Individual differences in music consumption are predicted by uses of music and age rather than emotional intelligence,neuroticism,extraversion or openness

Participants

In all, 535 participants (365 men, 170 women) took part in the current study. They were directed to the website for the study via different social networking sites (e.g., Facebook and MySpace), emails, and university-linked webportals; in addition, several banners advertising the study were placed at university campuses. Participants’ ages ranged from 18 to 64 years, with a mean of 24.2 (SD = 7.9); 2.1% described themselves as African/Black, 7.3% as Asian, 77.5% as European/Caucasian, 1.3% as Latino/Hispanic, 2.9% as Middle Eastern, 0.2% as Pacific Islanders, and 6.9% as ‘other’. With regards to educational level, 1.4% had completed up to high school or less, 9.7% were high school graduates, 24.9% had completed some college education, 21.8% were college graduates, 16.3% had some postgraduate education, 25.9% had completed postgraduate studies (e.g., MBA, PhD, etc.).

Measures

Uses of Music Inventory (Chamorro-Premuzic & Furnham, 2007). This 15-item scale was constructed to measure views about when and why one listens to music. All items were rated on a 5-point Likert-type scale (1 = Strongly disagree, 5 = Strongly agree), and five items were summed to compute scores for each of the inventory’s three components: Emotional use of music [M(emot); sample item: ‘Whenever I want to feel happy I listen to a happy song’]; cognitive use of music [M(cog); sample item: ‘Listening to music is an intellectual experience for me’], and; background use of music [M(back); sample item: ‘If I don’t listen to music while I’m doing something, I often get bored’]. Previous work has shown that the measure has high internal consistency and good internal reliability (Chamorro-Premuzic & Furnham, 2007; Chamorro-Premuzic, Swami et al., 2009, Chamorro-Premuzic, Gomà-i-Freixanet et al., 2009). Cronbach’s alpha coefficients for the subscales in the present study are reported in Table 1.

OPEN IN VIEWER

Table 1.

Descriptive statistics and inter-correlations for all target variables

	M (SD)	α	2	3	4	5	6	7	8	9	10	11
1 Music consumption	3.4 (.71)	.78	.06	.05	.13**	−.01	.52**	.29**	.41**	−.06	−.20**	−.02
Personality
2 Extraversion	9.8 (2.6)	.86		−.18**	.20**	.42**	.01	−.11**	.10*	−.06	.00	.09*
3 Neuroticism	9.1 (2.7)	.88			.02	−.52**	.01	.03	.20**	−.15**	−.07	−.01
4 Openness	11.8 (2.0)	.80				.25**	.07	.21**	.08	.11**	.08	.06
5 Trait EI	106.16 (12.9)	.86					−.12**	−.08	−.00	−.00	.13**	.10*
Uses of music
6 Background	3.01 (.95)	.68						.21**	.37**	.02	−.14**	−.09*
7 Cognitive	2.70 (.87)	.65							.20**	.20**	.02	.02
8 Emotional	3.40 (.83)	.67								−.14**	−.10*	.06
Demographics
9 Sex		–									.19**	.03
10 Age		–										.24**
11 Education		–

Note: N = 535; Sex coded 1 = Women, 2 = Men. * p < .05, ** p < .01.

Trait Emotional-Intelligence Questionnaire-Short Form (TEIQue-SF; Petrides & Furnham, 2006). The TEIQue-SF is a 30-item instrument designed to assess individuals’ emotional self-efficacy or ability to identify and manage their own and others’ emotions (e.g., ‘Others admire me for being relaxed’, ‘I often pause and think about my feelings’). It is based on the theory of trait EI, which regards the construct as a personality disposition at a higher order hierarchical level than the Big Five personality factors. All items were rated on a 7-point Likert-type scale (1 = Strongly disagree, 7 = Strongly agree). Excellent internal consistencies and broad coverage of the sampling domain of the construct have been reported in previous work (Petrides & Furnham, 2006). Cronbach’s alpha coefficient for the TEIQue-SF in the present study is reported in Table 1.

International Personality Item Pool (IPIP; Goldberg, 1999, 2001). The IPIP comprises 50 items assessing Extraversion (e.g., ‘I am the life of a party’), Agreeableness (e.g., ‘I feel little concern for others’), Conscientiousness (e.g., ‘I am always prepared’), Emotional Stability (low Neuroticism, e.g., ‘I get stressed out easily’), and Openness to Experience (or Intellect, ‘I have a rich vocabulary’) (Goldberg, 2001). In the present study only Openness, Neuroticism, and Extraversion were assessed for both time limitations and theoretical reasons – neither Agreeableness or Conscientiousness has been conceptually related to uses of music in past studies. Each item was rated on a 5-point Likert-type scale (1 = Very inaccurate, 5 = Very accurate). Cronbach’s α for the present sample (see Table 1) were in line with previously reported internal consistencies (which tend to average at.84). Goldberg (2001) reported strong correlations (corrected rs .85 —.92) between the IPIP factors and their equivalent factors as assessed by the NEO-FFI (Costa & McCrae, 1992), another widely-used measure of the Big Five.

