- Email: [email protected]
Is the general factor of personality based on evaluative responding? Experimental manipulation of item-popularity in personality inventories
Personality and Individual Differences 96 (2016) 31–35
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Is the general factor of personality based on evaluative responding? Experimental manipulation of item-popularity in personality inventories Martin Bäckström ⁎, Fredrik Björklund Department of Psychology, Lund University, Box 213, SE-221 00 Lund, Sweden
a r t i c l e
i n f o
Article history: Received 21 November 2015 Received in revised form 12 February 2016 Accepted 22 February 2016 Available online xxxx Keywords: Self-ratings Personality Big Five General factor of personality
a b s t r a c t The general factor of personality (GFP) is understood as a hierarchically superordinate factor, which suggests that it and the subordinate personality traits are mutually dependent on one another. If a personality inventory captures the subordinate traits the GFP should appear too. Likewise, manipulating the GFP should affect the subordinate traits and vice versa. The current study was an attempt to uniquely affect the size of the GFP by manipulating the evaluativeness of the inventory. First we estimated a general factor in a standard (evaluative) personality inventory, and found it to be robust. Then we estimated it in an inventory with evaluatively neutralized items, and found it to be unreliable. Finally, the neutralized inventory was made evaluative again. As expected, the GFP reappeared, suggesting the increased evaluative content to be the cause. Results are discussed in relation to personality assessment and to higher order factors in personality theory. It is suggested that for determining whether the GFP exists or not researchers should turn to other measures than personality inventories. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction The general factor of personality (GFP) is conceptualized as a higherorder factor causing some of the variation in lower-order personality traits (Musek, 2007; Rushton, Bons, & Hur, 2008; Van der Linden, te Nijenhuis, & Bakker, 2010). The correlation that exists between the different factors in personality inventories is seen as a consequence of the GFP by some (e.g. Musek, 2007; Van der Linden et al., 2010) but not by others (e.g. Revelle & Wilt, 2013; Riemann & Kandler, 2010). Research on the GFP has focused on extracting GFP from personality inventories. The current study concerns precisely this; the role of the test instrument in GFP research. Notably, personality instruments vary in the extent that they give support to a GFP. In instruments such as Jackson's PRF (Jackson, 1984) and the HEXACO (de Vries, 2011) there is little general correlation between the main scales, whereas in other instruments there is substantial correlation. We propose that this is due to some instruments being under greater influence than others from individual differences in how respondents approach the inventory. Some people respond identically to items that refer to the same content, regardless of wording. Other people are more sensitive to item wording. Arguably, a major factor behind the degree of correlation between scales is the instrument's level of evaluativeness (Pettersson, Turkheimer, Horn, & Menatti, 2012), which can be defined as the extent to which the inventory affords responses that reflect the cultural norm for desired behavior. As indicated by previous research, the variation related to a GFP should diminish in the inventory if the items are ⁎ Corresponding author. E-mail address: [email protected] (M. Bäckström).
http://dx.doi.org/10.1016/j.paid.2016.02.058 0191-8869/© 2016 Elsevier Ltd. All rights reserved.
made less evaluative (Bäckström, Björklund, & Larsson, 2009). However, to the extent that evaluativeness and support for a GFP are linked together, it should also be possible to redesign an inventory in the other direction. It should be possible to alter an instrument that shows little evidence of a GFP so that it, after the evaluativeness treatment, now does show evidence of a GFP. Making the inventory more evaluative should increase the correlation between the factors, which could effectively be seen as substantiating the general factor. Thus, the current study concerns how systematically increasing vs. decreasing the degree of evaluativeness in personality inventories, (while keeping other factors constant) affects the support for the GFP in the inventories. How should we understand the nature of the general factor in personality inventories? The advocates of the GFP suggest that the factor constitutes a hierarchically superior content factor affecting all underlying factors in the model in the same way. In a five-factor model context, this implies that the GFP influences each of the five underlying content factors. According to this way of reasoning, if the inventory actually captures the lower-order personality traits, then it will capture the GFP too. People who are high in GFP should tend to be higher on all factors, i.e. high in extraversion, agreeableness, conscientiousness, emotional stability and openness to experience. The standard five-factor model, where each factor is assumed to be independent of the other factors, is more parsimonious in the sense that there is one factor less, which also avoids the problem of where to place it in the factor structure. However, as has been shown several times, this is not a model that is generally supported by personality inventory data (Bäckström, 2007; Musek, 2007). Following the reasoning above, if GFP is verified by correlation among the big five factors of inventories alone, all instruments that do
32
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
not show such correlation would weaken the case for the GFP. In other words, an inventory which captures the big five but has no higher order factor, should be problematic from a GFP perspective. If it is possible to affect the pattern of correlation between factors by applying a simple strategy, such as the evaluative neutralization method (where item popularity is manipulated; Bäckström & Björklund, 2013), exclusive reliance on self-reported personality data in investigating the GFP hypothesis may be problematic. It should be noted that the reduction of the GFP by means of e.g. the neutralization method could be caused by other factors than evaluativeness, e.g. reduction in reliability or independent changes of all the factors by introducing content from other personality content factors (which would need to be separate from the big five, otherwise they would create correlation according to the GFP theory). To succeed with the removal of the GFP, the manipulation needs to affect items of all the sub-factors equally, otherwise the success would only be partial, e.g. there would be no general reduction in correlation or the factorial structure would break down. Granting these obstacles, this appears to be possible; in previous research using evaluative neutralization there has been no reduction of reliability and the factor structure has been intact. All five factors have been retained, while their intercorrelations decreased (Bäckström, Björklund, & Larsson, 2014). However, recreating the GFP by means of a reversal of the same method as was used to remove it would be an even more powerful demonstration of the influence of evaluativeness on GFP. This issue is to the core of the debate on the validity of the GFP as a personality content factor, and the main focus of the current study. In a similar vein as in a classic ABA-design study, we expect to be able to extract a robust GFP in the original (evaluatively loaded) version of a personality inventory, expect little evidence of a GFP after the inventory has been evaluatively neutralized, and a robust GFP again after it has been made evaluative again. We hypothesize that relatively small changes to the wordings of items affect their popularity and the correlation between scales that are based on the items. More precisely, we predict that reducing evaluativeness will mean that the GFP becomes smaller and increasing evaluativeness that it becomes larger. We also predict that the general factor that is expected to appear in the “re-evaluized” inventory will correlate substantially with the general factor of a typical Five factor model (FFM) inventory, which would be key with regard to concluding that it captures GFP-related variance. In other words, our hypothesis is that simple rephrasing of item to be more popular will influence the ratings of every item, from all the five scales, in the same way.
