The Vowel System of the Dialects of the Vipava Valley in Government Phonology

46
The Vowel System of the Dialects of the Vipava Valley in Government Phonology * Adrian Stegovec University of Ljubljana 1 Introduction It is frequently stated in Slovenian linguistic literature that Slovenian is one of the most, if not the most, dialectically diverse Slavic languages. However, most analyses of these variations are diachronic in nature. And especially for phonological analyses, there has been virtually no work done from a generative perspective on language variation. This paper has two major goals: To show the advantages of Government Phonology in analyses of phonological variation and to apply the framework to a specific case of dialectal variation, namely the two dialect groups of the Vipava Valley, in order to determine whether the division based on diachronic data is reflected in the current state of their vowel systems. The paper is organized as follows. Section 2 presents the traditional analysis of the vowel systems in question. Section 3 is a short overview of the principles and methods of Government Phonology. These are then applied in Section 4 to the analysis of the vowel systems of Vipava Valley dialects. Section 5 discusses some potential areas of further research. Section 6 concludes the paper with an overview of its findings. 2 The Dialects of The Vipava Valley The Slovenian dialects of the Vipava Valley (VVS hereafter) are normally not consid- ered a homogeneous group. Traditional dialectologists place them within Western Slove- nian/Littoral dialects based on recent language changes (Logar and Rigler 1983; Smole * This paper is a shortened English version of my BA thesis (Stegovec 2012). The full Slove- nian text is available at: https://sites.google.com/site/adrianstegovec/papers/stegovec_2012_ vipavavowels_bathesis.pdf 1

Transcript of The Vowel System of the Dialects of the Vipava Valley in Government Phonology

The Vowel System of the Dialects of the VipavaValley in Government Phonology∗

Adrian Stegovec

University of Ljubljana

1 Introduction

It is frequently stated in Slovenian linguistic literature that Slovenian is one of the most,if not the most, dialectically diverse Slavic languages. However, most analyses of thesevariations are diachronic in nature. And especially for phonological analyses, there hasbeen virtually no work done from a generative perspective on language variation.

This paper has two major goals: To show the advantages of Government Phonology inanalyses of phonological variation and to apply the framework to a specific case of dialectalvariation, namely the two dialect groups of the Vipava Valley, in order to determinewhether the division based on diachronic data is reflected in the current state of theirvowel systems.

The paper is organized as follows. Section 2 presents the traditional analysis of thevowel systems in question. Section 3 is a short overview of the principles and methods ofGovernment Phonology. These are then applied in Section 4 to the analysis of the vowelsystems of Vipava Valley dialects. Section 5 discusses some potential areas of furtherresearch. Section 6 concludes the paper with an overview of its findings.

2 The Dialects of The Vipava Valley

The Slovenian dialects of the Vipava Valley (VVS hereafter) are normally not consid-ered a homogeneous group. Traditional dialectologists place them within Western Slove-nian/Littoral dialects based on recent language changes (Logar and Rigler 1983; Smole

∗This paper is a shortened English version of my BA thesis (Stegovec 2012). The full Slove-nian text is available at: https://sites.google.com/site/adrianstegovec/papers/stegovec_2012_vipavavowels_bathesis.pdf

1

3 A Short Overview of Government Phonology

2000; 2004), but divide them into two groups based on older changes: the Karst dialectsin the west, and the Inner Carniolan dielects in the east.

Both groups have a vowel system with both monophthongs and diphthongs. Butwhat is especially interesting is the fact that despite individual vowels having a differentdistribution across lexical units (‘bicycle’ : ["ku:@lo] (Šempeter pri Gorici) / [k@"lu:] (Pod-nanos)), and different historical origins, the two vowel systems are nearly identical. InnerCarniolan VVS dialects are claimed to have the vowel system presented in Figure 1.1 Asseen in Figure 2, the Karst dialects have a very similar system. The two vowel systems do,however, exhibit some important differences, such as the presence of heavy diphthongs/e:i

“/ and /o:u

“/ in Inner Carniolan dialects. This will be crucial in their analysis within

the framework of Government Phonology (GP).

Stressed vowels:Long:i: u:

i:@ u:@e: o:

e:i“

o:u“i

“e: u

“o:

a:Short:

@Unstressed vowels:

i u

e @ o

a

Figure 1: Vowel system of Inner Carniolan VVS

3 A Short Overview of Government Phonology

Government Phonology is far from being a homogeneous framework, as it covers severaldistinct current approaches to phonology: CVCV phonology (Lowenstamm 1996; Scheer2004), VC phonology (Szigetvári 1999), x-bar phonology (Rennison and Neubarth 2003)and the developing framework sometimes called GP 2.0 (Pöchtrager 2006; 2009; 2010;

1Rustja (2001) claims that the Inner Carniolan dialect of the village Skrilje contrasts a long and shortstressed semivowel. This claim is unusual, as the long version is supposed to only be possible before /r/.A more sensible approach is to simply claim that this is merely a process of lengthening.

2

3.1 Goals and Principles

Stressed vowels:Long:i: u:

i:@ u:@e: o:

i“e: u

“o:

a:Short:

@Unstressed vowels:

i u

e @ o

a

Figure 2: Vowel system of Karst VVS

Pöchtrager and Kaye 2010; Živanović and Pöchtrager 2010). In this paper, however, Imostly deal with so-called Standard GP, as all other approaches stem from it.

3.1 Goals and Principles

In Kaye, Lowenstamm, and Vergnaud (1985), considered the beginning of GP, a researchprogram (Lakatos 1970; 1978) with the following goals was set up:

[ . . . ] phonology is to be regarded as a system of universal principles definingthe class of human phonological systems. [ . . . ] A complete phonologicalsystem consists, then, of these principles along with sets of parameter values(Kaye et al. 1985: 305).

This makes GP closer to generative syntax than to classic generative phonology(Chomsky and Halle 1968), as it does not assume traditional language specific phono-logical rules, but abstract universal phonological structures, expressions and processes,which are then set by parameters for each individual language

As a research program, it encompasses many distinct approaches to explaining thephonological component of human language. However, they all follow certain universalprinciples, that distinguish them from other phonological theories. For instance, GPis not grounded in acoustic and articulatory data, which is expressed in the followingprinciple.

The Phonological Epistemological Principle. The only source of phonological knowl-edge is phonological behavior (Kaye 2001a: 1).

3

3 A Short Overview of Government Phonology

GP also differs from most other approaches is that is seeks non-arbitrary connectionsbetween phonological phenomena and the environment in which they appear:

The Non-Arbitrariness Principle. There is a direct relation between a phonologicalprocess and the context in which it occurs (Kaye, Lowenstamm, and Vergnaud 1990: 194).

And finally, phonological processes in GP must follow non-violable principles andconstraints:

The Minimalist Hypothesis. [phonological] Processes apply whenever the conditionsthat trigger them are satisfied (Kaye 1995: 290).

3.2 Basic Formalizations

What follows is a short overview of the basic formalizations of standard GP that will beused in the analysis of the data in Section 4. It is based primarily on Kaye (2000; 2001b).

3.2.1 Elements

The basic melodic unit in GP is the element, which is independently interpretable, meaninga single element is already a pronounceable melodic expression (Harris and Lindsey 1995).In newer versions of GP elements are no longer considered internally complex bunches ofarticulatory-acoustic features (cf. Kaye et al. (1985)), but independent cognitive categories(Harris and Lindsey 1995). In this paper I follow the revised theory of elements (1).2

Because this paper deals with vowel systems A, I in U are of greatest importance.3

(1) E = {A, I, U, L, H, P}

But merely joining the elements together into sets does not produce all the necessarymelodic expressions to represent the sounds of human language. This is why the formingof expressions is governed by:

The Licensing Principle. All phonological positions save one must be licensed withina domain. The unlicensed position is the head of this domain (Kaye 1990: 306).

Thus a melodic expression has one head of the expression (H) position, which caneither be filled by an element or not, while other potential elements inside the expressionmust be operators (O). This way we get an ordered pair: (O.H). All possible expressionsare then as presented in Table 1.

2First referred to as such by Ploch (1998).3This does not mean that L in H never appear in vowels. Vowels contain them in tonal languages

(L — low; H — high) or in nasal vowels (L — nasality). Since such vowels do not exist in the dialects

4

3.2 Basic Formalizations

(A) (A.I) (A.I.U) (A. ) (A.I.U. )(I) (A.U) (A.U.I) (I. ) ( )(U) (I.A) (I.U.A) (U. )

(I.U) (A.I. )(U.A) (A.U. )(U.I) (I.U. )

Table 1: Possible combinations of elements in melodic expressions

This hierarchical relation makes the realization of the melodic expression more affectedby the head element, which is why closed [e]: (A.I), is closer to [i]: (I), while open [E]:(I.A), is closer to [a]: (A). The same goes for tense [i]: (I) and lax [I]: (I. ).4

But as vowel systems use different sets of expressions, they must be parametrized.Later versions of standard GP achieve this with licensing constraints (Kaye 2001b), whichlimit the licensing of elements inside melodic expressions. Below is an example of thepossible melodic expressions in Finnish (Table 2) derived with the constraints in (2).5

(A) [a] (A.I) [e] (A.I.U) [o] (A. ) (A.I.U. )(I) [i] (A.U) [o] (A.U.I) (I. ) ( )(U) [u] (I.A) [E] (I.U.A) (U. )

(I.U) [u] (A.I. )(U.A) (A.U. )(U.I) (I.U. )

Table 2: Melodic expressions of the vowel system of Finnish

(2) i. All expressions are headed

ii. U cannot be an operator

3.2.2 Structure and Constituents

In GP, like in one of its predecessors, Autosegmental Phonology (Goldsmith 1976), amelodic expression is pronounced only if connected to a skeletal point (3a), which is partof a skeleton of a phonological string (3b).6

I will be looking at I shall assume they are not present in the melodic expressions of vowels. Contrarilyit is true for P that it never appears in vowels, which is why Jensen (1994) believes it does not exist andshould be replaced with more structure. This is also assumed in GP 2.0 (Pöchtrager 2006).

4An empty expression is realized as a semivowel [1].5Because the term licensing has different roles in GP literature, the constraints will be stated in this

paper without actually using the term.6The letters in this and other diagrams are merely symbolic in nature and do not reflect the actual

content of phonological expressions.

5

3 A Short Overview of Government Phonology

(3) a. [ ] ×

a

[a] ×

a

b. ×

s

×

t

×

r

×

i

×

N

The realization of the melodic expression differs regarding what phonological con-stituent dominates it. If the skeletal point is dominated by an onset (O), the expressionis realized as a consonant, while if dominated by a nucleus (N) it is realized as a vowel.7

This means the element U will be [v] in an onset, but [u] in a nucleus.8

The third phonological constituent is called a rhyme (R). And the relationship betweennuclei and rhymes is universally constrained as presented in (4).

(4) i. Every rime must dominate a nucleus along its head path.

ii. Every nucleus must be dominated by a rime.

The distribution of nuclei and onsets is universally set by the constraints in (5).

(5) i. Every nucleus can and must license a preceding onset.

ii. Every onset must be licensed by a following nucleus.

iii. Every constituent licenser must dominate a skeletal point.

These constraints make (6a&b) grammatical and (6c) ungrammatical.

(6) a. O

×

R

N

×

b. O R

N

×

c. *O

×

R

N

The relationships between the components of each constituent are asymmetrical andonly one can and must be the head.9 The head governs all other components, andthe internal composition is determined by the two principles below, which predict theungrammaticality of the structures in (7).

Strict Directionality. The direction of government is universal (within constituents thehead is always on the left).10

Strict Locality. The two positions in a government relationship must be adjacent at therelevant level of projection.

7Heads of melodic expressions are underlined; if no element is the head nothing is underlined.8The difference in phonetic realization is not always necessary— I can essentially only get a different

function without a different realization.9The head is marked by a thicker line from the projection to its head.

10This is exactly opposite with government on higher levels. See the following section.

6

3.2 Basic Formalizations

(7) a. *X

Y Z W

b. *X

Y Z W

c. *X

Y Z W

From these two principles we can derive the following universal restriction on con-stituent structure:

The Binarity Theorem. No constituent may dominate more than two positions.

