Just as the existence of biological correlates of mental states does not imply that those biological processes impel the mental states with which they covary, the existence of trait correlates of mental states does not mean that these mental states are accounted for by trait descriptors. Inherent in the network approach of Borsboom et al. is that the “trait reductionism” of dimensional models of psychopathology cannot provide a complete explanation of psychological dysfunction and distress.
1. Neurobiological reductionism and trait reductionism
The Diagnostic and Statistical Manual of Mental Disorders (DSM) and International Classification of Diseases (ICD) have traditionally conceptualized psychological disorders categorically, but in recent years limitations of polythetic diagnoses based on combinations of symptoms have become increasingly apparent (e.g., heterogeneity within categories, excessive comorbidity across syndromes; see Herpertz et al. Reference Herpertz, Huprich, Bohus, Chanen, Goodman, Mehlum, Moran, Newton-Howes, Scott and Sharp2017). As a result, dimensional models of psychopathology have been presented as a psychometrically superior approach. Typically, these models conceptualize psychological syndromes using a series of trait dimensions, with patients assigned severity ratings that are combined into a trait profile to capture the central elements of a particular disorder. The Hierarchical Taxonomy of Psychopathology (HiTOP) is the most influential dimensional model currently in ascendance (Kotov et al. Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017), but others have been developed as well (e.g., Krueger & Markon Reference Krueger and Markon2014). It is clear that the next editions of the diagnostic manuals will implement dimensional frameworks for conceptualizing and diagnosing psychological disorders.
Like neurobiological models, trait models make intuitive sense to researchers and laypersons. Like neurobiological models, trait models capture one element of psychological dysfunction, but they cannot explain psychological disorders in all of their complexity. Borsboom et al.’s network model not only provides important context for understanding the strengths and limitations of neurobiological models, but of trait models as well.
2. Three neglected issues
Kotov et al.’s assertion that “not a single mental disorder has been established as a discrete categorical entity” (Kotov et al. Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017, p. 457) requires an exceedingly narrow definition of “categorical” (i.e., a wholesale shift in impairment once a threshold is met, along with virtually no symptomatic overlap or comorbidity with other syndromes). A number of categorical diagnoses in medicine (e.g., hypertension, autoimmune disorders, various pulmonary conditions) and psychology (e.g., anorexia, depression, narcissism) have imperfect thresholds and fuzzy boundaries, but are nonetheless clinically useful (see Hutchinson & Romero Reference Hutchinson and Romero2016; Lekkas & Mikhailov Reference Lekkas and Mikhailov2010). Kotov et al.’s (Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017) statement illustrates the importance of taking a more nuanced and critical stance in evaluating the strengths and limitations of different frameworks for conceptualizing and quantifying psychological dysfunction. Three issues are crucial in this effort.
2.1. A focus on process
Like neurobiological models, trait models do not capture cultural influences on the experience and expression of psychological distress, nor do they make a priori predictions regarding elements of the environment that exacerbate psychopathology and moderate its severity. Studies using ambulatory assessment techniques have documented trait-driven contextual variations in responding (Wright et al. Reference Wright, Hopwood and Simms2015); however, a complete understanding of psychological dysfunction requires use of experimental methods in which aspects of the environment are manipulated and the impact of these manipulations on cognition, behavior, and affect is assessed. Such studies exist (e.g., Bornstein Reference Bornstein2011; Horvath & Morf Reference Horvath and Morf2009), and while their results are not easily accommodated by trait models, they fit well with the network approach which specifies relations among elements in a network and generates predictions regarding how altering one feature of that network will affect other elements of the system.
2.2. Reification and misplaced concreteness
Just as the existence of biological correlates of mental states does not imply that those correlates account for the mental states with which they covary, the existence of trait correlates of mental states does not mean that mental states are accounted for by traits. Resistance to acknowledging the metaphoric underpinnings of neurophysiological constructs (e.g., “brain activity”; “cortical arousal”) and psychological concepts (e.g., “disinhibition”; “neuroticism”) can lead to “the error of misplaced concreteness” (Gargiulo Reference Gargiulo1998, p. 416) wherein descriptors are treated as immutable entities rather than descriptive labels. The search for useful metaphors is a key feature of scientific inquiry that cuts across disciplines (e.g., natural selection, molecular bonding, working memory), but these labels can have negative consequences as well (Bornstein & Becker-Matero Reference Bornstein and Becker-Matero2011). Just as certain neurobiological metaphors (e.g., “neural circuits”) are so familiar that they have taken on the appearance of concrete reality, the same is true of widely used trait metaphors (e.g., “facets of antagonism”).
