Philosophical Considerations Around HiTOP - Commentaries and Response
Philosophical case conference
Follow-up to:
In March 2026, I shared a summary of a new paper I had co-authored with folks from the HiTOP Revisions Workgroup. ‘Examining the Foundational Assumptions of the Hierarchical Taxonomy of Psychopathology’ was published in Philosophy, Psychiatry, & Psychology and provided a structured overview of the framework’s conceptual and philosophical underpinnings.
The paper was published as a “Philosophical Case Conference,” which means it is accompanied by 5 commentaries from authors across psychiatry, psychology, and philosophy. Our response to the commentaries was published online last week, so this is a good opportunity to bring the set to your attention. These commentaries were just the kind of conceptual and philosophical engagement we had been hoping to provoke, and we were excited to see that.
The commentaries and the response are open-access for a period of 2 weeks.
Sam Fellowes. HiTOP, Objectivity, and Logical Positivism.
Dost Öngür. HiTOP Enters Prime Time.
Nick Zautra. HiTOP 2.0 and Validity.
Miriam Solomon. A Time for Pluralism in Psychiatric Taxonomies.
S. Brian Hood. Realism Affords HiTOP Explanatory Power.
Awais Aftab, et al. (HiTOP Revisions Workgroup). Clarifying the Philosophical Foundations of HiTOP
Let me walk through the key themes of our response.
What HiTOP is and isn’t. Solomon suggests at one point in her commentary that HiTOP should be called a hierarchical taxonomy of psychology rather than psychopathology, because it models a broad range of psychological variation and lacks its own account of what makes something pathological. We push back on this suggestion. HiTOP restricts its scope to the traditional domain of what has been described by clinicians and researchers as “psychopathology.” It doesn’t include, say, the full range of personality traits in non-clinical populations or the non-clinical psychological aspects of perception, motivation, and so on. What is true is that HiTOP dimensions span from population-typical to pathological ranges, and we have not committed to a specific philosophical account of where the boundary of psychopathology lies or whether such a boundary even exists. The question of what makes something pathological is substantive and contested, involving dysfunction, distress, disability, social values, etc. However, we believe that the value of classifying clinically relevant features of psychopathology doesn’t depend on having resolved that philosophical question first.
Hierarchy and what the levels mean. Hood raises questions about the justification for HiTOP’s hierarchical arrangement, particularly the placement of symptoms, traits, and disorders at different levels. We address what we see as a fundamental misunderstanding: symptoms and traits are not placed at separate hierarchical levels (there isn’t a separate level for symptoms and a separate level for traits). At the bottom of the HiTOP hierarchy are “homogenous symptom components/maladaptive traits,” however, broadly speaking, symptoms and traits occupy every level of HiTOP. The distinction between symptoms and traits is temporal (symptoms are time-bound, traits are enduringdispositions), but any HiTOP dimension at any hierarchical level can be characterized as either symptom-based or trait-based by modifying the temporal framing of the assessment. The hierarchy itself reflects statistical relationships derived from latent variable models in cross-sectional data. Constructs at higher levels represent patterns of covariation among constructs at lower levels. This is a mathematical relationship based on covariance, not a causal or temporal one.
Latent variables and causation. Öngür notes, correctly, that we describe latent variables in statistical terms rather than as causalmechanisms andd suggested this may create tension with HiTOP’s aspirational goal of informing etiological research. The tension is indeed there. The consortium is contributing to mechanistic efforts, e.g. in the form of research on the genetic basis and neural correlates of HiTOP dimensions, but we remain cautious about baking causal interpretations into the model prematurely.
Dimensionality and pluralism. Fellowes identifies an apparent tension in our original paper: we appeal to assumptions as the basis for objectivity while also claiming that empirical evidence supports dimensionality over categorical approaches. We clarify in the response that our appeal to assumptions was in the spirit of methodological objectivity, emphasizing the theory-ladenness of data and the social dimension of knowledge production. Dimensionality, we argue in the response, is better characterized as a working assumption based on taxometric studies that fail to find discontinuities in the population-level distribution of psychopathological states and traits. The truly foundational assumption would be something more general: that we should choose the form of our variables (whether categorical or continuous) based on empirical observations of their distributional properties. We also agree with Solomon and Fellowes that categorical and dimensional frameworks can serve complementary functions, as long as we are mindful of the different notions of dimensionality at play. What we resist is the reification of DSM/ICD categories as statistically and ontologically coherent kinds when the evidence suggests they are not.
Validity and institutional process. Zautra offers a philosophical analysis of how validity functions in HiTOP and characterizes the thinking offered in our paper as “HiTOP 2.0” with relaxed ontological assumptions. From our perspective, our paper was a clarification rather than a turning point, but we are pleased that our clarifications were welcome. Zautra correctly observes that structural validity is prioritized in HiTOP’s validation process, and he notes that systematic external validation has focused primarily on the spectra level and has not yet been extended to other hierarchical levels. We acknowledge this as an active area of development. Solomon observes that both HiTOP and DSM rely on expert consensus and structured social processes, which is true, but there are also differences in the specifics of how these judgments are constrained and guided by evidence and how transparent the basis for changes is.
The paper closes by acknowledging what remains to be done: linking descriptive constructs to mechanistic understanding, more systematic validation across hierarchical levels, and stronger (though still defensible) ontological commitments about what the models represent. HiTOP does not claim to have all the answers; our collective understanding of psychopathology remains rudimentary. What we hope to offer is a framework built on systematic empirical research, explicit assumptions, and openness to revision.
See also:
