20 Novembre 2026

Psicologia Punto One | Psicologia e Psicologi nel Web

Menu

Menu

SCHEDA SCIENTIFICA2026Altra ricerca

Covariation among higher-order psychopathology dimensions is driven by nomorbidity and not comorbidity.

Watts AL, Rosen AFG, Greene AL, King KM, Meyer FAC, Trull TJ, Sher KJ, Joyner KJ

Nature. Mental health · 2026

Abstract originale

Emerging classification frameworks claim that covariation among psychopathology dimensions reflects comorbidity. We scrutinized this claim by applying Bayesian change point models to four waves of the Adolescent Brain Cognitive Development Study data (N = 11,868 youths; 48% female) and examined associations among higher-order psychopathology dimensions. Covariation between psychopathology dimensions was driven by nomorbidity, not comorbidity: we observed strong positive associations among psychopathology dimensions at relatively extreme low (asymptomatic) levels (for example, when three to five symptoms were endorsed) and weak positive associations at higher levels. For instance, at wave 1, a 1-point increase in internalizing was associated with a 32% increase in externalizing at nearly asymptomatic levels (B = 0.28), compared with only a 7% increase at higher levels (B = 0.07). These findings challenge existing assumptions about comorbidity in nonclinical samples, suggesting that covariation is more accurately explained by the absence of symptomatology rather than shared psychiatric conditions.

Parole chiave

PsychologySigns and symptomsSigns And Symptoms
Leggi il testo integrale originale

Testo estratto dalla pubblicazione: sono esclusi immagini, media, supplementi e parti soggette a diritti distinti.

Covariation among higher-order psychopathology dimensions is driven by nomorbidity and not comorbidity

Emerging classification frameworks claim that covariation among psychopathology dimensions reflects comorbidity. We scrutinized this claim by applying Bayesian change point models to four waves of the Adolescent Brain Cognitive Development Study data (N = 11,868 youths; 48% female) and examined associations among higher-order psychopathology dimensions. Covariation between psychopathology dimensions was driven by nomorbidity, not comorbidity: we observed strong positive associations among psychopathology dimensions at relatively extreme low (asymptomatic) levels (for example, when three to five symptoms were endorsed) and weak positive associations at higher levels. For instance, at wave 1, a 1-point increase in internalizing was associated with a 32% increase in externalizing at nearly asymptomatic levels (B = 0.28), compared with only a 7% increase at higher levels (B = 0.07). These findings challenge existing assumptions about comorbidity in nonclinical samples, suggesting that covariation is more accurately explained by the absence of symptomatology rather than shared psychiatric conditions.

This research utilized Bayesian change point models on Adolescent Brain Cognitive Development Study data to reveal that psychopathology dimension covariation stems from nomorbidity, indicating stronger associations at low symptom levels compared with higher symptomatology, challenging traditional comorbidity assumptions.

Main

Psychiatric comorbidity, the co-occurrence of two or more mental disorders, is often regarded as the ‘rule rather than the exception’1,2. But epidemiologic data suggest that it is statistically uncommon to be diagnosed with two or more conditions, let alone three or more, at the same time3. At a given time, most people in the general population do not exhibit appreciable levels of symptomatology to warrant a psychiatric diagnosis. Around 70% of the population exhibits what we term ‘nomorbidity’—wherein an individual is not diagnosed with a single mental disorder—when past-year diagnoses are assessed, and 50% when lifetime diagnoses are assessed3. Further, when followed over long stretches of time, many people (~17–40%) are not diagnosed with a single mental disorder at any time4,5. Of course, conditional upon receipt of one diagnosis, comorbidity is far more common. If a person receives one diagnosis, their odds of receiving two diagnoses increases, and so on.

Modern classification systems, such as the Hierarchical Taxonomy of Psychopathology6, argue that this ‘rampant comorbidity problem’7,8 results from the fact that the Diagnostic and Statistical Manual of Mental Disorders and related systems ‘do not carve nature at its joints’9–11. According to this criticism, comorbidity is largely conceivably artefactual, arising because we draw arbitrary borders around phenomena that overlap phenotypically and etiologically6,12. In response, these systems harness the supposed overlap among putatively distinct conditions using covariance-based modeling and reorganize psychopathology into a more parsimonious set of dimensions. Products of covariance-based models have, in turn, been widely promoted as capturing sources of psychiatric comorbidity13,14.

Across nearly half of a century and thousands of studies, these efforts have converged upon two broad psychopathology dimensions: externalizing and internalizing6,15–17, which were first studied in youth15 nearly thirty or so years before they were studied in adults17. Externalizing encompasses conditions of poor behavioral and emotional control18 (for example, conduct disorder, oppositional defiant disorder and substance use disorder), whereas internalizing encompasses conditions that are characterized by intense negative emotions19 (for example, generalized anxiety disorder, major depressive disorder and specific phobia). The idea behind these transdiagnostic dimensions is that, say, conduct disorder and substance use disorder exhibit high degrees of comorbidity because they are phenotypically and etiologically overlapping, with disinhibition explaining their overlap18,20.

On the basis of the correlation between externalizing and internalizing (r ≈ 0.5–0.6), various researchers further proposed that a single dimension21–23—the p-factor—is the etiologic source of psychiatric comorbidity: ‘a single dimension is able to measure and maybe even explain a person’s liability to mental disorder, comorbidity among disorders, persistence of disorders over time, and severity of symptoms’ (pp. 83121; see also refs. 24–29). This claim positively asserts that the covariation between externalizing and internalizing reflects some shared substantive mechanism (for example, poor emotional control30,31 and negative emotionality32,33).

We propose an alternative explanation, that evidence of so-called comorbidity arises owing to the considerable extent of asymptomatic cases in the population. That is, covariation among forms of psychopathology is driven by the inclusion of people who report little to no psychopathology, a typical presentation in large-scale epidemiologic and otherwise community-based samples that are most often used to model hierarchical structures of psychopathology18,19.

As the basis of this claim, we borrow from a larger literature that suggests that the inclusion of asymptomatic (or undiagnosed) cases in covariance-based analyses of the structure of psychological constructs encourages their unidimensionality34, or inflates the overlap among the phenomena of interest. Including asymptomatic cases produces dense network35 structures and coherent (unidimensional) factor34,36 structures that—although people assume they capture comorbidity—are driven by people who endorse few, if any, symptoms37. That is, what might glue together symptoms is a lack of comorbidity, or nomorbidity (if you find this claim counterintuitive, we expand on our argument in Appendix A).

We extend these findings to examine associations among higher-order psychopathology dimensions, whose covariation is interpreted as being synonymous with psychiatric comorbidity, using innovative statistical modeling techniques. Here, we focus on youth who took part in the Adolescent Brain Cognitive Development Study (ABCD)38 given that externalizing and internalizing were first observed among youth15 and because epidemiologic data are often used to characterize the structure of psychopathology6. The large size of the ABCD Study affords one of the best powered examinations of the interface between higher-order psychopathology dimensions in youth.

Results

Univariate and bivariate normality

At wave 1, 81% of the sample had nonclinical levels of internalizing and externalizing and only 5.3% of the sample exhibited some combination of at-risk or clinical levels of both internalizing and externalizing (Fig. 1a). All psychopathology dimensions were highly skewed, with a greater weight in the left tail than is expected in a normal distribution (Table 1; see Supplementary Fig. B8 and Supplementary Tables B4–6 for all other waves), kurtotic and not bivariate normal.Table 1Univariate and bivariate normality for higher-order psychopathology dimensions at wave one

Predictor Univariate normality Bivariate normality

Skewness Kurtosis Criterion Skewness Kurtosis

Externalizing 2.30 7.02 Internalizing 18,326 203

Attention problems 15,085 152

Thought problems 19,115 211

Internalizing 1.94 5.12 Externalizing 18,326 203

Attention problems 11,309 116

Thought problems 16,804 195

Attention problems 1.44 1.83 Externalizing 15,085 152

Internalizing 11,309 116

Thought problems 13,436 137

Thought problems 2.26 6.88 Externalizing 19,115 211

Internalizing 16,804 195

Attention problems 13,436 137

Univariate skewness values >1 or <−1 indicate a high degree of skewness, and univariate kurtosis values >2 or <−2 indicate substantial nonnormality. Higher estimates of Mardia bivariate skewness and kurtosis indicate deviations from bivariate normality, with 0 indicating true multivariate normality. Bonferroni-corrected two-tailed P values associated with the Mardia statistics were significant at P ≈ 2.220446 × 10−16, indicating that the data were significantly deviant from bivariate normality.

Associations between externalizing and internalizing

When we regressed externalizing onto internalizing, the change point occurred when internalizing was equal to a value of 4.85, corresponding to endorsement of approximately two to three internalizing symptoms. The pre-change point slope was strong (B1 = 0.28), with a 1-point increase in internalizing indicating a 32% increase in externalizing (exp[0.28] of 1.32) (Fig. 2a,b). By contrast, the post-change point slope was relatively weak (B2 = 0.07) (Fig. 2b), with a 1-point increase in internalizing only indicating a 7% increase in externalizing. The relative rate of the pre-change point slope was 4.6-times larger than that of the post-change point slope (32/7 = 4.6), suggesting far stronger slopes among largely asymptomatic cases who endorse fewer than three symptoms. Thus, nomorbidity, rather than comorbidity, drove the covariance commonly reported in literature.

In contrast with the one-change point model, the zero-change point model (that is, the ordinary least squares (OLS) estimate with a negative binomial link function) was associated with considerable prediction error (Fig. 3). The OLS estimate resulted in substantial prediction error, such that it overpredicted the association between externalizing and internalizing at nonclinical levels of internalizing (per the T scores), and particularly at symptomatic and clinical levels of internalizing.

We also considered two robustness checks: (1) the ordering of the predictor and criterion variables, and (2) replication across development. First, these patterns held when we regressed internalizing onto externalizing as opposed to the other way around (Fig. 2c). The change point occurred when externalizing was equal to a value of 2.5, corresponding to endorsement of one to two symptoms. Again, the pre-change point slope (B1 = 0.36) was far stronger than the post-change point slope (B2 = 0.06), with the former slope indicating that a 1-point change in externalizing was associated with a 43% increase in internalizing, and the latter indicating a 6% increase; the pre-change point slope was seven-times stronger than the post-change point slope.

