9: Perception & Design

Content for Monday, April 27, 2026

Before class

📖 Reading:

• Kang, M. (2016). Checklist: Does Your Data Visualization Say What You Think It Says?
• Rowe, S. (2022, May 24) Don’t hide the crucial findings.
• Cisneros, M. (2022, June 14). Showing off the insights vs. showing off.
• Esselman, A. (2022, June 23) Declutter a dual y-axis chart.
• Rowe, S. (2022, November 14) A diverging bar chart makeover: how to tackle the challenge of scope creep.
• Optional: Knaflic, Storytelling with Data, Ch 1–3

ImportantAssignment 3 is due today

Assignment 3: Tidying, Import & Quarto — due Sunday, April 26 at 11:59 PM.

During class

We’ll cover:

• Why perception matters — how the brain processes visual information
• Preattentive attributes: color, size, position, shape
• Gestalt principles and how they shape layout
• Color theory: sequential, diverging, and qualitative palettes
• Colorblind-friendly design (and why it matters)
• Decluttering: removing the visual noise
• Psychology-specific figures: error bars, interaction plots, raincloud plots
• Putting it all together with theme() customization

TipAssignment 4 is assigned today

Assignment 4: Visualization Deep Dive — due Sunday, May 3 at 11:59 PM.

Slides

View slides in new tab Download PDF

Embedded slides

After class

✅ Practice:

1. Take a plot you made earlier in the course. Identify 3 things that could be decluttered.
2. Recreate it using only theme_minimal() — what disappears? What stays?
3. Try switching your color palette to scale_fill_viridis_d(). Does it change the story?
4. Create a raincloud plot (violin + boxplot + jitter) for any grouped data you have
5. Make a “bad” version of a figure on purpose — violate as many design principles as you can. Then fix it.

NoteThe decluttering checklist

Before you call a figure “done,” ask:

• Is the background gray? → Switch to theme_minimal() or theme_classic()
• Are there gridlines? → Remove them
• Is there a legend when there’s only one group? → Remove it
• Do the axis labels say something meaningful? → If not, rename them
• Can someone understand the point without reading the caption? → If not, the title needs work

Colorblind-friendly palettes

# Viridis — works for continuous and discrete
scale_fill_viridis_d()   # discrete categories
scale_fill_viridis_c()   # continuous values

# ColorBrewer — curated palettes
scale_fill_brewer(palette = "Set2")      # qualitative
scale_fill_brewer(palette = "RdBu")      # diverging
scale_fill_brewer(palette = "Blues")     # sequential

These are all built into ggplot2 — no extra packages needed.