The Journal of Genetic Psychology Research and Theory on Human Development ISSN: 0022-1325 (Print) 1940-0896 (Online) Journal homepage: http://www.tandfonline.com/loi/vgnt20 The General Factor of Personality and Character: A Reanalysis Curtis S. Dunkel & Dimitri van der Linden To cite this article: Curtis S. Dunkel & Dimitri van der Linden (2017): The General Factor of Personality and Character: A Reanalysis, The Journal of Genetic Psychology, DOI: 10.1080/00221325.2017.1381070 To link to this article: http://dx.doi.org/10.1080/00221325.2017.1381070 Published online: 26 Oct 2017. Submit your article to this journal View related articles View Crossmark data Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=vgnt20 Download by: [University of Florida] Date: 28 October 2017, At: 07:29 THE JOURNAL OF GENETIC PSYCHOLOGY https://doi.org/./.. The General Factor of Personality and Character: A Reanalysis Curtis S. Dunkela and Dimitri van der Lindenb Downloaded by [University of Florida] at 07:29 28 October 2017 a Department of Psychology, Western Illinois University, Macomb, Illinois, USA; b Erasmus University, Rotterdam, The Netherlands ABSTRACT ARTICLE HISTORY Using data from the Texas Twin Project, it was recently reported that 7 measures of character covaried to the extent that they formed a general factor of character (Tucker-Drob, Briley, Engelhardt, Mann, & Harden, 2016). In turn the relationship between the general factor of character and the Big Five personality traits were examined. It was found that personality was associated with the general factor of character primarily through the traits of conscientiousness and openness. For several reasons we propose that a more accurate interpretation of the data is that a Big Five personality traits form a general factor of personality, and that the relationship between the general character factor and personality is primarily through the general factor of personality. The results lend some support to this contention and are discussed in relation to the growing interest in covariation among multiple personality traits. Received April Accepted September KEYWORDS Character; general factor of personality; grit; intelligence mindset Grit, defined as “perseverance and passion for long-term goals” (Duckworth, Peterson, Matthews, & Kelly, 2007, p. 1087), has recently received a great deal of consideration as an explanation for individual differences in achievement, offering the possibility that interventions to impart grit can, in turn, increase achievement. However, in tandem with the increased focus on the potential of grit has been the rise in skepticism as to its unique benefits and its malleability (e.g., Credé, Tynan, & Harms, 2016). In a recent study, Tucker-Drob, Briley, Engelhardt, Mann, and Harden (2016) examined grit in relation to six other measures of character, all thought to be predictive of academic achievement, and the Big Five personality traits (openness, conscientiousness, extraversion, agreeableness, and neuroticism). To do so they used data from the Texas Twin Project—a sample of third- through eighth-grade twins and triplets. Tucker-Drob et al. found that the seven measures of character, including grit, exhibited sizeable intercorrelations. Using both exploratory and confirmatory factor analysis, Tucker-Drob et al. showed that these intercorrelations yielded a substantial proportion of shared variance in the scales which led to what Tucker-Drob et al. referred to as a general factor of character. In examining the associations of the general character factor with the Big Five, Tucker-Drob et al. (2016) focused primarily on the traits of openness and conscientiousness. When adding openness and conscientiousness to the seven measures of character and rerunning factor analysis, two factors emerged. One factor included a strong loading for conscientiousness and positive loadings on all seven character measures. The second factor included a strong loading for openness and positive cross-loadings for the majority of character measures. Tucker-Drob et al. concluded that the seven separate character measures are best represented by a single character factor and this general character factor’s relationship with personality is such that a mixture of conscientiousness and openness (perhaps together also indicating the ability to self-regulate) best account for variance in the character factor. CONTACT Curtis S. Dunkel IL , USA. © Taylor & Francis firstname.lastname@example.org Department of Psychology, Western Illinois University, Waggoner Hall, Macomb, 2 C. S. DUNKEL AND D. VAN DER LINDEN