IE KloekhorstPronk-2019 (Precursors)

The Precursors of Proto-Indo-European
Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Downloaded from Brill.com11/07/2019 03:07:37PM
via free access
Leiden Studies in Indo-European
Editor
Alexander Lubotsky
volume 21
The titles published in this series are listed at brill.com/lsie

via free access
The Precursors of
Proto-Indo-European
The Indo-Anatolian and Indo-Uralic Hypotheses
Edited by
Alwin Kloekhorst
Tijmen Pronk
LEIDEN | BOSTON

via free access
The Library of Congress Cataloging-in-Publication Data
Names: Kloekhorst, Alwin, editor. | Pronk, Tijmen, editor.

Title: The precursors of Proto-Indo-European : the Indo-Anatolian and Indo-Uralic
hypotheses / edited by Alwin Kloekhorst, Tijmen Pronk.
Description: Boston : Brill, 2019. | Series: Leiden studies in Indo-European,
0926-5856 ; 21 | Includes bibliographical references and index.
Identifiers: LCCN 2019031210 (print) | LCCN 2019031211 (ebook) |
ISBN 9789004409347 (hardback) | ISBN 9789004409354 (ebook)
Subjects: LCSH: Proto-Indo-European language–History. | Proto-Indo-European
language–Grammar, Comparative. | Extinct languages.
Classification: LCC P572 .P74 2019 (print) | LCC P572 (ebook) |
DDC 417/.7–dc23
LC record available at https://lccn.loc.gov/2019031210
LC ebook record available at https://lccn.loc.gov/2019031211
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)
Copyright 2019 by Koninklijke Brill NV, Leiden, The Netherlands.

Koninklijke Brill NV incorporates the imprints Brill, Brill Hes & De Graaf, Brill Nijhoff, Brill Rodopi,
Brill Sense, Hotei Publishing, mentis Verlag, Verlag Ferdinand Schöningh and Wilhelm Fink Verlag.
All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise,
without prior written permission from the publisher.
Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill NV provided
that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive,
Suite 910, Danvers, MA 01923, USA. Fees are subject to change.
This book is printed on acid-free paper and produced in a sustainable manner.

via free access
Table of Contents
Preface vii
1 Introduction: Reconstructing Proto-Indo-Anatolian

and Proto-Indo-Uralic 1
Alwin Kloekhorst and Tijmen Pronk
2 The Proto-Indo-European Suffix *-r Revisited 15

Stefan Heinrich Bauhaus
3 Pronouns and Particles: Indo-Uralic Heritage and Convergence 30

Rasmus Gudmundsen Bjørn
4 Indo-Anatolian Syntax? 50
Dag Haug and Andrei Sideltsev
5 Daniel Europaeus and Indo-Uralic 74

Petri Kallio
6 Bojan Čop’s Indo-Uralic Hypothesis and Its Plausibility 88

Simona Klemenčič
7 Indo-European o-grade Presents and the Anatolian ḫi-conjugation 102

Frederik Kortlandt
8 The Proto-Indo-European mediae, Proto-Uralic Nasals from a Glottalic

Perspective 111
Guus Kroonen
9 Thoughts about Pre-Indo-European Stop Systems 115

Martin Joachim Kümmel
10 The Anatolian “Ergative” 131

Milan Lopuhaä-Zwakenberg
11 The Indo-European Suffix *-ens- and Its Indo-Uralic Origin 151

Alexander Lubotsky

via free access
vi table of contents
12 Headedness in Indo-Uralic 163

Rosemarie Lühr
13 Indo-Uralic, Indo-Anatolian, Indo-Tocharian 186

Michaël Peyrot
14 Proto-Indo-European *sm and *si ‘one’ 203

Michiel de Vaan
15 Indo-Uralic and the Origin of Indo-European Ablaut 219

Mikhail Zhivlov

via free access
Preface
‘The Precursors of Proto-Indo-European’ was the name of a workshop held at

the Leiden University Centre for Linguistics from 9–11 July 2015,1 during which
some thirty papers were presented by specialists from all over the world, deal-
ing with the reconstructible prehistory of Proto-Indo-European, with a focus on
the Indo-Anatolian and the Indo-Uralic hypotheses. The workshop turned out
to be an inspiring event, where scholars from different backgrounds exchanged
ideas and arguments, and where new data, analyses and visions were pre-
sented. In the course of the workshop, and particularly during the closing ses-
sion, it became clear that all participants, speakers as well as members of the
audience, felt the need for a dedicated volume dealing with the Indo-Anatolian
and Indo-Uralic hypotheses. Hence the idea was born to compile the work that
now lies before you.
This book contains adapted versions of a selection of the papers presented at
the Leiden workshop, but also several papers that were written especially for
this volume. It aims to provide an up-to-date overview and discussion of the
arguments for and against the Indo-Anatolian and the Indo-Uralic hypotheses
and related issues. We expect it to serve as a reference tool and inspiration for
new research into the prehistory of Proto-Indo-European.
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).

via free access
via free access
chapter 1
Introduction: Reconstructing Proto-Indo-Anatolian

and Proto-Indo-Uralic*
Alwin Kloekhorst and Tijmen Pronk
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.
1 The Nature of Proto-Indo-European
Before we try to answer the question whether any outer-Indo-European rela-

tives can be identified, we first need to be explicit about what exactly is meant
by the term Proto-Indo-European. In theory, the answer is straightforward:
what we call Proto-Indo-European should correspond to the proto-language
as it was spoken immediately before the first diversification took place that
resulted in its eventual dissolution into the Indo-European daughter languages.
However, in practice it is not always easy to determine what the proto-language
looked like at this stage. In part, this is due to the history of the field of compar-
ative Indo-European linguistics. Beginning with Sir William Jones’ observation
that Sanskrit bears “a stronger affinity” to Greek and Latin “than could possibly
have been produced by accident”, the field of Indo-European linguistics ini-
tially focused on the evidence from especially these three languages, resulting
in the classical reconstruction of late 19th century Proto-Indo-European as can
* 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).
© koninklijke brill nv, leiden, 2019 | doi:10.1163/9789004409354_002

via free access
2 kloekhorst and pronk
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
2 The Indo-Anatolian Hypothesis
Hittite also changed the reconstruction of Proto-Indo-European in another,

fundamental way. Although in some respects Hittite seems to be a very archaic
Indo-European language, e.g. by preserving verbal ablaut patterns better than
any other ancient Indo-European language, in other respects it turned out to be
radically different from the other languages. For instance, Hittite lacks a num-
ber of important linguistic categories that are present in Greek and Sanskrit,
like the feminine gender, the aorist, and the perfect, all of which had always
been regarded as core features of Proto-Indo-European. In order to account for
these facts, already in the 1920s, only a few years after its decipherment, it was
hypothesized that Hittite should not be viewed as another daughter language
of Proto-Indo-European, but rather as its sister language (Forrer 1921). This
would mean that Hittite and Proto-Indo-European both derive from an even
earlier proto-language, which was coined ‘Indo-Hittite’ by Sturtevant (1933:
30). Since we nowadays know that these special characteristics of Hittite are
found in the entire Anatolian branch, it is more appropriate to speak about
the ‘Indo-Anatolian’ hypothesis, and we will therefore use this term in this
book.
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).

via free access
introduction 3
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).

via free access
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:

via free access
introduction 5
24. Hitt. unreduplicated ḫi-conjugation vs. cl.PIE reduplicated perfect: gener-

alization of reduplication in the perfect (Kloekhorst 2018a) [but the pres-
ence of (traces of) unreduplicated perfects in ‘classical’ Indo-European,
esp. in Germanic and Balto-Slavic, may indicate that the generalization
of reduplication was either not absolute, or not shared by all bran-
ches].
25. Hitt. 1pl. -u̯ en(i) vs. cl.PIE dual *-u̯ e(-): development of a clusivitiy system
to a plural/dual system (Kloekhorst 2017b) [but it cannot be ruled out that
Hittite developed the plural ending from an original dual ending].
26. Hitt. lāḫu- ‘to pour’ < *leh₃u- vs. cl.PIE *leuh₃- ‘to wash’: laryngeal metathe-
sis (Oettinger 2013/14: 169) [but the details of possible laryngeal metathe-
sis in Anatolian are unclear].
27. Anat. *[tː], *[ʔt], *[t] vs. cl.PIE *t, d, *dh (*t, *ʔd, *d): consonant shift
(Kloekhorst 2016) [but cf. Kümmel, this volume, 115–130, for criticism of
Kloekhorst’s scenario].
28. Hitt. -(e)t < *-(e)t vs. cl.PIE *-(e)h₁ (instr. ending): development of PIH *-t
> cl.PIE *-ʔd > *-ʔ (Kortlandt 2010: 41) [but the exact conditions for the
proposed sound change remain unclear].
29. Anat. *mK vs. cl.PIE *nK in *h₂emǵh- > *h₂enǵh- ‘to tie, to restrict’ and
*temk- > *tenk- ‘to solidify, to coagulate’: assimilation (Eichner 2015: 1616)
[but it cannot be ruled out that the assimilation took place independently
in different branches; see Pronk 2010 for potential evidence for the *m of
*h₂emǵh- in Greek].
30. Anat. *-ms vs. cl.PIE *-ns (acc.pl. ending): assimiliation [but it cannot
be ruled out that the assimilation took place independently in different
branches].
31. Hitt. e-eš-ši < *h₁essi vs. cl.PIE *h₁esi ‘you are’: degemination of *ss to *s
(Kloekhorst 2016: 238–240) [but the ending *-si may have been restored
in Hittite].
32. Hitt. šiun < *diéum/*diḗum vs. cl.PIE *diḗm ‘god (acc.sg.)’: ‘Stang’s Law’
[but the Hitt. acc.sg. šiun may have been formed in analogy to the nom.sg.
šiuš].
33. Hitt. dā-i ‘to take’ vs. cl.PIE *deh₃- ‘to give’: semantic innovation (Norbruis
fthc.b) [but the innovation may not be shared by all non-Anatolian IE
branches].
34. Hitt. causatives of the shape *CóC-e(i) (e.g. lāki ‘he knocks down < he
makes lie down’ < *lógh-e(i)) vs. cl.PIE causatives of the shape *CoC-
eie/o- (e.g. *logh-éie/o- ‘to make lie down’): innovation of the *CoC-eie/o-
causative (Kloekhorst 2018a: 10028) [but this depends on the status of the
dūpiti-type in the Luwic languages].

via free access
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

via free access
introduction 7
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.
3 Indo-Tocharian and the Indo-Anatolian Hypothesis
‘Classical’ Proto-Indo-European, which above was taken as comprising all non-

Anatolian Indo-European languages, can be further divided into several
branches that split off in a certain sequence. It seems likely that the Tocharian
branch was the second branch to split off after Anatolian, as argued by Peyrot
(this volume, 186–202) and others before him. Peyrot proposes to use the term
Proto-Indo-Tocharian for the stage just preceding this split (with ‘core’ Indo-
European for the remaining languages). He rightly points out that arguments
in favour of the Indo-Anatolian hypothesis should always be weighed against
this Indo-Tocharian stage: if a certain hypothesized post-Anatolian innovation
cannot be shown to have affected Tocharian, it cannot in principle be used as
an argument for the Indo-Anatolian hypothesis, because the innovation could
also have taken place in post-Tocharian ‘core’ Indo-European.
In theory, this concept should be rigorously applied: for each linguistic inno-
vation it should be determined when it can be dated in relation to all nodes in
the family tree. In practice, however, the exact order of the splits in the Indo-
European family tree is uncertain, especially after the split of Tocharian, and
precise dating of innovations is often impossible. Therefore, it remains useful to
operate with larger, less specific entities with relatively vague names like ‘clas-
sical’ Proto-Indo-European (all or most Indo-European languages except Ana-
tolian) or ‘core’ Indo-European (all or most Indo-European languages except
Anatolian and Tocharian).
It is not fully clear to what extent Tocharian participated in all the post-
Anatolian innovations that were listed above: cf. Peyrot, this volume 188, who
points out the fact that e.g. *mer- (innovation no. 4 in the list above) is unat-
tested in Tocharian, as a result of which it cannot be determined whether the
semantic development of ‘to disappear’ to ‘to die’ was a post-Anatolian or a
post-Tocharian innovation. Similar reservations apply to other post-Anatolian

via free access
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
Another way of reconstructing one or more precursor stages of Proto-Indo-

European, without taking into account language material from outside the
Indo-European family, is internal reconstruction. Like any other language, also
Proto-Indo-European (or Proto-Indo-Anatolian) contained in its grammar
irregularities and other features that may be explained as the result of a rel-
atively recent development. For instance, the word for ‘hundred’ can be recon-
structed as *h₁ḱmtóm on the basis of e.g. Skt. śatám, Gr. ἑκατόν, Lat. centum,
OE hund, Lith. šim̃ tas. Nevertheless, on the basis of the assumption that ‘hun-
dred’ is derived from the numeral ‘ten’ (*déḱm, cf. Skt. dáśa, Gr. δέκα, Lat. decem,
Goth. taihun, etc.), it is usually assumed that *h₁ḱmtóm goes back to an ear-
lier *dḱmtóm (with *d > *h₁, cf. Garnier 2014). This latter form, which is the
result of internal reconstruction, must thus be assigned to a precursor stage
of Proto-Indo-European. In some cases, we can even distinguish several subse-
quent precursor stages. Take, for instance, the phenomenon of ablaut: already
Brugmann assumed different layers in the prehistory of Proto-Indo-European
to account for the PIE ablaut alternations (see also Kortlandt, this volume, 102–
110).
Of course, on the basis of internal reconstruction alone it is impossible to
reconstruct all details of these different precursors. Nevertheless, we view inter-
nal reconstruction as a vital way to penetrate as deeply into the prehistory of
Proto-Indo-European as possible, which is a prerequisite before one can start
with external comparison.

via free access
introduction 9
5 External Comparison
When it comes to comparing the Indo-European language family to one or

more non-Indo-European languages and/or language families, several sugges-
tions have been made for identifying possible relatives. Some of these sug-
gestions include large macro-families, like the ‘Nostratic’ family, which is usu-
ally thought to include Indo-European, Uralic, Kartvelian, Altaic, Japonic and
Koreanic (the latter three possibly forming a single ‘Transeurasian’ family), but
to which sometimes Afroasiatic, Dravidian, Chukotko-Kamchatkan, Eskimo-
Aleut, and other languages or language families are added as well (Peder-
sen 1903, Illich-Svitych 1971–1984, Starostin 1989, Dolgopolsky 2008, Bomhard
2008). Although we are not principally opposed to the concept of such macro-
families, we think it is methodologically preferable to start with one-to-one
comparisons in order to be able to reconstruct deeper in time step by step.
6 The Indo-Uralic Hypothesis
Already in the 19th century, the linguistic similarities between Indo-European

and Uralic led to the hypothesis that the Indo-European language family may
be related to Uralic (see Kallio, this volume, 74–87 about the earliest Indo-
Uralicists). We believe that this is still a valid point of view. The similarities
are found both in the morphology and in the lexicon. Kortlandt (2002) listed
no fewer than 27 morphemes of Indo-European and Uralic that are phoneti-
cally so similar to each other that he regards them as “definitely Indo-Uralic”.
This list includes pronominal morphemes (see also Bjørn, this volume, 30–
49), case markers (see also Bauhaus, this volume, 15–29), as well as verbal and
nominal suffixes (see also Lubotsky, this volume, 151–162). The lexical simi-
larities between Indo-European and Uralic are often attributed to borrowing
from Indo-European into Uralic (cf. Koivulehto 1994, 2001, 2003), but there are
reasons to believe that at least some lexical correspondences are due to inheri-
tence from a common source. The oldest layer of shared lexicon consists of
pronouns, nouns and verbs belonging to the part of the vocabulary that is least
prone to being borrowed (Napol’skix 1997: 147–148, Helimski 2001, Kümmel,
this volume, 115–130). This implies that the similarities are due to shared ances-
try and not to borrowing.
If Indo-European and Uralic are indeed related to each other, both should go
back to a common ancestor, Proto-Indo-Uralic, which can then be regarded as
a precursor of Proto-Indo-European (and of Proto-Indo-Anatolian). This is the
reason why the second part of the subtitle of this book refers to the Indo-Uralic
hypothesis.

via free access
Although we regard the Indo-Uralic hypothesis as very likely to be correct,

this does not mean it is easy to start reconstructing Proto-Indo-Uralic. There is
at this moment no consensus on the relationship between the phoneme inven-
tories of the two language families (see Klemenčič, this volume, 88–101, on
Čop’s attempts to find correspondences, and Kümmel, this volume, 115–130 and
Kroonen, this volume, 111–114, on possible correspondences in the consonant
system), nor on the shared lexicon (Illich-Svitych 1971–1984, I: 6–37, Helimski
2001: 19619), on the morphological relationships (see Zhivlov, this volume, 219–
235 for a possible connection between nominal paradigms in Indo-European
and Uralic), or on connections in other parts of grammar (see Lühr, this vol-
ume, 163–185 for a possible syntactic connection). This difficulty may be partly
explained by the possibility that, after the dissolution of Indo-Uralic, Indo-
European has undergone relatively strong substrate influence from North Cau-
casian (see Kortlandt 2018).
Another question regarding Proto-Indo-Uralic that remains to be answered
is where and when it was spoken. Post-Anatolian Proto-Indo-European vocab-
ulary is thought to reflect a Chalcolithic stage of development, while Proto-
Uralic vocabulary represents a Mesolithic society (Janhunen 2009). This does
not mean that Proto-Uralic must be dated much earlier than Proto-Indo-
European. The difference is more plausibly connected with the geographic
area in which the two proto-languages were spoken. There appears to be con-
sensus among Indo-Europeanists that Proto-Indo-European was spoken in the
Pontic-Caspian steppes in the middle of the fourth millennium BCE. Proto-
Indo-Anatolian can perhaps be dated to the middle or late fifth millennium
BCE in the same region (Anthony & Ringe 2015). The Proto-Uralic homeland
was probably located near the Ural mountains, either in the west between
the Volga river and the Central Ural mountains (Häkkinen 2009), or to their
east, in the vicinity of the rivers Ob and Yenisei (Napol’skix 1997: 135, Janhunen
2009). Traditionally, the time-depth of Proto-Uralic is estimated to be around
4000BCE (Napol’skix 1997, Helimski 2001), but a more shallow date of approx-
imately 3000BCE (Janhunen 2009) or 2000BCE (Kallio 2006, Häkkinen 2009)
now seems to be more plausible. There is thus a gap of up to 2500 years between
Proto-Indo-Anatolian and Proto-Uralic.
If Indo-Anatolian was indeed influenced by a North Caucasian substrate, it
stands to reason that its ancestor moved into the steppes north of the Caucasus
from somewhere else. The Uralic connection suggests that this somewhere else
must be sought more to the north and/or the east. As for the question when
Proto-Indo-Uralic was spoken, only a very rough estimate is possible on the
basis of the rather limited number of etymological correspondences between
the basic lexicons of Indo-European and Uralic. The relatively low number of

via free access
introduction 11
figure 1.1 The precursors of Proto-Indo-European
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.

via free access
Bomhard, A.R. 2008. Reconstructing Proto-Nostratic: Comparative Phonology, Morphol-

ogy, and Vocabulary. Leiden–Boston: Brill.
Damgaard, P. de B. et al. 2018. The first horse herders and the impact of Early Bronze
Age steppe expansions into Asia. Science 360, eaar7711.
Dolgopolsky, A. 20123. Nostratic Dictionary. Published online at http://www.dspace.cam
.ac.uk/handle/1810/244080.
Eichner, H. 2015. Das Anatolische in seinem Verhältnis zu anderen Gliedern der indo-
europäischen Sprachfamilie aus aktueller Sicht. In: T. Krisch; S. Niederreiter (eds.),
Diachronie und Sprachvergleich. Beiträge aus der Arbeitsgruppe “historisch-verglei-
chende Sprachwissenschaft” bei der 40. Österreichischen Linguistiktagung 2013 in Salz-
burg. Innsbruck: IBS, 13–26.
Forrer, E. 1921. Ausbeute aus den Boghazköi-Inschriften. Mitteilungen der deutschen
Orient-Gesellschaft 61, 20–39.
Garnier, R. 2014. Nouvelles réflexions sur l’effet-Kortlandt. Glotta 90, 140–160.
Haak, W. et al. 2015. Massive migration from the steppe was a source for Indo-European
languages in Europe. Nature 522, 207–211.
Häkkinen, J. 2009. Kantauralin ajoitus ja paikannus: perustelut puntarissa. Journal de
la Société Finno-Ougrienne 92, 9–56.
Helimski, E. 2001. Early Indo-Uralic linguistic relationships. In: C. Carpelan; A. Parpola;
P. Koskikallio (eds.), Early Contacts between Uralic and Indo-European: Linguistic and
Archaeological Considerations. Helsinki: Finno-Ugrian Society, 187–205.
Illich-Svitych, V.M. 1971–1984. Opyt sravnenija nostratičeskix jazykov (semitoxamitskij,
kartvel’skij, indoevropejskij, uralskij, dravidskij, altajskij). Moskva: Nauka.
Janhunen, J. 2009. Proto-Uralic—what, where, and when? In: J. Ylikoski (ed.), The
Quasquicentennial of the Finno-Ugrian Society. Helsinki: Société Finno-Ougrienne,
57–78.
Kallio, P. 2006. Suomen kantakielten absoluuttista kronologiaa. Virittäjä 110, 2–25.
Kloekhorst, A. 2008. Etymological Dictionary of the Hittite Inherited Lexicon. Leiden–
Boston: Brill.
Kloekhorst, A. 2016. The Anatolian stop system and the Indo-Hittite hypothesis. Indo-
germanische Forschungen 121, 213–247.
Kloekhorst, A. 2017a. The Hittite genitive ending -ā ̆n. In: B. Simmelkjær Sandgaard
Hansen; A. Hyllested; A.R. Jørgensen; G. Kroonen; J.H. Larsson; B. Nielsen White-
head; T. Olander; T. Mosbæk Søborg (eds.), Usque ad Radices: Indo-European Studies
in Honour of Birgit Anette Olsen. Copenhagen: Museum Tusculanum, 385–400.
Kloekhorst, A. 2017b. The Hittite verbal system and the Indo-Hittite hypothesis. Paper
presented at The Split. Reconstructing Early Indo-European Language and Culture,
University of Copenhagen, 13 September 2017.
Kloekhorst, A. 2018a. The origin of the Hittite ḫi-conjugation. In: L. van Beek; M. de
Vaan; A. Kloekhorst; G. Kroonen; M. Peyrot; T. Pronk (eds.), Farnah: Indo-Iranian

via free access
introduction 13
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.

via free access
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

via free access
chapter 2
The Proto-Indo-European Suffix *-r Revisited

Stefan Heinrich Bauhaus
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

via free access
16 bauhaus
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
The function of *-r in some pronominal interrogatives and demonstratives

is beyond dispute. I agree with Dunkel (2014) that this element is identical

via free access
the proto-indo-european suffix *-r revisited 17
to the one we find in formations of other kinds. The examples usually referred
to are the following:
Indo-Iranian: Skt. kár-hi ‘when’, tár-hi ‘(back) then’;

Germanic: Goth. ƕar ‘where’, þar ‘there’, hêr ‘here’;
Italic: Lat. cūr < quōr ‘why’;
Baltic: Lith. kur̃ ‘where’, Latv. tùr ‘there’;
Armenian: ur ‘where’;
Albanian: kur ‘when’.
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).

via free access
18 bauhaus
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.

via free access
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
4 Dunkels translation is “sich früh erhebend”.

via free access
20 bauhaus
formation would then be a hypostasis of a locative extended by *-i, which again

implies that at that stage the underlying locative was not regarded as a locative
anymore. The variation in ablaut in the quoted examples is a difficult topic. The
Armenian example points to zero-grade.
If *-r indeed is a case ending, there must be a regular pattern of ablaut. In
view of the accent / ablaut patterns of the nouns, I think that we are origi-
nally dealing with locatives that had the same oblique stem as the genitive.
Unless the root was acrostatic, the ending was accented and consequently
had e-grade. This assumption is supported by the reconstruction *dhǵhm-ér,
which is reflected in Av. zəmar-, but also by Skt. uṣar- and Arm. jmer̄n ‘win-
ter’.
The problem of double full grade in some formations (such as Skt. vasar-),
which Nussbaum (1986: 190, 291–292) mentions, could then be attributed to
analogy. Once the suffix *-r had lost its productivity as a locative case marker,
it was reanalyzed as an adverbial ending. In that new function its common
shape was *-er, with full grade. It could then be attached to full graded roots, as
well. This suggests that Skt. vasar- is younger then uṣar-. Nevertheless, it is also
possible that we are merely dealing with analogy due to paradigmatic leveling,
possibly displayed in Lat. hībernus and Gr. χειμέριος in contrast to Arm. jmer̄n. It
is not necessary to assume that Greek χειμέριος goes back to a locative ǵheim-er.
The full-grade in the root could have been arisen analogically to χειμών ‘winter’.
The scenario in an early stage of PIE, before the formations became lexicalized
to adverbs, probably looked like this:
a) loc. *dhǵhm-ér ‘in/on the earth’ vs. gen.*dhǵhm-és;
b) loc. *h₂us-ér ‘at dawn’ vs. gen. *h₂us-és;
c) loc. *ǵhim-ér ‘in winter’ vs. gen. *ǵhim-és;
d) loc. *nókwt-or ‘at night’.
Latin shows many more adjectives ending in -rnus. I consider it very likely that
the whole suffix is a derivative of the ending in *-r. For instance, we find Lat.
diurnus ‘daily’, vespernus ‘vespertine’, hesternus ‘of yesterday’, sempiternus ‘per-
petual’, aeternus ‘eternal’. If we understand this suffix as a secondary derivation
of the case ending *-r, we are perhaps dealing with some more locatives in
*-r. Considering Dunkel’s etymology of Gr. νυκτερινός ‘nocturnal’, we can ana-
lyze Lat. -rnus as -(e)r-i-no-, originating in locatives in *-r extended by locatival
*-i plus *-no-. For Lat. diurnus, Dunkel (2014, I: 164) reconstructs *di̭úr ‘dur-
ing the day’. De Vaan (2008 on Lat. diū, diu) rather assumes that diurnus is
analogical to noctūrnus. Whereas Lat. vespernus clearly derives from the noun
vesper ‘evening’, which Dunkel (2014, I: 166) also explains as a locative built
directly to a root, Lat. hesternus finds a striking and well know parallel in Ger-
manic, for instance OHG gestaron ‘yesterday’. Both show a similar extension as

via free access
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.
4 Local Adverbs and Adpositions
Dunkel assumes that an “Adverbialendung” *-r/*-er is also reflected in other

adverbs and adpositions, such as PIE *upér(i). He subsumes these under the
domain of particles. I agree with him on this latter point, but I consider them
to be nominal formations as well, at least on a diachronic level. Just as with the
aforementioned primarily nominal formations, these locatives can have two
major sources:
a) plain adverbs/adpositions
b) certain adjectives
I regard the plain adverbs/adpositions as locatives of nouns denoting locality.
One example for this is Gr. ὕπερ ‘over’ ~ Skt. upári ‘above’, which point to an
original formation *h₁uper. Adjectives like Skt. úpara- ‘lower, later’ should be
analyzed as hypostases to this formation. Their attributed comparative mean-
ing must be secondary. Formations like Lat. īnfernus ‘inferior’, on the other
hand, remind us of the nominal formations in -rnus treated above. Some of
the relevant adverbs/adpositions are the following:5
*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.

via free access
22 bauhaus
To these formations I would like to add a new reconstruction:
*ḱ(e)nter: Goth. hindar ‘behind’, Hitt. kattera- ‘lower, inferior’.
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).

via free access
PIE *h₁up- ‘top’: loc.*h₁up-ér ‘on top’ > *h₁upér (adv.) ‘above’ > * h₁upér
(adp.) ‘over X’.
Summarizing, I reconstruct the following set of locatives for an early phase of

PIE:
PIE *h₁up- ‘surface’: loc. *h₁uper ‘on the surface, on top’;

PIE *h₂ep- ‘back’: loc. *h₂eper ‘in the back’;
PIE *ḱent- ‘back, underside’: loc. *ḱ(e)nter ‘in the back, at the underside’;
PIE *ndh- ‘underside, bottom’: loc. *ndher ‘at the underside, bottom’.
Besides these formations there is a considerable number of adverbs, adposi-

tions and adjectives that point to a suffix *-ter rather than *-er. For instance,
we have Lat. inter ‘within, between’ and Skt. antár ‘inside, in between’ which
point to PIE *h₁en-ter.
Dunkel (2014, I: 180–181) reconstructs a suffix *-tér as a separate “Adverbial-
endung” that is attached to other particles. In this case, I agree that we are
dealing with an “Adverbialendung”. I think that this suffix came to rise at a
time when the locatives to nouns had already been lexicalized into adverbs. The
question is why there is this enlargement with -t- in the first place. We do find
similar variants of other suffixes as well, cf. *-ti vs. *-i but also *-tos vs. (gen.-abl.)
*-e/os. Dunkel (2014, I: 177) assumes that we are dealing with the Adverbialen-
dung *-t + locatival *-er. But I assume that the solution is much simpler. The
variant *-ter can just be understood as a false segmentation of stems ending
in -t, for instance *ḱnt-er → *ḱn-ter. This *-ter was then generalized and could
perhaps be attached to already lexicalized adverbs, like *h₁en-ter to *h₁en. This
idea could also explain the aforementioned Latin adjectives like sempiternus
and aeternus which point to adverbs in *-ter. They could be due to false seg-
mentations as well, maybe also furnished by *nokwt-e/or → *nokw-ter, and again
be paralleled to formations like internus.
The assumption that *-ter is more recent than *-(e)r and came into being
when the latter had lost its productivity explains why we have doublets like
Engl. after < *h₂epter vs. Goth. afar ‘after’ < *h₂eper but also Du. zonder ‘with-
out’ <*sn-ter vs. Av. hanarə ‘without’ < *sn̥ n-ér (reconstructions according to
Dunkel 2014, I: 163; II: 712). More recent formations like Lat. hesternus and
OHG gestaron could be added to these examples. Instances for the suffix *-ter
include:
*h₂ep-ter: Engl. after;

*h₁en-ter: Lat. inter ‘within, between’, Skt. antár ‘inside, in between’;

via free access
24 bauhaus
*sn-ter: Du. zonder ‘without’;

*pro-ter: Skt. prātár ‘early’.8
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.
5 The Connection to the Directive
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:
8 Other Skt. formations in -(t)ar are treated in Kümmel (2002: 44–47).

