Quiz: Layout, Composition, and Small Multiples

20 questions. Aim for mastery (18+). If you score below 14, revisit the relevant sections before moving to Chapter 9.

Multiple Choice (10 questions)

1. The chapter defines "composition" in the context of data visualization as:

(a) The process of computing statistical summaries before charting (b) The arrangement of multiple visual elements on a page or screen as a coherent whole (c) The act of combining multiple datasets into one (d) The choice of which data points to include in a chart

Answer **(b)** The arrangement of multiple visual elements on a page or screen as a coherent whole. Composition is a design discipline distinct from single-chart design. It addresses how charts, text, titles, annotations, and whitespace work together on the page to produce a unified visual experience. A collection of well-designed charts laid out badly is still a poor figure — composition does work that no individual chart can do on its own.

2. Which of the following is NOT one of the "four jobs of composition" identified in Section 8.1?

(a) Enabling comparison between related charts (b) Establishing visual hierarchy between panels (c) Maximizing the number of charts per page (d) Guiding reading order

Answer **(c)** Maximizing the number of charts per page. The four jobs are: enabling comparison, establishing hierarchy, guiding reading order, and creating visual unity. Maximizing chart density is explicitly not a goal — over-crowded figures fight against all four of the genuine jobs by reducing the whitespace needed for visual hierarchy and alignment.

3. Edward Tufte famously called small multiples:

(a) "The hardest design principle to get right" (b) "The best design principle" for information visualization (c) "An overrated technique from the 1980s" (d) "A useful tool for certain narrow use cases"

Answer **(b)** "The best design principle" for information visualization. Tufte's claim appears in *Envisioning Information* and has been cited widely since. The claim is strong and not universally accepted — "best" is hard to defend for any single principle — but small multiples have resolved more visualization problems than any other modern design technique, and Tufte's formulation remains the canonical statement of the case for them.

4. A small multiple is defined as:

(a) A single chart with many overlapping series (b) A set of charts with different chart types arranged in a grid (c) A set of charts that share the same visual encoding, differing only in which slice of the data they show (d) A small chart used as an inline annotation within text

Answer **(c)** A set of charts that share the same visual encoding, differing only in which slice of the data they show. The shared encoding is the defining feature. Every panel uses the same chart type, the same axes (usually), the same scales (usually), and the same design. Only the data subset varies. This consistency enables the reader to compare across panels pre-attentively, which is the whole point of the technique. Without shared encoding, you have a dashboard or a collage, not a small multiple.

5. Which of the following is the best aspect ratio for a line chart showing 150 years of global temperature data?

(a) Square (1:1) (b) Tall (1:3) (c) Wide (3:1 or wider) (d) Very wide (10:1)

Answer **(c)** Wide (3:1 or wider). Time series benefit from wide aspect ratios for two reasons: (1) the time dimension gets enough horizontal extent to show the full pattern without feeling cramped, and (2) Cleveland's "banking to 45 degrees" rule — the average slope should be around 45 degrees for human perception to best distinguish slope differences. Very wide (10:1) would stretch the slopes too shallow. Square or tall would make the slopes too steep. A ratio in the 3:1 to 4:1 range is usually about right for long time series.

6. Cleveland's "banking to 45 degrees" rule says:

(a) Line charts should always be viewed at a 45-degree angle (b) The average slope of line segments in a time-series chart should be around 45 degrees for optimal slope perception (c) All axis labels should be rotated 45 degrees (d) The chart aspect ratio should be 45:100

Answer **(b)** The average slope of line segments in a time-series chart should be around 45 degrees for optimal slope perception. The rule is based on empirical perceptual research: human perception is most accurate at distinguishing slope differences near 45 degrees. Too-shallow slopes blur together as "flat"; too-steep slopes blur together as "steep." The 45-degree region is where small slope differences remain perceptible. The practical implication is a design heuristic for choosing time-series aspect ratios: stretch or compress the chart until the average slope is in the 45-degree neighborhood.

