Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 75 of the 76 references it cites were found in a registry or an index.

Statistical Computations Underlying the Dynamics of Memory Updating

Open access
Gershman, Samuel J., Radulescu, Angela, Norman, Kenneth A., Niv, Yael · 2014 · PLoS Computational Biology

Abstract

Psychophysical and neurophysiological studies have suggested that memory is not simply a carbon copy of our experience: Memories are modified or new memories are formed depending on the dynamic structure of our experience, and specifically, on how gradually or abruptly the world changes. We present a statistical theory of memory formation in a dynamic environment, based on a nonparametric generalization of the switching Kalman filter. We show that this theory can qualitatively account for several psychophysical and neural phenomena, and present results of a new visual memory experiment aimed at testing the theory directly. Our experimental findings suggest that humans can use temporal discontinuities in the structure of the environment to determine when to form new memory traces. The statistical perspective we offer provides a coherent account of the conditions under which new experience is integrated into an old memory versus forming a new memory, and shows that memory formation depends on inferences about the underlying structure of our experience.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1371/journal.pcbi.1003939

Cites

75 of 75 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Simulations. (A) Simulated sensory measurements and inferred state variables. For the DP-KF, the colors indicate the mode assignment with the highest posterior probability, white circles = mode 1, black circles = mode 2. (B) Posterior probability of mode 1 as a function of morph index in the gradual and mixed protocols, using the DP-KF (averaged over multiple simulation runs). See text for details.
Figure 1 Simulations. (A) Simulated sensory measurements and inferred state variables. For the DP-KF, the colors indicate the mode assignment with the highest posterior probability, white circles = mode 1, black circles = mode 2. (B) Posterior probability of mode 1 as a function of morph index in the gradual and mixed protocols, using the DP-KF (averaged over multiple simulation runs). See text for details.CC BY · PubMed Central
Figure 2 Experimental task. (Left) Prediction trial: participants were asked to predict the orientation and length of the next line segment (prediction shown in the center of the screen). At the bottom of the screen, a black circle superimposed on a timeline (the black bar) was used to indicate the trial's serial position in the block. At the start of each block, the black circle started out in the leftmost position; after each trial, the circle's position shifted one position to the right. (Middle) After making a prediction, participants were shown the true line segment and received a point score based on their prediction accuracy. (Right) Reconstruction trial: at the end of each block, participants were asked to reconstruct from memory the line they saw on one of the first three trials (indicated by an arrow on the timeline). No feedback was given for these reconstruction trials.
Figure 2 Experimental task. (Left) Prediction trial: participants were asked to predict the orientation and length of the next line segment (prediction shown in the center of the screen). At the bottom of the screen, a black circle superimposed on a timeline (the black bar) was used to indicate the trial's serial position in the block. At the start of each block, the black circle started out in the leftmost position; after each trial, the circle's position shifted one position to the right. (Middle) After making a prediction, participants were shown the true line segment and received a point score based on their prediction accuracy. (Right) Reconstruction trial: at the end of each block, participants were asked to reconstruct from memory the line they saw on one of the first three trials (indicated by an arrow on the timeline). No feedback was given for these reconstruction trials.CC BY · PubMed Central
Figure 3 Example trajectories and hypothetical reconstructions. Each circle represents a line segment presented in a sequence, with the shaded circle indicating the first trial. The dimensions are standardized to a [0,100] range. The blue diamond represents a hypothetical reconstruction of the line segment indicated by the arrow. In solid black is the distance between the reconstruction and the starting point, while the dashed black line shows the distance between the reconstruction and the end point. (A) A gradual trajectory. Here we expected the reconstruction to be pulled away from the start point and towards the end point. (B) A jump trajectory. Here we expected the reconstruction to stay in the vicinity of the pre-jump points. As a result, we expected the distance between the reconstruction and the start point to be smaller in the jump condition as compared to the gradual condition, and the distance between the reconstruction and the end point to be smaller in the gradual condition.
Figure 3 Example trajectories and hypothetical reconstructions. Each circle represents a line segment presented in a sequence, with the shaded circle indicating the first trial. The dimensions are standardized to a [0,100] range. The blue diamond represents a hypothetical reconstruction of the line segment indicated by the arrow. In solid black is the distance between the reconstruction and the starting point, while the dashed black line shows the distance between the reconstruction and the end point. (A) A gradual trajectory. Here we expected the reconstruction to be pulled away from the start point and towards the end point. (B) A jump trajectory. Here we expected the reconstruction to stay in the vicinity of the pre-jump points. As a result, we expected the distance between the reconstruction and the start point to be smaller in the jump condition as compared to the gradual condition, and the distance between the reconstruction and the end point to be smaller in the gradual condition.CC BY · PubMed Central
Figure 4 Experimental results and model predictions. (A) Euclidean distance between participants' reconstructions and the observed (true) first and last lines in a block. Error bars represent within-subject standard error of the mean. The results show that participants were more accurate in their reconstructions in the jump condition as compared to the gradual condition. (B) Stationary Kalman filter (KF) model predictions. Data in (A) are represented by black circles. (C) Non-stationary KF model predictions. (D) Stationary Dirichlet process Kalman filter (DP-KF) model predictions. (E) Non-stationary DP-KF predictions.
Figure 4 Experimental results and model predictions. (A) Euclidean distance between participants' reconstructions and the observed (true) first and last lines in a block. Error bars represent within-subject standard error of the mean. The results show that participants were more accurate in their reconstructions in the jump condition as compared to the gradual condition. (B) Stationary Kalman filter (KF) model predictions. Data in (A) are represented by black circles. (C) Non-stationary KF model predictions. (D) Stationary Dirichlet process Kalman filter (DP-KF) model predictions. (E) Non-stationary DP-KF predictions.CC BY · PubMed Central
Figure 5 Model comparison. (A) Predictive log-likelihood for each model, on prediction trials, relative to the stationary KF model. Larger values indicate superior performance on held-out data from prediction trials. (B) Predictive log-likelihood for each model, on the reconstruction data, relative to the stationary KF model. Error bars represent within-subject standard error of the mean.
Figure 5 Model comparison. (A) Predictive log-likelihood for each model, on prediction trials, relative to the stationary KF model. Larger values indicate superior performance on held-out data from prediction trials. (B) Predictive log-likelihood for each model, on the reconstruction data, relative to the stationary KF model. Error bars represent within-subject standard error of the mean.CC BY · PubMed Central
Figure 6 Comparison of model and human reconstructions. Histogram of z-transformed correlations between human reconstructions and model reconstructions for (A) the orientation dimension and (B) the length dimension. Vertical black line indicates a correlation of 0.
Figure 6 Comparison of model and human reconstructions. Histogram of z-transformed correlations between human reconstructions and model reconstructions for (A) the orientation dimension and (B) the length dimension. Vertical black line indicates a correlation of 0.CC BY · PubMed Central
Figure 7 Model-based analysis. Number of modes (K) inferred by the DP-KF model for each condition.
Figure 7 Model-based analysis. Number of modes (K) inferred by the DP-KF model for each condition.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 76 references

