Entdecken Sie eBooks
Kategorien
Entdecken Sie Hörbücher
Kategorien
Entdecken Sie Zeitschriften
Kategorien
Entdecken Sie Dokumente
Kategorien
1. Teilband / Volume 1
Vorwort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXI
IV. Grundbegriffe
Basic concepts
23. Pierre Swiggers, Linguistic sign . . . . . . . . . . . . . . . . . . . ... 210
24. Jaap van Marle, Paradigmatic and syntagmatic relations . . . ... 225
25. José Luis Iturrioz Leza & Stavros Skopeteas, Variation und
Invarianz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
26. Laurie Bauer, Word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
27. William Croft, Lexical and grammatical meaning . . . . . . . . . . 257
28. Andrew Carstairs-McCarthy, Category and feature . . . . . . . . . 264
29. Linda R. Waugh & Barbara A. Lafford, Markedness . . . . . . . . 272
30. John Haiman, Iconicity . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
31. Wolfgang U. Dressler, Naturalness . . . . . . . . . . . . . . . . . . . . 288
32. Ferenc Kiefer, Regularity . . . . . . . . . . . . . . . . . . . . . . . . . . 296
33. Geert Koefoed & Jaap van Marle, Productivity . . . . . . . . . . . 303
VII. Allomorphie
Allomorphy
46. Hans Christian Luschützky, Morphem, Morph und Allomorph . 451
47. Martin Neef, Phonologische Konditionierung . . . . . . . . . . . . . 463
48. Martin Neef, Morphologische und syntaktische Konditionierung 473
49. Henriette Walter, Fluctuation and free variation (Translation:
Marie Landick) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
50. Tsutomu Akamatsu, Generalized representations . . . . . . . . . . . 489
51. Edmund Gussmann, Underlying forms . . . . . . . . . . . . . . . . . 499
52. Igor Mel’čuk, Suppletion . . . . . . . . . . . . . . . . . . . . . . . . . . 510
IX. Flexion
Inflection
62. Andrew Carstairs-McCarthy, Lexeme, word-form, paradigm . . . 595
63. Aleksandr V. Bondarko, Meaning vs. use in inflection . . . . . . . 607
64. Richard Coates, Exponence . . . . . . . . . . . . . . . . . . . . . . . . 616
65. Andrew Carstairs-McCarthy, Inflection classes . . . . . . . . . . . . 630
66. Silvia Luraghi, Synkretismus (Übersetzung: Simona Fabellini) . . 638
67. Fred Karlsson, Defectivity . . . . . . . . . . . . . . . . . . . . . . . . . 647
68. Martin Haspelmath, Periphrasis . . . . . . . . . . . . . . . . . . . . . 654
X. Wortarten
Word classes
69. Kjell-Åke Forsgren, Wortart, syntaktische Funktion,
syntaktische Kategorie . . . . . . . . . . . . . . . . . . . . . . . .... 665
70. Clemens Knobloch & Burkhard Schaeder, Kriterien für die
Definition von Wortarten . . . . . . . . . . . . . . . . . . . . . . . . . 674
71. Barbara Kaltz, Wortartensysteme in der Linguistik . . . . . . . . . 693
72. Nicholas Evans, Word classes in the world’s languages . . . . . . . 708
73. Christian Lehmann & Edith Moravcsik, Noun . . . . . . . . . . . . 732
74. D. N. S. Bhat & Regina Pustet, Adjective . . . . . . . . . . . . . . . 757
100. Numerus
101. Mass and collection
102. Case
103. Possession
Thus, we can have compounds such as com- gist needs to address, before reviewing the
puter manufacturer and phrases such as man- theoretical positions that have been taken up.
ufacturer of the kind of computer you can put The most obvious ways in which morphol-
in a small suitcase. However, we cannot con- ogy and syntax interact are seen in inflec-
struct compounds of the form *kind of com- tional systems. One of the primary functions
puter you can put in a small suitcase manufac- of inflection is to signal dependencies be-
turer. On the other hand, phrases are allowed tween syntactic units, for instance, the syn-
inside compounds to a limited extent, as is tactic relationships of agreement and govern-
discussed below. ment. Grammatical theory therefore needs to
Hence, in the real world we expect two find a way of explicitly relating the morpho-
types of deviation from ideality. First, there logical form with the syntactic construction
are cases in which parts of words seem to be or relationship.
manipulated by syntactic processes. Second, Syntactic constructions in one language
there are cases in which morphological pro- frequently correspond to morphological con-
cesses (or word formation processes) seem to structions in another. For instance, the de-
be defined over phrases which have been pendencies expressed by inflection in one lan-
formed by the syntax. guage may be expressed by other lexical and
At the same time, both syntax and mor- syntactic means, such as periphrasis, word
phology can be thought of as governing the order and so on, in another. Grammatical
distribution of formally and semantically theory must be able to state the commonali-
identifiable units, namely, words and mor- ties in such constructions, independently of
phemes. Just as syntax concatenates words morphosyntactic expression. This problem
into phrases, so morphology concatenates becomes particularly acute when similar rela-
morphemes into words. Even in the absence tionships are expressed by morphology in
of ‘deviations’ of the kind just mentioned, it one construction and syntactically in an-
is therefore possible that consideration of other, closely related construction. For exam-
point (iv) above will show that similar or ple, grammatical theory has to have some
even identical principles govern both compo- way of relating the periphrastic comparative
nents of grammar. This assumption is en- form of adjectives such as more bold with the
shrined in the notion the syntax of words (cf. synthetic, morphological construction, bolder
Art. 42), and it has been proposed by a (cf. Art. 68).
number of theorists that, for example, words Word formation processes may alter the
have a constituent structure with head-com- syntactic properties of words. This is trivially
plement and modifier-head constructions, true in the case of derivation which changes
just as in syntax. syntactic word class. However, there are
Further questions that arise from points more subtle changes possible when syntactic
(ii) and (iii) relate to the extent to which mor- word class is not affected. The most well-
phological rules and principles can appeal to known of these concern argument structure,
syntactic structures, and the extent to which manifested as valency changes and voice al-
syntactic rules and principles might be sensi- ternations (cf. Art. 107, 108). In many lan-
tive to morphological structures. For exam- guages a causative construction exemplified
ple, is it possible to find morphological rules by Tom made Harriet eat the apple is ex-
of the form “choose allomorph X of verb in- pressed by affixation of a causative mor-
flectional morpheme M if the verb precedes pheme to the verb, turning it from a two-
the subject in surface structure, and allo- place predicate into a three-place predicate.
morph Y otherwise” (cf. Hudson 1977). Like- On the other hand, many languages have a
wise, do there exist syntactic processes of the productive construction (for instance, a pas-
form “adjoin the verb to the left of the sen- sive or an antipassive) in which valency is de-
tence if and only if it has a monomorphemic creased by one argument. Other types of
stem”? Unlike most of the other interactions, word formation may affect argument struc-
these are controversial. In most theories, pu- ture. For instance, in English it is possible to
tative examples are likely to be analyzed prefix out- to certain intransitive (or intransi-
away as reflexes of something different. tively used) verbs to produce a transitive verb
In examining any particular set of con- with the meaning ‘to verb to a greater extent
structions, we often see more than one of than’, e.g. Tom outran all the other competi-
these three broad types of interaction. I shall tors, Tom outate Dick. At the very least gram-
therefore briefly survey some of the com- matical theory has to make available a frame-
moner sets of situation which the morpholo- work of description in which operations of
this sort can be stated, and their syntactic ef- Turkish, for instance, the so-called indefinite
fects predicted. izafet (Lewis, 1967) is a construction in which
It is common for one language to express the head of the expression appears to agree
modifier-modified relationships by means of with the modifier in person/number features.
compounding, where another language might For instance, in (5) -i is identical to the 3rd
employ syntactic means. For example, an ex- singular possessive formative, so that the
pression such as water-mill is translated as whole expression appears to mean ‘Istanbul
vodjanaja mel’nitsa in Russian, where vodja- its-mosques’:
naja is a relational adjective (with the -n- for- (5) Istanbul cami-ler-i
mative characteristic of adjectives) derived Istanbul mosque-pl-3.sg.poss
from the noun voda ‘water’. This typifies the ‘Istanbul mosques’
difference between the Slavonic languages However, in many respects such an expres-
and the Germanic languages with respect to sion is more like an English compound noun.
this sort of construction. Intermediate cases For instance, it is not possible to modify the
are also found, in which it is not entirely clear word camileri syntactically. In that case a dif-
whether we are dealing with syntactic con- ferent construction must be used. The modi-
structions or morphology. The construct (or fier Istanbul must appear in the genitive case
izafet) of languages such as Semitic, the Ira- (forming the so-called definite izafet), as seen
nian languages and the Turkic languages in (6), where ‘mosques’ is modified by an ad-
(among others) are typical in this regard. In jective:
Chukchee is an ergative language, and the tains a degree of referentiality. Thus, although
direct object in (7b) is in the absolutive case. (7a) would normally be used to mean some-
There is no doubt that it is a separate noun thing like ‘he is doing some knife-sharpen-
phrase. Equally, however, in (7a) the noun ing’, it could be used in principle to refer to
root wala forms part of the verb stem: it is some specific knife.
preceded by a person/number agreement pre- Incorporation of noun stems is typically
fix, t-, and it conditions vowel harmony in from noun phrases functioning as a direct
the verb root, mne-/mna. (7a) and (7b) are object. The verb then generally takes intran-
essentially synonymous, except that walat in sitive agreement morphology (Chukchee
(7b) is marked for plurality, whereas an in- verbs normally agree with their objects) and
corporated root can never be marked for the subject is marked with absolutive, not er-
number (or any other inflectional category). gative case. In other words, in Chukchee,
Interestingly, in both cases the noun root re- though not in all languages (e.g. Mohawk; cf.
driver is a synthetic or verbal compound. (cf. notion of ‘wordhood’ arises when we compare
Art. 87). Taxi is interpreted as the direct ob- words with lexicalized phrases (cf. Art. 26).
ject of the verbal base of the noun driver. A This problem has been discussed by Di Sci-
variety of theories have been proposed to ullo & Williams (1987), who point out that
handle such facts (cf. Spencer 1991: ch. 8). French is replete with constructions of the
Most observers are agreed, however, that it is type le porte-parole ‘the spokesman’, le va-et-
a kind of compound, and hence, more of a vient ‘the coming and going’, constructed out
word than a phrase. What, however, of Lon- of phrases which include verb forms with fi-
don taxi driver (on either interpretation)? nite inflections. That these are words, indeed,
London is clearly a noun, and one option nouns, is shown by the fact that they take the
would be to regard the whole expression as definite article and other modifiers character-
another compound. However, others might istic of nouns, and have a meaning which in
argue that the modification is essentially syn- some degree is idiosyncratic. This suggests
tactic and that London, though a noun, is be- that languages may be permitted to take a
ing used ‘like an adjective’ (cf. the notion of phrase, or even a whole sentence, and reana-
‘noun premodification’ in Quirk & Green- lyze it as a word. This is the opposite of the
baum 1973: 399). first of the interactions mentioned above, for
The first of these problems, how to de- here it is syntax which creates the objects
scribe syntactic relations between parts of which are then stored in the lexicon as words.
compounds, becomes especially acute in lan- Cases such as this, in which syntactic ex-
guages in which phrases and compounds are pressions are reanalyzed as words, represent
even less easily distinguished than in English. a particularly interesting, if poorly studied,
The synthetic compound taxi driver is distinct type of interaction between morphology and
from a phrase such as driver of taxis or even syntax. A language which illustrates some of
drives taxis amongst other reasons, because the complexities of word formation possible
of its characteristic word order. In particular, by such processes, and which makes particu-
such a compound has much the same struc- larly frequent use of them is Navajo. I rely
ture as any other endocentric noun ⫹ noun for my data on Young & Morgan (1980).
compound in English, such as houseboat. In a Navajo nouns consist of a simple stem,
number of languages, notably those of South with possibly a prefix and possible what
East Asia, we find a plethora of compound- Young & Morgan call a ‘nominalizing en-
ing types most or all of which correspond in clitic’ glossed ‘the (particular) one that …’.
nearly all respects to syntactic phrases. An- However, it is common to find inflected verbs
derson (1985), basing himself on Chao used as nouns. Consider, for instance, the
(1968), provides a useful survey of such con- literal glosses given by Young & Morgan for
structions in Chinese. He points out that an the words shown in (19) (the apostrophe ’ af-
expression such as (18) can have two mean- ter a consonant represents an ejective and
ings, one the regular syntactic meaning, the elsewhere represents a glottal stop, an acute
second a specialized compound meaning: accent means high tone, ł is a voiceless lat-
eral; a doubled vowel is long):
(18) da shou
strike hand (19) (a) ha-do
(i) ‘strike(s) the hand’; (ii) ‘hired rioter, 3:impers-warm
thug’ ‘heat’ (‘it is warm’)
In many respects expressions such as the (b) ha-k’az
Chinese compounds behave like frozen pieces 3:impers-cold
of syntax which have become lexicalized, ‘coldness’ (‘it is cold’)
rather than like a special morphological con- Another common way of forming nouns is
struction such as a true compound. One of by suffixing a nominalizer to an inflected
the problems we encounter in defining the verb form, as in (20):
Compound nouns are formed in a variety of sition ⫹ noun, and noun ⫹ possessed noun
ways, including the concatenation of noun (reminiscent of the Turkish izafet). Some ex-
stems, noun ⫹ postposition, noun ⫹ postpo- amples are given in (21):
These expressions are words in the sense One way in which such nouns behave like
that they express a particular meaning (de- syntactic constructions is seen in certain cases
spite the relatively wide range of referents with the possessed form. Possession is sig-
that might be suggested by the literal glos- nalled by a person/number agreement prefix.
ses). In other words, they share with words Young & Morgan (1980: 18f.) report that
specificity of reference. However, their com- most nouns of the type (21, 22) take posses-
sive morphology just like any other noun.
ponent parts appear to have been concate-
For instance, from béésh bee ’élzhéhé ‘knife
nated in part by syntactic rules, including in- with:it something:is:peeled-nominalizer (par-
flectional processes. It is an interesting, and ing knife)’, we form shi-béésh bee ’élzhéhé ‘my
potentially important question, whether such paring knife’. However, in other cases the in-
constructions should be regarded as the re- ternal structure of the complex noun deter-
sult of word formation or word creation (cf. mines how possession is expressed. Thus, the
van Marle 1985). word for ‘scholarship grant’ is given as (23):
However, the words for my scholarship grant These refer to current grants. The words for
and your scholarship grant are respectively my future grant and the grant I have just used
(24a, b), with a different verbal stem, reflect- up would be (25a, b) respectively:
ing the person/number of the possessor/sub-
ject of the verb (the stem of the verb ‘go to (25) (a) béeso bik’é ’ı́ı́déeshtahı́gı́ı́
school’ is ta’): (b) béeso bik’é ’ı́ı́łtahı́gı́ı́
(24) (a) béeso bik’é ’ı́ı́nı́shta’ı́gı́ı́ These include the future and perfective mark-
(b) béeso bik’é ’ı́ı́nı́łta’ı́gı́ı́ ers respectively. If these expressions are
words, then they are inflected by rules which logically and/or morphologically this fre-
belong to syntax, not morphology. A theory quently gives rise to a structural mismatch
of the morphology-syntax interface should between morphophonological form and syn-
give an account of such phenomena. tactic form (often referred to as a bracketing
In English, too, we sometimes see what ap- paradox). A striking example is provided by
pears to be a syntactic construction (i.e. a the fusion of a preposition with a definite ar-
phrase) inside an expression we would nor- ticle in French (e.g. du ⫽ de ⫹ le ‘of the
mally want to call a word, namely, a com- (masc. sg.)’. This simple case constitutes a
pound noun: car-of-the-month competition massive violation of the one:one correspon-
(cf. Lieber 1988; 1992). This violates what dence between form and function entailed by
Botha (1983) dubs the ‘No Phrase Con- the Lexical Integrity Hypothesis, because a
straint’ on compound formation: a com- single morphophonological word, du, corre-
pound cannot contain a phrase (though the sponds to two morphosyntactic words which
term No Phrase Constraint itself would ap- together do not even form a syntactic constit-
pear to violate this constraint). This type of uent.
construction is more tolerable in English Finally, a type of deviation from lexical in-
when the phrase can be perceived as lexi- tegrity which has not been widely publicized
calized (in some rather unclear sense), though is illustrated in the copiously documented de-
in Dutch, apparently, there is no such restric- scription of Slavonic possessive adjectives
tion (G. Booij, personal communication). presented by Corbett (1987), the most inter-
Moreover, in some cases we seem to be incor- esting examples coming from Upper Sorbian
porating a fossilized piece of genuine syntax, (a West Slavonic language closely related to
complete with specific expressive intonation Polish). The possessive adjectives have a
meaning roughly paraphrasable as ‘pertain-
contour in some cases (his why-do-I-always-
ing or belonging to N’ where N is the base
have-to-get-the-blame look), while in other
noun. They are thus similar to the form presi-
cases it is only a disjointed fragment which is
dential in an expression such as the presiden-
incorporated, which itself would not be syn- tial limousine meaning the President’s limou-
tactically well-formed (no-go area). sine. In Upper Sorbian they are formed by
The question of the boundary between suffixing -ow to masculine nouns and -in/-yn
words and phrases is discussed in more detail to feminine nouns: from muž ‘husband’ we
in Art. 40. Another way in which the mor- form mužow and from sotra ‘sister’ we form
phology-syntax interface gives rise to overlap sotrin. What is peculiar about these adjectives
is found with clitic systems (cf. Art. 41). It is that in certain constructions, the stem noun
is very common for function words such as behaves like an independent syntactic ele-
auxiliary verbs, articles, or pronouns to lose ment. Specifically, this stem can control
some of their status as words and become agreement on a further possessive adjective.
phonologically dependent on a ‘host’ word. Thus, Corbett (1987: 303) records examples
When the clitic and host fuse together phono- such as (26⫺28):
In (26) we see that the possessive adjective ‘sister’, even though this word itself agrees
takes feminine agreements, controlled by the with its masculine nominative singular head
head ‘sister’. However, the word for ‘my’ noun ‘fiancé’. Finally, (28) shows that the
takes genitive masculine singular agreement first possessive (in this case ‘our’) will take
morphology. The masculine singular features genitive case irrespective of the case of the
are controlled by ‘husband’, while the geni- nominal phrase as a whole (in this case geni-
tive case seems to be determined by the pos- tive, governed by the preposition). Corbett
sessive construction itself. Likewise, in (27) discusses other respects in which the stem
the possessive ‘my’ is in the genitive case and nouns of possessives appear to enjoy inde-
takes feminine agreements in concord with pendent syntactic existence.
2. Syntax and internal word structure usual percolation mechanisms. On this view
a word such as indecipherability would have
If we regard word structure as essentially the structure indicated in (29):
morphotactics, i.e. the concatenation of mor-
phemes, and syntax as word concatenation (29) N
(i.e. disregarding questions such as agree-
ment, government, the binding of pronomi-
nals and anaphors, the control of ‘under- A
stood’ subjects and other such phenomena)
then morphology resembles syntax. This is A
reflected in the difficulty in distinguishing be-
tween words and phrases using purely distri-
butional criteria of the sort found in Bloom- V
field (1933). In principle, word formation can
be as productive as phrase formation, mor-
phemes may concatenate in various orders
within the word, depending on meaning or
function, morphemes within a word can be N
‘interrupted’, and so on. (Bauer 1988: 45⫺55
provides a convenient textbook introduction
to these issues; cf. also Art. 29, 40.) in de cipher able ity
Nonetheless, there are several respects in
which morphological operations are distinct In this structure de- is the head of the verb
from syntactic operations. Thus, there is no decipher, while -ity is the head of the noun as
syntactic equivalent to purely phonological a whole.
or morphophonemic operations such as con- In the theory of Lieber (1980) the constitu-
sonant mutation or ablaut (cf. Art. 53). Nor ent structure is entirely free of node labelling.
do we find individual formatives in syntax Affixes as well as roots have categorial fea-
exhibiting allomorphic variation depending tures, which percolate according to a set of
on arbitrary classes in the formatives to conventions. This theory utilizes another syn-
which they are attached, though this is com- tactic device, that of subcategorization. In or-
mon in morphology. Syntactic processes fre- der to express the fact that, say, -able suffixes
quently involve discontinuous dependencies exclusively to verb bases, the lexical entry for
between one part of the sentence and an- this suffix is given a subcategorization frame
other, as for example in the case of the ‘front- which states that it attaches to the right of a
ing’ of wh-phrases to form questions in Eng- verb. The obligatoriness of this subcategori-
lish. Such ‘movement’ processes (however zation captures the fact that -able is a bound
they are formulated syntactically) are foreign morpheme.
to word formation, with the specific excep- In these theories, it is not possible to
tion of metathesis. distinguish morphologically between inflec-
Despite the difference, much effort has tional and derivational formatives. This leads
been put into showing that affixation pro- to problems with the percolation of features,
duces words with something akin to syntactic since inflectional affixes, which are regarded
constituent structure (Williams 1981; Selkirk as heads on these approaches, do not always
1982). Since such structures are described supply all the featural content to the entire
using a phrase structure grammar, we may word. For this reason, Di Sciullo & Williams
call these ‘phrase structure approaches’ (cf. (1987), for example, introduce a notion of
Spencer 1991: 198⫺205). It seems reasonably ‘relative head’: features may percolate from
clear that endocentric compounding gives what is strictly speaking the non-head (com-
rise to a simple binary branching constituent plement?) of the affix provided the ‘real’ head
structure, though what that structure is may is not specified for such features.
be controversial, particularly in the case of Further details of phrase structure gram-
synthetic compounds (cf. Art. 83, 87). More mar approaches can be found in Art. 42.
controversially this is also claimed for deriva- There has been no little debate over crucial
tional processes. A category changing affix is aspects of this approach. Doubts have been
regarded as the head of the word which it raised about the appropriateness of using a
creates. The features of this head are inher- syntactic notion such as ‘head’ for word
ited by the word as a whole by virtue of the structure (e.g. Zwicky 1985; Bauer 1990).
Williams (1981) has proposed that in words (though this impression can be created in
the head is universally to be found on the the limiting case, when a single formative
right, while in syntax phrases can be left or uniquely realizes a single meaning or func-
right headed. Moreover, the kind of constitu- tion). Nor is it necessary to push morpho-
ent structure found in word formation is ex- phonological processes such as mutation or
tremely simple and it is very difficult to see conversion into an agglutinative straitjacket
how the relatively rich structures found in, by assuming affixes whose effects are purely
say, the X-bar theory of syntax are reflected phonological, or a plethora of zero affixes.
in word structure. There is no obvious sense A strong version of this thesis is proposed
in which morphological formatives form by Beard (e.g. Beard 1987). It is widely rec-
modifier-modified structures, or governor- ognized that inflectional paradigms pose seri-
governee combinations (outside compound- ous problems for a strictly agglutinating view
ing, at least), and no evidence whatever for of morphology of the kind implied in word
the assumption current in Government-Bind- syntax approaches. Well-known problems of
ing syntax that there is a Specifier position. cumulation, fusion, syncretism and so on (cf.
Constituents of words do not generally be- Art. 38, 65, 66) have led some (e.g. Matthews
have like sentence constituents, in that the 1972; Anderson 1982) to propose a para-
usual tests of constituenthood for phrases are digm-based theory in which the inflectional
inapplicable for words. Thus, we do not find properties expressed by a set of morphologi-
word-internal constituents being replaced by cal operations on a word be regarded sepa-
single elements (e.g. pro-forms) and such rately from the formatives themselves. Beard
constituents do not undergo movement pro- observes that this many-many correspon-
cesses. Even where affix movement processes dence between form and function is not un-
are postulated (as in Pesetsky’s 1985 pro- commonly found in derivation, too. (In this
posal) it is only a single affix, not a constitu- respect, affixes behave more like functional
ent comprising strings of affixes, that is elements, such as prepositions, than as lexical
moved. Indeed, in general it is necessary to elements.) His proposal is the Separation Hy-
appeal to special principles to prevent phono- pothesis, which treats the more abstract no-
logical or morphological processes from tion of lexical derivation (e.g. ‘nominalization
gaining access to internal constituents (e.g. of a verb’) as a separate phenomenon from the
the Atom Condition (Williams 1981) or the morphological operation associated with it
Adjacency Condition (Siegel 1977); cf. Scalise
(say, adding -ment to the base or root ablaut).
1984: ch. 8; and Art. 42). Moreover, morpho-
In order to ensure that the mapping between
logical operations such as consonant muta-
the two is not entirely unconstrained, only the
tion or conversion, which have no counter-
features last added to a base are accessible to
part in syntax, require special treatment, e.g.
by recasting into an agglutinating form or by further word formation or to syntactic pro-
setting up special headless constructions. cesses. In this way, indecipherability behaves
The phrase structure approaches inherit much like the word ability, while decipherable
from structuralist progenitors an emphasis behaves much like readable.
on the view of the word as a linear string of One consequence of the separation of form
morphemes, and rely heavily on the view of and function is that morphologically com-
the morpheme as a formative which has plex words are at most linear strings of for-
meaning as well as form. Given this it is easy matives, in other words they cannot have a
to equate morpheme-in-a-word with word- constituent structure. This immediately ex-
in-a-phrase and attempt to use the same for- plains why certain so-called ‘bracketing para-
mal means of modelling them, particularly in doxes’ are possible (cf. Art. 81; see also Beard
those theories, like Lieber’s (1980) in which 1991). A much-discussed example of this is a
affixes are listed along with roots as lexical word such as ungrammaticality. The problem
entries in the lexicon. is that un- in English is a word level (‘native’,
In a different set of approaches a distinc- ‘Class II’) affix, while -ity is a root level
tion is drawn between morphological opera- (‘nonnative’, ‘Class I’) suffix. In general,
tions and morphological processes. Morpho- root-level affixation must precede word-level
logical formatives (affixes), morphophono- affixation. Now, if -ity is affixed first, we cre-
logical processes such as mutation or ablaut, ate a noun grammaticality. Hence, un- will
or lack of morphological operation as in con- have to be prefixed to that noun. However,
version, are all simply exponents of a given un- may only attach to adjectives such as
process. There is no sense in which an affix grammatical, and not to nouns. Therefore,
has a ‘meaning’ or a lexical entry of its own the morphological requirements on affixation
predict a constituent structure [un[gram- approach is that “the syntax neither manipu-
matical]ity], while morphosyntactic con- lates nor has access to the internal form of
straints demand an incompatible structure: words” (Anderson 1988: 165). Under this
[[un[grammatical]ity]. However, if the mor- model, derivational morphology has to be
phological structural requirements at the distinguished from inflectional morphology
level of morphological formatives is simply in a principled way, with each handled by dif-
that a root affix such as -ity may only attach ferent types of rule. This is known as the
to a latinate formative such as the -al of split-morphology hypothesis (cf. Anderson
grammatical then there is no paradox. A very 1982; Perlmutter 1988).
similar explanation has been proposed by It has also been claimed that all morpho-
Sproat (1985; 1988), discussed in some detail logical processes defining the form of words,
in Spencer (1991: 409⫺413, 417⫺420). (For whether traditionally derivational or tradi-
another approach see Stump 1991.) tionally inflectional, should be handled in a
unitary fashion by means of lexical opera-
3. Theoretical approaches tions. This is strong lexicalism. On this model
it is possible to understand why the morpho-
There are several different ways of interpret- logical operations subserving derivation (i.e.
ing the interaction between syntax and mor- affixation, ablaut, and so on) are in general
phology. A particularly important distinction indistinguishable from those subserving in-
is that between lexical approaches and what flection (cf. Lieber 1980; Marantz 1984). How-
we may call ‘syntactic’ approaches to mor- ever, it is then necessary to provide a mecha-
phology. nism for syntactic processes such as agree-
On a strictly syntactic approach to mor- ment to have access to just those aspects of
phology, all word formation and inflection is morphological structure which are relevant
the consequence of syntactic rules. Roughly to syntax. This is generally handled by ensur-
this position seems to have been taken by ing that properties of, say, agreement forma-
proponents of Generative Semantics (cf. tives become properties of the whole word,
Newmeyer 1980). Thus, the nominalization for example, by feature percolation (for ex-
destruction would be formed from the verb plicit discussion of this see Lapointe 1981;
destroy by means of a syntactic rule, in the 1988). When, say, a predicate and its subject
same component of grammar as that respon- agree in person/number features, this is
sible for, say, fronting a wh-word in a special achieved by a checking process in which the
question such as What did Tom eat? Likewise, features distributed over the syntactic repre-
the present tense agreement marker -s in Tom sentation must match the feature composi-
sings would be attached by a syntactic rule. tion of the fully inflected words inserted into
Chomsky (1970; cf. also Newmeyer 1980; the syntactic representation from the lexicon.
Scalise 1984; Spencer 1991) argued that no- On this view the lexicon in principle contains
minalizations such as destruction could not all the possible word-forms of a given lexeme,
be the result of regular transformational syn- though in many cases these have more the
tactic process, essentially because such for- status of ‘virtual’ word-forms, generated by
mations show too many idiosyncrasies to be productive inflectional processes.
so analyzed. In this respect they contrast with I shall illustrate some of these ideas with
genuinely regular, productive and transpar- some specific case studies. Consider noun in-
ent -ing nominalizations (destroying). Deriva- corporation, in which, for instance, the noun
tional morphology of this type came to be of a verb’s direct object forms a compound
handled by means of special morphological with the verb stem. Baker (1988 a) in an in-
rules applying specifically to words, and not fluential theory of grammatical function
to syntactic units. However, under this model changing argues that this is just one instance
inflection (including -ing nominalization) of a general syntactic process of head-to-head
could still be handled by syntactic rules. movement. In Government-Binding theory it
The place at which word-level derivational is assumed that a head may be raised and ad-
process apply has been identified with a spe- joined to another head under certain circum-
cific morphological level (‘component’, ‘mod- stances, leaving an empty category (‘trace’)
ule’) or the lexicon itself. Hence, we may refer with which it is coindexed. In Baker’s theory
to such a model as a Lexicalist model. How- the constraints on head-to-head movement
ever, since not all word forms are created by are a consequence of a syntactic restriction,
lexical processes this variant is known as the Empty Category Principle or ECP (see
weak lexicalism. The principal claim of this Haegeman 1991, for a textbook treatment of
these issues). For the purposes of incorpora- 1989). These accounts stress that in some of
tion, this states that the trace of the moved the languages discussed, particularly Mo-
head must be governed by the host head, and hawk, it is indeed possible to double the in-
the phrase from which the moved head origi- corporated noun. In addition, it has been
nates must be a complement of the host. claimed that when incorporation strands
Thus, it is possible for a verb to incorporate modifiers, the resulting headless phrase is one
the head of a direct object, but not the head which is permitted by the language even
of a subject, or an adjunct. This corresponds without incorporation. (Further details of
to other syntactic asymmetries between sub- these issues are found in Art. 88.)
jects and adjuncts on the one hand, and sub- Baker’s head-to-head movement theory of
categorized phrases on the other. incorporation is extended to an impressive
Because incorporation is movement leav- array of phenomena. He argues that all the
ing a trace, this theory predicts two things. main grammatical function changing pro-
First, it is impossible to ‘double’ the incorpo- cesses are instances of head-to-head move-
rated noun giving sentences corresponding ment (see Spencer 1991: ch. 7.4 for overview).
to, say, Tom horse-rode the stallion. Second, For instance, he argues that regular causativi-
it is possible to strand modifiers of the incor- zation, even when marked by a morphologi-
porated noun. So in some languages at least cal process, is the result of incorporation of a
it should be possible to say Tom horse rode verb by a governing causative verb (Marantz
the black meaning Tom rode the black horse. 1984 adopts a similar position). This can be
A number of lexicalist accounts have been illustrated in very simple terms (ignoring con-
proposed for such constructions (e.g. Mithun siderable technical detail) by the Chichewa
1984; Di Sciullo & Williams, 1987; Rosen example in (30) (Baker 1988a: 221f.):
Baker assumes an initial syntactic struc- joined to the matrix verb, the causative mor-
ture (D-structure) which is essentially like the pheme. In order to respect the Empty Cate-
English gloss to (30), namely (31) (simplify- gory Principle this has to involve movement
ing Baker’s account somewhat): of the whole embedded VP to a position un-
der the CP (Specifier position of C). From
(31) S there the verb stab is incorporated leaving be-
hind the NP elephant as the derived direct ob-
NP VP ject of the causative verb, and leaving the em-
bedded subject the hunters to be case marked
by a special rule introducing the preposition
hare V CP kwa. The result is shown in (32):
(32) S
made C'
NP VP
C S
hare V CP
NP VP
0
V V VP C'
hunters V NP
V NP S
stab elephant
stab made
Baker assumes that the causative mor- t elephant NP VP
pheme is represented syntactically as a verb
selecting a clausal complement (CP). The
complement clause verb ultimately gets ad- hunters t
Similar accounts are offered for applied any language. This is accounted for on a the-
verb formation (cf. Art. 107) which is re- ory such as Baker’s by general principles, in-
garded as a form of preposition incorpora- dependent of this kind of construction, which
tion. Thus, schematically, in a Bantu lan- prohibit movement across certain phrasal
guage such as Chichewa a structure underly- categories. In particular, it is impossible to
ing The boys hand an apple to Harriet would extract anything out of a relative clause. For
alternate with an expression of the shape The example, it is impossible to ask a question of
boys hand-to Harriet an apple, in which hand- an element of a relative clause. Hence, it is
to represents a specially suffixed form of the impossible to relate (35) to any non-emphatic
verb. More controversially Baker assumes version such as (36):
that passive morphology is the result of in-
corporation. The passive morpheme is re- (35) You met a man who has built WHAT?!
garded as a kind of nominal auxiliary to
(36) *What did you meet a man who has built?
which is assigned the subject argument role
of a transitive verb. This has the effect of re- Baker’s theory is able to exclude nonsense
ducing the valency of the verb by one argu- such as (34) by appeal to precisely the kinds
ment. The passive morpheme, which is in the of syntactic principles which exclude (36).
I(nfl) position, then incorporates the verb Moreover, in principle this pattern of expla-
stem. A more articulated version of this nation is open to any theory of valency oper-
theory is presented in Baker & Johnson & ations which can make appeal to syntactic
Roberts (1989). principles or syntactic structure in a compar-
There are several advantages to this type able fashion.
of syntactic approach. That advocated by Causative constructions illustrate an im-
Baker, in particular, proposes a number of portant sense in which morphological struc-
elegant ways of accounting for many of the ture seems to reflect syntactic structure. If we
syntactic properties of morphologically com- oversimplify somewhat we can say that there
plex words. The complementarity of distribu- are essentially two sorts of causative con-
tion between overt direct objects and incor- struction. The first the type is illustrated
porated direct object nouns is one such case
schematically by the English gloss to the
(essentially the same argument as that ad-
Chichewa example (30), represented schemat-
vanced by Kayne 1975 for assuming that pro-
ically in (31). This has rather different syntax
nominal clitics in Romance are attached to
from that of the Chichewa example to which
the verb by a syntactic movement rule). In
it corresponds. First, the embedded subject
addition, Baker explains the fact that noun
incorporation is restricted in essentially the behaves rather like a derived object of the
way that other types of movement are re- matrix verb make. (Such a verb is often re-
stricted. The trace left by movement of the ferred to as a ‘Raising-to-Object’ verb, be-
head must be in a strict relationship of cause the subject appears to have risen out
proper government with its antecedent, the of the embedded clause into the complement
moved head. For this reason, it is impossible position of the matrix verb. Alternatively, it is
to form a causative which ‘skips’ out of an- referred to as an ‘Exceptional Case Marking’
other, lower, clause. For example, it would be verb in Government Binding theory, because
impossible in Chichewa, or any other lan- the matrix verb is able, exceptionally, to as-
guage, to construct a verb form meaning sign objective case to the lower subject, as
‘cause to stab’ which could be used to trans- can be seen when the NP is a pronoun: The
form a sentence with the structure of (33) hare made them stab the elephant). At the
into a hypothetical sentence-type with the same time the embedded object, of course, re-
structure of (34): mains an object.
This is very different from the behaviour
(33) The hare made [CP[NPthe hunters of the Chichewa nominals, in which the em-
[CPwhich stabbed the elephant]] leave] bedded subject loses its argument status, and
(34) The hare made-stab [[the hunters which is treated more as a kind of adjunct phrase.
the elephant] leave] Here, the causativized verb behaves very
much like a derivationally simple verb in the
In (34) the verb has been incorporated from sense that it may only take one genuine ob-
a relative clause modifying the subject of the ject, namely the NP corresponding to the ob-
complement clause. The result is gibberish in ject of the embedded verb. It is this NP, for
instance, which is cross-referenced by the In other words, (38) would be glossed ‘The
verb in constructions in which the verb agrees hare made the hunters stab him (⫽ the hare)’.
with its object, and it is this NP which un- One way of summarizing these differences
dergoes passivization. Another difference is is to distinguish broadly two types of caus-
apparent when we replace the embedded ob- ative construction which we may call mo-
ject with a reflexive pronoun. The general noclausal and biclausal (cf. Aissen 1979). The
pattern with Chichewa and other such lan- biclausal type is illustrated by the English
guages is that the reflexive must refer back to ‘make’ examples. Here the embedded com-
the subject of its clause. In English, if we re- plement is syntactically autonomous in that
place elephant in (31) with a reflexive we get clause-level principles apply to it. However,
(37): in the Chichewa example the causative verb
seems to be the predicate of just a single
(37) The hare made the hunters stab self. clause. The point of crucial interest is that
Here, self must refer to the hunters (and there are languages with morphological caus-
hence assumes the form themselves). How- atives which behave like the English biclausal
ever, in the equivalent in Chichewa we would type. The Bantu language Chi-Mwi:ni is a
find that the reflexive morpheme in (38) case in point. (Indeed, there is a dialect of
would refer back to the hare, the subject of Chichewa itself which behaves like this.) In
the matrix clause: Chi-Mwi:ni a reflexive in the position of the
(surface) direct object of the causative verb
(38) The hare stab-made the hunters self. can refer to the subject:
However, if the reflexive is in the position of which, of course, is now the direct object of the
the embedded object then it can only refer causative verb (40); it cannot refer back to the
back to the (underlying) embedded subject, matrix subject (41) (cf. Baker 1988 a: 211):
(43) S (44) S
NP VP NP VP
I V CP I V CP
C S made hit C S
hit NP VP t NP VP
child V NP child V NP
t self t self
This analysis explains another feature of Constructions such as these pose a chal-
these constructions. The NP which behaves lenge to lexicalist theories, for such theories
like the true object of the derived causative have to find some non-syntactic way of ac-
verb is the embedded subject NP, not the counting for the apparently syntactic differ-
original object. Contrast the situation with ences between the two types of causative. Di
the monoclausal causatives. This again can Sciullo & Williams (1987) and Anderson
be related to the structural difference. The de- (1989) have addressed this problem. Their
rived verb can be said to govern the embed-
suggestion is a further radical lexicalization
ded subject in essentially the same way that
make governs its embedded subject. The re- of syntax. (See also Alsina 1996 for discus-
maining lower object then has to be specially sion within Lexical-Functional Grammar.)
marked (as a ‘second object’) by devices Anderson (1989: 3) notes the Georgian is a
which are employed in other constructions in language with morphological causatives which
the language. (But see Baker 1996 for a re- may pattern in the biclausal manner. Con-
vised account.) sider example (45):
In (45) it is possible for himself to refer to having the effect of binding the object to
Vano. Normally, however, it is impossible for the subject position. This would mean that
a reflexive of this kind to refer to a non-sub- only a reflexive pronoun could cooccur with
ject. Thus, it appears that at the level at such a verb. In many languages this feature
which anaphor binding is defined, Vano is a would take the form of affixation, and the
subject in (45). affix would then fulfil the role of the reflex-
Anderson proposes to handle this by as- ive direct object, in effect turning the verb
suming that there is a morphosyntactic fea- into an intransitive (see Spencer 1991: 248⫺
ture [reflexive] which accounts both for the 251 for discussion of such cases). Anderson
binding properties of independent words assumes that in Georgian there is no overt
(such as anaphoric elements) and for the morphological reflex of reflexivization. How-
behaviour of certain verb types. Citing Di ever, he does provide evidence from else-
Sciullo & Williams (1987), he suggests that where in Georgian grammar of the need for
transitive verbs may be marked so as to a clearly lexical rule to appeal to a [reflex-
require this feature on their direct objects, ive] feature.
Since the reflexivization process is lexical vization. In this case, the verb talk will un-
it can feed other lexical rules, such as causati- dergo the processes shown in (46):
This form will correctly account for the rele- (though this would have implications for case
vant reading of (45). marking and other aspects of syntactic struc-
Of some interest is the fact that (45) is actu- ture which are not entirely obvious). For An-
ally ambiguous. It can also mean that Vano derson, the ambiguity is what is to be expected
was persuaded to talk about the doctor. On if reflexivization and causativization are lexi-
this interpretation we have a monoclausal cal processes (though it is unclear on this ac-
causative. Baker does not discuss cases of this count why such ambiguity is not reported for
sort explicitly, and probably he would analyse other languages with morphological caus-
this as optional alternation between a mo- atives). For beside (46) we can just as easily
noclausal and a biclausal type structure construct a verb form along the lines of (47):
Clearly, such a form will underly the second Projection Principle, which says that the va-
reading of (45). lency of a predicate cannot change during a
A phenomenon related to the question of derivation, and the Theta Criterion, which
noun incorporation is the satisfaction of ar- says that all arguments must be satisfied.
gument structure in synthetic compounds. A This presupposes that the non-head of a com-
synthetic compound consists of a deverbal pound counts as an argument position. No-
noun head modified by a nominal non-head, tice that (48) counts as a type of ‘bracketing
as in truck driver. Much discussion has been paradox’ (see section 2), in that the morpho-
devoted to the observation that in such cases logically determined constituent structure of
truck seems to function as the direct object of the compound must presumably be [[truck]
the verb drive. In this respect the compound [drive ⫹ er]]. This is a further example of the
resembles a syntactic phrase such as drives a separation of morphological and syntactic
truck or more tellingly, (a) driver of trucks. well-formedness principles.
(See Spencer 1991: 324⫺340 for a review). An alternative approach, which we may
One approach is to claim that syntactic call ‘lexical’ is to regard synthetic compounds
principles governing the satisfaction of argu- as noun ⫹ noun compounds, but permit the
ment structure operate at the level at which argument structure of the verbal base to be
synthetic compounds are formed. Sproat ‘inherited’ by the deverbal nominal. An ac-
(1985), for instance, argues that truck driver count of this sort is proposed by Di Sciullo &
is constructed in the syntax by first com- Williams (1987), illustrated in somewhat sim-
pounding truck with the verb and then affix- plified form in (49):
ing -er, to give (48):
(49) N
(48) N
Nj N*Agi, Thj+
V er
V*Agi, Th+
N V
truck drive eri
truck drive
The symbol *Ag, Th+ represents the argu-
On the intended reading, in which drive is ment structure of drive, taking an Agent (sub-
used transitively, it must have a direct object. ject) and a Theme (object) argument. The
In the Government and Binding theory as- coindexation indicates that the Agent role is
sumed by Sproat this is encapsulated in the linked to -er, expressing the fact that this is
an agentive affix. The argument structure of consider the outwardly similar possessive ’s,
the verb base percolates to become a prop- as in the man responsible’s punishment. This
erty of the deverbal noun, and its Theme role will be final in its domain (of noun phrase),
is linked to the non-head truck. On this after and enclitic. The crucial point is that
analysis there is no ‘bracketing paradox’. the phonological parameter can induce a di-
However, this type of account needs to ex- rection of morphophonological attachment
plain what aspects of argument structure can which is entirely at odds with morphosyntac-
be inherited and which cannot. A variety of tic requirements.
refinements to the basic idea of inheritance A typical clitic has some of the properties
have been proposed, particularly where the of a word, in that it cooccurs with a great
interpretation of affixes such as -er is con- variety of different word and phrase types,
cerned. Booij (1988) provides an example and and some of the properties of an affix, in that
a good summary of the issues (cf. Art. 83). it is obligatorily bound to a host. Indeed, clit-
Related to the question of synthetic com- ics are often referred to as phrasal affixes for
pounding is the vexed issue of argument this reason (though see Art. 41 for a discus-
structure satisfaction in nominalizations. For sion of the possibility of distinguishing
instance, in a hallowed expression such as the phrasal affixes from bound words). In addi-
city’s destruction the ‘possessive’ expression tion, groups of clitics often form clusters
the city’s acts as the object of the verb stem which respect strict ordering constraints. This
destroy, just as in the expression the destruc- is particularly obvious in a language such as
tion of the city. On the other hand, in the bar- Serbo-Croat, where phrasal word order is
barian’s destruction of the city the ‘possessive’ rather free, but where the order of elements
acts like the subject of the verb stem. This in the clitic cluster is rigidly fixed. However,
depends on a variety of factors, including lex- many of the formatives which are often re-
icalization and on the extent to which aspec- ferred to as clitics have more of the character
tual districtions are preserved from the verbal of genuine affixes than clitics. Thus, the pro-
base (see Grimshaw 1990). nominal clitics of the Romance languages
Cliticization represents a particularly com- tend to cluster around the verb (not the verb
plex example of the interfacing of distinct lin- phrase) and in some varieties (most famously,
guistic components, in that it involves the Argentinian Spanish) the object clitic can be
mapping between morphology (and the lexi- ‘doubled’ by a full and overt noun phrase (cf.
con) on the one hand and both phonology Suñer 1988). In Macedonian, this type of
and syntax on the other. Particularly striking doubling is obligatory (see Spencer 1991: 358⫺
are the mismatches induced by cliticization 362), and Lyons (1991) has argued that the
between morphological structure and syntac- best way to accommodate these facts is to as-
tic structure, of the kind illustrated in section sume that all such clitics, whether they permit
1 (26⫺28). Klavans (1985) has argued that doubling or not, are actually (degenerate)
there are theoretically exactly eight types of agreement markers.
clitic possible (cf. also Sadock 1990: 72⫺77; Agreement itself represents one of the
Spencer 1991: 375⫺381). These are defined in more common aspects of the morphology-
terms of two syntactic parameters and one syntax interface. However, it appears in a
phonological parameter. All clitics are posi- great variety of guises and there is no consen-
tioned peripherally to a syntactic domain sus on how best to describe it. In its simplest
(e.g. sentence, verb phrase, noun phrase). form agreement is a morphophonologically
Given a domain a clitic may be either initial realized cross-referencing of a head and its
or final, and it may be placed either before complement or a head and its modifier. A
or after the peripheral constituent. Phonolog- checklist of cases is given in Lapointe (1988).
ically, it may be either enclitic or proclitic. The problem of distinguishing agreement
For instance, consider the clitic form of the from cliticization assumes intriguing dimen-
auxiliary verb ‘have’ in a sentence such as sions when we examine cases such as Celtic
The man responsible’s been punished. Assum- (cf. Doron 1988, and especially Sadler 1988:
ing the clitic ’s is attached syntactically to the 49⫺66). In Irish (and Breton) verbs may
verb phrase been punished, then it is initial, either bear inflections cross-referencing the
before (the verb) but enclitic. Hence, it atta- subject (the ‘synthetic’ form of the verb) or
ches phonologically to the previous word bear a default, non-agreeing inflection (the
(which will in practice mean the final word ‘analytic’ form). There is an absolute ban on
of the subject noun phrase, responsible). Now the appearance of subject-verb agreement
morphology (i.e. the ‘synthetic’ form) when form is used. In other words, there seems to
the subject is phonologically overt. Thus we be a principle of ‘complementarity’ at work:
may have (50a) but not (50b) (McCloskey & when there is an overt subject the analytic
Hale 1984): form of the verb must be used (i.e. the one
(50) (a) Chuirfinn isteach ar an phost sin. which lacks subject marking), and when the
put:cond:1.sg in on that job synthetic form is used there can be no overt
‘I would apply for that job.’ subject.
Welsh resembles Irish in that lexical NPs
(b) *Chuirfinn mé isteach ar an phost sin. are not allowed to cooccur with verbs in the
put:cond:1.sg I in on that job synthetic form. When there is an overt lexical
Not all person/number combinations have subject, the verb appears in the analytic form
inflected forms. However, where they exist and assumes default (3rd person singular)
they are obligatory. When the subject is a lex- morphology, as shown in (51) (from Sadler
ical NP then a default 3rd person singular 1988: 50f.):
are thought of as pronominal clitics are syn- are modules of grammar (cf. Chomsky 1986;
tactically agreement markers. However, on see Grimshaw 1986 for the terminological
Anderson’s analysis of Breton, verbal mor- distinction between ‘module’ and ‘compo-
phology which would normally be regarded nent’). As such it may have rather complex
as agreement inflection is treated as a species relationships to the components of grammar,
of cliticization. (One might add to this type and to other modules.
of argumentation the claim in Baker & John- Baker (1985: 375) has argued that lan-
son & Roberts 1989 that the passive partici- guages adhere to the Mirror Principle:
ple morphology of languages such as English (55) Morphological derivations must directly
is a type of pronominal clitic; see Spencer reflect syntactic derivations (and vice
1991: 384⫺390 for further data and discus- versa).
sion).
What this means is that if an affix A appears
closer to the base than affix B, then the syn-
4. Morphology as a module or a tactic operation associated with A occurs
earlier in the derivation than that associated
component with B. Other things being equal, this in prac-
tice only has any obvious interpretation
There is a further important aspect of mor-
where A and B are both prefixes or both suf-
phology-syntax interactions concerning the fixes. Baker takes the Mirror Principle to be
way that morphological phenomena are true, and much of Baker (1988 a) is given
viewed within the overall organization of over to providing syntactic analyses of gram-
grammar. Two principal alternatives can be matical function changing operations, such
discerned: either we can view the morpholog- as the causative constructions seen in the pre-
ical system as a separate component of gram- vious section, in which morpheme ordering
mar, ordered with respect to other compo- can be derived from the order in which syn-
nents such as phonology or syntax, or we can tactic processes apply. The argument is that
view morphology as a set of well-formedness such analyses would not be possible unless
principles which operate over structures cre- word formation and syntax were, in a certain
ated in grammatical components, such as sense, part of the same ‘component’.
syntax or phonology. On this second view, A crucial illustration of the Mirror Prin-
morphology is not a component in the sense ciple involves agreement in Chamorro. A
of a stage which a derivation ‘passes through’, prefix fan- appears if the clause contains an
rather it is a module of grammar, in much intransitive verb and the subject is plural.
the way that Case Theory or Binding Theory Baker draws attention to examples (56, 57):
These purport to show that fan- is found (58) (a) [[[verb] CAUS] PASS)
with derived (surface) subjects (example (56)) (b) [[[verb] PASS] CAUS]
but that fan- may also agree with a plural These are always accompanied by character-
embedded subject (the causee in (57)). The istic meaning differences: the first means ‘be
force of the example is lessened somewhat by caused to verb’ while the second means ‘cause
Durie’s (1986) painstaking analysis of this to be verbed’. In Baker’s theory this is auto-
and similar cases, in which he shows that matically accounted for, since both the caus-
number agreement is not, in fact, involved. ative and passive morphemes are heads
Another illustration, encountered in large which can incorporate their complements. In
numbers of languages can be cited schemati- (58a), the causative head incorporates the
cally. In a language with both passive and verb stem, then this complex verb base is in-
causative constructions we often find that it corporated by the passive morpheme. In
is possible to derive verbs of the form (58a) (58b) the opposite order is followed. Baker
or (58b): includes evidence from Bantu languages to
support this thesis. However, Hyman & niceties of morphological structure seriously
Mchombo (1992) offer examples in which undermines the program to reduce inflection
some morphemes are doubled (violating the to syntax.
strict interpretation of the principle) as well There are other problems with morpheme
as cases in which morphemes simply occur in ordering and the Full Functional Projection
the ‘wrong’ order. Hypothesis. First, the principle would pre-
Baker (1988 a: 487f., fn. 39) points out that sumably rule out a situation in which subjects
his theory can derive the Mirror Principle were syntactically external to the verb phrase,
without stipulation but only in the case of and in which a verb agreed with both its sub-
valency changing. It does not follow auto- ject and its object and in which the subject
matically in the case of agreement (or, in- marker appeared closer to the verb stem than
deed, other types of inflectional morphology, the object marker. In other words, structures
such as case marking, tense/aspect/mood such as the following should be excluded:
marking and so on). However, in the theory [OBJ [SUBJ [verb]]] or [[[verb] SUBJ] OBJ],
of verbal inflection and agreement proposed with the (possible) exception of cases in
by Pollock (1989) each (verbal) inflectional which the subject appears closer to the verb
formative or functional category is regarded in the syntax (e.g. in which the word order
as a syntactic head, each licensing its own is OSV or VSO). It is fairly rare for subject
phrasal projection. Inflected words are con- markers to be internal to object markers,
structed by a process of head-to-head move- though this is found for instance in Adigei, a
ment. Thus, in a language in which a finite language of the North Caucasus. Embarrass-
verb receives, say, a tense suffix followed by ingly for the Mirror Principle, the affix order
an agreement suffix the inflected verb form is [OBJ [… [SUBJ [… [verb]]]]] is found in Nav-
generated by first incorporating the verb root ajo, and this language maintains a rather
into the tense head, and then incorporating strict SOV word order.
this composite into the agreement head. For Another problem is discussed in detail in
instance, we might generate the Italian verb Spencer (1993) and concerns nominal inflec-
form parlavo ‘I was speaking’ from [-o [-av- tion. Presumably possessive agreement mark-
[parl-]]] via [-o [[parl-av] [t]]] and [[[parl-av]-o] t ers are attached within the nominal syntagma
[t]]], where t stands for the trace of the moved itself, since they express a dependency be-
constituent. Under these assumptions, inflec- tween phrases contained entirely within that
tion must then obey the Mirror Principle (see syntagma. Case marking of the syntagma as
Ohalla 1990 for explicit discussion of Berber a whole, however, must presumably follow
and Turkish verb inflection from this point of such phrase-internal inflection. This implies
view). In its strongest form, this thesis would that where possession and case marking ap-
claim that any inflectional formative must be pear as inflections on the head noun, the
a syntactic head (this is referred to as the morpheme order must be (for suffixation)
‘Full Functional Projection Hypothesis’ in [[[noun] POSS] CASE]. This is found in lan-
Spencer 1993). guages such as Hungarian and Turkish.
However, in inflectional morphology, re- However, in Finnish, together with the Sa-
plete as it is with cumulations, fusions, and moyed languages, Saami (Lappish), and
total suppletions, it is clear that the Mirror Mordvinian exactly the opposite order is
Principle should be weakened, to read ‘mor- found. In some Uralic languages (e.g. Mari,
phological derivations may not contradict Udmurt, Komi) the morpheme order de-
syntactic derivations’. Even this formulation pends on which case is used (cf. Hajdu & Do-
should be given a rider to accommodate the mokos 1987: 240).
not uncommon phenomenon of morpheme These facts are very difficult to reconcile
metathesis. This occurs particularly com- with obvious interpretations of the head-to-
monly in languages with template morphol- head movement analysis of inflectional mor-
ogy (cf. Simpson & Withgott 1986; Speas phology. Two other problems are cited in
1991), which frequently involves discontinu- Spencer (1993). First, there is the problem of
ous dependences between inflectional (or der- doubled formatives. A more subtle problem
ivational) formatives. It is very difficult to see is posed by the Turkish indefinite izafet dis-
how such phenomena can be reconciled in a cussed in section 1. This is marked by the
principled way with Full Functional Pro- same inflectional formative that is found in
jection Hypothesis. Joseph & Smyrnioto- the definite izafet (a phrasal construction).
poulos (1993) argue that inattention to the This should therefore project a syntactic
phrase. However, the indefinite izafet is lexi- called ‘constructs’ in Hebrew (cf. Art. 103,
cal and cannot contain phrases. Given other 125; cf. also Spencer 1991: 447⫺453) and the
doubts surrounding the manœuvre of regard- differential behavior of deadjectival caus-
ing agreement as a syntactic head this may atives and inchoatives (Borer 1991). Borer’s
count as evidence against this radically syn- proposals are reminiscent of Kiparsky’s
tactic approach. (1985) claim that a single phonological rule
Baker’s interpretation of the Mirror Prin- may apply lexically or postlexically with sys-
ciple effects are challenged by Grimshaw tematically different effects.
(1986) who points out that it is not necessary The idea that morphology might be a sep-
to assume that the affixes are syntactic heads, arate module interacting in complex ways
merely that they are morphological heads. with syntax but obeying its own domain spe-
Then, their crucial semantic properties will be cific principles is one which has attracted a
nested in precisely the same way that is pre- number of theorists (see Spencer 1991: ch. 11
dicted by the head-to-head movement analy- for review; note that it is explicitly endorsed
sis. Thus, Baker’s approach shows only that by Baker 1988 b). One of the more carefully
the syntactically based approach is compati- worked out approaches which takes this line
ble with the Mirror Principle, not that it is is that of Sadock (1990) (itself a reworking
entailed by it. Finally, within the ‘minimalist’ and refinement of Sadock 1985). Sadock ob-
program of syntax (Chomsky 1993) inflected serves that when we assume that morphology
words are constructed in the lexicon and is an autonomous module with distinct sets
checked against features resident on syntactic of organizational principles, we heighten the
head. On this view there is no necessary rela- need to provide explicit constraints on the in-
tionship between order of inflectional forma- teraction between morphological structure
tives and order of syntactic processes (or syn- and syntactic structure. In Baker’s theory of
tactic constituent structure). valency alternations the Mirror Principle to-
Grimshaw raises another point of impor- gether with the Head Movement Constraint
tance, concerning the notion of a ‘compo- (ultimately the Empty Category Principle) re-
nent’. She provides the following definitions strict the types of interaction. Sadock adopts
(Grimshaw 1986: 748): a model of syntax which does not include the
Empty Category Principle, namely General-
(59) “(i) A set of rules or representations de-
ized Phrase Structure Grammar (Gazdar et
fined over a certain vocabulary and gov-
al. 1985). He therefore proposes two basic
erned by principles of a particular type.
constraints, the Incorporation Principle, and
A component in this sense is a domain
the Cliticization Principle, which he states in
of the theory.
preliminary form as follows:
(ii) A set of rules/operations that act as
a block in the organization of the gram- “Incorporation Principle (IP): A lexeme that com-
mar.” bines with a phrase at some level may combine
with the head of that phrase at some other level.”
Let us use the term ‘module’ for compo- “Cliticization Principle (CP): A lexeme that com-
nent in the first sense. Grimshaw argues that bines with a phrase at some level may combine at
Baker’s arguments show only that affixation another level with a unit immediately to its right
processes and the morphosyntactic effects of or left.” (Sadock 1990: 43)
those processes must be found in the same By ‘level’ Sadock means essentially ‘semantic,
component (sense (ii)), not that they must be- syntactic or morphological level of represen-
long to the same module (component sense tation’. By ‘combines with’ Sadock means
(i)). In principle, then, the affixation and its ‘forms part of’. Thus, in the case of noun in-
associated morphosyntactic changes could corporation, the paradigm case of the Incor-
take place in a component called ‘the lexicon’ poration Principle, we would say that an ob-
rather than in the syntax. On the other hand, ject noun forms a noun phrase, which then
nothing precludes us from claiming that forms part of a verb phrase in the syntax, but
some (presumably non-productive) processes at the morphological level that noun may
occur in the lexicon, while other (presumably combine with the head of that verb phrase,
productive) processes occur in the syntax. namely the verb. In the case of the possessive
Borer (1988) has developed a promising the- clitic, ’s, in English, we would say that the
ory of ‘Parallel Morphology’ along exactly clitic combines with a noun phrase in the syn-
these lines, on the basis of a detailed compar- tax, but with the word immediately to its left
ison of lexicalized compounds and the so- in the morphology.
Baker, Mark & Johnson, Kyle & Roberts, Ian Everaert, Martin & Evers, Arnold & Huybregts,
(1989), “Passive Arguments Raised”. Linguistic In- Riny & Trommelen, Mieke (1988, eds.), Morphol-
quiry 20, 219⫺250 ogy and Modularity. In Honour of Henk Schultink.
