Chapter 2 Exercises: How the Eye Sees

Part A: Conceptual (Retrieval and Comprehension)

Exercise 2.1 Define pre-attentive processing in your own words. What three characteristics distinguish it from conscious visual analysis?

Exercise 2.2 A colleague argues: "It doesn't matter how we visualize the data as long as all the numbers are there — the audience can figure it out." Using concepts from this chapter, explain why this claim is wrong. Reference at least two specific perceptual mechanisms in your answer.

Exercise 2.3 List the six Gestalt principles covered in this chapter. For each, write a single sentence describing how it applies to chart design. Do this from memory before checking the text.

Exercise 2.4 Explain why Treisman's feature integration theory predicts that a conjunction search (find the red circle among red squares and blue circles) is slower than a feature search (find the red item among blue items). What is the chart design implication?

Exercise 2.5 The Cleveland-McGill hierarchy ranks position as the most accurate encoding and color saturation as the least accurate. Does this mean that color saturation should never be used? Describe two scenarios in which color saturation is the appropriate encoding despite its lower accuracy.

Exercise 2.6 Stevens's Power Law describes how perceived magnitude relates to actual physical magnitude. For the area channel, the exponent is approximately 0.7. In practical terms, what does this mean for a viewer interpreting a bubble chart? If a data point is twice the value of another, what happens to the bubble's perceived size?

Exercise 2.7 Distinguish between Bertin's classification of retinal variables as "selective," "ordered," and "quantitative." Why is this classification useful when choosing an encoding for a specific data type?

Exercise 2.8 Explain the difference between change blindness and inattentional blindness. Give one chart design scenario in which each phenomenon could cause a viewer to miss important information.

Part B: Applied (Encoding Decisions)

Exercise 2.9 You have the following data for Meridian Corp:

| Product Line | Q1 Revenue ($M) | Q2 Revenue ($M) | Q3 Revenue ($M) | Q4 Revenue ($M) | |---|---|---|---|---| | Cloud Services | 42.1 | 44.8 | 47.3 | 51.2 | | Hardware | 38.5 | 36.2 | 35.1 | 33.8 | | Consulting | 22.0 | 23.1 | 22.8 | 24.5 | | Support | 15.3 | 15.8 | 16.2 | 16.5 |

Your goal is to show the quarterly trend for each product line.

(a) Specify which visual encoding channel you would use for each variable (time, revenue, product line).

(b) What chart type does this encoding scheme produce?

(c) Which Gestalt principles support the viewer's ability to trace each product line's trend?

(d) A colleague suggests a stacked area chart instead. What perceptual problem does this introduce for the middle layers? Reference a specific principle from the Cleveland-McGill hierarchy.

Exercise 2.10 You are designing a dashboard showing vaccination rates across 50 countries, categorized into four income groups (low, lower-middle, upper-middle, high). The primary question is: "How do vaccination rates differ across income groups?"

(a) What encoding would you use for vaccination rate? Justify your choice using the hierarchy.

(b) What encoding would you use for income group?

(c) Would a choropleth map be a good choice here? Why or why not, in terms of the encoding channels it uses?

(d) Sketch (in words) the chart design you would recommend.

Exercise 2.11 A social media analytics report needs to compare engagement rates across five platforms (Instagram, TikTok, YouTube, Twitter/X, LinkedIn) for three content types (video, image, text).

(a) You have two spatial position channels (x and y) and several non-spatial channels available. Assign each variable to a channel and justify your assignment.

(b) How many distinct visual categories will the viewer need to track? Is this within the limits of working memory for visual information?

(c) If it is not within working memory limits, what design strategy would you use to reduce the cognitive load?

Exercise 2.12 A researcher presents you with a treemap showing the global distribution of CO2 emissions by country. The treemap has 195 rectangles of varying sizes, colored by continent.

(a) What encoding channel represents the emissions value for each country?

