Enhanced processing fluency leads to biases in source memory

Abstract

The present experiments were conducted to determine whether processing fluency affects source memory decisions. In the first three experiments, participants decided whether test items appeared in the same sensory modality (Experiments 1A, 1B) or perceptual form (font type, Experiment 2) at study and test. The results were consistent across the three studies and showed that perceptual priming leads to an increase in reports that stimuli were presented in the same sensory or perceptual form during the study and test phase. Experiment 3 showed that conceptual fluency affects source attributions in much the same way as perceptual fluency, and Experiment 4 showed that fluency is associated with a subjective experience of familiarity even when it might serve as a basis for source inference. These results are consistent with recent neuropsychological and neuroimaging evidence that familiarity-based processes contribute to source memory decisions under some circumstances, such as when items and contexts are unitized rather than merely bound together at encoding.

Keywords:

• Recognition memory
• Source memory
• Familiarity
• Recollection
• Processing fluency

Notes

¹Correct source attributions in this and subsequent experiments reflect the proportion of targets assigned to the correct source—for example, p(visual targets called “seen”) and p(auditory targets called “heard”) in Table 4—and incorrect source attributions reflect the proportion of targets assigned to the incorrect source—for example, p(visual targets called “heard”) and p(auditory targets called “seen”) in Table 4.

²Although predictive and nonpredictive sentence stems could be characterized as high cloze and low cloze probability sentences, respectively (Bloom & Fischler, 1980), the present terminology was adopted because it is in keeping with the terminology of other papers that have used this procedure to manipulate conceptual fluency during recognition memory.

³Note that this assumes that remember and familiar responses are based on the same or overlapping sources of information (Donaldson, 1996; Dunn, 2004; Wixted & Stretch, 2004).

⁴Confidence ratings were taken instead of yes/no responses because it was thought that the greater range provided by the confidence rating method might provide a more sensitive test of the claim that fluency elevates levels of recognition confidence more when it aids source inference than when it does not.

⁵A familiar response was substituted for the more conventional know response simply because participants seem to find this term more intuitive and meaningful.

⁶These results were confirmed by a set of analyses carried out on estimates of discrimination (d′) and criterion placement (C). Encoding condition had no effect on either discrimination or criterion placement, all Fs < 1. See Tables 2 and 3 for a summary of the effects of fluency on d′ and C.

⁷If participants in the auditory–visual group adopted a more conservative set of high confidence criteria without altering their placement of moderate and lenient decision criteria, then this could have obscured differences in the magnitude of the priming effect between the two conditions. For this reason, estimates of C were calculated for various levels of recognition confidence. However, because many participants were missing observations at several levels of confidence, it was not possible to compute a value of C for every confidence rating. Therefore, ratings were combined, such that high confidence reflected items assigned a rating of 7 or 8, moderately high confidence reflected items assigned a rating of 5 or higher, and moderately low confidence reflected items assigned a rating of 3 or higher. Analysis revealed that there was no main effect of encoding condition and no Confidence × Encoding Condition interaction, both Fs < 1. Thus, participants in the auditory–visual condition did not appear to adopt a more conservative set of decision criterion than participants in the all-visual encoding condition.

⁸The recognition confidence data are consistent with this interpretation. Priming did not differentially affect high confidence recognition in the auditory–visual group relative to the all-visual group.

⁹I thank an anonymous reviewer for suggesting this.

¹⁰Although source accuracy was somewhat lower for auditory targets than for visual targets in Experiment 1B, the difference was not significant. This may simply have been due to a lack of statistical power, given that Experiment 1B had fewer participants than the other experiments.