cogpsych-review-process

Review Process (cogpsych-review-process)

Cognitive Psychology combines selectivity for theoretical impact with deep **methodological and

modeling scrutiny**. Reviewers and editors weigh not only whether the finding is interesting, but

whether the model is well-specified, identifiable, and properly compared, whether the **experiments

discriminate the accounts, and whether the work is reproducible**. Knowing this lets you pre-empt

the common rejection reasons. Confirm the current process on the official page (检索于 2026-06；以官网为准).

When to trigger

• Before submitting, to stress-test the manuscript
• Interpreting a decision letter and setting expectations
• Deciding how to fit a long, model-driven program to a demanding review

How review works (typical Elsevier journal pattern)

1. Editorial triage. A handling editor assesses theoretical impact, scope, and fit; thin,

single-effect, or atheoretical submissions may be rejected without external review at this

long-form, model-driven venue.

1. Expert peer review. Typically multiple referees with cognitive-modeling and experimental

expertise. Expect detailed scrutiny of model specification, identifiability/recovery, model

comparison, experimental confounds, and the strength of the inference.

1. Reproducibility is checked. Reviewers may attempt to run model/analysis code; fits that don't

regenerate, or undocumented model choices, weaken the paper (see

cogpsych-open-science-and-transparency).

1. Decisions and cycles. Reject, major/minor revision, or accept; integrative model-driven papers

often go through substantial, sometimes multiple, revision rounds — added experiments, recovery

analyses, or model comparisons are common requests.

> Verify the review model (single- vs. double-anonymized), referee count, and timelines on the journal's

> current guide for authors — these are volatile (检索于 2026-06；以官网为准).

Shape the paper to pass

• Make the theoretical advance explicit and early; show the experiments discriminate the models.
• Fit and compare models under matched flexibility; include parameter and model recovery.
• Respect the data hierarchy (mixed/hierarchical models) and report effect sizes with intervals.
• Make the modeling reproducible from deposited code; complete Elsevier declarations.
• Separate confirmatory from exploratory model work honestly.

Desk-reject and decline-without-review patterns

The long-form, model-driven identity means many submissions never reach external review. Recognize these

shapes and pre-empt them:

| Pattern an editor sees | Likely outcome | Pre-empt it by |

|------------------------|----------------|----------------|

| One experiment, one effect, no model/theory | desk reject (wrong shape) | grow into a model-driven program or place in a short-report venue |

| Model fit but never compared to a rival | major revision or reject | fit rivals under matched flexibility; report criteria |

| Experiments don't discriminate the accounts | reject (non-diagnostic) | redesign for the discriminating signature |

| Aggregated analyses, ignored subject/item variance | methods flag | refit with mixed/hierarchical models |

| Fits not reproducible; no code | reproducibility flag | deposit seeded model code with a run log |

| Better-fitting but more flexible model claimed as winner | overfitting flag | add recovery + penalized comparison/cross-validation |

Worked micro-example (illustrative triage)

Manuscript: three preregistered recognition-memory experiments; UVSD vs.
            DPSD fit and compared (hierarchical Bayesian), recovery reported,
            open data + model code with DOIs, diagnostic z-ROC signature.
Editor read: theoretical impact (adjudicates a long-running debate), modeling
            rigor (comparison + recovery), reproducibility (code regenerates).
Likely route: external review, probable major revision for added robustness
            (alternative priors, a further model, more recovery).
Counter-case: same effect, one experiment, one model fit, request-only data,
            no recovery → likely declined without full review.

How reviewers weigh the evidence (calibration anchors)

• The strongest signal is a diagnostic experiment + a recovered, compared model that together pick

one account over a real rival — this converts "interesting fit" into "credible adjudication."

• Reviewers distrust a fit advantage without recovery and matched flexibility; a crossed qualitative

prediction is more persuasive than a smaller AIC.

• Reproducibility is part of the evidence, not a formality; a fit that doesn't regenerate reads as a

result that might not exist.

Anti-patterns

• A single-effect, atheoretical submission expecting full review at a model-driven venue
• A model fit with no rival, no comparison, and no recovery
• Aggregated analyses that ignore crossed subject/item variance
• Expecting acceptance without a substantial, modeling-heavy revision round
• Irreproducible fits or undocumented model choices

Output format

【Theoretical advance】clear early? [Y/N]
【Discrimination】do experiments separate the models? [Y/N]
【Modeling rigor】comparison + recovery + matched flexibility? [Y/N]
【Hierarchy + reporting】mixed/hierarchical + effect sizes/intervals? [Y/N]
【Reproducible】model code regenerates fits? [Y/N]
【Realistic outcome】reject / major revision / minor revision / accept
【Next】cogpsych-submission (or cogpsych-rebuttal if decided)

Supplementary resources

• [../../resources/official-source-map.md](../../resources/official-source-map.md) — review model, scope, and reproducibility expectations