Quiz — Chapter 3: Color

Answer all 20 questions. For multiple-choice questions, select the single best answer. For true/false questions, state whether the statement is true or false and provide a one-to-two sentence justification. For short answer and applied scenario questions, respond in the space indicated.

Multiple Choice (10 questions)

Question 1

Which color dimension is best suited for encoding nominal (categorical) data?

A) Luminance B) Saturation C) Hue D) All three dimensions are equally appropriate for nominal data

Question 2

The primary reason the rainbow (jet) colormap is considered harmful for continuous data is:

A) It uses too many colors, making the chart look busy B) It is perceptually non-uniform, creating false boundaries and masking real gradients C) It was designed before modern monitors could display it properly D) It uses colors that are culturally offensive in some regions

Question 3

Approximately what percentage of men have some form of color vision deficiency?

A) 1% B) 4% C) 8% D) 15%

Question 4

Which of the following color pairs is the MOST dangerous for visualizations viewed by a colorblind audience?

A) Blue and orange B) Red and green C) Blue and black D) Purple and yellow

Question 5

A perceptually uniform colormap is one in which:

A) All colors in the palette have the same hue B) Equal steps in data value produce equal perceived color differences C) The palette uses only primary colors (red, blue, yellow) D) Every color in the palette has the same saturation

Question 6

You are creating a heatmap showing year-over-year change in revenue for 50 stores. Some stores grew, some shrank. The most appropriate palette type is:

A) Sequential — because revenue is ordered from low to high B) Diverging — because the data deviates from a meaningful center (zero change) C) Categorical — because each store is a distinct entity D) Rainbow — because it provides maximum color variation

Question 7

The luminance-first principle states that:

A) Luminance should always be set to maximum in any chart B) Dark backgrounds are always preferable to light backgrounds C) A chart's essential information should be readable from the luminance channel alone, even without hue D) Low-luminance colors should always represent low data values

Question 8

Which tool was created by Cynthia Brewer and provides tested sequential, diverging, and qualitative palettes with colorblind-safety annotations?

A) Coolors B) ColorBrewer C) coblis D) Sim Daltonism

Question 9

A categorical (qualitative) palette should have which of the following properties?

A) A clear luminance gradient from light to dark B) A neutral midpoint between two hue anchors C) Distinct hues at similar luminance and saturation levels D) A single hue varying only in saturation

Question 10

You are designing a visualization for an international audience that includes viewers from both Western and East Asian financial contexts. Which approach to color encoding for profit/loss data is the safest?

A) Use red for loss and green for profit, matching Western convention B) Use red for profit and green for loss, matching East Asian convention C) Use blue for loss and orange for profit with text labels, avoiding culturally loaded red-green associations D) Use black for all values and rely entirely on text labels

True/False with Justification (4 questions)

Question 11

Statement: The HSL (Hue, Saturation, Lightness) color space is perceptually uniform, meaning that equal numerical steps in HSL produce equal perceived color changes.

True or False? Justify your answer in one to two sentences.

Question 12

Statement: A sequential palette that goes from light yellow to dark blue is appropriate for encoding a variable that ranges from "strongly disagree" to "strongly agree" on a 5-point Likert scale with a neutral midpoint.

True or False? Justify your answer in one to two sentences.

Question 13

Statement: Redundant encoding — using both color and shape to represent the same variable — is wasteful because it uses two channels to communicate what one could handle.

True or False? Justify your answer in one to two sentences.

Question 14

Statement: If a chart passes a deuteranopia simulation test, it is guaranteed to be accessible to all viewers with color vision deficiency.

True or False? Justify your answer in one to two sentences.

Short Answer (3 questions)

Question 15

Explain the difference between a sequential palette and a diverging palette. In your explanation, give one specific dataset example where each would be the correct choice, and state what would go wrong if you used the wrong type for each example.

(Answer in 4-6 sentences.)

Question 16

Describe the "grayscale test" for evaluating a color palette. What specific property of a palette does this test evaluate, and what does a failure look like? Name one widely used colormap that fails this test and one that passes it.

(Answer in 3-5 sentences.)

Question 17

A colleague argues: "I use the rainbow colormap because it shows the most detail — you can see more distinct color bands than in viridis." Explain why this argument is wrong, referencing the concept of perceptual uniformity and false boundaries.

(Answer in 3-5 sentences.)

