Methodology

Which instruments, and why these

Every instrument is a published measure of a trait or an interest whose licence allows any use: the International Personality Item Pool (public domain), the Ten-Item Personality Inventory (free for any use by its author), the Rosenberg Self-Esteem Scale (public domain) and the O*NET Interest Profiler Short Form (Creative Commons Attribution-NoDerivatives, reproduced without change). No instrument with a research-only, non-commercial or unclear licence is used, and nothing that screens for a condition. Items are reproduced character for character and the instrument is named as its authors name it.

How each instrument is scored

Scoring is a pure function in the open engine: the instrument file lists each item with its scale and its key (+1 or -1), and the response scale. A positively keyed item scores the option number chosen, counting from the lowest-scoring option; a negatively keyed item scores points plus one minus that number. A scale score is the sum (or for the TIPI the mean) of its items. A domain made of facets, as in the IPIP-NEO-120, sums all its facets' items. The O*NET profiler is a tick list: each tick counts one, exactly as the paper form is self-scored. The Rosenberg scale is published with Strongly agree first, so the option order is flipped before keying, and the site scores it from 10 to 40, one of the two conventions its owner describes. Nothing is weighted, transformed or adjusted.

A test in the repository asserts every file against its published key: the item count, each item's scale, and the set of reverse-keyed items. Worked examples (all highest, all lowest, all neutral, the pattern that reaches the maximum on every scale) are checked for each instrument.

Where the norms come from

No norm is estimated or invented. A norm table is a count of people at each possible raw score, per scale and per group, generated once by a committed script from a named public dataset. The script records the SHA-256 of the input file, so anyone can download the same file and reproduce the table. Instruments without a published dataset show raw scores against the possible range and say so on the page.

IPIP-NEO-120

Johnson, J. A. IPIP-NEO-120 public dataset (IPIP120.dat), OSF component wxvth of osf.io/tbmh5. https://osf.io/wxvth/. Generated 2026-11-03 by scripts/norms/ipip-neo-120.py; input SHA-256 526daf7ebe7d480ba71258cd20f0fc3b37ab3a43a2e4191c19a49e185a1df53b.

619150 rows read; 208774 dropped for a missing item response (coded 0) or a short line; 410376 kept. Age bands use the AGE field; the UK group uses the COUNTRY field. No licence is stated on the OSF component; the data are the public dataset the IPIP-NEO-120 paper describes.

The age bands cover people aged 18 and over who gave an age; the rest of the sample gave no age or one under 18 and are in the everyone group only, which is why the bands sum to fewer than the whole.

How percentiles are calculated

A percentile rank is the share of the norm group scoring below the person plus half the share scoring the same, times 100: the mid-rank definition, so a score in the middle of a symmetric sample is the 50th. The strip draws the share of the group at each raw score as a curve, with the middle 50% (the 25th to the 75th percentile scores) darker. Curve height is relative to a flat distribution over the same scores, so spread can be compared between strips of the same instrument. By default a result is compared with everyone in the sample; the tables also hold age bands and country groups where the dataset records them.

Reliability per scale

Published figures come from the paper or key page the instrument file names, and each file records where. Where the generating script could compute coefficient alpha from the norm dataset itself, that figure is shown separately and labelled as computed, never in place of a published one.

Assertiveness

Coefficient alpha as given on the key page, https://ipip.ori.org/newNEOKey.htm, accessed 3 November 2026.

Courage

Coefficient alpha as given on the key page, https://ipip.ori.org/newVIAKey.htm, accessed 3 November 2026.

Creativity

Coefficient alpha as given on the key page, https://ipip.ori.org/newVIAKey.htm, accessed 3 November 2026.

Curiosity

Coefficient alpha as given on the key page, https://ipip.ori.org/newVIAKey.htm, accessed 3 November 2026.

Emotional stability

Coefficient alpha as given on the key page, https://ipip.ori.org/newBigFive5broadKey.htm, accessed 3 November 2026.