Second, these patterns replicated across four waves (Fig. 2d). When we regressed externalizing onto internalizing across each wave, the location of the change points observed shifted slightly but not substantially across waves, with all change points being located between internalizing scores of 3 and 5 (corresponding to endorsement of one to five symptoms) and decreasing over time (τ1 = 4.85, τ2 = 4.61, τ3 = 3.57 and τ4 = 3.53). The pre- and post-change point slopes were consistent across waves (B1 = 0.28 (wave 1 and 2), 0.38 (wave 3 and 4) and B2 = 0.07 (wave 1–4)), with all waves suggesting much stronger slopes at low levels of internalizing. At waves 3 and 4, the pre-change point slopes were even stronger than previous waves, with a 1-point change in internalizing associated with a 46% increase in externalizing. These findings suggest that covariance among psychopathology dimensions may be increasingly driven by asymptomatic cases across development.

Associations among other higher-order psychopathology dimensions

We also considered externalizing and internalizing’s associations with other higher-order psychopathology dimensions: thought problems and attention problems (for wave 1 see Table 2 and Fig. 4; see Supplementary Tables B7–B11 for findings from all other waves). We observed the same trend across all models, whereby pre-change point slopes were far stronger than post-change point slopes, including when we regressed externalizing onto attention problems (B1 = 0.76, B2 = 0.15), externalizing onto thought problems (B1 = 0.54, B2 = 0.15), internalizing onto attention problems (B1= 0.25, B2 = 0.11), internalizing onto thought problems (B1 = 0.47, B2 = 0.13) and attention problems onto thought problems (B1 = 0.52, B2 = 0.12).Table 2Results from the one-change point negative binomial regressions for all higher-order psychopathology dimensions at wave 1

Predictor variable Criterion variable Change point location Pre-change point slope Post-change point slope Ratio of slopes

τ SE \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{{\boldsymbol{R}}}$$\end{document}R^ B1 SE (%) exp(B1) \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{{\boldsymbol{R}}}$$\end{document}R^ B2 SE (%) exp(B2) \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{{\boldsymbol{R}}}$$\end{document}R^

Internalizing Externalizing 4.85 0.07 1.04 0.28 0.00 32 1.05 0.07 0.00 7 1.03 4.5

Attention problems 3.82 0.05 1.02 0.27 0.00 31 1.03 0.06 0.00 6 1.01 5.0

Thought problems 5.26 0.06 1.04 0.27 0.00 31 1.03 0.07 0.00 7 1.03 4.3

Externalizing Internalizing 2.52 0.01 1.00 0.36 0.00 43 1.00 0.06 0.00 6 1.00 7.0

Attention problems 3.48 0.02 1.01 0.37 0.00 45 1.02 0.06 0.00 6 1.01 7.2

Thought problems 5.15 0.09 1.07 0.24 0.00 27 1.06 0.06 0.00 6 1.04 4.4

Attention problems Externalizing 1.50 0.00 1.00 0.76 0.00 114 1.00 0.15 0.00 16 1.00 7.0

Internalizing 1.07 0.05 1.02 0.25 0.05 28 1.03 0.11 0.00 12 1.02 2.4

Thought problems 1.51 0.01 1.00 0.54 0.00 72 1.00 0.15 0.00 16 1.00 4.4

Thought problems Externalizing 2.51 0.01 1.00 0.52 0.00 68 1.00 0.15 0.00 16 1.00 4.2

Internalizing 2.50 0.00 1.00 0.47 0.00 60 1.00 0.13 0.00 14 1.00 4.3

Attention problems 2.51 0.01 1.00 0.52 0.00 68 1.01 0.12 0.00 13 1.01 5.3

B, unstandardized beta (slope) coefficients; exp(B1), exponentiated beta coefficient; \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{{R}}$$\end{document}Rˆ, scale reduction factor (convergence criterion); SE, standard error; τ, the change point location. The exponentiated beta coefficients reflect the percent increase in the criterion variable per point of the predictor variable; for instance, when we regressed externalizing onto internalizing, B1 = 0.28, so exp(B1) = exp(0.28) = 1.32, or a 32% increase. The ratios of slopes are computed as follows: % exp(B1) / % exp(B2). Table C1 contains a translation between raw scores and T scores for all scales across all waves.

Discussion

Researchers commonly make the inference that the correlation between two higher-order psychopathology dimensions captures comorbidity16,21,39. Nevertheless, we provide empirical evidence here that covariation among psychopathology dimensions is largely driven by nomorbidity, or a lack of comorbidity. Associations between psychopathology dimensions were strong and positive among asymptomatic people and substantially weaker among people who experience more than a handful of symptoms with relative frequency. That is, psychopathology dimensions were more differentiated (or less overlapping) at higher severity levels, to the extent that associations were negligible at moderate to clinically severe levels of psychopathology. Thus, covariance among psychopathology dimensions does not necessarily reflect or capture comorbidity16,21,40. Our findings underscore important limitations of covariance-based analyses for studying comorbidity, the p-factor41–44 and the latent structure of psychopathology more generally.

Covariance-based similarity measures are dramatically influenced by the distributional qualities (for example, base rates and skewness) of the variables of interest45, an issue that has been better appreciated in other fields (for example, ecology, computer science and engineering). Ecology, for instance, relies on association measures that exclude the mutual absences of rare species’ locations46 for the very reason we demonstrate here: traditional association measures (such as the correlation or factor analysis34) erroneously indicate that two species, or forms of psychopathology, are highly related simply because they do not co-occur often. The use of association measures that are driven by the absence of relatedness is particularly relevant to textbook definitions of comorbidity, which specifically emphasize a shared presence—rather than shared absence—of two or more disorders47. Covariance-based analyses, which include factor analyses of psychopathology data6,16,48, in the general population may largely characterize the absence of different forms of psychopathology rather than their co-occurrence (or comorbidity).

A construct whose statistical and ontological status is particularly challenged by our findings is the p-factor, a dimension that is claimed to give rise to all forms of psychopathology and explains why different forms of psychopathology are associated with each other21,22. So it goes, the higher one’s p-factor score, the higher propensity to develop several forms of psychopathology21,22. Critically, our findings suggest that the p-factor appears largely irrelevant to clinical levels of psychopathology and the subpopulation of people who experience psychopathology in some appreciable form, which is inconsistent with the notion that it is a unitary dimension on which all humans can be placed21,22.

Common interpretations of the p-factor as the source of psychiatric comorbidity thus appear to rest on a faulty inference because the p-factor may not reflect comorbidity, but rather, an absence of clinically significant psychopathology. In this way, the p-factor appears more relevant to low as opposed to moderate or high levels of psychopathology. Although this conclusion may appear overly conjectural, related literature supports it. According to Spearman’s law of diminishing returns (also termed ‘cognitive ability differentiation’), the general factor of intelligence (or g)49, the supposed theoretical predecessor to the p-factor21,50, contains more explanatory power at lower as opposed to higher levels of intelligence51–54.

Our findings also have implications for structural models of psychopathology that form the bedrock of modern classification efforts6. Such models, which typically rely on the general linear model, assume rather than test linear (or additive) relationships among psychopathology variables. Factor analysis, which is most commonly used to construct taxonomic frameworks6,55, actually makes even stronger assumptions about linearity—that a latent normal distribution underlies the data56, that the statistical relationship between a given item and the latent factor is linear, and that the correlations among latent factors are linear6,57–59. Even though factor analysis assumes an underlying normal distribution, estimated factor scores are not required to be normally distributed. In fact, in these data, factor scores of psychopathology dimensions were highly skewed and kurtotic, often even more so than the sum scores (Supplementary Fig. B9 and Supplementary Tables B12–15), and assumptions of linearity between psychopathology dimensions were violated in the data considered here (Supplementary Figs. B9 and B10).

Given apparent violations of linearity, when researchers implicitly or explicitly adopt a model with one slope in the population (that is, a zero-change point model), the slopes simply capture the approximate average of the two slopes estimated in a one-change point model (Appendix D). Although it may be tempting to think that the ‘average’ is a sufficient guess for the typical person in the population, that average has no psychological meaning when there are substantially different slopes in the population. Such an estimate is not only an oversimplification of the true population relationship, but a biased estimate of it56,60–64. Specifically, we showed that these estimates are upwardly biased, over characterizing the statistical association between higher-order psychopathology dimensions at nonclinical levels of psychopathology, and particularly at clinical levels. Because the assumptions of the linear model were extremely violated in these data (Supplementary Figs. A9 and A10), we conducted a simulation study (Appendix D) to show that the linear model produces a biased, and often extremely misleading, estimate of the population association between two psychopathology dimensions in the presence of nonlinearity under many conditions.

Ultimately, there may be no single, ‘true’ structure of psychopathology in the population65,66. Instead, the assumption of population homogeneity may be inappropriate66, and there may be discrete subpopulations of people that are differentially susceptible to psychopathology, have differential comorbidity patterns and, in turn, have dramatically different structures of psychopathology. Related literature supports such a possibility: self-esteem appears to have different factor structures—one unidimensional and one multidimensional—depending on the subpopulation of interest67. Further, the structure of intelligence appears moderated by ability level, with the strength of correlations among intelligence tasks decreasing as intelligence levels increase68–70.

With the considerable degree of asymptomatic cases in most large-scale epidemiologic samples in mind, it may be the case that there is a large class with a relative absence of diagnosed psychopathology for whom psychopathology appears to conform to a unidimensional solution35,36,71, and one or more relatively symptomatic classes for which psychopathology appears multidimensional. We suspect that there is a subpopulation that comprises people with an exceptionally high threshold to experience and/or accurately report psychopathology, whose detection is obscured—even precluded—in most modern classification efforts72 owing to the widespread assumption of population homogeneity73.

