Printer Friendly

Turkish suspended affixation *.

Abstract

This article presents well-formedness conditions on Turkish coordinate constructions with suspended affixation (SA) , where certain bound morphemes are omitted from all conjuncts other than the final one while maintaining their semantic scope over the whole construction. It is argued that the legitimacy of verbal conjuncts with suspended affixation neither directly falls out from the conjunct's being the complement of the copula, nor is it due to the type of agreement paradigm. Instead, the article provides a unified analysis that accounts for SA in both verbal and nonverbal constructions based on the notion of morphological words. The morphological word is comprised of a stem plus optional affixes, the right edge of which can terminate a morphological string independently from agreement markers. Accordingly, SA is licit if the omission of inflectional affixes in nonfinal conjuncts results in a morphological word. It is further shown that the terminal morphemes must be overtly marked in nonfinal conjuncts although they can be null elsewhere, and derivational morphemes cannot be suspended. Finally, it is argued that affixes that exhibit tight phonological cohesion with their stems resist suspension, suggesting that the degree of morphological bonding correlates with the degree of phonological cohesion.

1. Introduction

Since the emergence of morphological typology, several different parameters have been proposed to characterize the morphological make-up of languages. While the separative / cumulative and the variant / invariant character of morphological elements constitute the most prominent parameters of the agglutination-flexion typology, other parameters remain relatively unknown (see Plank 1999 for an overview). One such parameter concerns the characterization of morphological bonding of elements as either loose or tight. Simply, bound elements are said to be loose if they are allowed to be omitted when multiple words within the same phrase carry the same exponent. The function of a single ending as a marker for a whole phrase is usually attributed to the phrase marking, as opposed to word marking, property of the ending. Word marking endings are internal to the word that is marked. For instance, the direct and oblique case endings in Hindi are word marking while the postposition ka 'of' is phrasal (Payne 1995). In a coordinate construction where two nouns are conjoined, both conjuncts must obligatorily be marked for direct and oblique case (la). By contrast, the postposition ka can attach at the end of the conjoined phrase (lb) (examples from Payne 1995: 285-286).

(1) a. bahno aur bhaiyo (*[bahan aur bhai]-yo) sister=OblP1Fem and brother=OblP1Masc 'sisters and brothers' (oblique)

b. [Ram aur Rani] ke bhai Ram and Rani of=OblSgMasc brother=OblSgMasc 'Ram and Rani's brother' (oblique)

In the Turkish linguistics literature, this sort of a phenomenon is commonly referred to as "suspended affixation". Despite its prominence in Turkish, very little work has been carried out to explore various constraints on suspended affixation. While the phenomenon was mostly used as a tool to support a number of independent observations and theoretical arguments in individual works (e.g., Orgun 1995, 1996; Kornfilt 1996; Yu and Good 2000), there is no study that has aimed to address the issue in its entirety and investigated various factors that enable a morphological system to adapt tight or loose bonding of its exponents. Since a full-fledged investigation of the limits of suspended affixation requires an extensive analysis on the various types of morphological elements that may be involved, I have chosen to investigate the phenomenon within a single language, namely Turkish, with the hope that the data and implications will carry over to future work concerning similar phenomena in other languages.

The primary goal of this article is to examine various structural conditions on the well-formedness of coordinate constructions with suspended affixation, where certain affixes can be omitted from all conjuncts other than the final one while maintaining their semantic scope over the whole construction. I provide a unified analysis that accounts for the phenomenon in both verbal and nonverbal coordinate constructions. The analysis crucially relies on the definition of the morphological word in Turkish and employs a special characterization of bound morphemes based on their ability to surface without further affixation. I demonstrate that suspended affixation is permissible both in Turkish verbal and nonverbal constructions if the omission of inflectional affixes in nonfinal conjuncts results in material that can qualify as a morphological word.

The article is organized in the following way. First, I demonstrate various suspended affixation facts in verbal coordinate constructions and discuss where previous analyses encounter problems. Specifically, I argue that the legitimacy of verbal conjuncts with suspended affixation does not directly fall out from the conjunct's being the complement of the copula (Kornfilt 1996). Furthermore, contrary to Yu and Good's (2000) claim, I show that the type of personal agreement paradigm that the final conjunct selects is not a direct predictor of the legitimacy of suspended affixation in verbal coordinate constructions. Inquiring what can minimally and maximally be left in nonfinal verbal conjuncts, I suggest that bare conjuncts are grammatical only if they constitute morphological words. A string of morphemes acquires morphological word status if the final morpheme is allowed to terminate a word. While all nominal constructions in Turkish constitute morphological words irrespective of the type of nominal suffix they are inflected for, only a small number of verbal inflectional categories create a morphological word. Accordingly, suspended affixation is illegitimate if the nonfinal conjunct ends in a nonterminal morpheme that cannot form a morphological word. By definition, nonterminal morphemes must ultimately be followed by a suffix that can end a word. It is further demonstrated that the ungrammaticality of coordinate constructions may be due to an agreement mismatch between the interpreted subject of the nonfinal conjunct and the intended subject of the whole construction. I argue that this is due to the fact that the default interpretation of the subjects of verbal constructions without overt agreement morphology is the 3rd person singular in Turkish, which may create an agreement clash if the intended subject of the entire construction is not 3rd person singular. This is taken to suggest that Turkish verbal constructions must be terminated by agreement morphemes in order to qualify for wordhood unless they end in other terminal morphemes such as aspect markers. Furthermore, terminal elements in nonfinal conjuncts must be phonologically overt, which is evinced by the fact that the aorist must not be null when the following affixes are suspended, although the same marker can normally be null elsewhere.

The second part of the article looks at nominal coordinate constructions and demonstrates that the same generalizations that apply to verbal-predicate conjuncts also hold for nonverbal ones. It is further shown that affix suspension exhibits a split behavior between derivation and inflection. Namely, the omission of derivational morphemes on nonfinal conjuncts is strictly prohibited. Accordingly, it is suggested that with one exception (the case of subsequent plural and possessive affixes), morphological wordhood also defines the limits of suspended affixation in Noun Phrase coordination. I suggest that the puzzling restriction on the inseparability of the plural morpheme from the possessive is far more complex than the description given in Orgun (1995, 1996), and show that complications may arise due to the semantic interpretation of the bare plural nouns in the first conjunct. Finally, I demonstrate that there is yet another constraint on the suspension of certain affixes which is purely morphophonological. Namely, affixes that exhibit tight phonological cohesion with certain kinds of stems (e.g., the dative suffix that causes vowel alternations on pronouns) resist suspension. As pronouns and the dative cohere more tightly, the extent to which the dative can be suspended on pronouns becomes restricted, suggesting that phonological cohesion and morphological bonding are interdependent but closely related parameters.

2. Suspended affixation

Lewis (1967: 35) characterizes suspended affixation as when "one grammatical ending serves two or more parallel words". Consider the sentences in (2) and (3) below, where the (a) sentences illustrate conjoined structures with suspended affixation. Only the final conjunct carries bound morphemes and has scope over the nonfinal conjuncts, as opposed to the (b) sentences, where both conjuncts are inflected for the same suffixes. (1)

(2) a. Zengin ve unlu-y-du-m. (2) rich and [famous-Cop-Past-1S.sub.(k)]

b. Zengin-0-di-m ve unlu-y-du-m. rich-Cop-Past-1S and [famous-Cop-Past-1S.sub.(k)] 'I was rich and famous.'

(3) a. gid-er, gor-ur ve al-ir-0-iz. go-Aor see-Aor and [buy-Aor-Cop-1P.sub.(z)]

b. gid-er-iz, gor-ur-uz ve al-ir-0-iz. go-Aor-1P see-Aor-1P and [buy-Aor-Cop-1P.sub.(z)] 'We go (there), see (it), and buy (it).'

Kornfilt (1996) argues that grammatical instances of suspended affixation are observed depending on the type of suffix the bare conjuncts (i.e. the nonfinal conjuncts with suspended affixation) carry. Accordingly, suspended affixation is legitimate if the suffix on the bare conjunct is a participle, and thereby forming a potentially complete form, which Kornfilt refers to as a "small word". To understand Kornfilt's observations, let us first discuss her assumptions about the morphosyntax of Turkish verbal constructions (Kornfilt 1996, 1997), to which I will mostly adhere in this article. Kornfilt suggests a morphosyntactic distinction between participle ("fake") and finite ("genuine") forms. Participle forms are constructed by aspect / modality markers such as the aorist, the future, or the evidential, that are obligatorily followed by the inflected forms of the copular verb -i (see also Goksel 2001, 2003 for similar assumptions; Lewis 1967; Lees 1962; Underhill 1976: 183, for earlier treatments of the copula). According to Kornfilt (1996), the copular verb i-, which is mostly used in its clitic form in Modern Turkish, is null in the present tense. When inflected for other tense-aspect-modality (TAM) markers, the copula i- is realized as /y/ except when it occurs interconsonantally (e.g., yap-ar i-di-m (do-Aor [Cop-Past-1S.sub.(k)]) vs. yap-ar-O-di-m 'I used to do (it).'; araba-da i-di-m (car-Loc [Cop-Past-1S.sub.(k)]) vs. araba-da-y-di-m 'I was in the car.'). Participle forms acquire any further inflectional categories such as tense and agreement via the intermediary of the copula. On the contrary, the "genuine" forms, which are constructed by the definite past and the conditional, are directly inflected by further TAM markers. The split between the participles and finite forms is further evinced by the type of subject pronominal suffixes they receive, which I will simply refer to as agreement markers following Kornfilt (1996) to reduce terminological differences to a minimum. While participles receive their agreement markers from the "z-paradigm", finite forms take their agreement markers from the "k-paradigm" (see Adamovic 1985 for the development of person marking paradigms in Turkish). (3) It should be noted that nonverbal predicates also take their agreement morphology from the "z-paradigm", which, for Kornfilt, further suggests that participle forms are "fake" verbal forms.

With these observations in mind, let us take a closer look at the coordinate constructions given in (2) and (3) above, where (a) sentences illustrate grammatical instances of suspended affixation. The well-formed conjuncts are further inflected via the intermediary of the copula. Specifically, the nonverbal predicate in (2) receives tense and agreement morphology via the present variant of the copular clitic (y / 0 (= null)). Likewise, the examples in (3) above contain an aspect marker, which forms a participle (rather than a finite tensed verb). This form also receives the agreement morphology from the "z-paradigm" via the copula. The "k-paradigm" and "z-paradigm" are illustrated in Table 1 below.