(b) Where does this channel rank in the Cleveland-McGill hierarchy?

(c) The researcher wants viewers to compare the emissions of Germany and Japan. Will the treemap support this comparison accurately? Explain why or why not.

(d) Propose an alternative visualization that uses a higher-accuracy encoding for the same comparison task.

Exercise 2.13 You are designing a chart to show the relationship between a country's GDP per capita (quantitative, continuous) and its life expectancy (quantitative, continuous), with each data point representing one country. You also want to encode the population (quantitative) and the continent (categorical, 6 categories).

(a) Assign each variable to a visual channel, prioritizing by the Cleveland-McGill hierarchy.

(b) Which variable gets the least accurate channel? Is this acceptable? Why?

(c) Identify one Gestalt principle that will help the viewer perceive continental patterns and one that might work against clear perception.

Part C: Real-World Analysis (Evaluate Existing Designs)

Exercise 2.14 Find a pie chart in a recent news article, business report, or dashboard (or use this scenario: a pie chart showing market share of five smartphone brands at 22%, 21%, 20%, 19%, and 18%).

(a) What encoding channel does the pie chart use for the quantitative variable?

(b) Can the viewer accurately rank the five segments? Why or why not?

(c) What alternative chart type would use a more accurate encoding channel? Would it change the message?

(d) Is there any argument for keeping the pie chart in this specific case? (Hint: consider the "part-of-whole" framing.)

Exercise 2.15 Consider a standard stock price line chart (date on x-axis, price on y-axis, with a line connecting daily closing prices). Identify which pre-attentive attributes and Gestalt principles the viewer relies on to extract the following information:

(a) The overall trend (up or down over the full period)

(b) The day with the highest closing price

(c) A period of high volatility

(d) The comparison between two stocks shown as two lines

Exercise 2.16 A health department publishes a choropleth map showing COVID-19 case rates by county. The map uses a sequential color scale from white (low) to dark red (high).

(a) What encoding channel is used for case rate?

(b) What Gestalt principle causes the viewer to perceive clusters of high-rate counties as a "hotspot"?

(c) Large rural counties dominate the visual area of the map but may have small populations. What perceptual bias does this introduce, and which encoding channel is responsible?

(d) The map is published in two versions: one static, one animated (showing the progression over 12 months). What perceptual risk does the animation introduce? Reference a specific phenomenon from Section 2.6.

Exercise 2.17 A corporate presentation uses a 3D bar chart to compare quarterly revenue across four divisions. The bars are rendered with perspective, so bars at the back of the chart appear smaller than bars at the front, even when they represent the same value.

(a) What encoding channel is the 3D chart using that a 2D bar chart does not?

(b) Where does volume/depth rank in the Cleveland-McGill hierarchy?

(c) How does the perspective effect distort the length encoding of the bars?

(d) What would you recommend instead, and which section of this chapter supports your recommendation?

Exercise 2.18 A data journalism piece uses a "bubble race" animation showing how country-level metrics change over time. Each country is a bubble (size = population, color = continent, position = x for GDP, y for life expectancy), and the bubbles move as the year advances.

(a) List all the encoding channels in use and the variable each encodes.

(b) How many channels is the viewer asked to decode simultaneously?

(c) Which channel is pre-attentive, and which requires conscious decoding?

(d) What cognitive limitations (from Section 2.6) are most likely to impair comprehension of this visualization?

Part D: Synthesis (Cross-Concept Integration)

Exercise 2.19 Design a "perception audit" — a systematic checklist of 5-7 questions that a chart designer could use to evaluate whether a chart is perceptually sound, based on the principles in this chapter. Each question should reference a specific concept (pre-attentive processing, Cleveland-McGill hierarchy, Gestalt principles, or perceptual limitations). Explain why you included each question.