Downloaded by [University of Florida] at 07:29 28 October 2017 While we concur with the conclusions of Tucker-Drob et al. (2016) that character measures are intercorrelated due to a general factor, in the present study we wish to emphasize and test an alternative possibility. Specifically, we suggest that the general character factor as described by Tucker-Drob et al. may be strongly related and possibly nearly identical to the so-called general factor of personality (GFP) instead of only being related to a mix of conscientiousness and openness. Although, the GFP was not referred to in the article by Tucker-Drob et al. there currently is a large amount of literature showing that personality traits consistently exhibit intercorrelations leading to a general factor. The concept of the GFP traces back to Galton (1884) and, indeed, Galton’s writing on the topic seem prescient of the Tucker-Drob et al. findings: We must guard ourselves against supposing that the moral faculties which we distinguish by different names as courage, sociability, niggardness, are separate entities. On the contrary, they are so intermixed that they are never singly in action. I tried to gain an idea of the more conspicuous aspects in the character by counting in an appropriate dictionary the words used to express them. Roget’s Thesaurus was selected for that purpose, and I selected many pages of its index here and there as samples of the whole, and estimated that it contained fully one thousand words expressive of character, each of which has a separate shade of meaning, while each shares a large part of its meaning with some of the rest. (p. 181) Yet, only recently has interest in the GFP rematerialized. This new interest in the GFP began with Figueredo, Vásquez, Brumbach, and Schneider (2004) and Musek’s (2007) analyses of Big Five trait scores. Figueredo et al. and Musek each found that Big Five trait scores covary to form a GFP. Shortly thereafter, Van der Linden, Te Nijenhuis, and Bakker (2010) solidified the finding that a GFP can be extracted from Big Five trait scores in a large meta-analysis. Likewise, echoing Galton’s (1884) speculation about the importance of the general factor in contrast to single traits, they also found that the GFP accounted for more variance in criterion variables than the individual trait variance. In a recent review of the quickly expanding GFP literature, Van der Linden, Dunkel, and Petrides (2016) concluded that the GFP is likely substantive and relates to the tendency and ability to regulate social behavior. Thus, we considered it likely that the GFP has statistical and conceptual overlap with the character general factor as reported by Tucker-Drob et al. (2016). Given the concept of parsimony in science, testing the overlap between those two constructs is a useful endeavor providing insight into the unity and diversity of various character and personality trait measures. Our hypothesis that the GFP may provide a better explanation for a common dimension in character is also supported by the following empirical findings. First, in our work on the GFP we have utilized several public datasets, including data from Pozzebon et al. (2013), which includes both a measure of the Big Five personality traits as well as measures of grit. Creating a GFP using the weights from Van der Linden et al. (2010) and correlating it with grit resulted in a correlation of r(1336) = .65, p < .001 between the two variables. Second, and more specifically, there also were several direct indicators in the Tucker-Drob et al. (2016) article suggesting that the presence and importance of the GFP. While the intercorrelations between the seven character measures ranged from r = .12 to r = .49, the absolute intercorrelations between the Big Five traits ranged from r = .10 to r = .50. In fact the mean intercorrelation for the character measures was r = .24 while the mean intercorrelation for the Big Five was r = .26 and when controlling for acquiescence the mean intercorrelation for the character measures was r = .24 and for the Big Five was r = .27. The average correlation between character measures and the Big Five traits was r = .20. As suggested by a reviewer, dividing the average correlation between the character measures and the Big Five by the average within-set correlation yields an estimated correlation between the general factor of character and the GFP of r = .82. Additionally, in examining the correlations of the seven character measures with the Big Five, each of the resulting 35 correlations is in the predicted direction; positively correlated with