<?page no="0"?> Jan Henrik Holst Advances in Burushaski linguistics <?page no="1"?> Advances in Burushaski linguistics <?page no="2"?> Tübinger Beiträge zur Linguistik herausgegeben von Gunter Narr 547 <?page no="3"?> Advances in Burushaski linguistics Jan Henrik Holst <?page no="4"?> Bibliographische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliographie; detaillierte bibliographische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliographie; detailed bibliographic data are available on the internet at http: / / dnb.dnb.de. © 2014 · Narr Francke Attempto Verlag GmbH + Co. KG Dischingerweg 5 · D-72070 Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Gedruckt auf säurefreiem und alterungsbeständigem Werkdruckpapier. Internet: http: / / www.narr.de E-Mail: info@narr.de Printed in Germany ISSN 0564-7959 ISBN 978-3-8233-6908-0 <?page no="5"?> 5 Preface This book investigates the Burushaski language, spoken in Asia near the Karakorum mountains, which offers many challenges to linguistics. Synchronic and diachronic approaches are made use of in this work. The aim is to advance knowledge on Burushaski in a variety of ways. The book has five chapters. Following a brief introduction, typological studies, historical-comparative analyses of the dialect differences, investigations employing internal reconstruction as well as studies on the vocabulary and semantics are presented. There are many individuals whom I would like to thank. In the summer terms of 2010, 2011, 2012 and 2013 I taught Burushaski linguistics at the Asien-Afrika-Institut of the University of Hamburg. I would like to express my gratitude to Ludwig Paul, head of the Iranian department, who made this possible, and most emphatically to the enthusiastic participants of these lectures. Concerning chapter 2, I would like to use the opportunity to thank Gregory D. S. Anderson, Elena Bashir, Carsten Bettermann, Robert M. W. Dixon, Matthew Dryer, Maren Fittschen, Pavel Iosad, Carsten Peust, Kristine Risløv, Elmar Ternes and Bertil Tikkanen for correspondence on Burushaski and related matters. I especially remember my close friend Carsten Bettermann who has gone from us much too early. As to chapter 4, I would like to thank Gregory D. S. Anderson, Lyle Campbell, Roland Kießling and Elmar Ternes for helpful comments. Also the other chapters have benefited from discussions with many colleagues <?page no="6"?> Preface 6 and friends. Although some contacts were only brief, they were all of great help. Naturally, the responsibility for the content of this work is entirely mine. I hope to have achieved a clear presentation of what I intend to convey in this book: why I support certain views on Burushaski, why I reject others or am skeptical about them, and how I arrive at my own conclusions about this language. It would be a pleasant experience if the field continues to flourish. Also research on other language isolates, e.g. Basque or Ainu, may benefit from some of the ideas found in this book. Especially the procedure to attempt to gather insights in the fields of typology, dialect comparison and internal reconstruction pursued here in chapters 2, 3 and 4 respectively may prove fruitful. Vantaa, June 2014 Jan Henrik Holst <?page no="7"?> 7 Contents Abbreviations and signs 9 1. Introduction 11 1.1. The language and its speakers 11 1.2. The dialects 13 1.3. The status as language isolate 14 1.4. Language contact 18 1.5. On the content of the monograph 20 2. Investigations into the typology of Burushaski 22 2.1. Introduction 22 2.2. Issues in morphophonology 22 2.2.1. A nasal alternation 22 2.2.2. The mutations 26 2.3. Issues in the morphology of the verb 33 2.3.1. The category of polarity 33 2.3.2. The negative prefix and the interrogative suffix 36 2.3.3. Verb forms without a stem 38 2.3.4. Person marking I: On the shape of the affixes 40 2.3.5. Person marking II: On the use of the affixes 43 2.4. Some afterthoughts 48 3. The dialects and the comparative method 51 3.1. Introduction 51 3.2. Consonants 53 3.2.1. Affricates 53 3.2.2. Clusters of nasal and plosive 55 3.2.3. Other topics 57 3.3. Vowels 62 3.3.1. The back vowels o and u 62 3.3.2. The front vowels e and i 70 3.3.3. Other topics 81 3.4. The “grass group” 85 3.5. Suprasegmentals 87 3.6. Vocabulary 89 3.7. Morphology 99 3.7.1. Introduction 99 3.7.2. Singulars from former plurals 100 <?page no="8"?> Contents 8 3.7.3. Some observations on possessor marking 102 3.8. Other levels of language structure 106 4. Internal reconstruction in Burushaski 107 4.1. Introduction 107 4.1.1. Internal reconstruction 107 4.1.2. Internal reconstruction in Burushaski so far 108 4.1.3. Objectives of the chapter 110 4.2. Phonology 111 4.2.1. An ancient vowel alternation 111 4.2.2. The consonant alternations n / y and s / S 115 4.2.3. Alternation of the retroflex fricative and h 117 4.2.4. Voicing after nasals and after l 120 4.3. Morphology 122 4.3.1. The numerals from ‘one’ to ‘ten’ 122 4.3.2. The remnants of the numeral ‘two’ in body parts 126 4.3.3. On ‘to be’ 130 4.4. Syntax 133 4.4.1. The rise of the system of local cases 133 4.4.2. The rise of an adjective type 135 4.5. On stress in Burushaski 136 4.5.1. Introduction 136 4.5.2. Main claims 137 4.5.3. Stress in morphological forms 140 4.5.4. Prehistory of the consonant group lt 143 4.5.5. Summary of the events 145 4.6. Final reflections 146 5. On the vocabulary of Burushaski 149 5.1. Introduction 149 5.2. Polysemy and homophony 151 5.3. The suffix -c 155 5.4. On some kinship terms 158 References 161 Indices 171 Subject index 171 Language index 173 Word index 176 <?page no="9"?> 9 Abbreviations and signs Abbreviations: A transitive subject AM approximant mutation C consonant f, f. feminine gen. genitive Hz. Hunza intr. intransitive m, m., M masculine NEG negative Ng. Nager nom. nominative NP noun phrase O (transitive) object pers. person pl. plural PM provective mutation PP postpositional phrase PRES present Proto-Bur. Proto-Burushaski PTCPL participle S intransitive subject sg., SG singular syll. syllable(s) TAM tense / aspect / mood tr. transitive UM suffix -um V vowel x gender x y gender y Ys. Yasin Signs: * reconstructed < is from > goes to <?page no="10"?> Abbreviations and signs 10 [ ] phonetic transcription ∅ loss (as a sound law result); zero shows position of a sound in a word (initial, medial, final); separates morphemes <?page no="11"?> 11 1. Introduction 1.1. The language and its speakers Burushaski is a language isolate spoken near the Karakorum mountains, in far northern Pakistan, by about 100,000 people. Due to recent migrations there are a few speakers also in Srinagar (a city in India) and elsewhere. Not only the language is unique, but also the people speaking it has a cultural heritage which does not exist elsewhere in exactly this form. Burushaski is usually an unwritten language (though there are some speakers today who are able to write it). This means that there are no older documents; the structure and the vocabulary of the language had to be explored by fieldwork. Such work started in the second half of the nineteenth century when colonization by the British began, see e.g. Biddulph (1880) and Leitner (1893). These sources, however, use English-based orthographies which are often inconsequent and unreliable, and the grammatical structures were not always understood: the language was thought to have two genders, for instance, while actually there are four. Only rarely are these old sources useful for historical linguistics, namely when slight changes in details occurred in the language between that time and now. Lorimer (1935a, 1935b, 1938) arrived at many new insights on Burushaski. Nowadays, however, this work is outdated, too, due to newer research by Hermann Berger, Étienne Tiffou and other scholars. Today the most important languages of publication on Burushaski are French, German and English, and to a lesser extent also Russian. It is possible to acquire a good working knowledge of the language with Berger (1974), Tiffou / Pesot (1989), Tiffou (1995a), Anderson (1997), Berger (1998), Tiffou (1999), Anderson (2007) and other publications. Research reports include Bashir (2000) and Tiffou (2004a). Publication of fieldwork results and traditional narrations, especially in the last decades and years, opens up various possibilities for research activities, and fieldwork itself may continue as well. The state of synchronic research on Burushaski is good, but some problems remain; on some points one does not know whether one should trust the data. Diachronic research has yielded fewer reliable results so far, but some progress has been made in this area, too; many claims have been put forward, but one has to pick out those which are methodologically sound. <?page no="12"?> 1. Introduction 12 The speakers of the language call themselves BurúSo (or BurúSu in the Yasin dialect). This is a plural form, containing the morpheme -o (or -u) which is a frequent plural suffix. In English, the designations Burusho and Burushos can be found, the latter one carrying the English plural suffix -s in addition. The singular of BurúSo is BurúSin. The name of the language, BurúSaski, is a derivative. As can be seen, in Burushaski itself the language name is stressed on the second syllable (the accent designates stress). In other languages, such as English or German, the stress may be placed there as well, or elsewhere; the German name of the language is Buruschaski or Burushaski; the French name bourouchaski. There are different opinions on whether Burushaski is an endangered language or not. Berger (1992: 25) is afraid that the language is going to disappear. In contrast, Anderson (2007: 1022) writes about the Burusho: “their language remains vital, spoken by all generations”. Cf. also Anderson (2009: 175): “In all communities where Burushaski is spoken, the language remains vital, with many women and children still monolingual speakers.” In fact, Burushaski is learned by children, and the number of speakers is rising. Moreover, the speakers of Burushaski are equipped with a positive attitude and healthy pride concerning their language. In view of these facts, in my opinion the language need not be considered endangered at present, and this is also the conclusion of Backstrom (1992: 54). The discussion will probably continue. Burushaski is an SOV language, and other basic constituent orders are AN, GN, NlN (i.e. numeral + noun), postpositions. The vowel system is i e a o u. The consonant system is large and contains e.g. aspirated and retroflex consonants. Stem shapes vary in their makeup; there are stems which consist of only one consonant, e.g. -s ‘heart’, -t- ‘to do’. The morphology is somewhere between agglutinating and fusional, and both prefixing and suffixing occur; most of the prefixing shows up in the rich verb inflection. Some verb forms are analytic and use the auxiliary ‘to be’. There is ergativity in Burushaski; with the noun this is simple: there is an absolutive case and an ergative case in the large case system, instead of a nominative case and an accusative case which are known from accusative languages. With the verb, in contrast, matters concerning ergativity and related topics require much more room for discussion (see 2.3.5.). There are four genders: masculine, feminine, gender x and gender y, abbreviated m, f, x, y in this book. Numerals have, in addition to gender differentiation, so-called z forms used for counting (hence five forms, but often there is <?page no="13"?> 1. Introduction 13 syncretism). Nouns are subdivided into two types: alienable and inalienable. While the former can express the possessor only with a separate word in the genitive, the latter are obligatorily marked for possessor. The language uses personal prefixes, e.g. a- ‘my’, and inalienable nouns are cited with a hyphen at the beginning, e.g. -rén ‘hand’. There are four sets with varieties of the personal prefixes, varying slightly in phonetic shape. 1 The brief typological profile just outlined shows that Burushaski certainly has a remarkable structure considering the area where it is spoken. This has to do with its status as a language isolate which will be treated in more detail in 1.3. The surrounding languages often do not share the grammatical features, although they frequently do share vocabulary items, and also phonetic traits recur, due to mutual influence. Burushaski would possibly make a less extraordinary impression if it was spoken, say, in the Caucasus or somewhere in the Americas. 1.2. The dialects The Burushaski speech area is divided into a western area with one dialect and an eastern area with two further dialects quite closely related to each other. The dialects are, with their usual abbreviations: Yasin Hunza Nager Ys. Hz. Ng. Hunza and Nager are spoken on both sides of the Hunza River, and Yasin is situated in the Yasin Valley, further to the west. The speech of Srinagar is Nager Burushaski which has very recently been subjected to strong language contact (Munshi 2010). Nager is also found with the spelling Nagar (and also Nagir, Nagyr in old sources). An outdated alternative term for Yasin is Werchikwar, a designation stemming from the neighbouring 1 Burushaski scholars have a system of diacritics indicating which set is used on an inalienable noun. This system will be omitted in this work, since it is not relevant for any point that will be made. If an inalienable noun does not carry a stress mark, e.g. -skil ‘face’, this means that stress is on the personal prefixes. <?page no="14"?> 1. Introduction 14 Dardic language Khowar; Tiffou (1995b: 159f.) argues that this term should be given up. Only rarely does one hear of differences within one of the three dialects mentioned. In this connection the word lists by Backstrom / Radloff (1992: 243-260), offering 210 words, are occasionally interesting since they sometimes show a deviation in the Yasin dialect of the village Thui from usual Yasin. According to a vernacular tradition the Yasin dialect came into being when a group of speakers migrated to the west in the 16th century AD (Berger 2008: 4f.). There does not seem to be any proof that this is correct, but the linguistic differences that can be observed can indeed be reconciled with about 500 years of separation. Hence, this is not an unrealistic theory. This book takes into account the whole language with all its dialect variation. Of course I made sure that no data from any dialect contradicts the conclusions drawn in my investigations. In particular this applies to the claims for internal reconstruction in chapter 4. In many regards the dialects are so similar anyway that the differences do not affect any conclusions. Data with no dialect specification is from Yasin (which does not exclude that the Hunza and Nager data may be the same, given the closeness of all dialects). 2 The Yasin data used in this book can for the most part be found in Berger (1974), Tiffou / Pesot (1989) and Tiffou (1999); the Hunza data can mostly be found in Berger (1998I, 1998II, 1998III). The transcription (of all dialects) is as Tiffou’s, but instead of háceks, acutes are used for alveolo-palatals, thus S C Ch. 1.3. The status as language isolate Most languages of the world can be grouped into families, but there are some with no known relatives for which Mary Haas coined the term language isolates. Haas (1965: 77) defines these languages as follows: “language isolates, that is, single languages with no demonstrable close relationship to any other single language or family of languages.” The status as language isolate need not remain unchanged for all time; work on language relationships may get ahead and provide new insights which put 2 Note that Berger (2008) proceeds in a different way: in his monograph data not specified for dialect is from Hunza. Linguists who are no specialists in Burushaski sometimes quote data without making clear from which dialect it is; frequently this does no harm, while occasionally the information would be useful. <?page no="15"?> 1. Introduction 15 an end to this status. However, it is often quite hard to find relatives for a language isolate, and researchers disagree on many issues. Burushaski is regarded as a language isolate by most scholars. For statements to this effect see e.g. Berger (1985: 36), Tiffou (1993: 1), Trask (1996: 191), Lyovin (1997: 125f.), Tikkanen (2001: 479, n. 1), Munshi (2010: 36, n. 6) and a very large number of other sources. There have been many attempts to find relatives for Burushaski. Morgenstierne (1935: XIIf.) reports that the earliest authors who thought about the question tried links with Munda, Dravidian, Andamanese, Caucasian and other languages. Bleichsteiner (1930) advocates relationship with “Caucasian” but we know today that there are actually three independent language families in the Caucasus which show hardly any signs of relationship with each other: Abkhazo-Adyghean, Nakh-Daghestanian and Kartvelian, see Deeters (1963), Klimov (1994), Hewitt (2004) and others. 3 Moreover, Bleichsteiner (1930) quotes words from various other languages (e.g. Basque, Elamite) as well if they look similar for him; he has hardly any other method than pointing out similarities. For criticism of Bleichsteiner see Morgenstierne (1935: XIII-XIX). The review by Bouda (1931) is less critical, but exactly that may be regarded as a shortcoming. The articles Berger (1956) and Berger (1959) contain some arguments which are intended to show that Burushaski is related to Basque. 4 However, later in his life Berger realized himself that these attempts were mistaken. Grolier (1995: 112) reports this and refers to personal communication with Berger. Moreover, Berger (2004: 13) also considers his own attempts as a young man a failure. For some further remarks on the topic see Berger (2008: 1). Starostin posited a macrofamily called Sino-Caucasian; his original article is in Russian, but a translation appeared in Shevoroshkin (1991): Starostin (1991). The term “Caucasian” refers only to the two northern 3 The three families are often called West, East and South Caucasian respectively (or Northwest, Northeast and South Caucasian). To my mind this usage is not recommendable since it makes the erroneous impression that these were branches of one family which is not the case. Designations with “West” etc. should be reserved for branches of a family or a subgroup. 4 Campbell / Poser (2008: 406) may give the impression that Berger (1956) advocates a relationship between Burushaski and Indo-European but this does not apply. Possibly they misinterpreted a statement by Bashir (2000: 1) and did not have access to Berger’s article itself. <?page no="16"?> 1. Introduction 16 families in the Caucasus here, i.e. not to Kartvelian. The idea was taken up by some other researchers. In several articles in Shevoroshkin (1991), e.g. Bengtson (1991a, 1991b), Bengtson adds Basque and also Burushaski to this macrofamily. Other publications by Bengtson followed, but Tiffou (1995b) points out flaws in them. Bengtson has since then tried to improve his case but not gained much support. Tuite (1996), drawing on ethnological evidence especially myths, folk tales and linguistic data, thinks that a relationship with Nakh- Daghestanian is possible but he does not exclude language contact either. He only offers three insignificant typological parallels, e.g. ergativity, and provides a few etymologies from those authors that advocate Sino-Caucasian. Tuite (1998) is a similar treatment in English. Wolfgang Schulze, an expert in Nakh-Daghestanian, kindly informed me that he is skeptical of a genetic link to Burushaski (personal communication). Some typological similarities with Yenisseian a small language family in Siberia of which only one language, Ket, is still spoken were pointed out by Toporov (1971). Later a relationship theory grew out of this approach but no-one ever adduced more than extremely little evidence. Grune (1998) tends to support this view. Furthermore, Driem (2001: 1198-1201) compares six affixes from the verbal system, e.g. Burushaski aand Yenisseian ba-, which are not even very similar. He also misunderstands and confuses the functions of some affixes that he compares. Casule (1998) claims that Burushaski is an Indo-European language, and there are further publications by Casule which are intended to back this view up. The Indo-European family provides him with a large number of languages to cull data from, and so for example multán ‘blood’ is allegedly cognate with the Germanic word: English blood, German Blut etc. Although this word is taken from Germanic, Casule sees a special affinity to some Indo-European languages from the Balkans. (The author himself stems from Macedonia.) However, Casule ignores the fruitful research on the subgroup that is called Balkan Indo-European today, see Klingenschmitt (1994) and his followers. Casule (2009) tries to declare the Burushaski numerals as having Indo-European origins, but these numerals are clearly entirely non-Indo-European, as anyone can see (for the data see 4.3.1.), and only by extremely far-fetched manoeuvres does Casule arrive at his claims. For example, thalé ‘seven’ is said to possibly consist of Avestan hapta ‘seven’ plus a particle used for addressing men le (Casule 2009: 174). Neither is it clear why the numeral should be such a <?page no="17"?> 1. Introduction 17 compound, nor is the segmentation of the Burushaski word possible in this way: in reality, -é is a suffix which becomes apparent when comparing other numerals, or, alternatively, other inflectional forms such as thaló (form for the four genders). Bengtson is correct in objecting to Casule’s theory, and see also Tiffou (2004a: 97f.). Unfortunately Casule does not have sufficient knowledge of several areas: neither of Indo-European linguistics, nor of Burushaski, nor of the methods of historical linguistics. The Journal of Indo-European Studies (JIES) recently devoted an issue to the topic (Vol. 40, 2012), although no Burushaski expert is among the authors. Finally, there are some other relationship claims about Burushaski, mostly with lexical comparisons, e.g. Bouda (1950). Joseph Greenberg, who classified many languages (e.g. Greenberg 1963), to my knowledge never ventured a statement on the classification of Burushaski. Nowadays there is a heated debate on the internet on the relationship hypotheses: on many pages discussants write both on the content of the claims and on matters of style when dismissing the work of someone whose views one does not share. It must be stated clearly that all attempts so far are not convincing. A basic problem is that they suffer from serious weaknesses in methodology. Many deal only with the vocabulary and not with the grammar, which would be desirable as well. The lexical material, then, is not treated as it should. Burushaski words are often quoted incorrectly. The fundamental insight of the Neogrammarians that there is regularity of sound change is hardly accounted for. Words are subjected to arbitrary segmentations, loanwords which entered Burushaski are treated as if they were old in the language, and unrealistic semantic deviations in the cognate sets are tolerated. In short, many explanations and etymologies are far-fetched and it is highly unlikely that they match the prehistoric facts. At the same time, important research on Burushaski is frequently ignored; many authors indeed know very little about the language (especially viewed in relation to the difficult task they set themselves). This also has to do with the fact that one should be able to study all revelant sources, also those in French and in German, and not only those in English; Burushaski linguistics is a multilingual enterprise. Hardly anyone of those who presented relationship claims attempted to advance research on Burushaski itself although there would have been many possibilities to do so: fieldwork, typological and areal <?page no="18"?> 1. Introduction 18 studies, dialect comparison, internal reconstruction etc. (Hermann Berger is an exception but he abandoned his hypothesis.) I am currently working on my own theory about the external relationship of Burushaski which is unlike any previous theory. My work is radically different from all previous attempts both in its content and in the methods; it meets the standards of historical linguistics which exist with full justification. What can be reported here is that the material I have at my disposal already now is much more compelling than that put forward so far for the other theories. The reason is, in my opinion, that I have found the correct combination of languages, which then enables the researcher to accomplish the task although still such work is hard. I intend to present my results in another monograph when time is ripe for that. The fact that I am determined to come up with such a work indicates that I will have something to offer which is definitely worth being studied by the linguistic community. Finally, it must be reported that the present monograph does not discuss the external relationships of Burushaski (except for what has been said in this section), nor has the work been influenced to a great extent by any such theories. The focus is exclusively on Burushaski itself; the monograph approaches this language as an object of research in its own right. In fact, Burushaski has a lot to offer for any linguist who is willing to delve into it. 1.4. Language contact Language contact is probably important when considering almost any language on earth and trying to develop a deeper understanding of it. Burushaski is no exception is this respect. Its neighbouring languages are mostly Indo-Aryan, to be more precise: Dardic, or Iranian, e.g. Wakhi. From the Dardic languages especially Khowar and Shina must be mentioned, the former being in contact with Yasin, the latter with Hunza / Nager. For a useful survey of the Dardic and Nuristani languages see Edelman (1983: 35-58). There is another nearby idiom, Balti, which is a variety of Tibetan. Information on language contact and the sociolinguistic situation in northern Pakistan can be gathered from Backstrom / Radloff (1992). Many inhabitants of the area are bilingual or have a command of even more than two languages. Burushaski has taken up numerous loanwords from languages of the region, but it has also supplied other <?page no="19"?> 1. Introduction 19 languages with words of its own. Most remarkably, the Dardic language Shina has taken up a large amount of loanwords from Burushaski. One rightly assumes that the speech area of Burushaski was once larger; speakers then went over to Indo-European or other languages, but substratum effects occurred. Sometimes when a neighbouring language shares a structural feature with Burushaski, this is addressed in this book. Turkic languages are situated a bit further away than those addressed so far. There must have been contact as well, however, since the Burushaski terminology which deals with horses is of Turkic origin, plus some other words; see also Rybatzki (2010). Even the word for ‘horse’ itself is from Turkic. It is Ys. ha6ór / Hz. ha6úr and can be compared to the Turkic word for ‘stallion’, which in Turkish is ayg I r. Berger (1998III: 185) and Rybatzki (2010: 158) express doubts about this etymology, but Berger (2008: 35) does not do so any longer. A brief diagram: Hz. h a 6 ú r ‘horse’ Turkish a y g I r ‘stallion’ Hunza is chosen here because with the vowel correspondence Ys. o / Hz. u it is u which represents the older state, see 3.3.1. Unlike Turkic, Burushaski does not have unrounded back vowels; therefore Turkic I or ï, as some scholars would write it was substituted by u. The y of Turkish is an innovation for an earlier dental, according to the Turkologists possibly *ð; it was such a sound which occurred in that Turkic language which gave the word to Burushaski, and the consonant group was simplified. Berger (2008: 35) gives a few examples of hwhich has been added in initial position. 5 Loanword studies are important because they make the stock of words shrink which is open to comparisons to the outside. Latin equus is one of the well-known indicators that this language is Indo-European, being related to Sanskrit áSvaH ‘horse’ etc.; Burushaski ha6úr, in contrast, does not have the same status. On the other hand, one must not invent too many etymologies with loanwords this mistake has been made as well. Some of the claims by Parkin (1987), for instance, may be correct, but others are not. Berger (2008) interprets too many Burushaski words as 5 In this book usually Turkish (of Turkey) is used in order to illustrate an etymology. For geographic and phonetic reasons the actual donor languages were different Turkic languages. For the state of the art which languages these may have been see Rybatzki (2010: 174f.). <?page no="20"?> 1. Introduction 20 loanwords from Sanskrit and other Indo-Aryan sources, often with farreaching phonetic and semantic changes. Urdu, now the official language of Pakistan, contributed a large number of words to Burushaski and is continuing to do so. Morin / Dagenais (1977) show how Urdu words are transformed when entering Burushaski; they are adapted to its phonetics and phonotactics. The Urdu words are in turn often loanwords; they are frequently of Persian or Arabic descent. For example: Ys. ketáp ‘book’ < Urdu kit á b < Persian < Arabic Ys. and Hz. úmur ‘age’ < Urdu umr < Persian < Arabic Someone with linguistic experience with the Islamic world will be able to spot many such words and often understand them correctly (although semantic change sometimes occurred). Nowadays there are also some loanwords from English via Urdu. The inherited core of the Burushaski vocabulary is quite different, naturally; these words do not “ring a bell” and are a new experience even for polyglots who have studied many languages of the world. The present monograph, whenever dealing with vocabulary, will not often be concerned with the large number of loanwords in Burushaski. Instead, the focus will more often be on those words that seem to be ancient in the language since they are more useful for detecting old structures (phonetic, phonotactic, morphological, semantic structures etc.). 1.5. On the content of the monograph Chapter 2 deals with structures of the Burushaski language which had not received sufficient attention, or no attention at all, so far. Within this context, typological comparisons are drawn, and issues in terminology and grammar theory are addressed. Chapter 3 uses the comparative method of historical linguistics and applies it to the dialects. Sound correspondences are presented and sound laws are concluded. Also morphological and other comparisons are made. The treatment gives an idea of what Burushaski was like before it split into dialects. In this way we move a few centuries into the past. Chapter 4 investigates what can be revealed about the linguistic past of Burushaski by the means of internal reconstruction. The chapter deals with methodological points, topics in phonology, morphology, syntax, and stress as an issue pervading the whole language system. This <?page no="21"?> 1. Introduction 21 chapter constitutes a considerable expansion of earlier work on Burushaski historical linguistics. Moreover, the treatment is intended to have an impact on investigating language isolates in general: it will be argued that in order to find out more about these languages (and ultimately their genetic links), it is rewarding to first subject them to internal reconstruction. Chapter 5 is a collection of observations that can be made on the vocabulary of Burushaski. For example, there are internal connections among words which deserve being pointed out, but also other approaches are pursued. Although each chapter constitutes a unit in itself, there are also various links between the chapters. A unit introduced by two numbers, e.g. 1.1., is referred to as a section; a unit introduced by three numbers, e.g. 2.2.1., is called a subsection. <?page no="22"?> 22 2. Investigations into the typology of Burushaski 2.1. Introduction The Burushaski language is full of features which are of interest to the typologist. Some of these have received detailed treatment, e.g. passive formations by Morin / Tiffou (1988), verbal number by Tiffou / Patry (1995) and converbs by Tikkanen (1995). The present chapter focuses on some typological properties which have not been discussed extensively yet, or not at all yet. The chapter therefore consists of a series of investigations, each of them presenting a new point about Burushaski which had not been made in the scholarly literature so far. The synchronic facts are laid out, they are put into a typological context, and questions of terminology and grammar theory are discussed as well. Occasionally diachronic or areal remarks are added, if possible. It turned out that the issues that something new could be said about fall under two areas: morphophonology (or morphophonemics, or phonology, depending on one’s terminological preferences) and the morphology of the verb. Sections 2.2. and 2.3. deal with these areas respectively. Section 2.2. contains two investigations which are independent of one another. One deals with an alternation of final -n and medial -min Burushaski. The other one lays out that Burushaski has mutations, similar to the ones in Celtic and other languages, and evaluates all information that can be obtained on them. Section 2.3. consists of five investigations. The first two are related in that they both refer to negation expressed in the verb. The treatment in the middle deals with verb forms without a stem. The last two investigations are related to each other as they both examine person marking in the verb. Finally, section 2.4. addresses a few general issues that appear to be important in the study of Burushaski. 2.2. Issues in morphophonology 2.2.1. A nasal alternation The investigation will first consider two Burushaski nouns, ‘eye’ and ‘liver’. To my knowledge these two nouns have not been connected yet, and a certain nasal alternation which can be observed in them has not been drawn attention to. Both are inalienable nouns; hence they are cited <?page no="23"?> 2. Investigations into the typology of Burushaski 23 with hyphens at the beginnings (cf. 1.1.). Stress is not marked here since it is on the possessive prefixes. What is needed for the discussion are the forms of the absolutive singular and of the absolutive plural (the absolutive has no case suffix). It is necessary to quote all three dialects. The data is: 1. ‘eye’: Yasin Hunza Nager sg. -l-Ci -l-Cin -l-Cin pl. -l-Cim-u -l-Cum-u-c -l-Cim-u-c Sources: Yasin: Berger (1974: 161), Hunza and Nager: Berger (1998III: 264) 2. ‘liver’: Yasin Hunza Nager sg. -ken -kin -kin pl. -ken-i9 -kim-i9 -kin-i9 Sources: Yasin: Berger (1974: 157), Hunza and Nager: Berger (1998III: 245) By presenting the data this way, the analysis has implicitly already begun, as all segmentations, indicated by the hyphens within the forms, are mine. First of all the plural suffixes should be mentioned. The -u (also -o) is a typical plural suffix of the language which occurs in genders m, f and x, and the lexeme ‘eye’ belongs to gender x. The expansion with -c in Hunza / Nager occurs also with other nouns. The most frequent plural suffix for gender y is -i9, and this indeed occurs in ‘liver’, which belongs to gender y. Next, -lin the item ‘eye’ is actually the remnant of an old dual prefix which has long lost its meaning and has made its way into the singular as well (Klimov / Èdel'man 1972, and see 4.3.2.), but this fact is not relevant to any point that will be made here. We are left with the stems then. The Hunza forms for ‘eye’ exhibit a vowel alternation i / u. This alternation crops up in various places in Burushaski, cf. the inflection of this noun: ‘man’: Yasin Hunza Nager sg. hir hir hir pl. hur-í hir-í hir-íkanc Sources: Yasin: Berger (1974: 18, 152), Hunza and Nager: Berger (1998III: 200) <?page no="24"?> 2. Investigations into the typology of Burushaski 24 The vowel alternation is therefore not an unusual feature and does not require discussion here. Sometimes one dialect retains it, whereas the other ones have given it up, i.e. they generalized one vowel by analogy. In ‘eye’ only Hunza has kept the alternation, and in ‘man’ only Yasin has done so. For a deeper account see 4.2.1. It is now possible to assess the nasal question. Looking first at ‘liver’, it can be seen that all singular forms have stem-final -n. In the plural two dialects have -nas well, but Hunza has an unexpected -m-. There is no sound law or other cause which could have produced this -m-. If the -mis assumed to be old, however, the -nof the two other dialects can be easily explained as arisen by analogy with the singular. The assumed oldest state is then supported by the inflection of ‘eye’ which also has -n in the singular and -min the plural in two dialects; Yasin exhibits a loss of -n according to this interpretation. Consider now three further examples: Hz. bácin ‘thigh’, pl. bácim-i9 (data from Berger 1998III: 28), bárSun ‘wooden sieve’ (certain type of sieve), pl. bárSum-u-c (data from Berger 1998III: 41) and phárcin ‘cap’, pl. phárcim-u-c (data from Berger 1998III: 324), Ys. phárcen, phárce ‘cap’. The analysis reveals that Burushaski has a nasal alternation n / m: the former appears in final position, the latter elsewhere. Secondary developments have sometimes blurred this state. Some of the items are core vocabulary items, however, and therefore the phenomenon seems to be ancient. It must now be discussed that there is an alternative analysis of ‘liver’ and ‘thigh’ by Lorimer (1935a: 39) and Berger (2008: 95) (they do not mention ‘eye’ or ‘wooden sieve’). According to their view there was a plural suffix *-mi9 in ‘liver’ and ‘thigh’, the m of which then made a stemfinal *n drop; i.e. there was a cluster simplification *-nm- > -m-. The idea of a cluster simplification is legitimate from a general typological perspective. Nevertheless, it is possible to adduce some considerations which render Lorimer’s and Berger’s theory less likely. The usual shape of the plural suffix of Burushaski gender y nouns is clearly -i9, as anyone agrees. There are indeed plural forms which end in the sound sequence -mi9, but they are by far not as frequent as those in -i9. Moreover, in some of these, although they are apparently not more than a handful, the -moccurs where the singular has -n-, cf. the examples above, and the basic idea may now be to identify the two nasals in a linguistic analysis. When this is done, forms such as Hz. -kim-i9 ‘livers’ simply contain the usual plural suffix. It is true that some nouns have <?page no="25"?> 2. Investigations into the typology of Burushaski 25 plurals in -mi9 but the singular does not have an n. In some instances there may once have been an n which has been lost. Others may be interpreted in the following way: by false segmentation the sound sequence -mi9 was reanalyzed as a plural suffix of the language and was then added to the nouns in question. The variant -mi9 is especially frequent after stems in vowels (Berger 1998I: 56). This fact can be interpreted in such a way that the m took on the behaviour of a “Bindekonsonant” here. Stems in vowels are a minority, stems in consonants being the rule. There are also variants such as Ys. cel ‘water’, pl. cel-í9 and cel-mí9 (Berger 1974: 136). Finally, consider the plural of den ‘year’ (all dialects), which is Ys. den-í9 (Berger 1974: 141), Hz. den-í9 / den-mí9 and Ng. del-mí9 (Berger 1998III: 118). This noun, having a genuine stem in n, does not reduce the consonant cluster when adding -mí9 (Nager has lm < *nm by dissimilation, Berger 2008: 57). To sum up, Lorimer’s and Berger’s analysis deserves an alternative next to it, namely the view proposed here. The analysis with the nasal alternation n / m proposed here has been made with other languages in mind. The most important of these is Finnish. Consider Finnish data such as the following, which constitute a parallel (mm / mp in the last item is a regular alternation, called “gradation”): ‘heart’ ‘key’ ‘warm’ nom. sg. sydän avain lämmin gen. sg. sydäme-n avaime-n lämpimä-n The Finnish alternation is well understood; it arose by a sound law *-m > -n in final position (Laanest 1982: 101). The consequence of this sound law is a phonotactic constraint which allows only -n and no -m in final position. Also Ancient Greek has this phonotactic constraint, and also in this language it is due to a sound law *-m > -n in final position. It has been argued recently that this is not a special Greek development but it is older, since the effect is shared by several related languages: Albanian, Armenian and some dead languages (Holst 2009: 78f.). Ancient Greek does not have paradigms with an alternation -n / -mas Finnish. They must have existed, however, and can be deduced by internal reconstruction e.g. in the declension of a word for ‘earth’ (Holst 2009: 78f.). Also the Digor dialect of Ossetic exhibits *-m > -n, as comparison with the Iron dialect shows (Abaev 1949: 378f., Bielmeier 1977: 44, <?page no="26"?> 2. Investigations into the typology of Burushaski 26 303, n. 217). Moreover, *-m > -n is reported from the Chunsàk dialect of Avar, a Nakh-Daghestanian language, by Dirr (1928: 163, n. 1). Finally, the Cushitic language Somali has the sound law *-m > -n as well, only -n of these two is tolerated by the phonotactics, and there is an alternation of final -n and medial -m-, e.g. in nin ‘man’, stem nim-, pl. nim-an. Returning to Burushaski, it is now possible to assume a sound law *-m > -n in final position as well, which accounts for the forms. It should be mentioned that, in contrast to Finnish, Ancient Greek and Somali, Burushaski phonotactics does not exclude final -m. However, this fact does not invalidate the hypothesis: new instances of -m may have arisen by secondary developments. Also Estonian, a close relative of Finnish, has developed a new final -m, and the same applies to Albanian and to Armenian (Holst 2009: 79). In these languages often the new -m arose through the dropping of final vowels, but it does not follow that this needs to be the case in Burushaski as well, other events being possible. 2.2.2. The mutations It is best to start this discussion with a simple example. Burushaski has a word baT ‘skin, leather’ and a word waT ‘bark (of tree)’ (both forms exist in all three dialects), moreover cf. Ys. -wáT ‘body, self’, Hz. and Ng. waT ‘body, person’. It makes sense to assume a connection between these words, and this is indeed done by Berger (1974: 185). The bark is the “skin of a tree”, cf. data such as Japanese ki-no-kawa ‘bark’: ki ‘tree’, no postponed genitive particle, kawa ‘skin’. Also a word for ‘body’ can be imagined to have meant ‘skin’ before, and according to Berger (1998I: 45) -wáT still means ‘skin’ in the Nager dialect (alongside baT). There is obviously a consonant alternation in initial position. Such a phenomenon is called a mutation in several philologies; it is most prominent in Celtic linguistics. There have recently been two independent attempts to give listings of where in the world such initial consonant alternations occur: Holst (2008: 40-46) and Iosad (2010: 109-116) (Iosad’s article is based on an older, unpublished, thesis in Russian). To the best of my knowledge the term mutations has not been used yet in dealing with Burushaski, except for a brief mention by Holst (2008: 44). <?page no="27"?> 2. Investigations into the typology of Burushaski 27 It makes sense to introduce this term to Burushaski studies, too. Mutations occur throughout this language. 6 In previous work on Burushaski there was not only no designation for these features, but there was no systematic attempt to deal with them either. No grammar or sketch of this language to date contains a section devoted especially to this topic. The longest passages I am aware of are the ones by Lorimer (1935a: 10f.) and by Tiffou (1999: 123f.); Tiffou (1999: 123) speaks of “un phénomène intéressant qui caractérise cette langue”. The most extensive modern work on Burushaski published is Berger (1998); it treats the Hunza dialect. Remarks on the mutations are scattered over various pages in volume I: pp. 24, 45, 106, 109, 113, 119, 127. Morgenstierne (1945: 73-82) has a slightly outdated section about morphophonemics which contains some information on the mutations as well. The aim in the following is to give a survey of the topic, to collect some material and to investigate some questions which arise. Languages with mutations differ in how salient this feature is in their structure. In some of them, such as the Atlantic language Fulfulde, the Celtic language Breton (and in fact all Insular Celtic languages), and the language isolate Nivkh, the mutations are ubiquitous. In other such languages, mutations are less frequent and often unproductive, and in some languages they make the impression of being in retreat; Wolof, a relative of Fulfulde, provides an example. There is of course no fixed boundary between these types. Burushaski can be located more towards the latter type. An attempt will now be made to classify the mutations of Burushaski; both systematic and isolated cases will be discussed. A. First of all, voiced plosives and voiceless plosives alternate, i.e. b / p, d / t, g / k and 6 / q. Additionally in one verb in the Hunza dialect two retroflex affricates, voiced and voiceless, alternate (Berger 1998I: 127). The symbol 6 calls for a comment. Descriptions of 6 vary according to author and dialect: it is actually often only a symbol for a voiced uvular plosive, IPA [G]. Uvular plosives tend to make a slightly fricative impression in their release phase. If this sound should turn out to have been shifted to a velar fricative in some dialects (to Yasin this may apply), this 6 Actually, Holst (2008: 33) makes a distinction between mutations and half-mutations, the latter term being applied when one of the two mutating consonants occurs only after other morphemes. In this sense, Burushaski mostly has half-mutations, but for the sake of simplicity only mutation is used here. <?page no="28"?