Distinctiveness, not dual coding, explains the picture-superiority effect.

The picture-superiority effect is the finding that memory for pictures exceeds memory for words on many tasks. According to dual-coding theory, the pictures' mnemonic advantage stems from their greater likelihood to be labelled relative to words being imaged. In contrast, distinctiveness accounts hold that the greater variability of pictures compared to words leads to their mnemonic advantage. Ensor, Surprenant, et al. tested these accounts in old/new and forced-choice recognition by increasing the physical distinctiveness of words and decreasing the physical distinctiveness of pictures. Half of the words were presented in standard black font, and half were presented in varying font styles, font sizes, font colours, and capitalisation patterns. Half of the pictures were presented in black and white and half in colour. Consistent with the physical-distinctiveness account but contrary to the dual-coding account, the picture-superiority effect was eliminated when comparing the black-and-white pictures to distinctive words. In the present study, we extend Ensor, Surprenant, et al.'s results to associative recognition and free recall. Results were consistent with physical distinctiveness. We argue that dual-coding theory is no longer a viable explanation of the picture-superiority effect.

The orthographic/phonological neighbourhood size effect and set size.

A growing number of studies have shown that on serial recall tests, words with more orthographic/phonological neighbours are better recalled than otherwise comparable words with fewer neighbours, the so-called neighbourhood size effect. Greeno et al. replicated this result when using a large stimulus pool but found a reverse neighbourhood size effect-better recall of words with fewer rather than more neighbours-when using a small stimulus pool. We report three registered experiments that further examine the role of set size in the neighbourhood size effect. Experiment 1 used the large pool from Greeno et al. and replicated their finding of a large-neighbourhood advantage. Experiment 2 used the small pool from Greeno et al. but found no difference in recall between the large and small neighbourhood conditions. Experiment 3 also used a small pool but the small pool was randomly generated for each subject from the large pool used in Experiment 1. This resulted in a typical large neighbourhood advantage. We suggest that set size is not critical to the direction of the neighbourhood size effect, with a large neighbourhood advantage appearing with both small and large pools.

Set size and the orthographic/phonological neighbourhood size effect in serial recognition: The importance of randomization.

The neighbourhood size effect refers to the finding of better memory for words with more orthographic/phonological neighbours than otherwise comparable words with fewer neighbours. Although many studies have replicated this result with serial recall, only one has used serial recognition. Greeno et al. (2022) found no neighbourhood size effect when a large stimulus pool was used and a reverse effect-better performance for small neighbourhood words-when a small stimulus pool was used. We reexamined these results but made two methodological changes. First, for the large pool, we randomly generated lists for each subject rather than creating one set of lists that all subjects experienced. Second, for the small pool, we randomly generated a small pool for each subject rather than using one small pool for all subjects. In both cases, we observed a neighbourhood size effect consistent with results from the serial recall literature. Implications for methodology and theoretical accounts of both the neighbourhood size effect and serial recognition are discussed. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Valence and concreteness in item recognition: Evidence against the affective embodiment account.

The Affective Embodiment Account posits that sensorimotor interactions play an important role in learning and processing concrete words whereas experiences from emotional states play an important role in learning and processing abstract words. Because of this, there should be greater enhancement of valence for abstract than for concrete words and therefore there should be an interaction between valence and concreteness. Although this prediction has been observed in a number of tasks, very few studies have looked specifically at memory. Three experiments are reported that assess whether valence interacts with concreteness in recognition. In Experiment 1, recognition of concrete words was better than abstract, but there was no difference as a function of whether the words were positive or negative and there was no interaction. Experiment 2 compared positive and neutral words and Experiment 3 compared negative and neutral words; in both, there was a concreteness effect but no effect of valence and no interaction. These results replicate previous findings that when positive and negative words are equated more fully, valence has no effect on recognition, and also suggest a limit on the scope of the Affective Embodiment Account.

Irrelevant speech, changing state, and order information.

