Interests are not competence.

An instrument built to help a nineteen-year-old choose a major now turns up inside hiring software: as a "motivation fit" score in a screening dashboard, as one of several inputs to a composite "culture fit" number, occasionally as a filter that decides which applicants a recruiter ever opens. In each of those places the instrument is being asked whether a candidate can do the job. That is not the question a vocational interest inventory measures, and it was never built to answer it.

What it was built to answer, it answers well. Interest measures tell you what a person is drawn toward, which fields they will enter, and which they will stay inside. They do not tell you what that person can do, and no amount of care in administering them will make them do so. The migration from guidance into selection carried the instrument across a boundary its scores do not survive.

This article makes a narrow claim, and the narrowness matters. The RIASEC framework is not a poor piece of measurement. Its six-type structure is one of the most successfully replicated in applied psychology, and the practical guidance built on it has helped an enormous number of people find work they can live inside. The failure mode here is a category error committed by the buyer, not a defect in the instrument. When an interest score is used to rank candidates for competence, an organization has substituted a good measure of attraction for a missing measure of capability, and the substitution is invisible because both arrive as numbers on the same page.

The size of the gap is worth stating up front. Across the meta-analytic record, vocational interests correlate with job performance at roughly .14 (Van Iddekinge, Roth, Putka, & Lanivich, 2011), and around .20 when interest is expressed as congruence with a specific job rather than as a general trait (Nye, Su, Rounds, & Drasgow, 2012). A structured interview sits at .42 and a job knowledge test at .40 (Sackett, Zhang, Berry, & Lievens, 2022). Those numbers do not indict interest inventories. They locate them.

Holland's hexagon describes direction, and describes it well

John Holland's account of vocational choice, set out most completely in Making Vocational Choices (Holland, 1997), rests on a deceptively plain idea: people and work environments can be described in the same vocabulary. Six types do the work. Realistic people are drawn to tools, machines, and tangible materials. Investigative people are drawn to analysis and to problems that yield to method. Artistic people are drawn to original expression and tolerate ambiguity well. Social people are drawn to teaching, helping, and developing others. Enterprising people are drawn to persuasion, leadership, and commercial risk. Conventional people are drawn to order, records, and systems that stay tidy.

The elegance is in the arrangement rather than the list. Holland placed the six types on a hexagon such that the distance between any two types corresponds to how similar they are. Adjacent types share a great deal, opposite types share the least, and the types in between fall in between. Realistic sits opposite Social, Investigative opposite Enterprising, Artistic opposite Conventional. This is a strong empirical claim, not a diagram drawn for convenience, and it is the claim that has held up. The same broad ordering keeps reappearing when researchers correlate items, factor responses, and scale distances in fresh samples.

On top of the structure sits the congruence hypothesis: people are more satisfied, perform better, and persist longer in environments that match their type. Because environments can be coded in the same six letters as people, congruence is computable, and a great deal of applied practice follows from it. A guidance counselor who knows a student's three-letter code and the codes of a hundred occupations can do genuinely useful work in twenty minutes.

Everything above stands unchallenged in this article. The hexagon has earned its place in career development, and an organization that throws it out on the strength of what follows has drawn the wrong lesson from it. The question is narrower: what happens when a structure designed to describe direction is asked to certify quality.

Notice what the hexagon does not encode. Distance around the perimeter tells you how similar two kinds of work are in the activities they involve and in the people they attract. It says nothing about how demanding either kind of work is, how much aptitude either requires, or how well any particular person will execute inside it. There is no vertical dimension. A candidate's three-letter code locates them on a map of terrain, and the map was drawn without altitude.

What vocational interests reliably predict is direction, not distance traveled

Interest measures are excellent at the things they were designed to forecast. They predict which fields people enter and which they avoid. They predict which majors students choose and which ones they abandon. They predict who leaves a job and who stays in it. They predict the affective texture of a working life, whether someone finds the daily content of the role tolerable, absorbing, or a slow grind.

The Van Iddekinge meta-analysis (2011), published in the Journal of Applied Psychology and a systematic accounting of what interest measures do in employment settings, makes this asymmetry explicit. The relations it found between vocational interests and turnover-related outcomes were stronger than those it found with performance. That pattern is exactly what Holland's theory predicts. Congruence is a theory about persistence and satisfaction. It was never a theory about competence.

This distinction dissolves a false controversy. Practitioners who defend interest inventories and practitioners who dismiss them are usually arguing about different criteria without noticing. A guidance counselor and a hiring manager evaluate the same instrument against different standards, and both are right within their own frame. A measure that predicts a decade of persistence in a field is doing serious work even if it says almost nothing about who will be in the top quartile of performers next year. The error is not in either use. It is in moving the instrument between them without changing the claim attached to it.