> 2. Investigations into the typology of Burushaski 28 is a secondary development. Some examples for this mutation, all using the negative prefix a-, are (Yasin dialect): bi ‘it is (gender x)’ / a-pí ‘it is not (gender x)’ (Tiffou 1995a: 159) d-í-ya ‘he came’ / a-t-í-ya ‘he did not come’ (Tiffou 1995a: 159) gu-wál-a ‘you (sg.) fell’ / a-kú-wal-a ‘you (sg.) did not fall’ (Grune 1998) In the first example a stem is affected, in the second one the so-called d-prefix (a morpheme the meaning of which is synchronically opaque), and in the third one a personal prefix (of the 2nd person singular). This mutation is one of the two most important types of mutation in Burushaski. It therefore deserves a name, and the proposal is made here to call it provective mutation; this is a term in current use for Breton, which has the same alternation, though without the uvular pair (Ternes 1990: 10). This mutation also exists in Breton’s close relative Cornish (Thomas 1992: 368), but not in the other Insular Celtic languages. What the other Insular Celtic languages do have (and what Breton and Cornish have additionally) is p / b etc., i.e. the opposite process. Henceforth, “provective mutation” will be abbreviated PM. It is essential to distinguish between morphemes that trigger PM and morphemes that undergo PM. PM mostly shows up in the verb. Verb forms consist of several prefixes, the stem and several suffixes (morphemes may be lacking). Anderson / Eggert (2001: 235) lay out the template for finite verb forms and list four prefix positions: NEG (negative), D (d-prefix), PERSON and CAUS (causative). Additionally, in certain non-finite verb forms there is a so-called n-prefix in the same position as the d-prefix. The d-prefix and the n-prefix mostly appear with a following vowel, which varies. The prefixes that trigger PM are (any dialect): anegative prefix dVd-prefix nVn-prefix -scausative prefix The causative prefix stands apart because the phonotactic rules would not tolerate s + voiced plosive anyway. Note that none of the person prefixes triggers PM. The morphemes that undergo PM are (any dialect): <?page no="29"?> 2. Investigations into the typology of Burushaski 29 dVd-prefix (hence -tV-) gu-, goprefix of the 2nd person singular (hence -ku-, -ko-) (various) various stems No other personal prefix than the one mentioned is affected by PM because there is no other personal prefix that begins with a voiced plosive (for a listing of the personal prefixes see 2.3.4.). Examples for verb stems which undergo PM are (Yasin dialect): ba- -pá- ‘to be’ bi- -pí- ‘to be’ gucár- -kúcar- ‘to walk’ girmín- -kírmin- ‘to write’ 6as- -qás- ‘to laugh’ To quote whole forms: bá-i ‘he is’ a-pá-i ‘he is not’, gucár-a ‘I walked’ nu-kúcar ‘having walked’ (source of the last two forms: Tiffou 1999: 126). PM can even occur twice in a verb form, e.g. in a-tú-ku-man-a ‘you weren’t born’ (Berger 1974: 33, see also Anderson / Eggert 2001: 241). The structure of this form is negative prefix, d-prefix, prefix of the 2nd person singular, stem, one suffix. The d-prefix, having the shape -túhere, causes mutation of the following prefix, and it is mutated itself due to the preceding negative prefix. These facts make the occurrence of two instances of PM inevitable. B. In another mutation voiced plosives and approximants alternate. Several cases show b / w. The introductory example baT ‘skin’ / waT ‘bark’ is of this kind, and there is also e.g. Ys. baland -wál- ‘to fall’ (Tiffou / Pesot 1989: 133), in Hunza the data is identical (Berger 1998I: 113). For more examples, from Hunza, see Berger (1998I: 113), Berger (2008: 26). Furthermore, g / y occurs, e.g. in Hz. gál- ‘to break (intransitive)’ vs. -yál- ‘to break (transitive)’ (Berger 1998I: 119). For more examples from Hunza see Berger (1998I: 113). The term approximant mutation shall be coined here, henceforth abbreviated AM. There is also an alternation g / ∅ in at least two words: the alienable noun gus ‘woman’ and the inalienable noun -us ‘wife’ clearly belong together, as has repeatedly been stressed, and there is a verb Hz. gurgín- / -úrgin- ‘to grind’, see Berger (1998I: 113), Berger (2008: 25). It may be significant that in both examples the vowel u is involved. It is difficult to decide whether some occasional cases of b / y <?page no="30"?> 2. Investigations into the typology of Burushaski 30 and b / ∅ can be included in AM. From a phonetic point of view a mutation b / y would hardly make sense. Berger (1998I: 113, n. 49) wonders whether in some verbs a -ydeveloped in hiatus after the personal prefix iand the -ythen spread to other forms. There are no reports of something like AM concerning d yet. In a different context, Tikkanen (1999: 295) observes: “Dand especially nare conspicuously rare as initial segments in native lexemes.” The facts may be connected: possibly there are no examples of AM for d yet because dis less frequent than other initials. However, Hunza has an alienable noun da9 ‘sleep’ (Berger 1998III: 114) and an inalienable noun -yá9 ‘sleep, dream’ (Berger 1998III: 472), and I would like to propose that they are etymologically identical; they would then exhibit a mutation d / y. C. The existence of an alternation of voiceless and voiceless aspirated plosives, i.e. p / ph, t / th etc., is uncertain. The author who claims its existence is Berger. According to Berger (1998I: 127), Hz. pus-ím-i means ‘he tied it’ (object of gender y), while i-phús-im-i means ‘he tied him’ (object masculine). There are also two Yasin examples given by Berger (1974: 30). Some caution is necessary here, however, because the notation of aspiration is not always reliable in the available data. This can best be seen with the affricates: the Yasin word for ‘meat’, for instance, is given as Cap by Berger (1974: 138, 202) and as C hap by Tiffou / Pesot (1989: 135) (their C = C); see more in 3.2.1. With plosives the notation of aspiration is much more reliable, but nevertheless further research on the question whether such a mutation exists may be useful. D. There are some miscellaneous isolated further phenomena, some of which may be regarded as mutations. The following isolated example involves the negative prefix aas the PM; it is, however, different in phonetic nature: Hz. hén- ‘to know’ negated a-khén- ‘not to know’ (data from Berger 1998I: 106). An alternation h / kh can be observed here; the Yasin dialect has given it up and has introduced h also to the negative forms. When there is a special verb or form for ‘not to know’ in a language, it sometimes preserves precious archaisms. Latin nesci ó ‘I don’t know’ (cf. sci ó ‘I know’), for instance, still contains an old prefix newhich is the equivalent of the Czech and Latvian prefix ne- ‘not’, while the new negation in Latin is n ó n, which from a historical point of view consists of *ne plus an additional element. It is undoubtedly the frequent use of ‘not to know’ which helps <?page no="31"?> 2. Investigations into the typology of Burushaski 31 such instances to survive in language history. The alternation h / kh in the Burushaski example may therefore represent something old. The inalienable noun -skir ‘father-in-law’ bears a relationship to hir ‘man’, but researchers disagree on the particulars. Tiffou (1999: 124) expresses the view that k and h may show another instance of a mutation. In principle then the mutation could be h / kh, as in the previous example, but after s no aspirated plosives are tolerated by Burushaski phonotactics (possibly with some exceptions in the Nager dialect, Berger 1998III: 381), hence h / k. The idea that a mutation is involved cannot be excluded. For more on these nouns, however, see the discussion in 5.4. Voiced and voiceless aspirated plosives may alternate in at least one instance. Consider the following pronouns (Yasin dialect): gender ‘that’ ‘this’ m ne khené f mo khomó x se gusé y te guté The forms for ‘this’ contain an element gVor khVplus an element which indicates gender, and thereby a mutation g / kh becomes apparent. The Nager dialect has only khin all genders (Tiffou 1999: 207). This appears to be an innovation of this dialect, since both the very closely related Hunza dialect and the more distantly related Yasin dialect exhibit the alternation. The numerals cend-í, cind-í ‘five’ and bi-Sínd-e ‘six’ have been connected in such a way that the latter consists of a prefix (or an old stem used for compounding here) plus the former (Berger 2008: 79). If this view is correct, an isolated example of a mutation c / S can be seen here. Finally, cer ‘time’ (in the sense of German Mal, French fois, Finnish kerta; see Berger 1974: 137) and hen-zér ‘once’ (cf. hen ‘one’) provide an isolated example of a mutation c / z (see also 4.2.4.). E. A phenomenon which stands somewhat apart is an alternation of m and b in Burushaski. Some interrogative pronouns are (Yasin): men ‘who’, but bo (occasionally be) ‘what’, bóta, bése, bésa ‘why’, báSa ‘when’. It is more typical, however, to find a difference m / b when comparing the dialects (Berger 2008: 28f.), although this is not a systematic correspondence. See also Anderson (1997: 1036). <?page no="32"?> 2. Investigations into the typology of Burushaski 32 The most important mutations of Burushaski are PM, treated under A., and AM, treated under B. They have in common that they both have the voiced plosives as their departure. It must be added that PM and AM can occur in combination in a paradigm; consider the following data: Hz. gán-imi ‘he took it’, a-yán-imi ‘he took me’, nu-kán ‘having taken it’ (Berger 1998I: 127, also mentioned by Holst 2008: 44). These forms exhibit g / y / k, which can be analyzed as the AM g / y and the PM g / k. In African linguistics the consonants which alternate in a position are called a series (Ternes 1990: 5, Holst 2008: 32). Using this term, g / y / k is a series in Burushaski. Diachrony can be addressed only briefly. Languages and language groups with mutations vary in how much can be said about the origin of this feature. For the Insular Celtic languages it has been possible to work out in detail how the mutations arose, which is due to the fact that Celtic is a branch of a large and well-known family, Indo-European. For the Atlantic languages in Africa the rise of the mutation system has been studied by Holst (2008). Burushaski is certainly in a less fortunate situation than some other languages. A distinction can be made, however, between laying out the exact ways of how the mutations arose (which is impossible at present), on the one hand, and the less ambitious goal, on the other hand, to determine which consonants are oldest in certain mutations. A statement on PM and AM is ventured here: it can be argued that in both of them the voiced plosives are oldest. This assumption is based on two observations. Firstly, the phonetic nature of the consonants involved points into this direction. Taking the series g / y / k as an example, g can be argued to be oldest because it could shift both to y on the one hand (in AM) and to k on the other hand (in PM), while a starting-point y or k would be harder to justify. Secondly, the voiced plosives occur in the forms with no prefix, and it is probably the prefixes which caused the mutations. The Burushaski mutations can be put into an areal context. Geographically close Tibetan has mutations as well (Beyer 1992: 112), and other Sino-Tibetan languages have remnants of them (Holst 2008: 44). Tibetan has voiceless, voiceless aspirated and voiced plosives, like Burushaski. Beyer (1992: 112) points out that there are pairs of intransitive and transitive verbs which are distinguished only by a mutation. He first adduces five examples with an alternation of voiceless aspirated plosives and voiced plosives: <?page no="33"?> 2. Investigations into the typology of Burushaski 33 khod ‘be placed, be put’ god ‘place, put’ t S hun ‘be tamed, be subdued’ d Z un ‘tame, subdue’ thon ‘come out, emerge’ don ‘throw out, eject’ tshugs ‘take root, become firm’ dzugs ‘plant, establish’ tshud ‘be put, enter’ dzud ‘put, lead’ Beyer then goes on with four examples with an alternation of voiced plosives and voiceless (unaspirated) plosives: gril ‘be twisted, be wrapped around’ kril ‘embrace, clasp round’ du ‘come together, assemble’ tu ‘gather, collect’ dor ‘be scattered, be dispersed’ tor ‘scatter, cast away’ bud ‘go away, disappear’ pud ‘transfer, put apart’ Although Beyer only speaks of intransitive and transitive verbs, a semantic analysis makes it likely that one can be more exact and speak of basic verbs and causatives. For don, for instance, one may think of a translation ‘to make come out, to make emerge’ which then reveals this verb as the causative of thon ‘to come out, to emerge’. This works in a more or less satisfying way with each pair. Occasionally one must allow for historically younger semantic change, of course, which could have set in at any time after the causative had been formed; this is a simple matter of relative chronology. Burushaski has mutations in causatives with the prefix -s-, as well as in other instances without a prefix, cf. ‘to break’ above under B. It follows that the two neighbouring languages Burushaski and Tibetan not only share the fact that they have mutations but also that they employ them in causative formation. 2.3. Issues in the morphology of the verb 2.3.1. The category of polarity From Indo-European languages such as Latin one is used to the categories of person, number, tense (and aspect), mood and voice in the verb. Whenever these categories occur in non-Indo-European languages, the grammatical terms for them already exist. There will be no challenge either when gender is marked in the verb, as e.g. in Arabic or Cree: Indo- European languages usually do not mark gender in the (finite) verb, but the term exists already because it had to be invented for the nominal system. <?page no="34"?> 2. Investigations into the typology of Burushaski 34 However, additional categories occur in the verb morphologies of the languages of the world. Sometimes one then does not have terms at one’s disposal. These are two verb forms from Turkish (stress added, kindly confirmed to me by native speaker Maren Fittschen): bil-íyor-um ‘I know’ know-PRES-1SG bíl-m-iyor-um ‘I don’t know’ know-NEG-PRES-1SG In Turkish there is a whole set of negative forms alongside the affirmative forms. This implies that verbs inflect not only for person, number, tense etc., but for another category. A general designation in linguistics for this category, however, is lacking, doubtlessly due to the Standard Average European heritage which does not know it. A term will certainly be useful, as this category is by no means rare on the earth; it also occurs e.g. in Japanese, Inuit and many more languages. The term polarity has now been used by Dravidianists (Steever 1998: 23), and by other philologies as well, e.g. African linguistics (Dimmendaal 2000: 172, 186). This term can also be introduced to Burushaski studies. As will be shown, Burushaski is indeed a language with polarity. The negative morpheme ais the first in any verb form, if it is present. The question may be asked why it could not be analyzed as an independent word ‘not’ then; the allomorph aywhich occurs before vowels (except i) could then be treated as a sandhi form. However, there are several arguments for a category of its own. Firstly, in many instances the subsequent morpheme undergoes a mutation. Usually this is the provective mutation, cf. ba ‘I am’, a-pá ‘I am not’; for further examples see 2.2.2. Another argument for a category stems from the fact that stress is usually different in the negative form and the affirmative form; cf. the following pairs of forms: gu-wál-a ‘you (sg.) fell’ / a-kú-wal-a ‘you (sg.) did not fall’ mu-wál-u ‘she fell’ / a-mú-wal-u ‘she did not fall’ Moreover, the verb ‘to see’ exhibits an irregularity; its stem is Ys. -yéc- (Hz. -yeéc-), but -ícafter the negative morpheme a- (Berger 1974: 187). The arguments discussed so far apply to all dialects. The dialects then add further evidence of their own. <?page no="35"?> 2. Investigations into the typology of Burushaski 35 In Yasin, an unstressed initial *igot lost, as e.g. in ne ‘he’, cognate with Hz. iné ‘he, she’ (see 3.3.3.). Burushaski verb forms use the prefix ifor the 3rd person of genders m, x and y. In Yasin this prefix is no longer present in the affirmative forms if it was unstressed, but it does still occur in the negative form because it was not in initial position in them. The result are pairs such as: wál-i ‘he fell’ / a-í-wal-i ‘he did not fall’ The first form cited is from earlier *i-wál-i. Moreover, consider the irregular relationship of Yasin lán- ‘to be able’ and aúl- ‘not to be able’ (data from Berger 1974: 35, 130, 161). Berger (2008: 30) reports that the Hunza dialect does not have this verb, and Nager uses ulánand ilán-. Berger (2008: 126) points out that -anis a suffix in several verbs and that it occurs also in this verb. It follows that the Yasin forms quoted can be segmented l-ánand a-úl-, and their roots belong together. The decisive point for the present discussion is the irregularity involved. From Hunza, the irregular hén- ‘to know’, a-khén- ‘not to know’ was already discussed in 2.2.2. Hunza underwent the monophthongizations *au > oo and *ai > ee, cf. Ys. 6aún - Hz. 6oón ‘melon’, which is a loanword from a Turkic language (Berger 2008: 22, Rybatzki 2010: 157f.), cf. Turkish kavun ‘melon’ (more on this in 3.3.3.). The monophthongizations affected certain Hunza negative forms as well because the morpheme boundary between the negative aand the following -uor -iof the next morpheme did not prevent the sound change (Berger 1998I: 86, 106f., 111, 113). Another complication is that intervocalic *h, which is still present in Yasin, was dropped in Hunza / Nager (see 3.2.3.), and due to this a positive form may still have h, while the negative form has not. Consider now the following Hunza forms (Berger 1998I: 107): hurúT- ‘to sit’ / oóruT- ‘not to sit’ huljá- ‘to rise’ / oólja- ‘not to rise’ It can be seen that in the negative forms the two effects mentioned occur combined. To use reconstructions: oóruT- < *a-húruT-, oólja- < *a-húlja-. Finally, the forms of gender y of the 3rd person of ‘to be’ are entirely irregular, in any dialect. In the Yasin dialect they are: affirmative negative sg. duá ‘it is’ a-pí ‘it is not’ pl. bicá ‘they are’ a-pí ‘they are not’ <?page no="36"?> 2. Investigations into the typology of Burushaski 36 It is hardly surprising that such an irregularity appears in the verb ‘to be’. There are typological parallels. The following three forms, all from Indo- European languages and all without any gender marking, mean ‘is not’ and are irregular: Latvian nav (cf. affirmative ir), Czech není (cf. affirmative je), Persian nîst (cf. affirmative ast). To sum up, there are many arguments which show the necessity to posit a category of polarity in a synchronic analysis of Burushaski. In order to express this category, by far not only simple affixation is made use of, but a variety of other means is used as well. It will therefore not suffice to treat negative forms as something that somehow exists in addition. It will only do to treat polarity on a par with other categories (and the only reason not to do so would be to think along European lines). 2.3.2. The negative prefix and the interrogative suffix The main allomorph for expressing negation in the verb, as has become clear, is a-, and for an older stage of Burushaski clearly a prefix *acan be reconstructed. For forming yes-no questions, Burushaski has a suffix (or enclitic) -a which from a typological point of view can be compared e.g. to Finnish -ko / -kö. Cf. data such as Ys. bái ‘he is’, a-pái ‘he is not’ (with mutation), bái-a ‘is he? ’. Both affixes thus consist of the vowel a, but at first there seems to be nothing remarkable about this fact: both affixes are as short as they could be, and a is a frequent vowel; the uniformity could be coincidental. There are more languages, however, which exhibit identity or similarity in their markers for negation and for questions. Sometimes these markers are independent words, sometimes affixes, but the latter may stem from independent words, of course. Thai, for instance, has mây negation, máy question marker (note the difference in tone). In Turkish negation is expressed by a suffix -m-, -me-, while the question marker is mi (plus vowel harmony variants). Latin once must have had a negation *ne ‘not’ from Indo-European *ne ‘not’, still present in nesci ó ‘I don’t know’ (despite the fact that the classical form is n ó n ‘not’, cf. 2.2.2.), and questions are marked in Latin with -ne. Moreover, in a large number of languages which fall under Niger-Congo or under Nilo-Saharan in the Greenberg classification of Africa there is an identity or similarity in markers for negation and for questions, as I was able to observe in data I had at my disposal from Matthew Dryer. Also Campbell (1997: 220) has <?page no="37"?> 2. Investigations into the typology of Burushaski 37 noticed the connection of negative markers and question markers. Looking across a wide range of languages, the pattern recurs. In some cases the observation may find an explanation as being due to chance, but coincidence cannot always be at work for statistical reasons. The statement is ventured here therefore that it can be a typological trait of languages to have a negative marker and an interrogative marker which are identical or similar. As I intend to argue now, a hypothesis about the cause for the similarity can be put forward. In many languages, e.g. German and Spanish, it is possible to add the word for ‘not’ or ‘no’ to a sentence in order to form a tag question, cf. Spanish ¿conoces al jefe, no? ‘you know the boss, don’t you? ’. It can be imagined that a construction of this type is reinterpreted in the syntactic history of a language as a more straightforward question ‘do you know the boss? ’. This development may have taken place in various languages. The result is a frequent identity or similarity of the negation and the question marker. Moreover, if a proto-language which has already undergone the syntactic reinterpretation splits up, numerous daughter languages may show the trait later as well; indeed they will do so unless either the expression of negation or the formation of questions undergoes a decisive innovation, be it phonetic or grammatical, which puts an end to the similarity. It can be argued that the assumed syntactic development occurred in the languages adduced above, such as Thai, or in predecessors of them. In each case the negative meaning must be the older one, and the interrogative meaning the younger one. Considering Burushaski now, it may be that negative aand interrogative -a are not identical by coincidence but that Burushaski is a language of the type under study, i.e. that the two affixes are historically identical. It is the typological background which makes this possible. The analysis would imply that the affix goes back to an independent word, because otherwise it could not occupy different positions in a sentence. The template of the Burushaski verb today has several prefix slots and several suffix slots (Anderson 2007: 1247). A general insight in morphologization is that usually the outermost layers of a template are the historically youngest. In the Burushaski verb there is never any prefix before a-, nor any suffix after -a. Therefore the above hypothesis has peeled away the content of the first slot and of the last slot of the Burushaski verb. When the elements for negation and interrogation became <?page no="38"?> 2. Investigations into the typology of Burushaski 38 parts of the verb forms, the rests of these verb forms, with all their wealth, were probably already in existence. 2.3.3. Verb forms without a stem A typological highlight of Burushaski is that there are some verb forms without a stem. In such forms the last prefix and the first suffix meet (are adjacent). The stem, with such verbs consisting of just one consonant, has disappeared due to phonetic change, and it can be recovered by internal reconstruction, or by synchronic analysis, depending on one’s viewpoint. The feature occurs in all dialects; all forms actually quoted in the following discussion are taken from the Yasin dialect. An example is éCum, a verb form approximately translatable as ‘doing’ or ‘doing it’, which can be segmented as é-C-um. The suffix -um is responsible for the effect that is translatable into English as -ing; it can be called a participle suffix. Before this, there is another suffix, the suffix for the present tense (or aspect) -C-. Finally, éis a prefix which means ‘it’ (also ‘him’), 3rd person object of gender m, x or y (this is the lowered allomorph of i-, mentioned e.g. in 2.3.1. and 2.3.4.). There is thus indeed no stem in this verb form. The stem of ‘to do’ is actually -t-, as can be seen e.g. in the form étum, approximately translatable as ‘having done’ or ‘having done it’, to be segmented é-t-um. This form has the same morphological structure as éCum, only with the difference that it does not contain a present tense suffix. The stem is kept here, whereas it disappears by assimilation in éCum. All forms of this verb which contain the suffix -Cdo not have a stem. It can be seen from the discussion that Burushaski has stems which consist of only one consonant (as already mentioned in 1.1.). This is known from other languages as well, e.g. in the Caucasus and in North America. Besides -t- ‘to do’, there are two other verbs in Yasin which behave in a similar way. One is -l- ‘to sting’, in the present of which the l is dropped in front of the present suffix -C-. The resulting forms are not homophonous with those of ‘to do’, however, because different allomorphs of the personal prefixes are used. Finally, there is the verb d-l- ‘to hit’; in line with the conventions of Burushaski studies this way of citing the verb is to say that the stem is also -l-, but there always occurs the d-prefix <?page no="39"?> 2. Investigations into the typology of Burushaski 39 before the personal prefixes. This verb loses its l before -Cas well; the allomorphs of the personal prefixes of d-lare those of ‘to do’. Hints of forms without any stem are even found in the literature, although only very rarely. Berger (1974: 31) claims that Ys. múCimi ‘he will give it to her’ does not contain a stem any longer, and for the Hunza dialect Anderson (2007: 1249) analyzes a “zero-stem allomorph” in ‘to come’. These instances require judgements about several issues and might benefit from further research. The question arises whether Burushaski is alone with the property of having verb forms without a stem, and in fact it is not. Some languages with this peculiarity are presented by Comrie (2003: 278f.). He discusses examples with the verb ‘to give’ from Amele and other languages spoken in New Guinea, as well as from Koasati (a Muskogean language, southeastern USA). Moreover, the Breton verb mont ‘to go’ has forms which consist only of suffixes (Ternes 2011: 477). Attention shall be drawn now to another language which has some verb forms without a stem: the Chukotko-Kamchatkan language Itelmen, spoken in the Russian Far East on the peninsula of Kamchatka. The following treatment is based on Georg / Volodin (1999). The verb affected is ‘to be’; its infinitive is L-ka-s, in which L-, a voiceless lateral, is the stem, -kais a suffix expressing intransitivity according to the authors (p. 142, although it occurs in many transitive verbs as well), and -s is the infinitive suffix. The present indicative of this verb has six forms, since there are three persons and two numbers; suffixes and partly also prefixes are employed to mark them. The forms are: sg. pl. 1st t-s-kiçen n-s-kiçen 2nd s0-ç s0-sx 3rd s0-n s0-ßn The authors do not give any segmentation (it has been added here), nor any glossing of the morphemes, but all information needed can be gathered from their grammar. All six forms contain a suffix for the present tense which is either -sor -s0-; that this is indeed a present tense suffix can be gathered from pp. 149, 151-153. All other morphemes (prefixes tand n-, and the remaining suffixes) indicate person and number (p. 142). The stem Lhas been assimilated to the sibilant of the present tense suffix, as the <?page no="40"?> 2. Investigations into the typology of Burushaski 40 authors state explicitly (p. 145). Therefore it does not appear in any of the forms. 2.3.4. Person marking I: On the shape of the affixes In Burushaski, person marking is a wide and subtle area, and it also shows some peculiarities in comparison with other languages. Some of these features will be described and put into a wider typological context here. Actually, the term person marking is a bit short: as usual, also number is marked in the verb. Moreover, Burushaski marks gender in the verb as well; this affects the 3rd person, and there are four genders. Thus, to be most exact, one would have to speak of person / number / gender marking. However, the term person marking is a convenient abbreviation. For the sake of bringing some order into the discussion, the treatment of person marking is divided into two parts. In 2.3.4. the focus will be on certain issues in the shape of the affixes used for person marking. Subsequently, 2.3.5. will deal with issues in the use of the affixes, i.e. such questions, for instance, whether they mark S, A or O in a morphological and syntactic context. Burushaski has both prefixes and suffixes that mark person. The person marking affixes are as follows (the Hunza dialect has -an for -en): prefixes suffixes sg. pl. sg. pl. 1st ami- -a -en 2nd guma- -a -en 3rd m iu- -i -en f muu- -u -en x iu- -i -en y ii- -i -i There is some minor allomorphy, and the choice of allomorphs partly depends on which stem they are attached to. For example, all prefixes containing i or u have variants with lowered vowels, i.e. e and o respectively, with certain verb stems. Moreover, stress can vary, and in Yasin prefixal iis dropped when unstressed (this was already seen with wál-i ‘he fell’ in 2.3.1. and will be seen again in 2.3.5. and 3.3.3.). The initial g of the 2nd person singular prefix alternates with k due to what has been called provective mutation (PM) here, cf. 2.2.2. The slight variation <?page no="41"?> 2. Investigations into the typology of Burushaski 41 in allomorphs is usually not relevant to the issues that will be discussed in the following. (There are forms of the 3rd person plural gender x in Hunza without final -n alongside the forms with -n, cf. Berger 1998I: 144; the reason for this remains unclear.) One point to note is that the gender distinctions in the 3rd person are not very elaborate. Other languages with four genders or classes, as some linguistic traditions prefer would exhibit four different affixes for a particular purpose. Burushaski is far from such a state of affairs in most parts of its structure. In the prefixes of the singular, for example, only two forms occur, namely mufor the feminine and ifor all other genders. Another point to note is that in some instances prefixes and suffixes may be identical. This affects: prefix suffix 1st sg. a- -a 3rd sg. m, x & y i- -i 3rd pl. y i- -i 3rd sg. f mu- -u The last line only constitutes an example of identity if one feels oneself able to account in some way for the presence of the m in the prefix or its absence in the suffix. Whether these two affixes are identical is indeed a question of historical linguistics, and the answer will become difficult given the limits of what can be told about the past of Burushaski. As to the first line, it may be argued that a suffix -a occurs also in the 2nd person singular, but it is a peculiar fact that in the Hunza dialect this suffix is in a slot different from the one for the suffix of the 1st person singular (Berger 1998I: 104, Anderson / Eggert 2001: 236), and this may speak against an identity. To sum up, in some instances prefixes and suffixes seem to be identical, whereas in other instances they are not. This is not uncommon from a general typological point of view; it applies also to other languages. An example is Swahili: when comparing the prefix series expressing subjects with the one expressing objects of this language, one also finds partly identities, partly differences. The longest discussion that will be presented here, however, will be about the suffixes in the plural. They are all -en except for one form, the last one, which is -i. Only little attention has been drawn to this fact, and it deserves further study. The next observation is that the suffix -i of the last form is identical to its counterpart in the singular. It has been suggested <?page no="42"?> 2. Investigations into the typology of Burushaski 42 that the plurals of nouns of gender y are actually old collectives (Tiffou / Pesot 1989: 50, n. 2). It would then be understandable both from a semantic and from a typological point of view if they required a singular verb form. A parallel is found in the Indo-European language family in which the syntax of some older languages (e.g. Ancient Greek) demands that neuter plurals occur with a verb form in the singular. If this one form is discarded, only -en remains as the suffix for all plural forms. Such a “unified plural” in person marking occurs in other languages as well. Cysouw (2003) is a cross-linguistic study of personal affixes in the verb. Cysouw (2003: 136-138) has found that a unified plural occurs also in Una, Waskia and Koiari, all spoken in New Guinea, as well as in Rongpo, a Tibeto-Burman language, and finally in some Germanic languages, of which Cysouw treats Modern English, Old English and Dutch. Modern English is the least impressive example because almost all forms are identical anyway; it is only the 3rd person singular in -s which deviates (except for to be). Other Germanic languages with a unified plural may be added. The first one is Faroese, in which vit ‘we’, tit ‘you (pl.)’ and teir ‘they’ always take the same verb form, e.g. hava ‘to have’ (Braunmüller 1991: 233). Another example is Platt, a West Germanic language spoken in northern Germany. In the Platt of Holsteen, for example, all plural forms take a suffix -t, which can be illustrated with don ‘to do’ (source: own competence of the language): sg. pl. 1st ik do wi do-t 2nd du deih-st ji do-t 3rd he deih-t se do-t Finally, Gothic has such an identity in the middle voice (Krahe 1961: 112f.), cf. the indicative present middle of the verb ‘to carry’: sg. pl. 1st baíra-da baíra-nda 2nd baíra-za baíra-nda 3rd baíra-da baíra-nda Also some varieties of Kurdish have a unified plural. There are in principle two possibilities why a language exihibits a “unified plural” in its person marking. One is that the forms were once different and language history then led to a unification, i.e. we are dealing <?page no="43"?> 2. Investigations into the typology of Burushaski 43 with syncretisms in the historical sense of the word. This is what is present in the Germanic languages adduced. There is another option, however: it is conceivable that the purpose of the suffix was to mark plural right from the start, and it is nothing but a logical consequence of this fact that all forms are identical. No decision will be made here on what to believe about Burushaski’s -en in this respect. It is a step forward, however, to become aware of the two alternatives just outlined, so that future research may return to the issue. 2.3.5. Person marking II: On the use of the affixes The previous investigation (2.3.4.) dealt with the shape of the affixes which are used for person marking. The following treatment is also about person marking, but now the focus shall be on when these affixes occur. This is by no means always straightforward in Burushaski but requires some explanations, many of which are not found yet in a clear way in the literature. First of all, it should be mentioned in passing that the prefixes occur not only in the verb, but also on inalienable nouns. Burushaski is, like Abkhaz, Lakhota etc., a language with a difference between alienable and inalienable nouns. Burushaski has its own way in what to do with this distinction. The inalienable nouns receive possessive prefixes: Ys. a-rén ‘my hand’, gu-rén ‘your hand’ etc., and it is even impossible for them not to express the possessor. The alienable nouns, in contrast, cannot occur with a possessive prefix. We will not dwell on this point: the investigation will only be concerned with the verb. Burushaski not only marks the subject in the verb, but also the object (provided the verb is transitive, of course). It makes sense to distinguish the subjects of intransitive and of transitive verbs. In the following, the terms S (intransitive subject), A (transitive subject) and O (transitive object) are used, following Dixon (1979, 1994) and other publications. There is a need to discuss intransitive verbs and transitive verbs. One might want to discuss intransitive forms first because their job is to encode only one argument (NP) and they may thus be expected to be simpler than transitive verbs, which encode two. Somewhat surprisingly, however, in Burushaski transitive verbs are in a certain way simpler, and they will therefore be dealt with first here. <?page no="44"?> 2. Investigations into the typology of Burushaski 44 Transitive verb forms are structured according to the template O-stem-A. This is to say that O is expressed by a prefix of the chart in 2.3.4., while A is marked by a suffix from that chart. From the stem -yéc- ‘to see’, for instance, examples can be formed such as mu-yéc-i ‘he saw her’, gu-yéc-a ‘I saw you (sg.)’, etc. (Yasin dialect; the tense of these and other forms is the aorist, which has no tense suffix). The fact that the prefix and the suffix are separated from each other by the stem has as a consequence that there is no need to discuss any fusional processes with person markers, such as often occur in languages with templates in which A and O are adjacent, e.g. A-O-stem, stem-O-A. It may be argued that person marking in transitive verbs in Burushaski is quite direct and simple. Turning now to intransitive verbs, an expectation that might arise is that S would be marked always in one and the same slot, i.e. either before or after the stem; thus the choice would simply be between the prefixes and the suffixes. With the first option an ergative pattern of marking would arise (S is then marked like O), and with the second option an accusative pattern would arise (S is then marked like A). Another conceivable pattern would be that some verbs use the prefixes, and others the suffixes; this is the pattern of active languages, or split-S in the terminology of Dixon (1994: 71-78). In fact Bashir (1985: 1) holds that Burushaski is reminiscent of an active language. Bashir (1985: 26, n. 1) gives credit to “Howard Aronson, to whom is due the important observation that Burushaski appears to fit the active typology”; Bashir reports that Aronson’s statement was made in a panel discussion. There are certain similarities to active languages in the structure of Burushaski. What we shall be concerned with here, however, is an investigation of person marking in the verb. Interestingly, this reveals that Burushaski is different from all three possibilities mentioned, i.e. it has neither ergative marking of person, nor accusative marking, nor active marking as it has been known so far. The basic point can best be outlined with the Yasin dialect. While active languages distinguish two types of intransitive verbs, there are in Yasin Burushaski three types of intransitive verbs; one type uses only suffixes, one both prefixes and suffixes, and one only prefixes. This can be illustrated with the examples 6as-á ‘I laughed’, a-wál-a ‘I fell’, and d-á-ya (with synchronically opaque d-prefix) ‘I came’: <?page no="45"?> 2. Investigations into the typology of Burushaski 45 suffixes both prefixes prefixes and suffixes sg. 1st 6as-á a-wál-a d-á-ya 2nd 6as-á gu-wál-a do-kó-ya 3rd m 6as-í wál-i d-í-ya f 6as-ú mu-wál-u do-mó-ya x 6as-í wál-i d-í-ya y 6as-í wál-i d-í-ya pl. 1st 6as-én mi-wál-en de-mé-ya 2nd 6as-én ma-wál-en da-má-ya 3rd m 6as-én u-wál-en d-ú-ya f 6as-én u-wál-en d-ú-ya x 6as-én u-wál-en d-ú-ya y 6as-í wál-i d-í-ya type -S type S--S type S- In the last line of the table the types have been given names which are to express where S is marked in relation to the stem; a name such as “-S” may be pronounced “hyphen S”. Taking transitive and intransitive verbs together, one arrives at the following table, which gives a summary: prefixes suffixes transitive O- -A intransitive -S (type -S) S- -S (type S--S) S- (type S-) Forms of intransitive verbs of the Type S--S may look similar to forms of transitive verbs at first glance. However, with type S--S intransitive verbs prefix and suffix always agree in person, number and gender, while in transitive verbs they cannot do this. A transitive verb form cannot mark 1st person or 2nd person twice. It can mark 3rd person twice, but then the arguments intended are not identical, e.g. you can say mu-yéc-u ‘she saw her’ which can only mean that one female person woman or girl saw another female person. (The similarity in structure to mu-wál-u ‘she fell’, from the paradigms above, is therefore not paralleled by the semantic facts.) When a reflexive meaning is intended (‘she saw herself’), this is <?page no="46"?> 2. Investigations into the typology of Burushaski 46 construed with an inalienable noun -khár which behaves like a direct object and takes a possessive prefix (Berger 1974: 27). In the following, some closer looks at the intransitive verbs will be taken. As the most remarkable type one may regard S--S due to its peculiar double marking. It could be maintained that this type marks person by circumfixes. However, such a viewpoint is not actually necessary. From a universal perspective it seems that circumfixes are actually always combinations of prefixes and suffixes if not synchronically, so at least historically; see Harris / Xu (2009: 287f.) for circumfixes marking possessor in Algonquian within a general discussion of circumfixes. In Burushaski there is no need to speak of circumfixes but it suffices to speak of a contemporaneous occurrence of a prefix and a suffix, because these affixes also occur on their own. On type Sit must be added that ‘to come’ is the only verb I have been able to find, and there is a remark by Tikkanen (1999: 284, n. 17) that this is the only verb of its type. This confirms the general rule that irregularities are often found in very central vocabulary items; in this case one may also point to the Hungarian verb jönni ‘to come’, which is irregular as well within an otherwise quite agglutinating morphology. It may be asked whether the fact that type Shas only one member does not constitute an argument for rejecting the classification into three types. There are several arguments against such a view. Firstly, if a morphological analysis enables one to identify a neat system, then why not subsume in it also classes that consist of few members, or only one? Secondly, other examples may be found in the future. Thirdly, one may bring diachrony into play. Experience shows that if an irregularity occurs in a central vocabulary item this is usually a hint of something old. English umlauted plurals men, feet, for instance, are known to be remnants of once more common formations. It is therefore possible to assume that the Burushaski type Sof ‘to come’ was once more widespread, and it is justified to account for it right from the start. The question can be asked whether a system can be detected in the distribution of type -S and type S--S, and if so, what kind of system this is. When investigating the intransitive verbs of Yasin, using the vocabularies and materials available, one gets to the conclusion that they are distributed by and large like in an active language. Type -S typically occurs with verbs where the S exerts control, e.g. gucár- ‘to walk’, hurúT- ‘to sit’, girát- ‘to dance’, and thus corresponds to the type in an active language in which <?page no="47"?