The effects of snap decision making on episodic recognition memory

The present study is interested in whether recognition memory errors made from snap decisions can be detected and corrected. Fifty two participants from the University of Tasmania made a two-stage recognition decision to words in a test list. The first decision, was a fast (<1sec) recognition memory (old/new) decision; the second a slower confidence response (high, low, change of mind). Words in the test list were manipulated for frequency (high, low), in the English language, and concreteness (high, low). Analysis of the change of mind responses showed an equal reduction from first response to second in the false alarm rate for both high frequency and low frequency words, and a significantly larger improvement in the hit rate for low frequency words from first response to second. Changed responses had a longer processing time (680ms) than unchanged responses (390ms). A significant word frequency mirror effect was also found, but not a significant word concreteness effect. The results indicate that participants can detect and correct errors made from quick decisions, and have important implications for the ability of decision making models' to predict correct responses, and for reducing the noise in speed-accuracy trade-off experiments (Rabbitt, 1969).