Chapter Six The Position of Anatolian 6.1 History of the Question 6.1.1 Discovery and Recognition Hrozný (1917) showed that the chief administrative language of the Hittite Empire, attested in cuneiform documents from Hattusha in central Turkey dating from the 16th-13th centuries BCE, was Indo-European. Through the work of a number of scholars it was known by the 1930s that Hittite was not alone in ancient Asia Minor, but part of a new sub-branch of Indo-European now called Anatolian, along with Palaic, Cuneiform and Hieroglyphic Luvian, Lycian, and Lydian. Indo-Europeanists had to reckon with a large new set of data to be integrated into the reconstruction of Proto-Indo-European. Due to its far more plentiful evidence and hence better understanding, Hittite remained for many decades the chief basis for comparison with the rest of Indo-European. Hittite presented a special challenge, because despite its antiquity it conspicuously lacked some key features of “classical” PIE as reconstructed chiefly on the basis of Sanskrit and Ancient Greek, with support from Latin and Avestan. In the noun there was no feminine gender distinct from the masculine. The verb showed no obvious trace of the aspectual contrast between “present” and “aorist” or of the “perfect” category at all. The subjunctive and optative moods were also missing. Hittite was typologically a synthetic and inflecting language 2 like those named above with recognizable Indo-European morphology, but it appeared to reflect either a more primitive or a more advanced stage of evolution than the other oldest attested representatives of the family. And precisely the choice between those alternatives quickly became the focus of a debate that has continued to the present. 6.1.2 First Responses Strictly speaking, there were nearly as many responses to the “Hittite problem” as there were Indo-Europeanists, and any generalizations run the risk of oversimplification. Nevertheless, most reactions may be fairly characterized as adopting one of three fundamental approaches. The first was to treat Hittite (respectively Anatolian) as merely one more subgroup of the Indo-European family like any other and to derive its features from the PIE already reconstructed, with a bare minimum of revisions to that model—as represented by the Grundriß of Karl Brugmann. Two articulate and nuanced presentations of this viewpoint may be found in Pedersen (1938) and Eichner (1975), but there have been many others. Since this account must assume that the features of “classical” PIE missing in Hittite are due to their having been lost there, it is often (simplistically) labeled the “Schwund-Hypothese”. While there have been important exceptions (see below), it is fair to say that this approach was dominant among Indo-Europeanists in Europe until the 1990s. 3 Some, however, adopted essentially a diametrically opposed position: namely, that the major features cited above (and arguably others) represent massive common innovations of non-Anatolian Indo-European in which Hittite did not take part. In terms of the family-tree (Stammbaum) model, Hittite (Anatolian) is thus not a descendant of “classical” PIE, but a co-equal branch, both being derived from an earlier prehistoric protolanguage. The most famous proponent of this view was Edgar Sturtevant, who argued in a series of works that Hittite and PIE are descended from what he labeled “Indo-Hittite” (see Sturtevant 1933a: 30 with diagram). The conception of “Indo-Hittite” found little favor in Europe, but Sturtevant’s ideas had rather more influence in North America, even if the explicit label “Indo-Hittite” was not always used.1 A third response to the new evidence of Hittite was exemplified by the work of Jerzy Kuryłowicz, in both phonology and morphology (see respectively 1927 and 1964 as representative). Kuryłowicz and others rejected both the “Schwund- Hypothese” and “Indo-Hittite”, contending that proper integration of the Hittite 1 The labels one chooses for the more remote parent language and for its immediate non-Anatolian descendant are of no consequence. What is crucial is the claim of large-scale common innovations that set off the latter protolanguage from Anatolian. See Cowgill (1979: 27) and compare the remarks of Eichner (1975: 722). 