Abstract

This paper investigates the realisation of marked exponents in a parallel derivational grammar. A parallel grammar model, in which the syntactic, semantic and phonological structures are built up in parallel, has the advantage that it can handle phonologically conditioned syntactic effects more efficiently than a sequential model. However, it is at a disadvantage with marked exponents whose realisation depends on features that are higher in the tree, since they are not available yet when the exponent is selected. This problem can be solved if we assume that marked exponents are stored in the lexicon with the relevant features, but this has to be done in such a way that Merge is still able to handle such elements.