The Luwian Language Ilya Yakubovich (Moscow/Marburg) The Luwian language belongs to the Luwic subgroup of the Indo-European Anatolian languages and is a close relative of Hittite. It is recorded in two scripts: an adaptation of Mesopotamian cuneiform and Anatolian hieroglyphs. The goal of this paper is to provide a concise description of the Luwian language. It contains both information on its structure, with an emphasis on phonology and morphology, and sociolinguistic data. The grammatical description is predominantly synchronic, but historical and comparative information is occasionally introduced if it has a potential to clarify the synchronic state of affairs. Keywords: Luwian, Hittite, Indo-European, Luwic, cuneiform, Anatolian hieroglyphs. 1. Name of the language The Luwian language is found in texts from central and southern Anatolia and northwestern Syria from approximately 1500 to 700 BC. Many American linguists refer to the same language as Luvian. The English name of the language goes back to the Hittite adverb luwili ‘in Luwian’, which introduces some of the Luwian passages embedded in Hittite texts. This adverb is, in turn, derived from the toponym Luwiya, which is mentioned in the Hittite Laws as a part of the Kingdom of Hattusa. The designation of the Luwian language by its native speakers is unknown. Since Luwian became the main written language of most Neo-Hittite States in the 1st millennium BC, it is possible that Luwian speakers referred to themselves as “Hittites” during this period (this is, at least, what they were called by their neighbors, the Assyrians and the Urartians). 2. Phylogenetic, areal, and sociolinguistic information The Luwian language belongs to the Anatolian group of Indo-European languages. The majority of Indo-Europeanists assume that the Proto-Anatolian language was the first one to separate itself from the Proto-Indo-European continuum, and some of them use the term Indo- Hittite or Indo-Anatolian for the common ancestor of the Anatolian (including Luwian) and “core” Indo-European languages. The closest relatives of Luwian within the Anatolian group are small corpus languages attested in alphabetic transmission in the 1st millennium BC, such as Lycian, Milyan (Lycian B), and Carian. The term “Luwic” is increasingly used for Luwian and its closest relatives. There is no up-to-date comparative grammar of the Anatolian languages, but, for phonological comparison between Luwian and its closest relatives, Melchert (1994) remains available for consultation. The Luwian language displays a number of non-trivial structural similarities with its neighbors which cannot be accounted for in phylogenetic terms, and therefore need probably be explained through language contact. The concept of an Ancient Anatolian linguistic area was discussed from a theoretical perspective in Watkins (2001). Among the likely contact-induced features of Luwian one may single out the neutralization of the inherited opposition between the historical voiceless and voiced stops in word-initial position (in intervocalic position it was probably reinterpreted in fortis/lenis terms). This feature, which is generally typical of the Indo- European Anatolian languages, finds a precise counterpart in the geographically adjacent Hurrian language, but not in its close relative Urartian, which was presumably spoken further to the northeast. The areal isogloss under discussion is quite unusual from the typological perspective, since the neutralization of consonant laryngeal features is generally more typical of syllabic codas than of onsets. Another likely contact-induced feature is the constraint on initial r-, which is overall typical of the languages of Ancient Anatolia and reconstructed for early Armenian and Greek, but not for Proto-Indo-European. In 1st millennium Luwian, however, initial r- becomes possible again due to the simplification of consonant clusters. On the morphological level, one may surmise that some sort of common influence is responsible for 2 case attraction in Hittite noun phrases, the proliferation of Luwian possessive adjectives at the expense of genitive case nouns, and the double case construction in Hurrian and Urartian. The common denominator here is the double marking of external case in possessive constructions (for details, see Yakubovich, 2008a). Unfortunately, the starting point of all these innovations, if they were indeed contact-induced, remains unclear; they may have been driven by a shared Anatolian substrate of unknown genetic affiliation. The local homeland of the Luwians can be reconstructed in the central part of Asia Minor (present-day Turkey), including the Konya Plain and Sakarya River valley. Already for the first part of the 2nd millennium BC one can postulate widespread Hittite-Luwian bilingualism, which speaks for the contiguity of the areas where these two languages were spoken (Yakubovich, 2010: 161-205). There are grounds to believe that Luwian functioned