(ly91-- ) CS't\~LL« h~~hu.

~ Ivtk~thlJ--1
h~(;.t.~e

9
The classification of
languages
Aleksandra Steinbergs

Everything it is possible for us to analyse depends all a clear method which distinguisht'';
the similar from the not similar.
~ Linn.ceus, Genera Plantarum (1754)

In the world today there are thousands of different languages. each with its
own sound patterns, grammar and vocabulary, However, regardless of how
different these languages are, they have important similarities that aIlo\\­
linguists to group them into a relatively small number of families and types.
This chapter describes the methods of classification linguists use. and some
of the findings which have resulted from this type of research.
I

1
SOME We will begin by considering two topics - the problem of distinguishing
PRELIMINARIES between a language and a dialect, and the chief methods of language
classification used in linguistics today.
1.1
Dialect and It is often difficult to determine whether two linguistic communities speak
language different languages or merely different dialects of the same language. One test
that linguists use to decide this involves the criterion of mutual intelligibility.
Mutually intelligible varieties of a language can be understood by speakers of
each variety. According to this criterion, the English of London, the English
of Edinburgh, the English of Cape Town, the English of Sydney, the English
of Toronto and the Englisb of Miami qualify as dialects of the same 1anguage.
On the other hand, if two speakers cannot understand one another, then
linguists normally conclude that they are speaking different languages. The
Italian of Florence and the French of Paris are examples of varieties of speech
that are not mutually intelligible.
Political, cultural, social, historical, and religious factors often intervene in
the determination of linguistic boundaries. For example, Serbs and Croatians
often maintain that they speak different languages. However, even though
their history, religion, and spelling systems differ, standard Serbian and
standard Croatian are mutually intelligible dialects of the same language.

____________________ J
 THE CLASSIFICATION OF LANGUAGES 373

which linguists call Serbo-Croatian. In contrast, we often speak of Chinese as

if it were a single language, even though, according to the mutual
intelligibility test, it is actually a number of separate languages (Cantonese.
Mandarin, Hakka, and so on), each with a multitude of dialects of its own.
In addition In the prohlel11~ prescnted hy thesc 1l(\Il-lingui:-;ti,' l','n~jdcr:~
tions. complications also arise when we trv to divide a continuulll or mutual"
intelligible dialects whose two end p()i,~ts arc not iJlte!li~ihk. Dutch and
German. for example. arc mutually intelligible around -the horder area
between Germany and Holland: however. the uutch of Amsterdam and the
German of Munich are not. Similarly. Palestinian Arabic and Syrian Arahic
arc mutually intelligible, but Moroccan Arabic and Saudi Arabian Arabic are
not.
Taking these considerations into account, how many distinct languages are
there in the world today? The best available estimate places the currem figure
at aboutsix thousand. However, many of these languages have onlY a few
hundred speakers and many others are in grave danger of de~ise as
indigenous peoples throughout the world lose their traditional culture and
homelands. Indeed, according to one estimate, only around three hundred of
the world's languages have a secure future.

1.2

Types of Within the field of linguistics, three different approaches to language

classification classification are used.
Genetic classification categorizes languages according to their descent.
Languages that developed historically from the same ancestor language are
grouped together and are said to be genetically related. This ancestor may he
attested (that is, texts written in this language have been discovered or
preserved, as in the case of Latin), or it may be a reconstructed hypothetical
proto-language for which no original texts exist (as is the case for Proto-Indo­
European). Genetic classification is discussed further in section 3.
Although genetically related languages often share structural character­
istics, they do not necessarily bear a close structural resemblance. For
example, Latvian and English are genetically related (both are descended
from Indo-European), but their morphological structure is quite different. An
English sentence like It has to be figured out/ascertained can be expressed in
Latvian by a single word.
I)
jaizgudro
(one) must out figure (it)
'It has to be figured out.'
Of course, Latvian and English are very distantly related, and languages that
are more closely related will typically share a larger number of similarities.
On the other hand, it is also necessary to recognize that even languages that
are totally unrelated may share some structural properties. Thus, English and
Swahili, which are unrelated, both employ subject-verb-ooject word order in
simple declarative sentences.
374 CONTEMPORARY LINGUISTICS