> 2. Investigations into the typology of Burushaski 47 S is marked like A. Type S--S typically occurs with verbs where the S cannot control the event, e.g. -wál- ‘to fall’, -wár- ‘to be tired’, -yúr- ‘to die’, -mal- ‘to be afraid’, and thus corresponds to the type in an active language in which S is marked like O. The Hunza dialect will briefly be turned to now. It has the same three types of intransitive verbs as Yasin, although the Smarking with ‘to come’ is not present in all TAM (tense / aspect / mood) combinations; Smarking does occur in what Berger (1998I: 147) calls the “Konativ” (Anderson 2007: 1256 uses this German term in italics as well). The most interesting point about Hunza, however, is that from this dialect a feature is reported which seems unknown in Yasin: certain verb stems are sometimes used with prefix + suffix, and sometimes only with the suffix. As to the grammatical or semantic motivation, the situation is not entirely clear. There is a discussion of this without any definite result by Berger (1998I: 118-120). Berger (1998I: 118) states explicitly that further research is necessary: “Der Gebrauch bei den Verben mit teilweiser Setzung [of the personal prefixes] ist oft schwankend und noch nicht in allen Einzelheiten erforscht”. An example from Berger (1998I: 118) is: Hz. 6urc-ím-i ‘he dove under’ i-6úrc-im-i ‘he sank, drowned’ (also quoted by Anderson / Eggert 2001: 241). With some caution it may be possible to come up with new interpretations, and this has in fact been done. Anderson / Eggert (2001: 241) see a parallel to the Salish languages and apply the terms “control” and “out-of-control” which they take from the Salish linguistic tradition. Moreover, the available descriptions remind of what Dixon (1994: 78-83) calls fluid-S characteristics. In fact, Tikkanen (1999: 284) has already realized this point and mentions the term fluid-S explicitly: “Many verbs display so-called fluid-S marking, depending on whether or not the subject can control the activity.” He gives the following examples with the verb man- ‘to become’: In wakíil man-íi ‘He became a lawyer (by his own effort)’ In Son i-mán-i ‘He became blind (owing to some external cause)’ Taking an areal perspective, what can be observed now is that Modern Tibetan is a fluid-S language (Dixon 1994: 80). Dialects of Tibetan are geographically not far away from Burushaski, especially from Hunza. <?page no="48"?> 2. Investigations into the typology of Burushaski 48 There may therefore be an areal connection to the fluid-S characteristics detectable in the Hunza verb. 7 Finally, attention should be drawn to the fact that in Burushaski split-S and fluid-S phenomena only occur in person marking on verbs. Case marking on nouns, in contrast, is fully ergative in Burushaski. 8 2.4. Some afterthoughts This chapter consists of several investigations each of which is intended to elucidate issues in the structure of Burushaski and each of which has its conclusions. There are however a few general points to be added here. One is that it clearly still pays to investigate Burushaski from a typological point of view this language is a goldmine. It seems that a large bundle of typologically interesting features, of which only some could be addressed here, comes together in this language; such features cluster and interact in a kind of unique celebration. And although situated in the heart of Asia, Burushaski has traits that are otherwise only known, or more typically present, in languages far away, e.g. in the Caucasus, in Siberia or in the Americas. In the future more areal considerations could be drawn into the picture as well, supposed enough reliable information can be retrieved on the neighbouring Iranian, Dardic and Tibeto-Burman languages. Another point is that the will to identify and describe systems ought to become still stronger in research on Burushaski. From time to time it is a bit of a pity that the most industrious fieldworker, Berger, who passed away in 2005, was not always so keen on detecting systems but concentrated more on making single statements. The systems now sometimes have to be detected in the material by others, who, however, may not be as familiar with the language as those who actually had the opportunity to do fieldwork on it and could have checked a thesis with informants. To be sure, a language does not only consist of systems; there are also all those peculiar single facts which are sometimes relics of a past to be unraveled 7 Cf. also the comparison of the mutation systems of Burushaski and Tibetan in 2.2.2. 8 As to pronouns, there are conflicting statements and data on their case marking. For discussion, see Tiffou / Morin (1982), Tiffou (1999: 188f.). Further research may refer to the Silverstein hierarchy or nominal hierarchy, as some call it and the splits it often causes, see Dixon (1994: 83-97). <?page no="49"?> 2. Investigations into the typology of Burushaski 49 by historical linguistics. All systematic traits which are interwoven into a language, however, should be laid bare by linguistics. Further remarks should be made on terminology when dealing with Burushaski, especially in relation to the linguistic discipline of typology. Many language structures repeat in the world, and linguists ought to account for this fact: by pointing out the parallels and by trying to make world-wide suggestions for terminology and ways of presentation. Berger was indeed interested in typology, as e.g. his well-written lecture Berger (1992) shows. He was just sometimes lagging a bit behind in the general developments of the field. In his book on Yasin, Berger (1974), the term Ergativ (ergative) does not appear yet, the ergative case being called Agentialis there (Berger 1974: 20). It is true that some classics on ergativity, e.g. Comrie (1978), Dixon (1979) and Plank (1979), were published only later, but the term had of course already existed for many decades, cf. Dixon (1994: 3). In the grammar of Hunza and Nager, Berger (1998I), the term Ergativ is present. Still not present in this 1998 work, however, is the usual designation for its counterpart, the Absolutiv (absolutive), this case being called Kasus absolutus therein (Berger 1998I: 58). Instead, the term Absolutiv is used for one of the non-finite verb forms (Berger 1998I: 143, 231). The term fluid-S has first been brought into play by Tikkanen (1999), cf. 2.3.5., and split-S or active language may be useful as well in order to speak about certain properties of the verb. In 2.3.1. the suggestion has been made to use polarity. Berger seems to be unacquainted with the term and the concept of verbal plural or verbal number; his approach to the topic in Berger (1998I: 130) could have profited from Tiffou / Patry (1995). Moreover, I would like to express my dissatisfaction with how the term vowel harmony is used in much work on various languages; often it seems that any vowel assimilation is designated in this way. The idea shall be put forward here to restrict the term to such phenomena as in Finnish, Hungarian or Turkish, where there are genuine phonotactic constraints on the distributions of vowels, and suffixes have allomorphs with different vowels in order to keep the laws of phonotactics. To my knowledge, this is indeed the original sense for which the term was invented. Anderson (1997: 1033) uses the term vowel harmony for Burushaski, but it may be noted that this language does not have vowel harmony in the narrower sense. (It can only be debated whether Burushaski once had vowel harmony; on this vague possibility see the end of 3.3.1.) Last but not least, the <?page no="50"?> 2. Investigations into the typology of Burushaski 50 term mutations has not been in use in Burushaski studies, as pointed out in 2.2.2., and it was high time to introduce it. All these remarks are not to detract from the achievements that have been made. The suggestion shall just be made here to keep the terminology with which Burushaski is treated in line with other linguistic traditions and with typology wherever this makes sense. There is no doubt that the investigation of single languages among them this gem of a language, Burushaski and the field of typology can enrich each other. <?page no="51"?> 51 3. The dialects and the comparative method 3.1. Introduction As laid out in 1.2., the Burushaski language has three dialects: Yasin, Hunza and Nager (although the last two are almost identical). It is reasonable to ask whether the dialect variation can help to reveal something about the past of the language. This is indeed the case. The dialect differences can be investigated with the comparative method which has been applied successfully to Indo-European, Uralic, Austronesian, Algonquian-Ritwan and many other language families in the world. The Burushaski dialects are then treated as if they were sister languages. However, the Burushaski dialects are relatively close to one another; the language is more uniform than e.g. Basque or many other languages, whether language isolates or not. Therefore one must not expect too revolutionary results from dialect comparison. The detailed insights that can be revealed in this way, however, are worthy of attention and often highly relevant for further work on the language. They are presented and discussed in this chapter. 9 Concerning the variety from which all Burushaski dialects arose, one can speak of “Proto-Burushaski”. However, this designation must not evoke the idea that the procedure takes us very far into the past. Very often words are simply identical, e.g. ‘stone’ is dan in all dialects, and originally this word must have been *dan as well. Such an item would then be an example for a reconstruction of a word in Proto-Burushaski. This is a methodologically “clean” way of approaching the matter, but the example did not yield anything remarkable, and for many purposes one could cite the actually attested data instead. Operating with many starred forms does not always inspire more confidence and constitute a real gain. Things become more interesting, naturally, if there are actual deviations between the dialects, so that decisions have to be made when reconstructing. In these cases almost always the form that must be reconstructed is identical with a form which actually occurs in a dialect. 10 9 Compare the approach here with the one in the following chapter, which uses internal reconstruction: the results of the latter are often more far-reaching and refer to a greater time depth. 10 As laid out in 1.2., the Yasin dialect Burushaski possibly split off from the rest of the speech area in the 16th century AD. There may have existed some dialect variation already at that time, however. What I call Proto-Burushaski here may be five centuries old or a little more. <?page no="52"?> 3. The dialects and the comparative method 52 There have already been a few comparisons between the Burushaski dialects, but they are different from what will be attempted here. Most of these treatments are short, and often they compare only synchronic facts and do not necessarily have diachronic goals; if they do, the question is to which methods they subscribe. Bashir (2000: 8) underlines that comparative work on the dialects is much needed; she especially recommends “work on the comparative phonology and morphology of the three dialect areas”. Anderson (1997: 1034-1040) presents a readable comparison of the dialects. However, he uses data which is old often from Lorimer’s works and often inaccurate (e.g. stress lacking or incorrect, erroneously long vowels); meanwhile many new sources are available. Moreover, Anderson does not focus on sound laws but presents also sound correspondences which turn out to be sporadic. Berger (2008) is sometimes very speculative. Nevertheless, some of the ideas in this chapter can already be found in Anderson (1997: 1034-1040), Berger (2008) or elsewhere. Other insights are new. The depth of the investigation is unpreceded. The basis for much work on this chapter were the cognate sets between the dialects. Almost all existing cognate sets were evaluated. It is very useful that the Hunza dictionary by Berger (1998III) notes Yasin equivalents in its entries. 11 The Nager facts are as a rule the same as the Hunza data. Therefore, when attempting to reconstruct Proto-Burushaski, it usually suffices to compare only Yasin and Hunza, and this is what is mostly done in the following. Of course I made sure that no important available data from Nager was ignored for certain differences in quantity between Hunza and Nager, for instance, see 3.5. There are regular sound correspondences (some of which recur many times), and there are sporadic differences with no systematic background. The focus here is on the regular sound correspondences. When preparing this chapter the question arose in what order to treat certain areas. It would have felt most natural to me to start with the vowels and treat the consonants after that; this would fit the synchronic treatments in many grammars. However, many problems in reconstruction are interrelated, and this sometimes causes problems. In Burushaski dialect comparison, when discussing vowel issues it is sometimes useful to have a 11 Berger missed hardly any cognates. I am merely able to add: Hz. hir-úm ‘sharp’ (data from Berger 1998III: 200) / Ys. hér-eS ‘edge’, her-éS-um ‘sharp’ (data from Berger 1974: 151). <?page no="53"?> 3. The dialects and the comparative method 53 grasp of certain consonant issues already, while it is of lesser importance whether one knows much about the vowels when considering the consonants. Therefore I chose to present the consonant issues first. The vowel system of all Burushaski dialects is as follows (Tiffou 1999: 117 and numerous other sources): front back high i u mid e o low a The correspondences between these vowels, however, do not always involve identities. There are problems concerning e and i and problems concerning o and u. I would have treated the front vowels first and the back vowels afterwards, thus proceeding “from left to right”, but since the situation with e and i is even more complicated than the one with o and u, the discussion starts with o and u. The consonants are treated in 3.2., the vowels in 3.3.; other topics follow. 3.2. Consonants 3.2.1. Affricates There are three points of articulation for the Burushaski affricates: dental, retroflex and alveolo-palatal, and after some initial wavering (see Morgenstierne 1935: XXVII) their identification in each lexeme was not difficult for the fieldworkers. The problem is, however, aspiration. Voiceless plosives clearly fall into two groups, unaspirated and aspirated, in all dialects. It is unclear, however, whether voiceless affricates exhibit this distinction as well in every dialect. In Hunza / Nager they probably do: all authors today write that there are both unaspirated and aspirated affricates, the word lists and dictionaries agree on their occurrences in the vocabulary, and I personally hear the difference myself. This amounts to the following Hunza / Nager system: dental retroflex alveolo-palatal c c6 C ch c6h Ch For Yasin the situation is more complicated. Without the contrast only the first row of the system would exist. Berger (1974: 7, 10f.) claims <?page no="54"?> 3. The dialects and the comparative method 54 that Yasin lacks the distinction and notes only unaspirated affricates, see also his vocabulary pp. 136-140. Berger (2008: 5) maintains that aspiration was lost in Yasin between Lorimer’s fieldwork and his own (similarly Berger 1974: 11). Other researchers attribute the contrast [ ± aspirated] to Yasin but beyond that they disagree: the question then is which words contain an unaspirated affricate and which contain an aspirated affricate. As already mentioned in 2.2.2., the word for ‘meat’ provides an example. For Yasin, it is claimed to be Cap e.g. by Morgenstierne (1935: XXIX) and by Berger (1974: 138, 202), while Chap is claimed to be correct e.g. by Tiffou / Pesot (1989: 135) and by Tiffou (1999: 270). 12 In Backstrom / Radloff (1992: 243-260) the aspirated affricates in Yasin largely appear in those words where the Hunza cognates have them as well; there are only two exceptions: ‘goat’ (p. 251) and ‘five’ (p. 256) (Yasin without aspiration, Hunza with aspiration). In publications with Tiffou as an author, e.g. Tiffou / Pesot (1989), Tiffou (1995a), Tiffou (1999), the aspirated affricates are remarkably rare, and their occurrence does not match the occurrence in Hunza. It is also possible that some subdialects of Yasin have the contrast and others lack it. Regardless of whether one takes Tiffou’s or Berger’s Yasin data, one arrives at frequent correspondences of unaspirated affricates in Yasin and aspirated affricates in Hunza: Yasin Hunza ceréS chiríS ‘root’ cel chil ‘water’ culá chulá ‘he-goat’ gacér gachír ‘vulture’ c6en c6hin ‘bird’ CiS ChiS ‘mountain’ Cúmu Chúmo ‘fish’ Cardé Chardá ‘stallion’ There are also correspondences of unaspirated affricates in both dialects, i.e. c / c etc. In final position they are frequent due to phonotactics (no aspirates can be finals); elsewhere there are only few examples: 12 In Hunza the same problem once existed: Morgenstierne (1935: XXVII) heard Cap ‘meat’, while Lorimer heard Chap; all researchers today e.g. Berger, Tiffou follow Lorimer. <?page no="55"?> 3. The dialects and the comparative method 55 Yasin Hunza gunc gunc ‘day’ phunc phunc ‘dew’ -c6o -c6o ‘sibling of same sex’ -lCi -lCin ‘eye’ If we have correct data here, the conclusion must be drawn that unaspirated and aspirated affricates merged in Yasin. The phonetic background of the entire problem is that when hearing an affricate, aspiration is more difficult to perceive than when hearing a plosive. This has to do with the fricative phase of the affricate which resembles the aspiration. The fact can lead to problems in fieldwork, but possibly also to loss of a contrast in language change. 3.2.2. Clusters of nasal and plosive Sometimes Yasin has a consonant group of nasal + homorganic plosive, whereas Hunza shows only the nasal (Nager joins Hunza). For a discussion see Berger (2008: 42f.). The correspondences Ys. mb / Hz. m, Ys. nd / Hz. n and Ys. 9g / Hz. 9 occur: Yasin Hunza humbák humák ‘quiver’ (for arrows) 6éndeS 6éniS ‘gold; queen’ 13 ga9gí gá9i ‘axe’ gí9giS gé9iS ‘small leather bag’ (Ys. also gí9iS) bá9gut bá9ut ‘fortification, entrenchment’ -thé9gi -thée9i ‘dowry’ (Ng. -thée9) Consider also the tribe name Hz. Husé9uc, discussed in 4.2.4., and Hz. miSí9uc ‘six days’: miSínd- ‘six’, gunc ‘day’ (Berger 2008: 42), Ys. biSí9kuc ‘six days’; a similar set exists for ‘eight days’. The groups represent the older state, and the simplification is an innovation in Hunza / Nager. However, there are also sets in which the consonant groups were not simplified in Hunza / Nager: Yasin Hunza altámbe altámbi ‘eight’ cendí chindí ‘five’ 13 On the connection of ‘gold’ and ‘queen’ see 5.2. <?page no="56"?> 3. The dialects and the comparative method 56 biSínde miSíndi ‘six’ gandál gandál ‘beam’ hundáres hundáris ‘roof joist, roof beam’ Ca9gú Cha9gú ‘crippled’ -9gu -9go ‘uncle’ As can be seen, out of the first ten numerals three contain such groups. Here belongs also Ys. and Hz. bundó which is the plural of Ys. and Hz. bun ‘rock, large stone’. At first glance it is unclear whether one will find exact conditions for the simplification in Hunza / Nager. The challenge would be to find a phonetic criterion according to which the consonant group remained or changed. Having Verner’s law in mind, I would like to propose that stress plays a role: before an unstressed vowel the plosive was lost; before a stressed vowel, in contrast, the plosive remained, which is not surprising because it is louder and more conspicuous in that position. Although there are some exceptions, this thesis looks promising, given the data, and some further details may support it as well. It is interesting to learn that for ‘axe’ Backstrom / Radloff (1992: 246) record the stress always on the first syllable; moreover, the Hunza word alone is regular anyway. Tiffou / Pesot (1989: 26) list altambé instead of altámbe ‘eight’, although this could be due to the French the fieldworkers speak and it may not help the Hunza equivalent; Tiffou (1999) has altámbe. Finally, Hz. miSíndi ‘six’ may have preserved its plosive because of chindí ‘five’. (As will be seen in 4.3.1., ‘six’ contains ‘five’ historically.) It must be added now that the material presented so far lacks certain words, but there are good reasons for this which have to do with the fact that they can be suspected to be loanwords from Dardic languages. These words are: Yasin Hunza gambúri gambúri ‘flower’ / ‘tuft of feathers’ - Khowar gambúri ‘flower’ (Berger 1998III: 144) phandár phindár ‘wart’ - Shina phináar ‘kind of skin disease’ (Berger 1998III: 329) sénde sínda ‘river’ - Shina sin (Berger 1998III: 379) boróndo buróndo ‘ring’ - Shina baróno (Berger 1998III: 64) <?page no="57"?> 3. The dialects and the comparative method 57 bar6óndo ba6úndo ‘yeast’ (Ng. bar6úndo) - Khowar br á 6undo (Berger 1998III: 30) búndas múndas ‘kind of insect’ / ‘tick’ For this set I am unable to find a reference to Dardic. There are various reasons why probably Dardic languages were the donor languages with these isoglosses and not Burushaski. Some are relatively long words for which no segmentation is possible within Burushaski. If one assumes the word for ‘yeast’ to be of Dardic origin this explains the metathesis of r; in this way Burushaski got rid of the initial cluster br-, unknown to its phonotactics. Furthermore, there are irregularities among the sound correspondences of Yasin and Hunza; for example, the final Ys. -e / Hz. -a in ‘river’ is a correspondence which will be found to be irregular in 3.3.3., and the first vowels of ‘wart’ do not match either. If Proto-Burushaski did not have these words yet and they entered Yasin and Hunza / Nager independently from Dardic languages, this may very well have caused irregular correspondences; these are thus independent evidence that the words are possibly not inherited from Proto-Burushaski. One can now assume that some words entered Hunza / Nager only after the cluster simplification. 14 In contrast, almost all words treated previously do not have equivalents in Dardic; they seem to have been present already in Proto-Burushaski. Note especially that the numerals are no loanwords. The word for ‘roof joist, roof beam’ does occur in Shina as hunáris (Berger 1998III: 206), but this is a loan from Burushaski and not vice versa since it contains the Burushaski word hun ‘wood’. Later, in 4.2.4., a topic will be treated that is to some extent interrelated: the development that voiceless obstruents became voiced after nasals. 3.2.3. Other topics There is a sound correspondence Yasin x / Hunza qh, as can be seen from cognate sets such as: 14 Note some details. Shina exhibits *nd > n, a sound law that Khowar does not have. The first two words in the list do not contradict the sound law but it would not be wise to list them as evidence for it if one already suspects that they are loanwords. The words ‘ring’ and ‘yeast’ were present in Yasin early enough to undergo some vowel shifts of this dialect that will be discussed in 3.3.1. <?page no="58"?> 3. The dialects and the comparative method 58 Yasin Hunza xos qhus ‘cough’ xorc qhurc ‘dust’ xork qhurk ‘chaff’ -xórpet -qhúrpat ‘lung’ -xát -qhát ‘mouth’ -xés- -qhís- ‘to tear’ xéli qhéli ‘a little bit’ Since a change from an aspirated uvular plosive to a fricative is more likely than the opposite development, *qh ought to be reconstructed in order to account for the correspondence. Moreover, the Yasin consonant system has p ph, t th, etc., but only q, no qh, and hence the reconstruction leads to a regular system with aspirated plosives at all places of articulation. Yasin underwent *qh > x (unconditioned), thereby removing *qh from its consonant system and acquiring x, which was unknown before. Note also the correspondence x / q in the position before another consonant in the set ys. baxtá ‘sheep’ / hz. baqtá ‘kind of sheep’. There appears to be a parallel with the voiced sounds written 6, cf. 2.2.2. (A.) on this topic. Although in Burushaski linguistics the symbol 6 is used frequently and for any dialect, there seems to be actually more than one sound represented in this way, and which one it is depends on the dialect or speaker. Probably Hunza / Nager has a plosive here and Yasin a fricative. The plosive is clearly uvular (and hence distinct from the velar g), in IPA it would be written [G]. The fricative is more probably velar, i.e. [V]; the description by Berger (1974: 10) is not clear. Also the word lists by Backstrom / Radloff (1992: 243-260) mostly indicate [G] for Hunza / Nager and [V] for Yasin. Morgenstierne (1935: XXVIII) reports that for Hunza he heard a voiced velar affricate from a speaker from Baltit. It is possible to make a statement about Proto-Burushaski even without going into details on the exact pronunciations today. Regardless of which dialects or which speakers have a plosive, an affricate or a fricative nowadays, obviously the original sound was a voiced uvular plosive, IPA *[G]. This holds because this sound can easily shift to a fricative (or to an affricate). This is also the reason why 6 and q can be interrelated in the mutation system in what I called the provective mutation; 6 and q have the same relationship as e.g. b and p, see 2.2.2. <?page no="59"?> 3. The dialects and the comparative method 59 Hunza has a fricative of which there is no really clear description in the literature (see Berger 1998I: 22); it is usually written as y with a dot under it. This is not a very frequent sound, and some words containing it are loanwords. Yasin does not have this consonant. Yasin words often have no sound in the positions where Hunza cognates have this sound (Berger 2008: 36), and this is probably the correspondence which ought to be regarded as regular. See also Anderson (1997: 1026f., 1038). The matter is irrelevant for any point that will be made in this book. The fricative h occurs in all dialects. In initial position hin one dialect corresponds regularly to hin the other dialects: Ys. har ‘ox’ - Hz. har ‘ox’. (For a few irregular losses or additions of initial hsee Berger 2008: 35, although some cases stem from the earliest fieldworkers and possibly merely represent mistakes.) In intervocalic position, however, and probably in general in the interior of words, Ys. -hcorresponds to - ∅ in Hunza and Nager: Yasin Hunza baháltbáalt- ‘to wash’ balhás bilás ‘witch’ bihái biái ‘illness’ 6uhá 6uá ‘waning moon’ / ‘new moon’ -híl -íl ‘lip’ -húT-is -úT ‘foot’ -yúhar -úyar ‘husband’ 15 -húles -úlus ‘brother of a female’ (see 5.4.) -híl-, -húl- -íl- ‘to dip, to soak’ do-hóndu-ún- ‘to seize’ behék biík ‘willow’ bahúr boór ‘salty earth’ A Hunza / Nager sound law *-h- > ∅ can be deduced from these and other cognate sets. In inalienable nouns such as ‘lip’ and verb stems such as ‘to dip, to soak’ the crucial position is non-initial in Hunza because there is always a prefix before the stem. As Berger (2008: 34) points out, by 15 This set clearly exhibits a metathesis of *y and *h either in Yasin or in Hunza / Nager. (The Nager form is identical with Hunza, Berger 1998III: 460.) Burushaski itself does not permit a decision in which branch the metathesis occurred. There are similar words in some Iranian languages, e.g. Wakhi S ohár ‘husband’ (Backstrom / Radloff 1992: 279). If these are etymologically identical, this speaks for Hunza / Nager as the innovator. <?page no="60"?> 3. The dialects and the comparative method 60 analogy -hwas restored in some grammatical forms, e.g. Hz. nu-hér ‘having wept’ from the stem hér- ‘weep’. There are also instances, however, in which such an analogical restoration did not occur, see examples with the negative prefix *ain 2.3.1. The last example, bahúr / boór, also shows a monophthongization which will be treated in 3.3.3. A brief comparison to Ancient Greek shall be drawn. This language (Attic and other dialects) underwent the same treatment of *h as Hunza / Nager: in initial position it was preserved but it was dropped in all other positions. The typological parallel is perfect as far as the phonetic development is concerned but a difference is that the analogical restorations which made new instances of internal h arise in Hunza / Nager are unknown to Greek. Plausible reasons can be found for the difference in behaviour. Burushaski has more prefixation than Greek, often causing allomorphy, and it often has shorter stems than Greek. Therefore the pressure for analogical restorations of h was higher in Hunza / Nager than in Greek. A velar nasal 9 in Yasin can correspond to a labial nasal m in Hunza, as evidenced by: 16 Yasin Hunza -nú9us -dúmus ‘knee’ -yú9us -úmus ‘tongue’ There are also Yasin / Hunza correspondences of 9 / 9 and of m / m. For 9 / 9 consider this set, for instance: Yasin Hunza Ti9án Ti9án ‘egg’ 17 16 The correspondence -n- / -din ‘knee’ is irregular (see also 3.6.). For y in the Yasin word for ‘tongue’ see 3.7.3. There is a third example which rhymes with the two ones given here (Berger 2008: 39). 17 This is one of the very few Burushaski words today on which there is disagreement whether it contains a retroflex or a dental (as the first sound). Berger writes Tin his publications for all dialects, and so does Morgenstierne (1935: XXVIII) for Hunza, while Backstrom / Radloff (1992: 250) have tfor all dialects and Willson (1999: 112, 140) has tfor Hunza. Lorimer (1962: 231) has tfor Yasin and Tfor Hunza, but Lorimer (1935a: 421) has tfor Hunza. Tiffou’s publications usually do not contain the word and thus evade the problem. Another word which is sometimes quoted with tand sometimes with Tis the numeral ‘nine’, Ys. hutí. In this book I follow Berger in having Tin ‘egg’ and tin ‘nine’. <?page no="61"?> 3. The dialects and the comparative method 61 Although the correspondence 9 / m is rare, it is regular nevertheless. It occurs between two rounded vowels. However, also m / m can occur in this environment, as evidenced by: Yasin Hunza Cúmu Chúmo ‘fish’ 6omór 6umór ‘hole’ For 9 / m it is not possible to reconstruct a labial nasal therefore. It is possible to reconstruct *9, however, and also Berger (2008: 39) advocates this. The reconstruction fits another consideration: a sound law must make sense from a phonetic point of view, and if *9 is posited, the shift to m can be interpreted as a takeover of the labiality of the two surrounding vowels. One arrives at the following diagrams: Proto-Bur. *9 *m Proto-Bur. *9 *m Yasin 9 m Hunza 9 m The oblique line refers to the position between two rounded vowels. In other environments *9 was preserved in Hunza. There may be a typological parallel for the law in the Turkic language family, cf. Old Turkish to9uz ‘pig’, Modern Turkish domuz ‘pig’. The treatment requires an afterword on a local development within Yasin which presumably exists. Lorimer (1962: 183), in his Werchikwar (i.e. Yasin) vocabulary, also has -númus alongside -nú9us for ‘knee’, although this form stems from only one informant, Shahzada Yusuf. (Initials in Burushaski dictionary entries often stand for informants.) Moreover, Lorimer (1962: x) reports that this was a speaker of Khowar. If the form is correct, it may mean that in one place within the Yasin speech area the shift *9 > m occurred as well. The form -númus is also listed in later literature: Morin / Tiffou (1989: 40), Tiffou / Pesot (1989: 145), Berger (1998III: 125) and Tiffou (1999: 287). Backstrom / Radloff (1992: 244) recorded -úmus for ‘tongue’ from the Yasin dialect of the village Thui (Northern Yasin). They do not have ‘knee’ in their word list. It must be emphasized that most Yasin subdialects employ 9 in both words. The next point see also Berger (2008: 137) is concerned with variation within Yasin, namely of the clusters rc6 (or rS in one item) and Sc6: <?page no="62"?> 3. The dialects and the comparative method 62 Yasin Hunza / Nager barc6, baSc6 baS ‘bridge’ harc6 18 , haSc6 harS ‘plough’ bisárS, bisáSc6 bisárS ‘sickle’ (Hz. also bicárS) In Yasin the clusters with r represent the older state. This is evident not only from the Hunza / Nager cognates in two cases but also from derivational facts: the noun ‘plough’ is a derivative of har- ‘to plough’, and ‘sickle’ belongs to Ys. and Hz. bisárk- ‘to reap, to harvest’. From a typological point of view the variation within Yasin reminds of the Basque dialect variants bortz and bost ‘five’. Finally, there is a correspondence of Ys. -yand Hz. - ∅ -. There are not many cognate sets which illustrate it: Yasin Hunza -yé -í ‘son’ -yék -ík ‘name’ buyé buí ‘shovel’ da6óye da6úi ‘unripe’ / ‘raw’ mayén meén ‘old’ (of things) -yúr- -ír- ‘to die’ Apparently Hunza lost *y before i. Most of these words also show a vowel correspondence Ys. e / Hz. i and will be taken up again in 3.3.2. because of this. The word for ‘old’ also shows the monophthongization which will be treated in 3.3.3.; the Proto-Burushaski shape may have been *mayín or *mayén. The vowel difference u / i in ‘to die’ will be dealt with in 4.2.1. 3.3. Vowels 3.3.1. The back vowels o and u The back vowels o and u require a somewhat longer discussion in Burushaski. When comparing Yasin and Hunza, it can be seen that these vowels enter into all four possible correspondences, i.e. o / o, u / u, o / u and u / o. In order to treat the question how to reconstruct Proto- Burushaski with regard to o and u, it is indispensable first to collect a certain amount of material a too small data base may lead to incorrect 18 Ys. harc6 is from Berger (1974: 150); Berger (1998III: 194, 2008: 137) has harS instead for Yasin. <?page no="63"?> 3. The dialects and the comparative method 63 conclusions. Subsequent work must then evaluate this material with the methods of historical linguistics and find the solution which accounts best for the data and which is most likely to meet the prehistoric facts. In the following, the cognate sets are arranged according to the correspondences (o / o, u / u, etc.) and within these in groups (group A, B, etc.). The purpose of the group distinctions may in some instances be immediately visible; in others it will become clear later. Sometimes the syllable is specified which is being referred to. Large groups are not always listed completely; in general, the smaller the group, the harder I tried to be exhaustive in listing all material that the language offers. Some cognate sets occur twice or (rarely) even three times, namely if they contain more than one rounded back vowel. The correspondence o / o between Yasin and Hunza can be illustrated by the following cognate sets: Yasin Hunza Group A gon gon ‘dawn (in the morning)’ Son Son ‘blind’ thoS thoS ‘new’ toq toq ‘mud’ -móq- -móq- ‘to pluck out’ manóT manóT ‘celebration’ 6omór 6umór ‘hole’ (2nd syll.) hósar hósar ‘pumpkin’ hol hol ‘army’ -xól- -qhól- ‘to hurt’ -ó -ó a frequent plural suffix -tóto -tóto ‘paw’ (1st syll.) horó6o huró6o ‘sweat’ (2nd syll.) boróndo buróndo ‘ring’ (2nd syll.) CoCór juCór ‘shaving’ (2nd syll.) Group B -tóto -tóto ‘paw’ (2nd syll.) horó6o huró6o ‘sweat’ (3rd syll.) boróndo buróndo ‘ring’ (3rd syll.) bar6óndo ba6úndo ‘yeast’ (3rd syll.) (Ng. bar6úndo) <?page no="64"?> 3. The dialects and the comparative method 64 This is material for the correspondence u / u: Yasin Hunza Group C gus gus ‘woman’ gur gur ‘wheat’ bur bur ‘hair’ culá chulá ‘he-goat’ Cúmu Chúmo ‘fish’ (1st syll.) un un ‘you (sg.)’ hun hun ‘wood’ huk huk ‘dog’ hurgás hurgás ‘thick’ humá humá ‘ford’ multán multán ‘blood’ Suá Suá ‘good’ phu phu ‘fire’ phunc phunc ‘dew’ -úl -úl ‘abdomen, belly’ -um -um a frequent suffix búpuS búpuS ‘pumpkin’ (both syll.) -nú9us -dúmus ‘knee’ (both syll.) -yú9us -úmus ‘tongue’ (both syll.) hurúThurúT- ‘to sit’ (both syll.) 6urpúS 6urpúS ‘barn’ (2nd syll.) More material exists for group C, but it would not help the investigation to list it. Group D 6uTúm 6uTúm ‘deep’ (1st syll.) 6usánum 6usánum ‘long’ (1st syll.) 19 6uhá 6uá ‘waning moon’ / ‘new moon’ 6uriáS 6uráS ‘feces’ 6urpúS 6urpúS ‘barn’ (1st syll.) -sá6un -sá6un ‘nephew, niece’ du-Cá6urdu-Chá6ur- ‘to cool down’ 19 Although Tiffou writes 6osánum for Yasin in several publications, e.g. Tiffou (1999: 276). <?page no="65"?> 3. The dialects and the comparative method 65 The correspondence o / u is shown by: Yasin Hunza Group E asqór asqúr ‘flower’ iSqórum Suqúrum ‘sour’ (2nd syll.) xos qhus ‘cough’ xorc qhurc ‘dust’ xork qhurk ‘chaff’ -xórpet -qhúrpat ‘lung’ -6ón- -6ún- ‘to give’ 6ól- -6úl- ‘to burn’ (intr.) 6órkun 6úrqun ‘frog’ (1st syll.) -6óya9 -6úya9 ‘hair’ 6al6ó 6al6ú ‘worm’ da6ó da6ú ‘glue’ da6óye da6úi ‘unripe’ / ‘raw’ do6ói du6úi ‘noon’ (2nd syll.) (Ng. du6ói) ha6ór ha6úr ‘horse’ ha6ós ha6úc ‘pass’ (in mountains) bar6óndo ba6úndo ‘yeast’ (2nd syll.) Some additional examples exist for group E. Group F doró duró ‘work’ (1st syll.) horgó hurgó ‘ascent, rise’ (1st syll.) horó6o huró6o ‘sweat’ (1st syll.) boróndo buróndo ‘ring’ (1st syll.) CoCór juCór ‘shaving’ (1st syll.) do6ói du6úi ‘noon’ (1st syll.) Group G xoró9 qhurónc6 ‘clouds’ / ‘cloud, fog’ (1st syll.) 6omór 6umór ‘hole’ (1st syll.) 6onó 6unó ‘seed, sperm’ (1st syll.) 6oró 6uró ‘stones’ (1st syll.) <?page no="66"?> 3. The dialects and the comparative method 66 Finally, this is some evidence for the fourth correspondence, u / o: Yasin Hunza Group H phéSu phéSo ‘pear’ Cúmu Chúmo ‘fish’ (2nd syll.) sáu sáo ‘sand’ qíu qío ‘scream’ -9gu -9go ‘uncle’ -u -o a frequent plural suffix -taru -taro a plural suffix brá6u birá6o ‘foal’ tul tol ‘snake’ For this group consider also Ys. páqu ‘bread’, Hz. pháqo ‘kind of bread; bread’, which must be a loanword from a Dardic language; in Shina a word páko ‘ripe, cooked’ exists (Berger 1974: 168, Berger 1998III: 323). The ultimate root is Proto-Indo-European *pek w - ‘to cook’ (in Indo-Iranian *e > a, *k w > k). A survey of correspondences and groups is as follows: correspondence groups (Ys. / Hz.) o / o A, B u / u C, D o / u E, F, G u / o H The question is now how many vowels, and which ones, ought to be reconstructed in order to account for the data. The possible maximum is four (one for each correspondence), but this would possibly lead to a strange vowel system, and the maximum is by no means necessary, as the data will reveal. It will certainly not work to posit just one vowel, but let us try assuming two ones, *o and *u; the correspondence o / o would then demand *o, and u / u would go back to *u in a straightforward manner. The challenge is whether the remaining two correspondences o / u and u / o can be reconciled with this preliminary idea. The answer is yes, but it takes a while to lay this out. To start with, an observation on the correspondence o / u can be made. This correspondence is dominated by the large group E. In this group, the vowels in question always occur after the uvular consonants <?page no="67"?> 3. The dialects and the comparative method 67 (plosives) of Proto-Burushaski, *q, *qh and *6. The symbol 6 often stands for a voiced uvular plosive [G] in Burushaski linguistics, and this sound was certainly present in Proto-Burushaski, see 2.2.2., 3.2.3. In Yasin, some of the uvular plosives were shifted to fricatives, and their place of articulation may be velar now, see 3.2.3. However, this shift can have been later in relative chronology, and this should be assumed. As known from many languages, e.g. Inuit and Quechua, uvulars can have a lowering effect on vowels. The phonetic background is that the uvular place of articulation is itself located relatively low in the vocal tract. It makes sense to assume that Hz. u represents the older state and in Yasin after uvulars *u was lowered to o. This is approximately also what Berger (2008: 46) advocates. In the following, I speak of the uvular law. In Hunza, *u was not affected by a preceding uvular; it remained u. It can be checked whether the assumption comes into conflict with any data. It turns out that group D exists which has preceding uvulars (or former uvulars), but the correspondence u / u. Even though group D is relatively small, it must not be ignored. (No group, of whatever size, will be ignored in this discussion.) It can be observed now that in group D the reconstructed *u was always unstressed, whereas in group E *u was always stressed (*ú). However, there exists a group G, even smaller than D, which has unstressed o after uvulars. Stress may certainly be of significance for the whole problem but the question is how. One may try assuming that only stressed *ú underwent the uvular law. This would explain group D but it would come into conflict with group G. If stress is considered irrelevant, this explains group G but it comes into conflict with group D. Group E would fit both accounts. Without going into details (since this would lead to an even longer discussion): limiting the uvular law to the stressed position would in the end lead to a scenario which could be reconciled with the data. However, there are phonetic and typological questions which would arise; it makes no sense why the lowering should be limited to the stressed position, and other languages, such as Inuit, show no connection to stress. Therefore I assume that the uvular law applied in all cases, i.e. independently of stress. As will be seen later on, this fits the data, too. Groups D and G will be discussed again. First the focus will now be on the correspondence u / o, for which only group H exists. It is striking that the vowels in question occur almost always in unstressed position. If one tried to posit unstressed *u > o in <?page no="68"?> 3. The dialects and the comparative method 68 Hunza, this would cause a conflict with various data in group C and elsewhere. However, it is possible to assume that Yasin underwent unstressed *o > u and this is more natural from a phonetic point of view anyway, cf. European Portuguese, where unstressed o is pronounced [u] today, and other languages. The law also explains some cases of allomorphy in Yasin such as stressed 6ól- / unstressed 6ul- ‘to burn’ (Berger 1974: 31, 148), Sol ‘avalanche’ / Sul- ‘to fall down’ (stress on suffix) (Berger 1974: 178). The development can also be seen with one of the most frequent plural suffixes, which in Hunza is -ó when stressed and -o when unstressed, but in Yasin it is stressed -ó, unstressed -u. The shift can be called the unstressed law. The only exception is tul / tol ‘snake’. For this set I would like to draw attention to the Yasin plural which is tul-iánc or tul-ánc (Berger 1974: 182): the plural form had unstressed *o which was shifted to u, and I assume that the singular received u due to analogy (leveling). 20 The question arises how the uvular law and the unstressed law relate to each other as regards time. Ys. páqu / Hz. pháqo (see above under H) shows that the unstressed law cannot have occurred before the uvular law since in that case the unstressed law, in the face of the *o, would have produced an u which then, however, would have been switched back to an o by the uvular law. The opposite chronology, however, works well since then the uvular law left the *o unaffected and later the unstressed law shifted it to u. Moreover, the opposite chronology explains group D which was already touched above: in words such as 6uTúm ‘deep’, first the uvular law shifted *u > *o, and later the unstressed law shifted *o > u. (As to group G, it will be discussed later.) 21 Something still remains to be done about the unstressed law: one should check whether it may become troublesome for some data. It turns out that group B had unstressed *o, but nevertheless the reflex in this group is not u, but o. An important observation is now that in all words of group B the stressed vowel is ó in Yasin. This fact leads to a new insight. Apparently an assimilation in vowel height was at work in Yasin: the 20 Another difficult example may be Ys. laphúT / Hz. laphóT (data from Berger’s works): this word is followed by the word for ‘finger’ (in both dialects), and the resulting phrase means ‘thumb’. But Lorimer noted the same vowel in both words. 21 The unstressed law reintroduced the combinations q, x, 6 + u, whose predecessors had dropped out of the system due to the uvular law, to Yasin phonotactics. Moreover, loanwords contain these sound sequences, e.g. Ys. xuk ‘pig’, which is from Urdu x ú k, and so do onomatopoetic items such as Ys. xur étor qur ét- ‘to snore’ (Berger 2008: 46). <?page no="69"?> 3. The dialects and the comparative method 69 group did undergo the unstressed law but later an assimilation made o arise again. Group H does not offer any contradicting cases for this view. 22 The cognate set Ys. bar6óndo / Hz. ba6úndo / Ng. bar6úndo ‘yeast’ (listed in groups B and E) shows that first the uvular law must have applied and later the assimilation: the assimilation needs the presence of the o in the stressed syllable in order to work. Consider now group F. Within the correspondence o / u, most words have a preceding uvular before their *u and this fact has already been addressed (the uvular law was posited). Group F deviates in this respect. Therefore the uvular law cannot be used for it; there must be another law. The decisive observation is that the next syllable always has stressed ó in Yasin. It must be assumed that in this group an assimilation *u > o was at work due to the stressed o. This is the same assimilation as the one in group B just discussed. The set Ys. do6ói / Hz. du6úi ‘noon’ (member of groups E and F) demands that the uvular law precedes the assimilation (since the assimilation requires the presence of an ó after the 6). In fact the same conclusion about the chronology was already drawn in the preceding paragraph when examining the set for ‘yeast’. The only group which remains to be discussed now is the very small group G. These words always show a stressed ó. Therefore, at first glance, it looks at if the Yasin words could have their o in the 1st syllable either due to the uvular law or due to the assimilation. However, three laws have been arrived at in the preceding discussion, and also their chronology has become apparent. Given this background, it turns out that group G apparently went through all three laws: the uvular law caused *u > *o, then the unstressed law *o > *u, and finally the assimilation *u > o again. To sum up, the results of the investigation are as follows. There were two vowels *u and *o in Proto-Burushaski. In Hunza no change affected these two vowels. In Yasin several laws applied which can be summarized as follows: uvular law: *u > o after uvulars unstressed law: *o > u when unstressed assimilations: *u > o if preceding or following syllable had o 22 There is relatively little material for unstressed *o in Proto-Burushaski. Apparently *o was a rare vowel in unstressed syllables in Proto-Burushaski (and by the way the same applies to *e). <?page no="70"?