1. Overview of the values

Negative indefinite pronouns like 'nobody', 'nothing', 'nowhere' may or may not co-occur with the ordinary marker of predicate negation. A language where they always co-occur with predicate negation is Russian. In (1a-b), omission of the predicate negator ne  would render the sentences ungrammatical.

(1) Russian

a.	Nikto		ne	prišel.
	nobody		neg	came.
	'Nobody came.'
b.	Ja	ne	videla	ničego.
	I	neg	saw	nothing
	'I saw nothing.'

A language where negative indefinites never co-occur with predicate negation is German. In (2a-b), use of the predicate negator nicht would render the sentences ungrammatical (*Niemand kam nicht; *Ich habe nichts nicht gesehen).

(2) German

a.	Niemand	kam.	b.	Ich	habe	nichts	gesehen.
	nobody	came		I	have	nothing	seen

Languages may also show mixed behaviour. For example, Spanish negative indefinites preclude predicate negation when they precede the verb (as in 3a), but predicate negation is required when they follow the verb (as in 3b).

(3) Spanish

a.	Nadie	vino.	b.	No	vi	nada.
	nobody	came		neg	I.saw	nothing

Finally, some languages require a negative existential construction with a negative or negated existential verb as the main predicate, and the word translating the English indefinite pronoun as the subject. An example is Nêlêmwa (Oceanic; New Caledonia):

(4) Nêlêmwa (Bril 1999: 84)

Kia	agu	i	uya.
not.exist	person	3sg	arrive
'Nobody came.' (Lit. 'There isn't a person who came.')

One could say that these languages, too, show co-occurrence with predicate negation, but they are assigned to a separate category here because these languages do not allow their counterparts of negative indefinites to occur in the standard nominal or adverbial positions.

Thus, four values are distinguished on the map:

Values of Map 115A. Negative Indefinite Pronouns and Predicate Negation

Go to map
	Value	Representation
	Negative indefinites co-occur with predicate negation	170
	Negative indefinites preclude predicate negation	11
	Negative indefinites show mixed behaviour	13
	Negative existential construction	12
Total:		206

2. What counts as a negative indefinite pronoun

Negative indefinite pronouns are defined as nominal or adverbial expressions that directly translate 'nobody', 'nothing', 'nowhere', 'never' (etc.) in sentences (1)-(3), independently of whether they co-occur with predicate negation. They fall into the same formal types as the indefinite pronouns that are discussed in Chapter 46:

1. interrogative-based negative indefinites;

2. generic-noun-based negative indefinites; and

3. special negative indefinites.

Typically, languages with interrogative-based specific indefinites also have interrogative-based negative indefinites, e.g. Georgian vin  'who', vi-ɣac  'somebody', ara-vin  'nobody' (Hewitt 1995: 81-82), and languages with generic-noun-based specific indefinites also have generic-noun-based negative indefinites, e.g. Kana (Benue-Congo; Nigeria) nū  'thing; something; nothing' (Ikoro 1996: 345).

Note that the definition of negative indefinite pronoun  that is used here does not make reference to any kind of "inherent" negative sense. Some linguists feel that English nobody  is "inherently negative", while anybody, which co-occurs with predicate negation (as in I did not see anybody ), is not "inherently negative" (e.g. Bernini and Ramat 1996: 118ff.). Such a claim would seem to be even more plausible with items such as Kana nū, which probably simply means 'thing' and does not even have a special sense in (5), where it occurs in the scope of negation.

(5) Kana (Ikoro 1996: 345)

ḿ-síì	yāɛ̄	nū.
I-neg	buy	thing
'I have not bought anything.'

However, in this chapter no such distinction between two types of "inherently negative" and "not inherently negative" indefinites is made, because it is unclear which criteria should be used to identify them systematically. One criterion, the occurrence in elliptical negative answers (Who did you see? Nobody/ *Anybody ), is not practical for a large-scale cross-linguistic study because of insufficient data, and it does not always coincide with another possible criterion, the non-occurrence in non-negative contexts. For instance, Modern Greek kanénas  'anybody, nobody' can be used in elliptical negative answers (unlike English anybody ), but it is widely used in non-negative contexts (like anybody ).

