’23::

.o

I:
..5
:1.
x
“.

 

 

 

“: -; ~
143.1,,‘Lm
.

u
m

g.) ”at
h‘ ’0.
.r,: .n, .

q
{ti-91%
9} 12;-
~ ~‘
0’ ‘ ‘
‘ g ‘3 ‘
- A
32")“ -
a: En . a 2' '
45,} 3"?“ .

“it

.39 ;

1:7
”-73:12.’
amt '

‘ 1 -"
Infilg‘u

”n

,

.. w
u
.vo-v
v

. is:

 

5.,
Hi:
‘i

 

Iiiiiiiiliiiiii\‘iifin‘s‘m'liii'i

mm

3 1293 01051 7906

 

 

 

 

 

 

 

 

 

 

 

 

 

I".
1

This is to certify that the

dissertation entitled

A STUDY OF TEMPORAL RELATIONS IN
NATURAL LANGUAGES WITH SPECIAL REFERENCE TO
CHINESE AND ENGLISH

presented by

Mingiang Zhou

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Linguistics

 

 

 

Major professor

Date 27 @515;- /973

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

LIBRARY
Mlchlgan State
UnlversIIy

 

 

 

PLACE II RETURN 80X to man this checkout from your wood.
TO AVOID FINES mum on or before duo duo.

DATE DUE DATE DUE DATE DUE

       

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A SH
\\

 

 

A STUDY OF TEMPORAL RELATIONS IN NATURAL LANGUAGES
WITH SPECIAL REFERENCE TO CHINESE AND ENGLISH

By

Minglang Zhou

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Linguistics and Germanic, Slavic,
Asian and African Languages

1993

 

A ST?

 

 

 

ABSTRACT

A STUDY OF TEMPORAL RELATIONS IN NATURAL LANGUAGES
WITH SPECIAL REFERENCE TO CHINESE AND ENGLISH

By

Minglang Zhou

It is tempting to assume that the study of temporal relations in
natural languages begins with verbs, but this is simply not true. Studies
have been focused on verbs in terms of theories of action rather than
temporal properties since Aristotle. In modern linguistics and philosophy,
how temporal relations are represented in natural languages is generally
ignored, while time is considered external as in possible world semantics,
where truth conditions for sentences are obtained against possible worlds at
a moment or interval of time. These approaches lead to paradoxes and
puzzles with respect to the representation of time in natural languages.

This study focuses on temporal relations represented in natural
languages. Linguistic evidence from Chinese, English and other languages
suggests that there are three dimensions of representation of temporal
relations in natural languages: Linear Time, Frame Time and Situation
Time. Linear Time is a set of points of time without duration but with
precedence ordered in relation to speech time in a linear structure.

Frame Time is a set of intervals of time, which are denoted by temporal

frame adverbials and within which 3 described event/activity takes place or

 

asmk
Time .
intcn.
Situat

comp?

I
LEI“ C;

adxefl

accour

IICIIFI

anapha
I

CIidcr
are un
ihOSC r
that it
ihCHIn

k.-

Parhci

 

 

a state holds. Frame Time represents temporal inclusion relations. Situation
Time is a set of instants or intervals of time denoted by verbs, where
intervals are designated as bound or non-bound by verb modifiers.
Situation Time represents part-of relations. These three dimensions exhibit
completely different logical behaviors that underlie entailment relations
between sentences with different verbs, tenses, aspects and temporal
adverbials.

The relations among Linear Time, Frame Time and Situation Time
account for a large range of linguistic phenomena, such as aspect
viewpoints, the present perfect puzzle, and the demonstrative use and
anaphoric behaviors of temporal expressions. For example, linguistic
evidence shows that the relations between Frame Time and Situation Time
are universal in verbal aspect interpretations, though grammaticalization of
those relations as aspect is subject to parameterization. This study claims
that these three dimensions of temporal relations and the relations among
them are universal in natural languages, though grammaticalization of a

particular relation is parameterized in a particular language.

Copyright by

Minglang Zhou
1993

iv

To my grandmother,

who cultivated a spirit in me so that I can achieve this much.

that ti

Barha
in m}
State I
ht‘gan
in \I h:
WSpc
had tir
Clarif}
Inade,
this “i
l
pr‘v‘ft‘ss

mm”1H

 

depam
dBCUSg
preSCnfl
Yen-Ht

Supeni

ACKNOWLEDGEMENTS

I am indebted to my dissertation committee members for the support
that they have given me since the first committee meeting.

My deepest gratitude goes to the chair of my dissertation committee,
Barbara Abbott. She has provided consistent guidance and encouragement
in my academic and professional life during my entire stay at Michigan
State University. In her class in semantics on noun phrase interpretations, I
began to get some ideas for this dissertation. It seems to be an unusual class
in which one gets interested in temporal relations, but it is where fresh
perspectives have arisen. In the whole process of writing this dissertation, I
had timely guidance from her so that I was able to avoid pitfalls and to
clarify myself. She read early drafts and revised drafts of each chapter, and
made numerous comments and suggestions which have greatly improved
this work.

I am grateful to Grover Hudson. He encouraged me in my
professional development, reading my papers and making helpful
comments. He was also supportive in his capacity as associate chair of the
department. I am also grateful to Carolyn Harford, with whom my
discussion about some syntactic phenomena in Chinese resulted in
presentations and publications. Of course, I want to express my gratitude to
Yen-Hwei Lin both as my committee member and as my graduate assistant

supervisor for her encouragement and confidence in me. She not only gave
vi

me sound advice in academics, but also gave me opportunities to gain
experience in teaching classes of different levels and experience in program
coordination.

My study at Michigan State University was made possible through
graduate assistantships from the Department of Linguistics and Germanic,
Slavic, Asian and African Languages, and a dissertation fellowship from
the Graduate School and College of Arts and Letters.

I also want to take this opportunity to express my thanks to Jeanette
S. DeCarrico. She not only introduced me to the world of linguistics in the
first linguistics class that I took back in China in 1985, but also helped me
get into the MA. and Ph. D. programs in linguistics in the United States.

Finally and most important of all, I want to thank my wife Wanli for
her love and support, and my daughter Mary for her understanding of my

excuses not to leave my desk and computer as often as she liked.

vii

 

List ..

 

TABLE OF CONTENTS

List of Figures

1

1.0
1.1
1.2
2

2.0
2.1

2.2
2.3

3.0
3.1
3.2

Introduction

Time, Actions and Verbs, Preliminaries

Problems Since Aristotle

An Overview of My Approach and Organization

Time, Verb Categorization, and Truth Conditions
Introduction

Previous Approaches to Verb Categorization

2.1.1 Vendler's Analysis and His Two Observations
2.1.2 Puzzles of Montague's Tense/Aspect Operator
2.1.3 Bennett & Partee's Interval Semantics

2.1.4 Dowty's Logical Operators and Inertia Worlds
2.1.5 Parsons' Event Semantics

2.1.6 Aqvist's Tense Logic Approach

General Problems with Previous Approaches

Summary of Chapter Two

Linear Time and Frame Time

Introduction

Linear Time: One Dimension of Time

Representation of Linear Time in English

3.2.1 Tense, the Perfect and Linear Time

viii

xi.

12
22
22
24
25

36
43
49
54
59
61
63
63

74
76

'JJ
.JJ

4.3
4.4

5.0
5.;

 

3.3

3.4

3.5

4.0

4.1

4.2

4.3
4.4
4.5

5.0
5.1

3.2.2 Linear Time and Temporal Adverbials
Representation of Linear Time in Chinese

3.3.1 The Dual Function of Aspect Markers in Chinese
3.32 Tense and Temporal Adverbials in Chinese

Frame Time: A Second Dimension of Time

3.4.1 Frame Time and Frame Time Expressions

3.4.2 Syntactic Functions of Frame Time Expressions
Summary of Chapter Three

Situation Time

Introduction

Situation Time: A Third Dimension of Time

4.1.1 Nominalization and Reference to Situations

4.1.2 Situations and Their Temporal Properties
Quantification and Situations

4.2.1 Inherently Bound Situations and Quantification
4.2.2 Inherently Nonbound Situations and Quantification
Situation Predicate

Some Logical Properties of Situation Time

Summary of Chapter Four

Some Implications of the Three-Dimensional Temporal Theory
Introduction

Mechanisms for Verbal Aspect as Viewpoints

5.1.1 In Search of Mechanisms for Aspect as Viewpoints

5.1.2 Mechanisms: Relations between Frame Time and
Situation Time

5.1.3 Situations Viewed Externally
5.1.4 Summary

ix

85

89
94
98
103
107
1 15
1 1,7

~117

119
120
123
134
135
139
145
149
155
158
158
160
161

166
173
177

 

 

5.4

6.1
()7

o-

Biblit

5.2 The Present Perfect Puzzle
5.2.1 Previous Approaches
5.2.2 Relations among the Three Dimensions: Solutions
5.2.3 Semantics vs. Pragmatics in the Present Perfect
5.2.3 Summary

5.3 The Problem of Structural Analogies between Tense and
Pronouns
5.3.1 Partee's Proposal
5.3.2 Analogies on the Dimension of Frame Time
5.3.3 Summary

5.4 Summary of Chapter Five

6. Conclusions

6.1 Conclusions

6.2 Recommendations

Bibliography

1 78
179
1 90
I 98
202

204
206
208
213
214
218
218
224
227

LIST OF FIGURES

Figure 1: A Linear Representation of Time
Figure 2: A Representation of Linear Time

Figure 3: Temporal Inclusion in Frame Time

xi

65
70
101

 

 

 

assu

The P1

bi? Int

 

[Ci-WV
313W 0;:

I
E‘t‘ntt

0” lhc

 

Chapter One
INTRODUCTION

1.1 TIME, ACTIONS AND VERBS: PRELIMINARIES
In a philosophical study of the relation between time and actions, I

assume that the set of questions in (1) need to be answered.

(1) a. Does an action take time ?

b. If it takes time at all, does an action take a period of time or a
moment of time ?

c. If an action takes a period of time, is the period of time open-
ended or with well-defined end-points ?

d. How is the time that one action takes related to the time that
another action takes ?

e. How is the time that an action takes related to the time of
the utterance of a sentence describing the action?

The purpose of this study, however, is to answer the set of questions in (2)
by investigating temporal relations in verb categorization and the
representation of time in natural languages in a linguistically motivated
approach with the questions in (1) underpinning the questions in (2).
Eventually, answers to the set of questions in (2) are supposed to shed light

on the questions in (1).

(2) a. Is there any linguistic evidence for (1a), (1b) and (1c) in
natural languages ?

b. What is the linguistic evidence, if there is any, concerning
whether a period of time is open-ended or with well-defined
end-points ?

c. What is the linguistic evidence for (1d) and (1e) ?

1

 

Tem;
philc'
We r:
inlcrt
not n
tneth
IOlltw
Speed

attion

 

11?);1
l967:j

gencn

2

d. How can the linguistic evidence regarding the above
questions be (formally or informally) represented in the
syntax and semantics of natural languages?

Temporal relations in different classes of verbs are of some interest in
philosophy and linguistics, since verbs are used to describe human actions.
We may naturally assume that our interests in verb classes begins with our
interests in descriptions of our actions. This assumption is logical, but is
not necessarily the reality in our inquiry of actions and verbs. It seems to
me that, with regard to actions, philosophers focus their attention on the
following aspects: the agent (WHO did it ‘2), the act-type (generic or
specific), the modality of action (HOW did he/she‘do it ‘2), the setting of
action (WHEN, WHERE, and UNDER WHAT CIRCUMSTANCES did he/she do
it ?) and the rationale of action (WHY did he/she do it ?) (Rescher
1967:215-9). The question of the relation between time and actions is
generally ignored. Few philosophers appear to ask the set of questions in
(1), and even fewer philosophers appear to ask the set of questions in (2) ,
though several philosophers, such as Kenny (1963) and Vendler (1967),
raise some questions regarding these aspects in one way or the other
following the tradition of Aristotle. Linguists have begun to show interest
in temporal expressions, temporal relations in different classes of verbs
only in the last two decades (Dowty 1972). These questions deserve more
attention than they have received so far, particularly from linguists in the
classification of verbs, in the study of linguistic representation of time, and
in providing truth conditions for sentences. I assume that those questions in
(1) are the philosophical foundation of the present study, though this study
focuses on the set of questions in (2) from the linguistic point of view. I

look at the problem of temporal relations in verb classes and linguistic

 

reprt
undc

lEpR‘.
ttptt‘
adeq
relat;

the st

 

3

representation of time in general with those philosophical questions as
underlying guidelines, show how these aspects of actions are linguistically
represented in natural languages, and try to provide informal and formal
representations of these aspects of actions in verbs, which in turn can
adequately characterize each category of verbs in terms of temporal
relations, the truth conditions for sentences containing them, and ultimately
the semantic relations between sentences in terms of temporal relations.
The interest in verb classes can be traced back to Aristotle, who
probably pays more attention to the classes of verbs than any other
philosopher or linguist. Aristotle basically categorizes verbs into two
classes: process, and activity/stative verbs. In Metaphysics, Aristotle
(1048b, translation 1966:188-90) discusses the distinction between a
process and an act with regard to activities. He notices that a process verb
involves changing from one state of affairs toward another state of affairs
which is not yet present. One of the examples he uses is reducing one '3
weight which has slendemess as an end. In this process, there is a
movement from fatness to slendemess. Such movement is not considered as
an action, since it is not an end itself. If one is reducing his weight, he may
become slender or he may not become slender, since the movement is not
an end, i.e., slendemess in this case. On the other hand, an activity verb
does not involve such a movement from one state of affairs to another state
of affairs, but involves a movement with an end in itself. One of the
examples given by Aristotle is seeing. In seeing, one has sight as an end.
Thus, Aristotle calls the former 'a process' and the latter 'an act'. In
Metaphysics, Aristotle lists learning, house-building, reducing, healing and
walking as process verbs, and thinking, knowing, living well, being happy
as activity verbs. In De Anima, Aristotle (417a - 417b, translation 1902:

64:7

 

acti\
N 0 5
0f it.
Statit

Verbs

4

647) seems to distinguish a third class of verbs, stative verbs, from
activity verbs, as pointed out by Kenny (1967: 173). Aristotle talks about
two senses of knowing: acquiring knowledge and possessing knowledge.
Of the two senses, the former has an activity reading, while the latter has a
stative reading. It is not entirely clear that what Aristotle discusses in m

Anima is applicable to the differences between stative verbs and activity

 

verbs, since his focus is on knowledge.

It seems that in his classification of verbs Aristotle does not explicitly
raise the questions in (1). In other words, the notion of time is not
explicitly employed in his criteria of verb categorization, since Aristotle is
mostly concerned with verbs in terms of theories of actions. However, the
notion of time seems implicitly to underlie Aristotle's verb categorization.
For example, the definition of verb classes is definitely temporally oriented
in that a process verb involves changing from one state of affairs toward
another state of affairs which is not yet present. In fact, the two or three
classes of verbs appear to be contrasted in a temporal frame, though
Aristotle only looks into one aspect of the action denoted by a verb,
namely, the end or goal of the action in question.

In short, the questions raised in (1) do not appear to have received
enough attention in the study of actions in philosophy, and the questions in
(2) are not satisfactorily answered in the study of linguistic representation
of time in verb categorization and in linguistics in general, though the
relation between the notion of time and actions seems to be implicitly
underlying verb categorization in Aristotle. In fact, Aristotle is more

interested in verbs in terms of theories of actions.

5
1. 2 PROBLEMS SINCE ARISTOTLE

In the study of verb categorization, I think that there are three
general problems with previous approaches to the problem of verb
classification and the semantics of each category of verbs. The first
problem is the absence of uniform criteria in categorization and its
consequence -- the absence of uniform results given such criteria, as
indicated in the disagreement about the number of categories of verbs. The
second problem is methodological in essence, characterized by attempts to
classify each category of verbs in terms of the semantic relation between
sentences in different tenses and aspects before the first problem is settled
in an appropriate way. Thus, the criteria appear to be inadequate, and the
methodology appears to be defective. These two problems appear to
originate from the failure to acknowledge the two sets of questions raised
in (1) and (2) as important aspects in the study of actions, verbs and verb
categorization with respect to temporal relations. The third problem is a
direct consequence of the first two problems. Without clear consideration
of the notion of time in verb categorization and without uniformly
categorized verbs, there have been problems for approaches to correctly
represent temporal relations in semantics and in tense and aspect logic for
natural languages, since these aspects of verbs in natural languages are
essentially temporal relations. The first two of these three problems are
discussed one by one in detail below, while the third problem deserves a
full treatment in Chapter Two.

The first problem has its origin in Aristotle's classification of verbs.
For example, Aristotle does not make a distinction between stative verbs
and activity verbs in Metaphysics, though he seems to do so in De Anima.

The confusion of these two classes of verbs appears to result from the lack

 

 

P3} 3
from
3Ctit
Verb:
affai'
mos.
Calls

TESU

6

of clearly formulated criteria in the categorization of the verbs. It seems
that the main criterion Aristotle uses is whether the action denoted by the
verb has an end (goal) in itself or not. When a verb has an end in itself,
then it is complete and is an activity/stative verb. When a verb does not
have an end in itself, then it is incomplete. An incomplete verb has its end
outside it, and thus it involves a movement from one state of affairs to
another state of affairs, the latter of which may be not yet present at all.
Aristotle's label for this class of verbs is 'process verb'. Aristotle's
approach apparently leaves much room for diversity in verb classification.

Among modern scholars, Ryle (1949) is one of the first authors to
pay attention to verb classification, and notes that verbs like win and find
from Aristotle's process verbs have logical behavior quite different from
activity verbs or the other process verbs. Like process verbs, this class of
verbs involves a movement from one state of affairs to another state of
affairs. However, this class of verbs does not express the complete
movement but only the achievement of the latter state of affairs. Thus, Ryle
calls this class of verbs 'achievement verbs', which are often labelled
'resultative verbs' in traditional grammar. Achievement verbs always have
a counterpart activity verb as a complement to part of the movement that
changes from one state of affairs to another state of affairs. For example,
there are pairs like treat and heal, listen and hear, and look for and find.
Ryle does not make it clear that there are four categories of verbs, though
he notices the difference between achievement verbs and process verbs, nor
does he clearly realize that his achievement verbs represent a completely
different temporal relation from that represented by other verbs in
Aristotle's process class. Kenny (1963) calls the process verbs

'perfonnance verbs', while still keeping Aristotle's activity verbs and state

 

tert-
dich.
four
esscr

ECCI‘.

 

are A
catcg
on th

termi

 

not C

7

verbs. Vendler (1957, 1967) breaks away from Aristotle's tradition of
dichotomy or trichotomy in verb categorization, and classifies verbs into to
four classes in time schemata, though he still follows Aristotle's tradition in
essence. Vendler's four categories are stative verbs, activity verbs,
accomplishment verbs and achievement verbs. The last two classes of verbs
are Aristotle's process verbs. Consequently, we see some diversity in verb
category terminology which is briefly summarized in (3). In the literature
on this topic, authors may choose to adopt some of these philosopher's

terminology with or without arguments.

(3) Aristotle: activity ?state process
Ryle: ? ? achievement process
Kenny: activity static performance
Vendler: activity state achievement accomplishment

For a better illustration of the verbs involved in these categories, I present
each author's categories with sample sentences below. First, Aristotle's
three categories of verbs are represented by the following three English

sentences respectively, though he focuses on Greek verbs in his study.

(4) a. John talked. (activity)
b. John loved Mary. (state)
c. John reduced his weight. (process)

Secondly, Ryle's distinction between his achievement verbs and process
verbs within Aristotle's process verbs are illustrated in (5). However, it is

not clear what he has to say about Aristotle's activity verbs and state verbs.

(5) a. John lost twenty pounds. (achievement)
b. John reduced his weight. (process)

Thcl
addt

 

mutt
thtfitl
hbd

 

 

HnaH

Sentei

Thes

lnSlg

and

8

(5') a.?John conscientiously lost twenty pounds. (achievement)
b. John conscientiously reduced his weight. (process)

The contrast between (5a) and (5b) is more clear when certain adverbs are
added to these sentences, as in (5'), where, though, other aspects of the
action are involved in addition to temporal relations. Thirdly, Kenny's
three categories of verbs are the same as Aristotle's, as in (6), though his

labels are different from Aristotle's.

(6) a. John talked. (activity)
b. John loved Mary. (static)
c. John lost twenty pounds. (performance)
e. John reduced his weight. (performance)

Finally, Vendler's four categories of verbs are demonstrated in the English

sentences in (7).

(7) a. John talked. (activity)
b. John loved Mary. (state)
c. John lost twenty pounds. (achievement)
e. John reduced his weight. (accomplishment)

These examples show that the differences in terminology are mostly
insignificant. However, the classification of Aristotle's process verbs into
achievement verbs and process verbs by Ryle and into achievement verbs
and accomplishment verbs by Vendler are of great significance in
semantics and logic, since achievement verbs and accomplishment verbs
have very different logical behaviors and involve completely different
temporal relations. This will be illustrated in detail in the next chapter.

The second problem lies in the way the semantics of each category of

verbs is characterized. The problem begins with Aristotle's work. In

disc

 

proc-
rtlat
and

disti.
lies 1
10 w

malt

 

 

or sit
impc
prOg'
entiti
Sent.

Sc m.
Atis

0f .
L.

9

discussing the verb categorizing criterion, Aristotle arrives at his classes of
process verbs and activity/stative verbs by calculating the entailment
relation between the sentences with verbs in question in imperfective forms
and the sentences with verbs in question in perfective forms. The
distinction between a perfective verb form and an imperfective verb form
lies in that the former presents the totality of the event or situation referred
to without reference to its internal temporal constituency, while the latter
makes explicit reference to the internal temporal constituency of the event
or situation (cf. Comrie 1976:3-4). In English, for example, the
imperfective is in the progressive form, while the perfective is in the non-
progressive forms. A sentence with an activity verb in imperfective form
entails a sentence with an activity verb in perfective form, whereas a
sentence with a process verb in imperfective form does not entail a
sentence with a process verb in perfective form. A typical example from
Aristotle is (8), where (8a) is an example exhibiting the entailment relation,
while (8b) is an example exhibiting the non-entailment relation
(Metaphysics, 1048b, translation 1966:189).

(8) a. One who is living well has at the same time achieved the good
living (lived well).
b. He who is learning has not yet learned.

This tradition is carried over in Kenny's approach in treating the semantics
of each class of verbs. Kenny (1963:171-86) proposes (9), in which (It

stands for a verb and A stands for a subject noun phrase:

(9) a. Performance verbs: "A is ¢ing" implies "A has not tied".
b. Activity verbs: "A is thing" implies "A has oed".
c. Static verbs: "A has ¢ed" implies "A (1)8".

 

 

Vcrt

 

 

 

ACCt
anzn
K)€F
anCt
nknh
t9ln
Qch

andt

entn
10p“;
1967
that

shug
dent
CaSe
dent

Phnh

“i”.

10

(10) a. If a man is building a house, then he has not yet built it.
b. If a man is living in Michigan, then he has lived in Michigan.
c. If a man has loved a woman for many years, then he still
loves her.

Verbs in sentences corresponding to those in (9) are illustrated in (10).
According to Kenny's proposal in (9), build is a performance verb, live is
an activity verb and love is a stative verb, since the first clause is supposed
to entail the second clause in (9). However, it is not clear in the literature
since Aristotle that (9) is the definition for the classes of verbs or the
methodology in identifying the classes of verbs. This ambiguous status of
(9) may result from the lack of clear acknowledgement of the set of
questions in (1) and (2), and has its consequence in approaches to semantics
and to to tense and aspect logic.

The methodological problem lies in the inappropriate application of
entailment relations in the categorization of verbs. In the literature on the
topic of verb categorization (cf. Binnick 1991, Dowty 1979, Vendler
1967), on the one hand, it is claimed in the case of accomplishment verbs
that a sentence with a verb or verb phrase that denotes part of a
situation does not entail a sentence with a verb or verb phrase that
denotes a whole situation, while, on the other hand, it is claimed in the
case of activity verbs that a sentence with a verb or verb phrase that
denotes a whole situation does entail a sentence with a verb or verb
phrase that denotes part of a situation. Those authors fail to realize
the semantic and temporal properties stated in the bold letters above, as I
will discuss in detail in section 4.4. In addition, as criteria, the statements in
(9) do not always work in verb categorization. For example, it is clear that

(9) is inadequate to distinguish achievement verbs and accomplishment

 

Verl
perr
und

I'erl

11

verbs, though it may be adequate to distinguish activity verbs from process
verbs. Both Aristotle and Kenny rely on criteria or a method like (9) in
verb categorization. As a result, neither of them was able to recognize the
logical and temporal differences between achievement verbs and
accomplishment verbs, and instead, they both classify achievement verbs
and accomplishment verbs into one package with the label of 'process
verbs'. In addition, (9c) may not always appropriately characterize the
logic behavior of static verbs.

Ryle (1949) comes to grasp the difference between achievement
verbs and the rest of process verbs from a completely different
perspective. He notices that the action denoted by a process verb may be
under deliberate control by the agent, while that denoted by an achievement

verb is not under deliberate control by the agent, as is illustrated in (l l).

(11) a. ?John carefully lost twenty pounds. (achievement)
b. John carefully reduced his weight. (process)

The perspective from which Ryle finds the difference between achievement
verbs and the other process verbs does not necessarily indicate that his
categorization is directly related to the different temporal relations between
these two categories of verbs, though one can hardly find adverbs like
carefully in the company of a verb that represents an instant of time. It is
Vendler who finds that the logical difference between achievement verbs
and accomplishment verbs (e.g. the rest of the process verbs) is directly
related to temporal differences in these two classes of verbs, as shown in
(12) on the next page (Vendler 1967:103-7).

 

His

HUM

\ert
the l
mctl
sent.
merr
Oftc
relati

Tend

1.3

Chap:
phliOL

ans“ (

 

rePrex
EQDEr,
cmeg.
Prob] ,. i
103%

I
”latch;

 

relaim

12

(12) a. At what time did John lose twenty pounds ?
b.?At what time did John reduce his weight ?

His observation indicates that achievement verbs are related to an instant of
time, while accomplishment verbs are related to a period of time.

In summary, of the three problems related to previous approaches to
verb categorization the first two are discussed in this section. The first is
the lack of clearly formulated criteria and its consequences. The second is a
methodological problem, in which the entailment relation between
sentences in question seems to be the criterion in identifying verb category
membership at one time, and to presuppose clear membership of a category
of verbs at another time, resulting in circularity. Moreover, temporal

relations were not seriously considered in verb categorization until
Vendler's pioneering work (1957, 1967).

1. 3 AN OVERVIEW OF THE APPROACH IN THIS STUDY

To give a brief summary of what I have discussed so far in this
chapter, I note that there is a set of questions in (l) to be answered in a
philosophical study of actions and there is a set of questions in (2) to be
answered in linguistic studies of verbs, verb categorizations and the
representation of time in natural languages. I point out that there are three
general problems with previous approaches to the study of verbs, verb
categorization and temporal relations in natural languages. These three
problems are the lack of uniform criteria and its consequences, methodo-
logical problems, and problems for semantics and for tense and aspect logic
related to verbs, verb categorization and the representation of time.

In this study, as claimed at the beginning, I plan to investigate the

relations among verbs, verb categorization and the representation of time

Qt
II
n:
35
pr
ti n
lilt
0b
0::

(BC

ling
\

fail.

13

in natural languages in a linguistically motivated approach with the set of
questions in (1) as the philosophical foundation and with the set of
questions in (2) as the specific goal. Two of the three problems concerning
verb categorization are discussed in this chapter, and the third problem,
namely, the problem of verb categorization for semantics and for tense and
aspect logic, is reviewed in detail in Chapter Two. With these three
problems reviewed, I focus on the investigation of the representations of
time in natural languages and the relation between such representations and
the set of questions in (2). The linguistic representation of time is usually
observed to appear in the form of tense and aspect (Comrie 1976, 1986,
Dahl 1985, Smith 1991), in the form of time adverbials (Bennett & Partee,
1972, Heny 1982, Richards 1982) and in the form of adverbial clauses
(Bennett & Partee 1972, Dowty 1982). Of course, I argue that the
linguistic representation of time also occurs in verbs and verb phrases, as
Vendler (1957, 1967), Parsons (1985, 1990) and others have observed but
failed to recognize its importance.

Reichenbach (1947) has studied the temporal relations in natural
languages from the perspective of logic. He puts forward three concepts:
point of speech, point of event and point of reference. These three points
involve all the above forms of linguistic representation of time. However,
Reichenbach only considers the order of the three points to account for
tense and aspect, with the assumption that these three are linearly ordered.
Reichenbach's theory is both too strong (cf. Comrie 1981) and inadequate.
This study will show that the temporal structure in natural languages is
much richer than Reichenbach's theory assumes.

If real time is considered as linear, and if it develops from the past to

the future, then it is usually represented as a line with an arrow pointing

 

IO

ar
if]
lir
re]
[ht
8Cl

US

F01
rel:
t 19
acti
We
the

lake

 

 

14

towards the right hand side, the left side being the past and the right side
being the future, as in Benthem (1983), Comrie (1985), ter Meulen (1983),
and Smith (1991). This linear representation of time usually provides
temporal orders and locations, though some authors fail to realize the
limitations of it. I consider, as the representation of Linear Time, the
representation of the temporal order of occurrences of actions, relative to
the point of time of utterance and relative to the occurrences of other
actions, along this temporal line. The representation of Linear Time gives
us relative locations or orders of the occurrences of actions and utterances
along the temporal line.

First, how do tense and aspect represent time in this linear picture?
For example, Comrie writes "Since tense locates the time of a situation
relative to the situation of the utterance, we may describe tense as deictic."
(19762). In this sense, tense locates points of time of occurrences of
actions along the temporal line, relative to the time of the speaker's
utterance. For example, (13a) indicates that read a book took place before
the utterance of the sentence, while (13b) indicates that read a book is to

take place after the utterance of the sentence.

(13) a. John read a book.
b. John will read a book.

In contrast to his view of tense, Comrie considers aspect different from
tense in that "Aspect is not concerned with relating the time of the situation
to any other time-point, but rather with the internal temporal constituency
of the one situation."(l976:3). This seems to suggest that aspect does not
locate points of time along the temporal line, but rather represents some

temporal structure within points of time. Whether Comrie's view of

15

representation of time by tense and aspect is accurate and adequate is
further discussed in Chapter Three and Chapter Five in relation to the set
of questions in (2). For my purposes here, the two aspects in the sentences
in (14) seem to indicate certain temporal order of the actions in question,
though they do more than just present the temporal order, as Dahl
(1981:23-5) points out.

(14) a. John had read a book when Mary came.
b. John was reading a book when Mary came.

The past tense in (14) shows that the actions described took place before the
utterance of these sentences. The perfective aspect in (143) indicates that
read a book took place before come, while the imperfective in (14b) tells
us that come took place in the middle of read a book. In short, both tense
and aspect do represent the temporal order or temporal locations along the
temporal line.

Secondly, temporal relations are found in temporal adverbial clauses.
For example, after and before are frequently used to express the order of
one action in relation to another action along the temporal line. Even and
may express some temporal relations when it is used to conjoin two clauses

or more, as in (15).
(15) John went home and watched television.

Moreover, tenses and aspects are usually involved in tense sequences in
embedded clauses, which also indicate order along the temporal line. I
think that these representations of temporal relations in languages belong to
the representation of Linear Time.

How does my Linear Time compare to Reichenbach's (1947:288)

 

poiri
Lint-
c0\c
N 0
refer

Rcic

 

refer
differ
Line.-
rt‘prc‘
Line
In or

rclar

16

point of speech, point of the event and point of reference? I think that
Linear Time definitely covers Reichenbach's point of speech, and may
cover point of the event in a certain sense, as my discussion in the previous
two paragraphs indicate. Does Linear Time cover Reichenbach's point of
reference? I think that the answer is 'yes' and 'no'. This is because
Reichenbach's point of reference sometimes is the same as point of speech
and point of the event, in which case Linear Time covers point of
reference. However, Reichenbach's point of reference sometimes is
different from the point of speech and the point of the event, in which case
Linear Time does not cover the point of reference, if it is not linguistically
represented, as my discussion in Chapter Three will show. Essentially,
Linear Time just covers linear temporal relations in Reichenbach's theory.
In other words, Linear Time in my theory only represents precedence
relations in natural languages.

Thirdly, temporal relations in natural languages are also represented
by temporal adverbials. Temporal adverbials can be grouped into frame
adverbials, e.g., this morning, frequency adverbials, e.g., twice a year,
durative adverbials, e.g., from two pm to five pm, and punctual adverbials,
e.g., at eight am on Tuesday (cf. Bennett & Partee 1978) Temporal
relations specified by temporal adverbials may be also deictic in nature.
This morning is an example of this kind. Then, what is the relation between
temporal relations specified by tense and temporal relations specified by
temporal adverbials, since both are deictic in some respects? This relation
is studied in detail in Richards (1982), Heny (1982) and Dowty (1982).
What concerns me here is the period of time denoted by temporal
adverbials. Most of these adverbials contain interval—referring temporal

expressions. For example, this morning not only specifies a temporal

als

311i

ma
dis.
tal.’
puz
St‘l

syn

Tin

by \
Verb
lepri
mm
the r
freqi
I0 lb.
(195'

l7

location along the temporal line but, as an interval-referring expression,
also has a period of time along this line as its reference. I refer to such a
period of time as Frame Time, a term used in Bennett and Partee (1978)
and Smith (1991), in contrast to Linear Time.1 Frame Time can be defined
as the period of time denoted by a referring temporal expression, which
may function as an adverb or adverbial, such as the ones that I have
discussed above, and as a period of time within which actions and events
take place. Frame Time is an important notion in this study, since it is a
puzzle in interval semantics and is of particular interest in relation to the
set of questions in (2). The nature and role of Frame Time in semantics and
syntax is fully treated in Chapter Three. I believe that the notion of Frame
Time is of importance in semantics and in tense and aspect logic.

Finally, temporal relations in natural languages are also represented
by verbs and by verb phrases with noun phrases. Temporal relations in
verbs do not get enough attention, as I have discussed in 1.2., while the
representation of temporal relations in verb phrases with noun phrases in
natural languages is generally ignored in linguistics and philosophy, though
the relation between a noun phrase and a verb phrase containing it is
frequently studied from other perspectives. Of all the previous approaches
to the problem of verb categorization, it seems to me that only Vendler's
(1957, 1967) work comes close to the realization of temporal relations by
noun phrases in verb phrases, but fails to pursue these aspects of temporal

relations further. Vendler's work is given a complete review in the next

 

1 Bennett and Partee (1978:22) use the term 'frame time' with a sense similar to
what I am talking about here, but they dropped it without further development. Smith
(1991:155) has mentioned the term 'frame time' in her discussion of temporal adverbials,
but she does not seem to give it any theoretical significance. Parsons (1990:211-2) seems
to take a more theoretical approach to frame adverbial phrases. I will give a detailed review
of these approaches in section 3.4 in Chapter Three.

18

chapter. Intuitively, we know the answer to the question (1a), namely,
'actions take time', but linguistically it is not so clear in the literature as
how the intuitive answer is represented in natural languages. I believe that
temporal relations, as represented by noun phrases in verb phrases, are of
more direct relevance to the questions in (2a) and (2b) than temporal
relations represented by other linguistic forms that I have mentioned
above. The noun phrases in question function grammatically as object or
complement in verb phrases, and function thematically as patient/theme or

goal, as shown in (16).

(16) a. Mary read a book.
b. Mary ran some distance.
0. Mary pushed the cart.

Temporally speaking, there is a relation between the action denoted by the
verbs and references of the noun phrases. Intuitively, I think that for each
and every chapter of a book that Mary read there is a period of time taken.
The time taken in reading each chapter of a book makes up the time taken
in reading a book. Similarly, for each and every part of some distance that
Mary ran there is also a period of time taken. The time taken in running
each part of some distance makes up the time taken in running some
distance. The temporal relations in this kind of verb phrases are predictable
in the sense that the period of time taken is specified by the noun phrases in
some way. On the other hand, this relation does not seem to be found in
(16c). First of all, Mary did not have to push each and every part of a cart
in order to push a cart. Secondly, it is possible that Mary pushed a cart by
just pushing an inalienable or unmovable part of a cart. Thirdly, Mary

might have pushed a cart for a period of time of arbitrary length so that

 

tempoi
the ser

the CU}
Time e

have

 

genera.
I976).
reprem
\I'Orkir
lt’mpo
a \‘erb
its PFC
repres
t0 the

Cale Qt

TemI
impo
dlStir
(197.

19

temporal relations in the latter kind of verb phrases are unpredictable in
the sense that the period of time taken is not specified by the noun phrase,
the cart, in (16c). In contrast to Linear Time and Frame Time, Situation
Time covers the temporal relations in processes, states, events and activities
that verbs and verb phrases denote in natural languages. Situation is a more
general cover term for processes, states, events and activities (cf. Comrie
1976). Situations are considered instantiations of temporal properties
represented in verbs and verb phrases (cf. Gabbay & Moravcsik 1980). A
working definition for Situation Time is that Situation Time represents the
temporal relations internal to an event, activity, process or state denoted by
a verb or verb phrase. The linguistic representation of Situation Time and
its properties is the main topic in Chapter Four. By studying linguistic
representations of Situation Time, 1 think that I am able to provide answers
to the questions in (2a) and (2b) and to some problems in verb
categorization.

In semantics and logic, verb phrases are termed 'predicates'.
Temporal relations in verb phrases, as shown in the discussion of (16), are
important in marking the mass-count distinction in predicates —- a
distinction which is parallel to that in temrs as discussed by Hoepelman
(1976), Hoepelman and Rohrer (1980), and ter Meulen (1980), among a
number of authors, but it is, however, never made clear how such a
distinction is made in predicates. I assume that predicates, depending on
their category, are specified as count or mass by certain classifiers as some
nouns are, provided the noun phrases in question specify a predictable
temporal relation in the verb phrases. The parallelism is that if terms are
specified as count and mass linguistically by the presence of certain

classifiers or packages, then so are the predicates, as in (17) and (18).

20

(17) a. some water ( mass )
b. a glass of water ( count ( package ) )
(18) a. John drank some water. (activity)
b. John drank a glass of water. (accomplishment)

The difference is that the distinction between references of count terms and
references of mass terms is made along the dimension of space, whereas
distinction between the extension of the count predicates and that of the
mass predicates is made along the dimension of time in terms of boundness
and nonboundness in situations if situations are instantiations of temporal
properties. The topic of the mass-count distinction is also to be fully treated
in Chapter Four in terms of bound and nonbound situations. The question
of how Situation Time as one of the three dimensions of the representation
of time is different from the other dimensions of the representation of time
is also considered in Chapter Four.
How the three dimensions of time, that is, Linear Time, Frame

Time and Situation Time interact with each other in natural languages is
treated in Chapter Five, where I discuss, in light of the three dimensions of
temporal structure, the consideration of aspect as viewpoints or
perspectives (Comrie 1976, Smith 1991, Thelin 1990). This three-
dimensional representation of time in natural languages will present a fresh
view on the present perfect puzzle (Klein 1992) and the fallacy of
structural analogies between tenses and pronouns (Partee 1973).
Conclusions from this study will be presented in the last chapter, Chapter
Six.

To review my main points briefly, I plan to distinguish Linear Time,
Frame Time and Situation Time and investigate their relations with each

other and their difference in linguistic representations. I believe that the

 

niaho'
t0 spclh
represtl
study I

Signifit|
011 lingI
Calegm
“ith dii
Slanlit‘r
Imated

“entua

 

21

differences in the linguistic representation of time along the three
dimensions, Linear Time, Frame Time and Situation Time, are of great
importance in semantics and in tense and aspect logic. Linear Time is
commonly mentioned and adopted in semantics and in tense and aspect
logic as well. However, there are misunderstandings concerning its
relations with the other dimensions of time. Therefore, one of my tasks is
to spell out clearly its properties so that Linear Time and its linguistic
representation may be better understood. As for Frame Time, my task is to
study linguistic representations of Frame Time and its theoretical
significance in the representation of time in natural languages. I will focUs
on linguistic representations of Situation Time in relation to verb
categorization and the characterization of the truth conditions for sentences
with different classes of verbs in different tenses and aspects, since
Situation Time is a dimension of temporal relations that are not fully
treated in natural languages. The problem of verb categorization is

eventually viewed in a temporal frame in this study.

Chapter Two
TIME, VERB CATEGORIZATION AND TRUTH CONDITIONS

2.0 INTRODUCTION

The purpose of this chapter is to review the problems of verb
categorization for the characterization of truth conditions of related
sentences in semantics, and in tense and aspect logic from the perspective of
time. This is the last of the three general problems that previous
approaches to time and verb categorization have, the first two of which
have already been reviewed in the preceding chapter.

What is the relation between verb categorization and truth conditions
for sentences or statements? In semantics, whether a sentence is true or
false is usually evaluated against time. For example, in model—theoretic
semantics, a sentence is evaluated against the pair of worlds and times
(Dowty 1979). In interval semantics, a sentence is specifically evaluated at
an interval of time, in an interval of time and for/throughout an interval of
time for its truth conditions (Bach 1980) if possible worlds are put aside.

A similar situation is found in logic. A statement is traditionally evaluated
at moments of time, but is now often evaluated in terms of intervals of time
(Benthem 1980). It is clear that time is a key condition for the truth of a
sentence or statement in semantics and logic. On the other hand, verb
categorization also involves the concept of time. For example, Aristotle's
classification of process verbs and activity verbs depends crucially on the

notion of time. If a process verb denotes an action that involves changing
22

 

fr rm (
01 tim.
m t} n
in case
on y at
im tort.
are clas
adtqua
qUt stio
acti )n t
trut l of
that actt
“at abl
Phrz ses ,
56m; llllt
80m. no
Bent Ett.
Bach 19>
Obse] Vaill
Verb . thrz.

 

 

23

from one state of affairs to another state of affairs, then a specific stretch
of time is needed for the changing to take place. However, a period of time
may not be required when no such changing of the state of affairs occurs as
in case of actions denoted by activity verbs. In the latter case there may be
only an arbitrary stretch of time involved. It is also apparent that time is an
important dimension along which verbs, or more specifically verb phrases,
are classified. If the truth of a sentence is evaluated in terms of time, an
adequate characterization of the temporal structure of the verb phrases in
question is necessary so that the relation between the time involved in the
action denoted by a verb phrase and the time used as a measurement for the
truth of the sentence can be fully considered. It seems that adequate
characterization of the temporal structure of verb phrases is not yet
available. Even available characterizations of temporal structures of verb
phrases do not appear to be fully considered in the treatment of the
semantics of sentences with the verb phrases in question. For example,
some notion similar to that of Frame Time is raised in early work by
Bennett and Partee (1978), but abandoned by themselves and others (cf.
Bach 1981, Smith 1991). Another example is Vendler's (1957, 1967)
observation of linguistic representation of distinctive temporal structures in
verb phrases which has also been ignored in semantics and tense and aspect
logic for a long time. This phenomenon of the ignorance of temporal
relations in natural languages is seen in previous formal approaches to
tense and aspect in natural languages. Of course, no consideration is ever
given in semantics and logic to all the three dimensions of the
representation of time, e.g., Linear Time, Frame Time and Situation Time
that I have outlined in section 1.3, though much attention is paid to one of

the three dimensions -- Linear Time in tense logic.

truth c
rapidl}
attenti.
With er
reviett
differer
work t I
the relal
categor'
rCpresei

I
author t.

