Truncating Bar Graphs Persistently Misleads Viewers

Abstract

Data visualizations and graphs are increasingly common in both scientific and mass media settings. How might graphs mislead? In the present work, we provide empirical evidence across five studies for a clear, persistent means of misleading readers: the truncation effect for bar graphs. This effect occurs when a bar graph with a truncated axis exaggerates the perceived size of an illustrated difference. In the first study, we present participants with bar graphs on a variety of topics and manipulate y-axis truncation within-subjects. We find evidence for a consistent effect of truncation. In four subsequent studies, we explore whether this effect is modified by a warning about the nature of the manipulation, a 24-hour delay, and in a specialized participant population (PhD students in the humanities and quantitative fields). This truncation effect persists in each of these contexts, but is attenuated after warnings and in specialized participant populations.

The Science of Visual Data Communication: What Works

“…The prevalence of this deceptive effect has led to quantitative prescriptions for how to set y -axis boundaries to produce accurate measures of statistical effect sizes by typical viewers (Witt, 2019; B. W. Yang et al, 2021).…”

Graphical integrity issues in open access publications: Detection and patterns of proportional ink violations

“…From the perspective of visual perception theory, the principle of proportional ink is based on how humans perceive the size of symbols in graphs [18] and the ratio of entities in graphs [19]. According to empirical visualization studies, violations of these principles can lead viewers to misunderstand the underlying data and results [13].…”

“…Compared to text and image integrity analysis, they are comparability much less studied. Still, graph issues can lead readers to misinterpret information [13], affecting their ability to make decisions [14]. Even if the readers are trained to detect problems, they can still be confused or misled [15].…”

Visualization aesthetics bias trust in science, news, and social media

“…The more beautiful versions of the graphs were plotted with high image resolution (300 dpi), legible font size, Sans font type, and saturated color. The less beautiful version of the graphs were plotted with low image resolution (18 dpi for maps; 26 dpi for other graph types), smaller font size, Comic Sans MS font type, and desaturated color (grayscale for line plots and scatter plots; 50% saturation for other graph types).The manipulation of graph misleadingness was based on previous research on best practices for data visualization(4)(5)(6)(7)(8)(9). Specifically, the more misleading version of the bar plot was truncated (starting from above zero) to exaggerate the relative group differences.…”