These principles limit the set of possible phonological structures to the ones in (8).They represent: a short vowel (8a-i), a long vowel (8a-ii), a rhyme with a vowel andcomplement (8a-iii),11 a simplex onset (8b-i) and a branching onset (8b-ii).

(8) a. Rhymes: b. Onsets:i. R

N

×

ii. R

N

× ×

iii. R

N

× ×

i. O

×

ii. O

× ×

But there are also other possible structures, which are crucial for the topic of thispaper. They are nuclei that branch under the skeletal point, or two melodic expressionsattached to a single point (9a),12 which realize as light diphthongs. In contrast to heavydiphthongs, which are phonologically long (9b), light diphthongs can also appear as ahead of a branching nucleus (9c) or rhyme (9d), without violating binarity.13

(9) a. R

N

×

α β

b. R

N

×

α

×

β

c. R

N

×

α β

×

γ

d. R

N

×

α β

×

γ

11The rhymal complement is not a coda. See the following section.12In Section 4.2.7 an alternative analysis that treats these structure together with regular nuclear

expressions is presented.13Kaye (2000) proposes that such structures are only possible in nuclei, making (a&b) ungrammatical.

(a) O

×

α β

* (b) R

N

×

α

×

β γ

*

Some authors, however, believe that at least (a) should be grammatical, and that it represents af-fricates. Rennison (1998) and Rennison and Neubarth (2003) also propose that this is the structure ofbranching onsets.

7

3 A Short Overview of Government Phonology

In GP, all phonological structures, as well as their relative distribution remain thesame throughout any derivation. This the consequence of:

The Projection Principle. Governing relations are defined at the level of lexical repre-sentation and remain constant throughout phonological derivation (Kaye et al. 1990: 221).

The importance of this principle will become evident in the next section.

3.2.3 Empty Expressions

In GP a word final consonant is not considered a coda (10). The argumentation for thisclaim is presented in Kaye et al. (1990); Kaye (1990); Gussmann and Harris (1998).

(10) CVC(C)#: a. O

×

r

R

N

×

E

O

×

p

b. O

×

m

R

N

×

a

×

r

O

×

s

However, GP does have a coda-like object — the rhymal complement (8a-iii; 10b),but its distribution is much more limited. As seen above, it cannot appear word finally:

The ‘Coda’ Licensing Principle. Post-nuclear rhymal positions must be licensed by afollowing onset (Kaye 1990: 311).

The rhymal complement is thus only possible if governed by the following onset (11).

(11) O

×

m

R

N

×

a

×

r

O

×

s

The skeletal point in the onset and the one in the complement are adjacent, butgovernment takes place in the opposite direction as in constituent internal cases. This isnot a violation of strict directionality, as the direction of government is fixed with respectto individual projection levels — in the case of transconstituent government it is alwaysright to left.

Slovenian, like most Slavic languages, has many cases of V/∅ (vowel-zero) alternation(12), which on the surface appear as a non-local morphological process, where a semivowel(empty expression) disappears when a suffix vowel is added.14

14In colloquial standard Slovenian this is almost without exception a semivowel. However, in othervariants of Slovenian and other languages, cases where non-empty vowels alternate are possible. This is

8

3.2 Basic Formalizations

(12) a. ["kam@n] ‘stone’ ∼ ["kamna] ‘stone (gen.)’

b. ["mac@k] ‘tomcat’ ∼ ["macka] ‘tomcat (gen.)’

Because of the projection principle, this alternation cannot be explained by a changein phonological structure, which means that underlyingly both alternating forms musthave the same structure. For the words in (12) this means: /"kam∅n∅/ and /"mac∅k∅/(‘∅’ marks empty expressions).15

For some reason, the empty expressions sometimes remain unpronounced, yet they arepronounced in the other form ([@]). GP regulates this with the Empty Category Principle(ECP), which correctly predicts which expressions are and which are not pronounced.

The (Phonological) Empty Category Principle. A p-licensed (empty) category re-ceives no phonetic interpretation.

p-licensing.

1. Domain-final (empty) categories are p-licensed (parameterised).

2. Properly governed (empty) nuclei are p-licensed.

Proper Government (p-government). α properly governs β if:

1. α and β are adjacent on the relevant projection,

2. α is not itself p-licensed, and

3. neither α nor β are not government licensers.

As transconstituent government takes place from right to left, the final nucleus canp-license the preceding nucleus, making it silent (13). Strict locality does not forbid this,as the two nuclei are adjacent at the level of nuclear projection.

(13) ["kamna]: O

×

k

R

N

×

a

O

×

m

R

N

×

O

×

n

R

N

×

a

normally explained with so-called floating elements, which are not connected to the skeletal point whenunpronounced.

15Because only phonological structure is relevant for GP, most cases in this paper will be transcribedphonetically. Rough phonological transcriptions will only be used in cases where the difference betweenthe underlying and surface form has to be pointed out.

9

3 A Short Overview of Government Phonology

Now only the empty nucleus at the end of the word has to be explained. Its presenceis predicted from the universal structure requirements, however, every speaker of Englishor Slovenian knows, that words also end in a consonant. But not every language behaveslike this. Italian and Japanese, for instance, do not allow words to end in a consonant.GP explains this with parametrized p-licensing of the Final Empty Nucleus (Fen). Thisexplains why some languages allow it, while others do not (Gussmann and Harris 1998).

["kam@n] is then a word, in which the final empty nucleus is p-licensed. And becausep-licensed positions cannot p-govern, they also cannot p-license the preceding nucleusand the empty nucleus remains pronounced (14).

(14) ["kam@n]: O

×

k

R

N

×

a

O

×

m

R

N

×

[@]

O

×

n

R

N

×

Fen

Proper government is also impossible when any of the involved positions is a governoritself, which can be illustrated with the word [p@r"vak] ‘champion’ (15), where the emptyexpression cannot be p-licensed, because there is an established governing domain betweenthe onset and the rhymal complement it is licensing.

(15) [p@r"vak]: O

×

p

R

N

×

[@]

×

r

O

×

v

R

N

×

a

O

×

k

R

N

×

Fen

As Charette (1988: 40) points out, this is nothing but a phonological version of TheMinimality Condition (Chomsky 1986: 42). This makes the connection between GP andgenerative syntax even more apparent.

In Section 3.2.2 it was said that constituents cannot contain more than two skeletalpoints. But the Slovenian examples in (16) seem to violate this restriction.

(16) a. [s"traX] ‘fear’ b. [s"krit] ‘hidden’ c. [s"krat] ‘dwarf ’d. [z"drawje] ‘health’ e. ["tekst] ‘text’ f. ["ekstra] ‘extra’

Such consonants clusters, which all begin with [s], [z], [s] or [z], are not only prob-lematic for Slovenian, as they are found in countless other languages. According to the

10

4.1 Basic Formalizations

theory presented so far, they cannot be branching onsets or rhymal complements fol-lowed by onsets. Kaye (1992) provides convincing empiric arguments16 that show SC(C)sequences17 are not branching onsets. He concludes they must be rhymal complementsfollowed by onsets, but admits that the p-licensing of the preceding nucleus is problem-atic, and introduces a third kind of p-licensing, he dubs magic licensing to make clear itis an unresolved problem. A less arbitrary solution has yet to be proposed either withinGP or any other approach.

p-licensing.

3. Magic licensing: SC(C) sequences p-license a preceding empty nucleus.

This type of licensing is present only with SC(C) clusters, and means that the emptynucleus within a rhyme with an S complement can be silenced (17&18).

(17) [s"treXa]: O R

N

× ×

s

O

×

t

×

r

R

N

×

e

O

×

X

R

N

×

a

(18) ["tekst]: O

×

t

R

N

×

e

O

×

k

R

N

× ×

s

O

×

t

R

N

×

The basic principles and mechanisms of GP presented in this section only partiallyreflect GP as a whole. However, they outline its effectiveness in dealing with differentphenomena using a small set of universals and parameters. These will be further put tothe test in the following section.

16Among them: phenomena connected to raddoppiamento sintattico and Italian definite articles, thepost-onset glide in some dialects of English. For more examples see Kaye (1992).

17The symbol S, SC(C) for clusters, covers all cases with [s], [z], [s] in [z].

11

4 The Vowel System(s) of the Vipava Valley Dialects

4 The Vowel System(s) of the Vipava Valley Dialects

4.1 The Data

4.1.1 Existing Sources

Existing discussions of VVS are either lexicographical (Rustja 2001) of historical (Logar1996b; Šekli 2009) in nature. Even when dealing with the current state of a dialect, itis mainly analyzed in relation to its historical origins, and the data reflects this. Somesources include longer transcriptions and have accompanying sound recordings. However,the recordings are mostly readings or recreations and not actual recordings from the field.The transcriptions are also problematic due to their inconsistency (see also Section 4.2.8).

4.1.2 Fieldwork

Because of this, additional data had to be gathered. Spontaneous speech of native speak-ers from both dialect groups was first recorded, and based on this data, a second roundof interviews was performed, specifically designed to test the hypotheses presented laterin this section. Importantly, the informants were not asked to pronounce specific words,but were asked questions where specific words and phrases were an expected part of theresponse. The data from existing sources is still listed for comparison. It is either markedwith an asterisk (*) before the name of the village/town or otherwise marked, usuallyciting its original source.

4.1.3 Transcription Conventions

Due to inconsistent, and potentially misleading, transcription conventions used for glidesin existing sources, I use my own conventions (Table 3).

El. [Ox] [Oxx] [R[Nx]x] [Nx] glide part of N / undeterminedI j j j i i

“U w w w u u“

Table 3: Transcription conventions for glides

I also depart from traditional conventions of stress marking, where it is usually markedin front of the last consonant of a cluster preceding the vowel. In this paper stress ismarked in front of the onset — branching or simplex — preceding the stressed nucleus.Both these conventions are also applied to existing transcriptions, except when the dif-ferences between the traditional and new conventions are relevant.

12

4.2 Analysis of the Vowel Systems

4.2 Analysis of the Vowel Systems

4.2.1 Monophthongs

Phonemic Analysis of Quantitative Contrasts

It is normally assumed that there are no quantitative contrasts in most dialects of VVS,and Smole (2002) mentions only a few cases of isolated dialects displaying limited con-trasts. However, even her analysis is based predominately on the retention of vowel lengthin historically long vowels, and not on current contrasts.

The semivowel seems to be problematic in this regard, as it is frequently listed as theonly short stressed vowel in systems with long stressed vowels (Logar 1993: 127–129) (seeFigures 1 and 2 in Section 2). This is problematic from both a phonetic analysis of lengthas well as a phonological contrast standpoint.

The average length of phonologically long stressed vowels is 60–150 ms, and 50–80 msin phonologically short ones (Pearce 2011: 12). Because of this partial overlap the onlyway to show that a language exhibits a contrast in vowel length is with minimal pairs.

The vowel diagrams in Figures 1 and 2 already point to a problem in the traditionalanalysis, as there are no long/short versions of the same vowel within the groups ofstressed or unstressed vowels. And there are no minimal pairs to back the claim in any ofthe data from previous analyses or in the data gathered during fieldwork. The anomaloussemivowel can be excluded the same way, as again there are no short/long minimal pairsto be found.

The length is attributed only to stressed vowels, which means it can be equated withstress itself. This is not surprising, since it is frequently assumed in GP that stressedvowels are structurally more complex that unstressed ones (Scheer and Szigetvári 2005;Pöchtrager and Kaye 2010), which could in principle also be expressed in length (see alsoSection 5.4).

Based on these findings the vowel systems of the two dialect groups can be simplifiedas presented in Figures 3 and 4, without a short/long distinction (see also Section 5.1).

4.2.2 The Melodic Content of Monophthongs

Determining the melodic content of the high vowels /i/ and /u/, and the low vowel /a/is straightforward in GP, as each of them is a realization of a single distinct element. Asthere is no additional tense/lax contrast, or any phonological processes which could beexplained by headless expressions, I conclude that /a/, /i/ and /u/ are realizations ofmelodic expressions headed by A, I in U without operators.

