Frequency effects on memory: A resource-limited theory.

Abstract

We present a review of frequency effects in memory, accompanied by a theory of memory, according to which the storage of new information in long-term memory (LTM) depletes a limited pool of working memory (WM) resources as an inverse function of item strength. We support the theory by showing that items with stronger representations in LTM (e.g. high frequency items) are easier to store, bind to context, and bind to one another; that WM resources are involved in storage and retrieval from LTM; that WM performance is better for stronger, more familiar stimuli. We present a novel analysis of preceding item strength, in which we show from nine existing studies that memory for an item is higher if during study it was preceded by a stronger item (e.g. a high frequency word). This effect is cumulative (the more prior items are of high frequency, the better), continuous (memory proportional to word frequency of preceding item), interacts with current item strength (larger for weaker items) and interacts with lag (decreases as the lag between the current and prior study item increases). A computational model that implements the theory is presented, which accounts for these effects. We discuss related phenomena that the model/theory can explain. 9 Whether the resource is spread amongst all items or whether a fixed amount is allocated for a limited number of items is currently under debate in the literature. See the General Discussion for more on this topic.

Memory resources recover gradually over time: The effects of word-frequency, presentation rate and list-composition on binding errors and mnemonic precision in source memory

“…Normative word frequency has played a key role in empirical research and theoretical development on human memory (Clark, 1992;Glanzer & Bowles, 1976;Hulme, 2003;Mandler, Goodman, & Wilkes-Gibbs, 1982;Popov & Reder, 2018). Its prominence is due to the fact that it can either facilitate or impair memory performance depending on the nature of the study task, the test, and the details of the study sequence.…”

“…The intricate effects of word frequency described above have led to the development, rejection, and revision of multiple memory models (Glanzer & Adams, 1990;McClelland & Chappell, 1998;Reder et al, 2000;Shiffrin & Steyvers, 1997). One memory model, the Source of Activation Confusion model (SAC; Popov & Reder, 2018;Reder et al, 2000), has proposed that these effects can be captured with a few simple assumptions: 1) memory formation depletes a limited working memory resource that recovers gradually over time; 2) HF words have stronger representations in memory compared to LF words; 3) the amount of resources required to process an item is an inverse function of its current strength 1 . As a result, processing LF words depletes more resources than do HF words.…”

“…As a result, processing LF words depletes more resources than do HF words. This model captures most word-frequency effects, including the interaction with list-composition, divided attention, presentation rate and age of participants (Popov & Reder, 2018). Despite SAC's success in modeling word-frequency effects, one unresolved question concerns what happens when there are insufficient resources to store a new item in memory.…”

Abstract concepts and the suppression of arbitrary episodic context

“…The synonym judgment task used at encoding may have worked to a disadvantage: as abstract concepts tend to have more diverse meanings, synonym judgments may be more difficult for abstract concepts, as it must be determined whether any particular sense of the word is a synonym to the target (Hoffman et al, 2013). Thus, an abstract concept like decision when paired with judgment might leave fewer resources available to process immediately available relational information (i.e., in the present study, the box color or the voice) because we must search for a context in which decision and judgment are in fact synonyms (a recent computational model makes this prediction; Popov & Reder, 2018). Relatedly, if abstract concepts are simply more difficult to process, and the context does not help with accessing the meaning of the word, it could render the immediate context less salient.…”

Does the motor system functionally contribute to keeping words in working memory? A pre-registered replication of Shebani and Pulvermüller (2013, Cortex)

“…As a positive control we will analyse the effect on recall of serial position of a word within a trial. Serial position effects are among the most robust effects in working memory research (see Popov & Reder, 2019 for a recent review). This check is orthogonal to our main hypothesis and merely serves as an outcome-neutral criterion to verify that we can replicate a pervasive effect in working memory tasks and that participants were engaged.…”