Professional Documents
Culture Documents
IE KloekhorstPronk-2019 (Precursors)
IE KloekhorstPronk-2019 (Precursors)
IE KloekhorstPronk-2019 (Precursors)
Editor
Alexander Lubotsky
volume 21
Edited by
Alwin Kloekhorst
Tijmen Pronk
LEIDEN | BOSTON
Typeface for the Latin, Greek, and Cyrillic scripts: “Brill”. See and download: brill.com/brill‑typeface.
ISSN 0926-5856
ISBN 978-90-04-40934-7 (hardback)
ISBN 978-90-04-40935-4 (e-book)
Preface vii
4 Indo-Anatolian Syntax? 50
Dag Haug and Andrei Sideltsev
The editors
1 Organized by Alwin Kloekhorst in the context of his research project ‘Splitting the Mother
Tongue: The Position of Anatolian in the Dispersal of the Indo-European Language Family’
(NWO project nr. 276-70-026).
Like any other natural language, the mother language of the Indo-European
language family did not originate out of nothing. It must have developed, as a
result of linguistic changes, from an earlier language, which in turn must have
developed from an even earlier language, and so on. It is therefore legitimate
to ask whether anything meaningful can be said about the nature of these pre-
cursors of Proto-Indo-European. The answer to this question naturally depends
on whether relatives from outside the Indo-European language family can be
identified and, if so, whether there are enough similarities with Proto-Indo-
European to set up hypothetical etymologies that can be used to reconstruct a
common proto-language.
* Part of the research for this article was financed by the research project ‘Splitting the Mother
Tongue: The Position of Anatolian in the Dispersal of the Indo-European Language Family’
(NWO project nr. 276-70-026).
be found in, e.g., Brugmann’s Grundriss. However, at the beginning of the 20th
century, two new Indo-European branches were discovered, Tocharian and
Anatolian, of which especially the latter had a huge impact on Indo-European
studies.
Hittite, the best known Anatolian language, famously provided conclusive
evidence in favour of what is today known as the “laryngeal theory”. In lex-
emes where de Saussure had predicted the presence of a coefficient sonantique,
Hittite turned out to have a consonantal phoneme ḫ. Anatolian thereby com-
pletely changed the reconstruction of the Proto-Indo-European phonemic sys-
tem. This, in turn, had important consequences for the reconstruction of Proto-
Indo-European morphology. Without the laryngeal theory, current views on,
e.g., the nominal ablaut-accent types, would not have existed. It has, however,
taken decades before all implications of the laryngeal theory were properly
understood and it was fully incorporated into the reconstruction of Proto-Indo-
European. Even today, its full impact is sometimes underestimated.1
1 E.g. when scholars fail to recognize that the ultimate consequence of the laryngeal theory is
that Proto-Indo-European did not possess a phoneme *a (Lubotsky 1989, Pronk 2019).
For a long time, the prevailing view was that the Indo-Anatolian hypothesis
was too radical. It was assumed that the aberrant character of Anatolian was
due to a massive loss of categories and other specific innovations within this
branch. As a consequence, no need was felt to assign a special status to the Ana-
tolian branch, or to alter the ‘classical’ reconstruction of Proto-Indo-European.
Over the last few decades, this point of view has started to shift and nowadays
the majority of scholars appear to accept the idea that the first split in the Indo-
European language family was between Anatolian and the other branches,
which at that point still formed a single language community that underwent
common innovations not shared by Anatolian.2 Nevertheless, no consensus
has yet been reached on the exact number or nature of these common non-
Anatolian innovations, nor on the amount of time that passed between the
‘Proto-Indo-Anatolian’ stage and the ‘classical Proto-Indo-European’ stage, as
one may refer to these stages now. In our view, the following examples are all
good candidates for cases in which Anatolian has retained an original linguistic
feature, whereas the other Indo-European languages have undergone a com-
mon innovation:
Semantic Innovations:
1. Hitt. participle suffix -ant-, which forms both active and passive partici-
ples, vs. cl.PIE *-e/ont-, which is only active (Oettinger 2013/14: 156–157).
2. Hitt. ḫarra-i ‘to grind, crush’ vs. cl.PIE *h₂erh₃- ‘to plough’ (Kloekhorst
2008: 9).
3. Hitt. lāḫu- ‘to pour’ vs. cl.PIE *leuh₃- ‘to wash’ (Oettinger 2013/14: 169).
4. Hitt. mer- ‘to disappear’ vs. cl.PIE *mer- ‘to die’ (Kloekhorst 2008: 8).
5. Hitt. nekutt- ‘twilight’ vs. cl.PIE *negwht- / *nogwht- ‘night’ (Melchert fthc.).
6. Hitt. šāḫ- ‘to fill up, to stuff’ vs. cl.PIE *seh₂- ‘to be satiated’ (Kloekhorst
2008: 9).
7. Hitt. šai-i ‘to impress, to prick’ < *sh₁-oi- vs. cl.PIE *seh₁- ‘to sow’ (Oettinger
2013/14: 168).
8. Hitt. ēš-zi ‘to sit’ < *h₁es- next to eš-a(ri) ‘to sit down’ < *h₁e-h₁s- vs. cl.PIE
*h₁e-h₁s-to ‘to sit’ next to innovated *sed- ‘to sit down’ (Norbruis fthc.a).
2 For recent discussions see Kloekhorst (2008: 7–11), Oettinger (2013/2014), Melchert (fthc.)
and, more sceptically, Rieken (2009), Eichner (2015) and Adiego (2016).
Morphological Innovations:
9. Anat. common/neuter vs. cl.PIE m./f./n.: innovation of the feminine gen-
der (e.g. Melchert fthc.).
10. Anat. *ti(H), *tu- vs. cl.PIE *tuH, *tu- ‘you (sg.)’: spread of obl. stem *tu- to
the nominative (Koekhorst 2008: 8–9).
11. Anat. *h₁eḱu- vs. cl.PIE *h₁eḱu-o- ‘horse’: thematization (Kloekhorst 2008:
10).
12. Anat. *iéug- (later replaced by *iéug-o-) vs. cl.PIE *iug-ó- ‘yoke’: themati-
zation (Kloekhorst 2014: 5031962).
13. Hitt. ḫuu̯ ant- < *h₂uh₁-ent- vs. cl.PIE *h₂ueh₁nt-o- ‘wind’: thematization
(Eichner 2015: 17–18).
14. Gen. *-om (number-indifferent) vs. cl.PIE gen.pl. *-om: formalization of
number distinction (Kloekhorst 2017a).
15. Anat. has no verbal suffix *-e/o- vs. cl.PIE has *-e/o- as subjunctive and
present marker: development of subjunctive *-e/o- to a present marker in
cl.PIE (and loss of the subjunctive in Anatolian) (Kloekhorst 2017b).
16. OHitt. conjunctions šu and ta vs. cl.PIE demonstrative pronoun *so/to-
(Watkins 1963).
17. The element *sm / *si in pronouns (De Vaan, this volume, 203–218).
18. Hitt. allative case -a < *-o vs. cl.PIE petrified *-o in the prepositions *pr-o
‘before’, *up-o ‘down to’ and *h₂d-o ‘to’.
Sound Changes:
19. Anat. *h₂ = *[qː] and *h₃ = *[qːw] vs. cl.PIE *h₂ = *[ħ] or *[ʕ] and *h₃ =
*[ħw] or *[ʕw]: fricativization of uvular stops (Kloekhorst 2018b).
20. Hitt. nekutt- < *negw(h)t- vs. cl.PIE *nokwt- ‘night’ and Hitt. šakuttai- <
*sogw(h)tH- vs. cl.PIE *sokwtH- ‘thigh’: voice assimilation (Eichner 2015: 15).
21. Hitt. amm- < *h₁mm- (< pre-PIA *h₁mn-) vs. cl.PIE *h₁m- ‘me’: degemina-
tion of *mm to *m (Kloekhorst 2008: 111234).
Syntactic Innovations:
22. The marking of neuter agents (Lopuhaä-Zwakenberg, this volume, 131–
150);
23. The syntax of bare interrogatives (Haug and Sideltsev, this volume, 50–
73).
There are several other arguments that are promising, though perhaps less
forceful than the ones mentioned above or requiring additional investigation
before it can be decided whether we are genuinely dealing with an innovation
of the ‘classical’ Indo-European languages:
Although it is quite possible that not each and every one of the arguments
listed above will eventually become generally accepted, it is to our mind very
unlikely that items 1–23 will all be refuted and we therefore regard the Indo-
Anatolian hypothesis as proven. Moreover, some of the arguments listed here
concern significant structural innovations, of which especially the rise of the
feminine gender (including the creation of the morphology that goes with it)
is something that cannot have happened overnight. Finally, it is important to
stress that the Indo-Anatolian hypothesis could be disproven by showing that
Anatolian shared its earliest innovations with some but not all other branches
of Indo-European. Thus far, no such counterevidence has surfaced. An attempt
to identify innovations that Anatolian shared with the western branches of
Indo-European, either at an earlier stage or after initial divergence (Puhvel
1994, Melchert 2016), has produced no evidence that would contradict the
Indo-Anatolian hypothesis.
In his treatment of the Indo-Anatolian hypothesis, Oettinger (2013/2014)
hypothesized that the time gap between Proto-Indo-Anatolian and ‘classical’
Proto-Indo-European may have been some 800 years. To our minds, this is a
conservative estimate, and we think that the gap may well have been in the
range of 1000–1200 years (depending, however, on the status of Tocharian, cf.
the next section). With the recent revolution in the genetic research on ancient
DNA, through which prehistoric migrations can be reconstructed in space and
time and therefore can be linked to the spread of archaeological cultures and
possibly of languages (cf. Haak et al. 2015, Allentoft et al. 2015, Damgaard et al.
2018, Kroonen et al. 2018), it is important to have a good idea about the time
depth of a reconstructed language. This is crucial for formulating hypotheses
about where that language may have been spoken, which in turn is important
when searching for a possible genetic relationship with other language fami-
lies.
Another important consequence of regarding the Indo-Anatolian hypothe-
sis as proven is that our view on the shape of the Indo-European proto-language
has to change, sometimes drastically. We already mentioned the topic of gen-
der: although for years it had been taken for granted that the Indo-European
mother language had three genders, it seems now inevitable that Proto-Indo-
Anatolian in fact had only two: common and neuter gender. This two-way
oposition is likely to reflect an original distinction between animate and inani-
mate gender. This is of course relevant knowledge when investigating possible
genetic ties with other languages or language families.
Another example concerns the phonetic nature of the laryngeals. The pho-
neme *h₂, which at the stage of ‘classical’ Proto-Indo-European may have been
a pharyngeal fricative, at the level of Proto-Indo-Anatolian may rather have
been a uvular fricative (Weiss 2016) or a uvular stop (Kloekhorst 2018b). Again,
this is relevant information when proposals for possible outer-Indo-European
cognate sets need to be assessed.
All this means that not only the time gap between Proto-Indo-Anatolian and
‘classical’ Proto-Indo-European may have been significant, but also that the lin-
guistic shapes of these two stages in some aspects differ dramatically. In a way,
we may therefore regard Proto-Indo-Anatolian as the first precursor of ‘classi-
cal’ Proto-Indo-European. It is for this reason that the first part of the subtitle
of this book refers to the Indo-Anatolian hypothesis.
innovations, e.g. in the word for ‘yoke’ (no. 12), the genitive plural ending *-om
(no. 14), the element *sm / *si in pronouns (no. 17), voice assimilation (no. 20)
etc. Therefore, the time gap between Proto-Indo-Anatolian and Proto-Indo-
Tocharian may have been less than the 1000–1200 years proposed above. There
are, however, still many cases for which it is clear that Tocharian did participate
in the post-Anatolian innovations (e.g. ‘you (sg.)’ (no. 10), ‘horse’ (no. 11), ‘wind’
(no. 13)), and since these include some major ones (e.g. the development of the
feminine gender (no. 9)), it remains attractive to assume that the Anatolian-
Tocharian time gap is substantial, and we would assign some 800–1000 years to
it. The relatively large number of shared Indo-Tocharian innovations contrasts
with the number of plausible post-Tocharian, ‘core’ Indo-European innova-
tions, which, according to our current knowledge, is “not overwhelming” (Pey-
rot, this volume, 186). It therefore seems unlikely that Proto-Indo-Tocharian
and ‘core’ Proto-Indo-European were separated by more than a few centuries.
4 Internal Reconstruction
5 External Comparison
probable cognates suggests that at least a couple of millennia must have passed
between the dissolution of Proto-Indo-Uralic and its daughter languages Proto-
Indo-Anatolian and Proto-Uralic (Napol’skix 1997: 143), even if we take into
account the temporal gap of up to 2500 years between Proto-Indo-Anatolian
and Proto-Uralic. Proto-Indo-Uralic would then have been spoken in or around
the 7th millennium BCE.
We can conclude that ‘core’ Proto-Indo-European, consisting of the Brug-
mannian branches of Indo-European, had a number of precursors. Its direct
ancestor was Proto-Indo-Tocharian, to which it was very similar. A more dis-
tant, but still clearly recognizable ancestor was Proto-Indo-Anatolian, which
seems to have been spoken at least a thousand years earlier. Finally, there was
a distant Proto-Indo-Uralic ancestor, with which ‘core’ Proto-Indo-European
shared only a limited number of words and a few dozen grammatical charac-
teristics and which must have been spoken at least several millennia before
Proto-Indo-Anatolian. The papers in this volume reflect the state of the art in
the research into these ancestors of ‘core’ Proto-Indo-European.
References
Adiego, I.-X. 2016. Anatolian languages and Proto-Indo-European. Veleia 33, 49–64.
Allentoft, M.E. et al. 2015. Population genomics of Bronze Age Eurasia. Nature 522, 167–
172.
Anthony, D.W.; D. Ringe 2015. The Indo-European Homeland from Linguistic and
Archaeological Perspectives. Annual Review of Linguistics 2015/1, 199–219.
and Indo-European studies in honor of Sasha Lubotsky. Ann Arbor–New York: Beech
Stave Press, 89–106.
Kloekhorst, A. 2018b. Anatolian evidence suggests that the Indo-European laryngeals
*h₂ and *h₃ were uvular stops. Indo-European Linguistics 6, 69–94.
Koivulehto, J. 1994. Indogermanisch—Uralisch: Lehnbeziehungen oder (auch) Urver-
wandschaft? In: R. Sternemann (ed.), Bopp-Symposium 1992 der Humboldt-Universi-
tät zu Berlin. Heidelberg: Winter, 133–148.
Koivulehto, J. 2001. The earliest contacts between Indo-European and Uralic speakers
in the light of lexical loans. In: C. Carpelan; A. Parpola; P. Koskikallio (eds.), Early
Contacts between Uralic and Indo-European: Linguistic and Archaeological Consider-
ations. Helsinki: Finno-Ugrian Society, 235–264.
Koivulehto, J. 2003. Frühe Kontakte zwischen Uralisch und Indogermanisch im nord-
westindogermanischen Raum. In: A. Bammesberger; T. Vennemann (eds.), Lan-
guages in Prehistoric Europe. Heidelberg: Winter, 279–316.
Kortlandt, F. 2002. The Indo-Uralic verb. In: R. Blokland; C. Hasselblatt (eds.), Finno-
Ugrians and Indo-Europeans: Linguistic and Literary Contacts. Maastricht: Shaker,
217–227.
Kortlandt, F. 2010. Studies in Germanic, Indo-European and Indo-Uralic. Amsterdam:
Rodopi.
Kortlandt, F. 2018. The expansion of the Indo-European languages. Journal of Indo-
European Studies 46, 219–231.
Kroonen, G.; G. Barjamovic; M. Peyrot 2018. Linguistic supplement to Damgaard et al.
2018: Early Indo-European languages, Anatolian, Tocharian and Indo-Iranian. DOI:
10.5281/zenodo.1240524.
Lubotsky, A. 1989. Against a Proto-Indo-European phoneme *a. In: T. Vennemann (ed.),
The New Sound of Indo European. Essays in Phonological Reconstruction. Berlin–New
York: Mouton de Gruyter, 53–66.
Melchert, H.C. 2016. “Western Affinities” of Anatolian. In: B. Simmelkjær Sandgaard
Hansen; B. Nielsen Whitehead; T. Olander; B.A. Olsen (eds.), Etymology and the Euro-
pean Lexicon. Proceedings of the 14th Fachtagung der Indogermanischen Gesellschaft,
17–22 September 2012, Copenhagen. Wiesbaden: Reichert, 297–305.
Melchert, H.C. fthc. The position of Anatolian. In: M. Weiss & A. Garrett (eds.), Hand-
book of Indo-European Studies. Oxford: Oxford University Press.
Napol’skix, V.V. 1997. Vvedenie v istoričeskuju uralistiku. Iževsk: RAN.
Norbruis, S. fthc.a. The etymology of PIE *h1es- ‘to be’ (ms.).
Norbruis, S. fthc.b. The etymology of IE *deh3- ‘to give’ (ms.).
Oettinger, N. 2013–2014. Die Indo-Hittite-Hypothese aus heutiger Sicht. Münchener Stu-
dien zur Sprachwissenschaft 67/2, 149–176.
Pedersen, H. 1903. Türkische Lautgesetze. Zeitschrift der deutschen Morgenländischen
Gesellschaft 57, 535–561.
Pronk, T. 2010. On Greek αὐχμός ‘drought’ and αὐχήν ‘neck’. Glotta 86, 55–62.
Pronk, T. 2019. Proto-Indo-European *a. To appear in Indo-European Linguistics 7.
Puhvel, J. 1994. West-Indo-European affinities of Anatolian. In: G.E. Dunkel; G. Meyer;
S. Scarlata; C. Seidl (eds.), Früh-, Mittel-, Spätindogermanisch: Akten der IX. Fachta-
gung der Indogermanischen Gesellschaft vom 5. bis 9. Oktober 1992 in Zürich. Wies-
baden: Reichert, 315–324.
Rieken, E. 2009. Der Archaismus des Hethitischen; eine Bestandsaufnahme. Incontri
Linguistici 32, 37–52.
Starostin, S. 1989. Nostratic and Sino-Caucasian. In: V. Shevoroshkin (ed.), Explorations
in Language Macrofamilies. Materials from the First International Interdisciplinary
Symposium on Language and Prehistory, Ann Arbor, 8–12 November, 1988. Bochum:
Dr. Norbert Brockmeyer, 42–67.
Sturtevant, E.H. 1933. A Comparative Grammar of the Hittite Language. Philadelphia:
Linguistic Society of America/University of Pennsylvania.
Watkins, C. 1963. Preliminaries to a historical and comparative analysis of the syntax of
the Old Irish verb. Celtica 6, 1–49.
Weiss, M. 2016. The Proto-Indo-European laryngeals and the name of Cilicia in the Iron
Age. In: A.M. Byrd; J. DeLisi; M. Wenthe (eds.), Tavet Tat Satyam: Studies in Honor of
Jared S. Klein on the Occasion of His Seventieth Birthday. Ann Arbor–New York: Beech
Stave Press, 331–340.
1 Introduction
Early in the history of Indo-European studies, a PIE element *-r was detected
in several adverbs of nominal and pronominal origin, which was already soon
identified as having a locatival meaning (among others Bartholomae 1889: 14–
25, Brugmann 1903: 456). Nevertheless, there has so far not been any extensive
survey on its history and its position within the realm of IE particles. However,
from time to time the idea of a locative suffix *-r is resumed, but then as addi-
tional supportive evidence for other theories. For instance, Nussbaum (1986:
237) mentions three different formations for the locative of the PIE word for
‘earth’, one being *dhǵh-(e)m-er ‘on earth’ (Av. zəmar-), which would stand next
to the formations *dhǵh-ém (Skt. kṣám-i) and *dhǵh-m-én (Skt. jmán) ‘id.’. There
are, however, many other locatives in *-r that do not have alternative, com-
peting formations. The question is how this element *-r can be qualified. The
literature gives basically two rivaling explanations for the formations at stake,
to which I would like to add a third one:
1. an “endingless” locative of heteroclitic r/n-stems;
2. an adverbial ending;
3. a case ending.
1. One theory is to assume that *-r represents an “endingless” locative of r/n-
stems or some other derivation related to these (e.g. Benveniste 1935: 87–99;
Hajnal 1992a: 211–213). Vanséveren (1999) provides a good overview of the his-
tory of this idea. She ultimately assumes that locatives in *-r are some kind
of “formes casuelles non marquées” (Vanséveren 1999: 117–118). We do in fact
find formations that come in doublets, such as Skt. áhar and áhan ‘in the day-
time’, which could be interpreted in this way. However, not all formations in *-r
have an attested heteroclitic r/n-stem next to them. We also find formations
like Gr. νύκτωρ ‘at nighttime’, which cannot be explained so easily. Since we are
dealing with a (feminine) root noun here, I think that the theory of an ending-
less locative of a hypothetical, unattested heteroclitic stem even obscures the
understanding of that formation. This is also true for many other formations
in *-r. In addition, we would then have to explain why this “endingless” loca-
tive appears in two shapes, namely ending both in *-en and in *-er. There is
the additional difficulty that the latter one contains a full grade: since the root
was accented, we would rather expect zero-grade *-r̥ as is usually attested in
heteroclitics. I think that this explanation is not suitable to explain most of the
formations at stake.
2. Another theory is to assume the existence of an independent “Adverbial-
endung”. Dunkel (2014, I: 162–167) reconstructs *-r, *-er, which he characterizes
as “lokativisch”. To his mind, “Adverbialendungen” can be attached to particles,
pronominal stems, adverbial endings and nominal stems. The problem of this
theory is, in my opinion, the opacity that comes with the term “Adverbialen-
dung”. Although the existence of such a category may be justified on a merely
synchronic level, in the case of formations in *-r that are clearly derived from
nominal stems such an analysis would imply that they had already been lexi-
calized to adverbs before they received the suffix *-r as an adverbial ending. I do
not consider this scenario particularly likely, and would like to claim a nominal
origin instead.
3. I will attempt to show that the element *-r is better understood if we
assume that it was originally confined to nominal and pronominal stems.
Hence, we are dealing with a real case ending, not an “Adverbialendung”. Loca-
tives of this kind can be detected indirectly in secondary thematizations either
prior or posterior to their lexicalization as adverbs, which makes them either
hypostases or deadverbial adjectives. Due to this lexicalization, which was
accompanied by the loss of productivity of *-r, the ending could be reanalyzed
as an adverbial ending in a later phase of PIE. This process gave also rise to
another suffix, *-ter, and ultimately to the Greek and Indo-Iranian compara-
tive suffix.
In the following three sections I will discuss the relevant attestations. These
can be found in three domains: pronouns (1), nouns (2) and other locati-
val adverbs (3). I will first treat the pronominal domain separately from the
nominal domain, for two reasons. First, pronouns often display archaisms as
opposed to nouns. Second, it is controversial whether the suffix *-r as found
with some pronominal stems is identical to the *-r we find in nominal forma-
tions (e.g. Hajnal 1992a: 211). Afterwards, I am going to take into account another
phenomenon, the PIE directive (4), before I present some additional implica-
tions that come along with the locative in *-r.
2 Pronominal Formations
to the one we find in formations of other kinds. The examples usually referred
to are the following:
The temporal meaning of Skt. kár-hi ‘when’ as well as the causal meaning of
Latin cūr ‘why’ can easily be attributed to a secondary semantic shift (cf. de
Vaan 2008 on Lat. cūr). These meanings can be contrasted with Goth. ƕar,
Lith. kur̃ and Arm. ur ‘where’. It is well established that a temporal meaning is
more likely to arise from originally spatial semantics than vice versa (cf. Hop-
per & Traugott 2003: 84–87). Whereas Indo-Iranian and Gothic clearly point
to a reconstruction *kwor and *tor, Baltic, Armenian and Albanian seem to
reflect the vowel -u-.1 Even within Germanic we find some vocalic alternation.
In OHG, for instance, we find wār ‘where’, dār ‘there’. One could argue that this
variation in vocalism shows that the formations themselves are more recent,
since it looks as if *-r has been attached to different stems. However, I think
that this assumption is not necessary, as PIE pronouns generally seem to be
inconsistent with regards to vowel quality and quantity. In particular for the
interrogative stem we find the variants *kwo- (lat. quod ‘what’), *kwi- (lat. quis
‘who’) and *kwu- (lat. ubī ‘where’). Note that the reconstruction of *kwur and
*tur for PIE is counterintuitive with regard to syllabification: we would rather
expect syllabic r̥ preceded by consonantal u̯ . The vocalism of Lith. kur̃ ‘where’,
Latv. tùr ‘there’, Arm. ur ‘where’ and Alb. kur ‘when’ could therefore be con-
sidered analogical, maybe furnished by other pronouns displaying the stem
variant *kwu-. Likewise, the long vowels of Latin cūr < quōr and OHG wār/dār
may be analogical. Therefore, I will follow the usual reconstruction of an early
PIE set of pronominal formations in *-r, namely interrogative *kwor vs. demon-
strative *tor.
If we assume that the suffix *-r lost its productivity towards the end of PIE, it
explains why we find so many different competing formations for interrogative
‘where’, like Greek ποῦ, Skt. kúha, OCS kъde and Lat. ubī. Since these are clear
innovations, I consider *kwor to be original. The opaqueness of this case for-
1 For the vocalism of the Baltic examples, see Forssman (2003: 87–90).
mation may have facilitated the semantics to shift to either causal or temporal
meaning as displayed in Latin and Albanian.
3 Nominal Formations
In the past, some adverbs of clear nominal origin were claimed to contain a
locatival *-r. For that reason, I would like to survey the forms at stake, which
have been treated separately by different scholars (Bartholomae 1889: 14–25,
Nussbaum 1986, Hajnal 1992b, Dunkel 2014, I: 162–167). The majority of these
appear in compounds or derived forms, that is, as hypostases of the locatives
to be reconstructed or adjectives derived from such adverbs. The ones that I
consider likely to be locatives in *-r, are the following:2
a) Av. zamarə-gūz /zəmar-gūz- ‘hidden in the earth’ < *dhǵh-(e)m-er (Nuss-
baum 1986: 236);
b) Skt. uṣar-budh- ‘awake early’, vasar-hā ́ ‘striking early’, Gr. ἠέριος ‘at dawn’,
ἦρι ‘early’ < *h₂us-ér(-i) (Dunkel 2014, I: 165);
c) Gr. νύκτωρ ‘at night’, νύκτερος ‘nightly’, Lat. nocturnus ‘nightly’ (e.g. Beekes
2010 on Gr. νύξ);
d) Lat. hībernus, Gr. χειμέριος, Arm. jmer̄n ‘winter’ < *ǵhei̭m-er(-i)3 (Dunkel
2014, I: 164–165).
a) As was mentioned above, the locatival formation in *-r as attested in Av.
zamarə-/zəmar- stands next to a locatival formation in *-en as attested in Skt.
jmán. To my mind, this is not enough evidence to assume the presence of an
original heteroclitic stem for ‘earth’, however. It is perhaps more likely that Skt.
jmán contains a postpositional *-en as Dunkel (2014, I: 164) suggests, whereas
our zəmarə- as well as Skt. kṣám-i reflect two competing locatives of the actual
root. I assume that the formation in *-er is the older one as it can be linked to the
abovementioned pronominal formations. The vocalism of Av. zamarə-, which
seems to reflect *dhǵhem-er, must be secondary, since otherwise we would have
double full grade. If we accept Av. zəmar- as the older form, it would show the
weak stem and contain the same ablaut structure as the genitive *dhǵh-m-és.
I would therefore rather reconstruct the locative case as *dhǵhm-ér, with zero-
grade in the root and accented e-grade in the ending.
2 Note that the reconstructions are not my own. I have just assembled a selection of the mate-
rial to give an overview.
3 I present here only the most prominent examples given by Dunkel.
b) Skt. uṣar-budh- ‘awake early’, Skt. vasar-han ‘striking early’,4 Gr. ἠέριος ‘at
dawn’ and Gr. ἦρι ‘early’ have already been discussed together by other scholars,
e.g. Nussbaum 1986: 236–247, Hajnal 1992b, Widmer 2004: 117–118. The vocal-
ism of the underlying locative is difficult to reconstruct. Dunkel (2014, I: 165)
considers the unexpected full grade in the root of vasar-hā ́ as secondary. How-
ever, we would have the same difficulty of double full-grade if we interpreted
it as an endingless locative. The assumption that we are dealing with a case
suffix could instead be a possible explanation, as other case formations, such
as the genitive, are also reflected with different ablaut grades in the individ-
ual languages, even if they derive from the same paradigm. Gr. ἠέριος and ἦρι
have been extensively surveyed by Hajnal (1992b) and these indeed seem to be
identical with Skt. uṣar-. I have doubts, however, whether Gr. αὔριον ‘tomorrow’,
which is usually derived from *h₂usri (for instance, Dunkel 2014, I: 165), should
be included here as well. This is not so much for semantic as for derivational
reasons. Whereas αὔριον < *h₂usri shows no full grade at all, ἠέριος and ἦρι point
to a formation *h₂us-s-ér that has been extended by locatival *-i. This indicates
that at a certain point this formation had been lexicalized and could not be
analyzed as a locative anymore, otherwise no additional *-i would have been
required.
c) The problem of ablaut is also obvious for Gr. νύκτωρ ‘at night’, νύκτερος
‘nightly’ and Lat. nocturnus ‘nightly’. The former shows an unexpected length-
ened grade, which is probably secondary, maybe analogous to masculine tor-
stems. The original formation could have been *nokwt-or. The o-grade in the
suffix could then be due to the accent on the root, if we assumed that it was
originally an allophonic variant of accented e-grade. I consider Gr. νύκτερος as
a hypostasis of a locative ‘at night’. An adjective with the meaning ‘being in the
night’ can easily attain the meaning ‘nightly’. In contrast to νύκτωρ, this forma-
tion rather reflects *nokwt-er. Additionally, Latin shows a similar formation in
nocturnus, which may point to original *nokwt-or, too. However, we also find
Gr. νυκτερινός ‘nocturnal’. According to De Vaan (2008: s.v. nox), Lat. noctur-
nus reflects *noctū-rino-, whereas Dunkel (2014, I: 165) reconstructs a preform
*nókṷter-i-no- to account for Gr. νυκτερινός and possibly also Latin nocturnus
and OE nihterne ‘by night’.
d) Just as nocturnus could be analyzed as a derivate of the underlying form of
the Greek examples, Lat. hībernus ‘wintry’ stands next to Gr. χειμέριος ‘wintry’
and Arm. jmer̄n ‘winter’ for which Dunkel reconstructs *ǵhei̯m-er(i). The Greek
compared to Latin herī ‘yesterday’. Dunkel (2014, I: 180) subsumes them under
his contrastive locatival suffix *-tér. As is clearly visible, also Lat. aeternus <
aeviternus and sempiternus rather point to an original ending *-ter as well. I
will return to this topic below, but at least for Latin one reconstructs *aeviter(i)
(see Walde/Hoffmann 1938 on Lat. aetās) and perhaps also *sempiter(i) as the
underlying forms. For diurnus I would rather reconstruct *diuér, again display-
ing the weak stem:
a) loc. *diuér ‘during the day’ vs. gen. *diués.
*h₁uper: Goth. ufar ‘over’, Skt. upári ‘above’, Gr. ὕπερ ‘over’, OIr. for ‘over’;
*h₂eper: Goth. afar ‘after’, OHG abar ‘but’, Skt. ápara- ‘later’;
*ndher: Av. aδairi ‘below’, Goth. undar ‘under’, Lat. īnferus ‘inferior’, Skt.
ádhara- ‘lower’.
5 The formations quoted are just a selection of commonly accepted reconstructions, however,
not everyone would reconstruct anlauting laryngeals. Dunkel, for instance, reconstructs none
for the first etymon. But for the sake of Goth. iup ‘up’ < *h₁eup-, which is extensively treated in
Kroonen (2010), I think an anlauting laryngeal is indeed more likely than the reconstruction
of a vocalic anlaut.
A link between these two words is not commonly recognized. If we are allowed
to connect the Germanic material, which has the semantic sphere of ‘behind’,
to Hitt. kattera- ‘lower, inferior’, which according to most scholars is related to
Gr. κατά ‘downwards’,6 the semantics of the root may be defined as ‘invisible
side’ which could have included the ‘bottom’ as well as the ‘back’.
The rise of PIE adpositions is a controversial topic. However, most would
agree that they are of a secondary nature. Their absence in Indo-Iranian and
Hittite reinforces the theory that they originally developed from adverbs, which
are attested in the oldest languages (e.g. Beekes 1995: 218, Hewson & Bubeník
2006: 1–27). I share the opinion that the grammaticalization of adverbs to adpo-
sitions is posterior to the widely accepted grammaticalization of nouns in a
certain case to adverbs. I think that in pre-PIE we are dealing with a situation
pretty similar to modern Japanese. Here, nouns of locality can be used in adver-
bial phrases, and as such correspond to situations where we would encounter
prepositional phrases, cf. tsukue no ue ni ‘on the table’, literally ‘on top of the
table’ where ue works as a relational substantive, which can be translated with
‘top’. Next to its nominal and its pseudo-prepositional usage, it can also take
the function of an adverb. This view is not new, as we have an oft-cited exam-
ple like this in PIE, too. PIE *h₂enti, which is most often analyzed as the locative
to a root noun *h₂ent- ‘front’, appears in its nominal function in Hittite. In San-
skrit, it has obviously already been grammaticalized to an adverb, whereas that
adverb is used as an adposition in Greek and Latin. According to the theory of
grammaticalization, as elaborated by Hopper & Traugott (2003), we witness a
cline, i.e. a pathway of grammaticalization, which looks like this:
noun + locatival case ending > local adverb > adposition7
Following this cline, we can now apply such a scenario to adverbs like
*h₁uper as well. If we assume that the root underlying h₁uper was originally
nominal and meant something like ‘surface, top’, just like Jap. ue preserves that
nominal meaning, we can easily understand that h₁uper, which would then be
the locative to the root noun h₁up- ‘top, surface’, originally meant ‘on top’ or ‘on
the surface’:
6 Cf. Dunkel (2014, II: 419–422), who reconstructs *ḱat-, and Beekes (1995: 21), who reconstructs
*ḱnt-.
7 A detailed discussion of the different stages of such a cline is offered by Hewson & Bubeník
(2006: 365–370).
PIE *h₁up- ‘top’: loc.*h₁up-ér ‘on top’ > *h₁upér (adv.) ‘above’ > * h₁upér
(adp.) ‘over X’.
I regard the comparative meaning of adjectives like Lat. īnferus ‘inferior’ and
Skt. upara- ‘lower, later’ as secondary. A productive comparative suffix *-tero
can only be reconstructed for Greek and Vedic. Dunkel derives this suffix from
his contrastive, locatival “Adverbialendung” *-tér. Already Brugman (1904: 321–
323) derives *-ero and *-tero from local adverbs in *-er and *-ter, treating both
formations equally, and Benveniste (1948: 114–163) discusses the connection
between *entér and the comparative suffix. It is easy to understand how a suf-
fix *-ero- from *-er-o- resp. *-tero- from *-ter-o could obtain comparative and
contrastive functions. The comparative meaning of any local adverb could be
viewed as inherent: when there is something ‘above’, there must be something
‘below’ as a reference point, cf. for example German der Obere ‘the one on
top’.
If we assume that at a certain stage of PIE we had *-er and *-ter as parallel
adverbial suffixes both originally denoting locality, this could explain why we
have reflexes of both *-ero- and *-tero- in many languages, however still limited
to local adverbs, whereas only Greek and IIr. developed a secondary compar-
ative suffix out of *-tero- (e.g. Risch 1973: 91–92). This means that we cannot
reconstruct a comparative suffix *-tero- for PIE. Originally formations contain-
ing *-(t)ero- were merely thematizations of local adverbs, from which some had
been lexicalized from hypostases to locatives ending in *-er.
Many of these hypothetical nominal roots form the base for other adverbs.
This, in itself, is nothing unusual; cf. prepositions like Latin pro, prae or sim-
ilar formations in the other languages. It is, however, striking that *-er seems
to interchange with *-o. This *-o, I assume, must be the directive ending (fol-
lowing Dunkel 2014, I: 154–161).9 To my knowledge, the comparatively frequent
parallelism of formations in *-er with those in *-o has received little attention.
Striking examples for this are:
Gr. ὕπερ, OIr. for < *h₁up-er vs. Gr. ὕπο ‘below, under’, OIr. fo ‘under’
<*h₁up-o;
Goth. afar < *h₂ep-er vs. af ‘from’, Gr. ἀπό ‘from’, Lat. ab ‘from’ <*h₂ep-o;
Goth. hindar, Hitt. kattera- < *ḱ(e)nt-er(-) vs. Hitt. katta ‘downwards’
<*ḱnt-o.
In the case of Goth. und ‘till’, which Dunkel derives from *n̥ dhí, whereas Lühr
(2000: 53) reconstructs *n̥ dhóm, I would rather suggest that we should recon-
struct *ndhó. For the meaning of *h₁upo ‘upwards’, which is counterintuitive, I
refer to García-Hernández (1995), who has convincingly argued that the orig-
inal semantics of the underlying root was not ‘under’. This is supported by
Germanic attestations such as Goth. uf ‘up’. In contrast to García-Hernández
(1995: 163), I do not consider *h₁uper as a derivative from *h₁upo, but view both
as equal formations. If we analyze Latin quō ‘whither’ as reflecting PIE *kwo-o
(e.g. Dunkel 1994: 21–22), we would have a striking parallel to that in the pro-
nouns as well. Perhaps we may also reconstruct a contrast between Skt. avár
‘below, downwards’ and áva ‘down, away’. Another possible example could be
found in the reflexes of PIE *ud, if we assume an opposition *ū ̆ dér, which may
be reflected by Skt. udára- ‘belly’ (with thematization) and OHG ūzar ‘out of’,
vs. *ū ̆ dó as in Goth. ūt ‘out’. According to my previous point, we could then
reconstruct early nouns in the directive case:
In my opinion, all this strongly suggests that we are dealing with a paradigmatic
opposition. However, this is only true for a certain early stage of PIE. Whereas
*-o in its allative function could keep its productivity at least until Anatolian,
*-r has apparently ceased to be productive in all attested languages. The cited
formations are merely lexicalized remnants of a once productive category.
6 Conclusions
The examples treated in the preceding sections show that *-r was originally a
case ending. The opposition of locative *-r and directive *-o, as in *h₁upér vs.
*h₁upó, was ultimately replaced by the already existing pair opposing the loca-
tive *-i and the accusative, which is generally assumed to incorporate directive
function. There is also evidence for an opposition between *-i and *-m from the
accusative. Next to *h₂enti, with its clear nominal origin, we also find *h₂entm,
which is reflected in Gr. ἄντα ‘against’ (e.g. Beekes 2010 on ἄντα).
As morphemes usually are not replaced that easily, one could think of two
originally separate but coexisting categories. For instance, *-r/-o could have
denoted an opposition adessive/allative and *-i/-m possibly inessive/illative or
vice versa, pretty much like we find it in Uralic languages. With the merger of
those distinctive categories into the simple opposition locative/directive one
of the morpheme-pairs disappeared.
For the history of *-r, we can posit the following stages:
i. *-r as a locative case ending;
ii. *-ero and other derivations as locatival hypostases;
iii. *-r loses its productivity as a locative case marker, *-i takes over its func-
tion;
iv. lexicalization of locatives in *-r to adverbs;
v. reanalysis of *-er as an adverbial ending;
vi. *-ter appears as a variant;
vii. deadverbial *-ero- and *-tero- attain contrastive and comparative func-
tion.