facets per factor. The items of the FFM-Evaluative were constructed iteratively by rephrasing all items from the FFM-Neutralized. They were rephrased to become more popular (as described in Bäckström and Björklund (2013)), i.e. made more evaluative in the sense that many participants in the population should find them attractive (and willing to rate high). For example, Want to constantly meet and enjoy the company of friends and colleagues was turned into Want to have people around me. The item Carry out all tasks, even when I see them as unimportant was turned into Important tasks can sometimes be put on hold, and Have to achieve everything I set my mind to do was turned into Often achieve what I set my mind to do. An item was categorized as popular when the mean rating was .3 steps above the midpoint of the five point Likert scale (negatively worded items were reversed). To check whether the changes were successful, we had a fairly large sample of respondents (between 86 and 190) make self-ratings on the revised items. The rephrasing was iterated until most items were clearly popular (mean above 3.3). The IPIP-NEO had 151 (out of 200) items that were popular, the FFM-neutralized had 20 (out of 160), and the FFM-Evaluative had 100 (out of 160). The item mean was 3.63, 3.03 and 3.41 for the IPIP-NEO, FFM-Neutralized, and FFM-Evaluative, respectively. In other words, although the FFM-Evaluative had items that were more popular, they were not as popular as the original IPIP-NEO. 2.2. Participants and procedure Participants were Swedish-speaking spontaneous visitors to the site www.pimahb.se, i.e. they were not actively recruited for the study. All participants volunteered and were provided with some feedback on their results. Across samples there was about 65% women and the mean age was ca 30 years. Visitors who register on the site report their educational level and if they work. Of those who have registered, about 40% have reported more than three years of college level education, 21% college studies for less than three years, 31% have reported high-school and the rest a lower level of education (e.g. secondary school). Of all registered about 61% have reported working more than 20 h a week. The IPIP-NEO was administrated separately, and items were presented in a random order. The FFM-Evaluative and FFM-Neutralized were administered at the same time, and items were presented together randomly.
2. Method 2.1. Materials This study is based on three different inventories measuring the FFM model. 2.1.1. IPIP-NEO The first one is the IPIP-NEO inventory from the International Item Pool (Goldberg et al., 2006). We used a 200-item Swedish version of this inventory which has been described elsewhere (e.g. Bäckström et al., 2014) and shown to be a valid instrument to measure the FFM, on par with the more well-known NEO-PI-R (Costa & McCrae, 1992). 2.1.2. FFM-Neutralized This 160-item inventory was developed in a project on evaluative neutralization of items from the IPIP-NEO. Validities are in the range of the original inventory (Bäckström et al., 2014). The factors of the inventory have lower intercorrelations than the IPIP-NEO (and the NEO-PI-R) and the facets have fewer cross-loadings than IPIP-NEO. 2.1.3. FFM-Evaluative This inventory was created for the sole purpose of the present study. It was based on the same items as the FFM-Neutralized and has four
2.2.1. Statistical analyses The hypothesis that evaluativeness brings about a general factor in personality inventories was tested with Confirmatory Factor Analysis. To extract the presumed general factor, we created a bi-factorial model where the FFM factors were defined by their five respective facets and the general factor, with all loadings fixed to 1, was defined as a common factor loading on all the 20 facets (see Fig. 1, panel 2). To test the hypothesis that making inventories more vs. less evaluative increases vs. decreases the GFP, we estimated two models for each inventory. The models included the Big Five factors that the inventories were designed to measure as well as a measurement factor with loadings to all subscales of the inventory. For each inventory, the first model defined the measurement factor to have zero correlation for all subscales (observed variables, equivalent to Fig. 1, panel 1), and the second model defined the measurement factor to have unit loadings for all subscales. Since we are interested in the exact proportion of variance that can be attributed to the general factor in different inventories we used the Normed Fit Index, which measures the proportion of covariance explained by the models. We supplemented NFI with the Comparative Fit Index that adds a penalty for larger models and is more common in the literature. In addition, as an alternative way of estimating the amount of systematic variance in the general factor, the
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
33
Fig. 1. Conceptual models of the analyses.
omega hierarchical (Revelle & Zinbarg, 2009) was calculated. This was done by means of exploratory factor analysis (Revelle & Wilt, 2013).
3. Results First, the two models were tested with the IPIP-NEO (N = 3826), revealing a large difference in fit. The first model had a NFI of .525 (CFI = .527) and the second model had a NFI of .599 (CFI = 601). The mean standardized loading to the measurement factor in the second model was .49, suggesting a fair amount of common covariation between the subscales. Another way to summarize the importance of the general factor is to estimate Omega hierarchical, which was .53 for IPIP-NEO. The next two models to be tested concerned the FFM-Neutralized (N = 2027), the evaluatively neutralized inventory loosely based on IPIP-NEO. The first model, with zero loadings to the measurement factor, had a NFI of .800 (CFI = .811). The second model, with loadings to the measurement factor, had a NFI of .808 (CFI = .819). The difference was statistically significant (p b .001), although not very large. The mean standardized loading to the measurement factor was .27. Omega hierarchical was .27, indicating very low reliability for the general factor. The final two models to be tested concerned the FFM-Evaluative. It was found (N = 1311) that the first model had a NFI of .787 (CFI = .803) and the second model a NFI of .818 (CFI = .835), which is a substantially larger difference than the one found for the FFMNeutralized inventory. The mean standardized loadings to the measurement factor was .39, substantially less than the original IPIP-NEO, but larger than that of the FFM-Neutralized. Omega hierarchical was .49, indicating a considerably higher reliability for the
general factor in the FFM-Evaluative than in the FFM-Neutralized inventory.1 The next question to address was if the variance taken away from the original inventory and then reintroduced in the evaluative inventory was in fact the same. If the general factor from the original IPIP-NEO and the one from the new FFM-Evaluative inventory correlate strongly, this would support the general hypothesis in this study, since it suggests that they are akin to one another. In a small study (N = 91, 68% females, mean age = 24.1) all three inventories were used. It was found that the general factor from IPIP-NEO correlated r = .72 (corrected for attenuation r = .92) with the general factor from FFM-Evaluative. Of course, the two general factors were estimated vis-à-vis the same (neutralized) inventory, that is, they were estimated in terms of the difference between each of the evaluative inventories and the one that was evaluatively neutralized. The problem is that the correlation could have its origin solely in the neutralized inventory. To test whether this was the cause of the strong correlation, we created two general factors based on 10 (instead of 20) randomly chosen facets from the neutralized inventory, ending up with two, now independent, general factors. One was based on the IPIP-NEO facets together with half of the facets of FFMNeutralized, and the other was based on the FFM-Evaluative and the other half of the FFM-Neutralized. These two general factors correlated .53. However, the reliability of both general factors was low (.58 and .47, respectively), and the corrected correlation was considerably higher, almost 1.00.
1 Fixing all loadings to 1 gave us the opportunity to estimate the common variance between all facets, but did not optimize the estimation of a common factor. For the neutral inventory, it was not possible to estimate a general factor, the pattern of loadings did not indicate a general factor at all (i.e. that all facets have loadings in the same direction). When a model without this restriction was applied to the original and the evaluative inventories, the pattern of loadings clearly supported a general factor.