There is a further reason interests deserve respect in workforce decisions: they are among the few things a candidate can report about themselves that concerns the long run. Most of a selection battery is aimed at the first year. Whether someone finds the substance of the work engaging shapes the fifth. A companion analysis in this series of what actually predicts early attrition treats that horizon directly, and interests belong in the conversation there far more comfortably than they do in a screening gate.

Against job performance, the coefficients are small and stay small

Van Iddekinge and colleagues put the interest–performance relation at about .14. Nye, Su, Rounds, and Drasgow (2012), summarizing more than sixty years of research in Perspectives on Psychological Science, arrived at roughly .20, with a consistent qualification: relations were stronger when interests were matched to the demands of the specific job or major rather than treated as general traits. Both figures come from careful work, and the gap between them is informative rather than embarrassing.

Why does congruence run higher than raw interest? Because a congruence index carries information about the job as well as the person. When you score the fit between a candidate's profile and a coded work environment, you are implicitly using knowledge of what the work involves, and job-specific measurement almost always outpredicts generic measurement. It is the same principle that makes a job knowledge test a stronger predictor than a broad ability test in many settings. The lift is real, and it is bounded. Moving from .14 to .20 changes the picture of interests from very weak to weak.

The benchmark marked in Figure 2 comes from Sackett, Zhang, Berry, and Lievens (2022), whose re-analysis in the Journal of Applied Psychology applied less aggressive range-restriction corrections than earlier syntheses and reordered the standard table of selection methods in the process. Their revised estimates put structured interviews at .42, job knowledge tests at .40, general mental ability at .31, and conscientiousness at .19. That last figure is worth sitting with: even conscientiousness, the most-studied Big Five domain in selection research, sits at .19, itself modest, though above where interests land. The full validity table and the reasoning behind the revised estimates are treated at length in the flagship piece in this series; the point here is only where interests fall on it.

Three cautions keep this fair. First, these coefficients come from separate research programs and are not strictly commensurable, which is why Figure 2 plots only the two interest estimates and marks the interview benchmark as a reference line rather than a rival bar. Second, a correlation of .14 is not zero; at the population level it carries genuine information, and pooled across thousands of hires it is not nothing. Third, and most importantly, the size of a coefficient only matters relative to the decision it is asked to support. For a decision about whether one candidate will outperform another, .14 is far too little to lean on when methods four times as informative are available and affordable.

Wanting and being able are related without being redundant

The intuition that interests should predict performance is not foolish. People who like doing something tend to do more of it, and practice builds skill. Ackerman and Heggestad (1997), in an integrative review in Psychological Bulletin, formalized that intuition and mapped its limits. Their finding was that abilities, personality traits, and interests are not independent domains but cluster into overlapping "trait complexes": recognizable bundles in which certain interests, certain temperaments, and certain aptitudes tend to travel together.

That review is often cited as evidence that interests and abilities are basically the same thing measured twice. It shows the opposite. The associations are modest, large enough that the domains travel together, far too small for one to stand in for the other. An interest score narrows the plausible range of a person's aptitude profile slightly and leaves nearly all of it unexplained.

The practical consequence is that the two facts most organizations want to know about a candidate stay separate. Some people love a field they cannot do especially well. Others are quietly excellent at work they have no particular feeling for. Both are common in any applicant pool of reasonable size, and no interest score distinguishes them, because whether someone is drawn to the work and whether they can do it are different questions requiring different evidence. Any process that collects one and infers the other is guessing, and it is guessing in a direction that feels reassuring, because enthusiasm is pleasant to encounter in an interview and reads as competence to an untrained evaluator.

There is a second-order effect worth naming. Because interests, temperament, and aptitude co-occur in complexes, an interest measure will appear to work in validation studies where the criterion is contaminated by any of the other two. If a supervisor's performance rating partly reflects how visibly engaged someone seems, an interest measure will correlate with it for reasons that have nothing to do with output. Trait complexes make interests look more predictive than they are whenever the criterion is soft.

The trait-complex account also explains the anecdote every practitioner has ready in defense of interest testing. Someone who is passionate about a field, competent at it, and durable in it is a real person, encountered often enough to feel like a rule. They are the visible case because all three properties happened to arrive together, which is what a complex means. The people who arrive with one property and not the others are equally real and far less memorable, because nothing about them prompts a story. Selection systems have to be built for the whole distribution, including its unremarkable middle, and not for the cases that stay in a hiring manager's memory.