 

models.
naIUra] l
analyzcg
Ihat has
Work On
BCnnen t
Montagu

(1977

I. It
that Don

 

in [his Sq!
applt‘mchL

ClChenbL-I

24

2.1 PREVIOUS APPROACHES TO THE PROBLEM

The literature on verb categorization and on the characterization of
truth conditions for sentences with different classes of verbs has grown
rapidly in the past two decades, though the topic did not receive enough
attention long after Vendler's (1957, 1967) pioneering work on the topic
with emphasis from a linguistic perspective. This section does not present a
review of all the literature but only to review major work on the topic in
different approaches. The first author to be reviewed is Vendler, whose
work (1957, 1967) not only draws linguists and philosophers' attention to
the relation between verb categorization and the notion of time in verb
categorization but also makes two important observations of linguistic
representations of temporal relation in natural languages. The second
author covered in this section is Montague (1970, 1973). Montague takes a
model-theoretic approach to the semantics of natural languages treating
natural languages like formal languages. In fragments of English, he
analyzes the present tense, future tense and the progressive in an approach
that has an impact on research on this topic, though he does not directly
work on verb categorization. Work on this topic was further developed by
Bennett (1977, 1981), and Bennett and Partee (1978) along the line that
Montague had initiated. Part of this section will focus on Dowty's work
(1977, 1979). Of all the recent authors on verb categorization, I believe
that Dowty is the most influential author who examines the problem from a
linguist's perspective and whose work draws more attention on this topic
from linguists than anyone else's work. One of the last two authors covered
in this section is Aqvist (1977), who, together with Giinthner (1978),
approaches the problem of verb categorization in tense logic based on
Reichenbach's (1947) approach. The other author is Parsons (1985, 1989,

 

 

1991‘,
actior

classir

2.1.1

seems

pay at!
97) not
form 01
conccp.
future,

pttSUpp
into iOu

8CCOmp

 

dlCltOlOr
ngress'
I'erbs. F.
m3} b6:

not be a,

ltlhg Ill l

25

1991) who looks at the problem from the philosophical perspective of
action (Davidson 1968, 1980) and proposes an analysis in the form of

classical tense logic.

2.1.1 VENDLER'S ANALYSIS AND HIS TWO OBSERVATIONS

Of all the approaches to verb categorization by modern authors, it
seems to me that Vendler (1957, 1967) is the first author who appears to
pay attention to some of the set of questions in section 1.1. Vendler (1967:
97) notes that verbs involve the concept of time in the commonly known
form of verb tenses. He goes on pointing out that consideration of the
concept of time in verbs is not limited to verb tenses like past, present and
future, since the particular way in which verbs are used also involves or
presupposes the notion of time. With this insight, Vendler classifies verbs,
into four categories, activity verbs, state verbs, achievement verbs and
accomplishment verbs, breaking away from the Aristotlian tradition of
dichotomy/trichotomy in verb categorization. Vendler first of all uses
progressive sentences as a criterion to distinguish some verbs from other
verbs. For example, a question in progressive form like the one in (la)
may be answered with some verbs in progressive form as in (lb), but may

not be answered with some other verbs in the progressive as in (1c).

(1) a. What is John doing?
b. John is running/writing a book.
c. ?John is knowing a friend/recognizing a friend.

Vendler's explanation for the the difference between (1b) and (1c) is that
verbs in (lb) denote processes that are going on in time, while verbs in (1c)

are not processes that are going on phase by phase in time (1967: 99-100).

 

Tum}

mzmi

Thtdt

 

rmnm‘
(ll ,) it
nfllnt
act. \‘it_\ I
[ha \fc
f0Utah
mmgh
\eflin
Pefltcrp
acgnm
”let! 0dr:
kmefr
(e.g. rt.“
Ume“hi

IOng< 1‘ pg

 

l
l
me I; ”Cr:
l
l

26

Vendler makes a further distinction between the kind of verbs in (1b) in

meaning relative to the notion of time as in (2).

(2) a. John is running.
b. John is writing a book.

The distinction, according to Vendler, is that for (2a) John may stop
running the next moment but it is still true that John did run, whereas for
(2b) it may not be true that John did write a book if John stops writing the
next moment. Accordingly, Vendler labels the former kind of verbs as
activity verbs and the latter kind of verbs as accomplishment verbs. We see
that Vendler uses the entailment relation between sentences in progressive
form and sentences in simple past tense in the classification of verbs,
though he does not make it explicit. It is apparent that if a sentence with a
verb in progressive form implies a sentence with the same verb in
perfective form then the verb in question is an activity verb. This becomes
a common practice in identifying verb classes in the literature, though it is
methodologically problematic, as I will argue in section 4.4. As for the
kind of verbs in (lo), Vendler (1967:102) realizes that some of the verbs
(e.g. recognize) may be predicated of a subject only for a single moment of
time while others (e.g. know) may be predicated of a subject for shorter or
longer periods of time. Thus, he calls the former achievement verbs and
the latter state verbs.

It seems to me that the most significant aspect of Vendler's
approach lies in his full use of linguistic evidence in his analysis of the verb
classes. In classifying verbs in time schemata, Vendler (1967) has made two
significant observations, which unfortunately he does not take full

advantage of to develop a more comprehensive theory of verb classifica-

 

tion
trith
vertr
jinrlz

adItfl

(Siren 3
'EF”} to
HONCeg.
"’tth (4h

SCHIEnc-c

 

(1rd“: Tl‘
aclll'llV \

Character

 

 

27

tion. One of the observations is that verbs of different classes appear to go
with different adverbial phrases of time in sentences. For example, activity
verbs may occur with temporal adverbial phrases such as for how long and
for half an hour, whereas accomplishment verbs may occur with temporal

adverbial phrases like how long and in half an hour, as shown below.

(3) a. For how long did he push the cart?
b. ?For how long did he draw the circle?
(3') a. ?How long did it take him to push the cart?
b. How long did it take him to draw a circle?
(4) a. He pushed it for half an hour.
b. ?He drew it for half an hour.
(4') a. ? He pushed it in half an hour.
b. He drew it in half an hour.
(4") a. ?It took him half an hour to push it.
b. It took him half an hour to draw it.

Given a verb like push, it sounds odd to ask a question like (3'a) and odd to
reply to (3'3) with answers like (4'a) and (4"a), as Vendler (1967:100-1)
notices. Given a verb like draw, it is equally odd to ask (3b) and to answer
with (4b). Vendler (1967:100) explains that the difference between the (a)
sentences and the (b) sentences lies in the difference between push and
draw. The verb push does not have a set terminal, which is typical of
activity verbs, whereas the verb draw has a set terminal or climax, which is
characteristic of accomplishment verbs. Vendler also finds that state verbs
appear to behave like activity verbs with regard to the adverbial phrases of
time in (5) and (5') ,but he fails to notice that they behave differently with

the adverbial phrases of time in (5").

(5) a. For how long did you push the cart?
b. For how long did you love her?

ln adc
and at
phrase
behave
the £1LI\

such as

t t

 

Verbs m
Itirh reg
achle\er|

Pitt/1) in

(8

Al the Sat

 

 

28

(5') a. ?How long did it take you to push the cart?
b. ?How long did it take you to love her?
(5") a. At what time did you push the cart?
b. ?At what time did you love her?

In addition, Vendler notices that achievement verbs, accomplishment verbs
and activity verbs behave differently with regard to a number of adverbial
phrases of time. For example, achievement verbs like reach in (6a) seem to
behave differently from accomplishment verbs like draw in a sentence with
the adverbial phrase at what time, and also differently from activity verbs,

such as push , with the adverbial phrase how long, as in (7b).

(6) a. At what time did you reach the top?
b. ?At what time did you draw the circle?

(7) a. How long did it take you to reach the top?
b. ?How long did it take you to push the cart?

However, it appears that Vendler did not realize that achievement
verbs may behave similarly to accomplishment verbs and activity verbs
with regard to the choice of some other time adverbials. For example, an
achievement verb like reach may behave similarly to an activity verb (e.g.

push) in sentences with adverbial phrases like at what time, as in (8).

(8) a. At what time did you reach the top?
b. At what time did you push the cart?

At the same time, an achievement verb (e. g. reach) and an accomplishment
verb (e.g. draw) appear to be found to behave similarly in (9a & 9b) with
the adverbial phrase for how long, in contrast to the behavior of an activity
verb like push in (9c).

ln spit
appear
PhllOSt
the qUt
\thethe
time. a
time in
ftprese
Concerr

I
With an
acting

(l

(l

 

 

 

29

(9) a. ?For how long did you reach the top?
b. ?For how long did you draw the circle?
c. For how long did you push the cart?

In spite of Vendler's failure to recognize the similarities, his observations
appear to suggest some linguistic evidence pertinent to answers to two
philosophical questions and one linguistic question raised in Chapter One:
the question whether actions take time in (1a) in section 1.1; the question
whether some actions take a period of time, some actions take a moment of
time, and some actions may take either a period of time or a moment of
time in (1b) in section 1.1; and the question whether there are linguistic
representations of temporal relations in natural languages, which is of
concern in the above two questions .

The other significant observation of Vendler's is that verb phrases
with an object noun phrase may completely change the behavior of an

activity verb in relation to the notion of time, as is contrasted in (10) and
(l 1) below.

(10) a. ? How long did it take John to run?
b. ?It took John half an hour to run.

(11) a. How long did it take John to run a mile?
b. It took John half an hour to run a mile.

The same phenomenon is found in verb phrases with prepositional

complements, as is contrasted in (12) and (13) below.

(12) a. ?How long did it take John to walk?
b. ?It took John five minutes to walk.

(13) a. How long did it take John to walk across the street?
b. It took John five minutes to walk across the street.

Vendler's explanation for the difference between (10) and (11) and between

 

(12)
or cl.
“e ll
accor
time.
to tim
obsen
1979. .
that is
phrases
switch
1.1: the
with \\ i
rEprese
In natur;
“(Toss l.
end-poii
S€ntcnct
ln
style am
reEtched
has alSo
different
“nhlefl
noun Phi
tarpon“
CalegOrl-C

30

(12) and (13) is an Aristotelian one. In (1 1) and (13), there is a set terminal
or climax as opposed to (10) and (12), respectively. Given this explanation,
we find that the verbs in (l 1) and (13) have the same explanation as
accomplishment verbs have for their behavior in relation to the notion of
time, and in fact they behave exactly like accomplishment verbs in relation
to time and in terms of some entailment relations. This phenomenon
observed by Vendler is later known as verb category switch (cf. Dowty
1979, Zhou 1991). Verb category switch is the kind of linguistic evidence
that is of direct relevance to questions of temporal relations in verbs
phrases. For example, the linguistic evidence involved in verb category
switch seems to be relevant to answers to two questions raised in section
1.1: the question whether the period of time may be open-ended or may be
with well-defined end-points, and the question whether there are linguistic
representations of open-ended period of time and of well-defined end—points
in natural languages. In sentences with a mile, as in (l l), and sentences with
across the street, as in (13), the period of time appears to have well-defined
end-points. On the other hand, the period of time seems to be open-ended in
sentences in (10) and sentences in (12) without those phrases.

In short, Vendler focuses on linguistic evidence in his Aristotlian
style approach to verb categorization. In his analysis, he not only has
reached four categories of verbs, which is well-known in the literature, but
has also made two important observations: different temporal structures in
different categories of verbs, as seen in the behavior of temporal adverbials
with verb classes, and verb category switch, as witnessed in adding certain
noun phrases to the verbs phrases in question. It seems to me that his two
important observations do not draw as much as attention as his four

categories of verbs in the study of verbs.

31
2.1.2 PUZZLES OF MONTAGUE'S TENSE/ASPECT OPERATORS

Montague approaches natural languages from the the perspective of
mathematics and logic, generalizes metamathematics to comprehend natural
languages, and treats natural languages, such as English, as formal
languages (T homason 1974). His approach to the semantics of natural
languages may be characterized as truth conditional, model theoretic,
possible worlds (Dowty, Wall and Peters 1981) as well as compositional
(Abbott 1992). Montague has not worked directly on the topic of verb
categorization, but his approach to tense logic and its application to
fragments of English has an impact on the research on this topic of verb
categorization in the same way as it has on the study of temporal relations
in natural languages in general.

Montague (1970,1973) has worked on the simple present tense, the
future tenses and the present perfect tense/aspect for fragments of English.1
Let us see how his analysis works. Montague considers the truth-value of a
sentence in relation to a point of reference, which is an ordered pair of a
possible world (w) and a moment of time (i).2 The set of moments of time
(J) is linearly ordered. Let 4) be Bill walks. Let a set of operators be W, H,
and N, where W is read as 'it will be the case that', H as 'it has been the
case that', and N as 'it is being the case that' (Montague 1970:125,
1973:257). Then, the truth condition for H4) or 4) in the present perfect
tense may be stated informally as (14) .

 

1 Montague does not seem to distinguish tenses and aspects here. This may be one of
the reasons that his analysis fails in this respect. In this study, I will use 'tense' just for
tense, and use 'tense/aspect' for aspect where such distinction is not made in the original
work.

2 For current purposes, I just consider the coordinate of a moment of time, and drop
that of the possible world in this chapter.

32

(14) It has been the case that Bill walks is true at i if and
only if there exists j such that j < i and Bill walks is true at j.
(14') Bill has walked.

(14) is supposed to capture the truth conditions for (14') in idiomatic
English. The truth condition for W¢ or it) in the future tense may be

informally represented in (15) in a similar fashion.

(15) It will be the case that Bill walks is true at i if and only
if there exists j such that i < j and Bill walks is true at j.
(15') Bill will walk.

(15) should be able to provide a sufficient truth condition for the semantics
of (15') in idiomatic English. As Bennett (1977:492) notes, progressive
tenses/aspects are conspicuously absent in Montague's treatment of tenses/
aspects in the fragment of English in The Proper Treatment of
Quantificatiop in glrdinag English (PTQ)(Montague 1973), though
Montague (1970:125) does discuss the issue. The truth condition for N4) or

(it in the present progressive tense/aspect may be informally stated in (16).

(16) It is being the case that Bill walks is true at i if and only
if there exist J such that i C J and Bill walks is true
throughout I, where J is a set of moments of time.

(16') Bill is walking.

(16) is expected to capture the truth conditions for the idiomatic English
sentence in (16'). In addition to the few tenses/aspects involved in the
fragment of English, Montague's analysis is supposed to be applicable, in
principle, to other tense/aspect operators in natural languages. It is clear
that consideration of verb categorization or consideration of the temporal

structure of verb phrases is not given in Montague's treatment of tenses/

33

aspects. I believe that Montague's analysis just involves one of the three
dimensions of time, Linear Time only.

In Montague's approach to tenses/aspects, there are two main puzzles
remaining to be solved, as pointed out by Bennett and Partee (1978) and
Bennett (1977). Let us examine the first puzzle. In his approach, Montague
uses a sentence in simple present tense, like Bill walks, as the base formula
underlying analyses of sentences in all other tenses and aspects. Generally
speaking, a sentence in simple present tense has two readings: a timeless/
habitual reading and a reportive reading (Jespersen 1965:258-60, Bennett
1977:493—4). The former reading involves a large stretch of Frame Time
including the 'past, 'present' and 'future', while the latter reading says
something just about Bill at present. 'Present' is considered durationless,
what goes before 'present' is the past and what is after 'present' is the
future, since time is considered as an ever flowing river so that no one can
bathe in the same river twice, as the ancient Greeks claim. According to
Bennett and Partee (1978), and Bennett (1977), it is apparent that
Montague uses simple present tense in the sense of the reportive reading in
his fragments of English. It is noticed that not every sentence with every

class of verbs in simple present tense may be used in the reportive sense, as
shown in (17) and (18).

(17) a. (Look.) Bill walks.
b. (Look.) Bill runs.

( 18) a. (Look.) ?Bill walks to the bookstore.
b. (Look.) ?Bill writes a book.

Sentences in (17) with the activity verb phrases are fine with the reportive

reading as used in Montague's analysis. However, it is questionable that

sente.
repon
occur
Gal to
expre:
in the
semen
The lit
also sc
accom;
is unac

read as

 

 

34

sentences in (18) with the accomplishment verb phrases may be used in the
reportive sense. These sentences can not be used to report a particular
occurrence of Bill's walking to the bookstore or Bill's writing of a book, as
Galton (1984: 1 1-2) claims. Such an occurrence may be either in the past, as
expressed by sentences like Bill wrote a book or Bill has written a book, or
in the future, as in the form Bill will write a book. When we have the
sentence Bill writes a book, its habitual reading is predominantly available.
The limitation of the simple present tense is not only seen in English but
also seen in Chinese and other languages. For example, a sentence with an
accomplishment verb phrase without any adverbial of time (19a) in Chinese
is unacceptable for a reportive reading, as in (19b), and is forced to be

read as a sentence in the present progressive aspect, as in (19c).3

(19) a. Kan, Lisi kan yi ben shu.
Look Lisi read a M(easure) W(ord) book.
b. Look, ?Lisi reads a book.
c. Look, Lisi is reading a book.

 

3 I have presented (19a) to a number of native speakers of Chinese, who claim that
they interpret it as the present progress, as in (i), when they hear it.

(i) Kan, Lisi zai kan yi ben shu.
Look, Lisi Asp. read a MW book
Look, Lisi is reading a book.

(ii) Zuotian Lisi kan yiben shu.
Yesterday Lisi read a book
Yesterday Lisi was reading a book.

(iii) Zuotian Lisi kan le yiben shu.
Yesterday Lisi read Asp. a book
Yesterday Lisi read a book.

In addition, it is claimed that in Chinese 3 sentence's tense is usually determined by
temporal adverbials or contexts, and that the marked le functions as an aspect mark rather
than a tense mark, as in (ii) and (iii). However, I will argue in Chapter Three that le and
guo function more like tense markers than aspect markers in section 3.3.

35

(19') a. Lisi tiantian kan yi ben shu.
Lisi every day read a MW book

b. Lisi reads a book every day.

On the other hand, the corresponding sentence with a time adverbial, as in
(19'a), is acceptable for a habitual reading (19'b). Then, what is the
difference between (17) and (18)? I think that the difference lies in the
different temporal structures that the verb phrases in (17) and (18) have.
The actions denoted by the verb phrases in (17) may just take an instant of
time that is compatible with the durationless 'present', whereas the actions
denoted by the verb phrases in (18) take a period of time which is too long
for the durationless 'present'. With this understanding, we see that it is
difficult for sentences like (18) and (19a) to be assigned 'a reportive
reading', since 'present' is a instant of time of very short duration.

The second puzzle also involves accomplishment verbs, as noted by a
number of authors (Bennett and Partee 1972, Bennett 1977, Parsons 1989),
and leads to what Dowty (1979:133) calls the 'imperfective paradox'. In
Montague's analysis, a sentence in present progressive tense/aspect is true at
a moment i, if and only if there exists a neighborhood about J such that the
corresponding sentence in present simple tense is true throughout J. In this
analysis, the sentence in present progressive tense/aspect implies that the
corresponding sentence in simple present tense is true too. This is where

the paradox arises in Montague's analysis.

(18') a. Bill is walking to the bookstore.
b. Bill is writing a book.

Given a scenario in which Bill is writing a book or is walking to a
bookstore, but he is to be killed in an accident before he finishes the book

36

or reaches the bookstore, we find that (18') is true at i without (18) being
true throughout J. This is the imperfective paradox.

In summary, Montague's rule of tenses/aspects is inadequate when it
is applied to the analysis of tenses/aspects in natural languages such as those
in English. There are two problems with his analysis. The first is the use of
sentences in simple present tense as the base formula from which
corresponding sentences in other tenses/aspects are derived by simply
adding tense/aspect operators. Sentences with activity verb phrases may be
used in simple present tense in the reportive sense without problems, while
sentences with accomplishment verb phrases and achievement verb phrases
may not be used in simple present tense in the reportive sense, according to
Bennett (1977) and Bennett and Partee (19978). The second problem is the
imperfective paradox in which a sentence in the present progressive
tense/aspect implies a corresponding sentence in simple present tense,
which is a wrong prediction for sentences with accomplishment verb
phrases and achievement verb phases. I believe that these two problems

result from the negligence of the two sets of questions raised in section 1.1.

2.1.3 BENNETT AND PARTEE'S INTERVAL SEMANTICS

Bennett and Partee (1978) are among the first authors to point out
problems in Montague's analysis of tenses/aspects in fragments of English.
In addition to the two puzzles discussed in the above section, Bennett and
Partee also find that Montague's treatment of the simple future tense in
English is inadequate and his analysis of the present perfect tense/aspect in
English incorrect. According to Bennett and Partee (1978:7-8), the
inadequacy of Montague's analysis of the simple future tense is shown in

the treatment of a pair of sentences like (20) on the next page.

Oil the
lmCfiti
zfifleto

Bennet

tense/a.

 

37

(20) a. Bill will walk to the bookstore.
b. Bill will walk to the bookstore some time.

On Montague's analysis, (20a) and (20b) are logically equivalent. Both
(20a) and (20b) are true sentences at i if and only if there exists j such that
i precedes j, and Bill walks to the bookstore is true at j. In any context,
linguistic or extralinguistic, (20a) is taken to assert that the event is to

obtain in a definite time in the future, as explicitly expressed below.

(20) a'. Bill will walk to the bookstore today.
a". Bill will walk to the bookstore next Saturday.

On the other hand, (20b) asserts that the event is going to obtain in an
indefinite future time. Montague's analysis of the simple future tense is not
able to capture the difference. The next problem seems to be more serious.
Bennett and Partee believe that Montague's analysis of the present perfect

tense/aspect makes wrong predictions.

(21) a. Bill has walked to the bookstore.
b. Bill walked to the bookstore.

In Montague's analysis, (21a) and (21b) are logically equivalent. They are
true sentences at i if and only if there exists j such that j precedes i, and Bill
walks to the bookstore is true at j. This is simply incorrect, as clearly
shown in the idiomatic English sentences in (21a') and (213"). (21a')
includes a temporal adverbial referring to a period of time that includes the

event and the utterance of the sentence, while (21a") does not.

(21) a'. Bill has walked to the bookstore today.
a". *Bill has walked to the bookstore yesterday.

38

The problem lies in the temporal structures of i and j for the present
perfect tense/aspect and for the simple past tense. i and j are points of time
and are linearly ordered that, say, j > i, where i is the moment of
utterance. I think that the interval of time for the present perfect
tense/aspect includes both i and j, but the interval of time for the simple
past tense does not include i, though (21b) is evaluated at j. It is apparent
that Montague did not consider this temporal relation in his analysis of
tense/aspect.4

Bennett and Partee (1978) propose to rescue Montague's analysis by
using interval semantics. An interval of time is a set of moments of time.
For interval semantics, a point of reference is an ordered pair of a possible
world and an interval of time, instead of Montague's ordered pair of a
possible world and a moment of time. This proposal appears to solve
Montague's two puzzles together with other problems. The problem of the
first puzzle is that a base sentence in the reportive simple present tense with
accomplishment and achievement verb phrases can not be evaluated in
terms of a moment of time as to its truth conditions. The true base sentence
is used in the description of the truth conditions for sentences in other
tenses/aspects. This puzzle is readily accounted for in interval semantics, as
in informally stated in (22) (Bennett and Partee 1978:38-9, Bennett 1977:
499-500).

(22) Bill has written a book a is true at interval of time I if and
only if I is a moment of time, (1 refers to an interval of time
I', I is a member of I', and there exists a subinterval of time I',

I " such that I is a final point for 1" or 1" < I, and Bill writes
a book is true at I ".

 

4The present perfect puzzle is discussed in detail in section 5.2.

Theh
book

Illllll

 

|
|
dauLti
Permh
oftthj
conditj

l

uuorn

 

39

Their intuition underlying this interval analysis is that Bill starts to write a
book at the initial of 1", he is writing the book through I " and he finishes
writing the book by the end of I ". This first puzzle is thus solved. In (22), a
clause is added to the truth conditions for the sentence in the present
perfect tense/aspect, that is, there is a super-interval I', as represented by a,
of which both I and I" are subintervals.5 Let us see how the truth
conditions for the corresponding sentence in the simple past tense are

informally stated in (23).

(23) Bill wrote a book a is true at interval of time I if and

only if I is a moment of time, (1 refers to an interval of time
I ' and there exists a subinterval of time I ', 1" such that I " < I,
and Bill writes a book is true at I ".

It is clear in (23) that I is not a subinterval of I ', thought I " is a subinterval
of I '. In other words, the difference between the present perfect and the
simple past in English lies in that I and 1" are subintervals of 1' the super—
interval for the present perfect, as in (22), while I is not a subinterval of I '
the super-interval, though I " is a subinterval of I ', as in (23). Given this
distinction, Bennett and Partee's analysis claims that a sentence in the
present perfect tense/aspect and its corresponding sentence in the simple
past tense are not logically equivalent. Bennett and Partee propose to
account for Montague's second puzzle, the imperfective paradox, in similar
fashion. Let us look at the same set of sentences (18) and (18') used in the

discussion of Montague's analysis in subsection 2.1.2.

 

5 Bennett and Partee (1978: 26-7) seem to take a for Frame Time in their analysis,
but they do no make it explicit in their work. I think that (I. certainly has its linguistic
motivation if it is taken to refer to Frame Time. Unfortunately, Bennett and Partee do not
elaborate this idea in their later work on this topic. This question will be fully treated in
Chapter Three.

40

(18) a. Bill walks to the bookstore.
b. Bill writes a book.

(18') a. Bill is walking to the bookstore.
b. Bill is writing a book.

In interval analysis (Bennett and Partee 1978:13), the base sentence in the
simple present tense (18) that underlie sentences in all other tenses/aspects
are supposed to refer to events that occur within an interval of time. An
event starts at the initial point of the interval, if it has an initial point, and
stops at the final point of the interval, if it has a final point. For example,
the truth conditions for (18'b) may be informally represented with (18b) as

the underlying sentence, as in (24).

(24) Bill is writing a book is true at I if and only if I is a
moment of time, there exists an interval of time I ' such that I
is in I', I is not an endpoint for I', and Bill writes a book is
true at I '.

In (24), the key clause that is different from Montague's analysis is about
the super-interval I ' and the relation between the super-interval I ' and
subinterval I, the latter of which is not the final point of the super-interval.
Bennett and Partee's (1978) analysis seems to have solved the paradox.
Moreover, the interval analysis of the progressive aspect appears to
provide some new criteria for the categorization of verbs (Bennett and
Partee 1978). In relation to the progressive aspect and truth conditions for
sentences in the progressive aspect, verb phrases may be classified into
three categories: + stative, + subinterval, and - stative & - subinterval.
Simply put, + stative verb phrases may not be used in the progressive

aspect, as shown in (25).

+ Sub
main
inten':
super-

membt

~Statix
aspect.
”3 ~ St;
the prof:
at the su

6“mil (2

Ho
imperfect

).

 

Partee's ar

(18,
(18-

 

41

(25) a. ? Bill is believing that John is wrong.
b. ? John is being happy.

+ Subinterval verbs may be used in the progressive aspect. If used as the
main verb of a sentence in the progressive aspect that is true at a super-
interval, then the sentence is true at each and every subinterval of the
super-interval, as in (26), so that (26a) entails (26b) and (26c), I ' and I " are

members of I.

(26) a. Bill is walking at I.
b. Bill walks at I'.
c. Bill walks at I ".

-Stative - subinterval verb phrases may also be used in the progressive
aspect, but they have logical behavior different from + subinterval verbs.
If a - stative & — subinterval verb is used as the main verb in a sentence in
the progressive aspect that is true at a subinterval I', the sentence is not true
at the subinterval's the super-interval I, as in (27), where (27a) does not
entail (27b).

(27) a. Bill is building a house at I.
b. Bill builds a house.

However, Bennett and Partee's (1978) analysis does not escape the
imperfective paradox, as Dowty (1979:145-146) and Parsons (1989:214—
215) note, among a number of authors. In particular, on Bennett and

Partee's analysis as stated in (24), (18') entails (18").

(18) a. Bill walks to the bookstore.
b. Bill writes a book.

(18') a. Bill is walking to the bookstore.
b. Bill is writing a book.

 

Spec
interr
will 1
will t
guara.
it no“

inaht

pl’0p03

in Eng

 

 

 

42

(18") a. Bill has walked to the bookstore.
b. Bill has written a book.

Specifically, if the sentences in (18') are true at 10 o'clock today, there is an
interval around 10 o'clock at which the sentences in (18) are true. There
will be some moment later than that interval. At that later moment, (18")
will be true. This is of course incorrect. We know that there is no
guarantee, for example, that Bill will reach the bookstore if he is walking to
it now. Bill may have an accident before he gets there, and he may end up
in a hospital.

To remedy Bennett and Partee's (1978) analysis, Bennett (1981:15)
proposes (24') as the truth condition for sentences in the progressive aspect

in English:

(24') Bill is writing a book is true at interval of time I if and
only if I is a moment of time, and there exists an interval of
time I ' such that l' is an open interval, I is included in I ',
and Bill is in the extension of writing a book at I'.

In this analysis, Bennett makes a distinction between an OPEN interval and a
CLOSED interval; the former has no end-points whereas the latter has end-
points. Without end-points, an OPEN interval describes a process which is
going on without ending. This is intended for sentences with activity verb
phrases, and also for sentences with accomplishment and achievement verb
phrases in the present progressive aspect. With end-points, a CLOSED
interval describes an event since end-points indicate the inception and
completion of the action. Therefore, the fact that (18'b) is true does not
imply that (18"b) is true. As Parsons (1989:215) comments, an OPEN
interval does not culminate so that Bennett's analysis eventually avoids the

imperfective problem. However, Bennett's analysis does not answer the

43

question of how one can tell if an interval is OPEN or CLOSED when facing
it (Parsons 1989:231-2). Bennett himself admits that almost everyone finds
it mysterious initially (1981:13). Does the distinction between OPEN
interval and CLOSED interval have any linguistic motivation? Bennett does
not try to answer this question. The question may not be fully answered
without appropriate consideration of the questions raised in section 1.1.

In short, Bennett and Partee's analysis of tense/aspect has solved, in
terms of interval semantics, the puzzle of the reportive use of the simple
present tense sentence with an accomplishment verb as the base sentence in
Montague's approach, and also remedied several other problems, such as
inaccurate analysis of the simple past tense and the present perfect tense/
aspect, in Montague's approach. However, the puzzle of the imperfective
paradox is not solved until Bennett comes up with the distinction between
OPEN interval and CLOSED interval in his later work. In addition, I believe
that Bennett and Partee put forward a significant notion of a dimension of
time a or frame adverbial, but fail to explore its linguistic and logical

significance in the study of the representation of time in natural languages.

2.1.4 DOWTY'S LOGICAL OPERATORS AND INERTIA WORLDS
The two important observations by Vendler (1957, 1967) virtually
had gone unnoticed in linguistic research until Dowty (1972, 1977, 1979)
treated these problems first in generative semantics and later in model
theoretical semantics. In Dowty's (1979: 62) opinion, Vendler's attempt to
classify verbs, once and for all, is misguided because his approach can only
takes surface verbs into account, though he notices that verb phrases may
be relevant. This is because problems arise when verbs take objects. The

same verb may behave like an activity verb when it takes a zero object or

44

zero complement, or takes a mass noun or indefinite plural noun as object
or complement as in (28a) and (28b), and behave like an accomplishment
verb when it takes those nouns with appropriate deterrniners or numerals
as in (28'a & 28'b) and when appropriate noun phrases ares added as
object or complement. As a matter of fact, Vendler (1967: 102) has noticed
the difference between the same verb phrases with and without object
nouns, but he does not provide any account for this observation except

different labels for them.

(28) a. John ate popcorn for/*in an hour.
b. John built houses for/*in a month.

(28') a. John ate a bag of popcorn in/*for an hour.
b. John built three houses in/*for a month.

As observed by Vendler (1957, 1967), an in -time-phrase can only modify
an accomplishment verb, whereas a for-time-phrase only modifies an
activity verb, as shown in (28) and (28'). Dowty's first solution to the
problems in (28) and (28') is along the line of generative semantics: to
postulate a single homogeneous class of predicates -- state predicates plus
sentential operators, DO and BECOME, and connectives so that (28) and
(28') have different underlying structures. DO is the highest operator
underlying activity verbs and allows for-time-phrases directly under it, but
does not allow in-time-phrases directly under it. On the other hand,
BECOME may be the highest operator underlying both accomplishment
and achievement verbs and permits in-time-phrases directly under it, but
does not allow for-time-phrase directly under it, though DO may also be
the highest operator in the logical structure of some of non-intensional

agentive accomplishment verbs. For example, the following underlying

45

structures may be postulated to account for the the difference between
(28a) and (28'a) respectively.

(29) [ DO [ John ate [ CAUSE [ BECOME [ pop corn is eaten DD
for an hour]

(30) [ DO [ John ate [ CAUSE [ BECOME [ a bag of pop corn is
empty ] in an hour ]]]]

The underlying structure in (29) indicates that the agent did the eating for
an hour, and for an hour is directly under the operator DO. On the other
hand, the underlying structure in (30) shows that in (28) the agent brought
the state that a bag of popcorn became empty in an hour, and in an hour is
directly under the operator BECOME. This seems to account for the
observation by Vendler (1957, 1967) that different verbs or verb phrases
go with different temporal adverbials. With these logical operators, Dowty
is able to give logical structures, in terms of the aspect calculus, to
Vendler's four categories of verbs, which Dowty further divides into
eleven groups, according to their logical representations in the aspect
calculus (1979:123-5).

In his first attempt to account for the semantics of the progressive
sentences, Dowty (1972) gives tenseless sentences with accomplishment
verb phrases the analysis in (31) and progressive sentences with same verb

phrases the corresponding analysis in (31').

(31) a. [tb CAUSE [BECOME 1p]]
b. [[Bill write] CAUSE [BECOME [there is a book]]]
(31') a. [PROG [(1) CAUSE [BECOME w]]]
b. [PROG [Bill write] CAUSE [BECOME [there is a book]]]

In this analysis, (I) and 1p are sentences. In (31), a sentence [BECOME LP] is

46

true at a moment of time t if and only the sentence '1’ is true at t and false at
a moment of time t' before t. Dowty suggests that the difference between
(31) and (31') is that in (31') PROG it: can be inferred but [BECOME 1p]
can not be, whereas in (31) both it) and [BECOME 112] can be inferred.
Realizing that this suggestion is rather arbitrary and the analysis in terms
of moments of time is problematic, Dowty (1977:57, 1979:146-7) provides
(32) as the truth conditions for progressive sentences with accomplishment

verbs and other verbs as well.

(32) [PROG b] is true at (I, w) if and only if there is an interval 1'
such that I C I' and I is not a final subinterval for I' and there
is a world w' for which it: is true at (I‘, w' ) , and w is exactly
like w' at all times preceding and including 1.

In (32), Dowty adopts Bennett and Partee's (1973) interval semantics, and
replaces the traditional moment of time, that is, a point of time, with
intervals of time in the ordered pair of times and worlds in model-theoretic
semantics. This does not seem to guarantee that the imperfective paradox
does not exist in this analysis. Therefore, Dowty (1979:148—9) proposes the
notion of inertia worlds —- a set of possible worlds, which are thought to be
exactly like the given world up to the time in question, and in which the
event's future course of development may mostly be compatible with the

past course. Given this function of inertia worlds, Inr in short, (32) is
restated as (32').

(32') [PROG 4)] is true at (I, w) if and only for some interval 1'
such that I C I' and I is not a final subinterval for I ', and for

all w' such that w' E Inr ((I, w)), (I) is true at (l', w' ).

According to (32'), PROG (I), such as (33b) and (34b), is not evaluated as

true
like
at It
thos
guar
SOUR

topit

point
3 nor
Struct
etany
SITUCI

lhOUg

[BEC‘
"'ht‘rt
(Don

are Co

 

Chis“. I:

Chen-3C

"hat is
Sense. .,

 

47

true in any possible worlds but only in inertia worlds in which (I) is most
likely to be true. This amounts to say that in all worlds like that actual one
at that time, John would eventually finish reading or writing the book, if in
those worlds nothing out of the ordinary happened. This of course does not
guarantee that (333) and (34a) are true when (33b) and(34b) are true. If
something unexpected happened, say, John found that he did not like the
topic, he might have abandoned reading or writing the book at the end.
There appear to be two major problems with Dowty's approach, as
pointed out by Bennett (1977), Declerck (1979) and Parsons (1989), among
a number of authors. The first problem is the way that underlying logical
structures are represented in the generative-semantics like approach. For
example, given Dowty's solution, it is not clear what distinctive underlying
structures can be postulated for (33) and (34) (Declerck 1979: 270-1),

though both read and write are accomplishment verbs.

(33) a. John read a book.

b. John was reading a book.
(34) a. John wrote a book.

b. John was writing a book.

[BECOME 1p] expresses the transition from an interval of time and a world
where It) is not true to an interval of time and world where w is true
(Dowty 1979:140). When accomplishment verbs in progressive sentences
are concerned, there are two kinds of verbs: one presupposes the pre-
existence of the object (33), and the other does not (34).6 Given Dowty's

characterization of the semantics, it is easy to postulate an underlying

 

6 It is commonly discussed in the literature that the second kind of verb gives rise to
what is called the 'imperfective paradox' (Dowty 1977, 1979, Parsons 1989). In a broader
sense, verb phrases like the one in (33b) also involve the 'imperfective paradox'.

48

logical structure for (34), while it is not clear what appropriate underlying
logical structures may be given (33). First, it seems that there is no
causative relation in (33). Second, there always exists a book in the non-
progressive sentence, as in (33a), and there always exists a book in the
progressive sentence, as in (33b), as well. Any analysis that allows the
inference of the existence of a book in (33a) but not in (33b) is simply
wrong, though it may be correct in the case of (34). It appears that
Dowty's analysis has difficulty to postulate uniform underlying structures
to account for the difference between sentences with an accomplishment
reading and sentences with an activity reading as well as the difference in
meaning between sentences with the same verb category reading.

The second problem involves the proposal of inertia worlds. The
proposal (32') seems to be both too strong and too weak. It is too strong
(Parsons 1989:215-6) in that it predicts that things actually develop in ways
most compatible with the past course of development. If this happens, then
the actual world becomes an inertia world for that interval of time. Then,
(32') requires that a progressive sentence is not true at that interval of time
unless its corresponding non-progressive sentence is true then or later. This
revives the 'imperfective paradox' for that interval of time (Parsons
1989:215). The proposal in (32') is also too weak (Bennett 1977:503-4,
Parsons 1989:237). It fails to account the fact that (35a) implies (35b) and
(350). Intuitively, we know that if (35a) is true at a moment of time t then

it is also true at a moment of time t' if t' is immediately following t.

(35) a. John is running.
b. John will be running.
c. John will run.

Sun

dit‘fe
it ith
appr
same

Sufie

IN.)
_‘
Ur

tnnh

apprc
( 196:.

 

Opera
Connl
emilltl
1963i

{Outhtl

 

“Pleat

eyetilt

 

49

(35') a. John is running to the bookstore.
b. John will be running to the bookstore.
c. John will run to the bookstore.

Similarly, (35'a) implies (35'b) in case of accomplishment reading, though
(35'a) does not imply (35'c).

In summary, Dowty's generative-semantics underlying structures for
different verb classes have difficulty in representing progressive sentences
with some verb phrases appropriately and have difficulty in representing
appropriately some sentences with the same verb category reading. At the
same time, truth conditions provided in model-theoretic semantics seem to

suffer from problems of either being too strong or too weak.

2.1.5 PARSONS' EVENT SEMANTICS

Parsons (1989, 1990) tackles the problem of verb categorization and
truth conditions for the progressive aspect in English with a combined
approach (1980:40-4). This combined approach adopts both Davidson's
(1968, 1980) event theory of action sentences and what Parsons calls the
operator approach. In Davidson's event theory, it is assumed that a sentence
contains a covert reference to an event. Events are thus introduced as
entities about which an indefinite number of things can be said (Davidson
1968:90—1). This point is illustrated in the logical representation of the

following example.

(36) a. John pushed the cart.
b. Push (j, c)
e. (3e) (Push (e ) & Of (c, e) & Agent 0, e))

Instead of being symbolized as (36b), (36a) has a logical form of (36c) in
event theory. (36c) states that there is an event of pushing, which is about

50

the cart and of which John is the agent.
In the operator approach, any modifiers of the verbs are treated as

part of complex predicates, which are formed by the verbs and their
modifiers, as in (37').

(37) a. Mary runs
b. R (m)

(37') a. Mary runs slowly
b. S (R) m

The simple verb phrase in (37a) is represented as a simple predicate in
(37b), while the verb phrase with an adverbial in (37'a) is represented as a
complex predicate in (37'b), in which the modifier is symbolized as an
operator. The representation of (37'a) in (37'b) avoids unwanted inferences
like (38c), as Parsons (1980) notes.

(38) a. Mary walks, and Mary runs slowly.
b. W m & S (R)m
c. Mary walks slowly.

(38c) is not available in the logical form in (38b), since the operator may
not be transferred from one predicate to another predicate. In combining
the two approaches, Parsons' approach gives (37'a) the logical form in

terms of event semantics in (37").

(37") a. Mary runs slowly.
b. (516) (R(e) & S(e) & Agent (m, e))

(37"b) may be read 'there is an event which is running which is slow and of
which Mary is the agent'.

Parsons argues that the difference between an activity verb phrase

 

 

 

51

and an accomplishment verb phrase can be represented in the same fashion,
that is, the difference can be represented by introducing an operator as he
does in the above sentences. Specifically, he treats to in to the bookstore as
an operator symbolized by T in the representation of the distinction

between an activity verb phrase and an accomplishment verb phrase in (39)
and (39').

(39) a. John walked.
b. (it) {t < now & (3e) [W(e) & Agent (j, e)]}
(39') a. John walked to the bookstore.
b. (it) {t < now & (3e) [W(e) &T(e, b) & Agent (j, e)]}

The difference in logical forms for (39b) and (39'b) is that (39'b) has the
clause T(e, b) which may be read 'the event is toward the bookstore', while
(39b) does not have such an operator. The T operator in (39'b) indicates
that there is a terminal so that (39'b) represents an event instead of an
activity. However, Parsons' approach has a problem in the representation
of the distinction between many activity verb phrases and accomplishment

verb phrases, where operators of Parsons' type may not be conveniently
introduced, as shown in (40) and (41) below.

(40) a. John read a book.

b. (Elt) {t < now & (3e) [R(e) & Obj (b, e) & Agent 0, e)]}
(41) a. John pushed a cart.

b. (3t) {t < now & (3e) [P(e) & Obj (c, e) & Agent (i, e)]}

No distinction in logical forms may be found in the representations of
(40a), a sentence with an accomplishment verb phrase, and of (413), a
sentence with an activity verb phrase.

Ontologically, Parsons (1985) believes that the problem of verb

syr
stat
det'
whi
cul r

repr

In an
aIK3Vc
PIOgn

a diff.

 

IIIUSlr

 

 

52

categorization should be examined in terms of eventualities. Eventualities
(Parson 1985:238-40) are classified into two categories: one that culminates
in time and one that holds in time. The former is represented as Cul (e, t),
which is read 'e is an event that culminates at t', while the latter is
symbolized as Hold (e, t), which may be read 'e is a state and e's agent is in
state e at t, or e is a process that is going on at t or e is an event that is in
development at t '. In Hold (e, t), t is the set of times at which e holds and
which is an open interval. In Cul (e, t), t is the moment of time at which e
culminates and which is the end of an interval. With this ontology, Parsons

represents the distinction between (40) and (41) in (40') and (41').

(40') a. John read a book.
b. (31) {t < now & (3e) [R(e) & Object (b, e) & Agent 0, e) &
Cul (e, t)]}
(41') a. John pushed a cart.
b. (3t) {t < now & (3e) [P(e) & Object (c, e) & Agent 0, e) &
Hold (e, t)]}

In an analysis of the progressive in English with the approach outlined
above, Parsons (1989, 1990) proposes that the difference between a
progressive sentence and a non-progressive sentence may be represented as
a difference between Hold (e, t) and Cu! (e, t) in logical forms, as
illustrated in (42) and (42').

