John (Jack) D. Mayer received his Ph.D. and M.A. in psychology at Case Western Reserve University, and his B.A. from the University of Michigan and was a postdoctoral scholar at Stanford University. He is presently Professor of Psychology at the University of New Hampshire. Dr. Mayer has served on the editorial boards of Psychological Bulletin, the Journal of Personality and Social Psychology, and the Journal of Personality, among others, and has been an Individual National Institute of Mental Health Postdoctoral Scholar at Stanford University. He has published over 120 articles, books, and psychological tests related to personality psychology, including on emotional intelligence, integrative models of personality, and more generally on the effects of personality on an individual’s life. He is widely acknowledged as the father of the science of emotional intelligence (together with Peter Salovey).
Andrea Scarantino (AS): Let me first thank you for agreeing to participate in the first “Topical Q&A” published in the Emotion Researcher. My job in this Q&A will be to raise some questions about your highly influential work on Emotional Intelligence (EI) that have emerged from the critical pieces published in this issue of ER, giving you an opportunity to respond. I will mainly focus on the two pieces by Antonakis and by Joseph and Newman, and occasionally raise some issues of my own. How does that sound?
John D Mayer (JDM): That sounds fine, Andrea. It’s a pleasure to take part in this Emotion Researcher Q & A.
AS: Great, let us get started then. What is your understanding of intelligence, the genus of which emotional intelligence is a species? And how many types of intelligence do you think we have scientific evidence for?
JDM: Well, I personally agree with the many intelligence researchers who regard intelligence as the capacity to carry out abstract reasoning: for example, to understand meanings, to grasp the similarities and differences between two concepts, to formulate powerful generalizations, and to understand when generalizations may not work because of contexts. There are many alternative definitions, including that intelligence is a system of mental abilities—with which I also agree—that it is an index of neurological integrity, and that it reflects the capacity to learn, or is in fact the ability to adapt. (The “abstract reasoning” definition and many of these others appear in the 1921 symposium of experts on intelligence and likely date back earlier; see Anonymous 1921).
Regarding the types of intelligence, I am content for the time being with what the intelligence researcher Kevin S. McGrew has labeled the Cattell-Horn-Carroll (CHC) model of intelligence (see McGrew, 2009). The CHC is a three-stratum model in which general intelligence or g—the capacity to reason abstractly —is at the top. Beneath it are a number of broad intelligences of a rather bewildering variety—including fluid reasoning, comprehension-knowledge (similar to verbal intelligence), visual-spatial processing, working memory, long-term storage and retrieval, and speed of retrieval. And at the bottom level are more specific mental abilities. For example, comprehension-knowledge includes the specific ability to understand vocabulary.
One reservation I have about this model concerns the exact nature of the broad intelligences and exactly what those intelligences are. The intelligences at this second level are all broad abilities that correlate with one another and with g. That said, they are quite diverse and seem to form conceptual subcategories—but what those subcategories are has not been settled. This concern over second level broad intelligences is shared by others (I particularly like the article by Schneider and Newman, 2015, in this regard).
To clarify the situation, I regard these broad intelligences as forming several groups, with g at the very top. If I sketched the top two levels of the intelligence hierarchy, they would look something like Figure 1:
Here, I have chosen to distinguish broad intelligences into three groups as a start. The first group concerns basic neuro-cognitive processing and includes abstract at-the-moment reasoning (fluid intelligence); it includes working memory, which involves the ability to retain and manipulate items in our attention as when we generate and recall a password; and it includes speed of mental processing. A second group is constituted by broad intelligences that are based on a sensory modality. Depicted here are visuo-spatial abilities and auditory-musical abilities.
Also quite important are content-focused intelligences including, in this abridged diagram, verbal-propositional and perceptual organizational intelligences. Verbal-propositional intelligence concerns understanding the meanings of words and sentences. Perceptual-organizational intelligence concerns putting together designs and puzzles—sometimes it is called mechanical ability.
AS: And how would emotional intelligence fit into this hierarchy? I am guessing it does somewhere?
JDM: Well, on first reflection it certainly seems to be a broad intelligence, so it could be added next to the verbal-propositional and the perceptual-organizational intelligences. This would yield the diagram in Figure 2.
I think, however, that it might make sense to add a new category of broad intelligences devoted to hot intelligences as compared to cool intelligences. Cool intelligences are those that deal with relatively impersonal knowledge such as verbal-propositional intelligence, math abilities, and visuo-spatial intelligence. Hot intelligences, by comparison, concern matters that affect our interests: they may warm our hearts or make our blood boil. Emotional intelligence appears to be of that hotter class, along with social intelligence, and a newly-proposed member of the group that I have called personal intelligence.
Personal intelligence involves reasoning not just about emotions, but also about a person’s motives, plans, traits, self and identity, and goals—about personalities—both one’s own and the personalities of other people. I reserve the term social intelligence for reasoning that is centrally concerned with relationships and larger group dynamics.
In Figure 3 I have revised the CHC model to include hot intelligences.
