This chapter shows the distribution of suppletion according to verbal number. The term verbal number requires some clarification: number as a verbal category can reflect the number of times an action is done or the number of participants in the action. When we say that a verb or verb phrase in an Indo-European language, say French, is in the plural, as in ils étaient fatigués ('they were tired'), what we mean is that the verb phrase étaient fatigués agrees with the subject ils in number. The term plural verbs (or pluractional verbs) in many of the languages to be discussed below, on the other hand, refers to any or all of the following senses: the action is performed several times (iterative), or at several places (distributive), or the action affects or involves several participants. For example, Mupun (West Chadic; Nigeria) uses derivational means such as infixation in (1a), suffixation in (1b) or completely different verbs in (1c) to express plural action or plural participants.
(1) Mupun (Frajzyngier 1993: 55-58)
a.
pūt
'go out.sg.action'
pú<á>t
'go.out<pl.action>'
b.
tù
'kill.sg.action'
tù-é
'kill-pl.action'
c.
cīt
'beat.sg.action'
nás
'beat.pl.action'
The derived verbs pú<a>t ‘go.out<pl>’ and tu-e ‘kill.pl’ can be used to indicate that several people went out and that someone killed many things, respectively. However, the use of the plural verbs in (1) is not a matter of syntactic agreement. For one thing, the plural marker mo on the noun that would trigger agreement is itself optional, as illustrated by (2b).
(2) Mupun (Frajzyngier 1993: 60)
a.
n-tu
joos
b.
n-tue
joos
(mo)
1.sg-kill.sg
rat
1.sg-kill.pl
rat
(pl)
‘I killed a rat.’
‘I killed rats.’
Secondly, while the singular form of a verb is generally ruled out with a plural object (cf. 3a), a plural verb may be used with a singular object to indicate an action performed with some intensity, or many times, as in (3b).
(3) Mupun (Frajzyngier 1993: 59)
a.
*wu
cit
mo
b.
wu
nás
war
*3.sg.m
hit.sg
3.pl
3.sg
hit.pl
3.sg.f
*'He hit them.'
‘He hit her many times.’
Thus, while it is a separate grammatical category, verbal number, in many instances, comes very close to other categories and at times becomes indistinguishable from them. In particular, in a number of cases, verbal number is akin to aspect; there are also cases where verbal number is not easy to distinguish from nominal number as reflected in agreement (Corbett 2000: 256).
There are a few other characteristics of verbal number as a grammatical category which need to be pointed out. As indicated above, the labels singular and plural here refer to quantification of the verbal action rather than the nominal arguments of the verb. However, in some languages such as Ainu, the quantity denoted by these lables differs in yet another way from what is usually meant by ‘singular’ and ‘plural’. In the agreement category of number, singular refers to ‘one’ and plural to ‘more than one’. In Ainu, however, if quantity is specified by a numeral and is small, such as ‘one’, ‘two’, ‘three’ and sometimes ‘four’, the singular verbal-number form of the verb is used. For larger numbers, normally, the plural form is used. This applies both to regularly derived and to formally unrelated verbal number pairs, and is illustrated by the pair ‘come’ in (4).
(4) Ainu (Tamura 1988: 40)
a.
tu
okkaypo
ek
two
youth
come.sg
‘Two youths came.’
b.
tupesaniw
ka
arki
ruwe
ne
eight
even
come.pl
nmlz
cop
‘Eight people came.’
Apart from Ainu, there are five other languages on Map 80A which behave in a similar way in that they group the dual and small quantities together with the singular and not with the plural. They are all from North America: Eastern Pomo (Hokan; California), Navajo, Hupa (Athabaskan; California), Wichita (Caddoan; Oklahoma and Kansas) and Kiowa (Kiowa-Tanoan; Kansas and Oklahoma).
Finally, the expression of verbal number in languages where this category is observed may be both lexical and morphological. The main focus of Map 80A is on lexical expressions, i.e. verb pairs such as the Ainu pairs ek vs. arki ‘come.sg' vs. 'come.pl’, or the Mupun pairs cit vs. nás ‘hit.sg’ vs. 'hit.pl’. However, inasmuch as such pairs can be seen as derivationally related, the morphological expression of verbal number is taken into account as well. It should be noted that the selection of singular vs. plural verb stem generally follows an ergative-absolutive pattern: for intransitive verbs, the stem reflects the number of the only argument, whereas with transitive verbs the stem reflects the number of the direct object.
(5) Ainu (Tamura 1988: 38)
a.
‘stand’
b.
