LEIDEN UNIVERSITY
FACULTY OF HUMANITIES
T
HE
T
HEMATIC
N
OMINAL
S
TEMS IN
A
NATOLIAN
AND
I
NDO
-E
UROPEAN
MA Linguistics Thesis by
Leire Sarto-Zubiaurre
1
To my family and my friends, for their invaluable support [a]mmell-a-mu-kan LÚ.MEŠ arus LÚ.MEŠTAPPI-YA[-ya] sarriskir1 'and my peers and partners separated from me'
2 Index 1. Abbreviations ... 3 2. Introduction ... 4 3. Materials, methodology and a note on Luwic ... 10
4. The thematic stems in the synchronic Anatolian languages ... 12
4.1. The inflection of the common gender thematic nouns ... 13
4.1.1. The singular cases ... 13
4.1.2. The plural cases ... 26
4.2. The inflection of the neuter thematic nouns ... 32
5. The thematic stems in Proto-Anatolian. ... 35
6. Conclusion ... 38
7. Excursus: Hitt. išḫā- ‘lord’ and the accent of the thematic nouns ... 39
8. Bibliography ... 44
3 1. Abbreviations abl. ablative acc. accusative CHD Chicago Hittite Dictionary CLL Cuneiform Luwian Lexicon
CLuw. Cuneiform Luwian comm. common (gender)
dat. dative
EDG Etymological Dictionary of Greek
EDHIL Etymological Dictionary of the Hittite Inherited Lexicon
EDL Etymological Dictionary of Latin
gen. genitive
Gr. Greek
HD hysterodynamic
HED Hittite Etymological Dictionary
Hitt. Hittite
HLuw. Hieroglyphic Luwian
IE Indo-European instr. instrumental Lat. Latin loc. locative Luw. Luwian MH Middle Hittite MS Middle Script neut. neuter NH Neo-Hittite nom. nominative NS Neo-script OH Old Hittite OS Old Script PA Proto-Anatolian Pal. Palaic PD proterodynamic PIE Proto-Indo-European pl. plural sg. singular Skt. Sanskrit voc. vocative
4 2. Introduction
Since the first handbooks of Indo-European linguistics,2 a twofold distinction was made in the realm of nominal classes, i.e. the distinction between thematic3 (Lat. dominus, Gr. θεός, Skt. deva-, etc.) and athematic (Lat. civis, Gr. πατήρ, Skt. ájas-, etc.) nouns. As already seen by Meillet, the thematic nouns are distinct from the athematic ones, including the stems in -i and -u,4 because they have the so-called thematic vowel and they do not show ablaut or accent shifts.5
The thematic vowel is a morpheme that for PIE is reconstructed as *-o-6 (on the basis of Gr. -ο-, Lat. -u-, Skt. -a-, etc.) and that it regularly appears between the root and the endings, so the nom.sg. is cited as *-o-s, acc. sg. *-o-m-, etc. (e.g. Meier-Brügger 2000:198-9). Much has been written about the origin and the function of the thematic vowel,7 because it does not have a distinct meaning in either the daughter languages or
PIE itself.
Additionally, a different set of endings is traditionally reconstructed for the thematic stems, compared to the athematic ones. Let us take, for example, the singular endings of the non-neuter nouns (Meier-Brügger 2000:198):
2 Meillet 1908:224 and ff.
3 I will use the term “thematic stems” rather than the also widespread “o-stems” since in the main language
of this thesis, Hittite, the “o-stems” are in fact a-stems. “Thematic” is therefore clearer and equally descriptive.
4 Which are athematic. The i- and u-stems behave like consonants stems as per the ablaut patterns they
show, cf. Beekes 1985:198-9.
5 “Cette voyelle change d’ailleurs tout l’aspect de la formation, car elle entraîne fixité du vocalisme de la
racine et de la place du ton dans la flexion” (Meillet 1908:224).
6 It is often claimed that the thematic stems do show ablaut, and as such they are often cited as the
*-o/-e-stems (Clackson 2007:92, Fortson 2010:126). However, Beekes showed that it is not necessary to reconstruct an ablauting inflection, since all of the cases that allegedly had the e-grade (the voc. and gen.sg.) in the thematic vowel can and should be explained differently (1985:184-191). Those alternative explanations are preferable because even if one considers the thematic stems to be ablauting, it is very difficult to justify the presence of ablaut in the vocative alone. Vid. infra for the discussion on the vocative.
5
athematic endings thematic endings
nom. *-s, *-ø *-o-s
voc. *-ø *-e
acc. *-m *-o-m
gen. *-és / -ós / -s *-o-sioi
abl. *-ōt < *-o-etii
dat. *-ei *-ōi < *-o-ei
instr. *-éh1 / -h1 *-o-h1 / *-e-h1iii
loc. *-ø / -iiv *-o-i / *-e-i
ii -es(y)o / -os(y)o (Szemerényi 1996:186) ii -es / -os / -s; -ed / -od (Szemerényi ibid.) iii -ē / -ō (Szemerényi ibid.)
iv -e/-o, -bhi/-mi (Szemerényi 1996:160)
As we can see, the main differences are in the oblique cases: a genitive and an ablative case with different endings in the thematic stems, but a genitive-ablative case for the athematic inflections; a separate vocative ending, and a long vowel in the ablative and dative, that is often explained as the result of a contraction between the thematic vowel and the ending —but not always: Clackson directly reconstructs abl.sg. *-ōd, dat.sg. *-ōi (2007:97)—.
Indeed, the presence of the thematic vowel, the lack of ablaut/accent shifts and the different endings has lead scholars to set the thematic nouns apart from the athematic ones. The thematic inflection is different, but why?
Already Meillet (1931:196) concluded that the thematic type (nominal and verbal) was more recent that the athematic. Since then, and particularly since Specht’s Der Ursprung der indogermanischen Deklination, the communis opinio has agreed to consider that the thematic nouns are of a younger age.8 A great number of scholars have
weighed in in the matter about the function of the thematic vowel. In the 1940s and the 1950s it was thought to have a pronominal origin; as for its function, Knobloch connected
8 What scholars have considered to be “recent” varies. For example, Haudry thought the thematic stems to
be more recent than the athematic ones, but still of considerable old age, since according to him the athematic endings, after being added to the thematic stems, were still understood as post-positions (1982:71-2). His theory however must be rejected for several reasons: it is impossible to prove that the thematic vowel was a post-posed article, such formation has not survived in any of the daughter languages. The relative chronology that he offers is very unclear, as is the motivation for the creation of a new inflection..
6
it with the ergative theory and considered it to be an object marker (1953:419-20). But the origin of the inflection itself remained a mystery.
Sturtevant was one of the first scholars to advance a hypothesis regarding the genesis of the thematic stems. Continuing Thurneysen’s idea that Latin reus ‘party to a lawsuit’ came from an old genitive *rēius (Skt. rāyās), connected to rēs ‘thing, legal case’, so reus would have come to being by a process of hypostasis: from *vir reus ‘man of the case’, vir would at some point be omitted by force of use and reus would have been reinterpreted as the nom.sg. of a new adjective, afterwards substantivized. Sturtevant saw that similar constructions were present in Hittite: a ‘vassal’ is literally linkii̯aš ‘(man) of the oath’ (1940:574); kūruraš ‘enemy’, lit. ‘man of enmity’ < kūrur- ‘enmity, hostility’ (Sturtevant 1940:575; EDHIL 496-7). All of this led him to conclude that “in Indo-Hittite [...] adnominal genitives in -os might give rise to o-stem adjectives.” (1940:578) Following Sturtevant among others, Villar fully adhered to the “hypostasis theory” (1974:99-122; 1995). As a whole, it is syntactically and semantically attractive, but it falls short when it comes to analyzing the relationship between the thematic stems and the other inflectional types as well as the development of the rest of the inflection. Brosman defended the hypostasis theory as opposed to9 another hypothesis that arose with Beekes’ 1985 book (which apparently Villar did not know) and that I will call the “hysterodynamic theory”.