as an acrolect in one or several Anatolian principalities during this period. Hence, for example, there are Luwian loanwords in Old Hittite which belong to the political and administrative sphere, for example Luw. ubadid- borrowed as OHitt. ubadi- ‘demesne’, or Luw. tummanti- ‘to hear’ representing the source of OHitt. tummantiya- ‘obedience’. The Kingdom of Hattusa, which was formed in central Anatolia in the 17th century BC, was traditionally called Kingdom of the Hittites in secondary literature, because Hittite was the main language used there for purposes of writing cuneiform. This kingdom, however, was multiethnic from the very beginning. Luwian, alongside Palaic, enjoyed there the status of a regional language, witness the Luwian and Palaic formulae embedded in the Hittite religious texts pertaining to the official state cult. Beginning in the 14th century BC one can trace the presence of large groups of Luwian speakers in Hattusa, although the Luwian language initially functioned there as a basilect, while Hittite was an acrolect. In the 13th century BC Luwian became the main vernacular of Hattusa, although Hittite retained its role in the official sphere. Direct evidence for Hittite and Luwian bilingualism in Hattusa during this period stems from growing lexical interference between Hittite and Luwian (Melchert, 2005), from a large number of foreign words preserving Luwian inflectional endings in Hittite official texts (van den Hout, 2006), as well as from the partial restructuring of New Hittite grammar under Luwian impact (Rieken, 2005[2006]). A Luwian language community was definitely present in the kingdom of Kizzuwatna (Classical Cilicia) in the 15-14th centuries BC. The Luwian dialect of Kitzzuwatna displays structural peculiarities that speak in favor of its genesis in the Luwian and Hurrian bilingual communities (cf. section 9). The expansion of the kingdom of Hattusa in 14th-13th centuries BC led to the further extension of the Luwian-speaking area into northwestern Syria. After the collapse of the kingdom of Hattusa and the end of the written transmission of the Hittite language in the early 12th century BC, Luwian takes over its administrative functions in the territory of the so-called Neo-Hittite states. The expansion of the Neo-Assyrian Kingdom led to the progressive shrinking of the area where the Luwian language was used in writing during the 1st millennium BC. While the latest Luwian inscriptions can be dated to about 700 BC, there are reasons to believe that Luwian vernacular dialects lingered on in southeastern Anatolia at least until the beginning of the Hellenistic period (late 4th century BC), and possibly even later. There is no way to define or even estimate the number of Luwian speakers at any given point in time. One should, however, assume that the Luwians were more numerous than the Hittites in the second half of the 2nd millennium BC, since otherwise it is hardly possible to account for the progressive language shift from Hittite to Luwian in the Kingdom of Hattusa in spite of the cultural dominance of Hittite in this polity. The first attempt at the comprehensive treatment of Luwians as an ethnic group is Melchert (2003). It was followed by Yakubovich (2010), where different perspectives on the Luwian homeland and the spread of the Luwian language are offered. The discussion of various approaches to the ethnic history of the Luwians became a topic of a recent conference, whose proceedings were published as Mouton, Rutherford, and Yakubovich (2013). 3 i 3. Philological information 3.1. Written texts The Luwian language was written in two syllabic scripts, namely, an adaptation of Mesopotamian cuneiform and Anatolian hieroglyphs. All the Luwian cuneiform texts are recorded on clay tablets, and the majority of them belong to the sphere of practical magic. The earliest Luwian cuneiform fragments were probably written down in the 15th century BC. These short texts do not feature a stable orthography, but rather represent transliterations of Luwian passages embedded within Hittite rituals. One can hypothesize that the reason for using Luwian magic incantations in religious texts from Hattusa, rather than translating them into Hittite, was a desire to preserve their efficacy. The frequent marking of Luwian foreign words with special signs (“gloss wedges”) in Hittite texts shows that the Hittite scribes were keenly aware of a difference between the two languages and perceived the marked words as deviating from the formal style. The cuneiform script stopped being used for any purposes in central Anatolia after the collapse of the Kingdom of Hattusa in the early 12th century BC, and so the tradition of rendering Luwian passages in cuneiform was also discontinued. The Luwian language gained official recognition in Hattusa chanceries in the 13th century BC, after it became associated with the Anatolian hieroglyphic script. In that period, however, the Luwian orthographic norm