The room-filler myth.

Every hiring debrief keeps a phrase ready for the candidate with presence: great energy, natural communicator, would be superb in front of clients. Few traits are hired on instinct as readily as extraversion, and no instinct survives the evidence worse. The foundational meta-analysis put extraversion at work at .13 against job performance (.15 even in sales, the trait's supposed home ground), and the 2022 re-estimates place generic extraversion measures at .10, with a credibility interval that dips below zero. The coefficient sent to confirm the instinct keeps coming back unable to find the effect.

The persistence of the instinct is no mystery. Extraversion is the trait a selection process sees first: it enters the room, speaks most, and is the one dimension of personality an interviewer can score from across a table in ninety seconds. Every other trait must be inferred; extraversion performs itself. An unstructured conversation is, functionally, an extraversion meter, one reason unstructured interviews predict so little. So organizations keep selecting for the trait their own outcome data declines to reward.

The result is a hiring culture with an unexamined default setting. Job advertisements ask for outgoing personalities in roles that have never been shown to reward them; panels code energy as motivation and fluency as competence; assessment programs that would never adopt an unvalidated cognitive test accept the folk theory that sociability sells. The folk theory is testable, and it has been tested for decades.

The difficulty is one of resolution. Extraversion is real, stable, and measurable, and in specific settings it carries genuine signal. But the domain label bundles two different people: an assertive, dominant half and a sociable, affiliative half. The evidence on extraversion at work keeps returning that split (the assertive half predicts, the sociable half barely does), and the interview responds mostly to the sociable half.

Extraversion's validity reverses between managers and professionals

The record starts with Barrick and Mount (1991), the meta-analysis in Personnel Psychology that made the Big Five a workable selection framework. Pooling studies across occupational groups, they estimated extraversion's corrected validity for job performance at .13, a correlation low enough to place the trait among the weakest tools on the scale used across this series (see what actually predicts job performance). Conscientiousness, in the same synthesis, reached .22. On its own, a correlation the size of extraversion's is nearly invisible at the level of any single hiring decision; pooled across a program, it is real but marginal. The trait that dominates a first impression landed well behind the trait an interview barely registers.

The occupational splits are more instructive than the average. Managers, at .18, gave extraversion its best conventional showing against job performance. Among professionals the sign flipped: −.09, meaning the more extraverted professionals in those samples performed slightly worse. A trait whose usefulness reverses across occupations behaves as a role variable, and treating it as a general virtue means accepting the professionals' penalty to chase the managers' premium. Nothing in an interview flags the reversal: the professional candidate and the managerial candidate present identically across a table, the panel scores them alike, and only the outcome data, years later and rarely joined back to the hiring file, records the difference.

One result in the 1991 tables outperformed everything else: training proficiency, at .26. Put people in a program that rewards participation, visibility, and verbal fluency, and the trait predicts; put them back at their desks, and the signal thins. That exception will matter later, because it marks the one setting where the undivided domain score genuinely helps.

Three decades on, Sackett, Zhang, Berry, and Lievens (2022) re-estimated the classic validities using less aggressive range-restriction corrections; the statistical reasons the estimates moved are traced in the account of the great validity re-estimate. Generic extraversion measures came out at .10, and the lower 80% credibility value sits at −.06. A credibility interval describes the range in which the true effect plausibly falls across settings, so an interval that dips below zero means there are plausibly workplaces where the trait's relationship to performance runs negative. Few instincts held that confidently rest on evidence that thin. Averages, though, are unkind to a trait whose sign flips across occupations; the rest of the record is what happens when the average is taken apart.

That synthesis also contains the record's most practical fact. When the items are contextualized to work (asking how a person behaves on the job instead of in general), the estimate doubles to .21, edging past the .19 that conscientiousness posts in the same tables. The same person answers differently about themselves in general and themselves at work, and the at-work answers are the ones that track the job. For practice, contextualization is close to free signal: no new construct, no longer instrument, only items that name the setting. Figure 1 plots the two records together: the occupational spread on top, the measurement contrast beneath it.

The sales record rewards assertion, not affiliation

Sales is where the instinct should be safest. If the extraverted ideal holds anywhere, it should hold on a floor where talking to strangers is the job description, and where personality plays a larger part in sales hiring than in almost any other screen. The ideal is old enough to feel like a law of nature: sales managers describe their best people as born talkers, advertisements ask for infectious energy, and screening tools are tuned to celebrate the highest score on the loudest trait. The evidence meets the instinct halfway, and the half matters.

