Printer Friendly

Syntax of information structure in Turkish.

Abstract

Three sentential positions have been ascribed a special status in the focus-background articulation of Turkish sentences. It has been argued that the sentence-initial, postverbal and immediately preverbal slots are three syntactic positions that are, respectively, allocated for topic, background and focus marking in Turkish. The analysis offered in this paper starts with a close examination of each of these positions in terms of the informational functions they are supposed to perform. It is shown that such a position-function mapping gives a noncomplete and noncorrect description of information structure in Turkish. The major claim made by this study is that Turkish does not employ any syntactic strategy to mark the informational status of a sentence element; but some sentence elements may undergo a syntactic operation of detachment to "clause-external" positions to satisfy a certain informational requirement of the sentence itself.

1. Introduction

The canonical/basic word order in Turkish is usually considered to be SOV (cf. Erguvanli 1984; Kural 1992, 1994, 1997; Kilicaslan 1994, 1998; Hoffman 1995; Kornfilt 1997; inter alia). (1) However, Turkish is a very flexible language in terms of word order variations. Even a simple sentence like the one in (1) may have six different permutations:
(1) a. Oya Kaya-yla evlen-di.
 Oya Kaya-com marry-pst
 'Oya married Kaya.'

 b. Oya evlen-di Kaya-yla.

 c. Kaya-yla Oya evlen-di.

 d. Kaya-yla evlen-di Oya.

 e. Evlen-di Oya Kaya-yla.

 f. Evlen-di Kaya-yla Oya.


Each of these is a perfectly grammatical sentence.

This flexibility, as Erguvanli (1984) suggests, seems to come from the fact that in Turkish, constituents are morphologically marked to signal their grammatical roles. Since the grammatical role of a constituent generally does not rely on its sentential position, each constituent gets a status relatively independent of its sentential context. Another common assumption is that the existence of so many permutations is not just the result of a stylistic variation but it has an interpretive value. Thus, the scrambling of constituents is not entirely free but constrained by some mechanisms affecting the interpretation. A straightforward observation is that word order variation (i.e. deviation from the SOV order) correlates with the information-structure dimension of sentence interpretation in Turkish. Three sentence positions have been assigned a special status in the formulation of this correlation, each being associated with a particular information-structure function (cf. Erguvanli 1984; Erku 1984; Kilicaslan 1994; Kornfilt 1997).
(2) POSITION: FUNCTION:
 Immediately preverbal Focus
 Postverbal Background
 Sentence-initial Topic


In this study, information structuring is taken to be a pragmatic phenomenon whereby the (semantic) content of a sentence is given a structure that is congruent with the context of use. (2) As will be exemplified throughout this article, different permutations of a sentence (such as those in [1]) may fit in different types of context. Focus, background, and topic are three informational primitives which are used in the information-structural analysis of the sentence. Focus denotes the portion of the sentence that encodes a piece of new information that bears a high degree of relevance to the discourse context. Background is used here as a mere complement of focus: all that is not in focus. Topic is a distinguished background element that refers to a discourse entity which the sentence is about.

In this article, I initially call the position-function mapping in (2) into question. Such a mapping, I argue, is problematic at least in the following respect:

(3) THE PROBLEM

A syntactic account mapping the topic, background, and focus functions onto the sentence-initial, postverbal, and immediately preverbal positions would fad to give a complete picture of information structure in Turkish, as it would be far from being congruent with many empirical and theoretical observations.

A mapping like (2) could be an appropriate formulation of a statistical tendency but not that of a structural necessity. Section 2 elaborates on this point. That is, it gives a specification of the problem by showing that no relation of syntactic dependence can be established between the information-structure functions and sentential positions in question. Section 3 comprises the actual contribution aimed by the article: a solution to the specified problem. The main objective in this section is to come up with a correct and complete characterization of the syntactic articulation of information structure in Turkish. This characterization rests on another proposal put forward in this section: a syntactic analysis of Turkish sentences which distinguishes between their "core" and "peripheries." The major proposal to be offered at the end of all this argumentation can be stated in most general terms as follows:

(4) THE PROPOSAL

Backgrounding via detachment to the peripheries is the only syntactic operation used in the informational structuring of Turkish sentences.

Let us now start with the justification of the problem.

2. Problem: a noncomplete description

2.1. Focus and the immediately preverbal position

There seems to be a statistical correlation between the immediately preverbal position and the surface realization of foci in Turkish. Turkish foci appear to come just before the verb in most cases. In fact, this position is generally considered to be the focus position in this language (Ahmet Cevat 1931; Erku 1983; Erguvanli 1984; Kural 1992, 1994, 1997; Hoffman 1995; Kennelly 1997). This section argues against characterizing the correlation between the immediately preverbal position and focus in Turkish in terms of syntactic necessity. It presents data and arguments disfavoring an approach to focus realization in Turkish that would postulate a relation of syntactic dependence between focus and the immediately preverbal position.

2.1.1. The immediately preverbal position is not the only position for focus. Many linguists have pointed out that the immediately preverbal position is not the only place where foci can appear in Turkish (cf. Gencan 1979; Goksel 1998; Goksel and Ozsoy 2000; Kilicaslan 1998). The example in (5) illustrates this fact. (Henceforth, the focal accent will be indicated by small caps and the whole of the focal segment will be delimited by F-labelled brackets.)
(5) Kaya-yla KIM evlen-di?
 Kaya-com who marry-pst
 'Who married Kaya?'

 a. Kaya-yla [F OYA] evlen-di.
 Kaya-com Oya marry-pst
 '[F OYA] married Kaya.'

 b. [F OYA] Kaya-yla evlen-di.


Throughout this study, the focus of a sentence will be identified by exploiting an idea originally proposed by work in the Prague tradition (cf. Sgall et al. 1986):

(6) FOCUS TEST

Given an appropriate question, the focus of a (declarative) sentence would be that part of it providing the answer.

According to this test, Oya serves as the focus of the response utterance in (5). Besides in the given context, either arrangement of that response utterance is acceptable, even though the one in (a) sounds more natural. That is, the focused NP felicitously appears in a non-immediately-preverbal position in (b).

Moreover, not only can foci appear in a slot not just before the verb, but sometimes they are obliged to do that. One such case is that of multiple-focus constructions. If foci were confined to the immediately preverbal position, one would naturally think that Turkish phrase structure allows only one constituent to be focused at a time. As the following example shows, this is not the case in Turkish:
(7) KIM KIM-LE evlen-di?
 who who-com marry-pst
 'Who married who?'
 [F OYA] [F KAYA-YLA] evlen-di.
 OYA Kaya-com marry-pst
 '[F OYA] married [F KAYA].'


In the given context, two separate constituents are felicitously interpreted and marked as focal. Obviously, in such cases it is obligatory for at least one focal constituent to occupy a position other than the immediately preverbal one.

It is also possible to have cases where a focal constituent cannot be placed in the immediately preverbal position in any possible arrangement of the sentence elements. These are the cases where the sentence element occupying the immediately preverbal position cannot be moved away from there, irrespective of whether it is focal or not. Consider the example below:
(8) Bahce-de KIM bir kopek gor-du?
 garden-loc who one dog see-pst
 'Who saw a dog in the garden?'

 a. Bahce-de [F OYA] bir kopek gor-du.
 Garden-loc Oya one dog see-pst
 '[F OYA] saw a dog in the garden.'

 b. *Bahce-de bir kopek [F OYA] gor-du.

 c. *Bir kopek bahce-de [F OYA] gor-du.

 d. *Bahce-de [F OYA] gor-du bir kopek.


In Turkish, an object NP not carrying case morphology (and not followed by a particle like dA 'also') is restricted to the position just before the verb. Its placement elsewhere would result in ungrammaticality. In the example above, bir kopek 'a dog' is such an object NP and, hence, necessarily intervenes between the verb and the focal NP Oya. For this reason, the (b), (e), and (d) sentences will be unacceptable in any possible discourse context.

In fact, there seems to be a way to treat this last type of cases as predictable in terms of the position of focus. In Turkish, direct objects without case morphology receive an indefinite nonspecific interpretation (cf. Erguvanli 1984; Kornfilt 1984, 1997; Enc 1991) and NPs having such interpretation tend to be immediately preverbal. Moreover, direct objects lacking case morphology appear to be restricted to this position also due to a requirement of case assignment: such objects receive structural case from the verb under adjacency (cf. Kornfilt 1984). Resting on these facts, one could take the verb and its case-morphemeless object to form a verbal complex. This step would allow for a looser definition of the immediately preverbal position such that it includes the immediate left of such a complex. Hence, utterances like (8a) would be predicted to be fine for a proposal confining the focus to the immediately preverbal position, as the focused constituent immediately precedes a verbal complex. However, it should be noted that weakening the association between the focus and the verb in this way would still be far from capturing the flexibility which Turkish loci display in terms of their sentential positioning. Foci separated from the verb by definite and/or specific NPs (e.g. examples [5] and [7]) would fail to be characterized as immediately preverbal.

As a further case where focus is not immediately preverbal, take the following example:
(9) Oya Kaya hakkinda NE hissed-iyor?
 Oya Kaya about what feel-prog
 'What does Oya feel about Kaya?'
 Oya Kaya-yi [F SEV-IYOR].
 Oya Kaya-acc love-prog
 'Oya [F LOVES] Kaya.'


It is the verb itself that is focused here. It is self evident that the immediately preverbal position will be occupied by a nonfocal constituent in such cases.

The evidence discussed up to this point shows that it is not possible to posit an association between the focus and the verb in Turkish as strong as one that requires the former to always appear just before the latter. Even if this association is weakened to an extent that allows for a treatment of foci preceding nonspecific NPs as immediately preverbal, many cases of focus realization would still remain to be captured. An interesting line of discussion to pursue after this point might be to call into question even a weaker form of this association: is there a type of focus that is confined to the immediately preverbal position for syntactic reasons? This question will be the subject matter of the ensuing discussion.

2.1.2. The immediately preverbal position serves as a syntactic marker for no type of focus. In fact, the observations presented above do not point to newly discovered facts. The ability of Turkish loci to appear somewhere other than the immediately preverbal position has been noticed by many other researchers. However, not all researchers take this to be a sign of the lack of a dependency relation between that position and the focus phenomenon. Some linguists call attention to the distinction between presentational focus and contrastive focus (3) and argue that only contrastive foci are free to move in the preverbal area, while presentational loci are still confined to the immediately preverbal slot (cf. Kural 1994, 1997; Kornfilt 1997; Issever 2001). However, this latter view, too, appears to be untenable under the light of some empirical and theoretical observations. Let us start with the empirical problems. Consider the following two examples from Kornfilt (1997: examples [693] and [694]):
(10) Kitab-i Ali-ye HASAN ver-di.
 Kitab-acc Ali-dat Hasan give-pst
 'HASAN gave the book to Ali.'

(11) Kitab-i HASAN Ali-ye ver-di (... Mehmet degil).
 Kitab-acc Hasan Ali-dat give-pst Mehmet neg
 'HASAN gave the book to Ali (and not Mehmet).'


In both (10) and (11), the subject carries stress and high pitch as a prosodic sign of its focal status. Kornfilt holds the view that the subject in (10) can be interpreted as either contrastive or noncontrastive, while the subject in (11) is restricted to a contrastive interpretation. These judgments directly follow from her characterization of the difference between noncontrastive focus and contrastive focus. She argues that the former "is marked, in addition to stress and pitch, also by placement of the emphasized constituent in immediate pre-verbal position," whereas the latter "can be marked by that same additional pre-verbal placement, as well, but does not need to be" (Kornflit 1997: 190). Admittedly, when interpreted out of the blue, these examples display the contrast indicated by Kornfilt most readily. That is, the most natural interpretation for (11) appears to be the one where the subject receives a contrastive reading. However, the same sentence could also be given an interpretation with no contrastive flavor with a context spelled out appropriately. Suppose that A has previously told B many times that it was Hasan that gave the book to Ali and that B has totally forgotten about it again. Being inquired about the same question, B gets irritated and responds angrily:
(12) Kitab-i HASAN Ali-ye ver-di (... Bunu daha once bircok
 Kitab-acc Hasan Ali-dat give-pst this more before many
 kez soyle-dim).
 times say-pst
 'HASAN gave the book to Ali (I've already said that many times).'


