Psychometric Properties of the Trait Emotional Intelligence Questionnaire: Factor Structure, Reliability, Construct, and Incremental Validity in a French-Speaking Population

Abstract

In this research, we investigated the psychometrical properties of the Trait Emotional Intelligence Questionnaire (TEIQue, Petrides & Furnham, 2003) in a French-speaking population. In summary, we found that (a) TEIQue scores were globally normally distributed and reliable; (b) the United Kingdom four-factor structure (well-being, self-control, emotionality, sociability) replicated in our data; (c) TEIQue scores were dependent on gender but relatively independent of age; (d) there was preliminary evidence of convergent/discriminant validity, with TEIQue scores being independent of nonverbal reasoning (Raven's [1976] matrices) but positively related to some personality dimensions (optimism, agreeableness, openness, conscientiousness) as well as inversely related to others (alexithymia, neuroticism); (e) there was also preliminary evidence of criterion validity, with TEIQue scores predicting depression, anxiety, and social support as well as future state affectivity and emotional reactivity in neutral and stressful situations; (f) TEIQue scores were susceptible to socially desirable responding; however, (g) TEIQue scores had incremental validity to predict emotional reactivity over and above social desirability, alexithymia, and the Five-factor model of personality. Such results constitute encouraging preliminary findings in favor of the use of the TEIQue.

Few psychological constructs have grabbed more scientific and popular attention than the construct of emotional intelligence (EI). Briefly, it aims to provide a scientific framework for the idea that individuals differ in the extent to which they attend to, process, and utilize affect-laden information of an intrapersonal (e.g., managing one's own emotions) or interpersonal (e.g., managing others' emotions) nature (Petrides & Furnham, 2003). Research on the EI construct has grown immensely over the past decade, and two conceptions of EI currently coexist: ability EI and trait EI. Both conceptions are based on the long-standing idea that cognitive abilities are not the unique predictor of successful adaptation and that emotional competencies or dispositions need to be considered. However, the two perspectives differ markedly with respect to their conceptualization of such emotional competencies/dispositions and their measurement. The ability perspective conceives EI as a form of intelligence best assessed via performance tests (Salovey & Mayer, 1990). In contrast, the trait perspective envisages it as a set of emotion-related dispositions best assessed through self-reports (Petrides & Furnham, 2000). Whereas performance-based measures capture maximal performance, self-report measures capture typical performance (Cronbach, 1949).

In this article, we focus on the latter perspective and present the psychometrical properties of the French translation of the Trait Emotional Intelligence Questionnaire (TEIQue; Petrides & Furnham, 2003). The TEIQue provides an operationalization for Petrides's (2001) model that conceptualizes EI in terms of personality. Basically, this model, whose initial sampling domain had been identified through a content analysis of early EI and related models, aims at organizing in a single framework all affect-related aspects of personality. The construct seems to encompass variance of two kinds: one portion that is scattered across the higher order dimensions of established personality taxonomies (e.g., Big Five, Giant Three) and one portion of variance that lies outside these dimensions. Two recent studies (Petrides, 2004; Petrides, Pita, & Kokkinaki, 2007) appear to support this conceptualization. Joint factor analyses (i.e., TEIQue/Big Five, TEIQue/Giant Three) that have been conducted separately on British, Spanish, New-Zealand, and Greek samples have consistently shown that trait EI formed a unique and composite construct that lies at lower levels of personality hierarchies. Namely, it could be isolated in both personality factor spaces (i.e., Big Five, Giant Three), it was partially determined by a cluster of traits across the existing personality taxonomies, and the trait EI factor was oblique rather than orthogonal to the Big Five and Giant Three. The benefit of gathering all affect-related personality traits under the same umbrella is twofold. Theoretically, it offers an advantage in terms of explanatory power. As pointed out by Petrides et al. (2007)

There are many constructs whose variance can be accounted for by some combination of the Giant three or the Big Five. Attempting to recast these constructs as blends of higher-order traits, however, fails to capture their essence and is not conductive to the development of personality theory (Funder, 2001). (p. 286)

For instance, variance that can be straightforwardly explained by the trait EI framework, such as the propensity to decode other's emotions (high scores process facial expressions quicker than low scores and are able to decode emotions at a lower threshold of intensity; Austin, 2004, 2005; van Kan, Mikolajczak, & Luminet, 2004), requires a complex function of Big Five factors to be predicted (e.g., low Extraversion, high Neuroticism, low Openness, and low Agreeableness). Moreover, such combinations make things far less easily intelligible, as their interpretation often involves to merge effects from four or five personality dimensions. Practically, gathering all affect-related personality traits offers an advantage in terms of predictive/incremental validity in that trait EI predicts a number of affect-related outcomes over and above the established personality taxonomies (e.g., perseverance in musical training as well as performance of ballet students: Petrides, Niven, & Mouskounti, 2006; deviant behaviors at school: Petrides, Frederickson, & Furnham, 2004; sensitivity to stress induction: Mikolajczak, Petrides, Luminet, & Coumans, 2006; coaching aptitudes of top managers: van der Zee & Wabeke, 2004).

Currently, the applications of the trait EI measures concern mostly the clinical/health, educational, and organizational domains. Regarding health, evidence has accumulated that trait EI is associated, for instance, with better mental/physical resistance to stress (e.g., Salovey, Stroud, Woolery, & Epel, 2002, using the Trait Meta-Mood Scale). With respect to education, research has suggested that trait EI is particularly relevant to both disadvantaged and gifted students. Regarding low-IQ pupils, those with higher trait EI scores performed better at school and exhibited lesser deviant behaviors (unauthorized absences, exclusions due to antisocial behaviors) than their lower trait EI peers (Petrides et al., 2004). Regarding gifted pupils, those with higher trait EI used more functional strategies to cope with their “difference” compared to their lower trait EI counterparts (Chan, 2003). Finally, much has been said about the purported role of EI in organizational settings, but few empirical research studies have actually been carried out. However, research has still suggested that trait EI is involved in job performance (see Van Rooy & Viswesvaran, 2004, for a meta-analysis), although not in the proportions that have been asserted in some popular publications. The variance of the correlations nevertheless suggests the presence of moderators. It is likely that trait EI is especially relevant to the professions having an affective component such as service workers who have to perform emotional labor (e.g., Mikolajczak, Menil, & Luminet, in press) and managers (e.g., Slaski & Cartwright, 2002). In all three domains, trait EI enables a better understanding and prediction of a number of phenomena. It is, however, open to question whether it is possible to improve trait EI and should it be the case, whether its correlates would be thereby also modified.

The operationalization of the construct through a self-reported measure is consistent with (a) the subjective nature of emotional experience and (b) the conception of EI as a set of dispositions, that is, preferences and typical—rather than maximal—performance. The TEIQue consists of 153 items rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) and encompasses 15 subscales organized under four factors: Well-Being, Self-Control, Emotionality, and Sociability. A detailed description of the factors and subscales is provided in the Appendix. The psychometric development of the instrument is described in Petrides (2001), and a full technical manual is currently in preparation. We conducted the studies we present in this article using a French translation of the TEIQue. In line with the International Test Commission guidelines for test adaptation (Hambleton, 2001), items were first translated into French and then back translated into English. The translators were fully bilingual, having French as a mother tongue but having lived several years in England and being currently employed in American organizations. The whole translation/back-translation process was supervised by the authors (C. Leroy and M. Mikolajczak). Items with problematic back translations were thoroughly discussed and appropriately amended. Most discrepancies were minor, involving the choice between two synonyms. Our translation was reviewed 18 months later by a French colleague from France. She proposed to rephrase 8 items to remove Belgian-French expressions. We implemented these modifications, as they did not change the meaning or readability of the items for Belgians.

OVERVIEW

First and foremost, we look at the distributional properties, the internal consistency, and factor structure of the TEIQue. Next, we examine the relationship between the TEIQue and demographic variables. Afterward, we focus on the TEIQue discriminant and convergent validity vis-à-vis nonverbal reasoning (i.e., an indicator of cognitive ability), alexithymia, optimism, and the five-factor model of personality. We also examine the liability of TEIQue responses to social desirability. Subsequently, we assess the criterion validity of the TEIQue regarding psychological distress, social support, state affectivity, and emotional responses to stressful situations. Finally, we investigate the incremental validity of the TEIQue over and above social desirability, alexithymia and the Big Five factors of personality.