was at its embryonic stage, as testified by frequent rebus writings and the minimal number of phonetic spellings (as in the SÜDBURG inscription, Hawkins, 1995). The normalization of Luwian hieroglyphic orthography becomes a reality only in the 1st millennium BC. The texts of this period contain a large number of historical and hypercorrect spellings, which bear witness to a developed orthographic tradition transmitted by generations of professional scribes. The most frequent variety of Luwian hieroglyphic texts is represented by monumental inscriptions belonging to the last kings of Hattusa and rulers of the Neo-Hittite states. A number of administrative documents written on lead strips have also been preserved. The Hittite sources inform us about a widespread use of wooden writing boards covered by wax. There is a strong suspicion that this medium was conducive to the proliferation of Luwian hieroglyphic texts in the Empire period (Waal, 2011), but an archaeological confirmation of this hypothesis is so far lacking. 3.2. Decipherment history and reference sources The decipherment of the Luwian language went through several stages and was a joint effort by many scholars. Already in 1880 the English scholar the Reverend Archibald Sayce succeeded in establishing the values of Anatolian hieroglyphic signs REX ‘king’ and REGIO ‘country, kingdom’. The subsequent process of decipherment was, however, slowed down by the unknown genetic connections of the language of Anatolian hieroglyphs. The discovery of the archives of Boğazköy and decipherment of Hittite in the early 20th century were quickly followed by the realization that the Luwian language of the Boğazköy cuneiform tablets represents a close relative of Hittite. The first steps toward the systematic study of Luwian hieroglyphic texts were made based on the assumption that they are written “Hieroglyphic Hittite”. The leading role at that stage in the decipherment belonged to Ignace Gelb (USA), Piero Meriggi (Italy), Emil Forrer (Switzerland /Germany), and Bedřich Hrozný (Czechoslovakia). The verification of their hypotheses became possible due to the discovery of a long Phoenician and Luwian bilingual of KARATEPE in 1947. In the 1950s and 1960s the French scholar Emmanuel Laroche demonstrated in a number of papers that “Cuneiform Luwian”, “Hieroglyphic Hittite” and Lycian constitute a subgroup within the Anatolian group of languages, which many scholars now call Luwic. But the true realization of the degree of proximity between the Luwian dialects of cuneiform and hieroglyphic texts came only after a team formed by David Hawkins and Anna Morpurgo-Davies in England and the German scholar Günter Neumann proposed the New Readings of several 4 high frequency Anatolian hieroglyphs (Hawkins, Morpurgo-Davies, and Neumann, 1974). Afterwards, the term “Hieroglyphic Hittite” came to be gradually replaced with “Hieroglyphic Luwian”. I have, however, tried to demonstrate in Yakubovich (2010) that the inner-Luwian dialectal isoglosses do not precisely align with the boundaries separating the cuneiform and hieroglyphic corpora. Therefore it is better from the methodological viewpoint to operate with the concept of one Luwian language having several dialects and written in two scripts. The decipherment of the Luwian language cannot be regarded as fully accomplished yet. Nevertheless, scholars have achieved a reliable understanding of the majority of hieroglyphic texts and a large number of cuneiform texts. Hawkins (2000) represents the main reference source for the hieroglyphic texts composed after the fall of the Kingdom of Hattusa. This monumental work contains their autographs, transliterations, and translations, accompanied by a detailed philological commentary. The longer hieroglyphic inscriptions of the Empire period (13th century BC) are published in the same format in Hawkins (1995). The very great majority of Luwian cuneiform texts are edited in transliteration in Starke (1985), but without translation or philological commentary. All the Luwian passages glossed in this paper were also treated in either Starke (1985) or Hawkins (2000), although the translation offered here reflects my own analysis. The most detailed grammatical treatments of the Luwian language that retain their value up to now are Plöchl (2003), which covers only for hieroglyphic texts, and Melchert (2003: 170- 210). In comparison with these works, the present one is shorter and therefore omits many details. On the other hand, it has the virtue of being able to incorporate the numerous discoveries of the last decade that contributed to the better understanding of the Luwian language. Melchert (1993) represents a glossary of Luwian lexical material in cuneiform transmission, which covers the entire corpus collected in Starke (1985) with an addition of isolated foreign words occurring in Hittite texts (the latter part now is in serious need of updating). Unfortunately, a comprehensive up-to-date lexicon of Luwian texts in hieroglyphic transmission still remains a desideratum. The most recent manual of Luwian in hieroglyphic transmission is Payne (2010), which also contains a glossary to the texts treated. 