This book is probably the first published doctoral thesis on Turkish syntax since Erguvanlı (Reference Kornfilt and Bengisu1984), and is thus very valuable for Turkish studies, for linguistic typology, and for theoretical syntax.
The focus of investigation includes a rather large area of Turkish syntax, given that the author addresses questions such as whether nominal phrases are Noun Phrases (NPs) or Determiner Phrases (DPs), whether the Turkish clause has a functional projection vP, and what the architecture of the clause is in Turkish. Some of the problems tackled are issues of long standing – for example, the fact that direct objects that are not referential cannot exhibit accusative case marking, and that similarly, subjects of embedded nominalized clauses reject the otherwise obligatory genitive case marking when they are non-referential. (This latter observation can be found in the literature much less often.)
The aims of the book in terms of coverage are thus very impressive. The question is whether the proposed edifice is quite as solid and well-founded in its entirety as claimed. My answer will be that it is not – but this is not necessarily a negative assessment; as a doctoral thesis, this is a very solid achievement, and my main criticism is directed towards the editor(s), who should have insisted on more thorough revisions of the dissertation than have been made.
The first chapter, the ‘Introduction’, proposes two conditions for argumenthood: (i) visibility via case marking and theta-role assignment, and (ii) the assignment of referentiality. The claim is that these conditions are not independent from each other but intimately connected, and that argumenthood and referentiality are assigned in the same domain. Configurationality is defined here as the availability of case-driven Agree. Whether a language is configurational or not is claimed to result from the interaction of case and referentiality.
The second chapter, ‘Referentiality in Turkish’, proposes that in Turkish, case interacts directly (i.e. not via any functional categories) with referentiality, claiming that Turkish has no evidence for a DP projection because it has no determiners. Nominal phrases in Turkish are argued to be NPs, which are predicates that can be turned into arguments by the case marker, which is thus viewed as a type shifter. The case marker is also a referentiality marker. Case and referentiality are both assigned (or licensed) by the same functional head. This holds for Turkish and other typologically similar languages, in contrast to languages where case-checking takes place in the domains of vP and Tense Phrase (TP), and where determiners express referentiality. In such languages, e.g. English, the conditions for case and referentiality are independent from each other.
The same chapter also addresses nouns devoid of overt case marking, a topic familiar in traditional, typological, and generative studies of Turkish. The author proposes that, in Turkish, case assignment cannot be carried out in the complement position of verbs, but only in higher positions where the NP is type-shifted into an argument, and that case and referentiality are assigned to that argument at the same time, as mentioned earlier. If the NP is merged in a lower position, such as the complement position, the NP cannot be type-shifted and instead forms a complex predicate with the verb. Complex predicate formation is handled via Massam's (Reference Massam2001) proposed pseudo-incorporation. This account, designed for direct objects by Massam, is claimed here to extend to subjects. This chapter also has an appendix on indefinites and case, which is further divided into a section on non-specific indefinites and another on specific indefinites.
The third chapter is entitled ‘Case, referentiality and non-configurationality’. The author characterizes her approach to clause structure as Neo-Davidsonian. She proposes that the positions in which NPs can undergo typeshifting to arguments are those in the functional clausal domain above the Verb Phrase (VP), where both case and referentiality are assigned. She also suggests that Turkish clauses have no vP layer, and that the TP-level does not interact with case licensing, either. Consequently, there is no case-driven Agree with the heads of the functional projections of vP and TP in Turkish, nor in Japanese and Hungarian; all arguments are equidistant with respect to the verb. This results, under the proposals made, in a non-configurational phrase structure. Öztürk further claims that non-configurational languages like Turkish, Japanese and Hungarian share typological properties such as pseudo-incorporation, scrambling, argument-drop, and lack of superiority effects.
In contrast, configurational languages whose referential nominal phrases are DPs exhibit case-driven Agree with the heads of vP and TP for case purposes, as in English. Such languages do not have free word order, argument-drop, or pseudo-incorporation, but they do exhibit superiority effects. Chinese is a third type of language which combines some features of each of the previous two types: on the one hand, it exhibits neither free word order nor pseudo-incorporation and is thus like English; on the other hand, it has argument-drop and lacks superiority effects and is thus like Turkish. Chapter 3 also has an appendix on case-driven Agree and language acquisition.