The idea that *-r was not just a mere particle attached to stems, but rather a
case ending, is supported by the fact that it shows ablaut. Whereas we find zero-
grade in the pronominal stems, like *to-r, we see accented e-grade in formations
like *h₁up-ér ‘on the top’ and many of the nominal examples, and ultimately
unaccented o-grade in νύκτωρ ‘at night’ and nocturnus ‘nightly’. This assump-
tion is apt to explain some of our PIE adverbs, which have until now remained
opaque. At least the ones ending in *-r can be better understood, if we consider
them as lexicalized locatives of nouns.
7 Further Implications
10 I hereby thank Rasmus Gudmundsen Bjørn who suggested this link after my lecture in
Leiden 2015.
References
Bartholomae, C. 1889. Arisches. Beiträge zur Kunde der indogermanischen Sprachen 15,
1–43.
Beekes, R.S.P. 1995. Comparative Indo-European Linguistics: An Introduction. Amster-
dam–Philadelphia: John Benjamins.
Beekes, R.S.P. 2010. Etymological Dictionary of Greek. With the assistance of Lucien van
Beek. Leiden–Boston: Brill.
Benveniste, E. 1935. Origines de la formation des noms en Indo-Européen. Paris: Adrien-
Maisonneuve.
Benveniste, E. 1948. Noms d’agent et noms d’action en Indo-Européen. Paris: Adrien-
Maisonneuve.
Brugmann, K. 1903. Kurze vergleichende Grammatik der indogermanischen Sprachen:
auf Grund des fünfbändigen ‘Grundrisses der vergleichenden Grammatik der indoger-
manischen Sprachen von K. Brugmann und B. Delbrück’ verfasst. Zweite Lieferung:
Lehre von den Wortformen und ihrem Gebrauch. Strassburg: Trübner.
Dunkel, G.E. 1994. The IE directive. In: G.E. Dunkel; G. Meyer; S. Scarlata; C. Seidl (eds.),
Früh-, Mittel-, Spätindogermanisch: Akten der IX. Fachtagung der Indogermanischen
Gesellschaft vom 5. bis 9. Oktober 1992 in Zürich. Wiesbaden: Reichert, 17–36.
Dunkel, G.E. 2014. Lexikon der indogermanischen Partikeln und Pronominalstämme.
Band 1: Einleitung, Terminologie, Lautgesetze, Adverbialendungen, Nominalsuffixe,
Anhänge und Indices. Band 2: Lexikon. Heidelberg: Winter.
Forssman, B. 2003. Das baltische Adverb: Morphosemantik und Diachronie. Heidelberg:
Winter.
García-Hernández, B. 1995. Die Evolution des lat. sub und die Urbedeutung des idg.
*(s)upo. Indogermanische Forschungen 100, 163–171.
Greenberg, J.H. 2000. Indo-European and its closest relatives: the Eurasiatic language
family. Vol. 1: Grammar. Stanford: Stanford University Press.
Hajnal, I. 1992a. Griechisch χαμαί—ein Problem der Rekonstruktion? In: R.S.P. Beekes;
A. Lubotsky; J.J.S. Weitenberg (eds.), Rekonstruktion und relative Chronologie. Akten
der VIII. Fachtagung der Indogermanischen Gesellschaft, Leiden, 31. August–4. Sep-
tember 1987. Innsbruck: IBS, 207–220.
Hajnal, I. 1992b. Homerisch ἠέριος, Ἠερίβοια und ἦρι: Zur Interrelation von Wortbedeu-
tung und Lautform. Historische Sprachforschung 105/1, 57–72.
Hewson, J.; V. Bubeník 2006. From Case to Adposition: The Development of Configura-
tional Syntax in Indo-European Languages. Amsterdam: John Benjamins.
Hopper, P.J.; Traugott, E.C. 2003. Grammaticalization. Cambridge: Cambridge Univer-
sity Press.
Kloekhorst, A. 2008. Etymological Dictionary of The Hittite Inherited Lexicon. Leiden–
Boston: Brill.
Kortlandt, F. 2002. The Indo-Uralic verb. In: R. Blokland; C. Hasselblatt (eds.), Finno-
Ugrians and Indo-Europeans: Linguistic and Literary Contacts. Maastricht: Shaker,
217–227.
Kortlandt, F. 2010. Indo-Uralic and Altaic. In: F. Kortlandt Studies in Germanic, Indo-
European and Indo-Uralic. Amsterdam: Rodopi, 415–418.
Kroonen, G. 2010. On Gothic iup and the Germanic directionals. North-Western Euro-
pean Language Evolution 58/59, 367–380.
Kümmel, M.J. 2002. Das Perfekt im Indoiranischen. Eine Untersuchung der Form und
Funktion einer ererbten Kategorie des Verbums und ihrer Entwicklung in den altin-
doiranischen Sprachen. Wiesbaden: Reichert.
Lühr, R. 2000. Gedichte des Skalden Egill. Dettelbach: J.H. Röll.
Nussbaum, A.J. 1986. Head and Horn in Indo-European. Berlin–New York: Walter de
Gruyter.
Risch, E. 1973. Wortbildung der homerischen Sprache. Berlin: Walter de Gruyter.
Vanséveren, S. 1999. Thèmes en *-r/-n, “locatif sans désinence” et histoire de la flexion
nominale. Indogermanische Forschungen 104, 110–119.
de Vaan, M. 2008. Etymological Dictionary of Latin and the Other Italic Languages.
Leiden–Boston: Brill.
Walde, A., J.B. Hoffmann 1938. Lateinisches Etymologisches Wörterbuch. Heidelberg:
Winter.
Widmer, P. 2004. Das Korn des weiten Feldes. Interne Derivation, Derivationskette und
Flexionsklassenhierarchie: Aspekte der nominalen Wortbildung im Urindogermani-
schen. Innsbruck: IBS.
Comparing Indo-European with Uralic makes sense for several reasons. Both
families are relatively well-defined (although some aspects of Uralic do sug-
gest wider affinities to Yukaghir) with widely accepted proto-stages; there is
no shortage of demonstrable layers of adstrate phenomena between individ-
ual branches (e.g. Indo-Iranian on Finno-Ugric, Germanic on the Balto-Fennic
languages, and possibly Tocharian on Samoyedic or vice-versa), and archaeo-
logical authorities abide in locating respective homelands in the immediate
vicinity of one another on the upper Volga more than 5,000 years ago (cf.
Anthony 2007 and Carpelan & Parpola 2001). On the proto-level, however, only
two lexical items are commonly accepted as shared, viz. ‘water’ and ‘name’, and
there is no convincing argument as to whether to regard them as shared her-
itage or contact phenomena.
The best point of departure to address these problems is by paying close
attention to the pronouns. Pronouns belong to the bedrock of any given lan-
guage; while certainly not impervious to change,1 their frequency in most
speech acts and proven track record in the uncontroversial histories of Indo-
European and Uralic languages attest to their resilience. Pronouns also provide
a closed system of correspondences between form and function that reduces
the risks of chance resemblances that is inextricably tied to lexical comparisons
(cf. Babaev 2009: 38). If an exclusive Indo-Uralic unity existed the pronouns
surely testify to their shared inheritance; if it did not, the similarities must have
come about through contact and provide other valuable insights. This discus-
sion aims to test the Indo-Uralic hypothesis based on three sets of pronouns:
* This article is based on a 2016 paper directly inspired by the 2015 Leiden conference. I am
indebted to Alwin Kloekhorst, Tijmen Pronk, Seán Vrieland, and Thomas Olander for valu-
able comments in the process.
1 Although pronominal borrowing does happen, e.g. Modern English 3rd person plural pro-
nouns (they, them, and their) from Old Norse, systemic pronominal borrowing constitutes an
extreme situation (cf. Cavoto 2004: 16); the phenomenon is treated in Thomason & Everett
(2001).
the interrogative complex (*kw-, §1), the anaphoric and relative pronouns (*i-,
§ 2), and the deictic animacy distinction (*s-/*t-, § 3).2 Each section sets out to
go beyond appearances and discuss the possible interpretations of the corre-
spondences, including implications of shared heritage—ultimately to test the
Indo-Uralic hypothesis on its bearings.
1 Interrogative *kw-
Both families agree to have the animate interrogative begin with a velar and
somewhat opaque ablaut. Internal evidence suggests that the PIE pronoun
continues a pre-PIE contraction of a general interrogative marker *ku (§ 1.2)
and the anaphoric pronoun *is (§3), which may be corroborated by certain
Uralic features.
1.1 Pronoun
This section deals with the correspondences between PIE *kwi- and PU *ki (see
forms in table 3.1) and between PIE *kwo- and PU *ku-(~*ko-) (table 3.2).
PIE can be reconstructed to have ablaut within the interrogative paradigm,
originally likely between i graded casus rectus (cf. OLat. quis, Av. ciš, Hit. kuis)
and o/e graded obliquus (cf. OLat. quō, Av. cahiia, Hit. kuel, cf. § 3.1). This ablaut
is qualitatively different from the verbal and nominal ablaut. There is no doubt
that the velar was labial at the point of departure of the Anatolian branch.
Paradigmatic levelling is common in the individual branches, e.g. Greek where
the i-grade is generalized. This distinction must necessarily echo a language
state earlier than PIE, the cause of which has been suggested by Olsen & Ras-
mussen as accentual variation, viz. that the oblique adverbial forms attract
the accent and are consequently “full-bodied”, whereas the nominal forms are
unaccented and accordingly “reduced” (1999: 422; 433).3 It is unclear at which
pre-PIE stage this accentual conditioning would apply. The Uralic material
seems to merit a similar conclusion, but the prevailing theory holds that the
system is suppletive, echoing two distinct stems that only secondarily came
2 The personal pronouns are uncontroversial: The 1st singular is trivial, at least for the PIE
oblique *me to PU *min, while a conditioned sound law is needed to align the 2nd per-
son singular pronouns, PU *sin with PIE *tu- (or, likely *ti, cf. Kloekhorst 2008a: 93, cf. also
§3.3).
3 An attempt to include the u-vocalism from labial rounding (Olsen & Rasmussen 1999: 433),
this author argues, does not appear to be warranted (cf. § 2.1).
PIE *kwi-
together (Tapani Salminen, p.c., cf. also Janhunen 1982: 28). The co-occurrence
of the different grades is most pronounced in Finnish, cf. kuka (nom.an.sg.)
and kenet (acc.an.sg.), but the variation permeates the interrogative complex
throughout the Uralic dialects, e.g. Enets (front vowel) sio ‘who’ and (back
vowel) hoke ‘which’.
1.2 Adverb
Related to the discussion of the pronoun is the adverbial particle that can
be traced back to PIE as *k(w)u- (table 3.3), but in the reconstruction of PU
the adverb is difficult to extricate completely from the pronominal forms (cf.
PIE *kwo-
the Uralic examples given in table 3.2; further examples are given in the text
below).
The basic function of the unmarked stem *ku must have been interroga-
tive, most likely with inherent locative meaning (although two common mor-
phemes seem to echo that meaning, viz. *-dh- and *-r); further extensions retain
the interrogative/indefinite character with spatial, temporal, or pronominal
content. The lack of labiality in these forms is traditionally ascribed to the βου-
κόλος rule, viz. that labiality is lost in the vicinity of other labial sounds, but this
explanation rests on the assumption that these adverbial forms automatically
had u-vocalism, and may be rendered superfluous since another interpretation
is possible at this pre-PIE stage: Dunkel thus proposes that the interrogative
pronoun *kwis is composed of the interrogative particle *ku and the anaphoric
pronoun *is (§3) to form *ku=is > *kwis in pre-PIE (2014, II: 441). This analy-
sis helps explain possible paradigmatic similarities between the interrogative
PIE *ku-
Vedic kū ‘where’ -Ø
Avestan kū ‘id.’ -Ø
Albanian ku ‘id.’ -Ø
Armenian owr ‘id.’ -r
Lithuanian kur̃ ‘id.’ -r
OCS kŭde ‘id.’ -dh-
Lydian kud ‘id.’ -dh-
Welsh cw(dd) ‘id.’ -Ø(-dh-)
Oscan puf ‘id.’ -dh-
Latin (c)ubi ‘id.’ -dh-
Gothic -hun ‘some’ -n(-)
Cypriotic Greek ὀπυι ‘whereto’ -i
Hittite kuššan ‘when’ (complex)
and anaphoric pronouns (cf. Rasmussen 1999: 323 and Szemerényi 1990: 208).
Furthermore, it gives a plausible internal development of the labiovelar from
pre-PIE *KuV- > PIE *KwV (cf. Kümmel 2007: 319). This scenario then turns
the tables on the common analysis of the interrogative complex: the labiovelar
forms arose under specific conditions and only secondarily spread to adverbial
forms through analogy, cf. Oscan puf (and similarly Greek -πυ-):
*ku=ís > *kwí-s > pis ‘who’ (cf. Latin quis, Welsh pwy)
*ku=dhi > **cuf ⇒ puf ‘where’ (cf. Latin [c]ubi, Welsh cw[dd])
PIE *-kwe
Uralic
and Finnish -ko,4 as in on-ko5 kissa musta? ‘is the cat black?’ (cf. kissa on musta
‘the cat is black’), and the proposition to include it in the general interrogative
stock does not seem far-fetched (SKES: 205). Closely related hereto, it appears,
is the suffix PIE *-kwe.
4 Or -kö; this alternation is due to the widespread Uralic phenomenon of vowel harmony and
only a single form is given here.
5 Note that the suffix always takes the second position of the sentence, cf. in PIE the enclitic
slot described in Wackernagel’s Law.
uum, too; the seemingly odd contradiction that inflectional IE should be better
at retaining an agglutinative feature is probably off-set by the discrepancy in
time of attestation. Note that similar agglutinative vestiges survive in the PIE
possessive pronouns still suffixal in Anatolian and are alive in Uralic today (cf.
Kloekhorst 2008a: 90ff. & Čop 1979: 18ff.).
A case for contact phenomena in the enclitic particle, PIE *-kwe, could be
made (cf. §1.3), although areal convergence seems a better hypothesis than
wholesale borrowing; the multifaceted Finnish use of the particle is too
nuanced to have been adopted completely from Proto-Germanic. Permic *-kö
could strictly speaking have been influenced from a later strain of IE in the
northern Caspian steppes, although a regular Uralic development is just as
plausible. Kamassian -go is difficult to explain directly from contact with an IE
language, although it belonged to the southernmost Samoyedic branch; a last-
ditch effort could thus have the suffix borrowed from a precursor of Tocharian
in the Sayan region, but the evidence for such a transfer is non-existing. The
complex as a whole, however, is securely reconstructable to both proto-stages
without any obvious signs or impetus for borrowing.
2 Anaphoric *i-/e-
PIE *i-
PU *i-?
with subsequent loss of the velar labiality is still possible, but requires signif-
icantly more complex developments, suggesting that there was no aboriginal
u-graded interrogative pronoun **kwu-. The stem is also widespread in Uralic,
although the pronominal function is deictic rather than anaphoric and some-
what marginalized.
Uralic
3 Demonstrative *so/to-
The PIE 3rd person pronoun has suppletive distinction in the casus rectus of
the animate gender (*so) as opposed to the inanimate (*tod), and by closer
inspection a similar distinction can be reconstructed for the Uralic proto-
language.
7 The suffixed pronoun is the Balto-Slavic marker of adjectival definiteness, cf. Lithuanian gẽras
‘good’ with geràs-is ‘the good (one)’.
8 This form may be a loan from Finnish, cf. Sammallahti (1998: 250) who teeters between inher-
ited and borrowed.
figure 3.1 Stylized visualization of established contacts events (b and c) and the period of
contact (a) suggested to account for the shared pronominal stock, whether inher-
ited or borrowed. The possible cloud of Caucasian influence on (pre-)PIE follows
the Uhlenbeck-Kortlandt hypothesis (cf. Kortlandt 2010: 155 f.).
Vedic sá ‘he’
Avestan hā, hō ‘id.’
Greek ὁ ‘id.’
Gothic sa ‘id.’
Latin sapsa ‘herself’
Toch. A səm ‘he’
Old Irish -so ‘this’
Hittite šu- (clause ptcl.)
PU *so(n)
9 In accordance with the Anatolian evidence that came to aid the suggestive hints already avail-
able from the classical languages, the feminine formation is treated as an innovation in Core
IE (cf. Szemerényi 1990: 155f.).
PIE *tod10
PU *to/tu
10 The rather elusive *-d ending of inanimate pronouns in PIE (*kwid, *tod, *id) might be
explained from the ablative, thus mirroring the more lucid accusative *-m formant in the
nominal system. The form may be cognate with the Uralic partitive *-tV, continued in
Finnish that notably distinguishes between objects in the accusative (“complete”, hän luki
kirjan ‘he read a book’) and partitive (“partial”, hän luki kirjaa ‘he read a book [but did not
finish]’). This hypothesis introduces a systemic alternative to internal explanations such,
e.g. “emphatic reduplication” *to-to > *to-d (Szemerényi 1990: 205).
PU *te/*ti
11 As a phonological corollary, and to account for the discrepancy between the larger number
of Uralic affricates and the PIE lack hereof, Collinder proposes that since the un-lenited
Uralic affricates are reflected by t in Samoyedic, it is not unconceivable that PIE went
through a similar process of deaffrication (1965: 130), and a piece of the 2nd person puz-
zle (i.e. from tentative PIU *tsi, cf. §0.1) could potentially be found here, thus:
Uralic
PIU? PIE
PU North Sam. (excl. Selkup)
*t *t *t
*t
*ts *ts *s / *t
*s *s *s *s
A similar explanation departing from the uneasy IE treatment of affricates was proposed
by Heller for the *so/*tod distinction (1956: 7f.), but as a completely internal process. I
argue that a potential phonetic split must have older bearings.
figure 3.2 Perspectives on proto-languages. (a) is hypothetical while (b) and (c) are uncon-
troversial (notwithstanding their respective places on the time axis). These points
of convergence represent shared non-trivial phenomena, e.g. animate *so vs.
inanimate *to. The (x)’s represent individually attested languages.
4 Conclusion
While ablaut certainly is not a stable feature of the Uralic languages, deic-
tic gradation is widespread in the pronouns and it cannot be ruled out that
the variation in the interrogative and anaphoric complex originates in ablaut
rather than from suppletion. The use of particles resonates well with the con-
joining stage of an agglutinative language (PU) and an inflectional language
(PIE) as either agglutinative or isolative, and it may, indeed, be in this unavoid-
able projection that Indo-Uralic unity becomes difficult to substantiate, i.e. in
the relative scarcity of inflectional morphology to extricate.
Two of the three pronominal complexes (interrogative § 1 and animacy § 3)
betray such intricate correspondences at the proto-level that chance resem-
blance alone must be ruled out entirely. There are, however, two promising
scenarios for convergence (the enclitic particle, § 1.3 and the relative pronoun,
§ 2.2), implying again that speakers of Uralic and Indo-European languages
have remained in contact for millennia; but these cases remain marginal and
do not have any bearings on the system as a whole. The major points of crit-
icism seem automatic (e.g. “lautsymbolismus”, Rédei 1986: 19 and “chance”,
Campbell 1998: 26), but ring hollow when compared to the multi-layered evi-
dence of sound correspondences and vowel gradation presented above, and
even Koivulehto, who is very efficient in demonstrating lexical loan relations,
admits the stronger case presented by pronouns in the Indo-Uralic hypothesis
Nach seiner beredten aussage würde sogar das ganze gebäude der verglei-
chende (indogermanischen) sprachforschung zusammenstürzen, falls
man die vorhandenen sprachlichen analogien nicht als beweis für den
gemeinsamen ursprung der indoeuropäer und finno-ugrier gelten lässt.
Paasonen 1907: 13
References
Abondolo, D. 1998. Introduction. In: D. Abondolo (ed.), The Uralic Languages. London–
New York: Routledge, 1–42.
Anthony, D.W. 2007. The Horse, the Wheel and Language. Princeton, NJ: Princeton Uni-
versity Press.
Babaev, K. 2009. Once again on the comparison of pronouns in Proto-Languages. Jour-
nal of Language Relationship 1, 37–48.
Bjørn, R.G. 2017. Foreign Elements in the Proto-Indo-European Vocabulary. MA thesis,
University of Copenhagen. Available at www.loanwords.prehistoricmap.com.
Bomhard, A.R. 2008. Reconstructing Proto-Nostratic. Leiden–Boston: Brill.
Buck, C.D. 1910. Greek Dialects. Boston: Ginn and Company.
Campbell, L. 1998. Nostratic: A personal assessment. In: J.C. Salmons; B.D. Joseph
(eds.), Nostratic: Sifting the Evidence. Amsterdam–Philadelphia: John Benjamins,
107–152.
Campbell, L.; W.J. Poser 2008. Language Comparison. Cambridge: Cambridge Univer-
sity Press.
Carpelan, C.; A. Parpola 2001. Proto-Indo-European, Proto-Uralic and Proto-Aryan. In:
C. Carpelan; A. Parpola; P. Koskikallio (eds.), Early Contacts between Uralic and Indo-
European: Linguistic and Archaeological Considerations. Helsinki: Finno-Ugrian
Society, 55–150.
13 Yet the borrowing party then would have refrained from taking on, e.g., numerals, that
would appear to be significantly more attractive items to borrow (cf. Bjørn 2017: 141).
Janhunen, J. 1998. Samoyedic. In: D. Abondolo (ed.), The Uralic Languages. London–
New York: Routledge, 457–480.
Janhunen, J. 2001. Indo-Uralic and Ural-Altaic. In: C. Carpelan; A. Parpola; P. Koskikallio
(eds.), Early Contacts between Uralic and Indo-European: Linguistic and Archaeolog-
ical Considerations. Helsinki: Finno-Ugrian Society, 207–220.
Joki, A.J. 1973. Uralier und Indogermanen: Die älteren Berührungen zwischen den urali-
schen und indogermanischen Sprachen. Helsinki: Finno-Ugrian Society.
Kloekhorst, A. 2008a. Some Indo-Uralic aspects of Hittite. Journal of Indo-European
Studies 36, 88–95.
Kloekhorst, A. 2008b. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden–
Boston: Brill.
Koivulehto, J. 2001. The earliest contacts between Indo-European and Uralic speak-
ers in light of lexical loans. In: C. Carpelan; A. Parpola; P. Koskikallio (eds.), Early
Contacts between Uralic and Indo-European: Linguistic and Archaeological Consider-
ations. Helsinki: Finno-Ugrian Society, 234–263.
Kortlandt, F. 2010. Indo-Uralic and Altaic revisited. In: L. Johanson; M. Robbeets (eds.),
Transeurasian Verbal Morphology in a Comparative Perspective. Wiesbaden: Harras-
sowitz, 153–164.
Kümmel, M.J. 2007. Konsonantenwandel. Wiesbaden: Reichert.
Künnap, A. 1999. Kamass. München: Lincom.
Olsen, B.A.; J.E. Rasmussen 1999. Indo-European -to-/-tu-/-ti-: A case of phonetic hierar-
chy. In H. Eichner; H.C. Luschützky; V. Sadovski (eds.), Compositiones Indogermani-
cae in Memoriam Jochem Schindler. Prague: Enigma, 421–435.
Paasonen, H. 1906. Die finnischen Pronominalstämme jo- und e-. Finno-Ugrische For-
schungen 6, 114–117.
Paasonen, H. 1907. Zur Frage von der Urverwandtschaft der finnisch-ugrischen und
indoeuropäischen Sprachen. Finno-Ugrische Forschungen 7, 13–31.
Rasmussen, J.E. 1999. Indo-European ablaut -i- ~ -e-/-o-. In: J.E. Rasmussen Selected
Papers on Indo-European Linguistics. With a Section on Comparative Eskimo Linguis-
tics. Part 1. Copenhagen: Museum Tusculanum, 312–326.
Rasmussen, J.E. 2005. Der Akkusativ auf *-m im Indogermanischen und Uralischen:
Kontakt oder Erbe? In: G. Meiser; O. Hackstein (eds.), Sprachkontakt und Sprachwan-
del. Akten der XI. Fachtagung der Indogermanischen Gesellschaft (13.–23. September
2000, Halle an der Saale). Wiesbaden: Reichert, 525–536.
Rédei, K. 1986. Zu den indogermanisch-uralischen Sprachkontakten. Wien: Verlag der
Österreichischen Akademie der Wissenschaften.
Rédei, K. 1988. Uralisches Etymologisches Wörterbuch. Wiesbaden: Harrassowitz.
Renfrew, C. 2003. Archaeology and Language: The Puzzle of Indo-European Origins.
Cambridge: Cambridge University Press.
Ringe, D. 1998. Probabilistic evidence for Indo-Uralic. In: J.C. Salmons; B.D. Joseph
Indo-Anatolian Syntax?
Dag Haug and Andrei Sideltsev*
(1) MH/MS (CTH 147) KUB 14.1+ rev. 45, cf. Beckman 1996: 146.
nu=wa=mu mān idālu-n memia-n
CONN=QUOT=me if evil-ACC.SG.C word-ACC.SG.C
kui-š [mema-i]
who-NOM.SG.C tell-3SG.PRS
‘If anybody tells me a bad word’.
(2) MH/MS (CTH 199) ABoT 1.65 obv. 11, cf. Hoffner 2009: 243.
ammug=a āššul UL kuit
me.DAT.SG=but greeting.ACC.SG.N NEG which.ACC.SG.N
ḫa[tr]ā-eš
write-2SG.PST
‘To me you did not send any greetings’.
Here we see wh-words/relative pronouns kuiš ‘who’ and kuit ‘what’ used instead
of expected indefinite pronouns kuiš-ki ‘some/anyone’, kuit-ki ‘some/anything’.
Such pronouns are termed bare interrogatives (Haspelmath 1993; Haug 2016)
and this will be the term we will be using in the paper. The use is instantly rem-
iniscent of that attested in all other Indo-European languages (3):
* We thank the audience at the Workshop for suggestions. Our particular gratitude goes to
I. Yakubovich for providing quite a few valuable suggestions and stimulating criticism. The
authors remain responsible for all possible errors of fact or interpretation.
1 Including those introduced by irrealis particle man or našma ‘or (if)’.
2 We will not treat here the distributive use of wh-words.
c. Latin
si quis
‘if anyone’
This use is attested in all the ancient languages of all the branches save Arme-
nian and Tocharian and is reconstructed for narrow PIE as the use of wh-words
en lieu of indefinite pronouns under specific licensing conditions, most com-
monly conditionals, questions and negations, to a much more restricted degree
in modal contexts (Haug 2016). Mind that the reconstruction is that of the
pattern—so it holds irrespectively of what wh-word is reconstructed. In most
cases the wh-words are reflexes of the bare interrogative/relative root *kwi/o-,
possibly expanded with one or more particles. When such a form is used as an
indefinite, it will be termed a bare interrogative, irrespective of whether it is
etymologically augmented or not, as long as the same form can also be used as
an interrogative and/or as a relative.
The correspondence between Hittite and narrow IE is held to be obvious
and of common ancestry, see, e.g., Hoffner & Melchert (2008: 149), cf. descrip-
tively from the inner Hittite perspective Huggard (2015: 34–35). However, it is
not so straightforward. As is well known, Hittite attests the use in (1–2) in post-
OH period (CHD sub mān). The oldest attested Hittite texts (OH/OS originals)
have only indefinite pronouns in conditional clauses (introduced by takku ‘if’)
and after negation markers, as in:
(4) OH/OS (CTH 291.I.a.A) KBo 6.2 obv. ii 33 (§ 44a), cf. Hoffner 1997: 52.
takku LÚ-an paḫḫuen-i kuiški
if man-ACC.SG.C fire-LOC.SG somebody.NOM.SG.C
peššie-zzi
throw-3SG.PRS
‘If anyone makes a man fall into a fire, …’.
(5) OH/OS (CTH 1.A) KBo 3.22 obv. 7–8, cf. Neu 1974: 10–11.
Ù DUMUMEŠ URU Neš[aš id]ālu natta kuedanikki
and sons Nesa evil.ACC.SG.N NEG someone.DAT.SG
takkiš-ta
set.up-3SG.PST
‘And he plotted no evil to any of the citizens of Nesa’.
Other attestations from the same text KBo 3.22 are lē kuiški (obv. 24, with dam-
aged lē), [l]ē kuiški (obv. 34). Indefinite pronouns licenced by operators (condi-
tional and negation) are termed in cross-linguistic literature negative polarity
items (NPIs) (Haspelmath 1997). This is the term we will use in the paper.
A count of the OH/OS corpus revealed 62× takku kuiški/kuitki vs 0× *takku
kuiš/kuit. According to (CHD sub mān), the latter is sporadically attested only
in later copies of OH texts and is very likely to reflect MH/NH usage. One of the
earliest examples of bare interrogatives is:
(6) MH/MS (CTH 262) IBoT 1.36 obv. ii 63, cf. Miller 2013: 112–113.
mān3=aš?=ši peran=ma kuwapi KASKAL–i-š ḫatku-š
if=it?=him ahead=but where road-NOM.SG.C narrow-NOM.SG.C
‘If the road ahead is at some point too narrow for him’.
As for Old Hittite bare interrogatives, there is only one case in the NH copy of
the OH edict of Telipinu, see (7), versus regular marked NPI s in the Old Hittite
originals, for which see (5) above:
(7) OH/NS (CTH 19.II.A) KBo 3.1+ obv. ii 44, cf. Hoffmann 1984: 32–33.
parkunu-ši=ma=za UL kuit
purify-2SG.PRS=but=REFL NEG which.ACC.SG.N
‘But you do not purify in any way’.
Example (7) occurs in the same text with three regular NPI s licenced by the
negation. One of them is in the next line of the text (obv. ii 45 lē kuinki), the
other one is in obv. 35 lē kuiški, and one is fragmentary (obv. ii 14). So Hoffmann
(1984: 32) probably had good reason to assess (7) as a copyist’s mistake and read
kuit as kuit⟨ki⟩.
This immediately raises the important problem of purely scribal errors, not
normally discussed in relation to bare interrogatives in Hittite, although cf.
Hoffner (2009: 196). As is known, indefinite pronouns and wh-words in some
forms differed only by one sign -ki/ka, e.g., NOM.SG.C relative kuiš vs indefinite
kuiš-ki. So, when the sign was not written due to a scribal mistake, some forms
of the indefinite pronoun were indistinguishable from those of the wh-word.
However, there are also cases which can only be interpreted as unambiguous
mistakes, and not as bare interrogatives. These are seen when the scribal mis-
take occurred in the forms which differed by more than one sign and thus the
omission of -ki/ka did not result in homonymy, see
(8) NH/NS (CTH 255.2.A) KUB 26.1+ rev. iv 20–21, cf. Miller 2013: 304–305.
[(mānn=a=)]ddu=za DUTU–Š=I kuedanik⟨ki⟩ memiyan-i
if=and=you=REFL Majesty=My som⟨e⟩.DAT.SG matter-DAT.SG
[(parā ui)]ya-mi
out send-1SG.PRS
‘[(And if)] I, My Majesty, [(se)]nd you [(out)] for som⟨e⟩ matter, …’.
Here omission of the KI sign did not produce the bare interrogative as the bare
interrogative is kuedani, not *kuedanik.
Turning back to OH bare interrogatives, another case which has been
claimed to be an OH bare interrogative occurs in a MS text:
CHD (L-N: 432–433) analyzes the context as “(1) I haven’t done anything against
you, my god, have I, (2) or sinned in some way, have I?”, followed by Hoffner
(1995: 92) and Huggard (2015: 35). However, the context has been convincingly
analyzed in an alternative way by Singer (2002: 35): “What have I ever done to
you and how have I sinned?”.
Bare interrogatives are never attested in NS copies and even later versions of
OH Laws. The statistics is impressive enough not to be just a matter of coinci-
dence.
All the rest of Anatolian languages pattern with OH/OS usage, i.e. they attest
NPIs and not bare interrogatives in conditional clauses and after negation,
even though the data are severely limited. See generally for the system of Ana-
tolian indefinite pronouns Sideltsev & Yakubovich (2016). Here we will review
only the evidence directly relevant for bare interrogatives.
According to the Luwian corpus of Yakubovich, Hieroglyphic Luwian attests
only one case of indefinite in a conditional clause. This is an NPI, and not a
bare interrogative:
Cuneiform Luwian attests more cases (Melchert 1993: 119), but all of them
involve NPIs:
(12) pre-NH/NS (CTH 760.II.1.A) KUB 35.43+ obv. ii 7, cf. Starke 1989: 144.
mān=ata īššar-ati kuiḫa
if=it hand-ABL something.NOM.SG.N
‘If it is something from the hand’.
4 Following Yakubovich 2015: 46. Cf. Hawkins 2000: 56: “If anyone from (among) kings”. The
indefinite pronoun translates Phoenician mlk ‘a king’ (Hawkins 2000: 66).
Only NPIs are used after negations in both Cuneiform and Hieroglyphic
Luwian (Melchert 2003: 204), as in KUB 35.79 rev. iv 13′ nawa kuiḫa. For Hiero-
glyphic Luwian see
5 Or =me( j)=.
6 Following Garrett 1990: 241 fn. 1; Melchert 2004: 19. Cf. Kloekhorst 2013: 149.
7 Or =me( j)=.
OH/OS data as well as the data from other Anatolian languages as inciden-
tal, the development from Proto-Indo-European would be narrowing of the
sphere of usage of bare interrogatives because the usage of bare interroga-
tives in narrow IE is considerably wider than that of Hittite. As shown by Haug
(2016), narrow PIE had several other contexts, besides conditional and neg-
ative clauses, where bare interrogatives were licenced. The most common of
them was questions, less common modal contexts. None of these attests any
bare interrogatives in Hittite. Even the use of bare interrogatives under nega-
tion is severely restricted in Hittite to a handful of cases. Besides (2, 7) above,
the only other attestations we are aware of are NH/NS KUB 21.38 obv. 48,
ATT 35, 12 and MH/MS HKM 52 rev. 35. The only sphere where bare interrog-
atives are consistently licenced in post-OH time is conditionals, both proto-
typical conditional clauses with overt subordinators as (1) above, condition-
als introduced by našma ‘or (if)’, see Huggard (2015: 34–35), and paratactic
conditionals, for which see generally Hoffner & Melchert (2008: 423), such
as:
(18) MS/MH (CTH 261.II) KUB 26.17 obv. i 5′–6′, cf. Miller 2013: 130–131.
1. māḫḫan=[ma] LÚKÚR-aš ak-i
when=but enemy-NOM.SG.C die-3SG.PRS
2. kūrur kui-š ḫar-zi
hostility.ACC.SG.N which-NOM.SG.C have-3SG.PRS
‘(1) As soon as the enemy has been vanquished, [though] (2) (if ) some
(enemy) retains hostility (then the occupation contingent that (is to
be left behind) I will leave behind for the occupation)’.
All the rest of licencing contexts for bare interrogatives in Hittite are closely
connected with conditionals. This would contradict Haspelmath’s (1997) gen-
eralization that bare interrogatives become more general in use over time as
indefinite pronouns. Thus even in this case it agrees better with the material to
speak of independent innovations in historical Hittite and in narrow PIE.
(19) MS/MH (CTH 261.II) KUB 26.17 obv. i 4′, cf. Miller 2013: 130–131.
mān DUTU–Š=I=ma kuwapi apāšila laḫḫiyai-zzi
when Majesty=My=but when himself go.on.campaign-3SG.PRS
‘When His Majesty himself, though, at any time goes on a campaign ([in
the land] of the e[nemy] there too, preparedness must obtain …)’.
(20) MH/NS (CTH 264.A) KUB 13.4 rev. iii 68, cf. Miller 2013: 260–261.
mānn=a=za MUNUS-i kui-š GAM-an šeš-zi
when=and=REFL woman who-NOM.SG.C down sleep-3SG.PRS
‘Also, when someone goes to sleep with a woman, …’.
Examples (19–20) are obvious off-shots from the common conditional licenc-
ing contexts. They are introduced by the same subordinator which in Middle
and New Hittite dominates in conditional clauses—mān, even though in (19–
20) it is used in a purely temporal meaning—‘when’. Thus (20) cooccurs in the
same text with the following prototypical conditional use of mān licencing a
bare interrogative:
(21) MH/NS (CTH 264.A) KUB 13.4 rev. iii 74, cf. Miller 2013: 260–261.
mān=ma=za ITTI MUNUS-TI kui-š šeš-zi
when=but=REFL with woman who-NOM.SG.C sleep-3SG.PRS
‘If, however, someone sleeps with a woman, …’.
ing semantics can be seen in (22a–c) which attest further analogical spread:
there bare interrogatives are licenced in Hittite by a purely temporal subordi-
nator māḫḫan:
(22) a. NH/NS (CTH 255.2.A) KUB 26.1+ rev. iii 50–51, cf. Miller 2013: 302–303.
[(š)]ummaš=ma kuit GIM-an išdamaš-ten
you=but which-ACC.SG.N when hear-2PL.PST
‘But when you have heard something, …’.
b. NH/NS (CTH 255.2.A) KBo 26.1+ rev. iii 16, cf. Miller 2013: 300–301.
nu=za kuit GIM-an kiš-ari
CONN=REFL which-ACC.SG.N when hear-3SG.PST
‘And should something happen …’.8
c. NH/NS (CTH 255.2.B) KBo 26.8 obv. ii 5′, cf. Miller 2013: 298–299.
[…]x=ma kuit GIM-an *u-š*ke-tteni
x=but which-ACC.SG.N when see-IMPF-2PL.PRS
‘But when you observe something, …’.
Actually, in these examples only the logographic writing GIM-an is attested and
the phonetic complement is compatible with both māḫḫan and mān. It is also
curious that in (22b) the context requires that the meaning of māḫḫan should
not be the common ‘when’, but quite unexpected and, according to CHD (L-N:
100), otherwise unattested ‘if’! In this light (22a–c) may simply attest writing
of mān with the Sumerogram GIM. Such confusion of mān and māḫḫan is
attested in other meanings, see CHD L-N: 146. But it is also possible to think that
(22a–c) may attest the analogy of māḫḫan after mān: māḫḫan does not have a
conditional meaning, thus it licences bare interrogatives by analogy after mān
which has both temporal and conditional meanings ‘when’ and ‘if’. The analogy
is all the more expected, as the same text KUB 26.1+ employs in obv. i 17 mān ‘if’
licencing a bare interrogative kuwapi ‘ever’. The word order is more compatible
with this hypothesis, although clause internal mān is also attested, see Sideltsev
(2015). Independent support for (22a–c) involving genuine analogy after mān
and not mān itself comes from the following example from the same period as
(22a–c)—the time of Tudḫaliya IV—which attests phonetic writing of māḫḫan
‘when’ licencing a bare interrogative:
8 The case belongs here if one follows Miller (2013: 300–301). Cf. CHD L-N: 102: “with regard to
(lit. like) what happens”. According to CHD, the clause rather contains complex subordinator
kuit māḫḫan.
The context is translated as: “Als aber mein Vater dann starb” (Otten 1988: 18–
19) or “And when my father died” (Beckman 1996: 112). Otten (1988) assesses
in his glossary kuwapi as an adverb “(irgend)wo, irgendwann”, thus the literary
English translation should be ‘When my father died at some point’. Under this
reading kuwapi is bare interrogative.
Example (23) is very obviously not generalizing and unparalleled by nar-
row Indo-European languages. Thus the first spread of licencing contexts for
bare interrogatives is inner-Hittite and unconnected with narrow PIE spread
of licencing contexts for bare interrogatives.
