1.1 The Concept

By grammatical interference, we refer to those situations of contact in which one language influences another language beyond the level of simple lexical exchange, with a potential impact on the structural layers of the target. Structural layers include the phonemic inventory, morphology, and syntax (see below, § 2). This type of interference is not uncommon but, contrary to the simple borrowing of words or the occasional calque of finite structures, requires a condition of stable and rather intense coexistence of different cultures to produce stable change. If this condition is met, it may alter the phonemic inventories of the languages involved; it may affect the morphology, producing changes in the typology if it goes beyond the occasional borrowing of morphemes; and it may affect the syntactic inventory, with shifts in the syntactic typology and phrase and clause architecture of idioms.

These types of phenomena are rare although not unprecedented¹ in simple borrowing scenarios but are frequent in multilingual contexts. In situations of bi- or multilingualism—either polarized binary systems in which two codes exist and one prevails over the other or more complex mixed societies in which multiple codes and varieties of codes generate fluid developments,—some features of different languages may become increasingly similar over time. Binary systems are traditionally described by the substrate-superstrate model, in which an endemic language (the substrate) enters into contact with an ‘intrusive’ language (the superstrate), which becomes sociolinguistically more prestigious.² This is what happened in colonial contexts throughout human history (cf. Matras 2009:300). In such contexts, the superstrate language carries prestigious new lexical labels that are borrowed by the local substrate, which, however, tends to maintain its grammatical features. When the coexistence lasts for long enough, the substrate and superstrate may gradually merge into a contact language such as a pidgin (a first-generation mixed language) or a creole (a stable, long-lasting mixed language).³

1.2 In and around the Ancient Near East and Anatolia

In the study of ancient languages and the ancient societies and cultures that employed them, the identification of grammatical interference can aid in the understanding of the social, cultural, and political context of peoples, groups, and sometimes even polities. Yet methodological issues and limitations make identifying these phenomena extremely difficult. A macroscopic problem complicates the identification of long-distance systems of interference, which operate indirectly and subtly. For the ancient world and the ancient Mediterranean in particular—of which the ancient Near East is one of the best-documented areas—a significant number of the languages that are well understood belong to the same genealogical families or branches. Many of the languages of Asia Minor, Syria, and Mesopotamia can be grouped into the Semitic group of the Afro-Asiatic family or assigned to the Indo-European family. When similar features emerge without obvious evidence for prolonged and intensive interaction, it is often difficult to disambiguate interference from inheritance.⁴

This does not mean that some widely shared features may not be hypothetically identified in the ancient Near East, but the idea of long-distance interference is usually easily challenged by the hypothesis of common inheritance in one or more language families or groups, meaning that long-distance interference can rarely be proven. A good example of this is an alleged long-distance case of convergence that involves Anatolian, Indo-European, and languages of the ancient Near East. An oddity of many languages used between northern Mesopotamia and the Aegean is the absence or near absence of words beginning with a polyvibrant rhotic consonant (cf. Bianconi 2015:139, followed by Romagno 2015:432). In the Anatolian branch of Indo-European, neither Hittite nor Luwian have words beginning with /r/ (the divine name Runtiya being in all likelihood the result of a former K(u)runtiya).⁵ There is no reason to assume that Palaic behaved differently, although the available thesaurus is not rich enough to allow a final determination. Lehmann (1951) observed that Hattian and Hurrian, which were areally close to Anatolian but did not belong to the Indo-European family, shared this phonotactic constraint. Romagno (2015:432) remarked that initial /r/’s also do not occur in Greek and Armenian (the initial ῥ of Greek is, in fact, /hr/). While a shared tendency of geographically close languages to eliminate the initial rhotic is not excluded, it is very difficult to prove. The reasons for the exclusion of the initial /r/ in Hattian and Hurrian (the latter also excluding initial /l/, while the related Urartian language excludes neither) is impossible to explain diachronically as there are no known proto-languages whose evolution we may reconstruct, but Lehmann (1951) argued against the reconstruction of Indo-European roots with initial /r/ by showing that most of the roots that were assumed to begin with an /r/ instead had an initial laryngeal or glide. Hence, the lack of initial /r/ may have been an areally shared feature, but this is not demonstrable. Whether Anatolian, in turn, influenced Hurrian and Hattian, is difficult to determine, but the Mesopotamian (and, perhaps earlier, Iranian?) provenance of Hurrian, as well as the limited protohistorical contacts with the Hittites, do not provide much support for this theory.