Music consumption. This was a purpose-designed 10-item scale to assess the frequency with which individuals used music in actual terms. Items (see Appendix 1 for a full list) included statements about the frequency with which individuals bought and downloaded songs and music videos, as well as how often they attended music performances or watched music programmes on television. The selection of items was based on pilot interviews with students (N = 16, 50% female, age M = 22.2, SD = 6.3), from which an initial item pool of 50 items was generated by the researchers. Several bibliographic sources, notably North et al. (2000) and Rentfrow and Gosling (2003) were also consulted and, in a second stage of the pilot study, a different set of participants (N = 20, 50% female, age M = 24.3, SD = 7.3) was given the option to generate novel items not present in the provisionally constructed scale. Items were intended to assess individual differences in music consumption (mp3s, CDs, videos), and music-related hobbies or activities other than buying or listening to music alone (e.g., reading about musicians, playing an instrument, attending concerts). A principal components analysis identified a clear single factor underlying the 10 items. The single factor was labelled ‘music consumption’ and explained 40.3% of the variance. All items were rated on a 5-point Likert-type scale (1 = Very rarely, 3 = Sometimes, 5 = Very often). Cronbach’s alpha for this scale is reported in Table 1.

Statistical analyses

All statistical analyses were conducted in SPSS v. 16. We initially computed descriptive statistics and bivariate correlations between all variables of interest. However, bivariate correlations do not take into account the simultaneous effects of multiple predictors on different criteria. Consequently, SEM was carried out using AMOS 4.0 (Arbuckle & Wothke, 1999) in order to: (1) account for the overlap between different predictors; (2) account for the overlap between different criteria, and; (3) assess the validity of a hierarchical model in which the same factors are both predictors and criteria.

To assess the fit of the model, the following indices were used: χ² (Bollen, 1989), which tests whether an unconstrained model fits the covariance/correlation matrix as well as the given model (although non-significant χ² values indicate good fit, well-fitting models often have significant χ² values); the parsimony goodness-of-fit indicator (PGFI; Mulaik et al., 1989), which measures power and is optimal around .50; the CFI (Bentler, 1990), which compares the hypothesized model with a model based on zero-correlations among all variables (values around .90 indicate very good fit); for the root-mean-square error of approximation (RMSEA; Browne & Cudeck, 1993), values <.08 indicate good fit; Akaike’s information criterion (AIC; Akaike, 1973) provides an estimate of the extent to which the parameter estimates from the original sample will cross-validate in future samples, and; Hoelter’s critical N (CN; Hoelter, 1983) provides the maximum sample size for which a model with same sample size and df would be acceptable at the .01 level.

Descriptive statistics and bivariate correlations

Descriptive statistics (Ms and SDs) and internal consistencies (Cronbach’s α) for all composite measures are reported in Table 1. The Uses of Music Inventory and trait EI scores and consistency values were in line with previous studies (Chamorro-Premuzic & Furnham, 2007; Chamorro-Premuzic, Swami et al., 2009, Chamorro-Premuzic, Gomà-i-Freixanet et al., 2009). Due to the nature of the IPIP, norms are not available for this personality measure. However, as shown in Table 1, all αs were acceptable at least, and mostly high.

Table 1 also reports the inter-correlations among target measures. As seen, music consumption was substantially correlated with background use of music and moderately correlated with the other two music uses. There were also modest but significant correlations between music consumption and Openness (positively) and age (negatively). As hypothesized, Openness correlated positively with cognitive use of music, whereas Neuroticism correlated positively with emotional use of music. The predictions of a positive link between Extraversion and background use of music, and trait EI and emotional use of music, were not supported. Additional significant correlations – not predicted – between personality and uses of music were the positive link between Extraversion and emotional use of music and the negative link between Extraversion and cognitive use of music. With regard to demographic factors, men were more likely to use music in a cognitive way, whereas women were more likely to use music in an emotional way. Age and educational level were both negatively related to background use of music.

