Philosophical Issues, 16, Philosophy of Language, 2006

DESCRIPTIONS WITH ADVERBS OF QUANTIFICATION

Delia Graff Fara

Princeton University

In “Descriptions as Predicates” (Fara 2001) I argued that definite and

indefinite descriptions should be given a uniform semantic treatment as
predicates rather than as quantifier phrases. The aim of the current paper
is to clarify and elaborate one of the arguments for the descriptions-as-
predicates view, one that concerns the interaction of descriptions with adverbs
of quantification.

I. First Preface: The Semantics of Descriptions as Predicates

I’ll take it that the semantic value of a predicate is a function of type

!e, t", that is, a function from entities to truth values. I’ll take the extension of
such a function to be the set of entities the function assigns the value true to,
and the extension of a predicate (denoted by putting double brackets around
the predicate) to be the extension of the function that is that predicate’s
semantic value. Assuming that there are only two truth values, and that the
semantic value of a predicate is defined for every entity, we can indicate what
semantic value a predicate has by saying what the extension of that predicate
is. Indefinite descriptions constitute the simplest case:

(1) [[‘a G’]] = [[G ]].1

On the descriptions-as-predicates view, the extension of an indefinite de-

scription is just the extension of the predicative complement of the indefinite
article.
If we follow Russell, as I will, in taking definite descriptions to have a
built-in uniqueness condition that is “proffered rather than presupposed” (to
use Craige Roberts’s nice phrase), then we could let the extension of a definite
66 Delia Graff Fara

description ‘the F’ be the extension of F, if that is a singleton set, and the

empty set otherwise. But in order to accommodate the use of the definite
article with mass nouns and plural count nouns, which typically do not have
singleton extensions, I prefer to adapt Richard Sharvy’s generalization of
Russell’s uniqueness condition to the descriptions-as-predicates proposal, as
follows:
(2) [[‘the F’]] = {x ∈ [[F ]]: ∀y ∈ [[F ]] (y ≤ x)}.

We may read this as saying that the extension of a definite description ‘the
F’ is the singleton set consisting of the maximal member of the extension of
F, if there is one, and is the empty set otherwise. The maximal member of
the extension of F is that member of the extension of F that has every other
member as a “part.”
The use of the word “part” here may be somewhat misleading, but I’ll
continue to refer to the relation expressed by ‘≤’ as a parthood relation, with
the understanding that a different relation might be involved for different
choices of F, and that calling this relation “parthood” might not be quite
accurate for some of these choices.
When F is a plural count noun such as women, the relation can be under-
stood as the relation of being a sub-plurality of . A singular count noun such
as ‘woman’ will have in its extension each individual woman, none of whom
will be, in any relevant sense, a “part” of any other; while its pluralization
‘women’ will, on Sharvy’s proposal, have in its extension every plurality of
women, each of which is a “part” of (in the sense of being a sub-plurality
of) the plurality of all the women there are.2 So on the predicate version
of Sharvy’s proposal, the extension of ‘the woman’ will either be empty or
contain exactly one woman if there are no others; while the extension of ‘the
women’ will either be empty or contain the one (maximal) plurality consisting
of all the women there are if there are any at all. If we further take numeral
words to be predicate-modifiers,3 so that the extension of ‘two women’, for
example, will be that subset of the extension of ‘women’ that contains every
two-membered plurality of women, then the semantic proposals above for
plurals, numeral words and the definite article yield the following desirable
results, assuming that there are just three women, Ana, Bea and Carlita:4

(3) [[‘woman’]] = {A, B, C},

(4) [[‘the woman’]] = ∅,
(5) [[‘women’]] = {A + B, A + C, B + C, A + B + C},5
(6) [[‘the women’]] = {A + B + C},
(7) [[‘two women’]] = {A + B, A + C, B + C},
(8) [[‘the two women’]] = ∅,
 Descriptions with Adverbs of Quantification 67

(9) [[‘three women’]] = {A + B + C},

(10) [[‘the three women’]] = {A + B + C}.

When F is a mass noun, the relation expressed by ‘≤’ can be understood

as the relation of being a sub-portion of . With Sharvy, I’ll take a mass noun
such as ‘water’ to be a predicate true of every portion of water, and a modified
mass noun such as ‘water in my bathtub’ to be a predicate true of every
portion of water in my bathtub. The definite descriptions ‘the water’ and ‘the
water in my bathtub’ then each have singleton extensions (assuming that my
bathtub has some water in it), containing, respectively, all the water there is,
and all the water in my bathtub.
There’s an interesting metaphysical question whether the relations being
a sub-plurality of and being a sub-portion of can be viewed as two subspecies
of the parthood relation, and if so, whether we can view the expression ‘≤’
as univocally standing for the parthood relation (perhaps better described as
the relation being some of ) in the above statement of the semantics for the
definite article. I’m inclined to think that the answer to the second question
here is ‘no’, no matter what the answer to the first is. One reason concerns the
following sort of example, discussed by Sharvy. Suppose we have in a seminar
room two tables A and B pushed together to form a third table C. Then the
extension of ‘table’, restricted to the things in this room, is {A, B, C}. A
question then arises what the extension of ‘the table’ is to be on the current
proposal. Table C has the property of being a parthood-maximal member of
the extension of ‘table’; so if we therefore take it to be a ≤-maximal member,
then the description ‘the table’ (restricted to the things in this room) will
be a “proper” description, with {C} as its extension. Sharvy was happy to
accept this consequence; but I am not, for reasons I won’t go into, since
they’re too distantly related to the point of this paper. My view is that ‘the
table’ has an empty extension here since there are three tables in the room in
question, and so that any felicitous use of ‘the table’ here has to be dealt with
on the general model of “incomplete” descriptions. Given this view, we’re
required to deny that ‘≤’ can be interpreted univocally as standing for the
parthood relation. A second reason for denying this will emerge in the next
section.

II. Second Preface: The Logical Form of Descriptions as Predicates

The present view provides sentences containing descriptions in pred-

icative position after the copula ‘be’ with a straightforward semantics. If
the subject term of such a sentence is an expression of type e, that is, an
entity-denoting expression, then the value of the sentence as a whole is the
value yielded by applying the function that is the semantic value of the
68 Delia Graff Fara

description to the semantic value of the subject expression; equivalently (given

assumptions mentioned at the outset), the sentence is true just in case the
semantic value of the subject expression is in the extension of the description.
The following serve as examples:

(11) Hillary is a woman,

(12) Hilary is not a woman,
(13) Ana is the woman I met yesterday,
(14) Max is not the owner,
(15) Ana and Bea are the two women I met yesterday.

