1. Find evidence for the causal power of symptoms
An urgent line of research is to test the proposal that symptoms have causal power for other symptoms. Although the network approach presented by the target article is plausible, we believe that this proposal has yet to be supported by sufficient empirical evidence. The most directly relevant evidence would be a demonstration of the causal power of symptoms. Evidence for the causal power of symptoms could be obtained from cross-lagged effects, either long-term effects from longitudinal designs or short-term effects from experience-sampling studies (Conner et al. Reference Conner, Tennen, Fleeson and Barrett2009). Evidence for causality might even be obtained from experimental manipulation of symptoms (if ethical and minimally distressing).
Recent research on “normal” personality has taken a similar line. For example, extraversion and positive affect are robustly correlated (Lucas et al. Reference Lucas, Diener, Grob, Suh and Shao2000), which is often interpreted as resulting from a latent extraversion trait factor causing a latent positive affect trait factor. One of us has conducted experiments showing instead that state manifestations of extraversion (becoming extraverted for a moment, analogous to a disorder's symptom) cause the state of positive affect (McNiel & Fleeson Reference McNiel and Fleeson2006). That is, the states of normal traits have causal power, suggesting that the symptoms of mental disorders may also have causal power, and that manipulating them in experiments or tracking their lagged effects in experience-sampling studies may be fruitful.
2. Examine symptoms thoroughly
The network approach suggests a more thorough assessment of symptoms. If symptoms are more than indicators, then their patterns and frequencies of occurrence should be assessed. For example, experience-sampling studies, in which participants report symptoms every few hours for a few weeks, would provide rich information about the frequencies, patterns, and co-occurrences of symptoms (i.e., evidence of symptom co-occurrences for a given person across time or situations). This information is key for determining node strength and edge strength in the network. For example, the frequencies and co-occurrences of emotional volatility (Trull et al. Reference Trull, Solhan, Tragesser, Jahng, Wood, Piasecki and Watson2008), idealization, suicidal ideation, and emptiness on a daily basis could help determine the node strengths and edge strengths in Borderline Personality Disorder (BPD)
3. Investigate individual differences in edge strength
The incorporation of edges or connections between symptoms opens up new possibilities for the conceptual definitions of disorders and of diagnosis. Rather than individual differences in the levels of the symptoms being the primary component of a disorder, individual differences in the interconnections among symptoms may become an additional or central component to the disorder. If symptoms affect one another, then different people might have different strengths of causation linking symptoms. For example, some people might have a strong connection between perceived feelings of abandonment and self-injury, such that abandonment is usually followed quickly by self-injury, whereas others may have a weak connection between these symptoms. Such individual differences in strengths of connections between symptoms might determine whether a given individual experiences only one symptom or descends into a cascade of mutually activated symptoms; since the difference between these two outcomes comes from the strength of the connections, it might mean that having the disorder or not is a matter of having or not having strong connections. That is, everyone might be vulnerable to occasional symptoms (e.g., anger, perceptions of rejection), but only some may show an entire collection of symptoms.
Carried still further, it is possible that different individuals, all with relatively strong edges among some symptoms of a disorder, nonetheless differ in which edges are strongest. Thus, they all suffer a cascade of symptoms when any one symptom is activated, but they suffer different cascades from each other. This kind of heterogeneity in symptoms would be identified by within-person analyses of experience-sampling data, which would reveal individual differences in co-occurrences of symptoms, and aid in risk assessment.
4. Discover etiological processes for each symptom
If there were only one causal factor for a mental disorder, then it would be more plausible that there is only one etiological process for that mental disorder; namely, the one that produces the latent factor within the individual. However, if each symptom has causal power to set up a chain of consequences, then each symptom likely has its own etiological history. Research that identifies the triggers and processes resulting in each symptom would be useful. Again, experience-sampling data would reveal co-occurrences between triggers and symptoms.
5. Determine whether and why symptoms cohere into distinct mental disorders
A concern about the network approach is that, without single latent causes, disorders may lose coherence and become replaced by countless individual symptoms, creating a situation too chaotic for research or for treatment. Cramer et al. have proposed some plausible technical rules for identifying coherent disorders, but they have not provided evidence that such rules produced distinct clusters or disorders. Thus, it is necessary to determine whether those rules produce coherence, whether other rules are needed, or whether distinct disorders are hard to identify. Distinct disorders may also arise from multiple, small latent causes, which affect each other causally.
We are embarking on a 5-year project on the symptoms of BPD that undertakes some of these lines of research. The evidence for or against the network approach is still to be produced, but we are convinced that the approach supports lines of research that are likely to produce new important insights into disorders. In turn, this should lead to new treatments focused on the causal power of symptoms.