Prior to the utterance of this response, B need not have in mind a closed set of individuals of which Hasan would constitute a contrasting complement subset. (4) This rather scornful utterance would serve to reactivate a piece of already acquired information but not to correct a piece of false information or pick out one of a set of alternative pieces of information. One fairly clear-cut difference between (10) and (11) (or [12]), which is also pointed out by Kornfilt (1997), is that the accent and stress falling on the focus of the latter is stronger than that falling on the focus of the former. In my view, a focal constituent is licensed to appear to the left of the immediately preverbal slot provided that its marked focal accent and stress (which is stronger than the unmarked one) can be attributed to a certain contextual factor, such as contrastive or emotional emphasis.

In fact, shifting a focal constituent towards the beginning of the sentence seems easier when this constituent is adjunctive. Consider the following question-answer pair:
(13) Hasan kitab-i NICIN Ali-ye ver-di?
 Hasan Kitab-acc why Ali-dat give-pst
 'Why did Hasan give the book to Ali?'
 Hasan kitab-i [F SEV-ME-DIG-I icin] Ali-ye ver-di.
 Hasan kitab-acc like-neg-fnom-3sg because Ali-dat give-pst
 'Hasan gave the book to Ali because he didn't like it.'


Here, the focus, which is not immediately preverbal, need not necessarily be interpreted as contrastive. The person asking the question might be entirely ignorant of possible reasons behind Hasan's act in question. That is, the set of such possible reasons might be open for him/her. Hence, the focus in the response might be providing entirely new information without any implication of contrast.

Theoretical considerations resting on crosslinguistic observations, too, seem to disfavor a view assigning a focus-marking status to the immediately preverbal position of a Turkish sentence. One straightforward assumption underlying such a view could be to take this position to be a derived position with a [+focus] feature where presentational foci have to move to receive that feature. Hungarian, for example, is argued to have such an overt focus position, which is also immediately preverbal, to store moved foci (cf. Kiss 1981, 1995; Horvath 1986; Puskas 1992; inter alia). However, Kiss (1995) points out that it is not presentational foci but [+exhaustive] foci, which presuppose a context with a dosed set of individuals as alternative focus values, that move to that structurally-defined focus position in Hungarian. Furthermore, considering the relevant observations made on many other languages (such as Rumanian, Greek, Catalan, Italian, Finnish, and Arabic), Kiss arrives at the following generalization (among others) that applies across languages:

(14) A CROSSLINGUISTIC OBSERVATION

Presentational focus remains in situ, while nonpresentational focus (defined with respect to a closed set) can move into a specifier position of a functional projection (Kiss 1995).

Therefore, postulating for Turkish a special position which presentational foci move to would conflict with this generalization.

Vallduvi and Engdahl (1996) point out another problem with describing Turkish as a language that resembles Hungarian in the way it realizes focus, that is, by means of a movement operation into a structurally-defined position. They note that Hungarian does not allow a leftward focus projection in the immediately preverbal focus position but focus in Turkish projects to the left in many cases. The following example taken from Vallduvi and Engdahl (1996: example [73]) illustrates this phenomenon in Turkish, where the question contextualizations in (15a)-(15d) highlight the intended focus readings:
(15) Bir hizmetci [F yemek-ten once [F masa-nin uzer-i-ne [F [F
 one servant meal-abl before table-gen3 top-poss3-dat
 NOT-U] birak-ti.]]]
 note-acc leave-pst
 'A servant put the note on the table before lunch.'

 a. What did a servant leave on the table before lunch?

 b. What did a servant do before lunch having to do with the
 table?

 c. What did a servant do before lunch?

 d. What did a servant do?


Vallduvi and Engdahl (1996) note that this flexibility contrasts radically with focus projection in Hungarian. In their view, if focus were syntactically realized in Turkish by movement to the immediately preverbal slot, we would expect this language to behave like Hungarian by not allowing projection of focus realized in this way. Resting on this observation, they suggest that Turkish is closer to languages like Catalan than Hungarian in that focal elements stay in situ, while background elements can be removed from within the focus domain. (5)

In effect, rejecting the possibility of presentational foci to move to the immediately preverbal slot need not necessarily lead one to also reject the focus-marking status of this position in Turkish. Kural (1992), for instance, maintains both that a constituent in a Turkish sentence has to appear in the immediately preverbal slot in order to be marked for the [+focus] feature (6) and that only nonfocal constituents can scramble in that language whereas focal ones stay in situ. He aptly points out that requiring both focal and nonfocal constituents to move to certain positions would create redundancy in the representation of focus in a theory that employs a structurally-defined unique focus position:

First, the focused constituent would move to some derived position, and then all defocused constituents would be forced out of the VP. Either mechanism would be sufficient to represent focus at S-structure; having both of them would be redundant (Kural 1992: 52).

In Kural's account, the required preverbal occurrence of the focus is ensured by forcing all defocused constituents to leave their VP-internal positions. This would obviously have the effect of removing all other materials between the focus and the verb. Avoiding redundant movements in this way implies some sort of assumption that linguistic structures conform to some "economy" criterion demanding that there be no extra operations in their derivations. Such ontological minimalism is the basic tenet of the minimalist program, which maintains that "there are no extra steps in derivations, no extra symbols in representations, and no representations beyond those that are conceptually necessary" (Lasnik 2002: 432).

Thus, it is worth having a look at focus positioning in Turkish also from a minimalist perspective. Kural's approach seems a good move towards a minimalist account of the focus phenomenon in Turkish. A direct conclusion that follows from the economy criterion of the minimalist program is that an account of a phenomenon postulating a smaller number of operations is preferred to one based on a larger set of operations. Thus, Kural's mechanism for bringing the focus just before the verb appears favorable to one that would require both focused and defocused constituents to move to their designated positions. Both would have the same effect in terms of the sequential positioning of the constituents, but the latter would do that with more operations.

However, the economy criterion of the minimalist program cannot be reduced to a criterion solely formulated in terms of the number of operations. The minimalist program, at least the particular version of it adopted here (cf. Chomsky 1993, 1995; Weinberg 2001; Lasnik 2002; Uriagereka 2002), has a more elaborate characterization of that criterion. As Weinberg (2001) puts it:

The two most salient features of [the Minimalist] system are its derivational character and the role that economy conditions play in regulating possible derived structures. At least at the level of competence, the model has moved away from the overgeneration-and-filtering character of its Government-Binding precursor. Structures that do not satisfy the economy conditions are simply not generated (Weinberg 2001: 285).

Those conditions that rule out overgeneration are the following:

LAST RESORT:

a) Operations do not apply unless required to satisfy a constraint.

b) The minimal number of operations is applied to satisfy the constraint (Weinberg 2001: 285).

GREED:

The operation cannot apply to a to enable some different element [beta] to satisfy its properties ... Benefiting other elements is not allowed (Chomsky 1995: 201).

Obviously, avoiding redundant movements in Kural's way is entirely in line with the b) clause of the last resort condition. However, in an account where focused constituents are constrained to be in the immediately preverbal position and that constraint is satisfied by moving defocused constituents out of the VP, the greed condition would be violated as defocused constituents would benefit focused ones in the satisfaction of that constraint. Thus, Kural's model does not appear to be compatible with the greed condition.

In brief, any attempt to postulate a relation of syntactic dependence between the immediately preverbal position and focus in Turkish appears to run into serious empirical and theoretical problems. We now turn to the role of that postverbal position in the marking process of nonfocused or backgrounded constituents.

2.2. Background and the postverbal position

2.2.1. Only backgrounded constituents can appear in the postverbal position. The postverbal position is another location that has been assigned a special status in the informational articulation of Turkish sentences. One common observation about the informational status of this position is that it can host only background elements (Kural 1992, 1994, 1997; Erguvanli 1984; Hoffman 1995; Kornfilt 1997; Goksel 1998; Kilicaslan 1998; Goksel and Ozsoy 2000; Kornfilt 2000). The following examples illustrate this fact:
(16) Kaya-yla KIM evlen-di?
 Kaya-com who marry-pst
 'Who married Kaya?'

 a. [F OYA] evlen-di Kaya-yla.
 Oya marry-pst Kaya-yla
 '[F OYA] married Kaya.'

 b. *Kaya-yla evlendi [F Oya].

 c. *Evlendi [F Oya] Kaya-yla.

 d. *Evlendi Kaya-yla [F Oya].

(17) Oya Kaya-yla NE zaman evlen-di?
 Oya Kaya-com what when marry-pst
 'When did Oya marry Kaya?'

 a. [F IKI yil once] evlen-di Oya Kaya-yla.
 two years before marry-pst Oya Kaya-com
 'Oya married Kaya [F TWO years ago].'

 b. *Evlen-di Oya Kaya-yla [F IKI yil once].

 c. *Evlen-di Oya [F IKI yil once] Kaya-yla.

 d. *Evlen-di [F IKI yil once] Oya Kaya-yla.


As background is taken to be a mere complement of focus in this study, the focus test suffices to identify the background of the sentence. Once the focus has been identified, the background will be, by definition, all that is not in focus. The examples above show that only backgrounded constituents (i.e. those not focused) can appear in the postverbal position. Besides, as exemplified by (17), this position can host more than one constituent.

Researchers observing the backgrounding status of the postverbal position also note that a constituent occupying this position cannot be a stressed and/or questioned phrase:
(18) a. *Kaya-yla evlen-di OVA.

 b. *Kaya-yla evlen-di KIM?

 c. *Kaya-yla evlen-di kim?


Neither of these sentences can be felicitously uttered in any possible context, since their postverbal constituents are either stressed or questioned, or both.

The inability of a postverbal constituent to receive stress is entirely compatible with the exclusion of foci from postverbal slots: as already mentioned, foci must receive primary stress in Turkish sentences. As for questioned constituents, it should have been noted that in the Turkish questions used in the examples up to this point, it is the question word that receives the primary stress (which is indicated by their appearance in small caps). As will be illustrated with many other examples in what follows, this is a general fact about Turkish questions. This fact can be considered as strong evidence form Turkish in favor of the commonly held assumption that the focus of a question is its questioned constituent. From this point of view, there is nothing surprising about the exclusion of questioned constituents from the postverbal position of Turkish sentences. As such constituents are focal by nature, they are categorically barred from occurring in postverbal slots of Turkish sentences, where only backgrounded elements can appear.

2.2.2. The postverbal position is not the only place where backgrounded constituents can appear. The postverbal position serves as a store for backgrounded constituents; but it is not the only position that such constituents can occupy in Turkish sentences (cf. Hoffman 1995; Kilicaslan 1998; Goksel and Ozsoy 2000):
(19) Kaya-yla KIM evlen-di?
 Kaya-com who marry-pst
 'Who married Kaya?'

 a. Kaya-yla [F OYA] evlen-di.
 Kaya-com Oya marry-pst
 '[F OYA] married Kaya.'

 b. [F OYA] Kaya-yla evlendi.


This example shows that a backgrounded NP, namely Kaya-yla, can appear preverbally, either before the focus (cf. [19a]) or between the focus and the verb (cf. [19b]).

The association between the postverbal position and background is similar to the association between the immediately preverbal position and focus in that the designated position is not the only place where the informational function in question can be realized. However, this time the correlation is stronger in the following sense. Not all immediately preverbal elements are focal; but all postverbal ones are backgrounded.

2.3. Topic and the sentence-initial position

2.3.1. Topic as a distinguished background element. It has been pointed out by many researchers that an approach resting solely on the focus-background dichotomy will not be capable of capturing all aspects of the informational analysis of sentences (cf. Halliday 1967, 1985; Dahl 1974; Valimaa-Blum 1988; Vallduvi 1990, 1993; inter alia). All these researchers have maintained that a notion of topic is required for such an analysis to be carried out thoroughly. (7)

Even though topic has been given various characterizations, such as what links/anchors the sentence to the bearer's mental world or the previous discourse (cf. Reinhart 1982; Vallduvi 1990, 1993; Vallduvi and Engdahl 1996) or the point of departure which the rest of the sentence unfolds on or is predicated of (cf. Halliday 1967; Strawson 1971; Gundel 1974; Danes 1974; Reinhart 1982; Sasse 1987; Erteschik-Shir 1997; Peregrin 1995; Rosengren 1995), the most pivotal characteristic of such a sentence element seems to be the feeling of aboutness it induces (cf. Mathesius 1915; Hockett 1958; Strawson 1971; Gundel 1974; Dahl 1974; Reinhart 1982). As a working definition, I take topic to be a distinguished background element that refers to what the sentence is about. That is, the key concept in the characterization of topic that I adopt is that of aboutness. Thus, the test that I will use for topic identification will be based on that criterion of aboutness:
(20) TOPIC TEST

 Speaker A: Tell me about / what about X.
 Speaker B: ... X ... (X = TOPIC)
 (cf. Reinhart 1982; Vallduvi 1990, 1993)


According to the request made by A, B's utterance will be about X. Therefore, X will serve as the topic. (8)

2.3.2. The sentence-initial position is neither the only position for topics nor a position only for topics. Topics have been argued to be marked in Turkish by movement to sentence-initial position (Erguvanli 1984; Kural 1994, 1997; Hoffman 1995; Kornfilt 1997). Consider the following example from Kornfilt (1997: example [724]):
(21) Istakoz-u Hasan Ali-ye ver-di.
 lobster-acc Hasan Ali-dat give-pst
 '(Speaking of) the lobster, Hasan gave (it) to Ali.'


