> ## Documentation Index
> Fetch the complete documentation index at: https://mehen.ophi.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Wording quality

> Style and register: passive voice, hedges, weasels, wordy phrases, adverbs, nominalizations, expletives, illusions, cliches, long sentences.

Wording quality is style and register. Where [readability formulas](/metrics/markdown/prose/english-readability)
score difficulty, wording quality flags writing patterns that consistently hurt clarity.

## What mehen emits

| Sub-metric                                                                                      | Default threshold                                                                            |
| ----------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------- |
| **Passive voice** ([write-good](https://github.com/btford/write-good) / retext-passive pattern) | Doc-type ratio from the [readability ensemble](/metrics/markdown/prose/english-readability). |
| **Hedge words** (Hyland 2005; \~165 entries)                                                    | Flag > 3 % in non-narrative docs.                                                            |
| **Weasel words** (write-good)                                                                   | Count-based.                                                                                 |
| **Wordy phrases** (\~240 entries from too-wordy / retext-simplify)                              | Per-match count / 100 words.                                                                 |
| **Adverb density** (`-ly` endings minus exceptions)                                             | Hemingway budget ≤ 1 per 100 words.                                                          |
| **Nominalizations** (`-tion`, `-sion`, `-ment`, `-ence`, `-ance`, `-ity`, `-ness`, `-ism`)      | Flag paragraph > 10 % of content words.                                                      |
| **Expletive constructions** (`^(there\|it)\s+(is\|are\|was\|were)`)                             | Per 100 sentences.                                                                           |
| **Lexical illusions** (`lower(t[i-1]) == lower(t[i])`)                                          | Zero-tolerance defect.                                                                       |
| **Clichés** (\~700 entries)                                                                     | Per 1,000 words.                                                                             |
| **Non-words** (`irregardless → regardless`, `thusly → thus`, …)                                 | Error-level flag.                                                                            |
| **Long sentences**                                                                              | Warning > 30 words, error > 40.                                                              |

## Wording Quality Score

```text theme={null}
WordingQualityScore = clamp01(
    1
  − 0.18 · sat(passive_ratio;          0.25, 0.60)
  − 0.15 · sat(hedge_density;          0.02, 0.08)
  − 0.12 · sat(weasel_density;         0.01, 0.05)
  − 0.12 · sat(wordy_density;          0.01, 0.05)
  − 0.10 · sat(adverb_density;         0.02, 0.06)
  − 0.08 · sat(nominalization_density; 0.08, 0.20)
  − 0.08 · sat(long_sentence_rate;     0.05, 0.30)
  − 0.07 · sat(cliche_density;         0.002, 0.02)
  − 0.05 · (lexical_illusions > 0 ? 1 : 0)
  − 0.05 · (nonword_count     > 0 ? 1 : 0)
)
```

WQS is deliberately orthogonal to [Filler / Lazy Risk](/metrics/markdown/filler-lazy-risk):
Filler covers repetition and specificity, WQS covers style and register.

## References

* Hyland, K. (2005). *Metadiscourse: Exploring Interaction in Writing.* Continuum.
* Williams, J. M. (1981). *Style: Lessons in Clarity and Grace.* University of Chicago Press.
* [write-good rules](https://github.com/btford/write-good).
* [proselint](https://github.com/amperser/proselint).
* [retext-passive](https://github.com/retextjs/retext-passive).
* [Hemingway editor](https://hemingwayapp.com).

## See also

* [English readability ensemble](/metrics/markdown/prose/english-readability) — orthogonal difficulty
  axis.
* [Inclusive language](/metrics/markdown/prose/inclusive-language) — separate flag list.