Because the realization of empty expressions is determined by government relations,

13

4 The Vowel System(s) of the Vipava Valley Dialects

Stressed vowels:i u

i“@ u

“@

e @ oi“e u

“o

aUnstressed vowels:

i u

e @ o

a

Figure 3: Vowel system of Karst VVS

Stressed vowels:i u

i“@ u

“@

e @ oei“

ou“i

“e u

“o

aUnstressed vowels:

i u

e @ o

a

Figure 4: Vowel system of Inner Carniolan VVS

14

4.2 Analysis of the Vowel Systems

the semivowel, if it is truly empty, will remain unexpressed in p-licensed positions.VVS dialects all display cases of V∼∅ alternation (19), however, there are also casesof semivowels that remain pronounced in p-licensed positions (Tables 4 & 5). This meansthat both stable and unstable semivowels are possible in these dialects.

(19) a. O

×

m

R

N

×

u“

o

O

×

č

R

N

×

[@]

O

×

n

R

N

×

Fenb. O

×

m

R

N

×

u“

o

O

×

č

R

N

×

O

×

n

R

N

×

a

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje

unstable

/"mu“oc@n∅/ /"k@s@n∅/ /"G@r@m∅/ /"toc@n∅/

‘strong (m.)’ ‘some (m.)’ ‘bush’ ‘punctual (m.)’/"mu

“oc∅na/ /"k@s∅na/ /"G@r∅me/ /"toc∅na/

‘strong (f.)’ ‘some (f.)’ ‘bushes (acc.)’ ‘punctual (f.)’

stable

/"@n∅/ /u"s@k∅/ /po"vi“@z@n∅/ /u"maz@n∅/

‘one (m.)’ ‘each (m.)’ ‘linked (m.)’ ‘dirty (m.)’/"@ni/ /u"s@ko/ /po"vi

“@z@nu/ /u"maz@na/

‘one (gen.f.)’ ‘each (acc.f.)’ ‘linked (n.)’ ‘dirty (f.)’

Table 4: Alternation in the pronunciation of the semivowel

Karst*Dornberk *Renče *Miren Miren Šempeter

unstable

/"k@s@n∅/ /"mi“@r@n∅/ /"mu

“oc@n∅/

‘some (m.)’ ‘Miren’ ‘strong (m.)’/"k@s∅nu/ /"mi

“@r∅na/ /"mu

“oc∅na/

‘some (acc.f.)’ ‘Miren (gen.)’ ‘strong (f.)’

stable

/po"vi“@z@n∅/ /u"maz@n∅/

‘linked (m.)’ ‘dirty (m.)’/"tak@/ /"k@s∅n@/ /"ri

“@tk@/ /po"vi

“@z@no/ /u"maz@na/

‘such (pl.f.)’ ‘some (pl.f.)’ ‘rare (pl.f.)’ ‘linked (n.)’ ‘dirty (f.)’

Table 5: Alternation in the pronunciation of the semivowel

Stable semivowels in stressed open syllables ([u"s@k] ∼ [u"s@ko]) are not expected fromthe basic principles of government (see Section 3.2.3). However, if additional abstractstructure is assumed with the stressed syllable as the head of the word (Harris 1997), theinability of this position to be p-licensed is predicted from the exact same principles of

15

4 The Vowel System(s) of the Vipava Valley Dialects

government.18 Or as simplified by Harris (1994: 169): ‘Heads are ungoverned.’The stability of some unstressed semivowels is slightly more problematic (["tak@];

[u"maz@n] ∼ [u"maz@na]). If exceptionless rules are to be maintained, we have to assumethat there are two kinds of semivowel which are realizations of two different melodicexpressions — of which only the unstable semivowel is truly empty, while the stable vari-ant is similar to the English [2/5] (["k2t], ‘cut’ ), which is phonetically very close to thesemivowel and has the underlying expression (A. ) (Harris 1994: 112, 115).19

The weakness of this solution lies in the fact that we now have two solutions for theproblem of stressed stable semivowels, which can now either be stable due to stress itselfor because they are not really empty expressions. But this is not only a problem of theanalysis presented here, as stable semivowels present a problem for every analysis of V∼∅alternation. The influence of stress is especially problematic. For instance, Scheer (toappear) claims that alternation cannot be predicted from stress, phonetic or contrastiveproperties of vowels; we can only be certain that it will be possible in properly governedenvironments. This are of course problems, that exceed the scope of this paper, which iswhy I propose, in the absence of evidence to the contrary, that both the empty (/1/: ( ))and the non-empty semivowel (/@/: (A. )) appear in stressed and unstressed positions.

At this point only the underlying representations of /e/ and /o/ are undetermined.Traditionally they are classified as open vowels, however, as there is no open/closedcontrast in these dialects, this must be a claim based exclusively on historical data andacoustic properties.

In standard GP /e/ and /E/ contain the elements A and I, while /o/ and /O/ containA and U. If we follow the traditional classification of these vowels the head of both theseexpressions must be A. In this case the entire set of monophthongs in both dialect groupscan be captured with only a small number of constraints (20). The derivation of possibleexpressions is presented in Table 6.

(20) i. I/U can only be operators if the melodic expression is headed.

ii. Melodic expressions headed by I/U cannot have operators.

iii. I and U cannot appear in the same melodic expression.

Since the supposed openness of the vowels in question is not determined from phono-logical properties, nothing excludes the possibility of A being the operator and /e/ and

18It should be noted here that some pairs with a perceived V∼∅ alternation within the stem, such as["p@s] ‘dog’ ∼ ["psa] ‘dog (acc.)’ and ["s@m] ‘be (1ps.sg.)’ ∼ ["smo] ‘be (3ps.pl.)’ in standard Slovenian,are merely separate lexicalized forms (cf. ‘is’∼‘are’ in English). What is important here is that ["psa]has the underlying form /p∅"sa/ with a p-licensed empty nucleus.

19The analysis of possible differences in the phonetic realization of the two variants exceeds the scopeof this paper.

16

4.2 Analysis of the Vowel Systems

*(I/U. ) = / *(E.I/U) = \ *(I&U) = —(A) /a/ XXXX(A.I) XXXX(A.I.U) (A. ) /@/ ���

���(A.I.U. )(I) /i/ XXXX(A.U) XXXX(A.U.I) ���

�(I. ) ( ) /1/(U) /u/ (I.A) /E/ (I.U.A) ��

��(U. )XXX(I.U) ���

�(A.I. )(U.A) /O/ ��

���(A.U. )XXX(U.I) ���

�(I.U. )

Table 6: Licensing constraints for monophthongs

/o/ being open only in comparison to their standard Slovenian closed equivalents. In thiscase only one, much simpler, constraint is needed (21). The derivation is presented inTable 7.

(21) i. I/U cannot be operators.

*(I/U.X) = —(A) /a/ (A.I) /e/ (A.I.U) (A. ) /@/ (A.I.U. )(I) /i/ (A.U) /o/ (A.U.I) (I. ) ( ) /1/(U) /u/ (I.A) (I.U.A) (U. )

(I.U) (A.I. )(U.A) (A.U. )(U.I) (I.U. )

Table 7: Licensing constraints for monophthongs

Of course simplicity does not guarantee the correctness of this analysis, however,what is important here is that no matter which solution is correct the whole set ofmonophthongs in both dialect groups can be described with a restricted set of elements,constraints and principles. After all, the whole list of expressions presented in Table 820

was derived from principles that were not introduced solely for this group of dialects andwith minimal parametrization.

4.2.3 The w∼v Alternation

Melodic expressions containing U may be expressed in a number of ways in onset posi-tions, or are at least limited to specific environments. In standard Slovenian, [w] cannotappear in onsets, where only [v] is found (cf. Toporišič et al. 2001: 73 (§640)) (22).21

20There is no phonological evidence at this point to determine the openness or closeness of /e/ and/o/ na tej točki sicer ni. However, due to the simplicity of the constraint in the case of their closeness Idecided to represent in this table the set represented in Table 8. Additional motivations for this choicewill be partially presented in Section 5.2.

21This difference is normally expressible in standard GP with a different function of U in an expression([w]: (U. ), [v]: (U)) (Cyran 2010: 40). While in GP 2.0 (Pöchtrager 2006) this is expressible with a

17

4 The Vowel System(s) of the Vipava Valley Dialects

Stressed monophthongs/a/ /i/ /u/ /@/ /1/ /e/ /o/(A) (I) (U) (A. ) ( ) (A.I) (A.U)

Unstressed monophthongs/a/ /i/ /u/ /@/ /1/ /e/ /o/(A) (I) (U) (A. ) ( ) (A.I) (A.U)

Table 8: Melodic content of monophthongs

(22) a. *["wiski]; ["viski] ‘whiskey’ b. *["wotka]; ["votka] ‘vodka’c. *["wermut]; ["vermut] ‘vermouth’ d. *["wat]; ["vat] ‘watt’e. *["twoj]; ["tvoj] ‘your’ f. *["hwala]; ["hvala] ‘thank you’g. *["kwackati]; ["kvackati] ‘to crochet’ h. *[cwe"kati]; [cve"kati] ‘to babble’

In addition [v] is not possible in word final positions (23), which seems problematic,as these are onsets in GP as well. This could be seen as evidence for U being the secondpart of heavy diphthong in these positions.

(23) a. *["pav]; ["paw] ‘peacock’ b. *["ziv]; ["ziw] ‘alive’c. *[s"pev]; [s"pew] ‘song/poem’ d. *["nov]; ["now] ‘new’

An alternative analysis is also available, as examples (22) and (23) prove only that [w]does not appear before pronounced nuclei. As we can see with V∼∅ alternation cases in(24) [v] changes into [w] when the following nucleus is p-licensed and silenced. The samealternation is observed when a suffix vowel is added after a final [Vw] sequence (25).

(24) a. ["briv@c] ‘barber’ ∼ ["briwca]; *["brivca] ‘barber (gen.)’b. ["dav@n] ‘ancient (m.)’ ∼ ["dawna]; *["davna] ‘ancient (f.)’c. ["ov@n] ‘ram’ ∼ ["owna]; *["ovna] ‘ram (gen.)’d. ["rev@n] ‘poor (m.)’ ∼ ["rewna]; *["revna] ‘poor (f.)’

(25) a. ["paw] ‘peacock’ ∼ ["pava]; *["pawa] ‘peacock (gen.)’b. ["ziw] ‘alive (m.)’ ∼ ["ziva]; *["ziwa] ‘alive (f.)’c. [s"pew] ‘song/poem’ ∼ [s"peva]; *[s"pewa] ‘song/poem (gen.)’d. ["now] ‘new (m.)’ ∼ ["nova]; *["nowa] ‘new (f.)’

If [Vw] sequences are really diphthongs this alternation cannot be explained becausethe projection principle of GP prohibits resyllabification, which would be required in caseswhere of alternation with [Vv]. The distribution of [v] and [w] can then be captured bysaying that [v] appears only in front of pronounced nuclei (26).

larger or smaller structure. This is not, however, relevant for the following test.

18

4.2 Analysis of the Vowel Systems

(26) a. O

×

u

[v]

R

N

×

V

b. O

×

C

×

u

[v]

R

N

×

V

c. O

×

u

[w]

R

N

×

Fen

d. O

×

u

[w]

R

N

×

O

×

C

R

N

×

V

e. O

×

C

R

N

×

V

×

u

[w]

O

×

C

R

N

×

(V)

Phenomena tied to this exact same criteria in other languages (cf. [r] in non-rhoticdialects of English appears only in front of pronounced nuclei (Harris 1994: chap. 5))further support this solution. This is why several authors, such as Cyran (2010), believethat: (a) pronounced nuclei are stronger licensers, (b) nuclei are stronger licensers thatonsets, and (c) headed expressions (like the Slovenian [v]) require a stronger licenser,which captures the representations in (26) perfectly. And as seen in (27), the w∼valternation is also present in Karst and Inner Carniolan VVS, which will be crucial inthe following analysis of diphthongs.

(27) a. Inner Carniolan:• Dobravlje: ["cav@n] ‘Čaven’ ∼ ["cawna]; *["cavna] ‘Čaven (gen.)’b. Karst:• Miren: ["rav@n] ‘straight (m.)’ ∼ ["rawna]; *["ravna] ‘straight (m.)’• Šempeter: ["glav@n] ‘head (m.)’ ∼ ["glawna]; *["glavna] ‘head (f.)’

4.2.4 Definitions of Diphthongs

Traditional Slovenian dialectological literature does not offer a consistent definition ofa diphthong, especially not one based on phonological behavior. The only criteria forsomething to be considered a diphthong is if its parts historically originate from a singlemonophthong, which is not an adequate definition for a synchronic analysis.

