Fig 1.
We used eight experimental conditions a): 4 color scales (rainbow, heated-body, univariate color intensity, bi-variate red-white-blue) × 2 image background conditions (black and white). Experimental conditions remained unchanged for a single data context. b) Participants were presented with a neuroscience (i.e., brain activity states) and a geographic (ecosystem states) data context: First, participants had to perform a trust rating task for two extreme states (“healthy brain|ecosystem” and “dead brain|ecosystem”); then they had to perform an interpretation task by ranking three randomly presented intermediate states between the two extreme states (one condition for each data context).
Fig 2.
Between-group comparisons of response variability.
Contrary to our hypothesis, overall response variability (higher values indicate greater variability) was highest for the neuroimaging experts in the neuroscience data context (a). This is mainly due to their greater trust variability (b). A tendency for experts’ interpretation variability is only discernible in the geographic data context (c). Non-experts show lower variability in the neuroscience data context compared to the geographic context.
Table 1.
Results across color scales.
Fig 3.
Comparing suitability with trust ratings.
General suitability ratings for color scales to be used in a domain (i.e., neuroimaging or geovisualization) do not align with trust ratings collected during the experiment. This is particularly true for the rainbow scale. This effect is strongest for the geovis experts, trained not to use the rainbow scale in the illustration of univariate data in a progression [13, 14].