Are There "Double Relative Clauses" in Korean?

It has been claimed that Korean allows the relativization of another relative clause, deriving the "double relative clause." The presence of "double relative clauses" has led some researchers to argue that Korean relative clauses do not involve any operator movement, but instead involve a mechanism such as unselective binding (Sohn 1980, Y.-S. Kang 1986), where an operator binds variables in situ. In this squib, we argue that there is no true "double relative clause," thus no real threat to the operator movement analysis for relative clauses in Korean. More specifically, we propose that "double relative clauses" are derived from double nominative constructions, through relativizing the first nominative NP that originates from an IP-adjoined position. Given our analysis, "double relative clauses" are not instances of island violations, and the operator movement analysis for relative clause formation in Korean can thus be maintained.

1 Issues

In Korean, the main verb in a relative clause is inflected with an adnominal morpheme -(n)un (glossed as ADN), which indicates that the clause is modifying a noun. The head noun occurs to its right, Korean being a head-final language. An example of a relative clause with a subject gap is given in (1).

(1) [_NP[_IP e ppang-ul   mek-nun] ai]

       e bread-ACC eat-ADN   kid

'the kid who is eating bread'

Although Korean does not have any overt relative pronoun, it is standardly assumed that there is an empty relative pronoun operator in Spec, CP that is syntactically associated with a gap in the relative clause (see D.-W. Yang 1989, H.-K. Yang 1990). The relative clause in (1) can be structurally represented as in (2) (throughout, English words are used in tree structures for convenience). The syntactic relation between the empty operator and the subject gap is instantiated by coindexation. Under the operator movement analysis, the subject gap ei is a trace of the empty operator Op_i.¹

(2)

"Double relative clauses," however, pose a problem for the operator movement analysis because they appear to involve relativization of another relative clause. For instance, in (3a) the object NP that is associated with kangaci-ka ('dog-NOM') has relativized and then the subject NP that is associated with ai 'kid' has relativized ('dog' is the subject of 'die'). The problem for the operator movement analysis is caused bythe subject gap (e_i): it appears to be a trace left byan island-violating movement out of another relative clause. This is illustrated in the tree structure in (3b).²

(3) a. [_RC1[_RC2 e_i e_j cohaha-nun] kangaci-ka_j cwuk-un] ai_i

         e_i e_j like-ADN    dog-NOM_j   die-ADN   kid_i

'the kid [_RC1 who_i the dog [_RC2 which_j e_i liked e_j]
died]'

'the kid who the dog which [he] liked died'

b.

More examples of "double relative clauses" are given in (4).

(4)

1. a. [_RC1[_RC2 e_i e_j tha-ko tani-nun]  cha-ka_j   mesci-n]

            e_i e_j   ride drive-ADN car-NOM_j stylish-ADN

   sinsa_i

   gentleman_i

   'the gentleman [RC₁ who_i the car [RC₂ which_j e_i is driving
   e_j] is stylish]'

   'the gentleman who the car that [he] is driving is stylish'

2. b. [_RC1[_RC2 e_i e_j kackoiss-nun] khemphwute-ka_j

            e_i e_j have-ADN     computer-NOM_j

   Mac-i-n]      kyoswu_i

   Mac-COP-ADN professor_i

   'the professor [_RC1 who_i the computer [_RC2 which_j e_i has
   e_j] is Mac]'

   'the professor who the computer which [he] has is [a]
   Mac'

Because of examples like (3a) and (4), some have taken an unselective binding approach to the analysis of relative clauses in Korean. Under such an approach, the gaps in the relative clause are pronominal variables and are bound by the empty operator in situ (Sohn 1980, Y.-S. Kang 1986). This approach then predicts that there should be no island...