The changing state effect is the finding that a stream of irrelevant sounds that change more (e.g., different digits in random order) disrupts memory more than a stream of irrelevant sounds that change less (e.g., a single digit repeated over and over). According to the Object-Oriented Episodic Record (O-OER) model, the changing state effect will be observed only in memory tasks that have an order component or which induce serial rehearsal or serial processing. In contrast, other accounts-including the Feature Model, the Primacy Model, and various attentional theories-predict that the changing state effect should be observable when there is no order component. Experiment 1 first demonstrated that the irrelevant stimuli created for the current experiments produced a changing state effect in immediate serial recall in both on-campus and online samples. Then, three experiments assessed whether a changing state effect is observable in a surprise 2AFC recognition test. Experiment 2 replicated Stokes and Arnell (2012, Memory & Cognition, 40, 918-931), who found that although irrelevant sounds reduce performance on a surprise recognition test of words presented previously in a lexical decision task, they do not produce a changing state effect. Experiments 3 and 4 used two different encoding tasks (pleasantness and frequency judgment) and also found no changing state effect. The results support the prediction of the O-OER model and provide additional evidence against the other accounts.

The block order effect in reconstruction of order tasks and metacognitive processing.

In serial reconstruction of order tasks, high-frequency words are better remembered than otherwise equivalent low-frequency words. Neath and Quinlan (2021) found that although the usual high-frequency advantage was observed when subjects received a block of low-frequency lists first followed by a block of high-frequency lists, there was no frequency effect when subjects received a block of high-frequency lists followed by a block of low-frequency lists. In order to assess whether the block order effect simply reflects the inherent changeability of frequency effects, we manipulated concreteness, a much more stable effect. Experiment 1 found a block order effect with concreteness: The usual advantage for concrete over abstract words was observed only when the abstract block came first and the concrete block second; when the block order was reversed, no concreteness effect was seen. In Experiment 2, subjects did not know whether the test would be serial reconstruction of order or immediate serial recall until after list presentation. This eliminated the block order effect, just as when frequency was manipulated. Experiment 3 found a block order effect with a free reconstruction of order task and with both open and closed stimulus sets. Given that the pattern of results with concreteness is the same as with frequency, it suggests the block order effect is not unique to frequency and that a more general explanation, such as a metacognitive account, is needed. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Using preexperimental familiarity to compare the ICE and cue-overload accounts of context-dependent memory in item recognition.

In recognition, context effects often manifest as higher hit and false-alarm rates to probes tested in an old context compared with probes tested in a new context; sometimes, this concordant effect is accompanied by a discrimination advantage. According to the cue-overload account of context effects (Rutherford, 2004), context acts like any other cue, and thus context effects should be larger with lighter context loads. Conversely, the Item, Associated Context, and Ensemble (ICE) account (Murnane et al., 1999) attributes context effects to two factors: subjects erroneously attributing context familiarity to the probe, and the formation of ensembles (mnemonic combinations of item and context). Context familiarity increases as exposure at study increases, and thus ICE predicts larger effects of context as context load increases. Relatedly, ICE predicts larger effects of context as context meaningfulness increases, as meaningful contexts are more likely to be bound to the target in an ensemble. In Experiments 1 and 2, rather than manipulate context load during the study phase, we relied on subjects' preexperimental context exposure to manipulate context load. Subjects studied words superimposed on photographs of their university campus or another university campus. At test, targets and distractors were evenly divided between study and novel contexts and between familiar and unfamiliar contexts. In Experiment 3, we manipulated context familiarity within the experimental session. Results supported ICE, suggesting that context does not act as a retrieval cue in recognition. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Additional evidence that valence does not affect serial recall.

In immediate serial recall, a canonical short-term memory task, it is well established that performance is affected by several sublexical, lexical, and semantic factors. One factor that receives a growing interest is valence, whether a word is categorised as positive (e.g., happy) or as negative (e.g., pain). However, contradictory findings have recently emerged. Tse and Altarriba in two experiments with one set of stimuli and fixed lists concluded that valence affects serial recall performance, while Bireta et al. in three experiments with three sets of stimuli and randomised lists concluded that valence does not affect serial recall performance. Two experiments assessed the experimental discrepancy between Tse and Altarriba and Bireta et al. For both experiments, in one block, every participant saw the exact same lists as those used in Tse and Altarriba, and in the other block, each list was randomly constructed for each participant, as was done in Bireta et al. In Experiment 1, with concrete words varying in valence, we replicated the results of Tse and Altarriba with fixed lists and the results of Bireta et al. with randomised lists. In Experiment 2, with abstract words with both fixed and randomised lists, we replicate the absence of effect valence like Tse and Altarriba and Bireta et al. Overall, we conclude that valence does not affect serial recall and the discrepancy was attributed to the peculiarity of the fixed lists used by Tse and Altarriba.