34
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
The final question to answer is whether a hierarchical model would receive support from CFA estimations of the two inventories. The analysis started with the basic FFM (see Fig. 1, panel 1). This model was compared with the typical hierarchical model, where the GFP is placed on top of the FFM factors, explaining common variation in the FFM factors (see Fig. 1, panel 3). First we analyzed the FFM-Neutralized. Adding the GFP to the basic model only increased the fit marginally (NFI increased from .811 to .825) and the standardized loadings were heterogeneous (E = .23, A = −.04, C = .27, ES = .94, and O = −.31). In other words, the results did not reveal a clear GFP factor. Given the variability of the loadings, this small increase in fit cannot be attributed to a general factor, but rather to some minor relations between a subset of scales. Analysis of the same hierarchal model on the FFM-Evaluative inventory showed stronger fit (NFI increased from .805 to .840), and the pattern of loadings clearly resembled the structure of a typical GFP factor (E = .58, A = .32, C = .64, ES = .63, and O = .29).2 4. Discussion The current study experimentally manipulated the evaluativeness of personality inventories. The results showed, as predicted, that more evaluativeness brings about a larger general factor. The general factor in the original (evaluative) inventory correlated strongly with the general factor of the re-evaluized inventory, which suggests that it is the same general factor. The same common variance (related to the evaluativeness) in personality ratings was produced with small changes to an evaluatively neutral inventory, without affecting the factor structure. This, along with the methodological implication that the findings have, is the unique contribution of the study. Crucially, the new evaluative inventory allowed for generating a general factor (correlating substantially with the corresponding factor of the IPIP-NEO), which suggests that it is the same factor as has been dubbed GFP in the past. Why are these results important? There are at least two implications of this research. 4.1. Implications for personality models with higher order factors We have shown that it is possible to reduce and reintroduce a general factor in personality inventories by manipulating evaluativeness. This is potentially useful information, not least since this general factor appears to be the same as what has been called the GFP, which some have suggested may be interpreted as a substantial personality factor (e.g. Musek, 2007). It is a fact that a general factor appears in most personality inventories, and that it has to do with the intercorrelations between factor scales. But a higher-order factor in a personality inventory is not equivalent to a hierarchically superordinate GFP. The current results suggest that it may rather be related to how respondents react to evaluative items, i.e. another factor that need not necessarily be superordinate, but can equally well be parallel to the five factors of the FFM (Bäckström & Björklund, 2014). However, the current knowledge of higher order personality factors is not such that we can conclude either the existence or non-existence of the GFP in the “real world”. How should we interpret the fact that a general factor often appears in personality measures, but can be reduced or increased by an evaluativeness manipulation? One implication of our results is that they problematize the interpretation of GFP in terms of substance. They do not show that such an interpretation is wrong, but contribute to weakening the conclusions that are possible to draw from inventory data and this is problematic for the view that the general factor in personality inventories is a substantial personality factor (GFP). On the other hand, this does not imply that the general factor is a nuisance factor. In fact, it is conceivable that the GFP exists, independently of 2 The correlation between factor scales from the neutralized and the re-evaluized inventory were .85, .76, 80, 92, 84, for extraversion, agreeableness, conscientiousness, emotional stability and openness, respectively.
personality inventories, and shows itself only in the more evaluative inventories, as these afford responses that are in line with the behavioral ideals of the culture at hand (Bell, Woodley, Schermer, & Vernon, 2012; Dunkel, 2013, for research on GFP and cultural standards). An alternative interpretation is that evaluative inventories might be seen as activating the motivation in the GFP-high person to rate in accordance with the normative behavioral standards of the culture. This would be similar to how self-enhancement motivation interacts with the phrasing of the items of a personality inventory (Bäckström & Björklund, 2013). The motivation to approach situations in the most adaptive way is conceptually closely related to self-enhancement, and it is not possible to distinguish them in the current research. Nevertheless, there may very well exist (also beyond the instrument itself) a motivationally related content factor which causes the general factor in personality inventories. Individual differences in the motivation to strive towards ideal behavior would be an example of such a content factor. A motivational factor of this kind may also explain why the general factor in personality measures shows relations to external criteria such as reported by van der Linden, Oostrom, Born, Van der Molen, and Serlie (2014). Research on the GFP has relied extensively on personality inventories and the current study shows the potential problems of this practice. To move forward on the issue of whether the GFP exists or not, researchers should shift focus to other methods than inventories. The GFP is an important and exciting idea, and to give it the attention that it deserves we recommend that researchers either turn to alternative methods, develop the methods currently used, or invent completely new ones. One important method is to analyze actual behavior, with the goal to identify any real relations between personality factors. For example, if it can be shown that actual extraverted behavior is systematically related to actual behaviors indicating openness to experience, then the case for higher order factors would be much stronger. Or put in the context of the GFP, if there is a systematic tendency for behaviors from not only two personality factors to go together, but all five of the FFM, this would be strong support for the GFP. To our knowledge, there are yet no studies indicating that actual behavior from different FFM-factors do go together. 4.2. Implications for personality measurement There is an intense and intricate scientific debate regarding how higher-order factors in personality inventories should be estimated. It has been shown, several times, that the correlation between the general factor (or GFP) of self-ratings and peer-ratings is near zero when the variability of other FFM-factors has been controlled for. In fact, the general finding from MTMM-studies is that the factor is substantially reduced when method factors are controlled for (e.g. Biesanz & West, 2004; Chang, Connelly, & Geeza, 2012; Danay & Ziegler, 2011; Riemann & Kandler, 2010), and this has been proof enough for many scholars to dismiss the GFP as a personality content factor (Just, 2011). In support of the other side, it has been shown that people who rate themselves in a desirable fashion in personality inventories often seem to act with higher confidence and are more successful in some areas. In the self-enhancement literature, Kwan, John, Robins, and Kuang (2008) showed that raters who self-enhance (a concept very close to social desirability) are more successful given that they also are highly appreciated by their peers. The same kind of relation has been suggested by researchers proposing a GFP. They suggest that people with mean ratings that are generally higher in extraversion, agreeableness, conscientiousness, emotional stability and openness are more successful (e.g. Dunkel & van der Linden, 2014) and find relations to behavioral criteria (e.g. work performance; van der Linden et al., 2014). Use of results from GFP research as proof in this debate is however complicated by the fact that the methods of operationalizing it has not always been controlled for content from the FFM factors (Chang et al.,
