Applying confidence-accuracy characteristic plots to recognition memory

Abstract

Confidence-accuracy relationships have been of great interest in the study of psychophysics, perception, and decision making, fields that have employed confidence-accuracy calibration plots using 100-point scales. More recently, they have been applied to the study of eyewitness memory and the technique has been extended to scales of many types by the development of confidence-accuracy characteristic (CAC) plots. We briefly describe the development of these two methods of analysis and the results obtained with them in eyewitness identification studies. We then extend the study of CAC plots to more standard recognition memory tasks, mostly those involving recognition of faces. We find that confidence is highly related to accuracy in recognition of 100 faces. We also use CAC plots to show that the “grain” of a confidence scale (e.g., 1–4, 1–20, 1–100) makes surprisingly little difference in judgments. CAC plots look similar in each scale type. We also show that CAC plots for correct rejections are quite different from those for hits, both in eyewitness identification experiments and in recognition memory of faces in a yes/no paradigm. Finally, we explore a different method of constructing CAC plots, using misses rather than false alarms, and discuss the advantages of each approach. The use of CAC plots is just beginning in the study of recognition memory, and the technique is useful for examining the critical issue of confidence-accuracy relations.