cogpsych-writing-style

Writing Style (cogpsych-writing-style)

Cognitive Psychology publishes longer, integrative articles, so the challenge is the opposite of a short-report journal: the danger is not the word limit but losing the argument across many experiments and a model. Writing well here means making a multi-experiment program plus a formal model read as one cumulative theoretical argument, in clear scientific prose with rigorous (APA-style) statistical reporting.

When to trigger

• Drafting the Introduction/General Discussion or doing a structural pass
• A series of experiments reads as a list rather than a single argument
• Integrating the model exposition with the empirical sections
• Aligning statistical reporting and references to the journal's style before submission

Writing the long-form, model-driven paper

1. Front-load the theoretical claim. State the question, the rival accounts, and the contribution in the Introduction (see cogpsych-literature-positioning). The reader should know what is at stake before Experiment 1.
2. Make each experiment a step in the argument. Open each experiment with the specific question it answers and what it adds (rules out a confound, tests a divergent prediction); close with a brief bridge to the next. Avoid a flat "Experiment N: Method/Results" list.
3. Integrate the model, don't append it. Introduce the model where it does explanatory work — often alongside the experiments it explains — not as a detached final section the empirical reader skips.
4. General Discussion earns its length. Synthesize across experiments and the model fit; state the theoretical advance and its scope; keep limitations honest. This is where the integrative payoff is delivered, not a restatement of each Results section.
5. Report statistics rigorously. Effect sizes with intervals for behavior, model-comparison criteria for fits, exact test statistics; APA-style author-date citations and reference formatting. Confirm the current style requirements on the official guide for authors.

Structure that keeps a multi-experiment paper coherent

Worked micro-example — making experiments cohere (illustrative)

For the recognition-memory program, the General Discussion turns three experiments into one claim:

Weak: "Experiment 1 found a linear z-ROC. Experiment 2 also found a linear
       z-ROC. Experiment 3 found a linear z-ROC."
Strong: "Across three experiments engineered to make z-ROC shape diagnostic -
       and after ruling out a list-composition confound (Exp 2) and testing a
       further divergent prediction (Exp 3) - the z-ROC was reliably linear,
       the qualitative signature UVSD predicts and DPSD forbids. The single
       continuous-strength account is thus favored not by a fit index alone
       but by a pattern the rival cannot produce."

Cohesion playbook

• Replace "Experiment N also showed X" with "having established X, we next asked whether..." — make the series a chain of questions, not a pile of studies.
• Put procedural detail in Method and lengthy derivations/recovery in an appendix/supplement so the main narrative stays about the theory.
• Replace dichotomous "significant" sentences with estimation/model-comparison phrasing — more precise and more venue-appropriate.
• If the General Discussion restates Results, cut it; lead with the cross-experiment synthesis and the theoretical advance.

Prose-stage reviewer pushback and the venue fix

• "Reads as a list of experiments" → add the question-and-bridge structure; synthesize in the General Discussion.
• "The model feels bolted on" → introduce it where it explains the data; connect each experiment to a model prediction.
• "Stats without effect sizes / model comparison" → report intervals and comparison criteria, not stars.
• "Too long for what it says" → length must buy integration; cut redundant per-experiment recaps.

Anti-patterns

• A multi-experiment paper written as independent, disconnected studies
• A formal model quarantined in a section detached from the experiments
• A General Discussion that restates each Results section instead of synthesizing
• Statistics reported without effect sizes/intervals or without model comparison
• Length used to pad rather than to integrate the program

Output format

【Theoretical claim front-loaded】question + rivals + contribution? [Y/N]
【Experiments cohere】each adds inference, with bridges? [Y/N]
【Model integrated】introduced where it explains, not appended? [Y/N]
【General Discussion】synthesis + advance + scope, not restatement? [Y/N]
【Reporting】effect sizes/intervals + model comparison + APA style? [Y/N]
【Next】cogpsych-open-science-and-transparency

Supplementary resources

• ../../resources/external_tools.md — reference managers, reproducible reporting (Quarto/R Markdown), papaja
• ../../resources/official-source-map.md — article length, style, and reference format