UCL Psychology and Language Sciences


Longitudinal Leicester Corpus

It has been suggested that children's learning and accuracy of speech sounds might be related to the number of times they hear those sounds. However, it is not clear if this is due to the frequency of individual speech sounds themselves, or the larger structures which these make up (i.e. words). In addition to this, there may be general patterns (or rules) in the ambient language which either facilitate or constrain the accurate learning and production of speech. To address these issues, this study takes place over three years, investigating the early language development of six children's language. This will video and audio record the natural linguistic development of the children and the language produced by their parents/carers whilst innatural spontaneous interaction with one another. To assess children's production accuracy of low probability/frequency sounds (to see if they can produce sounds that might not be recorded in the first part of the study), the spontaneous study will be supported by an experimental task taking place from 18 months until 48 months of age which elicits certain speech sounds from the infants, via a picture/object naming exercise designed to encourage infants to produce the sounds being studied. Parent-child interactions will be transcribed using Phon, and with parental permission the resulting corpora will be made publicly available on the CHILDES database. The infants will be children of speakers of British English, a language which perhaps surprisingly does not currently have any substantial longitudinal data of child and parent interactions available.