([Link]

info/)

Home ([Link] Features ([Link]

Chapters ([Link] Languages ([Link]

References ([Link] Authors ([Link]

(https%3A%2F%[Link]%2Fchapter%2Fs1)

Chapter Introduction References

by Bernard Comrie ([Link] and Matthew S. Dryer
([Link] and David Gil Moseley and Asher 1994
([Link]
([Link] and Martin Haspelmath and-Asher-1994)
([Link] cite
Asher and Moseley 2007
([Link]
and-Moseley-2007)

Table of contents. Lewis 2009

([Link]
2009)
1.  What and why
2.  Organization Foley 2000
2.1.  The features ([Link]
2000)
2.2.  The maps
2.3.  The feature values Schmidt 1926
3.  The languages ([Link]
1926)
3.1.  The WALS samples
3.2.  Language vs. dialect Grimes 2000
3.3.  Language names ([Link]
3.4.  Locations of languages 2000)
3.5.  Three-letter codes Goddard 1996
4.  The data sources ([Link]
5.  Disclaimer 1996)

Bradley 1997
1. What and why? ([Link]
1997)
The World Atlas of Language Structures (WALS) provides the reader with
Tindale 1974
160 maps showing the geographical distribution of structural linguistic ([Link]
features. When it was first published in book form in 2005, it was a quite 1974)
novel type of atlas. Linguists have long worked with maps showing the
geographical distribution of languages (i.e. the areas where most of their
speakers live), and the first complete world atlas was published as a book
in 1994 (Moseley and Asher 1994 ([Link]
and-Asher-1994), Asher and Moseley 2007
([Link] see also
Ethnologue ([Link] Lewis 2009
([Link] For well over a century,
linguists have also produced atlases that show the geographical
distribution of linguistic features in the dialects of a language. WALS is the
first feature atlas on a world-wide scale. It can be thought of as a kind of
dialect atlas of the “dialects” of Human Language. But it differs from dialect
atlases in an important way. While dialect atlases show the geography of
substantive linguistic features (such as particular cognate sounds, or
particular words), WALS shows only structural features, i.e. abstract
features of the language system that can be compared across unrelated
languages.

Linguists interested in linguistic typology — the systematic study of the

ways in which the languages of the world vary structurally and of the limits
to this variation — began in the 1980s to ask questions relating to the
geographical distribution of different values for structural linguistic features.
For instance, we may want to know whether languages with a particular
word order in the clause, say subject–verb–object as in English
([Link] (the farmer killed the
duckling), are found only in one part of the world, whether they are
distributed more or less evenly across different parts of the world, or
whether some in between scenario holds. Although previous work has been
able to provide some answers to some questions of this kind, these
answers were by and large unsystematic, often reflecting more the intuitive
feel that a particular linguist has for the geographical distribution of the
feature in question rather than a consistent sampling of the world’s
languages in order to answer the question. The World Atlas of Language
Structures aims to provide just this kind of systematic answer, since the
authors of the individual chapters, each dealing with a particular linguistic
feature, have set out to be as comprehensive (within their sampling limits)
as possible in the mapping of variants for that feature across the languages
of the world. The maps provide a visual overview of this distribution, by
using different colored dots for the different feature values. In addition, the
text that accompanies each map provides an explanation of the feature
values and of their assignment, as well as discussion of patterns of
geographical distribution and of the relevance of the chapter to theoretical
issues. The online maps allow users to zoom in on particular geographic
areas, and to see the bibliographical or other sources that underlie the
atlas’s database. The online version also enables the user to manipulate the
maps to some extent, for instance to see whether there is a correlation
between particular values of different features, and to export the data for
further analyses.

Although the main users of WALS will be those interested in linguistic

typology, the atlas is also relevant to the interests of other linguists. For
instance, certain theoretical approaches have been criticized for being
based too heavily on languages exhibiting particular geographically
restricted feature values, and WALS will enable those interested in testing
such criticisms to see whether they do indeed hold. Thus, much recent
work on relative clauses has been based heavily on the kind of construction
found in the major European literary languages, such as literary English
([Link] the man whom I saw, and
has been criticized for extending the analysis appropriate for such
languages to other languages with radically different relative clause types.
While the material on relative clauses in the atlas does not, of course,
directly address the issue of whether a particular approach to the analysis
of relative clauses is valid cross linguistically, it does show that the
distribution of the “European” type of relative clause (in the terminology
used here: the relative pronoun strategy) is by and large restricted to
Europe, thus at least calling into question theoretical approaches that rely
on so geographically restricted a typological variant (see chapter 122
([Link] and 123 ([Link]

Both typologists and other linguists are interested in questions relating to

correlations among different features, such as whether there is indeed a
tendency for occurrence of prepositions (rather than postpositions) to
correlate with verb–object order in the clause, and for postpositions to
correlate with object–verb order; whether there is a tendency for noun–
adjective order in the noun phrase to correlate with verb–object order in the
clause, and adjective–noun order with object–verb order. The relevant
WALS data, in particular the maps that can be generated in the electronic
version by combining data from different individual maps, suggest that
there is validity to the postulated correlation in the first case (adpositions
and verb position), but not in the second (adjectives and verb position) (see
chapters 95 ([Link] 96 ([Link]
and 97 ([Link]

Previous work in typology has provided extensive results regarding the

ways in which languages vary structurally and of correlations among
different features. As noted above, WALS provides significant further
contributions to these areas. However, because of the maps, WALS
provides an especially significant contribution to the field of areal typology,
which seeks to establish whether particular geographical distributions are
the result of language contact among neighboring languages.

