Dense sampling

This chapter describes the methods used to develop much denser samples of children’s naturalistic speech than have previously been available. Most longitudinal corpora capture an estimated 1–2% of children’s speech. Depending on the exact sampling regime, the new dense corpora capture an estimated 7–15%. Dense sampling is important in assessing the productivity of children’s grammars and in the collection of rarer structures. It also allows much more reliable quantitative comparisons between the input and the child’s developing system as well as the use of computational and modeling methods that cannot be used with corpora of smaller sizes. The collection and transcription of these denser corpora is complete or under way for English, German, Polish, Japanese, Estonian, and Finnish. The limitations of the method are the extensive resources that are required and, given this, the fact that corpora can only be collected from a very small number of children.