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
2012). Typically, the general factor in personality inventories has not been extracted from the common content of the FFM after taking away variance of the FFM, i.e. a bi-factorial model. Instead, the operationalization has been based on the first PCA (or Principal axis factor, or similar weighing together of the FFM scales). We agree that the estimations should be made by means of a bifactorial model rather than a hierarchical, since it provides a purer operationalization of the general factor. 4.3. Limitations and future research There are two notable limitations to the current research. First, the studies relied on a single personality inventory, based on the Five Factor Model. This puts limits on the generalizability of the results to other models of personality. Second, although an independent sample of respondents rated the manipulated items so that we could see whether item popularity was affected, there was no direct measure of how respondents in the main study reacted to item evaluativeness. Such measures would be useful in future research, to further track the processes involved in responding to evaluative vs. neutral personality items, which would give further insights into the content of the GFP. 5. Conclusion The findings from the current study reveal the limitations of exclusively relying on self-rated personality in research on the GFP. The fact that decreasing the evaluativeness of a personality inventory reduces the GFP, whereas increasing the evaluativeness makes it reappear, constitutes a challenge to the interpretation of the general factor in personality inventories as a content factor. Research on the GFP would benefit from turning to other methods than personality inventories. Acknowledgment This research was supported by grant P10-0922:1 from the Swedish Foundation for Humanities and Social Sciences. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.paid.2016.02.058. References Bäckström, M. (2007). Higher-order factors in a five-factor personality inventory and its relation to social desirability. European Journal of Psychological Assessment, 23(2), 63–70. http://dx.doi.org/10.1027/1015-5759.23.2.63. Bäckström, M., & Björklund, F. (2013). Social desirability in personality inventories: Symptoms, diagnosis and prescribed cure. Scandinavian Journal of Psychology, 54, 152–159. http://dx.doi.org/10.1111/sjop.12015. Bäckström, M., & Björklund, F. (2014). Social desirability in personality inventories: The nature of the evaluative factor. Journal of Individual Differences, 35, 144–157. http:// dx.doi.org/10.1027/1614-0001/a000138. Bäckström, M., Björklund, F., & Larsson, M. R. (2009). Five-factor inventories have a major higher order factor related to social desirability which can be reduced by framing
35
items neutrally. Journal of Research in Personality, 43, 335–344. http://dx.doi.org/10. 1016/j.jrp.2008.12.013. Bäckström, M., Björklund, F., & Larsson, M. R. (2014). Criterion validity is maintained when items are evaluatively neutralized: Evidence from a full-scale five-factor model inventory. European Journal of Personality, 28, 620–633. http://dx.doi.org/10. 1002/per.1960. Bell, E., Woodley, M. A., Schermer, J. A., & Vernon, P. A. (2012). Politics and the general factor of personality. Personality and Individual Differences, 53(5), 546–551. http://dx.doi. org/10.1016/j.paid.2012.04.027. Biesanz, J. C., & West, S. G. (2004). Towards understanding assessments of the Big Five: Multitrait-multimethod analyses of convergent and discriminant validity across measurement occasion and type of observer. Journal of Personality, 72, 845–876. http://dx. doi.org/10.1111/j.0022-3506.2004.00282.x. Chang, L., Connelly, B. S., & Geeza, A. A. (2012). Separating method factors and higher order traits of the Big Five: A meta-analytic multitrait–multimethod approach. Journal of Personality and Social Psychology, 102, 408–426. http://dx.doi.org/10.1037/ a0025559. Costa, P. T., & McCrae, R. R. (1992). Revised NEO personality inventory (NEO-PI-R) and NEO five-factor inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources. Danay, E., & Ziegler, M. (2011). Is there really a single factor of personality? A multirater approach to the apex of personality. Journal of Research in Personality, 45, 560–567. http://dx.doi.org/10.1016/j.jrp.2011.07.003. Dunkel, C. S. (2013). Evidence for the role of the General Factor of Personality (GFP) in enculturation: The GFP and self-construal in Japanese and American samples. Personality and Individual Differences, 55(4), 417–421. http://dx.doi.org/10.1016/j. paid.2013.04.002. Dunkel, C. S., & van der Linden, D. (2014). Evidence for the general factor of personality as social-effectiveness. Personality and Individual Differences, 64, 147–151. http://dx.doi. org/10.1016/j.paid.2014.02.030. Goldberg, L. R., Johnson, J. A., Eber, H. W., Hogan, R., Ashton, M. C., Cloninger, C. R., & Gough, H. G. (2006). The international personality item pool and the future of public-domain personality measures. Journal of Research in Personality, 40, 84–96. http://dx.doi.org/10.1016/j.jrp.2005.08.007. Jackson, D. N. (1984). Personality research form manual (3rd ed.). Port Huron, MI: Research Psychologists Press. Just, C. (2011). A review of literature on the general factor of personality. Personality and Individual Differences, 50, 765–771. http://dx.doi.org/10.1016/j.paid.2011.01.008. Kwan, V. S. Y., John, O. P., Robins, R. W., & Kuang, L. L. (2008). Conceptualizing and assessing self-enhancement bias: A componential approach. Journal of Personality and Social Psychology, 94, 1062–1077. http://dx.doi.org/10.1037/0022-3514.94.6.1062. Musek, J. (2007). A general factor of personality: Evidence for the Big One in the fivefactor model. Journal of Research in Personality, 41, 1213–1233. http://dx.doi.org/10. 1016/j.jrp.2007.02.003. Pettersson, E., Turkheimer, E., Horn, E. E., & Menatti, A. R. (2012). The general factor of personality and evaluation. European Journal of Personality, 26(3), 292–302. http:// dx.doi.org/10.1002/per.839. Revelle, W., & Wilt, J. (2013). The general factor of personality: A general critique. Journal of Research in Personality, 47, 493–504. http://dx.doi.org/10.1016/j.jrp.2013.04.012. Revelle, W., & Zinbarg, R. E. (2009). Coefficients alpha, beta, omega, and the glb: Comments on Sijtsma. Psychometrica, 74, 145–154. http://dx.doi.org/10.1007/S11336008-9102-Z. Riemann, R., & Kandler, C. (2010). Construct validation using multitrait-multimethodtwin data: The case of a general factor of personality. European Journal of Personality, 24, 258–277. http://dx.doi.org/10.1002/per.760. Rushton, J. P., Bons, T. A., & Hur, Y. (2008). The genetics and evolution of the general factor of personality. Journal of Research in Personality, 42, 1173–1185. http://dx.doi.org/10. 1016/j.jrp.2008.03.002. Van der Linden, D., Oostrom, J. K., Born, M. P., Van der Molen, H. T., & Serlie, A. W. (2014). Knowing what to do in social situations: The general factor of personality and performance on situational judgment tests. Journal Personnel Psychology, 13, 107–115. http://dx.doi.org/10.1027/1866-5888/a000113. Van der Linden, D., te Nijenhuis, J., & Bakker, A. B. (2010). The General Factor of Personality: A meta-analysis of Big Five intercorrelations and a criterion-related validity study. Journal of Research in Personality, 44, 315–327. http://dx.doi.org/10.1016/j.jrp. 2010.03.003. de Vries, R. E. (2011). No evidence for a General Factor of Personality in the HEXACO Personality Inventory. Journal of Research in Personality, 45, 229–232. http://dx.doi.org/ 10.1016/j.jrp.2010.12.002.