The maps vary with respect to the degree to which the features show areal
patterns. On some maps, clear geographical patterns emerge. On Map 83A
([Link] for example, one finds a
clear distribution of the two orders of object and verb among the languages
of Eurasia. In Europe, most of the languages are VO (placing the verb
before the object), while to the east of this is a huge area covering much of
Asia, where most of the languages are OV (placing the object before the
verb). And in Southeast Asia, and extending out into the Pacific, is a large
area where VO again predominates. Other maps show much less areal
patterning of this sort. Where one finds similarities of this sort within a
particular geographical area, there are three sorts of explanation. One is
that it is the result of contact between languages. There are a number of
well-documented instances of relatively small linguistic areas, such as the
Balkans and Mesoamerica, where many features are shared due to contact,
irrespective of their genealogical classification. However, some of the
WALS maps suggest the possibility of larger linguistic areas, such as one
covering much of northern Eurasia.

The second possibility is that it reflects a genealogical relationship among

at least some of the languages, involving a feature inherited from a
common ancestor. For example, the fact that VO order predominates from
Indonesia and the Philippines and extending eastward into the Pacific (but
excluding much of New Guinea) reflects the fact that almost all of these
languages belong to the Austronesian
([Link] family and the VO order is
apparently a feature that is inherited from Proto-Austronesian, the ancestor
language from which all of the Austronesian languages have descended.
Note that in many cases shared features within a geographical area may be
partly genealogical and partly due to contact. When languages within the
same family share certain features after a long period of time, contact
among speakers of different languages in the family may have reinforced
similarities that were features of the protolanguage so that after a long
enough period of time, the fact that the features are shared may be as
much due to contact as due to their common ancestry. It should also be
emphasized that features shared among languages in the same
geographical area which are classified genealogically as belonging to
separate families may reflect deeper genealogical connections. At first
sight, this sounds like a contradiction: if they are in separate families, then
how can they have a deeper genealogical connection? The answer is that it
must be understood that when linguists classify languages into language
families, they generally intend groups for which there is considered to be
strong evidence of common ancestry. But there is little doubt that many of
these families are in fact distantly related to each other but that the
available evidence means that, at least at this time (and perhaps at any
future time), there is a lack of convincing evidence for just which families
are in fact related to which other families. But the absence of convincing
evidence of a genealogical relationship does not mean that they may not
share a few features that are retentions from a common ancestor.

The third possible explanation for shared features within a particular

geographical area is that it is at least partly coincidence. Especially when a
map shows primarily two values, as does Map 83A
([Link] with VO and OV word order,
there are bound to be geographical areas in which one of these orders
predominates where the occurrence throughout this area is simply
accidental, where there may be two or more regions within this area which
are predominantly OV or VO due to contact or for genealogical reasons, but
where the fact that these regions are the same as each other is purely
coincidental. Along the north coast of Papua New Guinea
([Link] are a number of
languages in different families, including Austronesian
([Link] which are VO. While some
of these families may have acquired VO order due to contact with
Austronesian, this seems rather unlikely for one family, Torricelli
([Link] since, as noted by Foley (2000
([Link] 365), most the languages of this
family are inland and have apparently had relatively little contact with
Austronesian languages. Thus the contiguity of the VO languages in the
Torricelli ([Link] family with a few VO
languages in the Austronesian
([Link] family seems most likely to
be coincidental.

One way to determine whether a trait that is shared within a geographical

area does so only coincidentally is to see whether there are other unrelated
traits which are also shared among the same (or a similar) set of languages.
Evidence for multiple shared traits within a particular geographical area
provides evidence for a linguistic area. While the WALS database provides
the possibility of investigating correlations among linguistic features
throughout the languages of the world, it also provides the possibility of
investigating correlations within particular regions. If a number of feature
values correlate in a particular part of the world, but not in the world as a
whole, then this is particularly strong evidence that the current distribution
in that part of the world is the result of contact among its languages or
possibly deep genealogical relationship. Even brief perusal of the maps will
reveal, for instance, that Southeast Asia often stands out from neighboring
areas, and sometimes even from the world as a whole, with the result that
the materials provide a more solid basis than heretofore for the recognition
of Southeast Asia as a linguistic area (see, for instance, maps 13A
([Link] 51A
([Link] and 55A
([Link] (In addition, they show how
some feature values characteristic of Southeast Asia have a certain
distribution beyond this area, reflecting lower but nonetheless palpable
language contact between Southeast Asia and neighboring areas with
percolation of some feature values.)

Even a single trait, if relatively common in one area and relatively

uncommon outside that area, can provide evidence for a linguistic area.
The example cited above, of relative pronouns being common in Europe –
not only in Indo-European ([Link]
languages but also such Uralic ([Link]
languages as Finnish ([Link]
Estonian ([Link] and Hungarian
([Link] – but uncommon elsewhere
in the world, provides evidence for Europe as a linguistic area.