openness, conscientiousness, extraversion, and agreeableness and negatively correlated with neuroticism. Of the 35 correlations, 33 were significant at p < .05. Last, although two factors emerged when including openness and conscientiousness along with the seven character measures in exploratory factor analyses the two resulting factors still correlated at r = .44 indicating the presence of an even higher-order factor. THE JOURNAL OF GENETIC PSYCHOLOGY 3 Downloaded by [University of Florida] at 07:29 28 October 2017 Finally, although, in their original article Tucker-Drob et al. (2016) included a model that tested whether the associations between the individual character measures and the Big Five were entirely mediated through a general factor (Model 4 in Table 6, p. 798) this cannot be considered a reliable test of the overlap between the GFP and the general character factor. Their model that included an overall general factor showed relatively poor fit, which may not be unexpected given that the strict assumption was adopted that the individual character and personality measures could not have relevant unique variance or covariance. For empirical as well as theoretical reasons this assumption is not plausible (Van der Linden et al., 2016; 2017). Therefore, in the present study we test (a) whether a viable GFP is present in the personality data reported by Tucker-Drob et al., (b) whether a GFP in that data strongly overlaps or even converges with the general factor of character they reported, and (c) whether the assumption of a GFP fits the data better than a model in which character only relates to the shared variance of openness and conscientiousness, instead of to a GFP. To this end the correlation matrices presented in Tucker-Drob et al. (2016) were submitted to a series of exploratory and confirmatory factor analyses to test these alternative hypotheses. Method Tucker-Drob et al. (2016) described the sample as being ethnically and financially diverse and including 811 twins and triplets from 7 to 15 years old. The measures used in the reanalyses included self-report and rating scales of character and personality. The seven character measures include grit (Duckworth et al., 2007), need for cognition, intellectual self-concept, mastery orientation, educational attitudes, incremental mindset, and test motivation. The Big Five personality traits of openness, conscientiousness, extraversion, agreeableness, and neuroticism were measured using the child version of the Big Five Inventory. Results Extracting the first unrotated factor from an exploratory factor analysis (EFA) using principal axis factoring the Big Five dimensions yielded a general factor that had similar characteristic as the GFP reported in many previous studies. That is, the GFP comprised 29.38% shared variance among the lower-order traits and loaded on each of the Big Five dimensions in the expected direction, namely, .40, .66, .25, .76, and –.50, for openness, conscientiousness, extraversion, agreeableness, and neuroticism, respectively. Testing the GFP using confirmatory factor analysis (CFA) led to similar results as the EFA in the sense that the general factor loaded in the expected direction on each of the Big Five dimensions (.39, .65, .21, .77, and –.51, for openness, conscientiousness, extraversion, agreeableness, and neuroticism, respectively). In the initial model, the fit was not satisfactory (root mean square error of approximation [RMSEA] = .13, comparative fit index [CFI] = .89, Tucker-Lewis index [TLI] = .77); however, by allowing one correlation between the unique variances of openness and extraversion (r = .25; see also DeYoung, Quilty, & Peterson, 2007), the fit of the model became acceptable (RMSEA = .07, CFI = .97, TLI = .93). Note that adding a correlation between the unique variance influences the model fit but does not affect the structural model, as the factor loadings remain nearly identical. Given that the EFA and CFA clearly indicated the presence of the GFP, the next step was to test how the GFP in that dataset correlated to the general factor of character as reported by Tucker-Drob et al. (2016). To do so, we added the seven character measures as indicators of a general factor of character. This general factor was allowed to correlate to the GFP. For fair model comparisons, in the alternative Tucker-Drob et al. (2016) models we described we allowed exactly the same correlations between errors as in the target model. Although the correlation between the GFP and the general character factor was strong, the overall fit of this model was low, χ 2 (52) = 504.82, RMSEA = .08, CFI = .79, TLI = .74. Nevertheless, the fit improved considerably after allowing two additional correlations between the error variance of openness with grit and conscientiousness with intellectual self-concept. In this model, the correlation Downloaded by [University of Florida] at 07:29 28 October 2017 4 C. S. DUNKEL AND D. VAN DER LINDEN between the GFP and the general character factor was r = .87, confirming their strong overlap. The fit indices of the model were, χ 2 (50) = 259.41, RMSEA = .07, CFI = .90, TLI = .88. Although, still not reaching conventional values for acceptable fit, it is imperative, however, to note that this model that included a GFP, fit the data better than alternative models as reported in Tucker-Drob et al. (2016). For example, a model in which the general character model only correlated to the shared variance of openness and conscientiousness –assuming the remaining Big Five dimensions were unrelated to it- did not converge because it was not viable to model them as separate factors. Moreover, a model in which the general character factor loaded on the seven character measures and openness and conscientiousness – assuming unrelated remaining Big Five dimensions- showed a poorer fit, χ 2 (54) = 834.30, RMSEA = .13, CFI = .64, TLI = .56, compared with the models assuming a GFP as described previously. Using the chi-square test, the difference between the GFP model and the model with openness-conscientiousness loadings was significant, χ 2 = 574.9, df = 4, p < .0001, in favor of the former. Thus, confirming that models not including a GFP are inferior to ones that do so. As the GFP obviously also includes openness and conscientiousness, a question that may arise is whether the GFP-character association may be largely driven by these two traits being present in the GFP. Therefore, we conducted additional analyses with a GFP that excluded openness and conscientiousness. This model showed very similar fit values as the model that included the full GFP (with all Big Five dimensions) and were, χ 2 (34) = 149.21, RMSEA = .07, CFI = .91, TLI = .88. The GFP-general character correlation was somewhat lower compared with the full-GFP model, namely r = .67, yet still very substantial. These results show that the GFP-character factor association is not fully driven by the contributions of openness and conscientiousness to the GFP. Given the strong overlap between the GFP and the general character factor, a subsequent analysis we deemed logical and relevant was to test whether all personality and character variables would have a latent factor in common. To test this, we constructed a model in which a latent factor direct loaded on all 12 measures of personality and character. Similar to the previous models (and to the approach adopted by Tucker-Drob et al., 2016) we allowed for four correlated errors that capture additional relations between the remaining (unique) variances of the measures. In this model, the latent factor substantially loaded on each of the variables in the expected direction. For example, the loadings on the character measures were similar to those reported by Tucker-Drob et al. (2016) and ranged from .25 (test motivation) to .71 (need for cognition). The average factor loading was .49. The Big Five loadings ranged from .28 (extraversion) to .60 (conscientiousness), the loading of neuroticism was negative (–.33). The average (absolute) loading was .56. The fit of the model was acceptable, χ 2 (50) = 214.35, RMSEA = .06, CFI = .93, TLI = .90. In fact, compared with the model in which the two latent factors (GFP and character factor) correlated (see previous), the latter model with one general latent factor above all variables showed a significant better fit, χ 2 = 45.7, df = 0, p < .0001. Although the initial correlational analyses reported in the beginning of the Results section already revealed that controlling for acquiescence did not strongly affect the results, we also tested all CFA and structural equation models described previously, including acquiescence as an (observed) control variable. This had very little effect on the outcome (only some negligible changes in factor loadings) and led to the same conclusions. Details of these analyses can be provided upon request. Discussion Our premise is that independent of the label, the general factor of character, or GFP, a significant degree of unity can be found amongst many indices of personality, a pattern that has also recently been recognized as occurring across psychological disorders (Caspi et al., 2014) suggesting further unity. Tucker-Drob