T-Unit¶
A grammatical segment of written discourse containing one main clause and the subordinate material attached to it.
Core Idea¶
A T-unit is a segment of written discourse with one main clause and all subordinate clauses and other dependent material attached to that clause.[1][2] A second coordinated main clause starts another T-unit, even if the writer puts both in one punctuated sentence.[1] The grammatical boundary determines the unit; word counts and accuracy scores are later uses of it.
Scope of Application¶
The rule is directly supported for written language.[1][2] Hunt applied it to English school composition, and Gärdenfors used T-unit coding for Swedish student narratives.[1][2] An analyst must identify the main clauses and decide which dependent material attaches to each. Conversational fragments and backchannels require their own spoken-corpus convention; the written rule alone does not settle them.
Clarity¶
A punctuated sentence can contain several T-units.[1] Hunt shows a fourth-grader's 68-word sample written as one sentence but divided into six T-units.[1] This distinction helps separate subordination within a clause group from coordination of independent clause groups.
Manages Complexity¶
Once a passage is segmented, researchers can count T-units, words per T-unit, or subordinate clauses per T-unit without confusing those measurements with the unit's definition.[1][2] The segmentation keeps dependent material with its main clause while separating the next independent main clause.[1][2]
Abstract Reasoning¶
Find a main clause, keep its subordinate clauses and other dependent material attached, then begin a new T-unit at the next coordinated main clause.[1][2] Splitting off an attached because clause would leave a dependent fragment, whereas a new independent clause supplies its own head.[2] The rule is grammatical rather than a fixed word length or punctuation rule.
Knowledge Transfer¶
The same boundary rule applies to Hunt's English school-writing sample and Gärdenfors's Swedish narrative coding, despite their different languages and research settings.[1][2] This transfers the segmentation method, not a judgment that a longer T-unit is automatically better writing. Speech needs a separately stated convention for incomplete or elliptical turns.
Examples¶
Hunt's sample. In a 68-word passage punctuated as one sentence, Hunt identifies six T-units.[1] “I like the movie” heads one unit and its relative clause stays attached; “The captain said” begins the next.[1] Mapped back: written sample → one main-clause head per unit → dependents retained → coordinated heads separated.
Swedish narrative coding. Gärdenfors illustrates two T-units in “The panther went to the door / and he opened it,” but one in “The panther screamed / because he was in pain.”[2] These are translated method illustrations, not participant excerpts.[2] Mapped back: the second independent head starts another unit; the dependent because clause remains with its head.
Relationships to Other Abstractions¶
Current abstraction T-Unit Domain-specific
Parents (1) — outgoing typed relations
-
T-Unit includes Clause Domain-specific
Every Hunt-style written T-unit contains exactly one main clause as its grammatical head.
Hierarchy path (1) — routes to 1 parentless root
- T-Unit → includes Clause → is a kind of Constituent (linguistics)
Neighborhood in Abstraction Space¶
T-Unit sits in a sparse region of the domain-specific corpus (96th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2769 abstractions)
Nearest neighbors
- Digression — 0.77
- Discourse grammar — 0.77
- Type-Token Ratio — 0.77
- Independent clause — 0.76
- Rising declarative — 0.76
Computed from structural-signature embeddings · 2026-10-25
Not to Be Confused With¶
A Clause is an internal grammatical part of a T-unit: the written-core unit contains exactly one main clause with its dependents.[1] A clause can occur without itself defining the T-unit boundary. A punctuated sentence may contain multiple T-units. A Turn Construction Unit instead concerns a projected conversational turn, and a syntactic Constituent need not have a T-unit's one-main-clause rule.
References¶
[1] Kellogg W. Hunt, Grammatical Structures Written at Three Grade Levels, NCTE Research Report No. 3 (1965), printed pp. 20–21, PDF zero-index pp. 35–36 (viewer pages 36–37), https://files.eric.ed.gov/fulltext/ED113735.pdf. Primary report and worked 68-word, six-unit example. OCR is noisy; printed page numbers and the displayed segmentation were checked. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m
[2] Moa Gärdenfors, “Writing Development in DHH Students: A Bimodal Bilingual Approach,” Journal of Deaf Studies and Deaf Education 28, no. 2 (2023): 211–225, Methods / Linguistic Analyses and Table 2, https://academic.oup.com/jdsde/article/28/2/211/6901903. Primary study of Swedish DHH written narratives. Table 2's English sentences translate general Swedish grammatical examples and are not participant excerpts. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j