Subjective memorability and the mirror effect.

The mirror effect refers to the common finding that hit and false alarm rates on a recognition test are inversely related. The present research investigated the generality of the mirror effect (to rare words) and tested whether the effect might be grounded in accurate estimates of word memorability. The first 2 experiments showed that although high- and low-frequency words exhibit a mirror effect, rare words do not. Furthermore, contrary to expectations, Ss consistently (and mistakenly) predicted that memorability was directly correlated with frequency of usage. These findings weight against the idea that the mirror effect arises because of a S's ability to reject low-frequency lures on the grounds that such words would have been remembered had they appeared previously. Instead, the rejection of lures from different frequency categories may be determined by their semantic or phonemic overlap with list targets, and an analysis along these lines may help to explain why rare words constitute an exception to the otherwise ubiquitous mirror effect.     

The mirror effect and mixture signal detection theory

The mirror effect for word frequency refers to the finding that low-frequency words have higher hit rates and lower false alarm rates than high-frequency words. This result is typically interpreted in terms of conventional signal detection theory (SDT), in which case it indicates that the order of the underlying old item distributions mirrors the order of the new item distributions. However, when viewed in terms of a mixture version of SDT, the order of hits and false alarms does not necessarily imply the same order in the underlying distributions because of possible effects of mixing. A reversal in underlying distributions did not appear for fits of mixture SDT models to data from 4 experiments.     

The word-frequency paradox in recognition

High-frequency words are recalled better than are low-frequency words, but low-frequency words produce higher hit rates in a recognition test than do high-frequency words. Two experiments provided new date on the phenomenon and also evidence relevant to the dual process model of recognition, which postulates that recognition judgments are a function of increments in item familiarity and of item retrievability. First, recall and recognition by subjects who initially performed a single lexical decision task were compared with those of subjects who also gave definitions of high-, low-, and very low-frequency target words. In the second experiment, subjects initially performed either a semantic, elaborative task or an integrative task that focused attention on the physical, perceptual features of the same words. Both experiments showed that extensive elaborative processing results in higher recall and hit rates but lower false alarm rates, whereas word frequency has a monotonic, linear effect on recall and false alarm rates, but a paradoxical, curvilinear effect on hit rates. Elaboration is apparently more effective when the potential availability of meaningful connections with other structures is greater (as for high-frequency words). The results are consistent with the dual process model.     

Shades of the mirror effect: Recognition of faces with and without sunglasses

A mirror effect was found for a stimulus manipulation introduced at test. When subjects studied a set of normal faces and then were tested with new and old faces that were normal or wearing sunglasses, the hit rate was higher and the false alarm rate was lower for normal faces. Hit rate differences were reflected in remember and sure recognition responses, whereas differences in false alarm rates were largely seen in know and unsure judgments. In contrast, when subjects studied faces wearing sunglasses, the hit rate was greater for test faces with sunglasses than for normal faces, but there was no difference in false alarm rates. These findings are problematic for single-factor theories of the mirror effect, but can be accommodated within a two-factor account.     

Interactions between study task, study time, and the low-frequency hit rate advantage in recognition memory

In studies of episodic recognition memory, low-frequency words (LF) have higher hit rates (HR) and lower false alarm rates (FAR) than do high-frequency words (HF), which is known as the mirror pattern. A few findings have suggested that requiring a task at study may reduce or eliminate the LF-HR advantage without altering the LF-FAR effect. Other studies have suggested that the size of the LF-HR advantage interacts with study time. To explore such findings more thoroughly and relate them to theory, the authors conducted 5 experiments, varying study time and study task. The full mirror pattern was found only in 2 cases: the standard condition requiring study for a later memory test and a condition requiring a judgment about unusual letters. The authors explain their findings in terms of the encoding of distinctive features and discuss the implications for current theories of recognition memory and the word frequency effect.     

Models that allow us to perceive the world more accurately also allow us to remember past events more accurately via differentiation

Differentiation is a theory that originally emerged from the perception literature and proposes that with experience, the representation of stimuli becomes more distinct from or less similar to the representation of other stimuli. In recent years, the role of differentiation has played a critical role in models of memory. Differentiation mechanisms have been implemented in episodic memory models by assuming that information about new experiences with a stimulus in a particular context accumulates in a single memory trace and these updated memory traces become more distinct from the representations of other stimuli. A key implication of such models is that well encoded events are less confusable with other events. This prediction is particularly relevant for two important phenomena. One is the role of encoding strength on memory. The strength based mirror effect is the finding of higher hit rates and lower false alarm rates for a list composed of all strongly encoded items compared to a list composed of all weakly encoded items. The other is output interference, the finding that accuracy decreases across a series of test trials. Results from four experiments show a tight coupling between these two empirical phenomena such that strongly encoded target items are less prone to interference. By proposing a process model and evaluating the predictions of the model, we show how a single theoretical principle, differentiation, provides a unified explanation for these effects.     