The fourth chapter, entitled ‘Concluding remarks’, is a summary of the results of chapters 2 and 3.
I now turn to a discussion of some specific proposals.
Arguments against DP inTurkish. The arguments against a DP in Turkish are central to the book. They are based largely on the claim that Turkish does not have determiners. However, Turkish does have demonstratives, and it has an indefinite article. The author denies to the morpheme bir ‘one, a’ the status of an indefinite determiner. Her main argument is based on claims made in Crisma (Reference Crisma1997) and Longobardi (Reference Longobardi, Mark and Chris2001) that if a language has only one article, it will be a definite rather than an indefinite article. Since Turkish has no definite article, the author claims that bir cannot be an indefinite determiner. This reasoning is circular. Turkish is dismissed as a language having an indefinite, but no definite, article, based on cross-linguistic statistics. But in the very same statistics on which the author bases her argument that bir is not an indefinite article, Turkish could not appear as a counterexample, given the author's classification of bir. In addition, there are good reasons for characterizing bir as an indefinite article, because its syntactic properties as an article differ clearly from those when it is a numeral; such distinctions are to be found in any good grammar of Turkish and include differences in placement with respect to adjectives and relative clauses.
The author is right in stating that in a head-final language such as Turkish, determiners would be expected to follow the head noun if they occupy D, rather than precede the noun, as they do in Turkish. But this argues only against an analysis of these determiners as D, and is fully compatible with an analysis in which the determiners occupy the specifier position of DP.
Furthermore, the author's arguments against a DP-analysis for Turkish nominal phrases, with an abstract D-head, are invalid and inconsistent. Öztürk dismisses such an analysis, claiming that head-movement in a putative head-final DP cannot be diagnosed. However, exactly such head-movements to D have been proposed, e.g. in von Heusinger & Kornfilt (Reference Heusinger and Kornfilt2005), for partitive phrases and other nominal phrases in Turkish. While those analyses may be faulty, they do deserve discussion before the analysis is discarded. Moreover, the author herself assumes head-movement in clauses, i.e. raising of V to higher functional heads such as T – an assumption that is crucial for the book, as it offers motivation for the author's claim that all arguments in Turkish are equidistant to the predicate. Given that clauses are head-final in Turkish, it seems to me to be an inconsistent position to assume that head-movement can be diagnosed in clauses but not in the DP domain.
Relevance of TP inTurkish. The author assumes that Turkish has a TP, with T housing tense features, but that T (and its specifier position) are not involved with subject case, and that therefore there is no obligatory movement to the specifier of TP (SpecTP). Contrary to this claim, I would argue that firstly, there are instances with obligatory movement, motivated by case, to (a matrix) SpecTP, and that secondly, even in the absence of such movement, T (+Agr) is important in the licensing of case.
The first point can be argued with respect to Subject Raising, considered by the author to support her claim that movement to SpecTP is an optional phenomenon. However, the following pair of examples shows that there are instances where movement to SpecTP is obligatory.
(1)
When the embedded clause lacks person and number agreement marking for the local subject, raising of that subject to matrix subject position is obligatory, as demonstrated by the ungrammaticality of (1b). This shows that raising to SpecTP can be obligatory, and also that T (+Agr) is involved in case licensing.
The second point is that even in the absence of movements to specifier positions of functional heads, case licensing could be performed within a hierarchically structured clausal architecture, by functional heads such as T(+Agr), which can function as probes. Agree via probes has been widely used to account for a number of phenomena.
Problems with case as the direct licenser of referentiality. The first problem is linked to the unclarity of the notion of case used here, that is, whether it is a purely syntactic, abstract notion or a morphological one. When discussing the relationship between case and referentiality, the author appears to take morphological case to directly reflect syntactic case as the factor assigning referentiality to NPs. This leads us to expect that all nominal phrases with overt case should be referential. But this is so only for accusative and genitive, or, to use the terminology of Kornfilt (Reference Kornfilt1984), in the two ‘structural’ cases. In oblique cases, there is no relationship to referentiality. Thus, in examples such as (2), the dative object can be generic or non-referential.