The interpretation of (11) through (15) proceeds as just outlined given some
minimal assumptions. First, I assume that the the ‘to be’ verbs here are
the ‘be’ of predication and have no semantic value. Second, for negative
sentences such as (12) and (14) I assume that the ‘not’ appearing after ‘be’
is semantically an operator on the entire sentence, so that (12), for example,
has a logical form like ‘[not [Hilary [a woman]]]’, and is assigned the truth-
conditions represented by its standard predicate-calculus paraphrase, ‘¬Wh’.
The Russellian, in contrast takes ‘be’ here to be the ‘be’ of identity and, in the
absence of supplemental syntactic principles, assigns two logical forms to (12),
corresponding to the predicate-calculus paraphrases ‘¬∃x(Wx ∧ h = x)’ and
‘∃x(Wx ∧ ¬h = x)’. Only the first of these, which is equivalent to the simple
‘¬Wh’, represents appropriate truth conditions for (12).6 Third, I take it
that conjoined names, such as ‘Ana and Bea’ in (16), denote a plurality—
represented here as the sum Ana+Bea.
The treatment of descriptions as predicates occurring in the argument
position of other predicates, as in (16), is somewhat more involved.

(16) The singer smokes.

Before sketching this treatment, I need to step back temporarily to set out the
assumptions I’ll be making about the syntax and semantics of quantification
in English and how to represent it.
The logical forms of sentences containing quantifier phrases will be
represented using restricted quantifiers and complex predicates formed by
abstraction, following the model of Barwise & Cooper (1981). A quantifier
phrase is represented as a “restricted quantifier,” formed by prefixing a one-
place predicate with a quantificational determiner, as in ‘[some : man]’. I
enclose the restricted quantifier in square brackets and use a colon to mark
the separation between determiner and predicate. A complex predicate is
formed by prefixing an open sentence with a variable-binding abstraction
operator, as in ‘x̂ x smokes’. A restricted quantifier combines with a one-place
predicate (complex or not) to form a sentence.7 Examples will help to
 Descriptions with Adverbs of Quantification 69

illustrate; the gloss below the first logical form indicates how it is to be
read:

(17) Juan loves every man,

[every : man] x̂ Juan loves x,
Every man is an x such that Juan loves x;
(18) Some man smokes,
[some : man] x̂ x smokes.

There are two main advantages to using this notation in preference to the
restricted-quantifier notation familiar to readers of Stephen Neale’s (1990)
Descriptions, where the logical forms of (17) and (18) are represented as
follows:

(19) [every x : man x] Juan loves x,

(20) [some x : man x] x smokes.

One advantage is that by separating out the variable-binding function from

quantification, by means of predicate abstraction, the quantifier phrases
in (17) and (18) may straightforwardly be treated not only as syntactic
constituents but also as semantic constituents of the whole. I take it that
the semantic value of a quantifier phrase is a function of type !!e, t", t", a
“generalized quantifier” in Barwise & Cooper’s (1981) sense.8 The semantic
value of a quantifier phrase of the form ‘every F’ is that type !!e, t", t"
function that assigns the value true to a type !e, t" function just in case
its extension contains everything in [[F ]].9 Similarly, the semantic value of
a quantifier phrase of the form ‘some F’ is that type !!e, t", t" function
that assigns the value true to a type !e, t" function just in case its extension
contains something in [[F ]]. A predicate abstract such as ‘x̂ Juan loves x’ has a
predicate-type semantic value whose extension consists of those things that,
when assigned to the variable x, render the open sentence ‘Juan loves x’ true.
The truth values of (17) and (18) can now be computed compositionally from
their logical forms.
A second advantage is that the logical forms in (17) and (18) are virtually
identical to the LF-level syntactic structures assigned to these sentences on at
least one popular treatment of the syntax of natural-language quantification
(namely that defended in Heim & Kratzer (1998)), and so can be straightfor-
wardly conceived as being derived from the surface forms of English sentences
by transformational rules that are reasonably well accepted by linguists
working in the tradition of transformational grammar. In particular these
logical forms can be derived from surface structures by a rule of quantifier
raising (as initially developed by Robert May; see May (1985)) by which a
quantifier phrase moves out of its argument position to adjoin to a larger
70 Delia Graff Fara

phrase containing that quantifier phrase at the level of surface form, to

which (in Heim & Kratzer’s (1998) implementation) a predicate-abstraction
operator has also affixed to bind the variable left behind as a trace of the
movement. In short, this is the transformation that takes us from the English
sentences in (17) and (18) to the logical forms of them represented there.
With this background in place, I am now in a position to say what,
on the descriptions-as-predicates view, are the logical forms of sentences
containing descriptions in argument position. I propose that like other
determiner phrases, in particular quantifier phrases, descriptions undergo
a process of raising, leaving behind a variable that gets bound in the
process by a predicate-abstraction operator.10 The difference is that unlike
quantifier phrases, descriptions slot into the predicative portion of a restricted
quantifier:

(21) [. . . [a/the F ] . . .] =⇒ [DET : a/the F ] x̂ [. . . x . . .].

Here ‘[. . . ! . . .]’ represents a sentence11 containing an argument position,

represented by ‘!’, that’s occupied by a description at the level of surface form.
The occupant DET of the restricted quantifier is either an unpronounced de-
terminer with either existential or generic force or an adverb of quantification
occurring explicitly in the sentence. With respect to the latter case, I want to
add two qualifications. First: DET can be an adverb of quantification only
if the description occurs in an argument position of the verb modified by
that adverb at the level of surface form. Second: where DET is an adverb
of quantification explicitly occurring in the sentence, I take it to have moved
out of its position in ‘[. . .[a/the F] . . .]’, so that it will not occur in ‘[. . .x . . .]’.
An example should serve to illustrate:
(22) A runner hardly ever smokes.

Here we have the indefinite description ‘a runner’ occurring in the

argument position of the verb ‘smokes’, which is (superficially) modified by
an adverb of quantification ‘hardly ever’. The rule represented in (21) allows
for three logical forms, with their truth conditions spelled out and glossed as
indicated:
(23) [∅∃ : a runner] x̂ x hardly ever smokes,
Something in the extension of [[a runner]] is such that it hardly ever
smokes,
Some runner hardly ever smokes;
(24) [∅G E N : a runner] x̂ x hardly ever smokes,
It’s generally true of a runner that it is such that it hardly ever
smokes,
In general, runners hardly ever smoke;
 Descriptions with Adverbs of Quantification 71

(25) [Hardly ever : a runner] x̂ x smokes,

Hardly anything in the extension of [[a runner]] is such that it smokes,
Hardly any runners smoke.

Note that the predicate ‘smokes’ plays a very different semantic role in (23)
and (24) from the one it plays in (25). The difference can be captured by saying
that ‘smokes’ in (23) and (24) seems to be a predicate that’s true of a thing
at a time just in case it’s smoking at that time, whether or not that thing is a
smoker; while ‘smokes’ in (25) is a predicate that’s true of a thing at a time just
in case it’s a smoker at that time, whether or not that thing is smoking at that
time. I’ll describe the difference using the terms ‘episodic’ and ‘habitual’. The
predicate ‘smokes’ gets an episodic reading in (23) and (24), but a habitual
reading in (25). In saying this I do not take myself to be committed to the
claim that the word ‘smokes’ is ambiguous, only that predicates in which
it occurs may have either an episodic or a habitual reading. We might for
example say that ‘smokes’ invariably has the episodic meaning, but that it
may be modified (i) by an explicitly-occurring adverb of quantification—
leading to what I’ll call a “frequency of episode” reading for the modified
verb phrase “hardly ever smokes”; or (ii) by an unpronounced one with
something like the meaning of ‘generally’, ‘regularly’, ‘characteristically’ or
‘habitually’—leading to what I’ll call the “habitual” reading for the phrase
‘smokes’. I take it that (22) does have the three different truth conditions
displayed above, and that the descriptions-as-predicates view scores a point
against the Russellian in being able to account for them without positing any
ambiguity in the indefinite description.
A similar variation in quantificational force also occurs with definite
descriptions:

(26) The owner of an espresso machine rarely goes to bed early.