The Role of Recollection and Familiarity in the Context Variability Mirror Effect

Context variability refers to the number of preexperimental contexts that are associated with concepts. In four experiments, we investigated the basis for increased recognition memory for low context variability words. Low context variability was associated with greater recollection in the hit rates, and high context variability was associated with greater familiarity in the false alarms. Shortening the study time reduced recollection, but low context variability still influenced recollection in the hit rates. A modality change from study to test also reduced recollection but preserved recollective differences for low versus high context variability items. One interpretation of the results suggests that low context variability evokes more specific and, perhaps, idiosyncratic recollective associations during learning and that these associations support better recognition in the hit rates. By contrast, activating the larger number of associations for high context variability items may be mistaken for familiarity in the false alarm rates.     

Modeling the effects of verbal and nonverbal pair strength on associative recognition

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).     

The variance reaction time model

The variance reaction time model (VRTM) is proposed to account for various recognition data on reaction time, the mirror effect, receiver-operating-characteristic (ROC) curves, etc. The model is based on simple and plausible assumptions within a neural network: VRTM is a two layer neural network where one layer represents items and one layer represents contexts. The recognition decision is based on a random walk of nodes activated at recognition. VRTM suggests theoretical constraints on the distributions of nodes activated at recognition and the noise in the random walk. The variability in the net inputs to nodes depends on the item frequency (the number of times that the item has been encoded) and the list length. The essential mechanism that accounts for the empirical data is a non-linear activation function. The mean activation threshold in the non-linear activation function is placed to achieve efficient discriminability between new and old items and there is variability in the activation threshold. VRTM predicts the mirror effect for low and high frequency words, a strength based mirror effect between conditions but not within one condition, appropriate ROC-curves for old/new and high/low frequency items, and list-length effects. Furthermore, it predicts appropriate means and distributions of reaction times for old/new, correct/incorrect, and high/low frequency items as well as speed/accuracy tradeoffs. VRTM has an explicit mathematical solution, it is simulated in a neural network, and it is fitted to a number of datasets.     

Not all sources of familiarity are created equal: the case of word frequency and repetition in episodic recognition

Low-frequency (LF) words produce higher hit rates and lower false alarm rates than high-frequency (HF) words. This word frequency mirror pattern has been interpreted within dual-process models of recognition, which assume the contributions of a slower recollective process and a relatively fast-acting familiarity process. In the present experiments, recollection and familiarity were placed in opposition using Jacoby, L. L., Journal of Memory and Language, 30, 513–541 (1991), two-list exclusion paradigm with HF and LF words. Exclusion errors to LF words exceeded those to HF words at fast deadlines, whereas the reverse occurred at slow deadlines. In Experiments 2 and 3, false alarms to HF nonpresented lures were higher than to LF nonpresented lures, indicating the use of baseline familiarity for totally new items. Combined, these results indicate that in addition to baseline familiarity and recollection, a third process involving the assessment of a relative change in familiarity is involved in recognition performance. Both relative changes in familiarity and recollection processes have distinct time courses and are engaged when there is diagnostic list information available, whereas baseline familiarity is used when there is no diagnostic information available (e.g., for nonpresented lure items).     

Dissociative effects of phonological vs. semantic associates on recognition memory in the Deese/Roediger-McDermott paradigm

Using a variation of the Deese/Roediger–McDermott paradigm in four experiments, we examined hit rate, false alarm rate, and memory discriminability (d′) for critical items (e.g., sleep) after their semantic associates (e.g., dream, rest, and awake), phonological associates (e.g., bleep, sheep, and cheap) or unrelated items were studied. Replicating previous research (e.g., Neely & Tse, 2007), d′ was lower for critical items than for yoked associates in semantic lists. However, d′ was higher for critical items than for yoked associates in phonological lists, even when the same critical items were used for these two list types. The memory discriminability of critical items was enhanced by phonological lists via a larger increase in their hit rates than in their false alarm rates, relative to the same critical items in lists of unrelated words. These findings (a) could not be fully accounted for by the activation-based theories proposed in the false memory literature, and (b) may suggest that the memory discriminabilities of semantic and phonological lists are modulated by two distinct activation mechanisms.     

A grouping interpretation of the modality effect in immediate probed recognition

In a series of experiments on immediate probed recognition for eight 3-digit numbers, it was shown that if the target modality involved auditory components and the effect of the similarity of the modality of the probe to that of the targets was controlled, unequivocal evidence was obtained for an auditory superiority effect (modality effect) for hit rates for the final items of the list. Moreover, false-alarm rates were significantly lower following targets with an auditory component than they were following silently seen targets. It is argued that this pattern of hits and false alarms is consistent with the idea that targets that have an auditory component yield memory representations that are better grouped as units than are those for targets that are only silently seen; in particular, if a new probe has a first digit that accidentally matches the first digit of a target item, it is more likely that the subject will mistakenly identify this new probe as old (give a false alarm) if the target has only been partially encoded because it was only silently seen.     

Single-item recognition following a verbal discrimination study trial

A verbal discrimination (VD) study phase was followed by a single-item (i.e., one item at a time) test phase. Prior right (R) items from the VD list were found to have higher old-new hit rates on the test phase than prior wrong (W) items. However, prior W items had higher hit rates than prior R items in terms of recognition of their intrapair function. These results were interpreted in terms of a high criterion model whereby an old test item must prossess a large number of frequency units before it is identified as being R. In addition, false recognition effects were demonstrated for new test items that were homophones of prior R items, but not for new test items that were homophones of prior W items. These results were interpreted in terms of a feature analytic extension of frequency theory.     

