The Dangari tongue of choke and machoke: tracing the proto-language of Shina enclaves in the Hindu Kush.
Article Type: Report
Subject: Grammar, Comparative and general (Analysis)
Indo-Iranian languages (Varieties)
Indo-Iranian languages (Comparative analysis)
Grammar, Comparative and general (Phonology)
Author: Liljegren,Henrik
Pub Date: 01/01/2009
Publication: Name: Acta Orientalia Publisher: Hermes Academic Publishing Audience: Academic Format: Magazine/Journal Subject: Social sciences Copyright: COPYRIGHT 2009 Hermes Academic Publishing ISSN: 0001-6438
Issue: Date: Annual, 2009 Source Volume: 70
Geographic: Geographic Scope: Pakistan Geographic Code: 9PAKI Pakistan
Accession Number: 300652465
Full Text: Abstract

Data from four little-studied varieties of Indo-Aryan (Southern Palula, Northern Palula, Sawi and Kalkoti) spoken in the Hindu Kush is analyzed and discussed from a historical-comparative perspective. Evidence is presented showing that Kalkoti, until recently only tentatively classified, is part of this particular cluster of closely-related Shina varieties. An attempt is made at reconstructing some phonological and grammatical features of a common source speech, here named Proto-Dangari, and the order in which the present-day varieties may have split off. An important conclusion drawn is that Southern and Northern Palula probably are more distantly related than present-day similarities seem to indicate, the high degree of synchronic similarity instead being due to relatively recent convergence taking place in southern Chitral. It is hypothesized that the present speech communities are the result of two different westward routes of migration, one geographically linking Southern Palula (Ashreti) and Sawi with Chilas, the other linking Northern Palula (Biori) and Kalkoti with Tangir, both located in the same general area of the main Indus Valley.

Keywords: Indo-Aryan, Shina, Dangari, Palula, Sawi, Kalkoti, Gawri, Kohistani, convergence, vowel alternation, vowel raising, recon struction, comparative method, aspect, perfective, imperfective, tense, grammaticalization, aspiration, consonant cluster, chain shift.

1. Introduction

This paper is an attempt at applying some insights that can be drawn from historical linguistics on data from four little-studied varieties of Indo-Aryan Shina: Southern Palula, Northern Palula, Sawi and Kalkoti. (1) Each of these is spoken by a dispersed community in a region west of the main Shina belt. Data is also presented showing that Kalkoti, which has so far only tentatively been classified as belonging to Shina (Strand 2001: 254-5, 258; Joan Baart, p.c), is indeed part of this particular cluster.

In Section 2, the four speech communities are briefly introduced, followed, in Section 3, by a presentation of first-hand data supporting the classification of Kalkoti as a Shina variety. In Section 4,1 focus on issues pertaining to the relationship between Southern and Northern Palula, such as a common origin and shared development. After that, in Section 5, I present some features of a hypothetical source speech of these two varieties. In Section 6, I widen the scope to include the more distantly related varieties Kalkoti and Sawi in a discussion along comparative lines, in an attempt to trace a source speech (ProtoDangari) of all four of the Shina enclaves. In Section 7,1 sketch some lines of development from Proto-Dangari into the present-day varieties and suggest how these four may be grouped vis-[a]-vis each other. The most important conclusion drawn is that the apparently closely-related Southern and Northern Palula may in fact be a little more distantly related than the present-day similarities seem to indicate, these similarities instead being the result of more recent convergence. In Section 8, I hypothesize how two different routes of migration-one resulting in Southern Palula and Sawi, and another in Northern Palula and Kalkoti-took the speaker communities to where they are presently found. Finally, in Section 9, I point to the importance of a closer study of these, in many respects archaic, varieties to gain a better understanding of the early development of Shina at large.

