Supplemental Materials Distinguishing Between Demoralization and Specific Personality Traits in Clinical Assessment With the NEO-PI-R by A. Noordhof et al., 2014, Psychological Assessment http://dx.doi.org/10.1037/pas0000067 Appendix A. Exploring convergence and divergence between GFP and Demoralization As described in the introduction to the main paper, Demoralization is conceptually distinct from a General Factor of Personality (GFP). GFPs are factors derived from the intercorrelations between traits. There are many different ways to derive GFPs, which may result in different conclusions regarding their importance and meaning (Hopwood, Wright, & Donnellan, 2011; Loehlin, 2012; Revelle & Wilt, 2013). GFPs have been interpreted as artifacts by some (Ashton, Lee, Goldberg, and De Vries, 2009), and as a broad overarching personality trait by others (Rushton, Bons, & Hur, 2008). Demoralization by contrast is a more specific affective concept derived from a theory of self-reported affect (Tellegen, 1985). Demoralization is hypothesized to be one possible explanation for finding substantial covariation between scales in self-report questionnaires like the NEO-PI-R, but it is not directly derived from such between-scale covariation. Empirically it can be expected that Demoralization and general factors may be completely unrelated phenomena in some questionnaires, while being isomorphic in others. For example, if a questionnaire does not include affect-related items, a large general factor may still be found but no Demoralization can be expected. On the other hand, in self-report questionnaires that specifically concern emotionality, a general factor and Demoralization will likely coincide. For self-report questionnaires, like the NEO-PI-R, that include affective and non-affective content, a more mixed result can be expected. In those cases, a general factor would likely be a broader and more heterogeneous construct encompassing many specific traits, whereas Demoralization would be a more homogeneous construct explaining covariance between some but not all subscales. We expect this latter scenario to apply to the NEO-PI-R: GFP will be a more heterogeneous construct significantly related to, but different from, the more specific concept of Demoralization. Methods These analyses were based on data from the developmental sample described in the main paper. First, we calculated Ωh for GFP. As described in footnote 1 of the main paper, Ωh is based on an exploratory factor model. In this case we specified a 5-factor model, consistent with the domain-structure of the NEO-PI-R as well as a 30-factor model, consistent with the facet-structure of the NEO-PI-R. Subsequently, we used three different ways of deriving GFP, which are all regularly used in the literature (see Loehlin, 2012). First, GFP1 was derived as a first factor in EFA on the five NEO-PI-R domain-scores. Second, GFP2 was derived as a first factor in EFA on the 30 facet-scores. Third, GFP3 was derived from a bottom-up hierarchical analysis. In this case we started with an obliquely rotated factor analysis (PROMAX rotation) on facets, selecting all factors with eigenvalue >1. The results were used to calculate regression-based factorscores for each subject. These scores were saved and subsequently factor analyzed, again using the eigenvalue >1 criterion. This procedure was followed until only one factor obtained an eigenvalue >1. Finally, we compared GFP1, GFP2 and GFP3 with Demoralization. First, we computed correlations between these four. Second, we correlated them with a sum-score of all NEO-PIR items, expecting GFPs to be more strongly associated with a sumscore. Third, we computed correlations with NEO-PI-R domain-scales to investigate convergence and divergence of these constructs with the instrument from which they were derived. Fourth, we computed correlations with MMPI-2 RC-scales in order to investigate convergences and divergences of these constructs with external correlates. Results Assuming a 5-factor model Ωh was found to be small (Ωh=.34). Introducing more factors (e.g. 30) reduced it even more (to Ωh=.28). This shows that GFP is indeed a very heterogeneous construct compared to Demoralization (Ωh=.71 as reported in the main paper). Correlations between the different GFPs, Demoralization and a sumscore of all NEO-PI-R items (NEOsum) are reported in Table A.1. As can be observed from this table, GFP1 and GFP2 correlated more strongly with NEOsum than with NEOdem. Thus, GFPs based on construction methods 1 and 2 seem more comparable to a sumscore than to Demoralization, which is consistent with regarding them as broader and more heterogeneous constructs. For GFP3 the correlational pattern was not consistent with that expectation. It was only weakly related to Demoralization and not strongly related to NEOsum either. Correlations of these GFPs and Demoralization with NEO-PI-R domain scores are reported in Table A.2. As can be observed from this table GFP1 most strongly converged with the expectation that GFP is a general factor encompassing the multiple domains of the NEO-PI-R. GFP2 showed a similar pattern but appeared somewhat more specific, as it was associated with Neuroticism, Extraversion and Conscientiousness, but only weakly with Agreeableness. For GFP3 by contrast the term General