This project applied Jorgensen's three image retrieval tasks (1995) and Enser's four models of image retrieval (1995) to investigate end-users' image retrieval tasks and image query modes in the field of art history. Twenty-six undergraduate students majoring in art history were recruited in this study. They were required to finish a term paper including at least 20 images. The participants' image retrieval tasks were collected by pre- and post-search questionnaires, and a post-search interview. There was a significant difference between the mean number of search keywords or phrases participants planned to use and the mean number of search keywords or phrases they actually used. The study found a significant relationship between the level of success for the search results and the percentage of search keywords or phrases drawn from the topic title or topic description participants planned to use. Implications for image indexing tools and image retrieval system design are proposed.