7. The chapter recommends placing the "hero" panel (the most important chart) in which position in a multi-panel figure?

(a) Bottom-right, where the reader's eye ends (b) Center, to draw focus (c) Top-left, where the reader's eye lands first in the Z-pattern (d) Bottom-left, to leave the top for context panels

Answer **(c)** Top-left, where the reader's eye lands first in the Z-pattern. Western readers scan multi-panel figures in a Z-pattern: top-left → top-right → bottom-left → bottom-right. The top-left position gets the reader's freshest attention. Placing the hero there aligns the composition with the natural scanning pattern. Placing the hero elsewhere means the reader's best attention is spent on a less important panel, and the hero gets whatever attention is left.

8. The key distinction between a small multiple and a dashboard is:

(a) Small multiples are smaller than dashboards (b) Small multiples use the same chart type across panels; dashboards use different chart types for different questions (c) Small multiples are static; dashboards are interactive (d) Small multiples are for news; dashboards are for business

Answer **(b)** Small multiples use the same chart type across panels; dashboards use different chart types for different questions. Small multiples show the same kind of chart applied to different data slices, enabling comparison across slices. Dashboards show different kinds of charts covering different aspects of a system, enabling monitoring of multiple related questions. The two have different design rules — small multiples depend on consistency, dashboards depend on creating unity from diverse elements.

9. Which Gestalt principle is most directly responsible for making the reader perceive aligned panel edges as "intentional" and "organized"?

(a) Proximity (b) Similarity (c) Alignment (continuity) (d) Enclosure

Answer **(c)** Alignment (continuity). The continuity principle says that the eye follows implicit lines created by aligned elements. When panel edges line up at the same coordinates, the alignment creates invisible lines that the reader's eye follows, producing a sense of order and structure. Misaligned edges create broken lines that feel chaotic. Alignment at the pixel level matters — a two-pixel mismatch between panel edges is visible to the reader as "something is slightly off."

10. For a small multiple comparing the same metric across 20 countries, the chapter recommends:

(a) Using a free (per-panel) y-axis so each country can see its own data clearly (b) Using a shared y-axis so magnitudes are directly comparable (c) Using a shared y-axis when absolute comparison is the point, and a free y-axis when the shape of the pattern is the point (d) Mixing shared and free axes across panels to give variety

Answer **(c)** Using a shared y-axis when absolute comparison is the point, and a free y-axis when the shape of the pattern is the point. The choice depends on what comparison you want the reader to make. Shared axes enable direct magnitude comparison: country A's rate is higher than country B's rate, visibly. Free axes preserve the shape of each pattern regardless of absolute level, which matters when the groups have wildly different ranges. Neither is always right. The honest choice is to pick based on the comparison that matters and label the chart clearly so the reader knows which kind of comparison is possible.

True / False (5 questions)

11. "A figure with five perfectly designed individual charts will always be a good figure, regardless of how the charts are arranged on the page."

Answer **False.** Composition is a distinct skill. A collection of well-designed charts laid out badly is still a poor figure, because the arrangement fails to do the work composition is supposed to do: enabling comparison, establishing hierarchy, guiding reading order, creating visual unity. The reader experiences the figure as a whole, not as a sum of individual charts, and the experience depends on the arrangement as much as on the charts themselves.

12. "Small multiples work because every panel uses a different chart type, giving the reader variety."

Answer **False.** Small multiples work for the opposite reason: every panel uses the **same** chart type. The consistency is the point. Shared visual encoding enables the reader to compare across panels pre-attentively, because they do not have to re-learn how to read each panel. A layout with different chart types in each panel is a dashboard (if intentional) or a collage (if not), but it is not a small multiple.

13. "The reading order of a multi-panel figure should match the logical order of the argument the figure is making."

