There is a basic assumption surrounding the relationship between images and texts which has led to their specialised treatment in information retrieval systems. Images are pictorial and non-verbal, while texts are verbal. While there may indeed be a true image-text dichotomy, we would argue that the nature of the relationship is not well-enough understood to design image storage and retrieval systems around this assumption. Human information seeking in texts and images must be better understood before truly usable systems can be designed. The nature of human image processing and retention, which suggests that details are less important than an overall schema, has implications for the amount of detail necessary to enable searching. Images are also retained in a highly structured manner; this would suggest that textual indexing assigned to an image should maintain these structured relationships, rather than simply being assigned in a linear manner. By developing a practical application of a "visual thesaurus," a system which provides flexible search capabilities yet maintains authority control, we intend to explore this assumption and develop some new approaches to image indexing and retrieval which incorporate spatial and iconic representations.