You are on page 1of 18

Vaux, Bert. 1992. Adjarians Law and Consonantal ATR in Armenian.

In Proceedings of the Fourth International Conference of Armenian Linguistics, ed. John Greppin, 271-293. Delmar, NY: Caravan. Updated February 2010. Adjarians Law and Consonantal ATR in Armenian* Bert Vaux, Harvard University 1. Introduction In 1901 the Armenian linguist Hrachea Adjarian first observed a systematic development of from classical Armenian a in the modern Armenian dialect of Van, e.g. (1) classical bah spade > p classical danak knife > tnk classical gan sheep > kyr

Though at the time he was not able to explain this development, he continued to expand and refine his analysis in a series of articles (Adjarian 1901, 1909, 1911, 1953); by 1911 he was able to say that the change occurred almost exclusively in initial syllables, and was originally conditioned by the presence of a preceding voiced stop or affricate, as stated informally in (2): (2) a / #[+voice, -son, -cont] __ He later observed that several other dialects (Shamaxi, Salmast, Maragha, and Xoy) showed the same phenomenon, and some of these extended it to the back vowels o and u as well: (3) classical bak porch > Shamaxi pk classical bok radish > Shatax pk classical bukh chisel > Xoy pky M. Muradyan (1962) added Shatax and H. Muradyan (1973) added Kirzan, Meghri, and Karchevan to the list of dialects showing what had become known as Adjarians Law, which by that time could be expressed as in (4): (4) [+syllabic] > [-back] / #[+voice, -sonorant, -continuant] __ This is the classic form of Adjarians Law. We must make several important emendations to this formulation in order to capture all of the relevant data, however. First of all, Weitenberg (1986) observed that what had been initial ya- in classical Armenian seems to have triggered the rule as well. After examining the data in Adjarian (1927) and the various dialect grammars, I believe we must also add initial {z l v} to the list of triggering environments1, and Malatya, Musaler, I am indebted to Andrea Calabrese for his invaluable suggestions concerning the theoretical aspects of this paper. Research documented herein was supported by a grant from the National Science Foundation. 1 Khachaturyan (1982) includes r as a trigger as well, citing forms such as class. arbenal become drunk > Karabagh hrbil, but in these cases the r is always after the affected vowel,
*

and Goris to the list of dialects showing the rule. The basic distribution of the rule can then be represented as in (5)2: (5)

[There was a word list here in the original paper, but it has not survived on my computer. Representative words with initial voiced and voiceless stops can be found in ch. 6 of my 1998 OUP bookBV] That is to say, Adjarians Law fronts back vowels in initial syllables when preceded by and an initial h always appears in the dialect form, so we cannot be sure that the r is actually behaving like the other triggering consonants, or is even a trigger at all. Consequently, I have decided not to include it in our set of environments. 2 The base map comes from Hewsen 2001; the set of dialects showing Adjarians Law is taken from ahukyan 1972 and therefore does not include the dialects that I have added to the inventory (Malatia and Musaler). I have chosen a representative word for each triggering environment, and shown its outcome in the dialects which I believe show Adjarians Law; due to learned pronunciation and other interfering factors, the forms are not always consistent. For that reason I have also provided overall percentages (where available) of words undergoing the rule in the various vocalic environments: a 77 under Goris/a, for example, indicates that 77% of words beginning with {b g d z l j y v} changed following /a/ to []. I could not provide percentages for Salmast, Meghri, and Kirzan, as grammars of these dialects were not available to me; I have taken the forms for these dialects from Adjarian (1927) and Muradyan (1973). All other forms are taken from the respective dialect grammars, listed in the bibliography. I have not included {lo, zo, zu, o, u, yo, yu}, because there are too few relevant lexical items in the grammars to determine whether or not these environments trigger the rule.
2

