The effect of tree structure on search in a hierarchical menu selection system

Abstract Search processes in a hierarchical menu selection system were investigated in a study that varied the structure of the tree. A hierarchical data base was composed of 256 gift items grouped into different clusters and presented using menus. Depth of the tree was held constant while breadth varied with level. Five structures were explored with the following number of alternatives at each of four levels: constant (4 × 4 × 4 × 4), decreasing (8 × 8 × 2 × 2), increasing (2 × 2 × 8 × 8), concave (8 × 2 × 2 × 8), and convex (2 × 8 × 8 × 2). Subjects searched for either specifically named gifts (explicit targets) or gifts appropriate for a scenario situation (scenario targets). In general, explicit targets took less time to find and fewer frames to traverse than for scenario targets. For explicit targets, the increasing menu was slightly superior to the rest. Search time was about the same across the five tree structures for explicit targets, but differed greatly for scenario targets. The concave and increasing structures were faster than the constant structure and the convex and decreasing structures were slower. Similar results were found for the number of frames traversed. The patterns of search also differed greatly among the five structures in terms of the frequency of use of the ‘previous’ command and the ‘top’ command. The ‘previous’ command was used most frequently with the convex menu and least frequently with the concave menu. For scenario targets, the ‘top’ command was used most frequently with the decreasing menu and least frequently with the increasing menu. For explicit targets, the ‘top’ command was used most frequently with concave menu and least frequently with increasing menu. The pattern of search indicated that if subjects moved back up the tree they tended to move to a level with eight choices rather than two. Overall, it is concluded that the concave menu is superior when searching for scenario targets and the increasing menu is slightly superior when searching for explicit targets. A theory of menu uncertainty based on information theory is proposed which helps to account for some of the results.