As before, the predicate ‘goes to bed early’ gets an episodic reading in the
first two cases (leading to a “frequency of episode” reading for the modified
verb phrase ‘rarely goes to bed early’), but a habitual reading in the third
case:
(27) [∅∃ : the owner of an espresso machine] x̂ x rarely goes to bed early,
Some (unique) owner of an espresso machine rarely goes to bed
early;
(28) [∅G E N : the owner of an espresso machine] x̂ x rarely goes to bed
early,
In general, espresso-machine owners rarely go to bed early;
(29) [Rarely : the owner of an espresso machine] x̂ x goes to bed early,
Rare is the espresso-machine owner who goes to bed early.
72 Delia Graff Fara

The important thing to note about this example is that it contains a definite
description that can be true of more than one thing—despite the uniqueness
condition I’ve built into the semantics of the definite article—because the
uniqueness of ownership is relativized to individual espresso machines. This
can be made explicit in the logical form by representing the indefinite ‘an
espresso machine’ as taking its scope within the restricted quantifier but
outside the definite article:
(30) [DET : x̂ [∅∃ : an espresso machine] ŷ x (is) the owner of y].12

If the definite description here weren’t capable in principle of being true of

more than one thing then it wouldn’t make sense to quantify over those things
satisfying it using adverbial quantifiers like ‘generally’, ‘rarely’, ‘usually’
and ‘always’. The reason is that using one of these adverbial quantifiers
to quantify over things of a given sort pragmatically generates the strong
implication that there is more than one thing of that sort.
This brings to the fore an interesting feature of the treatment of generics
on the descriptions-as-predicates view, which I’ll briefly highlight before
moving on to the next section. So far, genericity has been treated as resulting
from a description’s restricting an unpronounced determiner ∅ GEN that has
generic quantificational force.13 But this treatment will not work for those
cases when a definite description is semantically required to have either a
singleton or an empty extension but which is nonetheless used generically:
(31) The tiger is an animal that breeds well in captivity.

The two logical forms assigned to this sentence (only two because it does not
contain an explicit adverb of quantification) are:

(32) [∅∃ : the tiger] x̂ x is an animal that breeds well in captivity,

(33) [∅G E N : the tiger] x̂ x is an animal that breeds well in captivity.

The second logical form here, the one with the generic, is perfectly well-
formed and is semantically interpretable, but is ruled out on pragmatic
grounds since ‘the tiger’ can be true of at most one thing. Yet the logical
form with the existential quantifier doesn’t seem to capture the generic truth
conditions of (31). For all that’s been said so far, given that ‘tiger’ is neither a
plural nor a mass term, (31) is wrongly predicted to be true just in case there
is exactly one tiger and it breeds well in captivity. But this is not obviously
what (31) means. After all, breeding is not something a tiger can do on its
own!
However, more can be said. Sharvy intended his generalization of Rus-
sell’s theory of descriptions to account for definite descriptions with plurals
and mass terms. But it, and its modification here, can also apply to the case
of kind definite descriptions: cases where the nominal complement of the
 Descriptions with Adverbs of Quantification 73

definite article, whether count or mass, has a “taxonomic” interpretation.

For example, the common noun ‘tiger’ can receive both an individual
interpretation and a taxonomic interpretation, as shown by the following:

(34) There are three tigers in this room; I hope they’re not hungry;
(35) Three tigers went extinct in the last century: the Caspian, the
Balinese, and the Javan.

If we take it that on its taxonomic interpretation the noun ‘tiger’ is a predicate

true of every tiger-kind—leaving open just what kind of thing a tiger-kind
is and also leaving open just which collections of tigers constitute a tiger-
kind, but allowing that a subspecies of tiger is a tiger-kind—then since kinds
can be ranked according to a parthood-like relation, the relation of being a
sub-kind of , Sharvy’s proposal provides us with a way of dealing with the
genericity of the sort exhibited in (31).14 The definite description ‘the tiger’
in (31) will on the present proposal be a predicate true of just one thing, that
tiger-kind of which every tiger-kind is a sub-kind—namely, the entire tiger
species.15 In order to say that we’ve thereby captured the “generic” truth
conditions of (31), we have to add that predicates like “has stripes” can be
true of tiger-kinds as well as individual tigers.
The point being emphasized here is that my analysis of descriptions
as predicates includes two different treatments of genericity for definite
descriptions, depending on whether they are at most singly applicable, i.e.,
true of at most one “thing,” or are multiply applicable, i.e., in principle
true of more than one thing, as a result (for example) of containing an
embedded indefinite. In the former case, we have existential quantification
over a kind-level predicate; in the latter case we have generic quantification
over an individual-level predicate:
(36) The mystery novel is popular,
[∅∃ : the mystery novel] x̂ x is popular,
(Taxonomic interpretation for ‘mystery novel’);
(37) The author of a mystery novel is (generally) popular,
[∅G E N : the author of a mystery novel] x̂ x is popular,16
(Individual interpretation for both ‘author’ and ‘mystery novel’).

I should at this point reiterate something mentioned only in passing

above, namely, that I also take bare plural noun phrases to be yet another
kind of predicate nominal, so that genericity with bare plurals is to be dealt
with in the same manner as indefinite descriptions and multiply-applicable
definite descriptions. The generic reading of ‘tigers have stripes’ (which is the
only reading in the absence of special intonational emphasis) is accounted
for by assigning to this sentence the following logical form:
(38) [∅G E N : tigers] x̂ x have stripes.
74 Delia Graff Fara

Although it’s not superficially apparent, it’s worth noting that there is also
an analogous difference, albeit less marked than in the case of singly- versus
multiply-applicable definites, between the treatment of singular-indefinite
generics and that of bare-plural generics. Both involve generic quantification
on the treatment proposed here, but in the singular-indefinite case this is
generic quantification over individuals, while in the bare-plural case this is
generic quantification over pluralities of individuals. I’d like to think that
this difference could be appealed to in accounting for at least one important
difference between the two.17 The distinction is between the following:

(39) Tigers are extinct,

(40) A tiger is extinct.

