Porn Use & the Sex Perspective Survey

The setup

Respondents rated 0–100 sliders in a fixed scenario: third time having sex with a new partner you like; the act hasn't been discussed beforehand. The design is asymmetric, which is the analytical gift: men's answers are beliefs about women, women's answers are ground truth (for this sample). So porn use gets tested two ways — for women, against what they actually like; for men, against miscalibration: the gap between belief and reality.

From 21,773 raw rows: removed non-cis/nonbinary respondents (the survey's branching makes their answers unanswerable-by-design), gay/lesbian respondents where detectable, repeat takers, joke ages, dropouts (<20 items), straightliners, and speeders. Two data bugs repaired: a variable written by two different questions, and duplicate-joined cell values. All adjusted results control for age, log lifetime partners, and political lean — the confounders that actually correlated with porn use (porn-using women are younger, r = −.31, and more liberal, r = +.22).

Men cluster at the top (76% several-times-a-week or more); women spread across the whole range. Among men, frequency therefore has little variance — the share of violent porn becomes the more informative variable for them.

FINDING 1Women who use porn more like kinky acts more — not everything more

Tender and communicative acts sit at ~77/100 for everyone. What porn frequency tracks is specifically the rough/degrading cluster — and the top per-item correlates read like a porn script: being called a dirty whore (+.28), pussy slapping, blowjobs, facials, light face slaps, consensual nonconsent, firm choking. The small negative end: sensual slow sex, an eager-to-please partner, explicit consent checks.

The gradient survives additionally controlling for violent-porn share (r = +.17). Earlier porn start and paying for porn also predict higher eagerness; written-vs-visual preference predicts nothing.

FINDING 2The average man underestimates women's kink — and porn frequency barely changes his beliefs

Comparing men's mean predictions against women's mean self-reports: men's dominant error is pessimism about kink, not porn-fueled optimism.

FINDING 3Violent porn content is the strong signal — for both sexes

Women who watch mostly-violent porn rate rough acts ~70/100 vs ~45 for porn users with no violent content — and it's the only porn variable that negatively predicts liking tenderness.

FINDING 4Matched on porn use: his predictions are flat, her tastes are not

Compare men and women at the same porn-use level — the calibration that matters if people date partners with similar habits. (Porn frequency is discrete and distributed very differently by sex, so exact decile slices don't exist; below are matched frequency bins, then each level plotted at its within-sex percentile midpoint.)

Men's predictions barely move with their own porn use (47 → 50); women's actual liking climbs 43 → 62. So Finding 2's "men underestimate women" is entirely concentrated in the porn-matched heavy users: a non-porn man is slightly too optimistic about a non-porn woman (+4.8), while a daily+ man underestimates a daily+ woman by ~13 points.

Read it this way: a woman's porn use carries strong information about her tastes; a man's porn use carries almost none about his beliefs about women's tastes. The heaviest male users are the most wrong about their female counterparts — in the pessimistic direction.

FINDING 5Your audience vs a paid general-population panel

The dataset turned out to contain a paid Positly batch (~1,100 panelists, task pa2999018, fielded May 23–30 2023, tagged by panel-provided device/education/gender/age fields the survey never asked). They're general-population: ~50/50 gender, older (mean 41 vs 34), and much lighter porn users. 866 survived cleaning. Splitting the eagerness curve four ways — gender × audience — lets us see whether the porn gradient is a quirk of the Twitter crowd. Here the Positly lines are adjusted (reweighted to your audience's age, partner count & political lean), so the comparison isn't just "they're older"; the faint dotted lines show raw Positly for reference:

Adjustment is OLS regression standardization to your-audience covariates. We confirmed separately that the audience gap survives full demographic adjustment (~7 points residual) and holds at every age band; demographics explain only ~a quarter of it. Positly women thin out at high porn use (daily+ omitted for n<15), so the right end of their curve is noisier. See audience_split_adjusted.py.

FINDING 6Three samples, one question: how much does "who answered" matter?

To sanity-check the audience effects above, here's the porn-frequency item compared across three independent straight-cis samples who answered the word-for-word identical question (same stem, same "for the purposes of getting aroused" qualifier, same 10 options, same 0–9 coding): your sex-perspective audience, the paid Positly panel, and the Big Kink Survey (n ≈ 390k cis+straight). Frequency collapsed to Daily (8–9), Weekly (6–7), and Monthly or less (0–5).

Men

Women

BKS population raking weights barely move its rates (men 41.3→41.8%, women 15.6→16.7% daily). "Positly (adjusted)" is g-computation-standardized to your audience's age, partner count & politics; raw Positly is lower still (men 17%, women 3%). BKS matched to cis + straight to mirror the sex-perspective sample.

Caveats