Vinchur, Schippmann, Switzer, and Roth (1998) meta-analyzed predictors of salesperson performance in the Journal of Applied Psychology and kept two criteria separate throughout: supervisor ratings of performance, and objective sales, the revenue column itself. Pooled extraversion looked modest against both (.18 for ratings, .22 for objective sales). The story is in the subtraits. Potency, the assertive, dominant, high-energy side of the trait, predicted ratings at .28 and objective sales at .26. Affiliation, the warm, sociable, relationship-seeking side, managed .12 and .15, and its objective-sales estimate rests on four studies with an 80% credibility interval running from −.19 to .45. An interval that wide, spanning zero, is evidence too weak and unstable to build a hiring profile on. The side of extraversion an interview warms to fastest is the side whose sales evidence is thinnest.

The two subtraits also describe recognizably different salespeople. A high-potency profile speaks first on a cold call, pushes to close, and holds position under pushback; a high-affiliation profile builds warm relationships and is liked by everyone, including the people who never sign. In an interview the two present almost identically: both read as classic extraverts, filling the room in different registers. The outcome data separates what the panel cannot. Assertion tracks revenue; warmth, in these studies, tracks both criteria weakly.

Figure 2 separates the two criteria, and the reference pair at the bottom is a caution about criteria in general: cognitive ability predicted supervisor ratings at .40 and objective sales at .04. Raters reward what raters can observe (fluency, polish, apparent effort) while the revenue column ignores nearly all of it. Any claim about extraversion and sales performance has to say which criterion it means, because the two columns are not scoring the same job.

The revenue curve bends in the middle

Even the subtrait split assumes the relationship is a line: more of the useful half, more revenue. The cleanest test of that assumption is Grant (2013), published in Psychological Science: 340 outbound call-center representatives, personality measured at intake, revenue tracked for three months. The linear effect of extraversion on revenue was β = .03, a standardized regression weight indistinguishable from zero. The quadratic term, by contrast, was significant at β = −.13 (p < .01). In this dataset the relationship is curvilinear: an inverted U, rising from the introverted end, cresting, and falling as extraversion keeps climbing.

The crest sits almost exactly in the middle. Predicted revenue peaked at 4.5 on the 1–7 extraversion scale, where representatives generated $151.38 an hour. Strong extraverts, 1.5 standard deviations above the mean, generated $114.96; strong introverts, an equal distance below, generated $126.80. Read those three numbers again in order: in the one sample with a clean revenue column, the strong extraverts finished last. Over the three months, the ambiverts in the middle out-earned strong extraverts by 32% and strong introverts by 24%.

Grant's title called it the ambivert advantage, and the plausible mechanism is conversational: selling requires assertion and listening in alternation, and the far right of the scale tends to supply the first at the expense of the second. Whatever the mechanism, the shape has a direct consequence for screening. A funnel that ranks candidates by extraversion score and takes from the top climbs away from the $151.38 peak toward the $114.96 end of the curve. The extraverts-versus-introverts framing that dominates hiring debates misses that in this dataset the winner was neither. The general pattern, in which more of a genuinely useful trait stops being useful, runs across many strengths and is examined in the briefing on strengths that derail.

The study's limits are worth stating, because the 32% figure invites overreach. It is one sample, in one selling channel, over a single three-month window; the relationship is curvilinear in this dataset, which does not certify it as curvilinear everywhere; and the peak's location near 4.5 on a 1–7 scale is an estimate from a fitted model, too soft to serve as a hiring cutoff. It is also a field result with a hard outcome, the kind of evidence hiring folklore rarely survives. What the result removes is a default. After Grant, "more extraverted, more sales" is a claim that requires evidence rather than a claim that supplies it.

Figure 3 traces the curve with the three published predicted values marked on it. The instructive region is the long right-hand slope: everything an interview panel would score as more impressive presence, past the midpoint, is movement downhill.

Extraversion at work finds leaders faster than it improves them

Leadership is the instinct's other stronghold, and here the trait genuinely delivers more. Judge, Bono, Ilies, and Gerhardt (2002) reviewed personality and leadership qualitatively and quantitatively in the Journal of Applied Psychology, drawing on k = 60 studies for the overall extraversion–leadership estimate, and found extraversion the strongest single Big Five correlate of leadership at .31, with the five traits jointly reaching a multiple correlation of .48; both numbers are charted in the Big Five versus personality types briefing. Taken alone, that result looks like vindication for hiring on presence.

The vindication dissolves once the criterion is split. Judge and colleagues separated leadership emergence, whether a person comes to be seen as a leader by others, from leadership effectiveness, how well the leader's unit actually performs. Extraversion predicts emergence at .33 and effectiveness at .24. The trait is better at putting people into the role than at making them good in it, and a promotion process that mistakes the first number for the second is measuring visibility and calling it capability.

Emergence is a group phenomenon, which is what makes the gap structural. In meetings, assessment exercises, and interview panels, the person who speaks early and often accumulates leadership attributions before any results exist, and organizations then treat the attribution as a result. The gap between .33 and .24 looks small on paper; its effect compounds because nearly every gate on the way to leadership (the interview, the talent review, the visibility that precedes a stretch assignment) is scored under emergence's home conditions. A pipeline can be perfectly consistent at every stage and still be consistently measuring the wrong criterion.

