Professional Documents
Culture Documents
August 1999
Contents
1 Aspectual composition 3
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 The verb as carrier of aspectual information . . . . . . . . . . . . . . 6
1.2.1 The Slavic tradition . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Vendler’s verb classes . . . . . . . . . . . . . . . . . . . . . 10
1.3 The VP as carrier of aspectual information . . . . . . . . . . . . . . . 13
1.4 The S as carrier of aspectual information . . . . . . . . . . . . . . . . 17
1.4.1 The aspectual scheme . . . . . . . . . . . . . . . . . . . . . 17
1.4.2 Aspectual asymmetry and inner/outer aspectuality . . . . . . 18
1.4.3 The Plus-principle . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.4 The construal of aspectual classes: . . . . . . . . . . . . . . . 21
1.5 How terminative is perfectivity? . . . . . . . . . . . . . . . . . . . . 22
1.5.1 Aspectual operators . . . . . . . . . . . . . . . . . . . . . . . 22
1.5.2 Comparing Slavic and Germanic sentences. . . . . . . . . . . 22
1.5.3 Thematic glue . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.5.4 The Finnish case . . . . . . . . . . . . . . . . . . . . . . . . 31
1.6 Plurality and aspectuality . . . . . . . . . . . . . . . . . . . . . . . . 34
1.6.1 Some difficult questions . . . . . . . . . . . . . . . . . . . . 34
1.6.2 Collectivity and distributivity . . . . . . . . . . . . . . . . . 36
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3
4 CONTENTS
This is a reader for the course on Tense and Aspect in Sentence and Discourse we are
teaching at the ESSLLI Summer School at Utrecht, August 9 - 13, 1999. We would
like to point out that the text offered is a rough draft of what eventually should become
a textbook. The aim of the book will be to bridge the gap between aspectual theories
focussing on the sentential domain (the micro-perspective) and theories developing dis-
course structure (the macro-perspective). In particular, we will try to connect interval-
or index-based semantics and event-based semantics in a systematic fashion.
Chapter 2 presents some background notions on time and tense and discusses
temporal ontology as used in various theories on tense and aspect.
Chapter 3 formalizes the semantics of so-called inner aspectuality in the index-
based compositional theory and establishes a comparison with event-based ap-
proaches. The conclusion we draw is that events are a more suitable notion to
use at the macro-level.
Chapter 4 deals with aspectual phenomena in the domain of so-called outer as-
pectuality. The event-based analysis developed here also accounts for logical
operators like negation and quantification, and outer aspectuality in Romance.
Chapter 5 extends the theory developed in chapter 4 to the discourse level. The
analysis takes into account both rhetorical structure and linguistic constraints
which particular tenses impose upon the temporal structure of discourse.
As said the text offered here is provisional in the sense that we want to use it to prepare
a book manuscript. That is why invite the reader and/or the participants in the course
to send us all their comments, reactions and suggestions for improvements.
1
2 CONTENTS
Chapter 1
Aspectual composition
1.1 Introduction
In the present chapter, we shall discuss a number of aspectual notions, which are in
need of a precise characterization, our general impression being that the progress of
theory formation in the field is hampered by a persistent lack of coined notions. It is
not so much that some authors use the term telic whereas others use terminative or
perfective or bounded, rather we see that on the semantic side of the analysis there
is a lot of trouble in determining what the authors mean when they use these terms.
For example, if authors speak about a perfective verb napisal in the Russian sentence
(1.1a),
(1.1) a. On napisal pismo
He Perf-wrote a/the letter
b. On pisal pismo
He Imp-wrote a/the letter
do they mean to say that there are perfective verbs like napisat’ and imperfective verbs
like Imp-pisat’ in that language, or do they actually purport to say that the (abstract)
verb pisat’ has two forms: a perfective one with na- in front of it and an imperfective
one with an abstract prefix Imp- in front of it? But why don’t they speak of verb forms
in that case? Is it persistent sloppyness? Or is it a measured use of the terms involved?
These apparently terminological questions are contentually loaden because they
are deeply connected with semantically relevant questions. For example, some authors
explicitly say that a verb may denote an event1 , which triggers the question of whether
they mean that interpretation of the abstract verb form pisat0 is in the domain of events
or rather that it is the interpretation of the perfective form napisal or pisal. In the
latter case, however, it should be determined whether or not the perfective prefix na-
in (1.1a) really belongs to the verb rather than to the sentential structure. There are
indeed reasons to assume that there is a close connection between the presence of
na- and the interpretation of the direct object pismo. At that very moment the notion
1 E.g. Kiparsky 1998
3
4 CHAPTER 1. ASPECTUAL COMPOSITION
of ‘denotation of the verb’ becomes rather dubious because higher levels of structure
seem to be involved. Moreover, one might wonder what it means for a verb to denote
an event without taking into account its arguments.
We have given these reflections in order to demonstrate the necessity to go care-
fully through a series of clarifying steps. The point at issue in this example will return
in more detail shortly. First we want to introduce another set of seemingly terminolog-
ical problems which turn out to be connected with contentual issues. These problems
concern the notions of compositionality and coercion. The former notion is around in
the field of aspectuality since the early seventies. The idea is that in sentences like (1.1)
it is not the verb alone which should account for its aspectual nature, but rather higher
level structures such as the VP (V plus its complements) or the S (roughly the subject
NP plus the VP). Here again there is a lot of confusion. First of all, should composi-
tionality receive a Fregean colour in the sense that the interpretation of sentences like
(1.1) proceeds from bottom to top in a compositional way along the lines of current
formal semantics? Or does it simply mean that one takes (somehow) into account that
the arguments of the verb need to be involved in establishing the aspectual information
conveyed by the sentence? In that case—in particular if one allows the verb to denote
events—, one might say that the Bulgarian perfective verb napisa in (1.2a) is coerced
into an (secondary) imperfective verb napisvaše in (1.2b).
(1.2) a. Ivan napisa edno pismo
Ivan Perf-wrote one letter
Ivan wrote one letter
b. Ivan napisvaše edno pismo
Ivan Imp-Perf-wrote one letter
Ivan used to write one letter
Or one might argue that due to the fact that the tenseless sentence Marie écrire une
lettre (Mary write a letter) expresses a bounded event, there is some need to coerce this
information into something unbounded in order to achieve the right interpretation of
the French Imparfait in (1.3).
(1.3) Marie écrivait une lettre
Mary was writing/wrote a letter
In the literature one does not find any attempt to compare the two notions involved here
in a systematic way. Our purpose is to clarify this issue as much as possible in the hope
of reducing the number of misleading notions associated with the issue of determining
at which level which sort of aspectual information is expressed.
Talking about levels at which aspectual information is expressed, let us first in-
troduce the leading figure of the present book in which different aspectual levels are
represented. Figure 1.1 determines in fact its composition in the sense that we start at
the bottom and proceed to the top. On our way to the top we will discuss the prob-
lems raised and illustrated so far and we will describe the compositional machinery
available. The higher we get, the more we meet places where discourse information
becomes relevant, the leading idea being that sentence structure provides information
for discourse by means of particular choices with respect to elements bearing aspectual
1.1. INTRODUCTION 5
Tadv S
Tense S
::: S
ASP S
NPext VP0
Det N ASP VP
V NPint
Det N
ality we are going to explain in the present part, the asymmetry between the two NPs
is crucial. Connected with this point is our wish to avoid Quantifier Raising as much
as possible. In other words, we will try to interpret Figure 1.1 as directly as possible
allowing traces in the generative fashion rather than letting interpretation proceed on
the basis of a Logical Form.
As a final clarifying terminological remark we will use the term aspectuality as the
term covering the two traditional notions Aktionsart and aspect. We will argue in the
next section that one should be cautious in using these terms as they are tied up with a
suspicious distinction.
Here, the idea underlying the distinction between Aktionsart and aspect is that the
Aktionsart of the verb zvenet’ (go-off, call, ring) is such that the verb expresses a com-
pleted event sui generis, that is, taken as a lexical unit having a specific “objective”
meaning. The same applies to the transitive verb zavodit’: to wind something up is
a bounded thing outside the sentence itself. The fact that the two verbs occur imper-
fectively (or in terms mostly found in Russian grammars, that the sentence contains
two imperfective verbs) is due to the sentential environment in which they are inserted.
The absence of the perfective forms prozvenet and zavesti makes it possible to express
that something is missing in the situation which is normally there as the result of the
Aktionsart-meaning. The same applies to sentences like (1.6),
(1.6) a. V sem’ časov Ivan rešal zadaču
Lit: At seven o’clock Ivan Imp-solved the problem
‘At seven o’clock Ivan was trying to solve the problem’
b. Včera v sem’ časov Ivan rešil zadaču
Lit: At seven o’clock Ivan Perf-solved the problem
‘At seven o’clock Ivan solved the problem’
where the imperfective rešal’ in (1.6) is used to indicate an attempt to do justice to the
“real” Aktionsart of the verb, expressing an effort rather than a result. Note in passing
that the opposition between perfective and imperfective aspect is now expressed in the
form of different infixes.
A final example which seems to support the fruitfulness of the distinction at hand
is given in sentences like (1.7).
(1.7) a. Olga spala
Olga Imp-slept
‘Olga was sleeping’
b. Olga pospala
Olga Perf-slept
‘Olga slept a while’
Here the Aktionsart is supposed to be unbounded because one may sleep indefinitely,
the perfective prefix po- restricting the period during which Olga slept. Many other
examples are given in which the distinction between imperfective and perfective verbs
is justified on the basis of a lexically scheme determining the Aktionsart. Sentences like
(1.6) are said to have a verb whose Aktionsart express boundedness or completeness
and on the basis of the scheme the imperfective has a certain semantic effect. Sentences
like (1.7) start from an Aktionsart which expresses unboundedness and here the effect
of the perfective prefix is to put a restriction on it. Many other variants on this theme
can given, such as the marking of the beginning or termination of events, iteration,
uncertainty, and so forth.
At this point, we should give some antidote against this seductive picture of a sys-
tem which separates the Slavic languages from, say, Germanic and Romance languages
in a way often considered as quite fundamental if not unbridgeable. Indeed, up till now
many Slavists claim the uniqueness of the aspectual system and they are inclined to
just flatly reject any attempt to generalize over the three language groups in order to
8 CHAPTER 1. ASPECTUAL COMPOSITION
find a common basis for making the proper generalisations. This antidote consists at
the present stage in two shots, which are necessarily to see a more promising picture in
which the data discussed so far can be handled on the basis of quite different insights.
Firstly, it should be observed that at the end of the nineteenth century the notion
of verb had quite a different status from the one it has now in most theories. In the
discussions about grammar in the second half of that century, attention was primarily
focussed on word classes. Verb morphology was around, but it was mainly restricted
to the verb itself, the notion of phrase structure not being available to systematically
express any influence of the verb on what in current linguistics are called its argument
NPs. Suppose we would informlly represent the argument structure of the sentences
discussed so far in term of the well-known logical scheme (1.8) in which the Verb is
taken as a predicate taking one or more arguments.
(1.8) P (a ; : : : ; an )
1
[T]
In that case, what the Slavic tradition does, is to separate the (a1 ; : : : ; an )-part of this
scheme from the P and to focus on the binary opposition between [+T] and [–T], where
[+T] stands for the expression of telicity or terminativity or boundedness and [–T] for
the expression of atelicity, durativity or unboundedness. If [T] concerns the Aktion-
sart, nothing in the literature indicates the urge to analyze the perfective-imperfective
opposition as including the whole structure in (1.8). On the contrary, all current Russian
grammars proudly distinguish between perfective and imperfective verbs: the depen-
dency of this opposition on the contribution of the arguments is not incorporated.3
In order to understand the reason why linguists followed this course, we have to
discern the force of the lexicographic tradition, not only in the past that we are dis-
cussing but also clearly visible in the current practice of making dictionaries. What
happens if one describes the meaning of verbs like write or kill in English? For write
dictionaries give a meaning description which amounts to something like ‘to form let-
ters or words ,etc.’ but this promising beginning to reveal the real nature of the activity
is soon put aside by the more prominent meaning ‘communicating with in writing’
where mostly examples such as write a letter or write a book are given. It is this
particular habit of providing examples together with the internal argument of the verb
practically always given in its singular form with an indefinite or definite article, which
is in the heart of what we called the lexicographic tradition of word meaning. We learn
to write letter, to kill a soldier, to solve a problem, to give a book to someone, and by
doing so we are nearly convinced that to write, to kill, to solve and to give inherently
express a bounded event, sui generis, qua Aktionsart, so to say. What is forgotten is
that one is factually learning a complete VP, which is construed from different smaller
parts, one being verbal, the other being nominal. If dictionaries would systematically
have given examples like to write letters, to kill innocent victims, to solve problems, to
give books to friends, etc., one would certainly have developed a very different feeling
for the sense of the verbs involved. In comparing the two sets of examples—one with
the singular NP, the other with the bare plurals—it is possible to derive the sui generis-
sense of the verbs involved, but this may very well be quite different from what the
3 Kolni-Balozky 1960, Shvedova 1982, Maslov 1959, among many others.
1.2. THE VERB AS CARRIER OF ASPECTUAL INFORMATION 9
Slavic tradition pulled out of them by being fixed on the singular NP examples. It is
not unreasonable to require some constant meaning in kill when occurring in to kill a
soldier and to kill soldiers.
Of course, there is a clear difference between to play in play a sonata and to kill in
kill a soldier in the sense that to play seems to focus on the activity of playing whereas
to kill seems to focus on the resulting death of the soldier. However, the crucial question
is not only whether this is true or not, but also whether it is relevant. Let us consider
the plural cases to play sonatas and to kill soldiers. Both pertain to something that
can go on indefinitely. Note, however, that if one holds that to kill soldiers means that
there is one killing for each of the soldiers involved, one can equally well maintain
that to play sonatas means that there is one playing for each of the sonatas performed:
one has to stop before the next play may begin. In fact, Verkuyl 1972 analyzed the
meaning of play in terms of an underlying notion of PERFORM, expressing that a sonata
comes into being by playing it. So, arguably the obvious difference between the lexical
meanings of the verbs play and kill might be a difference between ‘construction’ (say,
in a cumulative sense) in the case of play and destruction (its reverse) in the case of kill,
which entails that the verbs have something in common: both express a causal relation
between the activity expressed by the verb and some resulting state. What they share
can be seen as involving a processing, a dynamicity.
The easiest way to grasp the point at issue is that the verbs under discussion both
express ‘a going through a predication’. Someone who played a sonata performed some
activity to bring about the result of a sonata having been played. Someone who killed a
soldier was involved in some activity to bring about the result of a soldier having been
killed. If one accepts this as the general way to characterize the contribution of verbs
like play and kill, it is easy to take the next step and to allow the verb run or walk in to
run a mile or to walk a mile to express the activity bringing about the result of a mile
having been run or walked, the result notion following from the cumulative structure
created by the ‘going on’-sense of the verb walk.
It was exactly this point which led Verkuyl 1972:95f. to observe:
Summarizing, one could say that what we called the “lexicographic approach” to word
meaning has blurred the aspectual picture in the sense that the focus of the meaning
description was put on that aspect of meaning where at least some interaction with
the (singular) internal argument NP is visible. So either one admits that Aktionsart
is no longer a lexical matter and one takes into account the nature of the arguments,
or one focusses on the genuinely inherent meaning of the verb itself apart from what
its arguments contribute in phrase structure. This can be done by perceiving that the
verbs discussed above contribute cumulative information. The natural notion of Path of
some sort of movement emanating from the quotation may be called ‘nonstative’. It is
this meaning element that can be isolated from the verb meaning which can be said to
10 CHAPTER 1. ASPECTUAL COMPOSITION
play its role in the aspectual composition along the lines to be sketched below. Before
we will do that, we continue our historical analysis of what happened in the Slavist
theories on aspectual structure by embroaching our second point
This second point is related to the former in a suppplementary way. The emphasis
on the lexicographic approach to word meaning can be said to be due to the lack of a
satisfactory morphology and syntax. To say that aspect is a matter of morphology and
semantics amounted to analyzing aspectual phenomena without any notion of phrase
structure. This was, of course, due to the fact that syntax in the modern sense was
non-existent in the first decennia of the twentieth century. The Bloomfieldian notion of
constituent structure was not yet available, let alone the development of that notion in
the second half of the twentieth century. So, there were no technical tools available for
those who noticed that it was not only the verb that determines aspectuality. Even those
who considered the verb as governing the case of its arguments, could not express that
in a clear unequivocal way. For example, Poutsma 1926:291 observes that “the normal
aspect of a verb is often modified or even utterly changed by the context”. The basic
idea of Poutsma’s analysis seems to be that every verb has a “basic aspect” and that
this aspect can be modified by the presence of constituents with which the verb occurs.
In modern terms, we see here the idea of coercion and behind it the holy fear for the
cherished tenets of Slavist aspectology.
At the time of Poutsma’s analysis the notion of VP as a phrase structural con-
stituent was not available at all, so the appeal to a coercive strategy, also visible in
Jacobsohn 1933, was not an unreasonable move. Poutsma simply followed his infor-
mal intuition about the VP as a unit and not having available a formal semantics to
combine two sister nodes into a higher level structure, he fell back upon a sort of trans-
fer mechanism changing the value of a syntactic node on the basis of the presence of
another one. But having available the VP as the phrase structure level at which infor-
mation about the V and its internal argument come together, coercion is at best only
a second best solution, the more so if one accepts the fact that the verbs kill and play
and run have something in common which plays an aspectual role independent of their
sisters.
detailed introduction.4 Briefly, what Vendler says is that if someone loves somebody
between two points of time, then this love persists for all instants between them. This
characterizes a State. If someone is walking at a certain moment, this moment is part of
a walking interval. This characterizes an Activity. If someone is running a mile at a cer-
tain moment, that moment is part of an interval in which this person ran a mile. This
characterizes an Accomplishment. Finally, if someone wins a race between two mo-
ments, the winning took place at a moment in between. This is called an Achievement.
Table 1.1 exemplifies Vendler’s quadripartition.
This quadripartition is obtained on the basis of a scheme in which two oppositions play
a role. They are given in Table 1.2.
–Process +Process
–Definite State Activity
+Definite Achievement Accomplishment
Basically there are two sets of criteria to distinguish the four classes. One can dis-
tinguish States and Activities from Achievements and Accomplishments by observing
that (1.10a) containing a State verb and an Activity verb is acceptable whereas (1.10b)
is not well-formed under a one-event interpretation. This observation is complemented
by the fact that (1.9a) is unacceptable whereas (1.9b) containing an Accomplishment
verbs is fine.
(1.9) a. He slept and she drew circles for an hour
b. #He slept and she drew circles in an hour
(1.10) a. #He ran a mile and she drew a circle for an hour
b. He ran a mile and she drew a circle in an hour
We immediately recognize here the well-known observation made at least sixty years
before in the linguistic literature and being available to the linguistic community ever
4 For a detailed critical analysis of the role of Vendler classes in aspectual theory we refer to Verkuyl 1989
since. In other words, the criterion used is derived from the opposition between sen-
tences like (1.11) and (1.12).
(1.11) a. #Ivan vypil dva piva s čas
John Perf-drank two beer for about an hour
#‘John drank two beers for about an hour’
b. Ivan pil dva piva s čas
John Imp-drank two beer for about an hour
‘John was drinking two beers for about an hour’
The single event interpretation in which Ivan drank two beers is excluded in (1.11a).
Likewise, one may say (1.12a) whereas (1.12b) is not acceptable, unless it is taken to
express a habit.
Vendler uses the following sets of sentences to distinguish between verbs that ex-
press a process and verbs that do not do that.
(1.13) a. I am knowing, she is loving him
b. She was recognizing me, he was reaching the top
He rules out sentences like (1.13) on the ground that the verbs cannot have the Pro-
gressive Form, whereas the verbs in (1.14) can. This opposition sets the States and
Achievements apart from the Activities and Accomplishments.
It is not our intention to treat Vendler here in detail.5 Here the two tables, the
criteria and the examples are given in order to refresh the memory of the readers and
this should suffice to see the relevance of the two sets of examples, because they play
a very important role in the analysis of aspectual and temporal structure. They make it
possible for us to check whether the main claim made by Vendler—that there are four
verb classes—makes sense with respect to the two criteria given here. It is important to
note that in the examples Vendler is quite sloppy in his use of the term Verb. Sometimes
it covers genuine verbs like walk and run, sometimes in pertains to phrases of which
they are the verb, such as run a mile or draw a circle.
The title of Vendler’s paper is ‘Verbs and Time’. If we take him literally in the
sense that Verb pertains to the lexical verb category, then what Vendler actually did was
5 We refer for a very critical investigation of validity of the criteria to Verkuyl 1989. Moreover, there
are also empirical problems in the sense that Vendler’s judgments about the opposition between (1.13) and
(1.14) are generally not accepted.
1.3. THE VP AS CARRIER OF ASPECTUAL INFORMATION 13
to extend the Slavic scheme (1.8) into four verb classes, with the Verb -information
separated from the (a1 ; : : : ; an )-part:
(1.15) a. VState (a1 ; : : : ; an )
b. VAct (a1 ; : : : ; an )
c. VAcc (a1 ; : : : ; an )
d. VAch (a1 ; : : : ; an )
Like Vendler, Dowty 1979 distinguishes between four verb classes in English. He took
the term ‘Verb’ literally and followed Vendler by distinguishing between four lexical
classes. Given the Montagovian joy of doing genuine semantics—that is, interpretation
by relating language structurally to domains of interpretation—the course followed by
Dowty is understandable though misleading. In fact, Dowty fell victim to an ambigu-
ity introduced by Montague himself. For linguists in the seventies, it was extremely
uncommon to get used to the fact that an intransitive verb like walk was of the same
syntactic and semantic type as walk a mile. Both are syntactically of type IV and se-
mantically of type he; ti (which means to say that the verb denotes the set of walkers
in the domain). So, rather than settling the aspectual matter at the VP-level (of the
Montagovian type he; ti), he stuck to the traditional V. This ambiguity is inherent to
Vendler’s proposal. Being a philosopher and trained in the formal logic of the time
he may have used the notion of verb in a more broad sense, namely as ‘one-place
predicate’. One could take the notion of verb in the Montagovian way as a one-place
predicate at the level of VP. In that case, however, one may no longer speak of lexical
classes. As soon as Vendler-classes are treated at the level of the VP, i.e. as VP-classes,
the question arises what explanatory force may be attributed to such a quadripartition
for the analysis of aspectual phenomena. As observed repeatedly, the aspectual oppo-
sition terminative/telic/perfective vs durative/atelic/imperfective is crucially a binary
one.
Summarizing, Vendler not having the linguistic-syntactic notion of VP available
appealed in the same way as Poutsma and Jacobsohn to an informal intuition about
the cohesion between the verb and its internal argument, but he phrased, as they did,
what he wanted to express in terms of lexical notions. By doing so, he ended up in
the very traditional position of Slavic scholars who analyzed aspect in terms of lexical
classes. And this is why the influence of Vendler on current semantic analyses turns
out to produce conceptual anomalies.
VP VP
[+T] [-T]
V NP V NP
[+A] [+B] [+A] [-B]
features [A] and [B]. Here follows a list illustrating the four logically possible
different combinations, given two sister nodes:
These possibilities were first explored in this fashion in Verkuyl 1972. It should be
underscored that it would be wrong to take the features syntactically: they represent
essentially semantic information as purported to do the generative-semantic categories
employed in Verkuyl 1972.6 Semantically the features abbreviate semantic information
that will be presented formally in chapter 3. However, in view of the fact that we have
already discussed the dynamics involved in the meaning of verbs expressing change, we
will informally explain the notion of Path in terms of some figures which will express
the main idea of the feature combination in Figure 1.3.
In order to get as closely as possible at the semantic content of the features we will
relabel the feature A as ADDTO and the feature B as the label SQA. The verbal feature
[ADDTO] differentiates between stative and nonstative verbs. Nonstative verbs are
“dynamic” in the sense that they create a predicational Path: they pertain to a going
on in time during which the predication as expressed by the sentence is developing. It
opens up the possibility of taking into account the essentially atemporal structure intro-
duced by the internal argument NP by relating it to the cumulative structure introduced
by the verb. The feature [SQA] distinguishes NPs pertaining to a Specified Quantity
of A (where A is the set of sandwiches or persons in the domain of interpretation) from
NPs which express an Unspecified Quantity of A.7 The composite feature [+T] (for
6 The formal clothing of the features involved was different due to a choice in favour of generative-
semantic categorial nodes because they were considered more semantic than features. However,
Verkuyl 1972; 1976 made clear that whatever could be expressed by categorial nodes may be expressed
by features. This has caused some misunderstanding in the literature, e.g. in Dowty, Hinrichs, Krifka.
[+
7 Krifka’s term quantized is certainly modeled on the ]
SQA -feature, but Reniers has shown that there
are some subtle differences.
1.3. THE VP AS CARRIER OF ASPECTUAL INFORMATION 15
terminative, telic) represents information about what happens if you combine temporal
information ([+ADDTO]) with atemporal information ([+SQA]), the general idea being
that a predication “Path” is formed involving a set some entity is going through.
What happens if you are “going through” a set? In terms of set theory, this is an
odd question but the idea itself is not so odd. When we interpret the VP in sentences
like (1.16),
(1.16) a. Judith ate four sandwiches
b. The three girls wrote some letters
it contains an atemporal unit ([ four sandwiches] , [ some letters] ) which in some way
is involved in the development of temporal structure introduced by the verb. So, what
happens set-theoretically if you relate an atemporal set to a linear structure? This ques-
tion is not so odd any longer, the more so while you can send a line through a set
relating its elements to it by partitioning the set, as illustrated in Figure 1.3, where (a)
(a) (b)
s s3
s1 s2 s4
NP
s-
V
s s
0 1 2
-
3
shows the situation at which the verb and its argument are going to be structurally re-
lated into the VP; and where (b) illustrates the effect of amalgamating the verbal and
nominal information into a Path structure making (1.16) true just in case Judith first ate
one sandwiches, then two and in a third step the fourth one.
The main idea to get across at this moment is that the amalgamation of the features
[+ADDTO] and [+SQA] into [+T] involves the tuning of essentially temporal informa-
tion with information that can be captured set-theoretically in the case of Count Nouns.
Atemporal sets are unordered. Temporal structure forces us “to go through” a set im-
posing linear order to unordered partitions of the NP-denotation. The story is more
complex because in sentences like (1.16) what we do offer is a set of combinatorial
possibilities rather than just one configuration, but at this stage we aim at clarifying the
notion of Path in terms of one arbitrarily chosen configuration in Figure 1.3.
In order to prepare our discussion about the Partitive case in x 1.5.4 we will add
Figure 1.4 in which we disentangle the [+ADDTO] and [+SQA] of Figure 1.3 by putting
them in a mapping relation, but in which we also show that the set of four sandwiches
is embedded in the set A which is the semantic value of the Noun sandwiches in the NP
of sentences like (1.17). Actually, Figure 1.4 could represent what is expressed both
by (1.16a) and by (1.17).
(1.17) Judith ate four of the sandwiches
Thus, part of the [+T]-information expressed by a VP includes the information that the
sense of dynamic progress contributed by the verb affects the internal argument NP
with respect to just a part of the denotation of the Noun set A.
16 CHAPTER 1. ASPECTUAL COMPOSITION
s
s s s
A
s3 s
s1 s2 4
6
6 6-
SQA
0 1 2 3
At several places in Verkuyl 1972 it was observed that people like Poutsma in the
twenties and Jacobsohn in the thirties in fact had already departed from the traditional
Slavist position by paying more than incidental attention to the interaction between the
verb and its internal argument or other complements. Verkuyl’s way to shape their in-
tuition was the line just sketched, which ended up in recognizing the VP as the “molec-
ular” level at which atomic information was fused together into an essentially higher
level sort of semantic information. As a matter of consequence he rejected the coercive
strategy on the ground that when you transfer some information from the internal ar-
gument to the V you need an operator, say along the lines of Bennett and Partee 1978,
but also of Pustejovsky 1995.
(1.18) a: Mary walked WALK Activity
b: Mary walked a mile * WALK Accomplishment
c: Mary walked miles +* WALK Activity
To make walk an Accomplishment verb you need an operator, say *, so whatever re-
sults is, say *V. But this can only mean that * WALK 6= WALK and this means that the
sense of constancy in what is expressed by the verb is lost.8
It should be observed, though, that by the use of type logic in the analysis of
linguistic categories the idea of coercion as developed in e.g. Moens 1987 is quite
attractive. Along the lines of Moens one may think of coercive rules operating at the
S-level as in sentences like (1.19).
Pr
(1.19) Mary walk a mileterm ! Mary walk a miledur
P rog
! Mary was walking a mile
Here the idea is that in natural language we need type adaptation in order to meet
the demands of categories functioning as operators. That is, if Mary walked a mile is
a terminative sentence and Prog is an operator expressing a state, then one can either
require that Prog change the terminative sentence into a stative one, or one may assume
that there is some typeshifting operator Pr which first changes the [+T] into a [-T]
after which Prog has no problem in receiving its operand. In this way, the terminative
sentence is coerced into having the right property.
8 The criticism against Poutsma in Verkuyl 1972:42f. can easily be translated into a more modern objec-
tion against Pustejovsky’ strategy because Poutsma in fact followed exactly the same course as Pustejovsky
did.
1.4. THE S AS CARRIER OF ASPECTUAL INFORMATION 17
The obvious difference between the two examples of coercive strategies is, of
course, that the first one concerns the construction of the predication expressed by a
sentence, whereas the second one is used to accommodate a sentence into an environ-
ment requiring certain properties. In our view, the proper strategy should be to apply
compositionality in the Fregean sense in those cases where it is possible and to rely on
coercion only if that would be too bothersome. If that strategy is followed, it simply
follows that coercion with respect to the data in Figure 1.2 and Table 1.3 as proposed
by Poutsma and Jacobsohn and pursued by Pustejovsky is an inferior form of compo-
sitionality, the question being whether it makes sense to send information about the
internal argument to the verb which anyhow will become available at the VP-level. A
strictly compositional approach takes the A- and B-features of Table 1.3 together and
amalgamates them into a higher level feature. As said one may not blame Poutsma and
Jacobsohn for framing their intuitions in a coercive mode: they did not know about the
VP as a syntacto-semantic unit. The same seems to apply to Vendler 1957. We will
point out that the use of coercive rules in the sense of Moens are not well-motivated in
the realm of what we will call inner aspectuality, but that they can be of use in dealing
with phenomena in the domain of outer aspectuality. The two latter notions will be
explained in more in detail in the next section.
TVP [+ ADDTO ] TS
[ Children [ a sandwich]]
TS
b. VP ate durative
[ [–SQA] [ [+SQA]]] )[ ]
S
TVP [+ ADDTO ] TS
[ Judith [ sandwiches]]
TS
c. VP ate durative
[ [+SQA] [ [– SQA]]] )[ ]
S
TVP [+ ADDTO ] TS
[ Judith [ a sandwich]]
TS
d. VP wanted durative
[ [+SQA] [ [+SQA]]] )[ ]
S
TS TVP [– ADDTO ] TS
This scheme has only an explanatory function: it turns out to be very handy to use it
for organizing the presentation of some crucial notions in the compositional theory of
aspectuality we are going to present. We will use the scheme to discuss four important
issues:
the asymmetry involved in the scheme;
9 Of course, there are differences. At the technical level there is the difference between categories and
features but as observed in the thesis itself what was expressed generative-semantically could have been
expressed interpretatively in terms of features. Furthermore, the 1972-scheme served technically as a filter
to exclude certain transformations. Contentually the 1972-filter included also the indirect object, which we
leave outside our present scope. In spite of these differences, what is expressed by (1.20) is sufficiently
present in the 1972-book to justify (1.20) as a point of departure.
