Methodological considerations when using technology for automated vocal analysis (LENA) with young children.

TitleMethodological considerations when using technology for automated vocal analysis (LENA) with young children.
Publication TypeConference Proceedings
Year of Conference2013
AuthorsBlackwell, A, Babayigit, S, Roulstone, S
Conference NameCogDev (BPS)
Date Published09/2013
Conference LocationReading University
Abstract

Background: Traditionally, child language use and interactions have been explored through short observations in children’s homes or laboratory-based playrooms. However, using new technology (LENA: Language ENviroment Analysis) up to 16 hours of audio can be recorded from the child’s natural environment, using a small recorder worn by the child. The software provides automated data analysis.
Aims: To assess feasibility and reliability of using LENA to capture information about children’s communicative environments in preparation for a full-scale longitudinal study of children with language delay.
Methods: A pilot study was carried out with two typically developing children, aged 2;1 and 3;0, and their primary caregivers. LENA recordings were collected for each child, 16 and 3 ½ hours respectively. Automated analysis provided frequency histograms of child vocalisations, adult words, conversational turns and audio environment information (e.g. TV). These were used to identify high levels of language use from which five-minute samples were transcribed for detailed analysis of language use and parent-child interaction.
Results: LENA demonstrated clear, high quality audio. However, methodological considerations were highlighted, particularly the problem of unintelligible, or overlapping speech for accurate transcription of child language. Reliability of the automated analysis was also questionable in some instances, including incorrectly coding ‘TV speech’ as ‘real’ adult words.
Discussion: The LENA system is a useful, practical tool for collecting large amounts of data. However, there are problems with unintelligible speech, which could be more pronounced for children with language delay. Consequently, LENA may be more suitable for transcription of language input. Xu, Yapanel and Gray, (2009)demonstrated 71% sensitivity agreement for LENA TV coding with human transcription from 70 hours of audio. However, issues with reliability may be of greater importance to studies with a small samples using detailed analysis, therefore checks should be carried out according to individual aims and sample size.
Xu, D., Yapanel, U. and Gray, S., (2009) Reliability of the LENA Language Environment Analysis System in Young Children’s Natural Home Environment. Report number: LTR-05-2.Boulder, CO (Available from http://www.lenafoundation.org/Research/TechnicalReports.aspx): The LENA Foundation.