When bare conjuncts are finite forms, however, suspended affixation fails (4a). Furthermore, although the right conjunct is a participle form in (5a), the suspension of the aorist and other inflectional markers is ill-formed since the left conjunct is not a "complete form" in the sense that it does not have a suffix that can head a participle form.

(4) a. *Cahs-ti ve basar-di-k. work-Past and [succeed-Past-1P.sub.(k)]

b. Calis-ti-k ve basar-di-k. work-Past-1P and [succeed-Past-1P.sub.(k)] 'We worked and succeeded.'

(5) a. *Calis ve basar-ir-0-iz. work and [succeed-Aor-Cop-1P.sub.(z)]

b. Calis-ir-iz ve basar-ir-iz. work-Aor-1P and [succeed-Aor-1P.sub.(z)]

c. Calis-ir ve basar-ir-iz. work-Aor and [succeed-Aor-1P.sub.(z)] 'We work and succeed.'

Suspended affixation also takes place in complex verbal forms. Once again, only the complements of the copula, i.e., the participle forms, can form a legitimate conjunct with suspended affixation ([6], [7]).

(6) a. Calis-ir ve basar-ir-0-di-k. work-Aor and [succeed-Aor-Cop-Past-1P.sub.(k)]

b. Calis-ir-0-di-k ve basar-ir-0-di-k. [work-Aor-Cop-Past-1P.sub.(k)] and [succeed-Aor-Cop-Past-1P.sub.(k)] 'We used to work and succeed.'

(7) a. Calis-acak ve basar-acak-0-ti-k. work-Fut and [succeed-Fut-Cop-Past-1P.sub.(k)]

b. Calis-acak-0-ti-k ve basar-acak-0-ti-k [work-Fut-Cop-Past-1P.sub.(k)] and [succeed-Fut-Cop-Past-1P.sub.(k)] 'We were going to work and succeed.'

What seems to be crucial here is the kind of suffix that lodges at the right edge of the nonfinal conjunct. "If that suffix is of a type that cannot head a participle, but rather is a suffix that forms a genuine finite verb, then suspended affixation cannot take place" (Kornfilt 1996: 110). Since the conjuncts in the (a) sentences in ([6]-[7]) above contain a suffix that ends a participle form as a complement of the copula, suspended affixation is legitimate. Thus, Kornfilt (1996) generalizes the grammatical instances of suspended affixation to be an instance of coordination of participles with the inflected copula cliticized to the coordinate structure. In other words, suspended affixation is fully grammatical only if the bare conjunct is a "small word". Kornfilt (1996: 111) defines a "small word" as "a potentially complete form which can be the complement of the copula". Accordingly, "the only forms that can 'suspend' their inflections are forms that never receive those inflections directly anyway, because those inflections actually attach to the copula and not to the main verb" (Kornfilt 1996:111). Therefore, Kornfilt (1996:111-112) suggests that:

... the grammatical instances of suspended affixation in copular constructions and constructions involving verbal elements that appear to be finite main verbs [are] a coordination of adjectives or participles, with the inflected copula cliticized to the coordinate structure. In other words, suspended affixation is nothing else but the cliticization of the inflected copula to just those elements that it is allowed to cliticize in general, with the only difference that these complements are conjoined.

Kornfilt's observations on suspended affixation are formulized in (8) below.

(8) [[V.sub.participle] and [V.sub.participle]] [V.sub.Copula] + Inflectional Morphemes

An empirical question arises here as to whether suspended affixation is also legitimate with all kinds of affixes that can potentially be the complements of the copula. Furthermore, the notion of "small word" introduces questions pertaining to word size. Analogous to a "small word", could there be conjuncts that are too "big"? What would be the maximum and the minimum size of a conjunct with suspended affixation? The next section attempts to answer these questions.

3. Maximality and minimality in suspended affixation

The issue of word size in Turkish has received very little attention in the literature (but see Goksel 2001, 2003 from a morphosyntactic viewpoint; and Hankamer 1989 from the perspective of language processing). What is the maximum number of bound morphemes a verb can receive in Turkish? Figure 1, adapted from Goksel (2001), shows the schema for possible morpheme slots available for verbs (excluding derivational morphemes that may immediately follow the verb root) in main clauses. (4) Naturally, not all of the morphemes given in Figure 1 can co-occur with one another due to morphological (see Goksel 2001) and semantic constraints, which lie beyond the scope of this article. (5)

Along similar lines as Kornfilt's (1996) observations, Goksel (2001) argues that the second set of TAM markers (those in Slot 4) are actually inflected on the copular verb i- or its bound variants. Goksel gives a convincing account of the issue of word size in Turkish main verbs with reference to the maximum number of slots that can be realized without having to resort to a morphological buffer such as auxiliary verbs for further inflection. However, the classification of morphemes into slots as in Figure 1 only assumes the linear lineup of morphological elements, which is not sufficient to make a principled distinction between those markers that can stay in the nonfinal conjunct and those that cannot. This is primarily because, as far as morphotactic constraints are concerned, both the so-called "genuine" forms and the "fake" forms occupy the same morpheme slot, namely Slot 2. Since all Slot 2 markers are capable of being further inflected via the intermediary of the copula, Kornfilt's characterization of suspended affixation summarized in (8) above predicts that those conjuncts that end with any of the Slot 2 markers should yield well-formed instances of suspended affixation. However, the constructions with suspended affixation given in (9a) and (9c) are ill-formed although the past tense and the conditional markers are further inflected via the intermediary of the copula. That the copula follows these markers is indeed proven by the fact that the full (i.e., nonsuspended) forms of coordinate constructions given in (9b) and (9d) exhibit the overt marking of the inflected copula (-y). (6)

(9) a. * O yaz Avsa-ya git-ti ve deniz-e that summer Avsa-Dat go-Past and sea-Dat gir-di-y-di-k. [enter-Past-Cop-Past-1P.sub.(k)]

b. O yaz Avsa-ya git-ti-y-di-k ve deniz-e that sum. Avsa-Dat [go-Past-Cop-Past-1P.sub.(k)] and sea-Dat gir-di-y-di-k. [enter-Past-Cop-Past-1P.sub.(k)] 'That summer, we went to Avsa and swam in the sea.'

c. * Ev-imiz-i sat-sa ve bir dukkan house-Poss.1P-Acc sell-Cond and a shop al-sa-y-di-k, (iyi olurdu). [buy-Cond-Cop-Past-1P.sub.(k)]

d. Ev-imiz-i sat-sa-y-di-k ve house-Poss.1P-Acc [sell-Cond-Cop-Past-1P.sub.(k)] and bir dukkan al-sa-y-di-k, (iyi olurdu). a shop [buy-Cond-Cop-Past-1P.sub.(k)] '(It would be good) if we sold our house and bought a shop.'

Here it could be argued that the affix suspension in (9) is illicit because the left conjuncts in (9a) and (9c) end in a "genuine" verb marker (past and conditional, respectively). However, disambiguating participlehood from the copula criterion would not sufficiently account for the facts since not all Slot 2 morphemes qualify for participlehood according to Kornfilt's account. This is primarily because there is at least one morpheme, namely -mElI (necessitative), that can legitimately occur at the end of a conjunct with suspended affixation (10) although the morpheme tends to behave as a "genuine" form a 1a Kornfilt on several grounds.

(10) a. Calis-mali ve basar-mali-0-sin. work-Nec and [succeed-Nec-Cop-2S.sub.(z)]

b. Calis-mali-y-iz ve basar-mali-y-iz. [Work-Nec-Cop-1P.sub.(z)] and [succeed-Nec-Cop-1P.sub.(z)] 'We must work and succeed.'

If we take Kornfilt's (1996: 104-109) own assumptions to distinguish between genuine verbal forms from participle forms, those forms terminated by the necessitative marker tend to behave unlike participle forms (those that end in progressive, aorist, past participle, future markers) but more like "genuine" forms. First, such forms cannot be used as modifying participles in NPs (11). Second, they are not allowed to take the copular negation element, but the truly verbal one (12). Fourth, they cannot have the epistemological copula clitic (13).

(11) a. *calis-mali adam work-Nec man Intended reading: 'The man who must work.'

b. *kitab-i oku-mali adam book-Acc read.Nec man Intended reading: 'The man who must read the book.'

(12) a. *calis-mali degil-sin. Work-Nec [Neg-2S.sub.(z)] Intended reading: 'It is not the case that you must work.'

b. calis-ma-mali-0-sin. [work-Neg-Nec-Cop-2S.sub.(z)]

(13) a. */?calis-mali-dir-sin. (Variety A) (7) [work-Nec-EpCop-2S.sub.(z)] Intended reading: 'You must definitely work.'

b. */?Calis-mali-sin-dir (Variety B). [work-Nec-2S.sub.(z)]-EpCop Intended reading: 'You must definitely work.'

In so far as the placement of the question particle and the type of the following agreement marker are concerned, however, the very same marker seem to pattern with participles. In particular, the question particle appears immediately after the necessitative marker and crucially before an agreement marker from the z-paradigm (e.g., git-meli-mi-sin? ([go-Nec-Que-2S.sub.(z)]) 'Must you go?'). It could therefore be suggested that participlehood / finiteness may be a gradient property rather than discrete. As such, accounts that crucially rely on the distinction between finite forms and participle forms cannot provide sufficient grounds on which suspended affixation can be adequately explained.

In sum, we have observed two distinct sets of evidence working against (8). First, we have seen that conjuncts with suspended affixation can be ungrammatical despite the fact that the complements of the copula are conjoined. Second, there is at least one form that behaves as a genuine marker on several grounds but allows the suspension of affixes that follow it. Since participlehood is inexplicitly equated with the state of being able to receive inflection via the intermediary of the copula, Kornfilt's account fails to explain whether suspended affixation is a matter of: (i) omitting those morphemes that actually attach to the copula rather than to the verb stem, or (ii) leaving morphological material in the nonfinal conjunct that is potentially a complete form.

