Areal sound change and the distributional typology of affricate richness in Eurasia

Abstract

This paper makes a contribution to phonological typology by investigating the distribution of affricate-rich languages in Eurasia. It shows that affricate-rich and affricate-dense languages cluster areally within Eurasia and have area-specific histories. In particular, the affricate-rich areas of western Eurasia – a ‘European’ area and a Caucasian area – are not the result of contact-induced sound changes or borrowing, while the two affricate-rich areas of eastern Eurasia – the Hindukush area and the Eastern Himalayan area – are the result of contact. Specifically, affricate-dense areas depend on the emergence of retroflex affricates. Moreover, languages outside these affricate-dense areas tend to lose retroflex affricates.

Full text

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional
typology of aricate richness in Eurasia
Dmitry Nikolaev and Eitan Grossman
Hebrew University of Jerusalem
is paper makes a contribution to phonological typology by investigating the
distribution of aricate-rich languages in Eurasia. It shows that aricate-rich
and aricate-dense languages cluster areally within Eurasia and have
area-specic histories. In particular, the aricate-rich areas of western Eurasia–
a ‘European’ area and a Caucasian area– are not the result of contact-induced
sound changes or borrowing, while the two aricate-rich areas of eastern
Eurasia– the Hindukush area and the Eastern Himalayan area– are the result
of contact. Specically, aricate-dense areas depend on the emergence of ret-
roex aricates. Moreover, languages outside these aricate-dense areas tend to
lose retroex aricates.
Keywords: language typology, distributional typology, phonology, areal
linguistics, areal sound change, language contact, historical phonology,
consonant inventories, aricates
. Introduction
e goal of this paper is to investigate the distribution of Eurasian languages with
rich or dense inventories of aricate segments, as a probe for the distributional
typology (Bickel 2015) of complex consonant inventories. Distributional typology
aims to answer the question ‘what’s where why when?’ (Bickel 2007, 2015), i.e., to
account for extant linguistic diversity in the world’s languages, by accounting for
the historical processes that lead to observed distributions of linguistic properties
in space and time. On phonological typology, see the classic study by Maddieson
(1984) and, recently, Gordon (2016).
https://doi.org/./sl..nik
Studies in Language : (), –
 - / - - © John Benjamins Publishing Company

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
Aricates1 are a very common class of consonants (66.5% of languages in the
UPSID sample have one or more of them [Maddieson & Precoda 1992], most
typically /ts/), but on average they are limited in their total count in any given
language: only 22.8% of languages in the UPSID have 3 or more aricates, only
15.7% have 4 or more, and only 6.4% (29) languages have 5 or more (cf. more data
on the worldwide distribution of aricates below in §2.1.1). In contrast, 39.9% of
languages have 5 or more fricatives. is relative rarity of aricate-rich languages
makes them an interesting topic for a distributional-typological study, as we may
hope to uncover a limited set of diachronic scenarios that lead to their emergence.
ere are at least two ways of measuring the richness of segment inventories
with respect to a particular segment type. e rst is to simply count segments in a
given sound inventory. However, simple aricate counts are a rather unreliable way
of establishing the consonant richness of a language, since they are easily multiplied
by the phonologisation of secondary articulations. For example, a language can
easily phonologise palatalised or labialised allophones of pre-existing obstruents.
Furthermore, aricate inventories can be enlarged by the phonologisation of voiced
or aspirated allophones. ese processes can sometimes lead to signicant expan-
sions in the number of aricates in a given inventory, as it was considered to have
happened during the evolution of Abkhaz-Adyghe languages (Chirikba 1999); cf.
also the uctuations in the presence of voiced stops and aricates in varieties of
Chinese and Tibetan (Norman 1988; Denwood 1999).
A second, more robust measure of aricate richness, which controls for this
possible eect, is to count not the aricates themselves, but rather their places of
articulation, under the assumption that the existence of multiple places of articula-
tion is diachronically more stable than the existence of multiple aricates within a
single place of articulation. We will call a language aricate-rich if it has aricates in
at least three places of articulation.2 Moreover, we will call a language aricate-dense
if it has aricates in at least three places of articulation in the coronal-palatal range,
i.e., excluding labial, velar, and postvelar regions; the usefulness of this term will
become evident later.
. Characterised by Ladefoged and Maddieson as “stops in which the release of the constriction
is modied in such a way as to produce a more prolonged period of frication aer the release”
(1996: 90).
. However, we collapse dental and alveolar aricates if they are also distinguished by an addi-
tional articulation. us, for the Qiangic language Ersu, which opposes pure dental and alveolar
aricates, we count both places of articulation, while for Lithuanian, where dental aricates are
alveolar if they are palatalised, we collapse them into one.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
e aim of this paper is to investigate the distribution of aricate-rich and
aricate-dense languages in Eurasia based on the data contained in the Database
of Eurasian Phonological Inventories, EURPhon (Nikolaev etal. 2015).3 For the
EURPhon database, Eurasia– understood as including the Atlantic islands up to
Iceland and those along the eastern and southeastern coast of the main landmass,
but excluding the Indonesian archipelago and the Philippines– is sampled as
densely as possible. At the time of the writing of this paper, the database includes
data for 445 languages. Dialects are represented sporadically in the database, but
were excluded for the present analysis. e data are extracted from descriptions of
individual language varieties contained in grammars, phonological analyses, and
descriptive sketches published independently or as parts of overview works on
dierent language families or regions. Data from existing databases are not reused,
and no rst-hand analyses of raw data (such as wordlists and text collections) are
undertaken. e phonological entities logged in the database are phonemes, but
see Nikolaev (2018) for a detailed discussion of the precise criteria for data analysis
and collection.
In §2, we survey the geographical distribution of aricate-rich languages and
show that it has a clear and non-trivial areal structure. We also show that this
structure can be used to contextualise several claims about phonological areas in
Eurasia made in the literature, especially the Hindukush linguistic area hypothesis
(Edelman 1980, 1983; Tikkanen 1988, 1999, 2008; Liljegren 2017; Liljegren & Svärd
2017). In §3, we (i) survey the sound changes that were involved in the process of
emergence of some of these areas, (ii) investigate the role of retroex aricates in
their emergence and development, and (iii) interpret these processes in the frame-
work of the theory of contact-induced sound change (Blevins 2017). In §4, we pres-
ent evidence in favour of the hypothesis that aricate-rich inventories are unstable
outside linguistic areas that possess certain characteristics. Section5 concludes.
. Aricate richness in the languages of the world and in Eurasia
In this section, we rst provide an overview of the distribution of aricate-rich
languages in the world and in Eurasia. We then explore the aricate inventories of
Eurasian languages from dierent areal clusters. We show that of the four spatial
clusters seen on the map in Figure1, only two can be regarded as true linguistic areas4
. http://eurasianphonology.info/
. We use the term ‘(phonological) linguistic area’ in the restricted sense of a geographical re-
gion where a phonological feature is shared due to lexical borrowing or contact-induced sound
change. Cf. the discussion of the notion by Campbell (2017).

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
as far as aricate inventories are concerned. Furthermore, we demonstrate that the
overall distribution of aricate-rich languages is correlated with the distribution
of retroex aricates. e distribution of aricate-rich vis-a-vis non-aricate-rich
languages in Eurasia is shown on the map in Figure7 in AppendixA.
Placecount
Phylum
Hmong-Mien
Indo-European
Isolate
Kartvelian
Mongolic
Nakh-Daghestanian
Northwest Caucasian
Sino-Tibetan
Turkic
3
4
100 125755025
20
30
40
50
Figure1. Distribution of aricate-rich languages in Eurasia (Nikolaev etal. 2015)
. e overall distribution
.. Aricate richness in the world’s languages
Table2 presents the distribution of the number of places of articulation for aricates
computed based on the 2222 language sample pooled from EURPhon (Nikolaev
etal. 2015) and PHOIBLE (Moran etal. 2014) databases. e table shows that the
modal value is 1 (nearly half of all the cases) and that the overwhelming majority
of languages (93%) have less than 3 dierent places of articulation for aricates.
Languages having more than 3 dierent places of articulation for aricates are
extremely rare,5 and it is noteworthy that of 130 3-place ones nearly 66% (N = 81)
are located in Eurasia, while Eurasian languages comprise only around 30% of the
sample. erefore, Eurasian languages tend to be aricate rich not only in terms of
raw segment counts, but also in terms of the number of dierent places of articu-
lation for aricates. Relative frequencies of the aricates found in more than 10%
of the languages are shown in Table1. It may be noted that in 88% of the cases the
presence of an aricate at some place of articulation in an inventory coincides with
the presence of a fricative at the same place of articulation. e converse, however, is
not true: distributions of fricatives are in most cases much wider than distributions
of corresponding aricates. Eurasian languages on average have more than 4 dif-
ferent places of articulation for fricatives, and 25% of languages have more than 5.
. Both the 5-place ones, Banjun and Ronga, are African, while of 15 4-place ones, 2 are from
the Americas, 4 are from Africa, 8 are from Eurasia, and one is from the Pacic.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Table1. Relative frequencies of the most common aricates in the languages of Eurasia
Aricate tʃts dʒtsʰdz tɕ
Relative frequency 0.54 0.52 0.47 0.32 0.31 0.27
Aricate tʃʰ dʑtɕʰ ʈʂ ʈʂʰ ɖʐ
Relative frequency 0.24 0.19 0.18 0.17 0.13 0.11
Table2. Distribution of the number of dierent places of articulation for aricates
in the languages of the world
# of places of articulation 0 1 2 3 4 5
# of languages 663 961 451 130 5 2
Sources sometimes disagree on the number of places of articulation for dierent
languages. However, this does not have a signicant impact on the analysis pre-
sented below because (i) in the vast majority of cases a later description supersedes
an earlier one by providing a more precise description, and (ii) the sources never
disagree on the number of dierent places of articulation for aricates. In older
descriptions, an ambiguous notation is sometimes used (with aricates described
using characters such ‘c’ or ‘j’), but we have not encountered a case of an aricate
postulated by one source being described as a stop in another one.
.. Aricate-rich languages in Eurasia
In the EURPhon sample used in the study, 76 out of 445 are aricate rich. eir
distribution is presented in Figure1. e languages are colour-coded according to
phylum. Four areas can be discerned:
1. e European area (consisting of Standard and Zurich German, several Slavic
languages, Vlax Romani, and Gagauz)
2. e Caucasian area (languages from the Abkhaz-Adyghe, Nakh-Daghestanian,
and Kartvelian phyla)
3. e Hindukush area (Pamir Iranian and Dardic languages, Burushaski, and
one Western Tibetic variety)
4. e Eastern Himalayan area (Khams and Amdo Tibetic, Qiangic, rGyalrongic,
Qinghai-Gansu Mongolic, varieties of Mandarin Chinese, LoloBurmese, Salar,
Hmong-Njua, and possibly Daur)
It will be shown that this complex distribution of a seemingly structural prop-
erty– aricate richness– has a very strong and simple material correlate, namely,
the presence of retroex aricates in the inventory. First, however, we survey the
individual areas.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
. e European area
e aricate-rich languages of the European area in our sample are listed in Table3.6
e two German varieties in this area– Standard German and Zurich German– are
outliers: their aricate richness is due to peripheral segments (labiodental and velar
aricates), absent in all other languages from this cluster. As such, they do not meet
the criterion we have established for aricate density. In contrast, the other lan-
guages in this area are aricate-dense. Consequently, what we are probably dealing
with here is a cluster of Slavic languages of East and South Europe, augmented by
Table3. Aricate-rich languages of the European area
Phylum Language Aricate inventory
Indo-European (Germanic) Standard German Labiodental: /pf/
Denti-alveolar: /ts/
Postalveolar: /tʃ, dʒ/
Zurich German Labiodental: /pf/
Denti-alveolar: /ts/
Postalveolar: /tʃ/
Velar: /kx/
Indo-European (Slavic) Lower Sorbian Denti-alveolar: /ts/
Postalveolar: /tʃ, dʒ/
Alveolo-palatal: /tɕ, dʑ/
Standard Polish Denti-alveolar: /ts, dz/
Retroex: /ʈʂ, ɖʐ/
Alveolo-palatal: /tɕ, dʑ/
Rusyn (Lemko) Denti-alveolar: /ts, dz/
Retroex: /ʈʂ, ɖʐ/
Alveolo-palatal: /tɕ, dʑ/
Indo-European (Indo-Aryan) Vlax Romani Denti-alveolar: /ts/
Postalveolar: /tʃʰ, dʒ/
Alveolo-palatal: /tɕʰ, dʑ/
Turkic Standard Gagauz Denti-alveolar: /ts/
Postalveolar: /tʃ/
Alveolo-palatal: /dʑ/
. For a general overview of consonant inventories of European languages, cf. Stolz & Levkovych
(2017). Two additional languages that may be (arguably) analyzed as aricate-dense in the
European area are Skolt Sami and Hungarian (Juho Pystynen, p.c.). Mikhail Zhivlov points out
(p.c.) that Standard Basque and some Basque dialects might be considered aricate dense, with
three sets of aricates: dorso-alveolar (laminal), apico-alveolar, and prepalatal. However, since
we treat two of these as manners at the same place of articulation, these Basque varieties are not
technically aricate dense in our sense.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Vlax Romani and Gagauz. e status of Gagauz as a true aricate-rich language,
moreover, is dubious since the opposition in the place of articulation between /tʃ/
and /dʑ/ is augmented by that of VOT. In other words, the only clear case of possible
areal inuence here is that of several Slavic languages on Vlax Romani.
. e Caucasian area
e aricate-rich languages from the Caucasian area are shown in Figure2. e
language names and their aricate inventories are given in Table4. Several ob-
servations are immediately apparent from this table. First is the simple fact of the
prevalence of aricate richness in the languages of this area. Furthermore, unlike
the languages in the European area (see above), many of the languages in this area
have three or more aricates at the same place of articulation.
It may seem that this apparent geographical clustering lends support to a pop-
ular, albeit controversial, notion of the Caucasus as a linguistic area (for a recent
overview, see Grawunder [2017]). Indeed, all these languages have numerous places
of articulation for aricates and have ejective aricates, otherwise extremely rare
in Eurasia (to our knowledge, they are only attested in Itelmen). At the same time,
however, the picture is more nuanced, since languages from dierent families have
very dierent aricate inventories.
Starting from the west, there are the Abkhaz-Adyghe languages. All of them are
aricate-dense with no lateral aricates; two of them possess typologically unique
hissing-hushing aricates, which are described by Ladefoged and Maddieson
(1996) for the related language Ubykh as ‘laminal closed postalveolar’. Kartvelian
languages– whose aricate inventories are in fact identical– are in territorial prox-
imity to the Abkhaz-Adyghe languages, but their aricate inventories are not dense
and are rich only due to the postvelar aricate /qχ’/.
Postvelar aricates are also found in some of the Nakh-Daghestanian lan-
guages, whose common feature, however, is the presence of lateral aricates (Bats
is known to be heavily inuenced by Georgian). us, we see that each of the three
groups of languages from dierent families has its own way of achieving aricate
richness, and it is hard to argue that the specic aricate richness observed is a
contact-derived phenomenon in this region, since there is no evident segment bor-
rowing or contact-induced sound changes. is means that the sources of aricate
richness in the Caucasus are more or less unknown, since there is at present no
explanation for contact-induced preferences for large or small inventories that are
indierent to the actual segments or contrasts involved (cf. also §2.6).