Barlow, Michael & Ferguson, Charles (1988, eds.), Dordrecht, Providence/RI: Foris (Publications in
Agreement in Natural Languages: Approaches, The- Language Sciences)
ories, Descriptions. Menlo Park/CA: CSLI Gazdar, Gerald & Klein, Ewan & Pullum, Geof-
Bauer, Laurie (1988), Introducing Linguistic Mor- frey & Sag, Ivan (1985), Generalized Phrase Struc-
phology. Edinburgh: Edinburgh University Press ture Grammar. Oxford: Blackwell
Bauer, Laurie (1990), “Be-heading the Word”. Grimshaw, Jane (1986), “A Morphosyntactic Ex-
Journal of Linguistics 26, 1⫺31 planation for the Mirror Principle”. Linguistic In-
Beard, Robert (1987), “Morpheme Order in a Lex- quiry 17, 745⫺750
eme/Morpheme Based Morphology”. Lingua 72, Grimshaw, Jane (1990), Argument Structure. Cam-
1⫺44 bridge/MA: MIT Press
Beard, Robert (1991), “Decompositional Composi- Haegeman, Liliane (1991), Introduction to Govern-
tion: The Semantics of Scope Ambiguities and ment and Binding Theory. Oxford: Blackwell
‘Bracketing Paradoxes’”. Natural Language and Hajdu, Péter & Domokos, Péter (1987), Die Ura-
Linguistic Theory 9, 195⫺230 lischen Sprachen und Literaturen. Budapest: Akadé-
Beard, Robert (1995) Lexeme-Morpheme Base miai Kiadó
Morphology. Albany: SUNY Press
Hudson, Richard (1977), “The Power of Morpho-
Bloomfield, Leonard (1933), Language. New logical Rules”. Lingua 42, 73⫺89
York: Holt
Hyman, Larry M. & Mchombo, Sam (1992),
Booij, Geert (1988), “The Relation Between Inheri- “Morphotactic Constraints in the Chichewa Verb
tance and Argument Linking: Deverbal Nouns in Stem”. In: Buszard-Welcher, Laura & Wee, Lio-
Dutch”. In: Everaert et al. (eds.), 57⫺74 nel & Weigel, William (eds.), Proceedings of the
Borer, Hagit (1988), “On the Parallelism Between 18th Annual Meeting of the Berkeley Linguistics So-
Compounds and Constructs”. In: Booij, Geert & ciety. Berkeley: University of California at Berke-
van Marle, Jaap (eds.), Yearbook of Morphology ley, 350⫺364
1988. Dordrecht: Foris, 45⫺66 Joseph, Brian D. & Smirniotopoulos, Jane C.
Borer, Hagit (1991), “The Causative-Inchoative Al- (1993), “The Morphosyntax of the Modern Greek
ternation: A Case Study in Parallel Morphology”. Verb as Morphology and Not Syntax”. Linguistic
The Linguistic Review 8, 119⫺158 Inquiry 24, 388⫺398
Botha, Rudolf (1983), Morphological Mechanisms. Kayne, Richard S. (1975), French syntax: the trans-
Oxford: Pergamon Press formational cycle. Cambridge, Mass. etc.: MIT
Chao, Yuen Ren (1968), A Grammar of Spoken Press (Current studies in linguistics series, 6)
Chinese. Berkeley, Los Angeles: University of Cali- Kiparsky, P. (1985), “Some Consequences of Lexi-
fornia Press cal Phonology”. Phonology Yearbook 2, 83⫺136
Chomsky, Noam (1970), “Remarks on Nominali- Klavans, Judith (1985), “The Independence of Syn-
zation”. In: Jacobs, Roderick & Rosenbaum, Peter tax and Phonology in Cliticization”. Language 61,
(eds.), Readings in English Tranformational Gram- 95⫺120
mar. Waltham/MA: Blaisdell, 84⫺221
Lapointe, Stephen (1981), “General and Restricted
Chomsky, Noam (1986), Barriers. Cambridge/MA:
Agreement Phenomena”. In: Moortgat, Michael &
MIT Press
van der Hulst, Harry & Hoekstra, Teun (eds.), The
Chomsky, Noam (1993), “A Minimalist Program Scope of Lexical Rules. Dordrecht: Foris, 215⫺254
for Syntax”. In: Hale, Kenneth & Keyser, Samuel
(eds.), The View from Building 20. Essays in Lin- Lapointe, Stephen (1988), “Toward a Unified The-
guistics in Honor of Sylvain Bromberger. Cam- ory of Agreement”. In Barlow & Ferguson (eds.),
bridge/MA: MIT Press, 1⫺52 67⫺88
Corbett, Greville (1987), “The Morphology-Syntax Leontiev, V. V. & Ajnewtegin, M. V. (1957), Učeb-
Interface”. Language 63, 299⫺345 nik Leg?orawetl?ajelemel [Textbook of Chukchee].
Leningrad: Gossudarstvennoe Učebno-Pedagogi-
Di Sciullo, Anna-Maria & Williams, Edwin (1987), ceskoe Izdatel’stvo
On the Definition of Word. Cambridge/MA: MIT
Press Lewis, G. L. (1967), Turkish Grammar. Oxford: Ox-
ford University Press
Doron, Edit (1988), “On the Complementarity of
Subject and Subject-Verb Agreement”. In: Barlow Lieber, Rochelle (1980), The Organization of the
& Ferguson (eds.), 201⫺218 Lexicon. Unpublished PhD dissertation, Mas-
Durie, Mark (1986), “The Grammaticalization of sachusetts Institute of Technology [distributed by
Number as a Verbal Category”. Proceedings of the Indiana University Linguistics Club, 1981]
Annual Meeting of the Berkeley Linguistics Society Lieber, Rochelle (1988), “Phrasal Compounds in
12, 355⫺370 English and the Morphology-Syntax Interface”.
Proceedings of the Twenty-fourth Meeting of the Scalise, Sergio (1984), Morphology in Generative
Chicago Linguistics Society [Parasession on Agree- Grammar. Dordrecht: Foris
ment in Grammatical Theory], 202⫺220 Selkirk, Elisabeth (1982), The Syntax of Words.
Lieber, Rochelle (1992), Deconstructing Morphol- Cambridge/MA: MIT Press
ogy. Chicago: University of Chicago Press Siegel, Dorothy (1977), “The Adjacency Condition
Lyons, Christopher (1991), “An Agreement Ap- and the Theory of Morphology”. Proceedings of
proach to Clitic Doubling”. Transactions of the the Eighth Meeting of the North Eastern Linguistics
Philological Society 88, 1⫺57 Society, 189⫺197
Marantz, Alec (1984), On the Nature of Grammati- Simpson, Jane & Withgott, Margaret (1986), “Pro-
cal Relations. Cambridge/MA: MIT Press nominal Clitic Clusters and Templates”. In: Borer,
Hagit (ed.), The Syntax of Pronominal Clitics, Or-
Marle, Jaap van (1985), On the Paradigmatic Di- lando: Academic Press, Syntax and Semantics 19,
mension of Morphological Creativity. Dordrecht: 149⫺174
Foris
Skorik, Pjotr Ja. (1948), Očerk po Syntaksisu Ču-
Matthews, P[eter] H. (1972), Inflectional Morphol- kotskogo Jazyka: Inkorporatsija. Leningrad: Uč-
ogy. Cambridge: Cambridge University Press pedgiz
McCloskey, James & Hale, Kenneth (1984), “On Skorik, Pjotr Ja. (1961), Grammatika Čukotskogo
the Syntax of Person-Number Inflection in Modern Jazyka, Vol. I. Moscow: Izdatel’stvo Akademii
Irish”. Natural Language and Linguistic Theory 1, Nauk.
487⫺534 Skorik, Pjotr, Ja. (1977), Grammatika Čukotskogo
Mithun, Marianne (1984), “The Evolution of Noun Jazyka, Vol. II. Moscow: Izdatel’stvo Akademii
Incorporation”. Language 60, 847⫺894 Nauk
Nedjalkov, Vladimir P. (1976), “Diathesen und Speas, Margaret (1991), “Functional Heads and
Satzstruktur im Tschuktschischen”. In: Lötsch, Inflectional Morphemes”. The Linguistic Review 8,
R. & Růžička, R. (eds.), Satzstruktur und Genus 389⫺417
Verbi. Berlin: Akademie Verlag, 181⫺211 Spencer, Andrew (1991), Morphological Theory.
Newmeyer, Frederick (1980), Linguistic Theory in Oxford: Blackwell
America. Chicago: University of Chicago Press Spencer, Andrew (1993), “Nominal Inflection and
the Nature of Functional Categories”. Journal of
Ohalla, Jamal (1990), “Sentential Negation, Relati-
Linguistics 28, 313⫺341
vised Minimality and the Aspectual Status of Aux-
iliaries”. The Linguistic Review 7, 183⫺231 Spencer, Andrew (1995), “Incorporation in
Chukchi”. Language 71, 439⫺489
Perlmutter, David (1988), “The Split-morphology
Hypothesis: Evidence from Yiddish”. In: Ham- Sproat, Richard (1985), On Deriving the Lexicon.
mond, Michael & Noonan, Michael (eds.), Theoret- Unpublished PhD dissertation, Massachusetts In-
ical Morphology. Orlando: Academic Press, 79⫺ stitute of Technology
101 Sproat, Richard (1988), “Bracketing Paradoxes,
Cliticization, and Other Topics: the Mapping be-
Pesetsky, David (1985), “Morphology and Logical
tween Syntactic and Phonological Structure”. In:
Form”. Linguistic Inquiry 16, 193⫺246
Everaert et al. (eds.), 339⫺360
Pollock, Jean-Yves (1989), “Verb Movement, Uni- Stump, Gregory (1991), “A Paradigm-theory of
versal Grammar, and the Structure of IP”. Linguis- Morphosemantic Mismatches”. Language 67,
tic Inquiry 20, 365⫺424 675⫺725
Quirk, Randolph & Greenbaum, Sidney (1973), A Suñer, Margarita (1988), “The Role of Agreement
University Grammar of English. London: Longman in Clitic-doubled Constructions”. Natural Lan-
Rizzi, Luigi (1982), Issues in Italian Syntax. Dor- guage and Linguistic Theory 6, 391⫺434
drecht: Foris Williams, Edwin (1981), “On the Notions ‘Lexi-
Rosen, Sara (1989), “Two Types of Noun Incorpo- cally Related’ and ‘Head of a Word’”. Linguistic
ration: A Lexical Analysis”. Language 65, 294⫺317 Inquiry 12, 245⫺274
Young, Robert, W. & Morgan, William (1980), The
Sadler, Louisa (1988), Welsh Syntax: A Govern-
Navajo Language. A Grammar and Colloquial Dic-
ment-Binding Approach. Beckenham: Croom Helm
tionary. Albuquerque: University of New Mexico
Sadock, Jerrold (1985), “Autolexical Syntax: A Press
Proposal for the Treatment of Noun Incorporation Zwicky, Arnold (1985), “How to Describe Inflec-
and Similar Phenomena”. Natural Language and tion”. Proceedings of the 11th Annual Meeting of
Linguistic Theory 3, 379⫺439 the Berkely Linguistics Society, 372⫺386
Sadock, Jerrold (1990), Autolexical Syntax. Chi-
cago: University of Chicago Press Andrew Spencer, Colchester (Great Britain)
1. The relations between morphology and making crucial use of the syllable structure of
phonology the stem. Moreover, phonological alternations
2. The morpheme as a phonotactic domain may have a purely morphological function,
3. Allomorphy as is the case for ablaut phenomena in Ger-
4. The place of morphonology in the grammar
manic languages like the expression of tense
5. Morphological structure and the domain of
phonological rules in the pair sing ⫺ sang.
6. Phonological conditioning of morphological The study of the interaction between mor-
operations phology and phonology is traditionally re-
7. References ferred to as morpho-phonology or morphonol-
ogy for short (Trubetzkoy 1929; 1931). Tru-
betzkoy defined this area of investigation as
1. The relations between that part of linguistics which studies the mor-
morphology and phonology phological use of phonological differences
(Trubetzkoy 1929: 85). The following sub-
This article deals with the ways in which mor- tasks are distinguished by him for morpho-
phology interacts with phonology. The place nology:
of morphology in the language system in (1) “(1) die Lehre von der phonologischen
general is described in Art. 3. Here, we focus Struktur der Morpheme;
on its relation to a specific subsystem, pho- (2) die Lehre von den kombinatori-
nology. schen Lautveränderungen, welche
Phonology deals basically with two types die Morpheme in den Morphemver-
of phenomena: the distribution of sounds, bindungen erleiden;
and alternations in the phonological form of (3) die Lehre von den Lautwechselrei-
morphemes. hen, die eine morphologische Funk-
There are two ways in which morphology tion erfüllen.” (Trubetzkoy 1931:
and phonology interact. On the one hand, 161f.)
morphological information plays an impor-
tant role in the phonological system of a lan- The first area is that of the phonotactic prop-
guage because the distribution of sounds and erties of morphemes, the second area is that
the occurrence of alternations is partially de- of allomorphy, and the third one is that of
termined by the morphological structure of the morphological use of phonological al-
words. For instance, certain combinations of ternations. We will take Trubetzkoy’s classifi-
sounds may only occur within a morpheme, cation as the starting point of our discussion.
These three areas of investigation will be dis-
or only across morpheme boundaries. In
cussed in 2⫺4. Subsequently, I will discuss
Germanic languages, for example, a nasal
the relation between morphological structure
consonant is always homorganic with a
and the domains of phonological rules (5),
following obstruent within the same mor-
and the phonological conditioning of mor-
pheme, whereas this is not necessarily the
phological operations (6).
case across morpheme boundaries (compare
the English simplex word damp to the com-
pound timetable [taimteibl]). In Polish, a 2. The morpheme
word-final obstruent is always voiceless, un- as a phonotactic domain
like word-internal obstruents. Thus, in de-
scribing the alternative between [b] and [p] in As pointed out above, the first area of investi-
chleb [xlep] ‘bread-sg’ and chleby [xlebı] gation that belongs to morphonology is the
‘bread-pl’, reference to the notion word study of the phonological structure of mor-
boundary, i.e. to non-phonological, gram- phemes.
matical information, is crucial. This does not mean that the morpheme is
On the other hand, morphological pro- the most important domain of phonotactic
cesses may make use of phonological infor- constraints. There is general agreement that
mation. For instance, many languages make the syllable rather than the morpheme is the
use of reduplication processes in which one most important domain of phonotactic con-
or more syllable of the stem is copied, thus straints: a phonemic string is phonotactically
well-formed if it can be divided into one or we speak of the two allomorphs [hut] and
more well-formed syllables. [hud] of the morpheme hoed (cf. Ch. VII),
In early generative phonology phonotactic and observe an alternation in the phonetic re-
constraints were interpreted as morpheme alization of this morpheme.
structure rules or morpheme structure condi- The example given here is a case of non-
tions. However, since Hooper (1976) and suppletive allomorphy since the two allo-
Kahn (1976) it has been generally accepted morphs are phonetically similar. In the case
that phonotactic constraints should primarily of suppletive allomorphy, the phonetic sim-
be seen as constraints on syllable structure. ilarity is (almost) absent, as in the different
Nevertheless, in the literature one also allomorphs of the verb to be (am, are, is,
finds the observation that languages have was, were).
phonotactic constraints that hold specifically The distribution of allomorphs can be gov-
for the domain of the morpheme. An exam- erned by morphological structure (see 5), or
ple is that within Dutch lexical morphemes a by adjacent words. A case of selection by an
cluster of two obstruents is always voiceless, adjacent word is the choice in English be-
whereas voiced obstruent clusters do occur tween the determiners a and an, which de-
within words across morpheme boundaries, pends on the nature of the first segment of
as in the past tense of the verb tobben ‘to toil, the next word: an before a vowel-initial word,
tob-de [tcbde], with the past tense suffix -de. a before a consonant-initial one.
The morpheme is also relevant for pho- Allomorphy can be dealt with in a number
notactic generalizations in that different of ways (cf. Ch. VII). In the case of non-sup-
classes of morphemes obey different phono- pletive allomorphy one basic form or repre-
tactic constraints (Jakobson 1949 a [1971: sentation may be assumed for each mor-
106⫺108]; Nida 1949: 66). For instance, in pheme, from which the different phonetic
many languages the phonological form of af- forms are derived by means of phonological
fixes is shorter than that of lexical mor- rules. This basic representation is referred to
phemes. In Bantu languages roots are pre- as basic form (Bloomfield 1939), or as un-
dominantly CVCV, but affixes are almost al- derlying representation (in generative phonol-
ways of the CV type (Nida 1949: 66). Dutch ogy). In the case of the Dutch word hoed ‘hat’
affixes consist of at most two syllables, unlike discussed above, the underlying form is 兩 hud 兩.
lexical morphemes, which can consist of If it is not followed by another suffix, it will
more; suffixes can also be vowelless, unlike be syllabified as (hud)s, where s stands for
lexical morphemes (Booij 1995). In Semitic ‘syllable’. A phonological rule of syllable-fi-
languages lexical roots usually consist of a nal devoicing will then change the 兩 d 兩 into [t],
skeleton of three consonants, whereas the in- and hence the phonetic form is [hut]. If the
flectional morphemes consist of a sequence morpheme is followed by another one, as in
of vowels (cf. Art. 56). A survey of this kind the plural form hoed⫹en ‘hats’, the syllable
of data is given in Dressler (1985: 219⫺245). structure is (hu)s(den)s. Here, the /d/ will not
Prosodic restrictions may also pertain to devoice because it is not in syllable-final posi-
the domain of the morpheme. For instance, tion, and thus the phonetic form is [huden].
the rhymes of morpheme-internal Dutch syl- Such analyses imply that allomorphy is
lables consist of at most a long vowel or a partially accounted for by the phonological
short vowel plus one consonant; longer sylla- component of a language. Obviously, this
bles only occur at the end of morphemes. kind of allomorphy can be easily accounted
In sum, the morpheme functions as a do- for by the phonological component because
main of phonological generalizations. the relevant generalisation holds across the
board. The rule involved is an automatic pho-
nological rule (Bloomfield 1933: 211).
3. Allomorphy Trubetzkoy accounted for such alternations
which are the effect of the neutralisation of
A morpheme may have more than one pho- phonemic contrast in certain positions, by
nological form. For instance, the Dutch pho- means of the notion archiphoneme (cf. Art.
nemic strings [hut] and [hud] that occur in 50), and did not consider them to be part of
hoed [hut] ‘hat’ and hoeden [huden] ‘hats’ re- morphonology proper. A word like hoed
sepctively are different phonetic realisations would have been given the representation
of the same morpheme hoed ‘hat’. Therefore, /huD/, where the capital D stands for the
common properties of /t/ and /d/. Such a re- morphs are all listed can also be found in
presentation does not suffice, however, to Hudson (1974) and Lieber (1981).
predict that the plural form of this word is There are many alternations which, al-
[huden], and not [huten]. For Bloomfield, though they are regular, i.e. not lexically gov-
however, such alternations belong to morpho- erned, can only be accounted for by referring
phonemics (Bloomfield 1939). A historical to morphological information, and thus in-
survey of the development of these notions is stantiate the class of morphonological rules.
given in Kilbury (1976) and Anderson (1985). An example from Dutch is the allomorphy of
Post-Bloomfieldian phonology, which stress- the diminutive suffix, which has five different
ed the autonomy of phonology, and required phonetic shapes, depending on the phonolog-
a biunique mapping between the segments of ical properties of the final syllable of the base
the phonetic representation and those of the word: [je] as in huis-je ‘little house’, [pje] as in
phonological representation, did not allow riem-pje ‘little belt’, [kje] as in konin-kje ‘little
for such analyses because they make use of king’, [etje] as in lamm-etje ‘little lamb’, and
non-phonetic information (the existence of [tje] as in traan-tje ‘little tear’. The distribu-
two alternating forms) in the phonological tion of these five allomorphs is given in (2)
analysis. It was Pike (1947) who opposed this (Booij 1995):
rigid separation of phonology from the rest
of the grammar in post-Bloomfieldian pho- (2) je after obstruent consonants;
nology, and who pointed out that grammati- pje after /m/ preceded by a long vowel, a
cal information is often undispensable for a schwa or a consonant;
proper phonological analysis. kje after /n/ preceded by a stressless
Not all cases of allomorphy within com- vowel;
plex words can be accounted for in terms of etje after stressed short vowels (except
automatic phonological rules. To begin with, schwa) followed by a liquid or nasal;
allomorphy may be the historical relict of a tje in all other cases
once active phonological process like deletion
of word-final schwa. Many lexical mor- Suppose that we derive the allomorph /je/
phemes of Dutch, for example, have two allo- from the basic form /tje/ by means of a rule
morphs, one ending in schwa, and a shorter of /t/-deletion that deletes /t/ after an ob-
one, without schwa, which arose due to a his- struent. This rule will then be crucially condi-
torical process of schwa-apocope, which is no tioned by the morphological property dimin-
longer active. The word aarde [a:rde] ‘earth’ utive since there is no general phonological
has an allomorph aard [a:rt] which has to be rule of Dutch that deletes every /t/ after an
used when this word occurs in the non-head obstruent, witness past tense forms such as
position of compounds, as in aardbei ‘straw- hoopte [hopte] ‘hoped’, where the /t/ remains
berry’. That is, the rule of schwa-deletion is after the obstruent /p/. Also, in a sentence
not an automatic phonological rule: both al- like Koopt je moeder dat? ‘Does your mother
lomorphs must be listed, and the distribution buy that?’, the final /t/ of koopt does not de-
of the allomorphs can only be stated by refer- lete before je, which underscores the non-au-
ring to morphological structure and/or spe- tomatic character of the /t/-deletion involved
cific complex words. Thus, such allomorphy here. Therefore, there are two options. One
will not be accounted for by morphonology. is to assume one underlying form for the di-
The voiceless ⫺ voiced alternation in Eng- minutive suffix, and a set of phonological
lish singular-plural pairs such as house [haws] rules that crucially mention the property di-
⫺ houses [hawziz] and wife [waif] ⫺ wives minutive in their domains of application.
[waivz] is restricted to an enumerable set The other is to list the five allomorphs of the
of nouns (listed in Bloomfield 1933: 213f.). diminutive suffix, and to state their distribu-
Such lexically or morphologically governed tion in terms of the phonological properties
alternations are called morpholexical varia- of the stem to which they attach. The draw-
tions by Bloomfield (1939 [1970: 352]). They back of the latter approach is that it does not
can be relegated to morphology proper. express that the competing diminutive suf-
The common idea is that such allomorphs fixes discussed here are phonetically similar.
are all listed, and that they can be related In other words, this description does not dis-
by a kind of reduncancy rule, called via- tinguish these alternations from cases in
rule (Vennemann 1972), or relational rule which competing suffixes are phonologically
(Tiersma 1978). The idea that such allo- unrelated.
In some cases, the second option is the for German as in Apfel [apfel] ‘apple’ ⫺ Äpfel
only possible one even though phonological [ipfel] ‘apples’. In these cases this morpho-
conditioning is involved. As pointed out by logical function of phonological alternations
Carstairs (1988), languages may have com- is the effect of the morphologization of a
peting affixes which are phonologically unre- once regular phonological process. In Semitic
lated. Yet, their distribution can be stated by languages, on the other hand, the vowel al-
referring to phonological properties of the ternations in the different forms of the verbal
stem. For instance, in Hungarian the suffix paradigm are themselves ‘morphemes’: each
-ol is used for the 2nd sing. indef. pres. indic- vowel pattern has a particular meaning (cf.
ative form of the verb after sibilants and af- Art. 125).
fricates, and -(a)sz elsewehere. These suffixes
cannot be derived from a common underly-
ing form, although their distribution is stat- 4. The place of morphonology
able in phonological terms. in the grammar
Another example of a non-automatic al-
ternation is the alternation between front and In the linguistic literature the question has
back vowels in some French morphemes, been discussed whether morphonology is a
both in lexical roots and suffixes. The gener- separate component of the grammar (Marti-
alization is that the rightmost vowel of a net 1965; Linell 1977; Dressler 1981; 1985;
morpheme, if it is low, becomes [⫹back] if Wurzel 1981). The answer is usually negative:
that morpheme is immediately followed morphonology is claimed to belong either to
within the same word by a morpheme from morphology (Martinet 1965; Linell 1977), or
the class of so-called learned morphemes is divided between phonology and morphol-
(Dell & Selkirk 1978: 8). Some examples are ogy, where grammatically conditioned al-
given in (3): ternations, i.e. alternations without any pho-
nological conditioning, belong to morphol-
(3) vain [vĩ] ‘vain’ ⫺ vanité [vanite] ‘vanity’;
ogy, and the rest to phonology (Wurzel
clair [klir] ‘clear’ ⫺ clarifier [klarifje] ‘to
1981: 417). Other linguists consider morpho-
clarify’; immortel [imcrtil] ⫺ immortalité
nology as belonging neither to morphology
[imcrtalite] ‘immortality’
or to phonology; “it mediates between both
The class of morphemes that trigger vowel components without being itself a basic com-
backing cannot be defined in purely phono- ponent like morphology or phonology”
logical terms: they have to be marked with (Dressler 1985: 4).
the feature [⫹learned]. So we either assume a In the cases of ablaut and umlaut it is not
phonological rule that mentions [⫹learned] so easy to decide where to account for the
in its structural description, or we list both alternations involved. The alternation be-
allomorphs, and state which allomorph is to tween back and front vowels exemplified by
be used for which set of suffixes. German Apfel ⫺ Äpfel can be seen as a case
Allomorphy may also be conditioned by a of morphology, since vowel alternation is one
word class feature. For instance, within Rus- of the available morphological mechanisms
sian verbal stems “any morpheme which ends in natural languages. However, we might also
in a vowel loses that vowel before a suffix claim that this vocalic alternation should be
begining with a vowel” (Jakobson 1948 accounted for by a rule of phonology that is
[1971: 124]). This generalization crucially re- conditioned by the presence of the morpho-
fers to the word class of the stems involved. logical property plural. An argument for
As mentioned above, there are also al- the phonological approach in this particular
ternations that have a purely morphological case is that the back/front alternation also
function, the third area of morphonology. occurs in nouns that take the regular plural
For instance, whereas in the pair knife ⫺ suffix -e, e. g. Gast [gast] ⫺ Gäste [giste]. In
knives the plural character of knives not only this case the phonological alternation seems
manifests itself in the voicedness of the stem- to be a concomitant effect of pluralization
final fricative, but is also indicated by the rather than the expression thereof, and this
presence of a plural suffix, realized as [z], the suggests a phonological interpretation of
plural character of feet is only indicated by this alternation.
the fact that the vowel /i:/ stands in opposi- Similarly, ablaut in Germanic languages
tion to the vowel /u/ of the singular form (his- not only occurs as the sole marker of mor-
torically a case of umlaut). This also holds phosyntactic properties such as past tense
and past participle, but also in addition to within the phonological component, mor-
marking by affixation. Consider, for in- phonological rules must apply first. This is
stance, the following examples of inflection reminiscent of Dressler’s position that mor-
of Dutch strong verbs: phonology is an intermediate component be-
tween morphology and phonology. However,
(4) pres. past participle Anderson also argues that sometimes a pho-
(a) roep riep geroepen ‘to call’ nological rule may have to precede a rule that
loop liep gelopen ‘to walk’ accounts for morpholexical variation, i.e. a
(b) neem nam genomen ‘to take’ morpholexical rule.
vind vond gevonden ‘to find’ The most elaborated theory of the interac-
tion between phonology and morphology
In the past tense forms ablaut is the only
within the framework of generative phonol-
marker, whereas in the past participles we al-
ogy is that of Lexical Phonology (cf. Kipar-
ways get the discontinuous affix ge … en,
sky 1982; 1985). Its basic claim is that the
without (4a) or with (4b) ablaut. That is, the
rules of morphology and the rules of phonol-
same vowel alternation sometimes functions
ogy apply in tandem, i.e. morphological and
as the sole marker of a morphosyntactic
property, but as one of the markers in other phonological rules are interspersed (cf. Booij
cases. 1981). In other words, phonological opera-
The most reasonable position is that such tions may both precede and follow morpho-
purely grammatically conditioned sound al- logical operations. The basic idea is as fol-
ternations are morphological phenomena, lows: given an underived form A, we first ap-
whether symbolizing grammatical categories ply the rules of phonology to that form, giv-
alone, or along with other means (Wurzel ing us A⬘ (the first cycle). We may then apply
1981: 417; Anderson 1992). a morphological rule to A⬘, thus creating a
In standard generative phonology, mor- second cycle on which the rules of (word)
phonology is part of phonology, and there is phonology apply, and so on.
no separate morphonological component. A specific hypothesis about the role of
The following characteristic of generative morphological structure in the application of
phonology is given in Goldsmith (1995: 9): phonological rules within the framework of
Lexical Phonology is the so-called Revised
(5) “The motto ‘minimize allomorphy’ re- Alternation Condition which states that neut-
mains today’s watchword, in the sense ralizing phonological rules apply in derived
that in practice morphology continues to environments only (Kiparsky 1973: 65). A
be concerned with the linear order and phonological string counts as a derived envi-
constituent structure of words, and with ronment if it is the result of the application
making a choice of which morphemes are of either a morphological or a phonological
to be employed to realize a set of mor- rule.
phosyntactic features in a given sentence; Consider the Polish rule of palatalization
but contextually determined variations in that turns coronal consonants into prepala-
the realization of a given morpheme will tals before front vowels (Rubach 1984: 59⫺
be to the extent possible be accounted for 70). For instance, /s/ becomes a palatalized
phonologically.” [s⬘] in pasie ‘belt-loc-sg’ with the morpholog-
Generative phonology tries to account for as ical structure /pas⫹e/. The nom.sg. form is
much allomorphy as possible by making use pas [pas] ‘belt’. The Revised Alternation Con-
of underlying forms and derivations. By dition can now be invoked to explain why in
making use of complex derivations and the loc.sg. form of the word serwis [serv⬘is]
sometimes also of extrinsic ordering of rules ‘service’, serwisie [serv⬘is⬘e], the first /s/ does
it is possible to account for a considerable not palatalize, whereas the second does. The
amount of allomorphic relations in terms of first sequence /se/ is not a derived environ-
common underlying forms. ment, because it is already present in the un-
Anderson (1975) has pointed out that the derlying form of serwis. The morpheme-fi-
application of morphonological rules usually nal /s/, however, stands before an /e/ which
has to precede that of automatic phonologi- has been introduced by a morphological rule.
cal rules. Hence, the distinction between mor- Hence, this sequence /se/ is a derived environ-
phonology and phonology might be ex- ment, and thus the rule can apply: /se/ be-
pressed by a principle of rule ordering that, comes [s⬘e].
A consequence of the Revised Alternation one morpheme (i.e. it has an internal mor-
Condition is that both /s/ and /s⬘/ occur in phological structure), on the other hand it is
underlying forms. In sejm ‘parliament’ and hierarchically organized into one or more syl-
sekunda ‘second’ the initial consonant is [s], lables, i.e. it has a prosodic structure as well.
but in siew ‘sowing’ and siwy ‘grey’ it is [s⬘]. These two structures are not necessarily iso-
Therefore, there is a lexical contrast between morphic. For instance, the Dutch word
these two sounds, and thus the rule of pala- hoeden /huden/ ‘hats’ consists of the lexical
talization is a neutralizing rule: in certain morpheme /hud/ ‘hat’ and the plural suffix
contexts, it neutralizes an underlying distinc- /-en/, but the prosodic structuring is different:
tion between segments. it contains the syllables (hu)s and (den)s. In
In sum, the Revised Alternation Condition this example, the morphological structure of
makes use of morphological information in the word does not influence its syllabifica-
determining the applicability of a phonologi- tion: hoeden is syllabified as if it were a mor-
cal rule since one of the cases of ‘derived en- phologically simplex word. In other cases the
vironment’ is ‘derived by application of a
morphological structure of a word does co-
morphological rule’.
determine its syllabification, and thus it de-
A further development of this theory of
rule application is the idea that the class of termines indirectly the way in which phono-
rules subject to the derived environment con- logical rules that are sensitive to the prosodic
straint is the class of cyclic rules rather than structuring of words apply.
the class of (obligatorily) neutralizing rules. Compounds are classical examples of this
A cyclic rule is a rule which applies in a cyclic non-isomorphy of morphological and pho-
fashion, as determined by the morphological nological structure. Compounds are one
structure of the phonological string. For in- word from the morphological and syntactic
stance, given a word with the structure point of view, but in many languages each
[[[…]A…]B…]C the set of cyclic rules, which member of a compound corresponds with a
are linearly ordered, first applies to the pho- separate phonological (or prosodic) word.
nological material within the brackets A, Since the phonological word is usually the
then to the material within the brackets B, domain of syllabification, the boundaries be-
and finally to the material within the outer tween the members of a compound coincide
brackets C. with syllable boundaries. For instance, the
The assumption that certain phonological English compound meat eater is syllabified as
rules apply cyclically is also motivated by the (meat)s(ea)s(ter)s, not as (mea)s(tea)s(ter)s;
observation that morphological operations the latter syllabification would be expected if
may make use of derived phonological infor- the word had no compound structure.
mation (cf. 6). An additional type of relevant evidence is
Whereas the notion cyclicity refers to the available for German. Vowel-initial words of
hypothesis that phonological rules apply in a this language predictably begin with a glottal
cyclic fashion, the notion strict cyclicity that stop, and this glottal stop also occurs word-
is also used in Lexical Phonology refers to
internally at the beginning of a vowel-initial
the hypothesis that it is the set of cyclic rules
member of a compound, as in farb-echt [farb
that is subject to the derived environment ?
constraint (Mascaró 1976). ixt] ‘color fast’.
Thus, the theory of Lexical Phonology can Another illustration comes from Hung-
be seen as an attempt to account for morpho- arian vowel harmony. In Hungarian vowels
nological phenomena in the phonological are harmonic with respect to the feature
component, and to do away with a separate [back] within the domain of the phonological
morphonological component (cf. Booij 1994 word (Booij 1984). Therefore, the members
for a general survey of Lexical Phonology, of a compound may be disharmonic, as is il-
and Spencer 1988; 1991, for a discussion of lustrated by the compound word Budapest:
the position of morphology in generative Buda contains back vowels, and pest a front
grammar). vowel.
Prefixes and suffixes often differ in their
5. Morphological structure and the effects on the delimitation of phonological
domains. In Slavic languages “a word with a
domain of phonological rules prefix is conceived of as a kind of com-
The phonemic string that constitutes a word pound” (Jakobson 1949 a [1971: 107]). “The
is structured according to two dimensions. root plus derivational and inflectional suf-
On the one hand, it may consist of more than fixes forms a simple word. A root preceded
this can be expressed as the alignment condi- Another example is that in Dutch the adjecti-
tion that the stem boundary align with a syl- val suffix -ief only attaches to nouns in un-
lable boundary. stressed /i/ (e.g. agréssie ‘aggression’ ⫺ agres-
sief ‘aggressive’). Otherwise, -isch has to be
6. Phonological conditioning of selected, as in hysterı́e ‘hysteria’ ⫺ hysterisch
‘hysterical’ (Booij & Rubach 1987: 2).
morphological operations The phonological structure of base words
As pointed out in 1, not only may phonologi- also plays a role in partial reduplication. In
cal rules be conditioned by morphological Classical Greek perfective forms, the first
properties, but, vice versa, morphological segment of the verbal stem must be copied,
processes may also be conditioned by phono- and precedes the fixed vowel /e/, as in luo ⫺
logical properties. More specifically, a mor- leluka ‘to loose’. In Yidiny nominal reduplica-
phological process may impose phonological tion, the first two syllables of the input form
requirements on its input forms. For in- are copied (Dixon 1977):
stance, the Dutch nominalizing suffix -erd
(8) mulari ‘initiated man’ ⫺ mulamulari ‘ini-
/ert/ that derives personal names from adjec-
tiated men’
tives (e.g. viezerd ‘dirty person’ from vies
kintalpa ‘lizard’ ⫺ kintalkintalpa ‘lizards’
‘dirty’) cannot attach to adjectives ending in
/er/: dapper ‘brave’ ⫺ *dappererd (Schultink In this case the copying proces has to be de-
1962: 203). Similarly, the Dutch suffix -heid fined in terms of syllables, and this accounts
‘-ness’ does not combine with base words for the fact that we get a CVCV copy in the
ending in schwa (Van Marle 1985: 231). first example, but a CVCCVC copy in the se-
Many of these phonological restrictions on cond case. These examples thus show that
morphology can also be interpreted as pho- morphological operations must have access
nological output conditions on morphologi- to the phonological properties of the stems to
cally complex words. For instance, Dutch which they apply. The class of morphological
does not allow for the sequence /rer/. Such a operations that are defined in terms of pro-
sequence is avoided either by blocking the sodic properties of the stems and those of the
word formation process, as is the case with resulting complex words is also referred to as
-erd-affixation discussed above, or by choos- prosodic morphology (McCarthy & Prince
ing another allomorph of the affix. In the 1990).
case of comparative forms in Dutch, the allo-
morph -der is chosen instead of -er after ad-
jectival stems ending in /r/; thus we get duis- 7. References
terder ‘darker’ instead of the ill-formed *duis-
terer. Anderson, Stephen R. (1975), “On the Interaction
Besides segmental restrictions we also find of Phonological Rules of Various Types”. Journal
cases in which prosodic properties of the in- of Linguistics 11, 39⫺62
put forms play a role. For instance, the Eng- Anderson, Stephen R. (1985), Phonology in the
lish comparative suffix -er only attaches to Twentieth Century. Chicago: Chicago University
monosyllabic adjectives (nice ⫺ nicer), and to Press
bisyllabic adjectives with a light final syllable Anderson, Stephen R. (1992), A-morphous Mor-
(happy ⫺ happier), and to a few other disyl- phology. Cambridge: Cambridge University Press
labic adjectives including common, handsome, Aronoff, Mark & Kean, Mary-Louise (1980, eds.),
polite, quiet and wicked. In all other cases, the Juncture. Saratoga/CA: Anma Libri (Studia Lin-
periphrastic form with more has to be used. guistica et Philologica 7)
In many cases, the periphrastic construction
Basbøll, Hans (1978), “Schwa, jonctures et syllabi-
is also possible with adjectives that do allow fication dans les représentations phonologiques du
for -er-affixation. So we see that syllable français”. Acta Linguistica Hafniensia 16, 147⫺182
structure plays a role in morphology.
Stress properties may also play a role. For Bloomfield, Leonard (1933), Language. New
York: Holt
instance, the English deverbal nominal suffix
-al only attaches to verbs with main stress on Bloomfield, Leonard (1939), “Menomini Morpho-
the final syllable: phonemics”. Traveaux du Cercle Linguistique de
Prague 8, 105⫺115 [reprinted in: Hockett, Charles
(7) (a) betráy ⫺ betrayal, refúse ⫺ refusal F. (1970, ed.), A Leonard Bloomfield Anthology.
(b) órganize ⫺ *organizal, encóurage ⫺ Bloomington, London: Indiana University Press,
*encourageal 351⫺362]
Booij, Geert E. (1981), “Rule Ordering, Rule Ap- Hooper, Joan Bybee (1976), An Introduction to
plication and the Organization of Grammars”. In: Natural Generative Phonology. New York etc.: Aca-
Dressler et al. (eds.), 45⫺56 demic Press
Booij, Geert E. (1983), “Principles and Parameters Hudson, Grover (1974), “The Representation of
in Prosodic Phonology”. Linguistics 21, 249⫺280 Non-productive Alternation”. In: Anderson, John
Booij, Geert E. (1984), “Neutral Vowels and the M. & Jones, Charles (eds.), Historical Linguistics,
Autosegmental Analysis of Hungarian Vowel Har- Vol. II: Theory and Description in Phonology. Am-
mony”. Linguistics 22, 629⫺641 sterdam: North Holland, 203⫺229
Jakobson, Roman (1948), “Russian Conjugation”.
Booij, Geert E. (1985), “Coordination Reduction
Word 4, 155⫺167 [reprinted in: Jakobson
in Complex Words: A Case for Prosodic Phonol-
1971: 119⫺129]
ogy”. In: van der Hulst, Harry & Smith, Norval S.
H. (eds.), Advances in Non-linear Phonology. Dor- Jakobson, Roman (1949 a), “The Phonemic and
drecht: Foris, 143⫺160 Grammatical Aspects of Language in Their Inter-
relations”. Actes du Sixième Congrès International
Booij, Geert E. (1994), “Lexical Phonology: A Re-
des Linguistes, Paris, 1948. Paris: Klincksieck [re-
view”. Lingua e Stile 19, 525⫺555
printed in: Jakobson 1971: 103⫺114]
Booij, Geert E. (1995), The Phonology of Dutch.
Jakobson, Roman (1949 b), “Comparative Slavic
Oxford: Clarendon Press
Grammar”. In: Jakobson, Roman, Slavic Lan-
Booij, Geert E. & Rubach, Jerzy (1984), “Morpho- guages. New York: Columbia University Press [re-
logical and Prosodic Domains in Lexical Phonol- printed in: Jakobson 1971: 115⫺118]
ogy”. Phonology Yearbook 1, 1⫺27
Jakobson, Roman (1971), Selected Writings, Vol.
Booij, Geert E. & Rubach, Jerzy (1987), “Postcy- II: Word and Language. The Hague, Paris: Mouton
clic versus Postlexical Rules in Lexical Phonology”. Kahn, Daniel (1976), Syllable-based Generaliza-
Linguistic Inquiry 18, 1⫺44 tions in English Phonology. Bloomington/IN: Indi-
Carstairs, Andrew (1988), “Some Implications ana University Linguistics Club
of Phonologically Conditioned Suppletion”. In: Kilbury, James (1976), The Development of Mor-
Booij, Geert & van Marle, Jaap (eds.), Yearbook of phophonemic Theory. Amsterdam: John Benjamins
Morphology 1988. Dordrecht: Foris, 67⫺94 (Studies in the History of Linguistics 10)
Chomsky, Noam & Halle, Morris (1968), The Kiparsky, Paul (1973), “Phonological Representa-
Sound Pattern of English. New York: Harper & tions”. In: Fujimura, Osamu (ed.), Three Dimen-
Row sions of Linguistic Theory. Tokyo: TEC Cy, 87⫺136
Cohen, Abigail C. (1989), “Stress in Indonesian Kiparsky, Paul (1982), “From Cyclic Phonology to
and Bracketing Paradoxes”. Natural Language and Lexical Phonology”. In: van der Hulst, Harry &
Linguistic Theory 7, 167⫺216 Smith, Norval S. H. (eds.), The Structure of Phono-
Dell, François & Selkirk, Elisabeth (1978), “On a logical Representations, Part 1. Dordrecht: Foris,
Morphologically Governed Vowel Alternation in 131⫺175
French”. In: Keyser, Samuel J. (ed.), Recent Trans- Kiparsky, Paul (1985), “Some Consequences of
formational Studies in European Languages. Cam- Lexical Phonology”. Phonology Yearbook 2, 83⫺
bridge/MA: MIT Press (Linguistic Inquiry Mono- 138
graphs 3), 1⫺51
Lieber, Rochelle (1981), On the Organization of the
Dixon, R[obert] M. W. (1977), A Grammar of Yi- Lexicon. Bloomington/IN: Indiana University Lin-
diny. Cambridge etc.: Cambridge University Press guistics Club [reprint 1990, New York: Garland
Dressler, Wolfgang U. (1981), “Outlines of a Press]
Model of Morphology”. In: Dressler et al. (eds.), Linell, Per (1977), “Morphophonology as Part of
113⫺122 Morphology”. In: Dressler, Wolfgang U. & Pfeif-
Dressler, Wolfgang U. (1985), Morphonology: The fer, Oskar E. (eds.), Phonologica 1976. Akten der
Dynamics of Derivation. Ann Arbor: Karoma Pub- dritten Internationalen Phonologietagung, Wien,
lishers 1.⫺4. September 1976. Innsbruck: Institut für
Sprachwissenschaft (Innsbrucker Beiträge zur
Dressler, Wolfgang U. & Pfeiffer, Oskar E. & Sprachwissenschaft), 9⫺20
Rennison, John R. (1981, eds.), Phonologica 1980.
Akten der Vierten Internationalen Phonologieta- Marle, Jaap van (1985), On the Paradigmatic Di-
gung, Wien, 29. Juni ⫺ 2. Juli 1980. Innsbruck: In- mension of Morphological Creativity. Dordrecht:
stitut für Sprachwissenschaft (Innsbrucker Beiträge Foris
zur Sprachwissenschaft 36) Martinet, André (1965), “De la morphonologie”.
Goldsmith, John (1995), “Introduction”. In: Gold- La Linguistique 1, 15⫺30
smith, John (ed.), The Handbook of Phonological Mascaró, Joan (1976), Catalan Phonology and the
Theory. Oxford/UK, Cambridge/MA: Blackwell, Phonological Cycle. Bloomington/IN: Indiana Uni-
1⫺15 versity Linguistics Club
McCarthy, John & Prince, Alan (1990), “Foot and Siegel, Dorothy (1980), “Why There is No ⫽
Word in Prosodic Morphology”. Natural Language Boundary”. In: Aronoff & Kean (eds.), 131⫺134
and Linguistic Theory 8, 209⫺282 Spencer, Andrew (1988), “Arguments for Morpho-
McCarthy, John & Prince, Alan (1994), “General- lexical Rules”. Journal of Linguistics 24, 1⫺29
ized Alignment”. In Booij, Geert & van Marle, Spencer, Andrew (1991), Morphological Theory.
Jaap (eds.), Yearbook of Morphology 1993. Dor- Oxford: Blackwell
drecht: Kluwer, 79⫺153
Tiersma, Pieter M. (1978), “Bidirectional Leveling
Nespor, Marina & Vogel, Irene (1986), Prosodic as Evidence for Relational Rules”. Lingua 45,
Phonology. Dordrecht: Foris 65⫺77
Nida, Eugene A. (21949), Morphology: The De- Trubetzkoy, N[icolaj] S. (1929), “Sur la ‘morpho-
scriptive Analysis of Words. Ann Arbor/MI: Uni- nologie’ ”. Travaux du Cercle Linguistique de
versity of Michigan Press [11946] Prague 1, 85⫺88
Pike, Kenneth L. (1947), “Grammatical Prerequi- Trubetzkoy, Nikolaj S. (1931), “Gedanken über
sites to Phonemic Analysis”. Word 3, 155⫺172 Morphonologie”. In: Réunion phonologique inter-
Rubach, Jerzy (1984), Cyclic and Lexical Phonol- nationale tenue à Prague 1930. Prague: Jednota
ogy. The Structure of Polish. Dordrecht: Foris Československých matematiků a fysiků (Travaux
(Studies in Generative Grammar 17) du Cercle Linguistique de Prague 4), 160⫺163
Rubach, Jerzy & Booij, Geert (1990), “Edge-of- Vennemann, Theo (1972), “Rule Inversion”. Lin-
constituent Effects in Polish”. Natural Language gua 29, 209⫺242
and Linguistic Theory 8, 427⫺463 Wurzel, Wolfgang U. (1981), “Problems in Mor-
Schultink, Hendrik (1962), De Morfologische Va- phonology”. In: Dressler et al. (eds.), 413⫺434
lentie van het Ongelede Adjectief in Modern Neder-
lands. Den Haag: Van Goor Zonen Geert Booij, Amsterdam (The Netherlands)
1. Grammar versus the lexicon which dealt with the systematic means by
2. Lexicalist conceptions of the lexicon which new words were formed and added to
3. Where’s morphology? the lexicon; and syntax, which dealt with the
4. Blocking arrangement of words into phrases and sen-
5. Conclusion
6. References
tences (although compounding was some-
times included in syntax). In addition, both
inflection and word formation were often
1. Grammar versus the lexicon grouped together under the rubric of mor-
phology, inasmuch as they are both con-
Traditional grammarians divided a language cerned with bound formatives or affixes (as
into two major parts: grammar and lexicon. opposed to compounding and syntax). Tradi-
The lexicon contains the basic elements, tional grammar was thus word-based, as
while the grammar contains the rules for British scholars have emphasized (Robins
combining these basic elements and provid- 21979) and consisted very largely of morphol-
ing their proper forms in context. Both Jes- ogy, in practice and sometimes in theory.
persen (1965) and Chomsky (1965) approv- Saussure, for example, considered syntax to
ingly quote the same dictum from Sweet be part of parole and not part of langue, and
(1913: 13) on this dichotomy: “Grammar thus completely outside the purview of gram-
deals with the general facts of language, and mar. There was also little connection between
lexicology with the special facts.” From early morphology and the lexicon in this system,
on, the lexicon was thought of as simply an although it was acknowledged that word for-
unstructured list of words (the first lexicons mation was a mechanism for enlarging the
were lists of especially hard words). Gram- lexicon.
mar had three parts: inflection (or accidence), Sometime before 1881, Jan Baudouin de
which dealt with the systematic variations in Courtenay coined the term morpheme to refer
the forms that words took in context; word to the minimal meaningful forms of lan-
formation (derivation and compounding), guage. Baudouin defined the morpheme as
Martin, Jack (1988), “Subtractive Morphology as get (Instituttet for Sammenlignende Kultur-
Dissociation”. Proceedings of the West Coast Con- forskning B 85)
ference on Formal Linguistics 7, 229⫺240 Nida, Eugene A. (21949), Morphology. Ann Arbor:
Martin, Jack (1994), “Implications of Plural Redu- Univ. of Michigan Press [11946]
plication, Infixation and Subtraction for Musko-
gean Subgrouping”. Anthropological Linguistics Plank, Frans (1981), Morphologische (Ir-)regulari-
36, 27⫺55 täten: Aspekte der Wortstrukturtheorie. Tübingen:
Narr (Studien zur deutschen Grammatik 13)
Mel’čuk, I[gor’] A. (1967), “K ponjatiju slovoobra-
zovanija”. Izvestija Akademii Nauk SSSR, Serija li- Rainer, Franz (1993), Spanische Wortbildungslehre.
teratury i jazyka 26, 352⫺362 [German transl.: Tübingen: Niemeyer
“Zum Begriff der Wortbildung (Derivation)”. In: Scalise, Sergio (1994), Morfologia. Bologna: Mu-
Mel’čuk, I[gor’] A. (1976), Das Wort, Vol. I: Zwi- lino
schen Inhalt und Ausdruck. München: Fink (Inter-
nationale Bibliothek für allgemeine Linguistik 9), Spencer, Andrew (1991), Morphological Theory.
63⫺88] Oxford, Cambridge/MA: Blackwell
Mel’čuk, Igor A. (1991), “Subtraction in Natural Stonham, John T. (1994), Combinatorial Morphol-
Language”. In: Grochowski, Maciej & Weiss, Dan- ogy. Amsterdam, Philadelphia: Benjamins (Current
iel (eds.), “Words Are Physicians for an Ailing Issues in Linguistic Theory 120)
Mind”. For Andrzej Bogusławski on the Occasion of Truax, Catherine Elaine (1991), Transparency,
His 60th Birthday. München: Sagner (Sagners Analogy, and Phonological Change: Vowel Quantity
Slavistische Sammlung 17), 279⫺293 and Nominal Number in Selected German Dialects.
Morin, Yves-Charles (1981), “Où sont passés les s Ph.D. Dissertation, Stanford Univ.
finals de l’ancien français?”. In: Sankoff, David &
Cedergren, Henrietta (eds.), Variation Omnibus. Winter, Werner (1969), “Vocative and Imperative”.
Edmonton/Alberta: Linguistic Research (Current In: Puhvel, Jaan (ed.), Substance and Structure of
Inquiry into Language, Linguistics and Human Language. Berkeley, Los Angeles: Univ. of Cali-
Communication 40), 35⫺47 fornia Press, 205⫺223
Mosel, Ulrike & Hovdhaugen, Even (1992), Sa-
moan Reference Grammar. Oslo: Universitetsforla- Wolfgang U. Dressler, Vienna (Austria)
Tone may enter into a nominal paradigm, for neutral. Even where there is a systematic cor-
example marking case in Maasai (e.g. relation between grammatical meanings and
[c̀lk=´tìn] ‘ox(nom)’ vs. [c̀lk=´tín] ‘ox(abs)’) and stress/tone patterns, it may be attributable to
Turkana (Art. 138) or number and obviation the features of certain affixes (cf. Welmers
in some varieties of Montagnais (cf. Cowan 1973: 157f. on tone in ChiNyanja verbs) ⫺ or
1983; Art. 130). Similarly, tone may enter it may be secondary to segmental differences
into a verbal paradigm, as in the marking of (cf. Art. 64, 3.2 on affixal and accentual
tense/aspect in Gokana (e.g. [à dc̄?] ‘s/he subparadigms in Slavic languages). The
falls’ vs. [ā dc̄?] ‘s/he has fallen’) or person/ grammatical role of suprasegmental features
number in Chatino (cf. Pride 1963: 21). Be- must therefore be examined in the context of
sides its inflectional functions, tone may also the overall morphological system of the lan-
be the sole exponent of a derivational pro- guage in question.
cess, e.g. nominalization in Lendu ([]ò] ‘give’
vs. []ó] ‘gift’, [dhù] ‘insult [verb]’ vs. [dhú] ‘in-
sult [noun]’) or Classical Chinese (cf. Downer 2. The study of tonal morphology
1959: 267, 271⫺277). Compounding is an-
other example. In Mande languages, for in- Since tonal contrasts can perform the same
stance, non-initial elements of compounds functions as segmental contrasts, lexical tone
forfeit their underlying tones. The underlying has often been treated from a theoretical
tones are replaced by tones that characterize standpoint as being no different from any
tonally dependent morphemes, which, de- other segmental feature, i.e. by definition as-
pending on the language, are often spread sociated with a given segment or syllable. Yet
from the last tone of the nearest non-depen- there has also been some recognition that the
dent element in the construction, sometimes lexical and grammatical functions of tone
in combination with default tones (e.g. sometimes require that it be analytically di-
Mende [hàlè-wílì] ‘hospital’, from [hàlé] ‘me- vorced from segments. Examples of morpho-
dicine’ and [pílí] ‘house’, where the high tone logically relevant stress and tone, especially
on the derived form [wílì] is spread from in African and American Indian languages,
[hàlé] and the low tone of [wílì] is placed by were reported in the first half of the 20th cen-
default; cf. Leben 1978: 186⫺206). Tonal tury (cf. Sapir 1921: 78⫺81; Pike 1948: 22f.
morphology is particularly interesting because with further references). In the 1940s, Ameri-
it exhibits essentially the same range of mor- can structuralist works on morphological
phological properties as in all of segmental analysis explicitly mention suprasegmental
morphology. Even in tonal systems with a “morphemes”, including intonation contours
highly developed morphology, most morphs and stress or tone patterns of an additive
will consist of a pairing of segmental and character as well as stress or tone changes (cf.
tonal features. However, a morphological Hockett 1947: 337 [1957: 238]; Nida 21949:
property can be expressed solely through 62⫺65). The latter were regarded as a special
tone, i.e. by a tonal morph (often called tonal type of “replacive morph” (cf. Art. 58),
morpheme; cf. Art. 46). whereas additive elements were called super-
A suprasegmental contrast which coin- fix ⫺ despite the objection that they are nei-
cides with a difference in grammatical mean- ther “above” nor “below” the segments: “the
ing does not necessarily involve supraseg- term ‘superfix’ is no more apt than ‘subfix’,
mental morphs. Thus, the Italian minimal and (the bad Latin) ‘simulfix’ might be even
pair párlo vs. parló exhibits both a difference better” (Hockett 1954: fn. 4). The variant su-
in person (1st singular vs. 3rd singular) and prafix is also used (cf. Nida 21949: 69; Mat-
one in tense (present vs. past), neither of thews 1974: 133; Mel’čuk 1982: 87). Occa-
which is consistently signalled by stress; in sionally, these terms are applied to a “change
other inflection classes, 3rd singular past con- of tone or stress in a base, when this has the
trasts with 2nd singular present (dormı́ ‘s/he same effect as adding an affix” (Bauer
slept’ vs. dórmi ‘you sleep’) or 3rd singular 1988: 254), but this is not the prevalent us-
present (credé ‘s/he believed’ vs. créde ‘s/he age. Although the existence of suprasegmen-
believes’). This suggests that certain Italian tal morphs was frequently noted, they re-
suffixes (e.g. -o ‘3.sg.past’) are inherently ceived little attention in most varieties of
stressed or stress-attracting and others (e.g. American and European structuralism; well-
-o ‘1.sg.pres’) inherently unstressed or stress- developed attempts to understand their
unique status within phonological theory ceding tone results in contours that are not
were one of the hallmarks of John R. Firth’s otherwise attested in Cantonese.
“prosodic analysis” and of Zellig S. Harris’s The term floating tone or non-segmental
“phonemic long components” (cf. Firth 1949; tone has also been applied to an “abstract”
Harris 1951: 125⫺149). morphonemic entity which can trigger tonal
In the latter half of the twentieth century, changes:
work in African languages led to a more
“Non-segmental tones are tones which are posited
thorough understanding of tonal morphol- in the deep structure, but not attached to segmental
ogy. It was recognized that “certainly many phonemes. They might by late rules combine with
tone languages have some morphemes com- a preceding segmental tone or be deleted. In many
posed exclusively of pitch phonemes” (Wel- cases they have had an observable influence on the
mers 1959/60: 2; for examples, cf. Welmers surrounding tones.” (Voorhoeve 1971: 47)
1959/60: 6; 1969: 90f.; 1973: 126⫺128). In
For example, the Igbo associative construc-
Jukun, for instance, the hortative construc-
tion, which consists of a noun-noun se-
tion consists in the replacement of the normal
quence, was analyzed as containing a floating
pronoun tone by H: [kú bı̄] ‘he should come’
H tone between a noun and its possessive (cf.
(cf. [kū bı́] ‘he came’); [ḿ bı̄] ‘I should come’
Voorhoeve & Meeussen & DeBlois 1969).
