Methodology
What the model is built from, how confident to be in any of it, and exactly where that confidence runs out. This page exists so "not yet validated" is a visible, load-bearing statement, not a footnote.
The model
Resonance measures a four-layer trait model: Orientation (lead ↔ follow, an independent switch-pull measure, and how much control actually changes hands), Domains (~20 activity axes — sensation, restriction, roleplay, service, exposure — with giving and receiving kept separate), Drivers (eleven psychological needs a kink can meet — authority, surrender, intensity, transcendence, catharsis, devotion, ritual, transgression, play, intimacy, exposure), and Modifiers (experience, context-dependence, edge tolerance, arousal-onset, aftercare, attachment). The familiar archetypes — rope bunny, brat tamer, primal — are a derived match against your profile, not something measured directly.
The model and the driver taxonomy are our own design. Published research informs which constructs to measure and how items are worded; it does not hand us a ready-made model to adopt wholesale — no comparable four-layer instrument exists in the literature to borrow from.
Item design
- Balanced keying — roughly half of each dimension's items are reverse-worded, to cancel out "agree with everything" bias.
- Behaviourally anchored where possible ("I've lost track of time during…") over abstract self-labels ("I am submissive"), which are weaker and prime people toward an answer.
- A separate engagement tag (curious / want to try / have done / hard limit) on every item, so what draws you is measured separately from what you've actually done.
- Quality-control items — a planted near-duplicate, a reverse-keyed consistency check, and an instructed-response item — feed a per-layer confidence figure, shown faded on any axis it's low for. Nobody is told they "failed."
Evidence grading
Every factual claim in the Kink Library carries a grade, so a reader always knows how solid a statement is. Grades are never upgraded to make an entry look stronger — a C stays a C.
| A | Representative — a general-population or large representative sample, replicated or consistent across studies. |
| B | Community-sample research — a peer-reviewed study of a self-selected or community sample, not yet representative. |
| C | Theoretical, small, or preliminary — a model, a pilot, a single small study, or a contested finding. |
| C+ | Community consensus — no formal research, but a strong, long-standing consensus among experienced practitioners. |
| D | Forensic or clinical only — death reports or clinical caseloads. Tells you what can go wrong, not how common something is or what the typical experience looks like. |
| — | No research and no settled consensus. The entry is descriptive only, and says so. |
"No evidence" is treated as a valid, printable answer — an entry that says the research on a specific interest is essentially non-existent, and what follows is descriptive, is doing its job correctly.
Current honest status
The instrument has not been formally validated. Prototype vectors and item loadings start from theory and the published literature, not from a factor analysis of real response data — because there isn't enough real response data yet. The plan: once **N ≈ 250** completed individual takes exist, run an exploratory factor analysis on the anonymised response matrix (does the intended structure actually emerge — do Catharsis and Transcendence hold as separate factors, or collapse into one?), item-response-theory checks (which items carry real information), and reliability statistics per subscale. Weak items get pruned, loadings reweighted, and the result ships as a new, clearly version-stamped instrument — old results keep scoring under whichever version produced them, permanently.
A fuller, publishable validation study — pre-registration, a demographic block, a comparison sample, confirmatory factor analysis on held-out data, test–retest reliability, a preprint — is a further step past that internal pass, not a given. Until either exists, treat every result as a structured self-reflection prompt, not a validated psychological measurement.
Research base
Roughly 40 peer-reviewed sources, global (Sweden, Belgium, UK, Canada, Australia, Italy, the US, the Netherlands), inform item wording, the driver set, and the Kink Library's evidence grades. Representative examples: Richters et al. (2008, n=19,307, population BDSM prevalence), Wismeijer & van Assen (2013, established personality/attachment scales against a control group), Herbenick et al. (2017/2022, US-representative prevalence data), Holvoet et al. (2017, Belgian general-population survey), Ambler et al. (2017, flow and altered states in BDSM), Sagarin et al. (2009, cortisol response in scenes), Bancroft & Janssen (the Dual Control Model behind the arousal-onset module).
Nothing here is copied from any existing test or glossary — the model, the copy, and the scoring are original, informed by (not lifted from) the sources above. Citations summarise and link; they don't reproduce article text.