METHOD

Participants and Procedure

These data come from eight different samples. The data collection took place from September 2003 until May 2005. Altogether, 740 participants (M age = 25.5, SD = 11.31) completed the TEIQue along with one or several other measures (during the same or separate sessions). There were 512 women, 217 men, and 11 unreported. A total of 484 were students recruited on the campus (rewarded by lottery tickets or course credit) and 256 were lay people (snowball sample). Among the latter, 1 had no degree at all, 7 had a secondary school degree, 23 a college degree, 84 a bachelor's degree, and 141 a master's degree (or PhD). The students completed the questionnaire on a paper-and-pencil form, whereas the lay people completed it online. Responses were anonymous because participants identified themselves with codes. As the results we present following involve either one or several samples (depending on the relationships under investigation), we give only the total sample size, mean age, and gender repartition for each of the variables investigated.

One part of this article (i.e., examination of emotional reactivity under neutral and stressful conditions) involves an experimental design that we describe hereafter. All participants were tested individually by E. Roy. On arrival at the laboratory, we invited them to complete the TEIQue, and subsequently, they underwent a short relaxation procedure (1 min; based on Schultz, 1965). Afterwards, we randomly allocated participants either to a stress condition or to a neutral condition. The stress condition was based on a failure experience (Krohne, Pieper, Knoll, & Breimer, 2002) that had been approved by the Faculty Ethical Committee before the launch of the study. We told participants they would be tested on a recently developed test designed to predict occupational success. We informed them that as university students, they were expected to have a 75% success rate. We subsequently presented them with the 12 most difficult items from the Raven (1976) Advanced Progressive Matrices. The neutral condition involved reading a magazine article on the measurement of intelligence. Immediately after the mood induction (i.e., neutral vs. stress), participants completed the Emotional Reactivity Index (ERI; see following).

Measures

The means, standard deviations, and internal consistencies of the measures are presented in Table 1 (TEIQue's subscales and factors) and Table 2 (all other measures).

TABLE 1 Male and Female Univariate Statistics of TEIQue Subscales, Factors, and Global Score^a

		Minimum		Maximum		Males		Females			Skewness^b		Kurtoses^c		Cronbach's α
Subscale/Factor	Number of items	Males	Females	Males	Females	M	SD	M	SD	Effect size (r)	Males	Females	Males	Females	Males	Females
Self-Esteem	11	2.73	1.73	6.55	6.73	4.86	0.74	4.54	0.87***^d	−.19	−0.217	−0.279	−0.023	0.171	.79	.71
Emotion Expression	10	1.00	1.00	6.90	7.00	4.02	1.27	4.30	1.23**	.11	0.104	−0.241	−0.525	−0.537	.86	.89
Self-Motivation	10	2.20	2.10	6.90	6.90	4.70	0.83	4.67	0.80	−.02	−0.082	−0.175	0.041	0.137	.66	.74
Emotion Regulation	12	1.50	1.08	6.67	6.42	4.40	0.87	3.80	0.89***	−.32	−0.147	0.094	0.141	0.011	.77	.79
Happiness	8	1.25	1.00	7.00	7.00	5.47	1.16	5.55	1.15	.03	−1.205	−1.260	1.268	1.726	.91	.91
Empathy	9	2.44	2.22	6.89	6.89	4.78	0.89	4.91	0.76*	.08	−0.111	−0.203	0.035	0.101	.59	.73
Social Competence	11	2.36	2.36	6.73	6.55	4.85	0.81	4.74	0.80*	−.07	−0.212	−0.214	0.043	−0.468	.75	.75
Impulsiveness (low)	9	2.00	1.89	6.44	7.00	4.50	0.86	4.45	0.89	−.03	−0.147	−0.111	−0.319	−0.105	.66	.64
Emotion Perception	10	2.30	1.80	7.00	7.00	4.68	0.95	4.67	0.87	−.01	−0.073	−0.027	−0.480	−0.176	.74	.80
Stress Management	10	1.60	1.40	7.00	6.60	4.52	1.02	3.90	1.02***	−.29	−0.302	0.121	−0.114	−0.299	.77	.81
Emotion Management	9	2.44	1.67	6.89	6.56	4.83	0.86	4.52	0.87***	−.18	−0.334	−0.266	−0.121	−0.040	.72	.75
Optimism	8	1.63	1.13	7.00	7.00	4.86	1.03	4.80	1.10	−.03	−0.433	−0.525	0.038	0.096	.83	.81
Relationship Skills	9	3.22	2.56	7.00	7.00	5.46	0.72	5.64	0.78**	.12	−0.537	−0.730	0.378	0.535	.68	.59
Adaptability	9	2.22	1.44	6.78	6.78	4.55	0.81	4.26	0.85***	−.17	−0.116	−0.125	0.278	0.528	.69	.67
Asseriveness	9	2.22	1.00	6.44	7.00	4.70	0.92	4.41	0.97***	−.15	−0.265	−0.221	−0.577	−0.217	.75	.74
Well-Being	27	2.08	1.62	6.63	6.78	5.06	0.85	4.96	0.89	−.06	−0.892	−0.759	0.845	0.744	.91	.91
Self-Control	31	2.36	2.26	6.41	6.32	4.47	0.75	4.05	0.73***	−.27	−0.161	0.195	−0.067	0.068	.85	.87
Emotionality	38	2.91	2.79	6.59	6.49	4.74	0.73	4.88	0.64**	.10	0.173	−0.054	−0.213	−0.245	.86	.90
Sociability	29	2.86	2.47	6.48	6.34	4.80	0.72	4.56	0.72***	−.16	−0.212	−0.316	−0.063	−0.020	.86	.87
Global Trait El		3.05	3.06	6.12	6.20	4.75	0.56	4.61	0.53**	−.13	−0.007	−0.180	−0.051	−0.041	.94	.95
Note. TEIQue = Trait Emotional Intelligence Questionnaire; EI = Emotional intelligence.
^a N = 740; 217 men, 512 women, 11 unknown. ^bStandard error of skewness estimations is .17 for men and .11 for women. ^cStandard error of kurtosis estimations is .33 for men and .22 for women. ^d Asterisks correspond to p values for gender differences t tests.&dollar;*p < .10. **p < .01. ***p < .001.

TABLE 2 Means, Standard Deviations, and Internal Consistencies for the Scales Examined

	N	Number of items	Cronbach's alphas	M	SD
Convergent/discriminant validity
Optimism	91	6	.82	19.84	4.81
Alexithymia	457	20	.79	47.25	10.17
Difficulty in identifying Feelings	403	7	.78	16.81	5.45
Difficulty in describing feelings	403	5	.776	13.96	4.65
Extermally oriented thinking	403	8	.43	16.28	3.95
Emotional stability	80	11	.84	38.62	8.24
Introversoin	80	11	.79	39.65	8.89
Openness	80	11	.74	45.61	6.27
Agreeableness	80	11	.71	48.24	5.58
Connscientiosness	80	11	.77	44.92	9.73
Cognitive ability (IQ)	109	36	—^a	24.20	4.37
Timed 30 min
Socialdesirability	131	33	.67	15.51	4.46
Criterion validity
Anxiety	82	20	.92	42.77	10.33
Depression	82	13	.85	4.63	4.60
Perceivedquantityofsocialsupport	93	6	.87	33.30	10.80
Family members	93	6	.93	13.23	10.81
Friends	93	6	.87	16.53	8.36
Perceivedqualityofsocialsupport	93	6	.88	30.91	3.93
Positiveaffectivity	306	10	.84	2.87	0.60
Negativeaffectivity	306	10	.82	1.42	0.47
Emotionalreactivity	187	30	.90	51.16	13.61^b
				72.39	21.35^c
^aAs Raven' s matrices fo increasing difficulty, calculating internal consistencies does not make sense. ^bMeasured in the neutral condition. ^cMeasured in the stressful condition.

We appraised trait EI through the TEIQue described earlier.

