Transcript Why have-verbs can take bare nominals

prefinal version
Relational nouns and
existential have
Bert Le Bruyn, Henriëtte de Swart, Joost Zwarts
SALT 23
Our perspective
The Weak Referentiality Project
The Weak Referentiality project focuses on nominal
expressions that do not set up individual discourse
referents and more generally on 'non-standard'
arguments
> bare nominals
(John is in jail, John had pen and
paper to picture the whole event, ...)
> weak definites
(John was brought to the hospital,
John is doing the dishes, ...)
Henriëtte de Swart, Martin Everaert, Joost Zwarts, Ana
Aguilar-Guevara & Bert Le Bruyn
2
Our perspective
With and without
One of the subprojects we have been working on is the
analysis of bare nominals and weak definites in PPs.
Two prepositions that stand out across the languages we
have looked at (Dutch, French, Norwegian, Spanish,
English) are with and without.
They take bare nominals more productively than other
prepositions.
3
Our perspective
Link with have verbs
In a number of languages have verbs fairly productively
take bare nominals:
> Tengo coche (Spanish)
I-have car
See Dobrovie-Sorin, Bleam & Espinal (2006), Espinal &
McNally (2011)
> Similar phenomenon in all Romance languages,
Norwegian (Borthen 2003), Greek (Lazaridou-Chatzigoga
& Alexandropoulou to appear)
4
Our perspective
The start of this paper
> By looking at the literature on have with bare nominals
we hoped to find some deeper insight into why have
verbs productively take bare nominals so that we could
recycle it for with and without.
> Despite the detailed analyses provided in the literature,
there didn't appear to be such a deeper insight.
> This is when we started to look for a link between two
parts of the literature that developed independently,
hoping to find this insight...
Have verbs and bare nominals
Existential have
5
A proposal on the basis of
the literature on existential
have
Existential HAVE
Mary has a brother.
John has a friend.
relational
indefinite
7
The basic intuition in the literature
Mary has a brother.
This sentence is not about a 'possession'
relation holding between Mary and a brother.
It's about a brother relation holding between
Mary and someone.
> The brother relation - in some way or other has to make it to a higher level so that it can
directly combine with Mary.
8
Partee (1999)
Gist of the analysis
Direct connection between brother and Mary through quantifying-in.
Mary
have
v(exist(v))
a
brother
z
xy(brother of (y,x))
y(brother of (y,z))
Py(brother of (y,z)&P(y))
z y(brother of (y,z)&exist(y))
y(brother of (y,mary)&exist(y))
9
Landman (2004)
Gist of the analysis
Direct connection between brother and Mary by:
> adopting a modifier semantics for a
> analyzing have as a verb that selects relations and binds their
internal argument
Mary
have
a
Rpq(R(p)(q))
brother
xy(brother of (y,x))
xy(brother of (y,x))
pq(brother of (q,p))
q(brother of (q,mary))
10
Summary
> Both analyses make sure the indefinite article and
have don't stand in the way of Mary and brother.
> Existential quantification in Partee's analysis comes
from the indefinite article. This is less easily
transposable to constructions without indefinite article.
> Landman's analysis puts the existential quantification
at the level of the verb. We will - for the moment - follow
his insight that have is a verb that mediates between
relations at the nominal level.
> We will however not adopt Landman's modifier
semantics of the indefinite article as this would
eliminate the difference between bare and non-bare
nominals.
11
Putting things together
Our question was why have verbs would take bare
nominals more productively than other verbs.
It turns out that a potential answer has been in the
literature on existential have at least since 2004:
As a mediator between relations at the nominal and the
verbal level, have doesn't select normal arguments
but relational predicates. We consequently expect
the indefinite article to be a mere complication of the
derivation.
12
Why this isn't good enough
If the have we find with bare nominals is indeed a
mediator between relations at the nominal and the
verbal level...
... we would predict relational nouns to be the preferred
class of nouns to occur bare with have.
This prediction is not borne out.
> E.g. the nouns that combine (and combined) with
tener in Spanish don't exhibit a clear relational bias.
13
Our proposal
Our proposal
Landman's proposal
Existential have selects relations and consequently
mediates between relations at the nominal and the
verbal level.