18 CHAPTER 1. ASPECTUAL COMPOSITION
(1.21).10
(1.21) a. run a mile in an hour; #run a mile for an hour
b. John ran a mile in an hour; #John ran a mile for an hour
Secondly, it was a strong disagreement with Lindstedt 1984 which induced the position
that each structurally formed level has its own form of aspectuality which differs from
any other level unless it could be proven that they are identical. Lindstedt’s idea can be
easily demonstrated by the following list which applies to Finnish data but which can
be explained in terms of the corresponding English translations.
(1.22) a. Pekka read the paper Imp
b. Pekka read the paper one hour Perf
c. Pekka read the paper one hour a day Imp
d. Pekka read the paper one hour a day a week Perf
e. ... Perf
In Finnish the sentence under analysis contained the NP lehteä (the paper) with geni-
tive case. This gives (1.22a) an Imperfective interpretation: ‘Pekka read in the paper
or Pekka was reading the paper’. The idea is that a strict : : :(Perf(Imp(Perf(Imp))))-
structure is imperative, the alternative being the sequence : : :(Perf(Imp(Perf(Imp(Perf))))).
For Verkuyl 1988, this sort of analysis was the reason to sharpen the distinction be-
tween inner and outer aspectuality, his position being that aspect formation in sen-
tences in Pekka read the paper is crucially different from what is achieved by adding
adverbials and other adjuncts, due to the fact that inner aspectuality is derived from
interaction between the verb and its arguments. So, here is the outline of the position
emerging:
1. inner aspectuality is the level of argument structure;
2. the VP as part of the argument structure has its own aspectual properties which
deviates from the aspectual properties displayed at the S-level.
The first point invokes the Plus-principe that we will discuss shortly. It puts the first
position to test. The second point incorporates the view that sentences with a plural
external argument have distributive interpretations and one can perceive easily that the
distribution is sensitive for aspectuality in a structural way. That is, in the distributive
interpretation of sentences like (1.23),
(1.23) a. Three girls ate a sandwich
b. Three girls broke a glass
the telicity/terminativity expressed by the VP is to be distributed over each of the girls.
We will discuss the two points in much more detail when proceeding. The aspectual
analysis of sentences like (1.22b), (1.22c), etc. is left aside until chapter 4
10 In Verkuyl 1993:23–27 a number of similar arguments are given from which the VP emerges as an
warns against an unprincipled solution. It suggests that aspectualizers and the per-
fective prefix in this interpretation do not belong to the argument structure but rather
operate on it. Indeed in chapter 3 we will present a solution for aspectualizers and
Slavic prefixes expressing a beginning in terms of a modification on the predicational
Path.
Summarizing, we have illustrated how a feature-algebraic mechanism leads to the
position that terminativity as expressed by sentences like Judith ate three sandwiches
in (1.20a) is construed on the basis of plus-information only. The principle is restricted
to a certain area of inner aspectuality, but it protects the analysis of aspectuality against
arbitrary choices: begin and za- turn out to behave like operators on structures that
obey the Plus-principle rather than being part of them.
[-SQA ] [+SQA ]
NP
V
[– ADD TO ] [+ADD TO ]
+
(1.27) a. Judith loved sandwiches State
+ +
b. Judith wanted a sandwich State
+ +
c. Judith ate sandwiches Process
+ + +
d. Judith ate three sandwiches Event
22 CHAPTER 1. ASPECTUAL COMPOSITION
In the literature, this tripartition has been put forward, e.g. in Comrie 1976, Mourelatos 1978,
Bach 1981, among others. It runs counter to the bipartition found in Kamp and Reyle 1993,
de Swart 1995;1996 , among others. Evidently the present authors did not play on the
same side of the field. At this stage we shall not speak out on what we agree on. Here
it suffices to observe that whatever Figure 1.5 and Table 1.4 express, they offer the
possibility that it is the VP rather than the S that is said to pertain to States, Processes
or Events. So, they do not speak out on whether or not the external argument is to
be excluded. Never did Verkuyl so far, due to his reluctance to meddle in ontological
questions. In her analysis of tense, De Swart ended up in treating Processes as States.
In Chapter 4 we will pick up this theme again, observing that Figure 1.5 and Table 1.4
give away our view that it is language which determines ontological choices. To cloth
information in the form of (1.27c) indicates that one wants to speak about something
that Judith did (in principle) unboundedly, whereas by the choice of the determiner a in
(1.27d) a speaker presents the information as pertaining to a bounded event. We have
seen that the two sentences may be true in a domain of interpretation.
::: S
ASP S
NPext VP0
ASP VP
V NP/PP
ASP NPint
Det N
Secondly, sentences like (1.33) and (1.34) discussed in Młynarczyk 1998 show that
both the imperfective and the perfective forms of a verb may occur in the scope of
negation:
(1.33) Przez lata nie czytał zadnej ksiażek
-long/for years neg Perf-read-3sg no book-Gen
Sentence (1.33) means that for years it was the case that he did not read books, whereas
(1.34) expresses that for years he did not read a single book. The important thing to
note in this connection is that both (1.35a) and (1.35b) are durative.
(1.35) a. Przez lata nie czytał zadnej ksiażek
b. Przez lata nie przeczytał zadnej ksiażki
This is shown by the fact that they have the same behaviour in the For-test in (1.33)
and in the In-test in (1.36).
(1.36) a. ?W rok nie czytał zadnej ksiażek
b. ?W rok nie przeczytał ani jednej ksiażki
This is a very strong indication for the need to distinguish between terminativity and
perfectivity. However, it should be observed that this does not imply that there is no
overlap between the two notions. In other words, Germanic languages have a termi-
native VP generally without a prefix on the verb to mark this. In Slavic languages a
perfective prefix might be seen as encoding terminativity plus information concerning
definiteness or specificity, which in Germanic languages is expressed by the deter-
miner of the internal argument. One might argue that under negation the terminative
part of this information would in Slavic languages be neutralized as it is in Germanic
languages, the definiteness or specificity information being retained. We offer this sug-
gestion here as part of our strategy to accept the distinction between aspect and Aktion-
sart only when necessary and certainly not along the lines of the traditional distinction
discussed above.
A third consideration suggesting that one cannot simply equate Germanic termi-
nativity in Germanic languages with perfectivity in Slavic languages is induced by the
observation that in the absence of the determiner the NP pivo (beer) is marked as [–
SQA ]. A problem arises because the sentence (1.37b) means that John drank the beer.
The presence of vy- imposes a definite interpretation on the internal argument NP.
This point has been stressed by several authors, for example by Dahl 1984, Kabakčiev 1984,1989
and Filip 1994. For Czech Filip 1994:229 presents examples like (1.38).
(1.38) a. Pletla vetry
Imp-knitted-sg-fem sweaters-pl-Acc
‘She was knitting sweaters’
b. U-pletla vetry
Perf-knitted-sg-fem sweaters-pl-Acc
‘She knitted (all) the sweaters’
Here the perfective prefix u- contributes the sense of ”all-ness” and specificity of the
set of sweaters. Filip compares the effect of the presence of the prefix u- to the use of
the definite article in definite descriptions.
The idea that the perfective prefix in Russian and Czech restricts options made
possible by the absence of an overt determiner is confirmed by Bulgarian sentences
like (1.39), Bulgarian having a definite article suffixed to the Noun kafe as pointed out
in Guentcheva 1990.
(1.39) a. Az izpix kafe
I Perf-drank coffee
b. Az izpix kafeto
I Perf-drank coffee-the
‘I drank the coffee’
The definite article is required given the presence of the perfective prefix iz-. Bulgarian
is among the Slavic languages an important member from the point of view of aspec-
tual composition: it expresses perfectivity and imperfectivity just like the other Slavic
languages, but it is the only Slavic language having a definite article. By its presence
one can see formally what remains invisible in Russian cases like (1.37b).
The presence of vy- should be explained in a more general way, because Russian
has determiners like dva (two), tri (three), etc, nekiy (some, a certain), koy-kako’y
(some, a few) which are involved in the making of the [+ SQA]-feature information.
One option is that the tenseless (1.40)
(1.40) Ivan -pit’ dva piva
is terminative in the way of its Germanic counterpart, dva piva being [+SQA], and that
vy- is simply added to the verb stem as some sort of warrant, whereas the Imp-operator
(or, for that matter, the absence of vy-) would modify the terminative nature of (1.40) in
the same way as the Progressive Form in English in the b-sentence of (1.41) modifies
the terminative nature of its a-sentence.
(1.41) a. John was crossing the street
b. John crossed the street
We will pursue this line of thought concluding that the presence of the perfective prefix
has structurally something to do with information about the internal argument NP. This
was clearly shown by the two sentences in (1.35) which display a clear difference in
26 CHAPTER 1. ASPECTUAL COMPOSITION
interpretation: the (1.35a)-part of (1.33) simply expresses that the speaker wants to
remain neutral with respect to the situation reported by the sentence. The (1.35b)-part
of (1.34) stresses the fact that the person in question did not read a single book in spite
of the fact that he could have done that. The presence of the perfective prefix prze-
gives away the singling out of a specified quantity. This can be explained by assuming
a structural relation between the perfective prefix and the internal argument.
It should be observed at this point that it is not right to speak about the perfective
prefix in sentences like (1.37) Ivan vypil dva piva. For Czech, Filip 1994:231 points out
that the two perfective prefixes in (1.42b,c) have different effects on the interpretation
of the sentences.
(1.42) a. Pil kavu
Imp-drink-sg-masc coffee-Acc
‘He was drinking coffee’
b. Na-pil se kavy
Perf-drink-sg-masc refl coffee-Gen
‘He drank some coffee’
c. U-pil kavu
Perf-drink-sg-masc coffee-Acc
‘He drank a sip of coffee’
d. Vy-pil kavu
Perf-drink-sg-masc coffee-Acc
‘He drank all the coffee’
The use of na in (1.42b) restricts the interpretation of the internal argument NP to an
indefinite specified quantity ‘some’, whereas u- in (1.42c) imposes the more specific
interpretation of a small portion of the noun set denotation. Finally, vy- in (1.42d)
contributes the set of universal quantification and a sense of definiteness.
Two sorts of idiosyncrasies should be observed. Firstly, the use of the genitive
does not seem to bear on the absence or presence of perfective prefixes. Filip observes,
for example, that in (1.42c) the accusative kavu could be replaced by the genitive kavy
without any change of meaning. Secondly, the prefix u- in (1.42c) has clearly a different
contribution to make than in (1.38b), so one may assume that the choice of the verb
stem also determines the effect of the perfective prefix on the internal argument.
Summarizing, we may say that there are reasons to assume that the perfective and
imperfective operators can be seen as applying to the taking of the internal argument by
the verb. In this sense they overtly contribute to the making of the VP what in Germanic
languages (mostly) takes place without an appeal to morphological encoding. More-
over they contribute to the interpretation of the NP in the absence of quantificational
or referential information. So essentially, Imp and Perf may be seen as VP-operators
in Slavic languages. Before we try to characterize the semantic side of this idea, let us
first pay attention to structural matters. Where do we locate these aspectual operators?
Two proposals have been made to that effect. The first one is Verkuyl’s analysis in
(1.43).
(1.43) [Ivan [V P [V vyi [V stem -pil]][NP [Det ei ] [pivo]]]]
1.5. HOW TERMINATIVE IS PERFECTIVITY? 27
This expresses the idea that vy- in (1.37b) Ivan vypil pivo is in fact the determiner of the
NP pivo. It presumes a movement of the determiner to the prefix position. The other
position is adopted in Dimitrova-Vulchanova 1996. It assumes that iz- in the Bulgarian
sentence (1.44a) is a predicate of a small clause.
(1.44) a. Ivan izpija dve biri
Ivan Perf-drank two beers
b. [Ivan -pija[sc [dve biri][iz]]]
The first position presents some difficulties that are made visible in Figure 1.7. Figure 1.7a
VP VP [+T
VP ]
V NP 2 Perf VP [+T
VP ]
indeed gives the position of (1.43). The perfective prefix is sent from NP2 to its place
adjoined to the V-stem. But in this position it fails to account satisfactorily for the
relation between the V and the NP2 : it cannot be said to govern the process of amalga-
mation. Figure 1.7b, on the other hand, has a better structural position to offer for Perf.
It governs the lower VP in the sense that it operates on the information conveyed by its
sister node. It provides the quantificational plus-value at the upper VP-level requiring
that the NP2 be interpreted as [+SQA] in Slavic languages as Russian and Polish or
requiring it by the presence of the definite article as in the Bulgarian sentence (1.39b).
We will opt for the option provided by Figure 1.7b rejecting the option in (1.44b) on
the ground that a small clause analysis cannot explain why the Bulgarian sentence Az
izpix kafe (lit. I Perf-drank coffee) is unwellformed and the Russian sentence (1.37b)
is only well-formed in the interpretation that Ivan drank the beer. In other words, an
appeal to (1.44b) requires for these cases the stipulation that the subject of the small
clause be [+SQA]. Note that this stipulation is local in the sense that the process of
putting iz- into the right place is not governed at the level at which the verb and its
internal argument come together.
The presence of a perfective operator in a position governing the whole VP-domain
is very much compatible with the fact that in Slavic languages there are also verbs
which are considered perfective without having a perfective prefix, as in (1.45).
(1.45) a. Toj kupi edno kušta Bulgarian
‘He bought one house’
b. Dał pieć kanapek Polish
‘He gave five sandwiches’
c. On izdal pyat’ knigi Russian
‘He published five books’
28 CHAPTER 1. ASPECTUAL COMPOSITION
In all these cases the imperfective counterparts have an infix: kupuvaše, dawał, and
izdawal, respectively.
In order to get the picture clear, we will assume that the internal structure of a
VP in Slavic languages can be defined as in Figure 1.8. One drawback of this picture
VP VP
PERF VP IMP VP
V NP 2 V NP 2
is that Affix should be interpreted as the presence of a prefix, infix or suffix. So, its
structural position in front of the Vstem should be modified in this sense. For the im-
perfective operator IMP there are, in principle, two possibilities when the Verb itself
is [+ADDTO]. The first one is that the Verb combines with a [+ SQA]-NP to yield a
[+TVP ]-interpretation at the lower level VP which is to be undone by the IMP-operator.
The second one is that the Verb combines with a [ SQA]-NP to yield a [ TVP ]-
interpretation at the lower level VP which is transmitted to a [ TVP ]-interpretation
at the higher level VP by the IMP-operator. We will return to this matter later on. As
to the Perf-operator we observe that, dependent on the lexical nature of the [+ADDTO]-
Verb, the V has a perfective prefix (as vy- in vypit’) or not (as in kupit’) and that the NP
should have a [+SQA]-specification. In order to see how this is obtained, we need to go
into the internal structure of the NP.
NP
Det2 N
Spec Det1
j:j
Figure 1.9: NP structure
The leading idea is that an NP like a child in (1.46) receives its indefiniteness from
the SPEC-part of the Det2 -node and its singular information from the Det0 -part, the nu-
meral SG being interpreted as providing the information that in the set denoted by child
a singleton will be selected. By the indefiniteness feature of Spec this singleton is not
identified but rather introduced into the discourse along the lines of current Discourse
Representation Theory. In the children the collection of sets containing two or more
children is intersected with a context set in the sense of Westerståhl 1984, yielding
a definite set identified by discourse. The NP some children in (1.46c) in the well-
known sm-interpretation selects among sets of children meeting the quantificational
requirement 1. In (1.46d) the modifier next provides specific context-dependent in-
formation useful in the selection of the proper set containing three children. Note that
the at least-meaning can also be accounted for in terms of modification, as shown in
(1.46e).
We have discussed the NP-structure in some detail in order to prepare for our
discussion on Finnish case and on strategies used in Germanic languages in order to
bring about the same sort of information. In order to complete the picture we need to
embark fruitfully on issues of Finnish, we will briefly discuss the notion of thematic
role as applied to the internal argument. We will refer to this role as the int - or
simply -role. It is illustrated in Figure 1.10, where provides the information about
VP
V PP/NP
NP
Det N
Spec Det
Def j:j
Figure 1.10: -role information
30 CHAPTER 1. ASPECTUAL COMPOSITION
those in Figure 1.10. For example, the treatment of Finnish data in the next section will
suggest that it might be necessary to extend Figure 1.10. So, we simply postpone the
final evaluation of the ideas exemplified here until we have a full picture of the Slavic,
Finnish, Germanic and Romance data.
This shows that Jacobsohn’s point was correct. More recently, Kiparsky 1998 paid
attention to the Partitive case in Finnish from the point of view of its contribution to the
aspectuality of sentences. In the present section, we will discuss some of his examples
and use them to refine the picture of the internal VP-structure. We give the examples
and the glosses exactly as he gives them:
VP
V PP/NP
NP
Det N
Spec Det
Def j:j
Figure 1.11: Case information
pleteness.11 Given the proper subset specification for CaseP art the interpretation of
Ammuin karhuja brings about the existence of entities affected by the predicate. So,
I shot bears. This warrant cannot be given in the case of , which on our analysis
provides indeterminacy. Note that the definiteness or indefiniteness does not of any in-
fluence on this indeterminacy, just like the fact that we speak about the school in John
walked toward the school does not bear on the indeterminacy of what has taken place.
The present analysis also explains why Ammuin karhua with Partitive case for the
internal argument cannot mean ‘I shot a bear’ and why Ammuin kahta karhua can-
not mean ‘I shot two bears’. CaseP art operates on cardinality information by “taking
away” some part of a larger whole. But in the case of a bear one cannot partition the
cardinality 1 and taking away a part from a set of two entities does not guarantee that
one obtains a proper partition. In general, CaseP art may not affect quantificational
information expressed by Det0 because it is not sensitive to the nature of the entities
quantified over. It is neutral as to Mass or Noun.
From this analysis it follows that in cases where the verb is not [+ADDTO] the use
of the Partitive case should be okay, as in (1.56a). If CaseAcc is taken to operate on the
Det0 -information of the internal NP, then the unacceptability of (1.56b) follows too.
to three events. However, we cannot commit ourselves to the existence of such a clear-
cut distinction here, because the notion of event itself is quite obscure. May an event
spread out over three different days? May three times quantify over sub-events?
The same sort of difficult questions pop up as soon as the external argument of a
sentence is a plural NP such as given in (1.59).
(1.59) a. Three girls lifted a table
b. John and Mary walked three miles
c. Ivan i Maria napisali pismo
Ivan and Mary Perf-wrote a letter
Suppose that (1.59a) applies to a situation in which three tables were lifted. In other
words, each of the girls lifted her own table. Note that it is not necessary that they were
in the same room. They may have done what they did without having any knowledge
of each other. Do we speak about one event with three subevents? Or do we speak
about three events?
We consider these questions as irrelevant to a proper analysis of what is expressed
by the sentences in (1.59). Whatever is a proper ontological answer, at least what
should be observed is that the VP lift a table plays an important role because one may
say that each of the girls “had her own VP”. In more technical terms this amounts to
saying that each of the girls has her own predicational Path in the sense informally
introduced above. There is some interpretation in which Mary’s part in the walking is
complementary to what John walked within the total sum of three miles, but we will put
this interpretation aside here, because it will be discussed shortly in the next section.
On the interpretation that (1.59) is about six miles, both John and Mary undergo the
predication ‘walk three miles’. In the same way, Ivan and Mary both completed the
same letter or both Ivan completed a letter and Mary completed a letter, but if the
sentence applies to two letters it is excluded that Ivan completed his letter whereas
Mary did not, or reversely. So the information expressed by na- is distributed over the
two members of the external argument denotation.
We will take these facts as strong signal that a theory of aspectuality should include
a proper treatment of plural quantification. It should be noted that in sentences like
(1.60),
(1.60) Some girl lifted three tables
there is no need to distribute over the girls: the sentence is about three tables that were
lifted by some girls. Likewise in sentences such as (1.61),
(1.61) Three girls lifted four tables
each of the girls underwent the predication ‘lift three tables’, whereas it is impossible to
say of the tables that each of them is related to a different set of four girls. This shows
the asymmetry in the behaviour of the external argument and the internal argument.
In view of these observations, it is necessary to pay some attention to the distinction
between collectivity and distributivity.
36 CHAPTER 1. ASPECTUAL COMPOSITION
g1 - 4t 1 g1
- 4t q- 4t
g2 2 g2
1
g3 - 4t 3 g3
nal argument denotation (a set of individuals) and the VP-denotation (a set of factors)
, one may say that the distributive interpretations are governed by taken as an in-
jection, whereas the collective interpretations are determined by taken as a constant
function.12
We can now characterize more precisely what it means for a sentence like (1.62a)
to have a distributive interpretation. One can say that each of the girls receives “her
own VP-value”, in the sense that each of the three VP-denotations involved must be
different from one another. This requirement follows from the injectivity constraint on
. As to the collective interpretation one may say that all girls are mapped to the same
VP-value which amounts to saying that none of the participants in the external argu-
ment denotation may claim the satisfaction of whole predicate for herself. In Hitch-
cock’s movie Murder on the Orient Express the twelve participants all contributed to
the killing but their explicit purpose was to escape from the judgment they they as an
individual should be tried as in individual satisfying the predication ‘I have killed that
villain’. Neither Hans nor Uwe may claim ‘I wrote that book on DRT’ on the collective
interpretation. The functional modes offered by (1.12) seem to give exactly the right
interpretations.
As a final point of the present section, we would like to say a few words about
the function values in the case of a distributive interpretation. It is easy to see that a
12 The name has been chosen to evoke the notion of participancy
38 CHAPTER 1. ASPECTUAL COMPOSITION
sentence like (1.62) is true if it turns out to be the case that in the domain of discourse
there were only eight tables available for the lifting. It follows straightforwardly from
this restriction on the cardinality of the set of tables that some of them must have been
lifted twice. However, for (1.62) a table lifted twice counts as two lifted tables. Suppose
that table t1 was lifted by Mary at index i1 and t1 was lifted by Jane at index i2 , where
we stipulate that indices with different values can never be actualized simultaneously
in real time (even if there would be an overlap between the intervals involved). And
suppose that the following configuration would hold for the situation described to make
(1.62) a true sentence:
(1.67) a. Mary 7! fhi ; t i; hi ; t i; hi ; t ; ig
1 1 2 2 3 13
b. Jane 7 fhi ; t i; hi ; t i; hi ; t ; ig
! 2 1 3 2 4 14
c. Judith 7 fhi ; t ; ; ; ig
! 3 5678
This configuration covers a situation in which one table was lifted four times. Yet it is
possible to say that twelve tables were lifted in the obvious sense that there were twelve
table liftings. Note that (1.67) satisfies the injectivity constraint on : each of the girls
has a different function value, a different predicational Path.
It is obvious that if t1 in hi1 ; t1 i pertains to a singleton set, and t5;6;7;8 in hi3 ; t5;6;7;8 i
to a set of four members, all pairs in (1.67) make up functions, because function are
defined as sets of pairs fulfilling the proper conditions. In other words, given the fact
that the proper conditions for functionhood are fulfilled, one may write (1.67) in a more
general fashion:
(1.68) a. Mary 7! f 1
b. Jane 7 f
! 2
c. Judith 7 f
! 3
This is another way of saying that each of the distributed VP-factors can be taken as a
function assigned as value of to each of members of the external argument denotation.
In this way, the dependency of tables on indices can be modelled in a standard fashion
on the basis of the empirical fact that sentences like (1.62a) Three girls lifted four tables
may pertain to just eight object tables whereas each of the three girls may claim that
she lifted four tables. The distributive multiplications in (1.62) can be seen as yielding
index-dependent values.
Our discussion about distributivity and collectivity is not a deviation from the as-
pectual road we are following in the present chapter. On the contrary, it leads through
the centre of the domain which is under investigation in the present book: the domain
of aspectuality and temporality. In particular, the issue of the token-dependency raised
here is very important because what happens in the making of a VP out of a temporal
part (the verb) and an atemporal part (the internal argument) is the origin of a mapping
between them: atemporal entities are “drawn” into temporality. The functional format
seems to be a proper tool for accounting for this match.
1.6. PLURALITY AND ASPECTUALITY 39
sentences in (1.71)
(1.71) a. Tri momičeta izjadoxa pet sandvica
Three girls Perf-eat-Aor. five sandwiches
Three girls ate all together five sandwiches
b. Tri momičeta izjadoxa po pet sandvica (vsjaka)
Three girls Perf-eat-Aor. times five sandwiches (each)
Three girls ate all together five sandwiches
The collective interpretation can be fully understood in terms of -constant: none of
the girls may claim that she ate five sandwiches. The distributive reading is marked by
the presence of the distributive particle po (which occurs in the multiplication 2 po 3 =
6) and (for some) in the presence of the floating quantifier vsjaka. The thing to note,
however, is that in this interpretation for each of the girls the information expressed
by the perfective prefix iz- applies. In other words, the phenomena of collectivity
and distributivity so closely connected with the presence of plural NPs is relevant for
Slavic aspectuality too. As in the Western literature too much attention has been paid to
sentences in which Mary or John were involved in some bounded or unbounded event.
It seems time to open up the realm of plurality and to extend the domain of aspectuality
so as to include it.
1.7 Conclusion
The general picture offered in this chapter is that it is possible to proceed from the
bottom to the top of Figure 1.1. As pointed out in x 1.5 the VP structure is more
complex than shown in that figure which would provide us with a sequence of aspec-
tual operators: ASP - ASP - ASP - CaseAcc=P art - . This looks formidable, in
fact a little bit too much. But as suggested above, this sequence is available to us as
analysts. Languages select some of them to provide themselves with the appropriate
means to express aspectuality. Finnish may not need ASP and French may skip the
CaseAcc=P art - interaction because it uses, say,ASP - ASP . Our aim was to show
that if languages are investigated as to compare how they express aspectuality we need
to have the whole sequence available. Some parameters came up: commitment to the
whole or partial truth, completeness, indeterminacy and incompleteness. It also be-
came clear that there are different aspectual layers. It is important to see that all this
has been investigated abstracting away from tense. The VP and the S as carrier of as-
pectual information are tenseless semantic objects. So, there is some need to see what
tense contributes.
Chapter 2
2.1 Introduction
2.2 Subjective and objective time
Since Antiquity two images of time have been discussed: the flow of the river, and
the line made up of stationary points. On the one hand, time is linked to motion, i.e.
changes in the world, and on the other hand, time can be conceived as a stationary order
of events represented by numbers. In order to model these two pictures, MacTaggart
(1908) distinguishes between the A-series and B-series conceptions of time. The A-
series conception is based on the notion of past, present, and future. The B-series
is based on a simultaneity and earlier-later calculus defined over objectively existing
instants. The B-series embodies the ‘objective’ view on time.
It is independent of a point of view, and just gives a temporal ordering of events
by the precedence relation. If some event is ordered as preceding another, this relation
will never change as time goes by. So this type of temporal ordering is permanent; it
is not related to the point of speech, and it is consequently non-deictic. The A-series
embodies the ‘subjective’ (or intersubjective) view on time.
If we order propositions with the help of past, present or future, we introduce their
relation to some deictic center, the point of speech. Depending on the speech time, the
truth value of a proposition may change as time goes on. This then captures the flow of
time, since the present will become past, i.e. flow into the past. Both the static and the
dynamic view of time are part of our conception of time. Lakoff and Johnson (1987)
draw attention to metaphors built on the tension between static and dynamic views of
time. Metaphors are either built on the idea that time is a moving object (1) or time is
stationary and we move through it (2):
(1) a. Coming up in the weeks ahead . . .
b. The time has long since gone when
41
42 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
These metaphors are concerned with our interaction with the flow of time, but they
present different views of how we do that.
Temporal logic is mainly concerned with the A-series. Language is concerned
with both: tenses are typically concerned with the A-series, but many time adverbials
have to do with the B-series as well. Moreover, there is the interpretation of tensed
sentences in narrative discourse, where the construction of an ordering of events is at
issue.
The idea that the expression of tense in natural language is essentially deictic is
best demonstrated by showing that a statement couched in A-concepts cannot be re-
formulated into a statement in terms of B-concepts—or its truth conditions defined by
B-concepts. Take for instance the A-statement in (3a):
(3) a. It will rain in London
b. It rains in London at a moment of time after 1.51 pm on the 10th of November
in the year 1994
Suppose (3a) is uttered at 1.51 pm on the 10th of November in the year 1994. Ac-
cording to B-theorists, this statement is completely equivalent with the B-statement in
(3b). It should be noted that it is only possible to perform this reduction if the time of
utterance is known. For this reason, the suggested procedure is not perfect. Indeed, it is
not truly possible to express past, present and future in terms of the B-theory, without
some ‘hidden’ reference to the speech time.
Another problem for a ‘pure’ B-theory is the asymmetry between the past and
the future. In fact, one would have to claim that the future is just as real as the past,
and statements about the future are true or false today, exactly in the same manner
as statements about the past. This view makes it impossible to consider the idea that
it is impossible to change the past, but that different options are open to us for the
future. It is a fundamental part of our experience of reality that for instance today, one
can choose to travel to Los Angeles tomorrow, and one can also choose to stay home.
Furthermore, it is clearly also a part of our experience that one today cannot choose to
travel to Copenhagen yesterday. If you were not in Copenhagen yesterday, you have
lost any possibility you may have had of going there yesterday. If our experience of
freedom of choice as well as the passing of time are not illusions, we need a theory of
time which has certain properties of the A-series, not just the B-series.
confuse things, it is important to start by distinguishing time from tense. Tense is the
linguistic device which is used to express time relations.
Comrie (1985) provides an account of tense from the viewpoint of language uni-
versals and linguistic typology. This allows him to establish the range of variation that
is found across languages in tense, and what the limits are to that variation. Comrie
assumes that all cultures have a conceptualisation of time, even though not every lan-
guage has tense. When a language is traditionally described as ‘tenseless’, it does not
follow that there is no conceptualisation of past, present and future. According to Com-
rie, it would be equally logical to assert that speakers of languages lacking grammatical
gender categories have a radically different concept of sex from speakers of languages
with such grammatical categories. Not every culture seems to conceptualise progress,
though, so the picture is more ‘static’, and does not appeal to the notion of change. The
notion of change seems to be strongly present in modern Western culture.
All languages have ways of locating in time, but there are different classes of ex-
pressions. One class is the set of lexical items in a language that express location in
time. The information they express can bear on the A-series ( now, today, yesterday or
on the B-series (on May 13, 1998). But many languages express temporal information
by means of verbal morphology. Thus English, for instance, has grammaticalised ex-
pressions of location in time, which are realized on the verb or in the verbal complex:
present, past, future, pluperfect, future perfect.
Comrie assumes that tense is the grammaticalised expression of location in time.