‘Extinct’ is a predicate that can only be true of a species (or subspecies), not of
individual members of the species. Yet for some reason it is combinable with
the bare plural even on its non-taxonomic interpretation, while combinable
with the singular indefinite only on its taxonomic interpretation. This could be
handled on the present view by allowing “kind-level” predicates like ‘extinct’
to be true of pluralities as well as kinds. The bare-plural sentence would then
be interpreted on one of its false but sensible readings as generic quantifica-
tion over tiger-pluralities, and on its other false but sensible reading as generic
quantification over tiger-kind–pluralities;18 while the second sentence would
be interpreted on its sensible (and true) reading as existential quantification
over tiger-kinds.
Let me wrap up this section by summarizing the main points so far.
Definite and indefinite descriptions invariably have predicate-type semantic
values. When one occurs in an argument position of a predicate, it undergoes
a process that’s like quantifier-raising except that it slots into the predicative
(restrictor) portion of the restricted quantifier. The quantifier portion of this
restricted quantifier is either an unpronounced quantificational determiner
with either existential or generic force, or an adverb of quantification explic-
itly occurring in the sentence. Definite descriptions, however, except in those
special cases where they are multiply-applicable (for example because they
contain an embedded indefinite), can restrict only an existential quantifier,
the other quantifiers being ruled out on pragmatic grounds.19 Generic
readings for singly-applicable definite descriptions result not from generic
quantification but from a taxonomic reading for the predicative complement
of the definite article.

III. Descriptions with Adverbs of Quantification

I take it that one of the most forceful arguments for the descriptions-as-
predicates view stems from the fact that descriptions (now including bare
 Descriptions with Adverbs of Quantification 75

plurals under this heading) interact with adverbs of quantification in a

way that quantifier phrases do not. In “Descriptions as Predicates” (§VII)
I claimed that the a-sentences in the sets of examples below are multiply
ambiguous while the b-sentences are unambiguous.
(41) a. The owner of an espresso machine rarely prefers tea to coffee,
b. Most owners of an espresso machine rarely prefer tea to coffee,
c. Joe rarely prefers tea to coffee,
d. Joe prefers tea to coffee.

(41a) can be understood in any of the three ways so far predicted by

the analysis: as saying (i) that some one espresso-machine owner has the
“frequency of episode” property attributed to Joe by (41c)—represented as:
[∅∃ : the owner of an espresso machine] x̂ x rarely prefers tea to
coffee;
(ii) that espresso-machine owners in general have the frequency-of-episode
property attributed to Joe by (41c)—represented as:
[∅G E N : the owner of an espresso machine] x̂ x rarely prefers tea to
coffee;
or (iii) that few owners of an espresso machine have the “habitual” property
attributed to Joe in (41d)—represented as:

[rarely : the owner of an espresso machine] x̂ x (habitually) prefers

tea to coffee.

(41b), in contrast, is unambiguous, and can only mean that most espresso-
machine owners have the frequency-of-episode property.
Analogous claims were made about the interaction of indefinite descrip-
tions, bare plurals, and plural definites with adverbs of quantification:
(42) a. A Scandinavian rarely has brown eyes,
b. Every Scandinavian rarely has brown eyes,
c. Joe rarely has brown eyes,
d. Joe has brown eyes;
(43) a. Philosophers sometimes smoke,
b. Both philosophers sometimes smoke,
c. Joe sometimes smokes,
d. Joe smokes;
(44) a. The parents of a toddler usually have little time for relaxation,
b. Many parents of a toddler usually have little time for relaxation,
c. Joe usually has little time for relaxation,
d. Joe has little time for relaxation.
76 Delia Graff Fara

It is true that (42a) can’t equally naturally be read as having any of the three
readings (so far) assigned to it by the descriptions-as-predicates analysis. But
this is obviously to be explained on pragmatic grounds: having brown eyes
tends to be a permanent property of the things that have it, which is why we
most naturally read (42a) as saying that few Scandinavians have brown eyes.
The other readings (so far) assigned to this sentence by the descriptions-
as-predicates analysis involve attribution of the very strange frequency-of-
episode property attributed to Joe in (42c) (strange, at least, in the days
before tinted contact lenses) and are therefore less salient. (42b), in contrast
(I claimed), is unambiguous and can only mean that every Scandinavian has
the strange frequency-of-episode property.
I took these sets of examples to provide confirmation for the descriptions-
as-predicates view and a strong argument against Russellianism. The confir-
mation for the descriptions-as-predicates view stems from the fact that it
smoothly accounts for the “quantificational variability effects” of descrip-
tions (definite and indefinite, singular and plural) displayed in the a-sentences
in (41–44). Moreover, I take these sets of examples to provide confirmation
not just for the predicate view as contrasted with the uniform Russellian
view,20 but also for the predicate view as contrasted with a “type-shifting”
view, as proposed by Barbara Partee (1987) and accepted by virtually all
linguists who commit to a view on the matter—on which descriptions are
systematically and predictably ambiguous between a type !e, t" interpretation
when occurring in predicative position, and either a type e or a type !!e, t", t"
interpretation when occurring in argument position. For if descriptions in
argument position at least some of the time have to receive a type !e, t"
interpretation in order to serve as the restrictor of an adverb of quantification,
then since allowing them to uniformly take a type !e, t" interpretation provides
a smooth way of capturing the further ambiguity between existential and
generic readings in a way that, as we saw in the last section, provides a
nice account (given Sharvy’s important work and the extension of it here
to kind-level descriptions) of a wide variety of semantic facts about generic
statements, we have no good reason to adopt type-shifting.
The argument against Russell’s theory of descriptions stems from the
fact that however its proponents may ultimately explain the quantificational
variability effects that look troubling for it, the difference between the a-
and b-sentences in (41–44) with respect to the existence of quantificational
variability effects points to an important disanalogy between descriptions and
quantifier phrases. Call this the quantificational-variability argument against
Russellianism.
It’s been stressed to me that it was not quite right to say that quantifier
phrases don’t ever interact with adverbs of quantification in the way that
descriptions do.21 The adverbs of quantification in the b-sentences above
may obligatorily modify the verb phrase in those sentences, but there are
 Descriptions with Adverbs of Quantification 77

very closely related sentences that don’t place this restriction on the adverbs
in them.
Given the right conversational setup, the adverbs of quantification in (45–
47) can be understood as quantifying over some conversationally salient type
of situation or event rather than as modifying the main verb. For example,
imagine that each year I have a party to which I invite owners of various
types of kitchen machines: orange-crushing machines, apple-coring machines,
espresso machines, etc. In a conversation where it is under discussion that I
annually hold such a party, (45) can naturally be interpreted as meaning that
it’s occasionally the case that at one of these parties some of the espresso-
machine owners at that party prefer (habitually) tea to coffee:
(45) Some of the espresso-machine owners occasionally prefer tea to
coffee.

Given discussion of an annual party to which I invite people from various

European countries, (46) can with a little strain be interpreted as meaning
that it’s rarely the case at one of these parties that all of the Scandinavians
at that party have brown eyes:
(46) All of the Scandinavians rarely have brown eyes.

And, just to round things out, given discussion of an annual party to which I
invite two people from each academic department in my university, (47) can
easily be interpreted as meaning that it’s sometimes the case at one of these
parties that both of the philosophers at that party (habitually) smoke:
(47) Both of the philosophers sometimes smoke.

