A study was conducted examine whether the asymmetric confusability effect was generalized to an array of face photographs, and furthermore to investigate how the impression of faces affected the recognition of addition and deletion of the faces. In a preliminary investigation, 27 subjects rated the impression of 83 face photographs, and the photographs to be used in the present study were chosen on the basis of the impression scores. In the study, 40 subjects saw 14 photographs consisted of three or four faces and took a recognition test of unchanged photographs and changed photographs with a specific face added or deleted. The data showed that (a) the addition superiority was not found in the recognition of changes in face arrays; (b) the impression of faces differentially affected the recognition of addition and deletion changes in face arrays. These results suggest that the mechanism underlying the recognition of the deletion of a face may be different from that of addition.