A similar case, once again involving the structural level of phonology, is represented by the final -n of Greek and the Anatolian languages that derives from an etymological *-m (Bianconi 2015:139). Here, the identity of the sound laws could conceal a phenomenon of convergence if the convergence could be dated to a plausible moment in the relative chronology of the development of the phonemic inventories of the Anatolian and Hellenic branches. According to Melchert (1988), a few forms attested in the Luwian conjuration KBo 12.260 (tu-u-ri-im=ša-an, mu-ha-at-ra-am=ša-an ha-at-ta-ra-am=ša-an pí-iz-za-ar-na-am=ša-an, and the syntactically elusive pariyam=ša=tta, which contra Melchert is unlikely to represent an instance of the particle -ša/za) could testify to the retention of the original -m before a sibilant. Given that we are dealing with the boundary of a morpheme, this would have had to have happened synchronically, proving that the shift from -m to -n was in progress when the text was composed. However, this would contradict the Proto-Anatolian date of the change that Melchert himself (1988:214) supported. As the aberrant forms all occur in the same tablet in similar phonetic contexts, we believe that they are best explained as mishearings by a scribe taking dictation rather than retention of an archaic ending, which must have changed to -n centuries before the text was written down. Once these forms are eliminated from the equation, the change from -m to -n in the final position seems to be a regular Proto-Anatolian sound law that is identical to a sound law that must have existed in Proto-Greek or even in Greco-Phrygian (if the group is real, because Phrygian also exhibits the feature). Contact during the protohistory of the two branches could have triggered a convergence, but it would be difficult to date and localize it. Furthermore, we are dealing with two sound laws, meaning that independent inheritance and development remain possible, especially since, pace Bianconi (2015:139), Anatolian and Greek are not the only Indo-European languages to exhibit such a typologically common change. For example, Proto-Germanic has the final -m > -n, which emerges in the accusative ending just as it does in Greek, Phrygian, and Anatolian). These proposed areal features are perfect examples of unsolvable problems that, given our poor and speculative understanding of protohistorical geographical scenarios, may be explained in terms of convergence as well as inheritance.

While cases of wide convergence are certainly difficult to identify in the early stages of the linguistic and cultural geography of the Mediterranean world, the study of ancient contacts is challenging even when we reduce the scope. Subtle problems emerge when dealing with localized superpositions of codes. The conditions that trigger superstrate-substrate systems existed as a consequence of social shifts or geopolitical events. The Anatolian languages penetrated most deeply into Syria when Hatti’s political influence replaced that of Mittani as the northern counterpart of Egypt, in the late 14th century BCE. The Hurrians acted as sociolinguistic ‘newcomers’ in Syria from the Middle Bronze Age, and the presence of Hurrian words in the local Semitic grapholects is well documented.⁶ In Middle Bronze Age Anatolia, the Old Assyrian trading network acted similarly in Cappadocia, and forms of grammatical interference emerged in texts produced by Anatolian scribes, who had trouble dealing with the use of gender when writing Akkadian (see Chapter 4 for further discussion and references).

Before entering into the discussion of grammatical interference in Late Bronze Age Hittite texts, it is worth dwelling on the last two examples (Hurrian influence in Syria and Anatolian influence on the Assyrian language in the Kārum age) as they can demonstrate the fundamental methodological problem of dealing with binary substrate-superstrate systems in the ancient Near East. The role of Hurrian in Syria and, more specifically, in Ugarit, which is discussed by Andrason and Vita (2016:306–307), is that of a mixed code consisting of the use of a language, Ugaritic, that was spoken by the local population, combined with borrowings from a higher-ranking language, Hurrian, that was connected with the ritual sphere among other possible functions. What makes this case particularly unusual is the fact that the substrate, Ugaritic, is directly represented in the written records, which allows us to observe the phenomenon of lexical transfer from a superstrate to a substrate.

