Implications of the Flynn effect for age–cognition relations