Developmental correlates of emotional intelligence: Temperament, family environment and childhood trauma

Abstract

This study examined temperament, environmental factors (family environment and childhood trauma), and the interaction between them as developmental correlates of ability and trait emotional intelligence (EI). Using 97 university students, correlational analyses revealed that temperament characteristics were related to trait but not ability EI. Components of family environment and childhood trauma were not significantly related to ability or trait EI. Multiple regression analyses confirmed the importance of temperament in predicting trait EI, but provided little support for the role of environmental factors or their interactions with temperament in predicting either type of EI.

Key words:

• development
• emotional intelligence
• intelligence
• personality

Emotional intelligence (EI) has been conceptualised as an emotion‐related cognitive ability involving the ability to perceive, use, understand, and regulate emotion (Mayer & Salovey, 1997). Others define EI as a constellation of emotion‐related self‐perceptions at the lower levels of personality hierarchies (Petrides, Pita, & Kokkinaki, 2007). These two perspectives have been termed ability and trait EI, respectively. The former occupies the cognitive domain and is assessed via objective performance measures, and the latter the personality domain, assessed via self‐report.

Much research focuses on the validity of the ability and trait EI constructs. Trait EI studies have focused largely on distinctiveness and incremental validity in relation to higher‐order personality traits (e.g., Petrides, Pérez‐González, & Furnham, 2007; Petrides et al., 2007); ability EI studies focus on the relationships with intelligence and problems with the validity of scoring procedures (e.g., Brody, 2004a; O'Sullivan & Ekman, 2004). Recently, trait and ability EI have been validly measured in middle and late childhood (Mavroveli, Petrides, Sangareau, & Furnham, 2009; Rivers, Brackett, & Salovey, 2008).

In addition to validity studies, it is important for EI theory to identify developmental correlates. If EI can be validly measured, it should be related to theoretically relevant developmental variables. Such research on developmental correlates is complicated by the delineation of EI as either ability or trait based; the divergent psychometric domains means there may be different developmental trajectories (e.g., intrinsic influences such as temperamental disposition for trait EI and extrinsic environmental influences determining the acquisition of skills for ability EI). Ability EI is further complicated by the suggestion that the four EI skills are hierarchical (Mayer et al., 1997). As suggested by Zeidner, Matthews, Roberts, and MacCann (2003), the literature suggests parallel development of these skills. It is thus important for developmental theory of EI to show that these multiple abilities share common precursors.

In debating the underlying causes of individual differences in EI, researchers (e.g., Zeidner et al., 2003) have drawn upon the plethora of research pertaining to children's emotional development (e.g., Fox, 1994). In Zeidner et al.'s (2003) multi‐level investment model of the development of EI, temperament, environmental factors, and the interaction between them act as developmental determinants. The succeeding review considers how these may be differentially implicated in the development of ability and trait EI.

Temperament

Temperament reflects individual differences in emotional reactivity and self‐regulation (Rothbart & Derryberry, 1981). It is determined to some extent by genetic factors (Saudino, 2005; Zeidner et al., 2003). Temperamental differences during childhood and adolescence shape the child's adaptation to the world and continue throughout adulthood (Rothbart, Ahadi, & Evans, 2000). Rothbart's model (Rothbart et al., 2000) identifies four dimensions of adult temperament: negative affect, extraversion/surgency, effortful control, and orienting sensitivity; the first three are similar to temperament characteristics found in children (Rothbart et al., 2000). This model thus accounts for aspects of emotional reactivity and cognitive (effortful) control. It follows that temperament has direct influences on emotional experience, expressivity, and regulation (e.g., see Rothbart, Ellis, & Posner, 2004; Zeidner et al., 2003). Specifically, negative affect influences developing aspects of EI involving effective coping with stress, extraversion in developing emotional expressivity, and effortful control—which involves suppression of inappropriate behaviour—influences emotion regulation (Zeidner et al., 2003).