Applied Scenario (2 questions)

Question 18

Scenario: You are creating a dashboard for a hospital that shows patient wait times by department. The dashboard includes: - A bar chart comparing average wait times across 6 departments - A heatmap showing wait times by department and hour of day (24 hours x 6 departments) - A trend line showing average wait time over the past 30 days, with a target line at 20 minutes

For each chart, specify: a) The palette type you would use (sequential, diverging, or categorical) b) A specific named palette or described color scheme c) One accessibility consideration specific to that chart

(Answer in a structured format, addressing each chart separately.)

Question 19

Scenario: A public health agency publishes a choropleth map showing COVID-19 vaccination rates by county. The map uses a three-color scheme: green (above 70%), yellow (50-70%), and red (below 50%). A journalist points out that the map "makes it look like most of the country is doing fine" because the green category spans a wide range (70% to 99%) while the red category is narrow.

a) Explain the perceptual problem with this color scheme in terms of color steps and data resolution. b) Explain the ethical problem — how does this color choice affect public understanding? c) Propose a redesigned color scheme. Specify the palette type, the number of color steps, the breakpoints, and your choice of colors. Explain why your design is superior.

(Answer in 6-10 sentences.)

Analysis (1 question)

Question 20

Analysis Task: Consider the following three colormaps applied to the same continuous dataset (temperature readings from 0 to 50 degrees Celsius):

Colormap A: Rainbow (jet) — cycles through blue, cyan, green, yellow, red Colormap B: viridis — dark purple through blue, green, to bright yellow Colormap C: A two-color gradient from white to dark red (single-hue sequential)

For each colormap, analyze the following properties in a structured comparison:

Fill in the table with brief assessments (1-2 words or a short phrase per cell), then write a 3-4 sentence paragraph recommending which colormap is best for this dataset and why. If your answer depends on context (audience, medium, purpose), state the conditions under which each would be most and least appropriate.

Answer Key

Multiple Choice

1. C — Hue is an identity channel that communicates "this is different from that" without implying order, making it appropriate for nominal categories.

2. B — The rainbow colormap is perceptually non-uniform: equal data steps do not produce equal perceived color changes, which creates artificial visual boundaries in smooth data and can mask real gradients.

3. C — Approximately 8% of men have some form of color vision deficiency, predominantly red-green confusion (deuteranopia and protanopia).

4. B — Red and green are indistinguishable for viewers with deuteranopia (~5% of males) and protanopia (~2.5% of males), the two most common forms of CVD.

5. B — Perceptual uniformity means that a 10-unit step in data value at any point in the range produces the same perceived color difference as a 10-unit step at any other point.

6. B — Year-over-year change has a meaningful center (zero change), with values extending in two directions (growth and decline), making a diverging palette appropriate.

7. C — The luminance-first principle ensures that the chart's information survives colorblindness, grayscale printing, and other conditions where hue information may be lost.

8. B — ColorBrewer (colorbrewer2.org) was created by Cynthia Brewer and is the standard resource for perceptually tested visualization palettes.

9. C — Categorical palettes use distinct hues at similar luminance and saturation so that no category visually dominates and no ordering is implied.

10. C — Avoiding the red-green pair sidesteps both cultural ambiguity (red means different things in different financial cultures) and colorblind accessibility issues. Text labels provide unambiguous clarification.

True/False

11. False. HSL is more intuitive than RGB but is not perceptually uniform. Equal numerical steps in hue, saturation, or lightness in HSL do not correspond to equal perceived differences. The CIE LAB color space was designed specifically to address this problem.

12. False. A Likert scale with a neutral midpoint is not ordered in one direction — it has a meaningful center and extends in two directions (disagree vs. agree). A diverging palette with the neutral midpoint in the center would be more appropriate than a sequential palette.

13. False. Redundant encoding is a best practice, not a waste. It improves accessibility for colorblind viewers, provides resilience against printing in grayscale, and speeds comprehension for all viewers by confirming information through multiple visual channels.

14. False. Deuteranopia is the most common form of CVD, but other types exist (protanopia, tritanopia, anomalous trichromacy). A thorough accessibility check should test against at least deuteranopia and protanopia, and ideally also tritanopia and achromatopsia.

Short Answer

15. A sequential palette represents data ordered in a single direction (low to high), using a luminance gradient within one or a narrow range of hues. A diverging palette represents data that extends in two directions from a meaningful center, using two distinct hue anchors with a neutral midpoint. For a dataset of annual rainfall amounts (0mm to 300mm), a sequential palette is correct because the data has no meaningful center — using a diverging palette would falsely imply that some rainfall amount is a "baseline" from which values deviate. For a dataset of election margins (from -30 points to +30 points, where 0 is a tie), a diverging palette is correct because the center is meaningful — using a sequential palette would obscure the critical distinction between winning and losing.

