A temperament test is worth exactly as much as the method behind it. So here is ours in full: where the four types came from, what modern research supports and what it does not, how your answers turn into eight scales, two axes and four percentages, and where our instrument stops.
The four types are old, and their original explanation was wrong. Both halves of that sentence matter, so here is the short history without the usual reverence.
In the Hippocratic tradition and later in Galen, health and character were explained by four bodily fluids, and a person's character came from their mixture. The word temperament itself means a mixture, and that is the only part of the idea that stayed.
Kant, in his lectures on anthropology published in 1798, kept the four names but described them as patterns of feeling and activity rather than as fluids. The four types quietly became a way of describing behaviour instead of physiology.
Wundt made the decisive move: he arranged the same four types along two dimensions, the strength of emotional response and the speed with which states change. That is the structure we still use, and it is why our model has two axes.
Pavlov's work on the properties of the nervous system, strength, balance and mobility, gave the four types a physiological reading again, and in several countries this is still how temperament is taught at school. We name that tradition, and we do not borrow its instruments.
As physiology, the theory of four fluids is dead and has been for a long time. What survived is the observation underneath it: people differ in the intensity and speed of their reactions, stably and from an early age. We build on the observation, not on the explanation. Stelmack and Stalikas, 1991
Four findings from research on temperament that our instrument leans on. Each of them is a statement about temperament in general, not a claim that our test is validated. That difference is the subject of section ten.
Infants differ in how strongly and how fast they react to anything unfamiliar, and those differences are measurable in the first years of life and predict behaviour years later. This is where the idea of temperament as inborn dynamics comes from.
Kagan et al., 1988A meta-analysis of twin and family studies puts the heritability of personality traits at about forty percent. Inherited does not mean fixed: the same number says that the majority of the variance comes from elsewhere.
Vukasović and Bratko, 2015Rank-order consistency, meaning whether the same person stays higher or lower than others, rises steadily with age and is already substantial in childhood. Average levels still shift across the lifespan, so a result is a snapshot of a stable tendency, not a sentence.
Roberts and DelVecchio, 2000In the formal and dynamic tradition, temperament covers energy, tempo, reactivity, endurance and similar properties: the manner of behaviour rather than its content, values or goals. Our eight scales are built in that language.
Strelau, 1996What none of this proves: that humanity comes in exactly four kinds. The evidence supports measurable, partly inherited, fairly stable differences in the dynamics of behaviour. The four temperaments are a readable map drawn over those differences, and we use them as a map, not as a discovery.
Everything on your result is computed from these eight numbers. Each scale is 20 statements, roughly half of them worded in reverse so that agreeing with everything gets you nowhere. Both ends of every scale are normal: high is not better.
How much activity a day asks of you and how long you can hold that level.
How much you seek contact with people and how much effort contact costs you.
The speed of speech, movement and switching from one thing to the next.
How easily you change plans, routines and the way something is usually done.
How quickly and how strongly a feeling starts when something happens.
How fast you come back to your usual state after a hard episode.
The level of noise, light, tone or detail at which you begin to react at all.
How long you keep effort on one task after it stops being interesting.
Two of the eight, Flexibility and Persistence, stay out of the axes on purpose. They describe how a profile is coloured rather than where it sits on the map, and mixing them into the axes would blur the very thing the axes are for.
Four steps, no others. If you want to check the arithmetic on paper, everything you need is here.
166 statements on a five point agreement scale. Nothing is skipped, and no answer is weighted more than another.
Each scale gives a raw sum from 20 to 100, and reversed statements are flipped first. The raw sum becomes a position on the scale: pct = (sum - 20) / 80 * 100.
Activation is the mean of Energy, Sociability and Tempo. Stability is the mean of Recovery and the inverted Reactivity and Sensitivity. Both end up on the same 0 to 100 field.
Each temperament gets the product of its two conditions: sanguine is activation times stability, choleric is activation times the opposite of stability, phlegmatic and melancholic take the mirrored pair. The four products are then scaled to add up to 100.
In symbols, with A for activation and S for stability, both from 0 to 100: w(sanguine) = A x S, w(choleric) = A x (100 - S), w(phlegmatic) = (100 - A) x S, w(melancholic) = (100 - A) x (100 - S). The four weights are normalised to 100 and rounded so that the percentages always add up exactly.
There is no step in between where a human decides anything. The same answers always give the same result, your share link carries the raw sums rather than the conclusions, and an old link keeps working even if we later adjust a threshold.
A two axis model cannot produce every combination people expect, and pretending otherwise would be the first place to cheat. So here is the consequence, stated plainly.
