(Synthese Language Library 25) Gregory T. Stump (Auth.) - The Semantic Variability of Absolute Constructions-Springer Netherlands (1985)

THE SEMANTIC VARIABILITY OF ABSOLUTE CONSTRUCTIONS
SYNTHESE LANGUAGE LIBRARY
TEXTS AND STUDIES IN

LINGUISTICS AND PHILOSOPHY
Managing Editors.
ELiSABET ENGDAHL, University of Wisconsin
JAAKKO HINTIKKA, Florida State University, Tallahassee
ST ANLEY PETERS, Stanford University
Editorial Board:
EMMON BACH, University of Massachusetts at Amherst
JOAN BRESNAN, Stanford University
JOHN LYONS, University of Sussex
JULIUS M. E. MORAVCSIK, Stanford University
PATRICK SUPPES,Stanford University
DANA SCOTT, Carnegie-Mellon University, Pittsburgh
VOLUME 25
GREGORY T. STUMP
Department of English, University of Kentucky
THE SEMANTIC
VARIABILITY
OF ABSOLUTE
CONSTRUCTIONS
D. REIDEL PUBLISHING COMPANY

A MEMBER OF THE KLUWER ACADEMIC PUBLISHERS GROUP
DORDRECHT/BOSTON/LANCASTER
library of Congress Cataloging in Publication Data
Stump, Gregory T. (Gregory Thomas), 1954-
The semantic variability of absolute constructions.
(Synthese language library; v. 25)

Based upon the author's thesis (Ph.D.-Ohio State University,
1981) presented under title: The formal semantics and pragmatics'c;[free
adjuncts and absolutes in English.
Bibliography: p.
Includes indexes.
1. English language-Semantics. I. Title. II. Series.
PE1585.S665 1984 425 84-22277
ISBN-13: 978-90-277-1896-9 e-ISBN-13: 978-94-009-5277-5
DOl: 10.1007/978-94-009-5277-5
Published by D. Reidel Publishing Company,

P.O. Box 17,3300 AA Dordrecht, Holland.
Sold and distributed in the U.S.A. and Canada

by Kluwer Academic Publishers,
190 Old Derby Street, Hingham, MA 02043, U.S.A.
In all other countries, sold and distributed

by Kluwer Academic Publishers Group,
P.O. Box 322, 3300 AH Dordrecht, Holland.
All Rights Reserved

© 1985 by D. Reidel Publishing Company, Dordrecht, Holland
Softcover reprint of the hardcover I st edition 1985
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical
including photocopying, recording or by any information storage and
retrieval system, without written permission [rom the copyright owner
For Ma rcia
TABLE OF CONTENTS
PREFACE xiii
ACKNOWLEDGEMENTS xvii
I. THE SEMANTIC VARIABILITY OF FREE ADJUNCTS AND

ABSOLUTES
1. Introduction to Free Adjuncts and Absolutes in
English 4
1.1. The Free Adjunct Construction 4
1.2. The Nominative and Augmented Absolute
Constructions 8
2. Traditional Thoughts on the Semantic
Variability of Free Adjuncts and Absolutes 14
3. Plan of Discussion 22
4. Some Syntactic Conventions 27
4.1. Binary Features 27
4.2. Morphological Functions 29
4.3. Main Verb Marking 31
4.4. Subject Marking 32
4.5. Syntactic Subroutines 32
Footnotes 36
II. MODALITY AND THE INTERPRETATION OF FREE ADJUNCTS 41
1. The Semantic Bifurcation of Free Adjuncts in
Modal Contexts 42
2. Explaining the Entailment Properties of Strong
and Weak Adjuncts in Modal Contexts 44
2.1. Kratzer's Theory of Conditional Modality 45
2.2. The Roles of Strong and Weak Adjuncts
in Modal 3entences 53
2.2.1. Weak Adjuncts in Modal Contexts 56
2.2.2. 3trong Adjuncts in Modal contexts 58
viii TABLE OF CONTENTS
3. A Semantic Correlate of the Distinction

between Strong and Weak Adjuncts 64
3.1. Carlson's Ontology of Stages and
Individuals 66
3.2. A Categorial Distinction between Stage-
level and Individual-level Predicates 71
3.2.1. Be 1 f PIVi/PREDs 73
3.2.2. Be2 f PIVi/PREDi 75
3.2.3. Be 3 f PIVs/PREDi 76
79
3.3. The Stage/Individual Distinction as
a Semantic Correlate of the
Weak/Strong Distinction 84
3.4. Adjuncts with As and With 86
4. Chapter Summary 89
Footnotes 90
III. TENSE AND THE INTERPRETATION OF FREE ADJUNCTS 97

1. Preliminaries 100
1.1. Interval Semantics 101
1.2. Temporal Abstracts 103
1 .2.1. Rules of Tense 105
1.2.2. Shifting and Nonshifting Tenses 108
1.3. Two Categories of Time Adverbs 116
1.3.1. Time Adverbs of Category TA 116
1.3.2. Time Adverbs of Category MTA 117
1.4. The Semantics of Temporal Adverbial
Clauses 121
1.4.1. The AnalYSis of Tense in Temporal
Adverbial Clauses 123
1.4.2. Unbounded Dependencies in Temporal
1.4.3. Temporal Adverbial Clauses with
Main Tense Adverbs 137
1.4.4. Remarks on Some Unacceptable
Temporal Adverbials 144
1.4.5. Remarks on the Assumed
Truthconditions for Temporal
1.4.5.1. When 148
1.4.5.2. While 156
1.4.5.3. Before 158
TABLE OF CONTENTS ix
1.4.5.4. ACter 159

1.5. Summary of Tense and Time Adverb System 160
2. The Temporal Reference of Free Adjuncts 162
2.1. A-Abstracts 163
2.2. Deriving Conditional Adjuncts 165
2.3. Deriving Adsentential Adjuncts 166
3. Frequency Adverbs and the Distinction between
Strong and Weak Adjuncts 172
3.1. The Semantics of Relative Frequency
Adverbs 173
3.2. Adjuncts Restricting the Interpretation
of Relative Frequency Adverbs 184
4. A Generalization Operator 190
4.1. Two Sorts of Interpretations for
Temporal Adverbial Clauses 192
4.2. Generalization Operators 193
4.3. Adjuncts Restricting the Generali-
zation Operator G" 203
Footnotes 209
IV. ASPECT AND THE INTERPRETATION OF FREE ADJUNCTS 220

1. The Perfect Tense and the Interpretation of
Free Adjuncts 222
1.1. The Semantic Unspecificity of the Perfect 223
1.2. The Formal Semantics of the Perfect in
Finite Clauses 236
1.3. The Formal Semantics of the Perfect in
Free Adjuncts 242
2. An Argument for Free Adjuncts as Main Tense
Adverbs 247
3. The Progressive Aspect and the Interpretation
of Free Adjuncts 253
3.1. The Semantics of Present Participial
Phrases 255
3.2. The Formal Semantics of Present
Participial Adjuncts 261
Footnotes 265
x TABLE OF CONTENTS
V. THE FORMAL SEMANTICS OF ABSOLUTES 270

1. Modality and the Interpretation of Absolutes 272
1.1. Weak and Strong Absolutes 272
1.2. Formalizing the Distinction Between
Strong and Weak Absolutes 277
1.3. Deriving Conditional Absolutes 279
2. Tense and the Interpretation of Absolutes 280
2.1. Absolutes and Relative Frequency Adverbs 281
2.2. Absolutes and the Generalization Operator
~ ~3
3. Absolutes as Main Tense Adverbs 285

Footnotes 290
VI. INFERENCE AND THE LOGICAL ROLE OF FREE ADJUNCTS

AND ABSOLUTES 299
1. Summary of the Proposed Semantic Analysis of
Free Adjuncts and Absolutes 300
2. The Role of Inference in the Interpretation of
Free Adjuncts and Absolutes 303
2.1. Two Kinds of Inferences 304
2.2. Inferences, Adjuncts, and Absolutes 307
2.2.1. Carlson's Ontology 308
2.2.2. Instantaneousness 318
2.2.3. Hord Order 321
2.2.4. Knowledge of the World 321
2.2.5. Connective Adverbs 322
2.3. Pragmatically Admissible Values for Land
H 325
3. On the Possibility of Deriving Absolute
Constructions from Adverbial Subordinate
Clauses 329
4. On the Possibility that the Logical Role of an
Absolute Construction is Always Inferred 335
5. Theoretical Implications 340
Footnotes 344
TABLE OF CONTENTS xi
APPENDIX - A FORMAL FRAGMENT FOR FREE ADJUNCTS AND

ABSOLUTES 350
1. Intensional Logic 350
2. Syntax and Translation Rules for a Fragment of
English 360
2.1. Syntax 360
2.2. Translation 377
REFERENCES 390
INDEX OF NAMES 396
GENERAL INDEX 398
PREFACE
The goal of this book is to investigate the semantics of

absolute constructions in English; specifically, my object
is to provide an explanation for the semantic variability of
such constructions. As has been widely noted in traditional
grammatical studies of English, free adjuncts and absolute
phrases have the ability to playa number of specific
logical roles in the sentences in which they appear; yet,
paradoxically, they lack any overt indication of their
logical connection to the clause which they modify. How,
then, is the logical function of an absolute construction
determined? In attempting to answer this question, one must
inevitably address a number of more general issues: Is the
meaning assigned to a linguistic expression necessarily
determined by linguistic rules, or can the grammar of a
language in some cases simply underdetermine the
interpretation of expressions? Are the truthconditions of a
sentence ever sensitive to the inferences of language users?
If so, then is it possible to maintain the validity of any
really substantive version of the Compositionality
Principle? These are, of course, issues of great inherent
interest to anyone concerned with the formal syntax and
semantics of natural language, with the philosophy of
language, or with language processing.
The descriptive framework assumed throughout is the
semantic theory developed by Richard Montague (1970a, 1970b,
1973) and his followers. (For a very thorough introduction
to Montague semantics, the reader may refer to Dowty, Wall
and Peters (1981 ).) Montague' s theory, unlike the semantic
theories most familiar to lingui'sts, is a theory of
referential rather than psychological meaning: it
associates liaguistic expressions directly with their
denotations, without regard to processes of language
comprehension; such a theory lays no claim to psychological
reality, but affords a model-theoretic account of such
important notions as truth and entailment. In recent years,
it has nevertheless become clear that issues of language
xiii
xiv PREFACE
comprehension are not without relevance for referential

theories of meaning. For example, Putnam (1975) has
convincingly argued that a complete account of lexical
meaning must comprise complementary referential and
psychological components; and Partee (1979:10) has shown
that "the linguist's concern for psychological
representation may be relevant to every semanticist's
concern for an account of the sewantics of propositional
attitudes".
The analysis which I propose here provides further
evidence of the relevance of matters of language
comprehension for the referential conception of meaning. I
argue that the logical role played by an absolute
construction is in some cases wholly or partially determined
by the semantic properties of an accompanying modal or
temporal operator, but that in other cases--in fact, in the
majori ty of cases--the logical connection between an
absolute construction and the clause which it modifies is
determined only by the inferences of language users. In
such instances, the interpretation assigned to the sentence
by the grammar of English is indeterminate between a certain
range of choices; this referential indeterminacy is resolved
by nonlinguistic--inferential--means. As I shall show. a
number of diverse factors may give rise to such inferences.
This work began as my doctoral dissertation, ~ Formal
Semantics and Pragmatics of Free Adjuncts and Absolutes in
English (Ohio state University, 1981). A considerable
number of changes and improvements have, however, been
incorporated into the present version, including: a number
of formal ameliorations in the proposed f10ntague fragment;
an extension of the proposed analysis to account for the
semantic peculiarities of free adjuncts headed by as and
with; mention of relevant literature which has appeared
since 1981; and a number of expository improvements.
This work was completed during the summer of 1984 at
the University of Kentucky. I wish to thank my colleagues
in the Department of English for providing me with both an
ideal environment for finishing the book and the resources
and practical advice necessary for producing the camera-
ready manuscript; particular thanks go to Walt Foreman, Bob
Hemenway, Kevin Kiernan, Armando Prats, and Greg Waller.
lowe a great debt of gratitude to many people for
their willingness to discuss the issues addressed here. I
am especially indebted to Greg Carlson, David Dowty, Doug
Fuller, Mike Geis, Brian Joseph, John Nerbonne, Barbara
PREFACE xv
Partee, Mehmet and Feryal Yava§, Arnold Zwicky, and an

anonymous referee; also to Seiko Fukuchi ,f'Or kindly
supplying me with relevant literature not otherwise
available to me.
Special thanks to Martin Scrivener at Reidel for his
invaluable assistance.
Most of all, I wish to thank my wife, Marcia Hurlow,
whose contributions to the completion of this study are
beyond counting.
August, 1984 G. T. S.
The University of Kentucky
ACKNOWLEDGEMENTS
Excerpts from Elizabeth Drew, Portrait of an Election:

copyright © 1981 by Elizabeth Drew, InC.; reprinted by
permission of Simon & Schuster, Inc. All excerpts
originally appeared in The New Yorker: 9/8/80, pp. 41-
42, 46, 50, 68, 69, 77,80:-90, 99: 9/29/80, pp. 107,
119.
Excerpts from Robert van Gulik, The Phantom of the Temple
(Charles Scribner's Sons); copyright © 1966~bert H.
van Gulik; reproduced here by kind permission of Charles
Scribner's Sons.
Excerpts from Robert van Gulik, Poets and Murder (Charles
Scribner's Sons); copyright © 1968~obert van Gulik;
reproduced here by kind permission of Charles Scribner's
Sons.
Sentence taken from a Johnson's Dental Floss package;
copyright © 1980, Johnson & Johnson Products, Inc.;
reproduced here by kind permission of Johnson & Johnson
Products, Inc.
Reprinted by permission of Farrar, Straus and Giroux, Inc.
Excerpts from Oranges by John McPhee. Copyright
© 1966, 1967 by John McPhee. Excerpts from The Crofter
and the Laird by John t-lcPhee. Copyright 1969, 1970 by
John McPhee. The text of both books appeared originally
in The New Yorker.
Excerpts from Whitney Balliett, 'Mingus,' The New Yorker
6/18/79, p.100; 'Newport News,' The New YMker 7/16/79,
pp.74, 80, 81; 'Number Twenty-six,' The New Yorker
7/21/80, pp.90, 91, 92; copyright © 1979~980, The
New Yorker; reproduced here by kind permission of The New
Yorker and Whitney Balliett.
xvii
xviii ACKNOWLEDGEMENTS
Excerpt from Brendan Gill, 'The Horizontal City,' The New

