From long weak definites to short ones

1
Workshop on "Weak definite(nes)s"
17 février 2014
Université Paris 8
From long weak definites to short ones
Francis Corblin
Université Paris Sorbonne
IJN & IUF
The category « weak definites» was first introduced in the literature by Poesio (1994) for
complete definite descriptions of the form : the x of y .
Crucial example: the corner of a busy intersection
The term was used many years later by Carlson & Sussman (2005) for short definite NPs of
the form the x in restricted lexical environments.
Crucial example : to take the escalator.
In both cases definite NPs are said « weak » because they seem to lack (some ? all the ?)
attributes of well behaved definites.
The main issue I shall discuss is: are these two kinds of definite NPs related ; is it possible to
analyze them as variants of the same phenomenon ? Are they standard manifestation of
definiteness or are they indefinites in disguise?
General idea:
The so-called “weak definites” are not so special and they are a core manifestation of
definiteness.
Similarities and differences between some definites:
• Complete definite NP with a genitive: La montre de Marie (the watch of Marie).
• Long weak definites of the Poesio type : Le pied de la table (the leg of the table)
•“Associative” or “Bridging” definites : Dans cette pièce la fenêtre est petite. (in this
room the window is small).
• Short weak definites : Prendre l’ascenceur, aller à l’école (to take the escalator, to
go to the-school)
I propose to see short weak definites as a bound vrsion of bridging definites. This view
is developed in my RLV paper Corblin (forthcoming) and I will only sketch briefly this
approach.
I will discuss more at length some topics not addressed in the paper in connection to what I
call “discourse uniqueness” or “weak identification:
To be brief, the “weak identification” problem is the following:
Speakers accept sentences like (1):
(1) John was fixing the leg of the table.
Although they know that a table has four legs.
So we might think that they do not bother about uniqueness in some cases…
There is no doubt that (1) makes accessible a unique discourse referent, since (2) is
acceptable.
(2) John was fixing the leg of the table. It had been broken during the move.
But speakers do not accept very well the succession (3):
(3) John was fixing the leg of the table. Mary was fixing # another one (/leg of the
table)
2
This might suggest that those definites satisfy some kind of uniqueness, after all, since no
other individual satisfying the same descriptive content is accessible immediately after their
occurrence. This constrasts sharply with indefinite NPs as antecedents, which are fine:
(3’) John was fixing a leg of the table. Mary was fixing another one (/leg of the table)
I will try to show that a better analysis of this problem might solve some objections raised by
Bourmayan (2013) against my analysis of weak definites as bound relational definites.
Crucial to my view is the fact that short a weak definite introduces a specific discourse
referent:
(4) J’ai pris le train. Il est tombé en panne après 1km à Clichy-Levallois.
I took the train. It had a break-down after 1km at Clichy-Levallois
Bourmayan argues that this analysis cannot predict why (5) implies (6) :
(5) I did not take the train.
(6) I did not take any train.
I will show that “discourse uniqueness” is also a property of short weak definites:
(7) J’ai pris le train et Marie # un autre.
I took the train, and Marie another one
(8) Mon frère est allé à la banque et ma soeur dans # une autre.
My brother went to the bank and my sister to another one
This is another argument for considering that the two varieties are related, but also something
to keep in mind when discussing Bourmayan’s objection.
1. On long weak definites. The corner of a busy intersection
What is salient in these examples (see a.o. Poesio 1994, Barker, 1992, 2005, Corblin 2001):
There is no uniqueness
The head noun is relational
The preposition « of » plays a crucial role.
Claim : of denotes a free variable on relations.
Vs other preposition (e.g. near) introducing a specified relation.
The free variables analysis is needed for explaining standard cases like (9):
(9) Marie was sewing a dress. The dress of Marie was red.
Preferred reading = the dress Marie was sewing was red.
dress of Marie’ = λx (dress x & R (x, m))
The definite article in the most common view, presupposes that its linguistic content is
functional (denotes a single entity). The presupposition is thus that the free variable R can be
contextually specified in order to satisfy this demand.
As for first order pronouns, in van der Sandt’s terms , this can be done by:
a) binding : find in the close context a relation Ri such that: : λx (dress x & Ri (x, m))
denotes a single entity.
b) accommodation : there is a regular way to reduce the set : λx (dress x & R (x, m)) to a
unique individual, by accommodating a more specific Ri .