9 Provided we do accept *-o rather than *-h₂o, cf. e.g. Dunkel (1994) and Kloekhorst (2008: 126,
195).

via free access
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:
PIE *h₁up- ‘surface’: dir. *h₁upo ‘to the surface’;

PIE *h₂ep- ‘back’: dir. *h₂epo ‘to the back’;
PIE *ḱent- ‘back, underside’: dir. *ḱnto ‘to the back, underside’;
PIE *ndh- ‘underside, bottom’: dir. *ndho ‘to the underside, bottom’.
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

via free access
26 bauhaus
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).
Early PIE Late PIE
locative *-r *-i

directive *-o *-m = accusative
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.

via free access
7 Further Implications
Once a more archaic locative ending in *-r is reconstructed next to a more

recent one in *-i, one thing becomes apparent. One is reminded of the oppo-
sition in the verbal system, where we find the hic-et-nunc-i in contrast to the
mediopassive marker *-r.10 Dunkel (2014, I: 165) rejects the identification of the
latter one with his locatival *-r as it bears semantic problems. However, it is
striking that we are dealing with very archaic formations in both cases.
Kortlandt (2002) assumes a Proto-Indo-Uralic morpheme *ru which is also
reconstructed by Greenberg (2000: 147–150 on RU). He also sees evidence for
this element in Altaic (Kortlandt 2010: 416). Greenberg has even gone one
step further in discussing this element, detecting similar formations in, for
instance, Ainu and even Nivkh. However, he assumes an original directive
meaning (Greenberg 2000: 149). I am uncertain as to what degree I can follow
him, but his Komi-Zyrianic interrogative kor ‘when’ could be a likely candidate
for an Uralic equivalent to the pronominal formations in Indo-European lan-
guages. Just as with Skt. kar-hi ‘when’, the temporal meaning could then easily
be regarded as secondary. Greenberg also sees traces of RU in the Hungarian
dative suffix -ra/-re. To account for the directive meaning, this suffix could
be analyzed as reflecting locatival *-rV augmented by some directional ele-
ment.
It was the purpose of this article to discuss the evidence from the Indo-
European languages in order to provide an alternative theory for the locative
suffix *-r. The reconstruction of a case ending rather than an “Adverbialendung”
for at least an early stage of Proto-Indo-European does, however, by no means
contradict its relation to similar morphemes in Uralic or other, possibly related
language families. The postulation of a shared case marker *-r(u) might be one
promising piece of evidence for the assumption of an Indo-Uralic macrofamily.
Even if the wider implications of the suffix *-r as a reflex of an Proto-Indo-Uralic
morpheme could not be treated here, the Indo-European evidence points to a
much more complex paradigm of spatial relations in Indo-European, similarly
to the one we find in the Uralic languages.
10 I hereby thank Rasmus Gudmundsen Bjørn who suggested this link after my lecture in
Leiden 2015.

via free access
28 bauhaus
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.

via free access
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.

via free access
chapter 3
Pronouns and Particles: Indo-Uralic Heritage

and Convergence*
Rasmus Gudmundsen Bjørn
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).

via free access
pronouns and particles: indo-uralic heritage and convergence 31
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).

via free access
32 bjørn
table 3.1 Correspondences between PIE *kwi- and

PU *ki
PIE *kwi-
Vedic cid ‘even, at last’

Avestan ciš ‘who’
Greek τίς ‘id.’
Albanian qish ‘id.’
Latin quis ‘id.’
OCS čĭto ‘what’
Hittite kuiš ‘who’
Toch. A kus ‘id.’
Old Irish cía ‘id.’
Armenian zinčc ‘what’
PU *ki (SKES 181)
Finnish ken (dial.) ‘who’

Saami gii ‘who’
Mordvin ki- ‘who, someone’
Mari ke, kö, kü ‘who’
Udmurt kin ‘id.’
Mansi kwät ‘which’
Hungarian ki ‘who’
Samoyedic *ki- ‘id.’
– Enets sio ‘id.’
– Ngan. sïlï ‘id.’
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.

via free access
table 3.2 Correspondences between PIE

*kwo- and PU *ku-(~*ko-)
PIE *kwo-
Vedic kás ‘who’

Avestan kō ‘id.’
Greek πóθεν ‘where’
Gothic hwas ‘who’
Latin quod ‘what’
Lithuanian kàs ‘who’
OCS kŭto ‘id.’
Albanian kush ‘id.’
PU *ku- (~ko-) (UEW 191)
Finnish kuka ‘who’

Saami guhte ‘who, which’
Mordvin kona ‘which’
Mari kuδe ‘who, which’
Komi kudiz ‘who, one’
Hungarian hol ‘where’
Khanty χō, χoje ‘who’
Samoyedic *ku- ‘which, what’
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

via free access
34 bjørn
table 3.3 Adverbial particles derived from PIE *k(w)u-
PIE *ku-
Language Form Meaning Extension
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])
This scenario thus explains the labiality as a secondary function of original

interrogative *ku- in front of a vowel, as opposed to original labiality in *kw lost
in front of *-u (cf. Rasmussen 1999: 319).
An interrogative particle *ku is also well-known in the Uralic languages, even
in the exact same function as in IE, e.g. Udmurt ku, Estonian kus, Mari ku-
‘where’. It is, however, more difficult to distill the original function beyond the
fact that the interrogative pronoun was fully developed by Proto-Uralic (cf. Joki
1973: 274 and table 3.2). In addition, there is another, syntactically different, use
of an interrogative particle *kV [back] in the Balto-Fennic languages, cf. Saami -gŏ

via free access
table 3.4 Reflexes of PIE *-kwe and its possible cog-

nates in Uralic
PIE *-kwe
Hittite =kku ‘and, if, or’

Vedic -ca ‘and’
Greek τε ‘id.’
Gothic (-u)-h ‘id.’
Latin -que ‘id.’
Gaulish (eti)-c ‘id.’
OCS (ta)-kъ ‘thus’
Uralic
Finnish -ki(n); -kaan ‘too’; ‘either’

Saami -ge/-gâ ‘either, too; and’
Mordvin -ak ‘also, even’
Mari -ke ‘whole, all’
Permic *-kö ‘if’
(Mansi ke ‘if’)
Kamass. -go/-gö ‘yet, also’
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.
1.3 The Enclitic Particle *-kwe

This section discusses whether the following enclitic particles can reasonably
be compared at the proto-levels. Form and function are given in table 3.4.
The particle is securely established for PIE with attestations in most
branches, but is in rapid decline and survives only in vestigial forms in the
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.

via free access
36 bjørn
modern languages. The usage as a simple conjunction is widespread, but likely

innovated in PIE from a more complex modal function, which further strength-
ens its formal connection with the interrogative stem (cf. Szemerényi 1990: 210).
The Uralic material (table 3.4) is scattered, but with significant geographic and
dialectal distribution. Finnish has the most specialized assemblage with two
emphatic particles (positive *-kin and negative *-kaan) and an interrogative
(*-ko, cf. §1.2). The Mansi form is likely a borrowing from Permic (Toivonen
1956: 24) and should not be counted as Ugric evidence per se. The Kamassian
reflex, e.g. īne-gö ‘one more horse’ (Künnap 1999: 29), on the other hand, hardly
resembles the function of central Permic (and Mansi), but neatly echoes the
Finno-Saami suffix, and is best explained as shared retention. Wagner (1967)
provides an in-depth treatment of the functional similarities between some
of these forms in both families, and concludes that their presence in Uralic is
due to borrowings from various branchings of IE. There are two primary rea-
sons why this conclusion should not stand unchallenged: (1) He presumes that
PIE and Proto-Uralic ultimately are incompatible (1967: 76),6 and (2) he fails to
discuss the Kamassian forms that provide crucial evidence for the particle in
Samoyedic (though he seems to be aware of their existence, cf. 1967: 67). In sum-
mary, the existence of a Uralic proto-form is suggested by at least three distinct
reflexes in non-contiguous dialects with a syntactically unique (i.e. enclitic)
similar-sounding morpheme.
1.4 Concluding Remarks on the Interrogative Complex

The entire complex of both Uralic and PIE might then be based on an archaic
particle *ku. The pronominal forms arose through combination with the
vocalic anaphoric pronoun (§3) with subsequent change of *-u to *-w-. Fur-
ther neutralization of the glide occurred in satem IE and most of Uralic, while
labialization was retained in IE centum and, strictly tentatively, in Selkup, cf.
kut ‘who’ with qaj ‘what’, from *ku-C and *ku-V- (cf. Bjørn 2017: 87), perhaps
with typological aid from neighboring Ket (e.g. Abondolo 1998: 10).
The Uralic oscillation between the different vocalisms similarly suggests
analogical treatment of a graded paradigm, either vertical (like PIE) or horizon-
tal (this vs. that), possibly still visible in Finnish, and common provenance of
the vowel gradations of both PIE and PU cannot be excluded. The widespread
use of the enclitic particle suggest that the trait is shared at a proto-level unac-
counted for by later borrowing. Widespread loss is attested in the known histo-
ries of IE languages and must be assumed for large swaths of the Uralic contin-
6 Cf. also criticism in Joki (1973: 275).

via free access
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-
Often mentioned alongside the interrogative and deictic pronouns as evidence

for Proto-Indo-Uralic (e.g. Rasmussen 2005: 527), the uniquely relative pronoun
*io- is demonstrably a later innovation, likely diffused to the western Uralic
languages after the breakup of both proto-languages. The almost amorphous
anaphoric pronoun, on the other hand, is old and, with circumstantial evi-
dence, comparable at the Indo-Uralic level.
2.1 Anaphoric Pronoun

Correspondences between PIE *i- and PU *i- (table 3.5) are treated in the
present section.
The function in PIE may be considered an anaphoric 3rd person pronoun.
Notwithstanding the suggested paradigmatic similarities with the interrogative
pronoun, a perfect mirror image of the situation in the interrogative system is
found in the locative forms constructed on the bare stem formant, cf. Latin i-
bi ‘there’ and Vedic i-há ‘here’ (recall [c]u-bi and ku-ha, respectively, in § 1.2).
A notable difference between the paradigms of *kwis and *is is the lack of u-
vocalism in the latter. With otherwise parallel formations, the most economical
solution is to accept the vowel as inherently endemic to the particle *ku in
pre-PIE; if the interrogative formant was *kw-, rounding of the following vowel

via free access
38 bjørn
table 3.5 Correspondences between PIE *i-

and PU *i-
PIE *i-
Vedic iyám (f.) ‘this’

Cyp. (Hes.) ἲν ‘id.’
Gothic is ‘he’
Latin is ‘he; who’
PU *i-?
Mordvin iśt’a ‘such’

Komi e-sy ‘this, each’
Khanty it ‘this’
Hungarian itt ‘here’
Kamassian iidə ‘that one’
Nenets eke, eko ‘id.’
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.
2.2 Relative Pronoun

The correspondences between PIE *i̯o- and Finnish joka (see table 3.6) in both
form and function must be seen in the perspective of their respective develop-
ments.
The relative pronoun in PIE is derived from the anaphoric pronoun *i/e- (cf.
Szemerényi 1990: 210f.), and necessarily represents an extension where the i-
formant becomes a semi-vowel and gets a new e-o ablauting vowel. The rela-
tive pronoun is not original in Uralic either (e.g. Janhunen 1982: 39). According
to both Paasonen (1906: 116f.) and Rédei (1988: 637), Mordvin ju-za ‘here and
there’ represents a transitional form that connects the Finno-Permic relative
with the Uralic general demonstrative *i (cf. also Paasonen l.c. and Collinder
1965: 119f. for the IE connection). In both families the specialized pronoun
decreases polysemy in the original interrogative pronoun that also carried the
relative function, cf. the double function of modern English who (cf. § 1.1). The

via free access
table 3.6 Correspondences between PIE *i̯o- and

Finn. joka and cognates
PIE *i ̯o- ‘who (rel.)’
Vedic yás ‘who (rel.)’

Avestan yō ‘id.’
Greek ὄς ‘id.’
Phrygian ιος ‘id.’
Lithuanian -is [def.adj.]7
OCS iže ‘who (rel.)’
Uralic
Finnish joka ‘who’

Saami juokke8 ‘each’
Mordvin juza ‘here and there’
Mari južo ‘each’
development of the relative pronoun may thus be a convergence phenomenon

that Fenno-Volgaic underwent in contact with some (P)IE stock (sometime and
somewhere between stages [b] and [c] in the contact illustration, figure 3.1),
similar to, but not necessarily at the same time as, the possible semantic conver-
gence discussed above for the particle *-kwe (§1.3). Note that the Uralic forms
are missing the nominative *-s formant that is ubiquitous in the IE stock; if the
pronoun was borrowed, the Fenno-Volgaic recipient language must have lost
or removed the ending.
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.

via free access
40 bjørn
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.).
3.1 Animates in *so

The formal and functional similarities between PIE *so and PU *so(n) (table
3.7) are obvious to the naked eye.
In PIE the form is confined to the casus rectus of the animate gender while
the oblique and inanimate parts of the paradigm are based on the *t-variant
(cf. §3.2). The evidence of the Hittite form is disputed (cf. Kloekhorst 2008b:
772), but pre-PIE status is certainly implied by the lack of the regular nomina-
tive marker. The situation is markedly different in the Uralic languages where
the s-form permeates the paradigm of the animate gender. Hajdú demonstrates
that the lone Samoyedic form, Selkup tëp ‘he, she, it’, sufficiently establishes
the *so pronoun as Proto-Uralic (1990: 1f., cf. also Abondolo 1998: 25; con-
tra Déscy 1990: 57); widespread loss in the remaining Samoyedic languages is
thus needed, but this is hardly inconceivable with the phonetic confluence of
the inherited forms; note that Uralic *s regularly yields t in Samoyedic, sug-
gesting that the Uralic sound was closer to an affricate *ts (cf. discussion in
§ 3.3). The paucity of material in Samoyedic is probably, like the case with the
enclitic *-k(w)e, a consequence of late attestation and prehistoric conflation of
paradigms.

via free access
table 3.7 PIE *so and PU *so(n)
PIE *so ‘he, she’9
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)
Finnish hän ‘he, she’

Saami son ‘id.’
Mordvin son ‘id.’
Mari -žo, -žö ‘his, her’
Permic *so ‘each; he, she’
Hungarian ö, ön ‘he, she’
Mansi taw ‘id.’
Khanti tou, lou ‘id.’
Selkup tëp ‘he, she, it’
3.2 Inanimates in *t-

The corresponding inanimate pronoun, PIE *tod and PU *to/tu (table 3.8), is
more resilient and has survived in most languages.
Outside the animate casus rectus, the t-form is prolific in the PIE paradigm.
Already by Proto-Uralic a clear alternation between front and back vowel dis-
tinguishes proximity to the speaker, e.g. Mari ti ‘this’ ~ tu ‘that’ (table 3.9).
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.).

via free access
42 bjørn
table 3.8 PIE *tod and PU *to/tu
PIE *tod10
Vedic tát ‘this, that’

Avestan tat̰ ‘this’
Greek τό ‘this, that’
Gothic þat-a ‘that’
Latin (is-)tud ‘this here’
Armenian da ‘these’
Tocharian A täm ‘this’
OCS to ‘that’
PU *to/tu
Finnish tuo ‘that’

Saami duot ‘id.’
Mordvin tona ‘id.’
Mari tu ‘id.’
Komi tu ‘each’
Hungarian tova ‘away’
Mansi ton ‘that’
Khanti tŏmi ‘id.’
Samoyedic *to ‘this’
3.3 The Animacy Distinction

The salient IE feature of the s- animate casus rectus is thus echoed in a Uralic
opposition between the (animate) 3rd person pronoun and the (inanimate)
demonstrative.
Whatever the motivation behind the animate s and inanimate t alternation
in this pronoun, it is not obvious in any of the stocks, suggesting a shared irregu-
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).

via free access
table 3.9 Reflexes of PU *te/*ti
PU *te/*ti
Finnish tä-mä ‘this (here)’

Saami dāt ‘id.’
Mordvin tε, te ‘id.’
Mari ti ‘id.’
Permic *ta ‘id.’
Mansi ti ‘id.’
Khanti ta, te(-mi) ‘id.’
Samoyedic *tä- ‘this here’
larity (cf. figure 3.2).11 Although T as a deictic pronominal sound is widespread

in the Eurasian language families (cf. Greenberg 2000: 94), so that the identity
of the initial consonants alone is non-significant in terms of linguistic classifi-
cation, the systemic opposition that the animate s/inanimate t represents is by
no means trivial.
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.

via free access
44 bjørn
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
The beyond chance correspondences found in the interrogative complex (§ 1)

and the animacy distinction (§3) provide the theoretical basis of any further
inquiries into Indo-Uralic relations. The material presented in this article sug-
gests that parts of a Proto Indo-Uralic pronominal system can be reconstructed.
In addition to the personal pronouns, these include:
– *tso (animate 3rd person pronoun)
– *to (inanimate 3rd person pronoun)
– *i (anaphoric pronoun)
(with later development of a “full-grade” relative *io-)
– *ku (interrogative particle)
(before the split of PIU to combine with the anaphoric *i to produce *kwi)
(before the split of PIU to develop the particle *-kwe)
Other features of the Uralic pronouns that lack convincing counterparts in PIE
must then be treated accordingly, cf. the demonstrative plural stem *n- (e.g.
Fin. ne ‘they’) and the inanimate interrogative stem *m- (e.g. Fin. mikä ‘what’),

via free access
either as innovation in PU or loss in PIE. The systemic approach applied

here illuminates the unlikelihood of having a complete pronominal system
transplanted from one language to another. Although limited in number, the
phonological matches found in the pronominal category should be expected
to occur in other shared items. The following sound correspondences can be
deduced:
PIE PU < PIU?
*kw *k12 *ku /_V

*ku *ku *ku /_C
*s *s *s
*s / *t *ts *ts
*t *t *t
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
12 Possibly Selkup ku < *_C, but q < /_V.

via free access
46 bjørn
(2001: 258). It is in many ways of lesser consequence to propose borrowing

rather than common heritage, but:
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
Ultimately ancient borrowings and shared heritage can be impossible to dis-

tinguish, but even the most ardent criticism would have to contend itself with
the fact that the influence then eclipses the most basic linguistic material,13
practically stripping the recipient language of its skeleton.
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).

via free access
Cavoto, F. 2003. Supplétion et récurrence de thèmes pronominaux nostratiques. Dia-

chronica 20/2, 229–258.
Cavoto, F. 2004. Nostratic, Eurasiatic and Indo-European derivation. In: J. Clackson;
B.A. Olsen (eds.), Indo-European Word Formation: Proceedings of the Conference Held
at the University of Copenhagen, October 20th–22nd 2000. Copenhagen: Museum Tus-
culanum, 11–23.
Collinder, B. 1934. Indo-uralisches Sprachgut: die Urverwandtschaft zwischen der indo-
europäischen und der uralischen ( finnischugrisch-samojedischen) Sprachfamilie.
Uppsala: A.–B. Lundequistska.
Collinder, B. 1940. Jukagirisch und Uralisch. Uppsala: A.–B. Lundequistska.
Collinder, B. 1955. Fenno-Ugric Vocabulary. Stockholm: Almqvist & Wiksells.
Collinder, B. 1965. Hat das Uralische Verwandte? Ein sprachvergleichende Untersu-
chung. Acta Universitatis Upsaliensis 1/4: 108–180.
Čop, B. 1979. Indogermanisch-Anatolisch und Uralisch. 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, 9–24.
Déscy, G. 1990. The Uralic Protolanguage: A Comprehensive Reconstruction. Blooming-
ton: Eurolinga.
Dolgopolsky, A.B. 1984. On personal pronouns in the Nostratic languages. In:
O. Gschwantler; K. Rédei; H. Reichert (eds.), Linguistica et philologica. Gedenkschrift
für Björn Collinder. Wien: Wilhelm Braumüller, 65–112.
Heidelberg: Winter.
Fortson, B.W., IV 2010. Indo-European Language and Culture. Chichester: Wiley-Black-
well.
Greenberg, J.H. 2000. Indo-European and Its Closest Relatives, Vol. 1: Grammar. Stanford:
Stanford University Press.
Hajdú, P. 1990. Einiges über Fürwörter. Linguistica Uralica 1, 1–12.
Häkkinen, K. 2004. Nykysuomen etymologinen sanakirja. Helsinki: WSOY.
Heller, L. 1956. The I.E. so-/to- demonstrative suppletion or phonetic differentiation?
Word 12/1, 7–8.
Hyllested, A. 2014. Word Exchange at the Gates of Europe. Ph.D. dissertation, University
of Copenhagen.
Janhunen, J. 1977. Samoyedischer Wortschatz. Helsinki: Castrenianumin toimitteita.
Janhunen, J. 1982. On the structure of Proto-Uralic. Finnisch-Ugrische Forschungen 44,
23–42.

via free access
48 bjørn
Janhunen, J. 1998. Samoyedic. In: D. Abondolo (ed.), The Uralic Languages. London–
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

via free access
(eds.), Nostratic: Sifting the Evidence. Amsterdam–Philadelphia: John Benjamins,

153–190.
Sammallahti, P. 1998. The Saami Languages. Kárásjokka: Davvi Girji.
Schumacher, S. 2011. Middle Welsh. In: E. Ternes (ed.), Brythonic Celtic. Bremen:
Hempen, 85–236.
Sihler, A.L. 1995. A New Comparative Grammar of Greek and Latin. Oxford: Oxford Uni-
versity Press.
SKES = Toivonen et al. 1974–1981.
Szemerényi, O.J. 1990. Introduction to Indo-European Linguistics. Oxford: Oxford Uni-
versity Press.
Thomason, S.G.; C.L. Everett 2001. Pronoun borrowing. Berkeley Linguistics Society 27,
301–316.
Toivonen, Y.H. 1956. Über die syrjänischen Lehnwörter im Ostjakischen. Fenno-Ugri-
sche Forschungen 32, 1–126
Toivonen, Y.H.; E. Itkonen; A.J. Joki; R. Peltola; S. Tanner; M. Cronstedt 1974–1981.
Suomen kielen etymologinen sanakirja. Helsinki: Finno-Ugrian Society.
UEW = Rédei 1988.
Wagner, H. 1967. Indo-germanisch -kwe im Finnisch-Ugrischen? Münchener Studien zur
Sprachwissenschaft 20, 67–92.
Weiss, M. 2009. Outline of the Historical and Comparative Grammar of Latin. Ann
Arbor–New York: Beech Stave Press.

via free access
chapter 4
Indo-Anatolian Syntax?
Dag Haug and Andrei Sideltsev*
1 Hittite vs. Other Indo-European Languages
A well-known feature of Hittite is the use of wh-words en lieu of indefinite pro-

nouns in conditional clauses (1)1 and, considerably more seldom, after negation
marker (2):2
(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.

via free access
indo-anatolian syntax? 51
(3) a. Greek (John 11.9)

ἐάν τις περιπατῇ ἐν τῇ ἡμέρᾳ
‘If anyone walks in the daytime, …’.
b. Avestan (Yasna 50.1)

kat̰ mōi uruuā isē cahiiā auuaŋhō
question my soul.nom command.3SG.PRS wh.gen help.gen
‘Does my soul command any help?’
c. Latin
si quis
‘if anyone’
d. Gothic (John 7.4)

ni manna in analaugnein ƕa taujii
no man.nom in secret.dat wh.ACC do.3sg.prs
‘Nobody does anything in secret’.
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

via free access
52 haug and sideltsev
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
3 The signs are damaged.

via free access
(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:
(9) OH/MS (CTH 374.2.A) KUB 36.75 obv. ii 13–14

1. ūk=za neku DINGIR=YA tuk kuit iy[(a-nu)]n
I=REFL ever god=my you.DAT what do-1SG.PST
2. nu kuit waštā-ḫḫun
CONN what-ACC.SGN sin-1SG.PST
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.

via free access
2 Other Anatolian Languages
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:
(10) KARATEPE 1 Hu. §LIX 331–333

REX-ta-ti-i-pa-wa/i kwa/i+ra/i kwa/i-sa-há ¦¦
hantawatta-tti=ba=wa kwari kwisha
rule.as.king-3SG.PRS=but=PRTC if someone.NOM.SG.C
‘But if anyone rules as a king’.4
The usage of NPIs after ‘or’ is more frequent (5×):
(11) MALPINAR §20, cf. Hawkins 2000: 342.

ni-pa-wa/i-tà-’ POST+ra/i-i-sá kwa/i-sà-ha-’
niba=wa=ada appara-s kwisha
or=PRTC=them lesser-NOM.SG.C someone.NOM.SG.C
CAPUT-ti-sa ARHA MALLEUS
x-s ahha wallai-
(noble)man-NOM.SG.C away smash
‘Or (if) any inferior man shall erase them, …’.
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).

via free access
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
(13) KULULU 1 6 §14, cf. Hawkins 2000: 443.

wa/i-tà ¦¦ ¦ni-i ¦kwa/i-ti-i-ha ¦pi-i[a]-a-i
wa=ada ni kwadiha piya-i
PRTC=it NEG someone.DAT.SG give-3SG.PRS
‘And let him not give it to anyone (else)’.
This use actually dominates for Hieroglyphic Luwian. Out of 30 attestations of

NPIs in Yakubovich’s corpus it is attested in half of the cases (14×).
The same usage is attested for Lycian. Solely NPI s are licensed by negation
(14) and conditionals, both with overt subordinator ‘if’ (15) and in paratactic
conditionals (16–17):
(14) N 320, 34–35

se=we=ne: χtta-di: tike:
and=PRTC=NEG harm-3SG.PRS anyone
‘And no one shall do harm’.
(15) TL 89, 2–3

a-di=me=j5=ẽ: tike: χttbã:
do-3SG.PRS=CONN=him/her=if6 someone.NOM.SG harm.ACC.SG
tisñke:
any.ACC.SG
‘If anyone does any harm to him/her’
(16) TL 57, 7–9

1. [s]e=[ije] n[e hrppi tã]ti tike kbi:
and=him NEG on put.3PL.PRS someone.ACC.SG other
2. hrppi=(i)je=me=i: ta-di: tike:
on=thereupon=CONN=him put-3SG.PRS someone.ACC.SG
3. me=n=e: tubei-ti mãhãi huwedri: se
CONN=him=they strike-3PL.PRS god.NOM.PL all.NOM.PL.C and
5 Or =me( j)=.
6 Following Garrett 1990: 241 fn. 1; Melchert 2004: 19. Cf. Kloekhorst 2013: 149.

via free access
itlehi: trm̃ mili:

belonging.to.a.league.NOM.PL.C Lycian.NOM.PL.C
‘(1) And they will not put someone on top of him. (2) (If) one puts
someone on top of him, (3) they will strike him, all the gods and the
Lycian league’.
(17) TL 83, 10–11

ñte=me=j7=epi: ta-di: ti[k]e [tik]e:
in=CONN=him=upon put-3SG.PRS someone someone
‘(If) someone inters someone on top of him’.
Part of the clauses can in principle be understood as generalizing, as (17) ‘who-

ever puts …’. However, others, illustrated here by (16), cannot, because in this
case the resumptive pronoun would have the wrong reference. So, the tradi-
tional understanding of the clauses as paratactic conditionals, also attested in
Hittite, seems to be descriptively right.
Data from other Anatolian languages are even more ambiguous and frag-
mentary, see Sideltsev & Yakubovich (2016), and contribute no positive infor-
mation to the issue.
Thus the data that come from other Anatolian languages support the Old
Hittite usage vs. the Middle/New Hittite one and set the Anatolian languages
apart from other Indo-European languages.
3 Anatolian vs. Narrow IE Bare Interrogatives
In this light the Hittite post-OH/OS bare interrogatives in conditional clauses

and after negation marker cannot be directly equated with the seemingly iden-
tical usage in narrow IE languages and have to be an independent innovation
within historical Hittite. Narrow IE languages would then attest bare inter-
rogatives alongside NPIs in certain contexts, see Haug (2016), whereas Proto-
Anatolian reflected in OH/OS texts and other Anatolian languages attests only
NPIs in these contexts. The difference might constitute syntactic evidence for
the Indo-Anatolian hypothesis. Bare interrogatives would then have to be inde-
pendent innovations in Middle Hittite and in narrow IE.
Now, if, despite the evidence provided so far, one still traces the Hittite use
of bare interrogatives to the Proto-Indo-European one, writing off the lack of
7 Or =me( j)=.

via free access
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.
3.1 Hittite Contexts Other Than Conditionals

The fact that Middle Hittite and narrow IE bare interrogatives are independent
innovations is all the more likely if one views the modest broadening of the
licencing of bare interrogatives in Hittite. The direction of the spread is rad-
ically different from narrow PIE and closely tied in to the peculiarities of the
Hittite syntax. The first of them is that in Hittite bare interrogatives are licenced
by mān not only in the conditional meaning ‘if’, but also in the temporal one
‘when’:

via free access
(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, …’.
Actually, it is very clear why the distribution is exactly as attested. As is well

known, in Old Hittite texts conditional clauses were introduced by the sub-
ordinator takku ‘if’ whereas temporal clauses were introduced, among other
subordinators, by mān ‘when’. The state of affairs in Middle and New Hittite was
different: the conditional subordinator was mān ‘if’ whereas the main tempo-
ral subordinator was māḫḫan ‘when’. The transitional period occurred during
late OH time: “Temporal clauses in Old Hittite (OH) are frequently marked by
mān ‘when’, which in OH was only beginning to gain the secondary meaning
‘if’ that it acquired in NH” (Hoffner & Melchert 2008: 416). Now, at the time
when mān was taking over the conditional function of takku ‘if’, it was used
simultaneously, among other functions, for which see, e.g., (Hoffner & Melchert
2008: 416), both in temporal and conditional clauses. Thus it was highly likely
to “borrow” the syntax characteristic of takku ‘if’, more commonly in its new
conditional function, and, less commonly, in its temporal one. Actually, the
extention seen in (19–20) is not very surprising from the narrow Indo-European
point of view: both of the temporal clauses are generalizing, and generalizing
temporal clause licence bare interrogatives, e.g., in Latin. The same generaliz-

via free access
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.

via free access
(23) NH/NS (CTH 106.A.1) Bo 86/299 obv. ii 53.

maḫḫan=ma=za ABU=YA kuwapi DINGIRLIM-iš
when=but=REFL father=my when god-NOM.SG.C
kiš-at
become-3SG.PST
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

via free access
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)’.
Thus even in case of secondary spread independent innovations in historical

Hittite and in narrow PIE are more likely than common heritage. Even the mod-
est innovation of Hittite is different from the innovation of narrow PIE and is

via free access
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).
3.2 Bare Interrogatives outside of Conditionals?

It was observed above that our corpus does not attest bare interrogatives out-
side of conditionals / relative clauses / negative contexts. The contexts that
have been analyzed as such should rather be assessed otherwise:
(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’.

via free access
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:
(30) MH/MS (CTH 262) IBoT 1.36 obv. i 57

nu=war=an kuwapi au-tti
CONN=QUOT=him when see-2SG.PRS
‘Will you ever notice him?’.
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.

via free access
4 Bare Interrogatives in the Luwian Branch?
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.
4.1 Lycian tihe

The first deviating form is Lycian tihe, formally a genitive of the wh-word ti-
(Melchert 2004: 66). In the majority of contexts it is employed with indefinite
tike and thus is interpreted as a bare interrogative licenced in a conditional
clause (ibid):
(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’.
The translation of tihe follows a suggestion of I. Yakubovich, pers.comm. The

context and the structure are stereotypical and are virtually verbatim repeated
in TL 91 2–3, TL 95 2, TL 135 2. At face value the Lycian example is identical
to the following Hittite one and seems to finally provide a parallel from other
Anatolian languages for the Middle/New Hittite bare interrogatives:
(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.:

via free access
(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:
(34) KULULU 1 §§7–9, cf. Hawkins 2000: 443.