As the framers of the Diagnostic and Statistical Manual of Mental Disorders (DSM-6) are charged with capturing the continuum (or dimensionality) of mental health, and to explicitly consider severity and transdiagnostic features74, our findings are highly relevant to current efforts to incorporate dimensionality into future psychiatric nosology9,14. In particular, our findings suggest that dimensional structures of psychopathology should be evaluated in relation to their target population and intended clinical use75. A structure estimated from covariance in a community sample may be useful for population screening or describing overall symptom burden, but it should not automatically be interpreted as evidence of clinical comorbidity, shared etiology or treatment-relevant co-occurrence among symptomatic patients. In community samples overrepresented by asymptomatic cases, global covariance may primarily reflect shared absence of symptoms. Thus, studies used to support dimensional nosology should report the distributional qualities of the data (for example, skewness and kurtosis), confirm that assumptions of the general linear model are met in their data, examine nonlinear associations among forms of psychopathology and evaluate whether broad structures remain informative in the clinical range. Our data suggest that the general linear model results in overprediction of covariance (or so-called comorbidity) among psychopathology dimensions at clinically severe levels of psychopathology.

Given the strong implications of our findings, it is worth discussing several potential limitations or criticisms of them. First, because we examined nonlinear relationships among higher-order psychopathology dimensions, skeptics of our findings might argue that we have invoked Berkson’s bias76 or the like77,78 by restricting the sample post-change points to reflect higher degrees of severity (that is, conditioning on severity), thereby attenuating the associations between psychopathology dimensions at higher severity levels. That argument holds only if there is a single linear relationship between two psychopathology dimensions of interest in the population, or an absence of nonlinearity and population heterogeneity73 (Appendix D). It is also worth noting that our change points were consistently located at extremely mild levels of psychopathology (for example, when two or three symptoms were endorsed), meaning that our results are not artifacts of restricting the sample to high severity levels.

Second, because we allowed for a change point in the associations between higher-order psychopathology dimensions, skeptics might argue that we artificially produced different slopes and, in turn, evidence of nonlinearity. Change point models do not inherently produce evidence of nonlinearity. Indeed, our simulation study (Appendix E) demonstrates that the change point model will generate estimates identical to the linear model when there is no evidence of nonlinearity. We also observed evidence of this fact in our analysis of psychopathology data. For instance, in the Gaussian one-change point model that regressed externalizing onto internalizing (Appendix C), the pre- and post-change point slopes were almost identical to each other, and to the slope estimated in the zero-change point model. Thus, it is not always the case that modeling a change point will engender evidence of nonlinearity.

Third, we relied on sum scores of youth psychopathology data. In sensitivity analyses, we examined caregivers’ psychopathology data from the Adult Self-Report, the adult counterpart to the Child Behavior Checklist (CBCL), and our findings mirrored those reported here (Supplementary Table C9). Change point models indicated a preference for one (as opposed to zero or two) change points, and the pre-change point slopes were considerably stronger than the post-change point slopes. It is perhaps unsurprising that our findings replicated among adults because the adult data in the ABCD Study were highly skewed and kurtotic (Supplementary Fig. B11 and Supplementary Tables B16 and B17). As demonstrated in our simulation study (Appendix E), any data are liable to the problems we demonstrated here: the associations among higher-order psychopathology dimensions—in youths and adults—are driven up with the inclusion of asymptomatic cases.

With that said, here, we focused only on the associations among higher-order psychopathology dimensions and did not conduct factor analyses of psychopathology, which is typical in classification efforts (see Appendix C for subscale-level analyses). As yet, there is no implementation of the Bayesian change point model for factor analysis. Still, the sum score is a special case of a latent variable79, so our findings bear directly on factor models and the interpretation of correlations among higher-order psychopathology factors as evidence of comorbidity.

Fourth, given the highly positively skewed nature of our psychopathology dimensions, the ABCD data may be an inadequate representation of psychopathology at clinically severe levels. We offer several counterpoints to this potential conclusion. Although nonclinical cases reflected the plurality of data in the ABCD Study (Fig. 1), we had an appreciable amount of clinically severe levels of externalizing and internalizing, with 2.2% of youths (n = 257) falling into this category. Moreover, when we forced two-change point models, we did not find a change point at severe levels of psychopathology dimensions and did not find evidence of increasingly strong associations at severe levels (Appendix C). In one instance, when we regressed externalizing onto internalizing in the adult data, we detected a second change point at a score of 30 (Supplementary Table C9). Still, the three slopes detected in this model suggested a decreasing association between externalizing and internalizing as a function of severity (B1 = 0.29, B2 = 0.12 and B3 = 0.04).

Along those lines, we used an alternate estimation method, generalized additive multilevel models (Appendix C), that may be more sensitive to the upper tail of the distribution, and still found no support for this hypothesis. Thus, we are not convinced that our two-change point models were preordained to reject the notion that higher-order psychopathology dimensions become more strongly associated at severe levels21. Instead, there was no uptick for the third estimated slope, ruling out the possibility that comorbidity ‘emerges’ at severe levels of psychopathology.

We welcome investigations of nonlinearity among samples with more representation of severe levels of psychopathology, such as in clinical or mixed community–clinical samples, and differing assessments of psychopathology, including ones that assess finer-grained phenomena than we consider here. Although not of focus here, it is possible that the dimensionality of narrower phenomena is also driven by asymptomatic cases, and that higher-order psychopathology dimensions may appear more multidimensional at higher levels of psychopathology. For instance, perhaps depression and anxiety are most overlapping at low levels and less overlapping at higher levels (Appendix C).

Mitigating potential generalizability concerns, a recent meta-analysis80 indicates that there is strong reason to believe that higher-order psychopathology dimensions become more weakly—as opposed to more strongly—associated at higher levels of severity. The correlation between externalizing and internalizing is dramatically stronger in nonenriched (or more nomorbid) samples (Φ = 0.48) compared with psychopathology enriched samples (or samples with more comorbidity, Φ = 0.20). This pattern holds for assessment strategies that manipulate the extent of positive skew or nomorbidity in the data35,36,71; the correlation between externalizing and internalizing was much stronger when current diagnoses (Φ = 0.37), as opposed to lifetime diagnoses (Φ = 0.19), were modeled, and it was also much stronger when studies did not rely on hierarchical exclusion (or ‘skip-out’) rules (Φ = 0.41) compared with those that did (Φ = 0.24)81. Compared with their counterparts, current diagnoses and diagnoses that did not rely on hierarchical exclusion rules are more strongly skewed82,83 and hence contain more asymptomatic cases36,84.

Conclusion

Our findings challenge the prevailing assumption that statistical covariance among forms of psychopathology should be taken as evidence of psychiatric comorbidity. Just as correlation does not imply causation, correlation does not imply comorbidity. The inclusion of asymptomatic, or nomorbid, cases in covariance-based analyses of psychopathology created the illusion of strong covariation among higher-order psychopathology dimensions. At higher levels of psychopathology, higher-order psychopathology dimensions were far more differentiated than is implied by the existing literature16,17,85,86, providing weak evidence of a p-factor (or general factor of psychopathology). Ignoring the implications of the inclusion of largely asymptomatic cases in psychopathology research could jeopardize the integrity and utility of psychiatric classification systems6,9,14,74.

More generally, we call for a reconsideration of the uncritical adoption of the general linear model in psychopathology research, which often produces comorbidity-inference artifacts in community-level data, where most individuals do not exhibit clinically severe psychopathology. We encourage researchers to embrace nonlinearity and population heterogeneity and adopt conceptual and analytic frameworks that prioritize symptomatology over statistical correlation. An exaggerated sense of comorbidity among distinct forms of psychopathology has, and will continue to, perpetuate the reification of artifactual transdiagnostic dimensions42,87,88 and cease investigation of disorder-specific mechanisms21, fundamentally impeding the field’s detection of psychopathology’s etiologic mechanisms, treatment targets and consequences.

Methods

Data

Data were drawn from the first four waves of ABCD (data release 5.0), a collaboration between 21 sites across the USA38,89 that tracks a convenience sample of 11,868 nine- and ten-year-olds from the USA annually across development (wave 1: Mage = 9.94 years, s.d. 0.62 years; 48% female). The ABCD data are not nationally representative, but the study team endeavored to make the sample as representative of the US population as possible. Overall, 74% of youth participants identified as white (14% Hispanic), 21% as Black (2% Hispanic), 7% as Asian and 11% as ‘other’ (see Supplementary Table B1 in Appendix B for demographics of all other waves). In total, 20% identified as Hispanic. These data are available to researchers upon receipt of a data use agreement from the National Institutes of Health (https://nda.nih.gov/).

Measures

We used the caregiver-reported CBCL90, a 113-item assessment of psychiatric symptoms. The CBCL is among the most widely used broadband assessments of youth psychopathology, with over 10,000 publications reporting its use across over 50 different countries91. Symptoms are rated on a three-point scale (‘not true’, ‘somewhat true/sometimes true’ and ‘very true/often true’). We used summary score composites for all psychopathology variables. For readers interested in T scores, Supplementary Table C1 (Appendix C) contains a translation between raw scores and manual-based T scores for all scales across all waves. Of primary focus here were the two broadband psychopathology scores, externalizing and internalizing (at wave 1, coefficient omegas were 0.89 and 0.87, respectively) (see Supplementary Table B2 for omegas for all other waves).

We also considered two subscales that contain items not included in externalizing and internalizing: attention problems and thought problems (coefficient omegas were 0.86 and 0.70, respectively), because they map onto broad transdiagnostic dimensions in modern classification frameworks18,92. In addition to externalizing and internalizing spectra, modern frameworks also include neurodevelopmental93 and psychosis92 (or thought disorder) spectra, which map broadly onto attention problems and thought problems, respectively. As the field incorporates lesser studied higher-order psychopathology dimensions into hierarchical frameworks of psychopathology, we felt that broadening our focus to include these dimensions would be of relevance to the field.

Readers interested in the interface among lower-order psychopathology dimensions within a higher-order dimension (that is, anxiety and depression within internalizing) are directed to Supplementary Table C8 in Appendix C, where we show the results for all change point models based on CBCL subscales. In brief, results were consistent with what we report here, with a few exceptions: across models, the one-change point model was superior to others, and pre-change point slopes were far stronger than post-change point slopes.