‘kill’
as
'stand.sg'
rayke
'kill.sg.obj'
roski
'stand.pl'
ronnu
'kill.pl.obj'
It is a controversial issue whether verb pairs such as the Mupun pair cit vs. nás ‘hit.sg' vs. 'hit.pl’ are to be considered cases of suppletion according to verbal number, or simply as different words. There exists a discrepancy between grammars on the one hand, and theoretical work in morphology and suppletion on the other. While grammars usually treat pairs such as cit vs. nás as cases of suppletion, authors of theoretical works (e.g. Mithun 1988b: 214, Mel’čuk 1994: 386-387, Corbett 2000: 258-259) argue that such cases are to be considered separate lexical items which are related semantically but not paradigmatically. This conflict in views seems to originate from the way one defines suppletion. Apparently, grammar authors are content to focus on the presence of morphological patterns of any kind (including derivational patterns), and any exceptions to these patterns are seen as suppletive. Authors of theoretical works in morphology, on the other hand, appear to restrict the term suppletion to cases such as go vs. went, which represent blatant violations of general inflectional rules. For the purposes of the current chapter, cases of suppletion are considered those which satisfy either of the following criteria: (i) exceptions to very productive derivational patterns; (ii) exceptions to established agreement patterns. As we shall see below, there are languages where the expression of verbal number is purely lexical, that is, there are no morphological means for deriving pluractional verbs. However, in this group, there are languages where the verbal-number pairs/triples also appear integrated with the agreement system, in that the choice of verb stem has to agree in number with the agreement suffixes.
The languages of the sample were classified according to the following parameters: (i) the semantic distinction of singular-(dual)-plural made by the relevant words; (ii) whether these words can be seen as being in a relation of suppletion, in terms of the view of suppletion outlined above.
| Go to map | ||
|---|---|---|
| Value | Representation | |
 |
No singular-(dual)-plural pairs/triples | 159 |
 |
Singular-plural pairs, no suppletion | 12 |
 |
Singular-plural pairs, suppletion | 15 |
 |
Singular-dual-plural triples, no suppletion | 5 |
 |
Singular-dual-plural triples, suppletion | 2 |
| Total: | 193 | |
Mupun in (1)-(3) illustrates a language of the second type (no suppletion): the verbal number pairs do not form part of an agreement system, and the morphological means for deriving plural verbs are not only quite diverse, but also apply to a very limited number of verbs.
The third type covers languages such as Shipibo-Konibo (Panoan; Peru) and also languages such as Krongo (Kadugli; Sudan). In Shipibo-Konibo, almost all verbs have a single root and establish the singular/plural distinction only in the third person, by suffixing the plural marker –kan as illustrated in (6b). The verbs jo- ‘come’ and ka- ‘go’ are the only two verbs that make the number distinction with all persons by using different singular/plural roots, as illustrated by examples with jo- ‘come’ in (6c-d).
(6) Shipibo-Konibo (Valenzuela 1997: 48-49)
a.
jaskat-ax
rama-kama-bi
chii
ja-ke
so-intr
now-until-emph
fire
exist-cmpl
‘And it is so that even until now there is fire.’
b.
ani
jema-nko
icka
joni-bo
ja-kan-ke
big
village-loc
many
person-pl
exist-pl-cmpl
‘There are many people in the city.’
c.
ja-Ø-ra
Kako-nkoniax
jo-ke
3-abs-as
Caco-from.intr
come.sg-cmpl
‘(S)he came from Caco.’
d.
ja-bo-Ø-ra
Kako-nkoniax
be-kan-ke
3-pl-abs-as
Caco-from.intr
come.pl-pl-cmpl
‘They came from Caco.’
Thus in Shipibo-Konibo, the verbs jo- ‘come’ and ka- ‘go’ appear as exceptions to established agreement patterns and are, therefore, described as suppletive.
In Krongo, pluractional verbs (usually expressing iterative and habitual meaning) are derived by means of prefixes or tone changes or reduplication, as illustrated in (7a-c). These processes apply to a great number of verbs. However, there are fifteen verbs which do not use any of these morphological means but rather express iterative and habitual meaning by completely unrelated stems; one of them is shown in (7d).
(7) Krongo (Reh 1985: 201-9, Mechthild Reh, p.c.)
a.
prefix
sg
à-mà
’verbalizer-answer'
pl
t-èe-mà
'verbalizer-pl-answer'
b.
tone
sg
ò-kídò-ònò
'verbalizer -cut.off-suff'
pl
ò-kìdò-onò
'verbalizer -cut.off.pl-suff'
c.
reduplication
sg
t-áa-ʈà
'verbalizer-pref-go.away'
pl
à-ʈàa-ʈá
'verbalizer-rdp-go.away'
d.
suppletion
sg
à-pá-ánà
'verbalizer-hit.sg-suff'
pl
à-rwà-ànà
'verbalizer-hit.pl-suff'
In Krongo the functions of verbal number are very close to those of verbal aspect, such as iterative (i.e. repeated action) and habitual; the singular counterparts of the derived verbs usually express punctual action. The derivational devices illustrated above are, according to the grammatical description, lexically conditioned. However, the derivation as a whole is a very wide-spread one and the suppletive verbs illustrated in (7d) are easily perceived as exceptions to a pattern and hence as paradigmatically related, rather than simply as different words.