¦wa/i-ti ¦za-ia ¦DOMUS-na-’ ¦kwa/i-sá
wa=di zaya parna kwis
PRTC=REFL this.ACC.PL.N house.ACC.PL.N which.NOM.SG.C

via free access
¦ni-pa-wa/i ¦á-ma-ta-’ ¦tu-wa/i-ti-ia ¦wa/i-zi-ti-⟨i⟩

niba=wa amatta Tuwattiya wazzidi
or=PRTC estate.component.ACC.PL T.DAT.SG wish.3PRS.SG
¦ni-pa-wa/i ¦la-hi?-zi-i ¦ni-pa-wa/i ¦wa/i-ia-ni-[si?-]i
niba=wa lahinzi niba=wa wiyanissi
or=PRTC estate.component.ACC.PL or=PRTC vine.GEN
¦tu-wa/i[+ra/i]-sà-za-’ ¦kwa/i-sà-ha-wa/i-sa
tuwars-anza kwis-ha=wa=as
vineyard.ACC.SG someone.NOM.SG.C=PRTC=he
¦kwa/i-sa-⟨⟨pa⟩⟩9
kwis
who-NOM.SG.C
‘Whoever shall demand these houses from Tuwati, or the amatta-s, or the
lahi-s, or the vineyard of vine(s), whosoever he (be), …’
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:
(35) N 314b 1–3.

χupa ebe-hi ti( j)=j=a-di: tike zum̃ mẽ: ti-he
tomb this-GEN.ADJ who=it=do-3SG.PRS some harm who-GEN.SG
9 The particle -ba ‘but’ must be a scribal mistake here as it does not fit the context.

via free access
‘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
10 Following Melchert 2004: 11. Cf. Kloekhorst 2011.

via free access
constituent like (31). Thus tihe is not a synchronic or diachronic parallel to Mid-
dle/New Hittite bare interrogatives.
4.2 Luwian Bare Interrogatives in a Relative Clause?

Another form that can be claimed to be a Luwian bare interrogative occurs in
a relative clause
(36) BABYLON 1§4 … §8, cf. Hawkins 2000: 392.

1. | á-ma-za-pa-wa/i-’ kwa/i-a-za
amanza kwanza
my.NOM.SG.N which.NOM.SG.N
2. kwa/i-a-za kwa/i-i-ta PES-i
kwanza kwitta x-i
which.NOM.SG.N where be.located-3SG.PRS
‘(1) All that (is) mine, […] (2) (that) which is11 (any)where, (before him,
Halabean Tarhunzas, I gave (it) over)’.
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:
(37) ASSUR letter f-g §43–44, cf. Hawkins 2000: 537.

1. a-wa/i-i ¦LITUUS+na-ti-sa kwa/i-ta-ha ¦10 (“*78*”)a-ru-ti-na
a=wa mana-tis kwitta=ha 10 arudi-n
CONN=PART see-2SG.PRS where=and ten basket-ACC.SG
11 Following the understanding of Yakubovich, web-corpora.net/LuwianCorpus/search. Cf.

Hawkins 2000: 392.

via free access
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’.
The context warrants analysis of kwitta=ha as a free choice indefinite in a rel-

ative clause. It is significant that kwitta=ha is the only relative pronoun in the
clause and it is not licenced by another relative pronoun. Relative pronouns
in indeterminate relative clauses are virtually identical. It means that the dif-
ference between free choice indefinites and common relatives in indetermi-
nate relative clauses is minimal in Luwian in semantic and syntactic terms.
An important consequence is that Luwian, as different from Hittite, attests the
system where both indefinite free choice and relative pronouns are employed
in relative clauses, both as the only pronoun and when there is another pro-
noun. In the latter case they produce the impression of being licenced by the
first relative pronoun, but from the Luwian perspective this is not so. Another
important factor is that Luwian makes use of indefinite + relative (kwis-ha kwis)
as free choice items, see (34) above. This could have also affected the choice of
relative kwitta and not indefinite kwitta=ha in (36). In this context the use of the
relative pronoun in (36) occurs under very different circumstances in Luwian
than in Hittite and cannot be equated directly with bare interrogatives in rela-
tive clauses in Hittite which are illustrated by (25–27) above. I believe it is not
incidental, either, that the only Luwian form eligible for the bare interrogative
interpretation does not occur in a conditional or negative clause, predomi-
nantly licencing bare interrogatives in Hittite, but rather occurs in a relative
clause, which licences bare interrogatives only sporadically in Middle and New
Hittite.
4.3 Bare Interrogatives as Free Choice Items

So the forms which can be claimed to be bare interrogatives in the Luwian
branch should rather be assessed as free choice items.
4.4 Other Hieroglyphic Luwian Data

Several other Hieroglyphic Luwian forms have been compared to the post-
OH/OS Hittite usage of bare interrogatives. The first is kwiya in the following
context:

via free access
(38) KARKAMIŠ A6 §8–4, cf. Hawkins 2000: 124.

za-a-pa-wa/i (“MENSA.SOLIUM”)á-sa-na-’ ¦á-mi-i-’
za-n=ba=wa asa-n amiya
this-ACC.SG.C=PRTC seat-ACC.SG.C my.DAT.SG
(DOMINUS)na-ni ¦INFANS.NÍ ¦kwa/i-i-ia AEDIFICARE+MI-ha
nanni nimuwi kwiya tama-ha
master.DAT.SG son.DAT.SG when build.1SG.PST
‘And when I built this seat for Kamanis my lord’s child, …’.
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
(39) ASSUR letter e §7.

(We (are) to write no letter, (it is) you yourselves (who) are to write!)
a-wa/i ¦á-pi ¦u-zi-na ¦kwa/i-i ¦ha-tu+ra/i-na
a=wa appi unzi-n kwiya hattura-n
CONN=PART back your-ACC.SG.C when writing.ACC.SG
¦AUDIRE+MI-ta-ra+a-nu
tummanta-ranu
listen-2PL.IMP
‘When you hear your letter back’
The analysis which follows the understanding of Yakubovich in his corpus is

obviously preferable to unintelligeable “Hear your sort of letter back!” (Hawkins
2000: 535). Alternatively, we may understand the context following Waal (2016)
who proposed to translate the clause as “May you therefore(?) observe your
health/ hear your blessing”. It is important that the understanding does not
contain an indefinite reading of kwiya either. The same holds for ASSUR let-
ter e §10 which is even more obscure, cf. Hawkins 2000: 535 for an attempt at
a translation. An alternative analysis by Melchert (1988: 39–40) also dispenses
with an indefinite reading, interpreting kwiya as ‘nearly’, lit. ‘as it were/like’.

via free access
5 Conclusion
Thus post-OH/OS bare interrogatives in conditional clauses and after negation

marker in Hittite cannot be directly equated with the seemingly identical usage
in narrow IE languages and has to be an independent innovation within his-
torical Hittite and narrow IE. Narrow IE languages licence bare interrogatives
in conditional clauses, questions and after negation, but Proto-Anatolian, as
preserved by Old Hittite and the languages of the Luwian branch, attests only
indefinite pronouns—NPIs and free choice items—in these contexts.
References
Adiego, I. 2015. Lycian nasalized preterites revisited. Indogermanische Forschungen

120/1, 1–30.
Beckman, G. 1996. Hittite Diplomatic Texts. Atlanta: Society of Biblical Literature.
Brosch, C. 2014. Untersuchungen zur hethitischen Raumgrammatik. Berlin–Boston: De
Gruyter.
De Roos, J. 2007. Hittite Votive Texts. Leiden: Nederlands Instituut voor het Nabije
Oosten.
Garrett, A. 1990. The Syntax of Anatolian Pronominal Clitics. Ph.D. dissertation, Harvard
University.
Garrett, A. 1992. Topics in Lycian syntax, Historische Sprachforschung 105, 200–212.
Garrett, A. 1994. Relative clause syntax in Lycian and Hittite. Die Sprache 36, 29–69.
Güterbock, H.G.; T. van den Hout 1991. The Hittite Instruction for the Royal Bodyguard.
Chicago: Oriental Institute.
Haspelmath, M. 1997. Indefinite Pronouns. Oxford: Oxford University Press.
Haug, D. 2016. PIE *kwi-/kwo: Interrogative, indefinite or both? In: A.M. Byrd; J. DeLisi;
M. Wenthe (eds.), Tavet Tat Saryam. Studies in Honor of Jared S. Klein on the Occasion
of His Seventieth Birthday. Ann Arbor–New York: Beech Stave Press, 86–100.
Hawkins, J.D. 2000. Corpus of Hieroglyphic Luwian Inscriptions. Berlin–New York: Wal-
ter de Gruyter.
Hoffmann, I. 1984. Der Erlass Telipinus. Heidelberg: Winter.
Hoffner, H.A., Jr. 1995. About questions. In: T. van den Hout; J. de Roos (eds.), Studio His-
toriae Ardens. Ancient Near Eastern Studies Presented to Philo H.J. Houwink ten Cate
on the Occasion of his 65th Birthday. Istanbul: Nederlands Historisch-Archaeologisch
Instituut, 87–104.
Hoffner, H.A., Jr. 2003. The case against Ura-Tarhunta and his father Ukkura. In:
W.W. Hallo; K.L. Younger Jr. (eds.), The Context of Scripture. Volume 3. Leiden–
Boston: Brill, 57–60.

via free access
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.

via free access
Van den Hout, T. 1995. Der Ulmitešub-Vertrag. Wiesbaden: Harrassowitz.

Van den Hout, T. 2003. Apology of Hattusili III. In: W.W. Hallo; K.L. Younger Jr. (eds.),
The Context of Scripture. Volume 1. Leiden–Boston: Brill, 199–204.
Waal, W. 2016. Hi, how are you? Me, I am fine. A new interpretation of Luwian hatura-.
Talk presented at the 3rd Workshop on ‘Luwic’ Dialects: Inheritance and Diffusion,
Barcelona, 9 March 2016.
Werner R. 1967. Hethitische Gerichtsprotokolle. Wiesbaden: Harrassowitz.
Yakubovich, I. 2012. The reading of Luwian ARHA and related problems. Altorientali-
sche Forschungen 39/2, 321–339.
Yakubovich, I. 2015. Phoenician and Luwian in early Iron Age Cilicia. Anatolian Studies
65, 35–53.

via free access
chapter 5
Daniel Europaeus and Indo-Uralic

Petri Kallio
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):
Vilhelm Thomsen, as early as 1869, indicated the possibility of a relation-

ship with Finno-Ugrian, but he did not pursue the subject very far.
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).

via free access
daniel europaeus and indo-uralic 75
Uralic hypothesis rather than the idea of a genetic relationship between the
Indo-European and Uralic language families.
2 Before Indo-Uralic
As is well-known, language comparison as a hobby is much older than com-

parative linguistics as a science. Thus, the Indo-European and Uralic languages
were compared with each other long before the respective language families
themselves were established. For instance, already the 18th century Finnish
priest Nils Idman (1774) argued that Finnish and Greek are genetically related,
but this fact does not make him an Indo-Uralicist, even though he was among
the first to point out their similar personal pronouns which have since become
the cornerstones of the Indo-Uralic hypothesis. In any case, we cannot speak
of Indo-Uralic before we first have at least some preliminary version of both
Indo-European and Uralic.
Now the Indo-European language family goes back as far as the 17th century
Dutch scholar Marcus Zuerius van Boxhorn (1647) whose so-called Scythian
language family was already almost the same as the Indo-European language
family was before the early 20th century decipherments of Tocharian and Hit-
tite. The Uralic language family in turn started as Finno-Hungarian substanti-
ated by the 18th century Hungarian pioneers János Sajnovics (1770) and Sámuel
Gyarmathi (1799), but most other Uralic languages remained poorly docu-
mented until the expeditions of the 19th century Finnish field linguists Anders
Johan Sjögren (1861) and Matthias Alexander Castrén (1849–1862) throughout
the Russian North. In the end, it took until the early 20th century before the
position of Samoyed among the other Uralic subgroups was conclusively clar-
ified (see e.g. Setälä 1915).
As the early 19th century scholarly community already knew Core Indo-
European (i.e. Indo-European without Tocharian and Anatolian) and Finno-
Hungarian (a.k.a. Finno-Ugric), the idea of their relationship could already
have occurred to someone, but this was not the case. Indeed, even the foremost
pioneers in Indo-European linguistics compared Indo-European with almost
anything other than Uralic. For instance, the Englishman Sir William Jones
(1799) connected Indo-European with Austronesian and Tibetan, whereas the
German Franz Bopp (1841, 1847) with Austronesian and Caucasian. Meanwhile,
the Dane Rasmus Rask (1818, 1834) was among the few to compare Indo-
European with Uralic, but even he preferred to explain their similarities in
terms of language contact (cf. much more recently Rédei 1986; Koivulehto
1994). Instead, he came to be one of the forerunners of the Ural-Altaic hypoth-
esis later attributed to Castrén mentioned above.

via free access
76 kallio
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
David Emmanuel Daniel Europaeus was born on 1 December 1820 in Savi-

taipale, the Grand Duchy of Finland in the Russian Empire. His father Peter
Adolf Europaeus was a vicar whose family was probably originally from Äyrä-
pää, thus Latinized as Europaeus (Pulli 1984; Jurvanen 1988). Peter was one
of the most distinguished clerics in Old Finland (i.e. the southeastern part of
Finland that Sweden had already ceded to Russia in the peace treaties in 1721
and 1743), as best evidenced by the fact that in 1803 he was visited by Tsar
Alexander I himself (Väänänen 2003). Yet in 1806 the widowed Peter caused
quite a scandal by marrying Sofia Peijo, a milkmaid 30 years his junior. In spite
of the fact that their lastborn Daniel was of course totally innocent, his most
class-conscious contemporaries never really got over such a half-peasant back-
ground.
Moreover, although all the rest of Finland had been ceded by Sweden to Rus-
sia in 1809, the intelligentsia of the new Grand Duchy of Finland long remained
almost entirely Swedish speaking. Thus, Europaeus as a native Finnish speaker
was quite an oddity at the University of Helsinki where he began his studies
in 1844 but where he never graduated. Instead, he became a Jack-of-all-trades

via free access
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

via free access
78 kallio
4 Indo-Europaeus
When exactly Europaeus ended up with the Indo-Uralic hypothesis is hard to

say, because he did not breathe a word of it in his pre-1853 personal letters
to Castrén and others (or at least in those published in Niemi 1903; Salmi-
nen 1905; Timonen 1988), although in the same letters he openly advertised his
ideas of Uralic substrate toponyms which he did not publish until many years
later (Europaeus 1868–1870).4 In any case, the year 1853 saw his self-published
pamphlet Komparativ framställning af de finsk-ungerska språkens räkneord, till
bevis för Ungrarnes stamförvandtskap med Finnarne, och den indogermaniska
folkstammens urförvandtskap med den finskungerska, whose title tells it all: he
argued for the genetic relationships between Hungarian and Finnish5 as well as
between Indo-European and Finno-Ugric but, bizarrely enough, based on their
numerals.
Indeed, even though in passing Europaeus also mentioned pronouns, kin-
ship and body-part terms as well as grammatical morphemes, he instead re-
garded the numerals for ‘7’, ‘10’, ‘100’, and ‘1000’ as his strongest evidence for
Indo-Uralic. As most of them have since been considered Indo-European loan-
words in Uralic (cf. Honti 1993; Blažek 1999: 89–101), his etymological sugges-
tions were not entirely wrong, but yet his evidence for Indo-Uralic was drasti-
cally insufficient, to say the least. Even so, he became the first to suggest that
Indo-European and Finno-Ugric are more closely related to each other than to
any other language or language family, something that met with fierce opposi-
tion and even scorn, albeit privately rather than publicly (Salminen 1905: 118).6
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).

via free access
As a result, he spent years writing an extensive manuscript which he submit-

ted to the University of Helsinki in 1861 in order to finally graduate. Regrettably,
this magnum opus of Europaeus has not survived, but its central theses were as
follows (after Salminen 1905: 127–128):
1. Finno-Ugric and Indo-European are much more closely related than
thought.
2. Turko-Tatar is less closely related to Finno-Ugric than thought.
3. Samoyed is more closely related to Finno-Ugric than Turko-Tatar.7
4. Mongol-Tungus-Manchu is even less closely related to Finno-Ugric than
Turko-Tatar.
5. All the language families above go back to a widespread proto-language.
6. This macrofamily also includes Egypto-Semitic, Papuan-Northeast-Asian,
and some North American dialects judging from their numerals.8
7. Tamil and few other dialects have no analogy to the languages above, even
though one must leave open the possibility if language is as old as the
human race.9
Yet Europaeus still failed to graduate, because the committee blasted his ety-
mologies as Wortklauberei, to which he replied by criticizing the committee as
ossified and stuck in obsolete theories (Salminen 1905: 128–129). In fact, there
was more than a shred of truth in his reply because he was indeed the first
Finnish scholar to get acquainted with the works of the great German Indo-
Europeanist August Schleicher and the concepts of Lautgesetz and Stamm-
baum. Thus, Europaeus was theoretically well ahead of his contemporary com-
patriots. It is therefore no wonder that although he was constantly an outcast
in Finland, he had numerous foreign penfriends, such as the leading 19th cen-
tury Hungarian Uralicists Pál Hunfalvy, József Budenz, and József Szinnyei. Yet
even more importantly Europaeus first wrote about his Indo-Uralic hypothe-
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).

via free access
80 kallio
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.

via free access
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-

via free access
82 kallio
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

via free access
Europaeus to Thomsen’s student Pedersen and the Nostratic hypothesis, and

even today the name Europaeus can still occasionally pop up in Nostratic cir-
cles (see e.g. Dolgopolsky 2012: 3199). Yet who he was and what he did has long
been forgotten at least outside Finland, hence a good reason to bring his name
up now at the eve of his bicentennial birthday.
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).

via free access
84 kallio
Dolgopolsky, A. 20123. Nostratic Dictionary. Published online at http://www.dspace.cam

.ac.uk/handle/1810/244080.
Europaeus, D.E.D. 1847. Pieni runon-seppä eli Kokous paraimmista Inkerinmaan puolelta
kerätyistä runo-lauluista ynnä Johdatuksia Runon tekoon. Helsinki: J. Simeliuksen
perilliset.
Europaeus, D.E.D. 1852–1853. Svenskt-finskt handlexikon. Helsinki: Finska Litteratursäll-
skapet.
Europaeus, D.E.D. 1853. Komparativ framställning af de finsk-ungerska språkens räkne-
ord, till bevis för Ungrarnes stamförvandtskap med Finnarne, och den indogermaniska
folkstammens urförvandtskap med den finsk-ungerska. Helsinki: J.C. Frenckell.
Europaeus, D.E.D. 1854. Karjalan Kewätkäköinen: Runolaulukirja. Helsinki: J.C. Frenck-
ell.
Europaeus, D.E.D. 1857. Suomalaisten puustavein äännöskuvat ylös-ajatellunna. Hel-
sinki: D.E.D. Europaeus.
Europaeus, D.E.D. 1863. Vorläufiger Entwurf über den Urstamm der indoeuropäischen
Sprachfamilie und seine vor-indoeuropäischen Abzweigungen, namentlich die fin-
nisch-ungarische: Die Zahlwörter der finnisch-ungarische Sprachen. Helsinki: Friis.
Europaeus, D.E.D. 1868–1870. Tietoja suomalais-ungarilaisten kansain muinaisista
olopaikoista. Suomi 2/7, 1–190; 2/8, 27–105.
Europaeus, D.E.D. 1869.13 Die finnisch-ungarischen Sprachen und die Urheimath des
Menschengeschlechtes: Zur Beleuchtung der archäologischen Fragen in Betreff des
ältesten vorhistorischen Daseins der Menschen. Helsinki: J. Simeliuksen perilliset.
Europaeus, D.E.D. 1871. M.A. Castrén, försvarad för missförstånd från ett och för otill-
börliga tillvitelser från annat håll. Helsinki: J. Simeliuksen perilliset.
Europaeus, D.E.D. 1873. Ett fornfolk med långskallig afrikansk hufvudskålstyp i norden,
bestämdt till språk och nationalitet, jemte Finsk-ungerska omdömen. Helsinki:
J. Simeliuksen perilliset.
Europaeus, D.E.D. 1877. Die Stammverwandtschaft der meisten Sprachen der alten und
australischen Welt. St. Petersburg: Obščestvennaja Pol’za.
Forsman, A.V. 1896. Taavetti Emanuel Taneli Europaeus. Jyväskylä: Gummerus.
Gyarmathi, S. 1799. Affinitas linguae Hungaricae cum linguis Fennicae originis grammat-
ice demonstrata. Göttingen: Dieterich.
Häkkinen, K. 1996. Indouralilainen arvoitus ja sen mahdollinen ratkaisu. Sananjalka 38,
7–23.
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).

via free access
Halila, A. 1988. Näkymiä D.E.D. Europaeuksen aatemaailmaan. In: M. Kuusi; P. Laak-

sonen; S. Timonen (eds.), D.E.D. Europaeus: Suurmies vai kummajainen. Helsinki:
Suomalaisen Kirjallisuuden Seura, 11–22.
Honti, L. 1993. Die Grundzahlwörter der uralischen Sprachen. Budapest: Akadémiai
Kiadó.
Idman, N. 1774. Försök at visa gemenskap emellan finska och grekiska språken, såsom tje-
nande till uplysning i finska folkets historia. Åbo: J.C. Frenckell.
Itkonen, E. 1966. Suomalais-ugrilaisen kielen- ja historiantutkimuksen alalta. Helsinki:
Suomalaisen Kirjallisuuden Seura.
Janhunen, J. 1999. Critical remarks on current Indo-Uralic comparisons. In: C. Has-
selblatt; P. Jääsalmi-Krüger (eds.), Europa et Sibiria: Beiträge zu Sprache und Kultur
der kleineren finnougrischen, samojedischen und paläosibirischen Völker. Wiesbaden:
Harrassowitz, 211–215.
Janhunen, J. 2000. Reconstructing Pre-Proto-Uralic: Spanning the millennia of linguis-
tic evolution. In: T. Seilenthal; T. Palo; A. Nurk (eds.), Congressus Nonus Interna-
tionalis Fenno-Ugristarum I: Orationes Plenariae & Orationes Publicae. Tartu: Eesti
Fennougristide Komitee, 59–76.
Joki, A.J. 1973. Uralier und Indogermanen: Die älteren Berührungen zwischen den urali-
schen und indogermanischen Sprachen. Helsinki: Finno-Ugrian Society.
Jones, W. 1799. Eighth anniversary discourse: On the borderers, mountaineers, and
islanders of Asia. Asiatick Researches 3, 1–20.
Jurvanen, P. 1988. Vanhan Suomen kasvatti: Europaeuksen suku- ja kotiseututaustaa.
In: M. Kuusi; P. Laaksonen; S. Timonen (eds.), D.E.D. Europaeus: Suurmies vai kum-
majainen. Helsinki: Suomalaisen Kirjallisuuden Seura, 139–151.
Kalevala 1849. = E. Lönnrot (ed.), Kalevala: Toinen painos. Helsinki: Suomalaisen Kirjal-
lisuuden Seura.
Key, T.H. 1846. The Lapp and Finn tongues not unconnected with the Indo-European
family. Proceedings of the Philological Society 2, 180–187.
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.
Köppen, F.T. 1886. Materialy kъ voprosu o pervonačal’noj rodině i pervobytnomъ rod-
stvě indo-evropejskago i finno-ugorskago plemeni. Žurnal ministerstva narodnogo
prosveščenija 246, 199–233; 248, 21–64, 227–250.
Korhonen, M. 1986. Finno-Ugrian Language Studies in Finland 1828–1918. Helsinki: Soci-
etas Scientiarum Fennica.
Kuusi, M.; S. Timonen 1988. Suurmies? Kummajainen? Uhrilammas? Keskustelua Euro-
paeuksen elämästä ja työstä. In: M. Kuusi; P. Laaksonen; S. Timonen (eds.), D.E.D.
Europaeus: Suurmies vai kummajainen. Helsinki: Suomalaisen Kirjallisuuden Seura,
23–50.

via free access
86 kallio
Lehikoinen, L. 1986. D.E.D. Europaeus kirjasuomen kehittäjänä ja kielentutkijana. Virit-

täjä 90/2, 178–202.
Marácz, L. 2012. The “Ugric-Turkic War” and the origin of the Hungarian language. Inter-
national Review of Turkish Studies 2/4, 8–23.
Niemi, A.R. 1903. D.E.D. Europæuksen kirjeitä ja matkakertomuksia. Helsinki: Suoma-
laisen Kirjallisuuden Seura.
Paasonen, H. 1907. Zur Frage von der Urverwandtschaft der finnisch-ugrischen und
indoeuropäischen Sprachen. Finnisch-Ugrische Forschungen 7, 13–31.
Pedersen, H. 1931. Linguistic Science in the Nineteenth Century: Methods and Results.
Cambridge: Harvard University Press.
Pulli, H. 1984. Europaeus-suvun varhaispolvista. Genos 55, 60–70.
Ramstedt, G.J. 1947. The relation of the Altaic languages to other language groups. Jour-
nal de la Société Finno-Ougrienne 53, 15–26.
Rask, R. 1818. Undersögelse om det gamle Nordiske eller Islandske Sprogs Oprindelse.
København: Gyldendal.
Rask, R. 1834. Afhandling om den finniske Sprogklasse. In: Samlede tildels forhen utrykte
Afhandlinger af R.K. Rask: Første Del. København: Poppske bogtrykkeri, 1–46.
Rédei, K. 1986. Zu den indogermanisch-uralischen Sprachkontakten. Wien: Österreichi-
sche Akademie der Wissenschaften.
Ronimus-Poukka, P. 2005. Anders Johan Sjögren: En finskhetens väckare? Historisk Tid-
skrift för Finland 90, 221–240.
Sajnovics, J. 1770. Demonstratio, idioma Ungarorum et Lapponum idem esse. Trnava: Col-
legium Academicum Societatis Jesu.
Salminen, J. 1985. Daniel Europaeus ja “Suomen suwun” juuret. In: A. Kemiläinen (ed.),
Mongoleja vai germaaneja? Rotuteorioiden suomalaiset. Helsinki: Suomen Histori-
allinen Seura, 249–268.
Salminen, V. 1905. D.E.D. Europaeus. Helsinki: Suomalaisen Kirjallisuuden Seura.
Scaliger, J.J. 1610. Opuscula varia antehac non edita. Paris: Drovart.
Setälä, E.N. 1915. Zur Frage nach der Verwandtschaft der finnisch-ugrischen und samo-
jedischen Sprachen. Journal de la Société Finno-Ougrienne 30, 1–104.
Sjögren, A.J. 1861. Gesammelte Schriften. St. Petersburg: Eggers.
Sköld, H. 1927. Indo-Uralisch. Finnisch-Ugrische Forschungen 18, 215–231.
Slöör, K. 1872. Välikokouksessa Jouluk. 22. p. 1869. Suomi 2/10, 220–223.
Sulkunen, I. 2004. Suomalaisen kirjallisuuden seura 1831–1892. Helsinki: Suomalaisen
Kirjallisuuden Seura.
Sweet, H. 1900. The History of Language. London: J.M. Dent.
Thomsen, V. 1869. Den gotiske sprogklasses indflydelse på den finske: en sproghistorisk
undersøgelse. København: Gyldendal.
Thomsen, V. 1890. Berøringer mellem de finske og de baltiske (litauisk-lettiske) Sprog: en
sproghistorisk undersøgelse. København: Dreyer.

via free access
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.

via free access
chapter 6
Bojan Čop’s Indo-Uralic Hypothesis

and Its Plausibility
Simona Klemenčič
Bojan Čop (1923–1994) was a professor of comparative grammar of Indo-Euro-

pean languages at Ljubljana University. He is well known for his contribution
to the study of the Anatolian and Tocharian languages. His numerous articles
and books on the Indo-Uralic genetic linguistic relationship represent a sub-
stantial, albeit often overlooked, contribution to the subject of the precursors
of the Proto-Indo-European language.
In his work, Čop expressed his belief in the Nostratic—“Ureurasische” lan-
guage family, basing it mainly on works of Vladislav Illich-Svitych, but he him-
self seldomly attempted to delve that deep into the linguistic past. He some-
times wrote about what he felt to be a more reliable genetic connection, the
one between the Uralic and Altaic language families. There are proofs, he says,
of these two language families being genetically connected. An example can be
found in Fig. 6.1.
In connection with his Indo-Uralic hypothesis, Čop believed Proto-Indo-
European very likely to be genetically closely connected with both the Uralic
and Altaic proto-languages, as is illustrated by the word for ‘hair’, cf. Fig. 6.2.
He admits that a genetic relationship between Uralic and Altaic is “unsicher”,
but nevertheless goes on to claim that he believes that once upon a time there
must have existed a Proto-Eurasian language. The tree as given in Fig. 6.3 is
Čop’s illustration of this claim from Die indouralische Sprachverwandtschaft
und die indogermanische Laryngaltheorie.
Many of us believe that once there must have existed a common ancestor to
at least some of the proto-languages known to us today. Čop’s attempt at estab-
lishing one segment of this overwhelming hypothesis is and remains, to my
knowledge, the most systematic, extensive and meticulously built approach to
the Indo-Uralic question to this day.
The Indouralica series was a huge enterprise. Here is what Čop was planning
to publish (Indouralica I, Čop 1974a: 15–16):
I: ural. m, n, l, r = idg. m, n, l, r;
II: ural. j, w = idg. i̯, u̯ ; /…/
III: ural. Sibilanten = idg. s;

via free access
bojan čop’s indo-uralic hypothesis and its plausibility 89
IV: ural. anlautende Tenues = idg. anlautende Tenues (oder s + Tenues);

