Manipulating the way a graph is drawn influences viewers' ability to extract information from it. In a series of experiments with simple bar graphs, the authors varied the rendering characteristics and relative heights of the bars and asked participants to estimate the quantities portrayed. The addition of 3-dimensional (3D) perspective depth cues lowered accuracy. This accuracy disadvantage diminished when a short delay was introduced before judgments were reported. The height of the judged bar relative to nearby graphical elements also affected accuracy; this effect was about 1 order of magnitude larger and remained intact when the delay was introduced. Nearby elements also affected viewers' bias (under-or overestimation). These effects do not seem to be due to misestimation of object depth. The results suggest that warnings about accuracy decrements due to 3D shading may be overstated, whereas distortions due to neighboring elements should be of more concern.

No abstract

Diagrams for presenting quantitative data are an important component of print communication. Their rate of use is high and rising. This reflects in part the recent development of software tools for generating data graphics. These programs allow a wide range of choices for data visualisation -some of which may be ugly or ineffective. How has graph usage evolved during this period? A survey of graph usage in academic joul'llals, magazines, and newspapers during the years 1985-1994 revealed several dynamic trends in the characteristics of data graphics, as well as robust differences between media. However, graph features that have been singled out by experts as poor choices, such as "3-D" rendering, do not seem to be on the rise. IntroductionGraphical presentations of quantitative information are a striking and ubiquitous component of print media. A well-chosen graphic can effectively communicate a large amount of information efficiently [18], and can make that information perceptually salient and memorable. Many popular news publications, and most scientific journals, make extensive use of data graphics. Recently, three developments in the study and technique of graphing have impacted on our understanding of how graphics function.In the past two decades a cottage industry has developed around critiques of poor design in statistical graphics. Jibes at inefficient, misleading, or inelegant graphs come from designers [26], statisticians [8,31]' and perceptual psychologists [16]. In all these cases, criticism is accompanied by constructive insight into how to produce graphics that are efficient, perspicuous, and elegant.At the same time, the behavioural study of how people extract quantitative information from visual displays has begun to blossom. This tradition has its roots as far back as at least the 1930s (see [20] for an excellent review), but since the early 1980s there has been a flowering of both empirical research (e.g. [5,6,10,12,[23][24][25]30]) and theoretical analysis [2,9,19,22]. creating quantitative graphics were the exclusive province of major media design shops and well-funded scientists, many in private industry. Since then, the explosion of computing power and the development of the graphical user interface have put sophisticated tools for producing statistical graphics in the hands of everyday producers and consumers of data.These three developments have contributed substantially to what we know about how data graphics work, and have provided a wealth of new abilities to produce ever more sophisticated figures. They also raise a question: How are graphs actually being used in contemporary communications? This question is relevant to the critical enterprise, because there is little interest in critiquing that which does not occur; also, there is reason to worry that the new technologies have made it not simply easier to produce graphs, but particularly easier to produce bad ones. It is relevant to the behavioural study of graphs, because empirical and theoretical understanding is most valuable for t...