Use point marks instead of bars for nominal values
For categorical lookup on charts with nominal values, prefer point-mark charts over bar charts to prevent false ordering cues and address misleading length comparisons for readers interpreting unordered categories.
- purpose:select
- basis:empirical
- data:categorical
- chart:dotplot:use
- chart:bar:avoid
- quality:fidelity:use
- lever:chart-family
advice
Replace bars with points for nominal categories
Use point marks rather than bar length when the encoded values are nominal and unordered. For example, replace a bar chart with a plot chart that places a point at each category position, so the display does not suggest that one category is longer, larger, or better than another.
reason
Why bars misstate nominal data
Bars carry length, and length invites readers to interpret order or magnitude. That implication is harmless for ordered or quantitative values, but it is wrong for a nominal set.
Mechanism: Point marks show membership in a category without adding a length comparison that the data do not support.
Evidence: The paper gives an example where a bar chart incorrectly implies ordering for a nationality relation and contrasts it with a plot chart as the correct way to encode a nominal domain set (Mackinlay, 1986).
context
Use when the value set has no inherent order
- User Goal: Show which category each item belongs to without implying rank.
- Task: Read nominal categories correctly.
- Data: The encoded domain is nominal rather than ordered or quantitative.
- Chart Setting: A bar chart is a candidate only because it is familiar, not because the values have magnitude.
- Success Criterion: Readers do not infer a false ordering from the display.
exceptions
Use bars only when the values are ordered or quantitative
Break it when: The encoded values form an ordered or quantitative set. Why: Then length can legitimately encode the ordering or magnitude.
costs
Tradeoffs of removing bars
Sacrifice: You give up a length cue that can support ordered comparisons. Risk: A point-based display no longer provides magnitude judgments, which is correct here but may disappoint readers expecting bar lengths. Mitigation: Reserve bars for data that truly have order or quantity.
mistakes
Common nominal-data failure
Mistake: Draw bars for nominal categories and rely on the reader to ignore the lengths. Why it fails: The lengths still suggest an ordering relationship that is not in the data.
check
Check whether the chart implies an order that does not exist
Failure Sign: The viewer can describe one nominal category as longer, larger, or better than another from the marks alone. Quick Check: Compare a bar version and a point-mark version; if only the bar version creates a visible ranking, keep the point-mark version. Stronger Test: Ask whether any mark length in the chart encodes a magnitude that the data do not actually contain.
fix
Remove the misleading length cue
- Replace the bars with points positioned at the category values.
- Keep the category labels on the axes and let the point positions carry the relation.
- Reserve bar length for ordered or quantitative values only.