19

4 The Vowel System(s) of the Vipava Valley Dialects

On the other hand, GP does have a clear and restricted definition of diphthongs, asseen in (9) in Section 3.2.2. They are structurally further divided into light (28a) andheavy diphthongs (28b) based on the number of skeletal points in the nucleus.

(28) a. N

×

α β

b. N

×

α

×

β

Apart from these structural constraints, the melodic content of diphthongs is addi-tionally limited by the following requirements:

Head of light diphthong: no general requirement.

Governee of light diphthong: simplex, no A.

(Kaye 2000: 9)

The second requirement means that in light diphthongs the first position, which isgoverned, will only contain elements that are not A, meaning I and U. This makes thefollowing sequences all potential light diphthongs: [i

“@], [u

“@], [i

“a], [u

“a], [i

“u], [u

“i], [i

“e], [u

“o],

[i“o] in [u

“e].22 Their diphthong status will be tested in the following section following the

principles of standard GP.The list above does not include Inner Carniolan /ei

“/ and /ou

“/, traditionally also

assumed to be diphthongs. This is because they are excluded as light diphthongs by themelodic requirements presented above. This means that if they are truly diphthongs,they must be heavy diphthongs. This will be put to the test in Section 4.2.8.

4.2.5 Light Diphthongs

4.2.6 The Structure of Light Diphthongs

From all the candidates for light diphthongs established in the previous section traditionalliterature describes only /i

“@/, /u

“@/, /i

“e/ and /u

“o/ as diphthongs in VVS. Based on this

we can construct a working hypothesis:

Hypothesis 1. In all the dialects of VVS only the following sequences of glides andmonophthongs can be light diphthongs: [i

“@], [u

“@], [i

“e] in [u

“o].

If these sequences are truly light diphthongs, they have the structure in (28a). Fol-lowing the binarity theorem we can then conclude that: (a) only real light diphthongscan appear after branching onsets (29), and (b) in sequences that are not real diphthongsthe glide (I/U) must be an onset (30a) or the second position in a branching onset (30b).

22Disyllabic sequences such as [i"on] ‘ion’ will not be considered suitable candidates for diphthongs.

20

4.2 Analysis of the Vowel Systems

(29) a. O

×

C

×

C

R

N

×

i/u V

b. O

×

C

×

C

×

i/u

R

N

×

V

*

(30) a. O

×

i/u

R

N

×

V

b. O

×

C

×

i/u

R

N

×

V

In standard Slovenian [r] cannot govern, which means that in clusters where it is thefirst component, it must be a rhymal complement. This can be seen in cases where wordinitial [rC] cannot be interpreted as a branching onset (31). As [r] can only be governed,it is interpreted as a complement of the rhyme head [@] (cf. Toporišič et al. 2001: 137).

(31) a. *["rdec]; [@r"dec] ‘red’ b. *["rman]; [@r"man] ‘yarrow’

If we assume that in VVS, just like in standard Slovenian (32), [r] cannot govern thefollowing I/U, the sequence of [r] and a glide will have the structure in (33a) in case ofa light diphthong, and (33b) with a fake light diphthong.23

(32) a. *["rjava]; [@r"java] ‘brown’ b. *["rjuha]; [@r"juha] ‘bed sheet’

(33) a. O

×

r

R

N

×

i/u V

b. O R

N

×

@

×

r

O

×

i/u

R

N

×

V

Based on this and hypothesis 1 we can derive the following predictions:

Prediction 1.1. Out of the possible sequences of glides and monophthongs only [i“@], [u

“@],

[i“e] and [u

“o] will be found after branching onsets.

23In contrast to [r], the other two “bad” governers — [n] and [l] (*[#nCV]; *[#lCV]) — can be thehead of a branching onset, if the second position is filled by I ([p@s"ti

“@lja] ‘bed’, [n@d@lj@"vaNka] ‘TV series’,

[u"rinj@no] ‘inserted’, [s"vakinja] ‘daughter in law’ ). However, this is not possible with U, which can beused as additional motivation to look for universal differences in the distribution of I in U, as it is thepractice in GP 2.0 (Pöchtrager 2009; Živanović and Pöchtrager 2010).

21

4 The Vowel System(s) of the Vipava Valley Dialects

Prediction 1.2. [r] preceding sequences of glides and monophthongs which are not lightdiphthongs will always be a complement of the preceding rhyme — consequently it will notbe present in word initial positions unless preceded by [@] or some other vowel.

From the data we can see that both in Inner Carniolan (Table 9) and Karst dialects(Table 10) prediction 1.1 holds up. And prediction 1.2 holds as well with sequences withan I glide (Tables 11 and 12).

Inner Carniolan#[Oxx] + *Predmeja *Skrilje Podnanos Dobravlje

[i“@] ["Gri

“@sta] ["kli

“@t] ["tri

“@s@jo] ["kli

“@t]

‘they (dl.) are going’ ‘basement’ ‘they are shaking’ ‘basement’[u“@] ["Gru

“@zdje] ["pru

“@sis] ["pru

“@sjo]

‘grapes’ ‘you are asking’ ‘they are asking’[i“e] ["bri

“ezna] ["cvi

“ek] ["pri

“ese] ["cvi

“ek]

‘pit (gen.)’ ‘nail’ ‘small pig’ ‘nail’[u“o] ["Gru

“oznu] ["Gru

“ost] ["kru

“osnjo] [u"tru

“oci]

‘terrible (n.)’ ‘a grape’ ‘treetop (acc.)’ ‘children’*[i“a]

*[u“a]

*[i“u]

*[u“i]

*[i“o]

*[u“e]

Table 9: Sequences of glides and vowels following branching onsets

Karst#[Oxx] + *Dornberk *Renče *Miren Miren Šempeter

[i“@] ["pri

“@ci] [n@"pli

“@t@t] ["cri

“@s] [po"tri

“@s] ["tri

“@X]

‘immediately’ ‘to wind up’ ‘over’ ‘earthquake’ ‘three (gen.)’[u“@] ["pru

“@ti] ["Glu

“@da] ["kru

“@X]

‘against’ ‘he is munching’ ‘circle’[i“e] ["cvi

“ek] ["pli

“eX]

‘nail’ ‘sheet metal’[u“o] [o"tru

“oci] ["pru

“oc] ["Gru

“ost]

‘children’ ‘away’ ‘a grape’*[i“a]

*[u“a]

*[i“u]

*[u“i]

*[i“o]

*[u“e]

Table 10: Sequences of glides and vowels following branching onsets

22

4.2 Analysis of the Vowel Systems

Inner Carniolan#[r] + *Predmeja *Skrilje Podnanos Dobravlje[i“@] ["ri

“@ku] ["ri

“@ku] ["ri

“@ku]

‘he said’ ‘he said’ ‘he said’[i“e] ["ri

“ece] ["ri

“ece] ["ri

“ece] ["ri

“ect]

‘he says’ ‘he says’ ‘he says’ ‘to say’![i“a] [@r"i

“ave] ! [@r"i

“ava] !

‘brown (pl.f.)’ ‘brown (f.)’![i“u] [@r"i

“uXe] ! /@r"i

“uXe/ !

‘bed sheets’ ‘bed sheets’![i“o] [@r"i

“ovu] ! /@r"i

“ove/ !

‘he roared’ ‘he is roaring’

Table 11: Sequences of glides and vowels following word initial [r]

Karst#[r] + *Dornberk *Renče *Miren Miren Šempeter[i“@] ["ri

“@ku] ["ri

“@tk@] ["ri

“@s] ["ri

“@s] ["ri

“@ku]

‘he said’ ‘rare (pl.f.)’ ‘true’ ‘true’ ‘he said’[i“e] ["ri

“ece] ["ri

“ekla]

‘he says’ ‘she said’![i“a] [@r"i

“ava] ! [@r"i

“ava] !

‘brown (f.)’ ‘brown (f.)’![i“u] /@r"i

“uXe/ ! /@r"i

“uXe/ !

‘bed sheets’ ‘bed sheets’![i“o] /@r"i

“ove/ ! [@r"i

“ove] !

‘is roaring’ ‘he is roaring’

Table 12: Sequences of glides and vowels following word initial [r]

23

4 The Vowel System(s) of the Vipava Valley Dialects

With U glide examples the difference between real and apparent diphthongs is evenmore evident due to the w∼v alternation. Both with the testing of prediction 1.2 (Table13 and 14) as in other environments, (Table 15 and 16) [ua], [ui] and [ue] are impossible,and they realize in front of pronounced nuclei as [va], [vi] in [ve] (Table 15 and 16).The only potential counterexample is [u

“a] in Table 16 (Dornberk: ["u

“asm@]). It will be

discussed in more detail in Section 5.3. For now, suffice it to say it is only a differentphonetic realization of the diphthong /u

“o/, found only word initially.

Inner Carniolan#[r] + *Predmeja *Skrilje Podnanos Dobravlje[u“@] ["ru

“@cnu] ["ru

“@cni] ["ru

“@ze] ["ru

“@X]

‘by hand’ ‘hand (adj.m.)’ ‘flowers’ ‘horn’[u“o] ["ru

“oc] ["ru

“oka] ["ru

“osa]

‘handle’ ‘hand’ ‘dew’*[u

“a]

*[u“i]

*[u“e]

Table 13: Availability of the [ru] sequence

Karst#[r] + *Dornberk *Renče *Miren Miren Šempeter[u“@] ["ru

“@z@] ["ru

“@z@] ["ru

“@cnu]

‘flowers’ ‘flowers’ ‘by hand’[u“o] ["ru

“osa] ["ru

“oc]

‘dew’ ‘handle’*[u

“a]

*[u“i]

*[u“e]

Table 14: Availability of the [ru] sequence

We can also conclude that there are no differences between the two dialect groups interms of light diphthongs. In both groups we find only the following: /i

“@/, /u

“@/, /i

“e/ in

/u“o/. I propose they have the structure in (34). In (34a&b) I/U spreads to the position

of A and affects the phonetic realization of the head (I + A = [e]; U + A = [o]). Whilein (34c&d) the melodic material does not spread from the glide, and the head retains itssemivowel realization [@].

(34) a. /i“e/ N

×

i a

b. /u“o/ N

×

u a

c. /i“@/ N

×

i ∅

d. /u“@/ N

×

u ∅

24

4.2 Analysis of the Vowel Systems

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje

[u“@] ["Gu

“@zda] ["Gu

“@stu] ["ku

“@s] ["Xu

“@di]

‘forest (gen.)’ ‘dense’ ‘piece’ ‘he walks’[u“o] ["du

“ol] ["du

“obru] ["Gu

“or] ["ku

“oGa]

‘down’ ‘good’ ‘up’ ‘who (acc.)’*[u

“a] ["dva] ! ["dvakr@t] ! ["Xvala] ! ["Xvala] !

‘two’ ‘twice’ ‘thank you’ ‘thank you’*[u

“i] ["vidli] ! ["dviG@nlu] ! ["Gvis@n] ! ["dviG@nt] !

‘they saw’ ‘it lifted’ ‘sure (m.)’ ‘to lift’*[u

“e] ["vec] ! ["vec] ! ["vec] ! ["dvei

“] !

‘more’ ‘more’ ‘more’ ‘two (acc.)’

Table 15: Realization of (C +) u + V sequences

Karst*Dornberk *Renče *Miren Miren Šempeter

[u“@] [s"ku

“@zi] [o"ku

“@li] ["tu

“@ka] ["tu

“@rte] ["tu

“@i“]

‘through’ ‘around’ ‘current (gen.)’ ‘cakes’ ‘your’[u“o] ["ku

“osit] ["du

“osti] ["Xu

“odit] ["ku

“on] ["tu

“oc@no]

‘to mow’ ‘a lot’ ‘to walk’ ‘horse’ ‘tap (beer)’*[u

“a] ["u

“asm@] ? ["dva] ! ["bu

“@X "var] ! [s"tvar] ! ["dva] !

‘eighth (gen.)’ ‘two’ ‘god help us!’ ‘thing’ ‘two’*[u

“i] ["vidu] ! [na "vistu] ! ["vis@n] ! [od"visno] ! ["Gvisno] !

‘he saw’ ‘to the draft’ ‘sure (m.)’ ‘it depends’ ‘surely’*[u

“e] ["vez@] ! ["dve] !