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
fb
a
h
i
d
c
e
j
g
Name
a
b
c
d
e
f
g
h
i
j
Abkhaz (Bzyb)
Adyghe
Bats
Georgian
Godoberi
Kabardian (Ulyap)
Karata
Laz (Ardeşen)
Mingrelian (Zugdidi-Samurzaqano)
Tsez (Asax)
Phylum
Kartvelian
Nakh-Daghestanian
Northwest Caucasian
Figure2. Aricate-rich languages in the Caucasus (Nikolaev etal. 2015)

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Table4. Aricate inventories of the languages of the Caucasian area78
Phylum Language Aricate inventory
Abkhaz-Adyge Kabardian 7 Denti-alveolar: /ts, ts’, ts’ʷ, dz/
Postalveolar: /tʃ, tʃ’/
Alveolo-palatal: /tɕ, tɕ’, dʑ/
Abkhaz Denti-alveolar: /ts, ts’, dz/
Postalveolar: /tʃ, tʃ’/
Alveolo-palatal: /tɕ, tɕ’, dʑ/
Hissing-hushing: 8 /tŝʷ, dẑʷ/
Adyghe Denti-alveolar: /tsʰ, ts’, dz/
Postalveolar: /tʃʰ, tʃ’, dʒ/
Hissing-hushing: /tŝʰ, tŝ’, tŝʷʰ, tŝʷ’, dẑ, dẑʷ/
Retroex: /tʂʰ, tʂ’, dʐ/
Kartvelian Georgian, Mingrelian,
Laz
Denti-alveolar: /ts, ts’, dz/
Postalveolar: /tʃ, tʃ’, dʒ/
Uvular: /qχ’/
Nakh-Daghestanian Godoberi Denti-alveolar: /tsʰ, tsːʰ, ts’/
Postalveolar: /tʃʰ, tʃːʰ, tʃ’, dʒ/
Lateral: /tɬːʰ, tɬ’/
Uvular: /qχʰ/
Karata Denti-alveolar: /tsʰ, tsːʰ, ts’, tsː’/
Postalveolar: /tʃʰ, tʃːʰ, tʃ’, tʃː’, dʒ/
Lateral: /tɬːʰ, tɬ’, tɬː’/
Velar: /kxːʰ, kxː’/
Uvular: /qχː’, qχːʰ/
Tsez Denti-alveolars: /ts, ts’/
Postalveolar: /tʃ, tʃ’/
Lateral: /tɬ, tɬ’/
Bats Denti-alveolars: /tsʰ, ts’, dz/
Postalveolar: /tʃʰ, tʃ’, dʒ/
Uvular: /qχ’/
. It may be pointed out that the dialect of Kabardian spoken in Turkey retained only the
denti-alveolar aricates (Gordon & Applebaum 2006). Cf. the discussion on the loss of retroex
aricates outside specic areas below. We thank an anonymous reviewer for the reference.
. Cf. a note on their properties in the text.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
. e Hindukush linguistic area
e Hindukush seems to have been the rst linguistic area that was established
based on the aricates contained in the phonological inventories of the languages. A
representative sample of the languages from this area with their aricate inventories
is given in Table5 (the full list is given in AppendixB). e fact that languages from
dierent genera (Iranian, Dardic, Nuristani) found in this region have retroex af-
fricates, while other languages from the same genera outside this region lack them
prompted Edelman (1980, 1983) to hypothesise that this must be due to the areal
inuence of Burushaski. She also identied several lexical and grammatical features
uniting the languages of what she sometimes called the ‘Himalayan linguistic area’
or ‘Central Asiatic linguistic area’. An analysis of a fuller dataset showing the same
areal structure and the prominence of aricates and especially retroex aricates
in the Hindukush area was recently presented by Liljegren (2017).
Edelman’s argument was rened by Tikkanen (1988, 1999, 2008), who investi-
gated the distribution of dierent linguistic features in the South Asian subconti-
nent and on its borders. He noted that historical and distributional data point to the
northward spread of retroex segments, which possibly originated in South India.
erefore, Burushaski is an unlikely source for retroex aricates in the neigh-
bouring languages. It may be noted that the only argument in favour of Edelman’s
theory is the fact that, unlike for Iranian, Dardic, and Nuristani languages, we may
not be sure that proto-Burushaski did not have retroex aricates, while in other
languages their innovative character seems indisputable.
Unlike in the two previous cases, aricate inventories in the languages of the
Hindukush are absolutely uniform: they are all dense, consisting of a dental se-
ries, a postalveolar (palato-alveolar or alveolo-palatal) series, and a retroex series
(Dardic Gawarbati, situated at the centre of the area, can serve as a representative
r
l
n
q
u
f
h
e
j
o
t
k
m
i
s
d
g
p
bc
a
Name
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
u
Balti (Kharkoo)
Burushaski (Hunza)
Burushaski (Nagar)
Burushaski (Werchikwar)
Dameli
Gawarbati
Indus Kohistani (Kanyawali)
Ishkashimi
Kalam (Swat-Dir) Kohistani
Kalasha (Northern)
Kalkoti
Kati
Khowar
Munji
Palula
Shina (Gilgit)
Shumashti
Southwest Pashai
Ushojo
Wakhi
Wotapuri-Katarqalai
Phylum
Indo-European
Isolate
Sino-Tibetan
Figure3. Aricate-rich languages in the Hindukush linguistic area (Nikolaev etal. 2015)