In most languages that have tense, tense is indicated on the verb, either by the verb
morpholgy (English), or by grammatical words adjacent to the verb, as with certain
auxiliaries (Bamileke-Dschang). Example:
(4) a. à kè tán’ń [Bamileke-Dschang]
‘He bargained yesterday’
b. à lè tán’ń
‘He bargained some days ago’
c. à lè lá?ń’tán
‘He bargained a long time ago’
Tense is thus primarily a category of the verb. Noun phrase arguments of a verb are
often outside the scope of the tense, whereas the verb is necessarily within the scope of
the tense, e.g.:
(5) a. By 1999, every graduate student will have met a prime minister
b. All the fugitives are back in jail now
Escaped prisoners can no longer be described as fugitives when they are back in jail.
Similarly, (5a) can be true even if none of the people involved in the meeting are grad-
uate students of prime ministers at the time of the meeting.
Although tense is primarily a category of the verb or the sentence, one occasionally
finds tense expressed elsewhere, or with a different domain. In certain American Indian
languages, tense can be shown on noun phrases, thus distinguishing the ‘entity that was
an X’ from the ‘entity that is an X’, as in:
44 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
Tense describes the way in which events, processes and states relate to the time
axis. At least, this is what the basic meaning of tense does. Often, a grammatical cat-
egory has several secondary meanings, derived from the basic meaning. For instance,
the past tense in English is used in counterfactuals, so it does not refer to the past, but
to a situation not anchored in reality:
(7) If I were rich, I would buy a nice house
The past tense is also often used to indicate politeness. It is more polite to say (8b) than
(8a):
(8) a. I want to ask you something
b. I wanted to ask you something
Similarly, the future has modal uses, because of the close ties between future and
modality (see below):
(9) Will you do this for me?
= Are you willing to do this for me
There is also a distinction to be made between the core meaning of the tense, and
certain implicatures which can be derived from it. For instance, the past tense asserts
that a situation held at some time prior to the present moment. This often comes with
the implicature that the situation no longer holds, as in (10a):
(10) a. John was in the garden
b. John was in the garden when I came in, and for all I know, he still is
Grice’s maxim of relevance suggests that a statement reported in the past tense is not
about the present moment, otherwise the speaker would have used the present tense.
Thus the use of a form explicitly locating a situation in the past suggests that that
situation does not hold at the present. That this is an implicature, and not part of the
meaning of the past tense is shown by (10b), where the implicature is cancelled (or
suspended).
Similarly, there is a feeling that events reported in the present perfect are ‘closer’
to the speech time than events reported in the simple past. This implicature can arise
because the present perfect is defined in terms of current relevance. Naturally, recent
events typically have more current relevance than remote events, which gives rise to
the implicature. That it is not part of the meaning of the simple past to describe a
remote past is illustrated by (11a). That the present perfect can relate to an event which
happened a while ago is illustrated by (11b):
(11) a. John broke his leg five minutes ago
b. John has broken his leg. It happened six weeks ago, but it hasn’t healed yet.
2.4. PAST, PRESENT AND FUTURE 45
In similar ways, the pluperfect can come with the implicature that the event happened
in a remote past, because it happened before some other event. However, if that event
is close enough, there need not be a large distance:
(12) Susan came back a minute ago. But Bill had already left
This is not to say that languages cannot develop in such a way that these implicatures
become part of the meaning of the tense. In modern Eastern Armenian, an original
pluperfect has been reinterpreted as a remote past. The same is true for certain varieties
of English spoken in areas of Africa.
Comrie uses the traditional terms ‘absolute’ and ‘relative’ tense to refer to tenses
which do and which don’t take the present moment as their deictic centre. The three
basic absolute tenses are the present, the past and the future. The basic meaning of these
tenses is: present tense means coincidence of the time of the situation and the present
moment. Past tense means location of the situation prior to the present moment. Future
tense means location of the situation after the present moment.
The future exists now. (p. 101, immediately before Yu Tsun kills Albert)
The present tense is essentially deictic: it refers to the moment of speech of the utter-
ance situation. Just like ‘I’ always picks out the speaker, and cannot refer to anyone
else in the utterance situation, the reference of the present tense and of ‘now’ is fixed
by their reference to the speech time. But the flow of time means that the speech time
shifts from one day to the next, or from one utterance to the next, or even from one
word to the next. If we picture the moment of speech as being somewhere on the time
axis, this can only be an abstraction. There are typically two images for the relation
between the ‘eternal now’ and the flow of time. Either ‘now’ is like a person standing
in the river: the person has a fixed position, that the water in the river flows by. Or
the ‘now’ is like a particle of water in the river, and it flows by static objects such as
calendar terms. Both images show the tension between the static and the dynamic view
of time (the A-series and the B-series), and all the examples above play with this idea.
Although the present tense is described as locating the situation at the speech point,
it is relatively rare for a situation to coincide exactly with the present moment, i.e.
to occupy literally a single point in time which is exactly the same as the present mo-
ment. One set of examples is performative sentences:
(14) a. I promise to leave tomorrow
b. I bet you ten dollars that the Sharks will win
Although these situations are not strictly momentaneous, because it takes a certain pe-
riod of time to utter even the shortest sentence, they can be conceptualised as momen-
taneous, especially in so far as the time occupied by the report is exactly the same as
the time occupied by the act. Another set of examples involves the so-called reportive
use of the the simple present:
(15) John gets the ball, Bill takes over, and suddenly Max scores!
The report is simultaneous with the situation described, so there is literal coincidence
between the time location of a situation and the present moment. Hoewever, situations
2.4. PAST, PRESENT AND FUTURE 47
of this kind are relatively rare, and the more normal uses of the present tense, in lan-
guages where there is a separate grammatical category, go far beyond this restricted
range.
A more characteristic use of the present tense is in referring to situations which
occupy a much longer period of time than the present moment, but which nonetheless
include the present moment within them. In particular, the present tense is used to
speak of states and processes which hold at the present moment, but which began before
the present moment and may well continue beyond it as in:
(16) a. Sue is a student at Stanford
b. Mary likes fish
In each of these examples it is indeed true that the situation holds at the present moment,
i.e. at the speech time. But it is not the case that the situation is restricted only to
the present moment. As far as the present tense is concerned, in its basic meaning it
invariably locates a situation at the present moment, and says nothing beyond that. In
particular, it does not say that the same situation does not continue beyond the present
moment, nor that it did not hold in the past. Whether or not the situation is part of a
larger situation extending into the past or the future is an implicature, rather than part of
the meaning of the present tense. This implicatures is worked out on the basis of other
features of the structure of the sentence, and one’s knowledge of the real world. Of
other relevant features of sentence structure, aspect will be one of the most important
in deciding whether the larger situation is restricted to the present moment or not:
thus the use of the progressive or the habitual necessarily requires that the situation in
question be not momentaneous. Similarly, time adverbials may express the duration of
the larger situation:
(17) Since 1994, Sue is a student at Stanford
But this does not change the general definition of the present tense itself.
The truth of (18a) can be tested by seeing whether it does in fact rain tomorrow or
not. The truth value of (18b) cannot be assessed by observing whether or not it rains
tomorrow, since both presence and absence of rain are compatible with a modal possi-
bility operator. It is thus possible to have future time reference which is not necessarily
modal.
The clearest cases of languages having a future tense are languages which distin-
guish different degrees of remoteness in the future, i.e. have near and distant futures. In
Haya or Bamileke-Ngyemboon, for instance, the remote future tense can only be used
in referring to a situation which is located temporally subsequent to tomorrow, and
collocation with any other time reference (past, present, tomorrow) is ungrammatical.
One linguistic problem with studying the future tense is that most European lan-
guages have a clear grammatical distinction between past and non-past (the latter sub-
suming present and future time reference), but either no grammatical distinction or a
much less clear grammatical distinction between future and non-future, in particular
between future and present. Diachronically, most auxiliaries involved in the explicit
future tense construction are derived from modal expressions, e.g. of desirativity, such
as English will. This makes it hard to distinguish tense from modality. In many Eu-
ropean langauges, the so-called present tense is in fact the normal verb form used to
indicate future time reference, as for instance in German (19a), which is at least as
common as the specific future tense construction in (19b):
(19) a. Ich gehe morgen
I go tomorrow
b. Ich werde morgen gehen
I will go tomorrow
In English, the use of the present tense to refer to the future is restricted to ‘scheduled’
events. Thus, (20a) is felicitous, (20b) is not:
(20) a. The train departs at five o’clock tomorrow morning
b. It rains tomorrow
Our knowledge of the world as it is indicates that rain is not schedulable the way the
departure of trains is scheduled.
In some other languages, there is no specific future tense, but the conceptual dif-
ference between past and future leads to a system in which the tense reference is depen-
dent on the modal system. The future is subsumed under the irrealis, and the present
and the past under the realis. Dyirbal and Burmese are languages of this kind.
The question arises whether statements like (21a) and (b) can be said to be true (or
false) already today. Alternatively, is the truth value of a statement undetermined,
such that it cannot be said to have any actual truth value today? The answers to these
questions bear upon the interpretation of modality. For if we assume that (21a) is
true today, is the statement then not also necessary today? And further, if it turns out
that there is no sea-fight tomorrow, can (21a) be possible today? On grounds of his
basic assumption of indeterminism, Aristotle claimed that neither statement could be
necessary today. However, the same does not apply to statements about the past or the
present; they are either necessarily true or necessarily false. Aristotle is apparently a
‘past-determinist’ and a ‘present-determinist’, but a ‘future-indeterminist’.
The logicians of the Middle Ages in general took the Aristotelian view that a
statement can change its truth value with time. A proposition such as ‘Socrates runs’
is not true at all times. The truth value depends on the actual state of affairs. The idea
of ‘the truth of a proposition at a given time’ thus comes into the picture. The truth
value of the proposition was regarded as relative to the time at which it was put forth—
its ‘moment of utterance’. The problems of the future were also discussed. Buridan
(XIIIth century) presents a paradox of self-reference in relation to tense. Imagine the
following scenario: Socrates wants to cross a river and comes to a bridge guarded by
Plato, who says:
(22) Plato: Socrates, if in the first proposition which you utter, you speak the truth, I
will permit you to cross. But surely, if you speak falsely, I shall throw you into
the water.
Socrates: You will throw me into the water
Plato must admit that he cannot keep his promise. Buridan maintained that the sophism
when uttered by Socrates has a truth value, i.e. it is either true or false. It is how-
ever, “not determinately true or determinately false”. This means that we cannot know
whether it is true or false, until we have seen how Plato acts when Socrates is crossing.
Buridan thus holds the view that a statement about the contingent future is true or false,
although its truth value cannot be known by anybody now.
For the Medieval logicians, the problems related to the future were most intimately
connected to the relation between two fundamental Christian dogmas: human freedom
and God’s omniscience. God’s omniscience is assumed to also comprise knowledge
of the future choices to be made by men. But if God already now knows the decision
I will make tomorrow, then an inevitable truth about my choice tomorrow is already
given now! Hence, there seems to be no basis for the claim that I have a free choice, a
conclusion which violates the dogma of human freedom. So the argument goes from
divine foreknowledge to necessity of the future, and from there to the conclusion that
there can be no real human freedom of choice. Several solutions were proposed, which
roughly argue that divine knowledge is different from human knowledge. God relates
in the same way to all times. Time should be regarded as a system in which the basic
relations of succession and simultaneity are given in a timeless way—owing to the fact
that time is given to God in a timeless way. The conceptual difference between past,
present, and future is relevant only when humans are invoved, either as the subjects of
cognition, or as participants in communication.
50 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
A modern version of Anselm and Ockham’s ideas is worked out by Leibniz. Leib-
niz’s central idea is that God has chosen the best of all possible worlds, and made it
actual. But in actualising the creatures of that world, He did not change their free
natures. So it is not necessary for a man to do that which he will in fact be doing
according to the foreknowledge of God. It would have been possible for him to make
different decisions, leading to different acts. For Leibniz there is a difference in modal-
ity between the past and the future. For while it is not possible to cause a past event,
it is now possible to cause some of the future events. While there is no alternative to
the actual past, there are alternatives to the future. These alternatives correspond to the
concept of possible world.
The view of time as branching towards the future has some interesting conse-
quences for the interpretation of future tenses like (23a) and (b):
(23a) has a trivial reading which just claims that man is not immortal. It can also
have a strong “predicting” kind of reading, which may still be in the faraway future,
though. (23b) does not have the trivial reading. Moreover, it strongly suggests that
the dying is near and somehow unavoidable: we must have evidence from context
and situation that this is about to happen, otherwise we could not felicitously utter the
sentence. In (23a), the dying might be a mere possibility we are considering, in (23b)
it is not.
What is necessary is that which holds in all possible futures, and what is possible is
that which holds in at least one possible future. A future event is necessary if and only
if it is unchangeably caused by present or past events. Among the possible contingent
futures, there must be one which has a special status, simply because it corresponds to
the actual course of events in the future. This accounts for the differentation between
‘tomorrow’, ‘possibly tomorrow’, and ‘necessarily tomorrow’. The branching future
suggests an identification of the present with reality: what happens, happens now. But
according to McCawley (1993) for instance, we need to distinguish between actuality
and non-actuality in the future as well. The difference in natural language between
‘will’ and ‘possibly will’ suggests that the modal expression relates to a choice among
alternative future histories, whereas will relates to future time within a given future
history. Thus modality introduces branching, and future time refers to a later moment
on a particular branch. This means that a sentence like (24a) is different from (24b):
(24) a. Bill will finish his novel
b. Bill may finish his novel
c. Bill must finish his novel
d. George said in June that Mary would finish her thesis by January, but he was
wrong
The meaning of (24a) is clearly not that there is some time branch containing a future
point at which Bill finishes his novel, because this is the meaning of (24b). (24a) does
also not mean that on all branches leading into the future from the present there is a
point at which Bill finishes his novel. This is the meaning of the necessity statement
2.5. THEORIES OF TENSE 51
in (24c). What (24a) says is that at some point in the actual future, Bill will finish
his novel. This leaves it open whether there are other possible but nonactual futures
in which Bill does not finish the novel. McCawley agrees that the notion of actual fu-
ture is tricky because one normally has very little conception of which of the infinitely
many possible futures is the actual one. This is quite unlike one’s knowledge of the
past. Nonetheless, McCawley stresses the fact that speakers of natural languages “in-
dulge in the rashness of making statements that purport to describe the actual future”
(McCawley 1993: 433). That such statements as (24a) are interpreted as referring to
the actual future is confirmed by the kind of examples in (24c). If one can apply words
like right, wrong, true and false in talking about previous statements about the future,
this indicates that one can talk about how the actual future in the past turned out to be.
In the case of (24c), the actual future with respect to the evaluation time in June was
not what George said it would be.
post-future Cc after-future
future Cb future
ante-future Ca before-future
present B present
post-preterit Ac after-past
preterit Ab past
ante-preterit Aa before-past
2.5.2 Reichenbach
Reichenbach (1947) suggested that in order to understand how tenses work, we must
consider not only the time of utterance, and the time of the event in question, but also
52 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
For Reichenbach, the relation between S and R, and the relation between R and E are
fundamental. The different combinations bring us up to 9 fundamental tenses. English
does not distinguish between the Posterior Present and the Simple Future. Reichen-
bach suggests that, in French, the construction with aller ‘to go’ grammaticalizes the
Posterior Present, and the Futur Simple the Simple Future. Thus, ‘Je vais partir’ has
the structure S,R–E, and ‘Je partirai’ has the structure ‘S–R,E. Reichenbach points out
that the position of E relative to S is not important. This explains the three different
schemata associated with the Posterior Past and the Anterior Future. Comrie (1985)
proposes a slightly different notation which captures exactly that. He proposes to use
complex relations like E relevative R relative S, which are to be read as ‘E relative to R
and R relative to S’. This leads to representations like the following:
2.5. THEORIES OF TENSE 53
The Perfect of the Future is constrained by the fact that E must take place before R, but
R must be located after S. Whether E ends up being located before, simultaneous with
or after S is irrelevant, and the formula makes no claim about it. This modification is
an elegant generalization of the Reichenbachian schema.
The idea of these systems is to create a ‘present in the past’ around which is developed
the same system of tenses as in the present. So each ‘kernel’ has a future and a past. The
secondary reference points have only a past, not a future. The asymmetry between past
and future is thus present at two crucial levels in the system. Systems like these come
closest to the expression of tenses in natural languages such as English, French and
German. They can be considered as fine-tuning of the general Jespersen-Reichenbach
line of research.
Prior introduces two operators P for past, and F for future. These temporal operators
are defined over propositions. There is no operator for the present tense. A proposition
p is interpreted with respect to the given time of evaulations. The operators P and F
are interpreted by the following conditions:
(29) a. t j= P p if and only if 9t0 [t0 t ^ t0 j= p]
b. t j= F p if and only if 9t0 [t t0 ^ t0 j= p]
Two more operators can be defined in terms of P and F , namely H ‘has always been’,
and G ‘is always going to be’:
(30) a. Hp = :P :p
b. Gp = :F :p
P and F look suspiciously like the modal operator 3, whereas H and G are similar to
2. Indeed, the version of tense logic developed by Prior and others can be considered
as a variety of modal logic, with temporal relations playing the role of “alternativeness”
relations: instead of looking at alternative possible worlds, we look at alternative times
at which the proposition can come out true.
Moreover, tense operators can be freely iterated, and thus yield more complex
tenses. For instance:
(31) a. John left
Pp
b. John had left
PPp
c. John would leave
PFp
d. John would have left
PFPp
e. John will leave
Fp
f. John will have left
FPp
g. John has always loved Mary
Hp
h. John would always love Mary
P Gp
Prior develops a number of different tense logics, depending on the axioms of the
system. The most basic tense logic is Kt , defined by the following set of axioms:
(32) A1 p, where p is a tautology of the propositional calculus
56 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
see that all tenses are in fact derived from present tense statements. E.g. ‘John left’ is
represented as P p, which means that the statement ‘John leaves’ is true at some point
in time in the past. The only way to properly interpret this is to view this as a reportive
use of the Simple Present. But we already pointed out earlier that this is not the most
straightforward interpretation of the Simple Present. However, the use of the Simple
Past in ‘John left’ does capture the basic use of the tense. This means that the interpre-
tation of the basic use of the Simple Past is formulated in terms of an unnatural use of
the Simple Present. This is not very intuitive.
One way of giving the deictic nature of tense more content is to introduce a special
operator for the ‘now’ of the sentence. Kamp (1971) introduces the N -operator, which
gets the following interpretation for t0 the speech time:
(37) t j= Np if and only if t0 j= p
This interpretation guarantees the equivalence of (38a) and (b):
(38) a. John is asleep
b. John is asleep now
More importantly, the N-operator captures the deictic character of the present tense by
allowing us to go back to the speech time in embedded contexts. Consider an example
like (39):
(39) Someday you will be grateful for what I am doing now
The now in (39) refers to the time when the sentence is uttered. If we wouldn’t have
the N -operator, the sentence would be represented as in (40a):
(40) a. F (p ^ q)
b. F (p ^ Nq)
(40a) gives us the interpretation that ‘you are grateful’ and ‘what I am doing now’ are
both true at some future time t0 . The introduction of the N -operator in (40b) allows us
to relate the time of what I am doing to the speech time, rather than the time at which
the being grateful is true.
The N -operator also accounts for the difference between (41b) and (c), as repre-
sented in (42):
(41) a. A child is born who will rule the world
b. A child was born which would rule the world
c. A child was born which will rule the world
of the N -operator as bringing us back to the utterance time of the sentence, the F -
operator in (42b) determines some moment in the future which we now have, even
though it occurs within the scope of P .
Kamp carries the formal properties of indexical operators one step further and
introduces a system of double indexing. The idea is to interpret a sentence with respect
to two moments in time, say t and t0 . The first is called the moment of utterance, and
the second is called the moment of evaluation. We begin to evaluate a statement made
at time t with respect to this time t. But the evaluation process may lead us to consider
moments in time other than t. An example: in order to determine whether P ' is true
in a given model at time t, we must try and find a moment earlier than t at which '
is true. Now if ' contains an expression like now which necessarily refers back to the
original t, then we will need some way of keeping track of this t. We can do this by
always considering two moments in time, one of which will allow us to ‘remember’
the speech time.
One can carry this reasoning one step further (Vlach, 1973) with examples like
(43b) as the past tense variant of (43a):
(43) a. One day, all persons alive now will be dead
b. One day, all people alive then would be dead
The Priorian recipe for obtaining the past tense consists in prefixing the P-operator to
the formula for the present tense sentence. This would lead to (44a):
(44) a. P F 8x (N Alive(x)! Dead(x))
b. P KF 8x (N Alive(x) ! Dead(x))
(44a) gets the intended reading wrong, because ‘then’ should refer to some point of
reference defined by the sentence (or context), rather than to the point of speech. In
order to account for such complications, Vlach introduces another operator K (‘then’),
which resets the point of speech, equating it with the current point of evaluation. For-
mally:
(45) hs; ti j= K' iff ht; ti j= '
Thus, K acts as a substitution operator replacing the point of speech t0 by the current
reference point t. The correct rendering of (43b) is then as is (44b).
Kamp carries the formal properties of indexical operators one step further and
introduces a system of double indexing. The idea is to interpret a sentence with respect
to two moments in time, say t and t0 . The first is called the moment of utterance, and
the second is called the moment of evaluation. We begin to evaluate a statement made
at time t with respect to this time t. But the evaluation process may lead us to consider
moments in time other than t. An example: in order to determine whether P ' is true
in a given model at time t, we must try and find a moment earlier than t at which '
is true. Now if ' contains an expression like now which necessarily refers back to the
original t, then we will need some way of keeping track of this t. We can do this by
always considering two moments in time, one of which will allow us to ‘remember’
the speech time. We will not go into the formal details here. Kuhn (1979) discusses
some examples of tense logical systems involving double indexing.
2.5. THEORIES OF TENSE 59
But once you get the idea, you can go further. A sentence like (46) is still inex-
pressible in systems that include N and K :
(46) There will always jokes be told that were told at a certain time in the past
9t < t0 8t0 > t0 9x (Tell(x; t0 ) ^ Tell(x; t)
The relevant reading is that there is a time t preceding the speech time t0 such that for
all times t0 later than t0 there is a joke which will be told at t0 which was also told
at t. This leads to a proliferation of operators and reference times (e.g. Åqvist and
Guenthner, 1977).
Examples like these show that, when the complexity of natural language exam-
ples increases, one might be better off with a quantificational account of tense (similar
to predicate logic formulas), than a treatment of tense as operators on propositional
formulas. We will not go into the formal details here. Kuhn (1979) discusses some
examples of tense logical systems involving double indexing.
Prior was very concerned with the issue of expressibility. Kamp (1968) has demon-
strated that not every temporal operator can be expressed in Prior’s system. Consider a
sentence like (47):
(47) Bill has been watching little Alice ever since Mary left
This sentence is true at time t iff there is some time t0 preceding t such that (i) the
proposition that Mary leaves is true at t and (ii) the proposition that Bill is watching
little Alice is true at all times from t0 onwards, up to (and including) t. Sentences like
these cannot be expressed in a Priorean tense logic where the structure of time has
certain natural properties (like being like the real numbers). Kamp shows that every
temporal operator in a linear, dense, non-ending instant-logic can be defined in terms
of his two operators U and S . U ('; ) is like until and is read “ it will be the case that
until it will be the case that '” (or “ it was the case that until it was the case that
'”). S ('; ) is like since, and is read “ it has been the case that since it was the case
that '” (or “ it will have been the case that since it was the case that '”. The official
semantics is given in (48):
(48) a. t j= U ('; ) if and only if 9t00 [t t00 ^ t00 j= ' ^ 8t0 [[t t0 t00 ] !
t0 j= ^ t0 j= :']]
b. t j= S ('; ) if and only if 9t00 [t t00 ^ t00 j= ' ^ 8t0 [[t00 t0 t] !
t0 j= ^ t0 j= :']]
Although the system with S and U is more powerful than classical Priorean tense logic,
it is no more satisfactory when it comes to providing natural representations of tensed
sentences from languages such as English. As Kamp and Reyle (1993) point out, the
possibility of iterating tense operators in formulas like P P ', F P ' has no counterpart
in natural language. Spelling out such formulas in natural language quickly becomes
awkward, because English has no straightforward way of iterating tenses. A tense is
always associated with a verb, and tenses cannot just be piled on top of the other. Even
the complex tenses (such as the Pluperfect) can at best provide paraphrases for tense
logical formulas with two, or at most three layers of tense operators.
60 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
An even more important observation is that English does not only use operators,
it also refers to times explicitly with the help of nouns such as time, moment, and time
adverbials. A theory of temporal reference ought to explain how the tenses contribute
to the meaning of complex statements like the ones in (49):
(49) a. Alvin rang on Sunday
b. Alvin will ring on Sunday
The phrase on Sunday has the property that it refers to the nearest Sunday—in the future
or in the past—to some given, contextually salient time. In particular, in a context in
which no times have yet been made salient, the phrase will refer either to the next
Sunday after the day on which the utterance takes place, or to the last Sunday before
it. Given that the past tense of rang in (50a) indicates that the described event lies
before the time of utterance, the token of on Sunday must refer to the last Sunday
before the utterance time. Given the future tense in (50b), reference is made to the
first Sunday after the speech time. A theory of temporal reference in natural language
needs to combine the information provided by tense and time adverbials to determine
the temporal structure of sentences like (49). Priorean tense logic does not provide the
tools to do so. Consider maybe a simpler example like (50):
(50) John left yesterday
a. Yesterday[P[John leave]]
b. P[yesterday[John leave]]
If we treat the time adverbial and the tense separately, we have to decide which one
to give scope over the other. But both scopal configuations in (50) fail to capture the
meaning of the English sentence. (50a) claims that John’s leaving takes place in the
past of yesterday. So John may have left two days ago, or a month, or a year ago. (50b)
claims that it is true in the past that John’s leaving takes place yesterday. The informa-
tion provided by the tense operator is entirely redundant here, although it wasn’t so in
(49a). What we want is of course a combined action of the tense and the time adverbial,
such that the event takes place at (or within) the time period described by the time ad-
verbial. This is almost impossible to realize within Priorian tense logic, because there
is no way to refer to the time at which the proposition is true. In a Reichenbachian sys-
tem, one can of course exploit the notion of reference time in describing the semantics
of time adverbials.
Partee (1973) is one of the first to point out certain analogies between tenses and
pronouns in English, which suggest that we need a referential approach to natural lan-
guage tenses, rather than an interpretation in terms of operators. She argues that the
three main uses of pronouns, namely deictic, anaphoric and bound variable interpreta-
tions are found with tenses as well. The deictic use of the past tense morpheme appears
in a sentence like (51):
(51) I didn’t turn off the stove
When uttered, for instance, halfway down the turnpike, such a sentence clearly does not
mean either that there exists some time in the past at which I did not turn off the stove,
2.5. THEORIES OF TENSE 61
or that there exists no time in the past at which I turned off the stove. The sentence
clearly refers to a particular time, but it is not an instant. Most likely, it is a definite
interval, whose identity is generally clear from the extra-linguistic context, just as the
identity of she in a sentence with a deictic use of a pronoun.
Anaphoric uses of tenses are illustrated with sentences like (52):
(52) a. Sheila had a party last Friday, and Sam got drunk
b. When Susan walked in, Peter left
The time adverbial or temporal clause provides a descriptive specification of the time
of the main clause. Both main clauses in (52a) refer to the same time variable.
Bound variable interpretations of tenses are illustrated with examples like (53):
(53) a. If Susan comes in, John will leave immediately
b. Richard always gave assignments that were due the next day
c. Mary teaches every other day
Partee analyzes (53a) as ‘For all times t at which Susan comes in, John leaves in the
immediate future of t. The kind of sentence in (53b) is closely parallel to sentences
involving a universal quantifier like all.
Reference to times also plays a role in the phenomenon of discourse anaphora.
Compare the two-sentence texts in (54a) and (b):
(54) a. Last week Fred bought his ninth cat. He paid 75 ECU for it
b. Bill left the house at a quarter past five. He took a taxi to the station and
caught the first train to Bognor.
There is a very strong tendency to take the second sentence of (54a) as talking about the
same time as the first. However, the second sentence of (54b) is naturally understood
as describing two events that directly follow the one described in the first sentence.
The temporal relations between sentences brought together in a narrative discourse is
an important feature of the use of tense in natural language. Clearly, there is no way to
express such temporal anaphora in tense logic, because such a system does not capture
the notion of reference.
tion operators.
62 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
stage, this makes Priorean tense logic a less suitable system for the study of tense ex-
pressed by natural language. For example, an utterance of (55a) does not mean that
John ran at some completely unspecified past time, but that he ran at some particular,
contextually determined past time:
(55) a. John ran
b. P (John run)
c. P (i ^ John run)
The natural representation of this sentence in Priorean tense logic in (55b) fails to
mirror the referential nature of the English original. Blackburn’s solution is to make
Priorean tense logic referential by adding a new type of propositional symbol, called
nominals to the language. Nominals are typically written as i, j , k , etc. The point
of introducing nominals lies in their interpretation: we shall insist that in any model,
nominals are to be true at exactly one time. Nominals thus ‘name’ the unique time at
which they are true. The sentence in (55a) is now translated as in (55c). In order for
(55c) to be true, John must run at the past time picked out by i. This gives us a ‘definite’
time at which the situation takes place. Nominals thus mirror Reichenbach’s notion of
reference point. The table given above for Reichenbach’s logic, can now be extended
with one more column, in which the same tenses are described in nominal tense logic:
Blackburn sums up these entries by saying that he has factored the English into a ‘shift’
and a ‘refer’ component. The shift component is Prior’s contribution, and is performed
by the tense operator. The refer component is Reichenbach’s contribution, and is per-
formed by the nominals. Note in particular that, unlike in Prior’s system, the simple
past and the present perfect are represented differently. This does not yet solve all the
problems related to the interpretation of the present perfect, but it captures the idea that
the simple past is ‘definite’ in nature, whereas the present perfect is ‘indefinite’. Sen-
tences which require more than one point of reference can be represented by formulas
containing more than one nominal, e.g:
2.6. TEMPORAL ONTOLOGY 63
(59i) captures the directionality of time. It implies irreflexivity, which says that there
are not ‘stops’ in the flow of time: it keeps continuing. The flow of time is transitive,
as spelled out in (59ii). The combination of (i) and (ii) avoids ‘loops’ in time, with
structures like t1 < t2 < : : : tn < t1 : we clearly do not want time to be circular. The
linearity condition in (59iii) makes the order complete, that is, no instant is outside the
flow of time. If we thing of the flow of time as a river, it makes the time line stick to a
single (river)bed.
One can argue about the question whether time has a beginning (60i) or an end
(60ii):
(60) (i) 9x :9y y < x
(ii) 9x :9y x < y
(iii) 8x 9y y < x and
8x 9y x < y SUCCESSOR
The infinite structure induced by the successor relation in (60iii) is usually preferred
from a logical point of view.
Depending on whether one wants to conceptualize time as dense or discrete, one
can add a statement which creates infinite divisibility (61i) or step-wise succession
(61ii):
(61) (i) 8x 8y (x < y ! 9z x < z < y) DENSITY
(ii) 8x 9z (x < z ^ :9u x < u < z )) and
8x 9z (z < x ^ :9u z < u < x)) DISCRETENESS
Typical models associated with dense temporal structures are the rationals Q, or the
real numbers R. For the discrete temporal structures, the structure of the integers Z is
the most natural choice.