4. Writing systems and their interpretation 4.1. Cuneiform and hieroglyphic scripts A number of transcription and transliteration conventions will be used throughout the rest of this paper. The angular brackets < > mark narrow (sign-by-sign) transliteration. Square brackets are reserved for tentative phonetic transcription, and curly brackets are used for the morphophonemic representation of complex forms. The italic script without brackets is used for interpretative transliteration, which will be discussed in 4.2. The abbreviations (C) and (H) refer respectively to the cuneiform and hieroglyphic attestations of individual forms. The basic structural principles that characterize the Hittite adaptation of Mesopotamian cuneiform (Hoffner and Melchert, 2008: 9-24) are also generally applicable to the cuneiform transliteration of Luwian texts. For reasons of space, and because of the availability of many parallel treatments, my description of cuneiform orthography is necessarily cursory. The three main sign categories are phonetic signs, logograms, and determinatives/classifiers. The basic shapes of phonetic signs are <CV>, <V>, and <VC>, which are combined matching vowel quality within a syllable (e.g. <ta-a-ti-ya-an> = [ta:dijan]). Redundant signs of the shape <CVC> also occur rather frequently. The intervocalic fortis and lenis stops are respectively spelled <VC- CV> and <V-CV> in the cuneiform script (e.g. <a-pa> = [aba] vs. <ap-pa> = [appa]). A peculiarity of rendering the Luwian words in the cuneiform is the frequent use of extra <a> in word-initial position, i.e. <a-an-ta> ‘in’, <a-an-ni-i-> ‘to do, cause’, <a-ad-du-wa-a-li-> ‘evil’. Such spellings probably do not reflect either stress or vowel length, but possibly represent an attempt to render in writing the initial glottal stop (e.g. [Ɂanda] ‘in’), as was also the case in Assyrian cuneiform. 5 Logograms are signs for whole lexemes (usually nouns), which, in the instance of cuneiform orthography, are normally taken from the Sumerian language and transliterated into Sumerian in capital letters. The logographic (Sumerographic) spellings frequently feature phonetic complements, e.g. <ALAM-ša> = [tarussa] ‘statue’, where ALAM is the Sumerian word for statue. Determinatives are the same logograms, which are, however, not meant to be pronounced, being used instead as purely graphic classifiers. They are transliterated into Sumerian in capital superscript letters, e.g. <UZUŠÀ> ‘heart’, where <ŠÀ> is the Sumerian word for ‘heart’, while <UZU> is a classifier with the basic meaning ‘flesh’. The Luwian transliterated passages feature fewer Sumerograms in comparison with the general style of the Hittite rituals where they occur. Akkadograms, or words spelled in Akkadian but meant to be read in another language, are very rare in Luwian cuneiform texts. The Anatolian hieroglyphs represent an indigenous writing system that developed inside Asia Minor in the Hittite and Luwian bilingual environment (Yakubovich, 2008b). It was possible to reach such a conclusion through analysing the phonetic values of Anatolian signs, which may evoke the initial syllables of both Hittite and Luwian lexemes that correspond to their logographic values. Thus the logogram <PES>, used for the motion verbs, has the syllabic value <ti>, presumably reached through the mediation of Hitt. tiya- ‘to step’, while the <SIGILLUM> ‘seal’ has the syllabic value <sa >, extracted from Luw. sasanza ‘seal’. The earliest datable 5 hieroglyphic inscriptions containing phonetic signs are found on the seals of Hittite kings and high officials (14th c. BC) and represent short legends containing names and titles, which can in principle be read in both Hittite and Luwian. For a representative and up-to-date edition of (mostly) 13th century seals see Herbordt (2005). The choice of the new writing system for personal seals, alongside or instead of the traditional Mesopotamian cuneiform, was presumably prompted by nationalistic concerns. The complete inventory of Anatolian hieroglyphs includes more than 500 signs, although about one half of them still resist interpretation, largely due to their rare occurrence. Their most recent comprehensive discussion can be found in Marazzi (1990), but for practical purposes one can use the more up-to-date sign list in Payne (2010: 161-195). The most frequent template of phonetic (syllabic) signs is <(C)V>. The <CVCV> signs are, overall, less frequent in texts and fewer in number. Most signs correspond to syllables ending in one of the 3 basic vowels (e.g. <pa>, <ti>, <ru>), but syncretic vocalic values are also possible (e.g. <wa/i>, <lu/a/i>). Other categories of signs include logograms, and classifiers/determinatives (cf. above). The Luwian texts of the 1st millennium BC feature special signs marking word boundaries in the continuous text (transliterated as <|>) and markers of logograms and determinatives (transliterated as <“ ”>), although the use of neither sign is consistent. A peculiarity of the hieroglyphic script that sets it apart from the cuneiform is the absence of phonetic signs with the <VC> template, which makes impossible the unambiguous rendering of consonant clusters. In the majority of cases clusters C C are rendered as <С a-С V>, but /n/ is not reflected in writing before another consonant. 