1 MASS-COUNT DISTINCTIONS IN BARE PPS 1 Bert Le Bruyn, Henriëtte de Swart, Joost Zwarts Utrecht Institute of Linguistics OTS, Utrecht University Ms. January 2011 Word count: 11734 words Abstract Singular count nouns in Germanic and Romance languages usually require a determiner in regular argument position, but occasionally occur ‘bare’, e.g. in PPs like at school, in jail, without exception. This paper investigates countability issues in such bare PPs in the light of recent grammatical theories which claim that count interpretations of nouns require overt, disambiguating functional structure. Such structure is obviously missing in bare PPs. Evidence from productive subclasses of bare PPs reveals that we cannot ignore the mass/count distinction in bare PPs, and cannot reduce all noun interpretations in this environment to a mass use. Subclasses of nouns that are incompatible with a mass interpretation, selectional restrictions induced by prepositions, modification by shape/dimension adjectives, and the ambiguities generated by flexible nouns shows that count interpretations arise in bare PPs in similar ways as in other constructions. We conclude that countability is not an exclusively grammatical feature, and maintain that both mass and count meanings must be visible in the lexical semantics. 1. Mass-count distinctions with bare nominals: background and issues 1.1 Preliminary observations about article use in English and similar languages Germanic and Romance languages like Dutch, English, Italian, Spanish, etc. generally use full DPs in regular argument position. However, articles or determiners can be optionally left out with plurals or mass nouns. 2 This leads to the paradigm in (1): (1) a. I read a book/the latest book/every book by Chomsky. b. I read books/the latest books/two books/many books by Chomsky. c. I read poetry/much poetry by J.C. Bloem. d. *I read book/latest book by Chomsky. e. Johnny the termite prefers book over newspaper. Count nouns such as book can appear with an indefinite article (a), a definite article (the) or a quantifier (every) as illustrated in (1a). Plurals can appear with a definite article (the), a numeral (two), or some other determiner or quantifier (many), but they also be bare (i.e. lacking an article, a determiner or a quantifier) as in (1b). Mass nouns display a similar distribution as plurals, cf. (1c), but they do not tolerate numerals, and sometimes use a different form of the quantifier (many in 1b, much in 1c). Count nouns generally resist the bare form (1d), unless they get a mass interpretation (1e). The paradigm in (1) can be found in many standard grammars of the language. This paper is not concerned with languages such as Mandarin Chinese, Hindi, Hebrew, Russian, etc. in which bare (count) singulars freely occur in regular argument position, but is limited to English-type languages with extensive and 1 We gratefully acknowledge the Netherlands Organization for Scientific Research for financial support. 2 This is not the case in French, cf. de Swart & Zwarts (2009), Le Bruyn (2010) and references therein for discussion of the partitive articles du, de la, des in this language. Interestingly, we still find bare PPs and other bare constructions in French, cf. (3) below.
22
Embed
MASS COUNT DISTINCTIONS IN BARE S1 - let.uu.nl · PDF fileIf nouns are classified as mass or count in the lexicon, ... fails to account for mass-count distinctions in bare PPs (Section
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
MASS-COUNT DISTINCTIONS IN BARE PPS1
Bert Le Bruyn, Henriëtte de Swart, Joost Zwarts
Utrecht Institute of Linguistics OTS, Utrecht University
Ms. January 2011
Word count: 11734 words
Abstract Singular count nouns in Germanic and Romance languages usually require a determiner in regular
argument position, but occasionally occur ‘bare’, e.g. in PPs like at school, in jail, without exception.
This paper investigates countability issues in such bare PPs in the light of recent grammatical theories
which claim that count interpretations of nouns require overt, disambiguating functional structure.
Such structure is obviously missing in bare PPs. Evidence from productive subclasses of bare PPs
reveals that we cannot ignore the mass/count distinction in bare PPs, and cannot reduce all noun
interpretations in this environment to a mass use. Subclasses of nouns that are incompatible with a
mass interpretation, selectional restrictions induced by prepositions, modification by shape/dimension
adjectives, and the ambiguities generated by flexible nouns shows that count interpretations arise in
bare PPs in similar ways as in other constructions. We conclude that countability is not an exclusively
grammatical feature, and maintain that both mass and count meanings must be visible in the lexical
semantics.
