On the Concept of an Anatolian-Greek Language Area


This paper aims at presenting some thoughts on the hypothesis of an Anatolian-Greek language area in the second millennium bc comparing different approaches both in the theoretical frames and in the analysis of the linguistic facts. For this purpose, it is necessary to introduce some terminological premises, followed by a selection of methodological issues, which will help explore the putative features that characterize the Anatolian-Greek area (morphological traits such as actionality markers, particles, verbal prefixes as well as special morphological forms; morphosyntactic traits, such as modal particles, sentence particles, absolute participial constructions; lexical units and phonetic features).


Some Terminological and Conceptual Preliminaries
The birth of the debate concerning linguistic areas goes back to the nineteenth century, with specific reference to the work by Miklosich (1861) for the Balkan languages, and a few decades later to Baudouin de Courtenay (1904). In particular, the Neogrammarians -beside the development of the genealogical model based on shared hereditary traits -identified the problem that some languages shared traits that did not derive from a common matrix but rather from language contact, even if the study of these issues developed only in the Old World: Journal of Ancient Africa and Eurasia (2021) 1-19 twentieth century. Trubetzkoy (1928), following the Neogrammarians, introduced the classification of language groupings (Sprachgruppen) into two separate categories (language families or Sprachfamilien and language leagues or Sprachbünde; cf. Urban 2007: 140). He was followed by Jakobson (1931Jakobson ( , 1938 (for phonological leagues), and then, with Weinreich (1958), interest in areal phenomena has intensively arisen until the last 40 years, with new insight into the classification of linguistic similarities and interchanges in certain cultural patterns.
In fact, since the first investigations in the 19th century have been carried out within an anthropological cultural agenda, already at the beginning of the 1990s linguist became more and more aware that many linguistic patterns and typological and historical features required areal explanations. This new born approach led to a better clarification of many of the issues involved and to the identification of several linguistic areas.1 Consequently, the dedicated investigations produced a very large literature. As regards the ancient times and regions, and specifically the Ancient Near-East, cultural and linguistic contacts became obvious in those situations where both related (i.e. belonging to the same family and/or branch) and unrelated languages interacted. Having rightly noted that many of the works on language contact have generally paid little attention to the distinction between the behaviors of related linguistic codes and non-related ones, Epp -Huehnergard -Pat-El (2013) introduced important observations on the similarities and dissimilarities of ancient language contact phenomena. The case studies involved Semitic and non-Semitic as well as Indo-European Anatolian languages.2 Only in recent years, the identification a so-called "Greek-Anatolian" area has been proposed, which would constitute yet another example of such a contact scenario. However, the case has not yet been within the framework of a language contact in the Ancient Near East. The identification of the particular case under discussion goes back to the famous study by Watkins (2001), which has given rise to a number of follow-up works. Before dealing with such this topic, however, some terminological as well as methodological issues will need to be tackled, in order to provide a grid of parameters which enable a solid discussion.