‘connections’ ‘two (acc.)’

Table 16: Realization of (C +) u + V sequences

25

4 The Vowel System(s) of the Vipava Valley Dialects

/i“@/ in /u

“@/ in principle allow an alternative analysis, as similar diphthongs are an-

alyzed in English as heavy diphthongs (Harris 1994: 116). However, if they were trulyheavy diphthongs in these dialects as well, they would be impossible in branching rhymesdue to the binarity theorem, which we can see is not true in Table 17.

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje["Gu

“@zda] ["su

“@rta] ! ["du

“@Xt@rj@m] ! ["u

“@Xc@t] !

‘forest (gen.)’ ‘type’ ‘doctors (dat.)’ ‘wedding’Karst

*Dornberk *Renče *Miren Miren Šempeter[n@ "mi

“@sti] ["su

“@st@"rai

“] ["tu@rte] ! [pon"di

“@lk] !

‘on the spot’ ‘woodworks’ ‘cakes’ ‘monday’

Table 17: Diphthongs /i“@/ and /u

“@/ in heavy rhymes

Both sequences appear also in front of rhymal complements followed by onsets thatare not exceptional SC(C) sequences. This means that they must be light diphthongs.

4.2.7 Constraints on the Melodic Content of Light Diphthongs

The suggested structures correctly predict the distribution of light diphthongs. They donot, however, in themselves explain their melodic content and why it is so restricted.If we wanted to exclude the unattested expressions in (35), we would have to pose theconstraints in (36) in addition to universal requirements (see Section 4.2.4).

(35) a. */i“u/ N

×

i u

b. */u“i/ N

×

u i

c. */i“a/ N

×

i a

d. */u“a/ N

×

u a

e. */i“u/ N

×

i u

f. */u“u/ N

×

u i

g. */i“i/ N

×

i ∅

h. */u“u/ N

×

u ∅

(36) i. I and U cannot appear in the same expression.

ii. I/U spreads, when the expression is headed.

Constraint (36ii) is, however, somewhat unusual. Partially due to its very specificnature, partially due to the questionable mechanism of spreading (see Section 5.2).

26

4.2 Analysis of the Vowel Systems

Light diphthongs become more interesting when examined together with monoph-thongs (Table 18). Surprisingly there is no overlap in melodic expressions, which wouldbe expected as a consequence of different structures. This is only possible if we assume/e/ and /o/ are closed, which can be seen as additional support for such an analysis.This joint set of expressions is derived by a single set of constraints (37).

(A) /a/ (A.I) /e/ (A.I.U) (A. ) /@/ (A.I.U. )(I) /i/ (A.U) /o/ (A.U.I) (I. ) /i

“@/ ( ) /1/

(U) /u/ (I.A) /i“e/ (I.U.A) (U. ) /u

“@/

(I.U) (A.I. )(U.A) /u

“o/ (A.U. )

(U.I) (I.U. )

Table 18: Possible expressions for monophthongs and light diphthongs

(37) i. I and U cannot appear in the same expression.

ii. Operators must be simplex (cannot contain more than one element).

iii. I/U spreads in light diphthongs, when the expression is headed.

A description of a similar case cannot be found in existing GP literature. This is whyon this point I can only tentatively propose that the difference in realization of diphthongsand monophthongs is due to some set of parameters that determine the way melodicexpressions are realized structurally. Such an analysis makes the reasons for certaindialectal variations clearer, as /i

“e/ and /u

“o/ appear in similar environments/words as

the open /E/ and /O/ in standard Slovenian. Both pairs of vowels in fact share the samemelodic expressions: (I.A) in (U.A).

Another joint analysis is possible with /i“@/ and /u

“@/ as realizations of headless ex-

pressions ((A.I. ) in (A.U. )) with the structure in (38).

(38) a. /i“@/ N

×

i a

b. /u“@/ N

×

u a

The whole set of both groups of vowels (Table 19) is then derived with the constraintsin (39).

(39) i. I and U cannot appear in the same expression.

ii. Only melodic expression containing A can have operators.

iii. I/U spreads in light diphthongs, when the expression is headed.

27

4 The Vowel System(s) of the Vipava Valley Dialects

(A) /a/ (A.I) /e/ (A.I.U) (A. ) /@/ (A.I.U. )(I) /i/ (A.U) /o/ (A.U.I) (I. ) ( ) /1/(U) /u/ (I.A) /i

“e/ (I.U.A) (U. )

(I.U) (A.I. ) /i“@/

(U.A) /u“o/ (A.U. ) /u

“@/

(U.I) (I.U. )

Table 19: Possible expressions for monophthongs and light diphthongs

While this analysis also does not solve the problem of how to explain spreading, ithas the advantage of making A a part of all light diphthong structures. This is welcome,as A is frequently associated with additional structure (Pöchtrager 2006; 2010), and thedifference between monophthongs and diphthongs is structural. An attampt to developthis analysis within GP 2.0 is briefly discussed in Sections 5.2, 5.3 and 5.4.

4.2.8 Heavy Diphthongs

Traditional Analysis and Transcription Conventions

It was mentioned before that only Inner Carniolan dialects are said to have diphthongsthat would have to be considered heavy under a GP analysis — /ei

“/ and /ou

“/.

In addition, traditional analyses, and transcription conventions used in them, allowfor a difference between /ei

“/ and /ej/, and /ou

“/ and /ow/. In the examples taken from

this analyses in Table 20 we can take out that they at least acknowledge a differencebetween /j/ and /i

“/.

dialect group dialect (source) /ei“/ /ej/ /ou

“/ /ow/

Inner Carniolan

Predmeja (Logar 1993: 61) /"ne:i“s/ /"vo:u

“ce/

‘you are not’ ‘sheep’Skrilje (Rustja 2001) /t"re:i

“ba/ /"ce:jt/ /"ko:u

“ck/

‘need’ ‘time’ ‘suitcase’

Karst

Dornberk (Logar 1993: 63) /"ce:i“t/ /pou

“d"ni:@vi/

‘time’ ‘at noon’Renče (Logar 1993: 65) /o"be:i

“sju/ /"ko:u

“ci/

‘they hang’ ‘pales’Miren (Logar 1993: 65) /v@r"te:i

“b@nci/ /k"ra:sou

“ci/

‘Vrtojbans’ ‘Karstians’

Table 20: Differences in the pronunciation/transcription of heavy diphthongs/glides incodas

In the transcriptions of Karst dialects, which are not supposed to have heavy diph-thongs, we can find /e:i

“/ and /o:u

“/ as well. In the Skrilje dialect (Rustja 2001) we

even find both /e:j/ (/"ce:jt/) and /e:i“/ (/t"re:i

“ba/). Does this make them two separate

28

4.2 Analysis of the Vowel Systems

phonological entities? More importantly, does this mean that the Dornberk /"ce:i“t/ is

phonologically different from /"ce:jt/?The search for potential phonological differences is made even harder due to the use

of /i“/ and /u

“/, traditionally used to mark parts of diphthongs, in sequences that are not

assumed to be diphthongs in neither dialect group (Table 21).

*Predmeja *Skrilje *Renče *Miren/@i

“/ /de

˙:i“va/

‘let’s (dl.)’/@u

“/ /z@u

“"re:i

“lu/

‘it boiled’/ai

“/ /"fa:i

“n/ /"ma:i

“Xna/ /"ca:i

“t@/

‘nice’ ‘small (f.)’ ‘times’/au

“/ /sp"ra:u

“let/ /s"ra:u

“f/ /pozd"ra:u

“ju/ /p"ra:u

“j@/

‘to put in order’ ‘screw’ ‘they great’ ‘they say’/ui

“/ /"Xu:i

“si/

‘worse’/iu

“/ /"bi:u

“/ /skod"li:u

“ci/ /biu

“/

‘was’ ‘pests’ ‘was’/oi

“/

/eu“/ /"buo

˙snarjeu

“mi/ /p"re:u

“c/

‘Bošnar’s’ ‘too much’/u“@i“/ /"muo

˙i“/ /"su:@i

“/ /"w@i

“sku/

‘mine’ ‘one’s own’ ‘army (acc.)’/i“@u“/

/u“oi“/ /k"vo:i

“nem/

‘horse (dat.)’/i“eu“/

Table 21: /i“/ and /u

“/ in sequences not traditionally described as diphthongs

Due to these inconsistencies it is impossible to trust previous analyses. This is whythe analysis presented here will have as a goal the establishment of a consistent criterionfor determining heavy diphthongs.

4.2.9 GP Analysis and the Case of English

In GP, the problematic glides described above must either be a rhymal complement (40a)or the second skeletal point in a nucleus (40b).

(40) a. R

N

×

V

×

i/u

b. R

N

×

V

×

i/u

29

4 The Vowel System(s) of the Vipava Valley Dialects

But as the consequences of this distinction are hard to predict, analyses may differwith regard to which constituent (R or N) is immediately dominating the glide. Harris(1994), for instance, analyzes all such cases in English as heavy diphthongs (41).

(41) a. /ei“/ R

N

×

a

i

×

b. /ou“/ R

N

×

a

u

×

c. /oi“/ R

N

×

a

u

×

i

d. /au“/ R

N

×

a

×

u

e. /ai“/ R

N

×

a

×

i

We can see (41a&b) differ from the rest because I in U are not in the governed secondposition, but merely spread their melodic material there from the head position.24

4.2.10 Slovenian

Despite the difficulties with determining the underlying structure of vowel-glide sequencesbased on their surface realization, a comparison of English and Slovenian potential heavydiphthongs reveals discrepancies that indicate a difference in underlying structure.

The set of English heavy diphthongs with a I/U glide, presented in (41), is verylimited regarding possible heads and glides. The head cannot be a high vowel or a schwa.Even more interesting is the absence of /eu

“/.25 The situation is very different in standard

Slovenian (Table 22), where glides can appear after virtually every monophthong.This freedom of distribution is more in line with an analysis where the glides are not

part of the nucleus. Additional evidence is provided by the presence of the w∼v alterna-tion (see Section 4.2.3, examples (24) and (25)). There is no equivalent of this alternationin the examples with /j/ or /i

“/, however, since GP does not allow resyllabification, they

have to be analyzed the same way (42&43).26

24In his analysis of (southern) British English long vowels, Kaye (2000; 2001b) groups /ei“/ in /ou

“/

together with long monophthongs. This is probably due to the discussed difference in the internalconfiguration of elements, because of which these expressions must be restricted with constraints onlicensing and not government, just like long monophthongs.

25See Pöchtrager (2009) for an analysis of this restriction in terms of the GP 2.0 framework.26In Table 22 we can find apparent counterexamples like: *["fev@d]; ["fewd] ∼ ["fewda]. They are not

30

4.2 Analysis of the Vowel Systems

Slovenian〈ı〉 〈ı〉 〈u〉 〈u〉 〈a〉 〈a〉

I ["fui“] ["tui

“ka] ["nai

“b@rz] ["bai

“ta]

‘yuck’ ‘loanword’ ‘probably’ ‘house (pej.)’U ["ziu

“] ["biu

“si] ["prau

“kar] [astro"nau

“t]

‘alive’ ‘former’ ‘just now’ ‘astronaut’〈e〉 〈e〉 〈e〉 〈o〉 〈o〉 〈o〉

I ["pei“d] ["sei

“k] ["sei

“m] ["Xoi

“la] ["goi

“zar] ["boi

“ler]

‘go!’ ‘sheik’ ‘fair (noun)’ ‘howdy’ ‘a boot’ ‘boiler’U ["beu

“sk] ["feu

“d] ["leu

“ji] ["bou

“Xa] ["ou

“ca]

‘bark’ ‘fief ’ ‘lion like’ ‘flea’ ‘a sheep’

Table 22: Availability of V + i/u (+ C) sequences

(42) a. ["poz@n] ‘late (m.)’ ∼ ["pozna] ‘late (f.)’b. ["zej@n] ‘thirsty (m.)’ ∼ ["zejna] ‘thirsty (f.)’c. ["baj@n] ‘fabulous (m.)’ ∼ ["bajna] ‘fabulous (f.)’d. ["buj@n] ‘lush (m.)’ ∼ ["bujna] ‘lush (f.)’