Yorker 9/15/80, p.l09; copyright © 1980, The New Yorker;
reproduced here by kind permission of The New-rDrker and
Brendan Gill.
Excerpts from Emily Hahn, 'Eleventh Hour,' The New Yorker
9/1/80, pp.38, 55, 59; copyright © 1980, The New Yorker;
reproduced here by kind permission of The New-yQrker and
Emily Hahn.
Excerpts from Pauline Kael, 'Master Spy, Master Seducer,'
The New Yorker 8/4/80, p.68; 'Australians,' The New
YC>rker-9/15/80, p.154; 'The Man Who Made Howard Hughes
Sing and The Iron-Butterfly Mom,' The New Yorker
10/13/80, pp.174, 189; copyright ©--1980,- The New
Yorker; reproduced here by kind permission of The New
Yorker and Pauline Kael.
Excerpts from Calvin Tomkins, 'Autobiography,' The New
Yorker 6/30/80, pp.57, 60; copyright © 1980, The New
Yorker; reproduced here by kind permission of The New
Yorker and Calvin Tomkins.
Excerpts from Calvin Trillin, 'Regional Thoughts from Atop
the Gateway Arch,' The New Yorker 6/16/80, pp.l04, 107,
108, 109; copyright ~~80, The New Yorker; reproduced
he-re by kind permission of The New Yorker and Calvin
Trillin.
Excerpts from a brief review of Death Drop (by B. M. Gi 11),
The New Yorker 9/1/80, p.92; 'Rafting-Up,' The New Yorker
9/15/80, p.36j brief review of Island on the Edge of the
World: The Story of St. Kilda (by Charles Maclean)," The
New Yor"'k'er 10/13/80;-P.194j copyright © 1980, The New
YC>rkerj reproduced here by kind permission of The New
Yorker.
Sentence excerpted from Allan J. Mayer et al., 'A Tide of
Born-Again POlitics,' Newsweek, 9/15180, p.31; copyright
© 1980, Newsweek, Inc.; reproduced here by kind
permission of Newsweek, Inc.
Two sentences excerpted from the Ohio State University
Lantern, 8/26/80, pp.3,8; copyright © 1980, The Ohio
state University Lantern; reproduced here by kind
permission of the Ohio state University Lantern.
ACKNOWLEDGEMENTS xix
Sentence excerpted from The Oregonian, 9/14/80, p.C4,

reproduced here by kind permission of The Oregonian,
Portland, Oregon.
Sentence taken from the instructions to Pressman's Stack-
Ominos game; copyright © 1979, Pressman Toy Corporation;
reproduced here by kind permission of Pressman Toy
Corporation.
Excerpts from Robert J. Seidman, One Smart Indian (The
Overlook Press); copyright © 1977, Robert J. Seidman;
reproduced here by kind permission of G. P. Putnam's Sons
and Robert J. Seidman.
Sentence excerpted from Woodall's Trailer and RV Travel/West
12/77, p. 11; copyright © 1977, Woodall'S Trailer and RV
Travel; reproduced here by kind permission of WoO"dall'S
Trailer and RV Travel.
C HAP T E R I
THE SEMANTIC VARIABILITY

OF FREE ADJUNCTS AND ABSOLUTES
My purpose in this book is to investigate the semantic

properties of absolute constructions, under which heading
are subsumed the three constructions in (1).
(1) a. The free adjunct construction:
Walking home, he found a dollar.
b. The nominative absolute construction:
His Cather being a sailor, John knows all

about boats.
c. The augmented absolute construction:
With the children asleep, Mary watched TV.
These constructions have the bross syntactic characteristics

of adverbial subordinate clauses, but they are distinguished
by a special peculiarity: their logical connection with the
clause they modify is not overtly specified. This
peculiarity is obvious in the case of the free adjunct and
nominative absolute constructions, which lack any sort of
subordinating conjunction. The augmented absolute in (1c)
is, of course, headed by the preposition with; but the
contribution which this preposition makes to the
interpretation of (lc) is so vague and uncertain that it can
scarcely be viewed as an 'overt specification' of the
logical role of the absolute phrase.
Despite the fact that they lack any overt indication of
their logical function, absolute constructions may
nevertheless be felt to play very specific roles in the
sentences in which they appear. This paradoxical fact is
well illustrated by the free adjuncts in (2)_(5).1 (Here
and throughout the chapter, the use of boldface characters
to highlight a free adjunct or absolute is mine.)
2 CHAPTER I
(2) The school is determined to avoid a scandal. The

father is equally determined to find somebody to
blame. The reader, being more experienced in such
things, knows the truth: it was murder.
(NY 9/1/80, 92)
(3) Grabbing a newspaper from a guard, Tom went back

out, wiped up the dog shit and deposited it and
the day's news in a refuse can.
(OSI, 245)
(4) Transposed to a trumpet or saxophone, her

creations would probably herald a new school.
(NY 7/16/79,74)
(5) Dead, they have taught the living.

(Inscription at Lexington Cemetery in memory
of people who willed their bodies to the
University of Kentucky Medical Center)
The logical roles played by the free adjuncts in these

sentences are surprisingly diverse: the adjunct in (2)
intuitively serves as an adverb of causation or
explanation--because slhe is more experienced in such
things; that in en, as some sort of time adverbial--af'ter
he grabbed a newspaper from a guard, perhaps; that in (4),
as a conditional clause--if they were transposed to a
trumpet or saxophone; and that in (5), as an adversative
clause--though they are dead. Despite the absence of any
overt subordinating conjunction in these sentences, users of
English have no trouble at all picking out the logical
function of their free adjuncts. Moreover, the four logical
functions represented in (2)-(5) do not exhaust the
possibilities; further examples would reveal that free
adjuncts may play other logical roles as well. Analogous
remarks can be made about nominative and augmented
absolutes; for example, the absolute phrase in (6) is felt
to function as an adverb of causation or explanation, while
that in (7) intuitively serves as a time adverbial.
(6) Picasso, to himself and to many others~ was a god.

That being the case, he could afford to be a
monster.
(NY 6/30/80, 60)
SEHANTIC VARIABILITY 3
(7) Reginald, his breakfast finished, took his daily
morning stroll in Fleet Street.
(Visser 1972:1268)
As in the cases in (2)-(5), English speakers have no

difficulty naming the logical relation holding between the
absolute and the main clause in (6) and (7). This property
of free adjuncts and absolutes--the ability to play the role
of a number of different sorts of adverbial clauses--
constitutes their semantic variability.
My goal here is to investigate this special property of
absolute constructions, and to address the problems of
analysis which it raises. For example, what sort of formal
account can be provided for the semantic variability of
absolute constructions? How are the various logical roles
which English speakers attribute to the constructions in
(1 )-(7) determined? Are they uniquely determined by
linguistic rules, or are they the effect of non linguistic
inferences? The theoretical interest of these questions is
great, for they inevitably lead to a number of more
fundamental issues: Can the meaning of an expression depend
in a significant way on the contributions of language users'
inferences? If so, what are the implications of this fact
for the strict compositionality of Montagovian semantic
theory? These and other issues will be taken up in the
following chapters, where a detailed formal analysis of the
semantic variability of absolute constructions is developed.
A number of preliminary matters are discussed in the
present chapter. I begin with a brief summary of the
defining properties of free adjuncts and absolutes in
English (section 1). In section 2, I review the thoughts of
several traditional grammarians regarding the semantic
variability of absolute constructions; specifically, I
survey a number of opinions regarding (i) the range of
logical roles which an adjunct or absolute may assume, and
(ii) the factors that determine which of these logical roles
an adjunct or absolute will play in a given case. In
section 3, I briefly outline the analysis to be presented in
the succeedinb chapters. At the core of the proposed
analysis is a Hontague fragment for absolute constructions
(which is stated formally in the Appendix); in section 4, I
introduce some formal syntactic conventions to be employed
in this fragment.
4 CHAPTER I
1. Introduction to Free Adjuncts and Absolutes

in English
Here, I briefly revi~w the defining characteristics of the

free adjunct and absolute constructions in modern English.
1.1. THE FREE ADJUNCT CONSTRUCTION
A free adjunct is a nonfinite predicative phrase with the

function of an adverbial subordinate clause; it is typically
set off from the clause to which it is subordinate by a
pause or a fall in intonation (sometimes--though not
consistently--represented in writing with commas). Free
adjuncts are used very commonly in English, especially in
more formal speech and writing;2 they are extremely
heterogeneous from the point of view of both form and
interpretation 3 •
A free adjunct may be headed by a noun, an adjective, a
prepositional phrase, or a verb:
(8) A center for shoe factories and breweries early in

this century, it was industrialized at a time when
the cities west of it were still tied to the land.
(NY 6/16/80, 107)
(9) Unable to meet his eyes, Kate looks down at her

hands •••
(NY 8/4/80, 68)
(10) The Second World War began, and, still scarcely in

his teens, he was drafted into the coal mines of
Fife and Kent.
(CL, 19-20)
(11) The W. S. Q. imitated the Ellington reed section,

uaing a syrupy vibrato •••
(NY 7/16179, 80)
In the following chapters, attention will be concentrated

largely (though not exclusively) on verbal and adjectival
free adjuncts, since they pose the most significant problems
of analysis.
The main verb or head of a verbal free adjunct may be a
present participle, a past participle, or a marked
SEMANTIC VARIABILITY 5
infinitive form:
(12) Glancing up at the Arch over the downtown skyline

as I approached, I had to admit that it is an
impressive structure_.
(NY 6/16/80, 109)
(13) Published here in 1972, Thomas Keneally's novel is

no longer in print •.•
(NY 9/15/80, 154)
(14) To tell you the truth, I have never really thought

of them that way.
(CL, 73)
A verbal adjunct may be in the active voice, as in (12),

(14), and (16), or the passive, as in (13) and (15).
(15) ••• it was the home of a hardy race of Gaelic-

speakers from prehistoric times to 1930, when--
having been systematically destroyed or
demoralized by the zealotry of Christian
-missionaries,- the myriad diseases of
civilization, and the steamroller thrust of a
superior technology--the remnant survivors were
evacuated by a paternal British government.
(NY 10/13/80, 194)
(16) Arrived at the spot, the party lost no time in

getting to work.
(Visser 1972:1255)
It may show perfect 'aspect' alone, as in (17); or the

perfect combined with the progressive, as in (18) (thanks to
Doug Fuller for pointing out the latter example).
(17) Having come to this conclusion, I settled down to

enjoy the rest of the day.
(NY 9/1/80, 55)
(18) He was then drunk, having been all night taking

his leave.
(Scheffer 1975:253)
The progressive may not occur alone, as (19) suggests: this

6 CHAPTER I
might be attributed to Ross' (1972) Doubl-ing Constraint;4

another explanation will be noted in Chapter IV, where the
broader problems posed by the perfect and the progressive in
free adjuncts will be discussed.
(19) *Being eating, John couldn't answer.
Free adjuncts may be dependent upon a subordinate

clause:
(20) In the 1935 etching called "Minotauromachy~ the
monster seems to look for guidance to a little
girl who faces him fearlessly, holding a candle
aloft in one band and a buncb of flovers in tbe
otber.
(NY 6/30/80, 57)
(21) But when the man, baving backtracked, picked up

her glove. and put it on, touches her shoulder
with his gloved hand, she is so startled she bolts
away.
(NY 8/4/80, 68)
(22) He was called by Strauss, by Jordan, and by the