Answer **True.** If the argument goes "context → main finding → implications," the panels should be arranged so that the context comes first (top-left in a Z-pattern), the main finding is prominent (top, near the start of the reading path), and the implications are at the bottom (where the reader ends the scan). A figure whose panels are in a different order from the argument's logic fights the reader's natural scanning pattern and slows comprehension.

14. "Scatter plots should usually have a wide aspect ratio, like time series, because both axes show continuous variables."

Answer **False.** Scatter plots should usually be closer to square. The x and y axes of a scatter plot typically represent two equally-important continuous variables, and a square plotting area gives both axes the same visual weight. A wide or tall scatter plot implicitly tells the reader that one axis is more important. The exception is when one variable has a much wider meaningful range than the other, in which case some elongation is justified — but the default for scatter plots is roughly square.

15. "Dashboards and small multiples have the same design rules because they both involve multiple charts."

Answer **False.** They have different design rules because they solve different problems. Small multiples depend on **consistency** (same chart type, same scales, same design across panels) because the point is comparison. Dashboards depend on **unity from diversity** (different chart types for different questions, unified by consistent typography, accent color, whitespace, and alignment) because the point is monitoring multiple aspects of a system. Applying small-multiple rules to a dashboard would produce too much consistency; applying dashboard rules to a small multiple would produce too much diversity. The rules are not interchangeable.

Short Answer (3 questions)

16. In three to four sentences, explain why shared y-axes enable comparison in small multiples and why free y-axes sometimes sacrifice comparability for a different kind of insight.

Answer A **shared y-axis** means every panel uses the same scale, so the reader can directly compare the magnitudes of values across panels. A value of 50 in one panel is the same as a value of 50 in another — the visual positions are comparable. A **free y-axis** gives each panel its own scale, which can reveal the shape of the pattern in each panel (especially when groups have wildly different ranges), but sacrifices the ability to compare absolute magnitudes. The trade-off depends on what comparison matters for the specific chart: if magnitudes are the point, share the axis; if shapes are the point, free the axis and label clearly so the reader knows they cannot compare magnitudes.

17. Explain the Z-pattern for reader scanning of multi-panel figures, and state what design consequence follows from it.

Answer The Z-pattern is the sequence in which Western readers scan a multi-panel figure: the eye lands at the top-left, moves across to the top-right, drops diagonally to the bottom-left, and ends at the bottom-right. The design consequence is that the most important panel (the hero) should be placed in the top-left, where the reader's freshest attention lands. A hero placed elsewhere — bottom-right, center, or anywhere off the Z-path — receives leftover attention, and the reader's initial impression is shaped by whichever panel happens to be in the top-left instead.

18. Describe what is meant by "enforced comparison" in the context of small multiples, and explain why it depends on visual consistency across panels.

Answer **Enforced comparison** means that the layout of a small multiple makes comparison between panels automatic rather than optional. Because every panel uses the same visual encoding — same chart type, same scales, same design — the reader does not have to mentally compute "how do I read this panel vs. that panel?" They already know. The shared grammar means their attention can focus entirely on the differences in the data, which is the point of the comparison. If the panels used different encodings, the reader would have to re-orient for each panel, which is cognitively expensive and usually defeats the comparison. Visual consistency is therefore a precondition for enforced comparison — without consistency, the small-multiple layout becomes a gallery of individually-designed charts that cannot be compared efficiently.

Applied Scenarios (2 questions)

19. You are designing a multi-panel figure for a climate report that will show three related variables — global temperature anomaly, atmospheric CO2 concentration, and sea level — over the time range 1880-2024. Each variable has different units and a different meaningful range.

(a) Should you use a small multiple, a dashboard, or some other composition? Justify your choice. (b) How should the three panels be arranged (order, sizes, shared vs. free axes)? (c) What would you put in the overall figure title versus the panel titles?