Presumably, the reason these quantification-over-parties readings be-

come available—given the right conversational background—with the par-
titives above (which are, roughly speaking, “blank-of-the-blank”-phrases)
is that partitives for some reason already require some contextually salient
domain to quantify over. “Most of the philosophers called me last night,” said
in the absence of such a contextually salient domain’s having been provided is
rightly greeted with the question, “Most of which philosophers?” The domain-
providing property thing at my party that needs to be contextually salient for
the use of the partitive to be appropriate stands out as being also saliently
available to restrict (in some yet-to-be-spelled-out way) the domain of an
adverb of quantification.22
So the question is, does the existence of these “extra” readings—let’s call
them situation-readings—for the partitive counterparts of the b-sentences in
(41–44) undermine the quantificational-variability argument against Russel-
lianism? The point I wish to make here is that it does not. The reason is
that just as there is an extra, situation reading for the sentences in (45–47),
78 Delia Graff Fara

neglected in Fara (2001), on which the quantifier phrases can vary their
denotation with respect to the different situations being quantified over, there
is also an exactly analogous extra, situation reading—equally neglected in the
earlier discussion—for the a-sentences containing descriptions in (41–44). For
example, in a situation in which it’s under discussion that I have an annual
party to which I invite, for each of a number of kinds of kitchen machine,
one person who owns a machine of that kind, (41a) (repeated here) can be
understood as saying that it is rarely the case at one of these parties that the
owner of an espresso machine at that party prefers (habitually) tea to coffee.

(41) a. The owner of an espresso machine rarely prefers tea to coffee.

We might wish to describe this as a case in which the description “interacts”

with the adverb of quantification in the sense that this interpretation requires
us to evaluate the description in some or all of a variety of situations being
quantified over by the adverb. But it is not a case in which the description
interacts with the adverb in the special way provided by the descriptions-as-
predicates view, namely, as a restrictor of it.
In short, we have on the one hand that for each way of interpreting a
sentence with a quantifier phrase in the argument position of a predicate—
whether an adverb of quantification explicitly occurs in the sentence or
not, there is an exactly analogous way of interpreting the corresponding
sentence with a description in the place of the quantifier phrase. This is
just as we’d expect, since the Russell/Sharvy analysis of descriptions as
quantifier phrases yields truth conditions that turn out to be a special case of
those proposed by the descriptions-as-predicates view, namely, the case of a
description restricting an unpronounced determiner with existential force. On
the other hand, there are ways of interpreting a sentence with a description in
the argument position of a predicate that have no analogous interpretation
when the description is replaced with a genuine quantifier phrase. This fact
suggests that however the Russellian might account for the interpretations
we’ve accounted for here as an adverb being restricted by a description qua
predicate, that account will inevitably predict that other quantifier phrases
interact with adverbs of quantification in a way that they just don’t seem to.
A particularly striking minimal pair is the following:

(48) A Scandinavian usually has blue eyes,

(49) Some Scandinavian usually has blue eyes.

The preferred interpretation for (48) intuitively involves quantification over

individual Scandinavians, with the force of the quantification being that
given by the adverb ‘usually’. The descriptions-as-predicates view captures
that reading—which I’ll tendentiously call the quantification-over-individuals
reading—in exactly the way suggested by intuition: by allowing the adverb to
 Descriptions with Adverbs of Quantification 79

quantify over the individuals in the extension of the indefinite description by

forming a restricted quantifier with it: ‘[usually : a Scandinavian]’.23 Crucially,
there is no quantification-over-individuals reading for (49); it cannot mean
that most Scandinavians have blue eyes.
We’ll return to this latter fact in just a moment, but let’s first ask:
how might an ardent Russellian capture the quantification-over-individuals
interpretation for the sentence with the indefinite description? Our prediction
was that any proposal along these lines would incorrectly assign the same
interpretation to (49). One sort of proposal involves adverbial quantification
over “minimal” situations. Two proposals in this vein are found in Heim
(1990) and von Fintel (1994, 2004).24 One important feature of the “situa-
tions” deployed in these proposals is that they can have situations as parts
and can be parts of situations. A “minimal” situation in which it’s the case
that ! (call this “a !-situation”) is a !-situation that has no !-situation as a
proper part. The intuitive idea should be clear enough for the purposes of our
discussion here.25 The general form of von Fintel’s version of the proposal,
applied to a sentence that can be rewritten as one with the form ‘ADV: !’ is
that it gets the following truth-conditions: ADV-many minimal "-situations
are a part of a minimal (! ∧ ")-situation, where " is a restriction provided
by the conversational context. Now the paraphrase of the analysis of (48) is:

(50) Most minimal situations in which there’s a Scandinavian are part

of a minimal situation in which: a Scandinavian has blue eyes.26

Note that there’s a one-one correspondence between Scandinavians and

minimal situations in which there’s a Scandinavian. Quantification over
minimal situations in which there’s a Scandinavian for this reason mimics the
effect of regular quantification over individual Scandinavians. This is not a
mere trick, however; once quantification over situations is adopted, and once
it’s allowed that situations are parts of other situations, and, moreover, that
any situation that contains a !-situation as a part must itself be a !-situation,
the domain of quantification must be restricted to minimal situations in order
to avoid what we’d intuitively think of as counting the very same situation
multiple times. The reason for having the “larger” situation also be a minimal
situation is equally well motivated; for if the larger situation weren’t minimal,
the existence of a single blue-eyed Scandinavian would suffice to render the
entire claim true. Intuitively, we want the blue-eyed Scandinavian in the
larger situation to be the very Scandinavian that’s in the smaller situation.
The indefinite description ‘a Scandinavian’ can now be interpreted in the
Russellian manner as an existential quantification. The Russellian can in this
way account for the interaction of descriptions with adverbs of quantification.
It’s now time to return to the pressing question for the Russellian: why
is it that (49) does not allow a quantification-over-individuals reading while
(48) does?
80 Delia Graff Fara

(48) A Scandinavian usually has blue eyes.

(49) Some Scandinavian usually has blue eyes.

As far as I know, von Fintel is the only one to explicitly address this problem.
But his remarks are very brief, and meant only to suggest an avenue for
further investigation. He says that ‘some’, unlike ‘a’, is “inherently partitive”
and so that the situations being quantified over by the adverb each contain
all relevant Scandinavians. The quantifier-variability effect “will not arise,”
he writes, “because it is a side-effect of quantifying over very small situations
containing only one [Scandinavian] each” (2004, p. 163). The suggestion is
that ‘some Scandinavian’ is like ‘at least one of the Scandinavians’ in that it
requires for its felicitous use a contextually salient set of Scandinavians; but
unlike it, presumably, in allowing that set to be a singleton. In the absence
of such a contextually salient set of Scandinavians, a user of one of these
partitive phrases would be appropriately greeted with the question, “Which
Scandinavians are you talking about?” The further suggestion is that if,
for example, we’re talking about the Scandinavians at my annual party of
Europeans, then the minimal situations being quantified over by the adverb
must each contain enough of one of these parties to have all the Scandinavians
at that party in it (as well as enough of the party, I presume, to be a party).
The truth-conditions that result from this are just equivalent to what we
earlier called the situation-reading for the sentence. There’s no mimicking of
quantification over individuals because there’s not a one-one correspondence
between these minimal situations and Scandinavians.
My reply to this suggestion consists of two main parts. The first point
is that although I agree (I think) that there is a partitive-like use of ‘some’-
phrases, the availability of this partitive-like use is extremely sensitive to subtle
variations in intonation. In any case it’s not always there, since it’s perfectly
acceptable to initiate a conversation with an out-of-the-blue use of a ‘some’-
phrase. Compare the following.