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
example, with an /ts, tsʰ/ + /tʂ, tʂʰ/ + /tʃ, dʒ/) inventory, and there can be no doubt
of the contact origin for these commonalities (cf. the discussion of areally restricted
sound-change processes in the next section). e main question is where to look for
their origin. Edelman (1980, 1983) and Tikkanen (1988, 1999, 2008) were inclined
to look for a local source, either Burushaski or some other substrate language,
which would have disappeared aer having inuenced the mostly Indo-European
languages in the area. It is impossible to disprove such a scenario, but it is tempting
to nd another explanation that would also account for a very similar (but hitherto
unreported) aricate-rich linguistic area in the Eastern Himalayas.
Table5. Aricate inventories of the languages of the Hindukush area (sample)
Phylum Language Aricate inventory
Burushaski Hunza Burushaski Denti-alveolar: /ts, tsʰ/
Retroex: /ts, tsʰ, dz/
Alveolo-palatal: /tɕ, tɕʰ, dʑ/
Indo-European (Iranian) Wa k h i Denti-alveolar: /ts, dz/
Retroex: /ts, dz/
Postalveolar: /tʃ, dʒ/
Indo-European (Dardic) Khowar Denti-alveolar: /ts, tsʰ, dz/
Retroex: /ts, tsʰ, dz/
Alveolo-palatal: /tɕ, tɕʰ, dʑ/
Indo-European (Nuristani) Kati Denti-alveolar: /ts, tsʲ, tsʷ, dz/
Retroex: /ts, tsʷ, dz/
Postalveolar: /tʃ, tʃʷ/
Sino-Tibetan (Tibetic) Western Balti Denti-alveolar: /ts, tsʰ/
Retroex: /ts, tsʰ, dz/
Postalveolar: /tʃ, tʃʰ, dʒ/
. e Eastern Himalayan linguistic area
e Eastern Himalayan group of languages is the largest in the sample (both in
terms of number of languages and in terms of territory),9 but at the same time it
is also remarkably uniform. A representative sample of the languages from this
area and their aricate inventories is given in Table6 (the full list of languages
with their aricate inventories is given in AppendixC). Although some languages
display highly unusual inventories (cf. the already-mentioned Qiangic language
Ersu, which opposes dental and alveolar aricates: /t̪s̪, t̪s̪ʰ, d̪z̪, t̪s̪ʰ, d̪z̪/ + /ts, tsʰ,
. Several more northern Hmongic languages situated in the southern part of the area were
reported to be aricate rich, but the ambiguities in their descriptions made their inclusion in the
database undesirable.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
dz, tsʰ, dz/ + /tɕ, tɕʰ, dʑ, tɕʰ, dʑ/ + /ʈɽ, ʈɽʰ, ɖɽ, ʈɽʰ, ɖɽ/) most of them share the
‘basic’ set of dental, postalveolar, and retroex aricates, similarly to the Hindukush
languages. ere are exceptions (for example, Sangdam Tibetan: /ts, tsʰ, dz/ + /tɕ,
tɕʰ, dʑ/ + /cç, cçʰ, ɟʝ/), but all of them are aricate-dense and none involve peripheral
places of articulation.
Given the similarities between this area and the Hindukush region, it is tempting
to provide a unifying scenario for the emergence of aricate-dense inventories in
both areas, especially since we have reason to believe that the key ingredient for a
dense aricate inventory– the retroex aricate series– did not belong to any of
the proto-languages of the region (see the discussion of sound-change processes in
§3). As such, Burushaski or another substrate language from Western Himalayas is a
poor candidate for the progenitor of this areal structure as retroex-rich inventories
are conspicuously absent from the Sino-Tibetan and Indo-Aryan languages of Nepal,
Bhutan, North-West India (excluding Jammu and Kashmir) and central Tibet (cf.
Figure5; no less than 35 languages from the EURPhon database fall within this area).
A possible scenario accounting for both regions simultaneously is a general
diusion of retroex segments from South Asia, which could have led to the over-
lapping of the retroex area with two distinct areas: the Hindukush area and the
Eastern Himalayan area, both of which are characterised by large consonant inven-
tories comprising several aricates, and both of which are high-altitude ‘accretion’
zones [Nichols 1997]).
ε
η
u
δ
a
λ
r
f
βζ
ξ
i
θ
m
l
p
d
s
k
x
γ
o
h
q
n
y
v
g
z
w
j
π
b
c
ι
t
e
µ
Name
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
u
v
w
x
y
z
β
γ
δ
ε
ζ
η
θ
ι
λ
µ
ξ
π
Achang (Lianghe)
Ahi
Baima
Caodeng rGyalrong
Daur
Dongwang Tibetan
Dongxiang
Ersu
gSerpa
Hmong Njua (Green Hmong)
Japhug
Kami Tibetan
Khalong Tibetan
Kunming Chinese (Southwest Mandarin)
Labrang Tibetan
Lizu
Mangghuer (Minhe Monguor)
Melung Tibetan
Mongghul (Huzhu Monguor)
Nantong Chinese
Northern Cuona
Northern Qiang (Hongyan)
Northern Qiang (Yadu)
Nuosu (Black Yi)
Nyinpa Cone
Puxi
Rgyalthang Tibetan
Salar
Sa
ngdam Tibetan
Shigatse Tibetan
Shixing
Southern Cuona
Southern Pumi (Xiny
ingpan)
Standard Chinese (Beijing)
Themchen Tibetan
Xong
Yongning Na (Mosuo)
Zhongu Tibetan
Phylum
Hmong-Mien
Mongolic
Sino-Tibetan
Turkic
Figure4. Aricate-rich languages in the Eastern Himalayan linguistic area
(Nikolaev etal. 2015)

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Table6. Aricate inventories of the languages of the Eastern Himalayan area (sample)
Phylum Language Aricate inventory
Mongolic (Qinghai-Gansu) Monggul Denti-alveolar: /ts, tsʰ/
Retroex: /ʈʂ, ʈʂʰ/
Postalveolar: /tʃ, tʃʰ/
Sino-Tibetan (Qiangic) Northern Qiang (Yadu) Denti-alveolar: /ts, tsʰ, dz/
Retroex: /ʈʂ, ʈʂʰ, ɖʐ/
Alveolo-palatal: /tɕ, tɕʰ, dʑ/
Sino-Tibetan (rGyalrongic) Japhug Denti-alveolar: /ts, tsʰ, dz, dz/
Retroex: /ʈʂ, ʈʂʰ, ɖʐ, ɖʐ/
Alveolo-palatal: /tɕ, tɕʰ, dʑ, dʑ/
Sino-Tibetan (Lolo-Burmese) Achang Denti-alveolar: /ts, tsʰ/
Retroex: /ʈʂ, ʈʂʰ/
Alveolo-palatal: /tɕ, tɕʰ/
Hmong-Mien (Hmongic) Xong Denti-alveolar: /ts, tsʰ, ts, tsʰ/
Retroex: /ʈʂ, ʈʂʰ, ʈʂ, ʈʂʰ/
Alveolo-palatal: /tɕ, tɕʰ, tɕ, tɕ/
. Aricate-rich areas and retroex aricates
Before discussing the role of retroex aricates in the emergence of aricate-rich
areas, it is important to point out that they rarely appear in the inventories on their
own. In the pooled worldwide dataset of 2222 languages, only 3 (Eastern Khanty,
Tsou, and Pohnpeian) have only retroex aricates. Of the 113 languages with
retroex aricates, 94 also have denti-alveolar aricates, 102 also have postalveolar
or alveolo-palatal aricates, and 86 (76%) have aricates at both denti-alveolar and
post-alveolar places of articulation. Of this particular type of aricate-rich invento-
ries, 70% (60 languages) are found in Eurasia. As such, it is a very robust statistical
universal that the presence of retroex aricates implies the presence of aricates
at two additional coronal places of articulation.
On the other hand, out of 1261 languages with either postalveolar or alveolo-
palatal aricates and out of 741 languages with denti-alveolar aricates, only 505
have both types of aricates. is shows that denti-alveolar and postalveolar af-
fricates are much more likely to appear on their own, as compared to retroex
aricates, and therefore are not strongly linked to the emergence of aricate-rich
inventories.
e crucial role that retroex aricates play in the formation of the Hindukush
and Eastern Himalayan areas– and aricate-rich areas in general– is evident from
Figure5, which shows the distribution of retroex aricates in Eurasia. ey are
omnipresent in Hindukush and the Eastern Himalayas and appear both in the

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
Figure5. Distribution of retroex aricates in Eurasia (Nikolaev etal. 2015)

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Central European and Caucasian areas. ey additionally make an appearance in
several Uralic languages of the Russian Federation that have two series of aricates
in the inherited lexicon (these are the only languages that most probably inherited
retroex aricates from their proto-language, cf. §3.3.1). eir dierent densities
in the West and in the East, however, are in line with our observations that the
European and Caucasian aricate areas are very dierent from the eastern ones.
In fact, the European area is probably not a true area at all, but rather an outcome
of independent developments in Slavic languages (which are sketched in §3.3.2),
while the aricate richness of the Caucasus is not the result of borrowing of seg-
ments or of specic contact-induced sound changes. Rather, it appears to be the
outcome of a more general process of saturation of consonant inventories (which
may or may not be an areal feature in its own right) as analysed by Lindblom &
Maddieson (1988). erefore, in the subsequent discussion we will concentrate on
the Hindukush and Eastern Himalayan areas.10
. Sound-change processes giving rise to aricate-rich areas
. e Hindukush area
Even though the Hindukush area is a locus classicus for aricate-rich linguistic
areas, its sound-change history is rather poorly understood.
e situation is most dire with Burushaski, a language isolate or a group of
closely-related languages, since in the absence of comparative evidence we cannot
reconstruct its proto-history and cannot be sure even whether it inherited its nu-
merous aricates from its proto-language or, as most other languages of the region,
acquired them due to processes of sound change.
e early history of Nuristani, Iranian, and Dardic is much better known. ere
is no doubt that the denti-alveolar and postalveolar aricates in these languages
arose due to (i) historical development of the Proto-Indo-European segments that
are traditionally reconstructed as ‘palatal velars’ and (ii) processes of palatalisation
of velar and coronal stops. A full list of the relevant sound changes and of similar
sound changes in other Indo-European languages is given in Kümmel (2007). What
is le to show is how the languages in question acquired the third, viz., retroex,
. With this proviso, it is still remarkable to which extent aricate-rich languages in Eurasia
tend to come in rather tight areal clusters. is is apparently also the case in South America,
which shows the areal eects in the distribution of aricate-rich inventories, based on data from
the South American Phonological Inventory Database (Michael etal. 2015). See Figure8 in
AppendixD.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
series of aricates. e relevant collection in Kümmel (2007) is evidently incom-
plete,11 which reects a real gap in the literature.
e only existing comparative grammar of East Iranian languages (Edelman
1986) does not provide a clear overview of the process of emergence of retroex
aricates in the Iranian languages of Hindukush, although it does suggest that they
may have developed under the inuence of Burushaski.
ere is no general treatment of Dardic and Nuristani historical phonology,
but studies of the phonologies of individual languages of the region containing dia-
chronic analyses show that retroex aricates have diverse origins. e possible
sources, in addition to those listed by Kümmel (2007, cf. fn. 3), include *pr and *k(w)s
in Dardic (Konow 1936; Kogan 2015), *kr, *kl, and *ks in Nuristani (Hegedűs &
Blažek 2010; Hegedűs 2012), and aricates at other places of articulation that under-
went retroexivisation (Hegedűs 2012). Arsenault (2012) also argues that some ret-
roex aricates in Dardic may have arisen due to the process of retroex harmony.12
e phonological history of Balti, however, is straightforward in this respect.
It evidently preserved dental and alveolo-palatal aricates from a western dia-
lect of Old Tibetan and acquired some new ones by the process of palatalisation
(Backstrom 1994). Several of the Western Tibetic varieties (Eastern and Western
Balti, Purki, and Ladakhi) acquired retroex stops due to the process of simplica-
tion of consonant clusters.13 However, only one variety, Western Balti, also acquired
retroex aricates and consequently became aricate-rich. e process is described
as follows by Backstrom (1994: 49, emphasis ours):
Western Balti has recently begun undergoing a change similar, but not identical,
to the more advanced one in Ladakhi. So far, the change has only aected clusters
which began with proto velar stops, not the labials or dentals. Furthermore, instead
of producing retroexed stops, the change in Western Balti is producing retroexed
aricates. Whether this is only a stage toward stop production, or a more perma-
nent arrangement, only time will tell. Areal linguistic features, however, seem to lend
strength to the idea that this development may indeed be dierent, not only in its stage
of advancement, but also in kind, from that seen in Ladakhi. Baltistan, especially
. Only the following sound changes are listed (‘ṭ’ and ‘ṣ’ in Kümmel’s notation correspond to
/ʈ/ and /ʂ/): Iranian: *ṭʃ > ṭṣ /_ Wakhi; Nuristani: *ts > ṭṣ / _r>0 Common Nuristani; Dardic: ṣ >
ṭṣh /#_w Khowar; ṣṭ > ṭṣ /_ Bashkarik.
. I.e., sound change of the form tɕ, tʃ > ʈʂ / _Vʂ
. Cf. the typical sound changes:
pʰ.r, kʰ.r > ʈʰ / C,#V
p.r, k.r > ʈ / C,#_V
b.r, d.r > ʈ / #_V (or ɖ / #_V in the Western and Amdo varieties without devoicing)
b.r > ɖ / C_V