In most cases, I base the analysis on my own field data, collected since 1998 in collaboration with local researchers and informants, among whom I especially want to mention and give credit to Naseem Haider of Ashret and Muhammad Zaman Sager of Kalam, both associated with the Frontier Language Institute in Peshawar. (2) Thanks also to Ajmal Nuristani who assisted me in collecting Sawi data and obtaining information on this speech community. (3)

2. The speech communities

Southern Palula is spoken by 5-6,000 individuals in Ashret Valley in the southern part of Chitral District in northwestern Pakistan. Palula (paaiuulaa) is commonly referred to within the community as atshareetaa 'the speech of Ashret', and by the district's Khowar-speaking majority as Dangarikwar, the latter a designation inclusive of the speakers of Northern Palula. Ashret Valley is situated at the main entry point into Chitral through the Lowari Pass. Although often considered a single large village, it really is a long-stretched area consisting of seven separate settlements. There are few speakers of any other language residing in Ashret Valley and the language is vital and actively transferred to the next generation. The degree of multilingualism, however, must be considered high, with Pashto and Khowar as the most common second languages. Most Palula samples in the literature are taken from this variety. It was confirmed through Morgenstierne's pioneering fieldwork on this variety (Morgenstierne 1932; Morgenstierne 1941) that Palula is indeed part of the Shina linguistic cluster. Apart from the results of my own fieldwork (Liljegren 2008), Strand (1997/2008) offers a snapshot presentation of the variety, partly based on Morgenstierne, partly on a brief fieldstudy in the 1980s.

Northern Palula is spoken by 3-4,000 individuals, primarily in Biori Valley, in the villages Mingal, Dhamaret and Bhiuri. The valley is situated to the north of Ashret Valley, and as in Ashret, the speech serves as the sole instrument of communication within the valley. Here, too, we find a high degree of multilingualism, with Khowar as the second language of choice. This variety (with slight variations) is also spoken in Puri, (4) a village in the Shishi Koh Valley, although with considerably diminished vitality, (5) as well as in a portion of Kalkatak, a village in the main valley, about two kilometers south of the mouth of Biori Valley. (6) Northern Palula has not been the main subject of any study in the past, although one of Morgenstierne's informants belonged to Biori and another to Puri. Because Morgenstierne focuses on Southern Palula, he only offers fragmentary comparisons with Northern Palula (1941: 8). The variety is usually referred to as paaiuulaa or paaulaa by its speakers.

Sawi is the speech variety of Sau, a village situated on the east bank of Kunar River in Afghanistan, about 20 kilometers south of the border town Arandu in southern Chitral. (7) It is uncertain to what degree Sawi is spoken in this village today. According to K. D. Decker's (1992) informants there had been approximately 8-12,000 people living in the village before the long period of war and unrest. After that most of the people had moved out and settled in various refugee camps in Pakistan, primarily in Chitral and Dir. (8) That the variety spoken in Sau is closely related to the Palula varieties of Chitral was pointed out already by Morgenstierne in the first half of the last century (Morgenstierne 1941: 7), and was further confirmed by the more extensive study undertaken by Buddruss (1967: 11):

Many of my informants seemed to be aware of the speech of Ashret and its striking similarities to their own variety. However, no major interaction or contact between the two communities seem to have taken place in the recent past, and the population of Sau has already for a long time been included in the all-surrounding Gawar community, sharing their identity in all aspects save the language (Cacopardo & Cacopardo 2001: 231-2).

Kalkoti is spoken by approximately 6,000 people in the village of Kalkot in upper Panjkora Valley in Dir Kohistan (Pakistan). (9) As no systematic survey has been carried out in Dir Kohistan there may be other locations in the more inaccessible side valleys where this or similar varieties are spoken. Most other villages in the main Panjkora Valley, from Rajkot (Patrak) upstream, are Gawri-speaking, (10) and that the speech of this village may be something rather different was first hinted at in the sociolinguistic survey carried out by Rensch and his SIL colleagues (1992: 7): (11) "The linguistic variety spoken in the village of Kalkot in Dir Kohistan seems to be quite distinct from that spoken in the surrounding villages of Dir Kohistan and in Kalam, although it is obviously related." It was pointed out that Kalkotis understand the Gawri spoken in the same or in neighboring villages, but not vice versa. When carrying out intelligibility tests in Dir and Swat Kohistan, the survey team concluded that Kalkoti stood out as significantly different from the other speech varieties, in fact "generally considered to be a different language altogether" (Rensch 1992: 14). Gawri speakers from Swat Kohistan who listened to recorded texts from various locations in Dir Kohistan understood, for instance, most of the rather different speech variety of Rajkot whereas "when listening to the Kalkot text, some of the men in Kalam looked puzzled and asked what kind of language it was" (1992: 14). Strand (2001: 255, 258) tentatively classifies Kalkoti as one of a number of "dispersed dialects" of Chilasi Shina, based on the word list presented in the above-mentioned survey report (Rensch 1992: 159-76).

