9.1136, Sum: NPIs and FCIs

LINGUIST Network linguist at linguistlist.org
Tue Aug 11 15:57:15 UTC 1998

LINGUIST List:  Vol-9-1136. Tue Aug 11 1998. ISSN: 1068-4875.

Subject: 9.1136, Sum: NPIs and FCIs

Moderators: Anthony Rodrigues Aristar: Wayne State U. <aristar at linguistlist.org>
            Helen Dry: Eastern Michigan U. <hdry at linguistlist.org>

Review Editor:     Andrew Carnie <carnie at linguistlist.org>

Editors:  	    Brett Churchill <brett at linguistlist.org>
		    Martin Jacobsen <marty at linguistlist.org>
		    Elaine Halleck <elaine at linguistlist.org>
                    Anita Huang <anita at linguistlist.org>
                    Ljuba Veselinova <ljuba at linguistlist.org>
		    Julie Wilson <julie at linguistlist.org>

Software development: John H. Remmers <remmers at emunix.emich.edu>
                      Zhiping Zheng <zzheng at online.emich.edu>

Home Page:  http://linguistlist.org/

Editor for this issue: Martin Jacobsen <marty at linguistlist.org>


Date:  Tue, 11 Aug 1998 14:40:01 -0400
From:  Asya Pereltsvaig <aperel at po-box.mcgill.ca>
Subject:  Sum: NPIs and FCIs

-------------------------------- Message 1 -------------------------------

Date:  Tue, 11 Aug 1998 14:40:01 -0400
From:  Asya Pereltsvaig <aperel at po-box.mcgill.ca>
Subject:  Sum: NPIs and FCIs

About a month ago  I posted the following query:

I am looking for languages that use the same morphemes/items for both
Negative Polarity Items and Free Choice Items, like "any" in English.

(1) Kim didn't read any book. - NPI any
(2) * Kim read any book.
(3) Kim can read any book. - FCI any

Any information is greatly appreciated. I will post a summary.

I have received very useful answers from the following people (in no
particular order) to all of whom I am deeply grateful:

Vassilis Christodoulou  	Michael Israel
Ton van der Wouden    		Mika Kizu
Philip Franz Seitz 		Larry Horn
David Gohre 			Natalie Schapansky
Chungmin Lee 			Martin Haspelmath
David Gil 			Torsten Leuscher
Thomas Ernst

Several people have recommended Martin Haspelmath's dissertation (also
available as a book):

Martin Haspelmath's _Indefinite Pronouns_ (Oxford university Press

Other bibliography:

Lin, Jo-Wang (1996) Polarity Licensing and Wh-Phrase Quantification in
Chinese.  UMass dissertation.

Klein, Jared (1997) Indefinite pronouns, polarity, and related
phenomena in Classical Armenian.  Transactions of the Philological
Society 95: 189-245.

Chungmin Lee (1997) 'Negative Polarity and Free Choice: Where Do They
Come From?,' Proceed's of the 11th Amsterdam Colloquium, P.  Dekker et
al (eds.) ILLC/U of Amsterdam.  ---- a unified account.

Chungmin Lee (1996) 'Negative Polarity Items in English and Korean,'
Language Sciences, Vol.  18, Nos 1-2, pp 505-523 (Elsevier).

Gil, David (1991) "Universal Quantifiers and Universal Grammar",
EUROTYP Group 7 Working Paper.

Here are some extracts from the answers I've got:
Vassilis Christodoulou:

although I am only a translator and -let's say- an amateur "linguist",

hope that the following information will help you.

In Greek there are some indefinite pronouns, which, though they have a
negative meaning, they can be used as Free Choice Items, as you call
it, in casual speech. These are:

kanenas, kamia, kanena: nobody (masculine, feminine, neuter) /
somebody tipota *or* tipote: nothing / something pote: never/ some
time pouthena: nowhere / somewhere

Some examples: (Note that "den" is the negative particle and has to be
used, even if you already use the aforementioned pronouns)

-Kanenas den irthe sti sinandisi.
Nobody came at the meeting.
-Irthe kanenas sti sinandisi;
Did anybody come to the meeting?
-Thelo na pao kamia volta.
I want to go for a walk.

-Ides tipota kokkino na pernai;
Did you see anything red passing by?
-Ochi, den ida tipota.
No, I didn't see anything.

-Piges pote sti Gallia?
Have you ever gone to France?
-Den ergazome pote ta Savata.
I never work on Saturdays.

and so on...