I haven’t yet filled in the specific hot intelligences, but my guess is that the most likely division would be between personal and social intelligences with emotional intelligence nested within personal intelligence. Other arrangements also are possible: emotional intelligence may also stand apart from personal intelligence. Sorting out these relations in detail is a job for another day. I note in conclusion that the CHC model, augmented to include hot intelligences, broadens our perspective on emotional intelligence and locates it in a bigger picture of how the mental abilities function as a group within personality.
AS: Before we move on to that bigger picture, I was wondering what led you and Peter Salovey to introduce in 1990 the concept of emotional intelligence, which you defined back then as “the subset of social intelligence that involves the ability to monitor one’s own and others’ feelings and emotions, to discriminate among them and to use this information to guide one’s thinking and actions”?
JDM: At the time, Peter and I both were working in the area of cognition and affect—studying the way that emotion influences cognition and vice versa. Peter and I also had interests in clinical psychology and an interest in how people function healthily.
I trained in the doctoral program in clinical psychology at Case Western Reserve University where I was very interested in emotion and emotional people and how they fared in life. As I became increasingly interested in research, Doug Detterman became a key mentor of mine and I apprenticed in his intelligence laboratory and I later switched full time into research. That gave me several years to think about intelligence and its relation to emotion.
When I moved to Stanford University in the early 1980’s, Professor Noel Sharkey invited me to write a chapter for his edited book, Advances in Cognitive Science (Chichester: Ellis Horwood). I chose to write on why computers—expert systems, actually—might benefit from experiencing emotions. A few years later these ideas matured in a set of conversations I had with Peter Salovey and the result was our 1990 article “Emotional Intelligence.”
AS: You have characterized emotional intelligence as an ability, but there are alternative definitions that identify EI with a character trait or with a mix of abilities and traits. Since several other contributors mention these alternative models, it would be helpful to know your views on them, especially if you have some major reservations.
JDM: If you want to hypothesize and measure a new part of personality, it helps to develop a clearly focused idea of what you are speaking of, and to measure it properly. I believe emotional intelligence, conceived of as an ability, is a such a clear and focused idea.
Many alternative models of emotional intelligence are well-described as “mixed” because they draw together a diverse group of personality qualities: socio-affective traits such as optimism and persistence, motivational traits such as the need for achievement and traits of self-control such as persistence. The breadth of these alternative models led me to believe that they would correlate highly with many existing and well-conceptualized measures of personality and therefore fail to add to what psychologists already study. Two recent articles indicate that these alternative EI notions correlate between r = .69 and .72 with a “general personality factor” (GPF)—that is, the self-report of positive traits on the Big Five, a widely used set of traits that includes extraversion, neuroticism, agreeableness, openness and conscientiousness (Pérez-González & Sanchez-Ruiz, 2014; Van der Linden, Tsaousis & Petrides, 2012). These mixed-model scales of EI, in other words, are primarily duplicative of Big Five content and in my opinion there is little rationale for labeling them emotional intelligence.
This being said, I do believe it makes sense to speak of self-estimated emotional intelligence as some of these alternative models do, so long as that estimate is anchored to the ability model of emotional intelligence to keep it focused. That self-estimated trait is useful for examining discrepancies between actual (ability-based) emotional intelligence and what people believe about themselves. If you compare actual emotional intelligence with self-estimates, you find that people have only a hint of their actual emotional intelligence, with the correlation between ability and self-estimates about r = .19—and that is interesting (see Brackett et al., 2006).
AS: You are the author, jointly with Peter Salovey and David Caruso, of the most well respected and widely used measure of EI, the MSCEIT (see article by Ashkanasy and Dasborough in this issue for a description of the measure and of its alternatives). I would like to organize my questions around three aspects of the MSCEIT:
- Construct validity: Does the MSCEIT capture the latent ability it is intended to capture?
- Incremental validity: Does the MSCEIT add to the predictive value of general intelligence and personality measures with respect to various phenomena of interest?
- Commercialization and popularization: Are the commercial interests at stake in EI testing in danger of interfering with the scientific assessment of EI? Have popularizers of EI gone too far in their unfettered enthusiasm for EI?
Generally speaking, abilities tests (e.g. IQ tests) aim to present subjects with a battery of tasks designed to elicit responses that will reveal a latent and not directly observable ability. In the case of MSCEIT, the latent ability being measured is that of monitoring and discriminating between emotions, and using the information so obtained “to guide one’s thinking and actions”. One issue raised by Antonakis and several other critics is that the MSCEIT tests for emotional knowledge but fails to capture how such knowledge translates into action. For instance, one may know what mood might be helpful to feel when meeting in-laws for the very first time, to cite one of the 141 questions contained in the MSCEIT, but be unable to feel that very mood when actually meeting the in-laws. Why should we think that theoretical knowledge about emotions translates into emotionally intelligent action in real-world circumstances?
JDM: Intelligence tests measure the ability to reason, but not necessarily whether people use their intelligence effectively in social situations. We probably all know people who achieve very high scores on general intelligence tests and their proxies (e.g., SATs) but who may decide they are not interested in an intellectual life. IQ tests tell us what people can do but not what they like to do or decide to do. So I believe Antonakis’ and others’ criticism applies equally to all intelligence tests. These are mental tests: They assess inner properties of the individual and not necessarily what the person expresses in a social context. Measures of social expression are, I think, best conceptualized as distinct from personality and its underlying mental abilities, though our personal abilities and personality strongly determine how we act. (For more on distinguishing between personality and its expression see my article “The Personality Systems Framework: Current Theory and Development” in the Journal of Research in Personality).
