The generalizability of spontaneous speech data: the effect of occasion and place on the speech production of children

This study evaluated the effect of occasion and place on the generalizability of spontaneous speech data. Language samples were obtained from four normally developing children aged 3;8 and 3;9 from Dutch-speaking professional homes. Four samples of 100 utterances were gathered from each child, two samples at home and two in a clinical setting. All samples were recorded under unstructured conversational conditions. The time gap between first and last date of recording was no longer than 21 days. GRAMAT, a morphosyntactic model based on LARSP, was used for analysis at three levels: clause, phrase and morphological level. The data obtained were statistically analysed using loglinear model fitting. The results revealed a lack of generalizability at all three levels of morphosyntactic analysis. Some consequences of this outcome are discussed.