We measured professional sector via a single “multiple-choice” item. We invited participants to indicate which of the following 12 occupational categories they belonged to: social/healthcare (58), communication/medias (25), commerce (21), public service/administration (56), finance (20), transportation/tourism (2), telecommunications (6), agriculture/environment (2), justice (4), art (4), research and development (29), or currently unemployed (19).

We appraised optimism, which is defined as a generalized tendency to believe that one will generally experience good instead of bad outcomes in life, through the Life Orientation Test–Revised (LOT–R; Scheier & Carver, 1985; French translation: Régner, 2002). It contains six items and four fillers rated along a 5-point scale ranging from 1 (strongly agree) to 5 (strongly disagree).

We assessed alexithymia using the Toronto Alexithymia Scale (TAS-20; Bagby, Parker, & Taylor, 1994; French adaptation: Loas, Otmani, Verrier, Fremaux, & Marchand, 1996). This questionnaire consists of 20 items (rated on a 5-point scale ranging from 1 [strongly agree] to 5 [strongly disagree] targeting the core dimensions of the construct: difficulty in identifying feelings, difficulty in describing feelings, and externally oriented thinking.

We evaluated nonverbal reasoning by means of the well-known Raven's (1976) Advanced Progressive Matrices Test, which is one of the most robust predictors of the general cognitive ability score (supraordinate factor “g”; Spearman, 1927). This test consists of 36 problem series and is independent from language and formal schooling. Each problem consists of nine figures (arranged as a square) with a missing piece. Below the problem are eight alternative pieces to complete the figure, only one of which is correct. Each set involves a different principle for obtaining the missing piece, and problems are roughly arranged in increasing order of difficulty. The “Advanced” form of the test (originally designed for people having a university degree) spreads the scores of the top 20% of the population, and we used it here to increase the variance of the scores.

We appraised the Five-factor model of personality through the Description in Five Dimensions system (D5D; Rolland & Mogenet, 2001), which is a widely used French personality inventory based on the five-factor model (FFM; Costa & McCrae, 1992). This questionnaire assesses the Big Five dimensions of emotional stability, introversion, openness, conscientiousness, and agreeableness through 55 adjectives (e.g., nervous, reserved, cultivated, compassionate, tidy) rated along a 6-point scale ranging from –3 (does not describe me at all) to +3 (describes me perfectly).

We evaluated social desirability, which refers to a tendency to present oneself in an overly positive manner, using the Marlowe–Crowne Social Desirability Scale (Marlowe & Crowne, 1960; French translation: Blais & Lachance, 1992). It consists of 30 items rated on a dichotomous scale (true–false). Examples of items are “I am always courteous, even to people who are disagreeable”; “When I don't know something, I don't mind at all admitting it.” We measured trait anxiety through the Trait Anxiety Inventory (Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983; French adaptation by Bruchon-Schweitzer & Paulahan, 1990). This questionnaire consists in 20 items rated along 4-point scales.

We evaluated depression via the Beck Depression Inventory (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961; French adaptation by Collet & Cottraux, 1986). The short version, which we used in this study, encompasses 13 items consisting of four statements each. Respondents are instructed to circle the one statement in each group that best describes how they felt the last week.

We assessed positive and negative state affectivity through the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988; French translation: Gaudreau, 2000). The PANAS is currently the most widely used measure of affectivity. It consists of 20 adjectives rated along 5-point scales of which 10 measure positive affectivity (PA; e.g., inspired), and 10 measure negative affectivity (NA; e.g., guilty). We note that the focus here was on state (current) rather than trait (general) affectivity.

We evaluated perceived social support via the short form of the Social Support Questionnaire (SSQ; Sarason, Levine, Basham, & Sarason, 1983; French translation: Rascle et al., 1997). Respondents are first asked to list up to nine potential support providers for each of 12 hypothetical situations (e.g., “On who can you really count to console you when you are very upset?”). Then, they rate their overall satisfaction with support for each situation (rather than for each provider). The SSQ provides two scores: perceived quantity of social support and perceived quality of social support.

We estimated emotional reactivity by means of the ERI (created for this study; the full questionnaire may be requested from M. Mikolajczak). The ERI is based on Frijda's (1986) theory of emotions. According to Frijda (1986), an emotion is a three-component system. Any emotional episode begins by an evaluation of the situation as either relevant or irrelevant regarding a number of goals, each of them being evaluated in terms of importance for the self (self-integrity and self-accomplishment). The more relevant the situation is with regards to one's goals, the more intense the emotion will be. Then, the emotional episode leads to action readiness: Actions in progress are interrupted, and the individual prepares himself or herself to respond to the situation (e.g., fight or flight in negative situations, to sing, to jump for joy in positive situations). Although social constraints do not always allow one to concretely realize these actions, individuals still experience them as action tendencies (i.e., desire to flee or to jump). Finally, emotional episodes are accompanied by a number of behavioral or expressive manifestations (e.g., tensed face, shaking hands) and bodily changes (e.g., sweat, palpitations, etc.). The ERI measures the three facets of emotional experience: emotional intensity, bodily sensations, and action tendencies. We measured emotional intensity through four items adapted from Sonnemans and Frijda (1994). These items, respectively, referred to the emotion's duration, emotional intensity at the peak, emotional intensity out of the peak, and global intensity of the emotional experience. The emotion's duration item was rated along a 5-point scale ranging from 1 (1 to 5 minutes) to 5 (20 to 25 minutes); the other items were rated along a 5-point scale ranging from 1 (very weak) to 5 (very strong). We assessed bodily sensations via the Wallbot and Scherer's (1986) Bodily Sensations Checklist. This questionnaire consists of 13 bodily sensations (e.g., to have a lump in one's throat, difficulty to breathe, palpitations) rated along a 7-point scale ranging from 1 (not at all) to 7 (the strongest you can imagine). Finally, we evaluated action tendencies using Frijda, Kuipers, and ter Schure's (1989) questionnaire. Respondents were required to rate the intensity to which they wished to accomplish 11 actions (e.g., to swear, to disappear, to cry) on a 7-point scale ranging from 1 (not at all) to 7 (the most intense tendency you could experience).

Although the questionnaire was originally intended to be three factorial, principal axis-factoring revealed that it was unifactorial (all items loaded on the same factor). Reliability analyses as well as correlations among the facets (rs > than .75, ps < .001) fully supported this view: Although facets' reliabilities were acceptable (α _EI = .79, α _BS = .91, α _AT = .78; EI = emotional intensity, BS = bodily sensations, AT = action tendencies), internal consistency for the global scale was higher (α = .90), thereby supporting the validity of the construct as a whole. We thus computed a total score of emotional reactivity by summing the individual scores obtained from each of the three indicators. As findings were essentially the same across facets, we present results for the global ERI score only.

RESULTS

Distributional Properties

Minima, maxima, means, standard deviations, skewness, and kurtosis of the 15 subscales, four factors, and global trait EI score are given in Table 1 for men and women, respectively. Kolmogorov–Smirnov (KS) test revealed that normality was achieved for the global score, KS(740) = .036 and .033, ps = .05 for men and women, respectively; Self-Control, KS(740) = .056 and .034, ps = .05, respectively; Emotionality, KS(740) = .032 and .030, ps = .05, respectively; and Sociability, KS(740) = .036 and .033, ps = .05, respectively. Only the factor Well-Being was significantly negatively skewed, KS(740) = .091 and .062, ps = .05, respectively. It is noteworthy that the nonnormal distribution of the factor Well-Being constitutes an element of construct validity rather than a weakness. Indeed, research has shown that people are characteristically rather happy than neutral. Only in the very poor countries does the number of unhappy citizens equal the number of happy ones (for a review of the overwhelming dominance of happiness, see Veenhoven, 1984).

INTERNAL CONSISTENCY

Cronbach's alphas of the TEIQue subscales and factors are given in Table 1. Among the 15 subscales, 10 had acceptable to excellent reliability (varying between .71 and .91) among both men and women. Two subscales, namely, Self-Motivation and Empathy, had acceptable reliability among men but lower reliability among women. Finally, three subscales have Cronbach's alphas below .70 for both men and women: Impulsiveness, Relationship Skills, and Adaptability. Internal consistencies at the factor level were excellent for both men and women, and so is this the case for the global score.