(42) a. John was reading a book.
b. (it) {t < now & (3e) [Writing (e) & Subject (e, John) &
Object (e, a book) & Hold (e, t )]}
(42') a. John read a book.
b. (3t) {t < now & (3e) [Writing (e) & Subject (e, John) &
Object (e, a book) & Cul (e, t )]}

The difference between the progressive sentence in (42a) and the non-

pro
dill
rest
ace.
Pars
Thu
Pmé
actit

short

An a;
is tha
(43b;
Hi’a i
theort
btttt‘c
Verb p

 

t'erb t

Daria.

 

53

progressive sentence in (42'a) with accomplishment verb phrases is a
difference between holding and culminating in (42b) and (42'b)
respectively. Parsons (1989:235) claims that the difference between
accomplishment verbs and activity verbs (event verbs and process verbs in
Parsons' terms) lies in that the latter is a series or amalgam of events.
Thus, process verbs are treated as a special kind of event verbs. The
progressive sentence (43a) and the non-progressive sentence (43'a) with
activity verbs are represented by holding and culminating, respectively, as
shown in (43b) and (43'b).

(43) a. John was running.
1). (fit) {I < now & (3e) [Running (e) & Subject (e, John) &
Hold (e, t)]}
(43') a. John ran.
b. (it) {t < now & (3e) [Running (e) & Subject (e, John) &
Cul (e, t)]}

An apparent problem with Parsons' approach to the progressive in English
is that given the similarity between representations of (42a) and of (43a) in
(42b) and (43b), we can not tell from these logical forms why (43a) entails
(43'a), while (42a) does not entail (42'a). It seems to me that Parsons'
theory fails to make correct predictions about the semantic relation
between progressive and non-progressive sentences with accomplishment
verb phrases and progressive and non-progressive sentences with activity
verb phrases.

To briefly review this section, I notice that Parsons adopts both
Davidson's event semantics and the operator method to present the logical
structures of sentences with the assumption that a sentence contains a covert

reference to an event. In his approach, verbs/verb phrases are classified as

 

 

III

CVCI
whit
II hiti
repr
sentt
1112111

diffe

 

 

with

 

54

event verbs/verb phrases and non-event verbs/verb phrases, the former of
which are represented using the operator Cul (e, I), while the latter of
which are symbolized using Hold (e, t). The same approach is taken to
represent the difference between the progressive and the non-progressive
sentences with different categories of verb phrases in logical forms. As a
matter of fact, this approach seems to fail to capture the semantic
difference between progressive sentences and non-progressive sentences

with accomplishment and with activity verbs.

2.1.6 AQVIST'S TENSE LOGIC APPROACH

Aqvist and Giinthner (1978) develop an improved tense-logic to
provide a logical analysis of verb aspect and events. In the syntax of their
logic, the formal language contains a number of operators that specify what
takes place uninterruptedly throughout bound intervals of time, and what
takes place at designated or fixed moments of time. In the semantics of
their logic, models with multiple indexing are developed and applied to
characterize the interpretations of these operators. The truth or falsity of a
sentence is determined in a model or a family of models with at least three
indices, say, to, t, and t', which are members of a non-empty set of times or
moments. This system of indexing is in fact similar to Reichenbach's three-
reference-point account of tenses and aspect, except that it is enriched with
models (Aqvist, 1976).

Specifically, Aqvist (1977) proposes three imperfective operators to
account for accomplishment verb phrases and a simple operator to account
for activity verbs. The three imperfective operators together with A,
which is a sentence, yield the truth conditions for the sentences in (44). The

present perfective progressive operator for a sentence like (44a) is

Has
beet
ISCL
on c
the:

pres

beco

Leta
delin
it “a.
k)b¢
“incl
'nOnn
true [3
Aqtitl
Thus
defini

The EL

 

55

HasBecomeMoreAndMoreA, which is read 'it has up to this moment
become more and more the case that A'. The future progressive operator is
IsGoingToBecomeMoreAndMoreA, and is read 'it is from this moment
on going to become more and more the case that A', which accounts for
the sentence in (44b). The operator, IsBecomingMoreAndMoreA, is for the
present progressive, as in (44c). This operator reads 'it is in this moment

becoming more and more the case that A'.

(44) a. John has been drawing a circle.
b. John will be drawing a circle.
c. John is drawing a circle.

Let us look at the operator, IsBecomingMoreAndMoreA, in detail. By the
definition of IsBecomingMoreAndMoreA, the sentence A is interpreted as
it was at least once the case that A at t, it is the case that A at to and is going
to be the case that A at t'. This means that there is an open interval around
which A is evaluated, and in which t is the immediate past, to is this
moment and t' is the immediate future. By definition, the sentence A is
true throughout this open interval. To avoid the 'imperfective paradox,
Aqvist (1977) further enriches the improved tense logic with quantifiers.
Thus the present progressive operator for accomplishment verb phrases, by

definition, gives (44c) this logical form:

(44') c. 3x (i HasBeenDrawing x lnSuchAWayThatItHasBecome-
MoreAndMoreTheCaseThat x ISAC ircle & Draw(j, x) &
a Circle (x)) & j IsGoingToBeDrawing x lnSuchAWay-
ThatItIsGoingToBecomeAndMoreTheCaseThat x
IsACircle).

The existential quantifier binds all the occurrences of the variable x to

3m

Bet
leas
the -
i mrr
thro:
£h7?!

form

In cor

there

 

there j

VI'hich '

appIOEi
A‘ll'isr

dtanb,

 

a"ills:

56

guarantee that there is the one and the same object denoted by x. The agent
has been drawing that object in such a way that it was developing more and
more into a circle (Aqvist 1977:39). Given the logical form of (44c) in
(44'c), (44c) does not imply that there was a circle, and there is a circle
and there is definitely going to be a circle.

The logical operator for sentences with activity verb phrases is
BeingA, where A is a sentence. By definition, BeingA means that it was at
least once the case that A at t, it is the case that A at to, and it is going to be
the case that A at t', where t is the immediate past, to this moment and t' the
immediate future. It is apparent that BeingA asserts that A is true
throughout an open interval around the point of evaluation. The operator

BeingA with an existential quantifier gives (45a) the much simpler logical
form in (45b).

(45) a. John is pushing a cart.
b. 3x (Being (Push (j, x) & Cart (x))

In contrast to (44'c), (45b) implies that there was a cart, there is a cart and
there will be a cart.

One obvious advantage of Aqvist's approach is that his analysis
correctly captures the intuition that when we say (44c) and (44a) we know
there is a longer period than the evaluation point, e. g. this moment, around
which (38c) and (39a) must be true if (38c) and (39a) is true at the moment
evaluated. This is the intuition that Bennett has talked about and Dowty's
approach fails to capture. However, as Parsons (1989:229-30) notes,
Aqvist's analysis has several drawbacks. I think that the most serious
drawback is that Aqvist's approach is not able to provide a uniform

analysis for progressive sentences. In the above discussion two different

anal
accr
sent
cont

IIICZI

hisr
date.
tnayi
phras
to cor

"Cale.

(47);]I
IIII'Qh

andlh.

m to.

I47b)

. l

OIpop

 

“ithqt

prOOreS

b

 

57

analyses, a complicated one for the progressive sentence with
accomplishment verb phrases and a simpler one for the progressive
sentence with activity verbs, are mentioned. For example, in case of the

complicated operator, a third analysis is required to account for the

meaning of (46a), as in (46b) (Aqvist 1978:49).

(46) a. Mary is closing the door.
b. Being (Door (x) & m IsClosing x InSuchAWayThatltIs-
BecomingMoreAndMoreTheCaseThat x IsAClosedDoor)

It is noticed that the last part of the logical form in (46b) is completely
different from those in (44'c) and (45b). In this fashion, Aqvist's analysis
may require many different analyses based on the meaning of the verb
phrases involved. For instance, I believe that different analyses are needed
to correctly characterize the meaning of the sentences in (47) if they are

treated in Aqvist's approach.

(47) a. Mary is reading a book.
b. Mary is eating a bag of pop com.

(47) are progressive sentences with accomplishment verb phrases. (47a)
involves something that exists, while (47b) involves something that exists
and that may not exist in the future. In (47a), nothing happens to the book
in (473), and it only happens that Mary learns something by reading. In
(47b), on the other hand, something happens both to Mary and to the bag
of pop corn; Mary will be full and the bag will be empty.

To summarize briefly, we note that Aqvist's improved tense logic
with quantifiers seems to characterize some aspects of the semantics of the

progressive sentences correctly in terms of the notion of open intervals,

\the
Aqv

unfit

2.22

terb
tense.
anse
and n
have;
0116 It
thatl
lh€}’fi

“TUOI

OI Iimt

thenL
THn€.:
recOgnj
mOugh

MOIllag

 

andtoifl

pointed

 

58

where, for example, Dowty's inertia worlds fail. On the other hand,
Aqvist's approach suffers from the drawback that it is not able to provide a

uniform analysis of progressive sentences.

2. 2 GENERAL PROBLEMS WITH PREVIOUS APPROACHES

As the discussion in the previous sections indicates, approaches to
verb categorization and the truth conditions for sentences in different
tenses and aspects suffer from serious inadequacies or drawbacks, which
arise either in the way they view the relation between verb categorization
and truth conditions for sentences or in the specific techniques that they
have adopted. Generally speaking, how one views a problem determines in
one way or the other his or her approach to the problem. It is in this sense
that I believe that most authors have problems with their approaches, since
they fail to take the set of questions in section 1.1 into consideration. In
section 1.3, I have outlined the three-dimensional linguistic representations
of time in languages as responses to the set of questions in section 1.1. If
we view the problems that previous approaches have in terms of the three-
dimensional linguistic representation of time, e.g., Linear Time, Frame
Time, and Situation Time, we find that authors of previous analyses fail to
recognize one or the other dimension of linguistic representations of time,
though they all use the dimension of Linear Time. For example, in
Montague's (1970, 1973) analysis of tense/aspect, Frame Time is not taken
into consideration. As a result, Montague's approach incorrectly gives
logically equivalent analyses to (48a) in the present perfective tense/aspect
and to (48b) in the simple past tense, as repeated below on the next page, as

pointed out by Bennett and Partee (1978).

 

 

 

 

 

59

(48) a. Bill has read a book.
b. Bill read a book.

The incorrect analyses result from a model in which the description of truth
conditions is given only in terms of Linear Time. According to this model,
the event of Bill's reading of a book takes place at a time t which precedes
the speech time or the evaluation time t'. When truth conditions are given
based on the condition that t precedes t', then (48a) and (48b) are logically
equivalent. This analysis is incorrect, because intuitively we know that both
sentences in (48) have implicit time adverbials, which may be explicitly

stated in the sentences in (48').

(48') a. Bill has read a book (today/this week/this month/this year).
b. Bill read a book (yesterday/yesterday afternoon/last week).

In (48a), both the event of Bill's reading of a book and the utterance of
(483) take place within one frame of Frame Time, that is, the period of time
specified by one of a number of potential Frame Time adverbials in (48'a).
In (48b), on the other hand, the event of Bill's reading of a book and the
utterance of (48b) take place in two different frames of Frame Time. If
Bill's reading of a book takes place in the first frame of time which may be
specified by any of the adverbials in (48'b), the speech or evaluation takes
place in the second frame of time which is implicit in both (48b) and (48'b).
This is a simple example that illustrates one of the three dimensions of
linguistic representations of time where previous approaches fail. A full
treatment of the English present perfect will be presented in Chapter Five.
Another dimension of linguistic representations of time where the
previous approaches fail is Situation Time. Generally speaking, the failure

is seen in the repeated emergence of the 'imperfective paradox' in

Mom
and i
is a d
to pre
tense
of the
with 3

here.

011 Par-L
"metciil
lies In I
do“ Pa.

CharaClcl

 

Sikkim}

deslé'nat.
and the

.1
_

apfllod

 

60

Montague's (1970, 1973) analysis, in Bennett and Partee's (1978) analysis,
and in Dowty's (1972, 1977, 1979). Authors who seem to realize that there
is a dimension of linguistic representations of time, Situation Time, still fail
to present an accurate characterization of Situation Time in semantics, and
tense and aspect logic. A typical example of this kind is in Parsons' analysis
of the progressive sentences with accomplishment verb phrases (423) and
with activity verb phrases (43a) in subsection 2.1.5, which are repeated

here.

(42) a. John was reading a book.
b. (3t) {t < now & (3e) [Writing (e) & Subject (e, John) &
Object (e, a book) & Hold (e, t )]}
(42') a. John read a book.
b. (it) {t < now & (3e) [Writing (e) & Subject (e, John) &
Object (e, a book) & Cu! (6, t )]}
(43) a. John was running.
b. (it) {t < now & (Ele) [Running (e) & Subject (e, John) &
Hold (e, t)]}
(43') a. John ran.
b. (St) {I < now & (3e) [Running (e) & Subject (e, John) &
Cul (e, t)]}

On Parsons' analysis, (423) and (433) have logical forms with identical
operators as in (42b) and (43b) respectively. The problem with the analysis
lies in that (423) does not entail (42'a), whereas (423) entails (42'a). Where
does Parson's analysis fail? I think that his analysis fails in not correctly
characterizing the bound and unbound nature of Situation Time.
Specifically speaking, the event of writing a book takes a period of time of
designated length, as represented by the relation between the verb write
and the noun phrase a book. On the other hand, the event of running takes

a period of time of arbitrary length, as represented by the relation between

tilt? \t
thsd

repre~
truth i
dill-SCH

mpms

 

 

61

the verb run and the zero complement. Parsons' analysis does not capture
this distinction.

It seems that full considerations of the three dimensions of linguistic
representations of time is required in providing an adequate analysis of the
truth conditions for sentences in different tenses and aspects and with
different categories of verbs. Lack of consideration of any aspects of these

representations of time results in inadequacy or in incorrect analysis.

2.3 SUMMARY OF CHAPTER TWO

In this Chapter, I have examined the relation between verb
categorization and truth conditions for sentences in different tenses and
aspects. This relation exists, since both of them involve the notion of time.
I have reviewed a number of approaches to verb categorization or to the
description of truth conditions for sentences in different tenses and aspects,
or to both. As I noted in the review of previous approaches, this relation
between verb categorization and truth conditions for sentences in various
tenses and aspects is not considered at first in semantics, and tense and
aspect logic, regardless of Vendler's (1957, 1967) observations of different
temporal structures in verbs or verb phrases. The emergence of intervals
of time as the reference point, replacing the traditional moment of time, is
an indication that linguists and logicians finally recognize the existence of
this relation. However, interval semantics does not solve all the problems
concerning verb categorization and truth conditions for sentences in
various tenses and aspects, since linguistic representations of time are much
richer and more complicated than ever considered. If linguistic
representations of time are classified along the three dimensions: Linear

Time, Frame Time and Situation Time, as I propose in section 1.3,

prob!
of Frtl
resear
panic
can 0:
for se,

aspect

 

62

problems of previous approaches are generally found along the dimensions
of Frame Time and Situation Time. This indicates that we need further
research on the linguistic representations of time in natural languages,
particularly alone the lines of Frame Time and Situation Time, before we
can offer an adequate analysis of verb categories and of truth conditions

for sentences in various tenses and aspects in semantics, and tense and

aspect logic.

 

 

-——r
' 3" p: / _
_-——-_ ___ —

 

Chapter Three
LINEAR TIME AND FRAME TIME

3.0 INTRODUCTION

In this chapter, I focus on two dimensions of the structure of time in
natural languages and formally define the two notions Linear Time and
Frame Time that I informally put forward in Chapter One. I mainly
investigate time as represented by tense, aspect and temporal adverbials in
natural languages, and discuss how the linguistic representations of time by
tense, aspect and temporal adverbials relate to these two dimensions of
time, Linear Time and Frame Time, since these two dimensions of the
structure of time are essential in a description of the truth conditions for
sentences that involve temporal relations represented by tense, aspect and
temporal adverbials. In this chapter, I try to answer some of the questions
in (2) in section 1.1. with linguistic evidence from English and Chinese. I
believe that an understanding of the relationship between these two
dimensions of the structure of time and their linguistic representations
eventually contributes to the answers to the set of questions in (1) in section
1.1. The concepts of tense and aspect are ambiguous. It is particularly the
case when tense and aspect in English and Chinese are concerned. Some
authors (Montague 1970, 1973, Homstein 1991) do not distinguish tense
and perfect aspect in English, while others (Bennett and Partee 1978,
Comrie 1976, 1985) treat tense and aspect differently. However, I think

that my study of Linear Time and Frame Time will contribute to the

63

under
defin.

Thne

 

Fanng
and (1
their.
deteh'
Tnne.

torse;

3i

eontn.
IIInett
the strt
discus5
POPUla

[Cl Ca]

 

UlllI ed

1
COmen
i”distal
mbugh
braIIL‘hi
at“boys

adflpi I

64

understanding of tense and aspect, and that tense and aspect are eventually
definable in terms of the three dimensions of time: Linear Time, Frame
Time and Situation Time.

In my discussion of linguistic representations of Linear Time and
Frame Time, I make particular reference to temporal adverbials in English
and Chinese. I note that there is a lot of differences among languages in
their syntactical representations of time. However, I think that a theory
developed in terms of the three dimensions of time, Linear Time, Frame
Time and Situation Time, is essentially adequate to capture truth conditions

for sentences that exhibit different temporal relations in all languages.

3.1 LINEAR TIME: ONE DIMENSION OF TIME

Philosophically speaking, the structure of time has long been a
controversy since Aristotle, and the problem may not be settled for a long
time to come. In this section, therefore, I simply consider one dimension of
the structure of time as represented in natural languages, though my
discussion may involve some properties of the structure of time that are
popular topics in the philosophical study of time. This discussion may
reveal that time as represented in natural languages does exhibit some
universal properties.

Time, as I have already mentioned in the previous two chapters, is
conventionally represented as a straight line with an arrow on the right side
indicating its development from past to future (cf. Reichenbach 1947),
though it may well be represented in other ways, such as open time,
branching time and cyclic time (Newton-Smith 1980), depending on
authors' philosophical commitment to the notion of the structure of time. I

adopt the linear representation of time as a straight line in my discussion of

65

Linear Time, as in Comrie (1985z2), Jespersen (1965:256-7), Reichenbach
(1947:290-2) and Smith (1991:136-7). This is graphically illustrated in

Figure 1 below.

_______________________________________________________ >

Past Present Future

Figure l: A Linear Representation of Time

One view of this linear representation of time is that time is a set of points
without duration but with precedence (cf. Benthem 1983), as expressed in
tense and aspect in natural languages (Comrie 1985, Reichenbach 1947,
Smith 1991). Comrie (1985) treats points in time as location in time, which
may be grammaticalized in the form of tense. Reichenbach (1947)
considers the relationship among points of time as a linear ordering of
point of speech, point of reference and point of the event, which underlies
all the temporal relations expressed in tense and aspect. Homstein (1991)
adopts Reichenbach's linear ordering view and incorporates it in the
Govemment-Binding (GB) framework. Basically following Reichenbach,
Smith (1991) treats the relationship among points of time as that among
Speech Time, Reference Time and Situation Time.

My notion of Linear Time shares one common ground with previous
approaches mentioned above, in that time is a set of points without duration
but with precedence. In this sense, Linear Time is concerned with location
in time, as Comrie's (1985) approach is. The notion of points of time

without duration but with precedence is a relational view of time. A

 

 

repregt

leat ir.

one 0

HOr]

Spec

hfiit".

66

temporal expression related to a point of time may well have a duration,
but its duration is not of relevance when location in time is concerned, as

illustrated in (1).

(l) a. John came before yesterday.
b. John left school after the spring.

Yesterday in the expression of before yesterday in (13) denotes a period of
time, and so does the spring in after the spring. However, the duration of
the temporal expressions is irrelevant in (1), since the temporal relation
represented in (l) is one between the occurrence of John's coming or
leaving and the point of time yesterday or the spring. The relationship is

one of precedence. This is true with tense, as the examples in (2) indicate.

(2) a. John read a book.
b. John will read a book.

The sentence in past tense in (23) shows a temporal relation between the
time of the utterance of (23) and the time of the event of John's reading of
a book, that is, the event precedes the utterance of the sentence. A similar
relation is found in (2b), where the event follows the utterance of the
sentence. The event of John's reading a book and the utterance of the
sentence are both of some duration, but the duration is irrelevant here
either, since only the precedence relation is involved.

My notion of Linear Time departs from Reichenbach's (1947),
Homstein's (1990) and Smith (1991) approaches in avoiding introducing
event time into the picture, and avoiding linearly ordering event time with
speech time and other point of time in location in time, since event time

belongs to another dimension of the representation of time, Situation Time.

67

Reichenbach (1947:288-297) assumes that point of speech , point of
reference and point of the event are linearly ordered in determining the
temporal relations in tenses and the perfect aspect in English, as in (3)
where point of speech is symbolized as S, point of reference as R and point

of the event as E.

(3) Structure Examples
3. E--R--S--> John had read a book.

b. R, E--S--> John read a book.

c. E— -S, R--> John has read a book.

(I. S, R, E--> John reads a book.

e. S, R- —E--> John will read a book.

f. S- --E R---> John will have read a book.

Homstein (1990:14-5) further argues for the linearity of S, R, E, making it
explicit that the left point is interpreted as temporally earlier than the right
point when they are separated by a line, and that the points are interpreted
as contemporaneous when they are separated by a comma. I believe that
this linear representation of points of time makes wrong predictions about
location in time. For example, the sample sentence in (3f), as repeated in
(4) below, may describe two scenarios: one fits the temporal structure in

(3f) and one does not.
(4) John will have read a book.

In the first scenario, John may start to read a book after speech time, and
then finish reading it before reference time. This is the temporal structure
that Reichenbach's and Homstein's theories predict. In the other scenario,
John may start to read a book before speech time, and then finish reading it

after speech time but before reference time. The second temporal structure

68

is where Reichenbach's and Homstein's theories make wrong predictions.
A second piece of evidence against introducing event time into this
linear picture lies in the problem of the reportive use of a sentence, as
pointed out in Bennett and Partee (1978), and Bennett (1977).1 In
Reichenbach's (1947:290) and Homstein's (1990:14-5) theories, event time,
speech time and reference time may be contemporaneous in addition to the
linear structure. First let me show the problem of event time and speech
time being contemporaneous. It seems that sentences in (5) are fine with the

reportive use.

(5) a. (Look.) Bill moves.
b. (Look.) Bill runs.

It appears that it is acceptable for event time and speech time to be
contemporaneous, when the verbs do not represent a definite period of
time, as walk and run do not. However, it appears to be difficult for event
time and speech time to be contemporaneous, when the verbs or verb
phrases do represent a definite period of time, as walks to the bookstore

and write a book do in (6)

(6) a. (Look.) ?Bill walks to the bookstore.
b. (Look.) ?Bill writes a book.

The sentences in (6) do not readily allow a reportive interpretation but are
more readily available for a habitual interpretation. In Chinese, such
sentences always get a progressive reading, instead of a reportive reading,
as I have illustrated in section 2.1.1. The same problem exists in the

temporal relationship between event time and reference time. Event time

 

1 Subsection 2.1.2 contains a detailed discussion on this topic.

and n

St‘lllfil

Then
(1947
etent
interr
exist

defin

1n18
the S

readi

Pret

 

3°31
limp.
the .

POSi

 

 

69

and reference time may be contemporaneous, if the verbs involved in the

sentences are verbs that do not represent a definite period of time, as in (7).

(7) a. Bill moved when John came.
b. Bill ran when John came.

The when-clause in (7) is considered reference time in Reichenbach
(1947:293). It is possible to have an interpretation in which the point of the
event is concurrent with the point of reference in (7). However, an
interpretation of concurrent point of event and point of reference does not
exist, when the sentences involve verbs or verb phrases that represent a

definite period of time, as in (8).

(8) a. Bill walked to the store when John came.
b. Bill read a book when John came.

In (8), only the interpretation is available that the event of Bill walking to
the store follows the event of John's coming, and that the event of Bill
reading a book follows the event of John's coming.

In the above, I have illustrated my common ground with some
previous approaches about location in time. I have particularly argued
against the introduction of event time in the linear picture of location in
time. With my discussion and argument about the linear representation of
the structure of time in natural languages, I think that I am now in a

position to define more precisely the concept of Linear Time in (9).

(9) Linear Time is a set of points of time without duration but
with precedence ordered in relation to speech time in a linear
structure.

70

In Linear Time > is used to represent the temporal relation 'earlier than'.
Temporal order in Linear Time may be represented as in Figure 2, where

S stands for speech time and t with indexes for points of Linear Time.2

--------- nt > 2t > 1t > S > t1 > t2 > tn --------->

Past Present Future

Figure 2: A Representation of Linear Time

In this linear picture, any 1 stands in relation to S time and other ts,
establishing a relation between t and S, and between land t. Formally, t> 8
(past), t: S (present), S > t (future). Past, present and future are simply
temporal orders relative to S time in this linear representation of time in
natural languages. These relations are all expressible in natural languages. I
think that the differences between languages regarding Linear Time are a
problem of degree of grammaticalization of Linear Time. For example,
Linear Time is grammaticalized in the form of tenses and probably in the
form of 'aspects' as well, and is expressible in the form of temporal
adverbials in English. It is apparent that Linear Time is of deictic nature,
since temporal orders expressed by tense, aspect and temporal adverbials
are relative to S.

What formal properties does temporal order exhibit in Linear Time

 

2 I symbolize points of time earlier than S by indexing on the left upper side of t,
and symbolize points of time later than S by indexing on the right upper side of t. This
system of indexing is to be used henceforth in this study when it is needed. When indexing
does not involve speech time, I simply symbolize by indexing on the right lower side of t.

 

in na
ethn*
inrel
earth
FOlh‘

defin

”(1

Senit

 

71

in natural languages? Temporal order or precedence in natural languages
exhibits a number of formal properties. Those properties may be defined
in relation structures (T, R), where R stands for a temporal relation such as
earlier than, and T for a linear representation of a set of points of time.
Following Benthem (1983) and Landman (1991), I present the formal

definitions of those properties in (10).

(10) a. R is transitive iff Vt], t2, t3 ET :R(t1, t3) A R(t3, t3)—>

RUE t3)
b. R is asymmetric iff Vt], t2 ET: R(t1, t2)—a-R(t2, t1)

In (10), I assume that each temporal reference point is indexed with a
natural number. The sequence of natural numbers are discrete, since for
any number in the sequence there is a unique next number (Newton-Smith
1980:112). Temporal order with such index exhibits the characteristic of
discreteness. The notion that time is discrete is, of course, controversial in
the philosophical study of time. However, Linear Time as represented in
the structure of time in natural languages may exhibit the characteristic of
discreteness, since only the precedence relation is considered on this
dimension of the representation of temporal relations.

One of the formal properties in Linear Time, transitivity, as defined

in (103), is shown in the implications among the sentences in ( l 1).

(1 l) a. John came earlier than Mary did.
b. Mary came earlier than Bill did.
c. John came earlier than Bill did.

If (113) and (1 lb) are true sentences, then (11c) is necessarily a true

sentence, simply because of the transitivity property of the linear

repre
Mart
lingu
piece
the rt

form;

(197'

The L

“ hit‘.

 

72

representation of temporal order. The event John came precedes the event
Mary came which precedes the event Bill came, as expressed by the
linguistic structure earlier than in English. Therefore. the event John came
precedes the event Bill came. One formal relation of transitive relations in
the representation of temporal order is converseness. Converseness is
formally defined in (12), following Allwood, Anderson and Dahl
(1977:90) .

(12) A relation R is said to be the converse of another relation S, if
R (x, y) is true whenever S (y, x) is true. The converse of a

relation R is written K.

The converse of the temporal relation earlier-than is the later-than relation,

which is illustrated in the sentences in (13).

(13) a. Mary came later than John did.
b. Bill came later than Mary did.
c. Bill came later than John did.

These sentences in (13) have the same type of implications and the same
truth conditions as the corresponding sentences in (11) have. The converse
relation in Linear Time uniquely underlies certain entailment relations in
sentences in natural languages. A third property of Linear Time is
asymmetry, which is exhibited in the relations between the sentences in

(14), where (143) does not imply ( 14b) nor does ( 14b) imply (14a).

(14) a. John came earlier than Mary did.
b. Mary came earlier than John did.

Given the asymmetrical temporal relation expressed in the pair of sentences

in (14), either (143) is true or (14b) is true at a particular point of time. If

 

 

 

73

(143) is a true at a point of time, then (14b) is not a true at the same point
of time, assuming, of course, that the same events of John's coming and
Mary's coming are referred to in both sentences. These formal properties
of Linear Time are important in distinguishing Linear Time from Frame
Time and Situation Time.

It is apparent, as my discussions and definition indicate, that Linear
Time is different from Reichenbach's (1947) temporal structure. Linear
Time is completely linguistically motivated, while Reichenbach's temporal
structure is not. In the manipulation of S (speech time), R (reference time),
and E (event time), Reichenbach needs R to represent the temporal
structures, such as those in the English past perfect (E -- R -- S) and future
perfect (S -- E -- R), while R seems to be redundant and have no linguistic
motivation in the representation of temporal structures in other tenses like
simple past (E, R -- S), preSent (S, R, E) and simple future (S --R, E).
Homstein (1990:10-5) argues for Reichenbach's tense theory. On the one
hand, he recognizes that the temporal relationship between S and E remains
the same for the simple past tense, with or without R. On the other hand, he
argues that R obtains in syntactic relations even when it is not semantically
visible. In addition to the existence of R in the past perfect and future
perfect, Homstein (1990:90-1) argues that the existence of R accounts for
the upper limit of two temporal adverbs per sentence and the interaction
between the adverbs, and for the formation of complex tensed sentences. I
think that those problems can be accounted for more naturally from the
perspectives of Linear Time and Frame Time without R.

In short, I started with the common ground and the difference that
my notion of Linear Time has with other approaches to location in time. I

have particularly pointed to problems in introducing event time in the

 

Ur
_ 'T
— — — —

 

 

74

linear representation of time, as in Reichenbach (1947) and Homstein
(1990). I have defined Linear Time in (9), and discussed some formal
properties of Linear Time. In my approach, Linear Time represents the
dimension of temporal precedence relations in the structure of time in
natural languages, and is linguistically motivated, as compared with

Reichenbach's tense theory and theories along his line.

3.2 REPRESENTATION OF LINEAR TIME IN ENGLISH

I believe that Linear Time, as defined in (9), is represented in
English by three means, namely, tense, aspect and temporal adverbials,
while only the former two are grammaticalized morphologically. I first
examine the distinction between finite clauses and nonfinite clauses, which
is a result of the grammaticalization of Linear Time. It is generally
accepted that in English there are two types of clauses: finite and nonfinite
(Homstein 1990z8). The difference between a finite clause and a nonfinite

clause lies in the fact that the former bears a tense marker while the latter

does not, as in (15).

(15) a. Bill visited John.
b. Bill plans to visit John.

(153) is a finite clause, which is marked by -ed. On the other hand, there is
a finite clause Bill plans and an nonfinite clause to visit John in (15b). In
(15b), the finite clause is marked by a tense marker -s, while the nonfinite
clause is not. The tense marker in (153) simply indicates that the event
occurred at a time earlier than S. The tense marker in (15b) indicates that
the state of having a plan holds during S. The question is whether there are

any temporal relations between the nonfinite clause's tand S? The answer

 

 

 

to thi
nonli
the a
is rel
betttt
the ST
finite
the e1
[0 Ot-I
deno:
nont‘.
It ith
\t ith
aITtrr
lSanq
Perle
dentq
Claus
film I
COnIC

[IICCU

Ht)“ ‘

dentx

 

75

to this question depends on the verb forms (i.e. simple or perfective) in the
nonfinite clause. With respect to the simple verb form in nonfinite clauses,
the answer is affirmative, so long as the finite clause that a nonfinite clause
is related to has a temporal relation with S. However, the temporal relation
between the infinitival clause with simple verb forms and S is indirect in
the sense that a nonfinite clause has to establish a temporal relation with the
finite clause before its temporal relation with S is established. For example,
the event denoted by a nonfinite clause in the simple form (as in (15b)) is
to occur later than the event denoted by the finite clause. If the event
denoted by the finite clause occurs later than S, as in (15b), then the
nonfinite clause's event is also later than S. This is generally the situation
with nonfinite clauses in the simple form in English. On the other hand,
with respect to the perfective form in nonfinite clauses, the answer is not so
affirmative, since the temporal relation between the nonfinite clause and S
is ambiguous in some cases. An event denoted by a nonfinite clause in the
perfect form, such as to have met Mary occurs earlier than the event
denoted by the finite clause dominating it. If the event denoted by the finite
clause occurs earlier than S, then the event denoted by the nonfinite clause
also precedes S, as in (163). If the event denoted by the finite clause occurs
contemporaneously with S, then the event denoted by the nonfinite clause

precedes S, as in (16b).

(16) a. Bill was very happy to have met Mary there.
b. Bill is very happy to have met Mary there.
c. Bill will be very happy to have met Mary there.

However, when the event denoted by the finite clause follows S, the event

denoted by the nonfinite clause may precede S or follows S, as in (16c),

76

where exist two interpretations: i) Bill has already met Mary and he will be
happy, and ii) Bill will be happy after he meets Mary. In the first reading,
the event denoted by the nonfinite clause precedes S, while in the second
reading the event denoted by the nonfinite clause follows S.

Given the above observations, temporal relations in nonfinite clauses
depends on temporal relations between finite clauses and S in most cases. It
is in this sense that only finite clauses in English are considered in the

representation of Linear Time in this section.

3.2.1 TENSE, THE PERFECT AND LINEAR TIME

A finite clause in English is morphophonologically marked with
respect to tense in most cases. Generally speaking, tense is considered a
grammatical notion in the sense that tense is grammaticalized location in
time in English (cf. Comrie 1985:9-13). As for the number of tenses in
English, there seems to be a disagreement.3 Three tenses representing three
times appears to have become a part of the Western grammatical tradition
since Aristotle (Binnick 1991:8). These three tenses are past tense, present
tense and future tense, corresponding to the three times, as exemplified in

(17) respectively.

(17) a. Bill loved Mary.
3' (t > S)
b. Bill loves Mary.
b' (t = S)
c. Bill will love Mary.
c' (S > t )

 

3 I believe that the disagreement about number of tenses in English results from the
lack of a clear and specific definition of tense, as I argue later in this subsection.

77

On the other hand, some linguists and philosophers (Montague 1970,
Homstein 1990:15) implicitly or explicitly hold the view that there are six
tenses in English, namely, past, present, future, past perfect, present

perfect and future perfect, the last three of which are shown in (18).

(18) a. Bill had loved Mary.
b. Bill has loved Mary.
c. Bill will have loved Mary.

There is a great disagreement as whether the last three are tenses at all. I
will come back to this problem in the following. Let me first introduce the
controversy over the first three tenses: past, present and future in English.
Given Comrie's (1985) definition that tenses are grammaticalized location
in time, past tense and present tense in English are well accepted, since both
are represented by verbal inflections, as in (19) and (20) respectively, in

contrast to a verb's nonfinite form.

(19) a. Bill plans to study for the meeting.
a' (t = S)
b. Bill studies to plan the meeting.
b' (t = S)

(20) a. Bill planned to study for the meeting.
3' (t > S)
b. Bill studied to plan for the meeting.
b' (t > S)

Given the representation of the temporal relations in (19) and (20) in

Linear Time, past tense not only represents a point of time, but also

represents a relation: the point of time precedes S, as in (203') and (20b').
However, linguists and philosophers disagree as to whether there is a

future tense in English. The argument that future tense does not exist in

78

English runs as follows. First, auxiliary verbs in the so called future tense

sentences are not tense markers (Binnick 1991, Comrie 1985), as in (21).

(21) a. Bill will go there tomorrow.
3' (S > t )
b. Bill shall go there tomorrow.
b' (S > t )
c. Bill can go there tomorrow.
c' (S > t )

Go in (21) is a nonfinite verb. In contrast to the past forms would, should
and could, will, shall and can do not have future forms. In addition, will in
(213) expresses willingness, shall in (21b) obligation and can in (21c)
possibility or ability. These are modals instead of tense markers

(cf. Partee 1973). Secondly, the notion of future can be expressed without

the above auxiliaries, as in (22).

(22) a. Bill is about to go there tomorrow.
3' (S > t )
b. John is going to go there tomorrow.
b' (S > t )

(21) and (22) seem to suffice to show that future in English is not a tense in
the sense of the definition in Comrie (1985). Extreme cases of the non-
existence of a future tense in English are exemplified in the following
sentences in (23), where the verbs are in the present tense form but the

events denoted by these sentences are understood to take place in future.

(23) a. The sun rises at 6 o'clock tomorrow.
3' (S > t )
b. The train arrives at noon tomorrow.
b' (S > t )

79

The time adverbials seem to determine the temporal relations in (23)
instead of the present tense marker.

However, McGilvray (1991) argues that sentences like (23) are
limited to 3 small number of verbs. The sentences in (23) are related to
'reduced future' construction which are semantically or pragmatically
conditioned in a scheduling state (McGilvray 1991:41-2). 'Reduced future'
sentences may be hedged, as in (23').

(23') a. The sun rises at 6 o'clock tomorrow, I guess/think/believe.
a'. (S> t )
b. The train arrives at noon tomorrow, I guess/think/believe.
b' (S > t )

In terms of scheduling, for example, (23b) may have the form in (243).

(24) a. The train is scheduled to arrive at noon tomorrow.
b. (S < t )

Therefore, McGilvray argues that there is a future tense marked by will in
English. He supports his argument with the following pairs of sentences,
where the sentences in the progressive form contrast those with will in

acceptability, as in (25), (26) and (27).4

(25) a. *John is thinking tomorrow.
3' (S > t )
b. John will be thinking tomorrow.
b' (S > t )
(26) a. *The engine is misfiring tomorrow.
3' (S > t )
b. The engine will be misfiring tomorrow.
b' (S > t )

 

4 Barbara Abbott suggests that be going to may also be a future marker if will is a
future marker, as McGilvray argues. Is going to may replace will in (25), (26) and (27).

80

(27) a. *The rock is falling tomorrow.
3' (S > t )
b. The rock will be falling tomorrow.
b' (S > t )

When we compare the ungrammatical a—sentences with the grammatical b-
sentences, we find that verb types seem to be irrelevant with regard to the
progressive forms. It is apparent that generally speaking the progressive
forms can not be used to express future in English.

It should be stressed that in my approach tense is a grammaticalized
representation of temporal orders in Linear Time. By grammaticalization,
I mean the function-specific morphophonological marking of temporal
orders, such as the past tense and present tense morphemes in English.
Given this definition, it is clear that future that is specified by semantic and
pragmatic conditions is not future tense. In McGilvray's argument for the
existence of a future tense in English, it appears that future tense and future
are not distinguished. It is noticed that there are two verb forms, finite and
nonfinite, in (22) and (24). As I pointed out at the beginning, in most cases,
nonfinite clauses depend on finite clauses with regard to temporal relations
with S. Events and activities denoted by simple nonfinite verbs without an
aspect marker generally take place at a time later than the time at which
events and activities denoted by dominating finite verbs occur. I believe
that the nonfinite verb forms contribute to the specification of future time
in (22) and (24), in addition to time adverbials. On the other hand, I argue
elsewhere that grammaticalization is not categorical, but a matter of degree
(Zhou 1993a, 1993b). The evidence suggests that will in the sentences in
(26b) and (27b) is used as a tense marker and represents future, since the

agentless situations are less likely to express any willing, and future can not

81

be represented by other means except by will and probably be going to as
well. On the other hand, will not only just marks the temporal order but
also represent willingness. Thus, the future tense in English may be only
partly grammaticalized if it is compared with present tense and past tense
that are completely grammaticalized.

Another way in which Linear Time is grammaticalized in English is
in the perfect. What is traditionally called perfect 'aspect' in English
actually functions as tense and as aspect as well. This may be one of the
reasons that some authors (e.g. Montague 1970, Homstein 1990) simply
treat perfect in English as tense, while others (e.g. Comrie 1976:52) seem
to treat it as a special kind of aspect, and still others (Smith 1991:242-6)
hold that the perfect in English may function both as aspect and as tense. I
think that the perfect in English does function both as tense and as aspect if
it is viewed from the perspectives of Linear Time and Situation Time. I
argue that the English perfect is a tense when it represents Linear Time,
and it is an aspect when it involves Situation Time. For my purposes here, I
only look at the representation of Linear Time by the perfect in English,

and treat it as tense in this sense. The perfect involving Situation Time as
aspect will be dealt with in Chapter Five.

First, let me show how the present perfect functions as tense. In
terrns of Linear Time as defined in (9) and represented in Figure 2, tense is
a grammaticalized representation of Linear Time. It is apparent that the
English present perfect or the perfect in general is grammaticalized in the
form of a bound morpheme -ed or phonological alternations with a variety

of allomorphs, including Q, in addition to have, as in (28).

82

(28) a. Bill has cleaned (cf. clean) the table.
b. Bill has broken (cf. break) a glass.

The present perfect in (28) functions as tense in the sense that the past
participle of the main verb represents t, while the present form of have
represents the relation of t being earlier than S, in Linear Time. This is

formally represented in (28').5
(28') (t > S)

The representation of Linear Time in (28') clearly reflects the intuition
that the events of Bill clean the table and Bill break a glass take place
before S (cf. Smith 1991:146-7). The linear representation of (28) in (28')
is not any different from the linear representation of time in past tense.
This is exactly what Linear Time represents.6 Further linguistic evidence
for this representation of temporal relations comes with questions and
aflSWers concerning the same situations, as illustrated in (29) and (30),
which are when questions of the statements in (28) and are answers to the

questions of the statements in (28) (cf. Comrie 1976:54—55).

(29) a. *When has Bill cleaned the table ?

b. When did Bill clean the table ? (t > S)

c. Bill cleaned the table at three o'clock. (t > S)
(30) a. *When has Bill broken a glass ?

b. When did Bill break a glass ? (t > S)

c. Bill broke a glass earlier this morning. (t > S)

5 In Linear Time, forms of auxiliary verbs represents two kinds of temporal
relations: precedence relations with S and precedence relations with another point of time.

6 It is noticed that the English present perfect represents more than Linear Time. It
involves the dimension of Frame Time as well. I will discuss how Linear Time, Frame
Time and Situation Time work together to account for the so-called viewpoint analysis of
tense and aspect in semantic approaches to tense and aspect (cf. Smith 1991) and in
discourse approaches to tense and aspect (cf. Thelin 1990).

83

The when-question and its answer in English clearly indicate that the
present perfect has 3 Linear Time representation similar to that of the
simple past. Then, what is the difference in temporal structures between the
sentences in the present perfect in (28) and those in the simple past in (29)
and (30)? The answer is straightforward. There is no difference in
temporal relations between those in (28) and those in (29 & 30) in terms of
Linear Time, since Linear Time only represents precedence relations. This
is the tense perspective of the perfect in English. It is obvious that there are
differences in temporal structures between the sentences in the present
perfect in (28) and the sentences in the simple past in (29 & 30). The
differences do not lie in Linear Time but in Frame Time, as I will discuss
later in Chapter Five. This is where previous approaches (e.g. Montague
1970, 1973) fail to account for the differences between sentences in the
present perfect and simple past.

The representation of Linear Time is also found in past perfect

sentences, as in (31).

(31) a. Bill had cleaned the table when John called.
b. Bill had broken a glass when John called.

In (31), the events/activities of Bill clean the table and Bill break a glass
take place at a time before the event of John call. For this reason, the past
perfect is called anterior past (Reichenbach 1947:296) or past in the past
(Jespersen 1965:262-3). The past forms of the verbs in (31) indicate that
the events/activities all take place before the time when the sentences in
(31) are uttered. The temporal relations in (31) may be represented in
(31') in terms of Linear Time, where 2t is the point of time represented by

the past participle in the main clause and It is represented by the simple

84

past in the subordinate clause, while the relation of being earlier than is

represented by the auxiliary of the past perfect.
(31') (2t>1t> S)

(31') correctly predicts the intuition native speakers have about the past

perfect sentences in contrast to simple past tense sentences, as in (32).

