Zoe Howard¹ (), Anna Chinni¹, Ruth Kimchi², Karen B. Schloss¹; ¹University of Wisconsin-Madison, ²University of Haifa

In information visualizations (e.g., charts, maps), visual features such as texture and color are used to represent concepts. Visualizations may be designed with particular sizes in mind (e.g., a printed page at arm’s length), but they can be scaled down for smaller displays (phone screens) or scaled up for larger displays (projected slides). Thus, it is important to understand how interpretations of visualizations change with scale. For color, differences must be magnified to maintain discriminability for smaller visualization sizes (Stone et al., 2014; Szafir, 2017). Here, we studied effects of scaling on perceived texture similarity. Our approach was motivated by work comparing similarity between original patterns of local elements vs. enlarged versions that were either (a) proportional to the original pattern (i.e. dilation; “zoom-scaled”) vs. unproportional (i.e., tiled; “repeat-scaled”) (Goldmeier, 1936/1972; Kimchi & Palmer, 1982). Original patterns of few local elements were more similar to zoom-scaled stimuli, whereas original patterns of many local elements were more similar to repeat-scaled stimuli. Similar reversals occurred for changes in element size, reflecting a shift from part-based to texture-based representations (Kimchi & Palmer, 1982). Using Kimchi and Palmer’s (1982) paradigm, we tested for analogous effects of scaling type for 36 textures commonly used in data visualizations, which varied in shape/fill/size/number (n=34; within-subject). Effects of number and size were similar to Kimchi and Palmer (1982), but we observed a new effect of element fill that interacted with size (p<.001). For filled elements, participants responded near chance for large elements and chose repeat-scaled patterns for small elements. But, for unfilled elements, they chose repeat-scaled patterns for patterns of large and small elements. These results suggest that texture fill modulates a shift of pattern perception from part-based to texture-based representations. These findings could have important implications for interpretations of textures with various scales in data visualizations.

Acknowledgements: NSF award BCS-2419493 to K.B.S; Hilldale Undergraduate/Faculty Research Fellowship University of Wisconsin, Madison