(cf. [m̀ bı́] ‘I came’). Along with replacive
This floating H provided the conditioning for
tonal morphs, there are others that simply
certain changes, for example the change of a
function as prefixes and suffixes. Some are
L-L noun to L-H when the following noun
purely tonal allomorphs of segmental affixes
also begins with L. Thus, [àgbá ènwè] ‘jaw of
(cf. Welmers 1959/60: 9), but others have no
monkey’ was analyzed as underlying /àgbà+
segmental counterparts. In Urhobo, for in-
ènwè/ (cf. [àgbà] ‘jaw’).
stance, the negative is expressed by adding a
In every detail, a tonal morph is parallel
“floating” tone (cf. Welmers 1969: 90f.): Af-
to its segmental counterpart: It can occur as
ter a final L, the negative is a floating H, as
an affix or as a clitic, depending on whether
illustrated by (1a) and (1b), which differ only
it comes in at the word or phrase level (cf.
in the absence vs. presence of a floating H at Art. 41). In addition, a floating tone may
the end; after a final H, the negative is a either be added outside the tone of its base
floating LH, as in (2b) vs. (2a): or host, or it may replace that tone. A tonal
(1) (a) [c̀ tá kì bè] ‘he told me’ morph may be the sole exponent of a mor-
(b) [c̀ tá kì bè+] ‘he didn’t tell me’ phological expression, or it may cooccur with
another morph, as when a tense is defined by
(2) (a) [mè ta kí] ‘I told him’ a segmental prefix and a tonal suffix. In
(b) [mè ta kí ˇ ] ‘I didn’t tell him’ either case the same analytical problems arise
Other examples which were reported in the as with segmental morphs. Thus, in each case
1960s include the high-toned “zero” subject one must ask whether a tonal morph is a
concord prefix in Akan, which consists solely “thing” (i.e. present in an actual string of
of a tone (cf. Schachter & Fromkin 1968: 113, segments and/or tones) or a “process” (i.e.
where it is regarded as [⫹segment], with no the observable change that takes place). One
phonological features other than tonal ones), must also ask whether the tone spells out a
and the “known” or “specific” (definite) de- morphological or morphosyntactic feature or
terminer of Bambara, which is a floating L, is simply inserted in the presence of such a
recognizable by specific allotones of a preced- feature, and so forth (for a discussion of
ing L or a following H (cf. Bird 1966: 6; these possibilities in the context of morpho-
1968: 37f.). It was not only in African lan- logical theory, cf. Anderson 1992).
guages that such phenomena were being ob-
served. Cantonese, for instance, has a high- 3. Word-level tonal morphology
pitched suffix which adds a high-toned tail to
a preceding non-high tone as an expression Just as two morphs may differ solely in the
of endearment, familiarity etc. (cf. Whitaker value of a single segmental feature (e.g. the
1956: 193, 202f.). The diachronic origin of feature [⫾voice] in English minimal pairs
this floating tone appears to be the high- such as pear/bear), so can two lexical morphs
toned diminutive suffix -i; its status as a syn- differ only in a single feature of tone (e.g.
chronic suffix can be defended on the Lahu [tsà] ‘fierce’, [tsā] ‘feed’; Cemuhi [tı́i] ‘de-
grounds that the combination with the pre- stroy’, [tı̄i] ‘harvest’, [tı̀i] ‘write’). Similarly,
grammatical morphs may differ only in tone, nouns is marked by a H tone prefix (joining
e.g. Aghem [à] ‘with’ vs. [â] ‘to/for’; Haya the lexical L to create a surface ML falling
[ká] ‘if’, [kà] (‘distant past’). Also common tone):
is when a paradigm is characterized by tonal (5) singular plural
differences on several morphs. A case of such [Màm] [Mām̀] ‘animal’
multiple exponence (cf. Art. 64) is observed in [dzòm] [dzōm̀] ‘antelope’
Gokana verb forms such as [ā è dc̀?] ‘s/he [ndzèè] [ndzēè] ‘sheep’
fell’ vs. [à é dc̄?] ‘s/he will fall’. The past tense
marker [è] carries a L tone, while the future Tonal suffixes are frequently attested. In
tense marker [é] carries a H tone. However, Bamileke-Dschang, infinitives take the noun
there is also a difference in tone on the third class prefix /lè-/ and a L tonal suffix:
person singular subject marker ([ā] vs. [à]) as (6) (a) /lè-tòn-7/ ‘reimburse’
well as on the verb ‘fall’ ([dc̀?] vs. [dc̄?]). In (b) /lè-tón-7/ J [lè-!tón] ‘call’
other cases of multiple exponence, a segmen-
Frequently the same point in a paradigm may
tal affix may be accompanied by a tonal mel-
require both a tonal prefix and a tonal suffix.
ody. Thus, in Hausa, plurals are marked by
For instance, in Noni the imperative is
a segmental suffix and an accompanying tone marked by a L prefix and a H suffix:
pattern which is assigned to the resulting
stem ⫹ suffix combination. For example, (7) (a) /7-dc̀msè-+/ J [dc̀msě] ‘push!’
plurals in [⫺únàa] are formed by replacing (b) /7-tsı́msé-+/ J [tsı̄msé] ‘pacify!’
the final vowel of the base with this suffix Where they cooccur in African languages, the
and substituting an all H tone pattern over tonal suffixes generally tend to be more
the stem: tightly bound to their base than the tonal
prefixes. If introduced at different levels or
(3) (a) [kèeké] ‘bicycle’ [kéekúnàa] ‘bicycles’ strata, the analysis of such cases as tonal “cir-
(b) [àgóogó] ‘clock’ [ágóogúnàa] ‘clocks’ cumfixes” is less likely.
A similar, though more restricted, case oc- There are notorious methodological prob-
lems as well as gaps in the analysis of tones
curs in Haya. In this typical Bantu language,
as affixes. First, there is apparently no known
surface tones are derived by applying phono-
case of a tonal infix, say a L tone, which in-
logical rules to the underlying tones of con- terrupts the lexical tones of the base to which
catenated morphemes. The one exception oc- it is attached. There also are no clear cases of
curs when a Hn-L-H melody replaces the un- tonal transfixes. There are, however, sporadic
derlying tones on an imperative verb that has reports of tonal metathesis in the literature.
a first person singular object prefix, e.g. [ń- The Bamileke-Dschang infinitive [lè-!tón] in
kòmá] ‘tie me’, [ń-lı́mı̀rá] ‘cultivate for me’ (6b) has been analyzed as /lè-tón-7/ with the
(underlying /n-kóma/, /ń-limira/). Such in- infinitival suffix L metathesizing around the
stances of replacive tonology are exactly par- preceding H on the verb root [-tón] ‘call’.
allel to cases where an inherent segmental This causes the downstep on the H tone
feature or syllable structure is overwritten by marked by the raised exclamation point. In
a “prosodic template” based on the morphol- Bamileke-Ghomálá?, where such sequences
ogy. of linked H followed by unlinked L result in
In “concatenative” tonal morphology, LH rising contours, e.g. /tóm7/ J [tǒm] ‘fruit’
tonal morphs frequently turn up both as pre- (cf. [tóm] ‘[to] sprout’), the metathesis analy-
fixes and as suffixes. In Noni, a L tone prefix sis appears even more compelling (cf. Nis-
marks the singular of class 9/10 nouns, form- sim 1981: 288f.).
ing a LH rising tone when the noun stem is In other cases there is a problem of seg-
H: mentability. It is clear that a tonal morph is
present, but it is not possible to sequence the
(4) singular plural tones of a base and its affix. One such case
[bwě] [bwé] ‘dog’ comes from Mpi, where nouns and verbs be-
[dzčn] [dzćn] ‘star’ long to three non-intersecting tone classes. In
[jı̌n] [jı́n] ‘maggot’ each case the verb appears to have an addi-
tional contour that the noun lacks: [sı́] ‘four’,
On the other hand, nouns which are underly- [sı̄] a color, [sı̀] ‘blood’ vs. [sı̂] ‘die’, [sı̄7] ‘roll’,
ingly L show that the plural of class 9/10 [sı̌] ‘be putrid’. One would like to segment out
part of the tone as a tonal suffix on verbs, dialect of Mixtec whereby adjectives are de-
a morpheme marking verbhood perhaps, but rived from nouns: [bı́kó] ‘cloudy’ (cf. [bı̄kò]
there is no way to generalize across the ‘cloud’), [só?ó] ‘deaf’ (cf. [sò?ò] ‘hear’), [sı́nı́]
tonal contours. ‘pertaining to the head’ (cf. [sı̄nı̀] ‘head’). The
Other potential problems of tonal mor- derived adjective has all H tone, though it is
phology are reminiscent of those involving difficult to say whether this H comes in from
strictly non-tonal features. In many Tibeto- the left or from the right.
Burman languages ⫺ though usually not pro- At the lexical level, then, tone appears to
ductively ⫺ a tonal change can mark the dif- be available for every kind of morphological
ference between plain and causativized verbs. expression (with the possible exceptions of
In Tibetan the initial part of a tonal contour infixation and transfixation). A final illustra-
may be raised to form a causative: tion at the lexical level is the role of tone in
reduplication. In some languages, a morpho-
(8) [Mi14] ‘sleep’ [Mi44] ‘make sleep’ logical process may require that both tar-
[phap13] ‘descend’ [pap53] ‘make descend’ geted segments and their tones be redupli-
[sum15] ‘be firm’ [tsum54]‘make firm’ cated, e.g. Kinande /omu-rúme/ ‘man’, which
is realized [òmú-rùmè] by anticipating the H
While it would appear that the causative onto the preceding vowel, reduplicates as
morpheme is a tonal prefix, as part of the /omu-rúme-rúme/, realized [òmú-rùmé-rùmè].
process voiceless aspirates deaspirate and It is also possible for segments to be redupli-
fricatives affricate. The possibility thus arises cated without their underlying tones. In
that this tonal morph is instead a segmental Haya, for instance, the infinitive [oku-kı́n-
morph with an accompanying (non-low) gula] ‘to open’ reduplicates its stem as [oku-
tone. Or could this be a case of two separate kı́ngula-kingula] ‘to open here and there’, not
prefixes: a segmental one and a tonal one? as *[oku-kı́ngula-kı́ngula]. An additional
This latter kind of analysis suggests itself in possibility is reflected in the reduplicative
Slave, where alienable possession is marked process that derives adjectives from nouns in
by a H tone suffix [-é], while inalienable pos- the Changzhi dialect of Chinese: suan213-ti213
session is marked only by a H tonal morph: ‘sour’, xuan24-ti24 ‘yellow’, yan535-ti535 ‘soft’.
(9) (a) [dzı́je-tu-é] ‘water from berries’ In each case the first morpheme is the base
(b) [dzı́je-tú] ‘wine’ noun; the suffix -ti reduplicates (or spreads?)
the tone of this noun.
In these examples it is possible to identify the
H suffixal tone as ‘possessive’ and the [-é] 4. Phrasal tonal morphology
segmental suffix as ‘alienable’ (or perhaps as
‘genitive’) such that the underlying represen- A completely different kind of problem con-
tations of (9a) and (9b) become /dzı́je-tu-e-+/ cerns the basic issue of when a tonal morph
and /dzı́je-tu-+/, respectively. is an affix, as opposed to a clitic. One way to
These examples show that there is no a pri- look at this distinction is to regard an affix
ori reason why a morph should have to be as a grammatically bound morpheme intro-
strictly segmental or strictly tonal, rather duced in the morphology, while a clitic is in-
than combining tonal and non-tonal features troduced syntactically, i.e. at the phrase level.
(e.g. aspiration, nasalization etc.). In some One of the most often cited cases of a tonal
such cases it may be desirable to set up a seg- morph, the associative H tone of Igbo, falls
mental tone-bearing unit to which the tone is into this latter category. In cases where the
linked. In other cases the tonal and non-tonal possessed noun ends in L and the possessor
features may be semi-autonomous expres- noun begins with L, the H associative marker
sions of the same morpheme ⫺ or, upon will be assigned to the left: [àgbá ènwè] ‘jaw
closer examination, may turn out to be ex- of monkey’ (/àgbà ⫹ + ⫹ ènwè/). In all other
pressions of different morphemes. An addi- cases the H will be assigned to the right caus-
tional difficulty that does not frequently arise ing, for instance, the downstep in the se-
in segmental morphology may be in deter- quence [àgbà é!wú] ‘jaw of goat’ (/àgbà ⫹ +
mining the directionality of the tonal affixa- ⫹ éwú/; while a number of different accounts
tion. This is particularly in cases where a have been offered to derive this downstep,
morphological process completely replaces they all have in common that the H is as-
the underlying tones of the base. This is true signed to the right). Since the effect of the H
in the process in the San Miguel El Grande tonal morph can be felt on either the preced-
ing or the following word, a syntactic ac- In other words, where in some languages
count seems to be required: the associative H a tonal clitic adds a tone to a string, in other
will cliticize sometimes to the left, sometimes languages, tones are deleted.
to the right, depending on the surrounding In many languages, boundary tones are
tonal context. used to mark the edges of phrasal domains,
Virtually any kind of grammatical morph e.g. phonological phrase, intonational phrase,
can consist solely of a tone introduced syn- utterance. In Kinande, a lexically toneless
tactically: pronouns, determiners, tense-as- noun such as [omu-r=mi] ‘farmer’ is realized
pect auxiliaries, prepositions, conjunctions, with four (default) L tone syllables in envi-
and so forth. In Bambara, for instance, defi- ronments where no tonal morphs interact
niteness is indicated by a L tonal morph (cf. with it. When interacting with final boundary
2); in Gwari, serial verbs are marked by a tones, however, ‘farmer’ may surface with
preceding L that converts a H to a LH rising one of three additional tone patterns:
tone and a M to a lowered M, e.g. ([]é]
(11) (a) [òmù-rìmí] (— H%)
‘come’ / []ě] ‘and come’; [lō] ‘go’ / [!lō] ‘and
(b) [òmù-rímì] (— H% L//)
go’). In Gokana both locative and temporal
(c) [òmù-rímí] (— H% H//)
adjuncts are marked by a preceding H tone,
e.g. Here, a final H% boundary tone (the symbol
% indicates a boundary) marks a phonologi-
(10) (a) [āè mc̀n nc̄m] ‘he saw an animal’
cal phrase that ends in a lexically toneless
(b) [āè mc̀n nc̄ḿ tc̄] ‘he saw an animal in
vowel, e.g. a subject noun phrase as in (11a).
the house’
At the end of an intonational phrase, L//
(c) [āè mc̀n nc̄ḿ nı̄?ēı́] ‘he saw an animal
marks statements (11b) and H// marks ques-
today’
tions and lists (11c). While marking phrase
As (10) shows, tone may mark a variety of boundaries and intonations, these tonal
relationships within syntactic constituents. In morphs differ from their affixal and clitic
addition to Igbo, many other West African counterparts simply by being introduced late
languages use a H or L tone clitic to mark in the phrasal phonology.
possession. In other languages which lack a Tonal morphs of this type convey common
separate tonal clitic per se, tonal differences intonational meanings associated with de-
may still be used to subordinate one element clarative, interrogative, negative, imperative
to another phonologically. For example, in and vocative utterances. This obviously ap-
Shanghai Chinese compounds as well as se- plies to all languages that have intonations,
quences of a lexical item and following func- i.e. probably to all languages. Kinande pres-
tion word(s), only the initial element retains ents an example where boundary tones are
its phonological tones, which are mapped added to a lexically toneless form. In other
over the entire sequence, with any excess syl- cases the phrasal and intonational tones
lables receiving a default L tone. modify ⫺ and potentially neutralize ⫺ lexical
HL MH HL H L
‘sky’ ‘gas’ ‘weather’
LH MH LM LH LH L H L L
‘scold-exper-you int’ ‘[Has someone] scolded you?’
Fig. 61.1: Shanghai Chinese compounds and verb forms
tones. The most commonly encountered Bauer, Laurie (1988), Introducing Linguistic Mor-
cases involve using a tonal means to distin- phology. Edinburgh: Edinburgh Univ. Press
guish interrogatives from declaratives. In Bendor-Samuel, John T. (1960), “Some Problems
Hausa, a L is added after the rightmost lexi- of Segmentation in the Phonological Analysis of
cal H in a yes/no question, fusing with any Tereno”. Word 16, 348⫺355 [reprinted in: Palmer
pre-existing lexical L that may have followed (1970, ed.), 214⫺221]
the rightmost H (which is raised somewhat, Bird, Charles S. (1966), “Determination in Bam-
as are any following L tones whatever their bara”. The Journal of West African Languages 3.1,
source). As a result, lexical tonal contrasts 5⫺11
are neutralized. In statements, [káı̀] ‘head’ is Bird, Charles S. (1968), “Relative Clauses in Bam-
tonally distinct from [káı́] ‘you [masculine]’. bara”. The Journal of West African Languages 5,
But at the end of a yes/no question, they are 35⫺47
identical, consisting of an extra-H gliding Cowan, William (1983), “The Development of
down to a raised L. In Nembe, a final lexical Suprasegmental Inflections in Montagnais”. In-
L becomes H in statements, and a final lexi- ternational Journal of American Linguistics 49,
cal H becomes L in questions. Thus, L-L / L- 64⫺71
H contrasts such as [dı̀rı̀] ‘book’ / [bùrú] Downer, G. B. (1959), “Derivation by Tone-change
‘yam’ are neutralized as L-H in statements, in Classical Chinese”. Bulletin of the School of
but as L-L in questions. A similar case is Oriental and African Studies 22, 258⫺290
found in Isoko, where a final L marks posi- Firth, John R. (1949), “Sounds and Prosodies”.
tive questions, while a final H marks negative Transactions of the Philological Society 1948, 127⫺
questions. This causes a final lexical L to re- 152 [reprinted in: Palmer (1970, ed.), 1⫺26]
main L in a positively expressed question, Fromkin, Victoria A. (1978, ed.), Tone: A Linguis-
while this final L becomes a LH rise in a neg- tic Survey. New York etc.: Academic Press
atively expressed question: [ùbı̀] ‘book’ / [ùbı̌] Goldsmith, John A. (1995, ed.), The Handbook of
‘book? [negative]’. Phonological Theory. Oxford, Cambridge/MA:
In other languages tonal differences may Blackwell (Blackwell Handbooks in Linguistics 1)
mark emphasis or attenuations. In Hausa
Harris, Zellig S. (1951), Methods in Structural Lin-
raising the first H (or H sequence) in a word guistics. Chicago: Univ. of Chicago Press [later
to extra H signals emphasis. The same ges- printings entitled Structural Linguistics]
ture accompanied H ideophones, e.g. the ex-
Hockett, Charles F. (1947), “Problems of Morphe-
tra-high tone intensifiers such as [fa̋t] and mic Analysis”. Language 23, 321⫺343 [reprinted
[k’ı̋rı̋n] ini [fárı́ı́ fa̋t] ‘snow white’ and [bák’ı́ı́ in: Joos (1957, ed.), 229⫺242]
k’ı̋rı̋n] ‘jet black’. In Eastern Popoloc, a final
Hockett, Charles F. (1954), “Two Models of Gram-
upglide replaces the normal glottal stop to matical Description”. Word 10, 210⫺231 [reprinted
mark politeness, while imperatives in Ki- in: Joos (1957, ed.), 386⫺399]
nande receive a final L boundary tone (and
Joos, Martin (1957, ed.), Readings in Linguistics:
lack the phonological phrase-final H% tone).
The Development of Descriptive Linguistics in
while some distance from word-level tonal America since 1925. New York: American Council
morphology, these intonational boundary of Learned Societies [⫽ 41966 Readings in Linguis-
tones often have a complex grammar and tics, Vol. I. Chicago, London: Univ. of Chicago
phonology of their own and are not directly Press]
derivable from putative intonational univer- Leben, William R. (1978), “The Representation of
sals. Thus, contrary to the expectations of Tone”. In: Fromkin (ed.), 177⫺219
some linguists, there are cases where final H Matthews, P[eter] H. (1974), Morphology. Cam-
marks declaratives (Nembe, Luganda) and bridge etc.: Cambridge Univ. Press
cases where a final L marks interrogatives
Mel’čuk, I[gor’] A. (1982), Towards a Language of
(Dagbani, Isoko).
Linguistics. München: Fink (Internationale Biblio-
thek für allgemeine Linguistik 44)
Palmer, F[rank] R. (1970, ed.), Prosodic Analysis. tive Phrase”. The Journal of West African Lan-
London: Oxford Univ. Press (Language and Lan- guages 6, 79⫺84
guage Learning 25) Welmers, William E. (1959/60), “Tonemics, Mor-
Pike, Kenneth L. (1948), Tone Languages. Ann photonemics, and Tonal Morphemes”. General
Arbor: Univ. of Michigan Press Linguistics 4, 1⫺9
Pride, Leslie (1963), “Chatino Tonal Structure”. Welmers, William E. (1969), “Structural Notes on
Anthropological Linguistics 5.2, 19⫺28 Urhobo”. The Journal of West African Languages
Sapir, Edward (1921), Language: An Introduction 6, 85⫺107
to the Study of Speech. New York: Harcourt, Brace Welmers, W[illia]m E. (1973), African Language
Structures. Berkeley etc.: Univ. of California Press
Schachter, Paul & Fromkin, Victoria A. (1968), A
Phonology of Akan: Akuapem, Asante, and Fante. Whitaker, K[atherine] P. K. (1956), “A Study on
Los Angeles, Dept. of Linguistics, Univ. of Cali- the Modified Tones in Spoken Cantonese II: The
fornia at Los Angeles (UCLA Working Papers in Origin and Functions of the Modified Tones”. Asia
Phonetics 9) Major N.S. 5, 184⫺207
Voorhoeve, Jan (1971), “Tonology of the Bamileke Yip, Moira (1995), “Tone in East Asian Lan-
Noun”. Journal of African Languages 10, 44⫺53 guages”. In: Goldsmith (ed.), 476⫺494
Voorhoeve, Jan & Meeussen, A[chiel] E. & Blois,
K[ornelis] F. de (1969), “New Proposals for the De- Larry M. Hyman, Berkeley/CA (U.S.A.)
scription of Tone Sequences in the Igbo Comple- William R. Leben, Stanford/CA (U.S.A.)
1. Zur Rolle von Wörtern im Satz schen Funktion (2.1) und der syntaktischen
2. Syntaktische Funktionen Kategorie (3.1) verhält, und wie sich die be-
3. Syntaktische Kategorien treffenden Begriffe für die Beschreibung der
4. Zitierte Literatur
Rollen der Wörter im Aufbau des Satzes eig-
nen. Anschließend wird kurz und beispielhaft
auf einige einschlägige Theorien und Termi-
1. Zur Rolle von Wörtern im Satz nologien (2.2; 3.2) eingegangen, die methodo-
logischen Grundlagen dargestellt (2.3; 3.3)
Wörter sind die kleinsten Bausteine des Sat-
und schließlich die Entstehung und Entwick-
zes. (Unter Satz wird hier ein inhaltlich kohä-
lung der entsprechenden Unterscheidungen
renter Komplex von Wörtern verstanden, der
ein finites Verb enthält; zum Problem der und Begriffe nachgezeichnet (2.4; 3.4).
Satzdefinition vgl. z. B. Ries 1931; Müller
1985; Forsgren 1992 b). Im Satz Der Hund
biß den Briefträger beziehen sich die Wörter 2. Syntaktische Funktionen
Hund und Briefträger offensichtlich verschie-
2.1. Wortart und syntaktische Funktion
denartig auf das Wort biß (der Hund beißt,
der Briefträger wird gebissen), d. h. sie erfül- Es ist eine allgemein bekannte Tatsache, daß
len verschiedene syntaktische Funktionen ⫺ sich der Begriff Wort schwer definieren läßt
was auch die Formen der bzw. den signalisie- (vgl. Art. 26), weil die sprachlichen Einheiten,
ren. Die Wörter Hund und Briefträger gehö- die sich unter diesem Begriff zusammenfassen
ren aber zu derselben Wortart: Substantiv. lassen, fundamentale Unterschiede aufweisen.
Daraus wird ersichtlich, daß sich die Unter- Das betrifft einerseits den inneren Bau oder
schiedlichkeit dieser syntaktischen Funktio- die Morphologie der Wörter. Mit den Wörtern
nen mit Hilfe des Wortartbegriffs Substantiv Hund und bellen sind z. B. verschiedene Fle-
nicht eindeutig beschreiben läßt. xionsmuster (Paradigmen) verbunden (Hund/-
Im Satz Der kleine Hund bellt bezieht sich es/-e; bell-e/-st/-t etc.) Diese Wörter sind flek-
das Adjektiv klein direkt auf das Substantiv tierbar, andere aber wie heute und nach (Parti-
Hund und bildet mit ihm eine sprachliche keln) sind nicht flektierbar. Andererseits sind
Einheit (Konstituente), eine Phrase, deren auch die syntaktischen Eigenschaften der
Ganzes sich zu bellt verhält wie die Einheit Wörter verschiedenartig, weil diese sich nach
der Hund allein. Die Phrase ist eine syntakti- bestimmten Mustern oder Regeln mit anderen
sche Kategorie, die eine syntaktische Funk- Wörtern kombinieren lassen: Der/Ein Mann/
tion erfüllt. Hund/Fuchs kommt/schläft. Wörter mit ähn-
Es kommt nicht selten vor, daß die Be- lichen Flexionsparadigmen wie Hund können
griffe Wort und Wortart ungenügend von sol- demnach dieselbe Funktion im Satz erfüllen,
chen getrennt werden, die sich (a) auf die und an die Stelle von kommen lassen sich nur
Rollen (Funktionen) der Wörter im Satz be- Wörter mit entsprechenden morphologischen
ziehen, (b) auf die sprachlichen Einheiten Paradigmen einsetzen, wie z. B. schlafen.
(Konstituenten, Kategorien) beziehen, die Die Wörter haben somit unterschiedliche Fü-
diese Rollen spielen. gungspotenz oder Valenz (vgl. Stepanowa &
In diesem Artikel soll deswegen ausführ- Helbig 21981: 118⫺124; Admoni 41982: 216).
licher erklärt und erörtert werden, wie sich Der Begriff Fügungspotenz oder Valenz bezieht
der Begriff Wortart zu denen der syntakti- sich demnach auf die “nach außen” gerichte-
ten syntaktischen Eigenschaften des Wortes, Umgekehrt kann aber auch eine bestimmte
syntaktische Funktion hingegen auf den inne- Funktion durch verschiedene Wortarten er-
ren syntaktischen Aufbau des Satzes oder der füllt werden:
Phrase durch Wörter.
Um eine vollständigere Sprachbeschrei- (5) Das Problem / Es / Zu rechnen ist schwierig
bung zu ermöglichen, sind deshalb weitere In (5) erfüllt ein Substantiv und ein Artikel
Differenzierungen und eine Klassifikation (Problem und das) dieselbe syntaktische
der Wörter notwendig. Zu diesem Zweck hat Funktion wie ein Verb mit Partikel (rechnen
sich im Laufe der Zeiten das System der und zu) und wie ein Pronomen (es). Eine syn-
Wortarten (s. Art. 70⫺72) herausgebildet. taktische Funktion kann durch Einheiten
Ihre Anzahl kann auf Grund verschiedenarti- verschiedener Komplexität vertreten werden,
ger Abgrenzungs- und Organisationskriterien und zwar durch ein Wort, eine Phrase, oder
von Grammatik zu Grammatik stark variie- sogar einen ganzen Satz:
ren, beläuft sich aber häufig auf neun bis zehn
Klassen: Substantiv, Verb, Adjektiv, Adverb, (6) Die Nachricht / Der schöne Anblick / Daß
Numerale, Pronomen, Präposition, Konjunk- du heute kommst, ist erfreulich
tion, Interjektion. Zum System der Wortarten (7) Daß du heute / in zwei Wochen kommst,
gehört weiterhin ein Inventar von gramma- ist erfreulich / eine erfreuliche Nachricht
tischen Subkategorien: Kasus, Tempus,
Eine solche syntaktische oder funktionale
Komparation etc. (Die Artikel können entwe-
Synonymie der Wörter, Wortformen oder
der als eine selbständige Klasse oder als eine
Phrasen ist kennzeichnend für alle syntakti-
Subkategorie des Substantivs betrachtet wer-
schen Funktionen (vgl. Forsgren 1973: 44;
den.)
Gulyga 1978: 62) außer dem finiten Verb als
Daß die Wortarten für die Beschreibung Prädikat ⫺ z. B. biß in (1) ⫺ das nicht durch
syntaktischer Funktionen nicht ausreichen eine andere Form ersetzt werden kann. Auch
(vgl. 1), erklärt sich teils daraus, daß sich die die funktionale Synonymie begrenzt die Ver-
Gliederung auf das Vorkommen oder die wendbarkeit der Wortartenbegriffe für die
Distribution (vgl. Harris 1951: 146) der Wör- Beschreibung der syntaktischen Funktionen.
ter in allen möglichen Kontexten gründet, und Es ist somit nicht möglich, eindeutig etwa
teils daraus, daß eine Wortart in einem gegebe- von “dem Substantiv des Satzes” oder “dem
nen Satz mehrere Funktionen erfüllen kann: Adjektiv des Satzes” zu sprechen, und es
(1) Der Hund des Nachbarn biß den neuen kann ebensowenig behauptet werden, daß
Briefträger. eine bestimmte syntaktische Funktion von ei-
(2) Dieser Hund hat früher niemand gebissen. ner einzigen Wortart erfüllt wird. Zu diesem
(3) Der gebissene Mann wurde krank. Zweck steht indessen ein anderes Inventar
(4) Er wurde ins Krankenhaus gebracht. von Begriffen zur Verfügung, nämlich: Sub-
jekt, Prädikat, Objekt, Adverbiale, Prädikativ,
In (1) bezieht sich das Adjektiv (neu) auf ein Attribut. In Beispiel (6) betrifft die funktio-
Substantiv (Briefträger), dasselbe syntakti- nale Synonymie die Subjektfunktion, in Bei-
sche Verhältnis hat aber auch das Substantiv spiel (7) die Adverbial- bzw. die Prädikativ-
des Nachbarn zum Substantiv der Hund. In funktion. Die Funktionsbegriffe beziehen
(1) und (3) sind die Funktionen der Adjektive sich im Unterschied zu den Wortarten ein-
verschieden. Das Adjektiv neu bezieht sich di- deutiger auf die syntaktischen Funktionen.
rekt auf ein Substantiv, das Adjektiv krank Es ist durchaus plausibel, vom “Subjekt des
aber erst über die Vermittlung eines Verbs Satzes”, “Prädikat des Satzes” oder “Objekt
(wurde). Das Verb beißen erfüllt die gleichen des Satzes” zu sprechen. Dagegen ist es nicht
Funktionen in (1) und (2), nicht aber in (3), adäquat, den Ausdruck “Attribut des Satzes”
wo es dieselbe Funktion hat wie das Adjektiv zu verwenden. Attribute wie des Nachbarn,
neu in (1). Die Pronomina erfüllen verschie- neuen (4), dieser (5), gebissene (6) vertreten
dene Funktionen in (2) (dieser, niemand) und nämlich keine Satzfunktionen, sondern Funk-
(4) (er). Mit jeder Funktion verbindet sich in tionen innerhalb der Phrasen (vgl. 1 sowie 2.2,
der Regel eine bestimmte Flexionsform im 2.4, 3.1).
morphologischen Paradigma der flektierba- Die Funktionsbegriffe sind als reine Rela-
ren Wortarten, z. B. der (Hund), des (Nach- tionsbegriffe zu betrachten. Der Begriff Sub-
barn), gebissen-e etc. jekt bezeichnet beispielsweise nur eine Ein-
heit, die ein besonderes Verhältnis zum fini- tants (Aktanten), circonstants (Angaben) (vgl.
ten Verb hat, zu dem wiederum ein Objekt in z. B. Tesnière 1959: 102⫺105; Hays
einer anderen Relation steht usw. 1964: 513; Helbig 1982: 40f.). Das Verhältnis
Das finite Verb als Prädikat ⫺ z. B. biß zu den Wortarten entspricht weitgehend dem
in (1) ⫺ ist die einzige syntaktische Funk- der syntaktischen Funktionen, wie es in 2.1
tion ohne funktionale Synonymie. Es gibt dargelegt wurde. Nur ein finites Verb kann
zwei Definitionen des Begriffs Prädikat: eine, keine Leerstelle sein, da es als Satzkern die
durch die das Prädikat mit dem finiten Verb letzteren um sich eröffnet.
gleichgestellt wird, und eine, durch die es alle Argument und Prädikat sind ursprünglich
Wörter umfaßt, außer denen, die das Sub- formallogische Termini (vgl. Palmer 1994:
jekt bilden ⫺ d. h. auch Objekte, Adverbiale 2⫺4), die sowohl in der Dependenzgramma-
und vor allem Prädikative. Dadurch erhalten tik als auch in der generativen Transforma-
die letzeren Funktionen einen anderen Sta- tionsgrammatik (Chomsky 71993: 27, 35f.)
tus im Satz. Diese Schwankung ist historisch verwendet werden. Argument ist weitgehend
bedingt (s. 2.4), findet sich indessen auf synonym mit Leerstelle. Dagegen umfaßt
Grund unterschiedlicher theoretischer Posi- der Begriff Prädikat in diesem theoretischen
tionen auch in der heutigen Sprachwissen- Kontext alles, was eine Aussagefunktion er-
schaft wieder (s. 2.2). füllt, d. h. nicht nur das finite Verb, sondern
beispielsweise auch Prädikative. Im Satz Karl
2.2. Theorien und Terminologien ist groß ist Karl ein Argument, ist groß ein
Terminologien wechselnder theoretischer Pro- (einstelliges) Prädikat. In Karl ist größer als
venienz ⫺ und deshalb mit mehr oder minder Peter sind Karl und Peter Argumente und ist
abweichenden begrifflichen Abgrenzungen ⫺ größer ein (zweistelliges) Prädikat.
werden mit den syntaktischen Funktionen Die syntaktischen Funktionen können
verbunden. (Von der synonymen sogenann- auch terminologisch unbestimmt bleiben,
ten Verdeutschungsterminologie, z. B. “Satz- weil sie sich auf andere Weise darstellen las-
gegenstand” für das Subjekt, “Aussage” für sen. In der Phrasenstruktur-Grammatik und
das Prädikat, wird hier abgesehen.) Es ist der generativen Transformationsgrammatik
nicht das Ziel dieses Abschnitts, eine erschöp- in der Version der Aspects (vgl. Bierwisch
fende Übersicht über die betreffenden Termi- 1966: 38f.; Chomsky 1965: 63⫺74) werden
nologien und ihre theoretische Verankerung die Funktionen ebenso wie die Hierarchie der
zu geben; zur weitergehenden Information syntaktischen Funktionen bis zu den einzel-
wird auf einschlägige Literatur hingewiesen. nen Wörtern mit Hilfe von Einklammerun-
Die herkömmlichen Termini für die syn- gen und Baumdiagrammen veranschaulicht
taktischen Funktionen (Subjekt etc.) wurden (vgl. auch 3.1; 3.2).
zuerst in der Schulgrammatik (vgl. Art. 11) In Aktiv- und Passivsätzen wie Der Hund
verwendet, weil sich die Wortartenbegriffe als biß den Briefträger und Der Briefträger wurde
unzureichend erwiesen (s. 2.4). Diese Termi- vom Hund gebissen erfüllen Hund und Brief-
nologie wird häufig in der modernen Sprach- träger jeweils verschiedene syntaktische Funk-
wissenschaft benutzt, wobei die Benennung tionen, obwohl ihre semantischen Rollen in
(finites) Verb für das Prädikat oft vor- beiden Sätzen dieselben sind: Hund ist Agens
kommt. Eine Wortartenbezeichnung anstelle (agentive), Briefträger Patiens (objective,
eines Begriffs für eine syntaktischen Funktion goal, theme). Benennungen für solche Rollen
zu verwenden, ist nur dank der fehlenden sind (semantische) Kasus (Tiefenkasustheorie,
funktionalen Synonymie eben dieser Funk- vgl. z. B. Fillmore 1968; 1977; Stepanowa &
tion möglich (vgl. 2.1). Helbig 21981: 135⫺137; Jackendoff 1990: 50;
Leerstelle ist ein Terminus, der für die Fü- Palmer 1994: 4⫺11; Art. 102) oder Theta-
gungspotenz (Valenz) der Wörter verwendet Rollen (Theta-Theorie, vgl. Pleines 1978: 358;
wird, entspricht aber auch, insofern er mit Chomsky 71993: 34⫺48). Theta-Rollen wer-
dem finiten Verb verbunden ist, den Satz- den grundsätzlich Einzelwörtern zugeschrie-
funktionen (vgl. Bühler 1934: 173). Im letzt- ben. In Frage kommen natürlich nur Autose-
genannten Sinne wird er in der Valenztheorie mantika, nicht etwa “inhaltsleere” Konsti-
oder der Dependenzgrammatik für alle tuenten wie es in Es spielt das Orchester.
Funktionen außer der des finiten Verbs ver- Die Begriffspaare Thema und Rhema
wendet. Andere synonyme Termini sind ac- (funktionale Satzperspektive, vgl. Ammann
1928: 3; Mathesius 1929: 202f.; Boost 1955; Varianten dieser zwei Haupttypen sind bei-
Haftka 1981: 751f.; Welke 21994: 19⫺44), to- spielsweise:
pic und comment u. a.m. beziehen sich auf
den Informationswert der Funktionen, d. h. (a1) die Pronominalisierung: Der Zug fährt
ob sie ⫺ sehr vereinfacht ausgedrückt ⫺ in ab ⇒ Er fährt ab
der Darstellung “Neues” (Rhema) bringen (b1) die Frageprobe: Was kommt? ⫺ Der
oder schon “Bekanntes” (Thema) wiederge- Zug (und der Bus) kommen
ben. Das Verhältnis zu den herkömmlichen (b2) die Nominalisierung, z. B. Der Zug fährt
syntaktischen Funktionsbegriffen und zu ab ⇒ Die Abfahrt des Zuges.
Wort und Wortarten ist sehr frei und abhän- Wie die verschiedenen Kommutationen ver-
gig von Kontext und Situation, auch wenn wendet werden, zeigen die Ausführungen in
in einem normalen erzählenden Prosatext die 2.1 zu den Beispielen (1) bis (7). Die Trans-
Übereinstimmung Subjekt ⫺ Thema, Prädi- formation dient demselben Zweck, doch läßt
kat ⫺ Rhema nicht selten ist: der Hund (des sich durch sie besonders die Äquivalenz von
Nachbarn) (⫽Thema) biß den (neuen) Brief- Satz und Wort/Phrase nachweisen, vgl. 2.4,
träger (⫽Rhema). Je nach der Antwort auf (8) ⫺ (10). Die Methode ist für das Englische
durch Situation und Kontext motivierte zur Abgrenzung der Wortarten verwendet
Fragen können auch Konstituenten der ande- worden (vgl. Fries 1957: 74f.; Admoni 1971:
ren syntaktischen Funktionen die Rhema- 114).
funktion erfüllen: der Hund [des Nachbarn], In der Schulgrammatik wird besonders die
wenn die Frage lautet: Wer biß den Briefträ- Operation (b1) benutzt, um die syntaktischen
ger?, [den neuen] Briefträger bei der Frage Funktionen zu ermitteln, z. B. Jemand tut et-
Wen biß der Hund?, [der Hund des] Nachbarn was irgendwo. Frageprobe: Wer tut etwas? ⫺
bei der Frage Wessen Hund biß den Briefträ- Jemand (⫽ Subjekt); Was wird getan? ⫺
ger? etc. Etwas (⫽ Objekt); Wo wird das getan? ⫺
Irgendwo (⫽ Adverbial) etc. Auf diese Weise
2.3. Methodik können auch Thema-Rhema-Funktionen
Für die empirische Herausarbeitung von syn- nachgewiesen werden (vgl. Haftka 1981f.;
taktischen Funktionen und die Absicherung Welke 21994: 401).
der Begriffsbildung werden besondere Opera-
tionen verwendet, durch die beobachtet wird, 2.4. Geschichte
wie sich eine Konstante zu verschiedenen Va- Die Benennung Redeteile (partes orationis)
riablen verhält (vgl. z. B. Glinz 21961: 44f., für die Wortarten läßt vermuten, daß in der
85f.; Fries 1957: 75f.; Admoni 1971: 108⫺ zweitausendjährigen Geschichte der abend-
130). Dadurch können Begriffe voneinander ländischen Grammatik nicht von Anfang an
abgegrenzt und verborgene Relationen, Dis- zwischen Wortart und syntaktischer Funktion
krepanzen und Übereinstimmungen nachge- unterschieden wurde. Bedeutsam für die Aus-
wiesen werden. Operationen, die sich für die bildung der Unterscheidung waren besonders
Ermittlung syntaktischer Funktionen eignen, die Versuche der sog. Allgemeinen oder Phi-
sind: losophischen Grammatik, den sprachlichen
Satz mit dem logischen Urteil (Subjekt, Prä-
(a) Die Kommutation (Substitution, Er- dikat, Kopula) gleichzusetzen und die Wort-
satzprobe): Ein Glied in einem syntakti- arten nach ontologischen Kategorien (Sub-
schen Kontext wird durch ein anderes er- stanzwörter, Akzidenzwörter, Attributiva)
setzt, z. B. ein Wort durch eine Phrase: oder mit Bezug auf das logische Urteil (Sub-
Karl / mein kleiner Bruder kommt; eine jekts-, Prädikats-, Urteilswörter usw.) einzu-
Phrase durch einen Nebensatz: Wir er- teilen. Durch die Philosophische Grammatik
warten eine baldige Antwort / daß er bald wurde das Interesse auf syntaktische Fragen
antwortet; gelenkt. Häufig wurde hier Wortart und Satz-
(b) die Transformation: Ein Glied wird in funktion gleichgesetzt, vor allem Substantiv
bezug auf seine grammatische Form um- und Subjekt (vgl. Forsgren 1985: 48). Die
gewandelt, wobei das lexikalische Ele- Syntax der älteren Grammatik basierte im
ment und/oder sein Referent unverändert übrigen auf dem herkömmlichen Wortarten-
bleibt: Der Junge wirft den Ball ⇒ Der inventar. Johann Christoph Adelung (1732⫺
vom Jungen geworfene Ball. 1806) bediente sich in seiner Syntax einer auf
Von der Fügungspotenz der Wörter aus be- vermittelndes Glied zwischen Wort und Satz
trachtet gibt es demnach eine Übereinstim- zu verstehen sind (vgl. Lyons 1971: 174f.).
mung zwischen Wortart und Phrase, indem In Chomskys X-Bar-Theorie (Chomsky
sich die Phrase zu den syntaktischen Funktio- 71993: 48⫺55; Jackendoff 1977: 7⫺27) wer-
nen wie ihr Kernwort verhält. den die Phrasen nach ihrer Komplexität klas-
sifiziert. Die kleinste Stufe sind Wörter oder
3.2. Theorien und Terminologien “lexikalische Kategorien” (N[omen], V[erb],
Normalerweise wird der Terminus Satzglied A[djektiv]), die den “Kern” (“Kopf”, “head”)
in der heutigen Linguistik für die Bezeich- der Phrase bilden und mit X0 bezeichnet wer-
nung von verschiedenen Konstituentenkate- den. Je nach der Komplexität werden die Phra-
gorien (Wort, Phrase, Gliedsatz) verwendet sen danach als X1, X2 oder X⬘, X⬙ usw. mar-
(vgl. z. B. Engelen 1986: 33⫺49), insofern als kiert.
diese Satzfunktionen erfüllen. Dasselbe gilt
für den Terminus Gliedteil wodurch dieselben 3.3. Methodik
Kategorien (Wort, Phrase, Gliedsatz) begriff- Um Phrasen und Phrasenkonstituenten zu
lich subsumiert werden, insofern als diese ermitteln, werden in der Regel andere Opera-
Konstituenten Funktionen innerhalb einer tionen verwendet als Kommutation und
Phrase erfüllen (Attribut, Kernwort, Deter- Transformation (vgl. 2.3):
minant). Satzglied und Gliedteil sind dem-
nach den anderen Konstituentenkategorien (a) die Permutation (Verschiebeprobe): Ein
übergeordnet und deshalb selbst als (abstrak- Glied wird im Verhältnis zu den übrigen
tere) Kategorienbegriffe anzusehen. Gliedern im Kontext verschoben, z. B.
In demselben Sinne wie Satzglied wird Wir warten auf den Zug ⇒ Auf den Zug
auch Phrase als Terminus für die Konstituen- warten wir;
ten syntaktischer Funktionen benutzt. Die (b) die Reduktion (Weglaßprobe): Ein Glied
engere Bezogenheit der Phrasen auf die oder mehrere werden weggelassen, z. B.
Wortarten spiegelt sich darin wider, daß die Der verspätete Zug kam endlich an ⇒ Der
verschiedenen Phrasen nach den Wortarten Zug kam an; und sein Gegensatz, die
benannt werden, zu denen ihr Kernwort ge- Augmentation (Interpolierung, Inser-
hört: Nominalphrase, Adjektivphrase, Adver- tion): Ein Glied wird eingesetzt, z. B. Der
bialphrase. Der Begriff Verbphrase hat jedoch Zug kam an ⇒ Der verspätete Zug kam
einen anderen Status. Auf der ersten Verzwei- an.
gungsstufe des Satzes (S) werden nämlich in Durch Permutationen lassen sich vor allem
der Phrasenstruktur-Grammatik und der ge- Phrasen ermitteln:
nerativen Grammatik sowohl das finite Verb
als auch die davon abhängigen Träger der (14) Der Hund des Nachbarn biß den neuen
Satzfunktionen in den Begriff einbezogen Briefträger ⇒ Den neuen Briefträger biß
und von der Nominalphrase der Subjekts- der Hund des Nachbarn.
funktion unterschieden (Chomsky 1965: 72):
Dadurch werden demnach Phrasen als klein-
ste umstellbare Satzkonstituenten isoliert,
(13) S J [Karl liest ein Buch] J NP VP J vgl. *des Nachbarn der Hund; *der Briefträ-
[liest ein Buch] J V NP J [liest] [ein ger neue. (Zur Problematik der nicht in allen
Buch]. Sprachen gleichermaßen verwendbaren Per-
N[omen] und V[erb] sind Bezeichnungen für mutation vgl. z. B. Helbig 1972: 334f.; 1982:
Wortarten (lexikalische Kategorien) im Un- 45f.).
terschied zu Phrasen (syntaktische Katego- Mit Hilfe der Reduktion läßt sich das
rien), die nach der Wortart des Kernwortes Kernwort der Phrase feststellen:
bezeichnet werden (NP [Nominalphrase], VP (15) Der Hund des Nachbarn biß … ⇒ Der
[Verbalphrase]). Hund biß …
Ein anderer häufig verwendeter Terminus, (16) der sehr kleine Hund ⇒ der kleine Hund
der Bezeichnung für alle segmentierbaren
Einheiten vom Wort bis zum Satz sein kann, In (15) ist demnach Hund das Kernwort und
ist Syntagma. Besonders nach britischem in (16) klein, weil Hund und klein nicht weg-
Vorbild wird er auch für Phrasen gebraucht, laßbar sind, sofern der Satz nicht elliptisch
die kein finites Verb enthalten und somit als ist. Dagegen können in der Regel die Attri-
Grammatik von Karl Ferdinand Becker bis Konrad Knobloch, Clemens (1989 b), “Einige Entwick-
Duden 1830⫺1880. Münster: Nodus lungstendenzen der deutschen Schulgrammatik
Forsgren, Kjell-Åke (1992 b): “Zum Problem des nach Karl Ferdinand Becker”. Erlinger et al.
Satzbegriffes im Deutschen”. Beiträge zur Ge- (Hrsg.), 87⫺113
schichte der deutschen Sprache und Literatur (Tü- Lyons, John (1971), Einführung in die moderne Lin-
bingen) 114, 3⫺27 guistik. München: Beck
Fries Charles C. (1957), The Structure of English: Mathesius, Vilém (1929), “Zur Satzperspektive im
An Introduction to the Construction of English Sen- modernen Englisch”. Archiv für das Studium der
tences. London: Longman & Green [11952] neueren Sprachen und Literaturen 84.155 [NS 55],
202⫺210
Glinz, Hans (1947), Geschichte und Kritik der Lehre
von den Satzgliedern im Deutschen. Bern: Francke Müller, Beat Louis (1985), Der Satz: Definition und
sprachtheoretischer Status. Tübingen: Niemeyer
Glinz, Hans (21961), Die innere Form des Deut-
Palmer, Frank R. (1994), Grammatical Roles and
schen. Bern: Francke [11952]
Relations. Cambridge: Cambridge University Press
Grewendorf, Günther & Hamm, Fritz & Sterne- Paul, Hermann (1920), Prinzipien der Sprachge-
feld, Wolfgang (21989), Sprachliches Wissen. schichte. Halle/S.: Niemeyer
Frankfurt/M.: Suhrkamp [11987]
Pleines, Jochen (1978), “Ist der Universalanspruch
Gulyga, Elena (1978), “Die Satzgliedtheorie und der Kasusgrammatik berechtigt?”. In: Abraham,
die Hypotaxe”. In: Helbig, Gerhard (Hrsg.), Bei- Werner (Hrsg.), Valence, Semantic Case and Gram-
träge zu Problemen der Satzglieder. Leipzig: Enzy- matical Relations. Papers Prepared for the Working
klopädie, 56⫺78 Group “Valence and Semantic Case”, 12th Interna-
Haftka, Brigitta (1981), “Reihenfolgebezeichnun- tional Congress of Linguistics, Univ. of Vienna Aug.
gen im Satz (Topologie)”. In: Heidolph, Karl Erich 29 ⫺ Sept. 3, 1977. Amsterdam: Benjamins (Stu-
(Hrsg.), Grundzüge einer Deutschen Grammatik. dies in Language Companion Series 1), 366⫺375
Berlin: Akademieverlag, 702⫺764 Ries, John (21927), Beiträge zur Grundlegung der
Syntax, Bd. I: Was ist Syntax. Prag: Taussig und
Harris, Zellig S. (1951), Methods in Structural Lin-
Taussig [11894; unveränd. reprograph. Nachdruck,
guistics. Chicago: University of Chicago Press
Darmstadt: Wissenschaftliche Buchgesellschaft,
Haselbach, Gerhard (1966), Grammatik und 1967]
Sprachstruktur: Karl Ferdinand Beckers Beitrag zur Ries, John (1928), Beiträge zur Grundlegung der
Allgemeinen Sprachwissenschaft in historischer und Syntax, Bd. II: Zur Wortgruppenlehre. Prag: Taus-
systematischer Sicht. Berlin: de Gruyter sig und Taussig
Hays, David G. (1964), “Dependency Theory: A Ries, John (1931), Beiträge zur Grundlegung der
Formalism and Some Observations”. Language Syntax, Bd. III: Was ist ein Satz. Prag: Taussig
40, 511⫺525 und Taussig
Helbig, Gerhard (1972), “Zu Problemen des Attri- Sütterlin, Ludwig (21910), Die deutsche Sprache der
buts in der deutschen Gegenwartssprache (1)”. Gegenwart. Leipzig: Voigtländer [11900]
Deutsch als Fremdsprache 6, 332⫺341 Stepanowa, M[arija] & Helbig, G[erhard] (21981),
Helbig, Gerhard (1982), Valenz ⫺ Satzglieder ⫺ se- Wortarten und das Problem der Valenz in der deut-
mantische Kasus ⫺ Satzmodelle. Leipzig: Enzyklo- schen Gegenwartssprache. Leipzig: Bibliographi-
pädie sches Institut [11978]
Jackendoff, Ray (1977), X-Bar-Syntax: A Study of Tesnière, Lucien (1959), Éléments de syntaxe struc-
Phrase Structure. Cambridge/MA: MIT Press turale. Paris: Klincksieck
Vesper, Wilhelm (1980), Deutsche Schulgrammatik
Jackendoff, Ray (1990), Semantic Structures. Cam-
im 19. Jahrhundert. Tübingen: Niemeyer
bridge/MA: MIT Press
Weigand, Georg (1967), Karl Ferdinand Becker ⫺
Jellinek, Max Herman (1914), Geschichte der neu- ein hessischer Pädagoge und Sprachphilosoph des
hochdeutschen Grammatik von den Anfängen bis auf 19. Jahrhunderts. Frankfurt/M.: Diesterweg
Adelung, Zweiter Halbband. Heidelberg: Winter
Welke, Klaus, (21994), Funktionale Satzperspektive:
Knobloch, Clemens (1988), Sprache als Technik der Ansätze und Probleme der funktionalen Grammatik.
Rede. Frankfurt/M.: Lang (Theorie und Vermitt- Münster: Nodus [11992]
lung der Sprache 9) Wells, Rulon S. (1947), “Immediate Constituents”.
Knobloch, Clemens (1989 a), “Die deutsche Schul- Language 23, 81⫺118
grammatik vor dem Erscheinen von Karl Ferdi-
nand Beckers ‘Organism der Sprache’ (1827)”. In: Kjell-A˚ke Forsgren, Högskolan i Skövde
Erlinger et al. (Hrsg.), 62⫺86 (Schweden)
Schachter, Paul & Otanes, Fe T. (1972), Tagalog tive’”. In: Hawkins, John A. (ed.), Explaining Lan-
Reference Grammar. Berkeley: University of Cali- guage Universals. Oxford: Blackwell
fornia Press Underhill, Robert (1976), Turkish Grammar. Cam-
Shopen, Tim (1985, ed.), Language Typology and bridge/MA: MIT Press
Linguistic Description, Vol. III: Grammatical Cate-
Walter, Heribert (1981), Studien zur Nomen-Verb-
gories and the Lexicon. Cambridge: CUP
Distinktion aus typologischer Sicht. München: Fink
Simon, Walter (1937), “Has the Chinese Language
Parts of Speech?”. Transactions of the Philological Wierzbicka, Anna (1980), Lingua Mentalis. Syd-
Society 1937, 99⫺119 ney: Academic Press
Swadesh, Morris (1939), “Nootka Internal Syn- Wierzbicka, Anna (1986), “What’s in a Noun? (Or:
tax”. International Journal of American Linguistics How Do Nouns Differ in Meaning from Adjec-
9, 77⫺102 tives?)”. Studies in Language 10, 353⫺389
Thompson, Sandra (1988), “A Discourse Ap-
proach to the Cross-linguistic Category ‘Adjec- Nicholas Evans, Melbourne (Australia)
73. Noun
Latin nomen, so that the traditional and the markers, the closest syntactic environment is
modern meanings and hyponymy relations of chosen as criterial, e.g. a class of determiners
noun and substantive are exactly reverse. In or nouns of multitude.
what follows, the term noun s. l. ([sensu lato] The class of elements which thus figures in
noun in the wide sense) will be used for the the distributional definition of a word class
supercategory. could itself be conceived as a distributional
class, and its definitional environment could,
in principle, be the noun. This would obvi-
2. The concept ‘noun’ ously lead to circularity. If this is to be
avoided, the classes of elements which are cri-
Like any other grammatical category, the terial in the definition of a word class such as
word class ‘noun’ has no universal status a ‘noun’ must be grammatical categories which
priori; rather, it is a language-specific cate- are defined by direct recourse to the func-
gory. However, and again like with all other tions of language in communication and cog-
grammatical categories, there is a universal nition, such as are discussed in chapter XII
(cognitive or communicative) basis to it. Lan- of this handbook. The selection of the partic-
guages have a class of expressions which de- ular definitional criteria is, in theory, arbi-
signate entities, i.e. concepts that are reified, trary. In practice, only such word-class sys-
and which can be used to refer to specific en- tems have met with general acceptance whose
tities. The prototypical representatives of this definitions are heuristically guided by seman-
kind of concept are concrete individual phys- tic prototypes like the one mentioned above
ical objects such as a bird or an apple. The for nouns.
status of this class of expressions in the lan- Declension is the inflection of nouns. This
guage system may vary. It need not, in prin- concept has an ambiguous position in the
ciple, be a grammatical class, let alone one definitional hierarchy, since we can, in prin-
definable by morphological criteria. How- ciple, either define the noun on independent
ever, if there are grammatical criteria for as- grounds and then define declension as what-
signing words (lexemes) to a class with these ever inflection appears on nouns, or else we
semantic properties, then this is the class of can define declension as inflection for certain
nouns of the language. Up to now, probably morphological categories and then define the
all grammars have made use of a grammati- noun as the word which is declined. Given
cal category of noun. In particular, nouns the definitional procedure for nouns which
generally seem to form a distinct class by syn- we outlined in the preceding paragraph, it
tactic criteria, although not necessarily by appears that only the second way is passable.
morphological or phonological criteria. Un- This means that declension in Latin is inflec-
like other word classes like the adjective and tion for gender, number and case, while de-
the numeral classifier, the noun is therefore clension in Turkish is inflection for number,
universal in the sense that there is an empirical case and possessor. Analogous considerations
generalization that every language described apply, of course, to the concept of conjuga-
so far has a syntactic class which corresponds tion as the inflection of verbs.
to the notional definition of the noun.
The word classes of each language are de-
limited on distributional grounds. Essen- 3. Semantic properties
tially, if a class of stems such as that of nouns
is to be defined as a distributional class, then Concepts differ in their time-stability, i.e. in
a class of elements that constitutes the rele- the extent to which corresponding phenom-
vant criterial environment needs to be given ena are prone to change (cf. Givón 1979:
beforehand so that it can be used without cir- ch. 8). For instance, the assignment of an en-
cularity in the definition of the stem class. tity to a class is more stable than a property
Generally, this will be a class of elements with that it has, which in turn is more stable than
which the definiendum forms a close gram- a state that the entity is in, and so forth, mov-
matical (including morphological) construc- ing rightward in Fig. 73.1. The noun-verb
tion, such as those treated in 4. In the clearest distinction can be associated with time-sta-
cases, the criterial class is one of morphologi- bility in the sense that the most dynamic con-
cal markers, in the present instance, a nomi- cepts of all languages are manifested by
nal morphological category such as gender or verbs, whereas the most static concepts are
case. In the absence of such morphological manifested by nouns. For those languages
that have them, adjectives range between the tion of the term nomen substantivum intro-
two poles. duced in 1. Intuitively, the difference between
the meaning of a noun (‘substance’) and the
time-stability static ø¿¿¿¿¿¿¿¡ dynamic meaning of an adjective (‘property’) is that if
concept type class property state process event all the properties are subtracted from the ad-
part of speech noun adjective verb jective meaning, nothing is left, but if they are
subtracted from the noun meaning, something
Fig. 73.1: Scale of time-stability is left (viz. ‘person’, in the case of Toter).
Entities may be classified according to di-
Thus nouns typically designate (members of) verse criteria. The classification which is most
classes such as birds or apples. These are enti- relevant to linguistic structure is one accord-
ties which are not just features of an event, ing to the empathy which the speaker has
but conceptually independent participants of with the type of entity. Since this is a matter
it. Less typically, nouns designate properties, of degree, the classification takes the form of
like a beautiful one or a green one (cf. 5 be- an empathy hierarchy (cf. Kuno & Kaburaki
low and Art. 72: 2.3 on adjectives in 1977). One of the basic distinctions in the hi-
Quechua). The concept types further to the erarchy is between animate and inanimate
right in Fig. 73.1 are not directly expressed by beings, which is why the hierarchy has also
nouns. Abstract, typically derived nouns such been known by the name of animacy hierar-
chy (e.g. in Comrie 1981: ch. 9). In general,
as sickness, occupation and conquest may be
the speaker has most empathy with such enti-
used to designate states, processes and events
ties that are closest to and maximally like
as reified entities (cf. Art. 94). Thus, a noun himself. These are the speech-act participants
such as sickness does not designate a state or (SAP). In fact, for most purposes, the
property as conceived in Fig. 73.1. speaker himself forms the top of the hierar-
The noun is minimally distinct from the chy. The speaker has least empathy with phe-
adjective. For example, the German adjective nomena that are so little individuated and
tot designates the property ‘dead’, the noun thing-like that they cannot even properly be
Toter designates the class of dead persons, regarded as entities s. s. [in the strict sense],
whose intension is, of course, the property in viz. with propositions (abstract concepts) and
question. From Aristotle on, nouns were said locations. Other entities occupy intermediate
to designate ‘substances’, which here does positions on the hierarchy, as shown in
not mean ‘masses’, but ‘entities with ontolog- Fig. 73.2. Different languages may make
ical independence’. This provides the motiva- finer or grosser distinctions.
SAP Non-SAP
human non-human
animate inanimate
object location
The empathy hierarchy is relevant to diverse The prototypical noun designates a concept
structural features and grammatical rules in which comprises a class of concrete individ-
all known languages. Aspects of it also un- ual physical objects. This, however, presup-
derlie subclassifications of nouns (cf. 5), no poses that the object in question can be sub-
matter whether these manifest themselves at sumed under a class in the first place. This is
the morphological level, e.g. in the form of not so for those entities which form the top
noun classes, or only at the syntactic level, of Fig. 73.2. They are so highly individuated
e.g. in constructional differences between that they cannot even be assigned to a class.
mass and count nouns. The structural correlate of this fact is that
It was said in 2 that nouns such as apple they are normally not represented by nouns,
and bird are focal instances of their species. but by pronouns. On the other hand, con-
cepts at the bottom of the empathy hierarchy guage (cf. Seiler 1986). Some classifications
do not provide typical noun meanings either, of nouns are signalled by words, such as nu-
because they are not sufficiently individ- meral or possessive classifiers. If so, then any
uated. If we consider the empathy hierarchy noun that can partake in the relevant syntac-
as a continuum between the poles of the SAP tic construction at all (i.e., can combine with
and the proposition, it may be seen that the a numeral and a possessive pronoun, respec-
prototypical noun designates an entity which tively) is combined with one of the classifiers
occupies a central position on this contin- and to that extent belongs to a particular
uum. Speech act participants on the one class. However, these classes tend to be shift-
hand and abstract entities on the other hand able in the sense that the combination of a
are peripheral to the class of nouns. noun with a different classifier is not ungram-
matical, but, rather, it leads to a different
(possibly less usual) interpretation.