(32) a. Bill had broken a glass yesterday.
3' (2t > 1t > S)
b. Bill broke a glass yesterday.
b' (t > S)

(323) implies that Bill broke a glass before a particular point of time
during yesterday, since the past participle represents a point of time in the
past and the auxiliary had represents 3 relation of this point (2t) being
earlier than another point (It) which is not linguistically covert in this
sentence. The representation of Linear Time in (323') correctly
characterizes the relationship among 2t , 1t and S in (323). However, (32b)
does not have this implication (cf. McGilvray 1991:46—8), since simple
tense only represents the relation of I being earlier than S, as in (32b').

A similar relationship is found in the future perfect in contrast to the

simple future in English, as in (33).

(33) a. John will have cleaned the tables.
b. John will clean the tables.

(333) implies that John will have cleaned the table before a particular time
or a particular event in the future but after S, while (33b) simply says that

the event of cleaning will takes place after S. This implication with the

85

future perfect is correctly characterized in the representation of Linear

Time in (33').

(33')a.(S>tI>t2)or(1t>S>t2)
b.(S>t)

Linguistically, t1 is represented by past participle, and the relation of t1
being earlier than 12 is represented the auxiliary of the perfect, while the
relation of I32 being later than S is represented by will. It must be stressed
that it does not specify any temporal relation between t1 and S.7 Though t2

is linguistically covert in (333) t2 can be linguistically expressed, as in (34).

(34) a. John will have cleaned the tables by two o'clock this
afternoon.
b. John will have cleaned the tables by the time/when Mary
brings the food .

The temporal adverbials in (34) explicitly represent t2 in Linear Time.
The above examples illustrate that the perfect in English represents
Linear Time. It is by the perfect's function along the dimension of Linear
Time that the English perfect may be classified as tense. However, the
English perfect's role in Situation Time and relation with Frame Time
should not be neglected, if a complete picture of temporal relations is to be

considered. This picture will be presented in Chapter Five.

3.2.2 LINEAR TIME AND TEMPORAL ADVERBIALS
It was noticed in section 3.1 that the linear nature of the temporal

structure in language may be represented by temporal adverbials. This

 

7The event of cleaning may well be started before S or after S, and definitely ends
after S. This is what Reichenbach's (1947) tense theory fails to predict.

86

linear dimension of the temporal structure in natural languages is Linear
Time as defined in (9) and graphically illustrated in Figure 2. In English,

the representation of Linear Time is exemplified in the sentences in (35).

(35) 3. Today John has asked Mary to have lunch with him
tomorrow.
b. This morning John told Bill that he is going to watch a
movie this evening.
(36) 3. (1t> S > H)
b. (It > S > t1)

One of the functions that these temporal adverbials have in (35) is to
specify the temporal sequence of the activities expressed by the verbs
relative to S, as illustrated in (36). The relation of It being earlier than S is
specified by the present perfect, as discussed in section 3.2.1. I1 and the
relation of t1 being later than I t is specified by the nonfinite clause, but it
does not specify how much later t1 is than 1t. The relation of how-late is
specified by the temporal adverbial tomorrow. Other temporal relations

relative to S are found in the following examples.

(37) a. John left here a week ago yesterday.
3' (2t > 1t > S), where 2t = past tense, 1t = yesterday
b. John will leave here in a week from tomorrow.
b' (R: S > t1 > t2 ), where t1 = tomorrow, t2 = future tense

In general, temporal adverbials in English may represent the kind of
temporal relations that tenses and the perfect in English represent. They
never replace tense and aspect's function in English, but play the role of
specification in the sense that, for example, they specify how far in the
future or in the past a situation holds, as in (36). However, in some Bantu

languages such as Makaa and Nugunu, this role is played by tenses in the

it
It

in

Th

L11

ter

dis

SC

80

13

 

87

form of near and far future and near and far past (cf. Heath 1991, Orwig
1991).

However, temporal adverbials have more than one function in the
temporal structure of natural languages. In addition to their role in the
representation of Linear Time, temporal adverbials also represent Frame
Time. For example, in (35) and (36) both It and S are included in the
interval of time denoted by today and this afternoon respectively. This is

more clearly illustrated in (38), where E means 'temporal inclusion'.

(38) a. Mary left at three today.
b. (t > S ) (t 2 at three; F: Today; t E F)

This kind of temporal inclusion relations belongs not to the domain of
Linear Time, but to the domain of Frame Time. This double function of
temporal adverbials in representing Linear Time and Frame Time is not
distinguished in the studies of temporal relations (Prior 1967, Kamp 1979,
Dowty 1982), where both tense and temporal adverbials are treated as
sentential operators involving different scopes. This confusion results in the

so-called 'Adverbial Scope Paradox ' (Binnick 1992:310-9).

3.3 REPRESENTATION OF LINEAR TIME IN CHINESE

It is well accepted that there is no tense in Chinese. In other words,
verbs are not marked as to tense in Chinese. 'The language does not use
verb inflections to signal the relation between the time of the occurrence of
the situation and the time that the situation is brought up in speech' (Li and
Thompson 1981:184). Moreover, there is 3 notion that the kind of
temporal relations that tense represents do not seem to exist if a language

does not have tense markers. For instance, Wang (1985: 151) holds that

there
situat
DIR;
regar
21¢th
thong
mon
“on
not I
lack
type

up
For

his j

Russ
langi
For C
tteati,
Cnhei
“1th

Russc

88

there are two ways in which speakers treat the relations between time and
situations; focus on WHEN without regard to duration; and focus on
DURATION, its beginning and completion without regard to WHEN. Wang
regards Chinese as representative of the latter of the two. It is
acknowledged that Chinese is a language which is not rich in affixation,
though the language is developing from one with predominantly
monosyllabic words to one with an increasing proportion of disyllabic
words (cf. Chao 1968:194—198, Li and Thompson 1981:36). Chinese does
not represent plurality and gender morphologically either. However, the
lack of the affixation in Chinese does not necessarily mean that the WHEN
type of temporal relation, plurality and gender are neglected in Chinese.

However, the fact that Chinese does not have verb inflections to
express tense gives wrong impressions to linguists and philosophers as well.
For example, on tense, Russell (1956:248) made the following comment in
his 1918 lecture.

(39) The occurrence of tense in verbs is an exceedingly annoying
vulgarity due to our preoccupation with practical affairs. It
would be much more agreeable if they had no tense, as I
believe is the case in Chinese, but I do not know Chinese.

Russell intends to use language in a tenseless way in logic. However, to use
language in a tenseless way can not avoid the problem of temporal relations.
For example, Montague (1970, 1973) follows Russell's tenseless way in
treating language and runs into trouble, as Bennett and Partee (1972)
criticize, and I have reviewed in section 2.1.2. Moreover, a language
without tense inflections like Chinese is not used in a tenseless way as

Russell may expect, as is illustrated in the sentences in (40). The temporal

89

relations in (40) can be expressed in (41) in terms of Linear Time.

(40) 3. W0 kan le bao qu xuexiao.
I read Asp. newspaper go school
I will go to school after I read the newspaper.

b. Zuotian wo kan le bao qu xuexiao.
Yesterday I read Asp. newspaper go school
Yesterday I went to school after I read the newspaper.

(41) a. (S >t1>t2),

b. (2t > 1t > S)

In the utterance of (403), S is anchored at NOW, kan bao takes place at t1
and qu xuexiao at t2. In (40b), S is anchored at NOW, kan bao takes place at
2t and qu xuexiao at 1 t. The event denoted by the main verb tends to follow
S, unless temporal adverbials like zuotian indicate otherwise. I argue that a
sentence in Chinese expresses temporal relations relative to S when they are
uttered, as the sentences in (40) do, though Chinese does not have regular
morphological tense markers. The problem is only to what degree
temporal relations are grammaticalized in Chinese. I argue, however, that
some aspect markers in Chinese have a dual function as aspect markers and

as tense markers, as I show in the next two subsections.

3.3.1 THE DUAL FUNCTIONS OF ASPECT MARKERS IN CHINESE
The lack of regular morphological tense markers in Chinese appears
to render the distinction between finite clauses and nonfinite clauses
practically impossible. However, Tai (1985:50) argues that temporal
relations between clauses are represented in terms of the principle of
temporal sequence, which is stated as follows: 'the relative word order
between two syntactic units is determined by the temporal order of the

states which they represent in the conceptual world'. In other words, in the

90

unmarked case, temporal relations between clauses are reflected in the

syntactic order of the clauses, as in (42) and (43).

(42) a. [s[sl Zhangsan qu shudian] [52 mai san ben shu]].
Zhangsan go bookstore buy three C(lassifier) book
Zhangsan will go to the bookstore to buy three books.

b (S > t1 > t2 ), where t1 = go, and t2 : buy

(43) a. ?[s[51Zhangsan mai san ben shu ] [52 qu shudian]].

Zhangsan buy three C. book go bookstore
b. *(S > t7— > t1 ), where t1 : go, and t2 2 buy

(423) indicates that qu shudian syntactically and temporally precedes mai
san ben shu. Chinese relies on the match between syntactic order and
temporal order to correct represent Linear Time, as in (42b), since
Chinese lacks any morphological means. When the syntactic order in (423)
is reversed as in (433), the sentence seems to be odd. This oddness results
from the violation of the representation of the temporal relations in Linear
Time, as indicated in (43b). It is conceptually or semantically unacceptable
to reverse the temporal order, though it is not completely ungrammatical.
Thus, the unacceptability of (43a) is conceptual or semantic rather than
syntactical. The match between syntactic order and temporal order is not of
significance, when there are other grammatical means that may represent
the temporal order, as in English. Given the distinction between finite and
nonfinite clauses, English has more room for syntactical manoeuvre, as in

(44), than Chinese does, in contrast to (43).

(44) a. To buy three books, John will go to the bookstore.
b. (S > t1 > t2)

The morphological tense markers in English correctly represent the

temporal relation (44b) in the sentence in (443), while syntactic order of

91

constituents is of less importance.

Chinese has a few verbal aspect markers (Li & Thompson 1981,
Smith 1991). Of these aspect markers, le is considered to be a perfective
marker and guo as a perfective/experiential marker (Li & Thompson
1981:185, Smith 1991:344). I think that these two morphemes are related
to the notion of Linear Time as well. For example, both le and guo may

indicate that the event/activity in question takes place at a time earlier than
S, as in (45) and (46).8

(45) a. Zhangsan kan le zhe ben shu.
Zhangsan read Asp. this C. book
Zhangsan (has) read this book.

b (t > S)

(46) a. Zhangsan kan guo zhe ben shu.
Zhangsan read Asp. this C. book
Zhangsan read this book.

b (t > S)

 

8 Guo is also a verb and means 'to pass'. It is generally acknowledged that there are
two le's in Chinese. One of them is a verbal le (perfect aspect), and the other is a sentential

le . The latter is also Considered to be a manifestation of the perfect aspect (Li, Thompson
& Thompson, 1982). The sentential le, depending on the meaning of the matrix verb, has a
number of meanings, such as inchoative in (i) and present perfect in (iii), while the verbal
le generally represent past tense and the perfect, as in (ii) and (iv).

(i) Zhangsan kan shu le.
Zhangsan read book
Zhangsan began to read.
(ii) Zhangsan kan le shu.
Zhangsan read book.
Zhangsan read the book.
(iii)Zhangsan hui jia le.
Zhangsan return home
Zhangsan has returned home (He is home).
(iv) Zhangsan hui le jia.
Zhangsan return jia.
Zhangsan returned home/has been home (He is not home.)

92

(453) with le and (463) with guo appear to represent the same type of
temporal relations, as in (45b) and (46b) respectively, where the event
occurs before the utterance. However, [e and guo differ in a number of
ways in representing temporal relations. First le behaves more like a
perfect aspect marker in that it does not necessarily determine any

temporal relation relative to S, as in (47), in contrast to (48).

(47) a. Zhangsan kan le shu qu xuexiao.
Zhangsan read Asp. book go school
Zhang will go to school after he reads the book.
b. (S > t1 > t2)
(48) *Zhangsan kan guo shu qu xuexiao.
Zhangsan read Asp. book go school

In (473), le simply represents a temporal relation between a sequence of
two events so that the event kan le shu precedes the event qu xuexiao. (47a)
is taken to be anchored at NOW without any contrary context. On the other
hand, (48) does not have the temporal interpretation that (473) has. The
fact is that guo does not represent a temporal relation between a sequence
of two events. (48) may be grammatically rendered as (49) by means of a
lexical item, such as yiqian (before), which express a sequential temporal

relation.

(49) a. Zhangsan qu xuexiao yiqian kan guo shu.
Zhangsan go school before read Asp. book.
Zhangsan read the book before he went to school.

b. (2t > 1t > S)

(50) a. Zhangsan kan guo shu.
Zhangsan read Asp. book
Zhangsan read the book.

b. (t > S)

Second, guo is not a simple perfect marker in the sense that it does not

93

specify a time following S, as in (48), but systematically specifies a time
that precedes S, as in (49) or (50). Guo has a past tense function. Guo and
le represent temporal relations that differ in relation to S, as noted in Smith
(1991:348—53). Le may represent 3 continuity of an interval of time
extending from past to S, while guo involves a reference point that is

discontinuous with S.

(51) a. Zhangsan qu guo xuexiao.
Zhangsan go Asp. school
Zhangsan went to the school.
b. Zhangsan qu le xuexiao.
Zhangsan go Asp. school
(i) Zhangsan has gone to the school.
(ii) Zhangsan went to the school

(513) gives the interpretation that Zhangsan went to the school and he is no
longer at the school, while (51b) may have the interpretation that Zhang is
not back from school yet, though it may also have an interpretation similar
to that in (513).9 Discontinuity is characteristic of Linear Time as defined
in (9), while continuity belongs to the domain of Frame Time, which will

be discussed in the next section. Given the evidence that guo represents a

 

9 However, the difference between (513) and (51b) does not seem to exist or at
least is not obvious in the sentences below.

(i) Zhangsan kan guo zhe ben shu.
Zhangsan read Asp. this C. book
Zhangsan read/has read this book.

(ii) Zhangsan kan le zhe ben shu.
Zhangsan read Asp.this C. book
Zhangsan read/has read this book.

Both sentences in (i) and (ii) may have a present perfect interpretation and a simple past
interpretation, though read this book, like go to the school, is considered an
accomplishment verb and denotes a temporally bound situation. I think that the difference
in interpretations may originate from something other than temporal properties, but I do not
have an account now.

 

 

 

it?“

me
of

Ti

'4)

3F
C}

94

temporal relation that a time precedes S and is discontinuous with S, guo
functions more like a tense marker than an aspect marker. In addition, 1e
may function like a tense marker as well, since it also represents the order
of a sequence of situations and the order of a situation and S in Linear

Time.

3.3.2 TENSE AND TEMPORAL ADVERBIALS IN CHINESE

Chinese lacks regular tense morphemes so that temporal adverbials
appear to play a more important role in determining temporal relations in
Chinese. In this section, I examine the ways in which guo and 16 and
temporal adverbials interact in representing Linear Time.

These two morphemes, guo and le, may co-occur with a large range
of temporal adverbials, since they have a dual function for tense and aspect.
They may co-occur with a temporal adverbial that denotes a period of time

in which both S and tare anchored, as in (52).

(52) a. J intian Zhangsan kan le yi ben shu.
Today Zhangsan read Asp. one C. book
Today Zhangsan has read a book.
b. (t > S), where t, S E F , where E = temporal inclusion
and F = today
(53) a. J intian Zhangsan kan guo yi ben shu.
Today Zhangsan read Asp. one C. book
Today Zhangsan has read a book.
b. (t > S), where t, S E F, where E = temporal inclusion
and F = today

(523) and (53a) show that guo and [6 occur with a temporal adverbial that
is related to S. (523) and (53a) have an present perfect interpretation, since

there is a temporal continuity between tand S, as (52b) and (53b)

ind
ad\
lr’id

93

indicate.10 What is of interest here is that both guo and le occur with
adverbials that are not continuous with S, as in (54), if we assume that

today, this week, this month, etc. are continuous with S.

(54) a. Zuotian Zhangsan kan 1e yi ben shu.
Yesterday Zhangsan read Ts. one C. book
Yesterday Zhangsan read a book.

a' (t > S) '

b. Zuotian Zhangsan kan guo yi ben shu.
Yesterday Zhangsan read Ts. one C. book
Yesterday Zhangsan read a book.

b' (t > S)

The occurrences of [e in (54a) and guo in (54b) are like the English past
tense in the sense that the event/activity takes place in the past and is
complete, relative to S. In addition, the occurrence of temporal adverbials
discontinuous with S requires the occurrence of guo or le in some contexts,

as the sentences in (55) indicate.

(55) a. ?Zuotian Zhangsan kan yi ben shu.
Yesterday Zhangsan read one C. book
Yesterday Zhangsan read a book.

b. ?Qunian Zhangsan qu xuexiao.
Last year Zhangsan go school
Last year Zhangsan went to school.

The above evidence supports the argument that le and guo function like
tense markers. On the other hand, some sentences in Chinese may occur
with temporal adverbials discontinuous with S without [e or guo, as in (56)

on the next page.

 

10 Temporal continuity belongs to the domain of Frame Time, since Frame Time
concerns the duration of time and inclusions in time.

96

(56) a. Zuotian Zhangsan kan shu.
Yesterday Zhangsan read book
Yesterday Zhangsan read a book.

b. Qunian Zhangsan zai meiguo.
Last year Zhangsan be-in America
Last year Zhangsan was in America.

The sentences in (56) without le or guo are acceptable. These examples
suggest that the two tense morphemes le and guo together with temporal
adverbials represent temporal relations in Linear Time, and that the
grammaticalization of t > S in Chinese is not as complete as that in English.
With two tense morphemes, temporal adverbials in Chinese do not play as
important a role in the representation of t > S as they do in representing
S > t for the future.

Chinese does not have any future tense markers, though there are a
few modal verbs that may express the notion of future. One of these
modals is yao, which is like the English will in terms of function and
meaning. Yao expresses the notion of future and the willingness of the

agent, if there is an agent, as in (57a).

(57) a. Zhangsan yao qu xuexiao.
Zhangsan will go school.
(i) Zhangsan will go to the school.
(ii) Zhangsan wants to go to the school
b. J intian yao xia yu.
Today will fall rain
It will rain today

Hui is another modal that expresses possibility and also involves the notion

of future, as in (58).

97

(58) a. Zhangsan hui qu xuexiao.
Zhangsan may go school.
Zhangsan may go to the school.

b. Jintian hui xia yu.
Today may fall rain
It may rain today

These are modals rather than future tense markers, but yao is similar to the
English will in its function in representing relation in Linear Time. Yao
may be grammaticalized as a tense marker to some degree. The notion of
future can be expressed simply by temporal adverbials without these

modals, in contrast to the notion of past.

(59) a. Mingtian Zhangsan qu xuexiao.
Tomorrow Zhangsan go school
Tomorrow Zhangsan will go to the school.
b. Houtian xia yu.
The day after tomorrow fall rain
It will rain the day after tomorrow.

The temporal adverbials may determine the representation of Linear Time
with respect to future, but not as often in the case of past, in which case
both temporal adverbials and tense/aspect markers contribute to the
representation of Linear Time.

The above evidence shows that Chinese is not used in a tenseless way
nor is it as tenseless in the grammatical sense, as it is considered in the
literature. Chinese has some degree of grammaticalization of the temporal
relation t > S, but has no grammaticalization of the temporal relation t: S,
and probably some degree of grammaticalization of the temporal relation

S > t in Linear Time.

98

3. 4. FRAME TIME: THE SECOND DIMENSION OF TIME

Another aspect of the structure of time in natural languages is the
representation of the duration of time. Speakers impose frameworks on the
duration of time in terms of seconds, minutes, hours, days, weeks, months,
years, etc. in relation to events and activities. In the measurement of time
there are two distinctive temporal relations, one of which is temporal order
and the other of which is temporal inclusion. The former is the dimension
of Linear Time, and the latter is the dimension of Frame Time. Frame
Time is related to what are called 'frame adverbials' in Bach (1981),
Bennett and Partee (1978), Parsons (1990) and Smith (1991).

In classifying temporal adverbials, Bennett and Partee (1978: 22-30)
label a group of adverbials like this morning, yesterday, in I 973, etc. as
'frame adverbial phrases', as compared to other groups, such as 'durative
adverbial phrases' and 'adverbial phrases of number and frequency'. In
their definition (Bennett and Partee 1978: 22), 'Frame adverbial phrases
refer to an interval of time within which the described event is asserted to
have taken place'. They also include now, at noon, etc. in frame adverbial
phrases. However, they do not give their classification of temporal
adverbials any theoretical significance in their study of tense and aspect.
Smith (1991:155) has also mentioned frame adverbials in her treatment of
temporal adverbials. She gives frame adverbials the term 'locating
adverbials', which contribute to the specification of her Reference Time
and Situation Time. Smith considers the relation between the time specified
by locating adverbials and Situation Time vague. However, she does not
elaborate on the significance of frame or locating adverbials theoretically
either.

An in-depth treatment of frame adverbials is found in Parsons

99

(1990:209-23), who includes both 'temporal' and 'locative' adverbials, such
as in 1939 and in a dream, in this group. I just review his treatment of the
temporal frame adverbials here. What is a frame adverbial? Parsons
(1990:209) thinks that frame adverbials 'set a context within which the rest
of the sentence is to be interpreted'. On the other hand, Parsons (19902212)
notes that it is difficult to recognize and distinguish frame adverbials from
others, since these adverbials can function as frames (60a), as predicates of

time intervals (60b) and as predicates of eventualities (60c).

(60) a. From May to August Mary ran every day. (Frame)
b. From 2:00 to 3:00, Mary ran. (Pred. of interval)
c. Mary ran from 2:00 to 3:00. (Pred. of event)

(60') a. (x) |Day§x1 & Afterlx, May) & Beforegx, August) _.
(31)[I < now & On (I, x) & (3e)(3t)[t€l & Running (e) &

Agent-Theme (e, Mary) & Cul (e,t)]]]

b. (31)[I < now & From (2:QO,I) & To (3zm,!) &
(3e)(3t)[t El & Running (e) & Agent-Theme (e,Mary) &

Cul (e,t)]]
c. (31)[l < now & (3e)(3t)[t El & Running (e) & Agent-Theme

(e,Mary) & From (2:00,e) & To (3:00,e) & Cul (e,t)]]

The frame adverbial in (60a) constrains the noun phrase every day, as
shown in (60'a), while the predicate of the interval in (60b) constrains an
interval of time, as in (60'b). The predicate of event in (60c) constrains the
event, as in (60'c), which tells how long the event lasts. I think that
Parsons' distinction between frames and predicates of intervals is
semantically unnecessary, since both of these two temporally constrain
some period of time, whether they syntactically constrain a noun phrase
denoting time or not.

On the other hand, Bennett and Partee (1978:22) note that many of

these frame adverbials are indexical in character. They have a dual

100

function: they not only locate temporal orders, but also they locate
temporal inclusions. This dual function is exemplified by some temporal

adverbials that denote a period of time and express temporal order, as
shown in (61).

(61) a. Yesterday morning John read a book
b. John read a book at t, (t > S ).

c. John read a book within f, (f E F), where E : temporal
inclusion, f = morning and F = yesterday.

The properties and linguistic representations of temporal order have been
treated as Linear Time in sections 3.1, 3.2 and 3.3.. In this section, I just
look at the single dimension of Frame Time which deals with temporal
inclusion relations in natural languages.

It is apparent that my concept of Frame Time originates from the
above authors' classification and treatment of frame adverbials. However,
my Frame Time is different from the previous authors' frame adverbials in
that it is viewed as one of the three dimensions of the representation of
time in natural languages, instead of simply a subclass of adverbials. Frame

Time may be defined as (62).

(62) Frame Time is a set of intervals of time, which are denoted by
temporal frame phrases and within which a described event/
activity takes place or a state holds.

In contrast to Linear Time which is treated as a set of nondurative points,
Frame Time is a set of intervals of duration. In the measurement of the
duration of time, a relation exists among the units of measure that a smaller
unit of the duration of time f is temporally included in a larger unit of time

F, symbolically as f E F, where F symbolizes Frame Time and E stands

101

for temporal inclusion. Temporal inclusion is formally defined in (63) (cf.
Newton-Smith 1980:144).

(63) F includes f just in case f does not begin earlier than F
and f does not end after F.

Temporal inclusion in Frame Time may be graphically illustrated in Figure

4 below, where F stands for Frame Time, f stands for a frame of Frame

and { —\ stands for temporal inclusion.

 

 

 

Upper-
/ F \bound
( I V f \
{—r _\/_T ‘11—? —v—r —\/"f *xmwer-

bound

Figure 3: Temporal Inclusion in Frame Time

Within a frame of Frame Time, the largest interval of time denoted by a
temporal frame phrase may be said to be the upper-bound of Frame Time,
while the smallest possible subinterval of time is considered the lower-
bound of Frame Time. The bounds are context-relative. For example, in a
Frame Time expression like at 8 o'clock in the morning, at 8 o‘clock is
considered the lower bound and in the morning the upper bound. Each
interval of time denoted by a temporal expression is a frame of Frame
Time, and may temporally include another frame or be included in another
frame. What relations does the temporal inclusion in Frame Time show in

natural languages? One of the most important relations in Frame Time as

102

represented in the structure of time in natural languages is transitivity, as
defined in (64), where I = being temporally included, f: frame, and F =

Frame Time.

(64) I is transitive iff Vf], f2, ngF: I(f1, f2) A I(f2, f3)—->
1 (f1, f3)

Given that I is transitive, (64) says that if fl is temporally included in f2,
and f2 is temporally included in f3, then fl is temporally included in f3. The
transitivity property of the inclusion relation in Frame Time underpins the

entailment relationship in the sentences in (65).

(65) a. Mary walked to school at about 8 am yesterday morning.
b. Mary walked to school yesterday morning.
c. Mary walked to school yesterday.

Certain implications exist among the sentences in (65). (65a) implies (65b)
and (65c), and both (65a) and (65b) imply (65c). If (65a) is a true sentence,
then (65b) and (65c) are true sentences and so on. This entailment exists,
because 8 am is temporally included in yesterday morning, and yesterday
morning is temporally included in yesterday. I called this inferential
pattern 'increasing' in the sense that a sentence with a small frame of
Frame Time implies one with a larger frame of Frame Time. However, the
reverse is not necessarily true. For example, if (65b) is true, (65a) may not
be a true sentence, since it is possible that Mary walked to school at any
time in yesterday morning other than at about 8 am. This interesting aspect

of temporal inclusion is also seen in (66).

(66) a. John read a book Tuesday morning.
b. John read a book Tuesday.

103
c. John read a book this week.

The sentences in (66) exhibit the same properties as those in (65) do. If
(66a) is true, then (66b) and (66c) are true. If (66b) is a true sentence, then
(66c) is a true sentence too. However, converse implications do not exist in
Frame Time. For example, (66c) does not entail (66b) and (66a), and (66b)
does not entail (66a). In contrast to the entailment relations in Linear Time
which can be reversed, as discussed in section 3.1, the entailment relations
in Frame Time can not be reversed. I think that the difference in
inferential patterns between Linear Time and Frame Time lies in the
difference between linear ordering in Linear time and nonlinear ordering
in Frame Time, though I am not able to discuss the formal nature of linear
ordering and nonlinear ordering in this study. In the case of (66), if (66c)
is true, John may have read a any time during this week, not necessarily on
Tuesday or on Tuesday morning. The entailment relation in Frame Time

may be symbolically represented as (67).

(67) sf --> sF; sF -/-> sf, iff f E F, where sf = a sentence with a
smaller frame of Frame Time, SF 2 the same one with a larger
frame of Frame Time which contains the smaller frame of
Frame Time, -—> =implies, and -/-> = not imply

(67) characterizes an important relation that distinguishes Frame Time
from Linear Time, the latter of which exhibits entailment relations that can

be reversed.

3.4.1. FRAME TIME AND FRAME TIME EXPRESSIONS
Interval-referring temporal expressions may be categorized into two

Classes depending on their relationship with S, since some of these

104

expressions have a dual function in locating Linear Time and Frame Time,
as discussed above. The two classes are indexical expressions and non-
indexical expressions (Binnick 1991:307). The indexical expressions have
the dual function, while the non-indexical expressions do not have this dual

function. Indexical interval-referring temporal expressions are exemplified
in (68).11

(68) a. this morning, this afternoon, this evening, etc.
b. today, yesterday, tomorrow, the day before yesterday, etc.
c. this week, next week, this month, last month, etc.
d. this year, last year, the year before last year, etc.

These indexical interval-referring temporal expressions may form larger
units of expressions, such as since last Tuesday, before next month, etc..
These indexical interval—referring temporal expressions not only refer to
an interval of time but also locate the temporal order relations with S, as
their dual function. On the other hand, non-indexical interval-referring
temporal expressions, like those in (69), do not represent any temporal

relations with S.

(69) a. on May 15, on Christmas day, in a sunny afternoon, etc.
b. in March, in spring, in the fourth quarter, etc.
c. in 1993, in the 201h century, etc.

When expressions in (69) are used in sentences, Linear Time is usually

determined by tenses, as in (70).

 

. 11 I only give the English examples here for both indexical and non-indexical
Interval-referring temporal expressions, since their Chinese counterparts are similar. When
they are different, I will provide Chinese examples. I believe that natural languages
generally have these two classes of temporal expressions.

105

(70) a. John went to Washington D. C. on May 15.
a' (t > S)
b. John will go to Washington D. C. on May 15.
b' (S > t)

In (70a), May 15 precedes S, as represented by past tense, while in (70b)
May 15 follows 8, as represented by future tense. The expression May 15
itself does not represent any temporal order relations with S. In the
following, I refer to both indexical and non-indexical interval-referring
expressions as 'Frame Time expressions', since I am not interested in the
Linear Time function of these temporal expressions in this section.

There is an upper-bound and a lower-bound in Frame Time, as
illustrate in Figure 3 above. Frame Time bounds involve the definition of
Frame Time and the classification of Frame Time expressions. What
concerns me most is the lower bound of Frame Time, since larger intervals
of time may not be a problem for Frame Time. How small is an interval of
time that constitutes a frame of Frame Time? I plan to examine whether a
temporal expression is a frame of Frame Time or not with the definitions
provided in (62) and the properties in (64) and the relation in (67).

According to (64), Frame Time exhibits a temporal inclusion
relations, that is, a smaller frame is included in a larger frame. These
inclusion relations are transitive in the sense that a smaller frame is
temporally included in larger frames all the way upstairs towards the upper
bound, and this relation underpins the entailment relation of sentences with
Frame Time expressions, as stated in (67). Within such a frame denoted by
8 Frame Time expression an event or activity takes place, given the
definition of (62). If this kind of temporal relation exists, a temporal

expression represents Frame Time and is a frame of Frame Time. I will

106

look; at now and at 8 o'clock with respect to (62), (64) and (67). Now may
refer to point of time or an extended point of time (Binnick 1991:126 and

47 3 , Richards 1982:91). For example, the sentences in (71) involves the
extended now.

(71) a. John works at Michigan State University now.
b. Then, John was poor, but now he is rich.

The extended now in the sentences in (71) is not merely a point of time, but
a longer interval of time that covers some time past and some time to come
(B i nnick 1991 : 126). There is no question that the extended now can
furlCtion as a frame of Frame Time. I like to look into the now that refers
to an instant of time or to a short interval of time, as in (72), since I am

in te rested in the lower bound of a subinterval in a frame of Frame Time

(72) a. John is reading a book now (now: at this moment in this
morning).

b. John is reading a book this morning.
c. John is reading a book today.

(73) (f1 E f2 E F), where f1: now, f2: this morning, and F: today

Intuitively, we know that (72a) entails (72b) and (72c), and (72b) entails
(720). This intuition is supported by the formal representation of Frame
Time in (73), which shows that temporal inclusion relations exist in the
frame of Frame Time in the three sentences in (72). Now is a Frame Time
CXPression, though it is the lowest bound of a frame in Frame Time, as the
evidence indicates. Similar temporal relations are found in (74), where at 8
O'ClOck occurs. At 8 o'clock refers to an instant of time, which may be the

smallest subinterval or the lowest bound of a frame of Frame Time in any

context, if it logically behaves like a frame of Frame Time.

107

(74) a. John arrived at 8 o'clock (in the morning on that day) .
b. John arrived in the morning (on that day).
c. John arrived on that day.

(75) (f 1 E f2 E F), where f1: at 8 o'clock, f2: in that morning,
and F: on that day

(75) indicates that temporal inclusion relations formally exist among the
three frames of Frame Time, at 8 o'clock, in the morning and on that day.
This conforms with our intuition that (74a) entails (74b) and (74c), while
converse entailment does not hold. In this sense, at 8 o'clock is the lower
bound of a frame of Frame Time, which may be the smallest frame or the
lowest bound of Frame Time.

I have argued in the above that points of time are temporally
included in a frame of Frame Time, though points do not temporally
include any frames of Frame Time. In this sense, points of time may be
considered the lowest bound of Frame Time. A point in Frame Time is
different from a point in Linear Time in that the former is treated in terms
of temporal inclusion, while the latter is treated in terms of temporal
precedence. Frame Time and Linear Time represent two different

dimensions of the representation of time in natural languages.

3.4.2. SYNT ACT IC FUNCTIONS OF FRAME TIME EXPRESSIONS

In the previous section, I have examined a number of temporal
expressions and illustrated their semantic properties in representing Frame
Time, but I have not had a chance yet to discuss the relation between their
syntactic functions and their semantic representation of Frame Time. I
focus on this question in the present section.

Temporal expressions as adverbials are generally classified into

sentential adverbials and verb—phrase adverbials according to their

108

syntactical functions (Dowty 1979:323-36), as in (76) respectively.

(76) a. Yesterday John read a book.
b. John read a book from 2:00 to 3:00.

In (76), it seems that in English the temporal adverbials that occur in
sentence-initial positions are sentential and those that occur in verb phrase
positions are verb-phrase adverbials. However, mere syntactical positions
may not provide a complete picture, as shown in (76'), where yesterday is
moved to the non-sentence-initial position, while from 2:00 to 3:00 is

moved to the sentence-initial position.

(76') a. John read a book yesterday.
b. From 2:00 to 3:00, John read a book.

Is yesterday still a sentential adverbial? Is fi'om 2:00 to 3:00 a sentential
adverbial? In his treatment of adverbials as frames, predicates of intervals
and predicates of events, Parsons (1990:212) seems to take a semantic
approach to a syntactical problem. Adverbials that semantically function as
frames and predicates of intervals syntactically appear to be sentential
adverbials, as in (76a) and (76').12 Yesterday is a frame both in (763) and
(76'a), regardless of its positions. From 2:00 to 3:00, however, is a
predicate of an event in (76b) and a predicate of an interval in (76'b). This
suggests that some temporal adverbials like from 2:00 to 3.00 are position-
sensitive, while others like yesterday is not. However, Parsons concedes
that he is not aware of any operational tests that distinguish them.

Therefore, what is a sentential temporal adverbial and what is a verb-

 

12 I treat temporal adverbials that function as frames and as predicates of intervals
both as Frame Time adverbials, since my definition of Frame Time (62) holds for both of
Parsons' types of adverbials.

109

phrase adverbial remain an open question for now. His problem, I think,
lies in a semantic approach to a syntactical problem.

In Dowty's categorization, sentential temporal adverbials belong to
the category (t/t), forming sentences from sentences, while verb-phrase
temporal adverbials belong to the category (IV/IV), forming verb phrases
from verb phrases. Such categorizations appear to be supported by
syntactic tests like do-so pronominalization (cf. Binnick 1991:303). It is
said that only syntactic items that belong to one constituent may be
pronominalized in do-so pronominalization. This test suggests that the verb
and the verb-phrase temporal adverbial may form one constituent, as in

(78), compared with the sentences in (77).

(77) a. John ran for an hour and Bill did so for half an hour.
b. John ran for an hour and Bill for half an hour.
c. John ran for an hour and *Bill did so half an hour.
d. John ran for an hour and ?*Bill half an hour

(78) a. John ran for an hour and Bill did so too.
b. John ran for an hour and Bill did too.
c. John ran for an hour and ? Bill too.

(78) shows that ran and for an hour may constitute one constituent, though
they do not have to be treated as one, as in (77). Thus, temporal adverbials
like for an hour are supposed to be verb-phrase temporal adverbials.
However, it is already noticed that sentential temporal adverbials such as
yesterday may occur in verb-phrase position while verb-phrase temporal

adverbials may occur in sentence-initial position (Binnick 1991:303-4), as
in (76')

(76') a. John read a book yesterday.
b. From 2:00 to 3:00, John read a book.

l 10
The sentences in (76') are different from those in (76) in the positions of
temporal adverbials. Do we say that yesterday becomes a verb-phrase
temporal adverbial and for an hour becomes a sentential temporal
adverbial? Yesterday seems to become a verb-phrase temporal adverbial, if

do-so pronominalization test really reveals what it is supposed to, as

indicated in (79).

(79) a. John read a book yesterday, and Bill did so too.
b. John read a book yesterday, and ?Bill too.

In (79), read a book and yesterday appear to form one constituent, when
they undergo do—so pronominalization. I think that the do-so pronominal-
ization test may not really tell the difference between sentential temporal
adverbials and verb-phrase temporal adverbials, as it is expected.

I believe that the actual positions in a sentence may not matter much
in English with regard to the categorizations, as long as a temporal
adverbial is syntactically directly under an 8 node in terms of generative
grammar. Here, I propose a frequency-adverbial test for the distinction
between sentential temporal adverbials and verb-phrase temporal
adverbials. Frequency adverbials as sentential adverbials are unproblematic
in syntax. Frequency adverbials are treated as sentential operators in
representing temporal relations in semantics (Aqvist, Hoepelman and
Rohrer 1980, Richards 1982). Semantically, a temporal adverbial that has a
wider scope than a frequency adverbial should be considered sentential.
Syntactically, a temporal adverbial that is ordered outside the scope of a
frequency adverbial should considered sentential. First, I show what

temporal adverbials may appear within the scope of a frequency adverbial,
as in (80) and (81).

111

(80) a. ?John ran yesterday twice.
b. John ran for an hour twice.

(81) a. ?John read a book yesterday twice.
b. ?John read a book in an hour twice.

The sentences in (80) and (81) indicate that only the for-temporal adverbial
that functions as a predicate of an event may naturally occur with the scope
of a frequency adverbial, as in (80b), while other temporal adverbials may
not occur naturally within the scope of a frequency adverbial. Second, I

show what temporal adverbials may occur outside the scope of a frequency
adverbial, as in (82) and (83).

(82) a. John ran twice yesterday .

b. John ran twice for an hour .
(83) a. John read a book twice yesterday .
b. John read a book twice in an hour.

The sentences in (82) and (83) suggest that for-, in- and other temporal
adverbials may all occur outside the scope of a frequency adverbial
semantically as Frame Time adverbials. The test indicates that all those
temporal adverbials may appear in the sentence-final position syntactically
as sentential adverbials outside the scope of a frequency adverbial in
English. However, a for—temporal adverbial is ambiguous when it occurs in
the sentence-final position and there is no constituent between it and the
verb phrase. In such a case, a for-temporal adverbial may function as a
predicate of an event, if it is parsed as part of the verb phrase or function
as a frame of Frame Time, if it is parsed as a constituent directly under the
S node.

The above analysis provides a viable account for some observations

concerning the interpretation of for-temporal adverbials. For example,

112

Dowty (1979:250-60) notes that the sentence in (84) has three interpreta-
tions, namely, a durative reading, an internal reading and an iterative

reading, as in illustrated (84').

(84) The Sheriff of Nottingham jailed Robin Hood for four years.
(84') a. The Sheriff of Nottingham spent four years bringing it about
that Robin Hood was in jail.
b. The Sheriff of Nottingham brought it about that Robin Hood
was in jail for four years.
c. For four years the Sheriff of Nottingham brought it about
several times that Robin Hood was in jail.

The durative reading (84'a) and internal reading (84'b) are irrelevant here,
since they involve the different interpretations of the verb jail, where the
temporal adverbial is a verb-phrase temporal adverbial. When a for-
temporal adverbial functions syntactically as verb-phrase temporal
adverbial, the for-temporal expression represents temporal relations within
Situation Time, involving the temporal duration of an event or activity in
question, as in (84'a) and (84'b), and involving perfective and imperfective
interpretations.l3 That is what Parsons (1990) calls 'predicate of event'. In
(84'c), the temporal adverbial for four years functions syntactically as a
sentential adverbial in the sentence and semantically as a frame of Frame
Time so that it allows a 'reduced' frequency adverbial with the scope of the
for-temporal adverbial. This is how the iterative reading arises in (84'c).
Now I look at for-expressions as sentential in relation to the formal

definition of Frame Time in (62) and formal properties of Frame Time in

E

. l3For-temporal expressions as verb-phrase temporal adverbials representing
§ltmfion Time are seen in the following sentences, where perfective readings (i) and
lmperfective readings (ii) are concerned. This topic will be fully treated in Chapter Four.

(i) John read a book. (John finished reading it.)
(ii) John read a book for an hour. (John did not finish reading it.)

113

(64) and (67). Does afar-temporal expression exhibit the properties of
Frame Time defined in (64) and (67) when it functions as a sentential
temporal adverbial and as a frame of Frame Time? I answer this question

with an observation of the entailment relations in the sentences in (85).

(85) a. John walked 3t 8 o'clock on Tuesday morning.
b. John walked at 8:30 on Tuesday morning.
c. John walked twice for an hour on Tuesday morning, .
d. John walked Tuesday morning.

(853), (85b) and (850) all entail (85d). (853) and (85b) together imply
(85c). This kind of entailment relation exists, because the temporal relations

in (85) exhibit the properties of Frame Time, as illustrated in (86).

(86) 3. (f1 E f2 E F), where f1 = 8 o'clock, f2 2 an hour,
F = Tuesday morning.
b. (f1 E f2 E F), where f1: 8:30, f2: an hour,
F: Tuesday morning.
c. (f1 E f3 E F & f2 E f3 E F), where f]: 8 o'clock, f2: 8:30,
f3: an hour, F=Tuesday morning.
d. (f1 E F), where f1 2 some time, F: Tuesday morning.

(86) characterizes the temporal inclusion relations in the sentences in (85),
Where smaller intervals of time are temporally included in 3 frame of
Frame Time. The entailment relations that build on the temporal inclusion
relations can not be reversed, as stated in (67). Thus, it is predictable that
reverse entailment relations do not exist in the sentences in (85).

I have shown that in English Frame Time expressions function
syntactically as sentential temporal adverbials, while sentential temporal
adverbials function temporally as frames of Frame Time. This point is best

illustrated with the for-temporal expression which usually functions

114

syntactically as verb-phrase temporal adverbials in representing Situation
Time but which may function syntactically as sentential temporal adverbial
in representing frames of Frame Time.

In Chinese, sentential temporal adverbials occur in preverb or
sentence-initial positions only, while verb-phrase adverbials show up in
post-verb positions only.l4 Syntactically sentential temporal adverbials may
appear in the immediately preverb position, as in (873), or in the sentence-

initial position, as in (87b).

if!
(87) a. Zhangsan yi xiaoshi kan le yi ben shu.‘
Zhangsan one hour read Asp. a C. book
Zhangsan read a book in an hour. it
b. Yi xiaoshi Zhangsan kan le (yi ben shu..>
One hour Zhangsan read Asp. one C. book
Zhangsan read 3 book in an hour.

7". Ii ,2; / ii,
It is straightforward in Chinese that sentential temporal adverbials
uniformly occur in pre-verb positions, and only those kinds of Frame Time
expressions that I list in (68) and (69) may grammatically occur in pre-

verb positions.

 

14 Verb—phrase temporal adverbials appear in two forms: as object modifiers and as
unaccusative object , as in (i) and (ii).