1. Mass-count distinctions with bare nominals: background and issues
1.1 Preliminary observations about article use in English and similar languages
Germanic and Romance languages like Dutch, English, Italian, Spanish, etc. generally use full
DPs in regular argument position. However, articles or determiners can be optionally left out
with plurals or mass nouns.2 This leads to the paradigm in (1):
(1) a. I read a book/the latest book/every book by Chomsky.
b. I read books/the latest books/two books/many books by Chomsky.
c. I read poetry/much poetry by J.C. Bloem.
d. *I read book/latest book by Chomsky.
e. Johnny the termite prefers book over newspaper.
Count nouns such as book can appear with an indefinite article (a), a definite article (the) or a
quantifier (every) as illustrated in (1a). Plurals can appear with a definite article (the), a
numeral (two), or some other determiner or quantifier (many), but they also be bare (i.e.
lacking an article, a determiner or a quantifier) as in (1b). Mass nouns display a similar
distribution as plurals, cf. (1c), but they do not tolerate numerals, and sometimes use a
different form of the quantifier (many in 1b, much in 1c). Count nouns generally resist the
bare form (1d), unless they get a mass interpretation (1e). The paradigm in (1) can be found in
many standard grammars of the language. This paper is not concerned with languages such as
Mandarin Chinese, Hindi, Hebrew, Russian, etc. in which bare (count) singulars freely occur
in regular argument position, but is limited to English-type languages with extensive and
1 We gratefully acknowledge the Netherlands Organization for Scientific Research for financial support.
2 This is not the case in French, cf. de Swart & Zwarts (2009), Le Bruyn (2010) and references therein for
discussion of the partitive articles du, de la, des in this language. Interestingly, we still find bare PPs and other
bare constructions in French, cf. (3) below.
2
systematic article use along the lines of (1), i.e. basically Germanic and Romance languages.
Examples will be taken predominantly from English and Dutch.
There is an ongoing debate in the literature concerning the question whether the
mass/count distinction is located in the lexicon or the grammar. If nouns are classified as mass
or count in the lexicon, we can take book to be a count noun, which leads to a treatment of
examples like (1e) in terms of coercion from count to mass meanings. Alternatively, if the
mass/count distinction is in the grammar, nouns like book are underspecified in the lexicon,
and can appear in either count (1a) or mass environments (1e). This debate will be
summarized in Section 3 below.
Most discussions of countability issues relate to particular functional structure that
indicates a mass/count use, such as determiners, classifiers, etc. We take a radically different
approach by focusing on absence of functional structure in bare nominal constructions. There
is a range of special configurations in which bare (apparently count) singulars appear in
Germanic and Romance languages such as bare predication (Jones is chairman of the board),
bare coordination (The medicines will protect mother and child) or bare noun incorporation
(Terry watched television) play a crucial role in the argumentation. We refer to de Swart &
Zwarts (2009) for an overview of bare constructions in general, and Zwarts (2010) for an
exhaustive descriptive overview of bare constructions in Dutch. This paper takes one of these
configurations, namely prepositional phrases (PPs) as the empirical domain of our
investigation. There are PPs in which the object of the preposition is a bare nominal. We will
call these bare PPs.3 Such PPs provide a rich empirical domain for our investigation, because
this is a widespread construction in Germanic and Romance languages, which allows for the
application of a range of criteria concerning the mass/count distinction even within the bare
PP construction itself.
Bare PPs are intriguing, because they leave out the article on the nominal complement,
even if the noun looks like it might be a count noun (or is elsewhere used as such). Some
English examples are provided in (2):
(2) after school, at bay, at local level, by car, from hospital, from close range, in bed, in
greater detail, on television, on level ground, over dinner, per year, to hell
Article drop in PPs is neither completely free, nor limited to idiomatic constructions, as will
become clear in Section 2. In this paper, we argue that the mass-count distinction is operative
in bare PPs, which implies that we need to address the status of that distinction in the lexicon-
syntax-semantics interface (Section 3). We will argue that a purely grammatical approach
fails to account for mass-count distinctions in bare PPs (Section 4), and provide evidence
from within the bare PP that both mass and count interpretations can be elicited (Section 5).
Empirical support comes from monolingual and multilingual corpus investigations.
