Mean length of utterance (MLU) is a frequently used measure of the expressive language of young children. The suggested conventional, contemporary, clinical practice is to calculate it from a language sample of a minimum of 50 to 100 contiguous intelligible utterances. This practice places considerable strain on professionals working with young children with language disorders, for it is often impractical to devote the time needed to collect, transcribe, and analyse the recommended number of utterances. This research investigated the consistency of MLU calculated across language samples of different sizes for the same children. Transcripts of expressive language samples of research participants with developmental language impairment were analysed, with MLU being calculated on samples of varying sizes. The language samples ranged from 10 to 150 utterances. Measures of statistical differences and consistency of MLU across the various language samples were examined. Results demonstrate that, on the whole, one can reliably and efficiently determine MLU on much smaller language samples than that typically recommended, although, not surprisingly, there may be notable individual differences attesting to the vicissitudes of MLU. (Contains 3 tables, 1 note and 1 figure.)