Notably, temperament maps onto personality‐based models (Caspi, Roberts, & Shiner, 2005) and provides the basis for dispositions and orientations towards others (Rothbart et al., 2000). The fact that trait EI also involves self‐perceptions—which past studies show to be significantly associated with temperament (Klein, 1995)—also suggests a strong temperamental influence. Temperamental characteristics determine whether an individual meets environmental demands and thus develops positive self‐perceptions (Klein, 1995). Individual differences in the affective aspects of temperament may therefore move through intrinsic trajectories to more strongly predict trait rather than ability EI. To our knowledge no study has explored this; however, indirect support comes from studies where neuroticism and extraversion are moderately associated with trait, but not ability EI when investigated as higher‐order personality dimensions within the 5‐factor model (e.g., Brackett & Mayer, 2003; Saklofske, Austin, & Minski, 2003).

Environmental factors

Much empirical evidence suggests that the family environment and caregivers play a central role in the socialisation of emotional skills, via explicit instruction or observation and modelling (Denham, Zoller, & Couchoud, 1994; Field, 1994; Zeidner et al., 2003). For example, children reared in environments where parental conflict is rife, cohesion is low and anger is expressed, are more aggressive (e.g., Dadds & Powell, 1991; Fang et al., 2009), suggesting problems in the learning of adaptive emotion regulation strategies. Abusive environments are especially detrimental to the development of emotion perception, understanding, and regulation (e.g., Maughan & Cicchetti, 2002; Pollak, Cicchetti, Hornung, & Reed, 2000; Shipman & Zeman, 1999).

In contrast to trait EI then, which may be largely influenced by temperament (Zeidner et al., 2003), ability EI skills are likely to be learnt and acquired within the environment. Despite this, Goldenberg (2004) documented small positive relationships of 0.14 between ability EI managing emotions and psychological abuse, and between using emotions and sexual abuse in a community sample. These data are inconsistent with past literature on emotional development, and suggest that abusive environments are not detrimental to ability EI development. Inconsistent findings may result from some unmeasured third variable such as temperament, which may moderate the extent to which dysfunctional environments impact development. Further research will advance our understanding of these relationships.

Because trait EI is a lower‐order trait, it can be best understood as a biologically based basic tendency, and its long‐term development will be little determined by extrinsic paths (McCrae et al., 2000), such as the family environment. This notion is consistent with evidence of negligible effects of the shared environment (i.e., experiences within the family) on trait EI (Vernon, Petrides, Bratko, & Schermer, 2008). However, because the construct involves self‐perceptions, these views will be influenced directly by self‐beliefs, which will be affected by environmental influences and experience (Magnussun & Stattin, 2006). Trait EI is indeed related to unhealthy family environments, as measured by the perceived degree of family cohesion, expressiveness, and conflict (Devi & Rayal, 2004). However, relationships with abuse have been inconsistent: small significant negative associations with physical/psychological but not sexual abuse (Goldenberg, 2004), or non‐significant effects (Brown & Schutte, 2006). Inconsistencies may result from different assessments of abuse, or the impact of abuse may vary as a function of temperament (as noted below). Further research would help to clarify these inconsistent findings.

Interaction between temperament and environment

The impact of temperament on parents' socialisation efforts is emphasised in Zeidner et al.'s (2003) multi‐level investment model. Biologically based differences in the affective aspects of temperament interact with the environment to predict emotional and social competence. For example, children with poor self‐regulatory behaviour in the form of low effortful control may be more demanding, which in turn affects poor parental responses; this then socialises the emotional responses of the child. It is this interaction that may cause the development of emotionally intelligent or unintelligent behaviours/skills.

