Speech samples were taken from 21 children aged 16-40 months covering a wide range of mean utterance length. Presence or absence of 14 grammatical morphemes in linguistic and nonlinguistic obligatory contexts was scored. Order of acquisition of the morphemes was determined using two different criteria. The rank-orderings obtained correlated very highly with a previously determined order of acquisition for three children studied longitudinally. Age did not add to the predictiveness of mean length of utterance alone for grammatical development in terms of which morphemes were correctly used. The approximately invariant order of acquisition for the fourteen morphemes is discussed in terms of three possible determinants of this order. Frequency of use in parental speech showed no correlation with order of acquisition, but grammatical and semantic complexity both correlated highly with acquisition order.