What is factor analysis, and how did it give us the Big Five?

What is factor analysis?

Ask a large group of people to rate themselves on hundreds of trait words, and most of those ratings will not move independently. Someone who agrees they are talkative is more likely than chance to also agree they are outgoing, and less likely to deny being lively. Factor analysis takes a table of correlations like that and asks how few underlying dimensions could produce the pattern: if ten adjectives rise and fall together across thousands of raters, it treats them as one shared factor rather than ten separate traits.

A worked example makes it concrete. People who rate themselves high on talkative also tend to rate themselves high on outgoing and lively, and low on one tends to mean low on the others too. The three words move together often enough that a single dimension, roughly what gets called Extraversion, accounts for most of the shared movement. Reliable, dependable and careful form a separate cluster entirely apart from the first. Factor analysis turns that pattern into a small number of named dimensions, not a list of loosely related adjectives.

Why did the same five factors keep appearing?

Goldberg's (1990) three-study programme is the clearest single demonstration. First, 1,431 trait adjectives were sorted into 75 clusters, and ten analyses, each using a different factor-analytic procedure, recovered virtually the same five-factor structure every time. A second study narrowed 479 common trait terms to 133 synonym clusters and found the same five factors again, in self-rating and peer-rating samples alike.

One laboratory finding the same structure repeatedly is one kind of evidence. A field of researchers, working independently with different instruments, agreeing on it is another, and personality psychology spent much of the twentieth century without that agreement. Digman's (1990) review is the standard reference for the point a five-factor account became the field's emerging consensus, and Goldberg (1993) tells the convergence from the inside, arguing the model's standing owes much to each attempt to replace it having failed. The convergence crossed methods too: McCrae and Costa (1987) gathered self-reports and peer ratings, using adjective factors and NEO Personality Inventory questionnaire scales, across 738 peer ratings of 275 adults, and found the same five factors under both, with mean peer ratings agreeing substantially with self-reports, correlations ranging from .25 to .62.

What can factor analysis not decide?

Factor analysis narrows a large item set to a small number of dimensions, but it does not hand back one obviously correct number of factors. How much variance each successive factor explains offers a range of plausible stopping points, and where exactly to stop is partly a judgement call, weighing simplicity, interpretability and replication against one another; two careful researchers can reasonably choose four factors, five, or six. The clearest example is the sixth-factor debate: psycholexical studies run from scratch in seven languages, rather than translated from an English inventory, produced a similar six-factor solution, with a dimension separating out on its own called Honesty-Humility (Ashton et al., 2004). That does not mean the five-factor researchers were wrong; the same method, on a different vocabulary, can reasonably stop at a different number. The HEXACO personality model covers that sixth factor.

How is factor analysis used to test type systems like the Enneagram?

Factor analysis also checks claims that a personality system sorts people into a specific number of distinct types. If a system claims nine true types, and those nine are really just names for positions along four or five continuous dimensions, factor-analysing its test items ought to recover something close to nine separate factors; far fewer means the type claim is not supported by the evidence its own scoring rests on. That test has been run on the Enneagram: a review across 104 samples found mixed evidence, with some factor-analytic work partially aligning with the theory, but factor analyses typically recovering fewer than nine factors and no study deriving the nine types through clustering (Hook et al., 2021), a finding about statistical structure, not a verdict on whether people find the framework useful.

What is the psychometric background behind factor analysis?

Factor analysis sits inside psychometric theory, the study of how to measure something unobservable, such as a trait, from things that can be observed, such as questionnaire answers. Its standard graduate reference lays out the logic: a single item carries measurement error, a mix of the true trait and noise from wording or mood, and combining items measuring the same dimension averages that noise out, producing a steadier estimate than any one item alone (Nunnally & Bernstein, 1994). Factor analysis identifies which items belong together before that averaging can happen.

How does Wellington measure the traits factor analysis found?

Wellington is a science-based personality assessment from Therabot Labs LLC. It measures personality as continuous traits, built on the Big Five and HEXACO models, and writes the results back as a personal report rather than a type.

The free Snapshot asks 60 questions, about 7 minutes, covering six dimensions, the five this method kept recovering plus Honesty-Humility: Extraversion, Agreeableness, Conscientiousness, Emotional Stability, Openness to Experience, Honesty-Humility. Each question is a statement about behaviour on a five-point accuracy scale, not a word to sort. The first ten questions come before anything else is asked; after question ten you sign in with an email so your answers are saved and results can be shown. There is no card.

Every trait is reported as a percentile against the reference sample. The three bands (Quiet below the 30th percentile, Balanced to the 70th, High above) and the Portrait names exist to make the result easier to talk about; the percentile is the measurement, and it is always shown alongside. Norms are provisional until the norming sample is complete. The Wellington Membership, $4.99 a month, adds 336 questions and reaches 90 traits, including every Big Five facet and all 24 VIA character strengths, then carries the method toward 255 traits. Wellington is a wellness tool for reflection, not a medical or psychological diagnosis, and it does not replace care from a qualified professional. Not for hiring; data can be exported or deleted at any time.

Questions people ask

What is factor analysis, in simple terms?

A statistical technique for finding a small number of underlying dimensions that explain why many measurements move together. Words that rise and fall together across raters, such as talkative, outgoing and lively, point to one shared dimension rather than several unrelated ones. The Big Five was found this way, by analysing which of thousands of trait words moved together (Goldberg, 1990).

How was the Big Five actually discovered?

Large samples rated thousands of English trait adjectives, and factor analysis on the results recovered the same five-factor structure across three studies and ten procedures (Goldberg, 1990), a finding a widely read review described as the structure the field had converged on (Digman, 1990). Questionnaire researchers, working independently, arrived at the same five factors (McCrae & Costa, 1987).

Does factor analysis prove there are exactly five personality traits?

No. It finds the smallest number of dimensions that explains a set of correlations, and how many to keep is partly a judgement call, not a fixed output. Psycholexical studies run from scratch in seven other languages produced a six-factor solution instead, with Honesty-Humility separating out on its own (Ashton et al., 2004). Five is the answer English trait vocabularies give most often, not a mathematical certainty.

Can factor analysis test whether type systems like the Enneagram are real?

Yes. If a system claims nine distinct types, its test items should factor-analyse to something close to nine factors if the claim holds up. A review of 104 Enneagram samples found factor analyses typically recovering fewer than nine, with no study deriving the types through clustering (Hook et al., 2021), evidence about structure, not a judgement on whether people find the framework useful.