Yu and Good (2000) (8) provide a partial account of suspended affixation with respect to the type of agreement marker that heads the whole conjunct. Based on their phonological properties and a number of other diagnostic conditions provided in Zwicky and Pullum (1983), Yu and Good argue that the agreement markers from the z-paradigm are postlexical clitics compared to those from the k-paradigm, which are lexical suffixes. Accordingly, the suspension of an agreement marker is legal if it is from the z-paradigm. This explains why coordinate constructions with suspended affixation such as those given in (9) above are ill-formed. In particular, since the agreement markers on the final conjuncts are from the k-paradigm, they cannot be suspended. Those from the z-paradigm, however, can have scope over all the conjuncts (cf. example [10a] above). Although Yu and Good offer a convincing explanation based on a systematic distinction between the agreement paradigms, their analysis can only account for the inseparability of the past tense and the conditional marker from the personal agreement markers of the k-paradigm. However, if we generalize from Yu and Good's assumption that only those markers from the z-paradigm are allowed to be suspended in nonfinal conjuncts, we would make incorrect predictions. The problem is again related to the recursive nature of Turkish morphology. In particular, agreement markers from the k-paradigm can inflect verbs that contain aspect markers (i.e., those that require the z-paradigm) in order to form complex tenses. For instance, the examples in (14) and (15) are grammatical despite the fact that the conjuncts have agreement morphology from the k-paradigm. To the contrary, (16), (17) and (18), where the left conjunct ends in the question particle -mI, the negative morpheme -mE, and a bare verb root, respectively, are ungrammatical although the subject pronominal ending required in the full form of the left conjunct is from the z-paradigm.

(14) a. Calis-iyor ve kazan-iyor-du-k. work-Prog and [earn-Prog-Past-1P.sub.(k)]

b. Calis-iyor-du-k ve kazan-iyor-du-k. [work-Prog-Past-1P.sub.(k)] and [earn-Prog-Past-1P.sub.(k)] 'We were working and earning (e.g., money).'

(15) a. Calis-ir ve kazan-ir-sa-niz. (9) work-Aor and [earn-Aor-Cond-2P.sub.(k)]

b. Calis-ir-sa-niz ve kazan-ir-sa-niz. [work-Aor-Cond-2P.sub.(k)] and [earn-Aor-Cond-2P.sub.(k)] 'If you (pl.) work and earn (e.g., money).'

(16) a. *Calis-iyor-mu ve kalkin-iyor-mu-y-uz? (10) work-Prog-Que and pdevelop-Prog-Que-Cop-1P.sub.(z)]

b. Calis-iyor-mu-yuz ve kalkin-iyor-mu-y-uz? [work-Prog-Que-1P.sub.(z)] and [develop-Prog-Que-Cop-1P.sub.(z)] 'Are we working and developing?'

(17) a. *Calis-ma ve kalkin-ma-yacak-0-sin. work-Neg and [develop-Neg-Fut-Cop-2S.sub.(z)]

b. Calis-ma-yacak-0-sin ve kalkin-ma-yacak-0-sin. [work-Neg-Fut-Cop-2S.sub.(z)] and [develop-Neg-Fut-Cop-2S.sub.(z)] 'You (sg.) are not going to work and develop.'

(18) a. *Calis ve kalkin-ir-0-siniz. work and [develop-Aor-Cop-2P.sub.(z)]

b. Calis-ir-0-siniz ve kalkin-ir-0-siniz. [Work-Aor-Cop-2P.sub.(z)] and [develop-Aor-Cop-2P.sub.(z)] 'You (pl.) will work and develop.'

We observe here that the suspension of affixes leaves either too much (as in [16] and [17]), or too little (as in [18]) material in nonfinal conjuncts. We can systematically explore what the minimum and the maximum suspended material could be by combining two forms, namely (19a) and (19b), which carry several derivational and inflectional morphemes, and by suspending each affix at a time on the left conjunct. For the sake of illustration, I will only use the aorist marker although other aspect markers can also be used in the same position, yielding the same results.

(19) a. iyi-les-tir-il-ebil-ir-mi-y-di-k? [good-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)] 'Could we be able to be cured?'

b. uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)] 'Could we be able to be domesticated?'

(20) Intended meaning: 'Could we be cured and domesticated?' a. *iyi-les-tir-il-ebil-ir-mi-y-di ve good-Der-Caus-Pass-Pot-Aor-Que-Cop-Past and uysal-las-tir-ti-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

b. *iyi-les-tir-il-ebil-ir-mi-y ve good-Der-Caus-Pass-Pot-Aor-Que-Cop and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

c. *iyi-les-tir-il-ebil-ir-mi ve good.Der-Caus-Pass-Pot-Aor-Que and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

d. iyi-les-tir-il-ebil-ir ve good-Der-Caus-Pass-Pot-Aor and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

e. *iyi-les-tir-il-ebil ve good-Der-Caus-Pass-Pot and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

f. *iyi-les-tir ve good-Der-Caus and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

g. *iyi-les ve good-Der and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

h. *iyi ve good and uysal-las-tir-il-abil-ir-mi-y-di-k? [docile-Der-Caus-Pass-Pot-Aor-Que-Cop-Past-1P.sub.(k)]

Strikingly, among all possibilities, only (20d) is grammatical. This suggests that verbal conjuncts with suspended affixation are legitimate if and only if they end in aspect / modality markers, nothing more or less. It should be noted that Kornfilt's (1996) account also predicts (20d) to be grammatical. However, what remains unknown in her account is why those in (20a), (20b), and (20c) are ill-formed although they are all followed by the copula. Likewise, expanding on Yu and Good's characterization of Turkish affixes as phrasal vs. lexical suffixes, why can't postlexical clitics (or phrasal affixes) such as the question marker (unless -mI carries a disjunctive function--see Note 10), as illustrated in (21), be left in non-final conjuncts?

(21) a. *Calis-ir-mi ve basar-ir-mi-0-sin? work-Aor-Que and [succeed-Aor-Que-Cop-2S.sub.(z)]

b. Calis-ir-mi-0-sin ve basar-ir-mi-0-sin? Work-Aor-Que-Cop-2S and [succeed-Aor-Que-Cop-2S.sub.(z)]. 'Do you work and do you succeed?'

An alternative analysis based on the notion of clitic clusters can be proposed to account for the above facts. Typologically, clitic clusters cannot be split, being positioned with respect to a host as a single unit. This analysis would essentially require the classification of certain bound morphemes as clitics or phrasal affixes in Turkish. Accordingly, bound morphemes following aspect / modality markers, namely the copula and the agreement markers, as well as truly phrasal affixes such as the epistemological copula, the topic marker, and the question marker would need to be regarded as clitics. Furthermore, in order to account for the ungrammaticality of their separation from the following "clitics", this approach must crucially consider the past tense morpheme and the conditional (the genuine forms) as part of the clitic cluster while aspect / modality markers would need to be classified as suffixes. These assumptions, however, do not find any support in Turkish. Furthermore, they go against commonly held assumptions about Turkish morphology on at least three grounds. First, the treatment of the past tense and the conditional as clitics on the one hand, and aspect / modality markers as genuine suffixes on the other hand is false since they all truly behave as suffixes in Turkish with respect to their phonological properties. In particular, they receive stress, as opposed to clitics which do not fall within the stress assignment domain (e.g., Kabak and Vogel 2001). Second, just as agreement markers from the k-paradigm, those from the z-paradigm (cf. examples [16] and [17] above) may not be omitted due to the recursive nature of Turkish morphology. Consequently, this analysis would then have to assume that all agreement markers are clitics. However, as discussed above in the context of Yu and Good's (2000) analysis (see also Erdal 2000), the crucial distinction between the agreement markers from the z-paradigm and those from the k-paradigm is one of phrasal affixhood (clitichood). That is, agreement markers from the z-paradigm are argued to be phrasal affixes historically developed from pronouns, compared to the k-paradigm agreement markers, which are true affixes. Last but not the least, an analysis based on clitic clusters does not shed light on why bound elements such as the passive, causative, and abilitative that appear before aspect / modality markers cannot also be split (cf. the examples [20e]-[20h] above). To account for the ungrammaticality of (20e)-(20f), this analysis must stipulate that such bound morphemes also form clitic clusters, although they absolutely do not behave as clitics on any ground.

In sum, an analysis based on clitic clusters is unmotivated for Turkish and thus it cannot be maintained particularly because it needs to resort to an incorrect and unmotivated classification of bound morphemes into clitics and affixes. Below, I will propose an account of suspended affixation based on the notion of morphological words in Turkish. It should be noted that the present account essentially extends Kornfilt's observations on "small word"-hood to the notion of morphological words. Instead of alluding to notions such as "being a complement of the copula" and "participlehood", or banning particular affixes from suspension, the present account will establish the relevant unit of analysis as morphological words.

4. Proposal

The current proposal relies on the definition of the morphological word in Turkish with a characterization of morphemes as being terminal and non-terminal based on their ability to end a word. Accordingly, the suspension of affixes in a conjunct is legitimate if that conjuncts ends in terminal morphemes. Here, I adopt the Korean linguistics tradition in the classification of affixes, where a distinction between terminal (sentence enders) and nonterminal suffixes is made. Korean sentence enders (e.g., -sup-ni-ta (deferential declarative ender)), conjunctive enders (e.g., -myense 'while'), and relative clause enders (e.g., -nu-n (nonpast indicative)) must be present at the end of a sentence or a clause in order for a verb or an adjective to stand independently (e.g., Sohn 1999: 232). Nonterminal inflectional affixes, on the other hand, are optional in their occurrence before the enders. In the context of Turkish verbal inflectional morphology, the definition of a terminal suffix is given below:

Terminal suffix: A suffix that is allowed to appear at the end of a word, where further suffixation is not obligatory.

I argue that agreement markers are ultimate word enders in Turkish. As such, verbal constructions without agreement markers do not qualify as morphological words unless they are followed by other terminal suffixes. Accordingly, all aspect / modality markers constitute terminal elements since they are not directly inflected for agreement. This is further evinced by the placement of the question particle -mI. With "genuine" markers such as the simple past, the question particle must strictly follow the agreement markers, suggesting that the relationship between these markers and agreement markers is tight (22a). However, the same particle must obligatorily be inserted between the aspect / modality and the agreement markers (22b).

(22) a. Siz opera-ya git-ti-niz-mi? You (pl.) opera-Dat go-Past-[2P.sub.(k)]-Que 'Did you go to the opera?'

b. Siz opera-ya gid-er-mi-siniz? You (pl.) opera-Dat go-Aor-Que-[2P.sub.(z)] 'Do you go to the opera?'