4. Nominal categories A more grammaticalized form of nominal
This section surveys a range of morphologi- classification is found in noun classes (cf.
cal categories which are also included in the Art. 98). These are generally based on cogni-
onomasiological treatments of ch. XIII and tive categories such as individuality, animacy,
XIV. Here we focus on their nominal charac- humanness, etc. Any given noun stem is as-
ter, i.e. they are seen as categories which may signed to one of the classes. The class may be
provide the relevant context for a distribu- marked morphologically on the noun itself,
tional definition of the noun in the sense of as are the noun classes in Bantu languages
2. For surveys of the inflectional categories of (cf. Art. 140). A shift of noun class is a deri-
nouns, cf. Givón 1984: 57⫺64 and Anderson vational process which is subject to con-
1985 a: 174⫺189. straints that are partly semantically moti-
Although we concentrate on nominal cate- vated, partly idiosyncratic, as is typical for
gories qua morphological categories, it processes of word-formation. Noun classes
should be clear that all of the notional cate- appear on other constituents of the sentence,
such as determiners, adjectives and verbs if
gories coded in nominal morphological cate-
they agree with the noun. Class markers may
gories can also be coded in other ways. On
have different allomorphs for nouns and the
the one hand, they may appear as indepen-
various word classes that agree with them.
dent words, which may be grammaticalized
The most grammaticalized form of nomi-
to different degrees. For instance, the rela-
nal classification is gender. Gender is typi-
tions coded by case may also be coded by
cally sex-based, whereas noun class is mostly
adpositions, and these may be either concrete not. It is common for a language to have two
or grammatical. On the other hand, such no- or three genders. One type of twofold subdi-
tions may also be lexical-semantic features of vision yields animate and inanimate gender.
nominal stems. For instance, definiteness In Menomini, nouns designating animate be-
may be a feature of proper names; and argua- ings, including large plants, all belong to one
bly, denominal adverbs such as home (in go gender which Bloomfield (1962) calls ani-
home) embody a case. Similar considerations mate, while nouns designating inanimate en-
apply to all of the categories treated below. tities may be of either animate or inanimate
Subsections 4.1⫺4.8 review the range of no- gender. Hittite, too, has an animate and an
minal categories in inflection and derivation, inanimate gender, which, on the basis of rela-
while the remaining subsections make gener- tions to Indo-European cognates, are called
alizations about the distribution of nominal common gender (genus commune, i.e. indif-
categories over languages and within a lan- ferent to the masculine vs. feminine distinc-
guage. tion) and neuter. Another type of twofold
subdivision produces a masculine and a femi-
4.1. Nominal classification
nine gender, as in the Romance languages.
The entities of our cognitive world fall into Three genders are often masculine, feminine
many different categories, such as concrete and neuter, as in Latin, Russian and German.
and abstract objects, animate and inanimate Gender is an inherent grammatical cate-
beings, natural objects and artefacts. Such gory of a nominal lexeme, i.e. it is not as-
classes figure in the selection restrictions of signed by a syntactic rule. It need not be, and
verbs and adjectives perhaps of all languages often is not, marked separately on the noun
and to that extent are relevant to semanto- itself, but (just as noun class) appears on sen-
syntax. The classification of entities may be tence constituents that agree with a noun. On
grammaticalized to different degrees in a lan- the noun itself, gender is often bound to de-
clension class (cf. Art. 65). In Latin, for in- sification systems. First, the classification by
stance, the relation is twofold. First, a given gender is almost exhaustive, in the following
declension class may only contain nouns of sense: If a language has gender, then almost
one particular gender. For instance, the a-de- every noun has a determinate gender which
clension contains only feminine nouns (with shows up either on the noun itself or on some
few exceptions). Second, gender may deter- agreeing constituent. There may be some ex-
mine allomorphy within a declension class. ceptions to this; for instance in German, gen-
For instance, in all Latin declension classes der is neutralized in the plural, so that plu-
the nominative is syncretic (cf. Art. 66) with ralia tantum are of indeterminate gender. In
the accusative for neuters, but not for the other systems of nominal classification, the
other genders (cf. also Tab. 73.2 for Hittite). number of nouns of indeterminate class is
Moreover, the nominative has an -s ending much larger. In numeral classification, for in-
in most declension classes on masculine and stance, all the nouns that do not combine
feminine, but not on neuter nouns; e.g. in the with numerals never co-occur with a classifier
-u-declension: fructu-s (nom. m.) ‘fruit’, and thus fall into no corresponding class. Se-
domu-s (nom. f.) ‘house’, cornu (nom./acc. n.) cond, while the weakly grammaticalized sys-
‘horn’. Even more commonly than for noun tems of nominal classification may classify
classes, the gender of a noun is copied, by concepts, with relatively few arbitrary assign-
agreement, to other parts of speech, and then ments of nouns to classes, gender essentially
exponents of the genders may differ for classifies nouns. I. e., gender classifies not de-
nouns and those other parts of speech. In signata, but linguistic signs. In this sense, the
Russian, for instance, each gender has only classification is metalinguistic.
one morph in the singular of past verbs (-0, Most languages have at least one gram-
-a, -o), while on singular nouns there is abun- matical category that somehow classifies
dant allomorphy. nouns. Languages such as Turkish, which ⫺
Gender is inherent in a noun stem and so apart from syntactic reflexes of the empathy
highly grammaticalized that there are no gen- hierarchy ⫺ do not possess any grammatical
erally productive processes for its change. category of nominal classification, are rela-
The exception is what has traditionally been tively rare. At the pole of low grammaticali-
called motion (‘movement into a different zation, sometimes more than one classifica-
gender’): in nouns designating animate be- tory system may coexist in a language. An
ings, gender may designate sex, and then it admittedly extreme case is Yucatec Maya,
may be changed more or less productively, as which has numeral classification, possessive
in Latin lup-us (m.) ‘wolf’ ⫺ lup-a (f.) ‘she- classification and a prefixal sex distinction
wolf’; Span. muchach-o ‘boy’ ⫺ muchach-a (with metaphorical extensions reminiscent of
‘girl’. The high degree of grammaticalization gender). The three classifications are com-
of gender correlates with two other features pletely independent of each other and in (1)
which distinguish it from other nominal clas- co-occur in one noun phrase.
(1) ka1-túul in w-àalak1 h-taman
two-cl.anim poss.1.sg 0-cl.domestic m-sheep
‘two of my rams’
sen for a nominal on semantic grounds, from multitude as plural markers. In the evolution
where it may be transferred to other sentence of the Indo-Arian languages, nouns such as
components (cf. Art. 100). A typical case is (2). Sanskrit sakala ‘all’ > Old Bengali saēla, jana
(2) (a) The girl is there. ‘people’ > janø a, loka ‘world’ > lōa develop
(b) The girls are there. the grammatical function of plural markers.
They are combined with their host nouns in
However, it is not unusual for a noun to what is formally a compound, as in (4) from
have its number fixed lexically. Many nouns Old Bengali (Kölver 1982: 247).
do not form a plural, among them proper
nouns such as Iran and Saturn, collective (4) (a) manø dø ala-saēla bhājai
nouns such as cattle and police, mass nouns mandala-all broken
such as milk and rubbish, and abstract nouns ‘(all) the mandalas were broken’
such as advice and specificity. On the other (b) bidujanø a-lōa
hand, some nouns, including trousers and scholar-world
ashes, or German Einkünfte ‘revenue(s)’, do ‘the scholars’
not form a singular. A noun that is only used In Middle Bengali, more nouns of this kind
in the singular is a singulare tantum (Latin are recruited as plural markers, and several
‘singular only’), and one occurring only in of them appear as plural suffixes of present
the plural is a plurale tantum. In both cases, day Bengali.
the number opposition is neutralized. To the Number is not only copied to other parts
extent that these are nouns of the lower levels of speech by agreement, but ⫺ in contrast to
of the empathy hierarchy, it may be said that nominal class ⫺ it also has an independent
on its spreading downward the hierarchy, status on the verb (cf., again, Art. 100). On
number does not reach these subcategories. verbal plurality, see Dressler 1968; for a com-
Often the bare noun stem or at any rate a parison of nominal and verbal plurality, see
morphologically unmarked form serves as Mithun 1988 and Gil 1991. Just as with the
the singular, against which the plural is other nominal categories, the number para-
marked overtly. However, the opposite also digm may be the same on nouns and on other
occurs, for instance in Arabic. The unmarked word classes, or the allomorphy may differ.
stem may designate a collective, as in (3 a). Thus, while in Russian or German, adjectives
From this a singulative may be formed, as in and nouns take different suffixes for the same
(3 b), which may again be the basis for a plu- number-and-case categories, in Latin, for ex-
ral form, as in (3 c). Neither the collective nor ample, the affixes taken by adjectives form a
the singulative, but only the plural can di- proper subset of the pool of nominal affixes.
rectly combine with a numeral. Especially where verbal number is in a cross-
(3) (a) tßalātßa ru1ūs baqar reference relationship with the number of
three head\pl cow.coll some verb actant, the same number paradigm
‘three heads of cattle’ may be used on nouns and verbs. Turkish has
(b) baqar-a the suffix -ler on plural nouns and on third
cow-sglv person plural verb forms, and the same goes
‘cow’ for Yucatec -o1b and for the Hebrew (Gil
1991: 8f.) and Hungarian plural markers. Re-
(c) tßalātßa baqar-āt duplication is also often used for expressing
three cow-pl plurality in either lexical category.
‘three cows’ (Premper 1986: 4)
On the other hand, collective nouns may 4.3. Case and stem alternations
also be derived from stems which bear a Case is an inflectional category which ap-
number distinction. Portuguese, e.g., derives pears on a noun phrase or its constituents
collective nouns from individual nouns with and expresses the former’s syntactic or se-
the suffix -ada, as in menino ‘boy, child’ ⫺ mantic function in the construction. Case
meninada ‘group of children’. marking is the grammatical technique of sig-
Only highly grammaticalized number, such nalling case. In recent anglophone literature,
as in the Indo-European languages of the ar- the term may also include the marking of
chaic type, is obligatory, so that even nouns categories of actants (generally person and
combined with a numeral or quantifier are in number) on the verb in the form of pronomi-
the plural. Less grammaticalized number is nal affixes, and even verbal voice. However,
usually optional. At the pole of lowest gram- this usage is inappropriate (cf. Lehmann
matical status, we find the so-called nouns of 1988: § 5). Case is very widespread in the
world’s languages. In particular, languages the head noun as the representative of the
with verb-final word order almost always noun phrase, as in the Turkish (6).
have case (Greenberg 1963: 쒙41).
(6) küçük ev-den
Case systems are surveyed in article 102.
small house-abl
As indicated in the definition, the locus of
‘from the small house’
case is the noun phrase (cf. also 7 below). If
the noun phrase is governed by its depen- In Bété (Kwa, Ivory Coast), there is one case,
dency-controller, then its case will typically the genitive, and it is marked by lengthening
be a grammatical case (rather than a concrete the final vowel of the noun, as in (7).
or semantic one), and it may be analyzed as
assigned by the controller, as the accusative (7) lẃí-í gèı́
of seruam in (5) from Latin is assigned by elephant-gen tail
uerberat. If the noun phrase, instead, modi- ‘elephant’s tail’
fies its controller, then its case will typically In such languages, the noun is usually the last
be a concrete case, and it may be analyzed as constituent of the noun phrase. Grammatical
chosen for this NP on semantic grounds, as analysis has to ascertain whether case in this
the ablative of baculo. instance is a morphological category of the
(5) domina uerberat noun or is instead attached to the entire
mistress:nom.sg.f beats noun phrase.
Another possibility is for case to spread to
seruam baculo
the modifiers and determiners of a noun, as
servant:acc.sg.f stick:abl.sg.n
in the Latin (8). As far as marking external
‘the mistress beats the servant with a
syntactic or semantic relations of an NP is
stick’
concerned, this spreading of case only in-
In either instance, the case of the NP may creases redundancy. Like other forms of
percolate from there to its subconstituents. agreement, it signals coreference of the con-
One frequent possibility is case marking on stituents so marked.
(8) ill-um porc-um siluatic-um
d3-acc.sg.m pig-acc.sg.m forest:adj-acc.sg.m
‘that forest pig (acc.)’
(9) d-en
the-acc.sg.m/dat.pl
alt-en Frau-en
old-obl.sg.m//gen/dat.sg.f/n//pl woman(f )-pl
‘to the old women’
A slightly different form of marking case is based on a different stem from the nomi-
found in German. In noun phrases such as native. The oblique stem is essentially equal
(9), case is marked neither on the NP nor on to the genitive, except that this has an added
the noun, but only on the latter’s co-constitu- “enunciative vowel”.
ents (determiners and modifiers). The declen-
sion involves so much syncretism that the nominative maram
grammatical categories of the NP ⫺ dative genitive maratt-u
plural feminine in the instance of (9) ⫺ have accusative maratt-e
to be factored out as the intersection of the dative maratt-ukku
sets of possibilities contributed by each word instrumental maratt-aale
form (marked by the slash in the gloss of (9); comitative maratt-oote
cf. Werner 1979). locative maratt-ile
Noun stems frequently undergo al- ablative maratt-ile-runtu
ternations under case marking. Various lan-
guages have a morphological opposition be- Tab. 73.1: Declension of Tamil maram ‘tree’ (Asher
tween one casus rectus (lit. ‘upright case’), 1982: 103)
which is the absolutive or nominative, and
the remaining cases, which are casūs obliqui This kind of noun stem alternation is called
(oblique cases). The Tamil paradigm in heteroclisis and is also known from Hittite
Tab. 73.1 shows that the oblique cases are and other ancient Indo-European languages.
Here, heteroclitic neuter stems such as the Various languages of the Americas make a
one of Tab. 73.2 end in r in the nominative/ morphological distinction between that form
accusative, but in n in the other cases. of a noun which forms a syntactic constituent
and that form which is a morphological part
nominative/accusative watar-0 of the verb, either a stem incorporated in the
genitive weten-as latter or functioning as a derivational mor-
dative weten-i pheme (cf. Art. 88). In Nahuatl, the noun ap-
pears in the absolutive form (which here is
Tab. 73.2: Some singular case forms of Hittite wa- not a case ⫺ Nahuatl has no case) if it is in-
tar ‘water’ dependent (10 a), but as a bare stem if it is
incorporated in the verb (10 b).
A different kind of stem alternation which
may be found in declension is apophony (Ger- (10) (a) ni-ki-kwa in naka-tl
man Ablaut) familiar from Indo-European sbj.l-obj.3-eat def meat-abs
languages and shown in Tab. 73.3. In a cou- ‘I eat the meat’
ple of declension classes, the stem has full (b) ni-naka-kwa
grade (a) in the locative, lengthened grade (ā) sbj.l-meat-eat
in the nominative and zero grade (no vowel) ‘I eat meat’
in the genitive.
Among the entity concepts which often
nominative rājā-0
function in derivation, body parts are promi-
locative rājan-i nent. If a body part is used as an instrument,
genitive rājñ-as Yucatec Maya has the alternative of either
constructing an instrumental adjunct, as in
Tab. 73.3: Some singular case forms of Sanskrit (11 a), or incorporating the notion into the
rājan- ‘king’ verb, as in (11 b).
Just like any other bound personal mark- nouns and verbs. From among nouns, it in-
ers, possessive markers may either have full volves alienable nouns only if it also involves
pronominal function, i.e. occur without or inalienable nouns. From among verbs, it in-
even exclude a coreferential NP in the same volves intransitive, imperfective, modal, inde-
phrase ⫺ in this case: the possessor NP ⫺, or pendent verb forms only if it also involves
else co-occur with such an NP, agreeing with transitive, perfective, indicative, subordinate
it in pronominal categories such as person, verb forms.
number, gender, as in the Turkish (12) cf.
Art. 75). 4.4.2. Other possessive morphology
If the possessor is actually represented in the
(12) oǧlan-ın sœapka-sı
form of an NP, rather than by an affix, the
boy-gen hat-poss.3
relation of the possessor NP to the possessum
‘the boy’s hat’
NP must be somehow indicated. The relation
The paradigm of possessive affixes in a may be inherent in the possessed noun, as
language is mostly identical or similar to an when it designates a body part, a kin relation
affix paradigm that appears on the verb to or another relational concept (cf. 5). In this
cross-reference an actant (see Seiler 1983: case, the internominal relation is commonly
ch. 5). In Turkish, the paradigm of possessive not specifically marked. Otherwise, the pos-
suffixes is cognate, and in some forms iden- sessed noun may be equipped by a relator (a
tical, with the paradigm of person endings on relational marker) which converts it into a re-
the verb. In Quechua, the verb has personal lational noun. Or else the possessor NP may
suffixes cross-referencing the subject and the be equipped with a relator which converts it
object, and the former are virtually identical into a modifier or determiner. The relator in
with the possessive suffixes on the noun. In the last-mentioned technique is a genitive
Abkhaz, the verb has personal prefixes for case or, at any rate, a case marker or adposi-
absolutive, ergative and indirect object, the tion with a similar function, and is therefore
latter two paradigms being virtually identical included in the treatment of 4.3.
with each other and with the possessive pre- The distinction between relational and
fix paradigm on nouns. In Arabic, the suf- non-relational nouns appears as a grammati-
fixes cross-referencing the object on the verb cal one in many languages, in the form of
are from the same paradigm as the possessive the two classes of alienable and inalienable
suffixes on the noun. nouns. In Andoke (isolate), inalienable nouns
The similarity between possessive affixes only occur with a possessive pronoun or af-
on nouns and person affixes on verbs is best fix, as in (13 a), while alienable nouns may oc-
manifested for specific subcategories of cur without them, as in (13 b).
suffix example
past -pwer rók-wér-a ‘former house’
tenseless 0 rók-a ‘house’
future -ram rók-wám-a ‘future house’
rók-ám-wér-a ‘ex-house-in-spe
(what was to be a house)’
The tense morpheme situates the referent Even where both the noun and the verb have
of a noun in time relative to the time of the tense, tense is selected independently for a
clause containing it. verb and its nominal dependents, as in My ex-
While there may be agreement between a wife is visiting me, my future wife visited me,
nominal dependent and its verb in other cate- etc. Nevertheless, tense markers in nouns and
gories, tense is not an agreement category. verbs may be phonologically identical. Pota-
watomi (Hockett 1958: 238), Kwakw’ala (An- noodle), and again viòla ‘viol’ ⫺ violone ‘bass-
derson 1985 b: 30), and Tagalog (Schachter & viol’ ⫺ violoncello ‘viol-aug-dim (cello)’.
Otanes 1972: 153f.) for example, use the same Shift of gender or noun class (motion in
affixes, but they are completely distinct in Hix- the sense of 4.1) may be used to derive dimin-
karyana (Derbyshire 1985: 201, 196) (cf. also utives and augmentatives. Thus, Kxoe
the difference between the English verb mar- (Central Khoisan) has yı̀ı́ ‘tree’ ⫺ yı̀i-má
ried and the noun ex-wife). ‘tree-m (tall tree)’ -yı̀i-hí ‘tree-f (small tree)’
(Heine 1982: 191). In Swahili, noun class 5
4.8. Denominal derivation comprises paired and diverse other items,
Both nominal and denominal derivation are while class 7 comprises artefacts and other
extensively developed in many languages. objects. In derivation, class 5 creates aug-
While nominal derivation is treated in 6, we mentatives [e.g. m-tu ‘cl1-man (person)’ ⫺
will here survey a couple of denominal pro- ji-tu ‘cl5-man (giant)’], while class 7 creates
cesses. Languages such as Latin can derive diminutives [e.g. ki-tu ‘cl7-man (dwarf)’].
stems of all classes productively from nomi- We now turn to the semantically exocen-
nal stems. From the base milit- ‘soldier’, the tric type of denominal derivation of nouns
noun milit-ia ‘military service’, the adjective and adjectives. These derivations center
milit-aris ‘military’ and the verb milit-are around the notion of possession in the broad-
‘serve as a soldier’ are derived. And from the est sense. If X is the meaning of the base,
noun gutta ‘drop’ the adverb guttatim ‘drop- then the smallest common denominator of
wise’ may be formed. the derived meanings is ‘Y related to X’. Spe-
cific kinds of relation between X and Y may
4.8.1. Derivation of nouns and adjectives be distinguished. Since these are relations be-
Assume a nominal base designating X and a tween two nominal concepts, they can, in
noun derived from it and designating Y. Then principle, be expressed by adnominal cases,
Y may be a kind of X, or it may be altogether too. It will be seen that for each specific rela-
distinct from X. The first derivation is an in- tion, there may either be a clear alternative
stance of modification and therefore semanti- between a derivational process and case
cally endocentric. In the second derivation, the marking, or the distinction may be blurred in
derivational formative is the structural head of a particular language.
the construction, which is semantically exo- A possessive adjective is one which is de-
centric. We will treat these two types in turn. rived from a noun meaning X and which
Given a nominal base meaning X, a diminu- means ‘belonging to X, related to X’. Latin
tive is a noun meaning ‘little X’, and an aug- has various suffixes in this function, among
mentative is one meaning ‘big X’, as in Italian them -ilis, as in (23), and -arius, as in funus
libr-o ‘book’ ⫺ libr-ino ‘book-dim (booklet)’ ⫺ ‘burial’ ⫺ funer-arius ‘funerary’. In Russian,
libr-one ‘book-aug (big book)’ (cf. Art. 99). the suffix -nyj, as in želez-naja doroga ‘iron-
Common connotations of diminutives include related:f road:f (railway)’, is even more
‘cute, weak, unimportant, contemptuous’, widely used, competing with the genitive. The
common connotations of augmentatives in- semantic type of the possessive adjective is
clude ‘strong, important, ugly’. Italian has also instantiated by nouns. The Latin adjec-
special suffixes for some of these meanings, tives in -arius can be substantivized, and thus
e.g. libr-accio ‘trashy book’. Similar examples the same suffix can derive nouns of the se-
can be found in Baltic and Slavic languages mantic type ‘Y related to X’. Examples are
and all over the world. Since these derivations aqua ‘water’ ⫺ aquarius ‘Water Bearer’,
are structurally and semantically endocentric, herba ‘herb’ ⫺ herbarium ‘container of
they should be recursive, and in some lan- herbs’. This formation becomes highly pro-
guages they are. Thus, in Italian we have canna ductive, yielding, among others, the Ger-
‘tube’ ⫺ cannello ‘little tube’ ⫺ cannellone manic suffix -er as in potter (lit. ‘someone re-
‘tube-dim-aug (big little tube)’ (type of lated to pots’).
(23) domus eri / erilis
house:nom.sg master:gen.sg master:poss.adj:nom.sg
‘master’s house’
(24) dibirdibi-karran-ju dulk-u
Rock.Cod-gen-prop place-prop
‘with Rock Cod’s place’ (Evans 1995: 151)
Possessive adjectives are related to possessive achieved by converting the noun stem into a
noun attributes formed with the genitive, verb stem without any morphological means
both within a language and cross-linguisti- (cf. Art. 90). Here are some types (cf.
cally. For instance, Latin has the alternative Fleischer 21971: 288f.): X is subject comple-
of (23). In several Australian languages, the ment: Spitzel ‘spy’ ⫺ spitzeln ‘to spy’. X is
genitive is like derivational categories in al- object complement: Knecht ‘servant’ ⫺
lowing the addition of another case suffix, as knechten ‘reduce to servitude’; X is an (ef-
in (24) from Kayardild (Pama-Nyungan). fected) object: Knospe ‘bud’ ⫺ knospen ‘to
The semantic converse of the possessive bud’; X is an adverbial: Bürste ‘brush’ ⫺
noun or adjective is the type ‘Y having X, Y bürsten ‘to brush’. A subtype of the last-men-
provided with X’, as in Latin barba ‘beard’ tioned type are ornative verbs, which follow
⫺ barbatus ‘bearded’. If X is a body part, the pattern ‘to provide Y with X’ : Sattel
then the derivation normally presupposes ‘saddle’ ⫺ satteln ‘to saddle’. Their negative
some modification of X, as in long-legged, counterpart are privative verbs like Kopf
blue-eyed. Just as one normally does not say ‘head’ ⫺ köpfen ‘behead’.
‘the girl has legs’, there is also no derivation Denominal adverbs may be analyzed as an
legged (girl). This derivation, too, has a case adverbial case form of the base X, e.g. ‘in X’.
form corresponding to it: the proprietive. (25) German has an adverbializing -s suffix in this
is again from Kayardild (cf. also (24)). function, as in Morgen ‘morning’ ⫺ morgens
‘in the morning’, Anfang ‘start’ ⫺ anfangs ‘in
(25) dun-kuru-ya maku-y the beginning’. In analogous fashion, adposi-
husband-prop-loc woman-loc tions and conjunctions may be formed, e.g.
‘near a married woman’ (Evans 1995: Zweck ‘purpose’ ⫺ zwecks ‘for’, Fall ‘case’ ⫺
146) falls ‘if’.
The proprietive has a negative counter- 4.9. Distribution of nominal categories
part: a privative adjective is one meaning
‘lacking X, without X’. In English, such ad- It was mentioned on several occasions that a
jectives are derived with the suffix -less, as in concept that is expressed as a nominal cate-
hairless, verbless. In Kayardild, this is again gory may also be marked on other word
expressed by a case, to be seen in (26). The classes. This may be so either because such a
paradigmatic relationship of proprietive and grammatical category is selected indepen-
privative derivation is brought out by (27) dently for other word classes or because it is
from Mangarayi (non-Pama-Nyungan, Aus- copied on them by agreement. In general,
tralia). most of the nominal categories mentioned
may be marked on the noun s. l. In fact, as
(26) dangka-warri-wu dulk-u far as word classes (as opposed to nominal
person-priv-prop place-prop groups) are concerned, the functional locus
‘with uninhabitated places’ (Evans 1995: of several nominal categories, such as nomi-
158) nal class, determination and person, is not
(27) (a) ña-mawuj-(j)i the noun, but the pronoun. For more on this,
2.sg-food-prop see 5. Some grammatical categories are
‘you have got food’ shared between the noun s. l. and the verb.
These include the ones just mentioned plus
(b) ña-mi-wi number/collection and tense. Those cate-
2.sg-food-priv gories that are really pronominal (rather than
‘you have no food’ (Merlan 1982: nominal) are not shared between noun and
73) verb by virtue of some commonality of these
word classes. Rather, the noun carries these
4.8.2. Derivation of other parts of speech
categories to the extent it behaves like a pro-
A verb derived from another part of speech noun, while the verb carries these categories
consists of a base and a derivational element by virtue of carrying pronominal indices.
whose meaning is some verbal archisememe Number/collection covers a class of concepts
A such as ‘be’, ‘become’, ‘make’, ‘act’. Given that take different shapes on nouns and on
a nominal base meaning X, the derived verb verbs. Finally, tense is marked on nouns only
designates a situation whose core is A and in exceptionally. In sum, there are, at a univer-
which X is a participant. The semantic types sal level, no morphological categories that
of denominal verb derivation may then be are exclusive to nouns proper. A few, above
differentiated by the participant role of X. In all case and determination, appear on verbs
German, all of these verbalizations may be only to the extent that pronominal indices on
the verb may show them, but otherwise mark man ‘one’ is a case in point; it cannot be
the noun s. l. off from the verb. To this ex- used in the oblique cases and in their cor-
tent, typology confirms the approach to the responding syntactic functions. (In the
definition of the concept ‘noun’ taken by Di- latter, forms of einer ‘one’ have to be
onysios Thrax (cf. 1). used; but this is not an instance of sup-
There is a systematic dependency between pletion, since einer has a nominative of
verbal and nominal morphology: if a lan- its own.)
guage has nominal inflection, it has verbal in- (b) Zero marking: A grammatical category is
flection. Most languages have morphology in not marked on a certain noun or class of
both categories. Japanese is an example of a nouns (cf. Art. 45). This presupposes that
language with verbal, but without nominal the noun can be used in syntactic envi-
inflection. Thirdly, there are languages with- ronments that require the category in
out inflection, such as Thai, Burmese, Viet- question but the category is not ex-
namese and Yoruba. However, the fourth pressed. In French, large groups of nouns
logical type ⫺ languages with nominal inflec- have no number marking (even if one
tion but no verbal inflection ⫺ has not been takes liaison into account). What is or-
evidenced. For a similar dependency between thographically classe ‘class’, plural
pronominal and nominal inflection, see 5. classes, is phonologically /klas/ in both
Languages differ in the nominal categories numbers. In German, certain classes of
that they possess and in the extent to which feminine nouns have no case marking;
they develop subcategories of these cate- e.g. Frau ‘woman’, plural Frauen, in
gories. One systematic dependency seems to either number has the same form in all
hold for gender and number: If a language four cases [cf. also (9)]. This kind of fail-
has gender, then it has number (Greenberg ure to undergo inflection is frequent in
1963: 쒙36). Similarly, it seems that if a lan- loans. German Epos, plural Epen, which
guage has case, it also has number. In fact, is neuter and has no case marking, is a
number appears to be the most widespread typical instance.
nominal category. No implicational relation- In cases of defectivity, we say that the
ships involving possessive marking are noun in question does not have the category
known. in question. In cases of zero marking, one
The above observations involve the distri- would hesitate to propose this diagnosis. Al-
bution of nominal categories over word though it seems weird to distinguish, for a
classes and their subclasses. The question of noun like Frau, a paradigm of four cases
how nominal categories can co-occur in one none of which may be seen or heard, the syn-
noun form will be taken up in 6.5. tax treats such a noun just like one on which
4.10. Incomplete distribution of nominal the cases are marked. Nevertheless, it is
categories worth noting that if a grammatical category
gets lost altogether in diachronic change, this
Frequently, an inflectional category does not does not happen because defectivity spreads
apply equally to all the members of a given over the word class in question, but because
word class. Such lack of generality may range zero marking spreads. For instance, English
from some erratic exceptions to whole sub- and the Romance languages lost the category
classes of the word class in question. With of case not because fewer and fewer nouns
respect to nouns, two cases have to be distin- could be used in syntactic environments
guished (analogous considerations apply to which required oblique cases, but because
other word classes): fewer and fewer nouns were marked for case.
(a) Defectivity: A grammatical category is From this perspective, it does seem correct to
not available for a certain noun or class say that Frau has no case.
of nouns (cf. Art. 67). This means that
the noun cannot be used in a syntactic 5. Major subclasses
environment which would require this
category. For the category of number, Just like word classes themselves, their sub-
this is the situation for the singularia and classes may be grammatical and even mor-
pluralia tantum seen in 4.2, which can phological classes or may be recognizable
combine with articles or predicates of only on semantic grounds. In this section, we
only one of the numbers. For the cate- deal with the major classes of nouns that are
gory of case, this kind of restriction is manifested at least syntactically, if not mor-
rarer. In German, the generic pronoun phologically, in most languages. Classes
formed by such language-specific classifica- Latin and Russian where determiners are op-
tion systems as gender and noun class have tional. Moreover, pronouns can be substi-
been discussed in 4. Most of the distinctions tutes for nouns only if they are independent
to be reviewed here are based on the empathy words. Many languages have pronominal
hierarchy of Fig. 73.2. clitics or affixes which are, of course, not in
In many languages, the category of the a distribution class with nouns, but often
noun proper belongs to the supercategory of may co-occur with them.
the noun s. l., which commonly includes In most languages, the morphology of
nouns s. s., adjectives, numerals and pro- pronouns and nouns proper differs markedly.
nouns (cf. 1). In languages such as Latin, all In any given language, pronouns are at least
of these word classes are morphologically as richly inflected as nouns. In particular, if
alike in that they inflect for gender, number a language has gender or number or case or
and case. Moreover, they are syntactically any combination of these in the noun, then it
alike in some ways, for example, that a word has the same categories in the pronoun
of any of these classes can function as a noun (Greenberg 1963: 쒙43; Plank 1989: 298). For
phrase. Nouns and adjectives, in particular, instance, English and the Romance languages
form a natural class in many languages (cf. have case in the pronoun, but not in the
Art. 74). In Quechua, the grammatical poten- noun. This relationship is related to the ef-
tial of adjectives includes that of nouns. They fects of the empathy hierarchy, since pro-
differ by only two grammatical features: nouns essentially include person pronouns,
First, if an adjective and a noun are com- which refer to speech act participants, which
bined in an attributive construction, then the in turn occupy the top of the hierarchy (see
former precedes the latter. Second, adjectives, Art. 76 for the internal differentiation of the
but not nouns, undergo an inchoative deriva- class of pronouns). The universal is true for
tion. Apart from these two differences, adjec- the categories of nominal class, number and
tives and nouns share all their grammatical case at the generic level. Within the category
properties (cf. Schachter 1985: 17f.): they in- of number, the dual is special in that there
flect for number and case, they may consti- are a few languages, including Hopi, which
tute referential noun phrases (there is no pro- have dual in nouns but not in pronouns
cess of substantivization of adjectives), and (Plank 1989: 297f.). Nouns and pronouns
they require a copula if used as a predicate. may also differ in their declension class. The
Even more languages have derivational pro-
ancient Indo-European languages have a
cesses which apply to substantival and adjec-
pronominal inflection pattern which does not
tival bases alike. For example, in English,
appear on nouns. For instance, Latin has a
two different derivational affixes, -hood and
genitive singular allomorph -ius in pronouns
-ness, provide for abstract nouns depending
on whether the stem is adjective or noun, but (e.g. qui ‘who’ ⫺ gen. cuius, alter ‘the other’
in Persian, the same suffix -i can be used in gen. alterius), which no noun has.
both cases: mard-i ‘manhood’ and bozorg-i Lexical nouns are further subdivided into
‘greatness’ (Windfuhr 1989: 531). Similarly, proper nouns and common nouns. Proper
the Hungarian suffix -ság/-ség deriving ab- nouns are names, i.e. nouns whose designa-
stract nouns can join both nominal and ad- tion is not a concept, but an individual. Any
jectival stems: ember-ség ‘man-abstr (hu- other noun is a common noun or nomen appel-
manness)’ and szép-ség ‘beautiful-abstr lativum. The term nomen proprium was moti-
(beauty)’. The same goes for Hungarian -ta- vated by the ambiguity of the term nomen
lan/-telen ‘without’. noted in 1. Proper and common nouns often
Successively narrowing down on nouns belong to different distribution classes such
proper, the first subdivision is between lexical that the distribution of common nouns usu-
nouns and pronouns (see Art. 76 for the mor- ally includes the distribution of proper
phology of pronouns). By its name, a pro- nouns. First, names are semantically definite.
noun should be a substitute for a noun. How- Therefore, restrictions on the combination of
ever, since most pronouns incorporate cate- proper nouns with articles are common. A
gories of determination (definiteness, speci- proper noun does not generally combine with
ficity etc.), they are actually substitutes for the indefinite article (*a Linda); and if it does,
noun phrases. Insofar, a pronoun is a substi- it is thereby turned into a common one.
tute for a noun only if the noun can form a Proper nouns may require a definite article
noun phrase by itself. This, in turn, is true (e.g. English the Alps) or they may admit no
for proper nouns in most languages (see be- article (e.g. Linda). Second, proper nouns are
low) and for all nouns in such languages as generally defective for number (*the Alp,
*Lindas), and they undergo fewer deriva- (property, action, event etc.) itself while dis-
tional processes (mountainless vs. *alpless). regarding (“abstracting away from”) its argu-
Common nouns are subdivided into count ments. The grammatical/derivational opera-
vs. mass nouns, designating individual, sepa- tion that achieves this is the formation of ac-
rate objects and continuous substances, tion nouns by nominalization. A prototypical
respectively. A count or countable noun is one abstract noun is an action noun formed by
which combines with the indefinite article nominalization. All nouns which are gram-
and with numerals, such as a girl, two cook- matically or semantically like the prototypi-
ies. A mass noun does not so combine; cf. *a cal ones are abstract nouns. A concrete noun
milk, *two airs. If mass nouns can be plural- is then simply a common noun which is not
ized, the result is a sortal plural, as in three abstract in the sense defined. Grammatical
wines. Otherwise, a numeral must first be properties that abstract nouns have in com-
combined with a mensurative (or mensural mon include the same as for mass nouns:
classifier; cf. Art. 101), and the resulting They may occur without an article (cf. Linda
phrase can then combine with a mass noun, worked on the file / on grammaticalization)
as in three bottles of wine. Moreover, mass and hardly combine with the indefinite article
nouns may occur without an article in and with numerals (*a grammaticalization,
contexts in which count nouns may not: *three kindnesses). Thus, on grammatical
Linda ate butter/*cookie. Nevertheless, these grounds, abstract nouns may be a subclass of
classes do not have distinctive morphological mass nouns. This is motivated by the fact
properties in many languages. that they do not designate individual objects.
Common nouns are also subdivided into On the other hand, and in contrast with mass
concrete vs. abstract nouns. Girl, apple and nouns, abstract nouns often do have mor-
milk are concrete, time, grammaticalization phological characteristics, insofar as most of
and kindness are abstract. A semantic defini- them are formed by nominalization. Tab.
tion of this distinction presupposes a notion 73.6 presents some derived abstract nouns of
of abstraction, the operation of creating an Latin (in the genitive, for the sake of mor-
abstract concept. The standard way of doing phological clarity) from different base cate-
this is to orient the concept to the situation gories.
Other base categories, such as adverbs and may be delimited on grammatical grounds.
numerals, play a negligible role in nominal For instance, nouns like bunch and flock are
derivation. If a language has denominal deri- count nouns which form a specific possessive
vation, it has nominal derivation. Burmese is attribute construction, while police shares
a language which has verb-to-noun deriva- some features with mass nouns. In this area,
tion but, apparently (Wheatley 1989: 849), no many distinctions can be made (cf. Leech &
productive pattern of denominal verb deriva- Svartvik 21994: 39ff., Collins COBUILD
tion (cf. also Hopper & Thompson 1984: English Grammar 1990). Nouns of measure
737f.). It is possible that nominalization is may grammaticalize to mensuratives, unit
the most important derivational operation nouns may grammaticalize to numeral classi-
at all. fiers. Such classes are, however, of no known
Concrete nouns may be further subdivided morphological relevance.
into collective and individual nouns. A collec- Concrete nouns, whether individual or col-
tive noun designates a collection of similar lective, are subdivided into animate and inan-
entities which has a quality of its own, i.e. it imate nouns, and the former into human and
is more than a set of such entities. Examples nonhuman. Some aspects of grammatical
are bunch (of flowers), flock (of sheep), but structure in perhaps all languages are sensi-
also police. Some languages have productive tive to these distinctions. With certain well-
processes for the derivation of collective motivated exceptions, the principle is that
nouns (cf. 4.2). In English, subclasses of them within any of the nominal categories of 4,
more differentiation is found in nouns which cally or lexically conditioned. If it is, then the
are higher up on the empathy hierarchy. conditioning environment is usually one of
Thus, several Australian languages have a the subcategories of nouns reviewed in the
separate accusative only for nouns from a present section. For instance, allomorphy in
given position of Fig. 73.2 upward, e.g. only possessive marking may be conditioned by
for human noun phrases (Arabana) or only the alienable or inalienable character of the
for animate noun phrases (Thargari). While noun. Number marking is another example:
these are predominantly ergative languages, it is often different for human or animate
something similar is true for the (accusative) nouns as opposed to non-human or inani-
ancient Indo-European languages (cf. Tab. mate ones. For example, Latin and Greek
73.2), which syncretize the accusative with have allomorphy in the plural of masculine
the nominative in neuter nouns. Again, sev- and feminine nouns, but only the form -a for
eral languages mark number only on nouns neuter plural, which is not among the plural
from a given position of Fig. 73.2 upward. allomorphs of the other genders. The same
For example, Mandarin Chinese has obliga- subclasses which condition the applicability
tory number in pronouns, optional number of a nominal category in one language may
in human nouns and no number in nouns of condition its allomorphy in another lan-
less empathy.
guage. Cf. also 4.10 for the conditioning of
A stem which contains at least one posi-
zero allomorphs.
tion for a governed argument is called rela-
tional; otherwise, it is absolute. In this sense,
all verbs except the avalent ones are rela-
tional. Among nouns, those designating kin 6. Morphological structure
(like sister), body parts (hand), personal attri-
While the noun shares much of its morpho-
butes (name) and spatial regions (top) as well
as certain verbal nouns such as nomination are logical structure with words of other classes,
relational, while nouns designating physical the following subsections concentrate on
objects such as apple, woman are absolute. It those morphological properties which mark
will be seen that prototypical nouns as de- the noun off from other word classes. There
fined in 2 are absolute nouns, while relational are distinct processes of stem formation
nouns are more verb-like. Apart from differ- which have the noun as their target; and
ing syntactically from absolute nouns, rela- there are distinct processes of stem formation
tional nouns are often marked off from abso- and inflection which presuppose the noun as
lute nouns at the morphological level, too. their base. A given type of grammatical con-
The relevant facts are reviewed in 4.4. cept may be expressed by diverse kinds of
The marking of nominal categories is often morphological processes. Tab. 73.7 illustrates
subject to allomorphy that is morphologi- this for sex/gender.
In the following subsection, the levels of 6.1. Morphological levels in the noun
word structure in the noun are introduced. If ‘noun’ is taken as a syntactic (distribu-
Sections 6.2⫺6.4 review the formal processes tional) category, then nominals or phrasal
operative at the levels of compounding, deri- nouns and even nominalized clauses will
vation, inflection and theme formation. Fi- count as nouns. Following common practice,
nally, the mutual relationship of diverse the concept is restricted here to the word
markings on a stem is considered. level. As with other inflecting word classes,
(1) (a) word form J word form ⫹ word form periphrastic inflection
(b) word form ⫹ inflectional element synthetic inflection
(c) stem
(2) (a) stem J stem ⫹ stem compounding
(b) stem ⫹ derivational element derivation
(c) root
Again, a process of phrasal compounding pound, where one stem serves as the base and
could be posited between synthetic inflection is determined by the other one, is the most
and compounding. ‘Inflectional element’, common type in the languages of the world,
‘derivational element’ and ‘root’ are terminal, probably because it is in a virtually regular
i.e. unexpandable units. The rules apply in paradigmatic relationship with the syntactic
the usual way: Start with the unit ‘word construction of possessive attribution. For
form’ and apply expansion rules until only instance, there is a regular relationship ⫺
terminal units are left. Tab. 73.8 is, in prin- which has often been analyzed by means of
ciple, valid for any word class. Specialties of transformations ⫺ between linguistics student
nouns will be mentioned below. and student of linguistics. Examples of this
Tab. 73.8 accounts for the morphological type of compound may also be cited from
complexity that may be found in nouns. In Mandarin (Li & Thompson 1981: 48ff.): chu-
particular, since compound and derived ang-dan(zi) ‘bed sheet’, jiu-bei(zi) ‘wine
stems may be formed from compound and cup’. In these languages, the order of the
derived bases, nominal stems may get as stems is ‘determinans-determinatum’. The
complex as the English word decompartmen- opposite order is common in languages that
talization. Sanskrit is famous for its morpho- have had head-dependent order for a long
logical complexity in the nominal sphere. time. Yucatec Maya has chúumuk-k1ı̀in ‘mid-
However, such examples stand out against dle-day (noon)’, éet-kàahal ‘companion-place
the majority of the languages of the world, (compatriot)’, táan-ho1l ‘front-head (fore-
where morphological complexity is more de- head)’. Tagalog (Schachter & Otanes 1972:
veloped in the verbal sphere. 107ff.) has mata-ng-lawin ‘eye-linker-hawk
The alternation among morphological (keen eyes)’, bata-ng-lasangan ‘child-linker-
variants of noun stems is treated in 4.3. The
street (homeless child)’.
operations for the formation of entity con-
The determinans of a denominal nominal
cepts of various classes and, thus, of nouns,
compound may also belong to another cate-
are treated in Art. 94. Here the various for-
mal processes of producing a noun stem are gory, as in blackbird, washing machine and
discussed. jack-in-the-box. Further types of nominal
compounding may be distinguished by di-
6.2. Compounding verse criteria. First, the construction need not
Compounding as a morphological process be semantically endocentric. It is not, for in-
(by rule (2 a) of Tab. 73.8; cf. Art. 87) is most stance, in English redcap ‘suitcase carrier’ or
productive in nominal morphology. Ger- pick-pocket. Second, there does not need to
manic languages abound in compounds of be a nominal base in order to compose a
arbitrary complexity such as English 1997 noun stem. There is none in English see-saw
physics nobel prize winner or German Turm- and Mandarin kai-guan ‘open close (switch)’.
uhr-zeiger ‘tower-clock-pointer (hand of stee- Compound nouns constitute a specific for-
ple clock)’. The determinative nominal com- mal subclass of nouns, but there is, among
all the designata of nouns, no specific kind of i.e. nouns of multitude, have been seen in 4.2.
entities designated by compound nouns, wit- Words expressing case are adpositions. For
ness such cross-linguistic synonyms as eye- possession, we have possessive pronouns,
brow ⫽ French sourcil, fingernail ⫽ French attributors like of and possessive classifiers.
ongle etc. (cf. Andersen 1978). Grammatical words expressing determina-
tion and person are, of course, determiners
6.3. Processes of inflection and derivation and personal pronouns. By grammaticaliza-
The notion of periphrasis (see Art. 68) has oc- tion, all such grammatical words may de-
casionally been extended to noun inflection. velop into nominal inflections.
Just like Italian è venuto ‘has come’ is peri- Synthetic inflection of nouns may involve
any of the formal processes treated in
phrastic with respect to Latin uēnit, so Italian
ch. VIII. Affixation is most prominent. All of
del lago ‘of the lake’ may be called periphras-
the nominal categories mentioned in 4 may
tic with respect to Latin lacūs. However,
be expressed by suffixes. Nominal classes
there is a crucial difference in that è venuto may also be expressed by prefixes, e.g. in
consists of two verb forms, while del lago Bantu languages. Here, nominal number
does not consist of two noun forms. There fuses with the category of noun class. Case
are, in fact, no periphrastic noun forms in prefixes are exceedingly rare (Sanders 1978;
this narrow sense. Therefore, the notion of Hawkins & Gilligan 1988). They do occur in
periphrastic inflection as defined by rule (1 a) some Semitic and Bantu languages (cf. Hetz-
in Tab. 73.8 is not applicable to nouns. How- ron 1980: 278) and in Mangarayi, where they
ever, several of the nominal categories men- are fused with gender and combined with
tioned in 4 may be expressed by grammatical case suffixes, so that the case system might
words. Grammatical words which express a also be analyzed as consisting of circumfixes.
nominal class are classifiers as mentioned in Tab. 73.9 shows a major fragment of the de-
4.1. Words indicating plurality or a collective, clension paradigm.
accusative N nan- N
nominative nj a- N nalja- N
genitive/dative nj a- N -w1u naya- N
locative nj a- N -y1an naya- N -y1an
allative N -(ga)ljama naya- N -(ga)ljama
ablative N -w1ana naya- N -w1ana
Infixation (cf. Art. 55) is chiefly found in Here it suffices to mention that it also figures
verbal morphology and only very rarely ap- among the various processes of plural forma-
plies to noun stems. In Miskito (Misumalpan, tion in Arabic. Plurals formed by transfixa-
Central America), possessive affixes differ ac- tion, called broken plurals, are especially
cording to the alienability of the noun. In- common in non-human nouns. In each dia-
alienable nouns have possessive infixes of lect, there are more than a dozen different
patterns of broken plural formation. Tab.
first and second person possessors (the third
73.10 shows some examples from Gulf Arabic
is marked by a prefix), as in napa ‘tooth’ ⫺ (Holes 1990: 150⫺154):
nampa ‘your tooth’. The relative rarity of no-
minal infixation is probably a computable meaning singular plural
consequence of the overall rarity of infixation
and the relatively low average complexity of visitor zaayir zuwwaar
nominal (as compared to verbal) morphol- book kitaab kutub
house beet buyuut
ogy. mountain jabal jibaal
Transfixation and its importance in Semitic
languages is treated more fully in Art. 56. Tab. 73.10: Gulf Arabic broken plural formation
In nominal morphology, reduplication is nary nouns form a plural in galja, like malam-
mostly used for plural formation. Sumerian galja ‘man-pl (men)’. However, the plural of
has plural suffixes, but also total reduplica- gadj ugu ‘woman’ is gadj u-galja, with subtrac-
tion, as in kur ‘mountain’, kur-kur ‘moun- tion of the final stem syllable. Plural forma-
tains’. Mangarayi (Merlan 1982: § 2.1.1.8), tion of social subsection terms involves a suf-
too, uses mainly suffixing, but also has vari- fix -w2u plus various stem alternations, one of
ous forms of reduplication. Total reduplica- which is subtraction. Thus jamijin (a section),
tion is seen in bugbugbug ‘old people’. For plural jamij-bu, narijbalan, plural narijban-bu.
most nouns, reduplication is limited to their Apart from these formal processes, nouns
use as the basis for proprietive derivation, and may also be paradigmatically related to stems
there it is often partial reduplication, as in ma- of other categories by conversion and supple-
lam ‘man’, malalam-yi ‘having husbands’. tion. Conversion (or polycategoriality) is
Elsewhere, reduplication is also used as a pro-
common in English, as in dance, go, or fit.
cess of nominal derivation, e.g. Malay lamit
Suppletion is the general pattern in Kunjen,
‘sky’, lamit-lamit ‘cloth canope’, ‘palate of the
mouth’; Ewe fo ‘beat’, fo-fo ‘beating’. an Australian language of North Queenland.
The employment of internal modification Cf. egna ‘dance’ (verb), odnden ‘dance’
(cf. Art. 58) in nominal morphology is well- (noun) (Sommer 1972: 74).
known from Indo-European languages. Apo- As mentioned repeatedly, most languages
phony was exemplified in Tab. 73.3. Meta- use several of these processes in nominal
phony (German Umlaut) is common in Ger- morphology. Not infrequently, two different
man inflection. In plural formation, it may processes co-occur in one word form, and
be the sole mark of the plural, as in Mutter even as the (discontinuous) significans of one
‘mother’, plural Mütter, or it may co-occur grammatical meaning. Among all the dif-
with suffixation, as in Sohn ‘son’, plural ferent techniques, affixation is dominant.
Söhne. Other operations of nominal mor- While total reduplication may be found ⫺ in
phology involving metaphony include dimi- isolating languages ⫺ as the sole process of
nution, where the diminutive suffixes -lein nominal derivation and inflection, the use of
and -chen trigger metaphony, as in Mütter- all the other processes of nominal derivation
lein, Söhnchen. Consonant mutation is opera- and inflection in a language implies the use
tive, although again as a concomitant fea- of affixation.
ture, in English plural formation of the type
wolf ⫺ wolves. While metathesis does not ap- 6.4. Theme formation
pear as a productive process in nominal mor-
phology, all kinds of suprasegmental processes If no more rules of stem formation (number
may be employed. Accent shift occurs in Eng- 2 of Tab. 73.8) are applied, we should get to
lish nominalizations of the type contrást ⫺ the level of the root. It is, however, frequently
cóntrast, addréss ⫺ áddress. Tonal changes ex- the case that an elementary noun stem con-
pressing case are reported from Turkana sists of a root plus a stem-forming submor-
(Dimmendaal 1983: ch. 5.3). Vowel lengthen- phemic unit. Many Latin nouns, for instance,
ing in the same function was seen in (7). have the general structure displayed in the
Subtraction of final syllables in declension head row of Tab. 73.11 and exemplified with
occurs in Mangarayi plural formation. Ordi- nouns from three different declension classes.
significans turr -i -s
significatum tower N nom.sg.
significans mens -a 0
significatum tabel N nom.sg.
significans duc 0 -s
significatum leader 0 nom.sg.
together form the desinence. For stems of phology and sometimes not, and this varia-
such a structure, one may make an argument tion is both language-internal and cross-lin-
to the effect that the root signifies a pure, un- guistic.
categorized concept, while the significatum of
the thematic element is nothing but a cate- 6.5. Order of stem and affixes
gory, which it confers to the concept. For One question here concerns the position of
Latin, the argument would suffer from the an affixal morphological category relative to
extensive homonymy between noun-forming the stem. Some preferences concerning the
and verb-forming thematic vowels, as all of various kinds of affixes are stated in 6.3. A
the conjugation vowels have counterparts number of typological correlations exist be-
among the declension vowels. For Hausa, the tween inflectional affix order and syntactic
case would be much clearer (cf. Greenberg order (Hawkins & Gilligan 1988). These are
1978: § 5.3). Here, virtually all nouns end in summarized in Tab. 73.12.
a long vowel. Diachronically, this goes back
to a definite article, but synchronically, final implicans implicatum
vowel length is the significans of the gram-
matical meaning ‘noun’ (N in Tab. 73.11). Gender Affix N V O
In Latin, not all nouns have a thematic Gender Affix N Adp NP
vowel. On the one hand, a noun may have a NP Adp N Gender Affix
productive or unproductive derivational suf- SO V N Gender Affix
NP Adp N Indef Affix
fix that ends in a consonant, such as clam-or
SO V N indef Affix
‘shout-ing’. If these are discounted, the so-
called root nouns remain, words such as dux Tab. 73.12: Affix order and syntactic order
‘leader’ (in Tab. 73.11), fur ‘thief’ and lux
‘light’ which have the case/number suffix at- From this it follows that the default for
tached directly to the root. These may be an- gender and indefiniteness affixes is to be suf-
alyzed as resulting from the application of fixes. No exceptionless generalizations of this
rules 1. c and 2. c of Tab. 73.8. There are root kind have been found for definiteness and
words in every word class. One and the same number affixes. In general, nominal prefixa-
root may be restricted to one word class or tion tends to imply verbal prefixation.
may be used in more than one. The examples Another question concerns the co-occur-
just given are exclusively nominal. In such rence of markings on a single stem. Different
cases, the root itself is word-class specific. formal processes may apply to a given noun
Roots which are used in more than one word stem; no constraints on their combination are
class are frequent in German, e.g. Lauf known. Both agglutination and fusion of
(noun) ⫺ laufen (verb) ‘run’, krach (ideo- morphological categories are widespread. In
phone) ⫺ Krach (noun) ⫺ krachen (verb) Altaic languages, number and case receive
‘bang’. Refined and possibly historical analy- separate expression. Neither conditions any
sis may ascertain the direction of derivation allomorphy in the other, so this is typical ag-
in some such cases. For instance, the fact that glutinative morphology. In Niger-Kordofan-
laufen inflects as a strong verb (past lief ) may ian languages, gender and number are usu-
point to the basic character of the verb stem, ally fused in one morpheme. In ancient Indo-
while the weak inflection of krachen (past European languages, these are furthermore
krachte) argues for its derived status. This cumulated with case.
would still leave many cases unanalyzed, in- As regards the sequential order of different
cluding krach as an ideophone and as a noun. affixal categories on one stem, order con-
Here, one may once more consider that the straints could, in principle, pertain to any
root signifies an uncategorized concept. kind of formal marker relative to another
However, the word class would not be con- one or relative to the stem. In reality, how-
tributed by a distinct element, but would be ever, the only generalizations that can be as-
a purely distributional property of the root certained concern the order of an affix ex-
morpheme in question. The result of this dis- pressing some morphological category rela-
cussion is that the distinction between the tive to the stem and its relative closeness to
lexical and the grammatical portion of the the stem. That is, there is no rule of the kind
meaning of a stem, its individual vs. categor- that one inflection has to precede another
ial meaning, is sometimes reflected in mor- one (regardless of their position relative to
Fig. 73.3 is meant to represent relative has the opposite order of Hungarian (cf.
closeness of diverse categories to the stem, Comrie 1980):
but not their left vs. right ordering with re- (31) yslävä-lle-ni
spect to the stem or to each other. Naturally, friend-dat-poss.1.sg
relative ordering in the sense of Fig. 73.3 is ‘to my friend’
crucial only if the respective operators are af-
Similarly, while Turkish shows the canonical
fixed on the same side of the stem. The se- order of number and possessive, Chuvash has
mantic motivation for Fig. 73.3 cannot be these suffixes in the opposite order (Johanson
fully developed here. The following aspects 1973: 91). German has some erratic excep-
are important: No order is postulated among tions to Fig. 73.3 such as Kind-er-chen ‘child-
diverse derivational categories. However, as pl-dim (kids)’ (Dressler 1989: 8f.). Orders of
in other word classes, these are, on the whole, nominal categories that contradict the iconic
closer to the stem than inflectional cate- principle embodied in Fig. 73.3 are manifes-
gories. Of the latter, gender/noun class is tations of the fact that morphology is less
closest to the stem, since this is a lexical- iconic than syntax.
grammatical category of the stem that is pre- From the operator-operand model it also
supposed, but not changed by rules of syn- follows that there must be heavy restrictions
tax. All the following categories have the no- on the re-application of a morphological cat-
minal or even the entire noun phrase as their egory to the same stem. While this is indeed
semantic locus. The relative order of number rare, it is not excluded in general. First, a
and case is stated in Greenberg 1963, 쒙39. phenomenon which only appears to be an in-
stance of reapplication should be noted. In
Instantiations of Fig. 73.3 include Tab. 73.8
languages where the possessed noun agrees in
and Tab. 73.11. (30) from Hungarian also number and person with the possessor, pos-
contains some of the categories in the canoni- sessed nouns may end up with two number
cal order. markers, one showing the number of the pos-
sessed noun, the other the number of the pos-
(30) arc-ai-k-at
sessor, as is the case in Hungarian when the
face-pl-poss.3.pl-acc possessed noun is plural:
‘their faces (acc.)’
