Watkins and Peynirciog Ïlu (1990) found that, in recognition memory tasks, when items are preceded by a problemsolving task, such as doing an anagram, those items are more likely to be judged as "old," regardless of whether they have been presented in the study list (targets) or not (lures). This outcome was initially labeled the revelation effect, because the problem-solving task involved the probe word that was revealed in the solution to the task prior to the recognitiontest. Westerman and Greene (1998), though,have shown that the same effect occurs when the revelation task involves a word unrelated to the test probe. The revelation effect is thus quite puzzling since it is not predicted, nor easily explained, by current theories of recognition memory.The revelation effect has been found with a wide variety of problem-oriented tasks, such as revealing the test word one letter at a time, as an anagram, or rotating the individualletters of a word, or the word as a whole, by varying degrees (Watkins & Peynirciog Ïlu, 1990). Westerman and Greene (1998) also found that the revelation effect can be obtainedwhen item recognitionis precededby letter memoryspan tasks, letter-counting tasks, and synonym-generation tasks. The revelation effect occurs regardless of success on the problem-solving task (Westerman & Greene, 1998, Experiment 1), and the size of the effect is not influenced by the time and effort dedicated to the task (Luo, 1993;Peynirciog Ïlu & Tekcan, 1993;Westerman & Greene, 1998). To date, no revelation effect has been found when the type of information involved in the revelation task is different from the type of information used for the recognition test. Westerman and Greene found that simple arithmetic problems and a digit-span task presented before verbal recognition probes did not produce a reliable revelation effect.Recent studies (Cameron & Hockley, 2000;Westerman, 2000) have demonstrated another boundary condition for the revelation effect. It appears as though the revelation task influences episodic memory decisions based on familiarity, but not decisions that involve recall or recollection of the study episode. The assumption that the revelation task only influencesfamiliarity-baseddecisionsprovides an explanationof two aspects of the revelation effect. First, the revelation effect is typically greater for false alarms than for hits (cf. Hicks & Marsh, 1998). This would be expected if false alarms are most often the result of familiaritybased decisions, whereas a proportion of hits are based on recollection rather than on familiarity. Second, Bornstein and Neely (2001) examined the revelation effect for judgments of item frequency, and in two of three experiments, found that the revelation effect became larger as actual frequency increased. Brown (1995) has shown that frequency judgments can be based either on a count of retrieved instances (enumeration)or on a more general familiarity-based strategy (estimation). If, as Brown suggests, subjects use an enumeration strategy more often when frequ...

Glanzer and Adams (1985, 1990) showed that the mirror effect is a prevalent feature of recognition memory performance. This effect describes the relationship between hit and false alarm rates across two conditions that differ in terms of their level of discrimination or overall accuracy. More specifically, the hit rate is higher and the false alarm rate is lower in the more accurate condition than in the less accurate condition. In other words, the false alarm rates mirror the order of the hit rates. Two general types of manipulations give rise to the mirror effect. One involves comparing discrimination for different classes or types of stimuli that differ in memorability, and the other involves differentially strengthening one set of study items, relative to another set. Stimulus-Based Versus Strength-Based Mirror EffectsA variety of stimulus manipulations give rise to the mirror effect (Glanzer & Adams, 1985). The prototypical example is word frequency: Low-frequency words are usually associated with a higher hit rate and a lower false alarm rate than are high-frequency words in typical tests of yes-no and forced choice recognition (e.g., Glanzer & Adams, 1990; but see Criss & Shiffrin, 2004a, regarding the role of encoding on the generality of the word frequency mirror effect). Another way to produce a mirror pattern involves between-list manipulations of stimulus strength. The hit rate is higher and the false alarm rate is lower for studied items that were presented more often or more slowly than for items that were presented less often or at a faster rate at study (e.g., Cary & Reder, 2003;Murnane & Shiffrin, 1991;Stretch & Wixted, 1998).Glanzer and Adams's (1985) meta-analysis demonstrating the regularity of stimulus-based mirror effects motivated the development of a number of new theoretical accounts of recognition memory (e.g., Dennis & Humphreys, 2001;Glanzer & Adams, 1990;Joordens & Hockley, 2000;Malmberg, Holden, & Shiffrin, 2004;McClelland & Chappell, 1998;Reder et al., 2000;Shiffrin & Steyvers, 1997). The challenge provided by stimulusbased mirror effects, in signal detection theory (SDT) terms, was to explain why the order of the underlying distributions on the strength-of-evidence axis changes when the different classes of stimuli are old, as compared with when they are new-that is, why the weakest items, when new, become the strongest items, when old. Murdock (2003) has termed this a "leapfrog" effect because the weakest items (low-frequency new items) jump over stronger items (high-frequency new words) to be better remembered at test after a study presentation. Stretch and Wixted (1998) Strength-based mirror effects occur when the hit rate is higher and the false alarm rate is lower following strongly encoded study lists than when following more weakly encoded study lists. In Experiments 1A and 1B, strength-based mirror effects were observed in separate tests of single item and associative recognition for random word pairs. In Experiment 2, strength-based mirror effects were again seen when i...

The revelationeffect is evidenced by an increase in positive recognition responses when the test probe is immediately preceded by an unrelated problem-solving task. As an alternative to familiarity-based explanations of this effect (Hicks & Marsh, 1998;Westerman & Greene, 1998), Niewiadomski and Hockley (2001) proposed a decision-based account in which it is assumed that the problem-solving task displaces the study list context in working memory, leading subjects to adopt a more liberal recognition criterion. In the present study, we show that the revelation effect is seen when the stimulus materials are pure lists of very rare words or nonwords. In contrast, for mixed lists of common words and very rare words or nonwords, the revelation effect is found for common words but disappears for very rare words and nonwords. We argue that, in mixed lists, the liberal decision bias following the revelation task and the criterion changes between common words and very rare words and nonwords serve to offset each other.

No abstract