Picture-headline interaction (PHI) in print ads is further divided into Repetition, Substitution, Continuation, and Assistance in this study, and in each classification, relevance and originality are manipulated in four 2x2 between subject analysis of variance experiments to investigate their relationship with memory. The result reveals that the two variables have different effects depending on how picture and headline interact. In the Repetition group, relevance has a significant main effect, but originality does not. In the Substitution group, high originality (HO) ads are recalled more than those of low originality (LO) only in the low relevance (LR) condition. In the Continuation group, LO has a better memory effect than HO in high relevance condition (HC), and in the LO, subjects recall more HC ads than LC ones. The manipulation does not affect memory in the Assistance group. Research in the future can consider (1)taking Level of Processing as a possible explanation of PHI's memory effect, and (2) including involvement as an another variable.