Working Memory Has Better Fidelity Than Long-Term Memory: The Fidelity Constraint Is Not a General Property of Memory After All

Natalie Biderman*, Roy Luria, Andrei R. Teodorescu, Ron Hajaj, Yonatan Goshen-Gottstein

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

10 Scopus citations

Abstract

How detailed are long-term-memory representations compared with working memory representations? Recent research has found an equal fidelity bound for both memory systems, suggesting a novel general constraint on memory. Here, we assessed the replicability of this discovery. Participants (total N = 72) were presented with colored real-life objects and were asked to recall the colors using a continuous color wheel. Deviations from study colors were modeled to generate two estimates of color memory: the variability of remembered colors—fidelity—and the probability of forgetting the color. Estimating model parameters using both maximum-likelihood estimation and Bayesian hierarchical modeling, we found that working memory had better fidelity than long-term memory (Experiments 1 and 2). Furthermore, within each system, fidelity worsened as a function of time-correlated mechanisms (Experiments 2 and 3). We conclude that fidelity is subject to decline across and within memory systems. Thus, the justification for a general fidelity constraint in memory does not seem to be valid.

Original languageEnglish
Pages (from-to)223-237
Number of pages15
JournalPsychological Science
Volume30
Issue number2
DOIs
StatePublished - 1 Feb 2019

Funding

FundersFunder number
Israel Science Foundation1922/15

    Keywords

    • color memory
    • continuous-report paradigm
    • fidelity
    • long-term memory
    • open data
    • preregistered
    • replication
    • working memory

    Fingerprint

    Dive into the research topics of 'Working Memory Has Better Fidelity Than Long-Term Memory: The Fidelity Constraint Is Not a General Property of Memory After All'. Together they form a unique fingerprint.

    Cite this