Emotional Intelligence 1 Running head: EMOTIONAL INTELLIGENCE AND SOCIAL INTERACTION Emotional Intelligence and Social Interaction

Two studies found positive relationships between the ability to manage emotions and the quality of social interactions, supporting the predictive and incremental validity of an ability measure of emotional intelligence, the MSCEIT (Mayer-Salovey-Caruso Emotional Intelligence Test, 2002). In a sample of 118 American college students (Study 1), higher scores on the managing emotions subscale of the MSCEIT were positively related to the quality of interactions with friends, evaluated separately by participants and two friends. In a diary study of social interaction with 103 German college students (Study 2), managing emotions scores were positively related to the perceived quality of interactions with opposite sex individuals. Scores on this subscale were also positively related to perceived success in impression management in social interactions with individuals of the opposite sex. In both studies, the main findings remained statistically significant after controlling for Big Five personality traits. Emotional Intelligence 3 Emotional Intelligence and Social Interaction Emotional competencies are thought to be important for social interaction because emotions serve communicative and social functions, conveying information about people’s thoughts and intentions, and coordinating social encounters (e.g., Keltner & Haidt, 2001). Positive emotionality is associated with sociability (e.g., Argyle & Lu, 1990), whereas persistent negative affect keeps others at bay (e.g., Furr & Funder, 1998). Accordingly, people need to process emotional information and manage emotional dynamics intelligently in order to navigate the social world. Yet few studies to date have examined the relationship between individual differences in emotional competencies and social adaptation in adult, non-clinical populations. The idea that emotional competencies are crucial for adaptation in various realms of life has fueled interest in the concept of emotional intelligence (EI) and inspired numerous programs of social and emotional learning in school and work settings. Nonetheless, research on EI is still limited, and the construct has been criticized on several grounds, including lack of evidence of incremental validity, problems of assessment, conformity bias, and cultural differences in emotional expression (e.g., Roberts, Zeidner, & Matthews, 2001). The present studies were based on Salovey and Mayer’s (1990; Mayer & Salovey, 1997) theory of emotional intelligence. Whereas other authors have written about EI as a much broader construct (e.g., Bar-On, 2000), the Mayer and Salovey (1997) model focuses on emotion-related competencies that can be assessed through performance-based tests. The most recently developed ability measure of EI is the MSCEIT (Mayer-Salovey-Caruso Emotional Intelligence Test; Mayer, Salovey, & Caruso, 2002), which measures four core emotional competencies, including perceiving emotions, using emotions to facilitate thinking, understanding emotions, and managing emotions. We used the MSCEIT to examine the relationship between individual differences in emotional competencies and the quality of social interactions in the United States and Germany, and to evaluate the incremental validity of the MSCEIT in relation to Big Five personality traits. Evidence for the construct validity of EI, conceptualized as a set of abilities, is accumulating (for reviews see Brackett, Lopes, Ivcevic, Mayer, & Salovey, in press; Mayer, Emotional Intelligence 4 Salovey, Caruso, & Sitarenios, 2001; Salovey, Mayer, & Caruso, 2002). Confirmatory factor analyses have supported the four-branch model proposed by Mayer and Salovey in 1997 (Mayer, Salovey, Caruso, & Sitarenios, 2003). Higher scores on emotional intelligence tests have been associated with various indicators of social adaptation, including: more prosocial behavior among schoolchildren (Rubin, 1999); greater empathy (Ciarrochi, Chan, & Caputi, 2000); and fewer negative interactions with friends, among college students (Brackett, Mayer, & Warner, in press). In a recent study by Lopes, Salovey, and Straus (2003), college students scoring higher on the managing emotions subscale of the MSCEIT reported less conflict and antagonism in their relationship with a close friend, as well as more companionship, affection, and support in their relationship with a parent. These associations remained significant after controlling for Big Five personality traits. There is converging evidence from other lines of research that emotional competencies are associated with social adaptation. A large number of studies with children suggest that the capacity to decode, understand, and regulate emotions is associated with social and emotional adaptation (e.g., Eisenberg, Fabes, Guthrie, & Reiser, 2000; Halberstadt, Denham, & Dunsmore, 2001). Evaluations of school-based interventions emphasizing the development of emotional competencies also suggest that emotional learning contributes to social and academic adjustment (Greenberg, Kusché, Cook, & Quamma, 1995). Nevertheless, studies assessing a range of emotional competencies among adult, nonclinical populations have generally relied on criteria assessed through single-administration selfreport instruments. The present studies sought to extend previous research by collecting both selfand peer reports of