The choice of the model determines the validity of certain inference patterns. Con-
sider (62):
(62) P' ! PP'
In a dense time structure, this is a valid axiom, in a discrete time structure it is not.
Suppose we assume that P ' is true. Then there is a time t preceding the speech time
t0 at which ' is true. But in a dense model there is always a time t0 in between t
and t0 . This means that P P ' is true as well: there is a time in the past (namely t0 )
at which P ' is true. this means that we can only use the iterated temporal operator
P P as an interpretation of the Pluperfect if we assume a discrete time structure. After
all, we do not want the Simple Past to entail the Pluperfect. There is no doubt that
natural language refers to discrete orders of time: counting minutes or hours or days
is just that. But in fact, it is much more common to think of tense as having a dense,
than a discrete structure. After all, the flow of time is continuous, and there are no
‘gaps’ between times. In natural language, density is expressed by allowing ever more
fine-grained expressions: yesterday afternoon between twelve and two, I was sleeping;
I woke up when you called me; in fact, at the moment that the telephone first rang, I
2.6. TEMPORAL ONTOLOGY 65
woke up; in fact, I woke up at the moment the telephone first started ringing, etc. In
principle we can go on refining indefinitely: every moment that we reach can be split
into smaller moments by linguistic means. This means that we cannot expect there to
be minimal elements, and that we have to think of the tense structure as dense.
2.6.2 Intervals
Linguists and philosopher object against the use of durationless instants in a theory of
time and tense for various reasons. One observation is that even the strongest phrases
of our ordinary language refer to some (small) period of time (see the examples in 1).
Similarly, there is no punctual present. We can conceptualize the ‘now’ as an indivisi-
ble moment separating the past from the future, but if ‘right now’ refers to a period of
immediate awareness, it has to have duration, because the phenomenon of awareness
itself implies duration. Aside from the issue whether durationless instants are referred
to by natural language at all, we realize that instants cannot describe all natural lan-
guage phenomena. One of the problems which arises concerns the interpretation of
time adverbials which refer to longer periods of time. In examples like (63) we can
extend from instants to intervals in a trivial way, by claiming that the sentence is true
if and only if the state of raining holds at every instant of the interval or the leaving
occurs at some instant in the interval:
(63) a. It was raining yesterday afternoon
b. Jane left yesterday morning
If it was raining all morning, it does not make any difference to the truth value of (63a)
which instant we pick from the stretch of time denoted by yesterday morning. This
does not work for (64b): we cannot assume that the time adverbial covers only the
instant at which we locate the leaving. We interpret the sentence as describing John’s
leaving at some point within the period of time referred to by yesterday morning. This
means that we need to describe the denotation of the time
adverbial in terms of intervals rather than instants, but we can describe the mean-
ing of the proposition in terms of truth at instants within a certain interval. This use of
intervals does not really add any expressive force, because it can be reduced to instants.
There are also quite a few predicates the semantics of which cannot be described
in terms of instants. Examples are given in (64):
(64) a. The apples ripened
b. Susan and Bill waltzed at the party
c. Mary ran
The problem which such sentences raise is that it takes time for apples to ripen, it takes
three steps to be waltzing, and if we take a picture of Mary when she is running, we
won’t see the motion. Evaluating sentences describing actions, changes, and events in
terms of instants is inadequate, because they are simply not true at ‘frozen’ moments
in time. Rather than a static picture, they require a film (i.e. a succession of related
pictures) to capture the change over time. Sentences like the ones in (64) can be said
to be true at intervals, which cover longer periods of time. Intervals are defined as
(convex) sets of instants:
66 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
2.6.3 Events
So we see that interval semantics remains essentially static. The problem of the analy-
sis of change is an important motivation for the introduction of events in our ontology.
The intuition that the world is a dynamic system of changing processes then requires
68 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
the introduction of events as the basic constituents of time. We can locate a moment
where the door was open before the moment where the door is closed, because we
observed or are able to postulate an event of change leading from the state where the
door was open to the state where the door was closed. In fact, our awareness of time
passing by is closely related to things happening. This leads to the view, defended by
Davidson (1967), Kamp (1981) and others, that events are the real primitives of tempo-
ral structures. This view laid the ground for the developments of event semantics. The
definition of overlap, inclusion, and precedence relations between events in an event
structure can be used as the basis from which interval structure and instant structure
are derived. In this view, we take the dynamic changes to exist in their own right, and
we do not try to reduce them to a succession of stages.
Events are entities that inherently have duration. In this sense we cannot do with-
out a temporal ordering even if we can do without points, because duration is a temporal
concept. But we are able to take events and their temporal relations as primitive, and
construct the temporal entities, periods and moments, as abstractions out of them. The
approach of taking events as basic and constructing periods and points out of them goes
back to Russell, Wiener and Whitehead, and has been revived in modern semantics by
Hans Kamp (see in particular, Kamp 1979). We come to periods by looking at events
that are cotemporal. We can order a period before another by pointing at an event that
went on at one period, but no longer at the other.
Kamp’s event structures have precedence and overlap as primitive relations:
(71) An event structure E is a triple hE; <; Æi, where E is a set of events, < is a
strict partial order, and Æ is a reflexive symmetric relation such that:
(i) ! : (e < e )
e1 < e2 2 1
(iii) eÆe
(iv) e Æe ! e Æe
1 2 2 1
(v) e < e ! : (e Æ e )
1 2 2 1
points out, we may want to put more structure on events, for instance a relation of
(constituting) part-of. For example, we have the feeling that the invasion of Normandy
is not just temporally included in the Second World war, but is a constituting part of
it. We cannot use v for that, so to express such relations, we have to introduce a new
partial order. The definition of constituting part-of relations takes us deeper into the
domain of part-whole relations in general, see for instance Moltmann (1997) for more
discussion.
The introduction of events as ‘hidden’ arguments of predicates follows the tradi-
tion of Davidson (1967) and others. The advantage of treating events like individuals
is that we have a better way of handling the referential natural of tenses. The most im-
portant conceptual problem here is the problem of event identity: what is it that makes
up the event? When are two events identical? What parts does an event have? etc. All
this depends on what singles it out as an event, a conceptual scheme. How a conceptual
scheme singles out events is a mysterious matter, but for simplicity we can assume that
it does this by assigning descriptive content to events. Changes are primitives, so an
event can now be defined as reporting a change of a certain kind. (see Landman p.
229–233 for details).
Intervals and instants are defined on the basis of events. An instant of the event
structure E is defined as a maximal set of pairwise overlapping events in the following
way:
(73) Let E be an event structure. Then:
(i) i is an instant of E if:
(a) i E
(b) e1 ; e2 2 i ) e1 Æ e2
(c) if H E and i H , and for all e1 ; e2 2 H it is the case that e Æ e ,
1 2
then H i
(ii) An event e occurs at an instant i iff e 2 i
(iii) For all instants i1 ; i2 : i1 <i i2 iff there are e1 2i
1 and e2 2i
2 such that
e1 < e2
The moral of the story is that event semantics derives temporal structure from event
structure rather than the other way around. An example of a well-worked out theory
of time and events in this perspective is Kamp and Reyle’s implementation of event
semantics in Discourse Representation theory. One of the interesting notions they de-
velop is a discourse semantics of times and events (see chapter 5 below).
70 CHAPTER 2. TIME, TENSE AND TEMPORAL REFERENCE
Chapter 3
3.1 Introduction
This chapter consists of a selection of handout-like annotations, fragments taken out of
papers and material written for the course. It contains some of the formal machinery
of the PLUG+ -grammar of Verkuyl 1993 in order to make clear how a compositional
semantics may proceed formally. The PLUG+ -grammar involves higher order type
logic that we shall not discuss here in detail. What we shall do is to present properties
of the model structure to which the interpretation function operating on the presumed
logical language pertains. This turns out to facilitate the understanding of the basic
principles of the compositional system.
One important feature of PLUG+ is that it takes the VP as a central structural unit.
The external argument relates semantically to VP-information. This is because the VP:
is an intermediate station in aspectual composition;
is the predication Path;
is a factor in a multiplication;
reduces ambiguity;
plays a role in anaphoric reference;
is restricted by aspectualizers;
relates to tense.
Why asymmetry?
write a book , eat an apple, etc.
71
72 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
eat three apples in an hour, #eat apples in an hour; ?lift tables in a minute; lift
four tables in une minute.
Ivan procital poeziju = ‘Ivan read the poetry’ (bounded interpretation; Ivan cital
poeziju = ‘Ivan read from the poetry’; non-bounded.
idiomatic expressions.
VP[TV P ]
Utrecht CS Duivendrecht Amstel Amsterdam CS
Important is the idea of interaction between two number systems. Visible in Fig-
ure 3.2. Train schedule: stations in IN, lines in IR. Also in Figure 3.3: we “live” in
IR but use IN to speak about it, the 4th of July, week 32, etc. Only if necessary, we
month
3k n
*1 2 Y3 4
week
day
1 2 3 4 5 6 7
break open its IN-nature. Discretization is part of our way to use language. Odometer.
Rounding-off.
Indices are introduced as part of the information expressed by a [+ADD TO]-verb.
(3.2) : : : iY:[ lift] (i)(Y )(X )
Type-logically, lift is here of type hi; hhe; ti; tiii, the type of functions taking an index
and yielding a collection of sets.
y1 y2 y3
Q= t1 t3
t4
t2
the function on the union of the domain so as to be able to pick out individuals. The discussion about the
type-logic involved is irrelevant to the present argument, so to facilitate the exposition we will omit set braces
where possible.
3.2. THE INDEX-BASED APPROACH 75
In (3.5) there are twelve (3 4) tables provided that e.g. table as dependent
S on index
2
i differs from table as dependent on index k. If one takes the union i2I Ran(`xi ),
2
there are seven object tables.
Definition 3.2.4 X is distributive iff for all X; Y D:
(i) if X 2 X & Y X; then Y 2 X;
(ii) if X 2 X & Y 2 X; then (X [ Y ) 2 X
As to constant, the basic idea is that on this mode the individual Paths of the mem-
bers of the external NP denotation are blurred out: there is no individual Path satisfying
the predication. This may be called the kolkhoz-collective or totalizinginterpretation
of sentences like (3.1c). This interpretation is made true by several combinatorial pos-
sibilities, among which (3.6a) and (3.6b).
The obvious Orient Express interpretation for (3.7a) is that all passengers are mapped
to the same individual, the purpose of the collective murder being that a judge could
never pick out one of the twelve passengers for him or her to apply the kill-predicate
to. The -injective interpretation would have twelve individuals having been killed.
(3.7) a. The twelve passengers killed a horrible villain
b. Hans and Uwe wrote an introductory book about DRT
s
- -
`x1 s
- -
`xn
I I
6 DL :::
I I
6 DL
[[NPext ]] x1 ::: xn
c. g1 7! fh1; ft gig
1 d. g1 7! fh1; ft gig
1
g2 7! fh2; ft gig
1 g2 7! fh1; ft gig
1
g3 7! fh3; ft gig
1 g3 7! fh1; ft gig
1
(3.13)
fill
Dom( ) Ran( )
=
`fill
x
Dom(`) Ran(`)
jj pass
Dom( ) Ran( )
p1 7! i 7! s1 7! fhi 1; test p ig
1 1
p1 7 ! i+1 7 ! s2 7 ! fhi; test p ig
2 1
p2 7! j 7! s1 7! fhj 1; test p ig
2 2
p2 7 ! j 7 ! s3 7 ! fhj 1; test p ig
2 2
p2 7 ! j+1 7 ! s4 7 ! fhj; test p ig
2 2
p2 7 ! j+2 7 ! s5 7 ! fhj + 1; test p ig
2 2
p3 7! k 7! s1 7! fhk 1; test p ig
3 3
p3 7 ! k 7 ! s2 7 ! fhk 1; test p ig
3 3
p3 7 ! k 7 ! s3 7 ! fhk 1; test p ig
3 3
p3 7 ! k 7 ! s4 7 ! fhk 1; test p ig
3 3
In (3.1a) Mary lifted four tables = <, because of its Past Tense. When (3.16) applies
to it, it yields (3.17):
(3.17) 9I 9I [Mary lift four tables(I ) ^ I = Ent (I ) ^ I < i ]
V
+
V V
This says that there is a set of indices I (in the naturals) and an interval IV in the Reals
such that I is associated with Mary’s lifting of four tables and I represents the stretch
of intervals IV and this stretch is before the point of speech i .3 This means that the
actualization of the abstract I -structure into a real time event takes place by Tense.
The Progressive Form P ROG applies before INFL and is syntactically defined as
taking a tenseless S to form a tenseless S: S0 = PROG(S).4
(3.18) [ PROG] = SJ 0 9J [S (J ) ^ J 0 J ]
For (3.1a) application of PROG to the tenseless [Mary lift four tables] yields (3.19b),
after which INFL is applied, yielding (3.19c).
(3.19) a. Mary was lifting four tables
b. SJ 0 9J [S (J ) ^ J 0 J ](Mary lift four tables)
= J 0 9J [Mary lift four tables(J ) ^ J 0 J ]
c. 9I 9IV 9J [Mary lift four tables(J ) ^ I J
^ I = Ent+ (IV ) ^ IV < i]
This says that the Mary-lift-four-tables-predication associated with J can only be war-
ranted as to its truth as far as its proper subset I is concerned. That is, only the I -part
of J is given as actualized in the reals and Tense commits itself only for I and not for
its complement in J . The present analysis is a non-modal treatment of the Progressive
Form as far as the application of a rounding-off function is non-modal. It assumes that
a tenseless S contains the information that supplies a full Path. Tense takes only part
of it as the Progressive demands.
The correspondence between the Progressive Form in English and the imperfective
form in Slavic can be given a structural basis on the assumption that the Progressive
Form in English sentences like (3.20) and (3.21) operates on terminative information.
(3.20) a. Judith was eating those sandwiches
b. Prog-Judith eat those sandwiches
[ ] ( )( )([[
2 The sentential structure of (3.1a) is (roughly): I: lift I i ] )([[
four tables ])
Mary , where I , the
h i
collection of i’s associated with the verb, is of type i; t . So, this makes Tense as taking an S of type
i; t ; t to form an S0 of type t. For details: Verkuyl 1993:318–327.
hh i i
3 Ent+ is defined as a rounding-off function from the Reals into the Naturals. I= Ent+ IV says that I
( )
is the set of representatives from the intervals in IV .
hhh i i hh i ii
4 This means that it is of type i; t ; t ; i; t ; t , the modifying type.
3.3. APPROACHING OUTER ASPECTUALITY 79
In other words, (3.20a) and (3.21a) can be analyzed in terms of the presence of an
operator Prog modifying the information given in the tenseless phrases of (3.20b) and
(3.21b). There are many proposals in the literature on the Progressive Form involving
a very complex semantics, most of them based on modality.5 We will restrict ourselves
here to a more technical solution by assuming that the use of a Progressive operator
gives away the commitment of a speaker to the truth of what is in the scope of Prog. In
other words, given the terminative nature of the tenseless Judith eat some sandwiches
on the basis of the Plus-Principle, and given the fact that the truth can be determined
in terms of an interval (the running time of the event involved), the Prog-operator in
English to warrant the truth for just a proper subinterval. That is, for (3.20) to be
true, Judith may not have finished eating those sandwiches; in the same way, the truth
of (3.21) presumes that John did not yet had crossed the street. If Prog contributes
the subinterval Isub and I the interval of the terminative phrases, then one can say,
informally, that Prog requires Isub I . We will see Prog as an instance of an aspectual
operator ASP with = . If we put Prog on the place of PERF and IMP in Figure 1.8,
then we could make the following list:
(3.22) a. PERF = ASP =
b. IMP = ASP
c. PROG = ASP
These appear to be the three logically possible relations between the ASP -interval and
the I -interval of the tenseless phrase in the scope of the operator
Extending the scope of the analysis so as to include Romance languages we would
like to point out that it is possible to characterize some of the French tenses in these
terms. There is some debate possible and also going on about the question of whether
or not aspectual operators are in fact tense operators in these languages. We will return
to that point in detail in chapter ??. Here we can point out the possibility of analyzing
the Passé Simple in sentences like (3.23) along these lines.
(3.23) a. Marie écrivit une lettre
b. [ASP= [Marie-écrire-une-lettre]]
Sentence (3.23) is to be understood as expressing that Mary indeed produced a letter
by writing it. This is left in the dark in (3.24a).
(3.24) a. Marie écrivait une lettre
b. [ASP [Marie-écrire-une-lettre]]
This sentence can be understood as expressing that Mary did what she did in (3.23a)
but also as expressing the well-known in medias res of the Slavic Imperfective. Let
us say that (3.24a) is ambiguous as the meaning of . We would prefer to call this
underdetermination: (3.24a) simply does not want to give any commitment to either
of the two possibilities: and =. The term ‘commitment’ refers back to the point
5 Dowty, Asher & Lascarides, Landman, etc.,etc.
80 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
we made above, namely that the operators under analysis can be conceived of a ex-
pressing a commitment of the speaker. This fits the Slavic description of the difference
between Imperfective and Perfective aspect in which many scholars have observed that
the Imperfective aspect presents the situation medias in res, as opposed to the Perfec-
tive aspect by which a situation is presented as completed. As said there are different
opinions about the precise interaction between tense operators and aspectual operators
as well as about the criteria which decide whether some operator is in the temporal
domain or in the aspectual domain, but we shall not discuss them at this stage be-
cause formal machinery is necessary to have the discussion at the appropriate level of
preciseness.
3.3.2 Aspectualizers
In the past thirty years a lot of attention has been paid to the analysis of aspectualizers:
Newmeyer 1969, Perlmutter 1970, Freed 1979, among many others who based their
analyses on these studies. Recently, the attention for these verbs or rather to one of them
(begin) has been triggered by Pustejovsky 1995 and earlier work, Godard and Jayez 1993,
Pustejovsky and Bouillon 1995 who discuss the so-called coercion phenomena in sen-
tences like (3.25).
(3.25) a. Mary began a book
b. John finished his book
We will show now how aspectualizers are treated in the framework of Verkuyl 1993.
From this certain conclusions with respect to sentences like (3.25) may be drawn. So,
we will concentrate on aspectualizers occurring with to V and V-ing and try to solve
the problem how the information they contribute bears on the structure of `.
From a compositional point of view, the following question is relevant: why is
it that sentences like (3.26a)—when interpreted without the durational adverbials—
are durative and sentences like (3.26b) and (3.26c) are terminative —as shown by the
forced repetition in case they are combined with for hours, whereas the aspectual fea-
ture algebra of chapter 1 predicts that no [+T] may arise given a [–T] at the VP-level?
(3.26) a. Mary walked (for hours)
b. Mary began to walk (# for hours)
c. Mary began walking (# for hours)
One could solve this problem by simply stipulating that unergatives like walk escape
from this mechanism, but this would not help for sentences like (3.27).
(3.27) a. (For hours) Mary lifted tables
b. (# For hours) Mary began to lift tables
c. (# For hours) Mary began lifting tables
In these cases the well-known forced repetition occurs, which means that one has to
provide for a unified account of the terminativity expressed by b- and c- sentences in
(3.26) and (3.27). At this point it should be underlined that the notion of inchoativity
which is often associated with the sentences under discussion is aspectually without
3.3. APPROACHING OUTER ASPECTUALITY 81
This says that the application of yields a truncated Path ` because ` is restricted to
the first index. The restriction to index 1 is rather arbitrary because in sentences like
Mary began to lift hundred tables one might be more lenient as to the decision after
how many tables the use of begin is inappropriate. We will return to that point shortly.
At this stage, the point is that the restricted function `m ji1 yields just a finite number
of cells of the partition required by the VP-predicate. This explains the terminativity
of the sentences in (3.29).
(3.29) a. Mary began to lift tables
b. Mary began to walk
Sentence (3.29a) pertains to a bounded part of an otherwise unrestricted Path. The same
applies to (3.29b). Unergative verbs like walk can be analyzed as having a deficient
Path structure. The internal argument domain is empty. One may solve this problem
by modeling the unergative Path structure after the transitive Path structure discussed
in the preceding section. Let this co-domain of ` be the collection containing the empty
set and assume that all indices produced by the [+ADD TO]-verb walk are mapped onto
;. In this case, begin picks out the pair hi1 ; ;i, and this gives `ji1 a finite cardinality,
just as predicted.
This can be illustrated as in Figure 3.7
6 A restriction of a function f : A ! B , to a set C with C A is the set fhx; yi : hx; yi 2 f and
x 2 C g. The restriction is denoted by f jC .
82 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
begin
[ NP ] : p0 p1 p2 p3 :::
66 - 6 6
`Mary :
[V ]: 0 i i+1 i+2 :::
Aspectualizers like begin display a sort of neutrality as to their input: they slice blindly.
That is, if they receive a terminative VP the result will be terminative as well:
(3.30) a. Mary began to lift four tables
b. Mary began to walk a mile
In both cases, the restriction cuts out the first-index structure of a bounded Path, as
begin is not interested in the remainder of the Path. As soon as the restriction has been
applied, the process comes to a stop.
The point made for begin can be extended to other aspectualizers, but not for keep
and continue. These do not slice out pairs from `. Rather they come close to enhance
the working of the successor operation. We come back to them below in section ??.
The aspectualizers that qualify for being analyzed as restricting a function by selecting
a specific index or specific indices in the Path function ` are given in (3.31).
8 begin, start
< 1k<m<n
(x) = `jik ; s.t. for end, stop, quit, resume 1 < k = m < n
(3.31)
: finish, complete 1<m<kn
In general, aspectualizers modify the onset, the nucleus and the coda to use Freed’s
tripartition of the interval making up the Path. The aspectualizers in the second group
differ in many respects, but what they have in common is that they “slice out” a part of `
in the way indicated. It is also important that the value n contributed by the cardinality
information of the internal argument of the VP-predicate be not reached. Freed 1979
points out that verbs like end, quit, stop, resume, cease, etc. express the suspension
of an action or event and presuppose the prior occurrence or suspension of the action
involved. This distinguishes these verbs from verbs that indicate that the end of the Path
has been reached, like finish and complete and from verbs that indicate the beginning
of it like begin and start. In this sense, (3.31) provides for a correct account of the
interpretive data as far as restrictions on ` are concerned.
Returning to sentences like (1.26) Miša zakuril (sigaretu) (Miša started to smoke
(a cigarette)) in chapter 1, we propose that the prefix za- is an aspectualizer in the sense
just described. It operates on the VP and provides a restriction on Miša’s Path.
brought about by clicking the large cube, selecting it and moving it to the left. Now,
with Figure 3.10 on the screen, Denis comes in, is puzzled by the movement on the
screen and asks us what sort of game we are playing. Henriette’s answer is (3.33).
(3.33) I put the large cube to the left : : :
Sentence (3.33) is terminative: it expresses something bounded. The well-known as-
pectual tests as in (3.34) confirm this.
(3.34) a: #For hours I put the large cube to the left
b: I put the large cube to the left in a minute
If she had answered the question with (3.34a), Denis would have taken it to mean that
she was caught in the act of repeatedly putting that cube to the left side of the quan-
drangled base and he would probably have adviced us to get away from the computer
immediately.
84 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
But now suppose that Henriette said (3.34a) with Figure 3.8 on the screen of the
computer. Denis knew that we had the task of organizing all the world documents in
the Tarski folder in a certain way preparing them for the examination of some students,
so to put the large cube to the left in world documents having a large cube is a feasible
thing to do. With an eye on Figure 3.8 Denis would not be disturbed about our mental
health as he was in the first case. This is because in (3.34a) its part (3.33) is interpreted
duratively. How come?
The fascinating thing about this problem is that in both cases we use the same
language material. It is evident that each of the worlds in the folder has its own time
structure, so in order to bring about changes one has to open the world document,
which means that there are as many time structures as there are world documents. Our
going through all the documents is durative due to the plurality we are going through.
However, there is a third situation which makes the problem at issue even more fasci-
nating. At this moment, the general Macintosh operating system called OS 8.6 is not
so far developed as to allow us to operate in documents without opening them. For
instance, if one has written and preserved seven articles in which Jespersen is spelled
as Jesperson —as linguists regularly do—, then one cannot use a Find and Change
command replacing Jesperson by Jespersen in all documents at once. Each document
has to be opened separately because, as said, each document has a unique structure,
in particular a unique individual history. Suppose now that there is a system, say OS
12.0 enabling us to operate on the world documents of the Tarski folder simultaneously
without having to open them. And assume that we now say (3.34a) #For hours I put
the large cube to the left. Then we are immediately back in the first odd situation: the
sentence forces us into repetition or into stretching.
What can we learn from this? Well, we can observe that in the durative case we
do not (or cannot) take into consideration the cardinality information expressed by the
cube. Terminative aspectuality is closely tied up with cardinality information expressed
by a [+SQA]-NP. In the durative case, we have no immediate access to it. From this,
one may derive the position that the unique time structure of a document requires that
there be [+SQA]-information for it to be able to express terminativity. As soon as we
speak about more than one time structure (more than one model), we simply cannot
have terminativity. This is because the inspection of world documents with their own
time structure is subjected to the time structure of the inspection itself: folder time
dominates file time. It is only by compressing all these time structures—by obliterating
them in OS 12 that we are able to get back at terminativity: the desk top time structure
takes over, so all documents temporarily get the system 12.0 time. A situation that
comes very close to the OS 12 situation would be the well-known electronics shop
with the wall full of TV’s broadcasting the same program. In that case we would see,
say, twenty cubes but we would certainly speak about one cube because we attribute to
each screen the same time structure.
Another thing we learn is that we seem to operate at two levels: (a) the level of
terminativity; (b) a higher level in which we have to take into account so much of
the higher level time structure that we cannot have terminativity. A proper way to
account for this is again in terms of a multiplication. If we represent the information
expressed by the terminative sentence I put that cube to the left as a factor 1, one may
say that given n documents in the folder, the durative interpretation presumes n 1-
3.4. EVENTS OR INDICES? 85
occurrences of the same sort. This creates the sense of plurality which is inherent to
the notion of repetition. In terms of the structure in Figure 1.1, we may now assume
that the factor n in n 1 is located upstairs in the tree, say at the position of ASP .
We will see that the factor 1 might be taken as expressing an event e and in that sense
the habituality statement (at the folder level) provides information about n events. The
Tarski example discussed suppresses the notion of event a bit but this is to get at the
real thing happening, namely the fact that each document (event) has a unique time
structure making it different from any other document. This seems to be the heart of
the matter of so-called episodicity.
3.3.4 Conclusion
We have discussed some issues concerning the interaction between terminative seman-
tic objects and modifications of it: taking a part of it as in the Progressive Form and in
Imperfective sentences, giving a plurality of it in the form of habituality or frequency.
We have shown that levels of perspectives are involved and with these levels we see
that the type-token distinction tends to impose itself. This does not explain the issues
under investigation, it simply shows that some sort of multiplication is involved. Nega-
tion fits in this pattern. On the face of it, one could even argue that if the ASP-operator
would be represented as a factor, say as 1, then negation might be taken as bringing
about 01, whereas habituality would yield n 1, as opposed to the Progressive Form
having 0:n 1. We do not commit ourselves to such a perspective however tempting
it is to do so. But negation will turn out not to deny the existence of what is concealed
behind the factor 1 (the VP-terminative information). Before we tackle these very sub-
tle and intricate issues in chapter ?? we will first discuss the question of whether or not
the PLUG-approach of working with indices differs crucially from the event-semantic
approaches. The question will be, in short, whether we use numbers or events to model
our sense of time. In the next section we will compare the two approaches.
I(E) I(E )
6 k
I Int I(E) }(E) Int }(I(E))
E p - Ints(I(E)) E
? Int
(a) (b)
z }|i
2
{
z }|i
1
{
e1 e2 e3
The examples in (3.36) and (3.37) illustrate the nature of the mapping between the
structures E and Int(I(E)). This is done by the homomorphism p, which is structure-
preserving with respect to < and Æ. What it does is to provide for e1 in (3.36a) and
(3.37) the set ffe1 ; e2 gg and for e2 the set ffe1 ; e2 g; fe2 ; e3 gg, if we follow K&R.
In this way an event is associated with the instants at which it occurs.
The third option discussed by K&R is to abandon the idea that times should be
definable from events, which means that both events and times are taken as primitive
categories, that time has the structural properties of the reals IR, and that the event
structure and time structure are connected by some structure preserving function, which
is called LOC:
The most appropriate way of illustrating the third option may be Figure 3.13, in which
time and events are treated separately, the only direct connection being the LOC-function.
I(E) T I(E )
[] T
6 6
I Int I(E) }(E) Int }(T)
E LOC - Int(T?) E
? Int
(a) (b)
Whatever the merits of this third option, it is clear that it crucially uses the main in-
gredients of the Russell-Wiener construction, the difference being that rather than con-
struing time from events there is now an isomorphic mapping between event structure
and time structure.
(3.38) 0 =df ;
1 =df f;g = f0g
2 =df f;; f;gg = f0; 1g
3 =df f;; f;g; f;; f;ggg = f0; 1; 2g
4 = :::
Double perspective. For example, the fourth line of (3.38) says that 3 = f0,1,2g. In
other words, the number 3 is taken as a set with 0, 1 and 2 as its elements. We will
write : i3 = fe0 , e1 , e2 g, which simply says that 3 = f0; 1; 2g. The only goal of the
rewriting is to underscore the obvious parallel with an instant i as a set of e’s as defined
in Definition 5.6.1 and illustrated in Figure 3.12.
z }|i 3
{
z }|i {
2
s1 s2s3 s4
0 1 2 3
It is clear that the two perspectives on numbers play a role in Figure 3.14.
One difference between Figure 3.14 and Figure 3.12 is that Postulate 5 does not apply
to the former. To overcome this problem, we replace Æ by =. This leads to Defini-
tion 3.4.1, in which the labels of the model structures used by K&R are retained in
order to underscore the parallelism with Definition 5.6.1.
Definition 3.4.1
Let I be the set of indices isomorphic to IN with E = hI; i, obeying the postulates P1
- P7, with Æ replaced by =. Then I(E) = hI(E); <i i, with I(E ) the set of instants i as
3.5. EVENT-BASED APPROACHES 89
(i) i is an instant of E if
a. i I;
b. e1 ; e2 2 i ) e1 e2
(ii) For all instants i1 ; i2 : i1 <i i2 iff there are e1 2 i ;e 2 i : e < e
1 2 2 1 2
This produces the instant structure underlying the Path structure described above. From
this it should be possible to define a function Int: I(E) ! Int meeting the requirements
of Definition 5.6.2 and embedding Path structure into real time. In Verkuyl 1993 this
is done by an actualization function which connects the Reals to Figure 3.14 in much
the same way in which a performance of a sonata is related to the score. At that point
events take place in real time.
binary operation on its elements. D is partitioned into a set O of objects and a set of
events E and so there are two join-operations in these subdomains: hO; E; tO ; tE i.
He also assumes the (mereological) part of relation v. Thus, the resulting model struc-
ture is: hO; E; tO ; tE ; vO ; vE i. This move has clearly ontological consequences.
(3.39) a. Mary died
b. John and Mary died
Standardly one has Die(et) (me ) for (3.39a), which is true iff I (m) 2 I (Die). This
analysis leads to problems for (3.39b) because if John and Mary are treated as denoting
a set, a type clash results. There are two ways out to save the generalization over the
two sentences in (3.39):
to treat Mary as a singleton and take NPs as being of type ((et)t), which is the
set-theoretical solution;
to take John and Mary as an individual of type e.
