Use when building tables and figures for a Cognitive Psychology (Elsevier) manuscript. Exhibits here carry the experiment-to-model-fit argument — they should overlay model predictions on data, show distributions and uncertainty, and report parameter estimates, not just bars of means. Designs exhibits; it does not run the analysis or fit the model.
SkillScore100Maintained
Automatic
◆Research pack
Last verified 9/1/2026
What it changes
Passed 22 of 22 test cases with the skill, and 17 of 22 without it.
Without skill77%
With skill100%
Checks it passed that the model alone missed
Recommends adding explicit confidence or credible intervals
Plots observed z-ROC data points with uncertainty intervals
Requires reporting penalized fit criteria alongside parameter counts
and 3 more in the full report
5 fixed17 passed either way0 still failing0 broke
Pass-rate lift
+23 pp
Tokens
+60%
Agent turns
0%
Response time
+5%
Model gemini-3.6-flashMeasured 9/1/2026v122 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
I am preparing a primary results figure for a memory experiment destined for a cognitive psychology journal. I currently have a bar chart showing mean accuracy across three delay conditions with standard error bars. How should I adjust the visual design and plot elements to meet venue standards?
What was checked, without and with the skill
Checked
Without skill
With skill
Recommends showing underlying data distributions or points instead of bare bar plots
Passed
Passed
Recommends superimposing model prediction curves directly onto observed data
Missed
Passed
case-01 · gemini-3.6-flash · 9/1/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md114 linessingle file
Tables & Figures (cogpsych-tables-figures)
In Cognitive Psychology the central exhibit usually shows the model fitting the data — observed
patterns with the model's predictions overlaid — because the contribution is the model, not the bare
effect. Exhibits should reveal distributions and uncertainty, report parameter estimates with
intervals, and let a reader judge model comparison at a glance. Bars of means hide exactly what
this venue cares about.
When to trigger
Designing the main model-fit figure or a model-comparison table
Deciding what goes in the article vs. the supplementary material / appendix
A reviewer found an exhibit unclear, or said "show the fit, not just the means"
Visualizing distributions, individual data, model predictions, and uncertainty
Principles
Overlay model on data. The headline figure shows observed data (with uncertainty) and the
model's predicted curve/points superimposed, ideally for the rival model too, so the reader sees
which account tracks the data. This is the venue's signature exhibit.
Show the data and uncertainty. Prefer distributions/individual points with means and
confidence/credible intervals over bar-of-means plots; for model parameters, plot estimates with
intervals.
Make model comparison legible. A table reports each model's fit (AIC/BIC/BF or cross-validated
score), free-parameter count, and the winning criterion — so the comparison is checkable, not
asserted.
Self-contained. Titles, notes, axes, Ns, trial counts, units, and "intervals are 95% CIs/CrIs"
make each exhibit intelligible alone, following the journal's (Elsevier/APA-style) conventions.
Reproducible + accessible. Generated by the deposited model/analysis code so values match;
colorblind-safe and grayscale-legible.
Worked micro-example — the main model-fit figure (illustrative)
For the recognition-memory program, the primary figure must show the fit, not the means.
Figure 1. Observed and model-predicted z-ROCs, Experiments 1-3.
Geometry: observed confidence-ROC points with 95% CIs, UVSD predicted
curve overlaid (solid) and DPSD predicted curve overlaid
(dashed) — the reader sees UVSD track the linear z-ROC.