Our emotional reasoning is just that: reasoning. By comparison our emotional behavior is a product of myriad influences from our past behavioral conditioning, how emotionally reactive we are, and contextual factors including who we are interacting with. We would expect to see our emotional intelligence correlate with good emotional behavior and our well-being over time and situations—which it does. Any of us, however, may exhibit emotional failures in behavior at one or another time.
It’s possible that emotional intelligence is special case in the sense that, whereas understanding vocabulary and expressing it aren’t so different, knowing what’s best to do emotionally is one thing and doing it exactly as we would like may require considerable extra talent. For instance, even if we know full well that anger and sarcasm would hurt others at a given moment, we may be unable to refrain from expressing them. I regard the core of emotional intelligence as what we know is best to do. Its social expression involves other complementary skills including good self-management, fostered by personal intelligence, as well as understanding social provocation and resisting it, fostered by social intelligence.
People with high emotional intelligence are better liked and have better relationships than others. So I think of emotional intelligence as the mental ability that predicts good emotional behavior much of the time for many people. (For more on what emotional intelligence predicts, see (Mayer, Richards & Barsade, 2008, in the Annual Review of Psychology). But it does not predict good emotional behavior all the time for everyone.
AS: This clarification is helpful, but it appears to be somewhat in tension with the original definition of EI, which, as mentioned before, included the ability to “use…information [about emotions] to guide one’s thinking and actions”. Your last reply suggests instead that intelligent guidance of actions, although perhaps correlated with emotional intelligence, is not part of it. I noticed also that in later definitions of the concept of EI direct mention of action drops out entirely. For instance, in a 2008 paper we read that “emotional intelligence (EI) involves the ability to carry out accurate reasoning about emotions and the ability to use emotions and emotional knowledge to enhance thought” (Mayer, Richards & Barsade 2008), without any direct reference to enhancing the quality of the resulting actions. So I suppose at this point your view is that if we want to capture what we may commonsensically call emotionally intelligent behavior we need a measure different from, and additional to, EI. Is that right? If so, how should we go about coming up with a measure of one’s ability to carry out emotionally intelligent behavior? What practical tests could we rely on (e.g. role playing exercises, behavioral observations in emotionally charged settings, etc.)?
JDM: Yes, my thinking has shifted somewhat in regard to the place of action. Emotional intelligence concerns reasoning well about emotions, their meanings and their expressions. When a person actually guides her behavior, however, she needs to take into account far more than emotions alone and that requires a broader intelligence and other personality factors.
As soon as we begin to think about “what’s best” for us and others all things considered we draw on our personal intelligence. We use personal intelligence to guide ourselves and we do better if we keep in mind not only our emotions, but also our motives, our preferred ways of behaving, our capabilities and our goals—as well as the personalities and goals of others. When we make choices in a complex social world, in other words, there are far more capabilities involved than understanding emotion alone.
For example, let’s say a father of a high school senior watches his son open an acceptance letter to a college. The parent may feel happy and relieved for his son at his admission to college, but the father may also feel sadness at the thought of his son leaving home. There is no way to manage that feeling well without an appreciation of many factors that go beyond emotion alone. The father may elect to put the sadness aside for the moment to allow his son to experience an undiminished sense of achievement, pride and relief for the day. The father has a broad personal goal to protect his son’s moment of excitement; he further understands (back to emotional intelligence) that he can share his sadness with his son over leaving home at a later time in a loving way.
In our MSCEIT test we are very careful to explicitly state the goal of each instance of emotion management we inquire about so as not to tap into too much personal intelligence (which would reduce the validity of the test at assessing emotional intelligence specifically). In everyday life, however, we need personal intelligence—a sensitivity and understanding of our own and others’ goals, to balance our own needs with those of other people. (I describe this balance throughout the book Personal Intelligence).
Regarding the assessment of emotional expression, I believe that skills at expressing and controlling emotion are very difficult, but not impossible, to measure. Creating a role-playing exercise that makes people so mad they have trouble controlling themselves creates ethical issues, but sometimes such research is possible. In certain studies, for example, marital researchers ask couples to engage in common arguments they have while being observed in the laboratory. Assessing the behavior requires in-person testing and evaluation—and that is expensive. The costs run higher as journals demand larger samples than they have in the past. Perhaps in the future computer expert systems might automate these observations of people’s emotions. For the time being, however, there are many practical challenges facing anyone who is interested in developing measures of emotionally intelligent behavior.
AS: Finally, could knowledge of one’s own and other people’s emotions actually interfere with good decision-making in practical settings by making one “too sensitive”, as Antonakis suggests in his “curse of emotions” hypothesis? In other words, can emotion knowledge actually stand in the way of emotionally intelligent behavior?