We hope that WALS will provide the interested lay reader with an overview
both of the typological diversity of the world’s languages — including
reference to phenomena that might be judged highly unusual by the
standards of the world’s best known languages — and of the ways in which
that diversity is often patterned geographically, with particular areas often
being characterized by particular feature values or particular combinations
of feature values (whence the term “areal typology”).

Although The World Atlas of Language Structures is a pioneering effort, we

do acknowledge the efforts of earlier linguists who, on the basis of much
less extensive data and with far fewer resources, nonetheless achieved
significant results in mapping the world’s linguistic diversity. Among our
intellectual forebears, particular reference should be made to Schmidt
(1926) ([Link]

2. Organization
2.1. The features
The World Atlas of Language Structures contains 144
([Link] chapters, each consisting of a text and a main
map. Each of the 144 chapters shows the distribution of a particular
linguistic feature, reflected in the chapter’s title. In several cases, a single
text accompanies more than one map.

Each chapter was contributed by an author ([Link] (or

team of authors) who is an expert on the particular structural feature, and
who collected the world-wide cross-linguistic data from published materials
and other sources ([Link] Quite a few authors are
responsible for more than one chapter. Altogether 55 authors made a
contribution to this work.

The 144 chapters are grouped thematically ([Link] into

the following eleven sections: phonology, morphology, nominal categories,
nominal syntax, verbal categories, word order, simple clauses, complex
sentences, lexicon, sign languages and other. As suggested by the section
titles, the features span all of the major areas of language structure.

The first eight sections, phonology, morphology, nominal categories,

nominal syntax, verbal categories, word order, simple clauses, complex
sentences, encompass the major structural domains of grammar. Within
each of these sections, the features included provide a broad coverage of
the most important subdomains within each of these structural domains.
These eight sections constitute the core of the atlas.

The remaining three sections, containing a total of 14 chapters, are of a

more variegated nature. In the lexicon section, broad coverage would have
been impractical; instead, a somewhat arbitrary choice of features provides
a sample of the kinds of patterns that can be observed in this domain. In
the sign-language section, too, the two features that are included represent
arbitrary choices within the domain of sign-language morphosyntax. Finally,
the last section touches on two diverse domains that are generally
considered marginal to linguistic structure: para-linguistic sounds and
writing systems.

The choice of features included in the atlas was governed by several

competing considerations. As suggested above, the features were chosen
to represent as many as possible of the major domains and subdomains of
language structure. However, the desire to be as inclusive as possible
came up against some practical considerations.

First, the requirement that each map represent a geographically and

genealogically adequate sample of the world’s languages entailed that the
data for each map be obtained primarily from previously-published
descriptions of each individual language, typically in the form of a reference
grammar. However, this limited the choice of features to those for which
information is available in a typical reference grammar, and in a theory-
neutral form facilitating cross-linguistic comparisons.

Secondly, the magnitude of this work, coupled with the desire to achieve a
complete product within a limited time frame, entailed that the bulk of the
data in the atlas derive from work conducted over the course of a lengthy
period, extending back for years and in some cases even decades. Thus,
the choice of features was largely determined by what was already on offer
from the contributing authors.

Due to these practical considerations, many features of current or potential

future interest to linguistics had to be excluded. If the reader is
disappointed that his or her favourite feature is not to be found in the atlas,
chances are that this feature simply has not been described for a
sufficiently large number of diverse languages to have warranted its
inclusion. Hopefully, the absence of such features from the atlas will
motivate future linguists to go out and collect the necessary data.

2.2. The maps

The great majority of maps ([Link] show two hundred
languages or more. Map 83A ([Link]
tg_format=map) (“Order of object and verb”) shows the greatest number of
languages, while the two maps on sign languages (Maps 139A
([Link] and 140A
([Link] show a much smaller
number, for the simple reason that linguists have only recently begun to
study the grammatical structure of sign languages in a comparative
perspective. On average, the maps show about 400 languages. This is less
than 10 percent of the world’s languages, so the picture that we see in this
atlas is far from complete. However, not more than 10-15 percent of
languages have been described comprehensively, and many hundreds of
languages are still completely or almost completely unknown. But both
descriptive and comparative linguistics have made enormous progress in
recent decades, and these efforts are reflected in the current work.
Altogether 2662 languages, just little less than one half of the world’s
languages, occur somewhere in the atlas — we call these the WALS
languages. More than 6700 books and articles have been consulted by the
authors and the relevant bibliographical references can be accessed in the
online version of the atlas. In addition to the maps and accompanying
texts, the atlas contains a genealogically organized list of the languages
(the Genealogical Language List ([Link] by
Matthew S. Dryer ([Link] to facilitate
identification of each language. Issues having to do with the identification
and designation of languages and language families are discussed in detail
in §3.

One chapter, the chapter on writing systems (chapter 141

([Link] is somewhat special with respect to the
maps: it shows differently coloured areas rather than differently coloured
dots.

