Data Visualization: Preference and Use of Two-Dimensional and Three-Dimensional Graphs SAMUEL H. FISHER III JOHN V. DEMPSEY ROBERT T. MAROUSKY University of South Alabama study considered the interplay of simple versus perspective graphical information on aesthetic preference, instructional effectiveness, and retention. Students in an introductory U.S. government course were presented with examples of 2-D and 3-D graphs and asked to choose which was pleasing to the eye and which was most useful in answering questions about the graph’s content. The results of this study indicated that when visual appeal was the only criterion, subject choices overall were approximately evenly matched. When subjects were required to extract information from graphs, they used simple graphs almost 3 times more often than elaborate graphs. Information drawn from bar and circle graphs was extracted more accurately than the other three types of graphs. This Keywords: data visualization, graphing, charts, datagraphics that 11 of us, students and teachers, inundated with so much information in printed form and retention are becoming a hectic exercise in separating the substantive from the decorative. Just the increase in information itself has contributed to what Wurman (1990) has referred to as &dquo;information anxiety.&dquo; For example, more than a decade ago, famed data/graphic designer Edward Tufte (1983, p. 10) estimated that each year, 900 billion images of statistical graphics are printed. Given the enormous increase in technological wherewithal in that period, that estimate should have increased exponentially. This trend is surely accelerating with the popularity of Web sites on the Internet. And not surprisingly, the use of statistical graphs has become common in introductory U.S. government texts. In spite of the magnitude of this graphical activity, few research-based guidelines exist concerning ways to present graphical information to learners other than to assert unequivocally that 2-D graphs should be used. A manual for the popular statistical analysis program SYSTAT, for instance, states that, &dquo;We cannot think of a single instance in which a perspective [3-D] bar graph should be used.&dquo; Later the authors of this manual concede, &dquo;You may have noticed that SYSTAT has a full assortment of 3-D graphs. We had to make some [their italics] concession to the marketplace. Nevertheless, before you use them, please consider 2-D alternatives&dquo; (SYSTAT, Inc., 1992, p. 19). One fascinating designation in SYSTAT’s clear recommendation is the phrase, &dquo;concession to the marketplace.&dquo; An underlying assumption in the use of 3-D graphs for students in are comprehension, learning, AUTHORS’ NOTE: Research for this article Research Council. was supported by a grant from the Umversity of South Alabama Social Science Computer Review, Vol 15 No 3, Fall 1997 256-263 © 1997 Sage Publications, Inc 256 Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 257 science courses is that the graphs’ visual appeal will generate increased student interest and retention of course material. The unanswered question in this process, however, is do students, who presumably are the reason graphical displays are made in the first place, really like 3-D graphs, and will they use them to respond to questions about quantitative data? In essence, the controversy regarding whether or not to use 3-D graphs, then, seems to be between the advocates of the &dquo;less is more&dquo; dictum and those who look for more &dquo;prominence&dquo; or projection in their graphical displays. As with other areas of instructional technology, designers, given more powerful tools and capabilities, are faced with the basic paradox-embellishment versus embodiment. With respect to graphical displays, Cleveland (1985), for example, would place the elements of volume and color hue or density (distinguishing features of perspective graphs) along the &dquo;worse&dquo; end of the graphical elements hierarchy. Several questions arise in regard to this argument. Are simple graphs always better? Does it really matter if learners use perspective graphs? What do learners prefer to use? Do their preferences change based on their need to use the information contained in the graphical displays? How do the graphical display types interact with recall and retention? The ideal of instruction using graphical displays is to present or have the student construct information in such a way that the learner will most easily be able to use and remember the data. Answers to the questions above will help in the construction of graphs that more closely and consistently reach the ideal. It is generally agreed from an information processing perspective that graphical displays are passed through the visual receptors of the retina and briefly stored in the iconic memory. From there it is passed to the short-term memory, which Miller (1956), in his classic article, suggests is limited to a modest number of &dquo;chunks&dquo; of information. The short-term memory allows time for encoding into long-term memory (in the form of schemata), where it is thought to be4’permanently&dquo; stored assuming no physiological damage takes place in the brain. On the surface, it would appear that information should be presented in such a way that it is most simply assimilated into short-term memory. Perspective, or 3-D, graphs have more nonessentials and logically should require a more involved chunking process. In addition, it is to be expected that an increase in visual illusions (Coren & Girgus, 1978; Gregory, 1969; Jahnel, 1952) entering the iconic memory would arise with the use of perspective graphs. Some knowledgeable theorists and practitioners see value in perspective graphs and charts. For example, C. F. Schmidt (1983), a highly respected authority on the design of statistical graphics, challenges those who would abandon 3-D charts. &dquo;Such a view is shortsighted and self-defeating since 3-D charts, when properly designed and executed, do have a significant role to play in statistical graphics.