For “dress” for instance, our lexical knowledge and our knowledge of the situation referred to
in the sentence may contain propositions like:
If a dress is wore by someone at a given t, there is one and only one such dress ;
The notion of “Meaning enrichement” used by some authors (Aguilar & Zwarts 2019, Aguilar
& Schulpen 2011, Roberts 2003) for the analysis of weak definites applies perfectly in this
3
sandard case: it is the accommodation/binding of a functional specification for the R relation.
Note that this specification achieving uniqueness is triggered by the definite determiner, but
not for other determiners. For instance une robe de Marie is not interpreted as “a dress wore
by Marie at t “ but say, as a dress she owns, a dress she sometimes wear, she designed, etc…
The weak definites collected by Poesio as paradigmatic cases have the following special
properties:
- The head Noun imposes a specification of R (e.g: “part of”); neither binding nor
accommodation is licensed (in general).
- There are more than one such objects: leg of the table, side of the boat
I tend to see as desperate any attempt for maintaining that in those cases a strict uniqueness
presupposition is preserved. See Barker (2005) for a balanced discussion.
I cannot see how in the leg of the table, it is presupposed in some sense that there is (in some
world or in some mind) a unique leg of the table.
It looks more realistic to admit that there are some exceptions to the uniqueness requirement.
In predicative uses, definite descriptions do not require uniqueness. This is why (10) is fine:
(10) I am the son of the king, too.
This view has opponents, for instance Coppock and Beaver (2012) :
(10) a. I don’t know whether iguanas have hearts, but is that the heart?
(10) b. #I don’t know whether iguanas have bones, but is that the bone?
But if one replaces “the bone” by “the foreleg”, the sentence would be acceptable
(10) c. I do not know whether iguanas have forelegs, but is that the foreleg ?
If the number of the considered entities is small, predicative uses are easily acceptable and the
long weak definties of Poesio are just cases in which referential uses are acceptable as well.
But just saying that the uniqueness requirement can be relaxed is not enough. Consider (11):
(11) Marie épousa le fils du roi.
Marie married the son of the king.
The presupposition that the king has only one son is not very strong. But this does not predict
that (12) is not acceptable:
(12) ?Marie épousa le fils du roi en janvier et Jeanne aussi en février..
Marie married the son of the king in Janurary, and so did Jeanne in february
Any consulted speaker says it should be worded as:
(13) Marie épousa un (des) fils du roi et Jeanne aussi.
Marie married a son of the king, and so did Jeanne
Roughly: I can accept “the son of the king” without knowing how many sons he has, but I
cannot accept any reference to another son of the king in the close discourse context. See the
oddness of (14)-(16):
(14) Marie a épousé le fils du roi, et Jeanne en a épousé ? un autre.
Marie married the son of the king and Jeanne-of-married another (one)
(15) Le fils du roi m’aime, mais ? un autre me déteste.
The son of the king loves me, but another (one) hates me.
(16) Marie a réparé le pied de la table, # et Jean les autres.
Marie fixed the leg of the table and Jean the other (ones)
So there is some sort of uniqueness, after all, since the individual referred to cannot be
contrasted to other individuals accepting the same description in the close discourse context.
4
It is accessible for a pronoun, but not for other anaphora. We might call it “discourse
uniqueness”. This reference does not make accessible the “complement set” (if not empty) as
an indefinite reference does.
Because there are specific problems with the interpretation of the bare autre after a noun
followed by a complement (does un autre means “another N” or “another N of a”), the
crucial test for discourse uniqueness is defined as follows:
Discourse uniqueness is revealed by the fact that the succession le X… un autre X is illformed.
As the famous example of McCawley “The dog met another dog” reminds, this is not a
property of all definites.
To sum up :
Definites of the form « The N of x » in general:
- introduces a free variable on relations, R. (« of »)
- interpreted either by binding or accommodation.
- binding or accommodation enriches the relation up to a function
This general mechanism cannot explain why some definites of this form are fine although no
plausible meaning enrichment can turn the relation to a function (the leg of the table).
For these definites, a weak version of uniqueness is assumed labelled weak identification.
The test for weak identification is that a pronoun is licensed, but not other anaphora.
Not perfectly clear which property is responsible for a long definite to be classified as
“weak”.
• A. the litteral semantic content is not enough for achieving uniqueness
La robe de Marie
is weak
- Binding /accommodation is required for achieving uniqueness.
