People keep summarising this project back to me as "warmth declines as people get less close", every time I have to say no, that is not the claim.
The claim is narrower and stranger than that.
The facet is not one number. The value a standard inventory returns is an average, and for a lot of people it is the average of a range they never occupy. That is it. That is the whole thesis.
I have no prediction about which direction anybody's ranges run. Mine contract sharply on some traits and run backwards on others — Self-Consciousness and Anger both get stronger toward the people closest to me, not weaker. If yours load entirely on one facet nobody expected, that is data, not a problem with the data.
Why the distinction matters
Because it changes what would count as evidence against the project.
If the claim were "warmth declines with distance", then a respondent whose warmth holds steady would be a strike against it. Under the actual claim, they are not: what matters is whether a single number can stand in for their four readings. Somebody whose entire gradient lands on one unexpected trait is evidence for the thesis — they have a spread the point estimate hides — regardless of which trait it was or which way it went.
I got this wrong myself, in public, about a fortnight ago. Somebody reported that their whole gradient loaded on Self-Consciousness and that their Neuroticism ran higher with acquaintances than with strangers, and my first draft of a reply told them this was a point against my central hypothesis. It is not. It is a spread. It is the thing being measured.
The part that is a real prediction
The instrument does carry directional predictions for eight of its ten facets, and they were written down before any data existed. But those are falsifiability machinery — they exist so the project can fail — not the claim being argued. Confusing the two is how a hypothesis quietly becomes unfalsifiable while sounding more confident.
There is a separate, structural question underneath: do facets differ in how much they gate? Are some of them dispositional and others relational? That one is genuinely open, and the next few entries are about how you build an instrument that could tell you no.