=Paper=
{{Paper
|id=None
|storemode=property
|title="Some", Speaker Knowledge, and Subkinds
|pdfUrl=https://ceur-ws.org/Vol-954/paper19.pdf
|volume=Vol-954
}}
=="Some", Speaker Knowledge, and Subkinds==
https://ceur-ws.org/Vol-954/paper19.pdf
Some, Speaker Knowledge, and Subkinds✯
Andrew Weir
University of Massachusetts Amherst
Abstract. I provide an analysis of sentences with some combining with
subkind-denoting NPs, such as Some plant is growing through the wall
of my room. In such sentences, there is epistemic uncertainty concerning
the subkind, but not concerning the actual witness of the claim. I make
use of the semantics proposed by [AOMB10b] for the Spanish epistemic
indefinite algún, combined with the polysemy of common nouns between
being individual-denoting and subkind-denoting [Kri95, Kra08], to pro-
vide an analysis of such sentences.
Keywords: some, epistemic indefinites, kinds, subkinds
1 Introduction
The English determiner some, when combined with a singular noun phrase, seems to
carry a meaning of speaker ignorance as to the witness of the existential claim being
made [Bec99, Far02, AOMB03].
(1) a. Some ball bearing in this pile is actually made of a different material
from all the others. But they all look identical, so I can’t tell which
one./#Namely, that one there that’s a different color from the others.
b. The hackers implanted a virus into some file on this computer. But there’s
no telling which file./#It’s the file I’m pointing out to you right now.
In these cases, we can see that the speaker of the sentence cannot know who or what
the referent of the some NP phrase is. Following such a phrase with an identifying
statement (such as a ‘namely, . . . ’ statement) is infelicitous.
However, there are some cases of the use of some, exemplified in (2), where the
meaning appears to be subtly different. The intuition about these cases is that the
speaker’s lack of knowledge in saying some P is not really connected to knowledge of
which P is being referred to. Rather, the speaker does not know what kind of P the
referent is.
(2) a. I saw some contraption in the copy room this morning.
b. I came home to find some plant growing through a hole in my wall.
c. Doctor, some growth appeared on my arm. Should I be worried?
These cases, on the reading of interest here, cannot be paraphrased as meaning (for
example) ‘I saw a contraption in the copy room and I don’t know which contraption it
✯
I would like to thank Angelika Kratzer and two anonymous ESSLLI reviewers for
extremely useful comments on this material. All errors are of course mine.
R.K. Rendsvig and S. Katrenko (Eds.): ESSLLI 2012 Student Session Proceedings, CEUR Work-
shop Proceedings vol.: see http://ceur-ws.org/, 2012, pp. 180–190.
Some, Speaker Knowledge, and Subkinds 181
was’. Rather, the meaning seems to be something more like ‘I saw a contraption in the
copy room and I don’t know what kind of contraption it was’; and similarly for the other
examples in (2). This is not just a matter of knowing what the name of the contraption
is; I will argue that some can be sensitive to a lack of knowledge of a name in the case
when it is in construction with a human-denoting NP, but not when it combines with
a thing- or subkind-denoting NP. I will postulate a denotation for some based on the
semantics proposed by [AOMB10b] for Spanish algún. The two readings, the ‘unknown
entity’ and the ‘unknown kind’ reading, come about via polysemy of the common noun.
Following [Kri95] and [Kra08], I will propose that a noun like contraption can represent
either a property of individual contraptions or a property of subkinds of contraption.
I will then argue that some can quantify over either of these; it also has built into it a
partitive semantics, accounting for the fact that some contraption is in the office will
have the truth conditions that some part of a subkind of contraption is in the office
(and not a subkind itself). Firstly, I discuss the type of epistemic uncertainty that
holds when some is used, before turning to how [AOMB10b]’s semantics for algún can
be used to model some.
2 Uncertainty about things and people
We know from examples like the following (from [AOMB03]) that some in English is
not generally incompatible with ostension. That is, a sentence like (3) is acceptable
even if you can ‘point to’ the professor in question.