2.3. Statistical integration and cross-method generalizability
Studies examining the underlying structure of psychological dysfunction using taxometric procedures suggest that continuous latent variables outperform categorical latent variables with respect to model fit and replicability (Haslam et al. Reference Haslam, Holland and Kuppens2012). When other statistical methods are used (e.g., latent class analysis, finite mixture modeling), some symptom clusters and dysfunctional behavior patterns coalesce into replicable individual differences that have qualities of network structures (Hallquist & Pilkonis Reference Hallquist and Pilkonis2012; Yun et al. Reference Yun, Stern, Lenzenweger and Tiersky2013). Psychopathology researchers must use a broader range of methodologies to capture patient functioning on multiple levels, including dynamics not amenable to self-report. Use of contrasting assessment methods that engage different psychological processes should be the norm within, as well as between, studies (Hopwood & Bornstein Reference Hopwood and Bornstein2014).
3. The integrative potential of network structures: Beyond psychological dysfunction
As Borsboom et al. have noted, insofar as mental disorders arise from the causal interplay among symptoms, it is unlikely that psychological syndromes can be explained conclusively in neurobiological terms. The same is true of trait frameworks, which do not provide a complete explanation of complex psychological phenomena. Looking ahead, network structure models may be a useful integrative framework linking psychological dysfunction with adaptation and impairment in other domains as well. Findings have documented that psychological and physiological processes interact to influence illness risk and moderate disease outcome (e.g., Cromer & Villodas Reference Cromer and Villodas2017); as network structure models expand across multiple contexts, they have the potential to enhance our understanding of the intra- and interpersonal dynamics that moderate health and illness – biological as well as psychological.
Just as the existence of biological correlates of mental states does not imply that those biological processes impel the mental states with which they covary, the existence of trait correlates of mental states does not mean that these mental states are accounted for by trait descriptors. Inherent in the network approach of Borsboom et al. is that the “trait reductionism” of dimensional models of psychopathology cannot provide a complete explanation of psychological dysfunction and distress.
1. Neurobiological reductionism and trait reductionism
The Diagnostic and Statistical Manual of Mental Disorders (DSM) and International Classification of Diseases (ICD) have traditionally conceptualized psychological disorders categorically, but in recent years limitations of polythetic diagnoses based on combinations of symptoms have become increasingly apparent (e.g., heterogeneity within categories, excessive comorbidity across syndromes; see Herpertz et al. Reference Herpertz, Huprich, Bohus, Chanen, Goodman, Mehlum, Moran, Newton-Howes, Scott and Sharp2017). As a result, dimensional models of psychopathology have been presented as a psychometrically superior approach. Typically, these models conceptualize psychological syndromes using a series of trait dimensions, with patients assigned severity ratings that are combined into a trait profile to capture the central elements of a particular disorder. The Hierarchical Taxonomy of Psychopathology (HiTOP) is the most influential dimensional model currently in ascendance (Kotov et al. Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017), but others have been developed as well (e.g., Krueger & Markon Reference Krueger and Markon2014). It is clear that the next editions of the diagnostic manuals will implement dimensional frameworks for conceptualizing and diagnosing psychological disorders.
Like neurobiological models, trait models make intuitive sense to researchers and laypersons. Like neurobiological models, trait models capture one element of psychological dysfunction, but they cannot explain psychological disorders in all of their complexity. Borsboom et al.’s network model not only provides important context for understanding the strengths and limitations of neurobiological models, but of trait models as well.
