PUB - Publikationen an der Universität Bielefeld

This thesis is situated at the intersection of interactionist personality psychology and the increasing digitalization of everyday life. As individuals navigate both physical and digital environments that differ systematically in affordances such as temporality, visibility, persistence, and opportunities for self-presentation, this interactionist perspective raises the question of whether – and to what extent – personality traits generalize across contexts or show systematic contextual differences. Chapters 1 and 2 establish the theoretical and methodological foundation necessary for addressing this question. Chapter 1 outlines contemporary interactionist approaches to conceptualizing and integrating situations and persons – particularly the Big Five traits – and reviews prior findings on their offline-online differences. Chapter 2 discusses key limitations and methodological considerations in the operationalization of persons in situations using contextualized measures, including challenges related to single-method designs and measurement invariance. It then introduces multi-method latent state-trait models for random and fixed situations (MM-LST-RF) as a suitable, though complex, framework for disentangling stable person differences from situational variability and for modeling person-by-fixed situation interactions.

Building on this foundation, Chapter 3 formulates four overarching aims: (1) to evaluate the psychometric suitability of contextualized BFI-2 measures for cross-context comparisons, (2) to provide substantive evidence on cross-context consistencies of offline and online Big Five traits, mean-level differences, and person-by-context interactions; (3) to improve the accessibility of MM-LST-RF models; and (4) to develop a more parsimonious MM-LST-RF parametrization with comparable functionality and performance. These aims are addressed in Chapters 4 – 7, that all draw on the same underlying study design in which contextualized self- and other-rating versions of the German BFI-2 were administered twice in both offline and online contexts.

Manuscript I (Chapter 4) addresses the first aim by examining the dimensional structure, reliability, and measurement invariance of the contextualized BFI-2 across contexts and methods at both the domain and facet level. Analyses are based on data from the first measurement occasion of the overarching study design. The results indicate generally good reliabilities and largely replicable dimensional structures across contexts and raters, while also revealing systematic context- and method-specific deviations, particularly for Agreeableness and Negative Emotionality. Measurement invariance analyses demonstrate partial scalar invariance at the facet level – permitting cautious mean-level comparisons across contexts – whereas only metric invariance is supported at the domain level, allowing comparisons of associations but not means. Together, these findings establish the necessary psychometric foundation for subsequent chapters by clarifying when offline-online comparisons are meaningful and highlighting the importance of rigorous invariance testing for valid cross-context inference. Additionally, exploratory analyses predicting internet usage provide initial insights into unique and partially opposing predictive patterns of online and offline traits.

Manuscript II (Chapter 5) builds directly on this foundation and addresses the second, substantive aim of the thesis using self-rating data from the longitudinal sample analyzed with latent state-trait models for random and fixed situations (LST-RF). Results indicate substantial cross-context consistency for most Big Five facets, with approximately three quarters of trait-like variance shared between offline and online contexts, alongside meaningful context-specific variance for all traits – particularly for facets of Extraversion. Mean differences show generally lower trait levels online, most pronounced for Negative Emotionality, and person-by-fixed situation interactions reveal a buffering pattern whereby higher offline trait levels are associated with smaller offline-online discrepancies. Expanding on the exploratory analyses of Manuscript I, offline and online traits further exhibit distinct, and at times opposing, associations with different forms of internet activity. Overall, the manuscript demonstrates that personality traits are both highly stable and systematically shaped by contextual affordances of the digital world, underscoring the added value of contextualized assessment and LST-based modeling.

Manuscript III (Chapter 6) addresses the third aim of this thesis by improving the accessibility of MM-LST-RF models. Using self-ratings of positively and negatively keyed Negative Emotionality items as a motivating example, the manuscript provides a structured tutorial that introduces the MM-LST-RF framework through a modular decomposition into its multi-method (MM-LST) and random- and fixed-situation (LST-RF) components. It further clarifies model assumptions, key parameters, and the interpretation of trait, situation, and interaction effects. To facilitate model application, the manuscript also presents newly developed, user-friendly software consisting of an R function and a graphical user interface. These tools automate model specification and estimation, provide theoretically informed default settings, and offer flexible options for testing measurement invariance and equivalence assumptions. In doing so, Manuscript III lowers the methodological barrier to analyzing longitudinal multi-method, multi-situation data and supports more transparent and rigorous person-situation research.

Chapter 7 addresses the fourth aim by re-evaluating the reference-indicator parameterization used in MM-LST-RF models and introducing a more parsimonious alternative based on a reference-method approach, termed the MM(–1)-LST-RF model. Conceptually, the primary difference between the two approaches lies in the definition of method effects, which are modeled either as indicator-specific or as shared across all indicators of a method, rendering them suitable for different underlying assumptions. A comprehensive simulation study, aligned with the overarching study design, further provides an empirical comparison of the original and alternative parameterizations across varying sample sizes, numbers of measurement occasions, degrees of attrition, interaction effect sizes, and forms of model misspecification. Under correct specification, both parametrizations show comparable performance in terms of model fit, statistical power, and parameter recovery. However, the original model is more prone to convergence problems and inadmissible solutions, whereas the MM(–1)-LST-RF model is more robust overall but more sensitive to certain types of misspecifications. Taken together, Chapter 7 expands the methodological toolkit for person-situation research by clarifying the trade-offs between flexibility and parsimony and by providing evidence-based guidelines for model selection and implementation.

Chapter 8 synthesizes the central contributions and critically reflects on their limitations. In sum, this thesis advances the study of personality across physical and digital contexts by integrating theory, psychometrics, and methodological innovation. For applied research, it provides validated measurement tools, accessible modeling software, and evidence-based guidance. More broadly, it underscores both the importance of deliberate conceptualizations of personality and situations as well as the value of integrated, latent person-situation modeling approaches that can more fully capture the complexity of trait manifestations in an increasingly digitalized world.