quality of social relationships, in one case, and repeated measures of social interaction, in another. Moreover, we conducted these studies in two different cultures to enhance the generalizability of our findings. Among the four core emotional competencies proposed by Mayer and Salovey (1997), we expected that the ability to manage emotions would be most strongly associated with the quality of everyday social interaction for several reasons. First, the ability to regulate emotions is likely to influence the emotional valence of social interactions, as we infer other people’s Emotional Intelligence 5 intentions from their emotional cues, use others’ emotions as guides for our own behavior, or simply catch others’ emotions through emotional contagion (e.g., Hatfield, Cacioppo, & Rapson, 1994). Second, the ability to manage emotions may influence people’s motivation and expectations for social interaction (e.g., Cunningham, 1988), as well as their use of effective interaction strategies (e.g., Furr & Funder, 1998; Langston & Cantor, 1989). Third, the ability to manage emotions may facilitate a flexible focus of attention, which is important for smooth communication and social interaction. Negative affect can induce self-focused attention (e.g., Pyszczynski, Holt, & Greenberg, 1987), which is likely to make people less attentive to those around them. Fourth, the ability to manage emotions may facilitate executive functions associated with the coordination of numerous skills required for social behavior. This is apparent when unregulated social anxiety inhibits spontaneity and leads to overly constrained behavior. More generally, the capacity to regulate one’s own emotions seems to be linked to a broader capacity for self-control, including the control of impulsive behavior (e.g., Baumeister, Heatherton, & Tice, 1994). Additionally, the managing emotions subscale of the MSCEIT taps into emotional regulation in interpersonal contexts, and is therefore likely to be a more proximal predictor of the quality of social interactions than other MSCEIT subscales. The abilities to perceive, use, and understand emotions may influence the quality of social interactions more indirectly. For example, understanding emotional dynamics may help one to anticipate one’s own and others’ emotional reactions and thereby to manage emotions effectively during a tense encounter. Thus, it is likely that the abilities to perceive, use, and understand emotions will have only a weak effect on the overall quality of social relationships. This is all the more likely because people draw on many different skills when interacting with others, so that any one skill has only a diluted impact on social adaptation. In the present studies, we examined both EI and Big Five personality traits as predictors of the quality of social relationships because both emotional competencies and personality traits are likely to influence social adaptation (Lopes, Salovey, et al., 2003). Whereas personality theory emphasizes temperamentally-driven dispositions, the theory of emotional intelligence Emotional Intelligence 6 emphasizes acquired competencies that help people to regulate their emotions and manage social interactions. Competencies and dispositions thus provide distinct and complementary perspectives for understanding social and emotional adaptation. Yet traits and skills, or personality and intelligence (e.g., EI), are likely intertwined (see Sternberg & Ruzgis, 1994). In assessing the incremental validity of emotional intelligence in the present studies, we statistically controlled for Big Five personality traits because some self-report measures of EI, such as the EQi (Bar-On, 1997), have been found to overlap meaningfully with the Big Five (e.g., Brackett & Mayer, in press; Davies, Stankov, & Roberts, 1998). However, it is important to note that controlling for the Big Five is likely to be an overly stringent test of incremental validity. On the one hand, there may be conceptual overlap and common method variance between the Big Five (e.g., extraversion and agreeableness) and criteria such as self-perceived satisfaction with social interactions. On the other hand, there may be conceptual overlap between personality traits such as neuroticism and measures designed to assess emotional regulation (e.g., Watson, 2000). We view social adaptation as a multi-faceted construct, because people may be fairly well adjusted in one context but less well adjusted in others. Indeed, both theory and research suggest that social competence does not represent a cohesive domain of ability (e.g., Cantor & Kihlstrom, 1987; Hall & Bernieri, 2001). Conducting two separate studies allowed us to focus on relationships with friends in Study 1 and examine different types of social interactions in Study 2. Study 1 This study examined the relationship between emotional competencies, assessed by the MSCEIT, and the quality of relationships with friends, evaluated by participants and two friends. Friendships represent a crucial domain of social adaptation where emotional competencies are likely to play an important role because of the intimacy and emotional involvement associated with close friendships. Hypotheses In light of previous theory and research, we formulated two general hypotheses: Emotional Intelligence 7 1. Emotional intelligence (particularly the managing emotions subscale of the MSCEIT) will be positively associated with participants’ and friends’ reports of friendship quality. 