Structural Equation Modelling (SEM)

Next, SEM was applied to the data. Missing values (<5% per variable, in line with Tabachnik & Fidell, 2007) were first replaced with the series mean at the item level – prior to computing factor scores. Prior to analysis, variables (factor scores) were standardized across the whole sample to a mean of zero and unit variance. We used standardized variables in our analysis because personality and music preference scores lie on different scales, as suggested by Loehlin (1992). Using SEM, we treated participant demographics (age, sex, and educational level), the Big Five personality factors, and trait EI as exogenous variables; the uses of music factors were modelled as mediators, and music consumption was treated as the criterion or exogenous variable. The hypothesized model included paths from Openness to cognitive use of music, from Neuroticism, trait EI, and sex to emotional music use, from sex to cognitive music use, and from age and Extraversion to background music use. In addition, effects of the three uses of music factors on music consumption were hypothesized, but direct effects of the exogenous variables on music consumption were not allowed. The point of not allowing direct paths from the exogenous variables to music consumption was to test these effects via inspection of the modification indices; specifically, any improvements in fit suggested by modification indices indicating that direct effects of either personality (including trait EI) or demographic variables were necessary to improve model fit. It should however be noted that (as per Table 1) the only exogenous variables that were significantly correlated with music consumption were age and Openness (therefore, these would be the only mediated effects). In line with Feingold (1994), we allowed sex to covariate with the three Big Five traits, whilst in line with Petrides and Furnham (2006), we allowed sex to correlate with trait EI. Finally, inter-correlations among the Big Five were allowed in line with Chamorro-Premuzic (2007), Rushton and Irwing (2008), whilst inter-correlations among uses of music factors were allowed in accordance with Chamorro-Premuzic and Furnham (2007). Age and educational level were also allowed to covary, whereas (in line with Table 1) sex was allowed to affect both cognitive and emotional use of music.

The hypothesized model did not fit the data well: χ² (df = 35, N = 535) = 147.30, p < .01 CFI = .86, GFI = .94, AGFI = .91, PGFI = .52, RMSEA = .08 (.07 –.09). In line with modification indices (statistical indicators for model fit improvement) and Silvia and MacCallum’s (1988) recommendations, theoretically meaningful parameters were added – in decreasing order – until satisfactory fit could be achieved. These included correlations between sex and age (a sampling issue), and paths from trait EI, educational level and Openness to background use of music, and from Extraversion to cognitive use of music; in addition, a path from age to music consumption was also added.

The modified model (see Figure 1) explained the data well: χ² (df = 28, N = 535) = 44.31, p > .01 CFI = .98, GFI = .97, AGFI = .98, PGFI = .44, RMSEA = .03 (.01–.04). As shown in the model in Figure 1, there were significant effects of all uses music on music consumption, which was also affected by age. In combination, these predictors accounted for 36% of the variance in music consumption (most of this variance was explained by uses of music rather than age). In regards to uses of music, background use was affected by all exogenous variables except sex and Extraversion, cognitive use of music was affected by Openness, Extraversion and sex, whereas emotional use of music was affected by sex, trait emotional intelligence, and Neuroticism.

OPEN IN VIEWER

Figure 1.

Modified model of predictors of uses of music and music consumption.

Individual differences and uses of music

First, our results support previous findings in showing that Neuroticism positively predicted emotional use of music, an association that has been explained in terms of the higher emotional sensitivity of neurotics to music in relation to their emotionally stable counterparts (Chamorro-Premuzic & Furnham, 2007). In addition, we also replicated previous work showing that Openness positively predicted cognitive use of music. As discussed elsewhere (Chamorro-Premuzic & Furnham, 2007; Rentfrow & Gosling, 2003), open individuals have an interest in more sophisticated forms of music. This result is also consistent with the finding that cognitive use of music is positively associated with self-assessed intelligence (Chamorro-Premuzic, Gomà-i-Freixanet et al., 2009), which is explicable in terms of the positive correlation between Openness and intelligence scores (Ackerman & Heggestad, 1997).

Also supporting previous work with a British sample (Chamorro-Premuzic & Furnham, 2007), we found that Extraversion was negatively linked with cognitive use of music. However, this association was not found in other studies. Moreover, the current results did not replicate with the previously found positive associations between Extraversion and emotional use of music among Spanish and Malaysian samples (Chamorro-Premuzic, Swami et al., 2009, Chamorro-Premuzic, Gomà-i-Freixanet et al., 2009). Additionally, the present study failed to find a significant association between Extraversion and background use of music, which is consistent with previous work among British (Chamorro-Premuzic & Furnham, 2007), but not cross-cultural samples (Chamorro-Premuzic, Swami et al., 2009, Chamorro-Premuzic, Gomà-i-Freixanet et al., 2009). Taken together, these results suggest that a more in-depth examination of the influence of Extraversion on uses of music may be warranted, as there have been few consistent associations found in this respect.