Analyses

Univariate and bivariate skew and kurtosis

First, we estimated univariate skewness and kurtosis for all psychopathology dimensions to inform deviations from a normal distribution. A normal distribution’s skewness is 0, with positive values indicating positive skew, and its kurtosis is 3, with values greater than 3 suggesting leptokurtosis. Second, we examined bivariate skewness and kurtosis84, which inform how bivariate relationships deviate from a theoretical multivariate normal distribution. For bivariate estimates, we report Mardia’s chi-squared tests, whose P values indicate whether the data satisfy the assumptions of multivariate normality. We considered all unique pairwise comparisons (that is, six) among psychopathology dimensions across all four waves, and Bonferroni corrected all P values to penalize for the number of tests conducted (24, or 6 × 4).

Bayesian change point models

We used Bayesian change point models to examine whether the relationship between psychopathology dimensions changes at different points along their joint distribution, allowing for nonlinear relationships. Change point models have other common names, including broken-stick models94, or turning point95 or break point models96. They are used routinely in ecology to identify transitions in animal behavior97 and environmental thresholds98 in cross-sectional data, as well as in time-series analysis99. They are conceptually similar to regression discontinuity designs100.

The Bayesian change point model identifies locations, or change points, along the predictor variable where its relationship with the criterion variable may change. For instance, the relationship between x and y might be strong and positive before ten units of x, at which point a change point occurs, and the relationship may become weak and negative after ten units of x. By contrast, an OLS regression assumes that the relationship is equivalent, or does not change, across the joint distribution (in Appendix D we show, via a simulation study, that the OLS estimate is biased in the presence of nonlinearity).

The number of change points can range from zero to infinity. A one-change point model adheres to the following structure:\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${y}_{i}=\left\{\begin{array}{cc}{\alpha }_{1}+{\beta }_{1}X, & x < {\tau }_{r}\\ {\alpha }_{2}+{\beta }_{2}X, & x\ge {\tau }_{r}\end{array}\right.,$$\end{document}yi=α1+β1X,x<τrα2+β2X,x≥τr,where τr represents the location of the change point (or the value of the predictor variable where the slope changes) and where the α and β terms are the intercept and slope terms, respectively. That is, the one-change point model is a combination of two generalized linear models, with each model estimating an intercept and slope before and after the change point. Because the change point model is a generalized linear model, it can be extended to other extensions of the generalized linear model (for example, hierarchical and generalized linear models).

Model estimation

We allowed for up to two change points and compared two link functions, the Gaussian link and the negative binomial link. We used the Gaussian link given its overwhelming popularity, and we used the negative binomial link because it should be more appropriate given the positively skewed (or overdispersed) count data used here101,102.

In contrast with the Gaussian link-based regression, negative binomial regression assumes a nonlinear (log–linear) relationship between the predictor variables and the expected count of the criterion variable, whereby change in the predictor is multiplicative. The beta coefficient (or slope, B) represents the change in the log of the expected count for a 1-point increase in the predictor variable. To translate the nonlinear estimates, one can exponentiate the slope (exp[B]) so that it reflects the rate ratio or the multiplicative factor by which the expected count changes when the predictor increases by one point.

Every model used Markov Chain Monte Carlo sampling with a burn-in length of 10,000, a total chain length of 20,000 and a thinning of every second sample to control for potential autocorrelation. All sampling was performed using the Stan language and the No-U-Turn Sampler algorithm103,104. We selected diffuse and naive priors. Estimation of the change point(s) used uniform priors that ranged from the minimum to the maximum value of the domain, and estimation of the slope(s) and intercept(s) used normal priors centered on 0 with a variance of 1. Random effect variance used uniform priors ranging from 0 to 10. In terms of other covariates, we included a fixed effect for age and nested random effects for complex survey design features (family within site).

Model convergence

We examined model convergence using the potential scale reduction factor \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{R}$$\end{document}Rˆ (ref. 105) to determine if the chains mixed. \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{R}$$\end{document}Rˆ measures the ratio of the average variance of samples within each chain to the variance of the pooled samples across chains. If all chains are at equilibrium, the average variance of samples within each chain will be equal to the variance of the pooled samples across chains, producing an \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{R}$$\end{document}Rˆ of 1.0. Values less than 1.2 generally indicate convergence across chains105. To confirm strong model convergence, we also visually inspected the trace, density and autocorrelation plots (Supplementary Figs. B1–B4; Supplementary Table 2 contains \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{R}$$\end{document}Rˆ values for the slope parameters of interest here; and Appendix C contains all model parameters, including \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{R}$$\end{document}Rˆ values, for all models).

Model comparison

We prioritized the Leave-One-Out Information Criterion (LOOIC), which is commonly used for Bayesian model selection106. LOOIC facilitates the comparison of models within and across link functions, with lower values indicating a better-fitting model. We also used model R2 (or the proportion of variability explained).

Of the 288 models compared—Gaussian and negative binomial, each with either zero, one or two change points—LOOIC supported one-change point negative binomial models over Gaussian models (Supplementary Fig. B5 and Supplementary Table B3). Additional sources of evidence converged on the inappropriateness of the Gaussian models, including that the data were not univariate or bivariate normal, and that model diagnostics of linear regressions suggested substantial violations of linear regression assumptions, including linearity, homogeneity of variance and homogeneity in residuals (Supplementary Figs. B9 and B10). Thus, results from the Gaussian models are invalid.

Among the negative binomial models, the two-change point models were over extracted, with little to no increment in R2 above and beyond the one-change point models (Supplementary Fig. A7). For instance, when we regressed externalizing onto internalizing at wave 1, model R2 (or variance explained in externalizing) was 21% in the zero-change point model, 56% in the one-change point model and 55% in the two-change point model.

We anticipated that a two-change point model might reveal a change in the slope between internalizing and externalizing at high levels of internalizing (or high levels of severity), with the slope strengthening and reflecting increased comorbidity. Instead, the second change point in this model was located at τ2 = 5.09, corresponding to endorsement of around three or four internalizing symptoms. Further, the slope after the second change point was lower than the slope after the first change point (B1 = 0.17, B2 = 0.27 and B3 = 0.07, respectively), which is inconsistent with the notion that their association would increase as a function of presumed comorbidity. There were no instances in the two-change point negative binomial models where the slope after the second change point (B3) increased relative to the prior slope (B2). Given the lack of incremental R2 for the two-change point models, we carried forward all one-change point negative binomial models (Fig. 2a).

We considered an alternative estimation method, generalized additive multilevel models107,108, to examine nonlinear associations between higher-order psychopathology dimensions, as in previous work109. We detected the same nonlinear associations using this method and did not find consistent evidence that there is a strong positive association between externalizing and internalizing at extreme high levels of internalizing (Appendix E).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

References

1.First MB Mutually exclusive versus co-occurring diagnostic categories: the challenge of diagnostic comorbidity Psychopathology 2005 38 206 210 16145276 10.1159/000086093 First, M. B. Mutually exclusive versus co-occurring diagnostic categories: the challenge of diagnostic comorbidity. Psychopathology38, 206–210 (2005).10.1159/000086093 16145276

2.Widiger TA Frances A The DSM-III personality disorders: perspectives from psychology Arch. Gen. Psychiatry 1985 42 615 3890799 10.1001/archpsyc.1985.01790290097011 Widiger, T. A. & Frances, A. The DSM-III personality disorders: perspectives from psychology. Arch. Gen. Psychiatry42, 615 (1985).10.1001/archpsyc.1985.01790290097011 3890799

3.Kessler RC Chiu WT Demler O Walters EE Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the national comorbidity survey replication Arch. Gen. Psychiatry 2005 62 617 627 15939839 PMC2847357 10.1001/archpsyc.62.6.617 Kessler, R. C., Chiu, W. T., Demler, O. & Walters, E. E. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the national comorbidity survey replication. Arch. Gen. Psychiatry62, 617–627 (2005).10.1001/archpsyc.62.6.617 PMC2847357 15939839

4.Deighton J Lereya ST Wolpert M Enduring mental health in childhood and adolescence: learning from the Millennium Cohort Study J. Am. Acad. Child Adolesc. Psychiatry 2021 60 1030 1039 33264663 10.1016/j.jaac.2020.11.012 Deighton, J., Lereya, S. T. & Wolpert, M. Enduring mental health in childhood and adolescence: learning from the Millennium Cohort Study. J. Am. Acad. Child Adolesc. Psychiatry60, 1030–1039 (2021).10.1016/j.jaac.2020.11.012 33264663

5.Schaefer JD Enduring mental health: prevalence and prediction J. Abnorm. Psychol. 2017 126 212 224 27929304 10.1037/abn0000232 PMC5304549 Schaefer, J. D. et al. Enduring mental health: prevalence and prediction. J. Abnorm. Psychol.126, 212–224 (2017).10.1037/abn0000232 PMC5304549 27929304

6.Kotov R The Hierarchical Taxonomy of Psychopathology (HiTOP): a dimensional alternative to traditional nosologies J. Abnorm. Psychol. 2017 126 454 477 28333488 10.1037/abn0000258 Kotov, R. et al. The Hierarchical Taxonomy of Psychopathology (HiTOP): a dimensional alternative to traditional nosologies. J. Abnorm. Psychol.126, 454–477 (2017).10.1037/abn0000258 28333488

7.Hankin BL Understanding comorbidity among internalizing problems: integrating latent structural models of psychopathology and risk mechanisms Dev. Psychopathol. 2016 28 987 1012 27739389 PMC5119897 10.1017/S0954579416000663 Hankin, B. L. et al. Understanding comorbidity among internalizing problems: integrating latent structural models of psychopathology and risk mechanisms. Dev. Psychopathol.28, 987–1012 (2016).10.1017/S0954579416000663 PMC5119897 27739389

8.Clark LA Nuzum H Ro E Manifestations of personality impairment severity: comorbidity, course/prognosis, psychosocial dysfunction, and ‘borderline’ personality features Curr. Opin. Psychol. 2018 21 117 121 29291458 10.1016/j.copsyc.2017.12.004 Clark, L. A., Nuzum, H. & Ro, E. Manifestations of personality impairment severity: comorbidity, course/prognosis, psychosocial dysfunction, and ‘borderline’ personality features. Curr. Opin. Psychol.21, 117–121 (2018).10.1016/j.copsyc.2017.12.004 29291458

9.Conway CC A hierarchical taxonomy of psychopathology can transform mental health research Perspect. Psychol. Sci. 2019 14 419 436 30844330 PMC6497550 10.1177/1745691618810696 Conway, C. C. et al. A hierarchical taxonomy of psychopathology can transform mental health research. Perspect. Psychol. Sci.14, 419–436 (2019).10.1177/1745691618810696 PMC6497550 30844330

10.Beauchaine, T. P. & Klein, D. N. Classifying psychopathology: the DSM, empirically based taxonomies, and the research domain criteria. In Child and Adolescent Psychopathology 3rd edn (eds Beauchaine, T. P. & Hinshaw S. P.) 33–67 10.1002/9781394258932.ch2 (Wiley, 2017).

