CoChat
Sign Up
Cited by 270Decision-Making and Behavioral EconomicsNeural and Behavioral Psychology StudiesVisual perception and processing mechanisms

The DOI matches its Crossref record; 67 of the 67 references it cites were found in a registry or an index; none of the 43 sources its claims cite is checked for support yet.

A common mechanism underlies changes of mind about decisions and confidence

Open access
van den Berg, Ronald, Anandalingam, Kavitha, Zylberberg, Ariel, Kiani, Roozbeh, Shadlen, Michael N, Wolpert, Daniel M·2016·eLife
DOI 10.7554/elife.12192PMC PMC4798971
Open this paper in CoChat — free account

Abstract

Decisions are accompanied by a degree of confidence that a selected option is correct. A sequential sampling framework explains the speed and accuracy of decisions and extends naturally to the confidence that the decision rendered is likely to be correct. However, discrepancies between confidence and accuracy suggest that confidence might be supported by mechanisms dissociated from the decision process. Here we show that this discrepancy can arise naturally because of simple processing delays. When participants were asked to report choice and confidence simultaneously, their confidence, reaction time and a perceptual decision about motion were explained by bounded evidence accumulation. However, we also observed revisions of the initial choice and/or confidence. These changes of mind were explained by a continuation of the mechanism that led to the initial choice. Our findings extend the sequential sampling framework to vacillation about confidence and invites caution in interpreting dissociations between confidence and accuracy.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

Changes of mind about choice and confidence in a motion-discrimination task are explained by continued bounded evidence accumulation after the initial decision, using the same evidence stream. This post-decision processing can produce an apparent dissociation between confidence and accuracy without invoking separate metacognitive mechanisms.

Key findings

  • Four subjects made simultaneous direction and confidence reports by reaching to one of four targets; changes of mind occurred on 2.0–8.8% of trials, and confidence changes were more frequent than decision changes (p<1e-4 for all participants).
  • Changes of decision were more likely after errors and with stronger motion; for three of four subjects they corrected an initial error more often than they spoiled a correct choice (P<0.001, P<0.001, P=0.084, P<0.001). Changes of confidence were more likely after low-confidence initial choices, and changes of mind earned 0.72–1.12 additional points per change trial.
  • Motion-energy kernels showed that initial choice and confidence were supported by a common evidence stream, with information used until ~400 ms before movement onset. Post-decision information explained revisions: low-to-high confidence changes were supported by evidence for the initial choice, whereas high-to-low confidence changes were associated with evidence for the opposite direction.
  • Comparing odds ratios from initial versus final confidence produced a systematic discrepancy (P<1e-4, sign test; reliable for 13 of 19 points, P<0.05 bootstrap), which the authors argue can arise from continued first-order evidence accumulation rather than a separate second-order confidence process.

Methods

  • Six naïve right-handed subjects (ages 21–34) participated; two were excluded, leaving four analyzed. Subjects used a vBOT handle to judge random-dot motion direction and report confidence simultaneously, with coherence levels ±0, ±3.2, ±6.4, ±12.8, ±25.6, and ±51.2%; the stimulus was extinguished at movement onset. Each subject completed 12 sessions (9024–9216 trials), except Subject 4, for whom only the first 5000 trials were analyzed because of nonstationary behavior.
  • Behavioral analyses used logistic regressions, ANOVA, Fisher exact tests, and motion-energy psychophysical kernels. The main model was a race between two accumulators receiving bivariate Gaussian evidence, with decision bound B, non-decision time, and a log-odds threshold θ for confidence; changes of mind were modeled by continued accumulation for a post-initiation period tpip with shifted thresholds δθ1–δθ3.
  • Model parameters were fit by maximum likelihood to initial choice, confidence, and reaction-time data, then the change-of-mind model was fit with κ and θ carried over from the initial fits.

Limitations

  • Stated: the model is imperfect—it overestimates confidence at the lowest motion strength and shows discrepancies at 0% coherence for three of four subjects; the flat bound does not capture RT distributions or mean RT on error trials, so only correct-trial mean RTs and 0% coherence trials were used in the cost function.
  • Stated: basing changes of decision on log odds is a simplification, and the frequency of changes of mind was too small to evaluate more flexible post-initiation termination rules.
  • Evident: the main conclusions rest on four analyzed subjects in a specific random-dot motion task with stationary evidence samples; the paper notes that other perceptual and memory-based decisions involve less well-characterized evidence streams.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.7554/elife.12192

Cites

67 of 67 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 of 43 cited sources checked · 0 supported · 0 flagged

Show all 59 claims

  • Probably not a substantial claim. 32
  • It may state two separate claims as one. 26
  • The quote only partly states this claim. 1
45 of 45

  • The authors state this as an assumption.