4 evidence demanded a radical and far-reaching revision of reconstructed PIE— meaning PIE as the source of not only Hittite, but also the non-Anatolian languages including Sanskrit, Greek and the rest. Other representatives of this viewpoint include Watkins (1969), Meid (1963) and (1975), Neu (1976 and 1985), and Adrados (1963, 1982 and 2007). Unsurprisingly, the scholars just named disagreed, sometimes markedly, on just what radical revisions should be made. 6.1.3 Stalemate and Resolution It was not at all clear by the decade of the 1980s how any compromise could be reached between the opposing models of the “Schwund-Hypothese” and that of a radical revision of PIE, cast in terms of “Indo-Hittite” or not (see the strong statement of Eichner 1975: 72). Nevertheless, several factors have since significantly altered the terms of the debate. First, there has been a significant maturation of Anatolian philology. A crucial breakthrough in our ability to establish the relative chronology of Hittite texts and manuscripts has brought reevaluation of nearly every aspect of its synchronic and diachronic grammar and a much better grasp of the Hittite facts. At the same time evidence for and our understanding of the other “minor” Anatolian languages has also dramatically improved, to the extent that these languages now actively contribute to the debate over the position of Anatolian vis-à-vis the rest of Indo-European. 5 Second, there has been a serious retrenchment regarding some of the evidence cited from non-Anatolian languages in support of the radically revised model for PIE. For example, while debate still continues on the precise formal details, a consensus developed by the 1980s that the Insular Celtic contrast of absolute and conjunct verbal endings reflects in some fashion the “classical” PIE system of primary and secondary endings (following Cowgill 1975a) and does not justify the radically innovative accounts of Meid (1963) and Watkins (1969).2 Third, in response to proposals like those of Meid (1975) there has developed a widespread view that we need not view the problem as strictly a choice between Anatolian as another descendant of PIE like any other subgroup or Anatolian and PIE as representing branches of “Indo-Hittite” (see further 6.4 below). Some archaisms claimed for Hittite/Anatolian have withstood scrutiny, and any viable reconstruction of PIE must take these into account. For that reason there are now few defenders of the strict Schwund-Hypothese or of an unaltered “classical” PIE. However, continued analysis of Hittite and the other Anatolian languages brings them ever closer to the rest of Indo-European (see in detail e.g. Rieken 2009, who recognizes fewer “Indo-Hittite” features than the present essay). Furthermore, many of the remaining differences involve relatively minor 2 Most aspects of the very novel analyses of Neu and Adrados in the works cited in the preceding section have also failed to win broad acceptance. 6 reshaping of the formal expression of grammatical categories (or changes in productivity of particular formal patterns), not the creation of—or major functional changes in—the categories themselves. Seeing Anatolian as uniquely reflecting a very archaic “Indo-Hittite” or “Early Indo-European” is thus also now a minor viewpoint (respectively Lehrman 1998 and Adrados 2007). I do not mean to suggest that a full consensus has yet been achieved regarding just which features of Anatolian are archaisms and which are innovations. Nor are the differences in the conceptions of individual scholars by any means trivial. However, most Indo-Europeanists now agree on the basic terms of the debate: some revisions to “classical” PIE are required, in order to account for the genuine archaisms of Anatolian, whose prehistoric speakers “separated” (or became sufficiently geographically isolated) from the rest of the PIE speech community so as not to share in some common innovations. The goal of the debate is to identify which features constitute these common innovations and to determine just how radical the revisions to PIE need to be. On these points there is a broad spectrum of opinion. What follows in 6.3 reflects my own current best assessment, with due acknowledgement of divergent views. 6.2 Issues of Time and Space I follow here the long-standing majority view that the Indo-European languages of Anatolia are intrusive to Asia Minor, having moved there from some point farther north in Europe. I reject the claim of Gamkrelidze and Ivanov (1995: 757- 7 852, especially 791, and elsewhere) for a PIE “homeland” in Eastern Anatolia and the even more radical proposal of Renfrew (1987) of a central Anatolian location for the PIE speech community dating to 7000 BCE, associating the presumed movement of Indo-European languages into Europe with the spread of farming dated to that era. For further discussion of this point see Melchert 2011a. Contrary to earlier views, there has now developed a consensus among linguists that entry of Indo-European speakers into Asia Minor was much earlier than previously assumed. See Melchert (2003a: 23-6) with references to Carruba (1995), Oettinger (2002a) and others, and also Lehrman (2001: 116-7) and Yakubovich (2010: 6-7). The gist of the argument is that the attested degree of differentiation of the IE Anatolian languages such as Hittite and Luvian already by the beginning of the second millennium requires at a minimum that their divergence from Proto-Anatolian began by the middle of the third millennium. It may easily have begun as early as the end of the fourth. The further inference of “early entry” into Anatolian rests on the