11.Wright AGC Simms LJ A metastructural model of mental disorders and pathological personality traits Psychol. Med. 2015 45 2309 2319 25903065 PMC4498970 10.1017/S0033291715000252 Wright, A. G. C. & Simms, L. J. A metastructural model of mental disorders and pathological personality traits. Psychol. Med.45, 2309–2319 (2015).10.1017/S0033291715000252 PMC4498970 25903065

12.Carragher N Krueger RF Eaton NR Slade T Disorders without borders: current and future directions in the meta-structure of mental disorders Soc. Psychiatry Psychiatr. Epidemiol. 2015 50 339 350 10.1007/s00127-014-1004-z 25557024 Carragher, N., Krueger, R. F., Eaton, N. R. & Slade, T. Disorders without borders: current and future directions in the meta-structure of mental disorders. Soc. Psychiatry Psychiatr. Epidemiol.50, 339–350 (2015).10.1007/s00127-014-1004-z 25557024

13.Conway CC Forbes MK South SC A Hierarchical Taxonomy of Psychopathology (HiTOP) primer for mental health researchers Clin. Psychol. Sci. 2022 10 236 258 35599839 10.1177/21677026211017834 PMC9122089 Conway, C. C., Forbes, M. K. & South, S. C. A Hierarchical Taxonomy of Psychopathology (HiTOP) primer for mental health researchers. Clin. Psychol. Sci.10, 236–258 (2022).10.1177/21677026211017834 PMC9122089 35599839

14.Cicero DC State of the science: the Hierarchical Taxonomy of Psychopathology (HiTOP) Behav. Ther. 2024 55 1114 1129 39443056 10.1016/j.beth.2024.05.001 Cicero, D. C. et al. State of the science: the Hierarchical Taxonomy of Psychopathology (HiTOP). Behav. Ther.55, 1114–1129 (2024).10.1016/j.beth.2024.05.001 39443056

15.Achenbach TM The classification of children’s psychiatric symptoms: s factor-analytic study Psychol. Monogr. Gen. Appl. 1966 80 1 37 10.1037/h0093906 5968338 Achenbach, T. M. The classification of children’s psychiatric symptoms: s factor-analytic study. Psychol. Monogr. Gen. Appl.80, 1–37 (1966).10.1037/h0093906 5968338

16.Krueger RF Markon KE Reinterpreting comorbidity: a model-based approach to understanding and classifying psychopathology Annu. Rev. Clin. Psychol. 2006 2 111 133 17716066 PMC2242354 10.1146/annurev.clinpsy.2.022305.095213 Krueger, R. F. & Markon, K. E. Reinterpreting comorbidity: a model-based approach to understanding and classifying psychopathology. Annu. Rev. Clin. Psychol.2, 111–133 (2006).10.1146/annurev.clinpsy.2.022305.095213 PMC2242354 17716066

17.Krueger RF The structure of common mental disorders Arch. Gen. Psychiatry 1999 56 921 926 10530634 10.1001/archpsyc.56.10.921 Krueger, R. F. The structure of common mental disorders. Arch. Gen. Psychiatry56, 921–926 (1999).10.1001/archpsyc.56.10.921 10530634

18.Krueger RF Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): II. Externalizing superspectrum World Psychiatry 2021 20 171 193 34002506 PMC8129870 10.1002/wps.20844 Krueger, R. F. et al. Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): II. Externalizing superspectrum. World Psychiatry20, 171–193 (2021).10.1002/wps.20844 PMC8129870 34002506

19.Watson D Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): III. Emotional dysfunction superspectrum World Psychiatry 2022 21 26 54 35015357 PMC8751579 10.1002/wps.20943 Watson, D. et al. Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): III. Emotional dysfunction superspectrum. World Psychiatry21, 26–54 (2022).10.1002/wps.20943 PMC8751579 35015357

20.Krueger RF Markon KE Patrick CJ Benning SD Kramer MD Linking antisocial behavior, substance use, and personality: an integrative quantitative model of the adult externalizing spectrum J. Abnorm. Psychol. 2007 116 645 666 18020714 PMC2242625 10.1037/0021-843X.116.4.645 Krueger, R. F., Markon, K. E., Patrick, C. J., Benning, S. D. & Kramer, M. D. Linking antisocial behavior, substance use, and personality: an integrative quantitative model of the adult externalizing spectrum. J. Abnorm. Psychol.116, 645–666 (2007).10.1037/0021-843X.116.4.645 PMC2242625 18020714

21.Caspi A Moffitt TE All for one and one for all: mental disorders in one dimension Am. J. Psychiatry 2018 175 831 844 29621902 PMC6120790 10.1176/appi.ajp.2018.17121383 Caspi, A. & Moffitt, T. E. All for one and one for all: mental disorders in one dimension. Am. J. Psychiatry175, 831–844 (2018).10.1176/appi.ajp.2018.17121383 PMC6120790 29621902

22.Lahey BB Krueger RF Rathouz PJ Waldman ID Zald DH Validity and utility of the general factor of psychopathology World Psychiatry 2017 16 142 144 28498590 PMC5428167 10.1002/wps.20410 Lahey, B. B., Krueger, R. F., Rathouz, P. J., Waldman, I. D. & Zald, D. H. Validity and utility of the general factor of psychopathology. World Psychiatry16, 142–144 (2017).10.1002/wps.20410 PMC5428167 28498590

23.Smith GT Atkinson EA Davis HA Riley EN Oltmanns JR The general factor of psychopathology Annu. Rev. Clin. Psychol. 2020 16 75 98 32040926 10.1146/annurev-clinpsy-071119-115848 Smith, G. T., Atkinson, E. A., Davis, H. A., Riley, E. N. & Oltmanns, J. R. The general factor of psychopathology. Annu. Rev. Clin. Psychol.16, 75–98 (2020).10.1146/annurev-clinpsy-071119-115848 32040926

24.Romer AL Replicability of structural brain alterations associated with general psychopathology: evidence from a population-representative birth cohort Mol. Psychiatry 2021 26 3839 3846 10.1038/s41380-019-0621-z 31796893 PMC7266702 Romer, A. L. et al. Replicability of structural brain alterations associated with general psychopathology: evidence from a population-representative birth cohort. Mol. Psychiatry26, 3839–3846 (2021).10.1038/s41380-019-0621-z PMC7266702 31796893

25.Gluschkoff K Jokela M Rosenström T General psychopathology factor and borderline personality disorder: evidence for substantial overlap from two nationally representative surveys of U.S. adults Personal Disord. 2021 12 86 92 32584090 10.1037/per0000405 Gluschkoff, K., Jokela, M. & Rosenström, T. General psychopathology factor and borderline personality disorder: evidence for substantial overlap from two nationally representative surveys of U.S. adults. Personal Disord.12, 86–92 (2021).10.1037/per0000405 32584090

26.Schweizer TH Snyder HR Young JF Hankin BL The breadth and potency of transdiagnostic cognitive risks for psychopathology in youth J. Consult. Clin. Psychol. 2020 88 196 211 32068422 PMC7219569 10.1037/ccp0000470 Schweizer, T. H., Snyder, H. R., Young, J. F. & Hankin, B. L. The breadth and potency of transdiagnostic cognitive risks for psychopathology in youth. J. Consult. Clin. Psychol.88, 196–211 (2020).10.1037/ccp0000470 PMC7219569 32068422

27.Meier MA Meier MH Clinical implications of a general psychopathology factor: a cognitive–behavioral transdiagnostic group treatment for community mental health J. Psychother. Integr. 2018 28 253 268 10.1037/int0000095 Meier, M. A. & Meier, M. H. Clinical implications of a general psychopathology factor: a cognitive–behavioral transdiagnostic group treatment for community mental health. J. Psychother. Integr.28, 253–268 (2018).

28.Pettersson E Lahey BB Larsson H Lichtenstein P Criterion validity and utility of the general factor of psychopathology in childhood: predictive associations with independently measured severe adverse mental health outcomes in adolescence J. Am. Acad. Child Adolesc. Psychiatry 2018 57 372 383 29859553 10.1016/j.jaac.2017.12.016 Pettersson, E., Lahey, B. B., Larsson, H. & Lichtenstein, P. Criterion validity and utility of the general factor of psychopathology in childhood: predictive associations with independently measured severe adverse mental health outcomes in adolescence. J. Am. Acad. Child Adolesc. Psychiatry57, 372–383 (2018).10.1016/j.jaac.2017.12.016 29859553

29.McElroy E Belsky J Carragher N Fearon P Patalay P Developmental stability of general and specific factors of psychopathology from early childhood to adolescence: dynamic mutualism or p -differentiation? J. Child Psychol. Psychiatry 2018 59 667 675 29197107 10.1111/jcpp.12849 PMC6001631 McElroy, E., Belsky, J., Carragher, N., Fearon, P. & Patalay, P. Developmental stability of general and specific factors of psychopathology from early childhood to adolescence: dynamic mutualism or p -differentiation?. J. Child Psychol. Psychiatry59, 667–675 (2018).10.1111/jcpp.12849 PMC6001631 29197107

30.Carver CS Johnson SL Timpano KR Toward a functional view of the p factor in psychopathology Clin. Psychol. Sci. 2017 5 880 889 29057170 PMC5646702 10.1177/2167702617710037 Carver, C. S., Johnson, S. L. & Timpano, K. R. Toward a functional view of the p factor in psychopathology. Clin. Psychol. Sci.5, 880–889 (2017).10.1177/2167702617710037 PMC5646702 29057170

31.Shields AN Reardon KW Brandes CM Tackett JL The p factor in children: relationships with executive functions and effortful control J. Res. Pers. 2019 82 103853 10.1016/j.jrp.2019.103853 Shields, A. N., Reardon, K. W., Brandes, C. M. & Tackett, J. L. The p factor in children: relationships with executive functions and effortful control. J. Res. Pers.82, 103853 (2019).

