Skip to main content

DATASET 04 · Developmental linguistics

Children's Language Development Longitudinal Corpus

540 hours of natural interaction from 260 children tracked over 36 months, delivered with CHAT transcripts and developmental measures.

SpeechVideoTextDevelopmental linguisticsChild psychology
Children's Language Development Longitudinal Corpus · cover artwork
Scale
540 hours
Participants
260 children
Languages
Mandarin (children's spontaneous speech)
Annotation
CHAT-format transcription, Part-of-speech and syntactic annotation, Phonological error annotation, Parent-child interaction coding
Formats
WAV / MP4 / CHAT / JSON
Sync
Audio-video sync error ≤ 2 ms
License
Research license (ethics-approved projects only)
Updated
June 30, 2026

01 / Composition

Scale, key metrics and data streams

Recording the emergence and change of language along a developmental timeline

540 h
Total duration
260
Children
36 months
Tracking period
214
Households
Data composition and specifications
StreamSpecificationNotes
Audio48 kHz / 24-bitLavalier microphones, with child and caregiver on separate tracks
Video1920 × 1080 · 30 fpsSingle fixed camera in the natural home setting
TextCHAT / UTF-8CHAT transcript with part-of-speech and syntax layers

The table lists capture parameters for a sample batch; final specifications follow the documentation delivered with the data.

02 / Dataset notes

Collection, annotation and quality control

The notes below explain how the data was collected, annotated and verified, and can be used as a basis for methodological citation and proposal assessment.

Overview

The Children’s Language Development Longitudinal Corpus follows 260 children from 214 households for 36 months, yielding 540 hours of natural interaction recorded between February 2023 and February 2026. Children enrolled between 18 and 42 months of age, and each household completed a recording every three months, roughly 45 minutes per visit, covering three activity types: free play, shared picture-book reading and mealtime conversation.

The corpus is stratified by age, gender, home language environment (monolingual or bilingual) and caregiver education. Twelve children withdrew during the study, and both the reason and the timing of each withdrawal are recorded in the metadata so attrition bias can be analyzed.

Collection protocol

Collection happens mainly in the natural home setting, avoiding the influence a laboratory environment can have on children’s expression. Child and caregiver each wear a lavalier microphone and are recorded on separate tracks at 48 kHz / 24-bit. A single 1920 × 1080 · 30 fps camera covers the play and reading areas and is not moved during recording, to avoid distraction.

A research assistant sets up the equipment and checks audio calibration before recording, then leaves; a parent operates a simple start and stop flow. Every household signs an informed consent form, and recording stops immediately if a child shows distress. Segments that reveal the identity of household members are masked and voice-altered before ingest.

Annotation

Transcription follows the CHAT convention, with phonological and interaction layers added on top.

LayerContentScaleReview
TranscriptCHAT sentence-level transcription, pronunciation variants, repetitions and self-repairsAbout 600,000 sentences25% double-blind review
PhonologySegmental errors, syllable structure, tone realizationAbout 180,000 error instancesDual review by phoneticians
SyntaxPart of speech, sentence length, clause depthFull corpusAutomatic annotation + 20% manual correction
InteractionJoint attention, turn-taking, expansion and recast10 behaviour classesDual annotation of key segments

Quality & verification

The median audio-video sync error is 1.5 ms, with the 95th percentile at or below 2 ms. CHAT transcription agreement is κ = 0.90, phonological error agreement is κ = 0.84 and parent-child interaction coding agreement is κ = 0.86. Fifteen percent of each batch is double-blind reviewed, and 5% of that is finally reviewed by developmental linguistics researchers. Deliverables include age-norm tables, transcription conventions, annotation manuals and a longitudinal matching guide.

Research questions

  • The order in which Mandarin-speaking children acquire tones and consonants.
  • Longitudinal growth curves for vocabulary size and syntactic complexity.
  • How expansion and recasting during shared reading affect language development.
  • Early phonological and interactional markers of language delay.
  • How a bilingual environment shapes children’s code-switching patterns.

Access & licensing

This dataset is available only to research projects that pass ethics review; applicants must submit a study protocol, ethics approval and child data protection measures. Data is delivered as WAV, MP4, CHAT and JSON, filtered by age band and activity type, with longitudinal matching IDs irreversibly mapped. Commercial use is assessed case by case, and source video is generally not provided. You can inspect the CHAT field structure through the sample data page, or ask about the application process on the contact page.

03 / Typical use cases

Questions this dataset can answer

The directions below are typical research and product applications; a specific analysis plan depends on the study design and statistical method.

  • 01

    Child speech acquisition order research

  • 02

    Vocabulary growth and syntactic complexity analysis

  • 03

    Early markers of language delay

  • 04

    Parent-child interaction quality assessment

NEXT STEP

Need a different collection setting or annotation scheme?

An existing dataset can be licensed as it stands, or extended with additional speakers, scenarios and annotation layers around your study design.

Reply within one business day · NDA and data licensing supported · Dedicated academic track