Stimulus-based mirror effects in associative recognition revisited.

The mirror effect, the finding that a manipulation which increases the hit rate in recognition tests also decreases the false alarm rate, is held to be a regularity of memory. Neath et al. (in press) took advantage of the recent increase in the number of linguistic databases to create sets of stimuli that differed on one dimension but were more fully equated on other dimensions known to affect memory. Using these highly controlled stimulus sets, no mirror effects were observed; in contrast, using stimulus sets that had confounds resulted in mirror effects. In this article, we use their stimulus sets to examine associative recognition. Using confounded stimuli, Experiment 2 found a lower false alarm rate for high- compared to low-frequency words, replicating previous results, and Experiment 4 found a mirror effect when manipulating concreteness, also replicating previous results. Using highly controlled stimuli, Experiment 1 found no evidence that frequency affected associative recognition, and Experiment 3 found concreteness affected only the hit rate, not the false alarm rate. When highly controlled stimuli are used, frequency affects only the false alarm rate in item recognition and has no effect in associative recognition, whereas concreteness affects hit rates in both item and associative recognition. Implications for theoretical accounts are discussed. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Valence does not affect recognition.

Valence refers to the extent to which a stimulus is viewed as negative or positive. One recent model of valence, the NEVER model (Bowen et al., 2018), predicts that in general negative words will be better remembered than positive or neutral words. However, this prediction is difficult to validate for recognition tests because the literature reports inconsistent findings. Three experiments reexamined whether valence affects recognition of words by taking advantage of the recent increase in the number of high-quality norms and databases, which allow for the construct ion of three sets of stimuli that differ in valence, but are equated on numerous other dimensions known to affect memory. Experiment 1 found no difference in recognition performance between positive and negative words; Experiment 2 found no difference between positive and neutral words; and Experiment 3 found no difference between neutral and negative words. The results disconfirm a prediction of the NEVER model and suggest that previous demonstrations of an effect of valence are due to confounding other dimensions with valence. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Stimulus-based mirror effects revisited.

The mirror effect is the finding that in recognition tests, a manipulation that increases the hit rate also decreases the false alarm rate. For example, low frequency words have a higher hit rate and a lower false alarm rate than high frequency words. Because the mirror effect is held to be a regularity of memory, it has had a pronounced influence on theories of recognition. We took advantage of the recent increase in the number of linguistic databases to create sets of stimuli that differed on one dimension (contextual diversity, frequency, or concreteness) but were more fully equated on other dimensions known to affect memory. Experiment 1 (contextual diversity), Experiment 3 (frequency), and Experiment 5 (concreteness) found no evidence of a mirror effect. We also conducted parallel experiments which used previously published stimuli that could not avail of the new databases and which therefore contained confounds. Experiment 2 (contextual diversity), Experiment 4 (frequency), and Experiment 6 (concreteness) all resulted in mirror effects. If this pattern of results is replicable, it has broad implications for theories of recognition, which typically view the mirror effect as a benchmark finding. Unfortunately, few articles on the mirror effect include the stimuli, rendering the past literature of little use in testing this hypothesis. We encourage researchers to create and assess other pools of highly controlled stimuli to establish whether the stimulus-based mirror effect obtains when confounds are eliminated or whether it is due to the presence of these confounds. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Memory without retrieval: Testing the direct-access account of the missing item task.

In the missing item task, two short lists are presented. The test list contains all but one of the items from the study list in a new random order and the task is to report which item from the study list is missing. Murdock and Smith (2005) found that the time to correctly respond with the missing item was independent of the position of the missing item and was also independent of the list length. They argued that these data are difficult to accommodate by models that include a search process but are consistent with models that posit "direct access" such as the power set version of Theory of Distributed Associative Memory (TODAM). If direct access is occurring, redintegration cannot be occurring. Two experiments test the direct access account by determining whether two effects commonly ascribed to redintegration occur in the missing item task. Experiment 1 found a semantic relatedness effect and Experiment 2 found a word frequency effect. The presence of these effects is consistent with a redintegration account. Implications for TODAM and for an explanation based on the Feature Model are discussed. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Valence does not affect serial recall.