voiced obstruents, l, and y. This distribution is difficult to formulate within traditional phonological frameworks, such as that employed in (4); y and l are [+son, +cont], and {z v} are [+cont], so we cannot maintain the environment described in (4), and no other conglomeration of traditional features can capture our rule. Adjarian and Muradyan describe it as a type of palatalization, but this is only a description of what happens to the vowels; it does not help us understand why such a development would occur. They are justified in calling it palatalization insofar as the vowels (apparently) become [-back] (except in the Malatya dialect), but we obviously cannot say that the voiced obstruents and {z y l v} are [-back] and all other consonants are [+back]. Analyses so far have also not been able to explain why the rule appears to affect only back vowels (actually, there is good reason to believe that front vowels are involved as well, as we shall see in section 2.6). It seems clear that there is some kind of interaction between voiced consonants and following vowels in Adjarians Law; we can interpret this as a case of the consonants spreading some feature to the vowels, the vowels spreading some feature to the consonants, or as two independent developments. The third possibility, namely that the coocurrence of vowel fronting with voiced consonants is merely a coincidence, seems highly unlikely, though this has been proposed by most scholars studying this type of phenomenon in other languages (we will consider these proposals in section 2.4). The second possibility, namely that the vowels are affecting the consonants, is equally unlikely, for in that case we would expect all consonants to be affected, not just voiced consonants, and in fact we would not be able to account for the vowel fronting with this hypothesis. We are forced to conclude that Adjarians Law is actually a case of voiced consonants spreading some feature to following vowels; I propose that this feature is [ATR]. The feature [ATR], or advanced tongue root, refers to the general expansion of the pharyngeal cavity which often plays a phonological role in distinguishing the vowels traditionally labelled tense and lax; it also is often implicated in the production of voicing in obstruents. It is the interplay of these vocalic and consonantal features of [ATR] which seems to be at work in Adjarians Law. [ATR] is generally accepted for vowels (see section 2.4), and Westbury (1979, 1983) and Kohler (1984) have shown that it may also be present in consonants. Trigo (1987), based on evidence from Akan, Buchan Scots English, and Madurese, has proposed that [ATR] is active in consonants, so that for example [ATR] values of consonants can affect the [ATR] values of neighboring vowels. I argue in this paper that the data from Adjarians Law in the Armenian dialects provide compelling support for this hypothesis. I propose that we view Adjarians Law as a case of the feature [ATR] spreading from consonants to following vowels, as follows: (6) ## C V | [ATR]

I believe we can also use evidence from the behavior of consonantal [ATR] in Akan, Buchan Scots English, Madurese, Coeur dAlene, Sahaptian, Maasai, Jingpho, and Mon Khmer to explain the apparent anomalies in the behavior of Adjarians Law mentioned above. I examine the evidence for my proposal in section 2, and then consider some problems that Adjarians Law entails for the history and classification of the Armenian dialects in section 3.

2. Theoretical aspects of [ATR] and Adjarians Law 2.1. Before entering into the theoretical aspects of this paper, I will provide a brief introduction to the general theoretical framework in which I carry out my work, for the benefit of those who are not familiar with the types of concepts and terminology employed in contemporary phonology. The system which most American and European phonologists now employ is commonly referred to as non-linear phonology, or alternatively autosegmental or metrical phonology. It was initially developed in the early 1970s as a way of dealing with tonal phenomena in various African languages, but the explanatory potential of its three-dimensional structure quickly led to applications in other areas such as vowel harmony systems, syllable structure, and consonant mutations. Some simple examples of non-linear accounts of phonological phenomena are given below: (7) compensatory lengthening (Sezer (1985)) [O = onset, R = rhyme, N = nucleus, C = coda] i. Turkish kahya steward
O R O R N C N

k a h y a ii. /h/ deleted in syllable codas before continuants and nasal stops s s
O R O R N C N

h y

iii. vowel nuclei spread to empty rhyme slots s s


O R O R N C N

h y

iv. vowels occupying two moras are phonetically realized as long; segments not associated to segmental slots are deleted
O R O R N C N

a [kya]

(8) vowel harmony (Clements (1976)) i. Akan roots specified for melodic features, including [ATR] E-bu-O nest E-b-O stone | | [+ATR] [-ATR] ii. Melodic features of root spread to affixes E-bu-O E-b-O [+ATR] [-ATR]

iii. [-ATR] E, O [], []; [+ATR] E, O [e], [o] surface forms: ebuo nest, b stone For a basic introduction to non-linear phonology, one may consult Hogg and McCully (1987) and Goldsmith (1990). 2.2 Let us consider a case where non-linear notation allows us to account for a common phonological process which is difficult to represent within the framework of linear phonology. Voiced consonants and tense vowels clearly pattern together in many languages, for example, as in Buchan Scots English, which only allows tense vowels to follow voiced consonants. It is not clear how to explain this relationship in the framework of linear phonology; since voiced consonants are lax, we would have to postulate a rule something like (9): (9) V [+tense] / C[+lax]_

Both voiced consonants and tense vowels are [+ATR], on the other hand (that is to say, both sets are produced with the tongue root advanced), so we can easily account for these situations by using the feature [ATR] instead of [tense/lax] (see Westbury (1979, 1983)). It is also somewhat easier to see this interaction between consonantal and vocalic [ATR] if we use the formulation in (6) above, where the idea that [ATR] is actually spreading from the consonant to the vowel is made explicit. Adjarians Law in fact seems to be precisely such a case of [ATR] interaction between
5