Factor Structure of the TEIQue in the Belgian Data Set

We subjected the 15 TEIQue scales to a principal axis factor analysis. The scree plot and Kaiser eigenvalue extraction criteria suggested the presence of between two and five factors (the first six eigenvalues were, respectively, 5.261, 1.735, 1.378, 1.217, 1.025, 0.812). We applied parallel analysis ¹ to our data set, a method that is currently considered the most reliable procedure to determine the correct number of factors (see Zwick & Velicer, 1986, for a comparison of factor retention decision methods, and Hayton, Allen, & Scarpello, 2004, for methodological issues in parallel analysis). We obtained the eigenvalues and standard deviations generated from completely random data (and necessary to perform parallel analysis) through the “Marley Watkins Monte Carlo PCA for Parallel Analysis” program (Watkins, 2002a) using the following parameters: 15 variables, 740 participants, and 1,000 replications. We then compared our observed eigenvalues to the 95th percentile of the eigenvalues generated from these random data to reject factors that were most certainly artificial (at p = .05). Results supported our theoretical expectations in that four factors were retained. We thus rotated a four-factor solution via the Promax algorithm (κ = 4). The factor pattern matrix and factor intercorrelations are presented in Tables 3 through 5. Despite some cross-loadings, the four factors were substantively identical to the original British structure (Petrides, 2001, at the origin of the scoring key) and were thus labeled accordingly: Well-Being, Sociability, Self-Control, and Emotionality. This solution accounted for 63.95% of the total variance, and the strength of the intercorrelations was fully in line with the conceptualization as a multifactorial construct (Table 4).

TABLE 3 Factor Pattern Matrix for the TEIQue Subscales (Promax Principal Axis Factoring Four-Factor Solution)^a

Subscale	Well-Being	Sociability	Self-Control	Emotionality
Happiness	.99	−.07	−.15	−.01
Optimism	.86	−.04	−.07	−.12
Self-Motivation	.38	.07	.14	.24
Emotion Management (others)	−.18	.79	−.06	−.02
Assertiveness	.18	.69	.06	−.08
Social Competence	.02	.60	.07	.26
Self-Esteem	.33	.40	.11	.03
Emotion Control	−.19	.00	.92	.02
Stress Management	.14	.02	.76	−.10
Adaptability	.12	.15	.35	.04
Emotion Perception	−.13	.17	.00	.74
Emotion Expression	.05	.21	−.23	.58
Relationship Skills	.32	−.28	.00	.52
Empathy	−.11	.02	.08	.50
Impulsiveness (low)	.03	−.24	.35	.41
Note.    ^a N = 740. TEQue = Trait Emotional Intelligence Questionnaire. Factor pattern matrix: Coefficients that should theoretically define each factor are in boldface.

TABLE 4 Factor Correlation Martix Between Belgian Factor Scores and English Factor Scores^a

	Well-Beings^b	Sociability^b	Self-Control^b	Emotionality^b
Well-Being^a	.997 [.982]
Sociability^a	.544	.995 [.991]
Self-Control^a	.650	.495	.987 [.964]
Emotionality^a	.519	.626	.350	.984 [.969]
Note.    N = 740. Coefficients in brackets [] refer to congruence coefficients between factor structure (pattern matrix) in Belgian and English data set (computed with Watkins's (2002b) Coefficient of Congruence program).
^aFactor scores derived from the Belgian factors corecoefficients. ^bFactor scores derived from the English factor score coefficients.

TABLE 5 Factor Correlation Matrix Between Theoretical Factors and Factor Scores From the Belgian Factor Analysis

Factor	Well-Being	Sociability	Self-Control	Emotionality
Well-Being	.974	.502	.606	.545
Sociability	.438	.973	.449	.624
Self-Control	.549	.376	.945	.432
Emotionality	.479	.507	.364	.941
Note.    N = 740. The lower diagonal corresponds to the correlations among the theoretical factors scores (data scored according to the theoretical factor structure); the upper diagonal corresponds to the correlations among the factors scores derived from Belgian factor scores coefficients. The diagonal (boldface) corresponds to the correlations between the two types of factor scores (theoretically and factorially derived).

Comparison of the Factor Structure in the Belgian and English Data Sets

We then compared the Belgian and British factor structures at three levels: (a) congruence between Belgian factors scores independently derived from Belgian and British factor scores coefficients, (b) congruence—in the Belgian data set—between Belgian and British factor pattern matrices and (c) congruence between factors scores derived theoretically (i.e., from the British scoring key) and factors scores derived from Belgian factor scores coefficients. The first two comparisons were made possible thanks to the contribution of the author of the TEIQue (K. V. Petrides) who granted us access to the complete United Kingdom data set. His sample consisted of 1,139 participants (M age = 31.59, SD = 11.36), of which 49% were male and 51% female.

We first generated and correlated, in our own data set, the factor scores derived from the Belgian factor score coefficients with the factor scores derived from the English factor score coefficients (the latter being independently obtained on the British data set). The Pearson correlations between the factors were .997 for Well-Being, .995 for Sociability, .987 for Self-Control, and .984 for Emotionality (the whole matrix of correlations between these two sets of factors is presented in Table 4). Next, we computed coefficients of congruence ² between Belgian and English factor pattern matrices (using Watkins's [2002b] Coefficient of Congruence Program). Results indicate that congruence coefficients were .98 for Well-Being, .99 for Sociability, .96 for Self-Control, and .97 for Emotionality. Then, as it is still possible that the factor structure in Belgian and English data sets are highly similar, although different from the theoretical factor structure (on which the scoring key is based), we also decided to correlate the factor scores obtained according to the United Kingdom scoring key (see Appendix) with the factor scores based on the factor score coefficients derived from the factor analysis of the Belgian data set. Such analysis was imperative, as there were small discrepancies between our factor structure (see Table 3) and the British theoretical structure (see Appendix). Thus, Self-Motivation, which theoretically does not load on any factor, loaded on the factor Well-Being; Self-Esteem (factor Well-Being) cross-loaded on the factor Sociability; Adaptability, which theoretically does not load on any factor, loaded on the factor Self-Control; and Impulsiveness (low) cross-loaded on the factor emotionality. The foregoing correlations permit to evaluate the implications of these changes as well as the suitability of the UK factor scoring key (see Appendix) for the data at hand. The Pearson correlations were .97 for Well-Being, .97 for Sociability, .95 for Self-Control, and .94 for Emotionality (see Table 5). These values suggest a high level of compatibility between the United Kingdom scoring key and Belgian factor solutions and demonstrate that at least in this case, there is little to be gained by a purely data-driven scoring of the TEIQue factors (i.e., scoring as a function of the factor structure emerging from this particular data set).

Relationship With Demographic Variables

For age, both linear and curvilinear associations were tested, as a curvilinear relationship between age and the scores on the Bar-On (e.g., 1997) Emotional Quotient Inventory (EQ–I) has sometimes been observed (with a peak around 40). Because there was no evidence of a curvilinear relationship in our data, we report linear correlation coefficients. Trait EI as a whole was weakly correlated with age (r = .19, p = .005). At the factor level, correlations with age are quasi null regarding Well-Being (r = .07, p = .10) or Sociability (r = .06, p = .05) and weak regarding Emotionality (r = .13, p = .05). The only factor meaningfully correlated to age was Self-Control (r = .27, p = .005).

With the exception of the factor Well-Being, findings revealed significant gender differences (see Table 1). Women scored significantly higher on Emotionality, whereas men scored significantly higher on Self-Control and Sociability. As a result, men were found to score higher on global trait EI than women.

We asked only the lay people (not the students) to report their occupation. The number of individuals belonging to each sector is reported in Table 2. As some sectors were not chosen by enough individuals to permit valuable analyses, we retained and compared only the sectors selected by more than 20 individuals (i.e., social/healthcare, communication/medias, commerce, public service, finance, research and development, and unemployed people). Results are in Table 6.

