ebook img

TOPICS IN TURKISH PHONOLOGY Harry van der Hulst and Jeroen van de Weijer 0 ... PDF

49 Pages·2011·0.13 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview TOPICS IN TURKISH PHONOLOGY Harry van der Hulst and Jeroen van de Weijer 0 ...

TOPICS IN TURKISH PHONOLOGY Harry van der Hulst and Jeroen van de Weijer 0. INTRODUCTION In this chapter we offer a discussion of some aspects of the phonology of Turkish. Turkish phonology has played a significant role in theoretical discussions on the nature of phonological representation and rule formalism. In particular, the formal description of vowel harmony has attracted a considerable amount of attention in the phonological literature since the 1940s, and we, too, will devote a separate section to this topic. In section 1 we provide a synopsis of the general facts of Turkish phonology. Besides giving an overview of the phonemes of Turkish, we illustrate its syllabic structure and stress pattern. We also present a number of the phonological rules of Turkish, all of which have received earlier treatment in the literature, in particular compensatory lengthening (section 1.4.3). A number of linguists have provided analyses of the process of vowel harmony which pervades the Turkish language. In section 2 we lay out the basic facts, discuss some of the earlier analyses, and then provide our own account, which departs from the earlier approaches mainly by availing itself of unary elements which may extend over suprasegmental domains like the word. We believe that significant generalizations can be captured under this approach. 1. ASPECTS OF TURKISH PHONOLOGY 1.1 THE PHONEMIC INVENTORY 1.1.1 Vowels Turkish has eight vowel phonemes which may be plotted on the familiar triangular vowel diagram as follows (cf. Lass 1984: 145; Maddieson 1984: 277): (1) high i,y u,uu mid o lower mid e,oe low a Following all earlier writers (e.g. Jakobson 1942), we assume that the vowels phonologically pattern into a set of four high and four low vowels, in which /a/ is classified as back. We thus obtain the following rectangular vowel inventory (in which for reasons of typographical convenience we use /i ü u + o e ö a/ instead of the phonetic symbols above): (2) ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ front ³ back ³ ÃÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄ´ ³ non-round ³ round ³ non-round ³ round ³ ÚÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄ´ ³ high ³ i ³ ü ³ + ³ u ³ ÃÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄ´ ³ low ³ e ³ ö ³ a ³ o ³ ÀÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÙ There are also long vowels, which come from two sources (Underhill 1986a: 10): Arabic and Persian loans have introduced the long vowels /a: e: i: u:/, and thus we find sakin [sa:kin] `quiet' vs. sak+n [sak+n] `beware', etc. In native words, long vowels have also arisen through the loss of a voiced velar fricative, which is preserved as such in various dialects of Turkish and closely related languages, and appears in the current orthography as g. We might assume an underlying /g/ phoneme, which disappears intervocalically, creating a bisyllabic two-vowel sequence: agaç [aac] `tree', eger [eer] `if', etc, with merger into a long vowel in fast speech. In syllable-final position, then, the loss of this abstract phoneme causes lengthening of the preceding vowel: dag [da:] `mountain', tug [tu:] `banner', igne [i:ne] `needle'. We will return to this process in section 1.4.3. Underlying long vowels shorten in closed syllables (cf. section 1.4.4), although when derived they can occur in closed syllables (cf. 1.4.3). Long vowels do not occur before vowel-initial suffixes: vowels of suffixes are deleted in that situation, and vowels that have become long as a result of compensatory lengthening do not arise in that position. Morpheme-internally, however, long vowels may occur before short vowels (Sezer 1981: 380): (3) `poet' s,a:ir `always' da:ima This shows that there is no surface constraint against /V:/ followed by /V/ as such. The most well-known process with respect to vowels is vowel harmony, which is extensively discussed in section 2. 1.1.2 Consonants The consonantal system is as follows: (4) labial labiodent. dental palato- palatal velar glottal alveolar voiceless stop p t c, k voiced stop b d j, g voiceless fric. f s s, voiced fric. v z z, nasal m n liquid l,r approximant y h Some consonants, notably /k g l/, have two allophones, one palatal and one non-palatal. The distribution of these is determined by the frontness or backness of neighbouring vowels; we will discuss this in section 2.4.4. The evidence for recognizing phonemically palatal and non-palatal consonants is slight. In loans from Arabic, the original velar stop is consistently replaced by a front /k/ in Turkish, and the uvular Arabic /q/ by a back velar. As a result, front velars may appear with back vowels, as in /k,ar/ `profit', which contrasts with the native word /kar/ `snow'. The same goes for palatal /l/, as in hal /hal,/ `condition', which is a near-minimal pair with the native word /bal/ `honey'. Underhill (1986a) recognizes these three as phonemes of the language. 