3
As a matter of fact, when we come to the matter of defining concepts such as 'linguistic area' , 'linguistic league' , 'Sprachbund' , etc., in terms of language contact, the positions by several authors tend to remain vague.3 In defining a Sprachbund, scholars generally had no positive, shared criteria; rather, there is a tendency towards gathering common features that are specific to each situation and area. Nevertheless, in the conspicuous publications on this topic, theoretical models and various parameters have been proposed.
Therefore, in order to disambiguate the specificities of the different definitions of 'Sprachbund' , 'language union' , or 'area' and 'language / linguistic league' -the multiplication of the labels been often the result of metalinguistic translations -or again of 'convergence area' also with regard to the criterium of relatedness, a selection of chosen parameters will help clarify the differences (Table 1), (on relatedness criteria see Drinka 2013).
The history of these terms4 indeed shows how scholars have started from a definition of 'Sprachbund' as opposed to that of 'area of convergence' , to eventually arrive at a superimposition or neutralization of the differences between them (Aikhenvald 2006).5 A few notes on some concepts will help to better highlight the differences in the mechanisms of diffusion of the features and of realization of the convergences.
Under the premise that "Mutual influence is what significantly distinguishes between a league and an ordinary two-language contact situation. In the latter the influence is usually unidirectional, i.e. there is no partnership" (Urban 2007: 141), we consider hence the concept and the dynamic of 'language league' as different from the case of 'borrowing' between languages in contact. Moreover, in a league the main innovations of shared features are: the development of a new trait, or its loss; the unexpected retention of an old trait, the lack of development of an expected one. In this sense, the conditions identified and described by McMahon -Matras -Vincent (2006: 669) are noteworthy, namely that the shared traits cannot be typologically prevalent worldwide; should emerge in most languages in an area; should not emerge in related languages or in other languages of the geographical area that is affected; should not be traits that too easily emerge in languages. 3 We refer to Urban, (2007: 137ff), who discuss in detail the differences between Sprachbund and language area. 4 See Campbell (2017) with further references. 5 As a complement to this first methodological portion of this work, it is useful to clarify that the choice to use the term 'linguistic area' depends also on the considerations by Schaller (1975) and Joseph (1992 The term 'convergence' implies that the shared phenomena are a result of a mutual process of change among languages after a significantly long mutual exposition in specific geographic locations. 'Diffusion' on the contrary, is a term that merely refers to the mechanism of circulation of the shared trait or traits. The presence of related and/or non-related languages has also an impact on the definition of linguistic areas and on the way they may originate, an issue that is not irrelevant to the discussion on the Anatolian-Greek area. Basing their discussion on the methodological impossibility to project family languages indefinitely into the past, some contemporary linguists such as Dixon (1997) assumed that both the birth of language families and the emergency of linguistic areas are viable scenarios that may have shaped the general geography of mutation. According to Dixon (1997), the dynamic of convergence may be applied even to the model of the genealogical tree. The model would include "unstable" periods, which Dixon defines in terms of "punctuation" and during which, due to minor or major movements of peoples, former "equilibria" collapse, resulting in a differentiation of the languages. On the other hand, there may be long phases of convergence, which end up producing a new proto-language.6 Such considerations, while highly stimulating, appear in any case more suitable for macro-comparatively investigating the formation of language families, and the possibility to specifically apply them to the Indo-European one remains matter of debate. (See Francois 2014).
In a similar fashion, Watkins (2001: 62f.) adds, when referring to the Indo-European linguistic history, that both the diffusion by contact and the language-internal change occur very quickly even during phases of stability. Equilibrium, according to Watkins, is therefore a "theoretical construct" that is conceivable, at latest, for the Paleolithic period, "where nothing much was going on" (2001: 62). Diffusion and convergence, for Watkins, follow dynamics and mechanisms different from the ones proposed by Dixon (1997), who, on the other hand, wrote: "[…] the formation and development of genetic families and the formation and development of linguistic areas […] each […] [has] its own dynamic, its own history, and its own life and fall.
[…] Both […] would have their own distribution of rapid and abrupt and slow gradual change […]" (quoted from Watkins 2001: 62, 63). In defining the history of the Indo-European languages, Watkins goes on stating that "the formation of diffusional linguistic areas is, on the one hand, relatively rapid (a matter of half a millennium or less), and, on the other, coexists with normal and relatively rapid genetic differentiations and the formation of species". As a response to 6 Dixon's statements, he remarks that the Indo-European examples show that both "contact-induced linguistic change (i.e. diffusion) and system-internally driven linguistic change may occur with equal abruptness and rapidity -thus both counting as 'punctuation'" (Watkins 2001: 62).
While not being able to offer a final solution to the clarification of the different metalinguistic terms and concepts pertaining to the areal phenomena and to the issue of their classification, it is tempting to suggest that the term 'linguistic area' might define a static geographical situation (of equilibrium) where (both related and unrelated) languages interacted with each other through former or current migrations. The languages affected may share features, but these must not be the result of a shared and diffused language change.
As areas do involve shared traits, when trying to identify one it is also important to understand if there are types of features that are better predictors of proper areal dynamics. When comparing the views by a selected group of scholars, some differences emerge that are highlighted in Table 2. The different positions by the scholars seem to carry to the conclusion that the common features to be analyzed need to be identified case by case. The following sections, after these initial, general considerations, will move on to the specific case of the problem of the so-called Greek-Anatolian areal contacts.