Beekes sought to find the formation of the PIE inflection through the systematic analysis of the ablaut distribution in the languages. He reduced the main accent-ablaut paradigms of a very early stage of the proto-language to two: one static, one mobile. PIE would have been at this stage an ergative language:10 the mobile inflection had an absolutive case *CéC-R and an ergative *CC-R-ós (1985:202). In a second stage, out of the mobile paradigm, the proterodynamic (that of the neuters) and hysterodynamic (that of the animates) types would arise. At this point the language slowly shifted from
9 With unconvincing arguments: he dismisses Beekes’ theory solely on the grounds that “it would require
that the origin of the o-stems be relegated to an improbably early date.” (1998:66).
10 There are however many cautious voices in regard to the ergative theory (an overview of which can be
found in Bavant 2008, with references), for example Clackson and Matasović. Note however how Clackson’s rejection (2007:176-180) of the ergative theory starts from his analysis of the thematic stems, which, according to Beekes, arose only after (and, in part, because) the language shifted to nominative-accusative marking. Matasović (2017:164) has rightly argued that it is difficult to determine with certainty the alignment of the earliest stages of the language. While it is clear that Classical IE was a nominative-accusative language, the arguments for an ergative stage remain attractive, particularly in the light of the pre-history of the thematic stems.
7
ergative to nominative marking: in the HD inflection, the absolutive *CéC-R became the nominative (and the nominative-accusative of the neuters, i.e. the PD inflection), whereas the genitive-ablative *CC-R-ós (a case with agentive semantics) was interpreted as an ergative (1985:173-4). At the same time, the ergative *-s also became the nominative of the static inflection (1985:176) and of the newly formed animate PD nouns. This strong association between the mark *-s and agentive semantics, and now the nominative, made possible the reanalysis of old ergatives ending in *-os as new nominatives in *-o-s, with *-s as the nominative marker. On the basis of the new nominative in *-os, an accusative in *-om was created, while the genitive-ablative ending of the HD inflection was at this first stage maintained. The thematic stems, then, according to this theory, developed from the HD genitive-ablative (< ergative).
As for the neuter nouns, it could be envisaged that in a first stage all the newly created thematic stems were non-neuter and that the neuters were created in a second stage, taking their shape (in the nom.-acc., the only case where they differ from the common gender thematic stems) from the common gender accusative.11
This hypothesis answers the question of why the thematic stems developed from the HD genitive-ablative and not from the PD one: because the ergative case is strongly associated with animate nouns (i.e. the HD inflection), whereas the PD inflection developed animate nouns, with an ergative (> nominative) in *CéC-R-s only secondarily (Beekes 1985:176). Therefore, only *-os could be the model for the new inflection. As a whole, the hysterodynamic theory offers a cogent relative chronology for the development, as well as a sound model for the genesis of the inflection itself.
However, the relative chronology could also be seen as a disadvantage: if the thematic stems developed from the HD inflection right after the ergative > nominative marking shift in the proto-language (which must have happened at a reasonably early stage), how can they be considered recent? A pivotal point that will be explored in this thesis is whether the direct and the oblique cases are of the same age or not, which will help draw a clearer scenario of the development of the thematic stems, taking the Anatolian split as a terminus ante quem. Kortlandt’s claim that “the Proto-Indo-European
11 The neuters would be modelled after the common gender accusative because they would overwhelmingly
share the semantic role of patient (the common gender accusatives due to syntactic reasons, the neuters because the aforementioned connection between animacy and grammatical gender). For this general topic see Luraghi 2011.
8
thematic paradigm was probably uninflected except for the accusative in *-om [...]” (2010:40) will be put in perspective with the conclusions drawn from the Anatolian languages and the reconstruction of Proto-Anatolian.
What is, exactly, the relevance of the Anatolian material? In one word, its antiquity. I have already pointed out that the nominative-genitive homomorphy in Hittite is agreed to be an archaism rather than an innovation12. The other Indo-European languages have thematic genitive endings (Sanskrit -asya, Greek -οιο and -ου/-ω, Gothic -is) that differ not only from the nominative but also from the athematic genitive endings. This is interpreted as an innovation in all of the Indo-European languages excluding Anatolian, which is to say a common innovation.13 The analysis of the points where Anatolian differs from the rest of Indo-European (in this case, the thematic genitive) can be useful in determining the relative chronology of the grammatical change in both Anatolian and Indo-European in general. If the thematic inflection in Anatolian is more archaic than in the other IE languages, its reconstruction can shed light over the genesis and evolution of this thematic class in older stages of the proto-language.
Just a few years after the decipherment of Hittite, it was already suggested that Hittite was a sister, not a daughter, of Indo-European (Jasanoff 2017:233), a theory that is now more strongly supported by the number of innovations that are shared by the other Indo-European languages, but not by Anatolian.14 This was the kernel of the “Indo-Hittite hypothesis” (cf. Sturtevant 1962) and later renamed as the Indo-Anatolian hypothesis, once the other members of the Anatolian family became better known. It essentially posits that Proto-Anatolian was the first branch to separate from the parent language, and that this happened in a very early date. This event is referred to as the “Anatolian split”.15 The language common to Anatolian and the rest of Indo-European, including Tocharian, is Proto-Indo-Anatolian, although the term Proto-Indo-European (PIE) is widely used as 12 Beekes 2011:213: “the creation of an ending that is identical to another, here the nominative, is
improbable.” We do not expect to find an innovation that would create an ambiguity between the nominative and the genitive; both from a logical and from a formal point of view this homomorphy must be old.
13 The principle of common innovations(for which see François 2014:164 and ff.) allows comparatists to
establish the different daughter branches as well as their interrelation; if a group of languages shows the same innovation, that means they share a node, i.e. they have a common ancestor where that innovation first took place.
14 Cf. Kloekhorst&Pronk 2019:3-5 for a summary of those innovations.
15 The Indo-Anatolian hypothesis is further supported by archaeology: David Anthony has identified the
speakers of pre-Anatolian with the Suvorovo culture, that migrated from the steppes towards the lower Danube and then the Balkans around 4200 BC; later, perhaps around 3000 BC, they entered Anatolia. (Anthony 2007:100, 229, 251-260).
9
well. I will use “PIE” to refer to the common ancestor of Anatolian and the other Indo-European languages (also called “Proto-Indo-Anatolian”), “Classical Indo-Indo-European” to refer to the stage of Indo-European after the Anatolian split and prior to the Tocharian split (also called “Proto-Tocharian”) and “Core European” to refer to the Indo-European languages excluding Tocharian.16
Therefore, the reconstruction of Proto-Anatolian not only helps us to better understand the evolution of the Anatolian languages, but it is also a key element in the reconstruction of the parent language. Since the thematic stems are agreed to be a somewhat recent development, the reconstruction of the class for PA through the analysis of the relevant Anatolian material becomes of great importance, as it may allow us to better locate the genesis and evolution of the thematic inflection alongside the different stages of the proto-language.
The goal of this work can then be summed up in a seemingly simple question: what did the thematic nominal stems look like in Proto-Anatolian? In order to try and solve that question, I will first analyze the material from the major Anatolian languages, with Hittite as the main working language.
16 Kloekhorst&Pronk 2019:7. See for example Lundquist&Yates 2017:2080-1 for a slightly different
10 3. Materials, methodology and a note on Luwic
The data that will be treated in this thesis will mainly come from the consultation of reference works and specialized bibliography, particularly the EDHIL (which will be the main dictionary, since it is the most up-to-date), the Chicago Hittite Dictionary and Jaan Puhvel’s Hittite Etymological Dictionary. The Hittite grammar of reference will be Hoffner&Melchert 2008. For Luwic, the main reference works will be Starke 1990, Melchert 2003 and Payne 2010 for Luwian and Kloekhorst 2013 for Lycian. For Palaic, Carruba 1970 is still the most comprehensive reference. As for the reconstruction of PIE a number of handbooks and articles will be consulted, with emphasis on the more recent scholarship.