This latter is obtained by allowing for the extension of the predicate logic with an
operator , which is interpreted as the join t of two individuals. Thus, on a lattice-
theoretical approach (3.39b) can be represented as Die(John Mary), where [ John Mary]
is considered an individual of type e.
Some problems
1. Naumann 1996 observed that the cumulativity requirement on is empty. There
is no verb that is not cumulative. This means that any hope for making a distinction
between stative and nonstative verbs in terms of the notion of cumulativity must be
given up. It also means that all information about the role of the verb is relegated to the
(e; y)-part.
2. This holds also for the information expressed by Æ (y ) because in derivations like
(3.40) and those ending in (3.41), there is no Æ at all: the information about the Patient
is only expressed by Pat(e; y ). R(y; 3) or R(y; n) do not have the form Æ (y ). If Æ (y ) is
in fact an abbreviatory device for the “sandwich-part” of R(y; 3), i.e. its R(y; : : :)-part,
then a new problem arises: the Æ (y ) is the same in R(y; 3) and R(y; n), so if Æ (y ) is cu-
mulative in the latter it should also be in the former. And this means that the distinction
between Judith ate three sandwiches and Judith ate sandwiches cannot be made on the
basis of Æ (y ). The “sandwich-part” of R(y; 3) and R(y; n) is already expressed by the
92 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
between s and ` warrants that the structure of I and the internal argument denotation
are in a mapping-relation. But this relation is not a naive physical one but it can be
formulated at the set-theoretical level at which quantification is dealt with.
These objections against Krifka’s lattice-theoretical approach in an event seman-
tic framework strengthen our wish pay attention to the position in which cardinality
information and, for mass NPs, information about some appropriate measure along the
lines of Bunt 1985 are taken as primordial for the aspectuality of a sentence. Focussing
on cardinality means that one abstracts from (naive) physics. This results in some un-
naturalness in the treatment of Judith ate a sandwich because the partitioning of the
NP-denotation [ a sandwich] yields just a singleton and so there is little room left for
what seems at first sight so nice and natural in a mereological treatment of this sentence
in terms of MtO and MtE. In the PLUG-treatment, the eating is a one swoop-event so
to say and one may wish that there should be more to it. Whether or not PLUG+ should
be “mereologized” in this direction is not clear, because from the point of view of
aspectual composition the [+SQA]-information interpreted model-theoretically as the
presence of cardinality information (or some measure function information) could be
judged sufficient. In this sense, Verkuyl prefers the strategy of a step-by-step extension
of set theory to ensemble theory as a basis for capturing the relation between Mass and
Count.
This raises the question of how Krifka’s e’s relate to the indices in the PLUG-
framework. If he would follow Naumann’s advice to simply recast MtO and MtE in
terms of the monotonicity property, then it would follow that the conditions come closer
to what is inherent to the interpretation of nonstative verbs in terms of the s-function.
It would be another way of saying that Krifka’s e’s may in fact be taken as indices.
Evidently, that would make Krifka’s grammar part of the PLUG-framework as far as
the verbal information is concerned.
This brings us finally to the point raised earlier, namely that Krifka actually uses
two notions of event. From the fact that (e) in (3.42) does not differentiate aspectually
between verbs, if follows more or less straightforwardly that the aspectual differences
are to be located at the level of ' itself. That is, at the level of VP. It is at that level
that Krifka may distinguish between a terminative event and a durative event But this
means that Krifka looks at '-events as dividuals. This is actually not what he says. But
it seems the only plausible way to interpret him. It should be added that ' does not
capture the role of the subject and that (3.42) does not say anything about the aspectual
relation between VP and S.
3.5.3 Parsons
In this section, we focus on Parsons’ predicates Hold and Cul in order to see how and
how much they contribute to a better understanding of the differences between durative
94 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
and terminative sentences. The relevant rules of his grammar are given in (3.45):
(3.45) R1: If 2 Vintr ; then 2 VP
R2: If 2 Vtrans & 2 Name; then 2 VP
R3: 2
8 etName
If & 2 VP; then 2 S &
< [ 0 ( e ) ^ Agent( 0 ; e) ^ Hold(e; t)];
a.
: et
if State, Process VP
[ 0 (e) ^ Agent( 0 ; e) ^ Cul(e; t)] if Event VP
8 et[ 0 (e) ^ Agent( 0 ; e) ^ Hold(e; t)]; if State VP
or:
<
b.
: et [ 0 (e) ^ Agent( 0 ; e) ^ Cul(e; t)]
if Process, Event VP
::: :::
R10: If 2 Adv & 2 VP; then = e[ 0 (e) ^ 0 (e)]
Both Hold and Cul in R3 are treated as two-place predicates between events and times.
R3 is split up into two different options. The a-option is chosen before 1984, the b-
option after 1984. There is an intriguing replication of the transition from R3a to R3b
in Parsons’ work. The transition from R3a to R3b took place around 1984 and so one
expects R3b to be established from that time on. But in Parsons 1990 the rules R3a
reappear at the beginning of the book and it is only at the close of the book that they
are rejected and that R3b is reintroduced. This makes it necessary to have a look at the
two options.7
the event did not (necessarily) culminate. Cula expresses the sense of culmination that
Parsons attributes to (3.46c).
In the absence of any treatment of quantification in Parsons’ work, we must assume
that the difference between Judith ate three sandwiches and Judith ate sandwiches
would be accounted for as in (3.47).
(3.47) a. 9e[Eating(e) ^ Subj(e; j) ^ Obj(e; 3 sandwiches) ^ Cul(e; t)]
b. 9e[Eating(e) ^ Subj(e; j) ^ Obj(e; sandwiches) ^ Hold(e; t)]
The question arises: does Parsons present here a theory about aspectuality? If so, the
theory does not explain anything: Hold/Cula are at best descriptive predicates. That is,
some sort of meta-rules may be formulated:
if [ ::: Holda (e)]; then durative
if [ ::: Cula (e)]; then terminative
However, this does not work properly. Firstly, Lascarides 1988 pointed out devastat-
ingly that there is no logical relationship between (3.47a) and (3.47b). That is, there is
no Hold-part in Cula . The two predicates are unrelated, so there is no way to provide
for the one-way inference relation between sentences with a Hold-predicate and a Cul-
predicate, i.e. between (3.47b) and (3.47a). Secondly, from the purely aspectual point
of view it should be obvious that the aspectual difference between (3.47a) and (3.47b)
must be found in the difference between three sandwiches and sandwiches and it is
very hard to see what Hold and Cul could possibly contribute to explaining that differ-
ence, the more so because in (3.46) it seems to account both for the difference between
the Progressive Form and the simple Past, and for the difference between stative and
certain nonstative predications. At best, the addition of the Hold- and Cul-predicate is
a way of saying that the two sentences in (3.47) differ, but this is what we already knew
by looking at them. For Parsons there is no way of expressing that Hold sometimes has
to look at the subject NP—Nobody in Nobody ate a sandwich—, sometimes to the ob-
ject NP —three sandwiches, sandwiches—and sometimes to the verb (dislike in Judith
disliked a sandwich in order to establish durativity. In short, Hold and Cul do not do
anything to connect themselves to the spots where the relevant aspectual information
is. Not the slightest hint about which direction should be taken has been given in any
of Parsons’ articles nor in Parsons 1990.
verb and apply the theory as before” (p. 184). From then on we have:
(3.48) a. Mary ran
a.0 9e[Running(e) ^ Subj(e; m) ^ Cul(e; t)]
b. Mary ran to the store
b.0 9e[Running(e) ^ Subj(e; m) ^ To(e; the store) ^ Cul(e; t)]
c. Mary was running to the store
c.0 9e[Running(e) ^ Subj(e; m) ^ To(e; the store) ^ Hold(e; t)]
And we may add:
(3.49) a. Mary dislikes Fred
a.0 9e[Disliking(s) ^ Subj(s; m) ^ Obj(s; f) ^ Hold(s; t)]
b. Mary was crossing the street
b.0 9e[Crossing(e) ^ Subj(e; m) ^ Obj(e; the street) ^ Hold(e; t)]
c. Mary crossed the street
c.0 9e[Crossing(e) ^ Subj(e; m) ^ Obj(e; the street) ^ Cul(e; t)]
It is clear that some additional notions are required. It is also clear that the distinction
between Hold and Cul now is just the lexical distinction between stative and nonstative
verbs, or rather the way the events corresponding with them actualize in time. This
raises a problem in (3.48c0): Hold cannot hold for events any longer. This is solved by:
(3.50) a. Mary be running to the store
a.0 9e[Running(e) ^ Subj(e; m) ^ To(e; the store)
^ Hold(In-prog(e); t)]
The operator In-prog takes an e and makes an s out of it: Parsons 1990:234 describes
this operator as follows: “whenever an event e is in progress, there is a corresponding
state of affairs, ‘that e is in progress’. This is a state that holds while e is in progress
(and at no other time). Call this the ‘in-progress state’ of the event e, or ‘e’s IP-state’ ”.
This can only mean that for Parsons Holdb and Culb are no longer part of an aspectual
theory: they are now part of the Tense system in the sense that they can be considered
as accounting for the connection between States and Events with real time. They have
nothing (structurally) to do in making a sentence (in-)compatible with for an hour and
in an hour. Parsons distinguishes between states and events and for each of the two
ontological categories he needs a term to express two types of actualization in real time:
Cul for nonstative predications; Hold for stative ones. Then an aspectual notion comes
in: IP-state. The reasoning must be the following: Cul brings us in the domain of non-
stative semantic objects—events—and these have a development part, a culmination
part and a consequent state part, in Parsons 1990:235 called Resultant State. In-prog
always takes the development part and yields a state.
3.5.4 Jackendoff
Introduction
Jackendoff’s work on thematic structure has been very important over the years. It is an
extension of ideas generated by Gruber 1976, and from its beginning Jackendoff 1972
3.5. EVENT-BASED APPROACHES 97
3.5.5 Pustejovsky
Introduction
In a number of recent papers Pustejovsky’s work on the generative lexicon has drawn
considerable attention: Pustejovsky 1991, Pustejovsky 1993, among several others.
These papers are now worked into Pustejovsky 1995. An important part of his the-
ory is his notion of Qualia Structure, which spreads the burden of information over the
verbs and its arguments. Arguments contain information about event structure in which
they play a role. Pustejovsky thus articulated the notion of event structure.
Event structure
Pustejovsky distinguishes between States, Processes and Events. Events may have
eventual substructure. An event structure has the form in Figure 3.15.
e<a
e1 e2
3.5.6 Conclusion
We have discussed only some of the many event-semantic approaches available now,
the choice having been made dependent on whether a particular approach represents a
colour on the palette different enough from the other ones. For the present course this
should be sufficient. In the references we have included some of the most important
papers, collections of papers, dissertations or books that are around, most of them
having the noun event in their title.
100 CHAPTER 3. THE SEMANTICS OF INNER ASPECTUALITY
Chapter 4
4.1 Introduction
In the literature on aspectual class (‘Aktionsart’ or inner aspectuality as we call it in
this book) and grammatical aspect (or outer aspectuality as we call it), we can distin-
guish two general approaches. Smith (1991), Depraetere (1995) and others claim that
aspectual class and grammatical aspect are essentially different notions, which involve
distinct analytical tools. Smith talks about ‘situation type’ versus ‘point of view’, De-
praetere’s terms are ‘(a)telicity’ versus ‘(un)boundedness’. Smith and Depraetere pro-
vide a fine-grained analysis of the meaning effects which result from the combination
of different situation types with different aspectual markers. A disadvantage of their
approach is that it does not allow recursion, because there is exactly one situation type
and exactly one view point. This makes it hard to account for sentences in which com-
binations of expressions involving outer aspectuality play a role, e.g. both negation and
perfectivity/imperfectivity, or both progressive and aspectual adverbs or both iteration
and frequency, etc. Somehow, we need a notion of outer aspectuality which has recur-
sivity built into it. For constructions and languages in which there is no visible effect of
grammatical aspect it may be necessary to assume a third notion of ‘zero’ or ‘neutral’
aspect, which does not play an active role in the interpretation.
On the other hand, we find analyses in which the distinction between Aktion-
sart and aspect or inner and outer aspectuality is blurred. Hinrichs (1986) argues that
Progressive sentences in English introduce state variables, just like lexical states. His
analysis is based on Kamp and Rohrer (1983), who take Passé Simple sentences in
French to refer to events, and Imparfait sentences to describe states. Along similar
lines, Piñón (1995) states that Polish imperfective verbs denote processes, while per-
fective ones denote events. These analyses suggest that the situation type either does
not influence the denotation of the sentence as a whole, or is in some sense ‘overruled’
by the contribution of grammatical aspect. This view is motivated by a concern with the
temporal structure of narrative discourse, where we observe differences between states,
101
102 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
processes and events. Events move on the story line, whereas states typically provide
background information. Processes occupy an intermediate positon. If only the output
type matters at the discourse level, it may seem less important to know whether the
sentence denotes a state because it just has the Aktionsart of a state, or because it is a
derived (progressive or imperfective) state. The drawback of this focus on output type
is that we do not gain insight into the way specific meaning effects result from the com-
bination of aspectual class and aspectual operators. Moreover, we will see in chapter 5
that these subtle meaning effects actually do have consequences for the interpretation
of the temporal structure of narrative discourse.
Our general aim in this book is to develop a compositional analysis of Aktion-
sart and aspect which preserves insights from the two approaches we just described.
On the one hand, we keep the distinction between Aktionsart (inner aspectuality) and
grammatical aspect (outer aspectuality), so that we can study the meaning effects their
combination gives rise to. On the other hand, we assume that the model-theoretic no-
tions underlying inner and outer aspectuality are essentially the same. In chapters 1
and 3, we have seen that the Plus- principle is crucial for the calculation of inner as-
pectuality. The combination of a predicate and its inner and outer arguments at the
micro level yields an eventuality as its output at the macro level. In the study of outer
aspectuality, the Plus-principle does not play a role, for aspectual expressions can now
operate on a full-fledged eventuality. This does not mean that we cannot formulate an
analysis of outer aspectuality in terms of an interval-based semantics. In fact, we could
push the interval-based approach all the way up to the tense level, and claim that events
do not enter the picture until their actualization on the time axis (cf. chapter 2, section
5). However, the limit on the scope of the Plus-principle can also be taken to open the
door to an event-based analysis of outer aspectuality. In this chapter, we will interpret
the shift from micro to macro level as the challenge to develop an analysis of outer
aspectuality in terms of an event-based semantics. This chapter includes parts of de
Swart (1998) and de Swart and Molendijk (1999).
The Kleene star indicates zero, one or more operations. In terms of the syntactic
tree of the sentence as a whole, the eventuality description is the S-level, Aspect is the
4.2. AN EVENT-BASED SEMANTICS OF OUTER ASPECTUALITY 103
figure 1
Tadv S
Tense S
::: S
ASP S
NPext VP0
ASP VP
V NPint
figure 2
HOMOGENEOUS QUANTIZED
state process event
STATIVE DYNAMIC
Determiners like many and few are generalized quantifiers over atomic individuals
(in A), whereas much and little establish measurement relations between subsets of the
4.2. AN EVENT-BASED SEMANTICS OF OUTER ASPECTUALITY 105
domain D. (75) shows that a mass interpretation can easily be derived from a count
meaning: the determiner in (75a) operates on the count predicate Apple, the one in
(75b) on its mass counterpart m Apple. If we assume that such expressions as water,
gold, and beer translate as mass terms, we further need mappings from the mass to the
count domain to account for sentences like (76):
Beer in (76b) can be used to refer to a serving or a kind of beer. Starting with a
basic mass predicate P , we need to be able to define its count counterpart, which we
could call c P . In addition to the materialization function h, we thus need packaging
functions g which map quantities onto bounded entities. As a result, we have both
a ‘Universal Grinder’ and a ‘Universal Packager’ (Pelletier 1979; Bach 1986) in our
semantics.
The observation that states can be presented as events, and vice versa suggests
that the metaphor of the Universal Grinder and the Universal Packager extends to the
temporal domain, where it can be used to account for aspect shift. In this section
we will be mainly concerned with aspectual operators in a language like English. As
argued in chapter 1 above, the aspectual morphology of Slavic languages is interpreted
at the level of inner aspectuality (Asp ). The aspectual operators we are interested
in right now are interpreted at the level of outer aspectuality (Asp ). The aspectual
system of Romance languages will be argued to belong to the Tense node, rather than
the Asp node. This issue will be addressed in section 4 below. In English, aspectual
operators are optional. There are no aspectual operators in (77a) and (b):
figure 3
Mary was sick
106 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
n s t x
tn
Mary(x)
sÆt
s: x be sick
figure 4
Mary wrote a letter
n e t x y
tn
Mary(x)
Letter(y)
et
e: x write y
Figures 3 and 4 illustrate that events are included in their location time t, while states
and processes overlap with t. The past tense operator introduces a condition t n,
and locates t before the speech time n ‘now’.
The Perfect is an example of an aspectual operator in English. Consider (78a) with
the grammatical structure spelled out in (78b):
(78) a. Mary has met the president.
b. [ PRES [ PERF [ Mary meet the president ]]]
In our compositional analysis, we want a unified interpretation of tense operators
in sentences with or without aspectual operators. The optionality of aspectual operators
suggests an interpretation in terms of eventuality description modifiers: they map sets
of eventualities onto sets of eventualities. (cf. figure 1 above). This interpretation
guarantees that the input to the tense operator is always a set of eventualities. It also
opens the way to the iteration of aspectual operators, examples of which are given
below.
Sentences in the Perfect are stative, so the Perfect denotes a function PERF: E !
S , mapping any kind of eventuality onto a state. The Perfect introduces the consequent
state which starts when the eventuality ends:
figure 5
Mary has met the president
4.2. AN EVENT-BASED SEMANTICS OF OUTER ASPECTUALITY 107
n e s t x y
t=n
Mary(x)
President(y)
sÆt
e s
e: x meet y
The condition e s (e ‘abuts’ s) means that the result state starts right at the end of
the event. The Perfect is an extensional operator, which asserts the existence of both
the event e, and its consequent state s. It is the result state which provides the variable
for the tense to operate on, because that state corresponds to the output of the Perfect
operator. If the result state s holds now, we know that the event e took place in the past
of the speech time, and the consequent state has current relevance.
The Perfect operates on eventualities of any aspectual type.1 Other aspectual oper-
ators are more selective and come with special input conditions. Consider the Progres-
sive in English. Sentences in the Progressive are always stative, whether the underlying
sentence is characterized as a process (79a) or as an event (79b) (Vlach 1981, Hinrichs
1986). This characteristic they thus share with the Perfect. The Progressive does not
normally combine with stative eventuality descriptions (79c):
(1987, 70–76) and Kamp and Reyle (1993, 579–587) for discussion.
108 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
introduction of the operator PROG gives (80a) the grammatical structure (80b), and the
DRS representation in figure 6:
n s t x y
tn
Mary(x)
Book(y)
sÆt
e
s: PROG
e: x read y
In accordance with the intensional nature of the Progressive, there is no claim that the
event variable corresponding to the description ‘Mary read a book’ exists in the main
box of the DRS in figure 6.2 At the top level, only the state of the event in progress is
asserted to exist. The tense operator locates this state in the past.
The aspectual operator in the grammatical structure of (1) carries a Kleene star,
which allows for zero, one or more applications. The examples in (77a) and (b) do not
involve any aspectual operators. The examples in (78) through (80) have one aspectual
operator each. Examples involving both the Perfect and the Progressive require the
application of an aspectual operator to a structure which already contains an aspectual
operator:
(81) a. Jane has been writing a letter.
b. [ PRES [ PERF [ PROG [ Jane write a letter ]]]]
The Progressive in (81) maps a set of events onto the set of states of Jane’s writing
a letter being in progress. Given that aspectual operators are interpreted as eventual-
ity description modifiers, the output of the operation is of the same type as the input,
namely a set of eventualities. This allows for multiple aspectual operators which apply
recursively. In (81), the set of states which results from the application of the Pro-
gressive is taken as the input for the Perfect operator, which maps it onto the set of
consequent states of Jane writing a letter.3
2 Given that a book one reads already exists, its discourse referent is introduced in the main box. With
verbs of creation such as build, write, we have the problem of unfinished objects (Parsons 1991). In those
cases the discourse referent for the object is introduced in the embedded box (cf. figure 8 below).
3 In principle, event modifiers can be freely ordered with respect to each other. However, in (12) the
reverse order of the aspectual markers is not well-formed, because the output of the Perfect operator is
4.2. AN EVENT-BASED SEMANTICS OF OUTER ASPECTUALITY 109
Vet (1980), Moens (1987), Naumann (1995) and others point out that a for-adverbial
brings in a notion of boundedness. As a result, the inference from Progressive to Sim-
ple Past goes through in (84a), but not in (84b):
sentences involving quantifiers and negation. If we combine the analysis developed here with the assumption
that negative and quantified sentences introduce their own eventuality referent, as argued for in section 3
below, most of her objections disappear.
110 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
figure 7
Susan lived in Paris for two years
n s e t x y mt
tn
Susan(x)
Paris(y)
et
e = BOUND(s)
two years(mt)
dur(s) mt
s: x live in y
The amount of time (mt) introduced by the for-adverbial in figure 7 is two years, and
the duration (dur) of the state s of Susan’s living in Paris is at least as long () as
that. The application of the BOUND-operator to the state variable results in a bounded,
quantized portion of the state (see appendix, part D), which counts as an event because
it is quantized.
Given that time adverbials are extensional in nature, the construction rules for
for- and in-adverbials are quite similar to those for the Perfect. The interpretation of
both aspectual operators and duration adverbials as eventuality description modifiers
opens the way to recursive structures in which multiple operators are piled on top of
each other. (84b) combines the Progressive with a for-adverbial. The examples in (86)
contain a Progressive and an in-adverbial:
n s t x
tn
Susan(x)
sÆt
e e’ mt y
book(y)
s: PROG e’ = e
a month(mt)
dur(e) mt
e: x write y
The amount of time (mt) given by the temporal noun specifies the period within
which the event culminates, so that its duration (dur) is less than () a month. Given
that the output of the in-adverbial is again an event, it can serve as the input for the
Progressive.5 The interpretation of duration adverbials as eventuality description mod-
ifiers makes interesting predictions for the combination of their French counterparts
with the Passé Simple/Imparfait, as we will see in section 4 below.
4.2.4 Coercion
The interpretation of aspectual operators as eventuality description modifiers builds on
ideas developed by Moens (1987, 44f.). Moens constructs an aspectual network out of
a number of aspectual categories and relations between them. He labels a number of
aspectual transitions as established ‘routes’ through the network. Some of these shifts
are controlled by explicit aspectual operators. For instance, the change from event to
consequent state requires the use of the Perfect, and the transition from dynamic action
to state of the action in progress is expressed by the Progressive in English. Transitions
for which a language has no explicit markers are free as long as the context supports the
meaning effects associated with the aspectual change. Such transitions as inchoativity
and habituality in English are governed by a contextual reinterpretation process called
‘coercion’. Moens illustrates this with examples like (88):
(88a) measures the time during which the program ran. (88b) describes how long
it took for the program to start running (adding a preparatory phase to culminate in
the program running) or to do a complete run (turning the process into a culminated
process). We can coerce a state into an event by emphasizing the starting point of the
state, which gives the sentence an inchoative reading (89a). The event of reading a
5 Again, we do not get scope ambiguities, because the other scope order is ruled out as a violation of
aspectual constraints.
112 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
book can be coerced into an ongoing process (89b). Events can also be coerced into
states by giving the sentence an iterative (89c) or a habitual reading (89d):
a functor and its argument would lead to ungrammaticalities unless the argument is reinterpreted in such a
way that it meets the requirements on the functor; see Pustejovsky (1995) for a more general discussion.
7 States and events do not form a natural supercategory in figure 1, so no aspectual transitions relate the
[
set S E to the set P .
4.2. AN EVENT-BASED SEMANTICS OF OUTER ASPECTUALITY 113
which proceeds as usual. In examples like (88b) and (89), the eventuality description
does not satisfy the aspectual selection restrictions of the adverb. In such cases, the
verification procedure fails unless we coerce the description into an eventuality of the
appropriate type. For (89c), introduction of a coercion operator yields the grammatical
structure in (90) and the DRS in figure 9:
(90) John played the sonata for eight hours
[ PAST [ FOR eight hours [ Ceh [ John play the sonata ]]]]
figure 9
John played the sonata for eight hours
n e h t x y mt
tn
John(x)
Sonata(y)
et
e = BOUND(h)
eight hours(mt)
dur(h) mt
e’
h: Ceh e’: x play y
Combination of the event predicate with the for-adverbial fails, because the con-
dition ‘dur(h) mt’ in the semantics of the time adverbial cannot be met by the
quantized eventuality description. Application of Ceh maps the event of playing the
sonata onto a homogeneous eventuality description. This mapping introduces the de-
sired variable h into the DRS, which can then combine with for eight hours to result
in a bounded process. In figure 9, Ceh naturally gets the value of iteration: the event
takes place an indefinite number of times (compare the definition of ITER in part D of
the appendix).
The incompatibility of the homogeneous description with an in-adverbial in (88b)
triggers a coercion operator Che in the grammatical structure (91) and the DRS in figure
10, which introduces the necessary event variable e:
(91) The program ran in four minutes.
[ PAST [ IN four minutes [ Che [ the program run ]]]]
figure 10
The program ran in four minutes
114 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
n e e’ t x
tn
program(x)
e’ t
e’ = e
four minutes(mt)
dur(e) mt
p
e: Che p: x run
One interpretation for Che is ADD-PREP, which implies that the in-adverbial specifies
the amount of time (mt) within which the event of the program getting up and running
culminates. Another possible interpretation is ADD-CUL, which adds a culmination to
the process so that the complete run of the program takes no more than four minutes
(compare appendix, part D for definitions of the aspectual transitions).
A coercion operator also accounts for the felicity of sentences like (92):
The Progressive does not normally apply to state descriptions. But there are well-
known exceptions to this rule. The examples in (92) are taken to have a certain color,
and give a more vivid description than conventional statives. Smith (1991: 20) argues
that the sentences in (92) present states as dynamic situations, because they use the
viewpoint of the Progressive, which in English normally only occurs with non-statives.
In the present framework, we assume that the state has been coerced into a dynamic
eventuality by Csd before the Progressive applies, in order to satisfy the input condi-
tions on the aspectual operator. The grammatical structure of (92a) in (93) leads to the
DRS in figure 11:
n s t x y
t=n
Susan(x)
Play(y)
sÆt
d
s: PROG
d: Csd
s’
s’: x like y
The dynamic variable d, obtained by coercion, is the input for the Progressive
operator. The output of the Progressive is again a state, but the state of an event or
process being in progress is a lot more ‘dynamic’ than the underlying lexical state. The
more complex internal structure accounts for the vivid color of the description.8
As far as the interpretation of aspectual operators is concerned then, we only need
to make sure we have a variable of the right aspectual type in the DRS to satisfy the
aspectual constraints of the operators. We achieve this result by defining construction
rules which check the input conditions for aspectual operators, and introduce a coercion
operator Che , Ceh or Csd if the aspectual constraints are not satisfied. Such construc-
tion rules build DRSs like the ones in figures 9 to 11. Sentences can be ambiguous in
isolation (e.g. 19b), and the coercion operator leaves room for different interpretations.
Even though we can usually work out for each example what the semantic value of the
operator Che , Ceh or Csd will be in the context, we do not need to introduce specific
aspectual operators ITER, ADD-CUL or BOUND into the DRS. This is good, because we
want to preserve the insight that the interpretation in terms of coercion is fully com-
positional, but the value of the hidden operator is dependent on linguistic context and
world knowledge, and is determined only when the DRS is embedded in the model.
At this point, coercion of an event into a state or process leads to a choice among the
set of possible mappings fITER, HAB, PROC, . . . g, and coercion of a state or process
into an event leads to a choice among the set fBOUND, INCHO, ADD-CUL, . . . g.9 If
the linguistic and extra-linguistic context do not support any aspectual transition which
satisfies the aspectual restrictions on the operator, there is no proper embedding of the
DRS into the model, which means that it cannot be verified. Thus the well-formedness
of the DRS can be saved by the introduction of coercion operators, but its verification is
dependent on the felicity of specific aspectual transitions in the context.10 The appendix
8 English does not seem to use a coercion operator Cds , presumably because it lacks aspectual operators
which require non-dynamic eventualities as their input.
9 Not all interpretations are necessarily available for each eventuality description. See Pustejovsky (1995)
for observations about the semantic frame of different classes of eventuality descriptions, which constrains
coerced readings.
10 See Moens (1987, 106–107) for similar observations. If we had not wanted to make coercion operators
explicit in a theory like DRT, we could have developed an underspecification approach, instead of a mapping
analysis. However, as Bach (1986, 11) points out, the interpretation of common nouns is not completely
116 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
spells out the formal definition of the construction rules (part C), and the verification
procedure (part E), and gives an informal characterization of the semantic effect of the
aspectual transitions triggered by coercion (part D).
The claim is thus that logical operators are eventuality description modifiers, that
is, they introduce operations over eventualities, and build new, complex eventuality
descriptions out of other—atomic or complex—eventuality descriptions. The semantic
type of the expression does not change: the input for the logical operator is a set of
eventualities, and so is the output.
account situations in which Jane is alive. We assume that such restrictions are handled by the pragmatic
module.
4.3. LOGICAL OPERATORS: NEGATION AND QUANTIFICATION 117
discourse referents. As Kamp and Reyle (1993, 673) point out, it is straightforward to
construct states out of triples < t; x; P >, where t is a period of time, x an individual
and P a property such that t is an interval at which x has P . The state consists of
the individual having the property P at time t. What properties of individuals have
in common with quantified and negative statements is that they all introduce (static)
conditions into the DRS. We use a generalized version of Kamp and Reyle’s proposal
to build states out of pairs < t; > of a period of time t and a condition holding at
that time:12
(96) Recursive rule for the introduction of discourse referents:
For a condition holding at time t, s: and Max(s) and s =t t define the
maximal state consisting of the condition holding at time t.
can either be an atomic condition (a property ascribed to an individual) or a complex
condition (that is, a quantified condition). The application of this rule implies that
every quantified statement denotes a state description, namely the set of states which
consist of the negative or quantified situation holding at some time t. Fernando (1994)
suggests that there is a natural correlation between conditions and states. In DRT,
conditions require verification of embeddings in the model, and do not involve the
notion of change. DRSs, like events, involve change, and are interpreted as a relation
between input and output assignments. A condition can be understood as a special
kind of DRS, whose output assignment does not differ from its input assignment. This
makes it possible to reinterpret conditions as eventualities, as long as we label them as
states. For the condition on maximality, see below.
An example like (95) can now be interpreted as in figure 12:
figure 12
Jane always came late
n s t x
tn
Jane(x)
s =t t
Max(s)
e e’
s:
es
) e v e’
e: x come e’: x come late
12 The phrase ‘s holds at t’ suggests that s and t are coextensive rather than that they just overlap. Accord-
ingly, we strengthen the DRT condition and take states to be coextensive with their location time, written
= =
t . s t t is defined as the conjunction of s t and t s, where denotes temporal inclusion. This can
be viewed as a temporal version of the ‘strongest meaning hypothesis’ (Dalrymple et al. 1998).