V: ural. anlautende Tenues = idg. anlautende Mediae aspiratae;
VI: ural. Entsprechungen der indogermanischen an- und inlautenden
Mediae d, g, g̕, gu̯ ;
VII: ural. ŋ = idg. g, ng /…/;
VIII: indogermanische Entsprechungen der uralischen starken Tenues pp, tt,
kk;
IX: indogermanische Entsprechungen der uralischen inlautenden schwa-
chen Tenues p, t;
X: indogermanische Entsprechungen der uralischen intervokalischen
schwachen Klusile k und des Spiranten γ;
XI: indogermanische Entsprechungen uralischer inlautender Lautgruppen
vom Typus l, r, m, s, ś, ṣ̌, + k bzw. γ;
XII: indogermanische Entsprechungen der uralischen inlautenden Lautgrup-
pen vom Typus p bzw. k + Konsonant;
XIII: indogermanische Entsprechungen der uralischen anlautenden Affrika-
tae ć, č̣;
XIV: indogermanische Entsprechungen der uralischen inlautenden Affrikatae
ć, č̣ /…/;
XV: indogermanische Entsprechungen der uralischen Spiranten δ, δ’;
XVI: indogermanische Entsprechungen der uralischen Lautgruppen vom
Typus Nasal + Verschlusslaut;
XVII: einfache Vokale auf beiden Seiten;
XVIII: indogermanische Entsprechungen uralischer Verbindungen Vokal + j, w.
As can be seen from the list, Bojan Čop was systematically comparing seg-
ments of the Proto-Indo-European and Proto-Uralic phonetics, thus building
an image of the proposed Indo-Uralic phonetic system.
Although Čop sadly passed away at the age of 71 before completing the series
(Indouralica 3, 6, 11, 13 and 18 remain unpublished and no material that he might
have had prepared for these articles can be found today), he nevertheless left
a legacy of 14 articles and books of the Indouralica series that were published
between the years 1970 and 1990 plus seven more books and articles on the
Indo-Uralic theory, all published between 1970 and 1989. There are at least three
reasons why Čop’s work is not as known and popular as one believes it should
be:
– these articles are difficult to access even today;
– the topic is unpopular in linguistic circles;
– combing through the articles is hard work.
I set myself the task to collect and present his views on the matter. This is a
state-of-the-art report, all doubts and frustrations included.

via free access
90 klemenčič
figure 6.1 Taken from Indouralica I, Čop 1974a: 9
figure 6.2 Taken from Indouralica I, Čop 1974a: 11

via free access
figure 6.3 Language familiy tree as given in Čop 1970c
Čop believed that after the collapse of the Indo-Uralic proto-language it

was Uralic phonetics that remained more conservative, whereas Proto-Indo-
European would have gone through a series of changes; Proto-Uralic phonetics
therefore represents an earlier, and thus more authentic stage. His material has
by now unavoidably become outdated to some extent. His source for Proto-
Indo-European is Pokorny, for Uralic mainly Lehtisalo and Collinder—in his
articles Čop apologized for having a rather limited access to Uralistic literature.
It is his phonetic transcriptions used in the articles that I am following here—
at this point focusing primarily on logical consistency of his work. I took over
Uralic and Proto-Indo-European reconstructions as they were, without ques-
tioning them as long as it is undecided whether the hypothesis as a whole is
to be accepted as plausible or not. I did not question the outdated Proto-Indo-
European etymologies because even without delving into that question, there
is enough to be said about the inner consistency of Čop’s system as it is repre-
sented in his works.
There are hundreds of purported correspondences between Uralic and
Proto-Indo-European in the Indouralica series. I have unavoidably employed
some cherry-picking when I selected those etymologies that I found either par-
ticularly illustrative or problematic.
It seems that Čop changed his mind concerning some of the details now and
then in the process of publishing his work. This is understandable considering
the extent and complexity of the material, but confusing for a linguist trying
to outline his purported system. Nevertheless, all in all his system does present
itself as a whole and consistent, although necessarily very complex one.
As is evident from the titles of his articles, Čop was very well aware of the
dead ends that a researcher finds himself in when trying to establish long-range
genetic relationship in linguistics. Many of them are addressed in this volume.
Čop’s approach dealt with criticism aimed at attempts at long-range compar-
isons. He would

via free access
92 klemenčič
a. focus on phoneme clusters rather than single, isolated phonemes;

b. compare suffixes, thereby going beyond the basic root etymologies that
arguably fail to offer indisputable evidence, as is demonstrated in Ringe
(1999).
We will now look into his view of the prehistory of some of the Proto-Uralic
and Proto-Indo-European phonemes.
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
Where there is a diphthong in Proto-Uralic, there is a monophthong in Proto-

Indo-European.
As we can see, the Proto-Uralic system is more complex. Additionally there
would have been long vowels in Uralic and no opposition in quantity in Proto-
Indo-European. In accordance with his view that Proto-Uralic is the more
archaic language of the two (Indouralica XVII, Čop 1990: 21), Proto-Indo-Uralic
would have had nine short and five long vowels in the first syllable of bisyllabic
roots:
Short vowels: a, ä, e, e₂ (back vowel e), i, i₂, o, u, ü

Long vowels: e:, e₂:, i:, o:, u:
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

via free access
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.
Proto-Uralic > Proto-Indo-European
bisyllabic monosyllabic
aśe- ‘to be located, to dwell; place, spot’ es- ‘be’
His conclusion is that the Proto-Indo-European monosyllabic roots must have

been disyllabic at some point in the past. But before the vowel in root-final posi-
tion was lost, an umlaut (or instance of vowel harmony) took place (Indouralica
XII, Čop 1987: 138–139):
– if the Uralic root had a or ä in the second syllable, then the first vowel in the
matching Proto-Indo-European root became a;
– if the Uralic root had e in the second syllable, then the first vowel in the
matching Proto-Indo-European root became e:

via free access
94 klemenčič
Uralic Proto-Indo-European
Early stage With umlaut
*ki.ra- ‘curse’ *kera- > *kara- *kar- ‘abuse, punish’

*pitä- ‘hold, bind’ *peþa- > *paþa- *pas-ti- ‘firm’
*aśe- ‘dwell; spot’ *ase- > *ese- *es- ‘be’
*käte ‘hand’ *ǵhaþe- > *ǵheþe- *ǵhés-er/to- ‘hand’
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’
As for his postulated change of Proto-Indo-Uralic *e to Proto-Indo-European

*a—in view of the evidence he was forced to change it (Indouralica V, Čop

via free access
figure 6.4 Table from Čop 1975
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’:
*és- in *és-m(i) ‘I am’

*é- in *s-énti ‘they are’
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.

via free access
96 klemenčič
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:
Uralic PIE Hittite
k, ɣ H ḫḫ
ɣ H ḫ
An example containing an Indo-European laryngeal would be the following:

PU *näke- ‘see’: PIE *neH- ‘to be afraid, ashamed’, Hittite naḫḫ- ‘to be afraid’.
This parallel answers some of our questions considering laryngeals.
3 Proto-Uralic j, w, m, n, r, l = Proto-Indo-European i ̯, u̯ ( y, w), m, n,

r, l
The more complex situation of the Proto-Uralic is explained as more archaic,

whereas the Proto-Indo-European would have undergone a simplification (cf.
Indouralica I, Čop 1974a: 107–108).
PU PIE
Nasals
bilabial m-, -m- (s)m-, -m-

dental n-, -n- (s)n-, -n-
palatal ń-, -ń- (s)n-, -n-
retroflex ṇ-, -ṇ- (s)n-, -n-

via free access
(cont.)
PU PIE
Liquidae
dental l-, -l- (s)l-, -l-

palatal l̃-, -l̃- (s)l-, -l-
retroflex ḷ-, -ḷ- (s)l-, -l-
dental r-, -r- r-, -r-
3.1 Comparison of Roots

One type of roots is represented with VC(= j/w)V on the Proto-Uralic side and
with a diphthong VC(= i̯/ u̯ ) on the Proto-Indo-European side:
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:
3.1.1 Proto-Uralic C(m/n/r/l)VC( j/w)V
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’;

via free access
98 klemenčič
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’.
Moreover, according to Čop, in PIE, a root could be preceded by a mobile s-,

which he explains as a prefix meaning ‘up, upwards’.
In the other subtype the same consonants are found on the opposite posi-
tions as compared with the first type:
3.1.2 Proto-Uralic C(= j/w)VC(= m/n/r/l)V
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)?
3.2 Comparison of Suffixes

Uralic suffixes containing nasals, liquids and semi-vowels have the very sim-
ple structure -CV (Indouralica II, Čop 1972: 171). On the Proto-Indo-European
side the final vowel would be missing, so that what is left are merely these
consonants—j, w, m, n, r and l. These are very unspecific segments to work
with. Nevertheless, some of Čop’s attached evidence is very convincing. Let us
confine ourselves to some well known matching m-suffixes: the accusative case
ending -m in nominal forms, the 1st person singular ending -m in verbs and
-m- in formation of superlatives on both sides. But there are other premises

via free access
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

via free access
100 klemenčič
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. 1978. Indouralica V. Collectanea Indoeuropaea 1. Ljubljana: Univerza v Ljubljani,
145–196.

via free access
Čop, B. 1985. Indouralica X. Linguistica 25, 193–262.

Čop, B. 1987. Indouralica XII. Linguistica 27, 135–161.
Čop, B. 1989. Indouralica IX. Linguistica 29, 13–56.
Čop, B. 1990. Indouralica XVII. Razprave SAZU, II. razred, razred za filološke in literarne
vede 13, 21–46.
Matasović, R. 1998. Temeljni leksik i dubinska genetska srodnost. Folia onomastica
Croatica 7, 191–206.
Ringe, D. 1999. How hard is it to match CVC-roots? Transactions of the Philological Soci-
ety 97/2, 213–244.

via free access
chapter 7
Indo-European o-grade Presents and the Anatolian

ḫi-conjugation
Frederik Kortlandt
Elsewhere I have argued that Indo-European originated as a branch of Indo-

Uralic which was strongly modified under the pervasive influence of a North
Caucasian substratum, perhaps in the sixth millennium BC (cf. Mallory 1989:
192f., Kortlandt 2010: 387–428). I have proposed the following relative chronol-
ogy for the Indo-European branch of Indo-Uralic on the basis of the internal
evidence:
A. Indo-European vowel reduction, giving rise to full grade *e under the
stress and zero grade elsewhere;
B. phonetic lowering of *u (= syllabic *w) to *o, giving rise to a full grade (=
non-high) vowel in unstressed syllables;
C. analogical introduction of a full grade vowel in unstressed syllables (e.g.
in compounds), which automatically yielded new *o;
D. introduction of *o in stressed syllables (e.g. by decompounding), resulting
in a phonemic opposition between /e/ and /o/ under the stress;
E. analogical introduction of full grade *e in unstressed syllables, generaliz-
ing the opposition between /e/ and /o/;
F. rise of lengthened grade vowels *ē and *ō, yielding the conventional
Proto-Indo-European vowel system.
Under the assumption that the Indo-European laryngeals developed from a
glottal stop *q₁, a uvular stop *q₂, and a labialized uvular stop *q₃, my recon-
struction of the Proto-Indo-European perfect, stative (intransitive middle, e.g.
Vedic śáye ‘lies’) and (transitive) middle endings is the following (2010: 392 f.):

via free access
indo-european o-grade presents 103
Perfect Stative Middle
1sg. -q₂e -q₂ -mq₂

2sg. -tq₂e -tq₂o -stq₂o
3sg. -e -o -to
1pl. -me -medhq₂ -medhq₂
2pl. -e -dhq₂ue -tdhq₂ue
3pl. -(ē)r -ro -ntro
For an earlier stage I have proposed the following reconstruction (2010: 400 f.):
Perfect Stative Middle
1sg. -q₂-e -q₂ -m-q₂

2sg. -tq₂-e -tq₂ -s-tq₂
3sg. -e -o -t-o
1pl. -mq₂-e -medhq₂ -me-dhq₂
2pl. -q₂-e -dhq₂-ue -t-dhq₂-ue
3pl. -er -r -nt-r
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

via free access
104 kortlandt
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:
Stative Stage A1 Stage A2 Stages B–F
1sg. CeC-q₂ CeC-q₂ CeC-q₂

2sg. CeC-tq₂ CeC-tq₂ CeC-tq₂
3sg. CeC-u CeC-u CeC-o
1pl. CeC-mq₂ CC-medhq₂ CC-medhq₂
2pl. CeC-(t)q₂ CC-(t)dhq₂ue CC-dhq₂ue
3pl. CeC-r CeC-r CeC-r
Perfect Stages A–B Stages C–E Stage F
1sg. CC-q₂e CoC-q₂e CoC-q₂e

2sg. CC-tq₂e CoC-tq₂e CoC-tq₂e
3sg. CC-e CoC-e CoC-e
1pl. CC-mq₂e CC-mq₂e CC-mq₂e
2pl. CC-(t)q₂e CC-q₂e CC-q₂e
3pl. CC-er CC-er CC-ēr
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

via free access
with an ergative subject at stage E (for details I refer to my earlier treatment).

The developments can be summarized as follows:
Present D Present E Injunctive E
1sg. -o-q₁ -o-q₁ -o-m

2sg. -e-q₁ -e-q₁i -e-s
3sg. -e -e -e-t
1pl. -o-mq₁ -o-mq₁om -o-mo
2pl. -e-tq₁ -e-tq₁e -e-te
3pl. -o -o -o-nt
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-

via free access
106 kortlandt
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.

via free access
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)’.

via free access
108 kortlandt
– 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-.

via free access
– 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-
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.

via free access
110 kortlandt
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.
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.

via free access
chapter 8
The Proto-Indo-European Mediae, Proto-Uralic

Nasals from a Glottalic Perspective
Guus Kroonen*
Elsewhere, I have pointed out the possibility that the Proto-Indo-European

mediae, when envisioned as glottalized stops, can have developed from preglot-
talized nasals, i.e. *ɗ < *ˀn, *ɠ < *ˀŋ. This development is implied by the lexical
distribution of the participal no-suffix, which originally only occurred in roots
after *d, cf. Sanskrit bhinná- ‘split’, -chinna- ‘cut’, tunná- ‘struck’, Wakhi zü-bön
‘burst’, ra-sen ‘broke’, Welsh twnn ‘broken’ < PIE *bʰid-nó-, *skid-nó-, *tud-nó-.
Since this suffix is in complementary distribution with the more generic to-
suffix, it can be hypothesized that the former developed from the latter through
1) assimilation of *-ˀnto- > *-ˀnno- and 2) buccalization of *-ˀnno- > *-dno-,
where *d represents a doubly articulated stop consisting of a buccal and a glot-
tal closure (Kroonen 2018).
It now appears that this reconstruction of pre-PIE preglottalized nasals
can be brought in line with previous scholarship concerning the Indo-Uralic
Hypothesis. In 1972, Bojan Čop presented a number of Indo-Uralic isoglosses
in support of a regular sound correspondence between PIE *g (*ǵ, *gʷ) and
PU *ŋ. Recently, Martin Kümmel independently arrived at a similar conclu-
sion in a discussion of the most promising lexical matches between Proto-
Indo-European *ie⁽ǵ⁾- ‘ice’ and Proto-Uralic *jäŋe ‘id.’ (also cf. Collinder 1965:
124–125). Offering a diachronic typological perspective on the origin of the cor-
respondence, Kümmel suggested that “we might suspect a Proto-Uralic change
from implosives to nasals (or vice-versa?)” (2012: 305, see also this volume,
115–130). The sound correspondences can be observed in a small number of
potential cognate sets:
PIU *VnˀŋV-: PIE *ngʷ-ni- ‘fire’ ≈ PU *äŋ- ‘burn’1

PIU *ˀnVkV-: PIE *deḱ- ‘perceive’ ≈ PU *näki ‘see’
* 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.

via free access
112 kroonen
PIU *ˀɲVŋgV-: PIE *dnǵʰ- ‘tongue’ ≈ PU *ńï(ŋ)kćimi ‘palate, gills’2

PIU *jVˀŋV-: PIE *ie⁽ǵ⁾- ‘ice’ ≈ PU *jäŋi ‘ice’
PIU *pVˀnV-: PIE *ped- ‘step; fall’ ≈ PU *pane ‘put, place’3
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).

via free access
proto-indo-european mediae, proto-uralic nasals 113
form *ˀnui-ˀnḱmti- developed into *ˀnui-nḱmti- by dissimilation of the second

glottal stop, and then into *ʔui-nḱmti- by dissimilation of the first nasal. The
resulting *h₁uinḱmti- would have regularly developed into Skt. viṃśatí-.
It is relevant from this perspective, that some Germanic forms, too, can be
derived from the same nasalized form that appears to underlie Skt. viṃśatí-.
Old Norse tottogu, tuttugu and tyttugu, whose -tt- presupposed Proto-Germanic
*-nt-, can similarly be projected back into Proto-Indo-European as part of the
sequence *-ndeḱ- (Schmidt 1970: 128). Indeed, the variant tyttugu together with
Old Norwegian tuittugu (Hægstad 1915: 23) may simply continue PGm. *twin-
tegunþ < *duin-deḱmt-, where *deḱmt- represents the usual Germanic replace-
ment of PIE *-(h₁)ḱmti-. In conclusion, this variant, too, can be derived from PIE
*h₁uin(h₁)ḱmti-, although an additional explanation is required for the restora-
tion of the initial dental.
Given the wider dialectal distribution of traces of a nasal, it is actually
not inconceivable either that Lat. vīgintī similarly continues a nasalized form
*h₁uinḱmti-. This view was explicitly rejected by Thurneysen, who views the
nasal of Skt. viṃśatí- as a secondary, analogical intrusion from *saptãśati <
*septmḱmti and *navãśati < *neunḱmti (1883: 312). Instead, Thurneysen argues
that vīginti similarly acquired its voiced g analogically from the correspond-
ing cardinals *septm-(h₁)ḱmt-tmH-o- and *neun-(h₁)ḱmt-tmH-o-, where g would
have been regularly voiced between two nasals. The analogy becomes redun-
dant, however, by assuming that the inherited form was *h₁uinḱmti-, and that
in the resulting *vīngintī the first nasal was lost due to dissimilation.
Summing up, in this brief investigation I hope to have shown how the pos-
sibility of deriving PIE *ɗ, *ɠ from earlier *ˀ⁽ń⁾, *ˀŋ also opens potential new
inroads into the internal reconstruction of the Indo-European proto-language.
Firstly, it offers an alternative way to integrate seemingly unrelated problems
within Indo-European, namely that of the distribution of the no-participles and
origin of the nasal element of the numeral ‘20’. Secondly, the reconstruction
of such a series facilitates a more realistic phonetic interpretation of previ-
ously identified lexical similarities between Proto-Indo-European and Proto-
Uralic.
However, an important remaining question is whether the lexical material
displaying the correspondence of Proto-Uralic nasals and Proto-Indo-Euro-
pean mediae actually substantiates the Indo-Uralic Hypothesis or that it rather
reflects borrowing from Pre-Proto-Indo-European into Proto-Uralic. Future
studies will have to address this question, but regardless of the answer, we
are left with a small corpus of words suggesting that the PIE mediae devel-
oped from a series of some sort of prestopped, presumably preglottalized
nasals.

via free access
114 kroonen
References
Brugmann, K. 1911. Grundriss der vergleichenden Grammatik der indogermanischen

Sprachen. Zweiter Band: Lehre von den Wortformen und Ihr Gebrauch. Strassburg:
Trübner.
Collinder, B. 1965. Hat das Uralische Verwandte? Eine sprachvergleichende Untersu-
chung. Acta Universitatis Upsaliensis 1/4: 108–180.
Čop, B. 1972. Indouralica II. Ural-Altaische Jährbucher 44, 162–178.
Hægstad, M. 1915. Vestnorske maalføre fyre 1350. II. Sudvestlandsk, 2: Indre sudvest-
landsk. Færøymaal. Islandsk. Fyrste bolken. (Videnskapsselskapets skrifter II. Hist.-
Filos. Klasse. 1915. No. 3.) Kristiania: Dybwad.
Hübschmann, H. 1887. Etymologie und Lautlehre der ossetischen Sprache. Strassburg:
Trübner.
Kroonen G.J. 2018. Lachmann’s law, Thurneysen’s law, and a new explanation of the PIE
no-participles. In: L. van Beek; M. de Vaan; A. Kloekhorst; G. Kroonen; M. Peyrot;
T. Pronk (eds.), Farnah: Indo-Iranian and Indo-European studies in honor of Sasha
Lubotsky. Ann Arbor–New York: Beech Stave Press, 143–151.
Kümmel, M.J. 2012. 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.
Lubotsky, A.M. 1994. RV. ávidhat. 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, 201–206.
Opgenort, J.–R. 2004. The origin of implosive stops in Wambule Rai. Libju-Bhumju 25:
3–8.
Schmidt, G. 1970. Zum Problem der germanischen Dekadenbildungen. Zeitschrift für
vergleichende Sprachforschung 84/1: 98–136.
Thurneysen, R. 1883. Urspr. dn tn cn im lateinischen. Zeitschrift für vergleichende
Sprachforschung 26, 301–314.
Wei, J.X.; J.A. Edmondson. 2008. Sui. In: A.V.N. Diller; J.A. Edmondson; Y. Luo (eds.), The
Tai-Kadai languages. London–New York: Routledge, 585–596.

via free access
chapter 9
Thoughts about Pre-Indo-European Stop Systems

Martin Joachim Kümmel
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.

via free access
116 kümmel
2 Indo-Anatolian? Anatolian and PIE Stops
2.1 Two Scenarios

For PIA vs. PIE, more than one scenario is possible. The different systems will
be exemplified by dental-alveolar stops but these are meant to stand for the
whole class, i.e. *t = *p, t, k, …
A An Indo-Anatolian Scenario for the Stop Series

In a recent paper, Alwin Kloekhorst (2016, cf. Kloekhorst 2008a, 2010, 2013) has
reconstructed the PA stop system as a system with an opposition of voiceless
fortis (long) *tt and lenis (short) *t. In his view, this system is preserved from a
nearly identical PIA system with three types of stops, i.e. *tt | t | ˀt, the only dif-
ference being that the original preglottalized stops had lost their glottalization
and merged with the lenis series. However, the PIE system from which all other
branches of IE are descended, is reconstructed by him as showing an original
voicing contrast between voiceless fortis *t and voiced lenis *d plus preglottal-
ized voiced lenis *ˀd. Thus the main difference between these systems is the
one between the non-glottalized fortis and lenis: In PA and PIA, they were dis-
tinguished by length, in PIE by voicing. This implies a PIE innovation, a shift
from a length contrast to a voicing contrast, necessarily a chain shift tt > t > d =
t > d followed or accompanied by tt > t, while PA preserved the old system. Let
us call this model A. This reconstruction would clearly represent a good argu-
ment for PIA being distinct from PIE, since it requires a common innovation
of all IE branches other than Anatolian.
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.

via free access
thoughts about pre-indo-european stop systems 117
2.2 Evaluation of the Changes

To evaluate these two models, we can compare the probability of the changes
implied by them. In this respect, it is crucial to distinguish between con-
ditioned and spontaneous changes. Some sound changes are quite frequent
when conditioned by the environment but they are not at all frequent as a
default change in all phonotactic positions. For comparing our two models, the
probabilities of spontaneous changes are most relevant, since we are dealing
with unconditioned systemic shifts. Let us now take a look at the probabilities
of the single shifts implied by the models.
For the fortis stops, model A assumes degemination/shortening of long for-
tis stops in PIE. Such a change tt > t is generally quite frequent (Kümmel 2007:
133–136) and therefore quite probable. Model B assumes the opposite, i.e. a
lengthening of short voiceless stops in PA. Such a change t > tt is not infre-
quent as a conditioned change (e.g., in strong positions, cf. Kümmel 2007:
176–182), and it is attested as a spontaneous change in chain shifts (see below)
but less frequently, so it can be judged less probable than *tt > t. As for the
lenis stops, model B assumes devoicing of voiced stops in PA. This change d
> t is rare as a conditioned change but not infrequent in spontaneous shifts,
mostly in chain shifts (see below and cf. Kümmel 2007: 138–142); so its prob-
ability is not very high but also not very low. Model A assumes voicing of
voiceless stops in PIE. Such a change t > d is very frequent as a conditioned
lenition (in weak positions), but is extremely rare as a spontaneous change; in
fact, there is no really assured case (cf. Kümmel 2007: 47–54).2 The probability
of this change must thus be considered very low. This is not surprising con-
sidering the fact that voiced plosives are clearly less “natural” than voiceless
ones.
After looking at the individual probabilities of the changes, we may also try
to evaluate the probabilities of the whole chain shift. Model A presupposes a
shift tt > t > d for PIE. This type of change is well attested, but only as a condi-
tioned shift (lenition in weak positions), e.g. in Western Romance, and similarly
in some Uralic languages. Sometimes we rather find tt > t > (d >) ð (as in Lycian,
cf. also the Finnic “Stufenwechsel”). However, an unconditioned shift of this
type is not found anywhere, since initial *t is not expected to change (= yield
*d), as it would have to in PIE. Model B assumes a chain shift d > t > tt, and
while there is not much evidence for such a shift in conditioned environments,
it occurs as a spontaneous shift, though not very frequently:
2 The Western Armenian sound shift is only superfically similar, since it lead to a voiceless lenis
= /t/ > [d̥ ] vs. tʰ.

via free access
118 kümmel
In OHG, Pre-OHG *d / θ had changed via *d / ð into OHG t | d, and these

developed to Upper German tt | t (in all environments). The change is most
clearly attested with dentals, cf. Alemannic (Wallis/Zürich; gemination hap-
pened before open syllable lengthening) fater, betōn > fattər, pættu/pættə vs.
ledar, ladōn > lætær/lǣtər, latu/lātə; likewise in South Bavarian (Imst, cf. Schatz
1897; gemination after open syllable lengthening) fater, betōn > *fātər, *bɛ̄tən
> fɔ̄ ttər, pœi̯ttə; ledar, ladōn > *lɛ̄dər, *lādən > lœi̯tər, lɔ̄ tə. It may also be argued
that the somewhat different distribution of fortis pp, kk vs. lenis p, k also reflects
a parallel change of (originally allophonic) pre-OHG *b, g ~ *β, ɣ > early OHG
*p, k ~ *b, g.
The same change is found in IE loanwords into Fennic and Saamic (in non-
initial positions), where IE (Germanic or Slavic) t | d is generally reflected as tt
| t, cf. Germanic *katila- → Finn. kattila ‘kettle’ vs. *sadula- → Finn. satula ‘sad-
dle’; Germanic *salta- → Saamic *sālttē > Saami S(outh) saelhtie, N(orth) sálti,
I(nari) sältti ‘salt’; Germanic *haitaz → Saamic *hājttēs > Saami N háittis ‘hot’
or earlier *xaita- → *kajtta- > Saamic *kōjttē > Saami S guejhtie (Aikio 2006: 27)
vs. Germanic *sanda- → Finn. santa, Saamic *sāntē > Saami S saedtie ‘sand’;
Germ. *kaupa- > *kawppa > Finn. kauppa, (→) Saamic *kāwppē > Saami N gávpi
‘shop’ vs. Germanic (or Baltic) *naba- → Finn. napa, Saamic *nāpē > Saami S
naepie ‘navel’; Slavic *akuna (> CSl. *okъno) → Finn. akkuna ‘window’; *nīti (>
CSl. *nitь) → Finn. niitti ‘thread’; *paltina (> CSl. *poltьno) → Finn. palttina ‘can-
vas’ vs. *iges- (> CSl. ižes-) → *ikes- > Finn. ies, ikeen ‘yoke’; *babu (> CSl. bobъ)
→ Finn. papu ‘bean’.
This kind of chain shift and/or sound substitution appears to be quite typical
for the development of all such length-based “fortis-lenis” systems in European
languages.
2.3 Conclusion
The probabilities of change according to both models are summarized in the
following table:
A B
tt > t t > d tt > t > d t > tt d > t d > t > tt

conditioned ++ ++ + + + ?
spontaneous + – – + + +

via free access
While the probabilities for conditioned changes would appear to favour

model A, the probabilities for spontaneous changes (simple or connected)
clearly favour model B since model A would presuppose an improbable change
while model B does not. Therefore scenario B must be considered more prob-
able, and since it assumes a PIE archaism and a PA innovation, we must con-
clude that it does not provide an argument for an Indo-Anatolian model.
As additional evidence in favour of model A, Kloekhorst also adduced the
case of original *h₁éssi ‘you are’ which was apparently degeminated in PIE
*h₁ési > Vedic ási, Avestan ahī ̆ = Greek eĩ and thus is an independent instance of
degemination in PIE. Hittite e-eš-ši ‘you are’ does not show this degemination
and thus leaves the possibility that the degemination in *h₁éssi > *h₁ési was a
PIE innovation. However, the Hittite form cannot prove this since it could eas-
ily represent a trivial analogical innovation, as in Greek Homeric essí beside
*ehi, and Armenian es < *essi.
We can also adduce the following additional arguments for model B:
1) A length opposition in initial stops is difficult to maintain (and was in
fact lost in later Common Anatolian), so we might expect much more
variation in PIA (cf. the variation of initial lenis and fortis stops known
as “Notkers Anlautgesetz” in OHG) but evidence for such variation is
lacking. Even in PA, initial stops must still have preserved a contrast, cf.
Luwian j-, w- < *g-, *gʷ- vs. z-, kw- < *k-, *kʷ-. This would be easier to explain
if the original opposition was one of voicing and not only of length.
2) The famous PIE/PIA root structure constraints for stops (viz. T…T, D…D
vs. the absence of T…D, D…T) are much more plausible with a voicing dis-
tinction (cf. Miller 1977ab), since agreement of root-initial and root-final
stops in voicing is much more probable than agreement in length.
If the stop system thus does not show phonological innovations of PIE vs. PIA,
we may ask if there are other common innovations of IE languages vs. PIA.
However, potential cases are rather doubtful:
The good preservation of laryngeals in Anatolian is possibly the first point
coming to our mind, since this means that everywhere else they were weakened
and lost to a much higher degree than in Anatolian. However, most of these
changes must have been einzelsprachlich due to the differences in details that
we can find between the subfamilies, and partial preservation is also found else-
where (for Indo-Iranian see Kümmel 2016 and 2018). One may also assume that
many IE languages reflect a common change from uvular fricatives to (audi-
tively weaker) pharyngeals in the case of *h₂, *h₃ (cf. Kümmel 2007: 336; Weiss
2016: 331; 337 with ref.), but this is just a possibility for some of the branches, so
it does not qualify as a clear IE vs. IH innovation.
Another candidate might perhaps be the peculiar difference between PIE
*h₃ and a PA coronal sound reflected as Hittite s and Luwian d found in

via free access
120 kümmel
some few words like PIE *h₃okʷ- ‘eye’ = Hitt. sākʷ-, Luwian dāw(ī)- but this is
still much too unclear.3
3 Indo-Uralic? Implosives and Sonorants
On Indo-Uralic in general cf. Kortlandt 2002, 2004; Klingenschmitt 2005: 114–

116; Hyllested 2009ab; Kloekhorst 2008b; Kümmel fthc.; Kassian, Zhivlov &
Starostin 2015.
3.1 Stops in PIA/PIE/CIE

To assess the potential relationship to Uralic, it is important to start from the
earliest probable system for PIA. As mentioned above, there is some reason
to assume that the “classical” PIE contrast of “voiced aspirates” and voiced
stops was an innovation, and that the earlier system had something like voiced
explosives vs. implosives instead. To get from this earlier stage to the system
presupposed by Indo-Iranian and Greek, we can assume a “Central PIE sound
shift” that probably affected most languages, including Tocharian, but possibly
not Celtic, Balto-Slavic, and Anatolian (cf. Kümmel 2012ab; 2015), cf. the follow-
ing schema:
PIA CIE, IIr Greek
d > d̤ ʱ > th = chain shift ɗ > d > d̤ ʱ (> tʰ)

