Pre-submission checklist: Nature Communications (with Communications Psychology cascade)

[JamesPHoughton]

Submission target: Nature Communications first, with automatic cascade to Communications Psychology if NCom rejects.

Pre-submission edit checklist. Both journals share Nature Portfolio requirements; differences are flagged inline.

Blockers (must fix before submission)

Preregistration disclosure

• Add date of preregistration next to the AsPredicted link in Methods (HBSS hard requirement)
• Add a consolidated "Deviations from preregistered analysis plan" subsection (Methods or SI). Currently 3 deviations are scattered: cluster bootstrap vs. OLS (SI p.75), OLS vs. elastic net for variance comparisons (SI p.79), individual range scaling vs. elastic net feature selection (SI p.82–83)
• State explicitly that Wave 1 was exploratory (not preregistered) and Wave 2 was confirmatory
• In Table 1 footnote, state which specific analyses were preregistered (predictor list, model form, alpha) vs. exploratory robustness checks

Ethics / participant section

• Add explicit informed consent statement in Methods ("Informed consent was obtained from all participants")
• Add explicit compensation language in ethics statement (HBSS specifically requires this): "Participants received $10 (median hourly rate approximately $17), and could withdraw at any time without penalty"

LLM / AI disclosure

• Add a "Use of LLMs" subsection in Methods. LLMs were used for: (1) topic ideation (SI p.44, p.50), (2) transcript compliance verification (SI p.89, 97.8% accuracy). Need to disclose model name, version, dates, and purpose for each
• Add validation of the LLM compliance check: hand-code a random sample of transcripts, report agreement rate

Document structure

• Remove or rename "Significance Statement" (p.2) — this is a PNAS holdover; NCom doesn't use it. Either delete or fold into abstract/cover letter
• Split "Data and Materials Availability" into separate Data availability and Code availability sections (NCom requires this)
• Deposit data in a DOI-minting repo (Zenodo, Figshare, Dryad) and update the availability statement with the DOI. GitHub alone is not sufficient
• NCom-specific: consider depositing code in Code Ocean rather than Zenodo. NCom has explicit Code Ocean integration that gives reviewers a one-click reproducible run, and reduces the chance the editors invoke code review

Main text edits — null results language + ROPE references

Add ROPE references (~2–3 sentences total)

• Abstract: leave the framing as-is (decision recorded above — "felt" already governs the example list grammatically; "correlate" disclaims causality)
• Results summary paragraph (p.9): "none of our four dimensions of politicalness predicted which topics worked better" — add ROPE reference. Decision recorded: keep "predicted" wording (it is a statistical operation, not a causal claim, and the ROPE provides the credible-evidence basis the policy is actually about). Optional one-word insurance: change to "reliably predicted" if a reviewer flags it
• Figure 3 discussion paragraph: Add the ROPE sentence: "Bayesian ROPE analyses confirm that the practical effect of each predictor over its observed range is negligible for both outcome measures (SI Tables S3a–S3b)."
• Significance Statement (if kept): "these differences were not explained by the level of disagreement or politicalness of the topic" — soften

Language pass: absolute-null → "no credible evidence"

• "this variation does not appear to stem from their political nature" (p.12) → "we found no credible evidence that this variation stems from..."
• "no operationalization of politicalness... can explain more than 2%" (p.11) → soften
• "None of these variables reliably predicted changes" (p.15) → "None showed credible evidence of predicting..."
• Discussion (p.21): "none of our four dimensions of politicalness predicted which topics worked better" → soften
• Discussion (p.22): "these differences were not reliably explained by any of the topic-level dimensions" — verify wording
• Final scan of abstract, results, and discussion for any remaining absolute-null phrasing

Reviewer landmines (high priority)

Methodological clarity (the #1 reason for prior rejections)

• Rewrite the constrained randomization explanation in SI p.52. This is the section that tripped up Sci Adv reviewers. Insert a high-level reframing paragraph before the procedural details: "At a high level, our procedure performs stratified random assignment across a grid of 70 cells defined by the 10 topics and 7 levels of dyadic disagreement..." — see chat for full proposed text
• Fix the internal inconsistency where balls are described as "partner*topic combinations" in one place and "topic × disagreement cells" in another

ROPE presentation

• In SI Bayesian ROPE section, lead with the observed-range argument before showing Table S3a. The raw-scale ROPE (S3a) does not support the null for Topic Expected Identity Threat (5–16% posterior mass within ±1°–±3°), so a reviewer reading S3a first reaches the wrong conclusion. Present Table S3b first as the primary analysis; show S3a after, framed as the unadjusted version where the theoretical scale overstates the test
• Note: ROPE is on single-predictor regressions, not the multiple regression in Table 1. Reference it from the Figure 3 discussion paragraph (which is also single-predictor), not from the Table 1 paragraph

Conversation dynamics framing

• Abstract framing decision recorded: leave as-is. Rationale: grammatical "felt" governs the list of perception examples; "correlate" disclaims causality; further hedging is tautological since experience is by definition self-report

Topic measurements partly endogenous

• Add Methods statement acknowledging that contentiousness and partisanship measurements were refined using study-participant data after collection (SI p.47 mentions this, but it's not flagged in Methods). Argue why this doesn't bias inferences (e.g., quantify how much positions changed)

Wave designation

• State at the start of Results: "Wave 1 was exploratory; Wave 2 (preregistered) was confirmatory. We report both for transparency, but inferences are anchored on Wave 2."

