Tufte Visual Confections

Visual Confections: Juxtapositions from the Ocean of the Streams of Story

Overview

A confection is a single picture that gathers many separate image-events — some real, some imagined — and juxtaposes them on the flat page to make a point, narrate, explain, or list. It is neither a faithful scene (map, photograph) nor data poured into a standard format (chart, table); it is concocted on purpose, "showing all at once what never has been together" (p.127). The discipline matters because confections are the native form for explanatory posters, mixed-media dashboards, infographic narratives, and content-first interfaces — and they live or die on the quality of the thinking behind them, not the cleverness of the assembly.

§1. What a confection is — and is not

Tufte's model (the "Ocean of the Streams of Story," borrowed from Rushdie; definitions p.121):

• An event = the intersection of a noun and a verb (subject + action) — something happens.
• A plane of events = one time-slice holding every noun-verb combination at that instant.
• A story = a progression of noun-verb incidents — one strand running through time.
• An image sequence (Marey, Muybridge) = a short stretch of a single strand.
• A confection = an assembly of many image-events selected from different strands, lifted out and juxtaposed on the still flatland of paper.

What a confection does with that material: illustrate an argument, enforce visual comparisons, combine the real and the imagined, and tell another story.

Confection vs. the two things people confuse it with:

"What collage is for art, confections are for the design of information." — Tufte, Visual Explanations, p.139

Do/don't — the gateway test:

• Do (confection): assemble heterogeneous images to display information — usually expressible in words, often derived from words.
• Don't call it a confection if you are reproducing one real scene, or just formatting a dataset. That is illustration or charting; different rules apply.

§2. The two organizing structures

Every confection arranges its gathered images by one of two strategies — or both at once.

Leviathan as the model of "both" (Hobbes, 1651; pp.136–138): top half = an imagined scene (a giant artificial sovereign built from a multitude of tiny citizens, towering over mountains and spires to imply supreme power); bottom half = a grid of compartments cross-linked by both format and content. The compartments read downwards and across — each cell sized to differ from the one below but to pair with its opposite number (temporal rule left, ecclesiastical right; castle ↔ church, coronet ↔ mitre, cannon ↔ thunderbolt). It "adds up to something more than a statement of themes and something less than an argument," reasoning by analogy, metaphor, and visual-verbal parallelism. Concrete scale: a full verbal description of its elements ran ~31,200 words — an intensity of 91 words per square cm (579 per square inch).

Multifunctioning is the mark of a good confection (p.138): in Leviathan, citizens add up to the larger body, compartments link both vertically and horizontally, and the sovereign's head is a portrait of the author — single elements doing several jobs at once.

A confection can argue doctrine you should resist — Klutsis, 1930 (p.137): a Soviet election poster composes many citizens' hands merging into one giant hand — the same many-into-one figure as Leviathan, and like it an encoding of all-encompassing statist doctrine. Read the argument a confection is making, not only its craft; the imagined-scene structure serves ideology as readily as explanation.

§3. Confections as visual lists

A confection can be an inventory — a list rendered spatially.

• Pugin's churches (frontispiece, 1843; p.124): 25 churches, chapels, and schools gathered into one grand skyline, "ranged like a Gothic New Jerusalem" (Kenneth Clark). A scenic inventory.
• Scheiner's seven sunspot methods (Rosa ursina, 1630; p.122): a visual list of seven viewing techniques (darkened glass, projections, reflections) staged on a terrace of astronomers under five suns.

Lists can carry verbs, not just nouns — the "Ultimate Weed" (p.126): call-out circles around a central plant each state what the weed does (spreads quickly, resists removal, triggers allergies, poisons wildlife, resists herbicides). Drawings comment on a drawing; words and images blend into a coherent account of an imagined plant. The lesson: a confection portraying acts, verbs, and consequences is richer than a static parts-list.

Do/don't for visual lists:

• Do let the spatial arrangement itself carry meaning (Rousseau's rows of near-identical painters, p.131, express order in space — a queue — and order in time — a flow — simultaneously).
• Don't make the reader work to decode the list. Pugin's failure (p.124): building names sit in a legend six pages away, printed vertically while the frontispiece runs horizontally, so identifying one building means detecting a tiny number buried in engraving lines, turning pages, and rotating the book. A list whose key is unusable is a decoration, not information.
• Don't let the organizing scheme become its own labyrinth — Buno's 1673 mnemonic edition of the 50-book law Digest (p.125) turned each book into an emblem animal (Book 33 = a young pig) with chapters as superimposed mini-images (chickens perched on a virginal = bequests of musical instruments), all driven by an alphanumeric cipher. The memory aid required its own guide to navigate, and the scheme worked for exactly one book: the organizing apparatus became the obstacle. Tufte's antidote: organizing structures should be transparent, straightforward, obvious, natural, ordinary, conventional — demanding no hesitation from the reader.

