Meetup 2: Graphical Perception Theory

George Hagstrom

2026-09-07

Why Are Pie-Charts Terrible?

  • “Pie Charts are Evil”
    • Chapter 2 Storytelling with Data

Why Are Pie-Charts Terrible?

  • Lots of graphics advice is presented as rules
  • Are there principles behind these rules?

What to do this week

  1. Readings: Chapter 5 and 6 of “Fundamentals”, Section 1.4 of “Healy”
  2. Chapter 6 of Rougier (10 Simple Rules, helpful if you are struggling with your HW)
  3. Bonus Reading: Cleveland et al 1984 Perceptual Visualization Paper
  4. First HW due Sunday at midnight

What Makes a Graph Good

  • We use graphs to communicate quantitative data
  • A graph is good if it successfully conveys quantitative data to the reader
  • Can we break down visualizations into a series of elementary perceptual tasks and then test how good humans are at completing each one?
  • Research program led by many, William Cleveland most prominent

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Elementary Perceptual Tasks

  1. Position along a common scale
  2. Position along non-aligned scales
  3. Length
  4. Direction
  5. Angle
  6. Area
  7. Volume
  8. Curvature
  9. Shading
  10. Hue/Saturation

Some Perceptual Tasks are Easier

  • Position along a common axis is the best
  • Percentage of major errors:

Some Perceptual Tasks are Easier

  • Position along a common axis is the best
  • Error

Overall Ranking

  1. Position along a common axis
  2. Position along an uncommon axis
  3. Length, Direction, Angle
  4. Area
  5. Shading, Color Saturation
  6. Volume, Curvature

Munzner 2014

Implications

  • Plot the comparisons you want to make on the same axis
  • Position can usually substitute length, direction, and angle
  • Avoid 3D
  • Be cautious with area
  • Shading and color saturation are very imprecise

Dot-Plots for 1D Comparisons

  • Flights from each carrier leaving NYC in 2013

Dot-Plots for Comparisons

  • Flights from each carrier leaving NYC in 2013

Dot-Plots for Comparisons

  • Flights from each carrier leaving NYC in 2013

Show Differences Directly

  • Position is much easier to judge than length Robbins 2004

Show Differences Directly

  • Position is much easier to judge than length
  • These curves have a constant height difference Robbins 2004

Show Differences Directly

  • Adjacent Bars Versus Dots on the same Axis

UCR Bus. Analytics

Show Differences Directly

  • Adjacent Bars Versus Dots on the same Axis

UCR Bus. Analytics

What to do About Color?

  • Color is a poor choice for precise numerical comparisons
  • Consider whether you need to use it at all
  • It is very helpful for illustrating geographic patterns

Perceptually Uniform Colormaps

  • Colormaps attach color to number
  • Visual complexity makes it easy for the colormap to create features that aren’t there or to hide features colorcet

Colormap Example

  • Which colormap renders the surface height better?

colorcet

Colormap Types: Sequential/Perceptually Uniform

Colormap Types: Sequential

Colormap Types: Diverging

Colormap Types: Qualitative

Do you need a map?

Categories/Qualitative Features

  • Hue/Color and Shape are good for differentiating categories
  • Category limit 5 or 6

Small Multiples for Categories

  • Suppose we want to show a relationship between many variables
  • Make repeated plots with same axis that correspond to different categorical or numerical dimensions
  • Facets, Trellis plots, small multiples plots, etc

Small Multiples for Categories

  • Suppose we want to show a relationship between many variables
  • Make repeated plots with same axis that correspond to different categorical or numerical dimensions
  • Facets, Trellis plots, small multiples plots, etc

Small Multiples Example

Thanks!

DATA 608