(i) Zhangsan kan le yi xiaoshi de shu.
Zhangsan read Asp. one hour M(odifier) M(arker) book.
Zhangsan read one hour's book (Zhangsan read a book for an hour).
3: rt ‘5 .j ,2.“ 1.9.1.9
(ii) Zhangsan kan shu kan le yi xiaoshi.
Zhangsan read book read Asp. one hour
Zhangsan read the book for an hour.

Their syntactic and temporal functions are more straight forward, as I will discuss in detail
in section 4.3.

l 15
3.5. SUMMARY OF CHAPTER THREE

In this chapter, I have examined two dimensions of the temporal
structure in natural languages, namely, Linear Time and Frame Time.
Linear Time represents temporal orders or locations along a temporal line,
focusing on precedence relations. Frame Time represents temporal
inclusion relations, though it has a temporal order aspect. Each dimension
of these temporal relations has its formal properties. Temporal order
relations in Linear Time are transitive and have converses, as formally
defined in (10) and (13). Temporal inclusion relations in Frame Time are
transitive but the entailment relations with temporal inclusion underpinning
can not be reversed, as formally defined in (64) and (67). In other words,
the difference in the formal properties between Linear Time and Frame
Time leads to completely different inferential patterns in these two
dimensions of temporal relations in natural languages.

In English, Linear Time is linguistically represented by tenses, the
perfect and temporal adverbials, only the first forms are grammaticalized
representations of Linear Time. In Chinese, Linear Time is linguistically
represented by temporal adverbials and two 'aspect' markers, where the
latter forms are grammaticalized representations of Linear Time, while the
former forms are not. Both may function as tense markers, the latter of
which appear to behave more like tense markers than aspect markers,
regardless of the claim that Chinese is a tenseless language. Chinese
sentences are definitely not used in a tenseless way, though some of them
do not have any degree of grammaticalization of temporal relations in
Linear Time. Approaches that use tenseless sentences in capturing their
truth conditions are challenged (Bennett and Partee 1978). The evidence

that Chinese is not used in a tenseless way may support this challenge, since

116

the use of tenseless sentences in philosophical and linguistic study appears
to be justified by the belief that some natural languages like Chinese are
indeed used in a tenseless manner.

Frame Time is linguistically represented by Frame Time expressions
which are interval-referring temporal expressions, as listed in (68) and
(69). Frames of Frame Time have both upper bound and lower bound.
Their upper bound is the largest frame of Frame Time that is denoted by 3
Frame Time expression, while its lower bound can be as small as point of
time. Syntactically, the upper bound and the lower bound exhibit a
modifying-and-modified relationship, as in at 8 o'clock in the morning,
where in the morning modifies 8 o'clock. In terms of syntax, Frame Time
expressions function as sentential temporal adverbials in both English and
Chinese, though the syntactic function of afar-temporal phrase is
ambiguous in English. Frame Time expressions as sentential temporal
adverbials exhibit certain scope properties, which has implications for
approaches that treat temporal adverbials and tenses as sentential operators,

and that treat Frame Time as Situation in interval semantics.

Chapter Four
SITUATION TIME

4.0 INTRODUCTION

Philosophical studies of actions have generally ignored the temporal
properties of action, as I have discussed in sections 1.1 and 1.2.. For
example, Aristotle studied actions in terms of process, change and goal in
relation to verb categories, without any explicit reference to the notion of
time. In modern analyses of tense and aspect, for instance, Reichenbach
(1947:287-98) assumes, without examination of the temporal properties of
actions, that events play a role in his famous tense theory that consists of
speech time, event time and reference time. Thus, his tense theory has
unwanted consequences, as I have discussed in section 3.1. In section 2.1.1,
I have pointed out that it is Vendler (1957, 1967) who first made explicit
reference to the notion of time with supporting linguistic evidence in his
study of verbs. In this chapter, I will follow Vendler's tradition in studying
the temporal properties of verbs by examining linguistic evidence
regarding those properties and with explicit reference to the notion of
time. On the other hand, I will study the role of events, or more generally,
of situations in temporal representation in language, though I will not make
assumptions as Reichenbach (1947) did. I will not simply assume the
existence of time involved in linguistic representation of actions, activities,
processes and states, but support its existence with linguistic evidence.

Following Comrie (1976) and Smith (1991), I will use 'situation' as a
117

118

general term for 'event', 'activity', 'process' and 'state' that are denoted by
verbs or verb phrases, since I do not limit my study to action verbs. With
foe us on the notion of time, I consider a situation as an instantiation of
temporal properties, following Gabbay & Moravcsik (1980). Thus, I will
use 'Situation Time' in making reference to the temporal properties of a
s i tuation.
In philosophical and logical studies, there are three ways to treat the
assignment of denotations of verbs and nouns (Gabbay & Moravcsik
1 98,0:60): i) verbs denote situations and nouns denote objects with a
pro blem of how to link the two domains; ii) situations that are denoted by
Verbs are the basic domain and objects that are denoted by nouns function
as modifications of situations; and iii) objects that are denoted by nouns are
the basic domain and verbs denote properties of objects. I will study verbal
de notations in the spirit of the second approach: situations denoted by verbs
are the basic domain and objects denoted by nouns function as modifiers,
t1101.1 gh I will not work on a formal apparatus in the representation of this
rel ation. As a result, I will study verbal reference to situations, how
Situations are bound or nonbound in time and what objects temporally bind
a Situation and what objects do not temporally bind a situation. Further, I
assume that situation boundaries are Situation Time boundaries, since
situations are instantiations of temporal properties. I will examine the
relationship between temporal properties of situations and verb categories.
I believe that verb categories depend on the temporal structures of the
Situations involved, if verbs refer to situations. By doing this, I assume that
the questions in (23), (2b) and (2c) are answered.
In this study, the referential functions of verbs and verb phrases are

cOnsidered inherent properties, while the quantification interpretation

119

involved is considered a noninherent property of verbs and verb phrases. I
be] ieve that this is important in considerations of temporal properties of
situations.

Finally, I will define Situation Time and show how it is logically
different from the other two dimensions of the representation of time:
Frame Time and Linear Time. I will demonstrate that inferential relations

between sentences should be considered in terms of part-and-whole
re 1 ations.

4- 1 SITUATION TIME: THE THIRD DIMENSION OF TIME

Intuitively we know that a verb or verb phrase refers to an action,
ac“livity, process or state and that an action, activity, or process takes 3
Ce rtain period of time and that a state holds for a certain period of time.
Re ichenbach (1947) refers to this time as the 'point of the event', which he
assumes to be nondurative in his tense theory, but seems to treat it as
duI‘ative with respect to aspect. Parsons (1989, 1990) refers to this time as
'event', while Smith (1991) refers to it as 'situation'. In this chapter, I
Consider this time 'Situation Time', as an instantiation of temporal
PFOperties. Instead of following Reichenbach's assumptions, I am concerned
With the existence of situations and the temporal nature of situations, as
represented in linguistic forms, in this section. First, I will try to answer
the question 'Do verbs or verb phrases refer to situations in the same way
that nouns or noun phrases refer to objects ?'. Second, I will try to explore

the temporal nature of situations as whether they are instantaneous,
durative, bound or nonbound.

120
4- l - l NOMINALIZATION AND REFERENCE TO SITUATIONS

Verbs and verb phrases refer to situations in the same way that nouns
and noun phrase do to objects. This view is explicit or implicit in some
studies (cf. Davidson, 1968, 1980, Parsons 1985, 1989, 1990, Smith 1991).
Ev idence for verbal reference to situations is more clearly seen in the

nominal use of verbal forms, as in (1), where verbal forms have similar

referring functions as nouns do (cf. Parsons 1990).

(1) 3. Swimming is an interesting sport.
b. To walk alone is dangerous.

In English, some verbs have morphological nominal forms as in (2) below,

whereas in Chinese there is no morphological change between a verb as a

Verbal predicate and a verb as a nominal term.

(2) a. to destroy --> destruction
b. to move --> movement

The relationship between nouns and verbs has been a major question in
Syrltactic studies at least since Chomsky's work in nominalization (1970),
but little has been agreed upon about the similarities and differences with
res pect to the syntactic properties such as the argument structure and theta
theory, probably except for case theory (cf. Grimshaw 1990:45-6). I will
di Seuss only the referential properties of nominalization here. In nominal
reference to situations, verbal forms in nominal use share a number of
features with nouns in quantification interpretation. First, bare verbal

f0l‘ms in nominal use may have a generic interpretation as bare nouns do
(Cf. Carlson 1977), as shown in (3) on the next page.

121

(3) a. Skiing is popular.
b. Dogs are popular.

The nominal skiing refers to situations of skiing in general in (3a), and the
noun phrase dogs refer to dogs in general too in (3b). Secondly, verbal
forms in nominal us have an indefinite interpretation as nouns do, as in (4).

(4) a. Skiing was interesting last summer.
b. A dog was barking last night.

In (4), the nominal skiing does not refer to a particular situation of skiing,
nor does the noun phrase a dog refer to a particular dog. Thirdly, verbal

forms in nominal use have definite interpretations as nouns do, as in (5).

(5) 3. Swimming yesterday with my girlfiiend was fun.
b. The dog with spots was in the yard.

In (5), the nominal swimming yesterday with my girl friend refers to a
PafliCUIar situation of swimming, and the noun phrase the dog with spots
refers to a particular dog.

In Chinese, nominal reference of verbs to situations is more straight
forward. Nominalization adds nouns of behavior, activity, action, state or

thing to a sentence that is nominalized, as in (6)-

(6) 3. Zhang sheng bing de qingkuan

John be ill Part(icle). situation/state
(the state of) John's being ill

b. Zhang youyong de shi
John swim Part. thing
(the activity of or thing of) John's swimming

c. Diren huimie zhege chengzheng de xingwei
enemy destroy this city Part. behavior
(the behavior of) the enemy's destruction of the city

122

The italicized nouns in (6) directly refer to the situations that those
“om i nalized sentences refer to. The noun shi (thing/state) is most
cotfur-rionly used, though others may be used depending on the situation.
On the other hand, verbal forms in nominal use share a number of
features with verbs in sentences in terms of modification. The exact
5y ntactic nature of nominalization of verbs may still remain an open
question (cf. Newmeyer 1986:107-110), but evidence that both nominals
and verbs refer to situations appears to be very strong. Regardless of the
controversy on the syntactic nature of nominalization, event structures are
believed to underlie the relationship between nominals and verbs (cf.

Grimshaw 1990, Parsons 1990:72). Thus, the structure of a nominal and

that of a verb in a sentence are supposed to represent the structure of an
event or situation. For example, verbal forms in nominal use share with
verbs or verb phrases a number of features with respect to modification.
First, verbs in both nominal use and predicative use may take nouns as

modifiers- For example, they both take nouns as Agent, as in (7).

(7)

a. John's swimming
b. John swam.

John is the Agent both in (7a), where swimming is a nominal, and in (7b),

where SWim is a predicate of 3 sentence. Second, nominals and verbs both
take noun phrases as Patient or Theme, as in (8)-

(8) a. the enemy's destruction of the city
b. The enemy destroyed the city.

The noun phrase the city is considered to be the Theme in both (83) and
(8b)- Third, nominals and verbs both take adverbials of time, as in (9).

123

(9) a. John's swimming yesterday
b. John swam yesterday.

Yesterday is the temporal adverbial in the nominal phrase in (9a) and in the

sentence in (9b). Fourth, both nominals and verbs may take other phrases as
modifiers, as in (10) and (11).

( 10) a. John's swimming across the river
b. John swam across the river.

(11) a. John's swimming across the river with a life-preserver
b. John swam across the river with a life-preserver.

In ( l O), the prepositional phrase across the river functions as the adverbial
of Path in the nominal phrase and in the sentence as well. On the other
hand, the prepositional phrase with a life-preserver is the adverbial of
Instrument in the nominal in (l 13) and in the sentence in (11b).

In short, I have discussed two pieces of evidence that verbs refer to
situations in the same way that nouns do to objects. First, verbal forms in
nominal use share with nouns some common features in quantification
interpretations with respect to reference. Second, verbal forms as nominals
share some common features with verbs as predicates with respect to

mOdifiCation because of the structure of situations that they both refer to.

4.12 SITUATIONS AND THEIR TEMPORAL PROPERTIES
The time of the event or situation is treated as nondurative and is not

CORSidered to show a lot of temporal properties in Reichenbach's (1947)
tense theory. On the other hand, situations are considered instantiations of
teImPOI‘aI properties in Gabbay and Moravcsik (1980). Linguistic evidence
indicates that situations exhibit a range of temporal properties. Situations

that are denoted by verbs may be instantaneous or durative, the latter of

124

whiCh may be with or without gaps, and may be bound (telic) or nonbound
(atelic) (cf. Binnick 1991:189-97, Dahl 1981, Smith 1991). I will examine

relevant linguistic evidence regarding situations that exhibit those temporal
properties.

First, some verbs denote instantaneous situations. This category of
verbs covers what Vendler (1957, 1967) calls 'achievement' verbs. I will
refer to this category of verbs as 'instantaneous' verbs, since I am
concerned with the temporal properties of verbs, but not concerned with
theories of actions. The temporal properties of the situation denoted by an
instantaneous verb is seen in the relationship between that verb and some
Frame Time adverbials (cf. Vendler 1957, 1967). In section 3.3, I have
discussed the upper and lower bounds of Frame Time, and shown with
linguistic evidence that the lower bound can be an instant of time as
represented by Frame Time adverbials such as at six o'clock in English or
liudian (six o'clock) in Chinese. A verb that denotes an instantaneous
situation can occur with 3 Frame Time adverbial that denotes an instant of

time or an interval of time, as shown in (12), which are in the non-

ngreSSive form with a reportive reading.

( 1 2) a. John noticed the cat at six o'clock.
b. John noticed the cat yesterday.

I“ (123), the verb notice may cooccur with the Frame Time adverbial at six
O'CIOCk which denotes an instant of time. In (12b), at the same time, it may
3‘30 cOoccur with the Frame Time adverbial yesterday, which denotes an
interval of time, since an instant of time may be temporally included in an
interval of time. Instantaneous verbs may cooccur with temporal adverbials

33 1Orig as they represent Frame Time. Dowty (1979:58) notes that notice

125

may not cooccur with a temporal adverbial with for, as in (13a), though he
does not elaborate on an account for it.

(13) a. ??John noticed the cat for a few minutes.
b. John noticed the cat twice for a few minutes.

However, (13b) is acceptable with the repetition interpretation marked by
the frequency adverbial twice (cf. Dowty 1972, Carlson 1979:421-2). A
temporal adverbial phrase with for has two functions: it may represent
Frame Time and it may be predicated of Situation Time, as I have argued
in section 3.4.2. The verb notice denotes an instantaneous situation, while
the adverbial phrase for a few minutes denotes an interval of time. The
sentence is unacceptable if for a few minutes functions as what is called
'event predicate' (cf. Parsons 1990:212).l The sentence is acceptable if for
afew minutes function as a Frame Time adverbial, as in (13b), where the
function of the for-phrase is disambiguated by inserting a frequency or
count adverb like twice between the verb phrase and the for-phrase for
reasons that I have raised in section 3.4.2. The above evidence indicates
that verbs that denote instantaneous situations take Frame Time adverbials
but not Situation—predicate temporal adverbials. This distinction is a better
criterion than the traditional progressive form test (cf. Dowty 1979, Kenny

1963, Vendler 1967), since the progressive form may be used with some
insultltElneous verbs, as in (14).

(14) a. John is arriving at the station at six o'clock.
b. John is coming at six o'clock.

\

1 I may refer to an event predicate as a situation predicate, and regard it as
Tepmseming Situation Time, as I argue in section 4.3 below.

126

Those verbs in the progressive in (14) denote situations that are temporally
different from the ones normally denoted by instantaneous verbs. The
progressive forms in (14) may represent future tense, and may refer to
situations that form a chain of situations, in which case the progressive
forms refer to non-resultative part of the chain, whereas the non-
progressive form refers to the resultative part of the chain (cf. Binnick
199 l , Dowty 1979, McGilvray 1991:252-4, Parsons 1990, Smith 1991:33-
6). Therefore, this usage is limited to verbs that may denote a chain of

situations, and is not acceptable with verbs that do not refer to a chain of
situations, as in (15)

( l 5) a. ??John is spotting a bird at six o'clock.
b. ??John is discovering a bird at six o'clock.

(15) are not acceptable as sentences in future tense nor as sentences in the
progressive aspect, since those verbs are instantaneous verbs and do not
refer to the initial part of a chain of situations. It is clear that whether a
verb Can occur in the progressive form is not determined by temporal
pr0perties only of that verb, but rather by other properties in addition, at
least in English. Thus, progressive form is not a reliable test of the
temPOI‘al properties of a verb.

Second, some verbs denote situations that are durative. These verbs
include Vendler's activity verbs, accomplishment verbs and state verbs. I
W1“ refer to these verbs as 'durative' verbs for the same reason as stated
above. Traditionally, a verb that denotes a durative situation is supposed to
be able to occur in the progressive form, but it is also noted that state verbs
may not occur in the progressive form, as in (16), though they denote

durative situations (cf. Dowty 1979:55-60, Kenny 1963, McGilvray

127
199 1 :245-55, Vendler 1967).

(16) a. ??John is knowing the answer.

b. ??John is understanding me.
(17) a. John is a fool.

b. John is being a fool.
(18) a. The desk is square.

b. ?‘.Vl‘he desk is being square.

Of course, there is the idiosyncrasy in the usage of verbs. For example, the

verb phrase be a fool in (17a) refers to a state, but it may also appear in the

progressive form, while the verb phrase be square in (18a) also refers to a
state but can not occur with the progressive form in (18b) (cf. McGilvray
1991:255). The use of progressive forms involves other aspect of theories
of actions, such as agency and dynamics, which are not temporal properties.

Th us, I think that the relationship between verbs and temporal

adverbials is a better indicator of whether a verb denotes a durative
situation or not. For example, in contrast to instantaneous verbs, durative
VCrbS in non-progressive forms do not cooccur with 3 Frame Time

adverbi al denoting an instant of time, since such verbs refer to situations

that are durative, as in (19) and (20).

( 1 9) a. John ran yesterday.
b. ?John ran at six o'clock.
(20) a. John read a book yesterday.
b. ?John read a book at six o'clock.

(19b) and (20b) are unacceptable, if a completive reading is not available.
In English, this seems to be unlikely with most of the durative verbs. In
some Mandarin dialects, a completive reading may be forced upon durative

verbs in such sentences, though there is a completive marker, as in (21).

128
(21) a. Zhang liu dian kan wan le yi ben shu.

John six o'clock read Asp. Tns. one M book
John finished reading a book at six o'clock.
b. ?Zhang liu dian kan 1e yi ben shu.

John six o'clock read Asp./Tns. one M book
John finished reading a book at six o'clock.

(2 1 a) has a completive marker won and a completive reading, whereas

(2 1 b) does not have a completive marker but still has a completive reading,
which is acceptable to some speakers. Thus, a Frame Time adverbial
denoting an instant of time may force a completive reading in Chinese. In
addition, durative verbs may cooccur with temporal adverbials functioning

as situation predicates, as in (22a), but instantaneous verbs may not cooccur

with temporal adverbials functioning as situation predicate, as in (22b).

( 22) a. John ran for half an hour.
b. ??John arrived for half an hour.

Situation predicates denote a period of time. Thus, situation predicates can
only mOdify a verb denoting a durative situation.

Third, some verbs denote situations that are considered to have gaps,
and, 0n the other hand, other verbs refer to situations that are believed not
to have gaps (Binnick 1991:180-2, Gabbay & Moravcsik 1980, Mourelatos

1981)- For example, (233) is considered to denote a situation with gaps,
while (23b) refers to a situation without gaps.

(23) a. John read a book from Monday to Wednesday.
b. John was sick from Monday to Wednesday.

it is said that John could not be reading every minute within those three

days, Whereas John might be sick every minute within those three days.

129
Thus , (23a) does not entail that John was reading at each and every
subi nterval of those three days, but (23b) does entail that John was sick at
each and every subinterval of those three days (cf. Binnick 1991:181-8,
Gabbay & Moravcsik 1980). I want to argue that such an analysis is based
more on extralinguistic evidence than on linguistic evidence. Such an
analysis can not guarantee that (23a) is not true at each and every
subi nterval of those three days nor does it guarantee that John was sick at
each and every subinterval of those three days. For example, we may have a
scenario in which John is one of those Buddhist monks studying Buddhist
scripts without sleeping and eating for quite a long time. If (23a) refers to
this situation, then it could be true that John was reading at each and every
subinterval of those three days. Similarly, we may have a scenario in which
John had some stomach trouble on Monday and recovered, had a headache
on Tuesday and recovered, and had something else on Wednesday and
recovered. If (23b) refers to this situation, it is not necessarily the case that
10h“ Was sick at each and every subinterval of those three days, though it is
true that John was sick from Monday to Wednesday. To provide a viable
account of these differences, I think that we need an analysis based on the
diStin(ition between Frame Time adverbials and situation predicates. I
informally represent the distinction in terms of syntactic positions in (24),
With the capital letters as a metalanguage to represent the difference, where

Frame Time adverbials appear in the sentence-initial position while a

situation predicate occurs at the verb phrase position, .

(24) a. FROM MONDAY TO WEDNESDAY, John read a book.
a' John read a book FROM MONDAY TO WEDNESDAY.
b. FROM MONDAY TO WEDNESDAY, John was sick.
b' John was sick FROM MONDAY TO WEDNESDAY.

130

Linguistically speaking, (23a) has two interpretations as in (24a) and (24a').
When the temporal adverbial in (24a) functions as a Frame Time adverbial,
(24a) allows the interpretation of a series of (sub)situations. Therefore, it is
not true that John was reading a book at each and every subinterval of those
three days, according to (24a). (24a) denotes a situation with the so called
gaps, in which we find a layman John. However, the temporal adverbial in
(24a') functions as a situation predicate and is a representation of Situation
Time. Thus, (24a') allows the interpretation of a continuous situation. It is
true that John was reading a book at each and every subinterval of those
three days, according to (243'). (24a') refers to a situation without gaps, in
which we find a Buddhist John. (23b) has two interpretations too, as in
(24b) and (24b'). The analysis of (23a) applies to (23b). The temporal
adverbial functions as Frame Time in (24b), and an interpretation of a
series of situations is available. In contrast, the temporal adverbial functions
as situation predicate in (24b'), and only an interpretation of a continuous
situation is available. Given this analysis, gaps are taken to mean the
boundaries between (sub)situations.

In English, in-temporal adverbials and for-temporal adverbials seem
to exhibit completely different functional behavior to gaps. [rt-temporal
adverbials function as Frame Time adverbials that require verbs to denote
bound situations, as in (26a). Thus, the verb walk which refers to an

unbound situation can not cooccur with in an hour, as in (26b).

(26) a. John wrote a letter in an hour.
b. *John walked in an hour.

In addition to its function as a situation predicate, on the other hand, a for-

temporal adverbial may function as a Frame Time adverbial which requires

131

a multi-gap interpretation, as illustrated in (27).

(27) a. ?John read the letter for a whole day.
b. John read the letter again and again for a whole day.

For a whole day in (27a) can not function as a Frame Time adverbial
without an explicit expression to indicate multi-gaps, but can do so in (27b)
with an explicit expressions to indicate multi—gaps. Thus, gaps are accounted
for by the distinction between Frame Time adverbials and situation
predicates, instead of extralinguistic evidence regarding situations.
Linguistically, gaps appear to be represented between situations as situation
boundaries but not within situations. This explains some of the observations
of the in-temporal expressions and for-temporal expressions in English.2

Finally, verbs are observed to refer to situations that are temporally
bound (telic) and nonbound (atelic) (cf. Binnick 1991:179-83, Dahl 1981,
Dowty 1979, Smith 1991:28-33, Vendler 1967). As I have discussed above,
situations are either durative or instantaneous. Only durative situations
appear to exhibit temporal properties of boundness and nonboundness. In
Vendler's categorization, verbs that denote bound situations are called
'accomplishment' verbs, as in (28a), while verbs that denote nonbound
situations are called 'activity' verbs, as in (28b).

(28) a. John read a book.

b. John ran.

(29) a. John was reading a book.
b. John was running.

In (28a), the situation is said to have an inherent beginning and ending in its

 

2 In section 3.4.2, I have presented a detailed discussion on for-temporal
expressions as Frame Time Adverbials from a syntactical perspective.

132

temporal structure, when one starts to read a book and finishes reading a
book. In (28b), on the other hand, the situation is not considered to have an
inherent beginning and ending in its temporal structure, when one runs. In
the literature on this topic, the distinction is made on the basis that (29a)
does not imply (28a) so that the verb phrase denotes a bound situation,
whereas (29b) implies (28b) so that the verb phrase refers to a nonbound
situation.3

Syntactically, the distinction between boundness and nonboundness is
shown in the cooccurrence of verbs and Frame Time adverbials. It is
observed that in-temporal adverbials can cooccur with verbs denoting
bound situations but not cooccur with verbs denoting nonbound situations,

as shown in (30).

(30) a. John read a book in an hour.
b. *John ran in an hour.

The verb phrase read a book refers to a bound situation so that (30a) is
acceptable. On the other hand, the verb ran refers to a nonbound situation
and can not cooccur with in-temporal expression.

Semantically, the distinction between boundness and nonboundness is
seen in the logical behaviors of sentences with these two types of verbs. For
example, (29a) does not entail (28a), whereas (29b) entails (28b). This does
not seems to be determined by extralinguistic reality of situations, but
rather by the way situations are viewed and represented in language. Given

a situation, we may describe it in (31a) or in (31b).

 

3 This approach has problems and makes incorrect prediction in verb categorization,
as I will shown in section 4.4.

133

(31) a. John is reading a book.
b. John is reading.

(32) a. John has read a book.
b. John has read.

(31a) does not entail (323), though (31b) entails (32b). This further
suggests that bare verbs refer to more general situations and may be
modified to denote more specific situations. Further examples in this
regard are seen in the phenomenon of 'verb class switch or shift' (Dowty
1979:62-4, Vendler 1967:101—2, Vlach 1981:276-7, Zhou 1991). A verb
that usually denotes an nonbound situation may be modified to refer to a

bound situation, as shown in (33) and (34).

(33) a. John ran.
b. John ran two miles

(34) a. John walked.
b. John walked to the store.

Two pieces of evidence show that (33b) and (34b) behave like any other
verbs that refer to bound situations. First, (33b) and (34b) may cooccur

with in-temporal expressions as Frame Time adverbials, as in (33b') and

(34b'), though (33a) and (34a) can not, as in (33a') and (34a').

(33') a. *John ran in an hour.
b. John ran two miles in an hour.
(34') a. *John walked in an hour.

b. John walked to the store in an hour.

Second, there are different entailment relations between sentences in the
progressive forms and sentences in the non-progressive forms with respect
to reference to bound situations and nonbound situations. For example,

(35a) does not imply (33b), nor does (35b) imply (34b).

134

(35) a. John is running two miles.
b. John is walking to the store.

The above evidence indicates that verbs that denote bound situations exhibit
syntactical and logical behaviors different from those of the verbs that
refer to nonbound situations.

In short, 1 have discussed inherent temporal properties of situations
like instantaneousness, duration, gaps, boundness and nonboundness, as are
represented linguistically in English and Chinese. Based on temporal
properties, verbs may be categorized as instantaneous verbs and durative
verbs, the latter of which may further classified as bound and nonbound

durative verbs.

4.2 QUANTIFICATION AND SITUATIONS

In a comparison of reference to situations by verbs and reference to
objects by nouns in section 4.1.1, I have shown that they share a number of
features in terms of quantification. Reference to situations by verb phrases
may have generic interpretations, definite and indefinite interpretations
exactly as reference to objects by noun phrases. As I have argued in section
4.1.1, bare verbs refer to more general situations as bare nouns do to
objects (cf. Carlson 1977), while object noun phrases, subject noun phrases,
temporal adverbials, etc. modify bare verbs so that they may refer to more

specific situations, as reviewed here in (36).

(36) a. John reads.
b. John reads a book.
c. John read a book every evening in those days.
d. John read a book yesterday evening.

It is apparent that (363) refers to a most general situation among the four

135

in (36); (36b) refers to a more specific situation than (363) does but less
specific than (36c) and (36d) do; (36c) refers to a situation less general
than those in (36a) and (36b) but less specific than (36d) does; whereas
(36d) refers to the most specific situation. This is discussed in Dowty
(1979:84—88) in terms of the distinction between universal and existential
interpretations, following a discussion on this topic by Carlson (1977). The
object noun phrase in a verb phrase appear to play an important role in
quantification interpretation.4

However, I am mainly concerned with the distinction between
inherent and noninherent temporal properties of boundness and
nonboundness of situations denoted by verbs phrases. By inherent temporal
properties, I mean verbs or verb phrases referentially designate boundness
and nonboundness of situations, as I have discussed in section 4.1.2. By
'noninherent' temporal properties, I mean the effects of quantification
interpretation on boundness and nonboundness of situations. I will discuss
how quantification interpretation will affect the temporal properties of
inherently bound situations in section 4.2.1. In section 4.2.2, I will show the
relationship between quantification interpretation and inherently nonbound

situations.

4.2.1 INHERENTLY BOUND SITUATION AND QUANTIFICATION
Carlson (1977) argues for two types of predicates: individual and
stage, the former of which is of permanent nature while the latter of which

is temporally bound in nature. Carlson (1977:448) thinks that stages are

 

4 Hoop (1992) considers the relation between noun phrase interpretation and
grammatical cases, and argues that the interpretations of object noun phrases are related to
cases and weak case is related to verbs denoting nonbound situations while strong case is
related to verbs referring to bound situations.

136

conceived of as being more related to events (i.e., bound situations). There
are a few examples involving the interpretation of object noun phrases in
Carlson's work, as in (37), though he is more concerned with bare plurals

and other noun phrases grammatically functioning as subjects in sentences.

(37) a. John repairs a car for a living.
b. John repaired a car yesterday.

In (37a), a car has a generic interpretation and does not refer to a
particular car. Thus, the verb phrase repairs a car for a living is
considered an individual predicate. On the other hand, a car in (37b) has a
specific interpretation and denotes a particular car. Therefore, the verb
phrase repaired a car is considered a stage predicate. Carlson's study
clearly suggests that there is a relationship between the quantification
interpretation and temporal properties of situations, though Carlson's
attention is not on the topic of time. I think that the difference between
generic and specific interpretations lies in referentiality in the sense of
Fodor and Sag (1982), since the indefinite expression in (37a) does not
have a referential reading while that in (37b) does. Referentiality appears
to be important in reference to bound and nonbound situations. The
repairing of a particular car takes a definite interval of time, whereas the
repairing of a nonparticular car does not take a definite interval of time.
A noun phrase with a generic interpretation appears to be universal,
but it is not exactly universal (Allan 1986:136-7). Regardless of the
differences, the sc0pe phenomenon between universal interpretation and
existential interpretation is seen to be involved in temporal properties of
situations (cf. Carlson 1977:419, Dowty 1979:), as shown in (38) on the

next page.

137

(38) Every student read a book on linguistics.
(39) a- (VXXS (X)-->(3Y) (3 (Y) & X read y))
b. (3y)(Vx) (B(y) & (S(x)-->x read y))

Two interpretations are available for the sentence in (38): the universal
quantifier has a wide scope and the existential quantifier has a narrow
scope, as in (39a), or the universal quantifier has a narrow scope and the
existential quantifier has a wide scope, as in (39b). (38) with the
interpretation in (39a) seems to denote a nonbound situation involving
reading one or some book on linguistics, while (38) with the interpretation
in (39b) appears to denote a bound situation involving reading a particular
book. In the interpretation of (39a), it is unclear how the reading is carried
out temporally, while in the interpretation of (39b), there is a sequence of
single events of reading the book in question and there is a beginning and
end of each reading and of this sequence. In English, this distinction is not
morphologically or syntactically clear. However, in Russian where the
perfective occurs with verbs denoting bound situations and imperfective
with verbs referring to nonbound situations, the perfective can only occur
in a sentence with the interpretation of a sequence of single events, as in
(39b), but not in a sentence with the non sequential interpretation, as in
(393), in relation to universal quantification (cf. Merrill 1985). This seems
to suggest that referentiality indeed plays a role in the boundness and
nonboundness distinction, where scopes of universal and existential
quantifiers are involved.

If referentiality underlies the effects of interpretation of noun phrase
quantification on the distinction between bound and nonbound situations, it
may explain the interpretation of bare nouns as object noun in a verb

phrase. Bennett (1981:22-8) observed that the verb phrase in (40a)

138

functions like an activity verb while that in (40b) does not, regardless of
the fact that the verb phrase build a house inherently denotes a bound

situation.

(40) a. John is building houses.

b. John is building two houses
(41) a. John has built houses.

b. John has built one house.

These two sentences logically behave differently. Bennett (1981:22) pointed
out that (40a) implies (41a) but (40b) does not imply (41b). Bennett tried
to provide truth conditions for sentences like (40a) and (40b) in interval
semantics to account for the semantic difference, but failed to explain what
underpins the intervals and subintervals. I think that a more explanatory
account of (40a) and (40b) lies in referentiality. The bare noun houses in
(40a) has a generic interpretation, where the noun houses does not refer to
any particular group of houses (cf. Carlson 1977). It is in this
interpretation that (40a) may imply (41a). If the noun houses in (403)
refers to a particular group of houses, say John is in charge of a housing
development project in a suburb, John may not have finished building any
houses in the project, though he is building houses. Referentiality underlies
the logic behavior of (40b), in which two houses refer to particular houses
so that (40b) does not imply (41b).

In short, verbs that denote inherently bound situations may refer to
nonbound situations when the object noun phrase has a nonreferential
interpretation. Thus, referentiality is related to temporal properties

represented in natural languages.

139

4.2.2 INHERENTLY NONBOUND SIT UATIONS AND
QUANTIFICATION
Quantification plays a role in the bound-nonbound distinction in the
interpretation of inherently nonbound situations as well. Link (1983)
proposes in his theory of plurality that some verbs may contribute a

distributive reading while other verbs may contribute a collective reading,

as illustrated in (42).

(42) a. The children built the raft.
b. The children read a book.

(42a) involves what is called collective predication, where it is understood
that the children built the raft collectively. On the other, distributive
predication is seen in (42b), which means that each child read a book. I will
refer to verb phrases like build the raft as 'collective predicates' and to
verb phrase like read a book as 'distributive predicates'. In fact, (42b) is
ambiguous between a collective reading and a distributive reading. Thus,
(42b) may mean that the children read a book together in its collective
reading, which refers to a bound situation. Even in its distributive reading,
(42b) may either refer to bound situations in which each child read the
same book one after another or refer to nonbound situations in which each
child read a different book. In his study, Link is more concerned with the
interpretation of the plural terms in the subject position than in the
predicates themselves.

In my study, I am interested in the interpretation of noun phrases in
the object position so that I will restrict the subject noun phrases in my
examples to singular terms. In spite of the difference in focus, a collective

predicate refers to a single event, whereas a distributive predicate refers to

140

a sequence of situations or just a set of situations. As I have discussed in
section 2.1, a sequence of situations always gets a bound situation
interpretation.

I want to demonstrate here that some durative verb phrases that
denote inherently nonbound situations may have a collective reading or a
distributive reading when they cooccur with appropriate quantifiers. For

example, the durative verbs in (43) denote inherently nonbound situations,
as in (43).

(43) a. John pushed a cart.
b. John moved a desk.
(44) a. John is pushing a cart.
b. John is moving a desk.

These verbs denote situations that may last a few minutes or a few hours
and that are not linguistically represented with respect to their terminus. In
the literature (cf. Dowty 1979, Kenny 1963, Vendler 1967), the temporal
nonboundness is said to exhibit the implications between (43) and (44) :
(44a) implies (43a), and (44b) implies (43b). The question of concern in

this section is about the sentences in (45).

(45) a. John pushed two carts
b. John moved five desks.

(45a) may refer to two different situations: John pushed two carts at once
or John pushed one cart after the other. The same can be found in (45b): it
is possible that John moved five desks at once, if he is strong enough, or he
m0ved one after another. It appears that no account is ever provided for
the difference in meaning in (45). I think that this difference in meaning

can be accounted for in terms of distributive and collective predicates.

141

When the verb phrases function as a distributive predicate, there is a
distributive reading: each pushing of a cart by John or each moving of a
desk by John is counted in terms of the cardinals in the object noun phrase.
When the verb phrases function as collective predicates, there is a
collective reading: pushing of two carts by John or moving five desks by
John are not counted in terms of those cardinals. Countability is an
important feature of bound situations, since it is basic to event predication
(Armstrong 1981:10-11). Thus, each pushing or moving is viewed as a
single event and the pushing of the two carts or moving of five desks is
viewed as a sequence of single events. Sequences of single events are
considered temporally bound situations, as I have discussed in section 4.2.1.
As compared with the nonboundness in (43), the temporal boundness of the
situations denoted in (45) is supported by linguistic evidence regarding the
cooccurrence of verbs referring to bound situations and in-temporal

expressions as Frame Time adverbial, as in (46) and (47).

(46) a. John pushed a/the/one cart for an hour/*in an hour.
b. John pushed two carts for an hour/in an hour.

(47) a. John moved a/the/one desk for an hour/*in an hour.
b. John moved five desks for an hour/in an hour

In the b-sentences, the object noun phrases with a cardinal number larger
than one can occur with in—temporal adverbials, since those verb phrases
with cardinals larger than one allow a distributive reading. Of course, the
b-sentences can occur with for-temporal adverbials, since those verb
phrases have a collective reading. However, in the a-sentences, the object
noun phrases with quantifiers like a, the and one do not allow any

distributive reading and only allow a collective reading so that they can

142

only occur with for-temporal expressions as situation predicates.

If countability is a basic property of bound situations, cardinal count
adverbials should give a verb phrase 3 distributive interpretation, as they do
in Greek (cf. Armstrong 1981). I think that this is the case in both Chinese
and English. The English case is shown in (48) and (49).

(48) a. *John ran in an hour.
b. John ran twice in an hour.

(49) a. *John was reading a book in an hour.
b. John read a book in an hour.

(48b) not only patterns with (46b) and (47b) but also with (49b) in term of
cooccurrence with in-temporal expressions as Frame Time adverbials. The
predicate becomes a distributive predicate with the cardinal count adverbial,
in which case either running is viewed as a single event taking place in a

temporal sequence. The same phenomenon is found in Chinese too.

(50) 3. *Zhang yi xiaoshi pao le.
John one hour run Asp.
*John ran in an hour.
b. Zhang yi xiaoshi pao le liangci
John one hour run Asp. twice
John ran twice in an hour.

Frame time adverbials like yi xiaoshi (an/one hour), which syntactically
appear in pre-verb positions, do not grammatically cooccur with verbs
denoting nonbound situations, as in (503). When 3 cardinal count adverbial
like liangci (twice) is added, this kind of verbs can grammatically cooccur
with yi xiaoshi (an/one hour), as in (50b).

The above evidence suggests that countability in verbal reference is

as basic to bound situations as countability in nominal reference is to

143

objects. The parallelism in the mass-count distinction between verb phrases
and noun phrases is discussed in Hoepelman and Rohrer (1980) and ter
Meulen (1980, 1981, 1984). Ter Meulen proposes the notion of mass-count
distinction among verbs that coincide the distinction between 'accomplish-
ment verbs' (+ count) and 'activity verbs' (- count), but she does not
consider the mass-count distinction in terms of temporal properties, nor do
Hoepelman and Rohrer. As an extension of Hoepelman (1976), Hoepelman
and Rohrer (1980:85) propose that the accomplishments and achievements
be put in one main category, and the activity and states in another, the
division of which seems to have a parallel in the mass-count distinction in
nouns. However, they stopped short of linking the temporal properties of
bound and nonbound situations to the mass-count distinction, but tried to
explain the mass-count distinction in grammaticality of the sentences in

(51) in terms of properties of mass terms in a model theory .

(51) a. *John drank water in an hour.
b. John drank a glass of water in an hour.

In their theory, informally speaking, water is the name of an individual
concept. The extension of the individual concept at any point (i, j), where i
is a world and j a moment of time, is the union of all things which are
water, at (i, j). To form the union of them, the individuals are considered
as sets. Of course, (473) does not mean that John drank the individual
concept of water, but part of the extension of this concept. They propose a
condition of distributivity, which states that one and the same thing happens
to some subparts of, say, water and to all its subparts as well, and under
which for all parts of this part, John drank them all. Hoepelman and

Rohrer claim that this makes (513) unacceptable, since it cannot be the case

144

that John drank all the subparts of this subpart in an hour if he drank a
subpart of water in an hour (cf. Hoepelman & Rohrer 1980: 89-90).

I believe that the unacceptability of (513) can be more satisfactorily
accounted for in terms of the distinction of the temporal properties
between bound and nonbound situations. The verb phrase drink water does
not denote a situation that is bound and that is well framed in 3 Frame
Time adverbial. (513) can be completely acceptable if cardinal count

adverbials are added to give bound-situation interpretation, as in (52).

(52) a. John drank water just once in an hour.
b. John drank water twice in an hour.

The sentences in (52) indicate that the interpretation of mass terms is not as
crucial as Hoepelman & Rohrer thought, since water is still a mass term in
(52). What is important is whether a verb phrase or predicate denotes a
bound or nonbound situation. I think that the mass-count distinction in
predicates lies in the boundness and nonboundness of situations. Predicates
that denote bound situations are count, while those that denote nonbound
situations are mass. A large range of linguistic phenomena may be
accounted for if the mass-count distinction in predicates is considered in
terms of temporal properties.

To sum up, I have shown that quantification interpretation plays a
very important role in verb phrases that inherently denote nonbound
situations, which is generally ignored in the study of verb classes and their
temporal properties. Some phenomena concerning the mass-count
distinction in predicates can be satisfactorily accounted for in terms of

collective and distributive interpretations. Such interpretations are

145

temporally related to referentiality and situation types.

4.3 SITUATION PREDICATES

In section 3.4, I argued for a distinction of for-temporal expressions
as Frame Time adverbials and situation predicates, following Parsons
(1990:211-2). My discussion in section 3.4 was focused on for-temporal
expressions as Frame Time adverbials. In sections 4.1 and 4.2, I have
assumed this distinction in my discussion of temporal properties of
situations. In this section, I will further discuss some syntactical and
semantic properties of situation predicates in English and Chinese.

The double function of for-expressions as Frame Time adverbials and
as situation predicates is not adequately recognized in the literature on tense,
aspect and temporal expressions. Binnick (1991) and Moltmann (1992) treat
the distinction as a problem of scope phenomena. Dowty (1979:332-6) treats
the distinction as one of different syntactic categories in some cases but not
in other cases.5 On the other hand, Carlson (1981) treats temporal
expressions as some sort of measure of time without such a distinction. For
example, in Carlson's analysis, the distinction between for-temporal
expressions or others as Frame Time adverbials or as situation predicates,

as in (53), is not considered important to the semantics of aspect .

(53) a. John worked for a week/all the time.
b. John was in the room for a day/all the time.

(533) does not tend to refer to a continuous situation that last for a week or

all the time. On the other hand, (53b) tends to refer to a continuous

 

5 Dowty (1979) distinguishes temporal expressions like this week and for-temporal
expressions syntactically, but not in-temporal expressions and for-temporal expressions.

146

situation that may last for a day or all the time. Carlson argues that this
difference should be left to pragmatics. I believe that is wrong. I have
argued in section 3.4.2 that this should be accounted in terms of the
distinction between Frame Time adverbials and situation predicates.

I recognize that there is an ambiguity between for-temporal
expressions as Frame Time adverbials and for-temporal expressions as
situation predicates, as in (53). The ambiguity does not exist at all in the
semantics of temporal relations in natural languages. Moreover, this
ambiguity may be simply a language-specific problem in English, since it
does not exist in other languages like Chinese. Two pieces of evidence
firmly support this view. First, situation predicates logically pattern with
Situation Time in entailment relations. Second, situation predicates
unambiguously function as verbal predicates in Chinese, in contrast to
Frame Time expressions which syntactically function as pre-verb temporal
adverbials only. I will concentrate on the second piece of evidence in this
section, and leave the first to section 4.4.