The second spread is also typically Hittite. It was observed in the previous
section that bare interrogatives are often licensed after našma ‘or (if)’. Sporad-
ically the licencing is extended to the contexts where našma means just ‘or’ or
even to the contexts where another conjunction ‘or’, naššu, is used:
(24) NH/NS (CTH 261.I.B) KUB 13.2+ obv. ii 28′–30′, cf. Miller 2013: 226–227.
kēdani=wa=ššan URU–r-i naššu ŠA D10
this.LOC.SG=QUOT=LOC city-LOC.SG either of Storm.god
kuit⟨(ki)⟩ Ékarimmi našma tamēdaš DINGIR–LÌ-aš kuitki
some temple or other.GEN.SG god-GEN.SG some
Ékarimmi
temple
‘In this town there is either a temple of som⟨(e)⟩ Storm God or a
temple of some other deity (it is now neglected, and it is dilapidat-
ed).’
Actually, the form kuit in (24) is assessed by the editor as an error for kuit⟨ki⟩,
but due to the difficulty in distinguishing between errors and analogy it is
impossible to prefer one of the two options.
Another extention clearly traceable to the conditional use is that of relative
clauses:
(25) MH/MS (CTH 257.4.A) KUB 31.100 rev. 11′–13′, cf. CHD L-N: 388.
1. namma kui-ēš kui-ēš kueluwan-eš šer
further what-NOM.PL.C what-NOM.PL.C k.-NOM.PL.C up
É.G[AL?]
palace
2. kui-ēš namma kui-ēš kuwapi
what-NOM.PL.C further what-NOM.PL.C where
3. n=uš ḫūmand-u[š] wanalli-šk-andu ištalki-šk-and[u]
CONN=them all-ACC.PL.C w.-IMPF-3PL.IMP i.-IMPF-3PL.IMP
‘(1) Further, whatever k.’s are up in the pa[lace(?)] (2) (and) whatever
other k.’s are (any)where, (3) all those let them w. and i.’
(26) MH/NS (CTH 264.A) KUB 13.4 rev. iii 21–23, cf. Miller 2013: 256–257.
1. URUḪat⟨tu⟩š-i=ma=kan kuedani kui-š
Hattusa-LOC.SG=but=LOC who.DAT.SG which-NOM.SG.C
šaklāi-š šer
duty-NOM.SG.C up
2. mān LÚSANGA LÚGUDU₁₂ LÚ.MEŠḫaliyattallēš
3. kui[šš=a=aš] tarniškezzi
‘(1) He who is responsible for letting in someone who has some duty
up in Ḫat⟨tu⟩sa, though—(2) be he a priest, an anointed one (or) the
watchmen—(3) he must let only them in’.
The use easily falls into the extention of the original conditional usage: gener-
alizing relatives are very close to conditionals, see Garrett 1994: 44–45; Huggard
2015: 34. The connection between generalizing relatives and conditionals is so
trivial that generalizing relative clauses licence bare interrogatives in Greek and
Latin (Haug 2016). What does not seem to be attested in narrow PIE, however,
is that bare interrogatives are also sporadically attested in determinate relative
clauses, as in the following Hittite example:
(27) NH/NS (CTH 81.A) KUB 1.1(+) rev. iv 19–20, cf. Otten 1981: 24–25; van den
Hout 2003: 203.
nu mUrḫi-D[(U-u)]pa-š BELU ḪI.A kui-ēš kuwapi arḫa
CONN Urhitesub-NOM.SG.C lords who-ACC.PL.C where away
[(u)]iya-t
send-3SG.PST
‘To the generals whom Urḫitešub had dismissed to some place (Ištar
appeared in a dream)’.
tied in with the peculiarity of Hittite syntax: in Hittite bare interrogatives are
licenced by mān both in temporal and conditional functions ‘if’ and ‘when’ vs
only conditional clauses in narrow PIE; they are licenced in both indetermi-
nate and determinate Hittite relative clauses vs. only in generalizing relative
clauses in Greek and Latin.
Yet another indication that Hittite bare interrogatives are a purely inner
Hittite innovation comes from word order issues. Hittite attests syntactic dif-
ference between wh-words and indefinite pronouns: the former do not attest
constraint on the clause position (they can be clause first, second and prever-
bal) and on the position within DP whereas the latter are normally second, both
within the clause if they are in the left periphery and within the QP, see Sidelt-
sev (2015), cf. Huggard (2015). Bare interrogatives behave mostly like indefinite
pronouns, i.e. they are syntactically different from wh-words. However, in a
number of cases they retain the features typical of wh-words, i.e. they are clause
first:
(28) MH/NS (CTH 258.2) KUB 13.7 obv. 9, cf. Miller 2013: 140–141.
mān=an=za kuwapi=ma appezzian LUGAL–u-š EGIR-an
if=him=REFL when=but later king-NOM.SG.C back
kappūē-zzi
count-3SG.PRS
‘If, however, the king reassesses him (i.e., his case) at some point’.
The system is in stark contrast to the Greek or Latin systems of bare interroga-
tives, see Sideltsev, Molina, Belov (2015), cf. Huggard (2015).
(29) NH/NS (CTH 293) KUB 13.35+ obv. ii 6–8, cf. Werner 1967: 6–7.
1. tamēdani=ma=wa kuedani ANŠE.GÌR.NUN.NA
other.DAT.SG=but=QUOT which.DAT.SG mule
šarnikzil-aš EGIR–pa pe-ḫḫi
compensation-GEN.SG back give-1SG.PRS
2. nu=wa=šmaš SIG5–and-uš UL=pat pe-ḫḫi
CONN=QUOT=them good-ACC.PL.C NEG=EMPH give-1SG.PRS
‘(1) The other (person) to whom I give mules of compensation, (2) I do
not give the good ones to them under no circumstances’.
Hoffner (2003: 58) following Werner (1967: 6–7) translates the two last
clauses as ‘(1) To the other person I give mules of compensation, (2) but under
no circumstances do I give the good ones to them’. In this understanding
kuedani is used as an indefinite pronoun not licensed by anything. However,
there is no compelling reason to assess the context in the way Hoffner does. It
can be understand equally well as ‘(1) The other (person) to whom I give mules
of compensation, (2) I do not give the good ones to them under no circum-
stances’.
It has been suggested that bare interrogatives are licenced in Hittite in ques-
tions. But the two contexts that have been listed as evidence have also been
analyzed in a different way and cannot serve as unambiguous evidence.
Thus Miller (2013: 107) assesses the following subordinator as a bare inter-
rogative:
But Güterbock & van den Hout (1991: 11) assess it as a genuine subordinator:
“How will you see him?”, followed, e.g., by Brosch (2014: 239). The other context
which has been assessed as bare interrogative in a question, (9) above, cannot
serve as unambiguous evidence as it can contain a negation marker, or it can
easily be analyzed in a way not involving bare interrogatives.
Other contexts which have been assessed as bare interrogatives not licensed
by anything are even less probative. The absolute majority of them are in frag-
mentary contexts and the pronouns can be assessed as relative. Thus kuit in
NH/NS (CTH 590) KUB 15.30 rev. iii 8′–9′ was analyzed as “something” by de
Roos (2007: 198–199), but as relative “ce que” by Mouton (2007: 287). Simi-
larly divergent analyses have been offered for kuwapi in NH/NS (CTH 584.5)
KUB 48.126+ obv. i 44 (de Roos 2007: 126–127; Mouton 2007: 295). Only kuwapi
in NH/NS (CTH 584.7) KUB 48.118 10 is unanimously analyzed (de Roos 2007:
123–124; Mouton 2007: 272), but it is almost fully restored [kuwa]pi. Kuwapi in
KUB 48.126+ obv. i 31 which is translated as “at a certain moment” (de Roos
2007: 127; Mouton 2007: 295) occurs in a fragmentary context.
So, the only Hittite context licencing bare interrogatives which produces the
impression of being a real linguistic phenomenon and not an occasional error
of the scribe/copyist is conditional clauses.
An important part of the argument above was that Anatolian languages other
than Hittite do not attest any bare interrogatives. However, there are two forms
which have been expressly identified with the Middle/New Hittite bare inter-
rogatives. Now we will assess them in detail.
(31) TL 59 2–3
me=i( j)=a-di tike: ti-he
CONN=him/her=do-3SG.PRS someone.NOM.SG who-GEN.SG
zum̃ m[ẽ/ã]
harm.ACC.SG
‘(If) anyone does harm of any kind to him/her’.
(32) NH/NS (CTH 106.B.2) KBo 4.10+ rev. 18, cf. van den Hout 1995: 46–47; Beck-
man 1996: 107.
mān URULUM kui-š našma AŠRU kuitki ANA
if city which-NOM.SG.C or place some.NOM.SG.N to
mUlmi–DU–up LUGAL KUR URU.DU–tašša piy-anna UL
Ulmi-Tessup king land Tarhuntassa give-INF NEG
ZI–anza
soul.NOM.SG.C
‘If he does not wish to give some city or some locality to Ulmi-Tessup, …’.
However, Lycian tihe attests several peculiarities. As is seen in (31), it never func-
tions as synchronic genitive of wh-word ti-, it is rather used in the same contexts
as tike, an accusative form of NPI tike ‘some(one)’. This follows from contexts
which are similar to (31), e.g.:
(33) TL 56 3
se=ije ti e-di: tike: mẽtẽ:
and=him/her who.NOM.SG do-3SG.PRS some.ACC.SG harm.ACC.SG
‘who does any harm to him/her, …’, similar to fragmentary TL 72.
The parallelism of (31) and (33) makes it very likely that despite the difference
in form (genitive ti-he—accusative ti-ke), both forms are identical semantically:
tike zum̃ mẽ ‘any harm’ vs tihe zum̃ mẽ ‘harm of any kind’. A similar reanalysis
of originally genitive forms is attested in ebehi, which, according to Melchert
(2004: 11), was originally an accented possessive adjective of ebe- ‘of this’, but
synchronically equivalent to simply ebe- ‘this’. Thus, as different from Hittite
bare interrogatives, tihe is synchronically part of the paradigm of NPI tike, not
relative ti- and thus synchronically it is not a bare interrogative: bare interrog-
ative implies that the form which is synchronically a wh-word functions in a
certain context as an indefinite. Lycian tihe is synchronically an NPI, not a bare
interrogative. But can it be assessed as a bare interrogative diachronically, i.e.
equated with Hittite kuiš in (32)?
If we assess tihe at a previous stage of development as a bare interrogative,
the absolute majority of contexts (besides the cited TL 59 2–3, TL 91 3, TL 95
2, these are TL 135 2; N 314b 2–3; possibly also with negation in fragmentary
TL 45B 7–8), simultaneously display both indefinite tike and bare interrogative
tihe. Only in one context, TL 44C 17, tihe does not co-occur with tike, but the
context is fragmentary. This sets the Lycian form again apart from its presum-
able Hittite parallels. In Hittite, as clearly different from the Lycian pattern, if
two pronouns are simultaneously attested in the same Hittite clause, they are
either two indefinite pronouns in the same clause or two wh-words. In very
rare cases like (32) above it does attest an indefinite pronoun and a relative one
functioning as indefinite in the same context, but the use is never systematic.
In this light the Lycian recurrent use of ‘relative’ tihe alongside indefinite tike
within the same clause cannot be equated with Hittite (32), but is rather remi-
niscent of Hieroglyphic Luwian where kuisha coccurs with kuis as a compound
free choice item kuisha kuis ‘whoever’, see ex. (15) in Sideltsev & Yakubovich
(2016) repeated here as (34) which is attested alongside kuis kuis with the same
function:
The Luwian example (34) shows that the standard word order within the com-
plex free choice pronoun is indefinite pronoun—relative pronoun. Only once the
order is reversed—in KARKAMIS A6 §25 kwatti kwatti-ha REX-ti ‘to whatever
king’, see Sideltsev & Yakubovich (2016). It is highly instructive that the word
order indefinite pronoun—relative pronoun is also invariably the pattern found
in Lycian, as exemplied by (31): in Lycian it is always the indefinite pronoun tike
which comes first, followed by the would-be relative tihe. As (32) shows, the Hit-
tite word order is conspicuously different. In (32), as in all other Hittite clauses
simultaneously attesting a proper relative and a bare interrogative, the proper
relative precedes a bare interrogative. This is quite expected in the contexts
where the relative licences it, but even in cases when both pronouns are nega-
tive polarity items licenced by a conditional subordinator, it is commonly the
first pronoun which is a bare interrogative and the second is an indefinite of the
kuiški series, including the cases which are assessed as scribal errors, see (24).
This makes all the more likely that Lycian is comparable to Luwian, not Hit-
tite. Actually, there is one Lycian context which establishes the diachronical
identity between the Luwian free choice pattern in (34) and Lycian. In this con-
text tike and tihe do not just co-occur, they both modify the same noun:
9 The particle -ba ‘but’ must be a scribal mistake here as it does not fit the context.
‘The tomb of this (place),10 whoever does any harm of any kind whatso-
ever to it, (he will strike him, the Father of these)’.
C. Melchert (pers. comm.) suggests that the word order noun—tihe in (35) does
not allow our understanding of (35) and forces to analyze tihe not as modify-
ing the noun, but as genitive for dative (!) ‘to anyone’. Yet, none of these ad
hoc assumptions is necessary. Attributes are attested in Lycian both in pre- and
postposition to the head noun, see exx. (15, 31) above and Kloekhorst (2013).
In the majority of contexts, as seen from (31), supported by TL 44C 17, TL 91
2–3, TL 95 2, TL 135 2, tihe is in front of the noun. If the same noun is mod-
ified by tike in the preposition, tihe follows the noun, as in (35). Moreover,
another noun phrase in (35) shows that the word order noun—tihe is compat-
ible with the attributive understanding of tihe: the noun—tihe word order is
exactly paralleled by the word order of χupa ebe-hi ‘this tomb’ with the same
fossilized genitive form of the deictic pronoun ebe. This completely removes
any of Melchert’s arguments.
For us, the fact that Lycian tihe modifies the noun in (35) alongside another
indefinite pronoun tike is a clear indication that Lycian tike tihe is a free choice
item. It is clearly different from the Hittite free choice system where only rela-
tives were employed and it is suspiciously similar to the double marking of free
choice by both relative and specialized indefinite pronoun in the other mem-
bers of the Luwian branch, see (34) above and Sideltsev & Yakubovich (2016).
Thus we believe that Lycian tike tihe in (35) continued the Luwian pattern of
forming free choice items with the help of indefinite and relative pronouns,
even though the exact members of the pattern are different in Lycian and
Luwian: the Lycian indefinite pronoun tike is not an etymological match to
the Luwian indefinite kuisha, for which see Sideltsev & Yakubovich (2016). It
is curious that the other Lycian pronoun which continues the Luwian free
choice pattern, Lycian tise tise, which is an exact etymological match to Luwian
kuisha, see ibid., generalized the indefinite pronoun at the expense of the rel-
ative one. Thus it is tise tise which is an innovation upon the earlier Luwian
pattern indefinite—relative marking free choice. The pattern is preserved intact
in tike tihe. As is expected of an archaism, synchronically in Lycian the form
tihe was fossilized and lost its connection with the paradigm of relative/inter-
rogative ti-. After the older system of free choice pronouns broke down and
tike tihe was being replaced by the newer system of tise tise, the original dis-
tribution was preserved even in the cases where tike and tihe did not form a
constituent like (31). Thus tihe is not a synchronic or diachronic parallel to Mid-
dle/New Hittite bare interrogatives.
kwitta is analyzed as “wherever” by Hawkins (2000: 393), but at face value it pro-
vides an exact equivalent to the Hittite bare interrogatives in relative clauses,
as is seen above in (25). However, if seen in the context of Luwian syntax, it
is obvious that Luwian and Hittite are not identical. To understand the Luwian
context we will have to look at the wider picture of Luwian indefinite pronouns.
The main peculiarity of Luwian is that Luwian indefinite pronouns of the type
kuisha ‘anybody’ are employed not only in conditional clauses and after nega-
tion markers (as negative polarity items), but also where Hittite never employs
its negative polarity items—as a universal quantifier or as free choice items,
both in and out of relative clauses. This is seen in confronting exx. (9, 10, 12)
above where indefinite pronouns function as negative polarity items with (37),
see also exx. (10–16) in Sideltsev & Yakubovich (2016) where they function as
free choice items:
2. wa/i-mu-u ¦VIA-wa/i-ni-’
wa=mu harwanni
PART=me send.2SG.IMP
‘(Further, they (are) eleven baskets, or (if) you do not find(?) them,) (1)
wherever you see ten baskets, (2) send (them) to me’.
Hawkins (2000: 126) qualifies kwiya as obscure, but supposes that it may func-
tion inside a subordinate clause introduced by kuman ‘when’ as indefinite.
However, this still does not produce any sence and thus we follow Yakubovich
in his analysis of the form as ‘when’. Double kuman ‘when’ is attested in
HLuwian in POTOROO 7b. Kwiya is attested as ‘when’ in the same text (KAR-
KAMIŠ A6) in §8, 18.
An analogous analysis is advanced by Yakubovich for other forms which have
been claimed to represent bare interrogatives in Luwian. Hawkins (2000: 547)
observes that in ASSUR letter e §7 kwiya cannot be a relative pronoun intro-
ducing a subordinate clause; he thinks it is “perhaps indefinite” and compares it
with REL-i in ASSUR letter e §10. However, indefinite interpretation produces
a very poor understanding of
5 Conclusion
References
Hoffner, H.A., Jr.; H.C. Melchert 2008. A Grammar of the Hittite Language. Part 1: Refer-
ence Grammar. Winona Lake: Eisenbrauns.
Huggard, M. 2015. Wh-words in Hittite. Ph.D. dissertation, University of California, Los
Angeles.
Kloekhorst, A. 2011. The opening formula of Lycian funerary inscriptions: mẽti vs. mẽne.
Journal of Near Eastern Studies 70, 13–23.
Kloekhorst, A. 2013. Likijskij jazyk. In: Ju.B. Korjakov; A.A. Kibrik (eds.), Jazyki mira.
Reliktovye indoevropejskie jazyki Perednej i Central’noj Azii. Moscow: Academia. 131–
154.
Melchert, H.C. 1988. “Thorn” and “minus” in Hieroglyphic Luvian orthography. Anato-
lian Studies 38, 29–42.
Melchert, H.C. 1989. New Luvo-Lycian isoglosses. Historische Sprachforschung 102/1,
23–45.
Melchert, H.C. 1993. Cuneiform Luvian Lexicon. Chapel Hill.
Melchert, H.C. 1994. Anatolian Historian Phonology. Amsterdam–Atlanta: Rodopi.
Melchert, H.C. 2003. Language. In: H.C. Melchert (ed.), The Luwians. Leiden–Boston:
Brill, 170–210.
Melchert, H.C. 2004. A Dictionary of the Lycian Language. Ann Arbor–New York: Beech
Stave Press.
Melchert, H.C. 2012. Genitive case and possessive adjective in Anatolian. In: V. Orioles
(ed.), Per Roberto Gusmani. Studi in ricordo. Linguistica storica e teorica. Vol. II, tomo
1. Udine: Forum, 273–286.
Miller, J. 2013. Royal Hittite Instructions and Related Administrative Texts. Atlanta: Soci-
ety of Biblical Literature.
Mouton, A. 2007. Rêves hittites. Leiden–Boston: Brill.
Neu, E. 1974. Der Anitta-Text. Wiesbaden: Harrassowitz.
Otten, H. 1981. Die Apologie Hattusilis III: das Bild der Überlieferung. Wiesbaden: Har-
rassowitz.
Rieken, E.; I. Yakubovich 2010. The new values of Luwian signs L 319 and L 172. In:
I. Singer (ed.), Ipamati kistamati pari tumatimis. Luwian and Hittite Studies Presented
to J. David Hawkins on the Occasion of his 70th Birthday. Tel Aviv: Emery and Claire
Yass Publications in Archaeology, 199–219.
Sideltsev, A. 2015. Syntax of Hittite mān ‘if/when’. Aula Orientalis 33/1, 127–140.
Sideltsev, A. 2015. The riddles of Hittite indefinite pronouns. Altorientalische Forschun-
gen 42/2, 199–275.
Sideltsev, A.; M. Molina; A. Belov 2015. Syntax or phonology? Proclitics, enclitics, and
stress in Hittite. Journal of Language Relationship 13/2, 139–168.
Sideltsev, A.; I. Yakubovich 2016. The origin of Lycian indefinite pronouns and its
phonological implications. Münchener Studien zur Sprachwissenschaft 70/1, 75–124.
Singer, I. 2002. Hittite Prayers. Atlanta: Society of Biblical Literature.
1 Introduction
The Indo-Uralic hypothesis is usually credited to the great Danish linguist Vil-
helm Thomsen, not least because of what his no less great compatriot and
student Holger Pedersen had written in his historiographical masterpiece Lin-
guistic Science in the Nineteenth Century (1931: 336–337):
Indeed, Thomsen’s original formulation (1869: 1–2) was so cautious that even
today’s opponents of the Indo-Uralic hypothesis would have very little to dis-
agree about. On the other hand, Pedersen did not explicitly argue that Thomsen
would have been the first to mention the Indo-Uralic hypothesis. Meanwhile,
the Swedish Uralicist Björn Collinder was somewhat more explicit in his semi-
nal work Indo-uralisches Sprachgut (1934: 5):
Der erste, welcher die frage nach der urverwandtschaft der finnisch-
ugrischen sprachfamilie mit der indoeuropäischen auf streng wissen-
schaftlicher grundlage erörtert hat, ist Vilhelm Thomsen.
The fact that by Collinder’s account Thomsen was the first to formulate the
Indo-Uralic hypothesis “on a strong scholarly basis” suggests that Collinder was
well aware of even earlier pioneers whose scholarly basis was just not so strong,
although he did not name any of them in this connection. The present paper
therefore deals with one of these pre-Thomsen Indo-Uralicists who, inciden-
tally, were not yet called Indo-Uralicists since the term Indo-Uralisch itself was
not introduced until much later by another Swede, Hannes Sköld (1927).1 For
the sake of brevity, however, the present article goes on speaking of the Indo-
1 Thus also Collinder (1934: 3): “Der terminus “indo-uralisch” stammt von Sköld” (cf. similarly
Joki 1973: 165; Čop 1975: 2).
Uralic hypothesis rather than the idea of a genetic relationship between the
Indo-European and Uralic language families.
2 Before Indo-Uralic
As Castrén was arguably both the leading Uralicist and the leading Altaicist
during the mid-19th century, no one on the Uralic side dared to risk their reputa-
tion by embracing any other genetic affiliations. At the same time, the achieve-
ments of comparative Uralic linguistics remained more or less unknown on
the Indo-European side. For example, the English philologist Thomas Hewitt
Key (1846) was still comparing Indo-European with Finnish and Saami alone
without even mentioning Hungarian. On the other hand, although the Ger-
man lexicographer Lorenz Diefenbach (1851) was aware of a larger number of
Uralic languages, he was simultaneously comparing Indo-European to Basque,
Turkish, Semitic, Coptic, Malay, and even Polynesian, just to name a few. Thus,
nothing suggests that he supported the Indo-Uralic hypothesis, namely that
Indo-European and Uralic are not only related but also more closely to each
other than to any other language or language family.
Hereby we have reached the year 1853 when a Finnish undergraduate stu-
dent called Daniel Europaeus self-published a pamphlet truly launching the
Indo-Uralic hypothesis. Before moving on to this eccentric work, however, I
must say something about its even more eccentric author virtually unknown
outside his native Finland (see e.g. Forsman 1896; Salminen 1905; Timonen 2003
for his more detailed biographies).
3 Daniel Europaeus
who was active in an amazingly wide range of different scholarly fields with an
equally wide range of success and failure. Today he is no doubt best-known in
Finland as one of the major contributors to the Finnish national epic Kalevala
(1849) whose tragic Kullervo cycle of oral poems was largely collected during his
1847 fieldwork in Ingria. Retrospectively speaking, hardly any piece of Finnish
literature has inspired artists around the world as much as the Kullervo tale (see
e.g. Tolkien 2015).
Still, this was only the tip of the iceberg. Among other things, young Euro-
paeus was also the co-founder of the newspaper Suometar (even today contin-
ued as the online newspaper Uusi Suomi), the translator of the first Finnish-
language geometry textbook (at the time when Finnish geometrical terminol-
ogy was virtually nonexistent), the editor of two folk-poetry anthologies (1847,
1854), and, most of all, the main editor of the first extensive Swedish-Finnish
dictionary (1852–1853) in which he introduced numerous neologisms still in use
in standard Finnish (see e.g. Lehikoinen 1986). Any undergraduate student with
such achievements should really have been celebrated as a Wunderkind, and for
a while this was indeed the case. Soon, however, the Finnish scholarly establish-
ment turned its back on him for personal rather than scholarly reasons, namely
that for an undergraduate student he had allegedly become quite a Besser-
wisser. Finally, the last straw was his critical remark on the Kalevala, something
that was considered treasonous at that time (see e.g. Kuusi & Timonen 1988).
Even today there is an ongoing heated debate on whether Europaeus was
a victim (Sulkunen 2004: 107–118) or a villain (Apo 2009: 9–17). In any event,
before his 1853 Indo-Uralic hypothesis the two leading Uralicists of the era
already privately called him “the dumb Europaeus” (Sjögren apud Ronimus-
Poukka 2005: 235) and “an idiot equal to Gottlund”2 (Castrén apud Apo 2009:
2). By that time Europaeus merely had one publication remotely dealing with
comparative Uralic linguistics, an anonymously published newspaper article
(Suometar 1–8/1847), which still followed Castrén’s Ural-Altaic hypothesis.
Indeed, since Castrén had become a national hero in Finland even before his
premature death in 1852, any scholar questioning Ural-Altaic would have been
ostracized even without already being an outcast like Europaeus who, outra-
geously enough, was also going to replace Ural-Altaic with Indo-Uralic (cf. also
Salminen 1985).3
2 Carl Axel Gottlund was another 19th century Finnish outcast but more deservedly so. Even
though he single-handedly caused the national awakening among the Scandinavian Forest
Finns, he was hopelessly old-fashioned as a linguist even compared to his contemporaries,
and his excessively polemical style was not very helpful either.
3 Yet later in life Europaeus always praised Castrén as his last true friend and supporter (see
4 Indo-Europaeus
especially Europaeus 1871 defending the late Castrén against his critics). Hence, Europaeus
was apparently completely clueless of the fact that the feeling was not mutual.
4 This study described as “a remarkable mixture of intelligent, pertinent insights and indis-
criminate fantasy” (Korhonen 1986: 98) shows Europaeus both at his worst and at his best.
While his exact idea of a Permic and/or Ugric substrate in Finland and adjacent areas was
a failure based on reckless etymological guesswork, his general idea to systematically collect
substrate toponyms in order to reconstruct prehistoric language areas was groundbreaking at
that time. For this reason, he is now justifiably recognized as “one of the founders of Finnish
onomastic studies” (ibid.).
5 Although Europaeus was fully aware of Gyarmathi’s work on Finno-Hungarian, he simply felt
that he still had to offer further evidence, because most Hungarians remained reluctant to
accept any genetic relationship with northern Eurasian Untermenschen, as some of them do
even today (cf. Marácz 2012).
6 In general, Europaeus was no longer considered someone worth mentioning, but from now
on his studies were almost completely ignored in the scholarly literature. Besides, the only
exceptions did not really focus on constructive criticism but personal insults questioning his
mental stability (cf. especially Ahlqvist 1871: 36).
7 None of the earliest Indo-Uralic hypotheses by Europaeus and others (e.g. Wedgwood 1856a,
1856b; Thomsen 1869; Anderson 1879; Köppen 1886; Sweet 1900) included Samoyed, which
throughout the 19th century was connected with Altaic rather than Finno-Ugric. This being
the case, the first Indo-Uralic hypotheses including Samoyed did not emerge until the begin-
ning of the 20th century (e.g. Wiklund 1906; Paasonen 1907).
8 Here Europaeus may have been misquoted since in his later studies this macrofamily usu-
ally included Indo-European, Finno-Ugric, Samoyed, Turko-Tatar, Mongolic, Manchu-Tungus,
Georgian-Abkhaz, East Caucasian, Monosyllabische Sprachen, Malayo-Polynesian, Mela-
nesian-Papuan, Semito-Berberic, Old Egyptian, Basque, Hausa, Kole-Talaing, and Inner Afri-
can (1863, 1869, 1877). Even though he was otherwise constantly reconsidering the subgroup-
ings within his macrofamily, he never abandoned his favourite subgroup formed by Indo-
European and Finno-Ugric.
9 In addition to Tamil-Brahui, Europaeus later classified Japanese, Northeast Asian, American,
and Eskimo as isolated languages or language families (1863, 1869, 1877).
sis to Vilhelm Thomsen as early as 1862, after which these two kindred spirits
remained in touch for almost two decades (Timonen 1988: 240–244).
Europaeus never since extended or improved his 1853 study in Swedish,
but next he self-published largely the same arguments in German as Vorläu-
figer Entwurf über den Urstamm der indoeuropäischen Sprachfamilie und seine
vor-indoeuropäischen Abzweigungen, namentlich die finnisch-ungarische (1863).
This time, however, his pamphlet also included the family tree of the languages
in the world—or at least those known to him. This tree was among the first of
its kind because Schleicher’s first Indo-European tree had just been published,
whereas there were hardly any trees of the other language families involved
(e.g. the first Uralic/Finno-Ugric tree other than his own was not published
until Budenz 1878: 228). On the other hand, Europaeus had also ended up with
perhaps the earliest ever Nostratic/Eurasiatic family tree, which was quite an
achievement considering that his classifications were still mostly based on only
a handful of numerals.
The crucial problem with Europaeus was the fact that while his studies con-
tinually offered promising working hypotheses, he was too impatient to dig any
deeper, but he rather moved on to the next topics ranging from phonetics (1857)
to craniology (1873).10 Thus, his later works, such as Die finnisch-ungarischen
Sprachen und die Urheimath des Menschengeschlechtes (1869), already took
Indo-Uralic for granted and focused on even longer range comparisons in order
to show that the human race originated in Africa rather than Asia. As the Out
of Africa theory did not win the Out of Asia theory until the 20th century
palaeoanthropology, Europaeus was once again harshly ridiculed, and it was
about this time when he earned the nickname Indo-Europaeus africanus. Yet
as late as 1871 he was still planning to finally finish his dissertation apparently
titled Die finnisch-ungarische Sprachfamilie in ihrem verwandtschaftlichen Ver-
hältnisse zu anderen Sprachen, but ultimately nothing materialized (Salminen
1905: 132–133).
Since Europaeus never earned an academic degree, he never got a perma-
nent position either. The fact that his numerous odd jobs included archaeolog-
ical excavations in Olonets, Novgorod, and Tver Governorates caused him even
more trouble in Finland where he was now also considered too Russophilic. In
general, he typically found no academic publishers to print his studies, almost
all of which he was forced to self-publish. As he was always broke, his self-
10 Of course, it would now be easy to laugh at Europaeus who in the typical 19th century
fashion did not hesitate to link language families with cephalic indices (1873). Yet he was
no more silly than all those today who think that molecular genetics could immediately
provide ready solutions to linguistic problems.
publications were pamphlets rather than books. He was not lazy, as confirmed
by the fact that when he finally had a chance to be published in the academic
journal Suomi, he wrote an article of no less than 269 pages (1868–1870). Thus,
his Indo-Uralic hypothesis was not necessarily as poorly-founded as his publi-
cations suggest. Be that as it may, he fatally overestimated the probative force of
numerals in spite of similar recent numeral-based classifications (cf. Janhunen
2000: 60–61) which, however, do not offer much more in the 21st century than
those based on the word for ‘God’ (cf. Scaliger 1610: 119–122).
Europaeus spent much of his last years in Saint Petersburg where his reputa-
tion had always been greater than in Finland, as best evidenced by his 1879 sil-
ver medal from the Imperial Russian Geographical Society. At that time he was
apparently no longer actively following the most recent developments in com-
parative linguistics because even in his personal letters he never triumphantly
mentioned the flattering news that his contributions for Indo-Uralic were
finally acknowledged by the Baltic German philologist Nikolai Anderson in his
Studien zur Vergleichung der ugrofinnischen und indogermanischen Sprachen
(1879: 60–62). In general, Europaeus had by then already ceased to publish any-
thing other than concise newspaper columns. His health was gradually declin-
ing, and finally on 3/15 October 1884 (Julian/Gregorian calendar dates) this lone
drifter, who never married and had no children, passed away in a Saint Peters-
burg hospital for the poor.
5 After Europaeus
Europaeus was not allowed to rest in peace for long, but his remains were soon
dug up from his first grave in Saint Petersburg, and his second funeral service
took place in Helsinki where his tomb still stands. As often happens, he had
more friends now that he was dead than he ever had when he was alive. At first
he was rehabilitated in folkloristics as the savior of the Kullervo poems, but even
then he was patronizingly described as a natural talent whose promising career
tragically stalled due to a mental illness (Niemi 1903: I). True, his behavior was
often erratic and strange, not least because he suffered from epilepsy and stut-
tering. Still, the rumours about his insanity were mostly based on the fact that
in his countless newspaper pieces he openly advocated progressive ideals that
long used to be considered crazy, such as freedom of speech, pacifism, fem-
inism, animal rights, minority rights, racial equality, etc. (see e.g. Halila 1988;
Kuusi & Timonen 1988).
What Europaeus himself valued the most was what everyone else valued
the least, namely his comparative linguistic studies. Even though the rival Ural-
Altaic hypothesis had lost its appeal by the beginning of the 20th century (see
e.g. Ramstedt 1947), the Indo-Uralic hypothesis was still almost a taboo in Fin-
land, and when it was finally first discussed by the Uralicist Heikki Paasonen
(1907), he did not mention his compatriot Europaeus at all. This was certainly
not an accident, but Europaeus was deliberately ignored because of his infamy,
and only the later 20th century Finnish Uralicists (e.g. Itkonen 1966: 161–163;
Joki 1973: 25–27; Korhonen 1986: 96–98) were far enough from him in order to
objectively evaluate his studies almost inaccessible today. Yet the damage was
already done, since outside Finland his role as a founder of comparative Indo-
Uralic linguistics has gone more or less unnoticed.11
Even in Finland Europaeus is only occasionally remembered as “one of
the first supporters of the so-called Indo-Uralic hypothesis” (Korhonen 1986:
97), perhaps because he was also one of its last supporters. Indeed, the Indo-
Europeanist Raimo Anttila (1972: 320–321) is among the very few contemporary
Finns who have ever presented Indo-Uralic in a favourable light, whereas most
Uralicists are either agnostic (e.g. Häkkinen 1996) or sceptical (e.g. Janhunen
1999). Under these circumstances, it is no wonder that Indo-Uralic is not the
first thing that occurs to Finns when they think about Europaeus, but this fact
has not prevented him from finally becoming a hero especially in his home
municipality of Savitaipale where the visitors can find the Europaeus Memo-
rial (since 1970), Europaeus Plaza (since 1985), the Europaeus Society (since
1989), the Europaeus Museum (since 1999), Europaeus School (since 2000),
etc.
Ultimately, Europaeus was a child of his time. His limitations and shortcom-
ings as a scholar were obvious, but then again the same can also be said about
almost any other pre-Neogrammarian comparative linguist. Besides, what he
lacked as a scholar, he made up for as a visionary. Consider his three all-
time favourite comparative linguists: first Bopp, then Schleicher, and finally
Thomsen (Salminen 1905: 139). As Europaeus and Thomsen also remained pen-
friends throughout the 1860s and the 1870s,12 we can trace a straight line from
11 Over the past decades, there have been only a few exceptions, such as the Hungarian Urali-
cist Károly Rédei (1986: 7): “Bahnbrecher der Theorie der indouralischen Verwandtschaft
war Europaeus (1853).”
12 The year 1869 was particularly eventful and memorable since first on 23 March Thomsen
defended his epoch-making doctoral dissertation (1869), and then as soon as 14–17 July
Europaeus published its raving review (Finlands Allmänna Tidning 159–162/1869), even
though the latter was at first refused for publication for over two months (Salminen 1905:
139–140). Still, Europaeus was in time to become the first Finnish scholar to review Thom-
sen’s masterpiece, and for an oddly long time he was also the only Finn to truly realize
its value (Aalto 1987: 155–157). On the other hand, Thomsen despite his no-nonsense
References
Aalto, P. 1987. Modern Language Studies in Finland 1828–1918. Helsinki: Societas Scien-
tiarum Fennica.
Ahlqvist, A. 1871. Wäittelyä. Kieletär 1, 25–36.
Anderson, N. 1879. Studien zur Vergleichung der ugrofinnischen und indogermanischen
Sprachen. Tartu: Heinrich Laakmann.
Anttila, R. 1972. An Introduction to Historical and Comparative Linguistics. New York:
Macmillan.
Apo, S. 2009. Uusin Lönnrot-myytti ja kansanrunouden editiohistoria. Elore 16, 1–21.
Blažek, V. 1999. Numerals: Comparative-etymological Analyses of Numeral Systems and
their Implications. Brno: Masarykova universita.
Bopp, F. 1841. Über die Verwandtschaft der malayisch-polynesischen Sprachen mit den
indisch-europäischen. Berlin: Ferdinand Dümmler.
Bopp, F. 1847. Die kaukasischen Glieder des indoeuropäischen Sprachstammes. Berlin:
Ferdinand Dümmler.
Boxhorn, M.Z. van 1647. Antwoord van Marcus Zuerius van Boxhorn, gegeven op de
Vraaghen, hem voorgestelt over de Bediedinge van de afgodinne Nehalennia, onlancx
uytghegeven. Leiden: Willem Christiaens van der Boxe.
Budenz, J. 1878. Ueber die Verzweigung der ugrischen Sprachen. Beiträge zur Kunde der
indogermanischen Sprachen 4, 192–258.
Castrén, M.A. 1849–1862. Nordische Reisen und Forschungen. St. Petersburg: Impera-
torskaja Sankt-Peterburgskaja Akademija Nauk.
Collinder, B. 1934. Indo-uralisches Sprachgut: die Urverwandtschaft zwischen der indoeu-
ropäischen und der uralischen ( finnischugrisch-samojedischen) Sprachfamilie. Upp-
sala: A.–B. Lundequistska.
Čop, B. 1975. Die indogermanische Deklination im Lichte der indouralischen vergleichen-
den Grammatik. Ljubljana: SAZU.
Diefenbach, L. 1851. Vergleichendes Wörterbuch der gotischen Sprache. Frankfurt am
Main: J. Baer.
approach to historical linguistics never ignored Europaeus the same way the others did
(see e.g. Thomsen 1869: 8, 72, 143; 1890: 24, 38, 231).
13 The year of publication, not given on the pamphlet itself, was almost certainly 1869 when
Europaeus donated one copy to the Finnish Literature Society (Slöör 1872: 220). Note
also that in the text he referred to the Zeitschrift für vergleichende Sprachforschung 17
(1868).
Timonen, S. 1988. D.E.D. Europaeuksen kirjeitä vuosilta 1846–1882. In: M. Kuusi; P. Laak-
sonen; S. Timonen (eds.), D.E.D. Europaeus: Suurmies vai kummajainen. Helsinki:
Suomalaisen Kirjallisuuden Seura, 204–264.
Timonen, S. 2003. Europaeus, David Emanuel Daniel (1820–1884). Suomen kansallis-
biografia 2, 694–696.
