The Syntax Of Restrictive Relative Constructions

The restrictive relative construction (i.e. Rel-Con) has been a hotly discussed topic in the generative literature. Among the various approaches proposed to account for its derivation and interpretation, the two prevailing ones are the Adjunction Analysis and the Raising Analysis. Arguing that neither of them is optimal given the 'explanatory adequacy' requirement and the minimalist spirit, the author of the present study proposes a new analysis—the RelP Analysis—and argues that this analysis provides a more reasonable account of the derivation and the interpretation of Rel-Cons in typologically different languages.In the RelP Analysis, a Rel-Con is structurally a RelP or its extended projection (DP, GenP or NumP). In a RelP, the Rel° head carries an uninterpretable [+Rel] feature that has to be checked by a nominal or degree element (i.e. the relative head), which is either merged in or moved to Spec RelP, and the complement of the Rel° (i.e. the relative clause; the RC) can be any type of predicate (i.e. CP, TP, vP, PP, AdjP etc.). Theoretically, this analysis meets with the minimalist spirit, as it employs no operation other than Merge and Move. Empirically, it captures the fact that some types of Rel-Cons (i.e. those where the relative head is interpreted as an argument or a semi-argument within the R.C) exhibit properties of movement while others (i.e. those where the relative head is interpreted as an adjunct in the RC or a floppy nominal) do not. Moreover, it is proposed that the Rel° head denotes the connective marker '∧' or the Boolean operator AND; thus, the syntax of a Rel-Con is directly translated into its semantics, i.e., it denotes the intersection of the two sets denoted by the relative head and the RC respectively.When the RelP Analysis is applied to account for the structure and interpretation of the Rel-Cons in different languages, we will find that the syntactic and semantic variations with respect to the different forms of RCs can all be predicted and well accounted for. The author proposes that the specificity property of a Rel-Con isdetermined by the relativized element, while the form of the Rel-Con is detennined by both the relativized element and the properties of the RC. Specifically, when the relativized element is a degree phrase (DegP), the Rel-Con would contain no relative pronoun (RelPro) or resumptive pronoun (RP) and it is interpreted as a non-specific constituent. When the relativized element is a DP, the Rel-Con would be specific and its form is determined by whether the RC is a CP and the specification of the [wh] feature in the C° head. When it is not a CP, the Rel-Con can only have a gap or an RP as its relative marker; when it is a CP with a [+wh] C⁰ head, the Rel-Con would contain a (wh-form) RelPro; when it is a CP with a [-wh] C⁰, the Rel-Con would contain an overt or covert RP. Languages vary with respect to the following factors: the [±wh] feature in C° and relative D (D_rel), the overt/covert realization of the RP, and the possibility of relativizing from a topicalized element. As to the linear order between the RC and the relative head, it is proposed that the RC preceding the relative head is moved to Spec DP or Spec GenP while the RC following the relative head moves covertly. The landing site of the RC is determined by the kind of predicate it belongs to. Individual-level RCs raise to Spec DP, while stage-level RCs raise to Spec GenP.