Last update05:49:11 AM GMT


The Bar-On model of emotional-social intelligence (ESI) - The rigorous development of the EQ-i helped create a robust model of ESI

Индекс материала
The Bar-On model of emotional-social intelligence (ESI)
The theoretical foundation of the Bar-On model
Description of the instrument used to develop the Bar-On model (the EQ-i)
The rigorous development of the EQ-i helped create a robust model of ESI
The construct validity of the EQ-i confirms that the Bar-On model is describing ESI
The Bar-On model of ESI predicts various aspects of human performance
The Bar-On model is teachable and learnable
Все страницы

The rigorous development of the EQ-i helped create a robust model of ESI

The EQ-i was originally constructed as an experimental instrument designed to examine the conceptual model of emotional and social functioning that I began developing in the early 1980s (1988). At that time, I hypothesized that effective emotional and social functioning should eventually lead to a sense of psychological well-being. It was also reasoned that the results gained from applying such an instrument on large and diverse population samples would reveal more about emotionally and socially intelligent behavior and about the underlying construct of emotional-social intelligence. Based on findings obtained from applying the EQ-i in a wide range of studies over the past two decades, I have continuously molded my conceptualization of this construct; these changes have been mild and are ongoing in an effort to maintain a theory that is empirically based.

The development of the Bar-On model and measure of ESI proceeded in six major stages over a period of 17 years: (1) identifying and logically clustering various emotional and social competencies thought to impact effectiveness and psychological well-being based on my experience as a clinical psychologist and review of the literature; (2) clearly defining the individual key clusters of competencies, skills and facilitators that surfaced; (3) initially generating approximately 1,000 items based on my professional experience, review of the literature and input from experienced healthcare practitioners who were asked to generate questions they would ask in an interview situation guided by my definitions; (4) determining the inclusion of 15 primary scales and 133 items in the published version of the instrument based on a combination of theoretical considerations and statistical findings generated by item analysis and factor analysis; (5) initially norming the final version of the instrument on 3,831 adults in North America in 1996; and (6) continuing to norm and validate the instrument across cultures. The first normative sample of the EQ-i included individuals from every Canadian province and from nearly all the states in the US. The gender-age composition of the sample included 49% males and 51% females from 16 to 100 years of age, with an average age of 34.3 years. The sample was 79% White, 8% Asian American, 7% African American, 3% Hispanic, and 1% Native American.5 For more detailed demographic information, including the educational and occupational background of the original normative sample, the reader is referred to the instrument's technical manual (Bar-On, 1997b).

The EQ-i has been translated into more than 30 languages,6 and data have been collected in numerous settings around the world. Earlier versions of the inventory were completed by a total of 3,000 individuals in six countries (Argentina, Germany, India, Israel, Nigeria and South Africa). The first translation of the EQ-i was from English to Spanish to allow for extensive data collection in Argentina,7 which was followed by data collection in a number of other countries. In addition to providing cross-cultural data, this preliminary piloting of the EQ-i was important for item selection and alteration, continued scale development and validation, and establishing the final nature of the response format.

Numerous reliability and validity studies have been conducted around the world over the past two decades, a number of which will be referred to in the following sections to describe the reliability and validity of the EQ-i and the construct it measures.

The outcome of this rigorous development process has rendered psychometric properties that shed light on the validity and robustness of the model. After discussing the age-gender effect, factorial structure and reliability, I will focus primarily on the construct validity and predictive validity of the model. This approach of examining the validity of a concept by examining the psychometric properties of scales designed to measure that concept is not uncommon in psychology in general as well as in the specific area of ESI [e.g., Newsome et al., 2000; Petrides & Furnham, 2000; Salovey et al., 1995; Van Rooy & Viswesvaran, 2004].The impact of age, gender and ethnicity on the Bar-On model. An analysis of variance of the North American normative sample (n= 3,831) was conducted to examine the effect of age, gender and ethnicity on EQ-i scores (Bar-On, 1997b). It was thought that the results would also shed light on the underlying construct of ESI.

Although the results indicated a few significant differences between the age groups that were compared, these differences are relatively small in magnitude. In brief, the older groups scored significantly higher than the younger groups on most of the EQ-i scales; and respondents in their late 40s obtained the highest mean scores. An increase in emotional-social intelligence with age is also observed in children (Bar-On & Parker, 2000). The findings presented here, which are based on a cross-sectional comparison of different age groups, will eventually be compared with findings from an ongoing longitudinal study of the same cohort (n= 23,000) over a 25-year period from birth to young adulthood. This will provide a more accurate indication of how ESI develops and changes over time.8 Similar increases in ESI with age have been reported by others based on employing the EQ-i, MEIS9 and other measures of this construct (Goleman, 1998). These findings are interesting when one considers that cognitive intelligence increases up until late adolescents and then begins to mildly decline in the second and third decades of life as was originally reported by Wechsler (1958). The results suggest that as one gets older, one becomes more emotionally and socially intelligent.

With respect to gender, no differences have been revealed between males and females regarding overall ESI. However, statistically significant gender differences do exist for a few of the factors measured by the EQ-i, but the effects are small for the most part. Based on the North American normative sample (Bar-On, 1997b), females appear to have stronger interpersonal skills than males, but the latter have a higher intrapersonal capacity, are better at managing emotions and are more adaptable than the former. More specifically, the Bar-On model reveals that women are more aware of emotions, demonstrate more empathy, relate better interpersonally and are more socially responsible than men. On the other hand, men appear to have better self-regard, are more self-reliant, cope better with stress, are more flexible, solve problems better, and are more optimistic than women. Similar gender patterns have been observed in almost every other population sample that has been examined with the EQ-i. Men's deficiencies in interpersonal skills, when compared with women, could explain why psychopathy is diagnosed much more frequently in men than in women; and significantly lower stress tolerance amongst women may explain why women suffer more from anxiety-related disturbances than men (American Psychiatric Association, 1994).