Factor of Personality does not seem fully appropriate, as it appeared as a factor strongly dominated by Extraversion (r=.91). Correlations of these GFPs and Demoralization with MMPI-2 RC scales are reported in Table A.3. Again GFP3 appeared as very different from NEOdem, while GFP1 and GFP2 showed a more comparable pattern. GFP1 appeared more strongly correlated with a number of RCscales (e.g. RC3, RC4, RC6, RC8) than NEOdem, consistent with interpreting it as a broader factor. GFP2 did not clearly show a less specific pattern of correlates with RC-scales. It did correlate more strongly with RC2, RC7 and EID and less strongly with RCd, indicating GFP2 could be tentatively interpreted as a somewhat broader Emotionality Factor. Discussion Different ways of constructing GFPs provide different results. One way of deriving GFP clearly resulted in the broad general factor we anticipated and that in our view is most consistent with the term ‘General’. It appears to be a broad and heterogeneous factor, more strongly associated with a sumscore of all NEO-PI-R items than with Demoralization. A second way of deriving GFP resulted in a factor that was strongly associated with a sumscore as well. However, correlates with NEO-PI-R and MMPI-2 RC-scales indicated that it was a more specific affect-related factor. A third way of deriving GFP resulted in a rather different factor strongly associated with Extraversion. From these analyses, we conclude that the relation between Demoralization and GFP depends importantly on the way GFP is derived and defined. Without a clear theory that specifies a priori what should be measured by GFP, it is difficult to decide which of these methods provides the most reliable and valid results. If a GFP is indeed the broad construct that is suggested by the term general, then in appears to be a rather heterogeneous factor significantly associated with Demoralization. If a GFP is a much more specific construct dominated by Extraversion-related variance, than it appears less strongly associated with Demoralization. For these reasons, we believe that a conceptually derived Demoralization factor is a better alternative to extract this clinically relevant construct compared to any of these GFP approaches. References Ashton, M.C., Lee, K., Goldberg, L.R., and De Vries, R.E. (2009) Higher Order Factors of Personality: Do They Exist? Personality and Social Psychology Review, vol. 13(2), 79-91. Hopwood, C.J., Wright, A.G.C., Donnellan, M.B. (2011) Evaluating the evidence for the general factor of personality across multiple inventories. Journal of Research in Personality, vol. 45(5), 468-478. Loehlin, J.C. (2012). How general across inventories is a general factor of personality? Journal of Research in Personality, vol. 46(3), 258-263. Revelle, W., and Wilt, J. (2013). The general factor of personality: A general critique. Journal of Research in Personality, vol. 47(5), 493-504. Rushton, J. P., Bons, T. A., & Hur, Y.M. (2008). The genetics and evolution of the general factor of personality. Journal of Research in Personality, vol. 42(5), 1173-1185. Tellegen, A. (1985). Structures of mood and personality and their relevance to assessing anxiety, with an emphasis on self-report. In A. H. Tuma & J. D. Maser (Eds.), Anxiety and the anxiety disorders (pp. 681-706). Hillsdale, NJ, England: Lawrence Erlbaum Associates, Inc. Table A.1. Pearson Correlations Between Differently Derived GFPs, Demoralization and NEOsum NEOsum NEOdem NEOdem GFP1 GFP2 -.651 GFP1 .98 -.74 GFP2 .90 -.78 .94 GFP3 .55 -.38 .52 .72 Note. N=278; NEOsum = sumscore of all items from the NEO-PI-R; NEOdem = Demoralization scale constructed in the main paper; GFP1-3= General Factor of Personality constructed following three alternative procedures. 1 All correlations were significantly different from 0 (p<.01) Table A.2. Pearson Correlations Between Different Constructions of GFP, Demoralization and NEO-PIR Domain Scores With 99% Confidence Intervals N E O A C NEOdem .811 -.42 .10ns -.14 -.40 GFP1 -.79 .66 .23 .46 .52 GFP2 -.79 .77 .21 .19 .43 GFP3 -.30 .91 .48 -.24 -.02ns ns Not significant (p>.01) Note. N=Neuroticism, E=Extraversion, O=Openness, A=Agreeableness, C=Conscientiousness, NEOdem = Demoralization scale constructed in the main paper; GFP1-3= General Factor of Personality constructed following three alternative procedures. 1 All correlations were significantly different from 0 (p<.01), except for the two values indicated with ns. Table A.3. Correlations Between Different Constructions of GFP, Demoralization and MMPI-2 RCScales With 99% Confidence Intervals NEOdem1 GFP1 GFP2 GFP3 EID .65 -.66 -.69 -.40 THD .12 -.19 -.15 .05 BXD .12 -.19 -.05 .36 Rcd .66 -.60 -.60 -.26 RC1 .27 -.25 -.25 -.07 RC2 .42 -.49 -.58 -.56 RC3 .18 -.34 -.21 .09 RC4 .18 -.27 -.18 .19 RC6 .17 -.30 -.24 .01 RC7 .43 -.54 -.51 -.19 RC8 .14 -.20 -.16 .05 RC9 .07 -.07 .11 .53 Note. Critical values for significant correlations (p<.01) are r<-.16 and r>.16; N=278; NEOdem = Demoralization scale constructed in the main paper; GFP1-3= General Factor of Personality constructed following three alternative procedures; EID-RC9 = MMPI-2-RF scales; EID = Emotional/Internalizing Dysfunction, RCd = Demoralization, RC1-RC9 = Revised Clinical Scales, BXD = Behavioral/Externalizing Dysfunction, THD = Thought Dysfunction.