The input has two main functions in optimality theory (Prince and Smolensky 1993). First, the input defines the candidate set, in other words it determines which output candidates compete for optimality, and which do not. Second, the input is referred to by faithfulness constraints that prohibit output candidates from deviating from specifications in the input. Whereas there is general agreement concerning the relevance of the input in phonology, the nature of the input in syntax is notoriously unclear. In this article, we show that the input should not be taken to define syntactic candidate sets, and that all syntactic faithfulness constraints can straightforwardly be reformulated as constraints on outputs. In view of this, we contend that the input can be completely dispensed with in syntax, in contrast to what is the case in phonology; and we argue that this asymmetry between phonology and syntax is due to a basic, irreducible difference between these two components of grammar: Syntax is an information preserving system, phonology is not.
Linguistic Review, 2002, Vol 19, Issue 4, p. 345-376