From association to gist: Some critical tests.

We report the first evidence that the gist mechanism of fuzzy-trace theory and the associative mechanism of activation monitoring theory operate in parallel, in the recall version of the Deese/Roediger/McDermott illusion. In three experiments, we implemented a new methodology that allows their respective empirical indexes, gist strength (GS) and backward associative strength (BAS), to each be manipulated while the other is held constant. In Experiment 1, increasing GS increased false recall of missing words, but increasing BAS did not. In Experiments 2 and 3, however, increasing GS and increasing BAS both increased recall of missing words, and those effects were independent and additive. In all three experiments, GS and BAS affected true recall of list words in qualitatively different ways: (a) Increasing GS always improved true recall, regardless of whether BAS was high or low, but (b) increasing BAS impaired true recall when GS was high and improved true recall when GS was low. To pinpoint the retrieval loci of the two variables' effects, we analyzed the data of all experiments with the dual-retrieval model. Those analyses showed that the variables' respective effects were due to different retrieval processes. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

The big three: Accuracy, organization, and retrieval effects of latent semantic attributes.

Rating norms for semantic attributes (e.g., concreteness, dominance, familiarity, and valence) are widely used in many psychological literatures to study the effects of processing specific types of semantic content. Word and picture norms for many attributes are available for thousands of items, but there is a contamination problem in experimentation. When an attribute's ratings are varied, how the semantic content that people process changes is unclear because ratings of individual attributes are correlated with ratings of so many other attributes. To solve this problem, the psychological space that 20 attributes occupy has been mapped, and factor score norms have been published for the latent attributes that generate that space (emotional valence, age-of-acquisition, and symbolic size). These latent attributes have yet to be manipulated in experimentation, and hence, their effects are unknown. We conducted a series of experiments that focused on whether they affect accuracy, memory organization, and specific retrieval processes. We found that (a) all three latent attributes affected recall accuracy, (b) all three affected memory organization in recall protocols, and (c) all three affected direct verbatim access, rather than reconstruction or familiarity. The memory effects of two of them (valence and age-of-acquisition) were unconditional, but memory effects were only detected for the third at particular levels of the other two. The key implications are that semantic attributes can now be cleanly manipulated, and when they are, they have broad downstream effects on memory. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Memory effects of semantic attributes: A method of controlling attribute contamination.

Rating norms for semantic attributes (e.g., concreteness, familiarity, valence) are widely used to study the content that people process as they encode meaningful material. Intensity ratings of individual attributes have been manipulated in numerous experiments with a range of memory paradigms, but those manipulations are contaminated by substantial correlations with the intensity ratings of other attributes. A method of controlling such contamination is needed, which requires a determination of how many distinct attributes there are among the large collection of attributes for which published norms are available. Identification of overlapping words in multiple rating projects yielded a data base containing normed values for each word's perceived intensity (M rating) and ambiguity (rating SD) on 20 different attributes. Principal component analyses then revealed that the intensity space was spanned by just three latent semantic attributes, and the ambiguity space was spanned by five. Psychologically, the big three intensity factors (emotional valence, size, age) were highly interpretable, as were the big five ambiguity factors (discrete emotion, emotional valence, age, meaningfulness, and verbatim memory). We provide a data base of intensity and ambiguity factor scores that can be used to conduct uncontaminated studies of the memory effects of the intensity and ambiguity of latent semantic attributes.

How does attribute ambiguity improve memory?

The memory effects of semantic attributes (e.g., concreteness, familiarity, valence) have long been studied by manipulating their average perceived intensities, as quantified in word rating norms. The semantic ambiguity hypothesis specifies that the uncertainty as well as the intensity of semantic attributes is processed when words are encoded. Testing that hypothesis requires a normed measure of ambiguity, so that ambiguity and intensity can be manipulated independently. The standard deviation (SD) of intensity ratings has been used for that purpose, which has produced three characteristic ambiguity effects. Owing to the recency of such research, fundamental questions remain about the validity of this method of measuring ambiguity and about its process-level effects on memory. In a validity experiment, we found that the rating SDs of six semantic attributes (arousal, concreteness, familiarity, meaningfulness, negative valence, positive valence) passed tests of concurrent and predictive validity. In three memory experiments, we found that manipulating rating SDs had a specific effect on retrieval: It influenced subjects' ability to use reconstructive retrieval to recall words. That pattern was predicted by the current theoretical explanation of how ambiguity benefits memory.

A fundamental asymmetry in human memory: Old ≠ not-new and new ≠ not-old.