2)
Maria anapenda Ali
'Maria likes Ali.'
For these reasons, a second approach to language classification is required
Known as linguistic typology. it classifies languages ~ILLorJing to LlIL'iI
structural characteristics without regard for genelic relationships. Thu:-;.
typologists. might group together languages with similar sound pallcrns or.
alternatively. ones with similar word order. Typological studies also seck 1\\
identify linguistic universals, structuralcharactcristics which occur acros~
the languages of the world. We return to this type of classification in section
2.
Finally, areal classification identifies characteristics shared by languages
which are in geographical contact. Languages in close contact often bOITO\\
words, sounds, morphemes, and even syntactic patterns from each other. As
a result, neighbouring languages can come to resemble each other even
though they may not be genetically related. Due to space considerations, this
chapter will not deal with areal classification specifically; however, a
description of borrowing can be found in sections 1.2 and 5.1 of Chapter 8.

2
TYPOLOGICAL As already noted, the classification of languages according to their structural
CLASSIFICATION characteristics is known as linguistic typology. Typological research groups
together languages on the basis,' of similarities,'iJl their word order patterns.
morphological structures, or phonological systems. An important area of
research within the study of linguistic typology involves the search for
linguistic universals - structural patterns and traits that occur in all or most
human languages. Patterns or traits that occur in all languages are called
absolute unive...sals, whereas those that occur in mostlanguages are known
as universal tendencies;
Many typological generalizations involve implicational universals,
which specify that the presence of one.trait implies the presence of another
(but not vice versa). For instance, languages with fricative phonemes (such as
IfI and lsI) will also have stop phonemes (such as Ipl and Itl), although the
reverse is not necessarily true.
Another way to analyse linguist~c universals is through markedness
t-beury. Within this theory, marked traits are considered to be either more
complex and/or universally rarer than unmarked characteristics. In addition.
a marked trait is typically found in a particular language only if its unmarked
counterpart also occurs. Thus, markedness theory is closely related to the
study of implicational universals.
An example can help to clarify these terms. Nasalized vowels are said to
be marked, while non-nasalized (oral) ones are said to be unmarked.
Phonologically, oral vowels can be considered less complex: oral vowels
allow the airstream to exit only through the mouth, while nasal vowels allow
air to escape from both the mouth and the nose. Cross-linguistically, we find
 THE CLASSIFICATION OF LANGUAGES 375

that all languages have oral vowel phonemes, while only some languages
have nasal vowel phonemes. Moreover. even in the languages which have
both, there are usually fewer nasal vowels than oral ones. Thus. nasal vowels
(which are considered to be marked) are both rarer and phonologically more
complex than oral V()\\"('js (which are unmarked)
The following sections present some of the typologj.:al classifications and
universals that have been proposed in the areas of phonology. morphology.
and syntax.
2.1

Phonology In this section, we represent all vowel and consonant systems phonemically.
This simplifies the presentation; note. however, that the exact phonetic
realization of these systems may vary in the individual languages.

Vowel systems Languages are often classified according to the size and pattern of their vowel
systems. The most common vowel system has five phonemes - two high
vowels, two mid vowels, and one low vowel (Figure 9. I ). The front vowels
are unrounded, as is the low vowel, and the back vowels are rounded.
u

e o

a
Figure 9.1 The most common vowel system.

About half the world's languages, including Basque. Hawaiian, Japanese,

Spanish, and Swahili, have such a system.
The majority of the world's other languages have three, four, six, seven,
eight or nine different vowel phonemes (disregarding contrasts based on
length or nasalization. which can double or triple the number of phonemiC
vowels). Languages with fewer than three, or more than nine distinctive
vowels are rare. Some typical vowel systems are presented in Fi~ure 9.2.
u u

e o e

a a A

Three vowel system Four vowel system Seven vowel system

Gudanji (Australia) Navaho (Arizona) Geez (Ethiopia)

Figure 9.2 Common vowel systems.