(43) a. ["maj] ‘May’ ∼ ["maja] ‘in May’b. ["tuj] ‘foreign (m.)’ ∼ ["tuja] ‘foreign (f.)’c. ["mej] ‘borders (gen.)’ ∼ ["meja] ‘border’d. ["noj] ‘ostrich’ ∼ ["noja] ‘ostrich (acc.)’

The data presented here suggests that [Vu“] and [Vi

“] sequences are always a nucleus

followed by a consonantal (rhymal complement, onset) /w/ or /j/ glide in standardSlovenian.

4.2.11 A New Analysis of the Vipava Valley Dialects

If only /ei“/ and /ou

“/ have a heavy diphthong status in Inner Carniolan dialects, all other

similar sequences like [ai“], [au

“], [@i

“], [@u

“], [ui

“], [iu

“], [oi

“] and [eu

“], have to be sequences of

a vowel and consonantal glide. This is what [ei“] and [ou

“] must also be in Karst dialects.

Based on these assumptions we can set up a working hypothesis:

Hypothesis 2. The [ei“] and [ou

“] sound sequences are heavy diphthongs only in Inner

Carniolan VVS.

I argued that in standard Slovenian [Vu“] and [Vi

“] are a vowel and consonant based

partially on the freedom in distribution their components display. If we apply the samereasoning to VVS, we can make the following prediction based on hypothesis 2:

problematic, as /w/ is a rhymal complement licensing the following onset in these cases (cf. *["p@r@t];["p@rt] ‘prt’ ∼ ["p@rta] ‘prta’ ). The same can be said about similar cases with /j/.

31

4 The Vowel System(s) of the Vipava Valley Dialects

Prediction 2.1. Sequences of monophthongs followed by glides will not display freedomof distribution in the Inner Carniolan dialects. If this is not the case, a phonologicalcontrast between /ei

“/ and /ej/ or /ou

“/ and /ow/ is expected.

Tables 23 and 24 show the same freedom of distribution in both dialect groups. Se-quences with an on- and off-glide ([i@u]; [u@i]) are also present. This has to be understoodas either proof for the existence of triphthongs or support for the freedom of distributionclaim. Most importantly, minimal pairs distinguishing /ei

“/ from /ej/ or /ou

“/ from /ow/

cannot be found, which means that prediction 2.1 does not hold.

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje

[ei“] ["nei

“s] ["trei

“ba] ["lei

“t] ["vei

“du]

‘you are not’ ‘has to’ ‘years (gen.)’ ‘I/he knew’[ou

“] ["mou

“Gla] ["mou

“Glu] ["spou

“n@m] ["mou

“Gli]

‘could (dl.)’ ‘it could’ ‘I remember’ ‘could (pl.)’[@i“] [d@i

“va] ["m@i

“n]

‘let’s (dl.)’ ‘less’[@u

“] ["n@u

“s] [k@u

“"boi

“ke]

‘you won’t’ ‘jeans’[ai“] ["fai

“n] ["fai

“n] ["mai

“co] ["mai

“co]

‘nice’ ‘nice’ ‘t-shirt (acc.)’ ‘t-shirt (acc.)’[au

“] ["prau

“jo] [s"rau

“f] ["pau

“c] ["jau

“nu]

‘they say’ ‘screw’ ‘thumb’ ‘public (acc.f.)’[ui

“] ["Xui

“si] [nai

“"Xui

“si] ["tui

“c] ["Xui

“s@t]

‘worse’ ‘worst’ ‘foreigner’ ‘to diet’[iu“] [skod"liu

“ci] ["diu

“je] ["miu

“ka]

‘pests’ ‘wild (n.)’ ‘sand’[oi“] [po"koi

“ni] [od"boi

“ka]

‘the late (m.)’ ‘volleyball’[eu

“] ["bu

“@snarjeu

“mi] [p"reu

“c] ["peu

“c] ["peu

“ka]

‘Boršnar’s’ ‘too much’ ‘singer (m.)’ ‘singer (f.)’[u“@i“] /"u

“@i“ske/ [u"bu

“@i“nu] [f@"mu

“@i“]

‘vojne’ ‘both’ ‘my boy!’[i“@u“] [u"zi

“@u“]

‘I/he took’[u“oi“] /"ku

“oi“nem/ /"ku

“oi“nu/

‘konjem’ ‘konju’*[i“eu“]

Table 23: Availability of V + i/u (+ C) sequences

In standard Slovenian, the consonantal status of the glide in [ou“] can be shown

with cases that w∼v alternate. In languages with both long monophthongs and heavydiphthongs, both can appear in word final positions (for English examples see Harris(1994: Sections 2.4.3 & 2.4.4)). Based on hypothesis 2 we can make a prediction.

32

4.2 Analysis of the Vowel Systems

Karst*Dornberk *Renče *Miren Miren Šempeter

[ei“] ["cei

“t] ["pei

“dje] [d"vei

“zGa] ["rei

“c] [s@lo"tei

“p]

‘time’ ‘he’s going’ ‘the twentieth’ ‘say!’ ‘sticky tape’[ou

“] [pou

“d"ni

“@vi] ["kou

“ci] ["krasou

“ci] ["dou

“Go] ["dou

“X]

‘at noon’ ‘poles’ ‘Karstians’ ‘long (n.)’ ‘long’[@i“]

[@u“] [k@u

“"boi

“ke]

‘jeans’[ai“] ["Gmai

“nu] ["cai

“t@] ["pai

“k] ["fai

“n]

‘common (acc.)’ ‘times (acc.)’ ‘spider’ ‘nice’[au

“] [poz"drau

“ju] ["prau

“j@] ["mau

“rica] ["pau

“c]

‘they greet’ ‘they say’ ‘rainbow’ ‘thumb’[ui

“] [s"Xui

“s@t] ["Xui

“si]

‘to lose weight’ ‘worse (m.)’[iu“] ["miu

“ka] ["miu

“ka]

‘sand’ ‘sand’[oi“] [kau

“"boi

“k@] [po"koi

“ni]

‘jeans’ ‘the late (m.)’[eu

“] ["eu

“ro] ["preu

“c]

‘Euro’ ‘too much’[u“@i“] ["u

“@i“sku] ["tru

“@i“ka] [z@s"tu

“@i“n]

‘army (acc.)’ ‘three point shot’ ‘free’[i“@u“] ["pi

“@u“c] ["pi

“@u“ka]

‘singer (m.)’ ‘singer (m.)’[u“oi“] ["ku

“oi“n]

‘horse’*[i“eu“]

Table 24: Availability of V + i/u (+ C) sequences

33

4 The Vowel System(s) of the Vipava Valley Dialects

Prediction 2.2. Cases of [ou“] in word final positions where [u

“] alternates with [v] as

well as cases where this does not occur are expected (at least) in Inner Carniolan dialects.

But the w∼v alternation is present in all cases with word final [ou“] (Podnanos:

[nj@"Gow] ‘his (m.)’ ∼ [nj@"Gova] ‘his (f.)’ ). Due to the impossibility of resyllabificationprediction 2.2 does not hold, which means that [ou

“] is never a diphthong word finally. It

is also impossible, following the principles of GP, to formulate any constraints that wouldprohibit heavy nuclei to appear word finally but not elsewhere. In the absence of anyother evidence we must then conclude [ou

“] is always /ow/ in all dialects of VVS.

The same has to be assumed for [ei“] (Podnanos: [s"mej] ‘is allowed’ ∼ [s"mejo] ‘are

allowed’ ; [s"mejm] ‘am allowed’, [po"vej] ‘he tells’ ∼ [po"vejo] ‘they tell’ ; [po"vejm] ‘I tell’ ).As a diphthong, the final /ei

“/ would have to be analyzed as /ei

“.V/ when a suffix vowel

would be added. But there is no reason to pose this or evidence to base this on.We can conclude then that /ei

“/ and /ou

“/ do not have a heavy diphthong status, not

even in Inner Carniolan dialects. Their distribution is perfectly explained if we assumethey are always /ej/ in /ow/ — a sequence of a monophthong and consonantal glide. It istrue that they are more abundant in Inner Carniolan dialects, and they might originatefrom historical monophthongs, however, nothing in their phonological behavior supportstheir supposed diphthong status.

4.3 The Status of Dialectal Variation

Based on the analysis presented in the previous section I propose that all the dialects ofVVS have the same vowel system as presented in Table 25.

Stressed UnstressedI U I Ui u i ui“@ @/1 u

“@

∅@/1

∅A

e@/2

oA e @/2 o

i“e u

“o

a a

Table 25: Vowel system of VVS

I do not, however, dispute their different historical origin, which all diachronic analysessuggest, and on which the modern division of Slovenian dialects is based (Logar and Rigler1983; Logar 1993; 1996a; Smole 2004; Šekli 2009).

Apart from distinct historical origins, another standard motivation for the distinctnessof the two dialectal groups is the fact that their native speakers perceive them as such. But

34

5.1

a closer inspection shows that these are merely minute phonetic and lexical differences.This is essentially in line with Kaye (1997), who suggests such non-systemic differences areconsequences of the double function of the human vocal channel. One of the two functionsis transferring linguistic information, while the other serves for group recognition purposesand is manifested in non-systemic phonetic variations that have no effect on the speakersphonological system.

With this and the analysis from the previous section in mind, I conclude that in theircurrent state the dialects of VVS share the same vowel system, with minute non-linguisticdifferences restricted to potential phonetic variation and individual lexical units. Sincethe dialects have distinct historical origins, this means they must have underwent a seriesof convergent changes resulting in the current shared system.27

5 Further Research

5.1 Heavy and Superheavy Rhymes

5.1.1 Superheavy Rhymes and Heavy Diphthongs

The basic principles of GP make both structures in (44) equally impossible (See Section3.2.2). This is why the structural status of [Vu

“] in [Vi

“] in Vipava Valley dialects cannot

be straightforwardly tested using their absence or presence in heavy rhymes.

(44) a. R

N

×

V

×

i/u

×

C

* b. R

N

×

V

×

i/u

×

C

*

However, it has been long known that this restriction can be violated even in languageswith attested long vowels. These so-called superheavy rhymes are assumed to have thestructure in (44a), but since it is prohibited by basic principles of GP, the problem hasto be solved by explicitly stating the environments in which it is allowed to violate them.This is what Harris (1994) does in his analysis of English superheavy rhymes.

Aside from it being unexplanatory, this analysis has other problems discussed byPöchtrager (2010). The suggested alternative analysis is formulated withing GP 2.0.However, expanding the discussion here into this theoretic approach exceeds the scope

27Some of these changes might be very recent, as one of the informants for the Podnanos dialectreports that speakers who are only one to two generations older still use [u] (["suha] ‘thin (f.)’ ), whichshe does not.

35

5 Further Research

of this paper. Additionally, Pöchtrager (2010), just like Harris, deals only with Englishexamples, and Slovenian might have a completely different set of restrictions.

Within standard GP, we can only generalize that in standard Slovenian heavy rhymesand nuclei appear only before exceptional SC(C) sequences (Table 26).28 Such cases allowan analysis where the glide is an onset followed by a nucleus silenced by magic licensing(See Section 3.2.3, example (18)). Based on this we can conclude Slovenian does notallow superheavy rhymes.29

Slovenian〈ı〉 〈ı〉 〈u〉 〈u〉 〈a〉 〈a〉

I ["hui“skati] [e"nai

“st]

‘to instigate’ ‘eleven’U ["piu

“ski] [au

“f"biks] !? ["kau

“sniti]

‘drinking (adj.)’ ‘trouble’ ‘to peck’〈e〉 〈e〉 〈e〉 〈o〉 〈o〉 〈o〉

I ["dei“stvo] ["moi

“st@r] ["voi

“ska]

‘fact’ ‘skilled worker’ ‘army’U ["beu

“sk] ["reu

“sniti] ["pou

“ster] ["lou

“ski]

‘a yap’ ‘to yelp’ ‘pillow’ ‘hunting (adj.)’

Table 26: Availability of V + i/u + ‘CC’ sequences

In contrast, the data from the Vipava Valley dialects presented in Tables 27 and 28shows that at least some cases that do not follow the SC(C) generalization exist. Apartfrom one, all these cases match Harris’ generalizations for English. What is important inrelation to the analysis from Section 4.2.11, is that the exceptions appear with [au

“] and

[ai“], which are not traditionally not assumed to be diphthongs, and not with [ou

“] and [ei

“],

which are. The third exception (["kou“ck]) is interesting because of another unexplained

phenomenon found in Vipava Valley dialects — word final consonant clusters that seem-ingly violate the binarity theorem ([ce"t@rtk] ‘Thursday’, ["palck] ‘gnome’ ). But sincesuch clusters can also appear after other consonants, their appearances after sequenceslike [ou

“] cannot be used to motivate their diphthong status.