(32) (a) az én kabát-ai-m
However, orders that contradict Fig. 73.3 are the I coat-pl-poss.1.sg
not unheard of. Finnish, as illustrated in (31), ‘my coats’
we have seen, these differences tend to be son singular pronoun, was earlier a noun wa-
manifested in the relative closeness of the re- takusi meaning ‘privacy’ and boku, a mascu-
spective morphological categories to the line version of ‘I’, had the nominal meaning
noun stem. ‘male servant’. At the early stages of their
pronominal use, the element ‘humble’ was
8. Linguistic evolution part of their meaning. This meaning property
has dropped out later, and watakusi was
As we said in 2, the noun is a universal word shortened to watasi (Sugamoto 1989: 272f.).
class. It is not easy to tell why this should be Similarly, relational nouns such as front de-
so. Entity concepts can be co-lexicalized with velop into adpositions and finally into case
verbs, as the meaning of kick contains the affixes. Other nouns develop into classifiers
concept ‘foot’. Concrete nominal concepts of various sorts. Thus, the noun has an im-
can be formed freely by substantivization of portant function in feeding various classes of
adjectival expressions, as in the old one. This grammatical formatives.
would, in fact, allow that there be just one
concrete noun (‘one’), with all the others
9. References
formed by attribution. Abstract nominal con-
cepts can be formed with equal freedom by Andersen, Elaine. (1978), “Lexical Universals of
nominalization, as we have seen in 5. How- Body-Part Terminology”. In: Greenberg (ed.), Vol.-
ever, human language appears to be orga- III, 335⫺368
nized in such a way that the goal of such op- Anderson, Stephen R. (1985 a), “Inflectional Mor-
erations must exist in the form of ready-made phology”. In: Shopen (ed.), Vol. III, 150⫺201
Gestalten which can serve as a model. Anderson, Stephen R. (1985 b), “Typological Dis-
Nouns are relatively stable in aphasia tinctions in Word Formation”. In: Shopen (ed.),
(Dressler 1977). In the ontogeny of language, Vol. III, 3⫺56
nouns are prior to the other word classes Asher, R. E. (1982), Tamil. Amsterdam: North-
(Gentner 1982), which means they cannot be Holland (Lingua Descriptive Studies, 7)
regarded as primarily derived from adjectives Benveniste, Emile (21935), Origines de la formation
or verbs. The same can be assumed for the des noms en indo-européen. Paris: Adrien-Maison-
phylogeny of human language. The univer- neuve
sality of nouns equally entails that diachronic Bloomfield, Leonard (1962), The Menomini Lan-
change does not lead to the acquisition or guage. New Haven & London: Yale University
loss of the category ‘noun’ in a language. Press (William Dwight Whitney Linguistic Series)
What does happen, however, is the feeding Bybee, Joan L. (1985), Morphology: A Study of the
and bleeding of the class of nouns. Nominal Relation between Meaning and Form. Amsterdam:
and denominal stem formation is one of the J. Benjamins
ways in which this happens. But, in addition, Collins COBUILD English Grammar (1990), Lon-
various structures can turn into nouns or don & Glasgow: Collins
evolve from them in diachrony. Committee on Grammatical Terminology (1911),
Noun phrases can develop into nouns. A On the Terminology of Grammar. London: John
common process is for a nominal consisting Murray
of a head noun and a genitive attribute to Comrie, Bernard (1980), “The Order of Case and
develop into a compound noun. Thus, the Possessive Suffixes in Uralic Languages: An Ap-
German suffix -s appearing at the juncture of proach to the Comparative-Historical Problem”.
such compounds as Mannesmut ‘manly cour- Lingua Posnaniensis 23: 81⫺86
age’ goes back to the genitive suffix. Further- Comrie, Bernard (1981), Language Universals and
more, compound nouns can develop into de- Linguistic Typology: Syntax and Morphology. Ox-
rived nouns when the determinatum is gram- ford: B. Blackwell
maticalized to a derivational affix. Thus, Corrigan, Roberta & Eckman, Fred & Noonan,
English derived nouns of the structure X⫺Y, Michael (1989, eds.), Linguistic Categorization.
where Y 苸 {-hood, ship, -dom}, stem from Amsterdam: Benjamins (CILT, 61).
compound nouns in which X functioned as Derbyshire, Desmond C. (1979), Hixkaryana. Am-
determinans and Y as determinatum. sterdam: North-Holland (Lingua Descriptive
Again, nouns may fade out of their class Studies, 1)
by grammaticalization. Some, or perhaps all, Derbyshire, Desmond C. (1985), Hixkaryana and
Japanese personal pronouns have nominal Linguistic Typology. Dallas, Texas: Summer Insti-
origin. For example, wata(ku)si, a first per- tute of Linguistics
Dimmendaal, Gerrit J. (1983), The Turkana Lan- Hetzron, Robert (1980), “Universals of Human
guage. Dordrecht & Cinnaminson: Foris (Publica- Language: A Monteverdian Quartet in Forty-six
tions in African Languages and Linguistics, 2) Movements”. Lingua 50, 249⫺294
Dixon, Robert M. W. (1972), The Dyirbal Lan- Hockett, Charles F. (1958), A Course in Modern
guage of North Queensland. Cambridge & London: Linguistics. New York: MacMillan
Cambridge University (Cambridge Studies in Lin- Holes, Clive (1990), Gulf Arabic. London:
guistics, 9) Routledge (Croom Helm Descriptive Grammars)
Dressler, Wolfgang U. (1968), Studien zur verbalen Hopper, Paul J. & Thompson, Sandra A. (1984),
Pluralität: Iterativum, Distributivum, Durativum, “The Discourse Basis for Lexical Categories in
Intensivum in der allgemeinen Grammatik, im Latei- Universal Grammar”. Language 60, 703⫺752
nischen und Hethitischen. Wien: Hermann Böhlaus
Nachf. (Österreich. Ak. d. Wiss., Philos.-Hist. Kl., Humboldt, Wilhelm von (1994), Mexicanische
Sitzungsber., 259. Bd., 1. Abhandl.) Grammatik. Herausgegeben von Manfred Ring-
macher. Paderborn: F. Schöningh (Schöningh Wis-
Dressler, Wolfgang U. (1977), “Elements of a Poly- senschaft. W. v. Humboldt, Schriften zur Sprach-
centristic Theory of Word Formation”. Wiener Lin- wissenschaft, 3/2)
guistische Gazette 15, 13⫺32
Johanson, Lars. (1973), “Sprachbau und Inhalts-
Dressler, Wolfgang U. (1989), “Prototypical Dif- syntax am Beispiel des Türkischen”. Orientalia
ferences between Inflection and Derivation”. Suecana 22, 84⫺106
Zeitschrift für Phonetik, Sprachwissenschaft und
Kommunikationsforschung 42: 3⫺10. Kölver, Ulrike (1982), “Interaktion von nominalen
Kategorien am Beispiel der Entwicklung des mo-
Evans, Nicholas D. (1995), A Grammar of Kayar- dernen Bengali”. In: Seiler & Lehmann (eds.),
dild: With Historical-Comparative Notes on Tang- 244⫺251
kic. Berlin: Mouton de Gruyter (Mouton Grammar
Library, 15) Kuno, Susumu & Kaburaki, E. (1977), “Empathy
and Syntax”. Linguistic Inquiry 8: 627⫺672
Fleischer, Wolfgang (21971), Wortbildung der
deutschen Gegenwartssprache. Tübingen: M. Nie- Landaburu, Jon (1979), La langue des Andoke:
meyer [2. unveränd. Aufl.; 11969, Leipzig: VEB Amazonie colombienne. Paris: SELAF (Langues et
Verlag Enzyklopädie] Civilisations à Tradition Orale, 36)
Gentner, Dedre (1982), “Why Nouns are Learned Leech, Geoffrey N. & Svartvik, Jan (21994), A
before Verbs: Linguistic Relativity versus Natural Communicative Grammar of English. London:
Partitioning”. In: Kuczaj II., S. (ed.). Language de- Longman
velopment. Vol. II: Language, Thought, and Culture. Lehmann, Christian (1985), “On Grammatical Re-
Hillsdale, NJ: Lawrence Erlbaum Associates, lationality”. Folia Linguistica 19: 67⫺109
301⫺334 Lehmann, Christian (1988), “On the Function of
Gil, David (1991), Nouns, Verbs, and Quantifica- Agreement”. Barlow, Michael & Ferguson, Charles
tion. Strasbourg: European Science Foundation A. (eds.), Agreement in Natural Language: Ap-
(EUROTYP Working Paper, series 10, number 1.) proaches, Theories, Descriptions. Stanford: Center
Givón, Talmy (1979), On Understanding Grammar. for the Study of Language and Information,
New York etc.: Academic Press (Perspectives in 55⫺65
Neurolinguistics and Psycholinguistics) Li, Charles & Thompson, Sandra (1981), Mandarin
Givón, Talmy (1984), Syntax: A Functional-Typo- Chinese: A Functional Reference Grammar. Berke-
logical Introduction, Vol. I. Amsterdam & Philadel- ley: University of California Press
phia: J. Benjamins Manzelli, Gianguido (1990), “Possessive Adnomi-
Greenberg, Joseph H. (1963), “Some Universals of nal Modifiers”. In: Bechert, Johannes & Bernini,
Grammar with Particular Reference to the Order Giuliano & Buridant, Claude (eds.), Toward a Ty-
of Meaningful Elements”. In: Greenberg, Joseph pology of European Languages. Berlin & New York:
H. (ed.), Universals of language: Report of a confer- Mouton de Gruyter (Empirical Approaches to
ence held at Dobbs Ferry, New York, April 13⫺15, Language Typology, 8), 63⫺111
1961. Cambridge, Mass.: MIT Press, 58⫺90 Merlan, Francesca (1982), Mangarayi. Amsterdam:
Greenberg, Joseph H. (1978), “How Does a Lan- North Holland (Lingua Descriptive Studies, 4)
guage Acquire Gender Markers?”. In: Greenberg Mithun, Marianne (1988), “Lexical Categories and
(ed.), Vol. III: 47⫺82 Number Marking”. In: Hammond, Michael &
Greenberg, Joseph H. (1978, ed.), Universals of Noonan, Michael (eds.), Theoretical Morphology.
Human Language, Vol. 1⫺4. Stanford, Cal.: Stan- Approaches in Modern Linguistics. San Diego: Aca-
ford University demic Press, 211⫺234
Hawkins, John A. & Gilligan, Gary (1988), “Pre- Plank, Frans (1989), “On Humboldt on the Dual”.
fixing and Suffixing Universals in Relation to Basic In: Corrigan et al. (eds.), 293⫺333
Word Order”. Lingua 74, 219⫺259 Plank, Frans (1995, ed.), Double Case: Agreement
Heine, Bernd (1982), “African Noun Class Sys- by Suffixaufnahme. New York etc.: Oxford Univer-
tems”. In: Seiler & Lehmann (eds.), 189⫺216 sity Press
Premper, Waldfried (1986), Kollektion im Ara- Smith-Stark, Thomas C. (1974), “The Plurality
bischen. Köln: Institut für Sprachwissenschaft der Split”. Papers from the Tenth Regional Meeting of
Universität (Arbeitspapier Nr. 49) the Chicago Linguistic Society, Chicago, IL: Chi-
Sanders, Gerald A. (1978), “Adverbial Construc- cago Linguistic Society, 657⫺671
tions”. In: Greenberg (ed.), Vol. IV, 51⫺84 Sommer, Bruce A. (1972), Kunjen Syntax. Can-
berra: Australian Institute of Aboriginal Studies
Schachter, Paul (1985), “Parts-of-speech Systems”.
Shopen (ed.), Vol. I, 3⫺61 Sugamoto, N. (1989), “Pronominality: a Noun-
Pronoun Continuum”. In: Corrigan et al. (eds.),
Schachter, Paul & Otanes, Fe T. (1972), Tagalog
267⫺291
Reference Grammar. Berkeley: University of Cali-
fornia Press Wheatley, Julian K. (1989), “Burmese”. In: Com-
rie, Bernard (ed.), The World’s Major Languages.
Seiler, Hansjakob (1983), POSSESSION as an Op- London: Routledge, 834⫺854
erational Dimension of Language. Tübingen:
G. Narr (LUS, 2) Werner, Otmar (1979), “Kongruenz wird zu Dis-
kontinuität im Deutschen”. In: Brogyanyi, Bela
Seiler, Hansjakob (1986), Apprehension: Language, (ed.), Studies in Diachronic, Synchronic, and Typo-
Object, and Order. Part III: The Universal Dimen- logical Linguistics: Festschrift for Oswald Szeme-
sion of Apprehension. Tübingen: G. Narr (LUS 1, rényi on the Occasion of his 65th Birthday. Amster-
III) dam & Philadelphia: J. Benjamins (CILT, 11),
Seiler, Hansjakob & Christian Lehmann (1982, 959⫺988
eds.), Apprehension: Das sprachliche Erfassen von Windfuhr, Gernot L. (1989), “Persian”. In: Com-
Gegenständen. Teil I: Bereich und Ordnung der rie, Bernard (ed.), The World’s Major Languages.
Phänomene. Tübingen: G. Narr (LUS, 1, I) London: Routledge, 523⫺546
Shopen, Timothy (1985, ed.), Language Typology
and Syntactic Description, Vol. 1⫺3. Cambridge Christian Lehmann, Erfurt (Germany)
etc.: Cambridge Univ. Press Edith Moravcsik, Milwaukee, WI (U.S.A.)
74. Adjective
Premper, Waldfried (1986), Kollektion im Ara- Smith-Stark, Thomas C. (1974), “The Plurality
bischen. Köln: Institut für Sprachwissenschaft der Split”. Papers from the Tenth Regional Meeting of
Universität (Arbeitspapier Nr. 49) the Chicago Linguistic Society, Chicago, IL: Chi-
Sanders, Gerald A. (1978), “Adverbial Construc- cago Linguistic Society, 657⫺671
tions”. In: Greenberg (ed.), Vol. IV, 51⫺84 Sommer, Bruce A. (1972), Kunjen Syntax. Can-
berra: Australian Institute of Aboriginal Studies
Schachter, Paul (1985), “Parts-of-speech Systems”.
Shopen (ed.), Vol. I, 3⫺61 Sugamoto, N. (1989), “Pronominality: a Noun-
Pronoun Continuum”. In: Corrigan et al. (eds.),
Schachter, Paul & Otanes, Fe T. (1972), Tagalog
267⫺291
Reference Grammar. Berkeley: University of Cali-
fornia Press Wheatley, Julian K. (1989), “Burmese”. In: Com-
rie, Bernard (ed.), The World’s Major Languages.
Seiler, Hansjakob (1983), POSSESSION as an Op- London: Routledge, 834⫺854
erational Dimension of Language. Tübingen:
G. Narr (LUS, 2) Werner, Otmar (1979), “Kongruenz wird zu Dis-
kontinuität im Deutschen”. In: Brogyanyi, Bela
Seiler, Hansjakob (1986), Apprehension: Language, (ed.), Studies in Diachronic, Synchronic, and Typo-
Object, and Order. Part III: The Universal Dimen- logical Linguistics: Festschrift for Oswald Szeme-
sion of Apprehension. Tübingen: G. Narr (LUS 1, rényi on the Occasion of his 65th Birthday. Amster-
III) dam & Philadelphia: J. Benjamins (CILT, 11),
Seiler, Hansjakob & Christian Lehmann (1982, 959⫺988
eds.), Apprehension: Das sprachliche Erfassen von Windfuhr, Gernot L. (1989), “Persian”. In: Com-
Gegenständen. Teil I: Bereich und Ordnung der rie, Bernard (ed.), The World’s Major Languages.
Phänomene. Tübingen: G. Narr (LUS, 1, I) London: Routledge, 523⫺546
Shopen, Timothy (1985, ed.), Language Typology
and Syntactic Description, Vol. 1⫺3. Cambridge Christian Lehmann, Erfurt (Germany)
etc.: Cambridge Univ. Press Edith Moravcsik, Milwaukee, WI (U.S.A.)
74. Adjective
do not have any distinct category of adjec- fact that it establishes the adnominal, attribu-
tives as such. These also differ from one tive position as the most “natural” (Kuryło-
another with some like Sanskrit, Finnish, Ar- wicz 1936; Croft 1984) and categorial one for
abic, Hausa, Dyirbal, etc. merging the rele- adjectives. This point can be exemplified with
vant lexical items with nouns, and some like the help of Kannada, in which an adjective
Manipuri, Chinese, Lakhota, Thai, Samoan, like cikka ‘small’ can occur in its bare form
Indonesian, etc. merging them with verbs as an attribute in a noun phrase, whereas a
(Thompson 1988). There are gradations in noun like mara ‘tree’ requires the addition of
the case of both these types of languages in the genitive suffix; a verb like muri ‘break’
that the adjectival words, though forming a requires the addition of the participial suffix,
subclass of either nouns or verbs, show dif- and an adverb like mēle ‘above’ requires the
ferent degrees of differentiation from them. addition of a adjectival suffix in order to oc-
The occurrence of this enormous amount of cur as an attribute.
cross-linguistic variation (three major types (1) cikka petøtøige ‘small box’
of gradations, each merging with the others) mara-da petøtøige ‘wooden box’
in the encoding of adjectival words has led muri-da petøtøige ‘broken box’
many linguists to the conclusion that adjec- mēl-ina petøtøige ‘upper box’
tives form a “mixed” category.
It is possible, on the other hand, to take On the other hand, adjectives require some
care of this diversity by establishing a proto- markers to be affixed to them in several of
typical category of adjectives; actual lan- these languages in order to occur in non-ca-
guages can then be regarded as either having tegorial functions in which nouns or verbs
or not having such a category; in the case of can occur unmarked, as will be pointed out
languages which do possess such a category, in 4. This is one of the defining and distin-
we can make further differentiations by as- guishing characteristics of adjectives in sev-
suming that the category occurring in them eral languages like Kannada (and other
is more or less close to this prototypical ad- Dravidian languages); it is also used as the
jectival category. However, in view of the defining criterion for identifying adjectives in
complexity of the situation mentioned earlier, languages like Turkana (Dimmendaal 1983).
it would not be possible for us to provide a 2.1. Use of degree modifiers
simple definition for the adjective; we would
A criterion which does rely on a positive
have to use several criteria as the basis of our
morphological property of adjectives is de-
definition; further, all these criteria can be as-
gree modification. All languages that have
signed only to the prototypical category; in-
adjectives as a distinct category also have
dividual languages would only be showing morphological or syntactic constructions of
certain combinations of them. These criteria adjectives for indicating one or more types
would also have to be gathered from different of degree modifications. Even in the case of
areas of grammar, such as the morphological, languages which do not have a distinct adjec-
syntactic, semantic, pragmatic and func- tival category, we find subgroups of verbs or
tional. nouns (corresponding to the adjectives of
other languages) which participate in such
2. Morphological categories constructions. We can therefore regard de-
gree modification as an important part of the
The morphological criteria that can form the defining criteria of prototypical adjectives.
basis of a definition of prototypical adjectives The basis of this characteristic is the fact that
will have to be derived from the morphologi- adjectives denote a single property. They are
cal categories that occur (or do not occur) different from both nouns as well as verbs
with adjectives when they are used in their on this particular point. Nouns, for example,
categorial function of modifying a noun. The denote a cluster of properties and further,
most important of these is apparently a nega- they also denote an object or entity which
tive one, namely that adjectives occur in their possesses those properties (cf. Jespersen 1924;
unmarked, bare form when used as attributes Wierzbicka 1980). The possibility of modify-
of nouns, whereas words belonging to other ing the meaning of a noun by adjectives as
categories, like nouns, verbs, and adverbs, re- attributes apparently presupposes that they
quire some kind of modification, like affixa- indicate a single property. Such an indication
tion, in order to function as attributes. The is also necessary for them in order to func-
importance of this characteristic lies in the tion as the basis of degree modification or
comparison. Jespersen accounts for this latter partial reduplication in Obolo (Faraclas
constraint of comparatives occurring mainly 1984):
with adjectives (and adverbs) on the basis of (2) ı̂yòk ‘hot’ mı̂yòk ‘very hot’
the fact that degrees of comparison necessar- éfàk ‘read’ mı̂fàk ‘read:completive:pl’
ily deal with one quality at a time.
(3) sćntı́ı́k ‘small’ sćntı́tı́ı́k ‘very small’
Languages generally use several types of
constructions (both morphological as well as Reduplication is also a device used by
syntactic) for denoting degree modification. many languages for denoting plurality in the
For example, English uses the suffixes -er case of nouns and quantificational aspect in
‘comparative’ and -est ‘superlative’ with the case of verbs (Moravcsik 1978); it is used
monosyllabic adjectives and also with many for denoting the intensity of adjectives in
disyllabic adjectives; it also uses syntactic some languages like Basque (Saltarelli 1988)
constructions with more and most, respec- and Koromfe (Rennison 1997). The following
tively, for the same purpose (funny ⫺ more Basque sentence exemplifies this usage:
funny, funniest ⫺ most funny). We may regard (4) zopa bero-bero dago
this formation of comparatives and superla- soup hot-hot is
tives as part of the process of degree modifi- ‘The soup is very hot.’
cation or intensification in view of the fact
Manipuri uses the aspectual quantifier suf-
that the latter has a wider scope. There are
fix -men ‘much’ for denoting excessive degree
also languages in which the same marker is in the case of state verbs (which include
used with adjectives for forming compara- words denoting adjectival meanings). The
tives and also for denoting intensity or high suffix can also denote excess in one of the
degree of the quality that the adjective de- arguments that occur in the sentence, such as
notes. For example, Balinese uses the suffix the actor, theme, location, associate, etc.
-an for denoting the comparative meaning as There are also several other suffixes in this
well as the meaning of intensity; further, it language which occur along with the redupli-
also uses the process of reduplication for in- cation of the adjectival base, to denote inten-
dicating the latter meaning (Barber 1977). sity. Examples (Bhat & Ningomba 1997):
Another point worth noting here is that the
(5) mehakki yum cak-melle
majority of languages do not overtly mark
his house burn-much
the adjective (occurring as predicate) in com- ‘His house has burnt too many times.’ /
parative constructions (Stassen 1985), where- ‘His house has burnt too much.’
as marking adjectives for intensity is very
widespread. (6) phurit esi eynonde cin-melli
Languages appear to correlate degree shirt this me tight-much
modification (or intensity) in the case of ‘This shirt is too tight for me.’
adjectives with plurality in that of nouns, (7) (a) ey hotel edude ley-melle
and quantificational aspect in that of verbs I hotel this stay-much
for morphological marking. For example, in ‘I have stayed in this hotel too many
Obolo, the suffix -mi, which codes plural times.’
subjects in the completive aspect with verbs, (b) ey hotel yamne ley-melle
also denotes intensity with adjectives. Inten- I hotel many stay-much
sity may also be expressed by means of ‘I have stayed in too many hotels.‘
(8) nan ‘clean’ nan-thrik nan ‘very clean’
ten ‘short’ ten-thrik ten ‘very short’
khu ‘narrow’ khu-drik khu ‘very narrow’
cum ‘straight’ cum-drin cum ‘very straight’
pik ‘small’ pik-sek pik ‘very small’
Notice that the meaning of intensity pro- which has a sort of negative sense associated
vided by the suffix -men ‘much’ in the senten- with it. Basque uses the suffix -egi for denot-
ces given above involves “superabundance”, ing this meaning (Saltarelli 1988):
(9) oinetako horiek handi-egi akdira ni-retzat
shoe those big-too is I-benefactive
‘Those shoes are too big for me.’
adjectives”. The latter can, as a rule, be de- of lexical items belonging to those categories
rived in a fairly straightforward way from rather than as adjectives (see 4); this view is
underlying structures; they are in the true supported by the fact that unlike underived
sense only condensed sentences (Meys adjectives, these derived ones are generally
1975: xiii). This important difference between less prototypical with respect to their adjecti-
adjectives and nouns derives from the fact val characteristics; they show fewer of these
that in the case of nouns the relation between characteristics, and further, they also retain
an expression and its referent is established some of the characteristics of categories from
by convention rather than by the meaning of which they have been derived. For example,
constituent elements, whereas in that of ad-
adjectives can be derived from verbs in sev-
jectives, the meaning expressed is of crucial
importance. In view of this point, the term der- eral languages by changing verbs into partici-
ivation might not be adequate for describing ples; these forms generally retain some of
the morphological processes that are involved their tense-aspect distinctions and are less
in the formation of adjectives; the term “inflec- prototypical as adjectives; for example, most
tional” might be more appropriate here. of them cannot take degree modification.
We can divide these morphologically “de- This is also true of adjectives derived from
rived” adjectives into two distinct groups, nouns and adverbs. They behave differently
namely (a) those which involve lexical items from prototypical adjectives.
belonging to other categories like nouns, Another interesting point that may be
verbs, and adverbs, and (b) those which in- noted here is that the derivation of adjectives
volve an extension or modification of lexical is rather complex in the case of languages in
items that can by themselves function as ad- which adjectives form only a subgroup of the
jectives. Some of the derived adjectives of the nominal category. They apparently obtain
latter type are dealt with in 3.1. There are this derivational complexity from being part
some additional ones which do not character- of the nominal category. For example, San-
ize the adjectival category as a whole; these
skrit shows very little difference between ad-
are described in 3.2.
jectives and nouns. It has a very complex der-
3.1. Adjectives derived ivational system for nouns (almost all its
from other categories nouns can be derived from verbal roots), and
Most of the adjectives derived from other this system gets extended to adjectival words
categories like nouns, verbs and adverbs can as well, as shown by the following uses of
in fact be viewed as involving extended uses derivational affixes:
3.2. Adjectives derived from other adjectives tributive’ and -? o ‘privative’ which, when at-
In addition to degree modification, adjectives tached to nouns, provide the meanings ‘full
manifest certain other types of derivational of …’ and ‘lacking …’ respectively, and when
(or inflectional) processes; two of these may attached to verbs, they provide the meanings
be noted here: ‘having the tendency to …’ and ‘having the
(a) Several languages attach a negative tendency not to …’ respectively: 储 am’-’o ‘wa-
marker to adjectives in order to derive their terless’, 储 nàú-兩 nàm-’o ‘disobedient’, 储 xaé-xa
negative counterparts. This can be illustrated ‘sandy’, 储 ’oré-xa ‘deceitful’ (Hagman 1974:
with English examples like wise : unwise, 60).
fair : unfair, happy : unhappy, kind : unkind, (b) Another interesting distinction is re-
etc. The distinction is represented in Nama ported to occur in Athapascan (Hupa); in
Hottentot in its derivation of adjectives from this language, adjectives can occur with the
nouns and verbs; it has two suffixes, -xà ‘at- prefixes n or l which appear to classify them
according to the degree of connection of the In most languages, on the other hand, ad-
quality with the noun; n is used mostly of in- jectives are also used in functions belonging
herent qualities, such as dimension, and l is to other word classes such as predication
used for the more accidental qualities, such (verbal function), reference (nominal func-
as colour and condition of flesh (Goddard tion), modification of verbs (adverbial func-
1911). The use of adjectives in the nominal tion) and also in functions that do not belong
and verbal functions also involves some deri- to any word class as such, like compounding.
vational processes, but these would be de- We can regard these as secondary usages of
scribed in 4. adjectives; some of the categorial characteris-
tics are lost by adjectives in these usages, and
4. Syntactic functions further, some of the non-categorial ones are
acquired by them as these new characteristics
There are languages like Takelma in which are more suitable than their adjectival char-
adjectives cannot be used predicatively (ex-
acteristics for these secondary functions. We
cept that the simple form of the adjective
may regard the former change as represent-
may be predicatively understood for an im-
plied third person; Sapir 1922). This is also ing their decategorization (Hopper &
true of a small set of “true” adjective roots Thompson 1984), and the latter change as
denoting meanings such as ‘big’, ‘small’, representing their recategorization as verbs,
‘good’, ‘new’, ‘white’, ‘red’, ‘hot’ and ‘beauti- nouns or adverbs (Bhat 1994). There is also
ful’ in Supyire; they occur as attributes either a gradation among languages in the amount
by compounding with the head noun, or by or degree of decategorization or recategoriza-
taking the prefix min- (with its vowel and fi- tion that their adjectives undergo in these sec-
nal nasal showing assimilations with the ondary usages. This gradation appears to be
following root) as independent adjectives, correlatable with a gradation in their effec-
but not as predicates (Carlson 1994): tiveness in the functions concerned.
(20) kyaà niM-cinni 4.1. Referential use
thing prefix-good
‘a good thing’ Adjectives can be used either for referring to
the property that they denote or to the per-
(21) ninké num-bwćhe
tail prefix-big son or object to which the property belongs.
‘the biggest tail’ In English, for example, adjectives like black,
white, poor, rich, etc. can be used for denoting
Tinrin has a small, closed set of adjectives the person or object that has the relevant
that can occur only as attributives and not as property. In order to indicate the property
predicates. They occur immediately before itself, one will have to use nouns which are
the head noun in a noun phrase (Osumi derived from these adjectives, such as black-
1955): ness, poverty, etc. This is also true of Tzutujil
(22) wa drorro vajù in which a fairly large number of adjectives
det big sickness can be used as nouns for denoting the objects
‘the big disease’ or persons that have the relevant property; in
(23) saa wadrò hùwùnrâ fònrı̂moo order to denote the property itself, the adjec-
one tiny small history tives have to be used with a formative suffix
‘a wee wee history’ (Dayley 1985: 201):
In Kannada, on the other hand, adjectives ing as heads of noun phrases must take a def-
can be used as nouns for denoting the prop- inite determiner; they usually indicate a ge-
erty itself; in order to denote the possessor of neric meaning; further, they do not inflect for
that property, they need to be used with cer- number (but have plural concord), or for
tain pronominal markers or other deriva- genitive case (Quirk et al. 1972: 251):
tional affixes: (26) The poor are causing the nation’s leaders
(25) as a modifier as a noun great concern.
(a) dappa halige haligeya dappa (27) The very wise avoid such temptations.
‘thick plank’ ‘plank’s thickness’
Adjectives used as nouns differ from nouns
(b) bisi nı̄ru nı̄rina bisi in Moroccan Arabic in regularly having
‘hot water’ ‘water’s temperature’ either a definite article or a demonstrative ar-
(c) kurudø u bekku bekkina kurudø u ticle (Harell 1962: 204), whereas in Syrian Ar-
‘blind cat’ ‘cat’s blindness’ abic, adjectives in their nominal usage do not
occur as subjects (Cowell 1964: 382).
When used in the referential function, ad- Kannada, and several other languages of
jectives tend to lose some of their prototypi- the Dravidian family, have an echo-form
cal characteristics. In Kannada, for example, construction for nouns (and also for verbs)
adjectives used for referring to the property which is used in contexts in which the speak-
that they denote do not admit degree modifi- ers are not very sure about the nature of the
cation. Further, adjectives occurring in an- object or event that they are denoting, or
tonymous pairs like agala ‘broad’ vs. sapura when they do not want to be very specific
‘narrow’, udda ‘long’ vs. gidø dø a ‘short’, ettara about it. Morphologically, the construction
‘tall’ vs. taggu ‘low’ etc. lose their antony- involves reduplication of the noun or verb,
mous distinction in their nominal use, with with the initial vowel (or consonant-vowel
only one of the pair (namely the unmarked combination) in the second (reduplicated)
one) occurring as the referring expression for form being replaced by gi or gı̄, depending
denoting the quality indicated by the pair as upon whether the original vowel is short or
a whole. This is true of the nominals derived long respectively:
from such adjectival pairs in English as well;
for example, length refers to the dimension (28) (a) maløe banda-re baruvud-illa
rain came-if come-not
denoted by the adjective long as well as the
‘(I) will not come if it rains.’
one denoted by the adjective short; the latter
fact is shown by sentences such as Its length (b) maløe-giløe banda-re baruvud-illa
is short by three inches. rain-echo came-if come-not
The occurrence of adjectives with case and ‘(I) will not come if it rains or any-
number markers when used in their referential thing of that sort happens.’
function, in the case of languages in which ad- Attributive adjectives do not show this echo-
jectives occur in their unmarked bare form in form construction; this is apparently because
the attributive position, can be considered an the use of such forms goes against the
instance of their recategorization as nouns. requirement that adjectives need to be very
Notice, however, that adjectives differ from specific in their meaning in order to carry out
prototypical nouns in showing these inflec- their function of modifying the meaning of
tional characteristics. For example, they may their head noun; the use of echo-forms would
show less number or case distinctions as com- make their meaning less specific and vague.
pared to nouns. They may also show grada- When used as referring (or as predicating) ex-
tions in that some nominals derived from ad- pressions, however, adjectives are not con-
jectives may be more noun-like than others. strained in this fashion, and hence they do
In English, for example, adjectives function- allow echo-forms to occur with them:
curring on the other end, and adjectives oc- syntactically than concepts denoting ma-
curring in the middle. Cross-linguistic com- terial and sex;
parison of languages showing different mor- (c) concepts denoting shape and physical
phosyntactic characteristics reflecting this properties (‘hard’, ‘soft’, ‘heavy’, ‘light’,
continuum has allowed the establishment of ‘rough’, ‘smooth’, ‘hot’, ‘cold’, ‘wet’,
the following rules and tendencies (Pustet ‘dry’, etc.) tend to be included in the ver-
1989: 111f.): bal class (if not included in the adjecti-
val class);
(a) positionals and feelings/emotions (‘an- (d) properties of human character (‘proud’,
gry’, ‘sad’, ‘happy’, ‘hungry’) are never ‘cruel’, ‘clever’, ‘generous’, ‘wicked’, etc.)
more “nominal” morpho-syntactically tend to be included in the noun class (if
than concepts denoting colour, dimen- not included in the adjectival class).
sion, value, material and sex; From these rules and tendencies, an over-
(b) concepts denoting colour, dimension and all continuum such as in Fig. 74.1. may be
value are never more nominal morpho- derived:
nouns 再 material
sex
human-propensities
colour
dimension
value
positionals
feelings
emotions
冎 verbs
Fig. 74.1
Driem, George van (1987), A Grammar of Limbu. Press, Ian (1986), A Grammar of Modern Breton.
Berlin: de Gruyter Berlin: de Gruyter
Faraclas, Nicholas (1984), A Grammar of Obolo. Press, Margaret L. (1979), Chemehuevi: A Gram-
Bloomington/IN: Indiana Univ. Linguistics Club mar and Lexicon. Berkeley: Univ. of California
Foley, William A. (1980), “Towards a Universal Press
Theory of the Noun Phrase”. Studies in Language Pustet, Regina (1989), Die Morphosyntax des “Ad-
4, 177⫺199 jektivs” im Sprachvergleich. Frankfurt/M. etc.:
Lang (Continuum 7)
Fortescue, Michel (1984), West-Greenlandic. Lon-
don: Crom Helm Quirk, Randolph & Greenbaum, Sidney & Leech,
Geoffrey & Svartvik, Jan (1972), A Grammar of
Goddard, Pliny E. (1911), “Athapascan (Hupa)”.
Contemporary English. London: Longman
In: Boas (ed.), 85⫺158
Rennison, John R. (1997), Koromfe. London:
Harell, Richards (1962), A Short Reference Grammar Routledge
of Moroccan Arabic. Washington/D. C.: Georgetown
Univ. Press Saltarelli, Mario (1988), Basque. London: Rout-
ledge
Hopper, P. L. & Thompson, Sandra A. (1984),
“The Discourse Basis of Lexical Categories in Uni- Sapir, Edward E. (1922), “The Takelma Language
versal Grammar”. Language 60, 703⫺752 of Southwestern Oregon”. In: Boas (ed.), 1⫺294
Jespersen, Otto (1924), The Philosophy of Gram- Schachter, Paul & Otanes, Fe T. (1983), Tagalog
mar. London: Allen & Unwin Reference Grammar. Berkeley: Univ. of California
Press
Kuryłowicz, Jerzy (1936), “Dérivation lexicale et
Stassen, Leon (1985), Comparison and Universal
dérivation syntaxique (contribution à la théorie des
Grammar. Oxford: Blackwell
parties du discours)”. Bulletin de la Société de Lin-
guistique de Paris 37, 79⫺92 [reprinted in: Hamp, Stassen, Leon (1997), Intransitive Predication. Ox-
Eric P. & Householder, Fred W. & Austerliz, Ro- ford: Clarendon Press
bert (1966, eds.), Readings in Linguistics, Vol. II. Taylor, Charles (1985), Nkore-Kiga. London:
Chicago, London: Univ. of Chicago Press, 42⫺50] Routledge
Lehmann, Thomas (1989), A Grammar of Modern Thompson, Sandra A. (1988), “A Discourse Ap-
Tamil. Pondicherry: Pondicherry Institute of Lin- proach to the Cross-Linguistic Category ‘Adjec-
guistics and Culture tive’”. In: Hawkins, John A. (ed.), Explaining Lan-
Lorimer, David L. R. (1935), The Burushaski Lan- guage Universals. Oxford: Blackwell, 167⫺185
guage. Oslo: Aschehoug Wetzer, Harrie (1996), The Typology of Adjectival
Lyons, John (1977), Semantics. London: Cam- Predication. Berlin: de Gruyter
bridge Univ. Press Wierzbicka, Anna (1980), The Case for Case. Ann
Arbor: Karoma
Macawlay, Monica (1996), A Grammar of Chalca-
tongo Mixtec. Berkeley: Univ. of California Press Winfield, W. W. (1928), A Grammar of Kui Lan-
guage. Calcutta: Asiatic Society of Bengal
Meys, W[illem] J[ohannes] (1975), Compound Ad-
jectives in English and Ideal Speaker-Listener. Am- Bhat wishes to acknowledge his indebtedness to the
sterdam: North-Holland Research Council of Antwerp University and to
Moravcsik, Edith M. (1978), “Reduplication Con- the India Study Center for financial help in prepar-
structions”. In: Greenberg, Joseph H. (ed.), Univer- ing this article.
sals of Human Language. Stanford, California:
Stanford Univ. Press, 297⫺334 D. N. S. Bhat, Mysore (India)
Osumi, Midori (1995), Tinrin Grammar. Honolulu: Regina Pustet, Munich (Germany)
Univ. of Hawai’i Press and Boulder (USA)
75. Numeral
1. The numeral as a part of speech tion as a numeral modifier of the noun, e.g.
2. Kinds of numeral discourse four-legged table.
3. The unmarked status of cardinal numerals Both the syntactic relation of a numeral to
4. The mathematical structure of cardinal the noun it modifies, and the numeral’s mor-
numeral systems
5. The expression of mathematical functions in
phology often differ from those of adjectives.
cardinals For example, Latin numerals above ‘three’,
6. Extensions of the cardinal numeral system unlike adjectives, are not inflected for gender.
7. Cardinal numerals and substantival number Sometimes the morphosyntax of different
8. Non-cardinal series of numerals parts of the numeral system is diverse and id-
9. References iosyncratic. It has been stated already that
lower numerals often function as adjectives
while higher numerals are more noun-like.
1. The numeral as a part of speech However, the peculiarities of some numeral
systems extend beyond this dichotomy. In
In traditional grammar the numeral is con- Russian, for instance, the masculine nomi-
sidered, along with the adjective, pronoun native of animates and the masculine and
and noun, as a part of speech grouped in the neuter nominative-accusatives of ‘two’ are all
superordinate category of substantive and dva, while their feminine counterparts are
opposed especially to the verb and adverb. dve. No other noun modifiers have these in-
A number of observations indicate that a flectional endings. They are survivals of a
valid insight may underlie this traditional dual and govern what is synchronically a
classification. Like adjectives, numerals usu- genitive singular. They share this syntax but
ally modify nouns. Within complex numerals, not morphology with ‘three’ and ‘four’ while
expressions that designate bases (and some- forms for 5 to 20 are morphologically singu-
times other higher numbers) frequently have lar nouns which govern the genitive plural of
noun-like properties such as having a plural the noun designating the object counted.
or governing a genitive, e.g. Latin tres mi- Moreover, all of these rules differ when the
lia hominum ‘three:nom.pl thousand:nom.pl noun is in one of the oblique cases.
men:gen.pl (three thousand men)’. In gene- In certain languages, it might appear that
ral, the lower the number the more adjectival numerals are kinds of verbs, very much like
its properties; conversely, the higher the the claim made for certain languages that ad-
number the more it behaves as a noun. (cf. jectives are really stative verbs (Art. 72 and
Corbett 1978). Nominal properties are partic- 74). For example, in Yurok, a Penutian lan-
ularly characteristic of bases, especially guage of California, it is asserted that both
higher bases (cf. Stampe 1976). numerals and adjectives should be regarded
As nominal modifiers (their basic syntactic as subclasses of verbs (Robins 1958: 68).
use) they differ from adjectives in a number However, it is also noted that syntactically
of ways. The first concerns word order. Nu- numerals mostly precede nouns to form en-
merals frequently precede nouns in languages docentric groups. Further, most languages
in which the adjective follows, as indicated show a reluctance to predicate numerals. In
by the statistical universal AN J QN. When English, one says, There are three people in
both numeral and adjective precede the the house rather than The people in the house
noun, and when they both follow, the nu- are three.
meral is usually the distal modifier, as in three From the fact that the morphosyntax of
large houses. numerals in many languages differs from that
The probable reason for this is semantic. of the adjective and even varies within the
Whereas an adjective indicates a property of numeral system, it follows that numerals are
a noun, a numeral is not a property of the generally distinct from adjectives. This dis-
object itself but of a set of objects, often a tinction is most accurately described in se-
nonce-property, e.g. the five cards that hap- mantic terms.
pen to be in one’s hand at a particular point The distinctive characteristic of numerals
in a game of poker. Where a numeral does is ordering, or, more precisely, cardinal or-
act as a property of an object it will be dering. Although adjectives can be graded ⫺
treated differently and will no longer func- in that a particular object may have more or
less of a given quality ⫺ such grading re- sions such as dozen, score, and googol, which
mains confined to the semantic range of each are not part of the counting system in Eng-
adjective. In contrast, a fundamental charac- lish. In a few languages, there are differences,
teristic of cardinal ordering is the property of usually minor and sporadic, between expres-
“betweenness”. While it makes no sense say sions used in counting and those used in
that ‘beautiful’ is between ‘good’ and ‘true’, context usually as nominal modifiers. Wit-
it does make sense to assert that ‘four’ is be- ness Hungarian kettö ‘two’, used only in
tween ‘two’ and ‘six’. counting, versus két ‘two’, used as a noun
Even more fundamental among properties modifier. Such differences never affect the
of the natural numbers is “nextness”. This is mathematical structure of the system.
captured in Peano’s (1908) famous postulates One may say that counting is a special
for the natural numbers. Every integer has a kind of discourse. In fact, especially in li-
unique successor. Yet the grading within an terate societies, there are still other forms of
adjective as positive, comparative and super- discourse involving numerals which have dis-
lative, while possessing “betweenness”, lacks tinctive characteristics. In regard to counting
“nextness”. In other words, while the order- itself, one may distinguish abstract from con-
ing of quality across the semantic range of crete counting. The latter entails counting
an adjective is gradual, that of the natural objects of a specific kind. In languages with
numbers (‘one’, ‘two’, ‘three’ …) is discrete. numeral classifiers (cf. Art. 97) the appropri-
Contrastingly, the set of real numbers has, ate classifier is generally used in such in-
like adjectives, the property of “betweenness” stances. Numerals are also used in mathemat-
but not “nextness”. Between 1 and 3 on this ical discourse. Differences between numeral
continuum there is 2, yet this is but one of an expressions both in form and syntactic prop-
infinity of real numbers within the interval. erties sometimes occur within single lan-
In human language, the fundamental nu- guages. In Russian the instrumental case of a
meral set is the cardinal or natural numbers. few numerals, when used to indicate the mul-
To determine the numerical value of a set of tiplier, have a stress pattern different from
objects is to put its members into one-to-one that used in ordinary discourse, e.g. pját’ju trı́
relation with the words designating the natu- ‘five:instr.sg three:nom.pl (five times three)’
ral numbers in the order of counting. The nu- as against s pjat’jú továriščami ‘with five:
meral which completes the set then expresses instr.sg friend:instr.pl (with five friends)’.
the number of items in the set. Other uses of numerals distinct both from
What is outlined here is basically similar counting and ordinary discourse include nur-
to what has been called “the cardinality prin- sery rhymes, dates, telephone numbers and
ciple” (Hurford 1987: 305). names for numerals in written form. One
might say, You have written this “three” larger
than the other two “threes”. Note that in this
2. Kinds of numeral discourse case a plural of three is utilized, which is un-
acceptable in the numeral’s ordinary syntac-
What has been said above regarding the nu- tic use.
meral as a part of speech refers to what
mathematicians call “natural numbers” ⫺
those invoked in counting unit members of 3. The unmarked status of cardinal
a set. These are represented by the cardinal numerals
numerals. Their relation to other series such
as ordinals is discussed below in section 3. Everything asserted up to this point regard-
It is important to distinguish numerals ing numerals refers to cardinal numbers
from numerical expressions in general, of (those generated by the act of counting).
which the former are a subset. For example, However, there are other numeral series, e.g.
in English sixteen is a numeral expression ordinal (first, second…), adverbial cardinal
which is also part of the numeral system, (once, twice…), adverbial ordinal (the first
whereas the square of four is a numerical ex- time, the second time) and distributives (one
pression which designates the same number each, two each), etc. It is clear from a whole
(16) but is not part of the numeral system. series of facts that cardinals are the basic, un-
No one counts one, two, three, the square of marked numeral series, and that, when the
two, five… This distinction also allows one to numeral is said to be a part of speech, refer-
rule out certain marginal numerical expres- ence is to cardinals only. All of the general
works regarding numeral systems (e.g. ples, there are n! possible orderings ⫺ but
Brandt Corstius 1968, ed.; Hurford 1975; which one to use is a matter of indifference.
1987; Stampe 1976) examine mainly or exclu- One is not normally interested in the nth ap-
sively cardinals. However, studies of Indo- ple.
European numeral systems, (cf. Szemerényi An interesting example of the way in
1960; Gvozdanović 1992, ed.) usually include which ordinary language treats the relation-
non-cardinal series ⫺ especially ordinals. The ship between cardinals and ordinals (and of
reason for this is that there have been histori- the primary status of the former) is furnished
cal influences ⫺ mainly analogical ⫺ of one by Lotuxo, an Eastern Nilotic language of
set on another. the Nilo-Saharan family. Here the ordinal is
The non-cardinal series show the typical formed from the cardinal by means of a caus-
attributes of markedness (Trubetzkoy 1939; ative prefix. The ordinal has been interpreted
Greenberg 1966). Among these are the to be the number which causes the set to be
greater complexity of expression of the completed, e.g. ‘fifth’ means that which
marked series, as shown by affixation to the causes a set to have five members (Muratori
unmarked cardinal, e.g. English tenth as 1938: 18).
against ten, or by the use of specialized nouns The most important exception to the ab-
with cardinals, e.g. English time in the adver- sence of natural sequencing of entities in the
bial cardinal four times, or doubly marked in world is time, whether serial or absolute,
the adverbial ordinal the fourth time. Another since it is an inherently ordered asymmetrical
symptom of markedness is defectivity. In relation between the earlier and the later. It
Slavic languages the collective numerals are is precisely in time expressions that one finds
not formed above ‘ten’. In some languages variations within and between languages. In
higher ordinals involve partial neutralization English, dates in years are stated cardinally,
in which the unmarked cardinal takes the but in Russian ordinally. Yet in English, un-
place of the ordinal. An example is English like some other languages, ordinals are used
twenty-third in which the member with higher for days of the month. English uses ordinals
value uses the cardinal in place of the ordi- for successive monarchs of the same name,
nal twentieth. One may contrast Latin vicesi- e.g. Henry the Eighth; in French, however,
mus primus ‘twentieth:nom.sg.m first:nom.sg.m one finds Louis quatorze, which utilizes the
(twenty-first)’ in which both members are or- cardinal.
dinal. The data available show also that the
text frequency of cardinal numerals is over- 4. The mathematical structure of
whelmingly greater than that of any other
series. In languages with very restricted sys- cardinal numeral systems
tems lacking multiplicative bases, e.g. those All languages have finite systems of cardi-
of Australia, it appears that only cardinal nu- nals. For example, in American English the
merals exist. highest integer expressible is 1036⫺1 since the
The ordinals, which usually function as a lexical item in that numeral system with the
subclass of adjectives, deserve special con- highest value is decillion. Of course no one
sideration. Some philosophers of mathemat- starting with 1 will count up to this number.
ics have taken ordinals as primary. For exam- Counting, however, is a matter of compe-
ple, ‘five’ might be defined as the fifth tence rather than performance. The existence
number in the counting series of natural of rules, such that given a numeral designat-
numbers. Others, however, have accepted the ing a particular number, one can produce
cardinals as logical primitives. an expression designating the next higher
In natural languages the evidence is un- number, is sufficient. As has been shown, the
equivocal that the cardinals are the un- limits of the numeral system are narrower
marked and basic series. This, it may be sug- than those of numerical expressions in gene-
gested, derives from the fact that in the real ral. Among extra-systemic resources, by
world most things are not naturally ordered which one can express numbers larger than
sequentially. Sometimes one might want to the limit of the cardinal system proper, are
find out how many apples are in a sack, but addition, multiplication, and exponentials.
in what order individual apples are assigned For example, one can substitute a thousand
numbers is irrelevant. All that is required is decillions for (unacceptable) *undecillion.
that the count be complete and that no apple A system which assigned to every number
be counted more than once. If there are n ap- a simple and distinct lexical expression would
place great strain on the memory. Hence, ex- Even more strikingly, the following “system”
cept in very simple systems like that of the is reported by from the Kaliana in South
Botocudo, a Macro-Ge language of Brazil in America: meyakan ‘one’, meyakan ‘two’,
which a system ‘1, 2, many’ is reported, num- meyakan ‘three’, in which meyakan means
bers are expressed as arithmetical functions ‘finger’.
of a limited set of numeral expressions which Systems of the two major types mentioned
in are called the “atoms” or “frames” of the above (those without arithmetic functions
system (cf. Greenberg 1978: 256; Salzmann and those with sporadic use of addition and/
1950). or multiplication) are basically confined to
An example is ‘twenty-three’ expressed in hunting-gathering populations which do not
Mandarin by a succession of three atoms (2, possess goods (e.g. cattle) requiring counting
10 and 3 in that order) and which is interpre- to relatively high numbers, and which have
ted as a function (axb)⫹c in which the argu- no extensive commerce.
ment a has the value 10, b has the value 2 The vast majority of the world’s languages
and c has the value 3. use multiplicative bases. As can be seen from
From the examples of Botocudo and Eng- the earlier example of Yuma, the mere use of
lish one sees that numeral systems differ multiplication is not sufficient; requisite is the
vastly in their limits. This variation may be use of expressions formed by multiplication
the clearest example among very few in- of a base number successively by 1, 2, 3, etc.
stances of evolutionary factors at work If the base is 10, then these will be 10, 20, 30,
within language. etc., with intermediate numbers expressed by
Typologically, the most simple numerical some function (most commonly addition) in-
systems are those without any arithmetic volving non-base numbers. When the base
functions. The system of this type with the number itself is reached as a multiplier, e.g.
highest numerical limit noted is Guana, an 10x10, the language commonly uses a new
Arawakan language of South America which lexeme to indicate the square of the base, and
counts ‘1, 2, 3, 4, many’. so for some still higher powers of the base,
The next simplest systems make use of e.g. English thousand.
arithmetic functions ⫺ usually addition but
Although rarely if ever stated in such
sometimes multiplication ⫺ so that some-
terms, the invention of a base for counting
what higher numbers can be expressed pre-
is surely one of the great early inventions of
cisely. One dialect of Yuman, a California
humanity. The most common typology of nu-
Hokan language, is unusual in its use of spo-
meral systems is with reference to numeral
radic multiplication without addition in a
system which counts ‘1, 2, 3, 4, 5, 3⫻2, 7, bases, e.g. quinary (5), decimal (10), duodeci-
4⫻2, 3⫻3, 10’. However, in these systems the mal (12), vigesimal (20), sexagesimal (60). Of
crucial notion of a numeral base is lacking. these, by far the most common is the deci-
The upper limits of systems which use addi- mal. Its origin from finger counting is intu-
tion and/or multiplication without bases are itively obvious, and in a fair number of in-
difficult to determine. stances etymologically clear. In some lan-
This is because there are documented in- guages the words for ‘one’ are identical with
stances of languages in which numeral ex- or derived from those for ‘finger’ and terms
pressions involve counting first on the fingers for ‘five’ are derived in a similar way from
and then on the toes, and in which the words ‘hand’. The vigesimal system derives from
are actually the names of (anatomical) digits. counting both fingers and toes. In some in-
In these cases it seems that numerals above stances words for ‘twenty’ are identical with
the first few are never used without accompa- or derived from words for ‘man’. No satisfac-
nying gestures, and that the gestures are tory theories have been advanced for the oc-
often used without verbalization. currence of duodecimal and sexagesimal sys-
One indication of such a system is the use tems (cf. Ifrah 1985).
of the same word for more than one number. Many numeral systems are to some extent
Auetö, a Tupian language, uses the same “mixed”, in that, while a single base predomi-
word for 4 and 9; the word means ‘index fin- nates sufficiently for classification, other
ger’. In such cases it seems doubtful that bases exert sporadic influence within the sys-
these expressions are normally used attribu- tem, either as survivals from earlier baseless
tively with nouns in sentences, with the prob- systems or possibly as substratum of systems
able exception of the very lowest numerals. with different bases.
In regard to the former, the survival is a few cases where no overt mark of subtrac-
usually in the form of a serialized augend. tion seems to be in use, the process involved
In arithmetic terminology the augend is a is deletion, namely of ‘one’ or of a systemic
number to which another number is added; base (both alternatives are intuitively good
the latter number is called the addend. By a candidates for deletion). An example is Efik,
serialized augend is meant a base for addition a Benue-Congo language of Nigeria in which
to which successive numbers are added. Thus 9 is usuk-kiet. Since kiet is the word for ‘one’,
in a decimal system 10, 20, 30, etc. are usually one might be tempted to interpret usuk as
serialized augends, in that successive num- ‘ten’. However, the usual word for ‘ten’ is
bers are named by addition to a multiple of edip and usuk is clearly derived from suk ‘to
10, e.g. 41 is rendered as ‘forty plus one’, 42 be left over’. Yet with ‘ten’ deleted usuk might
as ‘forty plus two’, etc. However, some serial- be interpreted synchronically as a suppletive
ized augends are not themselves whole alternant of edip ‘ten’. A further restriction
multiples of a base, i.e. some serialized au- placed on lexical subtraction is that when a
gends are not serialized multiplicands. number is expressed by subtraction, or when
Using concepts of serialized augend and subtraction occurs as a constituent of a com-
power it is possible to define a “pure” system plex numeral expression, the subtrahend is
for a specific base. This has two require- never larger than the remainder. If a language
ments: every serialized augend must be a analyzed 2 as ‘10⫺8’, this rule would be
multiple of the base (including 1 as a violated, as the subtrahend, 8, would be
multiple) and all higher bases must be powers larger than the remainder. Subtrahends are
of the fundamental base. In this sense Man- never themselves mathematically complex,
darin is a pure decimal system. Examples of i.e. instances such as 12 being analyzed as
decimal systems with non-decimal deviations ‘20⫺(5⫹3)’ do not occur; addition is not sub-
include Welsh and French. In Welsh, 16, 17, ject to this constraint, as seen in French soi-
18 and 19 are lexically interpreted as ‘15⫹1’, xante-dix-sept ‘sixty-ten-seven’. In cases of
‘15⫹2’, ‘15⫹3’ and ‘15⫹4’. Therefore 15 is a subtraction, the minuend is almost always a
serialized augend which is not a whole multiple of the base and involves a con-
multiple of the system’s overall base, namely
tinuous sequence of subtrahends beginning
10. This evidently involves a quinary prin-
with 1. Latin undeviginti ‘one from twenty
ciple. However, the term for 15 itself, bym-
(nineteen)’, and duodeviginti ‘two from twenty
theg, consists of ‘five’ followed by ‘ten’, and
(eighteen)’ are typical examples. There are
18 has an alternate form involving multi-
very few exceptions to these rules (cf. Green-
plication of 6 and 3. In French, quatre-vingt
is ‘four-twenty’ and acts as an additive base berg 1978: 260f.).
for numbers 81⫺99. This clearly invokes a vi- The inverse of multiplication, division, is
gesimal principle, seen also in the forms for much rarer and subject to even greater re-
numbers 61 to 79. strictions than subtraction. Almost all exam-
There are pure decimal and pure duodeci- ples are of forms for ‘50’ expressed as ‘half
mal systems, but no pure quinary systems of 100’. The denominator of the fraction is
since 25, 125, etc. have not been documented always a power of 2 and probably does not
as higher bases in any language. Although go beyond 4, as in Oriya, an Aryan language
some vigesimal systems, e.g. Mayan, use of India, in which subtraction and division
powers of twenty for higher bases, expres- are both involved in expressions such as 275
sions below 20 ⫺ used additively also in as ‘300⫺((1/4)⫻100)’. The expression for 100
higher numerals ⫺ employ quinary and/or is in this case deleted.
decimal principles. Beyond the four standard arithmetic func-
The two basic lexical arithmetic operations tions is an operation called “overcounting”
are multiplication and addition; both are (cf. Menninger 1969; Hurford 1975: 235). An
used in all systems which utilize a numeral example is Ostyak, an Ugric language, in
base. The inverse of addition, subtraction, which 18 is expressed as ‘eight-twenty’, which
occurs widely but is highly marked, as shown is interpreted as ‘8 going-on 20’. Such con-
by the restrictions to which it is subject. Un- structions often involve an ordinal inter-
like addition, subtraction is almost never uti- pretation, as in Estonian, in which the nu-
lized without an overt marker indicating the meral expressing 18 can be interpreted as
operation itself, whereas such expressions as ‘eight of the second ten’ (with the term for
German drei-zehn ‘three-ten’ are common. In ‘ten’ optionally omitted).
In considering more systematically the the two addends and the order of the factors
analysis of complex numeral expressions, an 2 and 100 can only be ascertained by linguis-
important formal difference between addi- tic methods. A solution is usually (but not
tion and multiplication and their inverse pro- always) possible. For details concerning these
cesses (subtraction and division) becomes methods, see Greenberg (1978: 263⫺273).
crucial; commutative and associative laws Certain generalizations can be deduced
apply to the first pair of processes but not from such analysis. The most important is
to the last pair. Given that a certain numeral the following: wherever there are three sum-
expression in a language designates 8 and mands and at least one is a product, parenth-
does so in an expression whose numerical esization starts by separating the summand
values are 10 and 2, whether or not the oper- with the largest numerical value from the
ation of subtraction is indicated by a lexical rest. Thus in English three thousand four
item such as ‘take away’, the analysis 10⫺2 hundred and seventy three, one first parenthe-
is unambiguous. However, since by the com- sizes (3⫻1000) and subsequently adds to that
mutative law of addition a⫹b ⫽ b⫹a, and term [(4⫻100)⫹(7⫻10)⫹3)]. Within the se-
since such a conclusion can be generalized to cond summand one parenthesizes (4⫻100)⫹
any number of addends, then if 7 is expressed [(7⫻10)⫹3]. A number of generative analyses
by ‘five two’ one can deduce that 5 and 2 produce phrase structure trees for higher nu-
are addends but without further analysis, one merals that accord with this analysis (Brandt
cannot determine whether 5 has been added Corstius 1968, ed.).
to 2 or vice versa. The two possible inter- Note that the preceding generalization
pretations correspond to two different prag- does not assert anything about the linear or-
matic situations, e.g. placing two objects der of the constituents themselves. Regarding
onto a pile of five objects or placing the latter this there is another major generalization in-
onto the former. Similarly, if 30 is expressed volving a preference for putting the addends
as a function of 3 and 10, one cannot auto- with the higher numerical value before those
matically determine whether a 3 is taken ten with the lower. Whenever a language exhibits
times or a 10 is taken three times. The com- instances of both the higher preceding the
mutative law can be extended to any number lower and the lower preceding the higher, the
of factors, increasing these difficulties. latter construction is found in the lower num-
Where there are three or more addends bers and a distinct and definite switch point
and/or factors, an additional variable enters, or interval of variation occurs within the nu-
namely grouping in accordance with the as- meral sequence. There is never more than one
sociative laws (a⫹b)⫹c ⫽ a⫹(b⫹c) and switch point or interval.
(a⫻b)⫻c ⫽ a⫻(b⫻c). These laws are of For example, in Italian, given that 16 is
course generalizable to more than three ad- expressed by se-dici ‘six-ten’ one might pre-
dends and/or factors. Different associative dict that the same order of smaller plus larger
groupings have distinct pragmatic correlates. will hold for all smaller complex numerals,
In regard to addition, for example, a person e.g. quin-dici ‘five-ten’ for 15. In Italian, 16 is
may take three objects, add two more, and the switch point, since 17 is expressed as
then add six more to the resulting set of five, dicia-(s)sette ‘ten-seven’; all higher numbers
or he or she might first add two objects to have higher before smaller. For example, 375
a set of six and then add three more to the is tre-cento sett-anta cinque ‘three-hundred
resulting set. seven-ty five’. Note, however, that for 315
Sometimes the mathematical result shows Italian has tre-cento quin-dici ‘three-hundred
that only one grouping is possible. This hap- five-ten’, in accordance with a principle dis-
pens in complex numerals involving multi- cussed below.
plication and addition together. If, for exam- The generalization here is that there is
ple, 203 is expressed by functions with the ar- never more than one switch point or interval
guments 2, 100, and 3, then the two addends and that the switch always involves “lower⫹
must be (2⫻100) and 3. An analysis as higher” for lower values and “higher⫹lower”
2⫻(100⫹3) gives the wrong numerical result; for higher values. If a language always puts
one cannot indict a whole people of an arith- higher before lower (e.g. Mandarin), there is
metic error! Note, however, that in this in- no switch point. Alternately, one might say
stance it is only the parenthesization (i.e. the that the switch point is zero. Such languages
sequencing of analytic steps) which is deter- are fairly common. Languages with no switch
mined by the numerical result. The order of point in which lower always precedes higher
are very rare. The only noted example is say, as *cent-septante. In fact, some non-stan-
Malagasy. dard forms of French have septante for 70.