Weakly weak
• B. the litteral semantic content cannot achieve uniqueness
R is specified by the relational noun
The corner of an intersection
- only a weak form of uniqueness holds (discourse uniqueness).
Strongly weak
2. Short weak definites
Example:
to take the elevator, to go to the bank.
Some shared properties with long weak definites:
- no familiarity; used out of the blue.
- introduce a specific DR. See (4) above.
- discourse uniqueness.
(17) Elle a pris le train, # et moi un autre
She took the train, and me another (one)
# J’en ai pris un autre
I took another (one)
A missing link between the two varieties is what has been called “associative” or “bridging”
definites.
(18) [In a building]
Is the elevator on the right?
5
(19)
[In town]
Is the post-office far from here?
Not even very easy in some contexts to decide between a bridging case and a weak definite
occurrence, as in (20) and (21):
(20) Le docteur m’a dit d’arrêter de fumer.
The doctor asked me to stop smoking
(21) [an accident just happened] Appelez le docteur!
Call the doctor.
If one sees bridging definites and weak definites as short versions of weak long definites,
some common properties follow:
- no familiarity
- introduces a specific DR.
- implies discourse uniqueness
(22) Le facteur est passé, # puis un autre.
The postman passed, then another (one)
(23) Dans l’immeuble, l’ascenceur est tombé en panne # puis un autre
In the building, the elevator has a break-down, then another (one)
Some specific properties of short relational definites (bridging and weak) can also be related
to the missing genitive:
a) no “of” implies that a relational interpretation has to be triggered by the head noun itself.
(24) Marie va à l’immeuble.
No WD reading
Marie goes to the building
(25)
Marie va au supermarché
Marie goes to the supermarket
WD reading licensed
b) No argument of the relation is given in situ (by the genitive); it must be picked up in
context.
This derives two salient properties of short relational definites.
- they rely heavily on the lexical properties of their head.
- they are context dependent.
There are also differences between bridging definites and short weak definites:
Bridging definites impose constraints on their discourse context:
(26)
(27)
Ils entrèrent dans la banque.# Le garcon était souriant.
They entered the bank. The waiter was smiling
Ils entrèrent dans le restaurant. Le garcon était souriant.
They entered the restaurant. The waiter was smiling
No bridging def. reading
Short weak definites impose constraints on their clausal context:
(28)
(29)
Marie regarde le train.
Marie prend le train
No weak def. reading
6
Claim: Short weak definites are interpreted as relational definites picking up arguments in
context. The lexicon and the context play a crucial role for licencing the relational
interpretation, and for providing its arguments.
2.1 Free Vs bound short weak definites
The basis of the analysis is that definiteness can license a relational interpretation of short
definites (deprived of an overt genitive).
postman (x) : postman (x) & R (x,y,z…)
As for long definites, the use of the definite determiner takes for granted that by
accommodation this relation R can be turned up to a function. This process is driven by
meaning postulates: for a given place and a given t normally there is (if there is any) a single
person in charge of delivering mail.
Associative (or bridging) definites: On the model of free anaphora, the discourse context can
provide the required arguments. There is no strong constraint on the clausal context; it is only
required that it keeps the relational interpretation in focus. This is why (30) is not a good
candidate for a bridging reading and is best interpreted as denoting a familiar entity.
(30) Le facteur sortit de sa douche.
The postman get out from his shower
As for weak definites of the Carlson & Sussman type, they can be conceived on the model of
bound anaphora. The functional restriction of the relation must be found within the
boundaries of the clause itself. There are no constraint on the discourse context, but strong
constraints on the clausal context.
Consider the interpretation of (31):
(31) Pierre va à l’école
Pierre goes to the-school
Definiteness : there is one and only one school w.r.t. Pierre, and t he time reference of the
event.
When is it so in general? If one is a pupil, then there is a unique school at which she is at t.
Weak definites as “bound” relational definites.
Relational definite?
School (x) & R (x, y, z,…)
Accommodate R to Ri such that there is a unique school w.r.t. x,y, z,…
Bound?
The argument-variables must be found within the clause:
School (x) & Ri (x, Pierre, t)
The “clausal” constraint on weak definites just come from the fact that the clause must
contain material keeping in focus the relational meaning of the head noun. This is why
“regarder le train” does not trigger a weak reading, but “prendre le train”, “rater le train”can.
The clause must also contain plausible arguments for the relation: this is why “Mon chien est
à l’école” denotes a mere location, but “mon fils est à l’école” triggers a weak definite
reading.
I will leave most details aside in this presentation and focus on some potential or actual
objections one can raise against the analysis of short weak definites as bound relational
definites.
7
2.2 Some arguments against the analysis
1. One cannot predict which N and which clause will licence a bound definite reading.
True to a large extent but not a strong objection:
- the mechanism is conceived as “constructional”: only some Ns and some clausal contexts
will trigger it. And it needs strong uniqueness meaning postulates.
- It is in competition : with other interpretations of short definites (familiarity, bridging), and
with long weak definites (Marie consulted the doctor/ Marie consulted her doctor)
-It can be in competition with other ways of achieving the same meaning. In English to go to
the school does not licence a weak reading of the school, probably because of the existence of
to go to school. Cf. The French être en prison (= to be in jail) and the impossibility to get a
weak reading for to go to the prison. Asic & Corblin (2012) discusses similar issues in
Serbian
2. It is not always true that a unique individual is denoted.
See the discussion of Schwarz (2009, 2012) in Aguilar & zwarts (2010).
It is true, but it not a prediction of my proposal.
The proposal is that a unique individual is introduced for a given value assignment to the
argument variables, especially the time variable:
(31) Marie a pris le train pour aller de Paris à Dieppe
Marie took the train for going from Paris to Dieppe
Does not imply a unique train. It implies only that at any time of her travel to Dieppe, there
was a unique train she used.
And the same phenomenon holds for bridging definites:
(32) Dans ce bar, le garcon est très occupé.
In this bar, the waiter is busy.
That there is a unique waiter for each t does not imply that there is a unique waiter “in all”.
3. This proposal does not predict that: Je n’ai pas pris le train implies je n’ai pris aucun train.
Objection of Bourmayan (2013).
I do not think this is a prediction of the theory, and I am not sure it should be.
The spirit of Bourmayan’s objection is as follows:
Consider that the bank is interpreted as : “the unique bank which is the most relevant
at t for dealing with a’s money issues.” Rather close to my view.
Then if a does not go to the bank,is true, it should leave open the possibility that a goes to
another one (less relevant).
If a did not go to the bank entails that a did not go to any bank, then the theory makes the
wrong prediction. In that case, a theory considering a weak definite as an indefinite would
fare better.
There are cases establishing that not going to the bank does not entail not going to any bank.
(33) Je ne suis pas allé à la banque. Je suis allé à une banque près du bureau, mais elle était
fermée.
I did not go to the bank. I went to a bank near the office, but it was closed.
Examples of this kind would be difficult to accommodate in case “not V the N” entails “not V
any N”.
But it is fair to say that they may look artificial rather difficult to analyse in details and would
require further investigations.
8
A known source of difficulty is the interpretation of definites in the syntactic scope of
negation.
A specific problem regarding contrast for weak definites is discourse uniqueness or “weak
identification”.
Discourse uniqueness means that the succession le X… un autre X, un autre [X] is ill-formed.
It is verified by short weak definites:
(34) J’ai pris le train et # pas un autre
I took the train and not another (one).
Compare to :
(35) J’ai pris le train de Rouen et pas un autre.
I took the train of Rouen and not another one.
(36) I took this train and not another (one).
(37) I took a train and not another (one).
And the same holds for negation:
(38) Je n’ai pas pris le train,# j’en ai pris un autre.
I did not take the train. I took another (one)
This makes very difficult to get examples like (33) above.
As to the fact itself that short definites do not license other anaphora, it does not establish that
the train do not refer to a specific entity. The case of the leg of the table shows that a specific
entity can be referred to by a definite NP which does not license other anaphora.
It is instead, a confirmation that weak and long definites share the weak identification
property.
3. Concluding remarks and questions about weak identification
There is a famous example of McCawley supporting the view that some definite references do
accept other anaphora:
(39) The dog met another dog.
If we are right to observe that this option is not open for short weak definites what could
explain this difference?
There are many interesting puzzles. For instance:
(40 Ils allèrent à la banque, # puis à une autre.
(41) Ils allèrent à la banque. Elle était fermée. Ils allèrent à une autre.
What (41) indicates is that if there is no additional property of the entity referred to by the
definite NP except the ones grounding uniqueness, other anaphora is not licensed. This is
possibly what “weak identification” means: we have no mean to distinguish the bank they are
going to at t from any other bank, except that this bank is the only one they are going to for
their money issues.
It is typical that McCawley’s example is understood, by contrast as a familiarity definite : the
dog is a dog we know something about, so we know a property at least, distinguishing this
dog from others which is independent of the dog satisfying the meeting event.
A full explanation would require, of course, to know more also about the semantics of other
anaphora.
9
References
Aguilar-Guevara, A. & Schulpen., M. 2011. Understanding the meaning enrichment of
weak definites. Proceedings of the 2011 ESSLLI Student Session.
Aguilar-Guevarra, A. & Zwarts, J. 2010, Weak definites and references to kinds ,
Proceedings of SALT 20: 179–196.
Asic, T. & Corblin, F. 2012. Telic definites and their preposition(s): French Vs Serbian
Workshop Languages with and without articles 2012, Paris.
Barker, C. 1992. Definite Possessives and Discourse Novelty, Chicago Linguistic Society
Proceedings , p. 26-41.
Barker, Chris. 2005. 'Possessive weak definites'. In J.-Y. Kim, Y. Lander & B. Partee (eds.).
Possessives and beyond: semantics & syntax: 89-113.
Birner, Betty and Gregory Ward. 1994. Uniqueness, Familiarity, and the Definite Article in
English. Berkeley Linguistics Society 20. 93-102.
Carlson, Gregory & Rachel Sussman. 2005. « Seemingly Indefinite Definites », Linguistic
Evidence, S. Kepsar and M. Reis, eds. Berlin: de Gruyter.
Clark, H. H. 1975. Bridging. In R. C. Schank & B. L. Nash-Webber (Eds.), Theoretical issues
in natural language processing . New York: Association for Computing Machinery
Corblin, F. 1987. Indéfini, défini et démonstratif. Construction linguistique de la référence.
Paris/Genève : Droz.
Corblin, F. 2001. Défini et génitif: le cas des définis défectifs, Cahier Jean-Claude Milner,
Jean-Marie Marandin, ed., Editions Verdier, pp.19-54.
Corblin, F. 2011. Des définis para-intensionnels : être à l'hôpital, aller à l'école, Langue
française, nº 171 (3/2011) Détermination et prédication .p. 55-77.
Corblin, F. (Forthcoming) Weak definites as bound relational definites, RLV.
Epstein, R. 1999. Roles and non-unique definite , Berkeley Linguistics Society (BLS 20).
Berkeley (CA): 122-133.
Flaux, N. 1992. Les syntagmes nominaux du type le fils d'un paysan : référence définie ou
indéfinie 1 , le Français moderne 1 :113-140.
Flaux, N. 1993. Les syntagmes nominaux du type le fils d'un paysan : référence définie ou
indéfinie 1 , le Français moderne 2 :23-45.
Hawkins, John A. 1978. Definiteness and Indefiniteness. A Study in Reference and
Grammaticality Prediction. Londres: Croom Helm.
Kleiber, G., 1993. Anaphore associative, pontage et stéréotypie, , Lingvisticae 1nvestigationes
17: 38-82.
Löbner, Sebastian. 1985. Definites, Journal of Semantics 4 (4). 279-326.
Milner J.-C. 1982 Ordres et raisons de langue. ,Paris, Seuil.
Poesio, Massimo. 1994. Weak definites. In Proceedings of Salt IV, eds. M. Harvey and L.
Santelmann. Cornell: DMLL. 282-299.
Pustejovsky, James. 1995. The Generative Lexicon, Cambridge (MA): The M.I.T. Press.
Roberts, C. 2003. Uniqueness in definite noun phrases. Linguistics and Philosophy 26, 287350.
Schwarz, F. 2012. How weak and how definite are Weak Definites? Semantic Archives.
Stvan, Laurel Smith. 1998. The semantics and pragmatics of bare singular noun phrases.
Ph.D. dissertation, Northwestern University. Evanston, Illinois.
Zribi-Hertz, A. and Loic, J.-L. 2014. From Noun to Name : On definiteness marking in
Modern Martinikè, in Patricia Cabredo Hofherr & Anne Zribi­‐Hertz (eds.), Crosslinguistic
Studies on Noun Phrase Structure and Reference Leiden/Boston: Brill, pp. 268-315.