&dquo; Although admitting that an increase in skill and experience is necessary to prepare acceptable 3-D charts, he protests that inferior 3-D charts &dquo;are basically attributable to the ignorance and incompetence of the chartmaker, not the type or character of the chart per se&dquo; (p. 154). Undeniably, there are innate differences among different types of graphical displays. As long ago as the 1930s, for example, Croxton and Stein (1932) found that comparative judgments regarding size were more accurate for bar charts than circles and squares. Even so, all types of graphs and charts are used daily in such a way that perception is distorted because of unskillful design or, worse, purposely deceptive graphing. Many of these misrepresentations are the result of violations of what Norman (1993) refers to as the appropriateness principle, that is, &dquo;The representation used by the artifact should provide exactly the information acceptable to the task ... neither more nor less&dquo; (p. 97). The ready availability of 3-D graphical displays in off-the-shelf statistical programs and the increasing political Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 258 visual sophistication of learners only serve to confound designers’ efforts to follow the appropriateness principle. In the present study, our goal was to provide some practical information that may be of use to those faced with the thorny question of which tool to employ in order to show numerical comparisons or illustrate trends. SPECIFIC AIMS Specifically, we wish to consider the interplay of simple versus perspective graphical information (in essence, embodiment versus embellishment) on aesthetic preference, instructional effectiveness, and retention. Consistent with common usage, the terms &dquo;graphs&dquo; and &dquo;charts&dquo; are used interchangeably in this study (e.g., a pie chart is also referred to as a circle graph). The five types of graphical displays used in this study were chosen because they are among those commonly available microcomputer-based statistical software packages. Accordingly, this study addressed five major questions using five types of graphs. 1.Are 2-D or 3-D graphs preferred when it is not required that the viewer extract information from the graph? [visual preference] 2. Are 2-D or 3-D graphs used when the viewer must extract information from the graph to answer content-related questions? [use preference] 3. Do students answer content questions better when using one type of graphical display rather than another type? 4. Was there a difference between students who used simple graphs versus elaborate ones? 5. Which graphical displays promote retention? METHODS Undergraduate students taking an introductory U.S. government course, primarily freshand sophomores, were subjects in this experiment. Most of the students were not political science majors but were taking the course to fulfill a college core requirement. Participation in the experiment was voluntary; however, subjects did receive extra credit for participation. Sixty students participated in the first part of the experiment concerning preference and instructional effectiveness. Forty-six of these subjects chose to take the retention test (administered at a later time).’1 The graphical information, an analysis of the voting behavior of Congress’s Conservative Coalition, used in this study was taken from material that was normally presented in this course. Data on the Conservative Coalition’s voting behavior between 1976 and 1984 is used to present different graph styles. The fact that the data was not current or well-known to firstand second-year students meant that whatever they recall is from the charts and not a guess based on previous knowledge of the Conservative Coalition. Five types of common graphical displays were used to present the voting patterns over time. These were mixed bar/line, floating line, layered line, pie (circle graph), and bar. Consistent with what message design theorists (Wertheimer, 1938; Winn, 1992) consider to be good practice, materials presented by line graphs, illustrated trends, and bar and circle graphs showed comparisons. (To limit overload in the short-term memory, information was chunked in such a way that all graphs had five main parts or features, for example, five pieces of the pie in a circle graph.) As much as possible, the printed graphical displays were presented so that both simple and perspective charts were as clear as possible. All materials were displayed with black ink on opaque white paper (see Figure 1). men Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 259 Figure 1: Graph Examples First, we asked students to view 40 matched charts (20 simple/20 elaborate) (e.g., 2-D bar graph and 3-D bar graph) and indicate which of each pair they liked the most (or which was most pleasing to them). After the students answered these questions, the materials were collected. Next, we passed out a packet with the same 40 charts in a different order. Underneath each pair of charts were two questions. The first question concerned the substantive content of the charts. The second question asked which chart the students used to answer the content question. Then, these materials were collected. Four days later, a retention test, on the same content, but without the use of graphical displays, was administered. Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 260 RESULTS When visual appeal or preference was the only criterion, subject choices overall were rather evenly matched. Slightly more than 50% of the subjects in this study (51.8%) chose 2-D charts, whereas 48.2% chose 3-D charts. This was somewhat counter to our expectations based on the pilot study where we found the opposite. Looking at specific chart types (each was presented four times), layered line charts (50.8%, 2-D), pie charts (52.1%, 2-D), and bar graphs (52.9%, 2-D) closely followed this pattern. Subjects’ preference for simpler (2-D) floating line charts was stronger (72.4%, 2-D). On the other hand, a strong preference for 3-D mixed bar and line charts was observed. Specifically, only 30.8% of the subjects preferred 2-D mixed bar/line chart. Our second question was whether students would use simple (2-D) or elaborate (3-D) graphs in answering content questions. Overall, when it was actually necessary to extract information from the graphs to answer questions, students used simple graphs almost 3 times more often (73.6%) than elaborate graphs (25.7%). As Table 1 indicates, the partiality for 2-D graphs in this respect was very strong for all graph types except circle (pie) graphs. A Chi-square test of relationship among graphical displays and forms of graphs (2-D or 3-D) was significant (xz = 77.80, p < .000) but indicated a weak relationship (Lambda = .05, p < .05). question was concerned with which types of graphs helped students questions better. Overall, students in this study responded to 68.3% of the content-based questions correctly. As Table 2 indicates, information from bar graphs (82.5% correct) and pie charts (80% correct) was extracted more effectively than the other types of graphs. A Chi-square test of relationship among the type of graphs used to answer content questions and whether the correct answer was given was significant (x2 68.40, p < .000) but indicated a weak relationship (Lambda .05, p < .05). We were also interested to see if the percentage of correct answers from simple graphs was greater than that for elaborate graphs. Overall, students using simple graphs answered questions more correctly (69.9% correct) than those who answered using elaborate graphs (64.9%), although the difference was not significant. As Table 3 indicates, the lowest percentage of correct answers was rendered from the 3-D versions of floating line charts (45.2% correct) and the layered line chart (46.3% correct). There were no significant differences in any of the 2-D or 3-D comparisons for any of the graphical displays. Finally, we wished to know which graphical displays were best for retention. Unfortunately, even though the content was related to class material, the overall retention was only 31.5% correct. The best retention score (40.2% correct) resulted from information originally extracted from circle graphs. The worst retention resulted from information obtained through Our third research answer content = = bar charts (28.8% correct). IMPLICATIONS AND DISCUSSION Most of the students did not like 3-D floating line graphs. They did like 3-D mixed bar and line graphs more frequently. They were fairly even in preference of the other three types of 2-D and 3-D graphs. Based on this study, there appears to be little reason to use floating line charts where preference is a concern. Few individuals seem to like them for any reason. It is no surprise, however, that a reasonable percentage of individuals of college age do have a visual preference for other types of 3-D charts. Video games and stupendous television graphical effects are the stuff of everyday life for most young people. This study was purposely limited to black-and-white graphical displays to see if volume and shading alone, Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 261 TABLE 1 Form of Graph, 2-D or 3-D, Used to Answer Content Questions (percentages) TABLE 2 Percentage of Correct Answers to Content Questions by Graph Type TABLE 3 Percentage of Correct Answers to Content Questions by Graph Type and Form of Graph, 2-D or 3-D without full color, were sufficient to contribute to visual appeal. For almost half of the subjects in this study, they were. Students tended to use simpler (2-D) graphs much more frequently when it was necessary to extract information from the graph. A preference for 2-D charts as a more efficient form for answering questions about the content was in line with expectations. When it came to actually using the graphs to extract information, subjects tended to look for the clearest information giver, the 2-D graphs. The implication here is that, at least in the absence of color, simple graphs are preferred much more frequently than perspective graphs when the subjects had to actually use the information. Therefore, one would have to question why designers should use black-and-white perspective graphs for instructional purposes. This is not to suggest that they should never be employed for instruction, but, like a strong spice, they should be used sparingly for most individuals. Students tended to extract quantitative information more effectively from bar graphs and circle graphs. The use of bar graphs for effective representation of data has been promoted from many sources for some time (e.g., Hartley, 1987, p. 95). The finding of this study that circle graphs (or pie charts) were also effective was surprising. Some experts have long suggested that pie charts are less desirable because, among other reasons, they do not order Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 262 the data and numbers along a visual dimension (Bertin, 1983). Tufte (1983), never given to understatement, sums up this opinion when he states, &dquo;The only worse design than a pie chart is several of them&dquo; (p. 178). MacDonald-Ross (1977) is less skeptical yet still cautious about the use of pie charts, &dquo;Grouped bars seem the safest bet with range-graded keyed circles and pie charts as possibilities; but we still