ɗ > d > d
These changes can be compared with parallel chain shifts in Mon and Tai lan-
guages (cf. Diffloth 1984; Li 1977; Pittayawat 2009):
Old Mon **ɓ, **ɗ (preserved in Modern > Nyah Kur b, d

Mon)
Old Mon **b, **d, **g > **b̤ ʱ, **d̤ ʱ, **g̈ ʱ > Nyah Kur pʰ, tʰ, kʰ; Modern Mon
p, t, k
Proto-Tai **ɓ,**ɗ (preserved in Wu-Ming) > Shan m, l; Po-ai m, n
> b, d elsewhere
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.

via free access
Proto-Tai **b, **d, **g > “Cao Bang” b̤ ʱ, d̤ ʱ, g̈ ʱ (Weiss

2009)
> Thai, Lao, Saek pʰ, tʰ, kʰ (vs. pre-
served inherited p, t, k)
> p, t, k elsewhere
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.
3.2 Comparison with Uralic

One problem for the Indo-Uralic hypothesis is the strong divergence of the stop
systems of the protolanguages: While PIA must have had three series *t | d |
ɗ, as we have seen before, PU only had one series (and some geminates) of
voiceless stops. The possible correspondence of one type to three increases the
danger of finding accidental similarities and therefore weakens the possibilities
to prove a genetic relation. However, the reconstruction of implosives in PIA
increases the probability that one of the three series might have a different cor-
respondence in Uralic. It is well known that implosives = “non-obstruent stops”
can alternate with other non-obstruents, i.e., nasals or liquids (cf. Haider 1983:
86; Stewart 1989: 239f.; Clements & Osu 2002; Clements & Rialland 2005: 18).
Therefore we may consider the possibility that PIA implosives correspond to
PU sonorants. And indeed, PU also had a rather large inventory of phonological
nasals made at all places, i.e. *m | n | ń | ŋ, while PIA only had two nasals *m and
*n. So the question arises if these nasals could not be the equivalents of one of
the PIA series, most probably of the implosives, i.e. PU *m, *n, *ŋ = (pre-)PIA
*ɓ (> *m/w), *ɗ, *ɠ. This possibility was already mentioned in Kümmel 2012a:
305, based on the potential equations PIA *jeɠ- ‘ice’ = PU *jäŋə and PIA *ɗek-
‘to perceive’ = PU *näkə- ‘to see’, but it has never been explored any further.
3.3 Correspondences of PIA “Mediae”

To investigate this possibility, I have gathered potential Uralic cognates4 of
PIA / PIE words with “mediae”, i.e. former implosives. All of these potential
equations must be considered quite tentative at our present state of knowl-
edge.
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).

via free access
122 kümmel
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.

via free access
(cont.)
b) PIA *Ɗ = PU *N
?*peɗ- ‘to step, fall; foot’ ̈ -):

*panə- ‘to put’ (T *pïni̮ ̄ Finn. pane-, Mansi pon-,
Ngan. huan-
*jeɠ-i/o- ‘ice’ *jäŋə ‘ice’: PS *jieŋə, Finn. jää, Hung. jég
*juɠʷ- ‘to join’ *je̮ŋ-sə ‘bow’ (T *juŋus/*juŋsu): PS *juoksə, Hung.
íj, Selk. i̮nti̮
?*luɠ- ‘to loosen, break’ *luŋə- ‘to lift’: PS *loŋə-, Finn. luoda, Hung. lóg
*piɠ- ‘to gnaw?’ *piŋə ‘tooth’: Finn. pii, Hung. fog
*buɠ/g- ‘to bend’? *poŋə ‘bosom’ (T *puŋu): PS *puoŋə, Finn. povi,
Chanty puɣəl
?*Heʛ- in PIIr. *(H)ágra- *e̮ŋa- ‘to open’ (T *ɨ̄ŋa-): Finn. avaa, Mansi ē̮ŋk-,
‘beginning’? Chanty aŋə-
?*seʛ- ‘to stick, grab’ *seŋə- ‘to comb’ (T*sǖni-): Mordva sem-/sej-; Komi
sin̮ -
?*bleɠ- ‘to shine’ *pilŋə ‘cloud’ (T *pǟliŋ/*pǟlŋi): Finn. pilvi
?*seʛ- ‘to stick, fasten’ ?*saŋ-ća- ‘to stand’ (T *sańťa-): PS *čuoňčō-;
Chanty *ɬańť-
?*neɠ (-o)- ‘sister’ (only ?*niŋä ‘woman’ in Saami I niŋálâs, Chanty ni(ŋ),
Anatolian!)6 Hung. nő (?); cf. Čop 1979: 21 f.
?*beʛ- ‘to break’ *päŋə ‘head’: Finn. pää; Hung. fő, Ngan. feai
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.

via free access
124 kümmel
As far as we know now, a change from nasals to implosives or to any kind

of non-nasal stops is not a common type of sound change cross-linguistically.
In contrast, changes from implosives to sonorants, including nasals, are much
better attested (as already mentioned above). It follows that PIU implosives
would be more probable, if the equations given above reflect something real.
Thus, the most probable PIU system of stop articulation types would either be
identical to the one of PIA or contain only one series of explosives vs. one of
implosives, if the distinction of voice in PIA arose by a secondary split (as it
did, e.g., in Hungarian).
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:
table 8.1 Table of correspondences of

stop types
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

via free access
Appendix: Similarities of Indo-Anatolian and Uralic
1 Strong Similarities in Pronouns and Grammatical Markers

Pronouns
1st 2nd 3rd
PU *m- ~ -n- *t ~ -nt- *s-

PIA *m-, *n- *t- [~ *w-] *s-
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
PU *-t / *-i- *-k

PIA *-s / *-j- *-h₁
Accusative Ablative-partitive Lative Locative Genitive
PU *-m *-tA *-k *-nA *-n

PIA *-m *-t(i),*-s, *-tos *-h *-n/r [*-(ó)s] (*-n-?)

via free access
126 kümmel
Verbal Markers
1s 2s 1p 2p 3s 3p
PU *-m(ə) *-t(ə) *-mA- *-tA- *-Ø *-n

PIA *-m(i) *-s(i) *-me *-te *-Ø / *-t(i) *-r / -nt(i)
Participles Past Modal
PU *-ntA- *-tA- *-mə- *-i- imp. *-k-

PIA *-nt- *-tó- *-mo-? ? subj. *-ho-?
2 Additional Lexical Material
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-

via free access
(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
Abondolo, D. 1996. Vowel Rotation in Uralic: Obug(r)ocentric Evidence. SSEES Occasional

Papers, 31. London: University of London.
Abondolo, D. 1998. Introduction. In: D. Abondolo (ed.), The Uralic Languages. London–
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2006. On Germanic-Saami contacts and
Saami prehistory. Journal de la Société Finno-Ougrienne 91, 9–55.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2012. On Finnic long vowels, Samoyed vowel
sequences, and Proto-Uralic *x. In: T. Hyytiäinen; L. Jalava; J. Saarkivi; E. Sandman
(eds.), Per Urales ad Orientem. Iter polyphonicum multilingue. Festskrift tillägnad
Juha Janhunen på hans sextioårsdag den 12 februari 2012. Helsinki: Finno-Ugrian Soci-
ety, 227–250.
Aikio, A. (Luobbal Sámmol Sámmol Ánte) 2015. The Finnic ‘secondary e-stems’ and
Proto-Uralic vocalism. Journal de la Société Finno-Ougrienne 95, 25–66.
Clements, G.N.; Sylvester Osu 2002. Explosives, implosives, and nonexplosives: Some
linguistic effects of air pressure differences in stops. In: C. Gussenhoven; N. Warner
(eds.), Laboratory Phonology 7. Berlin: Mouton de Gruyter, 299–350.
Čop, B. 1979. Indogermanisch-Anatolisch und Uralisch. In: E. Neu; W. Meid (eds.), Hethi-
Diffloth, G. 1984. The Dvaravati Old Mon language and Nyah Kur. Bangkok: Chula-
longkorn University Printing House.
Haider, H. 1983. Der Fehlschluß der Typologie. In: W. Meid (ed.), Philologie und Sprach-
wissenschaft: Akten der 10. Österreichischen Linguisten-Tagung Innsbruck, 23.–26. Ok-
tober 1982. Innsbruck: IBS, 79–92.
Häkkinen, J. 2009. Kantauralin ajoitus ja paikannus: perustelut puntarissa. Journal de
la Société Finno-Ougrienne 92, 9–56.

via free access
128 kümmel
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.

via free access
germanische Forschungen 121, 213–247.
Kortlandt, F. 1985. Proto-Indo-European glottalic stops: the comparative evidence. Folia
Linguistica Historica 6/2, 183–201.
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.
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.

via free access
130 kümmel
Miller, D.G. 1977a. Some theoretical and typological implications of an Indo-European

root structure constraint. Journal of Indo-European Studies 5, 31–130.
Miller, D.G. 1977b. Bartholomae’s Law and an IE root constraint. In: P.J. Hopper (ed.),
Studies in Descriptive and Historical Linguistics: Festschrift for W.P. Lehmann. Ams-
terdam: John Benjamins, 365–392.
Nikolaev, A.S. 2014. Greek εἱαμενή, Vedic yávasa-. Münchener Studien zur Sprachwis-
senschaft 68, 127–140.
Olsen, B.A. 1992. Notulae Indogermanicae I. Copenhagen Working Papers in Linguistics
2, 13–18.
Olsen, B.A. 2006. Hittite š from *h₃? In: G. Carling (ed.), GIŠ.ḪARgu-ul-za-at-ta-ra, Fest-
schrift for Folke Josephson. Göteborg: Meijerbergs arkiv för svensk ordforskning, 237–
247.
Pittayawat, P. 2009. The Phonology of Proto-Tai. Ph.D. dissertation, Cornell University.
Rieken, E. 2008. Die Zeichen ⟨ta⟩, ⟨tá⟩ und ⟨tà⟩ in den hieroglyphen-luwischen In-
schriften der Nachgroßreichszeit. In: A. Archi; R. Francia (eds.), VI Congresso Inter-
nazionale di Ittitologia, Roma, 5–9 settembre 2005, Parte II, 637–648.
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.
Schatz, J. 1897. Die Mundart von Imst. Laut- und Flexionslehre. Strassburg: Trübner.
Steer, T. 2015. Amphikinese und Amphigenese. Morphologische und phonologische Unter-
suchungen zur Genese amphikinetischer Sekundärbildungen und zur internen Deriva-
tion im Indogermanischen. Wiesbaden: Reichert.
Stewart, J.M. 1989. Kwa. In: J. Bendor-Samuel (ed.), The Niger-Congo Languages: a Clas-
sification and Description of Africa’s Largest Language Family. Lanham: University
Press of America, 216–245.
Tálos, E. 1987. On the vowels of Proto-Uralic. In: K. Rédei (ed.), Studien zur Phonologie
und Morphologie der uralischen Sprachen: Akten der 3. Tagung für uralische Phonolo-
gie, Eisenstadt, 28. Juni–1. Juli 1984. Wien: Verband der wissenschaftlichen Gesell-
schaften Österreichs, 70–80.
Tálos, E. 2015. Az ősuráli vokalizmus természetes rekonstrukciója. Manuscript.
Weiss, M. 2009. The Cao Bang Theory. Available online at http://conf.ling.cornell.edu/
weiss/Cao_Bang_Theory.pptx (July 6, 2015).
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
Zhivlov, M. 2014. Studies in Uralic vocalism III. Journal of Language Relationship /
Voprosy jazykovogo rodstva 12, 113–148.

via free access
chapter 10
The Anatolian “Ergative”*

Milan Lopuhaä-Zwakenberg
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.

via free access
132 lopuhaä-zwakenberg
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.
2 Proposals for Analyses of the Agentive
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 -antis 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.

via free access
the anatolian “ergative” 133
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
Hieroglyphic Luwian -antis ? -adi

Cuneiform Luwian -antiš -antinzi -adi
Lycian ? -ẽti -edi, -adi
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/.

via free access
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 -antis 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 -antis 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 -antis 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

via free access
The difference between these two analyses, as is mentioned by Melchert,

is that under the ergative analysis the underlying noun X-antis of the com-
mon gender, whereas under the syntactical analysis the underlying noun X
is a neuter noun. Since a resumptive pronoun corresponding to X bears the
morphological gender of X, anaphoric reference can show us which analysis is
correct: if a noun in the agentive construction is resumed by a neuter pronoun,
this shows that the ergative analysis is correct, whereas if the resumptive pro-
noun is of the common gender, the syntactical analysis is correct. Furthermore,
if X governs an adjective Y, we expect Y to appear in the common nominative
if the suffix -anza is decomposable, and in the neuter ergative if it is indecom-
posable. Of course, a priori the adjectival common nominative ending might
be identical to the neuter ergative ending, so formally an attestation of the form
Y-aš X-anza does not allow us to distinguish between the two analyses. If, how-
ever, we find a construction of the form Y-anza X-anza, this is a clear argument
in favour of the ergative analysis.
The following table gives an overview of the characteristics that allow us
to distinguish between analyses. In particular, we can see that if a difference
in meaning between the base noun and the noun in the agentive position is
found, then it is certain that the derivational analysis is correct in that exam-
ple. The same holds if we find the agentive construction applied to a common
gender base noun. On the other hand, if a word in the agentive construction
has a modifying adjective with an ending that is not the common nominative
ending, or if it is resumed by a neuter pronoun, the ergative analysis must be
correct. In the next sections I will discuss the languages in which the agentive
construction is attested one by one.
Derivational analysis Inflectional analysis Ergative analysis
Difference in meaning Yes No No

Gender of base noun Common/Neuter Neuter Neuter
Adjectives Common nominative Common nominative Neuter ergative
Resumptive pronouns Common Common Neuter
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 -antis an inflectional
suffix as a synchronic description.

via free access
3 The Agentive in Hittite
I start by summarising the results of Goedegebuure (2013) on the Hittite agen-

tive construction. She distinguishes between Old, Middle and Neo-Hittite. Her
theory is as follows. In Old Hittite the agentive construction is purely semanti-
cal, and can be seen as an instance of the singulative use of the suffix -ant-. By
the time of Middle Hittite, however, the agentive construction had been gram-
maticalised and had lost its semantic value. In Middle Hittite, nouns in the
agentive construction are resumed by both neuter and common gender pro-
nouns; as such it could be ambiguously analysed as both an inflectional suffix
and an ergative ending. The diachronical development is clear: after the suffix
was grammaticalised, it could only appear in its nominative forms -anza and
-anteš. The lack of inflection led to a reanalysis as case endings of the underly-
ing neuter word rather than a suffix which changed the morphological gender.
We find the same ambiguous analysis for the adjectives governed by nouns
in the agentive construction. These either have a common gender nominative
ending -š or a neuter gender ergative ending -anza, also confirming the fact that
in Middle Hittite both the inflectional and the ergative analysis can be applied.
Here, the ergative adjective ending -anza was introduced from the nominal
inflection.
Although the agentive construction in Middle Hittite is in development
between a grammatical suffix and a morphological case, by the time of Neo-
Hittite, we only find neuter resumptive pronouns and adjectives in -anza. This
indicates that in Neo-Hittite the agentive construction can unambiguously be
analysed as an ergative case. Thus in Hittite we can clearly see a development
from a suffix with a semantic value to a grammatical suffix to the formation
of a new case. Still, however, the singulative suffix -ant- continued to exist in
Neo-Hittite as a semantic derivation.
While Goedegebuure provides evidence for the stages of Middle and Neo-
Hittite, she does not give any evidence for her claim that in Old Hittite the
agentive was only semantical. To investigate this claim one would have to con-
sider all Old Hittite attestations for the agentive. By my knowledge, there are
only two of such attestations, which we will discuss below.6
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:

via free access
Example 1: KBo 25.107, 4–6
4. [a]p-pu-uz-zi kar-ta-x[…]
5. ták-ku=uš še-e-er […]
6. [ap]-pu-uz-zi-an-za […]
‘… the animal fat …; if thus the animal fat … them’
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.
Example 2: KUB 36.106 rev. 5–7
5. [ … tu]p-pí-aš ut-ta-a-ar šar-ri-et-[ta]

6. [ … l]i-in-ki-i ̯a-an-te-eš ap-pa-an-tu
7. [ … ] n=a-aš ḫar-ak-tu
‘The word of the tablet is broken. The oaths must seize … . He must perish.’
7. [i-t]a-lu[(-u̯a) a-ra-i-iš ḫ]e-ẹn-ka-an ta-aš

8. [ … ]x[ … t]a-aš a-aš-šu ar-ḫa tu-uḫ-še-et
‘Evil arose; it took the plague; it took …; it has cut off the good’
However, as remarked by Weitenberg (1987: 227), since the actual ending of the word is
not attested, this fragment is too indecisive to constitute evidence for the use of neuter
nominative-accusative forms in the agent position in Old Hittite.
7 Alwin Kloekhorst (p.c.) informs me that this interpretation (by Puhvel) cannot be correct,
since the clitic =ma cannot follow the sentence-initial particle nu-.

via free access
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.
4 The Agentive in Cuneiform Luwian
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).
Example 3: KUB 9.6+ ii 14–16
14. a-a-aš-ša=ti e-el-ḫa-a-du tap-pa-ša-an-ti-iš

15. ti-i ̯a-am-ma-an-ti-iš ta-a-i-in=ti-i̯=a-ta a-i-i̯a-ru
16. ma-al-li=ti-i̯=a-ta [a-]i-i̯a-ru
‘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

via free access
sky themselves but to their personifications or deifications. Thus in this sen-

tence the semantical analysis is applicable.
Example 4: KUB 35.54 ii 49–iii 5
ii 49. [š]a-a-an-du-u̯ =a-ta pár-na-an-ti-in-zi

ii 50. [ḫ]u-u-um-ma-ti-iš ḫa-aš-ša-ni-it-ti-iš
ii 51. ḫu-u̯a-aḫ-ḫur-ša-an-ti-in-zi ti-i̯a-am-mi-iš
iii 1. ta?-ru-ša-an-ti-iš ad[-du-u̯ a-al-za ú-tar-ša]
iii 2. ḫal-li-iš-ša pa-ra-at-ta-an[-za]
iii 3. pu-u̯ a-ti-il-za [n]a-nu-un-tar-ri-š[a]
iii 4. ir-ḫu-u-u̯ a-aš-ša pa-ri-it-tar-u-u̯ a-a-aš-š[a]
iii 5. u-la-an-ta-al-li-i̯a-an ḫu-it-u̯ [a-li-i̯a-an]
‘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.
Example 5: KUB 35.107+ ii 7–12
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 […]
9. [tap-paš-š]a te-ra-a-im-ma-an-za dIŠKUR-za d[…]

10. [a-ri-]in-ta […]
11. [du-ú-]u̯ a-az-za-an ti-i̯a-am-me-in DINGIRLUM RA-BU-Ú

12. [ti-]i̯a-am-ma-aš-ši-iš=ḫa dUTU-u̯ a-za 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.’

via free access
In line ii 7 of Example 5 we find an agentive construction based on kur-

šau̯ ar ‘island’, a neuter r/n-stem. Unfortunately, the context is unclear. The next
line has a parallel construction, but āppuu̯ aninzi is a hapax. This presents us
with two possibilities for the interpretation of these lines: either the agen-
tive construction in kuršaunantinzi is grammatical, so that āppuu̯ aninzi is
another landscape feature, or the agentive construction denotes a personi-
fication or deification, so that āppuu̯ aninzi is another group of deities. The
next lines give us the answer: here we see deities as the subjects of the verb
arinta. Therefore it is best to view kuršaunantinzi and āppuu̯ aninzi as deities
as well, so that kuršaunantinzi ‘island deities’ would differ in meaning from
kuršau̯ ar ‘island’, and as such the semantical analysis fits this sentence the
best.
Example 6: KUB 35.107+ iii 15–18
15. [a=u̯ a=t]i ÍD.ḪI.A-in-za ḫa-pí-in-ni-in-za KI.MIN a=u̯ a=t[i]

16. [SAG.D]U-aš-ša-an-za IGI.ḪI.A-u̯ a-aš-ša-an-za GIG-an-za na-a-u̯ a
[KI.MIN]
17. [ … ] SAG.DU-aš-ši-iš IGI.ḪI.A-u̯ a-aš-ši-iš GIG-an-te-eš₁₇?

18. [ … ]x tar-pí-i-ta …
‘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.’
Line iii 17 of Example 6 contains the agentive GIG-anteš. Although we cannot

determine the meaning of the sentence it occurs in, it is useful to compare
it to lines iii 15 en iii 16 , where we see the same noun with the neuter sin-
gular nominative/accusative ending -anza, as well as with the same genitival
adjectives. This makes it likely that the two instances of the word GIG have the
same referent, which implies that the suffix -antis here grammatical rather
than semantical. It is also important to point out that the two genitival adjec-
tives SAG.DU-aššiš and IGI.ḪI.A-u̯ aššiš have the common nominative ending
-iš rather than a hypothetical ergative ending -antiš. This makes it likely that
GIG-anteš is a common gender nominative noun rather than a neuter gender
ergative noun. We cannot formally rule out the latter, since it is possible that -iš
is not only the common nominative plural ending, but also the neuter ergative
plural ending. As such, the syntactical analysis fits this sentence, although the
ergative analysis is possible as well.

via free access
Example 7: KUB 35.112, r.col. 3–7
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 -antis 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.
Example 8: KUB 35.65 iii 14–16
14. [ … ]=pa=an a-ap-pa ḫa-ra-at-na-an-ti-iš

15. [ … ]x-ti-iš ni-iš
16. [ … ma-a]l-ḫa-aš-ša-as-si-in EN-an
‘The […] offense must not […] him, the lord of the ritual.’
The word ḫaratnantiš ‘offense’ of Example 8 is otherwise also attested as a

neuter r/n-stem in the nom./acc.sg. ḫāratarša and the ablative/instrumen-
tal ḫaratnati. The word […]X-tiš is probably congruent to ḫaratnantiš, and it
could give us information about the congruence relations of the agentive in
Cuneiform Luwian. Unfortunately, only the very ambiguous part -tiš of the end-
ing has been preserved. This sentence therefore does not help us in determining
the nature of the agentive construction.

via free access
Example 9: KUB 35.86 ii 5
5. [ … -]zi ḫar-da-an-ti-in-zi
‘…’
Melchert glosses the word ḫardantinzi in Example 9 as an ergative because the

ablative-instrumental ḫartati has been attested as well. Given the lack of con-
text, however, we cannot determine whether this is an agentive construction,
or simply another derivation by means of the suffix -ant- (Melchert suggests
‘having ḫ.’).
From the examples it is clear that in Cuneiform Luwian the agentive suffix
-ant- could be found both as a semantic derivation, denoting personification or
deification, and as an inflectional suffix that has a purely grammatical function.
In Example 6, the two nominal adjectives corresponding to the agentive ending
in the common nominative ending -iš point to the fact that the agentive is a suf-
fix rather than an ergative ending in Cuneiform Luwian. Although formally we
cannot rule out the possibility that this is the neuter ergative ending, we would
expect the neuter ergative adjectival ending to be -antiš, similar to the neuter
ergative adjectival ending -anza of Neo-Hittite. Therefore it is best to suppose
that in Cuneiform Luwian there is no separate ergative case and that both the
semantical and the syntactical analyses occur.
5 The Agentive in Hieroglyphic Luwian
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.
Example 10: BOYBEYPINARI 2 §21
iii.B (“CAELUM”)ti-pa-sa-ti-sa=pa=wa/i=tu-u (“TERRA”)ta-

iii.C sà-REL+ra/i-ti-sa=ha || CAELUM-sa=ha TERRA-
iv.A REL+ra/i-sa=ha DEUS-ni-i-zi LIS-tà-ti || CUM-ni X-tu

via free access
‘The sky, the earth, and the gods of the sky and the earth must … him with
ligitation.’
In Example 10 the agentives (“CAELUM”)ti-pa-sa-ti-sa = tipasantis and

(“TERRA”)ta-sà-REL+ra/i-ti-s = taskwirantis are derived from the neuter noun
tipas ‘sky’ and the common gender noun taskwira/i- ‘earth’. Since in the latter
case the agentive is made from a common gender noun, it has to be a semanti-
cal derivation rather than a grammatical suffix, since the use of a grammatical
suffix would be triggered by the morphological neuter gender. This means that
we should understand the subject of this sentence as ‘The sky god, the earth
god, and the gods inhabiting the earth and the sky.’ We can conclude that -ant-
is present in Hieroglyphic Luwian as a personifying/deifying suffix. However,
since this is the only attestation of the agentive in Hieroglyphic Luwian, we
should not regard the absence of evidence of a grammatical function of the
suffix -ant- as evidence of its absence.
6 The Agentive in Lycian
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.
Example 11: TL 135
2 … s=ene teseti : tubeiti : trm̃ mili
‘And the Lycian oaths will strike him.’
Example 12: TL 149
10 … s=ẽne : tesẽti : qãñti : trm̃ milijẽt#i […]
‘And the Lycian oaths will seize him.’
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

via free access
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.
7 The Agentive in Proto-Anatolian
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

via free access
Proto-Anatolian. We will do this by considering the proposals for possible anal-

yses of section 2 and see how well they hold up for Proto-Anatolian. As in the
rest of this article we leave the ablative analysis aside.
The only Anatolian languages in which the agentive construction is a reflex
of an actual ergative case are Lycian and the later stages of Hittite. In Hittite
this ergative case in -anza was created as a reinterpretation of the nomina-
tive case of a suffix -ant- (Goedegebuure, 2013). The Lycian ergative is attested
only once, and we find a parallel construction with an inflectional suffix -et-i
(= -ẽt-i). The Lycian agentive is therefore structurally the same as the Middle
Hittite agentive, where the agentive can also be either an inflectional suffix or
a morphological case. Thus we can posit the same development for Lycian as
we can for Hittite: the agentive was originally an inflectional suffix that was
reanalysed as a case ending within the history of the language. Thus in both
languages the ergative is an innovation, and as such we cannot reconstruct
an ergative case into Proto-Anatolian. Since in Proto-Anatolian the agentive
is not an ergative case, we must conclude that every instance of the agen-
tive in Anatolian has developed from a Proto-Anatolian suffix *-ont-. There are
two grammatical roles of this suffix in the agentive construction present in
the Anatolian languages: there is a semantic suffix which denotes personifi-
cation, which is present in the earlier stages of Hittite, in Cuneiform Luwian
and in Hieroglyphic Luwian, and a grammatical suffix that obligatorily trans-
fers neuter nouns to the common gender in the agent position, which is present
in Hittite, Cuneiform Luwian and Lycian. The semantic suffix can be seen as a
specific instance of the Hittite suffix -ant- discussed in section 2. This suffix
has cognates in other branches of Indo-European (Oettinger, 2001), so we may
conclude that this suffix is old and that there existed a semantic suffix *-ont- in
Proto-Anatolian that had personification as one of its functions. Furthermore,
we can regard the syntactical suffix *-ont- as a degrammaticalisation of the
semantic suffix *-ont-. Such a development can be sketched in Lycian as follows
(but of course the same development must have taken place in all Anatolian
languages). In a non-attested stage of Lycian, the sentence sẽne tesẽti tubeiti
trm̃ mili must have meant ‘and the Lycian oaths [personified as deities] will
strike him.’ However, since the suffix *-ont- did not have a single, well-defined
semantic function in Anatolian (Josephson, 2004), the semantic component of
the derivation tesi/tesẽti was lost and the sentence simply came to mean ‘and
the Lycian oaths will strike him.’ As a result, the suffix now had a purely gram-
matical function, namely to be able to put neuter nouns in the agent position
by means of changing their grammatical gender. The semantical suffix, in con-
trast, did not only change the morphological gender, but also the semantical
animacy of the noun.

via free access
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.
8 The Proto-Indo-European Alignment System
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).

via free access
Proto-Classical Indo-European Proto-Anatolian
Masculine Neuter Common Neuter
Agent *-os *-0m *-os ×

Subject *-os *-0m *-os *-on
Patiens *-0m *-0m *-on *-on
The question is now what we should reconstruct for Proto-Indo-European. This

touches on the Indo-Anatolian hypothesis, which states that Anatolian was the
first branch to split off from Proto-Indo-European.8 Thus, if Proto-CIE is the lat-
est common ancestor of the CIE branches, then the Indo-Anatolian hypothesis
states that PCIE is a later language than PIE; in other words, the non-Anatolian
branches of Indo-European would share common innovations. Hence if we
find that PCIE has undergone an innovation with regards to PIE, then this con-
stitutes evidence for the Indo-Anatolian hypothesis.
The only point at which PCIE and PA differ is the ending for the neuter
agent. A priori there are two possible reconstructions for the neuter agent in
PIE:
1. The PIE neuter agent ending was *-om as in PCIE, and Anatolian inno-
vated in disallowing neuter nouns in the agent position;
2. As in Proto-Anatolian, PIE did not allow neuter nouns in the agent posi-
tion; CIE innovated by introducing *-om for neuter agents.
In order to accept the first reconstruction there has to be a reason why Ana-
tolian removed neuter agents from its grammatical systems. This can be ex-
plained by the fact that most PIE inanimate nouns were neuters. In practice,
most inanimate nouns could only appear in the agent position via the personi-
fying suffix *-ont-, which existed in PIE already, for example in *gérh₂-ont- ‘the
old one’ (Gr. γέρων, Skt. járanta-), which is derived from *gérh₂o- ‘old’ (Arm.
cer) (Oettinger 2001: 302–303). Thus, it is conceivable that the original way to
express neuter agents was lost, and the personifying suffix *-ont- was subse-
quently grammaticalised.
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.

via free access
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
Masculine Neuter Masculine Neuter Common Neuter
Agent *-os × *-os *-om *-os ×

Subject *-os *-om *-os *-om *-os *-on
Patiens *-om *-om *-om *-om *-on *-on
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

via free access
could be described by means of a mediopassive construction or by means of

the semantic suffix *-ont- (which would slightly alter the meaning of the sen-
tence). PIE would then be somewhat similar to the situation ascribed to Ana-
tolian according to the semantical analysis of section 2. It is, however, hard to
ascertain the precise construction used, since such a construction would have
disappeared in both CIE and Anatolian.
Since PIE did not have the neuter agent ending *-om, the CIE languages
share a common innovation. Thus, the development of the CIE alignment
system constitutes an argument in favour of the Indo-Anatolian hypothesis.
However, one might argue that these developments could have occurred inde-
pendently, as the absence of neuter agents constituted a gap in the PIE align-
ment system. Extending the neuter subject ending *-om to the agent func-
tion is a straightforward way to fix this gap. Thus, although this innovation
points towards a period of common innovation of the CIE languages and hence
towards an early separation of Anatolian from CIE, it by itself does not conclu-
sively prove it.
References
Benveniste, É. 1962. Les substantifs en -ant- du Hittite. Bulletin de la Societé de Linguis-

tique de Paris 57, 44–51.
Garrett, A. 1990. The origin of NP split ergativity. Language 66, 261–296.
Goedegebuure, P. 2013. Split-ergativity in Hittite. Zeitschrift für Assyriologie und vorder-
asiatische Archäologie 102, 207–303.
Hawkins, J.D. 2000. Corpus of Hieroglyphic Luwian Inscriptions Vol. 1: Inscriptions of the
Iron Age. Berlin–New York: Walter de Gruyter.
Josephson, F. 2004. Semantics and Typology of Hittite -ant-. In: J. Clackson; B.A. Olsen
(eds.), Indo-European Word Formation: Proceedings of the Conference Held at the Uni-
versity of Copenhagen, October 20th–22nd 2000. Copenhagen: Museum Tusculanum,
91–118.
Kammenhuber, A. 1993. Zum Modus Injunktiv und zum Drei-Genus-System im Ur-
Indogermanischen (ca. 3000–2500 v. Chr). In: A. Kammenhuber Kleine Schriften
zum Altanatolischen und Indogermanischen. Heidelberg: Winter, 698–729.
Laroche, E. 1962. Un ‘ergatif’ en indo-européen d’Asie Mineure. Bulletin de la Societé de
Linguistique de Paris 57, 23–43.
Melchert, H.C. 1993. Cuneiform Luvian Lexicon. Available online at http://www
.linguistics.ucla.edu/people/Melchert/LUVLEX.pdf (2018).
Melchert, H.C. 2001a. Cuneiform Luvian Corpus. Available online at http://www
.linguistics.ucla.edu/people/Melchert/CLUVIAN.pdf (2018).

via free access
Melchert, H.C. 2001b. Lycian Corpus. Available online at http://www.linguistics.ucla

.edu/people/Melchert/lyciancorpus.pdf (2018).
Melchert, H.C. 2007. The Problem of the Ergative Case in Hittite. In: M. Fruyt; M. Mazoyer;
D. Pardee (eds.), Grammatical Case in the Languages of the Middle East and Europe.
Chicago: The Oriental Institute of the University of Chicago, 161–167.
Oettinger, N. 2001. Neue Gedanken über das nt-Suffix. In: O. Carruba; W. Meid (eds),
Anatolisch und Indogermanisch/Anatolico e Indoeuropeo: Akten des Kolloquiums der
Indogermanischen Gesellschaft. Pavia, 22.–25. September 1998. Innsbruck: IBS, 301–
316.
Patri, S. 2007. Les structures d’alignement dans les langues indo-européennes d’Anatolie.
Wiesbaden: Harrassowitz.
Puhvel, J. 1984. Hittite Etymological Dictionary: Vol. 1: Words Beginning with A; Vol. 2:
Words Beginning with E and I. Berlin–New York–Amsterdam: Mouton.
Valério, M. 2009. Palaic fulāsinanza: One Anatolian suffix, two possible explanations.
Journal of Indo-European Studies 37, 421–429.
Weitenberg, J. 1987. Proto-Indo-European nominal classification and Old Hittite. Mün-
chener Studien zur Sprachwissenschaft 48, 213–230.

via free access
chapter 11
The Indo-European Suffix *-ens- and Its Indo-Uralic

Origin
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
2.1 PIE *ǵhh₂-ens-
– Gr. χήν m. ‘goose’ (nom.pl. χῆνες, Dor., Boeot. χᾱ́ν);

– Skt. haṃsá- m. ‘goose’ (RV+);
– Lat. ānser m. ‘goose’ (< *hānser < *ǵhh₂ens-(e)ro-);
– OIr. géis f. ‘swan’;
– PGerm. *gans- f. ‘goose’: OHG gans, ON gás, OE gōs, etc.;
– Lith. žąsìs f. ‘goose’ (gen.pl. žąsų̃ ), Latv. zùoss ‘id.’, OPr. sansy ‘id.’;
– PSlav. *gǫsь f. ‘goose’: Ru. gus’; OCz. hus; Slk. hus; Pl. gęś; Sln. gọ̑ s.
The word for ‘goose’ is likely to be a derivative of the PIE root *ǵheh₂- ‘to gape’
(Gr. χάος n. ‘chaos’ < *ǵhh₂-(e)u-, Gr. χάσκω ‘to gape, yawn, open the mouth wide’
< *ǵhh₂-n-, etc.; for the reconstruction of the root I refer to Lubotsky 2011: 107 f.).
This etymology is very attractive both from the formal and semantic point of
view, as the geese are ‘gaping’ birds.
The evidence points to an athematic stem, and we would certainly expect an
ablauting paradigm (Kortlandt 1985 and 2013 reconstructed nom.sg. *ǵheh₂ns,
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.

via free access
152 lubotsky
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
2.2 PIE *gwhr-ens(o)- ‘heat’
– 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.

via free access
the indo-european suffix *-ens- and its indo-uralic origin 153
2.3 PIE *dh₁-ens- ‘dense’
– 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
6 A similar solution is given by Nikolaev (2010: 241).