(3) Look! Some professor is dancing lambada on his table! [AOMB03, (9)]
Some other epistemic uncertainty is targeted by some in (3). [AOMB03] suggest the
professor’s name, which is possible but not the only possibility, as (4) shows:
(4) Look! Some professor wearing a name badge saying ‘John Smith’ is dancing the
lambada!1
[AP10] discuss cross-linguistic variation in epistemic determiners of the some kind (for
example, Spanish algún, German irgendein, Italian un qualche, Romanian vreun). There
is cross-linguistic variation in whether examples like (3) are licensed; Spanish algún, for
example, is not licit in examples like (3). However, in cases like (3) in English, which
are licit, there appear to be a variety of identification modes which some could indicate
uncertainty about. Possibly some is sensitive to all the ways that there are of ‘knowing
who’ [BL86] or ‘knowing what’, argued to be contextually determined by [Alo01].
If this were the end of the story, the examples of the form I saw some contraption in
the copy room, which are the ones which motivate this paper, would not be interesting.
Some in such examples might just denote lack of knowledge of the name of the subkind
of contraption. However, I will argue that when some combines with NPs which de-
note things, rather than people, the subtleties discussed above disappear. With things,
‘differentiation’ – being able to distinguish the witness of the claim from other things
in the extension of the NP – is the only identification mode which is relevant for some.
Having established diagnostics to show this, I will go on to show that the ‘unknown
subkind’ reading patterns with ‘unknown things’ rather than ‘unknown people’.
1
We assume that name badges are perfectly reliable. Some is still licensed even in
that situation.
182 Andrew Weir
Prima facie, ‘picking out’ seems to be the identification mode for thing-denoting
NPs as opposed to human-denoting NPs, as the below examples show.2
(5) a. Some professor is dancing the lambada!
b. I saw some guy hanging about outside.
(6) a. ??Some statue is in the middle of the square. [looking at it]
b. ??There’s some letter in my mailbox. [looking at it]
Being able to ‘pick out’ the referent appears to delicense some with thing-denoting
NPs in (6). One could argue that this simply represents there being many more means
of ‘knowing’ applicable to humans than to things; so for example humans have names
while things generally don’t. However, even when things have names, some does not
seem to be able to target uncertainty about the name, as the below examples show.
(7) Two diplomats from Peru are delegates to a conference you are at. One is a
man and one a woman. You see them both several times, and know that they’re
both from Peru, but never catch their names.
a. At dinner, I was sat across from a/some delegate from Peru.
(8) You are lost. You know that the city you’re in has only two squares. You keep
coming across both squares. You can tell them apart because one has a fountain
and the other doesn’t, but you can’t see any street signs. You end up in the
fountainless square in the city. Your friend phones you:
a. A: Where are you?
B: I’m in a/?#some square in the city.
Some appears able to signify lack of knowledge of the name in (7), but not easily in
(8), even though city squares do usually have names. Given these data, I propose that
the epistemic condition which some is sensitive to, when in construction with a thing-
denoting NP, is uniformly that in (9). This will delicense some in situations like (8),
where the speaker can differentiate the city squares, even though he does not know
their names.3
(9) Differentiation condition on ‘some NPthing ’
A speaker uses ‘some NPthing ’ to signal that she could not, if presented with
the extension of NP, ‘pick out’ the witness of the existential claim.
2
The examples in (6) are marginal rather than fully unacceptable, for reasons I will
discuss in section 4.
3
This ‘differentiation’ condition is subtly different from ostension, being able to
‘point at’ the referent. Ostension can’t be exactly what is at play, because of the
felicity of examples like:
(i) [A has been drugged and kidnapped; he wakes up, looks around, and exclaims:]
I’m trapped in some subway station!
Here, the poor kidnappee can point at the subway station perfectly well; what he is
not in a position to do is distinguish the one he’s in from other things in the extension
of ‘subway station’.
Some, Speaker Knowledge, and Subkinds 183
I do not want to speculate here about precisely how the contrast between human-
denoting and thing-denoting NPs is to be modeled. An obvious way would be to posit
homophony (or polysemy) between two somes, one which selects a [+human] argument
and one which does not. That is an unattractively unparsimonious solution and hope-
fully further work can shed light on whether and how the two somes can be unified.