As researching potential VVS superheavy rhymes is not one of the major goals of thispaper, the data was not gathered with them in mind. This is why the present analysis isfar from conclusive and calls for a more detailed analysis based on additional data.

28The only counterexample being [au“f"biks], which is not problematic, if we analyze it as an instance

of dummy morphology (See Section 5.4). Additionally, the listed pronunciation is one closer to thewritten form listed in the Dictionary of Standard Slovenian (Bajec 2005), which is not necessarily itstrue phonological form. My own judgment is that the pronunciation [au

“"biks] is much more common.

29Some newer loanwords from English that are not listed in the Dictionary of Standard Slovenian(Bajec 2005), words like ["poi

“nt] in ["Ãoi

“nt], exhibit what look like English superheavy rhymes. However,

a more detailed analysis would be needed to know for sure.

36

5.1 Heavy and Superheavy Rhymes

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje

[ei“] ["cei

“sti] ["vei

“ste] ["kei

“sv@m]

‘road (dat.)’ ‘you know (pl.)’ ‘what do I know!’[ou

“] ["kou

“ck] !

‘suitcase’[@i“]

[@u“]

[ai“] ["blai

“nce] ! [s"ti

“@r"nai

“st]

‘part of a grape press’ ‘fourteen’[au

“] ["plau

“Nk] !

‘grape pressing barrel’[ui

“]

[iu“]

[oi“]

[eu“] ["Xleu

“ski]

‘for barns’[u“@i“] ["u

“@i“ske]

‘war (gen.)’[i“@u“]

[u“oi“]

[i“eu“]

Table 27: Availability of V + i/u + ‘CC’ sequences

Karst*Dornberk *Renče *Miren Miren Šempeter

[ei“] [o"bei

“sju] [d"vei

“zGa]

‘they hang’ ‘twentieth’[ou

“] [t@r"nou

“ska]

‘of Trnovo’[@i“]

[@u“]

[ai“]

[au“]

[ui“]

[iu“]

[oi“]

[eu“]

[u“@i“] ["u

“@i“sku]

‘army (acc.)’[i“@u“]

[u“oi“]

[i“eu“]

Table 28: Availability of V + i/u + ‘CC’ sequences

37

5 Further Research

5.1.2 Long Monophthongs

In Section 4.2.1 I argued that there is no vowel length contrast present in The VipavaValley dialects. However, the lack of contrast does not yet show whether the vowels arephonologically long or short. In the case of stressed vowels actually being phonologicallylong, we expect them not to appear in branching rhymes. (with potential exceptionsmentioned in the previous section) But as we can see in Table 29, this is not so.

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje["tiskrat] [st@"vilke] ! ["palk] ! ["tistiX]‘takrat’ ‘numbers’ ‘spider’ ‘those (gen.)’

Karst*Dornberk *Renče *Miren Miren Šempeter["tolko] ! ["saldu] ! ["briXtni] ! [di"rekt] ! [zu"zelke] !‘toliko’ ‘very’ ‘smart (m.pl.)’ ‘directly’ ‘insects’

Table 29: Stressed monophthongs in heavy rhymes

In comparison to sequences containing glides, the data from monophthongs shows adifferent picture, as they can appear before a much greater number of rhymal complement-onset combinations. Even taking Harris’ exceptions into account, this difference is signif-icant, as English does not differ with respect to long monophthongs or heavy diphthongswhen it comes to allowed environments for superheavy rhymes.30 If we take this to meanthat stressed monophthongs in this dialects are short, the restrictions presented in theprevious section become more interesting, as it would be peculiar for a language to allowheavy diphthongs but not long monophthongs.

5.2 The Structure of Light Diphthongs

The GP structure for light diphthongs is, as mentioned, somewhat problematic. Thisis partially due to sub-skeletal branching and partially to the mechanism of spreadingneeded to explain how I and U transfer their melodic material to the head in cases like/i“e/ and /u

“o/ (45).

(45) a. /i“e/ N

×

i a

b. /u“o/ N

×

u a

30The alternative analysis in Pöchtrager (2010) primarily deals with monophthongs, and the differencesit distinguishes between the two classes are not of the same nature as in Slovenian.

38

5.2 The Structure of Light Diphthongs

Normally elements spread from positions dominated by one skeletal point to otherskeletal points. This happens with assimilation (46a), long monophthongs (46b) andheavy diphthongs (46c).

(46) a. ×

Ex

×

Ey

b. ×

Ex

× c. ×

Ex

Ey

×

The GP structure for light diphthongs comes from the autosegmental analysis ofrising (47a) and falling tone (47b), where this difference is expressed by associating twoelements to the same skeletal point. But for it to work, it is crucial that no “sharing” ofmelodic properties takes place between elements, as this would probably lead to risingand lowering towards a mid tone (47c&d).

(47) a. ×

l h

b. ×

h l

c. ×

l h

* d. ×

h l

*

This makes the type of spreading that has to be allowed in the case of light diphthongsvery unwanted. But if we assume, as in GP 2.0, that A is actually additional structureand not an element (48), I and U can spread to the head of this additional structure(49), as described for Putonghua in Živanović and Pöchtrager (2010).

(48) a. A = ×N

×N ×

(49) a. /i“e/ NP

×I

N

×N

×N ×

×

b. /u“o/ NP

×U

N

×N

×N ×

×

If we additionally assume that /i“@/ and /u

“@/ also contain A, as in the analysis pre-

sented in (39) in Section 4.2.7, the difference between the two types of diphthong can beexpressed with the possibility or inability of I/U to spread. The GP 2.0 structure for/i“@/ and /u

“@/ would then be something like (50).

39

5 Further Research

(50) a. /i“@/ NP

×I

N

×N

×N ×

×

b. /u“@/ NP

×U

N

×N

×N ×

×

Without further modifications, the analysis from Živanović and Pöchtrager (2010),is not suitable to apply directly to the Vipava Valled vowel system. In fact, the vowelsystem of Putonghua does not allow certain vowels found in the Vipava Valley dialects— /i

“@/ and /u

“@/ being among them. The parameters in Živanović and Pöchtrager (2010)

are set to exclude these sounds, which is why a direct application of the same analysis isimpossible. Stegovec (2011) presents an attempt to develop parameters suited to describethe vowel system of Vipava Valley dialects within GP 2.0, however, like the frameworkitself, it is still a work in progress.

5.3 The Beginning of the Word and Its Influence

In Section 4.2.5 a potential additional light diphthong variant was uncovered, namely/u“a/ (["u

“asm@] ‘eight (gen.)’ (Logar 1993: 63)). This sequence, interestingly, appears

only word initially and nowhere else, where light diphthongs are expected. And thisseems to be exactly the environment in which /u

“o/ is not found (Table 30). This can be

taken as a sign that [u“a] is only this diphthong’s word initial realization.

Inner Carniolan*Predmeja *Skrilje Podnanos Dobravlje

[#u“@] ["u

“@n] ["u

“@nde] ["u

“@n] ["u

“@truk]

‘out’ ‘there’ ‘out’ ‘child’[#u

“o] ["u

“ou“ce] !?

‘sheep (pl.)’[#u

“a] ["u

“aknu] ["u

“ast] ["u

“ani] ["u

“asu]

‘window’ ‘to drive’ ‘they’ ‘donkey’Karst

*Dornberk *Renče *Miren Miren Šempeter[#u

“@] ["u

“@p] ["u

“@s] ["u

“@i“sku] ["u

“@n] ["u

“@de]

‘beside’ ‘wagon’ ‘war (acc.)’ ‘out’ ‘water (gen.)’[#u

“o] ["u

“on] !?‘he’

[#u“a] ["u

“asm@] ["u

“an] ["u

“ani]

‘eight (gen.)’ ‘he’ ‘they’

Table 30: [u“@], [u

“o] and [u

“a] sequences in word initial positions

40

5.4 The Link Between Stress and Light Diphthongs

In existing transcriptions we find ["u“ou“ce] (Logar 1993: 61) and ["u

“on] (Logar 1993: 65),

which seem exceptions to this pattern. With the first example [u“o] is also attested in the

accompanying recording. The retention of this realization might be due to the presenceof U in the immediately adjacent [u

“]. But this being the only case, it is hard to make

generalizations. With the second problematic example, it seems to be a case of a tran-scription mistake, as in the recording all such cases are pronounced [u

“an], which was also

the pronunciation given by my informants.The glide [u

“] could in principle be analyzed as a word initial onset, however, it was

demonstrated already in Section 4.2.5 that [w] cannot appear in front of realized nuclei.In GP, the beginning of the word is an environment that is necessarily structurally

distinct in vowel initial words. Due to universal restrictions on constituent ordering, aword always begins with an onset, which is empty in vowel initial words.31 Lowenstamm(1999) even suggests, based on independent evidence that every word (or its extendedprojection) begins with an empty CV (OR) pair. There is, however, no set of GP prin-ciples that would tie this extra structure to the phenomenon described above. It is alsounclear, why this happens only with one type of diphthong. And since it is impossibleto differentiate between an initial /ja/ and /i

“a/ based on surface properties, it is hard to

figure out if this happens only to /u“o/. Further research is definitely needed.

5.4 The Link Between Stress and Light Diphthongs

It is noted already in traditional literature that diphthongs appear only in stressed po-sitions in Vipava Valley dialects, which is especially interesting from a GP standpoint.We can also see pairs of words derived from the same stem having different realizationsif the position of the stress is different (51).

(51) a. ["u“@trok] ‘child’ ; [u"tru

“oci] ‘children’

b. ["u“@zit] ‘to drive’ ; [voz"nik] ‘driver’

c. ["u“@i“ska]; ‘army’/‘war’ ; [vu"jak] ‘soldier’

There are seeming exceptions to this, however, they are mostly morphologically com-plex words, where unstressed prefix versions of prepositions maintain their diphthong re-alizations: {u

“@b-} ‘beside-’, {u

“@d-} ‘from/of-’ ali {u

“@-} ‘about-’ (52). Slavic prepositions

and prefixes are known for being resistant to phonological processes, which is evidenced incases of consonant clusters that are impossible anywhere else ([od"dremati] ‘to nap (ter-

31The opposite is true in the approach argued for by Szigetvári (1999), who suggests a VC unit forthe basic constituent pair.

41

5 Further Research

minative)’ ). This behavior is in line with analytic morphology (Kaye 1995: 302), wherephonological processes applied to morphemes individually before the word is formed.

(52) a. Dornberk (Logar 1993: 63): [u“@d"li

“@Glo] ‘it was relieved’ ; [u

“@"pazu] ‘he noticed’

b. Skrilje (Rustja 2001: 18–28): [u“@bo"lava] ‘is getting sick’ ; [u

“@"d@rsow] ‘scratched’ ;

[u“@"cistu] ‘he cleaned’ ; [u

“@tku"pav@li] ‘they were digging out’ ; [u

“@tko"palu] ‘it

dug out’ ; [u“@sl@"parli] ‘they cheated’ ; [u

“@st@r"Gav@t] ‘to scrape’ [u

“@d"rei

“z@li] ‘to

cut out/off’ ; [u“@"tisk@lu] ‘it pressed out’

Some exceptions do not fit in this group, as they are not created by a generativemorphological process (53). They can be equated to examples like ‘cranberry’, where‘cran-’ is not an individual lexical item, but the word is still formed like ‘straw’ + ‘berry’= ‘strawberry’. In GP this is referred to as dummy morphology (Pöchtrager 2006: 58–59),where the word behaves as morphologically complex only to phonological processes.

(53) a. Dornberk (Logar 1993: 63): [u“@"ku

“@li] ‘around’

b. Skrilje (Rustja 2001: 18–28): [u“@"kuli]; ‘around’ ; [u

“@b"ru

“@ci] ‘rings’ ; [u

“@"ru

“@dje]

‘tools’ ; [u“@s"talu] ‘the rest’ ; [u

“@"p@rtnik] ‘knapsack’ ; [u

“@d"zat] ‘from/at the

back’ ; [u“@k"ru

“@Gla] ‘round (f.)’