JDM: I agree there is a lot to our functioning beyond emotional intelligence. All things being equal, higher emotional intelligence is an advantage. But if someone is “too sensitive”—high in neuroticism and rejection sensitivity, for example—that individual will carry around a “curse of emotions.” People with high levels of empathy may feel other’s pain, and in some contexts that can be a drawback—for example, in a teacher who empathizes so much with each pupil that she may find it hard to enforce her classroom’s rules.
AS: An important feature of IQ testing is that it is generally clear what the correct answer to a particular cognitive task is (e.g. a mathematical task). Several critics have suggested that this is not the case for the tasks comprised in MSCEIT. Consider for instance the case of Debbie, who just came back from vacation feeling peaceful and content. Another one of the questions in the MSCEIT asks how well various actions preserve Debbie’s mood, with choices ranging from making a to-do list, thinking about her next vacation and ignoring her feeling of contentedness. In this case, it is far from clear what the correct answer is. The unspecified context seems to play a key role in determining which action would be most mood-preserving and different respondents are likely to fill in the details of the unspecified context in different ways (What kind of personality do they imagine Debbie to have? When and where are they assuming Debbie’s next vacation to be?). Aren’t you worried that there isn’t enough contextual information in the test questions, and that the answers given will be confounded by cultural biases and personal idiosyncrasies of the respondent, rather than revealing a latent and universal human ability? More generally, how can we capture the essential context-dependence of what emotional intelligence demands in complicated real-life circumstances with a battery of simple questions?
JDM: Regarding the “Debbie” item, by leaving her personality and context unspecified, we are asking people to answer the question by imagining someone of average personality in a general context.
The more we indicate specific personalities in items such as these, the greater the likelihood we measure personal intelligence rather than emotional intelligence. Remember that in our focal model, emotional intelligence involves perceiving emotions, using them to enhance thought, understanding them and managing emotions. Although no intelligence can be measured with absolute purity, we want to keep reasoning about personality and social situations to a minimum so as concentrate on emotional reasoning. Our measure requires some verbal intelligence to understand the questions and answer them and probably some personal intelligence as well. That said, we aim to keep emotional reasoning at the center, and I believe we have mostly succeeded.
Questions that veer into reasoning about motives, inner states, goals, plans, traits, and self-awareness are solidly in the territory of measuring personal intelligence. My colleagues David Caruso, Abigail Panter and I are now using the Test of Personality Intelligence (TOPI) to cover that ground (Mayer, Panter & Caruso, 2012).
I also should note regarding the Debbie question, that each test item goes through a multistage checking process. Elsewhere, Haladayna and colleagues (2002) have set down a helpful group of rules and concerns for writing test items. Beyond checking that the items conform to such guidelines, experts can have a difficult time spotting which items are best. One reason for this is because items may draw on multiple abilities—and are, as a consequence, read and solved in different ways by participants (see Ackerman, 1992). During our test-development phase, therefore, we study which items perform satisfactorily and drop those items that fail to perform. Although problematic items can and do slip through, the validity of the test indicates that the larger number of items work adequately.
AS: A related issue concerns the scoring system used to assess how emotionally intelligent respondents are. The MSCEIT uses consensus scoring (the more an answer matches the consensus, the higher its score) and expert scoring (the more an answer matches the experts, the higher its score). The use of consensus scoring is not unheard of in abilities testing, but it is generally reserved for tasks in which the consensus is indeed constitutive of correctness (e.g. consensus on the grammaticality of a sentence by a representative sample of native English speakers makes the sentence grammatical). What is the rationale for thinking that consensus indicates the correct answer in the realm of emotional intelligence? For instance, couldn’t ordinary people simply be wrong about what moods are helpful when meeting the in-laws for the first time, or about which specific mental practices are mood-preserving? Are we not simply tapping into ‘old wives’ tales’ about how emotions and moods work when we rely on respondents’ consensus?
JDM: Tests of verbal intelligence assesses people’s ability to recognize vocabulary words and to understand the meanings conveyed by brief passages. Those tasks evaluate the respondents’ answers against a consensus among experts as to what a word or brief literary passage conveys to an astute reader. Like English and other languages, emotions are a communication system and rely on people’s shared understanding of the meanings they convey. Not everyone can grasp the communicated consensual meaning, and those who fail to do so are considered low in communication ability in the given realm, be it our verbal language or emotional language. That is why consensus is a pretty good scoring method.
Other experts believe that the MSCEIT can be improved by using what Roberts and colleagues have called veridical scoring. We rely wholly on veridical scoring in the newly-published MSCEIT Youth Research Version (MSCEIT-YRV) —a test normed on youth between the ages of 10 to 18 years. To score the YRV, we first examined scientific articles that spelled out the rules and meanings of emotions based on laboratory and survey work, and provided it as a reference source to a group of emotion experts (doctoral-level and Ph.D. psychologists with specialty interest and/or training in the area). Next, we asked the experts to refer to the research and to identify the correct answers to the test. Our expert panel then identified correct and incorrect answers for all the possible items. Items for which they could not agree were removed from the test. This procedure is very similar if not identical to that used on standard intelligence tests such as the WAIS: Subject-matter experts evaluate each question for the correct answer and only those items that experts agree on are retained for the test. Going forward we expect to continue the transition to veridical scoring in the next version of the MSCEIT.