2.3. The feature values

Each feature is associated with a set of feature values, forming the basis
for distinguishing between languages of different types. The simplest maps
show just two different values. For example, Map 107A
([Link] on passive constructions
shows two language types, those possessing a passive construction and
those lacking a passive construction. Most maps distinguish between three
and five values, but some distinguish up to nine values (features 33A
([Link] 49A ([Link] and 51A
([Link] Although language types can usually be
subdivided into many more subtypes, the values have been limited to at
most nine in WALS because more types are difficult to represent by
different colours, and users would find them hard to distinguish anyway.
Each value is represented by a unique symbol, most often characterized by
a particular colour, but sometimes also by a particular shape.

Within each chapter, the feature values are indicated in a small box on the
map itself (which can be hidden), and in expanded form in a larger box
(called the feature value box) within the accompanying text. For each
feature value, these two boxes also show the colour and shape of the
associated dot, and the number of languages characterized by the feature
value in question. In addition, the text of each chapter provides a detailed
description of the set of feature values, including the criteria that were used
to assign feature values to individual languages.

One general requirement on the feature values of each chapter is that they
be exhaustive. What this means is that for each feature, each and every
language under consideration must be assigned one of the feature values;
there are no "blank" cases of languages that do not have a feature value. In
order to meet this requirement, many chapters include one or more feature
values that account for various cases involving non-applicability,
indeterminacy or mixing. For example, many maps characterize the
different kinds of a certain construction, such as the associative plural
(chapter 36 ([Link] but also include an additional
feature value corresponding to the absence of the construction in question.
Similarly, many of the word order maps (chapters 81-97), such as noun–
adjective order, also include an additional feature value corresponding to
the absence of a dominant word order. (However, the exhaustiveness
requirement is violated in Maps 14-17 on stress, which are silent on
languages lacking stress. There was no way around this, because these
chapters are based on a database collected earlier (StressTyp) that did not
include stressless languages.)

Each feature value is associated with a dot of distinctive colour and, in

some maps, also shape. The choice of default symbols is intended to
reflect, as accurately as possible, the logic underlying each set of feature
values. For example, if feature values form an ordered set, then their
symbols should ideally also form an ordered set. Similarly, feature values
that are closer to each other should ideally be associated with symbols that
are closer to each other. In many cases, however, it proved impossible to
come up with a set of symbols that would be completely isomorphic with
the logic of the feature values. Nevertheless, the reader will notice a
number of colour and shape schemes which recur throughout the atlas,
underscoring the presence of particular logical patterns each of which is
shared by a variety of different maps.

No attempt has been made to make chapters by different authors that

overlap in their features consistent, and it is not hard to find inconsistencies
between certain chapters. For example, there are chapters by three
different authors (or sets of authors) which deal with nominal case. There
are some languages which are shown as lacking case by one author and as
having case by a second author. These inconsistencies can arise for a
number of reasons. Some occur because different authors use different
criteria for identifying case, these differences hopefully being clear from the
text accompanying the maps. Some occur because different authors
classify particular borderline cases differently. Some occur because
different authors use different sources and the sources describe the
phenomenon differently. And some occur simply because of errors, either in
an author’s interpretation of their source or simply due to a coding error. In
fact, quite independently of inconsistencies between chapters, there are
bound to be errors of coding that will only be discovered after this atlas has
been published. Such errors can be reported to the editors via the contact
form ([Link] or the comment buttons associated
with each datapoint (i.e. language-feature pair).

3. The languages
3.1. The WALS samples
There is a total of 2662 languages which appear on at least one map in the
atlas. Some of these languages (262 in number) appear on only one map,
while some, such as English
([Link] appear on most of the
maps. There are 180 languages which appear on at least 80 maps, and 449
languages which appear on at least 40 maps. The choice of which
languages to include on particular maps was the choice of individual
authors. However, there is a set of 100 languages (hereafter the 100-
language sample ([Link] which authors
were asked to include on their maps if at all possible, and a further 100
languages which authors were encouraged to include on their maps
(hereafter these two sets of 100 languages together are referred to as the
200-language sample ([Link]

A general desideratum for a good language sample is that it maximize both

genealogical and areal diversity. Samples which include too many
languages from one area of the world or too many languages from one
family can provide a misleading picture of the relative frequency of different
types of languages. Typological studies in the past have often included a
disproportionate number of Indo-European languages or of languages of
Europe or Eurasia. While Eurasia has a larger land mass than any other
continental region in the world, fewer than twenty percent of the languages
of the world are spoken on mainland Eurasia (i.e. excluding the languages
of Indonesia and the Philippines and other islands). In fact, there are more
languages spoken on the island of New Guinea than in mainland Eurasia.
Furthermore, as a number of the maps in this atlas show, there are patterns
of similarity among languages of Eurasia that one does not find elsewhere
in the world. For example, Map 97A ([Link]
tg_format=map) shows that the vast majority of the OV languages of
Eurasia (i.e. ones that place the object before the verb) place the modifying
adjective before the noun. From this, linguists in the past erroneously
concluded that this was a normal feature of OV languages. But as Map 97A
([Link] shows, this is not true
outside of Eurasia, where OV languages more often place adjectives after
the noun. Only by using samples of languages which include many
languages from outside Eurasia can we avoid making erroneous inferences
of this sort.

