Most personality tests assume data follows a perfect Gaussian bell curve. Real data doesn't. We count.
When you're told you're at the 73rd percentile of Extraversion, what does that mean? With z-scores, it means a mathematical formula assumed symmetric distribution and estimated your position. With Afini's empirical percentiles, it means we counted: exactly 73 out of 100 people in your reference group scored below you. No assumptions. No shortcuts. The difference seems subtle — but in skewed distributions it can be several percentile points.
The normality assumption is false for most real psychometric distributions. Personality traits show skewness and kurtosis, especially in facets like Depression, Impulsiveness, or Excitement-Seeking. A z-score assumes the 84th percentile is always exactly +1 standard deviation from the mean. In skewed distributions, this is simply incorrect.
Afini calculates the individual's real position in the observed distribution. If 73 out of 100 people scored below you, your percentile is 73. Period. For scores between observed percentiles, we interpolate linearly — more conservative than parametric interpolation, but free of distributional bias. All scores are also transformed to T-scores (mean = 50, SD = 10) for cross-scale comparability.
Maximum facet resolution (10 items/facet). Stratified by Sex × 8 age groups. Primary reference for the 300-item test.
5-factor structure confirmed by Kajonius & Johnson (2019) on N = 320,128. Reference for the 120-item test.
Global dataset: 93 countries × 10 linguistic groups. Reference for 60-item test + geographic/cultural comparison.
Your PCP (Portable Cognitive Profile) automatically selects the most precise population: Johnson-300 > Johnson-120 > Open-FFM > language/country subgroup > cubic fallback.
A z-score converts raw score to distance from the mean in standard deviations. It's elegant, compact — and assumes Gaussian distribution. In facets like Depression (where most people score low and distribution is heavily skewed), a z-score can place you at the 84th percentile when your real position is the 79th. Five percentile points of error isn't noise — it's the difference between "slightly above average" and "clearly above."
Afini uses empirical percentiles with the same rigor demanded in high-precision clinical psychometrics. Norms are stratified by sex and age group, with comparisons across 93 countries and 10 linguistic groups — always with the caveat that cross-cultural percentiles should be interpreted with skepticism (Heine et al., 2002, "reference group effect").
Key references
Johnson, J. A. (2014). Measuring thirty facets of the FFM with a 120-item public domain inventory. J. Research in Personality, 51, 78-89. · Kajonius, P. J. & Johnson, J. A. (2019). Assessing the structure of the FFM in a Swedish population. · Heine, S. J. et al. (2002). What's wrong with cross-cultural comparisons of subjective Likert scales?
Every dimension of your Portable Cognitive Profile (PCP) includes raw score, T-score, empirical percentile, descriptive level, and confidence interval. It's not a loose number — it's a position calibrated against the real distribution of hundreds of thousands of people.
"Your Neuroticism score is 72. This places you approximately at the 84th percentile (assuming normal distribution)."
"Your Neuroticism score is 72. In the empirical distribution of your reference group (female, 25-34, N = 89,412), this places you at the 79th percentile — slightly above average, not at the extreme a z-score would suggest. 95% CI: 75th-83rd percentile."
Five percentile points of difference. Enough for AI to adjust its emotional sensitivity level — neither too cautious nor too direct.
Your PCP is built on empirical percentiles calculated against 1.46 million respondents. No statistical shortcuts, no distributional assumptions. The most honest measurement possible.