3. The classification of Kalkoti

Before comparing the four varieties, something more specific needs to be said about the relationship between Kalkoti (Klk) and Palula (Pal) on the one hand, and between Kalkoti and Gawri on the other. Since Pal and Gawri themselves are related to each other as IA languages, (12) it is rather challenging to determine the exact place of Klk in relation to these two varieties, but I primarily want to show that Klk in its essential parts is a Shina variety.

It is obvious that Klk, spoken for a long time in the vicinity of Kohistani varieties, would have been influenced by the latter. However, certain classes of words are much less likely to be borrowed, such as kinship terms, simple and basic verbs, lower numerals and pronouns (Trask 1996: 23); when taking these into account the Shina origins of Klk become obvious. Nearly all basic kinship terms in Klk have close cognates in Pal, whereas only a few such terms are similar enough between Klk and Gawri to define them as close cognates. The most striking in this comparison (see Table 1) is the close correspondence between Klk and Pal for the most frequent and important verbs 'be' and 'do', the latter also being the semantically "empty" component that together with e.g., a noun makes up many so-called conjunct verbs (Masica 1991: 326). The present tense form of 'be' in Klk represents a regular development of A-dropping and apocope from an earlier *hino, etc., with a stem hi- (or ha-) typical of the Shina copula, whereas in Kohistani languages the copula verb has a th-stem. A th-stem for 'do', on the other hand, is a typical Shina feature, while Kohistani follows the main IA languages with a karstem. With lower numerals, primarily '11' and '12' show significant differences between Pal and Gawri; here Klk clearly goes with Pal.

The personal pronouns (Table 2) also offer interesting points of comparison. Apart from form similarities, the Pal and Klk systems share most distinctions made, although Pal contains differentiations not found in Klk, such as between the masculine and feminine third person singular, and a three-way distance contrast. The forms and the distinctions used for first person singular and second person singular are also parallel to those in Pal. In the plural, however, only first person plural distinguishes between a nominative and an ergative form in Klk, as is the case with plural pronouns in Pal, whereas there is no such distinction made in the second and third person plural in Klk. The most obvious differences between the Pal and Klk systems on the one hand and the Gawri system on the other are the first and second person plural and the third person distal forms (see Table 2).

4. Northern and Southern Palula and their common source

Pal, that is Northern (NP) and Southern Palula (SP), is usually described as a single "language" or "dialect" (Morgenstierne 1941:7; K.D. Decker 1992:7; Masica 1991:21; Strand 2001:253, 258), as well as the speech shared by a single ethnic community (Cacopardo & Cacopardo 2001: 79-143). Although the former is not very surprising, from a synchronic perspective, the latter is a more complex issue.

To the outsider, particularly the Khowar-speaking majority of Chitral, the people and the speech of Ashret and Biori Valleys are indistinguishable, the people referred to as Dangarik and their speech-dramatically different from Khowar--as Dangarikwar. As there are no other closely related linguistic communities in Chitral, they are seen as a single community in much the same way as the neighboring Dameli community. Internally, however, the picture is less clear-cut. Indeed, the "southerners" find the speech of the "northerners" rather similar to their own and largely comprehensible, and vice versa, but both have the idea that the other variety has somehow deteriorated from its pure form. It is not uncommon for "southerners" to hold that they speak atshareetaa while the "northerners" speak another language, bhioorcaa or paaiuulaa. Among educated people in both valleys, however, there is a notion about a shared language, with minor dialectal differences, called Palula. (14)

Although speakers of the two varieties have interacted and also intermarried for a long time, the people of Ashret do not consider the people of Biori their own kin. They have no genealogy in common and their respective origins differ, no matter how well they understand one another. We will have reason to return to this issue, but for now we can imagine two different scenarios: Either we have a single speech community that geographically (i.e., settling in two separate mountain valleys) has been split in two and gradually diverged and over time become more different from one another, or we observe two speech communities with two distinct (but not too distantly related) source varieties that have merged due to prolonged contact.