Such a direct and generous recording of the local substrate in the ancient Near East is the exception, not the rule. Old Assyrian Anatolia (Chapter 4) is a perfect example of this issue. Due to the tight link between the cuneiform writing system and Akkadian, the cultures that adopted writing in the ancient Near East generally used Akkadian as the (sometimes exclusive) associated grapholect. This resulted in the development of very interesting codes that are sometimes called peripheral Akkadian⁷ but at the same time masked the substrata because the Akkadian that was written in areas such as the southern Levant and Canaan was itself a superstrate and a somewhat artificial code. Similarly, in Middle Bronze Age Anatolia, Hittite, Luwian, and Hattian texts were not composed, and the substrate role of the local vernacular emerges only in the errors made by some scribes because of their limited competence in Akkadian (a phenomenon that pertains to grammatical interference). Few loanwords existed (see Chapter 4), and they originated from the need to refer to items or concepts that were unknown to the Assyrians—for example, institutions such as the tuzzinnum social group⁸ or local realia such as the išhiulum belt.⁹ Whether, as would be expected, lexical transfers were made from the Assyrian superstrate to the local spoken language(s), is impossible to say, because we have no representation of the variety of Hittite spoken in Kaneš except personal names (on which see Kloekhorst 2019, Giusfredi 2020, and Yakubovich 2021, including the references to previous scholarship that they contain).

So far, we have discussed the issues that hinder the definitive identification of long-term, long-distance interference and those that complicate the analysis of local substrate-superstrate systems. The last case that requires methodological discussion is the one that is usually the most interesting for the linguists working with modern languages: the identification of mixed languages. To begin, it is appropriate to observe that there are no qualitative differences in the structures or functions of modern and ancient human languages that would prevent us from assuming that pidgins and creoles existed in the ancient world. The problem is twofold. Firstly, we need to ask ourselves if they were ever identified. Secondly, if they were not, we need to try and understand why.

The first part of the question is easily and negatively answered: we do not have evidence of something like Bislama or Tok Pisin in the ancient world,¹⁰ let alone in the ancient Near East. In light of the very sensible observations made by Andrason and Vita (2016:316–324), the mixing of Canaanite and Akkadian in the way the superstrate, Akkadian, was written in the Late Bronze Age southern Levant is indicative of significant interference, facilitated by the structural similarities of the two Semitic languages. However, it cannot be compared with the birth of a new language, as can be easily observed in modern Vanuatu. Similarly, the use of the term ‘creole’ to describe the form of Akkadian that was written in the Late Bronze Age Nuzi, in northern Mesopotamia, under influence of the local Hurrian elite, is improper, to say the least. Although some grammatical features of the language are Hurrian or Hurrianizing,¹¹ it remained Akkadian, at least in the written records.

The languages spoken in Canaan in the Final Bronze Age and Nuzi in the early Late Bronze Age were probably, respectively, a mix of Canaanite and Akkadian and Akkadian and Hurrian. Therefore, these unattested mix-languages could have been similar to modern pidgins and creoles. The problem, and the answer to the second part of our question, is, once again, the underrepresentation of substrata and spoken languages in the records. But if, as a general rule, definitive evidence for the existence of true creole languages cannot be produced for the ancient Near East, numerous small areas were multilingual or located at the interface of two linguistic cultures. Canaan, Nuzi, and Ugarit, studied by Andrason and Vita (2016), are three examples, but several others exist even if we limit the scope of the survey to Anatolia and its immediate neighborhood.

The Hittite capital city, Hattuša, during the Late Bronze Age will be the focus of the next pages, given the well-documented coexistence of Hittite and Luwian there and the presence of Akkadian and Hurrian, at least as written codes. However, we will also look at cases from Iron Age Syria, where Luwian arguably replaced Hurrian as the official language by the early 13th century BCE and became an endemic idiom in the following centuries—but, for reasons of periodization, these Iron Age cases will be discussed in the second volume of this work.

3.1 Hattian, Hittite, and Palaic

We will maintain the geohistorical periodization proposed in the previous chapter and consider first the circulation of grammatical features in the northwestern network formed by Hattian, Palaic, and Hittite. As previously discussed, this network was an early one. Palaic must have been extinguished or almost extinguished by the time that scribal activities began in the Hittite archives; most putative interferences would have occurred earlier. As for Hattian, the proposals that have been made for contact-induced grammatical change also pertain to a pre-Hittite late Middle Bronze phase,¹³ although Hattian was probably spoken for a longer time than was previously assumed.¹⁴

As mentioned in Chapters 4 and 9, Goedegebuure (2008) proposed that certain features present in the Hattian language derived from an Anatolian language before the age of Hatti (in her view, the model language was probably Luwian, but Hittite and Palaic appear to be good candidates for geohistorical reasons). These features are generally unexpected implicational correlates that should not emerge in a VO language in terms of typological prevalence. As mentioned in Chapters 4 and 9, some unexpected correlates may emerge because implicational universals, despite their name, are not mandatory, and some languages violate a few of them. Nevertheless, the number of irregularities identified by Goedegebuure makes it very likely that at least some resulted from the influence of neighboring languages, especially since historical contacts between Hattians and Anatolians are confirmed by the data that we possess on Middle Bronze Age Anatolia.