Note, that these morphemes are used as NPI's usually in casual speech,
in questions and after subjunctive (Modern Greek does not use
infinitive any more. Instead, it uses a subjunctive
construction. Other Southeast European languages do the same - the
famous Balkansprachbund). In affirmative sentences the correspondent
pronouns are:

kapios, kapia, kapio: somebody (M,F,N)
kati: something
kapou: somewhere
kapote: some time


-Ida kapion na pigeni pros to stathmo.
I saw somebody going towards the station.

The Greek I used is phonetically transcripted, i.e. orthographically
not correct.

Michael Israel:

Haspelmath looks at the expression of indefiniteness in a 40 language
sample and proposes an implicational map of 9 basic functions
indefinites can serve:

1) specific, known to speaker
2) specific, unknown to speaker
3) non-specific, irrealis (non-veridical) contexts
4) occurrence in conditionals
5)  occurrence in questions
6)  occurrence in standard of comparison
7) occurrence with indirect negation
8) free choice
9) occurrence with direct negation

All of the languages he surveys have at least two, and usually several
more than that, types of indefinites, each of which can serve one or
more of the nine functions.  English 'any' words, for example, occur
in functions 4-9. Serbo-Croatian has a series of indefinites, the
'bilo-' series, which can serve functions 4-8, but cannot occur with
direct negation; another series, the "i-NPIs", which can serve
functions 4-7, but cannot function as free-choice items,

and another series, the "ni-NPIs" which occur only in function 9, with
direct negation.

Haspelmath's map, given below, makes predictions about possible
indefinites in the world's languages. The idea is that while

can always serve multiple functions, any given indefinites can only
serve functions which form a contiguous chain on the map.

9       7
                5/4     3
8       6

Thus, to take an extreme example, you'll never find an indefinite
which can only be used with direct negation (9) and for specific
indefinites (1,2).  Other impossibilities would be direct negation and
comparatives (9, 6), without including indirect negation (7); or free
choice (8) and conditionals (4), without comparatives (6).

It's a neat result, and it looks pretty convincing in Haspelmath's
account.  One weakness is that things like "indirect negation" are
never really given an explicit definition, and one generally always
wants to know more about the details of the languages he talks about.
But that's what you get with typology: not enough information about
way too many languages.  Still, Haspelmath strikes me as one of the
best typologists around.

One thing worth noting, it seems to me, is that Haspelmath lists a lot
of languages with indefinites that only work in the free choice
function (e.g. Latin 'quisvis, quislibet'; Icelandic 'semer'; Russian
'ljuboj'); others with indefinites that work in both free choice and
comparative functions (Swedish 'som helst'; Italian '-unque'
indefinites; French 'n'importe' indefinites; and still others with
indefinites that work in the free choice function and a whole bunch of
other functions: English 'any' has functions 4-9; Hindi 'koii-bhii'
has functions 3-9; andTurkish 'hehangi' has functions 2-9.

But what Haspelmath apparently did not find (though his map predicts
he should) are indefinites that serve only in functions (7) and (9),
free choice and direct negation.

Anyway, the answer to your question is going to depend a lot on how
you define NPI. The typical NPI functions are Haspelmath's 4-7 and 9,
plus a few others that don't really fit on H's map ('before' clauses,
restriction on a universal quantifier, complement of non-negative
adversative predicate, etc.).  But many languages have forms which
occur in (4-6) and free choice contexts, but not with direct or
indirect negation.  And others have forms which occur in all of
functions 3-9. These might not really be NPIs, but then again they may
be subject to the some of the same crucial generalizations.

But it all depends on what you're looking for.
Mika Kizu:

I am wondering if Japanese is one of the languages that you are
looking for. I'm not so familiar with NPIs and FCIs in Japanese, but
my translation for your sentences would be something like the

(1)     Kim-wa  dono hon-mo    yom-anak-atta.
        Kim-TOP any  book-also read-NEG-PAST    'Kim didn't read any

(2)     Kim-wa  dono hon-mo    yon-da.
        Kim-TOP any  book-also read-PAST        'Kim read any books.'

(3)     Kim-wa dono hon-mo    yom-e-ru.
        Kim-TOP any book-also read-POT-PRES     'Kim can read any

What I'm not sure about the data above is that 'dono..mo' (any..also)
is really a free choice item in Japanese, because 'dono...mo' seems to
presuppose a list of books. Another thing is that Japanese doesn't
have a number system on nouns, which might be crucial to your

Philip-Franz Seitz:

In Vietnamese, one morpheme occurs in both "NPI any" and "FCI any"
constructions, as well as in "WH which" constructions.  Following are
some examples written in VIQR ("Vietnamese Quoted Readable"-- the
writing system we use for e-mail, etc.) with parses and translations.

