Human perception plays an important role in the field of visualization. Understanding perception can significantly improve both the quality and the quantity of information being displayed. The National Science Foundation’s experts on graphics and image processing cites the importance of perception, which proposed the term “scientific visualization”. The need for perception has been emphasized again recently by DOE/NSF panel on the directions for future research in visualization.This essay will briefly introduce the role of perception in visualization and why it is important.
Graph design for data analysis and presentation needs a scentific foundation. In the artical “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods”, Cleveland and McGill tried to formalize visualization studies scientifically. They indentified a set of elementary perceptual tasks when people received quantitative information from graphs. They introduced the theory that some types of graphical representations could have a significantly higher impact on viewers’ judgment of other forms of quantitative information. To do this, they used an ordered sorting to redesign old graphical forms to provide alternative graphical forms and assess whether a specific perception task was the key factor in determining chart performance. They found that as the distance between the two values being judged increased along the axis, the accuracy of the audience’s judgment decreased. Besides, location judgment is more accurate than the length judgment. They also found that the dot plot might be perceived more accurately than a pie chart or a bar chart, and the bar chart is better than a pie chart or split bar chart.
Legge, Gu and Luebker further used the concept of efficiency to study graphical perception. Their findings extended the results of Cleveland and McGill by quantifying the superiority of graphs over numerical tables. The way to evaluate the effectiveness is to measure the efficiency with which observers extract information from the graph.Perceptual efficiencies are very high for scatterplots, compared to numerical tables. The major contributor to the superiority of graphical displays is spatial parallel processing. Efficiency was lowest for the numerical tables and depended more strongly on sample size and viewing time.
Heer and Bostock, in their article “Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design”, reviewed the relevance of crowdsourcing in assessing visual and graphic design and found it particularly attractive because it is a low-cost, often immediate and effective way to get useful information about the visual design. However, they do need to validate the credibility of the participants, since crowdsourcing raises issues related to ecological effectiveness, motivation and experience, and lacks control over many experimental conditions. They assessed the viability of using Amazon Mechanical Turk to evaluate visualizations and then used these methods to obtain a new insight of visual design. They discovered that crowdsourcing of research with graphics could be feasible The scalability of the platform fills the difference in the results as more subjects can participate at the same cost.
Finally, to focus on the effects of chart size and layering on the graphical perception, Heer, Kong and Agrawala compared several graphs for visualizing time series data. They measured the speed and accuracy of subjects’ extract of value differences between charts using horizontal line plots of a range of chart sizes. They found that estimation error increases as chart height decreases. Error levels hold relatively stable at high virtual resolutions, but increase linearly at smaller resolutions.