2
The Features of the Anatolian-Greek Area

A Methodological Debate
A number of scholars (such as e.g. Puhvel 1991, Watkins 1997 onwards, followed in Italy by Lazzeroni 2006 and his school) proposed the existence of a Greek-Anatolian language area. The attempts were generally based on selected (putative) contact-induced changes, some of which will be discussed below. Also considering the genealogical relationship of the Greek language group and the Anatolian one, it is first necessary to face some specific questions: 2.1.1 What Are the Languages Involved within the Greek and Anatolian Groups? The hypotheses in this regard can be divided into two main types of approaches: -The first is represented by the hypothesis of involvement of a direct Eastern- Greek dialects − Hittite contact builds on the research by some scholars such as Puhvel (1991), Lazzeroni (2006), Bianconi (2015aBianconi ( , 2015b and forthcomings, Romagno (2015); -The second hypothesis is defended by scholars such i.e. Hajnal (2003Hajnal ( , 2011Hajnal ( , 2018, Gander (2010Gander ( , 2021, which paid more attention to the geo-historical collocation of the linguistic cultures in Bronze Age Anatolia and investigated the contacts between Luwic and Greek. To this regard, some considerations are in order. From a historical and geographical viewpoint we can nowadays easily all agree that some forms of Greek -Western Anatolian language contacts should be certainly expected, as the main zone of interface (modern South Eastern Turkey) was probably Luwic speaking (see Hajnal 2018;Cotticelli-Kurras -Giusfredi 2018: 180). From a chronological perspective, it has been supposed that the two languages which were in such a contact-situation and developed contact-induced features may have been Mycenaean and Luwian (or other Luwic languages).
However, one should also concede that contacts between the Greek and the Hittite language are not to be excluded per se (see Gander, 2021, i.p.). In fact, Greek and Hittite historical contacts are documented by the existence of a diplomatic correspondence between Ahhiya(wa) and Hatti, with letters written in Hittite that were found in the archives of Hattuša.7 From a cultural viewpoint, we may certainly quote to Hajnal (2018), who writes: "Mycenaean Greeks are in contact with people of southwestern Asia Minor. From the point of view of cultural history, this fact is hardly surprising: Archaic Greek mythology and the Greek epics show similarities to Bronze Age sources from the Near East" (with reference to Burkert (2005: 292) in the footnote, to which we wish to add West (1988West ( , 1997, also West (1978: 3-13) on Near Eastern influence in 7 For a comprehensive and detailed historical and geographical investigation we refer to the monography by Gander, (2021, i.p.) in which he discusses the identification of the since Forrer Hesiod). The historical facts and questions are best left to the historian: see the discussion by Gander, loc. cit, especially Chapter 9.4, where the author affirms that research in recent years has neglected Cilicia as a possible contact place between Mycenaeans and Hittites and that the connection via western Anatolia and the Aegean was given a too high priority. Whatever the details of the historical developments and events, for the second millennium bc there is generous documentation of the cuneiform texts from the Hittite archives. For the Greek one can rely only on the Mycenaean texts. However, these allow only limited insight into Bronze Age Greek. For this reason, records from the Homeric epics are consistently introduced in the discussion about Mycenaean-Anatolian language contacts. As for the choice to involve Luwian in the equation, this is based on the recent recognition (Yakubovich 2010) of the importance and large diffusion of the different varieties of Luwian (or, at least, of other Luwic languages) in Western and Southern Anatolia (a fact unknown in the mid twentieth century).

2.2
A Discussion of Some Linguistic Data Mycenaean Greeks and Anatolians were in contact towards the end of the Bronze Age. These political and cultural contacts, however, do not immediately entail a scenario of language contact so intensive to justify the assumption of a language league.8 Most of the attested features, indeed have been discussed in literature in terms of what was the specific source language that modeled a change (an approach viable for the description of borrowing), but one should also investigate the type of contact scenario that may have produced the shared innovations.
Keeping in mind the table of features presented in Table 2, we can state that structural borrowings on a phonological, morphological or syntactical level generally occur in intense contact scenarios or, more locally, in a bilingual situation. The question is, hence, how can we prove whether or not this was the case when we come to the topic of the Aegean-Anatolian interface? Also: which direction took the language contact, which was the model language? Did it consistently occur from the Anatolian group to the Greek one (as many studies seem to assume) or vice versa? Which are the possible comparanda, or, again, what are the linguistic documents one may compare? And finally, which are the features involved within the Greek and Anatolian groups? 9 After an initial review of the available literature, the hypotheses can also in this case be divided based on the two main approaches. Both approaches (Greek and Hittite relations vs Mycenaean and Luwian contact) take into consideration both the lexical and phraseological events as also the systematic contact-induced ones (including units of the morphological, grammatical and syntactic layers, see Dardano 2012 and. However, the central question remains: do the considered phenomena emerging in related but different languages depend on contact, or can common inheritance or typological prevalence have played a role? In Table 3 below, the peculiar and most striking features from the Greek an Anatolian corpora discussed in the quoted works by Puhvel, Watkins, Lazzeroni and Romagno are organized based on the traits listed above in Table 2. The following picture leads to the conclusion that those features are not exclusive traits of a contact-induce changes, but they might, in general, be explained as inherited structures. In the next section, a selected group of the traits which were used to propose a Greek-Anatolian area, will be briefly discussed. Morphological Traits The verbal endings -men vs. -mos for the 1 pers. Pl. in a complementary distribution both in the ie languages and among the Greek dialects since the Doric -mes ending (attested also in Latin, Sanskrit, and in the Baltic languages) represent one of the criteria for the dialectal or diatopic and diachronic categorization of the Greek dialects, as Ionic and Attic exhibit a variant -men, which is related to the ending present in Anatolian languages. Since Rix (1976, 243f. § 268 and 251f. § 274), the isogloss mirrors the distribution of the reconstructed forms for the 1pl. *-mes for the primary active endings and *-me for the secondary stative ones, whereby the ending *-me also attested in Vedic -ma, received the -n element in Greek and in Hittite. This distribution may very well reflect rather a geographical behavior of inherited and partially shared traits in earlier phases of the development of Indo-European than demonstrate that it occurred by contact during historical times.
As for the other allegedly shared morphological unit, the verbal suffix -sk-, common Indo-European inheritance is a solid explanation and grammatical interference does not need to be postulated (cf. the discussion in Cambi 2007 andDaues 2009).