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
western Baltistan, is partially surrounded and, to a considerable extent, inltrated by
speakers of Shina, a Dardic (Indo-Aryan) language which, like several other languages
of northern Pakistan, have an abundance of retroexed sibilants and aricates. is
areal feature could well have provided the pressure which is currently inuencing
the development of similar aricates in Western Balti.
e analysis of these sound changes shows that if denti-alveolar and postalveolar
aricates can arise from simple stops that underwent palatalisation, retroex af-
fricates in the absolute majority of cases arise from consonant clusters (especially
clusters with rhotics) or other aricates and are not the rst aricates in their re-
spective inventories. erefore, it may be hypothesised that presence of consonant
clusters and/or aricates in the inventory is a prerequisite for the emergence of
retroex aricates, which explains their strong link with aricate-rich inventories
both in Eurasia (cf. the map in Figure5) and South America (cf. the map and the
list of aricate-rich inventories in AppendixD).
e areal clustering of retroex aricates may be interpreted in the framework
of the Perceptual Magnet framework advanced by Blevins (2017). is framework
explains the appearance of segments previously absent from a given inventory as a
result of sound-change processes guided towards a particular end by a ‘magnet’–
a particular segment present in the neighbouring language or languages, present
in the pool of synchronic variation for production of particular sound sequences,
and winning the competition due to its being already established in the phonetic or
even phonological lexicon of the speakers. Due to the acoustic similarity between
fricatives and aricates it does not really matter in this case whether retroex af-
fricates are endogenous to the region or were brought to life by the presence of
retroex fricatives in some early Indo-Aryan variety. However, the dierence be-
tween Western Balti and other Western Tibetan varieties suggests that the presence
of retroex aricates as magnets clearly plays a role.
Unfortunately, there are no historical data on particular cases of contact-induced
sound changes establishing themselves in speech of individual speakers and then
propagating through the population. However, the Perceptual Magnet framework
seems to be the best way to account for spatial clustering of similar sound changes
and sound-system convergence in the absence of largescale lexical-borrowing
events, which were not recorded for the regions under discussion.
. e Eastern Himalayan area
is phonological history of the Eastern Himalayan languages is not completely
understood, but much ground has been gained in the recent years. Ongoing work
on Proto-rGyalrongic seems to indicate that denti-alveolar and alveolop alatal

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
aricates were inherited while retroex aricates stem from clusters with rhotics
(Guillaume Jacques, p.c.).
ere is also no reconstruction of Proto-Qiangic (this is a problematic linguis-
tic grouping overall, and as Evans [2004] puts it, ‘[for Qiangic] it has been (and con-
tinues to be) dicult or impossible to establish regular sound correspondences…’),
but preliminary work on some languages from this genus (Yu 2009) points to simi-
lar conclusions: ‘Retroexes in Lizu come mostly from two sources: complex velar
clusters [velar + l/r/j], and *sr clusters.’ ere is also one word exemplifying the
change *bl > dʐ / m V. Dental and postalveolar aricates partly arose due to pala-
talisation, but at least some of both kinds were inherited from the proto-language.
Proto-Lolo-Burmese (the proto-language of Achang, Ahi, Nuosu, and possibly
Yongning Na)14 is reconstructed with only denti-alveolar aricates (Bradley 1979).
Matiso (2003: 21) notes that retroex aricates in contemporary languages “seem
to be secondarily derived from proto-clusters with medial liquids”, but does not
give concrete examples. A change *rts(h) > ʈʂ(h) was proposed for Yongning Na
by Jacques and Michaud (2011). It may be noted that Lolo-Burmese languages
from outside the area of interest for the most part collapsed C+l and C+r clusters
in one series or even merge them with C+j clusters without producing retroex
aricates (Bradley 1979; Matiso 2003). e furthest ones– Rangoon and Myeik
Burmese– have only one series of aricates. As will be shown in the next section,
some Lolo-Burmese languages also show a tendency to lose retroex aricates aer
acquiring them.
e aricate history of Chinese consists of three stages (Baxter 1992):
1. At the Old Chinese stage, the language is thought to have had only dentialveolar
aricates.
2. Middle Chinese acquired (i) palatal aricates from the palatalisation of velar
and dental stops and (ii) retroex aricates (from clusters consisting of coronals
and rhotics).
3. e Middle Chinese aricate inventory was simplied in dierent ways, as seen
in dierent contemporary varieties; these processes will be described in more
detail in the next section.
erefore, it seems that Middle Chinese (the sources for which are dated to the
late 6th-early 7th c.) is the earliest documented case of the presence of retroex
aricates– or any retroex segments at all– in the region. Based on lexical and
grammatical similarities, it is oen assumed that as an early stage Chinese was
in contact with ‘Altaic’ or ‘Transeurasian’ languages in the north-west and with
. Which some scholars classify as Qiangic (Jacques & Michaud 2011).

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Hmong-Mien and possibly Mon-Khmer languages in the south (Starostin 2008;
Niederer 1998; LaPolla 2001). On Turkic, Mongolic, and Hmong-Mien see below;
Proto-Tungusic (Tsintsius 1949),15 and Proto-Mon-Khmer (Shorto etal. 2006) are
reconstructed without retroex segments.
e phonological history of the Eastern Himalayan Tibetic varieties is relatively
well known due to our knowledge of their proto-language (Hill 2010) and intensive
eldwork conducted in the region in the last two decades (Hongladarom 1996;
Makley etal. 1999; Sun 2003; Haller 2004; Huber 2005; Sun 2006; Bartee 2007;
Chirkova 2008; Suzuki & mTshomo 2009; Suzuki 2009; Chirkova 2010; Jacques
2011). Two main types of sound change that account for the emergence of retroex
aricates are recorded in the literature:
1. Retroex aricates arising from clusters with rhotics:
– k(ʰ)r > ʈʂ(ʰ) (emchen, rGyalthang, Kami)
– gr > ʈʂ (emchen, Kami, Zhongu)
– gr > ɖʐ16 (emchen, Kami, Zhongu)
– p(ʰ)r > ʈʂ(ʰ) (emchen, Kami, Zhongu)
– dr > ʈʂ (emchen, Kami, Zhongu)
– dr > ɖʐ (emchen, Kami, Zhongu)
2. Retroex aricates arising from alveolo-palatal aricates before non-front
vowels:
– tɕ(ʰ) > ʈʂ(ʰ) (rGyalthang, Melung)
– dɕ > ɖʐ (rGyalthang, Melung)
us, the Eastern Himalayan Tibetic varieties developed retroex aricates from
clusters with rhotics or by splitting the alveolo-palatal aricate series. e dier-
ence between eastern and more western Tibetic varieties is therefore the same as
between Western Balti and more eastern varieties: while the central Tibetic dialects
mostly changed the C+r clusters into retroex stops, those peripheral varieties that
did not retain them changed them into retroex aricates, presumably under areal
pressure. Melung Tibetan presents an interesting case since it aligns with central
Tibetic dialects in converting C+r clusters into retroex stops but nevertheless
acquired retroex aricates through conditioned shi of alveolo-palatal aricates
(cf. the discussion of the Mongolic data below).
. Some Tungusic languages of China, such as Kilen (Zhang 2013) acquired a series of retroex
aricates under the inuence of Mandarin Chinese.
. In most Tibetic varieties, the reexes of voiced obstruents in syllable initials are split accord-
ing to whether they were also word-initial (in which case they usually became voiceless with
the concomitant development of a low tone on the following vowel) or preceded by another
consonant.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
It is important to point out that, to our knowledge, sound changes transforming
stop-plus-rhotic clusters into retroex aricates (or any other aricates) have not
been recorded outside of the Hindukush and the Eastern Himalayan areas.
Earlier reconstructions of Proto-Hmong-Mien (Niederer 1998) included ret-
roex aricates in the inventory of proto-initials. However, in a more recent analysis
(Ratli 2010), they are also derived from clusters with rhotics and aricates at other
places of articulation.
Discussing the phonological history of Qinghai-Gansu Mongolic languages,
Nugteren (2011: 218) notes that “under the inuence of Chinese and/or Tibetan,
many QG languages have by now developed two or three sets of aricates”. e
processes leading to this were diverse. In Mongghul, the proto-Mongolic ari-
cates *č and *ǰ (characterised as ‘palatal’ by Janhunen [2003]) mostly became /tʃh/
and /tʃ/ with retroex aricates coming from borrowings. In Dongxian, *č and *ǰ
became /ʈʂʰ/ and /ʈʂ/, while /tʃh/ and /tʃ/ developed due to palatalisation of *t and
*d. In Mangghuer, *č and *ǰ split into /tʃh, tʃ/ and /ʈʂʰ, ʈʂ/ partly conditioned by the
subsequent vowels, with front vowels favouring postalveolar aricates. Finally, the
QG languages developed alveolar aricates by way of strengthening of *s in initial
and intervocalic positions, as well as acquiring them from borrowings.
e presence of retroex aricates in the Turkic language Salar is explained by
Dwyer (2007: 267–268) as a result of loanword incorporation:
e retroex series, now phonemic to Salar, was borrowed from Northwestern
Chinese (also with Amdo Tibetan influence) within the last century. …
Northwestern Chinese and Amdo Tibetan both distinguish three series of spirants:
apical/alveolar, palatal, and retroex. Salar already has a partial alveolar series (/s,
z/ but not /ts, dz, ʃ, tʃ/) and a full palatal series; since at least 15–25% of Salar’s vo-
cabulary consists of Chinese and Tibetan loans, it is unsurprising that the retroex
series has also become phonemic.17
Data from Salar and Qinghai-Gansu Mongolic languages show that, unlike lan-
guages from other families in the Eastern Himalayan aricate-rich zone, they de-
veloped retroex aricates exclusively from other aricates and/or acquired them
from lexical borrowings (which is logical given the aversion of Altaic/Transeurasian
languages to initial consonant clusters [Robbeets 2017]).
. It may be seen from this remark that the variety described by Dwyer diers from that de-
scribed by Tenishev (1976), which included a full alveolar series, including aricates, and was
used for our sample.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
. Retroex aricates outside the Hindukush and the Eastern Himalayan
area: Uralic and Slavic languages
In §2.6, we discussed the fact that the distribution of aricate-rich inventories in
Eurasia is correlated with that of retroex aricates (shown in Figure5), and our
discussion of the pathways to aricate richness would be incomplete if we did not
take into account the areal cluster of languages with retroex aricates outside of
the aricate-rich areas. We look at this in §3.3.1. It is also necessary to investigate
the emergence of aricate richness in Slavic languages in more detail (§3.3.2).
.. Uralic
e big outlying cluster of languages with retroex aricates comprises Finno-
Ugric languages (Komi-Zyrian, Komi-Yodzyak, Komi-Permyak, Udmurt, Eastern
Mari, Erzya), and Proto-Uralic has been reconstructed with a retroex aricate *ʈʂ
(Janhunen 1982).18 Some of these languages may have innovated an alveolo-palatal
aricate series, and due to recent borrowings from Russian (whose aricate inven-
tory consists of /t̪s̪/ and /tɕ/) acquired the dental aricate /t̪s̪/. ese borrowings
may potentially lead to another aricate-rich area with a reversed origin history:
(i)one original series of retroex aricates, (ii) an additional series developed due
to internal processes, and (iii) the areally-imposed dental aricate. It is also pos-
sible, however, that their aricate inventories will become even more Russianised
and will consist of a denti-alveolar and an alveolo-palatal series, as happened with
Standard Moksha and Hill Mari.
.. Slavic
Similarly to the Indo-Iranian languages, the Slavic languages entered the historical
era with denti-alveolar and postalveolar aricates resulting from what is tradi-
tionally called in the literature the rst, second, and third palatalisations of velars
(Carlton 1991). e subsequent history of individual languages included numerous
additional palatalisation processes, which for the most part did not result in the
emergence of additional places of articulation. However, there are several excep-
tions, of which Polish and Serbo-Croatian are representative examples.
Polish innovated an additional series of alveolo-palatal aricates /tɕ/ and /dʑ/
as a result of palatalisation of /t/ and /d/. By comparison, in contemporary Russian
palatalised /tj/ and /dj/ either retained their additional articulation or, in the Moscow
. We thank Ante Aikio for pointing out to us alternative views on the reconstruction of Proto-
Uralic alveolo-palatal aricates. In particular, traditional reconstructions contain an alveolo-
palatal aricate, which is disputed by Janhunen (1988). More recently, Zhivlov (2014) has ana-
lyzed the traditionally reconstructed alveolo-palatal sibilant as an aricate at the same place.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
dialect, started shiing towards /ts/ and /dz/. Aer that, according to Hamann
(2004), following the analysis proposed by Hall (1997) for early Indo-Aryan, the
Polish aricate inventory consisting of /ts/, /tʃ/, /tɕ/, and their voiced counterparts
became oversaturated due to the high acoustic similarity; consequently, postalve-
olars and alveolo-palatals and postalveolars unconditionally shied to more ‘grave
sounding’ retroexes (the original analysis was proposed for fricative segments,
but it naturally carries over to aricates). It must be noted that the analysis by Hall
(1997) was predicated on the thesis that it is impossible to have both /ʃ, ɕ/ in an in-
ve ntory,19 which is patently wrong (EURPhon has 15 languages with this combina-
tion, including Nantong Chinese, Standard Tatar, Ter Saami, and Luxembourgish).
However, a relaxed version of this hypothesis (the combination of palato-alveolars
and alveolo-palatals is likely to be unstable) seems tenable, albeit in need of addi-
tional testing.20
The scenario in Serbo-Croatian is very similar except for the source of
alveolo-palatals (they are supposed to stem from palatal stops /c/ and /ɟ/) and the
fact that the shi of palato-alveolars to a more ‘grave’ pronunciation was achieved
not by means of retroexion, but rather of lip rounding. e resulting acoustic sim-
ilarity of Serbo-Croatian labialised palato-alveolar aricates and retroex aricates
has led to a discussion in the literature and the competing analyses include /tʃ, dʒ/
vs. /tʃʷ, dʒʷ/, /tʃʷ, dʒʷ/ vs. /tɕ, dʑ/, and /ʈʂ, ɖʐ/ vs. /tɕ, dʑ/ (Petrović & Gudurić 2010;
Morén 2006).
ese data corroborate the thesis that retroex aricates are somewhat ‘marked’
in that they are usually latecomers to aricate inventories. ey also show that it is
possible for languages to acquire retroex fricative and aricate series without any
kind of areal pressure. As Lower Sorbian and Vlax Romani demonstrate, however, it
is equally possible to remain at the stage of a superdense palatal region (both these
languages contrast /tʃ, dʒ/ with /tɕ, dʑ/), and consequently, it seems that retroex
aricates are much more likely to emerge in a contact scenario.
. “One generalization that I assume to be exceptionless is summarized in (3):
(3) No language can contrast palatoalveolars and alveolopalatals.” (Hall 1997: 205)
. Cf. the theoretical discussion of the optimal structure of fricative inventories in (Boersma &
Hamann 2007).

