Traces of Pre-Indo-Iranian'

Traces of ‘Pre-Indo-Iranian’
Chronological Layers and Structural Characteristics of

Early Indo-Iranian Loanwords
Axel Palmér
S2091135
Thesis submitted in partial

fulfillment for the degree of
Master of Arts – Linguistics (Research)
Leiden University Centre for Linguistics
Faculty of Humanities
Leiden University
July 2019
Supervisor: Prof.dr. A.M. Lubotsky
Second Reader: Dr. A. Kloekhorst

Contents
Acknowledgements .................................................................................................................... v
Abstract ..................................................................................................................................... vi
Abbreviations ........................................................................................................................... vii
Languages ............................................................................................................................. vii
Other abbreviations.............................................................................................................. viii
Symbols ............................................................................................................................... viii
1. Introduction ......................................................................................................................... 1
1.1. The goal of the thesis ................................................................................................. 1
1.2. The Indo-Iranian languages ........................................................................................ 2
1.3. Methodology and hypotheses of previous literature .................................................. 2
1.3.1. Non-IE elements in IE languages ......................................................................... 2
1.3.2. Non-IE elements in Indo-Iranian .......................................................................... 4
1.3.3. Indo-Iranian origins: homeland and migration ..................................................... 6
1.3.4. The BMAC culture and the “Central Asian Substrate Hypothesis” ..................... 7
1.4. Research Questions .................................................................................................... 8
1.4.1. Identifying loanwords .......................................................................................... 9
1.4.2. Chronological layers ............................................................................................ 9
1.4.3. Structural characteristics .................................................................................... 10
1.5. Organization of the thesis ......................................................................................... 11
2. Proto-Indo-Iranian historical phonology........................................................................... 12
2.1. Proto-Indo-European phoneme inventory ................................................................ 12
2.2. Proto-Indo-Iranian phoneme inventory .................................................................... 13
2.3. Sound changes from PIE to PII ................................................................................ 13
2.3.1. Vowels ................................................................................................................ 13
2.3.2. Laryngeals .......................................................................................................... 15
2.3.2.1. Consonantal laryngeals ............................................................................... 15
2.3.2.2. Laryngeal vocalization ................................................................................ 15
2.3.2.3. Laryngeal metathesis .................................................................................. 18
2.3.2.4. Laryngeal accent shift ................................................................................. 19
2.3.2.5. Loss of intervocalic laryngeals ................................................................... 19
2.3.3. Stops ................................................................................................................... 20
2.3.3.1. Phonetics of the stops.................................................................................. 20
2.3.3.2. Phonetics of the palatals .............................................................................. 21
2.3.3.2.1. Indic ....................................................................................................... 21
ii
2.3.3.2.2. Iranian .................................................................................................... 22
2.3.3.2.3. Nuristani ................................................................................................ 22
2.3.3.2.4. Proto-Indo-Iranian ................................................................................. 22
2.3.3.3. Stops in contact with laryngeals.................................................................. 23
2.3.3.4. Palatalization of velars ................................................................................ 23
2.3.3.5. Bartholomae’s Law ..................................................................................... 24
2.3.4. The sibilant *s .................................................................................................... 25
2.3.5. Liquids ................................................................................................................ 25
2.3.5.1. General development .................................................................................. 25
2.3.5.2. Liquid + laryngeal clusters.......................................................................... 26
2.3.6. Nasals ................................................................................................................. 27
2.3.6.1. General development .................................................................................. 27
2.3.6.2. Nasal + laryngeal and nasal + resonant clusters ......................................... 28
2.3.7. Semivowels ........................................................................................................ 28
2.4. Relative chronology of Indo-Iranian sound changes ................................................ 29
3. Etymological analysis of proposed loanwords.................................................................. 31
3.1. Loanwords I: Pre-Proto-Indo-Iranian or early Proto-Indo-Iranian........................... 31
3.2. Loanwords 0: Proto-Indo-Iranian, but no further indication of date of borrowing .. 32
3.3. Loanwords II: Proto-Indo-Iranian, borrowed after certain sound changes .............. 48
3.4. Loanwords III: Post-Proto-Indo-Iranian................................................................... 52
3.5. Words with IE etymologies or insufficient evidence for borrowing ........................ 63
4. Chronological layers in early Indo-Iranian loanwords ..................................................... 73
4.1. Layer I: Pre-PII or early PII ..................................................................................... 73
4.2. Layer 0: PII (unspecified)......................................................................................... 73
4.3. Layer II: late PII ....................................................................................................... 74
4.4. Layer III: Post-PII borrowings ................................................................................. 75
4.5. Implications of chronological analysis ..................................................................... 76
5. Structural characteristics of Indo-Iranian loanwords ........................................................ 78
5.1. The CVCV̄CV-type ................................................................................................... 78
5.2. r/n-alternation ........................................................................................................... 79
5.3. The irregular correspondence Indic dh : Iranian t .................................................... 80
5.4. Non-initial mediae in clusters with *r or *n ............................................................ 81
5.5. Correlation between *i and affricates ....................................................................... 83
5.5.1. Dental stops ........................................................................................................ 85
5.5.2. Velar stops .......................................................................................................... 87
iii
5.6. The sequence *-ru̯- ................................................................................................... 88
6. Conclusions ....................................................................................................................... 89
6.1. Summary of main results.......................................................................................... 89
6.2. Identity of source languages ..................................................................................... 90
6.3. Implications for the Central Asian Substrate Hypothesis ........................................ 91
6.4. Directions for future research ................................................................................... 92
Bibliographical abbreviations .................................................................................................. 93
References ................................................................................................................................ 93
Appendix: Reference list of analyzed vocabulary.................................................................. 103
iv
Acknowledgements
First of all, I would like to thank Prof.dr. Sasha Lubtosky, who supervised this thesis and always
found time for our (sometimes lengthy) meetings in his busy schedule. I would also like to
thank Prof.dr. Guus Kroonen, who taught me crucial background knowledge relevant to this
thesis and offered me so many opportunities to realize my potential. I am grateful to Dr. Alwin
Kloekhorst, who supported my academic development during the ResMa programme in many
ways. Sincere thanks go to Dr. Tijmen Pronk, whose many excellent classes in Indo-European
linguistics I attended with pleasure.
I would also like to thank all the staff and students of LUCL, several of whom I consider
among my closest friends. Especially Oscar Billing, with whom I spent almost every day of
every week studying in the MPhil room, and enjoyed countless discussions about our work and
(thankfully!) other matters, has my sincere gratitude. I thank my parents Anne and Kjell for
their loving support throughout the years. Finally, I would like to express my loving thanks to
my partner Linde for moving with me to Leiden and building our lives together here.
v
Abstract
In this thesis, I study loanwords of unknown origin in Proto-Indo-Iranian and early Post-Proto-
Indo-Iranian. According to the Central Asian Substrate Hypothesis, Indo-Iranian speakers
migrated to Central Asia around 2000 BCE and came into contact with the agricultural BMAC
civilization, which resulted in a body of loanwords into Proto-Indo-Iranian, borrowed from the
language of the BMAC people. Following a methodology for identifying non-Indo-European
vocabulary in Indo-European languages, I argue that 74 out of 103 previously suggested
loanwords can plausibly be analyzed as loanwords (chapter 3). Only a handful of these may
have been borrowed from known languages. After establishing the relative chronology of Proto-
Indo-Iranian sound changes (chapter 2), I divide the 74 early Indo-Iranian loanwords into
chronological layers based on when they were borrowed (chapter 3-4). I argue that 21 words
were borrowed after the disintegration of Proto-Indo-Iranian. Moreover, I argue that many of
the remaining 53 loanwords that are reconstructable to Proto-Indo-Iranian were borrowed
towards the end of this stage. Finally, I integrate the chronological layers into my analysis of
structural characteristics of early Indo-Iranian loanwords and describe two new phonological
patterns of loanwords (chapter 5). The fact that many loanwords are shown to have been
borrowed in late PII or Post-PII, i.e. after Indo-Iranian speakers migrated to Central Asia, is
consistent with the timeline of the Central Asian Substrate Hypothesis. Second, the newly
discovered phonological characteristics provide additional support for the Central Asian
Substrate Hypothesis, since they increase the likelihood that most loanwords originate in the
same language.
vi
Abbreviations
Languages
A. = Ashkun N = Narym dialect (Selkup)
Akk. = Akkadian Nur. = Nuristani
Arab. = Arabic OAv. = Old Avestan
Av. = Avestan OBur. = Old Burušaski
Bactr. = Bactrian OCS = Old Church Slavonic
Bal. = Balochi OHG = Old High German
Baxt. = Baxtiari OIr. = Old Irish
Bur. = Burušaski ON = Old Norse
Elam. = Elamite OP = Old Persian
Finn. = Finnish Orm. = Ormuri
GAv. = Gathic Avestan Oss. = Ossetic
Germ. = German Par. = Parāčī
Goth = Gothic Parth. = Parthian
Gr. = Ancient Greek PCelt. = Proto-Celtic
Hitt. = Hittite PFU = Proto-Finno-Ugric
Hung. = Hungarian PFV = Proto-Finno-Volgaic
Ind. = Indic PGm. = Proto-Germanic
Ir. = Iranian Phl. = Pahlavi
Ke. = Ket dialect (Selkup) PIE = Proto-Indo-European
Khot. = Khotanese PII = Proto-Indo-Iranian
Km. = Kamviri PInd. = Proto-Indic
Kt. = Kataviri PIr. = Proto-Iranian
Lat. = Latin Pkt. = Prakrit
LAv. = Late Avestan Pre-II = Source language(s) of loanwords in
Lith. = Lithuanian Indo-Iranian
Mar. = Mari Pre-PII = Pre-Proto-Indo-Iranian
MHG = Middle High German Pto. = Pashto
MI = Middle Indic PToch. = Proto-Tocharian
MiP = Middle Persian PU = Proto-Uralic
MIr. = Middle Irish PUg. = Proto-Ugric
MoP = Modern Persian Rom. = Romani
vii
Ru. = Russian Other abbreviations
Sa. = Sami pl. = plural
Sam. = Samoyedic pres. = present
SCr. = Serbo-Croatian Rem. = Remarkable
Selk. = Selkup RV = R̥gveda
Skt. = Sanskrit sg. = singular
Sogd. = Sogdian
Symbols
Šu. = Šughni
> = regularly develops into
ToA = Tocharian A
< = regularly derives from
ToB = Tocharian B
>> = develops by analogy into, borrowed
Ty. = Tym dialect (Selkup)
into
W. = Waigali
<< = developed by analogy from, borrowed
V. = Vasi-vari
from
Wakh. = Wakhi
: = corresponds to
Veps. = Vepsian
Yagh. = Yaghnōbī
Yazg. = Yazgulyami
Y-M = Yidgha-Munji
aor. = aorist
AV = Atharvaveda
BL = Bartholomae’s Law
BMAC = Bactria-Margiana Archaeological
Complex
BrL = Brugmann’s Law
dat. = dative
dial. = dialectal
gen. = genitive
GL = Grassmann’s Law
inj. = injunctive
Irr. = Irregular
Lim. = Limited
nom. = nominative
viii
1. Introduction
The topic of this thesis is loanwords of unknown origin in early Indo-Iranian. In other words,
the thesis treats early Indo-Iranian words that are neither inherited from Proto-Indo-European
(PIE), nor innovated within Indo-Iranian based on inherited roots, but borrowed from languages
with which Indo-Iranian came into contact in prehistory. I use Pre-Indo-Iranian (Pre-II) as a
cover term for the unknown donor language(s) of early Indo-Iranian loanwords. Included within
the scope of “early Indo-Iranian” vocabulary is that of Proto-Indo-Iranian (PII). However, the
term also includes words shared between Indic1 and Iranian (and Nuristani) that cannot be
reconstructed to PII, but nevertheless must have entered the Indo-Iranian languages at an early
date, shortly after the disintegration of PII.
1.1. The goal of the thesis
The thesis has three main goals. The first goal is to establish which early Indo-Iranian words
are loanwords rather than inherited from PIE. With a few exceptions, all previously suggested
early Indo-Iranian loanwords are disputed, and alternative Indo-European (IE) etymologies
have been proposed. Therefore, an essential step of this study is to evaluate the proposals of
previous literature, to determine for each proposed loanword whether IE origin can be excluded
or not.
The second goal is to classify early Indo-Iranian loanwords into chronological layers. The
purpose is to determine how diverse vs. uniform the early Indo-Iranian loanwords are in terms
of relative time of borrowing. Based on established regular phonological correspondences, it
can be determined whether possible cognates in Indic and Iranian go back to PII or not. This
allows loanwords to be classified as PII or Post-PII.
However, the goal of the thesis is also to determine whether different chronological layers
of loanwords exist within PII. Based on the relative chronology of PII sound changes, I will
investigate whether, on the one hand, some words must have undergone certain PII sound
changes, and therefore must have been borrowed before these occurred, or, on the other hand,
some words cannot have undergone certain PII sound changes, and therefore must have been
borrowed after these occurred.
The third goal is to describe patterns in the phonology and morphology of early Indo-
Iranian loanwords. The purpose is to increase our understanding of the Pre-II language(s) with
which early Indo-Iranian came into contact. Besides being an intriguing question in itself, this
is a crucial step in the methodology of studying loanwords of unknown origin. If phonological
1
I use “Indic” instead of the more traditional term “Indo-Aryan”.
and morphological patterns in the loanword corpus are found, this in itself lends additional
support in favor of postulating an unknown source language. Since every language has a
phoneme inventory and a phonotactic system, phonological patterns among loanwords, which
cannot be explained by the phonology of the recipient language, imply that a foreign linguistic
system is fossilized behind them. Similarly, recurring morphological traits such as foreign
suffixes imply an underlying morphological system.
Below, previous literature on Indo-Iranian loanwords and related topics will be discussed,
followed by a more detailed formulation of research questions.
1.2. The Indo-Iranian languages
Indic and Iranian are the two major sub-branches of Indo-Iranian. For historical linguistic
purposes, the Old Indo-Iranian languages (Vedic) Sanskrit, Avestan and Old Persian are the
most important sources. However, since the Old Iranian corpus is limited, Middle and Modern
Iranian languages also play a crucial part in Indo-Iranian historical linguistics. Evidence from
Middle and Modern Indic languages is less commonly seen in the literature, but sometimes
preserves archaic features that Vedic Sanskrit had lost.
The Nuristani languages of Afghanistan are commonly considered to form a third sub-
branch of Indo-Iranian. However, the internal relationship between Indic, Iranian and Nuristani
remains unclear. Scholars have argued that Nuristani forms an intermediate subgroup with
Iranian (Mayrhofer, 1983) or Indic (Blažek & Hegedűs, 2012, p. 43), or that Nuristani was the
first branch to split off from PII, and that Indic and Iranian constitute a subgroup (Hegedűs,
2012, p. 145). As there is no general consensus, I will assume, for the purposes of this thesis,
that Nuristani is equally closely related to Indic as it is to Iranian.
1.3. Methodology and hypotheses of previous literature
1.3.1. Non-IE elements in IE languages
A series of publications have developed a methodology for identifying and systematically
studying non-IE vocabulary of unknown origin in ancient IE languages (cf. Kuiper, 1991, 1995;
Beekes, 1996, 2010; Schrijver, 1997; Lubotsky, 2001b). The methodology for identifying
prehistoric loanwords is based on five criteria:
1) Limited geographical distribution

2) Irregular phonological correspondences
3) Remarkable morphology
4) Remarkable phonology
5) Specific semantics
2
The first criterion applies if a word is restricted to one branch of IE or several branches
that are (or were in prehistory) spoken close to one another. This criterion stands out, since it is
in most cases a prerequisite for postulating borrowing in the first place. While in theory
straightforward, it is often the case that an IE etymology has been suggested, but that its validity
is disputed. Therefore, careful etymological analysis is always necessary.
The second criterion applies if a word is attested in two or more IE languages, but does
not show regular sound correspondences based on what we know from the inherited vocabulary,
and is therefore not reconstructable to PIE.
The third criterion applies if a word shows a derivational pattern or a suffix that is
marginal or absent in the inherited vocabulary. It is important to remember that loanwords
eventually adapt to the native morphology, generally following a productive pattern. Thus, a
nominal suffix *-bso- would be a clear indication of a loanword, in spite of the fact that it is
thematic, since *-bs- is clearly non-IE. In the case of verbs, loanwords are expected to belong
to a productive class, e.g. thematic rather than athematic.
The fourth criterion applies if a word contains phonemes or phonemic sequences that are
marginal or absent in the inherited vocabulary, e.g. two mediae in the root, the vowel *a
(depending on one’s views on PIE phonology).
The fifth criterion applies if a word is particularly “borrowable” due to its semantics. For
example, words for cultural phenomena as well as flora and fauna are more easily borrowed
than “basic” vocabulary (Tadmor et al., 2010).
As pointed out by Schrijver (1997, p. 296), none of these criteria is in itself decisive when
it comes to identifying loanwords. Limited geographical distribution may be accidental,
irregular phonological correspondences may be the result of analogy, remarkable morphology
and phonology may represent hitherto unknown inherited features, and words with specific
semantics may of course be inherited. Therefore, loanwords should ideally be identified based
on two or more of these criteria.
Besides the five criteria above, a crucial methodological principle of identifying and
studying loanwords of unknown origin is the notion of recurring irregularities and structural
characteristics (Schrijver, 1997, p. 296). In isolation, phonological and morphological features
can be used to identify loanwords, but when the same irregularity or foreign-looking structural
characteristic is found in several words, it drastically increases the plausibility that they are
loanwords. Recurring irregularities and structural characteristics can also indicate which
loanwords originate in the same language.
3
An aspect that set the studies cited above aside from other studies of lexical borrowing is
that the donor language, or “substratum” language as it is often called, is unrecorded and
unknown, and is thus only preserved in the loanwords themselves. The crucial step forward that
these studies represent, therefore, is the acknowledgement that such unknown prehistoric
languages can, and should, be studied in historical linguistics. To some, the notion of
postulating linguistic entities based on loanwords has seemed too methodologically problematic
to be taken seriously. Indeed, there is a certain risk that a new substrate language is used as a
magic wand each time a scholar is unable to explain an irregular correspondence or an obscure
lexeme. This would be similar to postulating an additional phoneme to explain a single cognate
set. However, this criticism is only valid when borrowing is used as an ad hoc explanation to a
particular problem. When, on the other hand, the methodology outline above is followed, the
situation changes, because if recurring irregularities and structural patterns are observed in
loanwords, postulating one substrate language can provide a solution to many unrelated
problems at once.
1.3.2. Non-IE elements in Indo-Iranian

Kuiper (1991) studied non-IE elements in Vedic Sanskrit. He identified hundreds of loanwords,
along with various morphosyntactic features which Sanskrit acquired in contact with non-IE
languages in South Asia. One of the most salient types of loanwords in Old Indic is the so-
called CVCV̄CV type, cf. Skt. caṣā̄́la- ‘knob’, trisyllabic words with a medial long vowel or
diphthong. This structure is rare in IE words, since these normally consist of a root and a suffix,
both usually monosyllabic.
In a series of publications, Witzel (1995; 1999a; 1999b; 2003; 2006; 2009) investigated
loanwords in Vedic Sanskrit, Indo-Iranian, and the linguistic (pre-)history of South Asia in
general. The main contribution of Witzel’s work lies in the early Indo-Iranian loanwords that
he proposes, as well as his discussion of some structural characteristics of these words.
Moreover, Witzel (2003) puts early Indo-Iranian loanwords in a broader perspective,
incorporating possible shared borrowings in other languages, such as Burušaski, Dravidian,
Anatolian, Greek, and languages of the Caucasus. A recurring irregularity in early Indo-Iranian
loanwords proposed by Witzel (2003, p. 45) is an r/n-alternation, argued to reflect dialectal
variation in the substrate language(s).
Although the significance of Witzel’s work should not be underestimated, it suffers from
the occasional inclusion of words with clear or likely PIE origin (e.g. PII *madhu- ‘honey’,
4
2003, p. 13) as well as the lack of methodological stringency in postulating common origins of
loanwords (cf. chapter 3 on *ganTuma-).
Lubotsky (2001b) systematically investigated vocabulary that is shared between Indic
and Iranian, but not found in other IE languages, i.e. words that fulfill the first criterion of the
methodology outlined above. In this material, he identified 55 loanwords. Most show regular
correspondences and can be reconstructed to PII, whereas others show irregular
correspondences. Additionally, 23 verbal roots isolated to Indo-Iranian were listed as possible
loanwords, although Lubotsky deemed it impossible to distinguish between inherited and
borrowed verbs (2001b, p. 310).
Lubotsky realized that several structural characteristics of PII loanwords are identical to
those of specifically Indic loanwords, as described by Kuiper (1991). These features include
the CVCV̄CV type, voiceless aspirates, frequent palatal stops, frequent clusters with *-s-, the
cluster *-ru̯-, and the suffixes -ig-, -pa-, and -h- (Lubotsky, 2001b, p. 305). Based on this
similarity, Lubotsky proposed that PII and Indic loanwords originate in the same language or
related languages, spoken in Central Asia on the one hand, and in the Punjab on the other. This
hypothesis will be further discussed below.
Furthermore, Lubotsky (2001b, p. 306) argued that loanwords with the irregular
correspondence Indic s : Iranian s were first borrowed into Indic and then transmitted to Iranian.
Kümmel (2017) collected Indo-Iranian vocabulary related to animal husbandry and
agriculture. He found that most terms for domesticated animals are inherited, whereas several
terms for cereals and other domesticated plants are not. Words in the latter group are potential
early Indo-Iranian loanwords.
As the literature review shows, non-IE vocabulary in Indo-Iranian has received some
attention from previous scholarship. However, it is not yet fully integrated into Indo-Iranian
lexicography, as is evident from the Etymologisches Wörterbuch des Altindoarischen (EWAia).
Although it sometimes acknowledges the possibility of borrowing, EWAia does not take into
account the systematic study of loanwords of unknown origin. Partly, this may be because some
of the aforementioned studies were not yet available at the time of publication, but the
dictionary also shows skepticism towards such proposals. This is expressed by the employment
of ad hoc explanations, such as how the s of Skt. sūcī̄́ ‘needle’ is said to be analogical from sīv-
‘to sew’, in order to explain the irregular correspondence to Ir. *ćūkā- / *ćūčī- (EWAia II, p.
739). In this case, assuming borrowing is preferable, since the Indo-Iranian word for ‘needle’
fulfills three of five criteria of a loanword: limited geographical distribution, irregular
5
phonological correspondences, and specific semantics. In other cases, EWAia simply dismisses
proposed borrowings as “unnecessary” (II, 241) or “implausible” (II, p. 151).
EWAia only considers borrowing as a possibility when a known source language exists.
For some words, an Austroasiatic, Dravidian or Uralic source has been suggested. However,
since these languages are known to have borrowed from Indo-Iranian, the direction of
borrowing is often difficult to prove.
1.3.3. Indo-Iranian origins: homeland and migration

Studying prehistoric language contact is one of the main pieces of linguistic evidence for
prehistoric migrations and language spread. A loanword from one language to another suggests
that speakers of the donor language and recipient language were in contact, which usually2
presupposes geographical proximity of the speaker communities. However, when it comes to
loanwords of unknown origin, the situation is somewhat reversed: the prehistoric location of
the recipient language delimits the possible locations of the donor language(s). Therefore, a
short review of the current views on the origin of the Indo-Iranian languages and their speakers
is due.
The Indo-Iranian branch originates in PIE. The question of when and where PIE was
spoken has generated two fundamentally different hypotheses. The Steppe Hypothesis places
the IE homeland in the nomadic Yamnaya culture on the Pontic-Caspian Steppe around 3500-
3000 BCE (Mallory, 1989; Anthony, 2007). This view has been rivalled by the Anatolian
Hypothesis (Renfrew, 1987), which claims that Proto-Indo-European dispersed with the spread
of agriculture from Anatolia around 7000 BCE.
Recently, strong evidence for large scale migrations from Yamnaya steppe populations
into Europe and Asia was offered by geneticists (Haak et al., 2015), favoring the Steppe
Hypothesis. The Steppe Hypothesis is also favored by the linguistic evidence, since PIE had
terminology for wheeled vehicles, which were invented after 4000 BCE, consistent with the
chronology of the Yamnaya culture (Anthony & Ringe, 2015). From the IE homeland on the
Pontic-Caspian Steppe, Indo-Iranian speakers eventually migrated all the way to South and
Western Asia, as evidenced by the high degree of Steppe Ancestry in the DNA of modern Indo-
Iranian speaking populations (Damgaard et al., 2018).
Kuz’mina (2007) approached the question of the Indo-Iranian migration and homeland
from an archaeological perspective, incorporating linguistic and anthropological evidence to
2
In some cases, words are more mobile, as it were, than the speakers who use them; Wanderwörter can spread
from one community to the other, without the original source language being in contact with all subsequent
recipients.
6
some extent. She argued that prehistoric Indo-Iranian speakers inhabited the Sintashta (2100-
1800 BCE) and Andronovo (2000-900 BCE) cultures. By retracing cultural development in the
archaeological record, Kuz’mina (2007, p. 205) found that the Economic and Cultural Type
(ECT) of the Indo-Iranian speaking Sauromatians and Saka cultures descends directly from the
Andronovo cultures, which in turn succeeded the older Sintashta culture. The pastoral Sintastha
culture, situated to the south-east of the Ural Mountains, is thus a plausible Indo-Iranian
homeland. This hypothesis is also supported by the many PII loanwords into (Proto-)Uralic
(Koivulehto, 2001). An archaeolinguistic argument is that chariotry, for which several terms
are reconstructable to PII (Witzel, 2001), originated in the Sintastha culture (Kuznetsov, 2006).
From their homeland in the Sintashta culture around 2000 BCE, Indo-Iranian speakers
spread southwards to the areas in which Indo-Iranian languages are still spoken today.
Moreover, Indo-Iranian languages continued to be spoken in Central Asia, with some groups
(e.g. the Alans) spreading westwards to Europe.
1.3.4. The BMAC culture and the “Central Asian Substrate Hypothesis”
The Bactria-Margiana Archaeological Complex (BMAC) denotes a Central Asian Bronze Age
civilization east of the Caspian Sea, to the south of the Andronovo horizon. With its origins in
the first half of the 3rd millennium, the BMAC civilization was at its peak ca. 2400-1700 BCE
(Francfort, 2005, p. 260). Around its fortified settlements, the BMAC people practiced
irrigation farming, cultivating wheat, barley, lentil, pea, grass pea, chick pea, grape, apple and
flax (Spengler et al., 2014).3 Domesticated animals include cattle, sheep, camels, pigs and
donkeys (Witzel, 2000, p. 4). Especially interesting is the archaeological evidence of groups of
mobile pastoralists, who lived outside of the fortified settlements, and whose animals may have
grazed the fields of the farmers after the harvest (Spengler et al., 2014, p. 808, 816). According
to Spengler et al. (ibid.), the fact that animal dung was used as fuel by the farmers indicates
non-hostile contacts between the groups. Since no written documents have been excavated from
the BMAC civilization, the identity of its language(s) is unknown.
Witzel (2003) and Lubotsky (2001b) have elaborated the hypothesis that most loanwords
of unknown origin in PII originate in an unknown language of the BMAC civilization. I refer
to this as the “Central Asian Substrate Hypothesis”. The hypothesis combines archaeological
and linguistic arguments into a plausible scenario.
3
Whether millet was cultivated is uncertain. Spengler et al. (2014, p. 817) did not find evidence for millet in
southern Central Asia earlier than the Iron Age, but notes that this could be accidental.
7
First, if the Sintashta culture is accepted as the PII homeland, the early Indo-Iranians must
have come into contact with BMAC groups as they spread southwards, their presence being
attested in the Near East in the 16-15th centuries BCE (Mallory, 1989, p. 38), perhaps even as
early as the 18th century (Kroonen et al., 2018, p. 12). Archaeological evidence for steppe
influence is seen in BMAC pottery (Witzel, 2000, p. 7). Moreover, as noted above, there is
evidence for temporary settlements of pastoralists in the BMAC area (Salvatori, 2008, p. 64),
which could very well have belonged to Indo-Iranian speakers from the pastoral Andronovo
cultures. Furthermore, the spread of Indo-Iranian to South and Western Asia has been connected
to the BMAC cultural influence in these areas in the second half of the 2nd millennium (Witzel,
2000, p. 8). According to Mallory (1998), the Indo-Iranians eventually assimilated to the culture
of the BMAC and transmitted it to the south, which would explain why there is little direct
cultural influence from the steppe south of the BMAC. Language contact between Indo-Iranian
speaking steppe populations and BMAC farmers is thus likely.
Second, the semantics of some early Indo-Iranian loanwords make a BMAC origin likely.
Witzel (2003, p. 25) mentions *Hustra- ‘camel’, *kHara- ‘donkey’ and *išt(i)- ‘brick’ as the
clearest cases, since the camel and donkey were present in the BMAC culture, but not in the
steppe where Indo-Iranian originates, and since BMAC settlements are built with bricks.
Lubotsky (2001b, p. 307) mentions *i̯ au̯īi̯ ā- ‘canal’, which can be connected to the irrigation
farming technique of the BMAC, as well as several other terms referring to building technology.
Thus, the Central Asian Substrate Hypothesis is supported both by the archaeological
evidence for contact between steppe populations and BMAC populations, and by the fact that
several loanwords seem to reflect BMAC material culture. However, many of the previously
proposed early Indo-Iranian loanwords cannot be directly compared to the BMAC culture, since
they denote abstract concepts or other notions not visible in the archaeological record. These
are instead hypothesized to originate in the BMAC language(s) simply because they are PII
loanwords without a known source.
1.4. Research Questions
As we have seen, previous research has identified a group of early Indo-Iranian loanwords,
discovered a number of structural characteristics of this group of words, and put them in an
archaeolinguistic context by proposing a plausible language contact scenario. The linguistic,
rather than archaeological perspective, is the main focus of this thesis, i.e. the loanwords
themselves and the language(s) they may have come from.
8
1.4.1. Identifying loanwords
The words recognized as early Indo-Iranian loanwords differ depending on the author, and it is
therefore necessary to reevaluate previous proposals. This serves two purposes: firstly, insights
from Lubotsky (2001b), Witzel (2003) and Kümmel (2017) will be synthesized to provide a
more complete picture of the material. Secondly, since some works are inexplicit with regards
to methodology, it is unclear whether all proposed loanwords have been analyzed under the
same criteria. By applying the methodology outlined in 1.3.1 to all previously proposed
loanwords, the analysis will become more explicit and the results more uniform.
Unlike in previous literature, verbs will also be taken into account. In principal, verbs can
be analyzed according to the same methodology as nouns. However, as the extensive Indo-
Iranian verbal morphology requires verbs to be analyzable as monosyllabic roots, borrowed
verbs often require more adaptation to the native system than nouns, and are thus more difficult
to differentiate from the inherited vocabulary. The presence of archaic derivations, e.g. nasal
infix present, strongly suggest IE origin. Borrowing will only be considered when archaic
derivations are absent.
In some cases, a known source language of a loanword has been proposed. Such proposals
will be evaluated, since a plausible known source would be a strong argument for postulating
borrowing.
1.4.2. Chronological layers
An aspect of early Indo-Iranian loanwords that has not been systematically taken into account
in earlier literature is the time of borrowing. The development of Indo-Iranian can be divided
into chronological stages following its separation from the rest of IE: Pre-PII, PII, and Post-
PII.4 Generally, the Central Asian Substrate is identified with loanwords in PII (Lubotsky,
2001b, p. 301). Yet, Witzel (2003) also assigns Post-PII loanwords, some of which may be very
late borrowings, to the Central Asian Substrate. This is problematic. Chronological stages
reflect temporal development, but indirectly often reflect geographical movement, since a
principal cause for the disintegration of PII must have been the geographical separation of
speaker communities that would later become Indic, Iranian and Nuristani. Therefore,
loanwords in different chronological layers should not a priori be lumped together.
Lubotsky reconstructs loanwords with irregular correspondences to PII, arguing that the
proto-language was a continuum of differentiated dialects that nonetheless underwent shared
4
After this, of course, follows Proto-Iranian and Proto-Indic and their respective historical development until the
modern period, but this goes beyond the scope of this discussion.
9
innovations (2001b, p. 302). On the one hand, it is true that the notion of clearly definable
linguistic entities such as “PII” does not fully capture the complexity of language development.
On the other hand, by not assuming a uniform PII, methodological stringency is decreased,
since the importance of regular sound change is downplayed. The possibility of variation within
PII should not be excluded, but irregular correspondences in loanwords nevertheless suggest
that, at the time of borrowing, the linguistic community was disintegrating, implying greater
distance between speaker groups and their dialects. Therefore, regular vs. irregular
correspondences is a relevant point of division when it comes to loanwords, and is one of the
main research questions of the thesis.
More specifically, I will investigate which loanwords can be reconstructed to PII, and
which cannot, based on current knowledge about the regular correspondences between the
branches of Indo-Iranian. Moreover, I will explore whether Post-PII loanwords were borrowed
independently from the same source, transmitted from one branch of Indo-Iranian to the other,
or borrowed from different sources. This will help to determine the likelihood of a Central
Asian Substrate origin of early Indo-Iranian loanwords.
Another question that will be addressed is whether there are different chronological layers
of loanwords within PII. The inherited vocabulary of Indo-Iranian has undergone all PII sound
changes, but this is not necessarily true for PII borrowings. If the structure of a loanword is such
that it cannot go back to Pre-PII, it must have been borrowed at a later stage. Besides allowing
for a more precise chronological stratification of early Indo-Iranian loanwords, showing that a
loanword cannot go back to Pre-PII would also provide a strong argument in favor of a non-IE
origin. Conversely, if a word must have gone through certain PII sound changes, it must have
been borrowed at an earlier stage.
The investigation of chronological layers hinges on establishing a relative chronology of
PII sound changes. Lubotsky (2018) proposes a chronology, but since several aspects of Indo-
Iranian historical phonology are debated, key points must be reviewed and revised.
A final purpose of dividing early Indo-Iranian loanwords into chronological layers is to
improve the analysis of structural characteristics of borrowed vocabulary (cf. below).
1.4.3. Structural characteristics
Previous literature has proposed several phonological and morphological characteristics of
early Indo-Iranian loanwords, which corroborate the hypothesis that many words originate in
the Central Asian Substrate. In this thesis, I will reevaluate previously proposed structural
characteristics, and examine the material for additional patterns. If more characteristics are
observed, it would strengthen the Central Asian Substrate Hypothesis.
10
Furthermore, chronological layers of early Indo-Iranian loanwords will be incorporated
into the study of structural characteristics. I will attempt to determine whether structural
characteristics hold for words within the same layer, and then compare the layers to each other.
Structural differences between the layers would point to multiple source languages, whereas
similarities would point to contact with the same language (or related languages) over an
extended period of time. This process serves to further test the Central Asian Substrate
Hypothesis, and could provide new insights into the Pre-Indo-Iranian linguistic landscape of
Central, South and Western Asia.
It is important to kept in mind that structural differences between chronological layers
may be due to changes in the recipient language (Indo-Iranian) rather than differences in the
donor language(s). Whether this is likely or not depends on the feature in question. For example,
the CVCV̄CV type looks equally foreign in PII as it does in Indic. On the other hand, voiceless
aspirates are less ‘marked’ in Indic than in PII, since such stops are fully integrated in the
phonology of the former but not of the latter.
A caveat is that loanwords attested in a single branch may still be inherited from PII.
Unless there is a clear argument against this (i.e. the attested phonological structure cannot
develop regularly from PII), it is difficult to disprove. The basic methodological principle,
however, is to only reconstruct loanwords with cognates in Indic and Iranian to PII.
Another caveat is that loanwords are only indirect attestations of the source language,
which have undergone adaptation to the structure of the recipient language. Accordingly, the
phonology and morphology of loanwords must have undergone some kind of change as the
loanwords are “nativized”, i.e. integrated into the Indo-Iranian linguistic system (Hock, 1991,
p. 390). However, structural patterns of the source language may still be carried over to the
recipient language in a (more or less) regular way (Hock, 1991, p. 394).
1.5. Organization of the thesis
The thesis is structured as follows. Chapter 2 describes the most important PII sound changes
and arranges them in a relative chronology. In chapter 3, previously proposed early Indo-Iranian
loanwords are analyzed etymologically, taking the relative chronology established in chapter 2
into account. Chapter 4 summarizes and discusses the findings regarding the chronological
layers of early Indo-Iranian loanwords. In chapter 5, previously proposed structural
characteristics of Indo-Iranian loanwords are discussed, and new patterns are presented.
Chapter 6 summarizes and concludes the thesis.
11
2. Proto-Indo-Iranian historical phonology
In this chapter, the historical development of PII phonology will be discussed. The goal is to
describe the sound changes from PIE to PII, focusing on their relative chronology. When
possible, the phonetic realization of PII phonemes will be described. This treatment will serve
as a basis for establishing chronological layers and analyzing structural characteristics of early
Indo-Iranian loanwords. Below, the PIE and PII phoneme inventories are given for reference,
as reconstructed by Beekes (2011, p. 119) and Lubotsky (2018, p. 1875), respectively.
2.1. Proto-Indo-European phoneme inventory