Maximizing genealogical and areal diversity were major considerations in

constructing the 100- and 200-language samples. However, there were a
number of other considerations that played a role in constructing these
samples that would not generally play the same role in constructing
samples of languages. First, most of the languages of the islands of the
Pacific fall within the Oceanic ([Link]
branch of the Austronesian ([Link]
family and thus are closely related to each other. For instance, one would
normally not include more than one of these languages in a sample of 100
or even 200 languages. However, because the sample used here is for an
atlas, we decided that we ought to include more of these languages, since
otherwise there would be few dots on the maps in the Pacific. For this
reason, there are two Oceanic languages in the 100-language sample and
seven in the 200-language sample. Similar considerations led to the
inclusion of three Bantu languages in the 100-language sample and five in
the 200-language sample. Without these, many of the maps would have
shown few languages in sub-Saharan Africa, and the majority of those
shown would have been non-Bantu languages that are in some ways
atypical of this region. A second consideration that would not normally play
a role in constructing a language sample is that we felt that we ought to
include a number of the major languages of Eurasia, even when this meant
including pairs of languages which are too close genealogically to be
otherwise included in a sample of 100 or 200 languages, including English
([Link] and German
([Link] French
([Link] and Spanish
([Link] and Modern Hebrew
([Link] and Egyptian Arabic
([Link]

A further consideration in choosing languages for the 100- and 200-

language samples was the ready availability of detailed grammatical
descriptions. In most cases, the choice of a language over genealogically
related languages was based on the availability of detailed descriptions.
Some of the languages that were included in the samples are ones for
which there was no detailed description at the beginning of the WALS
project (in 1999) but for which an expert on the language was willing to
answer questions from authors (see §4). Some of the languages in the 200-
language sample were chosen primarily for the purposes of maximizing
genealogical or areal diversity, despite the fact that the available
descriptions of these languages are somewhat meagre, thus making it
impossible for many authors to include them on their maps. One language
in the 200-language sample, Minica Huitoto
([Link] appears on only 32 maps;
however this was because we eventually realized the need to distinguish
this language from other Huitoto ([Link]
languages and some authors in attempting to include Huitoto used sources
for one of these other languages.

The choice of which languages to include in the 100-language sample and

which to include in the 200-language sample was based on the following
considerations. Languages which are more well-known were normally
placed in the 100-language sample. Languages with more readily available
or more detailed descriptions were also placed in the 100-language
sample. However, we also attempted to maximize genealogical and areal
diversity in both samples so that in some instances, a language was placed
in the 200-language sample but not in the 100-language sample, if there
was already a language that was close genealogically or geographically in
the 100-language sample.
The lists of languages in these samples is available as follows:

100-language sample ([Link]

200-language sample ([Link]

While the 100- and 200-language samples could be used as samples for
other typological studies, a genealogically more balanced sample, with only
one language per genus, would remove Kongo
([Link] Luvale
([Link] Nkore-Kiga
([Link] Zulu
([Link] Koromfe
([Link] Modern Hebrew
([Link] or Egyptian Arabic
([Link] German
([Link] or English
([Link] French
([Link] or Spanish
([Link] Bawm
([Link] Drehu
([Link] Kilivila
([Link] Kiribatese, Maori
([Link] Paamese
([Link] Rapanui
([Link] Karo Batak
([Link] Kobon
([Link] Usan
([Link] Ngiyambaa
([Link] Pitjantjatjara
([Link] Yidiny
([Link] Central Yup'ik
([Link] Navajo, Passamaquoddy-
Maliseet, Carib, and Urubu-Kaapor. Two languages (Makah, Kutenai) are
ones that were included because of the availability of language experts to
answer questions from chapter authors or because copies of unpublished
descriptions were made available to authors and which therefore might not
be included in a sample used for future typological studies.

3.2. Language vs. dialect

There are a number of instances in which distinct dialects are distinguished
in the atlas. For example, WALS distinguishes four dialects of Inuktitut, the
Eskimo-Aleut ([Link] language of
northern Canada ([Link] In
a more extreme case, Map 54A ([Link]
tg_format=map) shows a number of different dialects of German (because
dialects vary with respect to the phenomenon investigated). In other cases,
what is shown as a language is probably in fact a set of closely related but
mutually unintelligible languages. An example of this is Bikol
([Link] spoken in the Philippines
([Link] No systematic
attempt has been made here to distinguish dialects of the same language
from different languages. Where authors submitted separate data for more
than one dialect of a language, we generally maintained the distinction, and
where authors submitted data using a name that covers a number of
mutually unintelligible languages, we endeavoured to ascertain which
language their data was based on, but in some cases we were unable to do
so, and in some cases, their data was based on more than one source,
where the sources describe different varieties.

We originally attempted to have a set of WALS languages in which no

language was a variety of another WALS language. While we were able to
minimize this, it proved impossible to apply this consistently. For example,
as just mentioned, one map shows a number of varieties of German, while
all other maps that include German simply show German, without
specification of a particular variety. A more typical example is provided by
Irish. Here, most authors provide data for Irish, without specifying a variety.
But one author submitted data for Donegal Irish
([Link] and another author data for
Munster Irish ([Link] (since the
dialects differ from each other in some ways). The WALS set of languages
also occasionally includes pairs in which one is a variety of the other, where
the relationship is one of language to set of closely related languages rather
than one of dialect to language. For example, while we generally distinguish
different Huitoto ([Link] languages, a few
authors submitted data where it was not clear which Huitoto language their
data was based on and for their maps, the WALS language is simply called
Huitoto.

