Reading focus
Syllabus: What Would Justify a Causal Claim?
Match each design to the alternative explanation it helps address and state what remains assumed.
Why you’re reading it. Broaden the available comparisons when investigator assignment is unavailable or inappropriate.
A question to consider. What would a change in income policy help establish about children’s symptoms that an ordinary comparison of richer and poorer families would leave unresolved?
Why this reading
A correlation leaves several possible causal stories open. Rutter organizes natural-experiment designs around the alternatives they help address, from selection and reverse causation to shared genetic influences. The educational contribution is a discipline of matching a design to a specific threat, while making the remaining assumptions explicit.
Before you read
Be able to state the main threats to a causal reading of a correlation: reverse causation, confounding by a third variable, genetic mediation of an apparently environmental effect, and selection into the risk. Rutter's list overlaps with these but is organised for developmental questions.
Know what a counterfactual is: what would have happened to this person had they not been exposed. Every design in the paper is an attempt to approximate it.
Read for
- Identify the alternative explanations for a risk–outcome association.
- For each design, state which alternative it bears on and the assumptions required.
- Distinguish identifying a causal contribution from fully explaining an outcome.
- Notice where the author treats evidence as strengthening an inference rather than settling it.
Consider afterwards
- Choose a familiar correlational claim (screen time and adolescent depression, for instance). Which of Rutter's designs could be applied, what would each rule out, and what would remain?
- Randomised experiments are often called the gold standard. Which of Rutter's designs come closest, and what do they give up to be feasible?
Editorial context
Read this as a methodological synthesis, not a guarantee that a named design identifies a causal effect. Each application still depends on how exposure, comparison groups, timing, and outcomes are measured. The argument provides a bridge from prospective prediction to causal reasoning: specify the alternative explanation, explain why the design bears on it, and identify what the data cannot rule out.