2. Three neglected issues
Kotov et al.’s assertion that “not a single mental disorder has been established as a discrete categorical entity” (Kotov et al. Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017, p. 457) requires an exceedingly narrow definition of “categorical” (i.e., a wholesale shift in impairment once a threshold is met, along with virtually no symptomatic overlap or comorbidity with other syndromes). A number of categorical diagnoses in medicine (e.g., hypertension, autoimmune disorders, various pulmonary conditions) and psychology (e.g., anorexia, depression, narcissism) have imperfect thresholds and fuzzy boundaries, but are nonetheless clinically useful (see Hutchinson & Romero Reference Hutchinson and Romero2016; Lekkas & Mikhailov Reference Lekkas and Mikhailov2010). Kotov et al.’s (Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017) statement illustrates the importance of taking a more nuanced and critical stance in evaluating the strengths and limitations of different frameworks for conceptualizing and quantifying psychological dysfunction. Three issues are crucial in this effort.
2.1. A focus on process
Like neurobiological models, trait models do not capture cultural influences on the experience and expression of psychological distress, nor do they make a priori predictions regarding elements of the environment that exacerbate psychopathology and moderate its severity. Studies using ambulatory assessment techniques have documented trait-driven contextual variations in responding (Wright et al. Reference Wright, Hopwood and Simms2015); however, a complete understanding of psychological dysfunction requires use of experimental methods in which aspects of the environment are manipulated and the impact of these manipulations on cognition, behavior, and affect is assessed. Such studies exist (e.g., Bornstein Reference Bornstein2011; Horvath & Morf Reference Horvath and Morf2009), and while their results are not easily accommodated by trait models, they fit well with the network approach which specifies relations among elements in a network and generates predictions regarding how altering one feature of that network will affect other elements of the system.
2.2. Reification and misplaced concreteness
Just as the existence of biological correlates of mental states does not imply that those correlates account for the mental states with which they covary, the existence of trait correlates of mental states does not mean that mental states are accounted for by traits. Resistance to acknowledging the metaphoric underpinnings of neurophysiological constructs (e.g., “brain activity”; “cortical arousal”) and psychological concepts (e.g., “disinhibition”; “neuroticism”) can lead to “the error of misplaced concreteness” (Gargiulo Reference Gargiulo1998, p. 416) wherein descriptors are treated as immutable entities rather than descriptive labels. The search for useful metaphors is a key feature of scientific inquiry that cuts across disciplines (e.g., natural selection, molecular bonding, working memory), but these labels can have negative consequences as well (Bornstein & Becker-Matero Reference Bornstein and Becker-Matero2011). Just as certain neurobiological metaphors (e.g., “neural circuits”) are so familiar that they have taken on the appearance of concrete reality, the same is true of widely used trait metaphors (e.g., “facets of antagonism”).
2.3. Statistical integration and cross-method generalizability
Studies examining the underlying structure of psychological dysfunction using taxometric procedures suggest that continuous latent variables outperform categorical latent variables with respect to model fit and replicability (Haslam et al. Reference Haslam, Holland and Kuppens2012). When other statistical methods are used (e.g., latent class analysis, finite mixture modeling), some symptom clusters and dysfunctional behavior patterns coalesce into replicable individual differences that have qualities of network structures (Hallquist & Pilkonis Reference Hallquist and Pilkonis2012; Yun et al. Reference Yun, Stern, Lenzenweger and Tiersky2013). Psychopathology researchers must use a broader range of methodologies to capture patient functioning on multiple levels, including dynamics not amenable to self-report. Use of contrasting assessment methods that engage different psychological processes should be the norm within, as well as between, studies (Hopwood & Bornstein Reference Hopwood and Bornstein2014).
3. The integrative potential of network structures: Beyond psychological dysfunction
As Borsboom et al. have noted, insofar as mental disorders arise from the causal interplay among symptoms, it is unlikely that psychological syndromes can be explained conclusively in neurobiological terms. The same is true of trait frameworks, which do not provide a complete explanation of complex psychological phenomena. Looking ahead, network structure models may be a useful integrative framework linking psychological dysfunction with adaptation and impairment in other domains as well. Findings have documented that psychological and physiological processes interact to influence illness risk and moderate disease outcome (e.g., Cromer & Villodas Reference Cromer and Villodas2017); as network structure models expand across multiple contexts, they have the potential to enhance our understanding of the intra- and interpersonal dynamics that moderate health and illness – biological as well as psychological.