|
Full vs. Limited Context Hierarchical Menu Structures for Web Site Navigation with Frames |
|
1. INTRODUCTION In its relatively short existence, the World Wide Web (a.k.a. the Web) has, for many people, become a standard source for information of any kind. There are literally hundreds of thousands of web sites that could give one access to anything from the latest on Bill Clinton's woes to recent statistics the top ranked schools in the U.S. with just a few clicks of a mouse. But it is those "few clicks of a mouse" that this experiment is concerned with. Or more specifically, how individuals are accessing the information available on any given web site. Today's web sites are structured in a variety of ways. With respect to navigation, some of these web sites are constructed more effectively than others. The sites this experiment is concerned with are the ones that utilize frames as a method of navigation. This is usually achieved by incorporating a menu structure with one frame and using a second, larger frame as a display (see www.marvel.com). The most common practice for this method of navigation is to have one level of links to navigate an entire web site within the "menu frame". However, this method becomes problematic for large sites that have to be navigated with a hierarchical menu. Alternative methods include constructing a full context or limited context hierarchical menu structure. The full context menu is a kind of expanding menu (see www.virgin.com). This menu allows users to view the current and previous levels of the menu hierarchy as they browse through the categories they have selected. In layman's terms, once a user clicks on one link out of a set of links, another set of links (usually indented) that corresponds to their selection appears under it while the original set still remains displayed on the screen. The limited context menu, on the other hand, is a kind of sequential or reloading menu. This menu allows users to view only the current level of the menu hierarchy as they browse through the categories they have selected. In other words, once the user clicks on one link out of a set of links, another set of links that corresponds to their selection replaces the original set. There are definite advantages (as well as disadvantages) to both the full context and limited context structures. No matter how deep into the menu hierarchy a user may be, the full context menu allows the user to traverse the path of any visible category on higher level of the menu hierarchy. With the limited context menu, to traverse the path of another category on higher level, the user must backtrack to that level. However, by utilizing space more conservatively, the limited context menu can decrease the amount of links that appear inside of the frame. With the full context menu, as the menu expands with links, the frame may become cluttered up and very difficult to read. The purpose of this experiment is to evaluate these two structures to find which is more beneficial for web site navigation with respect to time, accuracy, and user preference. It is predicted that the results of this experiment will show a significant advantage for using full context menu structures over limited context menus. This is, in actuality, the same hypothesis that Zaphiris (1999) formulated when he conducted his experiment, which is the basis for this experiment. He attempted to evaluate the effectiveness of full context and limited context menus with respect to time, accuracy, and user preference at depths of two, three, and four. The results of Zaphiris' experiment suggest that by reducing the depth of the menu hierarchy, user performance will improve with respect to speed and search efficiency. His results also found that the full context menu structures resulted in poorer performance than the limited context structures, especially when the depth increased to four. However, it must be noted that he found no statistical significance in his results. This experiment will attempt to further Zaphiris' research by extending his findings on menu design for the Web and the findings of other researchers on hierarchical menu design to practical usage on the Web. In designing the interface for this experiment, the same researchers that Zaphiris consulted for his experiment will also be referenced. One such researcher includes Kiger (1984), who evaluated five treatments for menu design (2^6, 4^3, 8^2, 16x4, and 4x16) and found that reaction time and number of errors increased with the depth of the menu structure. Furthermore, the results of his experiment showed that the 4x16 (two levels with four items on the first level and sixteen items for each four on the second level) design had the faster reaction time and the fewest errors. And, out of the five designs, subjects preferred the 8^2 (two levels with a set of eight items on the first level and another eight for each the first set on the second level) design because of its relatively small depth. Based on these results, the hierarchical menu structure for this experiment was designed with a maximum depth of three and a maximum breadth of eight. Two other researchers Zaphiris consulted were Norman and Chin (1988). They recommend that along with small depth, greater breadth at the root and at the leaves of hierarchical menu structures should also be emphasized. In accordance with this recommendation, the menu structure for this site was designed with the maximum amount of links per level (eight) emphasized at the roots and at the leaves. Zaphiris then found a way to define and calculate user "lostness" by looking at the research of Larson and Czerwinski (1998). Their strategy in calculating "lostness" entailed comparing the number of unique and total links visited with the "optimal" path. In this experiment, that strategy is used to measure the users accuracy of information retrieval by defining the "optimal" path as the smallest number of visited links that are needed to complete the list of tasks for one treatment of the experiment (see independent variables). Lastly, Zaphiris completed the design of his interface by complying with observations made by Norman (1990), who commented that with previous studies, the focus of the experiment was the symmetrical menu structure. In a symmetrical hierarchical menu, each level has the same number of items, and each item has the same number of levels. However, the majority of real-world applications typically have asymmetrical structures. In an asymmetrical hierarchical menu, the number of items varies for each level as do the number of levels for each item. Therefore, the menu structure of this experiment was designed asymmetrically. As stated before, this experiment will further the research of Zaphiris' experiment. It will take into consideration Zaphiris' research and all of the aforementioned research. And despite Zaphiris' results (none of which prove to be statistically significant), it will attempt to prove that within a frame structure, the advantages offered by the full context menu structure make it more efficient for web site navigation than the limited context menu structure. |
 |
Department of Computer Sciences Direct questions and comments to the student editorial team |