An examination of the North American normative sample, upon which the EQ-i was normed, did not reveal significant differences in ESI between the various ethnic groups that were compared (Bar-On, 1997b, 2000, 2004; Bar-On & Parker, 2000). This is an interesting finding when compared with some of the controversial conclusions that have been presented over the years suggesting significant differences in cognitive intelligence between various ethnic groups (e.g., Suzuki & Valencia, 1997).

To summarize the above findings, the Bar-On model reveals that older people are more emotionally and socially intelligent than younger people, females are more aware of emotions than males while the latter are more adept at managing emotions than the  former,   and  that  there  are  no  significant  differences  in emotional-social intelligence between the various ethnic groups that have been examined in North America.

The factorial structure of the Bar-On model.Factor analysis was applied to study the 15-factor structure of the EQ-i to empirically evaluate the extent to which it is theoretically valid. Moreover, this statistical procedure was used to examine the factorial structure of the Bar-On model (i.e., to examine the extent to which the factorial components of this model structurally exist). This analysis was first performed on the normative sample, progressing from exploratory to confirmatory factor analysis (Bar-On, 1997b).

Based on a varimax rotation, a 13-factor solution afforded the most theoretically meaningful interpretation. These results provided a reasonable match with the subscale structure of the EQ-i. Nonetheless, the 13-factor empirical structure that emerged raised an important question that had to be addressed: Can the 15-factor model used in the Bar-On model and measure of ESI still be justified in light of the findings which suggested a 13-factor structure? The essential differences that were identified between the theoretical structure and the one that surfaced as a result of exploratory factor analysis were as follows: (a) two factors emerged from the Impulse Control items; (b) although Self-Regard, Self-Actualization, Optimism and Happiness represent four separate scales, most of their items loaded on two factors; (c) although Assertiveness and Independence are considered to be two separate subscales, items from both subscales loaded on one factor; and (d) although two separate experimental factors emerged from the Empathy and Social Responsibility items, they are the two highest correlating factors (.80).

A confirmatory factor analysis was initially applied to resolve the above-mentioned differences between the 15-factor structure of the Bar-On model and the 13 factors that emerged from the exploratory factor analysis. Although the results supported a 15-factor structure in the end, which fits the theoretical basis of the Bar-On model and measure (Bar-On, 1997b), an additional confirmatory factor analysis was subsequently applied to the same dataset (n= 3,831) in an attempt to explore an alternative factorial structure (Bar-On, 2000). The items from the above-mentioned problematic factors (Independence, Self-Actualization, Optimism, Happiness, and Social Responsibility) were excluded from the second analysis. Self-Actualization, Optimism and Happiness were excluded from this analysis in that a number of their items loaded on the Self-Regard factor while others loaded on an additional yet weaker factor; moreover, these three factors appear in the literature primarily as facilitators of ESI rather than actual components of the construct itself; Wechsler referred to them as «conative factors» (1940, 1943). Independence was excluded from the analysis because its items loaded heavily on the Assertiveness factor, and because it rarely appears in the literature as an integral component of ESI; however, assertiveness (the ability to express one's emotions and feelings) most definitely appears in the literature, from Darwin to the present, as an important part of this construct. For similar empirical and theoretical reasons, it was decided to exclude Social Responsibility items; moreover, this subscale was shown to correlate extremely high with Empathy as was previously mentioned, meaning that they are most likely measuring the same domain.

The results of this second analysis clearly suggested a 10-factor structure, which is both empirically feasible and theoretically acceptable as an alternative to the above-mentioned 15-factorstructure. In the order of their extraction, the ten factors that emerged are: (1) Self-Regard, (2) Interpersonal Relationship, (3) Impulse Control, (4) Problem-Solving, (5) Emotional Self-Awareness, (6) Flexibility, (7) Reality-Testing, (8) Stress Tolerance, (9) Assertiveness, and (10) Empathy. These ten factors appear to be the key components of ESI, while the five factors that were excluded from the second confirmatory factor analysis (Optimism, Self-Actualization, Happiness, Independence, and Social Responsibility) appear to be important correlates and facilitators of this construct. The ten key components and the five facilitators together describe and predict emotionally and socially intelligent behavior, as will be shown below.

The factorial validation of the EQ-i presented here compares favorably with that of the MSCEIT and ECI.i°

The reliability of the Bar-On model.The reliability of the EQ-i has been examined by a number of researchers over the past 20 years. A consensus of findings reveals that the Bar-On conceptual and assessment model is consistent, stable and reliable (Bar-On, 2004). More specifically, the overall internal consistency coefficient of the EQ-i is .97 based on the North American normative sample (Bar-On, 1997b). This well exceeds the .90 minimum for total scores suggested by Nunnally (1978). Internal consistency was recently reexamined on 51,623 adults in North America, revealing nearly identical results with a slight mean increase of .025 in consistency coefficients (Bar-On, 2004). An overall retest reliability examination of the EQ-i is .72 for males (n= 73) and .80 for females (n= 279) at six months (Bar-On, 2004). Other researchers around the world have reported similar findings regarding the reliability of the EQ-i (e.g., Matthews et al., 2002; Newsome et al., 2000; Petrides & Furnham, 2000). These findings compare favorably with those of other measures of this construct.11

To summarize, the findings presented here demonstrate that there is good consistency within the factorial components of this model as well as stability over time.