TABLE 6 Comparison of Mean Trait EI Scores as a Function of the Professional Sector

Trait El Factor	Occupation(s) Having the Highest Mean Scores (+ M; SDs)	Occupation(s) Having the Lowest Mean Scores (+ M; SDs)	Differences Reaching Statistical Significance
Global Trait EI	—Social/healthcare (4.86; 0.63)—Finance (4.88; 0.37)	—Unemployed (4.52; 0.60)	Social healthcare vs. unemployed: t(75) = 2.06, p < .05, r effect size = .26 Financiersvs. unemploye: t(37) = 2.26, p < .05, r effect size = .34
Well-Being	—Financiers (5.36; 0.65)	—Unemployed (4.78; 1.10)	Financiers vs. unemployed: t(37) = 2.00, p < .055,r effect size =.31
Self-Control	—Social/healthcare (4.51; 0.82)—Finance (4.49; 0.81)	—Commerce (4.10; 0.83)—Unemployed (4.06; 1.00)	Social/healthcare vs. unemployed: t(75) = 1.97, p < .055, r effect size = .25 Social/healthcare vs. commerce: t(77) = 1.96, p < .055, r effect size = .24
Emotionality	—Social/healthcare (5.13; 0.66)—Commerce (5.13; 0.63)	—Public service (4.87; 0.73)—Finance (4.83; 0.50)	Social/healthcare vs. finance: t(43, 76) = 2.074, p < .05, r effect size = .23 Social/healthcare vs. public serice: t(112) = 1.95, p < .055, r effect size = .18
Sociability	—Commerce (5.11; 0.74)	—Unemployed (4.40; 0.74)	Commerce vs. unemployed; t(38) = 2.97, p < .005, r effect size = .43
Note. EI = emotional intelligence.

Individuals belonging to social/healthcare as well as finance-related occupations had the highest mean global trait EI score, whereas, as expected, unemployed individuals had the lowest mean scores. The former differed significantly from the latter. Individuals in other occupations had mean trait EI scores situated between these extremes but did not statistically differ from either social/healthcare and finance occupations or unemployed.

As far as the factors are concerned, financiers exhibited the highest Well-Being scores, whereas unemployed people exhibited the lowest scores. The former differed marginally from the latter. Other occupations' means were in between and did not differ statistically from either financiers or unemployed.

Regarding the factor Self-Control, individuals in social/healthcare as well as in finance-related occupations had the highest scores, whereas individuals in commerce-related occupations and those who were unemployed had the lowest scores. The mean in social/healthcare occupations differed marginally from the mean in unemployed people and in commercial occupations. Other occupations' means were between the highest and lowest groups and did not differ statistically from any of them.

Concerning Emotionality, individuals in social/healthcare as well as in commercial occupations exhibited the highest scores, whereas those in public service and financial occupations exhibited the lowest scores. The mean in social/health occupations (but not in commercial ones) differed significantly from the mean in financial occupations and marginally from the mean in public service. Other occupations' means were in between and did not differ statistically from either social/health, commercial or financial occupations.

With respect to Sociability, commercial occupations had, as expected, the highest scores, whereas unemployed people had the lowest scores; the mean in these occupations were found to differ significantly from each other. It is noteworthy that the mean scores in commercial occupations were significantly higher than the mean scores in all other occupations (which did not differ significantly from each other).

Discriminant Validity ³ With Cognitive Ability (IQ)

Neither global trait EI nor any of its factors was related to nonverbal reasoning (Table 7).

TABLE 7 Convergent, Discriminant, and Criterion Validity of the TEIQue

		Age
Variables	N	M	SD	Balance Females/males (%)	Total Trait EI	Well-Being	Self-Control	Emotionality	Sociability
Convergent/discriminant validity
Optimism	91	18.36	1.52	85/15	.68****	.81****	.48****	.36****	.36****
Alexithymia	457	19.53	2.04	77/23	−.60****	−.36****	−.32****	−.66****	−.39****
F1: DIF	403	19.26	1.90	80/20	−.55****	−.41****	−.44****	−.43****	−.33****
F2: DDF	403	19.26	1.90	80/20	−.49****	−.29****	−.15***	−.67****	−.34****
F3: EOT	403	19.26	1.90	80/202	−.22****	−.07	−.02	−.30****	−.25****
Emotional Stability	80	18.57	1.00	77/23	.42****	.35****	.66****	.01	.15
Introversion	80	18.57	1.00	77/23	−.17	−.24**	−.27**	−.16	−.38****
Openness	80	18.57	1.00	77/23	.41****	.34***	.01	.31***	.43****
Agreeability	80	18.57	1.00	77/23	.47****	.48****	.33***	.26**	.15
Conscientiousness	80	18.57	1.00	77/23	.41****	.19*	.31***	.40****	.28**
Cognitive Ability	109	19.73	2.20	87/13	.04	.05	.13	.04	.00
Social Disirability	131	19.43	1.98	68/32	.44****	.32****	.46****	.23***	.12
Criterion validity
Anxiety	82	18.68	1.03	78/22	−.74****	−.72****	−.59****	−.30***	−.45****
Depression	82	18.68	1.03	78/22	−.55****	−.67****	−.31***	−.28**	−.24**
PQuantSS	93	18.36	1.52	85/15	.23**	.34****	.01	.25**	.09
PQualSS	93	18.36	1.52	85/15	.39****	.40****	.17*	.38****	.22**
State positive affectivity	306	20.01	1.90	67/33	.25****	.30****	.12**	.10*	.15***
State negative affectivity	306	20.01	1.90	67/33	−.37****	−.33****	−.37****	−.23****	−.12**
Emotional reactivity (N)	91	19.42	1.96	48/52	−.26**	−.25**	−.30***	−.11	−.09
Emotional reactivity (S)	96	20.11	2.17	40/60	−.41****	−.33****	−.48****	−.15	−.24**
Note.     TEQue = Trait Emotional Intelligence Questionnaire; F = Factor; DDF = difficulty in describing feelings; EOT = externally oriented thinking; N = sample size; PQuantSS = perceived quantity of social support; pQualSS = perceived quality of social support, (S) = measured in stressful conditions.
*p < .10. **p < .05. ***p < .01. ****p < .001.

Convergent Validity ⁴

Means, standard deviations, and reliability coefficients (Cronbach's alpha) of variables under examination are reported in Table 2. Trait EI was highly associated with optimism (see Table 7). However, as expected, this association seemed mostly attributable to the factor Well-Being, which conceptually overlaps with the construct of optimism. As shown in Table 7, alexithymia and trait EI were found significantly associated. At the factor level, the trait EI factor that is the most highly associated with alexithymia is Emotionality. We expected such overlap, as the trait EI factor Emotionality conceptually overlaps with the alexithymia construct.

In line with the theoretical conceptualization of EI as a personality trait, we observed several correlations between trait EI and the FFM (Table 7). Trait EI was positively associated with Emotional Stability, Openness, Agreeableness, and Conscientiousness. At the factor level, TEIQue and FFM factors that were theoretically expected to correlate with one another did correlate (e.g., Self-Control and Emotional Stability, Sociability and Introversion), whereas factors that were not expected to correlate did not correlate (e.g., Self-Control and Openness).

Susceptibility to Socially Desirable Responding

Overall, trait EI was moderately associated with social desirability scores but the various factors did not contribute equally to this relationship (see Table 7). Additionally, it is of note that gender differences need to be considered regarding the factors Emotionality and Sociability. Indeed, separate analyses checking for gender differences revealed that Emotionality was related to social desirability among women but not among men, whereas Sociability was related to social desirability among men but not among women.

Criterion Validity ⁵

As expected, trait EI was negatively correlated with indicators of anxiety and depression (see Table 7). As shown in Table 7, trait EI was more associated with perceived quality of social support than with perceived quantity of social support. Whereas all trait EI factors were associated with the perceived quality of social support, only Well-Being and Emotionality were associated with perceived quantity of social support.

Trait EI and State PA and State NA

Trait EI and state PA and state NA were assessed several weeks from one another using data from experimental studies on stress resistance in which participants completed the PANAS before and after a mood induction procedure to check the manipulation effectiveness. In this study, we only used the “before manipulation” data (measured on arrival at the laboratory) and examined whether the TEIQue, measured several weeks later was able to predict state affectivity. These data have never been presented as such. In the other studies involving those data (Mikolajczak et al., 2006b), a “before–after manipulation” difference score had been computed and subsequently used to test the success of our manipulation. The correlation between state PA and state NA was only –.17 (p = .01), thereby supporting the current view of PA and NA as distinct dimensions (for a review, see Cropanzano, Weiss, Hale, & Reb, 2003), and this explains why results might be different according to the dimension considered.

