Context, remember–know recognition judgements, and ROC parameters

Abstract

Recent work (e.g., Dunn, 2004; Heathcote, 2003) has questioned the necessity of postulating two processes to explain recognition memory. As part of this trend, strength theories of the remember–know methodology have gained in support. We present three experiments with pictorial material in which we force participants to use differential contextual information at test. Participants were required to give remember–know judgements and confidence ratings for each test stimulus. Hits, false alarms, remember–know data, and discrimination indices indicated systematic variations as a function of the availability and use of contextual information. Moreover, when we normalised the receiver operating characteristic data in terms of z-scores, the slopes were lower than 1 and slightly concave. Additionally, we computed the same set of statistical indices suggested by Wixted and Stretch (2004), with mixed results. Overall, we think that the data support a two-factor theory of remember–know and recognition, although many results fit well signal detection views of the task. Finally, the idea that remember and know responses are pure manifestations of recollection and familiarity seems difficult to sustain. We think that a productive use of the remember–know methodology involves the minimisation of the bias factors that may contaminate the responses, in addition to the introduction of the experimental manipulations needed to promote recollective and/or familiarity processes.

Acknowledgements

This research was supported by Grant SEJ2004-02541 from The “Dirección General de Investigación Científica y Técnica (Spanish Ministry of Education and Technology”).

We thank John T. Wixted for his thoughtful comments on an earlier version of this article.

Notes

¹Note the similarity between these z-ROC values estimated globally and the values estimated individually (see Table 2).