    Subjects adopt a consistent criterion on degree of belief (log odds) to decide in favor of high or low confidence.

    “For simplicity, we assumed that subjects adopt a consistent criterion on “degree of belief” (log odds, Figure 3b) to decide in favor of high or low confidence.”

  • The authors state this as an assumption.

    The same information was used to reach an initial choice about direction and confidence.

    “The critical assumptions are (i) the same information was used to reach an initial choice about direction and confidence, and (ii) there is additional information available to the decision maker after committing to a choice and confidence category, despite the disappearance of the random dot motion upon response initiation.”

  • The authors state this as an assumption.

    There is additional information available to the decision maker after committing to a choice and confidence category, despite the disappearance of the random dot motion upon response initiation.

    “The critical assumptions are (i) the same information was used to reach an initial choice about direction and confidence, and (ii) there is additional information available to the decision maker after committing to a choice and confidence category, despite the disappearance of the random dot motion upon response initiation.”

  • The authors state this as an assumption.

    Evidence was accumulated past the point of the initial decision for a fixed amount of time (tpip) that is less than the non-decision time.

    “We assumed that evidence was accumulated past the point of the initial decision for a fixed amount of time (a free parameter, tpip) that is less than the non-decision time.”

  • The authors state this as an assumption.

    The state of the belief at tpip determines the final confidence and choice.

    “The state of the belief at tpip determines the final confidence and choice (see Materials and methods).”

  • The authors state this as an assumption.

    Subjects might not apply identical criteria for confidence and choice in the pre- and post-initiation epoch (δθ parameters).

    “we allowed for the possibility that subjects might not apply the identical criteria for confidence and choice in the pre- and post-initiation epoch (the δθ parameters; see Materials and methods).”

  • The authors state this as an assumption.

    Choice and confidence are related but processed as if separate attributes, via a correspondence between decision variable and belief.

    “To explain our findings, we assumed that choice and confidence are related but processed as if separate attributes, via a correspondence between decision variable and belief.”

  • The authors state this as an assumption.

    The model parameterization requires the pink and red regions to be same for both initial high and low confidence decisions.

    “Note that the model parameterization requires the pink and red regions to be same for both initial high and low confidence decisions.”

  • The paper relies on this without support.

    Non-decision time typically exceeds 300 ms even in relatively fast perceptual decisions.

    “Even in relatively fast perceptual decisions, this so-called non-decision time typically exceeds 300 ms.”

  • The paper relies on this without support.

    The temporal stream of evidence in random dot kinematograms is effectively a sequence of independent, statistically stationary samples.

    “the temporal stream of evidence is effectively a sequence of independent, statistically stationary samples.”

  • The paper relies on this without support.

    There are processing delays between the arrival of information from the environment and updating its representation in working memory as a decision variable, and there are delays between commitment to a decision and communicating this decision through the motor system.

    “Specifically, there are processing delays between the arrival of information from the environment and updating its representation in working memory as a decision variable bearing on a proposition, and there are delays between commitment to a decision and communicating this decision through the motor system.”

  • The authors state this as an assumption.

    About half the noise is shared between the two accumulators (arises from the stimulus).

    “This choice of covariance assumes that about half the noise is shared between the two accumulators (i.e. arises from the stimulus) and is within the normal range of fitted covariances (Kiani et al., 2014a).”

  • The authors state this as an assumption.

    Once the initial decision has been made, the accumulators continue to integrate information that was not accessible at the time of the initial decision for a further post-initiation period tpip.

    “To model changes of mind, we assumed that once the initial decision had been made, the accumulators continued to integrate information that was not accessible at the time of the initial decision (due to latencies in the sensory and motor system) for a further post-initiation period tpip (constrained to be less than tnd).”

  • The authors state this as an assumption.

    A change in confidence or choice occurs if at the end of the post-initiation period the log odds crosses a confidence or choice threshold.

    “A change in confidence or choice would occur if at the end of this period the log odds crossed a confidence or choice threshold, respectively (Figure 3b).”

  • The paper relies on this without support.

    There are motor costs involved in changes of mind.

    “Since there are motor costs involved in changes of mind, we included additional parameters that could move the confidence and choice thresholds away from their initial values in the post-initiation stage by δθ1, δθ2 and δθ3 (Figure 3b).”