As shown in Table 7, trait EI is a significant predictor of both state PA and state NA, although it is more related to NA than to PA. The factors that best predicted state NA were (trait) Well-Being and Self-Control, and the factor that best predicted state PA was (trait) Well-Being.

Trait EI and Emotional Reactivity

Trait EI and emotional reactivity were assessed several weeks apart. Emotional reactivity was assessed during the individual session of an experimental study (i.e., mood induction: stress or neutral), whereas trait EI was assessed during a collective session several weeks later. Although these data are part of a larger research program of which some portion has been presented (Mikolajczak et al., 2006b, Study 2 and 3) or will be presented (Mikolajczak, Roy, Luminet, Fillée, & de Timary, in press), analyses involving emotional reactivity have never been presented.

Correlations (i.e., effect sizes) between trait EI and emotional reactivity are reported separately for the two conditions in Table 7. Multiple regression analyses (reported in Table 8) revealed (a) a main effect of condition indicating that emotional reactivity was higher in stressful than in neutral condition, (b) a main effect of trait EI indicating that higher trait EI scores reported less emotional reactivity than lower scores in both conditions, and (c) an interaction effect between trait EI and condition indicating that the effect of trait EI was even more pronounced in the stressful condition (i.e., higher scores reporting lower reactivity). All trait EI factors had a main effect on emotional reactivity (although it is only marginal in the case of Emotionality). However, in accordance with theoretical expectations, Self-Control was the factor that had the strongest main effect; it was also the only factor having a moderating effect (i.e., having a stronger effect in stressful condition than in neutral condition).

TABLE 8 Hierarchical Regression Analyses Predicting Emotional Reactivity by Condition, Trait EI, and Their Interaction

EI Factor Considered	Criterion Variable	Hierarchical Order	Predictor Variable	R	F Change	Semipartial Correlations^a
Trait EI	ERI	1	Condition	.525	69.87****	−.56
		2	Trait EI	.601	24.78****	−.29
		3	Condition × Trait EI	.614	4.41**	.12
Well-being	ERI	1	Condition	.525	69.87****	−.55
		2	Well-being	.581	17.35****	−.25
		3	Condition × Well-Being	.589	2.57	.10
Self-control	ERI	1	Condition	.525	69.87****	−.55
		2	Self-Control	.621	32.84****	−.32
		3	Condition Self-Control	.633	4.64**	.12
Emotionality	ERI	1	Condition	.525	69.87****	−.53
		2	Emotionality	.538	3.63*	−.13
		3	Eoncition × Emotionality	.541	0.88	.06
Sociability	ERI	1	Condition	.525	69.87****	−.54
		2	Sociability	.552	7.70***	−.15
		3	Condition × Sociability	.558	1.91	.09
Note.    N = 185. EI = emotional Intelligence; ERI = Emotional Reactivity Index. ^aThese correlations are the semipartial correlations, controlling at each step for the effects of the two others.
*p < .10. **p < .05. ***p < .005. ****p < .001.

Incremental Validity

One study provided the necessary data to test the incremental validity of trait EI to predict emotional reactivity over and above social desirability, the three factors of alexithymia, and the Five-factor model of personality. According to Judd and McClelland (statistical course based on Judd & McClelland, 1989), only the constructs that have previously been shown to be valid predictors of the dependent variable (DV) should be included in the multiple regression (principle of economy). Therefore, the first step consisted of performing independent regressions to identify the various predictors of our DV. To be as strict as possible in subsequent incremental analyses, we considered as a predictor any variable whose relation with the DV was significant or marginally significant. These preliminary analyses revealed that Emotional Reactivity was predicted by condition (neutral/stress), global trait EI, its factors of Well-Being and Self-Control, social desirability, and two dimensions of personality (i.e., agreeability and emotional stability). No interaction effect reached significance, which is not surprising, as this sample was smaller than the aggregated sample used to test predictive validity and that interactions are notoriously difficult to detect on small samples (e.g., Cohen, Cohen, West, & Aiken, 2003).

In a second step, we tested the incremental validity of EI over Social Desirability, Agreeableness and Emotional Stability using a partially stepwise procedure (Hunsely & Meyer, 2003). We entered condition as the first block; we entered Social Desirability, Agreeableness, and Emotional Stability in a stepwise fashion as the second block; and we entered Well-Being and Self-Control in a stepwise fashion as the third block.

As depicted in Table 9, trait EI significantly predicted the level of emotional reactivity over and above the effects of condition, Social Desirability, Agreeableness, and Emotional Stability. Among the predictors, the stepwise procedure only retained Emotional Stability as a significant predictor in the second block and Self-Control in the third block. It is noteworthy that complementary analyses entering all the predictors (condition, Social Desirability, Agreeableness, Emotional Stability, Self-Control and Well-Being) together in the equation showed that only condition (β = –.532, p = .001) and Self-Control (β = –.413, p = .005) remained significant when the influence of all others was controlled (i.e., Self-Control has incremental validity over and above the FFM, but the opposite is not true).

TABLE 9 Hierarchical Regression Analyses Testing the Incremental Validity of Trait EI Over Social Desirability, Agreeableness, and Emotional Stability to Predict Emotional Reactivity

Criterion Variable	Forced Hierarchical Order	Predictor Variable	R	F Change	Semipartial r^a
Emotional Reactivity	1	Condition (neutral vs. stress)	.519	22.084**	−.520
	2	Emotional Stability	.614	10.156*	−.053
	3	Self-Control	.709	11.530**	−.322
Note. EI = emotional intelligence. Procedure: Condition was entered as a first block; Social Desirability, Agreeableness, and Emotional Stability were entered in a stepwise fashion as a second block; and Well-Being and Self-Control were entered in a stepwise fashion as a third block. ^athese are the semipartial rs when all variables are considered together (Step 3).
*p < .01. **p < .001.

DISCUSSION

The number of findings involved in this article prevents us from discussing all of them in detail. As most of the results speak for themselves, we have chosen to devote this section to three findings that deserve a more extended discussion: the gender differences on TEIQue scores, the issue of social desirability, and the impact of trait EI on emotional reactivity.

Gender Differences on Trait EI Scores

Significant gender differences were found in the results, with women scoring higher on Emotionality and men scoring higher on Self-Control, Sociability, and the global score. Such results are consistent with those obtained with the English version of the TEIQue (Petrides & Furnham, 2000). They also echo those of Feingold (1994) who meta-analyzed gender differences in personality and found that women were more agreeable (more sensitive, etc.) and men more emotionally stable. Theoretically, the fact that women scored higher on Emotionality is coherent with Western norms according to which expressing emotions is generally viewed as “unmanly” (Brody, 2000, p. 26). Empirically, it is in line with findings obtained using the EQ–I (Slaski, 2001) and with data that showed that men scored slightly higher on alexithymia than women (Parker, Taylor, & Bagby, 2003). In the same vein, the fact that men scored higher than women on Self-Control is on one hand consistent with the divergent socialization of emotion as a function of gender (e.g., “Boys don't cry”) and on the other hand in line with findings obtained using the EQ–I (i.e., men > women on stress management; Bar-On, Brown, Kirkcaldy, & Thome, 2000). On the other hand, the finding that men scored higher than women on Sociability could seem odd at first sight, especially as women are commonly thought to have higher social skills than men (Pease & Pease, 2001); however, a closer look at the subscales that compose the Sociability factor revealed that this factor does not refer to dimensions such as empathy or ability to initiate/maintain close relationships (which abilities are part of the Emotionality factor) but rather to dimensions such as the ability to assert oneself or to influence other's emotions and decisions, which are more easily attributed to males (Kray & Thompson, 2005). Finally, although consistent with findings using the original version of the TEIQue, the fact that men scored higher than women on the global trait EI score is at odds with findings obtained using other trait EI measures such as the Bar-On EQ–I (no gender difference; Slaski, 2001) or the Schutte Emotional Intelligence Scale (EIS); Schutte et al., 1998; no gender difference: Chan, 2003; women > men: Ciarrochi, Chan, & Bajgar, 2001; Schutte et al., 1998). Future studies are thus sorely needed to determine to what extent these differences in self-perceptions (a) are attributable to biases in the test/questionnaire construction, (b) objectify real differences between men and women, (c) are due to women being less confident in their capabilities than men, or (d) are attributable to women being more demanding vis-à-vis themselves than men in the emotional domain. Meanwhile, such gender differences have to be taken into account in the establishment of norms and in the interpretation of scores, especially when women and men have to be compared (e.g., organizational or educational settings).

