Purchase this article with an account.
Madeline Parker, Allison Silverman, Audrey Wang, Karen Schloss; The role of perceived opacity in interpreting colormap data visualizations. Journal of Vision 2017;17(10):1179. doi: https://doi.org/10.1167/17.10.1179.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Interpreting data visualizations requires determining how perceptual features are assigned to semantic concepts. For example, interpreting colormaps requires inferring how dimensions of color correspond to quantities of a given measure (e.g., brain activity, correlation magnitude). This process should be easier when percept-concept assignments in visual displays (specified by legends/labels) match predicted percept-concept assignments in observers' minds (perceptual-cognitive fit). But, what are the predictions in observers' minds? Evidence suggests observers predict darker colors correspond to larger quantities—response times (RTs) to interpret colormaps are faster when legends specify dark-is-more rather than light-is-more assignments (Silverman et al., VSS-2016). This pattern was largely unaffected by contrast with the background. However, the background may have an effect when the colormap appears to vary in opacity (e.g., value-by-alpha maps; Roth et al., 2010). To test this hypothesis, we presented participants with fictitious data matrices in which columns represented time, rows represented animal species, and cell color represented number of animals sighted during each time window (as in Silverman et al., VSS-2016). Participants reported when more animals were sighted (early/late) and we measured RTs. We used 3 kinds of colorscales to construct the colormaps [black-white/black-blue/white-blue], which appeared on 3 possible backgrounds [black/white/blue], with 2 possible lightness-quantity assignments in the legend [light-is-more/dark-is-more], 2 legend orientations ["greater" labeled higher/lower in the legend] and 20 replications (720 trials). Under conditions in which the colorscale faded into the background (e.g., white-blue scale on a white or blue background), RTs revealed interactions between lightness-quantity assignments in the legend and background lightness (all ps< .01). RTs were generally faster when the legend specified dark-is-more on light backgrounds as shown before, but this difference either diminished or reversed on dark backgrounds (opaque-is-more bias). The results suggest the opaque-is-more bias overrides the dark-is-more bias when there is perceptual evidence for varying degrees of opacity.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only