Search

• For many years, memory items have been normed for descriptive properties. • We described how models can create norms for the processes that memory items activate. • Two examples provide retrieval-process norms for false memory items. • The norms allow researchers to manipulate retrieval processes in recognition studies. There is a long tradition of norming words and other materials for various descriptive properties (e.g., concreteness, emotional valence, imagery, meaningfulness), which allows those properties to be manipulated in memory experiments. We introduce a new approach to norming, in which measurement models are used to norm the underlying retrieval processes that memory items trigger. We report two worked examples of this approach, for the most commonly administered items in false memory experiments: DRM lists. In one project, 36 lists that had been previously normed for their levels of true and false memory were normed for their levels of 3 retrieval processes that control false memory (recollection rejection, phantom recollection, and semantic familiarity) and 3 parallel processes that control true memory (true recollection, erroneous recollection rejection, and semantic familiarity). In the other project, 72 new DRM lists, whose associative and semantic strengths had been systematically varied, were normed for the same retrieval processes. Together, these norms provide maximum likelihood estimates of all 6 retrieval processes for all 108 lists. Analyses of those estimates revealed that list differences in the three false memory processes are mutually dissociated, as are list differences in the three true memory processes. Thus, investigators can vary each retrieval process, without confounding it with other processes, by merely selecting appropriate groups of lists. That allows simple List × Treatment designs to be used to pinpoint the retrieval processes that are responsible for the effects of important memory manipulations (e.g., delay, divided attention, encoding instructions, repetition). [ABSTRACT FROM AUTHOR]

• A normed measure of the ambiguity of words' emotional valence has been developed. • With that measure, emotional ambiguity displays characteristic empirical effects. • We developed a similar normed measure of the ambiguity of other semantic attributes. • The ambiguities of other attributes displayed the same empirical effects as emotional ambiguity. • People process the ambiguity as well as the intensity of semantic attributes. The emotional ambiguity hypothesis posits that as items are encoded, people process the ambiguity as well as the intensity of their valence. The hypothesis predicts three signature effects, all of which have been reported: ambiguity-driven declines in valence-arousal correlations, a quadratic law relating perceived valence to valence ambiguity, and improvements in episodic memory as a function of increases in valence ambiguity. After reviewing evidence on these effects, we evaluated fuzzy-trace theory's proposal that the ambiguity hypothesis should apply to a broad range of semantic attributes other than valence (e.g., categorization, concreteness, meaningfulness). According to that proposal, all three effects should be observed for other attributes. They are. In Experiment 1, ambiguity-driven reductions in attribute correlations were identified for seven other attributes. In Experiment 2, 16 other attributes obeyed the quadratic law. In Experiments 3–5, three other attributes displayed ambiguity-driven improvements in recall. It appears that people process the ambiguity as well as the intensity of many semantic attributes, and hence, the memory effects of such attributes can be due to either or both. [ABSTRACT FROM AUTHOR]

• Some events are remembered as belonging to logically incompatible reality states (complementarity). • Complementarity comes in two forms, within-item and between-item. • The experiments provide the first evidence of within-item complementarity (impossible memory conjunctions). • Complementarity was predicted by the non-compensatory gist principle but selective retrieval may also be responsible. Conjoint recognition studies have revealed that episodic memory is subadditive over mutually exclusive reality states (e.g., old vs. new), which is known as overdistribution. Because overdistribution violates the additive law of probability, it has stimulated the development of quantum memory models, which implement the distinction between verbatim and gist traces of experience. Recent versions of those models treat verbatim and gist memory as incompatible knowledge states that obey the Bohr complementarity principle, which specifies that these states can only be accessed sequentially. An extreme form of subadditivity, super-overdistribution, falls out of this incompatibility assumption, and we report the first evidence of it. Super-overdistribution occurs when it is easier to remember that items belong to a subordinate reality state than to remember that they belong to a superordinate state that contains it. Quantum memory models identify conditions under which this paradoxical phenomenon should occur and the classes of items that are more vs. less likely to exhibit it. Super-overdistribution was detected in three lines of research and, as expected, was more pronounced for some classes of items than for others. [ABSTRACT FROM AUTHOR]