Methodology
How our tests are built, scored — and limited
This page is the site’s anchor of honesty. It explains, for every test we publish, where the questions came from, what model sits underneath, how your answers become a result, and — just as plainly — what a result cannot mean. If a claim appears anywhere on TestLark, its evidence trail starts here.
The six scoring models
Every TestLark test runs on one reusable engine and one of six scoring models. Tests are configuration, not bespoke code, so a scoring rule fixed once is fixed everywhere.
- Likert trait accumulator — statements rated on a five-point scale; each item is keyed to one trait, some are reverse-worded, and the trait score is the keyed average. Used by the Big Five and Dark Triad tests.
- Type accumulator — each answer votes for one type; the result is the full percentage split with a primary type and a blend. Used by the Temperament test.
- Forced-choice accumulator — paired statements force a preference between dimensions; the result is a ranked profile with a percentage split. Built for the love-languages-style tests.
- Dichotomy accumulator — answers push each of several dimensions toward one of two poles; the combination of poles forms the result type. Built for the 16-types test.
- Numeric estimator — each answer carries a numeric weight and the result is the clamped average. Used by the Mental Age test.
- Situational judgement — scenarios whose options carry quality weights per skill area; the result is a per-area percentage of the best available choices. Built for the EQ test.
Where the questions come from
Three sources, and only three. We never copy questionnaires, items or result text from other test sites.
- Public-domain instruments. The Big Five test uses the 50-item Big-Five Factor Markers from the International Personality Item Pool (IPIP, ipip.ori.org), a pool built by researchers and explicitly released for free use. Items and scoring keys are the pool’s own, and we say so on the test page.
- Classical frameworks. The four temperaments descend from Hippocratic and Galenic medicine — a public-domain model two millennia old. The framework is history’s; every question wording is ours.
- Original items on published constructs. For constructs whose reference scales are copyrighted academic work — the dark triad is the launch example — we write our own items measuring the same constructs and cite the literature (Paulhus & Williams, 2002; Jonason & Webster, 2010; Jones & Paulhus, 2014). Constructs are ideas and can be shared; another author’s sentences cannot, and are not.
Per-test sources
Temperament Test
Model: type accumulator · 24 items
Based on the classical four-temperaments model (Hippocratic/Galenic tradition, public domain). All questions were written by the TestLark team for this site; the framework is historical, the wording is ours.
Each of the 24 questions offers four answers, one written for each temperament. Behind the scenes, every answer you pick adds one point to its temperament. There are no right or wrong answers and no answer is 'better' — the four styles are equals with different strengths.
Dark Triad Test
Model: likert traits · 27 items
Constructs: Paulhus & Williams (2002); reference instruments: Jonason & Webster (2010), Jones & Paulhus (2014) — cited as sources for the constructs. All 27 items on this page are original TestLark items written to measure the same constructs; no published scale items are reproduced.
Each trait is measured by 9 statements on a five-point agreement scale. Your answers are averaged per trait and converted to a 0–100 score: below 40 is described as low, 40–65 as average, and above 65 as high, relative to the middle of the scale. Several statements are reverse-worded — agreeing with them lowers a trait score — so simply agreeing with everything produces a middling profile, not a dramatic one.
Mental Age Test
Model: numeric estimator · 20 items
An original TestLark questionnaire. Mental age is a folk concept, not a clinical measure — this test is a playful estimate of the age your answers act like, nothing more.
Each answer in this test carries an age weight — the age that answer most sounds like, from 9 up to 70. Your mental age is the average of the twenty weights your answers collected, rounded and kept inside that 9-to-70 range so one eccentric answer cannot drag the result somewhere absurd.
Big Five Personality Test
Model: likert traits · 50 items
Items: Big-Five Factor Markers, International Personality Item Pool (IPIP), ipip.ori.org — public domain, used freely as the pool permits. Scoring keys follow the published IPIP keys. Interpretation text written by the TestLark team.
Each trait is measured by 10 statements. For straightforward items, your rating counts as-is; for reverse-keyed items (marked in the item pool and listed on our methodology page), the score is flipped — so 'very accurate' on 'I keep in the background' lowers extraversion instead of raising it. Your trait score is the average of its 10 items, converted to a 0–100 scale.
Personality Types Test
Model: dichotomy · 24 items
This test uses an original four-dimension model created by TestLark for this site. It is not the MBTI® and TestLark is not affiliated with the Myers-Briggs Company or with 16personalities. All type names, codes, questions and descriptions here are TestLark's own writing, created for entertainment and self-reflection.
Each of the 24 questions offers two answers, one written for each pole of a single dimension — six questions per dimension, spread through the test. Picking an answer adds one point to its pole. There are no right or wrong answers: every pole is a strength in the right setting, and the test simply counts which way you lean most often.
Enneagram Test
Model: type accumulator · 27 items
The nine-pattern personality tradition has public-domain roots and many modern presentations. TestLark's type names, questions and descriptions are entirely its own writing, created for this site. This test is not the RHETI, and TestLark is not affiliated with the Enneagram Institute.
Each of the 27 questions offers four answers, each written for a different one of the nine patterns, and the options rotate so every pattern appears in every position across the test. Every pattern is offered exactly twelve times in total. Picking an answer adds one point to its pattern. There are no right or wrong answers and no pattern is better than another — each has its own gifts and its own growth edge.