> 3. The dialects and the comparative method 70 It is uncertain whether some of these laws applied at the same time. If they all fall into different times, the only relative chronology that works, and it works impeccably, is uvular law unstressed law assimilations. It can be evaluated which groups each law affected: law reminder groups uvular law: u > o E, D, G unstressed law: o > u H, D, B, G assimilations: u > o F, B, G The opposite perspective can be taken as well: each group can be considered and asked which laws it underwent. Groups A and C did not undergo any law at all. Groups E, H and F underwent one law each. The remaining three groups would show some switches to and fro between o and u. Group D would have *u > *o > u, group B *o > *u > o, and group G *u > *o > *u > o. Some further remarks on o and u in Burushaski shall be added which have nothing to do with the sound laws just mentioned. A few affixes have two allomorphs in both dialects, e.g. the personal prefixes o- / u-, mo- / mu-. This peculiarity also applies to the front vowels; there is e.g. a prefix e- / i-. Possibly an earlier vowel harmony is the cause which shows only tiny remnants today. There are a few cases in which a morphophonological alternation o / u exists both in Yasin and in Hunza. In these cases it is probably wisest to assume that the situation existed like this already in Proto-Burushaski. The adjective Ys. thoS ‘new’, pl. thu-á has an alternation o / u. This applies also to Hunza / Nager: thoS ‘new’, pl. thu-áa9 and thu-áan-c (Berger 1998III: 441); hence the alternation seems to be already of Proto-Burushaski age. (On the absence of S in the pl. see 4.2.3.) Another example is Ys. tórum ‘ten’, but turmain numerals such as turma-hék ‘eleven’; here, too, also Hunza has the alternation (Berger 1998I: 100). Theoretically the unstressed law would explain the Yasin facts but the point is that the Yasin data must not be judged alone in these cases; Hunza exhibits the same facts and hence they are probably inherited from Proto-Burushaski. 3.3.2. The front vowels e and i The front vowels vowels e and i present another challenge in Burushaski dialect comparison. At first glance the problem looks similar to <?page no="71"?> 3. The dialects and the comparative method 71 the one of o and u. In some respects this is true, but there are also numerous differences, and it soon shows that the overall situation is not parallel. Investigation of Yasin and Hunza reveals that e and i enter into three correspondences: e / e, i / i and e / i. There is thus one type less in comparison to the back vowel problem. 23 Material for the correspondences will be presented below. It could be asked whether it would work to simply take over the laws that apply to o and u (see 3.3.1.). It can be reported already here that this is not the case. The uvular law can hardly be made use of; there are only few instances of uvular + front vowel, and also i occurs after uvulars in Yasin: -móqiS ‘face’, di-Sqín- ‘to grow’. The unstressed law of the back vowels is based on the fact that there is a correspondence u / o, but there is no equivalent with the front vowels of this correspondence and thus the unstressed law cannot play a role. Systematic assimilations, however, will indeed be used later on. These are cognate sets which illustrate the correspondence e / e (there are many more examples): Yasin Hunza Group A sel sel ‘needle’ be be ‘not’ -hénhén- ‘to know’ hérhér- ‘to weep’ sénsén- ‘to say’ -mé -mé ‘tooth’ -mélc6 -mélc6 ‘jaw’ men men ‘who’ den den ‘year’ del del ‘oil’ ther ther ‘dirt’ thalé thalé ‘seven’ barés barés ‘blood vessel’ 23 Anderson (1997: 1037f.) claims that there is also a correspondence i / e, for which he gives two alleged examples, ‘needle’ and ‘oil’. However, it remains unclear where he has his Yasin data from. According to all sources known to me (e.g. Lorimer 1962: 78, 205, Berger 1974: 141, 175, Tiffou 1999: 271, 292), these words have e in Yasin, not i: sel ‘needle’, del ‘oil’. I have merely come across the set Ys. gí9giS, also gí9iS / Hz. gé9iS ‘small leather bag’ with i / e. <?page no="72"?> 3. The dialects and the comparative method 72 phéti9 phéti9 ‘ashes’ (1st syll.) béles bélis ‘sheep’ (1st syll.) 6éndeS 6éniS ‘gold; queen’ (1st syll.) khené khiné ‘this’ (2nd syll.) helés hilés ‘boy’ (2nd syll.) -e -e several case suffixes This is the correspondence i / i: Yasin Hunza Group B galgí galgí ‘wing’ halí halí ‘birch bark’ girí girí ‘ibex’ (both syll.) Siní Siní ‘summer’ (both syll.) milí milí ‘medicine’ (both syll.) hutí huntí ‘nine’ -lCi -lCin ‘eye’ -i -i suffix of 3rd pers. sg. m, x, y bi bi ‘is’ (3rd pers. sg. m, x, y) iiprefix of 3rd pers. sg. m, x, y mi mi ‘we’ -Cí- -Chí- ‘to give’ bicá bicá(n) ‘are’ (3rd pers. pl. y) Ti9án Ti9án ‘egg’ bihái biái ‘illness’ giSágiSá- ‘to weave’ biSábiSá- ‘to throw’ bisárkbisárk- ‘to reap, to harvest’ girátgirát- ‘to dance’ Group C -híl -íl ‘lip’ -díl -ndíl ‘breast (female)’ 6aCíS 6aCíS ‘awn, beard’ CiS ChiS ‘mountain’ tiS tiS ‘wind’ hir hir ‘man’ gírkis gírkis ‘rat, mouse’ (both syll.) <?page no="73"?> 3. The dialects and the comparative method 73 girmíngirmín- ‘to write’ (both syll.) -skil -skil ‘face; surface’ il il ‘eye of a needle’ díltar díltar ‘butter-milk’ hi9 hi9 ‘door’ -i9 -i9 a frequent plural suffix 24 Finally, these are cognate sets for the third correspondence, e / i: Yasin Hunza Group D hek hik ‘one’ -rek -rik ‘brother-in-law’ -ken -kin ‘liver’ -meS -miS ‘finger’ phen phin ‘fly’ (noun) khen khin ‘flea’ c6en c6hin ‘bird’ ten tin ‘bone’ menmin- ‘to drink’ bes bis ‘fat’ (noun) hesk hisk ‘back of the hand’ teSk tiSk ‘dagger’ delk dilk ‘dung’ -yáTes -yáTis ‘head’ béles bélis ‘sheep’ (2nd syll.) 6éndeS 6éniS ‘gold; queen’ (2nd syll.) ceréS chiríS ‘root’ (2nd syll.) behék biík ‘willow’ (2nd syll.) cel chil ‘water’ ses sis ‘person, people’ dasén dasín ‘girl’ phárcen phárcin ‘cap’ gacér gachír ‘vulture’ 24 There is a small group of cognate sets which could be interpreted as showing i / i (always in the 1st syllable) but which is not listed here. It consists of the majority of the words of a small special group which will be called the “grass group” later on in this book. This group will be discussed in 3.4. <?page no="74"?> 3. The dialects and the comparative method 74 mesqé bisqá ‘spittle, saliva’ (1st syll.) -xés- -qhís- ‘to tear’ 6en 6iín ‘thief’ More examples exist for group D. Group E barcé barcí ‘fur’ / ‘ibex skin’ haré harí ‘barley’ telé tilí ‘walnut; walnut tree’ (2nd syll.) hésa hísa ‘month’ 6amé 6amí ‘spiderweb’ Se6éSum Si6íSum ‘slimy’ (2nd syll.) thaméne thamíni ‘last year’ (2nd syll.) Group F khené khiné ‘this’ (1st syll.) helés hilés ‘boy’ (1st syll.) ceréS chiríS ‘root’ (1st syll.) behék biík ‘willow’ (1st syll.) telé tilí ‘walnut; walnut tree’ (1st syll.) Se6éSum Si6íSum ‘slimy’ (1st syll.) thaméne thamíni ‘last year’ (3rd syll.) Group G -yék -ík ‘name’ -yé -í ‘son’ buyé buí ‘shovel’ da6óye da6úi ‘unripe’ / ‘raw’ A survey of correspondences and groups is as follows: correspondence groups (Ys. / Hz.) e / e A i / i B, C e / i D, E, F, G The question is now what to conclude about the past in view of the three correspondences. It seems consequent to account for e / e by reconstructing *e and for i / i by assuming *i. This matches what has been done with the back vowels in 3.3.1. The crucial question is what lies behind e / i: <?page no="75"?> 3. The dialects and the comparative method 75 should another vowel be assumed, or can the data be reconciled with the reconstructed inventory *e and *i? Berger (2008: 8) underlines that the difficulties posed by the correspondence Ys. e / Hz. i are considerable, and he concedes that he does not know a satisfying solution. There are a couple of considerations and arguments which seem to indicate that the existence of another vowel is not likely. Firstly, the vowel system may then become asymmetric, having more front than back vowels, not counting a at the bottom of the triangle (although such vowel systems do occur). Secondly, the investigation of o / u revealed that no further vowel was needed and the situation may be similar here (although there is no guarantee that such reasoning works). These were two somewhat weaker hints. Consider now, however, a certain vowel alternation which Burushaski exhibits; in Hunza it shows up as u / i, but in Yasin one has partly u / i, partly u / e. Probably the difference in Yasin is secondary and uniformity, as in Hunza, represents the older state. This will only be possible, however, if Ys. i and e go back to one vowel in this alternation. (The vowel alternation will be treated at length in 4.2.1.) Finally, and this is the fourth hint, there are morphophonological alternations of e and i in Yasin. The noun Ys. phen ‘fly’ has the plural phí-u, and there are further nouns of this type (for the lack of n in the plural and other details see 4.2.2.). Consider the following singular / plural pairs: sg. pl. phen phí-u ‘fly’ khen khí-u ‘flea’ c6en c6í-u ‘bird’ -Sípen -Sípi-u ‘penis’ Hunza does not have such alternations of e and i; the equivalents are of the type phin ‘fly’, pl. phí-o etc., i.e. they have i in both numbers. (One could in fact add cognate sets of absolutive plural forms such as Ys. phíu / Hz. phío to group B above.) Also a few verb stems vary in Yasin: - Ys. -ser-, -sir- ‘to feed’ (data from Berger 1974: 175), cf. Hz. -sir- - Ys. 6arés-, 6arís- ‘to agree’ (data from Berger 1974: 147), cf. Hz. 6arís- - Ys. daldén-, daldín- ‘to sieve’ (data from Berger 1974: 141), cf. Hz. daldín- The Yasin alternations may have something to do with the problem under study. It would be most elegant to kill two birds with one stone and explain <?page no="76"?> 3. The dialects and the comparative method 76 both the alternations within Yasin and the correspondence Ys. e / Hz. i with one sound law, if possible. If it should turn out to be correct that no additional vowel needs to be assumed, there are also various hints which vowel probably lies behind the correspondence Ys. e / Hz. i: it may be *i. Firstly, since o / u led to *u, for e / i one would suspect *i. Secondly, for the alternation Hz. u / i, Ys. partly u / i, partly u / e addressed above one expects originally the same degree of openness, and hence e < *i would be required for Yasin u / e. Moreover, when there are morphophonological alternations within Yasin of e and i, Hunza has i exclusively, as has been seen. Finally, there is at least one loanword which indicates that some of Yasin’s e go back to *i: Ys. teSk / Hz. tiSk ‘dagger’ (in Ys. also ‘knife’) is from Sanskrit t í kSnaH ‘sharp’ with a metathesis of kS (Berger 2008: 8, 61). Therefore *i is reconstructed for the correspondence e / i. In Hunza the vowel always remained i. In Yasin it was lowered under certain conditions to e, and hence the correspondence e / i arose, but it remained i when the conditions did not apply, producing the correspondence i / i. The question is what these conditions were. Berger (2008: 8-10) attempts to explain the lowerings with neighbouring consonants, but no system appears. It is instructive to look at lexemes again which show morphophonological alternations, e.g. ‘fly’. Since phen ‘fly’ underwent *i > e, whereas phíu ‘flies’ did not, there must be something in the shape of the two forms which caused the difference. There are not many possibilities. One point is that with phen the syllable is closed, whereas with phíu it is open. (Closed syllables end in a consonant; open syllables end in a vowel.) It is a frequent experience that sound change depends on whether syllables are closed or open. The hypothesis is therefore that in Yasin *i underwent lowering to e in closed syllables (and in open syllables i remained). This law shall be called closed law. The hypothesis turns out to be a very fruitful idea. For the data base the closed law implies that the large group B is as expected, since it has Yasin i in open syllables. Also the large group D is regular, since it has Yasin e in closed syllables. The smaller group C, however, presents a problem, since it has Yasin i in closed syllables, and so does the smaller group E, which has Yasin e in open syllables. (The small groups F and G will be addressed later.) For solving the problem attention must be drawn to this: what such lists of cognate sets usually only show is one form per lexeme (e.g. ab- <?page no="77"?> 3. The dialects and the comparative method 77 solutive singular for nouns); there are however whole inflectional paradigms which are behind the entries. In many paradigms there is variation as to whether the decisive syllable is open or closed. In ‘fly’ and similar nouns, as already seen, the singular has e and the plural has i. In many other paradigms, in contrast, there is only one vowel, and this may mean that a well-known phenomenon of language history was at work: analogy (leveling). In fact, analogy was already addressed in 3.3.1. in order to account for the vocalism of the word ‘snake’. Analogy can now be invoked again when considering group C. There are, for example, the following two entries: Yasin Hunza -híl -íl ‘lip’ -díl -ndíl ‘breast (female)’ The Yasin forms do not show what the closed law would demand. However, consider the corresponding plural forms (data from Berger 1974: 142, 151, Berger 1998III: 212, 302): Yasin Hunza -híl-i9 -íl-i9 ‘lips’ -díl-i9 -ndíl-i9 ‘breasts (female)’ In the plural forms, the i of the first syllable was in an open syllable, and hence it remained. (The hyphens indicate morpheme boundaries, not syllable boundaries.) The closed law may have produced the singulars *-hél and *-dél but their vowels were replaced by the i that occurred in the plural. This is the decisive mechanism that produces some of the entries in group C which show a closed syllable, but i nevertheless. In the examples just mentioned the plural “won”, so to speak, in the leveling process. It may be asked now whether there are also instances in which the singular asserted itself, and indeed this occurred as well. For example, Ys. -ken ‘liver’, from group D, has the plural -ken-i9, and it can be seen that the singular is as expected considering the closed law, whereas the plural is not. The plural must be from earlier *-kin-i9 (or *-kim-i9; on the nasals see 2.2.1.) and must have taken over the vowel e from the singular. A critic may now ask this: given that both directions are possible for the takeovers, does this not create a situation in which too much of the data will be easy to explain, and this means that the scenario loses its <?page no="78"?> 3. The dialects and the comparative method 78 explanatory force? For certain reasons this is not the case. It is necessary to have a look at the semantics, and this leads to an important insight. The words for ‘lip’ and ‘female breast’ are very often used in the plural, since there are usually two of them, and hence it is natural that the singular, being of less frequent occurrence, adapts to the plural. The word for ‘liver’, in contrast, is more often encountered in the singular, and hence the plural may very well adapt to it. In sum, the directions of the analogies which were assumed are not arbitrary but they fit the meanings of the words and the frequency of the two numbers for each meaning. Let us look at two further examples. Ys. and Hz. 6aCíS ‘awn, beard’, from group C, has the plural Ys. 6aCíh-a9, Hz. 6aCí-a9 (or newer, with an assimilation: 6aCá-a9). In this item the plural gained the upper hand in Yasin; this makes sense since one usually speaks about many awns (a term from botany). In contrast, the singular won out in the 2nd syllable of 6éndeS ‘gold; queen’, pl. 6éndeha9, from group D, since this word was certainly used much more often in the singular than in the plural. There is an investigation by Tiersma (1982) which fits the results arrived at here for Burushaski. Using data from West Frisian and other languages, Tiersma demonstrates that sometimes the singular and sometimes the plural dominates in analogical processes, and which is the case depends on the meaning of the noun and on which number is used more frequently. Sihler (2000: 76) discusses some data from English. He shows that in the singular / plural pairs day / days, glass / glasses, mother / mothers, the singular is historically regular and the plural is analogical (which becomes apparent when considering Old or Middle English). In contrast, glove / gloves is based on a historically regular plural for which a new singular has been formed. This has undoubtedly got to do with the fact that gloves usually occur in pairs and mostly both gloves are spoken about. See also Haspelmath (2002: 246) for a discussion of Tiersma’s findings. Analogy will not be able to explain all data in group C; there is e.g. a problem with ‘mountain’ which has a plural form in which the syllable is closed, too: Ys. CiSkó, Hz. ChiSkó. But if a few problematic items remain this must not prevent one from assuming the sound law and defending the overall approach. 25 The group with ‘fly’ still shows the morphophonological alternation, which is strong evidence for its former existence in the 25 Moreover, the word for ‘mountain’ exists in Shina as well (Berger 1998III: 100) and hence this may be a loanword in Burushaski. <?page no="79"?> 3. The dialects and the comparative method 79 Yasin dialect. The reason why the alternation was preserved in this group of words is that singular and plural had gone their separate ways anyway, i.e. also in other respects (see 4.2.2.). In other nouns, in contrast, analogy could apply more easily. Possible explanations for two further items shall be added here. Ys. díltar ‘butter-milk’ may have its i because of -díl ‘breast’. Ys. hir ‘man’ may have i because of the indefinite form híren ‘a man’ (the plural cannot be used since it is hurí with u); it may occasionally be relevant that Burushaski nouns do not only inflect for number but can attach also other suffixes (also case suffixes). Finally, we do not exactly know which words with i may be loanwords in Yasin and do not show the lowering for this reason. Group E must be assessed now. In order to understand several sets in this group it is essential to know a peculiarity of a certain inflectional type in Hunza: some nouns, when forming their plural in -9, replace the finals -í and -ú with -éand -órespectively. The examples that Berger (1998I: 54) adduces are basí ‘garden’, pl. basé9, gundarú ‘kind of stone fruit’, pl. gundaró9. It turns out that the first three sets of group E form exactly such plurals; barcí ‘ibex skin’, for instance, has barcé9 (Berger 1998III: 39). Now the plural of Ys. barcé ‘fur’ is barcé9 (Berger 1974: 132); as can be seen, the plural forms of Ys. and Hz. are identical. We can assume now that the inflectional type existed already in Proto-Burushaski, and later the Yasin singular was influenced by the plural (i.e. analogy is involved again). In the same way Ys. haré ‘barley’ and telé ‘walnut tree’ received -é from their plurals haré9 and telé9 (Hz. haré9 and tilé9). Yasin abolished a rare inflectional type through the procedure. The remaining sets of group E have different causes, not all of which can be detected. Ys. hésa ‘month’ probably received e because it is a compound with he-k ‘one’ (see 5.2.). As to Ys. 6amé ‘spiderweb’, a word from Berger (1974: 147) and repeated by Berger (2008: 10), it is interesting to learn that Morin / Tiffou (1989: 26) found 6amí in their fieldwork; this would remove the set from group E. Ys. Se6éSum ‘sticky, slimy’ / Hz. Si6íSum ‘smooth, slimy’ may show onomatopoetic influence, and the last set I cannot explain. Group F is easy to explain: it shows assimilations in vowel height which work in the same way as those affecting *u (see 3.3.1.). If the following or the preceding syllable has e, an *i is lowered to e in Yasin. There is thus not only the closed law, but there are the assimilations as well which constitute another sound law. In relative chronology the assimilations follow the closed law, as evidenced by most Yasin words in the <?page no="80"?> 3. The dialects and the comparative method 80 group; a prerequisite for the assimilation in Ys. ceréS ‘root’, for instance, is that the vowel in the 2nd syllable has already become e, and for this the closed law is required. There are apparently also sometimes assimilations which imply *e > i. Besides cendí ‘five’, for example, there is also a form cindí. This may explain the 1st syllable of gírkis ‘rat, mouse’ and girmín- ‘to write’ in group C (also iskí ‘three’, not in the lists). On girmínone must bear in mind that it denotes a cultural achievement which may have been designated with a loanword. Morgenstierne (1935: XXI, n. 3) points out that girmín- ‘to write’ reminds of Armenian gir ‘writing’ and Finnish kirja ‘book’ (cf. kirjoittaa ‘to write’) but speaks of coincidence. If it should not be coincidence, Burushaski has a loanword here (and possibly Yasin and Hunza acquired it separately). The tiny group G is difficult to judge. The closed law could account for only one word, namely -yék ‘name’; the other words have e in an open syllable. The assimilations are not applicable. It may be that this group had the vowel *i, but *e cannot be ruled out either; in the latter case one can assume *e > i after *y in Hunza. Interestingly, in a 19th century source, Cunningham (1854), “goyak” (i.e. guyék) is written for what today is Hz. guík ‘your name’ (Berger 1998III: 3, 211). Since the group comprises so few sets the pressure to find further explanations is not high. To sum up, the investigation has arrived at two laws in the Yasin dialect which affect *i and result in e. Their relative chronology is known as well. The following diagram shows the developments: Proto-Bur. *e *i Proto-Bur. *e *i *e *i *e *i Yasin e i Hunza / Nager e i The first oblique line represents the closed law, the second oblique line represents the assimilations. Finally, there are some further facts to report about e and i in Burushaski. As already briefly mentioned toward the end of 3.3.1., some affixes have allomorphs with these two vowels in all dialects, so that <?page no="81"?> 3. The dialects and the comparative method 81 already Proto-Burushaski had this state of affairs. For instance, there is a suffix -i / -e in many numerals (see 4.3.1.), and there are personal prefixes i- / ewhich indicate 3rd pers. sg. of most genders (all except feminine) and 3rd pers. pl. of gender y. 3.3.3. Other topics There is a correspondence of Ys. e and Hz. a, although only few examples can be found: Yasin Hunza -xórpet -qhúrpat ‘lung’ 6óqares 6ókuras ‘raven’ -wáldes -wáldas ‘back’ (noun) papóres huk bapúras ‘boil, furuncle’ (huk ‘dog’) -qetara9 -qat ‘armpit’ jímele jímale ‘tomorrow’ (2nd syll.) -en -an verb suffix (more in 2.3.4.) -en- -ana suffix for forming verb stems sénde sínda ‘river’ mesqé bisqá ‘spittle, saliva’ Cardé Chardá ‘stallion’ With the word ‘tomorrow’ there is some variation in the sources. Ys. jímele is quoted from Tiffou / Pesot (1989: 141) and Tiffou (1999: 279) here; Berger (1974: 155) has jímale instead, and also Lorimer (1962: 142) has a, which would make the cognate set unusable. 26 The correspondence Ys. e / Hz. a must have a historical cause. It would be impossible to reconstruct *a since this clearly produces the frequent correspondence Ys. a / Hz. a. An attempt with *e can be made then. What is relevant now is that a long list for Ys. e / Hz. e was given in 3.3.2. (group A), and additions could be made. An important observation is, however, that e in 3.3.2. is almost always stressed. In contrast, e in the above list is almost never stressed. Only the last two words are exceptions, 26 There is an even more problematic word which has not been taken up into the list: Ys. gónSere / Hz. gónSare, góinSare ‘the whole night’ (2nd syllable). This information is from Berger (1974: 146) (Yasin) and Berger (1998III: 157) (Hunza) but other authors report other data. Both Tiffou / Pesot (1989: 138) and Morin / Tiffou (1989: 24) have gónSar ‘toute la nuit’ / ‘toute la nuit durant’ with a for Yasin, and Willson (1999: 57) has gónSere ‘at night’ with e for Hunza. <?page no="82"?> 3. The dialects and the comparative method 82 and of them Ys. Cardé / Hz. Chardá ‘stallion’ is a loanword from Turkic, although this fact alone does not yet clarify the correspondence, and ‘spittle, saliva’ remains a problem. The suspicion arises that possibly Hunza shifted unstressed *e > a. Only a small adjustment has to be made now. There is a small number of sets in which Ys. e / Hz. e correspond in unstressed syllables: Yasin Hunza jímele jímale ‘tomorrow’ (3rd syll.) -e -e locative suffix (more in 4.4.1.) -e -e ergative suffix -e -e genitive suffix (except f. sg.) The word ‘tomorrow’ probably contains the locative suffix -e; it is listed nevertheless because it must have been an independent word already when the dialects split up. 27 It seems that in final position the shift *e > a did not operate. This entails one exception, namely ‘river’, but this may be a loanword, cf. Shina sin (Berger 1998III: 379), and see the discussion in 3.2.2. No complete certainty is possible, but the tentative sound law is: unstressed *e > a in non-final position in Hunza. This is exactly what Berger (2008: 7) claims; see also Berger (1998I: 17). Additional evidence may stem from Hz. búran ‘seam, hem’ which has no Yasin equivalent but which in Shina is búren (data from Berger 1998III: 64). The Nager dialect preserves e at least in the verb suffix -en, as can be seen from the text presented by Skyhawk (1996). Yasin au corresponds to Hunza oo in a number of words (and in some grammatical forms as well): Yasin Hunza 6aún 6oón ‘melon’ thaún thoón ‘coriander’ haúlal hoólal-as ‘butterfly’ aúSin oóSin ‘guest’ Saú Soó ‘kind of rose’ Saúq Soóq ‘noose’ 27 Also the word for ‘the whole night’ could be listed here because of its 3rd syllable, and it too may contain the locative suffix -e. <?page no="83"?> 3. The dialects and the comparative method 83 Since a monophthongization *au > oo ( ó ) is known from many languages, e.g. Sanskrit and Tocharian A, it should be assumed here that Yasin represents the older state, and Hunza innovated. The opposite phonetic development is hardly possible. Moreover, as already explained in 2.3.1., ‘melon’ is a loanword from a Turkic language, cf. Turkish kavun ‘melon’, and the validity of the sound law is underlined again by the fact that this word shows a and u. The Nager dialect partly preserves au, partly it has oo. The front vowels provide a parallel, but the situation is less clear. Yasin ai sometimes corresponds to Hunza ee, as in this cognate set (data from Berger 1998III: 377): Yasin Hunza sái sée ‘chaste’ Here Hunza underwent the monophthongization *ai > ee (also known from Sanskrit, Tocharian A and other languages). Note, however, the following sets (the first two are homophones): Yasin Hunza bái bái ‘is’ (m.) bái bái ‘winter’ bihái biái ‘illness’ Why these exceptions exist remains unclear. In contrast, Hz. geéltin ‘ankle’ does show the innovation, cf. Ng. gaíltin, gaítin (Berger 2008: 14). Also the Hunza member of the following cognate set may have undergone the sound law: Yasin Hunza mayén meén ‘old’ Hz. meén could be traced back to an earlier *maín. From here on the loss of *y before i in Hunza can be made use of, cf. 3.2.3., and on the vowel in comparison to Yasin see 3.3.2. In the following set, mentioned already in 3.2.3., *h was regularly lost in Hunza / Nager: Yasin Hunza bahúr boór ‘salty earth’ <?page no="84"?> 3. The dialects and the comparative method 84 There are also some grammatical forms which show loss of *h and monophthongization, see 2.3.1. It is possible to venture some statements about relative chronology. Apparently two Hunza / Nager consonant sound laws precede the monophthongizations. The following diagram illustrates this: *-h- > ∅ *y > ∅ before i monophthongizations This relative chronology also becomes clear from the fact that not all subdialects of Nager underwent the monophthongization whereas they all exhibit the consequences of *-h- > ∅ and *y > ∅ before i. Sometimes Hunza / Nager has an initial unstressed i-, while Yasin has no vowel at all: Yasin Hunza ne iné ‘he’ (Hz. also ‘she’) 28 se isé ‘it’ (gender x) There are more pronouns of this type. It is obvious that the presence of the vowel must be reconstructed here: *i-. The dropping of unstressed *iin Yasin has consequences for the morphology. There were many verb forms with a prefix *iwhich stands for the 3rd person singular of genders m, x and y and the 3rd person plural of gender y. In Hunza the prefix is still present. In Yasin, in contrast, a form such as wál-i ‘he fell’ / ‘it fell’ / ‘they fell’ (gender y) is from *i-wál-i (cf. 2.3.1.), and this example is representative for many other verb forms. Possessor marking is affected in the same way, see 3.7.3. The dropping of an unstressed *ican also be observed in West Slavic, cf. the following data with Polish as a West Slavic language and Russian representing a different branch of Slavic (East Slavic): Polish Russian mieC  [i'm∆et∆] ‘to have’ graC  [i'grat∆] ‘to play’ Sometimes the term “apheresis” is used for the dropping of initial vowels. 28 In Yasin ‘she’ is mo. <?page no="85"?> 3. The dialects and the comparative method 85 3.4. The “grass group” There are a number of cognate sets which I propose to call the “grass group” since the word for ‘grass’, Ys. iSqá / Hz. Siqá, is a typical representative. The challenge these sets provide can hardly be classified as a vowel problem or as a consonant problem; it affects the order of certain vowels and consonants. The characteristic of the cognate sets in question is that Yasin shows the sequence vowel + sibilant, whereas Hunza / Nager shows the opposite order, i.e. sibilant + vowel (usually k or q follows): Yasin Hunza iSqá Siqá ‘grass’ iSqám Siqám ‘green, blue’ iSqár Siqár ‘wasp’ (Ng. Siqál) iSkárk Sikárk ‘yellow’ isqá siqáa ‘open’ iskím sikím ‘silk’ iSqórum Suqúrum ‘sour’ ispándur supándur ‘rue’ (a plant) The adjective ‘green, blue’ is derived from ‘grass’ (cf. the frequent adjective suffix -um), and ‘yellow’ is probably derived from ‘wasp’ (Berger 2008: 33). The derivations obviously had already occurred in Proto- Burushaski, however, and need not concern us here. (The connection ‘wasp’ - ‘yellow’ is one of the few cases in which a velar and a uvular alternate in Burushaski; normally this does not occur.) The question is what should be assumed in Proto-Burushaski. For a discussion see Berger (2008: 62). There are several options what to reconstruct. Obviously, the sequences vowel + sibilant and sibilant + vowel could be envisaged, and a metathesis must then be posited for one branch in each case. In addition, Berger (2008: 51, 62) points out that the words in question could also have begun with sibilant + plosive (mostly velar or uvular), i.e. with a consonant group, and then both dialects would have added a vowel, but in different places. The possibilities can be illustrated with the word for ‘grass’: option I: *iSqá, then metathesis in Hunza option II: *Siqá, then metathesis in Yasin option III: *Sqa, then prothesis of iin Yasin, epenthesis of -iin Hunza Advantages and disadvantages of these options must now be assessed. <?page no="86"?> 3. The dialects and the comparative method 86 It turns out that option I runs into a problem with Yasin’s “closed law” of 3.3.2., which shifts *i > e if the syllable is closed. If *iSqá really was the Proto-Burushaski form, one would not expect an iin Yasin since the syllable is closed in each example. (Only for iskím ‘silk’, one might be able to argue for an assimilation since the second syllable contains i, but this isolated possibility is of no help.) Moreover, note the following cognate set: Yasin Hunza iskí iskí ‘three’ If option I was correct, one would expect the opposite order of the vowel and the sibilant in Hunza. Hence, option I can be ruled out. Option II does not come into conflict with the “closed law” if a suitable relative chronology within Yasin is chosen: there may have been first the “closed law”, then the metathesis. However, the phonetic background may turn out to become a problem: the question is whether such types of metatheses occur at all (a criticism which similarly would be applicable to option I). This leaves us with option III; Berger (2008: 51, 62) favours it. The phonetic developments starting from *Sq-, *Sketc. are not problematic: both prothesis and epenthesis are understandable from a phonetic point of view, they have typological parallels and even areal parallels in this region. (For an epenthetic vowel, for instance, cf. Persian set á re ‘star’ with setfrom earlier st-.) Option III would entail that Burushaski had initial consonant groups when it split into dialects. The question must be addressed whether this is possible and likely, since initial consonant groups are not typical for this language. Any dialect today has only few instances of initial consonant groups, and then usually the word can easily be identified either as a loanword or as onomatopoetic (cf. any vocabulary or dictionary). Also to Proto-Burushaski a similar state of affairs could have applied. On the other hand there is no need that this was the case. Note now the verbs Ys. di-Sqín- ‘to grow (of plants)’ / d-Sqin- ‘to grow’ (tr.) (Berger 1974: 179), in which -inis a suffix, which are connected with Ys. iSqá / Hz. Siqá ‘grass’ (and with Ys. iSqám / Hz. Siqám ‘green, blue’). Similarly, the verbs Hz. du-Sqúr- ‘to become sour’ / d-Squr- ‘to make sour’ (tr.) (Berger 1998III: 410) are connected with Ys. iSqórum / Hz. Suqúrum ‘sour’ (-um is a suffix, see also 4.4.2.). The verb stems do not contain vowels before the S or between the S and the q. This is possible <?page no="87"?> 3. The dialects and the comparative method 87 since there are always prefixes, in any verb form, which provide vowels (the d-prefix or personal prefixes), and the interior of words permits consonant clusters in Burushaski still today. These verbs are strong evidence that the vowels in the “grass group” originally did not exist. 29 However, the verbs do not provide proof that Proto-Burushaski still had the initial groups *Sqetc. at that time when it split up. There is an alternative: already before Burushaski split into dialects, it may have developed either prothetic vowels or epenthetic vowels, and language history would then continue with option I or with option II respectively. In these cases the synchronic state of Proto-Burushaski would show internal *-Sqbut initial *iSqor *Siq-, which is of course not impossible, and internal reconstruction applied to Proto-Burushaski could then reduce this variation to earlier *Sq in all positions. Nevertheless, since option I and option II have certain weak points, as discussed above, it seems most likely that Proto-Burushaski still had the initial groups *Sqetc., i.e. that option III applies. Burushaski may have been within a process of abandoning its initial consonant groups, and some groups may still have existed before it split up into dialects. Moreover, the fact that in some sets, e.g. Ys. iSqórum / Hz. Suqúrum ‘sour’, the vowels (i vs. u) are not identical speaks against any metathesis; it speaks for option III. 3.5. Suprasegmentals The discussion will address stress, quantity and tone. Stress is on the same syllable within a cognate set, and one can take almost any example with two or more syllables in order to illustrate this, e.g. Ys. mathán ‘far’ - Hz. mathán ‘far’. Only very rarely exceptions can be found, e.g. Ys. ga9gí ‘axe’ - Hz. gá9i ‘axe’ (Berger 2008: 42, but see 3.2.2.), Hz. Sirijón ‘mushroom’ - Ng. Siríjon (Berger 1998III: 395). Since these are only sporadic examples, there is certainly no sound law in Yasin, Hunza or Nager which shifted stress. Proto-Burushaski simply had stress on that vowel where it is also found in the dialects today. Only for a few words such as ‘mushroom’ it may be difficult to tell where stress was originally. 29 By the way, in my opinion the noun Ys. asqór / Hz. asqúr ‘flower’ possibly belongs to the verb root ‘to grow’ as well despite the difference in the sibilant, which may be secondary. This connection is not mentioned by Berger (1998III) or Berger (2008). For the semantics cf. Albanian bimë ‘plant’ from a root which meant ‘to grow’ (Holst 2009: 95), Hungarian növény ‘plant’ from nÖ ‘to grow’ and Finnish kasvi ‘plant’ from kasvaa ‘to grow’. <?page no="88"?> 3. The dialects and the comparative method 88 The subdialect of Srinagar deviates more radically: stress has often been shifted to the first syllable, though details remain unclear, and subsequently a syncope has sometimes taken place (Munshi 2010: 52f.). Munshi accounts for these observations with language contact, which is certainly correct. Since this variety is historically only a subdialect of Nager, its deviating stress positions are clearly innovative and do not have any bearing on the consideration of Proto-Burushaski stress. Finally, the prehistory of Burushaski stress is a different matter which will be treated in 4.5. Quantity is not always an easy topic. Hunza / Nager has two quantities: short and long, and they occur with the vowels. (Consonants cannot be long in Burushaski.) Long vowels have been written in several ways in Burushaski studies, e.g. as aa, á , â or a.; at present double vowels are the notation that is preferred most often. Long vowels are much less frequent than short vowels. In Hunza / Nager the long vowels fall into two types if stressed: some have stress on the first element, some on the second, i.e. áa, aá. 30 The first type is more frequent. The second type sometimes arose by loss of *-h- (3.2.3.) or by monophthongization in the case of oó and eé (3.3.3.), but there are also a few words in which possibly neither origin applies, e.g. Hz. -yeéc- ‘to see’, cf. Ys. -yéc-; here an original short vowel in the root is more likely, and note the negative forms with -íc- (see 2.3.1.). In Yasin not even the synchronic facts are entirely clear; it is unclear whether long vowels exist (Berger 2008: 5). If they do (this is what many sources suggest), they are even rarer than in Hunza / Nager, and they do not show the distinction áa / aá. An example is Ys. háale ‘at home’ which is from *ha-ul-e (ha ‘house’ + suffixes, see more on these in 4.4.1.). In Burushaski, long vowels usually do not occur in unstressed syllables. In the Nager dialect, however, they sometimes do so because of morpheme boundaries: Ng. hísaan ‘a month’ consists of hísa ‘month’ and the indefinite suffix -an; Hunza has shortened the long vowel here: hísan ‘a month’ (Berger 1998I: 17). The most important diachronic point about quantity is that long vowels in Hunza / Nager usually correspond to short vowels in Yasin, as in: 30 This is apparently not unusual from a typological point of view. A parallel is provided e.g. by Somali in which long vowels can have stress on the first or on the second element as well: Somali waláal ‘brother’ / walaál ‘sister’. <?page no="89"?> 3. The dialects and the comparative method 89 Yasin Hunza tórum tóorimi ‘ten’ thánum tháanum ‘high’ -thé9gi -thée9i ‘dowry’ kharú kharúu ‘louse’ mac6í mac6híi ‘honey’ balt báalt ‘apple’ bardum báardum ‘red’ isqá siqáa ‘open’ -Sír- -Síir- ‘to take away’ 6on 6úun ‘quail’ The last set exhibits Yasin’s uvular law. The long vowels of Hunza / Nager probably represent the older state. Some of these words may be loanwords, though not all of them. As regards tone, a long time ago Lorimer (1935a: 13) wrote in a very straightforward manner: “I did not perceive tones in the language, but that is no guarantee that they do not exist.” Nowadays one can be more specific. Tone is not present in Yasin, and this also applied to Proto-Burushaski. In Hunza / Nager, the question is a matter of interpretation. The difference between áa and aá, etc., seen above, could be regarded as a tone difference, i.e. falling vs. rising, though the treatment as a stress difference is probably preferable. See also Anderson (2009: 176). For the areal context note the tone systems of many Sino-Tibetan languages, e.g. Modern Tibetan, and cf. what Edelman (1983: 43) writes on the Dardic and Nuristani languages: “In a number of languages (Ashkun, Dameli, Gawar, Khowar, Katarkalai, Bashkarik, Shina and, possibly, others) there are phonologically relevant syllable-tone oppositions, frequently appearing as the result of loss of aspiration in the voiced aspirated consonants of the syllables.” 3.6. Vocabulary Sometimes the vocabularies of the Burushaski dialects diverge. The word for ‘ant’ is pilíli in Yasin (Berger 1974: 169) and khon in Hunza / Nager (Berger 1998III: 256). The structure of the Yasin word, with its three open syllables, is not very typical for inherited words; besides the word exists also in Khowar (Lorimer 1962: 188, Berger 1974: 169) and in Shina (Backstrom / Radloff 1992: 334). The Hunza / Nager word, in <?page no="90"?> 3. The dialects and the comparative method 90 contrast, has a plural khóyo which represents an old formation, see 4.2.2., and apparently there is no neighbouring language with the same word: Berger (1998III: 256) does not mention one, and there are different words in Balti (Backstrom / Radloff 1992: 225) and in Wakhi (Backstrom / Radloff 1992: 278). One may conclude therefore that Proto-Burushaski exhibited the word which is now limited to Hunza / Nager, and Yasin acquired a loanword. With ‘kidney’, in contrast, no decision seems to be possible on which dialect has the older word: Ys. -riC ‘kidney’, Hz. -so ‘kidney’. It is more typical, however, to find etymologically identical words in Yasin and Hunza / Nager, and this applies especially to core vocabulary. In the following, Swadesh 100 word lists are given for Yasin and Hunza. The main source for Yasin is Berger (1974); the main source for Hunza is Berger (1998III). A hyphen indicates that no suitable word was found (not even in other sources). Sometimes synonyms or near-synonyms exist, and in such cases choices had to be made; then usually these were made in such a way that cognate sets were arrived at. This entails that the lists would not be suited for a lexicostatistical evaluation since this would make the two dialects seem slightly closer than they actually are. This is not a problem, however, since there is general agreement nowadays that glottochronology cannot provide us with reliable information about the time of separation anyway. There are various treatments of what Swadesh lists can and cannot do; for a recent one see Holst (2011: 169). The purpose here was simply to have some vocabulary to work on. In the last column information about the Proto-Burushaski reconstruction is added, giving an insight into this stage of the language and showing how the sound correspondences and sound laws treated in 3.2. to 3.5. are made use of in practice. The entries are organized as follows: = Hz. Proto-Burushaski reconstruction identical with Hunza = Ys. Proto-Burushaski reconstruction identical with Yasin = Proto-Burushaski reconstruction identical with both Sometimes no information on the reconstruction is given; this can apply e.g. when a sound correspondence is irregular and the older state difficult to decide, or when the Yasin word and the Hunza word are not cognates. In many cases further explanations are given. <?page no="91"?> 3. The dialects and the comparative method 91 Yasin Hunza Proto-Bur. 1. all har har = Loanword from Urdu. There is also Ys. khul / Hz. kul, also a loanword from Urdu. 2. ashes phét-i9 phét-i9 = A plural form like English ash-es. 3. bark waT waT = From baT ‘skin’ with a mutation, see 2.2.2. 4. belly -úl -úl = 5. big -nyú uy-úm Berger (1998III: 460) apparently regards the words as related. For his argumentation see Berger (2008: 67). 6. bird c6en c6hin = Hz. 7. bite gaTgaT- = 8. black mat-úm mat-úm = 9. blood multán multán = 10. bone ten -ltín, tin = Hz. Contains fossilized dual prefix, cf. 4.3.2. See also 5.3. In Burushaski medial -ltand initial talternate; the cluster lt does not occur in initial position. This alternation will be seen also elsewhere. 11. breast -díl -ndíl 12. burn 6ól- -6úl- = Hz. This is the intransitive verb. The transitive verb is Ys. and Hz. -squl- (the causative, with mutation). 13. claw -úru Ca9 The Hunza word is a loanword from Persian (Berger 1998III: 84). Ys. -úru also means ‘fingernail’ and is internally connected to Ys. -úri ‘peak, top, summit’, which in turn is cognate with Hz. -úri ‘fingernail’. These all are forms with approximant mutation (*b > *w > ∅ ); the unmutated form is shown in Hz. buúri ‘crest, peak, top’ (Berger 1998III: 66). 14. cloud xoró9 qhurónc6 *qhuró- The Yasin word is a plural ‘clouds’ (Berger 1974: 160). Nager has qhuróni9 (Berger 1998III: 359). The suffixation varies. 15. cold Ca6úr-um Cha6úr-um = Hz. From the verb Ys. du-Cá6ur- / Hz. du-Chá6ur- ‘to cool down’. <?page no="92"?> 3. The dialects and the comparative method 92 16. come d-yad-ya- = Suppletive; there are also other stems. 17. die -yúr- -ír- = Ys. See more in 4.2.1. 