Regarding the Anatolian material itself, it should be kept in mind that the weight that is going to be given to each of the languages is not equal. Hittite is going to be the main working language for several reasons: the volume of attestations and secondary literature is the highest; it is also the oldest language in the family (that is, the earliest attested). In addition to this, the Luwic languages show a phenomenon called i-mutation, which consists in the presence of an -i- in the direct cases of the common gender nouns, across inflections. The consonant stems add the -i- e.g. CLuw. nom.sg. comm. ādduwališ from ādduwal- ‘bad’, while in the thematic stems we see the substitution of the thematic vowel by this -i-: CLuw. ānniš ~ Hitt. annaš < PA *Hónno- ‘mother’ (EDHIL 174). Starke interpreted this -i- as a gender motion suffix, but the meaning of the suffix cannot be proven; according to Rieken it stems from a PIE derivational process of forming abstract i-stem nouns from thematic adjectives and vice-versa. The reason why the “i-mutated” forms show up in the common gender only would be due to differences in ablaut. However, Norbruis (fthc.) has shown that her analysis must be rejected.17 Instead, he has proposed a systematic takeover of the consonant and thematic stems by the i-stems, not due to morphological innovations but to phonetic developments: first the oblique cases of the i- and the consonant stems became identical, then the consonant stems adopted the i-stem direct case endings, and, finally, the massive number of new i-stems took over the much more reduced corpus of thematic nouns. In practice, this means that
17 As there were no thematic abstract nouns in PIE or Anatolian, the Hittite i-adjectives are clearly inherited
11
the Luwic material will be, comparatively, of little importance, since the etymologically thematic nouns (e.g. ‘mother’) inflect like synchronic i-stems. Whenever relevant, adequate attention will be given to the minor Anatolian languages, particularly Palaic.
The discussion will be organized as follows: I will present the reader with the thematic nominal paradigm of Hittite as given in the handbooks. The endings will be discussed case per case: what is their situation in Hittite, how can they be reconstructed for Proto-Anatolian and whether they should be projected back to PIE. At the end, a reconstruction of the thematic nouns in PA will be suggested. In the Conclusion I will go back to the pivotal question of this work –what did the thematic stems look like right before and after the Anatolian split?–, with the intent to offer a satisfactory answer. Finally, there will be an excursus on the accent of a small number of irregular-looking Hittite words that are traditionally reconstructed as thematic, with the objective of presenting a more comprehensive picture of the situation of the thematic nouns in PA.
12
4. The thematic stems in the synchronic Anatolian languages
Here I present an overview of the regular declension of the thematic nouns in Old Hittite. The examples have been taken from Hoffner’s and Melchert’s grammar (2008: 79-81) and the EDHIL (788).
lāla- ‘tongue; aruna- ‘sea’; atta- ‘father’, negna- ‘brother’ (c.)
sg. pl.
nom. lālaš lāleš
voc. attaš, atti -
acc. lālan lāluš
gen. lālaš arunaš, *negnan18
dat.-loc. lāli *arunaš
abl. lālaz(a) instr. lālit all. aruna - pēda- ‘place’ / šuppala- (n.) ‘cattle’ sg. pl.
nom.-acc. pēdan pēda
erg. šuppalanza *šuppalanteš
gen.
the same as the common gender dat.-loc.
abl. instr. all.
I follow Hoffner’s and Melchert’s “basic scheme” of the nominal endings in Hittite (2008:69), although, as we will see, the isolation and even more so the reconstruction of the endings is not always as clear-cut as the handbooks show.
13
4.1. The inflection of the common gender thematic nouns
4.1.1. The singular cases
Nom. sg. comm.: in Hittite the ending is -aš for the thematic nouns, e.g. alpaš from alpa- ‘cloud’, ḫimmaš from ḫimma- ‘imitation, replica’, lālaš from lāla- ‘tongue’ and so on (EDHIL 169, 343, 515).19 It can be analyzed as the thematic vowel -a- plus -s. The ending -s is the same in the common gender athematic nouns, e.g. kardimiyaz20 ‘anger’ from a stem kardimiyatt- (EDHIL 456). As for the other Anatolian languages, mainly Pal. -aš (Carruba 1970:43) and Lycian -e21 should be mentioned. Lycian shows a nom.sg. ending -e for the thematic nouns; the -e can only reflect PA *-o; all final -s were lost in Lycian (Kloekhorst 2013:142), so from Hittite, Palaic and Lycian, a PA proto-form *-os can be established. In the other Indo-European languages we have *-os: Greek -ος (λύκος), Latin -us (lupus), Sanskrit -as (vṛ́kas), Lithuanian -as (vil̃kas) ‘wolf’. The PA ending *-os can be clearly traced back to PIE *-os, analyzable as *-o-s.
Voc. sg. comm.: in Hittite we most commonly find the nom. sg. form (the most common procedure)22: ḫāššās, ‘O hearth!’ (EDHIL 377); the bare stem with no endings added: ṷappu ‘O riverbank!’ (ibid.), and an ending -i (and only marginally -e): LUGAL-u-i (EDHIL 176). In the case of the a-stems, we do not find the nom. sg., but we have the bare stem in examples like atta ‘O father!’ (Hoffner&Melchert 2008: 79), negna ‘O brother!’ (CHD 428), as well as a few forms in -i: atti (CHD 192), [ŠE]Š-ni(-mi) ‘O my brother!’ (CHD 429), perhaps lāli ‘O tongue!’ (Hoffner&Melchert 2008: 75, 81) and the (neuter) form pedanti ‘O place!’ (Hoffner&Melchert 2008:75)23. This i- ending spread to some athematic stems, like LUGAL-u-i (probably something like /Hasui/); the instances of e-spelling (e.g. LUGAL-u-e, dUTU-e) are specific to one scribe (EDHIL 377). In CLuw. we see “a mere handful of examples” of vocatives for which the bare stem is used (Melchert 2003:186), but generally the nominative is used. The other Anatolian languages
19 For the apparent ending -āš as shown by a small number of Hittite words, like išḫās ‘lord’, see the
Excursus.
20 “The cuneiform signs conventionally transcribed as containing a z represent [...] a sequence /t+s/”, among
others. (Hoffner&Melchert 2008:37).
21 Which is a very rare ending, as it only appears in the few common gender thematic stems that do not
show i-mutation; see Norbruis fthc.
22 Cf. Hoffner&Melchert 2008:244-246.
23 An i-vocative formed on the stem marked with the animatizing suffix -ant-, characteristic of the ergative,
14
do not have a separate case; as already in Hittite the nominative with vocative functions was already highly popular, it is safe to assume that the younger Anatolian languages lost the vocative as a separate case altogether.
The other Indo-European languages have an ending -e for the thematic vocative, usually unaccented: Gr. λύκε (Chantraine 1984:37), Lat. domine, Skt. déva (nom. sg. devás) (Burrow 1955: 255), Lithuanian vil ̃ke, all of which are traced back to PIE *-e24. This -e is traditionally interpreted as the e-grade of the thematic vowel (e.g. Sihler 1995: 256), but this is problematic in an inflectional type that has the lack of ablaut as one of its most remarkable features. Beekes’ alternative explanation, to consider *-e as a postposed appellative particle, is more attractive in regard to its semantics, although the formal side still poses problems.
The Hitt. ending -i can be compared to the *-e shown by the other IE languages, because there is good evidence pointing to Hitt. -i# < PIE *é# (EDHIL 377). On the other hand, Anatolian differs from PIE because the thematic nouns show the bare stem (Hitt. negna). There are two possibilities: either Anatolian inherited two mechanisms for forming the thematic vocative (the bare stem and *-e) and, later, Classical IE made *-e obligatory, or in PIE *-e was obligatory and Anatolian extended the bare-stem vocatives from the athematic nouns to the thematic type as well. Considering that we only have a handful of forms in -i, it seems like the bare stem ending in -a was the most popular one, but this says nothing about whether it is archaic or innovative. There are no unequivocal arguments for either one or the other: it seems likely that PA allowed both formations.