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Is the general factor of personality based on evaluative responding? Experimental manipulation of item-popularity in personality inventories Martin Bäckström ⁎, Fredrik Björklund Department of Psychology, Lund University, Box 213, SE-221 00 Lund, Sweden
a r t i c l e
i n f o
Article history: Received 21 November 2015 Received in revised form 12 February 2016 Accepted 22 February 2016 Available online xxxx Keywords: Self-ratings Personality Big Five General factor of personality
a b s t r a c t The general factor of personality (GFP) is understood as a hierarchically superordinate factor, which suggests that it and the subordinate personality traits are mutually dependent on one another. If a personality inventory captures the subordinate traits the GFP should appear too. Likewise, manipulating the GFP should affect the subordinate traits and vice versa. The current study was an attempt to uniquely affect the size of the GFP by manipulating the evaluativeness of the inventory. First we estimated a general factor in a standard (evaluative) personality inventory, and found it to be robust. Then we estimated it in an inventory with evaluatively neutralized items, and found it to be unreliable. Finally, the neutralized inventory was made evaluative again. As expected, the GFP reappeared, suggesting the increased evaluative content to be the cause. Results are discussed in relation to personality assessment and to higher order factors in personality theory. It is suggested that for determining whether the GFP exists or not researchers should turn to other measures than personality inventories. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction The general factor of personality (GFP) is conceptualized as a higherorder factor causing some of the variation in lower-order personality traits (Musek, 2007; Rushton, Bons, & Hur, 2008; Van der Linden, te Nijenhuis, & Bakker, 2010). The correlation that exists between the different factors in personality inventories is seen as a consequence of the GFP by some (e.g. Musek, 2007; Van der Linden et al., 2010) but not by others (e.g. Revelle & Wilt, 2013; Riemann & Kandler, 2010). Research on the GFP has focused on extracting GFP from personality inventories. The current study concerns precisely this; the role of the test instrument in GFP research. Notably, personality instruments vary in the extent that they give support to a GFP. In instruments such as Jackson's PRF (Jackson, 1984) and the HEXACO (de Vries, 2011) there is little general correlation between the main scales, whereas in other instruments there is substantial correlation. We propose that this is due to some instruments being under greater influence than others from individual differences in how respondents approach the inventory. Some people respond identically to items that refer to the same content, regardless of wording. Other people are more sensitive to item wording. Arguably, a major factor behind the degree of correlation between scales is the instrument's level of evaluativeness (Pettersson, Turkheimer, Horn, & Menatti, 2012), which can be defined as the extent to which the inventory affords responses that reflect the cultural norm for desired behavior. As indicated by previous research, the variation related to a GFP should diminish in the inventory if the items are ⁎ Corresponding author. E-mail address: [email protected] (M. Bäckström).
http://dx.doi.org/10.1016/j.paid.2016.02.058 0191-8869/© 2016 Elsevier Ltd. All rights reserved.
made less evaluative (Bäckström, Björklund, & Larsson, 2009). However, to the extent that evaluativeness and support for a GFP are linked together, it should also be possible to redesign an inventory in the other direction. It should be possible to alter an instrument that shows little evidence of a GFP so that it, after the evaluativeness treatment, now does show evidence of a GFP. Making the inventory more evaluative should increase the correlation between the factors, which could effectively be seen as substantiating the general factor. Thus, the current study concerns how systematically increasing vs. decreasing the degree of evaluativeness in personality inventories, (while keeping other factors constant) affects the support for the GFP in the inventories. How should we understand the nature of the general factor in personality inventories? The advocates of the GFP suggest that the factor constitutes a hierarchically superior content factor affecting all underlying factors in the model in the same way. In a five-factor model context, this implies that the GFP influences each of the five underlying content factors. According to this way of reasoning, if the inventory actually captures the lower-order personality traits, then it will capture the GFP too. People who are high in GFP should tend to be higher on all factors, i.e. high in extraversion, agreeableness, conscientiousness, emotional stability and openness to experience. The standard five-factor model, where each factor is assumed to be independent of the other factors, is more parsimonious in the sense that there is one factor less, which also avoids the problem of where to place it in the factor structure. However, as has been shown several times, this is not a model that is generally supported by personality inventory data (Bäckström, 2007; Musek, 2007). Following the reasoning above, if GFP is verified by correlation among the big five factors of inventories alone, all instruments that do
32
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
not show such correlation would weaken the case for the GFP. In other words, an inventory which captures the big five but has no higher order factor, should be problematic from a GFP perspective. If it is possible to affect the pattern of correlation between factors by applying a simple strategy, such as the evaluative neutralization method (where item popularity is manipulated; Bäckström & Björklund, 2013), exclusive reliance on self-reported personality data in investigating the GFP hypothesis may be problematic. It should be noted that the reduction of the GFP by means of e.g. the neutralization method could be caused by other factors than evaluativeness, e.g. reduction in reliability or independent changes of all the factors by introducing content from other personality content factors (which would need to be separate from the big five, otherwise they would create correlation according to the GFP theory). To succeed with the removal of the GFP, the manipulation needs to affect items of all the sub-factors equally, otherwise the success would only be partial, e.g. there would be no general reduction in correlation or the factorial structure would break down. Granting these obstacles, this appears to be possible; in previous research using evaluative neutralization there has been no reduction of reliability and the factor structure has been intact. All five factors have been retained, while their intercorrelations decreased (Bäckström, Björklund, & Larsson, 2014). However, recreating the GFP by means of a reversal of the same method as was used to remove it would be an even more powerful demonstration of the influence of evaluativeness on GFP. This issue is to the core of the debate on the validity of the GFP as a personality content factor, and the main focus of the current study. In a similar vein as in a classic ABA-design study, we expect to be able to extract a robust GFP in the original (evaluatively loaded) version of a personality inventory, expect little evidence of a GFP after the inventory has been evaluatively neutralized, and a robust GFP again after it has been made evaluative again. We hypothesize that relatively small changes to the wordings of items affect their popularity and the correlation between scales that are based on the items. More precisely, we predict that reducing evaluativeness will mean that the GFP becomes smaller and increasing evaluativeness that it becomes larger. We also predict that the general factor that is expected to appear in the “re-evaluized” inventory will correlate substantially with the general factor of a typical Five factor model (FFM) inventory, which would be key with regard to concluding that it captures GFP-related variance. In other words, our hypothesis is that simple rephrasing of item to be more popular will influence the ratings of every item, from all the five scales, in the same way.