18. dog huk huk = There is also a word Ys. kukúres / Hz. gukúrus, gukúro ‘young dog, little dog’. In my opinion both are loanwords from languages of India, cf. Sanskrit kukkuraH ‘dog’ (itself of non-Indo-European origin), kurkuráin the Atharva-Veda (an old text), Tamil kukka ‘dog’. Ultimately possibly onomatopoetic, relating to the barking of dogs. 19. drink menmin- = Hz. 20. dry béi qhaáo The Hunza word means ‘dry’ only approximately; Berger (1998III: 348) translates it with “verdorrt, geröstet”. 21. ear -ltúmal -ltúmal = Contains fossilized dual prefix, cf. 4.3.2. 22. earth tik tik = 23. eat Sé- Sé- = 24. egg Ti9án Ti9án = The noun phrases (or compounds) Hz. Tí9-e waT and Tí9-e puSi-9óro ‘eggshell’ (-e genitive suffix) reveal that -án is a former suffix; hence the original word for ‘egg’ is *Ti9 (Berger 2008: 53). 25. eye -lCi -lCin = Hz. Stem with final *m, see 2.2.1., and the vowel was *u, see 4.2.1. Contains fossilized dual prefix, cf. 4.3.2. 26. fat (grease) bes bis = Hz. 27. feather phol6ó phul6úuy = Hz. 28. fire phu phu = Berger (2008: 129) claims that this item is onomatopoetic, coming from the gesture of blowing. This is not convincing, however. With the same reasoning French feu [fP] could be regarded as onomatopoetic if French was a language isolate; the word is, though, a regular outcome of Latin focus. Short items with the structure CV are commonplace in Burushaski: sa ‘sun’, ha ‘house’, tha ‘hundred’, du ‘kid’ (animal), gu ‘soot’, etc. 29. fish Cúmu Chúmo = Hz. For an argumentation that the last vowel is a former plural suffix see 3.7.2. <?page no="93"?> 3. The dialects and the comparative method 93 30. fly du-wáldu-wál- = 31. foot -húT-is -úT *-húT In Yasin, -is is a suffix, cf. the plural Ys. -húT-i9. There is also a Hunza form -úT-is ‘foot’, pl. -úT-i9. 32. full hek hik = Hz. Probably same as ‘one’, see 5.2. 33. give -Cí- -Chí- = Hz. The verb is suppletive, and the other stems form cognate sets as well: Ys. -6ón- / Hz. -6ún-, Ys. -ú- / Hz. -ú-. 34. good Suá Suá = 35. green iSqám Siqám *Sqam Derivative of *Sqa ‘grass’, see 3.4. 36. hair bur bur = 37. hand -rén -ríi9 For further analysis see 3.7.2. 38. head -yáTes -yáTis = Hz. Ys. -es / Hz. -is is a suffix, see 4.4.1. There is also a word Ys. kapál / Hz. gapál, kapál ‘head, skull’ from Indo-Aryan. 39. hear d-yald-yal- = 40. heart -s -s = 41. horn tur -ltúr, tur = Hz. Contains fossilized dual prefix, cf. 4.3.2. 42. I ja je = Ys. The vowel correspondence is irregular; it appears in only very few further items: Ys. ka / Hz. ke ‘and’, as well as the comitative / instrumental case suffix Ys. -i9a / Hz. -a9e. The conjunction ‘and’ is in my opinion etymologically identical with the postposition Ys. and Hz. káa ‘with’, and then a would be the older vowel. Similarly, the ergative of ‘I’ is Ys. and Hz. jáa < *já-e, and this speaks for a as older also in ‘I’, but see Anderson (2007: 1265) for a different analysis. 43. kill -sqan- -sqan- = The causative prefix -sis present here. Causatives for ‘to kill’ are frequent crosslinguistically, and the root then usually means ‘to die’, cf. Turkish öl-mek ‘to die’ / öl-dür-mek ‘to kill’, and see Holst (2013: 262). The root indeed occurs in Burushaski, but with different (weaker) meanings: Ys. du-6án-, d-6án- ‘to become useless, exhausted, to be unable to take any more’ (Berger 1974: 147), Hz. du-6án-, d-6án- ‘to be exhausted, used up’ (Berger 1998III: 168). Note the mutation. <?page no="94"?> 3. The dialects and the comparative method 94 44. knee -nú9us -dúmus *-dú9us The reconstruction of the first consonant is difficult; it could be *n or *d. The choice of *d here is only tentative; it has at least the advantage that a distant assimilation can be assumed, i.e. in Yasin the feature [+nasal] spread. 45. know -hénhén- = Ys. 46. leaf tap tap = Probably this noun and the verb Ys. du-ltápi- / Hz. du-ltápu- ‘to wither’ belong together (Berger 1974: 162, Bengtson 1997: 90, Berger 2008: 82). 47. lie guChá- Probably a fossilized verbal plural form, cf. Berger (2008: 117). Same as ‘to sleep’, see 5.2. 48. liver -ken -kin = Hz. Stem *-kim-; for the *m see 2.2.1. 49. long 6usán-um 6usán-um = Berger (1974: 136) lists a Yasin word burúnum “lang (Zeit)” “long (time)”. In English semantics the word long can refer both to space and to time, and many other languages are similar. Latvian, in contrast, distinguishes a spatial ‘long’, gar S , and a temporal ‘long’, ilgs (Holst 2001: 228). Possibly Yasin makes such a distinction, too. 50. louse kharú kharúu = Hz. 51. man hir hir = 52. many buT buT = 53. meat (flesh) Cap Chap = Hz. 54. moon halánc halánc = In 5.3. it will be argued that -c may be a suffix. 55. mountain CiS ChiS = Hz. Irregular vowel correspondence for unknown reason, see 3.3.2. Possibly a loanword. 56. mouth -xát -qhát = Hz. 57. name -yék -ík If *e should be the older vowel (discussion see 3.3.2.), the reconstruction would be as in Yasin. 58. neck -S -S = 59. new thoS thoS = <?page no="95"?> 3. The dialects and the comparative method 95 60. night thap thap = 61. nose -múS -múpuS = Ys. See Berger (2008: 144f.), whom I follow. Hz. muS ‘breath’ (also present in Ys. muS ét- ‘to breathe’) is probably the root for these words. Ys. -múS ‘nose, nasal mucus’ and Hz. -múS ‘nasal mucus’ have a different sibilant, probably because of a morphophonological alternation. With an unknown second element one gets Ys. -múSpuT ‘beak, bill’, pl. múSpuS-i9 ‘nostrils’ (Berger 1998III: 294); Tiffou (1999: 286) has Ys. -múSpuT ‘beak, bill’. Hz. -múpuS ‘nose’ < *-múSpuS with dissimilatory loss. 62. not be be = In verb forms a prefix ais used, see 2.3.1. 63. one he-k hi-k = Hz. This is the z form. The -k is a suffix, cf. other forms such as Ys. he-n / Hz. hi-n (genders m and f). 64. person ses sis = Hz. Also used as a plural, ‘people’. 65. rain harált harált = A derivative of the verb Ys. hariá- / Hz. hará- ‘to urinate’, cf. Berger (2008: 73, 127f.). Although the suffixation has no parallel to my knowledge, this view is possible because words for ‘to urinate’ can come from ‘to pour’: Polish laC ‘to pour’, wy-laC ‘to pour out, to urinate’. 66. red bárd-um báard-um The root is from Iranian and represents the word for ‘rose’ (earlier vard-; several Iranian languages shift v > b). 67. road (path) gan gan = In old sources there is also a form gand, see references by Berger (1998III: 145). However, no final voiced plosives are tolerated by Burushaski phonotactics, and consonant groups of nasal + plosive are unknown in final position. Therefore this was probably a mishearing or a hypercorrect form. Modern fieldwork unearths only gan, and not even the old sources agree: Biddulph (1880: xxxiv) has gun (English spelling for gan). 68. root ceréS chiríS = Hz. 69. round pinDóro biDíro From Indo-Aryan languages, cf. the reconstruction *piN Daraby Turner (1966: 462); see also Berger (1998III: 51). 70. sand sáu sáo = Hz. The last vowel is a plural suffix, see 3.7.2. <?page no="96"?> 3. The dialects and the comparative method 96 71. say sénsén- = In Hunza there is also -s-, which represents a form of the stem without any suffix. 72. see -yéc- -yeéc- = Ys. 73. seed 6onó 6unó = Hz. 74. sit hurúThurúT- = 75. skin baT baT = 76. sleep guChá- 77. small -Cu6un joT 78. smoke thas thas = 79. stand caT man- 80. star asúmun asií The words belong together, but their exact relation is difficult to come by. For Berger’s thoughts on the matter see Berger (2008: 96). In my opinion possibly a loanword from the Persian word for ‘sky’, á sem á n, with a semantic shift, and Shervin Taheri-Kutanaee arrived at this idea independently as well (personal communication). 31 I agree with Berger (2008: 96) that Ys. asúmun corresponds to the Hunza plural form asiímuc; the Hunza singular asií can be due to a reanalysis of the plural form. Several details must remain unclear. 81. stone dan dan = 82. sun sa sa = 83. swim tam dél- 84. tail -Sílan -súmal 85. that te ité = Hz. This is the form of gender y. 86. this guté guté = This is the form of gender y. 87. thou un un = Some subdialects, especially of Nager, have um or u9 (Backstrom / Radloff 1992: 260, Berger 1998I: 79, Berger 1998III: 456, Berger 2008: 39). The view has been voiced that one of these forms contains an older nasal. Note, however, the clear 31 There is also Ys. and Hz. asmáan ‘sky’ (source for Yasin: Backstrom / Radloff 1992: 247). This is no problem, however, since languages can borrow the same word twice at different times. Cf. the doublets from French and Latin in English such as frail / fragile (Potter 1950: 45f.). <?page no="97"?> 3. The dialects and the comparative method 97 correspondence between Ys. n and Hz. n. In my opinion um and u9 are secondary forms which arose by sandhi when the nasal was assimilated to the initial of the following word. 88. tongue -yú9us -úmus *-ú9us On Yasin’s y as a secondary development see 3.7.3. 89. tooth -mé -mé = 90. tree draxt tom The Yasin word is a loanword from Urdu. 91. two alt-ó alt-ó = This is the z form. 92. walk néní- = Hz. Suppletive; there are other stems as well. 93. warm (hot) gar-úm garúr-um = Ys. Nager has gar-úm (Berger 1998III: 148). There are various proofs and hints that *gar-úm is the older form. Firstly, the identity of Yasin and Nager speaks for this. Secondly, arguments come from the noun garú ‘spring’ (all dialects), probably interrelated, and the verb root -yár- ‘to become warm’ (with mutation). Thirdly, Berger (2008: 117) discovered that Hz. garúrum acquired its longer shape because of the semantic opposite Hz. Cha6úrum ‘cold’. 94. water cel chil = Hz. 95. we mi mi = 96. what bo be Yasin sometimes has be as well; Berger (1974: 26) informs us that it seems to be restricted to idiomatic expressions. 97. white bur-úm bur-úm = 98. who men men = Since m and b sometimes alternate in Burushaski (see 2.2.2., E.), ‘who’ and ‘what’ probably share the same root. 99. woman gus gus = 100. yellow iSkárk Sikárk *Skark Derivative of *Sqar ‘wasp’, see 3.4. Though there are numerous small problems affecting single cases, it is fair to say that in general it is easy to reconstruct Proto-Burushaski vocabulary items. Position after position can be observed and the sound laws of 3.2. to <?page no="98"?> 3. The dialects and the comparative method 98 3.5. can be applied. As a final example, let us consider the word for ‘lung’ which has the peculiarity that as much as three positions differ: Ys. -x ó r p e t Hz. -qh ú r p a t Proto-Bur. * -qh ú r p e t Most cognate sets are much simpler of course. The Swadesh lists contribute to the question which Burushaski dialect is more conservative. Tiffou has sometimes favoured Yasin in this respect, and Berger Hunza. At first glance it looks now as if Hunza was clearly more conservative. A number of things must be taken into account here, however. If aspirated affricates are acknowledged for Yasin (cf. discussion in 3.2.1.), some entries “= Hz.” could be replaced by “=”, e.g. the one for ‘meat’. There are no examples with Hunza monophthongizations in the data; in such cases Hunza is innovative (see 3.3.3.). There are no examples with *e > a either which is another Hunza innovation (see 3.3.3.). Only once can the Hunza / Nager loss of *-hbe observed (‘foot’), although this is a far-reaching Hunza / Nager innovation (see 2.3.1., 3.2.3.). Most important of all, the material provides a glimpse only of phonetic issues, and there are other levels of language structure all of which are likewise relevant for an overall judgement. Especially the morphology is of interest here. Tiffou (1995b: 160) opines: “I am convinced that on many points and particularly from a morphological point of view, YB [Yasin Burushaski] is more conservative than HB [Hunza Burushaski]. Hence, it is wrong not to take this Burushaski dialect into consideration in historical linguistic studies.” The Yasin plural suffix -a (for a descriptive account see Berger 1974: 17f.) may be such a conservative feature. It occurs with a limited number of nouns, e.g. designations of animals, and with the irregular adjective thoS ‘new’, pl. thu-á which is certainly old (on some details concerning thoS see 3.3.1., 4.2.3.). Hunza has different plural formations for all these words; mostly it uses more widespread suffixes, e.g. -ánc. For some other conservative features of Yasin see Berger (1974: 5). I would like to point to some generalities which may seem trivial but they are important. It is the comparative method which reveals sometimes only after painstaking work what the common proto-language (or proto-dialect) was like, and on this depends the judgement <?page no="99"?> 3. The dialects and the comparative method 99 how conservative each modern dialect is. Moreover, without any doubt every Burushaski dialect exhibits both innovations and archaisms. Finally, the Burushaski Swadesh lists provide hints that many relationship theories are unlikely to be correct. Burushaski is certainly not an Indo-European language; note also that the typological profile of the language is radically different from any Indo-European tongue. Also a comparison with Basque will become difficult since its core vocabulary is entirely different, and many other attempts fare no better. For more on the topic see 1.4. 3.7. Morphology 3.7.1. Introduction Also the morphology of Burushaski can be subjected to historicalcomparative treatments. Often the affixes are the same in all dialects. The Yasin ergative suffix -e, for instance, corresponds to the Hunza (and Nager) ergative suffix -e, and this barely requires mentioning. It would be tedious (and pointless) to go through all identical affixes here especially since the morphological system of Burushaski is so large. Things become slightly more interesting, but hardly more difficult, if the shape of a morpheme is altered by sound change. A traditional king of the Burusho is called tham in all dialects; the plural is Hz. thám-o, but Ys. thám-u with -u due to *o > u in unstressed position, see 3.3.1. Sometimes the exact shape of the d-prefix differs: Yasin Hunza do-6óndu-6ún- ‘to ripen’ do-hóndu-ún- ‘to seize’ de-ncérdi-nchír- ‘to spread’ Clearly Yasin underwent the assimilations here which were discussed in 3.3.1. and 3.3.2. after having lowered the vowel in the stem 32 . Only rarely are there different suffixes for the same purpose. A possible example is provided by the dative which has -a with an allomorph -6a after vowels in Yasin, while Hunza uses -ar; Nager has -ar and -are. Even here, however, Berger (1998III: 20, 2008: 101) thinks that the Yasin suffix is 32 The first verb lowered the stem vowel due to the uvular law, the third verb due to the closed law. The second verb, do-hón- ‘to seize, to take’, has no regular reflex for *ú in the stem; it may have ó due to analogy with -6ón- ‘to give’, to which it is a semantic counterpart. <?page no="100"?> 3. The dialects and the comparative method 100 historically identical with the Hunza / Nager suffix, and this too is an option. Morphophonological alternations that Proto-Burushaski had are sometimes still present in one dialect but absent in another as far as a particular word is concerned. This applies, for instance, to the alternations n / m (see 2.2.1.) and i / u (see 4.2.1.). The dialects do not always have these alternations in the same vocabulary items; analogy can remove an instance of an alternation from the linguistic system of a dialect. Hence, morphological change is often concerned with sound law and analogy. Morphological reconstruction constantly must refer to these two fundamental concepts. However, several researchers, among them Arlotto (1972: 163f.) and Koch (1996: 218f.), have underlined that change in morphology is more than just sound law / analogy; morphological developments are of very diverse kinds and reconstruction procedures must account for this. In the following, some such more sophisticated issues in the diachronic morphology of the Burushaski dialects are discussed. 3.7.2. Singulars from former plurals Dialect comparison occasionally reveals changes which concern the category of number. The singular and plural forms of ‘hand’ are (data from Berger 2008: 97): Yasin Hunza sg. -rén -ríi9 pl. -ré-i9 -ríi9-C-i9 The hyphens provide a synchronic analysis. The Yasin plural -ré-i9 contains the frequent gender y plural suffix -i9, and the -n of the stem earlier alternated with *-ywhich was then lost in the vicinity of an *i, cf. 4.2.2. (The plural shows lowered e only due to analogy with the singular, for such processes see 3.3.2.) Now, given the vowel correspondences Ys. e / Hz. i and Ys. i / Hz. i, the conclusion must be drawn that the Hunza singular corresponds to the Yasin plural. The only possible interpretation then is that Hunza reanalyzed the plural form as singular and formed a new plural. This is what Berger (2008: 97) indeed infers, and he points out a dialect form -ríin ‘hand’ from Hasanabad which may still contain the n of the historical singular. In Hunza the former plural suffix -i9 has been fossilized. The new Hunza plural again has -i9 as its suffix since this is the <?page no="101"?> 3. The dialects and the comparative method 101 most typical choice for gender y nouns. The intervening -Ccan be interpreted as being from the suffix -c which may be a diminutive suffix, see 5.3. Reinterpretations of a plural form as a singular form are frequent in the histories of various languages. With ‘hand’ there is a parallel in the history of Armenian. Classical Armenian had a noun %e! -n ‘hand’ with a nom. pl. %e! -k. In Modern Eastern Armenian %e! k is the singular ‘hand’ and a new plural has been formed on this basis. A whole series of body part terms as well as some other nouns underwent this development (Karst 1901: 204). In the cognate set Ys. sáu / Hz. sáo ‘sand’ the Hunza word is still a plural (Berger 1998III: 374), while the Yasin form apparently is not; the suffix involved is the very frequent plural suffix -u / -o (on phonetic issues see 3.3.1.). It is now interesting to think one step further (although we will leave dialect comparison here). The example ‘hand’ showed how Hunza / Nager turned a plural form into a singular form, thus causing a dialect difference. It could be imagined that sometimes such “number shifts” occurred already in Proto-Burushaski. Then all dialects would exhibit a form now which is historically a plural. There are indeed some nouns which make this impression. These are the forms for ‘fish’: Yasin Hunza sg. Cúmu Chúmo pl. Cúmu-yu Chúmo Hunza shows an identity of singular and plural, and this is very rare in Burushaski; usually there is a suffix (and often there is allomorphy in the root in addition, or other complications). It can be imagined that the plural suffix -u / -o is present in today’s stem for ‘fish’ in a fossilized form. This would mean that the word is historically a plural, and it would explain how the two numbers can be identical in Hunza. Yasin has formed a new plural, see Tiffou / Pesot (1989: 18). Still another argument for the analysis comes from stress, but this will only become understandable on the basis of the discussion in 4.5. What will be laid out there implies that one would expect stress on the second syllable in the word for ‘fish’ if it had always been a pure stem. But many plural forms adopted the stress of the monosyllabic singular form (already before Proto-Burushaski split up), and if ‘fish’ is a former plural, the stress can be explained. <?page no="102"?> 3. The dialects and the comparative method 102 If pressed to give an etymology for Hz. -so ‘kidney’, I would speculate that it is a former plural of -s ‘heart’; this may have something to do with the fact that in some dialects the plural forms of ‘kidney’ and ‘heart’ are homophones (Berger 1998III: 381). Another former plural may be Hz. tulpó ‘radish’, which today has the pl. tulpó-i9, Ng. dulphó ‘vegetable with edible root’, pl. dulphó-mi9 33 (data from Berger 1998III: 432). The reason is that this word must be identical with a word in the Turkic languages, as Berger (1998III: 432) points out, cf. Turkish turp ‘red radish’, karaturp ‘radish’ (kara ‘black’). The addition of an -o can be explained when the plural theory is made use of, and in fact in this case also Berger (2008: 30) mentions this idea. The peculiar correspondence of Burushaski l and Turkic r before a consonant has a parallel in Hz. yult ‘country, empire’ - Turkish yurt ‘native country’ (etymology from Berger 1998III: 479). Some caution will be necessary, however, when continuing with this line of research. There are Dardic languages in which nouns and other nominals often end in -o. Shina is of this type, for example, as can be seen from the word lists by Backstrom / Radloff (1992: 301-369). If a Burushaski word is a loanword from a Dardic language, it can end in -o simply because the donor language had this vowel, too, e.g. Hz. múlo ‘rape, taproot’ < Shina múulo ‘rape’ < Sanskrit m ú laµ ‘root’ (Berger 1998III: 293, 2008: 11). Unfortunately in linguistics an explanation which may be correct in one case may be incorrect for the very next word. 3.7.3. Some observations on possessor marking When investigating possessor marking from a cross-dialectal perspective, mostly the correspondences are straightforward, e.g. Ys. a- = Hz. / Ng. a- ‘my’. However, there are some issues, all having to do with the 3rd person, which turn out to be a little more complex and which will be discussed here. Morphological change is sometimes the consequence of phonetic change. As briefly mentioned in 1.1., inalienable nouns are obligatorily marked for possessor. An example is -Sák ‘arm’, given here with all possessors in the singular in the Hunza dialect: 33 The Nager dialect sometimes has an aspirated plosive after l where both Hunza and Yasin have no aspiration: Ys. áltar, Hz. áltar, but Ng. álthar ‘twenty’. <?page no="103"?> 3. The dialects and the comparative method 103 1st a-Sák ‘my arm’ 2nd gu-Sák ‘your (thy) arm’ 3rd m i-Sák ‘his arm’ f mu-Sák ‘her arm’ x i-Sák ‘its arm’ y i-Sák ‘its arm’ There is no possibility to say ‘arm’ without any possessor, and there is usually no need in speech to do so either since every arm belongs to someone or something; the inalienable nouns usually are body part terms, kinship terms or some others such as e.g. ‘name’. If really in need to refer to an arm without specifying the possessor, the prefix of the 1st person plural mican be used, i.e. literally ‘our’ is employed; for an example in a text see Tiffou (1995a: 161). As a consequence of the typological facts the stem -Sák never occurs on its own. (Also listing the plural forms would not change this.) The paradigm given from Hunza must also be assumed for Proto-Burushaski in exactly this way. Now what shall be considered is the development in the Yasin dialect. In 3.3.3. the sound law that unstressed *iis dropped in Yasin was presented. This results in the following Yasin forms: 1st a-Sák ‘my arm’ 2nd gu-Sák ‘your (thy) arm’ 3rd m Sak ‘his arm’ f mu-Sák ‘her arm’ x Sak ‘its arm’ y Sak ‘its arm’ Note that the form Sak still means ‘his arm’ or ‘its arm’, and not ‘arm’ in general. The difference is, however, that in Yasin the stem does occur on its own in the paradigm, and this is a crucial morphological change due to phonetic change. Sometimes a dialect difference which cannot be accounted for by phonetic change, or at least not by phonetic change alone, finds an explanation when considering morphological change. This applies e.g. to this cognate set: Hz. imáguTi - Ng. mákuTi ‘wooden cupboard’ Yasin drops initial *i-, but Nager does not. In this case the solution is that Hunza exhibits a former personal prefix i- ‘his’ which has become part of <?page no="104"?> 3. The dialects and the comparative method 104 the stem; this is an instance of fossilization (Berger 2008: 69). Another cognate set shall be considered: Ys. -hálmun - Hz. -yaálmun ‘rib’ The initial consonants of these stems certainly do not match. What we do know is the law *-h- > ∅ in Hunza / Nager (3.2.3.). This would lead to Hz. *-álmun. Again it is necessary to bring the prefix iinto play, and Berger (2008: 26) does so. Proto-Burushaski *i-hálmun ‘his rib’ leads to Hunza / Nager *i-álmun, and it must now be assumed that in a sequence of two vowels the first one becomes an approximant: *y-álmun. As to the aá instead of *a (and similar cases), an explanation by Berger (2008: 25) is promising: the language was used to unstressed personal prefixes of the type Vor CV-, and this shape can be regained if an additional unstressed vowel with the same quality as the one in the stem is inserted. The cause for aá is not the decisive point, however. The prefix *yis hardly recognizable any longer as such; therefore it is generalized in the whole paradigm, i.e. employed also in ‘my rib’, ‘your rib’ etc., and y becomes part of the new stem. Berger (2008: 25f.) discusses further examples. Following in Berger’s footsteps, I now would like to deal with the following set (for the nasal, which was *9, see 3.2.3.): Ys. -yú9us - Hz. -úmus ‘tongue’ The initial consonant of Yasin does not match the absence of any consonant in Hunza. One must either assume a loss in Hunza or an addition in Yasin. The latter way is chosen here. At the time when Yasin still had the unstressed *i-, it became *y here because a vowel followed (similarly as with ‘rib’ in Hunza above). This leads to the form yú9us ‘his tongue’. Later then unstressed *iwas dropped, but here the prefix escaped the process because it had acquired the shape of an approximant. It was now inconsequent to have this approximant without any function only in certain forms but not in others. Therefore it was generalized in the whole paradigm: a-yú9us ‘my tongue’ etc. From a historical point of view this form contains both a prefix for the 1st person and one for the 3rd person: a-y-ú9us. Compare the following excerpts of two paradigms, ‘arm’ and ‘tongue’: <?page no="105"?> 3. The dialects and the comparative method 105 stage I stage II 1st *a-Sák > *a-Sák 2nd *gu-Sák > *gu-Sák 3rd m *i-Sák > *i-Sák 1st *a-ú9us > *a-ú9us 2nd *gu-ú9us > *gu-ú9us 3rd m *i-ú9us > *y-ú9us stage III stage IV > *a-Sák > a-Sák > *gu-Sák > gu-Sák > *Sak > Sak > *a-ú9us > a-yú9us > *gu-ú9us > gu-yú9us > *y-ú9us > yú9us At stage I (Proto-Burushaski) no difference in the inflection of ‘arm’ and ‘tongue’ exists. Stage II shows the result of *i- > y-; this stage, therefore, implies a slight difference in inflection. When now *i- > ∅ and stage III is reached, the difference is even greater. In order to abandon it, the paradigm of ‘tongue’ has to be reorganized. An easy way to accomplish this is to introduce y to all other forms, which has happened at stage IV, the present. Regarding these developments from a wider perspective, I would like to draw attention to a phenomenon which is sometimes called Watkins’ law. Watkins (1962: 90-96) points out a tendency which can often be observed in the development of verb paradigms: the 3rd person frequently has a pivotal role and influences forms of the 1st and 2nd persons. The influence can manifest itself in several ways: it may be that a morpheme for the 3rd person is fossilized and then appears elsewhere as well, but also other types of analogy may occur. See Collinge (1985: 239f.) for a discussion of “Watkins’ law” and further references. There is no absolute necessity that it is the 3rd person which dominates the process; also the 1st or 2nd person may do this, although this occurs less frequently. Therefore I would like to rename the phenomenon as Watkins’ tendency. In fact this fits Watkins’ opinion since he writes: “It is the 3 sg. which will tend to impose its form on the rest of the paradigm” (Watkins 1962: 90); note the word “tend”. <?page no="106"?> 3. The dialects and the comparative method 106 The discussion of Burushaski showed fossilizations of ior y-: they once marked a 3rd person possessor and then gave this function up in one dialect in each example. For a theoretical discussion of Watkins’ tendency the following point can be made. Not only person marking in verbs, but also possessive affixes typically behave in such a way that the 3rd person plays the decisive role. Watkins’ tendency is therefore not only a tendency for the evolution of verb paradigms, but also for noun paradigms if they mark person (possessor). 3.8. Other levels of language structure There are differences between the Burushaski dialects also at other levels of language structure, and one can attempt to evaluate these too for historical linguistics. In syntax only very rarely differences can be observed between the dialects. Tiffou (1999: 188f.) claims some tiny deviations concerning ergativity but the matter is too uncertain to be taken up here. Another problem for research is that sometimes a piece of syntactic information can only be found for one dialect and it is not entirely clear whether the other dialects behave in the same way or not. Burushaski syntax has many properties that are typical for SOV languages. In addition there are some more interesting syntactic topics, but what the present chapter is concerned with are differences between the dialects since only these entail the need to actually reconstruct. In all instances where the syntactic structures are the same, these can be assumed in Proto-Burushaski as well. In semantics there are occasional deviations such as in the following cognate set: Ys. pherán ‘moth’ - Hz. phirán ‘spider’; Nager has asqúr-i9-e phirán ‘butterfly’ (asqúr ‘flower’; the expression uses a genitive plural). Another example is Ys. -móqiS ‘face’ - Hz. -móqiS ‘cheek’ (and Ng. -móquS ‘cheek’); there is an interrelated Yasin word -móqoT ‘cheek’ (Berger 1974: 165, Berger 1998III: 291), with the second o < *u. Possibly sometimes semantic reconstruction can be ventured, e.g. by using typological information. Probably one would mostly choose one of the attested meanings and argue that this is the oldest one, but there are difficulties and pitfalls and the question is whether there are methods which help to overcome these. Be that as it may this chapter has shown that a lot can be done in order to arrive at insights on Proto-Burushaski and the essential features of this stage of the language can be considered regained now. <?page no="107"?> 107 4. Internal reconstruction in Burushaski 4.1. Introduction 4.1.1. Internal reconstruction Internal reconstruction seeks to make use of patterns in one language system, e.g. its morphophonological alternations, and to draw conclusions from this material about the past. Textbook treatments include Hock (1986: 532-541), Campbell (1998: 201-225), Sihler (2000: 150- 156) and numerous others; see also Campbell / Grondona (2007: 23-25) and references in Crowley / Bowern (2010: 134f.). 34 Internal reconstruction is not always as reliable as the comparative method. Why then engage in it at all? Crowley / Bowern (2010: 127f.), Joseph (2010: 58) and others have pointed to some types of languages for which this approach may be particularly useful. Indeed, languages are not all equal in this respect: some are more in need of treatment with this method than others. Internal reconstruction is recommendable where comparative evidence is not available or problematic. To be more specific, it is recommendable if a language: (a) is a language isolate, thus there are no relatives (yet) available, (b) has known relatives, but they are quite distant, and thus there are gaps to bridge, (c) has known relatives, but they are poorly known, (d) has a large gap in its documentation, (e) is a proto-language of an entire family; such a language seen by itself and within its time behaves like a language isolate (the point does not apply, naturally, to an intermediate proto-language of a branch of a family). In view of this survey it can be argued that one may regard language isolates as the classic case for applying internal reconstruction. Many other cases somehow relate to a kind of more or less factual isolate status. The method could be applied to all language isolates before hypotheses about relationships are put forward for them. Even if some work should have to 34 The evaluation of dialect variation (cf. chapter 3) also refers to one language, but it should not be subsumed under internal reconstruction: it treats dialects as sister languages and proceeds in the same way as the comparative method. There is therefore no difference in principle to comparing related languages. <?page no="108"?> 4. Internal reconstruction in Burushaski 108 be modified later, such enterprises may still be fruitful. This is done in the present chapter for Burushaski. Internal reconstruction has an inherent property which may be called time depth indetermination: it is unclear at what time in the past a reconstructed state of affairs should be located. This constitutes a difference from the comparative method which operates with fixed protolanguages to which a feature is reconstructed. What we can do in some instances is to say about two events that one occurred before the other, i.e. a relative chronology can be set up. Matters of time depth indetermination for certain reasons become particularly relevant with language isolates, as will be discussed in 4.6. 4.1.2. Internal reconstruction in Burushaski so far Research on Burushaski has so far concentrated mainly on synchronic linguistics (and on collecting oral literature). Considerably less has been done on historical linguistics, a field which for this language could be subclassified into dialect comparison, exchange of loanwords with contact languages, internal reconstruction, and other topics. Among the few publications treating Burushaski from a historical perspective not counting those which claim language relationships there is only one monograph: Berger (2008). Although it contains valuable and sometimes brilliant points and deserves detailed study therefore, there are also numerous ideas which are hardly convincing: too many core vocabulary items are declared loanwords from Sanskrit and other sources, often by unique phonetic and semantic changes, and this bias even affects basic morphemes. Various claims have not been arrived at by defensible methods. Occasionally statements in Berger (2008) are accompanied with an assertion that they are entirely certain, but the uninitiated reader must be informed that actually often they are not. As to internal reconstruction as a specific subfield of historical linguistics, it has seen proposals ranging from very appealing to less so. A case interesting in parts is an analysis of the numerals by Berger (2008: 77-80) which will be taken up in 4.3.1. Klimov / Èdel'man (1972) present a valuable insight: the numeral ‘two’ can be identified as a fossilized prefix in a number of body parts; we will return to this topic in 4.3.2. There are further achievements that can be made use of. In some areas, however, it will be necessary to start from scratch. <?page no="109"?> 4. Internal reconstruction in Burushaski 109 There is an issue in research so far which, if correct, would clearly fall under internal reconstruction, but which is not without problems. Burushaski has three types of plosives: voiceless, voiceless aspirated and voiced (e.g. t th d). There is a hypothesis claiming that these can be reduced to an earlier state with only two series: voiceless aspirated and voiced. This idea stems from Morgenstierne (1945: 75), who had investigated data by Lorimer (1935a, 1935b, 1938). Toporov (1970: 642) mentions this claim, and Berger (2008: 20, 37) shares Morgenstierne’s view. However, no sufficient arguments have been presented for the hypothesis. Morgenstierne (1945: 75) investigated the initials of verb stems and found voiceless plosives there rarer than the other two types. Berger (2008: 20) found the same result for inalienable nouns. The language consists of much more than only verb stems and inalienable nouns, however. The statistical observations by Morgenstierne and Berger may have some significance. It is not convincing, though, to jump from these observations to the conclusion that there should have been one less plosive type, as this would mean neglecting all options of other explanations. Given that historical investigations of Burushaski have so far often run into difficulties, it may be asked whether such enterprises are not more or less futile. Such a view has been expressed many decades ago by Morgenstierne (1935: XIX): “Our chances of ever being able to reconstruct older forms of Bu. [Burushaski; this term refers to Hunza and Nager alone here] either by comparisons with W. [Werchikwar; alternative term for Yasin] or by means of internal evidence in Bu. itself are not very promising.” However, new data and methodological advances have improved the situation considerably since then, and many scholars have already made lasting contributions to diachronic Burushaski linguistics. Moreover, chapter 3 has shown that the comparative method clearly yields a large number of results. A thoughtful and thought-provoking analysis has now been made by Klimov / Èdel'man (1995), who discuss what can and what cannot be achieved when trying to do historical linguistics on a language isolate such as Burushaski. Their conclusion is (Klimov / Èdel'man 1995: 37): “         -        <?page no="110"?> 4. Internal reconstruction in Burushaski 110    -      -   .          ,  -    .” “The genetically isolated language Burushaski, due to its comparatively weak dialect differentiation, supplies very little material for comparative-historical reconstruction of some time depth. This deficiency is compensated for to a certain extent by the possibilities of the typological method, especially by procedures of internal reconstruction.” Klimov / Èdel'man (1995) thus contrast two approaches: dialect comparison and internal reconstruction. None of these two is unimportant or negligible, but there are differences in what to expect. The former is more concerned with smaller and younger changes in details, while the latter is a field which can provide more radical insights, referring to a greater time depth. See also Bashir (2000: 3), who calls Klimov / Èdel'man (1995) an “important paper”. 4.1.3. Objectives of the chapter The aim in the following is to demonstrate that many diachronic points can indeed be made on Burushaski with the method of internal reconstruction. The pessimism by Morgenstierne (1935) is found to be obsolete. In contrast, the views by Klimov / Èdel'man (1995) are confirmed: internal reconstruction, often using typological evidence in order to back up a statement, furnishes us with numerous and at times sweeping insights. It will be seen that this language isolate speaks volumes. Often structures are analyzed which are the same in every dialect; thus dialect differentiation plays no role in these instances, and occasionally only Yasin data are quoted in order to keep the discussion short. If dialect differentiation does exist which has a bearing on the analysis, this is of course incorporated. The chapter is therefore about internal reconstruction supplemented by dialectal comparison whenever appropriate. References to chapter 3 are given, when useful (and also to other chapters). The treatment is also intended as a general contribution to historical linguistics in so far as it seeks to promote one view: internal reconstruction can be of great help in general in revealing something about the histories of language isolates. <?page no="111"?> 4. Internal reconstruction in Burushaski 111 The chapter investigates topics in phonology (or morphophonology), morphology, syntax and the stress system, in sections 4.2., 4.3., 4.4. and 4.5. respectively. As these areas often interrelate, sometimes a decision had to be made where to treat a topic. Section 4.6. gives a summary and then makes a general point on the relation between internal reconstruction and language isolates. 4.2. Phonology 4.2.1. An ancient vowel alternation There is an alternation i / u in Burushaski the historical background of which is unknown so far. The alternation is unproductive and rare nowadays, but it does occur in a number of striking instances. Consider the following singular / plural pair for ‘man’ in the Yasin dialect: sg. pl. hir hur-í ‘man’ These are the forms of the absolutive case; in the paradigm all singular forms contain i, and all plural forms contain u. Given that ‘man’ is a very basic item in the vocabulary of any language, we may suspect something old here, cf. the irregularity in English man / men. In the Hunza dialect the forms are hir ‘man’ and hir-í ‘men’, without a vowel alternation. Although the alternation i / u occurs in all three dialects, it does not always do so in the same items. It has apparently been given up in different words in different parts of the speech area. Berger (2008: 10f.) claims that the alternation i / u occurs only before l and r. Among his examples is ‘face’ in Yasin, which can be -skil or -skul; and he is right in seeing further traces in Hz. -ír- (< *-yír-, cf. 3.2.3.) compared with Ys. -yúr- ‘to die’, and in Hz. tilía9 ‘saddle’ vs. túl-, -ltúl- ‘to saddle’ (Berger 2008: 10, 81). Berger (2008: 11) touches on the question which of the two vowels is the older one in this alternation when he claims that Ys. c6úrk- ‘to milk’ is from the Dardic equivalent of Sanskrit kS í ram ‘milk’ (noun) and therefore represents “das einzige Wort, in dem u eindeutig später als i ist.” “the only word in which u is clearly later than i.” He seems to think that *i is the original vowel. However, the etymology is not convincing, and therefore the question of the older vowel is open. Moreover, Berger’s claim that the alternation i / u occurs only before l and r has two important drawbacks. One is phonetic: it is difficult <?page no="112"?> 4. Internal reconstruction in Burushaski 112 to find a reason why such an alternation should have arisen especially before liquids. Secondly, there is other evidence that could be subsumed under the alternation i / u as well. Consider this singular / plural pair from the Hunza dialect which shows another instance of the alternation: sg. pl. -lCin -lCum-u-c ‘eye’ 35 Yasin has these forms today (Berger 1974: 161): sg. pl. -lCi -lCim-u ‘eye’ But Lorimer (1962: 162), based on fieldwork which was carried out in 1923, 1924 and 1935 (as he explains on p. ix), hence several decades earlier, presents an older plural formation which exhibits the alternation: sg. pl. -lCi -lCum-o ‘eye’ Another example is (2nd syllable) Ys. aíSen ‘guest’, pl. aúSu (data from Tiffou / Pesot 1989: 131); Berger (1974: 129) has as the singular either aíSen or aúSin. Moreover, Hz. 6arís- ‘to agree’ and 6arús- ‘to be straight, to be fitting’ belong together (Berger 1998III: 171, 172). The Yasin equivalent of the first verb is 6arés- / 6arís- ‘to agree’ (Berger 1974: 147), and this can serve to illustrate a minor point: not only i and u can alternate, but sometimes also e and u if Yasin is concerned; this is due to Yasin’s “closed law” which lowered certain instances of *i to e, cf. 3.3.2. In addition, some forms of the numerals are treated entirely separately by Berger (2008: 38), but they show the same phenomenon (the slash separates forms with i and with u; letters in brackets indicate gender or z form): - Hz. hin (m and f) ‘one’ / hun-t-í (z) ‘nine’ (i.e. ten minus one, see 4.3.1.), hun-C-ó (form for all four genders) - Ys. hen (m and f) ‘one’ / hu-t-í (z) ‘nine’, hu-c6-ó (all four genders) - Hz. chind-í (z) ‘five’ / chund-ó (all four genders) - Hz. isk-í (z) ‘three’, isk-én (m and f) / usk-ó (x and y) - Ng. isk-í (z) ‘three’, isk-é (m and f) / usk-ó (x and y) 35 For the consonant alternation n / m see 2.2.1.; for -c in the plural form see 5.3. <?page no="113"?> 4. Internal reconstruction in Burushaski 113 For the data cf. Berger (1998I: 100, 1998III: 215) and Tiffou (1999: 334); Yasin has nothing but front vowels in the first syllables of ‘five’ and ‘three’. Berger (2008: 38) interprets the forms in such a way that i is older and is then assimilated to a following back vowel, but one may wonder if a phonetically more convincing explanation can be found. Now Yasin has an adjective huríski ‘manly, man-like, of a man’ (Berger 1974: 52, 152, 153); in Hunza it exists with leveling as hiríski (Berger 1998III: 200). If i really were the older vowel, it would have been changed to u both in the plural form and in the adjective. In hardly any language are adjectives derived from plural forms of nouns; especially in otherwise relatively agglutinating languages this would not make sense. Hence it is worth attempting to interpret the data, and all of them, the other way round, i.e. to regard u as older. Then hurí and huríski together represent the older state, and hir as the odd man out is the innovator. A hypothesis shall be ventured here: the Burushaski vowel alternation may be umlaut (vowel fronting due to the next syllable), known from various Germanic languages, the Brythonic branch of Celtic (Welsh, Cornish, Breton), Albanian (Demiraj 1993: 53-57), the Uralic language Livonian (Laanest 1982: 139f.), the Kartvelian language Svan (Klimov 1994: 100, 109) and others. The hypothesis would regard u as the older vowel and i as younger, i.e. the reconstructed change would be *u > i. Interestingly, already Morgenstierne (1935: XX) briefly mentioned the term “Umlaut” for the alternation of u and i in numerals. The umlaut hypothesis requires front vowels in subsequent syllables, and for all lexemes they can be found. In the three entries for the numerals ‘five’ and ‘three’, for example, the forms before the slash, which have umlaut, exhibit the suffixes -í, -én, -é. Consider also ‘saddle’ and ‘guest’ which can be interpreted in an equally straightforward manner. In other instances, the hypothesis requires some leveling and rearrangement of vowels in grammatical paradigms, but this is known from other languages with umlaut as well (Koch 1996: 233). The paradigm of ‘to die’ has many forms with a front vowel in the following syllable, cf. for example the 3rd person sg. aorist ‘he died’ and ‘it died’, which is yúr-i in Yasin (Tiffou 1999: 187), i-ír-imi in Hunza (Berger 1998I: 177); this form is valid for three of the four genders (all except feminine). Moreover, all plural forms have either -en or -i (Yasin and Nager), and there are further forms. Ys. yúr-i has u by analogy with forms which preserved u. For the nouns ‘eye’ and ‘man’ one may assume that their umlaut is due to the <?page no="114"?