AS: Another issue raised by Antonakis concerns the extremely high correlation between consensus scoring and expert scoring: “commoners” and “experts” seem to agree on the “correct” answer in the great majority of cases. The experts were 21 emotion scholars attending ISRE 2000 in Quebec. Why do you think that ISRE membership makes one an expert in emotional intelligence? Only half-jokingly, aren’t academics in general a peculiar bunch whose counsel in matters of emotional intelligence we should not especially seek? And isn’t the very fact that their answers largely overlap with the answers of regular folk a sign that they may be experts on, say, the amygdala, or Cartesian emotions, but not on how to use emotional information to guide one’s thinking?
JDM: Academics are surely more highly intellectual than average. The psychoanalytic tradition argues that intellectualization is sometimes a defense against feeling. That acknowledged, I suspect that academics who study emotion are on the whole above average in their emotional intelligence—although a few individuals among us are no doubt impaired in these areas.
An often-overlooked finding in our study of expert and everyday respondents on the MSCEIT is that the experts had a higher consensus among themselves than regular folk (Mayer, Salovey, Caruso & Siterenios, 2003). This finding is consistent with the idea that in many instances—such as understanding vocabulary and grammar—being an expert involves a heightened understanding of what the consensus is, along with other more rarified knowledge.
If we were developing an assessment of physics we’d poll physicists. In this case, the subject is emotions and my guess is that ISRE members are more knowledgeable about emotions than others. As noted, we have now moved to veridical scoring (asking experts to identify correct and incorrect answers) on the MSCEIT-YRV, and we will do the same for the next-generation MSCEIT.
AS: Additionally, in his article Antonakis argues that a by-product of using consensus scoring is that truly gifted individuals (call them emotional geniuses), who are by definition way above average, will score poorly. How do you answer these worries, which do not seem to apply equally to the case of IQ testing?
JDM: Yes, the consensus can be wrong at times—and for that reason we take seriously those commentators such as Antonakis (and, elsewhere, Roberts, Zeidner and others) who argue that test does less well at identifying truly gifted individuals in this area. We agree the test could be improved by adding more difficult items to help identify gifted individuals and hope to improve that in its revision.
AS: Joseph and Newman worry that EI may not neatly break down into the four sub-abilities posited by MSCEIT — perceiving, understanding, using and regulating emotions (a.k.a. “branches”) — suggesting that more or fewer sub-abilities may be needed (they doubt in particular the viability of the ‘using emotions’ branch). Why did you settle on exactly four sub-abilities? Additionally, Joseph and Newman wonder whether all and only abilities that are relevant to EI as you and Salovey originally defined it are assessed by MSCEIT. For instance, how does the MSCEIT test for one’s ability to express emotions or to perceive emotions in oneself (broader worry: are all relevant aspects of EI captured by MISCEIT)? And how is the ability to match emotions to colors and tastes relevant to the assessment of EI (broader worry: are aspects irrelevant to EI assessed by MSCEIT)? Finally, to answer an issue raised by Antonakis, what reason is there to think that the four abilities in question reflect a unified and underlying higher-order ability (EI) rather than being broadly independent?
JDM: Well, your first set of questions here ask whether the MSCEIT suffers from construct underrepresentation: What the Standards of Educational and Psychological Testing defines as a failure of a test to measure all that it ought to given its purpose. No intelligence test can include all possible measures of its target ability, and the MSCEIT is no exception. For example, we developed the Multifactor Emotional Intelligence Scale (MEIS) before the MSCEIT. The MEIS had a scale for measuring emotional perception in music, in which people listened to short original pieces of music and identified the emotional expressions they conveyed. A second MEIS subscale asked people to estimate the emotion of characters in brief stories. The MSCEIT excluded those scales in favor of other scales that better fulfilled our psychometric needs for the test. The issue then is whether we represent the ability with relative completeness through the specific abilities we measure on the test. An empirical argument to the contrary would require an ability-based measure of emotional-intelligence that measured factors of emotional intelligence not assessed by the MSCEIT. I am unaware of any such research or research finding so far.
In addition to construct underrepresentation, the Standards also point out that tests can suffer from construct irrelevant variance if they measure qualities unrelated to the construct. Some people might argue that our emotional synesthesia task — e.g. matching emotions to colors — falls into that area, although I wouldn’t agree because, in my opinion, being able to understand the sensory evocations of emotions is part of understanding and using them to think and to communicate with others.
One of the frustrating issues in psychometric work is that some of the concepts concerning test validity and the mathematical modeling of tests get very technical very quickly. For example the question of why we settled on four abilities involved our theoretical process and a number of measurement issues as we saw them in 1995-1996. The idea that we may measure too many or too few aspects of emotional intelligence has to do with theoretical issues involving what emotional intelligence is—that is, is our theoretical model a good one? The resolution of the debate is, therefore, partly theoretical. But the number of “branches” also rests on empirical matters such as the nature of our test’s items, reliability and covariance structure. To go into detail about these issues here would be beyond the scope of this interview. Those interested in the technical details can refer to a number of interesting articles on the topic (see, for example, Maul, 2011; Mayer, Salovey & Caruso, 2013). For general principles in this area, see the Standards for Educational and Psychological Testing, my forthcoming book, The Elements of Mental Tests, or any other general references in testing.