In selection, the desirable answer is printed on the job advertisement

Every self-report instrument used in hiring faces a strategic-responding problem, and applicants are not naive about it. Birkeland, Manson, Kisamore, Brannick, and Smith (2006) established the general form: job applicants score higher than non-applicants on precisely the traits a role advertises. That work concerns personality measures, and the detailed treatment of applicant faking in this series carries the numbers and the countermeasures. What matters here is that interest inventories inherit the problem in an unusually acute form.

On a personality questionnaire the socially desirable answer is often ambiguous. Should a project manager present as highly deliberate or highly decisive? A candidate can guess wrong. On an interest inventory the desirable answer is generally transparent, because the items describe activities and the job title names them. An applicant for a laboratory analyst role who is asked whether they enjoy running controlled experiments faces no puzzle at all. The item is a mirror of the job posting they read twenty minutes earlier.

This is precisely the property that makes interest inventories work well in guidance. When a student takes one, there is no incentive to distort, because the only person the answers can mislead is themselves. Interest measurement depends on the respondent having nothing to gain, and a hiring context removes that condition. The instrument is not being defeated by clever candidates so much as being used outside the conditions under which its scores mean anything.

One further caution deserves a plain statement. Interest patterns are shaped by exposure and socialization, by what someone has had the chance to try and what they were encouraged toward. That makes them a poor basis for gating access to opportunity, quite apart from what they do or do not predict. An interest score can record the shape of a person's prior options and then quietly reproduce it as a hiring decision.

Where interests earn their keep inside a hiring process

Removing interests from the competence decision does not mean discarding the data. It means routing it to the decisions it can carry. Figure 3 sets out the logic: capability and interest are separate axes, and a candidate's position on one says little about their position on the other.

The upper-left cell is the one organizations habitually mishandle. A capable candidate with little enthusiasm for the content of the work is usually a good hire and a foreseeable retention problem, and the two facts are not in tension. Knowing this at offer stage is worth more than knowing it in the exit interview. The lower-right cell is the campus case: a candidate with real appetite and an immature skill profile, who belongs in a development track or a different stream rather than in a rejection pile.

That last use is where interest data earns most of its value. In volume campus hiring, where one applicant pool has to be distributed across several functions, interests are a sound basis for allocating candidates across tracks after capability has established who is in. Two graduates who clear the same technical bar are not interchangeable when one is drawn to client-facing commercial work and the other to systems and analysis. Streaming on that basis is exactly the congruence hypothesis being used for what it was built for.

Interests also belong in the interview, not in the scoring model. A recruiter who can see that a finalist's profile leans hard toward investigative work, applying for a role that is two-thirds stakeholder management, has a specific and useful thing to ask about. That is a different act from subtracting points. The candidate may have a good answer, and the exchange informs how the role is framed at offer stage whichever way it goes.

Nothing about the measurement changes between the screening gate and the placement decision in Figure 4. What changes is the claim being made from it, and whether the candidate still has anything to gain by shading their answers. Once an offer rests on capability evidence already gathered, an interest inventory recovers the condition it needs to mean anything, which is the same condition that makes it trustworthy in guidance and untrustworthy at a gate.

Route the data to the decision it can carry

Four commitments follow from the evidence, and none of them require abandoning interest measurement.

Never let an interest score gate a shortlist. If a scoring model can reject a candidate on the strength of an interest scale, the model is treating a .14 predictor as though it were a .42 one. Capability evidence decides who advances; interest data enters after that decision, not before it.

Interrogate composite "fit" numbers. The most common way interests acquire unearned authority is by being folded, unlabeled, into a single culture-fit or motivation-fit index alongside personality and values content. Ask any vendor reporting one number what constructs sit inside it and at what weights, and ask for the validity evidence for the composite rather than for its parts. If the weights cannot be produced, the number cannot be defended in an adverse-impact review either.

Use interests where the criterion is persistence. Placement decisions, internal mobility, stream allocation on campus, and early-tenure retention planning are all decisions about whether someone stays and thrives, which is the criterion the research supports. An interest profile attached to a new hire's first ninety days is a genuinely useful artifact for their manager.

Match the instrument to the claim. This is the general principle beneath the specific one, and it applies well beyond interests. Every instrument in a battery predicts something, and the question is always whether that something is the thing you are deciding. The same discipline appears in this series' treatment of what conscientiousness does and does not forecast at work: a trait with real predictive weight still has to be pointed at the right criterion to be worth its place in a process.

An organization that follows these four commitments ends up with more use of interest data, not less. Freed from a job it cannot do, an interest profile becomes a normal part of onboarding, development planning, and internal mobility, where its actual predictive strengths do real work and where no candidate has an incentive to distort their answers.