Despite being the prototypical test of short-term/working memory, immediate serial recall is affected by numerous lexical and long-term memory factors. Within this large literature, very few studies have examined whether performance on the task is affected by valence, the extent to which a word is viewed as positive or negative. Whereas the NEVER model (Bowen, Kark, & Kensinger, 2018) makes the general prediction that negative words will be remembered better than positive words, two previous studies using serial recall have reported that positive words are better remembered than negative words. Three experiments reassessed whether valence affects immediate serial recall using stimuli equated on multiple dimensions, including both arousal and dominance. Over the 3 experiments, with 3 different sets of stimuli, we found no differences in either accuracy or various error measures as a function of valence. The data suggest that there is no effect of valence on an immediate serial recall task when potentially confounding dimensions are controlled. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Calculating semantic relatedness of lists of nouns using WordNet path length.

Lists of semantically related words are better recalled on immediate memory tests than otherwise equivalent lists of unrelated words. However, measuring the degree of relatedness is not straightforward. We report three experiments that assess the ability of various measures of semantic relatedness-including latent semantic analysis (LSA), GloVe, fastText, and a number of measures based on WordNet-to predict whether two lists of words will be differentially recalled. In Experiment 1, all measures except LSA correctly predicted the observed better recall of the related than the unrelated list. In Experiment 2, all measures except JCN predicted that abstract words would be recalled equally as well as concrete words because of their enhanced semantic relatedness. In Experiment 3, LSA, GLoVe, and fastText predicted an enhanced concreteness effect because the concrete words were more related; three WordNet measures predicted a small concreteness effect because the abstract and concrete words did not differ in semantic relatedness; and three other WordNet measures predicted no concreteness effect because the abstract words were more related than the concrete words. A small concreteness effect was observed. Over the three experiments, only two measures, both based on simple WordNet path length, predicted all three results. We suggest that the results are not unexpected because semantic processing in episodic memory experiments differs from that in reading, similarity judgment, and analogy tasks which are the most common way of assessing such measures.

The item/order account of word frequency effects: Evidence from serial order tests.

According to the item/order hypothesis, high-frequency words are processed more efficiently and therefore order information can be readily encoded. In contrast, low-frequency words are processed less efficiently and the focus on item-specific processing compromises order information. Most experiments testing this account use free recall, which has led to two problems: First, the role of order information is difficult to evaluate in free recall, and second, the data from free recall show all three possible patterns of results: memory for high-frequency words can be better than, the same as, or worse than that for low-frequency words. A series of experiments tested the item/order hypothesis using tests where the role of order information is less ambiguous. The item/order hypothesis predicts better performance for high- than low-frequency words when pure lists are used in both immediate serial recall (ISR) and serial reconstruction of order (SRO) tests. In contrast, when mixed (alternating) lists are used, it predicts better performance for low- than for high-frequency words with ISR tests, but equivalent performance with SRO tests. The experiments generally confirm these predictions, with the notable exception of a block order effect in SRO tasks: When a block of low-frequency lists preceded a block of high-frequency lists, a high-frequency advantage was observed but when a block of high-frequency lists preceded a block of low-frequency lists, no frequency effect was observed. A final experiment provides evidence that this block order effect is due to metacognitive factors.

Re-assessing age of acquisition effects in recognition, free recall, and serial recall.

Age of acquisition (AoA) refers to the age at which a person learns a word. Research has converged on the conclusion that early AoA words are processed more efficiently than late AoA words on a number of perceptual and reading tasks. However, only a few studies have investigated whether AoA affects memory on recognition, serial recall, and free recall tests, and the results are equivocal. We took advantage of the recent increase in the number of high-quality norms and databases to construct a pool of early and late AoA words that were equated on numerous other dimensions. There was a late AoA advantage in recognition using both pure (Experiment 1) and mixed (Experiment 2) lists, no effect of AoA on serial recall of either pure (Experiment 3) or mixed (Experiment 4) lists, and no effect of AoA on free recall of either pure (Experiment 5) or mixed lists (Experiment 6). We conclude that AoA does reliably affect memory on some memory tasks (recognition), but not others (serial recall, free recall), and that no current account of AoA can explain the findings.