Tolkien, J.R.R. 2015. The Story of Kullervo. London: HarperCollins.
Väänänen, K. 2003. Europaeus, Peter Adolf (1753–1825). Suomen kansallisbiografia 2,
696–698.
Wedgwood, H. 1856a. On the connexion of the Finn and Lapp with the other European
languages. Transactions of the Philological Society 1856, 1–18.
Wedgwood, H. 1856b. Further observations on the connexion of the Finnish and Indo-
Germanic classes of languages. Transactions of the Philological Society 1856, 172–179.
Wiklund, K.B. 1906. Finnisch-ugrisch und Indogermanisch. Le Monde Oriental 1, 43–65.
Simona Klemenčič
1 Vowels
Čop (1970c: 51) outlined the following postulated correspondences between the
Proto-Uralic and Proto-Indo-European vowels:
Proto-Uralic Proto-Indo-European
ä, e, i, ü e
a, o, u a, e
õ, y e
In the second syllable of bisyllabic roots, only e and a would have occurred
in Indo-Uralic, whereas the possibilities in the first syllable are based on the
comparatively rich vowel system postulated for the Uralic proto-language.
When one sees reconstructed phonemic or morphological inventories multi-
ply, Occam’s razor comes to mind, but there is no reason not to believe Čop’s
Indo-Uralic vowel system was possible.
There is a considerable gap between his postulated early Proto-Indo-Euro-
pean vowel system and what most of us tend to believe about it. Čop addresses
this problem in Indouralica XVII (Čop 1990: 24) and says that “la simplicité
suposée du côté indo-européen n’est qu’une illusion”. He maintains that the orig-
inal Indo-Uralic long vowels, still preserved in Uralic, were shortened in Proto-
Indo-European and subsequently a new system was established. We know from
Proto-Slavic that this is typologically entirely possible. One wonders neverthe-
less: could not a good explanation of the laryngeal theory type be offered on
the Uralic side to reduce its number of vowels to an originally less rich system?
Čop answers this question by pointing out that the Uralic system’s antiquity can
be confirmed by comparing it to the Altaic vowel system (Čop 1990: 25), but he
does not elaborate on this. He says specifically that the Uralic system cannot be
explained by secondary developments—neither by ablaut (apophony) nor by
laryngeal-like sounds in the vicinity of vowels. He offers a two-part explanation
for the simplification in Proto-Indo-European:
1) The Proto-Indo-European vowel system in its earliest phase underwent
changes in vowel height: high (closed) vowels became mid vowels and
mid vowels became low (open).
2) A Uralic root is always longer than the corresponding Proto-Indo-Euro-
pean root by one vowel, because this vowel was lost in Proto-Indo-
European. The Uralic root will always be disyllabic whereas the Proto-
Indo-European root will normally be monosyllabic.
bisyllabic monosyllabic
aśe- ‘to be located, to dwell; place, spot’ es- ‘be’
Uralic Proto-Indo-European
These rules only apply in the first syllable. In the second syllable of their disyl-
labic roots, Indo-Uralic had only e and a (elsewhere Čop nevertheless recon-
structs IU roots with second syllable vowels other than e and a).
There are three questions to be asked here:
– did Proto-Indo-European have a vowel *a?
– how do we then explain Proto-Indo-European roots with an *o (e.g., *bholo-,
Čop 1978: 151)?
– how does this relate to the Proto-Indo-European laryngeals?
As far as Proto-Indo-European *a is concerned, most Indo-Europeanists are
quite happy to accept that it was rare. But if we were to accept Čop’s view, this
would have to change—meaning that not as many late Proto-Indo-European
*a-s were reflexes of the sequence *h₂e as we now believe. Some of them would
be plain *a-s even before Proto-Indo-European became an autonomous unit.
As for problems with occurrences of Hittite ḫ in the vicinity of these a-s that we
believe to be reflexes of Proto-Indo-European *h₂, Čop maintains (in Indoural-
ica XVII, Čop 1990: 26) that if there were laryngeals in the Proto-Indo-Uralic
root in front of the vowel in question, then special rules apply to the Proto-
Indo-European vowel development. However, he did not seem to believe that
there was a laryngeal in these particular roots.
Here is an example of a root with a Proto-Indo-European laryngeal accord-
ing to Čop (how did he decide which Proto-Indo-Uralic roots had a laryngeal
and which did not? I cannot answer this question at this point of my research
nor can I answer whence the laryngeal would have come):
Uralic *ala ‘under’ (Finnish ala ‘area, territory’, ala-, ali- ‘sub-, lower’,
Saami vuolle ‘lower, under’ etc.) : Proto-Indo-European *Hal- ‘deep’, found
in Hittite hallu(u̯ a)- ‘deep’, Latin alveus ‘hollow, cavity, deep vessel’
1978: 147) by adding an observation that e remains unchanged “wenn der Ablaut
im Indogermanischen stark entwickelt ist”. He further argued (Indouralica
XII, Čop 1987: 139) that the final vowel was actually only lost in Proto-Indo-
European when the first syllable was stressed. With Proto-Indo-European zero-
grade in the first syllable, stress would fall on the second syllable, that is, the
second vowel, which would thus remain preserved. He sees an instance of this
rule in the above mentioned verb ‘to be’:
According to him, the 3rd person plural goes back to early Proto-Indo-European
*esé-nt- > Proto-Indo-European *sé-nt- (as opposed to *s-més etc.). The end-
ing is therefore a result of a secondary segmentation. This rule is further used
to explain the Proto-Indo-European Genitive-Ablative ending *-es, *-os, where
the vowels would be a match for the Uralic second vowel in bisyllabic roots. The
following example is used to illustrate the postulated rule: Indo-Uralic *ońtt́ á́ -ta
‘from the front’ (cf. Fin. otsa-a part. ‘of forehead’) = Proto-Indo-European *Hant-
és/ós gen.-abl. ‘from, of forehead’—where the correct segmentation should
have been *Hanté-s or *Hantó-s. Čop argues that the fact that Uralic has *-ta
where Proto-Indo-European has his reconstructed *-s speaks further in favour
of this rule. As compelling as this argumentation might seem with pieces of a
puzzle seemingly falling into places, it is still an example of circular reasoning.
Nevertheless, it does offer a possibility to explain in a convincing way how these
Proto-Indo-European endings came into being.
Čop goes into details in his book Die indogermanische Deklination im Lichte
der indouralischen vergleichenden Grammatik. A table from this book is given
in Fig. 6.4.
Čop goes into great detail, but as we know, the fact that there are many
details attached in support of a thesis does not necessarily prove that the con-
clusion of the argument is true.
2 Laryngeals
In Čop 1970c: 7, Čop wrote that he deliberately chose not to be more specific
about “das indogermanische Genus der “Laryngale””. Where there is a reflex in
Hittite we can assume that we are talking about the second laryngeal. Here is
what he proposed and argued in favour of:
k, ɣ H ḫḫ
ɣ H ḫ
PU PIE
Nasals
(cont.)
PU PIE
Liquidae
Uralic PIE
V j /w V V i̯ /u̯
In Čop’s view, the second vowel is missing in PIE (Indouralica II, Čop 1972: 166).
Examples: Finnic ujo—‘shy, modest, silly’: PIE *ai̯- in *ai̯gu̯ h- ‘to be ashamed’;
Finno-Ugric *uwa ‘current’: PIE *au̯ - ‘to flow’ (note how a nominal form on one
side is compared to a verb on the other).
The more common type, according to Čop, begins with a consonant and has
two sub-types:
m/n/r/l/w V j /w V
Examples:
Ural. mVjV : PIE mVi̯-: Finno-Ugric *muja ‘try, taste’: PIE *mei̯-no- ‘opin-
ion’;
Ural. nVjV : PIE nVi̯-: Ugric *naxjax ‘fire, sun’: PIE *nei̯- ‘glow’;
Ural. lVjV : PIE lVi̯-: Uralic *leje ‘juice’: PIE *lēĭ -̯ ‘flow’;
Ural. wVjV : PIE u̯ Vi̯-: Finno-Ugric *wajax- ‘sink’: Proto-Celtic * u̯ ai̯-lo-
‘low’.
j /w V m/n/r/l V
Examples:
Ural. wVnV : PIE u̯ Vn-: Finno-Ugric *wüńäx- ‘needle, thorn’: PIE *u̯ en-
‘hit, injure’;
Ural. wVrV : PIE u̯ Vr-: Finno-Ugric *wara, warta ‘slave, man’: PIE *u̯ er-
‘man’;
Ural. wVlV : PIE u̯ Vl-: Finno-Ugric *walax- ‘white, light’: PIE *u̯ el- ‘see’.
The reliability of these etymologies is, however, questionable. There are many
roots with nasals and liquidae and semi-vowels in many languages. If we would
compare Uralic and, e.g., Old Chinese, we would expect to find a similar num-
ber of possible cognates with resonants. And how can we tell shared vocabu-
lary apart from loanwords (compare Finno-Ugric *wara ‘slave’ and Proto-Indo-
European *u̯ er- ‘man’—semantics would speak in favour of borrowing just as
well)?
building his argument that I do not find equally convincing. Čop talks about
an -m-suffix used for building substantiva denominalia in both language fam-
ilies, extends his reasoning to -men stems by drawing parallels between the
Proto-Indo-European -men and Proto-Uralic -ma, but this is not consistent
with his own theory on vocalic development, so he says that this is “etymo-
logisch teilweise noch dunkel”. A great number of suffixes containing an m is
further discussed in connection with endings of nomina deverbalia, infinitive
endings, other verbal endings, etc. Summarizing, it can be concluded that there
are many suffixes containing an m on both sides. This means that one can sim-
ply pick and choose from those suffixes whichever one one needs to make a
case. Piling up examples this weak does not make for a strong case; a critic
can take the pile of cases apart one by one and dismiss all of them on the
grounds that this is an example of linguistic acrobatics instead of solid argu-
mentation. Besides, if both Proto-Indo-European and Proto-Uralic inherited
core words and suffixes from the mother language, I believe that in this case we
would be entitled to expect a larger-than-coincidental number of words with
matching both roots and suffixes. But then, how much exactly is “larger-than-
coincidental”?
Conclusion
Čop’s views are hesitantly and quietly accepted by Slovene scholars. The aca-
demic Etymological Dictionary of the Slovene Language (volume III, Bezlaj 1977:
301) says, for example, that Slovene spužva ‘sponge’ is related to the Uralic word
for mushroom:
spȗžva (f.) “Spongia officinalis”, adj. spȗžvast, -a. Borrowed from Croat.
spȕžva, which was, through Dalm. Rom. and Lat. spongia “idem” borrowed
from Gr. σπογγιά next to σπόγγος “mushroom” […] < IE *sphong(h)o-
[…], which is related to Uralic *paŋka- “mushroom”, cf. Mordw. paŋgo-,
Mari paŋgō- “idem”, Mansi paŋχ “poisonous mushroom” (Čop, Orbis XII,
16 […]).
By focusing on phoneme clusters and comparing suffixes Čop built a nice, well
thought out and extensive system, the building blocks of which fit together.
His work needs to be studied more thoroughly, but my present conclusion is
that the limited number of correspondences in the core vocabulary remains
the main problem of Indo-Uralic comparative linguistics. The basic and seem-
ingly convincing proposed Indo-Uralic roots and suffixes that Čop’s phonetic
system is built on have been assigned meanings that are rather vague and each
of them can be applied to mean almost anything: ‘tip’, ‘self’, ‘split’, ‘bind’, ‘flow’,
etc., whereas many of those that extend beyond this vague vocabulary do not
really fit into the phonetic system. This is like fishing in a sea with an abun-
dance of fish, where one can always count on catching one’s dinner. My fear
is that if we took any other language family instead of Uralic and applied the
same method, we could come up with a system just as elaborate and convinc-
ing and—I suspect—having roughly the same number of correspondences.
Čop’s reconstruction is too random and his method allows for cherry picking
too much. Therefore, I do not find it convincing.
Some of the examples in favour of a genetic relationship between Proto-
Uralic and Proto-Indo-European are nevertheless too good to ignore. When we
leave out those proposed correspondences that are unconvincing for the vari-
ous reasons stated above, we are left with the usually cited and very convincing
list of correspondences (see Helimski 2001). Čop’s approach, as methodical as it
may be, seems to have been doomed from the start. It is very much like compar-
ing Slovene and Danish. We need the Proto-Uralic and Proto-Indo-European
system to fit into a bigger picture. Whether or not this can be done at all is a
question that we have not been able to answer so far.
References
Bezlaj, F. 1977–2007. Etimološki slovar slovenskega jezika I–V. Ljubljana: SAZU ZRC.
Čop, B. 1970a. Indouralica VII. Zeitschrift für vergleichende Sprachforschung 84, 151–174.
Čop, B. 1970b. Indouralica XIV. Orbis 19/2, 282–323.
Čop, B. 1970c. Die indouralische Sprachverwandtschaft und die indogermanische Laryn-
galtheorie: Indouralsko jezikovno sorodstvo in indoevropska laringalna teorija. Ljubl-
jana: SAZU.
Čop, B. 1972. Indouralica II. Ural-Altaische Jährbucher 44, 162–178.
Čop, B. 1973a. Indouralica IV. Linguistica 13, 116–190.
Čop, B. 1973b. Indouralica XVI. Orbis 22/1, 5–42.
Čop, B. 1974a. Indouralica I. Ljubljana: SAZU.
Čop, B. 1974b. Indouralica VIII. Acta linguistica Academiae Scientarum Hungaricae 24,
87–116.
Čop, B. 1974c. Indouralica XV. Zeitschrift für vergleichende Sprachforschung 88, 41–58.
Čop, B. 1975. Die indogermanische Deklination im Lichte der indouralischen vergleichen-
den Grammatik. Ljubljana: SAZU.
Čop, B. 1978. Indouralica V. Collectanea Indoeuropaea 1. Ljubljana: Univerza v Ljubljani,
145–196.
Frederik Kortlandt
For an earlier stage I have proposed the following reconstruction (2010: 400 f.):
Moreover, I have argued that the Hittite ḫi-flexion comprises original perfects,
new perfects created on the basis of derived presents, and transitive zero grade
thematic formations such as Vedic tudáti ‘thrusts’. This merger obliterated the
semantic distinction between the original intransitive perfects and transitive
verbs in the Hittite ḫi-flexion and similarly between the 3rd sg. endings *-o and
*-to in the stative and the middle. As a result, the original distribution can no
longer be established on the basis of the Hittite evidence. At the earliest recon-
structible stage we expect e-grade of the root in the stative but zero grade before
the ending *-e in the perfect. If the apophonic alternation between e- and zero
grade was still automatic at the stage when the new 1st and 2nd pl. endings
*-medhq₂ and *-(t)dhq₂ue were introduced into the stative paradigm (stage A),
the new forms must have had zero grade in the root. The original 3rd sg. sta-
tive ending *-o arose phonetically from lowering of Indo-Uralic *-u (stage B).
The paradigmatic alternation between full and zero grade was then evidently
introduced from the stative into the perfect at a stage when the alternation
between stressed *e and unstressed *o was automatic (stage C). The stress was
eventually retracted in the singular forms of the perfect when stressed *o and
unstressed *e had become possible (stages D and E), probably on the analogy of
the athematic present and injunctive, which had root stress in the singular but
not in the plural. The rise of lengthened grade in the 3rd pl. ending *-ēr < *-er
was most recent (stage F). These developments can be summarized as follows:
It is clear that e-grade thematic presents such as *bhere cannot have arisen
before stage E, when both the root and the ending could have a full grade vowel
*e. Like the perfect, the original thematic conjugation had a zero grade root and
could only obtain an o-grade root vowel after stage C (e.g. Latin molō ‘to mill’).
In my view, the thematic present was originally an impersonal verb form with
a dative subject (cf. Kortlandt 2010: 101–103). The ending was *-e in the singular
and *-o < *-u [ǝw] in the plural, reflecting the Indo-Uralic demonstrative and
reflexive pronouns, respectively (see Kortlandt 2010: 399–403 for the develop-
ment of the endings). This differentiation may be compared with Dutch Het
wordt geregeld ‘It is arranged’, which implies that someone arranges something,
versus Er wordt gedanst ‘There is dancing’, which means that people are danc-
ing. A partial addition of the perfect endings yielded a full paradigm at stage D
(when both formations still had a dative subject) and the replacement of these
by the athematic secondary endings gave rise to a transitive thematic injunctive
The replacement of *q₂ by *q₁ in the thematic present can be explained by the
neutralization of the laryngeals before and after *o into a glottal stop *q₁ (cf.
Kortlandt 2010: 365–368 and passim), which was subsequently generalized in
the paradigm.
Most scholars have accepted Stang’s derivation of o-grade presents such as
Lith. kálti ‘to forge’, málti ‘to grind’, OCS bosti ‘to stab’, Latin fodiō ‘I dig’ from
the reduplicated intensive exemplified in Vedic jaṅghanti ‘strikes’ (1942: 41 f.,
cf. Kortlandt 2010: 216). Jasanoff’s alternative proposal to posit an alternating
paradigm with *o in the singular and *e in the plural (e.g. 1979, 2003) cannot
be maintained because no such paradigm can be reconstructed (cf. Kloekhorst
2008: 142f., Peyrot 2013: 497, Kortlandt 2015). In his lucid treatment of the prob-
lem, Kümmel observes that the Indo-Iranian cognates of proposed o-grade
presents are thematic, e.g. Vedic sphuráti ‘jerks’, tudáti ‘thrusts’, vijáte ‘trembles’
(2004: 150), and the same holds for Latin and Germanic. In my view, the Hittite
ḫi-verbs represent a merger of the original perfect and the original thematic
flexion with zero grade in the root, e.g. Vedic tudáti (Kortlandt 2010: 373–382,
where “q-” has unfortunately been printed as “-q” throughout the chapter). If
the root vowel of CeC-roots was introduced in this formation between stages
C and E of my chronology, it automatically became *o, as happened in the sin-
gular forms of the perfect. Thus, we arrive at o-grade in Slavic bosti ‘to stab’,
kopati ‘to dig’, kosnǫti sę ‘to touch’, kovati ‘to forge’ and the Germanic 6th class
verbs versus zero grade in the Vedic 6th class presents adduced above. The
new pattern could easily be extended to CeRC-roots. The athematic redupli-
cated intensive is evidently a derivative of this formation, e.g. Vedic jaṅghanti
‘strikes’, dediśam ‘point out’, Greek πορφύρω ‘boil’ (Vedic bhuráti ‘quivers’, jár-
bhurīti ‘sprawls’), Latin susurrō ‘whisper’, Gothic inreiraida ‘quaked’ (cf. Kort-
landt 2010: 237), OLith. barti ‘scolds’ (with loss of reduplication). In Anatolian,
the complementary distribution between o- and zero grade was brought into
line with the paradigmatic alternation of the perfect. In Indo-Iranian, the redu-
plicated intensive similarly adopted the alternation of the root vowel from the
3rd class reduplicated presents but preserved the zero grade root vowel of the
6th class presents in the subjunctive (cf. Schaefer 1994: 35–43).
When the ergative (with an ending *-s) and the absolutive (with a zero end-
ing) merged into a new nominative case, the old syntactic system broke down
and the original construction of the thematic present survived only in such
instances as English me dreamed a strange dream and German mir träumt,
which were eventually replaced by I dreamed and ich träume. The idea that
the thematic vowel was coreferential with an additional object in the thematic
injunctive is now supported by Eugen Hill’s analysis of the Indo-Iranian “aorist
presents” (2007). This “instrumental” object (cf. Hill 2007: 293–300) was dis-
tinct from the regular direct object (goal of the action) in the accusative in
*-m, which was a directive case (e.g. Latin ire Romam ‘to go to Rome’). The con-
struction may be reflected in Russian lodku uneslo vetrom ‘the boat (acc.) was
carried away by the wind (inst.)’, ego ubilo svin’ej ‘he (acc.) was killed by a pig
(inst.)’, viz. when it fell on him from a balcony, where the verb is impersonal
and the additional object is in the instrumental case. In this conception, the
original meaning of the thematic present *tude was ‘it (e.g. lightning) strikes
(me)’ or ‘it is a blow (to me)’, with the affected person in the dative, and the
meaning of the derived thematic injunctive *tudet was ‘he strikes (me)’ or ‘he
causes a blow (to me)’, with the agent in the ergative case. If my derivation
of the ending *-e from the Indo-Uralic demonstrative pronoun is correct, the
original structure of *tud-e was ‘it [is] a blow’, where *tud- is a verbal root noun.
The original syntax was apparently preserved in Greek δοκεῖ μοι ‘it seems to me’.
After the separation from Anatolian, the thematic present formation supplied
new presents to athematic injunctives in the other Indo-European languages
(cf. Peyrot 2013: 458 and Kortlandt 2015). When Tocharian had split off, the the-
matic and athematic injunctives yielded imperfects and aorists, respectively,
in the remaining languages and the addition of secondary endings to present
stems supplied new imperfects. The thematic present then became a subjunc-
tive when there was a competing athematic present.
In the Anatolian languages, the endings of the ḫi-conjugation are essentially
the Proto-Indo-European perfect endings. The Hittite preterit endings 3rd sg.
-š and 2nd pl. -šten were evidently taken from the s-injunctive, as they were
in Tocharian (cf. Kortlandt 2014: 83). It is therefore probable that the PIE per-
fect became a past tense in Anatolian at a relatively early stage. It supplied a
preterit to athematic injunctives in the same way as happened in Tocharian (cf.
Kortlandt 2015) and in Latin (e.g. dīxī ‘I said’, lēgī ‘I read’). On the other hand,
the thematic derivations in *-ie/o- and *-ske/o- joined the mi-conjugation in
Anatolian (cf. Kloekhorst 2008: 129–136). The endings of the ḫi-present were
evidently created on the analogy of the mi-present on the basis of the original
perfect. Among the ḫi-presents we expect to find underived thematic presents
and derived presents from athematic formations. There are only three ḫi-verbs
which semantically resemble original perfects:
– nāḫ-, naḫḫ- < *noq₂ei, *nq₂enti ‘to fear, to be(come) afraid, to be respectful,
to be careful’ (Old Irish nár ‘modest’);
– šākk-, šakk- < *sokq₁ei, *skq₁enti ‘to know (about), to experience, to recognize,
to remember’ (Latin secō ‘cut’, sciō ‘know’);
– āppa-, āppi- ‘to be finished, to be done’, which is a derivative of āppa ‘behind’
< *q₂opo (Kloekhorst 2008: 193).
In fact, all underived ḫi-verbs may be o-grade presents, which is also in accor-
dance with their semantics. Following Kloekhorst (2008), we arrive at a list of
48 underived ḫi-verbs in Hittite:
– āk-, akk- < *q(o)k- ‘to die, to be killed, to be eclipsed (of sun and moon)’.
– ār-, ar- < *q₁(o)r- ‘to come (to), to arrive (at)’, Greek ἔρχομαι ‘to come, to go’,
Vedic ṛcháti ‘to go’.
– ārr-, arr- < *q₁(o)rq₁- ‘to wash’, Tocharian A yär- ‘to bathe’.
– ārk-, ark- < *q₃(o)rgh- ‘to mount, to copulate’, Greek ὄρχις ‘testicle’.
– ārk-, ark- < *q₁(o)rk- ‘to cut off, to divide’, Latin (h)ercīscō ‘to divide (an
estate)’.
– au-, u- < *q₂(o)u- ‘to see, to look’, Greek ἀίω ‘to perceive’.
– ḫān-, ḫan- < *q₂(o)n- ‘to draw (liquids)’, Armenian hanem ‘to draw out’.
– ḫarra-, ḫarr- < *q₂(o)rq₃- ‘to grind, to splinter up (wood), to crush (bread)’,
Greek ἀρόω ‘to plough’.
– ḫāš-, ḫašš- < *q₂(o)ms- ‘to give birth (to), to beget, to procreate’.
– ḫāt-, ḫat- < *q₂(o)d- ‘to dry up, to become parched’, Greek ἄζω ‘to dry up’.
– ḫatk- < *q₂(o)dhgh- ‘to shut, to close’, Greek ἄχθομαι ‘to be burdened, to be
depressed’.
– ḫuwapp-, ḫupp- < *q₂u(o)pq₁- ‘to be hostile towards, to do evil against, to hurl,
to throw’, Vedic vápati ‘to strew, to scatter’.
– ḫuwart-, ḫurt- < *q₂u(o)rt- ‘to curse’, Old Prussian wertemmai ‘we swear’.
– iškalla-, iškall- < *sk(o)lq- ‘to slit, to split, to tear’, Greek σκάλλω ‘to hoe’.
– iškār-, iškar- < *sk(o)r- ‘to sting, to stab, to pierce’, Greek κείρω ‘to cut (off)’.
– išpānt-, išpant- < *sp(o)nd- ‘to libate, to pour, to sacrifice’, Greek σπένδω ‘to
libate’.
– išpār-, išpar- < *sp(o)r- ‘to spread (out), to strew’, Greek σπείρω ‘to spread
(out)’.
– išparra-, išparr- < *sp(o)rq- ‘to trample’, Vedic sphuráti ‘to kick’.
– ištāp-, ištapp- < *st(o)p- ‘to plug up, to block, to enclose, to shut’, Dutch stop-
pen ‘to plug up’.
– kānk-, kank- < *k(o)nk- ‘to hang, to weigh’, Gothic hahan ‘to hang’.
– karāp-, kare/ip- < *ghr(o)bq₁- ‘to devour, to consume’, Vedic grabh- ‘to seize’.
– lā-, l- < *l(o)q₁- ‘to loosen, to release, to untie, to relieve’, Gothic letan ‘to let’.
– lāḫu-, laḫu- < *l(o)q₂u- ‘to pour, to cast (objects from metal), to (over)flow’.
– lāk-, lak- < *l(o)gh- ‘to knock out (a tooth), to turn (one’s ears or eyes towards),
to train (a vine)’, Gothic lagjan ‘to lay down’.
– malla-, mall- < *m(o)lq₂- ‘to mill, to grind’, Latin molō ‘to mill’, Gothic malan
‘to mill’.1
– mālk-, malk- < *m(o)lK- ‘to spin’, Tocharian AB mälk- ‘to put together’.
– māld-, mald- < *m(o)ldh- ‘to recite, to make a vow’, Old Saxon meldon ‘to tell’.
– mārk-, mark- < *m(o)rg- ‘to divide, to separate, to distribute, to cut up’, Latin
margō ‘border’, Gothic marka ‘border’.
– mau-, mu- < *m(o)uq₁- ‘to fall’, Latin moveō ‘move’.
– nāḫ-, naḫḫ- < *n(o)q₂- ‘to fear, to be(come) afraid, to be respectful, to be care-
ful’, Old Irish nár ‘modest’.
– nai-, ni- < *n(o)iq- ‘to turn, to send’, Vedic náyati ‘to lead’.2
– para-, par- ‘to appear, to emerge’, which is a derivative of parā ‘out, further’
< *pro (Kloekhorst 2008: 630).
– pāšk-, pašk- < *P(o)sK- ‘to stick in, to fasten, to plant, to set up’.
– padda-, padd- < *bh(o)dhq₂- ‘to dig (the ground)’, Latin fodiō.
– šāḫ- < *soq₂- ‘to clog, to stuff, to fill in, to plug up’, Tocharian B soy- ‘to be
satisfied’.
– šākk-, šakk- < *s(o)kq₁- ‘to know (about), to experience, to recognize, to
remember’, Latin secō ‘to cut’, sciō ‘to know’.
– šārr-, šarr- < *s(o)rq₁- ‘to divide up, to distribute, to split, to separate’.
– šarāp-, šare/ip- < *sr(o)bh- ‘to sip’, Latin sorbeō ‘to slurp’.
– šarta-, šart- < *s(o)rdhq- ‘to wipe, to rub’, Middle High German serten ‘to vio-
late’.
– šuḫḫa-, šuḫḫ- < *suq₂- < *sq₂u- ‘to scatter’, Greek ὕω ‘to rain’.
– dā-, d- < *d(o)q₃- ‘to take, to wed, to decide’, Vedic dádāti ‘to give’, Greek δίδωμι
‘to give’, derivatives uda-, ud- ‘to bring (here)’, peda-, ped- ‘to take (some-
where), to carry, to transport, to spend (time)’.
– dākk-, dakk- < *d(o)kq₁- ‘to resemble’, Greek δοκεῖ ‘it seems’.
1 Dr Alwin Kloekhorst informs me that the stem malla- is the result of a younger development
and that the original stem was mall- from Old Hittite *māll- < *molq₁-.
2 But see now Kloekhorst & Lubotsky (2014), who argue for a derived stem *neq₁-i-.
– wai-, wi- ‘to cry (out)’, which is onomatopoeic (Kloekhorst 2008: 939).
– wāk-, wakk- < *u(o)q₂g- ‘to bite’, Greek ἄγνῡμι ‘to break’, Tocharian AB wāk-
‘to split, to burst’.
– warš- < *u(o)rs- ‘to reap, to harvest, to wipe’, Old Latin vorrō ‘to wipe’.
– wāš- < *uos- ‘to buy’, Latin vēnum dare ‘to sell’.
– wašta-, wašt- < *uosTq- ‘to sin, to offend’.
– zāḫ-, zaḫḫ- < *ti(o)q₂- ‘to hit, to beat’, Greek σῆμα ‘sign, mark’, σῶμα ‘corpse’.
There are no e-grade thematic presents in the Anatolian branch of Indo-Euro-
pean because these had not yet developed when it split off from the other
languages. Derived ḫi-verbs are based on nasal presents (e.g. tarna-, tarn- < *trk-
n(o)q- ‘to let go’; ḫamank-, ḫame/ink- < *q₂m(o)ngh- ‘to tie’), s-injunctives (e.g.
ḫārš- < *q₂(o)rq₃-s- ‘to till (the soil)’, Greek ἀρόω; paḫš- < *p(o)q₂-s- ‘to protect’,
Latin pāscō, pāvī), i-presents (e.g. arai-, ari- < *q₃r-(o)i- ‘to arise’, Latin orior;
išḫai-, išḫi- < *sq₂-(o)i- ‘to bind’, Vedic syáti; išpai-, išpi- < *spq₁-(o)i- ‘to get full’,
Vedic sphāyate), and reduplicated formations (cf. Kortlandt 2010: 378–380).
The third group of ḫi-verbs are the factitives in -aḫḫ- < *-eq₂-, which do
not show ablaut (cf. Kloekhorst 2008: 164). The model for this formation was
provided by the transitive thematic injunctive with secondary endings and an
ergative subject which originated at stage E (see above). If the original meaning
of *tud-e was ‘it [is] a blow’, where *tud- is a verbal root noun, and the mean-
ing of the thematic injunctive *tudet was ‘he causes a blow (to me)’, where the
thematic vowel was coreferential with the “instrumental” object which was
distinct from the regular direct object (goal of the action) in the accusative
case, the verbal root could easily be replaced by an abstract noun denoting a
property such as *neueq₂ ‘quality of being new’. In the other Indo-European
languages, this type of verb was replaced by the regular denominal formation
in *-eq₂-ie/o-, e.g. Latin novāre ‘to renew’.
References
Jasanoff, J.H. 1979. The position of the ḫi-conjugation. In: E. Neu; W. Meid (eds.), Hethi-
tisch und Indogermanisch. Vergleichende Studien zur historischen Grammatik und zur
dialektgeographischen Stellung der indogermanischen Sprachgruppe Altkleinasiens.
Innsbruck: IBS, 79–90.
Jasanoff, J.H. 2003. Hittite and the Indo-European Verb. Oxford–New York: Oxford Uni-
versity Press.
Hill, E. 2007. Die Aorist-Präsentien des Indoiranischen. Bremen: Hempen.
Kloekhorst, A. 2008. Etymological Dictionary of the Hittite Inherited Lexicon Leiden:
Brill.
Kloekhorst, A.; A.M. Lubotsky. 2014. Hittite nai-, nē-, Sanskrit nī-, and the PIE verbal
root *(s)neh₁-. In: H.C. Melchert; E. Rieken; T. Steer (eds.), Munus amicitiae Norbert
Oettinger a collegis et amicis dicatum. Ann Arbor: Beech Stave Press, 126–137.
Kortlandt, F. 2010. Studies in Germanic, Indo-European and Indo-Uralic. Amsterdam:
Rodopi.
Kortlandt, F. 2014. The Tocharian personal endings. Tocharian and Indo-European Stud-
ies 15, 79–86.
Kortlandt, F. 2015. Tocharian ē-grade verb forms. Tocharian and Indo-European Studies
16, 51–59.
Kümmel, M.J. 2004. Zur o-Stufe im idg. Verbalsystem. Indo-European Word Formation.
Copenhagen: Museum Tusculanum, 139–158.
Mallory, J.P. 1989. In Search of the Indo-Europeans: Language, Archaeology and Myth.
London: Thames & Hudson.
Peyrot, M. 2013. The Tocharian Subjunctive. Leiden: Brill.
Schaefer, C. 1994. Das Intensivum im Vedischen. Göttingen: Vandenhoeck & Ruprecht.
Stang, C.S. 1942. Das slavische und baltische Verbum. Oslo: Jacob Dybwad.
Guus Kroonen*
* I thank Ante Aikio, Sasha Lubotsky, Alwin Kloekhorst, Tijmen Pronk and Frits Kortlandt for
commenting on a draft version of this paper.
1 The Uralic proto-form is based on Hungarian ég and Komi i̮ń only, and represents just one
possible reconstruction.
Although the corpus of lexical similarities offered here is not by itself suf-
ficiently large to substantiate the Indo-Uralic Hypothesis, typological paral-
lels for linking the sound correspondences can be identified within a glot-
talic framework. A shift from PIU *ˀ⁽ń⁾, *ˀŋ to PIE *ɗ, *ɠ would be paral-
leled, for instance, by a similar shift that has been suggested for Wambule,
a Kiranti language spoken in eastern Nepal, where implosive ɗ developed
from *ˀn (Opgenort 2004). On the basis of this parallel, the Indo-Uralic sound
correspondence can at least theoretically be accounted for by assuming that
both the PIE mediae and the PU nasals derived from a series of PIU pre-
glottalized nasals (*ˀ⁽ń⁾, *ˀŋ,) or even implosive nasals: While in Proto-Indo-
European this series would have developed into glottalized stops, the preglot-
talization can simply have been lost in Proto-Uralic by a process of deplosiviza-
tion.4
Within the Indo-European family, the reconstruction of preglottalized
nasals is not actually ad hoc, as explained above, because it already accounts
for another, seemingly unrelated problem, i.e. the lexical distribution of the
no-participles. One may furthermore wonder whether the reconstruction of
preglottalized nasals can offer an explanation for what has been a perennial
question since the publication of Brugmann’s Grundriss, viz. that of the origin
of the aberrant nasal of Skt. viṃśatí- ‘20’, ostensibly from PIE *h₁uinḱmti-.
While the (ostensibly) non-nasal variants such as Avestan vīsaiti, Lat. vīgintī
and OIr. fiche etc. can be explained from the traditional glottalic proto-form
*dui-dḱmti- by dissimilation to *ʔuiʔḱmti- (Lubotsky 1994), the nasal of the
Sanskrit form is yet to be accounted for. It must demonstrably be of Proto-Indo-
Iranian age, at any rate, in view of the Ossetic form insæj (Hübschmann 1887:
104; Brugmann 1911, vol. 2: 31).
By reverting to a deeper reconstructional state, it is possible to assume that
in the stage prior to the buccalization of the preglottalized nasals, the Pre-PIE
2 This is a rather weak comparison in view of the evidence for *ŋ being restricted to Mari. It can
theoretically be saved by reconstructing the Uralic word as an obscured compound *ńï(ŋ)kV
‘?’ + *ćïmi ‘scales, fish skin’ (Ante Aikio, p.c.), perhaps within dissimilation of the first nasal.
3 The PIE and PU semantics can only be unified be reconstructing a meaning “put down”. How-
ever, this step requires an additional, unverifiable assumption.
4 Such a development is documented for variaties of the Sui language in southern China (Wei
and Edmondson 2008).
References
1 Introduction
If we assume that the Anatolian branch of Indo-European (IE) split off first, as
is frequently done, this implies that we can reconstruct two different stages of
the common ancestor of all other IE languages: Proto-Indo-European (PIE) is
the immediate ancestor of “Core IE”, the stage of the language before the first
(post-Anatolian) split occurred. From the comparison of Proto-Anatolian (PA)
and PIE we may then reconstruct the protolanguage of all IE languages, for
which the term Proto-Indo-Hittite has been used, but I would prefer the term
Proto-Indo-Anatolian (PIA). Since even the very fact that Anatolian split off
first is not completely assured, the distance between these two protolanguages
cannot have been too large. On the other hand, also PIA must have had its his-
tory, and there may be relatives of this language. One of the most promising
candidates is the neighbouring language family to the north of IE, i.e. Uralic.
From a comparison of PIA and Proto-Uralic (PU), we should then be able to
reconstruct their common ancestor, Proto-Indo-Uralic (PIU). This must be a
much more distant relationship, and it is far from undisputed. This means that
the quality and quantity of potential positive evidence is expected to be lower,
due to the rather large distance between the two protolanguages. Arguments
for both relationships and tree topologies have mainly been based on gram-
matical and lexical evidence. However, if these different protolanguages are
related, we should also be able to reconstruct their phonology and some fea-
tures of their sound systems, and the sound changes that have to be assumed
for these reconstructions may be used as evidence for their plausibility. In this
perspective, one main difference between PIE, PA, PIA and PU appears to lie
in their obstruent systems, especially in the number of stops: this number is
high in PIE, lower in PA and much lower in PU.
B An Alternative Scenario
If we accept both Kloekhorst’s PA reconstruction as such, and a (pre-)PIE
reconstruction with a “glottalic” or implosive media *ɗ (ˀd),1 we may also
reconstruct a PIA system identical with that of PIE, i.e. *t | d | ɗ (cf. below). In
this case, PIE had no change, and only PA changed the system by a shift from
a voicing contrast to a length contrast, again a chain shift, but in the opposite
direction: d > t > tt = t > tt, followed or accompanied by *d > t (cf. Kümmel 2007:
176). We may call this alternative model B, and in this case, the stop systems and
their changes would not support the reconstruction of PIA as distinct from PIE,
since there was only an Anatolian innovation, while the common ancestor of
all other branches was conservative.
1 Cf. Haider 1983; Weiss 2009; Kümmel 2012ab, 2014; cf. Kortlandt 1985.
2 The Western Armenian sound shift is only superfically similar, since it lead to a voiceless lenis
= /t/ > [d̥ ] vs. tʰ.
2.3 Conclusion
The probabilities of change according to both models are summarized in the
following table:
A B
some few words like PIE *h₃okʷ- ‘eye’ = Hitt. sākʷ-, Luwian dāw(ī)- but this is
still much too unclear.3
These changes can be compared with parallel chain shifts in Mon and Tai lan-
guages (cf. Diffloth 1984; Li 1977; Pittayawat 2009):
3 Cf. Ivanov 2001: 133; 2009: 3–5; Hart 2004; Olsen 1992; 2006; Kümmel 2008: 30–31; Kassian &
Jakubovič 2013: 22; Kassian, Zhivlov & Starostin 2015: 315 f., 327, 329.
Both cases show the same kind of parallel/chain shift, namely d | ɗ > d̤ ʱ | d. Such
a shift would plausibly explain the IE data, since it produced a rather unstable
system which was nowhere preserved as such.
