The Big Five is Incomplete
New Data vs Bryan Caplan, Gwern, Richard Hanania, and Jordan Peterson
Knowledge of the Big Five traits—Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism—is pretty well saturated through American highbrow culture. Bryan Caplan,1 Gwern,2 Richard Hanania,3 and Jordan Peterson4 all rely on the terminology and research from the Big Five, and I’m betting the majority of my readers do too, without worrying too much about how this personality model is by now over thirty years old.5
But listen, thirty years is a long time to go without advancing in a scientific field. Isn’t it? We might naturally wonder whether, after that much time, they’d have found some meaningful personality traits beyond the Big Five. And in fact, psychologists Michael Ashton and Kibeom Lee have recently discovered a new, sixth, major factor of personality, to establish that the classic five personality factors were, indeed, one factor short.
How recently? Well not that recently—just twenty years ago.6 But recently enough that Bryan Caplan, Gwern, and Richard Hanania don’t seem to know about it, which may be why nobody else seems to know about it, either. Not even Jordan Peterson—who makes a living talking about personality theory and sort of, maybe, really, should be the kind of person who knows about these things, but doesn’t seem to know about Honesty-Humilty, the sixth factor of personality found across multiple languages.7
For what it’s worth, the reason this sixth factor was found was not because Ashton and Lee were dispassionately examining the old studies and just happened to stumble upon it. They were actually reasoning from the basis of evolutionary psychology, which led them to think that there should be something else there:
Back in the 1990s, before there was any hint of the HEXACO model, one of us began trying to understand the functional or adaptive trade-offs behind the Big Five factors. The resulting suggestion was that kin-altruistic tendencies corresponded to traits that combine Big Five Agreeableness and high Neuroticism (e.g. sensitivity), whereas some cooperative or reciprocal-altruistic tendencies—those governing the tendency to tolerate exploitation by others— corresponded to traits that combine Big Five Agreeableness and low Neuroticism (e.g. patience). But then we realized that there was something confusing about this: why was there no dimension for another kind of cooperative reciprocal-altruistic tendencies, the ones involved in (not) exploiting others? Well, when we started looking at lexical studies of personality structure across various languages, everything began to make sense. Not only were there two factors that basically matched our suggested rotational variants of Big Five Agreeableness and Neuroticism, but there was also (that’s right) an Honesty-Humility factor, which is readily interpretable as a dimension of treating others fairly versus exploiting them. So now we had factors for both kinds of cooperative or reciprocal altruistic tendency and also for a kin-altruistic tendency.
And once we understood those three factors, we were better able to notice a conceptual parallel shared by the other three HEXACO factors: Extraversion, Conscientiousness, and Openness each involve a trade-off between more and less ‘engagement’ within a given area of endeavour: social, task-related, and idea-related, respectively.
I think this is fascinating, and exactly the kind of stuff Jordan Peterson and everyone else would be talking about if they knew about it.
No Look Apple Pie, They’re not Ignorant, They’re Just Skeptical
OK, to be fair, the fact that Ashton and Lee went looking for Honesty-Humility before finding it is a little suspicious. And no, it isn’t just those five pundits who don’t talk about this. Half the field of personality psychology feels as though it’s spent the last twenty years talking past the six-factor HEXACO model and the discovery of Honesty-Humility, a putative personality factor related to moral (and immoral) behaviors. Plenty of researchers do use the HEXACO, but many others ignore it. You can find a handful of critiques scattered across the literature, saying things like how the two traits correlate well enough that Honesty-Humility fails to even emerge as a separate factor from Agreeableness; in other words, some claim Honesty-Humility is a facet of (or even just another name for) Big Five Agreeableness.8
Twenty years ago, when I first read about Honesty-Humility, I was skeptical myself. I’d recently perfected a Big Five inventory that was optimized for community samples. I spent a lot of time annoying my neighbors and classmates carrying out independent research, just exploring the personality space and seeing if various research findings held up. I was young, and I harassed complete strangers loved the inventory I’d worked out; it was quick, easy to understand, and efficient. So when I first read about the HEXACO, I wasn’t just skeptical. I was annoyed. (Well maybe I deserved it, but you’d think karma would have inflicted hundreds of telemarketing calls to my number instead? IDK whatever)
But annoyed or not, I tested the new model out. And in a year or two, I was able to confirm for myself that six factors can be extracted from an independently designed HEXACO personality inventory. The whole thing seemed pretty cool, and I reasoned that eventually the entire field of trait psychology would follow suit, and reorganize their thinking from the Big Five to the HEXACO model.
Then twenty years passed, and researchers just keep on using the Big Five.
And I know from hanging around the rationalist community that it isn’t just the famous pundits who aren’t even familiar with the HEXACO. I could link to Wikipedia and call it a day, but I think it might help if I said the HEXACO is sort of like the MBTI, but it lets you take any position on the dimenstions (so you don’t have to be S or N, for example) and it also includes Honesty-Humility vs the Dark Triad, and Agreeableness vs Anger.