facets per factor. The items of the FFM-Evaluative were constructed iteratively by rephrasing all items from the FFM-Neutralized. They were rephrased to become more popular (as described in Bäckström and Björklund (2013)), i.e. made more evaluative in the sense that many participants in the population should find them attractive (and willing to rate high). For example, Want to constantly meet and enjoy the company of friends and colleagues was turned into Want to have people around me. The item Carry out all tasks, even when I see them as unimportant was turned into Important tasks can sometimes be put on hold, and Have to achieve everything I set my mind to do was turned into Often achieve what I set my mind to do. An item was categorized as popular when the mean rating was .3 steps above the midpoint of the five point Likert scale (negatively worded items were reversed). To check whether the changes were successful, we had a fairly large sample of respondents (between 86 and 190) make self-ratings on the revised items. The rephrasing was iterated until most items were clearly popular (mean above 3.3). The IPIP-NEO had 151 (out of 200) items that were popular, the FFM-neutralized had 20 (out of 160), and the FFM-Evaluative had 100 (out of 160). The item mean was 3.63, 3.03 and 3.41 for the IPIP-NEO, FFM-Neutralized, and FFM-Evaluative, respectively. In other words, although the FFM-Evaluative had items that were more popular, they were not as popular as the original IPIP-NEO. 2.2. Participants and procedure Participants were Swedish-speaking spontaneous visitors to the site www.pimahb.se, i.e. they were not actively recruited for the study. All participants volunteered and were provided with some feedback on their results. Across samples there was about 65% women and the mean age was ca 30 years. Visitors who register on the site report their educational level and if they work. Of those who have registered, about 40% have reported more than three years of college level education, 21% college studies for less than three years, 31% have reported high-school and the rest a lower level of education (e.g. secondary school). Of all registered about 61% have reported working more than 20 h a week. The IPIP-NEO was administrated separately, and items were presented in a random order. The FFM-Evaluative and FFM-Neutralized were administered at the same time, and items were presented together randomly.
2. Method 2.1. Materials This study is based on three different inventories measuring the FFM model. 2.1.1. IPIP-NEO The first one is the IPIP-NEO inventory from the International Item Pool (Goldberg et al., 2006). We used a 200-item Swedish version of this inventory which has been described elsewhere (e.g. Bäckström et al., 2014) and shown to be a valid instrument to measure the FFM, on par with the more well-known NEO-PI-R (Costa & McCrae, 1992). 2.1.2. FFM-Neutralized This 160-item inventory was developed in a project on evaluative neutralization of items from the IPIP-NEO. Validities are in the range of the original inventory (Bäckström et al., 2014). The factors of the inventory have lower intercorrelations than the IPIP-NEO (and the NEO-PI-R) and the facets have fewer cross-loadings than IPIP-NEO. 2.1.3. FFM-Evaluative This inventory was created for the sole purpose of the present study. It was based on the same items as the FFM-Neutralized and has four
2.2.1. Statistical analyses The hypothesis that evaluativeness brings about a general factor in personality inventories was tested with Confirmatory Factor Analysis. To extract the presumed general factor, we created a bi-factorial model where the FFM factors were defined by their five respective facets and the general factor, with all loadings fixed to 1, was defined as a common factor loading on all the 20 facets (see Fig. 1, panel 2). To test the hypothesis that making inventories more vs. less evaluative increases vs. decreases the GFP, we estimated two models for each inventory. The models included the Big Five factors that the inventories were designed to measure as well as a measurement factor with loadings to all subscales of the inventory. For each inventory, the first model defined the measurement factor to have zero correlation for all subscales (observed variables, equivalent to Fig. 1, panel 1), and the second model defined the measurement factor to have unit loadings for all subscales. Since we are interested in the exact proportion of variance that can be attributed to the general factor in different inventories we used the Normed Fit Index, which measures the proportion of covariance explained by the models. We supplemented NFI with the Comparative Fit Index that adds a penalty for larger models and is more common in the literature. In addition, as an alternative way of estimating the amount of systematic variance in the general factor, the
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
33
Fig. 1. Conceptual models of the analyses.
omega hierarchical (Revelle & Zinbarg, 2009) was calculated. This was done by means of exploratory factor analysis (Revelle & Wilt, 2013).
3. Results First, the two models were tested with the IPIP-NEO (N = 3826), revealing a large difference in fit. The first model had a NFI of .525 (CFI = .527) and the second model had a NFI of .599 (CFI = 601). The mean standardized loading to the measurement factor in the second model was .49, suggesting a fair amount of common covariation between the subscales. Another way to summarize the importance of the general factor is to estimate Omega hierarchical, which was .53 for IPIP-NEO. The next two models to be tested concerned the FFM-Neutralized (N = 2027), the evaluatively neutralized inventory loosely based on IPIP-NEO. The first model, with zero loadings to the measurement factor, had a NFI of .800 (CFI = .811). The second model, with loadings to the measurement factor, had a NFI of .808 (CFI = .819). The difference was statistically significant (p b .001), although not very large. The mean standardized loading to the measurement factor was .27. Omega hierarchical was .27, indicating very low reliability for the general factor. The final two models to be tested concerned the FFM-Evaluative. It was found (N = 1311) that the first model had a NFI of .787 (CFI = .803) and the second model a NFI of .818 (CFI = .835), which is a substantially larger difference than the one found for the FFMNeutralized inventory. The mean standardized loadings to the measurement factor was .39, substantially less than the original IPIP-NEO, but larger than that of the FFM-Neutralized. Omega hierarchical was .49, indicating a considerably higher reliability for the
general factor in the FFM-Evaluative than in the FFM-Neutralized inventory.1 The next question to address was if the variance taken away from the original inventory and then reintroduced in the evaluative inventory was in fact the same. If the general factor from the original IPIP-NEO and the one from the new FFM-Evaluative inventory correlate strongly, this would support the general hypothesis in this study, since it suggests that they are akin to one another. In a small study (N = 91, 68% females, mean age = 24.1) all three inventories were used. It was found that the general factor from IPIP-NEO correlated r = .72 (corrected for attenuation r = .92) with the general factor from FFM-Evaluative. Of course, the two general factors were estimated vis-à-vis the same (neutralized) inventory, that is, they were estimated in terms of the difference between each of the evaluative inventories and the one that was evaluatively neutralized. The problem is that the correlation could have its origin solely in the neutralized inventory. To test whether this was the cause of the strong correlation, we created two general factors based on 10 (instead of 20) randomly chosen facets from the neutralized inventory, ending up with two, now independent, general factors. One was based on the IPIP-NEO facets together with half of the facets of FFMNeutralized, and the other was based on the FFM-Evaluative and the other half of the FFM-Neutralized. These two general factors correlated .53. However, the reliability of both general factors was low (.58 and .47, respectively), and the corrected correlation was considerably higher, almost 1.00.
1 Fixing all loadings to 1 gave us the opportunity to estimate the common variance between all facets, but did not optimize the estimation of a common factor. For the neutral inventory, it was not possible to estimate a general factor, the pattern of loadings did not indicate a general factor at all (i.e. that all facets have loadings in the same direction). When a model without this restriction was applied to the original and the evaluative inventories, the pattern of loadings clearly supported a general factor.