1 2 1 2 Another peculiar feature of the Anatolian hieroglyphs is the frequent use of determinatives with verbal forms, while their use in the cuneiform is generally limited to classifying nouns. The third peculiarity is the use of phonetic indicators, i.e. special syllabograms that are used for specifying the reading of logograms without rendering the precise phonetic shape or even position of the respective syllable (cf. the next paragraph). Finally, an unusual feature of the Anatolian hieroglyphic script is the a-vowel spelled at the end but pronounced at the beginning (the so- called “initial-a-final”), e.g. graphic <mi-sa-a> corresponding to the phonological /amis/ ‘my’ (nom.sg.m). The modern transliteration of Anatolian hieroglyphs has a number of special conventions. Separate signs belonging to one word-form are connected by hyphens, graphemes forming a ligature (continuous/overlapping spelling) are linked by “plus” signs, while the ambiguity of the vowel in a syllabogram is marked by a slash, e.g. <i+ra/i-há-> = irha- ‘boundary’). The homographic syllabic signs are marked by subscript numbers, but the numeral 2 is replaced by an acute accent (<há> = <ha >), while the numeral 3 is replaced by a grave accent (<sà> = 2 6 <sa >), as in cuneiform. The better understood logograms and determinatives are transliterated 3 in capital letters, while their meanings or shapes are commonly rendered in Latin, e.g. <REX> ‘king’, <CRUX>: cruciform sign. Determinatives, unlike logograms, are not hyphenated to the lexemes they determine, but are written in parentheses instead. Thus tarrawannis ‘ruler’ can be written as a logogram with phonetic complements <IUDEX-ni-sa> or fully spelled out with the determinative <(IUDEX)tara/i-wa/i-ni-sa>. The less understood and usually less frequent logograms and determinatives are commonly transliterated with numbers under which they appear in the sign list of Anatolian hieroglyphs (Laroche, 1960). In this case they are accompanied by asterisks, e.g. <*501>. Capital italic letters are used for marking phonetic indicators, e.g. <PRAE-ia AUDIRE+MI-ma-ti-mi-i-sa> = parriya tummantimmis ‘glorified’, <INFANS.NI-sa> = nimuwizzas ‘son’. The same capital italic letters mark the few logograms whose meaning is rendered “in Anatolian” and not in Latin, e.g. <HALPA> ‘Aleppo’, <SARMA> ‘the god Sarruma’. The large hieroglyphic inscriptions are commonly subdivided into horizontal registers, which are filled in boustrophedically (first register right to left, second left to right, third right to left etc.). Each register tends to be two or three symbols “thick”. The columns are consistently filled in the downward direction, and attempts are frequently made to align word boundaries with column ends. Sometimes the sign shapes create lacunae or bulges that affect the order of subsequent signs. Most Anatolian hieroglyphs have monumental and cursive variants, the first being typical of bas-reliefs, the second scratched on stone or lead strips. There are, however, inscriptions where monumental and cursive signs occur side by side. From the paleographic viewpoint, one also distinguishes between the sign-shapes typical of the Empire of Hattusa and those of the 1st millennium BC, while the intervening 12-11th centuries BC left fewer written texts and can be characterized as “dark centuries”. In addition to paleographic differences, the signs of the two periods sometimes display functional distinctions. Thus the <ali> sign of the Empire period (*416) developed into the <la/i> sign (*319) in the 1st millennium, while the sign *445, which originally had the specific value <lu> developed the more inclusive reading <lu/a/i> in the Neo-Hittite period. The narrow transliteration of the hieroglyphic script described in this section reflects the system adopted in Hawkins (2000) with some minor changes proposed in recent scholarship (e.g. Rieken and Yakubovich, 2010). It is fair to say that this system reflects today’s academic mainstream. Note, however, that the principal editions of some hieroglyphic texts (e.g. Poetto, 1993) use different systems. 