The linguistic form at issue is: na`o 'which' (WHICH is standard

WH which
- ------
Co             Kim ddo.c sa'ch na`o  va^.y
father's^sister Kim read  book  WHICH thus
'Which book is Ms. Kim reading?'

NPI any
- -----
Co^             Kim kho^ng ddo.c sa'ch na`o  he^'t
father's^sister Kim NEG    read  book  WHICH all
'Ms. Kim isn't reading any book.'

FCI any
- -----
Co^        Kim co'  the^?      ddo.c sa'ch na`o  cu~ng ddu+o+.c

father's^sister Kim have possible read book WHICH also able all

'Ms. Kim can read any book.'

Larry Horn:

The best single resource on this is the dissertation (at the Freie
Universitt Berlin) of Martin Haspelmath, 1993, entitled A typological
study of indefinite pronouns.  It was later published as a book,
Indefinite Pronouns (Oxford U. P., 1997).  Although I supposed in my
1972 dissertation that English was unusual in this respect (having an
item like 'any' that fits both frames), it turns out that it's more
the rule than the exception, as Haspelmath demonstrates.  Languages
I've come across lately that share this trait are Chinese and
Classical Armenian, as is clear from the following two sources:

Lin, Jo-Wang (1996) Polarity Licensing and Wh-Phrase Quantification in
Chinese.  UMass dissertation.

Klein, Jared (1997) Indefinite pronouns, polarity, and related
phenomena in Classical Armenian.  Transactions of the Philological
Society 95: 189-245.

David Gohre:

Hi there.  I assume that you're familiar with the works of Ljiljana
Progovac.  She's done work on Serbo-Croatian which has the phenomena
you're looking for.  Maybe here 1994 book will contain more examples
from other languages, but that's a start.  You can e-mail her at
lprogov at cms.cc.wayne.edu

Natalie Schapansky:

As far as I know, no languages use the same word for the free choice
and negative polarity reading of anyone. From my study on comparative
constructions, languages tend to use quantifiers like all, the others
for the free choice (or universal) reading of anyone.

Chungmin Lee:

Korean shows a common part between NPIs and FCIs: amu N-to (NPI) vs.
amu N-i-ra-to (FCI). The determiner amu is indefinite and is like
'any,' and -to is a particle meaning 'even' (-i- 'copula, -ra- 'decl S

Martin Haspelmath:

See my recent book "Indefinite pronouns" (Oxford: Oxford University
Press 1997), where I discuss in some detail 40 languages and the way
they express PCIs and NPIs of various kinds, and why we find these
patterns of syncretism (and actually many other patterns). About half
of the languages of my sample show this particular syncretism.

David Gil:

Hebrew _kol_ can mean "all", "every", free-choice "any", and also
occur in NPI environments.

Here are a couple of examples:

Eyn kol corex laasot et ze

Harofe lo maca kol siman lamaxala

As you can see, the construction is a little bit formal, and
apparently restricted in its distribution, e.g.. you can't say:

?*Lo kaniti kol sefer

Here you would use the other NPIs:

Lo kaniti af / Sum sefer

>Is it true that in order to get "kol" as an NPI you need an existence
>verb of some kind (in the case of "maca" the sentence means roughly
>that there is no signs of illness)?

I'm not sure, but cf.

HabalaS lo maca kol siman lealimut bezirat harecax

?* HabalaS lo maca kol kadur bezirat harecax

It looks like the noun qualified by NPI kol must be abstract, not
concrete.  But I don't know if this is the correct generalization.

Torsten Leuschner:

It seems that the Scandinavian languages apply to your description,
i.e.  certainly Swedish and as far as I am aware also Danish and
Norwegian at least. These languages have a strongly grammaticalized
series of indefinite pronouns based on a WH pronoun and the typical
free-choice marker 'som helst' (lit. as favourite; exs.: vad som
helst, vem som helst, vilken som helst etc.) that, as far as I am
aware, can be used in all three (!) senses of 'any' shown in your

Actually my Swedish etc. is pretty bad so I give you references rather
than examples:

P. Holmes / P. Hunchliffe, Swedish: a Comprehensive Grammar. London:
Routledge 1993 (have a nice chapter on 'som helst' where they refer to
'any' for comparison)

There is a Danish grammar in Routledge too which I cannot consult at
the moment but would expect to be useful for any (!) comparison with


Tom Ernst have contacted me as well and told me that Chinese is one
such language.


Thanks again to all the people who answered my question. Best wishes
to everybody,

Asya Pereltsvaig
Department of Linguistics
McGill University
1001 Sherbrooke St. West
Montreal, Quebec, H3A 1G5, CANADA
aperel at po-box.mcgill.ca
tel. (514) 931-5046

LINGUIST List: Vol-9-1136

More information about the Linguist mailing list