2. Emotional intelligence (particularly managing emotions) will explain friendship quality over and above the variance accounted for by the Big Five. Method Participants and Procedure Participants were 118 college students (26 males and 92 females) between the ages of 17 and 24, who took part in a larger study conducted at the University of New Hampshire. Ninetyeight percent were white and from the New England area of the United States. All received course credit for participating in the study. The measures used in this study were completed at various times throughout the semester and included the MSCEIT, the Big Five, emotional regulation, and social desirability, as well as measures of friendship quality and interpersonal competence. In addition to completing these measures, each participant was asked to recruit two friends to evaluate the quality of their relationship with the participant. Informant surveys were sealed in envelopes, and participants were asked to give them to two friends or close acquaintances they had met at the University of New Hampshire (with whom they were not romantically involved). These instructions were intended to limit extraneous variability associated with the choice of informants, especially considering that romantically charged interactions may be qualitatively different from other friendships. The envelopes were coded and addressed to the principal investigators so they could be easily returned via campus mail. Complete data (i.e., peer reports from two friends) were received for 66 participants. Measures Emotional intelligence. The MSCEIT (Version 2.0; Mayer et al., 2002) measures the abilities to perceive, use, understand, and manage emotions. For each subscale, there are two tasks. For the Perceiving Emotions subscale, respondents identify the emotions in photographs of faces (Faces task), as well as in designs and landscapes (Pictures task). For Using Emotions, respondents describe emotions with non-emotional vocabulary (Sensations), and indicate the Emotional Intelligence 8 feelings that might facilitate or interfere with the successful performance of various cognitive and behavioral tasks (Facilitation). Understanding Emotions is assessed with questions concerning the manner in which emotions evolve and transition over time (Changes), and how some feelings are produced by blends of emotions (Blends). The ability to Manage Emotions is assessed through a series of scenarios in which people identify the most adaptive ways to regulate their own feelings (Emotion Management) and the feelings of others in social situations (Social Management). This last subscale will be referred to as “managing emotions”. The publisher of the MSCEIT requests that we not reproduce actual test items for fear of invalidating the test. The following abridged examples of items were considered when the MSCEIT was being developed. For Using Emotions: “Imagine feeling surprised because you got a birthday present that was totally unexpected. How much is the feeling of surprise like each of the following? Cold, blue, sweet...” For Understanding Emotions: “Tom felt anxious, and became a bit stressed when he thought about all the work he needed to do. When his supervisor brought him an additional project, he felt _______. Overwhelmed, depressed, ashamed, selfconscious, jittery.” For Managing Emotions in oneself: “Debbie just came back from vacation. She was feeling peaceful and content. How well would each action preserve her mood? (1) She started to make a list of things at home that she needed to do. (2) She began thinking about where and when to go on her next vacation. (3) She called a friend to tell her about the vacation...” For Managing Emotions in interpersonal situations: “Ken and Andy have been good friends for over 10 years. Recently, however, Andy was promoted and became Ken’s manager. Ken felt that the new promotion had changed Andy in that Andy had become very bossy to him. How effective would Ken be in maintaining a good relationship, if he chose to respond in each of the following ways? (1) Ken tried to understand Andy’s new role and tried to adjust to the changes in their interactions. (2) Ken approached Andy and confronted him regarding the change in his behavior...” The MSCEIT can be scored using both expert and consensus norms. Expert scores reflect the agreement between participants’ responses and those of an expert panel of 21 emotion researchers from various nations (43% United States, 19% Canada, 33% Western Europe, 5% Emotional Intelligence 9 Israel). For example, if someone answers “A” to the first question and 24% of experts also answered “A”, this person obtains a raw score of .24 for the first question. Consensus scores reflect the agreement between participants’ responses and those of the normative sample, which consists of 5,000 English-speaking people from various nations (63% United States, 17% Canada, 7% Western Europe, and 13% other). Split-half reliabilities for the normative sample, based on expert scores and corrected by the Spearman-Brown formula, range from .76 to .90 for the four subscales of the MSCEIT (Mayer et al., 2002). Reliabilities for the present sample were .80, .67, .60, and .73 respectively, for subscales 1 to 4. Scores based on consensus norms correlate highly (r > .90) with those based on expert norms (Mayer et al., 2001; Mayer et al., 2003). We report split-half reliabilities for the MSCEIT