34
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
The final question to answer is whether a hierarchical model would receive support from CFA estimations of the two inventories. The analysis started with the basic FFM (see Fig. 1, panel 1). This model was compared with the typical hierarchical model, where the GFP is placed on top of the FFM factors, explaining common variation in the FFM factors (see Fig. 1, panel 3). First we analyzed the FFM-Neutralized. Adding the GFP to the basic model only increased the fit marginally (NFI increased from .811 to .825) and the standardized loadings were heterogeneous (E = .23, A = −.04, C = .27, ES = .94, and O = −.31). In other words, the results did not reveal a clear GFP factor. Given the variability of the loadings, this small increase in fit cannot be attributed to a general factor, but rather to some minor relations between a subset of scales. Analysis of the same hierarchal model on the FFM-Evaluative inventory showed stronger fit (NFI increased from .805 to .840), and the pattern of loadings clearly resembled the structure of a typical GFP factor (E = .58, A = .32, C = .64, ES = .63, and O = .29).2 4. Discussion The current study experimentally manipulated the evaluativeness of personality inventories. The results showed, as predicted, that more evaluativeness brings about a larger general factor. The general factor in the original (evaluative) inventory correlated strongly with the general factor of the re-evaluized inventory, which suggests that it is the same general factor. The same common variance (related to the evaluativeness) in personality ratings was produced with small changes to an evaluatively neutral inventory, without affecting the factor structure. This, along with the methodological implication that the findings have, is the unique contribution of the study. Crucially, the new evaluative inventory allowed for generating a general factor (correlating substantially with the corresponding factor of the IPIP-NEO), which suggests that it is the same factor as has been dubbed GFP in the past. Why are these results important? There are at least two implications of this research. 4.1. Implications for personality models with higher order factors We have shown that it is possible to reduce and reintroduce a general factor in personality inventories by manipulating evaluativeness. This is potentially useful information, not least since this general factor appears to be the same as what has been called the GFP, which some have suggested may be interpreted as a substantial personality factor (e.g. Musek, 2007). It is a fact that a general factor appears in most personality inventories, and that it has to do with the intercorrelations between factor scales. But a higher-order factor in a personality inventory is not equivalent to a hierarchically superordinate GFP. The current results suggest that it may rather be related to how respondents react to evaluative items, i.e. another factor that need not necessarily be superordinate, but can equally well be parallel to the five factors of the FFM (Bäckström & Björklund, 2014). However, the current knowledge of higher order personality factors is not such that we can conclude either the existence or non-existence of the GFP in the “real world”. How should we interpret the fact that a general factor often appears in personality measures, but can be reduced or increased by an evaluativeness manipulation? One implication of our results is that they problematize the interpretation of GFP in terms of substance. They do not show that such an interpretation is wrong, but contribute to weakening the conclusions that are possible to draw from inventory data and this is problematic for the view that the general factor in personality inventories is a substantial personality factor (GFP). On the other hand, this does not imply that the general factor is a nuisance factor. In fact, it is conceivable that the GFP exists, independently of 2 The correlation between factor scales from the neutralized and the re-evaluized inventory were .85, .76, 80, 92, 84, for extraversion, agreeableness, conscientiousness, emotional stability and openness, respectively.
personality inventories, and shows itself only in the more evaluative inventories, as these afford responses that are in line with the behavioral ideals of the culture at hand (Bell, Woodley, Schermer, & Vernon, 2012; Dunkel, 2013, for research on GFP and cultural standards). An alternative interpretation is that evaluative inventories might be seen as activating the motivation in the GFP-high person to rate in accordance with the normative behavioral standards of the culture. This would be similar to how self-enhancement motivation interacts with the phrasing of the items of a personality inventory (Bäckström & Björklund, 2013). The motivation to approach situations in the most adaptive way is conceptually closely related to self-enhancement, and it is not possible to distinguish them in the current research. Nevertheless, there may very well exist (also beyond the instrument itself) a motivationally related content factor which causes the general factor in personality inventories. Individual differences in the motivation to strive towards ideal behavior would be an example of such a content factor. A motivational factor of this kind may also explain why the general factor in personality measures shows relations to external criteria such as reported by van der Linden, Oostrom, Born, Van der Molen, and Serlie (2014). Research on the GFP has relied extensively on personality inventories and the current study shows the potential problems of this practice. To move forward on the issue of whether the GFP exists or not, researchers should shift focus to other methods than inventories. The GFP is an important and exciting idea, and to give it the attention that it deserves we recommend that researchers either turn to alternative methods, develop the methods currently used, or invent completely new ones. One important method is to analyze actual behavior, with the goal to identify any real relations between personality factors. For example, if it can be shown that actual extraverted behavior is systematically related to actual behaviors indicating openness to experience, then the case for higher order factors would be much stronger. Or put in the context of the GFP, if there is a systematic tendency for behaviors from not only two personality factors to go together, but all five of the FFM, this would be strong support for the GFP. To our knowledge, there are yet no studies indicating that actual behavior from different FFM-factors do go together. 4.2. Implications for personality measurement There is an intense and intricate scientific debate regarding how higher-order factors in personality inventories should be estimated. It has been shown, several times, that the correlation between the general factor (or GFP) of self-ratings and peer-ratings is near zero when the variability of other FFM-factors has been controlled for. In fact, the general finding from MTMM-studies is that the factor is substantially reduced when method factors are controlled for (e.g. Biesanz & West, 2004; Chang, Connelly, & Geeza, 2012; Danay & Ziegler, 2011; Riemann & Kandler, 2010), and this has been proof enough for many scholars to dismiss the GFP as a personality content factor (Just, 2011). In support of the other side, it has been shown that people who rate themselves in a desirable fashion in personality inventories often seem to act with higher confidence and are more successful in some areas. In the self-enhancement literature, Kwan, John, Robins, and Kuang (2008) showed that raters who self-enhance (a concept very close to social desirability) are more successful given that they also are highly appreciated by their peers. The same kind of relation has been suggested by researchers proposing a GFP. They suggest that people with mean ratings that are generally higher in extraversion, agreeableness, conscientiousness, emotional stability and openness are more successful (e.g. Dunkel & van der Linden, 2014) and find relations to behavioral criteria (e.g. work performance; van der Linden et al., 2014). Use of results from GFP research as proof in this debate is however complicated by the fact that the methods of operationalizing it has not always been controlled for content from the FFM factors (Chang et al.,