Rationale, aims and hypotheses

The overall aim is to extend understanding of the developmental correlates of ability and trait EI. We examine biologically and environmentally based correlates and the interaction between them to help clarify the nature of these relationships. Consistent with Zeidner et al.'s (2003) model, and as suggested by the literature review, we examine temperament as a factor that is partly biologically based and focus our definition and operationalisation of temperament on Rothbart's four‐dimensional model (Rothbart et al., 2000). Family environment (conflict, cohesiveness and expressiveness) and childhood trauma/abuse were selected as potentially influential environmental factors. The study has three aims:

1	To test the hypothesis that temperament is more strongly related to trait than ability EI.
2	To explore whether environmental factors (family environment and childhood trauma) are more strongly associated with ability than trait EI.
3	To explore significant interactions between temperamental and environmental factors.

METHOD

Participants

Participants were 97 (80 females and 17 males) university students from across university disciplines (mean age = 22.29 years; range = 18–47). Eighty‐eight per cent of participants reported themselves as ‘white’, 8% as ‘Asian’, and 4% as ‘other’. Psychology students received course credit for participation.

Measures

Schutte Emotional Intelligence Scale (SEIS; Schutte et al., 1998)

The SEIS is a 33‐item self‐report measure that has shown to be reliable and valid measure of trait EI. Participants respond using a 5‐point Likert scale, ranging from ‘strongly disagree’ to ‘strongly agree’. Some studies support a 4‐factor structure: optimism/mood regulation, appraisal of emotions, social skills, and utilisation of emotion (e.g., Saklofske et al., 2003), although others have failed to confirm this structure (e.g., Gignac, Palmer, Manocha, & Stough, 2005). Our research and analyses were therefore restricted to trait EI as a global construct (calculated using the mean of the 33 items), which also kept the number of analyses low. Internal consistency in this sample was α = 0.88.

Mayer‐Salovey‐Caruso Emotional Intelligence Test Version 2.0 (MSCEIT; Mayer, Salovey, & Caruso, 2002)

The MSCEIT contains 141 items measuring four emotional skills: (1) perceiving emotions, (2) using emotions to facilitating thought, (3) understanding emotions, and (4) managing emotions (Mayer et al., 2002). However, recent studies have questioned this structure (e.g., Keele & Bell, 2008; Rossen, Kranzler, & Algina, 2008), and so the MSCEIT global score was used. The online version was used and the data scored by Multi‐Health Systems, the test distributor using expert norms (Mayer et al., 2002). Split‐half reliability with Spearman–Brown correction (Mayer et al., 2002) was 0.91.

The Adult Temperament Questionnaire Short‐Form (ATQ‐S; Rothbart et al., 2000)

This is a 77‐item questionnaire that assesses Rothbart's four temperament dimensions: negative affect, extraversion, effortful control, and orienting sensitivity. Items are rated on a 7‐point Likert scale, ranging from ‘extremely untrue’ to ‘extremely true’. Studies support the 4‐factor structure (Rothbart et al., 2000). Internal consistencies were α = 0.82, 0.74, 0.76, and 0.71, respectively.

The Family Environment Scale (FES; Moos & Moos, 2002)

This questionnaire assesses individuals' perceived family environment characteristics and functioning. Only the family relationship dimension was assessed, which comprises 27 true–false items and three subscales: family cohesion, expressiveness, and conflict. A ‘yes/no’ response format and standardised scores were used. There is evidence of construct (factorial) validity (Moos & Moos, 2002). Internal consistencies were α = 0.76, 0.62, and 0.76, respectively.

Childhood Trauma Questionnaire (CTQ; Bernstein & Fink, 1998)

This self‐report measure assesses histories of childhood trauma using 25 items. Subscales include physical, sexual, and emotional abuse, and physical and emotional neglect. Items are rated on a 5‐point Likert scale, ranging from ‘never true’ to ‘very often true’. The five‐factor structure has been replicated in numerous studies (see Bernstein & Fink, 1998). Internal consistencies were α = 0.93, 0.93, 0.87, 0.77, and 0.88, respectively.