Acc. sg. comm.: in Hitt. the ending is -an for the thematic stems (lālan from lāla-, alpan from alpa-). It can be analyzed as the thematic vowel plus -n, as is in the other vocalic stems: ḫalkin from ḫalki- ‘grain’ (EDHIL 274). The ending of the thematic stems spread to the consonant stems, where we also find -an (e.g. kardimmiyattan ‘anger’ from kardimmiyatt-), presumably to avoid the difficult cluster that would arise with the addition of -n to the stem ending in a consonant, and that would be regularly solved in -un (EDHIL 591). Lycian shows a thematic acc.sg. in -ẽ, which can only come from *-en <
24 It might be possible to think of an avoidance of a hiatus like *-o-e- or of a form ending in *-ō, as that
could be confounded with a different case-ending. Perhaps the influence from the (short) e-grade of the vocative of the consonant stems (e.g. πάτερ) played a role in the consolidation of *-e as the vocative ending of the thematic stems. Cf. Beekes 1985:101.
15
PA *-om (Kloekhorst 2013:140); likewise Luwian has -an in the thematic stems,25 clearly analyzable as -a-n; cf. the ending of the i-stems, e.g. tadin ‘father’, analyzable as -i-n (Payne 2010:23). Palaic points in the same direction with its acc.sg. ending -an (Carruba 1970:42). From this we can securely reconstruct PA *-om.
This ending is evidently cognate with the acc. sg. comm. ending in the other IE languages: Lat. dominu-m (< *domino-m) ‘lord’, Skt. agni-m ‘fire’, Gr. ἰχθύ-ν ‘fish’, Goth. guma-n ‘man’ (Beekes 2011: 194). Hittite, like Greek and Gothic, regularly shows -n instead of -m, as Hittite did not allow a final -m (Melchert 1994: 112). For Proto-Indo-European, this ending is reconstructed as *-m (*-m˳ allophonically); in the thematic stems, the thematic vowel is clearly present: *-o-m.
Gen. sg.: Hittite shows a gen.sg. ending -aš consistently across inflectional types (e.g. ḫalkii̯aš from ḫalki- ‘grain’; paḫḫuenaš from paḫḫur ‘fire’); we only have one example of a genitive in -s: the fixed expression nekuz mēḫur ‘at night, in the evening’ (EDHIL 602). Most importantly, the thematic nouns also have -aš: lālaš, arunaš, ḫuḫḫaš. This homophony of -aš (nom.sg.) and -aš (gen.sg.), as we have seen in the Introduction, indicates that the Hittite genitive preserves a very early stage within the development of the thematic inflection.
As for the other Anatolian languages, Palaic, like Hittite, has -aš (Carruba 1970:42). HLuwian shows two gen.sg. endings across inflections: -as(a) and -asi. From these, only -as(a) is relevant towards a reconstruction of the PA situation, as Palmér (2019) has recently shown that -asi is an innovation within Hieroglyphic Luwian.26 The ending -as(a) is problematic because the hieroglyphic script does not allow us to unambiguously decide whether it was pronounced as [-as], [-asa] or [-assa]. Lycian shows a similar, but slightly clearer, situation: a -ø ending that is traced back to PA *-s (in the case of the thematic nouns, *-os) (Kloekhorst 2013:142), with the difference that this ending is comparatively rare (Adiego 2010:1) and an ending -ehe (-he in the athematic stems).27 Additionally, in the Anatolian languages other than Hittite the use of derived adjectives to express adnominal or genitival relationships was very productive. In CLuwian it completely replaced the inherited genitive (Melchert 2003:171); it was
25 We also see -an in the consonant stems, which is explained as the regular reflex of *-m˳ by Starke
(1990:35).
26 An ending that is, furthermore, restricted to 9th-10th century texts originating in a specific area (South of
the Taurus mountain range), cf. Bauer 2014:142-148, Palmér 2019.
16
likewise extremely productive in Lycian, which only shows the genitive endings in personal names (Adiego 2010:1; Kloekhorst 2013:9).
Even though Hittite does not make productive use of genitival adjectives, the morpheme -assa- is attested in a handful of adjectives and deadjectival nouns, such as iugašša ‘yearling’ < iuga- ‘yearling’ and genušša- ‘knee-joint’ < genu- ‘knee’ are some examples (EDHIL 293, 423-4). The geminate in Hittite and Luwian can only be reconciled through a PA proto-form *-osso, that is most convincingly explained as deriving from PIE *-osio28. We have thus a rich scenario in PA with a maximum of three genitival suffixes: *-os, *-oso and *-osso.
In the other Indo-European languages we find a wide arrange of genitive marks: Latin shows as many as five gen.sg. morphemes: -ī and -osio in the thematic nouns, -is (< *-es), -os and -s in the athematic stems (Baldi 2002:313-4). In Greek we find, in the thematic nouns, -ου/-ω (depending on the dialect) from *-oso, -οιο, from *-osyo (like Skt. -asya);29 and –(ε)ς and -ος in the athematic stems. Mycenaean might have preserved a
thematic ending in -ος, just like Hittite (Bernabé&Luján 2006:144-6).
The details of the relationship of the gen. endings in Classical IE and what should be reconstructed for PA are not crystal clear. Palmér has argued that PA had “two genitive endings *-os and *-osio, as well as an adjectival suffix *-osio-” (2019). From a formal perspective, it works: *-os would yield Hitt. and Pal. -as (and, potentially, Luw. -as), *-osio would yield Luw. -assa and Lyc. -ehe, and *-osio- would yield Hitt. -assa-, Luw. -assa/i- and Lyc. -ehe/i-. From a functional perspective, however, I find it difficult to accept that PA inherited a fully productive inflectional morpheme, the ending *-osio, and, at the same time, also made use of it as a derivational morpheme. The idea that PA inherited two gen.sg. endings, *-os and *-oso, as well as the morpheme *-osio-, that would be specialized in deriving adjectives, does not seem excessive, particularly considering it would be a situation similar to that of Greek, where we find Myc. -os, and then -ου/-ω < *-oso as well as -οιο < *-osyo (García Ramón 2015:659). Formally it is perfectly possible to derive Lyc. -ehe from PA *-oso, and [-asa] was one of the theoretical readings of the HLuw. ambiguous ending.
28 But not *-osso-, contra Yakubovich 2008; see Melchert 2012:281.
29 There are two main theories in regard to this: either -ου/-ω (contractions of *-oo) and -οιο come from
*-osyo or they have different origins, in which case *-ου/-ω would come from *-eso/-oso, of pronominal origin.
17
I thus follow Melchert’s conclusion when he indicates that “the Anatolian languages as a whole preserve [...] the endings *-os [...] *-oso and *-osyo” (2012:284). PA, like Greek, inherited the very archaic gen.sg. ending *-os (which is, as stated in the Introduction, the probable origin of the thematic inflection in PIE as a whole), *-oso, of pronominal origin, and *-osio (see e.g. Lundquist&Yates 2015:2086-7). Then, *-os and *-oso were productive as genitive endings, while *-osio became an adjectival morpheme. In principle, one could argue that the Hitt., Pal. and potentially Luw. thematic ending -as is actually innovative, an extension from the athematic to the thematic paradigm (that would have taken place either before the Anatolian split or already in PA). This is methodologically unfortunate because linguistic innovations happen to avoid ambiguity, not to create it, which is why irregular forms are as a rule older than their regular counterparts: there would be no motivation for the speakers to replace a clear marker, such as *-osio or *-oso, with an ending that would create an ambiguity with the nominative. An attempt of explaining the thematic genitives in *-os as Anatolian innovations would then be required to find a motivation for the replacement, as well as a reason why the Anatolian situation seems to have been preserved also in Greek. In my opinion it is much preferable to posit that in PIE the thematic inflection had originally a gen. sg. *-os, preserved in Anatolian and in Greek; in Late PIE new and presumably more clear markers, *-oso and *-osio, began to appear, both of which were inherited in PA. In PA, though *-os was maintained, *-oso- gained popularity as a genitive ending, whereas *-osio was exclusively used as a derivational morpheme.