Susceptibility to Socially Desirable Responding

Our findings suggest that like most self-reports, responses to the TEIQue are not independent of social desirability. The various factors are, however, not equally subject to desirable responding. The weight of the factors as well as the magnitude of the correlations echoes previous findings obtained using the EQ–I (Bar-on et al., 2000; Hemmati, Mills, & Kroner, 2004) and Tett, Fox, and Wang's (2005) trait EI measure. It is noteworthy, however, that the association between social desirability and TEIQue scores has not seemed to pose serious threats to the validity of the findings that have been obtained using the latter, as predictions have remained fully significant after controlling for social desirability in both this article and others' (e.g., Coumans, 2005) studies.

The association between trait EI and social desirability appears mainly attributable to the factors Well-Being and Self-Control. With respect to Well-Being, its relationship with social desirability may be underlain by two reasons. First, it is possible that social desirability influenced self-reported well-being, with individuals higher in social desirability having reported higher levels of well-being. However, there was no apparent reason to present oneself as happier than one really was, especially as (a) the study was totally anonymous, and (b) participants were psychology students, that is, individuals who should have a priori no prejudice vis-à-vis the experience of psychological distress. Therefore, the hypothesis according to which Well-Being may influence responses to social desirability measures should also be considered. Indeed, several studies have shown that positive mood biases judgement positively (e.g., Isen, Shalker, Clark, & Karp, 1978) and promotes a heuristic processing of information (e.g., Chaiken, 1980). Thus, people scoring high on Well-Being may on one hand perceive themselves more positively than they actually are and on the other hand process information in a heuristic way (not going into much details) with the result that they will probably fail to consider a number of isolated events when responding to the desirability items. Considering, for instance, the item “I can remember playing sick to get out of something,” the aforementioned bias associated with positive mood may lead people high in Well-Being to report that they have never played sick. In fact, they did probably play sick once or twice; but and as they usually do not do it, and as their judgement is based on a heuristic examination of the episodic information, they may conclude that they never did.

Regarding the factor Self-Control, its relationship with social desirability is more surprising, especially as additional hierarchical regression analyses revealed that this relationship was neither mediated by conscientiousness, nor by self-esteem, nor by anxiety, nor by a fear of a negative evaluation. Studies that shed light on this association are thus sorely needed. Meanwhile, it is of note that, notwithstanding this partial overlap between Self-Control and social desirability, Self-Control was found to remain entirely predictive of all indicators of resistance to stress (the DV under consideration in our studies) when controlling for social desirability (e.g., Mikolajczak et al., 2006b). In addition, previous studies have also shown that self-control is underlain by clearly identifiable neuroendocrine (i.e., salivary cortisol secretion; Mikolajczak et al., in press) and cognitive processes (including unconscious ones such as early attentional processes; Mikolajczak, Luminet, Roy, & Vestrynge, 2006a), which would not have been the case if Self-Control scores were only the product of social desirability (especially as the association between Self-Control and these cognitive processes remained totally significant when social desirability was partialed out).

Trait EI and Emotional Reactivity

Although measured several weeks apart, Trait EI constituted a significant predictor of lesser emotional reactivity (and this effect held when the effect of social desirability and the FFM was partialed out). This finding is congruent with other analyses performed on the same data set that have shown that trait EI moderates mood deterioration subsequent to stress induction (higher trait EI scores are associated with smaller difference between affectivity at Time 1 and 2; see Mikolajczak et al., 2006b, Studies 2 and 3, as well as Mikolajczak et al., in press).

At first sight, the result that trait EI is associated with less emotional reactivity contrasts with conclusions of Engelberg and Sjöberg (2004) according to which EI would “draw on a disposition to experience more intense affect in response to emotional stimuli” (p. 538). However, Engelberg and Sjöberg's study was cross-sectional and required participants to forecast their emotional intensity in a number of fictitious situations (in brief, they correlated trait EI with the Affect Intensity Measure by Larsen, Diener, & Emmons, 1986) whereas ours is quasi experimental, meaning that people were put in real-life situations. Our findings appear also at odds with a study by Petrides and Furnham (2003), which showed that high-trait EI participants exhibited greater sensitivity to mood induction than their low-trait EI counterparts. However, it is noteworthy that Petrides and Furnham (2003) were not interested in resistance to stress but in sensitivity to emotion-laden stimuli. Accordingly, they manipulated mood through disturbing and cheerful video segments. As Petrides and Furnham's (2003) mood induction procedure clearly targeted another reality than this one, their findings are not necessarily incompatible with ours. It is theoretically conceivable that higher trait EI scores might be more resistant to stress than lower scores and at the same time be more responsive to emotion-eliciting movies (which do not represent a threat to one's integrity). In favor of this view, it is of note that our findings are in accordance with findings of an experimental study by Schutte, Malouff, Simunek, McKenley, and Hollander (2002) that showed that higher trait EI individuals lost less PA and self-esteem after a negative mood-induction through the Velten method (sentences meant to provoke a drop in self-esteem) than did lower trait EI individuals. Our findings are also in line with other experimental studies that have shown that high-trait EI individuals have presented lesser mood deterioration and lesser cortisol secretion after an experimentally induced stress than their lower counterparts (e.g., Mikolajczak et al., 2006b, Study 1; Mikolajczak et al., in press).

The finding that the factor Self-Control was the best predictor of emotional reactivity is in accordance with theoretical expectations (this factor specifically targets the ability to manage one's emotions, stress included). The effect size of its association with emotional reactivity in a neutral condition (r = –.30) and in stressful conditions (r = –.48) was meaningful and warrants mention, especially as the design was prospective. It is also noteworthy that Self-Control had incremental validity to predict emotional reactivity over and above the other trait EI factors, social desirability, and the five factors of personality. The increment in prediction was not only statistically significant, but it was substantial (semipartial r = .32) according to Hunsley and Meyer's (2003) proposed norms for social sciences, which is all the more interesting, as the DVs were measured several weeks apart from trait EI.

GENERAL DISCUSSION

This set of studies represents the most systematic published psychometric investigation of a test/questionnaire within the EI research field. It also constitutes the first rigorous investigation of the psychometric properties of the TEIQue since its introduction by Petrides and Furnham in 2003. Neither the psychometric properties of the original British TEIQue nor those of any other translations have ever been published. Thus, beyond their interest for French-speaking researchers and practitioners, these findings may be of interest to any researcher/practitioner using the TEIQue: They provide both a point of comparison as well as a first indication of what psychometric properties of other versions may look like (although cultural differences are possible).

Results lend preliminary support to the validity of the TEIQue. First and foremost, internal consistencies were generally good. Although they should be improved for 5 subscales (i.e., Impulsiveness, Empathy, Relationship Skills, Adaptability and Self-Motivation), they were satisfactory as far as the 10 others were concerned. Most important, internal consistencies of all factors and global scores were excellent.

Then, the British's four factor structure was globally replicated, confirming previous findings according to which the factor structure has been relatively stable across languages (K.V. Petrides, personal communication, August 2005). Congruence coefficients between Belgian and English factor structures were excellent according to the norms provided by MacCallum, Widaman, Zhang, and Hong (1999) and reflected “a practical identity of the factors” across countries (values greater than .95; Jensen, 1998, p. 99). It shows that a very similar structure of affect-related personality traits can be found in French and English (as well as in Spanish, New-Zealand, and Greek; see previous text). Although there may be cultural differences between these countries with respect to the way a trait is expressed, such cultural differences do not seem to affect the structure of traits in the individual. It thus appears that the TEIQue captures quite well the structure of affect-related traits. This finding is of particular importance, as it is the first EI test/questionnaire to show stability across studies and languages.