All other verbal bound elements are nonterminal since they necessitate the presence of a person agreement marker within the same word. Furthermore, bare verbs (e.g., git '[to] go', uyu '[to] sleep') are not morphological words because verbal forms must ultimately be followed by subject pronominal markers (or receive derivational morphemes) to qualify for a morphological word status. It should be noted that some imperative forms which have the phonological shape of a bare verb root (e.g., git! 'go!', kos! 'run!') indeed contain a phonologically null pronominal marker for the 2nd person singular because imepartive forms for other persons are overtly expressed (e.g., git-sin 'let him/her go!'; gid-in 'you (polite) go!'; gid-iniz 'you (pl./polite) go!'). If one follows a realizational approach (e.g., Stump 1998, 2001), this can be taken to suggest that the bare form fills a cell in the inflectional paradigm realizing the 2nd person singular agreement feature. As such, Turkish bare verbal roots never occur in isolation without a relevant agreement feature. Accordingly, it should be assumed that it is indeed the features that are terminal, irrespective of their concrete realization. (11) A list of verbal terminal morphemes is provided in Table 2. (12)

Table 2 only lists those inflectional morphemes that are crucially verbal (i.e., those that do not change the category of the verb). Therefore, deverbalizing suffixes such as -mEk, -Is, which can stand without personal agreement markers are not considered here since these markers turn verbal constructions into nominal constructions. As we will see in Section 6, all Noun Phrases constitute morphological words; therefore any nominalizing suffix can potentially end a nonfinal conjunct ([23a], [23b]). It must be pointed out that the omission of such suffixes from nonfinal conjuncts, however, is a different matter. When such elements are omitted, the nonfinal conjunct no longer qualifies as a morphological word, rendering the coordinate construction ungrammatical ([23c], [23d]).

(23) a. Resim yap-mak ve sarki soyle-mek-ten nefret picture do-Dvb and song sing-Dvb-Abl hate ed-er-im. Aux-Aor-[1S.sub.(z)] 'I hate drawing and singing.'

b. SSCB-nin kur-ul-us ve dagil-is-i-ni USSR-Gen assemble-Pass-Dvb and collapse-Dvb-Poss.3S-Acc 'The establishment and collapse of the Union of Soviet Socialist Republics.'

c. * Resim yap ve sarki soyle-mek-ten nefret ed-er-im. picture do and song sing-Dvb-Abl hate Aux-Aor-[1S.sub.(z)] Intended meaning: 'I hate drawing and singing.'

d. * SSCB-nin kur-ul ve dagil-is-i-ni USSR-Genassemble-Pass and collapse-Dvb-Poss.3S-Acc Intended meaning: 'The establishment and collapse of the Union of Soviet Socialist Republics.'

It should be pointed out that although the classification of affixes given in Table 2 above automatically regards cliticlike particles (or phrasal affixes), such as the question particle, epistemological copula, and the topic marker, as nonterminal, they can legitimately end an independent clause (24).

(24) a. kos-tu-n-mu? run-Past-[2P.sub.(k)]-Que 'Did you run?'

b. var-mis-iz-dir. (cf. var-mis-tir-iz (Variety A)) go-Evid-[1P.sub.(z)]-EpCop. 'It could be that we arrived (there).'

c. agla-mis-sin-da. cry-Evid-[2S.sub.(z)]-Top 'It is also that you cried (I heard).'

d. oyna-r-0-da play-Aor-(3S)-Top. 'He can also play/dance.'

Here it could be suggested that since clitics are not true affixes and usually they are considered to be phrasal in nature, the particles above cannot be expected to create word forms or stems since they must attach to already created morphological words at the postlexical level. Indeed, all the phrases in (24) carry a person marker, phonologically overt or not, suggesting that these particles attach to already established morphological words. Thus, such elements should not be expected to affect the wordhood of the forms that they adjoin to. It should be remembered that when nonfinal conjuncts end in the question particle -mI (again with the exception of -mI used as a disjunctive marker), suspended affixation is illegitimate (cf. [21] above). As discussed above, phrasal affixes (except for the question particle in participle forms and the epistemological copula that may precede agreement morphemes in Variety A) always come at the end, crucially after agreement markers. Therefore, the mere omission of such particles in the nonfinal conjunct does not render forms ungrammatical since the nonfinal conjunct is a complete morphological word with an agreement marker at the end (25a)-(25c). Notice, however, that the question particle and the epistemological copula -dIr cannot end nonfinal conjuncts (25d)-(25e).

(25) a. gel-mis-sin ve gor-mus-sun-dur. come-Evid-[2S.sub.(z)] and see-Evid-[2S.sub.(z)]-EpCop 'You must have definitely come here and seen (it).'

b. oyna-yacak-sin ve ic-ecek-sin-de. dance-Fut-[2S.sub.(z)] and drink-Fut-[2S.sub.(z)]-Top/Conn 'Also, you will dance and drink.'

c. gel-di-n ve gor-du-n-mu? come-Past-[2S.sub.(k)] and see-Past-[2S.sub.(k)]-Que 'Did you come and see (it)?'

d. * gel-mis-tir ve gor-mus-tur-sun come-Evid-EpCop and see-Evid-EpCop-[2S.sub.(z)] Intended meaning: 'You must have definitely come (here) and seen (it).'

e. * oyna-yacak-mi ve ic-ecek-mi-sin? dance-Fut-Que and drink-Fut-Que-[2S.sub.(z)] Intended meaning: 'Are you going to dance and drink?'

How can one explain why Turkish speakers do not accept the constructions in (25d) and (25e), where the epistemological copula (that precedes an agreement morpheme in Variety A) and the question particle head the nonfinal conjunct? I suggest that due to their phrasal nature, such particles always attach to phrases that are presumably complete words. Hence, the presence of such particles is rather unexpected in the absence of an agreement morpheme. When we closely investigate these ungrammatical instances of suspended affixation together with those that are headed by the past tense and conditional morphemes such as in (9), as well as the question particle such as in (16), we indeed find empirical support for the notion and status of morphological words in Turkish with crucial reference to agreement morphology. For convenience, (9) and (16) above are repeated below.

(9') a. * O yaz Avsa-ya git-ti ve deniz-e that summer Avsa-Dat go-Past and sea-Dat gir-di-y-di-k. enter-Past-Cop-Past-1P

b. O yaz Avsa-ya git-ti-y-di-k ve deniz-e that sum. Avsa.-Dat go-Past-Cop-Past-[1P.sub.(k)] and sea-Dat gir-di-y-di-k enter-Past-Cop-Past-[1P.sub.(k)] 'That summer, we went to Avsa and swam in the sea.'

(16') a. * Calis-iyor-mu ve kalkin-iyor-mu-y-uz? work-Prog-Que and develop-Prog-Que-Cop-[1P.sub.(z)]

b. Calis-iyor-mu-yuz ve kalkin-iyor-mu-y-uz? work-Prog-Que-[1P.sub.(z)] and develop-Prog-Que-Cop-[1P.sub.(z)] 'Are we working and developing?'

My informants consistently reported that they interpret the nonfinal conjuncts in the above constructions as denoting a 3rd person. This is in fact not surprising given the fact that the absence of an overt person agreement marker in a verbal construction with tense or aspect morphology almost always corresponds to the 3rd person singular in Turkish. Accordingly, the absence of an agreement marker in the nonfinal conjunct creates an agreement mismatch when the final conjunct is inflected for any person other than the 3rd person singular. That is, the subject of the nonfinal conjunct is automatically interpreted as the 3rd person singular which then yields an infelicitous interpretation together with the underlying subject of the final conjunct. This is illustrated in (26a) in a parallel construction to (9). When the intended subject pronoun of the nonfinal conjunct (overt/dropped) is indeed the 3rd person singular (26b), or the final conjunct is inflected for the 3rd person singular (26c), however, we have legitimate coordinate constructions.

(26) a. * [(Biz) kazan-di ve harca-di-k.] [(We) earn-Past-[3S.sub.(default)] and spend-Past-[1P.sub.(k)]] Intended meaning: 'We worked and spent (it).'

b. [(O) kazan-di] ve [(biz) harca-di-k.] (s/[he.sub.(default)]) earn-Past-[3S.sub.(default)]] and [(we) spend-Past-[1P.sub.(k)]] Intended meaning: 's/he earned and we spent (it).'

c. [(O) kazan-di] ve [(o) [(s/[he.sub.(default)]) earn-Past-[3S.sub.(default)]] and [(s/[he.sub.(default)]) harca-di.] spend-Past-[3S.sub.(default)] Intended meaning: 's/he earned and spent (it).'

Likewise, the presence of the question particle in the first conjunct given in (27) below renders the construction ungrammatical since the default interpretation of subject agreement is again the 3rd person singular.

(27) * [(O) calis-ir-mi] ve [(s/[he.sub.(default)]) work-Aor-Que-[3S.sub.(default)]] and [(sen) basar-ir-mi-0-sin]? [(you) succeed-Aor-Que-Cop2S] Intended meaning: 'Do you work and succeed?' Interpretation: 'Does s/he work and you succeed?'

Furthermore, with conjuncts that have the shape of a bare verb root such as in (18) above (i.e., calis 'work'), or a verb root followed by the negative morpheme as in (17) above (i.e., calis-ma (work-Neg)), the same informants interpret the first conjunct as an imperative form for the 2nd person singular (i.e., calis 'work!', calis-ma 'don't work!', respectively), where the agreement marker is phonologically null. This once again suggests that Turkish speakers need to recuperate a person agreement marker when processing verbal constructions as morphological words. In sum, we find legitimate grounds on which we can argue that nonfinal conjuncts must strictly end in overt terminal markers. (13) As such, the definition of a morphological word in verbal constructions is straightforward: a verb plus optional affixes that may never appear without terminal suffixes. (14) In the following, I will demonstrate that there is indeed yet another piece of evidence from aorist morphology that suggests that terminal elements must be overtly marked in nonfinal conjuncts although the same morphological feature is ordinarily null elsewhere.

5. The aorist must be phonologically overt in nonfinal conjuncts

The terminal element under investigation is the aorist marker, (/-(I)r/ or /-(E)r/), the various functions of which are not crucial for the purposes of this article. The aorist roughly corresponds to the simple present tense in English, and it typically marks the habitual aspect in Turkish (see Underhill 1976: 145). It should be remembered that, as a participle form, the aorist takes the copula for further inflection, as illustrated in (28).