In contrast, no evidence exists of grammatical influence by Hattian on Hittite.¹⁵ The unusual constructions that emerge in bilingual documents from the Late Bronze Age archives of Hattuša are best explained as instances of translation interference (or else result from the translation process); they do not reflect changes in the grammar of the target language. However, Hittite was not the only Anatolian language in contact with Hattian. As discussed in Chapter 12, Palaic was an obvious candidate for interference because it was spoken in an area that was near or even overlapped with that in which Hattian was spoken. Consequently, hypotheses exist that involve some peculiarities of Palaic being derived from Hattian. These (according to Goedegebuure 2008) include the presence of a morphophonemic assimilation of /n/ to /m/ before a bilabial stop, which (based on the way the Hittites wrote), could have been shared by Palaic and Hattian, and the use in Palaic of a contrastive particle, /pi/, that allegedly matched the Hattian /bi/ (a possibility, however, that is made unlikely by the possible Indo-European origin of the former).¹⁶ While Palaic is only attested in the texts composed in Hattuša during the Late Bronze Age, if any of these hypothetical interferences on the grammatical level occurred, they must have taken place during the Middle Bronze Age or earlier because, as discussed in Chapter 12, there is no solid reason to assume that Palaic was a living language after the 17th century BCE when the Hittites started writing it in cuneiform.

The alleged phenomena of grammatical interference that may have involved the area to the northwest of Hatti share three important features:

1. All must have occurred very early, probably during the Middle Bronze Age (cf. also Chapter 4).

2. In consequence, they have been reconstructed or observed based on their reflections in later texts that, significantly, derived from a Hittite scribal praxis.

3. Given the previous points, no evidence illustrates the diachrony of the changes.

Points 2 and 3 have significant methodological implications: that each and any of the phenomena summarized above was the result of a change—let alone a contact-induced change—is only a hypothesis based on comparative and contextual observations. No phase of Hattian with pure VO correlates is attested before the putative change that resulted in a mixed typology; no phase of Palaic is attested in which the contrastive particle -pa/pi is not yet present; and as was observed in Chapter 12 for the existence of a fricative /f/, there is no way to exclude the possibility that the notation of an assimilated nasal before bilabial stops was not extended to Palaic by Hittite scribes who had learned to employ it for Hattian.

3.2 Akkadian and Hittite

As discussed in Chapter 8, the label ‘Boğazköy Akkadian,’ frequently applied to the language employed by Hittite scribes for the composition of texts in Akkadian, should be regarded as an umbrella term because various Akkadian grapholects emerged in Anatolia in different phases and for different purposes. The waves of Akkadian grapholects used in Hattuša were functionally different and resulted in the production of different types of documents. The types of interference phenomena that emerged were also various and require a dedicated discussion.

During the Middle Bronze Age, the penetration of Assyrian produced a substrate-superstrate system. The substrata may have been multiple (Hittite, but possibly also Luwian or Palaic in less well-known areas of the trading network), but the superstrate was Assyrian in the written documents from the Kārum age. Assyrian received a limited number of borrowings from the local, lower-ranking vernacular(s), mostly names of local realia or institutions,¹⁷ and documents that were composed by local scribes seem to show influence on the grammatical level. Unsurprisingly, the features that derive from interference in the Old Assyrian documents are the typical results of competence errors by non-native speakers. Morphological and morphosyntactic categories of Semitic that do not exist in the Anatolian languages or Hattian, such as the gender marking of the inflected verb or the distinction between masculine and feminine pronominal forms, are often misused.¹⁸ It must be stressed that the rich prosopography of the families of archive owners from the kārum of Kaneš provides us with the final piece of evidence to confirm that grammatical interference was indeed at work: the grammatical mistakes not only match the main differences between the Anatolian and Assyrian grammars but also occur in documents that belonged to local families rather than foreign traders (see also Chapter 4).¹⁹