due to item heterogeneity, as each subscale comprises two tasks tapping into somewhat different abilities. In the present studies, all analyses were based on expert scores because expert norms may be less susceptible to cultural bias, and the expert panel was more internationally diverse than the normative sample. Note that expert and consensus scores were highly correlated (r’s > .90) in both studies. MSCEIT scores computed by the test publishers in North America are standardized (M = 100, SD = 15). These scores are based on 122 out of 141 items because psychometric analyses on the normative sample suggested exclusion of 19 items. These were not deleted from the actual test so as to preserve a balanced layout with the same number of items for all questions. Personality. The Big Five personality traits (Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness) were assessed with the 240-item NEO-PI-R (Costa & McCrae, 1992). Reliabilities were high for all five factors (α > .87). Socially desirable responding. Social desirability bias was assessed with the Paulhus Deception Scales (Paulhus, 1998), which include 40 items yielding two scores. Self-Deceptive Enhancement (SDE) measures the tendency to give inflated but honest self-descriptions reflecting lack of insight or narcissistic bias. Impression Management (IM) measures the tendency to provide consciously inflated self-descriptions (i.e., faking or lying) to produce a desired effect. The scale uses a 1 (not true) to 5 (very true) response format. Alphas were .83 (SDE) and .71 (IM) for the present sample. Emotional Intelligence 10 Self-perceived emotion regulation skills. The skills version of the Emotion Regulation Scale (ERS; Kovacs, 2002) assesses the degree to which people use adaptive strategies and avoid maladaptive strategies to regulate emotional experience (including physical/biological, behavioral, cognitive, and social-interpersonal strategies). This scale has 68 items using a 3-point response format: “not true of me,” “sometimes true of me,” and “many times true of me” (α = .81). Because this scale was administered as part of another study, data were available for only a subset of participants in the present study. Quality of interpersonal relationships. Both participants’ and friends’ reports on the quality of interpersonal relationships were obtained using an abridged version of the Network of Relationships Inventory (NRI; Furman & Buhrmester, 1985; see also Furman, 1996). The full measure includes 30 items and yields three factor scores: positive interaction (social support), negative interaction (conflict), and power imbalance. Given our hypotheses and time constraints, participants answered 20 items (14 for positive and 6 for negative interaction) about their relationship with each of two friends. Each friend also answered the same 20 items about his or her relationship with the participant. The positive interaction factor measures companionship, instrumental aid, intimacy, nurturance, affection, admiration, and alliance, whereas the negative interaction factor measures conflict and antagonism in the relationship. We used a 9-point Likert scale anchored at “not at all” and “extremely”. Reliabilities for both participants’ and friends’ reports on both scales were high (α > .88). Interpersonal competence. Both selfand peer-reports on two dimensions of interpersonal competence (emotional support and conflict resolution) were obtained using the Adolescent Interpersonal Competence Questionnaire (AICQ; Buhrmester, Furman, Wittenberg, & Reis, 1988). Each scale contained 8 items and respondents answered questions using a 5-item Likerttype scale ranging from “poor at this” to “good at this.” The reliabilities of both self-and peerreports on both scales were high (α > .84). Results Three participants were eliminated from all analyses because they scored more than 3 standard deviations below the mean for the present sample on the managing emotions subscale of Emotional Intelligence 11 the MSCEIT, suggesting they responded randomly on the test. Analyses involving friends’ ratings were based on 66 participants for whom complete data (i.e., reports from two friends) were available. Multivariate analysis of variance indicated that this subset of participants did not differ significantly from other participants on the predictor variables, including EI and Big Five scores (Wilk’s Lambda = .917, F(6, 108) = 1.65, p = .14). There were also no significant differences in age or gender. These analyses allayed concerns about attrition bias. The similarity of participants' reports about the quality of relationships with their two friends was assessed using intraclass correlations. The intraclass correlations between participants' reports for both friends were .62 and .75 for the NRI subscales of positive and negative interaction, respectively. Based on this, participants' reports about the quality of relationships with their two friends were averaged to yield a single score for each participant, for each subscale of the NRI. The similarity of friends' reports about the quality of their relationships with participants were also assessed using intraclass correlations. Intraclass correlations between both friends' ratings of each participant were .58 for NRI positive interaction, .43 for NRI negative interaction, .53 for AICQ emotional support, and .58 for AICQ conflict resolution. Based on this, the ratings provided by both friends about