(51) Guess what I just heard. One of the Scandinavians occasionally has
purple eyes.
(Appropriate response: which Scandinavians are you talking about?)
(52) Guess what I just heard. Some Scandinavian occasionally has purple
eyes.
(Appropriate response: Do you mean to say that he or she doesn’t
always have purple eyes?)
I’ve substituted a more news-worthy predicate in the example, in order to
render it plausible as a conversation-opener.
The second point is that the problem is not to be accounted for by
attributing special properties to ‘some’; the problem is much more widespread
than that. Consider the following:
 Descriptions with Adverbs of Quantification 81

(53) Few Scandinavians usually have blue eyes,

(54) Hardly any Scandinavians usually have blue eyes.

Both of these sentences clearly have a reading involving attribution of the

strange frequency-of-episode property. But if we apply the situation approach
to these sentences we get the following: Most minimal situations in which
there are few (hardly any) Scandinavians are a part of a minimal situation
in which few (hardly any) Scandinavians have blue eyes. But this is roughly
equivalent (it seems to me) to the claim that most Scandinavians have blue
eyes; clearly not an interpretation for (53) or (54)!
The upshot is that our prediction has been borne out. The situation-based
approach to preserving Russell’s theory of descriptions, although successful
insofar as it is capable of mimicking the quantification-over-individuals
readings for descriptions with adverbs of quantification, is unsuccessful in
that it does not account for the differential behavior between descriptions
and genuine quantifier phrases in this regard.
By way of closing I’d like to address one question still left unanswered,
namely, how on the descriptions-as-predicates view are we to handle those
uses of adverbs of quantification that do intuitively involve quantification
over situations? These were the “extra” readings for the a-sentences in (41–44)
as uttered, for example, against a background discussion of my annual parties
of this or that type. The question is more sharply focused by considering a
singly-applicable definite description with an adverb of quantification. The
following example is discussed by Jason Stanley (2002):

(55) The customer is always right.

One could, I suppose, utter this sentence to say of one particular (salient)
customer that that customer is perpetually right. But on its natural, almost
proverbial, reading, it’s a claim about any number of customers. Because this
is a singly-applicable definite description, it cannot sensibly restrict the adverb
‘always’, since this would yield something like a false “scalar” implicature to
the effect, in essence, that this singly-applicable definite description multiply
applies. So we cannot account for the natural reading of this sentence by
assigning it the logical form: ‘[always : the customer] x̂ x is right’. But note
that what’s being claimed here is not intuitively even like a quantification-
over-individuals reading for the adverb. We don’t say, in uttering this sentence,
that every customer is habitually right. Rather, we say that every customer in
a certain type of situation is right in that situation. What’s the relevant type of
situation? Something like a dispute between customer and employee. It looks
like we need to be able to accommodate something like the situation-based
approach in order to account for sentences like these (as well as the situation-
readings for the a-sentences in (41–44)). But there is nothing inherently
undermining in that. The semantics proposed here for descriptions will have
82 Delia Graff Fara

to modularly mesh with semantic accounts for any other type of phrase it
can co-occur with.27
Let’s look at Stanley’s proposal.28 He argues that we can account for
this sentence as a case of binding into a nominal restriction, a device which,
he proposes, accounts for a wide and somewhat disparate-seeming variety of
cases of context-dependence, including especially quantifier-domain restric-
tion:29

(56) Always i : the customer f (i) is right.

Here ‘f ’ occurs as a variable that in the context gets assigned a function

from objects to sets of individuals. Intuitively, it’s a function which, given a
dispute as argument, yields the set of parties to that dispute as value. The
extension of ‘customer f (i) ’, for a particular value of ‘i’, is the intersection of
the extension of ‘customer’ with the set f (i). Since the resulting intersection is
supposed to be a singleton set, the Russellian analysis of descriptions could
be maintained in face of this example. One potential difficulty here, though,
is that if ‘Always i ’ ranges over everything there is, then f (i) will either be
undefined for most values of ‘i’, in which case I’d think that the sentence
would be truth-valueless, or it will yield, for many values of i, sets that don’t
have any customers in them at all or even any things that could be right
or wrong, in which case I’d think that the sentence would be false. In any
case, putting this worry aside, this proposal in its bare-bones form needn’t
be viewed as an alternative to the situation-based approach, but could rather
be viewed as one way of implementing it.
The point I wish to make here by way of closing is that the descriptions-
as-predicates view can also make use of this form of nominal restriction in
order to capture situation-readings for adverbs of quantification. Let’s think
of the ‘i’ in a nominal restriction as occurring in an argument position; this
is natural if we think of the function f (!) assigned to ‘f ’ as doing the work
of a prepositional phrase such as ‘in (!)’. If ‘i’ is left free and is assigned
an entity as semantic value (in the context), then it can be interpreted in
situ. While if it is assigned something of type !e, t" or !!e, t", t", then it
must raise out of argument position. So suppose it gets assigned as value the
value of a description such as ‘a dispute between customer and employee’—‘a
dispute’ for short. Then on the description-as-predicates view, we can assign
the following surface and logical forms to the sentence, using expressions
in place of variables just to indicate what values those variables are getting
assigned by context:

(57) a. The customer in(a dispute) is always right,

b. [Always: a dispute] î the customerin(i) is right,30
c. [Always: a dispute] î [∅∃ : the customerin(i) ]x̂ x is right.

This says what we want it to say: that every dispute (between customer and
employee) is an i such that the customer in i is right.31 One advantage this
 Descriptions with Adverbs of Quantification 83

has over the situation-based approach is that in quantifying over disputes,

rather than minimal situations, we can retain a relatively simple semantics
for the adverb, rather than pack into its lexical meaning, as is typically done
on the situation-based approach, the notions of being a minimal !-situation
and being an extension of a situation. This will remain an advantage only to
the extent that nominal restriction is sufficiently widespread and flexible to
be used in this way.32

