It is argued that children's knowledge of the structure of multiple clause sentences and the rules governing control of missing complement subjects (PRO) can be described in terms of four grammar types which constitute separate developmental stages. These grammar types are believed to develop following an initial stage when children rely on a strategy to interpret multiple clause sentences. In order to test this hypothesis, 64 children ranging from 3;2 to 8;3 years of age were interviewed on four separate occassions. During the first interview a spontaneous language sample was collected and a Developmental Sentence Score (DSS), a measure of grammatical development, was obtained for each child. During the second and third interviews each child was askedto act out a total of 45 complex sentences. Fifty of the children returned for a fourth interview which included an acting out task and a judgment task. The five stages were associated, a priori, with specific patterns of control and the children were classified according to either grammar type or use of the initial strategy on the basis of their response patterns to a selected set of the 45 experimental constructions. The hypothesis of the four grammar types and their sequential development was supported by the fact that the children belonging to each grammar type differed significantly with respect to age and DSS scores. Furthermore, all the means were sequentially ordered in the predicted direction. Tests involving the relationship of grammar type to (1) reliability of response patterns across interviews, and (2) ability to identify semantically deviant sentences provided independent support for the theory of the four grammar types. There was only limited support for the existence of the initial stage.