Kornfilt (1997: 200) argues that "in this example, the topic is istakoz-u 'the lobster-acc,' whose original position is preverbal; however, as a topic, the constituent has undergone movement to initial position."

Admittedly, the sentence-initial position is the most appropriate place for topics to appear in. However, as Kornfilt herself suggests by recognizing an optional side to topicalization by movement to sentence-initial position, this appropriateness requirement on topics should not be taken in absolute terms. First of all, the topic does not have to occur at the very beginning of the sentence. (9) That is, it may be preceded by other sentence elements, as shown by the example below:
(22) Istakoz-dan ne haber? O-na ne ol-du?
 lobster-acc what news it-dat what happen-pst
 'What about the lobster? What happened to it?'
 Hasan [T istakoz-u] [F ALI-YE ver-di].
 Hasan lobster-acc Ali-dat give-pst
 'Hasan gave the lobster to Ali.'


Istakoz-u 'the lobster-acc' follows Hasan in the response sentence, though it is the topic of the dialogue according to our topic test. One, of course, might question the reliability of the topic test and argue that Hasan is the new shifted topic of the discourse. However, if this were necessarily the case, the cohesiveness of the discourse would be disrupted when continued with the lobster still being the topic without mentioning Hasan at all. But, the following sentence could be a fine continuation to (22):
(23) Zaten kimse o-nu yemek iste-mi-yor-du.
 In.fact nobody it-acc eat want-neg-prog-pst
 'In fact, nobody wanted to eat it.'


The topic of the response sentence in (22) could also be preceded by two DPs which are supposed to have entirely nonspecific referents, instead of Hasan referring to a specific individual:
(24) Birkac gun once birisi [T istakoz-u] [F ALI-YE ver-di].
 several day before someone lobster-acc Ali-dat give-pst
 'Several days ago someone gave the lobster to Ali.'


A common observation about the semantics of topics is that they must receive strong readings (cf. Jager 1995; Choi 1995; Kilicaslan 1998). A weak phrase such as birkac gun once 'several days ago' or birisi 'someone' in the above example, cannot function as the topic of the utterance it occurs in. It would, indeed, be counterintuitive to take the utterance in (24) as being about a nonspecific point in time or a nonspecific individual, as long as these were not fixed at least in mind.

Crucially, the rightward extendibility of the position which the topic can be dislocated to does not seem to be boundless:
(25) Istakoz-dan ne haber? O-nu kim ye-di?
 Lobster-abl what news it-ace who eat-pst
 'What about the lobster? Who ate it?'

 a. [T Istakoz-u] birkac gun once [F ALI] ye-di.
 lobster-acc several day before Ali eat-pst
 'Ali ate the lobster several days ago.'

 b. Birkac gun once [T istakoz-u] [F ALI] ye-di.

 c. */??Birkac gun once [F ALI] [T istakoz-u] ye-di.


(25c) would sound, at best, very odd in the given context. Apparently, contrary to nontopical background elements, topics are not allowed to occur between the focus and the verb in Turkish. That is to say, the focus appears to serve as a certain boundary in the preverbal position of Turkish sentences. We will return to this point in the next section.

The prefocal segment of the sentence is not the only position that can host topics in Turkish. Topics can also appear postverbally in this language:
(26) Birkac gun once birisi [F ALI-YE ver-di] [T istakoz-u].
 several day before someone Ali-dat give-pst lobster-ace
 'Several days ago someone gave the lobster to Ali.'


The utterances in (24) and (26) could be interchangeably used in a context where the lobster had already been established as the topic of the discourse. As topics are backgrounded elements, there should be nothing surprising in their ability to appear after the verb in an utterance like (26).

Notice that he postverbal topic has been characterized above as an already established one. This is because only such topics are allowed to come after the verb; but, newly introduced ones are barred from this position. For instance,
(27) [F ON milyon civarinda insan yasi-yor] [T Istanbul-da].
 ten million around person live-prog Istanbul-loc
 'Around ten million people live in Istanbul.'


can be felicitously uttered only in a context where Istanbul has already been set up as the topic of the discourse, for example, as a reply to a request like:
(28) Ban-a Istanbul-u anlatsana.
 I-dat Istanbul-acc tell
 'Tell me about Istanbul.'


If the topic is a new one (i.e. one that has not been established by the preceding discourse), the topical phrase must appear sentence-initially. For example, if the topic is one that is contrasted with the topic of the preceding utterance and, hence, a new topic, it will be restricted to the sentence initial position:
(29) [T Edirne] kucuk bir sehir.
 Edirne small one town
 'Edirne is a small town.'

(30) a. [T Istanbul-da] ise on milyon civannda insan yasi-yor.
 lobster-acc cond ten million around person live-prog
 'As for Istanbul, around ten million people live there.'
 b. *On milyon civarinda insan yasi-yor [T Istanbul-da] ise.


Only (30a) can be used as an acceptable continuation to (29). (30b) is unacceptable as its topic is contrastive and occurs postverbally.

As a final remark, the sentence-initial position can host backgrounded elements of any kind, not only those which are topical. The examples we have seen above provide evidence supporting this claim. Hasan in (22), and birkac gun (once 'several days ago' and birisi 'someone' in (24) are all nontopical backgrounded elements and can felicitously appear sentence-initially. It seems that the left and right peripheries of Turkish sentences are symmetrical in terms of their ability to host backgrounded elements, except that new topics are not allowed to occupy a right-peripheral position.

3. Proposal: a syntactic model for information structuring of Turkish sentences

3.1. An initial approach

In the previous section, we have made the following observations on the linear arrangement of focal, backgrounded, and topical segments of Turkish sentences:

(31) OBSERVATIONS ON THE LINEAR ARRANGEMENT OF INFORMATIONAL SEGMENTS OF TURKISH SENTENCES

Foci:

a. Not all loci occur in the immediately preverbal position; and

b. An account considering those appearing in this position as being there to license their focal status does not seem plausible under the light of some theoretical considerations.

Backgrounded constituents:

a. Only backgrounded constituents can appear in the postverbal position; but

b. Such constituents can also appear preverbally.

Topics:

a. Like all backgrounded elements, previously established topics can appear either sentence-initially or postverbally, but unlike nontopical backgrounded elements, no topics can occur between the focus and the verb; and

b. Like loci, new topics are barred from the postverbal position.

A straightforward interpretation of these observations is to deprive the immediately preverbal position of its privileged status of marking the focus and, instead, forefront the focus itself and the verb as two prominent positions serving a demarcation function in the informational segmentation of the sentence. This is, more or less, what Goksel and Ozsoy (2000) argue for in their characterization of focus position in Turkish. They call the area between the constituent that takes focal stress and the position that includes the verb complex the "focus field." It is in that part of the sentence that focal elements can appear:
(32) Focus FIELD

 ... {XP ... V} ...,
 where XP is the element receiving focal stress and marks the
 leftmost boundary of the focus field (Goksel and Ozsoy 2000:
 223).


Goksel and Ozsoy (2000: 227) make a distinction between focal stress and sentential stress and argue that "the immediately preverbal position is NOT the focus position in Turkish, but the position for sentential stress." They also add that although focal elements have to occur in the focus field, not everything that occurs there has to do with focal information. That is, nonfocal elements may appear in the focus field, as well as the positions falling to left and right of this field. These are all compatible with the observations in (31). However, these observations do not necessarily require the so-called "focus field" to be represented as in (32). Goksel and Ozsoy (2000: 227) conclude their study by suggesting that the representation of the focus field might be derived from a syntactic level of representation, as "not everything in any order can be focused at every position."

Interestingly, in an independent study Kilicaslan (1998) offers a phrase structural representation of the "focus field" in Turkish whose linear interpretation exactly matches that in (32):

(33) A PHRASE STRUCTURAL DESCRIPTION OF THE FOCUS FIELD IN TURKISH

[ILLUSTRATION OMITTED]

where the focus field is the domain rooted by S which is delimited by a stressed focal constituent on the left and the verb on the right (Kilicaslan 1998: 133).

According to Kalicaslan (1998), foci always remain within the "core clause" rooted by an S-node, while backgrounded elements can appear within either that core-clause or the left- and right-peripheral positions immediately dominated by an E-node. I will refine and revise that proposal to develop our syntactic model for information structuring in Turkish. Before going into this, however, some clarification is in order about the distinction between the "core" and "periphery" of the sentence.

3.2. The core vs. periphery of the sentence

I will exploit Banfield's (1982) notion of expression in making the distinction between "clause-internal" elements constituting the "core" of the sentence and "clause-external" ones occupying the peripheral positions. Banfield's primary motivation to introduce this notion is a desire to give a syntactic account of the interaction between direct and indirect speech, which can capture both their similarities and their differences. She starts her analysis by pointing out the following peculiarities of direct and indirect speech, which are identified as major differences between the two types of speech by traditional grammar:

(i) a subordinating conjunction (that or whether and if in English, que or si in French) introduces indirect speech; (ii) the verb of indirect speech is subject to concordance of tense rules which affect verbs in many types of subordinate clauses; (iii) the grammatical person of pronouns with the same referent in the main and embedded clauses must be identical in indirect speech alone--that is, there is a concordance of person as well as tense; and (iv) demonstrative elements referring to the time and place of the speech act must also show 'concordance' in indirect speech but not direct speech (Banfield 1982: 25).

Example (34) (from Banfield 1982) illustrates these differences. (34a) and (34b) contain, respectively, direct and indirect speech:

(34) a. Mary told me yesterday at the station, "I will meet you here tomorrow."

b. Mary told me yesterday at the station that she would meet me there today.

Working within a transformational framework of syntax, Banfield first raises the question of whether we can predict these differences between direct and indirect speech by deriving one from the other by means of a plausible transformation. (10) Then, she points out some further peculiarities of the two types of speech which, she argues, definitely make it impossible to relate a sentence of direct speech and a sentence of indirect speech by means of a transformational derivation. Banfield argues that the difference between direct and indirect speech is not the result of a derivation process but of a categorial nature. Some important tenets of the theory she puts forward can be stated, roughly, as follows:

1. All independent utterances are rooted by an E(xpression) node.

2. Ignoring some theory-internal details (such as the postulation of a separate projection for inflection markings or complementisers), the syntactic structure of a sentence is of the following form:

[ILLUSTRATION OMITTED]

3. The reported sentence of direct speech retains its status of being an independent utterance. Thus, it is rooted by an E-node. The reported sentence of indirect speech, on the other hand, loses its independent status by being subordinated to the matrix sentence. Therefore, it is not rooted by an E-node but an S-node.

The E-node/S-node distinction will be the cornerstone of my analysis. I will not go into the details of the line of reasoning which Banfield pursues in order to motivate an E-labelled node. But, in order to find out the distinguishing properties of the projection of E and that of S, I will sometimes refer to the differences which she detects between direct and indirect speech. Given a distinction between an E-node and an S-node dominated by it, it will be possible to give a more precise formulation of the operation of detachment of sentence elements from the core-clause to clause-external external peripheral positions. In my analysis, the S will be taken to be the core-clause and left and right peripheries of the S will be our clause-external positions. The proposal concerning the syntax of information structure in Turkish can now be formulated in general terms as follows:

(35) INFORMATION STRUCTURING OF TURKISH SENTENCES VIA SYNTAX (initial version)

Left- and right-detachment to clause-external positions are the only syntactic operations which are used in the informational structuring of Turkish sentences.

We will examine the right- and left-detachment operations separately. Let us start with the former.

