Global Extrema Bias Perception and Recall of Average Data Values in Line Charts

Keywords

Data visualization, Memory, Perception, Shape, Global Maxima and Minima

Abstract

Experiments in visualization perception demonstrate that people can perceive positions highly accurately. However, position encodings can be susceptible to systematic biases depending on the intrinsic properties of the visualized data, such as its shape and cognitive processes of memory and perception. Using line charts as a case study, we investigate how the shape of data, such as local and global extrema, can bias the perception and recall of average data values. In two studies, participants estimated the average data values in a line chart by adjusting a slider with their mouse. We found that participants’ estimates were systematically biased toward the direction of the global extremum. When multiple salient extrema were present, estimates appeared influenced by several extrema simultaneously but ultimately leaned toward the global extremum. Notably, the strength of this bias varied depending on whether participants were perceiving or recalling the mean. This work advances our understanding of how extrema influence perception and memory, potentially exaggerating or underestimating critical trends and contributing to a skewed interpretation of data. These findings offer valuable guidance for the design of narrative visualization tools and data storytelling strategies.