Consider anxiety. A clinical psychologist classifies it as a disorder, situating it within a diagnostic taxonomy alongside depression, PTSD, and OCD. A cognitive scientist treats it as an appraisal process—a computational state triggered by uncertainty detection. A social psychologist frames it as a culturally mediated emotional script. A neuroscientist maps it onto amygdala-prefrontal circuitry. Each classification places anxiety in a different genus, alongside different neighbours, implying different causal structures. These are not merely different labels for the same thing. They are different ontological commitments about what anxiety is.

This situation is not unique to anxiety. Across virtually every domain of psychological inquiry—memory, personality, emotion, motivation, psychopathology—distinct research traditions carve up the phenomena according to incompatible principles. The taxonomies don't merely differ in resolution or emphasis; they cross-cut each other, grouping and separating phenomena along orthogonal axes. What counts as a single category in one system fractures into several in another, and what appears unified from one theoretical vantage point dissolves into heterogeneity from the next.

Why does this happen, and what does it reveal? The persistence of cross-cutting classification in psychology is frequently treated as a temporary inconvenience—a symptom of the field's immaturity that will resolve as knowledge accumulates. But there is a more provocative possibility. Perhaps the domain itself resists the kind of univocal taxonomic structure that mature sciences are supposed to possess. If so, the failure of psychological categories to align is not an embarrassment to be overcome but a datum about the nature of mind and behaviour that demands theoretical explanation.

The most visible example of cross-cutting classification in psychology is the tension between psychiatric nosology and dimensional models of personality. The DSM carves psychopathology into discrete categories—major depressive disorder, generalized anxiety disorder, borderline personality disorder—each defined by polythetic criteria sets. Dimensional models like the Big Five or the Hierarchical Taxonomy of Psychopathology (HiTOP) instead arrange the same clinical phenomena along continuous spectra. These are not merely different levels of description. A categorical diagnosis groups together individuals who may share only a subset of features, while a dimensional profile can place diagnostically identical individuals at vastly different coordinates. The joints are located in incompatible places.

This is not confined to clinical psychology. In memory research, the distinction between episodic and semantic memory—Tulving's landmark contribution—cross-cuts the procedural-declarative divide from neuropsychology, which in turn cross-cuts the implicit-explicit distinction from cognitive psychology. Each taxonomy captures genuine regularities. Amnesic patients dissociate along one axis; priming experiments dissociate along another; neuroimaging data cluster differently still. No single classification scheme accommodates all the empirical patterns without residual anomalies.

Why is reconciliation so difficult? The core issue is that different research traditions are organised around different explanatory interests. Clinical taxonomy is optimised for treatment selection and prognosis. Cognitive taxonomy is optimised for information-processing architecture. Neuroscientific taxonomy is optimised for neural implementation. These are not competing answers to the same question—they are answers to different questions that happen to concern overlapping phenomena. As Kuhn recognised, paradigms do not merely offer different theories; they define different problems as worth solving.

The result is what philosophers of science call partial incommensurability. The categories of one system can be partially translated into those of another, but the translation is always lossy. Something is gained and something is lost in every mapping. Cognitive subtypes of depression do not map neatly onto neurobiological subtypes, which do not map neatly onto psychodynamic subtypes. Each carving captures structure that the others miss, and each obscures structure that the others reveal.

This taxonomic pluralism is frequently mistaken for mere terminological confusion—as though researchers are simply using different words for the same things. But the disagreement is deeper than language. It concerns which similarities among phenomena are theoretically significant and which are superficial. When a behaviourist groups phobias with conditioned taste aversions and a psychoanalyst groups phobias with obsessive rituals, they are not disagreeing about facts. They are disagreeing about which facts constitute the relevant basis for classification.

Takeaway

When classification systems cross-cut each other, the disagreement is rarely about facts—it is about which similarities matter, and that judgment is always relative to an explanatory purpose.

Plato's famous injunction was to carve nature at its joints—to classify in ways that reflect the real structure of the world rather than our arbitrary preferences. The success of chemistry and biology in developing unified taxonomies (the periodic table, Linnaean classification, later refined by cladistics) is often taken as evidence that mature sciences converge on a single correct way of categorising their domain. The persistent failure of psychology to achieve this convergence raises a foundational question: does the psychological domain have joints in the requisite sense?

There are reasons to think it might not—or at least, not joints of the kind that permit a single exhaustive taxonomy. Psychological phenomena are constituted at multiple levels of organisation simultaneously: molecular, neural, cognitive, behavioural, social, cultural. Unlike chemical elements, which are individuated by a single parameter (atomic number), psychological kinds are multiply realised and multiply constituted. The same computational process can be implemented in different neural architectures. The same behaviour can arise from different cognitive processes. The same subjective experience can serve different functional roles across cultural contexts.

