Clause Architecture: Subordination, Complexity, and Buried Verbs

Sentence length tells you how far a reader travels. These four readings tell you what the sentence is built out of. They measure structural choices that change how writing reads, and they say nothing about whether it is correct. Inkbreaker does not check grammar, and every one of these readings carries that note.

Subordination Rate

The share of your sentences that contain at least one dependent clause. The engine looks for 16 subordinating conjunctions (because, although, when, if) and five relative pronouns (that, which, who, whom, whose).

Read that definition carefully, because it is presence, not depth. A sentence carrying five stacked dependent clauses scores exactly the same as one containing a single when. The metric tells you how often you reach for subordination, never how far you take it.

• Fiction: 0.20 to 0.55. Dependent clauses give prose interiority and scene layering.
• Nonfiction: 0.25 to 0.60, the highest ceiling. Clause embedding is how argument builds.
• Technical: 0.25 to 0.55. Conditions and edge cases need dependent clauses.
• Journalism: 0.20 to 0.50. Attribution and context require them.
• Blogging: 0.15 to 0.45.
• Copywriting: 0.10 to 0.35, the tightest. Heavy subordination kills copy.

Poetry never sees this reading. Two deliberate carve-outs keep the count honest: as inside an as … as comparison is skipped, and demonstrative that at the start or end of a sentence is skipped. Than is excluded entirely.

Sentence Complexity

Every sentence is sorted into one of four buckets: simple, compound, complex, or compound-complex. Your score is the normalized entropy across those four proportions, which is a way of asking whether you use all four or live in one.

The classifier is strict about what counts as compound. It needs a semicolon followed by a word, or a comma followed by a coordinating conjunction. A bare and does not qualify, so I came and I saw is filed as simple. Comma splices land there too. This is a heuristic, not a parser.

Benchmarks are minimums only, with no ceiling anywhere: fiction 0.50, nonfiction 0.40, journalism 0.35, technical and blogging 0.30, copywriting 0.25. Copy is allowed to be heavy on simple sentences. Fiction is asked to move around.

Nominalization Density

Nominalization is the habit of turning a verb into a noun and then propping it up with a weak one. We made a decision instead of we decided. The engine counts words of nine or more characters ending in -tion, -sion, -ment, -ance, -ence, -ity, or -ness, minus a 69-word whitelist, per thousand words.

The nine-character floor exists to keep nation, vision, and moment out of the count. It works, and it has a cost: genuine short nominalizations are invisible to it. Treat the number as directional.

• Copywriting: 10 per thousand. Copy lives on verbs.
• Fiction and blogging: 15. Bureaucratic nouns slow narrative.
• Journalism: 18.
• Nonfiction: 25.
• Technical: 40, the highest tolerance of any type.

These are ceilings. There is no floor, because no genre needs a minimum number of buried verbs.

Syntactic Variety

A composite: the average of five structural diversity signals, each clamped between 0 and 1. It combines sentence complexity variety, sentence opener variety, paragraph opener variety, template entropy from Structural Predictability, and sentence rhythm variance. The benchmark is a minimum of 0.35 for every writing type.

Because it is a composite of composites, it is the one reading here you cannot act on directly. If it comes back low, the useful move is to look at the five components and find which one is dragging. Note that on short pieces the paragraph-opener component can be unavailable, and it feeds in as zero when it is, which pulls the composite down for reasons that have nothing to do with your sentences.

It measures variety. It does not measure quality. Prose can vary constantly and go nowhere.

Reading the four together

The combination that most often signals real trouble is high nominalization with low subordination: nouns doing the work verbs should do, in sentences that never subordinate one idea to another. That is the shape of prose assembled rather than thought through.

High subordination with low complexity variety is a different problem and a milder one. It usually means you have found one sentence shape you like and are running it repeatedly. Open a piece in the editor, run Grade this passage, and look at the complexity distribution rather than the entropy score. Seeing which bucket holds most of your sentences is more useful than the number summarizing it.