Our proposal
Existential have is not a mediator between relations but
rather a relation builder.
Pzn(transitivize(P)(z)(n))
It selects a one-place predicate and returns a two-place
predicate with an existentially bound internal
argument.
15
Our proposal
Our proposal
Existential have is not a mediator between relations but
rather a relation builder.
> If this is the correct analysis we no longer expect to
find relational nouns to be the preferred class of
nouns to occur bare with have.
Sample derivations
John has a book.
Mary has a brother.
17
John has a book
John have
a
Pzn(transitivize(P)(z)(n)) PQy(P(y)&Q(y))
book
x(book(x))
Qy(book(y)&Q(y))
Pzn(transitivize(P)(z)(n)) Qy(book(y)&Q(y))
BE(Qy(book(y)&Q(y)))
x(book(x))
zn(transitivize(x(book(x)))(z)(n))
18
The transitivization step in John has a book
input: a set of individuals
ex. x(book(x))
output: the set of pairs of which the first member...
1) ... belongs to the input set
2) ... stands in a pragmatically inferred relation to the
second member
ex. wx(book(x) & belonging to (x,w))
19
John has a book
John have
john
a
Pzn(transitivize(P)(z)(n)) PQy(P(y)&Q(y))
book
x(book(x))
Qy(book(y)&Q(y))
Pzn(transitivize(P)(z)(n)) Qy(book(y)&Q(y))
BE(Qy(book(y)&Q(y)))
x(book(x))
zn(transitivize(x(book(x)))(z)(n))
zn(wx(book(x)&belonging to (x,w))(z)(n))
zn(book(n)&belonging to (n,z))
n(book(n)&belonging to (n,john))
20
Mary has a brother
Mary have
Pqn(transitivize(P)(q)(n))
a
brother
PQz(P(z)&Q(z)) xy(B of(y,x))
PQz(P(z)&Q(z)) xy(B of(y,x))
detrans(xy(B of(y,x)))
yx(B of(y,x))
Qz(x(B of(z,x))&Q(z))
Pqn(transitivize(P)(q)(n)) Qz(x(B of(z,x))&Q(z))
BE(Qz(x(B of(z,x))&Q(z)))
z(x(B of(z,x)))
qn(transitivize(z(x(B of(z,x)))(q)(n))
21
The transitivization step for Mary has a brother
input: a set of individuals
the set of individuals that stand in a brother relation to
someone
output: the set of pairs of which the first member...
1) ... belongs to the input set
2) ... stands in the brother relation to the second
member (= lexico-pragmatically inferred relation)
22
Mary has a brother
Mary have
mary
Pqn(transitivize(P)(q)(n))
a
brother
PQz(P(z)&Q(z)) xy(B of(y,x))
PQz(P(z)&Q(z)) xy(B of(y,x))
detrans(xy(B of(y,x)))
yx(B of(y,x)))
Qz(x(B of(z,x))&Q(z))
Pqn(transitivize(P)(q)(n)) Qz(x(B of(z,x))&Q(z))
BE(Qz(x(B of(z,x))&Q(z)))
z(x(B of(z,x)))
qn(transitivize(z(x(B of(z,x)))(q)(n))
qn(rs(B of(s,r)&x(B of(s,x)))(q)(n))
qn(B of(n,q)&x(B of(n,x)))
n(B of(n,mary)&x(B of(n,x)))
23
Recap
Our proposal
Relational HAVE is not a mediator between relations but
rather a relation builder.
Pzn(transitivize(P)(z)(n))
It selects a one-place predicate and returns a two-place
predicate with an existentially bound internal
argument.
Given that have requires an <e,t> expression as input,
the indefinite article is nothing more than a
complication of the derivation. We consequently
expect the indefinite article - at least semantically
speaking - to be redundant.
24
Comparing the proposals
A new piece of data
Mary has the only nice brother.
