Language Sample Analysis (CHAT Transcription and Mean Length of Utterance)

Description

Transcribe natural speech in a standard format, segment it into utterances and morphemes, and compute measures such as mean length of utterance to compare over time or against norms.

Brown's team recorded children monthly at home, transcribed the tapes and analysed them in a weekly seminar. Mean length of utterance is the average number of morphemes per utterance in a spontaneous sample, an index of grammatical development up to about 4.0, with Brown's five stages in steps of 0.5. About half an hour of conversation yields roughly 100 eligible utterances. Utterances are segmented by a two-of-three rule (bounded by silence, terminal intonation, a natural grammatical unit); unintelligible utterances and repetitions are excluded; Brown's rules define what counts as a morpheme. CHAT is the standard transcription format, and CLAN programs compute MLU, vocabulary lists and grammatical profiles automatically, while the shared CHILDES database allows pooled, standardised comparison. The disciplined segmentation is what transfers: it makes length and complexity of spoken input comparable across people and sessions.

Attributes

Loss type
Unmigrated
Effort
Weeks

Steps

  1. Record about 30 minutes of natural conversation.
  2. Transcribe it exactly in a standard format such as CHAT.
  3. Segment utterances using the two-of-three criteria and exclude unintelligible utterances and repetitions.
  4. Divide utterances into morphemes using explicit rules (or count words for highly inflected languages).
  5. Compute MLU and other measures with software such as CLAN.
  6. Compare with normative data or across sessions and groups.

Sources

  • The Development of Language (Gleason & Ratner, 2024)

Also known as

  • Mean length of utterance
  • MLU
  • CHILDES
  • CHAT transcription
  • CLAN
  • Spontaneous speech sample analysis
  • Brown's stages

Categories

  • Content
  • Linguistics, Semiotics & Discourse