What Is Emotional Intelligence? A Clear Guide to a Fuzzy Term

Psychologists Peter Salovey and John Mayer set out emotional intelligence in 1990 as a set of skills: reading emotions accurately in yourself and others, regulating them, and using feelings to plan and act.¹ That skill-based version is still the clearest answer to what is emotional intelligence. The fuzzy part is that the same name now also covers an eclectic mix of positive traits such as happiness, self-esteem and optimism, which Mayer, Salovey and David Caruso described in 2008 as a split in the field.²

The two versions behave very differently in research, and the difference decides whether the popular claims hold up. In a 2010 meta-analysis drawing on 118 studies, tests of emotional intelligence as an ability accounted for only an extra 0.2 percent of the variation in supervisor-rated job performance, beyond what cognitive ability and personality already explained.³

What is emotional intelligence made of? Four linked skills

In the ability model of Salovey, Mayer and David Caruso, emotional intelligence is made of four skills: perceiving emotions, using them to help thinking, understanding them and managing them.⁴ Their 2008 definition, as a 2010 meta-analysis in the Journal of Applied Psychology quotes it, is the capacity to reason accurately about emotions and to use emotions and emotional knowledge to think better.³

The model calls these skills branches, and its main test, the Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT), is built to measure all four.⁴ Dana Joseph and Daniel Newman, who wrote the 2010 meta-analysis, argued that the branches build on one another: you have to perceive an emotion before you can understand it, and understand it before you can regulate it on purpose. Their data fit that ordering.³

1. 1Perceivenotice emotions in yourself and others
2. 2Uselet a mood help the thinking a task needs
3. 3Understandreason about what emotions mean
4. 4Manageregulate emotions in yourself and others

Perceiving emotions overlaps with reading the nonverbal cues that are part of everyday communication skills. Because the model treats emotional intelligence as a kind of intelligence, it predicts that scores should overlap with ordinary cognitive ability, and the 2010 meta-analysis found that ability test scores did (a corrected correlation of .25).³

Three models share one name

Researchers use three main models of emotional intelligence: the ability model, trait emotional intelligence and mixed models, and they differ in what they count as emotional intelligence and how they measure it.³⁵ Joseph and Newman summed up the split as a narrow, theory-based set of emotional abilities on one side and an umbrella term on the other.³

The authors of the ability model, who also designed the test that bears their names, defend the narrow version: Mayer, Salovey and Caruso argued in 2008 that emotional intelligence conceptualized as an ability predicts socially relevant outcomes beyond what other measures already capture, and that clarifying what it is and is not would separate research truly about it from research that is not.²

Trait emotional intelligence was set out by K. V. Petrides and Adrian Furnham in 2001, who concluded from two studies of 227 and 166 people that it can be treated as a distinct composite trait within existing personality models.⁵ Mixed models go wider. Bar-On defined emotional intelligence as a set of noncognitive capabilities, competencies and skills that help people cope with demands and pressures, and critics, as Joseph and Newman summarize them, object that such definitions include almost any desirable quality that is not cognitive ability and overlap too much with personality to count as something new.³ Joseph and Newman place Goleman’s 1995 account in the same mixed family, as a mixture of personality traits, motivation and mood.³

Tests that score answers versus questionnaires that ask about you

Emotional intelligence is measured in two very different ways, and a score from one tells you little about a score from the other. Ability tests present problems with better and worse answers; questionnaires ask you to rate yourself. In Joseph and Newman’s 2010 meta-analysis, self-rated and test-based ability EI, two measures meant to capture the same thing, correlated only .12.³ In practice, a person who rates their own emotional skills highly may score only average on a test of those skills, and the reverse, so the two kinds of score cannot stand in for each other.

