Vocational interests

A vocational interest is a stable preference for a kind of activity — not for a job title, and not for an outcome. Solving a puzzle rather than being called clever. Fixing the machine rather than being paid for fixing it. Interest inventories are the oldest surviving instruments in applied psychology; Strong's was in use in the 1920s, and the six-letter model most tests run on today is John Holland's, first set out in 1959 and refined by him for four decades after.

They are also the most consistently over-read results in the field, and always in the same direction. A person opens a profile with Investigative at the 90th percentile and hears you have an analytical mind. That is not what the instrument said. It said you would rather spend Tuesday on a problem than on a client.

Interest is a prediction about attention, not about ability. Everything worth knowing here follows from holding those two apart.

Holland's model sorts both people and work environments into the same six categories — Realistic, Investigative, Artistic, Social, Enterprising, Conventional — on the argument that a person is satisfied when the environment rewards what they already want to do. Most people are described by their top two or three letters rather than one, which is where Holland code comes from. If you want the six types walked through with entry-level roles attached, the article on using the RIASEC model to choose a first job does that and this entry will not repeat it.

The part worth adding is structural. The six are arranged as a hexagon because types next to each other correlate more than types across from each other, and Prediger showed in 1982 that the hexagon is mostly two underlying dimensions doing the work: things versus people, and data versus ideas. Realistic and Social sit opposite because one end is things and the other is people. Every type is a position on those two axes.

That matters because it tells you which parts of a profile carry information. The two axes are robust. The fine distinctions between adjacent letters — whether you came out IRA or RIA — are much less so, and a report that builds three paragraphs on the third letter is spending accuracy it does not have.

This is the distinction the whole category depends on, and the one that industry marketing erases, because "here is what you are good at" sells better than "here is what you will not get bored of".

An interest inventory asks whether you would like to do something. It does not sample the behaviour, score the output, or compare it against anyone's performance. Nothing in the item set could detect that you are bad at what you enjoy — a fact every amateur musician already knows.

The evidence supports the split. Ackerman and Heggestad's 1997 review of where intelligence, personality and interests overlap found interests and abilities in the same domain related positively but far from interchangeably — enough that the two travel together on average, nowhere near enough to read one off the other. Which means a substantial number of people are drawn to work they have no particular talent for, and a substantial number are talented at work they find dull. Both groups turn up in every large sample. Neither is a measurement error.

What interests do predict is worth having, and Nye and colleagues quantified it across sixty years of studies in 2012: interests relate to job and training performance at around the .2 mark once you correct for the usual artefacts — real, replicable, and nowhere near strong enough to be read as an aptitude score. The mechanism is not hidden, either. Interest predicts performance because interested people put in more hours, quit later and keep learning after the training ends. It works through effort and persistence, which is exactly why it cannot substitute for a measure of capacity.

So the honest reading of a high letter is: this is where your attention goes for free. Whether you are any good at it is a separate question, answered by having done it.

The promise attached to interest tests is that congruence — you in a matching environment — produces satisfaction. It does. The size of the effect is the problem.

Tsabari, Tziner and Meir's 2005 meta-analysis put the congruence–satisfaction correlation at about .16 to .17. Hoff and colleagues went back over the whole literature in 2020, covering 105 studies and nearly forty thousand people across sixty-five years, and landed in the same neighbourhood: interest fit predicts job satisfaction reliably and modestly, strongest for satisfaction with the choice of work rather than with the day-to-day job.

Put plainly: a matching job makes you somewhat more likely to be satisfied, and it is not the main thing determining whether you are. Your manager, your workload, your commute, your pay relative to the people around you and whether the work is designed to give you any control all compete for the same variance, and several of them win.

There is a second finding inside the first. Fit predicts staying better than it predicts liking. People in incongruent work change majors, change jobs and change fields at higher rates. If you are trying to decide whether a field is worth three years of training, the more answerable question is not "will this make me happy" but "will I still be doing this in three years", and that is the question the instrument is best at.

Three cautions, none of which are fatal and all of which are usually left out.

The hexagon is not universal. Rounds and Tracey compared the RIASEC structure across dozens of international samples in 1996 and found the circular order fits US data considerably better than it fits samples elsewhere. The six categories were derived from a particular labour market at a particular time, and where work is organised differently the distances between types come out differently.

An interest score carries your history as well as your disposition. Su, Rounds and Armstrong's 2009 meta-analysis found one of the largest sex differences in all of individual-differences research on exactly Prediger's things–people axis, with men higher on things and women on people, and a gap of that size across half a million people is not plausibly all disposition. You cannot be interested in work you have never seen. A low Realistic score in someone who was never handed a tool is a fact about their childhood at least as much as about their wiring.

The letters are a summary, not a mechanism. Holland himself treated the types as descriptions of typical patterns, not as six kinds of person. Reading them as kinds is the ordinary route into psychological essentialism, and the percentile underneath each letter is the part that actually contains the information — what a percentile means is worth knowing before you read one.

Two things make interest results more useful than they usually are.

Trust them on timing. Low and colleagues reviewed the longitudinal evidence in 2005 and found vocational interests are among the most stable things psychology measures — they firm up sharply through the late teens and change remarkably little from about twenty-two onward. This cuts both ways. A profile taken at fifteen is a weaker guide than it looks; a profile taken at twenty-five is more durable than most people expect, which means the discomfort you feel in a badly matched job is unlikely to resolve by waiting.

Use the code as a filter, not an answer. It is good at ruling large regions of work in or out and bad at choosing between two offers. If you want your own reading, the Career Interest Match scores the six dimensions as percentiles and sets them beside your personality results rather than handing back a job title — which is the correct amount of confidence for an instrument of this kind to have.

And then check it against two things it does not measure. What you are willing to be bad at for a while, which is where follow your passion goes wrong. And what you want the work to buy you, which is work values and is a different question with a different answer.