When, not earlier than the late 17th century, scribal offices begin to produce texts at the court of the Hittite kings in Hattuša, Akkadian is reintroduced as one of the many written languages (see Chapter 8). It is first employed in the so-called Old Hittite political texts, where, as already discussed in this book, it frequently accompanies Hittite in direct bilinguals. The Akkadian grapholect of these early documents generally contains structures that calque Hittite structures. Akkadian influence on the structure of the Hittite versions is harder to detect. The most common phenomenon of Hittite interference in Akkadian is the almost ubiquitous doubly marked genitival series ša X Y-šu ‘of X, his Y,’ which is not grammatical in Akkadian either because of the left collocation of the lexical genitive, which is typical of Hittite. Other examples, which were mentioned in Chapter 8, include calques of specific constructions, such as serial verbs or uses of the verb ‘to be’ (bašû) with participles. These phenomena can generally be detected easily. They stand out in a language that is, otherwise, a Syrianizing variety of Old Babylonian. They bear witness to the scholarization of the scribes in Hattuša in the early ages of the kingdom of Hatti and on that of the ‘original language’ in which bilingual documents (such as the Annals, Testament of Hattušili I, and the later Edict of Telipinu) were written.²⁰ Nevertheless, the local spoken language, presumably Hittite, does not seem to have undergone any significant influence on the structural level, even in bilingual texts; phenomena of interference were limited to the written code. This was also true later when Akkadian grapholects were used for other types of documents, including the land grants, international letters and treaties, and local versions of Mesopotamian literary and scientific texts.

In sum, in historical times—during the Middle and Late Bronze Age—Akkadian varieties were used in Anatolia as written codes by non-native speakers (local archive holders and local scribes in the kārum society who composed Akkadian texts in the age of Hatti), but no evidence emerges that Akkadian influenced Hittite or other Anatolian vernaculars beyond the lexical dimension of borrowings and calques (on which see Chapter 14). This scenario relegates Akkadian to the sociolinguistic role of a set of grapholects that were consistently employed by the Hittites for several reasons: the prestige of Akkadian; its technical utility in administrative, ritual, and scientific contexts; and its importance as a lingua franca in diplomacy. It never became a significant spoken language. These observations have implications for any hypotheses of grammatical influence that may be formulated based on mere structural observations. For instance, the isolated feature of the Hittite use of possessive clitics (a feature absent in Luwic) could tempt scholars to compare the forms with the ones that are quite ubiquitous in the Semitic group and of course present in Akkadian as well. Kloekhorst (EDHIL) reconstructed the clitic series in Proto-Anatolian; its presence would imply inheritance or very early contacts. But the forms show Indo-European morphs, proving that they were not borrowed, so the decisive factor to formulate a contact-based hypothesis would be the presence of very early documented contact between Proto-Anatolian and Semitic, which appears to be out of the question. A development in historical times is excluded by the fact the clitic possessive is present in Old Hittite.²¹

Another example of an alleged grammatical influence of Akkadian on Hittite is represented by the proposed role of the Semitic language in the development of conditional periods in Hittite (Zorman 2017). Zorman developed a complex and well-argued hypothesis to account for a range of phenomena. She argues that the Hittite takku ‘if’ was a structural calque of Akkadian šumma (thus a true case of morpheme induction by etymological calque), which, if one follows Speiser’s (1947) analysis, would involve the univerbation of the third person demonstrative pronoun šu and the particle -ma, thus matching the Indo-European etymology of the two components of takku, *to (a pronoun) and *-k^we (an additive marker). Assuming that Speiser’s etymology is correct (although it does not provide a clear explanation for the gemination of /m/, nor for the non-conditional meanings observed by Moran 1954), the structural similarity of the Akkadian and Hittite (pseudo-)subordinator (see Cohen 2012 on the syntactic status of šumma) would be striking. However, it is only fair to acknowledge that other hypotheses exist, so Speiser’s analysis should be presented as widely but not completely accepted (cf. Cohen 2012:115–116 and fn. 122; Cohen emphasizes that the presence of the -ma particle is virtually certain but does not commit himself to any interpretation of the first morph). Furthermore, the existence of a pronoun *to in Old Hittite is merely conjectural. Melchert (personal communication) suggests rather an etymology from *toh₂ with the meaning ‘further,’ which, combined with an early conditional value of the very clitic -kku (cf. Watkins 1985), provides an alternative explanation for the compound that would not require influence from Akkadian.

Zorman’s second proposal regards the use of the preterit and present in Hittite conditional clauses with mamman at the protasis and the clitic -man in the apodosis. She suggests that such clauses may use tenses to mark aspectual nuances in a manner deriving from the grammar of western peripheral Akkadian with, once again, a striking formal match between the Hittite clitic modal particle -man and the clitic Akkadian conditional marker -man.²² It is the similarity between the Hittite and Akkadian clitic elements that prompted Zorman’s third proposal: that the Hittite -man developed from the orthotonic subordinator mān ‘when, if’, which can be used hypothetically in the protasis and this development was facilitated by the existence of the homophone Akkadian -man, which has a near-identical function.

