The confounded signal hypothesis maintains that facial expressions of anger and happiness, in order to more efficiently communicate threat or nurturance, evolved forms that take advantage of older gender recognition systems, which were already attuned to similar affordances. Two unexplored consequences of this hypothesis are (1) facial gender should automatically interfere with discriminations of anger and happiness, and (2) controlled attentional processes (like working memory) may be able to override the interference of these particular expressions on gender discrimination. These issues were explored by administering a Garner interference task along with a working memory task as an index of controlled attention. Results show that those with good attentional control were able to eliminate interference of expression on gender decisions but not the interference of gender on expression decisions. Trials in which the stimulus attributes were systematically correlated also revealed strategic facilitation for participants high in attentional control.