Some memories are odder than others : Judgments of episodic oddity violate known decision rules
MetadataShow full item record
Current decision models of recognition memory are based almost entirely on one paradigm, single item old/new judgments accompanied by confidence ratings. This task results in receiver operating characteristics (ROCs) that are well fit by both signal-detection and dual-process models. Here we examine an entirely new recognition task, the judgment of episodic oddity, whereby participants select the mnemonically odd members of triplets (e.g., a new item hidden among two studied items). Using the only two known signal-detection rules of oddity judgment derived from the sensory perception literature, the unequal variance signal-detection model predicted that an old item among two new items would be easier to discover than a new item among two old items. In contrast, four separate empirical studies demonstrated the reverse pattern: triplets with two old items were the easiest to resolve. This finding was anticipated by the dual-process approach as the presence of two old items affords the greatest opportunity for recollection. Furthermore, a bootstrap-fed Monte Carlo procedure using two independent datasets demonstrated that the dual-process parameters typically observed during single item recognition correctly predict the current oddity findings, whereas unequal variance signal-detection parameters do not. Episodic oddity judgments represent a case where dual- and single-process predictions qualitatively diverge and the findings demonstrate that novelty is "odder" than familiarity.
O'Connor , A R , Guhl , E , Cox , J & Dobbins , I 2011 , ' Some memories are odder than others : Judgments of episodic oddity violate known decision rules ' , Journal of Memory and Language , vol. 64 , no. 4 , pp. 299-315 . https://doi.org/10.1016/j.jml.2011.02.001
Journal of Memory and Language
This is an author version of an article published in Journal of Memory and Language 64(4), available at http://www.sciencedirect.com