The cases on which Zorman’s three hypotheses rest are the only well-founded examples of proposed areal features involving Hittite and Akkadian. As all of the phenomena would be limited to Hittite and would not affect the Anatolian languages that were geographically more distant from Syro-Mesopotamia (Luwian and Palaic) and since the formal matches are as visible as the semantic ones, this hypothesis was well worth formulating. From the purely linguistic standpoint, it is difficult to propose a conclusive counterargument, but it is worth noting that the hypothesis about takku relies on speculative etymological explanations of both Hittite and Akkadian forms, the aspectual use of the tenses in the Hittite mamman + -man periods does not contradict the general uses of tenses in Hittite, and the development of modal -man in Hittite does not necessarily require a contact-based explanation. However, these points merely open the door to alternative explanations without disproving contact-based possibilities.

For Zorman’s theory to hold water historically, it is necessary to identify a stage in which Hittite and Akkadian were in contact. It must have followed the hypothetical Proto-Anatolian stage (because the innovations only occur in Hittite), preceded the early written records from Hattuša (because all of these features are present in texts from the Old Kingdom), and predated them by long enough for us to be able to assume that *-ku was still employed as a conjunction and ta, used only as a connective in Old Hittite, was still understood as a pronoun (provided, of course, that it ever had that function, which, as previously mentioned, it may not have had). The only reasonable candidate for such a stage and setting is the Old Assyrian age of pre-Hittite Anatolia, a possibility that Zorman herself seems to cautiously suggest.

The problem can be hence formulated as follows: could the Assyrian presence in Anatolia during the so-called Kārum age have been the factor that triggered a significant grammatical change and introduced some structural patterns of conditional clauses into Hittite? If this occurred, it would contradict the weak sociolinguistic pattern that we proposed relative to the interaction of Assyrian and the local vernaculars, which we characterized as limited in its effects and confined to specific portions of the lexical level. However, the detailed chronological inquiry by Barjamovic, Hertel, and Larsen (2012) reveals that the phase for which a strong presence of Assyrians could be assumed (based on documents from the archives of Kaneš) was limited to a period of approximately 40 to 50 years that corresponded to the last third of level II of the Kültepe site. As the three authors argue convincingly, very few Assyrians resided in the kārum before that date, and their business model did not entail a stable presence of traders in the city. Subsequently, in the period corresponding to level Ib of Kültepe, the presence of the Assyrians in the kārum again seems to have been very modest. Possibly local traders partly replaced Assyrians. Whether the significant presence of Assyrians lasted long enough to hypothesize that Assyrian was widely spoken and had a strong structural influence on the local vernacular rather than being a superstrate employed for the composition of technical texts is debatable. However, there was not a long period of coexistence between native speakers of Assyrian and Anatolian. A further consideration is the number of Assyrians who lived in the kārum. Although their presence was a phenomenon of the utmost historical importance, Barjamovic, Hertel, and Larsen (2012) estimate that fewer than 1,000 Assyrians resided in Anatolia or worked there for a part of the year when the Assyrian presence in Anatolia was at its height. Of these, a few hundred were probably stationed in Kaneš (which was inhabited by some 25,000 Anatolians, according to Barjamovic 2014, if we include the peripheral blocks and outskirts). Furthermore, the density of Assyrians was probably greater in Kaneš than in other gateways to the network: for instance, it would be unrealistic to assume that hundreds of Assyrians inhabited the kārum in Hattuša. If we add that Kloekhorst (2019) has shown that the Hittite dialect of Kaneš was not the same as that spoken in Hattuša, more problems become apparent. Did a few hundred Assyrians scattered across the trading network in the Old Assyrian age constitute a strong enough presence for their language to project grammatical features on Hittite in general? If not, should we assume that the innovations occurred specifically in Kaneš, where the Assyrian to Hittite ratio was low but at least higher than in the north or west? But if we assume this, how would these innovations have spread to Hattuša Hittite, which was the Hittite variety used for the documents written in Hatti during the Late Bronze Age?

In light of these issues, we can only conclude that the hypotheses of Akkadian grammatical influences on Hittite that were proposed by Zorman (2017) are not impossible per se but contain speculative steps. It is currently impossible to propose a convincing sociolinguistic setting in which they could have occurred. We would have to assume earlier phases of contact—but no such phases are documented and, geohistorically, it is difficult to conceive how they could have taken place.