Horizontal: Activation. Vertical: Stability.
Sanguine and melancholic sit in opposite corners of the field: one is high activation with high stability, the other is low activation with low stability. To be strongly both, a person would have to be on both sides of both axes at once. The arithmetic says the same thing: sanguine outranks choleric exactly when stability is above the middle, and melancholic outranks phlegmatic exactly when it is below. Those conditions cannot hold together, so if sanguine leads your formula, melancholic is never second.
What remains is four pure types, eight adjacent pairs in a fixed order, and one balanced profile for people close to the centre. Thirteen identities, all reachable, none invented to make the list longer. A profile counts as pure when the leading temperament is at least 14 points ahead of the second, and as balanced when the distance from the centre is smaller than 7 points.
The thirteen profilesLevels are the place where tests quietly start lying, so this is what our labels do and do not mean.
The lower end of the scale as this test defines it. It describes a pole, not a deficiency, and both poles have their own costs and advantages.
The widest band, and the most common place to land. A middle score means the property does not stand out in either direction, which is information, not a missing answer.
The upper end of the same scale. It says the property is pronounced, and it says nothing at all about whether that is good.
These positions are measured against the scale of this test, not against a population. We do not show percentiles, and we will not show them until we have enough completed and validated protocols to build norms honestly. When norms appear, the wording on your result will change to say exactly who you are being compared with: people of the same age and sex who took this test, not humanity.
Six of the 166 statements are not scored. They exist to tell us whether a set of answers can be read at all.
A few statements have an obvious intended answer. Missing one is a normal slip, missing several means the questionnaire was being clicked through rather than read.
Some statements come in pairs that say the same thing in different words, or the exact opposite. Answers that contradict each other by a wide margin are counted as a failure of that pair.
Two or more failures raise a reliability flag. Your result still opens in full, the flag is shown next to it in plain words, and the paid analysis is told not to build confident conclusions on shaky data.
We do not block anyone, and we do not hide the flag from the person it concerns. What we also do not do is store your individual answers as a public record: what leaves your browser in a share link is the eight raw sums, the flag and a checksum.
What we store and for how longFive things, written down so that nobody has to guess where our claims end.
This is not a clinical instrument, and none of our scales is a symptom. If something in your life feels beyond your control, a questionnaire is not the tool for it, and neither is a paid report from us.
Temperament describes the environment that costs you less energy, not the job title you should hold. Any test that lists occupations from four types is selling certainty it does not have.
A quantitative review of taxometric research found that personality differences are almost always continuous rather than categorical. Our four names are regions of a continuous field, which is why your result is a formula with percentages and not a label.
Haslam et al., 2012People accept vague, flattering descriptions as personally true, a demonstration first run in a classroom in 1949 and repeated ever since. That is why our text is tied to your numbers and why we tell you which scales produced each statement.
Forer, 1949Self-report is accurate for internal and easily observed traits and weaker for the ones your friends judge better than you do. That limit applies to every self-report instrument, ours included.
Vazire, 2010The honest position for a new instrument is not a claim of validation. It is a list of what has been done, what has not, and what will be published.
The item bank was written to explicit rules, with reversed statements on every scale and no double barrelled wording. The formula was tested on 100000 simulated response sets: the profile survives small answer changes in about 96 percent of cases, and no change of two answers ever moves a profile by more than one neighbour.
We have no population norms, no published internal consistency figures and no external validity study. Internal consistency will be reported per scale as the classical coefficient once enough valid protocols have accumulated, and it will be reported whatever the numbers look like.
Derived statistics only: norm tables, reliability coefficients, correlations between scales. Never individual protocols, and never data that could identify a person. Only sessions that pass the validity checks go into the numbers.
If a competing test states its accuracy as a single percentage, that number is either about something else or about nothing. There is no such figure for personality instruments, and we will not invent one for ours. Cronbach, 1951
Every reference here is a real publication, and every link is a DOI that resolves to it. Each one supports a specific statement on this page, and none of them says anything about our test.
Stelmack, R. M., Stalikas, A. (1991). Galen and the humour theory of temperament. Personality and Individual Differences.
The history of the four types and of the physiology that was abandoned.
10.1016/0191-8869(91)90111-NOpen the paperKagan, J., Reznick, J. S., Snidman, N. (1988). Biological bases of childhood shyness. Science.
Differences in reactivity to the unfamiliar are measurable early and persist.
10.1126/science.3353713Open the paperRothbart, M. K. (2007). Temperament, development, and personality. Current Directions in Psychological Science.
How temperament is defined in current research and how it relates to personality.