3.3. Right-detachment

Banfield notes that in English, right-detachment, which moves a constituent to the end of the S, is excluded from indirect speech, as shown in the following examples:

(36) a. She replied, "We may be parted for years, I and Peter."

b. *She replied that they might be parted for years, she and Peter.

(37) a. She exclaimed, "How awful they are, women!"

b. *She exclaimed that they were awful, women.

(38) a. "It was so with her--my wife," Mr Ramsay remarked.

b. *Mr Ramsay remarked that it had been so with her--his wife.

This means that in English right-detached constituents are restricted to an S-external position under E. It is for this reason that reported sentences of indirect speech, which are rooted by an S node, cannot contain such constituents. The same constraint, exclusion from reported sentences of indirect speech, applies also to postverbal constituents in Turkish. Turkish has two verbs that can be translated into English as 'to say': soyle-mek and de-mek. Interestingly, the former is used in indirect speech, whereas the latter is used in direct speech. In other words, soyle-mek subcategorizes for an S, while de-mek subcategorizes for an E. The following examples show that the sentential argument of the verb soyle-mek cannot have a postverbal constituent, but that of the verb de-mek can:
(39) a. Ali Oya-nin Kaya-yi ara-dig-i-ni soylu-yor.
 Ali Oya-gen3 Kaya-acc call-ger-poss3-acc say-prog
 'Ali says that Oya called Kaya.'

 b. *Ali Oya-nin [e.sub.i] ara-dig-i-ni Kaya-[yi.sub.i] soylu-yor.

 c. Ali 'Oya [e.sub.i] ara-di Kaya-[yi.sub.i]' di-yor.
 Ali Oya call-pst Kaya-acc say-prog
 'Ali says, "Oya called Kaya."'

(40) a. Ali Oya-nin bir arkadas-i-yla sinema-ya
 Ali Oya-gen3 one friend-poss3-com cinema-dat
 git-tig-i-ni soylu-yor.
 go-ger-poss3-acc say-prog
 'Ali says that Oya has gone to a movie with a friend of hers.'

 b. *Ali Oya-nin [e.sub.i] sinema-ya git-tig-i-ni bir
 arkadas-i-[yla.sub.i] soylu-yor.

 c. Ali 'Oya [e.sub.i] sinema-ya git-ti bir arkadas-i-[yla.sub.i]'
 Ali Oya cinema-dat go-pst one friend-poss3-com
 di-yor.
 say-prog
 'Ali says, "Oya has gone to a movie with a friend of hers."'


I argue that in Turkish all postverbal constituents occur in a position that is immediately dominated by the E. The verb is considered in this language to demarcate a certain sentential domain (cf. Goksel 1998; Kornfilt 1998, 2000; inter alia). In our phrase structure model, this verb-final domain is the S and, hence, any constituent dislocated to a position after the verb has to land on an S-external slot. The problem with the (b) sentences in the examples above is that the embedded clauses of these sentences are rooted by S nodes. For this reason, in contrast to the reported sentences in the (c) cases, these clauses do not have landing sites for their right-dislocated constituents. I should stress that what makes these sentences unacceptable is not the dislocation of a constituent of the embedded clauses. As shown in (41), a constituent of the embedded clause can be dislocated to a postverbal slot of the matrix clause:
(41) Ali Oya-nin [e.sub.i] ara-dig-i-ni soylu-yor
 Ali Oya-gen3 call-ger-poss3-acc say-prog
 Kaya-[yi.sub.i].
 Kaya-acc
 'Ali says that Oya called Kaya.'


As the matrix clause is an independent utterance, it is rooted by an E node. Thus, its postverbal slot is an E-dominated S-external position where the dislocated NP Kaya-yi can felicitously appear, as long as it conveys backgrounded information, of course.

To avoid possible misunderstandings, several points need to be clarified regarding the S- versus E-distinction. (11) Firstly, this distinction does not only apply to the phenomenon of reporting speech. The first clause of our proposal can be stated as follows:

(42) RIGHT-DETACHED CONSTITUENTS IN TURKISH--FIRST CLAUSE

Backgrounded constituents can be right-detached to a position in ANY E-domain.

Kornfilt (1998), for example, makes some observations similar to ours concerning clauses with the particle diye. This particle is a complementizer-like element that takes a fully finite, tensed embedded clause, which need not necessarily encode direct speech:
(43) Ali Oya [e.sub.i] ara-mis ol-mali Kaya-[yi.sub.i] diye
 Ali Oya call-infer.pst be-nec Kaya-acc "saying"
 dusun-uyor-du.
 think-prog-pst
 'Ali was thinking: "Oya should have called Kaya."'


In (43), what is reported is not speech but thought. Crucially, the diye-clause seems to be able to express this thought from the original thinker's point of view. This can be understood from the fact that this clause can contain a lexical item expressing the subjectivity of Ali, who was the original thinker:
(44) Ali bu aptal kadin [e.sub.i] ara-mis ol-mali
 Ali this stupid woman call-infer.pst be-nec
 Kaya-[yi.sub.i] diye dusun-uyor-du.
 Kaya-acc "saying" think-prog-pst
 'Ali was thinking: "This stupid woman should have called Kaya."'


In the given context, bu aptal kadin 'this stupid woman' could be used with a figurative meaning expressive of Ali's point of view. As another difference between direct and indirect speech, Banfield (1982) notes that the former but not the latter can contain such lexical items expressing the point of view or subjectivity of the speaker of the original utterance. In our case, the subjectivity-expressing item signals the fact that the diye-clause retains its independent status by "directly" representing the content of the original thinking process. That is, the diye-clause here introduces an E-domain. It is for this reason that it can host a postverbal constituent.

Secondly, not only embedded S-domains, but also embedded NP- or DP-domains do not allow for subconstituents to right-adjoin to them:
(45) a. Ali Kaya-nin kari-si-ni tani-yor.
 Ali Kaya-gen3 wife-3sg know-prog
 'Ali knows Kaya's wife.'

 b. *Ali [[[e.sub.i] kari-si-ni] Kaya-[nin.sub.i]] tani-yor.


Expectedly, such subconstituents can adjoin to the E-domain after the verb (provided that they convey backgrounded information, of course):

(46) Ali [[e.sub.i] kari-si-ni] taniyor Kaya-[nin.sub.i].

Apparently, backgrounded constituents are barred from right-adjoining to not only S-domains but any other non-E-domains. Thus, the first clause of our proposal needs to be complemented with the following:

(47) RIGHT-DETACHED CONSTITUENTS IN TURKISH--SECOND CLAUSE

Backgrounded constituents can be right-detached to ONLY a position in an E-domain.

Thirdly, once an NP, DP, or a nominalized subordinate clause is itself right-detached, its subconstituents can appear to the right of its verb:
(48) a. Ali soylu-yor [[Oya-nin [e.sub.i] ara-dig-i-ni]
 Kaya-[yi.sub.i]].

 b. Ali tani-yor [[[e.sub.i] kari-si-ni] Kaya-[nin.sub.i]].


(48a) and (48b) represent, respectively, grammatical transforms of the ungrammatical examples in (39b) and (45b). A question might arise about the grammaticality of such examples: does the subconstituent adjoin to the right-detached constituent or occur as an immediate daughter of the E? Proposing that the former alternative is the case would need to be supported by further argumentation. For example, it might be argued that the NP or the S are themselves now part of the E-level and, thus, can serve as hosts of adjunctions. The second alternative, however, does not necessitate any modification to the syntactic framework we have depicted up to this point: a backgrounded constituent can felicitously occur postverbally in an E-domain. It is this second technically more parsimonious alternative that I propose as an explanation of examples like those in (48). Hence, I assign the following (partially described) phrase structure to, for example, (48b):

(49) [ILLUSTRATION OMITTED]

3.4. Left-detachment

Having established the syntactic location of postverbal background elements within the adopted framework, we will now consider the case in Turkish where background elements appear before the verb. More specifically, our main concern will be the question of whether such elements occur S-externally under the E or within the S. This question cannot be given a straightforward answer. This is because Turkish lacks a boundary marker for the leftmost position of the S. In view of the fact that an element of an embedded clause seems to be able to move beyond the boundaries of that clause (e.g. example [41]), the structure for (40a), for instance, may be, at least in principle, any of those listed below: (12)
(50) a. [E Ali [S Oya-nin bir arkadas-i-yla sinema-ya git-tig-i-ni]
 soylu-yor].

 b. [E Ali Oya-[nin.sub.i] [S [e.sub.i] bir arkadas-i-yla
 sinema-ya git-tig-i-ni] soylu-yor].

 c. [E Ali Oya-[nin.sub.i] bir arkadas-i-[yla.sub.j] [s [e.sub.i]
 [e.sub.j] sinema-ya git-tig-i-ni] soylu-yor].

 d. [E Ali Oya-[nin.sub.i] bir arkadas-i-[yla.sub.j]
 sinema-[ya.sub.k] [s [e.sub.i] [e.sub.j] [e.sub.k]
 git-tig-i-ni] soylu-yor].


Thus, we need a linguistic argumentation to specify the location of preverbal background elements. Let us start with one of Banfield's observations about the difference between direct and indirect speech.

We have already seen that right-dislocated constituents and expressions of subjectivity cannot appear in indirect speech. Banfield observe that among the constructions that can appear only in direct speech (but not in indirect speech) are also addressee-oriented adverbials, such as between you and me, frankly, confidentially, to be honest, and if you ask me. She gives the following example to illustrate this fact:

(51) a. John said, "Between you and me, she is lying."

b. John said that {*between him and her / *frankly} she was lying.

The conclusion that follows from this observation is that addressee-oriented adverbials cannot occur under the S (i.e. they are restricted to an S-external position under the E). Assuming that the same constraint also applies to their Turkish counterparts, we can use such adverbials to detect the syntactic position of a constituent in Turkish. The idea to be exploited is that any constituent that appears to the left of an addressee-oriented adverbial must occupy an S-external slot directly dominated by the E node. Consider the examples below:
(52) Oya NEREDE?
 Oya where
 'Where is Oya?'

 a. Laf aramizda, Oya bir arkadas-i-yla [F
 word between.you.and.me one friend-poss3-com
 SlNEMA-VA git-ti].
 cinema-dat go-pst
 'Between you and me, Oya has gone to a MOVIE with a friend
 of hers.'

 b. Oya, laf aramizda, bir arkadas-i-yla [F SINEMA-YA git-ti].

 c. Oya bir arkadas-i-yla, laf aramizda, [F SINEMA-YA git-ti].

 d. *Oya bir arkadas-i-yla [F SINEMA-YA, laf aramizda, git-ti].

(53) Oya-nin KIM-I sev-dig-i-ni merak ed-iyor-um.
 Oya-gen3 who-acc love-ger-poss3-acc wonder do-prog-1sg
 'I wonder who Oya loves.'

 a. Ban-a sor-ar-sa-n, Oya [F KAVA-YI] sev-iyor.
 I-dat ask-aor-cond-2sg Oya Kaya-acc love-prog
 'If you ask me, Oya loves Kaya.'

 b. Oya, ban-a sor-ar-sa-n, [F KAYA-YI] sev-iyor.

 c. *Oya, [F KAYA-YI], ban-a sor-ar-sa-n, sev-iyor.


Laf aramizda and bana sorarsan are the Turkish equivalents of the English adverbials 'between you and me' and 'if you ask me,' respectively. A plausible inference that could be drawn from the examples above might be the following. In Turkish, background elements can be left-detached to S-external positions, while focal ones do not seem to enjoy this freedom of movement at least to the same extent. Before going into further substantiation of this inference, I would like to emphasize that the exclusion of the constituents sinema-ya 'cinema-dat' and Kaya-ya 'Kaya-dat' from S-internal positions in examples (52) and (53) is only because of their informational status. These dative constituents are not inherently restricted to S-external positions. The following examples show that they can felicitously appear outside the S, which means before an addressee-oriented adverbial in the given contexts, insofar as they are part of the background:
(54) Oya sinema-ya KIM-LE git-ti?
 Oya cinema-dat who-com go-pst
 'Who did Oya go to a movie with?'

 a. Oya sinema-ya, laf ararmizda, [F bir
 Oya cinema-dat word between.you.and.me one
 ARKADAS-I-YLA] git-ti.
 friend-poss3-com go-pst
 'Between you and me, Oya has gone to a movie with [F a
 friend of hers].'

 b. *Oya sinema-ya [F bir ARKADAS-I-YLA], laf aramizda, git-ti.