consonants and vowels. As I mentioned above, such interactions are not unusual; we find similar phenomena in some dialects of English, Akan, Madurese, Coeur dAlene, Sahaptian, Maasai, Jingpho, and various Mon Khmer languages. The mechanics of these interactions may not be immediately obvious to the reader, however, so we will briefly consider the reasons for positing a feature [ATR] in consonants and vowels now, and then see how such a system can account for the Armenian data. 2.3. In order to produce voicing, there must be a pressure differential of at least 2 cm H2O across the glottis, with the sub-glottal pressure being higher, so that air may flow up freely through the vibrating vocal folds (Catford (1977:74)). This is simple enough in sonorants, where the vocal tract is relatively unconstricted. Obstruents require a complete closure at some point in the vocal tract, however, which quickly neutralizes any pressure differential across the glottis. Consequently, other articulatory mechanisms must be employed in order to produce voiced obstruents. What tends to occur is an expansion of the pharyngeal cavity, implemented primarily by lowering of the larynx and advancing of the tongue root, which lowers the supra-glottal pressure sufficiently to allow voicing. The advancement of the tongue root often entails a fronting and raising of the tongue body as well, as we will consider in section 2.6. These maneuvers were detected as early as the nineteenth century, when Ellis (1877) and Jesperson (1889) observed that there was noticeable lowering of the larynx during the production of voiced stops. Then in our century, House and Fairbanks (1953) observed that vowels have lower fundamental frequencies and greater acoustic energy at high frequencies after voiced stopsthe same features which were later discovered to be the primary acoustic correlates of [+ATR] vowels (Lindblom and Sundberg (1971)). 2.4. These discoveries laid the groundwork for our theory of consonantal [ATR], but the importance of the tongue root itself was not actually noticed until 1967, when Stewart observed that the differences between the harmonic sets {} and {i e o u} in Akan could only be satisfactorily accounted for in terms of the relative advancement of the tongue root involved in producing the first set of vowels3. Many languages were shown soon thereafter to have vowel harmony systems based on tongue root oppositions, such as Lhasa Tibetan and the PaleoSiberian languages Koryak and Gilyak (Aoki (1968)), numerous Niger-Kordofanian and Nilotic languages (Hall et al. (1974)), Mon Khmer (Gregerson (1976)), Nez Perc (Rigsby and Silverstein (1969), Hall and Hall (1980)), DhoLuo (Jacobson (1977, 1978)), and Mongolian (Rialland and Djamouri (1983)). Halle and Stevens (1969) first proposed using [ATR] as a phonological feature for vowels, replacing the old tense/lax opposition, based on experimental observations and data from vowel harmony systems of the type just mentioned. Then, in 1976, Gregerson (following earlier observations by Haudricourt and MArtinet (1947)) noticed that the [+ATR] vowels in Mon Khmer always occurred after what had originally been voiced consonants (in the modern language these consonants became voiceless), but concluded that there was no causal relationship between the two. Similar observations were made by Diffloth (1980) for Wa (a Mon Khmer language) and by Maddieson and Ladefoged (1985) for Jingpho (a Tibeto-Burman language), but these analyses also stopped at pointing out the correlation between voiced consonants and [+ATR] vowels, and did not suggest that one might have led to the other.
3

Ladefoged (1964) made similar observations, but we need not consider these here.
6

Westbury (1979, 1983) and Kohler (1984) then provided experimental evidence that voiced consonants were phonetically [+ATR], but did not consider the phonological implications of that fact. 2.5. It was not until 1987 that the causal relationship between [ATR] and consonant voicing was made explicit, when Loren Trigo proposed that such cooccurrences in Mon Khmer, Buchan Scots English, and Madurese were actually cases of the feature [+ATR] spreading from consonants to vowels. In Buchan we find cases such as: (10) bed-I small bed wiskt-I little waistcoat bedI [+ATR] w i s k t - I [-ATR] bed wiskti

Similarly in Madurese we see: (11) budE back khusE gums bud E budi khus kh u s E

[+ATR]

[-ATR]

Madurese also has ATR harmony, so that [+ATR] values derived from consonants can spread beyond immediately adjacent vowels, unless blocked by a [-ATR] consonant. We should note that intervening {l r h y w} do not block [+ATR] spreading, which indicates that they are either [+ATR] themselves or unspecified for [ATR]: (12) bArAs health bAtO stone b A r A s b [+ATR] A t baras bat O

[+ATR] [-ATR]

Similarly, in Maasai (Tucker (1955), Cole (1987:41)) the glides {y w} spread [+ATR] onto immediately preceding and following vowels:

(13)

a-I-rOwA hot a--row ( = [+ATR] equivalent of a) a-I-wAn evade a-i-wn aI r OwA [-ATR] [+ATR] a I wAn [-ATR][+ATR]

These data may help us explain the behavior of {y l v} with respect to Adjarians Law, as we will see in section 2.6. In the American Indian language Coeur dAlene, all non-lax vowels become tense when followed by uvular and pharyngeal consonants (Cole (1987:90)). Consider the forms in (14): (14) kwn answer tsic tall : : ako-stq he answered back tsc-alqw he is tall

