One of contemporary psychology's more curious features, long noted (Cervone Reference Cervone1991; Mischel Reference Mischel1973), is that the strategies through which investigators pursue two of the fields' primary goals are discordant. One goal is to understand the workings of the mind. Although there may be disagreement on the details – the precise nature of mental mechanisms and abilities, and the degree to which they are innately specified or developed through interaction with the physical and social world – there is consensus at a broad strategic level. The mind is a complex system of distinct yet functionally related parts (e.g., Barsalou Reference Barsalou1999; McClelland & Rogers Reference McClelland and Rogers2003). Multiple interacting subsystems, both cognitive and affective, contribute to experience and action (Cervone Reference Cervone2004; Mischel & Shoda Reference Mischel and Shoda1995; Sander et al. Reference Sander, Grandjean and Scherer2005). “Observable behavior is an interaction effect par excellence” (Fodor Reference Fodor1983, p. 1).

The other goal is to assess qualities of the individual. Assessment efforts commonly describe people with respect to constructs identified in latent variable analyses. In clinical diagnosis, the constructs are diagnostic categories. In personality trait psychology, they are dimensions (Costa & McCrae Reference Costa and McCrae1992). In either case, as Cramer et al. highlight, the common cause hypothesis of latent variable modeling dictates that the observable behaviors indicative of a given category or dimension are not functionally related to one another. Local independence among the indicators is assumed.

At the level of substantive theory, few psychologists are likely to embrace the common cause hypothesis and its consequences. Clinicians commonly reject essentialist views in which a diagnostic category corresponds to a singular cause (Ahn et al. Reference Ahn, Flanagan, Marsh and Sanislow2006). Investigators who employ latent variable modeling in their research may, when pressed, abandon its common cause assumptions on theoretical grounds (see Cervone et al. Reference Cervone, Shadel, Smith and Fiori2006). “Problem and method,” then, “pass one another by” (Wittgenstein Reference Wittgenstein1953/2001, p. 197). Even investigators who recognize that actions, affects, and cognitions interact adopt latent variable methods that obscure these interactions from view.

The resulting conceptual contrast is just as sharp as Cramer et al. suggest. Consider a standard psychological science account of a person's tendencies to experience emotions such as anxiety and fear (“symptoms,” in the case of disorders). The affective scientist might explain them by reference to interactions among physiology, cognitive appraisals processes, and enduring beliefs (e.g., Sander et al. Reference Sander, Grandjean and Scherer2005). A developmentalist would add that temperament, which itself develops partly through environmental interaction, contributes to this interacting system of cognitive and affective elements (Schmidt & Fox Reference Schmidt, Fox, Cervone and Mischel2002). Importantly, in any such account, emotional tendencies such as anxiety and fear are the explananda, the phenomena to be explained. The interacting mental systems are the explanans.

The latent variable approach turns this world upside down (Cervone Reference Cervone, Cervone and Shoda1999). In comorbidity analyses, the high-level psychological tendency “internalizing” (Krueger Reference Krueger1999) does explanatory work: It explains correlations among lower-level locally independent tendencies such as anxiety and fear. The tendencies to experience anxiety and fear, in turn, do more explanatory work: They explain correlations among yet lower-level locally independent tendencies such as the experience of panic and social anxiety. And so it goes; it is locally independent tendencies all the way down. Not only are there no interactions among indicators of a given construct; there also is no explanation of experience and action by reference to well-defined systems of mind or brain.

This might be palatable if investigators were careful to describe latent variables merely as clusters of interrelated behaviors. But, instead, they commonly discuss them in terms normally reserved for structural entities with causal power (Cervone Reference Cervone2005). For example, internalizing, a construct that summarizes between-person correlations among indices of psychological distress, is said to be a “substrate” (Krueger Reference Krueger1999, p. 926) of mental disorders. This is not unlike a geologist positing a substratum of “destructiveness forcefulness” to explain a region's tendency to experience both volcanoes and earthquakes.

Explanations that reference abstract tendencies of the sort identified in latent variable analyses are seductive (Kagan Reference Kagan1998) – so much so that, once, even Cramer et al. are enticed. Did they really mean to say that “neuroticism” – a latent variable that reflects intercorrelations among dispositional tendencies to experience anxiety, hostility, self-consciousness, impulsiveness, vulnerability, and depression (Costa & McCrae Reference Costa and McCrae1992) – “can trigger the onset of depression” (sect. 4, para. 5, emphasis added)? Neuroticism is ripe for network analysis. One would not claim that it “triggers” depression, since depression is part of neuroticism, and neuroticism is conceptualized as a constant (see Borsboom et al. Reference Borsboom, Mellenbergh and Van Heerden2003).

The great virtue of Cramer et al.'s article is that they not only articulate a problem, but provide a solution: network analysis. Future work might expand their current scope. In principle, networks could include functional relations not only among observable behaviors, but among cognitive and affective components as well; empirical evidence documents numerous functional relations, such as the influence of self-consciousness on emotion (Mor & Winquist Reference Mor and Winquist2002), self-efficacy perceptions on motivation (Bandura Reference Bandura1997), knowledge structures on self-appraisals (Cervone et al. Reference Cervone, Caldwell, Fiori, Orom, Shadel, Kassel and Artistico2008), and mood on self-evaluations (Cervone et al. Reference Cervone, Kopp, Schaumann and Scott1994). One might account for personality traits in the manner Cramer et al. account for diagnostic categories and comorbidities. Functional relations among perceived self-efficacy, personal goal-setting, and disciplined, persistent behavior (e.g., Bandura & Cervone Reference Bandura and Cervone1986), for example, might enable one to view conscientiousness (whose components include competence, achievement striving, self-discipline, and dutifulness; Costa & McCrae Reference Costa and McCrae1992) as a cluster of functionally interrelated cognitive–affect elements and their behavioral effects.

Such an effort requires an assessment method that taps these cognitive and affective elements. A social-cognitive approach to assessment (Cervone et al. Reference Cervone, Shadel and Jencius2001) is apt in that it addresses “functional relations among affect and physiological arousal, cognition, and action” (Cervone et al. Reference Cervone, Shadel and Jencius2001, p. 41) rather than latent variables measured by independent indicators. Social-cognitive methods, and recent clinical assessment efforts (Haynes et al. Reference Haynes, Mumma and Pinson2009), are sensitive to individual idiosyncrasy, thus addressing Cramer et al.'s recognition of possible individual-level variability in network structure.

For more than four decades, psychologists have called for assessment and measurement strategies that align with the body of knowledge available in psychological science (Mischel Reference Mischel1968). Cramer et al.'s contribution is a most valuable step in this direction.