Dative-locative sg.: the ending in Hitt. is usually -i: atti from atta- ‘father’, pēdi from pēda- ‘place’. However, we have attested a handful of examples of a-stems with a dat.-loc.sg. in -ai: labarnai, tabarnai from l/tabarna-, a title of the Hittite kings (originally a personal name) (EDHIL 520), pittulii̯ai from pittulii̯a- ‘anguish, worry’ (EDHIL 680), plus a number of forms from other stems: ašaunai from ašāu̯ar, / ašaun- ‘sheepfold, pen’, a -r/-n-stem (EDHIL 220-1); once [ḫa-]aš-ša-a-i from ḫāššā- ‘hearth’ (EDHIL 322), taknai from tēkan ‘earth’, a n-stem (EDHIL 858), and u̯appuu̯ai from u̯appu- ‘riverbank’, an u-stem (EDHIL 958).
This poses, in itself, a problem of reconstruction. The ai-datives are scarcely attested and found in Old Hittite (EDHIL 376), which would suggest that they are archaisms; however all the attestations of the -ai forms are in MS/NS, not in OS, meaning
18
that they could easily be modern interferences rather than an archaic phenomenon.30 Indeed, the ai-datives of the consonant stems seem to be secondary, product of the influence of i-stem adjectives, where the ending -ai is regular (Hoffner&Melchert 2008:74). This led Neu to consider that all the ai-datives are secondary innovations (1979: 188-9).
From a historical perspective, it is relevant that dative and locative are one single case in Hittite, Luwian and Lydian, both in the singular and in the plural. Luwian regularly shows -i as a dative-locative ending in the thematic nouns, e.g. CLuw. parni from parna- ‘house’ (CLL 170) Occasionally we find -a, interpreted by Starke as an old allative (1990:36), e.g. CLuw. līla from līla- ‘pacification, conciliation’ (CLL 127). Lycian has a dat.-loc.sg. in -i, but it also shows -je in personal names, of unclear origin (Kloekhorst 2013:140). One could perhaps think of an analogical development to the genitive, -h : -he, so -i : -je, with an automatic glide. Another possibility is to take together the -a in Luwian and the -je ending in Lycian, and postulate a Luwic ending in *-o, originally the same as the allative ending (for which see below), that acquired locatival and then dative semantics, a shift that is not typologically uncommon. Palaic has a dative in -i (-ai in the ā-stems, in principle easily analyzable as -ā-i) and a locative in -a (Carruba 1970:42); this -a has been traced back to the allative ending (Melchert 2008b:43). The Carian material is extremely meager, but see Adiego 2007:318 for the forms that have been tentatively identified.
A clear point that can be established is that PA had a dative-locative case in *-i across inflections, and Luwic and perhaps also Palaic innovated another dat.-loc. ending (in the case of Luwic) and a locative (in the case of Palaic) from the PA allative in *-o. Since the ai-datives are secondary and the thematic dat.-loc. ends in -i, it seems clear that the dative did not show the thematic vowel.
The other IE languages paint a different picture: Greek has a thematic dative ending in -ωι (Sihler 1995:258). The few attestations of an Old Latin thematic dat.sg. in -ōi (Classical -ō) and especially Oscan -úi point to a Proto-Italic reconstruction in *-ōi (Baldi 2002:314). The Sanskrit -āya is explained as a hypercharacterization of the expected -āi plus a particle -ā (still independent in Avestan) or perhaps as a result of
30 Except for [ḫa-]aš-ša-a-i which is attested in OS. However, since it is an ā-stem, this is actually the
19
analogical levelling from the genitive -asya (Gotō 2017:346); the Indo-Iranian ending would thus be *-āi (Burrow 1955:256). To account for the long vowel shown in the languages, these forms are explained as coming from PIE *-ōi < *-o-ei, i.e. a contraction of the thematic vowel and the athematic ending *-ei, cf. e.g. Beekes 2011: 212, Meier-Brügger 2003: 198, Clackson 2007: 98. The locative is a distinct paradigmatic case in some IE languages, notably Sanskrit, and is thus sometimes projected to PIE as a case ending in *-i (the same element as in the dative). It is more likely that in origin, the locative was restricted to neuter nouns, while the dative (which requires an animate referent) was restricted to non-neuters, cf. Kloekhorst 2018:189-192.
In the case of PA, however, it is important to consider that the i-dat.-loc. forms (pēdi, ḫāšši) go back to PIE *-i, while the forms in -ī (išḫī, kišrī outside of the thematic stems) can only go back to PIE *-éi31 (Kloekhorst 2018: 190). Therefore, -i (as in atti) cannot be the regular outcome of PIE *-ōi.
From this, two possibilities arise. One is to consider that the thematic vowel was originally present and that PA had a thematic dat.-loc.sg. in *-oi that got replaced by the athematic ending in *-i. The other is to consider that the thematic vowel was simply never there; from a chronological perspective this would imply that the Classical IE ending *-ōi developed after the Anatolian split. The first possibility is favored by Hoffner and Melchert, who state that “in the case of a-stems, the stem-vowel is usually deleted before the d.-l. ending -i” (2008:74). The problem is that this “deletion” is merely a synchronic description, not a diachronic explanation of the development of the ending. Why would the thematic vowel be deleted if it was such a clear marker of a highly productive class of nouns?
Anatolian simply does not offer any support to a PIE reconstruction of the thematic dative in *-ōi; if it had inherited it, there would be no motivation for its disappearance. The scarcity of attestations of the secondary -ai ending further shows that it was not felt as a clear dat.-loc. marker by the speakers, or else it would have become widespread.
I thus believe that the PA thematic dat.-loc. ending -i is original and it continues the PIE situation. After the Anatolian split, the thematic vowel was extended to the
31 In this phonetic context, *-éi- was raised to *-ii- and then -ī- in open syllable (-i- in closed syllable).
20
oblique cases, including the dative: *-o-ei > *-ōi, whence Greek -ωι, Latin -ōi and Indo-Iranian *-āi.
Instr.: in Hitt., the ending is generally -it, across inflectional types: lālit in the thematic stems, but also ērḫuit in the i-stems (from ērḫui- ‘basket’) and išḫimanit in the athematic stems (from išḫiman- ‘cord, rope’) (Hoffner&Melchert 2008: 90, 112). This is already an indicator that the thematic instrumental does not show the thematic vowel, as we would expect pre-Hitt. *-o-it32 to yield Hitt. **-ait, not *-it.33 A later development of -ai- into -i- would be completely ad hoc, as the regular outcome of pre-Hittite *ai is [ai] or [āi] in certain contexts, but not [i].34
Alongside the more common ending -it, we have a good number of attestations of an instrumental ending -t: genut, ganut from genu-/ganu-35 ‘knee’ (EDHIL 467), kiššarat
and kiššarta (-a is a dead vowel) from keššar ‘hand’ (EDHIL 471-2). It is also the only instrumental ending in the demonstrative pronouns: kēt and later kēdanta (where -a is a dead vowel), with no trace of -i- (Hoffner&Melchert 2008:143). Often both -it and -t are attested within the same word: uddanit, spelled ni-it, and uddant, spelled ud-da-an-ta, are both attested as the instr. of uttar-/uddan- ‘word, speech’ (EDHIL 932). We have an attestation of a thematic noun ending in -at in šakuu̯at (MH/MS) (Hoffner&Melchert 2008:77); being only one MS attestation, it could easily be secondary.
Neu (1979:190) explained the ending -it as having arisen by anaptyxis: e.g. ektet, from ekt- ‘hunting net’< *ekt-t. (Hoffner&Melchert 2008: 77). However, Kloekhorst has rightly argued that the vowel in the ending -it must be original; that is, it comes from a full grade, because an anaptyctic vowel in this context is unexpected; the regular solution of a cluster such as *-t-t would have been -z. (2014:103-5) Since the distribution of the -et / -it spellings matches with the expected accentual distribution, I favor this view over Neu’s, which relied on a phonetic development and a subsequent paradigmatic levelling. Therefore Hittite shows the ending in the zero grade, -t, and in the full grade, both accented (with an outcome -et) and unaccented, -it, with regular /e/ > /i/ weakening (Kloekhorst 2014:105).