Answer **(a) Use a small multiple.** The three variables share the same kind of chart (line chart over time) and the same x-axis (year), making small multiples the natural choice. A dashboard would be wrong because there are no diverse chart types needed — all three panels answer the same kind of "how does this change over time?" question. A dual-axis single-panel chart would be dishonest (Chapter 4) because forcing different units onto one y-axis distorts comparison. **(b) Arrangement:** Three panels stacked vertically (three rows, one column). Same width for all panels, same height. Shared x-axis at the bottom (year), with the x-axis ticks on the bottom panel and the upper panels suppressing x-axis labels to avoid duplication. Free y-axes — each panel has its own y-axis with its own units (°C, ppm, mm) — because the units are fundamentally different. The shared x-axis enables temporal comparison; the free y-axes preserve each variable's meaningful range. Order: temperature on top (most familiar to readers), CO2 in the middle (the causal driver), sea level on the bottom (the consequence), telling the causal story top to bottom. **(c) Overall figure title (action):** "Three Measurements of a Warming Planet, 1880–2024" — states the finding that the three variables together tell a coherent warming story. **Panel titles:** "Temperature Anomaly," "CO2 Concentration," "Sea Level" — each names the specific variable shown in that panel. The overall title sets the frame; the panel titles label the pieces.

20. A colleague shows you a draft dashboard with nine panels in a 3×3 grid, each the same size. The panels contain: (1) a line chart of daily revenue, (2) a bar chart of regional sales, (3) a pie chart of product mix, (4) a 3D bar chart of customer types, (5) a line chart of churn, (6) a table of top products, (7) a map of sales by state, (8) a pie chart of expense categories, (9) a trend line of stock price.

Identify every compositional problem with this dashboard and propose a complete redesign. Your answer should address hierarchy, chart type choices, grouping, and consistency.

Answer **Problems identified:** 1. **No hierarchy.** All nine panels are the same size, so the reader cannot tell which is most important. At least one panel (probably daily revenue) should be the hero, larger and more prominent. 2. **Two pie charts.** Pie charts are a weak choice for the comparison questions both panels are trying to answer (product mix and expense categories). Horizontal bar charts would be clearer. 3. **A 3D bar chart.** 3D effects distort every comparison the reader tries to make (Chapter 4). The 3D bar chart should be flattened to a 2D bar chart. 4. **No grouping.** Related panels (revenue + regional sales + product mix) should be placed near each other (proximity); secondary metrics (churn + stock price) should be in a separate group. The flat 3×3 grid flattens the logical groupings. 5. **Stock price is probably out of place.** Stock price is a different question from operational metrics; it might belong on a different dashboard or a separate section. 6. **No figure title or action title.** The dashboard needs a top-level title stating the dashboard's purpose (e.g., "Q3 2024 Business Operations Dashboard") and ideally an action title for the hero chart. 7. **No source attribution.** Every panel should carry data source information, or the dashboard as a whole should have one attribution. **Redesign:** - **Hero (top-left, larger):** Daily revenue line chart with an action title showing the current value and trend. - **Top row supporting panels (to the right of the hero):** Regional sales bar chart (horizontal, ranked) and a 2D bar chart of product mix (replacing the pie). - **Middle row (secondary metrics):** Churn line chart, 2D bar chart of customer types (replacing 3D), top products table. - **Bottom row (geographic and context):** Sales-by-state map, expense category bar chart (replacing pie). Move the stock price trend to a separate section or remove it entirely. - **Consistency:** Single font family across all panels. One accent color (e.g., corporate blue) used consistently. Muted gray for reference elements. Same margin/padding around every panel. Aligned edges to a visible grid. - **Unity elements:** Dashboard title and subtitle at the top. Source attribution at the bottom. Consistent time range across the time-series panels (daily revenue, churn, stock price if kept).

Review your results against the mastery thresholds at the top. If you scored below 14, revisit Section 8.3 (small multiples) and Section 8.5 (reading order and hero placement) — those are the two most practical takeaways. Chapter 9 builds on the composition skills from this chapter to discuss narrative sequencing of charts in a data story.