In recognition memory, anything that is objectively new is necessarily not-old, and anything that is objectively old is necessarily not-new. Therefore, judging whether a test item is new is logically equivalent to judging whether it is old, and conversely. Nevertheless, a series of 10 experiments showed that old? and new? judgments did not produce equivalent recognition accuracy. In Experiments 1-4, wherein subjects made old? or new? judgments about test items, new? judgments yielded more accurate performance for old items than old? judgments did, and old? judgments yielded more accurate performance for new items than new? judgments did. This same violation of logical equivalence was observed in Experiments 5-10, wherein subjects made similar? judgments as well as old? and new? ones. In short, old? and new? judgments displayed consistent Judgment × Item crossovers, rather than equivalence. Response latencies were used to test the hypothesis that Judgment × Item crossovers were due to certain judgment-item combinations provoking more deliberate, thorough retrieval than other combinations. There was no support for that hypothesis, but the data were consistent with an earlier theory, which posits that latency depends on the extent to which judgments or items slant retrieval toward accessing verbatim traces. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Emotional ambiguity and memory.

The emotional ambiguity hypothesis introduced the principle that uncertainty about items' valence determines how emotional content affects memory and other psychological processes. It was formulated to explain why correlations between the perceived valence and arousal of memory items range from weak to unreliable, but it also makes novel predictions. Although data are consistent with those predictions, the hypothesis does not provide a process model of how valence ambiguity causes the valence-arousal relation to fluctuate. We tested 2 such models-a quantitative one, which assumes that increasing ambiguity lowers the reliability of valence judgments, and a categorical/quantitative one, which assumes that increasing ambiguity restricts the range of valence judgments. These models predict different mathematical relations between measures of ambiguity and intensity for valence and other semantic attributes (e.g., arousal, concreteness, familiarity, imagery, meaningfulness). In Experiments 1-3, tests of those predictions favored the categorical/quantitative model-showing that ambiguity is an inverted-U function for valence and other attributes. Experiments 4 and 5 were designed to investigate whether the memory effects of valence ambiguity are similar to the known effects of valence intensity. In both experiments, recall improved when ambiguity was increased, as well as when intensity was increased. A mathematical model revealed that increases in ambiguity produced large increases in items' familiarity, whereas increases in intensity produced smaller increases in both recollection and familiarity. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

From association to gist.

We removed a key uncertainty in the Deese/Roediger/McDermott (DRM) illusion. The mean backward associative strength (MBAS) of DRM lists is the best-known predictor of this illusion, but it is confounded with semantic relations between lists and critical distractors. Thus, it is unclear whether associative relations, semantic relations, or both foment the illusion. In Experiment 1, we developed a tool for investigating this question-a normed pool of materials in which subjects rated the gist strength of 120 DRM lists that varied widely in MBAS. This produced a mean gist strength (MGS) statistic for each list, which allowed MGS and MBAS to be manipulated factorially. In Experiment 2, we conducted the first MGS (high vs. low) × MBAS (high vs. low) factorial study of the DRM illusion. To measure how MGS and MBAS affect underlying retrieval processes, we implemented a conjoint recognition design. For raw memory performance, MGS affected both true and false recognition of critical distractors, and it affected both true and false recognition of list words. MBAS did not affect true or false recognition of list words or true recognition of critical distractors. With false recognition of critical distractors, it had a reliable effect in one condition when MGS was low, but it had no effect in another condition. At the level of retrieval processes, increasing MGS increased the familiarity of critical distractors' semantic content, and it also increased the familiarity of list words' semantic content. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Verbatim editing: A general model of recollection rejection.

Recollection rejection is traditionally defined as using verbatim traces of old items' presentations to reject new similar test cues, in old/new recognition (e.g., rejecting that couch is old by retrieving verbatim traces of sofa's presentation). We broaden this conceptualization to include (a) old as well as new similar test cues, (b) using verbatim traces for acceptance as well as rejection, and (c) using illusory verbatim traces of unpresented items (phantom recollection) as well as actual verbatim traces (true recollection). The expanded model describes how true recollection and phantom recollection generate memory decisions by creating matches and mismatches between comparisons of test cues to the content of retrieved verbatim traces versus comparisons of test cues to the content of test questions. This model generates a series of predictions about verbatim editing. Some are intuitive, such as the prection that performance will be more accurate for old cues than for new similar ones. Others are counterintuitive and conflict with an alternative model, such as correct rejections are easier than hits and that correct rejection rates will be more stable over time than hit rates. Meta-analyses of a corpus of conjoint recognition data sets provided support for the model's predictions. (PsycINFO Database Record (c) 2019 APA, all rights reserved).