The Bruce and Young (1986) model of face recognition specifies that names are stored in a separate, phonological code, distinct from other semantic information, and able to be recalled only after that information. This paper re-evaluates the evidence that has been used as the basis for Bruce and Young's distinction, and shows that the previous literature can be equally well explained by a model of semantic memory that specifies a lexicon containing words separate from the conceptual store (the semantic/lexical model). To distinguish between the two models three experiments were conducted. In Experiment 1, 36 subjects learnt names, occupations, and towns of residence for 16 previously unfamiliar male faces. The names and towns varied for meaningfulness. The subjects were then tested for both accuracy and speed of recognition. The meaningful items were recalled more accurately than the meaningless items, irrespective of whether they were names or not. The reaction time data showed no significant differences between the various stimulus item groups. In Experiment 2, names were varied for meaningfulness more precisely, and 36 different subjects learnt names and towns for 16 faces. Meaningful items were again recalled more accurately than meaningless items. It was also found that names were recalled more often in isolation than towns. Experiment 3 examined whether the recall of names was contingent upon retrieval of semantic information. Thirty-six naive subjects learnt names and occupations for 16 faces. Both names and occupations varied in terms of whether they were confusable with each other or not. It was found that when occupations were confusable, and hence more difficult to recall, names were recalled as accurately as when the occupations were non­ confusable. This is contrary to the predictions of the Bruce and Young model, which states that name recall can only occur after the recall of other semantic items. A possible version of the semantic/lexical model is outlined in the General Discussion.