Mind you, the HEXACO was originally just a reinterpretation of the Big Five. You submit trait adjectives in various languages to factor analysis, and find these factors; the early researchers just didn’t notice Honesty-Humility in their analysis, until Michael Ashton and Kibeom Lee pointed it out. But then a Russian study9 tried a different method. Rather than relying on factor analysis again, they instead checked adjective clusters across multiple languages, finding that Honesty-Humility terms always clustered together as distinct from Agreeableness terms. For what it’s worth, I found this a pretty convincing result. The study authors have no relationship to Ashton or Lee, their methods are different, and independent verification is really the gold standard across any science.
So why does everybody still ignore this?
Maybe, I thought, maybe it’s the fact that HEXACO Honesty-Humility really does correlate with Big Five Agreeableness across every inventory that’s been designed.
Two Factors, or One?
A recent meta analysis looked at the relationship between Honesty-Humility and Agreeableness across numerous different inventories. No matter what instruments were used, the two traits correlate at least r = 0.3, and overall, closer to r = 0.4.10 To get a sense of what a relationship of that size looks like, it looks like this:
Psychology isn’t like physics or medicine. You don’t stick a psychometer under somebody’s tongue and assume the reading will be quick, accurate, and precise. Generally speaking, any personality scale is as much as a third pure noise. So the fact that Honesty-Humility and Agreeableness go together this strongly is pretty suspicious. What we want to see, if those two factors are really, truly different, is clear orthogonality between them—we want to be able to describe, and measure, each one independently of the other.
If we can’t, that’s reasonable grounds for the distinction between these dimensions just being an artifact of the languages under study. Just because we have words for things doesn’t mean they exist—after all, any atheist will tell you we have words for dragons and elves, but Lord of the Rings doesn’t exist.11
So can Honesty-Humility and Agreeableness be measured separately, or not?
My Study
In 2021, I designed an Astral Codex survey with this question in mind. I had an existing HEXACO inventory that I’d been using on community samples, and I just needed to trim and modify it to see whether it was possible to find meaningful items that would pit H and A against each other. Scott was nice enough to include it in his 2021 survey, and I’m finally getting around to publishing the results here.
Due to space limitations in the survey, only 11 questions relevant to H and A could be included. But I created three key items representing disagreeable honesty (A- H+) and agreeable dishonesty (A+ H-): principled, uptight; ethical, unforgiving; and amoral, carefree. I had a clear prediction in mind—I expected all 11 items to produce something like the following under factor analysis:
Before you scroll further, you might stop for a moment and decide whether you think this is where the items actually fell after I analyzed them.
Originally I was only betting on obtaining a few hundred usable responses, so I only planned to run a factor analysis on these 11 key items. As it happened I received 3196 responses(!), but something something preregistered studies, so this is the intended factor analysis. The first two factors explained 44% of the variance (25% and 19% respectively); here they are in the same graphical format, given directly, and completely unrotated:
As you can see, the possibility of disagreeable honesty was cleanly recovered in these results. According to this, there is obviously a personality factor of honesty and morality here that is totally unrelated to agreeableness. If there were only one dimension of agreeableness there, the idea of a person who is, say, unforgiving and ethical at the same time wouldn’t be meaningful; it would be like describing a person who is nice but also mean, and the items would have just scattered around the space, or worse, produced only one major factor. Instead, everything is right there, as predicted. If Jordan Peterson is wanting to stay current and well informed, I think he really should start taking this HEXACO stuff more seriously.
But much more interesting: Do you see the way humble, unselfish is dragged away from honesty and towards agreeableness, and the way greedy, calculating is dragged towards disagreeableness? I think this is why the H and A traits always correlated. Being greedy and Narcissistic has always had an angry, disagreeable aspect to it; they don’t talk about Narcissistic rage for nothing. Ultimately the H factor of personality should probably never have been called honesty-humility. Humble people are agreeable! The true H, unrelated to Agreeableness, should probably rather be called something like Honesty-Honor; the name you use doesn’t really matter, but that’s the way I think of it, anyway.
For readers who really love the Five Factor Model, I wanted to make these results more legible by carrying out a complete factor analysis on the entire inventory, not just the 11 questions I was interested in. This time I extracted the first six factors to appear under varimax rotation, and then further rotated them manually to align as closely as possible with the familiar Big Five personality traits most of you know, leaving the H-factor at the end. (For anyone interested, the six factors explained 57% of the variance.) As you can see, I’ve highlighted all factor loadings greater than 0.25 for clarity:
If this is the first you’ve really ever seen of the HEXACO, and you wonder what all the fuss is about Honesty-Humility, you should be able to tell by looking at the last three items at the end that there’s clearly something missing from the Big Five. Specifically what’s missing is honesty, honor, and truthfulness, vs. the kind of happy-go-lucky unconcern for morality that one associates with big tobacco, or that neighbor who is always so nice to you whenever he comes begging for a loan, and then when you need the money back, he’s moved to a different country and changed his name to Rodriguez.