President, and then he decided that, baving given
it full consideration, he was not for an "open
Convention."
(NY 9/8/80, 68)
They normally occur either immediately before or immediately

after their superordinate clause, or after its subject,
provided this is not a pronoun; the order chosen is
sometimes iconic, but needn't be--
(23) He entered college at the age of fifteen,

graduating four years later at tbe bead of bis
class.
(Jespersen 1940:407)
(24) He occasionally lectured night-club audiences ten

or fifteen minutes at a stretch, issuing a mixture
of invective. musical pedagogy. autobiograpby. and
home.ade pbilosophy.
(NY 6/18179, 100)
The 'subject position' of a free adjunct is commonly

controlled by the subject of the superordinate clause; in

such an instance it is termed a related free adjunct, as in
all of (8)-(24) except (15). But frequently control of the
subject position of a free adjunct rests elsewhere--with a
nonsubject noun phrase in the superordinate clause or with
an extralinguistic controller; in this case the adjunct is
termed unrelated or (more prescriptively) misrelated. (Such
is the 'dangling participle'.) Besides (15), examples of
unrelated free adjuncts are:
(25) Having received only an elementary education, the

simple teachings and the colourful ritual had
appealed to her.
(PT, 76)
(26) Having done either of' the above his turn is ended
unless he uses all tiles on his rack ••• or if he
accumulates 9 points on one stack
(Instructions to stack-ominos 1M , Pressman Toy
Corporation)
(27) Her mother too had expressed incredulity, but

being the sort of' woman she was, that was only to
be expected.
(MMB, 24)
(28) I want them to vote their conscience on the great

issues of the platform and then on the nomination,
and I'm still very confident that, voting on their
conscience, we'll gain the nomination and go on.
(Ted Kennedy, cited in NY 9/8/80, 69)
(29) Considering the enemies of' citrus as a whole, it

is no wonder that the trees people keep in their
yards for ornamental effect seemed--to me, at
least--to be the sorriest-looking trees in the
state.
(Q, 43)
(30) A number of people, including some of Kennedy's

friends, and even some of his staff, have become
puzzled about what he thinks he is doing. Leaving
aside the style in which he is doing it, the
question is not so puzzling.
(NY 9/8/80, 46)
8 CHAPTER I
(31) The laird simply has no wish to be the animated

exchequer of an insular, private, picayune welfare
state--or, to give the situation its fUll setting,
a welfare state within a welfare state.
(CL, 110)
(32) The Carter people insisted then, as they insist

now, that the interest groups assembled were not
and are not reflective of the Party as a whole, to
say nothing of the country.
(NY 9/8/80, 80)
(33) now it has almost three million, making it the

largest city in the country after New York and
Chicago.
(NY 9/15/80, 109)
<34 ) Wit h d ail y flo s sin g , b 1 e e din g s h 0 u 1 d s top,

indicating that gums are getting healthier.
(Johnson's Dental Floss package, Johnson &
Johnson, New Brunswick, New Jersey)
The problems of control in free adjuncts are fascinating and

not simple,5 but are, regrettably, outside the scope of the
present study; in the remaining chapters, I will be dealing
almost exclusively with free adjuncts under the control of
the subject of the clause upon which they are dependent
(more correctly, with those interpretations of free adjunct
sentences which entail that the adjunct is subject-
controlled) •
1.2. THE NOMINATIVE AND AUGMENTED ABSOLUTE CONSTRUCTIONS
The nominative absolute construction consists of a 'subject'

noun phrase combined with a nonfinite 6 predicative
expression, the whole functioning as an adverbial unit
subordinate to an associated main clause; like free
adjuncts, nominative absolutes are normally set apart
intonationally. In modern English, the free adjunct and
absolute constructions show a number of similarities; and
although the latter is seen as somewhat more literary, it is
often observed that it is formally just like the former
construction but with an overt subject supplied for the
'free' predicate. 7 The predicative constituent of an
absolute phrase may be headed b

adjective, or a prepositional phrase:
Ba noun, a verb, an
(35) Joe Turner, seated in a press of admirers and

backed by Jay McShann, sings "Roll 'Em Pete," his
great voice a thunderhead.
(NY 7/21/80, 91)
(36) and there was Chuck Green--a square, flapping

scarecrow moving in half time, his head bent
forward in concentration, his arms loose.
(NY 7/21/80, 92)
( 37 ) At f i v e - t h i r t yin Car neg i eRe c i t.a I Hall,

Dardanelle sang and played a dozen or so songs,
among them "It Could Happen to You.- "Spring Can
Really Hang You Up the Most,· "Out of This World,·
"It's All Right with He." and "In the Evening.-
(NY 7/21/80, 90)
In the discussion which follows, attention will be focused

primarily on verbal and adjectival absolutes--like verbal
and adjectival free adjuncts, they pose the most interesting
problems.
The head of the predicative constituent of a verbal
absolute may be a present participle,9 a past participle,
or, infrequently, a marked infinitive:
(35) The eggs and vermiculite are placed in an earthen

vessel, the eggs just touching each other, in case
there is some inter-egg communication at hatching
tiDe.
(NY 9/1/80, 59)
(39) Communist Party leader Edward Gierek, his power

battered by striking workers and a corruption
scandal, was ousted from office today and replaced
by Politburo member Stanislaw Kania, a surprise
choice.
(Ashland (Ore.) Daily Tidings 9/6/80, 1)
(40) We shall assemble at ten forty-five, the

procession to start at precisely eleven.
(Visser 1972:1056)
10 CHAPTER I
An absolute phrase shows the same range of voice and

'aspect' as a participial adjunct: it may be active (as in
(38» or passive (as in (39»; perfect, or, in principle,
both perfect and progressive--
(41) The Arch has encouraged some new building

downtown--the old downtown that it was presumably
designed to anchor having drifted out into the
county by the time the Arch was completed.
(NY 6/16/80, 108)
(42) John decided to see a doctor, his health having

been steadily deteriorating.
The progressive alone does not occur:
(43) *John decided to see a doctor, his health being

deteriorating.
Similarly, absolute phrases may be dependent upon embedded

clauses:
(44) Ayla's musculature allows her great skill with the

sling, and though she must practice secretly,
hunting being forbidden to women under penalty of
death, she quickly becomes a sharpshooter capable
of firing two lethal stones in rapid succession.
(The Oregonian 9/14/80, C4)
(45) I made a quick tour of the room, pausing for a few

minutes to observe the source of much noise in one
corner: a TV set turned on full blast, on whose
screen an earnest man was talking about the
nurture of radiated tortoises (Testudo radiata),
the closeup of his face now and then supplanted by
closeups of the tortoises themselves.
(NY 9/1/80, 38)
(46) Some of Chertok's clips were very funny, among

them ••• an eccentric dance done in blackface in
1929 by James Barton, his body vibrating one
moment from head to foot, his legs spaghetti the
next, his arms whirligigs the next ...
(NY 7/21/80,
38)
(In this last example, the boldface absolutes are actually

embedded within a larger absolute.) As these examples show,
absolute phrases may be as freely ordered as free adjuncts
with respect to their superordinate clause.
The nominative absolute construction is so named
because its subject noun phrase is in the nominative case
(although this is verifiably the case only when the subject
is a pronoun). Exceptions to this rule can be found, as in
the following striking example cited by Visser (1972:1148);
(47) ••• as we strode along, I doing my best to keep

pace with him, and him reading aloud from some
political economist or other, he would drag out a
handful of nuts and munch them.
but among speakers of standard English, absolutes whose

subjects are oblique in case are generally regarded as
unacceptable. As a matter of fact, absolutes with
pronominal subjects are in general quite restricted in their
use; they strike many speakers as being much more formal
than absolutes of other kinds.
One problem of terminology and analysis arising in
connection with both free adjuncts and absolutes is that of
distinguishing present participles from gerunds. The
distinction comes into question in cases in which a
preposition or subordinating conjunction introduces what
might be otherwise construed as a free adjunct:
(48) When (while) fighting in France he was taken

prisoner.
(Jespersen 1940:407)
(49) She looked pleadingly at her parents as though

entreating forgiyeness.
(Visser 1972:1138)
(50) After leaving Interstate 75, I noticed a sign on a

roadside eating place •••
(Q, 19)
(51) ••• I naturally tried to make my peace with the

Gateway Arch in the years after its completion,
despite having been handed a second bitter pill to
swallow •••
(NY 6/16/80, 104)
12 CHAPTER I
(In examples like these, the subordinating conjunction or

preposition 'fills in' the semantic relation holding between
the 'adjunct' and its superordinate clause; in true free
adjuncts, of course, this relation is not overtly
specified.) When what looks like a free adjunct is
introduced by a subordinating conjunction, as in (48) and
(49), the verbal element heading the 'adjunct' is
participial (Jespersen 1940:407; Curme 1931:276); but when
the supposed adjunct is introduced by a preposition, as in
(50) and (51), it must be considered a gerund. This
difference can be appreciated by considering the sorts of
expressions which are substitutable for the -ing phrases in
(48)-(51): the full range of predicative expressions-past
participles, adjective phrases, prepositional phrases,
predicative noun phrases-is substitutable for the present
participles in (48) and (49), while only nonpredicative noun
phrases are substitutable for the gerunds in (50) and (51);
compare (52) and (53).
(52) While drunk,

While at the beach,
1 h was taken prisoner.
While stranded in enemy territory, e
While president,
(53) *After drunk,

*After at the beach,
*After stranded in enemy territory, ?I noticed a sign on a
*After president, ~ roadside eating place.
After dawn,
Absolute expressions can be introduced by with and

without; the latter being prepositions, one would at first
suppose chanting and shouting and affecting 'im to be
gerunds in (54) and (55);
(54) With the reeds now chanting and shouting, Bowie

released an assemblage of blats and growls and
yells.
(NY 7/16179, 81)
(55) He co u 1 d put a wayan a wf u 1 lot wit h 0 u.t it

affecting lim.
(Visser 1972:1158)
but a range of predicative expressions is substitutable for

these forms (as, for instance, in (56», which must
therefore be recognized as participles. 10
(56) With the reeds now silent, }

With the reeds now united in harmony, Bowie released
With the reeds now in utter cacophony,
an assemblage of blats and growls and yells.
(Notice that constructions like those in (54) and (55)

cannot simply be regarded as free adjuncts consisting of a
(predicative) prepositional phrase whose object has a
postnominal modifier; this is because such expressions
needn't be 'controlled' by the subject of their
superordinate clause, as the full acceptability of sentences
like (57) and (58) shows.
(57) With the water so cold, it was unlikely that we

would be allowed to swim.
(58) With John at the wheel, there wouldn't have been

any problem.)
Since I wish to be precise about the difference between

the constructions in (48)-(56) and the free adjunct and
absolute constructions, I shall maintain the following
terminological distinctions: 'adjuncts' introduced by
subordinating conjunctions (as in (48), (49» will be termed
augmented adjuncts; tho&e introduced by prepositions (as in
(50), (51» gerunds; and absolutes introduced by
prepositions1~ (as in (54), (55», augmented absolutes. The
semantics and pragmatics of augmented adjuncts and gerund
constructions are not a central concern here, because they
lack the semantic variability that is the focus of this
study; augmented absolutes, on the other hand, will be given
full consideration in Chapter V, because they are as
semantically variable as 'unaugmented' absolutes; the
preposition with in an augmented absolute like that in (54)
does nothing to narrow down the range of logical roles which
it may assume.
14 CHAPTER I
2. Traditional Thoughts on the Semantic Variability
of Free Adjuncts and Absolutes
Despite the theoretical importance of the problem posed by

the semantic variability of free adjuncts and absolutes,
these constructions have remained neglected in the formal
syntactico-semantic literature until only recently.12
Traditional grammarians, on the other hand, have devoted
much attention to these constructions and their semantic
heterogenei ty. Two questions recurrently guide their
discussion: What is the range of logical roles which may be
assumed by adjuncts and absolutes (i.e. what is the range of
logical relations which they may bear to their superordinate
clause)? What determines which role a given free adjunct or
absolute will play? I shall briefly describe how these
questions have been answered by Curme (1931), Kruisinga
(1932), Jespersen (1940), Visser (1972), and Quirk,
Greenbaum, Leech, and Svartvik (1972).
In their accounts of the inventory of logical roles
which may be assumed by free adjuncts and absolutes,
tradi tional grammarians show a fair degree of agreement.
Regarding absolutes in particular, Curme (1931:154-157)
suggests that six broad logical roles can be distinguished,
according to whether an absolute is intuitively linked to
its superordinate clause by a relation of:
a) time, e.g.
My task having been finished, I went to bed;
(p.154)
b) cause, e.g.
The rain having ruined my hat, I had to get a
new one; (p.155)
c) condition and exception, e.g.
As yet few have done their full duty, present
company excepted; (p.155)
d) attendant circumstance, e.g.
He entered upon the new enterprise
cautiously, his eyes wide-open; (p.156)
e) manner proper, e.g.
He put on his socks wrong side out; (p.157)
or f) concession, e.g.
Granted the very best intentions, his conduct
was productive of great mischief. (p.157)
Since Curme actually takes (p.158) free adjuncts to be

abridged absolute phrases, his taxonomy of logical roles is

meant to hold for them as well as for unabridged absolutes.
Kruisinga (1932:274-275) subsumes the whole range of
logical roles assumed by free adjuncts under the notion of
attendant circumstances, observing that 'the use of free
adjuncts in English is greatly promoted by the almost
complete absence of conjunctions that can introduce a
subordinate clause expressing attendant circumstances.' He
reasons that when a free adjunct seems to bear a more
specific type of relationship to its superordinate clause,
it is simply because this relationship is among the more
salient circumstances attending the event described; 'other
circumstances, though in the background, are thought of'
(p.276). With this qualification, Kruisinga exemplifies
some of the more prominent circumstances which may emerge in
the interpretation of a free adjunct:
a) reason or cause, e.g.

