Primary and secondary effects: some methodological issues