Sequential sampling from memory underlies perceptual decisions unyoked from actions
Sharma, P., Shadlen, M., Shushruth, S.
neuroscience
Quick verdict
Like macaques, humans postpone direction decisions until targets appear; diffusion fits support sampling from memory, though only six subjects were tested.
Summary
The paper asks whether human perceptual decisions, like those of macaques, can be postponed until the actions that express them are revealed. Six naive subjects viewed brief stochastic random-dot motion whose direction spanned 360° and whose coherence varied from 0–64%; after a 400 ms delay, two targets appeared at unpredictable locations—one at the true direction, one 40° away—and subjects saccaded to the closer match. The authors fit choice accuracy and go-reaction times with a bounded drift-diffusion model in which evidence samples are stored in memory during viewing and integrated only after targets appear, a fixed go-RT model in which the decision is made during viewing, and a hybrid mixture of the two strategies. All subjects showed lapses at 64% coherence, sensitivity varied widely across individuals, and sensitivity correlated with the coherence-dependence of go-RT (r = 0.82, p = 0.047). BIC comparisons favored the diffusion model over the fixed model in five of six subjects, and the hybrid over diffusion in all six; model-recovery simulations showed these comparisons were conservative. Extending the memory delay to 1200 ms in three subjects left accuracy intact, arguing against iconic memory. Limitations include six subjects with unreported trial counts, no control condition permitting during-viewing commitment, and the absence of stimulus-duration analyses.
Review complete
Assessment at a glance
6
risk flags identified
● high 1
● medium 3
● low 2
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Human subjects, like macaques, postpone formation of difficult direction decisions until response targets are revealed, deliberating during the action-selection epoch.
- Coherence-dependent go-reaction times in high-accuracy subjects reflect bounded evidence accumulation performed over samples of motion evidence retrieved from memory.
- A hybrid model in which strong evidence triggers commitment during viewing and weak evidence triggers memory-based integration fits all six subjects better than pure diffusion.
- The decision-relevant representation survives a 1200 ms delay without accuracy loss, implicating working memory rather than iconic memory.
- Working memory operates prospectively: sensory information is retained strategically according to its anticipated future utility.
Strengths
Design that defeats the saccade-remapping alternative
Randomizing motion direction over 360° in 10° steps and revealing targets only after a post-stimulus delay makes it impossible to pre-compile a small set of motor plans, directly answering the fast reorienting objection (tens of ms) the authors raise against prior action-independent decision studies in the Discussion.
Formal model comparison with model recovery
Table 1 shows diffusion- and fixed-generated data were almost never misclassified as hybrid (1% and 0% of simulations), so the hybrid's superiority over diffusion for all six subjects (Fig. 5B) cannot be an artifact of added flexibility — the comparison is conservative in the direction that matters.
Per-subject fitting with full uncertainty reporting
A second experiment targeting the memory substrate
Weaknesses
Small sample with a fragile headline correlation
The 'strategic' claim outruns the design
Working-memory attribution is inferential only
Asymmetric model-comparison ladder and recovery gaps
Nothing verifiable until acceptance
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
This preprint (bioRxiv, DOI 10.1101/2025.09.10.675475) tests whether humans, like macaques (Shushruth et al., 2022), postpone forming a perceptual decision until the actions that can express it are revealed. Six naive subjects judged the net direction of brief stochastic random-dot motion (RDM) spanning 360°; after a 400 ms delay, two targets appeared at unpredictable locations (one at the true direction, one 40° offset) and subjects saccaded to the closer match. The central claims: (1) humans store motion evidence in memory during viewing and integrate it only during target selection ('deliberation during action selection'); (2) go-reaction times scale with motion coherence in high-accuracy subjects, a signature of post-target accumulation; (3) a bounded drift-diffusion model operating over stored samples accounts for choices and go-RTs; and (4) the stored representation outlasts iconic memory, implicating working memory. The work is a human conceptual replication of the macaque study by two of the same authors, explicitly designed to rule out the 'monkeys cannot categorize direction' alternative.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 6 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What does it mean that perceptual decisions are 'unyoked from actions'?
In most laboratory tasks, people make perceptual decisions while already knowing the response action (e.g., which button to press). This study delayed the response options until after the stimulus ended: subjects watched noisy random-dot motion, and only afterwards saw two targets, choosing by eye movement the one closer to the perceived motion direction. This separates the moment of decision from the moment the action becomes available.
Did humans postpone decisions the way the monkeys did?
Largely yes. Subjects with better direction-discrimination accuracy showed longer reaction times after the targets appeared, and those times depended on motion strength — the same signature previously reported in macaques, indicating that evidence stored during stimulus viewing was being deliberated during the action-selection phase rather than committed during viewing.
How do the authors know the information was held in working memory rather than iconic memory?
Iconic (sensory) memory decays within roughly a second. In three subjects, extending the delay between stimulus offset and target onset from 400 to 1200 ms did not reduce decision accuracy, so the usable information outlived the iconic-memory timescale, consistent with working memory. The study did not directly manipulate memory capacity, so the attribution remains inferential.
Is the data or code from this study available?
Not at preprint stage. The paper states that behavioral data and the code used to generate the figures will be uploaded to a public repository upon acceptance of the manuscript.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figure panels could not be visually inspected; axis data in Figures 3–7 appeared as garbled label text, so plotted fit quality was judged from captions and the surrounding text only.
- Trial counts per subject are absent from the extracted text, limiting assessment of statistical power and of the adequacy of per-coherence go-RT distributions.
- The extraction appears to cover the full main text and references, but any supplementary material (e.g., fitting implementation details) was not included.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.