Judgments of recency and their relation to recognition memory

Subjects went through a list of 550 high- and low-frequency words (Experiment 1) or concrete and abstract words (Experiment 2) in which individual items were repeated at lags of 5 to 30 other items. They made old versus new recognition decisions on each word and followed each "old" response with a numerical judgment of recency (JOR). Recognition judgments displayed the mirror effect. Conditionalized on recognition, JORs were shorter for low-frequency words than for high-frequency words, and shorter for concrete words than for abstract words. This was true at every lag, suggesting that recognition and JOR may have a common basis. However, recognition confidence ratings obtained in Experiment 3 proved much less sensitive than JOR to test lag. Memory models applicable to multiple judgment tasks will be needed to account for such findings.     

Paradoxical effects of presentation modality on false memory.

Two experiments are reported in which the effects of presentation modality on false memory in recall and recognition are studied. False recall and recognition of critical targets are lower for non-presented items related to a study list when that study list is presented visually than when presented auditorily. This pattern of low levels of false memory for critical targets holds even when participants read the visually presented study items aloud. These results suggest that recollection of visual detail plays a role in the prevention of false memory. However, both the hit rates (true memory) and the false-alarm rates to weakly related distractors (non-critical targets) were higher for visual presentation than for auditory presentation, suggesting that more than one mechanism may underlie false recognition.     

Interrupting recognition memory: Tests of familiarity-based accounts of the revelation effect

The revelation effect is a puzzling phenomenon in which items on a recognition test are more likely to be judged as "old" when they are immediately preceded by a problem-solving task, such as anagram solution. The present experiments were designed to evaluate Westerman and Greene's (1998) and Hicks and Marsh's (1998) familiarity-based accounts of this effect. We found comparable revelation effects when probes were preceded by an anagram or a numerical addition task and when subjects performed either one or two of these tasks. Taken together, the results do not support familiarity-based accounts of the revelation effect but are consistent with a proposed decision-based interpretation (i.e., criterion flux), in which it is assumed that the revelation task displaces the study list context in working memory, leading subjects to adopt a more liberal recognition decision criterion, thereby increasing the hit and false alarm rates.     

On the nature of the decision axis in signal-detection-based models of recognition memory

Most models of recognition memory involve a signal-detection component in which a criterion is placed along a decision axis. Older models generally assume a familiarity-decision axis, but newer models often assume a likelihood ratio axis instead because it allows for a more natural account of the ubiquitous mirror effect. In 3 experiments reported here, item strength was differentially manipulated to see whether a mirror effect would occur. Within a list, the items from 1 category were strengthened by repetition, but the items from another category were not. On the subsequent recognition test, the hit rate was higher for the strong category, but the false-alarm rates for the weak and strong categories were the same (i.e., no mirror effect was observed). This result suggests that the decision axis represents a familiarity scale and that participants adopt a single decision criterion that they maintain throughout the recognition test.     

On the nature of associative information in recognition memory

In a typical associative-recognition task, participants must distinguish between intact word pairs (both words previously studied together) and rearranged word pairs (both words previously studied but as part of different pairs). The familiarity of the individual items on this task is uninformative because all of the items were seen before, so the only way to solve the task is to rely on associative information. Prior research suggests that associative information is recall-like in nature and may therefore be an all-or-none variable. The present research reports several experiments in which some pairs were strengthened during list presentation. The resulting hit rates and false alarm rates, and an analysis of the corresponding receiver operating characteristic plots, suggest that participants rely heavily on item information when making an associative-recognition decision (to no avail) and that associative information may be best thought of as a some-or-none variable.     

A bias in favor of the positive response to high-frequency words in recognition memory

A relatively liberal response bias for high-frequency words and a violation of the mirror effect for hit and false-alarm rates were found in a yes-no recognition-memory test. Subjects more frequently responded "old" to high-frequency words than to low-frequency words. Four experiments were conducted to determine the causes of the different response biases and of the violation of the mirror effect. The word-frequency effect on hit rates did not appear, whereas the false-alarm rate for low-frequency words was lower than that for high-frequency words. When low- or high-frequency words were presented separately in distinct halves of a recognition-memory test, the relatively liberal response bias for high-frequency words was diminished. A model for recognition judgment is proposed that assumes the use of a common criterion for low- and high-frequency words.     

错误记忆的产生是否依赖对词表的有意加工：无意识激活的证据

使用DRM范式考察了错误记忆产生过程中无意识激活的存在与否。通过操纵实验任务和呈现方式,导致被试对学习词表的无意加工和不同水平的有意加工。结果发现,无意加工时所产生的对关键诱饵的无意识激活足以导致错误记忆效应,说明对词表的有意加工并非错误记忆产生所必需。随着对词表有意加工水平的提高,多次无意识激活的累积可导致增强的错误记忆效应,且表现出与真实记忆的共变,而阻断连续激活累积的因素可有效降低错误记忆