Exercise 2.20 Consider two versions of the same chart showing monthly sales for Meridian Corp's Cloud Services division over two years (24 data points):

• Version A: A bar chart with 24 bars (one per month), colored alternately by year (blue for Year 1, orange for Year 2).
• Version B: A line chart with two lines (one per year), the x-axis showing months January-December, and the lines distinguished by color.

Analyze both versions using pre-attentive processing, the Cleveland-McGill hierarchy, and Gestalt principles. Which version better supports the comparison task "How did monthly sales patterns differ between the two years?" and why?

Exercise 2.21 A visualization researcher proposes a new encoding channel: "blur level." Data points with higher values are rendered sharply, while lower values are rendered with increasing Gaussian blur. The researcher claims this is pre-attentive (sharp items pop out from blurred ones).

(a) Is the researcher's claim about pre-attentive processing plausible? Why or why not?

(b) Using Bertin's framework, classify "blur level" as selective, ordered, quantitative, or some combination.

(c) Where would you estimate "blur level" falls in the Cleveland-McGill hierarchy? Justify your placement.

(d) What practical problems might arise in using blur as an encoding?

Exercise 2.22 Write a 300-word "principles memo" to a design team that has been producing charts with the following recurring issues: (1) extensive use of pie charts for comparing values, (2) rainbow color scales for continuous data, (3) charts with 10+ color-coded categories, and (4) 3D effects on bar charts. Reference specific principles from this chapter to explain why each practice is problematic, and provide a concrete alternative for each.

Part M: Mixed Review (Integrating Chapter 1)

Exercise 2.23 In Chapter 1, we discussed Anscombe's Quartet — four datasets with identical summary statistics but different visual patterns. Using the language of Chapter 2, explain why the visual patterns are immediately obvious in scatter plots but invisible in the summary statistics. Which pre-attentive attributes and encoding channels make the differences visible?

Exercise 2.24 Chapter 1 argued that visualization is a "cognitive tool" that offloads pattern recognition to the visual system. Chapter 2 provides the mechanism. Write a three-sentence explanation of visualization's cognitive benefit that specifically references pre-attentive processing and the Cleveland-McGill hierarchy.

Exercise 2.25 Chapter 1 introduced the idea that bad visualizations can mislead. Chapter 2 introduces the perceptual mechanisms that explain how misleading occurs. Pick one perceptual limitation from Section 2.6 (optical illusions, change blindness, inattentional blindness, or working memory limits) and describe a realistic scenario in which a chart designer — intentionally or accidentally — exploits that limitation to mislead the viewer.

Part E: Research and Exploration

Exercise 2.26 Read the original Cleveland and McGill (1984) paper, "Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods" (available in the Journal of the American Statistical Association, Vol. 79, No. 387). Alternatively, read the 2010 replication by Heer and Bostock, "Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design."

(a) How did Cleveland and McGill operationalize "accuracy" — what specific judgment did they ask participants to make?

(b) Did the Heer and Bostock replication produce the same ranking? Were there any notable differences?

(c) What are the limitations of both studies? (Hint: consider the types of judgments tested, the stimulus design, and the ecological validity of the tasks.)

Exercise 2.27 Jacques Bertin published Semiology of Graphics in 1967, but the English translation did not appear until 1983. Investigate the history: Why was the English translation delayed? How did Bertin's work influence (or fail to influence) the English-speaking visualization community before and after the translation? Write 200-300 words summarizing what you find.

Quick-Reference: Exercise Categories and Skills Tested

Answer Key Notes

Selected answers are provided in Appendix A. For Part B and Part D exercises, multiple valid answers exist; the appendix provides model responses with reasoning, not single correct answers. Part E exercises are open-ended research tasks; no answer key is provided.

Self-Assessment Guide: If you can answer all Part A questions from memory, you have solid recall of the chapter's key concepts. If you can complete Part B exercises with justified channel assignments and hierarchy references, you can apply the principles. If Part C analyses identify specific perceptual problems and propose improvements, you are reading charts critically. Parts D and E indicate deeper integration and scholarly engagement.