That the distinction should be grammatically encoded, however (rather than follow-
ing from some general constraint on epistemic relations towards people versus towards
things), seems to be supported by contrasts like that between (7) and (8). For present
purposes, it suffices to note that the contrast does exist.
3 Modeling the epistemic condition
How can we model in the semantics the ‘differentiation’ condition given in (9)? One way
is that developed by [AOMB10b] for the Spanish indefinite determiner algún. Algún
carries with it an implicature of speaker lack of knowledge, seemingly similar to the ‘un-
known entity’ readings of English some discussed above. For example, asking a speaker
to identify the referent of algún NP is an infelicitous conversational contribution, as
shown in (10) ([AOMB10b]’s (8)).
(10) a. Juan tiene que estar en alguna habitación de la casa.
Juan has to be in alguna room of the house
‘Juan must be in a room of the house.’
b. #¿En cuál?
in which
(‘In which one?’)
These contrast with examples where the determiner un is used. In such cases, there is
no epistemic uncertainty, as shown in (11) ([AOMB10b]’s (10)).
(11) a. Juan tiene que estar en una habitación de la casa.
Juan has to be in una room of the house
‘Juan must be in a room of the house.’
b. ¿En cuál?
in which
‘In which one?’
[AOMB10b] propose that algún is a standard existential quantifier, taking a restrictor
and a scope; as is un. However, both of these determiners also combine with a subset
selection function f . The purpose of these subset selection functions is to restrict the
domain of the quantifier to some subset, following proposals by [Sch02]. [AOMB10b]
propose that algún places a presuppositional restriction on the function in question: it is
an antisingleton function. Un does not do this. Below is a definition of an antisingleton
subset selector function, and [AOMB10b]’s definitions of un and algún ([AOMB10b]’s
(50, 54)).4
(12) Let f be a function which takes a set X and returns a subset of X.
f is an antisingleton function iff, for all X in the domain of f , |f (X)| > 1.
4
Following [AOMB10b], and [HK98], I place presuppositional restrictions on the
well-formedness of an expression between the colon and the period.
184 Andrew Weir
(13) a. JunK = λfhet,eti λPhe,ti λQhe,ti .∃x[f (P )(x) & Q(x)]
b. JalgúnK = λf λP λQ : antisingleton(f ).∃x[f (P )(x) & Q(x)]
[AOMB10b] argue that algún is in pragmatic competition with un. Un can in prin-
ciple allow for an exhaustivity inference to be drawn; a speaker hearing (11) could,
potentially, believe that the subset of rooms being quantified over is a singleton set.
Algún, with its extra presupposition, precludes this possibility; on hearing a sentence
like (10), the listener deduces that the speaker avoided using un specifically to avoid the
possibility that the listener could draw the inference that the speaker was restricting
the domain of rooms to a singleton set. Given this reasoning, it follows that algún is
used to indicate that the speaker is actively unable to restrict the domain of rooms to
a singleton set; that is, the speaker does not know which room Juan is in.
[AOMB10b, 16f.] argue that the ‘antisingleton’ presupposition is justified as algún
cannot combine with NPs which must denote singleton sets, as in ‘Juan bought un/#algún
book that was the most expensive in the bookstore.’ This seems also to be true of En-
glish examples such as Mary bought a/?#some ring that was the most expensive in the
jeweler’s. This ‘antisingleton’ presupposition can also, I argue, model the ‘differentia-
tion’ constraint on English some, at least when some pairs with thing-denoting NPs.
So, as a preliminary move, we can take over [AOMB10b]’s definition of algún to some:
(14) JsomeK = λfhet,eti λPhe,ti λQhe,ti : antisingleton(f ).∃x[f (P )(x) & Q(x)]
4 Kinds and subkinds
Having proposed this definition for some, I turn to the ‘unknown subkind’ reading of
some NP. As noted above, the initial contrast in e.g. (2b), repeated as (15) below, would
not be surprising if subkinds patterned with ‘people’ with respect to their combination
with some, rather than ‘things’.
(15) There’s some plant growing through the wall of my room.
In this section, I will argue that subkind-denoting NPs do, however, generally pattern
with ‘things’ in combination with some. The argument consists in showing that being
able to distinguish subkinds from one another is sufficient to de-license some, even if
other epistemic uncertainties (e.g. the name) remain. To show this, we turn to the
following examples.