32 That is the thematic vowel and the ending -it, the regular outcome of unaccented *-et in this position. 33 Cf. EDHIL 100.
34 Cf. Kloekhorst 2014: 395.
21
Another remarkable feature of the instrumental case (together with the ablative) is that it is number-indifferent, which means that a single ending is used with singular and plural referents. It is also relevant to keep in mind that the instrumental was a separate case only in Old Hittite; in New Hittite it has been fossilized in isolated expressions, outside of which the instrumental “was replaced by the ablative” (Hoffner&Melchert 2008: 70). This points to a relatively low productivity of the paradigmatic case within the history of the language. Finally, also relevant is the fact that in the pronominal declension the instrumental is not always present: the personal pronouns do not have it, while the possessive pronouns have an ablative-instrumental case ending in -it (Hoffner&Melchert 2008:139). As I am going to show, it is likely that the pronouns have preserved the oldest value of -t prior to the creation of a separate ablative.
Regarding the other Anatolian languages, Luwian has a number-indifferent ablative-instrumental case ending in -adi across inflections (Melchert 2003:186), clearly cognate with Lycian -(e)di (Kloekhorst 2013:140). These are the only Anatolian languages that have an ablative-instrumental (or indeed either one of them); the comparison between Luwian and Lycian forces a Proto-Luwic reconstruction in *-odi, that will be put in relation with the Hittite ablative (for which see below).
In most handbooks,36 the ending of the instr. sg. in PIE is reconstructed as *-h1 in
different ablaut grades, whence Sanskrit -ā (ápasā < ápas- ‘work’), Lithuanian -ù, e.g. vilkù from vil ̃kas, a thematic noun (Beekes 2011:212); Mycenaean -e37, and perhaps also the Latin ablative ending -e.38 Tocharian B might show traces of an old instrumental (of an n-stem) in the ending of the causal, -ñ < *-niT (Adams 2013:281), which would match nicely the Anatolian situation. The Kortlandt Effect (the dissimilation of *d [ʔd] > *h1 [ʔ]
in certain phonetic contexts) might allow us to reconcile the two endings in a single PIE proto-form, but more research is needed.39 The material from Old Hittite clearly indicates that we should reconstruct an instrumental ending in *-t for PA. Since the thematic nouns
36 Cf. for example Beekes 2011: 187-8, Meier-Brügger 2003: 198, Clackson 2007: 93 ff, Fortson 2010:116. 37 See Bernabé&Luján 2006: 151 for the difficulty of interpretation of the forms ending in -e.
38 Although the traditional explanation interprets -e (the abl.sg. ending of the consonant and i-stems) as
coming from the PIE locative in *-i, Sihler convincingly argues that a functional shift locative > ablative is unlikely (1995: 285); ablative-instrumental syncretism, on the other hand, is typologically common (Narrog 2014: 75). From that perspective, the Latin abl. in -e could also be explained as coming from *-h1. 39 The *h
1 ending would derive from *-t (via an intermediate stage *-d) For the Effect, see Kortlandt 1983
and especially Lubotsky 2013 and Garnier 2014. The relative chronology of the phenomenon is very unclear; there are isolated examples of it in Anatolian (Lubotsky 2013:162-3), but in this case it seems that the effect was not pervasive enough so as to affect the PA endings, as we have no trace of instr. *-(e)h1 or
22
overwhelmingly show the unaccented full grade ending, -it, I reconstruct it as such for PA: *-et.
Abl.: the oldest ending we have attested is -z, in certain consonant stems: ne-e-pí-iš-za [nēpists]. In the a-stems we only find -az (arunaz from aruna- ‘sea’). Already in OH an ending -az appears in the consonant stems (including those that previously had -z), with an accented variant -āz. (EDHIL 232). From a formal point of view, Hitt. -z can stem from a sequence /t/ plus /s/ (*-ent-s > -anza [-ants]) (EDHIL 26) and the result of the assibilation of PA *-ti-, i.e. ēšzi ‘he is’ < PIE *h1ésti.
If we compare it with the other Anatolian languages, Luwian has an ablative-instrumental ending in -adi, (Melchert 2003: 186) and -ādi in the thematic stems (Starke 1990:41); likewise, Lycian -(e)di (in the thematic and consonant stems; -adi in the astems) (Kloekhorst 2013:1412). Luw. ā̆di and Lyc. edi can only come from PA *óti, -d- being the regular result of the Proto-Anatolian lenition.40
From this, the conclusion is that the oldest ending, preserved in e.g. ne-e-pí-iš-za, was *-ti, at least for PA, and that it coexisted with *-oti (Hitt. -az) and *-óti (Hitt. -āz). PA thus clearly differs from the other IE languages, where we see a thematic ablative ending in *-ōt < *-o-et (Meier-Brügger 2000:198), with the same element *-et found in the Anatolian instrumental (Kloekhorst 2018:194).
The etymology of the suffix becomes clearer if one takes the instrumental *-(e)t into consideration. As it has been stated before, the functions of the instrumental and the ablative cases are close and it is not typologically problematic to have a single case for both, or ablative-instrumental syncretism. Based on this, and on the idea that the element *-t- is in both endings, I consider a reconstruction of abl. *-ti as the instrumental *-t plus *-i, following Kloekhorst (EDHIL 232) and Beekes (2011:187-8).41 The ending -az <
40 Cf. Kloekhorst 2014:356; 2016:223 and ff. Both PA *-óti and *-oti are contexts where lenition would
occur: after a long accented vowel (PIE *ó > PA *ṓ) and between two unaccented vowels in the post-tonic position. The non-lenited result in Hittite (showing the reflex of PA *-t- rather than the lenited *-d-) is due to analogy to the forms with the zero-grade of the ending (-ti) (EDHIL 232).
41 The consequences of this reconstruction regarding the prehistory of the ablative and the instrumental are
several. First of all, in a more general fashion, it solves the problem of having to establish different case systems for the athematic and the thematic stems in PIE, i.e. a genitive-ablative *-(o)s in the athematic nouns (Beekes 2011: 186), but a genitive in *-os and an ablative in *-et in the thematic stems, plus an instrumental in *-(e)h1. It is possible to consider a single case in *-(e)t, with original ablatival-instrumental
semantics, that is maintained in Hitt. –(i)t and e.g. Old Latin donōd. By means of the Kortlandt Effect, in certain positions *-(e)t became (via *-(e)d) *-(e)h1, the instr. ending in e.g. Sanskrit ápasā and perhaps
23
*-oti (and the accented variant) would have arisen from the o-grade of the instrumental ending plus the same i. Oettinger reconstructs two different ablative endings in PIE, *-d an*-d *-ti, both of which woul*-d have survive*-d in Anatolian (Hitt. -az an*-d -(i)*-d respectively) (Oettinger 2017:259, also Melchert&Oettinger 2009:57 and ff.). I think this reconstruction is uneconomical and unmotivated both semantically and functionally, considering that the alternative explanation manages to obtain both endings from just one, without needing to assume not one but two functional shifts.
Concerning the thematic stems, the reason why I reconstruct a PA ending *-oti (so with a segmentation arun-az) rather than *-o-ti (aruna-z) is because it would be odd for the thematic vowel to appear in the ablative, but not in the other oblique cases. If one accepts that the ablative ending is an off-shoot of the instrumental, where the thematic stems show a full-grade ending (-it), like the consonant stems (and not the expected **-a-t), it makes more sense for the ablative to behave in the same way. One would have to explain why the thematic vowel appears, of all the oblique cases reviewed so far, in the ablative alone. At any rate, even though methodologically this seems to be the sounder option, the attestations of the ablative as we have them do not allow us to unequivocally tell whether the thematic vowel was present or not.