3.3. Language names

Many languages are known in the literature under different names. We have
attempted to choose names for WALS that are the names by which the
languages are currently known. This means that our name is occasionally
different from that used in some sources on the language, where we have
reason to believe that the name used in the source does not conform to
more recent usage. For example, older sources on O’odham refer to the
language as Papago, but we use the more recent name O’odham
([Link] Older names are often
considered offensive by communities in which the language is spoken. On
each language page (see for example O’odham
([Link] we also give the name of
the language as it is listed in Ethnologue ([Link] as
best we can determine.
When names of languages involve two or more words with a modifier
followed by a head in the usual English name for the language, where the
head denotes a language or language group and the modifier identifies a
particular variety of that language or language group, we have two ways of
referring to the language. In the chapter texts we use the usual name, with
the modifier preceding the head, but in all lists, we place the head first with
the modifier following in parentheses. Thus, what is called Chalcatongo
Mixtec in chapter texts is called Mixtec (Chalcatongo) in lists. Note that we
follow this convention both when the head denotes a language and the
modifier identifies a dialect of that language (e.g. Irish (Donegal)
([Link] and when the head denotes
a language group and the modifier identifies a particular language in that
group (e.g. Mixtec (Chalcatongo)
([Link]

There are a number of cases that may look like instances of this, but where
the head does not denote a language or language group of which the
modifier identifies a variety. For example Upper Kuskokwim
([Link] remains in that form
because it is not the upper variety of a Kuskokwim language (there is no
Kuskokwim language or language group); rather it is a language spoken in
the vicinity of the Upper Kuskokwim river. Similarly, Tümpisa Shoshone
([Link] remains in that form since it
is not a variety of Shoshone
([Link] but just a closely related
language. Some language names in English already occur in the form Head
Modifier because their name reflects the syntax of some other language in
which the modifier follows the head. These remain in the form Head
Modifier, without parentheses. An example of this is Hmong Daw
([Link] Also we retain the order
Modifier Head if the Head denotes a type of language, such as creole,
pidgin, or sign language, rather than a genealogical category. An example
of this is Berbice Dutch Creole
([Link]

There are many instances of homophonous language names, where two

languages in different parts of the world happen to have the same name.
Where two such languages occur in the set of WALS languages, we
generally disambiguate them by adding a modifier of the form in plus
country name in parentheses, as in Baka (in Cameroon)
([Link] and Baka (in Sudan)
([Link] where the former is a
Niger-Congo ([Link] language and
the latter is a Nilo-Saharan ([Link]
language. We sometimes employ this usage for only one language, where
we use a different name for the second language, simply because the
ambiguous name is sometimes used by others as a name for the second
language. An example of this is Mono (in United States)
([Link] where Mono is also the
name used by Ethnologue for the Austronesian
([Link] language we call Mono-Alu
([Link] Occasionally this
approach does not suffice to disambiguate a language name when both
languages are spoken in the same country. In this case, we place the name
of the language family in parentheses. An example of this is Motilón
(Chibchan) ([Link] where the family
is needed to distinguish it from the Carib
([Link] language Yukpa
([Link] which is sometimes known
as Motilón and which is also spoken in Colombia
([Link] and Venezuela
([Link] Note that there are
many instances of homophonous language names where we do not add a
modifier because the other languages with this name are not WALS
languages. For example, Ethnologue lists two languages called Bulu, one
spoken in Cameroon, the other in Papua New Guinea. Since only the
former is a WALS language, we simply call this language Bulu
([Link]

While our names include standard diacritic symbols (as in á, ä, or ã), we

avoid using superscripts or symbols that are not standard symbols. Thus
we use Yidiny, rather than Yidinʸ or Yidiɲ.

3.4. Locations of languages

The languages are represented on the maps as dots, rather than as
regions, but it should be borne in mind that many languages are spoken
over areas larger than the dots. We attempted to locate the dots
somewhere near the centre of the region of where the languages are
spoken, although in some cases this was difficult because the region in
which the language is spoken is discontinuous. In these cases, we
generally located the dot within the larger region in which the language is
spoken. In some cases, the location of the dot is based on the location of a
major city, town, or village in which the language is spoken. For example,
Egyptian Arabic ([Link] is located
in the vicinity of Cairo, rather than in the middle of Egypt. For most
languages, the location of the dots is based on their location on maps in
Moseley and Asher (1994) ([Link]
Asher-1994) or Ethnologue ([Link] (Grimes 2000
([Link] For languages spoken in
Canada and the United States, the location is based on Goddard (1996)
([Link] For languages spoken in
Nepal, the location is based on maps in Bradley (1997)
([Link] For many of the languages
spoken in Australia, the location is based on Tindale (1974)
([Link] In some instances, the
location is simply based on an explanation in the specific sources for the
language. Future work will probably make clear that the location of some of
the dots is inaccurate. We encourage users to report inaccuracies to us.