118 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
The complex, quantificational condition holds over a period of time preceding the
moment of speech, so the tense operator constrains the domain of quantification. The
DRS is true iff every (contextually relevant) event of Jane coming within the domain of
quantification of the adverbial quantifier extends to her coming late (cf. De Swart 1991
for more details).
In essence, we want to apply the recursive rule for the introduction of discourse
referents to negative conditions as well as quantified conditions. Consider example
(97):
(97) Jane didn’t come.
According to (97) Jane has the property of not coming in at some time t in the
past. We view this as a complex property which is true if there is no embedding func-
tion which verifies the subordinate box (see appendix, part D) in figure 13:
figure 13
Jane didn’t come
n s t x
tn
Jane(x)
s =t t
Max(s)
e
s: : es
e: x come
This proposal immediately raises two important problems. First of all, it is obvious
that we cannot view the complex negative state as a minimal eventuality. Second, it
has been argued that that negative sentences do not describe eventualities (Kamp and
Reyle 1993, Asher 1993 and others). Before we develop an analysis of more complex
examples, we will address these two issues.
We agree that the complex negative state s that a sentence like (96) introduces
should not be viewed as a minimal situation. It is too weak a statement to say that we
can find at least one minimal situation holding at t at which Jane coming is not the
case. There can be many of such eventualities without them guaranteeing the truth of
(97). A similar problem arises in the interpretation of monotone decreasing indefinite
NPs like at most three N (cf. Krifka 1989). In order to make a sentence like (98) true,
it is not enough to find some group of students which has at most three members; we
have to rule out the existence of a larger group:
(98) At most three students failed the exam.
4.3. LOGICAL OPERATORS: NEGATION AND QUANTIFICATION 119
The non-persistent character of negative sentences like (97) and (98) implies that
we have to switch to maximal eventualities. Following Krifka, we interpret maximal
eventualities e (generalizing over states and events) as the sum of all eventualities at
some point in time:
(99) 8e [MAX (e) $ 9t [e = supE (e0 (e0 t))]]
supE is the supremum in the join semilattice of eventualities E , which gives us
the sum of all the eventualities temporally included in t. Krifka describes negative
eventualities as the ‘fusion’ of all eventualities at a given time t which are not of the
type described by the sentence. The dependency on t accounts for the intuition that
negative sentences typically presuppose a contextually determined period of time for
which the negative state of affairs holds. Krifka’s proposals lead to the definition of
event predicate negation in (100) (where e generalizes again over events and states):
(100) P s 9t [MAX (s) ^ s =t ^ : 9e [P (e) ^ e s]]
Negation is thus a modifier which operates on an eventuality description P and
yields the maximal state s holding at t such that no eventuality e of type P is included
in s at t.
The condition on maximality takes care of the technical problem surrounding the
introduction of a discourse referent for negative sentences. However, this is not enough
to really motivate the introduction of a discourse referent for negative eventualities per
se. We think that there are two reasons why it is useful to extend the recursive rule
for the introduction of discourse referents to negative sentences. The first is that this
allows us to deal with more complex examples than the ones Kamp and Reyle took
into consideration, including cases in which negation interacts with other aspectual
and logical operators. The second argument is that a treatment of logical operators as
essentially the same type of expressions as aspectual operators helps us understand the
cross-linguistic variation in the interaction between negation and aspect. This will be
illustrated for Romance in section 4 below. For the time being, we will limit ourselves
to English.
In order to appreciate the differences between Kamp and Reyle’s theory and the
analysis we are developing here, it is useful to run through their argumentation. Ac-
cording to Kamp and Reyle (1993, 546-555), negation is not an aspectual operator,
and we do not need to introduce a discourse referent for negative sentences as a whole
in order to obtain the correct interpretation for sentences containing frame adverbials
such as on Sunday, compare:
(101) a. Mary wrote a letter on Sunday.
b. Mary didn’t write a letter on Sunday.
In the Discourse Representation Theory (DRT) of Kamp and Reyle, these sentences
get the interpretation in figures 14 and 15. These Discourse Representation Structures
(DRSs) contain discourse referents, and conditions which assign properties to the dis-
course referents:
120 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
figure 14
Mary wrote a letter on Sunday
n e t t’ x y
tn
Sunday(t’)
t = t’
et
Mary(x)
letter(y)
e: x write y
figure 15
Mary didn’t write a letter on Sunday
n t t’ x
tn
Sunday (t’)
t = t’
Mary(x)
e y
: letter(y)
et
e: x write y
Temporal nouns like Sunday introduce an interval t0 into the DRS, and the con-
dition that t0 is a Sunday (typically the most recent Sunday in a past tense context).
The preposition on identifies t0 with the location time t of the eventuality e. The past
tense introduces the condition that t precedes the speech time now: t n. Events
are included in their location time, which is captured by the condition: e t. This
leads to an inclusive or existential interpretation of the frame adverbial in figure 14.
The negation operator in figure 15 is generated lower than tense, but high enough to
take scope over the event variable. As a result, we get the interpretation that no event
of Mary writing a letter is included in last Sunday. This corresponds with a durative
or universal interpretation of on Sunday in (101b): the negative sentence is true for
the entire Sunday. Given that stative sentences trigger durative readings of locating
time adverbials, the universal reading of (101b) has been taken as an argument in fa-
vor of the claim that negative sentences denote states (Verkuyl 1993, 163). However,
in Kamp and Reyle’s approach, it is not necessary to interpret negative sentences as
states to derive the correct meaning of (101b). Negation introduces a subordinate box,
4.3. LOGICAL OPERATORS: NEGATION AND QUANTIFICATION 121
but the temporal structure of the negative sentence is the same as that of its affirmative
counterpart.
Kamp and Reyle further argue that the interpretation of negative sentences as re-
porting states would yield the wrong results in discourse contexts. It would blur the
distinction between (103a) and (b), which are to be compared to their positive counter-
parts in (102):
(102a) suggests that Bill smiled after Mary looked at him, perhaps as a reaction
to her doing so. In DRT, this is derived by a rule for the interpretation of narrative
discourse which says that events usually happen in succession. Accordingly, a condi-
tion e e0 is introduced into the DRS. (102b) describes Bill as already smiling when
Mary looked at him. Kamp and Reyle treat the Progressive as an aspectual operator
which applies to the predicate smile to derive the corresponding state of the event of
smiling being in progress. They assume that states temporally include the event which
functions as their antecedent in the discourse, which introduces a condition e s. The
examples in (103b) show that the contrast between the Simple Past and the Progressive
is preserved under negation. This is reflected in the DRSs 16 and 17, which Kamp and
Reyle propose for (103a) and (103b):
figure 16
Mary looked at Bill. He didn’t smile
n e t x y t’ u
tn
et
Mary(x)
Bill(y)
e: x look at y
t’ n
u=y
e’
: e’ t’
e e’
e’: u smile
122 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
figure 17
Mary looked at Bill. He wasn’t smiling
n e t x y t’ u
tn
et
Mary(x)
Bill(y)
e: x look at y
t’ n
u=y
s
: s Æ t’
es
s: u PROG(smile)
If the non-arrival of the train denoted a state, we would expect to be able to mea-
sure this state and say how long it lasts, but that seems impossible. However, (104a)
contains a negated event nominal, which does not necessarily behave in the same
way as a negative sentence simpliciter. Although temporal measurement is apparently
blocked for negated event nominals, it is perfectly acceptable for regular negative sen-
tences such as (104b). Also, the duration of a negative state of affairs can be questioned
as in (104c). Although Asher is well aware of these observations, he does not provide
an explanation for the difference in behavior between negative sentences and negated
event nominals. In our view, it is the unacceptability of (104a), rather than the well-
formedness of (104b) and (c) which is in need of explanation. One possible hypothesis
is that the lack of tense and aspect precludes the introduction of a stative discourse
referent for the negated event nominals. As a result, only negative sentences which
come with tense and aspect would allow the introduction of a discourse referent (i.e.
an eventuality variable). This hypothesis would immediately account for the difference
between (104a) on the one hand, and (104b) and (c) on the other. However, given that
we do not intend to provide a semantics of event nominals in this book, we will not
pursue this line of research any further. Independently of the correct solution to the
problem of event nominals, we need to explain why negative sentences are compatible
with measurement phrases.
Amsili and Le Draoulec (1995) take sentences like (104b) and (c) to motivate a
reanalysis of for-adverbials as measurement phrases over intervals rather than states or
activities. They claim that negative sentences do not introduce a discourse referent (i.e.
an eventuality variable). However, they accept the idea that negative sentences have a
location time t outside the scope of negation. They take t to be an interval the length
of which can be measured by a durative adverbial. In their view, the assumption of
a location time t outside the scope of negation also explains the felicity of negative
temporal clauses:
(105) a. When John didn’t show up, Mary went to the movies by herself
b. When Mary didn’t finish her dissertation in time, she was fired
c. Mary asked John a question. When he didn’t answer, she walked out on him
Most analyses of (sentence-initial) when-clauses are set up in such a way that the
temporal connective establishes a relation between the two eventualities, and the subor-
dinate clause functions as the temporal antecedent for the main clause. As Heinämäki
(1984) points out, a negative sentence cannot be the temporal antecedent of the main
clause in sentences like (105) if it does not itself introduce an eventuality variable, or
if it denotes an atemporal fact. Amsili and Le Draoulec contest Heinämäki conclusion.
They refer to Hamann (1989) for the claim that the absence of an event can some-
times license the introduction of a reference time, in particular when the event was
expected or scheduled to occur. More generally though, the introduction of reference
times in theories like DRT and related frameworks is tied to the presence of an eventu-
ality variable. This makes the solution proposed by Amsili and Le Draoulec almost an
acknowledgement of the discourse status of negative eventualities.
124 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
Some further problems for the analysis of negative sentences in terms of time-
or fact-denoting expressions have been pointed out in the literature. Horn (1989: 55)
argues that this view cannot deal with examples like (106):
(106) What happened next was that the consulate didn’t give us our visa
Assuming with Vendler (1967: 141) and Horn (1989: 55) that events, but not facts
“happen”, we are forced to admit that we have a negative event here, and not something
that simply didn’t happen.
A related issue is brought up by Stockwell, Schachter and Partee (1973), who point
out that there are certain cases where the negation of an event may, loosely speaking,
itself be an “event”. They discuss cases like not paying taxes, not getting up early, not
going to church, not eating dinner, etc., and show that the negative “eventuality” can
even be embedded under frequency adverbs:
(107) a. He often hasn’t paid taxes
b. He sometimes doesn’t eat dinner
c. He doesn’t eat dinner two nights a week
These sentences allow a reading in which the frequency adverb has scope over
negation, meaning that they describe the frequency of a negative situation. As pointed
out by Stockwell, Schachter, and Partee, sentences like (29) are a counterexample
against the claim that “one cannot assert the frequency of an event that does not occur”
(Lakoff 1965: 172). Sentences like (29) are also hard to reconcile with an interpreta-
tion of negative sentences in terms of facts. If facts are abstract objects, which are not
located in time and space, one would predict that embedding under a frequency adverb,
which typically quantifies over eventualities according to De Swart (1991) is not possi-
ble. Under the analysis advanced here, the combination of negation and quantification
in examples like (107) is easily handled by the recursive nature of the Asp node.
The arguments discussed so far mainly concern the sentence-internal structure of
negative sentences. But Kamp and Reyle’s (1993) arguments were also crucially based
on the behavior of negative sentences in narrative discourse. However, we find that
they base their analysis on a relatively small set of examples. If we take into account
a wider range of data, we find that the temporal structure of negative sentences cannot
always be derived from their affirmative counterparts. Compare (108a) and (b):
(108) a. Paul switched on the light. The room was very bright.
b. Paul switched on the light. The room was not very bright.
As pointed out by Hinrichs (1986) and Lascarides and Asher (1993), the room be-
ing bright is a result of the event of switching on the light. This interpretation triggers
a temporal structure in which the state s immediately follows the event e. We will note
this with Kamp and Reyle’s ‘abut’ relation: e s. Admittedly, Kamp and Reyle’s
(1993) rules for the interpretation of narrative discourse do not generate this structure,
but we can assume that it is possible to develop a more refined version of DRT which
yields the desired interpretation. However, this is not sufficient to interpret (108b).
(108b) allows different temporal structures, depending on the rhetorical relation estab-
lished between the two sentences. The room not being very bright can be the result of
4.3. LOGICAL OPERATORS: NEGATION AND QUANTIFICATION 125
Paul switching on the light, if the light is not strong enough. This reading can be de-
rived by negating the abut relation between s and e: :(e s). This is the only option
within Kamp and Reyle’s system of negation. However, there is another, maybe more
likely interpretation of (108b), namely one in which the room not being very bright is
the reason for Paul to switch on the light. In that case, we have a relation of expla-
nation between the two sentences in Lascarides and Asher’s terms, and the temporal
structure is one of immediate precedence: the temporal structure is one of immediate
precedence: the room not being bright immediately precedes the event of switching on
the light, and ends there. We could represent this as :s e. Given that a relation
of explanation can be established in (108b), but not in (108a), it is impossible to derive
the temporal structure of the negative sentence from its affirmative counterpart.
Further problems arise when we consider the way the discourse continues after the
negative sentence has updated the context. Kamp and Reyle’s null hypothesis predicts
that the temporal structure of negative sentences does not play a role in the further
discourse. This prediction follows from the claim that the eventuality variable is em-
bedded under negation, and the negative sentence itself does not denote a state or an
event (compare figures 16 and 17). Accordingly, we expect subsequent sentences to be
attached to the same reference point as the negative sentence. Thus Kamp and Reyle’s
treatment of negative events as subordinated to the main storyline implies that negative
sentences should be dispensible as far as temporal structure is concerned. However,
the discourses in (109) become incoherent if we leave out the negative sentence:
(109) a. The plane exploded. But John didn’t panic. On the contrary, he was very
calm.
b. John invited all his friends. They didn’t show up. So John decided to go out
into the street and bring in all the homeless from the neighbourhood.
c. John invited all his friends. They didn’t show up, although they had promised
to come.
In (109a) the state of being calm overlaps with the not panicking, rather than with
the event of the plane exploding. (109b) relates a sequence of events, one of which
is a negative one. It is the friends not showing up which triggers John’s decision to
go out in the street and bring in all the homeless. This makes it impossible to directly
attach the last sentence to the first one. Finally, the pluperfect in (109c) locates the
promise before the not showing up, and world knowledge suggests that the promise
follows the invitation. Under the null hypothesis, it is unclear how we can account
for the way negative sentences update the context in narrative sequences like those in
(109), because negative sentences are not supposed to shift the reference time.
The conclusion we draw from these data is that there are quite a few case where
negative statements have a role of their own to play in the temporal structure of narra-
tive discourse. Given that the rules for the interpretation of narrative discourse in both
Kamp and Reyle (1993) and Lascarides and Asher (1993) are formulated in terms of
eventualities, this suggests that an analysis in which the external referent of a negative
sentence is an eventuality variable (a state or an event) rather than just a time is justified.
Note however that our claim represents a weaker ontological commitment than it might
126 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
seem. Kamp and Reyle (1993), Amsili and Le Draoulec (1995) and others do not deny
that we need an external referent for negative sentences. Their idea is that the external
referent of negative sentences is a time, rather than an eventuality (compare figures
15–17 above). The only respect in which we strenghten this claim is that we argue that
the negative condition holding for t can be labelled as a state which co-extends with
t. The recursive rule for the introduction of discourse referents that we propose allows
us to switch from times to eventualities whenever needed. Note that our treatment of
negation differs from the one proposed by Higginbotham (1983), who claims that the
only negative events which are allowed in perception reports are those that somehow
describe some positive, active behavior. But as argued by van der Does (1992, 133),
this constraint is the result of more general aspectual restrictions on perception reports.
Negative sentences are more limited in their use than affirmative sentences (cf. Horn,
1989 for discussion). We take these restrictions to be pragmatic, rather than semantic
in nature.
The switch from times to states is a minor one in the sense that states are like
conditions in that they just require verification of embeddings, and do not involve the
notion of change. It is a major one in the sense that it allows us to integrated negation
in the model of aspect structure that we have developed so far. Adopting the recursive
rule for the introduction of discourse referents for negative statements means that we
open the way to a systematic study of the interaction of logical operators and aspectual
operators in various languages.
In (110a), the leaving can be interpreted as taking place after Mary came home, for
instance as a reaction to this event. In (110b), the situation of Max leaving is interpreted
as already going on when Mary came home. Nothing changes in the temporal relation
between the main and the subordinate clause when we add an adverb of quantification
like always:
The same effects of succession of events (111a) or overlap (111b) can be observed
here as in (110), so the Progressive takes narrow scope with respect to the quanti-
fier. The Progressive cannot take wide scope, because the quantificational relation as a
whole describes a complex state, and the Progressive requires is an event (or a process)
as its input description. The interpretation of (111b) is spelled out in figure 18:
figure 18
Max was always leaving when Mary came home
n s t x y
tn
Max(x)
Mary(y)
s =t t
Max(s)
e s’
s: es ) e s’
e: y come home s’: PROG(x leave)
(112a) denies that we are now in the state which results from the event of Mary
meeting the president. This is naturally rendered by embedding the Perfect under the
negation operator. Examples in which the Perfect takes wide scope with respect to
negation are given in (113):
(113) a. Mary hasn’t seen her grand-child for over a year (now).
b. Mary hasn’t written a single poem for twelve years (now).
The reading of (113a) with an explicit or implicit now means that Mary is in the
state of not-seeing her grand-child, this state of affairs has been going on for more than
a year now, and it hasn’t ended yet. (113b) means that the last time Mary wrote a poem
is twelve years ago, and the state of not-writing a poem has lasted since then and up
until now. We propose (114) as the grammatical structure of (113a):
128 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
The peculiarity of this use of the Perfect is that it is limited to stative predicates,
as Kamp and Reyle (1993, 567) observe. Interestingly, the negative sentences in (113)
behave like the stative predicate in (115a) rather than the non-stative ones in (115b)
and (115c):
This is not just a motivation for the wide scope interpretation of the Perfect over
negation, it is also an additional argument in favor of the view that negative sentences
are stative.
This completes our analysis of outer aspectuality in an event-based semantics as
far as English is concerned. In the next section, we will see that there are interesting
differences between the aspectual systems of English and Romance languages. They
will be handled as an instance of paramatric variation: English carries its aspectual
information in Asp , whereas Romance carries part of its aspectual information in
Asp and part of it in Tense. Given that Tense is higher in the tree than Asp , this has
important consequences for the interaction between aspectual and logical structure.
The claims will be illustrated by a detailed discussion of French, but can be extended
(with minor variations) to other Romance languages.
The DRSs we construct for these sentences are similar to the ones given in figures
3 and 4 above, except for the condition n t which interprets the future tense as non-
past. But how do the simple past tenses in French fit into the picture sketched so far?
There are well-known differences between English and French in this respect.
13 We will leave the discussion of certain differences between the French Passé Composé and the English
Present Perfect for another occasion (cf. Kamp 1991 and Vet 1992).
4.4. ASPECT AND LOGICAL OPERATORS IN ROMANCE 129
The Simple Past tense in English is aspectually neutral and ‘transparent’, just like
the Futur Simple in French: it applies to states, processes and events alike, and the sen-
tence inherits its aspectual character from the eventuality description the tense operator
applies to. The two simple past tenses in French differ in aspectual nature. According
to Kamp and Rohrer (1983), the Passé Simple sentences in (117a), (118a) introduce
events into the DRS, and the Imparfait sentences in (117b) and (118b) introduce states.
Given our more fine-grained event ontology, it is more appropriate to say that the Passé
Simple/Imparfait contrast is sensitive to the homogeneous/quantized distinction, so that
the sentences in the Passé Simple describe events, and the ones in the Imparfait refer to
either states or processes:
The fact that the sentences in (117) have an underlying stative eventuality descrip-
tion, whereas the ones in (118) involve an event predicate does not influence the role
these sentences play in narrative discourse. In the larger context of (119) and (120),
the Passé Simple sentences move the story line forward, whereas the corresponding
Imparfait sentences describe the background to the action:
(119) a. Georges annonça sa résignation. Anne fut triste.
Georges announced-PS his resignation. Anne was-PS sad.
b. Georges annonça sa résignation. Anne était triste.
Georges announced-PS his resignation. Anne was-IMP sad.
the difference between the simple and the However, there are important differences
between an aspectual operator like the English Progressive and the French past tenses.
A first observation is that the Progressive is sensitive to the stative/dynamic distinc-
tion, whereas the Passé Simple and the Imparfait depend on the homogeneous/quantized
contrast. But even if we leave processes out of the discussion and just compare states
and events, there are a number of striking differences. Both the Perfect/Parfait and the
Progressive are independent of tense: they are compatible with Past, Present and Future
tenses. A compositional approach thus sorts out the contribution made by the aspec-
tual and the temporal operator. But the Passé Simple and the Imparfait morphology is
composite, and encodes temporal and aspectual information in one morpheme.
Furthermore, there are differences in meaning between the English Progressive
and the French Imparfait. With respect to (120b), we noted that an event predicate
with the Imparfait leads to a progressive interpretation. In other contexts, the same
combination may give us the choice between a progressive (121) or a habitual reading
(122):
(121) a. Un jour, je faisais mes courses chez l’épicier quand je rencontrai Jean.
One day, I got-IMP my groceries at the grocery store when I ran-PS into Jean.
b. One day, I was shopping at the grocery store, when I ran into Jean.
English uses the Progressive in episodic sentences like (121b), but the Simple Past in
habitual sentences like (122b). In other words, the meaning of the Imparfait is broader
than that of the Progressive.14
Finally, the Imparfait combines with a larger class of predicates. States do not
normally combine with the Progressive, unless we find some way to coerce the state
into a dynamic situation. (123a) is unacceptable, and only the simple tense is felicitous
(123b):
In French, the Imparfait is often considered the unmarked tense for statives (124a),
but statives may also occur with the Passé Simple (124b):
semantic contribution of the perfective and imperfective aspect reduces to zero are de-
scriptively characterized as ‘unmarked’ cases. The unmarked cases make it attractive
to develop a semantics of the French past tenses which does not involve vacuous as-
pectual operators. Such an analysis would also shed new light on the observation that
the perfective/imperfective contrast is not expressed by independent morphology, and
that it is limited to the past tense domain.
We therefore propose to treat the Passé Simple and the Imparfait are aspectually
sensitive past tense operators. Specifically, Passé Simple sentences refer to events, and
Imparfait sentences denote states or processes. The intuition behind this claim is illus-
trated by drawing a parallel with the nominal domain. As we have seen, determiners
like much and many, little and few come in pairs that have the same interpretation, but
different selection restrictions: they apply to either mass or count nouns. The deter-
miner itself does not convert the noun into a mass or a count expression; all it does is
to select for a mass or a count expression. In the unmarked case, determiners like many
and few combine with typical count nouns, such as chair, student, and determiners like
much and little operate on typical mass nouns, such as water, gold. If the predicate does
not fit the requirements on the determiner, coercion comes into play, giving rise to spe-
cial meaning effects. The French past tenses are similar to the determiners much/many,
little/few, in that they do not convert the description into an aspectually different one,
but that they select for a homogeneous or a quantized eventuality description.
This idea can be implemented in DRT in the following way. Both the Imparfait
and the Passé Simple introduce a condition t n into the DRS. The aspectual selec-
tion restrictions are treated as input requirements on the operator. The Passé Simple
requires a quantized event description as its input, and the Imparfait is felicitous only
if it applies to a non-quantized, homogeneous description (a state or a process). These
constraints are part of the construction rule (appendix, part C). The Passé Simple is
thus used exclusively to locate events in the past, whereas the Imparfait locates states
or processes in the past. In contexts in which the Imparfait applies to an eventuality de-
scription denoting a state or a process, or the Passé Simple to an event description, the
aspectual class of the description satisfies the selection restriction on the temporal op-
erator, and the well-formedness condition is fulfilled. As a result, the DRS-construction
of (125a) and (126a) proceeds as for their English counterparts in (125b) and (126b),
spelled out in the figures 3 and 4 in section 2 above. As far as the unmarked cases are
concerned, the French past tenses do not trigger any aspectual meaning effects beyond
the aspectual class of the eventuality description. Furthermore, we do not need vacuous
aspectual operators to derive the correct interpretation.
If the aspectual requirements of the Passé Simple or the Imparfait are not satis-
fied by the eventuality description, the tense operator cannot apply. A straightforward
extension of the analysis developed so far is to assume that a contextually governed
process of reinterpretation comes into play when the input conditions on an aspectu-
ally sensitive tense operator are not met. Thus I formulate construction rules which
check the input conditions of the tense operator and introduce a coercion operator if
the eventuality description does not satisfy these aspectual constraints (see appendix,
part C). If a state or a process description combines with the Passé Simple, we need
to come up with an appropriate aspectual transition which allows the state to be rein-
terpreted as an event. In such a situation, the construction rule for the Passé Simple
4.4. ASPECT AND LOGICAL OPERATORS IN ROMANCE 133
appeals to the operator Che in order to resolve the aspectual conflict. Similarly, the
combination of an event description with the Imparfait triggers the introduction of a
coercion operator Ceh , which reinterprets the event as a state or a process. As before, it
is the linguistic and extra-linguistic context which tells us how to interpret the coercion
operator. The embedding function requires the meaning effects of at least one of the
possible mappings to be supported by the context. If no appropriate reinterpretation
can be found, the DRS is not verified. The result is not an ungrammatical sentence (we
can build up a well-formed DRS), but an uninterpretable sentence (we cannot embed
it).
Consider the presentation of a stative description in the Passé Simple. The in-
troduction of a coercion operator Che leads to (127b) as the grammatical structure of
(127a). The interpretation of (127a) is given in figure 19:
(127) a. (Soudain,) Jeanne sut la réponse.
(Suddenly,) Jeanne knew-PS the answer.
b. [ PAST [ Che [ Jeanne know the answer ]]]
figure 19
Jeanne sut (PS) la réponse
n e t x y
tn
Jeanne(x)
Answer(y)
et
s
e: Che s: x know y
French does not have a morphological marker for inchoativity, but the value IN-
CHO of Che may be triggered by the combination of a state and the Passé Simple (127)
in the presence of soudain ‘suddenly’. The value BOUND is appropriate in (128a),
where Jeanne’s life has a definite beginning point and endpoint, which presents the
state as a quantized event. ADD-PREP adds a preparatory phase and an initial boundary
to the process in (128b):
In all cases, Che introduces the necessary event variable e to satisfy the input
condition on the Passé Simple.
134 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
English has aspectual operators like the Perfect and the Progressive which have
an output in the domain of states. However, there are no grammatical operators which
allow us to represent states as events. All aspectual transitions which have an output in
the domain of events are implicit and governed by a contextual process of reinterpre-
tation. Such aspectual transitions show up in the presence of an adverb like suddenly
or in five minutes, which are restricted to event descriptions, but the reinterpretation
of a state as an event is not visible in the morphology of the verb. French does not
have grammatical operators mapping states onto events either, but it has tense opera-
tors which apply only to homogeneous or only to quantized eventualities. Thus there
is a difference in the syntax-semantics interface for the two languages: the semantic
effects of sentences like (127)/(128) and (129) are the same, and their DRSs are identi-
cal, but only French makes the fact that the state is presented as an event visible in the
morphology of the verb, using an aspectually sensitive tense operator.
Given that French has no Progressive verb form, the transition of an event to the
state of the event being in progress is a free transition. A progressive reading may be
triggered by the combination of an event description and the Imparfait, as in (130a):
n h t x
tn
hÆt
Jeanne(x)
e y
h: Ceh letter(y)
e: x write y
English Progressive, because there are more free aspectual transitions that map events
onto homogeneous situations than just a progressive reading. The observation that
the French past tense operators are sensitive to the homogeneous/quantized distinction
does not lead us to expect that all tense operators in that language are specified along
this dimension. It is an idiosyncratic feature of certain determiners in English that they
are limited to either mass or count nouns, whereas other determiners are not specified
along the mass/count dimension. Similarly, it is an idiosyncratic feature of the French
past tenses that they are sensitive to the homogeneous/quantized distinction, whereas
other tenses in this language are aspectually neutral. Past tenses are more likely than
present or future tenses to develop such distinctions, because they often have a more
fine-grained semantics than the other tenses (Comrie 1976, 87). The general frame-
work developed here is thus rich enough to provide a fully compositional analysis of
aspect shift in French and English.
If we combine the analysis of duration adverbials as eventuality description mod-
ifiers with the interpretation of the French past tenses as aspectually sensitive tense
operators, we can make some interesting predictions concerning the combination of
duration adverbials and past tenses in French. Adverbials introduced by pendant have
the same selection restrictions as for-adverbials in English, and adverbials introduced
by en are similar to in-adverbials, as is clear from the examples in (132) and (133):
En-adverbials map events onto events that culminate within the period of time
specified by the temporal noun. Pendant-adverbials map states onto events consisting
of a bounded state, the duration of which is measured by the temporal noun.
If our analysis of the French past tense forms as tense operators, rather than as-
pectual operators, is correct, we do not expect to find duration adverbials taking wide
scope over Passé Simple and Imparfait. If duration adverbials take narrow scope with
respect to these tenses, and if the output of both en- and pendant-adverbials is an event
description, we predict that their combination with the Passé Simple will be unmarked.
This is indeed what we observe, as the following examples illustrate:
4.4. ASPECT AND LOGICAL OPERATORS IN ROMANCE 137
figure 21
Anne jouait (IMP) du piano pendant deux heures
138 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
n h t x y
tn
John(x)
Piano(y)
hÆt
e s’ mt
e = BOUND(s’)
h: Ceh two hours(mt)
dur(s’) mt
s’: x play y
A habitual reading is the most likely value for the operator Ceh in figure 21, but
other interpretations are not excluded, depending on the context.
If we accept the existence of aspectually sensitive tenses along with aspectual
operators, the grammatical structure in (1), repeated here as (136), predicts that the
combination of an aspectual operator and an aspectually sensitive tense operator leads
to an interpretation in which the tense operator takes wide scope over the aspectual
operator:
As observed in de Swart (1991, 221, 254), the role of the tenses changes when we
add a quantifier like toujours ‘always’. We can no longer use the combination of one
Imparfait and one Passé Simple, but we have to have either two Imparfaits or two Passé
Simples:
(138) a. Max partait toujours quand Marie rentrait,
Max always left-IMP when Marie came-IMP home
b. (Pendant deux mois) Max partit toujours quand Marie rentra
(For two months) Max always left-PS when Marie came-PS home
The Passé Simple/Imparfait contrast no longer applies to the relation between the
individual events of Marie coming home and Max leaving. In both (138a) and (b),
it remains vague whether the relation between the subordinate and the main clause
event is one of overlap or immediate succession, although in both cases there seems to
be a preference for a succession reading, due to the event-like character of the atomic
predicates. The tense difference affects the presentation of the quantificational relation.
The Imparfait in (138a) gives the relation an open, unbounded character and describes
it as an ongoing state of affairs. The Passé Simple in (138b) on the other hand, presents
the quantificational relation as a bounded situation which is closed off in time. This
suggests that the Passé Simple and the Imparfait take wide scope over the adverb of
quantification, and the requirement on parallel tenses is an agreement phenomenon.
