In this respect, the original clinically based Kraepelinian distinction between dementia praecox and manic-depressive illness endures in the current classification systems of psychosis. Also, despite the fact that it continues to be the focus of much criticism,¹ it still persists as the prevailing clinical model for neurobiological research.

Three main changes have notably modified the corresponding modern terms of the Kraepelinian dichotomy. First, Kraepelin's earlier unitary concept of manic depression, which included manic, circular, as well as recurrent depressive conditions, was changed to clearly differentiate bipolar from unipolar major depressive disorders.² Second, Jaspers’ hierarchical principle stating that “schizophrenic” symptoms have diagnostic prominence over “mood” symptoms for diagnostic and prognostic purposes was inverted.³ Third, avolitional and dissociative symptoms, which were described as distinctive manifestations of dementia praecox, as well as further nuclear manifestations of schizophrenia, were de-emphasized in favor of Schneiderian first-rank symptoms, which greatly influenced current consensus classifications, such as International Classification of Diseases (ICD)-10 and the third and subsequent Diagnostic and Statiscal Manual (DSM) editions.⁴ It is not unfeasible that these changes might result in an unexpected reduction of “points of rarity” among psychosis subtypes by emphasizing common instead of distinguishing features.

Little is known about why some illness concepts survive longer than others, but 2 main factors seem to be involved. First, sociopolitical factors might support an “authority” bias in the existence of particular models.⁵ Second, there may be a lack of new scientific models that are needed for a paradigm shift that could overturn the previous models. Moreover, although current diagnoses are useful concepts,⁶ mounting evidence confirms that psychiatric disorders still cannot be conceptualized as definitively valid.

Heterogeneity and pleiotropy are hallmarks of the complex nature of psychoses at the clinical level⁷ and have further confounded neurobiological research on psychosis. Elsewhere, we emphasized that “… as long as we are not able to disentangle the heterogeneity question at the clinical level, it is unlikely that heterogeneity at the etiologic and pathophysiologic levels may be resolved.”⁸

Extensive searches for clinical similarities and differences among psychosis subtypes (eg, between schizophrenia and bipolar disorders, and the categories in-between) have been carried out since the early formulations of psychosis, rendering a myriad of subtypes and classifications.¹⁰ In the case of schizophrenia disorder, at least 23 operationalized diagnoses have been reported. Four common shortcomings emerge in all classifications. First, there are no pathognomonic symptoms of schizophrenia.¹¹ Second, psychoses subtypes lack stability over time.¹² Third, a substantial proportion of patients conform to mixed or atypical groups, namely schizoaffective disorders and the “not otherwise specified” category. Finally, most classification systems, such as the DSM, are concerned with “diagnoses” but not with “disorders.”¹³ This means that current nosologies are intended more for achieving face or clinical validity than for neurobiological research.

Moreover, reexamination of patterns or chronicity within psychosis, ie, Kraepelin's “poor outcome” principle of schizophrenia, led many authors to also identify chronic deterioration in the course of bipolar disorders.¹⁴ Furthermore, a large subset of patients diagnosed with schizophrenia seemed to recover or significantly improve over the long term.¹⁵

Converging evidence from critical studies comparing categorical and dimensional models of psychosis demonstrated that symptoms and disease course, risk factors, endophenotypes, and putative neurobiological underpinnings are better explained in terms of continuous distributions.¹⁶ From a clinical perspective, studies from our own group failed to identify clear boundaries among psychoses in a large data set of psychotic patients. In fact, on the contrary, a monotonic dose-response relationship was found among the main schizophrenia-defining symptoms, main defining-associated features, and features not included in the definition for an empirically derived schizophrenia general factor.³ Moreover, psychotic symptoms did not possess a taxonic structure, which strengthens the concept of a clinical continuum among psychoses.¹⁷

Given its theoretical groundwork, great efforts have been made to address the problem of intermediate conditions in psychosis, such as schizoaffective disorder. By examining the heterogeneity underlying schizoaffective disorder in detail, we recently demonstrated that these “zones of rarity” can be distilled into a clear continuum of clinical manifestations, ranging from “nonaffective psychoses” at one extreme to “psychotic mood disorders” at the other.¹⁸

Regarding the number of putative dimensions underlying psychosis, there is some consensus that there are at least 6 symptom dimensions underlying the latent structure of psychosis: reality distortion, disorganization, negative, catatonia, mania, and depression. We also reported a hierarchical multidimensional model of psychosis comprising 10 dimensions and integrating the majority of psychopathological symptoms already present in psychosis.¹⁹ An insightful alternative paradigm currently under growing study holds the assumption that schizophrenia can be deconstructed into 3 latent constructs or “domains of psychopathology” with different etiopathogenic mechanisms and treatments.²⁰ This line of thinking led recently to the development of a consensus statement on the domain of negative symptoms within the National Institute of Mental Health-Measurement and Treatment Research to Improve Cognition in schizophrenia (NIMH-MATRICS) initiative,²¹ which clearly attempts to delineate between negative symptoms and other features of the illness and which advocates a distinct therapeutic indication area.

Setting the boundaries for psychosis is not a limiting problem in dimensional models as it does in categorical models, but whether there is a continuum from normality to psychosis is controversial. This field has made remarkable empirical progress in the last few years, identifying milder forms of expression of the core symptoms of psychosis in the general population. These milder forms are associated with the same demographic and environmental risk factors as the clinical disorder.²²

More research is needed to disentangle the underlying constructs of psychopathological symptoms, which continue to be inbred by conceptual and semiological links with “Gestalt Psychology” and outdated localizationism theories. A new psychopathology based upon current models of brain functioning, which focuses mainly on modular functioning, large-scale neurocognitive networks, and parallel processing, will be welcome. Deconstructing current complex psychopathological symptoms into a set of interacting subsymptoms and neurobiological signals by means of neuroimaging techniques would help in the development of these new fine-grained instruments. Furthermore, the scope of clinical information gathered should be widened to include not only psychopathological symptoms but also other domains of psychosis and its risk factors. There are many candidates for clinical markers (ie, alternative phenotypes), such as antecedent markers (eg, familial aggregation, obstetric abnormalities, neurodevelopmental abnormalities, premorbid functioning/personality, and precipitating factors), demographic markers (eg, age at onset, gender, ethnicity, and urbanicity), concurrent markers (eg, IQ, cognitive functioning, neurological soft signs, minor physical abnormalities, substance abuse, and community functioning), and predictive markers (eg, diagnostic consistency over time, rates of relapse and recovery, and response to treatment).

Reaching a consensus among experts about the clinical markers to be assessed would lead to delineation of the “psychosis phenome,” which has been advocated by the Human Phenome Project as an ideal strategy to deal with complex illnesses.²⁷ This strategy would be capable of generating a clinical phenome scan, which would be included in a database,²⁸ for each patient. This would be appropriate for large-scale phenotypic analysis, such as has recently been reported for bipolar disorder,²⁹ and for the advancing knowledge of the transdiscipline of phenomics in psychosis.³⁰