Inconsistencies / nits

• Title: Cover page says "cross-partisan conversation" (singular); SI title and file metadata say "conversations" (plural). Unify
• R² rounding: Main text (p.11) says R² = 0.021 / 0.011; abstract says "2%". Trait ratings is 1%, so abstract should say "1–2%" or pick one outcome
• SI topic table skips item #111 (jumps 110 → 112) — typo
• P-value reporting: Per Nature statistical guidance, report exact P values unless p<0.001. Abstract uses "p < .001" without sample sizes; "all p > .18" (p.17) hides individual values. Quick edit
• Reference 23 (Rossiter 2023) missing publication info — add SSRN/OSF URL or "manuscript"
• Two correlated outcomes (feeling thermometer + trait ratings) not corrected for in Tables 2/3 (BH adjusts across predictors but not across outcomes). Add one-sentence footnote justifying this
• Em-dashes in the introduction — already addressed. Also fix "sexual-identity based" → "sexual-identity-based" (compound modifier needs full hyphenation)

Required statements / files to prepare

• Reporting summary (BSS module) — separate fillable PDF (requires Adobe Reader). Mandatory before peer review
• Author contributions — already present (CRediT format), good
• Competing interests — already present, good
• Funding statement — present, but verify Templeton grant number is included (currently not shown)
• ORCID for corresponding author — done (0000-0002-6907-6973)

Cover letter (NCom-tuned)

• Affiliation and contact info for corresponding author
• Lead with the substantive headline + the methodological reframe. NCom's bar is "important advances of significance to specialists within the field." The cover letter should argue that the cross-partisan dialogue literature has been measuring the wrong thing (substantive) and that integrative experimentation is a generalizable framework for studying conversation interventions (methodological). The double argument is what justifies broad significance
• Argue scope without leaning on de Jong (2024) — that paper is in Comms Psych, not NCom. Cite Watts's prior NCom work or related NCom polarization papers if a comparator is needed
• Do not mention prior submissions
• Suggest 3–5 reviewers (diverse in geography, career stage, gender)
• Optionally request reviewer exclusions
• Indicate Wave 2 was preregistered
• Decide on Transparent Peer Review (NCom is opt-out per submission). Default is publish reviewer reports alongside the paper; opting out is a one-click choice in the portal. Default is recommended given the paper's open-data stance
• State preference for double-anonymized vs. single-blind review
• No need to suggest an editor or section — NCom assigns internally and has no section structure

If NCom rejects: cascade to Communications Psychology

• Use the automated manuscript transfer link in the rejection email (it carries reviews of record across)
• Edit cover letter to lead with substantive framing (Comms Psych's editorial bar is rigor + contribution to psychology, not broad significance). Cite de Jong (2024) in Communications Psychology as scope precedent
• Transparent peer review at Comms Psych is default-on, no opt-out — different from NCom
• No other structural changes required; the rewrite work is the same

Deferred to acceptance

• Cut main text to ~5,000 words (currently ~7,000–10,000)
• Reformat references to Nature style (already numbered)
• Overlay individual data points on bar charts in Fig 2 if not already
• High-res figure files as separate uploads (300 dpi, RGB, vector preferred)
• LaTeX/Word formatting to journal spec

[JamesPHoughton]

Re-review findings (after revised paper + cover letter)

Fixes verified (10/11)

All ROPE references, constrained randomization rewrite, methods statements, em-dashes, hyphenation, and most nits read correctly.

Fix incomplete: title singular/plural

• SI title page (p. 38) and SI table of contents still say "cross-partisan conversations" (plural). Cover page is fixed; SI is not.

Cover letter issues to address before submission

• Add a paragraph elevating the methodological reframe. Currently relegated to one sentence at the close. Suggested addition: "Methodologically, this study adapts the integrative-experiment framework to a domain where stimulus selection has been a persistent threat to inference, providing a template applicable across conversation, persuasion, and intervention research."
• Mention Wave 2 was preregistered. One clause is enough.
• State double-anonymized vs. single-blind preference.
• Add: "We have not previously discussed this work with a Nature Communications editor."
• Improve reviewer diversity — currently all 5 suggested reviewers are men, 4/5 at U.S. R1 institutions. Swap 1–2 (suggestions: Erin Rossiter, Emily Kubin, Hannah Schwarz, Yphtach Lelkes, Kristina Lerman).
• Clarify sample size — "805 U.S. participants engaged..." could be misread. Add "(of 1,031 recruited)".
• Fix topic-level vs. dyad-level conflation — cover letter says "none of our dimensions of politicalness predicted which topics worked better. Even the level of disagreement…" but disagreement is dyad-level, not a topic dimension. Reword: "…none of the three topic-level dimensions of politicalness, nor partner-level disagreement, predicted outcomes."

New paper issues found

• Topic-selection number rounding inconsistent between main p.7 ("0.09 / 0.197 / 0.102") and SI p.47 ("0.0907 / 0.196 / 0.102"). Pick one set and round consistently.
• Lorem ipsum text in SI p.75 screenshot — placeholder text in the participant UI. Add a caption clarifying it's intentional ("lorem ipsum is the form's placeholder text shown to participants, not study content").
• Table 1 † footnote dangles — Wave 2 columns are marked preregistered, but with no narrative anchor in Results. This will be resolved when the "Wave 1 exploratory / Wave 2 confirmatory" sentence is added (already pending on the checklist).