32.Brandes CM Herzhoff K Smack AJ Tackett JL The p factor and the n factor: associations between the general factors of psychopathology and neuroticism in children Clin. Psychol. Sci. 2019 7 1266 1284 10.1177/2167702619859332 Brandes, C. M., Herzhoff, K., Smack, A. J. & Tackett, J. L. The p factor and the n factor: associations between the general factors of psychopathology and neuroticism in children. Clin. Psychol. Sci.7, 1266–1284 (2019).

33.Tackett JL Common genetic influences on negative emotionality and a general psychopathology factor in childhood and adolescence J. Abnorm. Psychol. 2013 122 1142 1153 24364617 PMC3938013 10.1037/a0034151 Tackett, J. L. et al. Common genetic influences on negative emotionality and a general psychopathology factor in childhood and adolescence. J. Abnorm. Psychol.122, 1142–1153 (2013).10.1037/a0034151 PMC3938013 24364617

34.Christensen WF Wall MM Moustaki I Assessing dimensionality in dichotomous items when many subjects have all-zero responses: an example from psychiatry and a solution using mixture models Appl. Psychol. Meas. 2022 46 167 184 35528272 PMC9073639 10.1177/01466216211066602 Christensen, W. F., Wall, M. M. & Moustaki, I. Assessing dimensionality in dichotomous items when many subjects have all-zero responses: an example from psychiatry and a solution using mixture models. Appl. Psychol. Meas.46, 167–184 (2022).10.1177/01466216211066602 PMC9073639 35528272

35.Hoffman M The influence of sample selection on the structure of psychopathology symptom networks: an example with alcohol use disorder J. Abnorm. Psychol. 2019 128 473 486 31192638 PMC6614010 10.1037/abn0000438 Hoffman, M. et al. The influence of sample selection on the structure of psychopathology symptom networks: an example with alcohol use disorder. J. Abnorm. Psychol.128, 473–486 (2019).10.1037/abn0000438 PMC6614010 31192638

36.Steppan M Piontek D Kraus L The effect of sample selection on the distinction between alcohol abuse and dependence Int. J. Alcohol Drug Res. 2014 3 2 10.7895/ijadr.v3i2.100 Steppan, M., Piontek, D. & Kraus, L. The effect of sample selection on the distinction between alcohol abuse and dependence. Int. J. Alcohol Drug Res.3, 2 (2014).

37.Brusco, M., Cradit, J. D. & Steinley, D. A comparison of 71 binary similarity coefficients: the effect of base rates. PLoS ONE16, e0247751 (2021).10.1371/journal.pone.0247751 PMC8026075 33826612

38.Jernigan TL Brown SA Dowling GJ The adolescent brain cognitive development study J. Res. Adolesc. 2018 28 154 156 29460352 PMC7477916 10.1111/jora.12374 Jernigan, T. L., Brown, S. A. & Dowling, G. J. The adolescent brain cognitive development study. J. Res. Adolesc.28, 154–156 (2018).10.1111/jora.12374 PMC7477916 29460352

39.Lahey BB Is there a general factor of prevalent psychopathology during adulthood? J. Abnorm. Psychol. 2012 121 971 977 22845652 PMC4134439 10.1037/a0028355 Lahey, B. B. et al. Is there a general factor of prevalent psychopathology during adulthood?. J. Abnorm. Psychol.121, 971–977 (2012).10.1037/a0028355 PMC4134439 22845652

40.Forbes MK Implications of the symptom-level overlap among DSM diagnoses for dimensions of psychopathology J. Emot. Psychopathol. 2023 1 104 112 10.55913/joep.v1i1.6 Forbes, M. K. Implications of the symptom-level overlap among DSM diagnoses for dimensions of psychopathology. J. Emot. Psychopathol.1, 104–112 (2023).

41.Watts AL How robust is the p factor? Using multitrait-multimethod modeling to inform the meaning of general factors of youth psychopathology Clin. Psychol. Sci. 2022 10 640 661 36090949 10.1177/21677026211055170 PMC9454373 Watts, A. L. et al. How robust is the p factor? Using multitrait-multimethod modeling to inform the meaning of general factors of youth psychopathology. Clin. Psychol. Sci.10, 640–661 (2022).10.1177/21677026211055170 PMC9454373 36090949

42.Watts AL Greene AL Bonifay W Fried EI A critical evaluation of the p-factor literature Nat. Rev. Psychol. 2024 3 108 122 10.1038/s44159-023-00260-2 Watts, A. L., Greene, A. L., Bonifay, W. & Fried, E. I. A critical evaluation of the p-factor literature. Nat. Rev. Psychol.3, 108–122 10.1038/s44159-023-00260-2 (2024).

43.Watts AL Poore HE Waldman ID Riskier tests of the validity of the bifactor model of psychopathology Clin. Psychol. Sci. 2019 7 1285 1303 10.1177/2167702619855035 Watts, A. L., Poore, H. E. & Waldman, I. D. Riskier tests of the validity of the bifactor model of psychopathology. Clin. Psychol. Sci.7, 1285–1303 (2019).

44.Watts AL Lane SP Bonifay W Steinley D Meyer FAC Building theories on top of, and not independent of, statistical models: the case of the p-factor Psychol. Inq. 2020 31 310 320 33510565 10.1080/1047840X.2020.1853476 PMC7839945 Watts, A. L., Lane, S. P., Bonifay, W., Steinley, D. & Meyer, F. A. C. Building theories on top of, and not independent of, statistical models: the case of the p-factor. Psychol. Inq.31, 310–320 (2020).10.1080/1047840x.2020.1853476 PMC7839945 33510565

45.Warrens MJ On similarity coefficients for 2×2 tables and correction for chance Psychometrika 2008 73 487 502 10.1007/s11336-008-9059-y 20037641 PMC2792365 Warrens, M. J. On similarity coefficients for 2×2 tables and correction for chance. Psychometrika73, 487–502 (2008).10.1007/s11336-008-9059-y PMC2792365 20037641

46.Janson S Vegelius J Measures of ecological association Oecologia 1981 49 371 376 10.1007/BF00347601 28309999 Janson, S. & Vegelius, J. Measures of ecological association. Oecologia49, 371–376 (1981).10.1007/BF00347601 28309999

47.Lilienfeld SO Waldman ID Israel AC A critical examination of the use of the term and concept of comorbidity in psychopathology research Clin. Psychol. Sci. Pract. 1994 1 71 83 10.1111/j.1468-2850.1994.tb00007.x Lilienfeld, S. O., Waldman, I. D. & Israel, A. C. A critical examination of the use of the term and concept of comorbidity in psychopathology research. Clin. Psychol. Sci. Pract.1, 71–83 (1994).

48.Eaton NR A review of approaches and models in psychopathology conceptualization research Nat. Rev. Psychol. 2023 2 622 636 10.1038/s44159-023-00218-4 Eaton, N. R. et al. A review of approaches and models in psychopathology conceptualization research. Nat. Rev. Psychol.2, 622–636 (2023).

49.Spearman C General Intelligence,’ objectively determined and measured Am. J. Psychol. 1904 15 201 10.2307/1412107 14717630 Spearman, C. General Intelligence,’ objectively determined and measured. Am. J. Psychol.15, 201 (1904).10.1037/0022-3514.86.1.96 14717630

50.Caspi A The p factor: one general psychopathology factor in the structure of psychiatric disorders? Clin. Psychol. Sci. 2014 2 119 137 25360393 PMC4209412 10.1177/2167702613497473 Caspi, A. et al. The p factor: one general psychopathology factor in the structure of psychiatric disorders?. Clin. Psychol. Sci.2, 119–137 (2014).10.1177/2167702613497473 PMC4209412 25360393

51.Spearman, C. The Abilities Of Manhttp://archive.org/details/abilitiesofman031969mbp (Macmillan, 1927).

52.Detterman DK Daniel MH Correlations of mental tests with each other and with cognitive variables are highest for low IQ groups Intelligence 1989 13 349 359 10.1016/S0160-2896(89)80007-8 Detterman, D. K. & Daniel, M. H. Correlations of mental tests with each other and with cognitive variables are highest for low IQ groups. Intelligence13, 349–359 (1989).

53.Maxwell AE The WPPSI: a marked discrepancy in the correlations of the subtests for good and poor readers Br. J. Math. Stat. Psychol. 1972 25 283 291 10.1111/j.2044-8317.1972.tb00498.x Maxwell, A. E. The WPPSI: a marked discrepancy in the correlations of the subtests for good and poor readers. Br. J. Math. Stat. Psychol.25, 283–291 (1972).

54.Jensen AR Regularities in Spearman’s law of diminishing returns Intelligence 2003 31 95 105 10.1016/S0160-2896(01)00094-0 Jensen, A. R. Regularities in Spearman’s law of diminishing returns. Intelligence31, 95–105 (2003).