4.2. Interpretative transliteration IPA-based transcription is rarely used in Luwian Studies. Most cuneiform scholars use either narrow, sign-by sign transliteration, or the result of its compression known as broad transliteration. The latter uses a number of conventions that are rooted in the tradition of cuneiform studies, e.g. z ~ IPA /ts/, y ~ IPA /j/, ā ~ IPA /a:/ etc. In the instance of the hieroglyphic texts the broad transliteration is rarely used, but scholars try instead to transcribe the forms into something resembling the broad transliteration of the cuneiform texts. For example <zi-in-zi> (C) ‘these (nom.sg)’ is compressed into zinzi, while <za-zi> (H) ‘id.’ is not compressed into zazi but transliterated as zanzi on the fair assumption that the nasal is phonologically present in this form but not reflected in hieroglyphic orthography (cf. 4.1). At the same time, the precise rules of transcribing the hieroglyphic texts are rarely, if ever, spelled out. In this section I shall endeavor to provide principles of a unified interpretative transliteration of Luwian cuneiform and hieroglyphic texts, which will be used in the rest of this paper. The proposed unified interpretative transliteration includes only the three basic vowels a, i, and u, even though this may result in a loss of information. In contrast, its consonantal inventory faithfully reflects the phonological oppositions between Luwian fortis and lenis stops, since it is reasonably consistently rendered in cuneiform transmission (cf. 4.1). It is thus different 7 from the conventions of bound transliteration deployed, for example, in Payne 2010, which are adopted for hieroglyphic texts alone, and where the fortis and lenis pairs are not contrasted. Some comments on the vocalic inventory are in order. Based on the analogy with Hittite, it is usually assumed that the so-called plene spellings in cuneiform texts, which consist in adding an extra vocalic sign matching the neighboring <VC> and/or <CV> signs, are deployed for the iconic rendering of vowel length in Luwian. There are indeed quasi-minimal pairs, such as <a-ad-du-wa-al-za> = [Ɂat:uwal-tsa] ‘evil (nom./acc.sg.n)’ vs. <a-ad-du-wa-a-al> = [Ɂat:uwa:l] ‘evil (nom./acc.pl.n.)’, or <wa-a-šu-un> = [wa:sun] ‘good (acc.sg.c)’ vs. <ḫi-i-ru-ú-un> [xi:ru:n] ‘oath (acc.sg.c.)’, which speak in favor of contrastive vowel length in Luwian. On the other hand, there are cases of inconsistent plene spellings in the same form, e.g. <du-ú-pa-im-mi-in> vs. <du-pa-a-im-mi-in> ‘stricken (acc.sg.n.)’. The cuneiform plene spelling in word-initial position may have had an altogether different function (cf. next paragraph). As for the plene spellings in hieroglyphic texts, that have not been sufficiently studied, but at least in some cases they must have had an ornamental function, helping to align word-boundaries with ends of vertical columns. In view of these facts, it appears premature to include information about vowel length in the interpretative transliteration of the Luwian texts. The hieroglyphic texts, in turn, offer primary evidence for a hypothetical contrast between [a] and [ǝ], which would, however, be limited to word-initial position. Pursuing this hzpothesis one can surmise that the initial glyph <a> renders [ǝ], while the initial glyph <á> corresponds to [a(:)]. The unstable character of initial [ǝ] follows from the graphic phenomenon of “initial-a-final” (cf. section 4.1). One can hypothesize that the spellings such as <mi-sa-a> ‘my’, instead of the expected *<a-mi-sa> for phonetic [ǝmis], indicate the aphaeresis of /ǝ/ in allegro pronunciation. There is no doubt that [ǝ] historically goes back to *[а] at least in some cases, since one finds alternation between [a] and [ǝ] in forms of the same lexeme: e.g. <*a-mi- sa> = [ǝmis] ‘my (nom.sg.c)’ vs. <á-ma-za> = [amantsa] ‘my (acc.sg.n)’. The precise conditions of the development *[а] > [ǝ] would, however, remain unclear, and the status of [ǝ] as a contrasting phoneme is not obvious either. Alternatively, Simon (2010) attempted to establish a correlation between initial <á> in hieroglyphic transmission and initial plene spellings in cuneiform rendering of the same lexemes, which was then interpreted as evidence for a prothetic or etymological glottal stop ([Ɂa-] or /Ɂa-/). For yet another phonetic account of <a> and <á> see Melchert (2010). In view of all the uncertainties and debates regarding the interpretation of the Analolian hieroglyphs <á> and <a> their difference cannot be reflected in the interpretative transliteration at the present stage in Luwian scholarship. The transition from the narrow cuneiform transliteration to its broad (interpretative) counterpart remains automatic in most cases. It essentially involves removing hyphens, contracting homorganic vowels, “voicing” intervocalic stops, and “devoicing” them otherwise. For example, both <du-ú-pa-im-mi-in> and <du-pa-a-im-mi-in> must be recorded as tubaimmin. An additional convention requires removing all the diacritics, e.g. <wa-aš-ḫa-az-za-aš> = washazzas. In contrast, a degree of interpretation is required if one begins with the narrow hieroglyphic transliteration. Thus, the majority of hieroglyphic syllabograms do not distinguish between fortis and lenis consonants, and therefore the hieroglyphic