In section 3.4, 1 pointed out that temporal expressions as Frame
Time adverbials only appear syntactically in pre-verb positions in Chinese.
I review this syntactical phenomenon with two more examples in (54),

before I start to discuss temporal expressions as situation predicates.

(54) a. Zhege xingqi Zhang kan le liangben shu.

This week John read Asp. two book
This week John (has) read two books.

b. Zhang zhege xingqi kan 1e liangben shu.
John this week read Asp. two book
?John this week (have) read two books.

c. *Zhang kan le liangben shu zhege xingqi .
John read Asp. two book this week
John (has) read two books this week.

147

The temporal expression zhege xingqi (this week) grammatically appears in
the sentence-initial position in (543), and in the pre-verb position in (54b),
but its occurrence in the sentence-final position in (54c) is ungrammatical.
Semantically, this kind of temporal expression is classified as 3 Frame
Time adverbial, since it designates an interval or instant of time within
which the asserted situation holds, as defined in section 3.4. In Chinese,
some temporal expressions may occur in the sentence-final positions. In a
sentence with an intransitive verb, the temporal expression simply appears

right after the verb as in (55).

(55) 3. Zhang pao 1e liang xiaoshi.
John ran Asp. two hours
John ran for two hours.
b. Zhang zou le yi tian.
John walk Asp. one day
John walked for a whole day.

In a sentence with a transitive verb, the verb is copied and placed after the
object noun phrase, as in (55), where the temporal expression occurs

following the copy of the verb (cf. Li & Thompson 1981:442-50).

(55) 3. Zhang da lanqiu da le yige xiashi.
John play basketball play Asp. one hour
John plaid basketball for an hour.
b. Zhang kan shu kan le yige xiawu.
John read book read Asp. one afternoon.
John read a book/books for a whole afternoon.

In these structures, the temporal expressions must appear close to the verb
so that no elements may occur between the verb and the temporal
expressions in sentences with intransitive verbs, as in (563), nor can the

copy of a verb be deleted, as in (56b).

148

(56) 3. *Zhang pao le yi quan liang xiaoshi.
John ran Asp. a round two hours
John ran a round (of something) for two hours.
b. *Zhang da lanqiu 1e yige xiashi.
John play basketball Asp. one hour
John plaid basketball for an hour.

(563) is ungrammatical when yi quan (a round) is inserted between the
verb and the temporal expression. (56b) is ungrammatical when the copy
of the verb da is deleted, in which case the temporal expression and the
verb are separated by the object noun phrase. It is important to point out
that this pattern of verb copying is also seen in the cases of instruments and

locatives, as in (57).

(57) 3. Zhang xie xin xie gangbi.
John write letter write fountain pen
John writes letters with fountain pens
b. Zhang chi wufan chi shitang.
John eat lunch eat cafeteria.
John eats his lunch at cafeterias

Instrument and locative are thematic roles that form part of the argument
structure with event structures underlying (Grimshaw 1990:25-9). Thus,
instrument and locative noun phrases may appear after a copy of a verb as
some kind of unaccusative object noun, since they are arguments in the
argument structure.6 Given my assumption that time is an inherent
dimension of an event structure or situations are instantiations of temporal

properties, it is natural that situation predicates pattern after instrument

 

6 This kind of object is more like what Jespersen (1965: 159-80) refers to as 'object
of result' than accusative object in the sense of affectedness, though some such objects may
undergo passivization. However, passivization is not a reliable test of objecthood, since the
subject of a passive sentence in Chinese exhibits both properties of subjecthood and
topichood (cf. Zhou 19933).

149

and locative nouns as part of the argument structure. After all, event
structures underlie argument structures.

In summary, 1 think that the ambiguity between Frame Time
adverbials and situation predicates in English is an accident. Frame Time
adverbials and situation predicates represent two completely different
dimensions of temporal relations in natural languages. The distinction
between the two is necessary to capture the semantics of aspect and to
provide truth conditions for related sentences. Therefore, it can not be left
to pragmatics, as Carlson (1981) claims. My position is supported by
evidence in Chinese that situation predicates form part of the argument
structure, and will be further supported in section 4.4 by evidence that the

two dimensions of time have completely different logical behaviors.

4.4 SOME LOGICAL PROPERTIES OF SITUATION TIME

In the above two sections, 1 have shown that verbs denote situations
in the same way that nouns denote objects, and also shown that situations
are instantiations of temporal properties, such as instantaneousness,
duration, boundness and nonboundness. In this section, I will discuss how
Situation Time is logically or semantically different from the other two, in
particular from Frame Time, as one of the three dimensions of Linear
Time, Frame Time and Situation Time.

I think that I can now informally define Situation Time as (58), after
I have demonstrated with linguistic evidence for the existence of and some

properties of Situation Time.

(58) Situation Time is an instant or interval of time denoted by
verbs, where intervals of time are designated as bound or
nonbound by verb modifiers.

150

The definition in (58) is based on the discussions in previous sections on
instantaneousness, duration, boundness and nonboundness exhibited in the
temporal properties of situations. Situation Time may have a number of
logical properties. 1 am interested here in the properties that distinguish
Situation Time from Frame Time in particular. One such property is the
existence of the part-whole relation in Situation Time. In Situation Time,
the part-whole relation is transitive, as formally represented in (59),

where 5 means the part-whole relation or part-of relation (cf. Abbott
1992, Bunt 1985:56).

(59) (Vx, y, z)(x s y & y s z) -> x s z)

(59) means that if x is part of y, and y is part of z, then x is part of z. I am
not going to go into the formal apparatus of part-whole relations. Instead, I
intend to discuss some semantic properties of Situation Time with
transitivity as the background. Situation Time differs from Frame Time
not only in that the former is an instantiation of temporal properties of the
denotation of verbs while the latter is the denotation of some temporal
expressions, but also in that Situation Time and Frame Time have
completely different inferential patterns with respect to transitivity. In
Frame Time, for example, we see the following inferential pattern, as in
(60), where a sentence with a smaller frame of Frame Time entails one

with a larger frame of Frame Time.

(60) a. John read a book this morning.
b. John read a book today.
c. John read a book this week.

(603) implies (60b) and (60c), and (60b) implies (60c). I refer to this

151

inferential pattern as 'increasing', since we can inference from a sentence
with a smaller frame of Frame Time to one with a larger frame of Frame
Time which contains it, as illustrated in (60). However, the reverse is not
true, since John may have read the book on any day of this week or any
time during the day.7 In Situation Time, we see an inferential pattern that is

the reverse of that of Frame Time, as in (61) and (62).

(61) a. John read part of a book.
b. John is reading a book.
(62) a. John ran.
b. John is running.

Here, I assume that the verb phrases like read part of a book and ran
denote a subsituation of the situation that the verb phrases like is reading a
book and is running denote. This assumption is supported by the fact that
the b-sentence imply the a-sentence in (61), as the b-sentence does to the a-
sentence in (62). I will refer to this entailment relation in Situation Time as
'decreasing', since inference is made only from whole to part or larger
part to smaller part.

The question is why the reverse of the transitive property in
Situation Time underlies the entailment relations so that they are
decreasing, in contrast to the inferential pattern in Frame Time which is
increasing. I think that the answer lie in the distinction between linear
ordering in Linear Time and part-whole relations in Situation Time. In
part-whole relations, it is universal that sentences referring to part never
imply those referring to wholes, since certain parts may have some

property, while other parts of the whole do not. This relation in part-whole

 

7 A detailed discussion on the logical properties of Frame Time is found in section
3.4.

152

relations can be formally represented in ensemble theory (Bunt 1985:56-9),
as in (63)

(63) a. [zsx|P(z)]
b.ys[zsxlP(z)]-/->ysx&P(y)
c.ys[zsx|P(z)]-->ysx

(633) states that z is the part of x including all x-parts for which P is true.
(63b) says that if y is a part of x , it is not necessary that P is true for y,
though it can be deduced that y is a part of x, as in (63c). Therefore,

' temporally speaking, sentences with verbs denoting larger Situation Times
entail ones with verbs denoting smaller Situation Times, while sentences
with verbs denoting smaller Situation Times do not entail ones with verbs
denoting larger Situation Times. This is clearly reflected in that the 3-
sentences do not imply the b-sentences in (61) and (62) but the b-sentences
do imply the a-sentences in (61) and (62). In b-sentences parts of a situation
are realized in time, while parts of it have yet to be realized in time, as
indicated by the present progressive markers. On the other hand, in 3-
sentences, parts of a situation or subsituations are realized in time, as
indicated by the past tense markers.

In the literature, studies focus on the entailment of the a-sentence by
the b-sentence in (62), while they ignore the entailment of the a-sentence by
the b—sentence in (61), but emphasize the non-entailment of the a-sentence
by the b—sentence in (64) (cf. Chesterman 1991:133-8, Dowty 1979:57,
Kenny 1963:175, among many others on this topic).

(64) a. John read a book.
b. John is reading a book.

153

This approach may be due to the ontological distinction between
homogeneity and heterogeneity in analyses of the logical relations between
the b-sentences and a—sentences (cf. Declerck 1979, Dowty 1979, Moltmann
1992, Mourelatos 1978, 1981, Verkuyl 1989). (62b) implies (623), because
(62b) is said to refer to a homogeneous situation, while (64b) does not
imply (643), because (64b) is said to denote a heterogeneous situation. I
think that the distinction between homogeneity and heterogeneity in
situations may be related to theories of actions, but are not directly related
to the temporal properties in question. In (61) and (62), the b—sentences
entail the a-sentences, since there is a part—whole relation or smaller-part
and larger-part relation. In other words, the verb phrases in the b-
sentences refer to a whole situation while the verb phrases in the 3-
sentences refer to a subsituations which is part of the whole situation. This
accounts for the reason why the 3—sentences do not entail the b-sentences in
(61) and (62). On the other hand, this also explains why (64b) does not
entail (643). This explanation is the same for the reason why a-sentences do
not entail b-sentences in (61) and (62): the verb phrase in (643) refers to a
whole situation, whereas the verb phrase in (64b) only refers to part of 3
situation.

Further evidence for this part-whole relation in temporal properties
of Situation Time is found in Finnish (Cf. Chesterrnan 1991, Hoop 1992).
In Finnish, verbal representation of the distinction of part of a situation and
whole situation may be seen in the grammatical cases that the object noun
phrases bear, as illustrated in (65) on the next page, where the partitive
case is marked by the suffix a to the object noun. The partitive case that the
object noun phrase bears in (653) indicates that the situation referred to by

build a house is not complete and is partitive. The accusative case that the

154

object noun phrase bears in (65b) shows that the situation is complete.

(65) a. John rakensi taloa
John built housepART
John was building a/the house
b. John rakensi talon
John built house/(CC
John built a/the house.

We can say that (65b) implies that (653) is true at a subinterval of the

Situation Time denoted by (65b), since there is a part-whole relation.
In Chinese bare nouns have a partitive interpretation (cf. Zhou

1991). In progressive sentences, bare nouns occur more grammatically

than nouns with an indefinite quantifier, as in (66).

(66) 3. Zhang zai kan shu.

John Asp. read book
John is reading a book.

b. Zhang zai kan naben shu
John Asp. read that/the book
John is reading that book.

c. ??Zhang zai kan yiben shu.
John Asp. read a book
John is reading a book.

A bare noun with a partitive reading in (663) is more acceptable than a
noun with an indefinite quantifier in (66c), though a noun with a definite is
also grammatical in (66b), which also has a partitive reading.8

The part-whole relation may be further extended to other partitive
constructions as object noun phrase (cf. Abbott 1992 on partitive

constructions). For example, (673) with a partitive construction does not

 

8 I am not sure if the partitive interpretation here is related to the relation between
partitive structures and definiteness and indefiniteness (cf. Abbott 1992, Hoop 1992).

155

imply (67b), whereas (67b) implies (67a).

(67) a. John read some of each book.
b. John read each book.

This entailment relation exists, since there is a part-whole relation between
(673) and (67b). Some of each book is part of each book. The same relation
is found in (68).

(68) a. Five of those ten boys came.
b. Those ten boys came.

(683) does not entail (68b), since (683) only refers to a part, while (68b)
refers to the whole. On the other hand, (68b) implies (683).

In short, Situation Time, as instantiations of temporal properties of
verb denotation, exhibits completely different logical behaviors from those
of Frame Time which is the denotation of some temporal adverbials,
though the two may denote intervals or instants of time. Entailment
relations concerning Situation Time may be best captured in terms of part-
whole relations. In this approach, it is suggested that the entailment relation
between sentences with verbs in progressive form and sentences with verbs
in non-progressive form may be accounted for in terms of part-whole
relations. In this analysis, sentences with verbs referring to bound
situations or ones with verbs denoting nonbound situations are equally seen

to have the same logical behavior.

4.5 SUMMARY OF CHAPTER FOUR
This chapter begins with the assumption that verbs refer to events or

more generally to situations in the same way that nouns refer to objects,

156

which is vaguely seen in Reichenbach but made more clearly obvious in
Parsons (1985, 1990). Within the approach that situations denoted by verbs
are the basic domain and objects denoted by nouns function as modification
of situations (cf. Gabbay & Moravcsik 1980), this assumption is supported
by evidence in nominalization, similarities between nominals and nouns,
and similarities between nominals and sentences. Situations are considered
instantiations of temporal properties of verb denotations. Linguistic
evidence indicates that verbs may refer to instantaneous situations, or to
durative situations which may be bound or nonbound in their temporal
nature. A distinction between inherently bound situations and inherently
nonbound situations is made which is the basic referential function of verbs
and verb phrases. A further distinction is made between inherently
bound/nonbound situations and noninherently bound/nonbound situations. I
consider the latter as effects of quantification interpretations, which is
viewed in terms of distinctions between individual predicates and stage
predicates (cf. Carlson 1977) and collective predicates and distributive
predicates (cf. Link 1983). It is observed that there is a parallelism between
the mass-count distinction in predicates and that in nouns (cf. Hoepelman &
Rohrer 1980, ter Meulen 1980, 1984), which is often viewed in terms of
homogeneity and heterogeneity. I have tried to demonstrate that this
distinction is temporally related to the distinction between bound and
nonbound situations, rather than to mass terms in predicates per se.
Situation Time, as a dimension of the representation of time in
natural languages, exhibits distinctive logical behaviors that are different
from those of Frame Time and Linear Time. The primary relation in
Situation Time is in fact one between part and whole, as suggested by

linguistic evidence in Chinese and Finnish. A part-and-whole approach

157

opens up a more broad view of semantic relations among sentences in

different tenses and aspects.

Chapter Five
SOME IMPLICATIONS OF THE THREE-DIMENSIONAL
TEMPORAL THEORY

5.1 INTRODUCTION

1 have claimed in the Chapter One that we need a three dimensional
representation of the temporal relations in natural languages, after I briefly
reviewed some of the problems in the study of verbs and times. In Chapter
Three, I have treated two of the three dimensions that I have proposed.
One of them is Linear Time, which is defined as a set of points of time
without duration but with precedence ordered in relation to speech time in
a linear structure. Linear Time precisely characterizes tense systems in
natural languages: tenses only represent temporal precedence relations to
speech time and other points of time. Thus, Linear Time avoids some of
the pitfalls of being too rich and too strong as in Reichenbach's tense
theory and other tense theories (cf. Comrie 1981). The other dimension of
the two is Frame Time, which is defined as a set of intervals of time, which
are denoted by temporal adverbials and within which 3 described
event/activity takes place or a state holds. Frame Time has its origin in
frame adverbials, a class of temporal adverbials mentioned in Bach
(1981:72-75), Bennett and Partee (1978:22-30), Parsons (1990:209-223)
and Smith (1991:155). However, they all fail to give any theoretical
significance to this class of temporal adverbials. Frame Time, as I have

demonstrated with linguistic evidence from Chinese and English, represents
158

159

temporal inclusion relations which are logically different from precedence
relations in Linear Time. In Chapter Four, I have proposed Situation Time,
as a third dimension of the representation of time in natural languages,
which is defined as a set of instants or intervals of time denoted by verbs,
where intervals are designated as bound and nonbound by verb modifiers.
Situation Time has its origin in events (cf. Bach 1986b, Davidson 1968,
1980, Parsons 1980, 1985, 1989, 1990, Reichenbach 1947, Smith 1991),
but is only one dimension of events in the sense of instantiations of
temporal properties (Gabbay & Moravcsik 1980). Thus, Situation Time not
only represents temporal inclusion relations but also part-of relations.
Analysis in terms of part-of relations in Situation Time may avoid
unwanted consequences like the 'imperfective paradox' (cf. Declerck 1979,
Dowty 1977, 1979, Parsons 1989).

In those chapters, the three dimensions are viewed in their own
merits respectively in light of linguistic evidence and against problems in
representing and interpreting time in natural languages. In this chapter, I
will discuss how two or all three dimensions of the representation of
temporal relations interact to represent and interpret some temporal
phenomena in natural languages. In section 5.1, I will attempt to provide an
account of the mechanisms for verbal aspect as vieWpoints or perspective,
which are approaches to verbal aspect in syntax and semantics (cf. Comrie
1976, Smith 1991) and in discourse studies (cf. Hopper 1982, Thelin 1990)
but are not fully justified. I will tackle the present perfect puzzle (cf. Klein
1992) by illustrating how the three dimensions interact to underlie the
temporal relations in section 5.2. Finally, in section 5.3, I will show that
the structural analogies between tenses and pronouns (cf. Partee 1973) are

problematic when we view temporal relations in natural languages in terms

160

of this three dimensional theory.

5.1 MECHANISMS FOR VERBAL ASPECT AS VIEWPOINT S

Verbal aspect seems to be unknown and full of pitfalls and mazes,
compared with tense, though tense is confusing, as is seen in logical and
linguistic analyses (Binnick 1991). Aspect appears to lie in the fact that a
situation may be viewed differently. Thus, aspect is considered to consist of
viewpoints or perspectives in syntactic and semantic studies (cf. Binnick,
1991, Comrie 1976, Smith 1991) and in discourse studies (cf. Hopper
1982, Thelin 1990).

Figuratively, Smith (1991:91) provides the most vivid description

that 3 linguist has made about verbal aspect:

Aspect viewpoints function like the lens of a camera, making objects
visible to the receiver. Situations are the objects on which vieWpoint
lenses are trained. And just as the camera lens is necessary to make
the object available for a picture, so viewpoints are necessary to
make visible the situation talked about in a sentence.

What does the lens of an aspect camera focus on if aspect indeed functions
like the lens of a camera? There appear to be some different opinions.
Citing Holt's (1943) definition of aspect, Comrie (1976:3) seems to hold
the view that 'aspects are different ways of viewing the internal temporal
constituency of a situation'. On the other hand, Binnick (1991:136) writes
that 'the very same situation (event or state of affairs) may be viewed
either imperfectively or perfectively'. Binnick's point is illustrated in the

verbal forms in the English sentence in (l).

(1) John was reading a book when Mary came

161

The situation of reading a book is viewed internally, as the progressive
form indicates. However, the situation of coming is not viewed internally
but externally, since both its initial stage and ending stage are in the view.
In the former case, the situation is only viewed in part, while in the latter
case the situation is viewed in its totality. Moreover, aspects like the
habitual view even more than one situation in relation to Frame Time. The
next question is how the aspect camera adjusts its lens to get different
vieWpoints. More specifically, what are the mechanisms? These two

questions do not seem to be answered in the approaches mentioned above.

5.1.1 IN SEARCH OF MECHANISMS FOR ASPECT AS VIEWPOINTS
Grammatical morphemes that mark aspects definitely play a role in
presenting viewpoints, if aspect is considered as such. However, I think that
grammatical morphemes may play a passive role in this system, in the sense
that they simply mark viewpoints of situations as grammaticalized. For
example, we may get viewpoints of situations without any grammatical

morphemes marking aspects, as illustrated in (2).

(2) Zhang lai deshihou, wo kan shu.
John come when I read book
a. When John came, 1 was reading a book.
b. When John comes, I will begin to read a book.

Without any grammatical morpheme, the sentence in (2) may have two
interpretations: (23) in which the situation of John's coming is viewed
temporally within the situation of my reading a book, and (2b) in which the
situation of John's coming is viewed as temporally being followed by the
situation of my beginning to read a book (cf. Dahl l985:24~5). The Chinese

1, 62

sentence has the potential to present both views, since it is not
grammatically marked by any aspect morpheme. Grammaticalization of
aspect markers may be accounted for in terms of parameters (cf. Smith
1991). That is to say that the presentation of viewpoints may be
parameterized in terms of grammaticalization. However, I think that the
mechanisms which underlie viewpoints may be more universal. For

example, there are two ways to specify the interpretation of (2), as in (3).

(3) 3. Zhang lai deshihou, wo zheng kan shu.
John come when I Asp.(progressive) read book
When John came, I was reading a book.
b. Wanshang ba dian Zhang lai deshihou, wo kan shu.
Evening eight o'clock John come when I read book
When John came at eight o'clock in the evening, I was
reading a book.

In (33), a progressive marker is added to the verb kan (read) so that (33)
has the same interpretation as (2) does in (23). On the other hand, (3b) has
the same interpretation as (2) has in (23) too, when 3 Frame Time
adverbial is added. I believe that this is not an accident. In Chinese, Frame
Time adverbials always play a role in the interpretation of tense and aspect,
when there are no grammatical morphemes for them, as I have discussed in
section 3.3.2.

Frame Time adverbials represent Frame Time, while those verb
phrases refer to Situation Time. Are there any relations between Frame
Time and Situation Time? What are the relations between these two
dimensions of the representation of time if there are any? I think that there
is no doubt that those dimensions of the representation of time in natural
languages are related to each other. Let us look at (3b) again. It is noticed

that in (3b) the Frame Time adverbial wanshang ba dian (eight o'clock in

163

the evening) denotes an instant of time, which is the lowest bound of a
frame of Frame Time. Within an instant of time, we may find two
situations contemporaneous, but may not find two situations ordered one
after another, since an instant is not considered durative (cf. Benthem
1983). However, a precedence interpretation is available When 3 Frame
Time adverbial referring to a larger frame of Frame Time is added to (2),

as in (4) in contrast to (3b).

(4) Wanshang Zhang lai deshihou, wo kan shu.
Evening John come when I read book
a. When John came in the evening, I was reading a book.
b. When John comes in the evening, I will began to read a
book.

Like (2), (4) has two temporal interpretations: in (4a) the two situations are
interpreted as contemporaneous, where the instantaneous situation denoted
by the verb come is viewed in its totality, while the durative situation
denoted by the verb phrase read a book is viewed internally; in (4b) the
two situations are viewed as temporally ordered, where the instantaneous
situation of coming precedes the durative situation of reading. The point
here is not the relation between the two situations but that between the
Frame Time adverbial and the situations. Why can the durative situation be
viewed either internally or externally in (4) with the Frame Time adverbial
of wanshang (in the evening)? Why is the durative situation subject to only
the internal view in (3b) with the Frame Time adverbial of wanshang ba
dian (eight o'clock in the evening)? The evidence clearly points to the
relations between Frame Time and Situation Time, since the interpretations
in (3) and (4) involve situations types like instantaneous ones and durative

ones and sizes of frames of Frame Time.

164

Further evidence concerning the effects of Frame Time adverbials
on sentences with unmarked aspect and tense is seen in Mofu-Gudur, a
Chadic language spoken in Cameroon (cf. Hollingsworth 1991). In Mofu,
verb phrases may occur without any grammatical morphemes for aspect
and tense, though there are such morphemes in the language. The
unmarked sentences occur in conversations when Frame Time is known to
the speakers (linguistically covert), as in (53), or with 3 Frame Time

adverbial, as in (5b) (Hollingsworth 1991:241-3).

(5) a. A c3 zéna.
He weave cloth
He is weaving cloth.
b. Mandaw mandaw 3 ca zana.
Tomorrow tomorrow he weave cloth
Every day he weaves cloth.

Given the context of utterance, the covert Frame Time adverbial in (53) is
taken to mean 'now' so that (53) is understood in the progressive. On the
other hand, (5b) is interpreted as in the habitual aspect, since the temporal
adverbial denotes a large frame of Frame Time.

1 have illustrated with evidence from Chinese and Mofu-Gudur that
Frame Time plays an important role in specifying aspect interpretation in
sentences without marked grammatical morphemes for tense and aspect.
However, this does not mean that Frame Time does not play a role in
determining aspect interpretation in sentences with marked aspect or tense
morphemes. For example, the a-sentences in (6) and (7) are unacceptable,
since there seems to be a conflict between the interpretation of the marked

aspect morpheme and the Frame Time adverbials.

165

(6) a. ?John has built a house early this morning.
b. John has read a book early this morning.
(7) a. ?John read a book at six o'clock.
b. John arrived at six o'clock.

ln (6), (6b) is completely acceptable, while (63) is not. It is noticed that
(63) and (6b) have the same Frame Time adverbial early this morning.
Thus, there is no problem for the cooccurrence of the perfect with the
Frame Time adverbial early this morning. The difference between (63) and
(6b) lies in that the verb phrase in (6a) refers to a much larger situation
than that in (6b) refers to. Given that situations are instantiations of
temporal properties (Gabbay & Moravcsik 1980), there is a problem of the
relations between Frame Time and Situation Time in respect to aspect. In
the case of (6a), the size of Situation Time is too large for the frame of
Frame Time, when the perfective aspect is marked. A similar problem is
found in (7), where (73) is not acceptable but (7b) is acceptable. The frame
of Frame Time is too small, while the size of Situation Time is too large,
when a perfective interpretation is indicated by the past tense. Therefore,
the relations between Frame Time and Situation Time should be considered
as the mechanisms for the lens of the aspect camera in the presentation of
viewpoints, whether a sentence is or is not marked as to its aspect.
Markedness and unmarkedness in tense and aspect may be subject to
parameterization, as suggested in Smith (1991). However, the relations
between Frame Time and Situation Time may be universal in the

representation of aspect as viewpoints.

166

5 - l .2 MECHANISMS: RELATIONS BETWEEN FRAME TIME AND
SITUATION TIME
I claim that the relations between Frame Time and Situation Time
are the mechanisms for the aspect camera to focus its lens on viewpoints, in
order to answer the questions raised about those approaches to aspect as
viewpoints, and to account for the phenomena observed in (2), (3), (4), (5),
(6) and (7). Let us briefly review the concept of Frame Time (cf. (62) in

section 3.4) and Situation Time (cf. (58) in section 4.4) by repeating the
definitions here in (8) and (9).

(8) Frame Time (FT) is a set of intervals of time, which are
denoted by temporal frame phrases and within which 3
described event/activity takes place or a state holds.

(9) Situation Time (ST) is a set of instants or intervals of time

denoted by verbs, where intervals are designated as bound or
nonbound by verb modifiers.

By relations between Frame Time and Situation Time, I mean the temporal
inclusion relations between these two dimensions of the representation of
time. I look at the relations between the two dimensions from the point of
view of how Frame Time relates to Situation Time. Specifically, I view
Frame Time as adjustable like a lens, while Situation Time is the object to
be viewed. This is compatible with the traditional view that situations may
be viewed from different perspectives (cf. Binnick 1991:135-9, Comrie
1976:3-6, Smith 1991:914).

First, let us look at the relations between Frame Time and Situations

in the opposition between the perfective aspect and the imperfective aspect.

The relations between Frame Time and Situation Time are completely

167

different in the perfective aspect and the imperfective aspect. In the

perfective, Frame Time temporally includes Situation Time, as defined in

( 1 0a) and as formally represented in (10b).

(10) 3. Frame Time temporally includes the Situation Time of a

situation, iff Situation Time does not begin earlier nor end
later than Frame Time

b. PERF: STE FT, where E symbolizes temporal inclusion

This temporal relation is clearly illustrated in the English sentences in (11).

(l l) a. John read a book yesterday.
b. John read a book yesterday afternoon.
c. ?John read a book at three yesterday afternoon.

In (113) and (11b), the Frame Time adverbials yesterday and yesterday
afiernoon denote frames of Frame Time that temporally include the
Situation Time referred to by the durative verb phrases. Therefore, a
perfective viewpoint is presented. In (11c), the Frame Time adverbial at
three yesterday afiernoon denotes an instant of Frame Time that does not
temporally include the Situation Time referred to by the durative verb
phrase, since an instant of time is too short to temporally include an
interval of time. As a result, a perfective view point is not available. This
does not mean that 3 Frame Time adverbial referring to an instant of
Frame Time can not cooccur with the perfective aspect. A perfective
viewpoint is available when such 3 Frame Time adverbial cooccurs with an

instantaneous verb referring to an instantaneous situation, as in (12).

(12) a. John began to read a book at three yesterday afternoon.
b. John killed someone at three yesterday afiernoon.

168

The two sentences in (12) both have a perfective interpretation, since
Frame Time includes Situation Time. Given the definition of temporal
inclusion in (103), an instant of time temporally includes another instant of
ti me, when the latter does not begin earlier nor end after the former,
regardless of the nondurative nature of an instant of time. Therefore, the
perfective vieWpoint is presented in (12). In addition to the presentation of
vieWpoints, the temporal relation between Frame Time and Situation Time

constrains semantic interpretations of the temporal relations between

Clauses, as in (13).

(13) 3. ??Yesterday, John read a book, but did not finish reading it.

b. Yesterday, John read a book, but did not finish reading it
until eleven o'clock in the evening.

In (133), the temporal adverbial yesterday functions as a frame of Frame
Time for both clauses.1 The first clause in the perfective clearly states that
the frame of Frame Time yesterday temporally includes the Situation Time
denoted by the verb phrase read a book, but the second clause denies what
the first clause represents. Thus, the two clauses in (133) are semantically

contradictory, as illustrated in terms of temporal relations in (13'3).

(l3')a.STEFT&ST$FT
b.STEFT&STEFT

On the other hand, the sentence in (13b) is semantically acceptable, when
the second clause is modified by adding a temporal adverbial within the
bound of the frame of Frame Time, as in (13'b).

In the imperfective aspect, Frame Time only partially includes

 

1 This function of Frame Time adverbials is fully treated in section 5.3.2.

169

Situation Time temporally, as is defined in (143) and formally represented
in ( 14b) below.

(14) 3. Frame Time (FT) partially includes the Situation Time (ST)
of a situation temporally, iff Situation Time begins earlier
and ends later than Frame Time or begins earlier than Frame
Time or ends later than Frame Time.

b. IMPERF.: ST 0: FT, where 0: symbolizes partial temporal
inclusion.

This partial temporal inclusion relation is clearly seen in the sentences in

imperfective aspect in (15).

(15) a. John was reading a book yesterday.
b. John was reading a book last year.

In (153), the frame of Frame Time yesterday only partially includes the
Situation Time referred to by the verb phrase read a book. The same is
found in (15b), where the frame of Frame Time last year just partially
includes the Situation Time. The absolute size of a frame of Frame Time is
irrelevant in respect to the upper bound, as in (15), where (153) has
yesterday as its Frame Time, while (15b) has last year as its Frame Time.
What matters is the relation between Frame Time and Situation Time. It is
not clear whether Situation Time begins earlier or later than Frame Time,
but it is definite that Situation Time ends later than Frame Time, as

illustrated in the sentences in the imperfective in (16).

(16) 3. Yesterday, John was reading a book (, which he started in the
morning but didn't finish until today).
b. Yesterday, John was reading a book (, which he started the
day before but didn't finish until today).
c. ??Yesterday, John was reading a book (, which he started in
the morning and didn't finish until the evening).

170

When we look at the relative clauses in the parenthesis, (163) suggests that
Situation Time starts earlier than Frame Time and ends after Frame Time,
while (16b) indicates that Situation Time starts within Frame Time but ends
after Frame Time. Thus, (16a) and ( 16b) suggest that Situation Time may
or may not begin earlier than Frame Time, which does not seem to be
important in the imperfective. However, (16c) clearly shows that Situation
Time does not end earlier than Frame Time in the imperfective aspect. In
fact, sentences in the imperfective in Chinese and English always imply that
a situation is incomplete in the sense of something being unfinished.

Given the definition of partial temporal inclusion in (143), there are
more variations in temporal relations than represented by the relatively
Simple imperfective aspect in English. Some languages are finely tuned as
to those variations in temporal relations, while other languages are coarsely
tuned to those variations in temporal relations. This is a problem of
grammaticalization of temporal relations, and may be subject to
Parameterization (cf. Smith 1991). In a partial temporal inclusion, we may
find the relation stated in (17), where the initial of Situation Time is

fOcussed on, as in the inchoative aspect in Chinese.

(17) 3. Frame Time (FT) partially includes the Situation Time (ST)
of a situation temporally, when Situation Time begins at the
same time as Frame Time does or later than Frame Time,
but ends later than Frame Time.

b. INCH.: ST 2 FT, where z symbolizes temporal inclusion by
Frame Time of the initial stage of Situation Time but not the
end of Situation Time.

The temporal relation stated in (17) is seen in the inchoative aspect in

Chinese, as in (18) on the next page, where qi(lat) is an inchoative aspect

171

marker (cf. Chao 1968:251, Wang 1985:156-7).2

(18) 3. Zhang zuotian kan qi zheben shu.
John yesterday read Asp. this book.
John began to read this book yesterday.
b. Zhang jintian zaoshang pao qi.
John today morning run Asp.
John began to run this morning.

(183) means that John started the reading of the book yesterday and didn't
finish reading it yesterday. (18b) means that John started running this
morning and is still running up to the moment of speech. The temporal
relation is clearly seen in the sentences in (19), which are unacceptable,

since the second clause is a contradiction of the first.3

(19) 3. ??Zhang zuotian kan qi zheben shu, erque kan wan le.
John yesterday read Asp. this book and read finish Asp.

b. ??Zhang jintian zaoshang pao qi, erque pao dao le.
John today morning run Asp and run destination Asp.

I am not able to provide English translations for the two sentences in (19),

 

2 The inchoative aspect has two morphemes qi and lai. When used with intransitive
verbs, qi and lai appear together after the verb, as in (i). When used with transitive verbs,
qi occurs between the verb and the object noun while lai appear after the object noun, as in
(ii). In the latter case, lai may be omitted. as in (iii), but lai can not be omitted in the case of
intransitive verbs, as in (i).

(i) Zhang ku qilai.
John cry Asp.
John began to cry.

(ii) Zhang chang qi ge lai.
John sing Asp. song Asp.
John began to sing songs.

(iii) Zhang chang qi ge.

John sing Asp. song
John began to sing songs.

3 Wan is an completive aspect marker, which I translate as finish here in English.
The lexical items do not necessarily represent the same temporal relations as grammaticalized
aspect markers do, as I will discuss below.

172

because the meaning of the inchoative aspect is represent by lexical items
like begin or start in English. The lexical items do not exactly represent the
temporal relations defined in (173), since the translation of (19) in English

with lexical items is not a contradiction, as in (20).

(20) 3. Yesterday, John began to read this book, and finished
reading it.
b. This morning, John began to run, and reached his
destination.

This is true in Chinese too, when lexical items are also used, as in (21).

(21) 3. Zhang zuotian kaishi kan zheben shu, erque kan wan le.
John yesterday begin read this book and read finish Asp.
Yesterday, John began to read this book, and finished
reading it

b. Zhang jintian kaishi pao, erque pao dao le.
John today begin run and run arrive Asp.
Today, John began to run, and reached his destination.

The sentences in (21) are not contradictions either. This further supports
my view that the temporal relation represented by the inchoative aspect, as
defined in (17) is different from that represented by lexical items. By
definition, a sentence in the inchoative refers to an incomplete situation in
respect to the frame of Frame Time denoted by 3 Frame Time adverbial. I
think that the difference between English and Chinese regarding the
inchoative aspect also provides evidence for temporal relations as
universals in aspect. Whether all types of temporal relations are
grammaticalized as verbal aspects is a problem of parameterization (cf.
Smith 1991), which I am not going to discuss in this study.

In short, the above evidence suggests that the relations between

173

Frame Time and Situation Time provide the mechanisms for the focusing
on viewpoints, and correctly predict the meaning of aspect in the cases that

I have examined.

5.1.3 SITUATIONS VIEWED EXTERNALLY

With the relations between Frame Time and Situation Time as the
mechanisms for verbal aspect as viewpoints, I think that I am in a position
to examine the question whether aspects are only ways to view the internal
temporal constituency of a situation (cf. Comrie 1976:3). Given the
definition of the imperfect in (14) and one of its variations, the inchoative
in (17), it is clear that those aspects are ways of viewing the internal
temporal constituency of a situation. However, this is not the case when a
situation denoted by a verb is viewed in the perfective aspect, since by
definition (of. (10) above), the situation is viewed as a whole. In this
subsection, 1 will examine a more clear case, the habitual aspect, in which
situations are viewed externally.

The habitual aspect is traditionally considered as an aspect viewing
the internal temporal constituency of a situation, since a verb in the
habitual aspect is considered referring to one single situation (cf. Comrie
1976, Smith 1991). In Smith's (1991:38-42 and 86-9) classification of
situations, situations denoted by verbs in the habitual aspect are classified as
stative situations. I have claimed in section 4.1 that classification in this
manner is not based on the temporal properties of situations but on other
properties of situations in terms of theories of actions. Smith (1991:42)
recognizes that habitual situations are different from other stative
situations, again, in terms of theories of actions but not in terms of their

temporal relations. Therefore, her classification does not necessarily bear

174

directly on the temporal relations in situations. Temporally speaking, the
so-called stative verbs and verbs in the habitual aspect do differ, though

both refer to durative situations, as in (22).

(22) a. John was sick yesterday.
b. ??John used to watch TV many times yesterday.

(223), which refers to a stative situation, is acceptable, but (22b) in the
habitual aspect is not acceptable, though it is claimed to refer to a stative
situation. Thus, the so—called stative reading is not available. I think that the
unacceptability of (22b) lies in the relations between Situation Time and
Frame Time. The frame of Frame Time is simply too small for (22b) to
get a habitual interpretation. This relation is not recognized in Comrie
(19'76) either. In arguing for the distinction between the iterative aspect
and the habitual aspect, Comrie (1976:27) uses the evidence in (23) for his

argument.

(23) a. The lecturer stood up, coughed five times, and said
b. *The lecturer stood up, used to cough five times, and said

(233) is grammatical but (23b) is not. Comrie's argument is based on the
claim that a limited number of repetitions of a situation is not considered a
single situation. I want to point out that the number of repetitions of a
situation alone is irrelevant to the habitual aspect, if it is not considered in

terms of the relations between Situation Time and Frame Time, as in (24).

(24) a. *The lecturer stood up, used to cough a hundred times, and
said
b. *The lecturer stood up, used to cough thousand times, and
said

 

 

 

175

The number of repetitions of a situation may be increased to a hundred
times, as in (243), or to thousand times, as in (24b). However, the sentences
in (24) are still not acceptable. I believe that (23b) is not acceptable for one
single reason: there must be an agreement for the aspects involved in all
three clauses in (23b), since the covert frame of Frame Time has a

syntactic scope over all the three clauses, as in (25).4

(25) (In those days,) the lecture used to stand up, (used to) cough
five times, and (used to) say

When used to is added to each of those coordinated clauses, (25) is
completely acceptable. I think that the above evidence suffices to show that
the habitual aspect does not involve the concept of stative situations nor a
series of situations as one, but that the habitual aspect involves the relation
between the number of repetitions of a situation and the size of a frame of
Frame Time.

When looking closely, 1 find that the relation between the number of
repetitions of a situation and the size of a frame of Frame Time in the
habitual aspect exhibit these two features: i) the number of repetitions is
relative, i.e. it may be large or small, and ii) size of a frame of Frame
Time is absolute, i.e. it must be large. First, let us look at the size of a

frame of Frame Time for the habitual aspect.

(26) a. ???John used to get up at six last week.
b. ‘2? John used to get up at six last month.
0. ?John used to get up at six in the first half of the year.
d. John used to get up at six last year.

 

4 The covert Frame Time adverbial and its syntactic scope is treated as
demonstrative use and anaphoric behaviors of Frame Time adverbials in section 5.3.2.

176

In (26), (26b) is more acceptable than (263), and (26c) is more acceptable
than both (263) and (26b), while (26d) is completely acceptable. Thus, a
greater acceptability is found in (26), when a frame of Frame Time
becomes larger. Now, let us look at the relative number of a repeated

situation.

(27) a. Last year, John used to walk to the market once a month. (12
repetitions of a situation)
b. Last year, John used to read a book every day. (nearly 365
repetitions of a situation)

When counted, there are only twelve repetitions of the situation of walking
to the market in (27a), and, on the other hand, there are almost three
hundred and sixty-five repetitions of the situation of reading a book in
(27b). 1 think that the number of repetitions may be even smaller when a

frame of Frame Time becomes larger, as in (28).

(28) a. John used to go to his grandparents' house for Christmas,
during the years when he was in elementary school.
b. John used to celebrate his birthday with both his parents
before he was five. (Then, he either went to his mother's or
father's after they got divorced.)

In (283), there are fewer than ten repetitions of the situation of going to
grandparents' house in six or five years of elementary school. Moreover,
there are fewer than five repetitions of the situation of celebrating his
birthday with both his parents in (283). The evidence in (26), (27) and (28)
further supports my claim that the mechanism underlying the habitual
aspect is the relation between the number of the repetitions of a situation
and the size of a frame of Frame Time. Given this kind of temporal

relation, the habitual aspect clearly does not view a situation internally, as

 

177

claimed in Comrie (1976:3), but externally, as I have argued above.

5.1.4 SUMMARY

Aspect is considered to consist of viewpoints or perspectives in
syntactic and semantic studies (cf. Binnick, 1991, Comrie 1976, Smith
1991) and in discourse studies (cf. Hopper 1982, Thelin 1990), and
different ways of viewing the internal constituency of a situation (Comrie
1976:3). However, these approaches leave two questions unanswered: what
mechanisms underlie the aspect camera's lens for viewpoints and whether
aspects only view a situation internally. I have illustrated with evidence
from Chinese, English and some Bantu languages that the temporal
inclusion relations between Frame Time and Situation Time are the
mechanisms for verbal aspect as viewpoints. For example, Frame Time
temporally includes Situation Time in the case of the perfective aspect,
while Frame Time only partially includes Situation Time temporally in the
case of the imperfective aspect. I think that the relations between Frame
Time and Situation Time are universal in verbal aspect. On the other hand,
the evidence from the difference between the grammaticalized inchoative
aspect in Chinese and lexical items expressing the inchoative meaning in
English seems to suggest that the grammaticalization of aspect in natural
languages is subject to parameterization. There is also the difference
between grammaticalized temporal relations and nongrammaticalized ones.
In examining the temporal relation between Frame Time and Situation
Time in the habitual aspect, I have demonstrated with linguistic evidence in
English that verbal aspects do not only view situations internally but also
externally in relation to the size of a frame of Frame Time and to the

number of repetitions of a repeated situation. In the habitual aspect in

 

178

English, the size of a frame of Frame Time has to be absolutely large,
while the number of the repetitions of a situation is relative in the sense
that it does not have to be large. If the size of a frame of Frame Time is
marginal, the number of repetitions of a situation needs to be large,
whereas the number of repetitions of a situation can be very small, when

the size of a frame of Frame Time is absolutely large.

5.2 THE PRESENT PERFECT PUZZLE

The puzzle of the English present perfect is the apparent temporal
contrast between what tense the auxiliary have bears and what the verb
phrase refers to (cf. Brinton 1988:6-14, Comrie 1976:52-56, Klein 1992),

as shown in (29).

(29) a. John has read the book this week/*two days ago this week.
b. John has left school today/*at ten o'clock this morning.