4.1. Shared Palula features

Regardless of the position we take on divergence versus convergence, there are some important features that these two varieties share that are not documented in any other Shina varieties. One of them is the use of de, a grammaticalization of a form of 'give', as a tense marker, seen in examples (1) and (2). In many other varieties (including Klk and Sawi) we instead observe a grammaticalization of'be' or 'come'.

Another outstanding feature is the feminine plural suffix -(i)m, occurring with one of the noun classes, as in example (3), as the regular agreement suffix with verbs, as in example (4), and with many predicative adjectives, as in example (5). I know of no other Shina variety where the feminine plural is formed with a nasal consonant.

The extent to which the two varieties agree lexically and morphologically is also overwhelming. Their nominal as well as their verbal paradigms are, for instance, virtually identical. Before turning to the apparent differences between NP and SP and an application of the comparative method, we will take a look at some possibilities of internal reconstruction (Fox 1995: 146) in each of the two varieties.

4.2. Intravariety vowel alternation

One of the more promising sets of data we find in SP compounds or derivations (Table 3, following page), containing a number of vowel alternations between the form of e.g. a simple noun and the form of the noun when it occurs as part of a compound or a derivation.

From the following data we can already make some generalizations. First, there are vowel sounds in the derived words that alternate with other vowel sounds in the simple words: oo alternates with uu, aa with oo, ee with ii and short a with long aa. There are also a few examples of alternations between o and uu and between au and uu. Second, we observe that whereas these vowels are unstressed (or rather unaccented) in the polymorphemic words, they are stressed in the monomorphemic words. The latter observation should make us a bit cautious when interpreting our data; while it may be true that the forms in the first column represent an earlier pronunciation of these words (whether derived or nonderived), predating a possible vowel shift affecting only the vowels of the stressed syllables, it is equally possible that the vowels in the derived forms have lost their "original" quality or quantity due to destressing.

Not entirely surprising, some of these vowel alternations are resurfacing as we compare cardinal and ordinal numbers (Table 4). The aa~ oo alternation is seen in baasuma/b[o']os, the ee~ii alternation in treesuma/triis, and the a~aa alternation in, for instance, dasumal daas. There are also individual examples of the o~uu alternation.

In some noun paradigms, too, we find some of these alternations. The regular pattern of case and number inflection is through suffixes added to a noun stem, without any changes to the stem itself, such as in the following singular vs. plural forms: sun/siina 'string bed'; kud/kudi 'wall'. In some other cases, however, there are notable stem changes that are not phonologically motivated. For instance, while a shift of the accent from the root to a suffix is predictable with most nouns in SP, (15) this is not the case with the nouns in Table 5.

The alternation a~aa occurs with a group of nouns from two main classes, one that inflects with -a and one with -/ The example with aa ~oo alternation belongs to a small set of irregular nouns.

A study of the verb paradigms of SP confirms these observations. Some verbs show a paradigmatic stress alternation similar to that of the noun paradigms above. For instance, the future third person singular receives stress on the stem, whereas the stem of the present masculine singular is unstressed. For some verbs this results in vowel alternations (Table 6) that by now should strike us as familiar.

All alternations we have seen so far are various cases of stem alternations, and we have been able to identify a few common vowel alternations in SP that some way or another are related to accent. These are summarized in Table 7, with some possible developments.

Nothing is really controversial with those suggestions, even though they remain tentative at this point. The direction is the same for all of the suggested shifts: a historical vowel raising and/or tensing, mirroring a universal tendency (Trask 1996: 89). Data from NP, too, give evidence to raising/tensing, although not along exactly the same lines. Probably the most interesting in this regard is the extent of paradigmatic vowel alternations. While there are only a few examples of stem alternations in the noun paradigms of SP (as the ones in Table 5), NP is full of them, as shown with only a few examples in Table 8.

The corresponding forms for 'fire' in SP are, for instance, angoor/angoora/angooram/aagoorii, i.e., with a constant oo in all of its forms. On the other hand, some nouns that in SP show stem alternation do not alternate in NP: sar/sar/'lake'; hal/haM 'plough'.

4.3. Intervariety vowel correspondences

In spite of the differences between the two varieties, it is striking how systematic these (vocalic differences) seem to be. On basis of the regularity of the correspondences (Fox 1995: 65), we can therefore establish some equivalences and draw some tentative conclusions about proto-forms.

