Browsing by Subject "data visualizations"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    The role of numerical cognitive processes for bar graph comprehension
    (2023-05) Park, Jimin
    In the field of data visualization, there has been increasing recognition of the need to collaborate with the cognitive psychologists and investigate cognitive and mental processes that people recruit for graph comprehension. Previous studies have discovered that people perceive the length and width of bars to understand the numerical information in bar graphs. More generally, these two attributes are magnitudes. This raises the more general question of how people understand the magnitude information depicted in bar graphs. This dissertation research applies theories and experimental paradigms from numerical cognition, an area of study in cognitive psychology, to investigate whether people recruit number line representations and magnitude representations to understand the numerical values depicted in bar graphs. Across four experiments, undergraduate participants saw either two or three bars in a graph and judged which bar represents the greatest value. Experiment 1 investigated whether people prefer bar graphs that present bars horizontally, which aligns with the standard number line of mathematics, or graphs that present bars vertically, which is arguably a more conventional bar graph design. The results showed that participants judged the conventional vertical bar graphs faster than the number-line-congruent horizontal bar graphs. Experiment 2 further tested the conventionality versus number line congruity distinction by manipulating the orders of three bars in bar graphs. Participants made faster judgments when the three bars were presented in the conventional staircase order versus an irregular mixed order. Within the staircase order, participants showed comparable responses regardless of whether the bars were in ascending or descending order. This suggests that the number line representation, which places numbers left-to-right in increasing magnitude (i.e., the order effect) did not influence the performance. Experiment 3 examined whether people understand the magnitude information presented in bar graphs by the numerical values that the bars represent or by their physical sizes. The findings showed that participants’ comparison performance was driven by the sizes of the bars rather than the numerical values they represent. Participants’ performance was further predicted by the distance between the numerical values of bars and the scale of the axis. Experiment 4 investigated if gridlines serve as boundaries when comparing the bars in bar graphs. The results showed gridlines most benefit bar graph comprehension when it separates the bars on opposite sides. The findings of the current study introduce new theoretical constructs from numerical cognition to data visualization, inform how people parse magnitude information in bars graphs, and guide the design of better bar graphs for conveying the intended information.