There is evidently a cognitive principle in- In these, in accordance with the principle of
volved in preference for higher followed by base for predictable expression, 170 is in fact
lower. If one expresses a large number such expressed as cent septante. For a more rigor-
as 3625 with the thousands term first, then ous statement of this principle see Greenberg
the listener knows immediately that the (1978: 280f.).
number is within the range [3000⫺4000]. The
addition of six hundred further narrows the
possibilities. If, alternatively, one begins with 5. The expression of mathematical
five, then all the listener knows is that the functions in cardinals
number is greater than or equal to 5; fur-
ther confinement to a finite interval is not The mathematical structure of cardinal nu-
normally possible. Hence, while an order meral systems has now been examined, but
“higher⫹lower” orients the listener to ever not the methods employed by languages to
closer approximations of the final value, the express the mathematical functions of which
opposite order leaves him or her in the dark the individual numbers are arguments. In
until the last item is reached. And even this considering this topic one may distinguish
moment may be ambiguous; in order to close between the internal morphosyntax and the
the numerical expression under such a sys- external syntax of cardinal numeral expres-
tem, the speaker might have to utter a noun sions. The two are often intimately connected
or an inflection upon the last numeral. and therefore cannot have entirely separate
In light of this one can see why natural systems. The chief reason for this is that the
languages do not utilize place systems, which main external use of numerals is to qualify
of course require the use of numeral expres- substantives and that in systems which have
sions for zero. Under such systems, either the bases, the base itself is often treated as a sub-
powers of the base would ascend in order, stantive. For instance, in Russian the expres-
producing the cognitive processing problem sion for 100 is a neuter substantive sto. Just
just mentioned, or they might descend, in as the numeral pjat’ ‘five’ in the nominative
which case the listener would have the same and nominative-accusative governs the geni-
problem, since an initial 6 might indicate a tive plural of a noun as in pját’ dom-óv
multiple of any power of the base (down to ‘five:nom.sg house-gen.pl’, so in the numeral
n0 ⫽ 1), such that 600000 could not be dis- expression for 500, pjat’-sót ‘five:nom.sg-
tinguished from 60 until the entire numerical hundred:gen.pl’. However, while ‘five hou-
expression was completed. In natural lan- ses’ is a phrase, ‘five hundred’ has a single
guages, atomic expressions of the base and of word accent and is orthographically a single
at least some of its powers, such as hundred, word. There is even in Russian what might be
thousand, trillion, etc. orient the listener with called internal inflection, in that agreement in
maximal efficiency toward the intended nu- case can occur within the word. This can be
merical value. illustrated in the oblique cases of numerals
A further general principle underlies the like pjatı́desjati ‘fifty’, which is the genitive of
organization of all systems of cardinal nu- pjat’ ‘five’, and desjatı́, genitive of désjat’.
merals which utilize a numerical base. This Both of these are inflected for case like “soft”
principle was alluded to previously in discus- (palatalized) singular nouns. ‘Of five houses’
sion of Italian tre-cento quin-dici, 315. Given is pjatı́ domóv ‘five:gen.sg house:gen.pl’.
that 15 in Italian is a non-agglutinative union Similarly the genitive singular of ‘ten’ is used
of cinque ‘five’ and dieci ‘ten’, one finds that with the genitive plural of ‘house’. For ‘of
above some particular base ⫺ frequently, as fifty houses’ Russian has pjatı́desjati domóv,
here, the square of the fundamental base ⫺ in which both ‘five’ and ‘ten’ are in the geni-
there are, as it were, no surprises. Such a base tive singular while ‘house’ is in the genitive
may be called the base for predictable expres- plural. But once more, even with inflectional
sion. A further example is French, in which endings on both components, the example
the base for predictable expression is cent. In shows single-word stress and single-word or-
standard French, for instance, in which 70 is thography.
expressed as soixante-dix ‘sixty-ten’, one ex- One sees here, from a phonological point
pects that 170 will be expressed using the of view, a development (in this case histori-
same term, i.e. as cent soixante-dix, and not, cally documented) basically analogous to
grammaticalization. From ‘five’ and ‘ten’ lated with the adpositional type of the lan-
Russian has formed a single lexical unit. One guage in question and the permitted positions
might hesitate to call this process grammati- for absence of the link is like that of coordi-
calization, since there is no change in mean- nators in general. The term deletion is
ing; the earlier form ‘five ten’ and the later avoided here because its use in synchronic
term ‘fifty’ have identical meanings. In gene- grammar suggests strongly that it was for-
ral, lower complex numerals have a greater merly present and this is by no means always
tendency ⫺ presumably because of their the case. The following principles hold for
greater frequency ⫺ to undergo phonological links for addition:
changes, which can exceed simple unification
(a) A link for addition is never initial in a
by word accent with retention of inflectional
numeral expression;
categories of each member. Frequently, there
(b) if a link for addition is final, the language
is phonetic loss and change due to loss of ac-
is postpositional;
cent ⫺ especially in lower units ⫺ which
(c) when a link is medial it always goes with
shows that, synchronically, the word has be-
the following numeral in a prepositional
come a single lexical unit, e.g. French quinze
language and with the preceding numeral
[kẽz] which is entirely opaque in relation to
in a postpositional language.
cinq [sẽk] and dix [dis], as contrasted with
Latin quin-decim < quinque ‘five’, decem Note that coordinators are in general more
‘ten’. In Latin the mathematical components sparingly used in numeral expressions than in
have, of course, already lost their indepen- phrase- and clause-coordination. For exam-
dence and have undergone change and reduc- ple, although initial coordinators, as stated
tion in the process. above, do not occur in numeral expressions,
In all of the examples considered up to they may occur in non-numeral construc-
now there have been either original juxtapo- tions, e.g. Latin et … et … ‘both … and …’.
sitions of the mathematical components or In English, a prepositional language, coordi-
some syntactic bond of agreement or govern- nators are optionally used in higher numeri-
ment between them, but no morpheme or cal expressions, e.g. two hundred (and) fifty
word designating the mathematical operation six. Note that in English John and William
itself. This is common ⫺ but far from uni- there is an optional pause before and, but not
versal ⫺ with addition or multiplication, but after it; this is just so in numerical expres-
not with the highly marked operations of sions as well. A striking example of the valid-
subtraction, division, or overcounting. ity of this principle is observed in Bedauye,
A morpheme or word which explicitly in- a Cushitic language of Ethiopia, which has
dicates the operation itself will here be called borrowed the Arabic wa ‘and’ into its nu-
a link. For example, in German vier-und- meral system. In Arabic ⫺ a prepositional
zwanzig ‘four-and-twenty’, -und- functions as language ⫺ wa as a sentence- and phrase- as
a link for addition. Here, as in Greek treis- well as numeral coordinator, is always writ-
kaı́-deka ‘three-and-ten’ there is a coordina- ten with the following word, and may begin
tor (link) fossilized within a complex nu- a sentence. In postpositional Bedauye, how-
meral expression. ever, -wa is always part of the preceding
The two chief types of links for addition word. For example, 31 is expressed in Be-
are the associative ‘and’ (or ‘with’) and the dauye as mehei taman-wa gal-wa ‘three ten-
superessive, an expression meaning ‘upon’ or and one-and’. The last -wa can be omitted.
something similar. Forms meaning ‘under’ or The most common syntactic analysis of
the like are never used. The metaphor is multiplication treats the base as a noun so
clearly that of placing one object upon an- that, for example, 30 is expressed by ‘three
other. The associative link is often a mor- tens’, in a construction analogous to ‘three
pheme ‘and’ which is also used to coordinate houses’. Bantu, with noun class prefixes for
noun phrases or clauses. It can also be an singular and plural, shows many clear exam-
adposition meaning ‘with’, itself a frequent ples. In Ila, a Bantu language of Zambia, the
historic source for words meaning ‘and’. In numeral for 10 is i-kumi (sg.), ma-kumi (pl.),
some languages, e.g. Hausa da, the same just as the singular of ‘foot’ is i-tende and its
word functions as both ‘and’ and ‘with’. plural is ma-tende. Ila expresses 20 as ma-
Since both ‘with’ and ‘on’ are adpositions, kumi o-bili ‘pl-ten pl-two’ just as ‘two feet’
it is not surprising to find that with consider- is ma-tende o-bili. In other instances, multiples
able consistency its position is clearly corre- of the lowest base, (as in decimal 2⫻10, 3⫻10,
etc.) may be expressed in a non-transparent two with rare examples of fours. Almost all
manner although the identity of the lower instances are of 50 expressed as ‘half’ of 100,
element remains clear enough, e.g. Latin tri- and these usually appear in vigesimal sys-
ginta ‘three-ten’ in which the resemblance of tems. An example of division by 4 (involving
the first element to tre-s ‘three-pl’ is readily subtraction also) is found in Oriya, an Indo-
apparent while the second part -ginta reap- Aryan language which expresses 275 as
pears in other multiples of ten despite being ‘quarter from three hundred’ in which once
superficially unlike decem ‘ten’. This disparity more an occurrence of a base (here 100) is de-
between forms for 10 and its plurals is leted.
commonly attributed to an abstract *(d)k t In overcounting, the final arithmetic oper-
whose /g/ in Latin in place of /k/ is unex- ation encountered, overt expression of the
plained. Another alternative is the use of an operation is once more required while the
adverbial, also called multiplicative numeral, base in some language is deleted. Examples
e.g. ‘three times’. An example is found in are Vogul vat-nup lexurm ‘thirty-towards
higher Greek numerals, e.g. hex-akis-khı́li-oi three (twenty three);’ and Estonian numerals
‘six-times-thousand-nom.pl.m’. such as kaheksa-tei-st-(kümmend) ‘eight-se-
A common feature of the expression of cond-from-ten’, i.e. ‘eight of the second ten
multiplication is the deletion of 1. Thus in (eighteen)’, with optional base deletion.
English one does not say a ten or one ten as In the preceding discussion the basic topic
part of the counting system, but one does say has been the internal morphosyntactic ex-
one hundred or a hundred. The reason for the pression of cardinal numbers. As has been
deletion (which might not, of course, be a seen, this internal morphosyntax ⫺ akin to
true historical deletion) is the unique mathe- processes of grammaticalization ⫺ often in-
matical property of 1, namely that only for volves expressions which have become more
a ⫽ 1 is it true that for any n, a⫻n ⫽ n. Thus or less frozen and reduced in form to a basic
the use of 1 as a multiplicand is redundant. syntax similar to that of numeral expression
In fact, it is likely to occur with higher bases as a whole in relation to a qualified noun,
which, as has been seen, tend to act as nouns its most frequent use. In its external syntax,
and can therefore be counted. Thus English numeral expressions share many properties
one thousand is the initial step in counting normally associated with the major class of
thousands. As a statistical universal, the noun-qualifiers ⫺ namely adjectives ⫺ al-
overt expression of 1 with lower bases implies though such qualities may be manifest in pat-
its overt expression in all higher bases. ently reduced form, especially in higher nu-
In multiplication, bases may also be de- merals. Typically, lower numerals may be
leted if there is sufficient other information morphologically like adjectives, and, where
(e.g. inflectional) to allow it. In Efik, a adjectives agree in categories, such as gender,
Benue-Congo language of Nigeria which has case, and number (singularity/(duality/)plu-
a vigesimal system, ı́bà is ‘two’, but àbà is rality), numerals will agree also, though fre-
‘twenty’. Here, and for other multiples of quently with idiosyncratic morphology and
tens, the initial à- (a plural inflection) permits neutralization of categories. In regard to
the omission of édı́p, 20, without any result- number there may also be agreement. The
ing ambiguity. Presumably the use of plurals numeral expression for 1 may agree with a
of numbers 3 to 10 in Semitic languages, e.g. singular noun, that for 2 ⫺ where a dual ex-
classical Arabic, permits the conclusion that ists ⫺ itself showing morphological traits of
an expression for 10 has been omitted. a dual and agreeing with nouns which are
Unlike addition and multiplication, sub- themselves in the dual. Numerals larger than
traction normally involves a separate mor- 2 also sometimes show morphological plural
pheme ⫺ usually an expression meaning characteristics and agree with a noun in the
‘without’, ‘take away’, ‘from’ or even a nega- plural.
tive with the subtrahend ⫺ to indicate its op- A widespread peculiarity is the neutraliza-
eration. An example of the last is Somali tion of number in the noun, when it is modi-
laba-ton mid-la ‘two-ten one-not (19)’. Nau- fied by a numeral, so that it is in the singular
kan, an Eskimo dialect of Siberia, even has with all numerals, e.g. Hungarian, Turkish. In
(presumably with ‘one’-deletion in this case) some instances, as in Turkish, the noun plural
the expression of 9 as ‘ten-not’. seems not to be a compulsory category.
Division, which is even more marked than Higher numbers, especially bases, are, as
subtraction, is, when observed, primarily by has been seen frequently, morphologically
substantives. The most common syntactic Where integral fractions are expressed sys-
treatment is for the thing counted to be put tematically, the numerator is an expression
in the genitive (most often the genitive plu- for a cardinal, and the denominator is fre-
ral). Another more unusual syntactic device quently an ordinal as in English three fifths.
is affixation to the numeral rather than to the The reason for the ordinal seems to be the
noun, marking the syntactic connection following: suppose one has a collection of
much like the well-known izafet construction countable objects and wishes to discover the
in Farsi. An example is Wolof nyar-i nag numerical value of three fifths of the set. One
‘two-of cow (two cows)’, also employed for counts ordinally until reaching the fifth mem-
multiples of bases, as in nyar-i temer ‘two- ber, putting the already counted objects
of hundred’. aside, then repeats the operation until either
When complex cardinal expressions qual- the set is completely counted or a remainder
ify nouns there are two major alternatives re- smaller than five is left. Hence also the com-
garding the syntactic relation of the numeral mon use of a word like ‘part’ with the de-
to the noun, where higher expressions do not nominator.
simply take plurals, when they exist, e.g.
English twenty three houses. One is that it will 6.2. Indefinite and approximative cardinals
take the construction of the expression with In addition to the numerals in the counting
the lowest numerical order as in Russian series of cardinals, there are normally expres-
dvadcat’ odin celovek ‘twenty one:nom.sg.m sions for an undetermined small or large
man:nom.sg’. The other is that it will follow amount, i.e. words like ‘few’, ‘many’, inclu-
the construction of the numeral expression sives like ‘all’, interrogatives and expressions
closest to the noun. Thus in classical Arabic, for approximatives, such as ‘about twenty’,
in which multiples of ten govern the accusa- etc. Of these, the indeterminates and inclu-
tive singular indefinite of the noun, the nu- sives often share syntactic properties of nu-
meral expression for 347, expressed as ‘three- merals, usually of the highest set if there is
hundred seven and forty’, will govern the diversity of construction within the system.
accusative singular indefinite of the quali- Like higher numerals, indeterminates and in-
fied noun. clusives are often substantivized and may
govern genitives. These syntactic properties
are often shared with the interrogative(s) for
6. Extensions cardinal numbers. Examples are Russian
of the cardinal numeral system mnogo ‘many’ and skol’ko ‘how many’, both
of which govern genitive plurals in the nomi-
6.1. Mathematical extensions native-accusative, but have agreement like
In the development of mathematics, starting adjectives in oblique cases. This is the same
from the counting of integers, numerous ex- pattern observed in higher units of the
tensions of the numerical system have been cardinal system.
made, e.g. the inventions of zero, fractional Approximatives usually employ adverbials
numbers, negative numbers, imaginary num- or other qualifiers, as about in English about
bers, etc. Of these, outside of mathematical ten. However, in a few instances there are dif-
discourse in technologically advanced socie- ferences in word order. Here it seems that the
ties, the only extension commonly found is numeral follows the noun to express indefi-
to fractions. ‘Half’ is the unmarked fraction, niteness of the number. Examples include
sometimes present when there are no other Russian and Zyryan, e.g. Russian desat’
fractional expressions. Terms for ‘half’ are jablok ‘ten apple:gen.pl (ten apples)’ versus
often derived from those for ‘to split’ or ‘to jablok desat’ ‘apple:gen.pl ten (about ten ap-
break’ and are simple lexical expressions. ples)’. Similar examples are found in numeral
Sometimes a simple expression for quarter is classifier languages, such as Bengali, in
also present. Overcounting is sometimes which, however, the numeral precedes the
found in fractions. An example is Ainu, in classifier in the exact construction and fol-
which e, expressing negation, is used for lows it in the approximative.
subtraction in the cardinal system. In similar
fashion, 1½ is expressed as e-tup ‘not-two’, 6.3. Counting and countables
2½ by e-rap ‘not-three’, etc. One may com- It may also be considered an extension of the
pare such overcounting with German expres- cardinal system in a different direction when
sions of time like halb drei for 2: 30. numerals are used with mass nouns by the
intervention of units of measure. In such in- case some discernible emphasis can often be
stances there are various indications that this noted or inferred) or, finally, one could rely
is not counting in the countable-objects solely on ‘two’ in conjunction with the plural
sense. Often, with numeral expressions larger of the noun. This last usage derives from the
than 1, the singular of the measure word is fact that in such cases a dual is not necessary,
used rather than the plural, as in German drei since the fact that a set has two objects is
Pfund Kaffee ‘three pounds of coffee’. Al- already conveyed by the numeral. This infor-
though Pfund has a plural (Pfunde) its singu- mational redundancy is certainly one of the
lar is used in this construction. The syntax factors leading to the demise of the dual in
of measure expressions is apparently always many languages and to its subsequent re-
identical to that of numeral classifiers in lan- placement by the plural.
guages which use classifiers (cf. Art. 97). In Another aspect of this relationship is his-
some instances, as in Akkadian, it is the sin- torical: the numeral ‘two’ (often in reduced
gular of the noun, rather than of the mea- form) affixed to the noun is a common
sure-word, which is used, while the noun plu- source for expression of duality of the noun
ral is employed with non-measure expres- and/or pronoun. Lithuanian, for example,
sions. Weighing figures prominently in such has developed a dual category in the pronoun
expressions, since countable objects can often by this means, as seen in mù-du ‘we-two(m)’
be weighed in a separate analytic action, and mù-dvi ‘we-two(f )’. In Tahitian, as in
e.g. apples. other Polynesian languages, the dual pro-
nouns have a suffix which derives from the
numeral expression for 2, while the plural
7. Cardinal numerals suffix is a reduced form of ‘three’, showing
and substantival number that a trial category formerly existed, and
served as the basis for the modern plural.
The category of number in substantives gives
essentially the same kind of information as
do cardinal numeral constructions, but in a 8. Non-cardinal series of numerals
far less determinate way. For example, in a
language with the categories of the singular, Ordinals, as has been noted above, are com-
dual and plural, the first of these is used monly formed from cardinals with affixes.
when one is dealing with a set of objects However, the term for the ordinal related to
which has one member, the dual when the set ‘one’ is commonly not derived from the
has two members, and the plural when it has cardinal ‘one’, but from some word meaning
more than two. A few languages of the Pa- ‘foremost’, or ‘in front’, e.g. English first.
cific have in addition a “trial” in pronouns The ordinal corresponding to ‘two’ is usually
and even an extremely rare “quadrial”. derived from the cardinal but may be de-
Usually it is more convenient to use a plu- rived, like lower ordinals, from some root
ral of a noun rather than a defined numeral meaning ‘following’, as is the case with Latin
modifier > 2, either because one does not secundus ‘second’, which is derived from
know the exact numerical value or because it sequ-i ‘follow-inf (to follow)’ There is also a
is irrelevant to the discourse. Even for some- term ‘last’ to which no cardinal corresponds.
one beset by a plague of ants, knowledge of While nominal objects are usually those
the exact number of the attacking creatures things counted, additional numeral series
is usually both uncertain and unimportant. arise from the counting of verbal actions. The
In languages with nominal duals, the use series that arise in this way are called (vari-
of the numeral expression for ‘two’ is, strictly ously) multiplicative or adverbial. The divi-
speaking, redundant, although, of course, sion between the adverbial and cardinal
one needs a numeral designating 2 in higher series of numerals (the latter of which modify
complex numbers and for the act of counting nouns) is independent of the distinction be-
itself. The identity of referential meaning of tween cardinals and ordinals. There can thus
the cardinal number and the substantival cat- be both adverbial cardinals and adverbial or-
egory leads to a situation which, in the case dinals. For example classical Latin carefully
of ‘two’, may be resolved by three alternative distinguishes between quinquies ‘five times’
constructions. In Greek of the classic period, and quintum ‘for the fifth time’. While the
one could use the dual without a numeral former is a distinct formation with its own
‘two’, one could use both forms (in which interrogative (quoties), the latter is simply the
accusative neuter singular of the correspond- thus expects that its conceptual simplification
ing ordinal quintus ‘fifth’, and does not have into a grammatical system in which verbs dis-
a distinct interrogative. Adverbial ordinals tinguish single and plural acts (with a pos-
are relatively marked in comparison to the sible third category of dual) will be corre-
adverbial cardinals, as can be seen in by the spondingly marked. The category of verbal
defectivity in the absence of an interrogative plurality does exist in a number of the
and other ways. world’s languages, being particularly com-
The expression of adverbial cardinals may mon in North America. Swadesh (1946: 325),
involve a distinctive affix to the base of ordi- in a grammatical sketch of Chitimacha, a
nary cardinal, as seen above in Latin. Com- Gulf language of Louisiana, introduces the
mon methods include the use of adverbial term occurrence number. Plurality here in-
cardinals with particular words ⫺ like Eng- cludes temporal repetition, action by many,
lish time and French fois ⫺ and the incorpo- and action on many. In regard to these latter,
ration of adverbial cardinals expressions into an ergative pattern is common in that the
the corresponding cardinal numerals, as in plural category is employed for intransitive
German dreimal. subjects and transitive objects but not for
Adverbial cardinals and their associated transitive subjects. In some languages, a cate-
ordinals naturally modify stative rather than gory of dual actions is found (Karok). Where
active verbs, since only actions (and objects) it occurs, suppletion of singular, dual, and
can be counted. Thus to be red twice is un- plural stems is often found for a restricted set
grammatical or marginal in English. When of verb stems, with the common recurrence
such an expression occurs, the interpretation of various forms of reduplication for dual
of be must be ‘become’ rather than the cop- and plural (cf. Dressler 1968; Durie 1986;
ula; perhaps the above expression might be Greenberg 1991). Sporadic lexical expression
used to mean ‘blushed twice’. Such an obser- of verbal number often occurs within the lexi-
vation clearly serves to confirm the active con, e.g. English hit vs. beat. With indefinite
versus stative distinction. plurality, e.g. English many times, it occurs
When the adverbial cardinal is expressed in many languages as a frequentative deriva-
by a noun, this is likely to indicate a general tional formation.
and repeatable action, e.g. classical Arabic The remaining numeral series to be consid-
marrat-an ‘turn-acc.sg (once)’ which functions ered all depend in one way or another on the
as an adverbial accusative. Like cardinals, relativity of the concept of unit. This has al-
adverbials can be used for indefinite quan- ready been seen in regard to the more com-
tification, e.g. often, seldom, etc. Another plex numeral systems. At one point in a deci-
method, employed perhaps only in numeral mal analysis, ‘tens’ are in effect counted, so
classifier languages, involves the employment that these, rather than individual objects, are
of what has been called a verbal classifier the de facto units of counting. A simple ex-
(Killingley 1983: 100). An example from ample of this is what is sometimes called the
Mandarin is kàn-le liǎng-yán ‘look-comple- proportional numeral, as Latin simplus,
tive two-eye (looked twice)’. There is a possi- duplus, or triplus (corresponding to English
bility here that a language might classify single, double, triple) which answers the ques-
verbs by the nouns it uses with numerals to tion “how many times as great as x?” Here
express the number of times an action takes the conception of unit varies with the
place. These nouns are often just nominal context. If the whole of something is taken as
forms of the verb, as if one said in English the point of empirical reference, a question
jumped two jumps. Such a possibility seems to arises regarding the degree of internal com-
be realized in Cantonese and other Yüe dia- plexity whose common answer can be given
lects (Hashimoto 1972: 25). in multiplicatives, e.g. Latin simplex, duplex
Parallel to the relation of cardinal numer- ‘single’, ‘double/two-fold’, describing how
als to the category of nominal number is that many units compose a larger whole which
of the adverbial cardinals to the category of may again vary with the context.
so-called verbal plurality. Since the usual If one thinks of the set more abstractly as
function of cardinal numerals is to express any unit which has a certain number of mem-
the integral number of the members of a set bers which are units at a lower level, one ar-
of objects, and, as has been shown, the adver- rives at the notion of collective numerals like
bial cardinal, which counts the number of ‘pair’, ‘triplet’, etc. Collective numerals have
times an action occurs, is highly marked, one a variety of uses which vary interlinguisti-
cally. Frequently, no collective form corre- Frequently, the total which is distributed is
sponds to ‘one’. A characteristic use of col- unknown or unexpressed, as in The soldiers
lectives is to modify pluralia tantum or nouns marched in rows of five. Since each five is a
with different meanings in the singular and kind of group, the occasional use or histori-
plural. For example, in Latin littera refers to cal change to a collective meaning is under-
a single written letter; the plural is litterae. standable.
However, the plural form litterae also means From the foregoing exposition, it seems
a single epistle (viewed as a collection of the clear that the numeral should be viewed as a
letters of which it is composed). In this sense distinctive part of speech whose core consists
it has a singular meaning despite its plural of the cardinal series. The basic syntactic
form. The collective numeral binus is used to function of this set is in qualifying nouns.
pluralize litterae in the sense of epistle; the That non-cardinal series such as ordinals be-
resulting form, binae litterae, is unambiguous long to other parts of speech, and that some
and distinct from duae litterae, meaning two numerals are treated inflectionally and syn-
written letters, such as p and r. English, of tactically like nouns, does not nullify the no-
course, tolerates the ambiguity of two letters, tion of the numeral as a distinct part of
yet utilizes collectives in disambiguating speech any more than the existence of partici-
some of its own cases of pluralia tantum, such ples, infinitives, and verbal nouns destroys
as pair of scissors vs. two pair(s) of scissors the notion of the verb as a distinct part of
and pair of pants vs. two pair(s) of pants. speech. What is distinctive about the core
Collective numerals have a variety of other numerals is that the set is inherently ordered
uses. Besides being adjectival, as when modi- by the notion of unique succession, and that
fying pluralia tantum, they can be substanti- all other numeral expressions can be defined
val or adverbial. Russian dvoje ‘two’ (collec- in relation to the cardinals. As the great nine-
tive), distinct from dva ‘two’ (cardinal), is in- teenth century mathematician Kronecker re-
flectionally a neuter singular which governs marked “God made the natural numbers.
the following noun in the genitive plural: dva Everything else is man’s invention”.
brata ‘two brothers’ versus dvóje brát’jev ‘a
pair of brothers’, though in practice the dif-
ference between the expressions has become 9. References
increasingly negligible.
Brandt Corstius, H[ugo] (1968, ed.), Grammars for
Russian vdvojem, a fossilized instrumental Number Names. Dordrecht: Reidel (Foundations of
of the collective dvoje following v ‘in’, indi- Language Supplementary Series 7)
cates joint action like English together and
Brugmann, Karl (1907), Die distributiven und die
German zu dritt. Historically, collectives tend
kollektiven Numeralia der indogermanischen Spra-
to become distributives, and Brugmann chen. Leipzig: Teubner (Abhandlungen der philol.-
(1907), in a monograph concerning Indo- hist. Klasse der königl. sächsischen Gesellschaft
European distributives and collectives, claims der Wissenschaften 25.5)
that certain proto-forms generally labelled as Corbett, Greville G. (1978), “Numerous Squishes
distributives were really interpreted as collec- and Squishy Numerals in Slavonic”. International
tives. Review of Slavic Linguistics 3.1⫺2, 43⫺73
The basic notion of distributive is closely
Dressler, Wolfgang (1968), Studien zur verbalen
related to that of division. Consider, for ex- Pluralität. Wien: Böhlau (Österreichische Akade-
ample, an ordinary deck of 52 playing cards. mie der Wissenschaften, phil.-hist. Klasse, Sit-
If there are four players in a game, such as zungberichte 259.1)
bridge, then 13 cards are dealt to each player. Durie, Mark (1986), “The Grammaticization of
To say that the players receive 13 cards each Number as a Verbal Category”. Berkeley Linguis-
is a distributive numeral expression con- tics Society: Proceedings of the Annual Meeting
sisting, in this case, of the cardinal accompa- 12, 355⫺370
nied by each (or apiece). The two other com- Greenberg, Joseph H. (1966), Language Universals,
mon methods besides adverbs like ‘each’ are with Special Reference to Feature Hierarchies. The
reduplication of the cardinal numeral, as in Hague: Mouton
Hebrew š enayim š enayim ‘two two’ (used Bib- Greenberg, Joseph H. (1978), “Generalizations
lically to describe the animals entering the about Numeral Systems”. In: Greenberg, Joseph
ark, Gen. 7: 9), and adposition, as in Russian H. (ed.), Universals of Human Language, Vol. III:
po ‘according to’, which governs the cardinal Word Structure. Stanford/CA: Stanford Univ.
number in the dative. Press, 249⫺295
Greenberg, Joseph H. (1991), “The Semitic ‘Inten- Muratori, Carlo (1938), Grammatica Lotuxo. Ve-
sive’ as Verbal Plurality”. In: Kaye, Alan S. (ed.), rona: Missioni Africane
Semitic Studies. In Honor of Wolf Leslau, Vol. I. Peano, Giuseppe (1908), Formulario Mathematico.
Wiesbaden: Harrassowitz, 577⫺587
Torino: Bocca
Gvozdanović, Jadranka (1992, ed.), Indo-European
Numerals. Berlin, New York: Mouton de Gruyter Robins, R[obert] H. (1958), The Yurok Language.
(Trends in Linguistics: Studies and Monographs Berkeley: Univ. of California Press (Univ. of Cali-
57) fornia Publications in Linguistics 15)
Hashimoto, O. K. Y. (1972), Studies in Yüe Dia- Salzmann, Zdenek (1950), “A Method for Analyz-
lects, Vol. I: Phonology of Cantonese. London etc.: ing Numerical Systems”. Word 6, 78⫺83
Cambridge Univ. Press Swadesh, Morris (1946), “Chitimacha”. In: Lin-
Hurford, James R. (1975), The Linguistic Theory of guistic Structures of Native America. New York:
Numerals. Cambridge etc.: Cambridge Univ. Press Viking Fund (Viking Fund Publications in Anthro-
(Cambridge Studies in Linguistics 16) pology 6), 312⫺336
Hurford, James R. (1987), Language and Number. Stampe, David (1976), “Cardinal Number Sys-
Oxford, New York: Blackwell tems”. Papers from the Regional Meeting of the
Ifrah, Georges (1985), From One to Zero: A Univer- Chicago Linguistics Society 12.1, 594⫺609
sal History of Numbers. English transl. by Lowell Szemerényi, Oswald (1960), Studies in the Indo-
Blair. New York: Viking Penguin [orig. French European System of Numerals. Heidelberg: Winter
1981]
Killingley, K. S.-Y. (1983), Cantonese Classifiers, Trubetzkoy, N[ikolai] S. (1939), Grundzüge der
Newcastle-on-Tyne: Grevatt and Grevatt Phonologie. Prague (Travaux du Cercle Lin-
guistique de Prague 7) [⫽21958 Göttingen: Van-
Menninger, Karl (1969), Number Words and Num- denhoeck & Ruprecht]
ber Symbols. English transl. by Paul Broneer. Cam-
bridge/MA, London: MIT Press [orig. German
2
1958] Joseph H. Greenberg, Stanford (U.S.A.)
Greenberg, Joseph H. (1991), “The Semitic ‘Inten- Muratori, Carlo (1938), Grammatica Lotuxo. Ve-
sive’ as Verbal Plurality”. In: Kaye, Alan S. (ed.), rona: Missioni Africane
Semitic Studies. In Honor of Wolf Leslau, Vol. I. Peano, Giuseppe (1908), Formulario Mathematico.
Wiesbaden: Harrassowitz, 577⫺587
Torino: Bocca
Gvozdanović, Jadranka (1992, ed.), Indo-European
Numerals. Berlin, New York: Mouton de Gruyter Robins, R[obert] H. (1958), The Yurok Language.
(Trends in Linguistics: Studies and Monographs Berkeley: Univ. of California Press (Univ. of Cali-
57) fornia Publications in Linguistics 15)
Hashimoto, O. K. Y. (1972), Studies in Yüe Dia- Salzmann, Zdenek (1950), “A Method for Analyz-
lects, Vol. I: Phonology of Cantonese. London etc.: ing Numerical Systems”. Word 6, 78⫺83
Cambridge Univ. Press Swadesh, Morris (1946), “Chitimacha”. In: Lin-
Hurford, James R. (1975), The Linguistic Theory of guistic Structures of Native America. New York:
Numerals. Cambridge etc.: Cambridge Univ. Press Viking Fund (Viking Fund Publications in Anthro-
(Cambridge Studies in Linguistics 16) pology 6), 312⫺336
Hurford, James R. (1987), Language and Number. Stampe, David (1976), “Cardinal Number Sys-
Oxford, New York: Blackwell tems”. Papers from the Regional Meeting of the
Ifrah, Georges (1985), From One to Zero: A Univer- Chicago Linguistics Society 12.1, 594⫺609
sal History of Numbers. English transl. by Lowell Szemerényi, Oswald (1960), Studies in the Indo-
Blair. New York: Viking Penguin [orig. French European System of Numerals. Heidelberg: Winter
1981]
Killingley, K. S.-Y. (1983), Cantonese Classifiers, Trubetzkoy, N[ikolai] S. (1939), Grundzüge der
Newcastle-on-Tyne: Grevatt and Grevatt Phonologie. Prague (Travaux du Cercle Lin-
guistique de Prague 7) [⫽21958 Göttingen: Van-
Menninger, Karl (1969), Number Words and Num- denhoeck & Ruprecht]
ber Symbols. English transl. by Paul Broneer. Cam-
bridge/MA, London: MIT Press [orig. German
2
1958] Joseph H. Greenberg, Stanford (U.S.A.)
dent status, becoming unstressed pronominal Articles, likewise, may be free or bound;
clitics (cf. Art. 41), and ultimately lose their when bound, as with bound pronouns, it may
pronominal status as well, becoming agree- be difficult to determine when they are best
ment affixes (cf. Givón 1976). analyzed as bound articles and when they are
Since agreement affixes share inherent cat- best analyzed as definite nominal affixes. For
egory distinctions with pronouns, it is often example, in Danish, the definite morpheme
difficult to determine the precise analysis of a appears bound to an unmodified noun but
morpheme with typical pronominal category free with a modified noun: hus ‘house’, huset
information such as person and number. ‘house:def (the house)’, det lille hus ‘def little
However, accompanying the change in mor- house (the little house)’ (Anderson 1985:
phological independence are phonological 178). In Norwegian, the bound definite mor-
and functional changes. Phonological changes pheme remains attached to the noun in modi-
include loss of ability to carry stress and ulti- fied forms in which a free definite article also
mately fusion with other bound morphemes appears: det store huset ‘def big house:def
to create portmanteau inflections. Changes in (the big house)’. Using the same criterion
the communicative function of the mor- that is used to distinguish bound pronouns
phemes occur as well, with free pronouns from inflectional affixes, the Danish bound
having an emphatic, contrastive or discourse- form may be considered an article on the ba-
discontinuative function, while the less inde- sis of its complementary distribution with the
pendent bound pronouns and agreement in- free article, but the Norwegian bound form
flections have a discourse-continuative func- may be considered a nominal affix on the ba-
tion (Givón 1984: 354). A useful functional sis of its cooccurrence with the free definite
distinction is whether the marker is affected article.
by first mention in discourse: if it is, then it
functions pronominally, and if not, it func-
tions as an agreement marker (Ingram 3. Pronouns
1978: 232). In-depth syntactic analysis is
likely to be necessary to make the distinction 3.1. Personal pronouns
between bound pronoun and agreement in- Arguably, the most central members of the
flection. The primary syntactic diagnostic is category pronoun are the personal pronouns,
the presence or absence of a coreferential free which are a near-universal grammatical cate-
pronoun or noun phrase in the same simple gory, although some languages, such as Ma-
clause as the morpheme in question: if a free lay, are reported to prefer self-deprecating or
pronoun or noun phrase may not occur in honorific nouns over pronouns in polite con-
the same clause, then the morpheme in ques- versation (Schachter 1985: 27). Typologically,
tion functions pronominally, while if it may the personal pronouns typically exhibit more
occur, the morpheme functions as an agree- pronominal feature distinctions than other
ment inflection (cf. Bresnan & Mchombo pronominal systems. The examination of per-
1987). However, syntactic accounts which as- sonal pronouns will be relatively more de-
sume the presence of null pronouns may ana- tailed than the discussion of other pronoun
lyze this configuration differently (cf. Stump systems. This is justified by the relative
1984; Hendrick 1988). It may not always be centrality of this category, by the relatively
the case that a clear distinction can be made more attention and descriptive detail it re-
for all bound morphemes one way or the ceives in individual language studies, and the
other. For example, subject markers in relative attention it has received from lin-
Chichêwa may function either as bound pro- guists in typological studies. Person and
nouns or as agreement affixes (Bresnan & number are taken to be the essential categor-
Mchombo 1987). This is not surprising, given ial distinctions that are found in all systems
the gradual continuum of development from of personal pronouns (Greenberg 1963: 96;
free pronoun to bound pronoun to agree- Ingram 1978: 327), though they are not nec-
ment affix. The widespread occurrence of essarily contrastive in all subsystems of a
bound morphs with pronominal category in- pronominal system. Other distinctions com-
formation and multiple functions indicates a monly shown in pronominal morphology are
complex relationship between independent animacy, gender and case. Animacy and gen-
pronouns and the dependent forms at a par- der are considered to be inherently nominal,
ticular stage of development. since they are typically determined by the lex-
ical and semantic properties of the nominal However, there is also ample evidence in
antecedent, and case is considered to be a re- some systems to indicate that the language
lational property of nouns. treats 3 forms as part of the personal para-
digm. In Balante there is a two-way person
3.1.1. Person distinction of 1 vs. 2/3 which groups forms
Person (see Art. 96) refers to distinctions with narrative reference (3) and forms with
among the narrative participant-referents of speech act reference (2) in contrast to a form
the pronouns. Three persons are traditionally with exclusively speech act reference (1) (Wie-
distinguished: first person, defined by its semann 1986: viii). Furthermore, verbal per-
function of referring to the speaker; second son inflections often contain an overt three-
person, defined by its function of referring to way contrast that places 3 forms on a par
the addressee; and third person, defined by with 1 and 2 forms (Greenberg 1986: xixf). It
its function of referring to a participant in is somewhat common for languages to not
the events or states denoted in the discourse have independent morphemes for third per-
who is not a speech act participant. The nu- son. In this case, the gap in the paradigm
merals 1, 2, and 3 will be used to abbreviate may be filled by borrowing from the corre-
first, second and third person, respectively. sponding demonstrative paradigm, or it may
The personal pronouns are divided into speech not be filled, resulting in a clause with no
act participants (1 and 2) and non-speech act overt pronoun. For languages of this type, it
participants (3). Third person referents will may be best to analyze the paradigm as in-
be called “narrative participants” here, to in- complete rather than to consider a phonolog-
dicate their origin as arguments in the events ically null morpheme as a member of the
and states denoted in utterances of the first morphological paradigm (Mithun 1986).
person. This functional distinction between 2 is intermediate between 1 and 3 (Green-
speech act and narrative participants, the ab- berg 1986: xx). The two-way person contrast
sence of third person forms in the personal in Balante which distinguishes 1 from 2/3
pronoun system of some languages, and the would seem to support this claim. Other
use of a single morphological form for both similar evidence is the neutralization of the
distinction between 2 and 3 in Chitimacha
personal pronoun and deictic pronoun in
verb inflections for subject and the common
others, leads some linguists to conclude that
1 vs. non-1 opposition which is the basis of
third person is a “non-person” (Benveniste
two-term deictic systems (Greenberg 1986: xx).
1956: 347). There are also formal distinctions
Thus, the three-way person opposition in
which can be noted between 1 and 2 vs. 3.
personal pronouns and related verbal inflec-
Cross-linguistically, for example, 1 and 2 are tions is apparently structured rather than
more likely to carry overt number marking, simple, with formal distinctions opposing 1
while 3 is more likely to carry overt gram- vs. 2/3 as well as 1/2 vs. 3.
matical gender marking. Either 1 or 2 may Functionally, pronominal categories should
take precedence in determining the morpho- serve to allow the addressee to identify more
logical form of plural pronominal forms easily the narrative or speech act partici-
(Croft 1990: 136f.). For example, 1 takes pre- pant(s) referred to. Thus, it is often claimed
cedence in the well-known Indo-European that (personal) pronouns agree in number
systems ⫺ e.g., in Eng. she and I (3 & 1) or and gender/class with their antecedents or
you and I (2 & 1) both correspond to the with their referents. While this is generally
plural form we (1 ⫹ X), while you and she true, a more precise account of the matching
(2 & 3) corresponds to you (pl.) (2 ⫹ X). On in categorial information between a pronoun
the other hand, in Southwestern Ojibwe and its antecedent or referent must take ac-
(Schwartz & Dunnigan 1986) and other Al- count of the fact that both overt and covert
gonquian languages, you and I and you and grammatical features of the antecedent and
he/she correspond to the plural forms kiina- also semantic features of the referent may
wint (2 ⫹ 1) and kiinawaa (2 ⫹ non-1), while play a role in determining the appropriate-
she/he and I corresponds to the plural form ness of the form of the pronoun.
niinawint (1 ⫹ non-2). Apparently, in no sys-
tem does 3 take precedence over 1 or 2 to 3.1.2. Number
determine the selection of morphemes or the Number (see also Art. 100), like person, is as-
form of plural morphemes in a similar man- sumed to be present in virtually all systems
ner. of personal pronouns. In the encoding of
number in pronominal systems, a primary and all inanimates have no overt form
distinction is made between languages distin- (Schachter & Otanes 1972). In still other lan-
guishing number lexically and those distin- guages, the gender distinction is more exten-
guishing it morphologically (Forchheimer sive and includes at least some forms in the
1953). Some examples of systems in which speech act participant set. In Spanish, for ex-
number is encoded lexically are Eng. I vs. we, ample, first and second person plural familiar
or Swahili wewe ‘2.sg’ vs. ninyi ‘2.pl’. In (in dialects where these forms are used) dis-
forms where number is marked morphologi- tinguish two genders: exclusively feminine
cally, there may be an inflection for plural and other (traditionally called ‘masculine
only, as in Kottish ai ‘1 (I)’ vs. aj-on ‘1-pl plural’ but used from any set which is not
(we)’ (Forchheimer 1953: 48), or both singu- exclusively feminine; cf. Art. 25), as shown in
lar and plural may be marked morphologi- (1) to (3), using first person for illustration.
cally, as in Lappish m-un ‘1-sg (I)’ vs. m-in
‘1-pl (we)’ (Forchheimer 1953: 53). Lexical (1) Yo, Raúl / yo, Juanita
encoding of plurals seems to correlate with 1.sg Raúl / 1.sg Juanita
first person, while morphological encoding ‘I, Raúl / I, Juanita’
seems to correlate with third person when a (2) Nosotros, los obreros / nosotros, Raúl y
system has both types of plural formation. Juanita / *nosotros, las profesoras
An example of this is Osmali Turkish, in 1.m:pl def.m:pl worker:m:pl / 1.m:pl
which first and second person form plurals Raúl and Juanita / 1.m:pl def.f:pl profes-
with the suffix -iz, while third person shares sor:f:pl
the number suffix -lar with nouns (Forch- ‘We, the workers / we, Raúl and Juanita /
heimer 1953: 54). This is consistent with the *we, the (female) professors’
shared function of third person personal pro-
nouns and nouns as denoting discourse parti- (3) Nosotras, las profesoras / *nosotras, los
cipants. obreros / *nosotras, Raúl y Juanita
Within a system of reference tracking, the 1.f:pl def.f.pl Professor:f:pl / 1.f:pl
potential for encoding person and number def.m:pl worker.m.pl / 1.f.pl Raúl and
distinctions in a single pronominal form Juanita
leads to a set of formal oppositions in some ‘We, the (female) professors / *we, the
languages which distinguish between refer- workers / *we, Raúl and Juanita’
ence sets which include the speaker and ad- Other languages extend gender distinc-
dressee, called ‘inclusive’, and sets which in- tions into the speech act participants in the
clude the speaker and exclude the addressee, singular. In Hausa, for example, all para-
called ‘exclusive’. For example, in Samoan o- digms of personal pronouns distinguish gen-
tsaa-’ua (1 ⫹ 2) contrasts with o-maa-’ua (1 der in second and third person singular, but
⫹ 3), and ko-tsaa-suo (1 ⫹ 2 ⫹ X) contrasts gender is not distinguished in the plural. An
with o-maa-tsou (1 ⫹ 3 ⫹ 3 ⫹ …) (Givón example of this is the Hausa independent
1984: 358). pronoun paradigm shown in Tab. 76.1 (New-
man 1987: 714):
3.1.3. Gender
Given the referential function of pronouns sg. pl.
and given the potential for reference to any
of a set of narrative participants, forms 1st nı́i múu
which limit the choice of referent have a dis- 2nd kái
tinct communicative function. Thus, it is kúu
common but not universal that pronominal 2ndf. kée
forms will show gender distinctions (see 3rd m. shı́i
Art. 98), usually based on a combination of súu
semantic or grammatical categories. 3rdf. ı́tá
In some languages, a simple animate-inan-
imate distinction is made in third person by Tab. 76.1: Hausa independent pronouns
the contrast between an overt morpheme and
π. In Tagalog, for example, third person pro- The Spanish and Hausa paradigms show
nouns are used for human and pet animal re- that gender contrasts may occur in the singu-
ferents, while referents to non-pet animals lar but not the plural or vice versa for speech
act participants. (In Spanish, the extension of masculine/feminine grammatical gender dis-
the masculine/feminine gender contrast into tinction in nominals. A good example of the
the plural forms is quite obviously a function elaboration of such a system is the correla-
of the transparent morphology of these tion between pronominal and nominal gen-
forms, which are analyzable into first and se- der found in the Bantu languages. In these
cond person familiar bound object pronouns, languages, the nouns are divided into a large
nos ‘us’ and vos ‘you’ (pl.) combined with the number of classes which distinguish roughly
noun stem otro ‘other’ which shows morpho- human vs. non-human. For each human
logical variation for gender in its nominal class, there is a corresponding 3 personal pro-
form (otro m. sg., otros m. pl., otra f. pl., noun; for each non-human class, there is a
otras f. pl.). The use in Spanish of the same corresponding demonstrative pronoun which
form for sets whose members are all gram- may substitute where a personal pronoun
matically masculine and for sets whose mem- would appear (see the discussion and exam-
bers are mixed (both masculine and feminine ples in 3.3).
included in the set) is a common pattern, and It is sometimes the case that pronouns
represents a common strategy for resolving continue to maintain gender contrasts which
the problem of how to encode mixed sets, a are not found in the nominal system. In
“resolution rule” (cf. Corbett 1983; 1991). Modern English, for example, the three-way
The lack of a morphological gender contrast distinction of the third person singular per-
in the Hausa plural pronouns can also be sonal pronouns is not based on a grammati-
viewed as a strategy for resolving the gender cal categorization of nouns in English; rather,
conflict when referring to mixed sets, in the the current pronominal distinction is primar-
sense that this system simply encodes all sets ily notional: male referents high in animacy
with more than one member in the same way, are referred to using he, female referents high
irrespective of the gender of the members. in animacy are referred to using she and all
Third person can be considered the locus others are referred to using it, with a few con-
of morphological gender in personal pro- ventional exceptions such as reference to
nouns, since some languages have gender ships using she.
contrasts only in third person, and it is rarely It seems that there is a good general corre-
if ever the case that a language will have gen- lation between nominal gender and pronomi-
der distinctions in the first and second person nal gender, in that languages with morpho-
but not in the third person. There is another logically marked nominal gender strongly
sense in which the third person is the source tend to have pronominal systems in which
of gender in the personal pronouns, and that similar distinctions are found.
is that the gender of the personal pronoun is
based on the grammatical or semantic cate- 3.1.4. Case
gorization of its referent ⫺ that is the essen- Case (see Art. 102) in pronouns, like case in
tial distinctions of the pronouns into speaker, nouns, is not inherent, but rather is derived
addressee and other imposes only the prag- from the grammatical relations and semantic
matic condition that the speech act partici- roles the participant denoted by the pronoun
pants denote human beings (or, by extension, carries in the state or event of its clause. In
other entities to which human language use is pronouns, an example of this is the German
attributed). All other distinctions, then, come partial paradigm given in Tab. 76.2.
from the way in which the nominal system of
the language categorizes lexical items, or the nom. gen. acc. dat.
biological gender or animacy of the referent
of the pronoun. For example, Spanish nomi- 1st ich meiner mich mir
nals distinguish two grammatical genders, 2nd du deiner dich dir
masculine and feminine, as mentioned above. 3rd m. er seiner ihn ihm
The pronoun system likewise distinguishes 3rdf. sie ihrer sie ihr
two genders depending on the grammatical 3rd n. es seiner es ihm
gender of the lexical antecedent or the biolog- 1st pl. wir unser uns uns
ical gender of the referent. (Third person sin- 2nd pl. ihr euer euch euch
gular nominative pronouns have a third form 3rd pl. sie ihrer sie ihnen
ello traditionally called ‘neuter’, which is
used for abstract concepts.) Hausa also has a Tab. 76.2: German personal pronouns
As with gender distinctions in pronouns, a general iconic strategy for exploiting the
case distinctions are primarily derivative, in morphological features of the personal pro-
that they correlate closely with case distinc- noun system by using features typically asso-
tions in the nominal morphology. In Ger- ciated with referents removed from the ca-
man, for example, both nominal and pro- nonical second person singular addressee,
nominal paradigms distinguish four cases, either third person, a non-speech act partici-
and both have more distinctions in the singu- pant, or plural, thus distancing the linguistic
lar than the plural, although the paradigms form from association with the one-on-one
differ in which distinctions of the singular are nature of the interaction between speaker
neutralized in the plural. and addressee (Head 1978; Green 1987: 253).
One general area where a discorrelation The next three sections deal with other
between nominal and pronominal case occurs categories of pronouns which are closely re-
systematically is when a language has a split- lated to the personal pronouns: possessive
case system that is based on the Animacy Hi- pronouns (3.2), demonstrative pronouns
erarchy. This happens in a language where (3.3) and reflexive/reciprocal pronouns (3.4).
the nominal system shows an ergative-abso-
lutive case-marking pattern but where part or 3.2. Possessive pronouns
all of the pronominal system shows a nomi- Possessive pronouns carry person and number
native-accusative pattern. The favoring of a and are thus closely related to the personal
nominative-accusative pattern for pronouns pronouns; they may also show morphologi-
even when nouns show an ergative-absolutive cal case, when they are free pronouns. In ad-
pattern is claimed to be a consequence of the dition to person and number of the posses-
relatively higher position of personal pro- sor, they may also mark categories of the
nouns in the Animacy Hierarchy (cf. Sil- understood possessed, as in Span. el mı́-o
verstein 1976; Croft 1990: 111⫺117). ‘def.m.sg poss.1.sg-m.sg’, las suy-a-s ‘def
poss.3-f-pl’ Thus, el mı́o is interpreted as
3.1.5. Definiteness ‘that thing of mine which is masculine singu-
In a speech act, the referents of first and se- lar’, while las suyas is interpreted as ‘that
cond person pronouns are uniquely identifi- thing of his/hers/theirs which is feminine plu-
able and thus inherently definite. This is not ral’. Likewise, in Southwestern Ojibwe, the
so for third person pronouns, where the per- possessive pronouns are inflected for person
sonal pronouns contrast with indefinite and number of the possessor and animacy of
forms. The third person personal pronouns the possessed, as shown in intayi’iim ‘1.poss:
are used when the speaker expects their refer- 3.inan.sg (my inanimate object)’, kitaya’aa-
ents to be uniquely identifiable to the ad- mak ‘2.poss:3.anim.pl (your (sg.) animate ob-
dressee. Thus, in addition to the features of jects)’ (Schwartz & Dunnigan 1986: 300).
person and number, all personal pronouns Possessive pronouns may be bound to the
are definite. possessed nouns, as in Hausa lı́ttáafı̀-n-kà
‘book.m-linker.m-2.sg (your book)’ (New-
3.1.6. Social distinctions encoded in man 1987: 714).
pronominal morphology
A common distinction found in many pro- 3.3. Demonstrative pronouns
nominal systems is the distinction between Just as the personal pronouns encode primar-
familiar and respectful forms, where the se- ily the speech act and discourse status of enti-
lection of a form of address is determined by ties, the demonstrative pronouns encode pri-
the social status or relationship of the speech marily the spatial relations among entities,
act participants, with the exact circumstances and, derivatively, the psychological proximity
of usage assumed to be culturally deter- of entities in discourse. The common basis
mined. This distinction may be made lexi- for all such systems is spatial proximity to
cally, but it is commonly derivative on the speaker, illustrated in a very simple two-term
category information already encoded in per- system like English in the contrast between
son and number, as in the tu/usted contrast pronominal this and pronominal that, where
in Spanish or the du/Sie contrast in German, this refers to an object near the speaker, or in
where Span. usted is grammatically third per- discourse, to an object near to the speaker’s
son singular and Germ. Sie is grammatically attention, and that refers to an object not
third person plural. These examples illustrate spatially or attentionally near the speaker. A
simple three-way system which is based on languages, illustrated for Swahili for Class 7/
distance from speaker is Span. éste/ése/aquél, 8, where the noun class affix is suffixed to the
with éste being the nearest member of the proximal demonstrative base hi-, as in hi-ki
contrast, and aquél used only in contrast to ‘this-cl7 (this thing)’, hi-vi ‘this-cl8 (these
ése, with aquél denoting the farthest entity in things)’.
the contrast (cf. Art. 95).
In addition to the inherent deictic features 3.4. Reflexive and reciprocal pronouns
of person and space, demonstrative pronouns Reflexive pronouns are defined as members of
may encode morphologically other features a special pronominal paradigm whose ante-
such as number (Eng. this sg., these pl. / that cedent must appear in the most restricted
sg., those pl.), gender (Span. éste m., ésta f., syntactic domain, the simplest and most uni-
esto n.), and syntactic categories such as case. versal case being a special direct object pro-
All three of these variables are illustrated in noun which is coreferential to the subject of
the German paradigm in Tab. 76.3. its clause, as in He cut himself. There has
been much syntactic research on the defini-
m. f. n. pl. tion and universality of this restricted do-
main (Reinehart & Reuland 1991); there is
nom. dieser diese dieses diese some evidence that the domain may be corre-
gen. dieses dieser dieses dieser lated or tends to be correlated with the mor-
dat. diesem dieser diesem diesen phological form of the pronouns (Progovac
acc. diesen diese dieses diese 1993). Reflexive pronouns must be distin-
guished from other forms which occur in a
Tab. 76.3: German demonstrative paradigm of reflexive syntactic context (the limited do-
dieser ‘this’ main mentioned above) and which produce
a coreferential interpretation. Examples of
In the German paradigm, the demonstra- such forms are Russ. sebja, or Jap. zibun,
tive pronouns show the same morphological which show no morphological variation and
category distinctions as the personal pronouns can cooccur with all variants of person and
and nouns, but this is not always so. For in- number of its formal antecedent, or Span. se,
stance, in Spanish, personal pronouns show which occurs only with a third person ante-
morphological case variation but demonstra- cedent. These forms are thus most appropri-
tive pronouns do not, while in English, third ately classified morphologically as reflexive
singular pronouns show formal variations particles rather than reflexive pronouns
corresponding to notional gender while de- (however, a consideration of reflexive pro-
monstrative pronouns do not. In these in- nouns and reflexive particles together pro-
stances, demonstrative pronouns show less duces some typologically and functionally
variation than the personal pronouns. This significant results (Franks & Schwartz 1994).
may be related to the typical use of demon- An example of a reflexive pronominal par-
strative pronouns (in contrast to personal adigm is that of English, given in Tab. 76.4.
pronouns), which is to refer to entities on the
low end of the animacy scale ⫺ typically in-
sg. pl.
animates. These elements are less likely to fill
the same range of grammatical roles that 1st myself ourselves
higher animates do, and as a result, less likely 2nd yourself yourselves
to be in focus, be continuing discourse topics, 3rd m. himself
etc. Therefore, they may be less likely to 3rd f. herself themselves
carry the morphological marks of number 3rd n. itself
and case which are typically correlated with
the marking of higher animacy categories. Tab. 76.4: English reflexive pronouns
Languages in which nominal morphology
distinguishes gender for inanimates also are In English, each member of the reflexive
likely to have a robust gender morphology in paradigm consists of a person and number-
the demonstrative system, where the demon- inflected possessive determiner form (my, our,
strative pronouns may function in place of your) for speech act participants or pronomi-
third person pronouns for the inanimate nal form (him, her, it, them) for narrative
classes. This is a typical pattern in the Bantu participants, with the addition of gender in-
flection in third person singular forms. These (some, any), or negative (no) quantifier and a
combine with a number-inflected morpheme base of -body or -one for higher animates and
-self, the plural of which is the same as that -thing for lower animates and inanimates.
of the nominal paradigm. It is a common For non-counts and plurals, the quantifier
morphological pattern to form reflexive pro- alone fulfills the pronominal function, except
nouns from personal pronouns or possessives for the negative partitive none. We see, then,
plus a nominal stem like head or body; some that animacy and number are morphological
languages which form reflexives this way in- distinctions of the indefinite pronouns in
clude Fula and Akan (Schachter 1985: 28f.). English. In German, animacy is distinguished
Reciprocal pronouns also occur in a limited (jemand ‘anybody, somebody’ vs. etwas
syntactic domain. They require a notionally ‘something’ and niemand ‘nobody’ vs. nichts
plural subject and are interpreted as recipro- ‘nothing’) but number is not. We have seen
cal actions or events, as in we helped each in other German pronominal paradigms that
other. In some languages, like English, reflex- case inflection is very robust; however, this
ive and reciprocal forms are distinct and un- is not so in the indefinite pronouns. Rather,
related. In other languages, reciprocal rela- jedermann ‘everybody’ has a genitive form
tions are expressed using the same morpho- jedermanns but no other case inflection, and
logical means which the language uses to ex- jemand and niemand inflect for genitive and
press reflexive relations, so that, for example, only optionally for dative and accusative.
in Spanish, se ayudaron ‘refl help:perf.3.pl’ None show noun class variation.
may mean either ‘they helped themselves’ or In Spanish, on the other hand, the univer-
‘they helped each other’. Reciprocal forms sal pronoun todo ‘all’ shows number and gen-
may also be derivationally related to reflexive der distinctions, while the partitive pronoun
forms but distinct from them, as in Akan, alguien ‘somebody’ shows number but not
where the reciprocal is formed by doubling gender, and algo ‘something’ shows neither.
of the reflexive (Schachter 1985: 29). In the negative partitives, nadie ‘nobody’ and
nada ‘nothing’ show neither. The contrasts of
3.5. Indefinite pronouns alguien vs. algo and nadie vs. nada show an
animacy contrast. Thus, in Spanish, gender,
All of the pronouns discussed thus far have,
number and animacy are morphological cate-
in their pronominal use (i.e., excluding deic-
gories of the indefinite pronouns, with plural
tic uses where the referent is determined from
correlating with animacy in one part of the
the external context of the speech event), a
system but not another.
unique antecedent in the speech act partici-
pants or in the participants previously intro- 3.6. Interrogative pronouns
duced into the discourse context. Another Interrogative pronouns are defined as pro-
category, the indefinite pronoun, is used to re- nominal forms used in questions in which the
fer to non-uniquely identified members of a speaker requests that the addressee identify
set. Since the referents can only be narrative the referent. It is claimed that all languages
participants, one can conceive of the third have interrogative pronouns (Ultan 1978:
person pronouns as divided into definite (the 228). Their function as focus of the clause in-
personal pronouns) and indefinite. These can fluences their morphological form, in that
be further divided into “universal” and “par- they are almost always free rather than
titive” (Quirk & Greenbaum 1973: 108f.). bound. Their discourse function also tends to
The universal pronouns refer, collectively or limit their morphological distinctions to syn-
individually, to all members of the set, and tactic categories, because it is less likely that
the partitives refer to individual members of the speaker can anticipate the nominal fea-
the set. Another useful distinction, more tures of the referent. Animacy is a systematic
comprehensive than the universal/partitive, is exception to this claim. An animacy distinc-
the “ignorative” (cf. Art. 72). tion is almost always present in the interroga-
The category of quantifiers plays a large tive pronouns of a language (Ultan 1978: 229).
and relatively transparent role in the forma- For example, in German, the three gram-
tion and semantics of these pronouns. In matical genders are not distinguished in the
English, the quantifiers each, some and all interrogative pronouns, but animacy is: wer?
can function pronominally, and in addition, ‘who?’ vs. was? ‘what?’. Since case is a seman-
there are compound forms where the singu- tico-syntactic feature, we would expect that
lars consist of a universal (every), partitive interrogative pronouns will show case mor-
guages which do not, it is common for the Languages also vary as to whether they
demonstrative articles to function as overt have articles as a category; if no such category
markers of definiteness in contexts where this occurs, the discourse function of the definite
is important, just as the demonstrative arti- article may be taken over by a demonstrative
cles are sometimes used in the absence of article. When this happens, it is ordinarily a
third person personal pronouns (other non- distal demonstrative, with variation on which
morphological means such as word order distal demonstrative is favored when a lan-
also compensate to indicate definiteness in a guage has more than one (cf. Greenberg
language without definite articles). 1986: 282). This is the case in German, where
Indefinite articles function to mark noun there is no difference between the demonstra-
phrases which are ‘type identifiable’ in that tive determiners and the definite articles.
the addressee must mentally access a repre-
sentation of the type of object described by 4.2. Possessive articles
the expression (Gundel & Hedberg & Za- Possessive articles are closely related to the
charski 1993: 276). They are thus similar to personal pronouns and the possessive pro-
indefinite pronouns in that they are used to nouns. Like personal pronouns and posses-
identify type rather than individual. Not all sive pronouns, they have person and number
languages which have definite articles also morphology, and like possessive pronouns,
have indefinite articles; rather, in some, the they may be inflected for categories of the
contrast is between a definite article and an possessed noun, leading to the term ‘posses-
unaccompanied noun; this is the case, for ex- sive adjective’, which is sometimes used for
ample, in Welsh dyn ‘man/a man’, y dyn ‘the them. In Spanish, for example, the possessive
man’ (King 1996: 15). Another variation on articles are inflected for the number of the
this is found in a language like German, possessed noun: mi libro ‘poss.1(sg) book(sg)
where the indefinite article ein has a full case (my book)’, mi-s libro-s ‘poss.1-pl book-pl
paradigm in the singular but lacks a plural. (my books)’.