" ' - - - _ .. __ ._-­

376 CONTEMPORARY LINGUISTICS

Analysis of many languages has led to the discovery of a number of

universal tendencies with respect to vO\\'el systems. Some of these tendencies
are listed here. along with a description of thc most commonly occllrrjn~
vowels.
• The l11ust cUllImunly occurring Vll\\L"J pllUIlL·jlll' i:- ia!. \\hicli i:- iI/til,,; III
almost all of the languages of the world. The v(l\\L'ls Iii and lui are alllldS\
as common as la/.
• Front vowel phonemes (Ii, e. L. a:i) ~tre generally unrounded. whik 11')11·
low back vowel phonemes (h. o. u/) are generally rounded,
• Low \'owels (la:.. a. u/) are generally unrounucd.
Although English has an above-average numher of vowels. these phonemes
all comply with the above tendencies. Thus, RP English has only front
unfounded vowels, all the low vowel phonemes are unrounded, and all of the
back, non-low vowels are rounded. The only asymmetry found is that there
is no front unrounded vowel corresponding to 10/. The English vowel system
(excluding length, diphthongs and schwa) can be presented as shown in
Figure 9.3. (Here the vowel in set is represented as 1r:.1 rather than leI in order
to highlight the symmetry; see the discussion on pages 37 and 38.)
u

J\

a
Figure 9.3 The English vowel system.

The relationship among contrasting vowel types (such as oral vs nasal and
long vs short) can also be expressed interms of implicational universals, since
the presence of one vowel phoneme type implies the presence of another (but
not vice versa).

• If a language has contrastive nasal vowels, then it will also have

contrastive oral vowels. For example, "f'rencb contrasts' different nasa1
vowels ( 1151 'long' and /lfJJ 'slow'), and contrasts oral vowels with nasal
vowels (/laJ 'weary' and I1fJJ 'slow'). Predictably, French also contrasts
different oral vowels, as in /klol 'shut' and /klul 'nail'. English shows
contrasts among oral vowels but does not contrast nasal vowels with oral
vowels. There are no contrasts in English like /bot! and /bot!.
• If a language has contrasting long vowels, then it will also have
contrasting short vowels. For example, Finnish shows contrasting long
vowels, and, predictably, contrasting short vowels. The reverse is not
necessarily the case. Swahili shows contrasting short vowels (Ii e a 0 uI)

-.----------------~~

~ ."
 THE ClASSIFICATION OF LANGUAGES 377

but does nOl contrast long vowels with short ones, although the sccond­
to-last syllable of a word, which is stressed. is lengthened allophonically.

Table 9.1 Finnish vowel contrasts

Long versus long Ivi:lif 'junket" fya: Iif -l'icctinll'

Short versus short /sukaJ 'bristle' /suku/ . lam i Iy'
Short versus" long ftuli/ 'fire' ftu: Iif -wind-

Consonant systems It is not particularly useful to classify languages according to the number of
consonants that they contain, since languages may have as few as eight
consonant phonemes (as in Hawaiian), or more than ninety_ (!Kung. a
language spoken in Namibia, has ninety-six consonant phonemes.) Never­
theless, typological analysis of consonant systems has produced a number of
well-substantiated universals.
• AU languages have stops.
• The most common stop phonemes are fp, t, kI. Very few languages lack
anyone of these, and there are no languages that lack all three. If anyone
of these three stops is missing, it will probably be fpf; for example, Aleut,
Nubian, and Witchita have no Ipl phoneme. The .most commonly
occurring phoneme of the three is It/.
• The most commonly occurring fricative phoneme is lsi; few languages
lack it. If a language has only one fricative,itis most likely to be lsi. This
is the only fricative found in Nandi (a language of Kenya) and Weri (a
language of New Guinea). The next most common fricative is IfI.
• Almost every known language has at least one nasal phoneme; in cases
where a language has only one nasal phoneme, that phoneme is usually
In! (as in Arapaho, spoken in Wyoming). If there are two contrasting
nasals, they are normally 1m! and In!.
• The majority of languages have at least one phonemic liquid. However,
a small number of languages have none at all (Dakota, Blackfoot, which
is spoken in Alberta, Saskatchewan and Montana, Efik, which is spoken
in Nigeria, and Siona, which is spoken in Ecuador). English, of course,
has two: 111 and Ir/,
Consonantphonemes,,arealso subject to various implicational universals.
• If a language has voiced obstruent phonemes (stops, fricatives, or
affricates), then it will also have voiceless obstruent phonemes. The
reverse is not necessarily true; for example, Ainu (a language of northern
Japan) has only voiceless obstruent phonemes: Ip, t, k, tf, s/.
• Sonorant consonants (nasals, liquids, and glides) are generally voiced.
Very few languages have voiceless sonorants and those that do always
have voiced sonorants as well. For example, Burmese contrasts voiced
and voiceless nasals and laterals.
• If a language has fricative phonemes, then it will also have stop

a
380 CONTEMPORARY LINGUISTICS

• If a language permits sequen s of consonants in a syllable onset. then it

will also permit single con nants or zero consonants in an onset.
• If a language pennits s uences of consonants in a syllabic coda. then it
will also permit sin y consonants or zero consonants in a coda.
1..1