7 Only the names of diseases taššii̯atar n. (nom.-acc.sg. ta-aš-ši-i̯a-tar), taššii̯au̯ ar n. (nom.-
acc.sg. ta-aš-ši-i̯a-u-u̯ a-ar) and taššii̯ama- c. (acc.sg. ta-aš-ši-i̯a-ma-an), which have often been
etymologically connected with daššu- (see Kloekhorst 2008: s.v. daššu-), are spelled with ta-,
but this etymology is most probably wrong.
8 Van Beek (2013: 250f.) considers the semantic development from ‘dense’ to ‘experienced’ and
gives as a parallel Gr. πυκνός, πυκινός ‘hairy, dense’ and πυκιμήδης ‘shrewd’, lit. “with dense
plans”, πυκινόφρων ‘id.’, lit. “with dense mind”. To this it can be added that in Indo-Iranian, the
notion of ‘experienced, dexterous’ has further developed into ‘supernaturally dexterous’.

via free access
154 lubotsky
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’).
2.4 PIE *trh₂-ns
– 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

via free access
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’.
2.5 PIE *meh₁-ns- ‘moon, month’
– 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-
nsis 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

via free access
156 lubotsky
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’.
9 Either phonetically, or analogically (see about Germanic below).

10 Cf. already Pokorny 1959: 731: “mēnōt, Gen. mēneses, woraus mēnes-, mēns-, mēs-, mēn-”.
11 A similar case would be the PGerm. n-stem *nefan- ‘nephew’ on the basis of the nom.sg.
*nefō < PIE *nepōt.
12 Type *nepōt with secondary lengthening after the model of the stems in a resonant.

via free access
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?
3.1 PIE *dhuens- ‘to scatter, sprinkle’
– 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+).

via free access
158 lubotsky
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’.
3.2 PIE *ḱens- ‘to declare’
– 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.

via free access
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.
3.3 PIIr. *srans-
– 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.

via free access
160 lubotsky
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-
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.

via free access
Cheung, J. 2007. Etymological Dictionary of the Iranian Verb. Leiden–Boston: Brill.

Debrunner, A.; J. Wackernagel 1930. Altindische Grammatik III. Deklination der Nomina,
Zahlwörter und Pronomina. Göttingen: Vandenhoeck & Ruprecht.
Derksen, R.H. 2008. Etymological Dictionary of the Slavic Inherited Lexicon. Leiden–
Boston: Brill.
Collinder, B. 1960. Comparative Grammar of the Uralic Languages. Stockholm: Almqvist
& Wiksell.
Garcia-Ramon, J.L. 1993. Lat. cēnsēre, got. hazjan und das idg. Präsens *ḱéns-e-ti (und
*ḱn̥ s-éi̯e-ti?) “verkündigt, schätzt”, Stativ *ḱn̥ s-eh₁- “verkündigt, geschätzt sein / wer-
den”. In: G. Meiser (ed.), Indogermanica et Italica. Festschrift für Helmut Rix zum 65.
Geburtstag. Innsbruck: IBS, 106–130.
Gąsiorowski, P. 2010. The Germanic reflexes of PIE *-sr- in the context of Verner’s Law.
In: B. Nielsen Whitehead; T. Olander; B.A. Olsen; J.E. Rasmussen (eds.), The Sound
of Indo-European. Phonetics, Phonemics, and Morphophonemics. Copenhagen: Mu-
seum Tusculanum, 117–128.
Boston: Brill.
Kloekhorst, A. 2010. Initial stops in Hittite (with an excursus on the spelling of stops in
Alalaḫ Akkadian). Zeitschrift für Assyriologie 100, 197–241.
Kortlandt, F. 1978. I.-E. palatovelars before resonants in Balto-Slavic. In: J. Fisiak (ed.),
Recent Developments in Historical Phonology. The Hague: Mouton, 237–243.
Kortlandt, F. 1985. Long vowels in Balto-Slavic, Baltistica 21/2, 112–124.
217–227.
Rodopi.
Kortlandt, F. 2013. Palatovelars before syllabic resonants: another look. Baltistica 48/1,
13–17.
Kroonen, G. 2013. Etymological Dictionary of Proto-Germanic. Leiden–Boston: Brill.
Kubisch, P. 2012. Paippalāda-Saṃhitā, Kāṇḍa 20, Sūkta 1–30. Kritische Edition, Überset-
zung, Kommentar. Ph.D. dissertation, Bonn University.
Lubotsky, A. 1997. The Indo-Iranian reflexes of PIE *CRHUV. In: A. Lubotsky (ed.),
Sound Law and Analogy. Papers in honor of Robert S.P. Beekes on the occasion of his
60th birthday. Amsterdam–Atlanta: Rodopi, 139–154.
Lubotsky, A. 2011. The origin of Sanskrit roots of the type sīv- ‘to sew’, dīv- ‘to play dice’,
with an appendix on Vedic i-perfects. S.W. Jamison; H.C. Melchert; B. Vine (eds.),
Proceedings of the 22nd Annual UCLA Indo-European Conference. Bremen: Hempen,
105–126.
Matasović, R. 2014. Slavic Nominal Word-Formation. Proto-Indo-European Origins and
Historical Development. Heidelberg: Winter.

via free access
162 lubotsky
Mayrhofer, M. 1986–1996. Etymologisches Wörterbuch des Altindoarischen (EWAia). Hei-

delberg: Winter.
Mikola, T. 1988. Geschichte der samojedischen Sprachen. In: D. Sinor (ed.), The Uralic
Languages. Description, History and Foreign Influences. Leiden–New York: Brill, 219–
263.
Nikolaev, A. 2010. Issledovanija po praindoevropejskoj imennoj morfologii. St. Petersburg:
Nauka.
Oldenberg, H. 1912. Ṛgveda: Textkritische und exegetische Noten. Siebentes bis zehntes
Buch. Berlin: Weidmannsche Buchhandlung.
Pokorny, J. 1959. Indogermanisches etymologisches Wörterbuch. Bern–München: Fran-
cke.
Schindler, J. 1972. Das Wurzelnomen im Arischen und Griechischen. Ph.D. dissertation,
Würzburg University.
Zair, N. 2012. The Reflexes of the Proto-Indo-European Laryngeals in Celtic, Leiden–
Boston: Brill.

via free access
chapter 12
Headedness in Indo-Uralic
Rosemarie Lühr
In substantiating the claim of a relationship between Indo-European and Ural-

ic, a linguistic domain which has not been analyzed sufficiently is syntax. This is
surprising as a special word order, namely SOV, is assumed both for Uralic and
for Indo-European (Raun 1988: 569). Furthermore, SOV belongs not only to the
linguistic universals but also to the assumed implicational type, in so far as with
overwhelmingly greater-than-chance frequency, languages with normal SOV
order are postpositional. Also a modifier-before-headword order and a genitive
noun phrase before the possessor is connected to the SOV type, whereby the
underlying concept of all these relations is headedness. It is the head direction-
ality parameter we are dealing with here. In the following, phrases representing
head directionality in the oldest Indo-European languages Hittite and Vedic are
compared. For comparison purposes the earliest Uralic language documented
in writing, Old Hungarian, is used, since this language shows traces of head
finality in syntax.
This article is organized as follows. Firstly, we give an overview of the differ-
ent head phrases in Old Hungarian. Secondly, the comparison with correspond-
ing structures in Hittite and Vedic follows. Thirdly, the function of the subject in
Hittite and Vedic is examined. The reason for this is that in Hungarian a change
from Proto-Hungarian SOV to a Topic Focus Verb X* order occurred, while SOV
is maintained in Khanty and Mansi. Therefore our question is whether also in
Hittite and Vedic there are traces of the so called discourse configurational type.
The data for the Indo-European part comes mainly from our DFG-supported
projects “Information Structure in Older Indo-European sentences” and “Infor-
mation Structure in Complex Sentences—Synchronic and Diachronic”. For the
Uralic part the relevant literature is used.
1 Old Hungarian Phrases with Adpositions, Noun Phrases

with Adjectives and Genitives and Relative Structures
The basic word order of Proto-Hungarian is reconstructed on the basis of

archaic constructions of early Old Hungarian documents, and on the basis of
corresponding constructions of present-day Khanty and Mansi (Ostyak and

via free access
164 lühr
Vogul). These languages are thought to be most closely related to Hungar-

ian within the Ugric languages (Honti 1979: 7–19; 1998a: 353–355; 1998b: 179–
181).
Turning to Old Hungarian, one striking example for an older SOV order is the
following (Kiss 2013; Marcantonio 1985): whereas Old Hungarian already had
a general accusative case ending (the morpheme -t), the first surviving Hun-
garian codices, including books of the Bible (translated in 1416–1435, copied in
1450 and 1466), sporadically still contain a non-finite SOV construction whose
object bears no accusative case. The caseless object can be definite:
(1) Munich Codex, St Matthew 1,20

[ợ è gondoluan] yme vrnac angala ièlenec nèki
he this-Ø thinking Io Lord’s angel appeared he-DAT
‘while he thought on these, Io, the angel of the Lord appeared unto him.’
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):
(2) St Matthew 4,20, Munich C. (1416/1466)

És azok [legottan hálójok meghagyván] követék
and they immediately net-3PL-Ø PRT-leaving follow-PAST-3PL
őtet
him
‘And, straightway leaving their net, they followed him.’
Further evidence for an old SOV type comes from the verb-auxiliary order in
Old Hungarian. The auxiliary always immediately follows the verb:
(3) Funeral Sermon and Prayer (1192–1195)

es odu-tt-a vol-a neki paradisumut hazoa
and give-PERF-3SG be-PAST he-DAT Paradise-ACC house-for
‘and had given him Paradise for a house.’
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:

via free access
headedness in indo-uralic 165
(4) Munich C. 60rb (1416/1466)

az vy bor vý to̗ mlo̗ cbè èrèźt-ènd-o̗
the new wine new leather.bottle-PL-ILL pour-MOD-PART
‘new wine is to be put into new bottles.’
For a possessor preceding the possessum:

ig fa gimilcetvl
one tree fruit-3SG-from
‘from the fruit of one tree’
Also relative structures are preposed:1
(6) Kazinczy C. (1526–1541), p. 34

es ueǵed az [neko̗ d zo̗ rzo̗ ttem] Coronat
and take-IMP-2SG the you-DAT obtain-PASTPART-1SG crown-ACC
‘and take the crown I obtained for you.’
However, in (7) the relative clause is postposed. It is a non-restrictive one refer-

ring to a personal name:
(7) Funeral Sermon and Prayer (1192–1195) (Bácskai-Atkari 2013)

Eſ uimagguc || ſzent peter urot. Kinec odut hotolm
and pray-IMP1PL saint Peter lord-ACC who-DAT given power
ovdonia. eſ ketnie
bind-INF.3.SG. and unbind-INF.3.SG.
‘and let us pray to the lord Saint Peter, to whom the power was given to
bind and to unbind’
Finally, the adpositional phrase is head-final in Hungarian. Hungarian has post-

positions, not prepositions:

ív uimadsaguc-mia
they prayer-3PL-because.of
‘because of their prayer’.
1 There are also prenominal non-finite relative clauses in present-day Khanty (Nikolaeva 1999:
79; Bácskai-Atkári & Dékány 2014: 44; Csepregi 2012).

via free access
166 lühr
2 Head Phrases in Indo-European Languages
2.1 Phrases with Adpositions

2.1.1 Hittite
To continue with adpositions in the older Indo-European languages only a
short remark about Hittite is necessary for there is much research on this topic
(Brosch 2013; 2014a; 2014b; Melchert 2009: 613; Zeilfelder 2001: 224–230; Starke
1972). For the question whether Anatolian possessed both prepositions or post-
positions, Brosch (2013: 399) discusses the Lycian and Luwian evidence. While
Lycian has prepositions, in Luwian postpositions and prepositions are docu-
mented; for a preposition cf.
(9) KUB 35.29 iii 29′ (CLuw./NS)

a=duw[=an] annān patānza dūwandu
CONN=3SG.D/L=3SG.ACC.C under foot: D/L.PL put: IMP.3PL.ACT
‘They shall put it under his feet.’
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:
d. andan É-ri ‘in the house’ (with dative/locative)
Brosch (2013: 398) considers andan here as a preposition.
(11) KBo 6.2. iv 54 (OS)

andan=(m)a É-ri kuit
inside=CONN house: D/L.SG what(ever): NOM.SG.N
harkzi
get lost: PRS.ACT3SG
‘But what(ever) got lost inside the house.’

via free access
But respecting the context andan bears a constrastive stress. It is a top-

icalized contrastive topic in the function of a local adverb. Thus, in Hittite
there would be only postpositons. Nevertheless if the mentioned coexistence of
prepositions and postpositions in Luwian represents an old status of Anatolian,
in Lycian prepositions must have been generalized and in Hittite postpositions
(Brosch 2013: 39, 154).
In this case, Anatolian is an uncertain testimony for original postpositions
needed for a possible head final type of Indo-European.
2.1.2 Vedic
In Vedic prepositions besides postpositions can be found; for prepositions
cf.:2
(12) RV 1,30,19 (Speyer 1896: 24)

pári dyā ́m anyád
around: PREP heaven(M): ACC.SG other: NOM.SG.N
īyate
revolve: PRS.IND.MED/PASS3SG
‘The other [chariot wheel] revolves around the sky.’
(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.’
(14) RV 8,20,11, cf. Viti 2015: 61f.

ví bhrājante rukmā ́so ádhi
PFX glitter: PRS.IND.MED3PL jewel (M): NOM.PL on: PREP
bāhúṣu
arm(M): LOC.PL
‘Their jewels glitter upon their arms.’
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.

via free access
168 lühr
And for a postposition:
(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 Noun Phrases with Adjectives

The next topic is the position of attributive adjectives. According to language
typology, preposed adjectives are to be expected if a language has OV ordering
and, vice versa, postposed adjectives in the case of VO ordering.
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:
(16) Muwatalli (CTH 381, 1, 15)

DINGIR.LÚMEŠ DINGIR.MUNUSMEŠ ḫu-u-ma-an-te-eš ḪUR.SAGMEŠ
god: PL goddess: PL all: NOM.C.PL mountain: PL
ÍD MEŠ ŠA KUR URU.GIŠGIDRU-ti ḫu-u-ma-an-te-eš
river: PL of land Hatti all: NOM.C.PL
‘all the gods (and) goddesses, all the mountains (and) rivers of the land of
Ḫatti’.
Also Hittite dapiant- ‘all, entire’ regularly follows its head noun.
(17) KUB 5.1 iii 62 (Hoffner & Melchert 2008: 271)

LÚMEŠ Ga-aš-gaḪIA-ma-an-kán da-pí-an-te-eš GAM UGU RA-an-zi
‘But all the Kaska men will strike it (the city) up from below.’

via free access
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:
(19) Muwatalli (CTH 381 1,7–8)

NINDA.Ì.E.DÉ.A DUGÚTUL šu-u-wa-an me-[m]a-al
pound cake pot full: ACC.N.SG coarse meal(N): ACC.SG
=ma DUGDÍLIM.GAL šu-u-wa-an
but bowl full: ACC.N.SG
‘a pot full of pound cake, but a bowl full of coarse meal.’
A relict of the former prenominal position of ḫūmant- could be its preposition

with body parts:
(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:
(21) KUB 36.90 rev. 39

ḫu-u-ma-an-da-[az KU]R.KURMEŠ-za e-ḫu
‘come from every land’
is surely due to information structure. Here, ḫūmandaz is a contrastive topic

and is pragmatically highlighted. Returning to postposed ḫūmandaz, I argue
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).

via free access
170 lühr
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’
(23) a. Muwatalli (CTH 381 3,4)

dU ḫu-la-aš-ša-aš-ši-iš
Storm God belonging to the town Ḫulašša: NOM.C.SG
‘O Storm God of Ḫulašša’
b. Telepinu (CTH 19 1,30)

[ERÍNMEŠ] ḫur-lu-uš
troops: PL Hurrian: ACC.PL
‘Hurrian troops’
Comparing other languages with regard to the position of relational adjectives

(Bosque & Picallo 1996), it is worth noting that in languages which distinguish
word order of attributive adjectives with respect to what could be described as
their descriptive content, qualifying adjectives occur in prenominal and rela-
tional adjectives in postnominal position. Such a language is Polish for example
(Wągiel 2014).
As there is an essential distinction between the two adjectival classes, the
semantic difference could also have had an impact on word order in Hittite; cf.
(24) Muwatalli (CTH 381 3,46)

i-da-lu-uš me-mi-aš
evil: NOM.C.SG word(C): NOM.SG
‘the evil word’
As regards the position of adjectives, Hittite is broadly in line with the SOV-
type. Apart from some postposed relational adjectives and predominantly
5 Further research is needed.

via free access
postnominal hūmant- and dapiant-, being instances of a former quantifier

floating, in Hittite attributive adjectives precede their head noun. This is also
valid for numerals without exception (Hoffner & Melchert 2008: 165).
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:
(25) RV 4.42.8 (Royal Consecration 27)

té āsan saptá r̥ ṣ́ ayaḥ
the: NOM.M.PL be: IPF.IND.ACT3PL seven: adj.num R̥ ṣi(M): NOM.PL
daurgahé badhyámāne
Daurgaha(M): LOC.SG being captivated: LOC.M.SG
‘They were the seven R̥ ṣis, when the son of Durgaha was captive.’
(26) RV 10.86.14 (Indra and his monkey)

ubhā ́ kukṣī ́ pr̥ ṇanti me
both: ACC.M.DU cheeks(M): ACC.DU fill: PRS.IND.ACT3PL I: DAT.SG
‘They fill both my cheeks’
(27) RV 10.95.16 (Urvashi)

yát vírūpā ácaram
when in another shape: NOM.F.SG live: IPF.IND.ACT1SG
mártyesu ávasam rā ́trīḥ
mortal(M): LOC.PL spend: IPF.IND.ACT1SG night(F): ACC.PL
śarádaḥ cátasraḥ
autumn(F): ACC.PL four: ACC.F.PL
‘When I lived in another shape among the mortals, I spent the nights for
four autumns.’
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).

via free access
172 lühr
As quantifier floating with numerals is a living process in Vedic, quanti-

fiers as ‘all’, ‘every’ should be floating quantifiers, too. And indeed, viśva- occurs
before and behind its head noun:
(28) RV 4.42.1 (Royal Consecration)

máma dvitā ́ rāṣṭ(a)rám kṣatríyasya
I: GEN.SG still rule(N): ACC.SG king(M): GEN.SG
viśvā ́yoḥ víśve amr̥ t́ āḥ yáthā
swaying all life: GEN.M.SG all NOM.M.PL immortal: NOM.M.PL as
naḥ
I: GEN.PL
‘All immortals still (follow) my rule, that of the king as mine who is invig-
orating all life.’
(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:
(30) a. The only navigable river is to the north.

b. The only river navigable is to the north.
In (30a) “regularly navigable” is meant, in (30b) “navigable at present”. Reading

(30a) represents an individual-level predicate, reading (30b) a stage-level pred-
icate, where individual-level corresponds to an intrinsic/permanent property
and stage-level to a temporary property. For language typology it is of inter-
est that in SVO-languages the modification to the left of the noun is thought
to reflect nominal characteristics and modification to the right verbal char-
acteristics, namely in an iconic manner (Vogel 1996: 207; Lühr 2002a; 2002b;
2005).
Actually, in Vedic examples for prenominal modification with individual-
level meaning and those for postnominal modification with stage-level mean-
ing are documented:

via free access

nákiḥ mā daívyam sáhaḥ
never I: ACC.SG divine: NOM.N.SG power(N): NOM.SG
varate ápratītam
lock up: AOR.SUBJ.MED3SG irresistibel: ACC.M.SG
‘No divine power will lock up me, the irresistible.’
vs.
(32) RV 3.33.12 (River flood)

átāriṣuḥ bharatā ́ḥ
get across: AOR.IND.ACT3PL Bharata(M): NOM.PL
gavyávaḥ sám
desiring cows: NOM.M.PL together
‘The Bharatas got across together desiring cows.’
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)
(33) RV 10.86.7 (Indra and his monkey)

uvé amba sulābhike
see: PRS.IND.MED1SG woman(F): VOC.SG easy to be won: VOC.F.SG
yáthā iva aṅgá bhaviṣyáti
how somehow PART be: FUT.IND.ACT3SG
‘I see how it will be somehow, woman, who is easy to be won.’
But also other syntactic connections:
(34) RV 10.52.5 (Agni)

ā ́ vaḥ yakṣi amr̥ tatvám
here you: DAT.PL offer: AOR.INJ.MED1SG immortality (N): ACC.SG
́
suvīram
rich in heroes: ACC.N.SG
‘Here I offer you immortality which is rich in heroes.’
Another adjective use is documented in (35). The quantifier éka- ‘one’ appears
behind its head noun und is surely stressed.

via free access
174 lühr
(35) RV 10.51.1 (Agni)

víśvāḥ apaśyat bahudhā ́ te
all: ACC.F.PL see: IPF.IND.ACT3SG frequently you: GEN.SG
agne jā ́tavedaḥ tan(ú)vàḥ
Agni(M): VOC.SG Jātavedas(M): VOC.SG manifestation (F): ACC.PL
deváḥ ékaḥ
deity(M): NOM.SG one: NOM.M.SG
‘One deity, o Agni Jātavedas, saw all your manifestations frequently.’
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 Noun Phrases with Genitives

The next word order universal concerns the position of the genitive. When
adjectives are placed behind the noun they modify, also genitives should ap-
pear in this position and the other way round when a language has preposed
ones.
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.
(36) Muwatalli (CTH 381 1,6)

35 NINDA.GUR4.RA tar-na-aš
35 thick bread tarna-measure(C): GEN.SG
‘35 thick breads weighing each a tarnas’
A construction where a numeral refers to a measure word like in the following

Greek example is not documented in Hittite.

via free access
(37) Od. 4.129 (Viti 2008: 219)

δέκα δὲ χρυσοῖο τάλεντα
‘Ten talents of gold’
In Hittite, the denotation of what is counted immediately follows the numeral,

which is why the genitive tarnaš moves to the position behind the head noun.
Thus, it can be said that the word order is iconic, here. It follows Behaghel’s first
law according to which elements that belong close together intellectually will
also be placed close together: the number is connected to the expression for
the counted entity.
Also the genitive of material appears postposed:
(38) a. Ritual (CTH 443 1,1–5)

EME iš-na-aš
tongue clay(C): GEN.SG
‘tongue (made) of clay’
b. StBoT 8 i 22′ (Yoshida 1987: 72)

2 DḪa-an-ta-a-an-ta-še-pu-uš … GI[(Š-aš)]
‘2 DḪantašepa-deities … of wood’
Cf. the following examples from Homeric Greek:
(39) Il. 11.24 οἴμοι κυάνοιο ‘stripes of cyanos’

Od. 4.124 δέπας οἴνοιο ‘a cup of wine’ (Viti 2008: 219)
To explicate postposition of the designation of the material here, the kind of

reference has to be noted. For this purpose we compare possessive nominal
phrases with a noun as possessor. The possessor precedes always the posses-
sum.
(40) Edict of Telipinu i 66 (KUB 11.5 obv. 8′)

ad-da-aš e-eš-ḫar-še-et
father(C): GEN.SG blood(N): NOM.SG=his
‘the blood of the father’
(41) KUB 8.41 ii 7′

DIŠKUR-na-aš ša-ša-an-ti-iš-ši
Storm God(C): GEN.SG concubine(C): DAT./LOC=his
‘to the concubine of the Storm God’

via free access
176 lühr
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.:
(42) KBo 17.36+25.54 + (= StBoT 25 Nr. 54) iv 5′

[me-e-ma-]al še-ep-pí-da-aš
groats š.-grain(N): GEN.SG
‘groats from š.-grain’.
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:

máma … rāṣṭ(a)rám kṣatríyasya
I: GEN.SG rule(N): ACC.SG king(M): GEN.SG
‘my rule, that of the king’.
As regards the position of substantival genitives, a comparison of the num-

ber of records in Vedic shows that preposed ones are more often documented
than postposed ones. The proportion is 3 to 2. As it could be supposed that
preposition of the genitive is the normal word order, postposition must be
explained. All categories of adnominal genitives, the possessive, subjective,
objective, partitive genitive and the genitive of content can appear postnomi-
nally. But, admittedly, a ratio for postposition exists only for a part of the geni-
tival data.
Fixed word orders appear sometimes with kinship terms. The head noun
precedes:
(44) RV 2.028.03c putrā aditer

RV 4.042.04c putró áditer
RV 7.041.02a putrám áditer

via free access
RV 7.060.05c putrā ́ áditer

RV 8.018.05a putrā ́so áditer
RV 10.072.08a putrā ́so áditer
RV 10.185.03a putrā ́so áditeḥ
RV 9.069.03a naptīŕ áditer
Also in vocative constructions with an objective genitive the head noun ap-
pears first:
(45) RV 3.30.19; 10.47.1 vasupate vásūnām ‘lord of wealth’

RV 10.112.10 vasupate sákhīnām ‘lord of treasures’
Often the word order is inversed by hyperbaton. In (46) the head noun is in
front of the possessive genitive:
(46) RV 10.52.5 (Agni 37)

ā ́ bāh(u)vóḥ vájram índrasya
in arm(M): LOC.DU Vajra(M): ACC.SG Indra(M): GEN.SG
dheyām
put: AOR.OPT.ACT1SG
‘I would like to put the Vajra in Indra’s arms.’
Another genitival structure is represented in (47): three genitives exhibit extra-

position; this means that the heavy constituent appears to the right of its
canonical position.
(47) RV 1.165.15 (Marut 44)

eṣá vaḥ stómaḥ marutaḥ
this: NOM.M.SG you: DAT.PL praise(M): NOM.SG Marut(M): VOC.PL
iyám gīḥ́ māndār(i)yásya
this: NOM.F.SG hymn(F): NOM.SG Māndāriya: GEN.SG
mān(i)yásya kāróḥ
Māna(M): GEN.SG singer(M): GEN.SG
‘this is praise to you, O Maruts, this hymn of the singer Māndāriya, Māna’s
son.’
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-

via free access
178 lühr
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.
2.4 Phrases of Relative Clauses

Most Hittite relative clauses precede the main clause. There are two types, a
restrictive and a free relative clause in the function of a determiner phrase
(Lühr 2001; Hoffner & Melchert 2008: 424; Ott 2001).
In later Hittite also postposed relative clauses appear; cf. the following much
quoted sentence: the relative clause is a non-restricted one:
(48) Ullikummi (CTH 345 I 1, 5–6) (Garrett 1994: 47)

dKu-mar-bi-iš GA[LGA]-tar ZI-ni [kat]-ta-an da-aš-ke-ez-zi
Kumarbi: NOM wisdom: ACC mind: LOC into take=PRS3SG
UDK[AM-a]n ku-iš LÚ [ḪUL]-an šal-la-nu-uš-ke-ez-zi
day: ACC who: NOM being: ACC evil: ACC cause to grow: PRS3SG
‘Kumarbi takes wisdom into his mind, who raises the day as an evil
being.’
Clackson (2007: 175) understands the relative clause in (48) as an afterthought

added to the main clause, but as there is reference to a personal name, hence
to a semantic definite (Loebner 1985), the interpretation as appositive rela-
tive clause is more obvious; cf. the postposed corresponding structure in Vedic
referring to a personal name as well:
(49) RV 5.36.1 (Hettrich 1988: 689)

sa ā ́ gamad índro
he: NOM.M.SG PFX come: AOR.SUBJ.ACT3SG Indra(M): NOM.SG
yó vásūnāṃ cíketad
who: NOM.M.SG goods(N): GEN.PL know: PF.SUBJ.IND3ACT
dā ́tuṃ dā ́mano rayīṇā ́m
give: INF gift(M): GEN.PL treasure(M): GEN.PL
‘Indra may come to us, who knows rightly to give treasures to give
riches.’
But concerning the ratio of restrictive and non-restrictive relative clauses in

Hittite on the whole it is true that in this language restrictive relative clauses are
much more common than non-restrictive ones, while in Vedic non-restrictive
relative clauses outweigh restrictive relative clauses. The ratio is approximately

via free access
4 : 3, whereby in the case of non-restrictive relative clauses the order of matrix

clause—relative clause outnumbers the reverse order more than twice (Avery
1881: lxiv–lxvi; Hettrich 1988: 680; Holland 1991: 33; Lehmann 1984: 228 f.; for
Latin cf. Clackson 2007: 175; for Greek cf. Probert 2015). This shows that the dis-
tinction between background information and new information provided by a
non-restrictive and a restrictive relative clause respectively has effects on the
order of clauses.
3 The Function of the Subject in Hittite and Vedic Compared

to Proto-Uralic
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
(50) Proto-Hungarian → Old Hungarian

subject/topic → topic
object/focus → focus
right-dislocated elements → in situ arguments
↓↓↓ ↓↓↓
SOV → topic focus V X*
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.

via free access
180 lühr
table 12.1 Subjects and topics in Hittite
Text Muwatalli Ritual Telepinu Hitt. total
Subjects in MC total 116 200 53 369

Covert subjects in MC total 89 150 27 266
Subject in 1. position of MC 4 7 6 17
Subject in 2. position of MC 4 7 8 19
Subject final in MC 0 2 0 2
Subject second-to-final in MC 3 4 5 12
Topics as Subject in MC 11 10 9 30
thereof continuing topic 4 3 5 12
thereof shifting topic 6 6 4 16
thereof contrastive topic 1 1 0 2
covert Topic-Subjects in MC 21 58 18 97
thereof continuing topic 13 44 15 72
thereof shifting topic 7 12 3 22
thereof contrastive topic 1 0 0 1
table 12.2 Subjects and topics in Old Indic
Language Vedic Sanskrit OI. total
Subjects in MC total 1177 362 1539

covert subjects in MC total 451 93 544
Subject in 1. position of MC 279 75 354
Subject in 2. position of MC 114 60 174
Subject final in MC 51 28 79
Subject second-to-final in MC 133 48 181
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

via free access
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:
Hittite Verb-final Verb-first Verb-second Verb-middle
483 10 6 7
On the contrary, the relations are ambiguous in Old Indic, especially in Vedic:
R̥ gveda Verb-final Verb-first Verb-second Verb-middle
169 64 71 118
Pancatrantra Verb-final Verb-first Verb-second Verb-middle
70 8 13 9
Thus, Hittite is a verb-final language, Vedic probably not.