Shy, reserved, and proud, I would have died
rather than have breathed a syllable of my
secret; (p.27n
b) difference of time, e.g.
Finding Blanche determined, Father Andre
presently took his leave; (1931: 68)
c) contrast, e.g.
Like all craftsmen of the kind, he is at the
mercy of his material, which, abundant enough
in some respects, is disappointingly scanty
where the matters most provocative of
curiosity are concerned; (p.278)
d) alternative circumstances, e.g.
The young cavalier perused that letter again
in memory. Genuine, or a joke of the enemy,
it spoke wakening facts to him. (p.279)
Kruisinga assumes (p.280) that absolute phrases are a type

of free adjunct, distinguished from related free adjuncts in
having a subject; therefore, his remarks on the logical
roles played by free adjuncts apply to absolute phrases as
well.
Jespersen (1940:61-64), while admitting that 'it is
often difficult or even impossible to draw sharp boundaries
between the several applications,' suggests that absolutes
can be seen to play four sorts of logical roles, according
as they are intuitively linked to their superordinate clause
16 CHAPTER I
by a relation of:
a) cause or reason, e.g.

The wise men of an t i qui ty ••• were afraid
that--men being what they are--their
discoveries might be put to bad or futile
uses; (p.62)
b) condition, e.g.
Conciliation failing, force remains; but
force failing, no further hope of
reconciliation is left; (p.62)
c) time, e.g.
And the meal being over, he took Mr. Kaye
into the other room; (p.63)
or d) descriptive circumstances, e.g.
He remained in town, his idea being that he
wanted everything settled before his
departure. (p.M)
(In his discussion of verbal free adjuncts (1940: 406-411),

Jespersen does not consider the range of logical roles which
these may play; presumably, these would be the same as those
assumed for the absolute construction.)
Visser (1972:1054-1056; 1132-1139; 1149-1158; 1252-
1255; 1266-1271) discusses the range of logical roles played
by free adjuncts and by absolutes separately. With regard
to free adjuncts, he tentatively distinguishes eight roles;
these correlate with the relations of
a) attendant circumstances, e.g.

So they wandered carelessly through the
farmland and along the beaches, being polite
to the French people they met; (p.1134)
b) cause, reason, e.g.
He was thirty-seven when he died, having been
born the year before my brother Germanicus;
(p.1135)
c) time, e.g.
The king, hearing this, was stupefact;
(p.1132)
d) condition, e.g.
This same thing, happening in wartime, would
amount to disaster; (p.1137)
e) concession, e.g.
He is the more to be admired, that, labouring
under such a difficulty, his verses are so
numerous, so various, and so harmonious;
(p.1137)
f) means, manner, e.g.
Brownie ••• was lanky and limp--so limp that
Mother said he was becoming unnaturally
elongated, pulling that cart up those hills;
(p.1139)
g) hypothetical similarity, e.g.
He paused as if expecting her to answer;
(p.1138)
and h) purpose, e.g.
We planted a hedge for preventing the cattle
from straying. (p.1139)
(All of Visser's examples of adjuncts expressing

hypothetical similarity are what I have called augmented
adjuncts (see section 1); all of his examples of adjuncts of
purpose are preposition-gerund combinations.) His
categorization for absolutes is similar; it is based on the
relations of
a) attendant circumstances, e.g.

The Benbergs stood by, he clasping his hands
and watching her closely, she wiping a
plate round and round with a sodden cloth;
(p.1152)
b) reason, ground, cause, or motive, e.g.
The bus being crowded, James had to stand;
(p.1155)
c) time, e.g.
Gabriel having finished, the huge pudding was
transferred to the table; (p.1157)
and d) condition, e.g.
Other things being equal, the chances of any
man being hit in action vary ••• wi th the
rate of fire to which he is exposed.
(p.1158)
The most recent discussion of the semantic connections

holding between free adjuncts or absolutes and their
superordinate clauses is that of Quirk, Greenbaum, Leech,
and Svartvik (1972:759-763). Like Kruisinga, they assert
18 CHAPTER I
that, in general, free adjuncts and absolutes express

attendant circumstances or contingency, but assume a more
specific sense in context; they suggest (p.762) that the
relations which free adjuncts and absolutes most commonly
bear to their superordinate clause are those of:
a) cause, e.g.
All our savings gone, we started looking for
jobs; (p.762)
b) time, e.g.
Cleared, this site will be very valuable;
(p.759)
and c) circumstance, e.g.
A case in both hands, Mabel stalked out of
the house. (p.759)
This seems to be the minimal categorization--all sources

cited agree that at least these three roles may be assumed
by a free adjunct or absolute.
Thus, the sources cited offer five roughly similar
inventories of the logical roles which a free adjunct or
absolute may play. These are summarized in Table 1 (see
next page).
The second question addressed by these researchers is,
again: What determines which logical role a given free
adjunct or absolute will play? Two lines of explanation are
pursued in an attempt to answer this question; one is
syntactic, the other pragmatic. Al though nei ther sort of
explanation is developed much, I shall summarize the
tentative conclusions which they have afforded.
Jespersen goes the furthest in his attempt to uncover
syntactic correlates of the semantic variability of
absolutes. He offers the following observations: the
causal relationship is most usual with absolutes having
present participles, especially being, in their predicates;
absolute phrases bearing the conditional relationship
normally precede their superordinate clause, as do those
bearing the temporal relationship; absolute phrases having
past participles as constituents, as well as many of those
having adjectival or adverbial phrases, seem to bear the
temporal relationship to their superordinate clause (thus,
he remarks (p.63) that 'dinner over is temporal, dinner
being over (or finished) as a rule causal'); absolute
phrases expressing circumstances are 'generally added after
the main part of the sentence' (p.63), often having as
Table
The Logical Roles Played by Free Adjuncts and Absolutes
According to Five Traditional Grammars
A free adjunct or absolute may be linked to its superordinate clause by a

relation of:
Attendant or
Reason, Alternative Condition, Concession,
Time Cause Circumstances Exception Contrast Manner
Curme
( 1931) x x x x x x
Kruisinga
(1932) x x x x
Jespersen
(1940 ) x x x x
Visser a
(1972) x x x x (x)b
Quirk, et al.
(1972) x x x
aExcluded are 'hypothetical similarity' and 'purpose'; none of Visser's
examples of expressions bearing these relations are free adjuncts in the
sense assumed here.
b,(X)' indicates a logical role assumed for free adjuncts but not for
absolutes.
constituents adverbial or prepositional phrases, or

consisting of what Jespersen terms (p.60) 'condensed
constructions'--quasi-idiomatic absolutes such as hat in
hand, head first, and face down. In the same vein, Curme
(1931: 155-156) notes that the inversion of the subject and
predicate constituents occurs occasionally within absolute
phrases expressing a causal, conditional, or circumstantial
relation, and appears commonly in absolute phrases
expressing a concessive relation (this inversion is limited
to a small number of participles, some of which have taken
on a prepositional value: except (excepted), granted, given,
during, pending, notwithstanding; or the value of a
subordinating conjunction, when the subject of the absolute
is a clause: given that, granted that, provided that); and,
in agreement with Jespersen, he remarks that the predicative
constituent of a nominative absolute phrase expressing
attendant circumstances is commonly adverbial or
prepositional. As interesting as these remarks may be, they
are neither necessary nor sufficient correlates of the
20 CHAPTER I
semantic diversity of absolutes, as is suggested by the

tentative language in which they are proposed.
Quirk et a1. (1972) adopt a different, pragmatic
approach to accounting for the semantic heterogeneity of
free adjuncts and ,absolutes. They attribute this semantic
versatili ty to what they call (p.760) 'the chameleon-like
semantic quality of adapting to context,' a property which,
they claim, is shared by many other English constructions:
they liken free adjuncts and absolutes to nonrestrictive
relati ve clauses and clauses introduced by the conjunction
and--all four
are capable of assuming, according to context, a

more precise semantic role:
The girl, who was upset by the activities of the

ghost, decided to leave
The girl was upset by the activities of the

ghost, and decided to leave
Although the mode of clause connection does not

say so, we infer that the girl's emotional state,
as described in the relative clause and the
initial clause of the coordinate sentence, was the
REASON for her departure. Exactly the same point
could be made about the equivalent non-finite
clause:
The girl, upset by the activities of the ghost,

decided to leave (PP.759-760)
This insistence on the importance of context-dependent

inferences for the semantic variability of free adjuncts and
absolutes recalls Kruisinga's (1932:276) assertion that
basically all free adjuncts and absolutes express attendant
circumstances, and that modulations of this sense arise only
because some particular circumstance of the action or event
described is more prominent than others.
This pragmatic approach is promising; it takes account
of the fact that the relation holding between a free adjunct
or absolute and its superordinate clause is basically
'open', subject to a range of evaluations based on language
users' inferences in a given context. This, I think, cannot
be doubted. Consider, for instance, sentence (59):
(59) Noticing that a crowd had gathered, Bill

immediately called the fire department.
What relation is inferred to hold between the free adjunct

and the superordinate clause in this sentence depends
importantly on context. A causal relation might, at first
glance, seem most plausible--Bill called the fire department
because he saw the crowd, and expected some kind of crisis;
but in the context characterized in (60), a merely temporal
or circumstantial relation is inferred for the free adjunct
in (59), since a prior cause is established for his calling
the fire department.
(60) While Mary was at work, Bill and Jane smelled

smoke and, looking outside, saw that her house was
burning down. Jane screamed for Bill to do
something. [(59)]
Clearly, contextual inferences play an important role in the

determination of the relationship holding between a free
adjunct or absolute and its superordinate clause.
I do, however, have two reservations about the
pragmatic explanation of the semantic versatility of free
adjuncts and absolutes. first, even if it is granted that
the relation holding between a free adjunct or absolute and
its superordinate clause may be a function of context, the
possibility of other determinants cannot be ruled out.
Could there be aspects of the meaning of a free adjunct or
absolute (as well as, perhaps, of the meaning of the clause
to which it is attached) which systematically and
recurrently determine the logical role it plays,
independently of context? Quirk et a1. (1972:762)
themselves offer just such a semantic correlate of the
versatili ty of free adjuncts and absolutes: 'In -ing
clauses, dynamic verbs typically suggest a temporal link,
and stati ve verbs a causal link.' This modest observation
suggests a third, semantic approach to explaining how the
logical role of a given free adjunct or absolute is
determined; that is, it suggests that there may be quite
regular semantic reasons for people's intuitions about how
free adjuncts and absolutes function.
Second, it is not absolutely clear that inferences
regarding the logical role of a nonrestrictive relative
clause or a coordinate clause have precisely the same status
as those pertaining to the role played by a free adjunct or
22 CHAPTER I
absolute. One might well infer that someone using anyone

of (61 )-(63) is trying to establish some causal connection
between the fact that John is an Englishman and the fact
that he is brave.
(61) John, who is an Englishman, is brave.