premise that dialectal differentiation typically results from relative geographic separation of subsets of speakers of an original (putatively) unitary language (contact effects from different sub- or adstrate languages may play an important role). The default assumption in the present case is that this separation corresponds to the scattering of Proto-Anatolian speakers across Anatolia, after they had entered as a single group. I must stress, however, that we know virtually nothing about how Indo- 8 European speakers entered Anatolia. We cannot exclude the possibility that the dialectal differentiation began before such entry, with the speakers of pre-Hittite, pre-Luvian, etc. moving into the area at different times. See again Melchert 2011a. As indicated earlier, for many years the relatively poor attestation of the IE Anatolian languages other than Hittite severely restricted the amount of useful information they could provide us for recovering the linguistic features of Proto- Anatolian. As a result, despite the recognition of an Anatolian subgroup, there was almost no serious reconstruction of Proto-Anatolian. This situation fostered an unfortunate tendency to effectively project the features of (Old) Hittite back to Proto-Anatolian and to reflexively view any divergences in the other languages as due to innovation on their part. Due to the superior quantity and quality of its evidence, Hittite will perforce continue for the foreseeable future to occupy a privileged position within the study of Anatolian. However, we are now in some instances able to place it in its proper place as merely one of the languages that contribute to our picture of Proto-Anatolian, confirming that like all other natural languages Hittite reflects a mixture of archaisms and innovations. To the extent allowed by current knowledge, the following discussion of the relative position of Anatolian within Indo-European will be based on what we can reconstruct for Proto-Anatolian, not merely on the facts of Hittite. 9 6.3 Diagnostic and Non-Diagnostic Features 6.3.1 Basic Premises I adhere in what follows to the widely accepted principle that the crucial factor in linguistic subgrouping consists of non-trivial common innovations. Only when a subgrouping has been established on the basis of shared innovations can common retentions, if numerous enough, perhaps be adduced as supporting evidence. One well-known limitation on this procedure is that not all linguistic changes are clearly unidirectional. That is, if one set of languages shows state A of a given feature, and another state B, we cannot necessarily determine which represents the innovation. Defining a “non-trivial” innovation is also not always straightforward. For these and other reasons not every observed difference between the Anatolian and non-Anatolian IE languages will be useful for our purposes. Furthermore, like every other subgroup of IE, Anatolian is itself defined by certain shared innovations. Before we turn to the question of possible non-trivial innovations shared by all other IE languages, it is therefore useful to summarize those innovations that characterize Proto-Anatolian—particularly since the list given in Melchert (1994a: 6-8) requires serious revision. One fact has not changed since 1994: it is the apparently divergent development of Lydian, combined with our limited understanding of the language, that results in so few assured common 10 Anatolian innovations. If we limited ourselves to the other better-attested languages, the list would be substantially longer. Phonology. (1) the chief defining phonological innovation of Anatolian is the “lenition” (voicing) of PIE voiceless stops and *h between unaccented morae 2 (for this formulation see Adiego 2001, unifying what were previously treated as two separate changes, as per Eichner 1973: 79-83 and 10086 and Morpurgo Davies 1982/83).3 (2) unaccented long vowels inherited from PIE are shortened (Eichner 1973: 79 and 8615). (3) the sequence *h w became a unitary labialized 2 voiceless fricative ([xw] or similar): see Kloekhorst 2006: 97-101 and already in nuce Jasanoff 2003: 142-3. Evidence for the change to a unitary labiovelar in Hittite comes from alternate spellings tarḫu-/taruḫ- ‘conquer’ /tarxw-/ parallel to eku-/euk- ‘drink’ /egw-/, but more importantly from the fact that the labial articulation of the new /xW/ is absorbed by following /w/, just as in the case of the labiovelar stops: Pres1Pl tarḫḫueni just like akueni. A sequence of /xw/ before /w/ would have led via *tarḫuweni to *tarḫumeni, by the Hittite rule of dissimilation of *w > m before or after /u/ (Melchert 1994a: 92 with refs.). Furthermore, in Lycian the unitary labiovelar */xW/ “hardened” into a voiceless labiovelar stop /q/. 3 Virtually all evidence for the alleged Proto-Anatolian “limited Čop’s Law” proposed in Melchert (1994c) has been refuted, so the very existence of such a change must be regarded as extremely dubious.