Kenneth J. Malmberg

Recognition memory: A review of the critical findings and an integrated theory for relating them

The development of formal models has aided theoretical progress in recognition memory research. Here, I review the findings that are critical for testing them, including behavioral and brain imaging results of single-item recognition, plurality discrimination, and associative recognition experiments under a variety of testing conditions. I also review the major approaches to measurement and process modeling of recognition. The review indicates that several extant dual-process measures of recollection are unreliable, and thus they are unsuitable as a basis for forming strong conclusions. At the process level, however, the retrieval dynamics of recognition memory and the effect of strengthening operations suggest that a recall-to-reject process plays an important role in plurality discrimination and associative recognition, but not necessarily in single-item recognition. A new theoretical framework proposes that the contribution of recollection to recognition depends on whether the retrieval of episodic details improves accuracy, and it organizes the models around the construct of efficiency. Accordingly, subjects adopt strategies that they believe will produce a desired level of accuracy in the shortest amount of time. Several models derived from this framework are shown to account the accuracy, latency, and confidence with which the various recognition tasks are performed.

The "One-Shot" Hypothesis for Context Storage

In 3 experiments motivated by the implicit memory literature, the authors investigated the effects of different strengthening operations on the list strength effect (LSE) for explicit free recall, an effect posited by R. M. Shiffrin, R. Ratcliff, and S. E. Clark (1990) to be due to context cuing. According to the one-shot hypothesis, a fixed amount of context is stored when an item is studied for at least 1 or 2 s. Beyond the initial context storage, increases in study time or different orienting tasks do not influence the amount of context that is stored, and thus only spaced repetitions should produce a positive LSE. Consistent with prior findings, spaced repetitions always produced a positive LSE, but increases in depth of processing, study time, and massed repetitions did not. A model implements the one-shot hypothesis, and a role for context storage as a link between episodic and semantic memory is discussed.

On the form of ROCs constructed from confidence ratings.

A classical question for memory researchers is whether memories vary in an all-or-nothing, discrete manner (e.g., stored vs. not stored, recalled vs. not recalled), or whether they vary along a continuous dimension (e.g., strength, similarity, or familiarity). For yes-no classification tasks, continuous- and discrete-state models predict nonlinear and linear receiver operating characteristics (ROCs), respectively (D. M. Green & J. A. Swets, 1966; N. A. Macmillan & C. D. Creelman, 1991). Recently, several authors have assumed that these predictions are generalizable to confidence ratings tasks (J. Qin, C. L. Raye, M. K. Johnson, & K. J. Mitchell, 2001; S. D. Slotnick, S. A. Klein, C. S. Dodson, & A. P. Shimamura, 2000, and A. P. Yonelinas, 1999). This assumption is shown to be unwarranted by showing that discrete-state ratings models predict both linear and nonlinear ROCs. The critical factor determining the form of the discrete-state ROC is the response strategy adopted by the classifier.

The Effect of Normative Context Variability on Recognition Memory

According to some theories of recognition memory (e.g., S. Dennis & M. S. Humphreys, 2001), the number of different contexts in which words appear determines how memorable individual occurrences of words will be: A word that occurs in a small number of different contexts should be better recognized than a word that appears in a larger number of different contexts. To empirically test this prediction, a normative measure is developed, referred to here as context variability, that estimates the number of different contexts in which words appear in everyday life. These findings confirm the prediction that words low in context variability are better recognized (on average) than words that are high in context variability.

Feature frequency effects in recognition memory

Rare words are usually better recognized than common words, a finding in recognition memory known as theword-frequency effect. Some theories predict the word-frequency effect because they assume that rare words consist of more distinctive features than do common words (e.g., Shiffrin & Steyverss, 1997, REM theory). In this study, recognition memory was tested for words that vary in the commonness of their orthographic features, and we found that recognition was best for words made up of primarily rare letters. In addition, a mirror effect was observed: Words with rare letters had a higher hit rate and a lower false-alarm rate than did words with common letters. We also found that normative word frequency affects recognition independently of letter frequency. Therefore, the distinctiveness of a words orthographic features is one, but not the only, factor necessary to explain the word-frequency effect.

The word frequency effect for recognition memory and the elevated-attention hypothesis.

