The ICA study protocol
Draft protocol, version 0.1, published 3 November 2026. It states what we will test and how, before any study data is collected, so the results can be judged against a plan rather than against hindsight.
Status
This is a draft for review. The thresholds marked as planned will be fixed, and the protocol registered with a public timestamp, before data collection begins. Any change after that will be listed in the changelog below, with its reason.
Question 1: structure
Do responses support four dimensions, and the grouping of Integrative Synthesis and Systematic Analysis under Thinking, with Social Attunement as Feeling and Environmental Engagement as Willing?
- Planned method: a Thurstonian item response model for forced-choice data, comparing the four-dimension model with simpler and alternative structures.
- Planned support: the four-dimension model fits acceptably and better than the alternatives, by fit criteria fixed in the registration.
- Planned problem signal: a simpler or different structure fits as well or better.
Question 2: stability
Do results stay consistent over time?
- Planned method: participants take the ICA twice, several weeks apart. We compare dimension scores and how often the top-ranked dimension and the full profile stay the same.
- Planned support: agreement at or above levels fixed in the registration, judged against published conventions for self-report measures.
- Planned problem signal: profiles change often between sittings for most people.
Question 3: relationships to established measures
Do ICA results relate to established measures in the ways the model predicts?
- Planned measures: a public-domain Big Five inventory from the International Personality Item Pool, and the Rational-Experiential Inventory of thinking styles (Pacini and Epstein, 1999).
- Planned predictions: each ICA dimension is expected to relate to the established scales most similar in content, and less to the others. The full set of predicted directions will be listed in the registration.
- Planned problem signal: relationships that are absent, reversed or no stronger for similar scales than for dissimilar ones.
Participants and consent
- Adults, recruited through partner universities and organizations and through the free ICA, by explicit opt-in only.
- Results are used only as consented, and are analyzed without names.
- Partners contribute data under the terms in research partnerships.
Publication
We will publish the results of every planned analysis, whether or not they support the model, and update each affected line of the claims ledger. The standards we work toward are the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014).
Protocol changelog
| Version | Date | Change |
|---|---|---|
| 0.1 | 3 November 2026 | First public draft |
Questions
Why publish a protocol before the results?
Because a plan fixed in advance stops anyone, including us, from choosing the analysis that looks best after seeing the data.
Why not use a normal factor analysis?
Forced-choice scores always add to the same total, which distorts standard analyses. Methods built for this kind of data avoid the problem (Brown and Maydeu-Olivares, 2013).
Sources
Keep reading
Take the next step.
The draft ICA study protocol: the three questions, planned samples and analyses, and how support or failure will be judged, set before data is collected.