1.2 SYLLABLE STRUCTURE The canonical structure of Turkish syllables is (C)V(C). We will adopt the following view of internal syllabic structure: (5) syllable (σ) onset (o) nucleus (n) ³ x x x Turkish has more complex consonant clusters word-initially and word-finally. For instance, borrowings that have word-initial clusters may, especially in more casual speech styles, be made to conform to the phonotactics of the language by breaking up the clusters by vowel insertion. In (6) we give some examples, taken from Clements and Sezer (1982): (6) grup -> gurup `group' kral -> k+ral `king' prince -> pirens `prince' smok,in -> s+mok,in; simok,in `dinner jacket' The epenthetic vowel harmonizes in frontness with the following root vowel after labial and dental consonants. If the clusters are not broken up, as in more careful styles of speech, this velar is invariably back in quality, as Clements and Sezer (1982: 248) note. Another way of treating word-initial clusters appears to consist of the prothesis of a (usually harmonic) /i/ or /+/ before the onset (typically /sp- , st-, sk-/), shifting the syllable boundary. Thus the dictionary (Steuerwald 1972) lists: (7) +spanak `spinach' +statistik `statistics' iskelet `skeleton' istaka `billiards queue' Syllables can be closed or open. Word-finally the following consonant clusters are allowed (cf. Clements and Sezer 1982: 245): (8) (i) sonorant + obstruent k,ent `city' harf `letter' (ii) voiceless fricative + stop c,ift `couple' s,evk, `fervour' (iii) k + s raks `dance' boks `boxing' Other final clusters are broken up by epenthesis; cf. section 1.4.1. Kaye (1989) offers an analysis of syllable structure in which empty syllabic nuclei are postulated in such words as kaplar `containers', deriving the shortening of long vowels from the presence of such empty positions, rather than as a result of syllable closure. We will not participate in this discussion, and return to the effect of syllabic structure on stress in the next section. 1.3 STRESS In this section we offer a discussion of Turkish stress, which according to Underhill (1976) and Lewis (1985) is most accurately described as pitch accent, that is, a high tone on the accented syllable. We will continue, however, to use the term stress. 1.3.1 Regular and exceptional word stress Stress falls on the final syllable of a word, whether simplex or derived. We discuss exceptional cases below. The following examples (taken from Sezer 1983) nicely illustrate this point: (9) `know' tan+ `acquaintance' tan+ - d+k `acquaintances' tan+ - d+k - lar `my acquaintances' tan+ - d+k - lar - +m `our acquaintances' tan+ - d+k - lar - +m - +z Within metrical theory (cf. Hayes 1981, Halle and Vergnaud 1987) stress rules take the form of a recipe for assigning a binary branching tree structure to the syllables making up the word. For every pair of sister nodes, one node is dependent on (or, intuitively, 'is weaker than') the other, which we will call the head. In the case of Turkish the recipe is straightforward: (10) Assign a right-branching tree in which all left nodes are dependent Consider the following examples (in the graphs dependent nodes terminate a slanted line in the tree structure): (11) w / ³ / ³ / /³ / / ³ w / / ³ / ³ / / / ³ / ³ / / / ³ / /³ / / / ³ / / ³ / / / / ³ / / ³ / / / / ³ w / / / ³ / / / / /³ /³ / / / ³ / / / / / ³ / ³ / / / ³ / / / / / ³ (ta) (n+) (ta) (n+) (d+k) (lar) (ta) (n+) (d+k) (la) (r+) (m+z) In terms of a typology of stress systems, Turkish can be considered a quantity-insensitive, fixed stress system, since differences in syllable structure play no role for the assignment of regular stress, while its location is predictable. Although the above generalization correctly characterizes a large part of the vocabulary, there are also exceptions. Underhill (1979: 18) even states: "there are many pairs of words that are distinguished from each other only by the placement of accent". Some of the exceptions are morphologically conditioned (such as those involving the so-called unstressable affixes, which we discuss below), while others are said to be purely idiosyncratic. Sezer (1983) and Kaisse (1985) discuss a set of words which, although they are exceptional to the final stress pattern, show a subregularity which calls for further analysis. They leave it an open question whether all non-oxytones in fact fall into this class. The class is referred to as consisting of "place names (both Turkish and non-Turkish) and many words of foreign origin" (Kaisse 1985: 199). Sezer (1983) arrives at the following generalizations: (12) a. Stress never falls on the final syllable b. Stress falls on the penultimate syllable if it is either closed or contains a long vowel: `Samuelson' Samuelson `Washington' Vas,ington city in Turkey Antalya `restaurant' lokanta city in Turkey Istanbul `Eisenhower' Ayz+nho:ver c. Stress fall on the antepenultimate syllable if it is either closed or contains a long vowel and the penultimate is open and contains a short vowel: city in Turkey Ankara `window' pencere `buoy' samand+ra `Chevrolet' S,evrole d. Otherwise (i.e. if both penultimate and antepenultimate are open and contain short vowels) stress is on the penultimate: `Kennedy' Kenedi `Ptolemy' Pitolemi `jubilee' jübile city in Turkey Göreme city in Turkey Adana Lewis (1985: 21) makes a different generalization, saying that most exceptional place names have initial stress (noting in particular words like Zonguldak, which runs counter to generalization (12c)). Clearly this stress pattern raises interesting theoretical issues within the metrical approach, as Kaisse observes. The pattern can be analysed by allowing this class of words to undergo a special stress rule which has the following properties: (13) a. Mark the final syllable as 'extrametrical' (i.e. ignore it) b. Place a left-headed (i.e. trochaic) stress foot at the right edge of the word with the proviso that the syllable in head position must be closed or must contain a long vowel c. Assign a right-branching tree in which all left nodes are dependent (= 12) Contrary to the larger part of the vocabulary, then, in this class of words syllable structure does play a role. Syllables which are closed or contain a long vowel count as 'heavy', while open syllables with a short vowel are 'light'. Assuming that long vowels are represented phonologically as having two syllabic positions, we can characterize the heavy syllables as syllables having branching nuclei (cf. (5)). Formally, this class of words differs in two ways from the regular words. First, the final syllable is never stressed and, secondly, prior to the tree building instruction (13c) (which also applies to the regular words), a foot is built which respects syllable structure. Crucial here is that the stress foot cannot be assigned in the words in (12d), because of the condition that the head of the foot must be heavy. As a result no foot can be assigned in these words, and (13c) applies directly, as in regular words. In (14) we illustrate how the three types of word are assigned metrical structure: (14) (is) (tan) (bul) (an) (ka) (ra) (a) (da) (na) (by 13a) (is) (tan) [(bul)] (an) (ka) [(ra)] (a) (da)[(na)] F F ³ ³ \ (by 13b) (is) (tan) [(bul)] (an) (ka) [(ra)] (a) (da)[(na)] W W W /³ ³ ³ / F F ³ / ³ ³ / ³ (by 13c) (is) (tan) [(bul)] (an) (ka) [(ra)] (a) (da)[(na)] The final extrametrical syllable must still be incorporated into the structure as a dependent node. It has been proposed that the dependent character of formerly extrametrical syllables follows from a universal convention (Hayes 1982). A further interesting property of this class of words is that they remain exceptions under affixation: that is, when they are affixed main stress still falls on the originally stressed syllable (examples from Kaisse): (15) `from Ankara' Ankara - dan `from our Washington' Vas,ington - umuz - dan `from Göreme' Göreme - den `from our Ankaras' Ankara - lar - +m+iz - dan Secondary stress falls on the final syllable. We can interpret this as follows: the special class of words, after being stressed by rule (12), undergoes (with the rest of the vocabulary) the 'regular' stress rule, but the metrical structure already assigned is preserved: (16) ³\ / / / / ³ ³ \ / / / / ³ (an) (ka) (ra) (la) (r+) (m+z) (dan) To account for the fact that the earlier assigned exceptional stress remains the most prominent, Kaisse (1985) makes the proposal that the recipe in (10) (=12c) is slightly different: (17) The left node is dependent unless it branches This will make all syllables (ra), (la), (r+), and (m+z) dependent on (dan), but it will make the foot (an)(ka) the head. Support for this analysis comes from the fact that monosyllabic place names never preserve their main stress: (18) city in Turkey Of `from Of' Of - dan (*Of - dan) `from our Of' Of - umuz - dan (*Of - umuz - dan) The formal explanation is that in this case there is no branching foot over the stem. 1.3.2 Suffixation According to Lewis (1985: 23) all polysyllabic suffixes (except the adverbial suffixes -leyin and -cesine) are stressed on their first syllable: (19) `having gone' gid - ince `by doing' yap - arak Apparently, then, there is a rule marking final syllables of affixes as extrametrical. In accordance with another general convention, this rule will not apply to monosyllabic suffixes (Hayes 1982). There is furthermore a set of suffixes which are never stressed, and which also prevent main stress from being assigned to a following suffix. Stress will fall on the syllable preceding these unstressable suffixes (cf. Lewis 1985: 23, Underhill 1979: 34): (20) `with pleasure' memnuniyet - le `while writing' yazar - ken `bestially' hayvan - ca - sina `he did not understand' anla - ma - d+ Both Underhill (1979: 34) and Kornfilt (1988: xxx) note that a suffix following the 'unstressable' or 'pre-stressed' suffixes will receive a secondary stress, a phenomenon reminiscent of what we have seen above. Following Kaisse (1985), we could derive the behaviour of unstressable suffixes by ordering the main stress rule before the morphological level at which these suffixes are attached. Within the model of Lexical Phonology (Kiparsky 1982 and subsequent work) this is a valid procedure. However, to also derive the fact that secondary stress occurs on the final syllable (which seems a regular phenomenon for all words that have main stress somewhere inside the word), we will assume that the main stress rule in fact re-applies after the assignment of these unstressable suffixes. Since we have already assumed that main stress is assigned both as part of the special rule and, later, after affixation of '... suffixes', we arrive at the following overall picture: (21) I : Foot assignment to the special class of words rule (10) <---- II : Affixation up to and including the 'unstressable suffixes' III: further affixation In other words, we recognize three strata in the lexicon (following Kaisse 1985), all of which access the main stress rule in (10). Stratum I is the locus of exceptional patterns. Sezer (1983) points out that adverbs in -en are irregular:

Description:
`bestially' hayvan - ca - sina. `he did not understand' anla - ma - d+. Both Underhill (1979: 34) and Kornfilt (1988: xxx) note that a suffix following the
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.