each participant were averaged to yield a single score for each subscale of the NRI and AICQ. Correlational Analyses Table 1 presents correlations between predictor and criterion variables, as well as descriptive statistics. The perceiving, using, and understanding emotions subscales of the MSCEIT did not correlate significantly with criteria, and therefore we do not report any analyses with those subscales. Consistent with our hypotheses, the managing emotions subscale of the MSCEIT was positively correlated with participants’ self-reports of positive interaction with two friends (NRI positive interaction). It was also positively correlated with friends’ ratings of positive interaction, negatively correlated with friends’ ratings of negative interaction (NRI negative interaction), and positively correlated with friends’ ratings about participants’ tendency to provide emotional support. Emotional Intelligence 12 With regard to the Big Five, extraversion was positively correlated with higher selfreports and friends’ ratings of positive interaction with friends. Emotional stability and conscientiousness were negatively correlated with self-perceived conflict with friends. All the Big Five subscales except for neuroticism were positively correlated with self-reports of emotional support, but none was correlated with friends’ ratings of emotional support. Similarly, all the Big Five subscales except for extraversion were related to self-perceived conflict resolution skills, but only agreeableness was significantly correlated with friends’ ratings of conflict resolution skills. The Big Five traits and the self-report scale of emotional regulation did not correlate significantly with the managing emotions subscale of the MSCEIT. Self-reported emotional regulation correlated with self-reported conflict resolution skills but not with other criteria. Incremental Validity of MSCEIT Managing Emotions Considering that MSCEIT managing emotions scores were uncorrelated with the Big Five, we would expect that controlling statistically for Big Five traits would not undermine relationships between managing emotions and criteria. Multiple regression analyses confirmed this. We entered the Big Five first, using a forward-stepping strategy (i.e., entering independent variables only if they explained significant additional variance in the dependent variable). Managing emotions was entered last. In these analyses, the managing emotions subscale of the MSCEIT still explained significant variance in all four criteria for which significant zero-order correlations were found: self-perceived positive interaction with friends, and friends’ ratings of positive interaction, negative interaction, and emotional support. Managing emotions explained between 7% and 11% of additional variance in all four criteria. Entering gender, all of the Big Five, and managing emotions simultaneously in multiple regression models, managing emotions scores remained significant in the prediction of three out of four criteria (all but friends’ ratings of emotional support). These findings supported our second hypothesis, regarding the incremental validity of the MSCEIT. Further analyses revealed that controlling for socially desirable responding also did not undermine relationships between managing emotions and criteria. Emotional Intelligence 13 Discussion The findings support both of the hypotheses about the concurrent and incremental validity of the managing emotions subscale of the MSCEIT. This subscale was associated with higher self-perceived positive interaction with friends and friends’ reports of more positive interactions, less negative interactions, and higher emotional support. These associations remained statistically significant after removing variance explained by the Big Five personality traits. The present results replicate and extend previous findings on the positive relationships between emotional competencies and the self-perceived quality of relationships with friends (e.g., Lopes, Salovey, et al., 2003). To our knowledge, this is one of the first studies supporting the incremental validity of the MSCEIT using both selfand peer-ratings. The managing emotions subscale was unrelated to the Big Five, suggesting that this ability is distinct from personality traits. Managing emotions scores explained peer-rated criteria better than self-rated criteria, whereas the opposite was true for the Big Five. Managing emotions scores concurrently predicted 3 out of 4 peer-ratings but only 1 out of 4 self-ratings. For the Big Five, only 2 out of 20 correlations with friends’ ratings reached statistical significance, but 11 out of 20 correlations with self-ratings did. This may be due to conceptual overlap and common method variance between the Big Five and self-reported criteria. The pattern of findings also suggests that ability measures of emotional competencies may provide an important perspective for understanding social adaptation. Note that the managing emotions subscale also explained friends’ ratings of quality of social interaction better than a self-report measure of emotional regulation. The perceiving, using, and understanding subscales of the MSCEIT were not significantly correlated with criteria. This may reflect the fact that these abilities contribute more weakly or indirectly to the quality of social interactions, as hypothesized. Study 2 The previous study provided converging evidence that managing emotions