The facet story then repeats. In that review, dominance correlates with leadership at .37 and sociability at .37: either lower-order trait, measured alone, exceeds the global domain score that blends them. Note the contrast with sales, though. On the sales floor only the assertive side carried signal; for leadership, both halves do. Which half matters is a property of the role, not of the trait, and that is precisely the information a single domain score averages away.

The two halves are not an improvisation of the workplace literature; they have a structural name. DeYoung, Quilty, and Peterson (2007) showed that each Big Five domain divides into two aspects, with extraversion splitting into assertiveness, the agentic and dominant side, and enthusiasm, the sociable and positive-emotion side. The general argument for measuring below the domain level, the bandwidth-fidelity tradeoff, is made in the facet-level measurement briefing; extraversion makes that argument unusually vivid because its two aspects separate so cleanly against outcomes. An instrument that reports at aspect or facet resolution can see the separation. A domain-level instrument, by construction, cannot.

Sales, leadership, and training each need a different piece of it

For sales roles, the record on extraversion at work points away from the trait as commonly screened. Weight the assertiveness-side facets that potency tracks, alongside measured judgment, and treat sociability as neutral at best. Then remember the curve: past the midpoint, more of the trait went with less revenue in the cleanest data available. A personality screen for sales hiring that still rewards maximum scores is selecting for the sales floor of folklore. The outcome column describes a different floor. The remainder of the screen belongs to methods with stronger records than any personality domain: structured interviews built from the role's actual demands, and work samples that put the candidate on a simulated call and score what happens. Extraversion's contribution, where it has one, is a weighted facet inside a broader profile, sized to the evidence above.

Leadership pipelines face a different question: which criterion is the pipeline optimizing? Promotion decisions are made by panels, in person, on presence, and presence is an emergence signal. Extraversion's .33 with emergence says candidates high on the trait will keep arriving at the shortlist; the .24 with effectiveness says the shortlist is not thereby sorted by who will run the unit well. And the assertiveness that gets a candidate noticed can, unmanaged, harden into the derailment patterns cataloged in the evidence on strengths that become liabilities. A pipeline that screens for emergence and hopes for effectiveness is running on the .33 and billing itself on the .24.

Training-heavy entry roles are the genuine exception. Extraversion's .26 with training proficiency is the undivided domain score's one solid result, and the first year of a high-volume graduate program is mostly training. In campus cohorts (where assessment runs at cohort scale), a moderate weight on the broad trait is defensible in a way it is not for experienced sales or leadership hires.

Each of these designs turns on the shape of a candidate's profile: two candidates can share a domain score while differing on exactly the half the role depends on. The case for reading profiles whole is made in personality profiles versus personality scores; this article is that case confined to a single trait. Nor is extraversion special in splitting under scrutiny: the companion briefings on agreeableness and collaboration and on openness, innovation, and learning run this resolution argument on their own traits.

Treat presence as evidence about emergence

For organizations that use personality in screening at all, the extraversion record converts into four concrete changes. Each changes how existing information is read, and each fits inside an assessment program that already exists.

• Which half the role needs. Before extraversion enters a screening profile, the job analysis should name whether the work runs on assertion (directing, negotiating, closing) or on affiliation (service, warmth, cohesion). The two halves predict differently, and the sales evidence supports only one of them.
• Context in the items. Generic extraversion items produce the .10 estimate; items about behavior at work produce .21. Framing the trait in the job's own setting doubles the signal before anything else improves; even then, the .21 is modest, and context repairs the measurement while leaving the instinct untouched.
• Facets before the blend. Dominance and sociability each exceed the global trait for leadership; potency and affiliation diverge for sales. A domain score blurs the very distinction the decision needs. Facet-level reporting (how the platform structures traits) keeps the halves visible, and a sample report shows what that looks like for a live profile.
• Presence, reclassified. What a panel reads as command of the room forecasts emergence (.33) more strongly than effectiveness (.24). That is evidence the candidate will be noticed, not evidence the candidate will deliver; the measured profile carries the second question.

These changes do not retire the interview. A well-run conversation still verifies claims, probes judgment, and gives the candidate a hearing; the evidence on doing that with structure is collected in the briefing on running structured interviews. It retires only the interview's sideline as a personality instrument. Presence can be confirmed in person; prediction should come from measurement built for it.

The candidate the debrief calls a natural is giving the panel real information; the error is in what that impression is allowed to stand for. Presence forecasts who gets noticed, who emerges, who is remembered when the notes are compared — and almost nothing about the revenue column or the unit's results. The trait is real; the instinct aims at the wrong half of it. Name the half the role rewards, measure it at the resolution where the halves become visible, and let the middle of the scale win whenever the outcome column says it should.