These data are easily accounted for if we assume that uvulars and pharyngeals are specified [-ATR]; this feature then spreads to preceding vowels, in a fashion directly parallel to Madurese and Buchan. We can represent this process as follows: (15) k u s t q

[+ATR]

[+ATR]

[-ATR]

There is evidence that pharyngeals are specified [-ATR] in Arabic and Kurdish as well (see Djamouri (1984:342)); we shall consider the Kurdish case further in section 2.7. We should also note the parallel between tonogenesis and the development of distinctive [+ATR] vowel series in such languages. Many tonal languages, such as Mon Khmer and Jingpho, are believed to have developed their systems of distinctive tone through the loss of their voiced consonant series, which often produce lower pitch in following vowels than voiceless consonants do. When the conditioning environment for phonetic tone distinction disappeared, tone became phonemic in these languages. We can easily draw a parallel to western Armenian dialects which devoice the classical plain voiced series, and say that the advancement of the tongue root inherent in voiced consonants became a phonemic feature of the following vowels when this voicing distinction was lost. 2.6. The cases described in 2.5 are strikingly similar to the workings of Adjarians Law in the Armenian dialects. As we just mentioned, the languages with [+ATR] and [-ATR] harmonic series often develop this configuration through the loss of an old series of voiced consonants, just as all the Armenian dialects showing Adjarians Law (except Kirzan, Meghri, Karchevan, Goris, Musaler, and Varhavar) have devoiced the classical plain voiced series. Similarly, we can observe that in many languages showing [ATR] spreading, only the back vowels are affected (cf. Mongolian, Rialland and Djamouri (1984), Steriade (1987)), as with Adjarians Law.
8

We should expect this to be the case with Armenian, because i and e are already [+ATR], and should therefore not be affected by a rule spreading [+ATR]. For those to whom the feature specifications of the proto-Armenian vowel system are not clear, I will present my interpretation here: (16) feature specifications of the proto-Armenian vowel system4 i e a high + + low + back + + + ATR + + round + + We should note that (according to Khachaturyan (1982:88)) classical i and e show different reflexes after voiced and voiceless consonants in the dialects of Karabagh. Consider: (17) classical kir container > kr cit sparrow > ct ket insect > kt : : : gir letter > kyir jig taut > cikhy get river > kyet

We see a similar type of distribution in Meghri (Aghayan (1954:44,265)): (18) classical kem grass rope > km kir container > kr : : get river > get gin price > gin

It appears that the [-ATR] specification of voiceless consonants spreads to i and e in these dialects, just as the [+ATR] specification of voiced consonants spreads to following vowels in Adjarians Law ( (hard (Gharibyan (1953), p. 45) must be [-ATR]cf. Buchan i, u, Trigo (1987:8); I will henceforth refer to and i with what I consider to be their IPA values, and respectively). It should not surprise us then that in some dialects of Armenian we also get cooccurrence restrictions of the type in (19): (19) (a) (b) voiced consonants and resonants occur with { i e} voiceless consonants occur with {a u o } (data from Khachaturyan (1982:88))

We might say simply that series (a) is [-back] and series (b) is [+back], which seems to be true, but this would not explain why (a) occurs with voiced consonants and resonants, and (b) occurs with voiceless consonants. It makes more sense to interpret (a) and (b) as [+ATR] and [-ATR] respectively, with a rule which makes all [+ATR] vowels [-back], and all [-ATR] vowels [+back], as in (20): There is good reason to believe that schwa was not phonemic in early Armenian, but since this point is not relevant to our discussion I will not consider it further here. I have also left out , which I believe was not a simple short vowel, but rather a long vowel or diphthong (note that it alternates with i in unaccented syllables).
9
4

(20)

[ATR] [-back]

Such a rule is certainly plausible, for the tongue is an elastic mass, and consequently tongue root advancement often entails fronting and raising of the tongue body, as schematized in (21) [from an old handout by Doug Pulleyblank which I have now lost the reference to]: (21)

tongue root advanced

tongue body pushed up and forward

We find similar rules linking [+ATR] with [-back] in Nez Perc, Mongolian, and Somali (see Hall and Hall (1980), Rialland and Djamouri (1984), Hall et al. (1972)); I have represented this type of configuration in (24(ii)). Similarly, retraction of the tongue root often entails backing and lowering of the tongue body. In some dialects of Kurdish, for example, pharyngeal consonants can only occur with [+back] vowels (Kahn (1976:48)). Consider the forms in (22): (22) dr he built : DaR house (capital letters are pharyngeals) phn wide : Pan (pronunciation by Kurdish speakers in areas not influenced by Farsi)

