Language sample calculator
Paste a transcript, one utterance per line, and read off mean length of utterance in words and morphemes, number of different words and type-token ratio — with every morpheme the rules added shown token by token so you can check the calls.
- Measures
- MLU-w, MLU-m, NDW, TTR
- Rules
- Listed, switchable
- Stored
- Nothing, ever
Use the tool
The transcript is analysed in your browser and never saved or sent. Leaving the page clears it.
- MLU (morphemes)
- —
- 0 morphemes
- MLU (words)
- —
- 0 words
- Different words
- 0
- NDW
- Type-token ratio
- —
- 0 utterances
Language sample: no utterances
The counting rules, in full
- One utterance per line; a line ending in . ? or ! is also split.
- Each word is one morpheme. Hyphenated and compound words count as one.
- Contractions count as two: it's, can't, we're, I'll, I've, you'd, I'm.
- Inflections add one: plural or possessive -s, third-person -s, past -ed, progressive -ing. Irregular forms (went, children, was) stay at one.
- Fillers (um, uh, er, mm, oh) are dropped; an immediately repeated word ("I I want") counts once.
- Unintelligible segments marked xxx or (unintelligible) are dropped.
The rules, stated
Morpheme counting follows a stated approximation of Brown's conventions: one morpheme per word; contractions count two; plural, possessive and third-person -s, past -ed and progressive -ing add one, with a list of irregular forms that stay at one; fillers are dropped; an immediately repeated word counts once; unintelligible segments marked xxx are dropped. Each rule can be switched off, and the full list is on the page.
No automatic rule set matches a hand count on every transcript — over-regularised forms like "runned" are counted as past -ed, which is usually what you want, and an unusual word ending in -s may be split when it should not be. The utterance-by-utterance view exists so you can see each call and adjust by hand.
Type-token ratio and NDW
Types are distinct word forms after lower-casing and punctuation removal; tokens are the counted words. Both measures depend on sample length, so compare like with like — the same number of utterances, or the same number of words — across sessions.
Why nothing is stored
A child's language sample can identify them, so this page keeps nothing: there is no draft, no history, and a refresh starts clean. Copy the summary line into the note.
Questions, answered
How should I segment utterances?
One per line is safest. A line that contains sentence punctuation (. ? !) followed by a space is also split there.
Does it handle other languages?
The inflection rules are English. For another language, turn off "count inflections" and the page reports MLU in words, NDW and TTR, which do not depend on morphology.
Try SideClick free for 30 days
Full capability, no card required. Your data is sealed on our server and we cannot read it.