spelling <a-pa> can correspond to appa as well as aba. A salient exception here is the contrast between the coronal stop marked by the signs <ta> and <tá>, to be always transliterated as -(t)t-, and its counterpart marked by <tà>, to be always transliterated as -d- (this contrast was established in Rieken, 2008). In ambiguous cases, the interpretation of consonant laryngeal features requires comparison with cuneiform texts. For example, forms of the distal pronoun will be always transliterated as aba-, because it is spelled <a-pa->, not <ap-pa>, in the cuneiform. If a particular lexeme is only attested in hieroglyphic transmission, the discrimination between fortis and lenis consonants requires etymological analysis, or is outright impossible. Such instances can be marked by capital “archiphonemes” in the interpretative transliteration, e.g. aTuTinzi. Another interpretative decision involves the reconstruction of consonant clusters. Thus the hieroglyphic stem <(MANUS)i-sà-tara/i-> ‘hand’ can be in principle interpreted as istr(i)-, 8 istar(i)-, isatr(i)-, or even isadar(i)-. But the comparison with <iš-(ša)-ra/i-> ‘hand’ (C) suggests that the coronal stop is anaptyctic in origin and therefore vindicates the transliteration istr(i)-. As mentioned above, preconsonantal nasals are omitted in the hieroglyphic script, but cuneiform parallels help in their reconstruction, especially in grammatical morphemes. The most obvious instance where the hieroglyphic data can be invoked for improving the transliteration of cuneiform spellings concerns the labiovelars. Thus, the verb ‘to run’ is attested as <ḫu-u-i-(ya-)> in cuneiform transmission, which could be compatible with the transliteration huya as well as hwiya-, but <hwa/i-ia-> ‘to run’ (H) pleads for the second option. 4.3. Illustration It is appropriate to illustrate the principles of reading and transliterating the Anatolian hieroglyphic texts with a specific example (inscription MARAŞ 1, §2). The Luwian fragment provided below (to be read right to left) is followed by the linearized representation of normalized signs, narrow transliteration, and interpretative transliteration. The lack of correspondence between the original and normalized shapes of the signs <mu> and <ta> reflects their cursive writing in the passage under discussion. The grammatical analysis of this passage is given in 7.1. Figure 1: Fragment of a Luwian hieroglyphic inscription ¨Wì ¨ajIA ¨´TI GfIA ¨òai< |wa/i-mu |á-mi-zi-i |tá-ti-zi DEUS-ni-zi-i |(LITUUS)á-za-ta wa=mu aminzi tadinzi massaninzi azzanta 9 5. Phonology and morphophonemics 5.1. Inventory and phonotactics The phonological inventory postulated here for the Luwian language contains three vowels a, i, and u, and the consonants, which are summarized below. Note that the conventions of Table 1 are those of the interpretative transliteration, not the phonetic transcription. This does not exclude the possible existence of additional phonemes (such as long vowels), which are not consistently reflected in any relevant script. Table 1: Luwian consonantal system mode of articulation place of articulation labial coronal palatal velar labiovelar stops fortis p-, -pp- t-, -tt- k-ː -kk- kw-, -kkw- lenis -b- -d- -g- -gw- fricatives fortis s-, -ss- h-, -hh- hw-, -hhw- lenis -s- -h- -hw- affricates fortis z-, -zz- lenis nasals fortis m-, -mm- n-, -nn- lenis -m- -n- laterals fortis l-, -ll- lenis -l- rhotic fortis -r-ː -rr- lenis -r- glides w y A salient peculiarity of the Luwian consonantal system is a fortis vs. lenis phonological opposition, which is reconstructed for all the consonants except the glides (e.g. appa ‘back’ vs. aba- ‘that’, arra(ya)- ‘long’ vs. ar(a/i)- ‘time’, etc). Positions other than intervocalic, i.e. next to other consonants, or word-boundaries, do not display a contrast between the fortis and lenis features. The stops are neutralized in favor of their fortis counterparts (the sign <tà> is not used word-initially and in clusters in the hieroglyphic script). The consonants other than the stops and glides can also be assumed to undergo neutralization, but here its outcome is unclear. Neither the Anatolian adaptation of the cuneiform nor the hieroglyphic script has the capacity to mark the intensity of continuants and sonorants, except between vowels. Only indirect arguments can be invoked in some cases. Thus <iš-ra/i-> ‘hand’ (C) has the variant <iš-ša-ra/i->, which may be interpreted as an attempt to render the fortis character of preconsonantal s. But the interpretative transliteration *issr(i)- would be unusual and confusing, and so the continuants and sonorants are transliterated in positions of neutralization as if they were lenis. One should realize that this is no more than a convention. Laryngeal feature neutralization aside, the only phoneme that was proscribed word- initially is r (cf. section 2). This constraint apparently concerned only the inherited lexicon of the 2nd millennium BC, for the 1st millennium BC