You have
Access
The network approach proposed by Cramer et al. suggests fascinating new directions of research on mental disorders. This commentary highlights the advantages of the target article's approach and proposes directions for research supporting it. In particular, the proposition that symptoms causally affect one another, rather than being mere signs or outcomes of an underlying latent disorder, is exciting. Although symptoms would be powerful even as mere outcomes (e.g., by being intensely aversive for self or others), ascribing causal power to symptoms raises them to an even greater level of importance as the very heart of psychological disorders.
1. Find evidence for the causal power of symptoms
An urgent line of research is to test the proposal that symptoms have causal power for other symptoms. Although the network approach presented by the target article is plausible, we believe that this proposal has yet to be supported by sufficient empirical evidence. The most directly relevant evidence would be a demonstration of the causal power of symptoms. Evidence for the causal power of symptoms could be obtained from cross-lagged effects, either long-term effects from longitudinal designs or short-term effects from experience-sampling studies (Conner et al. Reference Conner, Tennen, Fleeson and Barrett2009). Evidence for causality might even be obtained from experimental manipulation of symptoms (if ethical and minimally distressing).
Recent research on “normal” personality has taken a similar line. For example, extraversion and positive affect are robustly correlated (Lucas et al. Reference Lucas, Diener, Grob, Suh and Shao2000), which is often interpreted as resulting from a latent extraversion trait factor causing a latent positive affect trait factor. One of us has conducted experiments showing instead that state manifestations of extraversion (becoming extraverted for a moment, analogous to a disorder's symptom) cause the state of positive affect (McNiel & Fleeson Reference McNiel and Fleeson2006). That is, the states of normal traits have causal power, suggesting that the symptoms of mental disorders may also have causal power, and that manipulating them in experiments or tracking their lagged effects in experience-sampling studies may be fruitful.
2. Examine symptoms thoroughly
The network approach suggests a more thorough assessment of symptoms. If symptoms are more than indicators, then their patterns and frequencies of occurrence should be assessed. For example, experience-sampling studies, in which participants report symptoms every few hours for a few weeks, would provide rich information about the frequencies, patterns, and co-occurrences of symptoms (i.e., evidence of symptom co-occurrences for a given person across time or situations). This information is key for determining node strength and edge strength in the network. For example, the frequencies and co-occurrences of emotional volatility (Trull et al. Reference Trull, Solhan, Tragesser, Jahng, Wood, Piasecki and Watson2008), idealization, suicidal ideation, and emptiness on a daily basis could help determine the node strengths and edge strengths in Borderline Personality Disorder (BPD)
3. Investigate individual differences in edge strength
The incorporation of edges or connections between symptoms opens up new possibilities for the conceptual definitions of disorders and of diagnosis. Rather than individual differences in the levels of the symptoms being the primary component of a disorder, individual differences in the interconnections among symptoms may become an additional or central component to the disorder. If symptoms affect one another, then different people might have different strengths of causation linking symptoms. For example, some people might have a strong connection between perceived feelings of abandonment and self-injury, such that abandonment is usually followed quickly by self-injury, whereas others may have a weak connection between these symptoms. Such individual differences in strengths of connections between symptoms might determine whether a given individual experiences only one symptom or descends into a cascade of mutually activated symptoms; since the difference between these two outcomes comes from the strength of the connections, it might mean that having the disorder or not is a matter of having or not having strong connections. That is, everyone might be vulnerable to occasional symptoms (e.g., anger, perceptions of rejection), but only some may show an entire collection of symptoms.
Carried still further, it is possible that different individuals, all with relatively strong edges among some symptoms of a disorder, nonetheless differ in which edges are strongest. Thus, they all suffer a cascade of symptoms when any one symptom is activated, but they suffer different cascades from each other. This kind of heterogeneity in symptoms would be identified by within-person analyses of experience-sampling data, which would reveal individual differences in co-occurrences of symptoms, and aid in risk assessment.
4. Discover etiological processes for each symptom
If there were only one causal factor for a mental disorder, then it would be more plausible that there is only one etiological process for that mental disorder; namely, the one that produces the latent factor within the individual. However, if each symptom has causal power to set up a chain of consequences, then each symptom likely has its own etiological history. Research that identifies the triggers and processes resulting in each symptom would be useful. Again, experience-sampling data would reveal co-occurrences between triggers and symptoms.
5. Determine whether and why symptoms cohere into distinct mental disorders
A concern about the network approach is that, without single latent causes, disorders may lose coherence and become replaced by countless individual symptoms, creating a situation too chaotic for research or for treatment. Cramer et al. have proposed some plausible technical rules for identifying coherent disorders, but they have not provided evidence that such rules produced distinct clusters or disorders. Thus, it is necessary to determine whether those rules produce coherence, whether other rules are needed, or whether distinct disorders are hard to identify. Distinct disorders may also arise from multiple, small latent causes, which affect each other causally.
We are embarking on a 5-year project on the symptoms of BPD that undertakes some of these lines of research. The evidence for or against the network approach is still to be produced, but we are convinced that the approach supports lines of research that are likely to produce new important insights into disorders. In turn, this should lead to new treatments focused on the causal power of symptoms.