Reading focus
Syllabus: What Is a Mental Disorder?
Trace how diagnoses that co-occur become evidence for comparing structural models. Keep the organization of the pattern separate from its explanation.
Why you’re reading it. Move from the performance of individual categories to relationships among them.
A question to consider. Several diagnoses tend to occur in the same people. Give two explanations for that pattern that could remain possible even if one factor model fits well.
Why this reading
Using common DSM-III-R diagnoses in the Dunedin birth cohort at ages 18 and 21, the authors examine models of internalizing and externalizing structure and their stability. The contribution is an influential longitudinal application: comorbidity becomes evidence to organize and explain, rather than only a complication for separate diagnoses. The paper did not originate the internalizing/externalizing distinction.
Before you read
A latent variable is not directly observed; researchers use it to represent a pattern among measured indicators. Confirmatory factor analysis compares a proposed pattern with the observed relationships among those indicators. Fit indices help assess how closely a model reproduces those relationships. Here, identify which diagnoses each model groups together and what differs between the models before interpreting their numerical fit.
It also helps to recall that the diagnoses being modelled are DSM-III-R categories, treated here as present or absent.
Read for
- The decision to treat diagnoses as fallible indicators of underlying dimensions rather than as the phenomena themselves. What does this assume?
- Which models are compared, and why a two-factor model rather than a one-factor or four-factor model is preferred.
- The stability argument: what it means that the same structure appears at 18 and 21, and what it does not mean.
- What the study cannot address. A good-fitting structure is silent about causes.
Consider afterwards
- Does a well-fitting factor model show that "internalizing" exists as a thing in the world? What alternative processes could produce the same pattern of co-occurrence?
- If the structure of disorders is dimensional and hierarchical, what follows for how a clinician should use a categorical diagnosis?
Editorial context
The study supports a structural description within its measures and population. A well-fitting latent model does not by itself establish a common causal entity, and a factor analysis of categorical diagnoses is not a direct test that each underlying condition is continuous. HiTOP develops a broader dimensional organization; taxometric evidence addresses a related but distinct question about latent categories. Keep those inferential steps separate.