(2)
The author mentions that generic phrases can be considered referential (following Carlson Reference Carlson1977). However, note that a generic interpretation is also possible for bare direct objects of transitive verbs, as seen in (3).
(3)
The direct object, which is missing accusative case, is interpreted generically, just as the dative object in (2) is. Thus, if a generic interpretation is considered to be referential, then the direct object in (3) would have to be referential, too. But since it lacks accusative marking, the impact of case on the referential interpretation, claimed to be so central and direct in this book, would be negated completely. Accordingly, Öztürk's insistence on a direct link between morphological case-marking and referentiality leads to both empirical problems and inconsistencies.
Null arguments. In connection with the book's focus on the configurationality ‘parameter’ and its claim that Turkish has a non-configurational phrase structure, the author addresses the issue of null arguments, claiming that pro in Turkish and other non-configurational languages, in contrast to pro in configurational null-subject languages, does not need to be licensed by local overt agreement, as previously claimed in the literature. Two main arguments are offered. First, Öztürk argues that both empty subjects and empty objects in Turkish are pronouns and that the absence of object agreement in Turkish suggests that an object pro can be licensed without overt agreement. She concludes that therefore, no agreement should be necessary for an empty subject pro either. Secondly, she claims that the empty subjects of certain adjunct clauses without local agreement are pro; again, if such instances of pro are licensed without local agreement, then the overt agreement found with most other instances of pro is not needed for purposes of licensing and identification of pro. I am unconvinced by these arguments, and I would analyze empty objects in Turkish as silent variables, and the silent subjects of the adjunct clauses in question as instances of PRO; some argumentation and examples follow.
While the book argues against the possibility of silent object variables (although not convincingly in my opinion; see below), the possibility of PRO as a subject of certain adjunct clauses is not addressed, cf. (4).
(4)
The silent subject of the adjunct clause in (4) can refer either to the matrix subject with its 1.SG features, or to the comitative phrase with its 3.SG features. If we analyze this silent subject as a pragmatically controlled PRO (cf. Kornfilt Reference Kornfilt, Uwe and Luka2003), these facts are accounted for. Note that this silent subject does not impose any inherent phi-feature combination, as made clear by the translation, and as opposed to clear-cut instances of pro, i.e. the empty subject licensed by local overt agreement. (For additional arguments distinguishing pro and PRO in Turkish, see Kornfilt Reference Kornfilt and Bengisu1996, where further arguments in favor of a configurational phrase structure in Turkish are presented, including effects of the Binding Theory conditions B and C.)
Turning to silent objects, the author claims that the ungrammaticality of (5) (Öztürk's example (211), 214) under a co-referential reading between the empty object and the subject is due to a violation of the Binding Theory condition B.
(5)
If condition B is violated, the empty category is pro, according to the author. However, the ill-formedness of the relevant reading could also be due to a condition C violation, if the silent object is a silent variable, bound by an abstract topicalization operator. There is evidence for the latter analysis, if examples are considered where condition B is irrelevant, and condition C explains ill-formedness:
(6)
The silent object in the embedded clause in (6) cannot be co-indexed with Mary, even though this would not give rise to a condition B violation, as Mary would not be a local binder of the silent object, not being its clause-mate. However, condition C would indeed be violated if the silent object is a variable, given that elements sensitive to condition C must be free everywhere, not just in their local binding domain.
Uncontroversial instances of pro, as in a silent embedded subject, are well-formed in similar contexts, as seen in (7).
(7)
The silent embedded subject is preferably interpreted as co-indexed with Mary (but can also be anteceded by a discourse referent). Since the silent object obviously has different syntactic properties than the silent subject, I thus continue to claim that the silent subject in (7), where it is licensed by local agreement, is an instance of pro, while the silent object in (6) is a silent variable. Therefore, the author's claim that local agreement is not the licenser of silent subjects in Turkish does not go through, and one of the bases for her proposed characterization of Turkish as a non-configurational language thus is lost as well.
These points of criticism should be viewed as positive observations about this book. They illustrate the wide coverage of the book, and show how strongly it can motivate further thought, observation, and explanation. It is a very welcome addition to generative studies of Turkish syntax, and interesting reading for any syntactician.