(55) Oya-nin Kaya icin NE hisset-dig-i-ni merak ed-iyor-um.
 Oya-gen3 Kaya for what feel-ger-poss3-acc wonder do-prog-1sg
 'I wonder what Oya feels about Kaya.'
 Oya Kaya-yi, ban-a sor-ar-sa-n, [F SEV-IYOR].
 Oya Kaya-acc I-dat ask-aor-cond-2sg love-prog
 'If you ask me, Oya [F LOVES] Kaya.'


Further evidence about the syntactic distribution of focal and backgrounded constituents in Turkish comes from structural differences that the E and S projections display. The E projection hypothesized by Banfield does not conform to the usual linguistic assumptions. For instance, there is no element that may head such a projection. That is, there is no [E.sup.0]. King (1993), approaching the problem from the point of view of government-binding theory, characterizes E as an anomalous projection: "the E projection does not conform to the usual [bar.X] schema: there is no head, no specifier" (King 1993: 99). S and E differ also in terms of the structural arrangement of their constituents. S has a configurational structure. Its constituents are organized in accordance with some dominance and precedence rules. All S-external constituents, on the other hand, are organized in a flat structure and not restricted to any particular word order. In Turkish, the part of the sentence that includes background elements that occupy the left and right peripheries of the sentence displays highly nonconfigurational characteristics, while the part of the sentence that starts with the focus and ends with the verb has a fairly rigid structure. That is, the former part appears to conform to the structure of E and the latter to that of S. Let us first observe the linear arrangement of constituents in the indicated portions of the sentence. Consider (56):
(56) Fido Oya-ya NE yap-ti?
 Fido Oya-dat what do-pst
 'What did Fido do to Oya?'

 a. Fido Oya-yi [F ISIR-DI].
 Fido Oya-acc bite-pet
 'Fido [F BIT] Oya.'

 b. Oya-yi Fido [F ISIR-DI].

 c. Fido [F ISIR-DI] Oya-yi.

 d. Oya-yi[F ISIR-DI] Fido.

 e. [F ISIR-DI] Fido Oya-yi.

 f. [F ISIR-DI] Oya-yi Fido.


All the permutations of the response sentence are equally fine in the given context. There is no restriction on the ordering of the background elements. Let us now embed the same sentence in a context where it will receive an all-focus interpretation:
(57) Parti-de ilginc birsey OL-DU mu?
 Party-loc interesting anything happen-pst Q
 'Did anything interesting happen at the party?'

 a. [F Fido OYA-YI isir-di].
 Fido Oya-acc bite-pst
 '[F Fido bit OYA].'

 b. *[F Oya-yi Fido isir-di].

 c. *[F Fido isir-di Oya-yi].

 d. *[F Oya-yi isir-di Fido].

 e. *[F Isir-di Fido Oya-yi].

 f. *[F Isir-di Oya-yi Fido].


In the context above, all the permutations except the first one sound, at best, very odd. This suggests that there is a word order restriction on the constituents that appear within the focus. Even though the oddity of the last four permutations might be attributed to the postverbal occurrence of a focal element, it seems that the unacceptability of (57b) can be accounted for only in terms of a deviation from a certain word order, which is SOV in the given example.

The background and focus of a Turkish sentence display contrasting behaviors also in terms of dominance relations. Consider the example below:
(58) Kaya parti-de NE yap-ti?
 Kaya party-loc what do-pst
 'What did Kaya do at the party?'

 a. [F KONUK-LAR-[I.sub.i] birbirleri-[ne.sub.i] tanit-ti].
 guest-pl-acc each.other-dat introduce-pst
 'He introduced the [guests.sub.i] to each [other.sub.i].'

 b. *[F Birbirleri-[ne.sub.i] KONUK-LAR-[I.sub.i] tanit-ti].


While the sentence in (58a) is perfectly fine, the one in (58b) is unacceptable in the same context. One might be tempted to attribute the contrast in acceptability between the two sentences to the difference between their word orders. Some researchers argue that the unmarked word order for such sentences is: subject-direct object-indirect object-verb (cf. Underhill 1972; Erguvanli 1984; Kalicaslan 1998). Therefore, the unacceptability of the (b) sentence might be argued to be the result of the marked ordering of its constituents. However, such an account would fail to predict the acceptability of an utterance like (59) in exactly the same context (i.e. as an answer to the question in [58]):
(59) [F Her yeni gel-en konug-a daha once gel-mis
 every newly arrive-part guest-dat further before arrive-pst
 OL-AN-LAR-I tanit-ti].
 be-part-3pl-acc introduce-pst
 'He introduced those who had arrived earlier to every newly
 arriving guest.'


The problem with (56b) seems to arise from the structural relation between the anaphoric NP and its antecedent. From the unmarked word order proposed for this sentence (i.e. S-DO-IO-V) it follows that the anaphoric NP is dislocated to a syntactic position that stands in the following relation to its antsecedent:

(60) [ILLUSTRATION OMITTED]

This structured relation can be described in terms of the syntactic notion of c-command, which is a configurational relation between two constituents: a node A c-commands a node B if A does not dominate B (i.e. the constituent labelled by B is not a constituent of the one labelled by A) and the first branching node dominating A also dominates B. Thus, in (58b) the anaphor c-commands its antecedent but not vice versa. However, it is usually argued that this is a strictly non-legitimate structural relation between an anaphor and its antecedent (cf. government-binding theory). Postulating a configurational structure for S and assuming that the focus of a Turkish sentence is confined to the boundaries of that projection seem to provide a way to give a plausible account of examples like this.

The discussion above suggests that the focus of a Turkish sentence is imposed a fairly rigid structure also in terms of dominance relations among the constituents. Interestingly, in Turkish background elements that occupy peripheral positions behave again in the opposite manner. That is, they do not seem to require any particular hierarchical organization among themselves. This fact shows up in the arbitrary placement of an anaphor and its antecedent in peripheral slots, when they are part of the background:
(61) Parti-de konuk-lar-[i.sub.i] birbirleri-[ne.sub.i]
 party-loc guest-pl-acc each.other-dat
 KIM tanit-ti?
 who introduce-pst
 'Who introduced the [guests.sub.i] to each [other.sub.i]
 at the party?'

 a. Patti-de konuk-lar-[i.sub.i] birbirleri-[ne.sub.i]
 Party-loc guest-pl-acc each.other-dat
 [F KAYA] tanit-ti.
 Kaya introduce-pst
 '[F KAYA] introduced the [guests.sub.i] to each
 [other.sub.i] at the party.'

 b. Parti-de birbirleri-[ne.sub.i] konuk-lar-[i.sub.i] [F KAYA]
 tanit-ti.

 c. Patti-de konuk-tar-[i.sub.i] [F KAYA] tanit-ti
 birbirleri-[ne.sub.i].

 d. Parti-de birbirleri-[ne.sub.i] [F KAYA] tanit-ti
 konuk-lar-[i.sub.i].

 e. Patti-de [F KAYA] tanit-ti konuk-lar-[i.sub.i]
 birbirleri-[ne.sub.i].

 f. Parti-de [F KAYA] tanit-ti birbirleri-[ne.sub.i]
 konuk-lar-[i.sub.i].

(62) [Kaya.sub.i] kendi-[si.sub.i] hakkinda NE dusun-uyor?
 Kaya self-poss3 about what think-prog
 'What does [Kaya.sub.i] think about [himself.sub.i]'

 a. [Kaya.sub.i] kendi-[ni.sub.i] [F COK
 Kaya self-acc very/much
 begen-iyor].
 be.pleased.with-prog
 '[Kaya.sub.i] is [F VERY pleased] with [himself.sub.i].'

 b. Kendi-[ni.sub.i] [Kaya.sub.i] [F COK begen-iyor].

 c. Kayai [F COK begen-iyor] kendi-[ni.sub.i].

 d. Kendi-[ni.sub.i] [F COK begen-iyor] Kayai.

 e. [F COK begen-iyor] [Kaya.sub.i] kendi-[ni.sub.i].

 f. [F COK begen-iyor] kendi-[ni.sub.i] [Kaya.sub.i].


Each of the (a)-(f) sentences in the examples above can be a felicitous answer to the related question, even though some may be preferred to the others due to some factors such as priming effect. (13) A plausible account of such examples could be given again by referring to the notion of c-command. It could be argued that all the utterances in these examples are acceptable because in each case the anaphor and its antecedent are organized in a flat structure such that they can c-command each other. C-commanding each other is mostly considered to be an appropriate structural relation between an anaphor and its antecedent. Besides, this would be entirely compatible with our claim that in Turkish background elements that appear at the peripheries of the sentence can occur outside of the S as immediate daughters of E, which is a nonembeddable top-level node (except for coordination). Consider, for instance, the structures below, which can be possibly assigned to sentences (62a)-(62f):

(63) [ILLUSTRATION OMITTED]

It is clear that in all these structures the anaphor and the antecedent, which are the background elements, will be able to c-command each other.

As an alternative account of examples like (61) and (62), one could argue that when an anaphor and its antecedent are detached to clause-external slots, they do not enter the relation that would normally impose on them a certain hierarchical configuration (which would be formulated in terms of c-command relation). Either of these alternatives allows us to reasonably assume that in Turkish peripheral background elements can felicitously occur as immediate daughters of the root node E.

To sum up, the observations made above reveal the following contrasts between backgrounded and focused constituents:

(64) SOME CONTRASTS BETWEEN BACKGROUNDED AND FOCUSED CONSTITUENTS IN TURKISH

1. Backgrounded constituents may precede addressee-oriented adverbials but focused ones cannot; and

2. Backgrounded constituents display highly flexible and nonconfigurational behaviors, but focused constituents are organized in a fairly rigid and configurational structure.

Resting on the assumptions that addressee-oriented adverbials appear S-externally in Turkish and that the E projection is flexible and nonconfigurational except the structure of its S daughter, Kilicaslan (1998) proposes the diagram in (33) as the general form of the syntactic realization of the focus-background articulation of Turkish sentences. The claims encoded by this diagram can be stated in words as follows:

(65) KILICASLAN'S (1998) CLAIMS ON THE SYNTAX OF FOCUS AND BACKGROUND IN TURKISH

1. Backgrounded constituents following the verb occupy S-external positions; likewise,

2. Backgrounded constituents preceding the focus have to occur S-externally under the E; and

3. Focused constituents cannot escape to an S-external position.

I do not see any reason to be dubious about the first claim. The examples we have discussed in Section 3.3 clearly show that sentences not rooted by an E node (i.e. those lacking S-external positions) cannot have postverbal constituents in this language. However, we are in fact far from having sufficient evidence to come to a conclusive decision about the last two claims. It is beyond doubt that no type of focus can be right-detached to a postverbal position in Turkish. But, the ban on left-detachment to a sentence-initial position does not actually seem to apply to all types-focus. The foci discussed above in relation to left-detachment have all been assumed to be presentational. Nonpresentational foci appear to display some flexibility in terms of detachability to a left-peripheral position. For example, when the focus is an exhaustively interpreted one, it seems to be able to appear to the left of an addressee-oriented adverbial:
(66) Calisan-lar patron-u sev-iyor MU?
 Employee-pl boss-acc love-prog Q
 'Do the employees like the boss?'
 [F Yalnizca HASAN], durust olmak gerekirse, patron-u sev-iyor.
 only Hasan to.be.honest boss-acc like-prog
 (Diger-ler-i o-ndan nefret ed-iyor.)
 other-pl-poss3 he-abl hate do-prog
 'To be honest, only Hasan likes the boss. (The others hate him.)'


In this example, the exhaustive focus felicitously appears before the addressee-oriented adverbial durust olmak gerekirse 'to be honest'. That is, the S-external occurrence of this focal constituent is acceptable. Such freedom of leftward movement can also be observed with contrastive foci, as illustrated in the following example:
(67) Oya MI Mehmet-i sev-iyor, Ayse MI?
 Oya Q Mehmet-acc love-prog Ayse Q
 'Does Oya loves Mehmet or Ayse?'

 [F AYSE], bana sorarsan, Mehmet-i sev-iyor. (Oya Kaya-yi
 Ayse if.you.ask.me Mehmet-acc love-prog Oya Kaya-acc
 sev-iyor.)
 love-prog
 'If you ask me, Ayse loves Mehmet. (Oya loves Kaya.)'