The MSCEIT is a 141-item test of the four branches.⁴ Its authors’ 2003 validation study found that 21 emotions experts, and separately 2,112 people in the standardization sample, picked many of the same answers as correct, and reported reasonable reliability.⁶ An independent 2014 analysis of the French version, taken by 729 adults in Belgium, found that the test separated people well mainly at the low end: many items were too easy for high scorers, and on several items the answer experts rated best did not go with the highest ability.⁷

1. Ability tests: answers are marked right or wrong against expert or consensus answers (for example the MSCEIT); scores overlap modestly with cognitive ability
2. Self-report questionnaires: you rate yourself (trait and mixed EI, for example Bar-On’s EQ-i); scores overlap more with personality traits than ability-test scores do

Self-report scales pick up personality. Mixed questionnaires correlated .53 with emotional stability and .46 with extraversion in the 2010 meta-analysis, and self-rated ability EI showed no correlation with cognitive ability at all.³ A 2015 follow-up meta-analysis by Joseph and colleagues found that the content of mixed EI questionnaires overlaps strongly with ability EI, self-efficacy, self-rated performance, three personality traits and general mental ability taken together (a multiple correlation of .79).⁸

How much emotional intelligence adds at work

Emotional intelligence scores are linked to better job performance ratings, but tests of emotional intelligence as an ability add very little once cognitive ability and personality are known, and self-report questionnaires add more, a gain a 2015 analysis traced to their overlap with self-perceptions, ability EI and personality. That is the combined picture from meta-analyses published in 2010, 2011 and 2015.³⁴⁸

The study

Moderate evidence

Does EI add anything once IQ and personality are known? The 2010 meta-analysis

Dana Joseph and Daniel Newman pooled 118 studies and combined them with earlier meta-analyses of cognitive ability and personality. Ability EI tests correlated .18 with supervisor-rated performance, self-rated ability EI .23 and mixed questionnaires .47. Once cognitive ability and the Big Five personality traits were in the model, mixed questionnaires explained an extra 14.2 percent of the variance in performance and self-rated ability EI an extra 2.3 percent, while ability tests added least of all.³

Performance here means ratings by a supervisor: the meta-analysis left out self-rated performance, student samples and studies that treated grades as job performance. The authors themselves warn that statistically significant increments like these are not automatically useful in practice. The main caveat is size: the ability-test estimate rests on 10 effect sizes from 887 people.³

In a split the authors added after the main analysis, ability tests predicted performance in jobs with heavy emotional labor, such as customer service (.24), and not in other jobs (.01), but each group had only three or four effect sizes. The authors’ reasoning is that jobs with frequent customer contact demand more emotion regulation, which can drain effort from the task itself unless a person regulates emotions well; they call the result preliminary evidence that needs more work.³

A 2011 meta-analysis by Ernest O’Boyle and colleagues, with 43 effect sizes from 5,795 people, found corrected correlations with job performance of .24 to .30 across the three kinds of measure. Ability tests added 0.4 percent beyond cognitive ability and personality, not a statistically significant gain, while self-report ability and mixed measures added 5.2 and 6.8 percent. Using a different method, dominance analysis, the authors still credited ability tests with 6.4 percent of the explained variance and read their results as support for emotional intelligence.⁴ How much that small share matters is a live disagreement.

A 2015 update by Joseph and colleagues offers one account of why questionnaires add more. In it, mixed EI correlated .29 with supervisor-rated performance, and the link dropped to nothing (a coefficient of −.02) once ability EI, self-efficacy, self-rated performance, personality and general mental ability were controlled. The authors’ interpretation is that mixed questionnaires measure a combination of ability EI and self-perceptions, plus personality and cognitive ability.⁸

For a sense of scale: a correlation of .28, the overall figure in the 2011 meta-analysis, means emotional intelligence scores account for about 8 percent of the differences in job performance (.28 squared, our calculation).⁴ That is a real link, and most of the variation lies elsewhere. Many popular claims concern managers, whose effectiveness has its own research base on what leadership is and how it is judged.

