This study investigates end-users' image queries by comparing the features of the queries to those identified in previous studies proposed by Enser and McGregor (1992), Jorgensen (1995) and Fidel (1997) in order to discover the utility of these existing features for the art history field and identify any expansions or new features. It also examines relationships between user search tasks and image query modes. Enser and McGregor's categories of Unique and Non-unique, and Jorgensen's classes of Location, Literal Object, Art Historical Information, People and People-related Attributes received high degrees of matching by three reviewers. This finding can be applied to add more details to Enser and McGregor's four categories (Unique, Non-unique, Unique with refiners, and Non-unique with refiners) and to re-group Jorgensen's 12 classes of image attributes. This study also found several significant relationships between the participants' retrieval tasks and query modes. The participants who used fewer keywords from their topic title and topic description had a larger number of keywords or phrases they planned to use. A significant difference was found between the mean of the search keywords or phrases participants planned to use and the mean of the search keywords or phrases they actually used. The participants who submitted a higher number of keywords or phrases they planned to use were able to draw more pictures representing their topic title and the participants who had a greater number of keywords or phrases actually used also generated a larger number of post-search drawings. A significant relationship was also found between the level of success and the number of keywords or phrases participants planned to use. Implications for 2 curriculum design in the art history field, library instruction, image indexing tools and image retrieval system design are proposed.