§4. Confection vs. collage

The distinction is the whole point of the chapter: same technique (cut, paste, juxtapose), opposite purpose.

A confection is a "miniature theater of information" — a cognitive art that illustrates an argument, makes a point, explains a task, shows how something works, lists possibilities, or narrates a story. (El Lissitzky's The Constructor, 1924, pp.140–141, sits on the seam: a photomontage that is also self-exemplifying — it depicts the process of graphic thinking, each overlapping image acting as a verb linking the nouns mind/eye/hand/compass/grid/paper. Cornell's boxes, p.139, hold down the pure-collage end.)

§5. Annotation, labels, captions, and the instructed viewer

Images alone are under-determined; text fixes the reading.

• Confections from texts (title pages of Rosa ursina, Anatomy of Melancholy, Leviathan) can sketch out complex writing and make visible what is "textually invisible, obscure, or beyond words." Some images are extreme reductions (a single emblem of a story); others enlarge the text, adding figures, details, and settings the source never specified.
• The under-determination problem (p.135): Genesis says Cain killed Abel but does not say how — so the picture must invent what the text omits.
• The instructed viewer (Meyer Schapiro, p.135): a few pictured elements — one or two figures, a single attribute — can evoke an entire known story for a viewer who already holds the text (Noah in the ark, Daniel between lions). Connotations not present in the bare text get fixed by surrounding commentary, ritual, captions.

Named failure mode — the out-of-towner (p.135): a display that relies on instructed viewers to supply the exegesis will only mystify "those viewers from out of town." Insider knowledge is not a substitute for legible annotation.

Do/don't for annotation:

• Do integrate labels, captions, and surrounding text directly into the image (Burton's stanzas keyed to compartments; the Anatomy couplet announces its own method). Rousseau weaves words in — the lion's scroll names individual artists, giving a verbal account inside the visual one.
• Don't assume the audience carries the source text. Annotate as if for the stranger; reward, don't require, prior knowledge.

§6. Confections argue — they don't decorate

The recurring theme: a confection's value equals the value of its underlying idea.

• Confectionary titles match confectionary images: Rosa ursina sive sol ("the bear-rose, or the sun") is a verbal melange — roses + bears + sun — because roses and bears were emblems of Scheiner's patron, the Orsini (Ursinus = bear) family; the title page concocts roses, bears, and suns to match. Title and image argue the same point.
• A confection can be its own lecture: Tansey's Myth of Depth (1984; pp.132–133) stages Pollock walking on water, Greenberg lecturing on flatness, Motherwell studying the surface — the painting is Tansey's argument about flatness vs. depth, illusion vs. reality, complete with a numbered key (1. Noland … 7. Pollock). Descriptions of confections "seem to provoke the language of miracles."
• Babar's Dream (de Brunhoff, 1933; p.127): winged-elephant virtues (each carrying its emblem — flowers of hope, candle of knowledge, saw of perseverance, clock of patience) drive out demon-vices; an imagined moral universe that argues for personal, mind-and-individual virtue rather than any corporate or nationalist cause.

"Excellence in the display of information is a lot like clear thinking." — Tufte, Visual Explanations, p.141

Closing principle (p.151): as 15th-century perspective let the mind see diverse objects in a correct spatial context, a confection places diverse images into the narrative context of a coherent argument — making reading, seeing, and thinking one act.

§7. Failure modes (named, p.141)

Tufte's explicit list of how confections fail:

Burton's own self-aware joke about bad assembly — "Marke well: If 't be not as 't should be, / Blame the bad Cutter and not me." (Robert Burton, The Anatomy of Melancholy; pp.134–135) — names the cutter as the point of failure. Diagrams themselves are not the problem: Tufte insists reading errors come from specific local explanatory failures or untruths, not from any inherent defect of the diagram (the Chemical Atlas CO₂-cycle confection, 1854, pp.142–143, wraps itself in apologetic text fearing "unduly literal readings" — the fear is misplaced; fix the local annotation instead).

The paired bad case — the "Ultimate Designer" (p.141): facing The Constructor in the book sits a design-annual jacket that sliced four award-winners' portraits into one grotesque recombined face, captioned as the ultimate designer — a confected photomontage Tufte calls a terribly wrong guess. Same cut-and-paste technique as the masterpiece opposite; the entire difference is the quality of thought behind the assembly, not the craft of it. Judge confections on the idea, never the technique.

§8. The interface as one-time confection

A computer can sort huge stockpiles and assemble a one-time confection for an immediate, local purpose — the museum kiosk being Tufte's worked example (pp.146–150).

