Turkish ATB Relativization: Implications on Relative Clause Formation

1. Introduction^[1]

Ever since Ross (1967: 107) defined the across-the-board (henceforth ATB) movement as an operation “which moves a constituent out of all conjuncts of a coordinate structure”, various properties of these constructions have been discussed in Generative literature. Discussions on the ATB movement in wh-movement constructions have been extended into the other A′-contexts involving topicalization, relativization, scrambling, and a number of A-movement contexts such as passivization and raising^[2]. An example of ATB relativization is given in (1).

1. The book that [_Conj.1 John likes _ ] and [_Conj.2 Mary hates _ ]

(1) above involves a NP the book which is interpreted as the complement of the verbs in both conjuncts. Similar structures are also found in Turkish where a single head noun is interpreted inside the two conjuncts Ahmet hates and Ali loves. Consider (2)^[3].

1. ^[4]

   1. Subject Relativization: [CP [TP ec_i [vP … verb-(y)An ] ] ] NP_i

      [Rel. Cl. ec_i koş-an] adam_i

      run-REL man

      ‘T he man who is running.’

   1. Non-subject Relativization: [CP [TP subject_k [vP … ec_i … verb-DIK-Agr_k ] ] ] NP_i

      [Rel. Cl. Ben-im_kec_i gor-duğ-um_k] adam_i

      I-GEN see-REL-POSS1SG man

      ‘The man that I saw.’

   [_Conj.1 Ahmet-in ec-ABL_inefret et-tiğ-i] [_Conj.2 Ali-nin ec-ACC_i

   Ahmet-GEN hate-REL-3SG Ali-GEN

   sev-diğ-i] kadın_i

   love-REL-3SG woman

   ‘The woman_i that Ahmet hates ec_iand Ali loves ec_i.’

In (2) gaps in the complement positions of the verbs in both conjuncts are interpreted with the head noun kadın ‘woman’. Note that a single item, the head noun which is interpreted in both conjuncts, has different case requirements: ablative in the first conjunct but accusative in the second. This “case mismatch” is also observed in complex relative clauses where the shared element is interpreted in the complement positions of the verbs in adjunct clauses which are embedded under relative clauses as exemplified in (3).

1. [_Conj.1 Ahmet-in _[Adj.Cl.ec-ABL_inefret et-tiğ-in-den] ec_ikov-duğ-u]

   Ahmet-GEN because.hate-3SG-ABL fire-REL-3SG

   [_Conj.2 Ali-nin _[Adj.Cl.ec-ACCisev-diğ-in-den] ec_i

   Ali-GEN because.love-3SG-ABL

   kolla-dığ-ı] kadın_i

   protect-REL-3SG woman

   ‘The woman_i that Ahmet fired e_i because he hates e_iand Ali protects e_i because he loves ec_i.’

In both conjuncts of (3), the shared material kadın ‘woman’ is interpreted inside the complement positions of the verbs in adjunct clauses. It is also interpreted as the complement positions of the intermediate verb in each conjunct, kov- ‘fire’ in Conjunct 1 and kola- ‘protect’ in Conjunct 2. Note that the structure is grammatical despite the “case mismatch” i.e. the shared constituent requires the ablative case in the first conjunct and the accusative case in the second.

The interesting point observed in these examples is related to the fact that the positions inside the conjuncts are not required to be the same in terms of case, i.e. relativization can target a position with the accusative case in one conjunct but the ablative in the other. In the rest of the paper, we discuss the implications of this fact on the analyses of relative clause formation in Turkish^[5],^[6].

2. Turkish ATB relativization and relative clause formation

As we have noted above, Turkish ATB relativization does not require positions inside the conjuncts be the same in terms of case. What does this imply for an analysis of relative clause formation which argues that the head noun is originated inside the relative clause and raises to Spec-CP? Consider (4) which exemplifies an ATB relativization.

1. _[DP[TP Ben-im ec_inefret et-tiğ-im] _[TP sen-in ec_i kız-dığ-ın]_k

   that.I.hate ec-ABL that.you.are.angry.at ec-DAT

   [D⁰_{[CP [NP} kadıni] [C′ ec_k]]]] woman

   ‘The woman that I hate and you are angry at.’

