Measurement and Interpretation of Longest Utterances in Child Language Samples

The measurement of the length of longest utterances (LLU) may offer a useful view of the upper boundaries of young children’s expressive grammar. In this study, LLU was examined in terms of its relationship to mean length of utterance (MLU) and other statistical properties of language samples, using the longitudinal data from Brown’s three participants as examples. Frequency distributions of the utterance lengths for each language sample were examined for skewedness, kurtosis, and modality. In addition, utterance length distributions were modeled with a best fit line to determine slope, which was found to be negative and became more level as participant age increased. Correlations between MLU and LLU within language samples and between subsequent language samples were calculated. The measures were found to be highly correlated and early measures were predictive of later measures. Models of different sizes of language samples were created by randomly selecting subsets of 50, 100, 200, and 300 utterances. Measures of LLU were relatively stable with low levels of variability at sample sizes of 100 utterances or more. Implications for measuring and interpreting LLU are discussed