RESEARCH

Personality continues to normatively develop with age, and this development has key implications. However, personality also changes in idiosyncratic ways that differ from normative mean-level trends. This requires approaches beyond the simple linear model to capture these more complex forms of development. We apply this in our research in a few ways:

(1) Use multiple metrics to explore development. For example, people differ in their Big Five profile consistency, but previously it was unclear if this was due to intrinsic (some people are just more consistent) or extrinsic forces (environments drive consistency). In 4 datasets, we tested this with multilevel nonlinear models and found some people are stably consistent and resilient to exogenous changes; stably inconsistent and always shift in who they are; and unstably (in)consistent and susceptible to external disruptions that briefly “shock” their personality systems (Wright & Jackson, 2023a, JPSP). We also found this pattern for personality pathology (Wright, Luhmann, et al., 2026, JOPACS), suggesting heterogeneity in consistency manifests similarly for different personality constructs.

(2) Test assumptions about personality change. It is often implicitly assumed that, for everyone in a sample, their patterns of personality change adhere to the same model form, typically first-order linear or sometimes quadratic. Using multilevel and N = 1 regression models of varying forms, we found this assumption is untenable across 4 datasets. Rather, there is sizable heterogeneity in model forms for person-level Big Five trajectories (Wright & Jackson, 2024, JPSP). People’s net and total changes across time were also directly related to their model form types and were substantially miscalculated with worse-fitting forms. Thus, people vary in their forms of change and in how much they change, but this insight is missed when their changes are constrained to be equal in form.

(3) Validate new individual difference metrics. Following the prior study, we quantified a new individual difference metric to explain heterogeneity in model forms: longitudinal within-person variability (Wright & Jackson, 2025, JPSP). In 5 datasets, we explored if people differ in how much they vary around Big Five trajectories using mixed-effects location scale models as these do not assume homogeneity in residual variance (sigma). Across datasets, we indeed found people vary in how well they adhere to the sample trajectory, thus supporting the prior study and providing a metric to quantify and leverage this heterogeneity. Collectively, this suggests many studies inaccurately depict change for many people, yet use it to predict outcomes, evaluate interventions, and claim what does (not) change personality. Further, it highlights the necessity of testing assumptions and shows that person-level insight is overlooked when novel sources of heterogeneity are not explored.

It is well established that personality predicts life outcomes, but this is primarily tested for mean levels of typical constructs, in isolation of other important factors, & with self-reports of the Big Five, leading to potential biases and neglected effects. To remedy this, our research heavily focuses on the consequences of personality using different metrics, frameworks, methods, and data sources.

(1) Utility of different metrics. Personality interventions are partly motivated by the idea that changes in traits should predict the same outcomes as trait levels. However, different processes link levels & changes with outcomes, meaning we cannot assume this is true. We first tested this by examining if changes in the Big Five had similar predictive utility to its trait levels for 13 outcomes in 7 datasets (Wright & Jackson, 2023b, JPSP). Although changes prospectively predicted outcomes, the number & magnitude of effects paled in comparison to trait levels–except when also predicting changes in outcomes. We are currently testing this same idea for levels and changes in life satisfaction and affect using 16 datasets from 5 continents (Wright & Weidmann, in prep). Similarly, we extended this idea for longitudinal within-person variability, finding that sigma has predictive value over trait levels and changes (Wright & Jackson, 2025, JPSP). In this study, there was also dependence among metric-specific effects, such that their interaction gave the most accurate view of how they influence an outcome. In total, this underscored that metrics differ in their utility and in the information they capture. Thus, much like optimizing data over the lifespan, doing so with different metrics is vital to understand the consequences of personality.

(2) Utility of multiple frameworks, such as temperament (Wright & Jackson, 2022, SR), life goals (Wright et al., 2023, JRP), and personality disorders (Boone et al., 2022, BPDED). For example, in comparing the prospective utility of early-life temperament and the Big Five in adolescence/young adulthood, we found child and adult-based personality have domain-specific acuity for future outcomes, demonstrating that personality in different developmental stages is nonredundant and distinctly matters for what people later experience or achieve. 

(3) Utility of multiple report methods, such as parent (Luking et al., 2025, JAACAP; Wright & Jackson, 2024, ICD) and spousal reports (Wright et al., 2022, HP; Wright et al., 2023, JRP). For instance, we used actor-partner interdependence models and dyadic response surface analyses to test if compensatory couple effects found for the Big Five also exist for life goals and health/career outcomes (Wright et al., 2023, JRP). We found spouses’ goals predicted people’s outcomes beyond their own self-reported goals, suggesting spouses can indeed “compensate” for what their partners may lack so they still have desirable outcomes. 

(4) Utility of multiple data sources, such as physiological (Wright et al., 2022, HP) and mobile sensing (Haehner, Wright, et al., 2026, JPSP). For example, in a multi-method and multi-rater study of older adults, we used structural equation models to test two mediators–health behaviors and inflammation–of the personality-health pathway and view how self- vs. informant-reports predict health. We found informant reports often out-predicted self-reports for future health, especially through physiological pathways.