EMT-SF Derived Corpus

	Ann Kaiser Special Education Vanderbilt University ann.kaiser@vanderbilt.edu		Pamela Hadley Speech and Hearing Science University of Illinois phadley@illinois.edu
	Megan Roberts Communication Sciences and Disorders Northwestern University megan.y.roberts@northwestern.edu

The complete corpus is available in this zip file.

Citation

Roberts, M. Y., Hadley, P. A., & Kaiser, A. P. (2018). Maximizing outcomes for preschoolers with developmental language disorders: Testing the effects of a sequentially targeted naturalistic intervention. Identifier NCT03782493. https://clinicaltrials.gov/study/NCT03782493

Roberts, M. Y., Hadley, P. A., & Kaiser, A. P. (2025). Maximizing Outcomes for Preschoolers with Developmental Language Disorders. Retrieved from https://osf.io/yad3f/

Hadley, P., Roberts, M., Harrington, E., Jones, M., Preza, T., Zanzinger, K., Kaat, A., Grauzer, J., & Kaiser, A. (in press). Short-term effects of a caregiver-mediated intervention for preschoolers at risk for developmental language disorder: A randomized controlled trial. Journal of Speech, Language, Hearing Research. Advance online publication. https://doi.org/10.1044/2026_JSLHR-25-00387n

Jones, M., Hadley, P., Preza, T., Harrington, E., Zanzinger, K., Kaat, A., Grauzer, J., Kaiser, A., & Roberts, M. (under review). Improving outcomes for preschoolers at risk for developmental language disorder: A randomized trial of Enhanced Milieu Teaching – Sentence Focused (EMT-SF).

In accordance with TalkBank rules, any use of data from this corpus must be accompanied by at least one of the above references.

Corpus Description

This dataset contains de-identified transcripts data from a randomized controlled trial examining the effects of the Enhanced Milieu Teaching–Sentence Focused (EMT-SF) intervention, an 18-month, multi-phase caregiver-mediated intervention for children at risk for developmental language disorder (DLD) (ClinicalTrials.gov: NCT03782493). The trial enrolled 108 children between 30 and 31 months of age and their caregivers. Children were randomly assigned to either the EMT-SF treatment group or a control condition.

Toddlers and caregivers were enrolled in the clinical trial between April 2019 and June 2023. Of these, three children received an autism diagnosis during the course of the study and were excluded from all analyses. Seven families did not consent to additional data sharing and are not included in this dataset. As such, published results from this database may differ from replication analyses using the archived transcripts.

The majority of families participated in all aspects of the trial via telehealth. Language samples were collected remotely via Zoom at seven timepoints across four different contexts. Sample size for each timepoint and context are detailed below:

Timepoint	Book	Snack	Play	SDPT
30 months	97	95	98	89
33 months	95	97	96	86
36 months	91	90	91	89
39 months	92	93	93	92
42 months	90	89	90	90
45 months	90	89	90	90
49 months	91	89	92	91

Participant Characteristics

English was the only language spoken to the child in the home. At recruitment, children were excluded if they:

• Had a diagnosis of a neurodevelopmental disorder other than developmental language disorder (DLD) or language delay.
• Had experienced a major medical event, medical condition, hearing loss, or had observable oral motor difficulties.

All toddlers demonstrated:

• Nonverbal cognition within normal limits as measured by a T score of 35 or greater (within 1.5 SD of the mean) on the Visual Reception subscale of the Mullen Scales of Early Learning (Mullen, 1995).
• no clinical levels of autism-related behaviors, as measured by the TELE-ASD-PEDS (Corona et al., 2020)
• Low to below average receptive language abilities, as measured by a standard score of 95 or less on the Preschool Language Scales, 5th Edition (PLS-5; Zimmerman et al. 2011) Auditory Comprehension scale.
• Below average expressive language abilities, as measured by a standard score of 80 or below on the PLS-5 Expressive Communication scale and/or two scores at or below the 10th percentile on the Words produced, Mean length of three longest utterances, or Complexity on the MacArthur-Bates Communicative Development Inventories: Words and Sentences (CDI; Fenson et al., 2007).

Language Sample Contexts

Language samples were collected via Zoom to capture the caregiver-child dyad interacting in a natural environment. A trained assessor was on the call to address any technology issues and ensure the caregiver and child stayed in frame. Children participated in a 30 min caregiver-child interaction (CCX) involving three different language sampling contexts and the SDPT AND T49BUS tasks:

• Book reading: At each timepoint, families received new books and were recorded engaging in book reading with their child for 5 minutes. Any utterance that is read verbatim from the book is marked with a [^ read] code in the transcripts.
• Snack: At each timepoint, caregivers provided a snack to their child and were recorded engaging as they normally would during a snack for 5 minutes.
• Free play: At each timepoint, families received a new set of toys and were recorded engaging in free play as they normally would for 20 minutes.
• SDPT: Children also participated in a structured elicitation task designed to assess sentence diversity in young children under supported conditions. Sentence Diversity Priming Task (SDPT; Krok et al., 2022; Krok et al., 2026): The SDPT is presented as an interactive picture book in which the adult and child take turns describing the actions of animated characters and objects. The task uses 24 pairs of prime-target sentences, including 12 with transitive verbs and 12 with intransitive verbs.
• T49BUS: At 49 months, children completed the Renfrew Bus Story (Glasgow & Cowley, 1994). Children’s story retelling was transcribed for later scoring.

Transcription Procedures

Prior to transcription, research assistants demonstrated adequate levels of agreement on three consecutive training transcripts, achieving at least 90% morpheme-by-morpheme reliability for adult transcription and 80% morpheme-by-morpheme reliability for child transcription.

Transcription was completed in multiple passes by trained research assistants.

• 1st transcription pass: 1st transcriber listened to adult and child utterances no more than 3 times. Words/morphemes that could not be transcribed within 3 repetitions were marked as unintelligible.
• 2nd transcription pass: 2nd transcriber completed a consensus pass, relistening to the recording and confirming words/morphemes initially transcribed as well as marking any part of the transcript that could not be confirmed as unintelligible.

Transcripts were originally created in Systematic Analysis of Language Transcripts (SALT) and converted to CHAT using the Salt2Chat program during the final years of the project. Because of the various deviations of SALT format from CHAT format and the lack of availability for audio, it was not possible to include these data in the main CHILDES database.

Further details of data collection and recording, participant demographics, examiner and transcriber training, play scenarios, and transcription conventions can be found in resources and publications made available on the trial’s OSF website: https://osf.io/yad3f/.

Dataset

Acknowledgements