3.3 Hurrian and Hittite (and Anatolian)

Despite the importance of the Hurrian documents from Hattuša for our understanding of the Hurrian language, the intensity of the historical contacts between the Hittites and Hurrians have probably been exaggerated in the literature, at least for the early ages. While Hurrians were important neighbors of the Anatolian peoples and polities by the Middle Bronze Age and the early kings of Hatti confronted them regularly during their own southeastern campaigns, there are no solid arguments to support the hypothesis that the Hurrian peoples and language had a widespread presence anywhere to the north of northwestern Syria and Cilicia. The Hurrians had penetrated beyond Upper Mesopotamia and the Jazira, but a Hurrian presence in the core area of the central Anatolian kingdom of Hatti seems to have been limited to specific historical phases. The first was the stage of the treaties with Kizzuwatna, and it is at this time that cultural and linguistic materials from the Hurrian world entered Anatolia. Significantly, Simon (2020b) observed that a vast number of Hurrian loanwords in Anatolia followed this path: they occur in Luwian, in texts that emerged from the mixed Hurro-Luwian environment of Late Bronze Age Cilicia. The second phase of Hurrian penetration into Anatolia can be dated to the mid-13th century when the Hurrian religious traditions were revamped at the time of Queen Puduheba. By then, however, the Hurrian materials were entering the Hittite archives in Hittite translation, so the presence of significant groups of Hurrian speakers in the capital city cannot be safely assumed. Therefore, the only stage that appears to be relevant to contact between languages is the early or proto-imperial one, roughly datable to the reigns of Tuthaliya I and Tuthaliya III. However, just as Luwian appears to be an important filter for the penetration of loanwords from Hurrian into Anatolian, it seems to have been the only Anatolian language that might have exhibited some level of grammatical interference with it.

As discussed in Chapter 10, § 4.1, the only credible case of a contact-induced morphosyntactic change in Luwian was identified by Yakubovich (2010:47–53), who spotted it in some Luwian texts from Hattuša that were considered to derive from a Kizzuwatna tradition. This change consisted of some idiosyncratic forms of genitival adjectives which are said to have exhibited a double inflection marker, thus expressing the number of the possessor in a way that recalls the agglutinative Suffixaufnahme of Hurrian. This hypothesis, not uncontroversial, has been criticized by Simon (2016), who proposes that the Kizzuwatna doubly marked genitival adjectives were an internal innovation (through reanalysis of the ending). The situation is further complicated by the fact that the construction seems to appear in a Luwian ritual tradition of Kuwattalla/Šilalluhi and documents connected with the central Anatolian city of Tauriša. For all these texts, a composition in Hattuša has been proposed (Mouton and Yakubovich 2021), which led the same Yakubovich to reconsider the role of the Hurrian influence in the development of the innovative genitival adjective. However, Hurrians and Luwians probably coexisted in Kizzuwatna by the early centuries of the Late Bronze Age, if not earlier. Luwian was certainly not endemic in the Hittite core-area before the imperial age of Hatti. It would not be unreasonable to assume that the penetration of forms that did not belong to the late local variety of the language, Empire Luwian, into some texts composed in Hattuša that had a clear connection to the Kizzuwatna tradition was due to the very origin of the composition. This might explain the unusual forms in the complex Kuwattalla/Šilalluhi ritual tradition, which is connected with the Tunnawiya ritual tradition that originated in the Lower Land (Miller 2004:454; Mouton 2015a) but was re-elaborated in Hattuša. While the Tunnawiya material notably lacks the usual Hurrian elements of other ritual texts and subcorpora, the situation is different for the texts composed or re-elaborated in the capital city. The name Šilalluhi, for instance, is Hurrian (more specifically, either a title or a noun of profession, as per Richter 2012:375). The Tauriša forms are more difficult to account for. Since the city was close to the Zuliya River (probably the modern Çekerek River), it must have been located to the east of the Hittite capital city, so one would expect the variety of Luwian written there to have been comparable to that written in Hattuša. However, Mouton and Yakubovich (2021) have pointed out differences that would indicate a dialectal diversification, which, given the MS date of some of the fragments of the rituals CTH 764–766, would have to have been completed no later than the early 15th century BCE. While some of its features are reminiscent of Empire Luwian, the Luwian of the Tauriša tradition shares with Kizzuwatna Luwian the presence of a handful of doubly marked genitival adjectives. Most of the occurrences are instances of the phrase ‘lord of the rituals’ (e.g., the dative malhaššaššanzan EN-ya in KUB 36.78 iv 7), which may be explained as forms of influence of the Kizzuwatna authoritative religious tradition. The tablets, after all, were written in Hattuša, so the scribes may have extended a label they found in Kizzuwatna rituals to the documents relating to Tauriša, without Tauriša playing a role, especially since the phrase ‘ritual lord’ probably replaced the name of the afflicted person and was added only in the written versions of the rituals. However, the one occurrence in KUB 35.102+103 iii 9 (with parallels at iv 9 and at KBo 8.130 iii 6), wayammana ulipnaššanza ‘the howlings of the wolves’ (n. nom.-acc. pl., although all of the other known occurrences are in indirect cases), cannot be explained in terms of interference between scribal traditions. It contains a genuine intratextual genitival adjective that marks the number of ‘wolves,’ and there is no proof that the phrase was ‘borrowed’ from a different Luwian tradition.