Already from a brief look at Table 9, we observe that there are in fact not very many one-to-one correspondences between individual vowel sounds in one variety as compared with the other. In effect we obtain a large number of correspondence sets. It does not help us a lot, when, for instance, a long oo in SP corresponds to three different long vowels in NP, aa, oo and uu. Two important factors in vowel change at large are accent placement and syllable structure (Trask 1996: 64), and we will therefore include them as conditioning factors to obtain useful sets of correspondence, as shown in Table 10.

The only item we are not able to deal with in the format of this table is item 14, as the syllable structure differs between the varieties. It is, however, more than likely, on phonological grounds, that the SP form has been subject to apocope after raising of the vowel: *naaw > *noow> noo 'name'. For the rest of the items we have a somewhat clearer picture, and we can already start making some hypotheses about proto-vowels and the vowel shifts that are likely to have taken place in each of the varieties. For the sake of simplicity we will limit our discussion to accented vowels. (16)

From an initial look it is clear that it is in the closed syllables we find most of the variation. We also note that the first-mora accented long vowels (such as oo) tend to display more intervariety diversification than the short vowels or the second-mora accented long ones (e.g. oo). Much of the differences in vowel quality and quantity between NP and SP can be attributed to three shifts taking place from the proto-language into SP. In closed syllables, *a was lengthened to aa, *aa was raised to oo, and *eeto ii. The two raising processes are relatively straightforward (SP: akoos 'eleven' < *akaas, and SP: chiitr 'field' < *cheetr), whereas the lengthening of *a happened to produce some irregularities in SP. Along with the lengthening of * a to aa, some of the vowels developed a first-mora (falling) accent and some of them a second-mora (rising) accent. The conditioning factors seem to have been aspiration and word structure. Second-mora accent evolved in those closed syllables that had an aspirated onset (including a single onset h) and in a syllable that was preceded by another unstressed syllable: khaar < *khar; basaand < *basand; haat 'hand' < *hat. (17) All other lengthened *a developed a first-mora accent: baat 'stone' < *bat Only first-mora accented long *aa and *ee were subject to raising, whereas second-mora accented aa and ee kept their quality: SP/NP kaal 'year' < *kaal; SP/NP dees 'day' < *dees. The contrast between NP aa and SP uu, as in the pair saarvs. suur 'father-in-law', must be attributed to a slightly different development, and it will be necessary to suggest a proto-form different from any of the two derived forms. I suggest the diphthong *au as a likely protophoneme, supported by internal evidence (cf. sa(w)ura 'household of father-in-law'), developing through raising and monophthongization into SP uu, and through monophthongization into NP aa. Another complicating issue is that NP e sometimes corresponds to SP e and sometimes to SP a, in roughly the same phonological environment. For reasons that will not become entirely clear until we discuss further internal reconstruction and comparison with other varieties, we will opt for a proto-phoneme *a/as the source of the e in NP/SP be 'we', whereas e~a in NP ate and SP ato 'bring!' probably go back on *e in the proto-language.

In the phonologically more archaic NP, there is only one notable case of vowel raising that has not taken place in SP, that from closed syllable *oo to uu in NP: truu 'three' < *troo. Again, this has only affected first-mora accented vowels, while the second-mora accented ones remained "untouched": phoo 'boy' < *phoo. All of this points in the direction of a 12-vowel system in the proto-language, as shown in Table 11.

[TABLE 11 OMITTED]

In Table 12, the developments that have taken place between a protolanguage and the modern-day varieties are summarized. (18)

[TABLE 12 OMITTED]

5. Tracing the precursor of Palula

The attempted reconstruction outlined in Section 4 is probably as far as the comparative method can take us in our search for an earlier stage of Pal. That, however, is not the end of the story, and we have already hinted at some other developments that are tracable through internal reconstruction. We will therefore make an attempt at sketching some features of an even earlier stage; one that we will call pre-proto-Palula, using a combination of the comparative method and internal reconstruction:

It is in the closed syllables in NP we find some of the more conservatively pronounced vowels, whereas in most open syllables we can in fact witness the results of an earlier raising or tensing in both varieties (stage 1 in Table 13). A lengthening of short accented *ain the open syllables of the pre-proto-language resulted in *aa in the proto-language (giving the pre-proto-form *kaku for NP/SP kaaku 'older brother'), while in closed syllables tensing took place only in SP (stage 2). That would simultaneously explain the alternations in the noun paradigms of both NP and SP, as seen in Table 13 and Table 14.