Setting technicalities aside, I can say further that I now prefer to interpret the four-branch model I developed with Salovey as representing the areas of problem-solving involved in emotional intelligence—but not necessarily a demarcation of four distinct mental abilities. That is, every intelligence has a class of problems it can be applied to. The four-branch model describes that class of problems as we envision it. To solve those kinds of problems, people construct a “problem space” — a mental representation of the problem to be solved, and then attempt to answer the specific questions within that problem space (see Hmelo-Silver, 2013; Keren, 1984).
On reflection, there is no particular reason to expect an exact correspondence between the areas of problems we solve — what is described by the four-branch model — and the number of distinct mental abilities we use to solve them. Psychologists determine the number of mental abilities involved through factor analysis and other related means. The exact number of mental abilities involved in solving problems in the emotion domain is an unsettled issue as of now. We ourselves have noted that two- and three-ability factor models might work as well or better than four-ability factor models. Whatever the final factor structure, the four-branch model may remain useful as a demarcation of the kinds of emotional problems people solve.
Regarding our rationale for the various abilities of emotional intelligence forming a unitary intelligence, the MSCEIT tasks generally correlate r = .25 to .65; these positive correlations provide a basis for the claim emotional intelligence is a unitary ability rather than a set of independent skills (more detail can be found in the technically-oriented publications cited earlier).
AS: Much of the excitement surrounding EI comes from the assumption that measuring it will give us a powerful predictive tool we would not otherwise have with respect to various phenomena of interest, most prominently work performance.
JDM: In case any of your readers may have missed it, let me begin by saying that Peter Salovey and I thought the idea of emotional intelligence was intriguing and might explain something and we hadn’t given much thought to the strengths of any specific relationships. So we didn’t make any claims about the strength with which emotional intelligence might predict outcomes. Journalists who first reported on our ideas made their own claims. We tried to guide the field toward more realistic perspectives. Writing in the American Psychological Association’s Monitor , I explained:
“Our own and others’ ongoing research indicates that emotional intelligence may well predict specific, important life variables at about the level of other important personality variables (e.g., 2 percent to 25 percent of the variance explained)…In contrast, the popular literature’s implication—that highly emotionally intelligent people possess an unqualified advantage in life—appears overly enthusiastic at present and unsubstantiated by reasonable scientific standards.” (Mayer, 1999)
AS: Both Antonakis and Joseph and Newman doubt that EI scores significantly add to the predictive power of measures of general intelligence and personality traits with respect to work performance. Joseph and Newman also suggest that EI scores may be better predictors of job satisfaction or the size of one’s social network than of work performance in particular. What is your answer to these worries? And what aspects of life’s success do you think EI scores are better predictors of, once we have accounted for personality and general intelligence?
JDM: In the 19th and 20th centuries, the quantifiable indices of success and productivity were relatively limited—grades in school, salary level and (for businesses) the bottom line profits. As we have transitioned into the 21st century, however, we are beginning to understand that our world—and our lives—are interrelated systems that are best assessed in multiple ways rather than by one criterion or another. Our measurement practices are beginning to catch up with these ideas.
Organizations, for example, are now gauged in part according to their social responsibility as well as their profits. Nations are assessed not only as to their wealth but also according to the freedom of their press and the well-being of their citizens. Individuals’ attainment is addressed not just by examining a person’s salary or the size of his family, but also in terms of his health, social relations and his longevity.
The profits of companies, the wealth of nations and the educational attainments of individuals all are very important. However, it may be that we won’t be able to fully appreciate the incremental predictive value of emotional intelligence unless we are examining criteria designed with emotional intelligence in mind. Broad intelligences predict criteria that often go unmeasured as of now and that we need to focus more on these heretofore unmeasured (or less emphasized) criteria.
In my Annual Review examination of the EI area with Roberts and Barsade (Mayer, Roberts and Barsade 2008), we concluded that the strongest relations between emotional intelligence and life outcomes concerned not salary or leadership position but rather people’s relationships with their colleagues, their social acceptance, and some aspects of job satisfaction.
AS: Joseph and Newman report that sex differences exist in abilities-based EI scores (e.g. MSCEIT), with women scoring significantly higher than men, but that such differences vanish when EI is measured through self-reports, which is to say that males and females assess themselves to be equally emotionally intelligent despite perhaps not being so. Does this result surprise you? More generally, do you expect EI to be differentially distributed across not only gender, but also socio-economic status, race, and age? Finally, do you expect psychopaths to score lower on EI than normal people do, or could their socially manipulative behavior actually rely on higher EI, as Joseph and Newman hint at? If so, is there a ‘dark side’ to EI?