labial dental palatal velar labiovelar
stops p t ḱ k kw
(b) d ǵ g gw
bh dh ǵh gh gwh
fricative s
laryngeals h1 h2 h3
liquids l r
nasals m n
semivowels u i
vowels5 e o
ē ō
5
Many scholars reconstruct a third vowel *ā̄̆, but there are many arguments against this (Lubotsky, 1989).
12
2.2. Proto-Indo-Iranian phoneme inventory
labial dental palatal velar
stops p t k
b d g
bh dh gh
affricate ć č
j̄́ ǰ
j̄́h ǰh
fricative s
laryngeal H
liquid r
nasals m n
semivowels u i
vowels a ā
Common allophones:
/s/ = [ s, š, z, ž ]
/i/ = [ i, i̯ ]
/u/ = [ u, u̯ ]
/r/ = [ r, r̥, l, l̥ ]
2.3. Sound changes from PIE to PII

2.3.1. Vowels
The PII vowel system was reduced in comparison to its PIE predecessor, showing a merger of
non-high vowels as *a and *ā. However, before the PII vowel merger, two important sound
changes must have occurred: Brugmann’s Law (BrL), i.e. lengthening of *o in open syllables,
and palatalization of velars.6 The chain of vowel developments can be described as follows:
6
In reality, the palatalization of velars was allophonic until it was phonologized as a result of the vowel merger.
13
Table 1. Vowel changes from PIE to PII
PIE → Pre-PII → PII: BrL → PII: vowel merger
*e *e *e *a
*ē *ē *ē *ā
*o *oC. *o *a
*o.C *ō *ā
*ō *ō *ō
*h2-3e *h2-3a *h2-3a *Ha
*eh2-3 *ah2-3 *ah2-3 *aH
In Pre-PII, laryngeals *h2-3 phonetically color an underlying *e. The resulting vowel must be
assumed to have been different from both *e and *o for two reasons: 1) *h3e did not undergo
BrL (Lubotsky, 1990), i.e. it was not phonetically identical to *o,7 and 2) *eh2-3 did not
palatalize a preceding velar (Ollet, 2014). Although the evidence that *eh2-3 was a non-
palatalizing context is scarce, there are no secure cases where *eh2-3 does palatalize a preceding
velar (cf. 2.3.3.4.). Thus, it seems likely that phonetic coloring did take place in Pre-PII. Ollet
(2014, p. 163) argues that this contradicts Lubotsky’s claim that *h3e did not undergo BrL, but
this is a false dilemma. It is perfectly possible that *h3e was phonetically different from *o, e.g.
*[h3a]. Thus, there is no contradiction between acknowledging laryngeal coloring as a phonetic
rule in Pre-PII and accepting that *h3e did not undergo BrL.
In spite of the above, evidence for phonological coloring of *h2-3e > *Ha will come from
the relative chronology of PII sound changes, which will be discussed in section 2.4.8
Note that under the scenario above, *h2 and *h3 appear not to have been phonemically
distinct in Pre-PII. One may object to the idea that *h2 and *h3 colored *e to the same vowel
*a, despite being phonemically distinct and giving different coloring effects (*h2a vs. *h3o)
elsewhere in Indo-European. Indeed, the only reason for arguing that *h3e yielded the same
vowel as *h2e in Pre-PII is to explain why *h3e does not undergo BrL, while at the same time
being distinct from *e. However, if Kloekhorst (2018, p. 89) is right that *h3 was the labialized
variant of *h2, an additional argument could be that the loss of distinction between *h3 and *h2
7
One of Lubotsky’s examples, Skt. ắvi- ‘sheep’ < PIE *h3eu̯i- is challenged by ToB āuwi, which seems to point
to *h2e/ou̯i- (Kim, 2000). However, other examples like Skt. ánas- ‘cart’ ~ Lat. onus- ‘burden’ < *h3enos- and
Skt. ápas- ~ Lat. opus ‘work’ < *h3epos- remain convincing.
8
In short, there is evidence that the merger of the laryngeals preceded BrL. Therefore, the colored vowel is
phonologized as *a before the general PII vowel merger.
14
is parallel to the loss of labialization of *Kw, which in Indo-Iranian merges with its non-labial
counterpart *K.
In the context *V̄̆HC, the vowel underwent compensatory lengthening when the laryngeal
was lost, cf. Skt. mā̄́tar- ~ Av. mātar- ‘mother’ < PII *maHtar- < PIE *meh2ter-.
2.3.2. Laryngeals
In this section, the development of the PIE laryngeals in PII will be described. The effect of
laryngeals on adjacent vowels and consonants, however, are discussed in the sections treating
those phonemes.
2.3.2.1. Consonantal laryngeals

The PIE laryngeals eventually merge into PII *H. This is deduced by the fact that the laryngeals
give identical reflexes in terms of vocalization, deglottalization of mediae, Lubotsky’s law, and
Indic aspiration.9 Due to its interaction with the mediae (cf. 2.3.3.3.), it is likely that PII *H was
a glottal stop. Since consonantal laryngeals seem to be preserved in some positions in Iranian
(Kümmel, 2018), phonemic *H must have been retained throughout PII.
2.3.2.2. Laryngeal vocalization

Laryngeal vocalization (LV) changed interconsonantal laryngeals to *i. While LV eventually
affects most interconsonantal laryngeals in Indic, for PII it is only securely reconstructable for
final syllables, i.e. *H > *i / C_(C)# (Lubotsky, 2018, p. 1882). Among the examples are Skt.
jáni- ~ OAv. jaini- ‘wife’ < PII *ǰani- < *genH- and Skt. sádhiṣ- ~ LAv. hadiš, OP hadiš ‘seat,
residence’ < PII *sadhis- < *sedHs-.
Some examples of LV in initial syllables can be reconstructed for PII, viz. Skt. pitár-, OP
pitar-, OAv. dat.sg. piθrē ‘father’ < PII *pitar- < *pHtar-, but OAv. nom.sg. ptā- does not
reflect a vocalized laryngeal. Another example is Skt. aśiṣat, OAv. sīšōit̰ ‘to instruct, command’
< PII *ćiša- < *ćHsa- < *ḱh2s- (Lubotsky, 2018, p. 1883).
In middle syllable position, Indic (and perhaps Nuristani) shows LV, but Iranian does not
(Ravnaes, 1981, p. 261). A reasonable hypothesis is thus that medial laryngeals remained
consonantal in PII, were lost in Proto-Iranian, but vocalized in Proto-Indic. However, there is
one case of PII LV in a medial syllable: Skt. duhitár- ~ Nur. Prasun lüšt ‘daughter’ < PII
9
Skt. piba- ‘to drink’ < PIE *pi-ph3-e- shows potential evidence that *h3 did not merge with the other laryngeals
in PII (Kümmel, 2018, p. 163). However, since *ph3 > *b seems to be a PIE development, reflected also in Lat.
bibō ‘to drink’ and OIr. ebait ‘they drink’, the laryngeal may have been lost at an early stage. An alternative
explanation of *pi-bh3-e- is that the Pre-PIE root was *beh3-, and that *b > *p in initial position (Kortlandt,
1996). In that case, the development to *pi-bhHe- may have been prohibited by the constraint against
tautosyllabic tenues and aspiratae. In general, the lack of clear examples of aspiration/deglottalization by *h3
may be due to the rarity of the phoneme in PIE.
15
*dhuǰhitar- < *dhugHtar- < PIE *dhugh2ter- (Lubotsky, 2018, p. 1883). PII *i is reconstructed
to account for the palatalization of *gh. Conversely, OAv. dugədar- and LAv. duxtar show no
trace of LV. To avoid projecting LV back to before PII palatalization, Kümmel (2016b, p. 220)
suggests that Skt. duhitár derives directly from *dughitar- with debuccalization of *gh > h, and
that the palatalization in Prasun lüšt is secondary, which is ad hoc.
Previous theories on Indo-Iranian LV can be divided into two types: those advocating a
single LV process and those operating with two separate LV processes.
The former type of theory assumes that interconsonantal laryngeals were vocalized once
in PII. To explain the lack of vocalic reflexes in medial syllables in Iranian, the proponents of
this theory introduce various phonological rules and analogical processes. Schmidt (1973, p.
54) proposed that laryngeals were lost in the sequence *CHCC, perhaps already in PIE. This
sound change would have given rise to paradigmatic alternations like *dhugh2ter- / *dhugtr- and
*ph2ter- / *ptr-, which according to Schmidt produced the attested Indo-Iranian paradigms
through levelling of different stems in Indic and Iranian. A reversed version of this theory,
whereby a laryngeal was vocalized in the sequence *CHCC is advocated by Beekes (1981, p.
285).
Lipp (2009, p. 356) argued that LV only affected pretonic laryngeals, *-CHC-ˊ. Since
laryngeals in initial syllables are by definition pretonic, an alternation between pretonic LV and
post-tonic non-LV would only be visible in medial and final syllables. It is immediately clear
that Lipp’s accent rule did not operate in final syllables, where LV is firmly attested, since such
laryngeals are by default post-tonic. Thus, the rule could in fact only predict the outcome of
laryngeals in medial syllables (i.e. CV̄́ CHCV > CV̄́ CCV, but CVCHCV̄́ > CVCiCV̄́ ).
Lipp’s accent rule would explain cases like Skt. déva-tta- ‘God-given’ < *dai̯ u̯á-dH-ta-
(no LV) vs. Skt. duhitár- < *dhugHtár- (LV). However, the expected regular outcome of PII
*-dH-ta- ‘given’ would be Skt. **-ddha-, since *d would be deglottalized to *dh by the
laryngeal,10 with subsequent progressive assimilation by Bartholomae’s Law. Therefore,
Skt. -tta- could be a secondary formation from the root dā, rather than a regular outcome. In the
latter case, Skt. duhitár- seems to follow the accent rule, but to explain the lack of LV in OAv.
dugədar- etc., Schmidt’s rule of laryngeal loss *CHCC > *CCC must be invoked. Another
problematic case is LAv. and OP Vištā(spa)- < *u̯i-sH-ta- ~ Skt. víṣita- < *u̯í-sH̥-ta-. The Skt.
form shows LV of a post-tonic laryngeal. It might of course be explained away as secondary,
but at face value, it suggests that accent did not influence vocalization. Ultimately, Lipp’s
10
Cf. section 2.3.3.3.
16
accent rule cannot replace any of the other phonological rules, and does not explain enough
material to be considered likely. In fact, there are so many exceptions that the accent rule can
just as well be reversed with Beekes (1981, p. 285), who assumed that only post-tonic *H was
vocalized.
All theories that operate with a single PII laryngeal vocalization must assume extensive
processes of analogy to explain the lack of LV in Iranian middle syllables. In the case of Lipp
(2009), unexpected reflexes of LV in Indic must be attributed to accent shifts or analogical
extensions. While it is not unexpected that paradigmatic alternations in PII would have been
levelled in the separate branches, the fact that Iranian always shows the forms without LV
suggests a phonological conditioning rather than analogical levelling.
The second type of theory, advocated by Kuiper (1976, p. 243) and Kümmel (2016b, p.
221), operates with two distinct processes of LV, one in PII that only affects laryngeals in initial
and final syllables, and another in Indic that affected remaining interconsonantal laryngeals.
This theory better accounts for the lack of LV in medial syllables in Iranian, and for the fact
that *H̥ can yield either ī or i in Indic, whereas only i is found in Iranian. The one clear
counterexample, Skt. duhitár-, however, is not easy to explain away. Thus, the theory of two
LVs has the advantage of being capable of explaining the divergent treatment of laryngeals in
middle syllables. The disadvantage is that the laryngeal in *dhugHtar- must have been affected
by the first LV, despite being in a medial syllable.
In addition to the above, there are different opinions regarding the phonetics of LV. Most
scholars, including Mayrhofer (1986), Schmidt (1973), Werba (2005), Lipp (2009) and
Kümmel (2016b) have assumed that LV was realized as an anaptyctic vowel adjacent to the
laryngeal, i.e. *Hi or *iH. Although the ‘anaptyxis theory’ is reasonable at first glance, since
laryngeals, being obstruents, cannot be vocalized in the same way as resonants, it faces
methodological problems and cannot explain the attested material.
The first problem regards the assumption that *H̥ = *Hi. The vowel is posited after the
laryngeal to explain the ‘double reflex’ of *H̥, yielding both aspiration of a preceding stop and
the vowel i, e.g. *dhughitar- < *dhugHitar- < *dhugHtar-. However, this cannot explain why PII
*pH̥tar- > *pHitar- ‘father’ becomes Skt. pitár- ~ OP pitā- and not **phitár- and **fitā-,
respectively. Mayrhofer (1986, p. 138) therefore assumed that *H̥ > *iH in initial syllables.
However, this does not explain why *iH did not yield long *ī, like other *VHC sequences. Byrd
(2015, p. 33) argued that the anaptyctic vowel in *iH was a non-moraic vowel that was not
lengthened by laryngeals like other vowels. However, since *H̥ elsewhere merges with PII *i,
there is no independent reason to assume that *iH was an “extremely short vowel” (Byrd, 2015,
17
p. 31). Kümmel (2016b, p. 223) instead assumes laryngeal loss in initial syllables, arguing that
the anaptyxis of *i in these cases is a separate process that affected clusters of three consonants,
e.g. *pHtr- > *ptr- > *pitr-. However, in the case of Skt. aśiṣat ~ OAv. sīšōit̰ < PII *ćHsa-,
there would be no phonetic motivation for anaptyxis after laryngeal loss, since *ćsa- is a
licensed cluster in PII. Thus, in Kümmel’s scenario, core evidence like *ćisa- < *ćHsa- must
be explained away as analogical.
Secondly, it is problematic to assume that a single phoneme *H̥ yielded aspiration and a
vowel *i, e.g. a ‘double reflex’. Moreover, the data does not require it. In the case of voiceless
aspirates, there are words where a laryngeal appears to yield both aspiration and i, cf. Skt.
pṛthivī̄́- ‘earth’ < *pr̥tHu̯ī- and pathíbhiḥ ‘with/from the road’ < *pn̥tH-bhis. Yet, in all such
cases, the aspiration may have been introduced by analogy from related words where the
laryngeal was not interconsonantal, e.g. Skt. pṛthú- < *pr̥tHu-. In the latter case, levelling of -th-
must have occurred in nom.sg. pánthās anyway, since LAv. paṇtā̄̊ preserves the regular reflex
of PII *pantaH-s. Thus, there is no reason to believe that the aspiration in Skt. pathíbhiḥ is old.
To my knowledge, the only case where analogical extension of aspiration is impossible is Skt.
átithi- ‘guest’ < *atH̥tHi-, and here aspiration is conspicuously absent.11
In the case of secondary voiced aspirates, Lubotsky (2018, p. 1882) has offered an
alternative explanation: in a cluster *DH, like in PII *dhugHtar-, mediae were not aspirated, but
deglottalized and thus merged with *Dh (cf. 2.3.3.3.). As this is a dissimilatory process, it does
not preclude that the laryngeal was later vocalized to *i. When LV occurred, the media had
already lost its glottalic feature.
Thus, in terms of phonology, LV should be viewed as *H > *i, without an intermediate
biphonemic stage. While the exact conditioning and chronology of Indo-Iranian LV remain
unclear, it is evident that it occurred in initial and final syllables in PII, as well as in *dhugHtar-.
2.3.2.3. Laryngeal metathesis

In the sequence CHiC and CHuC, the laryngeal and semivowel underwent metathesis, cf. Skt.
bhūtá- ~ LAv. būta- ‘become’ < *bhuHta- < *bhh2uto- and Skt. pītá- ‘drunk’ < *piHta- <
*ph3ito- (Lubotsky, 2018, p. 1884). That the laryngeal originally preceded the semivowel is
shown by Ru. (dial.) bávit’ ‘to linger’, Goth. bauan ‘to live, dwell’ < *bheh2u- (Kortlandt, 1986,
p. 90) and Skt. pāyáya- ‘cause to drink’ < *paHi̯ -ai̯ a- < *poh3i̯ -ei̯ e-.12
11
Unless the unaspirated t is explained by Grassmann’s Law; in that case, Skt. átithi- cannot be used as an
argument.
12
The semivowel in *bheh2u-, *peh3i- may be a fossilized suffix, if these roots derive from PIE *bheh2- ‘to shine’
and *peh3- ‘to drink’.
18
In addition to Indo-Iranian, examples of laryngeal metathesis exist in several IE branches.
Gr. πῖθι ‘drink!’ < *pih3-dhi seems to show the same development as in Skt. pītá-. Next to Goth.
bauan, ON búa ‘to dwell’ probably continues a metathesized zero-grade *bhuh2- (Kroonen,
2013, p. 71). Lat. grūs ‘crane’ < *ǵruh2- most likely shows metathesis (Vaan, 2008, p. 274),
since SCr. žȅrãv shows that the word was originally a u-stem (Kortlandt, 1985, p. 120). The
Anatolian evidence is scarce, but Hitt. šuḫḫa-i / šuḫḫ- ‘to scatter’ < *suh2-, if from the same
root as išḫuu̯ai-i / išḫui- ‘to throw, scatter, pour’ < *sh2u-oi- (Kloekhorst, 2008, p. 892), seems
to show a metathesized variant of *seh2u-.
Thus, laryngeal metathesis seems to have been a PIE, in any case Pre-PII, development.
2.3.2.4. Laryngeal accent shift

Lubotsky (1988, p. 50) observed that i- and u-stems derived from roots with laryngeals after
the root vowel (*Ce(C)H(C)-) are generally oxytone in Sanskrit. By contrast, roots without
laryngeals are barytone or oxytone depending on their ablaut pattern. To explain this, Lubotsky
postulated a laryngeal accent shift from the root to the suffix in PII, i.e. *Cé(C)H(C)-U- >
*Ce(C)H(C)-Ú-.
Two exceptions to this rule, Skt. íṣṭi- ‘sacrifice’ and yájyu- ‘eager to sacrifice’ < PIE
*Hieh2ǵ-, derive from a root where the laryngeal was lost before *-ǵC- with Lubotsky’s Law
(cf. 2.3.3.3.). Therefore, it seems that the laryngeal accent shift was posterior to Lubotsky’s
Law.
Two other counterexamples, Skt. bhū̄́mi- ‘earth’ < *bhúH-mi- and Skt. bhū̄́ri- ‘much’ <
*bhuH-ri-, were explained by assuming that the laryngeal accent shift preceded laryngeal
metathesis (Lubotsky, 1988, p. 53). Since the laryngeal originally preceded the vowel in the
root *bheh2u-, this would explain the barytonesis of Skt. bhū̄́mi- and bhū̄́ri-. However, as seen
in the previous section, laryngeal metathesis may have been a PIE development, whereas the
laryngeal accent shift clearly is not. Therefore, the accentuation of bhū̄́mi- and bhū̄́ri- must be
explained otherwise. Although no good alternative is available at present, bhū̄́mi- and bhū̄́ri-
should be treated as exceptional, since the laryngeal accent shift explains a clear majority of the
available evidence.
2.3.2.5. Loss of intervocalic laryngeals

In intervocalic position, laryngeals were lost in PII (Lubotsky, 1995, p. 229). However,
whenever the sequence -VHV- was separated by a morpheme boundary, i.e. VH- + -V or V-
+ -HV-, the laryngeal could be restored by analogy, often yielding a disyllabic long vowel or
diphthong in Vedic Sanskrit. As argued by Lubotsky (1995, p. 220), BrL must be anterior to
19
the loss of intervocalic laryngeals, cf. Skt. dāyi ‘was given’ < *dāHi < *doh3i. Moreover, loss
of intervocalic laryngeals must be anterior to *N̥ > *a, since *-aHn̥- yields a disyllabic long
vowel, even when there is no model for restoration of the laryngeal, cf. GAv. dat.sg. vātāi
‘wind’ /va’atai/ ~ Skt. vā̄́ta- ‘wind’ /va’ata/ < PIE *h2ueh1nto- (Lubotsky, 1995, p. 230).
2.3.3. Stops
In PII, the PIE stops developed into stops and affricates. The labial and dental series were
retained, whereas the labiovelar and velar series merged.13 Secondary affricates from
palatalized velars emerge from the merger of *e and *o. In this section, I discuss special
developments of stops in PII, as well as their phonetic interpretation. I leave out some early
changes like the depalatalization of *Ḱ before *r (Kloekhorst, 2011), which are not strictly
speaking Indo-Iranian but Post-PIE developments shared by several branches.
2.3.3.1. Phonetics of the stops

For the three series of PIE stops and their descendants in PII I use the terms tenuis/voiceless for
*T, media/glottalic for *D, and media aspirata/voiced (aspirate) for *Dh. The first set of terms
(tenuis/media/aspirata) is used as a cover term without phonetic implications. The second set
(voiceless/glottalic/voiced) is used when the phonetic interpretation of the stops is significant.
The cover symbols *T/*D/*Dh reflect the traditional (non-glottalic) phonetic interpretation of
the stops.
Although not yet commonly accepted by scholars of Indo-European linguistics, the
Glottalic Theory offers an explanation to a number of unrelated features of the stops in PIE and
in the daughter languages. The version of the Glottalic Theory employed here states that the
PIE mediae *D were pre-glottalized stops (Kortlandt, 2018), i.e. stops with an inherent glottalic
feature realized before the occlusion, [ʔt] etc.
In addition to offering a plausible phonetic explanation to the root constraint against the
type *De(R)D-, the scarcity of PIE *b and the make-up of the PIE stop system in general
(Kümmel, 2012, p. 299), the Glottalic Theory is supported by comparative evidence. For
example, Winter’s Law in Balto-Slavic (Winter, 1978), Lachmann’s Law in Latin (Kortlandt,
1985), Lubotsky’s Law in Indo-Iranian (1981), the Kortlandt effect (1983), and lengthening
before mediae in Anatolian (Kloekhorst, 2014, p. 230ff) all suggest that PIE mediae were not
plain voiced stops, but pre-glottalized.
13
Some scholars contend that labiovelars were preserved in PII, but this cannot be correct (cf. 2.3.5.2.).
20
A consequence of the Glottalic Theory is that the contrastive features of the tenues and
mediae aspiratae must be reinterpreted:14 since the mediae are not plain voiced stops but
glottalic, there is no reason to assume that the aspiratae contrasted with the tenues in both voice
and aspiration. In fact, voiced aspirates are only attested in Indic, where they are more properly
described as ‘breathy voiced’ (Kloekhorst, 2016, p. 234). The aspiration feature of Indic voiced
aspirates is not phonetically identical to the aspiration of voiceless stops (Kobayashi, 2017, p.
331). Thus, the aspiration feature in Greek voiceless aspirates need not be historically related
to the breathy voice feature of Indic. The PIE mediae aspiratae are therefore best interpreted as
plain voiced stops.
The stops and affricates in PII generally continue the PIE situation with tenues =
voiceless, mediae = pre-glottalized (cf. 2.3.3.3.), and aspiratae = voiced. In Iranian, the mediae
and aspiratae merged, probably as a plain voiced stop (Hoffmann & Forssman, 1996, p. 95). In
Indic, a new series of voiceless aspirated stops emerged from clusters of tenues + laryngeal
(Kuryłowicz, 1935, p. 46), and the aspiratae became voiced aspirated stops.
2.3.3.2. Phonetics of the palatals

In PIE, the palatals were most likely stops since they became velar stops in the centum
languages. However, in Indo-Iranian languages they emerge as affricates or fricatives. The goal
of this section is to approximate the phonetic quality of PII primary palatals by comparing the
reflexes in Indo-Iranian languages.
2.3.3.2.1. Indic
The outcome of PII *ć is Skt. ś. Synchronically, ś is a sibilant that does not only reflect *ć but
also *s and *š in external sandhi. Its place of articulation is either alveopalatal [ɕ] or palato-
alveolar [ʃ] (Kobayashi, 2004, p. 55). The voiced counterpart of Skt. ś is j, which continues PII
*j̄́ as well as *ǰ. Synchronically, j is a palato-alveolar affricate [ʥ] (Kobayashi, 2004, p. 74).
The outcome of *j̄́h is Skt. h, which is phonetically a voiced glottal fricative [ɦ] (Kobayashi,
2017, p. 331). However, forms like Skt. jáhāti ‘leaves’ < *j̄́ha-j̄́haH-ti show that h must have
been an affricate before the application of Grassmann’s Law.
In sum, the reflex of voiceless *ć and voiced aspirate *j̄́h are fricatives. While the latter
must have been an affricate in the prehistory of Indic, the reflex of *j̄́ is synchronically an
affricate. All are alveopalatal or palato-alveolar.
14
The contrast between *T and *Dh may have been realized as a length opposition in Proto-Indo-Anatolian,
which later developed into a voice opposition in ‘classical’ PIE (Kloekhorst, 2016).
21
2.3.3.2.2. Iranian
Traditionally, the PII palatals *ć, *j̄́ and *j̄́h were thought to yield dental fricatives *s, *z in PIr.
However, this reconstruction only accounts for the Avestan reflexes s/z, leaving crucial
evidence from other Iranian languages out of consideration.
In OP, *ć generally yields a labiodental fricative θ, whereas *j̄́(h) merges with the dental
stop d, perhaps pronounced as a voiced fricative [ð] (Cantera, 2017, p. 492). While θ could have
developed from a dental fricative [s] or affricate [ʦ], d can hardly go back to [z], but most likely
reflects an earlier dental affricate [ʣ]. Lubotsky (2001a, p. 49) argued that OP θ went through
an intermediate stage *s, since PII *sć > OP -s- in medial position but θ- in initial position. His
idea is that *sć yielded *ss, which was simplified to *s- in initial position, becoming θ, but not
in internal position, since a geminate was tolerated here. This seems reasonable, since if θ
derives directly from [ʦ], we would expect *sć to be pronounced [sʦ] or [tʦ], which could have
yielded -st- or -θ-, but hardly -s-.
Beyond Avestan and OP, the evidence speaks more clearly against reconstructing PIr.
dental fricatives. In Khotanese the reflex of *-ću̯- is -śś-, cf. aśśä ‘horse’ < PII *Haću̯a-, which
is to be interpreted as an alveo-palatal fricative [ɕ]. Elsewhere *ć > s. According to Kümmel
(2019, p. 15), dental *s could have been secondarily retracted to [ɕ] before *u̯. Sims-Williams
(1998, p. 136), on the other hand, argues that the pre-form of -śś- cannot have been dental, but
must have been pronounced further back, viz. alveo-palatal.
As for the manner of articulation, Khot. dasta, Parth. dst, Sogd. δst ‘hand’ < PII *j̄́hasta-
all underwent dissimilation of *j̄́h > *d due to the following *s, which shows that *j̄́h was an
affricate in Proto-Iranian. PII *j̄́h otherwise becomes a voiced dental fricative in these
languages, cf. Khot aysu [azu] ‘I’ < PII *Haj̄́hHam. Moreover, early Iranian loanwords into
Tocharian, e.g. ToB etswe ‘mule’ << *aʦwa, show evidence of an affricate (Peyrot, 2018).
In conclusion, the Iranian evidence suggests that PII *ć, *j̄́ and *j̄́h became PIr. dental
affricates [ʦ], [ʣ] or alveo-palatal affricates [ʨ] [ʥ].
2.3.3.2.3. Nuristani
The Nuristani reflexes of PII *ć, *j̄́ and *j̄́h are c, j, pronounced as dental affricates [ʦ], [ʣ]
(Blažek & Hegedűs, 2012, p. 46).
2.3.3.2.4. Proto-Indo-Iranian
Thus, all branches of Indo-Iranian suggest that the PII palatals were affricates. Moreover, in
Indic and perhaps Khotanese, reflexes of the palatals are alveo-palatal, whereas the remaining
languages have dentals. As an unconditioned change from dental/alveolar > alveo-palatal
22
affricate is barely attested cross-linguistically (Kümmel, 2007, p. 232), it seems highly probable
that the alveo-palatal pronunciation is original. The PII palatals may thus with some confidence
be interpreted as alveo-palatal affricates: *ć = [ʨ] (voiceless), *j̄́ = [ʔʥ] (glottalic) and [ʥ]
(voiced).
2.3.3.3. Stops in contact with laryngeals