It appears that the feature [+back] is spreading from pharyngeal consonants to following vowels; we should note that i and e are not affected (Kahn (1976:48)), just as in Armenian they are not affected by Adjarians Law. Evidence from Coeur dAlene (Cole (1987:94)) and Squamish (Kuipers (1967), Cole (1987)), however, indicates that pharyngeal consonants are specified [-ATR] (see discussion in section 2.5) and spread this feature to neighboring vowels (N.B. i and e are not affected, as in Kurdish). We have good reason to suspect, then, that it is actually [-ATR] which is spreading. We can represent the Kurdish developments as in (23): (23) i. [-ATR] spreads from pharyngeals to following vowels D r P n

[-ATR] ii. [-ATR] [+back]

[-ATR]

We can view (23(ii)) as a subset of the rule stated in (20), though it is not clear if the other
10

half of the rule (namely [+ATR] [-back]) applies in Kurdish. There seem to be three distinct stages of [ATR]/[back] interaction in the languages which show [ATR] systems. Many west and east African languages (see Hall and Hall (1980) for examples) have complex [ATR] systems, but no correlation with backness. Somali has a complex [ATR] system with a concomitant rule which fronts [+ATR] vowels. Several dialects of Armenian (Karchevan, Maragha, Cilicia, Agulis, Meghri, etc.; see Muradyan (1960)), as well as Mongolian, Nez Perc, and numerous other languages, fully implement the rule in (20). I have schematized these stages as follows: (24) Stages of [ATR]/[back] interaction i. Classical (original system) ii. Malatia [atr] spreads to V iii. Krzen [atr] [-back] iv. Van loss of voicing contrast {ba da ga}, {pa ta ka} {bA dA gA}, {pa ta ka} {b d g}, {pa ta ka} {p t k}, {pa ta ka}

These groupings may well reflect historical stages in the development of [ATR] systems: when [ATR] becomes phonemically distinctive in a language, we get languages of the type in (24(ii)); when for phonetic reasons the [+ATR] vowels are then fronted, we get languages of the type in (24(iii)); when this process is generalized to the form in (20), we get languages of the type in (24(iv)). Adjarians Law actually presents us with a nice case study of this phenomenonwe have to postulate precisely these stages to account for the results we see in the various dialects, as in (25): (25) Stages of Adjarians Law5 (o) proto-Armenian [+ATR] [-ATR] (i) [ATR] spreads (as in (6)) (ii) [+ATR] [-back] [+ATR] [-ATR] [+ATR] [-ATR] i i i i e e e e

a A a a a

(iii) [ATR] [-back] [+ATR] (as in (20)) [-ATR]

Stages (o) and (i) are basically the same; we assume that [ATR] values of consonants always spread phonetically to adjacent vowels, so that the vowels in (o) would have non-phonemic variants {uoA}. At stage (i) these oppositions are phonemicized. At stage (ii) the [+ATR] vowels are fronted, and we obtain the vowel system of Musaler, etc. Stage (iii) is reflected in the dialects which show [+back] equivalents of i and/or e, such as Karabagh, Meghri, Urmia, etc. We should note that and , the [+back] equivalents of i and e, do not always appear together: many dialects (e.g. Van, Shatax, eastern Mush, New Bayazed, Hadjin, Zeytun; see
5

merges with e after the classical period, as represented in (i).


11

Gharibyan (1953:110-114), Grigoryan (1957:82-84)) have , but not . Consider the cases in the Karchevan dialect: (26) [-ATR] [+ATR] classical esay ki bem gir Karchevan si ki bem gir gloss spouse beesting seat letter

The Armenian dialects actually reflect three levels of complexity, corresponding to the types of complex segments they allow: (27) (i) *[+high, -ATR, -round] * *[-low, +back, -round] (Goris, Musaler) (ii) *[+high, -ATR, -round] (Shamaxi, Urmia, etc.) * (iii) no filters (Karabagh, Meghri) *

One might object that there is no reason to say that the filter *[+high, -ATR, -round] is active in type (i) dialects, since the filter *[-low, +back, -round] alone can account for the vowel gap, but the fact that many dialects show the gap in (ii) indicates that we probably have two different filters overlapping (it is also possible that the cases in (27(i)) are direct products of (25(ii)), which have not undergone the stage represented in (25(iii))). The distribution in (27) indicates that the feature [ATR] is actually playing a role in the Armenian dialect vowel systems. Based on the considerations we have just discussed, we can now specify the general vowel system of the Armenian dialects as follows: (28) i e a high + + + + low + + back + + + + + ATR + + + + + round + + + + -