(28) a. yuz-er-0-iz / yuz-er-0-di-k. swim-Aor-Cop-[1P.sub.(z)] / swim-Aor-Cop-Past-[1P.sub.(k)] 'I swim. / I used to swim.'

b. yuz-er-0-sin / yuz-er-0-se-n. swim-Aor-Cop-[2S.sub.(z)] / swim-Aor-Cop-Cond-[2S.sub.(k)] 'You (sg.) swim. / if you (sg.) swim.'

c. yuz-er-0-iz / yuz-er-0-mis-iz. swim-Aor-Cop-[1P.sub.(z)] / swim-Aor-Cop-RepPast-[1P.sub.(z)] 'We swim. / We used to swim. (hearsay)'

With some person markers, the aorist does not phonologically surface if it follows the negative marker (-mE). This crucially takes place only if the 1st person singular or the 1st person plural agreement marker immediately follows the aorist, as shown in (29a) and (29b). In all other cases, such as with the negative or the question marker, the aorist marker takes the shape of -z ([29c], [29d], [29e]).

(29) a. yuz-me-0-0-m. swim-Neg-Aor-Cop-[1S.sub.(z)] 'I don't swim.'

b. yuz-me-0-0-yiz. swim-Neg-Aor-Cop-[1P.sub.(z)]. 'We don't swim.'

c. yuz-me-z-0-sin. swim-Neg-Aor-Cop-[2S.sub.(z)] 'You (sg.) don't swim.'

d. yuz-me-z-0-1er. swim-Neg-Aor-Cop-[3P.sub.(z)] 'They (pl.) don't swim.'

e. yuz-me-z-0-di-m. swim-Neg-Aor-Cop-Past-[1S.sub.(k)] 'I didn't use to swim.'

Essentially, the aorist has two allomorphs: (i) -z, which follows the negation marker -mE (e.g., Underhill 1976: 146-147), and (ii) -(I)r/-(E)r elsewhere. Just like other aspect markers, the aorist also takes the agreement markers from the z-paradigm unless "genuine" markers follow it. (15) In the examples provided below, the aorist morpheme is not phonologically realized when it is followed by the 1st person singular marker (30a). The same marker, however, appears as -z when it is in the past tense (30b). When -z is omitted in (30c), the meaning of the sentence changes from the habitual past to the simple past. That -z adds an additional meaning to the predicate strongly suggests that it must be considered as one of the surface variants of the aorist.

(30) a. oyna-ma-0-0-m. play-Neg-Aor-Cop-[1S.sub.(z)] 'I don't play.'

b. oyna-ma-z-0-dl-m. play-Neg-Aor-Cop-Past-[1S.sub.(k)] 'I didn't use to play.'

c. oyna-ma-di-m. play-Neg-Past-[1S.sub.(k)] 'I didn't play.'

Let us consider the following examples of suspended affixation, which involve the aorist and the negative morpheme. While the conjuncts inflected with the 2nd person and 3rd person (plural) are grammatical ([31b], [32b]), suspended affixation becomes illicit with the 1st person ([33b], [34b]).

(31) a. Kumsal-a git-me-z-0-sin ve top beach-Dat go-Neg-Aor-Cop-[2S.sub.(z)] and ball oyna-ma-z-0-sin. play-Neg-Aor-Cop-2S/z).

b. Kumsal-a git-me-z ve top beach-Dat go-Neg-Aor and volleyball oyn-ma-z-0-sin. play-Neg-Aor-Cop-[2S.sub.(z)]. 'You (sg.) don't go to the beach and play ball.'

(32) a. Kumsal-a git-me-z-0-ler ve top beach-Dat go-Neg-Aor-Cop-[3P.sub.(z)] and ball oyna-ma-z-0-1ar. play-Neg-Aor-Cop-[3P.sub.(z)].

b. Kumsal-a git-me-z ve top oyna-ma-z-0-1ar beach-Dat go-Neg-Aor and ball play-Neg-Aor-Cop-[3P.sub.(z)]. 'They don't go to the beach and play ball.'

(33) a. Kumsal-a git-me-0-0-m ve top beach-Dat go-Neg-Aor-Cop-[1S.sub.(z)] and ball oyna-ma-0-0-m. play-Neg-Aor-Cop-[1S.sub.(z)].

b. * Kumsal-a git-me-0 ve top beach-Dat go-Neg-Aor and ball oyna-ma-0-0-m. play-Neg-Aor-Cop-[1S.sub.(z)]. Intended meaning: 'I don't go to the beach and play ball.'

(34) a. Kumsal-a git-me-0-0-yiz ve top beach-Dat go-Neg-Aor-Cop-[1P.sub.(z)] and ball oyna-ma-0-0-yiz. play-Neg-Aor-Cop-[1P.sub.(z)].

b. * Kumsal-a git-me-0 ve top oyna-ma-0-0-ylz beach-Dat go-Neg-Aor and ball play-Neg-Aor-Cop-[1P.sub.(z)]. Intended meaning: 'We don't go to the beach and play ball.'

The mere presence of -z in (35) below renders the same sentence given in (34b) grammatical. Notice that -z would be ungrammatical if it co-occurred with the 1st person plural marker (cf. * oyna-ma-z-iz 'we don't play').

(35) Kumsal-a git-me-z ve top oyna-ma-0-0-yiz. Beach-Dat go-Neg-Aor and ball play-Neg-Aor-Cop-[1P.sub.(z)]. 'We don't go to the beach and play ball.'

It should be noted that the same is not true for the 1st person singular, as all my informants consistently judged (35) to be acceptable but (36) to be somewhat awkward.

(36) ?Kumsal-a git-me-z ve top oyna-ma-0-0-m beach-Dat go-Neg-Aor and ball play-Neg-Aor-Cop-[1S.sub.(z)]. Intended meaning: 'I don't go to the beach and play ball.'

There seems to be no apparent reason why the 1st person plural is better with /z/. It could, however, be suggested that the phonological similarity of the aorist with the agreement marker (/-(I)z/) rescues suspended affixation, adding an interesting dimension to the way phonology interacts with morphosyntax. In sum, the above examples suggest that the phonological presence of -z serves for a crucial purpose in rendering affix suspension grammatical, which can be further taken to indicate that -z is a morphologically conditioned surface variant of the aorist. Furthermore, the emergence of -z in suspended affixation for no apparent reason strongly suggests that the grammar exhorts forces on terminal elements to be overtly marked as an optimal way of realizing an inflected verbal form as a morphological word.

Having observed various conditions on the legitimacy of affix suspension in Turkish verbal coordinate constructions, we can make the following generalizations: (i) the ability of a verbal morpheme to terminate a word crucially correlates with its ability to stand on its own without the presence of agreement morphology; (ii) suspended affixation in verbal coordinate constructions is legal only if the conjuncts end in morphemes that can terminate a word in Turkish; (iii) when conjuncts end in cliticlike particles, although they may be preceded by terminal elements, they are interpreted as being followed by the default the 3rd person singular, causing agreement mismatches when the intended subject of the final conjunct is non-3rd person singular; and (iv) the terminal element in the nonfinal conjunct must be phonologically overt.

6. Suspended affixation in nonverbal coordinate constructions

Does the concept of the morphological word also account for suspended affixation facts in nonverbal coordinate constructions? Before we answer this question, let us first consider some of the morphosyntactic properties of Turkish nonverbal predicates. Just like participle forms, nonverbal stems are further inflected for categories such as tense and personal agreement via the intermediary of the copula. The same conditions on the copula in verbal constructions also apply to nonverbal predicates, namely that the copula is null in the present tense. Otherwise, it is realized as -y unless it occurs interconsonantally (Kornfilt 1996). The following examples illustrate suspended affixation in nonverbal predicates:

(37) a. hasta ve yorgun-0-uz. sick and tired-Cop-[1P.sub.(z)] 'We are sick and tired.' (Full form: hasta-0-yiz ve yorgun-uz)

b. guzel ve unlu-y-mus-sunuz. beautiful and famous-Cop-RepPast-[2P.sub.(z)] 'You were beautiful and famous (hearsay).' (Full form: guzel-mis-siniz ve unlu-y-mus-sunuz)

c. yoksul ve zavall-y-di-k. poor and miserable-Cop-Past-[1P.sub.(k)] 'We were poor and miserable.' (Full form: yoksul-0-du-k ve zavalli-y-di-k)

The nonfinal conjuncts in (37) are bare nouns, which are words that can occur in isolation independently from agreement morphology, suggesting that they constitute morphological words in Turkish. The suspension of inflectional categories on the bare nouns is legitimate since the left conjuncts constitute well-formed morphological words. It should be remembered that the same is not true for bare verbs, which need a terminal verbal marker in order to acquire morphological word status. However, just as with verbal coordinate constructions, if the nonfinal conjunct ends in morphemes such as the past tense marker, the conditional marker, or cliticlike particles, all of which may never appear without an agreement marker, nominal predicates with suspended affixation also become ungrammatical (38).

(38) a. * guzel-0-di ve zengin-di-0-niz. Beautiful-Cop-Past and rich-Past-Cop-[2P.sub.(z)] (Full form: guzel-0-di-niz ve zengin-0-niz.) Intended meaning: 'You were beautiful and rich.'

b. * hasta-y-sa ve fakir-0-se-k sick-Cop-Cond and poor-Cop-Cond-[1P.sub.(k)] (Full form: hasta-y-sa-k ve fakir-0-se-k) Intended meaning: 'If we are sick and poor.'

c. *hasta-mi ve yorgun-mu-sun? sick-Que and [tired-Que-2P.sub.(z)] (Full form: hasta-mi-sin ve yorgun-mu-sun?) Intended meaning: 'Are you sick and tired?'

d. *Unlu-dur ve zengir-dir-siniz. (Variety A) famous-EpCop and [rich-EpCop-2P.sub.(z)] (Full form: unlu-dur-sunuz ve zengin-dir-siniz) Intended meaning: 'You (pl.) are probably / must be famous and rich.'

When we look at other types of nonverbal constructions that involve the coordination of Noun Phrases, the morphological wordhood of non-final conjuncts also holds to be a crucial requirement for the suspension of all nominal affixes except for (i) derivational morphemes, and (ii) the possessive when it co-occurs with the plural morpheme. Before discussing these two exceptions, let us first take a close look at those cases where nominal affixes are free to be suspended:

(39) a. Almanya ve Amerika-dan Germany and America-Abl 'from Germany and America'

b. ev ve dukkan-lar-da house and shop-Pl-Loc 'in houses and shops'

c. Mert ve Can-la Mert and Can-Com 'with Mert and Can'

d. Arda ve Can-in Arda and Can-Gen 'Arda and Can's'

Since the nonfinal conjuncts are morphological words in (39), the suspension of nominal categories such as the case suffixes and the plural marker is permissible. Although the same condition holds for the following coordinate constructions in (40), there seems to be a strict constraint on the suspension of derivational morphemes.