4 Tentative Uralic reconstructions mainly follow Sammallahti 1988; Häkkinen 2009; Aikio 2012;
2015; Kallio 2012; Zhivlov 2014 (reconstructions based on Tálos 2015 are sometimes given in
brackets and marked by T, cf. Tálos 1987; Abondolo 1996,1998). Words often considered loans
from IE into Uralic are marked by (L).
1) There is some, but not much potential evidence for PU *t, k = PIA *ɗ, ɠ :
PIA *Ɗ = PU *T:
PIA PU/PFU/PFP
*ɗeh₃- ‘to give/take’ (L) *toxə- ‘to bring, get, give, sell’ (T *taɣi̮-): PS *tuokə-,
Hung. toj-
*ɠlH̥ -(e)w- *käläw ‘sister-in-law’: PS *kālō-, Finn. käly
*woɗ-/weɗ- (L) *wetə ‘water’ (T *wǟti): Finn. vesi, Hung. víz
*ge(n)ɗ- ‘to seize’ *kanta- ‘load, bring’: PS *kuontē-, Selk. quenda-
2) There is more possible evidence for PU sonorants = PIA *ɗ, ɠ/ɠ ʷ/ʛ, although
many cases are of course questionable.
a) PIA *Ɗ = PU *L
*seɗ- ‘to sit’ *sälə- ‘to sit down’: Komi se̮l-, Chanty jel-/tet-,
Selk. ti-
*meɗ- ‘to measure, think’ *mälə ‘mind’: PS *mielə, Finn. mieli, Komi mil ̮
?*kēr, kerɗ -, kr̥ɗ- ‘heart’ ?*ćüδä-mə ‘heart’ (T *śǟδmɜ): PS *čəðē, Finn.
sydän, Hung. szív [PU *δ, *δʲ have been interpreted
as laterals]
b) PIA *Ɗ = PU *N5
*ɗek- ‘to accept, perceive’ *näkə- ‘to see’: Finn. nähdä, Hung. néz-, Chanty
ni(ɣʷ)-
*ɗuk- ‘to draw, lead’ *ńükə- ‘to draw, tear’: Hung. nyű, Mansi ńüw-, Selk.
nek-
*ɗn̥ g-wéh₂- ‘tongue’ *ńaŋk-ćə ‘tongue, gums’: PS *ňuokčəm-, Chanty
ńaŋχšəm, Nen. ńińćīʔ
?*ɗig- ‘tick’ *ńiŋə ‘maggot’ (T *ńǟŋi): Saami N njivnja; Hung.
nyű, Mansi ńiŋʷkʷ
5 The Uralic distinction of *n and *ń appears not to be reflected in PIA. Either it was secondary
or two sounds had merged in pre-PIA.
(cont.)
b) PIA *Ɗ = PU *N
3.4 Conclusion
If his rather limited material shows us something real, it supports the idea that
PIA implosives can correspond to Uralic sonorants and not only to stops. The
evidence appears to be small for non-nasal sonorants and a bit better for nasals:
the numbers are roughly equal for PU *t = PIA *ɗ and for PU *N = PIA *ɗ ; in the
case of tectals the latter type seems to be a bit better attested. Of course, this
then raises a question about the PIU state of affairs: what should we reconstruct
here? Do PIA implosives = PU nasals (and laterals) and/or stops reflect PIU
implosives or nasals—or something else? So again, do we have a PIA archaism
and a PU change: PIU *ɓ, *ɗ, *ɠ > PU *p~m, *t~n, *k~ŋ? Or is a PU archaism
and a PIA change more probable, i.e., PIU *m, *n, *ŋ > PIA *m~ɓ, *n~ɗ, *ɠ ?
6 A new IA etymology is given by Steer 2015: 43–60; if correct, the comparison made here would
be impossible.
4 Summary
In these two studies, the prehistory of the PIE and the PIA system of stop artic-
ulation types has been investigated. In the first case, the most probable type
of distinction between fortis and lenis stops in PIA was investigated, with the
result that diachronic typology favours the reconstruction of a PIA system with
distinctive voicing, as in PIE, vs. an original system based on a quantity dis-
tinction, as in PA. The most archaic PIA system (including implosive stops)
was then compared to PU, with the result that the PIA implosives may corre-
spond at least partly to Uralic sonorants, especially nasals. This correspondence
is best explained as a PIA preservation of the original PIU sounds vs. a change
of implosives to sonorants and/or explosives in PU. The reconstruction of the
respective system types is given in the following table of correspondences:
Core IE *t *dʱ *d
PIE *t *d *ɗ
PA *t: *t
PIA *t *d *ɗ
PU *t *n (*l) ~ *t
PIU *t *d? *ɗ
Interrogatives Demonstratives
PU *ku, ko- *mi- sg. *tä-, **to- *ći/-e- *i/e- pl. *nä-, *no-
PIA *ku-, kwi/e– *mo *to- *ki/e- *(h₁)i/e– *no-
Nominal Markers
Plural Dual
Verbal Markers
1s 2s 1p 2p 3s 3p
PIA/PIE PU/PFU
*wed- ‘to lead’ (L) *wetä- ‘to lead’ (T *wǟtä-): Finn. vetä-, Hung.
vezet
*medu- ‘mead, honey’ (L) *metə ‘honey’ (T *mǟti): Finn. mesi, Hung.
méz/méze-
*h₁em- ‘to take’ *ɨm-ta- ‘to give’ (T *amta-): PS *wuomtē-, Finn. anta-,
Hung. ad
*Hjuh₂- ‘to eat’ (Nikolaev 2014) *juxə- ‘to drink’ (T *jaɣi̮-): PS *jukə-, Finn. juo-, Hung.
iv-
*Hjud- ‘to move’ *juta- ‘to wander, travel’: PS *juttē-, Mordva juta-,
Nen. jādā
*deh₁- ‘to do, put’ (L) *tekə- ‘to do, make’ (T *tǟki-): PS *təkə-, F. teke-,
Hung. tev-
*teq- ‘to weave’ *tekV- ‘to push’: Komi toj-, Hung. tűz-, Chanty tök-
*kerw-/kru- ‘horn’ (L) *ćorwa ‘horn’ (T *śārwa): PS *čoarwē, Finn. sarvi,
Hung. szarv
*ges- ‘hand’ *kätə ‘hand’: PS *kietə, Finn. käsi, Mari kit, Hung.
kéz/keze- ?*qelh₁- ‘to impel’ *kaδʲa- ‘to let, leave’: PS
*kuoðē-, Hung. hagy
?*qelh₁- ‘to bring, impel’ *kälä- ‘to wade’: PS *kālē-, Hung. kel-
(cont.)
PIA/PIE PU/PFU
*kʷek- ‘to look’ *kokə- ‘to see’: Finn. koke-, Selk. qo-
*kʷelh₁- ‘to turn’ *kulkə- ‘to move’: PS *kolkə-, Hung. halad, Nen. χūlā-
*leNd- ‘low’ *lamtə ‘low’ (T *lamta): Finn. lansi, Komi lud, Nen.
lamtū
References
Hart, G.R. 2004. Some problems in Anatolian phonology and etymology. In: J.H.W. Pen-
ney (ed.), Indo-European Perspectives. Studies in Honour of Anna Morpurgo Davies.
Oxford: Oxford University Press, 341–354.
Hyllested, A. 2009a. Internal reconstruction vs. external comparison: the case of the
Indo-Uralic larnygeals. In: T. Olander; J.E. Rasmussen (eds.), Internal Reconstruc-
tion in Indo-European: Methods, Results and Problems. Copenhagen: Museum Tus-
culanum, 111–136.
Hyllested, A. 2009b. PIE *-bh- in nouns and verbs: Distribution, function, origin. In:
R. Lühr; S. Ziegler (eds.), Protolanguage and Prehistory. Akten der XII. Fachtagung
der Indogermanischen Gesellschaft vom 11. bis 15. Oktober 2004 in Krakau. Wiesbaden:
Reichert, 201–214.
Ivanov, V.V. 2001. Southern Anatolian and Northern Anatolian as separate Indo-Eu-
ropean dialects and Anatolian as a late linguistic zone. In: R. Drews (ed.), Greater
Anatolian and the Indo-Hittite family. Washington: Institute for the Study of Man,
131–183.
Ivanov, V.V. 2009. K issledovaniju otnošenij meždu jazykami. Journal of Language Rela-
tionship 1, 1–12.
Kallio, P. 2012. The non-initial-syllable vowel reductions from Proto-Uralic to Proto-
Finnic In: T. Hyytiäinen; L. Jalava; J. Saarkivi; E. Sandman (eds.), Per Urales ad Ori-
entem: Iter polyphonicum multilingue. Festskrift tillägnad Juha Janhunen på hans sex-
tioårsdag den 12 februari 2012. Helsinki: Finno-Ugrian Society, 163–175.
Kassian, A.S.; I.S. Jakubovič 2013. Anatolijskie jazyki. In: J.B. Korjakov; A.A. Kibrik (eds.),
Jazyki mira: Reliktovye indoevropejskie jazyki perednej i central’noj Azii. Moscow:
Academia, 15–26.
Kassian, A.; M. Zhivlov; G. Starostin 2015. Proto-Indo-European-Uralic comparison
from the probabilistic point of view. Journal of Indo-European Studies 43, 301–347.
Klingenschmitt, G. 2005. Sprachverwandtschaft in Europa. In: G. Hauska (ed.), Gene,
Sprachen und ihre Evolution. Wie verwandt sind die Menschen, wie verwandt sind ihre
Sprachen? Regensburg: Universitätsverlag, 100–132.
Kloekhorst, A. 2008a. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden–
Boston: Brill.
Kloekhorst, A. 2008b. Some Indo-Uralic aspects of Hittite. Journal of Indo-European
Studies 36, 88–95.
Kloekhorst, A. 2010. Initial stops in Hittite (with an excursus on the spelling of stops in
Alalaḫ Akkadian). Zeitschrift für Assyriologie 100, 197–241.
Kloekhorst, A. 2013. The signs TA and DA in Old Hittite: Evidence for a phonetic differ-
ence. Altorientalische Forschungen 40, 125–141.
Kloekhorst, A. 2014. The Proto-Anatolian consonant system: An argument in favor of
the Indo-Hittite hypothesis? Talk at The Sound of Indo-European 3: Phonetics, Phone-
mics, and Morphophonemics, Opava, Nov. 13, 2014.
Kloekhorst, A. 2016. The Anatolian stop system and the Indo-Hittite hypothesis. Indo-
germanische Forschungen 121, 213–247.
Kortlandt, F. 1985. Proto-Indo-European glottalic stops: the comparative evidence. Folia
Linguistica Historica 6/2, 183–201.
Kortlandt, F. 2002. The Indo-Uralic verb. In: R. Blokland; C. Hasselblatt (eds.), Finno-
Ugrians and Indo-Europeans: Linguistic and Literary Contacts. Maastricht: Shaker,
217–227 [= 2010: 391–403].
Kortlandt, F. 2004. Indo-Uralic consonant gradation. In: I. Hyvärinen; P. Kallio; J. Korho-
nen (eds.), Etymologie, Entlehnungen und Entwicklungen. Festschrift für Jorma Koi-
vulehto zum 70. Geburtstag. Helsinki: Société Néophilologique, 163–170 [= 2010: 409–
414].
Kortlandt, F. 2010. Studies in Germanic, Indo-European, and Indo-Uralic. Amsterdam–
New York: Rodopi.
Kümmel, M.J. 2007. Konsonantenwandel. Bausteine zu einer Typologie des Lautwandels
und ihre Konsequenzen für die vergleichende Rekonstruktion. Wiesbaden: Reichert.
Kümmel, M.J. 2008. Review of John H.W. Penney (ed.), Indo-European Perspectives.
Studies in Honour of Anna Morpurgo Davies. Oxford: Oxford University Press, 2004.
Kratylos 53, 25–35.
Kümmel, M.J. 2012a. Typology and reconstruction: The consonants and vowels of Proto-
Indo-European. In: B. Nielsen Whitehead; T. Olander; B.A. Olsen; J.E. Rasmussen
(eds.), The Sound of Indo-European. Phonetics, Phonemics, and Morphophonemics.
Copenhagen: Museum Tusculanum, 291–329.
Kümmel, M.J. 2012b. The distribution of IE roots ending in IE *ND. In: R. Sukač; O. Šefčík
(eds.), The Sound of Indo-European 2. Papers on Indo-European Phonetics, Phonemics
and Morphophonemics. München: LINCOM, 159–176.
Kümmel, M.J. 2015. The role of typology in historical phonology. In: P. Honeybone;
J.C. Salmons (eds.), The Handbook of Historical Phonology. Oxford: Oxford Univer-
sity Press, 121–132.
Kümmel, M.J. 2016. Is ancient old and modern new? Fallacies of attestation and recon-
struction (with special focus on Indo-Iranian). In: D.M. Goldstein; S.W. Jamison;
B. Vine (eds.), Proceedings of the 27th Annual UCLA Indo-European Conference. Bre-
men: Hempen, 79–96.
Kümmel, M.J. 2018. The survival of laryngeals in Iranian. In: L. van Beek; A. Kloekhorst;
G. Kroonen; M. Peyrot; T. Pronk; M. de Vaan (eds.), Farnah. Indo-Iranian and Indo-
European Studies in Honor of Sasha Lubotsky. Ann Arbor: Beech Stave, 162–172.
Kümmel, M.J. fthc. Verwandte des Indogermanischen? Zur Frage des “Eurasiatischen”
und anderer Makrofamilien. In: H. Hettrich; S. Ziegler (eds.), Die Ausbreitung des
Indogermanischen. Arbeitstagung der Indogermanischen Gesellschaft, Würzburg, 24.
bis 26. September 2009. Wiesbaden: Reichert.
Li, F.K. 1977. A Handbook of Comparative Tai. Honolulu: University of Hawai’i Press.
1 Introduction
One of the characteristic features of the Anatolian languages that sets them
apart from the other Indo-European languages is the fact that neuter nouns
that are the subject of transitive sentences (the Agent position) do not appear
in their nominative/accusative form. Instead they receive a suffix, for example
Hittite sg. -anza, pl. -anteš, that is traditionally called the ergative suffix. Such a
terminology implicitly presupposes that Anatolian has an actual ergative case,
as has been argued for by Garrett (1990). This is, however, not the only theory
on the place of the suffix -anza/-anteš within the grammatical system of the
Anatolian languages. Laroche (1962) and Benveniste (1962) segment the suffixes
into -anza = /-ant-s/ and -anteš = /-ant-es/, so that they consist of a suffix -ant-
and a common nominative ending -š/-eš. Thus, under this analysis, there exists
a suffix -ant- that transforms a neuter word into a common gender one. Patri
(2007), on the other hand, considers the ergative construction to be a special
instance of the ablative case,1 which in Hittite has an allomorph -anza. Since
there is no consensus whether the “ergative” actually is a separate case in the
grammar of the Anatolian languages, in this article I will use the more neu-
tral term agentive2 and reserve the term ergative for a proper morphological
case.
The discussion on the function of the Anatolian agentive is primarily a syn-
chronic one. As such, the function of the agentive is to be decided in every
Anatolian language separately. The discussion has mostly centered on Hittite,
as this is the Anatolian language of which the most textual material has sur-
vived. Recently Goedegebuure (2013) has given a diachronic description of the
* This paper is based on a term paper for a course on Hieroglyphic Luwian taught at Leiden
University in 2013 by Alwin Kloekhorst, to whom I am greatful for helpful commentary. This
paper was written while the author was a Ph.D. student at Radboud University.
1 Garrett (1990) considers the ablative to be the origin of the ergative case in Anatolian, but in
his analysis the ablative and ergative are two separate cases in synchronic Anatolian.
2 Not to be confused, of course, with agentive nouns such as those formed by the suffix -er in
English, e.g. walk → walker.
agentive in Hittite. She shows that the agentive was a syntactic suffix -ant- in
Middle Hittite, whose nominative singular -anza and plural -anteš became fos-
silised in Neo-Hittite as endings of a new ergative case.
The aim of this article is to expand upon Goedegebuure’s research by deter-
mining the synchronic role of the agentive construction in the grammar of the
Anatolian languages in which it is attested, namely Hittite, Cuneiform Luwian,
Hieroglyphic Luwian and Lycian.3 After this is established, I will determine
the Proto-Anatolian origins of these ergative constructions. Finally, I will con-
sider its implications for the reconstruction of Proto-Indo-European and for
the Indo-Anatolian hypothesis.
As was mentioned before, there have been several proposals for the analysis of
the Anatolian agentive, all of which are neatly summarised by Melchert (2007).
One can distinguish between the following four analyses, presented here for the
Hittite agentive suffix -anza/-anteš:
1. The semantical or derivational analysis. Under this analysis, proposed by
Benveniste (1962), the agentive suffix -anza/-anteš consists of a suffix -ant-
and the nominative endings of the common gender. The suffix -ant- is a
derivational suffix that creates an animate noun from an inanimate noun
by means of personification. Hence, under this analysis, one cannot use
the word lingāi ‘oath’ in the agent position, and one has to resort to using
linkii̯anteš ‘oath gods’, which has a different but similar meaning.
2. The syntactical or inflectional analysis. This analysis, first proposed by
Laroche (1962), also considers the agentive suffix -anza/-anteš to consist
of an inflectional suffix -ant- and the nominative common gender end-
ings. The difference between this analysis and the previous one is that
under this analysis the suffix -ant- does not have any semantic value.
Instead this analysis posits that neuter nouns, as a rule, cannot be syn-
tactic agents. To express a situation in which a neuter noun is the actor of
an action with a patient, the suffix -ant- must be used to change the mor-
phological gender of the neuter noun into the common gender, without
affecting the semantics of the noun.
3 Valério (2009) raises the possibility that Palaic fulāsinanza is a possible example of the
agentive construction of a Hattic loan fulāsina ‘bread’, although he prefers an interpretation
fulāsin-ant-s ‘having bread’. At any rate the context is too scarce to provide enough informa-
tion about the role of the agentive in Palaic.
3. The ergative analysis. This analysis has been argued for by Garrett (1990).
Under this analysis the suffixes -anza and -anteš cannot be decomposed.
Instead, these suffixes are endings of a distinct morphological case, the
ergative, which only occurs in neuter nouns.
4. The ablative analysis. This analysis, proposed by Patri (2007), considers
the agentive suffix -anza to be a special use of the ablative case. In Hittite,
this case is indifferent to number. It occurs mostly as -az, but it has an
allomorph -anza. According to this analysis, neuter nouns with the suf-
fix -anteš are actually not agentives, but nominative plurals of derived
nouns.
Before I try to fit the Anatolian data on these four analyses it is useful to dis-
cuss how one can determine which of the analyses is correct for each Anatolian
language; this will streamline the discussion in the later sections. First, as was
shown by Melchert (2007: 163–164), the ablative analysis cannot hold in Hittite
for a number of reasons. First, if the agentive construction is a special use of the
ablative case, one would expect some attestations of the agentive construction
with the more common ablative allomorph -az.4 Furthermore, the Hittite abla-
tive is indifferent to number, whereas the agentive construction has a distinct
plural. Finally, if the agentive construction featured an ablative case, one would
expect accompanying adjectives and demonstratives to be in the ablative case
as well. As Melchert shows this does not happen in Hittite; the agreement rela-
tions of the agentive construction will be discussed shortly.
In the Luwic languages the same reasons hold. In fact, in these languages
the ablative ending is different from the agentive suffix, as can be seen from
the following table:
Agentive Ablative
Singular Plural
4 As Melchert notes, although there are some neuter nouns in which the agentive appears as
-az, this is due to “nasal reduction” and the suffix is underlyingly still /-ants/ rather than the
ablative ending /-ats/.
Thus in all Anatolian languages the agentive construction does not feature the
ablative case. From this point onwards I will not include the ablative analysis
in my discussion.
If the suffix -ant- is a semantical derivation, one expects its usage to be deter-
mined semantically rather than syntactically. As such, one expects a deriva-
tion X-ant- derived from a noun X to have a different meaning. Since their
distribution is determined semantically, one also expects the derived noun X-
ant- to appear in other cases than the nominative, and one also expects some
instances of X in the agent position. Also, the usage of the suffix would not
be determined by morphological gender, but by some semantic property. This
property would most likely be inanimacy, as is argued for by Benveniste (1962).
The correlation between the neuter gender and semantic inanimacy is not per-
fect, however, and we would expect a semantic derivation -ant- to appear on
inanimate common gender words as well.
It should be noted that there is a semantic suffix -ant- in Hittite with many
different uses, such as a singulative use and an abundative use; these uses have
been described by Josephson (2004). It is a priori possible that the agentive
is one of the functions of this suffix; this is especially attractive since one of
the functions of the suffix -ant- is to form agent nouns. It is very well possible
that this suffix is the historical origin of the agentive construction in Anatolian
(Oettinger, 2001). Before investigating this possibility, however, it is necessary
to determine the synchronic function of the agentive construction.
The remaining two hypotheses, that -anza reflects either an inflectional suf-
fix -ant- or a proper ergative ending, are harder to distinguish. In both of these
cases the use of the agentive construction would be determined by a grammat-
ical rather than a semantic feature of the sentence, namely the morphological
gender of the agent. Also, under both of these analyses we expect no difference
in meaning between a noun X and the form X-ant- in the agentive construction.
According to Melchert (2007) an inflectional suffix would be expected to
appear in all morphological cases. However, this reasoning is not fully correct;
if a neuter agent is the grammatical condition that defines the use of an inflec-
tional suffix -ant-, we would not expect to see it anywhere but in the nominative
singular and plural. The comparison Melchert makes with Latin dea ← deus is
not fully applicable, since dea differs in meaning from deus. If the inflectional
suffix -ant- is only used in the agent position, it only appears in the common
gender nominative forms -anza and -anteš. As such, the difference between the
syntactical analysis and the ergative analysis cannot be seen from the forms in
-anza and -anteš themselves.5
5 Of course, one might argue that a language with a grammatical suffix which can only occur in
the nominative case is a very unstable scenario, and that the suffix would be reanalysed into
a proper case soon. This, however, does not rule out the possibility that -ant- is an inflectional
suffix as a synchronic description.
6 A short discussion of the agentive construction in Old Hittite is found in Kammenhuber 1993,
who remarked that the argumentation of Laroche (1962) for the syntactical analysis of the
agentive does not take the Old Hittite evidence into account. She proposes that in Old Hittite
the neuter noun italu ‘evil’ occurs in its nominative-accusative form as the subject of both
transitive and intransitive sentences. She gives the following example from KBo 18.151, rev.
7–8:
4. [a]p-pu-uz-zi kar-ta-x[…]
5. ták-ku=uš še-e-er […]
6. [ap]-pu-uz-zi-an-za […]
Although the text from Example 1 is too fragmentary to completely make out
the meaning of the sentence, we can see that we are dealing with a derived form
of the neuter word appuzzi ‘animal fat’. Although the verb of this sentence is
lost, the fact that the accusative plural common enclitic pronoun =uš appears
in this sentence shows that the sentence must be transitive. The question is
now whether there is a difference in meaning between appuzzianza in this
text and the neuter word appuzzi. The text VBoT 58 is a new script copy of this
text. In this text we find the sentence (i 13–14) takku=aš t[innuzi] nu=ma=ašta
andurza UZUap[p]uzzii̯anza ḫarzi ‘If he paralyzes (the grains), the fat will keep
them within’ (Puhvel 1984: 103).7 In this sentence it is clear that there is no
semantic difference between appuzzi and appuzzianza. The same is probably
true in Example 1, since appuzzi in line 4 and appuzzianza in line 6 probably
have the same referent. This means that the suffix -anza has a grammatical role
in this sentence, and as such either the syntactical or the ergative analysis is cor-
rect in this instance. However, the context is too meagre to decide which of the
two is applicable.
‘The word of the tablet is broken. The oaths must seize … . He must perish.’
In Example 2 we find the agentive of the neuter word lingai- ‘oath’. To find
out what the semantics of linkiia̯ nteš is we have to consider the context. The
preceding and following lines show that the sentence in line 6 is part of a curse
formula. This makes one suspect that linkii̯anteš means ‘oath gods’ rather than
‘oaths’. It should be noted that we find NI-IŠ DINGIRMES ap-pa-an-t[u] ‘the
oath gods must seize’ earlier in the same text. Thus the most probable explana-
tion is that the word linkii̯anteš is semantically different from lingai-, so in this
instance the evidence points toward the semantical analysis.
Although the material is quite scarce, we see from Example 1 that it is prob-
able that the agentive suffix -ant- had a grammatical function in Old Hittite
already. This invalidates the suggestion of Goedegebuure (2013) that the suf-
fix -ant- only had a semantical function. We cannot see whether the agentive
consists of a syntactical suffix -ant- or of an unanalysable ending -anza from
the Old Hittite evidence itself. However, since we see an ergative ending -anza
develop from a syntactical suffix -ant- from Middle to Neo-Hittite, the Old Hit-
tite agentive is best to be interpreted as a syntactic suffix -ant- as well.
There are several attestations of the agentive in Cuneiform Luwian, but the
number is small enough to treat every attestation in detail. In this section all
Luwian forms are discussed that are marked as an ergative by Melchert (1993);
all texts are from Melchert (2001a).
‘The sky and the earth must wash their mouths; they must become oil;
they must become honey’
Example 3 contains the agentives of the nouns tappaš- ‘sky’ and tii̯amm(i)-
‘earth’. The latter is a common gender word, as its nominative singular tii̯ammiš
and its accusative singular tii̯ammin have both been attested. This already
shows that the suffix -ant- cannot have a grammatical function and must be
semantical. This can also be seen from the content of the sentence: the two
subjects have mouths and as such the words do not refer to the earth and the
‘The houses, the pediment, the hearth, the ḫuu̯ aḫḫurša-s, the earth, stat-
ues, the evil word, sickness, past (and) present impurity of irḫuu̯ a- (and)
of animals, of the dead (and) of the living must release them.’
Example 4 contains a rather large summation in which we find the two agen-
tives parnantinzi ‘houses’ and ḫuu̯ aḫḫuršantinzi ‘statues’. Although all ele-
ments of this summation are inanimate, the words that have the agentive con-
struction are precisely those that are of the neuter gender. In other words,
the determining factor for the use of the agentive construction is grammati-
cal rather than semantical, which shows that in this sentence the agentive is
either an inflectional suffix or an ergative case ending.
7. [ … ]x kur-ša-ú-na-an-ti-in-zi a-ri-in[-ta]
8. [ … ]x a-ap-pu-u̯ a-ni-in-zi a-a-ri-in-ta […]
‘The islands restrained […]; The āppuu̯ aninzi restrained […]; Tarḫunt and
[…] restrained the terāimman sky; the Great God and the Sun-god of the
Earth restrained the dūu̯ azza earth.’
‘Ditto (= he made) for him the rivers and streams. [Ditto] for him not the
illness of the heart and the eye. […] the illness of the head and the eye
[…] tarpī-ed.’
3. [ … ]x UZUNÍG.GIG-an-ti-iš KI.MIN
4. [ … ] KI.MIN
5. [ … ]x-u̯ a-an-ti-iš KI.MIN
6. [ … ]x-ta-ti-ti-in-zi KI.MIN
7. [ … ]x-u̯ a-an-ti-iš
8. [ … ḫal-l]i-na-i
‘[…] the liver ditto. […] ditto. [… the x]-u̯ ant- ditto; [… the x-]tati-s ditto;
[… the x]-u̯ ant- [… is s]ick.’
Unfortunately the context of the text of Example 7, which has been badly
preserved, tells us almost nothing about the use of the agentive. The only
other form of the word UZUNÍG.GIG- ‘liver’ that is attested is the ablative-
instrumental UZUNÍG.GIG-ti. Since the word underlying the sumerogram
UZUNÍG.GIG is unknown, it is possible that -ant- is just a part of the stem, so
we cannot be sure that UZUNÍG.GIG-antiš is indeed an agentive construction.
Even if it were, the context does not tell us anything about its use: the only word
preserved in the sentence is KI.MIN ‘ditto’, whose referent is unknown. As such
this attestation does not help us in distinguishing between the different analy-
ses of the agentive construction.
‘The […] offense must not […] him, the lord of the ritual.’
5. [ … -]zi ḫar-da-an-ti-in-zi
‘…’
Since the agentive construction in Cuneiform Luwian has the forms sg. -antiš,
pl. -antinzi, we expect the agentive to take the forms °a-ti-sa and °a-ti-zi in
Hieroglyphic Luwian. Unfortunately this is also what the nominative endings
for common gender nouns in -a(n)ti- would look like. In order to properly dis-
tinguish between these nouns and the agentive construction, I will only con-
sider nouns ending in °a-ti-sa and °a-ti-zi of which case forms of the noun in
-Ca- have been attested. However, in the material of Hawkins (2000) we find
only two such attestations, both in a single sentence in Example 10.
‘The sky, the earth, and the gods of the sky and the earth must … him with
ligitation.’
The last language in which the agentive is attested is Lycian. There are two attes-
tations in the material by Melchert (2001b). Since these attestations are in two
sentences that are very similar in nature, it is best to treat them both at once.
The sentences are given in Examples 11 and 12.
These two formulations are remarkably parallel (note that nasalisation may
drop in front of n or t in Lycian). This makes it probable that the function of the
agentive tesẽti is the same in both sentences. The word is derived from tese/i-
‘oath’. This word is of neuter gender, since a neuter accusative plural tasa is
attested in the sentence (TL 36 4–6) s=e=i( j)=e ñta tãtẽ tasa miñta meleime
se( j)=aladahali ada ///- ‘He placed the council oaths inside for meleime and
a fee of 3.5 ada’. In both of the examples given the noun tese/i- has a suffix
-ẽti that forms the agentive construction in Lycian. As was the case in the other
languages, there are three possible analyses to consider:
1. Semantical analysis. The suffix -ẽti consists of a semantical derivational
suffix -ẽt- and the common gender plural ending -i.
2. Syntactical analysis. The suffix -ẽti consists of an inflectional suffix -ẽt- and
the common gender plural ending -i.
3. Ergative analysis. The suffix -ẽti is the neuter plural ending of the ergative
case.
In both sentences the noun tesẽti governs the adjective trm̃ mili-. In Example 12
this adjective has the same suffix -ẽti as a case ending. Under both the seman-
tical analysis and the syntactical analysis we would expect the form trm̃ mili as
in Example 11. This shows that the adjective trm̃ milijẽti is in the ergative case
rather than the nominative. The same must be true of tesẽti, so in Example
12 the ergative analysis is correct. This again implies that tesẽti is semantically
identical to tese/i- ‘oath’, although one might suspect a personification in this
context.
Since the two sentences are so identical, we can assume that the meanings
are also similar. Thus we may suppose that in Example 11 there is no semanti-
cal difference between tese/i- and tesẽti. This rules out the semantical analysis.
Unlike in Example 12, however, the adjective is trm̃ mili rather than trm̃ mili-
jẽti, which has a common gender nominative plural ending. Hence the word
teseti is a common gender nominative as well, which means that the syntactical
analysis is correct.
We see that the agentive has two different analyses in Lycian. This could be a
dialectal or diachronical difference, but the amount of data is too small to draw
any conclusions about the distribution. At any rate, these two sentences show
that a proper ergative case was present in Lycian for at least some speakers, and
that this ergative case derived from an inflectional suffix -ẽt-, which may either
still be present in the language as such, or traces of this origin can be seen in
the allomorph -i of the adjectival ergative plural ending -ẽti.
Now that we have gathered all the synchronic information on the separate
Anatolian languages the historical development can be discussed. The central
question of this section is what the role of the agentive construction was in
By its nature this syntactic suffix could only occur in the nominative singu-
lar and plural. As it could not be fully inflected, the suffix and the nominative
ending together were prone to reanalysis as a single morpheme. Since there
was no longer a suffix to transfer the word to the common gender, this single
morpheme was considered a case ending of the neuter word. Since this case
was used precisely in the agent position, we can regard this as an ergative case.
This ergative case ending then spread to the adjectival system, which allowed
for sentences such as sẽne tesẽti qãñti trm̃ milijẽti ‘and the Lycian oaths will seize
him.’
Since we find the syntactical suffix *-ont- in all Anatolian languages in
which the agentive is attested, we should reconstruct this grammatical suf-
fix into Proto-Anatolian, which means that a suffix *-ont- that could only be
used syntactically must be of pre-Proto-Anatolian date. As mentioned before
the actual ergative case was a separate Hittite and Lycian development that
cannot be reconstructed into Proto-Anatolian. As such we see that Proto-
Anatolian reflects the analysis of the agentive as a syntactic suffix as put forth
by Laroche (1962) for Hittite, and that all the attested forms of the agentive
in the separate Anatolian languages can be derived from this. In particular,
this means that Proto-Anatolian, like all Anatolian languages except for Lycian
and Neo-Hittite, did not allow neuter nouns in the agent position in sen-
tences.
Having reconstructed the alignment system for Proto-Anatolian, I can now turn
towards Proto-Indo-European. As is well known, all branches of Indo-European
that retain the distinction between neuter and masculine/feminine, except
for Anatolian, display a partially accusative alignment system in which mas-
culine/feminine nouns have nominative and accusative forms, and in which
neuter gender nouns have a neutral declension, i.e. the Subject, Agent and
Patient forms are identical. On the other hand, in Proto-Anatolian, we have
seen that neuter nouns could not occur in the agent position, and a com-
mon gender noun had to be formed by means of the syntactic suffix *-ont-. If
the term Classical Indo-European (CIE) is used for the non-Anatolian Indo-
European languages, then one arrives at the reconstructions of the noun de-
clensions below (using o-stems as an example: the endings in C-stems are dif-
ferent, but the principle is the same).
8 In the context of the Indo-Anatolian hypothesis the ancestor of all non-Anatolian Indo-
European languages is occasionally called Proto-Indo-European, whereas the ancestor of PIE
and Anatolian is called Proto-Indo-Anatolian. These terms correspond to my PCIE and PIE,
respectively.
The problem with this explanation is that one would expect the loss of
agents to happen in semantically inanimate nouns, rather than in morpho-
logically neuter nouns. Although there is a strong correlation between inani-
macy and the neuter gender, this correlation is not perfect even in Proto-Indo-
European. For example, *dʰéǵ-m- ‘earth’ (Gr. χθών, Skt. kṣám, Hitt. tēkan) was
feminine in Proto-Indo-European, but refers to an inanimate object; on the
other hand *peḱu ‘cattle’ (Skt. páśu, Goth. faihu, Lat. pecū) is neuter, but refers
to something animate. Thus one would expect the agent of all inanimate nouns
to be lost, and one would expect the agentive construction to be determined by
animacy rather than by morphological gender. Example 4, however, shows that
this is clearly not the case.
On the other hand, if PIE did not allow for neuter agents, both the CIE
and the Anatolian situation can be explained. The lack of neuter agents would
constitute a gap in the system of the language. PCIE ‘fixed’ this gap by extend-
ing the neuter subject ending *-om to the agent, since the subject and agent
endings were equal in the masculine/feminine noun declension as well. Proto-
Anatolian, on the other hand, ‘fixed’ the gap by grammaticalising the person-
ifying suffix *-ont-. This reconstruction is considerably less problematic than
assuming a PIE neuter agent ending *-om, so the conclusion is that PIE did
not allow for neuter agents. This was, of course, a very unstable situation,
which was quickly resolved both in CIE and in Proto-Anatolian. The fact that
it was resolved in different ways, however, shows that we have to reconstruct
a gap in the system. We thus arrive at the reconstruction offered in the table
below.
PIE PCIE PA
One might wonder how PIE could express situations in which an object, ref-
erenced by a neuter noun, is the agent of an action. The suffix *-ont- was not
yet grammaticalised in PIE, since the only evidence for its grammaticalisation
is found in Anatolian. Thus PIE did not have either the CIE or the Anatolian
strategy for expressing neuter agents. One can imagine that such a situation
References
Alexander Lubotsky
1 Introduction
A nominal suffix *-ens- does not belong to the standard equipment of the
Indo-Europeanist. This suffix is not listed in the major handbooks and is but
rarely reconstructed. The only exceptions I am aware of are the word for ‘moon’
*meh₁-ns- and the word for ‘goose’, which Kortlandt (1978, 1985, 2013) has recon-
structed as *ǵhh₂-ens-. I believe, however, that this suffix is found in quite a few
Indo-European nominal and verbal formations.1
2 Nouns
1 I am grateful to Lucien van Beek, Frederik Kortlandt, Michaël Peyrot, Tijmen Pronk, Michiel
de Vaan, and Mixail Živlov for comments on an earlier version of this paper.
acc. *ǵhh₂ensm̥ , gen. *ǵhh₂n̥ sos), but the evidence for it is unfortunately rather
slim,2 based only on the necessity to account for the alternation of the ini-
tial consonants in Balto-Slavic. Slavic g- may be due to depalatalization in the
position before a laryngeal,3 but then it is not clear why Baltic does not attest
this development. This riddle would be solved by reconstructing an alternating
paradigm *ǵhh₂-ens- > BSl. *g- vs. *ǵheh₂-ns- > BSl. *ź-.
An alternative solution would be to assume that the Slavic word is a loan-
word from Germanic, which is a controversial issue. Borrowing has been sug-
gested, albeit hesitatingly, in some recent publications (Gąsiorowski 2012: 125,
fn. 14, Matasović 2014: 23). On the other hand, Tijmen Pronk has argued (apud
Kortlandt 2013: 14) that the Slavic word is unlikely to have been borrowed
because it has the same stem formation as in Baltic and because of the par-
allel between PSlav. *gǫserъ m. ‘gander’ (Cz. houser; Pol. gąsior; Sln. gosę́r; Bulg.
gắser) and Lat. ānser m. ‘goose’.4
– Skt. ghraṃsá- m. ‘blaze of the sun, summer heat’ (RV, AV, KauśS);
– MW gwres m. ‘heat (of the sun, fire), passion, lust’ (< *gwhrenso-); OIr. grís
‘heat, fire, embers, hot ashes’ (< *gwhrēnso-).
In an Atharvavedic formula, attested at AVŚ 7.18.2 = AVP 20.4.7,5 ná ghráṃs tat-
āpa ná himó jaghāna ‘not heat burned, not cold smote’ (Whitney), we encoun-
ter the root noun ghráṃs-, which is explained by Debrunner and Wackernagel
as due to a kind of haplology from ghraṃsás (Debrunner & Wackernagel 1930:
80f.), but this is improbable: the root noun is likely to be old. OIr. grís also indi-
rectly points to an athematic *gwhrēns with monosyllabic lengthening.
Although it is universally accepted that *gwhrenso- is a derivative of the root
*gwher- ‘to be hot’ (cf. Gr. θέρομαι ‘to become warm’, OCS grějati ‘to warm’, OCS
gorěti ‘to burn’, MIr. guirid ‘to warm’, etc.), its formation is considered unclear.
For instance, Mayrhofer writes in EWAia: 519: “Die Bildungsweise is nicht klar”,
with reference to Schindler 1972: 17.
2 Lucien van Beek suggests to me (p.c.) that the reason for the ubiquitous stem *ǵhh₂ens- may
be due to the fact that the word for ‘goose’ was very often used in the plural.
3 Depalatalization in the position before a vocalic nasal did not occur in Balto-Slavic (Kortlandt
2013). Depalatalization in Slavic in the position before a laryngeal may be attested in the word
for ‘grey’, cf. Cz. šerý, Pol. szary < PSlav. *śěrъ < *xěrъ < *ḱHoiro- (Derksen 2008: 447), if this is
not a loanword from Germanic, as suggested by Kroonen (2013: 201).
