Five Factors, Extracted from Dictionaries

In 1936 Gordon Allport and Henry Odbert sat over Webster’s New International Dictionary and wrote out everything the English language uses to describe people: 17,953 terms for traits and behaviour. The assumption behind this has been known ever since as the lexical hypothesis. Whatever differences between people matter to a language community eventually acquire a word; the more they matter, the sooner a single one.

That is an odd starting point for a science. It presumes that centuries of everyday observation have left behind a kind of raw data set that merely needs sorting. And sorting is exactly what then happened, over decades and with ever better computational methods.

From word list to five dimensions

Lewis Goldberg published the study in 1990 that gave the result its name. He had 1,431 adjectives rated, condensed them into 75 clusters and found the same structure in them again and again — across ten replications with different data and methods. Five broad factors stood their ground. They come not from a theory of how people work but from what repeatedly clustered together in the data.

Openness concerns curiosity and a preference for complexity. Conscientiousness describes order, persistence and self-control. Extraversion bundles social activity and positive emotionality. Agreeableness captures cooperation and compassion. Neuroticism means susceptibility to negative emotion. Each term contains facets; two people with the same overall score can have different profiles.

Types are the better story. Introvert or extravert, thinker or feeler, secure or anxious: a box promises wholeness. The Big Five are less satisfying and scientifically stronger precisely for that. They describe people along five dimensions without claiming natural dividing lines between sorts of people.

Description before explanation

The dimensions are a taxonomy, not a complete theory of why someone turned out as they did. Genes, development, culture, roles and situations act together. Scores show moderate stability and can change over a life.

Big Five traits predict relevant outcomes modestly: conscientiousness, for instance, performance and health behaviour; neuroticism, distress; extraversion, certain social experiences. “Modestly” is the important word. A correlation in groups permits no confident prediction of the next act. Christopher Soto systematically re-examined the links between traits and life outcomes in 2019: 87 per cent of the associations tested replicated, but the effects stayed small. Both halves belong together. The findings are dependable and they are weak.

Self-reports contain a self-image and a comparison norm; informant reports see something else. Good assessment uses several sources and concrete situations. A test score must never quietly become a moral grade: low agreeableness can mean a readiness for conflict, but also directness; high conscientiousness can combine reliability with rigidity.

The sixth factor

That there have to be five is no constant of nature. Michael Ashton and Kibeom Lee put forward a six-dimension model in 2007, the HEXACO model. It rests on the same lexical procedure, applied to several languages — and there a dimension appears that has no place of its own among the five factors: Honesty-Humility. It bundles sincerity, fairness, modesty of claim and the absence of any readiness to exploit others. The remaining five are called Emotionality, Extraversion, Agreeableness, Conscientiousness and Openness; Emotionality and Agreeableness are cut differently from their five-factor counterparts.

Ashton and Lee argue that their model explains things for which the Big Five have no language. Among them are the connection to biological models of kin and reciprocal altruism, and the pattern of sex differences in personality traits. In practice this means that anyone wanting to measure exploitation, corruptibility or modesty can measure them within the Big Five only by detours. The argument over the number of factors is not hair-splitting. It decides which differences between people count as fundamental and which as trimming.

Where the structure does not appear

The five-factor structure has been recovered in many countries, which long counted as evidence of its universality. Almost all those tests, however, took place in literate, urban populations.

In 2013 Michael Gurven and colleagues presented the first test in a largely non-literate indigenous society. 632 Tsimane men and women in the Bolivian lowlands worked through a translation of the 44-item Big Five Inventory. The structure could not be confirmed — not in internal consistency, not in stability over time, not in its relation to observed behaviour, and not in the factor analysis. A second sample of 430 adults rating their spouses changed nothing. What emerged instead were two principal factors that the authors connect with features of small-scale societies.

In 2019 came a second warning, this time from breadth. Rachid Laajaj and colleagues analysed 29 in-person surveys with 94,751 respondents in 23 low- and middle-income countries. The usual personality questions there largely failed to capture what they were meant to capture; their validity was poor. The contrast within the same paper is striking: in internet surveys with 198,356 self-selected participants from the same countries, the indices came out markedly better. Response habits, the handling of the interviewer and limited schooling can distort the measurement — so the structure may be disappearing not because it is absent there, but because the questionnaire does not work under those conditions.

Both lead to the same admonition that Joseph Henrich, Steven Heine and Ara Norenzayan formulated in 2010 under the acronym WEIRD: the samples of the behavioural sciences come overwhelmingly from Western, educated, industrialised, rich and democratic societies. Across many domains these populations are not typical representatives of the species but frequently outliers. Anyone inferring from them to human beings in general needs evidence for it, not merely habit.

Reading person and situation together

A profile becomes more informative when it is joined to situations. High extraversion may be visible in a familiar group and vanish under formal observation. Conscientiousness may show itself on important projects but not in trivial routine. Researchers therefore speak of trait expression: a tendency needs opportunities, incentives and sometimes safety in order to become behaviour.

For practice this means asking not only “How high is the score?” but “When does the pattern appear, when does it not, and what does maintaining it cost?” The exceptions in particular can make roles, motives and learned self-regulation visible.

What remains

The Big Five are the best available system for ordering differences in personality and at the same time a product of their origins. They come from dictionaries and from samples that cover neither the languages nor the ways of life of humanity. That a sixth factor emerges in languages beyond English, and that the structure did not appear among the Tsimane, are not marginal notes. They show that the number five is a finding about particular data and not a law of nature.

In everyday life the misuse announces itself in one sentence: “That’s just how I am.” A factor score says how someone describes himself relative to others and what that has to do, on average, with later outcomes. It does not say what someone will do tomorrow, whether he is a good person, or what is impossible for him. The value of the Big Five lies in their refusal of magic. They replace the question “What sort of person is this?” with a series of bounded questions — and leave enough room for people to be more than their regular tendencies.