(40) a. *fayans ve baca-ci gel-di tile and chimney-Der come-Past (Full form: fayans-ci ve baca-ci gel-di.) Intended meaning: 'The tile-layer and the chimney-sweeper arrived.'

b. *ruh ve toplum-sal aci-dan soul and society-Der angle-Abl (Full form: ruh-sal ve toplum-sal aci-dan) Intended meaning: 'from a psychological and societal viewpoint'

c. *guzel ve sade-lik konu-su beautiful and pure-Der topic-Pl-CmpMrk Intended meaning: 'the topic of beauty and simplicity'

It should be noted that derivational morphemes can be attached to certain tightly coordinated nouns, which on the surface may look like instances of affix suspension (e.g., meyva ve sebze-lik (fruit and vegetable-Der) 'a fruit and vegetable container'; badana ve boya-ci (whitewash and paint-Der) 'painter'; ana (ve) baba-lik (mother (and) father-Der) 'parenthood'; ay-yildiz-li bayrak (moon-star-Der flag), sari-kirmizi-li takim (yellow-red-Der team), two well-known instances of antonomasia to refer to the Turkish flag and the Galatasaray soccer team, respectively). However, as opposed to accidental coordination these forms should be regarded as instances of co-compounds or natural coordination, which express stereotypically conjoined entities (see Walchli 2005 for a detailed typological survey and analysis of the co-compound construction and natural coordination). Arguably, such constructions involve coordination of items that are expected to co-occur, and behave as a single conceptual unit with the derivational morpheme attached to it. Therefore, they should not be considered as representative of affix suspension.

Having excluded tightly coordinated nouns from consideration, we can conclude that the Turkish morphological system exhibits a split behavior between derivational morphemes and inflectional ones: unlike inflectional morphemes, derivational morphemes cannot have scope over conjuncts and hence they cannot be suspended. This could be attributed to the fact that derivational affixes show closer lexical affinity to the stems that they are attached to. As such, the ungrammaticality of suspended affixation in (40) should not be attributed to the morphological word status of left conjuncts. Instead, it results from the strong morphological bonding between the affixes in question and the bases that they attach to. Indeed, in light of compelling structural and psycholinguistic evidence that strongly suggests that inflection and derivation correspond to distinct systems in linguistic competence (e.g., Anderson 1992; Micelli and Caramazza 1988), the split between the two systems in Turkish morphosyntax should come as no surprise.

As for the relationship between the possessive and the plural in coordinate constructions, it is instructive to consider a number of observations made in Orgun (1995, 1996). Orgun observes that while it is possible to suspend all of the plural, possessive, and the accusative suffixes (41a), or the plural and the possessive (41b), the suspension of the plural and the possessive that leads to their separation is illegitimate (41c), (41d). When the possessive and the plural are realized together on the conjuncts (41e), however, the same construction becomes grammatical (examples from Orgun 1996).

(41) a. kedi ve kopek-ler-im-i cat and dog-Pl-1S.Poss-Acc 'my cats and dogs (Acc.)'

b. aci ve sevinc-ler-i sorrow and joy-Pl-3S.Poss 'his sorrows and joys'

c. *kedi-ler ve kopek-ler-im-i cat-Pl and dog-Pl-1S.Poss Intended meaning: 'my cats and dogs (Acc.)'

d. *aci-lar ve sevinc-ler-i sorrow-Pl and joy-Pl-3S.Poss Intended meaning: 'his sorrows and joys'

e. kedi-ler-im ve kopek-ler-im-i cat-Pl-1SPoss and dog-Pl-1S.Poss-Acc 'my cats and dogs (Acc.)'

Orgun claims that both suffixes at issue form a ternary branching (flat) structure with the base that they attach to, rather than a binary branching hierarchical structure. Accordingly, the plural and possessive must be sisters whenever they are both present; thus, splitting them up, as in (41c) and (41d), creates ill-formed structures. That is, suffixes can be separated if and only if they form a hierarchical structure. Those forming a flat structure have to be suspended en masse, or not at all.

At first sight, the above ungrammatical instances of suspended affixation seem to contradict what I have suggested above: if a given form qualifies as a morphological word, any affix that can attach to it can be suspended. For instance, as shown in (42) below, the plural marker constitutes a terminal suffix in Turkish since it can be realized in a nonfinal conjunct with the suspension of other suffixes that come after it.

(42) a. dil+bilim-ci-ler ve psikolog-lar-a gore language+science-Agn-Pl and psychologist-Pl-Dat PostP 'according to linguists and psychologists'

b. cevre il-ler ve ilce-ler-den around city-Pl and town-Pl-Abl 'from cities and towns around'

c. yuksek okul-lar ve universite-ler-de high school-Pl and university-Pl-Loc 'at high schools and universities'

Having shown that plural forms are morphological words in Turkish, the puzzling restriction on the suspension of the possessive marker that leads to its separation from the plural marker should then be regarded as a systematic exception to the rule. However, the suspended affixation facts concerning the possessive and plural morphemes are far more complicated than what Orgun suggests. For instance, there are cases where the separation of the plural and the possessive can be legitimate, as illustrated in (43).

(43) a. kahraman asker-let ve komutan-lar-imiz heroic soldier-Pl and commander-Pl-1P.Poss 'our heroic soldiers and commanders'

b. avukat-lar ve danisman-lar-iniz lawyer-Pl and advisor-Pl-2P.Poss 'lawyers and advisors'

c. aydin-lar ve bilim+adam-lar-imiz scholar-P1 and science+man-Pl-1P.Poss 'Our scholars and scientists'

d. yurt-lar, kampus-ler ve universite-ler-imiz dorm-Pl, campus-Pl and university-Pl-1P.Poss 'Our dorms, campuses and universities'

e. kasaba-lar ve koy-ler-imiz-de town-Pl and village-Pl-1P.Poss-Loc 'in our towns and villages'

A closer look at the semantic properties of the plural forms in (43) reveals that each coordinated Noun Phrase denotes a collection of individual entities that are either agentives or institutions. It seems that when the plurality of nouns that denote collectivity must be indicated, the apparent constraint on the separation of the plural from the possessive can be relaxed. Furthermore, Orgun's ungrammatical examples are somewhat awkward to some of my informants, including myself, but they are nevertheless considered to be marginally acceptable. Further examination reveals that the constructions illustrated by Orgun are even more acceptable if the plurality of the possessed things needs to be emphasized in speech. Interestingly enough, the same constructions render more acceptable structures when they are accompanied by overt possessor nouns (44).

(44) a. Herhalde sen-in [kedi-ler ve kopek-ler]-in cok Probably you-2S.Gen [cat-Pl and dog-Pl]-2S.Poss very sessiz-dir. quiet-EpCop

'Your cats and dogs must probably be very quiet.'

b. Fakir halk-in [aci-lar ve sevinc-ler-i]-ni Poor people-3S.Gen [sorrow-Pl and joy-Pl-3S.Poss]-Acc kim dusun-ur? who consider-Aor

'Who cares for the sorrows and joys of the poor?'

I suggest that, in the absence of an overt possessor, the strongest interpretation of a bare plural noun (i.e., a bare plural noun unmarked for a structural case such as the accusative or the genitive) such as kedi-ler 'cats' is generic or indefinite while a singular noun such as kedi 'cat' has a less strong generic reading; thus it can recuperate any semantic interpretation (i.e., definiteness or indefiniteness, etc.) from the context. Thus, when speakers process (45a) below, they construct (45b) as an initial resort, which leads to an infelicitous interpretation after the second conjunct has been processed.

(45) a. kedi-ler ve kopek-ler-im

b. [kedi-ler] (generic) ve [kopek-ler-im] ?'Cats and my dogs'

However, when it is in the singular, there is no a single dominant interpretation that may override any of the potential semantic interpretations of the NP. As such, when a noun with a possessive marker is encountered at the end of the coordinate construction, all semantic options are equally available for such bare singulars. Naturally, such an explanation on the basis of the semantics of plurals does not go beyond speculation without valid psycholinguistic investigation. As the primary goal of this article is to discuss morphosyntactic constraints of suspended affixation, I leave a full-fledged analysis of the semantics of suspended affixation for future research.

In summary, we have observed that any nominal inflectional category can be suspended in Turkish, suggesting that all nominal inflectional categories are terminal and all bare nominal bases, unlike bare verbal ones, constitute morphological words.

7. Tight phonological cohesion and suspended affixation

Yet another constraint on affix suspension seems to correlate with the extent to which affixes phonologically cohere with the base they attach to. It is widely assumed that Turkish employs a predominantly separatist and noncumulative morphological system where morphological boundaries are transparent. While morphemes causing stem alternations, thereby showing tight phonological cohesion with their bases, are rare, those that exist provide an instructive test case to further explore the interaction between phonological cohesion and morphological bonding. The facts are straightforward: the 1st and 2nd person singular pronouns change their root vowel when inflected for the dative case ([46a], [46b]), but they never do so when inflected for other cases, such as the ablative -D En. Other pronouns, however, do not change their vowel in the same context ([46c], [46d], [46e]).

(46) a. /ben/+/-(y)E/ [right arrow] [bana] 'to me' (cf. ben-den 'from me')

b. /sen/+/-(y)E/ [right arrow] [sana] 'to you' (cf. sen-den 'from you')

c. /biz/+/-(y)E/ [right arrow] [bize] 'to us'

d. /siz/+/-(y)E/ [right arrow] [size] 'to you (pl.)'

e. /onlar/+/-(y)E/ [right arrow] [onlara] 'to them'

Overall, Turkish speakers find the suspension of the dative marker on pronouns less acceptable, compared to, for instance, the suspension of the ablative or the locative, although the dative marker can be suspended in nonpronominal noun phrases. A detailed examination, however, reveals that speakers strictly reject the suspension of the dative marker on the 1st or 2nd person singular pronoun (47) while acceptability judgments vary from awkward to acceptable with other personal pronouns (48).