JDM: One reason I became a personality psychologist is because I prefer to understand individuals according to the specifics of who they are, rather than inferring what they are like from memberships in groups. I realize people have legitimate research interests that lead them to study group differences. For example, as test authors, my colleagues and I have the responsibility to ensure that the MSCEIT items are as free of bias toward any given group as we could make them. Beyond that responsibility, however, my chosen way of thinking emphasizes, to draw from Dr. Martin Luther King, the content of a person’s character. For those interested in group differences, MHS collects norms for different groups and these are been reported in our 2002 test manual (Mayer, Salovey & Caruso, 2002).
You also asked about psychopathy. In a 2012 article, Elsa Ermer and her colleagues, along with Peter Salovey, found that among incarcerated men, those with psychopathic symptoms scored lower on the MSCEIT, even after controlling for general intelligence. There probably is a dark side to emotional intelligence—emotional manipulativeness—but based on these results probably not psychopathy.
AS: In their paper for this issue, Mikolajczak and Peña-Sarrionandia argue that studies conducted in their lab show that EI can be successfully taught. What is your view on whether EI is an innate or a learned ability and, in light of your expertise, what do you consider the best ways to make people more emotionally intelligent?
JDM: Intelligence researchers have traditionally reserved a very specific meaning for “raising an intelligence”: It refers to raising the ability to carry out abstract reasoning. I don’t know whether EI can be raised or not—and I don’t think anyone else does either. In principle researchers could teach the answer to any intelligence test item and their students’ abilities would seem to rise, but this would not speak to a real rise in their ability—it would simply reflect teaching the answers to the test.
I do believe, however, that we can raise people’s emotional functioning by teaching them about emotions. Many educational curricula in this area have identifiably positive impacts. That said, their teachings often extend far beyond emotions. These programs may be effective and helpful because their teachings are very broad; we simply don’t know which of their teachings are effective, or whether it is a global effect; either way, their connection to ability models of emotional intelligence are uncertain in some instances (see, for example, Cherry et al., 2014). Elsewhere I have argued that many programs allegedly focused on teaching EI are likely teaching something closer to knowledge of personality—more relevant to personal intelligence than to emotional intelligence (see Mayer, 2014).
AS: Let us get to what may be the thorniest issues of all, namely the popularization of EI and the commercialization of EI testing. Five years after you introduced the concept of EI in your joint paper with Salovey, Daniel Goleman took the publishing world by storm with his New York Times bestseller Emotional intelligence: Why it can matter more than IQ. Ever since, popularizers of EI have claimed, contrary to what the scientific data appear to suggest, that EI is a better predictor of work performance than IQ. You have taken issue with hyperbolic claims of this sort in print, but in his piece Antonakis argues that the very publishers of MSCEIT in their advertising materials make scientifically unwarranted claims about the predictive power of EI in ways that closely echo Goleman’s own. This, Antonakis continues, raises an ethical problem, because unsuspecting human resource managers may be swayed to spend lots of money on EI testing under false pretenses. How do you answer this charge, and, more generally, how do you think scientists who study emotional intelligence like yourself can keep science separate from popular hype, and the trail of money that goes with it?
JDM: I write generally about the need to raise test literacy in my forthcoming book, The Elements of Mental Tests (Momentum Press, 2015). As you mentioned, I’ve also been active in encouraging people to be realistic about what EI might or might not predict in numerous popular publications and also in journal articles and book chapters.
I believe the public has become more realistic over the years about emotional intelligence. When I speak on the topic, I often begin by saying that general intelligence is a very powerful predictor of outcomes—more powerful than emotional intelligence, though (I believe) emotional intelligence is both theoretically important and practically useful. In the late 1990s, my audiences were often shocked to hear me speak that way and felt betrayed that I wasn’t joining the journalistic chorus that emotional intelligence was “the best predictor of success in life”. Nowadays I think most audiences are far more sophisticated—at least in terms of their moderated expectations. I like to think that my public speaking and writing, along with that of many other psychologists who promoted more realistic thinking about the topic, had something to do with that new realism.
We are regularly in contact with the research staff at MHS who have a quite clear-eyed view of the test. As opportunities have arisen we also have engaged with members of the sales and marketing area (and they have sought us out) to try to better explain the MSCEIT and its predictions. That said, when authors sign a contract with a book publisher or a test publisher, they agree that the publisher has the final say in how a given work will be marketed.
AS: On a related note, Joseph and Newman’s main complaint is that researchers interested in testing the scientific viability of EI are unable to do their job properly, due to the prohibitive cost of MSCEIT, which is copyrighted intellectual property. This adds to the ethical dilemma raised by Antonakis, because not only financial incentives may lead to hyperbolic claims about EI’s predictive success, but also the very cost of MSCEIT testing may stand in the way of assessing how predictively powerful it actually is. Joseph and Newman propose that MSCEIT testing is made available for researchers for a nominal cost or for free. What is your response to this proposal?
JDM: I agree it would be desirable for MHS to offer the test to researchers for a nominal cost or for free. But what is desirable and what is practical under a given system are not always the same because of the costs involved to an organization. We have a commitment from MHS, which they have honored, to offer researchers the MSCEIT at a considerable discount. My colleagues and I are aware that even with the discount some researchers cannot afford to work with the test. MHS has, in the past, tried to go beyond our agreement to make further arrangements with researchers (e.g., offering the test for free in exchange for the researchers adding other measures that would help further validate the MSCEIT). But I realize that not every researcher has that opportunity and the system doesn’t always work.