To sum up, the thematic stems in Old Hittite show an instrumental in -it and an ablative in -az, both identical to the endings of the consonant stems, and thus do not show the thematic vowel. The ablative ending in -(a)z is diachronically an off-shoot of the instrumental in -t. The other major Anatolian languages, including New Hittite, merged the two cases (the semantics of which were already close). It is then likely that we can reconstruct an original ablative-instrumental case in *-t for PIE, that in Anatolian crystalized as an instrumental in *-(e)t and an ablative in *-(o)ti.
All. Hittite knows a case that is usually called “allative”, “directive” or “terminative” (Hoffner&Melchert 2008: 76). The ending is -a (when unaccented) or -ā when accented, and its meaning is generally cited as ‘direction towards’. The case is attested only in the singular, and only in Old Hittite; in the later language, it is replaced by the dative-locative (Hoffner&Melchert 2008:70).
to explain why in Classical IE the Effect was triggered in the thematic instrumental (*-(e)h1) but not in the
24
The thematic allative ends in -a: anna from anna- ‘mother’ (EDHIL 174); aruna from aruna- ‘sea’ (EDHIL 212), as do the athematic stems: nēpiša from nēpiš- (EDHIL 603), with an accented variant -ā: taknā from tēkan- / taknaš ‘earth’. If the thematic vowel had been there, in synchronic Hittite we would find a long vowel as a result of contraction.42
The data from the other Anatolian languages is scarce, as none of them show an allative case; we have seen, however, that the Palaic locative and the Luwian -a datives might continue an original allative. The infinitive endings are taken to be old allatives as well. As such, Hitt. -ānna (infinitive II) is explained an original allative of a verbal noun in -ātar/-ānn (EDHIL 226). Luwian has an infinitive in -una, which, according to Melchert, reflects “an old dative-locative (originally allative) of such [verbal] noun” (2003:198). Palaic also shows an infinitive in -una: aḫuna ‘to drink’ (Melchert 2008b:44); Melchert also indicates, however, that “the locative ending /-a/ is cognate with the allative of Hittite continuing Proto-Indo-European ∗-h2e and ∗-oh2” (Melchert 2008b:43).
Unfortunately, his explanation is incomplete, as he does not offer a distribution for the different ablaut grades. Reconstructing an allative ending in *-h2 poses a problem of
homonymy, since that is already the marker for the nominative-accusative plural of the neuters in Anatolian as well as in the other Indo-European languages (Hoffner&Melchert 2008:71-2; Beekes 2011:188).
Both Luwian and Palaic -una- can continue PA *-un-o-,43 and there is thus no phonological necessity for a laryngeal. It seems as if the only reason to reconstruct an ending with a laryngeal is Lycian, where infinitives end in -ne and in -na, so both erijeine and erijeina are the infinitive of ‘to make’ (Kloekhorst 2013: 148).44 I believe that -na could easily be secondary, perhaps influenced by the nominal -a (< PIE *-eh2) stems,
42 E.g. tān ‘twice’ < PIE *duoióm (EDHIL 826-7), with loss of intervocalic *-i- and vowel contraction
(Melchert 1994:176-7; Kloekhorst 2014:521).
43 Lycian lacks the initial u- shown in Palaic and Luwian -una-, but this context would trigger the syncope
of -u- (Melchert 1994: 318). I am only dealing with the -a- segment of the ending, where both Luwian and Palaic -a continue PA *-o (Melchert 1994:218, 264). Luwian and Palaic point to an ending *-un-a-, whereas the etymon of Hittite -ānna is reconstructed as PIE *-ótn-o (EDHIL 174-5). However, Hittite also has the so-called “infinitive I” in -u̯anzi < *-u̯anz (with secondary -i-), most likely the (petrified) ablative ending of a deverbal noun in -u̯ar/-u̯aš (EDHIL 957-8). Therefore, the infinitive in Hittite and in Luwian and Palaic (and possibly Lycian) simply continue different fossilized cases of deverbal nouns formed with the same suffix *-u̯en-. I wonder if this is the same suffix as Skt. -van-, that forms perfect participles e.g. cakr̥vān ‘having done’ (Macdonell 1916:66).
44 This conclusion must be reached by elimination: the -u- cannot have been accented, or we would find
plene spellings in CLuwian (cf. CLL 120, Melchert 1994:262) and Palaic (Melchert 1994:215), and -no- cannot have been accented either as the result would have likewise been long in the languages (Melchert 1994:264, 218). The accent in the infinitive in PA must have been, then, in the previous syllables.
25
while -ne could continue PA *-uno-. The interpretation of the Lycian material is at any rate tentative.45
However, if I am correct, then the infinitive II in Hittite and the infinitives in Luwian, Lycian and Palaic, that derive from PA *-uno-, are all allatives of verbal nouns. A reconstruction with a laryngeal is not necessary, and it would weaken the connection with other Indo-European formations in *-o that also have directional meaning.
The same ending -a is found in adverbs and preverbs such as anda ‘into, inwards’, appa ‘behind, afterwards, further’, katta ‘downwards, alongside’, šarā ‘upwards’ (cf. šēr ‘above, on top’), parā ‘out (to), forth, towards’ (preverb) (EDHIL 185, 192-3, 463, 630, 745). In all of these, a clear sense of ‘directed movement’ (Hoffner&Melchert 2008:76) can be distinguished, which sits well with the meaning of the allative case.46
The allative is often connected47 to Gr. πρό ‘forth, forward, before, for’ (EDG
1235), which for Beekes is an “allative formation”. To explain the adv. πρωΐ ‘early, in the morning’ he reconstructs an adverb *prō, cognate with Latin prō ‘for, before’, OHG fruo ‘early’, Avestan frā ‘forward, in front’, “perhaps an old instrumental *pro-h1 [...]”
(EDG 1244). One could then think of a situation in PIE (right before the Anatolian split) along these lines: a particle *-o, that forms adverbs of direction, like *pro ‘forward’, *h2epo ‘away’, *upo ‘under’ (EDG 117, 1535) and that became productive as a case
ending in Proto-Anatolian, surviving in Old Hittite and disappearing in New Hittite and the other Anatolian languages, with significant remnants (like the Luwic dat. ending and the Pal. locative, as well as the infinitives).48 The allative fits the tendency shown so far: that the thematic nouns inflected like the athematic ones in the oblique cases, without the thematic vowel or a special set of endings.
45 The problem is that, first, a distribution of the two endings has not been found, and, second, that in
principle -na and -ne cannot go back to the same PA preform, as Lyc. -a < PA *-a (< PIE *-(e/o)h2), whereas
Lyc. -e < PA *-o (< PIE *-o). This forced Hajnal to conclude that “das Lyk. besitzt demnach mindestens zwei Infinitivbildungen.” (1995: 98). The two formations that he reconstructs are -e < PA *ō < PIE *-oH and -a < PIE *-ah2;45 one then wonders why not simply reconstruct the same ending with different ablaut
grades.
46 Melchert rejected the formal and semantic connection between the allative and these adverbs based on
semantic grounds: “the nominal case in -a expresses a goal, mostly an attained goal. However, the adverbs in -a mark merely direction.” (2017: 529). However, his claim that the allative case expresses only ‘goal’ is disputable, cf. Luraghi 1997:12. Melchert’s claim that the adverbs/preverbs in -ā̆ cannot have the same origin as the allative because the adverbs sometimes show locatival meanings is not an obstacle, as per Luraghi (2003:21): “In the Indo-European languages, there is often a fuzzy border between Location and Direction.”