For much of the discussion below, it is conve-
nient to combine definite and indefinite arti- 4.3. Demonstrative articles
cles in examining their morphology. Demonstrative articles are articles which nec-
Articles need not show morphological essarily encode features of spatial reference.
variation: in English, the definite article the There is a very high form correlation cross-
and the indefinite article a/an show no varia- linguistically between demonstrative pro-
tion for either features of the head noun they nouns and demonstrative articles, as well as
are in construction with or for the grammati- a close historical relationship. The spatial
cal relation of the nominal construction. system of demonstrative articles is essentially
Spanish articles are inflected for the gender the same as that for pronouns: relation to
and number of their head nouns, as shown speaker is a property of all systems, with the
in Tab. 76.5. possibility of relation to addressee being en-
coded as well. Likewise, the inflectional cate-
sg. pl. gories associated with demonstrative articles
correspond very closely to those found in the
def.m. el los parallel demonstrative pronouns.
def.f. la las Demonstrative articles may show distinc-
def.m. un unos tions in number and class, as illustrated in
indef.f. una unas the Spanish paradigm in Tab. 76.6:
One might expect that a language with a coverable antecedent, while indefinite pro-
more elaborate gender system may likewise nouns including interrogative pronouns, in-
have this reflected in its demonstrative arti- definite pronouns and negative pronouns
cles. Thus, in Swahili, the demonstrative arti- substitute for unknown or nonexistent parti-
cle is formed from the near demonstrative h- cipants. Definite articles, deictic articles and
plus the noun class marker or noun class possessive articles likewise require a dis-
marker plus the distal -le, as in ki-su hi-ki course-recoverable antecedent for their noun
‘cl7-knife hi-cl7 (this knife)’, ki-su ki-le phrase, while non-definite articles including
‘cl7-knife cl7-le (that knife)’, etc. Likewise, interrogative, indefinite and negative articles,
a language with a robust morphological case like the corresponding pronouns identify the
system likely has demonstrative articles noun phrase as unknown to the speaker or
which show case morphology, as in German, nonexistent. There is a significant form corre-
illustrated in Tab. 76.3; this example also il- lation between pronouns and articles: the
lustrates the very close morphological paral- same or similar morphemes are often used in
lel between demonstrative pronouns and de- the demonstrative, interrogative, and nega-
monstrative articles. tive systems of both, representing both his-
torical development and cognitive associa-
4.4. Interrogative articles tion. The categorial information encoded in
Interrogative articles function to ask for fur- the morphology of pronouns and articles is
ther identification of the referent of the head almost identical. Since both categories relate
noun, as which boy? what books? whose participants in discourse to participants in
gloves? Like interrogative pronouns, it is typ- the utterance in which they are used, person
ical for interrogative articles to formally dis- is found in the personal pronouns and the
tinguish animate from inanimate, as in the possessive articles. Both pronouns and arti-
English example given previously. cles are nominally-oriented, either substitut-
In Spanish, interrogative determiners may ing for noun phrases or occurring in noun
show number variation corresponding to the phrases; they may encode the nominal cate-
number of the head noun, but this isn’t uni- gories of number and gender/class, as well as
form: ¿quál? ‘which?’ shows number varia- the relational categories of case and deixis
tion only: ¿cuál libro? ‘which(sg) book:m.sg’ which are associated with noun phrases.
vs. ¿cuáles libros? ‘which:pl book:m:pl (which
books?)’, while ¿qué ‘what?’ shows no varia-
6. References
tion: ¿qué libro? ‘what book?’, ¿qué libros?
‘what books?’ There is no interrogative deter- Anderson, Stephen (1985), “Inflectional Morphol-
miner corresponding to Eng. whose in Span- ogy”. In: Shopen, Timothy (ed.), Language Typol-
ish; here, possessors are questioned using the ogy and Syntactic Description, Vol. III: Grammati-
nominal pattern ¿de quién? ‘of whom?’ rather cal Categories and the Lexicon. Cambridge: Cam-
than the pattern of the Spanish possessive de- bridge University Press, 150⫺201
terminers. Benveniste, Emile (1956), “Le nature des pro-
In Hausa, gender variation is found in the noms”. In: Halle, Morris (ed.), For Roman Jacob-
singular but not the plural: wànè which.m, son. The Hague: Mouton, 340⫺347
wàcè which.f, wàHànnè which.pl, correspond- Bresnan, Joan & Mchombo, Sam (1987), “Topic,
ing to the general pattern found throughout Pronoun and Agreement in Chichêwa”. Language
the nominal and pronominal system. As ex- 63.4, 741⫺782
pected, languages with strong case morphol- Comrie, Bernard (1987, ed.), The World’s Major
ogy, like German, the interrogative determin- Languages. New York: Oxford University Press
ers likewise show case morphology. Corbett, Greville (1983), “Resolution Rules: Agree-
ment in Person, Number and Gender”. In: Gazdar,
Gerald & Ewan Klein & Pullum, Geoffrey (eds.),
5. Parallels and correlations Order, Concord and Constituency. Dordrecht:
Foris, 175⫺206
Pronouns and articles differ in their distribu- Corbett, Greville (1991), Gender. Cambridge: Cam-
tions, but share a common basic function of bridge University Press
relating a noun phrase to participants in dis- Croft, William (1990), Typology and Universals.
course. Definite pronouns such as the per- Cambridge: Cambridge University Press
sonal pronouns, demonstrative pronouns, Forchheimer, Paul (1953), The Category of Person
and relative pronouns require a discourse-re- in Language. Berlin: de Gruyter
Franks, Steven & Schwartz, Linda (1994), “Bind- Mithun, Marianne (1986), “When Zero Isn’t
ing and Non-distinctness: A Reply to Burzio”. There”. Proceedings of the Berkeley Linguistics So-
Journal of Linguistics 30, 227⫺243 ciety 12, 195⫺211
Givón, Talmy (1976), “Topic, Pronoun and Gram- Newman, Paul (1987), “Hausa and the Chadic
matical Agreement”. In: Li, Charles (ed.), Subject Languages”. In: Comrie (ed.), 705⫺723
and Topic. New York: Academic Press, 149⫺188 Progovac, Ljiljana (1993), “Long-distance Reflex-
Givón, Talmy (1984), Syntax: A Functional-typo- ives: Move-to-INFL vs. Relativized SUBJECT”.
logical Introduction, Vol. I. Amsterdam: Benjamins Linguistic Inquiry 24, 755⫺772
Green, John (1987), “Spanish”. In: Comrie (ed.), Quirk, Randolph & Greenbaum, Sidney (1973), A
236⫺259 Concise Grammar of Contemporary English. New
Greenberg, Joseph (1963), “Some Universals of York: Harcourt, Brace, Jovanovich
Grammar with Particular Reference to the Order Reinhart, Tanya & Reuland, Eric (1991), “Ana-
of Meaningful Elements”. In: Greenberg, Joseph phors and Logophors: An Argument Structure Per-
(ed.), Universals of Language. Cambridge/MA: spective”. In: Koster, Jan & Reuland, Eric (eds.),
MIT Press, 73⫺113 Long-distance Anaphora. Cambridge: Cambridge
Greenberg, Joseph (1978, ed.), Universals of Hu- University Press, 283⫺321
man Language, Vol. III: Word Structure. Stanford: Schachter, Paul (1985), “Parts-of-speech Systems”.
Stanford University Press In: Shopen, Timothy (ed.), Language Typology and
Greenberg, Joseph (1985), “Some Iconic Relation- Syntactic Description, Vol. I: Clause Structure.
ships among Place, Time, and Discourse Deixis”. Cambridge: Cambridge University Press, 3⫺61
In: Haiman, John (ed.), Iconicity in Syntax. Am- Schachter, Paul & Otanes, Fe (1972), Tagalog Ref-
sterdam: Benjamins, 271⫺287 erence Grammar. Berkeley: University of Cali-
Greenberg, Joseph (1986), “Some Reflections on fornia Press
Pronominal Systems”. In: Weisemann (ed.), xvii- Schwartz, Linda & Dunnigan, Timothy (1986),
xxi “Pronouns and Pronominal Categories in South-
Gundel, Jeanette & Hedberg, Nancy & Zacharski, western Ojibwe”. In: Weisemann (ed.), 285⫺322
Ron (1993), “Cognitive Status and the Form of Re- Silverstein, Michael (1976), “Hierarchy of Features
ferring Expressions in Discourse”. Language 69.2, and Ergativity. In: Dixon, R. M. W. (ed.), Gram-
274⫺307 matical Categories in Australian Languages. Can-
Head, Brian (1978), “Respect Degrees in Pronomi- berra: Australian Institute of Aboriginal Studies,
nal Reference”. In: Greenberg (ed.), 152⫺211 12⫺171
Hendrick, Randall (1988), Anaphora in Celtic and Stump, Gregory (1984), “Agreement vs. Incorpora-
Universal Grammar. Dordrecht: Kluwer tion in Breton”. Natural Language and Linguistic
Hinnebusch, Thomas (1979), “Swahili”. In: Sho- Theory, Vol. II, 289⫺348
pen, Timothy (ed.), Languages and Their Status. Ultan, Russell (1978), “Some General Characteris-
Cambridge/MA: Winthrop, 290⫺293 tics of Interrogative Systems”. In: Greenberg
Ingram, David (1978), “Typology and Universals (ed.), 211⫺248
of Personal Pronouns”. In: Greenberg (ed.), 213⫺ Wiesemann, Ursula (1986, ed.), Pronominal Sys-
248 tems. Tübingen: Narr
King, Gareth (1996), Basic Welsh: A Grammar and
Workbook. London: Routledge Linda Schwartz, Bloomington (U.S.A.)
77. Verb
1. Introduction 1. Introduction
2. The lexical composition of verbs
3. Lexical classes that are morphologically
significant Verbs are semantically defined as describing
4. Morphology and the function of verbs events, actions and, in some languages,
5. The formation of verbs states. Syntactically, verbs constitute the pri-
6. Inflectional morphology mary relational focus of the clause, specifying
7. Implicational universals
8. Grammaticized verbs the roles and interactions of the participant
9. Uncommon abbreviations noun phrases. Morphologically, verbs can be
10. References the repository of clause-level morphology,
Franks, Steven & Schwartz, Linda (1994), “Bind- Mithun, Marianne (1986), “When Zero Isn’t
ing and Non-distinctness: A Reply to Burzio”. There”. Proceedings of the Berkeley Linguistics So-
Journal of Linguistics 30, 227⫺243 ciety 12, 195⫺211
Givón, Talmy (1976), “Topic, Pronoun and Gram- Newman, Paul (1987), “Hausa and the Chadic
matical Agreement”. In: Li, Charles (ed.), Subject Languages”. In: Comrie (ed.), 705⫺723
and Topic. New York: Academic Press, 149⫺188 Progovac, Ljiljana (1993), “Long-distance Reflex-
Givón, Talmy (1984), Syntax: A Functional-typo- ives: Move-to-INFL vs. Relativized SUBJECT”.
logical Introduction, Vol. I. Amsterdam: Benjamins Linguistic Inquiry 24, 755⫺772
Green, John (1987), “Spanish”. In: Comrie (ed.), Quirk, Randolph & Greenbaum, Sidney (1973), A
236⫺259 Concise Grammar of Contemporary English. New
Greenberg, Joseph (1963), “Some Universals of York: Harcourt, Brace, Jovanovich
Grammar with Particular Reference to the Order Reinhart, Tanya & Reuland, Eric (1991), “Ana-
of Meaningful Elements”. In: Greenberg, Joseph phors and Logophors: An Argument Structure Per-
(ed.), Universals of Language. Cambridge/MA: spective”. In: Koster, Jan & Reuland, Eric (eds.),
MIT Press, 73⫺113 Long-distance Anaphora. Cambridge: Cambridge
Greenberg, Joseph (1978, ed.), Universals of Hu- University Press, 283⫺321
man Language, Vol. III: Word Structure. Stanford: Schachter, Paul (1985), “Parts-of-speech Systems”.
Stanford University Press In: Shopen, Timothy (ed.), Language Typology and
Greenberg, Joseph (1985), “Some Iconic Relation- Syntactic Description, Vol. I: Clause Structure.
ships among Place, Time, and Discourse Deixis”. Cambridge: Cambridge University Press, 3⫺61
In: Haiman, John (ed.), Iconicity in Syntax. Am- Schachter, Paul & Otanes, Fe (1972), Tagalog Ref-
sterdam: Benjamins, 271⫺287 erence Grammar. Berkeley: University of Cali-
Greenberg, Joseph (1986), “Some Reflections on fornia Press
Pronominal Systems”. In: Weisemann (ed.), xvii- Schwartz, Linda & Dunnigan, Timothy (1986),
xxi “Pronouns and Pronominal Categories in South-
Gundel, Jeanette & Hedberg, Nancy & Zacharski, western Ojibwe”. In: Weisemann (ed.), 285⫺322
Ron (1993), “Cognitive Status and the Form of Re- Silverstein, Michael (1976), “Hierarchy of Features
ferring Expressions in Discourse”. Language 69.2, and Ergativity. In: Dixon, R. M. W. (ed.), Gram-
274⫺307 matical Categories in Australian Languages. Can-
Head, Brian (1978), “Respect Degrees in Pronomi- berra: Australian Institute of Aboriginal Studies,
nal Reference”. In: Greenberg (ed.), 152⫺211 12⫺171
Hendrick, Randall (1988), Anaphora in Celtic and Stump, Gregory (1984), “Agreement vs. Incorpora-
Universal Grammar. Dordrecht: Kluwer tion in Breton”. Natural Language and Linguistic
Hinnebusch, Thomas (1979), “Swahili”. In: Sho- Theory, Vol. II, 289⫺348
pen, Timothy (ed.), Languages and Their Status. Ultan, Russell (1978), “Some General Characteris-
Cambridge/MA: Winthrop, 290⫺293 tics of Interrogative Systems”. In: Greenberg
Ingram, David (1978), “Typology and Universals (ed.), 211⫺248
of Personal Pronouns”. In: Greenberg (ed.), 213⫺ Wiesemann, Ursula (1986, ed.), Pronominal Sys-
248 tems. Tübingen: Narr
King, Gareth (1996), Basic Welsh: A Grammar and
Workbook. London: Routledge Linda Schwartz, Bloomington (U.S.A.)
77. Verb
1. Introduction 1. Introduction
2. The lexical composition of verbs
3. Lexical classes that are morphologically
significant Verbs are semantically defined as describing
4. Morphology and the function of verbs events, actions and, in some languages,
5. The formation of verbs states. Syntactically, verbs constitute the pri-
6. Inflectional morphology mary relational focus of the clause, specifying
7. Implicational universals
8. Grammaticized verbs the roles and interactions of the participant
9. Uncommon abbreviations noun phrases. Morphologically, verbs can be
10. References the repository of clause-level morphology,
such as negative, interrogative, tense, mood, by some other word class, such as adjective
and evidentiality, as well as agreement and or noun. However, since verbs frequently do
the morphological categories of high rele- designate such non-prototypical percepts, the
vance to the verb such as valence, voice and term state-of-affairs can be used to cover the
aspect. Verbs are more often morphologically meanings expressed by verbs (see Art. 104).
complex than nouns, and in this article we In contrast to concrete objects which are
will examine the general morphological prop- easily isolated perceptually and individuated
erties of verbs and the way in which the lexi- (it is relatively easy to tell a table from a
cal semantics of the verb interacts with mor- chair), the boundaries of events are not so
phological categories. discrete. Thus what is included in the mean-
ing of a verb is not a perceptual given, but
rather subject to different interpretations in
2. The lexical composition of verbs different languages. In a series of experi-
ments, Gentner (1981) found that verbs are
No cross-linguistically valid delineation of less translatable and more mutable in context
the semantic domains of nouns and verbs is than nouns are, and thus more subject to lan-
possible because a concept that in one lan- guage-specific variation in their lexical
guage is expressed as a verb might in another content. Gentner also found that even within
language be expressed as a noun (see Art. 72). one language, verbs are more subject to vari-
Nonetheless a semantically-based distinction able interpretation than nouns are, as can be
between nouns and verbs is possible if we are seen in the following two sentences where the
willing to accept a prototype model of se- verb is the lexical element that is subject to
mantic properties (Hopper & Thompson an extended interpretation, while the nouns
1984). The semantics of nouns and verbs can maintain their ordinary interpretations:
be said to correspond to percepts in the real
world such that the prototypical noun corre- (1) John ran through the town.
sponds to a percept having time-stability and (2) The road ran through the woods.
a prototypical verb corresponds to an “ac-
tion” or “event”, that is, a percept that lacks 2.1. The conceptualization of event
time-stability (Givón 1979). From a func- In the following we examine some of the
tional perspective, a prototypical verb de- ways in which the lexical semantics of verbs
notes a concrete, kinetic, visible and effective can differ cross-linguistically. We begin by
action (Bates & MacWhinney 1982). To the considering the ways in which a sequence of
extent that the meanings of verbs deviate actions can be conceptualized. For instance,
from this prototype, and, for example, desig- English provides a means for describing the
nate more abstract events or actions, such as act of fetching firewood as one event, even
mental activities (think, feel), processes (grow, though it actually consists of a series of ac-
age, develop), internal conditions (itch, ache) tions. In the Papuan language Kalam this
or states (know, be old, be able), they are less same ‘event’ is described as a sequence of five
prototypical and more likely to be expressed discrete actions (Foley 1986: 113):
Most of what we conceive of as events are situations. They are used in combination with
actually sequences of actions, and languages other verbs and with nouns to designate com-
can divide these sequences in different ways plex actions.
to constitute the meanings of verbs. The divi- Given the freedom with which individual
sion of complex events into their component languages can interpret sequences of actions
actions is typical of verbal expression in Ka- as events and the mutability of the lexical
lam and correlates with the fact that the lan- meaning of verbs in context, it is not surpris-
guage has fewer than one hundred verb ing that the number of verbs a language has
stems, with only about twenty-five of these may differ widely cross-linguistically. The
being commonly used. Needless to say, the number of nouns a language needs is directly
commonly used verbs have very general related to the number of concrete objects
meaning and are applicable to many different (and abstract concepts) that are culturally im-
portant, but the number of verbs depends roll, float, bounce; or the shape of the moving
upon how relations between objects are con- object, as in Atsugewi -lup- ‘for a small shiny
ceptualized and lexicalized. We have just spherical object (e.g. a round candy, an
mentioned Kalam, with fewer than one hun- eyeball, a hailstone) to move or be located’,
dred monomorphemic verb stems, but even -qput- ‘for loose, dry dirt to move or be lo-
more extreme examples exist. Among the lan- cated’, -caq- ‘for a slimy lumpish object (a
guages of Australia, especially as one moves toad, a cow dropping) to move or be located’
away from the east coast, one finds Warlpiri (Talmy 1985). While the incorporation of di-
with around 110 simple verb roots, Walmatj- rection and manner are familiar from Euro-
ari with about forty simple verbs, Gurindji pean languages, the incorporation of proper-
with no more than thirty and some languages ties of the moved or located object are found
in the Kimberleys and Daly River area with in Native American languages, such as Atsu-
only about a dozen verbs to which inflections gewi (Hokan) and Navajo (Athapaskan), as
can be added (Dixon 1980: 280). These are of well as in American Sign Language. In these
course used in combination with other ele- languages, verbs that incorporate features of
ments to produce a wide variety of verbal the moved or located object are known as
meaning. classifier verbs.
across languages. Change of state events are The plurality of participants is often related
typically intransitive, since by definition the to other aspects of plurality of action, such
cause is not specified. Thus motion events as distributed or repeated action (see section
and events such as laughing, coughing, aging, 5. 3. 5).
ripening are usually lexicalized as intransi-
2.5. Lexical aspect
tive. Similarly, statives, which describe inher-
ent properties of entities, such as to be old, Another prominent lexical parameter for
to be heavy, to be red, are typically lexi- verbs is aspect or Aktionsart, since the mean-
calized as intransitive verbs, if they are lexi- ing of a verb specifies the temporal contours
calized as verbs at all. Thus verbs of commer- of the event it is describing (Comrie 1976;
cial activity, ingestion, manipulation and Dahl 1985; Art. 109). One distinction that is
creation strongly tend to be causative, while sometimes regarded as aspectual has already
verbs of motion with incorporated manner been mentioned ⫺ the distinction between
tend to be inchoative, and verbs of behavior stative and inchoative. This distinction is lexi-
or color tend to be stative. In many calized in such English pairs as know and re-
languages, a small subset of causative verbs alize. Another type of aspectual distinction
are ditransitive, that is, they obligatorily take occurs between verbs designating only one
three arguments ⫺ an agent, a patient and cycle of an action versus those designating re-
a recipient. Such verbs usually correspond peating cycles. Thus step constitutes one cycle
to meanings such as tell, give, send, and of which walk designates multiple repetitions.
Other examples are hit and beat, snort and
so on.
snore.
Despite these generalizations, languages
can still have some surprising valence values, 2.6. Dynamic vs. stative
as the Guugu Yimidhirr verb for laughing, A stative situation is one that is extended in
which is transitive: diinal ‘laugh at’ (Dixon time and which involves no change, while a
1980). More systematic differences are found dynamic situation brings about some kind of
for events that do not fit the prototypes well; change. Thus know and be tall are stative
they tend to vary in their valence cross-lin- predicates in English, while run and buy are
guistically (Croft 1990). Thus verbs that de- dynamic predicates (Comrie 1976: 48f.). Of
scribe emotional or mental states in some course there is a certain degree of indetermi-
languages take their human participant as nacy for situations such as sleeping or stand-
the subject, as in Eng. I like beans but other ing which might be conceptualized as stative
languages make the human participant an or dynamic, as well as for cases in which a
object argument or experiencer, as in Span. steady state is maintained actively, such as
Me gustan los frijoles. emitting a pure tone or holding one’s breath.
Besides the number and grammatical role The distinguishing factor that has been sug-
of the arguments, the meaning of a verb spec- gested is that stative situations are main-
ifies other characteristics of its arguments. tained until something happens to change
The incorporation of information about the them, while dynamic situations require a con-
physical attributes of an entity for motion stant input of energy to be maintained. Thus
and location verbs has already been il- sleeping and standing would qualify as sta-
lustrated. More common are specifications of tive under this criterion, but emitting a pure
animacy or humanness for certain arguments tone and holding one’s breath would qualify
of a verb. Such conditions on the use of verbs as dynamic.
with specific arguments are known as selec-
tional restrictions. 3. Lexical classes that are
Another property of participants that is morphologically significant
sometimes incorporated into the verb is plu-
rality of the absolutive argument, as in the The two semantic distinctions that are most
following pairs of verbs from !Kung (Sny- important for verbal morphology are the
man 1969: 124f.): transitive/intransitive distinction and the dy-
namic/stative distinction. Languages such as
(4) ⫽’wã ‘pick sg.’ ’kheu ‘pick pl.’ Nahuatl, which have subject and object
guº ‘take sg.’ n/’hwi ‘take pl.’ markers on the verb, have a different mor-
!o’a ‘break sg.’ kx’oma ‘break pl.’ phological structure for the verb if the verb
g//xõ ‘lay sg.’ g⫽a ‘lay pl.’ is transitive rather than intransitive. In some
languages this difference amounts only to the from Abkhaz illustrate the use of person/
addition of object markers to the verb along number prefixes for agreement with up to
with the subject markers, or ergative in addi- three arguments of the verb: with each added
tion to absolutive. The following examples argument a prefix is added to the verb.
(5) intransitive
(sarà) s-ce-yt’
(I) 1.sg.subj-go-fin
‘I go.’
(6) transitive
(sara barà) bzéya be-z-bó-yt’
(I you) well 2.sg.abs-sg.erg-see:dyn-fin
‘I like you.’
(7) ditransitive
s-àš-cwa àžwabž (ø-)-s⫹à-r-Øwe-yt’
my-brother:pl news (3.sg.abs-) 1.sg.io⫹to-3.pl-erg-tell-fin
‘My brothers told me the news.’
In some cases the extent of morphophone- or in the cooccurrence restrictions that they
mic fusion between subject and object mark- manifest. For instance, it is common for a
ers makes the transitive markers appear quite perfective morpheme to have a present state
different from the intransitive ones (e.g. in interpretation when used with a stative predi-
the Maung language of Australia). In Al- cate (e.g. in Island Carib [Arawakan], Tahi-
gonquian languages, the transitivity of the tian [Eastern Oceanic] and Slave [Atha-
verb determines the nature of the final suffix, paskan]). A progressive morpheme may have
and also the interpretation of the person/ a present state interpretation with a stative
number affixes as referring to the subject or predicate (as in Kanukuru [West Chadic]) or
the object. it may simply not be used with statives (as in
In addition, transitive and intransitive English or Cocama [Tupi]).
verbs differ in the extent to which they can be Cooccurrence restrictions based on the lex-
affected by morphological valence-changing ical aspect of the verb are common for deri-
processes, such as causative: in some lan- vational categories and for grammaticizing
guages only intransitive verbs may be causat- constructions that are near the beginning of
ivized (for instance, Nakanai [New Guinea]), the grammaticization chain. This is the case
while in others both transitive and intransi- for progressives, as mentioned above, but it
tive can be affected (as in Arabic). Passiviza-
also applies to iteratives (meaning repetition
tion is a morphological process that applies
on one occasion), which apply only to dy-
typically to transitive verbs, though passive-
namic verbs and at the beginning of their de-
like constructions are found with intransitive
verbs in some languages (e.g. German). velopment only to telic verbs. Similarly, re-
The stative/dynamic distinction is usually sultatives only apply to dynamic verbs, and
relevant only for morphological aspect. In in fact, are most appropriate with process
some languages stative verbs have fewer as- verbs. On the other hand, inchoatives are re-
pectual distinctions than dynamic verbs do. stricted to statives.
For instance, in Abkhaz (Caucasian) stative Lexical classes of verbs that are semanti-
verbs make no aspectual distinctions in the cally arbitrary also affect the morphology.
past, while dynamic verbs distinguish perfec- Some languages, such as Romance lan-
tive and imperfective aspects. (The past for guages, have different sets of allomorphs for
statives uses the same suffix as the imperfec- tense, aspect and mood depending on the ar-
tive for dynamics.) Moreover, in this lan- bitrary lexical classification of verb stems,
guage the present and future tense forma- commonly referred to as conjugation classes
tions are different for stative vs. dynamic (see Art. 65). Such affixal allomorphy arises
verbs. sometimes for phonological reasons, as for
It is more common, however, for the sta- instance when some stems end in consonants
tive/dynamic distinction to be important in and others in vowels, or for semantic reasons,
the interpretation of aspectual morphemes, as when different auxiliaries that give rise to
the inflections are used for different verbs (as regard the book as one to be read, I think
would be the case if Fr. être and avoir as used one should read the book’.
in the passé composé had become affixes). Other nonfinite forms belong to the gene-
ral category of verbal noun, since they bes-
tow on the verb to a greater or lesser extent
4. Morphology the properties associated with nominals al-
and the function of verbs lowing the verbs to be used in non-canonical
functions. The gerund is a non-finite substan-
In their most prototypical clause-level uses, tival verb form whose logical subject is usu-
nouns refer to participants, while verbs show ally coreferential with the subject of the main
the relation among those participants in the verb or more rarely, the object of the main
situation. On the discourse level, nouns pro- verb. For example, Coming to the top of the
ridge, we saw the valley stretch out before us
totypically introduce participants into the
where the logical subject of coming is we.
discourse, while verbs assert the occurrence
The infinitive form of the verb rarely has
of an event of the discourse (Hopper & person/number agreement and sometimes
Thompson 1984: 708). When a verb serves lacks tense or aspect (though it may appear
this asserting, reporting, function, it carries in both active and passive voice). As pre-
its fullest morphological marking ⫺ person, dicted, the infinitive does not appear as the
number, tense, aspect, modality or whatever main verb in the asserted clause of a sen-
categories are marked in the languages and is tence, nor does it present information in the
considered to be finite. On the other hand, main narrative line of the discourse. Rather
when it serves other functions its morpholog- an infinitive appears in a context in which it
ical markings are substantially reduced (Hop- does not bear a syntactic relation to its no-
per & Thompson 1984). Just as the functions tional subject (hence the lack of person/
of lexical verbs in context may vary along a number agreement), as for instance in com-
scale from true assertion to referring to a plement clauses such as I want him to do it
state of affairs as though it were an entity, so (Noonan 1985: 20). It is common for an in-
the formal properties of verbs are modified finitive to be formed with an adposition or
on a scale of finiteness (Givón 1990). case marker indicating allative, benefactive
Nonfinite verbal forms, which typically or dative relations (e.g. in Chuvash [Turkic],
lack some verbal inflection (most often person Udmurt [Finno-Ugric] and Hebrew [Semitic])
and number agreement) and may also have an (Haspelmath 1991). Cross-linguistic and dia-
affixal marking, occur in various functions chronic evidence suggest that infinitives de-
which are not prototypical for verbs. velop from the use of a locative element with
a verb in constructions indicating movement
Participles allow verbs to be used in adjec-
for a purpose (I went to see him) to general
tival functions modifying nouns. Aspectual
purpose (He saved money to buy a car) and
distinctions do occur in participles, as in the then to complement clauses (He wants to buy
present and past participles of English and a car) (Haspelmath 1991; Bybee et al. 1994).
other Indo-European languages (e.g. eating Thus infinitives are widely used in all of these
vs. eaten), and voice distinctions can also oc- contexts, as well as in other subordinate
cur as in the active and passive participles of contexts. While it is uncommon for infinitives
Latin. The gerundive, which is a non-finite to show person/number agreement, a notable
adjectival passive, conveys some modality exception is the personal infinitive of Portu-
sense: for instance Lat. liber legendus ‘a book guese, which has the forms shown in
to be read’ as in librum legendum censeo ‘I Tab. 77.1:
Another term for a type of verbal noun iary are commonly found in constructions
that is used in the Arabic grammatical tra- that give rise to new inflections (see Art. 68),
dition and also for the description of Cau- e.g. in the formation of the anterior or per-
casian languages is maSdar. In the Arabic fective (as in Spanish, French, German or
tradition the maSdar is a form lacking both Dutch), the progressive (as in English and
person/number agreement and tense but Spanish), the future (as in Spanish and
verb-like in that it can govern the nominative French) and the passive (as in English).
and accusative cases. In the description of Uncommon in Europe, but widespread in
Abkhaz, a Caucasian language, the maSdar is the remainder of the world is the use of serial
the only nominalized form of the verb or medial verb constructions in which a sin-
(Hewitt 1979). The notional subject appears gle event is described in terms of a series of
as a possessive marker on the Abkhaz verbs, as in example (3). In such construc-
maSdar. tions the verbs do not have independent ex-
Nominalizing a verb also allows the verb pression of tense, aspect and mood; rather, in
to appear in non-canonical functions. There one type of construction, exemplified in (8),
are two types of nominalizations, one in all the verbs in the construction have the
which the nominalization names the act, ac- same values for these categories, while in the
tivity or result that corresponds to the verb’s other type, exemplified in (9), only the first
action (e.g. destruction, arrival) and the other or last verb has access to the full range of
in which the meaning of the nominalization distinctions and the other verbs have no
reflects the relation between the verb and its markings for the inflection categories, or ex-
arguments. While the latter type includes tremely reduced markings (Hopper &
agentive (baker), instrumental (opener), and Thompson 1984):
patient (payee) nominalizations, by far the (8) Akan
most common of these in the languages of wo-a-didi a-nom
the world is the agentive nominalization, 3.pl-perf-eat perf-drink
which in fact occurs in almost all of the lan- ‘They’ve eaten and drunk.’
guage families of the world (Woodworth
1991). Nominalizing morphemes almost al- (9) Yoruba
ways occur contiguous with the verb stem mo ḿ-mú ı̀wé bòõ
and it is very unusual to find any inflectional 1.sg prog-take book come
distinctions made within the nominalized ‘I am bringing the book.’
verb (Woodworth 1991). In Papuan languages the medial verb takes a
Nominalizations, infinitives and partici- different person/number affix than the final
ples in combination with an inflected auxil- one, as shown in (10) (Davies 1981):
(10) Kobon
Yad be gau am-em kaj pak-nab-in.
I forest there go-1.sg.med pig strike-fut-1.sg.fin
‘I will go to the forest and strike a pig.’
Thus in all of these cases where the verb 5. The formation of verbs
stem is serving a function other than to assert
the occurrence of a separate event of the dis- 5.1. Compounding and incorporation
course, the verb carries fewer inflectional dis- One method languages exploit to create new
tinctions. When the verb is performing its ca- verbs is compounding ⫺ the combination of
nonical function it may be marked for those
two lexical stems into one verb (see Art. 87).
categories of high relevance to the verb, such
as aspect and tense, as well as for categories The two components may consist of any
that have clause-level or discourse-level func- word class plus a verb as head: noun and a
tions. Since the verb is central to the clause, verb (Eng. babysit), two verbs (Igbo túũ-fù
showing the relation among the participants, ‘throw-be lost ⫽ throw away’; Mandarin
it is also the repository of clause-level mor- lā-kāi ‘pull-be open ⫽ pull open’), or an ad-
phology in many languages, serving as the verb or adposition plus verb (as is common
bearer of markings of epistemic and speaker- in Indo-European languages: Lat. adēo
oriented mood, evidentials, honorifics, nega- ‘toward-go ⫽ approach’, Germ. mitfühlen
tion and interrogation. A discussion of these ‘with-feel ⫽ sympathize’). The entire forma-
inflectional categories follows in 6. tion is then inflected according to the (usually
regular) morphological categories of the lan- tion’, with ‘to become X’ and ‘to make X’
guage, and hence qualifies as a single verb. also occurring frequently as in the following
Directional adverbs that compound with a examples (Woodworth 1991: 190, 201):
verb may eventually come to signal aspectual (12) Modern Greek ‘to do X’s action’
meaning (see section 5. 3. 4). arx-iz-ø-o
Related to compounding is the process of beginning-vbz-imp-1.sg
noun incorporation, which involves the mor- ‘I begin’
phological incoporation of a noun into a ver- amerikan-iz-ø-o
bal complex (see Art. 88). Noun incorpora- American-vbz-imp-1.sg
tion differs from compounding in being re- ‘I mimic an American’.
stricted to a (sometimes rather large) subset
of the available nouns and verbs in the lan- (13) Karok ‘to be X’
guage. In particular, the nouns that can be ? a:s-hi
incorporated include body parts, food or water-vbz
game and cultural products. These items are ‘to be wet’
often at a generic rather than specific level of (14) Karok ‘to make X’
categorization; that is, proper nouns are ? uhru-hi
never incorporated and a word at the level of egg-vbz
categorization of tree is likely to be incorpo- ‘to lay eggs’
rated, but not a word that names a species of (15) O’odham (Papago-Pima) ‘to become X’
tree. A typical example of noun incorpora- e-hodaz-tcu’D
tion is shown in (11). refl-stone-vbz
(11) Tiwi (Osborne 1974: 47) ‘turned himself to stone’
ji-meni-alipi-ankina Some languages have verbalizers with
he-me-meat-steel more specific meanings, such as Inuit ‘to
‘He stole my meat.’ have few Xs’, ‘to hunt Xs poorly’ or O’od-
ham (Papago-Pima) ‘to break’ or ‘to take X
In some languages the incorporated version away forcibly’. These more specific verbaliz-
of the noun is a reduced or even suppletive ers are undoubtedly related to noun incorpo-
form of the independent noun. For instance, ration constructions (Woodworth 1991).
the free Tiwi noun corresponding to the in-
corporated form for ‘meat’ in (11) is punin- 5.3. Verbs derived from other verbs
kapa. At times the verb also appears in a re- A variety of processes of affixation or redu-
duced form in the incorporation unit. When plication exist that derive one verb from an-
the direct object is incorporated, the resulting other. These processes involve valence
word is an intransitive verb and does not take changes, aspectual changes or the addition of
an independent object (see Mithun 1984 and locational and directional meaning.
Art. 88).
5.3.1. Valence-changing morphology
5.2. Verbalization processes
The most common type of derivational mor-
Another way to form new verbs is to derive phology for verbs in the languages of the
them from nouns and adjectives (see Art. 89). world is valence-changing morphology (cf.
Verbs derived from adjectives usually mean Art. 107). Although some languages allow
‘to become Q’, as in Span. engrandecer ‘to the same verb stem to be used in intransitive,
enlarge’ from grande ‘big, large’, and are in- transitive or causative constructions without
choatives, which in general specify entry into a change in morphology (e.g. English verbs
a state. When such forms are used with a such as open and boil), most languages do
causative agent, they are termed factitives. have some morphological means of changing
An English verbalization such as blacken can valency. Causative morphology is by far the
be used as an inchoative or a factitive. most frequently-occurring, as in these exam-
Verbs may be derived from nouns by sim- ples from Syrian Arabic: nezel ‘go down’, naz-
ply adding verbal inflection to nouns, as in zal ‘bring down’, nām ‘go to sleep’, nayyam
English hammered, or they may be derived by ‘put to sleep’. Of course, the effect of caus-
adding affixes to the noun stem. These affixes ative on an intransitive verb is to make it
always occur closer to the stem than inflec- transitive, and the effect on a transitive verb
tion does. The most common meanings for is to make it ditransitive. Detransitivization
verbalizers are ‘to be X’ and ‘to do X’s ac- or anticausative formation also occurs, as in
this example from Turkish: acœ -tI ‘open-past’, jects that may occur in the clause. In Chis-
acœ -Il-dI ‘open-anticaus-past’. hona this added argument can represent a
A highly generalized valence-increasing goal, a motive, a prepositional phrase, or
process occurs in many Bantu languages. even an infinitive or clausal adjunct. The ap-
When an affix, called the applicative, is plicative suffix in Chishona is -ir- or -er-
added to verbs it increases the number of ob- (Harford 1993).
5.3.2. Attraction of adpositions to verbs rect objects has become a suffix on the verb.
It is also common for prepositions and post- For instance, in Dholuo, the preposition ni
positions to become affixes on verbs, giving occurs with indirect objects that are full no-
them directional or locational meaning, or minals, but when the indirect object is pro-
changing their valence or aspect. In Southern nominal, the preposition and pronoun are
Lwo (Western Nilotic) languages what was both suffixed (Reh 1986: 123f.):
previously a preposition occurring with indi-
In the related language Lango, the dative/ verb, they may become prefixes on the verb.
benefactive marker and the agreement marker The following examples from Abkhaz show
are suffixed in all cases. The /n/ of the suffix the instrumental postposition after the noun
has assimilated to the final consonant of the (19a) and alternatively incorporated into the
stem (Reh 1986: 126): verb (19b) (Hewitt 1979: 114):
Other examples of adpositions that be- jungle’ which also has completive meaning.
come associated with the verb are the separa- The Karok affix meaning ‘up to the height
ble prefixes of German and Dutch, and the of a man or less’ also means ‘to start to’
verb particles of English. while the affix meaning ‘down from the
height of a man or less’ is used with stative
5.3.3. Locational and directional predicates to indicate the action that results
morphology in that state (Bright 1957: 97, 102f.). Anda-
tive and venitive markings on verbs indicate
Many languages have locational morphemes motion away from and towards to the
that affix to the verb indicating the location speaker, or they indicate ‘go in order to
of the situation described by the verb. Such do’ and ‘come in order to do’. For instance
affixes most commonly use the location of in Karok, the suffix -ar̃ added to a verb
the speaker as the deictic reference point. gives the meaning ‘to go in order to’ (? ih-
The location of the hearer is only used in ar̃ ‘to dance-to go in order to ⫽ to go in
systems which also use the location of the order to dance’ [Bright 1957: 106]). The lat-
speaker. In the highly elaborated system of ter meanings sometimes give rise to future
Nimboran (North Papuan) such affixes indi- markers (as in the southern dialect of Si-
cate that the situation occurred ‘here’, erra Miwok).
‘there’ or ‘far away’, and additionally ‘below’
or ‘above’ (the speaker) (Anceaux 1965: 5.3.4. Aspectual derivation
62⫺64).
Systems of directional morphemes may One common type of aspectual derivation
be even more elaborate. Such systems are makes a verb completive. Completive action
based primarily on allative and ablative has been carried out thoroughly and com-
relations, again using the location of the pletely, often with an object that has been
speaker as the primary reference point. totally affected or consumed, or with the
Nimboran has eleven directional affixes implication that all objects have been af-
that specify the possible combinations of fected. For instance Tucano (Andean-Equa-
movement to and from the five positions torial) has a suffix péò which means comple-
listed above (Anceaux 1965: 70⫺79). Other tive, as in the following words (Sorensen
languages have even more specific meanings 1969: 173f.):
for directional affixes. For instance in (20) ba-péò-amÈ́
Karok (Hokan) a set of more than thirty eat-comp-evi:3.sg
directional affixes have meanings as specific ‘he ate all of it’
as ‘hence upriverward’ and ‘hither from
owhà-péò-apÈ́
upriver’; ‘horizontally away from the center
paddle-comp-evi:1.pl
of a body of water’ etc. (Bright 1957:
‘(we) paddled all the way over’
95⫺105). In Nicobarese directional affixes
indicate movement towards the jungle ver- Directional particles or adpositions can be
sus movement towards the village or the used to give completive meaning, as in the
sea (Braine 1970: 173⫺176). Of course, such English phrases eat up or burn down. If this
specific meanings for directionals reflect the use of directional morphemes is generalized,
geographical environment of the cultures in it may lead to the establishment of a deriva-
question. tional perfective, as found in the Slavic lan-
At times such directional affixes also guages. Thus the perfective forms of verbs in
take on aspectual meaning, as in the case Serbo-Croatian have prefixes that originally
of the Nicobarese affix meaning ‘into the indicated direction:
Another common aspect that may be ex- defined as the repetition of an action on a
pressed derivationally is the iterative aspect, single occasion. Iterative is often expressed
by reduplication and can have as other mean- following examples from Mwera (Eastern
ings habitual, continuative, frequentative and Sudanic) (Harries 1950: 77):
distributive (cf. Art. 109). Consider the
(22) lya ‘eat’ lyalyalya ‘eat and eat and eat’
gwa ‘fall’ gwagwagwa ‘fall and fall and fall’
tawa ‘tie’ tawa-tawa ‘tie over and over again’
pinga ‘want’ pinga-pinga ‘search for’
jenda ‘travel’ jenda-jenda ‘wander about’
markers on verbs may either constitute agree- tion only to clarify the referents of the pro-
ment or pronominal arguments of the verb. nominal affixes (Mithun 1986).
In the former case, the actual noun phrases Person/number markers on verb some-
are the arguments of the verb and the agree- times include distinctions for gender or noun
ment markers simply index these arguments. class. For instance in Krongo, the finite verb
In the latter case, the person/number markers has a prefix indexing the subject of the verb.
are pronouns bound to the verb and repre- In Tab. 77.2, n- is used for first and second
sent all the core arguments of the verb. The person singular and for neuter, m- is used for
verb with its inflections, then, represents a feminine, ø- for third singular and k- for plu-
complete clause, and full noun phrases func- ral (Reh 1985).
in fewer languages than aspect, tense and in a ‘helping’ status ⫺ as the finite verb in a
mood categories do. That it is not a strict or clause that contains a lexical non-finite verb.
causal implication is shown by the fact that Auxiliaries typically code a full range of as-
there are exceptions to this statement. For in- pectual and modal functions, though they
stance in Trukese (Oceanic) object markers can also be used for passive voice and future
are suffixed to the verb, but subject markers tense. They differ from lexical verbs in their
as well as all aspect, tense and mood markers inflections: in some cases they code more
are found in a preverbal auxiliary that is not morphological distinctions than full verbs (as
bound to the verb. Thus the verb has object with the English auxiliary verb be, which dis-
inflection but no aspect, tense or mood in- tinguishes first and second singular in the
flection. Similarly, O’odham (Papago-Pima) present, or the Basque auxiliaries); in other
has object markers prefixed to the verb, but cases they code fewer distinctions, (e.g. the
subject markers and aspect, tense and mood English modal auxiliaries lack third person
are contained in an auxiliary that is not inflection and do not have participles or in-
bound to the verb, but rather appears in sen- finitives) (cf. Steele et al. 1981).
tence-second position. Another type of ex-
8.2. Copulas
ception is exemplified by Dakota (Siouan),
which has an elaborate person-number sys- Copulas are grammatical words that serve
tem incorporated into the verb (prefixed or the predicating function in clauses with pred-
infixed), but most aspect, tense and mood icate nominals (Bob is a rancher). The same
categories are expressed by particles or auxil- form often occurs with predicate adjectives
iaries. (Bob is short) and locatives (Jane is down
One might expect implicational relations by the windmill). Semantically, copulas are
to hold among aspect, tense and mood cate- highly generalized stative predicates, but they
gories. However, no such relations can be are not always totally lacking in meaning, as
substantiated; rather these three categories evidenced by the fact that some languages
are so commonly occurring that almost all (e.g. Spanish, Portuguese and Irish) have two
languages that have one of them have the copulas, one indicating a permanent or stable
other two (see the numerical break-down in state and the other indicating a more tempo-
the preceding section). In fact, there is some rary state. Copulas often carry the morpho-
logical markers of verbs, indicating tense, as-
overlap in these categories and it might be
pect and person/number. However, they can
more useful to consider the meanings ex-
also be defective, lacking certain markers
pressed within these categories in looking for
that appear on other verbs. One source of
universals rather than looking for universals
this defectiveness is the fact that they are sta-
at the level of aspect, tense and mood. For tive verbs, and stative verbs often mark fewer
instance, it appears that if a language has in- aspectual distinctions than active verbs. In
flectional tense or aspect at all, then it has some languages (e.g. Russian) the copula is
either a perfective or a past morpheme (Dahl omitted when it occurs in unmarked verbal
1985; Bybee et al. 1994; Art. 109 and 110). categories, such as present tense, but it ap-
Similarly it might be said that a large major- pears in non-present contexts bearing the
ity of languages with inflection have an in- markers of tense.
flectional means of expressing the imperative. Another reason for their defectiveness in
Beyond these statements, however, it is very verbal categories is that some copulas are not
difficult to find more universal categories of derived from verbs at all, but rather from
aspect, tense or mood. pronouns. Even in these cases, however, per-
son/number or gender agreement with the
8. Grammaticized verbs subject may be present. For instance, in Da-
kota agreement with plural subjects is
Three types of verb-like entities occur that marked with the suffix -pi on the ‘defective
are more grammatical in their properties verb’ e, which is derived from an emphatic
than lexical: auxiliaries, copulas and pro- pronoun. Complete lack of inflection can
verbs. arise if one form of the pronominal copula
(usually the third singular masculine form)
8.1. Auxiliaries generalizes to be used with all subjects, as for
Auxiliaries (see Art. 78) are usually derived instance in Tigre (Semitic) where the singular
from lexical verbs diachronically but have ar- masculine pronominal copula tu can be used
rived at a stage in which they always occur with all subjects in copula constructions.
Meeting of the Berkeley Linguistics Society. Berke- Snyman, J. W. (1969), An Introduction to the !Xu
ley/CA: Soc., 195⫺211 (!Kung) Language. Cape Town: Balkema (Univer-
Noonan, Michael (1985), “Complementation”. In: sity of Cape Town School of African Studies Com-
Shopen, Timothy (ed.), Language Typology and munication 34)
Syntactic Description, Vol. II. Cambridge: Cam- Sorensen, Arthur P., Jr. (1969), The Morphology of
bridge University Press, 42⫺140 Tucano. Ph.D. dissertation, Columbia Univ.
Osborne, C. R. (1974), The Tiwi Language. Can- Steele, Susan. et al. (1981), An Encyclopedia of
berra: Australian Institute of Aboriginal Studies AUX: A Study of Cross-Linguistic Equivalence.
Cambridge/MA: MIT Press (Linguistic Inquiry
Parks, Douglas. R. (1976), A Grammar of Pawnee. Monograph 5)
New York: Garland Publishing
Talmy, Leonard (1985), “Lexicalization Patterns:
Reh, Mechthild (1985), Die Krongo Sprache (Nı̀inò Semantic Structure in Lexical Forms”. In: Shopen,
Mó-D̀i): Beschreibung, Texte, Wortverzeichnis. Ber- Timothy (ed.), Language Typology and Syntactic
lin: Reimer Description, Vol. III. Cambridge: Cambridge Uni-
Reh, Mechthild (1986), “Where Have All the Case versity Press, 57⫺149
Prefixes Gone?”. Afrikanistische Arbeitspapiere 5, Woodworth, Nancy (1991), From Noun to Verb and
121⫺34 Verb to Noun: A Cross-linguistic Study of Class-
Slobin, Dan I. (1997), “Mind, Code and Text”. In: changing Morphology. Ph.D. dissertation, SUNY
Bybee, Joan & Haiman, John & Thompson, San- at Buffalo
dra A. (eds.), Essays on Language Function and
Language Type. Amsterdam: Benjamins, 437⫺467 Joan Bybee, Albuquerque (U.S.A)
78. Auxiliary
1. Auxiliaries as a morphological class another verb, here monere ‘advise’, in its per-
2. Paradigmatic vs. syntactic auxiliaries fect participle form, it supplies a term in the
3. Auxiliaries and verbhood paradigm of the latter which is not expressed
4. References morphologically, the passive perfect; it helps
complete the paradigm (cf. Vincent 1987;
1. Auxiliaries as a morphological class Art. 62). (1b) and (1c) illustrate respectively
the (morphological) active perfect and pas-
Auxiliaries reflect in their morphology cate- sive present. (1d) is often described as a
gories which may also be marked morpho- periphrastic form of the verb monere (cf.
logically on (other) verbs. Their distinctive- Art. 68); and the construction is said to be
ness lies in their paradigmatic and syntag- grammaticalised (cf. Art. 145).
matic relationships with (other) verbs. In The notion of potential paradigm is impor-
much of the linguistic tradition the paradig- tant here, however. For the English verb form
matic relationship has been taken to be the is of (2b):
salient one.
(2) (a) He advises
1.1. A traditional view: helping verbs (b) He is advised
Traditionally, the term auxiliary is applied to is likewise described as an auxiliary, even
a verb which in combination with another though there is no morphological realisation
verb regularly supplies part of the paradigm of passive in present-day English; its expres-
or potential paradigm of the second verb. In sion always requires an auxiliary. Thus, in or-
such terms, the Latin verb form est in (1d): der to confer auxiliarihood, in this traditional
(1) (a) monet ‘he advises’ sense, we need to know the set of inherent
categories in principle available for expres-
(b) monuit ‘he (has) advised’
sion in the morphology of verbs. It is desir-
(c) monetur ‘he is advised’
able that this set should be well-defined. At
(d) monitus est ‘he was / has been advised’
any rate, it will include the categories of
can be described as manifesting an auxiliary: tense, aspect, voice, polarity, mood and mod-
this auxiliary has the distribution and mor- ality; these, in turn, will require definition, of
phology of a verb; and in combination with course, if they are to be identified cross-lin-
Meeting of the Berkeley Linguistics Society. Berke- Snyman, J. W. (1969), An Introduction to the !Xu
ley/CA: Soc., 195⫺211 (!Kung) Language. Cape Town: Balkema (Univer-
Noonan, Michael (1985), “Complementation”. In: sity of Cape Town School of African Studies Com-
Shopen, Timothy (ed.), Language Typology and munication 34)
Syntactic Description, Vol. II. Cambridge: Cam- Sorensen, Arthur P., Jr. (1969), The Morphology of
bridge University Press, 42⫺140 Tucano. Ph.D. dissertation, Columbia Univ.
Osborne, C. R. (1974), The Tiwi Language. Can- Steele, Susan. et al. (1981), An Encyclopedia of
berra: Australian Institute of Aboriginal Studies AUX: A Study of Cross-Linguistic Equivalence.
Cambridge/MA: MIT Press (Linguistic Inquiry
Parks, Douglas. R. (1976), A Grammar of Pawnee. Monograph 5)
New York: Garland Publishing
Talmy, Leonard (1985), “Lexicalization Patterns:
Reh, Mechthild (1985), Die Krongo Sprache (Nı̀inò Semantic Structure in Lexical Forms”. In: Shopen,
Mó-D̀i): Beschreibung, Texte, Wortverzeichnis. Ber- Timothy (ed.), Language Typology and Syntactic
lin: Reimer Description, Vol. III. Cambridge: Cambridge Uni-
Reh, Mechthild (1986), “Where Have All the Case versity Press, 57⫺149
Prefixes Gone?”. Afrikanistische Arbeitspapiere 5, Woodworth, Nancy (1991), From Noun to Verb and
121⫺34 Verb to Noun: A Cross-linguistic Study of Class-
Slobin, Dan I. (1997), “Mind, Code and Text”. In: changing Morphology. Ph.D. dissertation, SUNY
Bybee, Joan & Haiman, John & Thompson, San- at Buffalo
dra A. (eds.), Essays on Language Function and
Language Type. Amsterdam: Benjamins, 437⫺467 Joan Bybee, Albuquerque (U.S.A)
78. Auxiliary
1. Auxiliaries as a morphological class another verb, here monere ‘advise’, in its per-
2. Paradigmatic vs. syntactic auxiliaries fect participle form, it supplies a term in the
3. Auxiliaries and verbhood paradigm of the latter which is not expressed
4. References morphologically, the passive perfect; it helps
complete the paradigm (cf. Vincent 1987;
1. Auxiliaries as a morphological class Art. 62). (1b) and (1c) illustrate respectively
the (morphological) active perfect and pas-
Auxiliaries reflect in their morphology cate- sive present. (1d) is often described as a
gories which may also be marked morpho- periphrastic form of the verb monere (cf.
logically on (other) verbs. Their distinctive- Art. 68); and the construction is said to be
ness lies in their paradigmatic and syntag- grammaticalised (cf. Art. 145).
matic relationships with (other) verbs. In The notion of potential paradigm is impor-
much of the linguistic tradition the paradig- tant here, however. For the English verb form
matic relationship has been taken to be the is of (2b):
salient one.