It is worth weighing those frustrations against what MHS can and does provide that my colleagues and I would have been unable to do ourselves (without quitting our day jobs). MHS standardized the test on 5,000 participants worldwide, their research department checked our work, managed the proper scoring of the instrument, and served in the role of editor and publisher for the test manual. They supervised high-quality translations of the test in over 20 languages, provided near-world-wide distribution of the test, on-line administration, and data scoring and report generation. Moreover, our partnership has continued with the recent release of the MSCEIT Youth Research Version (MSCEIT-YRV), a test that addresses some of the issues raised here about scoring. We hope and expect to improve on the tests more in the future.
AS: Well, thanks much for your availability, Jack. I am convinced that reading your answers along with the papers in this issue of Emotion Researcher will allow our readers to get a good idea of the state of play in emotional intelligence research, along with a sense of the main challenges and opportunities currently afforded by EI testing.
JDM: Andrea, let me thank you for the opportunity to respond to your questions. The advancement of our understanding of intelligences and intelligence measures is an important factor in understanding human personality. The first intelligence tests came out in the early 1900’s and more than 100 years later the definition and assessment of general intelligence is still a work in progress.
It has been nearly 25 years since Peter and I published our theory of personal intelligence and the first demonstration of how it might be measured (Mayer, DiPaolo & Salovey, 1990; Salovey & Mayer, 1990), and just 13 years since the release of the MSCEIT. In that time, the field has greatly matured in its understanding of emotional intelligence and how to measure it. Let’s conduct a follow-up interview in about 75 years to see where this notion of an emotional intelligence has gone.
Ackerman, T. A. (1992). A didactic explanation of item bias, item impact, and item validity from a multidimensional perspective. Journal of Educational Measurement, 29, 67-91.
Anonymous (1921). Intelligence and its measurement: A symposium. Journal of Educational Psychology, 12, 123-147.
Brackett, M., Rivers, S. E., Shiffman, S., Lerner, N. & Salovey, P. (2006). Relating emotional abilities to social functioning: A comparison of self-report and performance measures of emotional intelligence. Journal of Personality and Social Psychology, 91, 781-795.
Ermer, E., Kahn, R. E., Salovey, P., & Kiehl, K. A. (2012). Emotional intelligence in incarcerated men with psychopathic traits. Journal of Personality and Social Psychology, 103, 194-204.
Cherry, G. M., Fletcher, I., O’Sullivan, H., Dornan, T. (2014). Emotional intelligence in medical education: A critical review. Medical Education, 48, 468-478.
Haladyna, T. M., Downing, S. M., & Rodriguez, M. C. (2002). A review of multiple-choice item-writing guidelines for classroom assessment. Applied Measurement in Education, 15, 309-334.
Hmelo-Silver, C. E. (2013). Creating a learning space in problem-based learning. Interdisciplinary Journal of Problem-Based Learning, 7,
Keren, G. (1984). On the importance of identifying the correct ‘problem space’. Cognition, 16, 121-128.
Maul, A. (2011). The factor structure and cross-test convergence of the Mayer-Salovey-Caruso model of emotional intelligence. Personality and Individual Differences, 50, 457-463.
McGrew, K. S. (2009). CHC theory and the human cognitive abilities project: Standing on the shoulders of the giants of psychometric intelligence. Intelligence, 37, 1-10.
Mayer, J. D., DiPaolo, M. T., & Salovey, P. (1990). Perceiving affective content in ambiguous visual stimuli: A component of emotional intelligence. Journal of Personality Assessment, 54, 772-781.
Mayer, J. D. (1999, September). Emotional intelligence: Popular or scientific psychology? Monitor: American Psychological Association.
Mayer, J.D., Salovey, P., & Caruso, D (2002). Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT) Users Manual. Toronto, Ontario: Multi-Health Systems.
Mayer, J. D., Salovey, P., Caruso, D. R., & Sitarenios, G. (2003). Measuring emotional intelligence with the MSCEIT V2.0. Emotion, 3, 97-105.
Mayer, J. D., Roberts, R. D. & Barsade, S. G. (2008). Human abilities: Emotional intelligence. Annual Review of Psychology, 59, 507-536.
Mayer, J. D. (2014). Personal intelligence: The power of personality to shape our lives. New York: Farrar, Straus & Giroux/Scientific American.
Mayer, J. D., Salovey, P., & Caruso, R. (2012). The validity of the MSCEIT: Additional analyses and evidence. Emotion Review, 4, 403-408.
Pérez-González, J., Sanchez-Ruiz, M (2014). Trait emotional intelligence anchored within the Big Five, Big Two and Big One frameworks. Personality and Individual Differences, 65, 53-58.
Van der Linden, D. Tsaousis, I. & Petrides, K. V. (2012). Overlap between general factors of personality in the Big Five, Giant Three and trait emotional intelligence. Personality and Individual Differences, 53, 175-179.
Salovey, P. & Mayer, J. D. (1990). Emotional intelligence. Imagination, Cognition, and Personality, 9, 185-211.