Research works printed but not linked: 1

  1. Bartlett FC (1932) Remembering: An Experimental and Social Study. Cambridge University Press.

Cited by

See all 109

Metadata

Authors
Gershman, S. J., Radulescu, A., Norman, K. A., & Niv, Y.
Affiliations
  • Massachusetts Institute of Technology (US)
  • Princeton University (US)
Corresponding author
Gershman, Samuel J.
Venue
PLoS Computational Biology
Published
2014-11-06
Volume
10
Issue
11
Pages
e1003939
ISSN
  • 1553-734X (ISSN-L)
  • 1553-7358
OpenAlex
W2008094377
Cited by
101
References
76
Citations by year
2026: 6 · 2025: 12 · 2024: 13 · 2023: 9 · 2022: 14 · 2021: 11 · 2020: 12 · 2019: 5 · 2018: 6 · 2017: 6 · 2016: 4 · 2015: 3
FWCI
2.42
Type
journal article
Language
en
Open access
Locations
  • PLoS Computational Biology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc)
Funders
  • National Science Foundation
  • John Templeton Foundation
  • National Institutes of Health
  • National Institute of Mental Health (R01MH098861)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory Processes and Influences
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Visual perception and processing mechanisms
Keywords
memory updating, memory formation, switching Kalman filter, dynamic environments, temporal discontinuity, visual memory
MeSH terms
Humans, Memory, Models, Neurological, Task Performance and Analysis, Models, Statistical, Computational Biology

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Statistical Computations Underlying the Dynamics of Memory Updating. https://app.cochat.ai/papers/10.1371/journal.pcbi.1003939
The paper
Gershman, S. J., Radulescu, A., Norman, K. A., & Niv, Y. (2014). Statistical Computations Underlying the Dynamics of Memory Updating. PLoS Computational Biology. https://doi.org/10.1371/journal.pcbi.1003939

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.