Similarly, long first-mora accented *aa were raised to oo in open syllables (further raised to uu in NP), resulting in the following forms: SP: troo, NP: truu< *traa 'three'. In some cases, an umlaut formation intervened to instead produce forms like SP/NP deeri 'beard' from *daari (cf panardooru 'village elder; lit. white-beard'). Second-mora accented aa on the other hand are likely to have kept their proto-vowel quality. For phoo we therefore have to assume a shift from a firstmora accent to a second-mora accent after raising had occurred: phoo 'boy' < *phoo < *phaa.

Since the processes of raising *ee to ii'and *oo to uu are more or less parallel, I will not go into any more detail here. Naturally, there is a difficulty from an exclusively comparative perspective to know which of the u and uu in the modern varieties are products of raising, and which are inherited high vowels. Having access to Old IndoAryan (OIA) cognates helps us differentiate niilu 'green, blue' as having an inherited long high front vowel (Turner 1996: 7563), while the same vowel quality in miisa 'men' most likely has developed from the proto-form *meesa. (19) As is the case with Greek (Trask 1996: 90), the high front position has possibly become a "sink" where vowels from other positions in the system have been collected and kept. It is only to be expected that the it, and to an almost equal extent uu, wherever they happen to occur in the lexicon of today's Pal varieties, are of a very mixed origin. The short vowels, other than *a, seem to have maintained their position in the system to a rather high degree, possibly with e acting as another "sink", akin to the high front position just mentioned.

The vowel system of Pre-Proto-Pal was probably not very different from the system suggested for Proto-Pal, i.e. one with five basic positions and length contrasts, but while the diphthongs au and ai are rather marginal and heavily restricted in the modern varieties, they may have played a role in Pre-Proto-Pal phonology on par with the long vowels. I trace the vowels in nuu 'nine', suur/saar 'father-inlaw', cuur/caar 'four', siluuk/silaak 'story' to the proto-diphthong *au, with a fair degree of confidence, whereas I prefer to be less dogmatic with my assigning *bhain as the proto-form of bheen, 'sister' and * bai as, the ancestor of be 'we'.

6. Reconstructing Proto-Dangari

The conclusions reached so far take us to a stage where we are dependent on comparative data from varieties more divergent than SP is from NP, to establish features of a more remote ancestor language.

6.1. Kalkoti findings

As yet no detailed study of Klk has been published, (20) and it will therefore be necessary to present a rough outline of its phonology and some of its morphological features. I hasten to add that this is a tentative analysis, and a future in-depth study will be needed to confirm these findings.

The consonant inventory (in Table 15, exemplified in Table 16) is similar to that of most languages in the immediate region, with its dental/retroflex contrasts for plosives and retroflex/palatal contrasts for fricatives. While aspiration is contrastive for voiceless stops, there is no such contrast for any voiced sounds.

A number of sounds may have been introduced via loans, primarily from Pashto and more recently from Urdu: q, ts, x, h, z, y, r. Some of these (ts, h and r) may also be examples of reintroduction of phonemes present (and later lost) in previous stages. Some phonemes are restricted. I have, for instance, not found any unambiguous t word-initially, (21) but that may of course be due to limitations of my data. I only have one single occurrence of th, and even that seems to alternate with th. ithyil~ithyil. The contrast between ch and c is questionable, lacking word-initial examples of the unaspirated member. The phonemic status of g is not entirely clear; there is at least a trace of a cluster ng alternating with this sound.

I suggest an analysis of Klk providing for a 10-vowel system (in Table 17, and exemplified in Table 18). (22)

[TABLE 17 OMITTED]

The primary contrastive features, however, are qualitative rather than quantitative, as the phonetic realizations of the vowels give at hand: (23)

The main motivation for not assigning instances of [e] phonemic status e is that a great deal of the instances of / fluctuate considerably in the vicinity of [i] and [e] in their pronunciation. The same holds for vowels that are heard as short [o], that they most likely are part of the u phoneme. I do not hold it impossible that previously contrasting [i] and [e], as well as previously contrasting [u] and [o], may have fused, perhaps quite recently. (24) With one of our main informants, the contrast between a and a was not entirely clear, both of them pronounced as central open vowels, whereas one of the other informants usually made a clear back [[c]] vs. front [[epsilon]] contrast. I have not been able to establish any minimal pairs with oral vs. nasal vowels and therefore refrain from introducing a series of nasal vowels along with an oral series, although nasalization does seem to be phonemic.

