Conventional methods for representing tree structures graphically emphasize the structure of nodes and edges rather than attributes such as node size. Shneiderman presents an algorithm for visualizing graphs as space-filling diagrams, with each node depicted as an area proportional to a selected attribute. The work originates in an attempt to find a better representation of disk storage space usage. A rectangle representing the entire tree is split vertically into rectangles whose areas are proportional to the sizes of the corresponding subtrees. The algorithm is recursive, splitting rectangles vertically for even levels and horizontally for odd levels. Care must be taken to ensure that abutting rectangles are of differing color or tone, or are otherwise distinguished by, say, a border. Shneiderman suggests that the representation provides a convenient way to identify anywhere in a directory hierarchy those large files that might be candidates for deletion in order to free up disk space, and that appropriate color coding can be used to identify other properties such as file ownership or category.
The representation emphasizes size at the expense of structure: it is possible to unravel the structure from the picture, and no doubt users would soon adapt to the different visualization. One possible development might be to combine the space-filling representation with an overlaid line drawing depicting the structure. The paper is brief: it would have been interesting to see some results from human factors testing.