- What is the effect to be studied? - What is the ideal experiment, if there were no external constraints (ethics, budget, etc.)? - What is the nature of the study (observations, surveys, or experiments)? - What is the method for statistical inference? \+ What are the FUQ (fundamentally unanswerable questions)? - E.g. it is impossible to control for age at which a child enters school when studying the effect of length of education on test scores (or vice versa).