Given that we analyzed the Passé Simple and the Imparfait as aspectually sensitive
tense operators, they necessarily operate on the highest discourse referent. Thus, they
cannot take narrow scope with respect to the adverb of quantification. The recursive
rule for the introduction of discourse referents introduces a state variable for the quan-
tificational relation as a whole. This satisfies the input conditions on the Imparfait (cf.
figure 22). In order for the quantificational relation to be compatible with the Passé
Simple in (138b), a coercion operator must be inserted in figure 23:
figure 22
Max partait (IMP) toujours quand Marie rentrait (IMP)
n s t x y
tn
Max(x)
Marie(y)
s =t t
Max(s)
e e’
s: es ) e e’
e: y come home e’: x leave
140 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
figure 23
Max partit (PS) toujours quand Marie rentra (PS)
n e t x y
tn
Max(x)
Marie(y)
et
s t’
Max(s)
s =t t’
e: Cse e’ e”
s: e’ s ) e’ e”
e’: y come home e”: x leave
The situation is presented as a single complete whole (whence the Simple Past), but
consists of an ongoing action (whence the Progressive). This suggests that the Progres-
sive is an aspectual operator, and both the Simple Past and the Imperfect are aspectually
sensitive tenses, similar to the Passé Simple and the Imparfait in French. Note that the
sentences in (139) cannot be used to describe a complete event as being in progress,
thereby confirming the hypothesis that the aspectual operator (the Progressive) takes
narrow scope with respect to the tense operator (the Simple Past).
Similar distinctions can be found in Slavic languages. Bulgarian has a morpholog-
ical opposition corresponding to the perfective/imperfective distinction. Furthermore,
there are two past tenses, the Aorist, and the Imperfective. The combination of the Im-
perfective with the Aorist leads to a meaning effect which is rather different from the
combination of the Perfective with the Imperfect. As Comrie (1976, 32) formulates it,
“the Imperfective Aorist takes a situation which is described by an imperfective form
(Imperfective), to give explicit reference to its internal complexity, and circumscribes
the situation by giving it a perfective form (Aorist). (...) The Perfective Imperfect, on
the other hand, takes a situation which would in itself be described by a perfective form
(Perfective), and then superimposes upon this imperfectivity.” Consider the example in
(140):
(140) Štom pukneše zorata, izkarvax ovcite navǎn.
as-soon-as broke.3sg.perf-imp dawn.the drove.1sg.imp-imp sheep.the outside
‘As soon as dawn broke (Perf Impf), I used to drive the sheep out.’
The Perfective Imperfect in (140) refers to a habitual situation using the Imperfect,
but each individual event is described as a complete whole by the Perfective. The
Imperfective Aorist in (141) describes a bounded action which is presented as a single
whole by the Aorist, but which has internal complexity referred to by the Imperfective:
(141) Pred mnogo ikoni ošte pop Stefan vodi djada Nedka. Te se spiraxa pred sv. Nikolaj
before many icons other father Stefan led.3sg.imp-aor old Nedko. They refl stopped.3pl.imp-aor before St. Nicholas
pred obrazite na arxangelite Gavrail i Mixail.
before images.the of archangels.the Gabriel and Michael.
‘Father Stefan led (Ipfv. Aorist) old Nedko in front of many other icons. They
stopped (Ipfv. Aorist) before St. Nicholas the miracle-worker, . . . , before the
images of the archangels Gabriel and Michael.’
There are no scope ambiguities in these cases: (140) gives us the ‘imperfective
of a perfective’, while (141) describes the ‘perfective of the imperfective’ in Comrie’s
terms. In combination with the restriction of the Aorist/Imperfect opposition to the
past tense, these observations suggest that Bulgarian (but also Georgian for instance)
has both aspectual operators (the imperfective/perfective, which we analyzed as sitting
in Asp ) and aspectually sensitive tenses (the Aorist and the Imperfect).
Further research on the relation between aspect shift in French and English on
the one hand, and other Romance languages such as Spanish and Portuguese, but also
Slavic languages like Bulgarian is needed to investigate these issues in more detail.
The data provided here support a general theory of aspect shift in which aspectual
142 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
4.5 Conclusion
In this chapter, we have developed a cross-linguistically valide event-based semantics
of outer aspectuality. This model involves a mixture of universal and language specific
characteristics. The interpretation of negation and quantification as eventuality de-
scription modifiers is taken to be universal. Thus, all logical operators in all languages
are expected to end up in Asp . The interpretation of aspectual information such as
the Perfect and the Progressive in English or the Pass/’e Simple/Imparfait contrast in
French depends on the aspectual system of the languages in question. We argued that
the Passé Simple and the Imparfait are not interpreted as aspectual operators (in Asp ),
but as tense operators (in Tense) which are sensitive to the aspectual nature of the
eventuality description they operate on. As a result, the French past tenses always take
wide scope with respect to logical operators like negation and quantification. We have
shown that the treatment of negative sentences as states does not exclude the possibil-
ity of them being presented as events, as long as we recognize the crucial role coercion
plays in both English and French. We can also distinguish between the negation of
events and the negation of states, if we define appropriate restrictions on the kind of
aspectual transitions which are permitted, and take into account the rhetorical structure
of the discourse. Our proposal can be viewed as a generalized version of the analysis
of aspect and logical operators developed by Kamp and Reyle (1993).
Appendix
In the Appendix we introduce the construction rules, the DRS-conditions, and a verifi-
cation procedure necessary to construct and interpret the DRSs given in the text, in so
far as they are different from the DRSs constructed by Kamp and Reyle (1993).
A. General notions
DRSs are confined to a given vocabulary V and a set R of discourse referents
The universe of discourse UK of the DRS K confined to V and R is a sorted do-
main. This is a set of discourse referents of possibly different types including:
individuals, times, amounts of time, events, processes, states, organized into their
supercategories of homogeneous/non-homogeneous and dynamic/stative eventu-
alities.
ConK is the set of DRS-conditions in K confined to a given vocabulary V and a
set R of discourse referents.
B. DRS-conditions for aspectual operators
4.5. CONCLUSION 143
Progressive:
If d: is a dynamic description (event or process) then
d
s: PROG
d:
is a state description.
Durative adverbs:
(i) If e: is an event description then e: IN x time( ) is an event descrip-
tion
(ii) If s: is a state description then e: FOR x time( ) is an event descrip-
tion
Coercion:
(i) If e: is an event description then
e
h: Ceh e:
is the description of a homogeneous situation (a state, or a process).
(ii) If h: is a description of a homogeneous situation (a state or a process),
then
h
e: Che h:
is an event description.
(iii) If s: is a state description, then
s
d: Csd s:
is the description of a dynamic situation (a process or an event).
C. Construction rules
Introduction of discourse referents for sentences modified by an aspectual oper-
ator:
(i) For progressive sentences of the form Prog(S ):
a. Introduce in UK : a new state discourse referent s
b. Introduce in ConK :
s: PROG K1
c. If is a dynamic description, introduce in UK1 : a new discourse ref-
erent p or e.
144 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
d. Introduce in ConK1 : p: or e: .
e. If is not a dynamic description, so does not satisfy the input condi-
tions on the Progressive, introduce in UK1 : a new discourse referent
d.
f. Introduce in ConK1 :
d: Csd K2
g. Introduce in UK2 : a new discourse referent s
h. Introduce in ConK2 : s:
(ii) For state/process sentences which undergo contextual reinterpretation Che (S ):
a. Introduce in UK : a new event discourse referent e
b. Introduce in ConK :
e: Che K1
c. Introduce in UK1 : a new discourse referent h.
d. Introduce in ConK1 : h:
(iii) For event sentences which undergo contextual reinterpretation Ceh (S ):
a. Introduce in UK : a new homogeneous discourse referent h
b. Introduce in ConK :
h: Ceh K1
c. Introduce in UK1 : a new discourse referent e.
d. Introduce in ConK1 : e:
Duration adverbials
(i) For sentences modified by a for-adverbial of the form (FOR x time(S )):
a. Introduce in UK : new discourse referents e and mt
b. Introduce in ConK : x-time(mt)
c. If is a state or a process description, introduce in ConK :
dur(s) mt or dur(p) mt and e = BOUND(s) or e = BOUND(p)
d. If is an event description, introduce in ConK :
h: Ceh K1
and introduce in ConK : dur(h) mt and e = BOUND(h)
e. Introduce in UK1 : a new discourse referent e.
f. Introduce in ConK1 : e:
(ii) For sentences modified by an in-adverbial of the form (IN x time(S )):
a. Introduce new discourse referents e0 and mt
b. Introduce in ConK : x-time(mt)
c. Introduce in ConK : e0 = e
d. If is an event description, introduce in ConK : dur(e) mt
4.5. CONCLUSION 145
Tense operators
(i) Past tense (English):
a. If S is the first sentence of the discoure, introduce in UK : a new dis-
course referent n which is identified with the speech time. If S is not
the first sentence of the discourse, continue with b.
b. Introduce in UK : a new time discourse referent t
where t is the location time of the sentence
c. Introduce in ConK : t n
d. If is a state description, introduce in ConK : s =t t
e. If is an event description, introduce in ConK : e t
(ii) Passé Simple (French)
a. If S is the first sentence of the discoure, introduce in UK : a new dis-
course referent n which is identified with the speech time. If S is not
the first sentence of the discourse, continue with b.
146 CHAPTER 4. THE SEMANTICS OF OUTER ASPECTUALITY
A function PredM which maps predicates onto their denotation such that:
(i) For each one-place predicate TAi over times, PredM (TAi ) is a subset of T
(ii) For each atomic event description Pi (ei , 1 ; : : : ; n ), PredM (Pi ) is a set of
tuples <ei , a1 , . . . , an > where ei 2 EM and a1 . . . an 2 UM .
(iii) For each atomic state description Pi (si , 1 ; : : : ; n ), PredM (Pi ) is a set of
tuples <si , a1 , . . . , an > where si 2 SM and a1 . . . an 2 UM .
(iv) <si , a1 , . . . , an > belongs to PredM (PROG(Pi )) for Pi an event description
iff si 2 SM and <si , a1 . . . an > belongs to (PROG(PredM (Pi ))).
(v) The coercion operator Ces is multiply ambiguous and has senses Ces1
. . . Cesn for the n free aspectual transitions defined as possible mappings
from events to states in the language under consideration. <si , a1 , . . . ,
an > belongs to PredM (Ces1 (Pi )) for Pi an event description iff si 2 SM
and <si , a1 , . . . , an > belongs to ITER(PredM (Pi )); <si , a1 , . . . , an > be-
longs to PredM (Ces2 (Pi )) for Pi an event description iff si 2 SM and <si ,
a1 , . . . , an > belongs to PROC(PredM (Pi )); <si , a1 , . . . , an > belongs to
PredM (Ces3 (Pi )) for Pi an event description iff si 2 SM and <si , a1 , . . . ,
an > belongs to HAB(PredM (Pi )), etc.
(vi) The coercion operator Cse is multiply ambiguous and has senses Cse1
. . . Csen for the n free aspectual transitions defined as possible mappings
from states to events in the language under consideration. <ei , a1 , . . . ,
an > belongs to Cse1 (PredM (Pi )) for Pi a state description iff ei 2 EM and
<ei , ai , . . . , an > belongs to ADD-TO(PredM (Pi )); <ei , a1 , . . . , an >
belongs to Cse2 (PredM (Pi )) for Pi a state description iff ei 2 EM and <ei ,
ai , . . . , an > belongs to INCHO(PredM (Pi )); <ei , a1 , . . . , an > belongs
to Cse3 (PredM (Pi )) for Pi a state description iff ei 2 EM and <ei , ai , . . . ,
an > belongs to BOUND(PredM (Pi )), etc.
Semantic effect of aspectual transitions (informal)
149
150 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
In (142a), the two events are consecutive: Marie starts making her phone call after
Pierre has come in. In (142b), Marie is already talking on the phone when Pierre comes
in.
Kamp and Rohrer propose to handle this moving on in time by assuming that ref-
erence times progress with the story line. All sentences are interpreted not only with
respect to the utterance time, but also with respect to some contextually determined
reference time. This makes it possible for the sentence to fit in with the event struc-
ture built up by the discourse so far. The dependency on the current reference time
is responsible for the anaphoric nature of independent clauses. In French, the current
reference time is provided by the last sentence in the Passé Simple. Kamp and Rohrer
assume that sentences in the Imparfait include their reference point while sentences in
the Passé Simple generally imply a relation of succession with their reference point.
This accounts for the “dynamics” of Passé Simple sentences which carry on the main
story line (142a), and it contrast with the “stative” nature of Imparfait sentences (142b).
We can give similar examples for English. An adequate theory of the temporal
structure of narrative discourse should capture the “dynamic” nature of (143a) versus
the “stative” nature of sequences like (143b):
(143) a. John bought a one-way ticket, hopped on the bus, and bought himself a new
pair of shoes.
b. He went to the window and pulled aside the soft drapes. It was a casement
windown, and both panels were cranked out to let in the night air.
(144) Sue switched off the light. It was pitch dark in the room.
The first sentence denotes an event, and thus provides the reference time for the
second sentence. The second sentence denotes a state, so we would expect the state of
it being pitch dark in the room to include the event of Sue switching off the light. This
is obviously not how we interpret the discourse. In order to solve this problem, Hin-
richs proposes that each new past-tense event sentence occurs within the then-current
reference time. It subsequently causes the reference time to shift to a new value, later
than the original one. The new value of the reference time becomes the current one
for the next sentence of the discourse to be processed. States, activities and sentences
in the progressive are required to include the current reference time. Unlike events,
states do not shift the reference time to a later time. The fact that event sentences shift
the reference time forward allows the stative sentence of (144) to include the current
reference time, without therefore including the event of switching off the light.
5.1. TEMPORAL STRUCTURE IN DRT AND DICE 151
The difference between the approach proposed by Kamp and Rohrer (1983) and
Hinrichs’ analysis has been described in terms of the contrast between “pulling” and
“pushing” the reference time. Aside from that difference, the two proposals share some
important properties:
(i) They use the Reichenbachian notion of reference time to account for progress in
time at the discourse level.
(ii) The analyses are procedural in the sense that tenses are interpreted as instructions
to develop certain discourse structures.
(iii) They use the aspectual distinction between states and events to account for the
difference in behavior between sentences in the Imparfait, statives and Progres-
sives on the one hand and sentences in the Passé Simple, and events on the other
hand.
The three features of the DRT analysis of tense and aspect in discourse have led to
very sophisticated analyses of English and other languages. However, in recent years,
researchers have pointed out some shortcomings of this model. This has led to a debate
about the claims in (i) and (iii), but what remains is the essential insight that discourse
interpretation is procedural, and tenses are interpreted as instructions to update the
discourse representation structure (ii).
(145) a. Our heros were playing a complicated game. Sue threw the ball at Phil. Phil
barely caught it, but quickly passed it on to Hilary. Hilary mad a wrong move
and missed the ball.
b. Our heros were going through a lot of changes. Sue got married and left for
Ohio. Bill graduated and got a job in San Francisco. Eve went on a trip
around the world.
(146) a. Hilary entered the room. Phil was reading in his chair.
b. Hilary entered the room. Phil was happy to see her.
Both (145a) and (b) present a series of events after an introductory stative (Pro-
gressive) sentence. Although the aspectual properties of the two discourses are the
same, we interpret (145a) as a sequence of events, and (145b) as an unordered list of
events. Given that the rules provided by Hinrichs (1986), Kamp and Reyle (1993), or
similar accounts only take into account strictly linguistic information about aspectual
class, it is impossible to explain this difference in interpretation on the basis of these
152 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
rules. Similarly, both (146a) and (b) present first an event denoting, and then a state
denoting (Progressive) sentence. However, in (146a), we interpret the state as overlap-
ping the event, whereas in (146b), we interpret the state as following the event. The
rules formulated by Hinrichs allow both temporal structures, but at the price of an un-
derspecification of the actual interpretation of the discourse. On the basis of examples
of this kind, Lascarides et al. argue that we need to take into account extra-linguistic
information, such as knowledge of the world framed in lexical rules. They go one step
further and argue that this actually allows us to dispense with the notion of reference
time altogether.
In DICE (discourse and common sense entailment), the system which Lascarides
et al. develop to describe the temporal structure of texts, temporal relations are not
dependent on the reference time, but are derived from rhetorical relations. The triple
< ; ; > expresses that the representation of the text so far (of which is already
part) is to be updated with the representation of the current clause via a rhetorical
relation which attaches to . functions as the temporal antecedent for in the
discourse. As the default case, Lascarides et al. assume that the descriptive order of
events matches the temporal order, unless there is information to the contrary. This is
the essence of the rhetorical relation of Narration:
(147) a. Narration
< ; ; > > Narration ( ; )
b. Axiom on Narration
For describes e1 and describes e2 :
2 (Narration ( ; ) ! e1 e2 )
The connective > in (147a) indicates the default character of the implication. (147b)
uses the strict inference !, and claims that, if Narration( ; ) holds, then the discourse
must describe two events happening in succession. This is illustrated in (148):
(148) Max stood up. Bill greeted him
: Max stood up
: John greeted him
Narration ( ; )
9e1 [Stand-up(max, e1 ) ^ 9e2 [Greet(john, max, e2 ) ^ e e ]]
1 2
Other discourse relations overrule the default case and describe more specific cases.
These are accounted for by means of (versions of) the Penguin principle:
(149) Penguin Principle
' > ; > ; ' > : ; ' j= :
but not: ' > ; > ; ' > : ; ' j=
If we know that penguins are birds, that birds fly and that penguins don’t fly, then know-
ing that Tweety is a penguin will support the (default) inference that Tweety doesn’t fly,
rather than that Tweety flies. According to the Penguin principle, the more specific rela-
tion wins. It is used for cases like (150), which involve Elaboration. Elaboration( ; )
holds if the event in is part of the preparatory phase of the event described by
(Moens and Steedman (1988):
5.1. TEMPORAL STRUCTURE IN DRT AND DICE 153
(150) Elaboration
a. h; ; i ^ e v e > Elaboration( ; )
b. 2 (Elaboration( ; ) ! e e )
The action of drawing up plans is taken to be part of what it means to build a bridge,
so we can infer a relation of Elaboration in (151):
(151) The council built the bridge. The local architect drew up the plans
: The council built the bridge
: The local architect drew up the plans
Elaboration ( ; )
9e1 [Build(council, bridge, e1 ) ^ 9e2 [Draw(arch, plans, e2 ) ^ e2 e ]] 1
The Penguin principle is also used in other rhetorical relations, such as Explanation.
Explanation ( ; ) holds if the event described in clause causes the event described
in clause . Typically, causes precede effects:
(152) Explanation
a. h; ; i ^ Cause (e ; e ) > Explanation( ; )
b. Causes precede effects:
2 (Cause(e1 ; e2 ) ! :e1e ) 2
c. 2 (Explanation( ; ) ! :e e )
Given that causes cannot precede their effects (152b), Explanation typically shifts back
in time. If Max pushed John and thereby caused him to fall, the pushing typically
precedes the falling:
(153) Max fell. John pushed him
: Max fell
: John pushed him
Explanation ( ; )
9e1 [Fall(max, e1 ) ^ 9e2 [Push(john, max, e2 ) ^ Cause(e2; e1 ) ^ e e ]]
2 1
The resulting state follows the event described by the first sentence, because of the
causal relation between switching off the light and a room being dark. In cases where
no such specific information is available, states are interpreted as overlapping the
state/event described by the first sentence. If a state takes an event as its temporal
antecedent, the default relation is Background. Background ( ; ) holds if the state in
describes the circumstances under which the event in holds:
(156) Max opened the door. The room was pitch dark
: Max opened the door
: The room was pitch dark
Background( ; )
9e1 [Open(max, door, e1 ) ^ 9s2 [Dark(room, s2 ) ^ e s ]]
1 2
The system Lascarides and Asher develop is flexible, and more discourse relations can
be defined if needed. For instance, the relation between the president asking for support
and Jill raising her hand is stronger than just Narration, but can be defined as Reaction,
which involves a causal connection. Temporal sequencing is derived as a result of the
interpretation of the second event as a reaction to the first one.
Information about causation is not usually encoded in temporal morphemes. DICE
is a system which uses a mixture of world knowledge and linguistic knowledge to
describe the temporal structure of a text. The notion of reference time is thus not
indispensable for a description of the temporal order between independent clauses in
a discourse. Note that the system as discussed here only accounts for a sequence of
sentences in the (English) simple past, and ignores all other tenses, the role of time
adverbials and temporal connectives, etc. As such, it provides the bare backbone of a
theory of tense and aspect at the discourse level. For an extension to time adverbials
and temporal connectives, see Lascarides and Oberlander (1993) and de Swart (1999).
Section 5.1.3 provides an extension to other tenses, in casu the Pluperfect.
(157) makes tenses like the (English) simple past the default case. However, not all
tenses are so liberal. A clear example of a tense which does not only impose an absolute
location in time, but also contributes to the relative location in time of a state/event is
the Pluperfect. The Reichenbachian structure E–R–S requires the event time to precede
some reference time which itself precedes the speech time. Given that DICE does not
have the notion of reference time, we need to use other means to impose anteriority.
The first problem we face, is that the interpretation of a sentence in the Pluperfect
is dependent on the tense form of its temporal antecedent. We need to make sure that
the event e2 in (158)a) is anterior to the event e1 , but we should allow the series of
events e2 – e5 to form a narrative sequence:
(158) The doctor received a call from Mrs Smith (e1 ). Her husband had taken all his
pills at once (e2 ), had drunk his cough lotion (e3 ) and had lighted a cigarette (e4 ).
And then there had been an enormous explosion (e5 ).
If we want to base our analysis directly upon the rhetorical structure of the discourse,
we could propose a discourse interpretation rule along the following lines (compare
Asher and Lascarides 1993b):
(159) Discourse interpretation rule for the Pluperfect (first version)
a. A sequence Simple Past + Pluperfect is compatible with the following rhetor-
ical relations: Explanation, Elaboration, . . .
b. A series of two Pluperfect sentences is compatible with the following rhetor-
ical relations: Narration, Explanation, Elaboration, . . .
Although the definition is compatible with examples like (158), it is somewhat
unsatisfactory. Note that the definition is purely extensional, which implies that it is
subject to change every time we add a new rhetorical relation to the general frame-
work. This is related to the fact that the disjunction of rhetorical relations does not give
us any insight into the reason why we have these constraints. In other words, (159)
is an extensional definition of the requirement of the Pluperfect to express anteriority
with respect to some moment situated in the past, whereas we would rather have an in-
tensional definition of this characteristic which allows us to predict whether a rhetorical
relation is going to be compatible with the Pluperfect every time we update the general
framework. We can obtain this insight once we realize that with every discourse rela-
tion corresponds a given temporal relation (cf. Molendijk and Vet 1995, Landeweerd
1998):
figure 1
Derivation of the temporal relation from the rhetorical relation
Narration ! posteriority
Result ! posteriority
Explanation ! anteriority
Background ! simultaneity
...
156 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
We observe that several rhetorical relations can imply the same temporal struc-
ture. If a tense imposes constraints upon relative location in time, the actual temporal
structure can thus be the result of different rhetorical relations, as illustrated by figure 2:
figure 2
Constraints the temporal relation imposes upon the rhetorical relation
We can use the schema in figure 2 to constrain the rhetorical possibilities of the
Pluperfect in a more insightful way. We propose to replace definition (159) with (160):
(160) Discourse interpretation rule for the Pluperfect (final version)
A sentence S in the Pluperfect needs to find a sentence S 0 in the Simple Past with
which S establishes a rhetorical relation R implied in the relation of anteriority:
R = Explanation or . . . .
As far as the sequence of Simple Past + Pluperfect is concerned, this definition gives
the same results as (159a). Nothing needs to be said about a sequence of Pluperfect
sentences as long as we add the following assumption to our framework:
(161) A sentence can establish rhetorical relations with more than one other sentence
If we accept (161), we allow a sentence in the Pluperfect to establish a relation with
another sentence in the Pluperfect as long as we also find a sentence in the Simple Past
with respect to which it establishes a rhetorical relation compatible with the temporal
relation of anteriority. The combination of the two assumptions in (157) and (161)
allows us to derive a narrative sequence for the series of Pluperfect sentences in (159).
The Pluperfect is just an example of a tense which shows us what happens when
the tense itself imposes restrictions upon the relative location in time of the state/event
denoted. The analysis outlined here is sketchy, but shows how we can strike a balance
between linguistic and non-linguistic information, and obtain an analysis of tense and
aspect at the discourse level which combines insights from both DRT and DICE. Our
analysis of the Pluperfect sets the scene for our treatment of the French past tenses, the
Passé Simple and the Imparfait in section 5.2.
4 above, and argued that this provided us with the right characterization of negative
and quantified sentences in the Passé Simple and the Imparfait. The question which
arises in this chapter is whether the aspectual characterization of Passé Simple and
Imparfait sentences is sufficient to account for their discourse behavior. In other words,
does a sentence in the Passé Simple have the same discourse properties as an English
event sentence in the simple past, and does a sentence in the Imparfait have the same
discourse properties as an English stative or Progressive sentence in the simple past?
If that is the case, we need not say anything more, given our assumption (157) about
unconstrained relative location in time being the default case. However, Landeweerd
(1998) and de Swart and Molendijk (1999) argue that the Imparfait excludes certain
temporal structures, and Molendijk and de Swart (1999) make the same point for the
Passé Simple.
Both driving too fast and running a red light are good reasons to get a fine. So if we
are able to infer Explanation in (162a), we would expect to be able to do so in (162b)
as well. However, the sequence in (162b) is incoherent, and no relation of Explanation
can be inferred. Note that the contrast disappears when we use the Plus-que-Parfait in
the second sentence:
(163) a. Karin attrapa une contravention. Elle avait roulé trop vite.
Karin got-PS a fine. She had driven-PQP too fast.
b. Karin attrapa un contravention. Elle avait brûlé un feu rouge.
Karin got-PS a fine. She had run-PQP a red light.
Both discourses in (163) are well-formed, and the two sentences are related by Ex-
planation. We conclude that the incoherence of (162b) must be due to the use of the
Imparfait. As pointed out by Landeweerd (1998: 180 sqq), a closer look at the exam-
ples reveals some crucial differences between the eventualities involved. Driving too
fast is a process, but running a red light is an event. Assuming that one only gets ticket
once the illegal behavior has been clearly established, this means that one can be fined
for driving too fast as soon as and as long as one is driving too fast, whereas one can
only be fined for running a stop light when when one has actually run the light. The
example in (164) illustrates this in the form of the classical imperfective paradox:
(164) Karin brûlait un feu rouge quand elle aperçut un gendarme de l’autre côté de la
rue. Elle s’arrêta juste à temps.
158 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
Karin was running-IMP a red light when she noticed-PS a policeman on the other
side of the street. She stopped-PS just in time.
Karin gets a fine in (163b), but not in (164). We cannot set up imperfective paradoxes
for processes, because of the homogeneous character of these eventualities (compare
chapter 2 above). We can use this to explain the difference in acceptability between
(162a) and (b) if we assume that the Imparfait requires that its state is temporally
contiguous with the state or event denoted by the sentence the Imparfait sentence es-
tablishes a rhetorical relation with. De Swart and Molendijk (199a) formulate this
constraint as a blocking of temporal discontinuity:
(166) a. Julie entra à la dernière minute. Elle portait un suberbe manteau rouge.
Julie entered-PS at the last minute. She wore-IMP a suberb red coat.
b. Julie rentra. Le soleil lui brûlait les épaules.
Julie came-PS in. The sun burned-IMP her shoulders.
c. Julie alluma la lampe. Il faisait très clair dans la pièce (maintenant).
Julie switched-PS on the lamp. It was-IMP very light in the room (now).
d. # Il se mit à pleuvoir. Jean était tout mouillé.
It started-PS raining. Jean was-IMP all wet.
The Background relation in (166a) is the default interpretation in which the state
provides background information about some event. As a result, the state temporally
surrounds the event. The Imparfait is not in conflict with the temporal structure of
5.2. FRENCH PAST TENSES AT THE DISCOURSE LEVEL 159
According to Vet (1996) and others, a reverse order interpretation is not available for
sequences of sentences in the French Passé Simple; the inherent temporal progress
implied by the use of the Passé Simple would be opposed to it. Thus, an example like
(168a) is interpreted as referring to a situation in which Max pushed Jean after he fell.
The only way to obtain a temporal structure of anteriority is to use the Plus-que-Parfait
as in (168b):
(169) a. Jean décida de partir sans guide. C’est ce qui le tua, le pauvre homme. Il
tomba dans un abı̂me.
Jean decided-PS to leave without a guide. This is what killed-PS him, the
poor guy. He fell-PS into a ravine.
b. Le jour suivant, les Martiens attaquèrent Londres. Ce qui provoqua cette at-
taque, on ne le sait toujours pas.
The next day, the Martians attacked-PS London. What provoked-PS this at-
tack, we still do not know.
160 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
Jean fell into a ravine before he was killed (169a); the attack was provoked before
it was actually carried out (169b), and the monkey disappeared before we could find
it (169c), so all these examples illustrate cases in which the last sentence mentions
an event which temporally precedes an event referred to by a previous sentence. The
question arises why a description in reverse order is possible in the examples in (169),
while it is clearly blocked in the paradigmatic case (168a).
Caenepeel and Moens (1994) argue that the possibility to tell events in reverse
order depends on the type of text in which they occur. They claim that non-narrative
discourse is compatible with reverse order, but narrative discourse is not. At first sight,
this might help us to explain the difference between the English sequence in (167)
and its French counterpart in (168a). The English simple past is easily used in deictic
contexts, which signal non-narrative discourse, whereas the French Passé Simple is
traditionally characterized as a narrative tense. Thus, (168a) would have to be part
of narrative discourse, which would explain why an interpretation in terms of reverse
order is blocked. However, such an approach would leave the felicity of the discourses
in (169) unexplained.
An alternative approach is developed by de Swart and Molendijk (1999) and Molendijk
and de Swart (1999). This analysis preserves the role of rhetorical structure, but impose
constraints upon the temporal structure with which the Passé Simple is compatible. In
section 5.2.2 above we have seen that the Imparfait requires temporal contiguity, which
is formulated as a blocking of temporal discontinuity. The characterizing property of
the Passé Simple is that it blocks temporal anteriority. This leaves open a relation
of posteriority or of (partial) simultaneity. The general formulation of the discourse
interpretation rule for the Passś Simple is given in (170):
(170) Discourse interpretation rule for the Passé Simple (blocking constraint):
If describes an eventuality e1 and is taken as the temporal antecedent of a
sentence in the Passé Simple, which describes an event e2 , then *e2 e1 .
This rule relies on the notion of temporal antecedent. The temporal antecedent of a
sentence S is normally identified with the sentence S 0 which functions as the rhetorical
antecedent in the structure h; ; i (compare section 5.1.2 above).
In order to spell out the blocking constraint in terms of positive options for the
temporal structure, we distinguish between the two cases of posteriority and (partial)
simultaneity. Remember that a rhetorical relation always implies a particular temporal
relation, but that a temporal relation can be the result of various rhetorical relations.