Apparently, the third claim in (65) needs to be weakened as follows:

(68) LEFT-DETACHMENT AND FOCUSED CONSTITUENTS IN TURKISH

Presentational foci cannot be left-detached to a position in the E-domain.

We will return to the pointed difference between presentational foci on the one hand and exhaustive and/or contrastive foci on the other in Section 4. Let us now question the validity of the second claim above. To start with the data leading to the first observation in (64), examples like (52)-(55) do not suffice to convincingly assert that all background elements preceding the focus necessarily occur S-externally in Turkish. The position of an addressee-oriented adverbial serves only to indicate that the elements falling to its left are S-external but it does not say anything about those falling to its right, like those in (52a), (52b), and (53a). So, only by exploiting the assumption that an addressee-oriented adverbial must be an immediate daughter of an E-labelled node, we cannot directly jump to the conclusion that background elements appearing between an addressee-oriented adverbial and a focal constituent (e.g. those in [52a], [52b], and [53a]) occur outside he boundaries of their Ss. Needless to say, we are now talking about foci in situ. We have just stated that nonpresentational foci can sometimes be left-detached to the E-domain. It would not, of course, be informative to add that background elements falling to the left of such left-detached foci are S-external. Taking also this latter point into consideration, I suggest that the second claim in (65) should be weakened as follows:

(69) LEFT-DETACHMENT AND BACKGROUNDED CONSTITUENTS IN TURKISH

Background elements preceding a focal constituent in situ are able to, but do not have to, appear S-externally.

This is a simple manifestation of the fact that not all constituents that occur under the S have to be focused in Turkish. We have already seen that it is possible to have backgrounded constituents that occupy a position between a focal constituent that is presentational (and, hence, probably in situ) and the verb (e.g. examples [12] and [13]). In fact, Kilicaslan (1998) admits that examples involving addressee-oriented adverbials do not provide sufficient evidence to assert that the focus serves as the left-most boundary of the S in Turkish and argues that more convincing evidence in favor of this claim comes from our second observation in (64). Actually, not only focal constituents but also backgrounded ones manifest rigid and configurational behaviors when they occur within the S. Consider the following examples, where the focal constituents are supposed to be interpreted as presentational:
(70) NE zaman Fido Oya-yi isir-di?
 What time Fido Oya-acc bite-pst
 'When did Fido bite Oya.'

 a. [F GECEN hafta] Fido Oya-yi isir-di.
 last week Fido Oya-acc bite-pst
 'Fido bit Oya LAST week.'

 b. Oya-yi Fido [F GECEN hafta] isir-di.

 c. *[F GECEN hafta] Oya-yi Fido isir-di.

(71) Patti-de KIM konuk-lar-[i.sub.i] birbirleri-[ne.sub.i]
 party-loc who guest-pl-acc each.other-dat
 tanit-ti?
 introduce-pst
 'Who introduced the [guests.sub.i] to each [other.sub.i] at the
 party?'

 a. Parti-de [F KAYA] konuk-lar-[i.sub.i] birbideri-nei
 Party-loc Kaya guest-pl-acc each.other-dat
 tanit-ti.
 introduce-pst
 '[F KAYA] introduced the [guests.sub.i] to each [other.sub.i]
 at the party.'

 b. *Parti-de [F KAYA] birbideri-[ne.sub.i] konuk-lar-[i.sub.i]
 tanit-ti.


The (c) sentence in (70) and the (b) sentence in (71) are problematic because their backgrounded constituents are not arranged in accordance with the unmarked word order, S-DO-IO-V. These examples show that the high degree of flexibility which backgrounded constituents that fall to the left of the focus or to the right of the verb enjoy in their ordering is not available for backgrounded constituents that appear between the focus and the verb. (70c) is unacceptable because, in such sentences, a deviation from the SOV order is not permitted within the S. (71b) is unacceptable because the dislocation of the dative marked anaphor from its base position causes it to occupy a position where it can c-command its antecedent but it cannot be c-commanded by it.

Is, however, the indicated contrast between prefocal and postfocal backgrounded constituents necessarily a sign of the S-external occurrence of all of the former? The data exemplifying the contrast in question do not actually allow for such a strong inference. The marked object-subject ordering in (70b), for instance, could be the result of the S-external occurrence of both arguments but also that of the following structural arrangement:

(72) [ILLUSTRATION OMITTED]

In this example, we do not have any evidence that would oblige us to place the NP Fido in a position outside the S, though this is a backgrounded constituent that comes before the focus.

3.5. Detachment in general

In light of the observations made up to this point, I propose the following as the general form within which the background-focus structuring of Turkish sentences must be realized:

(73) SYNTACTIC DISTRIBUTION OF FOCUS AND BACKGROUND IN TURKISH

[ILLUSTRATION OMITTED]

That is, the only syntactic constraint on the focus-background articulation of Turkish sentences is that presentational foci are restricted to the boundaries of the core clause, which is referred to as the S in this study, while [+contrastive] and/or [+exhaustive] foci can sometimes escape to the left periphery of the sentence. I will endeavor to account for this constraint in more explanatory terms in Section 4. Before going into this, we will finally look at the syntactic placement of topical constituents in Turkish within the established framework.

As discussed in Section 2.3, both sentence-initial and postverbal positions can host topics in Turkish. As topics are backgrounded elements, there should be nothing surprising about their ability to appear at these peripheral positions. But, as should be recalled, contrary to ordinary background elements, topics cannot occur between the focus and the verb. I take this latter observation as evidence suggesting that topics must be S-externalized in Turkish. Actually, clause-external placement of topics seems to be a strategy commonly used by various languages. For example, according to Vallduvi (1990, 1993, 1994, 1995), in Catalan, which he argues to be an underlyingly VOS language, topics (which he refers to as links) must always be left-detached to a sentence-initial position external to the core clause. The following may be an answer to the Catalan equivalent of a question like 'What about Pere? What did he do?':
(74) [T El [Pere.sub.1]] [F es va menjar els FESOLS [t.sub.1]].
 Pere be 3sg-pst-eat the beans
 '[T Pere] [F ate the BEANS].'


Vallduvi (1990), working within a government-binding framework, describes the core clause as the lowest IP and argues that topics are left-adjoined to IP. Vallduvi and Engdahl (1994) make a similar proposal, but this time within a head-driven phrase structure grammar framework: they postulate a language particular immediate dominance schema that introduces these phrases as sisters of S. (14)

Another proposal about the syntactic realisation of topics, which is more in line with the analysis I offer for Turkish, is made by Rudin (1985), Aissen (1992), and King (1993). Following Banfield (1973), these linguists, too, postulate an E-labelled root projection for the languages they examine. They argue that certain topics, which they call "external" topics, occur clause-externally under the E node. Let us have a brief look at their accounts. (15)

Rudin (1985) argues that Bulgarian has external topics which are dislocated to the left of S', as a daughter to the node E. Bulgarian left-dislocated topics, according to Rudin, are always definite and associated with a resumptive pronoun. Ivan in the following Bulgarian sentence is such a topic:
(75) [E [Ivan.sub.i] [s' [nego.sub.i] vidjah vcera]].
 Ivan him saw-1sg yesterday
 'Ivan, I saw him yesterday.'


Aissen (1992) makes a similar proposal for the syntactic structuring of topics in Tz'utujil, Tzotzil, and Jakaltek (which are Mayan languages). She takes the core clause to be rooted by CP. She claims that Tzotzil and Jakaltek have only external topics, while Tz'utujil is a language with internal topics. That is, in the former two languages, topics occur outside the core clause (i.e. CP) under E, whereas in the latter language, they can be within the CP. The most important evidence for the CP-external occurrence of topics in Tzotzil and Jakaltek is that topics cannot occur in subordinate clauses in these languages. In Tzotzil, topics are preceded by a topic marker, a, and usually take a definite determiner. As the following example illustrates, it is not possible for them to appear in embedded clauses:
(76) a. liyalbe li xun-e ti taxtal li petul-e. (Tzotzil)
 be.told.me det Xun-enc comp comes det Petul-enc
 'Xun told me that Petul was coming.'

 b. *liyalbe li xun-e ti taxtal [T a li petul-e].
 he.told.me det Xun-enc comp comes top det Petul-enc
 'Xun told me that [T Petul] was coming.'

 c. *liyalbe li xun-e ti [T a li petul(-e)]
 he.told.me det Xun-enc comp top det Petul-enc
 taxtal(-e).
 comes(-enc)
 'Xun told me that [T Petul] was coming.'


King (1993) makes similar observations for Russian. She argues that Russian is an underlyingly VSO language and all arguments of the verb, including the subject, are projected in the VP at D-structure. In Russian, according to her, external topics are not arguments of the verb, although they may be coreferential with one. Besides, they are in the nominative case, regardless of the case or grammatical function of pronouns with which they are coreferential. In King's account, such topics are detached to a clause-external position under the E. Below is a Russian sentence where the topic, namely Boris, has undergone such a detachment process:
(77) [E Borisi [CP [ja.sub.i] [ego.sub.i] h vcera]].
 Ivan him saw-1sg yesterday
 'Ivan, I saw him yesterday.'


I take the languages mentioned above to show close parallels to Turkish in terms of detachment of topics to clause-external positions. However, recall that this detachment process runs for old topics either leftwards or rightwards in Turkish. I suggest the following generalization for the syntax of topics in Turkish:

(78) SYNTAX OF TOPICS IN TURKISH

Newly established topics are left-detached, while already established ones are left- or right-detached to a clause-external position in the E-domain.

Before concluding this section, I would like to stress one point about left- and right-detached backgrounded constituents. What differentiates these two types of constituents is not that the left ones are topical whereas those on the right are purely backgrounded. Indeed, as already argued for, left-detached constituents might include nontopical but merely backgrounded constituents, and right-detached ones might include topical ones. Given that there is no restriction in principle on the number and type of left-detached constituents, labelling all such constituents as "topics" would turn this term into a mere cover term without any significant descriptive content. If the criterion of aboutness is to be preserved in the characterization of topics, this term should not be applied to, for example, nonspecific or anaphoric expressions, which might be encountered at the left periphery of the sentence. Besides, if the notion of topichood is to cover continuing (i.e. already established) topics, the possibility of topical interpretation should also be recognized for postverbal constituents, as such topics can mostly be shifted to the postverbal E-domain without requiring a change in the context.

4. Concluding remarks

4.1. Generalizations deriving from what has been said thus far

The main conclusion that derives from the discussion offered in this article up to this point can be stated in terms of two clauses, one being negative and the other positive.

(79) SYNTAX OF INFORMATION STRUCTURE IN TURKISH

a. Turkish does not employ any syntactic strategy to mark the informational status of a sentence element (cf. Section 2);

b. But, some elements with certain informational properties, namely backgrounded elements (including topics) and contrastive/exhaustive foci, may undergo a syntactic operation of detachment from the core clause to the peripheries of the sentence (cf. Section 3).

The peripheries of the sentence have been distinguished from the core-clause by exploiting Banfield's (1982) notion of E(xpression):

(80) TURKISH SENTENCE STRUCTURE

When analyzing the structure of a sentence, a distinction needs to be made between:

a. A fairly rigid and configurational S-DOMAIN, which constitutes the core-clause; and

b. A flexible and nonconfigurational E-DOMAIN, which comprises clause-external/peripheral positions.

Certainly, such a characterization of Turkish as both a configurational and a nonconfigurational language may have implications for the analysis of other aspects of the Turkish sentence. However, leaving such potentials to be explored in future research, I would like conclude this study with some other remarks showing how the proposed account can be further deepened along several lines.

4.2. A shift in the point of view

Given that elements remaining within the S can still be interpreted as backgrounded, why do some of backgrounded elements move to the E domain? The detachment of a backgrounded element does not seem to bring about any apparent change in the informational status of that element. This is in entire agreement with the negative side of my proposal. No syntactic strategy is used to mark the informational status of a sentence element. Nonetheless, this seems, at first glance, to be in conflict with the last resort condition of the minimalist approach: movement must happen for a certain reason.