If we can rely on dummy morphology with the examples in (53) this is impossiblewith the examples in (54). The number of exceptions is too great to ignore, despite thefact that none of them were observed during my own fieldwork.32

(54) a. Predmeja (Logar 1993: 61): [u“@"zovam] ‘wagons’ ; [u

“@"zili] ‘they drove’

b. Dornberk (Logar 1993: 63): [u“@"Gu

“@sta] ‘August (acc.)’ ; [u

“@"cmi] ‘eyes (instr.)’

c. Skrilje (Rustja 2001: 18–28): [u“@"zili]; ‘they drove’ ; [u

“@"tru

“oci] ‘children’ ;

[u“@"ci

“@smi] ‘eyes (instr.)’

Independent evidence shows that initial [u“] cannot be an onsets. Even with a hypoth-

esis where /w/ is possible only in front of semivowels, we immediately find counterex-amples: [v@"si

“@u“] ‘happy (m.)’, [v@"si] ‘villages’ in [v@"like] ‘big (pl.f.)’. Unstressed /u

“@/

could in principle be a consequence of secondary stress, however, a much simpler solutionis available if we take into account that all these examples appear word initially. In theprevious section we saw the effect this environment has on the realization of diphthongs,which makes the phenomenon discussed here less of an anomaly.

Most GP analysis do not deal with the effect of stress on phonological structure, butsome authors like Scheer and Szigetvári (2005); Pöchtrager and Kaye (2010) link stress to

32In the existing literature they avoid addressing this problem by writing these cases as sequences of/w/ and a vowel.

42

Conclusions

additional structural relationships between constituents or additional structure in itself.In this case, the suggestion of Pöchtrager and Kaye (2010) that stressed nuclei are morestructurally complex seems the solution to pursue.

This approach explains, among other things, why languages with open and closedvowels allow them only in stressed positions. Their proposal uses GP 2.0 structures,which is why the extra structure can be expressed hierarchically, as a specifier (55).

(55) NP

× N

×N ×

In the hypothetical GP 2.0 structures for light diphthongs (49) in Section 5.2, thespecifier position is the one, where the glide part I or U is located. If this is the positionassociated with stress, it becomes clear why light diphthongs are impossible withoutit. And since the beginning of the word seems to also be connected to some additionalstructure, the seeming exceptions become less problematic. The structure proposed forstressed nuclei in Pöchtrager and Kaye (2010) is motivated by different phenomena, whichmakes this analysis of diphthongs a novel piece of evidence in its support, opening newareas of research.33

6 Conclusions

The two dialect groups of VVS, are traditionally described as having two different vowelsystems. But a careful analysis following the principles of GP showed that there is noreason to believe this.

Using the standard GP definition of diphthongs and checking the predictions made bytheir structure and general theoretical principles, I was able to conclude that all dialectsof VVS have the same vowel system, with only the following (light) diphthongs: /i

“@/,

/u“@/, /i

“e/ in /u

“o/, and that the two sequences [ei

“] and [ou

“], which are also traditionally

described as diphthongs, must be sequences of monophthongs and consonantal glides.The collected data also helped reveal some previously unknown phenomena exhibited

by VVS, as well as point out some problems standard GP has with the description oflight diphthongs. However, there are also enough pointers in the data to how these

33An attempt of finding the relation between stress and light diphthongs was also briefly discussed inStegovec (2011).

43

REFERENCES

problems could be solved by further developing some existing frameworks, opening upnew possibilities of research.

References

Bajec, A. (Ed.) (2005). Slovar slovenskega knjižnega jezika. Ljubljana: Inštitut za sloven-ski jezik Frana Ramovša ZRC SAZU.

Charette, M. (1988). Some Constrainst on Governing Relations in Phonology. Ph. D.thesis, McGill University.

Chomsky, N. (1986). Barriers. Cambridge, MA: MIT Press.Chomsky, N. and M. Halle (1968). The Sound Pattern of English. New York: Harper &Row.

Cyran, E. (2010). Complexity Scales and Licensing in Phonology. Studies in GenerativeGrammar. De Gruyter Mouton.

Goldsmith, J. A. (1976). Autosegmental Phonology. Ph. D. thesis, Massachusetts Instituteof Technology.

Gussmann, E. and J. Harris (1998). Final Codas: why the west was wrong. In E. Cyran(Ed.), Structure and Interpretation, Studies in Phonology, pp. 139–162. Lublin: Folium.

Harris, J. (1994). English Sound Structure. Oxford UK, Cambridge, MA, USA: Blackwell.Harris, J. (1997). Licensing Inheritance: an integrated theory of neutralisation. Phonol-ogy 14, 315–370.

Harris, J. and G. Lindsey (1995). The elements of phonological representation. In J. Du-rand and F. Katamba (Eds.), Frontiers of Phonology, pp. 34–79. Harlow, Essex: Long-man.

Jensen, S. (1994). Is P an element? Towards a non-segmental phonology. In M. Cobband S. Jensen (Eds.), SOAS WP-LPhet 4, pp. 71–78. London: SOAS, University ofLondon.

Kaye, J. (1990). ‘Coda’ licensing. Phonology Yearbook 7, 301–330.Kaye, J. (1992). Do you believe in magic? The story of s+C sequencences. SOAS WorkingPapers in Linguistics & Phonetics 2, 293–313.

Kaye, J. (1995). Derivations and interfaces. In J. Durand and F. Katamba (Eds.),Frontiers of Phonology, pp. 289–332. London & New York: Longman.

Kaye, J. (1997). Why this Article is not about the Acquisition of Phonology. SOASWorking Papers in Linguistics & Phonetics 7, 209–220.

Kaye, J. (2000). A User’s Guide to Government Phonology (GP). [manuscript] http://lolita.unice.fr/~scheer/scan/Kaye00guideGP.pdf.

Kaye, J. (2001a). A Fresh Look at Putonghua Onset-Rime Pairs. GuangdongUniversity of Foreign Studies [neizdan rokopis] lolita.unice.fr/~scheer/scan/Kaye01ORputonghua.pdf.

Kaye, J. (2001b). Working with licensing constraints. In K. Dziubalska-Kołaczyk (Ed.),Constraints and Preferences, Number 134 in Trends in Linguistics: Studies and Mono-graphs, pp. 251–268. Berlin; New York: Mouton de Gruyter.

44

REFERENCES

Kaye, J., J. Lowenstamm, and J.-R. Vergnaud (1985). The internal structure of phono-logical elements: a theory of charm and government. Phonology Yearbook 2, 305–328.

Kaye, J., J. Lowenstamm, and J.-R. Vergnaud (1990). Constituent structure and govern-ment in phonology. Phonology Yearbook 7, 193–231.

Lakatos, I. (1970). Falsification and the Methodology of Scientific Research Programs.In I. Lakatos and A. Musgrave (Eds.), Criticism and the Growth of Knowledge, pp.91–195. Cambridge University Press.

Lakatos, I. (1978). The Methodology of Scientific Research Programmes, Volume 1 ofPhilosophical Papers. Cambridge University Press.

Logar, T. (1993). Slovenska narečja. Zbirka Cicero. Ljubljana: Mladinska knjiga.Logar, T. (1996a). Dialektološke in jezikovnozgodovinske razprave. Ljubljana: Inštitut zaslovenski jezik Frana Ramovša ZRC SAZU.

Logar, T. (1996b). Pripombe k obsegu in karakteristiki kraščine in notranjščine vRamovševih dialektih. In K. Kenda-Jež (Ed.), Dialektološke in jezikovnozgodovinskerazprave, Chapter 1, pp. 65–67. Ljubljana: Inštitut za slovenski jezik Frana RamovšaZRC SAZU.

Logar, T. and J. Rigler (1983). Karta slovenskih narečij. Ljubljana: DDU Univerzum.Lowenstamm, J. (1996). CV as the only syllable type. In J. Durand and B. Laks (Eds.),Current trends in Phonology: Models and methods, pp. 419–441. Salford, Manchester:ESRI.

Lowenstamm, J. (1999). The beginning of the word. In J. R. Renison and K. Kühnammer(Eds.), Phonologika 1996: Syllables !?, pp. 153–166. The Hague: Thesus.

Pearce, M. (2011). Methodology in Field Phonology. In N. C. Kula, B. Botma, andK. Nasukawa (Eds.), Continuum Companion to Phonology, Continuum companions,pp. 3–15. London, New York: Continuum International Publishing Group.

Ploch, S. (1998). Non-switch harmony in Yawelmani (and Turkish and Sakha). SOASWorking Papers in Linguistics and Phonetics 8, 209–238.

Pöchtrager, M. A. (2006). The Structure of Lenght. Ph. D. thesis, Universität Wien.Pöchtrager, M. A. (2009). Diphthongi, ei know thyselfi. Binding in Phonology. In Preda-vanje na konferenci 17th Manchester Phonology Meeting, 28.—30. maj, 2009, Manch-ester.

Pöchtrager, M. A. (2010). A Corset for Superheavy Rhymes. In Predavanje na konferenciGPRT7, 8. in 9. maj 2010, Ljubljana.

Pöchtrager, M. A. and J. Kaye (2010). In Quest of Pöchtrager’s Puzzle. In Predavanjena konferenci GPRT7, 8. in 9. maj 2010, Ljubljana.

Rennison, J. (1998). Contour segments without subsegmental structures. In E. Cyran(Ed.), Structure and Interpretation, Studies in Phonology, pp. 227–245. Lublin: Folium.

Rennison, J. R. and F. Neubarth (2003). An x-bar theory of Government Phonology. InS. Ploch (Ed.), Living on the Edge: 28 Papers in Honour of Jonathan Kaye, Studiesin Generative Linguistics, pp. 95–130. Berlin & New York: Mouton de Gruyter.

Rustja, I. (2001). Vinogradniška terminologija v Skriljah v Vipavski dolini. Diplomskodelo, Filozofska fakulteta Univerze v Ljubljani.

Scheer, T. (2004). A lateral theory of phonology. Vol 1: What is CVCV and why shouldit be? Berlin: Mouton de Gruyter.

Scheer, T. (to appear). Why Russian vowel-zero alternations are not different, and why

45

REFERENCES

Lower is correct. Journal of Language and Verbal Behaviour (St Petersburg).Scheer, T. and P. Szigetvári (2005). Unified representations for stress and the syllable.Phonology 22, 37–75.

Smole, V. (2000). Uvod v slovensko dialektologijo: del vsebine za predmet ZSD II.Ljubljana: Filozofska fakulteta, Oddelek za slovanske jezike in književnosti, Katedraza zgodovino slovenskega jezika in dialektologijo. [study material].

Smole, V. (2002). Kolikostna nasprotja v slovenskih narečjih. In Med dialektologijo inzgodovino slovenskega jezika: ob življenjskem in strokovnem jubileju prof. dr. MartineOrožen, pp. 52–61. Maribor: Slavistično društvo.

Smole, V. (2004). Narečne skupine — Pregled: del vsebine za predmet ZSD II. Ljubljana:Filozofska fakulteta, Oddelek za slovanske jezike in književnosti, Katedra za zgodovinoslovenskega jezika in dialektologijo. [study material].

Stegovec, A. (2011). The Vipava Valley Slovenian Vowel System in GP 2.x. Talk givenat GPRT8, Vienna on 7th May 2011.

Stegovec, A. (2012). Samoglasniški sistem narečij Vipavske doline v vezalni fonologiji.BA Thesis, University of Ljubljana.

Szigetvári, P. (1999). VC Phonology: a theory of consonant lenition and phonotactics.Ph. D. thesis, ELTE/MTA.

Šekli, M. (2009). Merila določanja mej med slovenskimi narečji in podnarečji. In V. Smole(Ed.), Slovenska narečja med sistemom in rabo: Obdobja 26 — Metode in zvrsti, Ljubl-jana, pp. 291–318. Znanstvena založba Filozofske fakultete Univerze v Ljubljani.

Toporišič, J. et al. (2001). Slovenski pravopis: I Pravila (Revised 6th ed.). Ljubljana:Inštitut za slovenski jezik Frana Ramovša ZRC SAZU.

Živanović, S. and M. A. Pöchtrager (2010). GP 2, and Putonghua too. Acta LinguisticaHungarica 57 (4), 357–380.

46