Hebrew Relative Clauses in HPSG

The form taken by the correct analysis of Hebrew RCs depends on the answer to the question: How alike are resumptives and gaps? Some previous researchers have posited mechanisms for the analysis of RPs distinct from those used to generate gaps. In their GB analyses, (Borer 1984b) and (Sells 1984) both analyze gaps as arising from movement, whereas RPs are treated as just normal pronouns, which special constraints force to be bound by an null operator in the COMP position of the RC. This pronominal coindexing provides the connection between the top of the RC construction and the RP. I will refer to this as the operator binding approach. We could mimic the essentials of this approach in HPSG as in (2) by using existential quantification to require that a RC contain a pronoun which is interpreted as the variable abstracted over (this is the meaning of the rel-ind feature; see section 4.1).1