I propose an analysis of the inflectional system of clauses that captures both crosslinguistic variation and differences between adult speakers and young children learning a given language. The phenomena of interest fall into two classes: 1) case marking and subject-predicate agreement; and 2) tense marking and the licensing of overt and null subjects. The major goals are:
I argue that structural case marking is a reflex of the same syntactic feature-checking relation as agreement; I label this conglomeration Accord. The presence in a clause of features involved in Accord is not an absolute convergence requirement. Rather, it is due to a preference among convergent derivations, expressed as the Accord Maximization Principle (AMP), which compares structures that differ only on uninterpretable features (in the sense of Chomsky 1995). Among those that meet all convergence requirements, only those with the most Accord relations are admissible. Children do not always successfully enforce this preference, sometimes reverting to a representation where structural case features have not been introduced. When this happens, arguments appear in the default case of the language, supplied in the Spell-Out component. Evidence from child corpus studies (both normal and Specifically Language Impaired) shows that children know both that case and agreement must be checked together and that default case must be supplied when case is absent.
With regard to subject licensing, I show that the absence of Tense features is often compatible with both PRO and overt subjects. Children's overt subjects of nonfinite clauses are thus consistent with adult grammars. The relationship between the distribution of syntactic Tense features and the meanings of clauses is governed by interface conditions on which adults and children apparently differ.