No abstract

824Unstudied test items that are judged to be perceptually or conceptually similar to study items produce increased false recognition for a wide range of stimuli (e.g., words, numbers, shapes, and pictures). The resulting negative correlation between similarity and accuracy, which we call the memory similarity effect, 1 is a general regularity in recognition memory tested by either yes/no or two-alternative forced choice (2AFC) procedures (e.g., Wickelgren, 1977). A second regularity, a positive correlation between recognition accuracy and retrospective confidence (i.e., confidence judgments given at test) is attributed to memory trace strength's being the common basis for both decisions (Hart, 1967); stronger traces cause greater confidence and are associated with more accurate recognition decisions.Hence, it was doubly surprising when Tulving (1981) reported a reversal of both regularities in 2AFC recognition decisions for scenic pictures followed by a confidence rating. The reversals were caused by choice similarity, the similarity between studied (target) and unstudied (lure) test alternatives. In higher choice similarity pairs (AA ), both items were halves of the same scenic picture, in which one half (A) was studied and the other (A ) was not. In lower choice similarity pairs (BC ), the lure (C ) was the unstudied half of a studied picture (C) that was not similar to the target (B). Since the lure in both choice similarity conditions was similar to a studied item, memory similarity was controlled. Tulving found that, for picture halves rated as higher in similarity, accuracy was increased but confidence was reduced for AA relative to BC test pairs. These results are all the more compelling because, in the same experiments, the usual memory similarity effect was obtained when choice similarity was controlled.Accuracy and confidence were greater for lower than for higher memory similarity pairs. However, for picture halves rated lower in similarity, only the choice similarity effect on confidence was reliable.Despite the surprising nature of Tulving's (1981) results, to our knowledge only one replication has been performed: Dobbins, Kroll, and Liu (1998) used the same type of stimuli that Tulving used, but their choice similarity effect was weaker for both accuracy and confidence, likely because memory similarity was lower for their stimuli. We know of no attempt to determine whether the choice similarity effect and the dissociation between confidence and accuracy apply to other types of stimuli. In the three experiments reported here, we investigated whether these phenomena also occur in recognition memory for faces. The experiments differed only in their test procedures. Experiment 1 followed Tulving, with a 2AFC response followed by a confidence rating on a 3-point scale. In Experiment 2, participants simultaneously indicated their choice and confidence on a 6-point scale. Experiment 3 followed Dobbins et al. in requiring a "remember/know" response after a simultaneous choice and confidence rating.Two mod...

a b s t r a c tWe used quantitative modeling of remember-know responses (Tulving, 1985) to investigate the processes underlying recognition memory for faces in the choice-similarity paradigm (Tulving, 1981). Similarity between recognition test choices produces opposite effects on confidence and accuracy in this paradigm. We extended existing models of this double dissociation to account for remember-know responses, by adding a variable recollection criterion to Clark's (1997) single-process model and by adding graded recollection strength to Dobbins, Kroll, and Liu's (1998) dual-process model. Both models provided an accurate and comprehensive account of objective and subjective judgments in an experiment we conducted on memory for faces, and of data from Dobbins et al. on memory for natural scenes. Model selection techniques were used to refine these accounts, providing insight into the psychological processes proposed by each approach and into their implications for the relationship between recollection and confidence in two-alternative forced choice recognition.

No abstract