Indic voiceless aspirates ph, th, kh derive from clusters of tenues + laryngeal, e.g. tistha- ‘to
stand’ < *ti-stH-a-, dat.sg. sákhye ‘companion’ < *sakHi̯ ai̯ (Kuryłowicz, 1935, p. 46). In
Iranian, tenues become fricatives before any consonant, including laryngeals, so there is no
reason to assume an intermediate stage of voiceless aspirates (pace Cantera, 2017, p. 21).
Nevertheless, in both Indic and Iranian, clusters of media + laryngeal merge with voiced
aspirates. This is most clearly seen in Skt. máhi < *maj̄́H < PIE *meǵh2, Skt. sadhiṣ- < *sadHs-
< PIE *sedh1s- and Skt. duhitár < *dhughHtar- *dhugHtar- < PIE *dhugh2ter-. Since OAv.
dugədar was affected by BL, it must go back to *dhughtar- < *dhugHtar- as well.
In the Sanskrit forms above, the laryngeal appears to show a double reflex, yielding both
aspiration of a preceding stop and *i. To explain this development, scholars have assumed that
*H̥ > *Hi. However, this hypothesis cannot explain the attested material (cf. 2.3.2.2.). An
alternative solution was offered by Lubotsky (2018, p. 1882). Under the assumption that the
mediae were pre-glottalized and the laryngeals had merged as [ʔ], the change *DH > *DhH may
be understood as a dissimilation of the glottalic element of the stop. The dissimilated *D merged
with *Dh, which was probably a plain voiced stop, later aspirated in Indic. As it is a
dissimilatory process, laryngeal deglottalization does not preclude that the laryngeal was
subsequently vocalized to *i.
The deglottalization process is phonetically paralleled by Lubotsky’s Law (1981), which
constitutes dissimilatory loss of laryngeals in the sequence *-HDC- > *-DC-, cf. Skt. păjrá-
‘firm’ < *paHj̄́ra-. Due to their phonetic similarity, it is likely that the two sound changes
occurred simultaneously.
2.3.3.4. Palatalization of velars

The outcome of the Pre-PII velars *k, *g, *gh depends on the phonological environment. Before
*e and *i, the velars are palatalized to affricates *č, *ǰ, *ǰh. As evidenced by Skt. duhitár- and
Prasun lüšt ‘daughter’ < *dhuǰhitar- < *dhugHtar-, secondary *i < *H̥ also caused palatalization.
This change is linked to the PII vowel merger, since the change *K > *Č was phonologized
when the conditioning factor, *e vs. *o, disappeared. Due to ablaut (e.g. *e ~ *o or *oi ~ *i),
alternations between velar stops and palatal affricates became a frequent phenomenon in verbs,
23
cf. Skt. 3sg.pres. hánti ‘slays’ < Pre-PII *ghenti but 3pl.pres. ghnánti ‘they slay’ < *ghnenti. The
original distribution is often blurred by analogy in the daughter languages, cf. Skt. gácchati ~
LAv. jasaiti ‘goes’ < *gm̥sćati.
Lubotsky (2001a) argued that *sk and *sḱ were not phonemically distinct in PIE.
Although the outcome of PIE *sk has three separate reflexes in Indo-Iranian, *sć, *sč and *sk,
they stand in complementary distribution. In a non-palatalizing context, *sk is retained. Before
*e and *i, *sk is palatalized to *sč as expected, but *sč later becomes *sć unless it is preceded
by an obstruent (Lubotsky, 2001a, p. 53).
A special problem is whether *eh2-3 caused palatalization of a preceding *K or not. As
shown by Ollet (2014), there is no good evidence for palatalization of velars before *eh2-3. The
difficulty lies in the fact that paradigmatic alternations were often levelled out, and so the
absence of palatalization before *eh2-3 does not necessarily reflect the original situation.
However, although Avestan usually generalizes the palatalized variant in paradigms, OAv.
3sg.aor. gāt̰ ‘went’ < *gweh2-t shows no palatalization. I conclude that *eh2-3 most likely did not
palatalize a preceding velar, because the vowel was lowered by laryngeal coloring.
2.3.3.5. Bartholomae’s Law

Bartholomae’s Law (BL) describes a progressive voicing assimilation which affected aspiratae
in Indo-Iranian. Clear examples are Skt. buddhá- ‘awoken’ < *bhudh-ta- and OAv. 3sg.inj.med.
aogədā ‘to announce’ < *HaHugh-ta.
In Iranian, clusters of the shape *Dhs were also affected, cf. OAv. (pairii-)aoγža <
*HaHugh-sa and diβža- < *dhi-(dh)b-zha-. In Indic, however, *Dhs becomes *Ts, cf. Skt. dipsa-
< *dhi-(dh)b-zha-15 (LIV, p. 133). The LIV explains the Sanskrit outcome as secondary
restoration of -s-. The reason for assuming that -ps- replaced earlier *-bzh- is that the initial d-
of Skt. dipsa- according to LIV has been deaspirated by Grassmann’s Law (GL). However,
other *Dhs clusters such as Skt. aor.inj. dhukṣa- ‘to milk’ < *dhugh-sa-, which can hardly be
analogical,16 appear to remain unaffected by GL. It seems more likely that Skt. dabh- originally
had an initial media, PIE *debh-. The only reason why LIV assumes PIE *dhebh- is Umbr. dat.sg.
fefure ‘damage’ < *dhebh-os-ei̯ . However, as Schirmer (1998, p. 64), who proposed the Umbrian
etymology, shows, there are several other possible interpretations of Umbr. fefure, and the
assumption of an s-stem requires analogical restoration of -u- < *-o- in the suffix, which does
not normally occur in the language.
15
The loss of root-initial *dh is regular due to cluster simplification, cf. Skt. nadbhyas ‘grandson’ < *naptbhyas
< *napt- (Kobayashi, 2017, p. 334)
16
Since other derivations of *dhugh- did undergo GL, cf. Skt. 3sg.pres. dogdhi < *dhau̯gh-ti.
24
If Skt. dipsa- instead goes back to PII *di-(d)bh-sa-, the development of *Dhs may be
explained differently. Within Indic, the PII plain voiced stops (= aspiratae) became aspirated.
However, the aspiration could not be realized phonetically in *Dhs clusters, since **[ʔdibzha]
was impossible. Therefore, the stop instead merged with the tenues, yielding [ʔdipsa]. This also
explains why aspiration was maintained in Skt. dhukṣa- < *dhugh-sa-, since *gh was never
aspirated here. Since Iranian never developed aspiration, the voicing of *Dhs = [Dz] in e.g.
diβža- was retained.
Since there are slight differences between Indic and Iranian in the operation of BL, it is
likely that the process was phonetic in PII and not phonologized until Post-PII.
2.3.4. The sibilant *s

PIE had a single sibilant phoneme *s. In PII, the phoneme *s has voiced [z] and palatalized [š]
allophones.
Already in Pre-PII *s > š / i, u, r, K_. Known as the RUKI-rule, this sound change is
shared by the satəm languages and was thus most likely active at a very early date. On the other
hand, the rule remained active for a long period, since secondary instances of i, u, r, K (e.g. *i
< *H̥) could trigger it (Lubotsky, 2018, p. 1881). Yet, there is one case of PII *š which is not
conditioned by RUKI, namely *šu̯ećš ‘six’ < PIE *sueḱs.17 Moreover, the clusters *tć and *ćs
may have merged as [tš] already in PII, although Khotanese evidence suggests that they were
kept separate (Cantera, 2017, p. 25). Thus, certain instances of *š could be analyzed as
phonological, but the RUKI-rule remained active until Post-PII.
2.3.5. Liquids
2.3.5.1. General development

The outcome of the PIE liquids *r and *l in PII is complicated. In Iranian languages *r and *l
merge (Cantera, 2017, p. 15). In Sanskrit, r and l are synchronically phonemically distinct, but
the distribution does not always match the etymological origins (Lubotsky, 2018, p. 1878). The
irregular distribution of r and l in Sanskrit is often explained as dialectal variation that was
adopted into the literary language (Burrow, 1973, p. 84). As a consequence, Skt. r and l are not
decisive for etymological analysis. For the purposes of this study, therefore, I will operate with
a single PII liquid *r.
PII *r has two main allomorphs conditioned by the phonotactic environment where it
occurs. Whenever *r is next to a vowel or precedes a vocalic resonant, it is consonantal
17
The initial *s- may be secondary, perhaps by contamination by *septm ‘seven’ (Kroonen, 2013, p. 431).
25
(Schindler, 1977). Conversely, whenever *r is interconsonantal, it is vocalic *r̥. Vocalic *r̥ is
retained in Sanskrit, and is generally rendered as ər in Avestan (Hoffmann & Forssman, 1996,
p. 91).
2.3.5.2. Liquid + laryngeal clusters

The cluster *r̥H shows a special development in Indo-Iranian. In Sanskrit, the reflexes are ī̄̆r
and ū̄̆r. The outcome is conditioned by three factors: 1) the following phoneme, 2) the accent,
and 3) whether the environment is labial or not. When *r̥H precedes a vowel, the anaptyctic
vowel is short, cf. Skt. híraṇya- ‘gold’ < *j̄́hr̥Hani̯ a- and tirá- ‘to cross’ < *tr̥Há-. The same is
true when *r̥H precedes *i̯ V or *u̯V, cf. Skt. turyā̄́ma ‘we shall conquer’ < *tr̥Hi̯ ā̄́ma and
bhurváṇi- ‘victorious’ < *bhr̥Hu̯áni-. However, if the liquid is accented, the vowel is long, cf.
Skt. tū̄́rva- ‘to cross’ < *tr̥̄́Hu̯a- and -śī̄́rya ‘having smashed’ < *ćr̥̄́Hi̯ a- (Lubotsky, 1997).
Elsewhere (*rHC where C is not *i̯ /u̯), the vowel is long, regardless of the accent, cf. Skt.
dīrghá- ‘long’ < *dr̥Hgha- and ī̄́rṣyant- ‘being envious’ < *r̥̄́Hsi̯ ant-. Finally, the quality of the
vowel is ū̄̆ if *r̥H is preceded by a labial consonant (p, b, bh, m, u̯) or followed by *Cu̯.
Exceptions to this rule (mainly verbal forms) can be explained as analogical to other forms in
the paradigm where a conditioning labial exists (Lubotsky, 1997, p. 139). Elsewhere, the vowel
is ī̄̆, cf. the examples above.
Burrow (1957), and more recently Clayton (2018), argued that *r̥H > ū̄̆ also when
preceded by a PIE labiovelar. However, counterexamples like Skt. girí- ‘mountain’ < *gwrH-i-
and gīrṇá- ‘swallowed’ < *gwrh3-no-, as well as the unlikelihood of preserved labiovelars in a
satəm language, render this analysis difficult.
The Indic development of *r̥H may be expressed by the following set of rules:
*r̥H > ur / C+labial _V, _u̯V
*r̥H > ūr / C+labial _C, _Cu̯, _́u̯V
*r̥H > ir / C-labial _V, _i̯ V
*r̥H > īr / C-labial _C, _C, _́i̯V
The Iranian reflexes of *r̥H also vary depending on the phonological environment. In most
Iranian languages, labial and non-labial contexts are differentiated, showing different
anaptyctic vowels (Clayton, 2018). In Avestan, the reflexes are ar, ər and ruu. The outcome is
dependent on 1) the accent, 2) the following phoneme, and 3) whether the environment is labial
or not (Cantera, 2001).
Under the accent, *r̥H becomes ar, cf. OAv. pauruua- ‘first’ ~ Skt. pū̄́rva- < *pr̥̄́Hu̯a-.
The same is true when *r̥H precedes a vowel, independent of the accent, cf. OAv. tarə̄
26
‘sideways’ < *tr̥Has and pouru-18 (< *paru-) ‘much, many’ < *pr̥Hú- (Skt. purú-). In
unaccented position before consonants, the development of *r̥H depends on the context. Before
*u̯, unaccented *r̥H yields ruu, cf. Av. zruuan- ‘life-time’ < *j̄́r̥H-u̯án- (Lubotsky, 1997, p.
144). In non-labial contexts, unaccented *r̥H becomes ar, cf. OAv. darəga- ‘long’ < *dr̥Hghá-.
In labial contexts, unaccented *r̥H becomes ər, cf. OAv. pərəna- ‘full’ ~ Skt. pūrṇá- < *pr̥Hná-.
A labial context is whenever *r̥H is preceded by a labial consonant (p, b, bh, m, u̯) or followed
by *Cu̯.
The Iranian development of *r̥H may be expressed by the following set of rules:
*r̥H > ar / _́, _V, C-labial _C,
*r̥H > ər / C+labial _C, _Cu̯
*r̥H > ruu / _u̯19
Cantera (2001, p. 25) argues that initially, *r̥H > ər in all contexts, but that ər later became ar
unless blocked by a labial environment. If so, there is no actual labialization, so much as
lowering of the vowel in non-labial contexts. Moreover, any ər that does not become ar
eventually merges with the reflex of *r̥ (without laryngeal). In other Iranian languages, the
anaptyctic vowels of original *r̥H and *r̥ are clearly labial in the appropriate contexts, cf. Phl.
purr ‘full’ < *pr̥Hná-, Pto. murγə̄́ ‘bird’, MoP murw ‘bird’ < *mr̥gá- and Y-M purs ‘to ask’ <
*pr̥sća-.
Although both Indic and Iranian show processes of labialization, they also show
significant differences that preclude projecting the vocalization of *r̥H to PII.
2.3.6. Nasals
2.3.6.1. General development

Like the liquids, the PIE nasals *n and *m, retained in PII, had consonantal and vocalic
allophones. The consonantal nasals are generally preserved in Indo-Iranian languages. Between
consonants and at word boundaries, the nasals are vocalic *n̥ and *m̥. They later merge with *a
in most positions (Lubotsky, 2018, p. 1876). An exception is *m in word-initial position before
resonants, which remains consonantal and yields *b before *r in Indic, cf. LAv. mraoiti ‘says’
~ Skt. brávīti ‘says’ < *mrau̯H-ti.
18
Here, ar was secondarily rounded (Hoffmann & Forssman, 1996, p. 90)
19
Lubotsky (1997, p. 144) did not find examples with *i̯ .
27
2.3.6.2. Nasal + laryngeal and nasal + resonant clusters
In some contexts, the outcome of the vocalic nasals is *an, *am. This occurs in the contexts
*N̥RV and *N̥HV, cf. Skt. mányate ~ OAv. mańiiete ‘thinks’ < PII *mani̯ a- < PIE *mn̥-i̯ e-,20
Skt. namrá- ‘loyal’ ~ LAv. namra- ‘respectful’ < *namrá < *nm̥ró-,21 Skt. -tama- ~ Av. -təma-
< *-tamHa- < *-tm̥Ho-,22 Skt. hanmás ‘we slay’ < *ǰhn̥-mas. The sequence *N̥HC instead
became *aHC > *āC, cf. Skt. jātá- ~ Av. zāta- ‘born’ < *ǵn̥h1-tó-. A sequence **N̥RC never
developed regularly, since the second resonant would have been vocalized instead of the nasal
(Schindler, 1977).
There is one major exception to the above. When *n̥ precedes consonantal *n, the
outcome is *an-, cf. Skt. tanóti ‘stretches’ < *tn̥-neu̯-ti. Most likely, the expected geminate
nasal *-nn- was simplified to *n (Kümmel, 2005, p. 322). In general, the vocalization of nasals
in nasal present formations is aberrant (Schindler, 1977, p. 56). For example, the plural *tn-nu̯-
énti gives Skt. tanvánti ‘they stretch’ instead of **tnanvánti < **tn-n̥u̯-énti, perhaps due to
paradigmatic levelling.
2.3.7. Semivowels
The PIE semivowels *i and *u are retained in PII and have consonantal and vocalic allophones.
Next to a vowel a semivowel is consonantal, whereas an interconsonantal semivowel is vocalic,
cf. Skt. 3pl. vidúr ~ LAv. vīδarə ‘they know’ < *u̯id-. Although PII *u in principle always
derives from PIE *u, PII *i also arose secondarily via laryngeal vocalization (cf. 2.3.2.2.).
The sequences *iHC and *uHC eventually yield long vowels *īC and *ūC, when the
laryngeal is lost with compensatory lengthening, cf. Skt. bhūtá- ~ LAv. būta- ‘become’ < PII
*bhuHta-. However, this change is probably quite late. In Nuristani, the sibilant in *uHs is not
affected by RUKI, cf. Kati mussā, Prasun mǖs’ū ‘mouse’ < PII *muHs-, indicating that the
RUKI-rule was phonologized before *uHs > *ūs. As seen in 2.3.4., the RUKI-rule was probably
not phonologized in PII. Since PII *muHs- > Skt. mūṣ- and Av. mūš- show the effect of RUKI,
the developments *iH > *ī and *uH > *ū form an isogloss between Indic and Iranian, excluding
Nuristani.
20
The zero-grade is paralleled by Gr. μαίνομαι ‘to be furious’.
21
-ró- derivatives normally take zero-grade of the root (Kümmel, 2005).
22
Cf. Lat. in-timus ‘inner’ (Lubotsky, 2018, p. 1876).
28
2.4. Relative chronology of Indo-Iranian sound changes
In this section, the relative chronology of the Indo-Iranian sound changes described in 2.3. is
discussed.
Lubotsky (2018, p. 1877) has argued that LV in final syllables is one of the earliest PII
sound changes, preceding BrL. The basis for this argument is the long vowel of Skt.
(yúva-)jāni-23 ‘(having a young) wife’ < *-gwonh2-, in contrast to the simplex Skt. jáni- <
*gwenh2-. If the laryngeal had been consonantal when BrL operated, it would have closed the
syllable and prevented lengthening of *o. While the o-grade is expected in a compound, not all
Vedic compounds follow this pattern, cf. pr̥ṣṇī-mātara- ‘having P. as mother’ < *-meh2ter-o-.
However, it is less likely that the long vowel of -jāni- is secondary than the short vowel
of -mātara-, since the latter conforms to the synchronic simplex form mātar-. In any case, LV
must precede the PII palatalization, since *i < *H̥ also caused palatalization (Skt. duhitár- ~
Prasun lüšt < *dhuǰhitar- < *dhugHtar-.
LV is not the earliest in the relative chronology of PII sound changes. Forms like Skt.
sadhiṣ- ~ Av. hadiš- ‘seat’ < *sadhis- < *sadhH̥s- < *sadHs- < PIE *sedh1s-, Skt. máhi ‘great’
< *maj̄́hi < *maj̄́hH̥ < *maj̄́H < PIE *meǵh2 and Skt. duhitár- ‘daughter’ < *dhuǰhitar- <
*dhughH̥tar-< *dhugHtar- < PIE *dhugh2ter- show that laryngeal deglottalization must precede
LV.24 The mediae must have been deglottalized already when LV occurred, since laryngeals at
that point merge with *i, losing the glottalic feature that caused deglottalization. Lubotsky’s
Law is most likely contemporary with laryngeal deglottalization.
Furthermore, by the time of laryngeal deglottalization, the three PIE laryngeals had
already merged into a glottal stop *H = [ʔ]. However, the difference in vowel quality must have
remained after the laryngeal merger, as *eH < *eh1 later caused palatalization, whereas *aH <
*eh2-3 did not. At this point, the phonetic coloring of *e was phonologized, since the
conditioning factor (*h1 ≠ *h2-3) was lost. In 2.3.1., I argued that laryngeal coloring could have
been sub-phonemic until the PII vowel merger, but since the laryngeals merged before the PII
vowel merger, laryngeal coloring must have been phonologized earlier than previously thought.
Thus, this conclusion is not based on direct evidence, but follows from considerations on
relative chronology.
23
The palatal j- is assumed to be analogical to the simplex form.
24
Only Skt. duhitár- is decisive, since in the paradigms of sadhiṣ- and máhi, the laryngeal would only have been
vocalized in certain case forms. Therefore, if the original nom./acc.sg.n was *maǰi < *maǰH̥, *ǰh could have been
levelled throughout the paradigm by analogy to the gen.sg. *maǰhHás < *maǰHás. In the case of *sadhis-, only
nom./acc.sg. is attested, but the original genitive could have been *sadH-as-as.
29
After LV, BrL caused lengthening of *o > *ō in open syllables. At this stage, PII had
three phonemic vowels *ē̄̆, *ō ̄̆ and *ā̄̆, of which the first (together with *i) caused phonetic
palatalization of preceding velars (/ke/ = [če]). Subsequently, the vowels merged as *ā̄̆, causing
the secondary palatals to become phonemic.
As discussed in 2.3.2.4., the laryngeal accent shift must have been posterior to Lubotsky’s
Law. The loss of intervocalic laryngeals is posterior to BrL. The vocalization of *N̥ > *a(N) is
posterior to the loss of intervocalic laryngeals.
The change V̄̆HC > V̄C is posterior to the vocalization of nasals, since *N̥HC > *āC, most
likely through an intermediate stage *aHC. As Nuristani preserved *i/uHC sequences, the
change V̄̆HC > V̄C seems to be an Indic-Iranian isogloss. In the strict sense, it is Post-PII, but
could be regarded as a shared innovation between Indic and Iranian, if Nuristani was the first
to split off from PII.
With these considerations, I arrive at the following relative chronology:
Table 2. Relative chronology of Proto-Indo-Iranian sound changes

Phase Sound change
1 (Pre-PII) Phonetic coloring of Laryngeal metathesis RUKI-rule
*h2-3e > *h2-3a *CHUC > *CUHC
2 (PII) Laryngeal merger to glottal stop *H, phonologization of *a
3 Lubotsky’s Law (ʔʔDC > ʔDC), laryngeal deglottalization (ʔDʔ > Dʔ)
4 Laryngeal vocalization Laryngeal accent shift *Cé(C)H(C)-U- >
*H > *i / C_(C)# *Ce(C)H(C)-Ú-
5 Brugmann’s Law
*o > *ō / _$C.
6 Loss of intervocalic Vowel merger Phonologization of
laryngeals *ē̄̆, *ō ̄̆ , *ā̄̆ > *ā̄̆ palatalized velars *č, *ǰ, *ǰh
7 *N̥ > *a(N)
8 (Post-PII) laryngeal loss with compensatory Vocalization of *r̥H clusters
lengthening *VH > V̄
30
3. Etymological analysis of proposed loanwords
In this chapter, previously proposed loanwords into Proto-Indo-Iranian are analyzed
etymologically according to the methodology outlined in chapter 1. Each entry includes a
discussion of potential etymologies, why a word can or cannot be considered a loanword
(applicable criteria are shaded), and when it was borrowed. The words are divided into two
main categories: loanwords and non-loanwords. The loanwords (sections 3.1.-3.4.) are divided
into chronological layers, which are further discussed in chapter 4. The non-loanwords (section
3.5.) are either inherited, i.e. have plausible or possible IE etymologies, or simply lack evidence
for borrowing.
3.1. Loanwords I: Pre-Proto-Indo-Iranian or early Proto-Indo-Iranian
1. PII *ućig- ‘sacrificing priest’

Ind. Skt. uśíj-
Ir. OAv. usig-
Lim. distribution Irr. correspondences Rem. morphology Rem. phonology Specific semantics
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 235) argues that it is derived from vaś-
‘to wish’. However, from a morphological perspective, the word looks non-IE, given the suffix
*-ig- (almost unique for this word, cf. AiGr. II, 2, p. 321).
Another word from the religious sphere, Skt. r̥tvíj- ‘priest’, is often analyzed as a
compound of r̥tu- ‘season’ + -ij- ‘sacrificing’ (EWAia I, p. 258). However, while the root yaj-
‘to sacrifice’ reflects a palatal stop (PIE *Hieh2ǵ-), the -k of nom.sg. r̥tvík reflects a plain velar
(Lubotsky, 2008). For this reason, it is likely that PII *ućig- and Skt. r̥tvíj- contain the same
suffix *-ig- and may have been borrowed from the same source. The same is likely for Skt.
vaṇíj- ‘merchant’ and bhuríj- ‘?’.
The suffix *-ig- is palatalized in forms like the genitive Skt. uśíj-as (< *ućig-es). The non-
palatalized form is reflected in Skt. nom.sg. uśik, instr.pl. uśigbhyas and OAv. nom.sg. usixš.
While it is not unthinkable that the paradigm could have been secondarily adapted to fit the
pattern of inherited palatalized velars, the most straightforward explanation is that *ućig- was
borrowed before the PII palatalization of velars.
31
3.2. Loanwords 0: Proto-Indo-Iranian, but no further indication of date of borrowing
2. PII *aka-
Ind. Skt. áka- ‘pain’
Ir. Av. aka- ’bad’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 39) takes it as a derivative of añc- ‘to
bend’ < PIE *h2enk-, which is semantically possible. However, this analysis is morphologically
problematic, since Sanskrit would reflect an accented zero-grade *Hn̥ ̄́ k-o-. The “suffix” *-ka-,
found in several loanwords, may also indicate non-IE origin.
3. PII *anću- ‘Soma plant’

Ind. Skt. aṃśú-
Ir. Av. ąsu-
Taken as a loanword by Lubotsky (2001b). No IE etymology (EWAia I, p. 37). Witzel (2003,

p. 37) identifies the Soma plant as ephedra, native to mountainous regions of Central and South
Asia. The word may share a common origin with ToA añcwaṣi ‘made of iron’, assuming that
the color of iron was associated with the color of ephedra (Pinault, 2003b).
4. PII *atHaru̯an- ‘priest’

Ind. Skt. átharvan-
Ir. LAv. āθrauuan-
Taken as a loanword by Lubotsky (2001b). The word had no IE etymology (EWAia I, p. 60),
but is sometimes connected to Av. ātar- ‘fire’, also of unknown origin. In reality, the connection
between the words is probably folk-etymological. This could explain the irregular
correspondence Indic ar : Iranian ra, if earlier *aθaru̯an was remodeled to āθrauuan- based on
the oblique stem āθr- ‘fire’. The “suffix” *-aru̯a- is found in other proposed loanwords.
5. PII *atka- ‘cloak’

Ind. Skt. átka-
Ir. LAv. aδka-, at̰ .ka-
Taken as a loanword by Lubotsky, since the suffix *-ka- is normally denominal (2001b, p. 304).
Attempts at IE etymology (e.g. PIE *tek- ‘weave’, EWAia I, p. 58) are unconvincing. LAv.
32
at̰ .ka- with ‘implosive’ t̰ , instead of **aθka-, as well as the variant aδka- could point to original
*adka-, cf. LAv. t̰ bi- ‘twice’ < *du̯i-.
6. PII *(H)āni-
Ind. Skt. āṇí- ‘linchpin, hip’
Ir. –
Taken as a loanword by Kuiper (1991, p. 89), Pinault (2003a, p. 132) and Witzel (2003, p. 33).
EWAia (I, p. 161) is agnostic as to the exact etymology of this word. The proposed connection
to Gr. ὠλένη ‘elbow’, Lat. ulna ‘forearm’, PGm. *alīnō- ‘forearm’ < PIE *Heh3l-én-eh2-, which
is not semantically clear, would require the assumption of Fortunatov’s Law.
According to Pinault (2003a), the word is found in the compound kalyāṇī- lit. ‘with
beautiful hips’. The connection between Skt. kaly- and Gr. καλός is difficult, since this requires
the reconstruction of PIE **kal-. Furthermore, Skt. -ṇ- remains unexplained, unless one
assumes that a dialectal variant *karyāṇi-, where -r- would cause the *n to become retroflex,
influenced the simplex form.
Lastly, the lack of an Iranian cognate puts it into question whether the word can be
reconstructed for PII. ToB oñi ‘hip’ cannot have been borrowed from Indic, since Indic ā is
adapted to Tocharian a in later borrowings (Pinault, 2003a, p. 131). Thus, if oñi was borrowed
from Indo-Iranian, it was most likely at a very early stage, perhaps PII. However, it is difficult
to exclude that Tocharian borrowed the word independently from a Central Asian source.
7. PII *bharu̯- ‘to chew, eat’

Ind. Skt. bharv- ‘to chew’
Ir. LAv. aš.baouruua- ‘where there is much to eat’
Suggested to be a loanword by Lubotsky (2001b). No IE comparanda are available (EWAia II,

p. 253). The cluster *-ru̯- is found in several PII loanwords but not in secure PIE roots. In 11
out of 12 cases of root final *-u in LIV, a laryngeal precedes *-u.25 The roots *melh2u- ‘to grind’
and *deh3u- ‘to give’ are clearly secondary roots from *melh2- ‘to grind’ and *deh3- ‘to give’,
25
The exception is *bheru- ‘to boil’, which is semantically clearly unrelated to PII *bharu̯-. Since it is isolated to
Italo-Celtic, it need not go back to PIE. It is strange that on the one hand, root final *-u- seems to be a fossilized
suffix, as evidenced by *melh2u- ‘to grind’ and *deh3u- ‘to give’, but on the other hand, it appears to correlate
phonologically with a preceding laryngeal. This suggests that the suffix *-u- was only reanalyzed as part of the
root in a specific phonological context, perhaps related to the loss of laryngeals in various IE branches.
33
perhaps originally u-stems. However, this is an unlikely origin of *bharu̯- ‘to chew’, since
*bhar- ‘to bear’ is semantically unrelated.
8. PII *bhiš-aj̄́- ‘healer’

Ind. Skt. bhiṣáj- ‘physician’
Ir. LAv. bišaziia- ‘to cure’, -biš- ‘healing’
Taken as a loanword by Lubotsky (2001b). IE origin, as argued by EWAia (II, p. 264), is very
unlikely, especially considering the foreign-looking suffix *-aj̄́- (with agentive function?). The
Sanskrit verbal form bhiṣákti ‘heals’ with -kt- (< *gt) instead of expected -ṣṭ- (< *j̄́t) suggests
that the verbal forms are secondary.
9. PII *bīj̄́a- ‘seed, semen’

Ind. Skt. bī̄́ja-
Ir. Sogd. byz’k, Par. bīz, Khot. bījä < *bīzya- (Bailey, 1979, p. 280)
Taken as a loanword by Witzel (2003, p. 33). IE origin is highly improbable as the root contains
two mediae, one of which is *b. Kent’s (1950, p. 29) attempt to connect OP (Bagā-)bigna- is
rightly rejected by EWAia (II, p. 227), since OP g cannot reflect PII *j̄́. As for the semantics,
either of the reconstructable meanings could have developed from the other, so it is not certain
that the word originally belonged to agricultural terminology.
10. PII *ćaru̯a- ‘Name of a deity’

Ind. Skt. śarvá-
Ir. LAv. sauruua-
Taken as a loanword by Lubotsky (2001b) and Witzel (2003). EWAia (II, p. 621) presents no
plausible IE etymology. Pinault (2003b) connects *ćaru̯a- to ToB śerwe, ToA śaru ‘hunter’,
arguing that Toacharian borrowed at an early stage from Iranian. Since the words are probably
non-IE, it is unclear whether this is the correct direction of borrowing (Adams, 2013, p. 695).
The “suffix” *-aru̯a- is potentially also found in *atHaru̯an- etc.
34
11. PII *ćyā- ‘to freeze, congeal’
Ind. Skt. śyā-
Ir. Oss. syjyn, sujun, Yagh. ši-
Suggested to be a loanword by Lubotsky (2001b). No IE comparanda adduced by EWAia (II,

p. 660) or LIV (p. 360). There are no other examples of anlaut *ḱi̯ - in LIV. The only PIE root
with a similar cluster is *ǵi̯ eu̯H- ‘to chew’ (LIV, p. 168).26 In view of this, it is more likely that
*ćyā- is a loanword than an isolated inherited root.
12. PII *gadā- ‘club’

Ind. Skt. (Su+) gadā-
Ir. LAv. gaδā-
As pointed out by Lubotsky (2001b, p. 303), *gadā- cannot be IE since it would reflect a root
with two mediae. EWAia (I, p. 460) does not propose an IE etymology.
13. PII *gr̥da- ‘penis’

Ind. Skt. gr̥dá- ‘penis’
Ir. LAv. gərəδō.kərəta- ‘cutting of the genitals’
As pointed out by Lubotsky (2001b, p. 303), *gr̥da- cannot be IE since it would reflect a root
with two mediae. EWAia (I, p. 494) mentions some previously suggested IE etymologies (e.g.
*geR-d-), which are implausible.
14. PII *Hustra- ‘camel’

