Visually Engaging, Analytically Poor , Use With Eyes Open Figure 78.1 , Visually Engaging, Analytically Poor

A word cloud , sometimes called a tag cloud or weighted list , is a text visualization in which words from a corpus are positioned within a bounded shape and sized in proportion to their frequency or importance. Larger words appear more frequently in the source text; smaller words appear less often. The most common implementations use a packing algorithm (Jonathan Feinberg's "Wordle" layout, which D3-cloud reimplements) to fit words into a rectangle or shape without overlap.

Word size encodes frequency. That much is honest. Almost everything else in the chart is decoration: Position is meaningless. Words are placed by an algorithm that prioritises packing density, not data relationships. Two words next to each other in the cloud are not more related than two words at opposite ends. Readers who infer proximity-as-relatedness are reading information the chart did not encode. Color is meaningless in most implementations. Color is chosen to make the cloud visually appealing, not to encode a second variable.

Stephen Few has been direct about word clouds in print: they are "amongst the most useless and absurd visualizations ever devised." Jacob Harris at The New York Times called them "the mullets of the Internet." The criticism is not aesthetic , it is mechanical. The chart's central encoding (size = frequency) is corrupted by the chart's own layout decisions (word length, orientation, packing position) in ways the reader cannot correct for.

Word clouds are appropriate in two contexts. The first is presentation theatre , a slide showing audience-submitted words after a brainstorming session, where the goal is engagement rather than analysis. The second is preliminary corpus exploration , a glance at a thousand reviews to see whether the dominant terms are about price, quality, or service before committing to a more rigorous frequency analysis. In both contexts, the word cloud should be paired with the underlying frequency table.

A horizontal bar chart of the top 20 (or top 50) terms by frequency, sorted descending, gives the reader length-along-a-shared-baseline , the most accurate quantitative encoding humans can read. The reader can compare any two terms directly. Word length does not confound the encoding because all bar lengths are measured the same way. There is no wasted real estate, no decorative position, no rotational confusion. Cleveland and McGill's perceptual ranking puts bar length above area; word clouds use area; the bar chart wins.

FRAMEWORK FT Visual Vocabulary: not classified , word clouds sit outside the FT's chart taxonomy because the encoding is too unreliable to defend. Tufte principle violated: the chart's data-ink ratio is poor; word length, orientation, and position all consume ink without encoding data. Few's position: the chart is "useless and absurd" for analytical work and acceptable only as decorative orientation.

Paste this into Claude Code to generate a working version of this chart, plus its data file. The result will not be a perfect replica , the goal is that the reader can run the prompt, get a chart of this type, and read its source. Generate a complete, self-contained word cloud in D3 v7 using the d3-cloud layout. Two files:

The ideas in this chapter didn't appear from nowhere. Marshall McLuhan argued in The Gutenberg Galaxy (1962) and Understanding Media (1964) that typography itself is a medium , that the size, the weight, the spacing of letters carry meaning independent of the words they spell. The word cloud is McLuhan's claim made literal. The frequency of a term is encoded in its visual prominence. The medium is the message, rendered as a chart. Marshall McLuhan, circa 1965. AI-generated portrait based on a public domain photograph (Wikimedia Commons).

The ideas in this chapter didn't appear from nowhere. Marshall McLuhan argued in The Gutenberg Galaxy (1962) and Understanding Media (1964) that typography itself is a medium , that the size, the weight, the spacing of letters carry meaning independent of the words they spell. The word cloud is McLuhan's claim made literal. The frequency of a term is encoded in its visual prominence. The medium is the message, rendered as a chart. Marshall McLuhan, circa 1965. AI-generated portrait based on a public domain photograph (Wikimedia Commons).