Modeling list-strength and spacing effects using version 3 of the retrieving effectively from memory (REM.3) model and its superimposition-of-similar-images assumption.

Shiffrin and Steyvers (1997) introduced a model of recognition memory called retrieving effectively from memory (REM) and successfully applied it to a number of basic memory phenomena. REM incorporates differentiation, wherein item repetitions are accumulated in a single mnemonic trace rather than separate traces. This allows REM to account for several benchmark findings, including the null list-strength effect in recognition (Ratcliff, Clark, & Shiffrin, 1990). The original REM treated massed and spaced repetitions identically, which prevents it from predicting a mnemonic advantage for spaced over massed repetitions (i.e., the spacing effect). However, Shiffrin and Steyvers discussed the possibility that repetitions might be represented in a single trace only if the subject identifies that the repeated item was previously studied. It is quite plausible that subjects would notice repetitions more for massed than for spaced items. Here we show that incorporating this idea allows REM to predict three important findings in the recognition memory literature: (1) the spacing effect, (2) the finding of slightly positive list-strength effects with spaced repetitions, as opposed to massed repetitions or increased study time, and (3) list-strength effects that have been observed using very large strong-to-weak ratios (see Norman, 2002).

Dynamic Visual Noise Does Not Affect Memory for Fonts.

Dynamic visual noise (DVN) selectively impairs memory for some types of stimuli (e.g., colors, textures, concrete words), but not for others (e.g., matrices, Chinese characters, simple shapes). According to the image definition hypothesis, the key difference is whether the stimulus leads to images that are ill-defined or well-defined. The former will be affected because the addition of noise quickly reduces the usefulness of the image in supplying information about the item's identity. The image definition hypothesis predicts that fonts should lead to ill-defined images and therefore should be affected by DVN, and although three previous studies appear to show this result, they lack a key control condition and report only proportion correct. Two experiments reassessed whether DVN affects memory for fonts, but, unlike the previous studies, both included a static visual noise condition and both were analyzed using signal detection measures. There was no evidence that DVN affected memory for font information, thus disconfirming a prediction of the original version of image definition hypothesis. We suggest a revised version that focuses on redintegration can explain the results.

Concreteness and disagreement: Comment on Pollock (2018).

Pollock (Behavior Research Methods doi:10.3758/s13428-017-0938-y, Pollock, 2018) points out that most memory experiments using abstract and concrete words have a potential confound: Raters express more disagreement, on average, about the rating for an abstract word than for a concrete word, as evidenced by the larger standard deviation of the rating (SDR). Therefore, past demonstrations of the concreteness effect could be explained by the disagreement hypothesis: Words that engender disagreement (i.e., have a larger SDR) are more difficult to remember than those that engender agreement (i.e., have a smaller SDR). Three experiments test predictions of the disagreement hypothesis. In Experiment 1, concreteness (abstract vs. concrete) and SDR size (small vs. large) were factorially manipulated. A concreteness effect was observed for both SDR sizes, but there was no effect of SDR and there were no interactions involving SDR. In Experiment 2, a concreteness effect was observed despite using abstract words with a small SDR and concrete words with a large SDR, the opposite of what the disagreement hypothesis predicts. Experiment 3 replicated Experiment 2 but with a larger set of stimuli. The results offer no support for the disagreement hypothesis.

Dynamic visual noise affects ill-defined, not well-defined, images.

Dynamic visual noise (DVN), a matrix of squares that randomly alternate between black and white, interferes with some but not all tasks that involve visuo-spatial processing. Although visual working memory is generally invoked to explain the detrimental effects of DVN, the failure of DVN to impair memory for some stimuli that should be processed via visual working memory has not been satisfactorily explained. The image-definition hypothesis proposes that DVN will impair only ill-defined, not well-defined, images. We report five experiments that test this hypothesis. Experiments 1 and 2 use stimuli that lead to well-defined visual images (matrices, photographs of common objects) and no effect of DVN was observed. In contrast, Experiments 3 and 4 use stimuli that lead to ill-defined visual images (textures, photographs of snowflakes) and DVN affected performance. Experiment 5 demonstrated that a potentially disconfirmatory result in the literature was due to analysing proportion correct rather than a measure of discriminability. These results offer initial support for the image-definition hypothesis, and we discuss the implications for theories explaining DVN.