The most useful document in this project is the one written to attack it. It is a literature map kept deliberately sceptical of my own construct, on the theory that a reviewer will be, and it is cheaper to find the problems myself.
Here is what it says I did not invent.
Clifton (2014)
Participants built a thirty-person social network and rated their own Big Five as experienced with each member. The finding: people report being more extraverted, more neurotic and less conscientious with more central — that is, closer — members.
That is a closeness gradient in Big Five expression, published, twelve years ago. It is also where my prediction that Anger runs backwards comes from. I did not derive it; I borrowed it.
And the rest
Self-concept differentiation has measured how much people vary across roles since the nineties. Social discounting has measured how generosity falls with social distance since the 2000s. Cognitive-affective personality systems — the if-then signature — has been the field's account of situational conditionality since 1995. Rating the same traits across relational targets is old.
There is even a cautionary precedent for exactly the mistake I might be making. Higgins' self-discrepancy theory was a beautiful difference-score theory that did not survive measurement: later work found no emotion-specificity, and the discrepancies added little over plain negative self-concept. A tidy construct built on a difference between two correlated self-reports has failed before, in a way that looked good for years first.
What might actually be left
Four things, and I hold all of them loosely:
- Facet-level resolution. The prior work is at domain level or adjective level. A thirty-facet map conditioned on target does not exist that I can find.
- Signed, directional profiles. Self-concept differentiation discards direction by construction — it measures how much you vary, not which way. Direction is exactly the information I want.
- Closeness as the ordering axis, rather than social role. Roles are categories; closeness is ordered, and ordering is what makes a slope meaningful.
- Validating a cheap questionnaire against experience-sampled behaviour. Nobody has done it. It is the highest-value and lowest-glamour contribution available here, and it is what would make the instrument credible rather than merely interesting.
Why say all this out loud
Because overclaiming is the fastest available way to lose the audience I most want. The people who could tell me this is wrong are the people who already know Clifton, and a page that quietly implies novelty it does not have reads to them as either ignorance or salesmanship. Neither gets you a second reading.
Noticing that a widely used instrument mismeasures a real pattern is a contribution. Showing that the replacement is reliable, valid and adds something is a different contribution, and it has not happened yet.