An open invitation to personality researchers
Personality psychology has produced numerous taxonomies of personality traits, but we still lack a widely accepted hierarchical measurement model that specifies how narrower traits relate to broader personality dimensions.
Although the Big Five are widely accepted as a descriptive framework, they lack a consensual measurement model. Moreover, other models like the HEXACO model exist without any attempt to create a unified model. This means that personality psychology still lacks a scientific taxonomy of personality traits. Instead, different researchers use different inventories, item selections, and statistical models, often producing different conclusions about personality structure.
How can we establish a common empirical foundation?
Like all other sciences, we have to start with a broad – not necessarily representative or exhaustive – list of objects that we want to classify. For personality psychologists these objects are specific personality traits. The key information for a classification system is how specific traits are related to each other. A key problem is that the standard method used by personality psychologists (EFA, PCA) is limited and produces new structures with every new set of traits. The solution to this problem is to use hierarchical structural equation modeling (or Hierarchical Factor Analysis, HFA) that imposes a theoretical structure on the data that does not change with the item set and correlations, but tests whether the data are consistent with the theoretical model.
I am inviting personality researchers, psychometricians, graduate students, and anyone interested in personality structure to participate in an open, collaborative many-analyst project.
This is a good time for such a project because data are openly available, software is free, and learning HFA is easier than ever with the help of AI. AI can even fit models for you. So, even personality scientists who are not familiar with structural equation models can formulate models and test them.
Participation does not require funding, new data collection, or specialized software. All you need is a computer. Call it theoretical or computational personality science.
The starting point for this project are data from the FacetMAP dataset (Irwing et al., 2024) that contains correlations for 77 facets, each measured with six items.
I created a GitHub repository that contains the data, syntax files for MPLUS and lavaan, and a model with five independent factors (the classic Big Five model). The model has a well fitting measurement model for the facets. Thus, researchers can focus on alternative models of the structural model that relates the facets to each other. However, modification to the measurement model are also possible. Starting with this model can save a lot of time and prevent convergence problems, but experienced users are free to start from scratch.
The immediate goal is to develop and evaluate measurement models that can help us answer several fundamental questions:
- Which facets can be measured reliably and distinctly?
- How are narrow personality traits related to one another?
- Can these relationships be explained by broader personality factors?
- How much of the shared variance reflects substantive personality characteristics versus method variance?
- Which structural conclusions remain robust across alternative, theoretically defensible measurement models?
Researchers are invited to contribute their own analytical approaches, challenge existing specifications, and propose alternative models.
Rather than trying to establish which researcher or existing taxonomy is correct, we want to identify results that withstand independent scrutiny and develop a measurement model that researchers with different theoretical perspectives can agree on.
All analyses, code, and findings can be shared openly, including results that challenge preferred theoretical models.
The FacetMAP dataset is only the beginning. Once we have established promising models, we can test whether the same structure emerges in other datasets and item pools, including large collections of personality-descriptive adjectives.
The ultimate goal is not another personality inventory. It is a cumulative, empirically grounded understanding of the hierarchical structure of personality traits.
Researchers interested in contributing are invited to explore the GitHub repository, run the existing models, and propose new analyses.
Ulrich Schimmack
University of Toronto Mississauga