The conclusion is: if there really existed a common Proto-Indo-Uralic lan-
guage, concerning word order its Indo-European representative must have
been of the Hittite type and not of the Vedic one, because this language clearly
is a mixed type relating to head directionality.
Actually, as the World Atlas of Language Structures (ONLINE) (http://wals
.info/) shows, the SOV-type is the most common in the languages of the world.

via free access
182 lühr
table 12.3 Distribution of the types of word order in WALS
Value Representation
Subject-object-verb (SOV) 565

Subject-verb-object (SVO) 488
Verb-subject-object (VSO) 95
Verb-object-subject (VOS) 25
Object-verb-subject (OVS) 11
Object-subject-verb (OSV) 4
Lacking a dominant word order 189
Total: 1377
It may be that convergent head directionality structures can be used as proof

of a common proto-language for Uralic and Indo-European with Hittite as
the main exponent of the Indo-European branch. However, more evidence is
needed.
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.

via free access
Brosch, C. 2014b. Räumlichkeit in Zentralanatolien. In: S. Kutscher; D. Werning (eds.),

On Ancient Grammars of Space. Linguistic Research on the Expression of Spatial Rela-
tions and Motion in Ancient Languages. Berlin: Mouton de Gruyter, 1–44.
Casaretto, A. 2011. Syntax und Wortarten der Lokalpartikeln des R̥ gveda. XII: ví. His-
torische Sprachforschung 124, 134–177.
Casaretto, A. 2014. Zum Verhältnis von Kasusfunktion und Lokalpartikel im R̥ gveda am
Beispiel des Akkusativs. Historische Sprachforschung 125, 49–67.
Clackson, J. 2007. Indo-European Linguistics. An Introduction. Cambridge: Cambridge
University Press.
Fitzpatrick, J.M. 2006. Syntactic and Semantic Routes to Floating Quantification. Ph.D.
dissertation, MIT.
Csepregi, M. 2012. Participiális jelzős szerkezetek két hanti nyelvjárásban. Nyelvtudo-
mányi Közlemények 108, 61–94.
Francia, R. 2011. La posizione degli aggettivi qualificativi nella frase ittita. In: O. Carruba;
W. Meid (eds.), Anatolisch und Indogermanisch. Innsbruck: IBS, 81–91.
Garrett, A. 1994. Relative clause syntax in Lycian and Hittite. Die Sprache 36, 29–
96.
Hoeksema, J. 1996. Floating quantifiers, partitives and distributivity. In: J. Hoeksema
(ed.), Partitives: Studies on the Syntax and Semantics of Partitive and Related Con-
structions. Berlin–New York: Mouton de Gruyter, 57–106.
Holland, G. 1991. Definiteness and relativization. In: J.P. Brereton; S.W. Jamison;
M.M. Deshpande (eds.), Panels of the VIIth World Sanskrit Conference. Volume IV:
Sense and Syntax in Vedic. Volume V: Pāṇini and the Veda. Leiden: Brill, 22–32.
Honti, L. 1979. Characteristic features of Ugric languages (observations on the question
of Ugric unity). Acta Linguistica Academiae Scientiarum Hungaricae 29, 1–26.
Honti, L. 1998a. ObUgrian. In: D. Abondolo (ed.), The Uralic Languages. London–New
York: Routledge, 327–357.
Honti, L. 1998b. Die Ob-Ugrischen Sprachen. In: D. Sinor (ed.), The Uralic Languages.
Description, History and Foreign Influences. Leiden–New York: Brill, 147–196.
Kiss, K.É. 2013. From Proto-Hungarian SOV to Old Hungarian Top Foc V X*. Diachronica
30/2, 202–231.
Kobuchi-Philip, M. 2003. Syntax and semantics of the Japanese floating numeral quan-
tifier and its implications for the theory of floating quantifiers. Rivista di Gramma-
tica Generativa 28, 57–70.
Laroche, E. 1982. Epithètes et prédicats en hittite. In: J. Tischler (ed.), Serta Indogerma-
nica. Festschrift für Günter Neumann zum 60. Geburtstag. Innsbruck: IBS, 133–136.
Lehmann, W.P. 1973. A structural principle of language and its implications. Language
49/1, 47–66.

via free access
184 lühr
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.),

via free access
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.

via free access
chapter 13
Indo-Uralic, Indo-Anatolian, Indo-Tocharian*

Michaël Peyrot
In this paper, I intend to illustrate the relevance of the Tocharian branch of

Indo-European for questions concerning the Indo-Anatolian hypothesis and
the possible macro-relationship of Indo-European with Uralic. To this end, I
discuss the relevance for the Indo-Anatolian hypothesis of the possibility that
the Tocharian branch was the second to split off after Anatolian, which I call
the “Indo-Tocharian” hypothesis; the case of the Anatolian and Tocharian verb
for ‘drink’ and its possible Uralic cognate; and the Anatolian and Tocharian
m-interrogative, which has a neat parallel in Uralic. As will become clear, the
results of these discussions are of uneven value, and the main aim of this paper
is principally of a methodological nature.
1 Indo-Uralic, Indo-Anatolian, Indo-Tocharian
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.
© michaël peyrot, 2019 | doi:10.1163/9789004409354_014

This is an open access chapter distributed under the terms of the CC-BY-NC-ND 4.0 License.
via free access
indo-uralic, indo-anatolian, indo-tocharian 187
figure 13.1 Tree of the Indo-European family; after Ringe,

Warnow & Taylor 2002: 87; simplified
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

via free access
188 peyrot
figure 13.2 Tree of the Indo-European family with labels for

the Indo-Anatolian, Indo-Tocharian and Indo-
Italo-Celtic nodes
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.
1) Hitt. mer-zi / mar- ‘disappear’ < *mer-
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.

via free access
2) Hitt. zīk / tu- ‘you’ < PAnat. *tiH / *tu-
In the other Indo-European languages the vowel of the allomorphs*tiH / *tu-

was levelled as *tuH / *tu-.
Tocharian goes together with the other non-Anatolian languages: Tocharian
A tu, B tuwe < *tuH-om.
3) Hitt. šāḫ-i ‘fill up, plug, stuff’ < *seh₂-
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
According to Kloekhorst, the Anatolian words for ‘daughter’ point to an origi-

nal paradigm *dhuégh₂tr, *dhugh₂térm. After Anatolian split off, this paradigm
was simplified to become the *dhugh₂tḗr, *dhugh₂térm known from the non-
Anatolian languages.
The Tocharian words for ‘daughter’ are A ckācar and B nom.sg. tkācer, obl.sg.
tkātär. The initial ck- of Tocharian A ckācar is not regular. In theory, it could
replace *cukācar < *t’ẃəkacer < *dhuegh₂t-, with elimination of the *-u- after
the non-nom.sg. cases with *tk- < *təkat- < *dhugh₂t- (this is what I suggested
apud Kloekhorst 2011: 241). However, the evidence is weak, and ckācar more
probably derives from earlier *tkācar through distant assimilation. In this lat-
ter, more probable case, Tocharian goes together with the other non-Anatolian
languages.
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

via free access
190 peyrot
‘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.’).
6) Hitt. mimma-i / mimm- ‘refuse, reject’ < *meh₁-
According to Kloekhorst, the prohibitive negation *meh₁ of the non-Anatolian

languages derives from the imperative of the verb *meh₁- still preserved as such
in Anatolian.
Tocharian clearly goes together with the non-Anatolian languages: Toch. AB
mā ‘not’.
7) The Anatolian words for ‘horse’ < *h₁eḱu-
Kloekhorst reconstructs the Anatolian words for ‘horse’ as a u-stem, from

*h₁eḱu-. The thematic noun *h₁eḱuo- found in the non-Anatolian languages is
the result of a later development.
Tocharian clearly goes together with the non-Anatolian languages: Tochar-
ian A yuk ‘horse’, B yakwe ‘id.’ < *h₁eḱuo-.
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 Indo-Tocharian and Indo-Uralic? The Case of the Verb ‘Drink’
One of the frequently cited Hittite-Tocharian matches is Hitt. eku-zi / aku-

‘drink’ ~ Toch.AB yok- ‘drink’ < *h₁egwh- (Pinault 2006: 93). Although Anatolian
and Tocharian are indeed the only two branches in which this verb is found,
and most other branches have reflexes of the more common *peh₃- ‘drink’, this

via free access
etymon is difficult to use as an argument for the Indo-Tocharian hypothesis.

Most importantly, even though Anatolian and Tocharian are the only branches
in which *h₁egwh- is attested, reflexes of this root are also found in Lat. ēbrius
‘drunk’ and Gr. νήφω ‘be sober’.
A lookalike of this Proto-Indo-European root is found in Uralic: compare
among others Fi. juo- ‘drink’, Norw. Sa. jukkâ-, -ǥ- ‘drink’, and Hu. iv- ‘drink’. In
the Uralisches Etymologisches Wörterbuch (Rédei 1988–1991: 103), this etymon is
reconstructed as *juγe- ( juke-). The correspondence between PIE *h₁egwh- and
the reconstruction *juγe- ( juke-) reminds of PIE *deh₃- ‘give’ ~ PU *toγe- ‘bring,
get, give’ (Rédei 1988–1991: 529). Kortlandt (1989: 83) explained this correspon-
dence assuming a Proto-Indo-Uralic preform *tagu-, which developed through
u-umlaut to PU *toγe- on the one hand, and with *gu > *h₃ to PIE *deh₃- on the
other. In the same vein, one might explain PU *juγe-, juke- from PIU *eku- with
u-umlaut, and the initial j- perhaps as a result of breaking. PIE *h₁egwh- would
derive from this reconstructed PIU *eku- through the change of *ku to *gwh.
Today, the reconstructions of the Uralisches Etymologisches Wörterbuch are
generally viewed as outdated. However, the more recent reconstructions of
Sammallahti (1988: 543, 550), PFU *toxi- ‘bring’ and PFU *juxi- ‘drink’, respec-
tively, have not changed this picture in any essential way. On the basis of these
reconstructions, it would still be possible to argue that PFU *toxi- and PIE
*deh₃- < PIU *tagu, and PFU *juxi- and PIE *h₁egwh- < PIE *eku- or *egu-.2
A more radical new reconstruction of the Uralic verb for ‘drink’ has been
proposed by Aikio (2002: 38–40): *ji̮xi.3 Obviously, the relevance of this revi-
sion is that no labial vowel is reconstructed anymore, which weakens the
comparison between the Proto-Indo-European and the Proto-Uralic roots con-
siderably. Revised reconstructions for Proto-Indo-Uralic are conceivable, for
instance *igu with -u > -i in Uralic, but the number of unproven sound laws
that has to be assumed increases, so that such reconstructions are hardly falsi-
fiable.
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.

via free access
192 peyrot
Aikio’s new reconstruction is based on 1) the inclusion of Proto-Samoyedic

*e̮- among the cognates, and 2) a different interpretation of the Permic and
Hungarian vocalism (on which see below).
The Proto-Samoyedic root *e̮- ‘drink’ is set up by Aikio on the basis of
*e̮r- ‘drink’ and *e̮kəl- ‘drink avidly’.4 According to him, the longer root *e̮kəl-
must be a derivative, and since the base is apparently *e̮-, the other root
*e̮r- must be a derivative as well. Indeed, so-called “augmentative” deriva-
tives in -r are quite well attested: they are frequently found in the individ-
ual languages and for Proto-Samoyedic Janhunen (1977) lists r-augmentatives
for *əm- ‘eat’, *cinɜ-/ci̮nɜ- ‘smell’, *jatə- ‘go’, *jäcə- ‘forge’, *ko- ‘see’ (as well as
*kont¹ə- ‘see’), *kot- ‘cough’, *kunə- ‘run away’, *mej- ‘make’ (*mir-), *nət- ‘scrape’,
*ńensə- ‘glide’, *ńim- ‘suck’, *pe- ‘seek’, *pit- ‘tan’, *sänə- ‘play’, *talä- ‘steal’, *te̮mta-
‘trade’ (*te̮mtəjr-) and *witɜ- ‘drink (water)’. Although the suffix -kəl is much
less frequent, it has parallels too. Janhunen reconstructs derivatives in -kəl
for *u- ‘swim’, *ü- ‘drag’, *je ‘heel’ (*je-kəl- ‘step’), *nek-/ne̮k- ‘pull’, *ni̮c- ‘tear’,5
*pən- ‘plait’, *pət- ‘dive’, *so- ‘scoop’, *tək- ‘hide’. Aikio then notes that Proto-
Samoyedic roots of the structure *(C)V usually continue PU *(C)Vxi and con-
nects the Finno-Ugric etymon previously set up as *juxi by Sammallahti (1988:
543).
The problem with this connection is that PSam. *e̮- suggests *i̮xi instead of
PU *juxi. As a solution, Aikio proposes that the protoform was *ji̮xi with spo-
radic loss of *j- in Samoyedic and a sporadic change of *i̮ to *u in Finno-Ugric.
The assumed loss of *j- in Samoyedic is in need of an exact conditioning, but
Aikio adduces a parallel that is convincing in itself and this development seems
acceptable. However, I have difficulties accepting his sporadic change *i̮ > *u in
Finno-Ugric. The reason is that his *ji̮xi invites to reconsider the reconstruction
of a number of other etyma, which in turn suggests that the change *i̮ > *u in
Finno-Ugric was not sporadic.
If Samoyedic needs *i̮xi or *ji̮xi and Finno-Ugric *juxi, it may at first glance
seem simpler to keep the PU reconstruction *juxi and assume a change of *uxi
to *i̮xi in Samoyedic. The reason why this is not possible is that there are good
examples for the correspondence PSam. *-u : PFU *-uxi, *-uji, which reflect PU
*-uxi, *-uji:
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.

via free access
– PSam. *tu- ‘row’, PFU *suxi- < PU *suxi-

– PSam. *u- ‘swim’, PFU *uji- < PU *uji- (Aikio 2002: 44, who reconstructs *j
instead of Sammallahti’s *x)
The following example of the correspondence PSam. *-u : PFU *-uwi < PU
*-uwi is more problematic, according to Aikio (2012: 247), because the *j should
not have disappeared in the Fi. cognate puo ‘anus’.
– PSam. *puə- ‘behind’, PFU ?*puji < PU ?*puwi
A similar corresponce is that between PSam. *-o and PFU *-uxi, *-uwi, *-uji. To
explain this correspondence, PU *o is reconstructed with raising to *u in an
open syllable before *i in Proto-Finno-Ugric (Sammallahti 1988: 486):
– PSam. *ńo- ‘pursue’, PFU *ńuxi- < PU *ńoxi (Aikio 2014a: 53)
– PSam. *so ‘mouth’, PFU *śuwi < PU *śowi- (Aikio 2002: 35, who reconstructs
*w instead of Sammallahti’s *x because of South Sami tjovve)
– PSam. *to ‘lake’, PFU *tuxi < PU *toxi
A further etymon to be added here is probably PSam. *tə- ‘bring’ and PFU
*tuxi- ‘bring’ (Sammallahti *toxi-). This root is not listed by Janhunen (1981)
because the correspondence is not regular. According to the established basic
sound correspondences (Janhunen 1981), there is no possible PU source form
that could yield the PSam. root. I assume that the PU form was *tuxi-, which
yielded *tu in Samoyedic. The *u of *tu was then weakened to *ə before *a in
the second syllable in the derivative *təta-, as in PSam. *kəpta- ‘extinguish’ < PU
*kupsa- (Sammallahti 1988: 484). Afterwards, the phonologically regular *ə of
the derivative spread to the underived root. Since only *u, not *o, is weakened
to *ə, PSam. *tə-, *təta- suggests *tuxi- for Proto-Uralic, not *toxi-.
The correspondence between PSam. *-e̮ and PFU *-uxi, *-uwi, *-uji is also
attested in more examples than just ‘drink’:
– PSam. *e̮- ‘drink’, PFU *juxi < PU *ji̮xi
– PSam. *le̮ ‘bone’,6 PFU *luwi < PU *li̮wi (Aikio 2002: 35, who reconstructs *w
instead of Sammallahti’s *x because of Mordvin lovaža with o and v)
– PSam. *je̮ ‘tree’,7 PFU *juxi < PU *ji̮xi8
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.

via free access
194 peyrot
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.

via free access
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.

via free access
196 peyrot
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-

via free access
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).

via free access
198 peyrot
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-.

via free access
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
In this paper, I have treated a number of independent problems that illus-

trate some of the methodological points relevant to the questions of the phy-
logenetic structure of the Indo-European language family and its supposed
genealogical relationship with the Uralic language family.
I have argued that the Indo-Anatolian hypothesis requires the next node
down to be meaningful. Assuming that the second branch to split off the Indo-
European protolanguage was Tocharian, all evidence for the Indo-Anatolian
node needs to be checked against this presumed “Indo-Tocharian” node in
order to see whether also Tocharian goes together with the other non-Anatolian
languages.
Since the Anatolian-Tocharian isogloss for *h₁egwh- ‘drink’ is not exclusive, it
can only be used as an argument for subgrouping if the meaning ‘get drunk’
found in the Greek and Latin reflexes is a common innovation of the other
branches. The resemblance of the root *h₁egwh- to Proto-Finno-Ugric *joxi
‘drink’ turns out to be only superficial because of Aikio’s addition of the Sam-
oyedic cognate *e̮-, which suggests a Proto-Uralic reconstruction *ji̮xi. Apart
from the meaning, this revised reconstruction shows no obvious similarity with
*h₁egwh-.
The Proto-Indo-European interrogative stem in *m- next to the well known
*kwi-, *kwe-, *kwo- has not received due attention in the literature, but can
nevertheless safely be reconstructed, and especially Tocharian provides strong
evidence for it, with for instance TB mäksu ‘which’, intsu ‘which’, TA äntsaṃ
‘which’, etc. The reconstruction of a closely parallel set of interrogative stems
for Proto-Uralic is uncontroversial; cf. for instance Fi. kuka ‘who’ next to mikä
‘what, which’ and Hu. ki ‘who’ next to mi ‘what’. Especially taken together with
other parallels in grammatical elements, the neat correspondence of two inter-
rogative stems in Proto-Indo-European and Proto-Uralic is further evidence in
favour of a genealogical relationship between the two families.

via free access
200 peyrot
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.
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.

via free access
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.
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-
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.

via free access
202 peyrot
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.

via free access
chapter 14
Proto-Indo-European *sm and *si ‘one’

Michiel de Vaan
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.

via free access
204 de vaan
table 14.1 PIE deictic *so/*to- and PIE anaphoric *ʔe/*i-
PIE deictic m n f PIE anaphoric m n f
Singular Singular
Nominative *so *tod *seʕ nom *ʔe *id *iʕ

Accusative *tom *tod *teʕm acc *im *id *iʕm
Oblique *to(sm)- *to(si)- obl *ʔe(sm)- *ʔe(si)-
Plural Plural
nom *toi *teʕ *seʕi? nom *ʔei *iʕ *iʕes

acc *tons *teʕ *teʕns acc *ins *iʕ *iʕns
obl *toi- *teʕ- obl *ʔei-
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

via free access
proto-indo-european *sm and *si ‘one’ 205
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).

via free access
206 de vaan
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:
Numeral ‘one’ Demonstrative in the oblique
*sm *ʔe-sm-, *to-sm-

*si *ʔe-si-, *to-si-
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?

via free access
5. The occurrence of *si as ‘one’ is restricted to the Anatolian branch, to Greek,

and possibly to Tocharian. In the latter two branches, *si had apparently be-
come a relic form, ousted by (derivatives of) *sem-. This distribution would
make sense in the light of the Indo-Anatolian hypothesis, according to which
Proto-Anatolian was the first branch of Indo-European to lose contact with the
other dialects (Sturtevant 1933: 30). In this view, Anatolian failed to share a
larger number of morphological innovations found in the other branches of
Indo-European, instead experiencing its own innovations but also retaining
a number of features which were subsequently lost from the other branches
of Indo-European. Whereas the Indo-Anatolian hypothesis has been explicitly
dismissed or implicitly put aside by many scholars in the second half of the
twentieth century, Kloekhorst (2016: 229) now judges:
“Yet, in the last few decades this viewpoint has shifted, and nowadays the
majority of scholars seem to support the hypothesis that Anatolian did split
off first from the mother language, and that the other branches at that point
in time still formed a single language community that underwent some com-
mon innovations that Anatolian did not share (Kloekhorst 2008: 7–11, Oettinger
2013/2014, Melchert fthc.; cf. Rieken 2009 for a more cautious view). This does
not mean, however, that there is at the moment any broad consensus on the
number or nature of these common innovations.”
One of the central issues in the discussion on an Indo-Anatolian prestage
is the rise of the Proto-Indo-European gender distinctions. It has repeatedly
been argued that the feminine gender only arose in PIE after the Anatolian
languages branched off (Matasović 2004). If the feminine gender arose at the
post-Indo-Anatolian stage (as I believe is likely), PIE *sm and *si did not have a
gender connotation at the Indo-Anatolian stage. They were two different deic-
tic stems. The form *si was probably the first to develop the meaning ‘one’: that
is its meaning in Hittite, and, residually, also in Tocharian and Greek. The stem
*sm may have come to mean ‘one’ at a later stage. Note that the facts concern-
ing *si ‘one’ would not contradict the theory, held by a number of scholars,
that Tocharian was the second branch to split away from the remaining Indo-
European languages, after Anatolian.
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

via free access
208 de vaan
table 14.2 PIE origins of Anatolian pronouns according to Kloekhorst

2012: 263
Singular Plural
nom.com. *Có-s *Cí-s *Có-i *Cí-es(?)

acc.com. *Có-m *Cí-m *Có-ms *Cí-ms
nom.-acc.n. *Có-d *Cí(-d) – *Cí-ʕ
genitive *Cé-l *Cé-(i-)nʔ-som
dative-locative *Cé-dh-(nʔ-)i *Cé-(nʔ-)dh-os
ablative-instr. *Cé-t
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.

via free access
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

via free access
210 de vaan
table 14.3 Early PIE pronouns according to Kortlandt 2010: 41
Animate Animate Inanimate Inanimate

nominative accusative absolutive ergative
deictic *so *tom *to *tod

anaphoric *ʔe *im *i *id
interrogative *kwe *kwim *kwi *kwid
paradigms) as a sigmatic nominative when the nominative/accusative sys-

tem was established. The o was reinterpreted as the stem affix, to which an
accusative in *-om was first formed, followed by the independent addition
of other case endings in Anatolian and in Late Proto-Indo-European (where
pronominal endings were often used). The o-stems remained productive in
most of the Indo-European daughter languages and have replaced many ath-
ematic formations. In the pronouns, too, it seems that only the nominative
and accusative (of animates) and the absolutive and ergative (of inanimates)
had developed, but no ablative = ergative (whence genitive) in *-s or any other
oblique case desinence.
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

via free access
table 14.4 Early PIE pronouns including deictic *se, si
Animate Animate Inanimate Inanimate

nominative accusative absolutive ergative
deictic *so *tom *to *tod

identificational *se *sim *si
anaphoric *ʔe *im *i *id
interrogative *kwe *kwim *kwi *kwid
‘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

via free access
212 de vaan
table 14.5 Late Pre-PIE pronouns, separation of *t- and *s-
Animate Inanimate
Nominative Nominative Accusative Absolutive Ergative

(stressed) (clitic)
deictic *to *tom *tod

identificational *sé *so *sim *si –
anaphoric *ʔé *ʔo *im *i *id
interrogative *kwé *kwo *kwim *kwi *kwid
the enclitic or unstressed use of the lexeme as a determiner. In fact, in view

of the evidence for asigmatic *kwo in compound pronouns of the type *kwo-so
(see below), we could go one step further: all pronominal monosyllables of the
structure *Co might be due to clitic usage of the animate nominative. This view
would allow us to postulate a systematic opposition between stressed *Cé and
clitic *Co in the animate nominative versus only *Ci in the inanimate of deictic
pronouns, as illustrated in the second, third and fourth row of table 14.5.
Within Proto-Indo-European, the intraparadigmatic consonant alternation
between s- and t- in the demonstrative *so, *to- is unique. As there is no sound
law which explains PIE *so from earlier *to, we might follow Kortlandt (2010:
398, 416) in assuming the existence of two different Proto-Indo-Uralic pro-
nouns, viz. *s- (compare Finnish hän ‘(s)he’) and *t- (Finn. tämä ‘this’, tuo ‘that’).
In this view, PIE *se, *so, *si reflects the former stem and PIE *to (as reflected in
Hitt. ta clause conjunctive particle < *to, takku ‘if’ < *to-kwe, Kloekhorst 2008:
925, 942), the latter. Table 14.4 can now be modified in such a way that the s-
and t-pronouns each have their own row, see table 14.5.
In view of the deviant vocalism of the to-pronoun in table 14.5, it is legiti-
mate to ask whether we can arrive at an even more symmetrical pronominal
system by means of internal reconstruction—while being fully aware that per-
fectly symmetrical morphological systems are rare at any synchronic stage of
language. The evidence for the accusative *tom is outer-Anatolian, and hence
it is conceivable—not certain, because we are building on an argumentum
ex silentio—that *tom was only created after *so and *to- had merged into a
single deictic paradigm. For similar reasons, it is possible that inanimate *tod
was created on the model of other pronouns in *-od. A more symmetrical ver-
sion of the t- and s-pronouns in table 14.5 would contain the forms *tim and
*ti:

via free access
table 14.6 Early Pre-PIE pronouns, symmetrical vocalism for *t- and *s-
Animate Animate Animate Inanimate Inanimate

nominative nominative accusative absolutive ergative
(stressed) (clitic)
deictic *to *tim *ti –

identificational *sé *so *sim *si –
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.

via free access
214 de vaan
table 14.7 Vowels of the PIE genitive and oblique deictics
e-grade o-grade
PIE genitive *ʔe-so, *kwe-so *to-si (>> *tosio)

