An alternate MLU calculation: magnitude and variability of effects.

This study investigated the effect of removing elliptical question responses, imitative utterances, and single-word Yes/No responses before calculating MLU. Forty-seven language samples from preschoolers with and without language impairment were analyzed. Initial MLUs ranged from 2.0 to 6.5 across both groups. The alternate calculation procedures resulted in an average 18% increase in the MLU index, with individual samples increasing as little as 3% or as much as 49%. The magnitude of the effect was primarily related to discourse properties, although in the SLI group there were secondary relationships to language level and proficiency. The observed variation in the discourse properties of language samples makes the alternate MLU index attractive to clinicians and researchers.