In (29), the verbs refer to situations that occurred in the past. It is clear to
us that the situation of John's reading a book and the situation of John's
leaving happened in the past or before speech time. For example, if speech
time is Friday, the situation of John's reading of a book could have
occurred on Monday, Tuesday, Wednesday, Thursday or Friday. It is quite
a few days ago, if it happened on Monday. However, the auxiliary verb
have is in the present, and these sentences can only occur with temporal
adverbials denoting a time related to the present. Thus, it is ungrammatical
for two days ago and at ten o'clock to cooccur with those sentences in the
present perfect, as in (29). This perplexity gives linguists, logicians and
philosophers a difficult time to determine whether the English present

perfect is a tense or an aspect (cf. Brinton 1988:10, Comrie 1976:52). It is

179

claimed that no successful treatment of the present perfect in English has
been formulated, though a few intuitive observations are available (cf.
Giinthner 1977).

In this section, I will first briefly review four previous approaches to
the English present perfect: the current relevance theory, the indefinite past
theory, the extended now theory, and the so-called pragmatic theory (cf.
Binnick 1991, Brinton 1988, Klein 1992, McCoard 1978).5 After the brief
review, 1 will examine the temporal relations in the English present perfect
in terms of the three dimensional temporal theory developed in this study. I
will classify the English present perfect as a combination of tense and
aspect, since it represents temporal relations in Linear Time and between
Frame Time and Situation Time, as defined in this study. In addition, I
think that the English present perfect represents a relation among the three
dimensions, Linear Time, Frame Time and Situation Time, which makes
the English present perfect different from the perfect in other languages,

such as German.

5.2.1 PREVIOUS APPROACHES

The current relevance theory is one of the four theories about the
English present perfect (cf. Comrie 1976:56-61, McCoard 1978:31-73). In
this theory, 'relevance' is technically understood to refer to certain
implications or suggestions (cf. McCoard 1978:31-32). How is the present
perfect relevant to the present? There are at least four ways that the present

perfect is considered to be relevant to the present (cf. Comrie 1976:56-61).

 

5 As a matter of fact, there is a fifth theory called the embedded past theory (cf.
Binnick 1991, McCoard 1978). This theory covers what was developed in Bach (1967),
Huddleston (1969) and McCawley (1971) in the generative grammar. I will not discuss this
theory here, since it is not directly related to my discussion on this problem here.

 

180

First, a present situation is the result of some past event, as in (30).

(30) a. John has come here.
b. John came here.

(303) in the present perfect implies that John is here, while (30b) in the
simple past does not imply that John is here. This is what Comrie (1976:56)
and Leech (1971:31-2) refer to as 'perfect of result' or 'resultative past'.

Secondly, present experience is related to past situations, as in (31).

(31) a. John has been to China twice.
b. John has never been to China.

(313) indicates that John has first-hand experience and information about
China, which results from past events, while (31b) suggests that John does
not have any first-hand experience or information about China, because of
his inaction in the past. This is termed 'experiential perfect' or 'perfect of
experience' (Comrie 1976:58). However, (31b) is considered a problem for
current relevance theory (McCoard 1978:34). Thirdly, a present situation
is a continuation of a past situation, as in (32), which is referred to as
'perfect of persistent situation' or 'universal perfect' (cf. Comrie 1976:60,
McCawley 1971:104).

(32) a. John has lived here for two years.
b. John has studied at Michigan State University since 1990.

Both (323) and (32b) indicate that the situations in question started in the
past and continue to the present. By the time of the utterance, the situations
still hold. Finally, a past situation is considered to be close to the present, as

shown in (33) on the next page.

181

(33) a. John has recently bought a new car.
b. John has just left.

(333) suggests that the event of buying a new car is so close that John still
has a brand new car (cf. McCoard 1978:35). (33b) appears to suggest that
the event of John's leaving is a few minutes from the present.

However, the current relevance theory as reviewed above does not
provide any clue to the puzzle in (29) in terms of any or all four ways that
'relevance' is viewed. The current relevance theory does not account for
why the present perfect cooccurs with certain temporal adverbials and not
with others nor does it account for the existence of the implications or
suggestions. It simply lists the implications. In addition, 'relevance' as
defined still appears to be too subjective and too vague. Given a situation
where John and Mary are bragging about their first-hand knowledge and
information about China, John went to China twice, once two years ago and

once a year ago, and Mary came back from China yesterday.

(34) a. I (John) have been to China twice.
b. I (Mary) just came back from China yesterday.

Doesn't (34b) suggest that Mary has more recent first-hand experience than
(343) suggests that John does? Thus, it is clear that the current relevance
theory of the present perfect can not solve the present perfect puzzle.

The next theory in question is the indefinite past theory. In this
theory, it is claimed that the present perfect locates situations somewhere
before speech time without specifying a particular point of time in the past
(cf. McCoard 1978:75-6). This theory draws a parallel between the
definiteness and indefiniteness opposition in noun phrases and that in

temporal expressions. Proponents of one or another version of this theory

 

182

are Allen (1966) and Diver (1963). In addition, according to McCoard
(1978:88-96), Reichenbach's (1947) tense theory and Bull's (1960)tense
theory both turn out to be variations of the indefinite past theory.

When definiteness and indefiniteness are understood in Hawkins'
(1978) location theory of definiteness, a temporal expression is definite if it
locates 3 particular point in 3 set of points of time shared by the speaker
and the hearer, and otherwise, the temporal expression is indefinite. The
indefinite past theory provide some clue to the present perfect puzzle in
accounting for why some temporal adverbials may not cooccur with the

present perfect, as in (35).

(35) a. John hasn't seen Mary for some time.
b. *John hasn't seen Mary at six o'clock.

The temporal expression for some time in (353) does not locate a point of
time that is known to the bearer/reader. Therefore, (353) is grammatical.
On the other hand, (35b) is ungrammatical, since the temporal expression
at six o'clock locates a point of time that is identifiable by the

hearer/reader. Further evidence is found in (36).

(36) 3. *John has left yesterday.
b. *John has come on Friday.

This theory may be extended to account for the grammaticality of the

sentences in the present perfect in (37).

(37) 3. John has arrived.
b. John has been to China.

The sentences in (37) do not have any overt temporal expressions. Covert

183

temporal expressions do not seem to locate any known point of time that is
not compatible with the present perfect, if we assume that there is any in
(37). However, the indefinite past theory has a problem, given the
definitions of definiteness and indefiniteness in Hawkins' location theory.
In (38) below, the sentences in the present perfect occur with definite

temporal expressions, and they are grammatical.

(38) a. John has left today.
b. John has been here since last Friday.

The temporal expressions today and since last Friday are definite in the
sense that they locate 3 point or interval of time that is known to the
hearer/reader. The theory can not explain the phenomenon observed in
(38), unless it adopts a theory of definiteness and indefiniteness that
excludes those temporal expressions as definite. That theory of definiteness
and indefiniteness may not be parallel to any theory of definiteness and
indefiniteness about noun phrases, if it is developed.

The third theory that I will review in this section is the extended now
theory (cf McCoard 1978, Mittwoch 1988). In this theory, the present
perfect is analyzed as the marker of prior events which are included within
the overall period of the present, the 'extended now' (cf. McCoard
1978:123). More specifically, Bryan (1936:366—7) notes that the present
perfect includes an action or state within 3 certain limit of time, which is a
period that began in the past and extends up to or into the present. This
theory pays particular attention to temporal adverbials with respect to their
relationship with 'now'. Temporal adverbials are categorized according to
their cooccurrence capabilities with the present perfect and the simple past

in terms of the features of +THEN, -THEN and +/-THEN, as in (39)

184
(cf. McCoard 1978:135, Koziol 1958).

(39) 3. -THEN: at present, up till now, so far, lately, since ...,
during etc.
b. +/-THEN: today, in my life, for three days, recently, just
now, already, etc.
c. +THEN: long ago, yesterday, those days, at six o'clock,
after ..., last night, etc.

The adverbials of the category of -THEN in (393) are supposed to require

the cooccurrence of the present perfect, as in (40).

(40) a. John has written three novels up till now.
b. *John wrote three novels up till now.

(403) is grammatical, since there is 3 cooccurrence of the present perfect
and the -THEN adverbial, up till now, whereas (40b) is ungrammatical,
since the -THEN adverbial, up till now, may not cooccur with the simple
past. On the other hand, temporal adverbials of the category of +/-THEN
are supposed to appear in sentences in either the present perfect or the

simple past, as in (41).

(41) a. John has lefi Detroit today.
b. John left Detroit today.

Both sentences in (41) are grammatical, though the +/-THEN adverbial
today cooccurs with the present perfect in (413) and the simple past in
(41b). Finally, temporal adverbials of the category of +THEN is supposed

not to cooccur with the present perfect but the simple past, as in (42).

(42) a. *John has left Detroit long ago.
b. John left Detroit long ago.

185

It is clear that it is ungrammatical for the +THEN adverbial long ago to
cooccur with the present perfect in (423).

The major problem for the extended now theory comes from
temporal adverbials of the category of +/-THEN, as in (41), since the
temporal opposition of the present perfect and the simple past may not be
captured in terms of binary features. Adding the category of +/-THEN to
temporal adverbials does not reveal the differences in temporal relations
between the present perfect and the simple past, though in formal terms
this category resolves the apparent conflict in cooccurrence phenomenon.
From the cooccurrence compatibility between temporal adverbials and the
present perfect, the extended now theory seems to point to the right
direction of solving the present perfect puzzle, but it stops short of
providing a final solution.

The last theory about the present perfect to be reviewed is the so-
called pragmatic theory of the present perfect. Klein (1992:525) argues
that the incompatibility of the present perfect and most past tense
adverbials has neither syntactic nor semantic causes, but follows from a
pragmatic constraint -- the position-definiteness constraint. In his theory,
Klein proposes a distinction of three types of time spans in the

interpretation of the English present perfect:

(43) a. TU: the time at which the utterance is made
b. Inf(inite)-time: the time which is related to whatever is
expressed by the nonfinite component of the utterance, i.e.
time of the event or situation (TSit for short)
c. Fin(inte)-time: the time which is related to whatever is
expressed by the finite component, i.e Topic Time (IT for
short), the time for which a claim is made

Klein's TU is equal to Reichenbach's (1947) point of speech, while his TSit

186

is equal to Reichenbach's point of the event. Thus, the first two types of
time spans in Klein's theory are straightforward, but his TT needs some
elaboration. According to Klein (1992:535), a witness in court may
truthfully provide the answer in (44b), when he is asked the question in
(443).

(44) a. What did you notice when you checked the cellar?
b. The door was open. It was wooden.

The witness makes a claim only for a limited span of time in the past,
namely, the time for which he is asked to make a claim, though the door
may have been open for a long time and may be still open. The span of
time in (44b) is the time when the witness checked the cellar. This span of
time is Klein's TT. TU, TSit and TT are supposed to interact with each
other to represent temporal relations in Klein's theory. For example, Klein
distinguishes three temporal relations regarding the simple past, the present
perfect and the future, when the relations between ”IT and TU are

concerned, as illustrated in (45).

(45) a. John was in Detroit.

3' PAST: TT before TU (symbolized as TT < TU or TT< for
short)

b. John is in Detroit.

b' PRESENT: TT includes TU (symbolized as TTO)

c. John will be in Detroit.

c' FUTURE: TT after TU (symbolized as TT>TU or TT< for
short)

In (453), a claim is made about John at a time before TU, as (453')
indicates. In (45b), 3 claim is made about John at a time that includes TU,

as represented in (45b'), while in (45c) a claim is made about John at a

 

187

time after TU as represented in (45c'). Klein notes that given two TTos,
TT.,' and TTO", they may differ in duration, but TTO' can not be before or
after TTO", while given two TT<s or TT>s, say TT<' and TT<", TT<' may
be before TT<", after TT<" or they may overlap. Klein thinks that the
present tense form specifies the former temporal relation. Thus, he refers
to the present tense form as 'position-definite or p-definite', while he
considers the past tense form and the future tense form non-position-
definite or non-p-definite. In my understanding, the present tense form is
p—definite, because TT includes TU in his theory. In addition, Klein treats

some lexical expressions representing TSit as p-definite, as in (46).

(46) a. P-definite expressions: <John be in Detroit at six>,
<John be in Detroit on May 24th, 1993>, etc.

b. Non-p-definite expressions: <John be in Detroit>,
<John leave Detroit>, etc.

According to Klein (1992:544), the lexical content of the expressions in
(463) explicitly specifies the time span (TT) in relation to TU so that those
expressions are p-definite, whereas the lexical content of the expressions in
(46b) does not specify the time span (T T) in relation to TU so that they are
not p-definite.6

In Klein's (1992:546) theory, the English present perfect puzzle is
accounted for by what he refers to as a pragmatic principle, the p-

definiteness constraint, as stated in (47) on the next page.

 

6 It seems to me that what are p-definite in (463) are those temporal expressions at
six and on May 24th, 1993, which are supposed to represent TT. Klein does not make this
concept clear in his paper, though he refers to such temporal expressions as p-definite in
passing (see Klein 1992:546). The relation between TT and TU specified lexically in (463)
is apparently different from that specified in the present perfect form.

188

(47) P-definiteness Constraint:
In an utterance, the expression of TT and the expression of TSit
can not both be independently p—definite.

(47) is supposed to explain why the sentence in (48) is not acceptable.
(48) ??John has left Detroit at six.

In (48), the present tense form of has is p-definite, whereas the expression
of TSit <John leave Detroit at six> is independently p-definite, as specified
by the lexical content of at six in the expression. Thus, (48) violates the
pragmatic principle in (47). Why does Klein refer to (47) as a pragmatic
principle? According to Klein (1992:547), semantically it is not false to
have both TT and TSit expressed by p-definite expressions and syntactically
no rule forbids the coocurrence of two p-definite expressions. Therefore,
in Klein's theory, (48) simply does not make sense, but it does not violate
any syntactic or semantic rules.

It seems to me that Klein's theory has at least three problems. The
first problem is his definition of p—definiteness. With regard to tense, he
classifies the present tense form as p-definite, where TT includes TU, as
shown in (45b'), while he classifies the past tense and future tense forms as
non-p-definite, because TT and TU exhibit precedence relations, as shown
in (453') and (45c'). However, with regard to lexical expressions, he
classifies a lexical expression as p-definite, if it specifies any relation
between 'TT and TU at all, as in (463). It is difficult to say that the
expressions with at six and on May 24th I 993 in (463) specify the same
temporal relation as the present tense form does in (45b'). Thus, it is clear
that Klein does not have a single definition for p-definiteness that is applied

universally. Second, his pragmatic account of the present perfect is not

 

189

viable. I think that (48) has a semantic problem, as decompositionally

shown in (49), where (48) is repeated.

(49) ??John has left Detroit at six.
3. John has left.
b. John left at six.

The problem is that (493) does not entail (49b), though (49b) may entail
(49a), given an appropriate context. Specifically, if John has left, it is
possible that he left at seven o'clock, eight o'clock or any time that is
related to the present. There is no guarantee that John left at six, when the
sentence John has lefi is true. Thus, there is a problem of semantic
incompatibility when (493) and (49b) are incorporated into one sentence as
in (48) and (49).7 Simply, if it is true that John has left, it is not necessarily
ture that John left at six. Thirdly, Klein's theory does not provide any clue
to implications or suggestions that the present perfect carries for recency,
result, and experience related to the present. I believe that those features of
the present perfect are pragmatic in nature, while the representation of
temporal relations are not pragmatic but semantic or probably syntactic
when they are grammaticalized. I think that Klein's theory is also
developed along the line of the indefinite past theory without full
recognition of the temporal relations represented in the English present
perfect.

To sum up, I have reviewed four previous theories, the current
relevance theory, the indefinite past theory, the extended now theory and
the so-called pragmatic theory, about the present perfect in English. Each

theory provides some insight into the present perfect puzzle from its own

 

7 The semantic nature of this problem will be further discussed in subsection 5.23.

190

perspective, but none of these four theories provides a complete solution to
the questions about it, such as the questions of the cooccurrence of the

present perfect with temporal adverbials and of its current relevance.

5.2.2 RELATIONS AMONG THE THREE DIMENSIONS: SOLUTIONS

Like tense and aspect, the present perfect in English can also be
accounted for in terms of the relations among the three dimensions of the
representation of temporal relations in natural languages. I have shown in
sections 3.1, 3.2 and 3.3 that tense is a grammaticalized representation of
temporal relations in Linear Time, and shown in section 5.1 that aspect is 3
grammaticalized representation of temporal relations between Frame Time
and Situation Time. I believe that the present perfect is a combination of a
tense and an aspect, when temporal relations are considered in terms of
Linear Time and in terms of Frame Time and Situation Time. I think that
the present perfect in English represents more than those temporal
relations. A solution to the present perfect puzzle and 3 full account for the
meaning and implicatures of the present perfect rests on the relation
between Linear Time and Frame Time, i.e. Frame Time temporally
includes both tand S in Linear Time.

First, let's look at the temporal relations in the present perfect in
terms of Linear Time. I have pointed out in section 3.2 that in English
temporal relations in Linear Time are grammaticalized in the form of verb
inflections. In the English present perfect, temporal relations in Linear

Time are grammaticalized in the verb inflections too, as in (50).

(50) a. John has broken a glass.
b. John and Mary have moved.

 

191

In (50), the past participle morphemes attached to the verbs represent the
relation of being earlier than. However, unlike the morpheme of the past,
the past participle morpheme does not specify what the point of time they
represent is earlier than. It is like the infinitive in the sense that it depends

on others for a specific relation, as in (51).

(51) 3. Being moved, John donated five hundred dollars.
b. Having moved, John and Mary have lost some mail.

In both (513) and (51b), the past participle morphemes indicate that the
situations referred to by those verbs being moved and having moved
precede the situation referred to by the verbs in the main clauses, and thus
precede the speech time. In the present perfect, the earlier-than relation
represented by the past participle morpheme is specified by the auxiliary
have as being earlier than S, since have is in the present form. Thus, the

temporal relation in (51) may be represented as (52) in Linear Time.
(52) (t < S)

The present perfect is similar to the simple past in Linear Time, since
sentences in the simple past represent exactly the same temporal relation in
Linear Time. This is the tense perspective of the present perfect in English.
Now, we can look at the aspect perspective of the present perfect in
English. Verbal aspect represent the temporal relations between Frame
Time and Situation Time in terms of temporal inclusions, as I have
discussed in section 5.1. What is the relation between the present perfect
and the perfective aspect? First, I review the definition for the perfective

aspect in (53), which is repeated from (10), and that for the imperfective in

 

192

(54), which is repeated from (14).

(53) 3. Frame Time temporally includes the Situation Time of a
situation, iff Situation Time does not begin earlier nor end
later than a frame of Frame Time

b. PERF: STE FT, where E symbolizes temporal inclusion

(54) 3. Frame Time (FT) partially includes the Situation Time (ST)
of a situation temporally, iff Situation Time begins earlier
and ends later than a frame of Frame Time or begins earlier
than a frame of Frame Time or ends later than Frame Time.

b. IMPERF.: ST 0: FT, where 0: symbolizes partial temporal
inclusion.

In English, the present perfect has two forms, the simple and the
progressive. I examine the simple form and the progressive form in
relation to the definition in (53). The present perfect in its simple form
clearly shows a perfective interpretation, in contrast to its progressive
form, when it is used with verbs referring to inherently bound situations,

as in (553).

(55) a. John has written a book. (STE FT)
b. John has been writing a book. (ST 0: FT)

The present perfect in its progressive form does not have a perfective
interpretation but an imperfective interpretation, as in (55b). Therefore,
we may say that the present perfect in its simple form is related to the
perfective aspect, while it is not when it is in its progressive form. This
seems to be fine with verbs referring to inherently bound situations (cf.
McCoard 1978:141-145). How does it fare in the case of verbs that refer to
inherently nonbound situations? Comrie (1976:60) refers to this type of

present perfect as 'perfect of persistent situation', as in (56).

193

(56) a. Mary has shopped here for three years.
b. Mary has lived here for ten years.

The situations in (56) are said to continue from the past up to the present.
Will those situations continue to the future? I think that the present perfect
in its simple form does not represent any temporal relations that indicate it

is the case, nor do those verbs indicate so. This is illustrated in (57).

(57) a. Mary has shopped here for three years, but she begins to
shop in the new mall today.
b. Mary has lived here for ten years, and today she is moving
to Ohio.

The added clauses in (57) apparently do not contradict what are stated in
the clauses in the present perfect. Thus, we can say that the present perfect
in those sentences represents a temporal inclusion relation, as stated in (53).
However, the present perfect in its simple form does not indicate either that

the situations will not continue, as in (58).

(58) a. Mary has shopped here for three years, and she will still do
so for another year.
b. Mary has lived here for ten years, and she will live here for
the rest of her life.

The added clauses state that the situations will continue, but the statements
do not contradict what are stated in the clauses in the present perfect either.
This is in a sharp contrast with the sentence in (553), which is in the .
present perfect in its simple form with a verb referring to an inherently

bound situations, as in (59).

(59) a. John has written a book, and he is reading it now.
b. ?John has written a book, and he is writing it now.

194

(593) shows that the situation is complete so that John may read the book,
which is the result of the situation of writing a book. (59b) further
indicates that the situation is complete so that John can not continue it.
Therefore, I think that I can state that in terms of grammaticalization
Frame Time temporally includes Situation Time in the present perfect in its
simple form, regardless of the distinction in the inherently bound and
nonbound situations. In the case of the present perfect in its progressive
form, as in (55b), Frame Time only partially includes Situation Time
temporally.

What makes the English present perfect different from the simple
past is not the temporal relations in Linear Time nor those between Frame
Time and Situation Time. The English present perfect represents a

temporal relation between Linear Time and Frame Time, as stated in (60).

(60) 3. Frame Time (FT) temporally includes t < S in Linear
Time, when neither point occurs earlier or later than one
single frame of Frame Time.

b. PRE(sent) PERF(ect): t E FT & S E FT, where t < S, two

FT s are identical and E symbolizes temporal inclusion

With (60), I can account for the syntactical cooccurrence puzzle of the
present perfect with temporal adverbials, where previous approaches fail.
They fail, since they do not recognize the grammaticalization of the
temporal inclusion relation between Frame Time and Frame Time, as
stated in (60). (60) can predict what temporal adverbial may or may not
cooccur with the present perfect. For example, (60) correctly predict that
the sentences in the present perfect in (61) are acceptable, since they do not

represent the kind of temporal relations stated in (60).

 

195

(61) a. John has written a book this year.
a'tEFT & S EFT, where t<S and FT =thisyear
b. John has seen Mary this morning.
b' t E FT & S E FT, where t < S and FT = this morning

The sentence in (613) is necessarily uttered within the interval of this year
for it to be true. The sentence in (61b) is necessarily uttered within the
interval of this morning, not even with other intervals of today like this
afternoon and this evening. On the other hand, (60) correctly predicts that
the sentences in the present perfect in (62) are not acceptable, since they do

not represent the kind of temporal relations stated in (60).

(62) a. ??John has written a book last year.
a'tEFT&SEFT,wheret<SandFT=lastyear

b. ??John has seen Mary yesterday.
b'tEFT & S EFT, wheret< S and FT = yesterday

In (62), S(peech time) is not temporally included in the frame of Frame
Time. S is in the present, while the frame of Frame Time is in the past.
There is a discontinuity between t and S in terms of Frame Time. Thus,
(60) accounts for part of the puzzle: the problem of the cooccurrence of
the present perfect and the temporal adverbials.

Two other questions about the present perfect also pose problems for
some of the previous approaches. These two questions are why sentences
may occur in the present perfect without any Frame Time adverbials and
why sentences in the present perfect may cooccur with temporal adverbials
like recently, lately, just and before. I think that these two questions can be
satisfactorily answered in my approach. First, let us look at some sentences

that may occur in the present perfect without any Frame Time adverbial, as
shown in (63).

 

196

(63) a. John has written a book.
b. John has seen Mary.

I want to argue that the sentences in (63) do not have overt Frame Time
adverbials but have covert Frame Time adverbials. The contexts in which
the sentences in (63) may occur must have established Frame Time that is
compatible with the present perfect. In other words, it may be said that the
Frame Time adverbials are omitted, because of the demonstrative use of
the Frame Time adverbials.8 In this sense, appropriate Frame Time
adverbials are recoverable from the contexts where the sentences in (63)
may occur. This is illustrated in (64) and (65), where (63b) occurs in an

appropriate context and an inappropriate context respectively.

(64) 3. Why is John so happy this evening?

b. He (John) has seen Mary (this evening/today).
(65) 3. Why was John so happy yesterday?

b. ??He (John) has seen Mary *yesterday/*last night.

Given the context in (643), Frame Time adverbials like this evening and
today may be recovered so that (64b) may occur appropriately in the
present perfect tense. On the other hand, given the context of (653), no
appropriate Frame Time adverbials may be recovered so that (65b) does
not occur appropriately in the present perfect. Second, the observation that
the present perfect cooccurs with recently, lately, just and before , as in

(66), poses a problem for some of the above authors' accounts of the
present perfect (cf. Klein 1992:547-9, McCoard 1978:129-35).

(66) a. John has recently/lately/just written a book.
b. John has been to Detroit before.

 

8 I will discuss the demonstrative nature of Frame Time adverbials in section 5.3.

 

 

197

I argue that this observation can also be accounted for in my approach. The
first three adverbials indicate a certain immediacy or closeness in
precedence relations. As a combination of tense and aspect, the present
perfect represents precedence relations as I have discussed above and
illustrated in (52). It is acceptable for the present perfect to cooccur with
adverbials that do not refer to a frame of Frame Time that is discontinuous
with the frame of Frame Time which includes S.9 The bare before in (66b)
also represents some precedence relation, which is allowed in the present
perfect, as I have argued above. Similarly, in (66b), the bare before does
not refer to a frame of Frame Time that is discontinuous with the frame of
Frame Time that temporally includes S. However, it is ungrammatical for
the present perfect to occur with a before expression, when that expression
denotes a frame of Frame Time that is discontinuous with the frame of

Frame Time temporally including 8, as in (67).
(67) ??John has been to Detroit *before five o'clock/yesterday.

In (67), the expressions before five o'clock and before yesterday indicate

that the frame of Frame Time referred to is discontinuous with the frame

 

9 For example, we can see a discontinuity with two frames in Frame Time in (623),
which is repeated here as (i).

(i) ??John has written a book last year.
(ii) ??This year, John has written a book last year.

(iii) *t E FI‘l & S E F19, where t < S, FT1 2 last year and FT2 : this year

The sentence in (i) has a covert Frame Time adverbial that temporally includes S, as is the
casein a sentence in the present perfect without an appropriate Frame Time adverbial.
Given a context, the sentence in (i) may recover this year as its Frame Time adverbial, as
shown in (ii). The temporal relations in (i)/(ii) are represented in (iii), which indicates that t
and S are temporally included in two discontinuous frames of Frame Time, last year and
this year, respectively. (iii) clearly shows that (i)/(ii) violates (60).

 

 

198

of Frame Time temporally including S. Thus, (67) is ungrammatical.

In short, I have shown that the present perfect in English is a
combination of tense and aspect, since it represents temporal relations in
Linear Time and relations between Frame Time and Situation Time. I have
pointed out that the English present perfect is unique in that it represents 3
temporal relation between Linear Time and Frame Time, in which case a
frame of Frame Time temporally includes both tand S. This temporal
relation is grammaticalized in the present perfect in English. This type of
temporal relation stated in (60) accounts for the cooccurrence of the
present perfect and different temporal adverbials. In a sentence, the present
perfect can cooccur with referring temporal expressions that denote frames
of Frame Time including both tand S in Linear Time. In addition, the
present perfect can cooccur in a sentence with non-referring temporal
expressions that do not indicate discontinuity in the frame of Frame Time
temporally including t and S. Thus, there is only one problem left
unanswered about the present perfect: the current relevance. This is to be

answered in the following subsection.

5.2.3 SEMANTICS VS. PRAGMATICS IN THE PRESENT PERFECT

It appears to me that there is a confusion about the present perfect.
Some theories, such as Klein's (1992) try to account for the present perfect
in terms of pragmatics and to deny its semantic significance. On the other
hand, others, such as the current relevance theory (cf. Comrie 1976,
McCoard 1978) try to explain it in terms of its semantic and pragmatic
consequences. In this section, I plan to sort out what are semantic problems
and what are pragmatic problems with respect to the present perfect.

First, I argue that the problem with the sentence in (68) is semantic

 

 

 

199

in nature, while Klein (1992:545-47) argues is pragmatic since it just
violates his p-definiteness constraint that he considers a pragmatic

constraint.

(68) 3. ??Today, John has read a book at six.
b. ??Today, John has seen Mary yesterday.

I think that the problem is semantic, since temporal relations are truth—
conditional. Both sentences in (68) violate (60) which defines what
temporal relations the present perfect represents. For example, (683)
involves two Frame Time adverbials which refer to two discontinuous
frames of Frame Time respectively. The Frame Time adverbial at six
refers to a smaller frame that is temporally included in a larger frame of
Frame Time referred to by the Frame Time adverbial today. In section
3.4, I have shown in detail that temporal inclusion relations in Frame Time
exhibit some logic behaviors as defined in (69), which is repeated from

(67) in section 3.4.

(69) sf --> sF; sF -/-> sf, iff f E F, where sf = a sentence with a
smaller frame of Frame Time, sF = the same one with a larger
frame of Frame Time which contains the smaller frame of
Frame Time, --> = implies, and -/-> = not imply

In addition to the violation of (60), the sentence in (683) also has a problem
with (69). In (683), the Frame Time adverbial today temporally includes
both tand S, while the Frame Time adverbial at six also temporally
includes t. According to (69), if it is true that John has read a book today,

it is not necessarily true that John read a book at six, though it is true that

 

200

John has read a book today, if he read a book at six today.” There is an
incompatibility in truth conditions, if (683) is decomposed. In (68b), the
Frame Time adverbial temporally today includes S, but the Frame Time
adverbial yesterday temporally includes t. If (68b) entails that John has
seen Mary today, then (68b) does not entail that John saw Mary yesterday
or vice versa, since those two Frame Time adverbials refer to two
discontinuous frames of Frame Time.

Second, current relevance associated with the present perfect appears
in the form of recency, result and experience related to the present (cf.
Comrie 1976:56-60, McCoard 1978:31-44). Current relevance is used to
explain the present perfect in some approaches (McCoard 1978, Comrie
1976). I think that those theories approach the present perfect in the wrong
direction. I believe that current relevance is some kind of generalized
conversational implicature associated with the present perfect (cf. Levinson
1985:126-32). This implicature is based on the grammaticalized temporal
relation that a continuous frame of Frame Time temporally includes both t
and S. Figuratively speaking, we tend to think that everything is related in
3 big building when we have that building in our mind. We tend not to
consider that everything in that building is related when we have a room or
some rooms in that building in our mind. More specifically, there is a
parallelism between the deictic nature of temporal expressions and place-
deictic expressions (Lyons 1977). A Frame Time expression that refers to a
frame of Frame Time temporally including both I and S is like the deictic

expression here, while 3 Frame Time expression that refers to a frame of

 

‘0 The sentence John has read a book today does not imply that John has read a
book this morning, since John could have read the book during any time of the day before
the utterance.

 

201

Frame Time temporally only including t is like the deictic expression there.
When we use here, there is the implicature that everything within the
domain of here is physically closer to us and more related to us than
everything within the domain of there is. This is also true for temporal
expressions. Thus, there is the implicature that tand S are more related and
temporally closer, when one frame of Frame Time temporally includes
both of them in a grammaticalized form, as in (703). They appear less
related and temporally less close when the temporal inclusion relation is
not grammaticalized, as in (70b), and even less related and temporally less
close when they are temporally included in different frames of Frame

Time, as in (70c).

(70) a. John has read a book this week.
b. John read a book this week.
c. John read a book last Monday.

Grammaticalization of this kind of temporal relation is very important to

the implicature of the current relevance, as shown in (71) (cf. Brinton
1988:11-12).

(71) 3. ??Today, John has read a book, but forgot what it says.
b. Today, John read a book, but forgot what it says.

The two sentences in (71) have the same truth conditions, but (713) has the
grammaticalized present perfect form with the implicature of recency,
while (71b) does not. However, (713) is not acceptable, since the speaker
violates pragmatic maxims in uttering those two clauses in a sequence (cf.
Grice 1975, 1978, Levinson 1983:100-18). (71b) is acceptable, since there

is no such implicature from the simple past. Chinese provides further

 

202

evidence that grammaticalization of the temporal relation in question makes

a difference, as in (72).

(72) Jintian Zhang kan le yiben shu, ke wan le shu de neirong.

Today John read Asp. one book but forget Asp. book '3 content
Today, John read a book, but forgot the content of the book.

(72) is as acceptable in Chinese as (71b) is in English. In Chinese, there is
no present perfect in the grammaticalized form, but only a simple past with
an aspect marker indicating completion.

In short, I have discussed the semantic-pragmatic distinction with
respect to the present perfect. I have argued that the unacceptability in a
sentence in the present perfect with two frames of Frame Time results
from a semantic incompatibility, in contrast to the claim (Klein 1992) that
it is pragmatic. On the other hand, I have argued that current relevance
associated with the present perfect is an implicature from the
grammaticalization of the temporal relation that a continuous frame of

Frame Time temporally includes t and S.

5.2.4 SUMMARY

The present perfect in English has been a puzzle for linguists,
logicians and philosophers. Among many theories about it, I have briefly
reviewed the current relevance theory, the indefinite past theory, the
extended now theory, and a so-called pragmatic theory (cf. Brinton 1988,
Comrie 1976, Klein 1992, McCoard 1978). I think that the current
relevance theory does not provide any insight into the temporal relations
represented in the present perfect, since current relevance is just an

implicature from the grammaticalization of the temporal relations in the

 

203

present perfect. In the indefinite past theory, advocators seem to realize
that a definite temporal expression does not cooccur with the present
perfect, but they do not recognize that it is a definite temporal expression
representing 3 discontinuous frame from the one that temporally includes
S. As a result, they have to assign definite temporal expressions like today,
this week, and this year an indefinite feature, giving up their parallelism
between definiteness and indefiniteness in noun phrases and those in
temporal expressions. I think that the extended now theory provides some
insight into the temporal relations in the present perfect in relating tto S.
However, this theory does not realize the exact nature of how I and S are
related, as witnessed in the categorization of temporal expressions in terms
of -THEN, +THEN, +/-THEN, the last of which includes today, recently, and
now. This last category is adopted to account for the cooccurrence of those
temporal adverbials with both the present perfect and the simple past. The
extended now theory does not recognize that the distinction between the
present perfect and the simple past lies in the grammaticalization of the
temporal inclusion of both tand S by one single frame of Frame in the
former and the lack of the grammaticalization of that relation in the latter.
Klein's (1992) pragmatic theory does not recognize either the importance
of the grammaticalization of the temporal inclusion of both tand S by one
single frame of Frame Time in the present perfect, since his solution to the
present perfect puzzle is basically developed along the line of the indefinite
past theory. In addition, Klein does not correctly distinguish what is
semantic and what is pragmatic with respect to the present perfect in his
theory.

Unlike all previous theories (cf. Brinton 1978, Comrie 1976, Klein
1972, McCoard 1988), my account of the present perfect does not create

 

204

any ad hoc categories or notions for that purpose only, but is based on the
three dimensions of the representation of temporal relations in natural
languages. These three dimensions, Linear Time, Frame Time and
Situation Time account for a large range of linguistic phenomena
concerning temporal relations in languages. Temporal relations among the
three dimensions are universal in all natural languages, though
grammaticalization of the relations among the three dimensions may be
subject to parameterization (cf. Anderson 1982, Smith 1991). In my three-
dimensional theory, the present perfect is considered a combination of
tense and aspect, since it represents precedence relations in Linear Time
and temporal inclusion relations between Frame Time and Situation Time.
What makes the present perfect different from the simple past is the
grammaticalization of the temporal inclusion of both tand S in Linear
Time by a one single frame of Frame Time in the present perfect. This
grammaticalized temporal relation is associated with an implicature in the
form of recency, result, experience, etc. that are considered to be related to
the present (of. Brinton 1988, Comrie 1976, Hirtle 1975, McCoard 1978).
Moreover, the theory developed in this study shows that problems of
cooccurrence of some temporal adverbials and the present perfect result
from semantic incompatibility concerning truth conditions for the

sentences in question (cf. Klein 1992, McCoard 1978).

5.3 THE PROBLEM OF STRUCTURAL ANALOGIES BETWEEN
TENSES AND PRONOUNS
It is generally agreed upon that tense in natural languages has a
deictic function (cf. Binnick 128-9, Comrie 1985:13-8, Heny 1982:109-13,
Mellor 1981:1-12, Partee 1973, Richards 1982, Smith 1991:36-9). Tense is

 

205

considered deictic or indexical in the sense that tense locates situations to a
point of time as reference, such as speech time, as a deictic system relates
entities to 3 reference point (Comrie 1985:14-5).

Analogies are generally made between tenses and English here and
there in their deictic function. Here refers to the location where the
speaker is, and there refers to any location that is away from the speaker.
In the tense system, the present tense refers to the location in time where
the speaker is at that point, while other tenses refer to any location of time
that is away from where the speaker is in time. It is noticed that in some
languages, such as Tuscan Italian (Comrie 1985:14), there is a second there
which refers to a location away from two speakers who are in different
locations. The English there may have this function as well. When two
people talk on the phone and refer to a third person's physical location,
there may be used to refer to a location away from both speakers. With
here and two theres, the analogy may better fit in the tense system, where
the present tense refers to a temporal location where the speaker is now,
the past tense refers to a temporal location away in the past and the future
tense refers to 3 temporal location away in the future. Thus, we may safely
say that there is an analogy in deictic functions between here-there-there
and now-then-then in tenses, since then can refer to a temporal location
away from now either in the past or in the future. There appears to be no
problem for that analogy in deictic functions.

Further, Partee (1973) proposes that some structural analogies can
be made between tenses and pronouns. I think that this notion is
problematic. I will briefly review Partee's proposal, and show in terms of
the three dimensional temporal theory that Partee's analogies are

problematic. I think that the deictic function of tenses is limited to the here-

 

206

there analogy discussed above. In addition, I will show that structural
analogies may be made between Frame Time adverbials and pronouns
along the dimension of Frame Time, instead of between tenses and

pronouns along the dimension of Linear Time.

5.3.1 PARTEE'S PROPOSAL

Partee (1973) claims that some structural analogies can be made
between tenses and pronouns. In her proposal, she notes that parallels exist
between tenses and pronouns in demonstrative use and anaphoric use.

In the demonstrative use, the parallelism of tenses and pronouns is
said to be found in the pair of sentences in (73), when he and the past tense

are considered.

(73) a. He shouldn't be in here.
b. I didn't turn off the stove.

According Partee, he in (733) does not refer to simply anybody, but to a
particular person. The referent of the pronoun he may be either identified
by a gesture to point out the person or by the (linguistic or extralinguistic)
context that the hearer shares with the speaker. By the same token, in
Partee's analysis, (73b) does not simply refer to some point in the past at
which (73b) holds or (73b) does not hold, when (73b) is uttered. Rather,
(73b) refers to a particular point or interval of time whose identity is clear
from the extralinguistic context. Thus, the past tense in (73b) refers to the
interval of time before the speaker left his or her house, when it is uttered
halfway down the turnpike. In its deictic use, the present perfect tense is
said to have a unique and unambiguous referent, like the pronoun I, while

the past tense, like the pronoun they, appears to be vaguer in its referent, as

207

in (74).

(74) a. They haven't installed my cable TV yet.
b. John went to a private school.

In (743), the pronoun they does not refer to some particular referents but
to whoever it is supposed to install the cable. The past tense in (74b) is said
not to refer to a particular interval of time, but to refer to whenever it was
that John went to school. Partee thinks that a temporal adverbial is like a

based-generated topic structure, when it occurs, as in (75).

(75) a. The woman in red, she almost ruined my life.
b. John visited her last Monday.

In (723), the pronoun is redundant, while in (75b) the past tense is
redundant, since nominal expressions have already specified their referents.
In the anaphoric use, the parallelism of tenses and anaphors is said to

be found in the pair of sentences in (76).

(76) a. Mary bought a new car, and John wrecked it.
b. Mary had a party last weekend, and John got drunk.

The pronoun it anaphorically refers to the referent designated by the noun
phrase a new car in the previous clause in (763), where the noun phrase a
new car serves as the antecedent for the pronoun it. In (76b), it is said that
the past tense in the second clause refers anaphorically to the interval of
time that is denoted by the temporal expression last weekend in the first
clause. Further, Partee notes that tenses and pronouns function as bound

variables in (77).

 

208

(77) a. If any arrow hits the target, it is mine.
a' Vx (x hits the target --> x is mine)
b. If Mary comes in, John will leave immediately.
b' Vt (Mary comes in (t) --> John will leave (Imm(fut ))(t)))

In (773), any arrow is not interpreted as referential but as universal. Thus,
the pronoun it is a variable bound by a universal quantifier, as in (773'). In
(77b), in a universal interpretation, the if-clause present tense does not
refers to a particular time, but to every time when Mary comes. Partee
treats the future tense in the main clause in (77b) as the present tense +
will, the former of which is bound, as a variable, by the present tense in
the if-clause, as in (77b'). In Partee's analysis, the same phenomenon can be
found in similar sentences with an existential interpretation.

In the above, I have briefly presented Partee's analysis in which a
parallelism is drawn between tenses and pronouns in terms of deictic
functions and anaphoric behaviors. I believe that it is problematic to
compare tenses to pronouns in this way, though a parallelism between tense

and here-there-there is generally right.

5.3.2 ANALOGIES ON THE DIMENSION OF FRAME TIME

In the three-dimensional temporal theory developed in this study,
tense belongs to the dimension of Linear Time. In Linear Time, tense
represents relations between points of time with speech time as the anchor.
Thus, tense does not refer to intervals of time nor instants of time, since it
does not have any referential function. Tense only represents temporal

relations, as defined in (78), which is repeated from (9) in section 3.1.

(78) Linear Time is a set of points of time without duration but
with precedence ordered in relation to speech time in a linear
structure.

209

The notion that tense represents relations between pairs of times is not new
(cf. Binnick 1991:128), though I have argued for a unique dimension of
representation of this type of temporal relation. The idea is clearly

illustrated in the sentences in (79).

(79) a. John left Detroit (yesterday).
a' (t < S)
b. John will leave Detroit (tomorrow).
b' (S > t)

The past tense and the future tense in (78) are not referring expressions but
represent relations of being-earlier-than or being-later-than. The question
may be what refers to t and S. The referential function belongs to the
dimensions of Frame Time and Situation Time, as I have argued in sections
3.4 and 4.1. From the point of view of Situation Time, verbs refer to
situations that instantiate t in terms of instants or intervals of time
depending on the type of verb involved. From the point of view of Frame
Time, Frame Time adverbials are referring temporal expressions that refer
to t , S or intervals of time including 1, S or both, depending on the
context. Therefore, tenses do not have referential functions. This is where
Partee's (1973) proposal goes wrong. What exhibits the demonstrative and
anaphoric behaviors in Partee's sample sentences if tenses do not ? I argue
that Frame Time adverbials have a demonstrative function and anaphoric
behaviors that are structurally parallel to those of pronouns in some
degree.

In the demonstrative use, Frame Time adverbials and pronouns are

generally omitted in here-and-now situations, as in (80) and (81).

 

210

(80) a. (You) wash the dishes (now).

b. (You) turn down the radio (now).
(81) 3. Where are you going (now) ?

b. (I am going) to school (now).

In (80) and (81), the pronouns are omitted, and so are the Frame Time
adverbials, as shown in the parenthesis. This omission is limited to here-
and-now situations, since interpretation of those pronouns and Frame Time
adverbials have already been established. When need arises, those pronouns
and Frame Time adverbials occur to specify who and when, as in (80') and

(81'), which are the emphatic forms of the sentences in (80) and (81).

(80') 3 You wash the dishes now!
b You turn down the radio now!
(81') 3. Where are you going now ?
b. I am going to school now.