Because of this last occurrence, it is impossible to doubt that the variety of Luwian used in the Tauriša ritual tradition featured this form. However, its other features (Mouton and Yakubovich 2021) do not fully coincide with Kizzuwatna Luwian, so it may have been a different variety of Luwian. Given the position of Tauriša, it is unlikely that any Hurrian-speaking areas were immediately geographically contiguous with it. Even if any were, the development of an identical outcome in two different Luwo-Hurrian bilingual areas would be unexpected.

Thus the problem should be tackled from a different perspective. What appears obvious is that, while the ‘Cilician’ geohistorical explanation given by Yakubovich (2010) is no longer valid, the doubly marked genitival adjective must be a monogenetic feature. This means that it was either 1) inherited from Proto-Luwian or common Luwic or 2) was introduced as an innovation in a dialect from which all of the dialects that contain it derived. Since no trace of similar developments is found in other Luwic languages, the first possibility appears unlikely.

That contact was the trigger of the innovation still appears to be the most likely explanation, as Mouton and Yakubovich (2021) also concede. One may add that, among the languages attested in and around Anatolia, Hurrian remains the best candidate to be the model language for this change, and Kizzuwatna, at some stage preceding the diffusion of the feature, remains a reasonable bilingual location in which it could have originated. All in all, while the new evidence testifies to a very complex diffusion of the varieties of Luwian in different areas of Anatolia over the centuries, it is hardly a conclusive piece of counterevidence regarding the possible Hurrian origin of the doubly marked genitival adjectives.

The arguments in support of this hypothesis, while structurally sound, are not conclusive, either. There appears to be no way to decide whether the contact-based explanation should be preferred, but it should at least be mentioned as the sole possible instance of proper grammatical interference of Hurrian on Anatolian during the history of the Hittite archives.

What can be stated with some degree of certainty and should certainly be emphasized to avoid misunderstandings, is that if the production of the doubly marked genitival adjectives was due to the contact of Hurrian and Luwian in Cilicia, the phenomenon was limited to these forms. An entirely different problem is whether the very emergence of the genitival adjective, regardless of the way number was marked, should be explained in terms of contact with Hurrian (cf. Stefanini 1969 and, for a much wider perspective on the Indo-European family and isolated languages of the Near East, Luraghi 2008).²³ While one cannot disagree with the historical reconstruction that postulates a coexistence of Luwians and Hurrians in Cilicia during an early phase of the Late Bronze Age, we remain unsure whether the pattern of relational adjectival agreement of the Luwic languages and the Suffixaufnahme of agglutinative Hurrian are so similar that they admit an areal explanation (especially considering that genitival adjectives also exist in Palaic and Lydian and are not uncommon in Indo-European; cf. the recent study by Melchert 2012). In conclusion, it seems that no traces of structural interference between Hurrian and Anatolian are safely documented except for the still partly obscure role that Hurrian may have played in the historical development of some of the features of relational adjectivation in Anatolian (or more precisely Luwian).²⁴

3.4 Luwian and Hittite

While the status of almost all of the languages attested in the Hattuša archives is open to debate—with Palaic and Sumerian being almost certainly the only written languages and Hattian, Akkadian, and Hurrian probably spoken by small minorities in specific phases of the Hittite history—it is now certain that Luwian was a language spoken by a significant number of speakers in central Anatolia at least during the imperial age and, in all likelihood, in the pre-imperial phase as well. Since the publication of Yakubovich’s study on the sociolinguistic status of Luwian (2010), contact phenomena have become apparent in the corpus. In the following sections, we will attempt to review and catalog the evidence on grammatical interference between Luwian and Hittite.