Kylie Lin¹, Maxwell Paik², Cindy Bearfield¹; ¹Georgia Institute of Technology, ²New York University

How do people mentally represent the visual complexity of scenes and visualizations? One possibility is that visualization complexity is perceived holistically, as a gist impression that compresses detailed structure into a global summary (Sun & Firestone, 2021; Sun & Firestone, 2022). Alternatively, visualization complexity may be perceived as more fine-grained, involving feature-based processing that integrates local structure and interactions among visual elements (Attneave, 1957; Rosenholtz, Li, & Nakano, 2007). To examine how visual complexity is perceived in visualizations, we compared two methods for measuring perceived complexity: a numerical rating task and a two-alternative forced-choice (2AFC) comparison. The 2AFC paradigm is typically more discriminative as it reduces individual differences in scale use and can reveal subtler distinctions between stimuli (So, 2022). If perceived complexity reflects only a coarse global impression, we would expect both methods to produce similar results; if it encodes finer visual structure, 2AFC should yield greater differentiation. We conducted an experiment where 43 participants judged the complexity of a diverse set of data visualizations using one of the two tasks. Using bootstrapped samples, we found a significant difference between the two methods (reduced chi-squared=1.36, p<0.05). This effect was largely driven by image pairs where all participants agreed on which image was more complex in the 2AFC task. Rating-scale data underrepresented these differences in perceived visualization complexity, suggesting that complexity perception preserves some fine-grained information beyond a global gist. These results demonstrate how methodological sensitivity can reveal the granularity of perceptual representations and inform broader understanding of how people process and communicate visual information.