Procedure

Participants were recruited via lectures and completed the study during their own time. They received a questionnaire booklet containing paper‐based questionnaires and information for accessing the online MSCEIT. A unique identification code was used in place of the participant's name to maintain anonymity and allow paper‐based and online responses to be matched.

RESULTS

Data screening

Univariate distributions showed no major deviations from normality, with the exception of the five trauma subscales, which were positively skewed. Three subscales (physical abuse, sexual abuse, emotional neglect) also had extreme univariate outliers. To overcome non‐normality problems, the five subscales were dichotomised via median split. The sexual abuse subscale had an extremely uneven split (93.8–6.2%) and was removed.

Main analyses

Descriptive statistics and correlations among the developmental and EI variables are reported in Table-1. To avoid Type I error, an alpha of 0.01 was used for the correlational and interaction analyses (Cohen & Cohen, 1983). Given the statistical difficulty of detecting interaction effects, more stringent alpha levels would be inappropriate (McClelland & Judd, 1993). Trait EI had moderate correlations with three temperamental characteristics, but was not significantly related to family environment or trauma. There was a modest relationship between ability EI and physical abuse (r = −0.22), but this was non‐significant, as was the relationship between ability EI and all other developmental variables.

Table 1 Descriptive statistics and correlations (r) between trait EI, ability EI, temperament, family environment, and childhood trauma

	1	2	3	4	5	6	7	8	9	10	11	12	13
1. Trait EI	–
2. Ability EI	0.26***	–
Temperament
3. Negative affect	−0.18	0.07	–
4. Effortful control	0.40**	−0.02	−0.42**	–
5. Extraversion	0.43**	0.02	−0.40**	0.11	–
6. Orienting sensitivity	0.47**	−0.02	0.03	0.21	0.14	–
Abuse/neglect
7. Physical abuse	0.05	−0.22	−0.05	0.14	−0.02	0.23	–
8. Physical neglect	−0.06	−0.02	0.10	−0.08	−0.11	0.13	0.25	–
9. Emotional neglect	−0.02	−0.07	−0.03	0.04	−0.14	0.24	0.36**	0.42**	–
10. Emotional abuse	0.01	−0.10	−0.03	−0.02	−0.05	0.22	0.47**	0.56**	0.62**	–
Family environment
11. Conflict	−0.16	−0.05	0.10	−0.13	−0.05	0.14	0.31*	0.28*	0.40**	0.43**	–
12. Cohesiveness	0.02	−0.09	−0.04	0.06	0.10	−0.23	−0.40**	−0.37**	−0.52**	−0.48**	−0.61**	–
13. Expressiveness	0.13	−0.14	0.00	0.15	0.07	0.09	−0.04	0.01	−0.25	−0.17	−0.25	0.30	–
M	3.71	97.23	4.12	4.26	4.74	4.63	6.59	6.41	9.67	9.24	48.76	45.80	49.62
SD	0.37	14.55	0.70	0.68	0.71	0.68	4.17	2.64	4.24	4.51	15.75	13.74	12.38

* p < .01,

** p < .001.

*** p = .011.

Two hierarchical multiple regression analyses were performed to explore significant interaction effects. Global trait and ability EI were the criterion variables. Temperament was entered at Step 1, and environmental factors at Step 2 to examine the contributions of the environment while controlling temperament. Interaction terms were entered at Step 3. Prior to analysis, continuous predictors (family environmental factors) and moderators (temperament dimensions) were centered to minimise multicollinearity between first‐order conditional effects and associated interaction terms. Dichotomous trauma predictors were subjected to weighted effects coding (West, Aiken, & Krull, 1996) with ‘high’ trauma being arbitrarily chosen as the reference category (West et al., 1996). Thus, positive and negative B coefficients indicate that the low trauma group scored higher or lower than the weighted grand mean, respectively. Effects coding allowed us to interpret the first‐order effects of abuse as average effects as in anlysis of variance (West et al., 1996) After centering and coding, product terms were created that represent the interaction between moderators and predictors (e.g., negative affect × physical abuse).