(47) a. *Ilk once sen ve ban-a bak-ti first you and I-Dat look Intended meaning: 'S/he first looked at you and me.'

b. *ben ve san-a gel-en paket-ler I and you-Dat come-Rel parcel-Pl. Intended meaning: 'the parcels that came to me and you'

(48) a. ?Izin sadece biz ve siz-e ver-il-ecek. Permission only we and you(pl)-Dat give-Pass-Fut. 'The permison will be granted only to you and us.'

b. ?Siz ve onlar-a karsi yap-il-an suclama-lar you(pl) and they-Dat against do-Pass-Rel accusation-Pl 'The accusations made against you and them'

The dative-conditioned allomorph of the 1st and 2nd person pronouns, namely san and ban can never be used in the absence of the dative. This suggests that these allomorphs are not morphological words, thus, they also cannot be left bare in nonfinal conjuncts (i.e., *san ve ban-a (you and I-Dat) is also ungrammatical). As shown in (49) below, the ablative and locative can freely be suspended with the 1st and 2nd person singular, suggesting that the ungrammaticality of the constructions in (47) is not due to the type of person markers but rather the phonological effect of the dative marker on the shape of the pronouns.

(49) a. Ben ve sen-den nefret ed-iyor. I and you-Abl hate Aux-Prog 'S/he hates me and you.'

b. Bu kitap bir tek sen ve ben-de var. this book one only you and I-Loc exists. 'Only you and I have this book.'

The overall unacceptability of the separation of the dative marker from pronouns can be attributed to the fact that the dative marker and the pronouns are frequently co-occuring elements, which ultimately paves way for their tight phonological cohesion.

8. Summary and conclusions

This article has focused on various constraints on suspended affixation. Investigating the type of material that can be left in nonfinal conjuncts in both verbal and nonverbal coordinate constructions, I have shown that suspension of affixes is legitimate if the bare conjunct constitutes a morphological word in Turkish. A morphological word is defined to be a form that is able to occur in isolation. Accordingly, I have suggested an account that classifies bound morphemes in Turkish depending on their ability to terminate a morphological word. While all nominal inflectional morphemes can terminate a word, only a small number of verbal inflectional categories can do so. Thus, in verbal constructions, the suspension of affixes is only legitimate if it renders a form ending in a terminal morpheme. It is shown that native speakers interpret the absence of personal agreement markers in nonfinal conjuncts as conveying the 3rd person singular. This is taken to further suggest that the morphological wordhood of a verbal construction is strongly correlated with its ability to survive without personal agreement morphology. Derivational morphemes, on the other hand, cannot be suspended in Turkish, leaving the suspended affixation phenomenon restricted to inflection classes only, and providing another piece of evidence for the crosslinguistically and psycholinguistically attested distinction between derivation and inflection. I have suggested that the inseparability of the plural from and the possessive in suspended affixation is far too complex to be accounted for merely on morphological grounds since the semantics of plurals seem to crucially interact with the acceptability of such forms. Finally, I have argued that the suspension of those affixes that exhibit high phonological cohesion with their bases leads to illicit structures, suggesting that loose morphological bonding, the hallmark of agglutinative languages with which suspended affixation is typically associated with, may strongly correlate with the degree of phonological cohesion. This observation can further be taken to provide an explanation for the relatively restricted occurrence of affix suspension in languages with more fusional properties. Future research should therefore focus on the extent to which morphological bonding of elements can be loose in flexive languages to arrive at universal generalizations.

University of Constance

Received 30 July 2004

Revised version received 14 December 2004

Appendix. Abbreviations

1/2/3S (k)/(z) 1st/2nd/3rd person singular from the k-paradigm / z-paradigm 1/2/3P (k)/(z) 1st/2nd/3rd person plural from the k-paradigm / z-paradigm
Abl Ablative
Abil Abilitative
Acc Accusative
Agn Agentive
Aor Aorist
Aux Auxiliary
Caus Causative
CmpMrk Compound Marker
Com Commutative
Cond Conditional
Conn Connector
Cop Copula
Dat Dative
Der Derivational morpheme
Dvb Deverbalizer
EpCop Epistemological Copula
Evid Evidential
Fem Feminine
Fut Future
Gen Genitive
Masc Masculine
Nec Necessitative
Neg Negative
Loc Locative
Obl Oblique
Pass Passive
Past Past
Pl Plural
Poss Possessive (e.g., 3S.Poss: 3rd person singular possessive)
PostP Postposition
Pot Potential
Prob Probability
Prog Progressive
Que Question
Rel Relativizer
RepPast Reported Past
Top Topic Marker


References

Adamovic, Milan (1985). Konjugutionsgeschichte der Turkischen Sprache. Leiden: E. J. Brill. Anderson, Stephen R. (1992). A-Morphous Morphology. Cambridge: Cambridge University Press.

Bassarak, Armin (2000). On some fusional features in Turkish. In Studies on Turkish and Turkic Language. Proceedings of the 9th International Conference on Turkish Linguistics, Ash Goksel and Celia Kerslake (eds.), 35-40. Wiesbaden: Harrassowitz.

Csato, Eva A. and Johanson, Lars (1998). Turkish. In The Turkic Languages, Lars Johanson and Eva Agnes Csato (eds.), 203-235. London and New York: Routledge.

Erdal, Marcel (2000). Clitics in Turkish. In Studies on Turkish and Turkic Languages: Proceedings of the 9th International Conference on Turkish Linguistics, Ash Goksel and Celia Kerslake (eds.), 41-48. Wiesbaden: Harrassowitz.

Good, Jeff and Yu, Alan (2005). Morphosyntax of two Turkish subject pronominal paradigms. In Clitic and Affix Combinations. Theoretical Perspectives, Lorie Heggie and Francisco Ordonez (eds.), 315-341. Amsterdam: Benjamins.

Goksel, Asli (2001). The auxiliary verb at the morphology-syntax interface. In The Verb in Turkish, E. Erguvanh-Taylan (ed.), 151-181. Amsterdam: John Benjamins.

--(2003). Syntactic head or morphological buffer? The auxiliary verb in Turkish. In Studies in Turkish Linguistics: Proceedings of the Tenth International Conference on Turkish Linguistics, Sumru Ozsoy, Eser Erguvanh-Taylan, Ayhan Aksu-Koc, Didar Akar, and Mine Nakipoglu-Demiralp (eds.), 49-57. Istanbul: Bogazici University Press.

Hameed, Jumah (1985). Lexical phonology and morphology of Modern Standard Turkish. Cahiers linguistiques d'Ottawa 14, 71-95.

Hankamer, Jorge (1989). Morphological parsing and the lexicon. In Lexical Representation and Process, William Marslen-Wilson (ed.), 392-408. Cambridge, MA: MIT Press.

Inkelas, Sharon and Orgun, Cemil O. (1995). Level ordering and economy in the lexical phonology of Turkish. Language 71, 763-793.

-- and Orgun, Cemil O. (1998). Level (non)ordering in recursive morphology: evidence from Turkish. In Morphology and its Relation to Phonology and Syntax, Steven G. Lapointe, Diane K. Brentari, and Patrick M. Farrell (eds.), 360-410. Stanford: CSLI.

Kabak, Baris and Vogel, Irene (2001). The phonological word and stress assignment in Turkish. Phonology 18, 315-360.

Kaisse, Ellen (1985). Some theoretical consequences of stress rules in Turkish. Chicago Linguistics Society 21, 199-209.

--(1986). Towards a lexical phonology of Turkish. In A festschrift for Sol Saporta, Michael Brame and Frederick J. Newmeyer (eds.), 231 239. Seattle: Noit Amrofer.

Kornfilt, Jaklin (1996). On some copular clitics in Turkish. In ZAS Papers in Linguistics 6, Artemis Alexiadou, Nanna Fuhrhop, Paul Law, and Sylvia Lohken (eds.), 96-114. Berlin: Zentrum fur Allgemeine Sprachwissenschaft.

--(1997). Turkish. London: Routledge.

Lees, Robert B. (1962). A compact analysis for the Turkish personal morphemes. In American Studies in Altaic Linguistics, Nicholas Poppe (ed.), 141-176. Indiana University Uralic and Altaic Series 13. Bloomington: Indiana University/The Hague: Mouton.

Lewis, Geoffrey L. (1967). Turkish Grammar. Oxford: Oxford University Press.

Micelli, Gabrielle and Caramazza, Alfonso (1988). Dissociation of inflectional and derivational morphology. Brain and Language 35(1), 24-65.

Orgun, Cemil O. (1995). Flat versus branching morphological structures: the case of suspended affixation. In Proceedings of the Berkeley Linguistic Society 21, Jocelyn Ahlers, Leela Bilmers, Joshua S. Guenter, Barbara A. Kaiser, and Ju Namkung (eds.), 252-261. Berkeley: Berkeley Linguistics Society.

--(1996). Suspended affixation: a new look at the phonology-morphology interface. In Interfaces in Phonology, Ursula Kleinhenz (ed.), 251-261. Berlin: Akademie Verlag. Payne, John R. (1995). Inflecting postpositions in Indic and Kashmiri. In Double Case: Agreement by Suffixaufnahme, Frans Plank (ed.), 283-298. Oxford and New York: Oxford University Press.

Plank, Frans (1999). Split morphology: how agglutination and flexion mix. Linguistic Typology 3, 279-340.

Sohn, Ho-Min (1999). The Korean Language. Cambridge: Cambridge University Press.

Stump, Gregory T. (1998). Inflection. In Handbook of Morphology, Andrew Spencer and

Arnold Zwicky (eds.), 13-43. Oxford: Blackwell Publishers.

--(2001). Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge University Press.

Underhill, Robert (1976). Turkish Grammar. Cambridge: The MIT Press.

Walchli, Bernhard (2005). Co-Compounds and Natural Coordination. Oxford: Oxford University Press.

Yu, Alan C. and Good, Jeff C. (2000). Morphosyntax of two Turkish subject pronominal paradigms. In Proceedings of the North Eastern Linguistics Society 30, M. Hirotani, A. Coetzee, N. Hall, and J.-Y. Kim (eds.), 759-773. New Brunswick: Rutgers University.

Zwicky, Arnold and Pullum, Geoffrey K. (1983). Cliticization vs. inflection: English n't. Language 59, 502-513.

Notes

* The initial stage of this research was written at the University of Delaware, USA. It was extended and completed at the University of Constance, as part of the SFB 471 "Variation and Evolution in the Lexicon" funded by the German Science Foundation (DFG). I would like to thank Nihan Ketrez and Meltem Kelepir for their comments on earlier drafts of this work. I am especially grateful to Frans Plank and Amanda Pounder for the long fruitful and inspiring discussions we had on affix suspension, and to the anonymous reviewers of Linguistics for their valuable criticism and suggestions, which have greatly improved this article. Naturally, all shortcomings and errors are my own. Correspondence address: Department of Linguistics, University of Constance, Fach D180, 78457 Constance, Germany. E-mail: Baris.Kabak@uni-konstanz.de.