4 On the word for ‘gander’ see Gąsiorowski 2012: 125, however. It cannot be excluded that the
Slavic word was also borrowed from Germanic (cf. MHG ganzer).
5 For the AVP text, see Kubisch 2012: 39.
– Hitt. daššu- ‘strong, powerful; heavy; well-fed; difficult; important’ < *dens-u-
(Kloekhorst 2008: 854);
– Lat. dēnsus [o/ā] ‘dense, thick, closely packed’ < *dens-o- / *dn̥ s-o-;
– Gr. δασύς ‘hairy, thick with leaves; aspirated’, Gr. δαυλός ‘thick, shaggy’ <
*dn̥ su-(lo-).
The Greek forms have recently been discussed by Lucien van Beek (2013: 250 f.),
who comes to the conclusion that Gr. δασύς must have got its -s- from the forms
with full grade in the root (*densu-).6 This presupposes an alternating paradigm
*dens-u-, dn̥ s-eu- within Greek. As to the semantics, van Beek assumes that all
meanings of Hittite, Latin and Greek adjectives can be explained from the orig-
inal meaning ‘dense’.
As demonstrated by Kloekhorst 2010, the cuneiform sign da is used in Old
Hittite texts for spelling clusters of a dental stop plus a laryngeal. Kloekhorst
did not treat Hitt. daššu- in his article because this word does not occur in Old
Hittite, but since the Middle Hittite texts show the same distribution and since
Hitt. daššu- is consistently spelled with da-, with very few exceptions,7 we are
bound to reconstruct the root as *dHens-. It seems then attractive to derive this
adjective from the root *deh₁- ‘to bind’ (Gr. δέω, Skt. -dyati) and to assume that
its original meaning was ‘bound, bundled’ and from there ‘firm, dense, strong’.
We may now consider the relationship of our adjective with the IE verb
*dens-, which is glossed in LIV2 as ‘kundig werden, kunstfertig werden’. In
the active, this verbal root means ‘to instruct, teach, make capable’, cf. Vedic
daṃsáyas 2sg.inj.act. (RV), daṃsayantu 3pl.impv.act. (AVP), OAv. didąs
3sg.inj.act., Gr. διδάσκω, aor. δέδαε, whereas in the middle it means ‘to learn,
be instructed’, cf. OAv. dīdaiŋ́ hē 1sg.pres.med., Gr. διδάσκομαι, aor. δαῆναι. These
meanings may have easily developed from ‘to make or to become able, strong’,
so that the verb is likely to be denominal in origin, derived from the adjective
*dh₁ens- ‘firm, dense, strong’.8
The nominal character of the root *d(h₁)ens- further follows from the fact
that the root is part of the Caland system, forming adjectives in *-mo- (Skt.
dasmá- ‘wondrous, masterly’, OAv. dahma- ‘wondrous, miraculous’) and *-ro-
(Skt. dasrá- ‘accomplishing wonderful deeds’, Av. daŋra- ‘wise, capable’); the
superlative in *-is-tHo- (Skt. dáṃsiṣṭha-, YAv. dąhišta-), an abstract in *-es- (Skt.
dáṃsas- n. ‘miraculous ability’, YAv. (hizuuō) daŋhah- n. ‘miraculous power,
dexterity (of the tongue)’, Gr. δήνεα n.pl. ‘counsels, plans’), and a compound
form in *-i- (Gr. δαΐφρων ‘artful, experienced’).
– PIIr. *trHas: Skt. tirás prep. ‘through, across, beyond, apart from’ (RV+), YAv.
tarō prep. ‘through, across, except’, OAv. tarə̄ -maiti- f. ‘overconfidence, pride’,
OP t[r] /tara/ prep. ‘through’ (DZc 12);
– PIt. *trāns: Lat. trāns prep. ‘across, through’, Umb. trahaf prep. ‘on the other
side of’;
– OIr. trá adv., conj. ‘then, therefore; but’, MW tra prep. ‘beyond, over, across’.
Although it is universally recognized that these Indo-Iranian and Italo-Celtic
prepositions are derived from the root *terh₂- ‘to go through, cross through’,
they are usually interpreted in a different fashion: IIr. *trHas is assumed to
reflect *trh₂-os (EWAia: 646), in parallel to PIE *prH-os (Skt. purás ‘in front,
before’, YAv. parō ‘before’, Gr. πάρος ‘before, formerly’) whereas PIt. *trāns,
together with OIr. trá and MW tra (for which see Zair 2012: 179), is usually inter-
preted as a nom.sg. m. of the PIE participle *trh₂-nt-s ‘crossing’ (cf. de Vaan
2008: 627 for a discussion and references).
It would, of course, be preferable to find a unified explanation for these
prepositions, the more so as the proposed etymological analysis of the Italo-
Celtic forms seems rather strained to me. The development from a nom.sg. m.
of a participle to a preposition would be fairly unusual, and, secondly, the for-
mation (with its double zero grade) and the meaning (for which see below) is
not what we would expect from an aorist (?) participle.
On the other hand, the reconstruction *trh₂ns accounts for all the forms. The
difference in vocalization is the same as in the word for ‘wind’, PIE *h₂ueh₁nto-,
where PIIr. has vocalised the n before the loss of the laryngeal (PIIr. *HuaHata-
> Skt. vā ́ta-, OAv. vāta- /vaʔata-/), whereas the order of developments was the
opposite in other languages (Lat. uentus m. ‘wind’; Goth. winds ‘wind’; Toch. A
want ‘wind’; Toch. B yente ‘wind’). The same vocalization pattern is found with
the word for ‘moon’, for which see below.
The double zero grade of *trh₂ns is probably due to its use as a preposition.
It is conceivable that OIr. tar prep. ‘over, across’ has preserved the form with the
expected full grade of the suffix, *trh₂-ens, although functionally, OIr. tar rather
matches MW tra.
The unextended IE verb *terh₂- is only preserved in Indo-Iranian (for the
apparent Hittite forms see Kloekhorst 2008: 835ff.), and it is basically transitive
there, meaning ‘to cross smth.’. Also the IE -u-present *terh₂-u- ‘to prevail, over-
come’ is transitive. Since the Indo-Iranian and Italo-Celtic prepositions refer to
the completed crossing, ‘beyond, across’, we can assume that the original mean-
ing of *trh₂ns was ‘crossed over, gone through’.
– Gr. (Att.) μήν m., gen.sg. μηνός ‘month, moon’, (Ion.) μείς, (Lesb.) gen.sg. μῆν-
νος;
– Lat. mēnsis m. ‘month’;
– OIr. mí m.f. ‘month’, MW mis m. ‘id.’ < PCelt. *mī(n)s-;
– PIIr. *maHas-: Skt. mā ́s- m. ‘moon, month’ (RV+), mā ́sa- m. ‘month’ (RV+),
omās(i)ya- ‘… months old, … monthly’ (AV+) || PIr. *maHah-: OAv. nom.sg. mā ̊
/maHah/, YAv. gen.sg. mā ̊ŋhō m. ‘moon’, mā ̊ŋha- m. ‘moon, month’; omāhiia-
‘… monthly’, OP māh- ‘month’; Sogd. m’γ ‘moon, month’; Oss. mæj/mæjæ
‘moon, month’ (< *māhi̯ā);
– PAlb. *mēn-: Alb. muaj / mu(e)j m. ‘month’;
– Arm. amis, gen.sg. ams-o-y ‘month’ (the initial a- is due to the influence of
am ‘year’);
– OCS měsęcь m. ‘moon, month’;
– Toch. A mañ ‘moon, month’; Toch. B meñe ‘moon, month’;
– Lith. mėń uo (acc.sg. mėń esį) ‘moon, month’; Latv. mẽness ‘moon’, mẽnesis
‘month’; OPr. menig (EV) ‘moon’;
– PGerm. *mēnan-: Goth. mena m. ‘moon’, ON máni m. ‘id.’, OE mōna m. ‘id.’,
etc.;
– PGerm. *mēnōþ-: Go. menoþs m. ‘month’, ON mánaðr m. ‘id.’, OE mōnað m.
‘id.’, etc.
In the recent literature, the connection of *meh₁-ns- with the root *meh₁- ‘to
measure’—under the assumption that its original meaning was ‘measure (of
time)’—is not questioned, but the original inflection is debated. PIE *meh₁-
ns- is different from the other formations in -ns- in that the root has full grade
and the suffix zero grade and that the full grade of the suffix in some languages
seems to appear as -nes-, rather than *-ens-.
Most languages point to *meh₁-ns- without any ablaut alternations. This is
clearly the case for Greek, Latin, Celtic, Indo-Iranian (with early vocalization
of n, cf. on *trh₂ns above), Albanian and Armenian. Slavic can also go back to
this stem where -n- was dissimilatorily lost because of an n in the suffix. Tochar-
ian may have lost the final *-s in the nom. *meh₁ns >9 PToch. *men, which then
joined the n-stems.
In order to account for the Germanic and Baltic forms, we can reconstruct a
PIE suppletive paradigm nom.sg. *meh₁nōt, acc.sg. *meh₁nes-m, gen.sg.
*meh₁ns-es (cf. Beekes 1982),10 but even this highly irregular paradigm would
require a lot of restructuring before we get the Germanic and Balto-Slavic
facts right. Therefore, it seems worthwhile to investigate whether a single stem
*meh₁ns- can account for the Germanic and Balto-Slavic forms, too.
Traditionally (cf. already Brugmann 1911: 126, fn. 1, 128), the Germanic forms
are explained out of a single t-stem paradigm—nom.sg. *mēnō, gen.sg.
*mēnōdiz—with subsequent split into two paradigms and concomitant seman-
tic specialization: on the basis of the nominative, an n-stem with the meaning
‘moon’ was created,11 whereas the oblique cases in -ōd- got a new nominative
and the meaning ‘month’. Brugmann (op.cit.: 128) further draws attention to
OE mónaþfyllen ‘time of full moon’ in order to show that the original t-stem
had both meanings.
This is an elegant and economical explanation, but there are still two ques-
tions to be answered: (1) what happened with the final -s of *meh₁ns-, and (2)
what is the origin of the Germanic suffix *-ōd-? We can envisage the following
scenario. At an early stage of Germanic, before the operation of Osthoff’s Law,
nom.sg. *meh₁ns could have been reanalyzed as the stem *meh₁n- + the end-
ing -s. The new stem *meh₁n- was enlarged with the suffix *-ot-, which is also
found in PGerm. *leuh-ad-a- n. ‘light’ (cf. Hitt. lukkatt- c. ‘dawn, next morning’,
also Hitt. šīu̯ att- c. ‘day’) and may have spread from there to the word for ‘moon,
month’. The nominative was *-ōt with a long vowel12 that then spread through
the whole paradigm.
We may suggest a somewhat similar scenario for Balto-Slavic. The gen.sg.
*meh₁ns-es was interpreted as the genitive of a hysterodynamic s-stem, in spite
of the fact that there are hardly any traces of this inflection left in Balto-Slavic
languages. It seems conceivable that the word for ‘dawn’, which was a promi-
nent hysterodynamic s-stem in PIE (nom.sg. *h₂éus-ōs, acc.sg. *h₂eus-ós-m,
gen.sg. *h₂us-s-és), has influenced the Baltic word for ‘moon, month’.
3 Verbs
As we have seen above with PIE *dh₁-ens- ‘dense’, it can sometimes happen that
a derivative in *-ens- becomes a verbal root. In the following we shall look at a
few IE verbal roots in -ens- from this perspective: can they possibly have the
same origin?
– PIIr. *dhuans-: Skt. dhvaṃs- ‘to pulverize, crumble’ (RV): Pres. I dhváṃsate
(AVP+), Pres. X dhvasayaḥ (RV); a-aor. -dhvasán (RV VIII1), pf. -dadhvase
(RV), dhvastá- (AVP+); ava-dhvaṃsá- m. ‘sprinkling’ (AV); dhvasáni- adj.
‘sprinkling (cloud)’ (RV); dhvasirá- ‘covered with dust’ (RV); dhvasrá- ‘ob-
scured’ (RV+); dhvasmán- m. ‘polluting’ (RV)
– ?ToB tänts- ‘to scatter, disperse, tear off’ < *dhuns- (Adams 2013: 307 f.);
– PGerm. *dunsta- (OE dūst n. ‘powder, dust’, MDu. donst, dunst ‘fluff, pollen’,
etc.)
As duly recognized by Mayrhofer (EWAia: 800), *dhuens- must be related to the
PIE root for ‘to shake’, *dheuH- (Skt. dhūnoti ‘to shake, move to and fro quickly’
(RV+), YAv. auui frā-δauuaite ‘to rub’, Gr. θῡ́ω ‘to dash, rush’, OIc. dýja ‘to tremble’,
etc.). Mayrhofer (loc.cit.) analyzes *dhuens- as *dhu̯ -en-s-, presumably assuming
two suffixes or enlargements, which does not help in elucidating its formation,
however.
In my view, we must seriously consider the possibility that the root contains
the suffix *-ens-, i.e. *dhuH-ens-. In Sanskrit, the laryngeal would probably dis-
appear quite early in this position or, at least, not be recoverable.13 In Germanic
we cannot see the difference, and in Tocharian, the laryngeal would probably
be lost in full grade *dhuHens- and then eliminated in zero-grade.
The root *dhuens- does not look like an archaic verbal root, all finite forms
being relatively productive,14 so that it can easily be of denominal origin.
13 Cf. Lubotsky 1997: 149ff. A possible trace of the laryngeal may be seen in the jagatī line
RV 10.113.7c, if we scan 3sg. middle pf. dadhvase as /dadhuvase/: dhvāntáṃ támó ‘áva dad-
huvase haté-. This scansion, however, creates a short 8th syllable, which is irregular. Most
probably, we simply have an 11-syllable jagatī line here, cf. Oldenberg 1912: 335.
14 Denominal verbs that are plain thematic, without a specific denominal suffix, are admit-
tedly rare, but not unknown: for instance, Skt. dyótate ‘to shine’ (RV+) is clearly derived
from a nominal t-stem (Hitt. šīu̯ att- c. ‘day’, Skt. dyút- f. ‘shine’, vi-dyút- f. ‘lightning’). Note
that this verb even attests a root aorist: 2.3sg. ádyaut, ptc. dyutant- (in dyutád-yāman- ‘with
shining driving’ (RV), dyutá(d)dyu- ‘with shining days’ (MS)), ptc.med. dyútāna-, dyutāná-
(RV+).
Semantically, this also makes sense: a derivative of the root *dheuH- ‘to shake,
rub’ with the suffix *-ens- would mean ‘shaken, rubbed off = dust, powder’, and
the denominal verb would mean ‘to dust, to powder’.
– Skt. śaṃs- ‘to praise, recite, declare’ (forms in the RV are: Pres. I śáṃsāmi,
iṣ-aorist áśaṃsīt, mediopass. aor.: śaṃsi, pass. śasyáte, caus. śaṃsaya, ta-
ptc.: śastá-, inf. anu-śáse, vi-śáse); śastí- f. ‘song of praise’; prá-śasti- f. ‘praise,
fame; instruction, guidance’; uktha-śaṃsín- ‘praising’, śáṃstar- m. ‘reciter’;
śáṃsya- ger. ‘to be recited’, śastí- f. ‘song of praise’, śásman- n. ‘praise’; śáṃsa-
m. ‘praise, judgment’.
– PIr. *sanh-: Av. saŋh- ‘to declare’: pres. YAv. saŋhāmi, OAv. sə̄ṇghaitī, aor.opt.
OAv. sax́ iiāt̰, inf. sastē, sazdiiāi, ta-ptc. YAv. aiβi.sasta-; OAv. sąstra- n. ‘teach-
ing’, sə̄ṇghana- n. ‘teaching, doctrine’, sə̄ṇghu- m. ‘preaching, doctrine’, Av.
fra-sasti- f. ‘fame, reputation, prestige’; YAv. sasti- f. ‘praise’; OAv. sə̄ṇgha- m.
‘declaration, judgment’ || OP θanh- ‘to declare, announce, call’: pres. aθanha,
pass. θa(n)hạyāmahạiy, inf. θa(n)stanaiy.
– Lat. cēnseō ‘to estimate, think; decide’.
The IE root *ḱens- ‘to declare’ can hardly be separated from two other, semanti-
cally very close, roots, viz. *ḱeH- and *ḱeHs-. The former root, which must be
the underived basis of the latter, is attested in Old Persian θā- ‘to proclaim’
(3sg.pres. ϑātiy, 2sg.subj. ϑāhạy, 2sg. impv. ϑādiy) and Alb. thom ‘to say’. If Gr.
ἀκήν ‘silently’ belongs here,15 the color of the laryngeal must be *h₁. The root
*ḱeHs- is found in Skt. śās- ‘to teach, chasten, command, order’, Av. sāh- ‘to
teach, instruct, command’; ToA kāṣ-iññ- ‘to scold, chasten’ (< *ḱHs-), Goth. haz-
jan ‘to praise’ (< *ḱHs-ie-), OHG harēn ‘to cry, call’ (< *ḱHs-eh₁-).
Assuming the suffix *-ens- allows us to analyze *ḱens- as *ḱh₁-ens- and direct-
ly relate it to the other two roots. The noun *ḱh₁-ens- may have had the meaning
‘smth. said, pronounced → statement, pronouncement’, and the verb derived
from it ‘to give a statement’, which is quite appropriate for the Latin and Indo-
Iranian verbs. This noun *ḱh₁-ens- may live forth in PIIr. *ćansa-, attested in Skt.
śáṃsa- m ‘praise, opinion, judgment’ (RV+); OAv. sə̄ṇgha- m. ‘pronouncement,
judgment’, YAv. saŋha- m. ‘prescription’; and, possibly, Khot. saṃja- ‘document’
(if from *sanha-čī-, cf. Bailey 1979: 417). Of course, formations in -a- have always
been productive in Indo-Iranian, so that it is difficult to prove that *ćansa- is
very old, but the precise correspondences in semantics and formulaic diction
15 Gr. (Pi.) ἀκᾶ, ἀκᾷ ‘id.’ can be hyperdorisms, cf. García-Ramón 1993: 127.
between Indo-Aryan and Iranian point to the archaic nature of this word. Suf-
fice it to mention the compounds and formulas given by Mayrhofer (EWAia:
s.v.): Skt. duḥ-śáṃsa- ‘wishing evil, malicious’ (RV+) ~ YAv. duš.saŋha- ‘slan-
dering’; Skt. várdhān naḥ śáṃsam ‘he will strengthen our praise’ (RV 5.41.9)
~ OAv. vərəzdāiš sə̄ṇghāiš ‘with strong teachings’ (Y 46.19); Skt. uru-śáṃsa- ‘of
far-reaching authority’ (RV) ~ OP *varu-θanha- (Hdt. ὀροσάγγαι ‘benefactors of
the king’); Skt. nárā-śáṃsa- m. epithet of fire (“who receives men’s praise”),
nárā … śáṃsam (in tmesis); narā ́ṃ ná śáṃsaḥ, śáṃso narā ́ṃ (RV+) ~ YAv.
nairiiō.saŋha- m. N. of a god, of the divine messenger, N. of the fire as a god,
also gen.sg. nairiiehe … saŋhahe, acc.sg. nairīm … saŋhəm.
– Skt. sraṃs- ‘to fall down, slip off’ (in the RV only a-aor. srasema and root-
nouns ava-srás- f. ‘slipping down’, vi-srás- f. ‘decay, dissolving’; in the AV
followed by pres. I middle 3sg.impv. sraṃsatām (AVP), caus. sraṃsayāmi,
redupl. aor. asisrasat, ta-ptc. -srasta-, and nouns sanisrasá- ‘defective, crip-
pled, weak’; asthisraṃsá- ‘causing the bones to fall asunder’);
– ?PIr. *(h)rah-: OAv. 3pl.caus.act. rā ̊ŋhaiiən ‘to deflect’ (Y 32.12); YAv. raŋha-
‘suffering of epilepsy (?)’ (Yt 5.93); (haca) ϑraŋhibiia du. ‘corners of the
mouth (?)’ (P 27 (28)).
The Iranian cognates are unfortunately very uncertain. Mayrhofer (EWAia:
s.v.) and Cheung (2007: s.v. *(h)rah) hesitatingly connect Av. rah- ‘to alienate,
deflect’, but this root has no nasal and an initial laryngeal (see Beekes 1979).
From the point of view of semantics, PIIr. *srans- clearly belongs together with
Skt. sridh- ‘to fail, err’ (RV) < PIE *sleidh- (OE slīdan ‘to slide, slip’, Lith. slýsti
‘to slip, slide, glide’, etc.) and Skt. srīv- ‘to be aborted, miscarried’ (RV+). As I
argued elsewhere (Lubotsky 2011: 119f.), these roots are enlargements of PIE
*sel- ‘to jump’, and *srans- can also be seen as an enlargement of the same root
with the suffix -ens-.
4 Conclusions
Let us now summarize our findings. The IE nominal suffix *-ens- is found in
*ǵhh₂-ens- ‘goose’, *gwhr-ens- ‘heat’, *dh₁-ens- ‘dense, strong’, *trh₂-(e)ns ‘across’,
*meh₁-ns- ‘moon, month’. Furthermore, the same suffix can be recognized in
the verbal roots *dh₁-ens- ‘to teach, make capable’, *dhu(H)-ens- ‘to pulverize’,
*ḱh₁-ens- ‘to recite, declare’, and IIr. *sr-ans- ‘to fall down, slip off’, all of which
are then likely to be of denominal origin.
What was the meaning of the suffix? As far as we can see, the suffix had
an adjectival function, except for *gwhr-ens- ‘heat’, which may have then devel-
oped out of ‘hot (sun)’. In *ǵhh₂-ens- ‘goose’ (= ‘gaping’), the adjective is agen-
tive, whereas it is patientive in *dh₁-ens- ‘dense, strong’ (= ‘bound’), *trh₂-(e)ns
‘across’ (= ‘crossed over’) and *meh₁-ns- ‘moon, month’ (= ‘measure, measured
time’). For the denominal verbs it is, of course, much more difficult to decide
what was the meaning of the basic noun, but, as we have seen above, *dhu(H)-
ens- ‘to pulverize’ is likely to be based on ‘powder’ = ‘rubbed off’, and *ḱh₁-ens- is
based on ‘statement’ = ‘pronounced’. It seems then that the meaning depends
on the transitivity of the verb: if the verb is transitive, the -ens-derivative is
patientive; otherwise, it is agentive. This distribution is also found with the
Hittite participles in -ant- and is usually assumed to be original for the IE
*-nt-participles.
These considerations make it probable that the two suffixes, *-ens- and
*-ent-, once belonged to one and the same paradigm, with an alternation s/t
that we also find in the suffix of the IE perfect participle *-uos-/-uot-. As was
argued by Kortlandt (2002 = 2010: 397, who developed an earlier suggestion of
Bojan Čop), Indo-Uralic *ti has become assibilated to *si. This would mean that
the suffix *-ens started as *-ensi < *-ent-i and reflects the original locative of the
Indo-Uralic participle.
The IE suffix *-ens-/-ent- might be identified with the Uralic nominal suf-
fix *nt (Collinder 1960: 269), cf. the Proto-Samoyed present participles in *-ntV
(Mikola 1988: 259), the Saami absolute gerund, etc. This Uralic participle suf-
fix may or may not be identical to the suffix *nt for deverbative verbs, which
“usually implies a continuative mood of action or a non-perfective aspect”
(Collinder 1960: 277).
References
Adams, D.Q. 2013. A Dictionary of Tocharian B. Revised and Greatly Enlarged Edition.
Amsterdam–Atlanta: Rodopi.
Bailey, H.W. 1979. Dictionary of Khotan Saka. Cambridge: Cambridge University Press.
Beekes, R.S.P. 1979. GAv. uzirǝidyāi and rārǝša-. Münchener Studien zur Sprachwis-
senschaft 38, 9–20.
Beekes, R.S.P. 1982. GAv. mā ̊, the PIE word for ‘moon, month’, and the perfect participle.
Journal of Indo-European Studies 10, 53–64.
Brugmann, K. 1911. Vergleichende Laut-, Stammbildungs- und Flexionslehre der indoger-
manischen Sprachen. 2. Band, 2. Teil: Lehre von den Wortformen und ihrem Gebrauch.
Strassburg: Trübner.
Headedness in Indo-Uralic
Rosemarie Lühr
Though Old Hungarian word order is, in general, flexible, the occurrence of a
morphologically unmarked object is always accompanied by a head-final OV
order.
A similar construction is the one with a participle on suffix -uan/uen (Mod-
ern Hungarian -ván/vén):
Further evidence for an old SOV type comes from the verb-auxiliary order in
Old Hungarian. The auxiliary always immediately follows the verb:
Also the other mentioned word orders being connected with SOV order match
the criteria of such a language type. The nominal modifiers such as adjectival
and genitival expressions precede the head noun (Lehmann 1973: 48); cf. for an
adjective:
1 There are also prenominal non-finite relative clauses in present-day Khanty (Nikolaeva 1999:
79; Bácskai-Atkári & Dékány 2014: 44; Csepregi 2012).
In Hittite there are different constructions. Dynamic place words with dative/
locative or allative appear always in front of this case form, anda parna ‘into
the house’, and are considered as pure adverbs (Brosch 2013: 398). In static con-
structions place words are postposed:
(10) a. É-ri andan and LUGAL-i peran ‘in front of the king’ (with dative/loca-
tive)
b. LUGAL-u̯ aš peran ‘in front of the king’ (with genitive)
c. peran(n)=mit ‘in front of me’ (with enclitic possessive pronoun) (Tjerk-
stra 2000: 6f.)
and preposed:
2.1.2 Vedic
In Vedic prepositions besides postpositions can be found; for prepositions
cf.:2
(13) RV 10,86,4
ś(u)vā ́ nú asya jambhiṣat
dog(M): NOM.SG now he: GEN.M.SG bite: AOR.SUBJ.ACT3SG
ápi kárṇe
in: PREP ear(M): LOC.SG
‘Soon may the dog bite him in the ear.’
2 However, according to Casaretto (2014: 59) in the R̥ gveda no prepositions are attested. Cf.
further Casaretto (2011): in an adnominal construction ví is an attributive or appositive satel-
lite.
(15) RV 10,51,6
agnéḥ pū ́ rve bhrā ́taraḥ
Agni: GEN former: NOM.M.PL brother(M): NOM.PL
ártham etám rathī ́ iva
object(M): ACC.SG this: ACC.M.SG car driver (M): NOM.SG like
ádhvānam ánu ā ́
path(M): ACC.SG along: POSTP back and forth: PFX
avarīvuḥ
move: INTENS.IPF.IND.ACT3PL
‘Agni’s elder brothers moved this object like a car driver along the path
back and forth.’
Thus, while Vedic has prepositions and postpositions, Hittite has postposi-
tions.
2.2.1 Hittite
In Hittite, attributive adjectives mostly appear in prenominal slots (Melchert
2003: 200; Laroche 1982: 134; Francia 2001: 91; Bauer 2015: 232–235).
But quantifier adjectives are postposed:
Also Hittite dapiant- ‘all, entire’ regularly follows its head noun.
Contrary to the communis opinio,3 I assume that these examples are in-
stances of an older quantifier floating. As English and German show, quantifier
floating is possible only with quantifiers that require a definite noun.4
(18) The Beatles and the Stones each made many hit records.
In English quantifiers other than all, both, and each cannot be moved. But in
Hittite, also the semantically related attributive participle šuwant- ‘filled’ is
postposed:
(20) KUB 30.10 obv. 27′ (Hoffner & Melchert 2008: 272)
ḫu-u-ma-an-te-et kar-di-it
‘with the whole heart’
If this structure was really the original one, postposition of ḫūmant- must have
been generalized.
However, preposition as in:
3 Yoshida (1987: 33) ascribes the postposition of ḫūmant- to its “unbestimmte Bedeutung” and
compares postposed kuelqa. According to Hoffner & Melchert (2008: 271) postposition of
ḫūmant- and dapiant- may be due to their meaning or to their formal resemblance to par-
ticiples in -ant-.
4 In English, only subject-related quantifiers can be separated from the subject and appear in
more than one position in a sentence (Hoeksema 1996; Maling 1976).
that emphasis also plays a crucial role with quantifier floating. As this phe-
nomenon is normally connected to stress it can be assumed that postposed
ḫūmant- and dapiant- are highlighted, too.5
The following instances of postposed adjectives are different. (22) and (23)
exhibit genitival relational adjectives:
(22) Muwatalli (CTH 381 1,41) (cf. Bauer 2014: 246 f. referring to Semenza 2006:
561)
dU pí-ḫa-aš-ša-aš-ši-iš
Storm God belonging to lightning: NOM.C.SG
‘O Storm God of lightning’
As regards the position of adjectives, Hittite is broadly in line with the SOV-
type. Apart from some postposed relational adjectives and predominantly
2.2.2 Vedic
Therefore, let us first prove the position of numerals in Vedic.
In Vedic there are examples for preposed numerals as well as for postposed
ones:
While in Hittite only quantifiers like ‘all’ and ‘every’ allow for quantifier float-
ing, in Vedic also numerals show floating into the position behind their head
noun. It is a remarkable typological feature for languages to vary in the extent
of quantifier floating.6
6 For example, while in Japanese numeral quantifiers are licensed to be floating, Chinese
numeral quantifiers are not (Kobuchi-Philipp 2003; Fitzpatrick 2006).
(29) RV 3.6.6
áthā ́ vaha devā ́n deva víśvān
‘Bring here all gods, o God.’
But as in Vedic also other adjectives than numerals and quantifiers are post-
posed, one has to examine whether this language belongs to those languages
which allow for prenominal and postnominal attributive adjectives whereby
the two word orders establish a conceptual difference. Such a language is
English for instance; cf. Bolinger’s (1967) example:
vs.
However, there are also a lot of text passages where adjectives denoting an
individual-level predicate are postposed:
There are not only adjectives following a vocative as in (33)
Another adjective use is documented in (35). The quantifier éka- ‘one’ appears
behind its head noun und is surely stressed.
Therefore, it can be supposed that in the Vedic NP a postnominal slot for focal-
ized adjectives exists independent whether the adjective is an individual-level
or stage-level predicate. This means that postnominality is a focus strategy.
By apposition-like structures the speaker wants to highlight an adjective the
meaning of which is unexpected or in another way important in the con-
text.
Hence, Vedic and Hittite differ fundamentally in the positioning of adjec-
tives. While Hittite has preposed adjectives apart from quantifiers and some
relational adjectives, Vedic has both prenominal und postnominal ones, where-
by the semantic class is not significant.
2.3.1 Hittite
So far, Hittite is of the second type: as for the position of genitives, Hittite nearly
fulfills the conditions of a SOV language. In the normal word order the genitive
precedes its head noun (Hoffner & Melchert 2008: 254), except for the genitive
of measure and for the genitive of material.
Phrases like (40) and (41) are always definite nominal phrases. The posses-
sum is either a sortal noun like house, or it refers to parts of entities like blood,
head, eyes, or to persons. In any case the referent of the possessum is uniquely
identifiable by the preposed genitive (Loebner 1985; Lühr 2002c). Hence, the
substantival genitive to the left of the possessum establishes the reference of
this word. On the contrary, the use of the genitive of material to the right of
the head noun (Yoshida 1987: 32, 75) is a signal that a specific interpretation is
disallowed (cf. Lühr 2004); cf.:
As word order does not have an influence on the reference of the whole nomi-
nal phrase the genitive of material can be postposed in Hittite.
2.3.2 Vedic
Turning to Vedic, only stressed pronouns denoting primarily a possessor pre-
cede the head noun, while unstressed ones appear in the Wackernagel position:
Also in vocative constructions with an objective genitive the head noun ap-
pears first:
Often the word order is inversed by hyperbaton. In (46) the head noun is in
front of the possessive genitive:
Comparing the Hittite and Vedic data containing an adnominal genitive, strik-
ing differences were stated. While in Hittite postposition of the genitive is
nearly an exception, Vedic allows for postposition of all kinds of genitives.
Often information structure is the motor for extraposition of the genitive, espe-
cially the hyperbaton, as well as a heavy weight of the genitival noun phrase or
vocative constructions with the head noun in the first position. As mentioned,
there are also records where none of these explanations holds. More research
is required.
If Khanty and Mansi have preserved the basic SOV order of Proto-Uralic, then
the change from SOV to Topic Focus Verb X* must have taken place in Hungar-
ian separately (Kiss 2013).7
Comparing the Hittite and Vedic data with this development, we examine the
position of subject and topic (see tables 12.1 and 12.2). A distinction is made
between a continuing, shifting, and contrastive topic. If the topic is a pronomi-
nal subject, it can be covertly enclosed in the verb. We only take main clauses
(= MC) into consideration.
The numbers clearly show that in Hittite the distribution of subjects in first
and second position is almost even, and in Old Indic a subject phrase is the
most common in initial position. When functioning as topic, the subject is usu-
ally a shifting topic. But a topic-subject, which is covert as a result of pro-drop,
is mainly a continuing topic.
However, both in Hittite and in Old Indic the records of subject-topic-
identity are too rare to give a hint on a development towards a discourse con-
figurational structure of the Hungarian type. Hittite and Old Indic are syntax
configurational languages (Lühr 2015).
7 According to Polo (2005) rightward extraposition can be responsible for the change from SOV
to SVO. An example is Latin.
Subjects as Topics in MC
Topics as Subject in MC 323 99 422
thereof continuing topic 97 29 126
thereof shifting topic 188 59 247
thereof contrastive topic 33 10 43
covert Topic-Subjects in MC 349 73 422
thereof continuing topic 178 34 212
thereof shifting topic 165 39 204
thereof contrastive topic 4 0 4
4 Conclusion
As the comparison of head structures in Hittite and Vedic shows, Hittite has
postpositions, Vedic pre- and postpositions, the adjective appears in Hittite
mostly in front of the head noun as well as the genitives. Exceptions are oper-
ators like hūmant- ‘all’, some relational adjectives and genitives of measure. By
contrast, in Vedic the position of adjectives and genitives fluctuates, also the
position of relative clauses does not agree with Hittite.
Of these two languages Hittite has more common features with an underly-
ing SOV type represented by Proto-Hungarian.
Including verb placement Hittite actually comes even closer to the SOV type:
In Hittite the finite verb appears mostly at the end of the clause:
483 10 6 7
On the contrary, the relations are ambiguous in Old Indic, especially in Vedic:
169 64 71 118
70 8 13 9
Value Representation
References
Avery, J. 1881. On relative clauses in the Rigveda. Proceedings of the American Oriental
Society 11, 64–66.
Bácskai-Atkári, J. 2013. The diachronic system of the left periphery of subordinate
clauses in Hungarian. In: B. Surányi (ed.), Proceedings of the Second Central Euro-
pean Conference in Linguistics for Postgraduate Students. Budapest: Pázmány Péter
Catholic University, 3–23.
Bácskai-Atkári, J.; É. Dékány 2014. From non-finite to finite subordination. The history
of embedded clauses. In: K.É. Kiss (ed.), The Evolution of Functional Left Peripheries
in Hungarian Syntax. Oxford: Oxford University Press, 147–299.
Bauer, A.H. 2014. Morphosyntax of the Noun Phrase in Hieroglyphic Luwian. Leiden–
Boston: Brill.
Bolinger, D. 1967. Adjectives in English: Attribution and predication. Lingua 18, 1–34.
Bosque, I.; Picallo, C. 1996. Postnominal adjectives in Spanish DP s. Journal of Linguis-
tics 32, 57–78.
Brosch, C. 2013. Untersuchungen zur hethitischen Raumgrammatik. Berlin–New York:
Mouton de Gruyter.
Brosch, C. 2014a. Eine Skizze der Räumlichkeit im Hethitischen. eTopoi. Journal for
Ancient Studies 3, 23–41.
Lehmann, C. 1984. Der Relativsatz: Typologie seiner Strukturen, Theorie seiner Funktio-
nen, Kompendium seiner Grammatik. Tübingen: Narr.
Loebner, S. 1985. Definites. Journal of Semantics 4, 279–326.
Lühr, R. 2001. Relativsätze im Hethitischen. In: G. Wilhelm (ed.), Akten des IV. Inter-
nationalen Kongresses für Hethithologie, Würzburg 4.–8. Oktober 1999. Wiesbaden:
Harrasowitz, 333–346.
Lühr, R. 2002a. Konzeptionierungen des Prädikativums in der Indogermania. Zeitschrift
für Sprachwissenschaft 21, 2–24.
Lühr, R. 2002b. Badal- und Genitivkonstruktionen. Historische Sprachforschung 115, 23–
36.
Lühr, R. 2002c. “Allgemeine Anaphora”. Zum Artikelgebrauch bei der Fügung “Substan-
tiv und adnominaler Genitiv” im Heliand. In: M. Fritz; S. Zeilfelder (eds.), Novalis
Indogermanica. Festschrift für Günter Neumann zum 80. Geburtstag. Graz: Leykam,
251–281.
Lühr, R. 2004. Der Ausdruck der Possessivität innerhalb der Determinansphrase
der ältesten indogermanischen Sprache. In: D. Groddek; S. Rößle (eds.), Šarnikzel:
Hethitologische Studien zum Gedenken an Emil Orgetorix Forrer. Dresden: Verlag der
TU Dresden, 415–446.
Lühr, R. 2005. Individuen- und Stadienprädikation im Altindischen. In: I. Balles; R. Lühr
(eds.), Indogermanische Nomina agentis. Leipzig: Institut für Linguistik der Univer-
sität Leipzig, 161–252.
Lühr, R. 2015. Traces of discourse configurationality in older Indo-European languages?
In: C. Viti (ed.), Perspectives on Historical Syntax. Amsterdam: John Benjamins, 203–
232.
Maling, J. 1976. Notes on quantifier postponing. Linguistic Inquiry 7, 708–718.
Marcantonio, A. 1985. On the definite vs. indefinite conjugation in Hungarian: A typo-
logical and diachronic analysis. Acta Linguistica Academiae Scientiarum Hungaricae
35, 267–298.
Melchert, H.C. 2003. Language. In: H.C. Melchert (ed.), The Luwians. Leiden–Boston:
Brill, 170–210.
Melchert, H.C. 2009. Local adverbs in Hittite: Synchrony and diachrony. Language and
Linguistics Compass 3/2, 607–620.
Nikolaeva, I. 1999. Ostyak. München: Lincom Europa.
Ott, D. 2011. A note on free relative clauses in the Theory of Phases. Linguistic Inquiry
42/1, 183–192.
Polo, C. 2005. Latin word order in generative perspective: An explanatory proposal
within the sentence domain. In: K.É. Kiss (ed.), Universal Grammar in the Recon-
struction of Ancient Languages. Berlin: Mouton de Gruyter, 373–428.
Probert, P. 2015. Early Greek Relative Clause. Oxford: Oxford University Press.
Raun, A. 1988. Proto-Uralic comparative historical morphosyntax. In: D. Sinor (ed.),
The Uralic Languages. Description, History and Foreign Influences. Leiden–New York:
Brill, 555–571.
Semenza, C. 20062. Impairments of proper and common names. In: K. Brown (ed.),
Encyclopedia of Language and Linguistics. Amsterdam: Elsevier, 561–564.
Speyer, J.S. 1896. Vedische und Sanskrit-Syntax. Strassburg: Trübner.
Tjerkstra, F.A. 2000. Principles of the Relation between Local Adverb, Verb and Sentence
Particle in Hittite. Groningen: Styx.
Viti, C. 2008. Genitive word order in Ancient Greek: A functional analysis of word order
freedom in the noun phrase. Glotta 84, 203–238.