(62) John is an Englishman, and he is brave.
(63) John, being an Englishman, is brave.
In the case of (61) and (62), this causal connection is

merely suggested--one could reply to either of these
sentences with (64), but hardly with (65).
(64) Are you implying that John is brave because he is

an Englishman?
(65) No, that's not why he's brave.
An utterance of (63), on the other hand, apparently

constitutes an assertion of the causal connection; here, an
appropriate reply would be (65) rather than (64). That is,
(63) appears to differ from (61) and (62) in that an
inference about the logical role of the free adjunct must be
drawn if the sentence is to be understood at all--in that
what is inferred in (63) is actually felt to be part of what
is asserted.
In the following section, I outline my account of the
semantic variability of free adjuncts and absolutes; my
central thesis is that both semantic and pragmatic factors
may determine the logical role which adjuncts and absolutes
are felt to play.
3. Plan of Discussion
The thesis which I shall defend in the ensuing chapters is

that semantic and pragmatic factors jointly determine the
relation felt to hold between a free adjunct or absolute and
its superordinate clause. I shall argue (i) that in certain
cases, a free adjunct or absolute may serve as the argument
of an expression whose interpretation either fully
determines its logical role or greatly limits the range of
logical roles which it may be felt to play; and (ii) that
when the logical role of a free adjunct or absolute isn't
fully determined by the interpretation of some other
expression, it may nevertheless be constrained to uniqueness
by the inferences of language users. I show that such

inferences may be based on a number of diverse factors,
including semantic properties of the adjunct or absolute
itself, semantic properties of the superordinate clause, and
knowledge of the world.
Chapters II-V are devoted to those cases in which a
sentence containing an absolute construction also contains
an operator whose semantic properties fully or partially
determine the logical role which the absolute construction
is felt to play. Chapter II is specifically devoted to
developing an explanation for the fact that in modal
sentences, free adjuncts may be felt to function as if-
clauses. It is argued that the logical role of the adjunct
in a sentence of this sort is determined by the semantic
properties of the accompanying modal verb. This claim is
elucidated in the context of Angelika Kratzer's theory of
modal semantics: given the assumptions of this theory, the
decidedly conditional character of the adjuncts in (66)-(71)
follows purely and simply from their status as arguments of
the boldface modal verbs.
(66) This same thing, happening in wartime, would

amount to disaster.
(Visser 1972:1137)
(67) Beth's only response is to look stricken. Who

wouldn't be stricken listening to this crap?
(NY 10/13/80, 189)
(68) Left to his own devices, he would do a lot of Q.

and A.
(NY 9/29/80, 107)
(69) dressed a little more elegantly, she would in

no way have jarred with the tone of average
middle-class society.
(Visser 1972;1255)
(70) It is not hard to envision how frightened a cow

might be, hanging in a sling from a crane over a
boat that was pitching and rolling at the edge of
an apparently limitless expanse of roiling water.
(CL, 26)
24 ~AnER I
(71) Unrestrained by Margery's presence, Frank would
have been rude to Mrs. Rowlands.
(Visser 1972:1255)
Only certain sorts of adjuncts can serve as arguments of

modals, however; the distinction between those that can and
those that cannot is shown to correlate exactly with Greg
Carlson's distinction between stage-level and individual-
level predicates.
Chapter III deals with certain similarities in function
between free adjuncts and time adverbs:
(i) First, i t is shown that free adjuncts, like time
adverbs, can be used to restrict the interpretation of an
adverb of frequency; thus, just as the when-clauses in (72)
and (73) specify the nature of the time intervals quantified
by the adverbs usually and sometimes, the free adjuncts in
(74) and (75) likewise specify the sorts of intervals
quantified by often and always.
(72) When the weekend arrives, he usually heads for

open water.
(NY 9/15/80, 36)
(73) When people walk past them, however, they

sometimes find them irresistible.
(,2., 6)
(74) Because of their great size, obvious age, and

relatively metropolitan setting, they are a
landmark among citrus men around Winter Haven,
who, giving directions, will octen say something
like, "Turn left at the stop light. You can't
miss it. You'll see a grove of big Seedlings on
the corner."
(Q, 22)
(75) Asked to explain how it was that he didn't

separate his characters into the good ones and the
bad ones, Renoir's answer was always "Because
everyone has his reasons,"
(NY 10/13/80, 174)
A semantics for frequency adverbs is developed in which an

expression overtly restricting the interpretation of a
frequency adverb serves as an argument of that adverb; when
a free adjunct is employed in this way, the semantic

properties of the frequency adverb suffice to guarantee that
the role of the adjunct will be perceived as temporal.
(ii) A second similarity between free adjuncts and
time adverbs appears in generic sentences such as (76) and
(77): just as (76) is understood as expressing a
generalization about the set of time intervals characterized
by its when-clause, (77) is likewise understood as
expressing a generalization about the set of time intervals
characterized by its free adjunct. A semantic analysis for
such sentences is proposed; according to this analysis, (76)
and (77) contain a nonovert 'adverb of generalization',
comparable in its semantic properties to a frequency adverb.
When a free adjunct serves as an argument of this operator,
its logical role is necessarily felt to be temporal in
nature.
(76) When their old steel teapot develops a leak,

Donald plugs up the hole with a wood screw.
(CL, 32)
(77) Talking to Reagan, one gets the sense that he does

not realize that he is evading, is not dealing
with the issues.
(NY 9/29/80, 119)
Not all free adjuncts occur naturally as arguments of

frequency adverbs or of the proposed generalization
operator; only those with stage-level predicates appear
to. 13
(iii) A third similarity between free adjuncts and
time adverbs is in their interaction with tense. It is
proposed that adjuncts not serving as an argument of a modal
or frequency adverb or of the proposed generalization
operator belong to the same category as main tense adverbs--
time adverbs which join with a simple tense or with the
perfect to pick out some interval of time. Examples of
adjuncts which I shall analyze in this way are those in
(78)-(80).
(78) Being a superb dancer, she was in great demand for

parties.
(PM, 111)
26 CHAPTER I
(79) Having previously had no reason to think of the

residents of st. Louis as anything other than
fair-minded Americans, I was naturally distressed
to find that they were not able to respond with
any graciousness at all to my suggestion •••
(NY 6/16/80, 104)
(80) Prone in the muck, we worked systematically,

following a small grid and sorting what we found.
(CL, 140)
Chapter IV deals with the semantics of the perfect and

the progressive as it pertains to the interpretation of free
adjuncts. A compelling argument in favor of the position
that the adjuncts in (78)-(80) are main tense adverbs is
presented: it is shown that a free adjunct whose
superordinate clause is in a perfect tense has the same
range of interpretations as a main tense adverb appearing in
the same context. In addition, new semantic analyses of the
perfect and the progressive are proposed which are
consistent with their use in free adjuncts. 14
Chapter V is devoted to the formal semantics of
absolutes in English. It is shown that, from a semantic
point of view, absolutes are similar to free adjuncts: they
too may serve as arguments of modals as well as of adverbs
of frequency and generalization; and when not serving in
this way, they may be convincingly analyzed as main tense
adverbs. Moreover, just as the semantic behavior of a free
adjunct depends on whether its predicate is stage-level or
indi vidual-level, the same is also true of absolutes. The
latter do, however, have a special peculiarity:. if an
absolute is to serve as an argument of a modal, frequency
adverb, or generalization operator, it must not only have a
stage-level predicate--it must also be prepositionally
augmented.
(In the Appendix, the semantics for free adjuncts and
absolutes developed in Chapters II-V is presented formally.)
In Chapter VI, I consider the nature of the inferences
by which language users determine the logical role of a free
adjunct or absolute in case this is not fully determined by
the semantics of English: in addition, I consider the means
by which language users arrive at such inferences. Five
factors which play an important role in such inferences are
discussed in turn: the presence of a stage-level vs. an
individual-level predicate in an adjunct or absolute; the
relative duration of the events or states of affairs to

which an adjunct/absolute and its main clause relate; the
order of an adjunct or absolute with respect to its
superordinate clause; language users' knowledge of the
world; and the presence of a 'connective adverb' in an
adjunct/absolute or its superordinate clause. In addition,
some alternative analyses of the semantic variability of
absolute constructions are scrutinized, and the theoretical
implications of the analysis advocated here are discussed.
4. SOme Syntactic Conventions
Throughout this work, I shall adhere to certain conventions

in the statement of syntactic rules. These conventions are
intended to make the syntactic rules more streamlined, but
also to eliminate any undefined notions from their
formulation.
Five basic conventions will be employed: the use of
binary features representing subcategories of one or several
categories of expressions; the use of morphological
functions; the marking of main verb~ in verbal phrases,
sentences, and certain temporal abstracts 15 ; the marking of
subject noun phrases in sentences and certain temporal
abstracts; and the use of syntactic subroutines.
4.1. BINARY FEATURES

In many kinds of situations, it will prove useful to
distinguish a coherent subset of some category of
expressions. This will be done with binary features; such
features are thus to be thought of as sets of expressions.
The features to be employed here are the following:
[±Adjunct] [±Perfect]
[±Plural]
[+Adjunct] will determine two complementary subsets of

the set {x: x is a temporal abstract or set-level time
adverb}. Any rule producing an expression in this set will
specify whether it is a member of [+Adjunct] or of
28 CHAPTER I
[-Adjunct]; all basic members of this set will belong to

[-Adjunct]. Adjuncts and absolute belonging to the category
TAB of temporal abstracts or to the category TA of set-level
time adverbs will belong to [+Adjunct], and will thus be
distinguished from other members of these categories. This
is necessary, since the applicability of certain rules
operating on temporal abstracts or set-level time adverbs
will depend upon whether or not the expression to which they
apply is an adjunct or absolute: for example, the rules
which derive adjuncts and absolutes of categories tl, TA,
and MTA apply only to temporal abstracts in [+Adjunct] (see
Chapter III, sections 2.2, 2.3, 3.2); the rules of tense
apply only to temporal abstracts in [-Adjunct] (see Chapter
III, section 1.2.1); and the rule converting set-level time
adverbs into main tense adverbs applies only to adverbs in
[-Adjunct] (see Chapter III, section 1.3.2).
[+Weak] will specify two complementary subsets of the
set {x: x is a temporal abstract in [+Adjunct]}; all rules
producing temporal abstracts belonging to [+Adjunct] will
specify whether they are [+Weak] or [-Weak]. As will be
seen in Chapters II, III, and V, the distinction encoded by
this feature is central to the semantics of free adjuncts
and absolutes; it conditions the rules deriving adjuncts and
absolutes of categories t' and TA, which apply only to
members of [+Weak] (see Chapter III, sections 2.2, 3.2).
[+Tense] will specify two complementary subsets of the
set {x: x is a declarati ve sentence or temporal abstract};
every syntactic rule giving rise to a sentence or temporal
abstract will specify whether it is a member of [+Tense] or
of [-Tense]. The feature [.:tTense] will be useful for
preventing unwanted iteration of rules of tensing; by
stipulating that such rules apply to members of [-Tense] but
yield members of [+Tense], multiple tensing of a single
clause can be prevented. (See Chapter III, section 1.2.1.)
[.:tPerfect] will specify two complementary subsets of
the category of temporal abstracts; every rule giving rise
to temporal abstract will specify whether it is a member of
[+Perfect] or of [-Perfect]. The reasons for the
introduction of this feature are somewhat technical, and
will be clarified in Chapter IV, sections 1.2, 1.3.
The feature [.:tPlural] will determine two complementary
subsets of the category T of noun phrases. Every rule
giving rise to a noun phrase will specify whether it is
subcategorized as [+Plural] or [-Plural]; both subcategories
will in addition contain basic expressions--for example,
theY4n+3 is a basic expression of [+Plural], and he4n a
basic expression of [-Plural]. The membership of a noun
phrase in one or the other subcategory determined by
[+Plural] will be exploited in the statement of rules
involving subject-verb agreement. (In the fragment
presented in this work, all singular noun phrases are third
person; I thus ignore features of person in the statement of
rules of subject-verb agreement.)
4.2. MORPHOLOGICAL FUNCTIONS
Morphological functions will be used to 'call' participles

and finite forms of verbs in the derivation of participial
phrases and tensed sentences. The use of such functions
depends on the assumption that the lexical entry for a given
verb will list its participles and finite forms, or at least
determine such a list; additionally, holes in the paradigm
of a verb will be indicated. Thus, the verbs [be], [go],
[walk], and [must] might have the partial lexical entries in
Table 2 (on the following page). Only one form in the
lexical entry of a given verb will be regarded as a
syntactically basic expression; the others will have no
independent status in the syntax. For the vast majority of
verbs, the base form ([be], [go], [walk]) will be the
syntactically basic expression; in these cases, the
participial and finite members of the lexical entry will
only be introduced into larger expressions by syntactic
operations replacing bracketed main verbs (see 4.3 below).
(Modals, on the other hand, have no base forms in their
maximally defective paradigms; ingeed, in the fragment
developed here, a modal is assumed 1 to have just a single
present tense form, which is, by necessity, its
syntactically basic form.) Syntactic operations will 'call'
the participles and finite forms of a verb a by means of the
following functions:
PRPL(a) is the present participle of a;

PSPL(a) is the past participle of Q;
PRES1(a) is the third singular present form of a;
PRES2(a) is the plural present form of Q;
PAST1(a) is the third singular preterit form of a;
PAST2(a) is the plural preterit form of a.
(A verb may, of course, be undefined for some or all of

30 CHAPTER I
Table 2
Partial Lexical Entries for Four Verbs
Base Pres. Past Present Preterit

form pple. pple. 3sg. 3pl. 3sg. 3pl.
[be] [being] [been] [is] [are] [was] [were]
[go] [ going] [gone] [ goes] [ go] [went] [went]
[walk] [walking] [walked] [walks] [walk] [walked] [walked]
II II II [must] [must] II II
these functions--cf. Pullum and Wilson (1977).) In some

cases, a syntactic operation may convert one nonbasic verb
form to another, or back to its base form. For this reason,
the functions listed above are defined for all expressions
in a verb's lexical entry, so that, for example, PSPL([be])
= PSPL([being]) = PSPL([been]) = PSPL([is]) = PSPL([are]) =
PSPL([was]) = PSPL([were]) = [been]. In addition, a
function BASE will be introduced which is such that for any
verb form a, BASE(a) is the base form of a. Thus, in the
derivation of the past tense sentence (81) from the sentence
(82), PAST1([is]) is substituted for [is].
(81) (John) [was] happy