PIE oblique *ʔe-sm-, *kwe-sm- *to-sm-, *to-si-h₂-
Kortlandt’s chronology is based on the absence from Anatolian of a geni-

tive in *-osio. Hence, the rise of this ending would postdate Indo-Anatolian.
Melchert 2012 argues that a number of Anatolian adjectival forms in -ašša-
reflect an o-stem ending *-osio. This would mean that its source form *tosio
must have existed in a prestage of Anatolian after all (but was lost without a
trace in the Anatolian pronouns), or that the suffix -ašša- < *-osso- had a differ-
ent origin. Clearly this issue will need to be further discussed among Anatolian
specialists. The existence of Early PIE *tosio would affect the relative chronol-
ogy but not the essence of what I propose here.
The vocalic difference between the e-grade in *ʔe-so, *kwe-so and the o-grade
in *to-si corresponds to the same vowel difference between *ʔe-sm-, *kwe-sm-
and *to-sm-, *to-si-h₂- (Beekes 1988a, 1988b: 4). We may therefore assume that
genitival *to-si (whence *tosio) is contained in the first two constituent ele-
ments of feminine *to-si-h₂-, see table 14.7.
The vowel distinction between *ʔe- and *to- in these juxtaposed pronouns
correlates with two different second members, *so versus *si. The animate ref-
erence of *ʔe and *so versus the inanimate usage of *tod suggests that *to may in
origin have applied specifically to inanimates, or that it developed this restric-
tion at some point in time. The inanimate reference of *si is also suggested by
tables 14.3–14.6.
A restriction of *tosi to inanimate or non-count nouns would explain its
reanalysis as a feminine marker in Late PIE. As is well known, the canonical PIE
feminine gender developed out of inanimacy markers, such as the collectives in
*-h₂. Addition of the productive feminine marker yielded the feminine obliques
*to-si-eʕ- etc. (Skt. tásyās, etc.). The reinterpretation of *to-si as feminine cannot
have been earlier than the creation of *tosi-o, which acquired animate (mascu-
line) reference. The explanation put forward here would imply that, in terms
of relative chronology, the feminine si-forms first arose in the pronoun *to- and
subsequently spread to other pronouns.
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

via free access
meanings ‘one’, ‘together’, ‘same’, ‘truly’ seen in section 1. A comparable use

of a derivative of *sm- is found in the Old Irish emphatic suffix -som from
*-som(H)o- (Schrijver 1997: 50). One is also reminded of the Hittite enclitic per-
sonal pronoun 2pl. dative and accusative ‘(to) you’ and 3pl. dative ‘to them’
-šmaš, as well as of the possessive pronoun 2pl. and 3pl. -šmi-, -šma-, -šme-
‘your (pl.); their’ (Hoffner & Melchert 2008: 135, 138–140). Their etymology is
regarded as unclear (Kloekhorst 2008: 888, 889). The combined use for 2pl.
and 3pl. might be interpreted as a general reference to plurals, in which case
a derivation from *sm ‘together’ may be envisaged.
Internal reconstruction suggests that the locative in *-sm-i was the starting
point for the addition of sm-endings (thus also Gotō 2013: 68): the dative and
ablative have added to *sm the Late PIE o-stem endings dat.sg. *-oʔei, abl.sg.
*-oʔed (Kortlandt 2010: 40), whereas the locative displays the older consonant-
stem ending *-i.2 Locatives such as *ʔe-sm-i and *to-sm-i may have arisen in
order to disambiguate the original locative forms *ʔei, *toi from the homony-
mous nominative plural forms. Such homonymy would have been particu-
larly disturbing after the verbal system had switched to nominative-accusative
marking. Before this change, the animate subject of intransitive verbs (that is,
the verbs which became the PIE perfect and the thematic presents) was proba-
bly expressed by the locative in *-i, which could be reinterpreted as an indirect
object in the dative (Kortlandt 2010: 102, 382). After the collapse of the ergative
construction, this dative-subject construction is gradually replaced by nomi-
native subjects.
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.

via free access
216 de vaan
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.
Benjamins.
Brugmann, K. 1904. Kurze vergleichende Grammatik der indogermanischen Sprachen.
Strassburg: Trübner.
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.
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.
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.
germanische Forschungen 121/1, 213–248.

via free access
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.

via free access
218 de vaan
Sturtevant, E.H. 1933. A Comparative Grammar of the Hittite Language. Philadelphia:

Linguistic Society of America/University of Pennsylvania.
Vaillant, A. 1936. L’ergatif indo-européen. Bulletin de la Société de Linguistique de Paris
37, 93–108.
Wackernagel, J. 1905. Altindische Grammatik II/1. Einleitung zur Wortlehre. Nominalkom-
position. Göttingen: Vandenhoeck & Ruprecht.

via free access
chapter 15
Indo-Uralic and the Origin of Indo-European

Ablaut
Mikhail Zhivlov*
1 Čop’s Ideas on the Origin of Indo-European Ablaut
In 1975 Bojan Čop published a book entitled “Die indogermanische Deklina-

tion im Lichte der indouralischen vergleichenden Grammatik” in which, inter
alia, he presented some thoughts on the development of Proto-Indo-European
ablaut. His ideas remained unnoticed, and I know of no reference to them in
the Indo-Europeanist literature.1
Čop’s thoughts on the origin of ablaut can be summarized as follows (see
Čop 1975: 89–93 and Čop 1990: 27–29). Proto-Indo-European monosyllabic
roots correspond to Proto-Uralic and Proto-Indo-Uralic disyllabic roots. As a
result of reduction of unstressed vowels, the second syllable vowel was lost
unless the root was followed by a suffix consisting of consonant + vowel.
Such suffixes caused the accent (that otherwise fell on the first syllable of the
root) to shift one syllable forward, so that it fell on the second syllable of the
root. When the original second syllable vowel was preserved, it was reinter-
preted as belonging to the following suffix (originally, all suffixes began with
consonants). Schematically this scenario can be represented as follows: PIU
*CV́ CV > PIE *CV́ C, PIU *CV́ CV-C > PIE *CV́ C-C, PIU *CVCV́ -CV > PIE *CC-
V́ C. The Proto-Indo-European genitive-ablative singular ending *-e/os, accord-
ing to Čop, corresponds to the Proto-Uralic ablative ending *-ta. The fact that
the Indo-European ending has an allomorph *-s may seem to contradict Čop’s
theory. Čop explains it in the following way: Proto-Uralic had two kinds of
stems: *a/ä-stems and *e-stems (now usually reconstructed as *i-stems). The
final vowel of *e-stems was deleted before certain suffixes, including the abla-
tive ending *-ta. Thus, Proto-Uralic and Proto-Indo-Uralic *e-stem nouns cor-
respond to Proto-Indo-European nouns with gen.-abl.sg. *-s, while original
* 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).

via free access
220 zhivlov
*a-stem nouns correspond to Proto-Indo-European nouns with gen.-abl.sg.

*-e/os: PIU *CVCá-ta > PIE *CC- és, PIU *CV́ C-ta > PIE *CV́ C-s.
Taking Čop’s ideas as a point of departure, we will attempt to sketch an out-
line of a theory of the origin of Proto-Indo-European ablaut. Below, we will
argue that the rise of different ablaut paradigms of the Proto-Indo-European
noun can be explained if we accept the following assumptions:
1) Proto-Indo-European and Proto-Uralic are genetically related;
2) Proto-Indo-Uralic phonotactics and morphophonology were largely pre-
served in Proto-Uralic, but drastically changed in Proto-Indo-European;
3) Proto-Indo-European ablaut must be studied separately from the accent
system that can be reconstructed for Proto-Indo-European on the basis of
Old Indic, Greek, Balto-Slavic and Germanic data. Ablaut reflects an ear-
lier, pre-Proto-Indo-European accent system, which, while superficially
similar in a number of points to the Proto-Indo-European one, was based
on entirely different principles.
Before we return to Čop’s ideas, we must clarify this last point.
2 Indo-European Accent and Ablaut
There are two approaches to the reconstruction of Proto-Indo-European ac-

cent and ablaut. According to the mainstream approach, also known as the
paradigmatic model, Proto-Indo-European quantitative ablaut is directly con-
nected with the Proto-Indo-European accent: zero grade results from vowel
reduction in unstressed syllables. The two most influential versions of this ap-
proach are the Erlangen model (summarized in Meier-Brügger 2010: 336–353),
and the Leiden model (summarized in Beekes 2011: 190–217 and Kloekhorst
2013).
The reconstruction of Proto-Indo-European accent in the paradigmatic
approach is based mainly on the evidence of archaic ablaut patterns. More
often than not accentological data of the daughter languages are treated as
secondary and irrelevant to the reconstruction simply because they do not fit
the picture based on ablaut. One of the main problems of the paradigmatic
approach is that the direct connection between accent and ablaut must be
ascribed to some pre-Proto-Indo-European stage, because reconstructed Proto-
Indo-European has numerous counterexamples to this generalization, cf. such
̥́
forms as *u̯ lkʷos ̥́
‘wolf’, *h₂ŕ̥tk̑os ‘bear’, *u̯ lh₁neh₂ ‘wool’, *septḿ̥ ‘seven’, *u̯ óide
‘knows’ etc. But if there was no synchronic connection between accent and
ablaut in Proto-Indo-European, how can we reconstruct the accent using the
evidence of ablaut? Another problem is the absence of a theory that would

via free access
indo-uralic and the origin of indo-european ablaut 221
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.

via free access
222 zhivlov
3 Uralic Morpheme Structure and Noun Morphology
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):
PU nom.sg. *-Ø ~ PIE nom.-acc.sg. *-Ø (in neuter athematic nouns)

PU acc.sg. *-m ~ PIE acc.sg. *-m
PU dual *-ki(-) ~ PIE nom.-acc.du. *-h₁
PU abl. *-tA ~ PIE gen.-abl.sg. *-es / *-os / *-s
PU nom.pl. *-t ~ PIE nom.pl. *-es
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.

via free access
table 15.1 Proto-Uralic case endings
Sg. Du. Pl.
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).

via free access
224 zhivlov
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-

via free access
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.
4 Uralic Stem Types and Indo-European Ablaut Paradigms
Now we can examine the possible combinations of Proto-Indo-Uralic stems,

suffixes and endings and their reflexes in Proto-Indo-European. We will first
discuss root nouns, then nouns with monosyllabic suffixes of the shape *-CV
and finally nouns with disyllabic suffixes of the shape *-CVCV. Unlike Čop, we
will not discuss actual word comparisons between Uralic and Indo-European.
The only direct comparison between Indo-European and Uralic morphemes
in our model is the comparison of noun endings. There are two reasons for
that. First, reliable Indo-Uralic etymologies are rather scarce. Second, and more
important, most of the Indo-European ablaut paradigms are preserved only
in a few stems in the ancient Indo-European languages. Therefore, words that
belong to more widespread ablaut types (mobile type of root nouns and hys-
terokinetic type of suffixed nouns) can actually be secondarily transferred to
these types from less productive paradigms.
Root nouns can be divided into two types: *a-stem nouns3 and *i-stem
nouns.
Type 1 yields the Indo-European mobile ablaut paradigm of root nouns,
attested in numerous words, e.g. the word for ‘man’: PIE nom.sg. *h₂nēr (<
*h₂ner-s), acc.sg. *h₂ner-m, gen.-abl.sg. *h₂n̥ r-es > Greek nom.sg. ἀνήρ, acc.sg.
ἀνέρα, gen.-abl.sg. ἀνδρός ‘man’. On the Uralic side, this type is represented by
the Proto-Uralic word *kala ‘fish’ (Finnish nom.sg. kala, part.sg. kalaa < PU abl.
*kala-ta).
Type 2 yields the acrostatic paradigm of root nouns. Although not attested
as such in the daughter languages, it is postulated for Proto-Indo-European on
the evidence of the fossilized gen.sg. form of the word for ‘house’, preserved
in the fixed phrase *dem-s poti-s ‘master of the house’ > Greek δεσπότης ‘mas-
ter (of the house), lord’, Old Indic dámpati- and pátir dán ‘householder, lord
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.

via free access
226 zhivlov
table 15.2 Type 1: *a-stem root nouns
PIU Pre-PIE PIE
nom.sg. *CVCa *CV́ Ca *CeC(-s)

acc.sg. *CVCa-m *CV́ Ca-m *CeC-m
(gen.)-abl.sg. *CVCa-ta *CVCá-ta *CC-es
nom.pl. *CVCa-t *CV́ Ca-t *CeC-es
table 15.3 Type 2: *i-stem root nouns
PIU Pre-PIE PIE
nom.sg. *CVCi *CV́ :Ci *CoC(-s)

acc.sg. *CVCi-m *CV́ :Ci-m *CoC-m
(gen.)-abl.sg. *CVC-ta *CV́ C-ta *CeC-s
nom.pl. *CVCi-t *CV́ :Ci-t *CoC-es
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

via free access
table 15.4 Type 3: *a-stem nouns with *-Ca suffixes
PIU Pre-PIE PIE
nom.sg. *CVCa-Ca *CVCá-Ca *CC-eC(-s)

acc.sg. *CVCa-Ca-m *CVCá-Ca-m *CC-eC-m
(gen.)-abl.sg. *CVCa-Ca-ta *CVCa-Cá-ta *CC-C-es
nom.pl. *CVCa-Ca-t *CVCá-Ca-t *CC-eC-es
table 15.5 Type 4: *a-stem nouns with *-Ci suffixes
PIU Pre-PIE PIE
nom.sg. *CVCa-C *CV́ Ca-C *CeC-C

acc.sg. *CVCa-Ci-m *CVCá-Ci-m **CC-eC-m
(gen.)-abl.sg. *CVCa-C-ta *CVCá-C-ta *CC-eC-s
nom.pl. *CVCa-Ci-t *CVCá-Ci-t **CC-eC-es
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).

via free access
228 zhivlov
table 15.6 Type 5: *i-stem nouns with *-Ca suffixes
PIU Pre-PIE 1 Pre-PIE 2 PIE
nom.sg. *CVC-Ca *CV́ C-Ca *CeC-C *CeC-oC(-s)

acc.sg. *CVC-Ca-m *CV́ C-Ca-m *CeC-C-m *CeC-oC-m
(gen.)-abl.sg. *CVC-Ca-ta *CVC-Cá-ta *CC-C-es *CC-C-es
nom.pl. *CVC-Ca-t *CV́ C-Ca-t *CeC-C-es *CeC-oC-es
table 15.7 Type 6: *i-stem nouns with *-Ci suffixes
PIU Pre-PIE PIE
nom.sg. *CVC-Ci *CV́ :C-Ci *CoC-C(-s)

acc.sg. *CVC-Ci-m *CV́ :C-Ci-m *CoC-C-m
(gen.)-abl.sg. *CVC-C-ta *CV́ C-C-ta *CeC-C-s
nom.pl. *CVC-Ci-t *CV́ :C-Ci-t *CoC-C-es
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.

via free access
M. Kümmel has convincingly shown that in Proto-Indo-Iranic i- and u-stems

the choice between ‘proterokinetic’ flexion (Old Indic nom.sg. sūnúḥ, gen.-
abl.sg. sūnóḥ) and flexion with consistent zero grade of -i- and -u- (Old Indic
nom.-acc.sg. mádhu, gen.-abl.sg. mádhvaḥ) is governed by the weight of the first
syllable (Kümmel 2014; see also Kiparsky 2010). The choice between zero grade
and full grade of the suffix in the subtype with ‘proterokinetic’ flexion depends
on the segmental shape of the ending: full grade of the suffix is found before
endings beginning with a vowel, whereas zero grade is found before endings
beginning with a consonant (Kiparsky 2010; Kümmel 2014). However, this rule
does not account for the gen.-abl.sg. form, where we find full grade before *-s.
Thus, although ablaut grades of the suffix were secondarily redistributed, the
forms of the genitive and nominative singular, crucial for our reconstruction,
preserve the original state of affairs.
Therefore, the ‘proterokinetic’ flexion in i- and u-stems (as opposed to the
proterokinetic ablaut paradigm in cases like *gʷen-h₂ / *gʷn-eh₂-s, see below)
tells us nothing about original ablaut alternations in the root. The real evi-
dence for proterokinetic ablaut comes only from nouns with simultaneous
ablaut alternation in root and suffix. These can be illustrated by the word for
‘woman’: PIE nom.sg. *gʷēn (< *gʷen-h₂), gen.sg. *gʷn-eh₂-s > Old Irish nom.sg.
ben,5 gen.sg. mná.
Type 5 yields amphikinetic nouns. Our model predicts that the suffix must
be in the zero-grade throughout the paradigm. In fact, the suffix has *o-grade in
the strong cases. The suffixal ablaut was probably introduced under the influ-
ence of the hysterokinetic type. Apparently, this happened at a stage when the
vowel *e in the unstressed position was restricted to certain environments;
therefore the analogically introduced vowel automatically became *o (Kort-
landt 2010b: 396). The amphikinetic type can be illustrated by the follow-
ing examples: 1) PIE nom.sg. *dʰeg̑ ʰ-ōm (< *dʰeg̑ ʰ-om-s), acc.sg. *dʰeg̑ ʰ-om-m,
gen.sg. *dʰg̑ ʰ-m-os > Hittite nom.-acc.sg. tēkan, gen.sg. taknāš ‘earth’ (the Hit-
tite word became neuter, but preserved the ablaut); 2) PIE nom.sg. *pent-oH-s,
acc.sg. *pent-oH-m, gen.sg. *pn̥ t-H-os, nom.pl. *pent-oH-es > Old Indic nom.sg.
pánthāḥ, acc.sg. pánthām, gen.sg. patháḥ, nom.pl. pánthāḥ ‘way, path, course’;
3) PIE nom.sg. *eh₁t-mō (< *eh₁t-mon-s), acc.sg. *eh₁t-mon-m, gen.sg. *h₁t-mn-
os > Old Indic nom.sg. ātmā ́, acc.sg. ātmā ́nam, dat.sg. tmáne ‘breath, soul, self’.
Due to the general tendency to eliminate ablaut alternations in the root, this
type is rather poorly attested, but G. Kroonen has shown that it was fairly fre-
quent in Proto-Germanic n-stems (Kroonen 2011: 133–334).
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).

via free access
230 zhivlov
table 15.8 Type 7: *a-stem nouns with *-CVCa suffixes
PIU Pre-PIE PIE
nom.sg. *CVCa-CVCa *CVCa-CV́ Ca *CC-CeC(-s)

acc.sg. *CVCa-CVCa-m *CVCa-CV́ Ca-m *CC-CeC-m
(gen.)-abl.sg. *CVCa-CVCa-ta *CVCa-CVCá-ta *CC-CC-es
nom.pl. *CVCa-CVCa-t *CVCa-CV́ Ca-t *CC-CeC-es
table 15.9 Type 8: *i-stem nouns with *-CVCa suffixes
PIU Pre-PIE PIE
nom.sg. *CVC-CVCa *CVC-CV́ Ca *CC-CeC(-s)

acc.sg. *CVC-CVCa-m *CVC-CV́ Ca-m *CC-CeC-m
(gen.)-abl.sg. *CVC-CVCa-ta *CVC-CVCá-ta *CC-CC-es
nom.pl. *CVC-CVCa-t *CVC-CV́ Ca-t *CC-CeC-es
The predicted outcome of type 6 is the acrostatic type of suffixed nouns.

Its reconstruction is rather doubtful, because it is not attested as such, but
merely postulated on circumstantial evidence. The main piece of evidence is
the Hittite expression nekuz mēḫur ‘in the evening’, literally ‘time of evening’,
where nekuz is taken to be a fossilized gen.sg. of the word for ‘evening’: *nekʷ-
t-s, related to Narrow Indo-European *nokʷ-t- ‘night’ (Schindler 1967). Since
otherwise Hittite generalized the allomorph *-os of the gen.sg. ending, the
form nekuz must be an archaism. Cf., however, an alternative interpretation in
Kloekhorst 2014. Another piece of evidence—the Indo-Iranian and Germanic
gen.-abl.sg. of the *r-stems in *-r̥-s—can have an alternative explanation (see
below). Note that the ‘acrostatic’ accentuation is a widespread phenomenon in
Indo-European, but it is irrelevant from our point of view: we look only at the
evidence for acrostatic ablaut pattern.
Up to this point we considered only combinations of roots with CV-type suf-
fixes (> VC-type in Indo-European). However, Proto-Indo-European had many
noun suffixes of the shape *-CeC-. In terms of our hypothesis, they must go back
to pre-Proto-Indo-European disyllabic suffixes of the shape *-CVCV- (it is not
clear whether such suffixes existed already in Proto-Indo-Uralic). The following
combinations of root and suffix are possible:
Types 7 and 8 yield the hysterokinetic paradigm. They require no further
comment.

via free access
table 15.10 Type 9: *a-stem nouns with *-CVCi suffixes
PIU Pre-PIE PIE
nom.sg. *CVCa-CVC *CVCá-CVC **CC-VCC

acc.sg. *CVCa-CVCi-m *CVCa-CV́ :Ci-m **CC-CoC-m
(gen.)-abl.sg. *CVCa-CVC-ta *CVCa-CV́ C-ta **CC-CeC-s
nom.pl. *CVCa-CVCi-t *CVCa-CV́ :Ci-t **CC-CoC-es
table 15.11 Type 10: *i-stem nouns with *-CVCi suffixes
PIU Pre-PIE PIE
nom.sg. *CVC-CVC *CV́ C-CVC **CVC-CC

acc.sg. *CVC-CVCi-m *CVC-CV́ :Ci-m **CC-CoC-m
(gen.)-abl.sg. *CVC-CVC-ta *CVC-CV́ C-ta **CC-CeC-s
nom.pl. *CVC-CVCi-t *CVC-CV́ :Ci-t **CC-CoC-es
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

via free access
232 zhivlov
neuter nouns type 4 yielded the proterokinetic paradigm): nom.sg. *CVC-C,

acc.sg. *CC-eC-m, gen.-abl.sg. *CC-eC-s. The only difference is in the genitive:
instead of **CC-eC-s, predicted by our model, we have *CC-C-es. This may be
viewed as a result of remodeling the original paradigm under the influence of
hysterokinetic nouns. Since this ablaut type is attested only in a single Hittite
word, we do not know whether this remodeling took place already in Proto-
Indo-European, or later in the separate history of Anatolian.
Another possible trace of this ablaut pattern can be discerned in the kinship
terms with the suffix *-ter-. Synchronically, they show hysterokinetic ablaut
paradigm, but with some peculiarities: 1) the words *meh₂-ter- ‘mother’ and
*bʰreh₂-ter- ‘brother’ have full grade of the root; 2) the root of the word for
‘daughter’ has zero-grade in Narrow Indo-European: *dʰugʰh₂-ter-, but full
grade in Anatolian: *dʰu̯ egʰh₂-ter- (Kloekhorst 2008: 902–904); 3) in Old Indic
and partly in Germanic these words have a genitive in *-tr̥-s. The latter is usu-
ally seen as originating in the word for ‘brother’ with its ‘acrostatic’ accent, but
there is no positive evidence for that. We can suppose that originally these
words had the following paradigm: nom.sg. *dʰu̯ egʰh₂-tr̥ (s), acc.sg. *dʰugʰh₂-
ter-m, gen.-abl.sg. *dʰugʰh₂-ter-s. Already in Proto-Indo-European, the ablaut
of the suffix was remodeled on the analogy of the hysterokinetic type, but the
shape of gen.-abl.sg. ending and the root ablaut (at least in the word for ‘daugh-
ter’) were preserved, yielding the following forms: nom.sg. *dʰu̯ egʰh₂-tēr, acc.sg.
*dʰugʰh₂-ter-m, gen.-abl.sg. *dʰugʰh₂-tr̥-s. After the separation of the Anatolian
branch, the root ablaut was leveled in favour of full or zero-grade.
The final, and most important, question is where does the limited corre-
lation between ablaut and synchronic accent (which can be easily seen in
Old Indic words with mobile stress paradigms) come from? The endings that
require a full ablaut grade of the preceding morpheme also do not attract stress:
they have “- valency” in terms of the morphophonological approach. According
to our hypothesis, such endings go back to Proto-Indo-Uralic endings without
a vowel, e.g., acc.sg. *-m and nom.pl. *-t. If the tonological interpretation of the
morphophonological approach is correct, the “+ valency” reflects the high tone.
Endings that originally did not have any vowels also evidently did not have tone,
so that their “- valency” reflects the absence of a high tone. This means that the
limited correlation between synchronic accent and ablaut, attested in ancient
Indo-European languages, is fully explained by our model.
We must emphasize that the hypothesis presented here involves no circu-
lar reasoning. We start from Proto-Indo-Uralic forms, whose shape is in no
way influenced by Indo-European data, because they are “back-projected” from
Proto-Uralic to our hypothetical Proto-Indo-Uralic. Applying to these forms a
set of simple phonological rules, we arrive at Indo-European ablaut paradigms.

via free access
Each of the possible combinations of roots and monosyllabic suffixes (types 1–

6 above) yields an ablaut paradigm that is either actually attested or at least (in
the case of acrostatic paradigms) postulated by Indo-Europeanists indepen-
dently of any external evidence.
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.
Benjamins.
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–

via free access
234 zhivlov
Jasanoff, J.H. 1989. Old Irish bé ‘woman’. Ériu 40, 135–141.

Kiparsky, P. 2010. Compositional vs. paradigmatic approaches to accent and ablaut.
In: S.W. Jamison; H.C. Melchert; B. Vine (eds.), Proceedings of the 21st Annual UCLA
Indo-European Conference. Bremen: Hempen, 137–181.
Boston: Brill.
Kloekhorst, A. 2013. Indo-European nominal ablaut patterns: The Anatolian evidence.
In: G. Keydana, P. Widmer; T. Olander (eds.), Indo-European Accent and Ablaut.
Kloekhorst, A. 2014. The Proto-Indo-European acrostatic inflection reconsidered. In:
N. Oettinger; T. Steer (eds.), Das Nomen im Indogermanischen. Morphologie, Sub-
stantiv versus Adjektiv, Kollektivum. Akten der Arbeitstagung der Indogermanischen
Gesellschaft vom 14. bis 16. September 2011 in Erlangen. Wiesbaden: Reichert, 140–163.
Kortlandt, F. 2010a. Proto-Indo-European tones? In: F. Kortlandt Studies in Germanic,
Indo-European and Indo-Uralic. Amsterdam–New York: Rodopi, 67–72.
Kortlandt, F. 2010b. The Indo-Uralic verb. In: F. Kortlandt Studies in Germanic, Indo-
European and Indo-Uralic. Amsterdam–New York: Rodopi, 391–403.
Kroonen, G. 2011. The Proto-Germanic n-Stems. A Study in Diachronic Morphophonology.
Amsterdam–New York: Rodopi.
Kümmel, M.J. 2012. Typology and reconstruction: The consonants and vowels of Proto-
Indo-European. In: B. Nielsen Whitehead; Th. Olander; B.A. Olsen and J.E. Ras-
mussen (eds.), The Sound of Indo-European: Phonetics, Phonemics, and Morpho-
phonemics. Copenhagen: Museum Tusculanum, 291–329.
Kümmel, M.J. 2014. Zum “proterokinetischen” Ablaut. In: N. Oettinger; T. Steer (eds.),
Das Nomen im Indogermanischen. Morphologie, Substantiv versus Adjektiv, Kollek-
tivum. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 14. bis 16.
September 2011 in Erlangen. Wiesbaden: Reichert, 164–179.
Lubotsky, A.M. 1988. The System of Nominal Accentuation in Sanskrit and Proto-Indo-
European. Leiden–New York–København–Köln: Brill.
Meier-Brügger, M. 2010. Indogermanische Sprachwissenschaft. 9., durchgesehene und
ergänzte Auflage. Unter Mitarbeit von Matthias Fritz und Manfred Mayrhofer. Berlin–
New York: Walter de Gruyter.
Pronk, T. 2013. On Indo-European tones, accentuation and ablaut. In: G. Keydana;
P. Widmer; T. Olander (eds.), Indo-European Accent and Ablaut. Copenhagen: Mu-
seum Tusculanum, 149–176.
Rasmussen, J.E. 1999. The make-up of Indo-European morphology. In: J.E. Rasmussen
Selected Papers on Indo-European Linguistics. With a Section on Comparative Eskimo
Linguistics. Part 1. Copenhagen: Museum Tusculanum, 244–255.
Salminen, T. 1996. Comments on László Honti’s paper “Zur Morphotaktik und Mor-
phosyntax der uralischen/finnisch-ugrischen Grundsprache”. In: H. Leskinen;

via free access
S. Maticsák; T. Seilenthal (eds.), Congressus Octavus Internationalis Fenno-Ugris-

tarum VIII. Jyväskylä: Gummerus, 25–27.
Schindler, J. 1967. Zu hethitisch nekuz. Zeitschrift für vergleichende Sprachforschung 81,
290–303.
Schindler, J. 1972. L’apophonie des noms-racines indo-européens. Bulletin de la Société
de Linguistique de Paris 67, 31–38.
Zhivlov, M. 2014. Studies in Uralic vocalism III. Journal of Language Relationship 12,
113–148.

via free access

IE KloekhorstPronk-2019 (Precursors)

Uploaded by

Copyright:

Available Formats

You might also like

IE KloekhorstPronk-2019 (Precursors)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IE KloekhorstPronk-2019 (Precursors)

Uploaded by

Copyright:

Available Formats

The Precursors of Proto-Indo-European

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

The titles published in this series are listed at brill.com/lsie

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Names: Kloekhorst, Alwin, editor. | Pronk, Tijmen, editor.

Copyright 2019 by Koninklijke Brill NV, Leiden, The Netherlands.

This book is printed on acid-free paper and produced in a sustainable manner.

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

1 Introduction: Reconstructing Proto-Indo-Anatolian

2 The Proto-Indo-European Suffix *-r Revisited 15

3 Pronouns and Particles: Indo-Uralic Heritage and Convergence 30

5 Daniel Europaeus and Indo-Uralic 74

6 Bojan Čop’s Indo-Uralic Hypothesis and Its Plausibility 88

7 Indo-European o-grade Presents and the Anatolian ḫi-conjugation 102

8 The Proto-Indo-European mediae, Proto-Uralic Nasals from a Glottalic

9 Thoughts about Pre-Indo-European Stop Systems 115

10 The Anatolian “Ergative” 131

11 The Indo-European Suffix *-ens- and Its Indo-Uralic Origin 151

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

12 Headedness in Indo-Uralic 163

13 Indo-Uralic, Indo-Anatolian, Indo-Tocharian 186

14 Proto-Indo-European *sm and *si ‘one’ 203

15 Indo-Uralic and the Origin of Indo-European Ablaut 219

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

‘The Precursors of Proto-Indo-European’ was the name of a workshop held at

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Introduction: Reconstructing Proto-Indo-Anatolian

Alwin Kloekhorst and Tijmen Pronk

1 The Nature of Proto-Indo-European

Before we try to answer the question whether any outer-Indo-European rela-

© koninklijke brill nv, leiden, 2019 | doi:10.1163/9789004409354_002

2 The Indo-Anatolian Hypothesis

Hittite also changed the reconstruction of Proto-Indo-European in another,

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

24. Hitt. unreduplicated ḫi-conjugation vs. cl.PIE reduplicated perfect: gener-

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

3 Indo-Tocharian and the Indo-Anatolian Hypothesis

‘Classical’ Proto-Indo-European, which above was taken as comprising all non-

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Another way of reconstructing one or more precursor stages of Proto-Indo-

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

When it comes to comparing the Indo-European language family to one or

6 The Indo-Uralic Hypothesis

Already in the 19th century, the linguistic similarities between Indo-European

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Although we regard the Indo-Uralic hypothesis as very likely to be correct,

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

figure 1.1 The precursors of Proto-Indo-European

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Bomhard, A.R. 2008. Reconstructing Proto-Nostratic: Comparative Phonology, Morphol-

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

Alwin Kloekhorst and Tijmen Pronk - 978-90-04-40935-4

The Proto-Indo-European Suffix *-r Revisited

14 Proto-Indo-European sm and si ‘one’ 203

Dunkel assumes that an “Adverbialendung” -r/-er is also reflected in other

PIE h₁up- ‘surface’: loc. h₁uper ‘on the surface, on top’;

PIE h₁up- ‘surface’: dir. h₁upo ‘to the surface’;

locative -r -i