Regarding demographic data, TEIQue scores were found independent of age but were influenced by gender, which suggests that separated norms for men and women would be useful. Although we could not construct a proper multitrait-multimethod matrix for this investigation, the TEIQue displayed preliminary evidence of convergent and discriminant validity. First, the various TEIQue factors correlated in meaningful and theoretically congruent ways with alexithymia, optimism, and the Big Five factors of personality. It is noteworthy that trait EI can, however, not be reduced to these constructs because (a) it has demonstrated incremental validity over them in the prediction of a number of variables and (b) it predicts variables that they do not predict (e.g., Mikolajczak, Luminet, & Menil, 2006; Mikolajczak et al., in press; Mikolajczak et al., 2006b).

Conversely, TEIQue scores were unrelated to nonverbal reasoning, which is consistent with theoretical considerations according to which trait EI is a personality trait instead of a form of intelligence (Petrides & Furhnam, 2001, 2003). These results echo previous findings according to which trait EI has been related to personality but not to indicators of IQ, whereas the opposite has been found with ability EI (e.g., for a meta-analytic review of personality and ability correlates of EI, see Van Rooy, Viswesvaran, & Pluta, 2005). Once again, such results stress the necessity of changing trait EI's label so that it does not contain the notion of intelligence anymore.

Although the foregoing findings are necessary conditions for a construct to be said valid, they are not sufficient. Ultimately, it is the construct's ability to predict outcomes of interest that will determine its utility. It is vital here to understand that “to predict” does not equal “to cause.” Saying that an instrument A is a good predictor of a variable B means that when you know the score of an individual on A, you have a good idea of what his or her score/behavior on B may be. This does not mean that A caused B (e.g., height and weight are related but they are both the product of a third variable).

The prediction of “objective” life outcomes such as professional success or medical status was beyond the scope of this study. All criteria we examined in this study were self-reported and pertained to subjective emotional adjustment. These are, however, not trivial criteria, especially as emotional welfare is a major goal in life for most people. Furthermore, examining such criteria was essential to ensure construct validity (i.e., if trait EI is a valid construct, it has to predict a higher emotional adjustment). In this respect, the TEIQue predicted a substantial and meaningful part of variance of depression and anxiety, perceived quality and quantity of social support, future state PA and NA, and emotional reactivity in stressful conditions. These findings, far from isolated, are consistent with previous studies that have been conducted with the French version of the TEIQue: TEIQue scores were found to predict students' resistance to examination stress (Mikolajczak et al., 2006), mood deterioration and neuroendocrine reactivity following a laboratory stressor (Mikolajczak et al., 2006b; Mikolajczak et al., in press), as well as the level of burnout and somatic complaints among nursing and call center employees (Mikolajczak et al., in press; van Kan, 2004). In conjunction with the findings accumulated using the original (English) version of the TEIQue regarding the prediction of both subjective and objective criteria (see previous text), the capacity for French TEIQue scores to predict resistance to stress in both applied and experimental settings constitutes evidence in favor of its validity and practical utility.

Although such evidence of predictive validity is a necessary condition for a test to be claimed useful, it is, however and once again, not sufficient. To be deemed useful, the TEIQue should also demonstrate that (a) findings are not attributable to social desirability and (b) that other tests/questionnaires cannot do the job, at least not as efficiently. This condition is of particular importance here because TEIQue scores were precisely found to be associated with social desirability. By showing that the TEIQue predicts emotional reactivity over and above social desirability, alexithymia, and the Five-factor model of personality, this article adds to the growing body of literature that has shown that the TEIQue does indeed predict variance over and above the Giant Three, the Big Five, and other cognate constructs (for examples in French, see Mikolajczak et al., in press; Mikolajczak et al., in press; and Mikolajczak et al., 2006b; for examples in other languages, see Furnham & Petrides, 2003; Petrides & Furnham, 2003; Petrides et al., 2004, 2006, 2007).

The foregoing represents modest but promising findings in favor of the validity and usefulness of the TEIQue. However, several limitations have to be acknowledged. First, most of these studies involve students, thus raising concerns about the generalizability of the findings to other population strata. It also restricts range, especially regarding age, for which results have to be interpreted with caution. Second, we measured all study variables through self-reports, which potentially introduced a “shared method covariance” bias. In this respect, it cannot be excluded that the absence of relationship between trait EI and nonverbal reasoning (indicator of IQ) was simply the product of divergent measurement methods (self-report vs. performance), just like the quasi-null relationship between ability and trait EI (which have been found uncorrelated, although their sampling domains are closely related). Third, the results of the investigation of mean trait EI differences across occupational sectors should be interpreted with caution, as only a limited sample of occupations was included in the analysis (i.e., we excluded those that were underrepresented in our sample). Furthermore, albeit their appealing character, it should be mentioned that the cross-sectional methodology adopted does not allow to draw conclusions in terms of causality. However, it is striking that most mean differences were in line with expectations. It is especially noteworthy that unemployed people had the lowest EI scores, which is in accordance with the current idea that EI is a necessary (although not sufficient) condition for professional success. Finally, in addition to addressing these issues, future studies would certainly benefit from considering other types of criterions such as behavioral (e.g., prosocial behaviors, performance) or medical ones (e.g., health status).

APPENDIX

Factorial and Subscale Structure of the TEIQue (Petrides & Furnham, 2003).

Table

Factor Facets	High Scorers Perceive Themselves as …
Well-Being
Self-esteem	Successful and self-confident
Trait happiness	Cheerful and satisfied with their lives
Trait optimism	Confident and likely to “look on the bright side” of life
Self-Control
Emtion regulation	Capable of controlling their emotions
Stress management	Capable of withstanding pressure and regulating stress
Impulsiveness (low)	Reflective and less likely to give in to their urges
Emotionality
Emotion perception (self and others)	Clear about their own and other people's feelings
Emotion expression	Capable of communicating their feelings to others
Relationship skills	Capable of having fulfilling personal relationships
Empathy	Capable of taking someone else's perspective
Sociability
Social competence	Accomplished networkers with excellent social skills
Emotion management (others)	Capable of influencing other people's feelings
Assertiveness	Forthright, frank, and willing to stand up for their rights
The following subscales do not belong to any particular factor and are directly included in the total score.
Adaptability	Flexible and willing to adapt to new conditions
Self-Motivation	Driven and unlikely to give up in the face of adversity

ACKNOWLEDGMENTS

Cécile Leroy is now at the ORBEM [Office Régional Bruxellois de l'Emploi], Brussels, Belgium. Preparation of this article was facilitated by an FSR grant from the Research Special Funds of the University of Louvain as well as a travel grant to London from the Belgian French-speaking Government accorded to M. Mikolajczak, and Grants 1.5.146.02, 1.5.123.04, and 1.5.175.06 from the Belgian National Fund for scientific Research accorded to O. Luminet.

We thank K. V. Petrides for granting us access to the United Kingdom data for the TEIQue Version 1.50.

Notes

¹The rationale underlying parallel analysis is that nontrivial components from real data with a valid underlying structure should have larger eigenvalues than parallel components derived from random data having the same sample size and number of variables (Hayton et al., 2004). The Marley Watkins program generates completely random data sets having the same sample size and containing the same number of variables than the real data set (up to 1,000 replications possible). Then it generates the correlations matrices for these data sets and provides their average eigenvalues. The researcher can then compare the eigenvalues generated from these random data sets to the real data set to ensure that the former are significantly lower than the latter.

²The congruence coefficient (rc) is an index of factor similarity. It is typically used to determine the factorial invariance of solutions across samples or studies.

³ Discriminant validity refers to the degree to which scores on a test do not correlate with (are independent of or orthogonal to) variables they are not supposed to correlate with given the nature of the construct.

⁴ Convergent validity refers to the degree to which scores on a test correlate with those on a test that is believed to measure a closely related construct (i.e., the two tests should end up ranking people in pretty much the same way).

⁵ Criterion validity refers to the ability of a test to predict some criterion it should theoretically be able to predict. Criterion-related validity can either be concurrent or predictive. Concurrent validity refers to the correlation between the predictor and criterion scores obtained at approximately the same time. Predictive validity refers to the degree to which scores on a test predict future behavior on a criterion variable.