The fourth type is illustrated by Navajo (Athabaskan; Arizona and New Mexico). In this language, we observe verb triples of the kind sg.du.pl as well as verb pairs of the kind sg.pl.
(8) Navajo (Young and Morgan 1987: 143, 133)
a.
‘go, walk’
b.
‘die’
-yá-
'go.sg'
-tsá-
'die.sg'
-’ash-
'go.du'
-ná-
'die.pl'
-ááł-
'go.pl'
In languages where verb triples of the kind sg.du.pl are observed, verb pairs of the kind sg.pl are also found; languages with triples only do not seem to exist. As to whether these verbs are in a relation of suppletion or not, Navajo belongs to the language type where the verb selects, rather than agrees with, its arguments (Durie 1986: 358) and as such cannot be described as suppletive, cf. (9) below.
(9) Navajo (Durie 1986: 358)
a.
shí
ashkii
bi-ł
yi-sh-’ash
I
boy
him-with
prog-1sg-walk.du
‘I am walking with the boy.’
b.
nihí
ła’
di-iid-ááł
we
subset
fut-1non.sg-walk.sg
‘One of us will go.’
The agreement affixes –sh- ‘1.sg’ in (9a) and –iid- ‘1.non-sg’ in (9b) follow the number of the subject, whereas the selection of stem follows the number of participants. Thus, postulating suppletion for the alternation of stems such as -’ash and –ááł is not justified.
The fifth and final type is illustrated in (10) with verbs from Slave (Athabaskan; Northwest Territories, Canada).
(10) Slave (Rice 1989: 790-792)
a.
‘be seated’
b.
‘die’
wʰida
‘be seated.sg’
łaninwe
‘die.sg’
wʰíkē
‘be seated.du’
łakenindé
‘die.du/pl’
c'edéhkʷ'i
‘be seated.pl’
Although closely related to Navajo, Slave behaves differently with respect to agreement and stem selection. The alternation of stems not only reflects the number of participants, but also agrees in number with the agreement affixes.
(11) Slave (Keren Rice, p.c.)
a.
ʔehts’é
thí-ke
reciprocal to
aspect/1du-sit.du
‘We two are facing each other.’
b.
kó
gá
de-í-kw’i
fire
near
aspect-1pl-sit.pl
‘We (pl) sat near the fire.’
Generally, in Slave, it is important that the verb stem agree in number with the overt subject marker. Combining a non-singular subject marker with a singular stem as in the Navajo example (9), is impossible in Slave for first and second person subjects (Keren Rice, p.c.). This is why the verb triples/pairs can be described as paradigmatically related and thus suppletive, given the view of suppletion adopted in this chapter.
The number of verbal number pairs/triples per language varies from 1 to as many as 18, as in !Xun (Ekoka) (Northern Khoisan; Namibia); however in more than half of the languages where they occur, they range between 1 and 4. Semantically, such verbs tend to center around the basic motion verbs ‘go/come’, ‘walk’, ‘fall’, position verbs such as ‘lie’, ‘sit’ and ‘stand’ as well as their transitive counterparts ‘lay’ and ‘put’; very frequent are ‘die’ and ‘kill’, less common ‘sleep’.
Verbal number pairs/triples are very common in North America. In South America, they occur with languages from the Macro-Gê, Panoan and Tucanoan families. In the Pacific region, verbal number pairs are found in Samoan (Polynesian) as well as in many languages of the Trans-New Guinea family, specifically the Northern branch and languages from the Madang-Adelbert Range branch. The phenomenon is represented in all four major families of Africa, but most strongly in the Nilo-Saharan language family, specifically in subgroups such as Eastern Sudanic and Kunama (both cover languages spoken in Sudan and Ethiopia). In Asia, verbal number pairs/triples are observed in Ainu as well as Ket and Burushaski. The western limit of this phenomenon appears to be the Caucasus.
This phenomenon bears on several matters of theoretical and descriptive importance: the category of verbal number and its various expressions treated as in (Gil 1993, Corbett 2000) as well in (Mithun 1988b) from a historical perspective; the distinction between semantic selection and syntactic agreement discussed in Durie (1986); the notion of suppletion (Mel’čuk 1994); and the emergence of paradigmatic relations between semantically related words (Veselinova 2003).