47 For which see Dunkel 1994:28 and ff., with refs. 48 For this also see Norbruis fthc.
26 4.1.2. The plural cases
Nom.pl. The thematic stems show in synchronic Hittite an ending -eš: laleš, antuḫšeš ‘men’, ÍDḪI.A-eš ‘rivers’ (Hoffner&Melchert 2008:81; EDHIL 294), etc. This ending is in principle the same as the one found in the consonant stems: ḫāššeš from ḫāšš- ‘ash’ (EDHIL 318), ḫuu̯anteš from ḫuu̯ant- ‘wind’ (EDHIL 368), etc. This ending is indicated in the handbooks as -ēš (e.g. Oettinger 2017:259), in view of the plene spelling of the ending shown in the other vocalic stems: the nom.pl. of ḫalki- ‘grain’ is spelled ḫal-ki-e-eš (EDHIL 274), ar-ki-i-e-eš from arki- ‘testicle’ (EDHIL 203), ḫé-e-a-u-e-eš and ḫé-e-i̯a-u-e-š from ḫēu-, / ḫē(i̯)au̯- ‘rain’, an ablauting u-stem (EDHIL 340-1), and so on. However, it is important to take into account that the plene spelling of e after -i and -u did not indicate vowel length but it was an orthographical device to spell the glide (Kloekhorst 2012:248-51; Kloekhorst 2014:170). In both the consonant and the vocalic stems, the ending is in principle short.
A hypothetical PA nom.pl. ending *-es / *-és would yield Hitt. -iš for the unaccented variant, instead of -eš (Kloekhorst 2014:170-1). This necessarily forces a reconstruction with a long vowel, as only PA *-ēs / *-ḗs can consistently yield pre-Hitt. *-ēš, which is afterwards shortened to -eš.49 PA *-ēs has been traced back to the original nom.pl. ending of the -i stems, *-ei-es (cf. EDHIL 249-50 for the relative chronology of the development). In the later language, the ending -eš (i.e. the unaccented variant) is sometimes weakened to -i-iš /ɨs/ (EDHIL 249), which is unproblematic, as is a phonetic, not a phonological (or morphological) development. From MH onwards, we see both the original acc.pl. -uš (Hoffner&Melchert 2008:70) and the dat.-loc.pl. -aš being used as a nom.pl.comm. marker as well (EDHIL 214).
The other Anatolian languages point towards a different formation. Luwian has a nom.pl.comm. ending -nzi across inflections: huhanzi from huha- ‘grandfather’, a thematic noun, tadinzi from tadi- ‘father’, an i-stem (Melchert 2003: 186). Lycian shows two endings: -ẽi and -ãi in the thematic and a-stems, and a zero ending in the i-stems (including i-mutated nouns50). We do not have attestations of a consonant-stem noun in the nominative plural. Both sets of endings have the same origin: -ẽi comes from
pre-49 In closed syllables, PIE unaccented *-ē is shortened to /e/ in Old Hittite, whereas the long vowel is
preserved if it was also accented (again, Kloekhorst 2014: 171, also cf. Melchert 1994: 142-3).
50 The term “i-mutation” refers to a partial merger of the common gender thematic and i-stems that takes
place in the Luwic languages, for which see Starke’s already classic work (1990:59-93) but, most importantly, Norbruis fthc.
27
Lycian *-enhi, while the ending of the i-stems comes from *ĩi < *-inhi: the /n/ nasalizes the previous vowel, and the intervocalic /h < *s disappears. The Luwian and Lycian endings point to Proto-Luwic *-nsi (Melchert 1994:278; Kloekhorst 2013:141). From a PIE perspective, this ending can be analyzed as the accusative plural ending of the common gender, *-ms plus a morpheme *-i. Therefore, Luwic does not show a specific nom. pl. ending of the thematic stems.
Palaic, however, offers a much more positive outlook: it has -es for the consonant stems, that seems to continue PA *-es,51 the same ending as e.g. Gr. φύλακες, and -aš in the thematic stems. Like Hittite, Palaic merged the PA vowels /o/ and /a/; so Pal. -aš < PA *-os. According to Melchert, this ending *-os must continue PIE *-ōs (2008b:43); the shortening of unaccented long vowels in PA is regular (Melchert 1994:76). He does not elaborate any further on the origin of *-ōs, but it is reasonable to venture a contraction of *-o-es, i.e. the thematic vowel plus the nom.pl. ending of the athematic stems. If this is the case, then Palaic preserves a very old scenario: the original nominative plural endings in PA.
In the non-Anatolian Indo-European languages we see an ending *-ōs < *-o-es, preserved in Indo-Iranian and Germanic; several branches substituted it with the pronominal ending *-oi (Beekes 2011:213). The substitution had to happen in very late and in each of the individual languages, because, for example, Latin has -ī < PIE *-oi, but Oscan and Umbrian continue PIE *-ōs; likewise Old Irish maintained the old nom. ending *-ōs for the vocative plural (Sihler 1995:261).
As we have seen, the Hitt. ending must continue the PA nom.pl. ending of the i-stems, namely *-es < *-ēs (accented and unaccented) < PIE *-ei-es (Oettinger 2017:259). This ending was generalized to all the nom.pl.comm. endings regardless of their inflection, so we do not have any examples of the presumably inherited inflection prior to the generalization of *-ēs. The Luwic ending is likewise innovative. However, if we can project the archaic-looking Palaic ending to PA, then PA had a thematic nom.pl. in *-ōs < PIE *-o-es.
51 But not *-ēs, the result of which would have been Pal. **-īš or perhaps **-ēš, never -eš; cf. Melchert
28
Acc.pl.: in Hittite the acc.pl. ending is -uš across inflections; attuš from atta-, ḫalkiuš from ḫalki- (Hoffner&Melchert 2008:80, 88), with an accented variant, e.g. ku-u-uš from kāš ‘this one’ (EDHIL 425). Hitt. -ku-u-uš continues both PIE *-oms and *-ms (Melchert 1994:182, 185-6), i.e. the acc.sg. ending pluralized with *-s.52 As Kloekhorst notes, it could not continue *-(o)ns, or we would see **-aš or perhaps **-anz (EDHIL 928-9). Since the outcome of the thematic and the athematic endings is the same, it is reasonable to think that the thematic nouns continue *-o-ms, but proving it is impossible.
However, CLuw. -nz and HLuw. -nzi (the -i presumably extended from the nominative to the accusative) (Melchert 2003:186) as well as Lycian -Vs (< *-Vns) (Kloekhorst 2013:141) point towards an ending *-ns in Proto-Luwic; since we do not have secure attestations of the acc.pl. ending in Palaic or Carian, from what material we have it is reasonable to consider that PA had *-(o)ms, inherited by Hittite, while in Proto-Luwic the nasal changed its point of articulation *-(o)ms > *-(o)ns. Therefore, the evolution of *-(o)ms would be one of the Luwic common innovations that show that it was a separate dialect from Hittite.
The same development took place in Classic Indo-European, after the Anatolian split: Cretan Doric -ονς (Miller 2014:217), Sanskrit -āns53, Gothic -ans (Beekes 2011:188, 194). It seems therefore clear that the PA ending *-oms directly continues PIE *-oms.
Gen. pl.: two genitive plural endings (both with accented and unaccented variants) are attested already in Old Hittite: -an and -aš (see below), which is the dat.-loc. pl. ending and that will replace -an completely in NH (Hoffner&Melchert 2008:73). In the case of -an, it is possible to tell from the attestations that we have that it remains the same across inflections: nepišan from nepiš- (EDHIL 603-4), u̯ittan from u̯itt- ‘year’ (EDHIL 1014), patān from pat- ‘foot’ (EDHIL 653-4), and likewise in the thematic nouns, where we do not have instances of the accented variant -ān: šīunan from (d)šīuna-54 ‘god’ (EDHIL 763), URU-i̯a-an from ḫāppirii̯a- ‘town’, NIN.MEŠ-n, which must represent /neknan/, from nekna- ‘brother’ (Hoffner&Melchert 2008:73; EDHIL 601).
52 See now Melchert 2020:269-271 for an in-depth treatment of the phonetic reality of the the acc.pl. ending. 53 With analogical -ā-, cf. Burrow 1955:236.
54 Originally the stem was /sīu-/, with a nom.sg. ši-ú-uš, but from OH times on we find a thematic stem