According to the head raising analysis, (4) involves two steps of movement: (i) the head noun moves from the complement position of the relativized verbs into the Spec-CP position, and (ii) the whole TP moves into the Spec-DP position in order to yield prenominal modification (Kornfilt 2000). The relevant tree representation is given in (5) below.

(5)

The problem with (5) is that the head noun originates inside the relative clause conjuncts, i.e. a single item kadın ‘woman’ merges with the two verbs with different case requirements. This means that the shared element in an ATB relativization checks ablative case in the first conjunct, but dative case in the second. Given that a single item cannot check different case features of different verb heads, the structure above should be ungrammatical. However, the structure in (4) where the positions inside the two conjuncts are not the same in terms of case is perfectly grammatical. This makes the head raising analysis of relative clause formation is problematic for Turkish ATB relativization. Now, let us move into the empty operator movement analysis of Chomsky (1977) and see how it works with ATB relativizations.

According to the empty operator movement analysis, an empty operator which is merged in the relativized position moves to the Spec-CP in relative clauses. The head noun which is interpreted inside both conjuncts is base generated in its surface position and gets its case from the matrix clause (6).

1. _{[NP [CP} OP_i[TP Ahmet-in [ec_i] nefret et-tiğ-i]

   that.Ahmet.hates ec-ABL _[TP sen-in [ec_i] kız-dığ-ın]] kadın_i] that.you.are.angry.at ec-DAT woman

   ‘The woman that Ahmet hates and you are angry at.’

According to the operator movement analysis, an empty operator merges with the verbs inside the conjuncts and satisfies the case requirements of both verbs simultaneously^[7] Then, OP moves to the Spec-CP position. Following Aoun and Li (2003) who states that the operator moves to the Spec-CP position in order to take the whole clause under its scope, we assume that OP moves to the Spec-CP position in order to check some feature on the C head. The representation of (6) is given in (7) below.

(7)

We propose that the grammaticality of “case mismatch” follows from the idea that case requirements of the head noun are satisfied not inside the relative clause but in the matrix clause. Since what moves is not the head noun itself but an empty operator, it is not surprising that the positions inside the conjuncts are not required to be the same in terms of morphological case.

A question that naturally follows from the discussion above is what if a resumptive pronoun appears in the position of the gap inside the conjuncts? Since the resumptive pronoun which is interpreted with the head noun in a relative clause has case requirements, its behavior in ATB relativization becomes crucial for the discussion here. If ATB relativization favors the empty operator movement analysis, resumptive copies of the head noun inside the conjuncts are predicted to be ungrammatical. The example in (8) below shows that this prediction is borne out.

1. *[_Conj.1 Ahmet-in kendisin-den_inefret et-tiğ-i] Ahmet-GEN herself-ABL hate-REL-3SG [_Conj.2 Ali-nin kendisin-i_i sev-diğ-i] kadın_i Ali-GEN herself-ACC love-REL-3SG woman

   ‘*The woman_i that Ahmet hates her_i and Ali loves her_i.’

The example in (8) above is ungrammatical if the head noun kadın ‘woman’ is interpreted inside both conjuncts, i.e. there is one single woman that Ahmet hates and Ali loves. This indicates that head noun is not originated inside the relative clause where it undergoes case checking operation. Now, let us consider a further example where a resumptive pronoun appears only in the first conjunct but not in the second.

1. [_Conj.1 Ahmet-in kendisin-den_i nefret et-tiğ-i] Ahmet-GEN herself-ABL hate-REL-3SG [_Conj.2 Ali-nin ec-ACC_isev-diğ-i] kadın_i Ali-GEN love-REL-3SG woman

   ‘*The woman_i that Ahmet hates her_i and Ali loves ec_i.’

The example where an optional resumptive pronoun appears in the first conjunct but not in the second is grammatical. The grammaticality of this example further shows that the head noun is not originated inside the conjuncts. If it were, the example should be ungrammatical due to “case mismatch” problem^[8]

3. Conclusion

The paper concludes that Turkish ATB relativization presents evidence against the head raising analysis of relative clause formation given that the positions inside the conjuncts are not required to be the same in terms of case. The grammaticality of such cases speaks for an analysis where the head noun is base generated in its position where it gets its case from the matrix syntactic environment.