Ind. Skt. úṣṭra-, uṣṭár-
Ir. Av. uštra-, OP uša-(bāri-), Sogd. xwštr-
EWAia (I, p. 237) discusses some attempts at IE etymology but admits that the word looks non-
IE (the suffix *-tro- is not normally used for animates). Moreover, it is unsurprising that a word
for ‘camel’ would be borrowed by Indo-European speakers, since the camel was domesticated
only in the mid-3rd millennium in Iran (Heide, 2011, p. 367). Initial laryngeal is strongly
suggested by Av. Zaraθuštra- < *j̄́arat-Hustra- ‘having aging camels’. Its presence either
means that the source language had an initial consonant that was adopted as PII *H-, or that at
26
The only other case in LIV, *ǵi̯ eH- ‘to bereave’, is isolated to Indo-Iranian and need not go back to PIE.
35
the time of borrowing, any vowel-initial word would automatically be pronounced with an
initial laryngeal, like in PIE.
15. PII *indra- ‘name of a God’

Ind. Skt. índra-
Ir. LAv. iṇdra-
As noted by Lubotsky (2001b, p. 311), if the word was IE, *n should have vocalized, giving
**i̯ adra-. EWAia (I, p. 192) discusses several unlikely etymologies. Parpola (2015, p. 66)
suggests that *indra- was borrowed from PU *ilmar / *inmar ‘thunder god’, derived from PU
*ilma- ‘sky’, reflected in Finn. Ilmarinen and Udmurt Inmar.27 However, since the suffix *-ri- is
a Finnic innovation and cannot be reconstructed for PU (Frog, 2012, pp. 215-6), this is unlikely.
16. PII *išt(i)- ‘brick’

Ind. Skt. iṣṭakā (VS, Br +), iṣṭikā (Sū+)
Ir. LAv. ištiia-, OP išti-, MiP xišt-
Taken as a loanword by Lubotsky (2001b) and Witzel (1999b, p. 54). EWAia (I, p. 201)
discusses possible connections with the PIE *ies- ‘to boil’. This is semantically problematic,
since a brick is burnt, not boiled. Moreover, a -ti- derivation should yield an abstract noun.
Lastly, if Persian x- reflects a laryngeal (Kümmel, 2016a, p. 83), the word cannot reflect PIE
*is-ti-.
Indic has both -akā- and -ikā- suffixes, whereas Iranian has -i- and -i̯ a-. For PII, an
original i-stem seems the most likely, but variation may have existed already at this stage. A
related word is ToB iścem ‘clay’, which may have been borrowed from Indo-Iranian (Witzel,
2003, p. 30).
27
Probably borrowed from Finnic, cf. Frog (2012).
36
17. PII *j̄́harmii̯ a- ‘house’
Ind. Skt. harmyá-
Ir. Av. zairimiiāuuant-
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 807) suggests an IE etymology from
*ǵher- ‘to cover’, which does not explain the PII suffix *-mii̯ a-.28 Semantically, the word fits
together with other loanwords pertaining to permanent structures (e.g. *išt(i)- ‘brick’).
18. PII *kāća- ‘grass’

Ind. Skt. kā̄́śa-
Ir. MoP kāh, MoP kašk, Munji kosk < PIr. *kaćaka-
Nur. Km. kaċo, Kt. kċo, W. kac
Taken as a loanword by Kümmel (2017, p. 284). The connection between the Indic and Iranian
forms is dismissed by EWAia (I, p. 345) for unclear reasons. The variant with suffix -ka- in
Iranian is aberrant in that the root vowel is short. In any case, *kāća- can be reconstructed to
PII based on Skt. kā̄́śa- and MoP kāh.
19. PII *kaći̯ apa- ‘tortoise’

Ind. Skt. kaśyápa-
Ir. LAv. kasiiapa-
Taken as a loanword by Lubotsky (2001b) and Witzel (2003, p. 35). EWAia (I, p. 331) provides
no IE etymology. The “suffix” -pa- is non-IE and is found in other loanwords, e.g. *pāpa-,
*stupa-. If borrowed from a Central Asian language, *kaći̯ apa- may have referred to the
Russian tortoise, Testudo horsfieldii, native to the area of the BMAC culture.29
28
Rather, one would have to assume a root structure *ǵherm-, which is unparalleled in PIE (LIV, p. 708)
29
Cf. The Reptile Database (http://reptile-database.reptarium.cz/species?genus=Testudo&species=horsfieldii)
37
20. PII *kadru- ‘brown’
Ind. Skt. kádru- ‘reddish brown’
Ir. MoP kahar ‘light brown’, LAv. kadruua.aspa ‘with brown horses’ (name of a
mountain)
Taken as a loanword by Lubotsky (2001b) and Witzel (2003, p. 33), who notes that words for
secondary colors are often borrowed. EWAia (I, p. 295) rightly rejects previous attempts at
connecting the Indo-Iranian forms to Gr. Κόδρος.
21. PII *kapāra- ‘bowl’

Ind. Skt. kapā̄́la- ‘dish, bowl’
Ir. MiP kabārag, MoP kabāra ‘vessel’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 300) suggests a connection to Lat.
capiō ‘to take’, caput ‘head’ and OE hafola. For the Latin words, de Vaan (2008, p. 90)
reconstructs a root PIE *kh2p-, but argues that substrate origin in *kap- is equally likely.
Kroonen reconstructs PGm. *hafe/alan- < *kap-ola- for the OE form and *ha(u)bu/eda- ‘head’
based on other Germanic cognates (2013, p. 215). Beekes (1996, p. 220) considers these to
originate in a European substrate language, but does not discuss *kapāra-. The main problem
with connecting the Indo-Iranian and European material is that the only way to explain why *k-
remains non-palatalized in Indo-Iranian is to reconstruct PIE *a.30 The problem is not solved
by assuming a laryngeal in the root, because *kh2ep- would have given Skt. kh-, Ir. x-, and
*keh2p- a long vowel.
A possible scenario, if the word is non-IE, is to assume parallel Post-PIE borrowing by
Italic, Germanic and Indo-Iranian from a common European substrate language.31 This scenario
presupposes that the word entered Indo-Iranian after *a had been phonologized. It also
presupposes a relative geographical proximity of Indo-Iranian, Italic and Germanic speakers.
However, since *kapāra- has the CVCV̄CV structure, characteristic of Sanskrit loanwords, it
more likely derives from an Asian language. Therefore, I will treat the similarity between PII
*kapār/la- and OE hafola, Latin caput etc. as a chance similarity.32
30
The only other possibility for Indo-Iranian is PIE *knp- / *kmp-, but this is not supported by the European
words.
31
This scenario is based on the possibility that speakers of Indo-Iranian first migrated to Europe, before turning
eastwards to Asia.
32
Alternatively, the word diffused as a Wanderwort from Europe to Asia or the other way around, but as it is not
a prototypical ‘culture word’ prone to spreading over large areas (like e.g. ‘wheat’), this seems unlikely.
38
22. PII *kapau̯ta- ‘pigeon’
Ind. Skt. kapóta- ‘pigeon’
Ir. OP kapautaka- ‘blue’, MiP kabōd ‘grey-blue, pigeon’, Khot. kavūt ‘pigeon’
Taken as a loanword by Lubotsky (2001b) as one of the trisyllabic CVCV̄CV words in PII.
EWAia (I, p. 303) considers IE origin likely based on the suffix -ta- which is often found in
colors, but here the meaning ‘pigeon’ is probably primary. The -ka- suffix in OP is a trivial
innovation. According to Berger (1959, p. 58), the source of the word is Austroasiatic, e.g.
Santali potam, Mundari pudām ‘dove’, with ka- as a prefix (also Kuiper, 1991, p. 42).
PII *kapau̯ta- cannot with certainty be assigned to late PII, since the non-palatalized PII
kap- could go back to Pre-PII *kN̥p.
23. PII *kapHa- ‘phlegm, mucus’

Ind. Skt. kapha- ’phlegm’
Ir. LAv. kafa-, Khot. khavȧ ‘mucus’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 303) offers no IE etymology, but
suggests on the basis of Khot. khavȧ, which seems to reflect PIr. *χafa-, that it could be
borrowed or represent a ‘Kraftwort’. Burrow (1973, p. 26) suggests borrowing from Uralic, cf.
Hung. hǎb ‘foam, froth’, Veps. kob̄́ e ‘wave, foam’, and Sam. (Kam.) khòwü ‘foam’. However,
since these words go back to PU *kompa- ‘wave’ (Sammallahti, 1988, p. 537), it is unclear why
the *m would not be reflected in Indo-Iranian.33 Moreover, the Uralic word does not refer to
bodily fluids like in Indo-Iranian but rather to frothy water. In view of the formal and semantic
problems, a Uralic origin must be regarded as speculative.
24. PII *kHā- ‘well, source’

Ind. Skt. khā̄́-
Ir. LAv. xā-
Taken as a loanword by Lubotsky (2001b), due to its semantics and phonology. EWAia (I, p.
451) offers no IE etymology, and denies a connection to the verb Skt. khan-, LAv. kan- ‘to dig’.
LIV (p. 344), however, reconstructs PIE *k(u̯)eh2- ‘to dig’ with a nasal present *k(u̯)-né/n̥-h2-,
which allegedly yielded a secondary root *kanH- and the PII word for ‘well’. The aspirated kh-
33
It is of course possible that PII borrowed from a Uralic language where *m had been lost.
39
in Sanskrit is explained as secondary from the zero-grade *kh(i)-, but this is not attested
anywhere. Even if the connection between ‘well’ and ‘to dig’ is maintained, the problem
remains that both are isolated to Indo-Iranian.
25. PII *kHara- ‘donkey’

Ind. Skt. (AVP+) khara-
Ir. LAv. xara-
Taken as a loanword by Lubotsky (2001b), Witzel (2003, p. 35), and hesitantly EWAia (I p.
447). A relation to Akk. (Mari) ḫāru (ḫârum, ayarum) ‘donkey foal’, as proposed by Eilers
(1959, p. 467), is plausible, but it is difficult to prove that Akkadian was the direct source of
the PII word. The Akkadian word, in turn, is likely borrowed from West Semitic ʿai̯ r ‘foal’
(CAD H, p. 118), which is semantically and formally further removed from *kHara-.
Witzel (2003, p. 29) believes PII *kHara- to be connected to Skt. garda-bhá- ‘donkey’
and ToB kercapo ‘donkey’ < PToch. *kercäpā-. The Tocharian word is likely borrowed from
Indo-Iranian, since *-bha- is a common “animal-suffix” in Indo-Iranian.34 Tocharian probably
borrowed the word before the PII vowel merger as *gord(h)ebho-, since *d(h) was palatalized to
*c within Tocharian (Adams, 2013, p. 210). If *gord(h)e(bho)- was borrowed into early PII, it is
possible that *kHara- was borrowed into PII at a later stage. Possibly, PII *g- (= [gɁ] / [kɁ]) and
kH- (= [kɁ] or [kh]) represent different adaptations of the same sound in the source language,
since both contain a glottalic element.
26. PII *kšīra- ‘milk’

Ind. Skt. kṣīrá-
Ir. MiP šīr, Y-M xšīra
Taken as a loanword by Lubotsky (2001b), partly due to the cluster *kš-. EWAia (I, p. 433) does
not offer an IE etymology. Kroonen (2013, pp. 261-2) argues for a connection to PGm. *hwaja
~ *huja- ‘whey’, which he derives from an i-stem adjective *tkw-ōi-. According to him, PII
*kšīra- would derive from *tkwih2-ro-, a *-ro- derivation of a *h2-collective of *tkw-ōi-. A
similar derivation would be Alb. hirrë ‘whey’ < *tkwiH-r-neh2-, although this may be a
loanword from Hungarian (Szemerényi, 1958, p. 171). To explain the semantics, Kroonen
argues that *tkw-ōi- has the same root as PGm. *Þinhla- ‘curdled milk’ and Skt. takrá-
34
Cf. Skt. vr̥ṣa-bhá- ‘bull’
40
‘buttermilk’ < *t(e)mk-lo-. This is difficult, however, since the root *temk- ‘to thicken’ has a
nasal which is absent in *tkw-ōi-. Furthermore, it is implausible that a word for raw milk would
be derived from a root meaning ‘to thicken’, which is clearly associated to processed milk.
Based on these considerations, Kroonen’s etymology of PII *kšīra- seems unlikely.
27. PII *kućsi- ‘round side of the body’

Ind. Skt. kukṣí- ‘cheek’
Ir. Sogd. qwšy- ‘side of the body’
Nur. W. küc ‘belly’
Taken as a loanword by Lubotsky (2001b). The morphology is not easily explainable from an
IE perspective. EWAia (I, p. 360) argues for a derivation based on an unattested s-stem *kuć-
as-. However, the root of *kuć-as- itself is unknown: the proposed root cognates LAv. kusra-
‘arching’, Skt. kuśá- ‘grass’ or Skt. kuśī̄́ ‘?’ also lack IE etymologies and are semantically
dissimilar.
28. PII *matsi̯ a- ‘fish’

Ind. Skt. mátsya-
Ir. LAv. masiia-
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 298) supports a connection to PGm.
*mati- ‘food’. According to this etymology, *matsi̯ a- is a *-i̯ o-derivative of an unattested s-stem
*med-s-. However, such a derivational chain is improbable and it is unclear why the word would
mean ‘fish’.
PU *maća ‘fish net’, reflected in Mar. mača ‘fish net’, Sam. Selk. Ke. maazeng, N mā̄́šek,
Ty. mā̊̄sa ‘net, trawl’ (UEW, p. 263), is a possible source of the Indo-Iranian word, in which
case the meaning ‘fish’ could have developed metonymically from ‘fish net’.35 However, UEW
concedes that the reconstruction is uncertain due to the scarcity of cognates. Indeed, neither
Sammallahti (1988) nor Aikio (2015) reconstruct PU *maća. According to them, Mari -č-
regularly corresponds to Selk. -tc-, -c- or -tč-, cf. PU *ceca ‘uncle’ > Mar. čəčə ~ Selk. Ke.
citca, cica, N četčeka. It is thus unlikely that PII *matsi̯ a- was borrowed from Uralic.
35
I thank Niels Schoubben for bringing this possibility to my attention. Décsy (1990, p. 89) reconstructs *matja,
where tj is simply an alternative way to write ć.
41
29. PII *mr̥ga-
Ind. Skt. mr̥gá- ‘wild forest animal’
Ir. LAv. mərəγa- ‘bird’
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 371) mentions a proposed connection
to Gr. μάργος ‘mad, furious’. However, this word shows so much irregular variation in within
Greek that it is most likely Pre-Greek in origin (Beekes, 2010, p. 905). As the semantics relate
to flora and fauna, a loanword seems likely.
30. PII *muska- ‘testicle’

Ind. Skt. muṣká- ‘testicle’
Ir. MiP mušk ‘musk’
Taken as a Wanderwort by Lubotsky (2001b), who believes that MiP mušk was borrowed from
Indic. However, the MiP form could also have been inherited from PII, especially since the
suffix *-ka- is characteristic of other proposed PII loanwords. EWAia (II, p. 363) argues that
Skt. muṣká- ‘testicle’ is derived from PIE *mū̄́s- ‘mouse’, evolving from a literal meaning ‘little
mouse’, similar to Lat. mūs-culus ‘muscle’. As muṣká- has a short ŭ, this would require the long
*ū of *mū̄́s- to be explained by monosyllabic lengthening. However, the acute accent of SCr.
mȉš ‘mouse’ < *muHs-, ToB maścitse < *mu̯H̥s- (Beekes, 2010, p. 985) and Nur. Prasun mǖs’ū
‘mouse’ (without RUKI) all point to PIE *muHs-. Therefore, Skt. muṣká- cannot be derived
from ‘mouse’. Gr. μόσχος ‘musk’ is likely borrowed from an Iranian source, and further spread
to Lat. muscus ‘musk’.
31. PII *nagna-

Ind. Skt. (AVP+) nagnáhu- ‘yeast’
Ir. Sogd nγny, Pashto naγan, Bal. nagan, naγan, MiP nān ‘bread’ < PIr. *nagna-
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 6) assumes that the Sanskrit word was
borrowed as *nagna-hvā- from Iranian *nagna-xvada-, cf. MoP nānxvāh- ‘bread spices’. This
is rather ad hoc, since the change from *nagna-hvā- > *nagna-hu- in Sanskrit is left
unexplained.
Bailey (1979, p. 179) derives *nagna- ultimately from PIr. *ni-kana- lit. ‘put down (into
the ashes)’. Semantically, this is acceptable, but the change from *ni- > *na- is irregular. More
likely, this is a loanword.
42
32. PII *pāpa- ‘bad’
Ind. Skt. pāpá-
Ir. LAv. pāpa-
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 120) offers no IE etymology, rejecting
a connection to Skt. pāmán- ‘skin disease’ and Gr. πῆμα ‘disaster, sorrow’. Morphologically,
*pāpa- looks non-IE due to the suffix *-pa-, found in other proposed loanwords. Assuming
reduplication of a root *peH- (*pe-pH-o-) cannot explain *pāpa-.
33. PII *parsa- ‘sheaf’

Ind. Skt. parṣá-
Ir. LAv. parša-
Taken as a loanword by Witzel (2003, p. 33). EWAia (II, p. 101) gives no IE etymology. The
suffix *-sa- is perhaps found in another loanword, *pīi̯ ūša-, although in *parsa-, the *s could
also be analyzed as part of the root. The fact that ‘sheaf’ is an agricultural term increases the
likelihood that *parsa- is a loanword.
34. PII *pau̯asta-

Ind. Skt. pavásta- ‘cover, garment’
Ir. OP pavastā- ‘clay envelope’
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 105) dismisses the idea that the Iranian
word is borrowed from Indic and concludes that the word has no etymology. The long final -ā
in OP is regular. Since the morphology (suffix *-asta-?) is inexplainable from an IE perspective,
*pau̯asta- is likely a loanword.
35. PII *rāći-

Ind. Skt. rāśí- ‘heap, mass’
Ir. Pashto ryāša- ‘heap (of grain)’
Taken as a possible loanword by Lubotsky (2001b), who does not exclude a connection to Skt.
raśmí- ‘reins, rope’. This etymology is hesitantly advocated in EWAia (II, p. 449), who
postulates a root PII *rać- ‘to bind’ < PIE *laḱ-, with a supposed cognate in Lat. laqueus ‘loop
of rope’ and lacio ‘to entice’. However, as de Vaan (2008, p. 327) points out, Lat. laqueus
43
reflects a labiovelar *kw rather than a palatal *ḱ. As for lacio, it is likely connected to Lat. lacer
‘mutilated’ and Gr. ἀπέληκα ‘I have torn off’ and λακίς ‘tatters of clothes < PIE *l(e)h2k- ‘to
tear’ (Beekes, 2010, p. 826), in which case the a-vocalism in Latin reflects a laryngeal36,
incompatible with Skt. raśmí-. Formally, PII *rāći- ‘heap’ could derive from *leh2ḱ-i- but the
semantics are difficult to explain. The word may have a connection to agricultural terminology.
36. PII *ringa- ‘mark’

Ind. Skt. liṅga-
Ir. LAv. (haptō-)iringa-
Taken as a loanword by Witzel (2003, p. 33). EWAia (II, p. 479) argues that the word is related
to Lith. lýgus ‘alike’ < PIE *leig-, also reflected in PGm. *līka- ‘alike’. The foremost problem
of this etymology is the *-n- in Indo-Iranian, resembling the IE nasal infix, which is entirely
unexpected in a nominal form. Comparative evidence for this nasal has been argued to exist in
Gr. ἐναλίγκιος ‘like’ (Pisani, 1981, p. 207), but this is impossible due to the irregular
correspondence Gr. -nk- : PII -ng-. A root of the structure *Reing-/*Rieng- is unparalleled in IE
(cf. LIV).
37. PII *r̥si- ‘seer’

Ind. Skt. r̥̄́ṣi-
Ir. Av. ərəši-
Taken as a loanword by Lubotsky (2001b), due to the abnormal initial accent in Sanskrit. As
shown by Lubotsky (1988, p. 54), Sanskrit. i-stems with zero-grade in the root are generally
oxytone.
38. PII *sćāga- / *sćaga- ‘male goat’

Ind. Skt. chā̄́ga-, chagalá-
Ir. Oss. sæǧ/sæǧæ, Wakh. čəy
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 558) argues that the long root vowel
in Sanskrit is a back formation from the feminine chā̄́gā- ‘female goat’, itself a
vr̥ddhi-derivation of *sćaga-. However, the variation may also be an indication of borrowing.
36
And not PIE *o > Lat. a / l_
44
A connection to PGm. *skēpa- ‘sheep’ has been suggested. However, that the Germanic
*-p- would be the result of dissimilation from earlier *skēka- (cf. EWAia I, p. 559) is ad hoc.
One could postulate a parallel borrowing of a preform *skē̄̆gwo- into Germanic and Indo-Iranian,
but it is improbable that Germanic would nativize *gw as *b, since Germanic retains the
phoneme *gw. Moreover, if the long vowel of Skt. chā̄́ga- is secondary, the PII form *sćaga-
looks less similar to PGm. *skēpa-.
39. PII *spāra- ‘ploughshare’

Ind. Skt. phā̄́la-
Ir. MoP supār, Išk. uspir, Wakh. spūndr
Taken as a loanword by Lubotsky (2001b), since the anlaut Skt. ph- : Ir. sp- is irregular.37
Semantically, a loanword is not unexpected, since ‘ploughshare’ is an agricultural term. EWAia
(II, p. 204) argues that Skt. phā̄́la- < *spāla- and that PII *spāla- < PIE *spelH- ‘to split’ can
be reconstructed. However, the proposed cognate OCS plěvǫ ‘to weed, separate the husk from
the grain” (LIV, p. 577) is semantically unconvincing and requires the assumption of s-mobile,
which is not synchronically attested in either Slavic or Indo-Iranian.
Another problem is that *sp- > ph- is a Middle Indic development, which requires the
assumption that Skt. phā̄́la- is a ‘prakritism’, i.e. a colloquial form which had undergone a
‘Middle Indic’ sound change already in Vedic times. However, this hypothesis may be
supported by the context in which Skt. phā̄́la- is attested. The word occurs in RV X.117 (the
‘Praise of Generosity’), which, according to Jamison & Brereton is “unusual in both subject
matter and tone” and makes “no mention of divinities […], an almost unique situation in the
R̥gveda” (2014, p. 1586). Importantly, the style is “colloquial and conversational” (ibid.). In
this context, a colloquial word is not unexpected. Since Vedic is not equal to Proto-Indic, it is
not impossible that certain lower-prestige dialects parallel to Vedic had already undergone a
sound change *sp- > ph- at this time. Based on these considerations, *spāra- will be treated as
a PII loanword.
37
The chronology of the development of PII sP-clusters in Indic is complicated. As pointed out by Kobayashi
(2004, p. 72), the change of PII *sć > Skt. (c)ch precedes the change of Skt. sp- > MI (p)ph by hundreds of years.
45
40. PII *stuka- / *stupa- ‘tuft of hair’
Ind. Skt. stúkā-, stupá- ‘hair’, stū̄́pa- ‘hair, top beam of house’
Ir. Oss. styg/stug, Y-M stūγ ‘long hair’
Taken as a loanword by Lubotsky (2001b), due to the suffixal variation -ka-/-pain Sanskrit.
Although EWAia (II, p. 760) agrees that Skt. stúkā- and stupá- are connected, it suggests that
the variation can be ascribed to dissimilation or different extensions of an unattested root *stu-,
neither of which seems likely. Moreover, the suffixes *-ka-/*-pa- are attested in other proposed
loanwords.
41. PII *u̯āćī-

Ind. Skt. vā̄́śī- ‘axe’
Ir. LAv. vāsī- ‘pointed knife’, Oss. was ‘axe’
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 548) assumes IE origin in connection
to OHG wahs ‘sharp’, or as an Indo-Iranian innovation based on the root Skt. vāś- ‘to roar’ (cf.
Germ. Klinge ‘blade’ ~ klingen ‘to sound’). The latter explanation is semantically highly
uncertain and further unhelpful since vāś- lacks IE comparanda. In the AV, Sanskrit has a
parallel form vāsī- with an unexpected dental s, which could indicate non-IE origin.
Parpola (2012, p. 161) and Kümmel (2019) argue that PII *u̯āćī- is borrowed from PU
*weŋći ‘knife’, reflected in Finn. veitsi ‘knife’, veitsä ‘to cut’, Hung. vés ‘to chisel’ (cf. UEW,
p. 565). However, the reconstruction of PU *weŋći is uncertain, since the regular outcome of
*ŋć38 is Hung. gy (Pystynen, 2014). The word is not reconstructed by either Sammallahti
(1988), Aikio (2015) or Zhivlov (2014). Until a solution for the problematic Uralic
correspondences is found, the connection to Indo-Iranian remains speculative.
42. PII *u̯and(H)- ‘to praise’

Ind. Skt. vandi-
Ir. Av. vaṇd-
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 502) and LIV (p. 681) project
the Indo-Iranian forms to PIE *u̯end-. However, the seṭ-character of the Sanskrit root points to
*u̯endH-, which we would expect to give Skt. **vandh-. This increases the likelihood that PII
38
As Pystynen (2014) points out, the proper reconstruction is *ńć.
46
*u̯and(H)- is a loanword. The root has no archaic derivations that would indicate that it is old
(Macdonell, 1916, p. 416). As it is often used in religious contexts, *u̯and(H)- could have been
borrowed along with deity names such as *indra- etc.
43. PII *u̯arāj̄́ha- ‘wild boar’

Ind. Skt. varāhá-
Ir. LAv. varāza-
Taken as a loanword by Lubotsky (2001b), due to the trisyllabic CVCV̄CV structure. EWAia
(II, p. 514) mentions a proposed connection to PCelt. *ǵhoru̯o- with irregular
“Konsonantenvertauschung”, which can hardly be correct. Although PFV *orase ‘boar’ looks
related, it is probably borrowed from an Indo-Iranian language (Rédei, 1986, p. 54) .
44. PII *umā-(kā)- ‘flax, linseed’

Ind. Skt. úmā- ‘flax’
Ir. Y imoγō, ümoγō, M yimaγå ‘linseed’
Taken as a Post-PII Wanderwort by Lubotsky (2001b). EWAia (I, p. 225) dismisses non-IE
origin but offers no alternative etymology. The initial vowels of the Yidgha-Munji words derive
from *u- with umlaut (Morgenstierne, 1938, p. 96). Since the suffix -kā- is productive in
Iranian, there is nothing against assuming that this word entered the language in PII times. Skt.
kṣumā- ‘flax’ may be indirectly related. It is noteworthy that flax was cultivated in the BMAC
culture (Spengler et al., 2014).
45. PII *u̯rīj̄́hi- ‘rice’

Ind. Skt. vrīhí-
Ir. Pto. wr’iže, OP *vrīzi- < PIr. *u̯rīj̄́i-, Khot. rrīysū < PIr. *u̯rīj̄́uka, Orm. rízan < PIr.
*u̯rīj̄́ana-, Sogd. rysk < PIr. *u̯rīj̄́aka-
Taken as a Wanderwort by EWAia (II, p. 598). Kümmel (2017, p. 283) reconstructs PII *u̯rīj̄́hi-
based on Skt. vrīhí-, Pto. wr’iže and OP *vrīzi-, the latter being reconstructed based on Elam.
mi-ri-zi-iš ‘rice’. The Elamite form is probably borrowed from “Median”, where PIr. *j̄́(h) > z.
The other Iranian forms show various (productive) suffixes -uka-, -aka-, and -ana-,
neither of which precludes that they were inherited from PII and later reshaped.
47
It is often assumed that Gr. ὄρυζα ‘rice’ was borrowed from an Iranian language (Beekes,
2010, p. 1112). If so, Greek probably adopted the word when υ = [i] and ζ = [z], as the ancient
Greek pronunciation would not reflect the Iranian phonology.
The fact that rice was not cultivated in the BMAC culture (Spengler et al., 2014) reduces
the likelihood that *u̯rīj̄́hi- originated in the Central Asian Substrate, even though it is
reconstructable to PII.
In addition to the above, Iranian has another similar word for rice, reflected by MiP blnc,
brynz, Sogd. βrync ‘rice’, reconstructable as PIr. *brinǰa- (Kümmel, 2017, p. 283). A
descendant of PIr. *brinǰa- is probably the source of Gr. ὀρίνδης ‘rice flour bread’ (Beekes,
2010, p. 1102). Furthermore, Arm. brinj ‘rice’ seems to be borrowed from a Middle Iranian
source. Since it only exists in Iranian, *brinǰa- was most likely borrowed at a later stage than
PII *u̯rīj̄́hi-. However, PIr. *brinǰa- contains an *-n- which is absent in the older word. Since *ī
is unlikely to develop into *-in-, this indicates that *brinǰa- was borrowed from a different
source language than *u̯rīj̄́hi-, rather than the a later historical stage of the same language.
46. PII *u̯r̥tka- ‘kidney’

Ind. Skt. vr̥kká-
Ir. OAv. vərəδka-, MiP gurdag, Khot. bilga-, Y-M wulγa
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 571) argues that it is derived from
*vart- ‘to turn’. However, since the suffix -ka- is denominal, whereas vr̥t- is a plain root, this is
improbable. Also, the etymology is not semantically convincing.
3.3. Loanwords II: Proto-Indo-Iranian, borrowed after certain sound changes
47. PII *čāt(u̯āla)- ‘pit, well’

Ind. Skt. (Br.) cā̄́tvāla-
Ir. LAv. cāt-, Sogd. čʾt, Bactr. σαδο
Taken as a loanword by Lubotsky (2001b). No IE etymology has been suggested (EWAia I, p.