2.7. If we accept that Adjarians Law is a spreading of [+ATR], we still must explain why {z y l v} trigger the rule, as only voiced stops and affricates are normally considered to be specified [+ATR]. The voiced fricatives are not difficult to account for, as they also require some mechanism to create the pressure differential needed for voicing (Catford (1977:111)); I see no problem with specifying them as [+ATR]. The glides {y v} are not too problematic, either: we have already seen that glides in Maasai (and perhaps Madurese) are [+ATR] (though for different reasons). Unfortunately, l remains problematic. I have been assuming that the change of classical ya- > modern (h)- observed by
12

Weitenberg (1986) was actually triggered by y before it became h, because h is if anything [-ATR]. We could account for h triggering the rule if it were voiced, however, which is basically what Weitenberg proposes. If this were the case, we would have more evidence for an extension of the rule to all voiced continuants. Unfortunately, the issue of whether or not certain Armenian dialects have a voiced h, and what the phonetic properties of such a sound might be, are unresolved. Haneyan (1985) states that the so-called voiced h is a voiced laryngeal sonant in all of the dialects in which it appears, except in the Shatax dialect, where it is a voiced pharyngeal sonant. Further experimental study of the articulation of these consonants is required before we can truly determine what is happening. We should also note that in the Shatax dialect, which supposedly has voiced h, Adjarians Law applies to what were originally initial as, which now are preceded by an initial voiced h. Khachaturyan (1982) attributes this to the presence of a following resonant {l r} (see footnote 1), which in fact seems to account for all the data, though it is not clear how l and r would motivate such a change. We have now accounted for the various consonantal triggers of the rule, and the distribution of its vocalic targets; the only remaining issue is why our rule tends to occur in initial syllables. One might suspect that this could be related to stress, but unfortunately there is not much information available on the stress patterns of the dialects showing Adjarians Law. This question will require further study. 2.8. Now that we have considered the basic facts of Adjarians Law, I would like to consider the related issue of palatalization in the Armenian dialects, which we mentioned briefly in section 2.6. The type of palatalization we see in the dialects showing Adjarians Law is of significant theoretical interest on its own merits, but it also bears on our discussion of ATR. Consider the forms in (29): (29) classical gan sheep go thief gund heap gir letter gerezman tomb > > > > > Maragha Maragha Van Maragha Van kyr ky kynd kyir kyerezman

We can represent Armenian dialect palatalization as follows: (30) C


X

V
X

place place dorsal dorsal

[+back] [-back]

We clearly have to order this palatalization after Adjarians Law and the rule in (20) have applied, as { }, which would be back vowels before these two rules had applied, also trigger palatalization, as shown in (29). This is precisely the form of palatalization we find in innumerable languages; for discussion of this representation of palatalization, see Calabrese
13

(1991). 3. Philological and dialectological aspects of Adjarians law Now that we have accounted for the phonetic and phonological aspects of Adjarians Law, I would like to consider briefly a few more general problems that arise when we look at this phenomenon. First, let us consider the locations of the dialects showing Adjarians Law: (31)
Tiflis 4 8 1 2 6

Yerevan 3 9 11 7

10

1-Goris; 2-Karchevan; 3-Xoy; 4-Kirzan; 5-Maragha; 6-Meghri; 7-Salmast; 8-Shamaxi; 9-Shatax; 10-Musaler; 11-Van The dialects showing Adjarians Law basically cluster together in southern Armenia and southeastern Turkey, except for Musaler, which is in the extreme south of Turkey, on the Mediterranean. We must assume that the speakers of the Musaler dialect originally came from the same area as the other dialects showing our rule; consequently, ascertaining the time at which Armenians moved to Musaler might help us refine our dating of Adjarians Law. Unfortunately, it is extremely difficult to find historical records concerning the Armenian community in Musaler. It seems to have been well-established by the 11th century (Sanjian (1964:53)), and according to Jacquot (1931), there is an Armenian monastery in Musaler dating from the 5th century, but I am not sure these facts tell us much about when the linguistic community was established. Adjarians Law itself has been dated between the 7th and 11th centuries by Adjarian, based on the behavior of Arabic and Turkish loanwords, whereas Muradyan (1962) dated it to the 5th century (his reasons for postulating such an early date are unclear). Sanjians evidence that the Armenian community in Musaler was well-established by the 11th century supports Adjarians upper limit for the rule, but I am not sure we can accurately establish a lower limit before which the law could not have applied. Adjarians Law raises another problem for our study of the history of the Armenian dialects, in particular our reconstruction of the historical grouping of the dialects, and the times at which these groups separated from each other. The type of phonological change that Adjarians Law shows is rare enough in the worlds languages that we would not expect it to develop
14

