Four types in 1.5 million people

What the study did

'Are there types of people?' has a rocky history in personality research. For decades, studies kept finding clusters in small samples that vanished as soon as anyone looked at a new sample, and most researchers had written types off. In 2018, Gerlach, Farb, Revelle and Nunes Amaral tried again with the one thing earlier attempts never had: scale. They analyzed four large online questionnaire datasets — more than 1.5 million people in total, on questionnaires ranging from 300 items down to 44 — reduced each person's answers to their scores on the five-factor ('Big Five') trait model, and asked a precise question: are there regions of that trait space where people genuinely pile up? At that scale, density becomes measurable, and a real answer has to clear two bars — denser than chance, and found again in the other datasets.

The four types

Four regions cleared both bars, and the authors gave them plain names. 'Average' combines above-typical extraversion with easily triggered negative emotion (the Big Five's 'neuroticism') and below-typical openness — and it is by far the most populated region, which is the whole point of the label. 'Reserved' is the steady, conventional region: low on both neuroticism and openness, neither loud nor closed off, reasonably agreeable and conscientious. 'Role model' pairs that same emotional steadiness with high scores on every other trait — extraversion, agreeableness, conscientiousness and openness alike — the dependable people groups organize themselves around, and a region whose population grows markedly with age. 'Self-centered' — the paper's own bluntest label — pairs high extraversion with below-average agreeableness, conscientiousness and openness; it is most common in the young and thins out sharply with age.

Peaks, not boxes

It matters what kind of thing a 'type' is here. These are not boxes that sort everyone — they are density peaks in a continuous landscape: places where somewhat more people stand than chance would put there. Most people do not stand near any of the four centers; they live on the slopes in between, a mix of tendencies with no clean label. The machinery deserves the same honesty: the standard statistical criterion (BIC) actually favored thirteen clusters, and about a third of the clusters the algorithm proposed sat in regions of lower-than-random density — the method will happily draw a circle around near-empty space. Only four peaks passed the paper's meaningfulness test, denser than chance and repeatable across datasets. That gap between thirteen found and four believed is the paper's quiet lesson in how carefully clustering results have to be read.

Shorter tests see fewer types

The peaks also need a long lens. From the 300-item and 120-item datasets, all four types were recovered. From the 100-item and 44-item datasets, only three. From a simulated 44-item dataset, just two. Resolution collapses as the questionnaire shrinks — and Ethea's assessment is 60 items, chosen so you can finish it in one sitting. A test that length sits squarely inside the zone where this method demonstrably loses types, which on its own rules out using the study's approach to classify you here.

It never looked at HEXACO

There is a second, cleaner reason: the entire analysis ran on Big Five factor scores. HEXACO — the six-dimension model this app measures — was never analyzed. The authors say so themselves, naming HEXACO's six domains as an alternative representation of personality in which the same procedure would yield different types. Nobody yet knows what the crowded regions of HEXACO space look like at this scale, because nobody has done that study.

What Ethea does with it

So this research sits in your Learn tab as knowledge, not as your label. Read one way, it is the best evidence so far that personality space has real geography — some regions genuinely are more crowded than others. Read the other way, it is equally good evidence that no short test can honestly place you on that map. Ethea's eight types therefore come from somewhere else entirely: our own rule, splitting your scores high/low on the three dimensions the app actually trains — Extraversion, Emotionality, and Honesty-Humility. That is a plain summary of three measured scores, not a scientific category, and the app says exactly that wherever your type appears. The six-dimension profile underneath is always the real picture.