 Mary is the only one with a nice brother.
the only
quantificational (non-presuppositional)
the
cf. e.g. McNally (2008)
nice brother xy(nice_brother_of(y,x))
26
An analysis à la Landman
Step 1
A modifier semantics for the:
for non-relational predicates
the man
x (man(x)
&y(man(y)
y=x))
for relational predicates
the brother
zx(brother of (x,z)&y(brother of (y,z)y=x))
27
An analysis à la Landman
Mary have
Step 2
the only nice brother
mary Rpq(R(p)(q)) zx(nice_brother_of (x,z)&y(nice_brother_of (y,z)y=x))
pq(nice_brother_of (q,p)&y(nice_brother_of (y,p)y=q))
q(nice_brother_of (q,mary)&y(nice_brother_of (y,mary)y=q))
28
An analysis à la Landman
Mary have
Evaluation
the only nice brother
mary Rpq(R(p)(q)) zx(nice_brother_of (x,z)&y(nice_brother_of (y,z)y=x))
pq(nice_brother_of (q,p)&y(nice_brother_of (y,p)y=q))
q(nice_brother_of (q,mary)&y(nice_brother_of (y,mary)y=q))
> Mary has a nice brother and all nice brothers of
Mary are identical to this brother.
> This doesn't exclude that other people have nice
brothers too.
> This is a non-existing interpretation.
29
An analysis à la Landman
Mary have
Evaluation
the only nice brother
mary Rpq(R(p)(q)) zx(nice_brother_of (x,z)&y(nice_brother_of (y,z)y=x))
pq(nice_brother_of (q,p)&y(nice_brother_of (y,p)y=q))
q(nice_brother_of (q,mary)&y(nice_brother_of (y,mary)y=q))
> What crucially goes wrong is that Mary is specified
as the internal argument of brother.
> Given that this is the gist of Landman's whole
analysis, it would appear to be impossible to remedy.
> The same holds for an analysis à la Partee.
30
The derivation
Our analysis
M
have
the only
nice brother
mary Pqn(transitivize(P)(q)(n)) PQz(P(z)&v(P(v)z=v)&Q(z)) xy(n_b_of(y,x))
PQz(P(z)&v(P(v)z=v)&Q(z))
detrans(xy(n_b
_of(y,x)))
yx(n_b_of(y,x))
Qz(x(n_b_of(z,x))&v(x(n_b_of(v,x))z=v)&Q(z))
Pqn(transitivize(P)(q)(n)) Qz(x(n_b_of(z,x))&v(x(n_b_of(v,x))z=v)&Q(z))
BE(Qz(x(n_b_of(z,x))&v(x(n_b_of(v,x))z=v)&Q(z)))
z(x(n_b_of(z,x))&v(x(n_b_of(v,x))z=v))
qn(transitivize(z(x(n_b_of(z,x))&v(x(n_b_of(v,x))z=v)))(q)(n))
qn(x(n_b_of(n,x))&v(x(n_b_of(v,x))n=v)&R(n,q)))
n(x(n_b_of(n,x))&v(x(n_b_of(v,x))n=v)&R(n,mary)))
31
Our analysis
Evaluation
n(x(n_b_of(n,x))&v(x(n_b_of(v,x))n=v)&R(n,mary)))
> There's a single nice brother and it's Mary's.
> This is the required interpretation.
32
Our analysis
Evaluation
n(x(n_b_of(n,x))&v(x(n_b_of(v,x))n=v)&R(n,mary)))
> The reason this goes right is because we build a
new relation on top of the original one.
> Let's look at which assumptions are crucial to
ensure that this is the only option.
33
Our analysis
>
Especially
quantifying-in
has
important
applications in other parts of the literature. It's
consequently undesirable to assume that it would be
unavailable here.
Evaluation
Quantifying-in
would
eliminate the
need
to
detransitivize.
(the
Partee
problem)
Having
a
determiner
that
is
compatible
with relations
would
also
eliminate the
need
to
detransitivize.
(the Landman
problem)
34
Our analysis
Last step
> In order to maintain quantifying-in as a generally
available mechanism we propose to analyze relational
nouns - at least in English - not as expressions of type
<e,<e,t>> but as expressions of type <e,t>.
xy(brother of (y,x))
yx(brother of (y,x))
35
Implications (1)
> We have to reconsider the (few) restrictions that
have been ascribed to the distinction between
relational and non-relational nouns.
the mother of John
*the pen of John
A refinement however seemed necessary anyway on
the basis of (attested) examples like the following:
the church of the
monastery of Asteri
the café
harbour
of
the
old
36
Implications (2)
> We open up a new avenue for the analysis of
different kinds of relational nouns within languages
closely related to English:
Those nouns that are like mother in coming with an
existentially closed internal argument.
