Academic journal article Human Factors

Multiple Factors That Determine Performance with Tables and Graphs

Academic journal article Human Factors

Multiple Factors That Determine Performance with Tables and Graphs

Article excerpt

INTRODUCTION

Graphic and alphanumeric displays of quantitative data are a major component in software applications, such as statistical packages, spreadsheets, and decision support systems. Despite the abundance of these displays, no appropriate guidelines yet exist for the choice of the best display for a given purpose (DeSanctis, 1984; Thomas, 1991). The results of the numerous studies on this subject are inconclusive. Regarding the merits of graphic displays over tables, some studies reported an advantage of graphs (e.g., Benbasat & Schroeder, 1977), others found an advantage of tables (e.g., Powers, Lashley, Sanchez, & Shneiderman, 1984), whereas a third group of studies revealed no difference between the two types of displays (e.g., Lucas & Nielsen, 1980). Comparisons among different graphic displays also fail to show clear advantages of one type of display over the others (e.g., Casali & Gaylin, 1988).

Apparently no display is optimal for all purposes, and instead of looking for a single generic best display, the choice of the optimal display might depend on a variety of variables. Seven categories of variables are necessarily involved whenever quantitative data are displayed.

1. The displays. Any application or study with quantitative displays uses one or several displays, which may be alphanumeric (such as tables) or graphic (such as line graphs, bar graphs, or pie charts). Numerous display types have been designed, and new ones are suggested frequently.

2. The graphic characteristics of the display. Displays of the same type (e.g., line graphs or bar graphs) may differ in features such as the width of the lines, patterns used to mark the different bars, spacing of the legends, use of a grid in the display, and use of colors. A number of books contain recommendations for the optimal design of the different displays (e.g., Kosslyn, 1994; see also Kosslyn, 1985, for a review of some earlier books), and attempts were made to use existing knowledge in cognitive psychology to avoid design errors in displays (Kosslyn, 1989). However, empirical studies of some generally accepted guidelines, such as the recommendation to order slices in pie charts or bars in bar charts according to their sizes, have not shown that adherence to the recommendations made any difference in terms of performance (Spence & Lewandowsky, 1991).

Thus what constitutes the optimal design for each type of display remains unclear. This causes one of the major problems in the empirical comparison of different displays. Whenever a study shows the advantage of one display over another, this advantage might be attributable to one display's design being closer to optimal, whereas the other display suffers from major design flaws. Unless the design of all displays is optimal, conclusions drawn from the comparisons must be regarded as tentative.

3. The conditions of presentation. Identical displays may be presented under very different conditions. Some of the conditions pertain to the visual conditions under which the display is seen, such as the visual angle, the illumination in the room, and the display-background contrast. Other conditions that might be relevant concern the circumstances in which a display is seen, such as time pressure, the presence of large quantities of additional information, and other situational factors. A display that is preferable under one set of conditions might not be appropriate under different conditions.

4. The complexity of the displayed data. Data sets, displayed in tables or graphic displays, differ in their complexity, which is defined in terms of three factors: (a) the number of points in the display (displays with fewer data points are less complex); (b) the configuration of the display (i.e., 12 data points may appear on one line of 12, two lines of 6, three lines of 4, or 4 lines of 3 points); and (c) the regularity of the displayed data (whether the different values exhibit some ordered pattern, such as clear linear trends, or whether they are dispersed erratically). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.