Signal‐detection models as tools for interpreting judgements of recollections

Abstract

This paper considers how the two-criterion signal-detection model can be used to interpret judgements of recollection from the remember-know paradigm. We propose that, among other uses, the model can be applied to discriminate results that provide strong evidence for the influence of multiple memory processes from results that are merely consistent with the assumption of such influences. The specific logic motivating this approach is that results falsifying the two-criterion signal-detection model provide strong evidence for the influence of multiple-memory processes on judgements of recollection. We believe that focusing theory construction on results that provide strong evidence of multiple memory processes will lead to the construction of coherent, parsimonious models of the relationship between memory processes and consciousness. We review recent papers by Conway and Dewhurst (1995), Hirshman andLanning (1999), andConway et al. (2001) to provide examples of the usefulness of using the two-criterion signal-detection model. Conway et al. (2001) present a series of experiments examining the effects of selfjudgement during study on remember-know judgements in recognition memory. A central finding of their experiments is that when hit rates are equated across encoding conditions, self-judgements during study produce more remember responses than control-encoding judgements in later recognition memory. Importantly, this finding occurs when the retention interval is greater than one hour, but not at shorter retention intervals (e.g. 5 minutes). At these shorter retention intervals, the proportion of remember responses tends to be equal across the self-and control-encoding conditions when hit rates are approximately equal across these encoding conditions. Conway et al. emphasize that their findings have important implications for conceptions of the working self, consolidation, and longterm memory.We are extremely pleased to see continuing interest in conceptions of the self and its role in episodic memory. We are also gratified to see a replication of Hirshman and Lanning's (1999) finding that remember responses are approximately equal across the self-and control-encoding conditions at short retention intervals when overall hit rates are equated across these conditions. We concur with Conway et al. that the contrast between the results at long and short retention intervals provides an important constraint on theories

Perceptual priming enhances the creation of new episodic memories

“…Second, the results of the ROCs analyses strongly support the idea that the distinct states of consciousness gauged with R and K responses in our experiment represent two distinct sources or components of memory that do not result from variations in response criteria/decision processes along a single memory trace (e.g., see also Conway, Dewhurst, Pearson, & Sapute, 2001;Reder et al, 2000;Rotello et al, 2004 who have reached the same conclusions). While this conclusion only applies to our paradigm and does not rule out definitively one-dimensional SD models (see Hirshman et al, 2002), it reinforces the idea that perceptual priming modulates the creation of new episodic memories independently of familiarity or know judgments, especially when the nature of the recollected details (as the sound context in our study) becomes critical (Banks, 2000). Table 3 Cumulative proportions of hits and FAs and sensitivity measures (d 0 ) for each z-ROCs point (R and RK) according to the three priming conditions (''no-priming'', ''low-priming'' and ''high-priming'') Furthermore, the words that were associated with the creation of new episodic memories in the subsequent recognition task were subject to more perceptual priming during incidental encoding (RT differences between ''no-priming'', ''low-priming'' and ''high-priming'' conditions) than words that were subsequently judged to be known or forgotten, which did not yield any significant differences between priming conditions at encoding.…”

Dissociating familiarity from recollection using rote rehearsal

“…However, it is useful to consider whether or not the data are also consistent with a single-process, signal detection account. In the case of remember/know (familiar) paradigms, the most widely cited model is the two-criteria signal detection model (Donaldson, 1996;Hirshman & Henzler, 1998;Hirshman, Lanning, Master, & Henzler, 2002;Inoue & Bellezza, 1998). Indeed a recent search of the Social Science Citation Index indicated that the Donaldson (1996) article had been cited 119 times.…”

Transferring voice effects in recognition memory from remembering to knowing

“…These memory-strength/decision process accounts have now been shown not to provide an adequate general account of remembering and knowing (see, e.g., Conway et al, 2001;Gardiner, Ramponi, & RichardsonKlavehn, 2002), although more sophisticated dualcomponent accounts based on a similar kind of approach are proving more successful (e.g., Rotello, Macmillan, & Reeder, in press;Yonelinas, Dobbins, Szymanski, Dhaliwal, & King, 1996). Nonetheless, as Hirshman and his colleagues have suggested (Hirshman, Fisher, Henthorn, Arndt, & Passannante, 2002;Hirshman, Lanning, Master, & Henzler, 2002), there may still be some theoretical value in applying the single trace-strength model in some circumstances-for example, to obtain evidence for whether one can discount an explanation in terms of decision processes in those particular circumstances.…”

“…Individual t tests for voice congruency effects as measured by these A9 estimates (which we will also not report in detail) merely reflected those for the simple response proportions and did not materially alter the conclusions to be drawn. Thus, the results of the signal detection analyses serve mainly to allow us to discount the possibility that the changes in awareness with respect to voice congruency effects might be explained simply in terms of decision processes (Hirshman, Fisher, et al, 2002;Hirshman, Lanning, et al, 2002).…”