Those nouns that were analyzed as relational before
but could actually also get a purely non-relational
semantics. Hand would be an example (an object
consisting of a palm and five fingers).
This might help to lead to a better understanding of
the kinship/body-part split we find in inalienable
37
possession constructions in e.g. French.
Implications (3)
> We open up a new avenue for cross-linguistic
variation:
Languages like English in which relational nouns are
non-relational and are consequently not expected to
come with an obligatorily realized internal argument.
Languages like e.g. Dakaaka (von Prince 2012) in
which relational nouns are relational and are
consequently expected to come with an obligatorily
realized internal argument.
38
Implications (4)
> Assuming relational nouns to be non-relational
allows us to restore type-shifting within the relational
domain to normalcy, i.e. type-shifting would only apply
if required to avoid type-clashes.
John's mother
the mother that John is taking care
of (not necessarily his own)
If we assume relational nouns are non-relational, and
we assume John's builds or selects relations, the fact
that John's mother needn't refer to his actual mother
is expected. This is due to the fact that relations
would always have to be derived.
39
Implications (4)
> Assuming relational nouns to be non-relational
allows us to restore type-shifting within the relational
domain to normalcy, i.e. type-shifting would only apply
if required to avoid type-clashes.
John's mother
the mother that John is taking care
of (not necessarily his own)
If we however assume relational nouns are relational,
we have to make an (unmotivated) detransitivization
type-shift followed by a transitivization type-shift.
40
Summary and conclusion
Summary and conclusion
> We have proposed a new analysis of existential
have as a relation building verb
> We have shown how the addition of an
indefinite article is a mere complication of the
analysis and we consequently have found a
rationale behind the cross-linguistic tendency to
allow for bare nominals behind have.
> We have argued that relational nouns like
mother are best analyzed as expressions of type
<e,t> in which the internal argument of their
traditional <e,<e,t>> interpretation is existentially
closed.
42
Summary and conclusion
> The analysis we proposed for have can easily
be extended to (at least one of the senses of)
with and without:
hat with(out) feather
with(out)
PQy()x(Q(y)&P(x)&transitivize(y)(x))
with(out) feather
Qy()x(Q(y)&feather(x)&present-with(y)(x))
hat with(out) feather
y()x(Hat(y)&Feather(x)&Present-with(y)(x))
43
References
References
> Borthen, K. (2003). Norwegian bare singulars
(Doctoral dissertation, Norwegian University of
Science and Technology).
> Dobrovie-Sorin, C., Bleam, T., & Espinal, M. T.
(2006). Bare nouns, number and types of
incorporation, in Vogeleer, S., & Tasmowski, L.
(Eds.), Non-definiteness and plurality, Amsterdam:
John Benjamins, 51-79.
> Espinal, M. T., & McNally, L. (2011). Bare nominals
and incorporating verbs in Spanish and Catalan.
Journal of Linguistics, 47(01), 87-128.
> Landman, F. (2004). Indefinites and the Type of Sets.
Blackwell.
32
References
> Lazaridou-Chatzigoga, D. and Alexandropoulou, S.
(2013). A corpus study of Greek bare singulars:
implications
for
an
analysis.
Revista
da
Abralin/Associação Brasileira de Linguística, special
issue “Weak Definiteness and Referentiality”,
organized by Roberta PIRES DE OLIVEIRA, número
XII/1, janeiro a junho de 2013.
> McNally, L. (2008). DP-internal only, amount relatives,
and relatives out of existentials. Linguistic Inquiry,
39(1), 161-169.
> Partee, B. (1999). Weak NPs in have sentences. In
JFAK [a Liber Amicorum for Johan van Benthem on
the occasion of his 50th birthday], ed. J. Gerbrandy et
al.
32
References
> von Prince, K. (2012). Nominal possession in
Daakaka: Transitivizing vs. linking. In Proceedings of
AFLA (Vol. 18).
32
48