How to (not) disentangle personality constructs: scrutinizing longitudinal, multi-rater, and behavioral genetic approaches

To disentangle the characteristic patterns that shape the uniqueness of individuals, personality is often regarded as a dynamic system composed of some more basic, stable, and consistent person-only characteristics as well as more specific and environmentally malleable person-in-environment characteristics. Prior research proposed testable criteria for examining the extent to which personality constructs might rather reflect person-only or person-in-environment characteristics. These criteria involve various empirical properties of constructs, such as their stability, observability, or heritability. Appropriate examinations hereof require comprehensive empirical data, encompassing longitudinal data from multiple measurement occasions, multi-rater data from different rater perspectives, and behavioral genetic data from twins and their family members. In the current work, I scrutinize selected methodological approaches using these three data sources in terms of their benefits and pitfalls and thus the conclusions they allow to be drawn (or not). To this end, I focus on the structural equation models that were applied within the three empirical research papers of the present dissertation, and illustrate their application by the aim of disentangling whether the potential person-only characteristic emotional stability differs from the three self-related schemata self-esteem, self-efficacy, and internal locus of control. It crystallizes that the benefits and pitfalls of any methodological approach also depend on the nature of the constructs examined. Longitudinal approaches allow to decompose variance in state measures into stable and state-specific components. Both components, however, might be more or less informative for more or less stable characteristics. Associations between constructs, as examined in cross-lagged panel models, could thus differ depending on the part of variance between which the links are estimated. Multi-rater approaches allow to decompose variance in personality measures into intersubjectively objective and rater-specific components. However, intersubjectively objective components might represent valid variance in easily observable characteristics, while rater-specific components might contain relevant and valid information on more contextualized characteristics. Overlaps of intersubjectively objective components between constructs, as examined in multitrait-multimethod models, could thus question distinctiveness or hint that constructs overlap in a specific rater-consistent part. Behavioral genetic approaches allow variance decomposition in genetic and environmental parts, whereas these components might be estimated differently based on classical twin or nuclear twin family models. Besides, genetic variance in more contextualized characteristics may be explained by more basic ones, making sole heritability estimates only conditionally insightful. Examining constructs in terms of their unique and common heritability, using common pathway models, may thus be additionally relevant. Overall, it proves essential that variance in personality measures results from an interplay of stable and state-specific, construct- and rater-specific, as well as genetic and environmental variance. Combining multiple perspectives on the sources of individual differences could serve to examine and disentangle the properties of constructs more accurately. The current dissertation therefore reveals the need for more integrative approaches using longitudinal, multi-rater, and behavioral genetic data within future research into personality differences.