So despite what some have tried to insist, there is a difference between factors A and H which seems to lie somewhere between the sweetness of FFM Agreeableness on the one hand, and the impartial morality of the HEXACO H-factor on the other.
When I look over the table above, I see the classic Big-Five Agreeable person being pretty well-described as soft, humble, and unselfish. But look at the negative loadings for items containing terms like unemotional, uptight, or unforgiving; Agreeable people probably don’t like cold-hearted, inflexible, or dehumanized moral principles which would create friction or prevent reconciliation with others. People high in Factor H, on the other hand, seem to be concerned with ethics regardless of personal content, and while they may prefer to work with others, they could just as easily be antisocial when they see that as right.
Two Dimensions of Morality
I talk a lot about morality, and we can see by looking at all of this that there are two ways of being good that are baked into personality variation. We can of course talk about morality as a philosophical concept, but the propensity to behave in moral ways is also a part of personality, just like any other human tendency. Right? Some people are “good” in the sense of being humble, unselfish, and easygoing, while others are “good” in the sense of refusing to go along with unethical ideas or to compromise their principles for the sake of getting along. And some people are, well… bad.
For deontologists, this highlights an interesting distinction between, for example, the morality of Jesus driving money changers out of the temples, and the morality of Paul, who wrote, “If it be possible, as much as lieth in you, live peaceably with all men.”12 Which is right, forgiving others and getting along, or fighting evil and demanding virtuous behavior from others?
The same question applies to utilitarians who wonder whether they should argue with their friends about animal cruelty and harm to the environment, or preserve the general good humor and just go along with that meat-lovers’ pizza everybody always wants to order. Or if you’ve heard of a better moral system called ethica pomi, then it raises the question of whether it’s more important to bake pies for people (H), or to not bother them while they’re trying to eat their pie (A).
Or maybe you’re a psychopath who’s pathologically low in both A and H, and you can’t decide whether to lie, cheat, and steal just for the money, or for the music of your victims’ suffering. I guess life is full of dilemmas no matter who you are.
Well anyway, I hope everybody found this worthwhile. It was very illuminating for me, and resolved any lingering doubts I had about the HEXACO model. If you helped me out by taking the Astral Codex survey, or by being Scott Alexander, then hey: Thank you very much!
Caplan, B. (2016, November 15). The Silent Suffering of the Non-Neurotic. Econlib. Retrieved September 5, 2023, from https://www.econlib.org/archives/2016/11/the_neurotic_he.html
About Gwern · Gwern.net. (2009, August 5). Gwern.net. Retrieved September 5, 2023, from https://gwern.net/me
Hanania, R. (2017). The personalities of politicians: A big five survey of American legislators. Personality and Individual Differences, 108, 164-167.
DeYoung, C. G., Quilty, L. C., & Peterson, J. B. (2007). Between facets and domains: 10 aspects of the Big Five. Journal of personality and social psychology, 93(5), 880.
Goldberg, L. R. (1990). An alternative" description of personality": the big-five factor structure. Journal of personality and social psychology, 59(6), 1216.
Lee, K., Gizzarone, M., & Ashton, M. C. (2003). Personality and the likelihood to sexually harass. Sex roles, 49, 59-69.
Ashton, M. C., Lee, K., Perugini, M., Szarota, P., De Vries, R. E., Di Blas, L., ... & De Raad, B. (2004). A six-factor structure of personality-descriptive adjectives: solutions from psycholexical studies in seven languages. Journal of personality and social psychology, 86(2), 356.
Lynam, D. R., Crowe, M. L., Vize, C., & Miller, J. D. (2020). Little evidence that honesty-humility lives outside of FFM agreeableness. European Journal of Personality, 34(4), 530-531.
Gill, C. H. D., & Berezina, E. (2019). Modeling Personality structure using semantic Relationships: is the heXaco honesty-humility a Distinct trait?. Psychology in Russia: State of the art, 12(1), 89-103.
Howard, M. C., & Van Zandt, E. C. (2020). The discriminant validity of honesty-humility: A meta-analysis of the HEXACO, Big Five, and Dark Triad. Journal of Research in Personality, 87, 103982.
Don’t mind me; I’m just bitter there were no Lloyd Alexander, Robert Howard, or HP Lovecraft books in my parents’ libraries, so I ended up reading Tolkein’s trilogy more than three times as a kid.
Romans 12:18, King James Version.
I tentatively agree. I'm trying to pry it apart: What if honesty-honor, even if distinct, is not a character trait? One of the things Peterson mentioned with Big 5 factor analysis is that those traits are stable in time. At first glance it doesn't seem so to me. Could you definitively reject this argument?
You don't seem to have that many in the amoral, carefree quadrant. I wonder if that's selection bias from surveying rationalists, who seem like they would be in the uptight/ethical quadrant.
Maybe hit an anime board?