- Method
- Utterances are split with the shared SLP utterance parser (src/lib/slp/utterance-parser.ts) and tokenised with lowercase + leading/trailing punctuation stripped. Each non-empty utterance contributes one main clause. The detector walks tokens after the first token; each token that matches the subordinator list (because, when, while, if, although, since, before, after, until, that, who, whom, whose, which, where, why, how, what, as, …) adds one subordinate clause. Clausal density = total clauses / total utterances. Developmental bands follow Loban (1976) and Nippold (2009) reference values.
- Last Validated
- 2026-04-08