competencies are related to the quality of interactions with friends. Study 2 sought to extend previous research linking emotional competencies and the quality of social interactions by examining everyday Emotional Intelligence 14 interactions with different types of people using an intensive, repeated measures design. To address concerns that cultural differences might limit the generalizability of previous research findings with the MSCEIT, we conducted the present study in Germany. Participants completed the MSCEIT, a measure of the Big Five, and maintained a social interaction diary for two weeks. These data also allowed us to examine relationships between EI and daily social interaction controlling for Big Five traits. Because participants completed measures of self-esteem, adult attachment, growth orientation, and self-presentational style for another study, we were also able to examine correlates of EI that have not been examined previously. The main criteria for this study were the descriptions of social interactions that participants provided over a period of two weeks using a variant of the Rochester Interaction Record (RIR; Wheeler & Nezlek, 1977). Participants’ reactions to everyday social encounters served as indicators of self-perceived quality of interactions. The balance between the impressions that participants wanted to convey and the impression they thought they had made served as an indicator of self-perceived success at impression management. We expected that emotional competencies, and in particular the ability to manage emotions, would be related to both the quality of social interactions and the capacity to balance goals and perceived achievement in the sphere of impression management. Using a social interaction diary allowed us to overcome the problems inherent in single assessments of sociality, including distortion due to memory loss and the influence of unusual events. The advantages of such techniques are discussed in detail by Reis and Gable (2000). Moreover, RIR-based studies have successfully examined relationships between social interaction and various constructs that may be conceptually related to EI, such as social skills (Nezlek, 2001), perceived risk in intimacy with others (Nezlek & Pilkington, 1994), and depression (Nezlek, Imbrie, & Shean, 1994). With regard to cultural differences, considerable theory and research suggest that there are both cultural universals and differences in the experience, expression, and interpretation of emotions (e.g., Ekman, 1972; Mesquita & Frijda, 1992; Scherer & Wallbott, 1994). For example, Emotional Intelligence 15 people recognize facial expressions of emotion at above chance levels across cultures, but they also have a significant advantage in decoding the expressions of individuals belonging to their own cultural group (Elfenbein & Ambady, 2002). Accordingly, we would expect that basic emotional competencies are important in all cultures, even if these competencies are applied and expressed somewhat differently. A measure of emotional competencies such as the MSCEIT was designed to tap into the more universal aspects of emotional information-processing so as to be relatively free of cultural bias. Accordingly, we expected that cultural differences would not undermine the predictive validity of the MSCEIT, even though this test may need to be adapted for use in different cultures. Hypotheses Two hypotheses guided this study: 1. Emotional intelligence will be positively related to people’s reactions to social interactions and to self-perceived success in impression management. This effect was presumed to be strongest for the managing emotions subscale of the MSCEIT, because emotional regulation is crucial for managing social interactions and is also likely to contribute to people’s capacity to balance expectations and actual experience in social interactions. 2. Emotional intelligence, and particularly the managing emotions subscale of the MSCEIT, will explain variance in these criteria over and above the variance accounted for by the Big Five personality traits. Method Participants One hundred and six undergraduate students enrolled at Chemnitz University of Technology in Germany participated in this study initially, receiving course credit for participation. Of the 103 participants who provided sufficient data, 86% were women and 98% were Caucasian. Participants’ ages ranged from 18 to 36 years, with a mean of 22.4 (SD = 3.2). Procedure In an introductory session, participants received oral instructions about the diary study and completed the MSCEIT, following guidelines for group administration outlined in the test Emotional Intelligence 16 manual (Mayer et al., 2002). Two experimenters were present to answer questions. Following this, participants completed a battery of personality and other self-report measures online. For the third part of the study, participants described their social interactions each day for two weeks using a web-based variant of the RIR (Wheeler & Nezlek, 1977). During the introductory session, participants were told that the diary study concerned patterns of social interaction, and that they would use a structured questionnaire to describe these interactions. They were asked to report every social interaction they had that lasted 10 minutes or longer. An interaction was defined as any encounter with one or more other people, in which the participants attended to one another and