55.Haeffel GJ Folk classification and factor rotations: whales, sharks, and the problems with the Hierarchical Taxonomy of Psychopathology (HiTOP) Clin. Psychol. Sci. 2022 10 259 278 35425668 10.1177/21677026211002500 PMC9004619 Haeffel, G. J. et al. Folk classification and factor rotations: whales, sharks, and the problems with the Hierarchical Taxonomy of Psychopathology (HiTOP). Clin. Psychol. Sci.10, 259–278 (2022).10.1177/21677026211002500 PMC9004619 35425668

56.Reise SP Du H Wong EF Hubbard AS Haviland MG Matching IRT models to patient-reported outcomes constructs: the graded response and log-logistic models for scaling depression Psychometrika 2021 86 800 824 10.1007/s11336-021-09802-0 34463910 PMC8437930 Reise, S. P., Du, H., Wong, E. F., Hubbard, A. S. & Haviland, M. G. Matching IRT models to patient-reported outcomes constructs: the graded response and log-logistic models for scaling depression. Psychometrika86, 800–824 (2021).10.1007/s11336-021-09802-0 PMC8437930 34463910

57.Flora DB Curran PJ An empirical evaluation of alternative methods of estimation for confirmatory factor analysis with ordinal data Psychol. Methods 2004 9 466 491 15598100 PMC3153362 10.1037/1082-989X.9.4.466 Flora, D. B. & Curran, P. J. An empirical evaluation of alternative methods of estimation for confirmatory factor analysis with ordinal data. Psychol. Methods9, 466–491 (2004).10.1037/1082-989X.9.4.466 PMC3153362 15598100

58.Preacher KJ MacCallum RC Repairing Tom Swift’s electric factor analysis machine Understand. Stat. 2003 2 13 43 10.1207/S15328031US0201_02 Preacher, K. J. & MacCallum, R. C. Repairing Tom Swift’s electric factor analysis machine. Understand. Stat.2, 13–43 (2003).

59.Emmons, O. Linearity In Factor Analysis10.58809/VCNR9865 (Fort Hays State Univ., 1939).

60.Wall MM Park JY Moustaki I IRT modeling in the presence of zero-inflation with application to psychiatric disorder severity Appl. Psychol. Meas. 2015 39 583 597 29881029 PMC5978495 10.1177/0146621615588184 Wall, M. M., Park, J. Y. & Moustaki, I. IRT modeling in the presence of zero-inflation with application to psychiatric disorder severity. Appl. Psychol. Meas.39, 583–597 (2015).10.1177/0146621615588184 PMC5978495 29881029

61.Kirisci L Hsu Tchi Yu L Robustness of item parameter estimation programs to assumptions of unidimensionality and normality Appl. Psychol. Meas. 2001 25 146 162 10.1177/01466210122031975 Kirisci, L., Hsu, T. chi & Yu, L. Robustness of item parameter estimation programs to assumptions of unidimensionality and normality. Appl. Psychol. Meas.25, 146–162 (2001).

62.Azevedo CLN Bolfarine H Andrade DF Bayesian inference for a skew-normal IRT model under the centred parameterization Comput. Stat. Data Anal. 2011 55 353 365 10.1016/j.csda.2010.05.003 Azevedo, C. L. N., Bolfarine, H. & Andrade, D. F. Bayesian inference for a skew-normal IRT model under the centred parameterization. Comput. Stat. Data Anal.55, 353–365 (2011).

63.Field AP Wilcox RR Robust statistical methods: a primer for clinical psychology and experimental psychopathology researchers Behav. Res. Ther. 2017 98 19 38 28577757 10.1016/j.brat.2017.05.013 Field, A. P. & Wilcox, R. R. Robust statistical methods: a primer for clinical psychology and experimental psychopathology researchers. Behav. Res. Ther.98, 19–38 (2017).10.1016/j.brat.2017.05.013 28577757

64.Reise SP Rodriguez A Spritzer KL Hays RD Alternative approaches to addressing non-normal distributions in the application of IRT models to personality measures J. Pers. Assess. 2018 100 363 374 29087217 10.1080/00223891.2017.1381969 PMC6252010 Reise, S. P., Rodriguez, A., Spritzer, K. L. & Hays, R. D. Alternative approaches to addressing non-normal distributions in the application of IRT models to personality measures. J. Pers. Assess.100, 363–374 (2018).10.1080/00223891.2017.1381969 PMC6252010 29087217

65.Haslam N The latent structure of mental disorders: a taxometric update on the categorical vs dimensional debate Curr. Psychiatry Rev. 2007 3 172 177 10.2174/157340007781369685 Haslam, N. The latent structure of mental disorders: a taxometric update on the categorical vs dimensional debate. Curr. Psychiatry Rev.3, 172–177 (2007).

66.Meehl, P. E. & Golden, R. Taxometric methods. In Handbook of Research Methods in Clinical Psychology Ch. 5 (eds Kendall, P. & Butcher, J.) 127–181 (Wiley, 1982).

67.Kam CCS Bifactor model is not the best-fitting model for self-esteem: investigation with a novel technique Assessment 2021 28 1799 1812 10.1177/1073191120949916 32917100 Kam, C. C. S. Bifactor model is not the best-fitting model for self-esteem: investigation with a novel technique. Assessment28, 1799–1812 10.1177/1073191120949916 (2021).32917100

68.Molenaar D Dolan CV Wicherts JM van der Maas HLJ Modeling differentiation of cognitive abilities within the higher-order factor model using moderated factor analysis Intelligence 2010 38 611 624 10.1016/j.intell.2010.09.002 Molenaar, D., Dolan, C. V., Wicherts, J. M. & van der Maas, H. L. J. Modeling differentiation of cognitive abilities within the higher-order factor model using moderated factor analysis. Intelligence38, 611–624 (2010).

69.Molenaar D Kő N Rózsa S Mészáros A Differentiation of cognitive abilities in the WAIS-IV at the item level Intelligence 2017 65 48 59 10.1016/j.intell.2017.10.004 Molenaar, D., Kő, N., Rózsa, S. & Mészáros, A. Differentiation of cognitive abilities in the WAIS-IV at the item level. Intelligence65, 48–59 (2017).

70.Reynolds MR Keith TZ Beretvas SN Use of factor mixture modeling to capture Spearman’s law of diminishing returns Intelligence 2010 38 231 241 10.1016/j.intell.2010.01.002 Reynolds, M. R., Keith, T. Z. & Beretvas, S. N. Use of factor mixture modeling to capture Spearman’s law of diminishing returns. Intelligence38, 231–241 (2010).

71.Hoffman M Steinley D Trull TJ Sher KJ Estimating transdiagnostic symptom networks: the problem of ‘skip outs’ in diagnostic interviews Psychol. Assess. 2019 31 73 81 30160496 10.1037/pas0000644 PMC6312735 Hoffman, M., Steinley, D., Trull, T. J. & Sher, K. J. Estimating transdiagnostic symptom networks: the problem of ‘skip outs’ in diagnostic interviews. Psychol. Assess.31, 73–81 (2019).10.1037/pas0000644 PMC6312735 30160496

72.Kazdin AE Kagan J Models of dysfunction in developmental psychopathology Clin. Psychol. Sci. Pract. 1994 1 35 52 10.1111/j.1468-2850.1994.tb00005.x Kazdin, A. E. & Kagan, J. Models of dysfunction in developmental psychopathology. Clin. Psychol. Sci. Pract.1, 35–52 (1994).

73.Muthén BO Latent variable modeling in heterogeneous populations Psychometrika 1989 54 557 585 10.1007/BF02296397 Muthén, B. O. Latent variable modeling in heterogeneous populations. Psychometrika54, 557–585 (1989).

74.Oquendo MA Initial strategy for the future of DSM Am. J. Psychiatry 2026 183 292 300 41593833 10.1176/appi.ajp.20250878 Oquendo, M. A. et al. Initial strategy for the future of DSM. Am. J. Psychiatry183, 292–300 (2026).10.1176/appi.ajp.20250878 41593833

75.Ruggero CJ Integrating the Hierarchical Taxonomy of Psychopathology (HiTOP) into clinical practice J. Consult. Clin. Psychol. 2019 87 1069 1084 31724426 PMC6859953 10.1037/ccp0000452 Ruggero, C. J. et al. Integrating the Hierarchical Taxonomy of Psychopathology (HiTOP) into clinical practice. J. Consult. Clin. Psychol.87, 1069–1084 (2019).10.1037/ccp0000452 PMC6859953 31724426

76.de Ron J Fried EI Epskamp S Psychological networks in clinical populations: investigating the consequences of Berkson’s bias Psychol. Med. 2021 51 168 176 10.1017/S0033291719003209 31796131 de Ron, J., Fried, E. I. & Epskamp, S. Psychological networks in clinical populations: investigating the consequences of Berkson’s bias. Psychol. Med.51, 168–176 10.1017/S0033291719003209 (2021).31796131

77.Hill G Neyman’s bias re-visited J. Clin. Epidemiol. 2003 56 293 296 12767404 10.1016/S0895-4356(02)00571-1 Hill, G. Neyman’s bias re-visited. J. Clin. Epidemiol.56, 293–296 (2003).10.1016/s0895-4356(02)00571-1 12767404

78.Meredith W Notes on factorial invariance Psychometrika 1964 29 177 185 10.1007/BF02289699 Meredith, W. Notes on factorial invariance. Psychometrika29, 177–185 (1964).

