Use bar charts to support main-effect inferences from familiar multivariate data

Choose a grouped bar chart instead of a line chart when familiar multivariate content should lead readers to an overall effect that ignores one of the two independent variables. For example, use grouped bars rather than connected lines when the key message is the average effect of one variable across the levels of the other.

Grouped bars help readers keep the relevant values together while mentally averaging or scanning across one variable, which makes main-effect summaries more available than they are in a line chart.

Mechanism: Proximity and similarity support the mental computation needed to collapse across one variable, so readers can more easily derive a main effect when they already expect it and know how to compute it from the graph.

Evidence: Viewers made more main-effect inferences for familiar data than unfamiliar data, and high-skilled viewers were most likely to produce those inferences when the familiar data were shown as bar graphs rather than line graphs. (Shah & Freedman, 2011)

Notes: Bar charts alone did not make low-skilled viewers infer main effects.

• User Goal: The reader should state an overall effect across one variable rather than only describe interactions.
• Task: Open-ended explanation of the main point in a three-variable graph.
• Data: The variables are familiar enough that readers have expectations about likely overall patterns.
• Chart Setting: One variable is on the x-axis, one is in the legend, and the same data could be drawn as either grouped bars or lines.
• Audience: Readers have relatively high graph-comprehension skill.
• Success Criterion: Readers explicitly mention a main effect without being prompted to compute it.

Break it when: The audience has low graph-comprehension skill or the content is unfamiliar to them. Why: In the study, bar charts did not reliably produce main-effect inferences under those conditions.

Sacrifice: You reduce the strong interaction emphasis that a line chart provides.
Risk: You may assume the grouped bar chart alone will make all readers compute the intended main effect.
Mitigation: If the audience is low-skilled, explicitly represent the important inference on the display instead of relying on mental computation alone.

Mistake: Using a grouped bar chart alone and assuming low-skilled readers will derive the intended main effect. Why it fails: Low-skilled viewers in the study did not make those inferences even when the data were shown as bar graphs.

Failure Sign: Readers describe only interaction patterns and omit the intended overall effect.
Quick Check: Show matched bar and line versions of familiar content and ask, “What is the main point?” Choose the grouped bar chart only if high-skilled reviewers mention the main effect more often from the bar version.
Stronger Test: Test with the target audience; if low-skilled readers still omit the main effect, the chart is not carrying the inference on its own.

• Replot the same three-variable data as grouped bars instead of connected lines.
• Use familiar content examples when you expect readers to bring prior expectations to the graph.
• If low-skilled readers still miss the overall effect, explicitly represent the main-effect takeaway on the display instead of relying on the grouped bar chart alone.