Graphical Perception Lab

Complete all seven estimates before submitting. The blue value is printed on each display; estimate the value represented by orange.

Blue: known reference Orange: value to estimate

Your results

The treatments are ordered by your log error. Lower values indicate more accurate proportional judgments.

Your observed ranking

Rank	Treatment	Estimate	Actual	Absolute error	PP error	Log error

Published benchmark

T1 · Single bar chart
T3 · Separate bar groups
T6 · Pie angle
T8 · Centered rectangles
T9 · Treemap area
T7 · Bubble area

This is the crowdsourced mean-error order for the six selected paper treatments in Heer and Bostock (2010). Luminance was not part of Figure 4.

How the scores work. First, both the estimate and true orange value are expressed as a percentage of the known blue value. Proportional error is the absolute difference between those two percentages.

log error = log₂(proportional error + 1/8)

Because every trial uses a different pair of values, this normalized error is more comparable than raw absolute error. A single response per treatment is still a classroom demonstration, not a stable estimate of perceptual accuracy.