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
. Languages outside the Hindukush and the Eastern Himalayan linguistic
areas tend to lose retroex aricates
e data on the spontaneous emergence of retroex aricates in Slavic (and pre-
sumably in Uralic) suggests that the observed areal patterning has two aspects. e
rst is positive, so to speak: retroex aricates tend to spread horizontally within
inventories that already have aricates. e second is negative: retroex aricates
are predicted to be unstable without areal support. Data for the total disappearance
of certain types of segments is understandably scarce, as there is usually no evidence
to suggest they were there in the rst place, but what data we do have suggests that,
unlike coronals– which are easily acquired and preserved in sound inventories, and
which are relatively hard to lose (Blevins 2009)– retroex aricates are both hard
to acquire and hard to preserve. More precisely, retroex aricates tend to merge
with aricates at other places of articulation.21
. Croatian
In some varieties of Serbo-Croatian spoken in Croatia and Bosnia, the tripartite
system /ts, ʈʂ, tɕ/ was simplied into a bipartite one: /ts, tɕ/. A phonetic analysis
of this merger and a theoretical interpretation in the framework of Functional
Optimality eory was presented by Ćavar (2011).
. Uralic
Although the retroex aricate *ʈʂ (Janhunen 1982) is condently reconstructed
for Proto-Uralic, it was lost in the Finnic, Hungarian, and Saamic branches and
several languages of the Samoyedic branch, that is, nearly everywhere except for
a small region west of the Urals, where its preservation may be due to mutual
. We have not systematically studied dearication processes in the languages of Eurasia, but
there is evidence for processes turning aricates into stops (as in Japonic and in the Southern
Tungusic language Uilta spoken on Sakhalin island) or fricatives (as in Burmese and Yanghao
Hmong). Numerous examples are collected in Kümmel (2007: 67–75). Absent a systematic study
of dearication, we take these examples as possible and very tentative evidence of the relative
instability of aricates in general, although given further data and analysis, it may turn out that
such dearication processes themselves show areal eects, as perhaps suggested by the relative
proximity of, e.g., Japonic and Uilta, on the one hand, and Yanghao Hmong, Burmese, and
Yunnan Sinitic varieties, on the other (see §4.4 below). We thank Guillaume Jacques and José
Andrés Alfonso de la Fuente for discussing these issues with us.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
areal support between languages of three dierent branches of the family (Ugric,
Mordvinic, Permic, and Mari).
. Lisu
e most conservative Central Lisu varieties have retroex aricates in three places
of articulation (a plain voiceless series /ts, ʈʂ, tɕ/ is accompanied by a voiceless as-
pirated and a voiced one). In Southern Lisu, they merged with alveolar fricatives
and aricates, while in Northern Lisu they became allophones of alveolo-palatal
fricatives and aricates (Bradley 2017).
. Chinese
e most extensive data on the loss of retroex aricates is provided by varieties
of Chinese, both at the level of large dialect groups (some of which lost retroex
aricates completely) and at the level of the varieties inside the rest of the groups.
We will give an overview of the distribution of retroex aricates among the major
dialect groups, and then turn to a more detailed description of several varieties of
Mandarin Chinese. e process of retroex-aricate loss in Chinese was in many
cases a part of a more general process of loss of all retroex segments, but at the
moment there is not enough data to estimate the relationship between these two
processes.
.. Loss of retroex aricates in non-Mandarin varieties
Retroex aricates were completely lost in Wu (Simmons 1999), Min22 (Norman
1974), Hakka (except for the dialect of Sandu), and Yue (Yan 2006), and were lost
in some varieties of Xiang, Gan, and Ping (Yan 2006). e map of the distribution
of major Chinese dialect groups reproduced in Figure623 shows that retroex af-
fricates progressively disappear along the north-west– southeast axis, that is, as we
leave the area close to the Eastern Himalayas and approach the South China Sea.
. e Min dialects are considered to contain a pre-Middle-Chinese lexical layer, which com-
plicates their phonological history (Norman 1991).
. e status of Jin as a major dialect group is disputed (Kurpaska 2010), and it patterns with
Mandarin in its inventory of initials.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Figure6. Geographical distribution of major Chinese dialect groups
(Wikimedia Commons)
.. Loss of retroex aricates in the varieties of Mandarin Chinese
Proto-Mandarin, the ancestor language of the contemporary Mandarin varieties,
is considered to have retained Middle Chinese retroex aricates (Baxter 1999,
2000). e Beijing variety of Mandarin Chinese, the most widely known dialect
from this group, which became the basis for the standard language, not only re-
tained Middle Chinese retroex aricates, but also saw some of the previously
palato-alveolar aricates and retroex stops become retroex aricates24 as well
. It is oen remarked in the literature that Beijing Mandarin retroexes are not ‘real’ retroexes,
but rather ‘apical post-alveolar’ sounds (Lee & Zee 2003). However, this characterisation falls under
the broad denition of retroex segments as ‘bunched’ non-laminal postalveolars, which is actually
used in practice to describe consonant inventories of languages outside South India where ‘real’
tongue-curled retroexes are abundant (Hamann 2003; Ladefoged & Maddieson 1996).

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
(Chen 1976). However, many other, especially southern, varieties of Mandarin are
known to have merged retroex aricates with alveolar aricates (Norman 1988).
Two representative case-studies are the Mandarin varieties of the southern province
of Yunnan and the partial or complete loss of retroex aricates in Xi’an, Hefei,
and Yangzhou as analysed by Baxter (Baxter 1999).
... Retroex initials in the Mandarin varieties of Yunnan
Gui (2001) analysed the geographical distribution of retroex aricates in the
Yunnan varieties of Southwest Mandarin Chinese and traced their history in
the dialect of Kunming, the provincial capital. Diachronically, it is interesting to
note that the transition between the so-called ‘Old Kunming Chinese’ and ‘Con-
temporary Kunming Chinese’ that, among other things, involved merger of retro-
ex aricates with the alveolar ones, took place relatively recently– in the second
half of the 20th century.
e geographical distribution of retroex-less varieties in Yunnan shows a clear
centre-vs-periphery pattern. e map in Gui (2001: 42) shows that the varieties that
retained retroex aricates are concentrated in the middle part of the province,
while the varieties on the periphery, spoken in the regions bordering on Burma,
Laos, Vietnam, and Guangxi and Guizhou provinces– and thus in contact with
retroex-less Sinitic and non-Sinitic languages– lost them.
e recent phonological shi in the speech of the provincial capital, which is
situated in the eastern part of the central region, can be explained by the inux
of immigrants from peripheral regions speaking non-Sinitic languages or more
deviant varieties of Mandarin during the 20th century.
... e fate of retroex aricates in Xi’an, Hefei, and Yangzhou
In his paper on the reconstruction of retroex aricates in Proto-Mandarin, Baxter
(1999) reconstructs, among other processes, partial or complete loss of retroex
aricates in several Mandarin varieties.
1. Xi’an, a variety of Central Plains Mandarin, “was aected by… retroex loss
before *ə, *a, and the reex of original *u” (Baxter 1999: 28).
2. Hefei, a variety of Lower Yangtze Mandarin (a dialect group situated further
to the east and inuenced by Wu dialects), was aected by the same processes
of retroex loss, but it had not previously underwent the change i > ə, which
aected northern dialects, and therefore had fewer contexts for this process.
3. Finally, Yangzhou, another Lower Yangtze Mandarin variety situated even fur-
ther to the east from Hefei, underwent retroex loss in all environments.
It may be added that the dialect of Jinan, a Jilu Mandarin (previously known as
Northern Mandarin) variety, situated not far from Beijing, retained retroex af-
fricates in all environments. erefore, we again witness a north-west– southeast