Note that in identifying the location of languages, we use locations prior to

European colonial expansion. This means that the dot for English is located
in England, and not in some other country where English is spoken.
Similarly, Spanish is located in Spain, despite the fact that the majority of
speakers are in the Americas. Thus, the languages shown in the Americas
are for indigenous languages and for creoles and sign languages (since
instances of the latter two types in the Americas only have locations in the
Americas). Analogously, indigenous languages that are now spoken in
locations different from where they were spoken at the time of European
contact are located in their location at the time of European contact rather
than their present location (in contrast to the practice in Ethnologue
([Link] For example, a number of indigenous
languages of the United States are now only or primarily spoken on
reservations in Oklahoma, often far from where their speakers originally
lived. An example of this is Yuchi
([Link] originally spoken much
further east, in what is now Tennessee.

3.5. Three-letter codes

On the maps of the printed atlas, the languages are identified by means of
a three-letter code (WALS code). This code is also used in the URL of the
language pages in the online version. We considered using the three-letter
codes employed by Ethnologue ([Link] (which have
meanwhile become the ISO 639-3 ([Link] standard)
but decided not to, both because there are many instances in which the
languages in WALS either represent varieties of languages in Ethnologue or
correspond to an entire set of languages in Ethnologue and because we
wanted to use three-letter codes that are more mnemonic. On each
language page, we give the Ethnologue three-letter code for each WALS
language, as best we can determine. In assigning three-letter codes to
languages, we first attempted to use the first three letters of the language
name, unless the language name includes two or more words, in which
case we attempted to use three letters based on initial letters in the
different words. In other cases (where there is more than one language with
the same three initial letters), we used the first three consonants in the
name of the language (ignoring vowels). In some cases we had to use
some other sequence of letters appearing somewhere in the name of the
language and in a few instances had to resort to adding as the third letter
some letter that does not appear in the name of the language. But we
always use the first letter in the WALS language name as the first letter in
the three-letter code. With language names which are represented as
Modifier Head in the chapter texts but as Head (Modifier) in the
Genealogical Language List ([Link] and in
the electronic version, we use the latter for determining the three-letter
code so that the first letter in the three-letter code matches the first letter in
the head. For example, the three-letter code for Chalcatongo Mixtec, called
Mixtec (Chalcatongo) ([Link] in the
Genealogical Language List ([Link] and in
the electronic version, is mxc. For languages whose names consist of two
letters, such as Ik ([Link] we use a
two-letter code rather than a three-letter code.

4. The data sources

The maps of the World Atlas of Language Structures are largely based on
published primary sources that provide information about the languages in
question. These include full grammars and dictionaries, but also more
specialized articles that are confined to particular aspects of the language
structure (e.g. only the phonological structure, or only certain syntactic
constructions). Unpublished dissertations have also been used as sources,
because these are often accessible to typologists. In a few cases (for
languages like English, Spanish or Russian), the authors have relied on their
own knowledge of the language. Secondary sources, i.e. published
typological surveys based on primary sources, have also occasionally been
used where it seemed hard to avoid, although it is now generally
recognized that comparative linguists should ideally work with primary
sources.

Many dialect atlases have worked with an entirely different method of data
collection, based on questionnaires. These are drawn up by the atlas
editors and filled in by different fieldworkers on location for each data point.
This is practical for dialect atlases because the fieldwork is restricted to a
relatively small area, but for an atlas of global scope like WALS, this
method would have required a budget a thousand times larger. Moreover,
the questionnaire method is problematic for more sophisticated cross-
linguistic work, because identifying certain phenomena (e.g. iambic rhythm
or applicative constructions) in different languages requires detailed
knowledge of the phenomenon in question.

The editors briefly considered the possibility of basing the atlas on

questionnaires sent to a set of experts who know their respective
languages so well that they would be able to answer structural questions
without additional fieldwork. However, it quickly became clear that while
this method would have the advantage of showing a uniform sample of
languages on all maps (as in dialect atlases), it would be difficult to find
enough experts willing to collaborate on such an enterprise, and it would
not be wise to leave the rich data sources of published descriptions
untapped. (However, this method was later used for the APiCS
([Link] (Atlas of Pidgin
and Creole Language Structures) project, which is scheduled to appear in
2012.)
As was mentioned in §3.1, the editors did encourage authors to try and
provide data on a fixed sample of 100 core languages and a further sample
of 100 additional languages. For quite a few of these languages, the editors
contacted experts, asking them to serve as consultants for the WALS
authors. The following 83 people generously agreed to serve as expert
consultants:

Abkhaz
([Link]
George Hewitt
Alamblak
([Link]
Les Bruce
Amele
([Link]
John Roberts
Apurinã
([Link]
Sidi Facundes
Armenian (Eastern)
([Link]
Jasmine Dum-Tragut
Awa Pit
([Link]
Timothy Curnow
Bambara
([Link]
Valentin Vydrine
Basque
([Link]
José Hualde
Batak (Karo)
([Link]
Geoff Woollams
Beja
([Link]
Richard Hudson
Brahui
([Link]
Josef Elfenbein
Chamorro
([Link]
Sandra Chung
Chukchi
([Link]
Michael Dunn
Coos (Hanis)
([Link]
 Matthew Dryer
Cree (Plains)
([Link]
Rose-Marie Déchaine, Christoph Wolfart
Evenki
([Link]
Igor Nedjalkov
Finnish
([Link]
Pirkko Suihkonen
Georgian
([Link]
George Hewitt
Gooniyandi
([Link]
William McGregor
Greek (Modern)
([Link]
Brian Joseph
Greenlandic (West)
([Link]
Jeanette Sakel
Hausa
([Link]
Mahamane L. Abdoulaye
Hebrew (Modern)
([Link]
David Gil
Hindi
([Link]
Anvita Abbi
Hixkaryana
([Link]
Desmond Derbyshire
Hungarian
([Link]
István Kenesei
Hunzib
([Link]
Helma van den Berg
Indonesian
([Link]
David Gil
Ingush
([Link]
Johanna Nichols
Japanese
([Link]
 Akiko Comrie
Kannada
([Link]
S.N. Sridhar
Kayardild
([Link]
Nicholas Evans
Ket
([Link]
Edward Vajda
Khasi
([Link]
Gracious (Grace) Temsen
Khmu'
([Link]
Jan-Olof Svantesson
Kilivila
([Link]
Gunter Senft
Kobon
([Link]
John Davies
Korean
([Link]
Ho-Min Sohn
Koromfe
([Link]
John Rennison
Koyraboro Senni
([Link]
Jeffrey Heath
Krongo
([Link]
Mechthild Reh
Kutenai
([Link]
Matthew Dryer
Lak
([Link]
Konstantin Kazenin
Lakhota
([Link]
Robert Van Valin
Lango
([Link]
Michael Noonan
Latvian
([Link]
 Aleksandra Steinbergs
Lavukaleve
([Link]
Angela Terrill
Lezgian
([Link]
Martin Haspelmath
Makah
([Link]
Matthew Davidson
Malagasy
([Link]
Charles Randriamasimanana
Martuthunira
([Link]
Alan Dench
Maybrat
([Link]
Philomena Dol
Meithei
([Link]
Shobhana Chelliah
Mixtec (Chalcatongo)
([Link]
Monica Macaulay
Ndyuka
([Link]
George Huttar, Mary Huttar
Nez Perce
([Link]
Noel Rude
Ngiyambaa
([Link]
Tamsin Donaldson
Nivkh
([Link]
Elena Gruzdeva
Nunggubuyu
([Link]
Jeffrey Heath
Oneida
([Link]
Karin Michelson
Oromo (Harar)
([Link]
Jonathan Owens
Paamese
([Link]
 Terry Crowley
Pirahã
([Link]
Daniel Everett
Pitjantjatjara
([Link]
Heather Bowe
Quechua (Imbabura)
([Link]
Pieter Muysken
Rama
([Link]
Colette Grinevald
Russian
([Link]
Bernard Comrie
Semelai
([Link]
Nicole Kruspe
Shipibo-Konibo
([Link]
Pilar Valenzuela
Slave
([Link]
Keren Rice
Taba
([Link]
John Bowden
Tagalog
([Link]
Carl Rubino
Trumai
([Link]
Raquel Guirardello
Turkish
([Link]
Jaklin Kornfilt
Vietnamese
([Link]
Mark Alves
Wambaya
([Link]
Rachel Nordlinger
Wari'
([Link]
Dan Everett
Wichí
([Link]
 Kenneth Claesson
Wichita
([Link]
David Rood
Yagua
([Link]
Doris Payne, Tom Payne
Yidiny
([Link]
R.M.W. Dixon
Yimas
([Link]
William Foley
Yoruba
([Link]
Ayo Bamgbose
Yukaghir (Kolyma)
([Link]
Elena Maslova
Yup'ik (Central)
([Link]
Anthony Woodbury

Some authors also contacted other experts and received relevant data by
personal communication from them.

More than 58,000 data points are shown on the WALS maps. Of these,
about 2000 (or 3.4%) are based on personal communications from experts,
about 700 (or 1.2%) are based on the authors’ own knowledge or own
data, and about 400 (or 0.7%) are based on secondary sources. For many
of the data points, the updated online version also provides an example of
the phenomenon in question.

For each data point (i.e. language-feature pair), the source used by the
author is given in the online version. And each bibliographical record in the
WALS Reference Database ([Link] lists the
chapters it is referenced in and links to related languages. Thus, while
being an extensive linguistic bibliography, the reference database does not
aim to contain as many works as possible, but is limited to works which
have actually been cited by authors of the WALS chapters. References can
be exported in various formats (XML, BibTeX, RIS, !EndNote, RDF+DC) and
thus integrated in other (e.g. personal) collections of bibliographical data.

5. Disclaimer
In identifying the status of speech varieties as languages or dialects, in
assigning names to languages and dialects, in identifying countries, and in
locating languages in countries, we have been guided solely by practical
considerations and by current scholarly practice. In no instance should our
usage be taken as implying a particular political stance or as insulting the
speakers of a particular speech variety.

Privacy Policy
([Link]
([Link]
[Link])
WALS Online edited by Dryer, Matthew S. &
Disclaimer
Haspelmath, Martin
([Link] ([Link]
is licensed under a Creative Commons Attribution
  Application sourc
4.0 International License
(v2014.2-187-g835fb56
([Link]
on

([Link]