The only two definite syllable-internal consonant clusters found are word-initial tr and dr. There is also a general preference for closed syllables word-finally, a word-structure feature shared with neighboring Gawri (Baart 1997: 37).

I do not have enough data to account for Klk noun morphology, an area therefore needing extensive follow-up. A main objective of a future study would be to determine whether Klk (like Pal) forms plural and non-nominative case forms primarily by adding suffixes to the stem, or if it (like Gawri) employs stem modification (Baart 1999: 15, 35). (25) Neither do I claim to have data to give a complete picture of Klk verb morphology, but a somewhat sketchy account will nevertheless be given. The two most commonly occurring TMA categories that correspond to distinct forms can be characterized thus:

1) Activities, states or actions in the present (and possibly in the future) or those that can be considered continuous. I regard this as primarily an imperfective. The regular endings are: -uun (MSG), -aan (MPL), and -Un (F). In this category, the verb always agrees with the subject, whether transitive (6) or intransitive (7).

These forms are found with almost all verbs in the material, the only obvious irregularity, in this respect, being the copula, with a single form for this category: in 'am, is, are'. With some verbs, -oon instead of -uun is heard, but that could also be related to stress: 'maroon 'kills, is killing' (contrasting with ma'niun 'dies') and Moon 'aches'.

2) Activities, states or actions in the past that are completed. I regard this as primarily a perfective. The two most frequently occurring endings are -il and -aal. The difference between the two is conjugational, exemplified in (8) and (9).

For at least one verb, the ending is -aalrather than -aal: thaal 'did'. Usually no gender or number differentiation occurs, but a number of verbs show irregular formations (Table 19), some of them radically suppletive, and a number of them forming perfective with -t. (26)

Apart from the imperfective and perfective verb forms, there are isolated examples of oth . forms. It is quite likely that Klk has a mode category that displays at least partial person agreement (see Table 20).

Although the primary TMA contrast in Klk is aspectual, the data suggests a more peripheral layer of tense distinction. What apparently is a past tense -s can be suffixed to the aspectual forms. A fragmentary picture of this process is given in Table 21. The last segment of the perfective is dropped if consonantal and replaced by -s to form past tense (cf. dris and dris).

6.2. Sawi revisited

Turning to Sawi (Sw), I follow Buddruss to give a brief outline of this variety, focusing on phonology.29 Again, the consonant inventory (Table 22) is rather similar to many other such systems in the region.

Buddruss (1967: 15-6) expresses some uncertainty as to the phonemic contrast between some aspirated and unaspirated stops, such as between ch and c , between ch and c, and more generally between unaspirated and aspirated voiced plosives. In fact, in none of the words that Buddruss (B) tentatively transcribes with an aspirated voiced plosive do I detect any aspiration (L, my own data): L: dorim 'pomegranate', damn 'earth', bervhe(e)n 'sister' vs. B: d(h)aarim, dheereen, bhyeen. I do not exclude that these words were indeed still weakly aspirated by B's informants half a century ago,30 but is altogether lost as a feature of modern-day Sw. The ts is marginal, occurring almost exclusively in Pashto or Gawarbati loans. Other sounds that are primarily found in loans are x, y, q, z and f. Whether n, r as well as d have full phonemic status is also questioned (Buddruss 1967: 16-7). The I is a voiceless lateral, contrasting in voice with I, but is phonetically a fricative. (31)

As for the vowels, B suggests a symmetric 8-vowel system (Table 23). Although he includes length in his transcription, he questions its phonemic relevance, as almost all vowels show a quantitative fluctuation. The only relevant length contrast is between a and aa.