Multicollinearity was inspected via pairwise correlations (r; see Table-1) and variation inflation factors (VIF). All predictor intercorrelations and VIFs were below the recommended 0.80 and 10, respectively (Kleinbaum, Kupper, & Muller, 1988; Ryan, 1997). One exception was the extraversion‐cohesiveness interaction (VIF = 11.39). Extraversion × conflict also had a VIF of 8.41 and a large correlation with extraversion × cohesiveness. Running analyses excluding the latter altered the contribution of the former interaction term, and also reduced the VIF of extraversion × conflict to 3.89. Thus, the extraversion–cohesiveness interaction was excluded from our analyses.

Table-2 shows the results, displaying unstandardised regression coefficients (B), adjusted R² (R²_adj) for Step 1 (temperament), adjusted R² change (ΔR²_adj) for Steps 2 (trauma and family environment) and 3 (interaction terms), and the overall significance of the model. Unstandardised regression coefficients are interpreted because standardised coefficients are not properly standardised for interaction terms (West et al., 1996). The overall model was significant for trait EI, and 41% variance was accounted for by temperament on Step 1; Steps 2 and 3 were non‐significant. For ability EI, the overall regression model was non‐significant (p > .01), although this approached significance, as did Step 3 of the model.

Table 2 Multiple hierarchical regression of trait EI and ability EI on temperament (Step 1), family environment, childhood trauma (Step 2), and interaction terms (Step 3)

Predictor	Trait EI	Ability EI
	B	B
Step 1
Negative affect	0.05	2.14
Effortful control	0.17	0.62
Extraversion	0.19	1.24
Orienting sensitivity	0.19	−0.78
Step 2
Physical abuse	0.01	2.54
Physical neglect	0.01	−0.42
Emotional neglect	0.01	1.11
Emotional abuse	−0.02	0.34
Conflict	−0.01	−0.16
Cohesiveness	0.00	−0.28
Expressiveness	0.00	−0.15
Step 3 ×
Negative affect × physical abuse	0.00	1.77
Negative affect × physical neglect	0.00	−2.28
Negative affect × emotional neglect	0.11	7.65***
Negative affect × emotional abuse	0.01	0.63
Negative affect × conflict	0.01	0.30
Negative affect × cohesiveness	−0.01	−0.78*
Negative affect × expressiveness	0.01	0.22
Effortful control × physical abuse	0.00	4.82
Effortful control × physical neglect	0.04	0.22
Effortful control × emotional neglect	0.01	11.62**
Effortful control × emotional abuse	−0.05	−3.11
Effortful control × conflict	0.01	0.28
Effortful control × cohesiveness	0.01	−0.46
Effortful control × expressiveness	0.00	−0.26
Extraversion × physical abuse	−0.06	−2.21
Extraversion × physical neglect	0.03	−1.62
Extraversion × emotional neglect	0.15	1.49
Extraversion × emotional abuse	−0.08	2.26
Extraversion × conflict	0.01	0.38
Extraversion × expressiveness	0.01	0.64*
Orienting sensitivity × physical abuse	0.01	−3.38
Orienting sensitivity × physical neglect	0.00	−10.88*
Orienting sensitivity × emotional neglect	0.00	1.94
Orienting sensitivity × emotional abuse	−0.06	5.06
Orienting sensitivity × conflict	0.00	0.09
Orienting sensitivity × cohesiveness	0.00	0.11
Orienting sensitivity × expressiveness	0.00	−0.22
R^2 adj Step 1	0.41**	−0.04
ΔR^2adj Step 2	−0.01	0.05
ΔR^2adj Step 3	0.06	0.24†
Overall R^2adj	0.46	0.26
Final ANOVA	F(38, 58) = 3.19**	F(38, 58) = 1.89‡

* p < .01,

** p < .001.

*** p = .011;

† p = .014;

‡ p = .012.