Empirical tests were conducted on theelevated-attention hypothesis that low-frequency (LF) words are better recognized than high-frequency (HF) words because LF words attract more attention than do HF words (e.g., Glanzer & Adams, 1990). The elevated-attention hypothesis predicts that the hit rate advantage for LF words should be reduced by increases in attentional strain at study. We first tested this prediction in two experiments by varying the amount of experimenter-controlled study time (on the basis of the assumption that a decrease in study time would reduce the amount of resources available for studying a word). The elevated-attention hypothesis was confirmed, but only when words were studied for relatively short durations. This finding led us to formulate anearly-phase elevated-attention hypothesis that proposes that more attentional resources are allocated to LF words than to HF words only during the early phase of encoding (which produces the LF hit rate advantage in subsequent recognition) and that the allocation of attentional resources during the late phase of encoding is not greater for LF words than for HF words. An additional empirical test of this revised hypothesis was conducted: Experimenter-controlled study time and the composition of the to-be-remembered pairs of words were varied orthogonally. The results confirmed the early-phase elevated-attention hypothesis.

Turning up the Noise or Turning Down the Volume? On the Nature of the Impairment of Episodic Recognition Memory by Midazolam.

E. Hirshman, J. Fisher, T. Henthom, J. Amdt, and A. Passanname (2002) found that Midazolam disrupts the mirror-patterned word-frequency effect for recognition memory by reversing the typical hit-rate advantage for low-frequency words. They noted that this result is consistent with dual-process accounts (e.g., R. C. Atkinson & J. F. Juola, 1974; G. Mandler, 1980; A. P. Yonelinas, 1994) of the word frequency effect for recognition memory (S. Joordens & W. E. Hockley. 2000; L. M. Reder et al. 2000). The present authors show that this finding is also consistent with a variety of single-process, retrieving effectively- from-memory (REM) models (R. M. Shiffrin & M. Steyvers, 1997), the simplest of which assumes that Midazolam decreases the accuracy with which memory traces are stored. These findings therefore do not discriminate between single- and dual-process models of recognition memory.

A Global Theory of Remembering and Forgetting From Multiple Lists

Forgetting is frustrating, usually because it is unintended. Other times, one may purposely attempt to forget an event. A global theory of recognition and free recall that explains both types of forgetting and remembering from multiple list experiments is presented. The critical assumption of the model is that both intentional and unintentional forgetting are often due to contextual interference. Unintentional forgetting is the natural result of contextual changes between study and test. Intentional forgetting is accomplished by a rapid, metacognitively instigated change in mental context that renders to-be-forgotten information relatively inaccessible and renders to-be-remembered information more accessible (L. Sahakyan & C. M. Kelley, 2002). This occurs for both recognition and free recall. Implications for item-method directed forgetting, exclusion recognition, source memory, and encoding operations are discussed.

List Composition and the Word-Frequency Effect for Recognition Memory

The attention/likelihood theory (ALT; M. Glanzer & J. K. Adams, 1990) and the retrieving effectively from memory (REM) theory (R. M. Shiffrin & M. Steyvers, 1997) make different predictions concerning the effect of list composition on word recognition. The predictions were empirically tested for two-alternative forced-choice, yes-no, and ratings recognition tasks. In the current article, the authors found that discrimination of low-frequency words increased as the proportion of high-frequency words studied increased. The results disconfirm the ALT prediction that recognition is insensitive to list composition, and they disconfirm the predictions of the REM model described by R. M. Shiffrin and M. Steyvers (1997). The current authors discuss a slightly modified version of REM that can better predict our findings, and we discuss the challenges the present findings pose for ALT and REM.

Discriminating between changes in bias and changes in accuracy for recognition memory of emotional stimuli

A debate has emerged as to whether recognition of emotional stimuli is more accurate or more biased than recognition of nonemotional stimuli. Teasing apart changes in accuracy versus changes in bias requires a measurement model. However, different models have been adopted by different researchers, and this has contributed to the current debate. In this article, different measurement models are discussed, and the signal detection model that is most appropriate for recognition is adopted to investigate the effects of valence and arousal on recognition memory performance, using receiver operating characteristic analyses. In addition, complementary two-alternative forced choice experiments were conducted in order to generalize the empirical findings and interpret them under a relatively relaxed set of measurement assumptions. Across all experiments, accuracy was greater for highly valenced stimuli and stimuli with high arousal value. In addition, a bias to endorse positively valenced stimuli was observed. These results are discussed within an adaptive memory framework that assumes that emotion plays an important role in the allocation of attentional resources.