Thus, for this chapter all nominal expressions that correspond to 'nobody' and 'nothing' are regarded as "negative indefinite pronouns", even though in many or most languages the negative sense is contributed exclusively by the predicate negation (see Haspelmath 1997: 193-200 for further discussion).

3. What counts as predicate negation

For a language to qualify as belonging to the first type (coccurrence with predicate negation), the negator must be the same as in ordinary negative clauses lacking a negative indefinite pronoun. This is indeed by far the most common situation, illustrated in (6) from Abun (West Papuan; Papua, Indonesia). This language has bipartitite predicate negation (yo...nde ).

(6) Abun (Berry and Berry 1999: 131, 143)

a.	Isak	yo	ben	suk	mo	nggwe	nde.
	Isak	neg	do	thing	loc	garden	neg
	'Isak did not do anything at the garden.'
b.	Án	yo	ma	mo	nu	nde.
	they	neg	come	loc	house	neg
	'They did not come to the house.'

In most cases, there is no danger of mistaking the predicate negator for a part of the indefinite, because the negator has a fixed position in the clause, generally close to the verb (see Dryer 1988b). However, sometimes the position of the negator is independent of the position of the verb, as in Gooniyandi (Bunaban; Western Australia):

(7) Gooniyandi (McGregor 1990: 479, 389)

a.	Mangaddi	ngoorndoo-ga-ngaddaya		wardnginbidda.
	neg	who-erg-indef		they.took.me
	'No one gave me a lift.'
b.	Mangaddi	yangbala-ngga	goornbu	doownga.
	neg	young.man-erg	woman	he.got.her
	'Young men didn't take wives.'

In (7a), one might conceivably interpret mangaddi ngoorndoo-ga-ngaddaya  as a single negative indefinite that does not co-occur with predicate negation; but as (7b) illustrates, Gooniyandi's negator typically occurs initially and need not be adjacent to the verb. Thus, mangaddi  in (7a) is just the ordinary predicate negation, and we have the same negative construction in (7a) and (7b).

The situation is different in Tzutujil (Mayan; Guatemala):

(8) Tzutujil (Dayley 1985: 245, 321)

a.	Majuun	x-in-tz’at.
	nothing	comp-1sg.erg-see
	‘I saw nothing.’
b.	Ma	x-b’e	ta	jar	Aa	Lu7.
	neg	comp-go	irreal	the	young	Pedro
	‘Pedro did not go.’

In (8a) ma-  in ma-juun  [neg-one] does not count as predicate negation, because (i) the predicate negator is generally immediately preverbal, as in (8b), (ii) it is a separate word rather than a prefix, and (iii) it is generally accompanied by the irrealis particle ta  in the completive aspect. Thus, (8a) represents a different negative construction from (8b).

Besides clear cases like Gooniyandi and Tzutujil, there are also truly doubtful cases like Chalcatongo Mixtec (Oto-Manguean; Mexico). In this language, predicate negation is ordinarily expressed by the pre-predicate clitic tu=, as in (9a). However, this clitic may also precede a focused noun phrase, as in (9b), so that it is possible that we are dealing with the same construction in (9c). If this were the case, Chalcatongo Mixtec would be classified as type 1.

(9) Chalcatongo Mixtec (Macaulay 1996: 120, 124)

a.	Ndežu	tú=žaʔu.
	food	neg=expensive
	'The food is not expensive.'
b.	Tú=šūʔū	ñabaʔa=rí.
	neg=money	have=1
	'I didn't have any money.'
c.	Tú=ndéú	ní=kii.
	neg=who	comp=come
	'Nobody came.'