Morphology Words and the morpheme arc both cross-linguistic;!l!) kgililllale L'alcgoril'~
Howcvcr. there arc clear differences in the W;t\" ill which indi\"idLl~lI
languages combine morphemcs to form v'ords. FnllJ types of svstems arl'
usually distinguished.

The isolating type A language which is purely isolating oranalyticu'ouldcomain only words
that consist of asi l1gle (root) morpheme. In such a languag..::there would be
no affixes, and categories such as number and tense would therefore have to
be expressed by a separate word. For instance, in Mandarin Chinese, which
is primarily an isolating language, the morpheme Ie is often used to indicate
that an action was done in the past, or is completed. Although this morpheme
is thus semantically similar to a past tense morpheme in English. it acts jU~I
like an independent word since its position in a sentence may vary - it can
be separated from the verb with which it is semantically associated.
3)
a. Ta chi Ie fan.
he eat past meal
'He ate the meal.'
b Ta chi fan leo
he eat meal past
'He ate the meal.'
Other languages that are primarily isolating include Cantonese,Vietnamese.
Laotian, and Cambodian.

The polysynthetic In a polysynthetic language, single words can consist of long strings efroots
type and affixes that often express meanings associated with entire sentences in
other languages. The following word from Inuktitut illustrates this.
4)
Qasuiirsarvigssarsingitluinarnarpuq.
Qasu -iir -sar -vig -ssar -si -ngit -luinar -nar -puq
tired not cause-to-be place-for suitable find not completely someone 3/sg
'Someone did not find a completely suitable resting place.'
Polysynthetic structures are found in many native languages of North
America, including Inuktitut, Cree, and Sarcee.
The terms isolating and polysynthetic refer to two extremes: words
consisting only of single morphemes versus words which can be complete
 THE CLASSIFICATION OF LANGUAGES 381

sentences. Few if any languages are either purely isolating or purely

polysynthetic. Rather, the vast majority of languages are synthetic. in that
they admit words containing several morphemes. Two types can bc dis­
tinguished.

The agglutinating An agglutinating language has words which canc()l1lain several morphemcs.
type but the words are easily divided into their component parts (normallv a rool
and affixes). In such languages, each affix is clearly idcl1lifiahle and t)'pically
represents a single grammatical category or meaning.
Turkish words can have a complex morphological structure, but each
morpheme has a single clearly identifiable function. In the words in Table 9.5,
for instance, -ler marks plurality, and -ill marks the genitive (which is used
to indicate the possessor).

Table 9.5 Affixes in Turkish

k~j •village'
k~j-Ier 'villages' (plural)
k~j-Ier-in •of the villages' (plural, genitive)

The fusional type Words in a fusional, language can also consist of several morphemes.
However, in contrast to agglutinating systems, the affixes in fusional
languages often mark several grammatical categories simultaneously. In
Russian, for example, a single inflectional affix simultaneously marks the
noun's gender class (masculine, feminine, or neuter), its number (singular or
plural), and its grammatical role (subject, direct object, and so on). The suffix
-u, for instance, can be used to indicate that a noun belongs to the feminine
gender class, is singular, and functions as direct object.

5)
mi vidjim ruk-u.
we see hand-femlsg/Ac
'We see althe hand.'
This situation is typical of the entire Russian case system.
1ile-distinctien between agglutinating and fusional is sensitive to the
number of 'bits' of infonnation nonnally packed into an affix; in an
agglutinating language, each affix nonnally contains only one element of
grammatical or lexical meaning, while in a fusional language, affixes
nonnally have several simultaneous functions.