2. Lexical, syntactic, semantic and pragmatic features of bare PPs
Bare PPs constructions are prepositional phrases with a nominal complement that is lacking
an article, a determiner or a quantifier, and appears without plural marking. Bare PPs occur in
languages like English, Dutch, German, French, Spanish, and many other languages that have
an obligatory use of a determiner with singular count nouns in regular argument position.
Section 1 provided some English examples in (2). In (3), we provide some triplets in English,
3 Bare PPs are also called determinerless PPs or P+N combinations. Strictly speaking it is not the PP that is bare
or determinerless, of course, but the noun phrase inside it, but we trust that the short term ‘bare PP’ will not lead
to confusion.
3
Dutch and French, showing that all three languages have bare PPs, but do not always use
them in the same way:
(3) a. by train (En) – per trein (Du) – en train (Fr)
b. at school (En) – op school (Du) – à l’école (Fr)
c. in prison (En) – in de gevangenis (Du) – en prison (Fr)
d. at the office (En) – op kantoor (Du) – au bureau (Fr, au < à le)
e. without a hat (En) – zonder hoed (Du) – sans chapeau (Fr)
(3a) illustrates the bare PP configuration of a preposition combining with a noun that indicates
a means of transportation to express the way of travelling in all three languages under
consideration. The comparison of (3b), (3c) and (3d) reveals that all three languages use bare
PPs consisting of a locative P and a noun indicating some sort of institution, but are subject to
different lexical constraints on prepositions and nouns. Articles are underlined, and indicate
that all three languages alternate between bare and non-bare PPs. (3e) indicates the
configuration of with/without type prepositions that Le Bruyn et al. (2010) observe to
frequently occur with articleless nouns in Dutch and French. The examples in (3) show that
the study of bare PPs is relevant to lexical semantics and lexicography, because of the lexical
restrictions involved, and the cross-linguistic variation we find.
The investigation of bare PPs is also relevant to the semantics-pragmatics interface.
On the one hand, we typically find enriched, stereotypical meanings in bare PP
configurations, as observed by Horn (1993) and Levinson (2000) and studied in detail by
Stvan (1998). For examples such as (4), Stvan (1998) argues that the location of the figure
(my husband) at the ground (the prison) is more than a physical location, but comes with
associations about activities or role of the person in this location.
(4) a. My husband is in prison. (≈ as a prisoner)
b. My husband is in the prison
(≈ as a visitor or as the handyman who came to do some repairs)
As a result of the focus on the more abstract institution rather than the concrete location itself,
the bare nominal displays weak referentiality features, as observed by Stvan (2009). Weak
referentiality indicates that the nominal is less likely to serve as the anchor for an anaphoric
pronoun in subsequent discourse. This leads to the contrast in (5):
(5) a. John is in prison. #It is is brick building.
b. John is in the prison. �It is a brick building.
The contrast in (5) has been empirically tested (for Dutch), by Scholten & Aguilar (2010).
Weak referentiality is a typical ingredient of bare constructions in English-type languages, as
highlighted by de Swart & Zwarts (2009).
The lexical restrictions on bare PPs, combined with their enriched meanings and
weakened discourse referentiality features might lead one to think that the examples in (2) and
(3) are highly idiomatic, and should be treated in the lexicon. Computational linguists have
been working on extracting multi word expressions from large corpora to build extended
theories of the lexicon. Baldwin et al. (2003, 2006), Van der Beek (2005), and Dömges et al.
(2007) followed by Kiss (2008) and Kiss et al. (2010) have applied these computational
methods to bare PP configurations in English, Dutch, and German respectively. After careful
investigation, they come to the conclusion that a listing approach is not tenable, because there
is too much productivity. As Baldwin et al. (2006) point out, the by+N[means of
4
transportation] construction in English is conventionally used in by train, by bus, by car, by
bike… but tolerates a creative use as in by flying carpet. Dömges et al. (2007) statistically
argue that bare PPs are productive. The method they use is inspired by the productivity
measure developed by Baayen (2001) and the statistical models in Evert (2004). Informally, a
bare PP involving a given preposition X is said to be productive if the likelihood of finding
new X+N combinations stabilizes after n words at some positive value (significantly)
different from 0. This – Dömges et al. (2007) claim – is e.g. what we find for bare PPs with
unter (‘under’) when browsing through large German newspaper corpora. The conclusion
they draw is that the creative use of bare PPs is too common to be ignored. Accordingly, the
starting point of our investigation is that bare PPs need to be accounted for at the interface
between lexicon, syntax and semantics.
Lexical constraints on prepositions and nouns are investigated in Le Bruyn et al.
(2010). De Swart & Zwarts (2009) drew attention to restricted modification possibilities:
(6) a. in jail – in county jail - in *(the) new jail
b. to college – to technical college – to *(the) best college
Only compound-like modifiers that form a fixed combination with the noun are allowed in
bare PPs. Examples (3) – (6) taken together show how bare PPs differ from full PPs in
different respects.
If bare PPs need to be investigated in the grammar, one issue that arises is the question
of countability. Recall that bare plurals and bare mass nouns are fully regular in all kinds of
argument position. This includes the nominal complement position of a preposition. So the
only ‘special’ instances of bare PPs in Germanic and Romance languages involve singular
count nouns. It is easy to exclude bare plurals on formal grounds, and restrict our
investigation to nouns without number marking. However, that leaves us with a large set of
prepositions followed by a mass noun, as illustrated in (7) (examples from the British
National Corpus):
(7) a. in water, to music, under pressure, with money, without food.
If bare PPs with singular count nouns are productive, even in a somewhat limited way, then
the question arises how we distinguish the ‘special’ constructions in (2)-(6) from the fully
regular PPs in (7). We need to resolve the issue of the mass-count distinction in bare PPs in
order to develop the syntax-semantics interface of bare PPs in a comparative perspective. In
order to see what the issues are, and how we should address them, we propose to take a step
back, and review the debate on countability in the linguistics and philosophical literature more
in general. As we will see in Section 3, our take on the mass-count distinction in bare PPs
depends to a large degree on whether we take this distinction to be operative in the lexicon or
in the grammar.
3. The mass-count distinction: lexicon or grammar?
The classical view locates the mass-count distinction in the lexicon (Pelletier 1975, Link
1983, Bunt 1985, Gillon 1992). There is a lexical distinction between mass and count nouns,
which is reflected in the ontology. A particularly influencing view is the one advanced by
Link (1983), who maintains that count nouns denote join semi-lattices with minimal parts, and
mass nouns involve semi-lattices without minimal parts. There are fairly reliable criteria for
classifying nouns as count nouns, for instance the possibility to combine them with an
5
indefinite article (a book vs. #a sand). There is also the possibility of plural marking (pen-
pens) that is not available to mass nouns (sand - #sands). Count nouns combine with a
numeral without the need for a unit of measurement (two tables), but mass nouns do not (#two
sands vs. two kilos of sand). Some English determiners are sensitive to the mass/count nature
of the noun (many dollar bills vs. much money, few successes vs. little time).
Not all languages implement the mass/count distinction in the same way. Doetjes (to
appear) offers a thorough cross-linguistic comparison of number marking languages and
classifier languages. Since the empirical scope of our paper is restricted to Germanic and
Romance languages, we will not be dealing here with classifier languages, but focus on
languages with a singular/plural distinction.
From the beginning of the debate, it has been clear that the mass-count distinction
needs to be flexible. Many mass nouns have count uses: wine-a wine, coffee-two coffees, love-
a love, thread-a thread, stone-a stone. And similarly, many count nouns have mass uses in the
right context: there is dog/ stone/ chicken on the floor, that’s quite a bit of table/carpet for the
money, quite a bit of/a lot of/too much dog/chicken/table/carpet. Pelletier (1975) proposed the
Universal Grinder to model mappings from count interpretations onto mass ones (in particular
substance interpretations), see also Pelletier & Schubert (1989). Bunt (1985) added the
Universal Sorter and Jackendoff (1991) the Universal Packager to model mappings from mass
to count interpretations, in particular conventionalized portions of a substance as in a coffee, a
cake or kind interpretations as in a beautiful red wine from Napa Valley). The lexical view is
still widely accepted, and gets interesting cross-linguistic applications, as for example the
analysis of restrictions on the Universal Grinder in Mandarin Chinese by Doetjes et al. (2009).
Notwithstanding its success, the lexical view also faces problems, for instance when it
comes to nouns like furniture and cattle, which are mass nouns, even though they have