M. Bäckström, F. Björklund / Personality and Individual Differences 96 (2016) 31–35
2012). Typically, the general factor in personality inventories has not been extracted from the common content of the FFM after taking away variance of the FFM, i.e. a bi-factorial model. Instead, the operationalization has been based on the first PCA (or Principal axis factor, or similar weighing together of the FFM scales). We agree that the estimations should be made by means of a bifactorial model rather than a hierarchical, since it provides a purer operationalization of the general factor. 4.3. Limitations and future research There are two notable limitations to the current research. First, the studies relied on a single personality inventory, based on the Five Factor Model. This puts limits on the generalizability of the results to other models of personality. Second, although an independent sample of respondents rated the manipulated items so that we could see whether item popularity was affected, there was no direct measure of how respondents in the main study reacted to item evaluativeness. Such measures would be useful in future research, to further track the processes involved in responding to evaluative vs. neutral personality items, which would give further insights into the content of the GFP. 5. Conclusion The findings from the current study reveal the limitations of exclusively relying on self-rated personality in research on the GFP. The fact that decreasing the evaluativeness of a personality inventory reduces the GFP, whereas increasing the evaluativeness makes it reappear, constitutes a challenge to the interpretation of the general factor in personality inventories as a content factor. Research on the GFP would benefit from turning to other methods than personality inventories. Acknowledgment This research was supported by grant P10-0922:1 from the Swedish Foundation for Humanities and Social Sciences. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.paid.2016.02.058. References Bäckström, M. (2007). Higher-order factors in a five-factor personality inventory and its relation to social desirability. European Journal of Psychological Assessment, 23(2), 63–70. http://dx.doi.org/10.1027/1015-5759.23.2.63. Bäckström, M., & Björklund, F. (2013). Social desirability in personality inventories: Symptoms, diagnosis and prescribed cure. Scandinavian Journal of Psychology, 54, 152–159. http://dx.doi.org/10.1111/sjop.12015. Bäckström, M., & Björklund, F. (2014). Social desirability in personality inventories: The nature of the evaluative factor. Journal of Individual Differences, 35, 144–157. http:// dx.doi.org/10.1027/1614-0001/a000138. Bäckström, M., Björklund, F., & Larsson, M. R. (2009). Five-factor inventories have a major higher order factor related to social desirability which can be reduced by framing
35
items neutrally. Journal of Research in Personality, 43, 335–344. http://dx.doi.org/10. 1016/j.jrp.2008.12.013. Bäckström, M., Björklund, F., & Larsson, M. R. (2014). Criterion validity is maintained when items are evaluatively neutralized: Evidence from a full-scale five-factor model inventory. European Journal of Personality, 28, 620–633. http://dx.doi.org/10. 1002/per.1960. Bell, E., Woodley, M. A., Schermer, J. A., & Vernon, P. A. (2012). Politics and the general factor of personality. Personality and Individual Differences, 53(5), 546–551. http://dx.doi. org/10.1016/j.paid.2012.04.027. Biesanz, J. C., & West, S. G. (2004). Towards understanding assessments of the Big Five: Multitrait-multimethod analyses of convergent and discriminant validity across measurement occasion and type of observer. Journal of Personality, 72, 845–876. http://dx. doi.org/10.1111/j.0022-3506.2004.00282.x. Chang, L., Connelly, B. S., & Geeza, A. A. (2012). Separating method factors and higher order traits of the Big Five: A meta-analytic multitrait–multimethod approach. Journal of Personality and Social Psychology, 102, 408–426. http://dx.doi.org/10.1037/ a0025559. Costa, P. T., & McCrae, R. R. (1992). Revised NEO personality inventory (NEO-PI-R) and NEO five-factor inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources. Danay, E., & Ziegler, M. (2011). Is there really a single factor of personality? A multirater approach to the apex of personality. Journal of Research in Personality, 45, 560–567. http://dx.doi.org/10.1016/j.jrp.2011.07.003. Dunkel, C. S. (2013). Evidence for the role of the General Factor of Personality (GFP) in enculturation: The GFP and self-construal in Japanese and American samples. Personality and Individual Differences, 55(4), 417–421. http://dx.doi.org/10.1016/j. paid.2013.04.002. Dunkel, C. S., & van der Linden, D. (2014). Evidence for the general factor of personality as social-effectiveness. Personality and Individual Differences, 64, 147–151. http://dx.doi. org/10.1016/j.paid.2014.02.030. Goldberg, L. R., Johnson, J. A., Eber, H. W., Hogan, R., Ashton, M. C., Cloninger, C. R., & Gough, H. G. (2006). The international personality item pool and the future of public-domain personality measures. Journal of Research in Personality, 40, 84–96. http://dx.doi.org/10.1016/j.jrp.2005.08.007. Jackson, D. N. (1984). Personality research form manual (3rd ed.). Port Huron, MI: Research Psychologists Press. Just, C. (2011). A review of literature on the general factor of personality. Personality and Individual Differences, 50, 765–771. http://dx.doi.org/10.1016/j.paid.2011.01.008. Kwan, V. S. Y., John, O. P., Robins, R. W., & Kuang, L. L. (2008). Conceptualizing and assessing self-enhancement bias: A componential approach. Journal of Personality and Social Psychology, 94, 1062–1077. http://dx.doi.org/10.1037/0022-3514.94.6.1062. Musek, J. (2007). A general factor of personality: Evidence for the Big One in the fivefactor model. Journal of Research in Personality, 41, 1213–1233. http://dx.doi.org/10. 1016/j.jrp.2007.02.003. Pettersson, E., Turkheimer, E., Horn, E. E., & Menatti, A. R. (2012). The general factor of personality and evaluation. European Journal of Personality, 26(3), 292–302. http:// dx.doi.org/10.1002/per.839. Revelle, W., & Wilt, J. (2013). The general factor of personality: A general critique. Journal of Research in Personality, 47, 493–504. http://dx.doi.org/10.1016/j.jrp.2013.04.012. Revelle, W., & Zinbarg, R. E. (2009). Coefficients alpha, beta, omega, and the glb: Comments on Sijtsma. Psychometrica, 74, 145–154. http://dx.doi.org/10.1007/S11336008-9102-Z. Riemann, R., & Kandler, C. (2010). Construct validation using multitrait-multimethodtwin data: The case of a general factor of personality. European Journal of Personality, 24, 258–277. http://dx.doi.org/10.1002/per.760. Rushton, J. P., Bons, T. A., & Hur, Y. (2008). The genetics and evolution of the general factor of personality. Journal of Research in Personality, 42, 1173–1185. http://dx.doi.org/10. 1016/j.jrp.2008.03.002. Van der Linden, D., Oostrom, J. K., Born, M. P., Van der Molen, H. T., & Serlie, A. W. (2014). Knowing what to do in social situations: The general factor of personality and performance on situational judgment tests. Journal Personnel Psychology, 13, 107–115. http://dx.doi.org/10.1027/1866-5888/a000113. Van der Linden, D., te Nijenhuis, J., & Bakker, A. B. (2010). The General Factor of Personality: A meta-analysis of Big Five intercorrelations and a criterion-related validity study. Journal of Research in Personality, 44, 315–327. http://dx.doi.org/10.1016/j.jrp. 2010.03.003. de Vries, R. E. (2011). No evidence for a General Factor of Personality in the HEXACO Personality Inventory. Journal of Research in Personality, 45, 229–232. http://dx.doi.org/ 10.1016/j.jrp.2010.12.002.