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
cline (Jinan– Xi’an+Hefei– Yangzhou) along which dialects progressively lose
aricates, although on a smaller scale (it must be noted that the centre of gravity
for the retroex initials in this case is not only the Eastern Himalayan region, but
also the inuential Beijing Mandarin area).
.. Retroex aricates in Chinese, an overview
In order to have a more complete understanding of the processes of the develop-
ments of retroex segments in Chinese dialects, we need more diachronic analyses,
but the data presented above point to two cycles of retroex loss in the history of
Chinese:
1. Early Post-Middle-Chinese dialects groupings situated further to the southeast
from the Eastern Himalayas and the centre of gravity of Middle Chinese in the
Central Plains lost retroex initials, either at the level of local proto-languages,
or later as a result of an areally-spreading sound change.
2. Aer the secondary southward spread of the speakers of initially retroex-rich
Mandarin Chinese varieties, southern varieties of Mandarin, which came into
contact with retroex-less non-Sinitic languages and dialects of Chinese that
had already lost retroex initials by that time, also started losing them, a process
that seems to be still ongoing.
. Conclusion
e analysis of the distribution of languages with aricates at three or more places
of articulation shows that there two general pathways that lead to their emergence.
Languages can either (i) develop aricates at peripheral places of articulation:
labiodental and/or velar and post-velar (Standard and Swiss German, Kartvelian
languages); or (ii) develop aricate-dense inventories with three or more places of
articulation in the coronal-palatal range. ree additional observations can be made:
1. In the vast majority of cases, languages follow only one of these pathways.
2. e pathway involving aricate-dense inventories is much more widespread.
3. Aricate-dense inventories are extremely prone to contain retroex aricates,
which makes the latter a very strong correlate of aricate richness in general.
e analysis of diachronic pathways leading to aricate-rich inventories show that
the emergence of retroex aricates is in the majority of cases the last step towards
aricate denseness: retroex aricates are diachronically ‘marked’ in that they usu-
ally appear in inventories that already have aricates at two or more dierent places
of articulation. Two scenarios were identied.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
In the case of several Slavic languages and possibly some early Indo-Aryan
varieties, tripartite fricative and aricate systems with a super-dense palatal region
/s, ʃ, ɕ, ts, tʃ, tɕ/ saw the acoustically-based dissimilation of their middle elements,
which led to the emergence of a more ‘grave’-sounding retroex fricative and af-
fricate series /s, ʂ, ɕ, ts, ʈʂ, tɕ/.
In many other cases, however, the change was not spurred by any systemic fac-
tor and seems to have a purely areal basis. Numerous languages in the Hindukush
and the Eastern Himalayan linguistic areas innovated a series of retroex aricates
starting from a ‘maximally dispersed’ system consisting of /ts/ and /tɕ/. e sound
changes involved in this process (simplication of consonant clusters, especially
those with rhotics, yielding retroex aricates and retroexivisation of /tɕ/) are not
attested outside these regions.
Retroex aricates, therefore, provide some of the most compelling evidence
to date of areally-induced sound change. Compared to the relatively few cases of
the internally-motivated appearance of retroex aricates in Slavic languages, there
is an impressive array of genera (and even phyla) whose languages underwent
certain processes of sound-change that led to the emergence of retroex aricates
exclusively inside the Hindukush and the Eastern Himalayan regions.25 us, our
conclusions support the position of Edelman (1980), Tikkanen (1988, 1999, 2008),
and Liljegren (2017), who argued that Hindukush is a bona de linguistic area.
Finally, we were able to show, although only in a preliminary fashion, that lan-
guages outside of areas where retroex aricates are common tend to lose them.
is thesis is corroborated by the history of Croatian and Bosnian varieties of
Serbo-Croatian, most branches of the Uralic family, several varieties of Lisu, and a
large number of Chinese varieties situated to the south and south-east of the core
retroex areas in the Eastern Himalayas and North China Plain.
us, it is highly probable that aricate-rich areas not only facilitate the emer-
gence of retroex aricates in the nearby languages, but also help retain them.
It is an interesting open question whether aricate richness or aricate density
as abstract features of inventories can also be areally induced or strengthened.
For instance, in the Eastern Himalayan area, there are several languages, such as
Sangdam Tibetan, that are aricate-dense even though they do not have retroex
aricates. e processes leading to the emergence of such inventories are unclear.
Even more problematic is the situation in the Caucasus where no less than three
types of aricate richness are attested, and a general trend toward oversaturated
. However, we still do not know the origins of retroex aricates in Burushaski (inside the
Hindukush area) and of the voiceless *ʈʂ reconstructed for Proto-Uralic, which is responsible for
the single pocket of non-aricate-rich languages with retroex aricates outside the Himalayas.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
consonant inventories is evident (Grawunder 2017), but both facts at present lack
any convincing areal interpretation.
More broadly, this paper has been a probe for distributional typology in the
domain of phonology, asking ‘what’s where why when?’. e property of interest,
aricate-rich and aricate-dense phonological inventories, has been shown to have
dierent subtypes, to be geographically limited to a number of areas within the
Eurasian macro-area, and to have developed through a limited number of pathways.
is last in fact constitutes an area-specic diachronic typology of sound inventory
complexication in a particular domain. Importantly, while the strong areal signal
points to a contact-based explanation (‘event-based’, in Bickel’s [2015] terminol-
ogy), a ner-grained examination leads to the conclusion that these areas are not
equally amenable to the same type of contact-based explanation. As a result, we
conclude that distributional typologies can and should address the actual histories
of languages and their speakers.
Of course, it remains to be seen to what extent these results can be generalized
beyond Eurasia to macro-areas with aricate-rich and aricate-dense inventories
(for instance, they also show a spatially clustered distribution in South America),
a question we leave to specialists in these areas. In essence, however, such dense
areal studies are necessary in order to provide an empirical basis for teasing apart
the relative contributions of inheritance, language contact, and sound change to
the observed diversity (and uniformity) of sound systems in the world’s languages.
Acknowledgements
We are grateful to Ante Aikio, Nino Amiridze, Marko Crnobrnja, José Andrés Alfonso de la
Fuente, Guillaume Jacques, Peter Piistanen, Juho Pystynen, Willem Vermeer, Rémy Videraz, eo
Yeh, and Mikhail Zhivlov for their comments on a dra of this paper. We also thank the audience
at the 2017 (Zurich) meeting of the Societas Linguistica Europaea, the members of the Dynamics
of Language Lab of the Hebrew University of Jerusalem, and the members of the Israel Phonology
Circle, in particular Outi Bat-El, Evan-Gary Cohen, and Hadas Yeverachyahu.
References
Arsenault, Paul Edmond. 2012. Retroex consonant harmony in South Asia. Toronto: University
of Toronto dissertation.
Backstrom, Peter C. 1994. A phonological reconstruction of Proto-Western Tibetan. Arlington:
University of Texas at Arlington MA thesis.
Bartee, Ellen Lynn. 2007. A grammar of Dongwang Tibetan. Santa Barbara: University of Cali-
fornia, Santa Barbara dissertation.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
Baxter, William Hubbard. 1992. A handbook of Old Chinese phonology (Trends in Linguistics 64).
Berlin/New York: Mouton de Gruyter. https://doi.org/10.1515/9783110857085
Baxter, William Hubbard. 1999. Reconstructing Proto-‘Mandarin’ retroex initials. In Richard
VanNess Simmons (ed.), Issues in Chinese dialect description and classication (Journal of
Chinese linguistics: Monograph series 15), 1–35. Berkeley, California: Project on Linguistic
Analysis.
Baxter, William Hubbard. 2000. Did Proto-Mandarin exist? Journal of Chinese Linguistics 28(1).
100–115.
Bickel, Balthasar. 2007. Typology in the 21st century: Major current developments. Linguistic
Typology 11(1). 239–251. https://doi.org/10.1515/LINGTY.2007.018
Bickel, Balthasar. 2015. Distributional typology. In Bernd Heine & Heiko Narrog (eds.), e
Oxford handbook of linguistic analysis, 901–923. Oxford University Press.
Blevins, Juliette. 2009. Another universal bites the dust: Northwest Mekeo lacks coronal pho-
nemes. Oceanic Linguistics 48(1). 264–273. https://doi.org/10.1353/ol.0.0033
Blevins, Juliette. 2017. Areal sound patterns: From perceptual magnets to stone soup. In
Raymond Hickey (ed.), e Cambridge handbook of areal linguistics, 88–121. Cambridge
University Press. https://doi.org/10.1017/9781107279872.006
Boersma, Paul & Silke Hamann. 2007. e evolution of auditory contrast (unpublished manu-
script). http://www.fon.hum.uva.nl/paul/papers/ EvolutionOfContrast.pdf.
Bradley, David. 1979. Proto-Loloish (Scandinavian Institute of Asian Studies monograph series
no. 39). London: Curzon Press.
Bradley, David. 2017. Lisu. In R. P. E. Sybesma, Wolfgang Behr, Yueguo Gu, Zev J. Handel,
Cheng-Teh James Huang & James Myers (eds.), Encyclopedia of Chinese language and lin-
guistics, Leiden/Boston: Brill.
Campbell, Lyle. 2017. Why is it so hard to dene a linguistic area? In Raymond Hickey (ed.), e
Cambridge handbook of areal linguistics, 19–39. Cambridge: Cambridge University Press.
https://doi.org/10.1017/9781107279872.003
Carlton, Terence R. 1991. Introduction to the phonological history of the Slavic languages.
Columbus, Ohio: Slavica Publishers.
Ćavar, Malgorzata. 2011. Merger of the place contrast in the posterior sibilants in Croatian (un-
published manuscript).
Chen, Matthew Y. 1976. From Middle Chinese to Modern Peking. Journal of Chinese Linguistics
4(2/3). 113–277.
Chirikba, Viacheslav. 1999. e West Caucasian material in “e North Caucasian etymological
dictionary” by S. L. Nikolayev and S. A. Starostin. In Helma van den Berg (ed.), Studies in
Caucasian linguistics, 152–170. Leiden: CNWS Publications.
Chirkova, Ekaterina. 2014. Phonological prole of Kami, the Tibetan dialect of Muli (manu-
script). In Jackson T. -S. Sun (ed.), Special issue of Language and Linguistics on lesser known
dialects of Tibetan. In press.
Chirkova, Katia. 2008. On the position of Baima within Tibetan: A look from basic vocabulary.
In Alexander Lubotsky & Frederik Kortlandt (eds.), Evidence and counter-evidence: Essays
in honour of Frederik Kortlandt. Volume 2: General linguistics, 69–91. Amsterdam: Rodopi.
Denwood, Philip. 1999. Tibetan (London Oriental and African language library v. 3). Amsterdam/
Philadelphia: John Benjamins. https://doi.org/10.1075/loall.3
Dwyer, Arienne M. 2007. Salar: A study in Inner Asian language contact processes (Turcologica
37). Wiesbaden: Harrassowitz.

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Edelman, Dzhoi. 1980. K substratnomy naslediyu Tsentralnoaziatskogo yazykovogo soyuza [On
the substrate heritage of the Central Asian linguistic area]. Voprosy yazykoznaniya 5. 21–33.
Edelman, Dzhoi. 1983. e Dardic and Nuristani languages. Moscow: Nauka.
Edelman, Dzhoi. 1986. Sravnitelnaya grammatika vostochnoiranskikh yazykov. Fonologiya
[Comparative grammar of East Iranian languages. Phonology]. Moscow: Nauka.
Evans, Jonathan. 2004. e reconstruction of Proto-Qiang verb inection. In Ying-chin Lin,
Fang-min Hsu, Chun-chih Lee, Jackson T. -S. Sun, Hsiu-fang Yang & Dah-an Ho (eds.),
Studies on Sino-Tibetan languages: Papers in honor of professor Hwang-cherng Gong on his
seventieth birthday, 201–238. Taipei: Instute of Linguistics, Academia Sinica.
Gordon, Matthew & Ayla Applebaum. 2006. Phonetic structures of Turkish Kabardian. Journal of
the International Phonetic Association 36(2). 159. https://doi.org/10.1017/S0025100306002532
Gordon, Matthew Kelly. 2016. Phonological typology. Oxford: Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780199669004.001.0001
Grawunder, Sven. 2017. e Caucasus. In Raymond Hickey (ed.), e Cambridge handbook of
areal linguistics, 356–395. Cambridge: Cambridge University Press.
https://doi.org/10.1017/9781107279872.014
Gui, Ming Chao. 2001. Yunnanese and Kunming Chinese: a study of the language communities,
the phonological systems, and the phonological developments (LINCOM studies in Asian
linguistics 28). München: LINCOM Europa.
Hall, T. Alan. 1997. e historical development of retroex consonants in IndoAryan. Lingua
102(4). 203–221. https://doi.org/10.1016/S0024-3841(96)00050-2
Haller, Felix. 2004. Dialekt und Erzählungen von emchen: sprachwissenschaliche Beschrei-
bung eines Nomadendialektes aus Nord-Amdo (Beiträge zur tibetischen Erzählforschung
Bd. 14). Bonn: VGH Wissenschasverlag.
Hamann, Silke. 2004. Retroex fricatives in Slavic languages. Journal of the International Phonetic
Association 34(1). 53–67. https://doi.org/10.1017/S0025100304001604
Hamann, Silke Renate. 2003. e phonetics and phonology of retroexes (LOT 75). Utrecht: LOT.
Hegedűs, Irén. 2012. e RUKI-rule in Nuristani. In Benedicte Nielsen Whitehead (ed.), e
sound of Indo-european: phonetics, phonemics and morphophonemics (Copenhagen stud-
ies in Indo-European vol. 4), 145–168. Copenhagen: Museum Tusculanum, University of
Copenhagen.
Hegedűs, Irén & Václav Blažek. 2010. On the position of Nuristani within IndoIranian. Paper
presented at the conference Sound of Indo-European 2 (Opava, Oct 2010).
Hill, Nathan W. 2010. An overview of Old Tibetan synchronic phonology. Transactions of the
Philological Society 108(2). 110–125. https://doi.org/10.1111/j.1467-968X.2010.01234.x
Hongladarom, Krisadawan. 1996. Rgyalthang Tibetan of Yunnan: A preliminary report. Lin-
guistics of the Tibeto-Burman Area 19(2). 69–92.
Huber, Brigitte. 2005. e Tibetan dialect of Lende (Kyirong): A grammatical description with
historical annotations (Beiträge zur tibetischen Erzählforschung Bd. 15). Bonn: VGH
Wissenschasverlag.
Jacques, Guillaume. 2014. A phonological prole of Cone. In Jackson T. -S. Sun (ed.), Phono-
logical proles of little-studied Tibetic varieties (Language and Linguistics monograph series
55), 269–330. Taipei: Institute of Linguistics, Academia Sinica.
Jacques, Guillaume & Alexis Michaud. 2011. Approaching the historical phonology of three
highly eroded Sino-Tibetan languages: Naxi, Na and Laze. Diachronica 28(4). 468–498.
https://doi.org/10.1075/dia.28.4.02jac

