Reading focus
Syllabus: How Do Psychological Disorders Develop and Persist?
Compare the different routes from an observed association to a causal claim.
Why you’re reading it. This supplies the standards used to examine the studies that follow.
A question to consider. If an exposure predicts symptoms, what comparison would distinguish its effect from a shared family influence, and what could that comparison still miss?
Why this reading
A risk factor can predict a disorder without causing it. Thapar and Rutter explain how naturally occurring comparisons can help separate an exposure from other influences. Their central contribution is to ask what different designs rule out, and why agreement across designs with different weaknesses can strengthen a causal explanation.
Before you read
A confounder helps account for both an exposure and an outcome. Reverse causation means the developing difficulty influences the supposed risk. Selection concerns why some people experience an exposure while others do not. In a natural experiment, researchers use variation they did not randomly assign. The paper uses this term broadly, including family comparisons and instrumental-variable approaches. No design becomes assumption-free because it is called natural.
Read for
- Use Table 1 to distinguish an associated characteristic, a predictor that precedes an outcome, and a causal risk factor. What evidence moves a claim from one category to another?
- For two designs, identify the comparison and the alternative explanation it addresses. Table 2 also shows why a design useful for postnatal exposures may not separate prenatal influences.
- Follow the argument for triangulation. Ask whether converging findings have different potential biases, rather than simply counting studies with similar results.
Consider afterwards
- A stress exposure predicts later symptoms among unrelated people and within sibling pairs. Which explanations become less plausible, and which could survive both comparisons?
- What would be lost by replacing a carefully chosen comparison with a longer list of statistical control variables?
Editorial context
This is a methodological review using psychiatric examples available in 2019. It supports scrutiny of causal assumptions, not a claim that every example establishes an enduring causal verdict. Family comparisons can retain measurement error and individual-specific confounding; instrumental variables require their own exclusion assumptions. The authors also discuss limited representativeness and differences between unusually severe exposures and everyday conditions.