Viti, C. 2015. Variation und Wandel in der Syntax der alten indogermanischen Sprachen.
Tübingen: Narr.
Vogel, P.M. 1996. Wortarten und Wortartenwechsel. Zu Konversion und verwandten Er-
scheinungen im Deutschen und in anderen Sprachen. Berlin–New York: Walter de
Gruyter.
Wągiel, M. 2014. From kinds to objects. Prenominal and postnominal adjectives in Pol-
ish. In: I. Veselovská; M. Janebová (eds.), Complex Visibles Out There: Proceedings of
the Olomouc Linguistic Colloquium 2014: Language Use and Linguistic Structure. Olo-
mouc: Palacký University, 457–476.
Yoshida, D. 1987. Die Syntax des althethitischen substantivischen Genitivs. Heidelberg:
Winter.
Zeilfelder, S. 2001. Archaismus und Ausgliederung. Studien zur sprachlichen Stellung des
Hethitischen. Heidelberg: Winter.
For the important but still disputed question of Indo-Anatolian the relevance
of the Tocharian branch is in my view not so much its informativeness on the
original state of affairs in Proto-Indo-European, since it has undergone so many
sound changes, and lost and replaced so much of the original lexicon. Rather,
its relevance is due to its presumed position in the Indo-European family tree:
it is often assumed to have been the second branch to split off after Anatolian,
as in the tree reconstructed by Ringe, Warnow & Taylor (2002: 87; see figure 13.1,
next page).
Several authors have argued that Tocharian was the second branch to split
off, e.g. Carling (2005: 48–49), Jasanoff (2003: 204), Kim (2007), Kortlandt (e.g.
2016: 81–82), Schindler (apud Jasanoff 2003: 46), Schmidt (1992) and Winter
(1997). Yet, the evidence is not overwhelming, and, strikingly, many authors
strongly differ in the arguments they adduce for this position of Tocharian in
* This research was first supported by a Marie Curie Intra European Fellowship within the
7th European Community Framework Programme (project number 626656) and then by the
Netherlands Organisation for Scientific Research (NWO, project number 276-70-028). This
article is an adaptation of a talk at The precursors of Proto-Indo-European: The Indo-Hittite
and Indo-Uralic hypotheses. 9–11 July 2015, Leiden University. I am grateful to the editors Alwin
Kloekhorst and Tijmen Pronk for valuable comments on an earlier version.
the family tree. There are also critical voices in the literature, for instance by
Malzahn (2016), who argues that the lexical arguments of e.g. Schmidt (1992)
and Winter (1997) are not strong and should not be used. I agree with Malzahn
that we need more and better evidence before we can consider the early split-
off of Tocharian proven, but that is no reason to discard the lexical evidence
that we have. With Kloekhorst (2008; see below), I consider the potential of
lexical evidence for subgrouping to be actually rather good. Although I admit
that more work needs to be done, the hypothesis that Tocharian was the second
to split off seems to me the most likely so far.
The possibility that Tocharian was the second branch to split off is rele-
vant for the Indo-Anatolian hypothesis. Since the Indo-Anatolian node in the
tree is defined by common innovations of the non-Anatolian branches, it is
only of value in contrast to the next node in the tree. Thus, if the next node
down from Indo-Anatolian is the node that we may term “Indo-Tocharian” (see
figure 13.2, next page), Indo-Anatolian can only be established in contrast to
Indo-Tocharian; if Tocharian evidence for a given common innovation of the
non-Anatolian languages is lacking, the position of Tocharian for this feature is
not clear, and what seems to be an argument for Indo-Anatolian could as well
be in fact an argument for Indo-Tocharian. Obviously, if Tocharian was not the
second branch to split off, the same argument applies to the branch that was in
fact the first to split off after Anatolian, for instance Italo-Celtic. In light of the
Indo-Tocharian hypothesis, therefore, evidence for Indo-Anatolian requires a
systematic check against Tocharian data.
Below, I will systematically review Kloekhorst’s seven arguments for Indo-
Anatolian (2008: 8–10), because these are in my view well selected and clear
evidence. The main purpose of this review is not to challenge Kloekhorst’s argu-
ments, but to illustrate the methodological case I want to make.
The Hittite reflex of *mer- means ‘disappear’, while in the other Indo-European
languages the meaning ‘die’ is found. While the shift of meaning of ‘disappear’
to ‘die’ is commonplace according to a well established metaphor, the opposite
semantic change is difficult to imagine.
In Tocharian, the verb for ‘die’ is wäl- in Tocharian A and srəwka- in Tocharian
B; there is so far no reflex of *mer-.1
1 Tocharian has the apparently related *mers- ‘forget’ as A mräsā- ‘forget’, B mərsa- ‘id.’; this verb
has not been influenced by the semantic shift of *mer- and is therefore irrelevant here.
The Hittite reflex of *seh₂- means ‘fill up, stuff’, but those of the other Indo-
European languages mean ‘satiate’. Again, a semantic change of ‘stuff’ to ‘sati-
ate’ is more likely than the converse.
Tocharian goes together with the other non-Anatolian languages: Tocharian
A säy(n)- ‘satiate; be satiated, depressed’, B soy- ‘be sated’, səyn- ‘be satiated,
depressed’.
4) HLuw. tuwatra/i- ‘daughter’ and Lyc. kbatra- ‘id.’ < *dhuégh₂tr, *dhugh₂térm
5) ḫarra-i / ḫarr- ‘grind, crush’ < *h₂erh₃- and ḫārš-i ‘harrow, till the soil’
Kloekhorst argues that Hittite ḫarra-i ‘grind, crush’ preserves the older mean-
ing, and that the meaning ‘plough’ of the same etymon in the non-Anatolian
languages is a common, later development.
The verb *h₂erh₃- is not attested as such in Tocharian. The regular expres-
sion for ‘plough’ seems to be TA pātā- ‘plough’, pate ‘ploughing’ < *bhodhh₂-
(Lat. fodiō, -ere ‘pierce, dig’, Hitt. padda-i, padd- ‘dig (the ground, a pit)’, OCS
bodǫ ‘stab’, Lith. bedù ‘stick, dig’). However, Tocharian A āreñ surely means
‘plough’. Apparently this is in origin the plural of an abstract noun āre ‘plough-
ing’ derived from a verb *ār- or *ārā-; this verb is obviously to be set up as *ārā-
< Proto-Tocharian *ara- < *h₂erh₃- (Peyrot 2018b: 262–263).
The replacement of the original verb for ‘plough’ by the verb for ‘dig’ is a little
peculiar. Perhaps the reason is the phonological merger of *ara- ‘plough’ with
*ara- ‘cease’ (Tocharian B ara- ‘cease’, A arā- ‘id.’).
Of these seven arguments for Indo-Anatolian, six easily stand the test because
Tocharian goes together with the other non-Anatolian branches. Only number
1), ‘disappear / die’, is not attested in Tocharian and should therefore strictly
speaking not be used as an argument for Indo-Anatolian because the change
of ‘disappear’ to ‘die’ could theoretically also have taken place in the core Indo-
European languages after Tocharian split off. Nevertheless, I do not think that
we should discard this argument for Indo-Anatolian completely. It is in fact a
good argument. We should just keep in mind that the position of Tocharian for
this item cannot so far be decided.
2 Obviously, this derivation has to cope with the difficulty that PU *x would correspond to
PIE *h₃ in ‘give’ but to *gwh in ‘drink’. However, a more serious problem, in my view, is dis-
cussed directly below. Kortlandt later offered a revised explanation for this etymon: “The
rounded laryngeal *q₃ of Indo-European *deq₃- < *toqi- suggests that the non-initial vowel
was rounded as a result of Indo-Uralic vowel harmony in this root.” (2002: 217–227). I inter-
pret this to mean that PIU (= PU) *toxi- > *toxu- > *texu- > *deqw- > PIE *deh₃-. In my view,
this revision only brings Proto-Indo-Uralic closer to Proto-Uralic. Whether this is the right
route to take is questionable, and I find his earlier derivation more convincing.
3 In the following, the symbols “i̮” and “e̮” denote a high unrounded back vowel (alternatively
sometimes “ï”, IPA “ɯ”) and a mid unrounded back vowel (alternatively sometimes “ë”, IPA
“ɤ”), respectively.
4 For Proto-Samoyedic a weak vowel is reconstructed that is noted with “ə̑” in Janhunen 1977,
“ɵ” in Sammallahti 1988 and “ø” in Janhunen 1998; here it is noted with “ə”.
5 4 out of 8 are monosyllabic roots ending in a vowel. *ni̮c-kəl- ‘tear’ may have been formed
after *ü-kəl- ‘drag’, cf. the variant *nüc- of *ni̮c-, probably through influence from *ü-. Perhaps
the suffix arose in roots in -k, of which there are two, and then spread to mostly monosyllabic
roots ending in a vowel.
6 The preservation of the initial *l in Samoyedic is irregular. The expected outcome is rather *je̮.
Conceivably, the *l was reintroduced from compounds, in which it was regularly preserved;
cf. *kuŋkəlä ‘shin bone’ from *kuŋkə ‘bend’ + *le̮ and *puəjlɜ from *puəj- ‘knee’ + *le̮ (Janhunen
1977). Obviously, if *le̮mpara ‘breast’ contains *le̮ as its first member, it must be a later for-
mation. Another option is that loss or preservation of initial *l- depends on the following
vowel. Aikio (2014c: 86) argues that *l- is regularly preserved before PU *i̮, and adduces three
additional examples fitting this conditioning.
7 Janhunen (1977: 42) reconstructs this word as *je1 = *je̮, *je.
8 The reconstruction of PFU *juxi follows Sammallahti (1988: 537), who sets up PU *joxɨ
instead.
Since these examples include all three “weak” consonants *x, *w, *j as well as
the PSam. reflexes *-u, *-o and *-e̮, there seems no other option than to recon-
struct different vowels for Proto-Uralic, i.e. *u, *o and *i̮, respectively, and to
assume merger of all three into *u in Proto-Finno-Ugric. An alternative hypo-
thetical development of, for instance, *uxi to *i̮xi > *e̮ in PSam. (which could
be a case of assimilation) is contradicted by PSam. *tu- ‘row’, PFU *suxi- < PU
*suxi-.
Aikio’s interpretation of the correspondence between unrounded vowels in
Samoyedic and rounded vowels in Finno-Ugric is that sporadic rounding took
place in several words in Finno-Ugric, and at several stages. This is unlikely.
In view of the correspondences above, the rounding in Finno-Ugric was more
probably a regular development, which occurred at an early stage in Finno-
Ugric, not several times in the separate branches. In support of his idea that
the rounding of the vowel of *ji̮xi ‘drink’ was a late development in this word
in particular, he adduces Hu. iszik, iv- ‘drink’ and Hu. íj ‘bow’ < *ji̮ŋsi (Aikio
2002: 40). However, since both words have initial *j-, it seems best to return
to the earlier explanation that the Hungarian vocalism in these words is due
to a secondary development caused by the initial, and derive them from PFU
*joxi and *joŋsi, respectively. I will not discuss the Permic evidence for survival
of *i̮ into Finno-Ugric that Aikio adduces, because, according to him, it is uncer-
tain.
The word for ‘bow’ is one of another small group of etyma in which Samo-
yedic shows unrounded vowels for rounded vowels in Finno-Ugric:
– PSam. *ji̮ntə ‘bow’, PFU *joŋsi < PU *ji̮ŋsi (Aikio 2002: 39)9
– PSam. *ki̮nsV- ‘star’, PFU *kunśa < PU *ki̮nśa
– PSam. *ki̮j ‘moon, month’, PFU kuxi10 < PU *ki̮xji
The last two words are further evidence against Aikio’s idea of a survival of old
*i̮ in Hu. iv- and íj, since Hungarian shows reflexes of rounded vowels in both
cases: Hu. húgy ‘star’ and hó, hava- ‘moon’.
If indeed Samoyedic preserves the contrast between PU *o, *u and *i̮ so
much better than Finno-Ugric, this further confirms that the primary split in
the Uralic family was between Samoyedic on the one hand and Finno-Ugric
on the other: the merger of these vowels is then a common innovation of the
Finno-Ugric languages. Nevertheless, it must be noted that, apart from the posi-
tion before *xi, *wi, *ji, the exact conditions of these vowel changes are not
clear; that the total number of examples is modest by all standards; and that
9 In a later article, Aikio reconstructs PU *joŋsi (2014b: 11), perhaps for PFU *joŋsi.
10 Aikio (2002: 39) notes that the reconstruction of *x in this word is uncertain. Other options
are *kuwi and *kuŋi.
the Proto-Samoyedic words are so short that not all relevant conditions are
necessarily clear at this point. Another problem is that with the small number
of accepted Uralic etymologies for Samoyedic it is largely unclear which root
structures were absent for structural reasons and which happen to be unat-
tested by chance. For instance, all roots with *x are reconstructed with final *i,
while with other root types final *a is found as well; it is theoretically possible
that contrasts now seen as archaisms of Samoyedic are to be attributed to an
earlier contrast between roots in *-xi and *-xa.11
With the revised reconstruction of the Uralic verb for ‘drink’ by Aikio, the
comparison with PIE *h₁egwh- loses much of its initial appeal. If Aikio’s revision
turns out to be mistaken, or if more evidence for phonological correspondences
between Proto-Indo-European and Proto-Uralic is found, the connection may
eventually be revived—the semantic side, at least, is good. But at this point the
more detailed reconstruction within Uralic overrules the macro-comparison
with Indo-European.
For the subgrouping of Indo-European the verb *h₁egwh- is difficult to use
since the etymon is not confined to Anatolian and Tocharian, as mentioned
above, even though it is best attested there. Lat. ēbrius ‘drunk’ and Gr. νήφω ‘be
sober’ (see also Weiss 1994) simply prove that the root survived into core Indo-
European. The only possible way to use this root for subgrouping is to assume
that the original meaning was ‘drink’, which acquired the meaning ‘get drunk’
after Anatolian and Tocharian had left the speech community.
3 The m-interrogative
Next to the widespread PIE interrogative stem *kwi-, *kwe-, *kwo-, there is an-
other interrogative in *m- (Dunkel 2014: 518–523). Although the existence of
this interrogative stem is recognised in the literature, it is not well known, and
not mentioned in standard introductions such as Beekes (2011: 227–231), Fort-
son (2004: 130) and Meier-Brügger (2003: 227–229). This is certainly due to the
fact that this stem is clearly attested only in Anatolian and Tocharian. In Ana-
tolian the following forms are found:
– Hitt. maši- ‘how many; however many’
– Hitt. mān ‘if, how, when, like’, possibly also the modal particle man
– Hitt. mānḫanda ‘just as’ (Kloekhorst 2010)
– Pal. maš ‘as much as’
11 According to Janhunen (2007: 216–217), the lack of roots in *-xa may be due to a sound
change of *-ki to *-xi: while roots in *-ka are well attested, roots in *-ki are extremely rare.
In addition, Hitt. =ma ‘and, but’ may be related, but the semantics are not
compelling; several steps of development would have to be assumed, e.g. ‘how’
> ‘as’ > ‘as well as’ > ‘and’. All forms are apparently based on a stem *mo-, e.g.
Hitt. maši- < *mo-s-i- and Hitt. mān < *mó-n (Kloekhorst 2008: 552, 564).
The relationship of these Anatolian forms with a number of interrogatives,
relatives and indefinites in Tocharian has been discussed by Hackstein (2004),
Pinault (2010) and myself (Peyrot 2018a). Since the relevant formations are
treated in some detail in Peyrot (2018a), I will here give only a brief presen-
tation of the material.
The basic elements found in the Tocharian interrogatives, relatives and
indefinites are:
– PToch. *kwə- < PIE *kwi-
– PToch. *mə- < PIE *mo-
– PToch. *ən- < PIE *mo-
– The PToch. demonstrative stem nom.sg.m. *se, obl.sg.m. *ce, nom.sg.f. *sa,
etc. < PIE *so, *seh₂, *tod
– The PToch. clitic markers *-nə, *-w, probably from PIE *nu, *u, and the
emphatic particle *kə
About PToch. *mə- < *mo- it should be noted that the reconstruction of the
vowel *o is based on Anatolian. As far as the Tocharian evidence is concerned,
*mu would also be possible, or, with loss of the expected palatalisation of *m,
also *mi or *me. This needs to be stressed, since the *m-interrogative may have
had different stem variants, e.g. *mi-, *me-, *mo-, parallel to *kwi-, *kwe-, *kwo-,
as pointed out to me by Tijmen Pronk. The derivation of *mə- < *mo- requires
a special reduction of the vowel, since the normal reflex of *o is PToch. *e. A
parallel for this reduction is found in the demonstratives, where next to *se <
*so and *te < *tod we also find *sə and *tə.
In my analysis, PToch. *ən- is a further development of *mə-. Since *ən- is
always followed by a demonstrative element, the *n can be explained from
assimilation of original *m to a following *t- (or *c-), which then spread
throughout the paradigm, that is, to forms with a demonstrative in *s-. Al-
though the details remain to be settled, the rise of forms of the type *ən-te < *m-
te < *mə-te < *mo-tod through syncope of the *ə of *mə- is probably due to the
accent or the syllable structure. PToch. *ən- is a unique initial, since no other
word in the language begins with *ə-. In Toch.A, this situation is preserved, with
the unique initial än- < *ən-, while in Toch.B we find it changed to in-.
It is not clear at which stage the demonstrative elements have been added,
but since all three interrogative, relative and indefinite elements are unin-
flected, it is very likely that the inflected demonstrative could compensate for
the loss of the inflexion of the elements *kwə-, *mə- and *ən-. The demonstra-
tives ending in *-e have reduced variants in *-ə: nom.sg.m. *sə for *se < *so and
sg.n. *tə for *te < *tod (see also above).
Most of the complex formations listed below were probably found in Proto-
Tocharian, but not in Proto-Indo-European. The PIE reconstructions are only
meant to illustrate the derivation of the separate elements.
With the formative *mə- we find:
– Toch.B mäksu ‘which’ < *mə-kwə-sə-w < *mo-kwi-so-u
– Toch.B mäkte ‘how’ < *mə-kwə-te ‘what’ < *mo-kwi-tod
– Toch.A mänt ‘how’ and Toch.B mant ‘so’ < *mə-ən-tə ‘how’ < ‘what’ < *mo-
mo-tod
– Toch.B mantsu ‘some’ (?) < *mə-ən-sə-w < *mo-mo-so-u
In addition, Toch.B manta /mə́ nta/ ‘never’ may derive from ‘ever’ and reflect
*mə-ən-ta < *mo-mo-teh₂m.
With the formative *ən- we find:
– Toch.B intsu ‘which’ < *ən-sə-w < *mo-so-u
– Toch.A äntsaṃ ‘which’ < *ən-se-nə < *mo-so-nu
– Toch.A äntā ‘where’ (with a reduced variant tā) < *ən-ta < *mo-teh₂m
And further *ən is found in second position in Toch.A mänt and Toch.B mant,
mantsu and manta (see above).
With the formative *kwə- we find:
– Toch.B kuse ‘who, what’ and Toch.A kus, id. < *kwə-se < *kwi-so
– Toch.B ksa, indefinite pronoun, a reduced form of kuse < *kwə-se < *kwi-so
And further with *kwə in second position we find Toch.B mäksu and mäkte (see
above).
As argued by Hackstein (2004: 280–283), in most of these formations the
first element must have been originally interrogative, and if there was a sec-
ond interrogative-relative element it was relative or “connective”. As I see it, the
demonstrative element provided the necessary inflexion. Thus, the interroga-
tive value of *mə- is preserved in Toch.B mäksu ‘which’, mäkte ‘how’ and Toch.A
mänt ‘how’; for *ən- it is preserved in Toch.B intsu, Toch.A äntsaṃ ‘which’, and
in Toch.A äntā, tā ‘where’; and for *kwə- it is preserved in Toch.B kuse, Toch.A
kus ‘who, what’.
In my view, Tocharian thus provides a solid piece of evidence for the recon-
struction of the interrogative pronoun stem *mo-. Interestingly, the co-
existence of PIE *kwi-, *kwe-, *kwo- and *mo- is parallel to the two interroga-
tive pronouns found in Uralic; cf. Fi. kuka ‘who’, Hu. ki ‘who’ and Fi. mikä ‘what,
which’, Hu. mi ‘what’.12 The comparison of the Proto-Indo-European and Proto-
12 Yet the vocalism is difficult to reconstruct. For instance, Rédei reconstructs PU *mɜ ‘what’
with “ɜ” as an “unbestimmbarer Vokal” (1988: 296).
Uralic m-interrogatives has been made a.o. by Pedersen (1938: 71–72), Collinder
(1965: 113, 127, 149) and Greenberg (2000: 229–231, № 62). Although the equation
is missing from Kortlandt’s list of 27 Proto-Indo-Uralic grammatical elements
(2002), it seems solid enough to be added.
Nevertheless, the neat contrast observed between Hu. ki ‘who’ and mi ‘what’
is not found in Indo-European. In view of Hitt. maši- ‘how many’ and Pal.
maš ‘as much as’, Hackstein (2004: 281–282) suggests that *mo- was a quanti-
fying interrogative, ‘how many’. However, I find this unattractive for the more
basic meanings attested in Tocharian. If any more precise meaning should be
reconstructed for *mo-, it would rather be something like ‘which’, a more spe-
cific, restrictive interrogative; cf. Hackstein (2004: 281–282), “TB mäksu is best
described as an adjectival interrogative for restricting reference, “which one of
a given class or group.””
For the internal subgrouping of Indo-European, the Tocharian-Anatolian
match in the m-interrogative is of limited value at most. First of all, in the
words of Pedersen, “Es genügt hervorzuheben, dass Verlust des Alten (in diesem
Falle des m-Pronomens), der allmählich in jedem Sprachzweige für sich einge-
treten sein kann, nicht als eine gemeinsame Neuerung gewertet werden darf.”
(1938: 72). This statement can be relativised a little, since it is obviously more
economical to assume loss of an archaic feature once, as a common innova-
tion of the languages that have not preserved it, than several times, for each
branch independently. This is all the more true in the case of a grammatical ele-
ment such as the m-interrogative. At the same time, it is clear that a common
innovation that involves a loss is a much weaker argument for subgrouping
than a common innovation consisting of an indisputably traceable “positive”
change.
In the case of the m-interrogative, the evaluation of its usefulness for sub-
grouping is further complicated by the existence of further possible traces in
Celtic, where we find OIr. má ‘if’ and related forms.13 Although these appar-
ently have no interrogative value, original interrogative value may be suggested
by the Breton and Cornish local conjunction ma (Pedersen 1913: 230). If these
conjunctions derive from the stem *mo-, the only possible common innovation
that remains for the Indo-European languages including Celtic and exclud-
ing Anatolian and Tocharian is a development of the original interrogative
to a relative or a local relative. Since shifts of this kind are commonplace, as
13 Dunkel lists many more cognates of the *mo-stem, including reflexes as indefinite in Vedic
and Armenian, but these are in my view not convincing (Dunkel 2014: 518–523). The pos-
sible sources of indefinites are manifold (Haspelmath 1997) and not each and every -m-
can be derived from the same interrogative *mo-.
shown by the Tocharian interrogatives and relatives listed above, but also by
similar developments in other branches of Indo-European and in non-Indo-
European languages, this would make the evidence of the m-interrogative for
Indo-European internal subgrouping practically useless.
4 Conclusions
References
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2002. New and Old Samoyed etymologies.
Finnisch-Ugrische Forschungen 57, 9–57.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2012. On Finnic long vowels, Samoyed vowel
sequences, and Proto-Uralic *x. In: T. Hyytiäinen (ed.), Per Urales ad Orientem. Iter
polyphonicum multilingue. Festskrift tillägnad Juha Janhunen på hans sextioårsdag
den 12 februari 2012. Helsinki: Finno-Ugrian Society, 227–250.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2014a. The Uralic-Yukaghir lexical corre-
spondences: genetic inheritance, language contact or chance resemblance? Fin-
nisch-Ugrische Forschungen 62, 7–76.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2014b. Studies in Uralic etymology II: Finnic
etymologies. Linguistica Uralica 50, 1–19.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2014c. Studies in Uralic etymology III: Mari
etymologies. Linguistica Uralica 50, 81–93.
Beekes, R.S.P. 2011. Comparative Indo-European Linguistics. An Introduction. Second edi-
tion. Revised and corrected by Michiel de Vaan. Amsterdam–Philadelphia: John
Benjamins.
Carling, G. 2005. Proto-Tocharian, Common Tocharian, and Tocharian. On the value
of linguistic connections in a reconstructed language. In: K. Jones-Bley; M.E. Huld;
A. Della Volpe; M. Robbins Dexter (eds.), Proceedings of the Fifteenth Annual UCLA
Indo-European Conference. Washington: Institute for the Study of Man, 47–70.
Collinder, B. 1965. Hat das Uralische Verwandte? Eine sprachvergleichende Unter-
suchung. Acta Universitatis Upsaliensis 1/4: 108–180.
Dunkel, G.E. 2014. Lexikon der indogermanischen Partikeln und Pronominalstämme.
Band 2: Lexikon. Heidelberg: Winter.
Fortson, B.W., IV 2004. Indo-European Language and Culture. An Introduction. Oxford:
Wiley-Blackwell.
Greenberg, J.H. 2000. Indo-European and its Closest Relatives: the Eurasiatic Language
Family. Vol. 1: Grammar. Stanford: Stanford University Press.
Hackstein, O. 2004. From discourse to syntax: The case of compound interrogatives in
Indo-European and beyond. In: K. Jones-Bley; M.E. Huld; A. Della Volpe; M. Robbins
Dexter (eds.), Proceedings of the Fifteenth Annual UCLA Indo-European Conference.
Washington: Institute for the Study of Man, 257–298.
Haspelmath, M. 1997. Indefinite Pronouns. Oxford: Clarendon.
Janhunen, J. 1977. Samojedischer Wortschatz. Helsinki: Finno-Ugrian Society.
Janhunen, J. 1981. Uralilaisen kantakielen sanastosta. Journal de la Société Finno-
Ougrienne 77, 219–274.
Janhunen, J. 2007. The primary laryngeal in Uralic and beyond. In: J. Ylikoski; A. Aikio
(eds.) Sámit, sánit, sátnehámit. Riepmočála Pekka Sammallahtii miessemánu 21.
beaivve 2007. Helsinki: Finno-Ugrian Society, 203–227.
Jasanoff, J.H. 2003. Hittite and the Indo-European Verb. Oxford–New York: Oxford Uni-
versity Press.
Kim, R.I. 2007. The Tocharian subjunctive in light of the h₂e-conjugation model. In:
A.J. Nussbaum (ed.), Verba Docenti. Studies in Historical and Indo-European Linguis-
tics Presented to Jay H. Jasanoff by Students, Colleagues, and Friends. Ann Arbor:
Beech Stave, 185–200.
Kloekhorst, A. 2008. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden–
Boston: Brill.
Kloekhorst, A. 2010. Hittite mān, maḫḫan, māḫḫan, māḫḫanda and mānḫanda. In:
R. Kim; N. Oettinger; E. Rieken; M. Weiss (eds.), Ex Anatolia Lux: Anatolian and
Indo-European Studies in Honor of H. Craig Melchert on the Occasion of his Sixty-Fifth
Birthday. Ann Arbor–New York: Beech Stave Press, 217–226.
Kloekhorst, A. 2011. The accentuation of the PIE word for ‘daughter’. In: T. Pronk;
R. Derksen (eds.), Accent Matters. Papers on Balto-Slavic Accentology. Amsterdam–
New York: Rodopi, 235–243.
Kortlandt, F.H.H. 1989. Eight Indo-Uralic verbs? Münchener Studien zur Sprachwis-
senschaft 50, 79–85.
Kortlandt, F.H.H. 2002. The Indo-Uralic verb. In: R. Blokland; C. Hasselblatt (eds.),
Finno-Ugrians and Indo-Europeans: Linguistic and Literary Contacts. Maastricht:
Shaker, 217–227.
Kortlandt, F.H.H. 2016. Baltic, Slavic, Germanic. Baltistica 51, 81–86.
Malzahn, M. 2016. The second one to branch off? The Tocharian lexicon revisited. In:
B. Simmelkjær Sandgaard Hansen; B. Nielsen Whitehead; T. Olander; B.A. Olsen
(eds.), Etymology and the European Lexicon. Proceedings of the 14th Fachtagung
der Indogermanischen Gesellschaft, 17–22 September 2012, Copenhagen. Wiesbaden:
Reichert, 281–292.
Meier-Brügger, M. 2003. Indo-European Linguistics. In Cooperation with Matthias Fritz
and Manfred Mayrhofer. Berlin–New York: Walter de Gruyter.
Pedersen, H. 1913. Vergleichende Grammatik der keltischen Sprachen. Zweiter Band. Göt-
tingen: Vandenhoeck & Ruprecht.
Pedersen, H. 1938. Hittitisch und die anderen indoeuropäischen Sprachen. København:
Munksgaard.
Peyrot, M. 2018a. Interrogative stems in Hittite and Tocharian. Indogermanische For-
schungen 123, 65–90.
Peyrot, M. 2018b. Tocharian agricultural terminology: Between inheritance and lan-
guage contact. In: G. Kroonen; J.P. Mallory; B. Comrie (eds.), Talking Neolithic: Pro-
ceedings of the workshop on Indo-European origins held at the Max Planck Institute
for Evolutionary Anthropology, Leipzig, December 2–3, 2013.
Pinault, G.-J. 2006. Retour sur le numéral “un” en tokharien. Indogermanische Forschun-
gen 111, 71–97.
Pinault, G.-J. 2010. Le pronom d’ipséité en tokharien. In: I. Choi-Jonin; M. Deval; O. Sou-
tet (eds.), Typologie et comparatisme. Hommages offerts à Alain Lemaréchal. Leuven–
Paris: Peeters, 351–365.
Rédei, K. 1988–1991. Uralisches etymologisches Wörterbuch. Budapest: Akadémiai Kiadó.
Ringe, D.; T. Warnow; A. Taylor 2002. Indo-European and computational cladistics.
Transactions of the Philological Society 100, 59–129
Sammallahti, P. 1988. Historical phonology of the Uralic languages with special refer-
ence to Samoyed, Ugric, and Permic. In: D. Sinor (ed.), The Uralic Languages. Descrip-
tion, History and Foreign Influences. Leiden–New York: Brill, 478–554.
Schmidt, K.T. 1992. Archaismen des Tocharischen und ihre Bedeutung für Fragen der
Rekonstruktion und der Ausgliederung. In: R.S.P. Beekes; A. Lubotsky; J.J.S. Weiten-
berg (eds.), Rekonstruktion und relative Chronologie. Akten der VIII. Fachtagung der
Indogermanischen Gesellschaft, Leiden, 31. August–4. September 1987. Innsbruck: IBS,
101–114.
Weiss, M. 1994. On the non-verbal origin of the Greek verb νήφειν ‘to be sober’. His-
torische Sprachforschung 107, 91–98.
Winter, W. 1997. Lexical archaisms in the Tocharian languages. In: H.H. Hock (ed.), His-
torical, Indo-European, and Lexicographical Studies. A Festschrift for Ladislav Zgusta
on the Occasion of his 70th Birthday. Berlin: Mouton de Gruyter, 183–193.
1. The Proto-Indo-European root noun *sem-, *sm- (Beekes & de Vaan 2011: 210)
is reflected in the cardinal ‘one’ in a number of ancient Indo-European lan-
guages: Greek εἷς, accusative ἕνα < *sem-, Armenian mi < *miyo- << f. *smiʕ-
(Martirosyan 2010: 468), Tocharian A sas, Tocharian B ṣe < Proto-Tocharian
*ṣæns << *ṣäns (Pinault 2006: 91).1 The locative singular *sēmi ‘in one (of two)’
is reflected in Latin sēmi-, Old High German sāmi-, Greek ἡμι- ‘half’, Sanskrit
ásāmi ‘completely’. From a collective *som-éʕ-, the thematic adjective *somʕo-
‘the same’ can be explained, as in Skt. samá-, Avestan hama-, Gr. ὁμός, Proto-
Germanic *sama- ‘the same’, the Old Irish emphatic particle -som, and the
Tocharian oblique forms of ‘one’, ToA som-, ToB m. ṣem-, f. ṣom- (Pinault 2006:
89). Other derivatives include the preverb *som ‘together’ (Skt. sám, Old Church
Slavic sŭ, Lithuanian sam-), and, possibly, the particle *sme(ʕ) (Skt. sma, which
according to Mumm 2004 reinforces an assertion or a request; Gr. μήν ‘truly’).
In compounds, the zero grade appears as the first member in words such as Skt.
sakr̥ t́ , Av. hakǝrǝt̰ ‘once’ from Indo-Iranian *sm̥ -kr̥ t, in the numeral ‘thousand’,
Skt. sahásra-, Av. hazaŋra- < PIE *sm-ǵhes-lo-, Lat. mīlle < PIE *sm-ǵhes-lio-, and
in many other words.
A further likely occurrence is the element *sm in the oblique cases of the
demonstrative pronouns PIE *ʔe- and *to-: ablative *ʔesmōd, locative *tosmi,
dative *tosmōi, etc. There is no other candidate for the etymology of the ele-
ment *sm, and it is not difficult to imagine that the same morpheme can be
used to build deictic pronouns while also developing the identificational mean-
ing ‘one’. The restriction of *sm to the oblique cases—we can reconstruct dative
*-smōi, abl. *-smōd, loc. *-smi—will be addressed in the final section of this
paper.
1 For esthetic and comparativistic reasons (viz. to facilitate the comparison with other lan-
guage families), I write ʔ (glottal stop), ʕ (pharyngeal fricative) and ʕw (labialized pharyngeal
fricative) for the canonical reconstructions *h₁, *h₂, *h₃. Whereas I regard the phonetic iden-
tity of *h₁ with [ʔ] as quite certain, I concede that *h₂ and *h₃ may have developed different
realizations, such as velar fricatives (Kümmel 2007: 327–336). Furthermore, I define “Early
Proto-Indo-European” as Proto-Indo-Anatolian, that is, the stage before the Anatolian lan-
guages split off, and “Late PIE” as the subsequent common stage of the remaining branches
of Indo-European.
Singular Singular
Plural Plural
Table 14.1 provides a summary of the main deictic forms that can be recon-
structed for Late PIE in the nominative, accusative and in the oblique case
forms. The survey is based on Beekes & de Vaan 2011: 228–229. The only dif-
ference with their interpretation is that I reconstruct the feminine oblique
with o-vocalism, thus *to(si)-, with Beekes 1988b: 3, rather than *tesi-. The latter
reconstruction was based in particular on Gothic gen.sg. ϸizos, which can be
analogical to gen.sg.m. ϸis.
2. Kloekhorst (2008: 865–867) reconstructs the Hittite word for ‘one’ as /si-/
or /sī-/: Old Hittite nom. 1-iš /sis/, acc. ši-an, gen. ši-el. He explicitly rejects
the interpretation as a thematic stem /sia-/, based on the Old Hittite prepon-
derance of the spelling 1-iš. According to Kloekhorst, “nom.sg.c. /sis/ was in
younger times secondarily changed to /sias/, spelled 1-aš.” Hoffner & Melchert
2008: 154 regard /sis/ as an i-stem form with unclear relationship to /sia-/. Pos-
sibly, Hittite šī-̆ is cognate with Greek ἴα ‘one and the same, the one’ (Homer,
dialects). The latter has been explained as an inflected form of the anaphoric
pronoun *ʔi- (Beekes 2010: 571) or as the result of reduction of a consonant clus-
ter *smj- to *sj- in a preform *smiéʕ- to the feminine PIE *smiʕ- ‘one’ (Schmidt
1898: 399, Gippert 2004, Hackstein 2005). But if a PIE stem *si- ‘one’ existed,
we could assume that it is continued as such in Hittite and, as a relic form, in
Greek too.
Other possible cognates of Hittite sī-̆ are Tocharian A ṣī ‘for the first time’,
ṣyak ‘together’, both of which may continue Proto-Tocharian *ṣyæ (Pinault
2006: 84). Pinault explicitly compares Hittite thematic šia- and reconstructs
a pronoun *sio, cognate with Skt. demonstrative syá/tyá- ‘that one concerned’.
But the latter is probably an Indo-Aryan innovation on the basis of sá/tá- (Gotō
2013: 69). Hackstein (2005: 178) traces Tocharian A ṣī back to a similar reduc-
tion of PIE *smiéʕ- to *siéʕ- as proposed by Schmidt for Greek ἴα. In the next
section I will show why this solution must be refuted.
3. The identification of the element *sm in *ʔe-sm- and *to-sm- with *sem- ‘one’
has naturally led scholars to search for a similar origin of *si in the feminine
oblique stems *ʔe-si- and *to-si-. Many have tried to derive -si- from the femi-
nine *sm-iʕ- ‘one’, starting with Schmidt 1898. For a prestage of Greek, he posits
nom. *smía, gen. *smjā ́s, dat. *smjā ́i, acc. *smíam. In the genitive and dative,
the interconsonantal m would have disappeared “as in Skt. tásyai”. The resulting
forms *sjā ́s, *sjā ́i, which Schmidt dates to the PIE stage, would have developed
within Greek to *isjā ́s, dat. *isjā ́i by a process with Schmidt does not explicitly
define, but which he probably regards as vowel epenthesis before the cluster
sj-. Schmidt’s main argument for his explanation is the near-complementary
distribution in Homer between the m-forms in the nominative and accusative
versus the i-forms in the other two cases: nom. μία 12x, ἴα 3x; acc. μίαν 7x, ἴαν 1x;
gen. μιῆς 1x, ἰῆς 2x; dat. μιῆι zero, ἰῆι 4x. Schmidt’s theory has been modernized
but maintained by Gippert 2004 in all its details except for the initial i-.
As suggestive as the distribution between μία, μίαν and ἰῆς, ἰῆι may be, a num-
ber of counterarguments can be mustered against it. Firstly, the scenario does
not explain the pervasive psilosis of ἴα in Greek. Schmidt himself solves this by
assuming Proto-Greek epenthetic *i-, which would not yield *hi-. But this road
is closed to us, as Schmidt’s comparandum, viz. the imperative ἴσθι ‘be’, allegedly
from *zdhi, is now explained by vocalization of the initial laryngeal in *ʔs-dhi.
We could assume that ἴα was an Aeolism, but that would mean that the whole
feminine of ‘one’ (on the complementary distribution of which the argument
largely builds) must have been an Aeolism. Secondly, the assumed simplifica-
tion of *smj to *sj is an ad hoc rule, devised only to explain the forms in which
it is allegedly found. The reference to Skt. áśman- ‘stone’, gen.sg. áśnaḥ, with
loss of the middle consonant in the cluster *-ḱmn-, is of no support, since the
loss of m was probably restricted to the context *-Cmn- (Schmidt 1895: 87–159,
Mayrhofer 1986: 159, Nussbaum 2010). In general, PIE does not seem to have
problems with consonant clusters containing only one obstruent. Thirdly, the
morphemes *sem- and *sm- ‘one’ were quite productive in Indo-European, and
one would expect *sm- to have been restored immediately if there ever was
such an early tendency to reduce *smiéʕ- to *siéʕ-. Fourthly, Schmidt’s theory
separates the oblique feminines from the pronominal genitive form *tosio (as
seen by Gippert 2004: 163 himself).