539). The Sanskrit word is not directly comparable to the Iranian forms due to the
“suffix” -vāla-, itself of unknown origin. Skt. cā̄́tvāla- has the trisyllabic CVCV̄CV structure,
which is unparalleled in Iranian. Bactr. σαδο reflects either *čātā̄̆- or *čāti- (Davary, 1982, pp.
137, 264). The morphological variation between Indic and Iranian suggests a more recent time
of borrowing.
48
48. PII *i̯ au̯īi̯ ā- ‘canal’
Ind. Skt. *yavīyā̄́- (metrically restored), yavyā̄́-
Ir. OP yauviyā-
Taken as a loanword by Lubotsky (2001b), on account of its CVCV̄CV structure. EWAia (II, p.
405) does not offer an IE etymology. According to Witzel (2003, p. 32), the Sanskrit and OP
words cannot go back to the same proto-form: this is incorrect, however, because a > au /_v
occurs elsewhere in OP, and the length of i cannot be deduced from the script, cf. tauviyah- ~
Skt. távyas- / távīyas- ‘stronger’. For the inclusion of *i̯ au̯īi̯ ā- in layer II, I refer to section 4.3.
49. PII *ǰaǰha/ukā̄̆- ‘hedgehog’

Ind. Skt. jáhakā- (YV+), Lahnda jahā-
Ir. LAv. dužaka-, Bal. ǰaǰuk, dužux, MoP žūža-
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 582) mentions no convincing IE

etymology. Besides, it has reduplication, which is morphologically remarkable, and words for
animals are easily borrowed.
The initial d- of LAv. dužaka- and Bal. dužux (instead of expected *j-) could be explained
as dissimilation of ǰ… ǰ > d…ǰ39, although OAv. jījišəṇtī ‘they conquer (repeatedly)’ < PII *ǰai-
(Cheung, 2007, p. 222) renders this unlikely. Another question is whether the u-vowel in LAv.
dužaka- is old. AiWB (p. 755) analyses dužaka- as a compound duž- + -aka- ‘having bad hooks’.
Given the related Indic words, this cannot be the actual etymology (Av. duž- ‘bad’ : Skt. duḥ-
< *dus-), but folk-etymologically this analysis is conceivable. Thus, it is possible that Iranian
speakers remodelled *ǰažaka- (with regular ž < *ǰ) as *duž-aka- ‘having bad hooks’.
Although the -u- in the initial syllable of some Iranian forms may be secondary, Bal. ǰaǰuk
points to PIr. *ǰaǰuka-, whereas the Indic forms point to *ǰaǰhakā-. The suffixes -uka- and -aka-
could be secondary innovations within Iranian and Indic, in which case Lahnda jahā- preserves
an older form. However, the morphological and phonological variation suggests that the word
was borrowed in late PII. Since it occurs before *u and *ā, *ǰh was most likely not palatalized
within PII but borrowed as such.
39
Cf. Khot. dasta- ‘hand’ < *j̄́asta- with dissimilation of the fricative element of *j̄́.
49
50. PII *kárus- ‘damaged’
Ind. Skt. kárū-ḍatin- ‘with damaged teeth’
Ir. Sogd. krwʾ ‘gap’, krw δntʾk ‘with damaged teeth’, MoP karve ‘decayed teeth’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 313) gives no IE etymology. MoP
karve is part of a larger group of borrowings from Sogdian (Henning, 1939, p. 96) and thus
does not inform on the PII situation.
Sogd. krw δntʾk is read /karw dandāk/. Importantly, the first part of the compound is also
attested as simplex karw’ /karwā/ (Gharib, 1995, p. 194), for which reason EWAia (I, p. 313)
reconstructs PIr. *karu̯a- ‘gapped, damaged’. This reconstruction is homophone to the preform
of LAv. kaurva-40 ‘thin-haired’ < PIr. *karu̯a- ‘thin-haired’, cognate to Skt. kū̄̆lva- ‘thin-haired’
< PII *kl̥ ̄́ H-u̯o-, and further related to Lat. calvus ‘bald’.41 Henning (1939, p. 96) proposed an
etymological relation between PIr. *karu̯a- ‘thin-haired’ and PIr. *karu̯a- ‘gapped, damaged’.
However, if the latter is linked to Skt. kárū-ḍatin-, this is impossible, since the retroflex -ḍ- and
long -ū- show that Skt. kárū- goes back to *karuž- < *karuš-. The suffix *-us- (s-stem derived
from u-stem?) is rare but not unparalleled in Indo-Iranian (AiGr. II, 2, p. 477).
The fact that Skt. kárū- goes back to *karuž- precludes a scenario where Skt. kárū-ḍatin-
was borrowed from (proto-)Sogdian, as Sogdian preserves -z- in this position, e.g. ’ztyw <
uzdahyu- (Gershevitch, 1961, p. 44). The reverse scenario, that Sogd. krw δntʾk (with alternative
reading as /karu dandāk/) was borrowed from Sanskrit, would explain the lack of -z in Sogdian,
but cannot explain the existence of Sogdian simplex krw’ /karwā/ ‘gap’. To save the latter
scenario, one would have to separate Sogd. krw’ ‘gap’ from krw δntʾk ‘with damaged teeth’,
and instead connect it to LAv. kaurva- ‘thin-haired’ and Skt. kū̄̆lva- ‘id.’, which seems rather
ad hoc.
As both possible directions of borrowing seem impossible, the remaining possibilities are
to assume that Sogd. krw δntʾk was remade based on a synchronic stem *krw which had
somehow lost its -s, or that Sogdian and Sanskrit borrowed from slightly different sources. As
40
For the meaning ‘thin-haired’, cf. Lubotsky (1997, p. 42). Naturally, the reading of LAv. kaurva- as ‘thin-
haired’ and not ‘damaged’ hinges partly on the etymological identification with Skt. kū̄̆lva- ‘thin-haired’. For the
compounds kaurvō.gaoša- ‘with … ears’, kaurvō.dūma- ‘with … tail’ and kaurvō.barəša- ‘with … neck’ (AiWB,
p. 456), all said of horses, kaurvō- could theoretically be read as either ‘damaged’ or ‘thin-haired’, although the
latter seems slightly more likely in the context.
41
Cf. de Vaan (2008, p. 85). To account for the a-vowel of Lat. calvus, we must reconstruct a thematicized weak
u-stem *klH-eu-o-. In Indo-Iranian, the word was thematicized as *kl̥ ̄́ H-u̯o-, with accented zero-grade in the root.
This explains the long *ū in Skt. kūlva- and non-labialized PIr. *karu̯a-. EWAia (I, p. 449) suggests that Skt.
kharvá- ‘mutilated’ should be equated to PIr. *karu̯a- ‘damaged’, but this is impossible since Skt. kh- < *kH- and
corresponds to Ir. x-.
50
the compound with the IE word for ‘tooth’ is quite specific, it seems better to assume the former
scenario, i.e. a single borrowing event with a subsequent loss of -s in Sogdian.
The structure of PII *kárus- is difficult to derive from Pre-PII. With its non-palatalized
anlaut, it could only reflect Pre-PII *kórHus-.42 However, this is difficult, since Pre-PII
*kórHus- should have undergone the laryngeal accent shift (> **korHús-) and laryngeal
metathesis (> **koruHs-). Accordingly, as *k- is not palatalized, PII *kárus- must either have
been borrowed with an *a (after the phonologization of *a), with an *ŏ (after BrL), or after the
palatalization of velars.
51. PII *mai̯ ūkHa- ‘peg’

Ind. Skt. mayū̄́kha- ‘peg for stretching the woof’
Ir. OP mayūxa- ‘doorknob’, Sogd. myγk ‘peg’, MiP and MoP mēx ‘peg, nail’, Oss.
mīx/mex ‘stake’
Taken as a loanword by Lubotsky (2001b), who argues against a derivation from the root may-
‘to erect, build’ (EWAia II, p. 317), since the “suffix” *-ūkHa- is left unexplained. For the
inclusion of *mai̯ ūkHa- in layer II, I refer to section 4.3.
52. PII *pīi̯ ūša- ‘beestings’

Ind. Skt. pīyū̄́ṣa-
Ir. Wakh. pyix̄̆, Munji fə̄́yū
EWAia (II, p. 138) considers this word to be related to Skt. páyas- ‘milk’ and pay- ‘to swell’.
Lubotsky (2001b, p. 303) rejects this due to the unusual derivation (suffix -ūsa- / -sa-?) and the
unexpected long ī, and argues that the word is a characteristic loanword with the structure
CVCV̄CV. The idea that Skt. pīyū̄́ṣa- is a compound of pī- ‘to swell’ + yū̄́ṣa- ‘broth’ (AiGr. II,
2, p. 500) is semantically unlikely and formally problematic, since pī- is a bare root. For the
inclusion of *pīi̯ ūša- in layer II, I refer to section 4.3.
42
The laryngeal closes the syllable, accounting for the lack of BrL. Other input forms are impossible: Pre-PII
*kerus- > **čarus-, *korus- > **kārus-, *kN̥rus- > **kaNrus-, *keh2-3rus- > **kaHrus-, and *kHerus- > Ind.
**kharus- ~ Ir. **xarus-.
51
53. PII *pusća- ‘tail’
Ind. Skt. púccha- ‘tail’
Ir. LAv. pusa- ‘head dress’
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 140) maintains the possibility that the
word is related to PGm. *fuhsa- ‘fox’. This etymology is supported by Kroonen (2013, p. 158),
who reconstructs *puk-so- for the Germanic word and *puḱ-sk-o- for Indo-Iranian. However,
the fact that Indo-Iranian and Germanic do not reflect a single PIE form, and the abnormal
suffixes *-so- / *-sko- weakens the plausibility of the etymology. Furthermore, the
correspondence Skt. -cch- : Av. -s- reflects PII *-sć- < *-sč-. Yet the paradigm of *puḱ-sk-o- (>
*pu-sk-o-43) would not provide a palatalizing context for the *-sk- cluster, except in the vocative
*puḱ-sk-e, which, for a word ‘tail’ is a highly unlikely model of analogy. Thus, PII *pusća-
must have been borrowed after the PII phonologization of palatalized *sč.
A possible source of *pusća- is PFU *ponci ‘tail’.44 However, since it is unclear why the
PFU *-n- is not reflected in PII, the connection remains speculative.
3.4. Loanwords III: Post-Proto-Indo-Iranian
54. Pre-II *?
Ind. Skt. áṇu-
Ir. MiP ʾrzn, MoP arzan, Pto. zdən, Wakh. yirzn < PIr. *(H)arj̄́aná-, MiP ʾlwm, Baxt.
halum
Nur. V. üǰ’ü̃, A. az’̣ũ , W. ə̃zü < *(H)arj̄́aná-
Kümmel (2017, p. 283) proposes that the Iranian and Nuristani words are loanwords, and adds
Skt. áṇu- ‘millet’ as a possible cognate. This connection presupposes PII *-rj̄́n- > Indic -ṇ-, for
which I know of no other examples. Skt. áṇu- is thus unlikely to be related to the Iranian and
Nuristani words. MiP ʾlwm and Baxt. halum have been influenced by *ganTuma- ‘wheat’.
43
*-ḱsk- was simplified to -sk- probably already in PIE, cf. *prḱ-ske- > *pr-ske- ‘to ask’.
44
For the reconstruction cf. Sammallahti (1988, p. 547).
52
55. Pre-II *banγα- ‘hemp’
Ind. Skt. bhaṅgá- ‘hemp’
Ir. LAv. baŋha- ‘a plant, narcotic’, MoP bang ‘hemp’
Taken as a loanword by Witzel (2003, p. 34). EWAia (II, p. 241) argues that MoP bang is
borrowed from Sanskrit. For Skt. bhaṅgá-, a semantically unconvincing connection to bhañj-
‘to break’ is suggested. LAv. baŋha- (< (virtual) *bhansa- / *bhasa-) is semantically and
formally close but does not correspond regularly to Skt. bhaṅgá-. Kümmel (2019) reconstructs
PII *bhanga-, which he interprets as a borrowing from PU *pe̮ŋka- ‘mushroom’. However, the
reconstruction is based only on Sanskrit. Given the synchronically close form and meaning of
Skt. bhaṅga- and LAv. baŋha-, it seems more plausible that they represent parallel Post-PII
borrowings, with different adaptations of a Pre-II *banγα- (vel sim.).
56. Pre-II *ćika(tā)-

Ind. Skt. síkatā- ‘sand, gravel’
Ir. OP θikā- ‘gravel’, Sogd. šykth, Khot. siyatā-
Taken as a loanword by Lubotsky (2001b) and EWAia (II, p. 728) due to the irregular
correspondence Ind. *s- : Ir. *ć- and the absence of an IE etymology. Lubotsky (2001b, p. 306)
argues for an Indic >> Iranian direction of borrowing, but this need not be the case. Since OP
and Sogdian seem to reflect PIr. *ćika(tā)-, where *ć is still an affricate, it was probably not
borrowed from Indic, which has initial *s-. Khot siyatā- reflects *ćii̯ atā-.
To assume an independent parallel borrowing by Indic and Iranian is complicated by the
difficulty of determining a plausible common source form of the word. If the source language
had an initial dental affricate [ʦ], we would expect PIr. *ć- but Skt. ts- or possibly kṣ-.
Reconstructing a palatal affricate [ʨ] might be compatible with PIr. *ć- but can hardly explain
Skt. s-. Ultimately, however, parallel borrowing cannot be excluded, since irregular adaptations
to the native linguistic structure may have occurred.
The other direction of borrowing, Iranian >> Indic, is possible if one assumes that Skt.
síkatā- was borrowed from an Iranian language where PIr. *ć- > *s-. Given the difficulty of the
two other scenarios, this might be the most plausible one.
53
57. PII *dū̄̆rća- / *dr̥ća- ‘(goat’s) wool, hair’
Ind. Skt. dūrśá- ‘(large) garment’
Ir. Wakh. δirs/δɪrs/δürs/dərs ‘wool of goat/yak’, Šu. δox̄̆c ‘body hair, course cloth’,
Y-M lirs/līrs/lurs ‘goat’s hair’
EWAia (I, p. 740) reconstructs PIE *dr̥H-ḱo-, but the non-IE-looking suffix *-ḱo- is not
reflected in any of the proposed root cognates. Lubotsky (2001b, p. 311) takes it as a loanword
based on the irregular correspondence between Indic and Iranian, reconstructing PII *dr̥Hća- /
*dr̥ća-. However, PII *-r̥H- > Skt. -ūr- only regularly occurs in labial contexts, i.e. C+labial _ or
_Cu̯, which means that Skt. dūrśá- rather reflects PII *dūrća-.
Wakh. δirs/δɪrs/δürs/dərs reflect dialectal variants (Morgenstierne, 1938, p. 481). PII
*dr̥ća- is a possible ancestor of Wakh. dərs (ibid.). However, -ə- may also reflect *-u-, e.g.
Wakh. dəγd ‘daughter’ < PIr. *dugdar-. The other forms, Wakh. δirs/δɪrs/δürs, do not show the
normal outcome of *r̥, nor the outcome of *r̥ in a labialized context (e.g. Wakh. pʊrs ‘to ask’ <
PII *pr̥sća-). Instead, the vowels of Wakh. δirs/δɪrs/δürs seem to reflect delabialized *-u-
(Morgenstierne, 1938, p. 480). Thus, *durća- is a more likely origin of the Wakhi forms.
The same is true for Y-M lirs/līrs/lurs, where the vocalism of the attested dialectal
variants could go back to *ū̄̆ (Morgenstierne, 1938, pp. 96-7), thus being compatible with a
reconstruction PIr. *durća- or *dūrća-.
In Šughni, -o- can be the outcome of *r̥, but only when a long *ā in the following syllable
causes a-umlaut (*r̥ > *ūr > *ār > ox̄̆, cf. Sokolova (1967, pp. 56, 58)). Moreover, *r̥ > *ūr
occurs in stressed position, whereas unstressed *r̥ becomes *ir or *ar (Sokolova, 1967, p. 61).
Since Skt. dūrśá- has oxytone accentuation, this would mean that either Indic or Iranian
underwent an accent shift. On the other hand, the normal outcome of PIr. *ū̄̆ is Šughni u, unless
affected by i- or a-umlaut, becoming i or a (Sokolova, 1967, p. 49). It is thus difficult to connect
Šu. δox̄̆c the other Indo-Iranian forms by regular sound changes.
In conclusion, the attested words variously point to PII *dūrća-, *dŭrća-, or *dr̥ća-. This
suggests that the word was borrowed after the disintegration of PII.
54
58. Pre-II *ganDəru̯a- ‘a mythical being’
Ind. Skt. gandharvá-
Ir. LAv. gaṇdərəβa-
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 462) offers no IE etymology but
mentions the irregular correspondence with Gr. Κένταυροι. As pointed out by Lubotsky (2001b,
p. 303), the Sanskrit and Avestan forms are irregular correspondences, since Skt. -arvá- would
reflect PII *-aru̯a- whereas LAv. -ərəβa- would reflect PII *-r̥b(h)a-. The most likely
explanation is that LAv. gaṇdərəβa- was borrowed after the sound change *b > β / V_V. This
indicates that the source language had a fricative sound, which was adapted as -β- in Avestan,
since PIr. *u̯ was still a glide. In Indic, however, the fricative was adopted as Skt. -v-. Skt.
gandh- may reflect earlier *gandh- or *ghandh- with Grassmann’s Law.
59. Pre-II ganTi- ‘smell’

Ind. Skt. gandhá- / -gandhi- ‘smell’
Ir. LAv. gaiṇti- ‘bad smell’, OP gasta- ‘evil, repugnant’, Khot. ggañu ‘stench’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 461) concedes that the origin is
uncertain but does not consider non-IE origin. Bailey (1979, p. 79) postulates a root *gan- ‘to
smell’, which would be the basis of the Iranian forms (e.g. -ti- derivation in Av.), with the
variant *gan-d- in Indic. However, the dental stop must have been part of the root, as shown by
OP gasta- < *gn̥t-ta- and other Iranian forms, which all show a root-final stop (Cheung, 2007,
p. 103). However, Khot. acc.sg. ggañu seems to point to *d rather than *t (Bailey, 1979, p. 79).
Skt. gandhá- has a variant -gandhi- in compounds, which bridges the gap to LAv. gaiṇti-.
Due to the irregular correspondence Indic dh : Iranian t, the word was most likely
borrowed in Post-PII times. Especially noteworthy is that the same irregular correspondence is
found in *ganTuma-.
60. Pre-II *ganTuma-

Ind. Skt. godhū̄́ma- ‘wheat’
Ir. LAv. gaṇtuma- ‘wheat’, Pto. γan|əm, Parth. gndm, Wakh. ɣ̌ədim, MiP gnm, Khot.
ganama, Bal. gandūm, Yazg. γwont
Taken as a loanword by EWAia (I, p. 498) and Lubotsky (2001b). Skt. godhū̄́ma- has probably
been affected by folk etymology, reanalyzed as a compound go-dhū̄́ma- lit. ‘cow-smoke’
55
(EWAia I, p. 498; Kümmel, 2017, p. 282). Thus, it could very well reflect earlier *gandhū̄́ma-.
The many Iranian cognates variously point to either PIr. *t or *d and long or short *ū̄̆ (Kümmel,
2017, p. 281). Yazg. γwont < *gantu- lacks the “-m-suffix” seen elsewhere in Indo-Iranian. The
variation clearly points to a parallel Post-PII borrowing.
Conversely, Witzel (2003, p. 31) reconstructs a single PII form *gantuma-, from which,
according to him, the forms with d and ū developed via folk etymology. However, there is good
reason for reconstructing a second preform *gandū̄̆ma-: firstly, forms reflecting *d are also
found in Iranian languages that otherwise show no apparent trace of folk-etymological
restructuring (since they preserve short ŭ and anlaut gan-); secondly, the same irregular
correspondence Indic dh : Iranian t is independently attested in Skt. gandhá- / gandhi- ‘smell’
~ LAv. gaiṇti- ‘bad smell’.
Similar words for ‘wheat’ appear outside of Indo-Iranian. EWAia (I, p. 499) mentions
Burušaski gur, pl. guri/eŋ ‘wheat’ < *γorum, as well as Hitt. kant- ‘wheat’ and Arab. ḥiṇtatun
‘wheat’ < *hnt-, to which Gr. χόνδρος45 ‘grain’ may be added. ToB kanti ‘bread’ probably
belongs here as well (Adams, 2013, p. 146).
According to Berger (1970, pp. 40-42), Burušaski is the source of all Indo-Iranian forms,
since the -m-suffix is native to this language. However, according to Berger’s data, a
suffix -m- never occurs as a separate morpheme in Burušaski, but always as part of a plural
morpheme -miŋ, e.g. ǰi ‘soul’ ~ ǰimiŋ ‘souls’ (1970, p. 37). Additionally, -miŋ is not directly
attested for gur in Burušaski, but is according to Berger indirectly attested in Rom. kharmin
‘wheat’ << OBur. *γor-miŋ. Berger furthermore argues that the r in Burušaski would have
developed in the plural *γor-miŋ < *γun-miŋ < *γund-miŋ and spread analogically to the
singular. The Indo-Iranian forms would have been borrowed from the earliest stage when the
dental stop was still preserved. The main problem with the whole scenario is the many
unverified steps of Burušaski historical development. It cannot be excluded, for example, that
Bur. gur was borrowed from an Indic language as *godum-, and then lenited -d- to -r- within
Burušaski.
Berger (1970, p. 40) argues that Rom. kharmin must have been borrowed from a form
with initial *γ-, and finds such a form in compounds like Bur. sauriŋ ‘ration’ < *sa-γuriŋ lit.
‘wheat of the day’. However, the fact that *-γ- is only found in compounds means that it could
represent an original *-g- that was lenited in intervocalic position.
45
However, the -ρ- remains unexplained. Also, Gr. χόνδρος could reflect earlier *χόνρος with epenthetic -δ-.
56
Witzel (2003, p. 31) further adduces Basque gari ‘wheat’, but this is most likely a chance
similarity since gari may come from *wari (Berger, 1970, p. 43).
In conclusion, the direction of borrowing of *ganTuma- is difficult to determine.
The -m-suffix more likely originates in an unknown source language than in Burušaski. Due to
the irregular correspondences, the word will be treated as a Post-PII borrowing.
61. Pre-II *Kai̯ ća- ‘hair’

Ind. Skt. kéśa- ‘head hair’, keśavá- ‘with long hair’
Ir. LAv. gaēsa- ‘curly hair’, gaēsu- ‘with curly hair’
Taken as a loanword by Lubotsky (2001b). To explain the irregular correspondence Skt. k- :

LAv. g-, EWAia (I, p. 401) postulates a contamination of original *geśa- with the semantically
close Skt. késara- ‘hair, mane’, yielding kéśa- by analogy. However, this scenario fails to
explain why ś was retained, while *g was substituted, even though the model késara- has a
dental s. Furthermore, Skt. késara- itself may be a loanword, given the absence of the RUKI-
rule in this word (EWAia I, p. 401). Perhaps Skt. kéśa-, késara- and LAv. gaēsa- are indirectly
connected as parallel Post-PII loanwords.
62. Pre-II *?
Ind. Skt. khaḍgá- (JB) ‘rhinoceros’
Ir. MoP karkadān ‘rhinoceros’
Taken as a loanword by Witzel (2003) and EWAia (I, p. 443). Arab. karkaddan ‘rhinoceros’
and Gr. καρτάζωνος ‘rhinoceros’ also belong here, probably as a borrowing from Persian.
Kuiper (1948, p. 137) adds Akk. kurkizānu, which is formally similar but means ‘pig, piglet’
(CAD K, p. 561). In the more western languages, the words contain the “suffix” *-d(z)an,
whereas Sanskrit lacks this element. Kuiper (ibid.) identifies the prefix kar- as evidence for
Proto-Munda origin. In that case, Skt. khaḍ- reflects *khar-. In any case, the attested Indo-
Iranian words are so dissimilar that they must be classified as Post-PII borrowings, probably
from different sources.
57
63. Pre-II *?
Ind. Skt. masū̄́ra- ‘lentil’
Ir. MiP mycwk/myšwk < PIr. *mižuka- ‘lentil’
EWAia (II, p. 335) gives no IE etymology. Kümmel (2017, p. 284) suggests that the Indic and
Iranian forms are Post-PII borrowings from the same source. Indeed, they share some features
that could support this hypothesis (*mVsu-), but differ in the suffix. In Sanskrit, unaccented -ra-
can form denominal adjectives (AiGr. II, 2, p. 849-858). However, there is no indication that
masū̄́ra- was originally a denominal adjective. It was more likely borrowed as a trisyllabic word
with the structure CVCV̄CV. On the other hand, Iranian *mižuka- could be analyzed as
*miž-uka-, as -uka- is a common denominal suffix. This decreases the likelihood that *mižuka-
and masū̄́ra- have the same source. Both words were likely borrowed Post-PII.
64. Pre-II *mVša- ‘bean’

Ind. Skt. mā̄́ṣa-
Ir. MiP māš, Šu. max̄̆, Sogd. mwškh, Yagh. mušk < *mušakā-
Taken as a Wanderwort by Lubotsky (2001b). EWAia (II, p. 352) concludes that there is no
satisfactory IE etymology. The word is clearly non-IE since there is no regular source for š after
ā. The variant *māša- has a wider distribution than *muša-, occurring in both Indic and Iranian,
as well as in ToB māśak ‘mung bean’ (Adams, 2013, p. 483) and Arabic māš, and may have
spread westwards from India. In that case, MiP māš could also be borrowed from Sanskrit.
However, Ir. *muša(-kā)- is more likely a parallel Post-PII borrowing (Kümmel, 2017, p. 284).
65. Pre-II *nai̯ Ts(a)- ‘skewer’

Ind. Skt. nikṣ- ‘to pierce’, nī̄́kṣaṇa- / nékṣaṇa- ‘skewer, fork’
Ir. LAv. naēza- ‘sharp point of needle’, MiP nēzag ‘lance’, MoP neš ‘skewer’
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 41) considers IE origin likely, but
argues against explaining the root final *-s as a remnant of an old desiderative. Indeed, a
desiderative seems semantically unmotivated. Thus, it seems more likely that the attested
variation is due to parallel Post-PII borrowing. An affricate in the source language may have
been adopted as PIr. *j̄́ [ʣ], but PInd. *tš46 (Skt. > kṣ). MoP neš (< *nai̯ tš-?) looks to be closer
46
Although we might have expected PInd. *ǰ = [ʥ] (< PII * *j̄́ and *ǰ) (Kobayashi, 2017, p. 331).
58
to the Indic form. The long vowel of Skt. nī̄́kṣaṇa- is unexpected and could be an additional
argument for non-IE origin.
66. Pre-II *pərd-a(n)k- ‘leopard, panther’

Ind. Skt. pr̥̄́dāku- ‘snake’
Ir. MoP palang < PIr. *pard-, Sogd. pwrδnk, Pto. pṛāng ‘leopard’
Taken as a loanword by Witzel (2003, p. 35). EWAia (II, p. 163) offers no IE etymology and
doubts the connection between the Indic and Iranian forms on semantic grounds. However, the
semantics of Skt. pr̥̄́dāku- ‘snake’ may very well be secondary, in which case Skt. (JüS) pr̥̄́dāku-
‘tiger, panther’ probably is closer to the original meaning. Furthermore, the structure of Skt.
pr̥̄́dāku- may have been influenced by sr̥dāku- ‘lizard’ (Witzel, 1999a, p. 44), which potentially
hides the original form of the word. The Iranian forms are generally connected to Gr. πάρδαλις
‘panther’ (Beekes, 2010, p. 1152), although they seem to go back to *pərd-ank-. The “suffix”
*-ank- is reminiscent but not identical to the Sanskrit desinence -āku-.
Additionally, Witzel (2003, p. 35) argues that Gr. πάνθηρ ‘panther’ ultimately has the
same origin as Gr. πάρδαλις and the Indo-Iranian words. While alternations in
voicing/aspiration of stops is common in early Greek loanwords, the alternation r/n is not.
However, it is possible that earlier *πάρθηρ >> Gr. πάνθηρ due to folk etymology (πᾱν ‘all’
θηράω ‘to hunt’, cf. Beekes, 2010, p. 1150), or by irregular dissimilation.
67. Pre-II *pinda-

Ind. Skt. píṇḍa- ‘lump’
Ir. Khot. piṇḍaa47
Taken as a loanword by Witzel (2003, p. 33). EWAia (II, p. 128) considers the possibility that
Skt. píṇḍa- is a loanword. In any case, it seems likely that it existed in Iranian, since it was
probably borrowed as Arm. pind ‘firm, dense’.48 Given the distribution of the word and the
unexplained retroflex in Sanskrit, it is possible that píṇḍa- was borrowed into Sanskrit and then
spread to Armenian via Khotanese, or that Indic and Iranian borrowed the word independently.
47
Bailey (1979) does not include Khot. piṇḍaa in his dictionary.
48
Martirosyan (2010, p. 552) considers Arm. pind to be IE, derived from PIE *bhendh- ‘to bind’. However, since
the expected outcome of this root would have been **bind, the etymology requires the assumption of
Grassmann’s Law in Armenian, for which there are no further examples.
59
68. Pre-II *?
Ind. Skt. śāli- ‘unhusked rice’
Ir. Pto. šole, Orm. šōl, Par. šēl, YM šālē
Nur. Km. šāl’i-
EWAia (II, p. 632) presents no IE etymology. Due to the initial š-, Kümmel (2017, p. 283)
assumes that the Iranian forms are borrowed from Nuristani or Sanskrit. Also, since l is
secondary in Iranian languages, this seems likely. Neither Nur. š- corresponds regularly to Skt.
ś-. The word was likely borrowed into Indic in Post-PII times and later diffused to other Indo-
Iranian languages.
69. Pre-II *?
Ind. Skt. śaṇá- ‘hemp’
Ir. MiP šan, MoP kanab, Khot. kaṃha ‘hemp’
Taken as a loanword by Witzel (2003, p. 34) and EWAia (II, p. 605). This old word for hemp
is likely connected to Gr. κάνναβις ‘hemp’ (which was borrowed into many European
languages), Sumerian kunibu ‘hemp’ (Beekes, 2010, p. 636) and Akk. qunnabu ‘an aromatic’
(CAD Q, p. 306). According to Bailey (1979, p. 52), Khot. kaṃha goes back to *kanfa- <
*kanaba-, which would bring it very close to the Gr. and Near Eastern words. Somewhat
confusing is the relationship between the k-initial words and Skt. śaṇa-, which from an IE
perspective looks like a centum-satəm distribution. However, given the retroflex -ṇ-, śaṇa-
looks like a more recent borrowing. Moreover, an alternation of k/ś is found elsewhere in
Sanskrit loanwords (Witzel, 1999a, p. 34). In light of this, the words for ‘hemp’ most likely
reflect Post-PII loanwords from different sources.
70. Pre-II *?
Ind. Skt. sarṣapa- ‘mustard’
Ir. Khot. śśaśv(a)-āna-, Sogd. šywšp-δn, MiP span-dān
Taken as a Wanderwort by Lubotsky (2001b) and EWAia (II, p. 712). Gr. σίνᾱπι ‘mustard’
probably belongs here as well. The Iranian words reflect a compound with dāna- ‘seed’, but the
first part of the compound presents irregular correspondences. Khotanese and Sogdian show
disyllabic words whereas MiP span- is monosyllabic. Khot. śś- and MiP sp- could be taken as
regular reflexes of *ću̯-, but given its position in the word MiP sp- seems to correspond to Khot
60
-śv-, which is irregular. The extra initial syllable of Khot. śśa- and Sogd. šy-, which Persian
lacks, is comparable to Skt. sa- and Gr. σί-. However, Skt. sarṣapa- is quite dissimilar from the
Iranian forms. This suggests that the Indic and Iranian words were borrowed from different
source languages.
71. Pre-II *?
Ind. Skt. siṁhá- ‘lion’
Ir. Parth. šarg, Khot. sarau ‘lion’
Taken as a loanword by Witzel (2003, p. 45). Perhaps belonging here is Arm. inc/j ‘lion’.
Semantically, it is not surprising that IE languages from the Eurasian steppe would borrow a
word for ‘lion’.
The words show variation between *-r- and *-n-. Moreover, the anlaut correspondence
Skt. si- : Khot. sa- is irregular. Parth. šarg points to *či̯ - or *kš- (Kümmel, 2019). The -h- in
Skt. siṁhá- reflects a primary or secondary palatal *j̄́h/ǰh, whereas Iranian points to *g(h). Due
to the variation, the words were clearly borrowed Post-PII.
72. Pre-II *šu̯ai̯ pa- ‘tail’

Ind. Skt. śépa-, Pkt. cheppā- ‘tail, penis’
Ir. LAv. xšuuaēpā- ‘tail’
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 654) gives no IE etymology, but
rejects non-IE origin. Moreover, Pkt. cheppā- is explained as a generalized sandhi variant
(°c chépam) of Skt. śépa-. However, since the anlaut of Pkt. cheppā- (*chepyā-) corresponds
to LAv. xšuuaēpā-, it need not be secondary. The correspondence is paralleled by Pkt. cha ‘six’
~ Av. xšuuaš ‘six’ < *šu̯aćs (Lubotsky, 2000). Skt. śépa- could reflect earlier *śvepa- with
dissimilation of -v- before a labial consonant, cf. Skt. śiti-pád- ‘with white feet’ < *śviti- <
*ḱu̯iti-.
The origin of the anlaut cluster is complicated. Lubotsky (2000, p. 260) tentatively
suggests PII *pću̯ai̯ pa- (< PIE *pḱu- ‘cattle’ + *u̯ei̯ p- ‘to swing’?), but this is unlikely since
Avestan preserves initial *pć- as fš-, cf. Av. fšūmant- ‘having cattle’ < *pću-mant-.
Furthermore, *pću̯ai̯ pa- could hardly yield Skt. *śvepa- in view of kṣumánt- ‘having cattle’.
It appears that two variants must be reconstructed: *šu̯ai̯ pa-, continued by Iranian and
Prakrit, and *ću̯ai̯ pa- continued by Sanskrit. The anlaut of *šu̯ai̯ pa- cannot be secondary from
earlier *su̯-, because unlike *šu̯aćs ‘six’, where original *s >> *š by assimilation to *ćs (= [tš]),
61
*šu̯ai̯ pa- does not provide any conditioning for a similar assimilation.49 Rather, *šu̯ai̯ pa- was
borrowed with anlaut *š-. In Sanskrit, the source word was adapted as *śvepa-, indicating that
*ć had probably become a fricative in Indic at this point. The irregular correspondences indicate
that the word was borrowed Post-PII.
73. Pre-II *(t)sūkV̄- ‘needle’ >> Indo-Iranian *ćūkā- / *sūčī- / *ćuči / *ćaučani̯ a-
Ind. Skt. sūcī̄́- ‘needle’
Ir. LAv. sūkā-, Wakh. sic ‘needle’, MiP sozan, Oss. sūʒīn/soʒīnæ, Khot. suṃjsañu
‘needle’
Taken as a loanword by Lubotsky (2001b), due to the irregular correspondence Ind. *s : Ir. ć-.
EWAia (II, p. 739) instead assumes that earlier Skt. *śūcī̄́- >> sūcī̄́- by analogy to sīv- ‘to sew’.
As this is rather ad hoc, one might entertain Lubotsky’s (2001b, p. 306) hypothesis that Indic
borrowed the predecessor of Skt. sūcī̄́- from an unknown source, which was then borrowed into
Iranian. However, since PII *ć remained an affricate in PIr., Skt. sūcī̄́- is an unlikely source of
PIr. *ćūkā- etc. Moreover, the fact that Sanskrit has -c-, never -k-, makes it improbable that
LAv. sūkā- was borrowed from Indic.
While Wakh. sic < *ćuči resembles Skt. sūcī̄́-, the MiP, Ossetic and Khotanese words
reflect the rather divergent form *saučani̯ a- (Bailey, 1979, p. 427). This much variation points
to a Post-PII borrowing.
However, as the palatalization of velars was a PII development, this requires the
assumption of two source forms, one with palatalized *-c- and one with *-k-. Yet, the alternation
of *-c- and *-k- seems to correlate with the quality of the following vowel (-cī- vs. -kā-, except
in *ćaučani̯ a-), indicating regular palatalization of *kī > *čī. This correspondence gives the
impression of an old borrowing, whereas the irregular correspondence Indic s- : Iranian s- points
to the opposite. This paradox has no easy solution, but one may speculate that the variation č/k
arose as the loanwords for needle were adapted to the native linguistic structure, where velars
were normally palatal before *ī.
49
Cf. Skt. svápna- ~ Av. xvafna- < PII *su̯apna-.
62
74. Pre-II *u̯īna- ‘lute’
Ind. Skt. vī̄́ṇā-
Ir. Khot. bīna, Sogd. wyn’, MiP win
Taken as a loanword into PII by Witzel (2003, p. 33). EWAia (II, p. 568) gives no etymology
but states that the word may have diffused amongst the Indo-Iranian languages, although the
direction of borrowing is not clear. The irregular retroflex ṇ in Sanskrit indicates that the word
was not inherited from PII. It may have been borrowed into Sanskrit and then spread to Iranian
and eventually also to Arm. vin, or into Indic and Iranian independently.
3.5. Words with IE etymologies or insufficient evidence for borrowing
75. PII *āćā- / *aćas-