This leads to a disjunction of rhetorical relations along the following lines:
(171) The interpretation of the Passé Simple (positive options, first version)
The temporal antecedent of a sentence S in the PS is a preceding sentence S 0
with which S establishes:
5.2. FRENCH PAST TENSES AT THE DISCOURSE LEVEL 161
Examples which illustrate the cases in (171i) and (ii) are provided in (172):
In (172a), we find a narrative structure. The turning pale and fainting in (172b) can be
construed as results or consequences of the tyrant’s words. Both discourses exemplify
the temporal structure of posteriority. (172c) and (d) are cases of (partial) simultaneity.
In (172c), this is the result of a relation of Incidence, (172d) illustrates a relation of
Elaboration. The examples of partial simultaneity are crucial to demonstrate that the
Passé Simple does not require posteriority, but comes with the weaker constraint of
blocking anteriority.
The examples in (172) illustrate cases in which a preceding PS sentence provides
the temporal antecedent for the PS sentences at hand, and a rhetorical structure is cre-
ated. In some cases, the temporal antecedent of a sentence is not another sentence in
the linguistic context, but a deictic time (usually the speech time). This situation fre-
quently arises with deictic tenses such as the Présent (173a) or the (resultative) Passé
Composé (173b), but we also find it with the Passé Simple, as argued by Molendijk
(1996) in relation to examples like (173c):
The speaker of (173a) directly relates the general statement to the utterance time. Sim-
ilarly, (173b) locates the falling of the speaker prior to the speech time. In (173c), we
observe that the PS sentence ‘fut un grand compositeur de musique orchestrale’ does
not establish a rhetorical relation with the rest of the fragment. We do not locate Du-
moulin’s being a great composer before his visit to Paris or his great success, or even
simultaneous with either of these events. Rather, we interpret the sentence as meaning
‘we agree now that Dumoulin was a great composer of orchestra music’. This deictic
use of the Passé Simple requires a small revision of the rule (171). The final version is
given in (174):
(174) The interpretation of the Passé Simple (positive options, final version)
The temporal antecedent of a sentence S in the PS is a preceding sentence S 0
with which S establishes:
(i) a rhetorical relation R implied in the relation of posteriority: R = Narration
or Result or . . . ;
(ii) a rhetorical relation R implied in the relation of (partial) simultaneity: R=
Elaboration or Incidence or . . .
(iii) If S does not establish any rhetorical relation with its textual environment,
the temporal antecdent of the sentence is associated with the speech time
(deictic PS).
(175) a. Jean décida de partir sans guide. C’est ce qui le tua, le pauvre homme. Il
tomba dans un abı̂me.
Jean decided-PS to leave without a guide. This is what killed-PS him, the
poor guy. He fell-PS into a ravine.
b. Jean décida de partir sans guide. C’est ce qui le tua, le pauvre homme. Dix
minutes après son départ, il tomba dans un abîme.
Jean decided-PS to leave without a guide. This is what killed-PS him, the
poor guy. Ten minutes after his departure, he fell-PS into a ravine.
The last sentence of the discourse in (175a) cannot establish a rhetorical relation with
the immediately preceding sentence without violating the constraint against anteriority
imposed by the Passé Simple. However, the discourse interpretation rule in (174) does
not require the temporal antecedent of the sentence to be the immediately preceding
sentence. A closer look at the discourse reveals that the third sentence is logically
5.2. FRENCH PAST TENSES AT THE DISCOURSE LEVEL 163
connected to the first one, rather than the second one. This is confirmed by the felici-
tous insertion of a time adverbial picking up the event denoted by the first sentence in
(175b). The reverse temporal order is thus not the result of the rhetorical structure of
the discourse.
The choice of the temporal antecedent is also at stake in example (??b), repeated
here as (176):
(176) Le jour suivant, les Martiens attaquèrent Londres. Ce qui provoqua cette attaque,
on ne le sait toujours pas.
The next day, the Martians attacked-PS London. What provoked-PS this attack,
we still do not know.
In (176), the second PS sentence does not take the first PS sentence as its temporal
antecedent, but the utterance time. The fact that the main clause is in the Présent makes
it possible to use a deictic Passé Simple in this context. Again, the reverse temporal
order is not the result of the rhetorical structure of the discourse, and no violation of
the discourse interpretation rule for the Passé Simple is found.
The last case (??c), repeated here as (177) is a slightly more complicated version
of (175):
(177) Le singe s’échappa. Nous ne le retrouvâmes plus, car il disparut (aussitôt) dans
la forêt épaisse.
The monkey escaped-PS. We did not find-PS it anymore, for it disappeared-PS
(immediately) into the thick forest.
The third sentence cannot establish a rhetorical relation with the second sentence, for
this would create a temporal structure of anteriority. The fact that we can felicitously
add ausssitôt (‘immediately’) to the last sentence supports an analysis in which the
third sentence is connected to the first one by Narration (or a stronger version of this
relation like Occasion or Consequence). What complicates this case is the presence of
the causal connector car, which establishes an explanatory relation between the not-
finding and the disappearing. If the causal link between the two clauses is so strong,
how can we defend that the temporal antecedent of the car-sentence is found in an
earlier sentence of the discourse? The work on argumentative structure by Ducrot
(1983), Anscombre (1989), Nolke (1995) and others suggests a possible answer to
this question. They observe that causal relations can be asserted (by connectives like
parce que ‘because’) or presupposed (by connectives like car ‘for’). Examples like
(177) suggest that if the causal relation is presupposed, the argumentative structure
can be preserved, even if the explanandus does not take the explanans as its temporal
antecedent. Indeed, it is easy to find examples of reverse temporal order with car or en
effet (‘indeed’), whereas we hardly ever encounter parce que in this construction. For
more discussion of the relation between argumentative and temporal structure, we refer
to Molendijk and de Swart (1999). We just conclude here that the examples in (175)–
(177) actually confirm our claim that the Passé Simple is incompatible with a temporal
relation of anteriority. Both the Imparfait and the Passé Simple are thus examples of
tenses which constrain the possible rhetorical relations which can be established by the
discourse by imposing restrictions upon the temporal structures with which the tense is
compatible.
164 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
The Passé Simple (PS) in (178a) suggests that Paul smiled after Marie looked at
him, possibly as a reaction to her looking at him. The Imparfait ( IMP) in the second
sentence of (178b) describes a state of Paul smiling which overlaps with the event of
Marie looking at him. This is in line with the view that Passé Simple sentences refer
to events in the DRS, and Imparfait sentences denote states. We can consider this the
well-established view of the PS/IMP contrast, independently of the question whether
this information is analyzed in terms of aspectual operators or in terms of aspectually
sensitive tense operators. Defenses of this view can be found in Kamp (1981), Kamp
and Rohrer (1983), Molendijk (1990), etc. The difference between states and events
has consequences for the temporal structure of the discourse. States include the current
reference time provided by a preceding event sentence. In a sequence of event sen-
tences, an event typically moves the reference time, and gives a further development
of the story line. There is no difference between affirmative and negative sentences in
this respect. The sequence of Passé Simple sentences in (179a) describes succession
of events, whereas (179b) states that Paul’s not-smiling temporally includes Marie’s
looking at him.
In order to represent these examples in DRT, we have to decide where the infor-
mation contributed by the Passé Simple and the Imparfait comes in. We assume with
Vet (1994) that French also follows the general schema in (180):
(181) a. Paul alluma la lampe. Il faisait très clair dans la pièce (maintenant).
Paul switched-PS on the light. It was-IMP very bright in the room (now).
b. Paul alluma la lampe. Il ne faisait pas très clair dans la pièce.
Paul switched-PS on the light. It NEG was-small IMP NEG very bright in the
room.
Molendijk (1990, 80) points out that the Imparfait sentence in sequences like
(181a) describes a state which coincides with the result state of the event of turning
on the light, described by the Passé Simple sentence. Thus we have the same temporal
structure e s as in (177a). Just as (177b), the discourse involving the negative
sentence in (181b) is ambiguous. It can either take the room not being very bright to
describe the result state of Paul switching on the light (under the assumption that the
lamp is not strong enough), or describe it as the state which explains why Paul switched
on the light. Thus, in French as well as in English, not all examples allow derivation of
the temporal structure of the negative sentence from its affirmative counterpart.
Furthermore, we already pointed out that the null hypothesis faces serious prob-
lems when we consider how the temporal structure of a discourse is updated after the
negative sentence has been processed. The examples can be repeated for French:
(182) a. Il y eut une énorme explosion. Mais Jean ne paniqua pas. Au contraire, il
avait l’air tout calme.
There was-PS an enormous explosion. But Jean NEG panicked-PS NEG. On
the contrary, he seemed-IMP quite calm.
166 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
b. Paul s’énerva. Ses invités ne venaient pas. D’ailleurs, ils ne vinrent pas. Paul
fut déçu.
Paul got-PS worked up. His guests NEG came-IMP. In fact, they NEG came-
PS. John be-PS disappointed.
c. Jean invita tous ses amis. Ils ne vinrent pas. Pourtant, ils avaient promis
d’être là.
John invite-PS all his friends. They NEG come NEG. However, they had
promised to be there.
(184) a. Depuis quelque temps, Jean courait après Pauline. ??Il l’attrapait.
Since some time, Jean run-IMP after Pauline. He catch-IMP her
b. Depuis quelque temps, Jean courait après Pauline. Il ne l’attrapait pas
Since some time, Jean run-IMP after Pauline. He NEG catch-IMP NEG her
(185) a. Depuis le début de l’année, Pauline était une étudiante assidue. ??Elle man-
quait un cours.
Since the beginning of the year, Pauline be- a diligent student. She miss-IMP
a class.
5.3. NEGATION IN DISCOURSE 167
b. Depuis le début de l’année, Pauline était une étudiante assidue. Elle ne man-
quait pas un cours.
Since the beginning of the year, Pauline be-IMP a diligent student. She NEG
miss-IMP NEG a class.
(186) a. This morning, Jean was running after Pauline. ??He was catching her.
a. This morning, Jean was running after Pauline. He caught her.
c. This morning, Jean was running after Pauline. ??He was not catching her.
d. This morning, Jean was running after Pauline. He didn’t catch her.
(186a) describes a situation of global temporal overlap, just like (184a) above. For
(186b) we assume that the event sentence takes the location time of the progressive
sentence as its reference time. The inclusion of the event in its reference time implies
that the event of catching occurs this morning. We can add a negation and claim that
there is no event of Jean catching Pauline during the entire time of Jean’s running after
her. In English, this would be expressed by (186d), rather than (186c). The aspec-
tual difference between the two sentences can be accounted for by the null hypothesis,
as is reflected in the DRSs constructed for (186c) and (186d) in figures 3 and 4. The
observation that the Imparfait sentence in (184a) has the same feeling about it as the
Progressive sentence in (186a) is compatible with the idea that both types of sentences
are stative. According to the null hypothesis, the treatment of Imparfait sentences as
referring to states automatically leads to the introduction of state variables under nega-
tion. In other words, if we would treat the Imparfait as an aspectual operator similar to
the Progressive, we would give the discourse in (184b) the temporal structure depicted
in figure 3, and we would expect it to have the same degree of relative inacceptability
as (186c). But this is not what we find. (184b) is perfectly coherent, and its meaning
is more like the one of (186d), depicted in figure 4. Under the null hypothesis it is not
possible to derive this interpretation, for this analysis gives aspectual operators neces-
168 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
figure 3
This morning, Jean was running after Pauline. He was not catching her.
n s t t’ x y t”
tn
sÆt
this morning(t’)
t = t’
Jean(x)
Pauline(y)
s: x PROG(run) after y
t” n
s’
s’ Æ t”
: s’ Æ t
s’: x PROG(catch) y
5.3. NEGATION IN DISCOURSE 169
figure 4
This morning, Jean was running after Pauline. He did not catch her.
n s t t’ x y t”
tn
st
sÆt
this morning(t’)
t = t’
Jean(x)
Pauline(y)
s: x PROG(run) after y
t” n
e
et
: e t”
e: x catch y
What these observations suggest is that the interaction between the Progressive
and negation is different from the interaction between the Imparfait and negation. It
may be true that the Progressive takes narrow scope with respect to negation, but this
is not the case for the Imparfait. In fact, we will argue that the Passé Simple and the
Imparfait always take wide scope with respect to negation. As a result, we cannot derive
the temporal structure of a negative sentence in the Passé Simple or the Imparfait from
their affirmative counterpart. Moreover, the examples discussed in this section show
that negative sentences in the Imparfait refer to states, whereas negative sentences in
the Passé Simple describe events. The difference in aspectual nature combined with the
claim that the Passé Simple and the Imparfait take wide scope with respect to negation
shows that negative sentences must refer to eventualities just like affirmative sentences.
We can combine the different parts of the picture and argue that the eventuality
variable for the negative sentence is introduced by the recursive rule for the introduc-
tion of discourse referents, and that the wide scope effects are due to the fact that the
Passé Simple and the Imparfait are analyzed as aspectually sensitive tense operators,
rather than as aspectual operators. This fits in with our claim that the PS/IMP take
wide scope with respect to aspectual adverbs introduced by in and for and quantifica-
tion (compare chapter 4, section 4.4). The data discussed here suggest that this claim is
also valid for negation. In this view, the Passé Simple and the Imparfait apply to event
and state variables respectively. According to the recursive rule for the introduction of
discourse referents (compare chapter 4), negative sentences refer to states. Building on
these two insights, we predict that the combination of negation with the Imparfait is
straightforward, whereas combination with the Passé Simple requires coercion of the
negative state. The addition of boundaries is sufficient to give the eventuality quantized
170 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
reference, and to treat it as an event in the DRS. By default, a sequence of two event
sentences relates temporal succession, whereas a sequence of an event sentence fol-
lowed by a stative sentence describes a situation of temporal overlap. The combination
of these claims accounts for the French version of the ‘smile’ examples in (??) and
(??), repeated here as (187) and (188):
(187) a. Marie regarda Paul. Il lui sourit
Marie looked-PS at Paul. He smiled-PS at her
Assuming that the Passé Simple and the Imparfait are tense operators, which take wide
scope over negation, we predict that the negative event in (188a) follows Marie’s look-
ing at Paul, whereas the negative state in (188b) includes Marie’s looking at Paul. The
fact that the Passé Simple triggers the reinterpretation of the negative state as an event
is reflected in the grammatical structures for (188a) and (b) in (189). The DRSs these
sentences give rise to are derived in a straightforward way from the results in previous
sections. They are given in figures 5 and 6:
(189) a. PS: [ Past [ Cse [ : [ Paul smile at Marie ]]]]
b. IMP: [ Past [ : [ Paul smile at Marie ]]]
5.3. NEGATION IN DISCOURSE 171
figure 5
Marie regarda (PS) Paul. Il ne lui souriait (IMP) pas
n e t x y s t’
tn
et
Mary(x)
Paul(y)
e: x look at y
t’ n
s =t t’
es
Max(s)
e’
s: : e’ s
e’: y smile
figure 6
Marie regarda (PS) Paul. Il ne lui sourit (PS) pas
n e t x y e’ t’
tn
et
Mary(x)
Paul(y)
e: x look at y
t’ n
e’ t’
e e’
s t”
Max(s)
s =t t”
e’: Cse e”
s: : e” s
e”: y smile
Other French examples are analyzed along the same lines. Consider (??b), re-
peated here as (190a) with its grammatical structure in (190b):
172 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
figure 7
Jean courait (IMP) après Pauline. Il ne l’attrapait (IMP) pas
n s t x y s’ t’
tn
s =t t
Jean(x)
Pauline(y)
s: x run after y
t’ n
s’ =t t’
Max(s’)
t s’
e
s’: : e s’
e: x catch y
The DRS captures the intuition that there is no event of catching during the time of the
run. Thus, this DRS resembles the one in figure 4, rather than the one in figure 3, as
desired.2
We have shown so far that the recursive rule for the introduction of discourse
referents combined with the interpretation of the Passé Simple and the Imparfait as as-
pectually sensitive tenses provides the right interpretation for all the French sentences
discussed. Given that all the components of this analysis have been independently mo-
tivated, this is an interesting result. The examples discussed so far have a relatively
simple discourse structure. They illustrate the default situation in which two event sen-
tences describe a succession of events (188a), an event sentence followed by a stative
sentence describes a structure of temporal inclusion of the event in the state (188b), and
two stative sentences describe global simultaneity of two states (190a). These temporal
2 The English examples come out slightly different in our theory than the DRSs constructed after the
null hypothesis. In particular, the negated boxes will be labelled as states. However, this does not affect the
comparison.
5.3. NEGATION IN DISCOURSE 173
structures are compatible with the rules for the discourse interpretation of the Impar-
fait and the Passé Simple, so they are easy to account for. However, as we already
observed above, the default relations can be overruled. Accordingly, Lascarides and
Asher (1993) and others have argued that we need to take into account the rhetorical
structure of the discourse.
Lascarides and Asher (1993) develop a framework in which a mixture of linguistic
and non-linguistic information involving causal relations and other world knowledge
determines the temporal structure of narrative discourse. Their work on English has
been extended to French by Asher and Bras (1993) and Landeweerd (1998). With
these researchers, we assume that aspectual information constrains the set of admissible
rhetorical relations. In particular, we claim that the Passé Simple does not allow us to
go back in time, whereas the Imparfait blocks temporal discontinuity. This is reflected
in the discourse rules we formulated in the previous sections, and which are repeated
in (191):
(192) a. Paul alluma la lampe. Il ne faisait pas très clair dans la pièce.
Paul turned-PS on the lamp. It NEG was-small IMP NEG very bright in the
room.
b. Il y eut une énorme explosion. Mais Jean ne paniqua pas. Au contraire, il
avait l’air tout calme.
There was-PS an enormous explosion. But Jean NEG panicked-PS NEG. On
the contrary, he seem-IMP quite calm.
c. Jean invita tous ses amis. Ils ne vinrent pas. Pourtant, ils avaient tous promis
d’être là.
John invite-PS all his friends. They NEG come NEG, although they had all
promised to be there.
174 CHAPTER 5. TENSE AND ASPECT AT THE DISCOURSE LEVEL
d. Paul s’énerva. Ses invités ne venaient pas. D’ailleurs, ils ne vinrent pas. Il
fut déçu.
John got-PS worked up. His guests NEG came-IMP NEG. In fact, they NEG
came-PS NEG. He was-PS disappointed.
e. Paul s’énerva. Ses invités ne venaient pas. ?? Il fut deçu.
Paul invited-PS all his friends. They NEG came-IMP NEG. He was-PS disap-
pointed.
In (192a), we can either derive a relation of Explanation (the room not being very bright
causes Paul to turn on the light) or a relation of Result (Paul turned on the light, but
the room is still not very bright). Both relations are compatible with the discourse rule
for the Imparfait, so it depends on the context which interpretation is most appropri-
ate. In (192b) the last Imparfait sentence establishes a Background relation with the
preceding Passé Simple sentence. The observation that Jean was all calm is not the
Background for the explosion, but for the event of not panicking. We can easily extend
our framework to deal with discourses like (192c). When the Plus-que-Parfait is used
to shift back in time from a point in the past, it requires a relation of anteriority with
respect to some moment situated in the past. In (192c), this moment is provided by
the friends not coming, rather than by John’s invitation. The analysis also provides a
straightforward explanation of the contrast between (192d) and (192e). The negative
state of the guests not-coming reported by the Imparfait sentence in (192e) provides the
Background for and thus overlaps with the event of Paul getting worked up reported in
the Passé Simple. The unbounded nature of the negative state of affairs suggests that
Paul is still waiting. In that context, Paul’s disappointment is not naturally related to his
getting worked up or him still waiting, so the discourse is not coherent. In (192e), the
negative Imparfait sentence also establishes a Background relation, but it is followed
by another negative sentence, this time in the Passé Simple. We can establish an Occa-
sion relation between the last two Passé Simple sentences: it is the event of his friends
definitely not showing up which leads to his disappointment. The quantized nature of
the negative event closes off the period of waiting. The natural interpretation is that
Paul is disappointed as a reaction to the guests definitely not-arriving, and follows it in
time.
The analysis sketched in this section combines a theory of rhetorical structure
with a set of temporal constraints on the French past tenses to account for the temporal
structure of narrative discourses involving affirmative and negative sentences alike. We
will argue now that the English ‘smile’ examples can be interpreted in a similar way.
The main difference is that the English simple past tense is not sensitive to the aspectual
nature of the eventuality description. The means that rhetorical structure opens up more
possibilities of different temporal relations. But this makes the situation harder to deal
with in some sense. We can coerce negative states into events, just like in French, but
the result is not visible in the morphology.
(194):
(193) a. Mary looked at Bill. He smiled.
b. Mary looked at Bill. He was smiling.
interface it results from which is shown in the DRS, and it need not come as a surprise
that figure 6 is the appropriate representation for both (187a) and (192a). If we compare
figure 6 with Kamp and Reyle’s representation of (194a) in figure 16 of chapter 4, we
observe that the two proposals lead to different but equivalent DRSs. This confirms our
claim that an event-based semantics of negative sentences incorporates all the insights
of an analysis which treats the external referent as a time.
This still leaves the past progressive version of the smile example to be accounted
for. The introduction of rhetorical relations triggering coercion might suggest that
posteriority should be a possible interpretation in (194b) as well, but of course it is not.
We base our explanation of this fact on Asher’s (1993) observation that states are closed
under negation, but events are not. We take the fact that states are closed under negation
to mean that negation applied to states has no aspectual effects. We assume that the
negation of a state is ‘transparent’ in the sense that the negative sentence inherits its
aspectual nature from its affirmative counterpart. Given that aspect constrains the set
of admissible rhetorical relations, this means that negative sentences inherit constraints
on discourse structure from their affirmative counterparts. This is formulated in the
following inheritance principle:
(196) Negation of states
The discourse relations that a sentence describing the negation of a state enters
into constitute a (not necessarily proper) subset of the admissible rhetorical rela-
tions for its affirmative counterpart.
As Dowty (1986) points out, the Progressive tends to express overlap with the
events described in the previous discourse. He suggests that this is due to the seman-
tics of the Progressive, which abstracts away from the start and finish of the situation.
The feeling of being ‘in the middle of things’ does not easily allow an inchoative or
a bounded event reading.3 If a Progressive sentence is not easily reinterpreted as an
event, it will not normally describe a response to another event. Thus, unlike (193a),
(193b) forces temporal overlap. If it is true that stative sentences preserve aspectual
structure under negation, we expect the same relation of temporal overlap to hold for
the negative counterpart of the progressive sentences in (194b). This interpretation of
(194b) is spelled out in figure 8:
figure 8
Mary looked at Bill. He wasn’t smiling
3 One exception Dowty notes concerns the kind of example in (i). An anonymous reviewer suggests a
n e t x y s t’ u
tn
et
Mary(x)
Bill(y)
e: x look at y
t’ n
s =t t’
Max(s)
es
u=y
s’
s: : s’ s
s’: PROG(u smile)
Figure 8 is not identical to the representation of the French Imparfait in figure 6. This
reflects the fact that we treat the Imparfait and the Progressive in essentially different
ways. However, the DRSs in figures 8 and 6 both rule out an occurrence of smiling over
a period of time including the event of looking. Although they differ in grammatical
structure, the sentences can therefore be used as translational equivalents. The reason
that (194b) expresses temporal overlap, whereas (194a) does not (necessarily) is that
(194b), as the negation of a progressive sentence, does not allow any other temporal
relation than its affirmative counterpart in (193b). Kamp and Reyle’s claim that the
restrictions on the interpretation of the negative sentence in (194b) are related to its
affirmative counterpart in (193b) is thus preserved under our analysis. However, unlike
Kamp and Reyle, we allow more room for variation. Aspect constrains the set of admis-
sible rhetorical relations, but it does not usually fully determine the discourse structure.
In a particular context, this allows a negative sentence to make a different choice from
among the set of admissible discourse relations than its affirmative counterpart. This
was illustrated above in (5), repeated here as (197a and b):
(197) a. Paul switched on the light. The room was very bright.
b. Paul switched on the light. The room was not very bright.
c. Paul switched off the light. The room was very bright.
means that both Result and Explanation are among the possible discourse relations the
affirmative stative sentence can establish. The discourses in (197) confirm our view
that the inheritance principles needs to be defined in terms of admissible, rather than
actual rhetorical relations.
5.4 Conclusion
In this chapter, we studied the discourse semantics of a number of English and French
tenses. We started with a discussion of the importance of aspectual information, con-
tinued with the relevance of rhetorical structure, and ended with an attempt to strike
a balance between linguistic and non-linguistic information. The resulting theory was
applied to negative sentences in French and English discourse contexts. It turns out that
our general approach to interpret negation lower than the Passé Simple/Imparfait but
higher than the Progressive in combination with a discourse theory of the French and
English past tenses yielded results equivalent to or better at certain points than com-
peting theories. Our proposal can be viewed as a generalized version of the analysis of
aspect and logical operators developed by Kamp and Reyle (1993) in combination with
mixed theories of discourse structure as advocated by Lascarides and Asher (1993) and
others.
References 179
References
[Bach 1981]Bach, Emmon. 1981. The Algebra of Events. Linguistics and Philosophy
9:5–16.
[Filip 1994]Filip, Hana. 1994. Aspect and the Semantics of Noun Phrases. In Tense
and Aspect in Discourse, ed. Co Vet and Carl Vetters. Trends in Linguistics. Stud-
ies and Monographs, Vol. 75, 227–256. Berlin and New York: Mouton de Gruyter.
[Freed 1979]Freed, Alice F. 1979. The Semantics of English Aspectual Complementa-
tion. Synthese Language Library, Vol. 8. Dordrecht: D. Reidel Publishing Com-
pany.
[Godard and Jayez 1993]Godard, Danièle, and Jacques Jayez. 1993. Towards a Proper
Treatment of Coercion Phenomena. In Proceedings of the Sixth Conference of
the European Chapter of the Association for Computational Linguistics, 168–177.
OTS Utrecht, The Netherlands.
[Gruber 1976]Gruber, Jeffrey S. 1976. Lexical Structures in Syntax and Semantics.
North-Holland Linguistic Series, Vol. 25. Amsterdam, etc.: North-Holland. Part
I. Studies in Lexical Relations (1964); Part II. Functions of the Lexicon in Formal
Descriptive Grammars (1967).
[Guentcheva 1990]Guentcheva, Zlatka. 1990. Temps et Aspect: L’exemple du Bulgare
contemporain. Collections Sciences du Langage. Paris: Editions du CNRS.
[Herbig 1896]Herbig, G. 1896. Aktionsart und Zeitstufe. Beiträge zur Funktionslehre
des Indogermanischen Verbums. Indogermanische Forschungen 6:157–269.
[Hermann 1927]Hermann, E. 1927. Objektive und subjektive Aktionsart. Indoger-
manische Forschungen 45:207–228.
[Jackendoff 1972]Jackendoff, Ray S. 1972. Semantic Interpretation in Generative
Grammar. Current Studies in Linguistics Series, Vol. 2. Cambridge (Mass.):
The M.I.T. Press.
[Jackendoff 1991]Jackendoff, Ray S. 1991. Parts and Boundaries. Cognition 21:9–45.
[Jacobsohn 1933]Jacobsohn, H. 1933. Aspektfragen. Indogermanische Forschungen
51:292–318.
[Kabakčiev 1984]Kabakčiev, Krasimir. 1984. The article and Aorist/Imperfect Dis-
tinction in Bulgarian: An Analysis based on cross-language ‘aspect’ parallelisms.
Linguistics 22:643–672.
[Kabakčiev 1989]Kabakčiev, Krasimir. 1989. On ‘Telicity’ and Related Problems. In
Proceedings of the Second Scandinavian Symposium on Aspectology, ed. Lars-
Gunnar Larsson, Studia Uralica et Altaica Upsaliensia, Vol. 19, 13–32. Uppsala.
Acta Universitatas Upsaliensis.
[Kamp and Reyle 1993]Kamp, Hans, and Uwe Reyle. 1993. From Discourse to Logic.
Introduction to Modeltheoretic Semantics of Natural Language, Formal Logic and
Discourse Representation Theory. Studies in Linguistics and Philosophy, Vol. 42.
Dordrecht: Kluwer Academic Publishers.
References 181
[Kiparsky 1998]Kiparsky, Paul. 1998. Partitive Case and Aspect. In The Projection
of Arguments: Lexical and Compositional Factors, ed. Miriam Butt and Wilhelm
Geuder. 249–291. CSLI Lecture Notes. Stanford: CSLI Publications.
[Kolni-Balozky 1960]Kolni-Balozky, J. 1960. A Progressive Russian Grammar. Lon-
don: Pitman & Sons. 6th edition. [1938].
[Krifka 1989]Krifka, Manfred. 1989. Nominal Reference, Temporal Constitution and
Quantification in Event Semantics. In Semantics and Contextual Expression, ed.
Renate Bartsch, Johan van Benthem, and Peter van Emde Boas. Groningen-
Amsterdam Studies in Semantics, Vol. 11, 75–115. Dordrecht: Foris Publications.
[Lascarides 1988]Lascarides, Alex. 1988. A Formal Semantics of the Progressive.
Doctoral dissertation, University of Edinburgh, Edinburgh.
[Lasersohn 1995]Lasersohn, Peter S. 1995. Plurality, Conjunction and Events. Studies
in Linguistics and Philosophy, Vol. 55. Dordrecht: Kluwer Academic Publishers.
[LePore and McLaughlin 1985]LePore, Ernest, and Brian P. McLaughlin (ed.). 1985.
Actions and Events. Perspectives on the Philosophy of Donald Davidson. Oxford:
Basil Blackwell.
[Leskien 1919]Leskien, A. 1919. Grammatik der altbulgarischen (altkirchenslavis-
chen) Sprache. Heidelberg. 2 edition.
[Lindstedt 1984]Lindstedt, Jouko. 1984. Nested Aspects. In Aspect Bound. A Voyage
into the Realm of Germanic, Slavonic and Finno-Ugrian Aspectology, ed. Casper
de Groot and Hannu Tommola. 23–38. Dordrecht: Foris Publications.
[Link 1983]Link, Godehard. 1983. The Logical Analysis of Plurals and Mass Terms.
In Meaning, Use and Interpretation of Language, ed. Rainer Baüerle, Christoph
Schwarze, and Arnim von Stechow. 302–323. Berlin and New York: De Gruyter.
[Maslov 1959]Maslov, Ju. S. 1959. Glagol’nyj vid v sovremennom literaturnom bol-
garskom jazyke; Znachenie i upotreblenie. In Voprosy grammatiki sovremennogo
literaturnogo bolgarskogo jazyka, ed. S.B.Bernshtejn. 157–312. Moscow.
[Młynarczyk 1998]Młynarczyk, Anna. 1998. Aspectualiteit in het Nederlands en het
Pools. Master’s thesis, Instytut Germanistyki, Uniwersytet Warszawski. 111 p.
[Moens 1987]Moens, Marc. 1987. Tense, Aspect and Temporal Reference. Doctoral
dissertation, University of Edinburgh, Edinburgh.
[Mourelatos 1978]Mourelatos, Alexander P.D. 1978. Events, Processes and States.
Linguistics and Philosophy 2:415–434.
[Naumann 1996]Naumann, Ralf. 1996. Aspectual Composition and Dynamic Logic.
Doctoral dissertation, Heinrich Heine Universität, Düsseldorf. Habilitationss-
chrift.
182 References