Principles of the content-first ("flat") interface:

• Information becomes the interface. The opening panel shows the scope of available information immediately; only a small corner is computer administration (touch-screen hint, language options). ~90% of the screen is substance — no decorative logotypes, no navigation chrome.
• Distribute in space, not time. Surface ~45 options at once rather than sequentially unveiling little bits down a decision tree.
• Each technology does its own job. The computer selects, organizes, and customizes; paper then prints a high-resolution, portable, permanent record (map + directions + a video snapshot of the visitor) — a memory the screen cannot be.
• Borrow real-world gesture. A red pointer on the live video image, linked to a red line and footprint on the map, mimics the human gesture of "go around and down that way."

Quantitative interface measures (p.150 — use these to audit any screen):

Worked numbers — the "Yearbook" book-metaphor screen (From Silver to Silica, 1991; pp.149–150): only 18% of the screen showed substance (photographers and their work); 82% was administrative debris or nothing. The spread carried 53 typographic characters where a real book spread carries 1,000–50,000. A 1990s screen already had ~5–10% the resolution of a printed map; a wasteful metaphor squanders what little it has.

Named bad-interface failure modes (p.148; "television-disease": thin substance, contempt for audience and content, short attention span, over-produced styling):

The Pioneer plaque — the eternal confection (1972/1973; pp.150–151): on a 15 × 23 cm (6 × 9 in) gold-anodized aluminum plate aboard Pioneer 10 and 11, an intensely quantified assembly — hyperfine hydrogen transition as the base unit of time/distance, a map of 14 pulsars locating the sun, planet scales spanning atoms to galaxies, outline human figures (the woman's height annotated as the binary of decimal 8 × the 21.11 cm hydrogen wavelength = 169 cm, p.150). It demonstrates the ceiling of the form: a self-contained confection legible to an unknown viewer across "the eons and light-years," carrying its own complete annotation because there is no instructed viewer to assume.

§9. Application — building your own confection

Decision order when composing an explanatory poster, mixed dashboard, onboarding scene, or cover:

1. Find the idea first. Confections stand or fall on how deeply they illuminate ideas and the relations among them — name the argument before sketching the layout (§6, §7).
2. Pick a structure (§2): imagined scene, compartment grid, or both. Use both when you must show a unifying whole and itemized parts (Leviathan, Traps and Pitfalls).
3. Gather heterogeneous image-events — maps, photos, diagrams, drawings, real and invented — from different "strands." Mixing media is the form working, not a flaw.
4. Show verbs, not just nouns (§3): depict acts and consequences, not a static parts list.
5. Make elements multifunction (§2): one image earning several jobs is the signature of a good confection.
6. Annotate for the stranger (§5): integrate labels/captions; never require insider knowledge.
7. Audit content share (§8): keep substance dominant; strip chrome, decoration, and org-chart-shaped structure.
8. Check for the cutter's art (§7): did you select to illuminate, or to win? If to win, you have propaganda, not explanation.

Reference gallery (paraphrased examples):

The Potomac lesson on medium choice (pp.144–145): the printed confectionary diagram explains why the river is deceiving — dangers beneath a tranquil surface — using cut-away and multiple views with focused annotation (20–50 words tied to each of 9 collaborating story-pictures). It beats a TV account, which would be jumpy, low-resolution, cover under a third of the material, and impose a fixed one-dimensional order; the printed page lets the reader control order, pace, and entry point. Tufte's rule: whenever possible give the audience words and images on paper, even just to supplement speech. And to activate viewers, add fine detail they can search and edit — e.g., 57 small portraits of the drowning victims and 57 dots on the map locating each death, so readers can particularize the danger by seeing where the victims walked.

Maker's knowledge (fn.24, p.145): the footnote records the graphic's provenance — its maker witnessed a drowning, spent years researching the river, and had a ranger lowered on a rope into the killer hydraulic to understand it; the finished graphic was later cast into permanent metal signs along the trail. Excellent confections come from intensity of knowledge and passion for the particular content — the "find the idea first" rule (§9, step 1) embodied.

§10. Cross-book notes

• Seeing with Fresh Eyes extends the form with modern working confections: a visitor guide to outdoor artworks arranged north/south by east/west with objects of known size (a 0.7 m dog) as the self-representing scale (p.33), the Palace Garage 3-D-pointer collage (pp.132–133), and Ch.9's recursive quilt-of-quilts visual index (pp.162–171).
• Envisioning Information pp.70–71: the Antupit-designed Lichtenstein mural spread — isolating a detail and re-placing it in full context — is a proto-confection worth studying for detail-plus-overview composition.