Notes

1. Throughout, I’ll use single quotes for both name-forming quotation and quasi-
quotation, and will drop the quotes in some cases for the sake of readability.
2. For a more developed and formal presentation of this idea, consult Link (1983,
for example).
3. For this proposal, along with independent motivations for it, see for example
Ladusaw (1982), Link (1987), and Krifka (1999).
4. For definite–numeral combinations I assume the syntactic structure [the [num F ]]
rather than the structure [[the num] F].
5. Here I represent pluralities as sums, which would have annoyed Sharvy given his
insistence (Sharvy 1980, p. 620) that, e.g., the men in Auckland are men, and hence
essentially a plurality, rather than some one single, mysterious sort of entity.
6. This is noted by George Wilson (1978, p. 51f.) and James Higginbotham (1987),
and is discussed in section IV of Graff (2001).
7. The syntax is reminiscent of that for Robert Stalnaker’s (1977) language L̂, in
which a logical formula such as ∀x̂F x is parsed as consisting of the quantifier ∀
and the complex predicate x̂F x formed by predicate abstraction. It extends the
syntax of L̂ in allowing restricted as well as unrestricted quantifiers.
8. In saying this, I blur the distinction between a set and its characteristic function
on the relevant domain, since Barwise & Cooper take a generalized quantifier to
be a set of sets of individuals.
9. This generalized quantifier can be obtained compositionally, moreover, from the
semantic value of the determiner and that of F, by assigning the determiner ‘every’
the appropriate type !!e, t", !!e, t", t"" function. Consult Barwise & Cooper (1981,
§2) or Heim & Kratzer (1998, ch. 6) for further details and elaboration.
10. A brief remark about terminology: in calling descriptions “determiner phrases”
I assume that the definite and indefinite articles, as well as numeral words, are
determiners, even though on my proposal they have the semantics of a predicate
modifier. I reserve the word ‘quantifier’ for those determiners with semantic type
!!e, t", !!e, t", t"", and ‘quantifier phrase’ for phrases of type !!e, t", t". ‘Restricted
quantifier’ is the term I use for those expressions that represent quantifier phrases
in the language being used to represent logical forms.
11. Or other phrase that has a type t semantic value at the level of logical form.
12. Essentially, we have allowed the (restricted quantifier restricted by the) indefinite
description to adjoin to something other than a sentence, namely, to the definite
description in which it is embedded. This sort of quantifier raising had already
been proposed by May (1985) under the heading of “inverse linking”. Allowing
it requires some complications of either the syntax or the semantics. Rather than
84 Delia Graff Fara

complicate the semantics by allowing for type-shifting of quantifier phrases, I

prefer to complicate the syntax. I was rather hand-wavy about this in Graff (2001);
so let me be more explicit here. I adopt the kind of solution for interpreting these
embedded determiner phrases that’s outlined by Heim & Kratzer (1998, ch. 8).
The solution involves positing an empty subject PRO in the definite description,
which, although it has no semantic value itself, can act as if it had one by raising
and leaving a bound variable as trace. So the internal logical form of the definite
description here is derived as follows:
[ DP PRO the owner of an espresso machine] =⇒
[DP [∅∃ : an espresso machine] ŷ[DP PRO the owner ofy]] =⇒
[DP PRO x̂ [∅∃ : an espresso machine] ŷ[DP x the owner of y]].
This is perfectly interpretable given the semantic rules already provided. Together
with the assumption that ‘PRO’ is semantically vacuous, these rules yield a
predicate that is true of a thing just in case there’s some espresso machine it
uniquely owns. Some may balk, of course, at the uniqueness requirement here;
thinking that (26), on any of its readings, is as much about those who jointly own
an espresso machine as it is about those who uniquely own one. I sympathize
with the worry, but I suspect that it can be assuaged without overhauling the
view, yet I won’t attempt to do that here since I think that the issues involved are
detachable from those of primary concern in this paper. For interesting discussion
of the history of treatments of inverse linking and the importance of it to the
development of syntax and its interface with semantics, see May & Bale (2006).
13. I don’t take a stand on what exactly the meaning of the generic quantifier is,
at least not as stated in terms that don’t involve using it; but I take myself
nonetheless to be providing truth conditions for logical forms containing it by
using it itself, since I take it that we have, to a sufficiently large extent anyway,
an intuitive grasp of its meaning. I call ∅ GEN a quantifier, since by this I just
mean an expression of type !!e, t", !!e, t", t"", that is, an expression that combines
with a one-place predicate to form an expression of type !!e, t", t". Ascertaining
its contribution to truth-conditions is a notoriously difficult matter. I’ll refer
the reader to two reasonably plausible recent accounts, both containing good
overviews of the literature on this issue: Cohen (1999) and Leslie (n.d.).
14. I take it that the relation of being a sub-kind of is definitely not a subspecies of
the parthood or some-of relation, and that for that reason we cannot take ‘≤’ in
our statement of the semantics of the definite article to univocally express this
relation. Rather, we must let it variably stand for different sorts of parthood-like
relations depending on the structure of the extension of the predicate it operates
on. We cannot let it stand univocally for some more general parthood-like relation
of which the parthood relation as well as the sub-kind relation are both subspecies,
because of the fact discussed above—that tables, for example, can have other tables
as parts.
15. The discussion of this in Fara 2001 (pp. 30–31) is flawed in that it does not
explicitly make use of Sharvy’s maximality operation, as it should have. The
discussion there was a holdover from earlier versions of the view on which I’d
not yet adopted Sharvy’s generalization of Russell’s uniqueness condition, and I
recognized the oversight only after it became too late to make revisions. Roberto
 Descriptions with Adverbs of Quantification 85

Zamparelli (1998) had already proposed the same idea in work I’ve only since
become aware of.
16. Here I suppress the internal logical form of the definite description inside the
restricted quantifier for the sake of readability. Spelled out, the whole thing
becomes:
[∅GEN : ŷ([∅∃ : a mystery novel] ẑ y ((is) the author of z))] x̂ x is popular;
which is worth emphasizing so that it is clear that this is a multiply-applicable
definite description.
17. This much-discussed difference is observed in Lawler’s (1973) dissertation. For
more recent discussion of it, see for example Greenberg (2002) and the references
cited there.
18. Three subspecies of tiger went extinct in the last century; there are five remaining
ones (all endangered, three on the brink of extinction).
19. The fact that definite descriptions are subject to quantificational-variability effects
has pretty much been ignored or gone unnoticed. I suspect that this is because it’s
only multiply-applicable definite descriptions that are capable of exhibiting this
genericity.
20. As vigorously defended by Stephen Neale (1990).
21. This has been stressed to me in discussion by Jeffrey King and Jason Stanley, and
also in a presentation by Berit Brogaard (2005).
22. Partitives aren’t required in order to get a quantification-over-situations reading
for quantifier phrases with adverbs of quantification. For example:
• All philosophers in the department sometimes attend an APA meeting;
• Some student who procrastinates usually fails.
The tentative generalization is that we get a quantification-over-situations reading
for these sentences because the sentences themselves raise to salience a type
of situation, either by mentioning it explicitly (APA meetings) or by using an
implicitly relational phrase (‘fail’) that has a type of situation (a class, say) as a
relatum. The first example was given to me by King, the second by Brogaard.
23. The descriptions-as-predicates view shares this advantage with Lewis’s (1975)
original proposal, and the developments of it instigated by Kamp (1981) and
Heim (1982).
24. The proposal Neale sketches (1990, pp. 247–251) can be viewed as a start in this
same direction. I should mention that Heim and von Fintel explicitly commit to
Russellianism only about indefinite descriptions.
25. It seems to me that these situations are very much like states of affairs that obtain.
26. Since ! in this case entails ", we can just re-write (! ∧ ") as !. I should mention
that von Fintel himself would probably not accept this paraphrase (though I
suspect that Heim would) since he implies that a bare minimal situation in which
there’s a Scandinavian is not sufficiently “situation-like” to serve as a contextual
restriction for an adverb of quantification, and would apparently prefer instead
to quantify over minimal situations in which we’re “encountering” or “sampling”
Scandinavians (2004, p. 163). I don’t think this is a promising tactic, in part for the
reason discussed by Lewis with respect to his example about quadratic equations,
but also because it requires that we the encounterers in these situations are not
ourselves blue-eyed Scandinavians.
86 Delia Graff Fara

