Abstract
Graphical perception studies have largely focused on how accurate participants can be at extracting quantitative information from glyphs in various visualization types, including infographics. However, most studies assume that there is only one correct way to encode information, for example to compare two circles the encoding mechanism must be relative area. We suspect that this assumption is over-simplistic, especially for glyphs that have an obvious two- dimensional structure, and that some participants use relative height instead. This is a crucial, but it seems often overlooked, aspect of graphical perception studies, and indeed of data visualization in general. If participants are not using the same encoding mechanism that the designer has used, this represents a clear systematic bias that can induce large errors in viewers’ estimates. We therefore assessed which encoding mechanism was being used for several glyphs, including bars, squares, circles, triangles and human icons. We found that almost half of participants’ estimates were likely based on the encoding mechanism of relative height, not area, for all two-dimensional glyphs. A strong relationship was found between the relative size of the two glyphs being compared and the encoding mechanism used. Participants were much more likely to choose relative height if the smaller glyph was over 60% of the size of the larger glyph. In contrast for the one-dimensional glyph (bar), almost all participants chose relative height as the encoding mechanism. This study therefore shows the importance of considering what encoding mechanisms participants could use, and thus tailoring infographics and visualizations to avoid such ambiguity.