16. The grayscale test involves converting a visualization to grayscale (removing all hue and saturation information, leaving only luminance) and checking whether the data ordering or category distinctions remain readable. It evaluates whether a palette's luminance gradient is monotonic and sufficiently varied to carry the data's information independently of hue. A failure looks like a chart where previously distinct colors collapse to similar gray tones, making the data unreadable or creating a non-monotonic pattern. The rainbow (jet) colormap fails this test — its luminance profile peaks at yellow and dips at blue, producing a confusing grayscale image. The viridis colormap passes — its luminance increases monotonically from dark to light.

17. The "extra bands" visible in a rainbow colormap are not additional detail — they are perceptual artifacts. Because the rainbow is perceptually non-uniform, the human eye perceives sharp transitions where none exist in the data (for example, at the yellow-to-green boundary) and perceives smooth regions where there may actually be variation (for example, across the blue-to-cyan range). viridis appears to show fewer "bands" precisely because it is perceptually uniform: the visual change is proportional to the data change at every point in the range. What the rainbow presents as "detail" is actually noise injected by the colormap into the viewer's perception.

Applied Scenario

18. Bar chart (6 departments): a) Categorical — each department is a nominal category. b) Set2 (ColorBrewer) or tab10 — 6 distinct, muted hues at similar luminance. c) Ensure all 6 colors are distinguishable under deuteranopia simulation; consider adding department name labels directly on or adjacent to bars to provide redundant encoding.

Heatmap (24 hours x 6 departments): a) Sequential — wait time is an ordered variable going from low to high. b) viridis or YlOrRd (ColorBrewer) — monotonic luminance gradient, colorblind-safe. c) Ensure sufficient luminance contrast between adjacent steps so that fine-grained time-of-day patterns are visible, not just the extremes. Test at the display size it will actually be used.

Trend line (30 days with target): a) This chart does not need a multi-color palette. Use a single color for the trend line and a contrasting color (or dashed line) for the target. b) Dark blue for the trend line, medium gray dashed for the target line. If adding a fill to indicate "above/below target," use a light diverging fill (light red for above target, light blue for below, or vice versa depending on context). c) The target line at 20 minutes is the critical reference — it must have sufficient contrast against both the trend line and the background. Do not rely on color alone; add a text label "Target: 20 min."

19. a) The three-color scheme compresses a wide range of data into only three visual categories. A county at 71% and a county at 98% receive the same green color, hiding a 27-percentage-point difference. Meanwhile, a county at 49% (red) and a county at 51% (yellow) are shown as dramatically different despite a 2-point gap. The perceptual difference between the color steps does not correspond to the actual data differences.

b) This creates an ethical problem because the map communicates "three levels of reality" when the underlying data is continuous. By making green the largest bucket, the map visually minimizes the number of areas that appear to need attention. Viewers form impressions from the dominant color on the map, and a green-dominant map suggests that the situation is mostly under control — which may not reflect the underlying variation within that 70-99% range.

c) A redesigned approach: use a sequential palette (e.g., YlOrRd or viridis) with 7 to 9 discrete steps at evenly spaced breakpoints (e.g., <40%, 40-50%, 50-60%, 60-70%, 70-80%, 80-90%, 90%+). This reveals gradation within both the low and high ranges. Use a perceptually uniform palette so that each step looks equally different from its neighbors. Add a prominent marker or boundary line at the 70% threshold to preserve the policy-relevant breakpoint without collapsing all data above it into a single color. Ensure colorblind safety by choosing a palette that progresses in luminance, and test in grayscale for the printed version.

Analysis

20.

For this temperature dataset, viridis (Colormap B) is the best overall choice. It provides perceptual uniformity, ensuring that a 5-degree change looks the same at any point in the range; it is colorblind-safe and grayscale-readable; and its moderate hue variation (dark purple to bright yellow) provides enough color contrast to distinguish fine gradations while avoiding the false boundaries of the rainbow. Colormap C (white-to-red) would be a reasonable alternative if semantic association (red = hot) is important for the audience, but it provides less perceptual discrimination in the mid-range than viridis. Colormap A (rainbow) should not be used — the apparent "detail" it shows is perceptual distortion, not data.