(82) (John) [is] happy
Similarly, one of the operations necessary for the

derivation of the modal sentence (83) from sentence (84) is
the substitution of BASE([walks]) for [walks].
(83) Bill [might] walk home

(84) (Bill) [walks] home
Morphological functions will also be used to 'call'

plural nominal forms. Thus, where a is a basic common noun,
PLUR(a) is the plural form of a.
As will be seen in 4.5 below, these eight morphological
functions will figure importantly in the statement of a
number of syntactic subroutines.
4.3. MAIN VERB MARKING
All sentences and verbal phrases will have their main verb
enclosed in brackets, as will temporal abstracts in the
subclass [-Adjunct]. If a is a basic expression of any
verbal category, it will be of the form [S], for some S.
Rules combining a sentence, temporal abstract, or (basic or
derived) verbal phrase y with another expression will retain
or delete the brackets in y according as the main verb of
y is or is not to be the main verb of the resulting
expression. Thus, the rule 335 combining an intransitive
verb phrase Y with a verb-phrase adverb will retain the
bracketing in Y:
S35. I f a e: P1VslIVs and S e: PIVs, then F35 (a,S)

e: P1Vs, where F 35 (a, S) is the concatenation
of sand a.
Example: F35 (very dangerously, [drive]) = [drive] very

dangerously
The main verb of the bare verb phrase is the same as the
main verb of the adverbially modified verb phrase in this
example. On the other hand, the rule 327 joining a verb
with a sentential complement ¢ will delete the bracketing in
S27. I f a e: P1Vs/ t and ~ E Pt , [+Tense], then

F 27 (a, ¢) e: P1Vs, where F27(0I.,~) is the
concatenation of a and that and ~', where ~,
is the result of deleting all brackets and

parentheses from ¢.
Example: F27 ([say ], (John) [likes] Mary) = [say] that

John likes Mary17
The main verb of the subordinate clause doesn't survive as

the main verb of the resulting verb phrase in this example.
A number of rules will refer to the main verb of a
sentence,
temporal abstract, or verbal phrase--those accomplishing
subject-verb agreement and tense marking, as well as those
32 CHAPTER I
introducing modals and perfect have and those converting
intransitive verb phrases into infinitive and participial
phrases. 18
4.4. SUBJECT MARKING

All sentences will have their subject enclosed in
parentheses, as will temporal abstracts in the subclass
[-Adjunct]; these parentheses will be introduced by the
rules joining a noun phrase with an intransitive verb phrase
to form a sentence. Rules operating on a sentence or
temporal abstract will either retain or delete the
parentheses around its subject according as it is or is not
to be regarded as the subject of the resulting expression.
For example, the rules which introduce tense, modals, and
perfect have into an expression ~ will retain the
parentheses around the subject of ~; but as was just seen,
the rule S27 joining a verb with a sentential complement
~ deletes the subject-parentheses of~. A number of rules
will refer crucially to the subject of a sentence or
temporal abstract: for example, rules introducing modals
and perfect have will place these immediately after the
subject of their sentential argument. It might seem that
such rules could be equivalently formulated as placing
auxiliary elements immediately before the (bracketed) main
verb of their sentential argument; but such a formulation
would yield the wrong results for sentences with preverbal
adverbs, producing (86) from (85) instead of (87).
(85) (John) Quickly [douses] the flame

(86) *(John) quickly [might] douse the flame
(8'7) (John) [might] quickly douse the flallle
4.5. SYNTACTIC SUBROUTINES
In order to avoid repetition in stating the effects of the

structural operations to be employed, and to capture the
similarities between certain sorts of operations, a number
of structural subroutines will be employed; the effects of a
given structural operation will be stated in terms of one or
several such subroutines. The subroutines to be employed
are:
EN DEL
ING ROS ('right of subject')
NONF ('nonfinite') RB ('rebracketing')
DB ('debracketing') PRB ('perfect rebracketing')
DP ('deparenthesizing RWRAP ('right-wrap'
and debracketing') --Bach 1980)
SSDELn ('n-subscript AGR ('agreement')
deletion') SUB
PRET CON
SUBJ ('subjunctive') CONWIB ('concatenation with
OBJ ('object case') intonation break')
These are defined as follows: for any expressions a. 8. y.
EN(a) is the result of replacing every expression of the

form [6] in a with PSPL([6]) (and is undefined i f some
bracketed expression in a lacks a past participial
form) :
EN([walk] home) = [walked] home;
ING(a) is the result of replacing every expression of the

form [6] in a with PRPL([6]) (and is undefined i f some
bracketed expression in a lacks a present participial
form) :
ING([walk] home) = [walking] home;
NONF(a) is the result of replacing every expression of the

form [6] in ex with BASE([6]) (and is undefined i f some
bracketed expression in a lacks a base form):
NONF«John) [walks] home) = (John) [walk] home;
DB(a) is the result of deleting all brackets from ex:
DB«John) [walks] home) = (John) walks home;
DP(a) is the result of deleting all brackets and parentheses

from ex:
DP«John) [walks] home) = John walks home;

34 CHAPTER I
SSDELn(a) is the result of deleting the subscript n from

every occurrence of hen' him n • his n • she n • her n , it n ,
its n • they n • them n , and their n in a:
SSDEL 1 «she 1 ) [likesJ himu) = (she) [likesJ himo;
PRET(a) is the result of replacing each expression of the

form [6J in 0: with PAST1([6]) i f [eI] is PRES1([6J) and
with PAST2([elJ) i f [6] is PRES2([6]) (and is undefined
if some expression of the form [6J in a is such that
PRES1([6]). PRES2([d]). PAST1([6]), or PAST2([6J) is
undefined):
PRET«John) [is] happy) = (John) [was] happy;
SUBJ(o:) is the result of replacing each expression of the

form [6J in 0: with PAST2([d]) (and is undefined if some
expression of the form [6J in a is such that PAST2([6])
is undefined):
SUbJ«John) [is] happy) = (John) [were] happy;

OBJ(a) is him n • her n • them n • or us if a is hen' she n , theYn'
or we, respectively; is a if a is any other member of
PT; and is otherwise undefined:
OBJ(heo) = himo;
OBJ(John) = John;
DEL(o:.B) is the result of deleting every instance of 0: in B:
DEL(at-that-time o' (John) [leftJ at-that-timeO) =

(John) [left];
ROS(o:, S) is the resul t of placing 0: immediately after the

first parenthesized expression in B (and is undefined
if B has no parenthesized expressions):
ROS(often, (John) [walks] home) = (John) often [walksJ

home;
RB(o:,B) is ROS(a, DB(NONF(B»):

SEl'1ANTIC VARIABILITY 35
RB( [will], (John) [walks] home) = (John) [will] walk
home;
PHB(a,S) is ROS(V(a), DB(EN(S»), where V is the

morphological function such that the first expression
of the form [6] in S is V([6]) (and is undefined i f 8
has no bracketed expressions):
PRB( [have], (John) [walks] home) = (John) [has] walked

home;
RWRAP(a,8) is the result of pI-acing 8 after the first word

of a:
RWRAP( [persuade] to leave, John) = [persuade] John to

leave;
AGR(at,8) is the resul t of replacing each expression of the

form Cd] in B with PRES1([6]) if a £ [-Plural] and with
PRESZ([6]) if a E [+Plural] (AGR is undefined if IX i
[-Plural] U [+Plural] or if there is some expression
[6] in S such that PRES1([6]) and PRES2([d]) are
undefined):
AGR(theY3' [be] happy) = [are] happy;
SUB(a,S,y) is the result of replacing the first occurrence

of 8 in y with a (and is undefined if y has no
occurrences of s):
SUB(John, him O' (Hary) [likes] him O) = (Mary) [likes]

John;
CONCa,S) is the result of concatenating a and s:

CON([drive], very dangerously) = [drive] very
dangerously
CONWIB(a,8) is the resul t of concatenating at and S with an

intervening intonation break:
36 CHAPTER I
CONWIB(yesterday, (John) [saw] Mary) = yesterday,

(John) [saw] Mary
The uses of these subroutines should be obvious from

the examples. EN and ING can be used to convert
intransitive verb phrases into past and present participial
phrases. NONF can be used to convert a finite clause ¢ into
a nonfinite one in preparation for the placement of a modal
verb into ¢. DB can be used to debracket any expression
whose main verb won't end up as the main verb of the
expression of which it is a constituent. DP can be used to
deparenthesize and debracket any expression whose subject
and main verb won't become the subject and main verb of the
expression of which it is a constituent (thus, rule S27
above can now be abbreviated as follows: ••• where F:Z7(ex,<P )
is CONCa, CON(that, DP( ¢))). SSDELn can be used 1n the
statement of rules of quantification. PRET can be used to
introduce past tense marking into a sentence. SUBJ can be
used to introduce subjunctive mood into a conditional
clause. OBJ is used in the statement of rules combining
verbs or prepositions with their objects. DEL can be used
to delete instances of the adverbial proform at-that-time n
(see Chapter III, section 1.4.2) as a concomitant of
binding. ROS can be used to insert adverbs into full
clauses. Similarly. HB can be used in the placement of
modals within a clause, and PRB can be used to place perfect
have. RWHAP can be used to place a direct object in the
appropriate spot within a complex transitive verb phrase
(cf. Bach 1980). AGR can be used to effect subject-verb
agreement as part of the rule combining a noun phrase with
an intransitive verb phrase. SUB can be used in the
statement of rules for binding term variables. CON and
CONWIB will, of course, be indispensable for the statement
of any rule effecting the concatenation of two or more
coconstituents.
Footnotes
1. Throughout this chapter and Chapter VI, I will rely

heavily on naturally occurring example sentences. Many of
these are from the following sources, whose abbreviations
are as indicated.
NY The New Yorker

o John Mc Ph e e, 0 ran g e s. Far r a r , St r a u s &
Giroux, New York. 1966.
CL John McPhee, The Crofter and the Laird.

Farrar, Straus & Giroux,~w York. 1969.
OSI Robert J. Seidman, One Smart Indian. Putnam,

New York. 1977.
MMB Georges Simenon, Maigret and the Man on the

Boulevard, trans. by H. Hamilton.
Penguin, New York. 1978.
PT Robert van Gulik, The Phantom of the Temple.

Scribner's, NewYork. 196~
PM Robert van Gulik, Poets and Murder.

Scribner's, New York. 1968.--
2. A careful reading of Poets and Murder, a mystery novel

by Robert van Gulik, turned Up:306 free adjuncts in 174
pages of text.
3. The free adjunct is native to English; examples can be

found in abundance from Old English through Modern English.
See Visser (1972:1132f;1252f).
4. For the standard statement of this constraint, see

Pullum (1974).
5. Unrelated adjuncts show extremely diverse control

properties. An unrelated adjunct may, as in (25)-(28), be
controlled by some contextually salient individual who may
or may not be referred to in the superordinate clause; it
may be controlled by the speaker(s), as in (31) and (32); it
may be controlled by the fact or event described by the
superordinate clause, as in (33) and (34); or it may express
an epistemic condition in whose light the superordinate
clause is or would be true, as in (29) and (30). A formal
semantics of 'control' in free adjuncts would clearly be a
complicated undertaking: it would have to draw upon
auxiliary theories of contextual salience, self-reference,
propositional attitudes, and speech acts.
6. Because absolutes are nonfinite, the appearance of

38 CHAPTER I
overt subject NPs poses difficulties for the Government and
Binding theory set forth by Chomsky (1982), according to
which subject position is normally only assigned abstract
Case in finite clauses (so that, by virtue of the Case
Filter, only finite clauses normally have overt subjects).
Reuland (1983) suggests that absolutes are in fact finite--
that they have phonetically unrealized agreement markers
which assign nominative Case to their governees and in this
way get them past the Case Filter. This claim obviously
cann'ot be falsified, but would appear to reduce the
substantive content of Case theory to vacuity.
7. A number of researchers working within the

transformational framework have assumed that the two
constructions are derivationally related by Equi-NP
Deletion, including Berent (1975), Ruwet (1978), Beukema
(1980), Beukema & Verheijen (1982), and apparently also Ross
(1973:112-113) and Cantrall (1974:21). As the examples in
(25)-(34) show, this cannot be the case for all adjuncts; I
doubt that it is for any adjunct, although I won't attempt
to justify this position here.
The slightly formal character of absolutes is perhaps
responsible for the relati vely small number--seventy--of
absolutes found in Poets and Murder; cf. footnote 2.
The Modern English absolute construction stems from the
Old English dative (occasionally accusative or nominative)
absolute construction, which is widely assumed to derive
from the Latin ablative absolute construction. Most often,
it 'occurs in those Old English translations from Latin in
which the writers allow themselves to be influenced by the
syntax of the original' (Visser 1972: 1073). See Jespersen
(1940:45f) and Visser (1972:1073f).
8. Reuland (1983:127) claims that absolutes whose
predicates are headed by -ins forms are 'the standard
case'--'that a grammatical association specifically with
-ing has been developing' in absolute constructions. This
conjecture does not accord with my own findings: in my
sampling of sixty absolutes, thirty-six did not have present
participial phrases as their predicates.
9. Reuland (1983:114f) holds that the predicative

consti tuent in an -ins absolute is not participial, a
position not easily reconciled with the fact that it can be
freely replaced with other sorts of predicative expressions,
as in (35)-(37).
10. Jespersen (1940:38-44) notes this peculiarity of these

prepositions; oddly, he later declares (p.145f'): 'The ing
in combinations after without I take to be a gerund'. Cf.
also Quirk, et al. (1972:762).
11. 'Absolutes' appear not to be commonly introduced by

subordinating conjunctions in Modern English; cf. Visser
(1972:1158;1271-1277).
12. Perhaps early generative linguists generally just

assumed that these constructions arise through the deletion
of an underlyingly explicit subordinator. Postal (1970:490)
casually refers to a rule of 'Since Subject Erasure',
'involved in the derivation of sentences such as [(i)] from
structures similar to sentences like [(iU]', as though it
were self-evident that these sentences are to be
transformation ally linked.
(i) Baving no boat, Harry was forced to swim.