cf. e.g. Late Luwian ruwan ‘formerly’. The word- final phonemic inventory is considerably more restricted. The only consonants that occur before the word boundary are s, n (frequently) and z, l, r (more rarely). The orthography does not allow one to decide whether consonant clusters are allowed in word-initial position, but the only combination that is likely on etymological grounds is obstruent + sonorant (СR-). For example, the preverb <pa-ri-i (C)>, <pa+ra/i-i> (H) is likely to be cognate with Latin prae, but this does not resolve the question whether the Luwian form should be transliterated as pri or pari. It the form <pa-a-ri> (C), which occurs in an unclear context, represents a variant of the same preverb, this would tip up the scales in favor of epenthesis *pri > pari in Luwian. Luwian, unlike Hittite, drops the inherited word-initial *s- before stops, cf. e.g. Luw. tummanti- vs Hitt. (i)stamass- ‘to hear’. 10 The only reconstructable word-final cluster is -nz, e.g. in the Kizzuwatna Luwian accusative plural ending (cf. 6.2). The cuneiform spelling <lu-u-la-ḫi-in-za-as-ta> interpreted as lulahinz=tta pleads against the pronunciation **[lulaxintsa] of lulahinz ‘mountaneers (acc.pl)’ in Kizzuwatna Luwian (cf. Starke, 1990: 44). The maximal syllable in Luwian is thus possibly CCVCC, e.g. Late Luwian is-tranz ‘hands (dat.pl)’. It is not certain, however, whether syllable- initial clusters were stable or whether excrescence isranz > istranz was shortly followed by anaptyxis istranz > istaranz. It is likewise uncertain whether the final cluster -nz was preserved in the 1st millennium BC. The phonetic realization or phonological type of the Luwian accent is not clear. One can, however make a clear distinction between fully accented words, separated by spacing (C) or word boundaries (H), and enclitics, which are attached in writing to the preceding forms. Although the marking of word boundaries in hieroglyphic texts is haphazard, the high frequency of clitics facilitates the conclusion that the consistent absence of word-dividers within clitic complexes is not accidental. The clitic boundary is marked by = in broad transliteration. 5.2 Alternations 5.2.1. Free/dialectal variation The contraction *iya>i and *uwa > u is very common, cf. e.g. 3sg.prs <a-ri-it-ti> (C) vs. 3sg.imp <a-ri-ya-ad-du> (C) from ari(ya)- ‘to raise’ or 3sg.prt <tu-wa/i-ta> (H) vs. 3pl.imp <du-ú-un-du> (C) from tu(wa)- ‘to put’. The contracted variants become more frequent in the 1st millennium BC. There are instances of a synchronic alternation ai ~ a, e.g.. 3 pl. imp. <a-ru-na- in-du> and <a-a-ru-na-an-du> (C) from arunai- ‘?’, 3pl.imp <ú-i-da-a-in-du> vs. 3pl.prt <wi -ta- 5 an-da> from widai- (C) ‘to strike’. The synchronic simplification rule a+i → a is also convenient for the synthesis of nominal paradigms (cf. section 6.2). At the same time, there are many instances if synchronic -ai- in Luwian, which normally goes back to the contraction of historical *-ayV-. It is likely that the combination “vowel+n” was sporadically realized as a nasal vowel before afficates. One indication of this phenomenon is the frequent omission of n in such a position in cuneiform orthography, e.g. <ḫu-u-up-pa-ra-za> (C) < *hupparanza ‘belts (acc.pl)’, <ti-wa-an-na-al-li-zi> (C) < *tiwannallinzi ‘? (nom.pl)’. The optional nasal vowel formation in Luwian is all the more likely, since the nasal vowels are directly attested in the closely related Lycian language. In addition, the nasal element before a stop could develop a labial co- articulation, which is reflected through the vowel <u> in both cuneform and hieroglyphic orthographies, e.g. < na-ak-ku-uš-ša-a-u-un-ta> (C) < *nakkussanta ‘they offered a scapegoat’, <wa/i-la-u-ta> (H) < *walanta ‘they died’. The attestations of this optional development appear to be limited to the verbal endings of 3pl.pret. The cuneiform texts show examples of -h- dropping before w and u, e.g. siwal and sehuwal ‘stiletto (vel sim.)’, lahuni- and launai- ‘to wash’. On “rhotacism”/flapping in Late Luwian, see section 9, on the likely reduction of initial a- see section 4.2. 5.2.2. Phonetically conditioned alternations The combination of two coronal stops across a morphemic boundary develops epenthesis and is realized as [tst]: cf. 2pl.pres.med az-tuwari vs. 3pl.prs ad-antu from ad- ‘to eat’. The combination of n or l with the following s on a morphemic boundary develops t-epenthesis. This is manifested above all in the formation of nom.-acc.sg.n of nominal forms, which are endowed with the -sa extension: e.g. {tarud-sa} → tarusa ‘statue’, {udar-sa} → udarsa ‘word’, but {parnan-sa} → parnanza ‘house’, {attuwal-sa} → attuwalza ‘evil’. The coronal stops fall out in word-final position according to the rules of Luwian phonotactics (cf section 5.1), e.g. nom.- acc.sg annarumahi from annarumahid- ‘virility’. There are cases of assimilation -n > -m before the clitic =pa: {man=pa} → mam=pa ‘but if’, {nanun=pa} → nanum=pa ‘but now’.