79.McNeish D Wolf MG Thinking twice about sum scores Behav. Res. Methods 2020 52 2287 2305 32323277 10.3758/s13428-020-01398-0 McNeish, D. & Wolf, M. G. Thinking twice about sum scores. Behav. Res. Methods52, 2287–2305 (2020).10.3758/s13428-020-01398-0 32323277

80.Ringwald WR Forbes MK Wright AGC Meta-analytic tests of measurement invariance of internalizing and externalizing psychopathology across common methodological characteristics J. Psychopathol. Clin. Sci. 2022 131 847 856 36326626 10.1037/abn0000785 Ringwald, W. R., Forbes, M. K. & Wright, A. G. C. Meta-analytic tests of measurement invariance of internalizing and externalizing psychopathology across common methodological characteristics. J. Psychopathol. Clin. Sci.131, 847–856 (2022).10.1037/abn0000785 36326626

81.Conway CC Brown TA Evaluating dimensional models of psychopathology in outpatients diagnosed with emotional disorders: a cautionary tale Depress. Anxiety 2018 35 898 902 29637665 10.1002/da.22740 Conway, C. C. & Brown, T. A. Evaluating dimensional models of psychopathology in outpatients diagnosed with emotional disorders: a cautionary tale. Depress. Anxiety35, 898–902 (2018).10.1002/da.22740 29637665

82.Kessler RC Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States: results from the National Comorbidity Survey Arch. Gen. Psychiatry 1994 51 8 8279933 10.1001/archpsyc.1994.03950010008002 Kessler, R. C. Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States: results from the National Comorbidity Survey. Arch. Gen. Psychiatry51, 8 (1994).10.1001/archpsyc.1994.03950010008002 8279933

83.Kotov R Ruggero CJ Krueger RF Watson D Zimmerman M The perils of hierarchical exclusion rules: a further word of caution Depress. Anxiety 2018 35 903 904 30178498 10.1002/da.22826 Kotov, R., Ruggero, C. J., Krueger, R. F., Watson, D. & Zimmerman, M. The perils of hierarchical exclusion rules: a further word of caution. Depress. Anxiety35, 903–904 (2018).10.1002/da.22826 30178498

84.Mardia KV Measures of multivariate skewness and kurtosis with applications Biometrika 1970 57 519 530 10.1093/biomet/57.3.519 Mardia, K. V. Measures of multivariate skewness and kurtosis with applications. Biometrika57, 519–530 (1970).

85.Forbes MK Three recommendations based on a comparison of the reliability and validity of the predominant models used in research on the empirical structure of psychopathology J. Abnorm. Psychol. 2021 130 297 317 33539117 10.1037/abn0000533 Forbes, M. K. et al. Three recommendations based on a comparison of the reliability and validity of the predominant models used in research on the empirical structure of psychopathology. J. Abnorm. Psychol.130, 297–317 (2021).10.1037/abn0000533 33539117

86.Ringwald WR Forbes MK Wright AGC Meta-analysis of structural evidence for the Hierarchical Taxonomy of Psychopathology (HiTOP) model Psychol. Med. 2021 14 1 14 10.1017/S0033291721001902 33988108 Ringwald, W. R., Forbes, M. K. & Wright, A. G. C. Meta-analysis of structural evidence for the Hierarchical Taxonomy of Psychopathology (HiTOP) model. Psychol. Med.14, 1–14 (2021).10.1017/S0033291721001902 33988108

87.Littlefield AK Lane SP Gette JA Watts AL Sher KJ The ‘Big Everything’: integrating and investigating dimensional models of psychopathology, personality, personality pathology, and cognitive functioning Personal. Disord. 2021 12 103 114 32915005 10.1037/per0000457 PMC8318803 Littlefield, A. K., Lane, S. P., Gette, J. A., Watts, A. L. & Sher, K. J. The ‘Big Everything’: integrating and investigating dimensional models of psychopathology, personality, personality pathology, and cognitive functioning. Personal. Disord.12, 103–114 (2021).10.1037/per0000457 PMC8318803 32915005

88.Watts AL Sher KJ Heath AC Steinley D Brusco M ‘General addiction liability’ revisited Clin. Psychol. Sci. 2025 13 242 260 40241784 10.1177/21677026241245070 PMC12002852 Watts, A. L., Sher, K. J., Heath, A. C., Steinley, D. & Brusco, M. ‘General addiction liability’ revisited. Clin. Psychol. Sci.13, 242–260 (2025).10.1177/21677026241245070 PMC12002852 40241784

89.Garavan H Recruiting the ABCD sample: design considerations and procedures Dev. Cogn. Neurosci. 2018 32 16 22 29703560 PMC6314286 10.1016/j.dcn.2018.04.004 Garavan, H. et al. Recruiting the ABCD sample: design considerations and procedures. Dev. Cogn. Neurosci.32, 16–22 (2018).10.1016/j.dcn.2018.04.004 PMC6314286 29703560

90.Achenbach, T. M. & Rescorla, L. A. Manual for the ASEBA School-Age Forms & Profiles (Univ. of Vermont, Research Center for Children, Youth, & Families, 2001).

91.Achenbach TM Ruffle TM The child behavior checklist and related forms for assessing behavioral/emotional problems and competencies Pediatr. Rev. 2000 21 265 271 10922023 10.1542/pir.21.8.265 Achenbach, T. M. & Ruffle, T. M. The child behavior checklist and related forms for assessing behavioral/emotional problems and competencies. Pediatr. Rev.21, 265–271 (2000).10.1542/pir.21-8-265 10922023

92.Kotov R Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): I. Psychosis superspectrum World Psychiatry 2020 19 151 172 32394571 PMC7214958 10.1002/wps.20730 Kotov, R. et al. Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): I. Psychosis superspectrum. World Psychiatry19, 151–172 (2020).10.1002/wps.20730 PMC7214958 32394571

93.Michelini G Where do neurodevelopmental conditions fit in transdiagnostic psychiatric frameworks? Incorporating a new neurodevelopmental spectrum World Psychiatry 2024 23 333 357 39279404 PMC11403200 10.1002/wps.21225 Michelini, G. et al. Where do neurodevelopmental conditions fit in transdiagnostic psychiatric frameworks? Incorporating a new neurodevelopmental spectrum. World Psychiatry23, 333–357 (2024).10.1002/wps.21225 PMC11403200 39279404

94.Van Buuren S Broken stick model for irregular longitudinal data J. Stat. Softw. 2023 106 1 51 10.18637/jss.v106.i07 Van Buuren, S. Broken stick model for irregular longitudinal data. J. Stat. Softw.106, 1–51 10.18637/jss.v106.i07 (2023).

95.McArdle, J. J. & Wang, L. Modeling age-based turning points in longitudinal life-span growth curves of cognition. In Applied Data Analytic Techniques for Turning Points Research (ed. P. Cohen) 105–127 (Taylor & Francis Group, 2008).

96.Muggeo VMR Modeling temperature effects on mortality: multiple segmented relationships with common break points Biostatistics 2008 9 613 620 18308697 10.1093/biostatistics/kxm057 Muggeo, V. M. R. Modeling temperature effects on mortality: multiple segmented relationships with common break points. Biostatistics9, 613–620 (2008).10.1093/biostatistics/kxm057 18308697

97.Gundermann KP Change-point models for identifying behavioral transitions in wild animals Mov. Ecol. 2023 11 65 10.1186/s40462-023-00430-0 37864238 PMC10589947 Gundermann, K. P. et al. Change-point models for identifying behavioral transitions in wild animals. Mov. Ecol.11, 65 (2023).10.1186/s40462-023-00430-0 PMC10589947 37864238

98.Qian SS King RS Richardson CJ Two statistical methods for the detection of environmental thresholds Ecol. Model. 2003 166 87 97 10.1016/S0304-3800(03)00097-8 Qian, S. S., King, R. S. & Richardson, C. J. Two statistical methods for the detection of environmental thresholds. Ecol. Model.166, 87–97 (2003).

99.Aminikhanghahi S Cook DJ A survey of methods for time series change point detection Knowl. Inf. Syst. 2017 51 339 367 10.1007/s10115-016-0987-z 28603327 PMC5464762 Aminikhanghahi, S. & Cook, D. J. A survey of methods for time series change point detection. Knowl. Inf. Syst.51, 339–367 (2017).10.1007/s10115-016-0987-z PMC5464762 28603327

100.Imbens GW Lemieux T Regression discontinuity designs: a guide to practice J. Econom. 2008 142 615 635 10.1016/j.jeconom.2007.05.001 Imbens, G. W. & Lemieux, T. Regression discontinuity designs: a guide to practice. J. Econom.142, 615–635 (2008).

101.Hilbe, J. M. Negative Binomial Regression 2nd edn (Cambridge Univ. Press, 2011).

102.Gardner W Mulvey EP Shaw EC Regression analyses of counts and rates: poisson, overdispersed poisson, and negative binomial models Psychol. Bull. 1995 118 392 404 7501743 10.1037/0033-2909.118.3.392 Gardner, W., Mulvey, E. P. & Shaw, E. C. Regression analyses of counts and rates: poisson, overdispersed poisson, and negative binomial models. Psychol. Bull.118, 392–404 (1995).10.1037/0033-2909.118.3.392 7501743

103.Stan Development Team. The Stan Core Libraryhttp://mc-stan.org/ (2023).

104.Hoffman MD Gelman A The no-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo J. Mach. Learn. Res. 2014 15 1593 1623 Hoffman, M. D. & Gelman, A. The no-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo. J. Mach. Learn. Res.15, 1593–1623 (2014).

105.Gelman, A. & Rubin, D. B. Inference from iterative simulation using multiple sequences. Stat. Sci. 7, 457–472 10.1214/ss/1177011136 (1992).

106.Vehtari A Gelman A Gabry J Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC Stat. Comput. 2017 27 1413 1432 10.1007/s11222-016-9696-4 Vehtari, A., Gelman, A. & Gabry, J. Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Stat. Comput.27, 1413–1432 (2017).

107.Wood S. N. Generalized Additive Models: An Introduction with R 2nd edn 10.1201/9781315370279 (Chapman and Hall/CRC, 2017).

108.Hastie, T. & Tibshirani, R. Generalized Additive Models10.1201/9780203753781 (Taylor & Francis, 2017).

109.Pines A Psychiatric symptoms, cognition, and symptom severity in children JAMA Psychiatry 2024 81 1236 39196567 PMC11359114 10.1001/jamapsychiatry.2024.2399 Pines, A. et al. Psychiatric symptoms, cognition, and symptom severity in children. JAMA Psychiatry81, 1236 (2024).10.1001/jamapsychiatry.2024.2399 PMC11359114 39196567

110.Haist, F. & Jernigan, T. L. Adolescent Brain Cognitive Development Study (ABCD) - Annual Release 5.0. NDA10.15154/8873-zj65 (2023).

Come citare questo studio

Watts AL, Rosen AFG, Greene AL, King KM, Meyer FAC, Trull TJ, Sher KJ, Joyner KJ. (2026). Covariation among higher-order psychopathology dimensions is driven by nomorbidity and not comorbidity.. Nature. Mental health. DOI: 10.1038/s44220-026-00727-0

Consulta la pubblicazione originale ↗