(ii) Since hei had no boat, HarrYi was forced to swim.
(Babby (1975;32,fn2) analogously suggests, without argument,

that Russian gerunds (functionally equivalent to free
adjuncts) derive from deep structure APs labelled according
to their logical role.) In Chapter VI, I argue against such
syntactic approaches to the semantic variability of free
adjuncts and absolutes in English.
Among recent works pertaining to the syntax of absolute
constructions are Berent (1975), Beukema (1979, 1980, 1982),
Beukema & Verheijen (1982), Fukuchi (ms), Grady (1972),
Ishihara (1981,1982), McCawley (1983), Reuland (1983),
Riemsdijk (1978), Ruwet (1978), Sakakibara (1982), and
Williams (1975).
13. Chapter II and parts (1) and (ii) of Chapter III are an
elaboration of Stump (1980a).
14. This chapter is an elaboration of Stump (1980b).
15. These are expressions with the form of a sentence, free

adjunct, or absolute, but which denote sets of time
intervals. The existence of such a category of expressions
will be motivated in Chapter III.
40 CHAPTER I
16. This is merely a simplifying assumption, since some

modals (e.g. can) do, of course, have past tense forms; but
others (e.g. must) are genuinely defective in the way
described, from a synchronic perspective.
17. The subject parenthesization will be explained

immediately below.
18. As David Dowty has pointed out to me, the bracketing of

main verbs can probably be dispensed with if a recursive
definition of the head of an expression is introduced into
the syntax. Elegant though such an approach would be, I
will employ bracketing here, in the interests of simplicity.
C HAP T E R I I
MODALITY AND TIlE INTERPRETATION

OF FREE ADJUNCTS
In modal sentences, a free adjunct may be interpreted as

though it were an if-clause. Examples are the adjuncts in
(1 a)-(4a), whose semantic similari ty to the corresponding
conditional clauses in (1b)-(4b) is at least intuitively
apparent.
(1) a. Wearing that new outfit, Bill would fool

everyone.
b. If he wore that new outfi t, Bill would fool
everyone.
(2 ) a. Standing on a chair, John can touch the
ceiling.
b. If he stands on a chair, John can touch the
ceiling.
a. Taken in the prescribed dosage, it must be
very effecti ve.
b. If it is taken in the prescribed dosage, it
must be very effective.
(4 ) a. In first gear, the truck might reach the top
of that hill.
b. If it were in first gear, the truck might
reach the top of that hill.
In this chapter, I argue that the condi tional character of

free adjuncts such as those in (1 a)-(4a) is not the effect
of language users' inferences; instead, I argue that in
modal sentences, free adjuncts may function as members of
the same syntactic category as if-clauses, and that when
they do, the conditional interpretation which they receive
derives entirely from the semantic rules necessary for the
interpretation of modals.
Not all free adjuncts may function as conditional
clauses, however; consider sentences (5a)-(8a).
(5) a. Being a master of disguise, Bill would fool

everyone.
41
42 CHAPTER II
b. If he were a master of disguise, Bill would

fool everyone.
(6) a. Having unusually long arms, John can touch
the ceiling.
b. If he has unusually long arms, John can touch
the ceiling.
a. Containing twice the iron in a pound of
calf's liver, i t must be very effective.
b. If it contains twice the iron in a pound of
calf's liver, i t must be very effective.
(8) a. Weighing only a few tons, the truck might
b. If it weighs only a few tons, the truck might
Intuitively, the adjuncts in (5a)-(8a) do not play a

conditional role; that is, (5a)-(8a) are very different in
interpretation from (5b)-(8b). Thus, a second important
issue addressed in this chapter is the distinction between
two subclasses of free adjuncts: those which roay receive a
condi tional interpretation in modal sentences, as in (1 a)-
(4a), and those which may not, as in (5a)-(8a).
Section 1 is devoted to an informal, preliminary
discussion of this bifurcation of the class of free
adjuncts. In section 2, the nature of the bifurcation is
more carefully elucidated within the formal framework of
Kratzer's (1977, 1978, 1979) theory of conditional necessity
and possibility. In section 3, an independent criterion for
distinguishing the two sorts of adjuncts is uncovered; this
criterion is based on Carlson's (1977) ontological
distinction between stages and individuals.
The formal syntactic and semantic rules developed in
this chapter appear in the general fragment' for free
adjuncts and absolutes presented in the Appendix.
1. The Semantic Bifurcation of Free Adjunct3

in Modal Contexts
In this chapter, a free adjunct will be said to be in a

modal context iff its superordinate clause is headed by a
modal auxiliary verb, as for example in (9)-(12).
(9) a. Being a master of disguise, Bill would fool

everyone. (= (5a»
MODALITY AND FREE ADJUNCTS 43
b. wearing that new outfit, Bill would fool
everyone. (= (1 a»
(10) a. Having unusually long arms, John can touch
the ceiling. (= (6a»
b. Standing on a chair, John can touch the
ceiling. (= (2a»
( 11) a. Containing twice the iron in a pound of
calf's liver, it must be very effective.
(= (7a»
b. Taken in the prescribed dosage, it must be
very effective. (= (3a»
(12 ) a. Weighing only a few tons, the truck might
reach the top of that hill. (= (8a»
b. In first gear, the truck might reach the top
of that hill. (= (4a»
As was noted above, the class of free adjuncts is

bifurcated with respect to the range of interpretations
which they may receive in modal contexts. Thus, careful
examination of (9)-(12) reveals a striking difference
between the -a sentences and the -b sentences. The a
sentences entail the actual truth of their adjuncts:
sentence (9a), for example, entails that Bill is a master of
disguise; (lOa) entails that John has unusually long arms;
and similarly for (lla) and (12a). None of (9a)-(12a) could
be true unless its adjunct were also true. The ~ sentences,
on the other hand, don't necessarily entail the actual truth
of their adjuncts; each has a reading on which i t could be
true at some interval even if there were no time durin~ that
interval at which its adjunct were true.
The a sentences of (9)-(12) are just like the b
sentences,- apart from their adjuncts. This distinction in
entailments must therefore be regarded as the reflection of
some kind of important difference between two varieties of
free adjuncts. In the discussion which follows, adjuncts
like being a master of disguise and having unusually long
arms, whose actual truth is uniformly entailed in sentences
like (9)-(12), will be termed strong adjuncts; those like
wearing that new outfit and standing on a chair, whose
actual truth can fail to be entailed in sentences like (9)-
(12), will be called weak adjuncts. As will be demonstrated
in the present chapter and in Chapter III, this bifurcation
of the class of free adjuncts is quite systematic.
(Moreover, I will show in Chapter V that absolutes show a
similar bifurcation in English.)
44 CHAPTER II
In each of (9)-(12), the difference in entailment

between the strong and the weak adjunct is clearly linked to
the presence of a modal verb in the main clause; sentences
like those in (13) and (14), which contain no modal in their
main clause, uniformly entail the truth of their free
adjunct.
a. Being a master of disguise, Bill is fooling

everyone.
b. Wearing that new outfit, Bill is fooling
everyone.
( 14) a. Weighing only a few tons, the truck reaches
the top of the hill.
b. In first gear, the truck reaches the top of
the hill.
In addition, the difference in entailments between the

strong and the weak adjuncts in (9)-(12) is linked to an
intuitive difference in logical role: the strong adjuncts
are fe1 t to function as adverbs of explanation (because he
is a master of disguise, since it weighs only a few tons),
while the weak adjuncts are felt to serve as conditional
clauses (if he wore that new outfit, if it were in first
gear).
An understanding of this difference in entailments
between strong and weak adjuncts in sentences like (9)-
(12)--and of the apparently correlative difference in their
intuitive logical ro1e--requires a fine appreciation of the
semantics of moda1s. In the following section, I shall show
that Kratzer's (1977, 1978, 1979) theory of moda1s affords
an important insight into the source of these differences.
2. Explaining the Entailment Properties of Strong

and Weak Adjuncts in Modal Contexts
In her recent paper 'Conditional Necessity and Possibility',

Ange1ika Kratzer develops a novel theory about the meaning
of modal expressions in conditional sentences. This theory
yields a very satisfactory explanation of the entailment
differences between strong and weak adjuncts in modal
contexts. In this section, I shall briefly characterize her
theory, and then discuss its implications for the analysis
of free adjuncts.
2.1. KRATZER'S THEORY OF CONDITIONAL MODALITY
Traditionally, modals have been assumed to have several

di fferent kinds of meanings. Must, for example, has been
thought to have (at least) a deontic, an epistemic, a
dispositional, and a preferential meaning, as exemplified in
(15)-(18).
( 15) Deontic: Wi tnesses must swear to tell the

truth.
(16) Epistemic: It must be snowing at higher
elevations.
(17) Dispositional: Mary must lie down to keep from
fainting.
(18) Preferential: We must have that hat.
Kratzer (1977,1978), however, argues that modals don't

really have many different sorts of meanings; that must, for
example, has a single, neutral interpretation which remains
constant across sentences like (15)-(18), the apparent
variance in its meaning being a consequence of the fact that
its interpretation is relativized to a highly variable
aspect of the context of use.
According to Kratzer's theory, the interpretation of a
modal sentence is a function of more than just the meanings
of the modal verb and the sentence it modifies. To
interpret a modal sentence used in some context, we must
also consider the contribution of what Kratzer calls the
conversational background of that context. A conversational
background is, in model-theoretic terms, just a function
from possible worlds to sets of propositions; every context
of use will have one of these as one of its components. In
some modal sentences, the conversational background is
explicitly invoked; in sentences (19)-(22), for 'example, the
underlined phrases each invoke a conversational background.
In (19), this is the function which, for any possible world
w, gives the set of propositions expressing what is legally
prescribed in w; this might be called a deontic
conversational background.
(19) In view of what is legally prescribed, witnesses

must swear to tell the truth.
(20) In view of what is known, it must be snowing at
higher elevations.
46 CHAPTER II
(21) In view of what her disposi tions~, Mary must

lie down to keep from fainting.
(22) In view of what our preferences are, we must have
that hat.
In (20), the conversational background is a function which,

for any world w, gives the set of propositions expressing
what is known in w; this is thus an epistemic conversational
background. Other sorts of conversational backgrounds may
also be invoked, as in (21) and (22).
Modals, according to Kratzer, specify some logical
relation between the meaning of the sentence they modify and
the extension of the conversational background of the
context in which they are used. Must, for example
specifies a relation something like logical consequence;f
thus, sentence (19) is taken to mean, approximately, that
the proposition expressed by (23) follows logically from the
extension of the deontic conversational background which it
invokes (i.e. the set of propositions expressing actual
legal prescriptions); sentence (20), to mean that the
proposition expressed by (24) follows logically from the
extension of the epistemic conversational background which
it invokes (the set of propositions expressing what is
actually known); and so on.
(23) Witnesses swear to tell the truth.