Ind. Skt. ā̄́śā- f. ‘space’
Ir. LAv. asah- n. ‘region’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 178) offers no IE etymology. The
irregular correspondence Indic ā : Iranian ă could be taken as an argument for non-IE origin.
However, morphologically, the alternation between Skt. ā̄́śā- (< *Hóḱ-eh2-?) and LAv. asah-
(< *Heḱ-os?) looks old.
76. PII *ćan- ‘to ascend’

Ind. Skt. adv. śanaiḥ ‘gradually’
Ir. LAv. san-, Khot. san- / sata- ‘to rise’
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 608) and LIV (p. 324) assume
an IE verbal root *ḱen- also reflected in Arm. snanim ‘to be raised, grow’. Since both Iranian
and Armenian have nu-presents, IE origin is likely.
77. PII *ću̯itra- ‘white’

Ind. Skt. sī̄́sa- (AV) ‘lead’
Ir. SW-Iranian *siça- ‘white’
EWAia (II, p. 734) argues that Skt. sī̄́sa- is borrowed from an Iranian word *siça- ‘white’, which
cognate to Skt. śvitrá- ‘white’ < PIE *ḱu̯it-. Against this, Witzel (2003, p. 33) argues that a
borrowing from SW-Iranian, does not fit with the fact that Skt. sī̄́sa- is attested already in the
63
AV. Yet, this counterargument seems to be built on a misunderstanding, as Witzel (ibid.) notes
that “[t]he Persians moved into the Persis and Anšan from NW Iran only after c. 700 BCE”.
However, the (pre-)historical location of the Persian speech community is not crucial for the
hypothesis that Skt. sī̄́sa- was borrowed from a SW-Iranian dialect: the only relevant
assumption is that the sound change PIr. *ću̯ > *s and *θr > *ç had already occurred in this
language. Borrowing of an originally IE word from SW-Iranian thus remains a possible
explanation of Skt. sī̄́sa-. In any case, Skt. sī̄́sa- must be a late borrowing, as it does not show
the effect of the RUKI-rule.
78. PII *daćā-

Ind. Skt. daśā- ‘hem’
Ir. Khot. dasa-, Bal. dasag- ‘thread’
Taken as a loanword by Lubotsky (2001b). EWAia (I, p. 710) mentions possible connections to
PGm. *tagla- ‘hair’ and OIr. dual ‘tuft, plait’ < PIE *doḱ-lo-, to which Matasović (2009, p.
102) adduces SCr. dlàka ‘single hair’ < PSl. *daḱlā- (with depalatalization). Kroonen (2013, p.
504) points out that *tagla- could be an inner-Germanic formation from a different root,
however. PII *daćā- (< PIE *deḱ-eh2-) is derived differently from the proposed cognates. If the
original meaning was ‘hair’, an *-eh2-derivation could yield ‘single hair’ >> ‘thread’. It is
possible that the word is inherited.
79. PII *dhu̯aǰ- ‘to flutter’

Ind. Skt. dhvajá- ‘banner’
Ir. LAv. dβōža- ‘to flutter’, Sogd. wy-δβys- ‘to bloom’
Suggested to be a loanword by Lubotsky (2001b). EWAia (I, p. 800) derives it from the root PII
*dheu̯H- ‘hin und herbewegen’ with a suffix *-eg-. However, *dhu̯aǰ- does not reflect a
laryngeal, and the supposed suffix does not look IE. Since the original distribution of palatalized
and plain velars in verbs is often distorted, the existence of *ǰ before non-palatalizing vowels
(e.g. Skt. dhvajá-) does not prove non-IE origin.
64
80. PII *ghas- ‘to devour’
Ind. Skt. ghas-
Ir. LAv. gah-
Suggested to be a loanword by Lubotsky (2001b). No IE comparanda given by EWAia (I, p.

514). However, it may be connected to PGm. *gamman- ‘stall, hut’ and Arm. gom ‘fold (for
cattle)’ < *ghos-mo- (Kroonen, 2013, p. 166), although the semantic connection is weak. IE
origin cannot be excluded.
81. PII *ghau̯s- ‘to make sound, hear’

Ind. Skt. ghoṣ-
Ir. Av. gaoš-
Suggested to be a loanword by Lubotsky (2001b). EWAia (I, p. 518) mentions a possible

connection to PGm. gauma- ‘heed, attention’ (< *ghou-mo-), in which case the PII verb would
be an original s-present/desiderative of the same root. In this scenario, the PIE s-stem would
originally have meant ‘to wish to be heard’, contrasting with the s-less stem ‘to be heard’.50
Kroonen (2013, p. 171) proposes a different etymology for the Germanic words, which he
connects to Skt. gū̄́hati ‘to hide’. This is formally possible, but semantically less attractive than
the above. In any case, there is not enough evidence to postulate substrate origin.
82. PII *Hat-

Ind. Skt. at- ‘to wander’
Ir. LAv. xvāθra- ‘well-being’ (< *su̯-at-ra), a-pairi.āθra- ‘unavoidable’
Suggested to be a loanword by Lubotsky (2001b). EWAia (I, p. 56) suggests a connection to

Lat. annus ‘year’ (< *atno- ‘which goes’), reconstructing PIE *h2et- ‘to wander’ (cf. LIV, p.
273). Although this may not be correct, there is no further indication that PII *Hat- is a
loanword.
50
A similar situation would be reflected in Skt. śru- ‘to hear’ but śroṣa- ‘to be obedient’ (lit. ‘to wish to hear’)
65
83. PII *Hu̯ap- ‘to strew, scatter’
Ind. Skt. vap-
Ir. OAv. (vī-)uuāpat̰
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 503) gives no convincing IE

cognates. Kloekhorst (2008, pp. 430-1) argues that Hitt. ḫuu̯app-i / ḫupp- ‘to throw, be hostile
towards’ is related to the Indo-Iranian forms and reconstructs PIE *h2uóph1-ei, with a final
laryngeal to explain the Hittite geminate -pp- (since -p- would otherwise have been lenited).
The initial laryngeal reflex in Hittite fits with the lengthened vī- in OAv. A root final laryngeal,
however, is not compatible with the Indo-Iranian evidence, since that should have given Skt.
*vaph-, Av. *vaf-. A solution is to reconstruct PIE *h2uep-, and assume that the geminate -pp-
in Hittite was levelled from the 3pl.pret. ḫuppēr, where -p- would have escaped lenition since
it was preceded by a short vowel. In conclusion, this verb is likely of IE origin.
84. PII *Hu̯ap- ‘to shave, shear’

Ind. Skt. vap-
Ir. Khot. patävutta-
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 504) assumes that the verb is
etymologically identical to *Hu̯ap- ‘to strew, scatter’. As a semantic development from ‘to
scatter (hair) > ‘to shear’ seems plausible, the verb may be treated as IE.
85. PII *Hu̯i̯ dhH- ‘to split in two’

Ind. Skt. vyadh- ‘to wound, hurt’
Ir. LAv. vīδ- ‘to pierce’
Suggested as a possible loanword by Lubotsky (2001b). Conversely, EWAia (II, p. 591) and
LIV (p. 294) reconstruct PIE verbal root *h2u̯i̯ edh- based on Indo-Iranian. An initial *h2- is
postulated based on Gr. ἠίθεος / ᾄθεος ‘unmarried youth’ (Tichy, 1993, p. 15), but since ᾄθεος
is likely a hyperdorism, Greek could also reflect *h1- (Beekes, 1992, p. 172). Lubotsky (1994,
p. 204) explains Gr. ἠίθεος and Skt. vidhávā- ’widow’ from PIE *du̯i-dhh1-u- ‘widow(er)’ (lit.
‘bereft of its half’, with Kortlandt effect *d > *h1), and connects this compound to Skt. vidh-
‘to allot, apportion’.
66
In my opinion, Skt. vyadh- belongs to the same etymon51 as a full grade of the secondary
root PII *Hu̯idhH- ‘to split in two’. In PII, this root underwent semantic change from ‘to split
in two’ >> ‘to divide, allot’ on the one hand, and >> ‘to pierce’ on the other, and finally, within
Sanskrit >> ‘to wound’. This analysis is supported by the fact that a) Skt. vidh- ‘to allot,
apportion’ never takes the full grade (EWAia II, p. 555) and b) the Iranian evidence. In LAv.,
we find a root vīδ- with the present stem viθiia- ‘to pierce’ (Kellens, 1995, p. 55). The same
root likely underlies vaēδa- ‘Wurfgeschoss, Name einer bestimmten Angriffswaffe’ (AiWB, p.
1320) and a-šəmnō-vīd- ‘das Ziel nicht erreichend, verfehlend’ (AiWB, p. 257). It is also found
in MiP wistan ‘to shoot, throw’, Pto. wīštəl ‘to shoot, hit’, Šu. wēδ-d ‘to throw’ (EWAia II, p.
592). The Iranian forms show a different full grade (~ *Hu̯ai̯ dhH-) than Indic (~ *Hu̯i̯ adhH-),
indicating the secondary nature of this formation. In sum, I regard this verb as IE in origin.
86. PII *i̯ ātu- ‘black magic’

Ind. Skt. yātú-
Ir. LAv. yātu-
Taken as a loanword by Witzel (2003, p. 38). EWAia (II, p. 411) suggests connections to either
Skt. 1yā- ‘to travel’, 2yā- ‘to request’ or 3yā- ‘to attack’, the first two of which have IE cognates.
Lubotsky (1988, p. 47) holds 3yā- as the most likely base of *i̯ ātu- on semantic grounds. Perhaps
3
yā- is etymologically identical to 2yā-, if a semantic development ‘to pursue’ >> ‘to attack’ is
assumed. In any case, IE origin seems likely.
87. PII *j̄́hai- ‘to incite’

Ind. Skt. hi-
Ir. LAv. frazaiiaiiāmi
Suggested to be loanword by Lubotsky (2001b). EWAia (II, p. 802) suggests connection to

PGm. *gaiza- ‘spear, tip’ and OIr. gae ‘spear’, but these reflect a root *ǵheis- (Kroonen, 2013,
p. 164), which might be comparable to Skt. héṣas- ‘weapon’ and heṣ- ‘to damage’, but not to
Skt. hi-. Even if isolated, the nu-present derivation in Sanskrit suggests IE origin.
51
Cf., less explicitly, Melchert (1977, p. 113).
67
88. PII *ǰhas- ‘to laugh’
Ind. Skt. has-
Ir. LAv. jahī, jahikā- ‘prostitute’
Suggested to be a loanword by Lubotsky (2001b). No IE comparanda given by EWAia (II, p.

811). Given the formal similarity to PII *ghas- ‘to devour’, one might consider an etymological
connection between the two. Starting from *ghes- with an original meaning ‘to open the jaws’,
it is possible that the palatalized variant *ǰhas- < *ghes- was lexicalized as ‘to laugh’, whereas
the non-palatalized *ghas- < *gh(o)s- was lexicalized as ‘to devour’. Whether they are related
or not, there is not enough evidence to postulate substrate origin.
89. PII *kuč- ‘to crook, bend’

Ind. Skt. kuc-
Ir. MiP n-gwč-, Khot. us-kuj- ‘to rise up’
Suggested to be a loanword by Lubotsky (2001b). EWAia (I, p. 361) regards the proposed
connection to OIr. cúar ‘bent’ uncertain. However, Matasović (2009, p. 228) has shown that
MIr. cúar ‘curved’ may derive from *kukro- or *koukro- (with regular loss of *k before *r).
This has further cognates in BSL, cf. Lith. kaũkas ‘lump’ and PSl. *kùka-52 ‘hook’ (Derksen,
2008, p. 256). In view of the formal and semantic correspondences, it is in my opinion likely
that PII *kuč- is IE.
90. PII *magha- ‘gift, offering, sacrifice’

Ind. Skt. maghá-
Ir. Av. maga-
Taken as a loanword by Lubotsky (2001b) due to its connection to ritual practices. EWAia (II,
p. 289) assumes IE origin in connection with Goth. magan ‘to be able’ and OCS mogǫ ‘id.’.
While the semantic match is not perfect, it is not farfetched enough to exclude IE origin.
52
The Slavic acute is analogical to *kl̨ȕka- (Derksen, 2008, p. 256).
68
91. PII *marj̄́ha- ‘udder’
Ind. Skt. malhá- ‘with hanging belly/udder’
Ir. LAv. mərəzāna- ‘belly’, gen.sg. maršuiiā̄̊- ‘paunch’
According to Lubotsky (2001b, p. 312), the proposed connection to Lith. mìlžtis ‘to swell up’
(EWAia II, p. 334) is impossible since the Baltic acute reflects PIE *ǵ, whereas PII *j̄́h must go
back to *ǵh. However, since PII *j̄́h could also go back to PIE *ǵH with PII deglottalization, the
etymology is possible if PIE *(h2)melǵH- ‘to swell’ is reconstructed.
LAv. mərəzāna- seems to reflect *ml̥ ǵH-ono-. The gen.sg. maršuiiā̄̊- looks related, but š
< *j̄́ only regularly occurs before *n.
Lubotsky (2001b, p. 312) furthermore suggests a connection to Skt. bárjaha- ‘udder’,
also entertained in EWAia (II, p. 211, 334). Skt. bárjaha- cannot reflect the same PII form as
Skt. malhá-, since m > b only occurs before consonantal r (cf. Skt. bravīti < *mreu̯H-ti). Neither
is there an analogical model for initial b-. The only way to connect the words would be to
assume parallel borrowings *malj̄́h- ~ *barj̄́(h)- from a non-IE source.
92. PII *monH-i-

Ind. Skt. maṇí- ‘necklace’
Ir. LAv. (zarənu-)maini- ‘with golden neck-jewel’
Taken as a loanword by Witzel (2003, p. 33). Due to the irregular retroflex in Sanskrit, the word
cannot readily be reconstructed for PII. EWAia (II, p. 293), however, argues for IE origin in
*monh2-i-, further reflected in PGm. *manja- ‘necklace’ and derived from *mon-eh2- ‘neck’,
reflected in PGm. *manō-. Note that the Indo-Iranian forms could derive from *mn̥h2-i- or
*monh2-i-. Although the retroflex in Sanskrit remains unexplained, IE origin cannot be
excluded.
93. PII *nard- ‘to hum, complain’

Ind. Skt. nr̥d-
Ir. Sogd. nrδ-, MiP nāl
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 22) proposes no IE etymology

but suggests that the word could be onomatopoeic. This is difficult to disprove, but neither a
particularly compelling hypothesis.
69
94. PII *raj̄́h-
Ind. Skt. rah- ‘to be abandoned’
Ir. MiP rāz ‘mystery’
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 442) concludes that there is no
IE etymology.
95. PII *sagh-

Ind. Skt. sagh- ‘to be able to bear’
Ir. LAv. azgatō ‘unbearable’
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 686) assumes origin in PIE
*segwh- on the basis of Gr. σθένος ‘strength’. However, due to the unclear suffix *-eno-, σθένος
may not be IE (Beekes, 2010, p. 1326).
Alternatively, *sagh- could derive from the same PIE root as Skt. sáhate ‘conquers’,
sáhas- ‘victory’, Av. hazah- ‘victory’ < PIE *seǵh- ‘to hold’, cf. Gr. ἔχω ‘to have, hold’. In the
zero-grade *sǵh-, *s would depalatalize *ǵh > *gh. From the zero-grade, a secondary root *segh-
was formed, continued in PII *sagh-. The lexical split would only have occurred in satəm
languages, where *ǵh and *gh were distinct. This analysis allows LAv. azgatō to be connected
to Gr. ἄσχετος ‘irresistible’ (AiWB, p. 228).53
96. PII *srans-

Ind. Skt. sraṃs- ‘to fall apart’
Ir. OAv. rā̄̊ŋhaiiən ‘they make fall away’
Suggested to be loanword by Lubotsky (2001b). EWAia (II, p. 783) offers no IE etymology.
97. PII *stHūna- ‘pillar’

Ind. Skt. sthū̄́ṇā-, sthū̄́nā-
Ir. LAv. stunā-, OP stūnā-
Taken as a loanword by Witzel (2003). The irregular Sanskrit retroflex -ṇ- is unexplained, but
may be secondary given the non-retroflex variant form. EWAia (II, p. 768), argues that the word
53
However, since *n-sgh-eto- should have yielded *n-zǰh-eto- > *a-zj̄́hata-, the velar of LAv. azgatō must be
secondary.
70
is related to Skt. sthūrá- ‘big, strong’ and Gr. στῦλος ‘column, pillar’, all from PIE *sth2u- (itself
from *steh2- ‘to stand’ with a u-extension). The long ū in Skt. sthūrá- and sthū̄́ṇā- presupposes
a proto-form *stuH-, presumably from metathesis of *stHu-, in which case the aspirate -th-
must be analogical from some other form of the same root, e.g. *stHeu-. IE origin is possible.
98. PII *su̯ag- ‘to embrace’

Ind. Skt. svaj-
Ir. LAv. pairiš.xvaxta
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 788) and LIV (p. 610)
reconstruct PIE *su̯eng- (*n reflected in Skt. pári-ṣvañjalya-), a root also reflected in MHG
swanc ‘movable’ and OIr. seng ‘skinny’.
99. PII *u̯i̯ ak- ‘to encompass’

Ind. Skt. vyac-
Ir. MoP gunǰidan
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 590) argues for an IE origin in
relation to Lat. vinciō ‘to bind’. LIV (p. 696) reconstructs *u̯i̯ ekw- ‘to encompass’ and adds Gr.
(Thess.) ἴμψας ‘having yoked’ as evidence for the labiovelar. However, there is reason to doubt
that ἴμψας belongs here. Firstly, the -s- in the Greek word is also found in ἴμψιος ‘(of the) yoke’
(Beekes, 2010, p. 591), implying that it is part of the root. Secondly, in this scenario, the -m-
would be a relic of the nasal present, which should be absent in the aorist. Furthermore, Lat.
vinciō ‘to bind’ could very well be related to Lat. vincō ‘to conquer’ (Vaan, 2008, p. 679), which
does not reflect a labiovelar but goes back to PIE *u̯ei̯ k- (LIV, p. 670). PII *u̯i̯ ak-, cannot derive
from PIE *u̯ei̯ k- since the root structures differ.
Despite the lack of convincing cognates outside of Indo-Iranian, the fact that PII *u̯i̯ ak-
formed a nasal present indicates IE origin.
100. PII *u̯i̯ atH-

Ind. Skt. vyath- ‘to be unsteady’
Ir. OAv. a-iβiθura- ‘unshakeable’
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 591) takes it as a secondary

root from *ui- + -eth2. This analysis is problematic since there is no semantically fitting root
71
*teH- to explain the second part of the root (cf. LIV, p. 616: *teh2- ‘to steal’, *teh2- ‘to thaw’.
However, as PII *u̯i̯ atH- may be cognate to PGm. *witt/dōn- ‘to tremble’, it could be IE.
101. PII *u̯ik- ‘to separate’

Ind. Skt. vic-
Ir. LAv. vic-, MiP wēxtan/wēz-
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 577) proposes an etymological

connection to Hitt. ḫuek-zi ‘to slaughter’, which is formally impossible (Kloekhorst, 2008, p.
407). However, other possible cognates are Lat. victima ‘sacrificial animal’, lit. ‘the separated
one’ (Vaan, 2008, p. 675), and PGm. wīha- ‘holy’. For this reason, IE origin cannot be excluded.
102. PII *u̯r̥ćsa- ‘tree’

Ind. Skt. vr̥kṣá-
Ir. LAv. varəša-
Taken as a loanword by Lubotsky (2001b). EWAia (II, p. 572) hesitantly supports a connection
to Skt. válśa- ‘twig’, LAv. varəsa- ‘hair’, OCS vlasъ ‘hair’ < PIE *uolḱ-o-. This explanation is
possible under the assumption that PII *u̯r̥ćsa- derives from an s-stem *uelḱ-es- ‘twig’, from
which a thematic possessive derivative *ulḱ-s-ó- ‘having twigs’ >> ‘tree’ was formed. While
this scenario requires the assumption of an unattested s-stem, the connection between Skt.
vr̥kṣá- ‘tree’ and válśa- ‘twig’ is semantically likely, and the postulated derivational chain
explains their semantic relationship in a morphologically convincing way. 54 As such, PII
*u̯r̥ćsa- ‘tree’ could be of IE origin.
103. PII *u̯riH- ‘to oppress, collapse’

Ind. Skt. vlī-
Ir. LAv. uruuīnaitīš (acc.pl.)
Suggested to be a loanword by Lubotsky (2001b). EWAia (II, p. 598) derives it from an

extended root *u̯R-eiH- for unclear reasons. Although the verb is isolated to Indo-Iranian,
archaic-looking derivations like Skt. vlināti < *uli-ne-H-ti means that IE origin cannot be
excluded.
54
Cf. Skt. vatsá- ‘yearling’ < *uet-s-ó-.
72
4. Chronological layers in early Indo-Iranian loanwords
In chapter 3, early Indo-Iranian loanwords were classified into chronological layers. In this
chapter, each layer is discussed as a whole. The goal is to summarize the findings from chapter
3 and to discuss potential implications of the chronological layers for the Central Asian
Substrate Hypothesis.
4.1. Layer I: Pre-PII or early PII

Layer I consists of loanwords borrowed into Pre-PII or into early PII before the operation of
certain PII sound changes. A single word (*ućig-) was argued to belong to this layer.
The descendants of PII *ućig- ‘sacrificing priest’ show a regular alternation between velar
-k- / -g- and palatalized *-ǰ- in the paradigm. The most likely reason for this alternation is that
*ućig- underwent palatalization before the PII vowel merger. However, it cannot be excluded
that it was secondarily adapted by analogy to inherited stems with a similar alternation, as these
were common in the morphological structure of early Indo-Iranian.
Thus, the evidence for Pre-PII borrowings is scarce or non-existent. However, it must be
kept in mind that any loanword in layer 0 may have been borrowed at the Pre-PII or early PII
stage as well. Accordingly, the scarcity of words in Layer 1 does not in itself prove that Indo-
Iranian borrowed less during the Pre-PII or early PII period.
4.2. Layer 0: PII (unspecified)

Layer 0 refers to 45 loanwords that are reconstructable for PII, but for which there is no further
indication of time of borrowing. Theoretically, these words may have been borrowed at any
time after the disintegration of PIE up until the split of PII. However, certain words in layer 0
have features that suggest a more precise classification.
Four words55 in layer 0 contain voiceless aspirates (*ph, *th, *kh). I have chosen to
consistently analyze these as clusters of voiceless stops + laryngeal (*pH, *tH, *kH), since
phonemic voiceless aspirates are an Indic innovation. However, given that *H was phonetically
a glottal stop,56 it seems likely that *PH-clusters in loanwords represent adaptations of
monophonemic stops or fricatives in the source language(s), rather than clusters. Stops in
tautosyllabic clusters with glottals automatically adopt the laryngeal feature of the glottal, since
this becomes a feature of the cluster as a whole (Kehrein, 2002). Since *H was in the process
of being lost as a segmental phoneme in late PII (having been vocalized, lost intervocalically
55
*atHarvan-, *kapHa-, *kHā-, *kHara-.
56
Or, according to Kümmel (2018), a glottal fricative [h].
73
etc.), the sequence *PH was probably phonetically equivalent to *Ph. Therefore, it is likely, but
not provable, that loanwords in layer 0 with voiceless aspirates were borrowed in late PII.
Another four words57 in layer 0 contain long vocalic semivowels *ī and *ū. In inherited
words, these long vowels arose regularly from the sequences *iH and *uH. This development
is shared by Indic and Iranian but not Nuristani (cf. chapter 2). Although it is impossible to
prove, it seems unlikely that loanwords containing *ī or *ū were borrowed as *iH and *uH.
These sequences are connected to the ablauting morphophonology characteristic of Indo-
European, and none of these loanwords show ablaut. It is more likely that loanwords containing
*ī or *ū were borrowed at a time when *iH and *uH had already become long vowels, i.e. late
PII. However, this should be regarded as a tentative conclusion.
Lastly, the semantics of some words provide additional indications of their time of
borrowing. PII *Hustra- ‘camel’ and *kHara- ‘donkey’ both denote animals that are absent
from the Indo-European homeland and rather associated with Central Asian cultures (Witzel,
2000, p. 4). PII *parsa- ‘sheaf’, *spāra- ‘ploughshare’ and *u̯rīj̄́hi- ‘rice’ all relate to
agriculture, which was not practiced in the Sintashta culture (Judd et al., 2018) or on the
Eurasian steppe until after 2000 BCE (Anthony, 2007). Therefore, it is unlikely that any of these
words were borrowed into Pre-PII, before Indo-Iranian speakers had migrated to Central Asia.
Thus, out of 45 loanwords in layer 0, 13 show indications of being borrowed towards the
end of Indo-Iranian linguistic unity.
4.3. Layer II: late PII

Layer II consists of 7 loanwords borrowed into late PII, i.e. after certain PII sound changes had
already occurred. Below, the rationale for including words in layer II based on the relative
chronology of PII sound changes will be discussed.
PII *pusća- ‘tail’ and *ǰaǰha/ukā̄̆- ‘hedgehog’ are included in layer II, since they contain
palatalized velars before non-palatalizing vowels. Therefore, they are unlikely to have been
borrowed before the PII palatalization of velars. In the case of *ǰaǰha/ukā̄̆-, the morphological
variation among the attested Indo-Iranian forms also suggests a late time of borrowing. The
non-palatalized *ka- in PII *karus- also indicates borrowing after the palatalization of velars,
but could also be because it was borrowed after the phonologization of *a. In any case, it cannot
be Pre-PII.
57
*bīj̄́a-, *kšīra-, *u̯āćī-, *u̯rīj̄́hi-. In the case of *u̯āćī-, the long *ī is synchronically a suffix, and could
theoretically be secondary due to adaptation to the native morphology.
74
PII *čāt(u̯āla)- ‘pit, well’ is included in layer II due to the morphological variation in the
attested forms, which suggests that the Indo-Iranian speaker communities were disintegrating
at the time of borrowing. It is thus likely that the word was borrowed after the palatalization of
velars, i.e. initial *č- was borrowed as such and not palatalized within PII.
The inclusion of *mai̯ ūkHa- ‘peg’, *pīi̯ ūša- ‘beestings’ and *i̯ au̯īi̯ ā- ‘canal’ in layer II is
based on indirect evidence. The long vowels *ī and *ū go back to *iH and *uH. Thus, if e.g.
*pīi̯ ūša- goes back to Pre-PII, it must be reconstructed as *piHi̯ uHsa-. There is no direct
counterevidence against reconstructing *piHi̯ uHsa- etc., but the fact that the attested words
seem to belong to the CVCV̄CV type makes it highly likely that they were borrowed with a
medial long vowel. Since the sound change *i/uH > *ī/ū / _C is the latest change shared by
Indic and Iranian, it is likely that *mai̯ ūkHa-, *pīi̯ ūša- and *i̯ au̯īi̯ ā- were borrowed quite late.
Strictly speaking, this is circular reasoning, since the analysis of *pīi̯ ūša- as a CVCV̄CV word
and the rejection of the reconstruction *piHi̯ uHsa- are logically co-dependent. However, the
CVCV̄CV type is so pervasive in Indo-Iranian loanwords that I consider it very likely that
loanwords attested with this structure belong to the same group.
4.4. Layer III: Post-PII borrowings

Layer III consists of 21 loanwords that are attested in both Indic and Iranian, but cannot be
reconstructed for PII due to irregular sound correspondences that cannot be explained by
secondary processes. However, the formal and semantic similarity of 19 of these loanwords
puts it beyond reasonable doubt that they share a common origin.
In chapter 3, I argued that these 19 loanwords have been borrowed along four different
paths: 1) source language >> Indic and Iranian, independently, 2) source language >> Iranian
>> Indic, 3) source language >> Indic >> Iranian, and 4) source language A >> Indic, source
language B >> Iranian.
Group 1 (parallel borrowings) consists of 13 words: *banγa- ‘hemp, narcotic’, *dū̄̆rća-
‘(goat’s) wool, hair’, *ganDaru̯a- ‘a mythical being’, *ganTi- ‘smell’, *ganTuma- ‘wheat’,
*Kai̯ ća- ‘hair’, *mVša- ‘bean’, *nai̯ Ts(a)- ‘skewer’, *pinda- ‘lump’, *pərd-a(n)k- ‘panther’,
*šu̯ai̯ pa- ‘tail’, *(t)sūkV̄- ‘needle’, and *u̯īna- ‘lute’. For justification of each case, I refer to
chapter 3.
As an example of group 1, consider Skt. kéśa- ‘head hair’ (*kai̯ ća-) next to LAv. gaēsa-
‘curly hair’ (< *gai̯ ća-). The irregular initial stop correspondence can hardly be explained by
assuming that Sanskrit or Avestan borrowed the word from the other: Indic *kai̯ ća- would most
likely have been adopted as Iranian **kai̯ ća- since voiceless /k/ exists in the Iranian phoneme
75
inventory. Similarly, Iranian *gai̯ ća- would most likely have been adopted as Indic *gai̯ ća-.
The most plausible scenario is therefore parallel borrowing from a Pre-II source that I write as
*Kai̯ ća-. This “reconstruction” is not necessarily accurate, but merely an approximation of the
source form. Since the Sanskrit and Avestan words are so similar, it is likely that they were
borrowed only shortly after the split of PII, during a period of dialectal differentiation.
A single word constitutes group 2: *ćikatā- ‘sand, gravel’ may have been borrowed into
Iranian and then spread to Indic. However, parallel borrowing is also possible.
Similarly, a single word constitutes group 3: Skt. śāli- ‘unhusked rice’ likely spread from
Indic into Iranian and Nuristani.
Finally, four words constitute group 4: khaḍgá-/*karkadan- ‘rhinoceros’, sarṣapa- (etc.)
‘mustard’, śaṇa-/kanaba- ‘hemp’, and siṁhá-/sarau ‘lion’. All words refer to animals and
plants common in South Asia and the Middle East. They may ultimately derive from a common
source,58 but the Indic and Iranian reflexes are so dissimilar that their immediate source
languages must have been different. This does not exclude the possibility of a Central Asian
Substrate origin, but it implies that at least one of the Indo-Iranian variants was borrowed from
another language.
The two remaining words in layer III are *(H)arj̄́aná-/áṇu- ‘millet’ and masū̄́ra-/*mižuka-
‘lentil’. In these cases, the proposed Indic and Iranian cognates are too phonologically divergent
to plausibly share a common origin. However, the structure of Skt. masū̄́ra- indicates that it
belongs to the group of Indic CVCV̄CV words.
4.5. Implications of chronological analysis