  • The paper relies on this without support.

    Under sequential sampling models, meta-d′ ≠ d′ even when choice and confidence are explained by the same decision variable.

    “Under sequential sampling (e.g., bounded drift diffusion), meta-d′ ≠ d′, even when the choice and confidence are explained by the same decision variable.”

  • The authors state this as an assumption.

    Momentary evidence is drawn from a bivariate Gaussian with mean (κC, -κC).

    “Momentary evidence was modeled as draws from a bivariate Gaussian with a mean that depends on the coherence (C) such that drift rate (/s) is κC for the rightward and -κC for the leftward accumulator, giving a mean of (κC,-κC).”

  • The authors state this as an assumption.

    The two accumulators share noise with negative covariance ρ.

    “The bivariate Gaussian has a negative covariance ρ which determines the extent to which the two accumulators share noise (for example arising from fluctuations in the stimulus).”

  • The authors state this as an assumption.

    A decision is made when one of the two races crosses a decision bound B.

    “A decision is made when one of the two races crosses a decision bound B.”

  • The authors state this as an assumption.

    Reaction time is determined by the time to reach the decision bound plus a non-decision time modeled as a normal distribution with mean μtnd and standard deviation σtnd.

    “The reaction time is determined by the time to reach the decision bound and an additional non-decision time (e.g., due to sensory and motor latencies) which was modeled as a normal distribution with mean μtnd and standard deviation σtnd.”

  • The authors state this as an assumption.

    Confidence is modeled with a log-odds threshold θ separating high from low confidence judgments.

    “To model confidence, we included a log-odds threshold θ which separated high from low confidence judgments.”

  • The authors state this as an assumption.

    The confidence threshold is time-invariant.

    “For simplicity, we chose a time-invariant threshold.”

  • The paper relies on this without support.

    Fluctuations have a stronger impact when motion is weak.

    “Because fluctuations have a stronger impact when motion is weak, only the lowest motion strengths (≤6.4% coh) were included in these analyses.”

  • The paper relies on this without support.

    Due to the stochastic nature of the motion stimuli, the strength of motion varies from trial to trial and within a trial.

    “Due to the stochastic nature of the motion stimuli, the strength of motion will vary from trial to trial, and even within a trial.”
24 of 24

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?

Citations

Cites

Flagged first, then by citations

Show all 67 references

Research works printed but not linked: 1

  1. Wald, 1947, Sequential Analysis

Cited by

Sorted by citations

Items 201–269 of 269

Related

See all

Metadata

Authors
van den Berg, R., Anandalingam, K., Zylberberg, A., Kiani, R., Shadlen, M. N., & Wolpert, D. M.
Affiliations
  • Computational and Biological Learning Laboratory, Department of Engineering, Cambridge University, Cambridge, United Kingdom
  • University of Cambridge (GB)
  • Howard Hughes Medical Institute (US)
  • New York University (US)
Corresponding author
Wolpert, Daniel M
ORCID
Venue
eLife
Publisher
eLife Sciences Publications, Ltd
Published
2016-02-01
Volume
5
Pages
e12192-e12192
ISSN
2050-084X (ISSN-L)
DOI
10.7554/elife.12192
PMC
PMC4798971
OpenAlex
W2261997804
Semantic Scholar
8a3457c13959cfb9435daf0710bf5eac7eacbc9b
Cited by
270
References
67
Citations by year
2026: 19 · 2025: 20 · 2024: 23 · 2023: 38 · 2022: 24 · 2021: 26 · 2020: 32 · 2019: 26 · 2018: 31 · 2017: 16 · 2016: 14
FWCI
25.07
Type
journal article
Language
en
Open access
Licence
Locations
  • eLife (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • KTH Publication Database DiVA (KTH Royal Institute of Technology) (submittedVersion, other-oa)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Cambridge University Engineering Department Publications Database (submittedVersion)
Funders
  • Wellcome Trust
  • Howard Hughes Medical Institute
  • Royal Society
  • Human Frontier Science Program
  • National Eye Institute
  • Simons Foundation
  • Alfred P. Sloan Foundation
Topics
  • Social Sciences › Decision Sciences › General Decision Sciences › Decision-Making and Behavioral Economics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Visual perception and processing mechanisms
Keywords
decision confidence, sequential sampling, vacillation, change of mind, perceptual decision making, motion direction discrimination, reaction time, processing delay
MeSH terms
Choice Behavior, Decision Making, Humans, Models, Neurological, Reaction Time, Healthy Volunteers

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.

A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

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