Psychiatric classification is often framed as a choice between categorical and dimensional approaches. Traditional diagnostic categories are often seen as less able to accommodate heterogeneity, graded variation and overlap, whereas dimensional models are valued for their greater flexibility in capturing severity and individual differences. This is not a new debate. Current discussions of the future DSM and of psychiatric classification more broadly have revived a longstanding tension, with a renewed emphasis on the limits of traditional descriptive categories. ^{Reference Öngür, Abi-Dargham, Clarke, Compton, Cuthbert and Fung1}

The difficulty is that this debate is often framed in overly binary and unnecessarily oppositional terms, narrowing a complex theoretical question into a misleadingly reductive choice. The implicit assumption is that greater descriptive precision reduces or eliminates the need for psychiatric categories. Yet dimensions do not inherently negate categories; rather, they often supplement and refine them. Indeed, almost all differences in nature could be said to exist dimensionally. But we distinguish between lions and tigers, and between night and day, not because there are no gradations or blurred boundaries, but because these categories capture phenomenologically distinct patterns that are both real and meaningful. We do not believe that psychiatry is any different.

Historically, the dimensional nature of some psychiatric concepts was used by antipsychiatrists to deny the existence of psychopathology. In their classic text The Reality of Mental Illness, Martin Roth and Jerome Kroll ^{Reference Roth and Kroll2} observed that many physical diseases – including epilepsy, diabetes, respiratory disease and hypertension – also lack sharp boundaries, yet this does not call their reality into question. As they note, ‘the observation that a concept has indefinite boundaries and that unequivocal judgements in all parts of the terrain to which it refers are not possible merely reflects the absence of sharp lines of demarcation in nature. Furthermore, a concept which has uncertain application in borderline cases may have a wide range of cases for which application is entirely beyond doubt’.

This point remains important for current debates in psychiatric classification. We agree and suggest that dimensional and categorical approaches are best understood as complementary rather than competing frameworks. Dimensions can capture continua of severity, trait distribution and other variations within and across disorders, thereby enhancing sensitivity to individual differences. At the same time, categories remain indispensable for identification of coherent syndromes that exhibit characteristic patterns of symptoms, course, familial aggregation (genetics) and treatment response. ^{Reference Robins and Guze3} The presence of fuzzy boundaries does not undermine the validity (reality) of such categories; it simply reflects the structure of natural phenomena.

This distinction is also methodological. Dimensional measures are often preferable for description of variation and for hypothesis testing, as dichotomisation of continuous phenomena may reduce statistical power and obscure clinically relevant gradations. But the superiority of dimensions for some research purposes does not mean that categories are dispensable in practice. Rather, it suggests that dimensional and categorical approaches serve different functions depending on the task at hand. ^{Reference Kraemer, Noda and O’Hara4} Categorical diagnoses also have consequences beyond treatment, including access to services and social protection, so where diagnostic thresholds are drawn can have practical and social consequences. These administrative and social functions should, however, be distinguished from the scientific validity of a diagnostic distinction.

One significant limitation of dimensional models of psychopathology, at least for the purposes of the present discussion, is that they treat symptoms as the primary data of psychiatry. The Hierarchical Taxonomy of Psychopathology, for instance, organises psychopathology on the basis of patterns of symptom covariance, emphasising statistical associations among features rather than clinically based identification of distinct syndromes. ^{Reference Cicero, Ruggero, Balling, Bottera, Cheli and Elkrief5} Such models are useful for capturing severity, overlap and heterogeneity. Yet classical psychiatric nosology, beginning with Kraepelin, has emphasised that mental illnesses are best understood through their course and outcome over time. ^{Reference Kraepelin6} The well-known Kraepelinian dictum that ‘diagnosis is prognosis’ reflects a methodological ideal: a valid diagnosis should carry prognostic information by identifying clinically meaningful differences in longitudinal course and outcome that extend beyond cross-sectional symptom structure.