adjusted their behavior in response to one another. Examples were provided to clarify what was an interaction (e.g., a conversation) and what was not (e.g., sitting silently with another person watching TV). Participants were asked to describe only face-to-face interactions. Telephone and internet conversations were excluded because the self-presentation aspects of these interactions are different from those of face-to-face interactions. All instructions were available on the Internet, and participants were encouraged to contact the experimenters by email if they had any problems. Participants described whom they interacted with (using unique initials for each person) as well as the sex of each individual involved in the interaction, for up to three people. For interactions with more than three people, participants indicated how many men and women were present instead of recording individual initials. Participants also rated how they felt about the interaction, how much they wanted to make certain impressions on their interaction partners, and to what extent they had achieved these impressions. In response to the question, “How did you feel during the interaction?”, participants rated each interaction along nine dimensions: enjoyment, interest, intimacy, dominance, and feeling calm, safe, wanted, and respected (singleitem measures). To assess participants’ impression management goals, and the impression they thought they had made upon others, they responded to the questions “how did you want to be perceived?” and “how were you perceived?” Participants answered these questions in terms of seven dimensions: friendly, likable, competent, intelligent, interesting, honest, and attractive (also single-item measures). These dimensions reflect Jones and Pittman's (1982) selfEmotional Intelligence 17 presentation categories of ingratiation, self-promotion, and exemplification. Physical appearance was included because it is an important dimension of interaction among college students (e.g., Nezlek & Leary, 2002). The response categories were discussed until participants understood the definitions, forms, and procedure. They were asked to complete an interaction record as soon as possible after each interaction, or at least once a day. Paper forms were made available to all participants in case they had no access to the internet for a particular day. After completing the study, participants answered questions about how they had maintained the diary (e.g., regarding the accuracy of the diary, the reactivity of the procedure, and technical problems). Based on these answers and inspection of the dates and times participants provided data (recorded by the server), the data from 3 original participants were excluded from the analyses. Emotional Intelligence: The MSCEIT Emotional intelligence was measured using a German translation of the MSCEIT (Version 2.0; Mayer et al., 2002; Schütz, Hertel, & Schröder, 2002). This test was described in Study 1. Given that cultural differences may undermine the validity of some test items, MSCEIT scores may need to be based on slightly adjusted norms in different socio-cultural contexts. Accordingly, we computed scores based on 123 items with positive item-total correlations in the present sample. Therefore, we report raw scores on a scale from 0 to 1, rather than standardized scores with a mean of 100 and standard deviation of 15. Correlations between the 4 subscale scores based on these slightly modified scales and those calculated by the test publishers were all above .94. Other Individual Difference Measures To examine correlates of EI, participants completed the following measures, all of which were translated from the original English versions into German. All translations that were not published scales or previously validated by other authors were back-translated into English to ensure accuracy. 1. The BFI-44, a 44-item measure of the Big Five personality traits using a 1-5 response format (John & Srivastava, 1999; Lang, Lüdtke, & Asendorpf, 2001). Emotional Intelligence 18 2. A multidimensional self-esteem scale (Fleming & Courtney, 1984; Schütz & Sellin, 2003) with 26 items using a 1-7 response format, and three subscales: Self-Regard, Social Confidence, and School Abilities. 3. A scale measuring Self-Acceptance and Acceptance of Others with 24 items using a 1-5 response format (Berger, 1952; Bergemann & Johann, 1985). 4. A measure of adult attachment (Doll, Mentz, & Witte, 1995) based on the four-category model proposed by Bartholomew and Horowitz (1991): secure, preoccupied, dismissive, and fearful attachment. Using 7-point, single-item scales, participants indicate the extent to which prototypical descriptions of attachment styles correspond to their relationships with former and present romantic partners. 5. The Goal Orientation Inventory (Dykman, 1998), a 36-item measure using a 1-6 response format to assess two motivational factors: validationand growth-seeking. 6. The Self Presentational Style Inventory (Leary, 2000), an 80-item measure of selfpresentational motives and concerns, with a 1-5 response format. Using an adjective checklist, participants indicated how they wanted to be perceived with respect to five self-presentational motives/concerns suggested by Jones and Pittman (1982): ingratiation, self-promotion, exemplification, supplication, and intimidation. Respondents indicated: “To what degree do you want other people to see you as... (e.g., cheerful, intelligent, honest, cowardly, forceful ....)