Relative Clauses

next up previous
Next: The Analysis of Up: Linguistics in MiMo2 Previous: Spanish Verb Movement

Relative Clauses

The implementation of relative clauses, both restrictive and non-restrictive, covers NP- and PP-extraction.

Figure 4: Relative clauses

The implementation differs from the analysis proposed in [p. 77]hpsg in two respects. First, [34] argues that the relation between the antecedent noun (phrase) and the relative pronoun, marked by the subscript above, is the sharing of semantic information only. This may suffice for English but it does not for Dutch. Dutch requires agreement of syntactic gender as well, as the following examples show:

het meisje dat/die ik gisteren ontmoette

the (neuter) girl that I met yesterday

de jongen die/dat ik gisteren ontmoette

the (nonneuter) boy that I met yesterday

It shows that the information shared between antecedent and relative pronoun should not only consist of semantic but also of syntactic information.

The second difference between the HPSG-analysis and ours concerns unbounded dependencies like in Figure 4, topicalisation and wh-movement. The SLASH feature employed in HPSG brings about full unification between the antecedent and the gap, thus identifying gap and antecedent entirely. It is however not desirable to have full unification between gap and antecedent, as is shown by the following data:

the man whom/who/[ ]/that I spoke to

the man to whom/who/[ ]/that I spoke

The differences between the pied-piping examples and the data that show preposition stranding can only be accounted for if we distinguish between gaps and overt antecedents. In our analysis, the relation between gap and antecedent is brought about by the unification of head features only. This allows gaps and antecedents to be different in other respects.

As to structural differences between English and Dutch w.r.t. relative clauses, the latter only allows preposition stranding in some special cases whereas English allows it quite freely. P-stranding in Dutch is only allowed when the complement of the preposition is a non-human pronoun, in which case it obligatorily appears in the so-called R-form and precedes the P of which it is a complement (cf. [53]).

de katedraal waarnaar Marie kijkt / waar Marie naar kijkt

the cathedral which_at Mary is looking / which Mary at is looking (lit.)

the cathedral at which Mary is looking / which Mary is looking at

de jongens naar wie Marie kijkt / wie marie naar kijkt

the boys at whom Mary is looking / whom Mary at is looking (lit.)

the boys at whom Mary is looking / whom Mary is looking at

Unification of gap and antecedent in unbounded dependencies is achieved by means of the gap-threading technique [30]. The method allows a straightforward analysis of the differences described. The Dutch PP consisting of an R-pronoun and a preposition allows the pronoun to be gapped and unified with an antecedent. The same holds for English PPs. Dutch PPs consisting of a P and a non-R pronoun are islands to gap-threading. The P-complement cannot be a gap, hence no preposition stranding will occur in these cases. The PP can of course be gapped in its entirety, yielding the pied-piping variant.

next up previous
Next: The Analysis of Up: Linguistics in MiMo2 Previous: Spanish Verb Movement

Gertjan van Noord
Thu Nov 24 19:09:23 MET 1994