independently in separate dialects; consequently, we must assume that all dialects showing Adjarians Law were originally members of a single dialect group. It is also possible that the rule spread from an original dialect group to neighboring dialects, but this cannot account for dialects such as Musaler, Shamaxi, and Kirzan, which are substantially distant from what appears to be the center of the rule, in southern Armenia. Based on these considerations, we might expect to find other isoglosses shared by the dialects showing Adjarians Law, but further research on this issue should be the subject of a separate investigation. We should also be skeptical of prior groupings of the Armenian dialects, such as those of Gharibyan (1959, 1969), Djahukyan (1972), and Kortlandt (1978), which do not include all dialects showing Adjarians Law in a single group. Further research on this matter would be useful. 4. Conclusions We have seen that there is compelling evidence to view Adjarians Law as a case of the feature [ATR] spreading from voiced consonants to following vowels, as schematized in (6). In this way we are able to account for the range of consonants that trigger the rule, and why it tends to affect back vowels. The reasons why it tends to occur primarily in initial syllables, and is triggered by l, remain unclear. Our theory also enables us to explain the cooccurrence restrictions we find in Karabagh and Meghri, and motivate the palatalization that takes place in many of the dialects showing Adjarians Law. The dialects we have added to the list showing Adjarians Law may be helpful in understanding certain aspects of the rule; in particular, ascertaining the date at which the Armenian linguistic community was established in Svedia might help us date the approximate time at which Adjarians Law took place. It is hoped that the materials presented in this paper will serve as a springboard for further research. References Adjarian, H. (1901) Lautlehre des Van-Dialekts. Zeitschrift fr Armenische Philologie 1, pp. 74-86, 121138. Marburg. (1909) Classification des dialectes armniens. Paris. (1911) Hay Barbaagituthiwn [Armenian Dialectology]. Moscow. (1926) Khnnuthiwn Maraayi Barbai [Analysis of the Maragha Dialect]. Erevan. (1927) Hayeren Armatakan Baaran [Armenian Root Dictionary]. Erevan. (1953) Khnnuthiwn Vani Barbai [Analysis of the Van Dialect]. Erevan. Aghayan, E. (1954) Meru Barba [The dialect of Meghri]. Erevan. Andreasyan, T. (1967) Svediayi Barba [The dialect of Svedia [Musaler]]. Erevan. Asatryan, M. (1962) Urmiayi/Xoyi Barba [The dialect of Urmia/Xoy]. Erevan. Aoki, H. (1968) Toward a typology of vowel harmony. International Journal of American Linguistics 34. Baghramyan, R. (1961) Kzeni Barba [The dialect of Kirzan]. Erevan.
15

(1964) amaii Barba [The dialect of Shamaxi]. Erevan. Calabrese, A. (1991) Palatalization. Manuscript, Harvard University. Catford, J. (1977) Fundamental Problems in Phonetics. Bloomington: Indiana University Press. Clements, G.N. (1976) Vowel Harmony in Nonlinear Generative Phonology: an Autosegmental Model. Published in 1980 by the Indiana University Linguistics Club. Cole, J. (1987) Planar Phonology and Morphology. PhD dissertation, MIT. Diffloth, G. (1980) The Wa Languages. Fresno: California State University. Djahukyan, G. (1972) Hay barbaagituthyan neracuthyun [Introduction to Armenian Dialectology]. Erevan. Ellis, A. (1877) Pronunciation for Singers. London. Gharibyan, A. (1941) Hamaotuthyun Hay Barbaagituthyan [Short Course in Armenian Dialectology]. Erevan: Erevan State University Press. (1959) Ob Armyanskom Konsonantizme [On Armenian Consonantism]. Voprosy Yazykoznaniya, vol. 5. (1969) A propos de la premire mutation des consonnes occlusives dans larmenien. In Studia classica et orientalia Antonino Pagliaro oblata, vol. 2. Rome. Goldsmith, J. (1990) Autosegmental and Metrical Phonology. Oxford: Blackwell. Gregerson, K. (1976) Tongue root and register in Mon Khmer. in Austroasiatic Studies I, P. Jenners, ed. University of Hawaii. Greppin, J. and Khachaturian, A. (1986) A Handbook of Armenian Dialectology. Caravan. Hall, B., R. Hall, M. Pam, A. Myers, S. Antell, and G. Cherono. (1974) African vowel harmony systems from the vantage point of Kalenjin. Afrika und bersee 57. Hall, B. and Hall, R. (1980) Nez Perc Vowel Harmony. in Vago, ed. (1980). Halle, M. and K. Stevens. (1969) On the Feature Advanced Tongue Root. Quarterly Progress Report #94, Research Laboratory of Electronics, MIT. Haneyan, A. (1985) Grabari Baaskzbi Y Jaynordi Artachoum Ardi Hayereni Barbanerum [The Pronunciation of Grabar Initial Voiced Y in the Modern Armenian Dialects]. In Hayereni Barbaagitakan Atlas, vol. 2. Haudricourt, A. and A. Martinet (1947) Assourdissement et sonorisation docclusives dans lAsie du Sud-Est. Bulletin de la Socit de Linguistique de Paris 43.
16

