Form class and intralist similarity in paired-associate learning