However, I have classified it as type 2 (precluding predicate negation), because ndéú  'who' does not seem to be otherwise used in an indefinite sense, and in the corresponding tú=kwɨtɨ  'nothing', the element kwɨtɨ  is unique. Thus, tú=ndéú  and tú=kwɨtɨ  seem to be non-compositional expressions, but ideally one would have to have more information to be able to confirm this analysis.

4. Types of mixed behaviour

There are several different ways in which a language can show mixed behaviour. One type is the Spanish type, where the use of the predicate negator depends on the position of the indefinite with respect to the verb (see (3) above). Another type is represented by Swedish, which has two different sets of indefinites, one of which co-occurs with predicate negation, while the other precludes it (as the translations show, English is mixed in the same way):

(10) Swedish

a.	Jag	vet	inte	något.
	I	know	neg	something
	'I do not know anything.'
b.	Jag	vet	ingenting.
	I	know	nothing
	'I know nothing.'

Third, a language may have a bipartite negator, only one of whose parts is precluded by the negative indefinite. This is the case in Maltese, whose ordinary predicate negator consists of the preverbal particle ma  and the postverbal suffix -x. When the clause contains a negative indefinite, only ma  appears (French, with its bipartitite negator ne...pas, behaves in the same way).

(11) Maltese (Haspelmath 1997: 200)

a.	Patricia	ma	rat-x	lit-tifel.
	Patricia	neg	she.saw-neg	acc.the-boy
	‘Patricia did not see the boy.’
b.	Patricia	ma	rat(*-x)	xejn.
	Patricia	neg	saw	nothing
	‘Patricia did not see anything.’

5. The issue of "double negation"

In a number of Germanic languages whose normative varieties have indefinites precluding predicate negation, these indefinites do co-occur with negation in non-standard varieties. For example, in Bavarian German, structures like (12) are completely normal (see, e.g., Weiß 1999).

(12) Bavarian German

Gesdan	han	e	neamd	ned	gseng.
yesterday	have	I	nobody	not	seen
‘Yesterday I saw nobody.’

Such sentences are often criticized from a normative point of view as being "illogical", because according to logic, two negations should cancel each other out. Two problems with this point of view are (i) that the negative indefinites in question cannot easily be characterized as "inherently negative" because this concept is difficult to define (see §2 above), and (ii) that the co-occurrence of negative indefinites with predicate negation, as in (12), is by far the most common type in the world's languages. One would not, after all, want to say that only a few Germanic languages are "logical". In fact, it seems that ultimately the normative criticism of "double negation" is not based on logic, but on the prestige of Latin, which served as the model for the standard varieties of several western European languages, and which had negative indefinites precluding predicate negation (cf. Haspelmath 1997: §8.2).

Although it is not possible to say whether a negative indefinite is "inherently negative", we can ask whether it contains a morpheme that functions as a negator elsewhere in the language. And here an interesting generalization emerges: in the great majority of languages where negative indefinites preclude predicate negation, the indefinites themselves contain a negation marker. Some examples are given in (13).

(13)

Ossetic	ma-cy	[neg-what]	'nothing'
Chamorro	ni hayi	[neg who]	'nobody'
Tzutujil	ma-juun	[neg-one]	'nobody'
Chalcatongo Mixtec	tú=ndéú	[neg=who]	'nobody'
Dutch	n-iets	[neg-something]	'nothing'

Exceptions to this generalization are found exclusively in western European languages (e.g. German kein  'no', Swedish ingen  'no(body)', and several Romance expressions which preclude verbal negation only when occurring preverbally, such as Spanish nada 'nothing', Italian mai  'never'). The probable diachronic explanation for this is that the pattern where negative indefinites preclude predicate negation typically arises by "attraction" of the negator to the negative indefinite pronoun (see (9) above for a possible intermediate stage in this process).

6. Geographical distribution

Negative indefinites precluding predicate negation are concentrated in Western Europe and in Mesoamerica, and languages with existential constructions are found particularly among the Oceanic languages of the small Pacific islands. The great majority of the world's languages have negative indefinites co-occurring with predicate negation.