(2) (a) He advises
1.1. A traditional view: helping verbs (b) He is advised
Traditionally, the term auxiliary is applied to is likewise described as an auxiliary, even
a verb which in combination with another though there is no morphological realisation
verb regularly supplies part of the paradigm of passive in present-day English; its expres-
or potential paradigm of the second verb. In sion always requires an auxiliary. Thus, in or-
such terms, the Latin verb form est in (1d): der to confer auxiliarihood, in this traditional
(1) (a) monet ‘he advises’ sense, we need to know the set of inherent
categories in principle available for expres-
(b) monuit ‘he (has) advised’
sion in the morphology of verbs. It is desir-
(c) monetur ‘he is advised’
able that this set should be well-defined. At
(d) monitus est ‘he was / has been advised’
any rate, it will include the categories of
can be described as manifesting an auxiliary: tense, aspect, voice, polarity, mood and mod-
this auxiliary has the distribution and mor- ality; these, in turn, will require definition, of
phology of a verb; and in combination with course, if they are to be identified cross-lin-
guistically. This is, however, a far from trivial paradigm it contributes to) non-finite (but cf.
question (cf. 2.4, Art. 108⫺111, 113, and the 2.3.2 on Greek etc.); and, as the finite form,
extensive discussions in Lyons 1977; Ted- an auxiliary will also typically show person/
eschi & Zaenen 1981, eds.; Hopper 1982, ed.; number (and sometimes gender) concord
Siewierska 1984; Dahl 1985; Palmer 1986). with a particular argument or arguments,
Moreover, given that it is often difficult to typically the subject (as in (1) and (2)), if this
associate particular auxiliaries uniquely with is found in the language concerned (cf.
one of these categories, and that the cate- Art. 98, 100; see too Art. 77 on finiteness).
gories themselves, as well as marking seman- Where different non-finite forms are avail-
tic oppositions, may also serve to indicate able, modal auxiliaries are frequently accom-
discourse function and/or viewpoint (cf. e.g. panied by the infinitive, whereas auxiliaries
Hargreaves 1986, on Newari), these tradi- expressing aspect and voice (as in (1)) tend to
tional categories may themseves reflect recur- govern participles: perfects and passives se-
rent conjunctions of more primitive univer- lect a perfect/passive participle, imperfects or
sal properties. progressives an active participle ⫺ or a ger-
und. Participles also often agree with a spe-
1.2. Morphological categories cific argument in respect of number/gender/
What seem to be relatively uncontroversial case (see 2.3.2). But such observations, again,
cases of auxiliary expression of relevant cate- are not without exception (particularly out-
gories are illustrated in the sections which side the Indo-European group for whose de-
follow. However, tense, aspect, voice and po- scription such terms were introduced), and
larity are not universally present as verbal in- they again depend on cross-linguistic identifi-
flexions or auxiliaries. Negative polarity, as cation of infinitives, participles and gerunds
well as being expressed in verbal morphology (cf. Art. 77). Notice, for instance, that even
(as in the Agaw languages ⫺ Appleyard in Greek the modern perfect in -i is formed
1984) or by an auxiliary (as in Finnish ⫺ see by exi (‘have’) plus what is historically, at
2.2), is also typically expressed by non-verbal least, the aorist infinitive (cf. e.g. Browning
grammatical items (particles or adverbs). 1969: 84). Non-finiteness is one indicator of
Tense, which is the most likely to be ex- subordination; but an auxiliate may bear no
pressed morphologically, is perhaps, along overt marker of subordination (see 2.4; and,
with polarity, least widely claimed to be rep- on the complexities associated with subordi-
resented as an auxiliary. It may be that this nation, cf. Harman & Thompson 1984).
reflects the empirical situation; but it also, at In some languages, such as Wunambal,
least partly, has to do with the difficulties in only an auxiliary can be finite and show con-
establishing whether a future tense should be cord:
attributed to a particular language (given
that modality may also be involved ⫺ see (3) (a) Yala ngu-wanban
Comrie 1985 a: ch. 2.3), as well as the fact hunt I-aux:pres
that past tense is often cumulated with some (b) Ngu-wanban
aspectual category. (See, however, Comrie I-aux:pres
1985 a: ch. 11 on past tense auxiliaries in
Bamileke-Dschang.) (3a) illustrates auxiliary use of wanban; (3b)
Auxiliaries of mood and modality are re- its use as a ‘lexical’ verb (Blake 1987: ch. 7.2).
ferred to collectively in what follows as In other languages, the auxiliary, though ar-
‘modal auxiliaries’, given that how such a guably syntactically finite, in guaranteeing
distinction is to be drawn remains highly con- the sentential status of the construction, is
troversial (cf. e.g. Lyons 1977: ch.17); and ‘impersonal’, and concord is shown on the
often the same item can be said to be in- auxiliate (only): see e.g. Lehmann (1993) on
volved in the expression of either according Yucatec Maya, and also 2.3.2.
to the traditional differentiation of subjunc- Much thus remains tentative concerning
tive or optative as moods and necessity or the distribution of auxiliaries in language and
possibility as modalities (cf. May he perish! the properties they display. On the one hand,
vs. He may perish; and see 2.3.1); cf. Art. 111. as we have noted, their identification and
It seems that verbal expression of mood/ character depend on the establishment of a
modality is also not universal: see further 3.2. theory of verbal categories, including the par-
Often, as in (1d) and (2b), an auxiliary ren- adigmatic (tense etc.) and syntagmatic (infin-
ders its auxiliate (the form of the verb whose itive etc.) categories. Differentiation between
syntagmatic categories of the verb (gerund, (c) They say Neil might have feared it,
冦
participle) and the marking of deverbal might
nouns and adjectives is also problematical: might have
and he
witness the debate over ‘syntactic’ vs. ‘lexical’ *might have feared
passives (e.g. Wasow 1977). did / might have done
On the other hand, the drawing of a dis- (d) Maggie WAS laughing
tinction between affix and auxiliary depends Maggie HAD laughed
on the cross-linguistic identification of the Maggie COULD laugh
syntactic word (cf. Art. 26). Maggie LAUGHED
Maggie DID laugh
2. Paradigmatic vs. Only auxiliaries take the sentential Nega-
syntactic auxiliaries tive marker (“N”) to their right and have it
cliticise to them (4a); except in some residual
Auxiliaries may also exhibit distinctive syn- cases, only auxiliaries Invert (“I”) as in the
tax and/or morphology. But sets of items questions of (4b); only auxiliaries appear in
sharing particular morphosyntactic proper- ‘Code’ positions (“C”), such as finally in the
ties need not coincide with the class of auxil- truncated second conjuncts in (4c); only aux-
iaries in the traditional sense illustrated in iliaries coincide with sentential, rather than
1.1. It may be, then, that it is necessary to lexical, Emphasis (“E”) or contrast (4d). do
recognise two kinds of auxiliarihood, sharing supplies the expression of the relevant parts
the feature of expressing certain categories, of the otherwise non-auxiliary paradigm in
but distinguished as to other means of identi- such circumstances. In these respects, it pat-
fication, paradigmatic vs. (morpho)syntactic. terns syntactically with the auxiliaries. And,
Paradigmatic auxiliaries are suppletive with in so far as it thereby supplies part of the
respect to the morphology of the auxiliate; potential paradigm, it might be counted as
syntactic auxiliaries are a syntactically dis- fulfilling the traditional requirements for
tinctive subset of verbs. The notion of syntac- auxiliaries too. We return to this in 2.3.1 (and
tic auxiliary is illustrated in 2.1 in relation to see further e.g. Huddleston 1979; 1984:
the much discussed English situation. A simi- ch. 4.4).
larly well-defined set of auxiliaries can be be is also morphologically distinctive, in
established for e.g. Wunambal (Blake 1987: showing more paradigmatic distinctions than
ch. 7.2), Ngäbére (Young & Givón 1990) and other verbs:
Mandarin (Li & Thompson 1981: 173⫺188),
but not universally (see 3.1). (5) (a) am, is, are vs. advise, advises
(b) was, were vs. advised
2.1. The morphosyntax of auxiliaries in
English (c) I / (s)he / it were (subj.), was (indic.)
vs. I / (s)he / it advised
English have and be, which (paradigmati-
cally) are associated respectively with the ex- and have is at least morphophonologically ir-
pression of the perfect and progressive, as regular (have : has : had). Indeed, all of the
well as, in the case of be, of the passive il- auxiliaries show irregular third person singu-
lustrated in (2b), are distinguished syntacti- lar marking (cf. does [dvz], may, etc.); but
cally from other verbs, except do and the then so does the non-auxiliary says ([siz]), so
modals, by what have come to be called the that such distinctiveness is not criterial.
‘NICE’ properties, as exemplified in (4): have and be share other aspects of their
(4) (a) Maggie wasn’t leaving distribution with other verbs, as non-finites
Maggie hadn’t left (6a) and as finites (6b) in non-NICE posi-
Maggie couldn’t leave tions:
*Maggie leftn’t/left not (6) (a) to advise / be / have, advising / being /
Maggie didn’t leave having, advised / been / had
(b) Was Maggie leaving? (b) He advises / is advising / has advised
Had Maggie left?
Could Maggie leave? They remain verbal in these respects; and, as
*Left Maggie? auxiliaries in the sense of 1.1, they belong to
Did Maggie leave? the same paradigm as the simple form.
The modals, whose auxiliary behaviour is kamba 1979 on Lingala; Palmer 1986: ch.
illustrated in (4) (and at least some of whose 1.5.1).
members express appropriate categories), ad- There may be found within a single lan-
ditionally lack the third-person singular in- guage a distinction between (paradigmatic)
flexion found with all (other) verbs, their in- auxiliaries which do and do not show excep-
herited tense variation is no longer transpar- tional syntactic behaviour. The English verb
ently such in most instances, and they occur get, for example, is arguably, along with be,
only as finites: a periphrastic marker of passive voice, but it
is not distributionally like an auxiliary; it be-
(7) *Dan is canning cope,
haves, as illustrated in (11):
*Dan has could cope,
*Dan should can cope, (11) Did Kenneth get caught out?
*Dan’s canning cope is a miracle, *Got Kenneth caught out?
*For Dan to can cope would be a miracle like a non-auxiliary. As noted above, this
The English modals thus share with the non- might lead us to draw a distinction between
auxiliaries, as part of their basic distribution, those auxiliaries which are simply paradig-
only the position illustrated in (6b). This matically defined (as in Finnish) and those
raises the question, more than in the case of which also meet syntactic criteria. It also
the other auxiliaries, of whether they are in- raises the question of whether there might be
deed verbs. See further 3. items which meet the syntactic criteria associ-
ated with auxiliaries in a particular language
2.2. The non-coincidence of paradigmatic but are not appropriate on paradigmatic
and syntactic auxiliarihood grounds: see further 2.4.
Auxiliaries (paradigmatically defined) need A further apparent discrepancy arises
not display marked distributional differences from the observation that in languages with
from other verbs such as are illustrated in (4) a paradigmatic ‘be’ auxiliary the same verb
(whose precise character is, of course, Eng- is typically used as a copula with predicative
lish-specific); though they tend, as commonly nominals/adjectivals. And, in English, for in-
used verbs, to be morphologically irregular. stance, the copula displays the NICE proper-
Indeed, in many languages it is difficult to ties associated with paradigmatic be. How-
motivate syntactically a class of auxiliaries. ever, the copula can perhaps be regarded as
Thus, the Finnish verb which typically would an auxiliary for non-verbal predicators, al-
be translated as ‘be’ and which is used as an lowing to non-verbs occurrence in finite and
adjectival copula in (8): non-finite constructions. ‘Have’ auxiliaries,
too, typically can take nominal complements
(8) Kivi on kova (He has a car). And in English there is varia-
stone is hard tion as to whether in this use have displays
‘The stone is hard’ the NICE properties (Has he a car? vs. Does
is also used to form compound tenses: he have a car?). Perhaps this is a reflection of
its status as a ‘marked’ copula, with locative
(9) Olette erehtynyt subject (see further 2.3.2, and cf. Lyons
you:are err:past part 1968: ch. 8.4).
‘You have made a mistake’ Concurrent or antecedent use as a copula
Finnish also has a negative verb, as il- is one cross-linguistic characteristic of tense/
lustrated in (10): aspect/voice auxiliaries. Another tendency is
for auxiliaries, as grammatical elements, to
(10) Ette ole erehtynyt cliticise (cf. Art. 41), particularly to a preced-
You:not be err:past part ing element, as in English; indeed, in some
‘You have not made a mistake’ traditions, cliticisation is criterial for auxiliar-
But these (paradigmatic) auxiliaries are not ihood.
distributionally distinct from other Finnish In so far as they are otherwise syntactically
verbs, some of which can appear in the same distinct, auxiliaries can be said to be syntacti-
constructions, governing infinitives and par- cally ‘more active’. One way of making such
ticiples. And there seem to be particular diffi- a notion more precise is illustrated by the
culties in establishing a set of syntactic mod- following definition (Emonds 1985: 174):
als for most languages (cf. e.g. Huber & “A verb is an auxiliary if and only if it occurs in
Kummer 1974 on German; Mufwene & Bo- some syntactic position where main verbs do not
occur, or if it cannot appear in some position Paradigms may also show syncretism (cf.
where main verbs do occur (deep structure subca- Art. 66), whereby in a certain inflexional
tegorization for phrasal complements permitting).” class two otherwise distinct terms of a partic-
though a set so defined would in some re- ular category have the same expression.
spects be more inclusive than the traditional Thus, the form in (13b) also expresses accusa-
(either paradigmatic or syntactic) auxiliaries; tive as well as nominative. Non-finite and
note too that the definition is accompanied past forms of have in English show syncre-
by an assumption that the English modals tism. Finite non-past have contrasts with the
are not verbs. Alternatively, a more tradi- simple past as perfect vs. past, and as such
tional delimitation of syntactic auxiliaries is usually rejects definite past-time reference:
arrived at if they are said to be preferred to (14) (a) Bilbo went there (last Tuesday)
non-auxiliaries in finite use (Anderson 1990).
(b) Bilbo has gone there *(last Tuesday)
2.3. Paradigmatic properties of auxiliaries But (15) shows syncretism of perfect and past:
As periphrases, traditional auxiliaries exhibit (15) (a) Bilbo seems to have gone there (last
the paradigmatic (and possibly derivational) Tuesday)
properties of terms of morphological cate-
gories, otherwise expressed in the form of a (b) Bilbo had gone there (last Tuesday)
word. with both interpretations being available, the
latter permitting definite past-time reference.
2.3.1. Basic properties
Similarly, cumulation (cf. Art. 64), simulta-
Auxiliaries are traditionally associated with neous expression of two or more categories,
paradigmaticity. As members of a paradigm, as evidenced, again, in (13), or in the French
auxiliaries are typically in opposition with (past) imperfect (simultaneously past and im-
other members of the paradigm; the members perfect) of (16a):
are semantically mutually exclusive. Even
when the meaning distinction between two (16) (a) Il nageait ‘he was swimming/(habitu-
members is contextually attenuated, as by the ally) swam’
addition in brackets in (12): (b) He used to swim
(12) (a) Beryl knits (a lot these days) is also found in auxiliary constructions, such
(b) Beryl is knitting (a lot these days) as English (16b): used (though perhaps only
a ‘marked’ periphrasis) is simultaneously past
they remain distinct; the simple form cannot and habitual. We can, of course, here as else-
be used to mean the same as the progressive. where, associate the signalling of past with
In this way a distinction might be drawn be- the final coronal plosive; but in this instance
tween ‘true’ auxiliary periphrases like the there is no contrast with non-past (*He uses
progressive of I am writing and ‘marked’ pe- to swim); used is a paradigmatic unit.
riphrases like Italian Sto scrivendo: the latter Conversely, inflexional paradigms may
merely insists on an interpretation available show multiple exponence (Art. 64), whereby a
to the simple form (cf. Comrie 1976: ch. 6.1). single category is expounded at more than
However, some auxiliary constructions are one place in structure, as in the Old English
merely suppletive variants (cf. Art. 52) of form of (17):
particular members of the paradigm: this is
the status of periphrastic do in present-day (17) luf-o-d-on
English, which occurs in certain circum- love-weakii.past-weak.past-past.pl.
stances (especially the NICE contexts) in indic
place of the simple form. Suppletion is one ‘we / you / they loved’
characteristic of inflexional systems, not mer- -o- signals that the form is associated with a
ely of stems (as in the familiar go / went ex- verb belonging to a particular conjugational
ample, or Greek vlépo / ı́da ‘see (present / class (weak, sub-class II) and that it is past
aorist)’, but also of inflexions; this is what in (-o- being absent from non-past parts of the
part characterises difference of declensional paradigm ⫺ cf. infinitive lufian); -d- that it is
or conjugational class, as illustrated by the weak and past (strong verbs showing past by
Latin forms in (13): ablaut); -on that it is plural and past (no non-
(13) (a) fili-i son-nom.pl.masc pasts or non-plurals end thus), and indica-
tive. Past is expounded (cumulatively) in
(b) host-es enemy-nom.pl.masc three different places in structure; weak in
This is another reflexion of paradigmaticity. two. Multiple exponence can also be attrib-
uted to auxiliary constructions such as (12b): structions (cf. Meillet 1924; Kuryłowicz 1931;
progressive is expounded, in this case collec- Anderson 1973; Comrie 1976: ch. 5.2; Har-
tively, by both is and the -ing suffix. greaves 1986; Green 1987; Pinkster 1987).
Further, as with the members of inflex- ‘Have’-perfects are typically associated with
ional categories, the occurrence of auxiliaries transitive verbs, in involving a bleaching of
may be determined rectionally. Thus the sub- the construction type illustrated by I have it
junctive in the English subordinate clause in built, with passive complement. In English,
(18a) (indicated by the absence of the third perfect have has been generalised to intransi-
person singular -s) is triggered by the pres- tives (including passives, so that (2b), unlike
ence of a governing verb of a particular class: (1d), is not perfect). But in a number of other
Indo-European languages (including earlier
(18) (a) They insisted that she leave
English), only some intransitives take a
(b) They insisted that she should leave ‘have’-perfect, while others require ‘be’, as il-
As an alternative to (18a), however, there is lustrated by (20):
an auxiliary construction (18b) (sometimes (20) (a) Francesca ha telefonato ‘Frances has
interpreted as ‘more tentative’). In other cir- phoned’
cumstances the English modals have replaced
the subjunctive, with the archaic (19a) giving (b) Francesca è arrivata ‘Frances has
way to (19b): arrived’
(19) (a) If she had had wheels, she had been a This discrepancy in Italian, and other phe-
bicycle nomena (such as agreement), has been attrib-
uted to the status of the subject of the (20b)
(b) If she had had wheels, she would have type as an underlying object, analogous to
been a bicycle the subject of passives, which display similar
The ‘past subjunctive’ (except partially in the properties (cf. Burzio 1986: 1.7, who rather
case of be ⫺ recall (5c) ⫺ now formally iden- misleadingly refers to the (20b) construction
tical with the past indicative) in (19b) is in a as “ergative”, one more generally known as
suppletive relation with a modal auxiliary. unaccusative). In Basque, in both perfect and
imperfect constructions, transitives take a
2.3.2. Bleaching and transparency ‘have’ auxiliary and intransitives a ‘be’ (see
The reflections of paradigmaticity illustrated example (32) in 3.1; and cf. Saltarelli 1988:
in 2.3.1 can be taken as evidence for the 1.2.1).
traditional idea of auxiliary constructions as ‘Have’ perfects and the ‘have’ copula are
‘forms of’ the simple verb, as periphrases. As not widespread, and, in many cases, at least,
members of a paradigm, so grammaticalised, they are preceded historically by construc-
auxiliaries typically show semantic depletion tions with ‘be’ ⫹ locative/dative or other
or bleaching; semantically they express sim- oblique marker (Vendryes 1937; van Gin-
ply a (admittedly possibly complex) term of a neken 1939; Benveniste 1952; 1960), con-
category that might otherwise be expounded structions which, as in (1d), combine perfect
inflexionally. Historical depletion in the evo- and passive. ‘Have’ constructions provide a
lution of perfect auxiliaries based on ‘have / periphrastic active parallel to the passive per-
own / grasp / put’ verbs in various Indo- fect of (1d) etc., though, initially at least, con-
European families and elsewhere is well-doc- cord on the participle continues to reflect the
umented, as it is in the development of imper- predicative status of the participle vis-à-vis
fects / progressives based on verbs such as the object, as in the Latin of (21) (from
‘exist / remain / stand’ or (other) locative con- Pinkster 1987: 210):
namese, the verbs bi (‘suffer something un- The development of clitics and morpho-
pleasant’) and duoc (‘undergo something not logical forms from earlier verbal auxiliary
unpleasant’) form alternative passives (cf. Le constructions seems to be well attested (cf.
1976, who also discusses similar construc- e.g. (on Latin and Romance) Pinkster 1987;
tions in other languages). Vincent 1987; (on Muskogean) Haas 1977;
The main sources of future/predictive aux- Munro 1984), and this ‘syntactic depletion’
iliaries are in distal goal-oriented directionals can indeed be seen as confirming the para-
(‘go to’) and volitionals and obligationals digmaticity of these and other auxiliary con-
(Bybee & Pagliuca 1987); ‘recent past’ and structions.
perfect auxiliaries often derive from a distal Depletion might also be said to be re-
source-oriented directional (‘come from’). flected in the typical transparency of auxilia-
Perfects also evolve from completive (‘use ries to argument structure. The subject of an
up’ / ‘finish’) verbs (cf. e.g. Munro 1984, on auxiliary, for instance, typically does not
Muskogean). have its semantic function determined by the
Epistemic senses for modals and the mark- auxiliary but by the non-auxiliary with which
ing of mood (as in (18b) and (19)) seem to that auxiliary is associated; auxiliaries typi-
develop later than the less-depleted other cally do not impose selectional restrictions on
senses; and the latter show a range of ‘natu- their subjects or other potential arguments in
ral’ sources: ‘know’ / ‘have power’ / ‘be fitted the overall construction. As already observed
to’ ⇒ ‘be able’ (⇒ ‘be allowed’), ‘owe’ ⇒ (1.2), in some languages (some) auxiliaries
‘have to’, etc. (Standop 1957; Givón 1982; are indeed overtly impersonal, as in construc-
Shepherd 1982; Traugott 1989). tions involving Greek prepi:
(22) Prépi na ton do
must:3.sg prt 3:acc.sg.m see:aor.subj.1.sg
‘I must see him’
which is invariable for person/number (and (b) Menà yap⬘a doõm-k‘wa-k‘
governs a finite (subjunctive) clause ⫺ cf. e.g. bear man kill:irr-3.human.obj-in-
Mackridge 1985: 2.4.1). 2.4, however, looks ferential
at some consequences and limitations of ta- ‘It seems that the bear killed the man’
king transparency as criterial for auxiliari-
hood. (Chung & Timberlake 1985: 3.3; cf. too An-
derson 1985: 2.2.1), where the irrealis stem
2.4. Paradigmatic, derivational and and the final inferential suffix mark the prop-
syntactic auxiliaries osition expressed as not necessarily being
If we associate transparency with auxiliari- vouched for by the speaker. (Comrie 1976:
hood, this would include within the class of 5.2.2.1) discusses inferential auxiliary con-
auxiliaries all raising-to-subject verbs, such as structions in Bulgarian, Georgian and Esto-
seem, appear, happen etc. in English. This nian.)
would mean that an even larger number of Further, among this more inclusive set of
(paradigmatic) auxiliaries in English would auxiliaries would be a subset, particularly of
fail to meet the syntactic criteria discussed verbs of ‘objective aspect’ or ‘Aktionsarten’
above (in 2.1) as NICE phenomena; there (Deutschbein 1939), such as the English ones
would be a marked discrepancy between the in (24a), which express notions also often ex-
sets of syntactic and paradigmatic auxiliaries, pounded morphologically ⫺ cf. (24b) (from
the latter set being much more inclusive. And Comrie 1985b: 343f.):
the range of ‘auxiliary notions’ would like- (24) (a) It started / continued / ceased to fall
wise be more extensive.
But perhaps this is appropriate, on (b) Russian: za-pet’ (‘start-singing’)
grounds of paradigmaticity. A verb like Eng- Gombe-Fulani: ? o-dø aani-dø aaninto
lish seem could be seen as the periphrastic (‘He keeps on falling asleep’)
equivalent of the (arguably ‘modal’) inferen- Swahili: lev-uk-a (‘cease to be drunk’)
tial verb forms of a language like Takelma, However, the morphological relations ex-
as in (23b): pressed here are usually regarded as deriva-
(23) (a) Menà yap⬘a t⬘omõ-k‘wa tional rather than inflexional (cf. Art. 38). If
bear man kill-3.human.obj the finite verbs in (24a) are to be regarded as
‘The bear killed the man’ auxiliaries, they are derivational rather than
(30a) illustrates the use of the auxiliary with indeed go. [It] may equally indicate that the
intransitive verbs; (30b) the transitive auxil- referent of the subject is obliged to go, that
iary. The auxiliary is the locus of tense varia- he is permitted to go, or it is possible that
tion and person and number marking, while he will go” (Comrie 1991: 31). The intended
the non-auxiliary occurs in a non-finite form, interpretation may be made more specific
a form which elsewhere is adjectival or nomi- pragmatically, or by paraphrase, but distinc-
nal. The Basque auxiliaries, as well as form- tive modal periphrases are not available. It
ing constructions with verbs (as in (30)), also may be that there are languages which lack
take non-verbal complements, as in (31): auxiliaries altogether: Dyirbal (Dixon 1972)
is perhaps such a language.
(31) (a) Hirian da
town-in (s)he-is 3.2. Non-verbal auxiliaries
‘(S)he is in town’
The question of the universality of auxiliari-
(b) Liburua du hood, however, is intimately bound up with
book-the (s)he-has-it the status of the notion auxiliary, and in par-
‘(S)he has the book’
ticular its association with verbhood. A
(cf. e.g. Saltarelli 1988: 1.2.1.1), in which sit- somewhat weaker position, whereby it is
uation the Basque forms would be translated merely claimed that in any language some of
as ‘be’ and ‘have’, respectively, with no dis- the auxiliary notions (assuming again this set
tinct main verb. The Basque auxiliary verbs to be well-defined) will be expressed gram-
may be appropriately described as such: they matically (paradigmatically), by items that
are distributionally distinctive (with respect are at least historically verbs or by verbal in-
to finiteness) and mark distinctions in tense flexion or by non-verbal closed class items,
or aspect; and they share with non-auxiliary has a greater chance of universal validity.
verbs the possibility of occurring as non-fi- Particularly in the case of modality, whose
nite, as shown in (32) for the ‘be’ (auxiliary) notional connexion with verbhood is (per-
verb: haps along with tense ⫺ but see Comrie
1985 a: 1.4) arguably less close than in the
(32) Alegera izaiten da case of aspect and voice, non-verbal expres-
lively be:gerund (s)he-is sion of ‘auxiliary’ notions (by ‘particles’) is
‘(S)he is usually lively’
apparently common (cf. e.g. Palmer 1986:
This distribution as non-finite is common to 1.5.3). Such a claim is made rather more vul-
all Basque verbs. What is striking is that the nerable and interesting by the additional stip-
auxiliaries are the only finite verbs (with the ulation that expression is associated with a
few exceptions noted before), despite being universal syntactic constituent, AUX (cf.
non-prototypical verbs. Steele 1981, ed.; Ramat 1987), despite such
This striking distributional distinctiveness apparent diversity of realisation.
is in contrast with the situation in Finnish One consequence of this hypothesis
(exemplified in 2.2), where auxiliary periph- (though this can and has been proposed inde-
rases are not distributionally distinct from pendently of it) is that even in relation to a
verb ⫹ complement structures. But a lan- language like English, wherein the auxiliaries
guage may lack not merely syntactic distinc- are clearly diachronically verbs, it has been
tiveness for its auxiliaries; it may fail to show argued that these items (or at least the first in
a set of auxiliaries altogether. Comrie (1991), any sequence ⫺ the one displaying the NICE
for instance, argues that Haruai (Highland properties) are not verbs. Similarly, it has
Papua New Guinea) lacks a set of items cor- been proposed (Manandise 1987) that the
responding to the modals of English etc.; i.e. Basque auxiliaries illustrated in (30) ⫺ which
a set of items grammaticalised (in the proto- appear to lack a verbal root ⫺ are not verbs;
typical cases) to express modalities. The use and, for example, it has been claimed that in
of the verb form in (33): Hausa “a synchronic analysis of auxiliaries
as verbs seems questionable” (Schachter
(33) nagö dy-n-o⁄ -a
you go-fut-2.sg-decl 1985: 41). The category status of auxiliaries is
particularly opaque in the absence of overt
“does not necessarily commit the speaker to markers of (non-)finiteness, as with the Eng-
the view that the referent of the subject will lish modals.
Consider also e.g. the Greek ‘future’ Controversy concerning the AUX hypoth-
marker ua, which is accompanied by a finite esis involves in part the nature of constitu-
subjunctive verb: ency (cf. e.g. Pullum 1981 on Akmajian &
(34) ha su g rápso Steele & Wasow 1979, replied to in Steele
‘I shall write to you’ 1981 ⫺ e.g. note 8 to ch.2). It is not clear, for
instance, that arguments for AUX in English
Historically, this form is apparently a con- (e.g. Oehrle 1981) establish a syntactic con-
traction of héli na (héli representing an imper- stituent any more viable than the AUX of
sonal use of heli ‘want’) ⫺ cf. (22); but in (35) rather than a class of items individually
Modern Greek the form is invariant and satisfying the properties attributed to AUX
bears no morphological mark of finiteness. ⫺ and constituting paradigmatic auxiliaries.
It has nevertheless been maintained, Similarly, the semantic argument for AUX in
though controversially, that the auxiliaries of Luiseño might be regarded rather as evidence
Basque and the modals of English show a of paradigmaticity. More generally, there has
verbal distribution, even though it is shared been criticism both of the semantic character-
only partially with (other) verbs; the differ- isation of AUX (Thrane 1983: ch. 3) and of
ences can be related to finiteness. However, it equivocation in the use of criteria for its es-
does not seem that a set of syntactic auxiliary tablishment in individual languages (4), as
verbs can be established for every language; well as suggestions (cf. Reuland 1983) that
and even periphrastic auxiliaries may be lack- AUX is simply the finite-clause manifestation
ing in particular languages (see 3.1). But the of the INFL of Chomsky (1981), which
AUX hypothesis poses the question of uni- therein dominates tense and agreement. The
versality in a different form: the putatively character of INFL and its relation to auxili-
universal AUX may be realised inflexionally arihood have been in turn much debated (cf.
or as an element or elements with no verbal e.g. Pollock 1989; Iatridou 1990).
‘connexion’.
AUX is to be identified cross-linguistically
in terms of the following definition: 4. References
“Given a set of language-internal analyses, those Akmajian, Adrian & Steele, Susan & Wasow,
constituents which may contain only a specified Thomas (1979), “The Category AUX in Universal
(i.e. fixed and small) set of elements, crucially con- Grammar”. Linguistic Inquiry 10, 1⫺64
taining elements marking tense and/or modality,
will be identified as non-distinct.” (Steele 1981: 21) Anderson, John M. (1973), An Essay Concerning
Aspect. The Hague: Mouton
Thus AUX is a constituent realised by gram- Anderson, John M. (1990), “On the Status of Aux-
matical (rather than lexical) elements some of iliaries in Notional Grammar”. Journal of Linguis-
which express modality and/or tense, a con- tics 26, 241⫺262
stituent which will, further (Steele 1981: ch. 2), Anderson, Stephen R. (1985), “Inflectional Mor-
display non-definitional properties, some ab- phology”. In: Shopen (ed.), 150⫺201
solute, some probabilistic. So, the sequence
Appleyard, D. L. (1984), “The Morphology of the
of second-position clitics in the Luiseño (35): Negative Verb in Agaw”. Transactions of the Philo-
(35) Noo xu n po hunwuti pati logical Society 1984, 202⫺219
I prt prt prt bear:obj shoot Benveniste, Émile (1952), “La construction passive
‘I should shoot the bear’ du parfait transitif”. Bulletin de la société de lin-
guistique de Paris 48, 52⫺62
is argued to be a constituent realised by
grammatical elements some of which (xu and Benveniste, Émile (1960), “ ‘Être’ et ‘avoir’ dans
leurs fonctions linguistiques”. Bulletin de la société
po ⫺ while n marks subject concord) express
de linguistique de Paris 55, 113⫺134
modality and tense, and which also display
the non-definitional properties of e.g. occur- Blake, Barry J. (1987), Australian Aboriginal Gram-
ring in a fixed internal sequence not predicta- mar. London etc.: Croom Helm
ble from other regularities of the language Brinton, Laurel (1988), The Development of English
and showing attachment (cliticisation) to the Aspectual Systems. Cambridge etc.: Cambridge
preceding element. Part of the argument for Univ. Press
the sequence in (35) constituting a constitu- Browning, Robert (1969), Medieval and Modern
ent is the interaction of the two marginal par- Greek. London etc.: Hutchinson
ticles in determining the global modality of Burzio, Luigi (1986), Italian Syntax: A Govern-
the sentence. ment-Binding Approach. Dordrecht etc.: Reidel
Bybee, Joan L. & Pagliuca, William (1987), “The Harris, Martin & Ramat, Paolo (1987, eds.), His-
Evolution of Future Meaning”. In: Ramat, Anna torical Development of Auxiliaries. Berlin etc: Mou-
Giacalone & Carruba, Onofrio & Bernini, Giu- ton de Gruyter
liano (eds.), Papers from the 7th International Con- Hopper, Paul (1982, ed.), Tense-Aspect: Between
ference on Historical Linguistics. Amsterdam: Ben- Semantics and Pragmatics. Amsterdam: Benjamins
jamins, 109⫺122
Huber, Walter & Kummer, Werner (1974), Trans-
Caldwell, R. (1875), A Comparative Grammar of formationelle Syntax des Deutschen, Vol. I.
the Dravidian or South Indian Family of Languages. München: Fink
London: Trübner
Huddleston, Rodney D. (1979), “Criteria for Aux-
Chomsky, Noam (1981), Lectures on Government iliaries and Modals”. In: Greenbaum, Sidney &
and Binding: The Pisa Lectures. Dordrecht, Cinna- Leech, Geoffrey & Svartvik, Jan (eds.), Studies in
mon/NJ: Foris English Linguistics. For Randolph Quirk. London,
Chung, Sandra & Timberlake, Alan (1985), “Tense, New York: Longman, 65⫺78
Aspect, and Mood”. In: Shopen (ed.), 202⫺258 Huddleston, Rodney D. (1984), Introduction to the
Comrie, Bernard (1976), Aspect. Cambridge etc.: Grammar of English. Cambridge etc.: Cambridge
Cambridge Univ. Press Univ. Press
Comrie, Bernard (1985 a), Tense. Cambridge etc.: Iatridou, Sabine (1990), “About Agr(P)”. Linguistic
Cambridge Univ. Press Inquiry 21, 551⫺577
Comrie, Bernard (1985 b), “Causative Verb Forma- Joseph, Brian D. & Zwicky, Arnold M. (1990,
tion and Other Verb-deriving Morphology”. In: eds.), When Verbs Collide: Papers from the 1990
Shopen (ed.), 309⫺348 Ohio State Mini-Conference on Serial Verbs. Co-
lumbus/OH: Department of Linguistics, The Ohio
Comrie, Bernard (1991), “Modality without Mod- State Univ.
als”. In: Kakietek, Piotr (ed.), Problems in the Mod-
ality of Natural Language. Opole: The Pedagogical Kuryłowicz, Jerzy (1931), “Les temps composés du
University of Opole, 29⫺33 roman”. Prace Filologiczne 15, 448⫺453
Dahl, Östen (1985), Tense and Aspect Systems. Ox- Le, Tam Duy (1976), “Vietnamese Passives”. Pa-
ford, New York: Blackwell pers from the Annual Regional Meeting of the Chi-
cago Linguistic Society 12, 438⫺449
Deutschbein, Max (1939), Aspekte und Aktionsar-
ten im Neuenglischen. Leipzig Lehmann, Christian (1993), “The Genesis of Auxil-
iaries in Yucatec Maya”. Proceedings of the Interna-
Dixon, R. M. W. (1972). The Dyirbal Language of tional Congress of Linguists 15.2 (1992), 313⫺316
North Queensland. Cambridge: Cambridge Univ.
Press Li, Charles N. & Thompson, Sandra A. (1981),
Mandarin Chinese: A Functional Reference Gram-
Emonds, Joseph E. (1985), A Unified Theory of mar. Los Angeles, London: California Univ. Press
Syntactic Categories. Dordrecht, Cinnamon/NJ:
Foris Lyons, John (1968), Introduction to Theoretical Lin-
guistics. Cambridge etc.: Cambridge Univ. Press
Ginneken, Jacobus van (1939), “Avoir et être (du
Lyons, John (1977), Semantics, Vol. II. Cambridge
point de vue de la linguistique générale)”. In: Mé-
etc.: Cambridge Univ. Press
langes de linguistique offerts à Charles Bally. Ge-
nève: Georg, 83⫺92 Mackridge, Peter (1985), The Modern Greek Lan-
guage. Cambridge etc.: Cambridge Univ. Press
Givón, Talmy (1982), “Tense/Aspect/Modality: The
Creole Prototype and Beyond”. In: Hopper (ed.), Manandise, Esméralda (1987), “AUX in Basque”.
115⫺163 In: Harris & Ramat (eds.), 317⫺344
Green, John N. (1987), “The Evolution of Ro- Matthews, Peter H. (1981), Syntax. Cambridge
mance Auxiliaries: Criteria and Chronology”. In: etc.: Cambridge Univ. Press
Harris & Ramat (eds.), 257⫺267 Meillet, Antoine (1924), “Le développement du
Haas, Mary R. (1977), “From Auxiliary to Inflec- verbe ‘avoir’”. In: Antidoron: Festschrift J. Wacker-
tional Suffix”. In: Li, Charles N. (ed.), Mechanisms nagel. Göttingen: Vandenhoeck & Ruprecht, 9⫺13
of Syntactic Change. Austin: Univ. of Texas Press, Mufwene, Salikoko S. & Bokamba, Eyamba G.
525⫺537 (1979), “Are there Modal-auxiliaries in Lingala?”
Hargreaves, David (1986), “Independent Verbs and Papers from the Annual Regional Meeting of the
Auxiliary Functions in Newari”. Berkeley Linguis- Chicago Linguistic Society 15, 244⫺255
tics Society: Proceedings of the Annual Meeting Munro, Pamela (1984), “Auxiliaries and Auxiliari-
12, 401⫺412 zation in Western Muskogean”. In: Fisiak, Jacek
Harman, John & Thompson, Sandra A. (1984), (ed.), Historical Syntax. Berlin: Mouton de
“ ‘Subordination’ in Universal Grammar”. Berke- Gruyter, 333⫺362
ley Linguistics Society: Proceedings of the Annual Oehrle, Richard (1981), “AUX in English”. In:
Meeting 10, 510⫺23 Steele (ed.), 226⫺259
Palmer, Frank R. (1986), Mood and Modality. Standop, E. (1957), Syntax und Semantik der mo-
Cambridge etc.: Cambridge Univ. Press dalen Hilfsverben im Altenglischen ‘magan’, ‘motan’,
Pinkster, Harm (1987), “The Strategy and Chro- ‘sculan’, ‘willan’. Bochum-Langendreer: Boppingh-
nology of the Development of Future and Perfect aus
Tense Auxiliaries in Latin”. In: Harris & Ramat, Steele, Susan (1981, ed.), An Encyclopedia of AUX:
193⫺223 A Study of Cross-linguistic Equivalence. Cam-
Pollock, J.-Y. (1989), “Verb Movement, Universal bridge/MA, London: M. I. T. Press
Grammar, and the Structure of IP”. Linguistic In- Strong, David R. (1980), “Relational Grammar
quiry 20, 365⫺424 and the Diachrony of Periphrastic Constructions
Pullum, Geoffrey K. (1981), “Evidence against the in Latin and Romance”. Papers from the Annual
‘AUX’ Node in Luiseño and English”. Linguistic Regional Meeting of the Chicago Linguistic Society
Inquiry 12, 453⫺463 16, 298⫺310
Ramat, Paolo (1987), Introductory Paper to Har- Tedeschi, Philip J. & Zaenen, Annie (1981, eds.),
ris & Ramat (eds.), 3⫺19 Tense and Aspect. New York etc.: Academic Press
(Syntax and Semantics 14)
Reuland, Erik J. (1983), “Government and the
Search for AUXes: A Case Study in Cross-linguis- Thrane, Torben (1983), “On the Universality of
tic Category Identification”. In: Heny, Frank & AUX”. Acta Linguistica Hafniensia 18, 154⫺200
Richards, Barry (eds.), Linguistic Categories: Aux- Trask, R. Larry (1977), “On the Origins of Ergati-
iliaries and Related Puzzles, Vol. II: The Scope, Or- vity”. In: Plank, Frans (ed.), Ergativity. London,
der, and Distribution of English Auxiliary Verbs. New York: Academic Press, 385⫺404
Dordrecht etc.: Reidel, 99⫺168
Traugott, Elizabeth Closs (1989), “On the Rise of
Saltarelli, Mario (1988), Basque. London etc.: Epistemic Meaning in English: An Example of
Croom Helm Subjectification in Semantic Change”. Language
Schachter, Paul (1974), “A Non-transformational 65, 31⫺55
Account of Serial Verbs”. Studies in African Lin- Vendryes, J. (1937), “Sur l’emploi de l’auxiliaire
guistics, Supplement 5, 253⫺270 ‘avoir’ pour marquer le passé”. Mélanges de lin-
Schachter, Paul (1985), “Parts-of-speech Systems”. guistique et de philologie offerts à J. van Ginneken.
In: Shopen, Timothy (ed.), Language Typology and Paris: Klincksieck, 85⫺92
Syntactic Description, Vol. I: Clause Structure. Vincent, Nigel (1987), “The Interaction of Periph-
Cambridge etc.: Cambridge Univ. Press, 2⫺61 rasis and Inflection: Some Romance Examples”.
Seuren, Pieter (1990), “Serial Verb Constructions”. In: Harris & Ramat (eds.), 237⫺256
In: Joseph & Zwicky (eds.), 14⫺33 Wasow, Thomas (1977), “Transformations and the
Shepherd, Susan (1982), “From Deontic to Episte- Lexicon”. In: Culicover, Peter W. & Wasow,
mic: An Analysis of Modals in the History of Eng- Thomas & Akmajian, Adrian (eds.), Formal Syn-
lish, Creoles, and Language Acquisition”. In: tax. New York etc.: Academic Press, 327⫺360
Ahlqvist, Anders (ed.), Papers from the Fifth In- Young, Philip D. & Givón, Talmy (1990), “The
ternational Conference on Historical Linguistics. Puzzle of Ngäbére Auxiliaries: Grammatical Re-
Amsterdam: Benjamins, 316⫺323 construction in Chibchan and Misumalpan. In:
Shopen, Timothy (1985, ed.) Language Typology Croft, William & Denning, Keith & Kemmer, Su-
and Syntactic Description, Vol. III: Grammatical zanne (eds.), Studies in Typology and Diachrony:
Categories and the Lexicon. Cambridge etc.: Cam- For Joseph H. Greenberg. Amsterdam: Benjamins,
bridge Univ. Press 209⫺243
Siewierska, Anna (1984), The Passive: A Compara-
tive Linguistic Analysis. London etc.: Croom Helm John M. Anderson, Edinburgh (Great Britain)
1. Introduction 1. Introduction
2. Adpositions
3. Conjunctions 1.1. Minor word classes
4. Particles
5. Adverbs Although inflection is most typical of the
6. Uncommon abbreviations major word classes noun, verb, adjective and
7. References closely related classes such as pronouns,
Palmer, Frank R. (1986), Mood and Modality. Standop, E. (1957), Syntax und Semantik der mo-
Cambridge etc.: Cambridge Univ. Press dalen Hilfsverben im Altenglischen ‘magan’, ‘motan’,
Pinkster, Harm (1987), “The Strategy and Chro- ‘sculan’, ‘willan’. Bochum-Langendreer: Boppingh-
nology of the Development of Future and Perfect aus
Tense Auxiliaries in Latin”. In: Harris & Ramat, Steele, Susan (1981, ed.), An Encyclopedia of AUX:
193⫺223 A Study of Cross-linguistic Equivalence. Cam-
Pollock, J.-Y. (1989), “Verb Movement, Universal bridge/MA, London: M. I. T. Press
Grammar, and the Structure of IP”. Linguistic In- Strong, David R. (1980), “Relational Grammar
quiry 20, 365⫺424 and the Diachrony of Periphrastic Constructions
Pullum, Geoffrey K. (1981), “Evidence against the in Latin and Romance”. Papers from the Annual
‘AUX’ Node in Luiseño and English”. Linguistic Regional Meeting of the Chicago Linguistic Society
Inquiry 12, 453⫺463 16, 298⫺310
Ramat, Paolo (1987), Introductory Paper to Har- Tedeschi, Philip J. & Zaenen, Annie (1981, eds.),
ris & Ramat (eds.), 3⫺19 Tense and Aspect. New York etc.: Academic Press
(Syntax and Semantics 14)
Reuland, Erik J. (1983), “Government and the
Search for AUXes: A Case Study in Cross-linguis- Thrane, Torben (1983), “On the Universality of
tic Category Identification”. In: Heny, Frank & AUX”. Acta Linguistica Hafniensia 18, 154⫺200
Richards, Barry (eds.), Linguistic Categories: Aux- Trask, R. Larry (1977), “On the Origins of Ergati-
iliaries and Related Puzzles, Vol. II: The Scope, Or- vity”. In: Plank, Frans (ed.), Ergativity. London,
der, and Distribution of English Auxiliary Verbs. New York: Academic Press, 385⫺404
Dordrecht etc.: Reidel, 99⫺168
Traugott, Elizabeth Closs (1989), “On the Rise of
Saltarelli, Mario (1988), Basque. London etc.: Epistemic Meaning in English: An Example of
Croom Helm Subjectification in Semantic Change”. Language
Schachter, Paul (1974), “A Non-transformational 65, 31⫺55
Account of Serial Verbs”. Studies in African Lin- Vendryes, J. (1937), “Sur l’emploi de l’auxiliaire
guistics, Supplement 5, 253⫺270 ‘avoir’ pour marquer le passé”. Mélanges de lin-
Schachter, Paul (1985), “Parts-of-speech Systems”. guistique et de philologie offerts à J. van Ginneken.
In: Shopen, Timothy (ed.), Language Typology and Paris: Klincksieck, 85⫺92
Syntactic Description, Vol. I: Clause Structure. Vincent, Nigel (1987), “The Interaction of Periph-
Cambridge etc.: Cambridge Univ. Press, 2⫺61 rasis and Inflection: Some Romance Examples”.
Seuren, Pieter (1990), “Serial Verb Constructions”. In: Harris & Ramat (eds.), 237⫺256
In: Joseph & Zwicky (eds.), 14⫺33 Wasow, Thomas (1977), “Transformations and the
Shepherd, Susan (1982), “From Deontic to Episte- Lexicon”. In: Culicover, Peter W. & Wasow,
mic: An Analysis of Modals in the History of Eng- Thomas & Akmajian, Adrian (eds.), Formal Syn-
lish, Creoles, and Language Acquisition”. In: tax. New York etc.: Academic Press, 327⫺360
Ahlqvist, Anders (ed.), Papers from the Fifth In- Young, Philip D. & Givón, Talmy (1990), “The
ternational Conference on Historical Linguistics. Puzzle of Ngäbére Auxiliaries: Grammatical Re-
Amsterdam: Benjamins, 316⫺323 construction in Chibchan and Misumalpan. In:
Shopen, Timothy (1985, ed.) Language Typology Croft, William & Denning, Keith & Kemmer, Su-
and Syntactic Description, Vol. III: Grammatical zanne (eds.), Studies in Typology and Diachrony:
Categories and the Lexicon. Cambridge etc.: Cam- For Joseph H. Greenberg. Amsterdam: Benjamins,
bridge Univ. Press 209⫺243
Siewierska, Anna (1984), The Passive: A Compara-
tive Linguistic Analysis. London etc.: Croom Helm John M. Anderson, Edinburgh (Great Britain)
1. Introduction 1. Introduction
2. Adpositions
3. Conjunctions 1.1. Minor word classes
4. Particles
5. Adverbs Although inflection is most typical of the
6. Uncommon abbreviations major word classes noun, verb, adjective and
7. References closely related classes such as pronouns,
auxiliaries and numerals (cf. Art. 73⫺78), it the present tense and the progressive aspect.
is occasionally observed in function words. If ka in (1) is considered the root of the auxil-
On the basis of the different functions they iary, one might argue that forms such as kula-
express, the following classes can be distin- ka-a are inflected particles. Such an analysis
guished: is indeed plausible for the so-called “negative
⫺ adpositions (pre- and postpositions which auxiliaries” in Arabic (cf. 4), but elements
express grammatical relations, cf. 2), which carry tense/mood/aspect distinctions
⫺ conjunctions (sub- and co-ordinators together with subject and object agreement
which express the relation between clauses markers are better regarded as auxiliary
and between phrases, cf. 3), verbs (cf. Art. 78).
⫺ particles (elements which express gram- From a morphological point of view, mi-
matical distinctions such as aspect, nega- nor word classes have the following proper-
tion, and illocution, cf. 4), and ties:
⫺ adverbs (words which further specify (a) Unlike verbs, which can have inflec-
events for temporal, locational and other tional paradigms to mark aspectual, tempo-
distinctions, cf. 5). ral, modal, and person distinctions, or nouns,
which can have case inflection, minor word
In a system of morphologically defined classes do not have their own inflectional
parts of speech, such inflected function words paradigms. When inflectional forms in minor
constitute minor word classes. word classes are found they always belong to
There is not always a clear-cut distinction the paradigms of major word classes such as
between one of the minor word classes and verbs, nouns and adjectives. In Erza Mordvin
other word classes. There may be an overlap the postposition can even be inflected with
between content words and function words. both verbal and nominal affixes (cf. (31) be-
For instance in Arawak loko can be used as low). However, most elements of the minor
a noun meaning ‘cavity’ and it can be used as word classes do not allow any inflection at
a postposition to express inner location ‘in’. all.
Hungarian kivéve ‘except’ is an adverbial (b) Typically, inflectional categories are
participle which is developing into a preposi- fully productive within a given class. How-
tion. In both cases it can take an object. ever, they are not fully productive within mi-
When the form is used as a participle it can nor word classes. The main reason for that is
have an object in the accusative, as in kivéve that minor word classes are rather hetero-
János-t ‘excepting John-acc’. However, the geneous. For instance, the class of particles
object does not have a case marker when the may encompass elements which are neither
form is used as a preposition, as in kivéve semantically, syntactically nor formally re-
János ‘except John’. lated. Nevertheless it is quite common to
There may also be an overlap between classify elements referred to as verbal par-
classes of function words. In Dutch there are ticles, question particles, and negative par-
elements which can be used as prepositions ticles under one heading. On the other hand
on the one hand and as particles on the other. the class of adpositions does not seem to
Consider for instance op in op de tafel ‘on the have this kind of diversity. However, lan-
table’ and op-bellen ‘to call up’. Different and guages may have adpositions which originate
more complicated is the distinction between from verbs and adpositions which originate
particles and a subclass of auxiliaries. There from nouns. In these cases the deverbal ad-
are auxiliary elements which do not have a positions may show different inflectional
verbal stem and which consist of one or more morphology than the denominal adpositions.
grammatical morphemes. The auxiliary in Another reason for the fact that inflectional
Warlpiri, for instance, can take the following categories are not fully productive within mi-
sequence of markers: relative ⫺ tense/mood/ nor word classes is that these word classes
aspect ⫺ subject agreement ⫺ object agree- do not have their own inflectional paradigms
ment, or negative ⫺ tense/mood/aspect ⫺ (cf. (a)).
subject agreement as in (Hale 1973: 312): There are instances of derivation within
minor word classes. However, these are rare.
(1) kula-ka-a wawiri pura-mi
One example constitutes diminutive forma-
neg-pres-1.sg kangaroo cook-nonpast tion within the class of prepositions in Ara-
‘I am not cooking the kangaroo.’ bic. Prepositions in Arabic can be combined
The element ka in the “auxiliary” together with the diminutive infix in the same way as
with the element mi on the finite verb mark nouns do (Fischer 1987: 134). For that reason
the prepositions listed in the second column Less clear is the following example from
can be considered forms derived from the Ngiyambaa. In Ngiyambaa there is a system-
prepositions presented in the first column: atic semantic difference between adverbs
(2) ba’da ‘after’ bu’yada ‘little after’ which have the suffix -bu and adverbs which
qabla ‘before’ qubayla ‘little before’ do not have this suffix (Donaldson
fawqa ‘over’ fuwayqa ‘little over’ 1980: 76f.). Consider:
(3) biyal ‘sometimes’ biyalbu ‘always’
maradhal ‘far back (in time)’ maradhalbu ‘as far back as possible’
dhalan ‘close (in time)’ dhalanbu ‘as close as possible’
dhuga:y ‘really, indeed’ dhuga:ybu ‘absolutely’
yu:dhi ‘excessively’ yu:dhibu ‘absolutely excessively’
The difference between these forms can be ity. It is also possible to repeat the ideo-
taken to be the result of some derivational phones three or four times. For instance:
process. However, the suffix -bu is also found (4) è-lòs-i’ ciiim ciiim ciiim ciiim
on elements which belong to other word 3.s-go-asp ideop ideop ideop ideop
classes, such as nominals, determiners, and ‘He walks with pride.’
particles. The unified meaning of the suffix
-bu seems to be that of universal quantifica- Non-durative verbs do not allow redupli-
tion. In that case, the application of -bu may cation of ideophones, because these verbs
rather be analyzed as the result of a gram- cannot express ongoing activity; they are mo-
mentary verbs. Note that (4) shows redupli-
matical operation, hence -bu can be consid-
cation of a minor word.
ered an inflectional morpheme.
1.2.2. Relational category
1.2. Inflection of minor word classes Relational categories reflect the position the
1.2.1. Inherent category word occupies in larger structures, for in-
The distinctive aspect of inherent categories stance subject or object, or agent or patient.
is that they are not imposed by the structural Prototypically, minor word classes do not oc-
position occupied by the word and they do cur in this type of position. There are, how-
ever, instances in which an element can be
not depend on the properties of other words
involved in more than one relation. In those
in the structure, for instance the difference
cases it is possible to find the expression of
between singular and plural. It is not likely to the different relations realized in one word.
find many examples of this type of category Consider for instance Arabic in which the
within minor word classes. However, the prepositions mitl- ‘like’ and ġayr- ‘unlike’,
application of the universal quantifier in which express the relation between entities,
Ngiyambaa (see (3) above) may be an exam- can be marked for different cases to express
ple of an inherent category. A second ex- the relation between the verb and (one of) the
ample may be reduplication of ideophones entities. The preposition mitl- ‘like’ takes the
in Turkana (cf. Art. 138 and Dimmendaal accusative case in (5a) and the genitive in
1982). Each verb in Turkana has its specific (5b), whereas the preposition ġayr- ‘unlike’ is
ideophones. Ideophones of durative verbs marked with the accusative case in (6a) and
can be reduplicated to express ongoing activ- the nominative case in (6b):
(5) (a) qul la-hū mitßl-a dß ālika
say:imp.sm to-3sm like-acc that
‘Tell him the same.’
(b) ‘atā bi-nāqat-i-n mitßl-i nāqatı̄
come:perf.3.sg.m with-she-camel-gen-indef like-gen she-camel:1.sg
‘He brought a she-camel like mine.’
(6) (a) dß ahaba n-nās-u ġayr-a-nā
go:perf.3sm the-people-nom except-acc-1p
‘The people left except for us.’
(b) man ’ilāh-u-n ġayr-u llāh-i
who god-nom-indef except-nom God-gen
‘Who can be a god except for Allah?’
Different from the Arabic examples is a se- ticle in Arabic: laysa (3sg.m.), laysat (3sg.f.),
cond example which comes from Arawak lasta (2sg.m.), lasti (2sg.f.), lastu (1sg.), etc. See
(Baarle et al. 1989). The location of events 4 for further discussion of these data.
can be expressed by the inner locative post-
position loko. Since this postposition is asso-
2. Adpositions
ciated with stative states of affairs, the suffix
-di is added to the postposition to indicate 2.1. Nominal inflection
that the state of affairs designated by the verb 2.1.1. Introduction
is dynamic and not stative. The suffix -di trig-
gers the aspiration of /k/ in lokhodi. Inflected adpositions ⫺ pre- and postposi-
tions ⫺ are found in many languages of the
(7) lo-kojwa-bo-ka li-sikwa lokho-di world, as for instance in Abkhaz (Caucasian)
3.m-go.back-asp-tns 3m-house in-dyn (10a), Hixharyana (Amerind; Ge-Pano-
‘He is going up and down in his house.’ Carib) (10b), Hungarian (Uralic-Yukaghir)
Note that in the examples of both languages, (10c), Tarafit (Afro-Asiatic; Berber) (10d),
it is not the adposition which is the element that Breton (Indo-Hittite; Celtic) (10e).
is involved in one of the relations. The adposi- (10) (a) a-q’⫹nè ‘3.sg.nonhum-at (at it)’
tions seem to be the appropriate place to mark (b) k-hoko ‘pl-incl.concerning
the expression of the second relation. (concerning us)’
1.2.3. Agreement (c) alatt-od ‘under-2.sg (under you)’
(d) gar-s ‘to-3.sg (to him)’
The third inflectional category arises by agree- (e) evid-om ‘for-1.pl ‘(for us)’
ment with properties of some other word or
phrase. The category of agreement seems to Adpositions which can be inflected for per-
be the most applicable category within the son distinctions are also called “prepositional
minor word classes. However, if one con- pronouns” in Irish (Bammesberger 1983: 56),
siders referential affixes a category distinct “prepositions allowing personal endings” in
from the category of agreement, genuine ex- Breton (Press 1986: 116⫺119), and “conju-
amples are rare. One good example presents gated prepositions” in Welsh (Williams 1980:
Arawak in which there is gender agreement 127⫺129).
between the conjunction and the second Adpositions that are inflected for person
member in a coordination. distinctions are referential. In most cases the
referential element can be identified as a pre-
(8) (a) li wadili matho hijaro fix or suffix attached to an adposition in a
art.masc man and:fem woman similar way as referential affixes or person
‘the man and the woman’ marking affixes are found on verbs and
(b) to hijaro mathi wadili nouns as for instance in Hungarian:
art.fem woman and:masc man
‘the woman and the man’ (11) (a) Sétál-ok. (walk-1.sg ‘I walk.’)
(b) könyv-ed (book-2.sg ‘your book’)
A second example comes from Erza Mord- (c) alatt-a (under-3.sg ‘under him / her /
vin in which postpositions can be used as it’)
predicates (Erdődi 1968). First and second
person distinctions of the subject are marked Historically, these types of affixes arise by
on the postpositions themselves. Third per- the spelling out of subject, direct object, indi-
son subjects are expressed by personal pro- rect object and possessor. The motivation to
nouns (cf. 3.2). In constructions with third consider referential affixes a distinct category
person subjects the postpositional predicate and not a type of agreement is the following.
agrees in number with the subject. Consider: Along the lines of Givón (1976) and De
Groot & Limburg (1986) it can be argued
(9) (a) son udalo that unstressed pronouns cliticize onto verbs,
he after-sg then become affixes which are however still
‘He is after.’ clearly pronominal in nature, and in the last
(b) si‹ ń udalo-t stage wear down to agreement affixes or zero.
they after-pl This diachronic process can schematically be
‘They are after.’ presented as follows:
A third example may be person and number (12) free pronoun J clitic J referential affix
agreement of the subject on the negative par- J agreement marker J zero
2.1.3. The rise of inflected adpositions must have become an adposition before
Adpositions may arise historically from bound pronominal forms could be added. In
nouns and verbs (cf. Mallinson & Blake (19) for instance, the noun hely in Hungarian
1981: 385f.). A straightforward explanation is followed by an old locative case-ending
for the occurrence of referential affixes on and a pronominal affix.
adpositions and the strong similarities be- (19) hely-ett-em
tween these affixes with other person mark- place-loc-1.sg
ing paradigms may then be the following: ad- ‘instead of me’
positional affixes arise from a reinterpreta-
tion of verb ⫹ (object) pronominal affix and If a construction of noun ⫹ pronominal
noun ⫹ (possessor) pronominal affix. It ap- affix had been the basis of this construction,
pears that denominal adpositions are better one would have expected the order noun ⫺
candidates for having pronominal affixes person affix ⫺ locative.
than deverbal adpositions. The reason for Postpositions may develop into case af-
that may be that many verbal forms which fixes. In some of the languages which show
develop into adpositions are verbal stems, this development, case affixes may still serve
infinitives, and participles, which are not in- as hosts for person marking affixes. This is
flected for person distinctions. Nouns which for instance the case in Erza Mordvin and in
develop into adpositions are typically in the Hungarian. An example of an inflected post-
position to be inflected for person distinc- position in Hungarian is (19). The dative case
tions, because they are nouns which occur in, suffix as in Peti-nek ‘Pete-dat (to Pete)’ oc-
mostly inalienable, possessive constructions. curs in inflected forms as in nek-em ‘dat-1.sg
The development runs allong the following (to me)’ or nek-ed ‘dat-2.sg (to you)’. A se-
lines, where (18b) is the reinterpretation of cond example is the comitative/instrumental
(18a): case as in Petivel ‘Pete-com (with Pete)’ vel-
ünk ‘com-1.pl (with me)’ or vel-ük ‘com-3.pl
(18) (a) NP NP-possessor (with them)’. Finnish, another Finno-Ugric
house inside-3.sg language, shows a distribution which is dif-
‘the house its inside’ ferent from Hungarian (20a). Finnish only al-
house roof-3.sg lows appositonal constructions with postpo-
‘the house its roof’ sitions, but does not allow inflected case af-
fixes. Instead pronouns are marked with case
house back-3.sg affixes (20b)
‘the house its back’
(20) Postposition Case
(b) NP adposition-person/number (a) (én⫽)vel-em (én⫽)ben-em
house inside-3.sg I⫽with-1.sg I=inessive-1.sg
‘in the house’ ‘with me’ ‘to me’
house roof-3.sg (b) (minu) kanssa-ni minu-ssa
‘on the house’ I with-1.sg I-inessive
house back-3.sg ‘with me’ ‘inside me’
‘back of the house’ The different degrees of morphological
A relation of the type between (18a) and boundedness of case markers is also reflected
(18b) is visible in Turkish, where a number of in the order of affixes. Hungarian takes the
postpositions are based on forms which can order person marker ⫺ case marker (üveg-
be used as nouns which are used with their em-ből bottle-1.sg-elative ‘from the bottle’),
own lexical meaning. For instance alt in masa- whereas Finnish takes the order of case
nın alt-ı ‘table-gen underside-3.sg (the un- marker ⫺ person marker (pullo-sta-ni bottle-
derside of the table / under the table)’ (Lewis elative-1.sg ‘from the bottle’).
1967: 80). It is not the case that all adposi-
tions which can be inflected in a language 2.1.4. The fall of inflected adpositions
arise in the way sketched in (18). When a per- The diachronic process sketched in (12) to-
son marking paradigm has been established gether with the hierarchy (17) predicts that in
in a language, the paradigm may be applied languages with appositional constructions
to other new adpositions which originate the referential affixes on adpositions are first
from other constructions than possessive in line to wear down to agreement markers
constructions. There are cases where a noun and zero. This seems to be correct. The loss
of person marking affixes goes together with be similar to that of constructions with an
a reinterpretation of the appositional con- noun phrase as in (22):
structions as non-appositional constructions. (22) (Pronoun) Adpos⫹PM J Pronoun
This goes as follows. In the development of Adpos⫹AM J Pronoun Adpos
a language the appositional construction, in
which the pronominal affix is the head and Constructions without emphasis on person
the noun phrase is the optional extension of may develop into constructions in which the
the head, is reinterpreted as a construction in pronoun is not used any longer. There re-
which the noun phrase (no longer optional) mains the adposition with pe