[TABLE 23 OMITTED]

Following is an approximate IPA representation of Buddruss, partly drawn from his prose description:

The aa and o are acoustically very similar and Buddruss (1967: 12-13) expresses a difficulty in differentiating consistently between them. The o is slightly labialized, but not as rounded as its Persian equivalent. Both o and oo can be heard as a short [o], but there is never a variation between [d] and [o:] in the same word; they are very definitely realizations of two different phonemes, oo and u may on the other hand be neutralized, and it is only in word-final position that the phonemic contrast between them is beyond doubt. The primarily qualitative contrast e~ee plays a role in morphology, although the latter is quantity-wise very variable. In my own data, the pair sen 'roof and seen 'string-bed' illustrates this contrast. A sound [e:] (varying with [e:] and [e]), B considers an allophone of aa, rather than an independent phoneme, as it occurs only as an umlaut counterpart to [e:] in masculine-feminine pairs such as [ghe:nu] (M) ~ [ghe:ni] (F) 'big' or the present tense endings [-e:nu] (M) ~ [e:ni] (F). A neutralization analogous to that between oo and u, is also found between / and ee, leaving monosyllabic words ending with vowels as the only fully contrasting environment for these two vowels. B identifies two distinct diphthongs at and aw(Buddruss 1967: 12-14).

The only word-initial clusters in Sw are plosives followed by either r or y, but a few loanwords with the clusters pi and bl are also noted. Word-finally, clusters with a nasal followed by a stop are found, and in a few loans some other cluster types as well (Buddruss 1967: 17).
Dagegen ist die nahe Verwandtschaft des Sawi mit dem Phal. bereits
   durch einen Blick in Grammatik und Vokabular evident und wird
   iiberdies durch die Angabe meines Gewahrsmannes bestatigt, dass
   er die Sprache der Leute von Ashret verstehen konne. Dennoch
   sind die beiden Sprachen keineswegs identisch mit einander.


(1) biidu  gaadu    tesee     dabdaba   de
     very   big      his      dignity   be.PST

'His was a most dignified person.' (NP)

(2) xaamaar             ba    mheeril-u      de
dragon/big.snake       PRT    kill.PFV-MSG  PST

'And the dragon was killed.' (SP)


(3) bhiira       kimaan-a      ceeli-m          kimaan-a
 he.goat.PL   sell.PRS-MPL  she.goat(F)-PL   sell.PRS-MPL

'They sell their he-goats and their she-goats.' (NP)

(4) bijeeli dhiya         tas/i     heensil-im      de
several daughters         his       stay.PFV-FPL    PST

'He had several daughters.' (SP)

(5) ani     peeruun-a      pureenim
  these    shirt(F)-PL    old.FPL

'These shirts are old.' (NP)


Pre-proto-Palula
[up arrow]
Internal reconstruction
[up arrow]
Proto-palula
[up arrow]
comparative method
[up arrow]    [up arrow]
NP              SP


ii is pronounced [i:]/[i]
i is pronounced [i]/[e]
ee is pronounced [e:]/[e]
a is pronounced [a]/[[epsilon]]/[ae]/[e]
aa is pronounced [ae:]
aa is pronounced [a:]
a is pronounced [c:]/[D]/[a]
oo is pronounced [o:]/[o]
u is pronounced [u]/[[??]]
uu is pronounced [u:]/[u]


(6) ma     tipa     tusaa~     ta   a   qisa   th-uun
    I.NOM  now      you.OBL    to   a   story  do-lPFV.MSG
    'I'll  now tell you a story.'

(7) su          y-iin
    3SG.NOM     come-iPFV.F
    'She's coming.'


(8)  asi     mukha    a    puu    y-aal
   Ipl.gen   front    a    boy    come-PFV
'A boy came up to us.'

(9)  ta     a     phit    mar-il
   3SG.ERG  a     fly     kill-PFV
'He killed a fly.'


i(B: i) is pronounced [i]/[i:]/[c]
ee(B: e) is pronounced [e:]/[e]
e(B: [epsilon]) is pronounced [ae]/[a]
a(B: a) is pronounced [[epsilon]]/[e]
aa (B: a) is pronounced [a:]/[[epsilon]:]
o(B: c) is pronounced [o]/[o]
oo(B: o) is pronounced [o:]/[o]
w(B: u) is pronounced [u]/[o]/[u:]
Gale Copyright: Copyright 2009 Gale, Cengage Learning. All rights reserved.