The division of early Indo-Iranian loanwords into chronological layers has consequences for
the analysis of the loanword corpus and the Central Asian Substrate Hypothesis.
As stated above, the small size of layer I does not in itself prove that most early loanwords
entered Indo-Iranian at a later stage, since words from layer 0 could in theory be very old.
However, considering the 7 words of layer II, 21 of layer III, and the fact that 13 words in layer
0 show features that suggest a late time of borrowing, the general trend is clear: most loanwords
were not borrowed into Pre-PII, but into the later stages of PII and Post-PII. This result supports
the Central Asian Substrate Hypothesis, since the time of borrowing coincides with the period
when PII is believed to have been spoken in the Sintastha and Andronovo cultures.
58
Either because the source languages were distantly related or because the source languages, in turn, had
borrowed the word from a common source.
76
Another consequence is that loanwords in different layers cannot a priori be considered
to originate in the same language(s). This is especially true for Post-PII loanwords (layer III)
vs. PII loanwords (layers I, II and 0). Instead, they will (at least initially) be treated as
originating in different languages. Crucially, however, this does not necessarily imply
“different languages” in the phylogenetic sense of mutually unintelligible linguistic
communities. Rather, “different languages” may represent different chronological stages of the
same language, i.e. “Old Pre-II”, “Middle Pre-II” etc.
Loanwords belonging to the same layer may in theory originate in different languages or
be separated by hundreds of years of linguistic development. In other words, a layer may be
more diverse, in terms of absolute chronology, than is discernible by the available historical
linguistic methodology. However, with the current methodology, it is these layers that must be
the basic units of analysis.
The next step in our analysis is to determine whether the two situations (borrowing from
phylogenetically different languages vs. borrowing from different chronological stages of the
same language) can be differentiated based on the data itself. As discussed in chapter 1,
Lubotsky argued, based on structural similarities between PII loanwords and Indic loanwords,
that “a substratum of Indo-Iranian and a substratum of Indo-Aryan represent the same language,
or, at any rate, two dialects of the same language” (2001b, p. 306). In other words, he found
similar structural characteristics in separate chronological layers, which demonstrate a link
between the source languages of both layers. Now, structural characteristics of the layers
proposed in this study can be analyzed in a similar fashion.
77
5. Structural characteristics of Indo-Iranian loanwords
This chapter discusses the phonological and morphological structure of Indo-Iranian loanwords,
including characteristic word structures, recurring sequences of phonemes, and recurring
irregular correspondences. The goal is to show how the structure of loanwords differs from
inherited words and to determine what the structure of loanwords reveals regarding the structure
of the substrate language(s). In accordance with the results of chapter 4, the chronological layers
0-III will be kept apart in the analysis, in order to determine whether different layers represent
borrowings from the same language(s) or different language(s).
5.1. The CVCV̄CV-type
The recurrence of non-IE trisyllabic words with a medial long vowel or diphthong in Sanskrit
was described by Kuiper (1991). The existence of CVCV̄CV words in PII was demonstrated by
Lubotsky (2001b, p. 306), who put forward the hypothesis that the source of these words in PII
and Sanskrit was the same language, or at least related languages. Therefore, the treatment of
CVCV̄CV words is crucial for understanding the Pre-II linguistic landscape of Central and South
Asia.
Table 3. CVCV̄CV words in early Indo-Iranian loanwords

CVCV̄CV
Layer I (Pre-/early PII)
Layer 0 (PII) *kapāra-
*kapau̯ta-
*u̯arāj̄́ha-
Layer II (late PII) *i̯ au̯īi̯ ā-
*mai̯ ūkHa-
*pīi̯ ūša-
*čāt(u̯āla)-
Layer III (Post-PII)
Total: 7
Evidently, four CVCV̄CV words belong to the late PII layer, although *čāt(u̯āla)- is uncertain
since the trisyllabic structure is only attested in Indic. The three remaining CVCV̄CV words
belong to layer 0, since it could not be demonstrated that they have or have not undergone
certain PII sound changes. However, due to their structural similarity, it is probable that all PII
CVCV̄CV words were borrowed from the same language. The fact that other CVCV̄CV words
78
were borrowed in late PII indicates that *kapāra-, *kapau̯ta- and *u̯arāj̄́ha- too were borrowed
towards the end of Indo-Iranian linguistic unity. This would also explain why the initial *k- of
*kapāra- and *kapau̯ta- is not palatalized, without requiring the reconstructions Pre-PII
*kN̥pāra- vs. *kN̥pau̯ta-.
While 6 CVCV̄CV words are securely reconstructable for PII, Kuiper’s (1991, pp. 90-93)
list of non-IE words in Vedic includes 63 CVCV̄CV words.
In the PII group, all CVCV̄CV words are thematic a-stems. Conversely, in the Indic group,
23 of 63 words belong to other stem types, e.g. 9 u-stems.59 Witzel (2003, p. 33) considers all
thematic loanwords as original consonant stems (e.g. *kapau̯t-) that were thematicized within
PII. If this is true, PII CVCV̄CV words, which are all thematic, could derive from disyllabic
words. On the other hand, it is unlikely that the many athematic stems among the Indic
CVCV̄CV words are secondary. Therefore, under Witzel’s analysis, the link between PII and
Indic CVCV̄CV words is challenged. However, since there are consonant stems among the PII
loanwords (*ućig-, *bhišaj̄́-), there is no reason to assume that all thematic loanwords were
athematic in the source language(s) and subsequently thematicized within PII. Thus, the link
between PII and Indic CVCV̄CV words should be maintained.
Yet, the morphological difference between the groups could originate in a morphological
difference in the source languages. It is therefore more likely that CVCV̄CV words in PII and
Indic originate in slightly different, though related, languages, than in the same language.
5.2. r/n-alternation
Witzel (2003, p. 45) proposed that the Central Asian Substrate had a dialectal variation of r/n,
reflected in Indo-Iranian loanwords as well as other words in languages of Asia Minor, the
Middle East and the Caucasus. The evidence consists of the following words:
Table 4. Evidence for r/n-alternation in loanwords

meaning r-variants n-variants
leopard, panther Skt. pr̥̄́dāku-, PIr. pard-, Gr. πάρδαλις Gr. πάνθηρ
lion Parth. šarg, Khot. sarau Skt. siṁhá-, Arm. inc/j
wheat Bur. gur, Basque gari LAv. gaṇtuma-
water, river Bur. hur, Macro-Caucasian *(t)sir- Skt. sindhu-, LAv. həṇdu-
mustard Skt. sarṣapa- MiP span-dān, Gr. σίνᾱπι
59
Namely: ikṣvākú-, jábāru-, jarā̄́yu-, kiyā̄́mbu-, kúṇāru-, pr̥̄́dāku-, urvārū̄́-, viṣṇāpū̄́-, to which palāṇḍu- may be
added.
79
First, that there would be any etymological relation between Burušaski hur and PII *sindhu- is
by no means certain. As shown in chapter 3, it is impossible to ascertain that the r of Bur. gur
did not develop secondarily from *γund- or *γud-. Likewise, the n of Gr. πάνθηρ could be
secondary.
Secondly, one may question whether “r/n-alternation” is a proper way to describe the
variation. In ‘wheat’ and ‘water, river’, -r- corresponds to the cluster -nd(h)-. In ‘leopard,
panther’, one the other hand, both variants show a cluster, -rd- vs. -ndh-. The word for ‘mustard’
has -rs- in Indic but -n- in Iranian, although Khot. and Sogdian have neither -n- nor -r(s)-. Only
in the word for ‘lion’, does -r- in Iranian seem to correspond to -n- in Indic and Armenian, but
here the root vowels are also different. Thus, the evidence for r/n-alternation is more
heterogeneous than has previously been acknowledged.
Of the Indo-Iranian evidence, only Skt. sindhu-, Av. həṇdu- can be reconstructed for PII,
whereas the rest show irregular correspondences pointing to Post-PII borrowing. Indic and
Iranian have the same r/n-variant for ‘leopard, panther’, ‘wheat’, ‘water, river’, but different
variants for ‘lion’ and ‘mustard’. As for the geographical distribution, the r-variants are found
in both Indic and Iranian, as well as Greek, Burušaski and Caucasian languages. The n-variants
are absent from Burušaski and Caucasian languages, but seen in Indic, Iranian, Greek and
Armenian, the latter being geographically close to Caucasian languages. If the r/n-alternation
originates in dialectal variation, we would expect a clearer geographical distribution.
Thus, the words used as evidence show more variable elements than r vs. n, i.e. the r/n-
alternation is not their lowest common denominator. As such, it is methodologically hazardous
to use the concept of r/n-alternation to equate words that in reality are very different. Secondly,
their geographical and chronological distribution offers no reason to assume that the variation
would originate specifically in a Central Asian language.
5.3. The irregular correspondence Indic dh : Iranian t
A recurring irregular correspondence in Post-PII loanwords can be observed based on the dental
stops in Skt. godhū̄́ma- ‘wheat’ ~ LAv. gaṇtuma-, Khot. ganama, Bal. gandūm ‘wheat’ and Skt.
gandhá- ‘smell’ ~ LAv. gaiṇti- ‘bad smell’, Khot. ggañu ‘stench’. The Sanskrit, Khotanese,
and Balochi words point to PII *d(h), whereas LAv. points to *t. Although Witzel (2003, p. 31)
reconstructs a single PII form of ‘wheat’, I argue that the irregular variation cannot be explained
by secondary developments (cf. chapter 3).
80
It is interesting to note that the first syllable of *ganTuma- and *ganTi- is identical.60 This
similarity led Witzel (2003, p. 31) to assume a folk-etymological relationship between the
words, viz. Skt. godhū̄́ma- << *gandha-dhū̄́ma- ‘perfume smell’. However, other than their
formal similarity, there is no indication that such a folk-etymological influence would have
taken place, since Skt. godhū̄́ma- synchronically looks like a compound of go- ‘cow’ and
dhū̄́ma- ‘smoke’. A folk-etymological reanalysis of *ganTuma- based on *ganTi- and *umā-
‘flax’ is not unthinkable for Iranian, but in that case the expected outcome would be
**ganTi̯ uma-. Finally, if the similarity is due to folk etymology, original *t must have been
generalized in Avestan, but original *dh in Indic, which is unnecessarily complicated.61 It seems
preferable to project the reason behind the phonological similarity of the words to their source
language.
The recurring irregular correspondence dh : t allows for some interesting observations.
First, the fact that the same irregular correspondence is found twice makes it highly likely that
these words were borrowed from the same language.
Second, in the source language, the words contained a sound which was adapted as *dh
in Sanskrit and some Iranian languages, but *t in LAv. Based on the reconstructed phonology
of Indic and Iranian, it is difficult to find a plausible explanation for this irregularity. Both
Proto-Indic and Proto-Iranian have voiceless and voiced stops, implying that if the source had
[t] or [d], they should have been adapted as such. The source words might have contained a
breathy stop, close to Indic dh, but that would most likely have been adapted as Iranian d, not t
as in LAv. Thus, it is possible that the source language had a different sound altogether, or,
more precisely, a sound, alien to the synchronic Indic and Iranian phonologies, that was
interpreted as t by some Iranian speakers but dh (which may still have been [d] at this point) by
Indic speakers. Alternatively, one may assume that Indic and Iranian borrowed at different
points in time, and that the source language underwent a change from *t > *d (vel sim.) in the
meantime, but this is unlikely, given the relatively short time span from the disintegration of
PII until the attestation of the separate branches.
5.4. Non-initial mediae in clusters with *r or *n
A previously unnoticed feature of Indo-Iranian loanwords is the tendency for non-initial62
mediae to co-occur with n, r (sometime both) or r̥. Here is the evidence:
60
A similar anlaut is found in Skt. gandharvá- ~ LAv. gaṇdərəβa-, although here Indic and Iranian both point to
voiced *d(h).
61
If, for example, originally *ganti- vs. *ganduma-, then the influence must have gone in opposite directions in
Sanskrit and Avestan.
62
Initial mediae occur in: *dū̄̆rća-/*dr̥ća, *gadā-, *bīj̄́a-, *gr̥da-, LAv. gaēsa-, gaṇtuma-, gaiṇti-.
81
Table 5. Non-initial mediae in early Indo-Iranian loanwords
Elsewhere r-clusters n-clusters
Layer I (Pre-/early PII) *ućig-
Layer 0 (PII) *bīj̄́a- *gr̥da- *u̯and(H)-

*bhiš-aj̄́- *mr̥ga- *ringa-
*gadā- *kadru- *indra-
*sćāga- *nagna-
Layer II (late PII)
Layer III (Post-PII) *(H)arj̄́aná *banγα-
khaḍgá-/*karkadan *ganTuma-
*pərd-a(n)k- *pinda-
Total: 18 5 6 7
Table 5 shows that most (13/18) word-internal mediae co-occur with *n, *r or *r̥. Since the
words for ‘rhinoceros’ may be borrowed from different sources, they are best left out of this
discussion. Likewise, *(H)arj̄́aná- has no Indic equivalent.
In most cases, the media occupies the coda position of the cluster, but in *kadru- and
*nagna-, the media occupies the onset. The mediae are mostly dental or velar, rarely palatal,
and never bilabial.
In layer 0, 7/11 words follow the pattern. Of the 4 words with mediae that do not, two
(*bīj̄́a-, *bhiš-aj̄́-) contain palatals. By contrast, mediae in clusters with *r and *n are never
palatals in PII loanwords. This potentially indicates that palatals stops, i.e. affricates, adhered
to different phonotactic rules than stops in the source language(s). PII *gadā- may in principle
go back to *gn̥dā-, in which case it would fit into the pattern.
The high frequency of mediae in clusters with *n, *r or *r̥ is contrasted by the low
frequency of other stops in these positions in PII loanwords. Three cases are attested: *anću-,
*Hustra- and *u̯r̥tka-. Since palatal stops do not seem to be part of the pattern, *anću- may be
disregarded. PII *Hustra- has a variant *Hustar-, implying that the *r may originally not have
been part of the cluster. The *t in *u̯r̥tka- could in principle have been a media *d originally,
since it would have been devoiced by sandhi anyway.
In Indo-Iranian inherited vocabulary, all three stop series occur in clusters with *n, *r or
*r̥, cf. Skt. vártate ~ LAv. varətata ‘to turn’, Skt. pardate ~ LAv. pərədən ‘to fart’, Skt.
spárdhate ‘to contest’, Skt. pánthā- ~ Av. paṇtā ‘way’, Skt. skándati ‘jumps’, Skt. bandhaya-
82
~ LAv. baṇdaiie- ‘to bind’, etc. It is thus likely that the co-occurrence of dental and velar mediae
with *n, *r or *r̥ in loanwords reflects a feature of the source language(s). In other words, it is
likely that the source language(s) of these PII borrowings only allowed one type of stop in
clusters with *n, *r or *r̥, which was nativized as PII mediae.
Depending on the phonetic interpretation of PII mediae, the above feature appears more
or less salient. If the mediae were pre-glottalized, the correlation with *n and *r is quite salient.
On the other hand, if the mediae were plain voiced stops at the time of borrowing, the co-
occurrence with nasals is quite trivial, since voicing of stops after nasals is very common cross-
linguistically (Kümmel, 2007, p. 53). By implication, the feature would be less likely to reflect
the phonological system of a single substrate language. However, the same does not apply to
the co-occurrence of mediae and *r.
In layer III, *pərd-a(n)k-, *pinda- and *banγα- follow the pattern. The Iranian reflexes
of *ganTi- and *ganTuma- show either voiceless or voiced stops after *n (cf. chapter 3). The
Indic equivalents have voiced aspirates. One possibility is that the variation is caused by the
changes in the stop systems of Iranian and Indic. Another is that *ganTi- and *ganTuma-
were borrowed from a different language than the loanwords with mediae in clusters with *n,
*r or *r̥. A third possibility is that the correlation of mediae with *n and *r only holds for the
PII layers. With so few examples in Post-PII, the correct scenario cannot be determined with
any degree of certainty.
5.5. Correlation between *i and affricates
Lubotsky (2001b, p. 304) observed the high frequency of palatal stops and clusters containing
*s in Indo-Iranian loanwords. Another tendency is the correlation between *i and affricates.
The evidence is presented below:
83
Table 6. Loanwords containing *i, as well as loanwords containing palatals
Palatal + *i Other obstruents + *i *i elsewhere Other palatals
Layer I (Pre- *ućig-
/early PII)
Layer 0 (PII) *u̯rīj̄́hi- *bhiš-aj̄́- *j̄́harmii̯ a- *anću-
*kaći̯ apa- *kućsi- *āni- *bīj̄́a-
*rāći- *išt(i)- *ringa- *ćaru̯a-
*u̯āćī- *kšīra- *indra- *kāća-
*ćyā- *matsi̯ a- *bhiš-aj̄́-
*r̥si- *j̄́harmii̯ a-
*bīj̄́a- *sćāga-
*u̯arāj̄́ha-
Layer II (late *pīi̯ ūša- *i̯ au̯īi̯ ā- *pusća-
PII) *mai̯ ūkHa-
Layer III *ćika(tā)- *(t)sūkV̄- (*sūčī-) *mižuka- *dū̄̆rća-
(Post-PII) *ganTi- *nai̯ Ts(a)- *Kai̯ ća-
*pinda- śāli- (Skt.) śaṇá-/kanaba-
*šu̯ai̯ pa-
*u̯īna-
Total: 41 7 11 11 12
Focusing on the PII words (layer I, 0, II) where *i follows an obstruent (n=13), we see that 5
co-occur with a primary palatal. Of the 8 remaining cases, three (*kućsi-, *kšīra-, *matsi̯ a-)
show clusters with *s, another (*r̥si-) a simple *s. Note that the cluster of *kućsi- could go back
to *tć. Three cases (*pīi̯ ūša-, *bhiš-aj̄́-, *bīj̄́a-) show labial stops preceding *i. Only *išt(i)- has
a dental stop before *i, but the original derivational suffix is not certain for this word (cf. chapter
3).
Thus, non-labial obstruents before *i are predominantly palatal affricates or clusters with
*-s, which are phonetically close to affricates. There is no inner-Indo-Iranian explanation for
this phenomenon, since the PII palatalization before *i only affects velars and does not produce
primary palatals, except in the case of PIE *ske/i > PII *sć. Given this distribution, a reasonable
hypothesis is that non-labial stops were affricated before *i in the source language(s). The
original stops may have been dental or velar.
84
That said, not all primary palatals in PII loanwords can be analyzed in this way: 9
loanwords contain palatals that are not followed by *i. Cases like *sćāga-, *j̄́harmii̯ a-, and
*ćaru̯a- could reflect palatals before a high vowel Pre-PII *e, which could be assumed to have
caused palatalization in the source language(s), but this is mere speculation. Moreover, the
palatal in *anću- appears in a distinctly non-palatal context. Thus, we must assume that the
source language(s) of these words also possessed phonemic palatals (that were not conditioned
by a following *i), or at least, a phoneme that was nativized as PII palatals.
In Post-PII loanwords, *ćika(tā)- seems to follow the above pattern. In the case of *sūčī-,
the *č may be secondary. On the other hand, *ganTi- contradicts the distribution. The evidence
is too scarce to allow for a clear analysis.
5.5.1. Dental stops
Above, the high frequency of palatal stops and clusters with *s before *i in PII loanwords was
explained by postulating a process of affrication of dental or velar stops in the source
language(s). If this hypothesis is correct, we would expect dental and velar stops in loanwords
to occur in non-palatalizing context. The evidence for dental stops is given below:
Table 7. Loanwords containing dental stops

Before *u Before *r Before thematic vowel Elsewhere
or *-ā-
Layer I
(Pre-/early
PII)
Layer 0 *stuka- *Hustra- *gadā- *atHarvan-
(PII) *kadru- *gr̥da- *atka-
*indra- *kapau̯ta- *matsi̯ a-
*pau̯asta- *išt(i)-
*u̯and(H)-
*u̯r̥tka-
Layer II *čāt(u̯āla)-
(Late PII)
Layer III *dū̄̆rća- *ćikatā- *ganTi-
(Post-PII) *ganTuma- *pinda- *ganDəru̯a-
*pərd-a(n)k-
Total: 22 4 3 7 8
85
In the PII layers (0, II), dental stops almost exclusively occur in non-palatalizing contexts. This
is most clearly seen in the 5 cases were a dental stop precedes *u or *r. In 4 cases, a dental stop
precedes the suffixes -a- (< *-o-) or -ā- (< *-eh2-), which were both non-palatalizing in PII. Of
course, the suffix vowels may have been added after the words were borrowed, but in any case,
there is no indication that the stops were followed by an *i or another palatalizing vowel in the
source language. Of the 6 remaining cases, *atHarvan-, *atka- and *u̯r̥tka- show *t in clusters
with *H and *k, which may be treated as non-palatalizing contexts. Conversely, *matsi̯ a-
belongs to the affricated group described in the previous section. For *išt(i)-, the original
derivation may not have been an i-stem. In the case of *u̯and(H)-, the dental occurs in different
contexts depending on the derivation.
The fact that dental stops occur in non-palatalizing contexts provides indirect support for
the hypothesis that PII palatals and clusters before *i reflect affricated stops in the source
language(s) of PII loanwords. The source language(s) seems to show a complementary
distribution of *t, *d / _*u, *r, *ā̄̆ and *ć, *j̄́h, *tć, *ts / _*i, which most likely reflects a historical
process.
In Post-PII loanwords, there is one case (*ganTi-) of a dental stop in a palatalizing
context. This suggests that the above analysis only holds for the PII layers, or that *ganTi- was
borrowed from a different source language.
86
5.5.2. Velar stops
As stated above, affricates before *i could in principle also reflect original velar stops, in which
case we would expect velar stops in to occur in non-palatalizing contexts. The evidence is given
below:
Table 8. Loanwords containing velar stops