Another noteworthy aspect of our study was the lack of a significant correlation between trait EI and emotional use of music. Previous work has shown that trait EI was positively correlated with emotional use of music (Chamorro-Premuzic et al., in press), but that its effects on uses of music was negligible once other personality traits had been taken into account. Taking these results together, it is possible to conclude that trait EI does not have an incremental validity over the Big Five in relation to explaining the uses of music. This is probably due to the strong inter-relation between trait EI and the Big Five personality factors of Extraversion, Neuroticism, and Openness (Chamorro-Premuzic & Bennett, 2007), all of which were included in our model. Indeed, the significant effects of trait EI on emotional use of music shown in Figure 1 may be regarded as ‘suppressor’ effect (given that trait EI was not significantly correlated with this variable – as shown in Table 1 – but became a significant predictor once its overlap with other Big Five factors was accounted for). However, the negative association between trait EI and emotional use of music may be more than a mere statistical ‘artefact’ and warrants further investigation. It is possible that the ambiguous nature of the emotional use of music factor – which, as said, comprises items referring to both positive and negative emotional experiences – may partly account for the inconsistent associations found here with trait EI. This suggests it may be beneficial to break down the emotional use of music factor into two components, one referring to the use of music for negative affect regulation (which may be hypothesized to be negatively correlated with trait EI) and one referring to use of music for positive affect refulation (which may be hypothesised to be positively correlated with trait EI). Clearly, this would require an expansion of the uses of music inventory (with new items added).

On the other hand, trait EI was found to have negative effects on background music use, an association which was not hypothesized. Although further replications are needed, this finding is counter-intuitive because trait EI is positively correlated with Extraversions, and extraverts have been found to be more likely to use music as background to other activities (though not in the present study). That said, the fact that the effects of trait EI on background use of music were found even when individual differences in Extraversion were accounted for suggests that there are other underlying causes to the significant correlation between trait EI and background use of music (furthermore, in the present study Extraversion had no significant effects on background use of music). This begs the question of why individuals with lower trait EI would be more likely to listen to music as background to other activities. Although this may require further studies, particularly some which look at the different aspects or dimensions of trait EI, it is possible that music as background to other activities could help individuals combat boredom and other negative thoughts (e.g., worry, lack of meaning or intrinsic motivation in everyday activities and routine, etc.) which can be expected to be associated with low rather than high trait EI.

Our results also showed that participant demographics were significant predictors of certain uses of music. First, men were more likely than women to use music for cognitive or intellectual reasons. Recent work (Chamorro-Premuzic et al., in press) has explained this association both in terms men’s greater cognitive musical skills, though see Furnham and Bunclark (2006) for a different account. In terms of self-presentational styles, where men may report greater use of music for cognitive reasons to create a gender-congruent external image for others (cf. Rentfrow & Gosling, 2003). In contrast, women were more likely to use music for emotional reasons, which is consistent with previous work showing that men and women differ with regard to their perception of emotional expression in music (e.g., Kamenetsky, Hill, & Trehub, 1997; but see Lundqvist, Carlsson, Hilmersson, & Juslin, 2009).

In the current study, we also found that older participants were less likely to use music for background purposes. One possible explanation for this finding is that music is typically more important in the lives of younger people, particularly adolescents (e.g., Arnett, 1991; Christenson & Roberts, 1988; Sikkema, 2005; ter Bogt et al., 2003). As postulated by Chamorro-Premuzic et al. (in press), increasing age may also be associated with a decline in the appreciation of music, which may result in greater distraction experienced by older individuals in the presence of background music. Finally, we also found that education level was negatively associated with use of music for background situations, independently of age. The fact that less educated participants were more likely to use music as background to other activities may highlight the less intellectual, cognitive and creative nature of these participants’ everyday jobs or occupations, which may be more suited to distraction and even benefit from the arousing properties of music if their jobs are somewhat more monotonous, physical, or non-cerebral (Chamorro-Premuzic, 2007). However, this interpretation will remain speculative until further evidence is provided.

Music consumption

An important independent contribution of the current study to the extant literature was the finding that most individual difference variables, with the exception of age, did not affect music consumption. On the other hand, our results showed that all three uses of music factors were significantly and positively correlated with actual use of music, explaining a good deal of variance in this outcome even when other predictors, such as age, are taken into account. This finding is important because it suggests that the different reasons that music is used for are stronger predictors of music consumption than dispositional and even demographic individual difference variables.

These results may have important implications for researchers seeking to understand actual music-related behaviour. Specifically, our results suggest that, in order to understand music consumption, it is more important to focus on the reasons why individuals use music rather than individual difference factors such as personality or demographics (other than age). Indeed, although uses of music are affected by demographic and less proximal/more dispositional individual differences, such as personality traits, they do not account for any dispositional or demographic effects on music consumption because such variables have trivial and non-significant effects on that outcome. Of course, it is important to note that, in our study, music consumption was operationalized using self-reports, which may be affected by socially desirable responses or self-presentation and shares method variance with the predictors examined. Future work could extend the current study by utilising more objective measures of music consumption, such as the weekly number of music downloads or concert attendance, though this information is harder to obtain, particularly for large samples.