This multi-level constitution means that there is no single dimension along which psychological phenomena can be exhaustively ordered. The domain is high-dimensional, and any classification system projects this high-dimensional space onto a lower-dimensional taxonomy. Different projections preserve different structural features—just as different map projections preserve either area, shape, or distance, but never all three simultaneously. The cross-cutting of psychological categories may therefore be an inevitable consequence of the domain's dimensionality, not a failure of scientific progress.

John Dupré's concept of promiscuous realism is instructive here. Dupré argued that even in biology, there exist multiple legitimate classification systems that cross-cut each other—taxonomies based on phylogeny, ecology, morphology, and genetics do not always converge. If this is true even for organisms with relatively stable material constitutions, it is unsurprising that psychological phenomena—which are defined relationally, functionally, and contextually—resist univocal classification even more strongly.

The implication is radical but worth entertaining seriously. If the psychological domain is genuinely high-dimensional and multiply constituted, then the search for a single correct taxonomy is not merely premature—it is misconceived. Cross-cutting classification would not be a symptom of ignorance but a structural feature of any adequate engagement with psychological reality. The question shifts from 'which classification is correct?' to 'what does each classification reveal and conceal, and under what conditions is each most useful?'

Takeaway

If psychological phenomena are constituted across multiple levels simultaneously, no single classification can capture all the relevant structure—and the search for one correct taxonomy may itself be the wrong project.

If cross-cutting classification is a durable feature of psychology rather than a temporary inconvenience, then the practical question becomes: how do we work productively within this condition? The answer requires abandoning the assumption that integration means unification—the reduction of multiple classification systems to a single master taxonomy. Instead, what is needed is a form of pragmatic integration: the disciplined use of multiple systems, with explicit awareness of what each one is optimised to capture.

The first element of pragmatic integration is purpose-relative classification. Rather than asking which taxonomy is correct simpliciter, researchers and practitioners should specify which explanatory or practical goals their classification serves. A clinician selecting a pharmacological intervention may be best served by a neurobiological taxonomy. A clinician designing a psychotherapy protocol may be better served by a cognitive-behavioural taxonomy. A researcher studying cultural variation in emotional expression may need a constructionist taxonomy. None of these is more fundamental than the others; each is more or less apt relative to a specified purpose.

The second element is translational mapping—the systematic charting of correspondences and divergences between classification systems. Where do two taxonomies agree? Where do they diverge? What phenomena fall into the gaps between them? This kind of meta-taxonomic work is unglamorous but essential. The Research Domain Criteria (RDoC) framework represents one attempt at this, mapping clinical phenomena onto neurocognitive dimensions. Its limitations are instructive: the mapping is always partial, always contested, but the exercise of mapping reveals structural features of the domain that no single taxonomy captures alone.

The third element is epistemic humility about category boundaries. When categories cross-cut, the boundaries of any given category are revealed as conventional rather than natural—useful conventions, perhaps, but conventions nonetheless. This awareness should temper the reification of psychological categories. 'Depression' is not a natural kind waiting to be discovered in its final form; it is a pragmatic grouping that serves certain purposes well and others poorly. Treating it as more than this leads to fruitless debates about whether particular cases 'really' count as depression—debates that dissolve once we recognise that the question has no answer independent of the classification system being employed.

Pragmatic integration is intellectually demanding because it requires holding multiple frameworks simultaneously without collapsing them into premature synthesis. It resists the cognitive comfort of a single authoritative taxonomy. But it is also more honest about the current state of psychological knowledge—and, if the arguments about domain structure are correct, more honest about the permanent state of any science that takes the full complexity of mind and behaviour seriously.

Takeaway

Integration without unification is not intellectual weakness—it is the appropriate epistemic stance toward a domain whose structure exceeds the capacity of any single classification to represent.

The cross-cutting of psychological categories is typically framed as a problem to be solved—evidence that the discipline has not yet matured sufficiently to achieve taxonomic consensus. But this framing smuggles in an assumption: that the domain permits a single correct decomposition. The evidence increasingly suggests otherwise.

What emerges instead is a picture of psychological reality as irreducibly multi-perspectival. Different classification systems are not competing approximations of the same hidden structure; they are complementary projections of a domain whose dimensionality exceeds what any single taxonomy can represent. This is not relativism—some classifications are demonstrably better than others for specified purposes. It is pluralism disciplined by pragmatic accountability.

The mature response to taxonomic cross-cutting is neither despair nor forced unification. It is the cultivation of taxonomic literacy—the capacity to move fluently between classification systems, to understand what each reveals, and to resist mistaking any single map for the territory it partially represents.