It is my opinion that the reason behind the movement of some elements to the E-domain can be better understood if we shift our point of view from the moving sentence element to the sentence itself. Such movement happens to satisfy some informational property of not the moving item(s) but the whole of the sentence. I will shortly return to this property. Let us now have a minimalist look at the formal side of this approach. Lasnik (2002) maintains that movement needs to be "seen from the point of view of the target rather than the moving item itself." To give an example, "suppose that Infl has a feature that must be checked against the NP. Then as soon as that head has been introduced into the structure, it 'attracts' the NP or DP that will check its feature.... The Case of the NP does get checked as a result of the movement, but that is simply a beneficial side effect of the satisfaction of the requirement of the attractor" (Lasnik 2002: 435). In the earlier minimalist approach, the satisfaction of a property of the moving item was taken to be the driving force of movement. This view was called "greed" by Chomsky (cf. Section 2.1). The later view, where the driving force is the satisfaction of a requirement of the target, has been called "enlightened self interest" by Lasnik.

Speaking in terms of the enlightened self interest, the operation that detaches some elements to the peripheries of the sentence can be described as follows:

(81) "ATTRACTION" TO THE E-DOMAIN

As soon as the sentence has been projected to the level of E(xpression), some sentence elements may be attracted to the E-domain to satisfy a certain informational requirement of that domain.

So, what is that requirement that needs to be satisfied by means of the elements attracted to the E-domain?

4.3. Two layers of sentence interpretation

According to Banfield (1982), the function of S is to express the purely truth-conditional content of the utterance, whereas that of E is to deal with non-truth conditional aspects of interpretation entailing reference to the pragmatic context of utterance. It is for this reason that addressee-oriented adverbials or expressions of subjectivity, referring to the bearer's and speaker's mental states, are excluded from the S of indirect speech (cf. Section 3.4). To be more specific, an important function of the E-domain seems to ground or embed the truth-conditional content encoded by the S in the context of utterance. This task is performed by linking that content to the heater's mental state via the textual and situational environment.

I hold the view that elements detached to the E-domain serve to carry out the function of linking the sentence to the discourse context. That is to say, the driving force behind the process of detachment to clause-external positions is that linking requirement.

Backgrounded elements are the most appropriate items that could link the sentence to the context, as they encode discourse-linked information. As the driving force behind their detachment is to satisfy the requirement of grounding the sentence, not all backgrounded elements need to undergo this process. Once the sentence has been grounded with the backgrounded elements detached to the E-domain, the rest will remain in situ.

Another question arises at this point as to what is common to detached backgrounded constituents and left-detached foci, if the former are invested with the function of linking the sentence to the discourse-context?

4.4. Backgrounding via foci

It seems possible to reconcile detached backgrounded constituents and foci by having recourse to Enc's (1991) account of specificity. On her account, an NP/DP is specific if its referent(s) is (are) linked to already established discourse referents through:

--either an identity relation

--or a subset relation.

We have seen that left detached foci must be contrastive/exhaustive. Such foci introduce a closed set of individuals known to the participants of the discourse and identify the subset of it of which the predicate holds. I suggest that these foci are Enc-specific, and thereby, rink the sentence to the discourse-context via the relevant subset relation. Presentational foci do not come with a superset that can serve as a linking element and, hence, are not licensed to appear in the E-domain, where only elements with a discourse-linking function can occur.

Ward (1985) offers an account of focus preposing in English which seems entirely in the same spirit as our account of left-detached foci in Turkish. He suggests that preposing has two functions in English: 1) marking the referent of the preposed constituent as the BLC (backward looking center) of an utterance, and 2) encoding the focus. Neglecting the details, the BLC is defined as that element which "links up" the current sentence with the preceding discourse. As to how a sentence element can perform such a dual function, consider the following example from Ward (1985: example [207]):

(82) I think she was Japanese. No--KOREAN she was.

Ward points out that preposed foci serve to "call up" a scale into the salient discourse and specify a value in that scale. What Korean does in (82) is "call up" the scale '(Asian) nationalities' and specify the value 'Korean' in that scale. This scale is marked as the BLC, and the specification as focus. Thus, Ward's "scales" are like our supersets in that they both enable a focused element to also carry out a linking function.

Interestingly, in Turkish foci can take on a discourse-linking function only through left-detachment. Right-detachment is not an operation available for Turkish foci in any possible way. A question might naturally come to mind as to what the relevant difference is between left- and right-detachment in Turkish.

4.5. Asymmetry between left- and right-peripheral positions

To avoid possible misunderstandings, a remark is in order about the symmetry between left- right-peripheral positions in Turkish sentences. These positions are symmetrical in terms of serving as nonconfigurational and flexible stores for discourse-linking sentence elements. But, this symmetry is distorted when these elements are further subclassified. As already stated in Section 2.4, already established topics can appear in either the left- or fight-periphery. However, newly established topics are restricted to the former position. As just stressed, the same restriction also applies to contrastive/exhaustive loci. They can be left-detached but are barred from right-peripheral positions. I suggest that what is relevantly common to a newly established topic and left-detached focus is a factor of newness. The former presupposes a shift from an old topic to a new one, while the latter presupposes an operation of narrowing down an old set of elements to a new set of elements (of which the predicate holds). Apparently, the left- and right-peripheries of a Turkish sentence contrast with each other in that only the former is capable of dealing with such factors of newness.

I would like to conclude this study with a last question about the immediately preverbal position in Turkish: why has this position been ascribed a focus-marking status?

4.6. A fallacious impression

Recall that the point of departure for this study was a doubt on the informational status of three sentential positions: sentence-initial, postverbal, and immediately preverbal. The sentence-initial and postverbal positions have retained their status of being information-structurally significant positions by being redefined as respectively left- and right-peripheries. The immediately preverbal position, however, has not been recognized any syntactic role in the information structure of Turkish sentences. This position might be the position for sentential stress, which, as Goksel and Ozsoy (2000) show, need not necessarily be the same as focal stress. But, it is not a syntactically allocated position for focus. There might be statistical correlation between the immediately preverbal position and the surface realization of foci in Turkish. It appears that Turkish foci come just before the verb in most cases. Does this, however, necessarily mean that there is a syntactic relation of dependence between focus and this position in Turkish? As discussed in Section 2.1, a positive answer to this question runs into serious empirical and theoretical difficulties. I argue that in Turkish, focus marking never triggers a syntactic operation, such as the displacement of a constituent. The frequent appearance of foci in the immediately preverbal position is the result of the displacement of background elements. The strong tendency which nonfocal constituents (excepting the verb) display to leave their position within the "core" clause (by means of left- or right-detachment to clause external positions) mostly leads to a surface arrangement of sentence elements where no constituent remains between the focal constituent and the verb. To sum up, the frequent appearance of Turkish foci just before the verb is a surface manifestation of a backgrounding strategy, not a focusing operation.

Received 12 July 2001

Revised version received

10 December 2003

Notes

* I am indebted to Elisabet Engdahl, Euric Vallduvi, Sheila Glasbey, and Evan Klein for helping me shape my ideas about information structuring and its realization in Turkish. I am also grateful to the reviewers for their very valuable comments that allowed for a vast improvement of this paper during the review process. Correspondence address: University of Trakya, Faculty of Engineering and Architecture, Dept. of Computer Engineering, Edirne, Turkey. E-mail: yilmaz_kilicaslan@hotmail.com.

(1.) An alternative view on what is the basic word order in Turkish might be derived by having recourse to Kayne's (1994) "linear correspondence axiom" which dictates that the universal word order is specifier-head-complement. Kayne's system forces the derivation of OV languages from VO languages. In this article, I will not go into the question of whether Turkish could be considered an underlyingly SVO language. However, the interested reader may refer to Kural (1994), who argues against that possibility.

(2.) This view is found in Vallduvi's (1993) or Chafe's (1976) characterization of (information) packaging, where emphasis is placed on the communicative aspect of the discourse context:

I have been using the term packaging to refer to the kind of phenomena at issue here, with the idea that they have to do primarily with how the message is sent and only secondarily with the message itself, just as the packaging of toothpaste can affect sales in partial independence of the quality of the toothpaste inside (Chafe 1976: 28).

A sentence, in one of its facets, may be viewed as a structural vehicle used to transfer some piece of knowledge (a proposition) from speaker to hearer. Information packaging is the speaker's tailoring of this structural vehicle to suit some 'communicative' aspect of the transfer of knowledge (propositional content) to the hearer (Vallduvi 1993: 2).

(3.) The distinction between presentational focus and contrastive focus goes back to at least Halliday (1967). This distinction has also been treated under different rubrics, such as weak vs. strong forms of focus (Rullman 1995) or information focus vs. focus operator (Kiss 1995).

(4.) As Kiss (1995: 189-190) points out, the focus is contrastive if the set it operates on "is a closed set of individuals whose members are known to the participants of the discourse ... as in this case the identification of a subset of this set goes together with the formation of a contrasting complement subset." See also Szabolcsi and Zwarts (1993).

(5.) Kural (1992) notes another relevant dissimilarity between Hungarian and Turkish: unlike the former, focus by stress is an available strategy in the latter language. This could be taken as one of the reasons why Turkish employs more flexible syntactic strategies to realize focus.

(6.) As already mentioned, when making this statement, Kural disregards cases where focus appears in a place other than the immediately preverbal slot as cases of contrastive focus.

(7.) Various other terms have been used to refer to the notion of topic, such as "theme" (Halliday 1967, 1985), "S-topic" (Valimaa-Blum 1988), and "link" (Vallduvi 1990, 1993).

(8.) It is noteworthy that the criterion of aboutness is a very slippery one and that the tests based on this notion are sometimes too strong in that they fail to recognize actual topics and sometimes too weak in that they identify more than one element as topics of the sentence (cf. Gundel 1974; Prince 1984; Ward 1985; and Vallduvi 1988). As our concern in this study will be what syntactic position a possible candidate of topic will occupy, rather than which element of a contextualized sentence is the topic, we will be able to ignore the indicated defects of the topic tests.

In fact, other than the vague criterion of aboutness, topical elements linguistically manifest some properties which distinguish them from other background elements. Linguistic strategies used to mark topics vary across languages. Some languages have developed syntactic mechanisms to distinguish topics from other sentence elements. To give an example, according to Vallduvi (1990, 1993, 1994), in Catalan all background elements must be detached away from the core clause. However, Vallduvi argues that topics are left-detached to a sentence-initial position while all other background elements are right-detached sentence-final slots. That is, they are syntactically separated from each other. Proposals postulating a separate syntactic location for the topic have also been made for other languages like Bulgarian (Rudin 1985), Russian (King 1993), Tz'utujil and Tzotzil (Aissen 1992). Another distinguishing property of topics is that they are associated with a special accent in some languages. For example, topics in English are associated with a LH * accent, which is called a B accent in Jackendoff's (1972) terminology. It is a common assumption that English topics require this accent irrespective of whether their sentential position (cf. Steedman 1991; Vanduvi and Zacharski 1994; Vallduvi and Engdahl 1994, 1996). It is worth noting that the topic-related pitch accent is not the same in all languages. For instance, German topics are like English ones in that they require a pitch accent, but not a LH * one. Fery (1992) observes that the topical accent may be either a raising tone (L *) or a falling tone (H *). A third strategy used by some languages to signal that a sentence element is distinguished from others as a topic is to associate it with a specific particle. The Japanese particle wa is perhaps the morphological item that has been studied the most in its relation to the marking of topics (cf. Kuno 1972, 1973; Hinds et al. 1987). It is a frequently recognized fact that this particle is used in a fairly systematic way to signal the topic of a Japanese sentence. Korean is another language that has a particle invested with the function of marking a topical element. It has been suggested that the topic and subject information are encoded, respectively, by the topic marker nun/un and the subject or nominative case marker ka/i in this language (cf. Kim 1991; Lee 1987; Choi 1995). A third language using a morphological strategy to mark the topic of the sentence is Tzotzil. In this language, topics are preceded by the topic marker a (Aissen 1992).

(9.) This is also pointed out by Erku (1984).

(10.) Transformation is a notion that belongs to generative grammars developed within a Chomskyan framework. In this framework, a division is made between two kinds of rules or formal operations: base rules and transformations. Base rules are phrase structure rules and together with rules of lexical insertion constitute the base component. The base component generates deep structures. The transformational component, which consists of rules of transformation, takes a deep structure as input and transforms it to an S-structure, which is akin to the surface structure of an expression. The actual surface form of the expression is obtained in two steps: First, stylistic rules are applied to the S-structure to generate the phonetic form. Then, the phonological component takes the phonetic form as input and yields the surface phonetic representation.

(11.) These points were suggested by one of the anonymous reviewers.