10.1111/j.1467-8721.2007.00505.xOpen the paperZentner, M., Bates, J. E. (2008). Child temperament: an integrative review of concepts, research programs, and measures. International Journal of Developmental Science.
A review of what different research programmes actually mean by temperament.
10.3233/DEV-2008-21203Open the paperStrelau, J. (1996). The regulative theory of temperament: current status. Personality and Individual Differences.
The formal and dynamic language our eight scales are written in.
10.1016/0191-8869(95)00159-XOpen the paperTrofimova, I., Robbins, T. W. (2016). Temperament and arousal systems: a new synthesis of differential psychology and functional neurochemistry. Neuroscience & Biobehavioral Reviews.
Why energy, tempo and endurance are treated as separate properties rather than one.
10.1016/j.neubiorev.2016.03.008Open the paperVukasović, T., Bratko, D. (2015). Heritability of personality: a meta-analysis of behavior genetic studies. Psychological Bulletin.
The heritability figure of about forty percent quoted in section three.
10.1037/bul0000017Open the paperRoberts, B. W., DelVecchio, W. F. (2000). The rank-order consistency of personality traits from childhood to old age: a quantitative review of longitudinal studies. Psychological Bulletin.
Stability of relative position across the lifespan, and its limits.
10.1037/0033-2909.126.1.3Open the paperCaspi, A., Roberts, B. W., Shiner, R. L. (2005). Personality development: stability and change. Annual Review of Psychology.
Average levels of traits keep shifting with age, so a result is a snapshot.
10.1146/annurev.psych.55.090902.141913Open the paperDigman, J. M. (1997). Higher-order factors of the Big Five. Journal of Personality and Social Psychology.
Evidence that broad traits group into two higher order factors, as our axes do.
10.1037/0022-3514.73.6.1246Open the paperDeYoung, C. G. (2006). Higher-order factors of the Big Five in a multi-informant sample. Journal of Personality and Social Psychology.
The same two factor structure confirmed with ratings from other people.
10.1037/0022-3514.91.6.1138Open the paperHaslam, N., Holland, E., Kuppens, P. (2012). Categories versus dimensions in personality and psychopathology: a quantitative review of taxometric research. Psychological Medicine.
Personality differences are continuous, which is why we report percentages.
10.1017/S0033291711001966Open the paperGerlach, M., Farb, B., Revelle, W., Nunes Amaral, L. A. (2018). A robust data-driven approach identifies four personality types across four large data sets. Nature Human Behaviour.
Data driven types can be found, and they are dense regions rather than boxes.
10.1038/s41562-018-0419-zOpen the paperForer, B. R. (1949). The fallacy of personal validation: a classroom demonstration of gullibility. The Journal of Abnormal and Social Psychology.
Why a description can feel accurate and still mean nothing.
10.1037/h0059240Open the paperVazire, S. (2010). Who knows what about a person? The self-other knowledge asymmetry (SOKA) model. Journal of Personality and Social Psychology.
Which traits self-report reads well and which ones other people read better.
10.1037/a0017908Open the paperCronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika.
The classical internal consistency coefficient we will report per scale.
10.1007/BF02310555Open the paperTwo things this list is not. It is not a claim of endorsement: none of these authors knows about this project. It is not a bibliography for decoration: if a source stopped supporting a statement on the page, it would be removed.
Text last updated:
As a theory of bodily fluids, no, and it has not been for centuries. As a way of describing differences in the energy, speed and intensity of behaviour, the underlying observations are studied seriously and hold up. We use the four names as a readable map over eight measured scales, and we do not claim more than that.
Both, in a specific sense. Twin and family studies put heritability at roughly forty percent, and relative position stays fairly stable over years. Average levels still shift with age and circumstances. A result describes a stable tendency, not a fixed trait, and it is a snapshot of the person you are now.
Yes, and that is not a flaw of this test alone: a self-report questionnaire measures what you report. The control items catch inattentive or contradictory answering, and the simulation shows that small honest differences rarely move the profile further than one neighbouring option.
Percentiles need norms, and norms need a large, filtered sample. Until we have one, every level is stated against the scale of this test, and the wording says so. A single accuracy percentage does not exist for personality instruments at all, so anyone showing one is describing something else.
No. All 166 statements were written for this test to our own rules, and no protected instrument is reproduced, translated or adapted anywhere in it. The classical tradition is named on this page as history and as context, which is what it is.
The method above is the same for everybody. What it produces is not: eight scales, two axes and a formula that belongs to one person. It takes about twenty minutes, and the result costs nothing.
Take the test