Three recognition memory experiments were conducted using modified Deese-RoedigerMcDermott (DRM) and DRM paradigms. In Experiment 1, the reaction time (RT) of the false alarms to critical nonpresented words 1 (false memory) was compared with the RT of hits to the critical presented words and with the RT of hits to the studied list words (true memory). The RT of the false alarms to the critical nonpresented words was significantly longer than that of the hits to the critical words and than that of the studied list words. In Experiment 2, in addition to RT, participants' confidence level was measured on a 4-point scale for a yes or no response. Confidence rating was significantly higher for the hits to the critical presented words and to the list words than for the false alarms to the critical nonpresented words. Experiment 3 further showed that how similar false memory experience was to that of true memory was a function of retention size (number of lists of words retained in memory). In all three experiments, the participants' recognition RTs distinguished false memory from veridical memory, and in Experiments 2 and 3, so did their confidence ratings. Therefore, false memory and veridical memory differ at both the objective and the subjective levels. The results are consistent with a single familiarity dimension model of recognition memory.

A common response bias in psychophysical judgments is regression toward the mean (overestimation of small and underestimation of large values, or the response contraction bias). The same bias is observed in magnitude estimation from memorized quantities. Participants estimated alphabetic interval distances between 2 letters for different levels of interletter distances. The underestimated and overestimated values and the point of least error changed, depending on the level of alphabetic distances judged; furthermore, their estimation showed a progressively increasing tendency toward the mean, rendering the estimation progressively less accurate as the estimation task was repeated. We conclude that the regression toward the mean in memorial quantifying judgment derives from a cognitive adaptation process rather than from a permanent, compressed memory representation of the stimuli. Two opposing views on the adaptive meaning of this judgment bias are discussed.

No abstract