ALL-WRIT0004 · paragraph_length_uniformity

Measure uniform paragraph length without inferring authorship.

This is a deterministic rule for all languages. Read its implementation.

Definition

Split each prose section at blank lines once the non-prose blocks are removed, and keep the paragraphs holding at least minimum_words words. For every section holding at least minimum_paragraphs of them, compute 100 * max(0, 1 - MAD / mean) over the paragraph word counts and return the highest value any section reaches.

Paragraph rhythm is the same observation as sentence rhythm one level up. A writer working through an argument spends more words where the idea is harder, so paragraphs of identical weight suggest a template rather than a train of thought. It is a pacing measurement and never an authorship claim.

Evidence

The finding names each section, its paragraph count, and its bounded uniformity percentage. The value is the highest section uniformity in the document, and equal paragraph lengths produce one hundred.

Exceptions

A section holding fewer than minimum_paragraphs qualifying paragraphs is skipped rather than measured. Paragraphs under minimum_words are dropped first, so a run of one-line notes cannot drive the score. Templates, reference manuals, release notes, and deliberately parallel explanations are uniform because the form calls for it, so a high value there is a description rather than a defect.

Examples

Four paragraphs of fifty words each return 100. Paragraphs of 30, 45, 70, and 100 words have a mean of about 61 and a mean absolute deviation of about 23, so they return about 62. A section holding three qualifying paragraphs is skipped under the default minimum_paragraphs.

References

• Cites “Vale AI Tells”, experimental ParagraphLengthVariance rule. Open reference
• Cites “Pangram documentation”, AI writing patterns. Open reference