In addition, the missing pronouns and Frame Time adverbials in (80) and
81) are grammatical constituents of those sentences so that they can be
recovered in questions, since Wh—expressions are underlying grammatical
elements of either the main clause, or a subordinate clause, if there is one,
in those questions (cf. McCawley 1988:468-469). The questions in (82) and
(83) are made from those statements in (80) and (81).

(82) 3. Who do you ask to wash the dishes ?
b. You.

(83) a. When do you ask me to wash the dishes ?
b. Now.

In response to (803), the bearer may ask the speaker to specify who he asks
to wash the dishes, as in (823), if the bearer is not sure who the speaker

asks. The missing pronoun is recovered in the question form and then in

 

211

the answer in (82). The Frame Time adverbial can be recovered in the
same manner, as in (83), where the hearer simply asks the speaker when he
asks him to do it in response to (803). Thus, the evidence appears to
indicate that the missing pronoun and Frame Time adverbial in (80) are at
least underlying grammatical elements of those sentences in question. The
examples show that in demonstrative use Frame Time adverbials and
pronouns share a number of features in syntax and semantics.

In the anaphoric use, Frame Time adverbials exhibit anaphoric
behaviors that are parallel to those of pronouns in several ways. In a

coordinate structure, like (843), an anaphor may not violated the condition

stated in (84b), as is pointed out by McCawley (1988:336-40).

(84) a.
C

m

conj C

b. An anaphoric device X may not precede its antecedent Y
if either (i) X is in one conjunct of a coordinate structure
and Y in a later conjunct or (ii) X commands Y.

In a coordinate structure as described in (843), Frame Time adverbials like
pronouns observe (84b) for an anaphoric interpretation, as in the sentences

in (76), which is repeated here as (85).

(85) a. Mary bought a new car, and John wrecked it.
b. Mary had a party last weekend, and John got drunk f.

In (853), the nominal antecedent a new car precedes the anaphoric it, and in

(85b) the nominal antecedent last weekend precede an anaphoric Frame

 

212

Time f which is not linguistically overt. Thus, the anaphoric it refers to the
referent of a new car, and the anaphoric Frame Time f refers to the
referent of last weekend. When (84b) is violated, this interpretation does

not exist, as shown in (85').

(85') a. ?Mary bought it, and John wrecked a new car.
b. ?Mary had a party f, and John got drunk last weekend.

In (85'3), backward pronominalization is odd, if it is not ungrammatical
(cf. McCawley 1988:338).The backward 'pronominalization' for the
anaphoric Frame Time f and its antecedent last weekend is strange too in
(85'b). In addition to the problem of acceptability in backward
pronominalization, the pronoun it in (85'3) does not necessarily refer to the
referent of a new car nor does the anaphoric FrameTime f in (85'b)
necessarily refer to the referent of last weekend.

Like a pronoun, an anaphoric f of Frame Time adverbials may

identify with its antecedent several clauses away in a discourse, as in (86).

(86) John lived in Massachusetts during the thirties. He was taken
care of by his grandparents who wanted him to have 3 good
education 1. He went to a private school I.

In (86), John is the antecedent for he, his and him in the following
sentences, while the Frame Time adverbial during the thirties is the
antecedent for the anaphoric Frame Time f in the following sentences as

well.11 However, in discourse, an anaphoric f of Frame Time is not as

 

11 In section 5.1, I have briefly reviewed approaches to verbal aspect as viewpoints
in syntax and semantics (cf. Comrie 1976, Smith 1991) or as perspectives in discourse
studies (cf. Hopper 1982, Thelin 1990). It still remains a question how syntactic and
semantic studies of aspect and discourse studies of aspect may be unified at some level. In
my study, the temporal relations between Frame Time and Situation Time are considered

 

213

vague as pronouns, as Partee (1973) claims in discussion of sentences like

those in (74) above, which is repeated here as (87).

(87) 3. They haven't installed my cable TV yet.
b. John went to a private school f.

The pronoun they is vague in its referents in (873), but the anaphoric f of
Frame Time is not vague at all given a context in which this sentence is
uttered or written. For example, (87b) may occur in the context of (86),
where the anaphoric f of Frame Time has a definite antecedent in
discourse.

In short, I have demonstrated with evidence from English that Frame
Time adverbials can be used demonstratively and anaphorically like
pronouns. In the demonstrative use, omitted Frame Time adverbials like
pronouns may occur overtly in emphatic forms or may be recovered in
questions and answers. In the anaphoric use, anaphoric Frame Time f is
constrained, by rules that govern the relation between an anaphor and its
antecedent, in its relation with the antecedent referring Frame Time
adverbial. On the other hand, anaphoric Frame Time f may establish an
anaphoric relation with its antecedent several clauses away in discourse, as

a pronoun does in the same context.

5.3.3 SUMMARY

I have briefly reviewed the notion that tense in natural languages has

a deictic function (cf. Binnick 128-9), Comrie 1985:13-8, Heny 1982:109-

 

the mechanisms for the aspect camera to get viewpoints. Given this assumption, Frame
Time may be the missing link between verbal aspect proper (syntactic and semantic) and
discourse functions of verbal aspect This is a promising direction of study in verbal
aspect, but I am not able to pursue this topic in the present study.

 

214
I3, Mellor 1981:1-12, Partee 1973, Richards 1982, Smith 1991:36-9). It is

generally agreed upon that the deictic function of tense exhibits a
parallelism with here-there-there (cf. Comrie 1985:14-5). However, I do
not agree with Partee (1973) on her proposal that structural analogies exist
between tenses and pronouns in terms of demonstrative use and anaphoric
behaviors. Her proposal is problematic for a single reason that tenses are
not referential expressions but simply represent temporal relations, such as
being earlier than or later than. In my theory of the three dimensions of
representation of temporal relations, tenses belong to the dimension of a

representation of Linear Time. In fact, structural analogies do exist

 

between pronouns and Frame Time adverbials in demonstrative use and

anaphoric behaviors on the dimension of Frame Time.

5.4 SUMMARY OF CHAPTER FIVE

My concern in this chapter is the significance of the three
dimensional theory of the representation of temporal relations in natural
languages. Among many problems about temporal relations in natural
languages, I have discussed three problems, i.e aspect as viewpoints, the
present perfect puzzle and the fallacy of structural analogies between tenses
and pronouns, within the framework of the three dimensional theory
developed in the present study.

Aspect is considered to consist of viewpoints or perspectives in
syntactic and semantic studies (cf. Binnick, 1991, Comrie 1976, Smith
1991) and in discourse studies (cf. Hopper 1982, Thelin 1990), and also as
different ways of viewing the internal constituency of a situation (Comrie
1976:3). However, these approaches leave two questions unanswered: i)

what mechanisms underlie the aspect camera's lens for viewpoints and ii)

215

whether aspects only view a situation internally. I have illustrated with
evidence from Chinese, English and some Bantu languages that the the
temporal inclusion relations between Frame Time and Situation Time are
the mechanisms for verbal aspect as viewpoints. In the perfective aspect,
Frame Time temporally includes Situation Time, while in the imperfective
aspect, Frame Time only partially includes Situation Time temporally. I
think that the relations between Frame Time and Situation Time are
universal in verbal aspect. On the other hand, the evidence from the
difference between the grammaticalized inchoative aspect in Chinese and
lexical items expressing the inchoative meaning in English seems to suggest
that the grammaticalization of universal temporal relations as aspect is
subject to parameterization. Finally, I have demonstrated with linguistic
evidence in the habitual aspect in English that verbal aspects do not only
view situations internally but also externally in relation to the size of a
frame of Frame Time and to the number of repetitions of a situation.

The present perfect in English has been a puzzle for linguists,
logicians and philosophers. Among many theories about it, I have briefly
reviewed the current relevance theory, the indefinite past theory, the
extended now theory, and a pragmatic theory (cf. Brinton 1988, Comrie
1976, Klein 1992, McCoard 1978). The current relevance theory does not
provide any insight into the temporal relations represented in the present
perfect, since current relevance is just an implicature from the
grammaticalization of the temporal relations in the present perfect. In the
indefinite past theory, advocators seem to realize that a definite temporal
expression does not cooccur with the present perfect, but they do not
recognize that it is a definite temporal expression representing a

discontinuous frame from the one that temporally includes S. I think that

216

the extended now theory provides some insight into the temporal relations
in the present perfect in relating t to S. Unfortunately, this theory does not
realize the exact nature of how tand S are related, as demonstrated in
the categorization of temporal expressions in terms of -THEN, +THEN,
+/-THEN, the last of which includes today, recently, and now. The extended
now theory does not recognize that the distinction between the present
perfect and the simple past lies in the grammaticalization of the temporal
inclusion of both I and S by one single frame of Frame in the former and
the lack of the grammaticalization of that relation in the latter. The last
theory, Klein's (1992) pragmatic approach, does not recognize either the
importance of the grammaticalization of the temporal inclusion of both I
and S by one single frame of Frame Time in the present perfect, since his
solution to the present perfect puzzle is developed basically along the line
of the indefinite past theory. Klein does not appropriately distinguish
semantics and pragmatics with respect to the present perfect in his theory.
Unlike previous approaches (cf. Brinton 1978, Comrie 1976, Klein
1972, McCoard 1988), my account of the present perfect does not create
any ad hoc categories or notions for that purpose only, but is based on the
three dimensional theory with Linear Time, Frame Time and Situation
Time, which accounts for a large range of linguistic phenomena concerning
temporal relations in languages. In my three-dimensional theory, the
present perfect is considered a combination of tense and aspect, since it
represents precedence relations in Linear Time and temporal inclusion
relations between Frame Time and Situation Time. What makes the present
perfect different from the simple past is the grammaticalization of the
temporal inclusion of both tand S in Linear Time by a one single frame of

Frame Time in the present perfect. This grammaticalized temporal relation

 

 

217

is associated with an implicature in the form of recency, result, experience,
etc. that are considered to be related to the present (cf. Brinton 1988,
Comrie 1976, Hirtle 1975, McCoard 1978). On the other hand, the theory
developed in this study shows that temporal relations are semantic in nature
so that problems of cooccurrence of some temporal adverbials and the
present perfect result from a semantic incompatibility concerning truth
conditions for the sentences in question (cf. Klein 1992, McCoard 1978).

Tense in natural languages has 3 deictic function (cf. Binnick 128-
9), Comrie 1985:13-8, Heny 1982:109-13, Mellor 1981:1-12, Partee 1973,
Richards 1982, Smith 1991:36-9). It is generally agreed upon that the
deictic function of tense exhibits a parallelism with here-there-there (cf.
Comrie 1985:14-5). However, I do not agree with Partee (1973) on her
proposal that structural analogies exist between tenses and pronouns in
terms of demonstrative use and anaphoric behaviors. Her proposal is
problematic, since tenses are not referential expressions but simply
represent temporal precedence relations in Linear Time. Structural
analogies do exist between pronouns and Frame Time adverbials in
demonstrative use and anaphoric behaviors on the dimension of Frame
Time.

However, it does not mean that the three dimensional theory
developed in this study is limited to those three problems, though I have
just reviewed those three in this theory. Rather, I believe that this theory
provides insights into universals in temporal relations in natural languages.
Those three problems are simply samples dissected in the framework of

this theory.

 

 

 

Chapter Six
CONCLUSIONS

6.1 CONCLUSIONS

In this study, I began with a set of philosophical questions about the
representation of time in natural languages and a set of more specific
linguistic questions about the representation of time in natural languages.
Rather than answer the set of philosophical questions directly, I have tried
to answer the set of specific linguistic questions in this study. 1 think that
my answers to the set of specific linguistic questions shed light on the set of
philosophical questions.

When considering the representation of time in natural languages,
one would naturally think about verbs, tense and aspect. A review of
research on this topic since Aristotle reveals that studies of verbs do not
begin with the notion of time. In Aristotle's (cf. Metaphysics, translation
1966, De Anima translation 1902) approach to verbs, the focus is on the

 

modes of actions rather than on the temporal properties of actions. It is
Vendler (1957, 1967) who first specifically points out the temporal nature
of verbs in the linguistic cooccurrence compatibility between types of verbs
and temporal adverbials, though he still follows an Aristotelian tradition.
Thus, verbs are generally examined in terms of theories of actions rather
than in terms of temporal theories. Even when the notion of time is applied
in research on this topic in linguistics and semantics, it is used as an

independent measure in terms of instants or intervals of time in possible
218

 

219

world semantics, where a sentence with a certain verb is measured against
an instant or interval of time in a possible world for its truth conditions
(cf. Abbott 1991,Aqvist 1976, 1977, Bennett 1977, 1981, Bennett &
Partee 1978, Dowty 1977, 1979, 1982, Montague 1970, 1973, Parsons
1985, 1989, 1990).

Without proper consideration of time, previous approaches to verbs
generally suffer from three problems. The first problem is the absence of
uniform criteria in verb categorization and its consequence -- the lack of
uniform result, as witnessed in the disagreement about the number of

categories of verbs. For example, there are two or three classes of verbs in

 

Aristotle's categorization, and there are three classes in Kenny's (1963),
whereas there are probably four classes in Ryle's (1949) and definitely
four classes in Vendler's (1967) categorizations. These are not just
differences in terminology but in temporal relations and logical behaviors.
The second problem is methodological, characterized by attempts to
classify each category of verbs in terms of the semantic relation between
sentences in different tenses and aspects before the first problem is settled
in an appropriate way. The most problematic is the entailment relation
approach, where given 3 pair of corresponding sentences with one in the
past tense and one in the present progressive, a sentence with one class of
verbs in the progressive referring to a whole or larger situation is
measured against its corresponding sentence in the simple tense denoting a
subsituation, while a sentence with another class of verbs in the progressive
referring to a subsituation is measured against its corresponding sentence in
the simple tense denoting a whole or larger subsituation. I think that the
methodological problem is in essence an indication of lack of consideration

of temporal properties of verbs. Without clear consideration of the notion

220

of time in verb categorization and without uniformly categorized verbs,
there have been problems to represent adequately temporal relations in
semantics and in tense and aspect logic for natural languages, since these
aspects of verbs in natural languages are essentially temporal relations. A
typical problem is the 'imperfective paradox' that has plagued theories
about verbs, tenses and aspects in instant-semantics, interval-semantics and
event semantics from Montague (1970, 1973), Bennett and Partee (1978),
Dowty (1979) to Parsons (1989, 1990).

If linguistic representations of time are classified along the three
dimensions: Linear Time, Frame Time and Situation Time, as I have
argued for in the current study, problems of previous approaches are
found along these three dimensions and the relations between the three.
What are exactly those three dimensions of the representation of temporal
relations? Linear Time is a set of points of time without duration but with
precedence ordered in relation to speech time in a linear structure. Frame
Time is a set of intervals of time, which are denoted by temporal frame
phrases and within which 3 described event/activity takes place or a state
holds. Situation Time is an instant or interval of time denoted by verbs,
where intervals are designated as bound or nonbound by verb modifiers.

Each dimension of these temporal relations has its formal properties.
Temporal order relations in Linear Time are transitive and reversable, as
formally defined in (10) and (13) in section 3.1. Temporal inclusion
relations in Frame Time are transitive but the entailment relations with
temporal inclusion underpinning can not be reversed, as formally defined
in (64) and (67) in section 3.4. In other words, the difference in the formal
properties between Linear Time and Frame Time leads to completely

different inferential patterns in these two dimensions of temporal relations

 

221

in natural languages. Situation Time exhibits distinctive logical behaviors
that are different from those of Frame Time and Linear Time, as formally
defined in (59) and (63) in section 4.4. The logical behaviors of Situation
Time can be characterized by a transitive relation, of which the temporal
inferential pattern is reversed. This temporal relation is in fact one between
part and whole, as suggested by linguistic evidence in Chinese and Finnish.
In addition to avoiding the imperfective paradox (cf. Declerck 1979,
Dowty 1979), a part—and-whole approach opens up a more broad view of
semantic relations among sentences in different tenses and aspects.

The set of linguistic questions in (2) in section 1.1, with the set of
underpinning philosophical questions in (1) in section 1.1, are essentially
answered in my analysis of temporal relations in natural languages in the
three dimensional theory developed in this study. The first question in (2)
in section 1.1 is whether there is any linguistic evidence regarding the
philosophical questions whether action takes time, whether it takes a period
of time or moment of time and whether a period of time it takes is open-
ended or well-defined. In Chapter Four, I have demonstrated that situations
as denotations of verbs instantiate temporal properties of verbs. Linguistic
evidence in this regard suggests that actions as part of situations do take
time. Further, in subsection 4.1.2, I have shown that the phenomenon of
the cooccurrence between verbs and temporal adverbials indicates that
some verbs denote instantaneous situations, while other verbs denote
durative situations. Thus, I refer to verbs that denote actions or situations
taking a moment of time as instantaneous verbs and to verbs that denote
actions or situations taking a period of time as durative verbs. The second
question in (2) in section 1.1 is what linguistic evidence there is regarding

the nature of a period time that an action takes. I have discussed boundness

222

and nonboundness of Situation Time in terms of cooccurrence between
verbs and in-temporal adverbials and for-temporal adverbials in section
4.1.2, and further in terms of referentiality and quantification in section
4.2. The third question in (2) in section 1.1 is what linguistic evidence
there is for the relations between the time one action takes and that another
action takes and between the time an action takes and the time of speech. In
Chapter Three, I have shown that these relations are represented in Linear
Time in terms of tense and in terms of relations between finite clauses and
nonfinite clauses. The last question is how the linguistic evidence regarding
all the above questions can be represented in syntax and semantics of
natural languages. This study shows that linguistic evidence regarding the
temporal relations in those questions can be considered in the three
dimensional theory and in terms of relations among the three dimensions
with respect to semantics and in terms of grammaticalization of the
relations among the three dimensions with respect to syntax.

What is the significance of the three dimensional theory of the
representation of temporal relations in natural languages? In Chapter Five
of this study, I have discussed three problems, i.e aspect as viewpoints, the
present perfect puzzle and the fallacy of structural analogies between tenses
and pronouns, within the framework of the three dimensional theory
developed in the present study.

Aspect is considered to consist of viewpoints or perspectives in
syntactic and semantic studies (cf. Binnick, 1991, Comrie 1976, Smith
1991) and in discourse studies (cf. Hopper 1982, Thelin 1990), and also as
different ways of viewing the internal constituency of a situation (Comrie
1976:3). However, these approaches leave two questions unanswered: i)

what mechanisms underlie the aspect camera's lens for viewpoints and ii)

223

whether aspects only view a situation internally. I have illustrated with
evidence from Chinese, English and some Cameroon languages that the the
temporal inclusion relations between Frame Time and Situation Time are
the mechanisms for verbal aspect as viewpoints and are not limited to
situations internally. I think that the relations between Frame Time and
Situation Time is universal in verbal aspect though the grammaticalization
of those relations as aspect is subject to parameterization.

The present perfect in English has been a puzzle for linguists,
logicians and philosophers with respect to the cooccurrence phenomenon of
the present perfect and temporal adverbials and to its current relevance.
Among many theories about it, I have briefly reviewed the current
relevance theory, the indefinite past theory, the extended now theory, and a
pragmatic theory (cf. Brinton 1988, Comrie 1976, Klein 1992, McCoard
1978). I have pointed out that all these approach fail to recognize an
essential temporal relation between Frame Time and Linear in the present
perfect. In the current theory, the present perfect is considered a
combination of tense and aspect, since it represents precedence relations in
Linear Time and temporal inclusion relations between Frame Time and
Situation Time. What makes the present perfect different from the simple
past is the grammaticalization of the temporal inclusion of both I and S in
Linear Time by one single frame of Frame Time in the present perfect.
This grammaticalized temporal relation gives rise to an implicature in the
form of recency, result, experience, etc. that are related to the present (cf.
Brinton 1988, Comrie 1976, Hirtle 1975, McCoard 1978).

Tense in natural languages has a deictic function (cf. Binnick 128-9,
Comrie 1985:13-8, Heny 1982:109-13, Mellor 1981:1-12, Partee 1973,
Richards 1982, Smith 1991:36-9). It is generally agreed upon that the

 

224

deictic function of tense exhibits a parallelism with here-there-there (cf.
Comrie 1985:14—5). However, I do not agree with Partee (1973) on her
proposal that structural analogies exist between tenses and pronouns in
terms of demonstrative use and anaphoric behaviors. Her proposal is
problematic, since tenses are not referential expressions but simply
represent temporal precedence relations in Linear Time. Structural
analogies do exist between pronouns and Frame Time adverbials in
demonstrative use and anaphoric behaviors on the dimension of Frame
Time. For example, Frame Time adverbials are constrained, in its
anaphoric behavior, by anaphoric rules governing anaphors in coordinated
structures.

However, it is not the case that the theory developed in this study is
limited to the three problems just reviewed. I believe that this theory
provides insights into universals in temporal relations in natural languages.
Those three problems are samples dissected in the framework of this

theory.

6.2 RECOMMENDATIONS

There are four things which I am not able to accomplish in the
current study but I believe are of great significance for further study.

First, I am not able to incorporate systematically these three
dimensions, Linear Time, Frame Time and Situation Time, of the
representation of temporal relations in natural languages into one single
logic system in representing temporal relations in any sentences. Thus, I
think that further work can be done to provide a formal apparatus that may
elegantly represent temporal relations revealed in the three dimensional

theory.

 

225

Second, I have made claims that temporal relations along those three
dimensions are universal and the grammaticalization of those relations are
subject to parameterization in individual languages. Evidence for those
claims mainly comes from Chinese and English, though I have occasionally
cited evidence from Bantu languages and Finnish. It is desirable if further
research within the framework of the three dimensional theory can be
carried out in more languages or more extensively in a particular language.

Thirdly, I think that the theory that I have developed in the present
study has broad implications for existing problems, puzzles and paradoxes
on temporal relations in natural languages. For example, I have shown in
the three-dimensional theory that it is fallacious to draw an analogy
between tenses 3nd pronouns (cf. Partee 1973) and that such analogy exists
between Frame Time adverbials and pronouns instead. The insight
provided by the current theory into the analogy problem has significant
implications for the adverbial scope paradox originated in Pior's (1957,
1967, 1968) treatment of both tense and temporal adverbials as logic
operators (cf. Binnick 1991:310-3, Dowty 1982). Solutions to this paradox
are challenges to formal semantics (cf. Binnick 1991:311). Therefore, I
think that further research on temporal puzzles and paradoxes can be
carried out fruitfully in the three dimensional theory developed in this
study.

Finally, I think that the current theory provides a missing link
between semantic and syntactic studies of verbal aspect and discourse
studies of verbal aspect. In discourse studies of verbal aspect, semantic and
syntactic approaches are generally criticized for their failure and for their
lack of significance for discourse studies (cf. Thelin 1990). In my

discussion of aspect in section 5.1 and of structural analogies between

226

pronouns and tense in section 5.3, I have mentioned in passing that Frame
Time adverbials are related to discourse. Given the observations that the
relations between Frame Time and Situation Time are the mechanisms for
verbal aspect as viewpoints and perspectives and that Frame Time
adverbials are related to discourse, significant further study of the relations
between semantic and syntactic perspectives of verbal aspect and discourse

perspective can be done in this three dimensional theory.

BIBLIOGRAPHY

BIBLIOGRAPHY

Abbott, Barbara. 1991. Montague grammar. International encyclopedia of
linguistics, ed. by William Bright, vol. 2, 438-440. New York:

Oxford University Press. 1

Abbott, Barbara. 1992. Doing without a partitive constraint. ms. Michigan
State University.

 

Allan, Keith. 1986. Linguistic meaning. London. Routledge & Kegan Paul.

Allen, Robert. 1966. The verb system of present-day American English.
The Hague: Mouton.

Anderson, Lloyd. 1982. The 'perfect' as a universal and as a language-
specific category. Tense-aspect: between semantics and pragmatics,
ed. by Paul Hopper, 227-264. Amsterdam: John Benjamins.

Aqvist, Lennart. 1976. Formal semantics for verb tenses as analyzed by
Reichenbach. Pragmatics of languages and literature, ed. by Ttin A.
van Dijk, 229-236. Amsterdam: North-Holland Publishing.

Aqvist, Lennart. 1977. On the analysis of some accomplishment and
activity Verbs. On the logical analysis of tense and aspect, ed. by
Christian Rohrer, 31-65. Tiibingen: TBL Verlag Gunter Narr.

Aqvist, Lennart, and Franz Giinthner. 1978. Fundamentals of a theory of
verb aspect and events within the setting of an improved tense logic.
Studies in formal semantics, ed. by Franz Giinthner and Christian
Rohrer, 167-200. Amsterdam: North-Holland publishing.

Aqvist, Lennart; Jaap Hoepelman and Christian Rohrer. 1980. Adverbs of
frequency. Time, tense, and quantifiers: proceedings of the Stuttgart
conference on the logic of tense and quantification, ed. by Christian
Rohrer, 1-18. Tiibingen: Niemeyer.

227

228

Aristotle. 1966. Metaphysics. Translated by Richard Hope. Ann Arbor,
University of Michigan Press.

Aristotle. 1902. De anima. Aristotle's psychology: a treatise on the
principle of life, translated by William Alexander Hammond, 1-143.
London: Swan Sonnenschein & Co..

Armstrong, David. 1981. The ancient Greek aorist as the aspect of
countable action. Syntax and semantics: tense and aspect, ed. by Philip
Tedeschi & Annie Zaenen, 13—30. London: Academic Press.

Bach, Emmon. 1967. Have and be in English syntax. Language 43.462-85.

Bach, Emmon. 1980. Tense and aspect as functions on verb-phrases. Time,
tense, and quantifiers: proceedings of the Stuttgart conference on the
logic of tense and quantification, ed. by Christian Rohrer, 19-38.
Tubingen: Niemeyer.

Bach, Emmon. 1981. On time, tense, and aspect: an essay in English
metaphysics. Radical Pragmatics, ed. by Peter Cole, 63-81. New
York: Academic Press.

Bach, Emmon. 1986a. The algebra of events. Linguistics and philosophy 9:
5-16.

Bach, Emmon. 1986b. Natural language metaphysics. Logic, methodology,
and philosophy of science VII, ed. by Ruth Barcan-Marcus, George
J. W. Don and Paul Weingartner, 573-595. Amsterdam: North
Holland Publishing.

Bennett, Michael. 1977. A guide to the logic of tense and aspect in English.
Logique et analysis. 20.491-517.

Bennett, Michael. 1981. Of tense and aspect: one analysis. Syntax and
semantics: tense and aspect, ed. by Philip Tedeschi & Annie Zaenen,
13-30. London: Academic Press.

Bennett, Michael and Barbara Partee. 1978. Toward the logic of tense and
aspect in English. Indiana University Linguistics Club.

Benthem, Johan van. 1980. Points and periods. Time, tense, and quantifiers:
proceedings of the Stuttgart conference on the logic of tense and
quantification, ed. by Christian Rohrer, 39-58. Tiibingen: Niemeyer.

229
Benthem, Joan van. 1983. The logic of time. Dordrecht: Kluwer Academic.

Binnick, Robert. 1991. Time and the verb: 3 guide to tense and aspect. New
York/Oxford: Oxford University Press.

Brinton, Laurel. 1988. The development of English aspectual systems:
aspectualizers and post-verbal particles. Cambridge: Cambridge
University Press.

Bryan, W. 1936. The preterite and the perfect tense in present-day English.
Journal of English and Germanic philology 35.363-382.

Bull, William. 1960. Time, tense and the verb: 3 study in theoretical and
applied linguistics with particular attention to Spanish. Berkeley:
University of California Press.

Bunt, Harry. 1978. A formal semantic analysis of mass terms and amount
terms. Amsterdam papers in formal grammar, Vol. 11, ed. by
Groenendijk en Stokhof, 81-93. Amsterdam: University of
Amsterdam.

Bunt, Harry. 1985. Mass terms and model-theoretic semantics. Cambridge:
Cambridge University Press.

Carlson, Greg. 1977. A unified analysis of the English bare plural.
Linguistics and philosophy 1.413-457.

Carlson, Lauri. 1981. Aspect and quantification. Syntax and semantics:
tense and aspect, ed. by Philip Tedeschi & Annie Zaenen, 31-64.
London: Academic Press.

Chao, Yuen Ren. 1968. A grammar of spoken Chinese. Berkeley: 1
University of California Press.

Chesterman, Andrew. 1991. On definiteness. Cambridge: Cambridge
University.

Chomsky, Noam. 1970. Remarks on nominalization. Readings in English
transformational grammar, ed. by R. Jacobs and P. Rosenbaum.
Waltham, MA: Blaisdell.

Clifford, John. 1975. Tense and tense logic. The Hague: Mouton.

230

Comrie, Bernard. 1976. Aspect: an introduction to the study of verbal
aspect and related problems. Cambridge: Cambridge University
Press.

Comrie, Bernard. 1981. On Reichenbach's approach to tense. Papers from
the 17th regional meeting of Chicago linguistic society, 24-30.
Chicago: Chicago Linguistic Society.

Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press.

Cresswell, M. J. 1977. Interval semantics and logical words. On the logical
analysis of tense and aspect, ed. by Christian Rohrer, 7-29. Tfibingen:
TBL Verlag Gunter Narr.

Dahl, Osten. 1981. On the definition of the telic-atelic (bounded-
nonbounded) distinction. Syntax and semantics: tense and aspect, ed.
by Philip Tedeschi & Annie Zaenen, 79-90. London: Academic Press.

Dahl, Osten. 1985. Tense and aspect systems. New York: Basil Blackwell.

Davidson, Donald. 1968. The logical form of action sentences. The logic of
decision and action, ed. by Nicholas Rescher, 81-95. Pittsburgh:
University of Pittsburgh Press.

Davidson, Donald. 1980. Essays on action and events. Oxford: Clarendon
Press.

Declerck, Renaat. 1979. On the progressive and the 'imperfective paradox'.
Linguistics and philosophy 3.267-272.

Diver, William. 1963. The chronical system of the English verb. Word
19.141-181.

Dowty, David. 1972. Studies in the logic of verb aspect and time reference
in English. Department of Linguistics, University of Texas, Austin.

Dowty, David. 1977. Toward a semantic analysis of verb aspect and the
English 'imperfective' progressive. Linguistics and philosophy 1.45-
78.

Dowty, David. 1979. Word meaning and Montague grammar: the semantics
of verbs and times in generative semantics and Montague's PT Q.
Boston: D. Reidel Publishing Company.

231

Dowty, David. 1982. Tenses, time adverbs, and Compositional Semantic
Theory. Linguistics and philosophy 5. 23-55.

Dowty, David, Roberte Wall, and Stanley Peters. 1981. Introduction to
Montague Semantics. Dordrecht: Reidel.

Fodor, Janet and Ivan Sag. 1982. Referential and quantification indefinites.
Linguistics and philosophy 5. 355-398.

Gabbay, Dov and Julius Moravcsik. 1980. Verbs, events, and the flow of
time. Time, tense, and quantifiers: proceedings of the Stuttgart
conference on the logic of tense and quantification, ed. by Christian
Rohrer, 59-84. Tubingen: Niemeyer.

Galton, Antony. 1984. The logic of aspect. Oxford: Clarendon Press.
Grimshaw, Jane. 1990. Argument structure. Cambridge, Mass: MIT Press.

Grice, H. P. 1975. Logic and conversation. Syntax and semantics 3: speech
acts, ed. by P. Cole & J. L. Morgan, 41-58. New York: Academic
Press.

Grice, H. P. 1978. Further notes on logic and conversation. Syntax and
semantics 9: pragmatics, ed. by P. Cole, 113-28. New York:
Academic Press.

Giinthner, Franz. 1977. Remarks on the present perfect in English. On the
logical analysis of tense and aspect, ed. by Christian Rohrer, 83-98.
Tiibingen: TBL Verlag Gunter Narr.

Hawkins, John. 1978. Definiteness and indefiniteness: 3 study in reference
and grammaticality prediction. London: Cromm Helm.

Heath, Danial. 1991. Tense and aspect in Makaa. Tense and aspect in eight
languages in Cameroon, ed. by Stephen Anderson and Bernard
Comrie, 3-15. Texas: the Summer Institute of of Linguistics and
University of Texas at Arlington.

Heny, Frank. 1982. Tense, aspect and time adverbials II. Linguistics and
philosophy 5.109-154.

Hirtle, Walter. 1975. Time, aspect and the verb. Quebec: Les Presses de
L'universite’ Laval.

232

Hoeksema, Jack. 1983. Plurality and conjunction. Studies in modeltheoretic
semantics, ed. by Alice ter Meulen, 63-84. Dordrcht: Foris.

Hoepelman, Jaap. 1974. Tense logic and the semantics of the Russian
aspects. Theoretic linguistics 1 . 158- 180.

Hoepelman, Jaap. 1976. Mass nouns and aspects, or: Why we can't eat
gingercake in an hour. Amsterdam papers in formal grammar, Vol.
I. 137-153.

Hoepelman, Jaap. 1978. The analysis of activity verbs in a Montague-type
grammar. Studies in formal semantics, ed. by Franz Gtinthner and
Christian Rohrer, 121-166. Amsterdam: North-Holland publishing.

Hoepelman, Jaap, and Christian Rohrer. 1980. On the mass-count distinction
and the French imparfait and passe simple. Time, tense and
quantifiers, ed. by Christian Rohrer, 85-112. Tiibingen: Niemeyer.

Hollingsworth, Kenneth. 1991. Tense and aspect in Mofu-Gudur. Tense and
aspect in eight languages of Cameroon, ed. by Stephen Anderson and
Bernard Comrie, 239-55. Arlington, TX: the Summer Institute of
Linguistics and University of Texas at Arlington.

Holt, Jens. 1943. Etudes d'aspect. Acta Jutlandica 15.2.

Hoop, Helen de. 1992. Case configuration and noun phrase interpretation.
Groningen: Grodil.

Hopper, Paul. 1982. Tense-aspect: between semantics and pragmatics.
Amsterdam: John Benjamins.

Homstein, Norbert. 1977. Toward a theory of tense. Linguistic inquiry 8.
521-557.

Homstein, Norbert. 1990. As time goes by. Cambridge, Mass: MIT Press.

Huddleston, Rodney. 1974. Further remarks on the analysis of auxiliaries
as main verbs. Foundations of language 11.215-229.

Jespersen, Otto. 1965. The philosophy of grammar. New York: Norton.

233

Kamp, Hans. 1979. Events, instants and temporal reference. Semantics
from different points of view, ed. by Bauerle et 31. Berlin: Springer
Verlag.

Kenny, Anthony. 1963. Action, emotion and will. London: Routledge &
Kegan Paul.

Klein, Wolfgang. 1992. The Present perfect puzzle. Language 68.525-52.

Koziol, H. 1958. Zum Gebrauch des Present Perfect und des Past Tense.
Die Neuren Sprachen 7.497—506.

Landman, Fred. 1991. Structures for semantics. Norwell, MA: Kluwer
Academic Publishers.

Langacker, Ronald. 1982. Remarks on English aspect. Tense-aspect:
between semantics & pragmatics, ed. by Paul J. Hopper.
Amsterdam: John Benjamins.

Leech, Geoffrey. 1971. Meaning and the English verb. London: Longmans.
Levinson, Stephen. 1983. Pragmatics. Cambridge: Cambridge University.

Li, Charles and Sandra Thompson. 1981. Mandarin Chinese: 3 functional
reference grammar. Berkeley: University of California Press.

Li, Charles, Sandra Thompson, and R. McMillan Thompson. 1982. The
discourse motivation for the perfective aspect: the Mandarin particle
Le. Tense-aspect: between semantics & pragmatics, ed. by Paul J.
Hopper. Amsterdam: John Benjamins.

Link, Godehard. 1983. The logical analysis of plurals and mass terms: a
lattice-theoretical approach. Meaning, use, and interpretation of
Language, ed. by R. Bauerle, C. Schwarz, & A. von Stechow, 302-
323. Berlin/New York: Walter de Gruyeter.

Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press.
McCawley, James. 1971. Tense and time reference in English. Studies in

linguistic semantics, ed. by C. J. Fillmore and D. T. Langendoen, 96-
113. New York: Holt, Rinehart and Winston.

 

234

McCawley, James. 1988. The syntactic phenomena of English, vol. 1.
Chicago: University of Chicago Press.

McCoard, Robert. 1978. The English perfect: tense-choice and pragmatic
inferences. Amsterdam: North-holland.

McGilvray, James. 1991. Tense, reference, & worldmaking. Montreal &
Kingston: McGill—Queen's University Press.

Mellor, D. H. 1981. Real Time. Cambridge: Cambridge University Press.

Merrill, Peter. 1985. Universal quantification and aspect in Russian. Issues
in Russian morphosyntax, ed. by Michael S. Flier and Richard D.
Brecht, 58-72. Columbus, OH: Slavica publishers.

Meulen, Alice ter. 1980. Substances, quantities, and individuals: a study in
the formal semantics of mass terms. Bloomington: Indiana
University Linguistics Club

Meulen, Alice ter. 1981. An intensional logic for mass terms. Philosophical
studies. 40.105-125.

Meulen, Alice ter. 1983. The representation of time in natural language.
Studies in modeltheoretic semantics, ed. by Alice ter Meulen, 177-
192. Dordrcht: Foris.

Meulen, Alice ter. 1984. Events, quantities and individuals. Varieties of
Formal semantics, ed. by Fred Landman and Frank Veltman, 259-
280. Dordrecht: Foris.

Mittworth, Anita. 1987. Aspects of English aspect: on the interaction of
perfect, progressive and durational phases. Linguistics and
philosophy 11.203-254.

Moltmann, Friederike. 1992. Measure adverbials. Linguistics and
philosophy 14. 629-660.

Montague, Richard. 1970. Pragmatics and intensional logic. Formal
philosophy, ed. by Richmond Thomason, 119-147. New Haven: Yale
University Press.

235

Montague, Richard. 1973. The proper treatment of quantification in
ordinary English. Formal philosophy, ed. by Richmond Thomason,
119-147. New Haven: Yale University Press.

Moravcsik, Julius. 1973. Mass terms in English. Approaches to natural
language, ed. by J. Hintikka, J. Moravcsik and P. Suppes. Dordrecht:
Reidel.

Mourelatos, Alexander. 1981. Events, progresses, and states. Syntax and
semantics: tense and aspect, ed. by Philip Tedeschi & Annie Zaenen,
191-212. London: Academic Press.

Newmeyer, Frederick. 1986. Linguistic theory in America. New York:
Academic Press.

Newton-Smith, W. H. 1980. The structure of time. Boston: Routledge &
Kegan Paul.

Orwig, Carol. 1991. Relative time reference in Nugunu. Tense and aspect
in eight languages in Cameroon, ed. by Stephen Anderson and
Bernard Comrie, 147-62. Texas: the Summer Institute of Linguistics
and University of Texas at Arlington.

Parsons, Terence. 1970. Analysis of mass and amount terms. Foundations
of language 6.362-388.

Parsons, Terence. 1980. Modifiers and quantifiers in natural languages.
Canadian journal of philosophy, supplementary vol. 6. 29-60.

Parsons, Terence. 1985. Underlying events in the logical analysis of
English. Actions and events: perspectives on the philosophy of
Donald Davidson, ed. by Ernest Lepore and Brian Mclaughlin, 235-
267. Oxford: Blackwell.

Parsons, Terence. 1989. The progressive in English: events, states and
process. Linguistics and philosophy 12. 213-241.

Parsons, Terence. 1990. Events in the semantics of English: A study in
subatomic semantics. Cambridge, Mass: MIT Press.

Partee, Barbara. 1973. Some structural analogies between tenses and
pronouns in English. Journal of philosophy 69.601-690.

 

 

236
Prior, Arthur. 1957. Time and modality. Oxford: Clarendon Press.

Prior, Arthur. 1967. Past, present and future. Oxford: Clarendon Press.
Prior, Arthur. 1968. Papers on time and tense. Oxford: Clarendon Press.

Reichenbach, Hans. 1947. Elements of symbolic logic. New York:
Macmillan.

Rescher, Nicholas. (ed.) 1968. Aspect of action. The logic of decision and
action, ed. by Nicholas Rescher, 215—219. Pittsburgh: University of
Pittsburgh Press.

Richards, Barry. 1982. Tense, aspect and time adverbials I. Linguistics and
philosophy 5. 58-107.

Russell, Bertrand. 1956. The philosophy of logical atomism. Logical and
Knowledge, ed. by Robert Charles Marsh, 175-282. New York:
Macmillan.

Ryle, Gilbert. 1949. The concept of mind. London: Hutchinson's
University library.

Scha, Remko. 1984. Distributive, collective and cumulative quantification.
Truth, interpretation and information, ed. by Jeroen Groenendijik,
Theo M. V. Janssen, and Martin Stockhof, 131-158. Dordrecht:

Foris Publications.

Smith, Carlota. 1980. Temporal structures in discourse. Time, tense, and
quantifiers: proceedings of the Stuttgart conference on the logic of
tense and quantification, ed. by Christian Rohrer, 355-375. Tiibingen:
Niemeyer.

Smith, Carlota. 1981. Semantic and syntactic constraints on temporal
interpretation. Syntax and semantics: tense and aspect, ed. by Philip
Tedeschi & Annie Zaenen, 213-238. London: Academic Press.

Smith, Carlota. 1991. The parameter of aspect. Norwell, MA: Kluwer
Academic Publishers.

Tai, James. 1985. Temporal sequence and Chinese word order. Iconicity in
syntax. ed. by John Haiman, 49-72, Amsterdam/Philadelphia: John
Benjamins.

237

Taylor, Barry. 1977. Tense and continuity. Linguistics and philosophy 1.
199-220.

Taylor, Barry. 1985. Modes of Occurrence. Oxford: Basil Blackwell.

Thelin, Niles. (ed.) 1990. Verbal aspect in discourse: Contributions to the
semantics of time and temporal perspective in Slavic and non-Slavic
languages. Amsterdam/Philadelphia: John Benjamins.

Thomason, Richmond. (ed.) 1974. Formal philosophy: selected papers of
Richard Montague. New Haven: Yale University Press.

Tommola, Hannu. 1990. On Finnish 'aspect' in discourse. Verbal aspect in
discourse: contributions to the semantics of time and temporal
perspective in Slavic and non-Slavic languages, ed. by Niles Thelin,
349-64. Amsterdam/Philadelphia: John Benjamins.

Vendler, Zeno. 1957. Verbs and times. The Philosophical review. 66.143-
160.

Vendler, Zeno. 1967. Linguistics in philosophy. Itbaca, NY: Cornell
University Press.

Verkuyl, Henk. 1989. Aspectual classes and aspectual composition.
Linguistics and philosophy 12.39-94.

Vet, Co. 1983. From tense to modality. Studies in modeltheoretic
semantics, ed. by Alice ter Meulen, 193-206. Dordrcht: Foris.

Vlach, Frank. 1981. The Semantics of the progressive. Syntax and
semantics: tense and aspect, ed. by Philip Tedeschi & Annie Zaenen,
271-292. London: Academic Press.

Vuyst, Jan de. 1983. Situation-descriptions: temporal and aspectual
semantics. Studies in modeltheoretic semantics, ed. by Alice ter
Meulen, 161-176. Dordrcht: Foris.

Wang, Li. 1985. Modern Chinese grammar. Beijing: Commercial Press.

 

 

238

Zhou, Minglang. 1991. Count nouns used as mass nouns in perfective
sentences in Chinese: 3 study of verb category switch and and the
'imperfective paradox'. WECOL91, ed. by Katharine Hunt, Thomas
Perry and Vida Samiian, 320-31. Fresno, CA: California State

University.

Zhou, Minglang. 19933. A lexical and cognitive view on language change:
the relation between some predicate argument structure variations
and change in Chinese. Paper presented at Kentucky Foreign
Language Conference, University of Kentucky, Lexington, KY,
April 22-24, 1993.

Zhou, Minglang. 1993b. The bei-construction in Chinese: 3 construction
with multistage properties. The proceedings of the 1992 Mid-
America linguistics conference & conference on Siouan/Caddoan
languages, ed. by Evan Smith and Flore Zéphir, 287-301. Columbia,
MO: University of Missouri-Columbia.