> 4. Internal reconstruction in Burushaski 114 many suffixes in the singular with front vowels, e.g. genitive, ergative and locative -e, as well as the indefinite suffix -en; i.e. the absolutive singular form has its i due to leveling. Within this approach the plural form hurí ‘men’ provides the greatest challenge. It is necessarily regarded as having its -í secondarily, for reasons beyond our grasp. It may stem from a back or mid vowel (rounded or unrounded) or a diphthong, cf. Latin í < *-oi in declensional forms. The irregular singular / plural pair hir / hurí is only a single pair, not more: it would be methodologically mistaken to value it over the broader dataset. What counts is the overall picture. Umlaut in Burushaski is probably very old, since it is only visible in the remnants discussed. It is understandable, therefore, that since then sometimes the conditioning factors have undergone changes. Burushaski umlaut has some parallels with English. In both languages the phenomenon is residual today. In both languages in the nouns that have umlaut, it is linked to number; thus hir / hurí is so to speak the goose / geese, tooth / teeth, foot / feet of Burushaski. Note, however, that English has the umlauted vowel in the plural, while Burushaski has it in the singular. In both languages the vocabulary item for ‘man’ is affected. Finally, in both languages the conditioning has undergone changes subsequently even more so in English than in Burushaski. Whether the Burushaski development went via *ü (IPA [y]), as in the Germanic languages, cannot be established; it is well possible, of course, from a phonetic point of view. In this connection it is interesting to note that the choice is strictly speaking not only between *u > i or *u > *ü > i: in English examples such as goose / geese the alternation formerly was one of two vowels which were one degree more open ( ó , é ), as is still reflected in the orthography. Finally, a check of the hypothesis from an areal point of view reveals many encouraging facts. Languages with ü the possible intermediate stage indeed exist not far from Burushaski, namely the Nuristani languages, and also Modern Tibetan (DeLancey 2003: 271). The term umlaut turns out to be known in Indo-Aryan studies. Masica (1991: 208f.) mentions historical changes of such a kind and focuses on Sinhalese. This language is spoken in Sri Lanka, but Masica (1991: 223) points out umlaut also in Dardic, which is geographically contiguous to Burushaski. From Bashkarik, of the Kohistani group of Dardic, Masica (1991: 223) cites ucuT ‘small’ (m.) / iciT ‘small’ (f.) as umlaut; cf. the Sanskrit feminine <?page no="115"?> 4. Internal reconstruction in Burushaski 115 suffix í . Moreover, geographically close Iranian languages of the Pamir group join the picture. Rushan has kud ‘dog’ and kid ‘bitch’, the latter form “by a process of umlaut triggered by endings which are now lost” (Comrie 1981: 168). Schmitt (2000: 96) presents exactly the same data, but for a different language or dialect, Shughni, and he traces the vocalism of kid to umlaut caused by the Proto-Iranian feminine suffix *í , related to Sanskrit í . 4.2.2. The consonant alternations n / y and s / ś Burushaski has alternations of n / y and s / S in all its dialects, and in several parts of its morphological system. For n / y consider these singular / plural pairs from the Yasin dialect (there are many more examples): sg. pl. 6órkun 6órkuy-u ‘frog’ Ti9án Ti9áy-u ‘egg’ damán damáy-u ‘owner, master’ khápun khápuy-u ‘spoon’ mayén mayéy-u ‘old’ (of things) -sán -sáy-u ‘chin’ -Sílan -Sílay-u ‘tail’ This is the usual treatment for nouns in -n (unless of gender y). In the Hunza dialect the n / y alternation works the same way; to quote a word that Yasin does not have (cf. 3.6.): Hz. khon ‘ant’, pl. khóyo. (The -o here represents an older state than -u, see 3.3.1.) Tiffou / Pesot (1989: 18) point to the suffix -yu and quote the unique pair Cúmu ‘fish’, pl. Cúmu-yu. It is indeed appropriate to reconstruct something like *-yo as a plural suffix; compare these forms with stems in r (the two last ones are accidental homophones): sg. pl. har har-ió ‘ox’ sar sar-ió ‘hare’ sar sar-ió ‘thread’ It is of little phonetic importance that the fieldworkers noted i here and not y. The reconstructing of *-yo, which has in fact already been undertaken in Burushaski studies, cf. Morgenstierne (1935: XX), Anderson (1997: 1033), entails that there was a consonant group *ny. Hence the alternation n / y <?page no="116"?> 4. Internal reconstruction in Burushaski 116 goes back to *n / *ny. The *ny became y, possibly via a palatal nasal (proposed by Anderson 1997: 1032) or a palatalized nasal. A minor point that can be added is illustrated by only a few words (on the vowel alternation e / i see 3.3.2.): sg. pl. phen phí-u ‘fly’ khen khí-u ‘flea’ c6en c6í-u ‘bird’ -Sípen -Sípi-u ‘penis’ Cidín Cidí-u ‘cooking pot’ Hunza equivalents are phin ‘fly’, pl. phí-o, etc. One can reconstruct earlier plural forms with *íy instead of í and assume that the *y was lost after i for simple phonetic reasons; see also Berger (1974: 17). The development thus went a tiny step farther than in nouns such as 6órkun. Further evidence for *iy > i comes from the Hunza verbs gáarc- ‘to flow’ and di-áarc- ‘to rain’ (data from Berger 1998I: 109); the latter verb probably once contained *y since this consonant is interrelated with g in the mutation system (*di-yáarc-). The alternation n / y also occurs in other parts of speech. The forms of the adjective thánum ‘high’ are thán-um (sg., all genders), tháy-u (pl., genders m, f, x), thán-um-i9 (pl., gender y) (Tiffou 1999: 165). The verb stem -mán- ‘to be, to become’ acquires the shape -máyfor certain TAM forms (Tiffou / Pesot 1989: 38). The same holds for lán-, lái- ‘to be able’ and other verbs (Berger 1974: 34). Cf. also sénand sí- ‘to say’; sícan be assumed to have lost a *y after i like phí-u ‘flies’. The fact that the alternation n / y occurs also in verbs shows that it is not possible to analyze the last consonant -n of the nouns as a singular suffix. As to the alternation s / S, it comes as no surprise now that it is regarded as a development from *s / *sy (cf. Anderson 1997: 1033). These are examples for s / S: sg. pl. dúlas dúlaS-u ‘boy’ helés heléS-u ‘boy’ balhás balháS-u ‘witch’ gírkis gírkiS-u ‘rat’ -nú9us -nú9uS-u ‘knee’ <?page no="117"?> 4. Internal reconstruction in Burushaski 117 -yú9us -yú9uS-u ‘tongue’ -yáTes -yáTeS-u ‘head’ -wáldes -wáldeS-u ‘back’ 6óqares 6óqareS-u ‘raven’ 6árqas 6árqaS-u ‘lizard’ -phénas -phénaS-u ‘plait, braid’ hurgás hurgáS-u ‘thick’ (adjective) It should be addressed that there are also plural formations in *-o in Burushaski, e.g. gaT ‘knot’, pl. gaT-ó, and it can be asked how these relate to the plurals in *-yo. An analysis reveals that there is a connection to the columns of the consonant system, and thus to the place of articulation. After dentals, such as n, r, s, *-yo can be reconstructed, after the other places of articulation *-o. A remark on numerals in this regard. All numerals from ‘two’ to ‘nine’ have certain forms in -o (or partly -u in Yasin), and this is the plural suffix. Now some numerals do not obey the dental vs. non-dental rule, e.g. alt-ó ‘two’. But ‘nine’ does and exhibits an alternation t / C in Hunza: hunt-í (z form), hunC-ó (form for all four genders). This instance of t / C can be compared to n / y and s / S. Furthermore, occasionally c / C occurs, e.g. in Hz. huyéltarc ‘shepherd’, pl. huyéltarC-o (Berger 1998I: 49). 4.2.3. Alternation of the retroflex fricative and h There is a group of nouns and adjectives with the retroflex fricative S in the singular and h in the plural in Yasin: sg. pl. -ltéS -ltéh-a ‘eyebrow’ buláS buláh-a ‘childless, sterile’ ceréS ceréh-a9 ‘root’ teléS teléh-a9 ‘piece of a sinew’ (pl. also ‘saddle’) manyóS manyóh-a9 ‘big tube or sack’ noS nóh-a9 ‘young tree’ puS púh-a9 ‘shirt’ hariáS hariáh-a9 ‘urine’ 6uriáS 6uriáh-a9 ‘feces’ 6aCíS 6aCíh-a9 ‘awn, beard’ <?page no="118"?> 4. Internal reconstruction in Burushaski 118 -móqiS -móqih-a9 ‘face’ 6éndeS 6éndeh-a9 ‘gold; queen’ As with other alternations, it is reasonable to assume that this alternation goes back to one sound. The phonetic histories of the languages of the world show that many sounds can develop into h, but h can hardly develop into anything else (it can be dropped). Therefore *S can be regarded as the older state, and there was a sound law *S > *h in intervocalic position. This is also the conclusion that Berger (2008: 35) draws. Two comments ought to be made. Firstly, the hypothesis implies that today’s intervocalic S cannot be old. It turns out that S is rare in intervocalic position; it does appear e.g. in Ys. phéSu / Hz. phéSo ‘pear’ (possibly a loanword from Dardic) and in Ys. Se6éSum ‘sticky, slimy’ / Hz. Si6íSum ‘smooth, slimy’ (possibly onomatopoetic). Secondly, the sound law does not imply that h always must have *S as its basis; there may be other origins as well (outside the group of words under study). In initial position, for instance, h cannot be from *S, since *Swas not shifted in this position and there are many words with S-. The Hunza / Nager dialect has no sound in the plural forms, cf. data from this dialect such as: sg. pl. chiríS chirá-a9 ‘root’ 6éniS 6én-a9 ‘gold; queen’ -móqiS -móqi-a9 ‘cheek’ -miS -mi-an-c ‘finger’ In 3.2.3. we saw that Hunza / Nager underwent *h > ∅ intervocalically (and possibly also under other conditions). First the Proto-Burushaski sound law *S > *h took place, and later Hunza / Nager *h > ∅ . Berger (2008: 35) points out the regularity of the latter development. Some assimilations and losses occurred subsequently. Interestingly, however, also in Yasin quite a number of words can be found which lack h in the plural: sg. pl. pháreS pháre-a ‘duck’ búpuS búpu-a ‘pumpkin’ -ní-kiS -ní-ki-a ‘bearded’ (-kiS suffix) <?page no="119"?> 4. Internal reconstruction in Burushaski 119 -lkiS -lki-a9 ‘lair, nest’ ílkiS ílki-a9 ‘needle’ gí9iS gí9i-a9 ‘small leather bag’ bisárS bisár-a9 ‘sickle’ -meS -m-a-c ‘finger’ When examining this list and comparing it to the Yasin list with h above, in my opinion a connection to stress must be drawn: if *h immediately followed a stressed vowel, it was preserved, while later on in the word it was lost. This explains the presence and absence in almost all instances in Yasin. Only ‘face’ and ‘gold; queen’ are exceptions; the cause for their h may be analogy to other words with h or a high register of speech. (There are a few other words, see Berger 1974: 19, but they confirm the picture.) Finally, it would be welcome to find an explanation for the following pair from Yasin: sg. pl. thoS thu-á ‘new’ Hunza has thoS, pl. thu-áa9 and thu-áan-c. A former *h must be assumed in the plural form again. Its absence in Yasin cannot be accounted for by the sound law just posited or a different type of reference to stress; Yasin preserves h in words such as bihái ‘illness’, 6uhá ‘waning moon’ and forms such as ahíl ‘my lip’, ahúTis ‘my foot’, cf. 3.2.3. Possibly *h was lost already in Proto-Burushaski here due to a kind of Grassmann’s law since the first consonant, th, is an aspirate. In this book several sound shifts have been explained with a reference to stress, and it seems convenient to sum them up here: - In Hunza nasal + plosive > nasal unless before stress (3.2.2.) - In Yasin unstressed *o > u (3.3.1., “unstressed law”) - In Hunza unstressed *e > a in non-final position (3.3.3., following Berger) - In Yasin *h > ∅ unless directly after stress (4.2.3., i.e. present subsection) The question may be asked whether this does not mean returning too often to one phenomenon, stress. This is by no means the case, however. The data seem to indicate such a connection in each case. Moreover, experience shows that if stress is present in a language, it can very well influence <?page no="120"?> 4. Internal reconstruction in Burushaski 120 various developments of its vowels and consonants. In the Indo-European family stress is crucial for Verner’s law, Winter’s law, Brugmann’s law and the development of the vowels in Russian, Portuguese and Classical Armenian, for instance. As a geographically close example, Pashto can be named. Schmitt (2000: 101) reports on the stress “daß er in ganz besonderem Maße die gesamte lautgeschichtliche Entwicklung des Paschtu dominiert hat und zahlreiche Besonderheiten und Unterschiede in ihm ihre Erklärung finden” “that it has dominated the entire sound history of Pashto to a very special degree and numerous peculiarities and differences find their explanation in it”. Stress also played a role in the history of Tocharian A and B, which were located not so far away either. 4.2.4. Voicing after nasals and after l Often observations on the phonotactics of a language give clues to the past. It can be revealing, for instance, which consonant clusters are present and which are absent, and it may also be significant if some clusters, although they do occur, are rare. In Burushaski, the clusters mb, nd and 9g, i.e. homorganic nasal + voiced plosive, are common, cf. words such as altámbe ‘eight’, phandár ‘wart’, wá9ga ‘perhaps’. (Also the corresponding retroflex cluster could be taken into consideration here, but retroflexes are rarer than other consonants.) In contrast, the combinations of nasal + voiceless plosive, e.g. nt, are rare. There are other languages with this state of affairs, e.g. Greek (Modern Greek), Albanian and Mordvin (a Finno-Ugric language). In these languages the diachronic background is that voiceless plosives were shifted to voiced ones after nasals, i.e. *mp > mb etc., which is a widespread sound law. It can be asked whether Burushaski underwent this sound law as well. Berger (2008: 90) demonstrates that the Hunza plural suffix -taro regularly has the allomorph -daro after nasals, as seen e.g. in: Hz. guSpúr ‘prince’, pl. guSpúr-taro Hz. aqhón ‘priest’, pl. aqhón-daro Correspondingly, in Yasin the plural suffix -taru has the allomorph -daru after nasals, see the data listed by Berger (1974: 16). The allomorphy is evidence that indeed voiceless plosives underwent voicing after nasals; this happened already in Proto-Burushaski, since all dialects exhibit the effect. Another example can be adduced. In Hunza there is a suffix -kuc which designates the inhabitants of a place, e.g. Hz. Húnzukuc ‘inhabitants of <?page no="121"?> 4. Internal reconstruction in Burushaski 121 Hunza’, Nagérkuc ‘inhabitants of Nager’ (Berger 1998I: 206, Berger 1998III: 499, Berger 1998III: 504). In the irregular name Hz. Husé9uc, the designation of a tribe, the suffix can be assumed as well, and hence the development 9 < *9g < *9k can be deduced (Berger 1998I: 206f., Berger 1998III: 499). On *9g > 9 in Hunza see 3.2.2. One does find the cluster nt in Hz. huntí ‘nine’ (z form) which has the Yasin equivalent hutí ‘nine’ (z form). In this case the n in Hunza may be secondary; for more on this numeral see 4.3.1. Another counterexample is Ys. baldá9kuin ‘carrier, porter’ from baldá9 (pl.) ‘load’ (data from Berger 1974: 52), but it does not provide a problem: in this case analogy is the cause; there are other nouns for persons with the suffix -kuin, and baldá9kuin has kept or restored the k by analogy. Affricates after nasals can be investigated as well. In final position instances of nasal + voiceless affricate occur: gunc ‘day’, halánc ‘moon’. This is rather meaningless, however, since in final position only voiceless obstruents can occur and no voiced obstruents. (There are no voiceless aspirated obstruents in final position either.) It is the internal position which is relevant. Now in 2.2.2. we saw that a mutation can be found in Ys. cer ‘time, turn’ / hen-zér ‘once’; the Hunza equivalents are chir ‘line, row’ / hin-zír ‘one set’. The conclusion can be drawn that also affricates became voiced after nasals; see also Berger (2008: 44). Another example is provided by Hz. hunc ‘arrow’, pl. hunz-é (Berger 1998III: 205). Yasin has hunc ‘arrow’, pl. hunc-é (Berger 1974: 152), hence no alternation in the paradigm; since the sound law of voicing inevitably must have applied, the conclusion is that analogy must have affected the plural form. There are only very few instances of internal -ncin Burushaski, and then the explanations for their existence are simple: máncel ‘whey’, for instance, is a compound, see 5.2., and cencén et- ‘to cut into pieces’ obviously has reduplication. Thus, in some cases a cluster such as nd can be traced back to *nt. However, an occurrence of nd may also have been *nd already before. In many words which contain nd or a different cluster of the type under study, the cluster is historically ambiguous if no related language can be pointed out; it is unknown whether the type *nt or *nd was present. Various words are affected, among them three numerals: cendí ‘five’, biSínde ‘six’ and altámbe ‘eight’. Obstruents probably underwent voicing also after l. To take a simple example, the plural allomorph -daru, mentioned above, is also <?page no="122"?> 4. Internal reconstruction in Burushaski 122 used for Ys. -yál ‘friend’: -yál-daru (Berger 1974: 16, 186). Berger (2008: 44) speaks of voicing after l as well, but he regards this development only as a tendency. In my view historical linguistics must try to operate with fixed sound laws and explanations for deviant cases in a Neogrammarian fashion. As will be illustrated briefly, it is perfectly possible to assume a normal sound law here and then explain exceptions, when they appear, with analogy. If in contrast one did not assume the sound law, some instances of voiced obstruents after l would be very hard to account for. A derivative suffix employed for forming abstract nouns is -kuS. It occurs in all dialects; examples include Ys. Suá-kuS ‘goodness’ from Ys. Suá ‘good’ (Berger 1974: 52), Hz. daltás-kuS ‘beauty’ from Hz. daltás ‘good, beautiful’ (Berger 1998I: 203). The evidence for voicing after l can be seen in Hz. hól-guS ‘war’ from Hz. hol ‘army’ (Berger 1998I: 203, Berger 2008: 44). There is also a variant Hz. hól-kuS ‘war’ (Berger 1998I: 203, Berger 2008: 44) and this can be interpreted as having restored k analogically. A derivative suffix designating persons is Hz. -kuin, seen e.g. in Hz. ha6úr-kuin ‘rider, horseman’ from Hz. ha6úr ‘horse’ (Berger 1998III: 185). The plosive appears voiced in Hz. il-gúin ‘owner of water’ from Hz. il ‘eye of a needle, hole, opening of an irrigation canal’ (Berger 1998III: 211f.). Another derivative of il ‘eye of a needle’, which exists also in Yasin, is Ys. íl-kiS ‘needle’ (Berger 1974: 153). Here analogy must have been at work; as Berger (1974: 52) reports, the suffix -kiS occurs also in other words. There are further examples for voicing after l; Berger (2008: 44f.) adduces some but a few of them would need critical re-examination. The consonant group lt is frequent in Burushaski today, which may look contradictory at first sight. However, in Yasin the l is voiceless in this group, and in 4.5.4. it will be shown that lt usually goes back to a different type of sound. In relative chronology the voicing after l may be older than the rise of lt. 4.3. Morphology 4.3.1. The numerals from ‘one’ to ‘ten’ There is a discussion of the numerals from ‘one’ to ‘ten’ by Berger (2008: 77-80). The following treatment leaves out his most doubtful <?page no="123"?> 4. Internal reconstruction in Burushaski 123 claims (e.g. on p. 78 the suffix -i / -e is supposed to be via Ys. hek / Hz. hik ‘one’ from Sanskrit ékaH) and concentrates on the remainder. The first ten numerals in Burushaski are (Yasin, z forms): he-k ‘one’ alt-ó ‘two’ isk-í ‘three’ wált-e ‘four’ cend-í, cind-í ‘five’ biSínd-e ‘six’ thal-é ‘seven’ altámb-e ‘eight’ hut-í ‘nine’ tórum ‘ten’ Some simple morpheme boundaries have been indicated by hyphens. Instead of -i / -e the gender forms show different suffixes. The reason for the allomorphy -i / -e is not well understood (see also the end of 3.3.2.); alt-ó ‘two’ contains the plural suffix -o. One point by Berger (2008: 79) results from a comparison of ‘five’ and ‘six’. These numerals can be arranged in slots as follows: c i n d -í ‘five’ b i S í n d -e ‘six’ Probably the consonants c / S can be identified with one another, cf. 2.2.2.; the difference may be due to an old phonetic development impossible to recover in more detail. It then appears that ‘six’ contains ‘five’ and has an additional element. There is much typological evidence which supports such an analysis, e.g. in West Africa: Efik, Bariba (Welmers 1973: 298f.). 36 The Hunza dialect has a form with m-: miSíndi; there are occasional interchanges of b and m (Berger 2008: 28f., and see 2.2.2., E.). Berger (2008: 79) thinks that the meaning of bi- / mimay have been 36 An alternative would be to assume that ‘six’ was altered secondarily to look more like ‘five’ (or vice versa), and this is in fact assumed by Morgenstierne (1935: XX) who holds that a “secondary adaptation” may be involved. It is true that numerals sometimes undergo such changes in which they adapt to adjacent numerals, cf. Kurdish he S t ‘eight’ with unetymological hfrom heft ‘seven’ or the well-known dinstead of nin the word for ‘nine’ due to ‘ten’ in Slavic and in East Baltic: Czech devêt ‘nine’. Usually, however, only a single segment is involved in such cases. In the case of Burushaski ‘five’ and ‘six’ the hypothesis is unlikely, since in that case these numerals would probably not have so many segments in common. <?page no="124"?> 4. Internal reconstruction in Burushaski 124 something like ‘on’, but it may also be an old word for ‘one’ different from hek. Furthermore, there is a connection between ‘one’ and ‘nine’ pointed out by Berger (2008: 79). The data is: ‘one’ ‘nine’ m, f x, y z all genders z Yasin hen han hek hu-c6-ó hu-t-í Hunza hin han hik hun-C-ó hun-t-í What makes the assumption strong right from the outset is the typological background. There are many languages which designate ‘nine’ as ‘one away from ten’ and often also ‘eight’ as ‘two away from ten’. Berger (2008: 79) quotes Tamil onpatu ‘nine’ from onru ‘one’ and pattu ‘ten’, and Finnish yhdeksän (misspelled yhdeksan) from yksi, stem in weak grade yhde- ‘one’, kahdeksan ‘eight’ from kaksi, stem in weak grade kahde- ‘two’. One may add Ainu sine-pesan ‘nine’ and tu-pesan ‘eight’, cf. sine ‘one’, tu ‘two’; and Basque bederatzi ‘nine’ can probably be segmented bede-ratzi because of bat ‘one’ (t < *d, in final position no voiced plosives are possible) and zo-rtzi ‘eight’. Moreover, some Omotic languages (spoken in East Africa) show such numerals for ‘eight’ and ‘nine’ (Amha 2012: 481). See also Greenberg (1978: 259). Even Latin has ú nd é v í gint í ‘nineteen’, literally ‘one from twenty’, duod é v í gint í ‘eighteen’, and close to Burushaski there are Dardic languages with similar numerals (Edelman 1983: 51). In Burushaski, one will therefore compare (quoting this time not Yasin, but Hunza data): h i n ‘one’ (m, f) h u n -t -í ‘nine’ (z) The Yasin forms for ‘nine’ either lost n or never had it. The alternation i / u is accounted for by Berger (2008: 38, 79) in assuming *i to be older, which then underwent an assimilation in the proto-form of Hz. hunCó (and Hz. huntí must then be analogical). In the context of more data, however, i / u was identified as umlaut in 4.2.1. If this is correct, *u is older, and umlaut can be seen in Hz. hin (and the non-occurrence of umlaut in Hz. huntí must then be analogical). <?page no="125"?> 4. Internal reconstruction in Burushaski 125 Now the elements in ‘nine’ between ‘one’ and the suffixes -ó, -í shall be addressed, i.e. -c6-, -Cand -t-. In view of the typological background ‘ten’ would not be surprising here. In Burushaski ‘ten’ is tór-um, and this may indeed be present, given that in compounds of numerals second parts are often contracted. As Berger (2008: 79f.) points out, it is especially interesting to consider the retroflex -c6-. The neighbouring language Shina underwent *tr > c6 (Masica 1991: 210, Berger 2008: 18), and there are similar developments, also elsewhere in Dardic. There is no reason why Burushaski c6 should not have arisen in a similar way as in Shina, which is even known to be in contact with Burushaski (e.g. Morgenstierne 1935: XXIII). Thus, Ys. huc6ó ‘nine (all genders)’ may go back to *hu(n)-tr-ó, and *-trcould be interpreted as the remnant of *tor- ‘ten’. Berger (2008: 79f.) remarks: “Das wäre freilich der einzige Fall, in dem der dardische Wandel tr > c6 auch für ein heimisches Bur.-Wort nachweisbar ist.” “Admittedly, this would be the only case in which the Dardic change tr > c6 is demonstrable also for an indigenous Burushaski word.” But the fact that this is currently the only case need not concern us overly because we do not know the development of large parts of the vocabulary of a language isolate anyway. The next topic is that a connection between ‘two’, ‘four’ and ‘eight’ exists (cf. Morgenstierne 1935: XX, Berger 2008: 79): a l t -ó ‘two’ w á l t -e ‘four’ a l t á -m b -e ‘eight’ As a background consider also Edelman (1999: 226f.) on “octonary” and “quaternary-octonary” systems, as she calls them, in this part of the world. The most difficult member of the set is wálte ‘four’ (wáltu form for the four genders) due to its w-; according to Berger (2008: 79) wmay be the plural marker u, although it would be somewhat awkward that it is a prefix here. The semantics is identified by Berger as “die (zwei) Zweier” “the (two) twos” or something similar. There are two possible paths for interpreting the membership of altámbe ‘eight’ in the set. One appeals to the fact that ‘eight’ doubles ‘four’. The other refers to the fact that ‘eight’ is two less than ten, and a parallel to ‘nine’ can be drawn then. The latter opinion is advocated by Berger (2008: 43, 79). However, no ‘ten’ can be identified this time; Berger (2008: 43) speaks of altámbe, “dessen <?page no="126"?> 4. Internal reconstruction in Burushaski 126 Hinterglied aber nicht mehr zu identifizieren ist.” “the second component of which cannot be identified any longer, however.” Summing up, the discussion arrived at language-internal explanations (or parts of such explanations) for ‘four’, ‘six’, ‘eight’ and ‘nine’. I do not attempt to etymologize the other numerals here. In general it is rewarding to check, for instance, whether ‘five’ is interrelated with ‘hand’; in Burushaski, however, this does not yield any result. According to an opinion by Berger (2008) a “dental suffix” *t or *d can be identified in the numeral system. Berger (2008: 77, 79), convinced of this, even claims that thalé ‘seven’ once had this dental suffix but it was lost by dissimilation to initial th-. I find myself unable to agree with this hypothesis and think that the necessary evidence is not there. It is true that five numerals contain a dental as the second-to-last sound: alt-ó ‘two’, wált-e ‘four’, cend-í, cind-í ‘five’, biSínd-e ‘six’ and hut-í ‘nine’. But consider the results of Berger’s own analyses. In ‘nine’ the dental is a remnant from ‘ten’; ‘six’ must not be counted as it depends on ‘five’, and ‘four’ not either as it depends on ‘two’. We are thus left with a t in ‘two’ and a d in ‘five’. These need not bear a relationship; the fact that both are dental plosives may be coincidence. Moreover, the alleged loss of a dental *t or *d in ‘seven’ by dissimilation caused by th- (Berger 2008: 77, 79) is phonetically questionable. 37 4.3.2. The remnants of the numeral ‘two’ in body parts In Burushaski, suspiciously many designations for body parts that occur in pairs begin with -lt-. Examples include (Yasin dialect): sg. pl. -ltúmal -ltúmal-i9 ‘ear’ -ltán-c -ltá-i9 ‘leg’ -ltéS -ltéh-a ‘eyebrow’ -ltáltar-i9 ‘(female) breasts’ -lCi -lCim-u ‘eye’ These nouns are inalienable, as can be seen from the hyphen; to quote whole forms: a-ltúmal ‘my ear’, gu-ltúmal ‘your (sg.) ear’, etc. That ‘eye’ 37 Some analyses elsewhere in this book affect the discussion. In 4.2.4. it turned out that since the d in ‘five’ is found in the position after a nasal, an earlier *t is an option, though no necessity. Later on, 4.5.4. will be relevant for the t of ‘two’. <?page no="127"?> 4. Internal reconstruction in Burushaski 127 exhibits no t has to do with the fact that another consonant follows; this is regular, cf. Hz. wált- ‘four’ vs. wál-kuc ‘four days’, wál-sa ‘four months’. There are additional examples for the prefix -ltin the Hunza dialect: sg. pl. -ltín -ltí-o ‘bone’ (Ys. ten) -ltúr -ltúri-a9 ‘horn’ (Ys. tur) -lpur -lpur-a9 ‘eyelid’ Again, the last item regularly contains no t. Though most examples are inalienable, there are also alienable instances, and sometimes there is free variation, i.e. a noun can be used as inalienable or alienable: Hz. tin, pl. tinjó ‘bone’, Ys. ten, pl. tenjó, and Hz. tur ‘horn’ (with regular t instead of lt in initial position). Body-parts which do not occur in pairs do not exhibit the prefix: -meS ‘finger’, -móqiS ‘face’, -múS ‘nose’, -s ‘heart’, -skil ‘face’, -xát ‘mouth’, -yáTes ‘head’. Klimov / Èdel'man (1972) discovered that the -ltmust be historically the numeral altó ‘two’ (-ó suffix) and that therefore here we have traces of a dual. This holds although the prefix occurs in the singular as well in the data; this must have been different once. A language the authors refer to, as a typological parallel, is Basque. They name five Basque body parts beginning with be- / bi-, cf. the numeral bi ‘two’ (Klimov / Èdel'man 1972: 160). Note that the fact that Basque is another language isolate does not play any role in their reasoning; it just happens that this language provides a parallel for the phenomenon. In the following the views of Klimov / Èdel'man (1972) on Burushaski and Basque will be supported with further evidence, and the prehistory of the body part forms will be investigated. There are in fact several opinions about the Basque body part prefix question. Uhlenbeck (1927) draws attention to the fact that many body parts begin with b-, especially be-, but he does not put forward an explanation. Holmer (1947: 26) mentions the identification as ‘two’ but does not claim to be the originator of this view. The identification as ‘two’ is supported by Jacobsen (1995: 126) and Zabaltza (1995: 170). Another opinion regards the prefix as a relic of bere ‘his, her’. Bengtson (1991b: 150-153) claims that the prefix is a former class prefix, but he does so only because he advocates a relationship to Nakh-Daghestanian. Trask (1997: 63, 288) denies that there is any prefix at all. <?page no="128"?> 4. Internal reconstruction in Burushaski 128 It is a bit surprising that while all Burushaski experts agree today that -ltis a prefix ‘two’ in body parts (see e.g. Tiffou 2004a: 97f., Berger 2008: 80f.), Basque experts are divided about be-. It may be possible now to settle the matter. Consider the following list of Basque body parts (segmentation added): be-larri ‘ear’ be-gi ‘eye’ be-so ‘arm’ be-laun ‘knee’ be-koki ‘forehead’ be-hatz ‘finger, toe’, also hatz without prefix bi-ri, bi-rika ‘lung’ bi-hotz ‘heart’ bi-zkar ‘back’ The list does not include bepuru ‘eyebrow’, betazal ‘eyelid’, betile ‘eyelash’ and bekain ‘eyebrow’ because some scholars regard them as compounds with begi ‘eye’ (Saltarelli 1988: 293, Trask 1997: 186f., 284). The sources are not very explicit, however, on how these compounds were formed. 38 As to the allomorphy be- / bi-, although its conditioning is unknown today (remnants of vowel harmony would be an option), it can be compared to the allomorphy in the verb prefix e- / i-. One must allow the speakers to have regarded some organs, e.g. a ‘heart’, as consisting of two halves. Moreover, the univerbation is old and semantic shifts could of course occur subsequently; note that behatz means ‘thumb’ in dialects, and Kühnel (1999: 14) translates bizkar as “Schulterblatt” (shoulder blade). In addition, it may be that the prefix spread by analogy to some words which did not have it originally. As a rule, Basque body-parts which do not occur in pairs do not show the prefix: aho ‘mouth’, buru ‘head’, gibel ‘liver’, hagin ‘tooth’, sudur ‘nose’. It is fair to say that the Burushaski evidence and the Basque evidence support each other in the way that they provide a typological background for each other. Consider now another language, geographically 38 As an irrelevant aside, the reader might notice the resemblance of Basque -puru ‘eyebrow’ and Burushaski (Hz.) -l-pur ‘eyelid’. In fact historical identity has been claimed (Bengtson 1991a: 79, 87). One has to be careful here, however. According to Basque historical phonetics intervocalic r (the flap) < *l (as in zeru ‘sky’ < Latin caelum); old *r yielded rr (the vibrant). Moreover, the semantic difference is problematic. <?page no="129"?> 4. Internal reconstruction in Burushaski 129 not far away from Basque: Breton. This language has the numeral daou (feminine div) ‘two’ and singular / dual forms of paired body parts such as the following (Ternes 1992: 416, 2011: 470; segmentation added, in the last form v < *b): sg. du. lagad m. ‘eye’ daou-lagad dorn m. ‘hand’ daou-arn glin m. ‘knee’ daou-lin skouarn f. ‘ear’ div-skouarn gar f. ‘leg’ div-har bronn f. ‘female breast’ di-vronn For Basque and Burushaski, Breton turns out to be the missing link from a diachronic-typological point of view: on the one hand ‘two’ is already part of forms such as daoulagad, on the other hand these forms are semantically still dual forms. It is now possible to reconstruct the development, for all three languages. At the first stage no incorporation of the numeral ‘two’ is present. NPs consisting of numeral + noun may occur, but this is as in many languages if constituent order is NlN (and not NNl, as e.g. in Inuit or Swahili). Then usual NPs with ‘two’, due to their frequent use, become univerbated, and minor phonetic changes occur. The consequence is that dual forms with prefixes exist. This is the stage shown by Breton. Basque and Burushaski go one step further: the singular is lost and the phonetic material of the former numeral ‘two’ becomes a synchronic part of stems; the forms are used as singular forms and obtain a new plural. Corbett (2000), the standard work on number, mentions neither Basque benor Burushaski -lt-; the two languages appear in the index (p. 351), but for other reasons. Corbett is of course well aware of the fact that the numeral ‘two’ may be univerbated to form a dual. This seems to be especially frequent with suffixes; it is less common with prefixes. Corbett mentions Breton daou- (p. 36), but in a different context. The shift to singular, however, is not present in this language. Something can be observed with ‘hand’ and ‘foot’. Burushaski -rén ‘hand’ and -húTis ‘foot’ are not in the lt-group, although they designate pairs, too. Before turning to possible reasons for this, let us look again at Basque: neither are esku ‘hand’ and oin ‘foot’ in the group of body parts with be- / bi-. Jacobsen (1995: 126) draws attention to this fact <?page no="130"?> 4. Internal reconstruction in Burushaski 130 and expresses the view “that ‘hand’ and ‘foot’, being at the extremities of the limbs, would be placed so often in disparate positions that they would not count as symmetrically opposed.” Although this is an intriguing thought, an alternative explanation shall be proposed. The causes may lie in frequency, since this is a crucial factor with analogies and rearrangements of patterns. For applications of this general fact to the category of number see Tiersma (1982), Haspelmath (2002: 246) and the discussion in 3.3.2. which quotes these sources. Two types of frequency must be considered. Firstly, the frequency of the lexemes may explain why ‘hand’ and ‘foot’ could retain their old singular, in contrast to e.g. ‘arm’ or ‘knee’. Secondly, the frequency of the numbers interferes: ‘ear’ and ‘eye’ are often used in the dual. Finally, it should be pointed out how important these insights may become both for Basque and for Burushaski with regard to their status as language isolates. Basic body parts are among the key items in detecting language relationships. It makes a difference whether it is possible to identify a prefix in a whole series of body parts or not. If, for instance, a relative for Basque is found, its body part terms to be more precise: only those which are etymologically identical, of course may contain equivalents only for the segments after the be-. A similar conclusion applies to Burushaski and its -lt-. Note, however, that for both languages we do not know the age of the univerbation; hence it is possible that a common proto-language already had the prefix, and in this case a related language may contain an equivalent to the prefix in question. 4.3.3. On ‘to be’ In Burushaski, as in many other languages, the verb ‘to be’ is the most irregular of all verbs. For Berger’s thoughts on this verb see Berger (1998I: 129), Berger (2008: 105f.); it must be emphasized that one should not follow him in regarding this verb as a loanword from Shina, because for such a central vocabulary item this hardly makes sense. The following analysis draws attention to a crucial point, hitherto not mentioned anywhere, which apparently contributed to the shaping of the Burushaski paradigm for ‘to be’ as it exists today. The usual person / number / gender suffixes in Burushaski are (cf. 2.3.4.): <?page no="131"?> 4. Internal reconstruction in Burushaski 131 sg. pl. 1st -a -en 2nd -a -en 3rd m -i -en f -u -en x -i -en y -i -i These are Yasin forms, but they can also be posited for Proto-Burushaski. In Hunza but not in Nager there is mostly -an instead of -en (Berger 1998I: 136); this innovation is recent. Berger (1998I: 17) accounts for it with a sound law that shifted unstressed *e > a, see 3.3.3. The Hunza dialect has the most conservative forms of ‘to be’; their present indicative forms are charted by Berger (1998I: 144) and Tiffou (1999: 208). These are, supplied with hyphens for some assumed old morpheme boundaries (the main difference in Yasin is that the long vowels have been shortened, cf. 3.5.: ba ‘I am, thou art’, ban ‘we are, etc.’): sg. pl. 1st bá-a bá-an 2nd bá-a bá-an 3rd m bá-i bá-an f bo bá-an x bi bi-é(n) y dilá bicá(n) Negative forms require a prefixed aand devoicing of the initial consonant: Hz. a-pá-a ‘I am not’, ‘you (sg.) are not’, a-pá-i ‘he is not’, etc. (Berger 1998I: 144). The forms of gender y, however, are a-pí in both numbers, thus identical to the form of gender x sg. The first point to note is that the negative forms of gender y must have as their basis an earlier *bi for gender y in both numbers. Therefore, dilá and bicá(n) are probably innovations, the origin of which will not be assessed here, and *bi survived only in apí. 39 Since the ergative form Hz. já-a ‘I’ is < *ja-e by assimilation (-e is ergative suffix), bá-an may be < *ba-en, and this is a good assumption 39 In fact, dilá is a form used by older speakers, deliberately chosen here, and still more widespread in the Nager dialect, whereas younger speakers use bilá with bby analogy with the other forms (Berger 1998I: 144). <?page no="132"?> 4. Internal reconstruction in Burushaski 132 because it makes the suffix -en appear; internal reconstruction seeks to discover patterns, and the assumption is another step to arrive at an earlier stage at which this verb was less irregular. (The assimilation mentioned is older than the split of Burushaski into dialects and has nothing to do with *e > a in Hunza.) Moreover, bi may be assumed to be < *bi-i. All in all, it is not far-fetched to posit the following earlier paradigm: sg. pl. 1st *ba-a *ba-en 2nd *ba-a *ba-en 3rd m *ba-i *ba-en f *ba-u *ba-en x *bi-i *bi-en y *bi-i *bi-i This reconstruction has the same person / number suffixes as other verbs. The decisive point to observe now is that this paradigm is divided into forms with ba- (first four lines) and forms with bi- (last two lines). Forms with baoccur in the 3rd person in the two genders which always designate humans (m, f), as well as in the 1st and 2nd persons, which inherently designate humans, too; in contrast, bioccurs in genders x and y, which never designate humans. Thus the division has a semantic basis; it is human / non-human and employs human ba-, non-human bi-. The distinction human / non-human is present in the structure of Burushaski also elsewhere, of course; it suffices to recall that genders m and f are human, often abbreviated h, and are never distinguished in the plural, while genders x and y are non-human. Cf. also ‘one’: Ys. hen m f, han x y. The discussion has focused on the present indicative. It turns out that other TAM forms (there are not many for this verb) support the thesis. The past tense forms (Berger 1998I: 144) yield nothing new, since they consist of present form + past tense marker as an additional morpheme, and the S-optative (Berger 1998I: 145) confirms the distribution of a and i. There is typological support for using two stems or stem varieties for ‘to be’ according to a semantic feature of the NP. Japanese has two verbs for ‘to exist, to be’ which are animate i-, inanimate ar-. The criterion here is not exactly human / non-human, but animate / inanimate; this, however, is a minor difference (animals are classed differently). <?page no="133"?> 4. Internal reconstruction in Burushaski 133 4.4. Syntax 4.4.1. The rise of the system of local cases Many languages have a system of local cases which indicate position relative to something ( • ), motion towards something ( → ) and motion away from something ( ← ). These languages fall into two groups. One group is satisfied with three cases for the three purposes; this applies e.g. to Turkish and Greenlandic (Holst 2005: 93). The second group makes distinctions such as ‘in’ and ‘on’ something, ‘into’ and ‘onto’ something etc., i.e. the cases specify both the threefold distinction mentioned and more specific spatial relations. Such languages are e.g. Finnish and Hungarian (Holst 2005: 93). It will be argued that Burushaski moved from the first type to the second type, and it did so by univerbation of NPs consisting of noun + postposition. The decisive points of the development that should be reconstructed have already been presented by Anderson (2007: 1238f., 1241). Consider first the interrogative pronouns Ys. áne ‘where’, ána ‘where (to)’ and ánum ‘from where’. Taking them as a point of departure, three cases may be posited: locative • -e dative → -a (-6a after vowels) ablative ← -um The interrogative pronouns are not typical, however; most Burushaski case endings consist of two parts today. Tiffou (1999: 164) speaks of a system of “surdéclinaison” (“superdeclension”): “la surdéclinaison consiste à juxtaposer deux désinences concrètes qui conservent chacune leur valeur sémantique”. With the element -ul-, for instance, the case endings -ul-e ‘in’ and -ul-um ‘from within’ are composed; with -yaTone obtains -yaT-e ‘on’, -yaT-a ‘onto’ etc. The three old (“pure”) cases all survive in the interrogative pronouns (in Yasin). Apart from that, their fate differs. The dative is still entirely alive, the ablative is no longer so (the dative suffix quoted is limited to Yasin; Hunza has -ar instead). The old locative survives in various fossilized forms and phrases such as: - Ys. jér-e hurúCa- ‘to sit in a row’ (Berger 1974: 155), cf. jer ‘line, row’ - Hz. tér-e ‘on the mountain pasture’ (Berger 1998III: 425), cf. ter ‘mountain pasture’ <?page no="134"?> 4. Internal reconstruction in Burushaski 134 - Ys. Yásin-e ‘in Yasin’ (Tiffou 1995b: 162) 40 - Ys. tháp-e ‘in the night, at night’ (Berger 1974: 183), e.g. in thápe gucáren ‘we went at night’ (Berger 1974: 98, sentence 36), cf. thap ‘night’ - Ys. han sán-e ‘(on) one day’ (Tiffou / Pesot 1989: 20), cf. sa ‘sun; day’ As can be seen, both spatial and temporal uses are attested. The longer case endings go back to postpositions which consisted of stems and the shorter case endings. The whole forms therefore go back to PPs consisting of noun + postposition. In an old source there are even still some genitive markers on the nouns: Biddulph (1880: iii) lists for instance siss-é-ùloo ‘in a man’ (Nager dialect, which has -o < *-e by assimilation of rounding; why there are two o’s is not clear and it may be a misprint). This form contains a genitive suffix -e which has now been lost, see also Berger (2008: 102). The feminine nouns do not use -e as genitive suffix, but -mo or -mu, and in fact this morpheme is even still present today: here we have the reason why most cases show an “expansion” of the stem with this syllable. The synchronic grammars describe this expansion, e.g. Tiffou / Pesot (1989: 23f.) for Yasin, Berger (1998I: 58) for Hunza. Consider, for instance, Ys. gús-mo-ul-e (Tiffou / Pesot 1989: 24) / Hz. gús-mu-l-e (Berger 1998I: 58) ‘in the woman’ from the feminine noun gus ‘woman’. A minor point to clarify is why then also the dative has -mo or -mu, since as one of the original cases it would not demand such a morpheme. This is certainly due to analogy: the dative acquired this extra morpheme because so many other oblique cases had it: Ys. gús-mo-6a ‘to the woman’ (form quoted from Tiffou / Pesot 1989: 23). The postpositions in turn go back to nouns. In some instances these can be identified as body parts. For example, -ul- ‘in’ can be linked to -ul ‘belly’, and -yáT- ‘on’ is present in -yáTes ‘head’ (Berger 2008: 101f.). The developments body part > postposition, noun + postposition > noun inflected for case are known from other languages and have been studied in the framework of grammaticalization, e.g. with African data (Heine / Claudi / Hünnemeyer 1991: 126). See also Blake (1994: 166-168). The important conclusion can be drawn that large parts of the Burushaski case system are young; its impressive size is secondary. 40 Also Yásin-ul-e ‘in Yasin’ exists (Tiffou 1999: 134, 140, 142, 150, 152). <?page no="135"?> 4. Internal reconstruction in Burushaski 135 4.4.2. The rise of an adjective type Adjectives in Burushaski fall into two types. One type does not have any special marking. Adjectives of the other type, in contrast, have a suffix -um; cf. a simple sentence such as: bárdum bái red- UM be- 3SG.M ‘he is red’ Now several tenses in Burushaski are analytic, and among them both the present and the perfect use the present of ‘to be’ as an auxiliary verb; its forms are combined with two different participles, but both end in -um. A simple sentence is (Berger 1974: 35): cúrum bái come- PTCPL be- 3SG.M ‘he comes, he is coming’ It can be seen that the sentence with the um-adjective is identical in structure to the analytic verb form. Therefore it makes sense to assume that um-adjectives arose via a reinterpretation of analytic verb forms with ‘to be’. This means that um-adjectives earlier were participles. This has parallels in other languages, e.g. English, cf. adjectives such as interesting, boring, tired, excited etc., even old (cf. Latin altus ‘high’, from alere ‘to nourish’). There are some adjectives which already existed in the language without -um and then acquired this suffix later, see Berger (2008: 66f.), from whom the following examples are taken. Consider for instance the words for ‘soft’: Ys. 6er6éT, Hz. 6ir6íT and 6ir6íT-um; it makes sense to assume that the form without -um, being present in every dialect (also Nager), is older, and the Hunza form with -um is an innovation. For Ys. iSqór-um / Hz. Suqúr-um ‘sour’ cf. the expression Hz. Suqúr sis ‘Europeans’, literally ‘sour people’ (Berger 1998III: 411, Berger 2008: 67) and Hz. Suqúriki ‘sour’ (Berger 1998III: 410). Sometimes the inflection of an adjective reveals that -um was added secondarily: bién-um ‘thin’ has a plural form biéy-u which points to earlier *bién. The explanation of um-adjectives is not spectacular, but attention should be drawn to a less obvious implication that the foregoing theory seems to have if it is correct. It would follow that in the prehistory of <?page no="136"?> 4. Internal reconstruction in Burushaski 136 Burushaski first the analytic verb forms existed, and only then the umadjectives arose. There is a third occurrence of -um in the language, namely it is the suffix of the ablative case (cf. 4.4.1.). Whether this -um is historically identical to the participle suffix and adjective suffix as well is difficult to tell. Berger (2008: 66) takes such a connection for granted, but he does not point out a way in which one use could have originated from the other which would in my opinion be required. 4.5. On stress in Burushaski 4.5.1. Introduction A very surprising insight can be gained on the (at first quite bewildering) stress in Burushaski. In order to make the following investigation accessible, it proceeds step by step. (This section contains one of the longest and most intricate discussions of the whole book.) First some generalities on stress (or accent). I distinguish three types of languages with stress: (1) In some languages stress is fixed, i.e. always on the same syllable, e.g. on the first as in Finnish or Hungarian, on the last as in French, or on the penultimate syllable as in Polish. (2) In other languages, stress is predictable from the segments by a simple set of rules, e.g. Latin, cf. below; no reference to the morphology or any other part of the language is required. (3) In still other languages stress is distinctive (or free, a possibly misleading term). This means that no system exists which could predict stress in general. Examples are Russian as well as Ancient and Modern Greek. Often in such languages stress is also mobile, which designates that it can be on different syllables in morphological paradigms. There is no need, of course, for every paradigm to have mobile stress; usually there are both mobile and immobile paradigms. In languages in which a set of rules determines stress (the second type), these rules often refer to syllable structure. A crucial concept is that of weight. One distinguishes light and heavy syllables: the former end in a short vowel, while the latter may end in any more complex structure, e.g. a long vowel, a combination of short vowel + consonant or a vowel plus several consonants. To put it with morae, the time units of language: “light” syllables end in one mora, whereas “heavy” syllables end in two or <?page no="137"?> 4. Internal reconstruction in Burushaski 137 more morae. Latin is a well-known example of a language in which syllable weight determines stress. Considering words with three syllables or more, if the penultimate syllable is heavy, it bears stress, while if the penultimate syllable is light, the antepenultimate syllable bears stress, hence am í cus [-'--] ‘friend’ with heavy -m í -, digitus ['---] ‘finger’ with light -gi-. In diachrony any stress system may undergo changes. The Latin system, for instance, is not preserved in any Romance language. In Burushaski some synchronic facts are as follows. Stress is distinctive; there are minimal pairs such as galí ‘he (or: it) went (away)’ gáli ‘it (or: he) broke’ (Berger 1974: 7). Stress is also mobile, i.e. its position can be on different syllables in different forms of one word, cf. gaT ‘knot’, pl. gaT-ó (mentioned in 4.2.2.), mu-wál-u ‘she fell’ / a-mú-wal-u ‘she did not fall’ (Berger 1974: 25). All words consisting of only one syllable and they are frequent in Burushaski need not be marked for stress, of course. But alienable nouns with one syllable fall into two sets: those which shift the stress to the plural suffix and those which retain it in the stem, thus gaT ‘knot’, pl. gaT-ó, but tham ‘king’, pl. thám-u. Alienable nouns with two or more syllables never shift stress when forming the plural. Inalienable nouns can be divided into two types: those which have the stress on the first syllable of the stem such as a-nú9us ‘my knee’, a-rén ‘my hand’, and those which have the stress on the possessive prefix such as á-skil ‘my face’; the latter group is much smaller. All this looks at first quite discouraging, and most previous researchers have accepted the language as “messy” with regard to stress. In the following investigation, however, it will be argued that internal reconstruction reveals an unexpected result on Burushaski stress. The treatment will consist of three parts. First, the basic claims are presented (4.5.2.). Subsequently, a survey of stress in the morphology is given (4.5.3.). Next, an insight on the consonant group -ltwill be arrived at (4.5.4.); this topic could be regarded as a matter of consonantism, but it can only be understood in connection with stress. A brief summary follows (4.5.5.). 4.5.2. Main claims Words with stress on the third syllable or later almost always turn out to be loanwords, e.g. hukumát ‘government’. Examples for such words which may not be loanwords are extremely difficult to find; an example is <?page no="138"?> 4. Internal reconstruction in Burushaski 138 Hz. Sirijón ‘mushroom’ (but Ng. Siríjon, data from Berger 1998III: 395, see also 3.5.). This implies that inherited words in Burushaski with two or more syllables have their stress either on the first syllable or the second. Consider now some examples: stress on first syllable stress on second syllable sórqon ‘thunder’ garú ‘spring’ 6órkun ‘frog’ akhúin ‘today’ 6árqas ‘lizard’ telé ‘walnut’ 6éndeS ‘gold; queen’ gacér ‘vulture’ wá9ga ‘perhaps’ Siní ‘summer’ sáaTik ‘yesterday’ da6óm ‘flour’ bérkat ‘top, peak’ 6aSú ‘onion’ ílkiS ‘needle’ halánc ‘moon’ phárce ‘cap’ behék ‘willow’ 6óqares ‘raven’ haúlal ‘butterfly, moth’ (trisyllabic) It turns out that there is a connection of stress to the nature of the first syllable. The examples in the leftmost column mostly have a “heavy” first syllable: sór-, 6ór-, etc. (Only as the last word I intentionally chose one of the exceptions, in order to underline that these exist as well.) Words of the other type have an open syllable with a short vowel, i.e. a “light” first syllable: ga-, a-, etc. Due to the connection to syllable weight the pattern can be called weight regularity. The weight of the non-first syllables is irrelevant. The system occurs also in other languages, e.g. Ainu (Shibatani 1990: 12). However, it has not been pointed out so far in Burushaski studies. 41 Although the correlation does have exceptions, it is so striking in the whole lexicon that it should not be regarded as coincidental. 41 Recently Munshi (2010: 52) has claimed the following: “In Burushaski primary stress in underived polysyllabic words falls on the rightmost heavy syllable, if there is one, or on the second syllable.” This statement makes the same predictions as the system proposed here for more than 90% of the vocabulary. (The most important difference arises when a word consists of two heavy syllables.) Munshi’s idea works worse on large data bases, however, and this being of paramount importance it is less simple and less natural than the system advocated here. There are also some statements by Anderson (1997: 1028f.) which connect syllable type and stress. The fact that several researchers recently looked for a system in stress data is a turn of events in the history of Burushaski studies, since it had been usual practice among scholars not to see any system in Burushaski stress. <?page no="139"?> 4. Internal reconstruction in Burushaski 139 An older state can be imagined in which the pattern of weight regularity was exceptionless, just as the Latin system referring to weight was. This is an issue of internal reconstruction: a pattern which can be observed today in a slightly imperfect form, though clearly still showing through, is reconstructed for an older stage in a strict form. Certain events have blurred the pattern; among them are loanword acquisitions (such as e.g. úmur ‘age’, see 1.4., duSmán ‘enemy’) and some other kinds of events that will be presented later on (in 4.5.3., 4.5.4. and 4.5.5.). Internal reconstruction can even try to go another step back in time and ask how the system of weight regularity could have come into being. Although the system may have existed like this for a very long time, it must have arisen once in the past, at what time ever. Correlations ought to be accounted for, if possible. Facing initial pand medial -bin complementary distribution, one reconstructs *p. In the present case the correlation is the one between weight of the first syllable and stress position. The crucial idea is that light syllables may lose stress, and therefore the words with stress on the second syllable may have undergone this process; this would mean that garú ‘spring’ < *gáru, etc. The shift can be called the weight shift. Note now what this hypothesis entails. If words of the type gacér once had stress on the first syllable and words of the type sórqon ‘thunder’ had stress there as well, and there were only these two types of words in the untouched system of weight regularity, the consequence is that all words once had stress on the first syllable. We thus arrive at a language with fixed stress on the first syllable. It is now possible to describe the development. Originally there was fixed stress on the first syllable. Then the weight shift made stress move to the second syllable if the first was light. This implies that stress remained on the first syllable if that was heavy. 42 It turns out that there is no alternative to this scenario. One might at first envisage interpreting the data the other way round and reconstruct fixed stress on the second syllable, which then shifts to the first if that is heavy. This would be a “mirror image theory”. However, this would not be wise due to typological reasons: languages with fixed stress on the second syllable do not seem to 42 Attention should be drawn to the difference between the diachronic term “weight shift” which denotes a sound law in the area of suprasegmentals, and the synchronic term “weight regularity” which denotes the resulting state in the language. <?page no="140"?> 4. Internal reconstruction in Burushaski 140 exist. In contrast, languages with fixed stress on the first syllable abound: Icelandic, Czech, Latvian, many Uralic and Dravidian languages, etc. 4.5.3. Stress in morphological forms In the discussion so far, only citation forms were mentioned, such as e.g. absolutive singular for nouns. Naturally, however, there are also various affixed (inflected) forms. This leads to a new part of the investigation, which deals with stress in the morphology. Inevitably, the weight shift often caused different stress positions in morphological forms that belong together. However, this state of affairs could be changed: by analogy. If an affixed form should not have the stress in the same position as the non-affixed form, analogy (leveling) could make the stress of the affixed form adapt to the affixless form. This is indeed what happened in certain parts of the grammar and what makes morphological criteria play a certain role today for the position of stress in paradigms. (One could also speak of “morphological reorganizations” with regard to stress or similar, but this amounts to the same as speaking of analogy; the latter is the proper term within the classic dichotomy sound law / analogy.) Many Burushaski forms with affixes are indeed historically regular, which is to say that they still retain the effect of the weight shift; other forms, in contrast, have been subjected to analogy. Compare these two singular / plural pairs, for instance: sg. pl. gaT gaT-ó ‘knot’ tham thám-u ‘king’ 43 From a historical point of view gaT-ó is regular, while thám-u is interpreted as analogical, i.e. it acquired its stress from the singular form. (It may be that the specific item gaT is a loanword from Indo-Aryan but that does not affect the matter in principle.) It would now be possible to go through the whole morphological system of Burushaski, deduce where the weight shift had to occur, and then explain each form found either as historically regular or as analogical. 43 All hyphens indicate morpheme boundaries, as in the rest of this book. Division into syllables would demand thá-mu etc. Syllable divisions in Burushaski are so simple that they do not require marking; if a cluster contains two consonants, the syllable break is between these two. <?page no="141"?> 4. Internal reconstruction in Burushaski 141 For reasons of space only a brief survey of the morphology with respect to stress will be given. The main objective is to underline that the interpretation of the Burushaski data as stemming from a system with weight regularity (which in turn is from former stress on the first syllable and then weight shift) is indeed feasible, and even highly explanatory. Besides, it becomes apparent what can be concluded or should be assumed about some details. I proceed according to parts of speech. Alienable nouns can be divided into monosyllabics and polysyllabics for the purpose of the discussion. Monosyllabics never change their stress when adding case endings. Many cases are secondary and quite young (see 4.4.1.), thus decisive are the older cases. For example, hir ‘man’ has the genitive hír-e and the dative hír-a; the data can be interpreted as analogical. The analogy also applies to some other types of suffixes, e.g. the indefinite suffix -en, -an: hír-en ‘a man’. In number marking, in contrast, there are still many effects of the weight shift, namely with those plurals of monosyllabic nouns which stress the plural suffix (such as gaT-ó, cf. above). As to polysyllabics, it was mentioned in 4.5.1. that nouns with two or more syllables never have mobile stress. This rule is now easily explained: the weight shift either affected no form in the paradigm (if the first syllable is heavy) or all forms (if it is light). The result is immobile stress either way. Alienable nouns constitute the part of speech which was affected most by analogy; the phenomenon needs to be assumed much less often with the other parts of speech. Inalienable nouns usually reveal, with data such as a-nú9us ‘my knee’, a-rén ‘my hand’ and á-skil ‘my face’, the expected regularity; note that -skil is a stem beginning with two consonants, and s closes the first syllable. The prefix a- ‘my’ is of course only one of several possessive prefixes, but all possessive prefixes have either the shape Vor CVand behave in the same way as adoes for each particular lexeme. The possessive prefixes must already have been present when the weight shift occurred. A challenge to the theory is presented by the lt-words such as a-ltúmal ‘my ear’ in which it is not the first syllable which is stressed (for the data see 4.3.2.); this problem, however, will later be solved (in 4.5.4.). When inflecting inalienable nouns for case, number etc., stress is always immobile; this is what had to be expected. Adjectives are divided into those without any special marking and those with the suffix -um, see 4.4.2. Adjectives without -um but with two <?page no="142"?> 4. Internal reconstruction in Burushaski 142 syllables include Suá ‘good’, 6uSán ‘glad’, qaqám ‘bitter’, mathán ‘far’ etc., all regular; 6er6éT ‘soft’ is an exception. From the um-type consider: cul-úm ‘heavy’ Ca6úr-um ‘cold’ gar-úm ‘warm’ humálk-um ‘fast’ 6uT-úm ‘deep’ Se6éS-um ‘sticky, slimy’ bur-úm ‘white’ mat-úm ‘black’ bié-n-um ‘thin’ gagá-n-um ‘short’ gilí-n-um ‘thin’ 6usá-n-um ‘long’ da6á-n-um ‘thick’ thá-n-um ‘high’ This supports the theory, and we now also understand why the suffix -um is sometimes stressed and sometimes not. The last item looks like an exception but it is tháanum (with long vowel) in Hunza (Berger 1998III: 435). These are some additional examples from Hunza: ha6-úm ‘moist, wet’ dalén-um ‘fat, thick’ hir-úm ‘sharp’ tharén-um ‘narrow’ Pronouns behave like nouns; the analogy with case endings mentioned with the nouns applies also to them: ja ‘I’ já6a ‘to me’, men ‘who’ méni9a ‘with whom’. Numerals (for a list up to ‘ten’ see 4.3.1.) mostly adhere to the system as well. For some which do not, one may argue that numerals often influence each other; in this way iskí ‘three’ may have its deviating stress from altó ‘two’ (on which in turn see the lt-discussion in 4.5.4.). For Ys. tórum ‘ten’ cf. Hz. tóorimi with a long vowel. Verbs have numerous forms, and these have their stress extremely often on the second syllable. It turns out that this is perfectly regular because the forms mostly consist of prefixes (or prefix chains) of the shape Vor CVfollowed then by a stem with one initial consonant (and then suffixes), e.g. a-yéc-i ‘he saw me’. If the stem begins with a consonant group, usually the prefix is stressed: é-sqan-u ‘she killed him / it’. For judging certain forms it becomes relevant that the Yasin dialect lost unstressed *i-, preserved in Hunza-Nager (see 2.3.1., 3.3.3.). Of special interest are certain pairs of verb forms such as these: a-wál-a ‘I fell’ / ay-á-wal-a ‘I did not fall’ gu-wál-a ‘you (sg.) fell’ / a-kú-wal-a ‘you (sg.) did not fall’ wál-i ‘he fell’ (< *i-wál-i) / a-í-wal-i ‘he did not fall’ mu-wál-u ‘she fell’ / a-mú-wal-u ‘she did not fall’ <?page no="143"?> 4. Internal reconstruction in Burushaski 143 In descriptive accounts of Burushaski this has so far been described as a peculiar shift of the stress one syllable towards the beginning in the form with one prefix more, see e.g. Berger (2008: 53). From a synchronic point of view this would be legitimate. It can be seen now, however, that in language history there actually was no such shift backwards: originally both forms had the stress on the first syllable, and both shifted it regularly to the second. Particles often have only one syllable, but if they are polysyllabic, the claim is supported again: amútuk ‘now’, akhúin ‘today’, wá9ga ‘perhaps’, sáaTik ‘yesterday’. The adverb jímel, also jímele, ‘tomorrow’ is no real exception because it is a fossilized case form. There are some pairs of related alienable nouns and inalienable nouns. Considering those with a stem which has more than two syllables of which the first one is light, it is revealing to observe the position of the stress (Yasin data from Berger 1974: 164, Hunza data from Berger 2008: 52f.): - Ys. mamú ‘milk’ / -mámu ‘female breast’ - Hz. 6uyá9 / -6úya9 ‘hair of the head’ - Hz. manóT / -mánuT ‘celebration’ It can be seen that the alienable nouns have stress on the second syllable, which is regular, and the inalienable nouns have stress on the first syllable of the stem, which is regular as well, since a prefix of the shape Vor CVprecedes. The alternating stress is a consequence of the former stress on the first syllable on each form. The situation can be compared to that of the verb forms discussed above such as a-wál-a ‘I fell’ / ay-á-wal-a ‘I did not fall’. 4.5.4. Prehistory of the consonant group lt Finally, we now come to a real highlight of internal reconstruction. In the discussion of the inalienable nouns in 4.5.3. we came across the problem that those with -ltpresent a challenge for the theory: a-ltúmal ‘my ear’, a-ltánc ‘my leg’, etc. Although the first syllable is closed (by the l) and should therefore carry the stress, the effect of the weight shift is observed. If a different consonant follows the l in such words, the stress is on the prefix: á-lCi ‘my eye’, so also the Hunza term for ‘eyelid’ with -lp- (for all data see 4.3.2.). An unexpected stress is also found in altó ‘two’, as well as in multán ‘blood’, maltáS ‘butter’, bultú ‘day’, galtár ‘twig’, hultáS <?page no="144"?> 4. Internal reconstruction in Burushaski 144 ‘barefoot’, Hz. daltás ‘good’, Hz. giltír ‘pod, husk’, Ng. daltán- ‘to thresh’ and many other items. All these stress irregularities apparently co-occur with the consonant group -lt- (whereas for other consonants groups this does not seem to be typical even if the first element is l). First, note in passing that no semantic connections exist between the words (except, of course, the connection between ‘two’ and the body parts), and this means that the search for a solution ought to focus on phonetics. If one were to give a description of the situation, it could be said that the group -ltbehaves as one consonant. One is now led to wonder whether at the time of the weight shift it actually was one consonant. There is indeed a way to carry on with this assumption: one can swap the elements and reconstruct a lateral affricate, IPA [  ]. Both Caucasian and American Indian linguistics use the symbol ", and I herewith want to introduce this to Burushaski studies as well. An affricate is a single unit, not a sequence of two units. In this case *" dissolved in a way that the two components were metathesized to lt; after the process two independent segments exist. The relative chronology implied is that first the weight shift occurred, producing forms such as e.g. *mu"án ‘blood’, and later *" > lt. Cases such as á-lCi ‘my eye’ have the stress on the first syllable because they always had a consonant group. The reconstruction can be checked in several ways. Synchronic systems with " often also contain a voiceless lateral L. However, there is no implicational universal that the presence of " also required L, as can be gathered from Maddieson (1984) who gives a list of languages with " (p. 225) and a list of languages with L (p. 234). Interestingly, the cluster lt even has voiceless L in Yasin (Tiffou / Pesot 1989: 10, data by Backstrom / Radloff 1992: 243-260), and the voicelessness of the lateral can be interpreted as a relic of former *". Considering the claim from a diachronic point of view, there may be no attestation of " > lt so far; there is " > l or " > t in modern Nahuatl dialects, for instance. The question is whether this is problematic. First, no linguist can know every attested sound change. Moreover, some sound changes are for the first time found in work on reconstruction (demanded by evidence which permits hardly any other way out), and not because they are directly attested which requires long stretches of documentation, and languages with lateral affricates usually lack that. This is in fact a methodological point. One often uses typology in backing up reconstructions, <?page no="145"?> 4. Internal reconstruction in Burushaski 145 and I do so in this book as well. However, typology should not be overused. This is because something may have existed (in synchrony) or happened (in diachrony) that we happen to have no parallel for this is logically possible and ought to be accounted for by us linguists. Changes similar to the one reconstructed here do occur with more common types of affricates (sibilants involved), i.e. [  ] > [St], etc. Modern Irish has examples in which affricates in loanwords from English have been treated in such a way: cisteanach ‘kitchen’ (st [St j ]) < English kitchen with a suffix. There may be another parallel in Slavic. Bulgarian no S t ‘night’ has a cluster S t where almost all other Slavic languages have affricates: Serbo-Croat noC, Polish and Czech noc, etc. The Balto-Slavic ancestor is *kt, cf. Lithuanian naktìs ‘night’, Latvian nakts ‘night’. The traditional view requires several independent changes to affricates: in West Slavic, East Slavic and parts of South Slavic. It may make sense to assume an intermediate affricate stage in Proto-Slavic, which then entails a change to a cluster in Bulgarian. It must be added, though, that this idea about Slavic is innovative and would need further examination, for instance with how it relates to Macedonian, which has a palatal plosive, and to the system of affricates in general. Checking the reconstruction *" from an areal point of view, it turns out that the affricate " exists in some nearby Dardic languages, namely “Gawarbati, some Pashai dialects, and partly Bashkarik” (Masica 1991: 210); Masica writes it $. Thus, similarly as in the case of umlaut and other claims in this chapter, an areal background for the reconstruction can be adduced. 44 4.5.5. Summary of the events I maintain that in a predecessor of Burushaski there was fixed stress on the first syllable. Then occurred a change, the “weight shift”, which moved stress to the second syllable if the first was light. The resulting 44 I originally developed the idea of an earlier lateral affricate and a metathesis in Burushaski in a computer file from 2005. In 2013 I saw to my surprise that Bengtson (2008: 246) has had this idea, too. He reports that details are found in “a paper yet to be published” of his. Bengtson apparently arrived at the idea on the grounds of his relationship theory which I do not support, cf. 1.3. In contrast, the discussion above focused on arguing with internal evidence in Burushaski, typology, areal considerations, etc. It is remarkable that two researchers entirely independently assumed the lateral affricate and the metathesis, despite having completely different backgrounds. <?page no="146"?> 4. Internal reconstruction in Burushaski 146 system was “weight regularity”. After that, the language moved away from weight regularity to some extent by a number of developments. These are: adoption of loanwords (4.5.2.) analogy (reorganization of stress in morphological patterns, 4.5.3.) subsequent sound laws The last point shall be elaborated upon. Already before it split up into dialects, Burushaski underwent the following law: single consonant > consonant group: *" > lt (4.5.4.) The Yasin dialect then added further complications: dropping of unstressed *i- (3.3.3. and 4.5.3.) rise of some cases with i + sibilant such as iSqá ‘grass’ (3.4.) shortening of long vowels (3.5.) For Hunza / Nager, in contrast, the following innovation must be recorded: rise of a few cases with nasals from former consonant clusters such as Hz. and Ng. 6éniS ‘gold; queen’ (3.2.2.) All these developments made the state of affairs arise which Burushaski has today. That there once was fixed stress on the first syllable in a forerunner of Burushaski constitutes a remarkable new insight. 4.6. Final reflections It is useful to summarize what internal reconstruction seems to be able to reveal about Burushaski. If the results of the foregoing are correct, predecessors of Burushaski had a system of umlaut, consonant groups with *y existed which then merged into single consonants, there was a law *S > *h in intervocalic position, and obstruents became voiced after nasals and after l (4.2.). The numerals from ‘one’ to ‘ten’ show various traces of interrelations, the numeral ‘two’ can be discerned in a number of body parts, and two types of ‘to be’ existed, their usage depending on whether the referent is human or not (4.3.). There was a much smaller local case system once, and the adjective type in -um apparently goes back to participles (4.4.). Furthermore, it was argued that a predecessor of Burushaski had stress on the first syllable, and connected to this a case has been made to reconstruct a lateral affricate *" (4.5.). 45 45 There are further points mostly smaller or less certain ones that can be made on Burushaski with internal reconstruction. They could not be discussed in this book. <?page no="147"?> 4. Internal reconstruction in Burushaski 147 It is now essential to put these results into some perspective. For this purpose consider first a recent investigation by Campbell / Grondona (2007) into Chulupí, a Matacoan language from South America. The authors first get to a number of results on Chulupí by means of internal reconstruction. In order to check these results, they then draw on the other Matacoan languages and apply the comparative method. An important observation is made then. Some of the changes that had been worked out by internal reconstruction were indeed changes which are peculiar to Chulupí; an example is what the authors denominate “delateralization” (p. 24). In contrast to that, other changes are older; they took place at earlier times in relation to language splits, and therefore the consequences of these changes affect also other languages. An example is provided by what the authors denominate “vowel deletion”, the effects of which are also present in Chulupí’s close relative Maká (p. 23, n. 14, p. 24f.). Campbell / Grondona (2007) deal with the question whether internal reconstruction or the comparative method should be applied first when investigating a language. I would like to focus on an interrelated aspect of the matter here. As is known, internal reconstruction is usually unable to tell something about the age of older states and of subsequent events. This phenomenon I call time depth indetermination (cf. 4.1.1.). In particular, what I would like to draw attention to is that internal reconstruction does not provide any information on which relation its results have to language splits. There is now a crucial difference between Chulupí and Burushaski. Relatives of Chulupí are known, and therefore it was possible to detect which developments are confined to Chulupí, and which are not. Burushaski, in contrast, is a language isolate, and therefore languages which could provide this type of information are lacking. A general point about language isolates can now be made which seems to be new in historical linguistics. In the long run a goal of historical linguistics is to develop a good hypothesis about what an isolate is related to: language presumed isolate related under study language(s) <?page no="148"?> 4. Internal reconstruction in Burushaski 148 The point is now to realize what time depth indetermination implies in this context. A change unraveled by internal reconstruction may be a change in the left branch of the diagram, i.e. the isolate underwent it at some earlier time. This possibility is easy to grasp. There is a second option, however: it is just as possible that the change had already occurred in the common proto-language of the isolate under study and the presumed related language(s). Take an example that was already addressed at the end of 4.3.2.: the remnants of ‘two’ in Burushaski body part terms may be absent in related languages, but they may just as well be present in them. We have to be aware of both possibilities, and neither one is less important. If ever a language, or languages, related to Burushaski should be found, each event reconstructed in this chapter may be after the split of the common protolanguage, or before this split. This study has dealt with Burushaski. It could be investigated what can be achieved by internal reconstruction with the other language isolates of the world. Such work could contribute to detecting their origins one day. There are several regions of the world which could be envisaged. A promising one is North Asia (Siberia) with its adjacent areas, i.e. Korea and Japan; an advantage is that good documentation is available for most of the isolates in this region. Another example is Mesoamerica, an area for whose classification Suárez (1983: xiv-xvii) is still the best starting-point, as has recently been argued by Holst (2011: 164); especially Tarascan is so well documented that it should be tackled. There are also some dead language isolates in the Near East and Asia Minor, e.g. Sumerian and Hattic, but some of them are not well known and this is likely to become an obstacle. And there is Basque. There are further language isolates in the world that would benefit from investigations with internal reconstruction. <?page no="149"?> 149 5. On the vocabulary of Burushaski 5.1. Introduction The vocabulary of a language often allows various interesting observations. Linguists frequently focus on the grammatical structure when approaching a language, but in fact the vocabulary should not be neglected either. Burushaski contains large amounts of loanwords (cf. 1.4.), but it also has many words which seem to be inherited. A particularly fruitful area of study is what I would like to call language-internal etymologies. Thinking of the term etymology, the first association will usually be that a word is connected to another word in another language. English hand, for instance, is of course cognate with German Hand, Danish hånd, etc. It is not always possible, however, to link a word to something in a different language, and for a language isolate this will even be impossible for any inherited word as long as its relatives are not known. Nor is there always a need to look at other languages. There is something that is applicable to any language: there are visible etymologies within the language itself. Regarding Modern English, for instance, someone might infer that wo-man may somehow be connected to man. In fact this is correct: the first element was what now is wife, meaning ‘woman’ in general, like its obsolete German counterpart Weib, and man was a designation without any reference to sex, i.e. it was not limited to males. The word once meant “woman-person” literally. Let us consider another example, the Classical Armenian noun anasown ‘animal’. Taking a look at other branches of Indo-European, one could hypothesize that this is a derivative of the verb ‘to breathe’, seen e.g. in Sanskrit ániti ‘he breathes’. There are no obstacles in the phonetic history of Armenian for this proposal. It would even be possible to point to typological parallels for such a derivation: consider e.g. Latin animal ‘animal’ from the verb root mentioned (further nouns: animus ‘spirit’, anima ‘soul’). It would only remain unclear what other morphemes follow the root. However, knowledge of Armenian reveals that there is a prefix a- ‘not’ with an allomorph anbefore vowels, a verb root as- ‘to speak’ and a former present participle suffix -own. This makes it possible to interpret anasown literally as ‘not speaking’, i.e. ‘creature not having speech’; from a typological perspective cf. Modern Greek älogo ‘horse’, literally ‘not supplied with reason or with speech’. Both etymologies have been <?page no="150"?> 5. On the vocabulary of Burushaski 150 suggested in the history of Armenian studies, but the first one has been given up because the second one is clearly superior: while the first proposal only supplies a root, the second one explains all morphemes. The point is that internal analysis of a language pays off. For language isolates, language-internal etymologies are particularly important. As laid out in 4.6., a goal of linguistics is to find relatives for isolates in the long run. For this purpose, beside comparing grammars, one will also have to compare vocabulary items across languages. It is of utmost importance then that the comparisons do not contradict internal evidence of the language isolate. Therefore language-internal etymologies provide a field to engage in, if the topic is handled with care. 46 At issue are not so much all the trivial cases such as synchronically simple compounds or derivations. For instance, Ys. bárcel ‘small watercourse’ consists of bar ‘valley’ and cel ‘water’ (Berger 2008: 62). What is more interesting are the more opaque instances. In the research procedure one can use insights from semantics, typological parallels and areal considerations, for example. All analyses, if correct, may later have a bearing on comparisons to the outside. Research on Burushaski has already brought many cases of language-internal etymologies to light. For example, some colour adjectives have been linked to nouns (Berger 2008: 66), e.g. burúm ‘white’ to burí ‘silver’ (Berger 1974: 135, Tiffou / Pesot 1989: 78, Berger 2008: 66, 99). The adjective may once have meant *‘bright, shining’. For the semantics cf. Tocharian A á rki B á rkwi ‘white’, Hittite 4arki ‘white’, Latin argentum ‘silver’ (Pinault 1989: 39). On the sources of iSqám ‘green, blue’ and of iSkárk ‘yellow’ see Berger (2008: 33) and 3.4. The noun Hz. da6án-us ‘pig’ comes from the adjective Hz. da6án-um ‘thick’ (Berger 1998I: 77, Berger 1998III: 110), and Hz. 6usán-us ‘snake’ comes from Hz. 6usán-um ‘long’ (Berger 1998I: 77, Berger 1998III: 181). Furthermore, there is probably ancient identity between har ‘ox’ and har- ‘to plough’ (Berger 2008: 141). Berger thinks that the verb is older, but there is no need to regard this as the only option, and the derivation could also have been the other way round, i.e. ‘to plough’ could literally be “to ox”. There are further instances that can be added. The stems Ys. cr-, cer- ‘to go’ and cur- ‘to come’ may be connected; this is a simple observation, but I am not aware of a source which points this out. Umlaut, 46 It is not always handled with care by Berger (2008). <?page no="151"?> 5. On the vocabulary of Burushaski 151 on which see 4.2.1., may be involved (and lowering of *i). Both verb paradigms exhibit suppletion. Moreover, the adjectives Suá ‘good’ and Sum ‘bad’ may share the same stem. An idea concerning Hz. tatápal ‘bat’ shall be added not for the sake of insisting on it but in order to mention the possibility. Words for ‘bat’ are sometimes derived from ‘night’, cf. Ancient Greek nukteríß ‘bat’ from núx ‘night’, gen. sg. nukt-óß, Modern Georgian 6am-ura ‘bat’ from 6am-e ‘night’ (-e is a suffix), employing the same suffix as nat-ura ‘lamp’ from nat-eb ‘to shine’, and a different derivative is Old Georgian m-6am-ob-i ‘bat’ (word quoted from Sardshweladse / Fähnrich 2005: 853). Therefore Hz. tatápal ‘bat’ may be connected to thap ‘night’. The hypothesis requires reduplication which could be assumed anyway and a suffix: ta-táp-al. A problem is the absence of aspiration that tatápal exhibits in contrast to thap; onomatopoetic influence could be imagined. The following studies on the vocabulary of Burushaski deal both with language-internal etymologies and with other topics. The sections are concerned with polysemy and homophony (5.2.), the suffix -c (5.3.) and certain kinship terms (5.4.). 5.2. Polysemy and homophony As is well known, one speaks of polysemy if a word has several meanings, and homophony designates the state of affairs when accidentally two or more words have the same phonetic shape. Polysemy and homophony are thus quite different phenomena. The reason why it makes sense to treat them together is that sometimes it is rather difficult or even impossible to tell whether a given case constitutes polysemy or homophony. Especially the fact that Burushaski is a language isolate makes research difficult; with other languages one would clarify at least some cases with the help of language history. It must be remarked, though, that for the native speakers of any language the situation may not be clear in each case either; this is a psycholinguistic question. It may happen that one native speaker feels an example to be polysemy, whereas for a different native speaker it is homophony. The following treatment makes use of typological parallels and areal considerations, where possible, in order to argue for polysemy. In some cases there are good grounds to argue for polysemy. An example is this one: <?page no="152"?> 5. On the vocabulary of Burushaski 152 Ys. 6éndeS ‘gold’ (gender y) - 6éndeS ‘queen’ (feminine gender) Hz. 6éniS ‘gold’ (gender y) - 6éniS ‘queen’ (feminine gender) For the Hunza words Lorimer (1935a: 23) points to Shina son ‘gold’ and so.ni ‘queen’; the latter word is given as soóni by Berger (1998III: 175). Additional arguments appear when considering the plural. For both meanings this is Ys. 6éndeh-a9 / Hz. 6én-a9, and given the fact that Burushaski has a large number of ways to form the plural, the identical formation is probably not fortuitous. Plurals in 9 usually occur exclusively with gender y. Therefore apparently the word’s original meaning was ‘gold’, because this has gender y, and it was then used as ‘queen’ as well, maintaining its plural marking. This fits the fact that in Shina the word for ‘queen’ seems to be derived from ‘gold’, and not vice versa. Ys. hek ‘one’ (z form) and hek ‘full’ are treated as homophones by Berger (1974: 151). Later, however, considering the Hunza equivalents hik ‘one’ (z form) and hik ‘full’, Berger (1998III: 198) pointed out himself that they are probably (“wohl”) identical. In this context attention must be drawn to the Nuristani language Prasun, in which i / u püün means ‘one’ (Edelman 1983: 118, written pün with a macron). Given that many Indo- Iranian languages use words which are etymologically identical with Sanskrit ékaH ‘one’, this deviation is remarkable. I would like to argue that Prasun püün is etymologically the word for ‘full’ (Sanskrit p ú rNáH ‘full’, Latin pl é nus, Gothic fulls, etc.) and that ‘one’ was designated as “a full entity” in Prasun. If now ‘one’ and ‘full’ bear a relationship in a geographically close language, this speaks for polysemy in Burushaski. Berger (1974: 173) lists two homophones sa ‘sun’ and sa ‘month’. There may have been an original polysemy here as well. First of all, there is an obvious connection of ‘month’ to ‘moon’. In some languages they are identical, e.g. Estonian kuu ‘moon; month’, Mingrelian tuta ‘moon; month’, Swahili mwezi ‘moon; month’, i.e. a polysemy exists. In other languages one of the two is derived from the other or compounded with the other; this applies to many Germanic languages, including English: moon, mon-th. In Burushaski, ‘moon’ is halánc today. It is possible, however, that sa ‘month’ goes back to an earlier word which meant both ‘moon’ and ‘month’. A development of this kind is found in Latin: m é nsis ‘month’ is the former word for ‘moon’ and ‘month’, cf. Latvian m é nes-s ‘moon’, m é nes-is ‘month’ (Holst 2001: 114) and other cognates, while Latin l ú na ‘moon’ is an innovation (from the same stem as l ú x ‘light’). Now in some <?page no="153"?> 5. On the vocabulary of Burushaski 153 languages ‘sun’ and ‘moon’ are the same, and there are also beliefs that one deity represents the two. (What possibly contributes to this is that, seen from the earth, the sun and the moon have almost identical size.) It may be added that sa ‘month’ is rarely used; the usual word for ‘month’ is Ys. hésa / Hz. hísa, which Lorimer (1935a: 181) and Berger (1974: 151) analyze as a compound of Ys. hek / Hz. hik (see the preceding paragraph) and sa. In expressions such as Hz. wál-sa ‘four months’ -sa occurs. In many languages ‘mother, mummy’ is something like mama, a “Lallwort” (or “hypocorism”, as Sihler 2000: 257 calls it). Since the mother feeds the baby at the breast, also ‘breast’ is often a similar word, cf. Basque mama, Latin mamma, Turkish meme, all ‘breast’. As already pointed out in 4.5.3., Yasin has two words -mámu ‘female breast’ and mamú ‘milk’. The Hunza forms are -mámut ‘female breast’ and mamú ‘milk’ (Berger 1998III: 277). Despite the u there is clearly a connection to the mama-type. Semantically the Burushaski words are interesting since they introduce another related meaning, ‘milk’, designating what the child receives from the breast. A universal scheme of possible connections can be posited: ‘mother, mummy’ - ‘female breast’ - ‘milk’ In Burushaski we mostly encounter the two latter concepts, although one Hunza informant remembered a word mámo ‘mother, mummy’ (Berger 1998III: 277), and also Hz. máma ‘mother’ can be found (Berger 1998III: 276). Clearly an earlier polysemy existed. The connection was blurred, e.g. due to the use of the personal prefix and the development of stress. While at it, Ys. máncel / Hz. mánchil ‘whey’ shall be discussed. The second element herein is Ys. cel / Hz. chil ‘water’. The first element is not connected to mamú ‘milk’ by Berger (1974: 164); later he does note the possible link but he remains uncertain, putting a question mark (Berger 1998III: 279). It would be possible to leave mánunanalyzed. Berger (1998I: 220) uses the term “halbverdunkelte Komposita” (literally “halfobscured compounds”) for compounds in which only one component can still be identified with an independent word. He points out that these are frequent in Burushaski. In English “cranberry morph” is sometimes used for that part of a word which is synchronically opaque, such as cranin cranberry (Koch 1996: 243). The mánof ‘whey’ would fall under this definition then. However, in my opinion it is highly likely that ‘whey’ contains ‘milk’ in Burushaski, since words for ‘whey’ often have a link to <?page no="154"?> 5. On the vocabulary of Burushaski 154 ‘milk’: French petit-lait ‘whey’ lait ‘milk’, German Molke ‘whey’ - Milch ‘milk’. As to the phonetic development, of course mánmay be assumed to be from *mámsince only an assimilation of the second nasal to the dental affricate needs to be assumed (*m > n). The greater challenge is provided by the lack of the second vowel; this may be due to a syncope. Hz. guCháis both ‘to lie’ and ‘to sleep’ (Berger 1998III: 158). This reminds of Latvian gul é t ‘to lie; to sleep’ (Holst 2001: 217, 218, 227). The Yasin cognate of Hz. guCháis guCá- ‘to lie down to sleep’ (Berger 1974: 146). The original meaning of the verb was probably ‘to lie’ or ‘to lie down’, not ‘to sleep’. Also the older meaning of Latvian gul é t is ‘to lie’, cf. Lithuanian gul \ ti ‘to lie’, but miegóti ‘to sleep’. Note also the cognate set Estonian magada ‘to sleep’ - Finnish maata ‘to lie’. If we are looking for an older root for ‘sleep’ in Burushaski, we may find it in the noun Ys. and Hz. da9 ‘sleep’ (see also 2.2.2.). In the cases treated so far, polysemy was argued for. There are also numerous cases in Burushaski in which homophony should be assumed, or is present with certainty. This applies, for example, when the semantic differences are great: tal ‘key’; ‘birch’; ‘pigeon, dove’ sar ‘hare’; ‘thread’ khen ‘flea’; ‘time’ Homophony is clearly present when the part of speech is different: bái ‘is’ (m.); ‘winter’ gáli ‘he broke’; ‘blanket’ Investigating the available dictionaries and vocabularies of Burushaski leads to the insight that homophony is rather frequent in this language. This is a trait of Burushaski that ought to be recorded. In fact, languages vary in how many homophones they have. English and French are well-known examples from Western Europe for languages with many homophones, and Chinese is an equally well-known instance from Asia. (Of course the writing spelling / characters of the words in question often differs.) It could be argued that Burushaski, with its wealth of homophones, fits Sino-Tibetan from an areal point of view. It is now revealing to consider what the causes for homophony usually are. These are four French words, all phonetically [vER], and their Latin predecessors: <?page no="155"?> 5. On the vocabulary of Burushaski 155 French Latin ver ‘worm’ < vermis verre ‘glass’ < vitrum vers ‘against’ < versus vert ‘green’ < viridis It can be seen that all four words were once different, and sound change caused the homophony. Similarly English homophones (e.g. see and sea, night and knight) can be traced back to earlier forms with different phonetic shapes, and the same applies to Chinese. The point is now what this may mean for Burushaski, the history of which is yet to be unraveled since it is a language isolate. One may speculate whether simply the same cause lies behind the frequent homophony in Burushaski. Although we do not know yet which sound changes Burushaski underwent, the homophones may be a hint that some mergers or losses took place. In a stimulating discussion Fox (1995: 210-214) argues that internal reconstruction usually recovers splits, and it does not enable the researcher to detect mergers. I would add the detail, though, that a special case arises when a language has frequent homophony: this may indeed be due to mergers, although one may remain unable to reconstruct the particulars. 5.3. The suffix -c Burushaski (Ys. and Hz.) gunc ‘day’ has been linked to (Ys. and Hz.) gon ‘dawn (in the morning), daybreak’ (Bengtson 1991a: 98). Disregarding the slight difference in the vowel, this opinion requires a suffix -c. It turns out that such a suffix appears also elsewhere in the language. Berger (1998III: 30) records the words Hz. ba6ú “Doppelarmvoll”, ba6úc “kleines Doppelarmvoll”. These words not only support the existence of the suffix, but they give a hint to the semantics: it may be a diminutive suffix. I would like to argue for a relationship between hun ‘wood’ and hunc ‘arrow’ 47 , which had hitherto not been connected. The background is provided by certain data from the Indo-European language family. The noun which is continued in Sanskrit d3ru ‘wood’, Avestan d á u ru ‘wood’ and Tocharian A, B or ‘wood’, appears in Ancient Greek as dóru 47 According to Berger (2008: 93) hunc is from an older form of Wakhi wuc6, which I do not believe. Apart from various phonetic difficulties, the irregular plural suffix -e used for this word, seen in Ys. hunc-é, Hz. hunz-é ‘arrows’, does not speak for a loanword. <?page no="156"?> 5. On the vocabulary of Burushaski 156 ‘tree-trunk, wooden spear, spear’. Apparently spears were originally made of wood, and in Greek the name for the material was used for the device, too. (Cf. the use of iron in English for the tool to iron clothes.) I assume that a transfer similar to the one in Greek occurred in Burushaski. The ‘arrow’, being smaller than a spear, justifiably obtained a diminutive suffix. Bow and arrow indeed played a role: the Burusho shot at the British invaders with arrows (Schaefer 1978: 40, 42). There is a compound or derivative Ys. humbák, Hz. humák ‘quiver’ (for arrows) which also has to be considered here. (On *mb > m in Hunza see 3.2.2.) Berger (2008: 43, 93) sees the connection of this word to hunc ‘arrow’ and assumes the full form hunc in the word for ‘quiver’; this is a straightforward interpretation but it would imply a loss of -c-. We are provided with an alternative now: it is possible that the word for ‘quiver’ was formed at a time when no -cwas yet present. The relative chronology implied is that ‘wood’ was already used in the sense of ‘arrow’ or ‘spear’, then ‘quiver’ came into being, and only later -c became obligatory in hunc ‘arrow’. Also the noun hunCá ‘wood-pile, stack of wood’ is likely to contain *-c-; -á is probably the same suffix which occurs in various Yasin plural formations (such as huk-á ‘dogs’ from huk ‘dog’, for instance), since many pieces of wood make up the pile. The noun 6anc ‘look, glance’ must be connected to the verb -6án- ‘to look’ (Berger 1974: 147). It can be assumed that first a noun *6an ‘look’ existed, not documented, which had the same stem as the verb. What we only have is that noun which from a historical point of view is the diminutive. The fact that a suffix -c can be detached in several nouns has an important implication: any noun ending in the affricate -c may or may not contain this suffix from a historical point of view, even if synchronic analysis does not permit any segmentation. For example, Ys. xorc / Hz. qhurc ‘dust’ as well as Ys. and Hz. phunc ‘dew’ may contain the suffix, but on the other hand this cannot be proved either. Ys. and Hz. halánc ‘moon’ is a noun for which the probability is high that the suffix is present. Some languages use a historical diminutive for ‘sun’: French sol-eil, similarly some varieties of Rheto-Romance, and the Slavic languages, cf. Polish sLoN-c-e (Holst 2009: 225). If the analysis can apply to ‘sun’, it may also do so to ‘moon’. Indeed there are examples, e.g. Svan do S d-ul ‘moon’ (Klimov 1994: 131). Another example comes from an Albanian dialect. While in Standard Albanian ‘moon’ is hënë, in the dialect of Mandrica, <?page no="157"?> 5. On the vocabulary of Burushaski 157 an Albanian village in Bulgaria, hënë has disappeared and the diminutive hënezë has become the usual word (Sokolova 1983: 172). There are thus typological parallels, which makes it easy to assume a diminutive suffix in Burushaski halánc. The following analysis will arrive at an etymology of Ys. and Hz. -ltánc ‘leg’. As is well known, German Bein ‘leg’, Dutch been ‘leg’ are cognate with English bone, Icelandic bein ‘bone’, etc. The meaning ‘bone’ existed also in German, cf. the collective Ge-bein-e ‘bones, skeleton’, and there are many particular bones with designations which contain -bein as a second element; examples include Schienbein ‘shinbone’, Schlüsselbein ‘collarbone, clavicle’, Jochbein ‘cheek-bone’ and Steißbein ‘coccyx’. Considering these facts, Burushaski -ltánc ‘leg’ may be connected with ‘bone’ in this language: Hz. -ltín, tin, Ys. ten (cf. data in 3.6. and 4.3.2.). The vocalism involved here, a vs. Hz. i and Ys. e, reminds of the alternation which occurs in the inflection of the numeral ‘one’: Hz. and Ys. han (x and y), Hz. hin / Ys. hen (m and f), Hz. hik / Ys. hek (z). As seen in 4.3.2., -ltis a former dual prefix. Klimov / Èdel'man (1972: 161) write about Hz. -ltín ‘bone’: “  ,      -   -    .” “possibly it was originally linked to the designation of some sort of paired bones.” It can be seen now that the connection to ‘leg’ is no obstacle for this reasonable statement. The plural of -ltán-c ‘leg’ is -ltá-i9 (both Ys. and Hz.), and this confirms once more that -c is a suffix, since it is absent from the plural. The loss of n in the plural is the same as in -rén, pl. -réi9; it has to do with the alternation n / y and subsequent loss of y, cf. 3.7.2., 4.2.2. This leads to an important interrelated topic: indeed -c can occur in only part of a paradigm. In the pair mentioned it is only present in the singular: sg. pl. -ltán-c -ltá-i9 ‘leg’ In the following singular / plural pairs from Hunza, in contrast, -c occurs only in the plural: 48 sg. pl. -lCin -lCum-u-c ‘eye’ bárSun bárSum-u-c ‘wooden sieve’ 48 On the nasal alternation n / m see 2.2.1., on the vowel alternation i / u see 4.2.1. <?page no="158"?> 5. On the vocabulary of Burushaski 158 Similarly with ‘eye’ in Nager: sg. pl. -lCin -lCim-u-c ‘eye’ Compare this to Yasin, however, where ‘eye’ has no -c in any number: sg. pl. -lCi -lCim-u ‘eye’ Suppose now the identification as diminutive suffix is correct. The German term teildeminutive Paradigmen has recently been coined (Holst 2009: 224, 241-244); it refers to paradigms in which some forms are diminutives from a historical point of view and others are not, e.g. Greek guný ‘woman’, nom. pl. guna-îk-eß (Holst 2009: 243f.). It seems that some such paradigms exist also in Burushaski. 5.4. On some kinship terms A challenge for linguistic analysis is provided by four Burushaski nouns designating persons which can be arranged in a table as follows: hir ‘man’ -skir ‘father-in-law’ gus ‘woman’ -skus ‘mother-in-law’ It has been pointed out that the nouns with ‘-in-law’ contain the simple nouns, which is doubtlessly correct. Moreover, researchers often argue that the -sof the nouns with ‘-in-law’ is probably the remnant of -us ‘wife’. Beyond that things become more uncertain. In the second line one may identify an assimilation *g > k after s. For -skir Tikkanen (2001: 479) deems it possible that it received k by analogy with -skus, and Berger (2008: 141) agrees with this view. However, this is not the only option. It may also be that the first line of the table shows an instance of a mutation, and indeed Tiffou (1999: 124) expresses that this is possible, see 2.2.2. Despite all efforts, however, questions remain. I would like to propose still another analysis of the nouns with ‘-in-law’ here: they may contain the suffix -iski, -ski, which forms possessive adjectives, as an additional element that did not occur in previous analyses. The hypothesis would then demand some contractions: -skir ‘father-in-law’ < -us ‘wife’ + -ski + hir ‘man’ -skus ‘mother-in-law’ < -us ‘wife’ + -ski + gus ‘woman’ If this idea should be correct, the words would not exhibit any mutations. <?page no="159"?> 5. On the vocabulary of Burushaski 159 Another issue in kinship terminology is provided by the terms for ‘brother’ and ‘sister’. English has just two terms, namely brother and sister; they express the sex of the sibling and nothing else. Many languages, however, arrive at more terms because they encode certain distinctions; the semantic systems have been investigated by Nerlove / Romney (1967). The matter is in fact a question of typology, since typology is not confined to phonetics and grammar, but it can also be applied to the vocabulary or to semantics. Languages may express, for example, whether a sibling is older or younger; i.e. there are four words with the meanings ‘older brother’, ‘younger brother’, ‘older sister’ and ‘younger sister’. Such a system occurs in Hungarian, for instance, and it can be presented in a table as follows: older younger brother bátya öcs sister nÖvér húg Further examples are provided by Chinese (Goddard 2005: 76) and by some languages which have been counted as “Altaic”. There is another type of system in which the sex of the referent plays a role. This means that the system distinguishes, for instance, whether someone is the brother of a male person or the brother of a female person. Such a system is found in Basque: of a male of a female brother anaia neba sister arreba ahizpa The same type of system is present in Svan, a Kartvelian language (Gamkrelidze / Macavariani 1982: 22): of a male of a female brother muxwbe &umil sister da C wir udil Turning now finally to Burushaski, the system is similar to the one of Basque and Svan, but there is one term less; data from Yasin: of a male of a female brother -c6o -húles sister -yást -c6o <?page no="160"?> 5. On the vocabulary of Burushaski 160 The Hunza forms are: of a male of a female brother -c6o -úlus sister -yás -c6o As can be observed, there is a single word for a brother of a male and a sister of a female: -c6o (in Yasin also -c6u exists). A possible translation for this word would be ‘sibling of the same sex (as the referent)’. When a man or boy utters á-c6o, this would be rendered in English as ‘my brother’, whereas the same form spoken by a woman or girl would be translated as ‘my sister’. For a brief typological comparison of the Basque and Burushaski systems see also Tiffou (1995b: 160). The Burushaski terms are inalienable nouns, and this entails that they require a possessive prefix. If the possessor is a third person, due to the gender system certain combinations of possessors and stems are lacking. There is a form muhúles ‘her brother’, for instance, but it would not make sense to try to combine the prefix i- ‘his’ and -húles. The word -húles may be related to helés ‘boy’, if umlaut is involved. Furthermore, one may speculate whether there is another relationship to hir ‘man’, pl. hurí. The other two terms, -c6o and -yást, do not seem to have any links within Burushaski. This leads to the final question: who are the brothers and sisters of the Burushos; whose language is akin to their language? There were enough other questions to be investigated in the present book. The issue will have to be clarified by future research. <?page no="161"?> 161 References Abaev, Vasilij I. (1949): Osetinskij jazyk i fol'klor. Moskva / Leningrad: Izdatel'stvo Akademii Nauk SSSR. Amha, Azeb (2012): Omotic. In: Frajzyngier, Zygmunt / Shay, Erin (eds.): The Afroasiatic languages. Cambridge: Cambridge University Press. 423-504. Anderson, Gregory D. S. (1997): Burushaski phonology. In: Kaye, Alan S. (ed.): Phonologies of Asia and Africa (including the Caucasus). Vol. 2. Winona Lake: Eisenbrauns. 1021-1041. Anderson, Gregory D. S. (2007): Burushaski morphology. In: Kaye, Alan S. (ed.): Morphologies of Asia and Africa. Vol. 2. Winona Lake: Eisenbrauns. 1233-1275. Anderson, Gregory D. S. (2009): Burushaski. In: Brown / Ogilvie (2009). 175- 179. Anderson, Gregory D. S. / Eggert, Randall (2001): A typology of verb agreement in Burushaski. In: Linguistics of the Tibeto-Burman Area 24 / 2. 235-254. Arlotto, Anthony (1972): Introduction to historical linguistics. Boston: Houghton Mifflin Company. Backstrom, Peter C. (1992): Burushaski. In: Backstrom / Radloff (1992). 31-54. Backstrom, Peter C. / Radloff, Carla F. (eds.) (1992): Sociolinguistic survey of Northern Pakistan. Vol. 2: Languages of northern areas. Islamabad: National Institute of Pakistan Studies, Quaid-i-Azam University / Summer Institute of Linguistics. Bashir, Elena (1985): Toward a semantics of the Burushaski verb. In: Zide, Arlene R. K. / Magier, David / Schiller, Eric (eds.): Proceedings of the Conference on Participant Roles: South Asia and adjacent areas. Bloomington: Indiana University Linguistics Club. 1-32. Bashir, Elena (2000): A thematic survey of Burushaski research. In: History of Language 6 / 1. 1-14. Bengtson, John (1991a): Notes on Sino-Caucasian. In: Shevoroshkin (1991). 67- 129. Bengtson, John (1991b): Macro-Caucasian phonology (Part I). In: Shevoroshkin (1991). 142-161. Bengtson, John D. (1997): Ein Vergleich von Buruschaski und Nordkaukasisch. In: Georgica 20. 88-94. Bengtson, John D. (2008): The languages of northern Eurasia: inference to the best explanation. In: Bengtson, John D. (ed.): In hot pursuit of language in <?page no="162"?> References 162 prehistory. Essays in the four fields of anthropology in honor of Harold Crane Fleming. Amsterdam / Philadelphia: Benjamins. 241-262. Berger, Hermann (1956): Mittelmeerische Kulturpflanzennamen aus dem Burusaski. In: Münchener Studien zur Sprachwissenschaft 9. 4-33. Berger, Hermann (1959): Die Burusaski-Lehnwörter in der Zigeunersprache. In: Indo-Iranian Journal 3. 17-43. Berger, Hermann (1974): Das Yasin-Burushaski (Werchikwar). Grammatik, Texte, Wörterbuch. Wiesbaden: Harrassowitz. Berger, Hermann (1985): A survey of Burushaski studies. In: Journal of Central Asia 8. 33-37. Berger, Hermann (1992): Das Burushaski - Schicksale einer zentralasiatischen Restsprache. Sitzungsberichte der Heidelberger Akademie der Wissenschaften. Philosophisch-historische Klasse. Heidelberg: Winter. Berger, Hermann (1998): Die Burushaski-Sprache von Hunza und Nager. Vol. I: Grammatik. Vol. II: Texte mit Übersetzungen. Vol. III: Wörterbuch. Wiesbaden: Harrassowitz. Berger, Hermann (2004): Reconstruction d’anciennes formes linguistiques en bourouchaski. In: Tiffou (2004b). 13-16. Berger, Hermann (2008): Beiträge zur historischen Laut- und Formenlehre des Burushaski. Wiesbaden: Harrassowitz. Beyer, Stephan V. (1992): The Classical Tibetan language. Albany: State University of New York Press. Biddulph, John (1880): Tribes of the Hindoo Koosh. Calcutta: Office of the Superintendant of Government. (Reprint Graz 1971.) Bielmeier, Roland (1977): Historische Untersuchung zum Erb- und Lehnwortschatzanteil im ossetischen Grundwortschatz. Frankfurt am Main: Peter Lang. Blake, Barry J. (1994): Case. Cambridge: Cambridge University Press. Bleichsteiner, Robert (1930): Die werschikisch-burischkische Sprache im Pamirgebiet und ihre Stellung zu den Japhetitensprachen des Kaukasus. In: Wiener Beiträge zur Kulturgeschichte und Linguistik 1. 289-331. Bouda, Karl (1931): Review of Bleichsteiner (1930). In: Caucasica 9. 141-143. Bouda, Karl (1950): Die Sprache der Buruscho. In: Eusko-Jakintza / Revue d’études basques 4. 37-50, 337-346. Braunmüller, Kurt (1991): Die skandinavischen Sprachen im Überblick. Tübingen: Francke. <?page no="163"?> References 163 Brown, Keith / Ogilvie, Sarah (eds.) (2009): Concise encyclopedia of languages of the world. Oxford: Elsevier. Campbell, Lyle (1997): American Indian languages. The historical linguistics of Native America. New York / Oxford: Oxford University Press. Campbell, Lyle (1998): Historical linguistics. An introduction. Edinburgh: Edinburgh University Press. Campbell, Lyle / Grondona, Verónica (2007): Internal reconstruction in Chulupí (Nivaclé). In: Diachronica 24. 1-29. Campbell, Lyle / Poser, William (2008): Language classification. History and method. Cambridge: Cambridge University Press. Casule, Ilija (1998): Basic Burushaski etymologies. The Indo-European and Paleo-Balkanic affinities of Burushaski. München: Lincom Europa. Casule, Ilija (2009): Burushaski numerals of Indo-European origin. In: Central Asiatic Journal 53 / 2. 163-182. Collinge, Neville E. (1985): The laws of Indo-European. Amsterdam: Benjamins. Comrie, Bernard (1978): Ergativity. In: Lehmann, Winfred P. (ed.): Syntactic typology. Studies in the phenomenology of language. Sussex: Harvester Press. 329-394. Comrie, Bernard (1981): The languages of the Soviet Union. Cambridge: Cambridge University Press. Comrie, Bernard (2003): Recipient person suppletion in the verb ‘give’. In: Wise, Mary Ruth / Headland, Thomas N. / Brend, Ruth M. (eds.): Language and Life. Essays in memory of Kenneth L. Pike. Dallas: SIL International and The University of Texas at Arlington. 265-281. Corbett, Greville (2000): Number. Cambridge: Cambridge University Press. Crowley, Terry / Bowern, Claire (2010): An introduction to historical linguistics. 4th edn. Oxford: Oxford University Press. Cunningham, Alexander (1854): Lad á k, physical, statistical and historical; with notes on the surrounding country. London: Sagar Publications. Cysouw, Michael (2003): The paradigmatic structure of person marking. Oxford: Oxford University Press. Deeters, Gerhard (1963): Die kaukasischen Sprachen. In: Spuler, Bertold (ed.): Handbuch der Orientalistik. Abteilung 1, Band 7: Armenisch und kaukasische Sprachen. Leiden: Brill. 1-79. <?page no="164"?> References 164 DeLancey, Scott (2003): Lhasa Tibetan. In: Thurgood, Graham / LaPolla, Randy (eds.): The Sino-Tibetan languages. London: Routledge. 270-288. Demiraj, Shaban (1993): Historische Grammatik der albanischen Sprache. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Dimmendaal, Gerrit J. (2000): Morphology. In: Heine, Bernd / Nurse, Derek (eds.): African languages. An introduction. Cambridge: Cambridge University Press. 161-193. Dirr, Adolf 1928. Einführung in das Studium der kaukasischen Sprachen. Leipzig: Asia major. Dixon, Robert M. W. (1979): Ergativity. In: Language 55. 59-138. Dixon, Robert M. W. (1994): Ergativity. Cambridge: Cambridge University Press. Driem, George van (2001): The languages of the Himalayas. Vol. II. Leiden: Brill. Edelman, Dzoj I. (1983): The Dardic and Nuristani languages. Moscow: Nauka. Edelman, Dzoj I. (1999): On the history of non-decimal systems and their elements in numerals of Aryan languages. In: Gvozdanović, Jadranka (ed.): Numeral types and changes worldwide. Berlin / New York: Mouton de Gruyter. 221-241. Fox, Anthony (1995): Linguistic reconstruction. An introduction to theory and method. Oxford: Oxford University Press. Gamkrelidze, Thomas V. / Macavariani, Givi I. (1982): Sonantensystem und Ablaut in den Kartwelsprachen. Eine Typologie der Struktur des Gemeinkartwelischen. Tübingen: Narr. Georg, Stefan / Volodin, Alexander P. (1999): Die itelmenische Sprache. Grammatik und Texte. Wiesbaden: Harrassowitz. Goddard, Cliff (2005): The languages of East and Southeast Asia. Oxford: Oxford University Press. Greenberg, Joseph H. (1963): The languages of Africa. Bloomington: Indiana University Press. Greenberg, Joseph H. (1978): Generalizations about numeral systems. In: Greenberg, Joseph H. (ed.): Universals of human language. Vol. 3: Word structure. Stanford: Stanford University Press. 249-295. Grolier, Eric de (1995): Comment on Trask’s critique of “Dene-Caucasian- Basque relationship”. In: Mother Tongue 1. 111-114. <?page no="165"?> References 165 Grune, Dick (1998): Burushaski: an extraordinary language in the Karakoram mountains. Pontyporidd: Joseph Biddulph Publisher. Haas, Mary R. (1965): Is Kutenai related to Algonkian? In: Canadian Journal of Linguistics 10. 77-92. Harris, Alice C. / Xu, Zheng (2009): Diachronic morphological typology. In: Brown / Ogilvie (2009). 287-293. Haspelmath, Martin (2002): Understanding morphology. London: Arnold. Heine, Bernd / Claudi, Ulrike / Hünnemeyer, Friederike (1991): Grammaticalization. A conceptual framework. Chicago / London: The University of Chicago Press. Hewitt, George (2004): Introduction to the study of the languages of the Caucasus. München: Lincom Europa. Hock, Hans Henrich (1986): Principles of historical linguistics. Berlin: Mouton de Gruyter. Holmer, Nils M. (1947): Ibero-Caucasian as a linguistic type. In: Studia Linguistica 1. 11-44. Holst, Jan Henrik (2001): Lettische Grammatik. Hamburg: Buske. Holst, Jan Henrik (2005): Einführung in die eskimo-aleutischen Sprachen. Hamburg: Buske. Holst, Jan Henrik (2008): Reconstructing the mutation system of Atlantic. Neuried: Ars Una. Holst, Jan Henrik (2009): Armenische Studien. Wiesbaden: Harrassowitz. Holst, Jan Henrik (2011): Forschungsfragen zu den mesoamerikanischen Indianersprachen. In: Amerindian Research 6 / 3. 164-172. Holst, Jan Henrik (2013): Wie die Inuit die Dinge beim Namen nennen: Zu Semantik und Onomasiologie ihrer Sprache. Amerindian Research 8 / 4. 261- 263. Iosad, Pavel (2010): Right at the left edge: initial consonant mutations in the languages of the world. In: Wohlgemuth, Jan / Cysouw, Michael (eds.): Rethinking universals. How rarities affect linguistic theory. Berlin / New York: Mouton de Gruyter. 105-137. Jacobsen, William H. (1995): Comment on R. L. Trask’s “Basque and Dene- Caucasian: A critique from the Basque side”. In: Mother Tongue 1. 120-142. <?page no="166"?> References 166 Joseph, Brian (2010): Internal reconstruction. In: Luraghi, Silvia / Bubenik, Vit (eds.): The Continuum companion to historical linguistics. London / New York: Continuum. 52-58. Karst, Josef (1901): Historische Grammatik des Kilikisch-Armenischen. Straßburg: Trübner. Klimov, Georgij A. (1994): Einführung in die kaukasische Sprachwissenschaft. Hamburg: Buske. Klimov, Georgij A. / Èdel'man, Dzoj I. (1972): K nazvanijam parnych castej tela v jazyke burusaski. In: Ètimologija 1972. 160-162. Klimov, Georgij A. / Èdel'man, Dzoj I. (1995): K perspektivam rekonstrukcii istorii izolirovannogo jazyka (na materiale jazyka burusaski). In: Voprosy Jazykoznanija 1995 / 5. 27-38. Klingenschmitt, Gert (1994): Die Verwandtschaftsverhältnisse der indogermanischen Sprachen. In: Rasmussen, Jens Elmegård (ed.): In honorem Holger Pedersen. Kolloquium der indogermanischen Gesellschaft vom 25. bis 28. März 1993 in Kopenhagen. Wiesbaden: Reichert. 235-251. Koch, Harold (1996): Reconstruction in morphology. In: Durie, Mark / Ross, Malcolm (eds.): The comparative method reviewed. Regularity and irregularity in language change. New York / Oxford: Oxford University Press. 218-263. Krahe, Hans (1961): Germanische Sprachwissenschaft. Vol. II: Formenlehre. 4th edn. Berlin: De Gruyter. Kühnel, Helmut (1999): Wörterbuch des Baskischen. Wiesbaden: Reichert. Laanest, Arvo (1982): Einführung in die ostseefinnischen Sprachen. Hamburg: Buske. Leitner, Gottlieb W. (1893): The Hunza and Nagyr handbook. Being an introduction to a knowledge of the language, race, and countries of Hunza, Nagyr, and a part of Yasin. Woking: Oriental University Institute. Lorimer, David L. R. (1935a, 1935b, 1938): The Burushaski language. Vol. I: Introduction and grammar. Vol. II: Texts and translations. Vol. III: Vocabularies and index. Oslo: Instituttet for sammenlignende kulturforskning. Lorimer, David L. R. (1962): Werchikwar English vocabulary (with a few Werchikwar texts). Oslo: Instituttet for sammenlignende kulturforskning. Lyovin, Anatole V. (1997): An introduction to the languages of the world. New York / Oxford: Oxford University Press. MacAulay, Donald (ed.) (1992): The Celtic languages. Cambridge: Cambridge University Press. <?page no="167"?> References 167 Maddieson, Ian (1984): Patterns of sounds. Cambridge: Cambridge University Press. Masica, Colin P. (1991): The Indo-Aryan languages. Cambridge: Cambridge University Press. Morgenstierne, Georg (1935): Preface. In: Lorimer 1935a. VII-XXX. Morgenstierne, Georg (1945): Notes on Burushaski phonology. In: Norsk Tidsskrift for Sprogvidenskap 13. 61-95. Morin, Yves-Charles / Dagenais, Louise (1977): Les emprunts ourdous en bourouchaski. In: Journal Asiatique 265. 307-343. Morin, Yves-Charles / Tiffou, Étienne (1988): Passives in Burushaski. In: Shibatani, Masayoshi (ed.): Passive and voice. Amsterdam / Philadelphia: Benjamins. 493-524. Morin, Yves Charles / Tiffou, Étienne (1989): Dictionnaire complémentaire du bourouchaski du Yasin. Paris: Peeters / Selaf. Munshi, Sadaf (2010): Contact-induced language change in a trilingual context: The case of Burushaski in Srinagar. In: Diachronica 27. 32-72. Nerlove, Sara / Romney, A. Kimball (1967): Sibling terminology and cross-sex behavior. In: American Anthropologist 69 / 2. 179-187. Parkin, Robert J. (1987): Tibeto-Burman and Indo-European loans in Burushaski kinship terminology. In: Bulletin of the School of Oriental and African Studies 50. 325-329. Pinault, Georges-Jean (1989): Introduction au tokharien. In: Chanet, Anne-Marie et al. (eds.): Lalies. Actes des sessions de linguistique et de littérature d’Aussois 7. Paris: Presses de l’école normale supérieure. 5-224. Plank, Frans (ed.) (1979): Ergativity. Towards a theory of grammatical relations. London: Academic Press. Potter, Simeon (1950): Our language. Harmondsworth: Penguin. Rybatzki, Volker (2010): Türkische Lehnwörter im Burushaski. In: Studia Orientalia 108. 149-175. Saltarelli, Mario (1988): Basque. London: Croom Helm. Sardshweladse, Surab / Fähnrich, Heinz (2005): Altgeorgisch-deutsches Wörterbuch. Leiden / Boston: Brill. Schaefer, Hermann (1978): Hunsa. Ein Volk ohne Krankheit. Düsseldorf / Köln: Eugen Diederichs Verlag. <?page no="168"?> References 168 Schmitt, Rüdiger (2000): Die iranischen Sprachen in Geschichte und Gegenwart. Wiesbaden: Reichert. Shevoroshkin, Vitalij (ed.) (1991): Dene-Sino-Caucasian languages. Materials from the First International Interdisciplinary Symposium on Language and Prehistory, Ann Arbor, 8.-12. November 1988. Bochum: Brockmeyer. Shibatani, Masayoshi (1990): The languages of Japan. Cambridge: Cambridge University Press. Sihler, Andrew (2000): Language history. An introduction. Amsterdam: Benjamins. Skyhawk, Hugh van (ed.) (1996): Libi Kisar. Ein Volksepos im Burushaski von Nager. Wiesbaden: Harrassowitz. Sokolova, Bojka (1983): Die albanische Mundart von Mandrica. Berlin: Osteuropa-Institut. Starostin, Sergej A. (1991): On the hypothesis of a genetic connection between the Sino-Tibetan languages and the Yeniseian and North-Caucasian languages. In: Shevoroshkin (1991). 12-41. Steever, Sanford B. (1998): Introduction to the Dravidian languages. In: Steever, Sanford B. (ed.): The Dravidian languages. London: Routledge. 1-39. Suárez, Jorge A. (1983): The Mesoamerican Indian languages. Cambridge: Cambridge University Press. Ternes, Elmar (1990): Initial mutations in Celtic and in West African languages. Synchrony and diachrony. In: Afrika und Übersee 73. 3-17. Ternes, Elmar (1992): The Breton language. In: MacAulay (1992). 371-452. Ternes, Elmar (2011): Neubretonisch. In: Ternes, Elmar (ed.): Brythonic Celtic - Britannisches Keltisch. From Medieval British to Modern Breton. Bremen: Hempen. 431-530. Thomas, Alan R. (1992): The Cornish language. In: MacAulay (1992). 346-370. Tiersma, Peter Meijes (1982): Local and general markedness. In: Language 58. 832-849. Tiffou, Étienne (1993): Hunza proverbs. Calgary: University of Calgary Press. Tiffou, Étienne (1995a): Yasin, vallée heureuse de l’Himalaya. Étude sur les Bourouchos du Yasin (Pakistan septentrional). Paris: Peeters. Tiffou, Étienne (1995b): About Trask: Basque and Dene-Caucasian: A critique from the Basque side. In: Mother Tongue 1. 159-162. <?page no="169"?> References 169 Tiffou, Étienne (1999): Parlons bourouchaski. État présent sur la culture et la langue des Bourouchos. Paris: L’Harmattan. Tiffou, Étienne (2004a): Les priorités de la recherche en bourouchaski. In: Tiffou (2004b). 97-109. Tiffou, Étienne (ed.) (2004b): Bourouchaskiana. Actes du colloque sur le bourouchaski organisé à l’occasion du XXXVIème congrès international sur les études asiatiques et nord-africaines (Montréal, 27 août - 2 septembre 2002). Louvainla-Neuve: Peeters. Tiffou, Étienne / Morin, Yves-Charles (1982): A note on split ergativity in Burushaski. In: Bulletin of the School of Oriental and African Studies 45. 88-94. Tiffou, Étienne / Patry, Richard (1995): La notion de pluralité verbale: le cas du bourouchaski du Yasin. In: Journal Asiatique 283. 407-444. Tiffou, Étienne / Pesot, Jurgen (1989): Contes du Yasin. Introduction au bourouchaski du Yasin avec grammaire et dictionnaire analytique. Paris: Peeters / Selaf. Tikkanen, Bertil (1995): Burushaski converbs in their South and Central Asian areal context. In: Haspelmath, Martin / König, Ekkehard (eds.): Converbs in cross-linguistic perspective. Berlin / New York: Mouton de Gruyter. 487-528. Tikkanen, Bertil (1999): Concerning the typology of Burushaski and the roots of its prefixes dand n-. In: Studia Orientalia 85. 277-300. Tikkanen, Bertil (2001): Burushaski --skir ‘father-in-law’ and --skus ‘mother-inlaw’. In: Studia Orientalia 94. 479-482. Toporov, Vladimir N. (1970): About the phonological typology of Burushaski. In: Jakobson, Roman / Kawamoto, Shigeo (eds.): Studies in general and Oriental linguistics presented to Shirô Hattori. Tokyo: TEC Company, Ltd. 632-647. Toporov, Vladimir N. (1971): Burushaski and Yeniseian languages: some parallels. In: Poldauf, Ivan (ed.): Études de la phonologie, typologie et de la linguistique générale. Travaux Linguistiques de Prague 4. Prague: Academia. 107- 125. Trask, Robert Larry (1996): Historical linguistics. London: Arnold. Trask, Robert Larry (1997): The history of Basque. London / New York: Routledge. Tuite, Kevin (1996): Der Kaukasus und der Hindukusch: eine neue Betrachtung des Beweismaterials für frühe Verbindungen. In: Georgica 19. 92-108. Tuite, Kevin (1998): Evidence for prehistoric links between the Caucasus and Central Asia: the case of the Burushos. In: Mair, Victor H. (ed.): The Bronze <?page no="170"?> References 170 Age and Early Iron Age peoples of Eastern Central Asia. Vol. I: Archeology, migration and nomadism, linguistics. Washington: Institute for the Study of Man. 448-475. Turner, Ralph L. (1966): A comparative dictionary of the Indo-Aryan languages. London: Oxford University Press. Uhlenbeck, Christian Cornelius (1927): Die mit banlautenden Körperteilnamen des Baskischen. In: Boas, Franz et al. (eds.): Festschrift Meinhof. Hamburg: L. Friederichsen & Co. 351-357. Watkins, Calvert (1962): Indo-European origins of the Celtic verb. Vol. I: The sigmatic aorist. Dublin: Institute for Advanced Studies. Welmers, William E. (1973): African language structures. Berkeley / Los Angeles / London: University of California Press. Willson, Stephen R. (1999): Basic Burushaski vocabulary. Islamabad: National Institute of Pakistan Studies, Quaid-i-Azam University / Summer Institute of Linguistics. Zabaltza, Xabier (1995): Comments on R. L. Trask’s article “Basque and Dene- Caucasian: A critique from the Basque side”. In: Mother Tongue 1. 165-171. <?page no="171"?> 171 Indices Subject index Terms which are too frequent (e.g. “vowel”, “verb”) have not been taken up. Note also that many topics can be located with the help of the table of contents. active language 44, 46, 47, 49 allomorph(s) 34, 36, 38, 39, 40, 41, 49, 70, 80, 99, 120, 121, 149 allomorphy 40, 60, 68, 101, 120, 123, 128 analogy 24, 60, 68, 77, 78, 79, 100, 105, 113, 119, 121, 122, 128, 134, 140, 141, 142, 146, 158 apheresis 84 approximant mutation 9, 29, 91 Atharva-Veda 92 Brugmann’s law 120 causative(s) 28, 33, 91, 93 causative prefix 28, 93 circumfixes 46 citation forms 140 closed law 76, 77, 79, 80, 86, 112 colonization 11 colour adjectives 150 comitative 93 compound(s) 17, 79, 92, 121, 125, 128, 150, 153, 156 compounding 31 constituent order(s) 12, 129 control 46, 47 cranberry morph 153 diminutive(s) 156, 157, 158 diminutive suffix 101, 155, 156, 158 dissimilation 25, 126 dissimilatory loss 95 distinctive 136, 137 doublets 96 (n. 31) d-prefix 28, 29, 38, 44, 87, 99 dual 127, 129, 130 dual prefix 23, 91, 92, 93, 157 <?page no="172"?> Indices 172 endangered language 12 epenthesis 85, 86 ergative 12, 44, 48, 49, 82, 93, 99, 114, 131 ergativity 12, 16, 49, 106 fieldwork 11, 17, 48, 54, 55, 79, 95, 112 fluid-S 47, 48, 49 gradation 25 grammaticalization 134 Grassmann’s law 119 halbverdunkelte Komposita 153 heavy 136, 137, 138, 139, 141 hypocorism 153 immobile 136, 141 implicational universal 144 instrumental 93 IPA 27, 58, 114, 144 Lallwort 153 language-internal etymologies 149, 150, 151 lateral affricate 144, 146 leveling 68, 77, 113, 114, 140 light 136, 137, 138, 139, 141, 143, 145 locative 82, 114, 133 metathesis 57, 76, 85, 86, 87 metathesized 144 mobile 136, 137, 141 monophthongization(s) 35, 60, 62, 83, 84, 88, 98 mora(e) 136, 137 Neogrammarian(s) 17, 122 number shifts 101 onomatopoetic 68 (n. 21), 79, 86, 92, 118, 151 oral literature 108 participle(s) 9, 38, 135, 136, 146, 149 prothesis 85, 86 provective mutation 9, 28, 34, 40, 58 reanalysis 96 reanalyzed 25, 100 reduplication 121, 151 relative chronology 33, 67, 70, 79, 80, 84, 86, 108, 122, 144, 156 <?page no="173"?> Indices 173 retroflex 12, 27, 53, 60 (n. 17), 117, 120, 125 sandhi 34, 97 sound correspondence(s) 20, 52, 57, 90 split-S 44, 48, 49 Standard Average European 34 substratum 19 suppletion 151 suppletive 92, 93, 97 syncope 88, 154 syncretism(s) 13, 43 synonyms 90 template 27, 38, 44 time depth 51 (n. 9), 110 time depth indetermination 108, 147, 148 transcription 10, 14 umlaut 113, 114, 115, 124, 145, 146, 150, 160 umlauted 46, 114 univerbation 128, 130, 133 unstressed law 68, 69, 70, 71, 119 uvular(s) 27, 28, 58, 66, 67, 69, 71, 85 uvular law 67, 68, 69, 70, 71, 89, 99 (n. 32) Verner’s law 56, 120 vowel harmony 36, 49, 70, 128 Watkins’ law 105 Watkins’ tendency 105, 106 weight regularity 138, 139, 141, 146 weight shift 139, 140, 141, 143, 144, 145 Winter’s law 120 Language index Specifications such as “Modern”, “Old”, etc. are omitted. Burushaski itself has not been evaluated, being too frequent, but see the word index. With lesser known languages usually the first occurrence in the text provides some information about genetic affiliation or geographic location. Abkhaz 43 Ainu 124, 138 Albanian 25, 87 (n. 29), 113, 120, 156, 157 <?page no="174"?> Indices 174 Amele 39 Arabic 20, 33 Armenian 25, 26, 80, 101, 120, 149, 150 Ashkun 89 Avar 26 Avestan 16, 155 Balti 18, 90 Bariba 123 Bashkarik 89, 114, 145 Basque 15, 16, 51, 62, 99, 124, 127, 128, 129, 130, 148, 153, 159, 160 Breton 27, 28, 39, 113, 129 Bulgarian 145 Chinese 154, 155, 159 Chulupí 147 Cornish 28, 113 Cree 33 Czech 30, 36, 123 (n. 36), 140, 145 Dameli 89 Danish 149 Dutch 42, 157 Efik 123 Elamite 15 English 11, 12, 16, 17, 20, 38, 42, 46, 78, 91, 94, 95, 96 (n. 31), 111, 114, 135, 145, 149, 152, 153, 154, 155, 156, 157, 159, 160 Estonian 26, 152, 154 Faroese 42 Finnish 25, 26, 31, 36, 49, 80, 87 (n. 29), 124, 133, 136, 154 French 11, 12, 17, 31, 56, 92, 96 (n. 31), 136, 154, 155, 156 Frisian 78 Fulfulde 27 Gawar 89 Gawarbati 145 Georgian 151 German 11, 12, 16, 17, 31, 37, 47, 149, 154, 157, 158 Gothic 42, 152 Greek 25, 26, 42, 60, 120, 136, 149, 151, 155, 156, 158 Greenlandic 133 <?page no="175"?> Indices 175 Hattic 148 Hungarian 46, 49, 87 (n. 29), 133, 136, 159 Icelandic 140, 157 Inuit 34, 67, 129 Irish 145 Itelmen 39 Japanese 26, 34, 132 Katarkalai 89 Ket 16 Khowar 14, 18, 56, 57, 61, 89 Koasati 39 Koiari 42 Kurdish 123 (n. 36) Lakhota 43 Latin 19, 30, 33, 36, 92, 96 (n. 31), 114, 124, 135, 136, 137, 139, 149, 150, 152, 153, 154, 155 Latvian 30, 36, 94, 140, 145, 152, 154 Lithuanian 145, 154 Livonian 113 Macedonian 145 Maká 147 Mingrelian 152 Mordvin 120 Nahuatl 144 Nivkh 27 Ossetic 25 Pashai 145 Pashto 120 Persian 20, 36, 86, 91, 96 Platt 42 Polish 84, 95, 136, 145, 156 Portuguese 68, 120 Prasun 152 Quechua 67 Rheto-Romance 156 Rongpo 42 Rushan 115 Russian 11, 15, 26, 84, 120, 136 <?page no="176"?> Indices 176 Sanskrit 19, 20, 76, 83, 92, 102, 108, 111, 114, 115, 123, 149, 152, 155 Serbo-Croat 145 Shina 18, 19, 56, 57, 66, 78 (n. 25), 82, 89, 102, 125, 130, 152 Shughni 115 Sinhalese 114 Somali 26, 88 (n. 30) Spanish 37 Sumerian 148 Svan 113, 156, 159 Swahili 41, 129, 152 Tamil 92, 124 Tarascan 148 Thai 36, 37 Tibetan 18, 32, 33, 47, 48 (n. 7), 89, 114 Tocharian A 83, 120, 150, 155 Tocharian B 120, 150, 155 Turkish 19, 34, 35, 36, 49, 61, 83, 93, 102, 133, 153 Una 42 Urdu 20, 68 (n. 21), 91, 97 Wakhi 18, 59 (n. 15), 90, 155 (n. 47) Waskia 42 Welsh 113 Wolof 27 Word index The index registers a reasonable selection. Hunza equivalents can often be found when consulting the pages with the Yasin words (and sometimes vice versa). Yasin aíSen ‘guest’ 112 altó ‘two’ 97, 117, 123, 125, 126, 127, 142, 143 asúmun ‘star’ 96 aúSin ‘guest’ 82, 112 ba- ‘to be’ 29; bá-i ‘he is’ 29, 36, 83, 154 barc6 ‘bridge’ 62 baT ‘skin’ 26, 29, 91, 96 baxtá ‘sheep’ 58 <?page no="177"?> Indices 177 bo ‘what’ 31, 97 burí ‘silver’ 150 burúm ‘white’ 97, 142, 150 cel ‘water’ 25, 54, 73, 97, 150, 153 cer ‘time, turn’ 31, 121 cer- ‘to go’ 150 c6úrk- ‘to milk’ 111 Cap ‘meat’ 30, 54, 94 Cúmu ‘fish’ 54, 61, 64, 66, 92, 101, 115 dan ‘stone’ 51, 96 den ‘year’ 25, 71 gacér ‘vulture’ 54, 73, 138, 139 garú ‘spring’ 97, 138, 139 gaT ‘knot’ 117, 137, 140 girmín- ‘to write’ 29, 73, 80 gon ‘dawn’ 63, 155 gunc ‘day’ 55, 121, 155 gus ‘woman’ 29, 64, 97, 134, 158 -6án- ‘to look’ 156 6arés- ‘to agree’ 75, 112 6as- ‘to laugh’ 29; 6as-á ‘I laughed’ 44, 45 6aún ‘melon’ 35, 82 6éndeS ‘gold; queen’ 55, 72, 73, 78, 118, 138, 152 6er6éT ‘soft’ 135, 142 6órkun ‘frog’ 65, 115, 116, 138 ha ‘house’ 88, 92 ha6ór ‘horse’ 19, 65 halánc ‘moon’ 94, 121, 138, 152, 156, 157 har ‘ox’ 59, 115, 150 har- ‘to plough’ 62, 150 hésa ‘month’ 74, 79, 153 -híl ‘lip’ 59, 72, 77 hir ‘man’ 23, 31, 72, 79, 94, 111, 113, 114, 141, 158, 160 huk ‘dog’ 64, 92, 156 hukumát ‘government’ 137 humbák ‘quiver’ 55, 156 hun ‘wood’ 57, 64, 155 <?page no="178"?> Indices 178 hunc ‘arrow’ 121, 155, 156 hunCá ‘wood-pile’ 156 huríski ‘manly’ 113 -húTis ‘foot’ 59, 93, 129; a-húTis ‘my foot’ 119 iSqá ‘grass’ 85, 86, 146 ja ‘I’ 93, 142 jer ‘line, row’ 133 -ken ‘liver’ 23, 73, 77, 94 ketáp ‘book’ 20 lán- ‘to be able’ 35, 116 -lCi ‘eye’ 23, 55, 72, 92, 112, 126, 158; á-lCi ‘my eye’ 143, 144 -ltánc ‘leg’ 126, 157; a-ltánc ‘my leg’ 143 -ltúmal ‘ear’ 92, 126; a-ltúmal ‘my ear’ 126, 141, 143 mamú ‘milk’ 143, 153 máncel ‘whey’ 121, 153 mathán ‘far’ 87, 142 mayén ‘old’ 62, 83, 115 men ‘who’ 31, 71, 97, 142 -móqiS ‘face’ 71, 106, 118, 127 multán ‘blood’ 16, 64, 91, 143 -nú9us ‘knee’ 60, 61, 64, 94, 116; a-nú9us ‘my knee’ 137, 141 páqu ‘bread’ 66, 68 pilíli ‘ant’ 89 phen ‘fly’ 73, 75, 76, 116 phéSu ‘pear’ 66, 118 phunc ‘dew’ 55, 64, 156 -rén ‘hand’ 13, 93, 100, 129, 157; a-rén ‘my hand’ 43, 137, 141 -s ‘heart’ 12, 93, 102, 127 sa ‘month’ 152, 153 sa ‘sun’ 92, 96, 134, 152 -skil ‘face’ 13 (n. 1), 73, 111, 127; á-skil ‘my face’ 137, 141 -skir ‘father-in-law’ 31, 158 -skus ‘mother-in-law’ 158 sórqon ‘thunder’ 138, 139 <?page no="179"?> Indices 179 Suá ‘good’ 64, 93, 122, 142, 151 Sum ‘bad’ 151 -t- ‘to do’ 12, 38 tul ‘snake’ 66, 68 tham ‘king’ 99, 137, 140 thánum ‘high’ 89, 116, 142 thap ‘night’ 95, 134, 151 thoS ‘new’ 63, 70, 94, 98, 119 Ti9án ‘egg’ 60, 72, 92, 115 úmur ‘age’ 20, 139 -wál- ‘to fall’ 29, 47; wál-i ‘he fell’ 35, 40, 45, 84, 142 waT ‘bark’ 26, 29, 91 -wáT ‘body, self’ 26 xorc ‘dust’ 58, 65, 156 -xórpet ‘lung’ 58, 65, 81, 98 -yéc- ‘to see’ 34, 44, 88, 96; a-yéc-i ‘he saw me’ 142 -yúhar ‘husband’ 59 -yú9us ‘tongue’ 60, 64, 97, 104, 117; a-yú9us ‘my tongue’ 104, 105 -yúr- ‘to die’ 47, 62, 92, 111; yúr-i ‘he died’ 113 Hunza bácin ‘thigh’ 24 bárSun ‘wooden sieve’ 24, 157 da6ánus ‘pig’ 150 daltás ‘good’ 122, 144 da9 ‘sleep’ 30, 154 gáarc- ‘to flow’ 116 gán-imi ‘he took it’ 32 guChá- ‘to lie; to sleep’ 94, 96, 154 gurgín- ‘to grind’ 29 6éniS ‘gold; queen’ 55, 72, 73, 118, 146, 152 6usánus ‘snake’ 150 ha6úr ‘horse’ 19, 65, 122 hén- ‘to know’ 30, 35, 71, 94 huyéltarc ‘shepherd’ 117 khon ‘ant’ 89, 115 -lpur ‘eyelid’ 127, 128 (n. 38) <?page no="180"?> Indices 180 phárcin ‘cap’ 24, 73 -so ‘kidney’ 90, 102 Sirijón ‘mushroom’ 87, 138 tatápal ‘bat’ 151 tilía9 ‘saddle’ 111 túl- ‘to saddle’ 111 tulpó ‘radish’ 102 Nager álthar ‘twenty’ 102 (n. 33) daltán- ‘to thresh’ 144 gaíltin ‘ankle’ 83 Siríjon ‘mushroom’ 87, 138 ulán- ‘to be able’ 35 <?page no="181"?> Near the Karakorum mountains, an offshoot of the Himalayas, one of the most enigmatic languages of Asia is spoken: Burushaski. This book provides a wealth of investigations into this language. It deals with typology and grammar theory, the comparative method when applied to the dialects, internal reconstruction, as well as vocabulary and semantics. Due to its thorough research and thoughtful discussions, this work contributes significantly to ongoing linguistic debates.