Mixed types Many languages, including English, do not belong exclusively to any of these
four categories. English employs isolating patterns in ma.'1Y verbal construc­
tions, where each notion is expressed by a separate word. Futurity, for
382 CONTEMPORARY LINGUISTICS

instance, is indicated by the free morpheme will (rather than an aftix) in

structures such as I will leave. In contrast. English exhibits considerahle
agglutination in derived words, such as re-el1-acl-ment, which consist of a
series of clearly identifiable morphemes. each with its own unique llleaning
and fUllctioll. TIll' English pronoun systelll. on tIll' other han(l. is lar!!l'h
fusional since a single form can be lIscd to indicate person. number. gentle!.
and casco The \\"()rd him. for instance. is lIsed to express a third persoll.
singular. !l1asculine direct object.

Implicational A variety of generalizations can be made about \\-ord structure in humall

universals: language.
morphology
• Ifa language has intlectional affixes. it will also have derivational affixes.
For example, English not only has inflectional affixes such as the past
tense -ed and plural -s, but it also contains derivational affixes like 111/·
(unhappy, unwanted) and -ly (quickly, slowly).
• If a word has both a derivational and an inflectional affix (Table 9.6). the
derivational affix is closer to the root. (DA = derivational affix: IA =
inflectional affix.)

Table 9.6 The ordering of derivational and inflectional affixes

English
friend-ship-s *friend-s-ship
Root DA fA Root fADA

Turkish
if -tfi -ler *if -ler -tfi
work -er -pi work -pI -er
Root DA fA Root fA DA

• If a language has only suffixes, it will also have onlypostpositions. (As

noted in Chapter 5, postpositions are the equivalent of prepositions in
languages that place the head at the end of the phrase.) Turkish, for
example, has only suffixes; as expected, it also has postpositions rather
than prepositions. This is illustrated in the following sentence.
6)
ahmet ajJe itJin kitab-w aI-dw
Ahmet Ayshe for book-Ac buy-Pst
'Ahmet bought the book for Ayshe.'
2.3
Syntax Because we lack detailed descriptions for most of the world's languages.
much of the work on syntactic universals has been restricted to the study of
word order in simple declarative sentences such as The women built the house.
Patterns are classified in terms of the order of the subject (S), direct object
(0), and verb (V). The three most common word orders (in descending order

--------- ------- -- ------- -----_._------ - ~

r
THE CLASSIFICATION OF LANGUAGES 383

of frequency) are SOY, SVO. and VSO. Over 95 per cent of the world's
languages usc one of these patterns as their basic word order.
7)
SOY (Turkish I:
Hasan oktiz-u al-dw
Hasan ox-Ac buy-Pst
'Hasan hought (he ox.'
8)
SYO (English):
The athlete broke the record.
9)
VSO (Welsh):
Lladdodd y ddraig y dyn.
killed the dragon the man
'The dragon killed the man.'
SOY, SVO, and VSO patterns have one property in common: in all three
cases, the subject occurs before the direct object. The prevalence of this type
.
of pattern may stem from the fact that the subject usually corresponds to the
.
topic (what the sentence is about - Chapter 7, section 4.3), and therefore is
more useful at an early point in the utterance.
Although an overwhelming majority of the world's languages place the
subject before the direct object in their basic word order. this practice is not
universal. There is a small number of VOS.languages, the best-known
example of which is Malagasy.
10)
VOS (Malagasy):
Nahita ny mpianatra ny vehivavy.
saw the student the woman
'The woman saw the student.'
As well, there are a very few OVS and OSV languages. all of which seem to
be spoken in the Amazon basin of South America.
11)
OVS (Hixkaryana):
Kana yanimno biryekomo
fish caught boy
..
.:
'The boy caught a fish.'
12)
OSV (Apurina):
Anana nota apa
pineapple 1 fetch
'1 fetch a pineapple.'

--_ .. - ---_. - - - - --_. -----------­

 ·,
384 CONTEMPORARY LINGUISTICS

Word order Sometimes, the order of elements within one kind of structure has implica­
universals tions for the order of elements in other structures. Many of these implications
concern the relationship helween the verh and its (direct) ohject.

• If a language has YO word order. then it will have prep()~itiolls rather 111;111
postposilions. Languages of this type include Berbcr \ spokcn in
Morocco), Tagalog (the Philippines), Maori (0:ew Zealand). Maasai
(Kenya). Welsh. and Irish Gaelic.