Extraversion

Coefficient alpha as given on the key page, https://ipip.ori.org/newBigFive5broadKey.htm, accessed 3 November 2026.

IPIP Big-Five Factor Markers

Coefficient alpha for each 10-item scale as reported on the IPIP key page (https://ipip.ori.org/newBigFive5broadKey.htm), accessed 3 November 2026. No test-retest figure is published there.

IPIP emotional intelligence components

Coefficient alpha per scale from the IPIP key page, which gives two figures per scale (women and men); the lower of the two is stored here. Women/men: positive expressivity .81/.79, negative expressivity .75/.74, attending to emotions .81/.83, emotion-based decision-making .73/.67, responsive joy .73/.70, responsive distress .59/.63, empathic concern .64/.80.

IPIP Interpersonal Circumplex

No single reliability figure is published for this instrument in a form the file can carry; the test page quotes what its source says.

IPIP-NEO-120

Coefficient alpha per facet from the IPIP key page (https://ipip.ori.org/30FacetNEO-PI-RItems.htm, internet sample N = 619,150), accessed 3 November 2026. The key page gives no domain alphas; the norms file records domain alphas computed from the same public dataset by scripts/norms/ipip-neo-120.py. Cronbach's alpha from the 410376 complete rows, computed by the generating script. Published facet alphas are in the instrument file.

IPIP-VIA-R

Coefficient alpha per scale for the UK sample (n = 474) from the IPIP key page, which gives Germany/UK figures for alpha, McDonald's omega and two-to-three-week test-retest reliability. The authors say alpha understates the reliability of these balanced four-item scales and recommend omega.

O*NET Interest Profiler Short Form

No single reliability figure is published for this instrument in a form the file can carry; the test page quotes what its source says.

Orderliness

Coefficient alpha as given on the key page, https://ipip.ori.org/newNEOKey.htm, accessed 3 November 2026.

Patience

Coefficient alpha as given on the key page, https://ipip.ori.org/newHEXACO_PI_key.htm, accessed 3 November 2026.

Perseverance

Coefficient alpha as given on the key page, https://ipip.ori.org/newVIAKey.htm, accessed 3 November 2026.

Rosenberg Self-Esteem Scale

No single reliability figure is published for this instrument in a form the file can carry; the test page quotes what its source says.

Ten-Item Personality Inventory

Coefficient alpha per scale from Gosling, Rentfrow and Swann (2003), Table 1 (N = 1,813). The same table gives six-week test-retest reliabilities of .77, .71, .76, .70 and .62. Two-item scales are expected to have low alpha; the authors recommend judging the TIPI by its test-retest reliability and its correlations with longer Big Five measures.

Trust

Coefficient alpha as given on the key page, https://ipip.ori.org/newNEOKey.htm, accessed 3 November 2026.

The known limits of self-report

Every instrument here asks people to describe themselves. That measures self-image as much as behaviour, is shaped by mood on the day and by how the person wants to be seen, and can be faked in either direction by anyone who wants to. The norm samples are people who chose to take a test online and are younger, more female and more English-speaking than any national population; they are not representative samples, and the site says so on every page that uses them. Short scales are coarse: a two-item TIPI scale or a four-item facet moves more between attempts than a 24-item domain. Interests are not abilities, traits are not destinies, and a percentile is a comparison with a sample, not a judgement.

What is stored where

Responses and results are stored only in the browser's local storage on the device used, under versioned keys, so a changed instrument file never mis-scores an old attempt. Nothing about answers or results is sent to this site: there is no server-side scoring, no account and no cookies. Page views are counted with Google Analytics in its consent mode with storage denied, as the privacy page says. A share link encodes the scores in the link itself, after the hash, which browsers never send to a server; whoever receives the link sees the scores. Deleting the browser's site data deletes everything.

Teachers and researchers: the page for teachers and researchers says how to use the instruments in teaching or research, with the licence for each, and how to cite the site and the instruments.