Love Languages Test
Model: forced choice · 20 items
The idea that people prefer different ways of expressing and receiving love was popularised by Gary Chapman's books on the five love languages. TestLark is not affiliated with Gary Chapman or Moody Publishers, and this site carries a non-affiliation disclaimer to that effect. The style names used here (Kind Words, Helpful Hands, Thoughtful Tokens, Unhurried Time, Warm Closeness), their descriptions and all 20 questions are TestLark's own original writing.
Each of the 20 questions offers two ways a partner could show love, and each choice adds one point to that style. The ten possible pairings of the five styles each appear twice — once in each option order, with different scenarios — so no style benefits from going first, and every style can earn a maximum of 8 points. There are no right answers; a low score does not mean a style means nothing to you, only that you reach for it after the others.
Attachment Style Test
Model: type accumulator · 24 items
Informed by attachment research in the Bowlby and Ainsworth tradition and the two-dimension model behind the ECR questionnaire (Fraley and colleagues). All items on this page are original scenario questions written for TestLark; no published questionnaire items are reproduced. This test is a self-reflection tool, not a clinical assessment.
Attachment researchers commonly describe adult attachment with two underlying dimensions. The first is closeness anxiety: how strongly you monitor a bond for signs of loss, and how much reassurance you need to feel settled in it. The second is closeness avoidance: how comfortable you are depending on someone and being depended on, versus keeping your needs private and self-managed. Every style in this test is a combination of those two dials.
EQ Test
Model: situational · 20 items
An original situational-judgment test written for TestLark, informed by the four-branch ability model of emotional intelligence (perceiving, using, understanding and managing emotions — Mayer & Salovey) as a concept only. It is not the MSCEIT or any other commercial EQ assessment, and no published test items are reproduced. The answer key is TestLark's own editorial key, written for an entertainment and education instrument.
Each scenario has one strongest response (scored 3), one good-but-incomplete response (2), one well-meaning but unhelpful response (1), and one response that dismisses, escalates or avoids in a harmful way (0). Your choices are added up separately for each of the four areas, and each area is shown as a 0–100 score: 40 or below is described as low, 41–65 as average, and above 65 as high. The positions of the strongest answers are rotated through every slot, so there is no pattern to game — only the situations to read.
Light Triad Test
Model: likert traits · 24 items
Constructs: the light triad — faith in humanity, humanism and Kantianism — described by Kaufman and colleagues (2019) as the positive counterpart of the dark triad. TestLark uses its own dimension names for these concepts (Everyday Compassion, Everyday Integrity, Valuing People) and all 24 items on this page are original TestLark items written for this site; no published scale items are reproduced. For entertainment and self-reflection.
Each tendency is measured by 8 statements on a five-point agreement scale. Your answers are averaged per tendency and converted to a 0–100 score: 40 or below is described as low, 41–65 as average, and above 65 as high, relative to the middle of the scale. Several statements are reverse-worded — agreeing with them lowers a score — so simply agreeing with everything produces a middling profile rather than a glowing one.
Career Test
Model: type accumulator · 24 items
The six interest areas come from John Holland's RIASEC model as used by the U.S. Department of Labor's O*NET Interest Profiler, which is U.S. government public-domain material. All activity items, descriptions and result profiles on this page are TestLark's own writing. This test is a career-exploration aid, not vocational advice or a placement service.
Each of the 24 questions offers four activities, and every activity belongs to one of the six interest areas. Behind the scenes, every choice you make adds one point to its area. Across the whole test each area is offered exactly 16 times, so no area has a head start and your result reflects your preferences alone.
Color Personality Test
Model: type accumulator · 24 items
An original TestLark test. Colour–personality associations are a popular cultural and entertainment framework, not a validated scientific instrument — and we say so cheerfully and honestly. All questions, profiles and descriptions were written by the TestLark team for this site, for fun and self-reflection.
Each of the 24 questions offers four answers, and every answer belongs to one of the six colour profiles. Behind the scenes, every choice you make adds one point to its colour. Across the whole test each colour is offered sixteen times, so no colour has a head start and your result reflects your own taste and style alone.
How we keep results honest
- Reverse-keyed items. Trait tests include statements worded in the opposite direction, so agreeing with everything cannot manufacture an extreme profile.
- Bands, not fake percentiles. Where a test shows low/average/high bands, the bands are cut points on the answer scale itself. We do not present population percentiles we do not have the norming sample to support.
- Full interpretation, always. The result screen shows the complete written interpretation — never a score with the explanation paywalled.
- No ratings theatre. We publish no star ratings, review counts or “accuracy percentages” for tests, because we have no evidence base that would make such numbers true.
The limits — read this part too
These tests are self-reflection tools. Self-report has real limits: mood, honesty, self-knowledge and the wish to look a certain way all move scores, and a 24-to-50 question instrument is a sketch, not a portrait. Research instruments like the IPIP pool are validated for group research and personal insight; none of the tests here are clinical assessments, none diagnose any condition, and none should be used for hiring, admissions, clinical or other high-stakes decisions. Where a test is openly playful — mental age is the clearest case — the page itself says so.
If you are struggling with your mental health, a quiz is the wrong tool entirely: please talk to a qualified professional or a trusted person in your life. And if you spot an error in a question, a score or a source on this page, tell us — corrections are part of the methodology, not an embarrassment to it.