et al. (2016) uncovered a general factor of character and they assumed its relationship with personality is such that it is primarily a function of conscientiousness and openness. The alternative premise we introduce differs little from that of Galton (1884), as we hypothesized a GFP would emerge from the personality trait correlation matrix and that the GFP would be strongly associated with Tucker-Drob et al. general factor of character. To that end we used the correlation matrix presented in Tucker-Drob THE JOURNAL OF GENETIC PSYCHOLOGY 5 Downloaded by [University of Florida] at 07:29 28 October 2017 et al. to reanalyze the data to test for a GFP and examine the strength of its relationship with the general factor of character. We found that the correlation between the GFP and the character factor was no less than r = .87, which in social science often is considered a very good value for test–retest reliability of a single construct or measure. Moreover, we found that a model in which a single latent factor is on top of all personality and character measures showed an even better fit than one in which the two separate factors (GFP and character factor) correlated. This finding is in line with the idea that the GFP is a general factor in the noncognitive domain, representing socially valued aspects of personality as well as character (Galton, 1884). Buttressing the this idea, the results of a recent meta-analysis on the relationship between a GFP based on Big Five trait scores and trait emotional intelligence exhibited a similar level of overlap, r .85 (Van der Linden et al., 2017). At a minimum the findings support the notion stated in the hypotheses that there is a GFP and it is strongly associated with the general character factor. However, although these general factors appear largely equivalent, the results do not preclude a role for further associations among neither lower-level traits nor the possible role of lower-level traits accounting for variance in outcome variables abilities. References Caspi, A., Houts, R. M., Belsky, D. W., Goldman-Mellor, S. J., Harrington, H., Israel, S., …, & Moffitt, T. E. (2014). The p factor one general psychopathology factor in the structure of psychiatric disorders? Clinical Psychological Science, 2, 119–137. doi:10.1177/2167702613497473. Credé, M., Tynan, M. C., & Harms, P. D. (2016). Much ado about grit: A meta-analytic synthesis of the grit literature. Journal of Personality and Social Psychology, 113, 492–511. DeYoung, C. G., Quilty, L. C., & Peterson, J. B. (2007). Between facets and domains: 10 aspect of the Big Five. Journal of Personality and Social Psychology, 93, 880–896. doi:10.1037/0022-3518.104.22.1680. Duckworth, A. L., Peterson, C., Matthews, M. D., & Kelly, D. R. (2007). Grit: Perseverance andpassion for long-term goals. Journal of Personality and Social Psychology, 92, 1087–1101. doi:10.1037/0022-3522.214.171.1247. Figueredo, A. J., Vásquez, G., Brumbach, B. H., & Schneider, S. M. R. (2004). The heritability of life history strategy: The K-factor, covitality, and personality. Social Biology, 51, 121–143. Galton, F. (1884). Measurement of character. The Forthnightly Review, 36, 179–185. Musek, J. (2007). A general factor of personality: Evidence for the Big One in the five-factor model. Journal of Research in Personality, 41, 1213–1233. doi:10.1016/j.jrp.2007.02.003. Pozzebon, J., Damian, R. I., Hill, P. L., Lin, Y., Lapham, S., & Roberts, B. W. (2013). Establishing the validity and reliability of the Project Talent Personality Inventory. Frontiers in Psychology, 4, 1–13. doi:10.3389/fpsyg.2013.00968. Tucker-Drob, E. M., Briley, D. A., Engelhardt, L. E., Mann, F. D., & Harden, K. P. (2016). Genetically-mediated associations between measures of childhood character and academic achievement. Journal of Personality and Social Psychology, 111, 790–815. doi:10.1037/pspp0000098. Van der Linden, D., Dunkel, C. S., & Petrides, K. V. (2016). The general factor of personality (GFP) as social effectiveness: Review of the literature. Personality and Individual Differences, 101, 98–105. doi:10.1016/j.paid.2016.05.020. Van der Linden, D., Pekaar, K. A., Bakker, A. B., Schermer, J. A., Vernon, P. A., Dunkel, C. S., & Petrides, K. V. (2017). Overlap between the general factor of personality and emotional intelligence. Psychological Bulletin, 143, 36–52. doi:10.1037/bul0000078. Van der Linden, D., Te Nijenhuis, J., & Bakker, A. B. (2010). The general factor of personality: A meta-analysis and a criterion-related validity study. Journal of Research in Personality, 44, 315–327. doi:10.1016/j.jrp.2010.03.003.