 11pm” rather than “10pm” allowing for a margin of error and increasing the likelihood that the answer will be correct. Research on the grain-size option has been narrowly focused in that the materials used have been restricted to quantitative ques- tions involving numeric answers (e.g., age, time, distance, or height; Goldsmith et al., 2002) or gradients (e.g., dark color versus navy blue; Weber & Brewer, 2008; although see Evans & Fisher, 2011). In response to this limitation, Luna, Higham, and Martin- Luengo (2011) (see also Luna & Martin-Luengo, in press) investi- gated a means for regulating accuracy that is closely related to the grain-size option but which better lends itself to situations in which there are qualitatively different candidate answers to contend with. For example, when asked about the type of get-away car used for a robbery, an uncertain eyewitness might offer three plausible possibilities such as “a Vauxhall, Peugeot, or VW” instead of just one alternative. Luna et al. called this new mechanism the plurality option because it involves control over the number of answers being offered as opposed to the specificity of a single answer. Regardless of the regulatory option that is used, any accuracy advantages come with a cost of low informativeness (see Goldsmith & Koriat, 2008 for a review). In particular, eyewitnesses who respond “I don’t know” to too many questions in a courtroom, or those who provide answers that are too vague, convey their igno- rance in other ways. Consequently, although moderate use of these regulatory options can be used to improve memory accuracy rela- tive to cases in which they are unavailable, their overuse can produce uncommunicative, potentially useless, reports. 1.1. Accuracy regulation in educational contexts As the above examples suggest, more work on accuracy regu- lation has been conducted in forensic than educational contexts (e.g., Evans & Fisher, 2011). As an exception, Higham (2007) showed that students writing the SAT Reasoning Test omit answers in q Portions of this research were presented at the 5th Meeting of The International Conference on Memory, York, UK, August 1e5, 2011. Thanks to Adam Pawley for research assistance. * Tel.: þ44 23 80595942. E-mail address: higham@soton.ac.uk. Contents lists available at SciVerse ScienceDirect Learning and Instruction journal homepage: www.elsevier.com/locate/learninstruc 0959-4752/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.learninstruc.2012.08.001 Learning and Instruction 24 (2013) 26e36