Fluency-based production and memorability-based reduction of false alarms in recognition memory

The production of false alarms in recognition memory tests has long been of interest to memory researchers. A recent paradigm devised to demonstrate false recognition was the "hension" effect paradigm (Whittlesea & Williams, 1998), where the false alarm (FA) rate for regular nonwords (e.g., HENSION) was found to exceed that for natural words (e.g., CURTAIN) and for irregular nonwords (e.g., STOFWUS). The hension effect has been cited as empirical evidence for the discrepancyattribution hypothesis, which assumes that the high FA rate for regular nonwords arose because the processing of these fluent, yet meaningless items is discrepant. Discrepancy in tum prompts fluency misattribution (i.e., false alarms) to occur. An objective ofthis thesis was to examine the suitability of the discrepancyattribution hypothesis in explaining the hension effect. In Experiments 1 - 4, the sense of discrepancy associated with the hension effect materials was manipulated. These experiments found that discrepancy did not appear to underlie false recognition. As an alternative explanation for the hension effect, it was argued that recognition judgments are dependent on fluency-based processes for regular and irregular nonwords. However, the low FA rate observed for natural words was due to their high memorability levels (as substantiated by ratings data in Experiment 5), which allowed participants to correctly reject these items when they acted as lures. Compelling evidence for the involvement of a memorability-based, metacognitive strategy in lure rejection came from the finding of a FA rate decrease for items whose memorability levels have been experimentally enhanced (Experiments 7 - 8). These results were interpreted from the perspective oftwo signal-detection models, one based on criterion shifts and one based on distribution shifts (a multi-process model). Support for the multi-process model was found in Experiments 9 - 10, where it was demonstrated that lure groups of differing intrinsic (item-based) and extrinsic (experimentally-manipulated) memorability levels are located on distinctly separate points on a hypothetical strength-of-evidence scale