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
Janhunen, Juha. 1982. On the structure of Proto-Uralic. Finnisch-ugrische Forschungen 44. 23–42.
Janhunen, Juha. 2003. Proto-Mongolic. In Juha Janhunen (ed.), e Mongolic languages (Routledge
language family series 5), 1–29. London: Routledge.
Kogan, Anton. 2015. Nekotorye voprosy geneticheskoj klassikacii dardskih jazykov po dannym
istoricheskoj fonetiki [Some issues of the genealogical classication of the Dardic languages
based on historical phonology]. Journal of Language Relationship 13(12). 1–21.
Konow, Sten. 1936. Note on the ancient North-Western Prakrit. Bulletin of the School of Oriental
and African Studies 8(2–3). 603–612. https://doi.org/10.1017/S0041977X00141230
Kümmel, Martin. 2007. Konsonantenwandel: Bausteine zu einer Typologie des Lautwandels und
ihre Konsequenzen für die vergleichende Rekonstruktion. Wiesbaden: Reichert.
Kurpaska, Maria. 2010. Chinese language(s): A look through the prism of the Great Dictionary of
Modern Chinese Dialects (Trends in linguistics: Studies and monographs 215). Berlin/New
York: De Gruyter Mouton. https://doi.org/10.1515/9783110219159
Ladefoged, Peter & Ian Maddieson. 1996. e sounds of the world’s languages. Oxford, UK/
Cambridge, MA: Blackwell Publishers.
LaPolla, Randy J. 2001. e role of migration and language contact in the development of the
Sino-Tibetan language family. In Alexandra Y. Aikhenvald & Robert M. W. Dixon (eds.),
Areal diusion and genetic inheritance: problems in comparative linguistics, 225–254. Oxford/
New York: Oxford University Press.
Lee, Wai-Sum & Eric Zee. 2003. Standard Chinese (Beijing). Journal of the International Phonetic
Association 33(1). 109–112. https://doi.org/10.1017/S0025100303001208.
Liljegren, Henrik. 2017. Proling Indo-Aryan in the Hindukush-Karakoram: A preliminary
study of micro-typological patterns. Journal of South Asian Languages and Linguistics 4(1).
107–156. https://doi.org/10.1515/jsall-2017-0004
Liljegren, Henrik & Erik Svärd. 2017. Bisyndetic contrast marking in the Hindukush: Additional
evidence of a historical contact zone. Journal of Language Contact Online preprint.
https://doi.org/10.1163/19552629-01002010
Lindblom, Björn & Ian Maddieson. 1988. Phonetic universals in consonant systems. In Victoria
Fromkin, Larry M. Hyman & Charles N. Li (eds.), Language, speech, and mind: studies in
honour of Victoria A. Fromkin, 62–78. London/New York: Routledge.
Maddieson, Ian. 1984. Patterns of sounds. Cambridge: Cambridge University Press.
https://doi.org/10.1017/CBO9780511753459
Maddieson, Ian & Kristin Precoda. 1992. UPSID and PHONEME (version 1.1). Los Angeles:
Uni ver sity of California at Los Angeles.
Makley, Charlene, Keith Dede, Hua Kan & Wang Qingshan. 1999. e Amdo dialect of Labrang.
Linguistics of the Tibeto-Burman Area 22(1). 97–127.
Matiso, James A. 2003. Handbook of Proto-Tibeto-Burman: system and philosophy of Sino-
Tibetan reconstruction (University of California publications in linguistics v. 135). Berkeley:
University of California Press.
Michael, Lev, Tammy Stark, Emily Clem & Will Chang (eds.). 2015. South American phonologi-
cal inventory database v1.1.4. Berkeley: University of California. http://linguistics.berkeley.
edu/~saphon/.
Moran, Steven, Daniel McCloy & Richard Wright (eds.). 2014. Phoible online. Leipzig: Max
Planck Institute for Evolutionary Anthropology. http://phoible.org/.
Morén, Bruce. 2006. Consonant-vowel interactions in Serbian: Features, representations and con-
straint interactions. Lingua 116(8). 1198–1244. https://doi.org/10.1016/j.lingua.2005.04.003

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
Nichols, Johanna. 1997. Modeling ancient population structures and movement in linguistics.
Annual review of anthropology 26(1). 359–384.
https://doi.org/10.1146/annurev.anthro.26.1.359
Niederer, Barbara. 1998. Les langues Hmong-Mjen (Miáo-Yáo): phonologie historique (LINCOM
studies in Asian linguistics 07). München: LINCOM EUROPA.
Nikolaev, Dmitry. 2018. e Database of Eurasian Phonological Inventories: A research tool for
distributional phonological typology. Linguistics Vanguard 4(1).
https://doi.org/10.1515/lingvan-2017-0050
Nikolaev, Dmitry, Andrey Nikulin & Anton Kukhto (eds.). 2015. e database of Eurasian pho-
nological inventories (beta version). http://eurasianphonology.info/.
Norman, Jerry. 1974. e initials of Proto-Min. Journal of Chinese Linguistics 2(1). 27–36.
Norman, Jerry. 1988. Chinese (Cambridge language surveys). Cambridge, UK/New York:
Cambridge University Press.
Norman, Jerry. 1991. e Min dialects in historical perspective. In William S. -Y. Wang (ed.),
Lan guages and dialects of China (Journal of Chinese Linguistics Monograph Series 3), 323–
358. Berkeley, California: Project on Linguistic Analysis.
Nugteren, Hans. 2011. Mongolic phonology and the Qinghai-Gansu languages. Utrecht: LOT.
Petrović, Dragoljub & Snežana Gudurić. 2010. Fonologija srpskoga jezika [Phonology of Serbian].
Beograd: Institut za srpski jezik SANU.
Ratli, Martha Susan. 2010. Hmong-Mien language history (Studies in language change 8).
Canberra: Pacic Linguistics.
Robbeets, Martine. 2017. e Transeurasian languages. In Raymond Hickey (ed.), e Cambridge
handbook of areal linguistics, 586–626. Cambridge: Cambridge University Press.
https://doi.org/10.1017/9781107279872.023
Shorto, Harry Leonard, Paul Sidwell, Doug Cooper & Christian Bauer. 2006. A Mon-Khmer com-
parative dictionary (Pacic linguistics 579). Canberra: Pacic Linguistics, Research School
of Pacic and Asian Studies, Australian National University.
Simmons, Richard VanNess. 1999. Chinese dialect classification: A comparative approach to
Harngjou, Old Jintarn, and common northern Wu. Amsterdam/Philadelphia: John Benjamins.
https://doi.org/10.1075/cilt.188
Starostin, Sergei. 2008. Altaic loans in Old Chinese. In Alicia Sanchez-Mazas, Roger Blench,
Malcolm D. Ross, Ilia Peiros & Marie Lin (eds.), Past human migrations in East Asia:
Matching archaeology, linguistics and genetics (Routledge Studies in the Early History of
Asia 5), 254–262. London/New York: Routledge.
Stolz, omas & Nataliya Levkovych. 2017. Convergence and divergence in the phonology of the
languages of Europe. In Raymond Hickey (ed.), e Cambridge handbook of areal linguistics,
122–160. Cambridge: Cambridge University Press. https://doi.org/10.1017/9781107279872.007
Sun, Jackson T. -S. 2003. Phonological prole of Zhongu: A new Tibetan dialect of Northern
Sichuan. Language and linguistics 4(4). 769–836.
Sun, Jackson T. -S. 2006. Special linguistic features of gSerpa Tibetan. Linguistics of the Tibeto-
Burman area 29(1). 107–126.
Suzuki, Hiroyuki. 2009. Tibetan dialects spoken in Shar khog and Khod po khog. East and West
59(1/4). 273–283.
Suzuki, Hiroyuki & Tshering mTshomo. 2009. Preliminary analysis of the phonological history
of Melung Tibetan. Language and Linguistics 10(3). 521–537.
Tenishev, Edgem R. 1976. Stroj salarskogo jazyka [Grammar of Salar]. Moscow: Nauka.

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
Tikkanen, Bertil. 1988. On Burushaski and other ancient substrata in northwestern South Asia.
Studia Orientalia 64. 303–325.
Tikkanen, Bertil. 1999. Archaeological-linguistic correlations in the formation of retroex ty-
pologies and correlating areal features in South Asia. In Roger Blench & Matthew Spriggs
(eds.), Archaeology and languages IV: Language change and cultural transformation, 138–
148. Abingdon, UK: Taylor & Francis. https://doi.org/10.4324/9780203208793_chapter_6
Tikkanen, Bertil. 2008. Some areal phonological isoglosses in the transit zone between South
and Central Asia. In Israr-ud-Din (ed.), Proceedings of the third international Hindu Kush
cultural conference, 250–262. Oxford/New York: Oxford University Press.
Tsintsius, Vera. 1949. Sravnitelnaya grammatika tunguso-manchzhurskikh yazykov [Comparative
grammar of Tungus-Manchu languages]. Leningrad: Gosuchpedgiz.
Yan, Margaret Mian. 2006. Introduction to Chinese dialectology (LINCOM Studies in Asian
linguistics 22). München: LINCOM Europa.
Yu, Dominic. 2009. Lizu and ProtoTibeto-Burman. Unpublished manuscript. University of
Berkeley.
Zhang, Paiyu. 2013. e Kilen language of Manchuria: Grammar of a Moribund Tungusic Lan-
guage: University of Hong Kong dissertation. https://doi.org/10.5353/th_b4985881
Zhivlov, Mikhail. 2014. Studies in Uralic vocalism III. Journal of Language Relationship 12.
113–148.
Appendices
A. Distribution of aricate-rich and non-aricate-rich languages in Eurasia
Figure7. Aricate-rich (red) vs. non-aricate-rich (grey) languages in Eurasia
(Nikolaev etal. 2015)