Before Before thematic Before other Word-final
consonant vowel vowels position
Layer I (Pre-/early *ućig-
PII)
Layer 0 (PII) *kHā- *aka- *gadā-
*kHara- *atka- *gr̥da-
*kšīra- *mr̥ga- *kaći̯ apa-
*nagna- *muska- *kāća-
*ringa- *kadru-
*sćāga- *kapāra-
*stuka- *kapau̯ta-
*u̯r̥tka- *kapHa-
*kućsi-
Layer II (Late PII) *mai̯ ūkHa- *ǰaǰha/ukā̄̆- *kárus-
*čāt(u̯āla)-
Layer III (Post-PII) *banγα- *ćikatā- *pərd-a(n)k-
khaḍgá-/*karkadan *ganDəru̯a-
siṁhá- (Skt.) *ganTi-
*ganTuma-
*Kai̯ ća-
*kanaba-
*(t)sūkV̄-
Total: 36 5 12 18 2
Pre-vocalic velar stops in PII and Post-PII loanwords mostly occur before *ā̄̆, *u or *r̥, which
are non-palatalizing contexts. Here, PII *ā̄̆ cannot go back to a high vowel Pre-PII *ē̄̆, since the
velar would have been palatalized within PII. Some velars occur before *H or *n, which are
also non-palatalizing contexts. PII *kšīra- is part of the affricated group described above.
87
There are five cases of palatalized velars: three in PII (*ǰaǰha/ukā̄̆-, *čāt(u̯āla)-, *ućig-)
and two in Post-PII (*(t)sūkV̄-, Skt. siṁhá-). Only *ućig- (with palatalized *ǰ in the gen.sg.) is
likely to have undergone PII palatalization. The palatalized velar of Skt. sūcī̄́- (<< *(t)sūkV̄-)
may have arisen secondary within Indo-Iranian. In the remaining cases, the palatalized velar
could reflect the original form of the source language.
As with the dentals, the absence of velar stops before *i is compatible with the hypothesis
that PII palatals and clusters before *i reflect affricated stops in the source language(s).
However, the existence of palatalized velars in PII loanwords could indicate that affricated
velars in the source language(s) were adapted as PII *Č. In that case, dental stops are the most
likely origin of affricates + *i. Another indication of this is that *matsi̯ a- and *kućsi- (if
< *kutći-) contain dental clusters.
5.6. The sequence *-ru̯-
The recurring sequence *-ru̯- in PII and Sanskrit loanwords was observed by Lubotsky
(2001b, p. 304). Among early Indo-Iranian loanwords, the evidence consists of *atHaru̯an-,
*ćaru̯a-, and *ganDəru̯a-, to which I have added *bharu̯-. Although *-ru̯- is not absent in the
inherited vocabulary, the sounds are generally separated by a morpheme boundary, e.g. Skt.
sárva- ‘all’ ~ Av. hauruua- ‘whole’ < PIE *solh2-u̯o-, which is probably a thematicized u-
stem (Pronk, 2011, p. 189). The loanwords with *-ru̯- could in principle contain a morpheme
boundary as well, but there is no indication that that is the case.
The word *ganDəru̯a- is Post-PII, whereas the three remaining words are PII. Although
the number of words is quite small, it is noteworthy that the *-ru̯- cluster is found in both
chronological layers.
88
6. Conclusions
In this chapter, the results of the thesis will be summarized and discussed.
6.1. Summary of main results
The 103 words of possible non-IE origin that have been analyzed in this study may be divided
into two groups. 29 words do not fulfill the necessary criteria to make substrate origin likely.
In the majority of cases, this is due to the existence of possible or plausible IE comparanda. For
8 words, however, there is no IE etymology, but no other criteria make a non-IE origin likely.63
Although borrowing is in principle as likely as inheritance in such cases, they were left out of
the loanword corpus to avoid interference with the results.
The remaining 74 words can be considered as loanwords according to the applied
methodology. Besides lacking IE etymologies, these words show structural peculiarities that
separate them from the inherited lexicon and/or specific semantics that make them particularly
liable to borrowing.
It has furthermore been demonstrated that the 74 early Indo-Iranian loanwords cannot be
ascribed to the same chronological layer in the history of Indo-Iranian. The majority, 53 words,
are reconstructable to PII (layers 0, I, II). Within this group, only one word (*ućig-) shows
evidence of an early Pre-PII time of borrowing, although this is not absolutely certain. On the
other hand, 7 words (layer II) were borrowed in late PII, after the operation of various sound
changes, such as the phonologization of *a, the palatalization of velars, and the lengthening of
short vowels preceding laryngeals (*V̄̆H > *V̄). The remaining 45 words (layer 0) could
theoretically have been borrowed at any point during Pre-PII or PII, but 13 words have features
that indicate a late PII time of borrowing.
21 loanwords showing irregular correspondences were classified as Post-PII. 19 of these
loanwords are most likely related, and 13 of those are so similar that they reflect parallel
borrowings by Indic and Iranian from the same source language. The two remaining words are
too dissimilar to have any etymological relation.
The analysis of structural characteristics of early Indo-Iranian loanwords generally
supports the conclusions of previous literature. However, the evidence for Witzel’s r/n-
alternation is very scarce. Even if it is accepted, the wide geographical distribution of the r/n-
variants do not support the idea that the words originated in the Central Asian Substrate.
63
These are *āćā-/*aćas-, *dhu̯aǰ-, *j̄́hai-, *nard-, *raj̄́h-, *srans-, *u̯riH-, *u̯i̯ ak-.
89
Two new structural characteristics were proposed in chapter 5. In PII loanwords, velar
and dental stops are absent before *i. In contrast to this, non-labial obstruents before *i are
palatals (phonetically affricates) or clusters with *s (that are phonetically close to affricates).
Thus, obstruents seem to be complementarily distributed depending on the following vowel. I
argued that this reflects a feature of the source language, probably affrication of dental stops
before *i.
Another characteristic of PII loanwords is that non-initial dental and velar mediae
co-occur with *n and *r. Conversely, tenues and aspiratae almost never appear in this position.
Since Indo-Iranian allows all series of stops in clusters with *n or *r, I argued that this reflects
a phonotactic feature of the source language of the loanwords.
Neither of the new structural characteristics seem to hold for the Post-PII layer.
Interestingly, for both characteristics, the counterexamples are the words *ganTi- ‘smell’ and
*ganTuma- ‘wheat’, which consequently appear increasingly isolated from the rest of the
loanword corpus. In fact, other Post-PII loanwords are generally in line with the PII pattern.
The *-ru̯- cluster of Post-PII *ganDəru̯a- also shows a link between the layers. It may thus be
best to view *ganTi- and *ganTuma- as outliers, perhaps originating in a different language
than the rest.
Together with the patterns proposed in previous literature, the evidence for affricates + *i
and *n/r + mediae lend additional evidence to the hypothesis that most early Indo-Iranian
loanwords originate in the same unknown substrate language.
6.2. Identity of source languages
A minority of early Indo-Iranian loanwords have been proposed to originate in known
languages.
The possibility of a Uralic origin has been discussed for six words:
1) Skt. bhaṅgá- ‘hemp’ << PU *pe̮ŋka- ‘mushroom’
2) *indra- << PU *ilmar / *inmar ‘thunder god’
3) *matsi̯ a- ‘fish’ << PU *maća ‘fish net’
4) *u̯āćī- ‘axe’ << PFU wäŋći ‘knife’
5) *kapHa- ‘phlegm’ << PU *kompa ‘wave’
6) *pusća- ‘tail’ << PFU *ponci ‘tail’.
The first word is not reconstructable to PII, and since it refers to a domesticated plant, a
BMAC origin seems more likely. Moreover, the cluster -ng- has been shown to be characteristic
of PII loanwords of unknown origin. Words 2, 3, and 4 are not reconstructable to Proto-Uralic,
90
which is the stage at which contact between Uralic and PII otherwise seems to have taken place.
In 5, the semantics of the proposed source does not match the Indo-Iranian word. While
semantic change may be considered, a Uralic origin of *kapHa- remains speculative. The 6th
case is semantically plausible, but the correspondence Uralic *n : PII *Ø presents a formal
problem. Interestingly, words 4 and 5 also show the correspondence Uralic *n : PII *Ø, but
since they present other problems, it is unlikely that this reflects a regular adaptation strategy
of Uralic loanwords into Indo-Iranian.
Ultimately, it seems unlikely that Uralic was a major donor language of early Indo-Iranian
loanwords.
A Near Eastern source has been proposed for *kHara- ‘donkey’, *ganTuma- ‘wheat’,
*ganDəru̯a- ‘a mythical being’, and *kanaba- ‘hemp’, based on similar words in languages of
the Near East and, to some extent, Greek. While the ultimate source could be a known language
of the Near East (e.g. Sumerian for *kanaba-), a direct source of borrowing cannot be
determined for any of these words. Therefore, an intermediary language located in Central Asia,
transmitting Near Eastern (agricultural) vocabulary, remains equally likely.
Thus, to account for the 74 early Indo-Iranian loanwords treated here, it remains necessary
to assume unknown donor language(s).
6.3. Implications for the Central Asian Substrate Hypothesis
The Central Asian Substrate Hypothesis places the unknown donor language of early Indo-
Iranian loanwords in the BMAC culture. As we have seen, a core argument of the hypothesis is
the semantics of certain PII loanwords, which can be connected to the material culture of the
BMAC. However, many loanwords cannot be linked to material culture. Using the structural
characteristics advanced in this study, words without reference to material culture can be
connected to words with reference to material culture. For example, since *kaći̯ apa- ‘tortoise’
and *u̯āćī- ‘axe, knife’ can be connected to the BMAC (Lubotsky, 2001b, p. 307), it becomes
increasingly likely that other loanwords with affricate + *i, like *ćyā- ‘to freeze, congeal’,
*ućig- ‘sacrificing priest’, and *matsi̯ a- ‘fish’, also originate in the Central Asian Substrate. In
this way, the study offers new ways to bridge the gap between linguistics and archaeology in
support of the Central Asian Substrate Hypothesis. However, a complicating factor in this
particular case is that *u̯rīj̄́hi- ‘rice’, also with affricate + *i, cannot easily be connected with
the BMAC culture, since rice was not cultivated here.
Another contribution of the present study to the Central Asian Substrate Hypothesis is
that the time of borrowing of many loanwords has been shown to be late PII or shortly Post-
91
PII, rather than Pre-PII. This supports the hypothesis, since the language contact between Indo-
Iranian and the Central Asian Substrate is believed to have occurred after the founding of the
Sintashta culture, where PII was probably spoken, at a time when Indo-Iranian speakers,
identified with the Andronovo cultures, spread over a larger area in Central Asia.
6.4. Directions for future research
Several questions remain open for future research. The division of loanwords into chronological
layers could be supplemented by a detailed integration of archaeological data, to investigate
whether chronological layers of linguistic development can be connected to archaeological
layers of cultural development. On the linguistic side, one could investigate to what extent the
newly proposed structural characteristics of loanwords hold for the corpus of Indic loanwords
proposed by Kuiper (1991). Lastly, future research would benefit from incorporating potentially
crucial evidence from Middle and Modern Iranian, Indic and Nuristani languages to a greater
extent.
92
Bibliographical abbreviations
AiGr. = Debrunner, A. & Wackernagel, J. (1954). Altindische Grammatik Band II, 2: Die
Nominalsuffixe. Göttingen: Vandenhoeck & Ruprecht.
AiWB = Bartholomae, C. (1904). Altiranisches Wörterbuch. Berlin: Walter De Gruyter & Co.
CAD = Gelb, I. J., et al. (1956-2010). The Assyrian Dictionary of the Oriental Institute of the
University of Chicago. 21 vols. Chicago: The Oriental Institute at the University of
Chicago.
EWAia = Mayrhofer, M. (1992-2001). Etymologisches Wörterbuch des Altindoarischen I-III.
Heidelberg: Universitätsverlag Carl Winter.
LIV = Rix, H. & Kümmel, M. (2001). Lexikon der indogermanischen Verben. Wiesbaden: Dr.
Ludwig Reichert Verlag.
UEW = Rédei, K. (1988). Uralisches Etymologisches Wörterbuch. Budapest: Akadémiai
Kiadó
References
Adams, D. Q. (2013). A Dictionary of Tocharian B. Amsterdam/New York: Rodopi.
Aikio, A. (2015). The Finnic ‘Secondary e-stems’ and Proto-Uralic Vocalism. Suomalais-
Ugrilaisen Seuran Aikakauskirja 95, 25-66.
Anthony, D. W. (2007). The Horse, the Wheel, and Language: How Bronze Age Riders from
the Eurasian Steppe Shaped the Modern World. Princeton/Oxford: Princeton University
Press.
Anthony, D. W. & Ringe, D. (2015). The Indo-European Homeland from Linguistic and
Archaeological Perspectives. Annual Review of Linguistics 1(1), 199-219.
Bailey, H. W. (1979). Dictionary of Khotan Saka. Cambridge: Cambridge University Press.
Beekes, R. S. (1981). The Neuter Plural and the Vocalization of the Laryngeals in Avestan.
Indo-Iranian Journal 23, 275-287.
Beekes, R. S. P. (1992). Widow. Historische Sprachforschung 105 (2), 171-188.
Beekes, R. S. P. (1996). Ancient European Loanwords. Historische Sprachforschung 109 (2),
215-236.
Beekes, R. S. P. (2010). Etymological Dictionary of Greek. Leiden: Brill.
Beekes, R. S. P. (2011). Comparative Indo-European Linguistics: An Introduction (2nd ed.).
Amsterdam: John Benjamins Publishing Company.
93
Berger, H. (1959). Deutung einiger alter Stammesnamen der Bhil aus der vorarischen
Mythologie des Epos und der Purana. Wiener Zeitschrift für die Kunde Süd- und
Ostasiens und Archiv für indische Philosophie III, 34-82.
Berger, H. (1970). Die Burušaski-Lehnwörter in der Zigeunersprache. Indo-Iranian Journal 3,
17-43.
Blažek, V. & Hegedűs, I. (2012). On the position of Nuristani within Indo-Iranian. In R.
Sukač & O. Šefčík, The Sound of Indo-European 2 (pp. 40-66). München: Lincom.
Burrow, T. (1957). Sanskrit “gṝ-/gur-” ‘to Welcome’. Bulletin of the School of Oriental and
African Studies 20(1), 133-144.
Burrow, T. (1973). The Sanskrit Language. London: Faber and Faber.
Byrd, A. M. (2015). The Indo-European Syllable. Leiden: Brill.
Cantera, A. (2001). Die Behandlung der idg. Lautfolge (C)RHC- im Iranischen. Münchener
Studien zur Sprachwissenschaft 61, 7-27.
Cantera, A. (2017). The Phonology of Iranian. In J. Klein, B. Joseph & M. Fritz, Handbook of
Comparative and Historical Indo-European Linguistics (pp. 481-503). Berlin: De
Gruyter Mouton.
Cheung, J. (2007). Etymological Dictionary of the Iranian Verb. Leiden: Brill.
Clayton, J. (2018). Rounding of Indo-Iranian *R(H). Retrieved from UCLA: Program in Indo-
European Studies:
https://pies.ucla.edu/IEConference/IEChandouts/clayton_j_2018h.pdf
Damgaard et al. (2018). The First Horse Herders and the Impact of Early Bronze Age Steppe
Expansions into Asia. Science 360 (6396).
Davary, G. D. (1982). Baktrisch: ein Wörterbuch auf Grund der Inschriften, Handschriften,
Münzen und Siegelsteine. Heidelberg: Groos.
Décsy, G. (1990). The Uralic Proto-Language: A Comprehensive Reconstruction.
Bloomington: Eurolingua.
Derksen, R. (2008). Etymological Dictionary of the Slavic Inherited Lexicon. Leiden: Brill.
Eilers, W. (1959). Akkad. kapsum “Silber, Geld” und Sinnverwandtes. Die Welt des Orients
2, 322-337 + 465-469.
Francfort, H.-P. (2005). La civilisation de l’Oxus et les Indo-Iraniens et Indo-Aryens en Asie
Centrale. In G. Fussman, J. Kellens, H.-P. Francfort, & X. Tremblay, Aryas, Aryens et
Iraniens en Asie Centrale (pp. 253-328). Paris: Collège de France.
Frog. (2012). Confluence, Continuity and Change in the Evolution of Mythology: The Case of
the Finno-Karelian Sampo-Cycle. In A.-L. Siikala & E. Stepanova, Mythic Discourses:
94
Studies in Uralic Traditions. Studia Fennica Folkloristica 20. (pp. 205-254). Helinki:
Finnish Literature Society.
Gershevitch, I. (1961). A Grammar of Manichean Sogdian. London: Basil Blackwell.
Gharib, B. (1995). Sogdian Dictionary. Tehran: Farhangan Publications.
Haak et al. (2015). Massive Migrations from the Steppe were a Source for Indo-European
Languages in Europe. Nature 522, 207-220.
Hegedűs, I. (2012). The RUKI-rule in Nuristani. In B. Nielsen Whitehead, T. Olander & B. A.
Olsen, The Sound of Indo-European 1 (pp. 145-167). Copenhagen: Museum
Tusculanum Press.
Heide, M. (2011). The Domestication of the Camel. Ugarit-Forschungen 42, 331-382.
Henning, W. B. (1939). Sogdian Loan-Words in New Persian. Bulletin of the School of
Oriental Studies 10, 93-106.
Hock, H. H. (1991). Principles of Historical Linguistics. Berlin: Mouton de Gruyter.
Hoffmann, K. & Forssman, B. (1996). Avestische Laut- und Flexionslehre. Innsbruck: Institut
der Sprachwissenschaft der Universität Innsbruck.
Jamison, S. W. & Brereton, J. B. (2014). The Rigveda: The Earliest Religious Poetry of India.
Oxford: Oxford University Press.
Judd et al. (2018). Live in the Fast Lane: Settled Pastoralism in the Central Eurasian Steppe
during the Middle Bronze Age. American Journal of Human Biology 30, 1-23.
Kehrein, W. (2002). Phonological Representation and Phonological Phasing. Tübingen:
Niemeyer.
Kellens, J. (1995). Liste du verbe avestique. Wiesbaden: Dr. Ludwig Reichert Verlag.
Kent, R. G. (1950). Old Persian: Grammar, Texts, Lexicon. New Eaven: American Oriental
Society.
Kim, R. (2000). Reexamining the History of Tocharian B ‘ewe’. Tocharian and Indo-
European Studies 9, 37-43.
Kloekhorst, A. (2008). The Hittite Inherited Lexicon. Leiden: Brill .
Kloekhorst, A. (2011). Weise’s Law: Depalatalization of Palatovelars before *r in Sanskrit.
Indogermanistik und Linguistik im Dialog: Akten der XIII. Fachtagung der
Indogermanischen Gesellschaft vom 21. bis 27. September 2008 in Salzburg (pp. 261-
270). Wiesbaden: Reichert Verlag.
Kloekhorst, A. (2014). Accent in Hittite: A Study in Plene Spelling, Consonant Gradation,
Clitics, and Metrics. Wiesbaden: Harrassowitz.
95
Kloekhorst, A. (2016). The Anatolian Stop System and the Indo-Hittite Hypothesis.
Indogermanische Forschungen 121, 213-247.
Kloekhorst, A. (2018). Anatolian Evidence Suggests that the Indo-European Laryngeals *h2
and *h3 were Uvular Stops. Indo-European Linguistics 6, 69-94.
Kobayashi, M. (2004). Historical Phonology of Old Indo-Aryan Consonants. Tokyo:
Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of
Foreign Studies.
Kobayashi, M. (2017). The phonology of Indic. In J. Klein, B. Joseph & M. Fritz, Handbook
of Comparative and Historical Indo-European Linguistics (pp. 325-344). Berlin: De
Gruyter Mouton.
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, Early Contacts
between Uralic and Indo-European: Linguistic and Archaeological Considerations (pp.
235-263). Helsinki: Suomalais-Ugrilainen Seura.
Kortlandt, F. H. H. (1983). Greek Numerals and PIE Glottalic Consonants. Münchener
Studien zur Sprachwissenschaft 42, 97-104.
Kortlandt, F. H. H. (1985). Long Vowels in Balto-Slavic. Baltistica 21(2), 112-124.
Kortlandt, F. H. H. (1985). Proto-Indo-European Glottalic Stops: the Comparative Evidence.
Folia Linguistica Historica 6(2), 183-201.
Kortlandt, F. H. H. (1986). Posttonic *w in Old Irish. Éiru 37, 89-92.
Kortlandt, F. H. H. (1996). The High German Consonant Shift. Amsterdamer Beiträge zur
älteren Germanistik 46, 53-57.
Kortlandt, F. H. H. (2018). Proto-Indo-European Glottalic Stops: The Evidence Revisited.
Münchener Studien zur Sprachwissenschaft 71(1), 147-153.
Kroonen, G. (2013). Etymological Dictionary of Proto-Germanic. Leiden : Brill.
Kroonen, G., Barjamovic, G. & Peyrot, M. (2018). Linguistic Supplement to Damgaard et al.
2018: Early Indo-European Languages, Anatolian, Tocharian, Indo-Iranian.
https://doi.org/10.5281/zenodo.1240524.
Kuiper, F. B. J. (1948). Proto-Munda Words in Sanskrit. Amsterdam: Noord-Hollandsche
Uitgevers Maatschapij.
Kuiper, F. B. J. (1976). Old East Iranian Dialects. Indo-Iranian Journal 18, 241-253.
Kuiper, F. B. J. (1991). Aryans in the Rigveda. Amsterdam: Rodopi.
Kuiper, F. B. J. (1995). Gothic bagms and Old Icelandic ylgr. NOWELE 25, 63-88.
Kuryłowicz, J. (1935). Études Indoeuropéennes I. Krakow: Polska Akademija Umiejetnosci.
96
Kuz’mina, E. E. (2007). The Origins of the Indo-Iranians. Leiden: Brill.
Kuznetsov, P. F. (2006). The Emergence of Bronze Age Chariots in Eastern Europe. Antiquity
80, 638-45.
Kümmel, M. J. (2005). Vedisch tand- und ein neues indoiranisches Lautgesetz. In G.
Schweiger, Indogermanica: Festschrift Gert Klingenschmitt (pp. 321-332). Taimering:
Schweiger VWT-Verlag.
Kümmel, M. J. (2007). Konsonantelwandel: Bausteine zu einer Typologie des Lautwandels
und ihre Konsequenzen für die vergleichende Rekonstruktion. Wiesbaden: Dr. Ludwig
Reichert Verlag.
Kümmel, M. J. (2012). Typology and Reconstruction. In B. Nielsen Whitehead, T. Olander &
B. A. Olsen, The Sound of Indo-European: Phonetics, Phonemics, and
Morphophonemics (pp. 291-329). Copenhagen: Museum Tusculanum Press.
Kümmel, M. J. (2016a). Is ancient old and modern new? Fallacies of attestation and
reconstruction (with special focus on Indo-Iranian). In D. M. Goldstein, S. W. Jamison
& B. Vine, Proceedings of the 27th Annual UCLA Indo-European Conference, Los
Angeles, October 23rd ad 24th, 2015 (pp. 79-96). Bremen: Hempen Verlag.
Kümmel, M. J. (2016b). Zur ‘Vokalisierung’ der Laryngale im Indoiranischen. In D. Gunkel,
J. T. Katz, B. Vine & M. Weiss, Sahasram Ati Srajas: Indo-Iranian and Indo-European
Studies in Honor of Stephanie W. Jamison (pp. 216-226). Ann Arbor/New York: Beech
Stave Press.
Kümmel, M. J. (2017). Agricultural Terms in Indo-Iranian. In M. Robbeets & A. Savelyev,
Language Dispersal Beyond Farming (pp. 275-290). Amsterdam: John Benjamins
Publishing Company.
Kümmel, M. J. (2018). The Survival of Laryngeals in Iranian. In L. Beek, A. Kloekhorst, G.
Kroonen, M. Peyrot & T. Pronk, Farnah: Indo-Iranian and Indo-European Studies in
Honor of Sasha Lubotsky (pp. 162-172). Ann Arbor/New York: Beech Stave Press.
Kümmel, M. J. (2019). Early Indo-Iranic Loans in Uralic: Sounds and Strata. Contacts:
Archaeology, genetics, languages joining forces to shed light on early contacts (4000
BC – 1000 AD) between Indo-European and Uralic speakers. Sveaborg.
Kümmel, M. J. (2019). Einführung ins Ostmitteliranische. URL:
https://www.academia.edu/30130317/Einführung_ins_Ostmitteliranische_Introduction_
to_Eastern_Middle_Iranian_.
97
Lipp, R. (2009). Die indogermanischen und einzelsprachlichen Palatale im Indoiranischen:
Band II, Thorn-Problem, indo-iranische Laryngalvokalisation. Heidelberg:
Universitätsverlag Winter.
Lubotsky, A. M. (1981). Gr. pḗgnumi : Skt. pajrá- and loss of laryngeals before mediae in
Indo-Iranian. Münchener Studien zur Sprachwissenschaft 40, 133-138.
Lubotsky, A. M. (1988). The System of Nominal Accentuation in Sanskrit and Proto-Indo-
European. Leiden: Brill.
Lubotsky, A. M. (1989). Against a Proto-Indo-European Phoneme *a. In T. Venneman, The
New Sound of Indo-European (pp. 53-66). Berlin: Walter de Gruyter.
Lubotsky, A. M. (1990). La loi de Brugmann et *H3e-. La reconstruction des laryngeales
(Bibliothèque de la Faculté de Philosophie et Lettres de l’Université de Liège 253),
129-136.
Lubotsky, A. M. (1994). RV. ávidhat. In G. E. Dunkel, G. Meyer, S. Scarlata & C. Seidl,
Früh-, Mittel-, Spätindogermanisch. Akten der IX. Fachtagung der Indogermanischen
Gesellschaft vom 5. bis 9. Oktober 1992 in Zürich (pp. 201-206). Wiesbaden: Dr.
Ludwig Reichert Verlag.
Lubotsky, A. M. (1995). Reflexes of Intervocalic Laryngeals in Sanskrit. In W. Smoczýnski,
Kurylowicz Memorial Volume. Part One (pp. 213-233). Krakow: Universitas.
Lubotsky, A. M. (1997). The Indo-Iranian reflexes of PIE *CRHUV. In A. M. Lubotsky,
Sound Law and Analogy: Papers in honor of Robert S.P. Beekes on the occasion of his
60th birthday (pp. 139-154). Amsterdam: Rodopi.
Lubotsky, A. M. (2000). Indo-Aryan ‘six’. In M. Ofitsch & C. Zinko, 125 Jahre
Indogermanistik in Graz (pp. 255-261). Graz: Leykam.
Lubotsky, A. M. (2001a). Reflexes of Proto-Indo-European *sk in Indo-Iranian. Incontri
linguistici 24, 25-57.
Lubotsky, A. M. (2001b). The Indo-Iranian Substratum. Early Contacts between Uralic and
Indo-European: Linguistic and Archaeological Considerations. Papers presented at an
international symposium held at the Tvärminne Research Station of the University of
Helsinki 8-10 January 1999 (pp. 301-317). Helsinki: Mémoires de la Société Finno-
ougrienne.
Lubotsky, A. M. (2008). Vedic ‘ox’ and ‘sacrificial cake’. In A. M. Lubotsky, J. Schaeken &
J. Wiedenhof, Evidence and Counter-evidence. Essays in Honor of Frederik Kortlandt.
Volume 1: Balto-Slavic and Indo-European Linguistics (pp. 351-360). Amsterdam:
Rodopi.
98
Lubotsky, A. M. (2018). Proto-Indo-Iranian Phonology. In J. Klein, M. Fritz & M. Fritz,
Handbook of Comparative and Historical Indo-European Linguistics (pp. 1877-1888).
Berlin: De Gruyter Mouton.
Macdonell, A. A. (1916). A Sanskrit Grammar for Students. New Delhi: D.K. Printworld.
Mallory, J. P. (1989). In Search of the Indo-Europeans: Langauge, Archaeology and Myth.
New York, N.Y: Thames and Hudson.
Mallory, J. P. (1998). A European Perspective on Indo-Europeans in Asia. In V. Mair, The
Bronze Age and Early Iron Age Peoples of Eastern Central Asia (pp. 175-201).
Washington/Philadephia: The University of Pennsylvania Museum Publications.
Martirosyan, H. (2010). Etymological Dictionary of the Armenian Inherited Lexicon. Leiden:
Brill.
Matasovic, R. (2009). Etymological Dictionary of Proto-Celtic. Leiden: Brill.
Mayrhofer, M. (1983). Lassen sich die Vorstufen des Uriranischen nachweisen? Akten der
Österreichischen Akademie der Wissenschaften, 249-255.
Mayrhofer, M. (1986). Indogermanische Grammatik. 2. Halbband: Lautlehre. Heidelberg:
Carl Winter Universitätsverlag.
Melchert, H. C. (1977). Tocharian Verb Stems in -tk-. Zeitschrift für vergleichende
Sprachforschung 91, 93-130.
Morgenstierne, G. (1938). Indo-Iranian Frontier Languages Vol. II: Iranian Pamir
Languages (Yidgha-Munji, Sanglechi-Ishkashmi and Wakhi. Oslo: H. Aschehoug & Co.
Ollet, A. (2014). Evidence of Laryngeal Coloring in Proto-Indo-Iranian. Historische
Sprachforschung 127, 150-165.
Parpola, A. (2012). Formation of the Indo-European and Uralic (Finno-Ugric) Language
Families in the Light of Archaeology: Revised and Integrated ‘Total’ Correlations. A
Linguistic Map of Prehistoric Northern Europe. Suomalais-Ugrilaisen Seura
Toimituksia 266, 119-184.
Parpola, A. (2015). The Roots of Hinduism: The Early Aryans and the Indus Civilization.
Oxford: Oxford University Press.
Peyrot, M. (2018). Tocharian B etswe ‘mule’ and Eastern East Iranian. In L. Beek, Farnah:
Indo-Iranian and Indo-European Studies in Honor of Sasha Lubotsky (pp. 270-283).
Ann Arbor/New York: Beech Stave Press.
Pinault, G.-J. (2003a). Skt kalyāṇa- interprété à la lumière des contacts en asie centrale.
Bulletin de la socièté de linguistique de Paris 98, 123-161.
99
Pinault, G.-J. (2003b). Futher links between the Indo-Iranian substratum and the BMAC
language. Paper given at the 12th World Sanskrit Conference. Helsinki.
Pisani, V. (1981). Greco ἐναλίγκιος ed ἄντην. Indogermanische Forschungen 86, 206-211.
Pronk, T. (2011). The Saussure Effect in Indo-European Languages other than Greek. Journal
of Indo-European Studies 39, 176-193.
Pystynen, J. (2014). Palatal Unpacking in Finnic. Papers presented at the 19th Conference of
the Finno-Ugric Association of Canada.
Ravnaes, E. (1981). The Development of ə/Interconsonantal Laryngeal in Iranian. Indo-
Iranian Journal 23, 247-273.
Rédei, K. (1986). Zu den indogermanisch-uralischen Sprachkontakten. Wien: Österreichische
Akademie der Wissenschaften.
Renfrew, C. (1987). Archaeology and Language, the Puzzle of Indo-European Origins.
London: Jonathan Cape.
Salvatori, S. (2008). The Margiana Settlement Pattern from the Middle Bronze Age to the
Parthian–Sasanian: a Contribution to the Study of Complexity. In S. Salvatori, M. Tosi
& B. Cerasetti, The Bronze Age and Early Iron Age in the Margiana Lowlands: Facts
and Methodological Proposals for a Redefinition of the Research Strategies (pp. 57-74).
Oxford: Archaeopress.
Sammallahti, P. (1988). Historical Phonology of the Uralic Languages with Special Reference
to Samoyed, Ugric and Permic. In D. Sinor, The Uralic Languages: Description,
History, and Foreign Influences (pp. 478-554). Leiden: Brill.
Schindler, J. (1977). Notizen zum Sieversschen Gesetz. Die Sprache 22, 56-65.
Schirmer, B. (1998). Studien zum Wortschatz der Iguvinischen Tafeln: Die Verben des Betens
und Sprechens. Frankfurt am Main: Peter Lang Europäischer Verlag der
Wissenschaften.
Schmidt, G. (1973). Die iranischen Wörter für “Tochter” und “Vater” und die Reflexe des
interkonsonantischen H (ə) in den idg. Sprachen. Zeitschrift für vergleichende
Sprachforschung 87 (1), 36-83.
Schrijver, P. (1997). Animal, Vegetable, and Mineral: Some Western European Substratum
Words. In A. M. Lubotsky, Sound Law and Analogy (pp. 293-316). Amsterdam:
Rodopi.
Sims-Williams, N. (1998). The Iranian Languages. In A. Giacalone Ramat & P. Ramat, The
Indo-European Languages (pp. 125-153). London: Routledge.
100
Sokolova, V. S. (1967). Genetičeskie otnošenija jazguljamskogo jazyka i šugnanskoj
jazykovoj gruppy. Leningrad: Izd-vo “Nauka,” Leningradskoe otd-nie.
Spengler, R. N., Cerasetti, B., Tengberg, M., Cattani, M. & Rouse, L. M. (2014).
Agriculturalists and Pasoralists: Bronze Age Economy of the Murghab Alluvial Fan,
Southern Central Asia. Vegetation History and Archaeobotany 23, 805-820.
Szemerényi, O. (1958). Greek γἀλα and the Indo-European term for „milk“. Zeitschrift für
vergleichende Sprachforschung auf dem Gebiete der Indogermanischen Sprachen 75,
170-190.
Tadmor, U., Haspelmath, M. & Taylor, B. (2010). Borrowability and the Notion of Basic
Vocabulary. Diachronica 27:2, 226-246.
Tichy, E. (1993). Kollektiva, Genus femininum und relative Chronologie im
Indogermanischen. Historische Sprachforschung 106 (1), 1-19.
Vaan, M. (2008). Etymological Dictionary of Latin and the other Italic Languages. Leiden:
Brill.
Werba, C. H. (2005). Sanskrit duhitár- und ihre (indo-)iranischen verwandten. In G.
Schweiger, Indogermanica: Festschrift Gert Klingenschmitt (pp. 699-732). Taimering:
Schweiger VWT-Verlag.
Winter, W. (1978). The Distribution of Long and Short Vowels in Stems of the Type Lith. ésti
: vèsti : mèsti and OCS jasti : vesti : mesti in Baltic and Slavic Languages. Recent
Developments in Historical Phonology, 431-446.
Witzel, M. (1995). Early Indian History: Linguistic and Textual Parameters. In G. Erdosy,
Language, Material Culture and ethnicity. The Indo-Aryans of Ancient South Asia (pp.
85-125). Berlin: De Gruyter.
Witzel, M. (1999a). Early Sources for South Asian Substrate Languages. Mother Tongue
(special issue), 1-76.
Witzel, M. (1999b). Substrate Languages in Old Indo-Aryan (Rgvedic, Middle and Late
Vedic). Electronic Journal of Vedic Studies 5-1, 1-67.
Witzel, M. (2000). The Home of the Aryans. In A. Hintze & E. Tichy, Anusantatyai.
Festschrift für Johanna Narten (pp. 283-338). Dettelbach: J. H. Röll.
Witzel, M. (2001). Autochtonous Aryans? The Evidence from Old Indian and Iranian Texts.
Electronic Journal of Vedic Studies 7(3), 1-118.
Witzel, M. (2003). Linguistic Evidence for Cultural Exchange in Prehistoric Western Central
Asia. Sino-Platonic Papers 129, 1-70.
101
Witzel, M. (2006). Early Loan Words in Western Central Asia: Indicators of Substrate
Populations, Migrations, and Trade Relations. In V. H. Mair, Contact and Exchange in
the Ancient World (pp. 158-190). Honolulu: University of Hawai’i Press.
Witzel, M. (2009). The Linguistic History of Some Indian Domestic Plants. Journal of
BioSciences 34(6), 829-833.
Zhivlov, M. (2014). Studies in Uralic Vocalism III. Journal of Language Relationship 12,
113-148.
102
Appendix: Reference list of analyzed vocabulary
Layer Indo-Iranian Meaning #
PII I *ućig- sacrificing priest 1
PII 0 *aka- bad 2
*anću- Soma plant 3
*atHaru̯an- priest 4
*atka- cloak 5
*(H)āni- linchpin, hip 6
*bharu̯- to chew 7
*bhiš-aj̄́- healer 8
*bīj̄́a- seed, semen 9
*ćaru̯a- Name of a deity 10
*ćyā- to freeze, congeal 11
*gadā- club 12
*gr̥da- penis 13
*Hustra- camel 14
*indra- name of a God 15
*išt(i)- brick 16
*j̄́harmii̯ a- house 17
*kāća- grass 18
*kaći̯ apa- tortoise 19
*kadru- reddish brown 20
*kapāra- dish, bowl 21
*kapau̯ta- pigeon 22
*kapHa- phlegm 23
*kHā- well, source 24
*kHara- donkey 25
*kšīra- milk 26
*kućsi- ~ side of the body 27
*matsi̯ a- fish 28
*mr̥ga- wild animal 29
*muska- testicle 30
103
*nagna- bread 31
*pāpa- bad 32
*parsa- sheaf 33
*pau̯asta- cover 34
*rāći- rope 35
*ringa- mark 36
*r̥si- seer 37
*sćāga- / *sćaga- goat 38
*spāra- ploughshare 39
*stuka- / *stupa- tuft of hair 40
*u̯āćī- axe 41
*u̯and(H)- to praise 42
*u̯arāj̄́ha- boar 43
*umā-(kā)- flax 44
*u̯rīj̄́hi- rice 45
*u̯r̥tka- kidney 46
PII II *čāt(u̯āla)- pit, well 47
*i̯ au̯īi̯ ā- canal 48
*ǰaǰha/ukā̄̆- hedgehog 49
*kárus- damaged 50
*mai̯ ūkHa- peg 51
*pīi̯ ūša- beestings 52
*pusća- tail 53
Post-PII *(H)arj̄́aná-/áṇu- millet 54
*banγα- hemp 55
*ćika(tā)- sand, gravel 56
*dū̄̆rća- (goat’s) wool, hair 57
*ganDəru̯a- a mythical being 58
*ganTi- smell 59
*ganTuma- wheat 60
*Kai̯ ća- hair 61
khaḍgá-/*karkadan rhinoceros 62
104
masū̄́ra-/*mižuka- lentil 63
*mVša- bean 64
*nai̯ Ts(a)- skewer 65
*pərd-a(n)k- leopard, panther 66
*pinda- lump 67
śāli- (Skt.) unhusked rice 68
śaṇá-/*kanaba- hemp 69
sarṣapa- (Skt.) etc. mustard 70
siṁhá- (Skt.) etc. lion 71
*šu̯ai̯ pa- tail 72
*(t)sūkV̄- needle 73
*u̯īna- lute 74
Inherited *āćā- / *aćas- space, region 75
*ćan- to ascend 76
*ću̯itra- white 77
*daćā- thread, hem 78
*dhu̯aǰ- to flutter 79
*ghas- to devour 80
*ghau̯s- to make sound, hear 81
*Hat- to wander 82
*Hu̯ap- to strew, scatter 83
*Hu̯ap- to shave, shear 84
*Hu̯i̯ dhH- to split in two 85
*i̯ ātu- black magic 86
*j̄́hai- to incite 87
*ǰhas- to laugh 88
*kuč- to crook, bend 89
*magha- gift, offering, sacrifice 90
*marj̄́ha- udder 91
*monH-i- necklace 92
*nard- to hum, complain 93
*raj̄́h- to be abandoned 94
105
*sagh- to be able to bear’ 95
*srans- to fall away/apart 96
*stHūna- pillar 97
*su̯ag- to embrace 98
*u̯i̯ ak- to encompass 99
*u̯i̯ atH- to be unsteady 100
*u̯ik- to separate 101
*u̯r̥ćsa- tree 102
*u̯riH- to oppress, collapse 103
106

Traces of Pre-Indo-Iranian'

Uploaded by

Copyright:

Available Formats

Traces of Pre-Indo-Iranian'

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

Traces of Pre-Indo-Iranian'

Uploaded by

Copyright:

Available Formats

Traces of ‘Pre-Indo-Iranian’

Chronological Layers and Structural Characteristics of

Thesis submitted in partial

Leiden University Centre for Linguistics

Supervisor: Prof.dr. A.M. Lubotsky

Second Reader: Dr. A. Kloekhorst

1) Limited geographical distribution

1.3.2. Non-IE elements in Indo-Iranian

1.3.3. Indo-Iranian origins: homeland and migration

2.1. Proto-Indo-European phoneme inventory

2.3. Sound changes from PIE to PII

2.3.2.1. Consonantal laryngeals

2.3.2.2. Laryngeal vocalization

2.3.2.3. Laryngeal metathesis

2.3.2.4. Laryngeal accent shift

2.3.2.5. Loss of intervocalic laryngeals

2.3.3.1. Phonetics of the stops

2.3.3.2. Phonetics of the palatals

2.3.3.3. Stops in contact with laryngeals

2.3.3.4. Palatalization of velars

2.3.3.5. Bartholomae’s Law

2.3.4. The sibilant *s

2.3.5.1. General development

2.3.5.2. Liquid + laryngeal clusters

2.3.6.1. General development

Table 2. Relative chronology of Proto-Indo-Iranian sound changes

3.1. Loanwords I: Pre-Proto-Indo-Iranian or early Proto-Indo-Iranian

1. PII *ućig- ‘sacrificing priest’

3. PII *anću- ‘Soma plant’

Taken as a loanword by Lubotsky (2001b). No IE etymology (EWAia I, p. 37). Witzel (2003,

4. PII *atHaru̯an- ‘priest’

5. PII *atka- ‘cloak’

7. PII *bharu̯- ‘to chew, eat’

Suggested to be a loanword by Lubotsky (2001b). No IE comparanda are available (EWAia II,

8. PII *bhiš-aj̄́- ‘healer’

9. PII *bīj̄́a- ‘seed, semen’

10. PII *ćaru̯a- ‘Name of a deity’

Suggested to be a loanword by Lubotsky (2001b). No IE comparanda adduced by EWAia (II,

12. PII *gadā- ‘club’

13. PII *gr̥da- ‘penis’

14. PII *Hustra- ‘camel’

15. PII *indra- ‘name of a God’

16. PII *išt(i)- ‘brick’

18. PII *kāća- ‘grass’

19. PII *kaći̯ apa- ‘tortoise’

21. PII *kapāra- ‘bowl’

23. PII *kapHa- ‘phlegm, mucus’

24. PII *kHā- ‘well, source’

25. PII *kHara- ‘donkey’

26. PII *kšīra- ‘milk’

27. PII *kućsi- ‘round side of the body’

28. PII *matsi̯ a- ‘fish’

30. PII *muska- ‘testicle’

31. PII *nagna-

33. PII *parsa- ‘sheaf’

34. PII *pau̯asta-

35. PII *rāći-

36. PII *ringa- ‘mark’

37. PII *r̥si- ‘seer’

38. PII sćāga- / sćaga- ‘male goat’

75. PII āćā- / aćas-