(16) Able to distinguish subkinds
Katniss, having grown up on her wits, is intimately familiar with all the plants
in her district, and how they can be used for medicinal purposes. She’s never
had any formal schooling or parental teaching of herbal lore, though, so she
doesn’t know any of their names. She applies one to heal Gale’s burns.
Gale: What’s that?
Katniss: A/?#some plant that’s good at soothing burns.
(17) Unable to distinguish subkinds
Katniss is in the Hunger Games Arena, far from home, where there are new
types of plant that she’s never seen before. She discovers through experimenta-
tion that one type is good for healing burns. She applies it to heal Rue’s burns.
Rue: What’s that?
Katniss: A/some plant that’s good at soothing burns.
Some, Speaker Knowledge, and Subkinds 185
Here, the crucial point is that, in (16), Katniss is able to differentiate subkinds from each
other, and categorize the entities (the actual plants) into each subkind, as appropriate.
Even if she doesn’t know the subkinds’ names, some is still not licensed. However,
if Katniss cannot differentiate the subkinds with certainty, as in (17), some on the
subkind reading is again licensed.
I now turn to an analysis of common nouns which will allow us to derive the
‘unknown subkind’ reading. Nouns like plant appear to be polysemous between denoting
individual plants and subkinds of plants, as shown in (18):
(18) That plant {was watered yesterday/is widespread}.
I will summarize a means of achieving this polysemy compositionally, proposed by
[Kri95] and [Kra08]. We start from the assumption that a kind, for example the kind
PLANT, is no more than the mereological sum of all plants in the world. See e.g. [Chi98,
349]: ‘It seems natural to identify a kind in any given world √ (or situation) with the
totality of its instances.’5 [Kra08] argues that the noun root plant denotes this sum of
all plants, which I will notate with ΣPlant. This root is not pronounced alone, however.
The word plant which we actually pronounce includes a classifier, which in English is
silent. There are two classifiers in English: one combines with a kind and returns a
property of individuals which are parts of the kind; the other combines with a kind
and returns a property of subkinds of the kind. (19), (20) show how this works. ‘Π’ is
the part relation, as in [Lin83].
(19) (Kratzer’s (2), adapted)
√
a. J plantK = ΣPlant
b. JCLind K = λxλy.kind(x) & individual(y) & yΠx
(takes a kind and returns the property of being an individal member of
that kind)
c. JCLkind K = λxλy.kind(x) & kind(y) & yΠx
(takes a kind and returns the property of being a subkind of that kind)
In (20), we see how this system can provide a subkind
√ reading for a DP like that plant.
The NP formed by the combination of the root plant and the classifier CLkind is
pronounced as plant.
(20) a. DP
D NP
√
that CLkind plant
5
In fact, Chierchia argues that kinds are not entity-type, type e, but individual
concept type, type hs, ei – a function from situations to entities. Here, I work with a
fully extensional semantics, and so treat kinds simply as the entities which would, in
a fuller treatment, be the result of applying a Chierchia-type kind to the world (or
situation) of evaluation w0 .
186 Andrew Weir
√
b. Function Application on JCLkind K and J plantK:
λy.kind(ΣPlant) & kind(y) & yΠ(ΣPlant)
= λy.kind(y) & yΠ(ΣPlant)6
c. JthatK = λPhe,ti .ιx[P (x)]7
d. Function Application on (b) and (c): ιx[kind(x) & xΠ(ΣPlant)]
‘The contextually salient subkind of plant’
The word plant can either denote the property of being an individual plant, or the
property of being a subkind of plant. Given this, it is not surprising that some might
combine with plant and yield an ‘unknown kind’ reading, rather than lack of certainty
as to the witness of the existential claim. In a sentence like Some plant is growing
through my wall, we want some to make an existential claim about a plant, the witness
to which claim the speaker might well be able to distinguish from other members of
JplantK; and at the same time express speaker ignorance about the kind of plant. Below
is a revised proposal for the denotation of some, based on [AOMB10b]’s proposal for
algún, but with a crucial underlined addition.
(21) JsomeK = λfhet,eti λPhe,ti λQhe,ti : antisingleton(f ).∃x[(f (P ))(x)
& ∃y[yΠx & Q(y)]]
In this denotation, the quantifier’s scope, Q, is not applied to the x that the restrictor
P applies to. Rather, there is a subpart of x which Q applies to. So some P is a
Q means that there is a part of a P that is a Q. The work done by the underlined
addition is to allow some to felicitously combine with kind-type arguments, but to end
up quantifying, not over subkinds themselves, but over instantiations of subkinds (here
equated with parts of those subkinds). If we apply this some to a noun root combined
with CLkind , we get the reading that is at issue here, as shown below. For perspicuity, I
have indicated some places where I replace lambda-abstraction notation for properties
(understood extensionally as sets of entities) with set notation, as the lambda notation
for the set of subkinds in (23a) is not very transparent when embedded in a larger
expression such as (23b).
(22) DP
D’ NP
√
D f CLkind plant
some
(23) a. JNPK = λy.kind(y) & yΠ(ΣPlant)
= {ΣIvy, ΣCreeper, ΣRhododendron, . . .} (using set notation)
6
I abbreviate by removing from the truth conditions the restriction on ΣPlant that
it be a kind. I simply assume from now on that CLkind combines only with kinds,
without writing it explicitly into its denotation.
7
I am ignoring the deictic contribution of that here and simply identifying it with
the.
Some, Speaker Knowledge, and Subkinds 187
b. JDPK = λQ.∃x[(f (λy.kind(y) & yΠ(ΣPlant))(x)) & ∃z[zΠx & Q(z)]]
= λQ.∃x[x ∈ f ({ΣIvy, ΣCreeper, ΣRhododendron, . . .}) & ∃z[zΠx & z ∈
Q]] (using set notation)
Presupposition: antisingleton(f )
(24) Jsome plant is growing through my wallK =
∃x[(f (λy.kind(y) & yΠ(ΣPlant))(x)) & ∃z[zΠx & growsInWall(z)]]
‘There is some subkind of plants x, and there is something z that is a part
of that subkind of plants, and z is growing through my wall, and the speaker
wants to signal that the set of subkinds of plants to which x belongs is not a
singleton.’
The pragmatic effect of (24) is such that there is no signal that the speaker does not
know which plant is at issue. It is not the case that the speaker is signaling that she
cannot narrow the set of plants down to a singleton set. Rather, she is signaling that
she is not whittling the set of subkinds of plant ({ΣIvy, ΣCreeper, ΣRhododendron,
. . . }) down to a singleton. We use the same reasoning as used by [AOMB10b] to analyze
algún: on hearing some plant, the listener deduces that the speaker chose some (rather
than a, which has no anti-singleton presupposition) in order to signal that she actively
could not restrict the set of possible subkinds that the plant could fall into to a singleton
set.
This analysis also predicts the marginal (not fully ungrammatical) status of sen-
tences like ??There’s some statue in the town square (while looking at the statue). Such
sentences are good exactly to the extent that we can imagine uncertainty about the
subkind involved; that is, they are good to the extent that statue can mean kind of
statue (see [Car77] for discussion of which nouns can easily receive a kind reading).
5 Some with individuals
In cases like some file is infected, there does seem to be an epistemic effect concern-
ing individuals, very much parallel to Spanish algún. Do we need yet more somes in
English; one to combine with properties of kinds, and one combining with properties
of individuals? I will argue that this is not the case. There is only one some (at least
when it combines with thing-denoting NPs), with the semantics given above. Below, I
show the result of combining this some with a property of individuals, another possible
meaning for a common noun.
(25) a. DP
D’ NP
√
D f CLind file
some
b. JNPK = λy.individual(y) & yΠ(ΣFile)8
= {file1 , file2 , file3 , . . .} (in set notation)
8
Again, I abbreviate by omitting the statement that ΣFile is a kind.
188 Andrew Weir
c. JDPK = λQ∃x[(f (λy.individual(y) & yΠ(ΣFile)))(x) & ∃z[zΠx & Q(z)]
= λQ∃x[x ∈ f ({file1 , file2 , file3 , . . .}) & ∃z[zΠx & z ∈ Q]]
(in set notation)
Presupposition: antisingleton(f )
(26) Jsome file is infectedK = ∃x[(f (λy.individual(y) & yΠ(ΣFile)))(x)
& ∃z[zΠx & infected(z)]]
‘There is some individual x which is a mereological part of the file-kind (i.e. x is
a file), and there is a z which is a mereological part of x, and z is infected, and
the speaker wants to signal that the set of individual files to which x belongs is
not a singleton.’
If we say there is an individual x which is a file, and a z such that zΠx, then – because
x is an individual and so has no proper mereological parts – z must be an improper
mereological part of x, that is, z = x. The semantics then is precisely equivalent to the
semantics proposed by [AOMB10b] for algún. A sentence like some file is infected means
that there is an x which is a member of a subset of individual files, and x is infected,
and the speaker wishes to signal that the subset of individual files is not a singleton set;
that is, the speaker cannot specify which files is the witness to the existential claim.
We therefore derive both the ‘unknown individual’ and ‘unknown kind’ readings with
the same denotation for some.
6 Notes on plurality
Consider the semantics for a sentence like some contraption is in the office (on the
‘unknown kind’ reading).
(27) Jsome contraption is in the officeK =
∃x[(f (λy.kind(y) & yΠ(ΣContraption))(x)) & ∃z[zΠx & inTheOffice(z)]]
Let us say that one choice of contraption is ‘hole punch’. Then we could have, for
example, the situation where ∃z[zΠ(ΣHolePunch) & inTheOffice(z)] is true as one
verifying instance of (27). We do not actually have any requirement that z in (27) be
atomic, despite the singular morphology on the noun contraption. On the basis of the
following examples, I suggest that in fact the ‘unknown-kind’ reading of some is indeed
number-neutral.
(28) A: What’s this warehouse for?
B: There’s some contraption in there. There are shelves upon shelves of the
things, all the same. I don’t know what they are, though.
(29) [I take a delivery of 100 plants, but they are not the type I ordered; they are
all the same type of plant, but I don’t recognize what type.]
A: Did you get the plants you ordered?
B: They did deliver some plant. I have 100 of the things clogging up the office.
But I’ve no idea what they are, they’re not what I ordered.
Some can range over pluralities; the restriction on cases like (28), is that all the things
in the plurality belong to the same subkind. If this is not the case, then the examples
above become sharply bad.
Some, Speaker Knowledge, and Subkinds 189
(30) A: What’s this warehouse for?
B: #There’s some contraption in there. Three shelves for three different things,
but I don’t know what any of them are.
(31) [I take a delivery of 100 plants, not the type I ordered, and not all the same
type of plant; I don’t recognize any of the types of plants.]
A: Did you get the plants you ordered?
B: #They did deliver some plant.
These examples suggest that the semantics for some proposed here is on the right track;
on the ‘unknown kind’ reading, some NP is number-neutral with respect to the entities
quantified over, but makes reference to one specific subkind of the denotation of the
NP. Note that some file is infected, the individual reading, is not number-neutral with
respect to entities in the same way:
(32) Some file is infected. I don’t know which one/#which ones.
The (unknown) referent of some file cannot be plural. However, this is accounted for by
the
√ semantics of the individual classifier CLind , which, when combined with a root like
file, returns a set of individual files. Some file is infected asserts that there is (some
part of) some member of that set which is infected. We thereby achieve the result that
some file is infected only makes reference to individuals, in contrast with the ‘unknown
kind’ reading of a phrase like some plant.
7 Conclusion
I have argued that the ‘unknown individual’ meaning of some file is infected and the
‘unknown subkind’ reading of some contraption is in the office can be accounted for by
a unitary analysis of some, whose denotation includes a partitive semantics allowing it
to combine with properties of kinds. The ambiguity is not a property of some itself, but
rather due to the polysemy of noun phrases like plant, contraption between subkind-
type readings and individual-type readings, following [Kri95] and [Kra08].
Various questions remain. For example, does some with plural NPs (as in some
files are infected, namely these ones), which has no epistemic effect, admit of the same
analysis as some with singular NPs? [AOMB10a] propose an analysis of Spanish algunos
(the plural form of algún), which also does not have an epistemic effect, where the
antisingleton constraint is retained in the denotation but the epistemic effect is not
present.9 Whether the account can be transplanted to English is a question I leave for
future work.
Furthermore, the question raised in section 2 concerning the source of the differ-
ence between some when paired with human-denoting NPs and thing-denoting NPs
remains open. This ‘split’ does not seem to be cross-linguistic – for example, algún is
9
[AOMB10a] assume a number-neutral semantics for plural, in which case the
set {John, Mary, John+Mary} can be the subset of ‘students’ picked out by ‘algunos
students’. Crucially, while this set contains only one plural individual (John+Mary),
the set is not a singleton, and so is licit as the domain of algunos. It is possible, then,
for the speaker to not narrow down the set of students to a singleton set and yet have
only one witness (here, John and Mary) in mind. Algunos is then predicted not to have
an epistemic effect. See [AOMB10a] for the full details.
190 Andrew Weir
not felicitous in [AOMB03]’s ‘lambada professor’ example. That these splits are not
reproducible across languages seems to indicate that the nature of the constraint is
grammatical, rather than a ‘deeper’ (mental/cognitive) constraint on epistemic rela-
tions. Cross-linguistic work will be crucial here (see [AP10] for an overview). I leave
these as open questions, however.
References
Alo01. Maria Aloni. Quantification under conceptual covers. PhD thesis, Univer-
sity of Amsterdam, 2001.
AOMB03. Luis Alonso-Ovalle and Paula Menéndez-Benito. Some epistemic indefi-
nites. In Makoto Kadowaki and Shigeto Kawahara, editors, Proceedings
of NELS 34, pages 1–12. Amherst, MA: GLSA, 2003.
AOMB10a. Luis Alonso-Ovalle and Paula Menéndez-Benito. Domain restriction,
modal implicatures and plurality: Spanish algunos. Journal of Semantics,
28(2):211–40, 2010.
AOMB10b. Luis Alonso-Ovalle and Paula Menéndez-Benito. Modal indefinites. Nat-
ural Language Semantics, 18:1–31, 2010.
AP10. Maria Aloni and Angelika Port. Epistemic indefinites cross-
linguistically. Presented at NELS 41. http://staff.science.uva.nl/
~maloni/NELS2010-handout.pdf, 2010.
Bec99. Misha Becker. The some indefinites. In Gianluca Storto, editor, Syntax at
Sunset 2, number 3 in UCLA Working Papers in Linguistics, pages 1–13.
1999.
BL86. Steven E. Boër and William G. Lycan. Knowing who. Cambridge, MA:
MIT Press, 1986.
Car77. Gregory N. Carlson. Reference to kinds in English. PhD thesis, University
of Massachusetts Amherst, 1977.
Chi98. Gennaro Chierchia. Reference to kinds across languages. Natural Language
Semantics, 6:339–405, 1998.
Far02. Donka Farkas. Varieties of indefinites. In Brendan Jackson, editor, Pro-
ceedings of SALT 12, pages 59–83. Ithaca, NY: CLC Publications, Cornell
University, 2002.
HK98. Irene Heim and Angelika Kratzer. Semantics in generative grammar.
Malden: Blackwell, 1998.
Kra08. Angelika Kratzer. On the plurality of verbs. In Johannes Dölling, Tatjana
Heyde-Zybatow, and Martin Schäfer, editors, Event structures in linguistic
form and interpretation, pages 269–300. Berlin: Walter de Gruyter, 2008.
Kri95. Manfred Krifka. Common nouns: a contrastive analysis of Chinese and
English. In Gregory N. Carlson and Francis J. Pelletier, editors, The
generic book, pages 398–411. Chicago: University of Chicago Press, 1995.
Lin83. Godehard Link. The logical analysis of plurals and mass terms: A lattice
theoretic approach. In Rainer Bäuerle, Christoph Schwarze, and Arnim
von Stechow, editors, Meaning, use and the interpretation of language.
Berlin: de Gruyter, 1983.
Sch02. Roger Schwarzschild. Singleton indefinites. Journal of Semantics,
19(3):289–314, 2002.