27. This point is well-made throughout Neale’s (1990) Descriptions (with respect to
his favored theory, of course).
28. See pp. 386f.
29. See Stanley & Szabó (2000).
30. In order for this to be in keeping with the requirement that the description re-
stricting an adverb occur in the argument position of the verb phrase superficially
modified by that adverb, we have to think of ‘right’ as it occurs here as really
meaning something like ‘right in i’, which seems desirable.
31. Or: ‘is right in i’; see the preceding footnote.
32. For discussion of matters relevant to this paper I am especially indebted to Berit
Brogaard, Paul Elbourne, Michael Fara, Irene Heim, Jeffrey King, Robert May,
Daniel Rothschild and Jason Stanley.

References

Barwise, Jon and Cooper, Robin. (1981) ‘Generalized Quantifiers and Natural Language’,
Linguistics and Philosophy 4: 159–219.
Brogaard, Berit. (2005) Russell’s Theory of Descriptions vs. the Predicative Analysis: a Reply
to Graff. Presentation to the Eastern Division meeting of the American Philosophical
Association, New York, December 2005.
Cohen, Ariel. (1999) Think Generic!, CSLI Publications, Stanford, CA.
Fara, Delia Graff. (2001) ‘Descriptions as Predicates,’ Philosophical Studies 102(1): 1–42.
Originally published under the name ‘Delia Graff’.
von Fintel, Kai. (1994) Restriction on Quantifier Domains, PhD thesis, University of Mas-
sachusetts, Amherst.
———. (2004) ‘A Minimal Theory of Adverbial Quantification’, in H. Kamp and B. Partee,
eds., Context-Dependence in the Analysis of Linguistic Meaning, Elsevier Ltd, Oxford,
pp. 137–75.
Greenberg, Yael. (2002) ‘Two Types of Quantificational Modalized Genericity, and the Inter-
pretation of Bare Plural and Indefinite Singular NPs’, in B. Jackson, ed., Proceedings
from SALT XII, Cornell University, CLC Publications, Ithaca, NY, pp. 104–23.
Heim, Irene. (1982) The Semantics of Definite and Indefinite Noun Phrases, PhD thesis,
University of Massachusetts, Amherst.
———. (1990) ‘E-Type Pronouns and Donkey Anaphora,’ Linguistics and Philosophy 13(2):
137–77.
Heim, Irene and Kratzer, Angelika. (1998) Semantics in Generative Grammar, Blackwell, Oxford.
Higginbotham, James. (1987) ‘Indefiniteness and Predication’, in Eric J. Reuland and Alice G. B.
ter Meulen, eds., The Representation of (In)definiteness, MIT Press, Cambridge, MA.
Kamp, Hans. (1981) ‘A Theory of Truth and Semantic Representation’, in J. Groenendijk, T.
Janssen and M. Stokhof, eds., Formal Methods in the Study of Language, Mathematical
Centre Tracts 135, Mathematisch Centrum, Amsterdam, pp. 277–322.
Krifka, M. (1999) ‘At Least Some Determiners Aren’t Determiners’, in K. Turner, ed., The
semantics/pragmatics interface from different points of view, Elsevier, pp. 257–91.
Ladusaw, William. (1982) ‘Semantic Constraints on the English Partitive Construction’, in
D. Flickinger, M. Macken and N. Wiegand, eds., Proceedings of the First West
Coast Conference on Formal Linguistics, Linguistics Department, Stanford University,
Stanford, pp. 231–42.
Lawler, John. (1973) Studies in English Generics, PhD thesis, University of Michigan, Ann
Arbor.
 Descriptions with Adverbs of Quantification 87

Leslie, Sarah-Jane. (n.d.) Generics: Cognition and Acquisition. Unpublished MS., Princeton
University.
Lewis, David. (1975) ‘Adverbs of Quantification’, in E. Keenan, ed., Formal Semantics of Natural
Language, Cambridge University Press, Cambridge, England, pp. 3–15.
Link, Godehard. (1983) ‘The Logical Analysis of Plurals and Mass Terms’, in R. Bäuerle, C.
Schwarze and A. von Stechow, eds., Meaning, Use, and Interpretation of Language,
Walter de Gruyter, Berlin, pp. 302–23.
———. (1987) ‘Generalized Quantifiers and Plurals’, in P. Gärdenfors, ed., Generalized Quanti-
fiers, Studies in Linguistics and Philosophy, D. Reidel, Dordrecht, pp. 151–80.
May, Robert. (1985) Logical Form, Linguistic Inquiry Monographs, MIT Press, Cambridge,
MA.
May, Robert and Bale, Alan. (2006) ‘Inverse Linking’, in M. Everaert and H. C. V. Riemsdijk,
eds., The Blackwell Companion to Syntax, Blackwell, Oxford.
Neale, Stephen. (1990) Descriptions, MIT Press, Cambridge, MA.
Partee, Barbara. (1987) ‘Noun Phrase Interpretation and Type-Shifting Principles’, in J. Groe-
nendijk, D. de Jongh and M. Stokhof, eds., Studies in Discourse Representation Theory
and the Theory of Generalized Quantifiers, Vol. 8 of Groningen-Amsterdam Studies in
Semantics, Foris Publications, pp. 115–43.
Roberts, Craige. (2003) ‘Uniqueness in Definite Noun Phrases,’ Linguistics and Philosophy 26:
287–350.
Russell, Bertrand. (1905) ‘On Denoting,’ Mind 14: 479–93.
Sharvy, Richard. (1980) ‘A More General Theory of Definite Descriptions,’ Philosophical Review
89(4): 607–23.
Stalnaker, Robert. (1977) ‘Complex Predicates,’ The Monist 60(3): 327–39.
Stanley, Jason. (2002) ‘Nominal Restriction’, in G. Preyer and G. Peter, eds., Logical Form and
Language, Oxford University Press, Oxford.
Stanley, Jason and Szabó, Zoltán Gendler. (2000) ‘On Quantifier Domain Restriction,’ Mind
and Language 15: 219–61.
Wilson, George. (1978) ‘On Definite and Indefinite Descriptions,’ Philosophical Review 87: 48–
76.
Zamparelli, Roberto. (1998) ‘A Theory of Kinds, Partitives and OF/Z Possessives’, in A.
Alexiadou and C. Wilder, eds., Possessors, Predicates and Movement in the Determiner
Phrase, Vol. 22 of Linguistics Today, John Benjamins, Amsterdam, pp. 259–301.