Other explanations for the pronominal feminine obliques *ʔesi-, *tosi- also
have their disadvantages. Lane (1961: 473) derives them from a deictic pro-
noun *sio-, which does not explain its restriction to the feminine or to the
oblique cases, not to mention the question of whether a PIE *sio- existed at
all (see above). Beekes (1988a: 79–81) explains Gr. ἴα as a direct reflex of PIE *siʕ
‘she’. Its oblique cases, gen. *siéʕs, dat. *siéʕi, would be the direct input for the
oblique feminines of PIE such as *ʔesiéʕi. This would still leave the question
unanswered, why the si-forms were only used the oblique cases. Moreover, a
feminine nominative *siʕ is unlikely to have existed in PIE. Gothic si and OIr.
sí ‘she’ can be explained as recent remakes of nom.sg.f. *iʕ plus *s- (Schrijver
1997: 56). Brugmann (1904: 404) assumes that “*tesi̯o” was originally used for
all genders, and then served as a basis to create the feminine “*tesi̯ās” by the
addition of the productive feminine endings. From the genitive = ablative, the
pattern would have spread to the dative and locative. This explanation is not
so far removed from what I will propose.
4. My own proposal is based on the observation that *sm and *si are both
attested in the meaning ‘one’, on the assumption that this may have devel-
oped from a different, earlier meaning, and on the well-known complementary
distribution of *sm and *si in the pronouns. The starting hypothesis is there-
fore that *sm is present in the masculine oblique forms and *si in the femi-
nine:
This distribution may imply that the stems *sm and *si were not equivalent
at an earlier stage of Proto-Indo-European; that the masculine/neuter and the
feminine forms of the demonstrative pronouns *ʔe and *to did not arise at the
same stage of Proto-Indo-European; or both. The main questions to be dis-
cussed in the next sections will therefore be: How did the extended pronominal
forms arise? Why was *sm used in the masculine/neuter and *si in the femi-
nine? Why were these extended forms restricted to the oblique case forms? To
what extent did they differ from unextended forms? What is the relationship to
the genitive singular *tosio? Why were there two different stems meaning ‘one’
in PIE?
6. It is entirely possible that PIE *si was a separate word with an unknown his-
tory, unrelated to other words of the reconstructed vocabulary. But it is equally
legitimate to hypothesize that *si is cognate with PIE deictic *so (Dunkel 2014
II: 741). If *so and *si were related, their vowel alternation is reminiscent of that
between the deictic stems *ḱo and *ḱi ‘this (here)’ (Hitt. m. kās < *ḱos, n. kī < *ḱí;
further in PIE *ḱi ‘here’), *bho ‘that’ and *bhi (Hitt. apā- ‘yon’, PIE inst. *-bhi), or
*kwo ‘who’ and *kwi(d) ‘what’. Unlike these sets of pronouns or particles, whose
Singular Plural
mutual relationship is still easy to see, *so and *si became dissociated. Probably,
this already happened in Proto-Indo-European. The paradigmatic connection
may have been lost so early due to the fact that *so became the nominative
singular of the demonstrative paradigm of *to-, whence canonical PIE *so/*to-
‘that’.
Kloekhorst 2012: 262–263 discusses the Anatolian evidence for the deictic
pronouns. He finds a regular distinction between alternating i- and o-vocalism
in the nominative and accusative versus e-vocalism in the oblique case forms,
as shown by table 14.2.
Kloekhorst stresses that the vowel distribution is so remarkable that it must
be original: Proto-Anatolian must have inherited it from PIE. An important
difference with reconstructions based on non-Anatolian material is that the
latter lead us to postulate an e-grade in the nom.sg. *ʔe and *kwe (Beekes & de
Vaan 2011: 227, 231), based on IIr. *ayam and Germanic *e for *ʔe, and based on
Old Irish cía ‘who’ < *kwei for PIE *kwe. Since it is unlikely that an asigmatic
nom.sg. in *-e would have arisen in Late PIE on the basis of the oblique case
forms, Anatolian has probably secondarily restricted the e-grade to the oblique
cases.
7. In order to provide the background for the subsequent analysis of the pro-
nouns, I must briefly explain some of my assumptions of the prestages of Proto-
Indo-European. I subscribe to the view that Proto-Indo-European and Proto-
Uralic share such a large number of core morphological forms that we have
enough reason to posit a common Proto-Indo-Uralic stage (Kortlandt 2010:
387–403). Because of the greater time depth, our reconstruction of the PIU
linguistic system will inevitably remain much less detailed than that of PIE.
Nonetheless, the Pre-PIE period cannot be viewed as a ‘black hole’ anymore.
In its entirety, it can be defined as the period between Indo-Uralic and Indo-
Anatolian. This period can be further subdivided into different stages with a
specific relative chronology of events.
Two main differences between Proto-Uralic and Early Proto-Indo-European
concern the phonological system and the verbal syntax. It has therefore been
assumed that PIE arose from an earlier Indo-Uralic stage which was trans-
formed by a Caucasian substrate (Kortlandt 2010: 391). I provisionally call the
initial outcome of this transformation, by which the vowel system was reduced
and the verbal syntax reorganized, “Early Pre-PIE”. The subsequent phase of
development, which saw, among other features, the introduction of full vow-
els in unstressed syllables and the rise of new inflexional classes, can be called
“Late Pre-PIE”. The later stages “Early PIE” and “Late PIE” have been explained
in footnote 1.
My conception of Pre-PIE and Early PIE morphology is built on Kortlandt
2010 and Beekes & de Vaan 2011. For an accessible introduction to the theory of
PIE nominal morphology, see Kloekhorst 2013. The earliest stage of ablaut alter-
nations that we can reconstruct for Pre-Proto-Indo-European distinguished
between accented e-grade and unaccented zero grade of any syllable. The PIE
o-grade arose later: it represents the introduction of a full vowel (which became
*o) in an unstressed syllable (Schindler 1975, Beekes 1985: 157, Kortlandt 2010:
396). A famous example is the suffix in the nom.acc.sg. of neuter s-stems, *mén-
s >> *mén-os.
Pre-PIE had an ergative case system, in which the agent of a transitive clause
was marked differently from the agent of an intransitive clause, the latter shar-
ing the marking of a patient (Vaillant 1936). This assumption explains, among
other things, why the PIE nominative equals the accusative in the neuter gen-
der: these were inanimate nouns, which did not normally occur as an agent.
This theory entails that the PIE nominative continues the Pre-PIE endingless
absolutive (the patient marker). The Pre-PIE genitive ending *-s was the erga-
tive marker for animates, which became the animate nominative marker *-s in
canonical PIE. The Pre-PIE instrumental ending *-t, which is preserved in Ana-
tolian *-ti and became “ablative” *-d but also, by phonetic reduction, “instru-
mental” *-h₁ elsewhere, served as the ergative of inanimate nouns (Kortlandt
2010: 40). The accusative marker *-m probably goes back to an earlier allative
affix of Indo-Uralic. The locative ending *-i may also go back to Indo-Uralic, but
the genitive, dative and oblique plural cases were only created within the new
PIE nominative-accusative system (Kortlandt 2010: 41).
Before the rise of the o-stems, nouns could be root nouns or consonant
stems. The canonical PIE o-stems arose from the reinterpretation of the gen-
itive ending *-os (that is, the variant of the ending *-s which arose in mobile
8. We can now continue our analysis of the pronominal system. The only cat-
egory in which PIE e- and o-vocalism alternates with i-vocalism are the pro-
nouns, of which we have seen the Anatolian evidence above. Kortlandt (2010:
41) thinks that the difference between e- and i-vocalism reflects an animacy
difference in Early PIE. He reconstructs the basic forms as shown in table 14.3.
If inserted into this table, Early PIE *si can be interpreted as an inanimate
absolutive form. It may be cognate with the Late PIE reflexive pronoun *se,
attested in the accusative *se (Gr. ἕ, Latin sē, Gothic sik), gen. *sei besides *seue,
dat. *sebhio besides *soi (Beekes & de Vaan 2011: 234). Kortlandt (2010: 369–371)
traces it back to the same demonstrative stem *s- from which *so derives. In
*seue, the element *ue would have been added to *se to disambiguate “a per-
son who is contrasted with another third person” (p. 371). To the pronouns of
table 14.3, I therefore add an s-pronoun which I provisionally call identifica-
tional:
9. The animate accusative *sim can be reconstructed on the basis of its
reflexes in Indo-Iranian and Celtic (Beekes 1983). Rigvedic anaphoric sīm
‘him/her/it/them’, indifferent to gender, has replaced earlier *sim with a short
vowel. In its usage, it has almost become a particle (Jamison 2002, Kupfer 2002:
252–260). The long vowel may have been introduced on the model of nom.sg. sī
‘she’ for *ī. In that case, we can posit feminine nom. *iʕ, acc. *im which replaced
earlier common gender nom. *ʔe, acc. *im. Sanskrit simá- ‘he himself’ (mascu-
line only) may be due to thematization of *sim (Beekes 1983: 203).
PIE anaphoric *sim is also reflected in Proto-Celtic deictic *sim, which pro-
vided neuter forms of the paradigm of *so. Among its reflexes, we find Old Irish
sin ‘the aforementioned’ < *sim, OIr. inso ‘this’ < *sim *so ‘the one / such one’,
OIr. in(d) ‘the’, Gaulish (s)indo- < *sindo- << *sim-de, Gaulish sosin n. ‘this’ <
*so-sim (Schrijver 1997: 39–48).
Beekes (1983: 225) suggests that *sim may have arisen through the addition
of *-s of a preceding word to the pronominal acc.sg. *im, in the same way that
Avestan dim ‘him, it’ can be explained from the meta-analysis of a final *-d of a
preceding word, or Gr. μίν, νίν ‘him’ from the addition of the final nasal of a pre-
ceding word in -m > -n. But these are branch-specific developments, whereas
*sim must already have existed in Late PIE.
10. As presented in tables 14.3 and 14.4, the vocalism of the so/to-pronoun devi-
ates from the e/i-vocalism of the other stems. The o in so/to- could be compared
with pronominal stems such as *kwo- ‘who, what’, *ḱo- ‘this’, *io- ‘which’, *ʔo-
‘this’, *bho- ‘that’, which have long been regarded as adjectival. Their o would
stem from thematization of the stem consonant or from vowel replacement in
adverbial forms such as *ḱi ‘here’, *bhi ‘at’ (Kortlandt 2010: 41). While themati-
zation is conceivable for *kwo-, *ḱo-, *ʔo-, etc., it is less obvious that nom.sg. *so
replaces earlier *se or *si, since *so belongs to a paradigm which otherwise has
only t-forms. PIE *so may instead have arisen at an earlier stage, due the auto-
matic vocalization as *o of an unstressed vowel in Late Pre-PIE. The rise of *so
as an enclitic variant of *se would then be comparable to the rise of o-grade in
nominal compounds (e.g., Gr. ἀ-πάτωρ ‘fatherless’ < PIE *-pʕtor-, Skt. bhádra-
jāni- ‘having a beautiful wife’ < *-gwon-ʕ-, maybe Latin extorris ‘exiled’ if from
*-tors-; Wackernagel 1905: 100–101), for which it may be assumed that it reflects
Animate Inanimate
table 14.6 Early Pre-PIE pronouns, symmetrical vocalism for *t- and *s-
Kortlandt (2010: 397) has proposed an Early Pre-PIE sound law *ti > *si in
order to explain, among other elements, the PIE 2sg. verbal ending *-si versus
the 2sg. pronoun *t- (Indo-Uralic *ti ‘you’) and the nominal plural ending *-es
beside *-i (Indo-Uralic *-t+i). If Early Pre-PIE had the forms *tim and *ti for the
animate accusative and inanimate absolutive, they would have become *sim
and *si by this sound law and merged with existing *sim and *si. That would
have left *to an orphaned pronoun which could merge semantically with *so,
creating the familiar so/to-pronoun of PIE which included the forms *tom and
*tod. In the process, *to seems to have acquired inanimate reference. In this
way, the less symmetrical system of table 14.4 would have come into being.
Thus, in chronological order, I reconstruct the systems given in tables 14.6, 14.5
and 14.4.
11. Now I turn to the use of *sm and *si in the oblique forms of the demonstra-
tives. Kortlandt (2010: 41) holds that, in Early PIE, “genitival and adjectival rela-
tionships were apparently expressed by simple juxtaposition and partial agree-
ment.” A clear example is the genitive plural in PIE *-om, explained by Kort-
landt 1978 and 2010: 40 from an originally predicative adjective. The recognition
of mere juxtaposition as a means to express appurtenance provides a syntactic
justification for the supposition—shared by many scholars—that the mascu-
line and neuter genitive singular *tosio arose as a compound. For instance,
Beekes 1992 concluded that *sio must have been the original form, which was
then reshaped into *tosio. The recognition of *si as a separate pronominal ele-
ment now allows us to refine this analysis. Kortlandt (2010: 42) concludes that
the non-Anatolian languages “created a pronominal gen.sg. form by composi-
tion: *kwe-so, *ʔe-so, *to-si with addition of *-o from *-so.” Such concatenations
of pronouns can be regarded as juxtapositions expressing a relationship of
appurtenance or another type of pragmatic modification, e.g. *kwe ‘who?’, *kwe-
so ‘who-this’ = ‘whose’. If -o in *tosio was indeed taken from *ʔeso, *kweso, this
would point to a more recent age of *tosio.
e-grade o-grade
12. Chances are that *ʔe-sm- was a compound deictic form of the same type
as *ʔe-so, *to-si. It contains the deictic element *sm which has developed the
References
Beekes, R.S.P. 1983. On laryngeals and pronouns. Zeitschrift für Vergleichende Sprach-
forschung 96, 200–232.
Beekes, R.S.P. 1985. On the Origins of the Indo-European Nominal Inflection. Innsbruck:
IBS.
2 Another argument for the locative as the starting point is the absence of sm-forms in the
instrumental singular in Old Indo-Iranian. To Dunkel (2014 II: 196), these facts suggest exactly
the opposite, viz. that the instrumental ending *-sm-eh₁ had already been replaced by *-na in
Indo-Iranian. Yet Avestan inst.sg. tā and aēta are the only oblique singular forms which do
not have hm-endings, nor have they introduced *-na as in the inst.sg. ana to a-/i-. Since there
is no reason why an existing inst.sg. *tasmaH would have been replaced by *taH, when all
other forms of the paradigm had *-sm-, it seems more likely that tā and aēta preserve an
archaism.
Beekes, R.S.P. 1988a. The origin of the Indo-European pronominal inflection. In: M.A. Ja-
zayery; W. Winter (eds.), Languages and Cultures. Studies in Honor of Edgar C. Polo-
mé. Berlin: Mouton de Gruyter, 73–87.
Beekes, R.S.P. 1988b. The pronominal genitive singular in Germanic and PIE. Beiträge
zur Geschichte der deutschen Sprache und Literatur 110, 1–5.
Beekes, R.S.P. 1992. The genitive in *-osio. Folia Linguistica Historica 11. 21–25.
Beekes, R.S.P. 2010. Etymological Dictionary of Greek. With the assistance of Lucien van
Beek. Leiden–Boston: Brill.
Beekes, R.S.P. 2011. Comparative Indo-European Linguistics. An Introduction. Second edi-
tion. Revised and corrected by Michiel de Vaan. Amsterdam–Philadelphia: John
Benjamins.
Brugmann, K. 1904. Kurze vergleichende Grammatik der indogermanischen Sprachen.
Strassburg: Trübner.
Dunkel, G.E. 2014. Lexikon der indogermanischen Partikeln und Pronominalstämme.
Band 1: Einleitung, Terminologie, Lautgesetze, Adverbialbildungen, Nominalsuffixe,
Anhänge und Indices. Band 2: Lexikon. Heidelberg: Winter.
Gippert, J. 2004. Ein Problem der indogermanischen Pronominalflexion. In: A. Hylle-
sted; A. Jørgensen; J. Larsson; T. Olander (eds.), Per Aspera ad Asteriscos. Studia
Indogermanica in honorem Jens Elmegård Rasmussen sexagenarii Idibus Martiis anno
MMIV. Innsbruck: IBS, 155–165.
Gotō, T. 2013. Old Indo-Aryan Morphology and its Indo-Iranian Background. In co-
operation with J. Klein and V. Sadovski. Vienna: Österreichische Akademie der Wis-
senschaften.
Hackstein, O. 2005. Archaismus oder historischer Sprachkontakt: Zur Frage westindo-
germanisch-tocharischer Konvergenzen. In: G. Meiser; O. Hackstein (eds.), Sprach-
kontakt und Sprachwandel. Akten der XI. Fachtagung der Indogermanischen Gesell-
schaft (13.–23. September 2000, Halle an der Saale). Wiesbaden: Reichert, 169–184.
Hoffner, H.A., Jr.; H.C. Melchert 2008. A Grammar of the Hittite Language. Part 1: Refer-
ence Grammar. Winona Lake: Eisenbrauns.
Jamison, S. 2002. Rigvedic sīm and īm. In: M.M. Deshpande, P.E. Hook (eds.), Indian
Linguistic Studies. Festschrift in Honor of George Cardona. Delhi: Motilal Banarsidass
Publishers, 290–312.
Kloekhorst, A. 2008. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden–
Boston: Brill.
Kloekhorst, A. 2012. Pronominal morphology in the Anatolian language family. Altori-
entalische Forschungen 39, 254–264.
Kloekhorst, A. 2013. Indo-European nominal ablaut patterns: The Anatolian evidence.
In: G. Keydana, P. Widmer, T. Olander (eds.), Indo-European Accent and Ablaut.
Copenhagen: Museum Tusculanum, 107–128.
Kloekhorst, A. 2016. The Anatolian stop system and the Indo-Hittite hypothesis. Indo-
germanische Forschungen 121/1, 213–248.
Kortlandt, F. 1978. On the history of the genitive plural in Slavic, Baltic, Germanic and
Indo-European. Lingua 45, 281–300.
Kortlandt, F. 2010. Studies in Germanic, Indo-European and Indo-Uralic. Amsterdam–
New York: Rodopi.
Kümmel, M.J. 2007. Konsonantenwandel. Wiesbaden: Reichert.
Kupfer, K. 2002. Die Demonstrativpronomina im Rigveda. Frankfurt am Main: Peter
Lang.
Lane, G.S. 1961. On the formation of the Indo-European demonstrative. Language 37,
469–475.
Martirosyan, H. 2010. Etymological Dictionary of the Armenian Inherited Lexicon. Lei-
den–Boston: Brill.
Matasović, R. 2004. Gender in Indo-European. Heidelberg: Winter.
Mayrhofer, M. 1986. Lautlehre (Segmentale Phonologie des Indogermanischen). In:
M. Mayrhofer (ed.), Indogermanische Grammatik. Band I, 2. Halbband. Heidelberg:
Winter, 75–216.
Melchert, H.C. 2012. Genitive case and possessive adjective in Anatolian. In: V. Orioles
(ed.), Per Roberto Gusmani. Studi in ricordo. Linguistica storica e teorica. Vol. II, tomo
1. Udine: Forum, 273–286.
Melchert, H.C. fthc. The position of Anatolian. To appear in: M. Weiss; A. Garrett (eds.),
Handbook of Indo-European Studies. Oxford: Oxford University Press.
Mumm, P.-A. 2004. Altindisch sma. Teil 1: Rig- und Atharvaveda. International Journal
of Diachronic Linguistics and Linguistic Reconstruction 1, 19–68.
Nussbaum, A.J. 2010. PIE -Cmn- and Greek τρανής ‘clear’. In: R. Kim; N. Oettinger;
E. Rieken; M. Weiss (eds.), Ex Anatolia Lux: Anatolian and Indo-European Studies
in Honor of H. Craig Melchert on the Occasion of his Sixty-Fifth Birthday. Ann Arbor–
New York: Beech Stave Press, 269–277.
Oettinger, N. 2013/2014. Die Indo-Hittite-Hypothese aus heutiger Sicht. Münchener Stu-
dien zur Sprachwissenschaft 67, 149–176.
Pinault, G.-J. 2006. Retour sur le numéral “un” en tokharien. Indogermanische Forschun-
gen 111, 71–97.
Rieken, E. 2009. Der Archaismus des Hethitischen—eine Bestandsaufnahme. Incontri
Linguistici 32, 37–52.
Schindler, J. 1975. Zum Ablaut der neutralen s-Stämme des Indogermanischen. In:
H. Rix (ed.), Flexion und Wortbildung. Akten der V. Fachtagung der Indogermanischen
Gesellschaft. Wiesbaden: Reichert, 259–267.
Schmidt, J. 1895. Kritik der Sonantentheorie. Weimar: Böhlau.
Schmidt, J. 1898. Das Zahlwort μία, ἴα. Zeitschrift für Vergleichende Sprachforschung 36,
391–399.
Schrijver, P. 1997. Studies in the History of Celtic Pronouns and Particles. Maynooth:
Department of Old Irish, Saint Patrick’s College.
Mikhail Zhivlov*
* I would like to express my sincere gratitude to Simona Klemenčič, who kindly furnished me
with offprints of rare and unpublished works by Bojan Čop.
1 Except a brief mention in a footnote in Rasmussen (1999: 251).
predict which accent-ablaut paradigms are possible and which are not. See the
detailed critique of the paradigmatic approach in Kiparsky 2010.
An alternative approach, termed “morphophonological” by V. Dybo (2006)
and “compositional” by P. Kiparsky (2010), was developed mainly in Balto-
Slavic studies, although it can also be applied to Old Indic and Greek. Its main
premise is that the surface stress of any word in the languages that preserve
original Indo-European accent can be derived from the underlying accentual
properties of its constituent morphemes. All Proto-Balto-Slavic morphemes
belong to one of the two classes: a “+ valency” class and a “- valency” class. The
ictus falls on the first “+ valency” morpheme; if there are no such morphemes in
the word, the ictus falls on the first syllable (Dybo 2000: 5–14). According to the
hypothesis proposed by V. Dybo, morphophonological “valencies” reflect Proto-
Indo-European tones: “+ valency” goes back to high tone, while “- valency” goes
back to low tone (Dybo, Nikolayev & Starostin 1978). The tonal interpretation
was supported by F. Kortlandt (2010), A. Lubotsky (1988) and T. Pronk (2013).
The role of ablaut in the morphophonological/compositional approach re-
mains unclear. According to P. Kiparsky (2010), zero grade appears before an
immediately following underlyingly accented morpheme (= “+ valency” mor-
pheme in Dybo’s terminology). While plausible as a synchronic statement
when applied to Old Indic (or, more precisely, only to paradigms with ablaut
alternations in this language), Kiparsky’s rule is not easily interpretable in a
diachronic sense. Moreover, it explains only the hysterokinetic ablaut pattern,
leaving the amphikinetic ablaut unaccounted for (Kiparsky has an alternative
explanation for proterokinetic ablaut).
The fact that there was no synchronic direct connection between accent and
ablaut in Proto-Indo-European means that it is methodologically incorrect to
use ablaut data in the reconstruction of accent and vice versa. The only way to
avoid circular reasoning is to reconstruct accent and ablaut systems separately.
Therefore, we are justified in proposing a theory that accounts for the origin of
ablaut without reference to the Proto-Indo-European accent system. Of course,
such a theory must explain the limited correlation between ablaut and accent,
which is attested in Old Indic paradigms with mobile stress.
Below, we will use the terms ‘acrostatic’, ‘hysterokinetic’, ‘proterokinetic’ and
‘amphikinetic’ as names for ablaut paradigms only, without any connection to
accentology whatsoever. In order to emphasize this, we will not mark stress in
Proto-Indo-European forms.
Proto-Uralic roots have a minimal shape *CV(C)CV, i.e. they are at least disyl-
labic (see Aikio 2014: 45 for arguments in favour of this constraint). The typical
shape of suffixes and endings is *-CV(C) or (for endings) simply *-C (Janhunen
1982: 25–26). Two types of stems can be reconstructed for Proto-Uralic: *A-
stems, ending in a low vowel, realized as *a or *ä according to rules of vowel har-
mony,2 and *i-stems, ending in a high vowel *i. This high vowel “could alternate
with zero before suffixes comprising a whole syllable, provided the restrictions
on consonant distribution were not violated” (Janhunen 1982: 27). Both roots
and suffixes can end in *-A or *-i. The suffixes ending in *-i lose this vowel under
the same conditions as roots. Trisyllabic *i-stems lose the stem vowel not only
before certain suffixes, but also word-finally (cf. Helimski 2000). This peculiar-
ity is preserved in Finnic and Saami, i.e., in branches most faithfully preserving
the Proto-Uralic syllabic structure. It cannot be due to a recent apocope of final
*-i, because no such apocope can be postulated for Saami: Proto-Saami pre-
served the Uralic trisyllabic structures with final *-i, if this vowel belonged to
an ending (Salminen 1996). Proto-Uralic had fixed accent on the first syllable.
Although we cannot prove it, we can suppose that this feature was inherited
from Proto-Indo-Uralic.
The crucial element in our hypothesis is a set of comparisons between Proto-
Uralic and Proto-Indo-European noun endings. For Proto-Uralic, the following
case endings can be reconstructed (table 15.1 below is based on Janhunen 1982:
30 with minor changes).
Janhunen reconstructs the ablative ending as *-ti, because of its Proto-
Samoyed reflex *-tə. We prefer the traditional reconstruction *-tA (based on
West Uralic reflexes), since Proto-Uralic *a can yield Proto-Samoyed *ə in non-
first syllables.
The following Proto-Uralic endings can be etymologically equated with their
Proto-Indo-European counterparts (Čop 1975: 74–75):
2 Possibly, we must reconstruct more than one type of stems with non-high vowels for Proto-
Uralic (Zhivlov 2014: 117–121; Aikio 2015: 37–38). This does not affect our hypothesis, since it
is not vowel quality, but the presence or absence of vowel deletion that matters for our argu-
ment.
nom. -Ø -ki(-) -t
gen. -n -ki(-) -j
acc. -m -ki(-) -j
loc. -nA
abl. -tA
lat. -ŋ
The last two equations presuppose the sound change of a word-final *-t to *-s
at some stage between Proto-Indo-Uralic and Proto-Indo-European (after the
reduction of the final vowel in abl. *-ta). An additional example of this change
is the 2 sg. verbal ending: Proto-Uralic *-t ~ Proto-Indo-European *-s.
The Proto-Uralic dual ending is attested only in Samoyed and the Ob-Ugric
branches. In Samoyed it comes with an added *-ń of unclear origin, resulting
in Proto-Samoyed *-kəń (Janhunen 1998: 470). Since Proto-Uralic and Proto-
Samoyed do not tolerate word-final clusters, *-kəń may result not only from
PU *-ki- + *-ń, but also from PU *-k + *-ń with an epenthetic vowel. The Ob-
Ugric languages cannot clarify the picture either. Khanty has added another
nasal to the dual ending, resulting in Proto-Khanty *-γən, whereas the Proto-
Mansi dual ending *-γ is uninformative since the rules governing the fate of
word-final vowels in this branch are not sufficiently well studied. Thus, the PU
dual ending can be reconstructed either as *-ki or as *-k. Since the presence or
absence of a final vowel is crucial for our argument, below we will not discuss
dual forms.
The Proto-Indo-European nom.sg. ending *-s has no counterpart in Proto-
Uralic. We cannot accept the widespread idea that this ending is originally the
same as gen.-abl.sg. *-es/*-os/*-s, because the morphophonological properties
of these two endings differ (nom.sg. has “- valency”, does not trigger zero grade
of the preceding morpheme and is subject to Szemerényi’s law, whereas gen.-
abl.sg. has “+ valency”, triggers zero grade of the preceding morpheme and is
not subject to Szemerényi’s law).
Now let us assume that Proto-Indo-Uralic had the morpheme structure rules
described above for Proto-Uralic. Somewhat modifying Čop’s suggestions, we
postulate the following rules leading from Proto-Indo-Uralic to Proto-Indo-
European.
A. The accent became fixed on the penultimate syllable.
B. First syllable vowels were lengthened before second syllable *i. A similar, but
more restricted, change has taken place in Finnic (Aikio 2012). For the
sake of convenience, we will call the stage immediately after the applica-
tion of the rule B “pre-Proto-Indo-European”.
C. Unstressed vowels were reduced to zero. Judging by Proto-Indo-European
nom.pl. *-es (< Proto-Indo-Uralic *-t), the reduction was blocked at least
in the position before Proto-Indo-Uralic word-final *-t. In those cases
where the stem vowel was not subject to reduction, it was reanalyzed as
belonging to the following morpheme. Thus, Proto-Indo-European roots
became monosyllabic.
D. Short vowels yielded PIE *e, long vowels yielded PIE *o. Within Indo-
European, there is evidence that PIE *o goes back to an earlier long vowel
(Kümmel 2012: 307–317).
E. Word-final *-t became *-s (see above). This rule applies not only to the
original word-final *-t in nom.pl. ending, but also to *-t in the Proto-
Indo-Uralic ablative ending *-ta, which became word-final as a result
of unstressed vowel reduction. The rule was apparently blocked after
pronominal stems of the shape CV-: another reflex of the PIU abl. *-ta
in Proto-Indo-European is the abl. ending *-d of personal pronouns (e.g.
PIE *me-d—abl. of 1st sg. pronoun), while the Proto-Uralic acc. ending
of personal pronouns *-t (e.g. Finnish minu-t, Vakh Khanty män-t—acc.
of 1st sg. pronoun) corresponds to the PIE nom.-acc. neuter ending *-d
in pronominal declension (e.g. PIE *kʷi-d—nom.-acc.sg. of interrogative
pronoun). The latter restriction was proposed (in a somewhat different
wording) by P. Kallio in his presentation on the conference “The precur-
sors of Proto-Indo-European: The Indo-Hittite and Indo-Uralic hypothe-
ses”, held at Leiden University in 2015.
F. Rise of a new accent system. The accent shifted to the first syllable with
high tone (or simply to the first syllable, if there was no high tone in the
word). Whether the tones that served as input to this rule were inherited
from Proto-Indo-Uralic or developed later is irrelevant to our argument.
Lubotsky’s discovery that the accentuation of primary nouns in Old Indic
correlates with the voicedness/voicelessness of root consonants supports
the latter view (Lubotsky 1988).
The main differences between Čop’s ideas and our reformulation of them are
the following. 1) According to Čop, the accent in Indo-Uralic fell on the first syl-
lable of the stem, but if the stem was followed by a suffix containing a vowel,
the accent shifted onto the stem-final vowel immediately preceding this suffix
(Čop 1975: 92). We postulate instead a simpler rule of penultimate accent. 2)
We do not identify the accent system that gave rise to Indo-European ablaut
with the Proto-Indo-European accent system. 3) Our model also accounts for
the o-grade.
3 For simplicity’s sake, we will reconstruct the Proto-Indo-Uralic precursors of Proto-Uralic *A-
stems (where *A = *a or *ä) simply as *a-stems.
of the house’, Avestan də̄ṇg paitiš ‘lord’ (Schindler 1972: 32). The nominative
of the root noun ‘house’ is possibly preserved in the Armenian nom.sg. tun
< *dōm.
The crucial difference between Type 1 and Type 2 is that in the latter the
final vowel of the Proto-Indo-Uralic root was subject to deletion before suf-
fixes which contained a vowel. On the Uralic side, this type can be illustrated
by the Proto-Uralic word *käli ‘tongue’ (Finnish nom.sg. kieli, part.sg. kiel-tä <
PU abl. *käl-tä). The pre-Proto-Indo-European penultimate accent would fall
in such nouns on the root in gen.-abl.sg. as well as in nom.sg. The difference
between the two forms lay in the quality of the second syllable vowel in Proto-
Indo-Uralic: *i in the nominative, but *a (as a part of the ending *-ta) in the
ablative. This difference is reflected in the acrostatic ablaut: *o before original
Indo-Uralic *i in the second syllable / *e before original Indo-Uralic *a of the
second syllable.
The form *dem-s is one of the main counterexamples to Szemerényi’s law.
Another such counterexample is *-oms—the acc.pl. ending of thematic stems.
Both these forms have *s from Proto-Indo-Uralic *t: gen.-abl. sg. *-s corresponds
to Uralic abl. *-tA, while acc.pl. ending can be analyzed as acc.sg. *-m plus *-s
taken from the nom.pl. ending, going back to PIU *-t. We can suppose that these
forms escaped Szemerényi’s law because at the time of its action the reflex of
Proto-Indo-Uralic *t did not yet coincide with *s (it could have been a kind of
affricate).
Now we may turn to suffixed nouns. Since both roots and suffixes in Uralic
(and by our hypothesis, in Indo-Uralic) can end in *-a or *-i, the following com-
binations are possible.
Type 3. Root ending in *-a- plus suffix ending in *-a-: *CVCa- + *-Ca- >
*CVCa-Ca-. On the Uralic side this type can be illustrated by the Proto-Uralic
deverbal noun *elä-mä ‘life’, derived from PU *elä- ‘to live’ (Finnish nom.sg.
elämä, part.sg. elämää < PU abl. *elä-mä-tä).
Type 4. Root ending in *-a- plus suffix ending in *-i-: *CVCa- + *-Ci- >
*CVCa-Ci- / *CVCa-C-. The result is an *i-stem that will lose its final vowel
before certain affixes and word-finally (see above). The Uralic representative
of this type is PU *śüδ´ä-m ‘heart’, derived from PU *śüδ´ä ‘id.’ (Finnish nom.sg.
sydän, part.sg. sydäntä < PU abl. *śüδ´ä-m-tä, nom.pl. sydämet < PU *śüδ´ä-mi-
t).
Type 5. Root ending in *-i- plus suffix ending in *-a-: *CVCi- + *-Ca- > *CVC-
Ca-. The root loses its *-i- before any affix that contains a vowel. The Uralic reflex
of this type can be exemplified by PU *kal-ma ‘death’—an archaic derivative
from PU *kali- ‘to die’ (Finnish nom.sg. kalma, part.sg. kalmaa < PU abl. *kal-
ma-ta).
Type 6. Root ending in *-i- plus suffix ending in *-i-: *CVCi- + *-Ci- > *CVC-
Ci- / *CVC-C-. Here the root also loses its final vowel before a suffix, but the
outcome is an *i-stem that in its turn is capable of losing its final vowel. The
existence of this type is somewhat doubtful, mainly because it has no clear
parallel in Uralic. On the other hand, it is easy to understand why this type
could have been lost: Proto-Uralic normally did not tolerate clusters of more
than two consonants. In some *i-stems this constraint led to simplification of
resulting clusters (cf. Finnish lapsi ‘child’, part.sg. las-ta), but in most cases the
vowel deletion was given up.
Let us now look at possible development of types 3–6 in Indo-European.
The outcome of type 3 is the hysterokinetic ablaut paradigm: nom.sg. *uks-
ēn (< *uks-en-s), acc.sg. *uks-en-m, gen.-abl.sg. *uks-n-os, nom.pl. *uks-en-es >
Old Indic nom.sg. ukṣā ́, acc.sg. ukṣáṇam, gen.-abl.sg. ukṣṇáḥ, nom.pl. ukṣáṇaḥ
‘bull’.
Type 4 yields the proterokinetic paradigm. It appears to be originally restrict-
ed to neuter nouns, so acc.sg. and nom.pl. forms are not attested4 (expected, but
unattested reflexes here and below are marked with a double asterisk).
4 However, a trace of the expected ablaut paradigm is apparently preserved in the Greek pair
μέγα nom.-acc.sg.n. ‘great, big, large’ < *meg̑-h₂ vs. ἄγᾱν adv. ‘much, too much’ < *m̥ g̑-eh₂-m
acc.sg.m./f.
5 Nom.sg. ben < *gʷenā replaces earlier bé (neuter, later feminine) < *gʷēn. The latter is pre-
served as a poetic and legal term (Jasanoff 1989).
The expected outcomes of types 9 and 10 are not attested. We can suppose
that early in the prehistory of Indo-European, these types were given up and
replaced by proterokinetic and amphikinetic types, created on the analogy of
stems with VC-type suffixes. The model for such an analogy was provided by the
hysterokinetic type, which was present in stems with both VC-type and CVC-
type suffixes.
5 Conclusion
The hypothesis presented above does not account for the so-called acrostatic
type II with *ē / *e ablaut in the root and zero grade in the suffix. This type,
however, is not attested in any Indo-European language, and Kloekhorst con-
vincingly argues against its reconstruction (Kloekhorst 2014).
One more type remains unaccounted for: the so called ‘hysterodynamic’
type of the Leiden school: nom.sg. *CeC-C, acc.sg. *CC-eC-m, gen.sg. *CC-C-
es (Beekes 1985: 154). Kloekhorst argues that this ablaut type is synchronically
attested in the Hittite word for ‘hand’: nom.sg. keššar < *g̑ ʰesr, acc.sg. kiššeran <
*g̑ ʰserom, gen.sg. kiš(ša)raš < *g̑ ʰesros or *g̑ ʰsros (Kloekhorst 2013: 111–115). This
type looks like the predicted outcome of our type 4 in non-neuter nouns (in
References
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2012. On Finnic long vowels, Samoyed
vowel sequences, and Proto-Uralic *x. In: T. Hyytiäinen; L. Jalava; J. Saarikivi and
E. Sandman (eds.), Per Urales ad Orientem: Iter polyphonicum multilingue. Fest-
skrift tillägnad Juha Janhunen på hans sextioårsdag den 12 februari 2012. Helsinki:
Suomalais-Ugrilainen Seura, 227–250.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2014. The Uralic-Yukaghir lexical correspon-
dences: genetic inheritance, language contact or chance resemblance? Finnisch-
ugrische Forschungen 62, 7–76.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2015. The Finnic ‘secondary e-stems’ and
Proto-Uralic vocalism. Suomalais-Ugrilaisen Seuran Aikakauskirja / Journal de la
Société Finno-Ougrienne 95, 25–66.
Beekes, R.S.P. 1985. The Origins of the Indo-European Nominal Inflection. Innsbruck: IBS.
Beekes, R.S.P. 2011. Comparative Indo-European Linguistics. An Introduction. Second edi-
tion. Revised and corrected by Michiel de Vaan. Amsterdam–Philadelphia: John
Benjamins.
Čop, B. 1975. Die indogermanische Deklination im Lichte der indouralischen vergleichen-
den Grammatik. Ljubljana: SAZU.
Čop, B. 1990. Indouralica XVII. Razprave SAZU, II. razred, razred za filološke in literarne
vede 13, 21–46.
Dybo, V.A. 2000. Morfonologizovannye paradigmatičeskie akcentnye sistemy: Tipologija
i genezis. Tom 1. Moskva: Jazyki russkoj kul’tury.
Dybo, V.A. 2006. Sravnitel’no-istoričeskaja akcentologija, novyj vzgljad: po povodu
knigi V. Lefel’dta “Vvedenie v morfologičeskuju koncepciju slavjanskoj akcentologii”.
Voprosy jazykoznanija 2006/2, 3–27.
Dybo, V.; S. Nikolayev; S. Starostin 1978. A tonological hypothesis on the origin of
paradigmatic accent systems. Estonian Papers in Phonetics 1978, 16–20.
Helimski, E. 2000. Bisyllabic consonantal and trisyllabic vocalic stems in Finno-
Permian and further. In: E.A. Xelimskij Komparativistika, uralistika: Lekcii i stat’i.
Moskva: Jazyki russkoj kul’tury, 191–195.
Janhunen, J. 1982. On the structure of Proto-Uralic. Finnisch-ugrische Forschungen 44,
23–42.
Janhunen, J. 1998. Samoyedic. In: D. Abondolo (ed.), The Uralic Languages. London–
New York: Routledge, 457–479.