?” 7. A measure of self-deceptive enhancement (Paulhus, 1984), a 20-item measure using a 1-7 response format. Paulhus (1984) argued that self-deception and impression management are two components of socially desirable responding, and suggested that self-deceptive enhancement may be viewed as an aspect of self-definition associated with self-esteem, ego-resiliency, and perceived control. Results The research questions required four types of analyses: analyses of MSCEIT scores per se, correlations between the MSCEIT and other individual difference measures, analyses of relationships between MSCEIT scores and measures of social interaction, and analyses of these Emotional Intelligence 19 relationships controlling for individual differences in the Big Five personality traits. Descriptive statistics for trait-level measures are presented in Table 2. Analyses of MSCEIT Scores As can be seen from Table 2, the German version of the MSCEIT was reliable, although reliabilities were slightly lower than those reported for the normative sample (Mayer et al., 2002). MSCEIT subscales were positively correlated (rs ranging from .20 to .41, p < .05), as would be expected for an interrelated set of abilities, with one exception: the correlation between perceiving and managing emotions (r = .04, ns). To check whether cultural differences might seriously distort responses to the test and undermine the validity of the MSCEIT in a German context, we compared the response profile of the present sample to the response profile of the predominantly North American normative sample. Response profiles were defined as the percent of participants endorsing each of five response options for the 141 questions in the MSCEIT. Because the percent endorsing the fifth response for any item was dependent on the percent who endorsed the first four, only the first four responses for each item were used. This created a sample of 564 responses. The correlation between the two response profiles was r (564) = .90, reflecting very high agreement. We repeated this analysis for each subscale of the MSCEIT separately. The correlations between the response profiles for this sample and those of the normative sample were all above .86. Separate analyses for men and women found no differences in the similarity of the two sets of profiles. Correlations Between the MSCEIT and Trait Measures Correlations between the MSCEIT and other trait measures are presented in Table 3. Overall, the MSCEIT showed limited overlap with the Big Five and other trait measures, supporting previous findings regarding the discriminant validity of ability measures of emotional intelligence (Brackett & Mayer, in press; Brackett, Mayer, et al., in press; Lopes et al., 2003; Roberts et al., 2001). The managing emotions subscale of the MSCEIT was significantly (although modestly) correlated with the Big Five (except conscientiousness), all the self-esteem scales, acceptance of others, secure attachment, a growth-seeking goal orientation, and positive self-presentational strategies. All of these correlations were positive except for the correlation Emotional Intelligence 20 with neuroticism, as expected. The other subscales of the MSCEIT did not correlate significantly with the measures of individual differences included in this study, except for the associations between using and understanding emotions and self-regard, and those between perceiving and using emotions and positive self-presentational strategies. Emotional Intelligence and Daily Social Interaction The present data constituted a hierarchically nested data structure in that one set of observations (social interactions) was nested within another set of observations (persons). There is growing consensus that multilevel random coefficient models (MRCM), also called multilevel or hierarchical linear models, provide more accurate analyses of such data than ordinary least squares (OLS) analyses (e.g., Bryk & Raudenbush, 1992; Kreft & de Leeuw, 1998). Accordingly, relationships between EI and social interaction were analyzed using MRCM as implemented by the program HLM (Raudenbush, Bryk, Cheong, & Congdon, 2000). See Nezlek (2001, 2003) for detailed discussions about this type of analysis. The present data set comprised a two-level structure. Level 1 observations were social interactions, and level 2 observations were persons. Of the 106 participants who began the study, 99 provided sufficient data for multilevel analyses. These participants described a total of 4553 interactions (M = 46, SD = 17.2). Prior to analysis, all person-level variables were standardized. We followed a forward-stepping strategy (i.e., initial models contained a single predictor and additional predictors were added one at a time) because multilevel models estimate a large number of parameters and entering too many predictors can tax the carrying capacity of the data (e.g., Kreft & de Leeuw, 1998; Nezlek, 2001). All coefficients were modeled as random. Reactions to social interaction. The first set of analyses examined relationships between single-item measures of reactions to interactions (enjoyment, intimacy, feeling interested, important, wanted, respected, safe, calm, and dominant), as indicators of the perceived quality of social interactions, and scores on each of the four subscales of the MSCEIT. The level 1 model was an unconditional model – there were no level 1 predictors: