Operations that improve the accuracy of associative recognition can do so in qualitatively different ways. Increasing repetitions and study time increases hit rates but has small effects on false alarm rates, and the specific patterns of false alarms are dependent on the stimuli (e.g., pairs of words, pseudowords, faces, or Chinese characters). In contrast, manipulating the type of stimuli that make up pairs produces a robust mirror effect: The hit rate is greater, and the false alarm rate is lower, for better recognized stimuli. To explain these findings, a model of single-item recognition is extended to associative recognition. Within this dual-process framework, the present results suggest that words are encoded more extensively than nonverbal stimuli and that recognition of frequently encountered stimuli (words and faces) is more likely to be based on recollection than is recognition of uncommon stimuli (pseudowords and Chinese characters).
In an associative recognition (AR) task, subjects study pairs of items. Items that are studied together are intact pairs, items that are not studied together are rearranged pairs, and subjects discriminate between them at test. Positive endorsements of intact and rearranged pairs are respectively referred to as hits andfalse alarms. Generally speaking, accuracy improves as the difference between hit rates (HRs) and false alarm rates (FARs) increases, and thus, factors can improve AR via qualitatively different patterns of HRs and FARs: HRs can increase and FARs can decrease, HRs can increase and FARs can remain steady, or HRs can remain steady and FARs can decrease.
Different operations affect AR in different ways. For instance, varying the type of stimuli that make up to-beremembered pairs (referred to here as pair type) produces a mirror effect (Glanzer & Adams, 1985): HRs are greater for words than for nonverbal stimuli, but FARs are less for words than for nonverbal stimuli (Criss & Shiffrin, 2004; Greene, 1996,2004; Hockley, 1994). In contrast, a mirror effect is not always observed with increases in repetitions or study time: When strength is varied within a list, the probability of endorsing intact pairs increases with target repetitions, but the probability of endorsing rearranged pairs remains fairly steady (e.g., Kelley & Wixted, 2001); but when strength is varied between lists, a mirror effect is observed (Clark & Shiffrin, 1992). Thus, some operations have robust effects on both HRs and FARs, whereas other operations primarily affect HRs. (To the best of our knowledge, no factors have been identified that produce steady HRs and decreasing FARs.)
It is unknown why these variables affect AR in different ways or how pair type and standard strengthening operations interact. In addition, it is unknown how generalizable the pair type effect is, because words have traditionally been used in recognition memory experiments. To address the empirical questions, we present an AR experiment in which repetitions, study time, and pair type (words, pseudowords, Chinese characters, or faces) were varied. Using an extension of a dual-process memory model (Malmberg, Holden, & Shiffrin, 2004), we hypothesized that strengthening different types of stimuli might affect the encoding of pairs, the contribution of recollection to AR, or both.
We will begin with a discussion of when mirror effects occur and do not occur and the constraints these findings place on models of recognition memory. In the next section, we will generate predictions of an extension of Malmberg, Holden, and Shiffrin's (2004) recognition model in order to understand how the pair type mirror effect and the effect of repetitions can be explained by the same model. Finally, we will report the results of an experiment in which standard strengthening operations and pair type were varied. Given these results, we then will fit the three competing models to the data in order to quantitatively choose among them. …