Hewsen, Robert. (2001) Armenia: A Historical Atlas. Chicago: University of Chicago Press. Hogg, R., and C. McCully (1987) Metrical Phonology: A Coursebook. Cambridge. House, A., and G. Fairbanks (1953) The influence of consonant environment upon the secondary acoustical characteristics of vowels. Journal of the Acoustical Society of America, vol. 25. Jacobson, L. (1977) Voice-quality harmony in Western Nilotic languages. In Vago, ed. (1980). Jacobson, L. (1978) DhoLuo vowel harmony. UCLA Working Papers in Phonetics 43. Jacquot, P. (1931) Antioche. Beirut. Jesperson, O. (1889) The Articulation of Speech Sounds Represented by Means of Analphabetic Symbols. Marburg. Kahn, M. (1976) Borrowing and Variation in a Phonological Description of Kurdish. PhD. dissertation, University of Michigan. Khachaturyan, A. (1982) Aayan ew hay barbaagituthyun [Adjarian and Armenian Dialectology]. In Hayereni Barbaagitakan Atlas, H. Muradyan, ed. Kohler, K. (1984) Phonetic explanation in phonology: the feature fortis/lenis. Phonetica 41:150-174. Kortlandt, F. (1978) Notes on Armenian Historical Phonology II (The Second Consonant Shift). Studia Caucasica 4. Kuipers, A. (1967) The Squamish Language. Mouton, the Hague. Ladefoged, P. (1964) A phonetic study of West African languages. West African Language Monograph no. 1, Cambridge. Lindau, M., L. Jacobson, and P. Ladefoged (1972) The feature Advanced Tongue Root. UCLA Working Papers in Phonetics 22. Lindau, M. (1976) Larynx height in Kwa. UCLA Working Papers in Phonetics 31. Lindblom, B. and J. Sundberg. (1971) Acoustical consequences of lip, tongue, jaw, and larynx movement. Journal of the Acoustical Society of America 50. Maddieson, I. and P. Ladefoged (1985) Tense and lax in four minority languages of China. Journal of Phonetics 13:433-454. McCarthy, J. (1991) The Phonology of Semitic Pharyngeals. Manuscript, University of Massachusetts, Amherst.

17

Muradyan, H. (1960) Karevani Barba [The dialect of Karchevan]. Erevan. (1973) Inhps haskanal Aayani orenkh [How to understand Adjarians Law]. Banber Erevani Hamalsarani, 1973.2.161-170. (1982) Hayereni Barbaagitakan Atlas [Armenian dialect atlas], vol. 1. Erevan. (1986) How to Interpret Adjarians Law. In Greppin (1986). Muradyan, M. (1962) atai Barba [The dialect of Shatax]. Erevan: Erevan State University Press. Rialland, A. and Djamouri, R. (1984) Harmonie vocalique, consonantique et structures de dpendance dans le mot en Mongol Khalkha. Bulletin de la Socit de Linguistique de Paris. Rigsby, B. and M. Silverstein. (1969) Nez Perc vowels and Proto-Sahaptian vowel harmony. Language 45. Sanjian, A. (1964) Armenian Communities in Syria under Ottoman Dominion. Harvard University Press. Sezer, E. (1985) An Autosegmental Analysis of Compensatory Lengthening in Turkish. In Studies in Compensatory Lengthening, Wetzels and Sezer, eds. Steriade, D. (1987) Redundant Values. Papers from the 23d annual meeting of the Chicago Linguistic Society, Parasession on Autosegmental and Metrical Phonology, A. Bosch, B. Need, and E. Schiller, eds. Stewart, J. (1967) Tongue Root Position in Akan Vowel Harmony. Phonetica 16. (1971) A theory of the origin of Akan vowel harmony. Proceedings of the sixth international congress of phonetic sciences, Prague: Academia Publishing House. Trigo, L. (1987) Voicing and Pharyngeal Expansion: ATR Harmony in Buchan and Madurese. Manuscript, MIT. Tucker, A. and J. Mpaayei. (1955) A Maasai Grammar. Longmans Green and Co. Vago, R., ed. (1980) Issues in Vowel Harmony. Amsterdam, John Benjamins. Weitenberg, J. (1986) Additional h-, initial y- and Indo-European *y- in Armenian. In La place de larmnien dans les langues indo-europennes, M. Leroy and F. Mawet eds. Louvain. Westbury, J. (1979) Aspects of the temporal control of voicing in consonant clusters in English. Department of Linguistics, University of Texas at Austin. (1983) Enlargement of the supraglottal cavity and its relation to consonant voicing. Journal of the Acoustic Society of America 73:1322-1336. Williamson, K. (1974) Some reduced vowel harmony systems. Afrika und bersee 57.

18

You might also like