Where “twice as important as IQ” and “90 percent of top performers” come from

The “twice as important as IQ” line goes back to a claim in Daniel Goleman’s 1998 book Working with Emotional Intelligence, which a 2010 meta-analysis found untrue.³ The “90 percent of top performers” figure comes from TalentSmart, a company that sells emotional intelligence tests; a 2022 article on its site gives it with no method.⁹

Big claims came early. Goleman’s 1995 book carried the subtitle Why It Can Matter More Than IQ, and by a September 2008 count cited by Joseph and Newman, at least 57 consulting firms were devoted mainly to emotional intelligence, alongside 90 organizations offering training or assessment and 30 certification programs.³

Goleman’s 1998 book said that for star performance in every field, emotional competence was “twice as important as purely cognitive abilities.” Joseph and Newman quote the line and state that their meta-analytic results show it is not true.³ O’Boyle and colleagues, whose results favor emotional intelligence, likewise found no support for claims that it is the most important single predictor of performance.⁴

Myth

Emotional intelligence matters twice as much as IQ, and 90% of top performers are high in it.

Fact

The 'twice' line is a 1998 claim by Daniel Goleman that meta-analyses contradict, and the 90% figure comes from a test company that gives no sample or method for it.

TalentSmart sells emotional intelligence tests, training and trainer certification. A 2022 article on its site by its co-founder, Travis Bradberry, states that 90 percent of top performers it studied are high in emotional intelligence, against 20 percent of bottom performers, and that emotional intelligence explains 58 percent of success in all types of jobs. Beyond saying that emotional intelligence was tested alongside 33 other workplace skills, the article gives no sample, no definition of top performer, no method and no published study.⁹ The same article states that there is no known connection between IQ and emotional intelligence, while the 2010 meta-analysis found a modest one for ability tests.⁹³

Frank Landy made the wider point in 2005: much of the data behind claims about emotional intelligence at work sits in proprietary databases, which prevents rigorous tests of the measures and their predictive value, so those claims cannot be made as science.¹⁰

Scores rise after training, but evidence of better work is thin

Emotional intelligence scores rise after training, according to a 2019 meta-analysis of 58 published and unpublished studies by Victoria Mattingly and Kurt Kraiger, which found a moderate positive effect whether studies compared people before and after training or against a control group. The effect was similar for ability and mixed measures, and larger in published studies than in dissertations. The abstract gives no reason for that gap; one possibility, our reading rather than the authors’, is that weaker results were less likely to be published. The authors add that the size of the effect is in line with other meta-analyses of competency-based training programs.¹¹

The outcome these reviews pooled is the emotional intelligence score, and a score is not the same as better work. A 2024 systematic review of 17 studies of training for healthcare workers found that every study reported higher emotional intelligence afterwards, but about 88 percent were at high risk of bias in how outcomes were assessed, and the authors caution that methodological flaws may have inflated the effects. Work outcomes were secondary in that review, such as self-reported quality of care in one controlled before-and-after study.¹²

Using the idea without the hype

The meta-analyses from 2010 to 2019 support a narrower version of emotional intelligence than the popular one: a set of emotional skills whose test scores rise after training, whose link with job performance looked stronger in customer-facing jobs in one small post hoc analysis, and which makes a poor shortcut for judging people.

When you see an emotional intelligence or EQ score, ask which kind of measure produced it. A self-report questionnaire, especially a mixed one, overlaps heavily with personality and self-perceptions; an ability test measures something narrower and different.³⁸ The two can disagree about the same person.

Be wary of emotional intelligence scores in hiring. Ability tests added almost nothing beyond cognitive ability and personality in the 2010 meta-analysis, and women outscored men on those tests by about half a standard deviation, a gap large enough that the authors judged adverse impact against men mathematically very likely if the tests were used for selection.³

The bottom line

Before you trust an emotional intelligence or EQ score, ask whether it came from a test with right answers or from a questionnaire about yourself, which overlaps heavily with personality and self-perceptions. In its original, best-defined sense, emotional intelligence is a set of abilities for perceiving, using, understanding and managing emotions, and in the 2010 meta-analysis tests of those abilities accounted for only an extra 0.2 percent of the variation in job performance beyond IQ and personality. The skills are worth practicing; “twice as important as IQ” is a claim the meta-analyses contradict.