Correlational selection does not explain the evolution of a behavioural syndrome