Subsequent efforts to establish diagnostic validity in psychiatry, most notably the Robins and Guze ^{Reference Robins and Guze3} validators, have extended this longitudinal perspective. In addition to symptom clusters, they have emphasised course of illness, family aggregation, treatment response and laboratory findings as converging lines of evidence for disease entities. Although these validators have not yet yielded a complete solution to the problem of psychiatric classification, they remain essential for distinguishing clinically meaningful syndromes from descriptive groupings based solely on symptom correlations. As Guze argued, progress in psychiatric nosology requires a certain ‘toughmindedness’, ^{Reference Guze7} resisting premature theoretical enthusiasm until diagnostic constructs have been supported by converging empirical evidence. Contemporary genomic evidence demonstrates pervasive genetic overlap across psychiatric disorders, while also identifying partially differentiated genomic structures. ^{Reference Grotzinger, Werme, Peyrot, Frei, de Leeuw and Bicks8} These findings caution against the assumption that existing categories necessarily represent sharply separated biological entities, but they do not by themselves establish that categorical distinctions lack clinical or scientific validity.

The question is thus not whether dimensional approaches should inform psychiatric understanding. They should. The question is whether added dimensional precision makes categorical diagnosis dispensable, or whether diagnosis and formulation belong to different levels of clinical reasoning. The answer, we suggest, is the latter. Dimensions may improve formulation (case conceptualisation), but a clinical formulation is not a diagnosis.

Precisely because the two are not the same, the value of dimensional approaches lies elsewhere. Categorical diagnosis does not exhaust understanding of the patient, but it often provides the starting point from which more individualised formulation becomes possible. ^{Reference Weinberg and Huprich9} This is where dimensional approaches offer their most significant contribution – in clarifying aspects of the clinical picture that a categorical diagnosis alone may not fully specify. A diagnosis identifies the type of disorder or syndrome present, even when refined by severity, specifiers, course, impairment or other qualifiers. Dimensional characterisation contributes something different: it can clarify which features are most prominent, how they are configured within the presentation and how far the clinical picture overlaps with adjacent syndromes or broader dimensions of psychopathology. There is therefore no reason to assume a priori that psychopathology as a whole must be either dimensional or categorical. Whether a particular condition is better represented categorically, dimensionally or through a combination of both should depend on the available evidence.

These distinctions matter, because patients who share the same diagnosis may nonetheless differ substantially in their clinical profile. Such differences are not incidental details. They may influence disease burden, lived experience, functioning, treatment selection and the forms of support most likely to be useful. The presence of a categorical diagnosis should not, by itself, determine the intensity of intervention. Dimensional assessment of severity is particularly important in cases in which differences in symptom burden, functional impairment or risk have implications for clinical need and treatment intensity. A dimensional account can therefore make visible aspects of psychopathology that would remain obscured if the patient were to be described only in categorical terms.

For that reason, dimensional approaches are well suited to a more individualised and humanistic psychiatry. They do not replace diagnosis, but they can improve clinical formulation by linking syndrome-level recognition to the particular ways in which distress, dysfunction and vulnerability are expressed in the individual patient. In this respect, they can sharpen description, enrich formulation and support more precisely individualised treatment planning.

A central question in psychiatric nosology is whether categorical distinctions organise clinically and scientifically relevant differences. In at least some cases, they do. ^{Reference Ghaemi10} This does not imply that every current or historical psychiatric category is equally valid; categorical distinctions must remain open to empirical scrutiny and revision. For instance, lithium is an established treatment for bipolar disorder, whereas its role in schizophrenia is limited and not well supported. For this reason, reducing all distinctions to dimensional variation is not necessarily a more scientific act. In some cases, it may obscure meaningful differences by treating gradation at the margins as if it nullified distinction at the centre. Some distinctions may deserve to be retained, not only for clinical utility but because they continue to mark differences of scientific and practical importance. ^{Reference Ghaemi10}

We believe that the most defensible future direction is the integration of dimensions with categories. Future psychiatric classifications, including DSM and ICD, may rightly incorporate dimensions more fully, but they should do so in a way that preserves categorical diagnosis as a distinct clinical and scientific act. Richer formulation is a welcome advance, but it should not be confused with elimination of diagnosis itself.

In sum, the opposition between categorical and dimensional approaches reflects a false dichotomy. Psychiatry requires both categorical and dimensional approaches. Whether a condition is best understood categorically, dimensionally or through some combination of the two should be determined by the available evidence, including the extent to which dimensional assessment adds clinically useful information about severity and individual variation. The task for future classification efforts is not to choose between these approaches, but to integrate them in a way that preserves diagnostic validity while enhancing clinical understanding.