13) Irish Gaelic

a. VSO pattern:
Chonaic me rno mhathair
saw I my mother
'I saw my mother.'
b Preposition pattern:
sa teach
in house
'in the house'

• Ifa language has OV wordprder, then it will probably have Pos(positions

rather than prepositions. Languages with this structural pattern include
Basque, Burmese, Hindi, Japanese, Korean, Quechua, Turkish. and
Guugu Yimidhirr, an aboriginal language of Australia.
14) Guugu Yimidhirr
a. S-OV pattern:
Gudaa-ngun yarrga dyinday.
dog-Erg boy bit (Erg= Ergative, see Chapter 4, section 6.3)
'The dog bit the boy'

b. Postposition pattern:
yuwaal nganh
beach from I
II
'from the beach'
• PPs almost always precede the verb in OV languages, and usually follow
the verb in VO languages.

15) Japanese
a. S-OV pattern:
Gakusei-ga hon-o yonda
I
student-Nom book-Ace read
'The student read a book'
b. PP precedes verb:
Taroo-ga [pp nitiyoobi nil tsuita
Taroo-Nom Sunday on arrived
'Taroo arrived on Sunday.'

j
~
.~

---------------------------------------------------------------- J
 THE ClASSIFICATION OF LANGUAGES 385

16) English
o. s-vo pattern:
Harnish likes haggis.
1>, pp ycrh:
f()ll()\\'~

Georgina left I""on Sundayl.

• Manner adverbs overwhelmingly precede the verb in OV languages and
generally follow the verb in VO languages,
17) Japanese (S-OV pattern, as seen in 15a)):
Manner adverb precedes verb:
hayaku hasiru
fast run
18) English (S-VO pattern, as seen in 16a)):'
Manner adverb follows verb:
Sally runs fast.
• With respect to possessive structures, there is an oYerwhelming prefer­
ence for Genitive + N order in OV languages, and a (somewhat weaker)
preference for N + Genitive order in VO languages.
19) Japanese (S-OV pattern, as seen in 15a)):
Genitive structure precedes head N:
Taroo-no hon
Taroo-Gen book
'Taroo's book'
20) French
a. S-V 0 pattern
Pierre aime Marie
'Pierre likes Marie. '

b. Genitive structure follows head N:

la maison de Marie
the house of (Gen) Marie
'Marie's house'
English, although an SVO language, exhibits both Gen + Nand N + Gen
'Patterns:
21)
a. Gen + N pattern:
Mary's friend
b. N + Gen pattern
a friend of Mary
Examples like 21) do not invalidate the universals we have presented, since
these universals are intended to capture tendencies only.

-------- .'--- -," '-' '

 386 CONTEMPORARY LINGUISTICS

Grammatical Implicational universals are often stated in terms of hierarchies of categories

hierarchies or relations. One of the most important hierarchies of this type refers to the
grammatical relations of suhject and direct ohjcct (see Chapter 5),
Hierarchies represent degrees of markedness. with the least marked option
al the tup and tk most lll:u-kcd at the hnlloll1 According tn the hierarchy ill
Figure 9.4. then. a process that applies only to subjects is less marked than a
process that applies to direct objects, and so on. Given the definition of
markedness outlined at tlk' beginning of section ::'. this entails thai if :1
particubr phenomenon applies to direct objects. it should also apply In
subjects. The converse. hO\\'c\,er, need not be truc: il would not he surpri\ill~
to find a process that applies to subjects but not to direct objects.

Least marked j Subject

Direct object
Other
Most marked

Figure 9.4 Hierarchy of grammatical relatiolls.

Among the many typological phenomena that comply with this hierarch:­
is verb agreement. first mentioned in Chapter 4 (section 6.4). As the following
examples show, there are languages in which the verb agrees with just the
subject and there are languages in which it agrees with both the subject and
the direct object. (3 = 3rd person, sg = singular, pI = plural, pst= past)
22)
Agreement with subject only (Spanish):
Subject I
Juan parti-6.
Juan leave-3/sglPst
'Juan left. '

23)
Agreement with subject and direct object (Swahili):
Subject I I Direct ~bject
Juma a- li- wa- piga watoto
Juma 3/sg Pst 3/pl hit children
'J uma hit the children. '
However, as predicted by the hierarchy, however, there are no languages in
which the verb agrees with just the direct object.