know too little about pie charts to feel entirely comfortable at their widespread use&dquo; (p. 375). Perhaps, in this decade when abominations such as 3-D floating line charts are not uncommon, circle graphs are comprehended more easily by visually sophisticated young adults. There was a larger percentage of correct answers from information obtained using 2-D graphs than from 3-D graphs. When it came to abstracting information from a graph, the results of this study support the notion that embellishment, in this case placing the chart in three-dimensional form, does not more accurately or more effectively convey the visual display’s data. Although there was a greater percentage of correct answers for simple than for perspective graphs for all types of graphical displays, the differences were not significant. CONCLUSIONS indicate that simple or 2-D graphs are more effective for information than are 3-D graphs when using black ink on white transmitting quantitative results of the retention test were poor, this may be in part the consequence the paper. Although of the experimental procedure and not lack of impact by the graph. What is especially pertinent to the presentation of the graphical information, however, is the broader issue of what questions graphs, charts, and other visual displays may be used to answer. Bertin (1983) and Wainer (1992) have proposed a three-level model of graphicity: (a) elementary questions involving data extraction, (b) intermediate questions involving trends seen in parts of the data, and (c) overall-level questions involving an understanding of the data’s deep structure. Another perspective would be to consider the types of information extracted from data to be rote, rule-driven, or evaluative. The poor retention of graphical information in this study may be due in part to the level of graphicity, which in this and most cases, was level one and two only. Another explanation for the poor retention evident in this study when graphic displays were used alone is information density. Graphs are a very information-dense system compared with the verbal language of text (Brasell, 1990). Although much time and effort can be saved viewing masses of data represented in graphical displays, the sheer amount of data absorbed through this process may reduce the ability to retain the information without the assistance of other media or symbol systems. Political science has a wealth of quantitative information that is useful for describing and explaining a wide range of concepts. It is important, therefore, to consider the potential shortcomings that translation of text into graphs has on the success of learners’ comprehension and retention. The results of this study NOTE 1. A pilot study was conducted with 19 students in an introductory U.S. government course. Results from this study suggested that 2-D graphs are preferred over 3-D graphs. However, the pilot study did not include a retention test. Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016 263 REFERENCES Bertin, J. (1983). Semiology of graphics. Madison: University of Wisconsin Press. Brasell, H. M. (1990). Graphs, graphing, and graphers. In M. B. Rowe (Ed.), What research says to the science teacher: The process of knowing. Washington, DC: National Science Teachers Association. . Monterey, CA: Wadsworth Advanced Books. Cleveland, W. S. (1985). The elements of graphing data . Hillsdale, NJ: Lawrence Coren, S., & Girgus, J. S. (1978). Seeing is deceiving: The psychology of visual illusions Erlbaum. Croxton, F., & Stein, H. (1932). Graphic comparison by bars, squares, circles, and cubes. Journal of the American Statistical Association , 27 , 54-60. . New York: McGraw-Hill. Gregory, R. L. (1969). Eye and brain Hartley, J. (1987). Designing instructional text. London: Kogan Page. Jahnel, F. (1952). Cheating with charts. In R. Modley & D. Lowenstein (Eds.), Pictographs and graphs: How to make and use them. New York: Harper and Brothers. MacDonald-Ross, M. (1977). How numbers are shown: A review of research on the presentation of quantitative data in texts. Audio Visual Communication Review , 25 , 359-409. Miller, G. (1956). The magical number seven, plus or minus two. Psychological Review, 63, 81-97. Norman, D. A. (1993). Things that make us smart. Reading, MA: Addison-Wesley. SYSTAT, Inc. (1992). SYSTAT for windows: Graphics (Version 5 ed.). Evanston, IL: Author. Schmidt, C. F. (1983). Statistical graphics: Design principles . and practices New York: John Wiley. Tufte, E. R. (1983). Visual display of quantitative information . Cheshire, CT: Graphics Press. Wainer, H. (1992). Understanding graphs and tables. Educational Researcher, 21, 14-23. Wertheimer, M. (1938). Laws of organization in perceptual forms in a source book for gestalt psychology. London: Routledge and Kegan Paul. Winn, W. (1992). Perception principles. In M. Fleming & W. H. Levie (Eds.), Instructional message design (2nd ed.). Englewood Cliffs, NJ: Educational Technology. Wurman, R. S. (1990). Information anxiety . New York: Bantam Books. Samuel H. Fisher III is an associate professor of political science. Address: Department of Political Science and Criminal Justice, University of South Alabama, Mobile, AL 36688. E-mail: [email protected]. John V. Dempsey is an associate professor of behavioral studies and education technology at the University of South Alabama. Address: College of Education, 3100 UCOMM, University of South Alabama, Mobile, AL36688.E-mail:[email protected]. Marousky is a Ph.D. student in the Department of Behavioral Studies and Education Technology of South Alabama. Address: College of Education, 3100 UCOMM, University of South Alabama, Mobile, AL 36688. E-mail: [email protected]. Robert T. at the U. Downloaded from ssc.sagepub.com at PENNSYLVANIA STATE UNIV on May 12, 2016
© Copyright 2026 Paperzz