(12.) As we do not have any evidence indicating to the contrary, we assume that in Turkish sentences, the verb never leaves its original position within the clause.

(13.) Priming refers to the phenomenon whereby the processing of an expression affects the subsequent processing of another expression in a way that they will share some features. In (61) and (62), the (a) sentences are the most preferred ones. This is probably because their constituents are arranged in the same order as those of the questions which they are supposed to be answering. That is, it is likely that this preference stems from a syntactic priming effect created by the preceding utterance of the questions.

(14.) Immediate dominance schemata are simply phrase structure rules, which in effect serve as templates for permissible configurations of immediate constituency.

(15.) The exposure of these linguists' accounts of the syntactic realization of topics is based on King (1993).

References

Ahmet, Cevat (1931). Yeni Bir Grumer Metodu Hakkinda Layiha. Istanbul: Devlet Matbaasi.

Aissen, Judith (1992). Topic and focus in Mayan. Language 68, 43-80.

Banfield, Ann (1973). Narrative style and the grammar of direct and indirect speech. Foundations of Language 10, 1-39.

--(1982). Unspeakable Sentences. Boston: Routledge and Kegan Paul.

Chafe, Wallace L. (1976). Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In Subject and Topic, Charles N. Li (ed.), 25-55. New York: Academic Press.

Choi, Hye-Won (1995). Topic, focus, and strong NPs in Korean: the information partition by phrase structure and morphology. Paper presented at the Workshop on Theoretical East Asian Linguistics, University of California, Irvine.

Chomsky, Noam (1993). A minimalist program for linguistic theory. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and Samuel Jay Keyser (eds.), 1-52. Cambridge, MA: MIT Press.

--(1995). The Minimalist Program. Cambridge, MA: MIT Press.

Dahl, Osten (1974). Topic-comment structure revisited. In Topic and Comment, Contextual Boundedness and Focus, Osten Dahl (ed.), 1-24. Papers in Text Linguistics 6. Hamburg: Helmut Buske.

Danes, Frantisek (1974). Functional sentence perspective and the text organization. In Topic and Comment, Contextual Boundedness and Focus, Osten Dahl (ed.), 106-128. Papers in Text Linguistics 6. Hamburg: Helmut Buske.

Enc, Murvet (1991). The semantics of specificity. Linguistic Inquiry 22, 1-25.

Erguvanli, Eser E. (1984). The Function of Word Order in Turkish Grammar. University of California Publications in Linguistics, Vol. 106. Berkeley and Los Angeles: University of California Press.

Erku, Feride (1983). Discourse pragmatics and word order in Turkish. Doctoral dissertation, University of Minnesota.

--(1984). Discourse pragmatics and syntactic description in Turkish. In Proceedings of the Turkish Linguistics Conference, Ayhan Aksu-Koc and E. Eser Erguvanli (eds.), 179-194. Istanbul: Bogazici University Publications 400.

Erteschik-Shir, Nomi (1997). The Dynamics of Focus Structure. Cambridge: Cambridge University Press.

Fery, Caroline (1992). Focus, topic and intonation in German. Arbeitspapiere des Sonderfirschungsbereichs 340, Nr. 20, Universities of Stuttgart und Tubingen.

Gencan, Tahir Nejat (1979). Dilbilgisi (=Grammar). Ankara: Turk Dil Kurumu [the Turkish Language Institution].

Goksel, Asli (1998). Linearity, focus and the postverbal position in Turkish. In The Mainz Meeting: Proceeedings of the Seventh International Conference on Turkish Linguistics, Lars Johanson (ed.), 85-106. Wiesbaden: Harrasowitz Verlag.

--; and Ozsoy, Sumru A. (2000). Is there a focus position in Turkish? In Studies on Turkish and Turkic Languages, Asli Goksel and Celia Kerslake (eds.), 219-228. Wiesbaden: Harrassowitz Verlag.

Gundel, Jeanette (1974). The role of topic and comment in linguistics theory. Doctoral dissertation, University of Texas.

Halliday, Micheal A. K. (1967). Notes on transitivity and theme in English II. Journal of Linguistics 3, 199-244.

--(1985). An Introduction to Functional Grammar. London: Arnold.

Hinds, John; Maynard, Senko K.; and Iwasaki, Shoichi (1987). Perspectives on Topicalization: The Case of Japanese wa. Amsterdam: John Benjamins.

Hockett, Charles F. (1958). A Course in Modern Linguistics. New York: Macmillan.

Hoffman, Beryl (1995). The computational analysis of the syntax and interpretation of "free" word order in Turkish. Doctoral dissertation, University of Pennsylvania.

Horvath, Julia (1986). FOCUS in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris.

Issever, Selcuk (2001). Information structure in Turkish: the word order-prosody interface. Unpublished ms., Ankara University.

Jackendoff, Ray (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press.

Jager, Gerhard (1995). Topic, scrambling, and aktionsart. In Proceedings of the Goettingen Focus Workshop, Inga Kohlhof, Susanne Winkler and H. Bernhard Drubig (eds.), 19-34. Arbeitsbericht Nr. 69 des Sonderforschungsbereichs 340, Universitat Tubingen.

Kayne, Richard (1994). The Antisymetry of Syntax. Cambridge, MA: MIT Press.

Kennelly, Sarah D. (1997). The P-focus position in Turkish. Abstract, 20th GLOW Colloquium, Rabat.

Kilicaslan, Yilmaz (1994). Information packaging in Turkish. Unpublished M.Sc. thesis, University of Edinburgh, Edinburgh.

--(1998). A form-meaning interface for Turkish. Unpublished doctoral dissertation, University of Edinburgh, Edinburgh.

Kim, Kyu-hyun (1991). The role of the Korean topic marker in foregrounding speaker stance. Paper presented at the Seventeenth Annual Meeting of the Berkeley Linguistics Society, University of California, Berkeley, California.

King, Tracy H. (1993). Configuring topic and focus in Russian. Doctoral dissertation, Stanford Universtiy.

Kiss, Katalin E. (1981). Structural relations in Hungarian, a "free" word order language. Linguistic Inquiry 12, 185-213.

--(1995). Focus is a non-uniform phenomenon. In Proceedings of the Goettingen Focus Workshop, Inga Kohlhof, Susanne Winkler and H. Bernhard Drubig (eds.), 175-196. Arbeitsbericht Nr. 69 des Sonderforschungsbereichs 340, Universitat Tubingen.

Kornfilt, Jaklin (1984). Case marking, agreement, and empty categories in Turkish. Doctoral dissertation. Harvard University.

--(1997). Turkish. London and New York: Routledge.

--(1998). On rightward movement in Turkish. In The Mainz Meeting: Proceeedings of the Seventh International Conference on Turkish Linguistics, Lars Johanson (ed.), 107-123. Wiesbaden: Harrasowitz Verlag.

--(2000). Directionality of identical verb deletion in Turkish Coordination. In Essays in Honor of Jorge Hankamer's 60th Birthday, UCSC.

Kuno, Susumu (1972). Functional sentence perspective: a case study from Japanese and English. Linguistic Inquiry 3, 269-320.

--(1973). The Structure of the Japanese Language. Cambridge, MA: MIT Press.

Kural, Murat (1992). Properties of scrambling in Turkish. Unpublished ms., UCLA.

--(1994). Postverbal constituents in Turkish. Unpublished ms., UCLA.

--(1997). Postverbal constituents in Turkish and the linear correspondence axiom. Linguistic Inquiry 28(3), 498-519.

Lasnik, Howard (2002). The minimalist program in syntax. TRENDS in Cognitive Sciences 6(10), 432-437.

Lee, Hyo Sang (1987). Discourse presupposition and the discourse function of the topic marker nin in Korean. Indiana University Linguistics Club.

Mathesius, Vilem (1915). O passive v moderni anglictine. Sbornik filologicky, 5, 198-220.

Peregrin, Jaroslav (1995). Topic, focus, and quantification. In Proceedings of the Goettingen Focus Workshop, Inga Kohlhof, Susanne Winkler and H. Bernhard Drubig (eds.), 49-57. Arbeitsbericht Nr. 69 des Sonderforschungsbereichs 340, Universitat Tubingen.

Prince, Ellen F. (1984). Topicalization and left-dislocation: a functional analysis. In Annals of the New York Academy of Sciences 433: Discourses in Reading and Linguistics, Sheila J. White and Virginia Teller (eds.), 213-225. New York: The New York Academy of Sciences.

Puskas, Genoveva (1992). The wh-criterion in Hungarian. Rivista di Grammatica Generativa 17, 141-186.

Reinhart, Tanya (1982). Pragmatics and linguistics: an analysis of sentence topics. Philosophica 27, 53-94.

Rosengren, Inger (1995). The thetic/categorical distinction revisited once more. In Proceedings of the Goettingen Focus Workshop, Inga Kohlhof, Susanne Winkler and H. Bernhard Drubig (eds.), 89-111. Arbeitsbericht Nr. 69 des Sonderforschungsbereichs 340, Universitat Tubingen.

Rudin, Catherine (1985). Aspects of Bulgarian Syntax: Complementizers and WH Constructions. Columbus, Ohio: Slavica Publishers.

Rullman, Hotze (1995). Maximality in the semantics of wh-constructions. Doctoral dissertation, University of Massachusetts, Amherst.

Sasse, Hans-Jurgen (1987). The thetic/categorical distinction revisited. Linguistics 25, 511-580.

Sgall, Petr; Hajicova, Eva; and Panevova, Jarmila (1986). The Meaning of the Sentence in its Semantic and Pragmatic Aspects. Dordrecht: Reidel.

Steedman, Mark (1991). Structure and intonation. Language 67, 260-296.

Strawson, Peter F. (1971). Identifying reference and truth-values. In Semantics, Danny D. Steinberg and Leon A. Jakobovits (eds.), 86-99. Cambridge: Cambridge University Press.

Szabolsci, Anna; and Zwarts, Frans (1993). Weak islands and an algebraic semantics for scope taking. Natural Language Semantics 1(3), 235-284.

Underhill, Robert (1972). Turkish participles. Linguistic Inquiry 3, 87-99.

Uriagereka, Juan (2002). Derivations: Exploring the Dynamics of Syntax. London and New York: Routledge.

Valimaa-Blum, Riitta M. (1988). Finnish existential clauses: their syntax, pragmatics and intonation. Doctoral dissertation, Ohio State University.

Vallduvi, Enric (1988). Constituent focusing and Catalan right dislocation. Paper presented at the 1988 Annual Meeting of the Linguistic Society of America, New Orleans.

--(1990). The informational component. Doctoral dissertation, University of Pennsylvania.

--(1993). Information packaging: a survey. Unpublished ms., Centre for Cognitive Science and Human Communication Research Centre, University of Edinburgh, Edinburgh.

--(1994). Detachment in Catalan and information packaging. Journal of Pragmatics 22, 573-621.

--(1995). Structural properties of information packaging in Catalan. In Discourse Configurational Languages, Katalin E. Kiss (ed.), 122-152. New York: Oxford University Press.

--; and Engdahl, Elisabet (1994). Information packaging and grammar architecture. In Proceedings of the Twenty-Fifth Annual Meeting of the North Eastern Linguistics Society 25(1), Jill N. Beckman (ed.), 519-533. University of Pennsylvania.

--; and Engdahl, Elisbaet (1996). The linguistic realization of information packaging. Linguistics 34(3), 459-519.

--; and Zacharsky, Run (1994). Accenting phenomena, association with focus, and the recursiveness of focus-ground. In Proceedings of the Ninth Amsterdam Colloquium, Paul Dekker and Martin Stokhof (eds.), 683-702. ILLC/Department of Philosophy, University of Amsterdam.

Ward, Gregory L. (1985). The semantics and pragmatics of preposing. Doctoral dissertation, University of Pennsylvania.

Weinberg, Amy (2001). A minimalist theory of human sentence processing. In Working Minimalist, Samuel D. Epstein and Norbert Hornstein (eds.), 283-315. Cambridge, MA: MIT Press.

YILMAZ KILICASLAN

University of Trakya
COPYRIGHT 2004 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2004 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Kilicaslan, Yilmaz
Publication:Linguistics: an interdisciplinary journal of the language sciences
Date:Jul 1, 2004
Words:17166
Previous Article:Karmiloff, Kyra and Karmiloff-Smith, Annette: Pathways to Language: from Fetus to Adolescent.
Next Article:Combinatory conditions on Halkomelem causatives (1).
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters