Tone and stress in North-West Indo-Aryan: A survey

Tone and Stress in North-West Indo-Aryan: A Survey Joan L.G. Baart SIL International and University of North Dakota PRE-PUBLICATION DRAFT. 1. Introduction The northwestern corner of the Indo-Aryan language territory is rich in languages that display contrastive lexical tone in some form. In Map 1 below, the oval shape roughly indicates the area where most of these IA tone languages are spoken. Map 1: South and Central Asia. The oval shape roughly indicates where most of the IndoAryan tone languages are spoken. In addition to being tonal, many languages in this area share another feature: they have lost the breathy-voiced consonants that are distinctive of Hindi-Urdu and many other IA languages. Not all languages in this area have lexical tone, but roughly two out of three do. Neither have all these languages lost their breathy-voiced consonants. There are languages with breathy voice as well as tone, languages with breathy voice and no tone, and languages with neither. Still, the typical language in this area has tone and no breathy-voiced consonants. These languages can be divided into three groups on the basis of the number of contrasting tones in a language. Quite a few languages have a three-way tonal contrast: Punjabi, Hindko, Pahari-Pothwari, Gojri, Sansiboli (and some other Rajasthani languages), Bangani, Kangri, and others. The following languages have only two contrasting tones: Shina, Burushaski (not itself an IA language, but on IA territory), Indus Kohistani, Palula, Kundal Shahi, Batera, Domaaki, Ushojo, Khowar, Gawar-Bati, Wotapuri, Chilisso, Gowro, Dameli, Pashai. And finally there are a few languages that have more than three contrasting tones: Kalam Kohistani, Torwali, and Khalkoti. Examples of these various systems of tonal contrast will be presented as we go along. Using the names of the three languages for which we have the most detailed descriptions of tone to date, I use the label “Shina-type tone languages” for IA languages that have a two-way tonal contrast, “Punjabi-type tone languages” for IA languages that have a three-way tonal contrast, and “Kalami-type tone languages” for IA languages with more than three contrasting tones. Geographically, the three groups of languages correspond to three distinct areas on the map, with the small Kalami-type group constituting an enclave within the much larger Shinatype group. This is shown in Map 2. Map 2: North-West Indo-Aryan tone languages. Dots represent a rough geographical centre of gravity for each language. Ovals delineate the three groups. The Kalami-type area constitutes an enclave within the Shina-type area. In what follows we first take a brief look at a number of word-prosodic features of HinduUrdu, the major lingua franca and literary language in the northwestern parts of the SouthAsian subcontinent. Hindi-Urdu is not a tone language itself, but belongs with the tone languages to a larger continuum of related word-prosodic systems. We then go on and look in more detail at the three types of IA tone languages mentioned above. 2. Hindi-Urdu In many cases, the location of word stress in Hindi-Urdu is predictable on the basis of syllable weight and the position of the syllable in the word, according to the following rules (Hussain, 1997): (i) Stress falls on the rightmost heavy syllable in the word; (ii) If there is no heavy syllable, stress falls on the penultimate syllable; (iii) Word-final segments are extrametrical (invisible to the stress rules). A number of examples are listed in (1). Here, extrametrical segments are shown in parentheses, and vowel length is indicated by means of a doubled vowel symbol. (1) is.laa.maa'baa(d) pi'šaa.wa(r) ka'raa.či(i) hi'maa.la.ya(a) 'an.ju.ma(n) 'di.li(i) ‘Islamabad’ ‘Peshawar’ ‘Karachi’ ‘Himalaya’ ‘association’ ‘Delhi’ Sometimes morphology plays a role. The causative suffix -aa, for instance, is inherently stressed and overrules the weight-based system of stress assignment. The minimal pair in (2) illustrates the role of the inherently stressed causative suffix -aa. (2) 'pakaa ‘cooked’ pa'kaa ‘cook!’ In both items, the actual verb root is pak- ‘cook’. The first item is a perfective participle formed with an unstressed suffix -aa. Stress is on the first syllable in accordance with the general rules laid out above. The second item, however, is a causative verb stem, derived with the inherently stressed suffix -aa. (Bare verb stems in Hindi-Urdu and many other IA languages can be used as imperatives, hence the gloss, ‘cook!’.) Perceptually, stress in Hindi-Urdu is noted in the first place through the fact that it provides a landing site for prominence-lending pitch movements that are supplied by sentence intonation (i.e. post-lexical pitch accents). The by far most common shape of these pitch accents is a rise of the kind that is transcribed as L*+H (or simply L*H) in an autosegmentalmetrical notation system (Harnsberger, 1994; Ladd, 2008, p. 87ff and references cited there). Pitch starts out low on the stressed syllable and rises through the syllable boundary, so that it is the syllable following the stressed syllable that has the highest pitch. There have been claims, backed up by some empirical research, to the effect that duration and intensity are not phonetic correlates of lexical stress in Hindi-Urdu (Dyrud, 2001). If these claims can be confirmed, then Hindi-Urdu is to be classified as a non-stress accent system in terms of Beckman’s (1986) phonetic typology. Morpho-phonologically, on the other hand, we do observe a correlation of length and stress in Hindi-Urdu. A phonemically long vowel may become a phonemically short vowel when the syllable within which it occurs loses its stress (is de-stressed). An example is given in (3). (3) baat ‘word’ ba'taa ‘tell!’ The second item in (3) contains the same inherently stressed causative suffix -aa that was seen in (2) above. When combined with the causative suffix, the stem baat is de-stressed and its vowel is shortened. 3. Punjabi-type tone languages “Variations in the tone of the voice form a very remarkable feature of Panjabi pronunciation. There are two special tones, apart from the ordinary tone of speaking. They occur in stressed syllables only.” (Bailey, 1914, p. xv) Many of the Punjabi-type tone languages have lost the breathy-voiced consonants bh, dh, ḍh, ɡh, and others. These have merged with their regular-voiced counterparts, b, d, ḍ, ɡ. In some languages, including standard Punjabi, the breathy-voiced consonants have merged with their voiceless counterparts, p, t, ṭ, k, in word-initial position, while merging with their regularvoiced counterparts in other positions in the word. Breathy voice is characterized by a relatively low rate of vibration of the vocal folds and hence by low pitch. Low pitch, then, was a sub-phonemic feature of the breathy-voiced consonants in an earlier stage of the language. When breathy voice was lost, low pitch stayed around and became phonemic. This phonemic low pitch remained linked to the vicinity of the consonant that lost its breathiness. This low tone (L), in turn, creates a contour with the relatively high pitch (H) associated with word stress. Depending on the position of L tone vis-à-vis stress, this state of affairs results in a rise (LH) or a fall (HL). The location of stress itself is usually predictable in Punjabi-type languages along lines similar to the Hindi-Urdu stress rules. In (4), a minimal triplet is presented from Northern Hindko, which is spoken in Pakistan in the Abbottabad and Mansehra districts and surrounding regions, to the north of the capital Islamabad. (4) Hindko Urdu a. 'kóóṛàà b. 'kooṛaa c. 'kòòṛáá gloss 'kooṛhii ‘leper’ 'kaṛwaa ‘bitter’ 'ɡhooṛaa ‘horse’ In the Hindko words in (4), stress falls on the first syllable, and the same is the case for the Urdu cognates in the column to their right. The Urdu item in (4a) includes the breathy-voiced consonant ṛh, while the one in (4c) includes the breathy-voiced consonant ɡh. Hindko does not have breathy-voiced consonants, so the breathy-voiced ṛh in Urdu corresponds to a plain ṛ in Hindko. At the same time, the breathy-voiced ɡh in Urdu corresponds to a voiceless k in Hindko (in word-initial position). However, while the breathy voicing seen in Urdu does not carry over to Hindko, the low pitch associated with breathy voice does carry over to Hindko, causing a falling contour (HL) on the Hindko item in (4a), and a rising contour (LH) on the Hindko item in (4c). The item in (4b) is not associated with a low pitch stemming from historical breathy voice; in a sentence-medial position this word is pronounced on a fairly level or slightly rising pitch. In summary, stress placement in Punjabi, Northern Hindko, and in several other Punjabitype tone languages is determined by syllable structure and morphology, in a way that is very similar to Hindi-Urdu stress placement. Furthermore, in some roots a mora is pre-associated in the lexicon with an L tone. Usually this is a mora in the immediate vicinity of a historical breathy-voiced consonant. Finally, a default H tone is assigned to the nucleus of a stressbearing syllable. The resulting tonal configurations are LH (when lexical L tone precedes stress), HL (when lexical L tone follows stress), and H (when there is no lexical L tone). This kind of interaction of the location of stress and lexical tone, producing different contrastive tonal configurations, can be seen at work in synchronic processes in Punjabi-type tone languages. To illustrate, a number of Gojri verb stems are listed in (5). Gojri is the mother tongue of around one million Gujars who live in the northern, mountainous parts of Afghanistan, Pakistan and India. Traditionally they are nomadic pastoralists, whereas nowadays significant numbers of them are settled or semi-settled agriculturalists (Losey, 2002, p. 1). (5) p ṛ ‘read!’ b n ‘tie’ s mǰ ‘understand!’ The final consonants in the items in (5) are historically, but not currently, breathy voiced, as we can still see in related Urdu words such as paṛh ‘read!’, bandh ‘tied’, samaǰh ‘understand!’. In isolation, the words in (5) are spoken with falling pitch contours. Like Hindi-Urdu, Gojri has an inherently stressed causative suffix -aa. When this suffix is attached to the forms in (5), stress shifts to the suffix and a default H tone now follows, rather than precedes, the lexical L tone associated with the historical breathy-voiced consonant, as is shown in (6). Accordingly, rising contours are produced instead of falling contours (Losey, 2002, p. 67ff). (6) pa'ṛ ba'n sam'ǰ ‘teach!’ ‘get (someone) to tie!’ ‘get (someone) to understand!’ 4. Shina-type languages Gilgiti Shina is spoken in the Gilgit-Baltistan territory in northern Pakistan, in the town of Gilgit and the wider area around Gilgit. The place of stress in Gilgiti Shina simplex stems is not entirely predictable, but the following description may serve as a rule of thumb (Radloff, 1999, p. 65f.): (i) A long vowel attracts stress (there is usually not more than one long vowel in a stem). (ii) If there is no long vowel, then stress falls on the rightmost heavy syllable. (iii) If there is no heavy syllable, then stress falls on the penultimate syllable. Stress in Shina correlates with relatively high pitch on the vowel in the syllable that is stressed, while absence of stress correlates with relatively low pitch. A particular feature of Gilgiti Shina, and of Shina-type tone languages in general, is the fact that on long vowels (sometimes on bimoraic syllables in general) a contrast may occur between a falling and a rising pitch contour. In the literature on Shina it is generally assumed that the stress-bearing unit in Shina is the mora rather than the syllable, and that a rising tone on long vowels is due to stress on the second mora in the syllable, while a falling tone is due to stress on the first mora in the syllable. Examples of the various possibilities for the placement of mora stress in Gilgiti Shina words are shown in (7). Two minimal pairs are presented in (8). The Shina data cited here and below have been taken from Radloff’s (1999) work. (7) bambula karka muš kh oǰan turm k t ši bul ‘male cat’ ‘hen’ ‘inquiry’ ‘gun’ ‘roof’ ‘polo’ (8) k či ‘near’ k am ‘relative’ kač ‘scissors’ ka m ‘a vegetable’ Another distinctive feature that occurs in Gilgiti Shina, as well as in a range of other Shinatype languages (e.g. Palula, Ushojo, Kundal Shahi), is that of stress advancement, where stress shifts from the stem to a suffix (that is itself not inherently stressed). More precisely, stress shifts from a short vowel, or from the second mora of a long vowel, onto the first mora of an immediately following suffix. This is a fairly general process in these languages, although there are several classes of exceptions as well. A few examples from Gilgiti Shina where stress shifts to a suffix are given in (9). (9) k ṇ ‘ear’ ad t ‘Sunday’ ḍe r ‘belly’ koṇ ṭ ‘to ear’ adit y ‘of Sunday’ ḍer ǰo ‘from belly’ In the case of the examples in (10), stress does not shift away from the root as there is an intervening mora between the stressed mora and the suffix. (10) luṣṭ ak ‘tomorrow’ m ɡar ‘goat’ luṣṭ akeṭ ‘to tomorrow’ m ɡareṭ ‘to goat’ If stress shifts away from a long vowel, this vowel may be shortened or even deleted, as in (11) where the long vowel in the stem is shortened or deleted altogether in the inflected form. This reduction process is also seen at work in the third example in (9) above. (11) ɡili t ‘Gilgit’ ɡilit ṭ/ɡilt ṭ ‘to Gilgit’ In Gilgiti Shina, several verbal suffixes are inherently stressed. Examples are - iǰ (passive) and -o k (infinitive). These suffixes may be attached to verb stems that are themselves inherently stressed. In such cases, the rightmost stress supersedes the others. The other stresses are deleted and that may once again be accompanied by vowel shortening, as illustrated in (12). Note how the long vowel in a r is shortened to ar when - iǰ is attached, and how the latter is shortened in turn when -o k is attached. (12) ča r ‘graze’ čar iǰ ‘be grazed’ čariǰo k ‘to be grazed’ Shina-type languages appear to lend themselves well to an analysis in terms of mora stress, along the lines illustrated just above for Gilgiti Shina. Indeed, such an analysis was first introduced many years ago by Hermann Berger (1960) for Burushaski and applied to Shina by Buddruss (e.g. 1993, 1996), Schmidt and Kohistani (2008), and others, and to Palula by Liljegren (2008). In fact, the concept of mora stress goes back quite a bit further than its application by Berger to Burushaski. In a discussion of Indo-European accent and the practice of dividing languages into those with “dynamic accent” and others with “musical” or “pitch accent”, Szemerénye (1999:73) remarks: Jakobson has shown that the essential difference between the two traditional types of accent is that in the one the extent of the accent is equal to the duration of the whole syllabic phoneme, while in the other the accent affects only a part of the syllable, the mora. The former type is perceived as an accent of intensity, as in English or Russian. The other can be illustrated from Lithuanian, in which a long vowel, which consists of two morae, can take an accent either on the first mora (falling accent) or on the second mora (rising accent). Lithuanian accent, then, is strikingly similar to the rising vs. falling contrast seen in Shina; and like Shina, Lithuanian has been analyzed in terms of mora stress. The process of stress advancement constitutes a further similarity between Shina and Lithuanian. According to “Saussure’s law,” stress is advanced in Lithuanian from a short or circumflex (rising) syllabic nucleus to an immediately following acute one (Collinge, 1985, p.149). The first part of the structural condition of Saussure’s law (stress shifts from a syllable whose rime is short or has rising pitch) matches exactly the structural condition for stress advancement in Shina. A difference between Gilgiti Shina and Lithuanian is that in Gilgiti Shina the rising vs. falling contrast is possible on long vowels only, whereas in Lithuanian the contrast may also occur on a combination of a short vowel and a sonorant consonant. However, we do have Shina-type tone languages that agree with Lithuanian in this respect. Indus Kohistani, for example, has minimal pairs such as k l ‘grain’ vs. k l ‘people’ (Zoller, 2005, p. 25). A question that needs to be addressed at some point is whether in the lexical representation of Shina words, a particular mora is marked as stressed, or whether a better approach is one where a particular mora is pre-linked with an H tone. Under the second approach we would abandon the notion of mora stress and explain Shina word prosody in tonal terms only. For a tonal approach the challenge is to account for the stress shift data and the accompanying deletion of a de-stressed mora with its associated segmental material shown in (11) and (12) above. When prosody affects segments, which is what is happening in these cases, we normally assume that stress is involved rather than tone, following Hyman’s insight that “Tones affect tones. Tones are not expected to affect consonants or vowels or cause any of the mutations affiliated with stress” (2001:1378). At the same place, however, he also says, “The stress-bearing unit is the syllable. Any metrical structure built up out of units smaller than the syllable, e.g. the mora, should not be viewed as stress.” This, then, leaves us with a bit of a dilemma, a dilemma that I will leave unresolved for now. 5. Kalami-type languages As most of the research reported in this section is based on my own fieldwork, I would like to insert a brief autobiographical note. The discovery of a complex tone system in Kalam Kohistani was a total surprise. In SIL International, married couples who move overseas to be involved in a long-term field project are expected to learn the local language, both husband and wife. During pre-field training it became clear that my wife had major difficulty with the recognition and production of pitch distinctions. This factor played a role in the decision as to where in the wonderful world of SIL we would serve. In August 1990, when I came back from an orientation visit to the gorgeous Swat valley in northern Pakistan, deeply impressed by the beauty and magnificence of the mountain scenery surrounding the town of Kalam in the very north of the valley, I reassured my wife that we would not be going to such tone-ridden parts of the world as Central America, South-East Asia, or Africa, but that we would be working on an Indo-Aryan language, and that Indo-Aryan languages belong to the Indo-European language family, just like Dutch, English and German. Nothing to worry about! Fairly soon after starting fieldwork in Kalam in 1991, I noticed that something seemed to be happening with pitch in the language. However, as I had already promised my wife that Kalam Kohistani was not to be a tone language, I tried as hard as I could to analyze these pitch phenomena in terms of sentence intonation. A first draft that I produced of a report on Kalam Kohistani phonology discussed word-prosodic phenomena in such terms and ignored the possibility of lexical tone. Only some 18 months after starting fieldwork the penny dropped and I realized that we were dealing with lexical tone. One day I was sitting in a shop in the Kalam bazaar and saw a man swatting flies. I asked the shopkeeper what this man was doing. He answered: ~ m š ph t m r nt this man flies is.killing ‘This man is killing flies’ Whatever I tried, I could not make sense of the low pitch on ph t, especially as I had collected examples of very similar constructions in very similar contexts where the object had high pitch. So finally, to my embarrasment, I had to begin to entertain the possibility that Kalam Kohistani had lexical tone. Once I had made this mental switch, I sat down with my language consultant and began to systematically record and analyze the tones of Kalam Kohistani. Of course this whole episode could have been avoided if I had paid more attention to the literature on the northwestern IA languages before starting my fieldwork. For instance, the tones of Punjabi had already been accurately described early in the 20th century by the prolific Rev. T. Grahame Bailey (1872-1942) in a language learning course book for British administrators (Bailey 1914). Bailey also discovered the existence of tone in Shina, and reported on tonal phenomena in many other languages and dialects in the western Himalayan region as well. While the existence of a complex tonal system in Kalam Kohistani was a surprise for me, this was not the case at all for Dr. Georg Buddruss, one of the leading experts on the languages of this part of the world. In the summer of 1980, Buddruss made a short field trip to Kalam with a number of German colleagues. In the few days that he was there he learned enough about the local language to conjecture that it had five contrastive tones. Years later, upon receiving a copy of my published phonology of Kalam Kohistani (Baart, 1997), he was delighted to see that my work confirmed his conjecture. Again, I could have saved myself a lot of time and effort had I taken the trouble to contact Dr. Buddruss and pick his mind before going to Pakistan. Kalam Kohistani (also known as Kalami, Gawri, or Bashkarik) is spoken by approximately 100,000 people who live in the upper reaches of the Swat and Panjkora valleys in Pakistan’s Khyber Pakhtunkhwa Province (formerly known as the North-West Frontier Province). Many of these people practice seasonal migration: in the winter they move to the cities in the lower parts of the country to work and earn some cash. In the summer they return to their home areas in the mountains to attend to their fields. Table 1 below presents the five contrastive tonal melodies of Kalam Kohistani. The column headings are mostly self-explanatory, but it is probably useful to explain that there is a distinction between the “falling tone” and the “delayed falling tone” in that the delayed falling pattern typically falls, so the speak, from the last syllable of a word onto the first syllable of the next word, while the regular falling tone is fully executed within one and the same word. high level H falling HL delayed falling H(L) low level L b r ‘lion’ b r ‘lions’ b r ‘deaf’ b r ‘Pathan’ ɡ r ‘partridge’ ɡ r ‘partridges’ b n ‘joint’ b n ‘excuse’ rising LH ɡ r ‘horse’ b n ‘bowl’ ch r ‘loss’ b r ‘many’ ch r ‘milk jet’ b r ‘Open!’ Table 1: Minimal sets of words illustrating contrastive tone in Kalam Kohistani Within an autosegmental-metrical framework the high-level and low-level tones are represented by the symbols H and L, respectively. The regular rising and falling tones are written as LH and HL, whereas the delayed falling tone is written as H(L). The five tones of Kalam Kohistani are “word melodies”: the pitch patterns belong to whole words rather than to single syllables, as in b b y ‘apple’, where the rise spreads over two syllables. Tones associate from right-to-left in Kalam Kohistani: the last tone of a melody associates with the last vowel of the morpheme, the other tone associates with the second to last vowel and with any previous vowels. If a word is monosyllabic, then both tones associate with the vowel of that syllable. In the case of H(L), the parentheses indicate that the L does not take part in the initial association process (it is “inert”). The H in H(L) is initially associated with the last vowel and any previous vowels in a morpheme, while the L remains unassociated or “floating”. When words are put together in a sentence, the floating L tone may associate with the first vowel of the following word. If there is no following word, the floating L tone is realized as a glottal stop or as creaky voice in the final syllable. For an extensive description of Kalam Kohistani tone, the reader is referred to Baart (1999 and 2004). Stress in Kalam Kohistani is somewhat of a puzzling phenomenon. In my phonetic transcriptions of Kalami data, I marked word stress as I perceived it, using as my criterion that a syllable is stressed if it sounds prominent relative to the other syllables. Of the polysyllabic words in my data that were pronounced in isolation, I marked over 80 percent as being stressed on the final syllable. The other 20 percent, that is all words with non-final stress, are words with an HL lexical melody. For words spoken in sentence contexts it was found that the place of perceived prominence may sometimes shift. An example is the word išpo ‘sister’ that has a H tone that spreads over both syllables. In utterance-medial position I often hear this word as stressed on the first syllable. However, when spoken in isolation or in utterance-final position, the final syllable is spoken with falling pitch (due to a final L added by sentence intonation) and in that case stress is perceived to be on that final syllable. In view of these observations, a reasonable conclusion is that stress does not function independently in Kalam Kohistani; rather the perception of a particular syllable as prominent is predictable from the tone patterns. However, before we leave the matter of Kalami word stress, we need to look at a class of words where the alignment of the HL melody deviates from the default pattern. Two examples are given in (13). (13) 'č č n 'š ṭ ṭ r ‘common Hawthorn’ ‘bat’ These words have high pitch on the first syllable, and low pitch on the other two syllables. The lexical melody is HL, and the regular way of associating tones and syllables in this language is from right-to-left, linking the L tone to the final syllable and the H tone to the preceding two syllables. So what we would expect for the words in (13) is a surface melody HHL, but what we see in actual fact is a surface melody HLL. An example of a three-syllable word with a lexical HL melody and regular alignment is presented in (14). Here, indeed, the final syllable is low, and the preceding two syllables are high. (14) s r ‘finger’ What we see, then, is that the alignment of the HL melody shows variation and this variation does not seem to be predictable. It follows that in a lexical representation of words such as those in (13), a diacritic mark is needed to indicate the non-standard alignment of the HL melody. A further complication concerns the claim cited above that Kalam Kohistani lexical melodies stretch out over the entire word; this is only true for H, L, and LH. In the case of HL and H(L), syllables preceding the H, if any, receive a neutral, mid pitch instead of being linked to the H. The impression we are left with, then, is one of Kalam Kohistani as a mixed system, with one set of words bearing a falling pitch accent, the placement of which is lexically determined, while other sets of words bear lexical tonal melodies, the alignment of which does not refer to stress or accent. To what extent this is true is a subject for further research. 6. Conclusion As I hope to have shown in this brief survey, the word prosodic systems found in the northwestern Indo-Aryan languages constitute a rich subject for linguistic study. For many of these languages adequate descriptions have not yet been produced at all. Of the few people who work on these languages, fewer still analyze prosody in any depth. My hope is that more interested linguists will join the few of us that are active in this study. References Baart, Joan L.G. (1997). The Sounds and Tones of Kalam Kohistani. Islamabad: National Institute of Pakistan Studies and Summer Institute of Linguistics. Baart, Joan L.G. (1999). Tone rules in Kalam Kohistani (Gawri, Bashkarik). Bulletin of the School of Oriental and African Studies 62/1 (February 1999), p. 88-104. Baart, Joan L.G. (2004). Contrastive tone in Kalam Kohistani. Linguistic Discovery 2/2, pp. 1-20. Bailey, Thomas Grahame. (1914). A Panjabi Phonetic Reader. London: University of London. Beckman, Mary E. (1986). Stress and Non-Stress Accent. Dordrecht: Foris Publications. Berger, Hermann (1960). Bericht über sprachliche und volkskundliche Forschungen im Hunzatal. Anthropos 55, pp. 657-664. Buddruss, Georɡ (1993). Muhammad Amin Zia: An meine Lebensgefährtin. In: Günther D. Sontheimer (ed.), Südasien Anthologie: 44 Übersetzungen aus Südasiatischen Literaturen (pp. 1.-4), Stuttgart: Franz Steiner Verlag. Buddruss, Georg (1996). Shina-Rätsel. In: Dieter B. Kapp (ed.), n vi haikat : Festschrift für Hermann Berger (pp. 29-54), Wiesbaden: Harrassowitz. Collinge, N.E. (1985). The Laws of Indo-European. Amsterdam: John Benjamins. Dyrud, L. (2001). Hindi-Urdu: Stress Accent or Non-Stress Accent? Grand Forks: University of North Dakota. MA Thesis. [Online: http://arts-sciences.und.edu/summer-institute-oflinguistics/theses/2001-dyrud-lars.cfm] Harnsberger, James (1994). Towards an Intonational Phonology of Hindi. Unpublished manuscript. [Online: www-personal.umich.edu/~jharns/hindi.html] Hussain, Sarmad (1997). Phonetic Correlates of Lexical Stress in Urdu. Northwestern University. Ph.D. Thesis. Hyman, Larry M. (2001). Tone systems. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology and Language Universals: an International Handbook, vol. 2, 1367-1380. Berlin & New York: Walter de Gruyter. Ladd, D. Robert (2008). Intonational Phonology: Second Edition. Cambridge: Cambridge University Press. Liljegren, Henrik (2008). Towards a Grammatical Description of Palula: an Indo-Aryan Language of the Hindu Kush. Stockholm: Stockholm University. Losey, W. (2002). Writing Gojri: Linguistic and Sociolinguistic Constraints on a Standardized Orthography for the Gujars of South Asia. Grand Forks: University of North Dakota. MA Thesis. [Online: http://arts-sciences.und.edu/summer-institute-oflinguistics/theses/2002-losey-wayne-e.cfm] Radloff, Carla F. (1999). Aspects of the Sound System of Gilgiti Shina. Islamabad: National Institute of Pakistan Studies and Summer Institute of Linguistics. Schmidt, Ruth L. and Razwal Kohistani (2008). A Grammar of the Shina Language of Indus Kohistan. Wiesbaden: Harrassowitz. Szemerénye, Oswald J.L. (1999). Introduction to Indo-European Linguistics. Oxford: Oxford University Press. Zoller, Claus Peter (2005). A Grammar and Dictionary of Indus Kohistani, Volume 1: Dictionary. Berlin: Mouton de Gruyter.