2.4
Explaining universals Linguists are still uncertain about how to explain the existence of many
linguistic universals. Nonetheless, a number of interesting proposals hay\!
been made, and it is worthwhile to consider some of them here.

-.----- ---_.- . ----_._-_._------­ _. _ _ _ _ _ _ _ _ -;I

 THE CLASSIFICATION OF LANGUAGES 387

Phonology Perceptual factors play a role in shaping phonological universals.. For

example, the fact that lsI is the most commonly occurring fricative may ha\'(~
to do with its acoustic prominence: varieties of [sJ are inherently louder than
other kinds of fricatives.
VO\',cl systems (discussed in SL'L'tioll 2, I) <.In"clop \(1 ;!S to kl'C!' \"()\\l:
phonemes as different from each other as possible. A three-vowel system stich
as that shown in Figure 9.5 allows for plenty of 'space" around eaeh vowel.
\\"hich may make each phoneme easier to distinguish from the others.

LI

Figure 9.5 A three-vowel system.

The same holds true of stop phonemes. It may be that Ipl, ItI, and Ik/ are
the three most common stops because they occur at three maximally distant
places of articulation within the supralaryngeal vocal tract. These stops are
perceptually much easier to distinguish from each other than a sequence of
dental, alveolar and palatal stops, for example, all of which are produced in
the central region of the oral cavity. (See Figures 2.1 and 2.4 in Chapter 2.)
It has been recently suggested that consonant systems in general respond
to the articulatory pressures that give rise to unmarked sounds and systems.
Thus, for example, articulatorily basic obstruents such as [p], [t], and [k] are
found much more widely than more complex articulations such as [til and
[qW). Table 9.7 shows the set of obstruents that is exploited first cross­
linguistically.

Table 9.7 Obstruents found cross-linguistically

P t k ?
b d g
f s h
tf

Languages tend to have consomrnt"systeutS'dlat corrsisrdfm,out 7fJ percent

obstruents and 30 per cent sonorants, no matter what the total size of their
consonant inventories may be. These figures appear to reflect the kind of
space that is available for consonant contrasts: more distinctions can be made
among obstruents than among sonorants. There are, for example, no nasal
fricative sonorants, because the air pressure needed to force air through the
narrow articulatory opening characteristic of fricatives cannot be built up
when so much air is flowing through the nasal passage at the same time. For
reasons such as this, the number of obstruent consonants in any language is
potentially larger than the number of sonorant consonants. This is just one
 T
J<.~.. ~
388 CONTEMPORARY LINGUISTICS ·1
.t':
example of how considerations involving articulatory complexity can playa
role in the shaping of consonant systems.
t .'

Other Iypes of explanations Illay he appropriate for morphnlogicalunivcrsals.

Morphology
For example. the fact that languages with sulfixes but no prdixl..~s always h;!vL'
postpositions (section 2.2) may have a historical explanation. III these
languages. some postpositiolls may have gradually hecome attached to a
preceding word and thus eventually became suffixes. Because suffixes in such
languages have evolved from postpositions, the link between the two
elements can be traced to their common origin.
The requirement that derivational affixes occur closer to the root than
inflectional affixes has another type of explanation. As noted in Chapter 4.
derivation typically changes the category of the base to which it applies
whereas inflection simply marks the subclass (for example, plural for Ns, past
tense for Vs) to which a word belongs. Given that a word's category must be
known before its subclass can be determined, it follows that derivational
processes should precede inflection. This is reflected in word structure, where
derivational affixes appear closer to the root than inflectional markers. In
Figure 9.6, for example, the verbal root treat is converted into a noun by the
affix -ment before the plural inflectional marker is added.
N

v
A~ Af Af
I
treat
I
ment
I
s
Figure 9.6 The structure of a word containing a derivational affix and an inflectional affix.

Syntax At least some syntactic universals may be explained in terms of the way that
the human brain processes sentence structure. Consider the summary of word
order patterns in Table 9.8, which is based on the implicational universals
discussed in section 2.3.

Constituents Order in VO language Order in OV language

P&NP preposition-NP NP-postposition

V&PP verb-PP PP-verb
V & manner Adv verb-mannerAdv Manner Adv-verb
Gen&N noun-genitive genitive-noun

On~ recent explanation as to why the word order properties in the second
and thIrd columns cluster together involves contrast between right-branching