(1.) It should be noted that suspended affixation can be observed in coordinate constructions with multiple conjuncts as in (2). For space limitations, however, I will only illustrate coordinate constructions with two conjuncts. The constraints governing the legitimacy of suspended affixation with two conjuncts are also generalizable to those with three or more. Furthermore, the phenomenon is not necessarily specific to the conjunctot ve 'and'. Semantics permitting, other conjunctors such as ya 'either', hem 'both', ne 'neither', ya da 'or' can also be used in coordinate constructions with suspended affixation. For consistency, however, I will only give examples using re.

(2.) In this article I adhere to the conventions of Turkish orthography. Accordingly, i and t represent the high front unrounded and high back unrounded vowel, respectively. u is the high front rounded vowel, o the non-high front rounded vowel. s represents the voiceless palatal fricative; f and c indicate voiceless and voiced palatal affricates, respectively. Phonologically null morphemes, such as the copula, are indicated with "0". Following the Turkish linguistics tradition, capital letters are used to indicate phonemes that undergo assimilation processes, where necessary. Accordingly,/E/represents vowels that undergo harmony for backness, and /I/ for both backness and rounding.

(3.) Erdal (2000) considers that the pronominal suffixes from the z-paradigm are clitic pronouns, not subject agreement suffixes. As will be shown below, this assumption is also in line with Yu and Good's (2000) analysis, which, however, alone makes empirically wrong predictions with respect to suspended affixation.

(4.) The slot numbers given in Figure 1 do not correspond to those given in Goksel (2001) who considers -(y)Ebil as consisting of two suffixes. It is true that -(y)Ebil has been grammaticalized from a converb marker (-(y)E) plus an auxilary bil 'to know'. However, representing this form as two morphemes is tangential to my purposes in this article.

(5.) A few statements with respect to the placement of the question particle -mI and the topic or connector marker -dE are in order here. The question particle normally comes after the TAM suffixes and before the agreement marker unless the verb is in the definite past, where the particle follows the agreement marker instead of preceding it (e.g., al-ir-mi-sm (take-Aor-Q-2S) 'Do you take (it)'; al-acak-mi-y-di-k (take-Fut-Q-Cop-Past-1S) 'Were we going to take (it)' vs. al-di-n-mi (take-Past-2S-Q) 'Did you take (it)?'; al-di-k-mi (take-Past-lP-Q) 'Did we take (it)?'. The marker -dE behaves just as the epistemological copula -DIr, which goes after the agreement marker. There is, however, some dialectal variation in the placement of such particles between the abilitative -(y)Ebil, which historically developed from a converb suffix plus an auxiliary verb bil 'to know'. The detailed discussion of this variation is however beyond the focus of the present article.

(6.) All of the grammaticality judgments in this article are based on 4 native speakers of Turkish, including myself, who speak the Istanbul variety.

7. One anonymous reviewer suggests that (13a) is unacceptable because it shows the wrong order of clitics, where -DIr should follow the agreement markers (13b). However, other participle constructions where -DIr precedes aspect markers are as acceptable as the same constructions with the reverse order of the clitics for my informants (e.g., yap-mis-tir-sm (do-Evid-EpCop-2S(z)) 'You must definitely have done it.'; bekli-iyor-dur-sunuz (wait-Prog-EpCop-2P(z)) 'You (pl.) must be waiting.'. It is therefore necessary to assume that the placement of the epistemological copula in relation to the agreement marker shows variation in Turkish, in that for some speakers (including myself) it is also legitimate to place them before agreement markers. However, this option is mostly unavailable with the necessitative marker for the same speakers. I will refer to these speakers as belonging to "Variety A" throughout this article.

(8.) For an extended version of this article, see Good and Yu (2005).

(9.) For simplicity, the 2nd person plural agreement morpheme (-nIz in the k-paradigm and -sInIz in the z-paradigm) is shown as a single decomposable suffix although it is usually assumed that the suffix is diachronically composed of two suffixes: the 2nd person marking -n (k-paradigm) or -sIn (z-paradigm) plus the plural -Iz. The reader is referred to Bassarak (2000) for synchronic evidence from suspended affaxation suggesting the inseparability of the individual components of this morpheme.

(10.) It should be noted that the doubling of the particle -mI has disjunctive meaning (i.e., either ... or) in Turkish, as suggested by an anonymous reviewer. Accordingly, (16) becomes acceptable if ve is taken out: Calis-iyor-mu kalkm-iyor-mu-yuz?, where the intended meaning is "Are we either working or developing?". In this case, -mI on the nonfinal conjunct is a disjunctive marker rather than a question particle. The evidence for this comes from the fact that -mI with the disjunctive function cannot co-occur with the conjunctor ve 'and', and it also seems to be somewhat unacceptable with disjunctors such as veya or ya da 'or'. In those cases where the question particle has a disjunctive function, our focus of interest is the final morpheme on the nonfinal conjunct that precedes -mI, instead of -mI itself. Interestingly, in all such cases, the nonfinal conjunct ends in an aspect / modality marker (in the above example, it is -Iyor [progressive]). Accordingly, the present analysis correctly predicts such constructions to be grammatical since aspect / modality markers can legitimately terminate a conjunct with suspended affixation.

(11.) This point has been suggested to me by an anonymous reviewer. Although it is possible to modify the analysis proposed here in line with the premises of realizational approaches to morphology, I will leave this for future research since (i) it is largely tangential to the issue of suspended affixation, and (ii) the applicability of realizational approaches to Turkish morphology needs to be tested on independent grounds.

(12.) It should be noted that the classification of Turkish morphemes based on their ability to terminate a word is not analogous to classifying morphemes into levels according to the premises of Lexical Phonology. Indeed, such an account has been offered in the context of Turkish suspended affixation by Inkelas and Orgun (1995), extending on Orgun (1995), which will be discussed below. The stratification of Turkish morphemes into different levels in Inkelas and Orgun (1995) is, however, arbitrary and incompatible with other level stratification of suffixes based on the phonological properties of Turkish affixes in the literature (e.g., Hameed 1985; Kaisse 1985, 1986). As extensively discussed in Kabak and Vogel (2001), level ordering is unrealistic for Turkish as it often leads to ordering paradoxes. Indeed, Inkelas and Orgun in their subsequent work (Inkelas and Orgun 1998) admit that strict level ordering such as the one they offer in Inkelas and Orgun (1995) is "illusory" for a language like Turkish since it employs recursive morphology.

(13.) It could be argued that a grammatical construction such as (i) below could be used as a counter example to this argument. The nonfinal conjunct ends in a non terminal marker such as the past tense morpheme while the null 3rd person singular morpheme takes the scope over the whole conjunct.

(i) Ali gel-di ve git-ti-0. Ali came-Past and go-Past-3S 'Ali came (in) and went (out).'

However, sentences such as in (i) do not constitute a coordinate construction with suspended affixation. That is, there is no evidence that the null 3rd person singular marker is omitted from the nonfinal conjunct. Therefore, we can assume that sentences such as (i) are grammatical because both conjuncts carry the null 3rd person singular agreement morphology, as shown in (ii).

(ii) Ali gel-di-0 ve (Ali) git-ti-0. Ali came-Past-3PS and (Ali) go-Past-3S 'Ali came (in) and went (out).'

(15.) It should be pointed out that mE in mEz differs from other tokens of the negative -mE with respect to stress in Turkish. Namely, stress normally falls on the syllable preceding the negative morpheme. However, mE receives stress in mEz. Furthermore, the shape of the 1st person singular marker when the negative co-occurs with the aorist is intriguing. If it were from the z-paradigm, it would have the shape -yIm. Rather, the agreement marker looks like the 1st person singular from the k-paradigm, as one anonymous reviewer suggests. While these complications necessitate a detailed morphological analysis of the aorist in Turkish, I leave the discussion of this for future studies since they are tangential to the question of suspended affixation.
Table 1. Person agreement paradigms

 k-paradigm z-paradigm

 do-Past do-Cond do-Aor-Cop lawyer-Cop

1S yap-ti-m yap-sa-m yap-ar-0-im avukat-0-im
2S yap-ti-n yap-sa-n yap-ar-0-sin avukat-0-sin
3S yap-ti-0 yap-sa-0 yap-ar-0-0 avukat-0-0
1P yap-ti-k yap-sa-k yap-ar-0-iz avukat-0-iz
2P yap-ti-n yap-sa-n yap-ar-0-siniz avukat-0-siniz
3P yap-ti-lar yap-sa-lar yap-ar-0-lar avukat-0-lar

Table 2. Verbal terminal morphemes

(i) Agreement markers

(ii) Aspect/Modality markers -(I)r/(E)r--Aorist
 -Iyor--Progressive
 -(y) EcEK--Future
 -(y)ml--Evidential/Rep. Past
 -mElI--Necessitative

(iii) Converb markers Such as:
 -(y) IncE
 -(Y) IP
 (see Csato and Johanson 1998:
 231-232 for their function).

Figure 1. Morpheme slots for Turkish verbs
(adopted from Goksel 2001)

 1 2 3 4

Verb -(y)Ebil -Iyor (Prog) i-/-y/-0 -DI (Past)
 (Abil) -(I)r/(E)r (Cop) -mIs
 (Aor) (Rep.Past)
 -(y)EcEK -sE (Cond)
 (Fut)
 -mElI (Nec)
 -mIs (Evid)
 -sE (Cond)
 -DI (Past)

 5 6 7

Verb -sE AGR -DIr
 (Cond) (Prob /
 Ep.Cop.)
COPYRIGHT 2007 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2007 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Kabak, Baris
Publication:Linguistics: an interdisciplinary journal of the language sciences
Geographic Code:7TURK
Date:Mar 1, 2007
Words:12814
Previous Article:Syntactic lexicalization as a new type of degrammaticalization.
Next Article:A handbook of historical linguistics.
Topics:


Related Articles
Trading Post: Bank on Germanyto go cold in T urk ey; sports play of the day Lay Germany (in Turkey).
Turkey roasted over violence at play-off.
Officer was threatened.
Rights watchdog criticize Turkey's hosting of Sudan president.
Rights watchdog criticize Turkey's hosting of Sudan president.
Rights watchdog criticize Turkey's hosting of Sudan president.
Rights watchdog criticize Turkey's hosting of Sudan president.
Even monkeys have their grammar in place.
Turkish delegation meets Duhok governor.

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