Two surveys. The tree’s overall shape comes from the Mini Kink Survey — a compact set of 105 broad kink items. The fine detail (the ~500 leaves) comes from the much larger Big Kink Survey (~1.05 million responses), whose specific items are mapped onto the matching mini-kink category.
The clustering. The 105 mini-kink items were grouped by how often the same people endorse them: a polychoric correlation between every pair → a distance (1 − |correlation|) → hierarchical clustering (Ward’s method). That builds the trunk and main branches. Then, within each mini-kink category, its Big Kink Survey sub-items are clustered the same way (here using Pearson correlations) and grafted on as the twigs and leaves. A graded-response (IRT) model picked each cluster’s most representative items for naming.
The region names are written by hand to summarize what’s on each major branch — the branches are data-driven, the names are editorial. The branching layout is drawn for legibility and carries no meaning beyond the cluster structure.
The colors. In cluster mode, color is assigned by position in the tree, so similar kinks share similar hues. In gender mode, each kink is colored by the share of the people who endorse it who are men. The sample is ~38% male, and because men endorse more items the average kink’s fans are ~45% male — so colors show whether a kink leans more male (blue) or female (pink) than the average kink. Toggling Doms or Subs recomputes that split among only people aroused by dominating or submitting (from the survey’s dominance/submission items), each centered on that group’s own average.
Responses are self-reported fantasy and arousal, not behavior. Many items are fantasy themes only; items involving minors, animals, or non-consent are research categories about fantasy and are not endorsements of real-world acts.