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
B. Aricate-rich languages of the Hindukush area
Phylum Genus Language name Aricate inventory
Indo-
European
Indo-Iranic Southwest Pashai ts, tʃ, dʒ, tʃh, dʒh, ʈʂ, ʈʂh
Kati ts, tsj, tsw, dz, tʃ, tʃw, dʒ, ʈʂ, ʈʂw
Munji ts, dz, tʃ, dʒ, ʈʂ, ɖʐ
Shumashti ts, tʃ, ʈʂ, ʈʂh
Wotapuri-Katarqalai ts, tsh, dz, ʈʂ, ʈʂh, dʒ
Gawarbati ts, tsh, tʃ, dʒ, ʈʂ, ʈʂh
Ishkashimi ts, dz, tʃ, dʒ, ʈʂ
Dameli ts, tsh, tʃ, tʃh, ʈʂ, ʈʂh
Northern Kalasha ts, tsh, dz, dzh, tʃ, tʃh, dʒ, dʒh, ʈʂ, ʈʂh,
ɖʐ
Palula ts, tɕ, tɕʰ, ʈʂʰ
Wakhti ts, dz, tʃ, dʒ, ʈʂ, ɖʐ
Kalkoti ts, tɕ, tɕʰ, dʑ, ʈʂ
Khowar ts, tsʰ, dz, tɕ, tɕʰ, dʑ, ʈʂ, ʈʂʰ, ɖʐ
Kalam Kohistani ts, tsʰ, tʃ, tʃʰ, dʒ, ʈʂ, ʈʂʰ
Ushojo ts, dz, tɕ, dʑ, ʈʂ, ɖʐ
Indus Kohistani ts, tsʰ, tʃ, tʃʰ, dʒ, ʈʂ
Shina ts, tsʰ, tʃ, tʃʰ, dʒ, ʈʂ, ʈʂʰ
Burushaski – Werchikwar, Hunza, Nagar ts, tsʰ, tɕ, tɕʰ, dʑ, ʈʂ, ʈʂʰ, ɖʐ
Sino-Tibetan Tibetic Balti (Kharkoo) ts, tsʰ, tʃ, tʃʰ, dʒ, ʈʂ, ʈʂʰ, ɖʐ
app-sc
C. Aricate-rich languages of the Eastern Himalayan area
Phylum Genus Language name Aricate inventory
Sino-Tibetan Tibetic Shigatse Tibetan ts, tsʰ, ʈʂ, ʈʂʰ, tɕ, tɕʰ
Southern Cuona ts, tsʰ, dz, ʈʂ, ʈʂʰ, ɖʐ, tɕ, tɕʰ, dʑ
Northern Cuona ts, tsʰ, dz, ʈʂ, ʈʂʰ, ɖʐ, tɕ, tɕʰ, dʑ
Sangdam Tibetan ts, tsʰ, dz, tɕ, tɕʰ, dʑ, cç, cçʰ, ɟʝ
emchen Tibetan ts, tsʰ, dz, ʈʂ, ʈʂʰ, ɖʐ, tɕ, tɕʰ, dʑ
Melung Tibetan ts, tsʰ, dz, tɕ, tɕʰ, dʑ, ʈʂ, ʈʂʰ, ɖʐ
Dongwang Tibetan ts, tsʰ, dz, dz, tɕ, tɕʰ, dʑ, dʑ,
ʈʂ, ʈʂʰ, ɖʐ, ɖʐ
Rgyalthang Tibetan ts, tsʰ, dz, dz, ʈʂ, ʈʂʰ, ɖʐ, ɖʐ,
tɕ, tɕʰ, dʑ, dʑ
gSerpa ts, tsʰ, dz, ʈʂ, ʈʂʰ, ɖʐ, ɖʐ, tʃ,
tʃʰ, dʒ, dʒ
Khalong Tibetan ts, tsʰ, tsʰ, dz, tʃ, tʃʰ, tʃʰ, dʒ,
dʒ, ʈʂ, ʈʂʰ, ɖʐ, ɖʐ

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
Phylum Genus Language name Aricate inventory
Kami Tibetan ts, tsʰ, dz, dz, tʃ, tʃʰ, dʒ, dʒ, tɕ,
tɕʰ, dʑ, dʑ, ʈʂ, ʈʂʰ, ɖʐ, ɖʐ
Labrang Tibetan ts, tsʰ, dz, dz, ʈʂ, ʈʂʰ, ɖʐ, ɖʐ, tɕ,
tɕʰ, dʑ, dʑ
Nyinpa Cone ts, tsʰ, dz, dz, tɕ, tɕʰ, dʑ, dʑ, ʈʂ,
ʈʂʰ, ɖʐ, ɖʐ
Zhongu Tibetan ts, ʰts, tsʰ, tsʰ, dz, dz, ʈʂ, ʈʂʰ, ʈʂʰ,
ɖʐ, ɖʐ, tʃ, ʰtʃ, tʃʰ, tʃʰ, dʒ, dʒ
Baima ts, tsʰ, dz, tʃ, tʃʰ, dʒ, dʒ, tɕ, tɕʰ,
dʑ
Sinitic
Kunming Chinese (SW Mandarin)
tɕ, tɕʰ, ʈʂ, ʈʂʰ, ts, tsʰ
Standard Chinese (Beijing) ts, tsʰ, tʃ, tʃʰ, tɕ, tɕʰ
Nantong Chinese ts, tsʰ, tʃ, tʃʰ, tɕ, tɕʰ
Qiangic Southern Pumi ts, tsʰ, dz, tʃ, tʃʰ, dʒ, ʈʂ, ʈʂʰ, ɖʐ
Lizu ts, tsʰ, dz, tʃ, tʃʰ, dʒ, tɕ, tɕʰ, dʑ,
tsʰ, dz, tʃʰ, dʒ, tɕʰ, dʑ
Ersu t̪s̪, t̪s̪ʰ, d̪z̪, t̪s̪ʰ, d̪z̪, ts, tsʰ, dz,
tsʰ, dz, tɕ, tɕ, dʑ, tɕʰ, dʑ, ʈɽ,
ʈɽʰ, ɖɽ, ʈɽʰ, ɖɽʰ
Northern Qiang (Hongyan) ts, tsʰ, dz, tʂ, tʂʰ, dʐ, tɕ, tɕʰ, dʑ
Northern Qiang (Yadu) ts, tsʰ, dz, ʈʂ, ɖʐ, tɕ, tɕʰ, dʑ
Puxi ts, tsʰ, dz, ʈʂ, ʈʂʰ, ɖʐ, tɕ, tɕʰ, dʑ
rGyalrongic Caodeng rGyalrong ts, tsʰ, dz, dz, tʃ, tʃʰ, dʒ, dʒ, ʈʂ,
ʈʂʰ, ɖʐ, ɖʐ
Japhug ts, tsʰ, dz, dz, tɕ, tɕʰ, dʑ, dʑ, ʈʂ,
ʈʂʰ, ɖʐ, ɖʐ
Lolo-Burmese
Achang (Lianghe) ts, tsʰ, tɕ, tɕʰ, tʂ, tʂʰ
Nuosu (Black Yi) ts, tsʰ, dz, dz, ʈʂ, ʈʂʰ, ɖʐ, ɖʐ, tʃ,
tʃʰ, dʒ, dʒ, tɕ, tɕʰ, dʑ, dʑ
Ahi ts, tsʰ, dz, tɕ, tɕʰ, dʑ, ʈʂ, ʈʂʰ, ɖʐ
Naxi Yongning Na (Mosuo) ts, tsʰ, dz, tɕ, tɕʰ, dʑ, ʈʂ, ʈʂʰ, ɖʐ
– Shixing t̪s̪, t̪s̪ʰ, d̪z̪, ʈʂ, ʈʂʰ, ɖʐ, tɕ, tɕʰ, d̪z̪,
ɖʐ, dʑ, dʑ
Turkic Oghuz Salar tsʰ, dz, tʃʰ, dʒ, tɕʰ, dʑ, ʈʂʰ, ɖʐ
Mongolic
Qinghai-Gansu
Mongghul ts, tsʰ, tʃ, tʃʰ, ʈʂ, ʈʂʰ
Mangghuer ts, tsʰ, tʃ, tʃʰ, ʈʂ, ʈʂʰ
Dongxiang ts, tsʰ, tʃ, tʃʰ, ʈʂ, ʈʂʰ
Daur Daur ts, tsʰ, tʃ, tʃʰ, ʈʂ, ʈʂʰ, tʃw, tʃwʰ
Hmong-Mien Hmongic Hmong Njua ts, tsʰ, ts, tsʰ, tʃ, tʃʰ, tʃ, tʃʰ, ʈʂ,
ʈʂʰ, ʈʂ, ʈʂʰ
Xong ts, tsʰ, ts, tsʰ, ʈʂ, ʈʂʰ, ʈʂ, ʈʂʰ, tɕ,
tɕʰ, tɕ, tɕʰ

Uncorrected proofs -  John Benjamins Publishing Company
 Dmitry Nikolaev and Eitan Grossman
D. Aricate-rich languages in South America
Figure8. Distribution of aricate-rich languages in South America (Michael etal. 2015)
Table9. Aricate-rich languages in South America
Phylum Language Aricate inventory
Panoan Matís ts, tʃ, ʈʂ
Matsés ts, tʃ, ʈʂ
Shipibo ts, tʃ, ɖʐ
Arawak Chamicuro ts, tʃ, ʈʂ
Yánesha ts, tʃ, ʈʂ
Yine ts, tʃ, cç
Quechua Ancash Quechua ts, tʃ, ʈʂ
Barbacoan Guambiano ts, tʃ, ʈʂ
Uru-Chipaya Chipaya ts, ts’, tsʰ, tʃ, tʃ’, tʃʰ, ʈʂ, ʈʂ’, ʈʂʰ
Isolates Candoshi-Shapra ts, tʃ, ʈʂ
Muniche ts, tʃ, ʈʂ
Camsá ts, tʃ, ʈʂ

Uncorrected proofs -  John Benjamins Publishing Company
Areal sound change and the distributional typology of aricate richness in Eurasia 
E. Overview of sound changes producing retroex aricates
Table10. Sound change processes leading to retroex aricates
Typ e Source Context Language
Unconditioned shi dʒ, tʃ_ Polish; SerboCroatian; Waxi
(Iranian)
dʑ, tɕ_ Dongxian (Qinghai-Gansu,
Mongolic); ProtoMandarin
Shi under the inuence
of a rhotic segment
ts r_ Yongning Na (Qiangic,
SinoTibetan)
ts _r Common Nuristani
Shi under the inuence
of a nonfront vowel
tɕ_V[−front] Tibetic
Retroex harmony tɕ, tʃ_C[+retroex] Dardic
Simplication of
velar+sibilant clusters
ks, kjs,
kws
 Common Nuristani, Common
Dardic
Simplication of clusters
with rhotics
tr #_V Proto-MiddleChinese,
ProtorGyalrongic, Hmong-Mien
br, p(h)r _ Common Dardic, Tibetic,
HmongMien
gr, k(h)r _ Tibetic, Nuristani
Simplication of clusters
with laterals
kl _ Common Nuristani
Aricativisation of
retroex stops
ʈ#_V Proto-Mandarin
Aricativisation of
retroex fricatives
ʂ_w Khowar (Dardic)
Metathesis ʂʈ _ Bashkarik (Dardic)
Author queries
Q0001
Please provide a complete reference for the citation ‘(Chirkova 2010), (Jacques 2011)’ in this
article.
Q0002
Please provide a citation for the reference id “CIT0019 (Chirkova, Ekaterina. 2014), CIT0040
(Jacques, Guillaume. 2014)” since citation is missing in the article.
Authors’ addresses
Dmitry Nikolaev (corresponding author)
Department of Linguistics
Hebrew University of Jerusalem
Mt. Scopus
Jerusalem, 9190501
Israel
dsnikolaev@gmail.com
dmitry.nikolaev@mail.huji.ac.il
Eitan Grossman