This study explores how study context and perceptual processing demands in picture recognition affect attention allocation and, in turn, memory encoding, as well as the applications of current findings in teaching. The repetition priming paradigm was adopted in two experiments. Relative magnitudes of picture-picture (P-P) and word-picture priming effects were compared. The results showed that when pictures and words were mixed in the study phase or when the recognition of the picture required processing of local perceptual features, attention was perceptually directed and resulted in perceptual encoding of the picture. If, however, the recognition of the picture did not require local processing and the surface format of pictures and words did not lead to contrast in subjects, no evidence of perceptual encoding for pictures was found. These findings suggest that: (1) memory encoding of pictures is not a simple process even though pictures are customarily regarded as an‘easy-to-learn’text adjunct; (2) pictorial encoding is determined by whether and how study context and processing demands affect the manner in which attention is allocated; (3) in devising a pictorial learning aid, learning activity should be so arranged that it properly directs students’attention allocation and the adeptness of the learner in pictorial processing should also be understood.