2.2.2
Morphosyntactic Traits (Modal Particles, Sentence Particles, Verbal Prefixes) as Grammatical Units Criteria 2.2.2.1. Hittite verbal prefixation, often considered to be a shared feature, is not comparable with the same phenomenon in Ancient Greek. First of all the verbal "prefixation" is a category that is undergoing constant change in Hittite, since the involved verbs do generally not display compounds and the "preverb" could still work as an adverbial element. Secondly, non-inherent telicity is mostly expressed in Hittite through the particle -kan, which has no relevant counterpart in Greek. A very clear example is given by the Hittite verb simple kuen-'to hit' vs. kuen-+ -kan 'to kill' .9 Moreover, verbal prefixes can encode telicity also in other Indo-European languages, like the German ones, which suggests that this could be an inherited strategy (cf. πίνω "to drink" vs ἐκπίνω "to drink up" and German trinken vs. austrinken "to drink" vs "to drink up/completely"; cf. also Romagno 2015: 7ff.).
2.2.2.2. Modality could be encoded through particles as well as through other morphological devices. Hittite had only two finite moods (the Indicative and Imperative), which means that the subjunctive and optative moods were not available. The past tense in combination with the particle -man (ma-an or ma-n=) was hence employed to convey a contra-factual meaning as well as a potential one (s. Romagno 2015: 6f.). Hittite man has been compared with gr. ἐαν and ἄν, (see Beck et al. 2012), from a functional viewpoint, but the Greek particles are syntactically dependent on the use of a morphological optative mood, which is unavailable in Hittite. All in all, the comparison is quite weak.
Furthermore, in Hittite, the particle is always required in order to convey the modal meaning (in any type of clause, either main or subordinate). This is not the case in Homeric Greek.10 Below, I quote a potential optative and a counterfactual indicative, both without a modal particle but with modal meaning. ῥεῖα θεός γ' ἐθέλων καὶ τηλόθεν ἄνδρα σαώσαι (Odyssey 3,231). "Easily, a god could save a man even from afar, if he so wishes." In this instance the optative σαώσαι has potential meaning, and yet no modal particle has been used. ὦ γέρον, ἦ ὀλίγου σε κύνες διεδηλήσαντο ἐξαπίνης, καί κέν μοι ἐλεγχείην κατέχευας (Odyssey 14,(37)(38). "Old man, suddenly, the dogs had almost torn you into pieces and you would have brought great shame on me." In this instance both the indicatives διεδηλήσαντο and κατέχευας have counterfactual meaning, but in the first instance this meaning is signalled by ὀλίγου "almost" and in the second instance by the modal particle κεν.
Finally, when comparing the modal particle use in Hittite with that in Greek, scholars have traditionally used Attic Greek as comparandum, but there is a substantial geographical and chronological difference between Hittite (xvii -xii bc) and Attic Greek (v -iv bc) and, contrary to Ionic and the Western Anatolian languages, neither of them has ever been geographically contiguous in historical times. Apart from this, it is also noteworthy that the use of the modal particle in Attic Greek is rigidly regulated (as result of a grammaticalisation process), but this is not (yet) the case for epic Greek, in which the particle seems to have a more deictic value and not a proper modal meaning.11 12