ANOVA = analysis of variance.

DISCUSSION

This study investigated relationships between temperament, family functioning, reports of child trauma, and emotional intelligence. As predicted, and consistent with previous theoretical work (Zeidner et al., 2003), bivariate analyses showed moderate correlations between three dimensions of temperament (extraversion, orienting sensitivity, and effortful control) and trait, but not ability EI. The small non‐significant relationship with negative affect suggests that negative emotional reactivity may not account for individual differences in trait EI. However, our analyses only generalise to global trait EI and do not address causality. Trait EI was also not significantly associated with the environmental variables, findings confirmed by the multiple regression analysis. Taken together, these data support the conceptualisation of trait EI as a personality variable, and suggest that it may be largely driven by intrinsic developmental paths. Findings are also consistent with recent behavioural genetics evidence that trait EI is largely heritable with little influence of the shared family environment (Vernon et al., 2008).

The final step of the ability EI regression model approached significance, suggesting that it would be fruitful to isolate a smaller, specific set of interactions for investigation in future studies. However, ability EI per se was not significantly related, either in the correlational or regression analyses, to any study variable. The low and patchy correlations question the contribution of trauma and family environment to the development of ability EI, at least at the global EI level. However, given that factors such as family cohesion and expressiveness are key components of the family system and important for emotional development (Bennun, 1988), this questions what ability EI is measuring. One explanation for the weak relationships is that ability EI, as operationalised through the MSCEIT, cannot be measured validly. The MSCEIT utilises consensus and expert scoring techniques that are inappropriate for the assessment of a cognitive ability and also fraught with various problems (Brody, 2004a; MacCann, Matthews, Zeidner, & Roberts, 2003). Other problems are also evident. For example, the test contains tasks that involve selecting appropriate responses for ill‐defined social situations that do not have correct answers, and also assesses knowledge of emotion management strategies rather than actual performance of these actions (Brody, 2004a, 2004b).

There are implications of this research. Findings suggest that trait EI is largely a biologically based tendency and not related to the environmental variables in this study. Thus, interventions to improve the quality of the family environment and interactions with caregivers may have little value. This latter conclusion also applies to ability EI, although ability EI may be more malleable. Given the evidence that both trait (Chang, 2006; Nelis, Quoidbach, Mikolajczak, & Hansenne, 2009; Qualter, Whiteley, Hutchinson, & Pope, 2007; Qualter, Whiteley, Morley, & Dudiak, 2009) and ability EI can be improved (Chang, 2006; Nelis et al., 2009), future research is needed to identify factors that contribute to their development and understand how such changes come about.

This study has limitations. It uses a small predominantly female student sample and the developmental trajectories of ability and trait EI remain uncertain given the cross‐sectional design. Analyses were also restricted to global trait EI,and the SEIS may yield weaker effects relative to more comprehensive measures of trait EI (e.g., Freudenthaler, Neubauer, Gabler, Scherl, & Rindermann, 2008). In addition to addressing these limitations, future studies should examine the importance of other environmental factors (e.g., child–caregiver attachment). They should also investigate the relevance of the non‐shared environment (cf. Vernon et al., 2008) and how this interacts with neural and genetic influences, which appear to be key developmental pathways to ability and trait EI, respectively (e.g., Jausovec & Jausovec, 2005; Vernon et al., 2008). Longitudinal designs will also help clarify whether EI development is a process of continuous change or whether it develops in a hierarchical fashion as a function of age and cognitive maturation.

CONCLUSIONS

This study expands work in the EI field by investigating correlates of ability and trait EI. Trait EI findings are consistent with the conceptualisation of a construct that is intrinsically determined. Childhood trauma and family environment factors appear to hold little importance for both types of EI. Analyses involving a smaller set of predictors and interaction terms should be the focus of future research.