Continuous Feedback in Recognition Memory

Abstract

Signal-detection models of recognition memory assume that S's decision as to whether or not he recognizes a stimulus is a function of a criterion value. In selecting his criterion, S takes into consideration the a priori probability of an old item and the costs and rewards of a hit or false alarm. In the present experiment, Ss were given feedback during recognition testing in an effort to determine whether it would aid S in selecting his criterion. The results showed that the feedback improved recognition performance by significantly reducing the number of false alarm errors. Evidence was presented to support the claim that S's criterion was affected by this manipulation.

Can corrective feedback improve recognition memory?

“…These results are consistent with those of Titus (1973) and the digit and letter conditions of Estes and Maddox (1995) but are in contrast to the findings in the word condition of the latter study. Although the reason for this disparity is not obvious, the strengthening of the base-rate manipulation from .67-/.33-old in the Estes and Maddox experiments to .75-/.25-old in Experiment 2 is likely a factor.…”

“…Feedback significantly increased response criterion, but no compelling evidence of an impact on sensitivity was obtained. These results demonstrate the potential utility of feedback in cases of highly uneven base rates, although the use of only 15 old items (each of which was viewed six times during study) likely limited the variance of the signal distribution relative to that of the noise distribution, complicating the interpretation of the reported computations of sensitivity and bias (Titus, 1973). 1 Estes and Maddox (1995) used a continuous recognition test and manipulated three between-subjects variables: proportion old (.33 or .67), stimulus type (digits, letters, or words), and trial-by-trial accuracy feedback (provided or withheld).…”

“…In early work on the factors affecting response bias in recognition, Titus (1973) gave trial-by-trial feedback during recognition of CVC trigrams. Only 20% of the test items were old.…”

“…In the experiments reported here (and in a few previously published studies described below; Titus, 1973;Verde & Rotello, 2007), a single test list followed a single study list, and each test item was presented only once.…”

“…Feedback would seem most likely to facilitate criterion placement when base rates of old and new items are unequal, a condition not typically included in recognition experiments but more common in studies of category learning (e.g., Kruschke, 1996). By conveying the correct responses on a trial-by-trial basis, feedback should enable participants to tune in to underlying probabilities of old and new items and to adjust criterion accordingly (Estes & Maddox, 1995;Titus, 1973). The impact of feedback on response bias when base rates of old and new items are equal has been the subject of more recent work (Rhodes & Jacoby, 2007;Verde & Rotello, 2007), which we review below.…”