2.2.2.3
Absolute Participial Constructions Since Romagno 2015: 10f. brought this example as an open question regarding the possibility to be a shared trait of the Anatolian-Greek language contact, it is worth noting that in the whole Hittite tradition there are only four known examples of a so-called absolute participial construction, and all of them are referred to the same two verbs ('to stay' and 'to sit'), written in their sumerographic form, gub and tuš respectively, in the phrases: 'the king sitting/staying, libates (the gods)' . These isolated forms are still debated in the literature12 and it is not possible to compare their function (maybe participium conjunctum) with that of the Greek absolute participle construction with the genitive. It can, therefore, hardly be considered a shared trait. Furthermore, the Genitive Absolute (ga) might very well have originated within Greek itself, being and independent phenomenon, perhaps starting from constructions in which the subject of the ga was first an object of a verb of the main clause, after which the construction grammaticalised into an independent syntagma. In epic Greek we find examples of both. For a detailed analysis, see Classen (1879: 134-188) and Ruppel (2013).13

2.2.3
Phonetic Features Assibilation in Anatolian was thoroughly described by Kloekhorst 2008, edhil: 91f., and its counterpart represents a new classificatory element for the categorization of the Greek dialects (northwest vs. (south) east).
Regarding the phenomenon of the constraints of assibilation, the ending of the 3rd person sg. *-ti becomes -zi in Hittite and -σι in the Greek dialects of the Circum-Aegean area, i.e. Ionic-Attic, Lesbian, Mycenaean and Arcadian-Cypriot; Hittite <z> probably represents an affricate (Kronasser, 1955: 61) and in Greek -σι the sibilant outcome implies an affricate stage (see Romagno 2015: 4). This phenomenon has been considered a possible shared feature. but it is worth noting that the corresponding Luwian isogloss in a direct neighboring context retains -ti and does not share the innovation: affrication before final /i/ is, indeed, merely a Hittite sound change. This forces us to doubt the areal connection because of the geographical distance and the unlikelihood of direct contact.

2.3
The Distribution of Selected Features: An Overview All in all, considering also the other direction of the language contact, from Anatolian to Mycenaean, based on Hajnal's discussion (2018) we can describe the following alleged phenomena in the given distribution (Table 4):

Summing Up on the Greek-Anatolian Area
The hypothesis of an Anatolian-Greek language area has been a much debated one, also for theoretical reasons. First of all, it is often treated as an area consisting of only two related languages, though, in fact, the reference to "Hittite" 14 For the accusativus graecus see also the discussion in Smyth-Messing (1956: 360-361) where it is catalogued under "The Free Uses of the Accusative". It exists in Attic Greek, but is used very limitedly.
on the concept of an anatolian-greek | 10.1163/26670755-01010006 Old World: Journal of Ancient Africa and Eurasia (2021) 1-19 14 should often be replaced by a wider reference to the languages of the Anatolian group. Indeed, direct Hittite-Greek language contact is unlikely to have happened for geographical and chronological reasons. Still, the phenomena analysed were often typical of Hittite but not of the Luwian language, or vice versa, which represents an obvious problem. Another anomaly is represented by the direction of the supposed shared phenomena, which seems to be unidirectional. Thus, scholars systematically tried to find influence from Anatolian (i.e. Hittite) to Greek, ignoring the opposite direction. For the importance of mutual influence in areal contexts, see also Campbell (1985: 20f., and. Finally, as already highlighted in the work by Hajnal (2018) Anatolian appears to be consistently the putative source language, with little or no trace of mutual influence within the alleged area. Gander (2003, 2011) do not exclude all kinds of Mycenaean-Anatolian contacts, but assumes that they may have existed only for a limited period of time and in a localized geographical region. (On the Ahhiyawa-debate see Fischer 2010).
Summing up, if we check the above mentioned linguistic traits, we can conclude, concerning the Greek-Anatolian area, that: a. Some traits are indeed typologically common to many languages in the world; b. The alleged shared traits emerged only in part of the languages of the area; c. Other languages that belong to the same (Indo-European) family of those involved also share some traits. In light of these considerations, it appears that the available evidence may call for some cases of interference between Western Anatolian and the Greek languages, but does not justify the hypothesis of a language league.