(24) It is snowing at higher elevations.
(Because a conversational background may have an

inconsistent set of propositions as its extenSion, the
relation specified by must actually has to be slightly more
complicated than logical consequence. See Kratzer
(1977:347ff); cf. also below.)
Sentences (15)-(18), unlike (19)-(22), don't explicitly
indicate the nature of their conversational background; this
will, however, be implicit in the context of their use.
Thus, if (15) is understood deontically in some context, it
is because this context implicitly contains a deontic
conversational background; an epistemic understanding of
(16) in some context implies a covert epistemic
conversational background in this context; and so on. From
one context to another, the modal verb remains invariant in
interpretation; what varies is the conversational background
relative to which it is interpreted.
Now, this general analysis of modal sentences has
interesting consequences for the analysis of modal
conditionals like (25)-(28), as Kratzer (1979) shows. In
particular, it affords a very appealing account of the
function of the if-clause in such sentences.
(25) If they are convicted, defendants must appear in

court for sentencing.
(26) If John is a native of Medford, he must resent the
arrival of Californians in Oregon.
(27) If she exercises too hard, Mary must lie down to
keep from fainting.
(28) If we have enough money, we must have that hat.
In these sentences, as in (15)-(18), the modal

specifies a logical relation between the meaning of the
sentence it modifies and a certain set of propositions.
Here, however, the relevant set of propositions is not
merely the extension of the contextually-specified
conversational background-rather, i t is the union of this
set with the (unit set of the) proposition expressed by the
if-clause.
An example might help make this clear. Let H be a
conversational background whose extension is the set H' of
propos i t ions expressed by (29)-(32).
(29) Every native of the Pacific Northwest resents the

arrival of Californians in his home state.
(30) Every native of Oregon is a native of the Pacific
Northwest.
(31) Oregon is the home state of every native of
Oregon.
(32) Every native of Medford is a native of Oregon.
Suppose now that sentence (26) is used in a context in which

the conversational background is H. (26) will in this case
be true--not because the proposition expressed by (33)
follows from H' (it doesn't), but because it follows from
the union of H' with the unit set of the proposition
expressed by (34).
(33) John resents the arrival of Californians in

Oregon.
(34) John is a native of Medford.
In this example, must can be safely thought of as

48 CHAPTER II
specifying a relation of logical consequence. But the

proposi tion expressed by the if-clause of a sentence like
(26) may turn out to be incompatible with the extension of
the relevant conversational background (or, again, this
extension may itself be inconsistent); this possibility
complicates the kind of logical relation which a given modal
may be thought to specify in the interpretation of a modal
conditional.
An example will help to bring this out. Suppose that
the extension B' of a deontic conversational background B
consists of the propositions expressed by (35) and (36);
(35) No one drives down a one-way street the wrong way.

(36) Anyone driving down a one-way street the wrong way
loses his license.
The union of B' with the unit set of the proposition

expressed by (37) is an inconsistent set.
(37) John drives down a one-way street the wrong way.
Since any proposition follows logically from an inconsistent

set of propositions, all of (38a-d), if used in a context
with B as its conversational background, would be true under
the assumption that must specifies a relation of simple
logical consequence. This undesirable consequence shows
that must cannot simply be taken to specify a relation of
logical consequence between the proposition expressed by the
sentence it modifies and the relevant set of propositions.
(38) If John drives down a one-way street the wrong

way,
a. he must lose his license.
b. he must not lose his license.
c. his license must be displayed at the National
Gallery.
d. he must be distinct from himself.
Parallel examples would support analogous conclusions for

other modals.
One way to adjust the interpretation of must is as
follows. Let schema (39) stand for any modal construction
with must as its modal verb;
(39) If S" must S2

let p and q be the propositions expressed by S1 and S2'
respectively. Suppose (39) is used in a context whose
conversational background A has extension A'. Then must
specifies that q follows from every maximal consistent
subset of A' U {p} which contains p. As the reader can
demonstrate to him/herself, this adjusted interpretation for
must will make <38a) true if used in a context whose
conversational background has as its extension the set of
propositions expressed by (35) and (36), but will make (38b-
d) false, as desired.
Kratzer (1979:126) points out, however, that 'as soon
as we consider sets of propositions which have a more
respectable size, we cannot always be sure that all their
consistent subsets can be extended to a maximal consistent
subset.' For this reason, she proposes a slightly weaker
relation for must. Thus, suppose that (39) is used in a
context whose conversational background A has extension A',
and that p and q are as above. Let S be the set of all
consistent subsets of A' U {p} which contain p. Then must
specifies that for every s e: S, there is a superset of s in
S from which q follows logically. As the reader may verify,
this way of interpreting must yields exactly the same
results as the maximal-consistent-subset analysis for those
cases in which every s e: S can be extended to a maximal
consistent subset of A' U {p}: but it also covers the cases
in which some s e: S cannot be so extended.
Other modals can be analogously interpreted. Can, for
example, may be treated as follows. Let schema (40) stand
for any modal conditional with can as its modal verb;
(40) If S" can S2
let p and q be the propositions expressed by S, and S2'

respectively. Suppose that (40) is used in a context whose
conversational background A has extension A', and that S is
the set of all consistent subsets of A' U {p} which contain
p. Then can specifies that there is some s e: S such that q
is compatible with all supersets of s in S.
Would and might, according to Kratzer, are interpreted
just like must and can, respectively. This may, at first,
seem highly implausible; but careful consideration of the
differences between would and must, and between might and
can, reveals that they are not differences between the
logical relations which these modals specify; rather, they
are differences (i) in the sorts of conversational
50 CIlAPTER II
backgrounds with which these modals are conventionally

associated, and (ii) in the sorts of if-clauses with which
these modals are conventionally associated. For example,
the very distinctive sort of conversational background usual
for counter factual conditionals is associated with would and
might, but not with must and can; see Kratzer (1979:128-133)
for a discussion of such conversational backgrounds. 2 Would
and might, furthermore, can be used with if-clauses whose
propositions are incompatible with 'common knowledge', or
the presuppositions of language users; must and can require
if-clauses whose propositions are compatible with what is
presupposed (cf. Kratzer 1979: 133-137). These conventional
associations of certain modals with certain sorts of
conversational backgrounds and conditional clauses are of
great intrinsic interest; but as they are of marginal
relevance to the specific problem of interpreting free
adjuncts in modal contexts, I will have nothing more to say
about them here.
Although other sorts of modals may be interpreted in
accordance with Kratzer's theory, the purposes of the
present discussion can be satisfied most economically if
consideration is restricted to these four: must, can,
WOUld, and might.
In the light of Kratzer's theory) the following formal
account of the syntax and interpretation of modals is
assumed. (This account diverges from Kratzer's own formal
analysis in certain ways, all of them tri vial. One reason
for these departures is that I wish to have a single
interpretation for each modal, valid in both conditional and
nonconditional contexts; Kratzer's (1979) analysis involves
two interpretations for each modal, no doubt in the
interests of expository convenience.)
Modals are regarded as basic expressions of a basic 4
category MOD; conditional clauses are assigned to a special
category t'; and declarative sentences, as usual, belong to
the category t.
Category Description Basic expressions
MOD the basic category of [must], [can],

modals [would], [might]
t' the basic category of

conditional clauses
t the basic category of {( it) [rains]}

declarative sentences
Modals combine with tenseless sentences,5 either with or

without an accompanying conditional clause:
(41) 317. If cx t: P MOD and ¢ t: Pt , [-Tense], then

F 17 (cx,¢) £ P t , [-Tense], where F 17 (cx,¢) is
RB(cx, ¢).
Example: F 17 ([must], (John) [loses] his license) =

(John) [must] lose his license
(42) S18. If a t: P MOD ' ¢ t: Pt " and ljJ <: Pt' [-Tense],
then F 18 (a, ¢,ljJ) E P t , [-Tense], where
F 1S (cx,¢,ljJ) is CONWIB(¢, RB«(y',ljJ».
Example: F 18 ([must], if we have enough money, (ve)

[have] that hat) = if we have enough money,
(ve) [must] have that hat
Expressions of category MOD translate as intensional

logic expressions of type «s,«<s,t>,t>,t»,«s,t>,t», to
be interpreted in the following manner.
must', would' denote that function f of type

«s,«<s,t),t>,t»,«s,t>,t» such that for any
index <w,i>, any function h of type
<s,«<s,t>,t>,t», and any proposition p, f(h)(p)
= 1 (Le. truth) at <w,i> iff for every set s in
h«w,D), there is a superset of s in h«w,D)
from which p follows logically.
can', might' denote that function f of type

«s,«<s,t>,t>,t»,«s,t>,t» such that for any
index <w,i>, any function h of type
<s,«<s,t>,t),t», and any proposition p, f(h)(p)
= 1 at <w,i> iff there is some set s in h«w,i»
such that p is compatible with every superset of s
in h«w,i».
In sentences in which the modal is unaccompanied by a

52 CHAPTER II
conditional clause, the modal is to be interpreted relative

to the set of all consistent subsets of the extension of the
conversational background of its context of use. Thus, the
translation rule corresponding to (41) is stated as follows:
(43) T17. If ex E: PMOD ' ¢ Pt , and a,¢ translate as a',

E:
¢', then F 17 (ex, ¢) translates as

(X'(C(cb»(~¢') •
Example: (John) [must] lose his license translates as

must'(C(cb»(AJohn-loses-his-license,)6
Here, cb is a constant (of type <s,«s,t>,t») which, in any

context of use, is to denote the conversational background;
C (a constant of type «s,«s,t>,t»,<s,«<s,t>,t>,t>>» is
to denote that function C' such that for any function g of
type <s,«s,t>,t» and any index <w,i>, C'(g)«w,i» is the
set of all consistent subsets of g«w,i». As the patient
reader can verify, (43) has the effect of relativizing the
interpretation of a modal to all consistent subsets of the
extension of a conversational background.
In modal conditionals, the modal is to be interpreted
relative to the set of all consistent subsets of A' U {pI
containing p, where A' is the extension of the relevant
conversational background and p is the proposition expressed
by the accompanying if-clause. This is accomplished by
stating the translation rule for (42) as follows:
(44 ) T18 • I f a E PM 0 D' ¢ E P t I, 1/1 e; P t' and a, ~ 11/1

translate as a',Ij>',1/!', then F 18 (a, 1j>,1/J)
translates as a'(D(cb)(AIj> ,»(A1/!').
Example: if we have enough money, (we) [must] have

that hat translates as
must'(D(cb)(Awe-have-enough-money'»
(Awe-have-that-hat ' )
Here, cb is as above. D (a constant of type

«s,«s,t>,t»,«s,t>,<s,<<<s,t>,t>,t»») is to denote that
function D' such that for any function g of type
<s,«s,t>,t», any propOSition p, and any index <w,i>,
D'(g)(p)«w,i» is the set of all consistent subsets of the
union of g«w,i» with the unit set of p which contain p.
Where A' is the extension of a conversational background and
p is the proposition expressed by an if-clause, (44) has the
effect of relativizing the interpretation of a modal to all
consistent subsets of A' U {pI that contain p.7
The applications of this analysis to the semantics of
free adjuncts in modal contexts can now be considered.
2.2. THE ROLES OF STRONG AND WEAK ADJUNCTS IN MODAL

SENTENCES
Recall that the free adjuncts in sentences (9)-(12)

(repeated here) were seen to be of two types: strong, those
whose actual truth is entailed by the truth of any (present
tense) sentence in which they occur (as in the ~ examples);
and weak, those whose truth isn't always entailed in this
way ~in the b examples). If the foregoing analysis of
modality in conditional sentences is assumed, then the
entailment differences between strong and weak adjuncts can
be explained.
(9 ) a. Being a master of disguise, Bill would fool

everyone.
b. Wearing that new outfit, Bill would fool
everyone.
( 10) a. Having unusually long arms, John can touch
the ceiling.
b. Standing on a chair, John can touch the
ceiling.
( 11) a. Containing twice the iron in a pound of
calf's li ver, it must be very effecti vee
b. Taken in the prescribed dosage, it must be
very effecti vee
(12 ) a. Weighing only a few tons, the truck might
b. In first gear, the truck might reach the top
of that hill.
Consider, first, sentences (9a,b). In the most likely

understanding of (9b), the adjunct wearing that new outfit
is felt to express a (counter factual) condition;
intui ti vely, its function is exactly like that of the
subjunctive if-clause in sentence (45)--both expressions
54 CHAPTER II
condition the interpretation of the modal would.
(45) If he wore that new outfit, Bill would fool

everyone.
Thus, just as (45) translates as (46), it is plausible that

(9b) should likewise receive a translation of the form of
(47), in which p denotes the conditioning proposition
expressed by the adjunct wearing that new outfit.
(46) would'(D(cb) (ABill-vears-that-new-outfit'»

(AHill-fools-everyone')
(47) would'(D(cb)(p»(ABlll-fools-everyone')
Besides accounting for the conditional character of the weak

adjunct in (9b), such an analysis directly explains why (9b)
doesn't entail the truth of its adjunct: given the assumed
semantics for would, vp cannot be validly inferred from
(In). (More generally, for any modal a, a'(D(cb)(p»(q)
fails to entail either vp or vq.)
In (9a), on the other hand, being a master of disguise
doesn't express a condition on would. (9a) doesn't mean 'if
Bill were a master of disguise, he would fool everyone';
indeed, there is no explicit condition on would in this
sentence. The adjunct has some other, inferred role (most
likely, that of an adverb of explanation: because he is a
master of disguise); moreover, because its truth is
entailed, it must be assumed to be outside the scope of the
modal.
(10a,b) are analogous. In the most probable reading of
(10b), the adjunct standing on a chair has the intuitive
function of a conditional clause; it serves to condition the
interpretation of the modal can in precisely the same way as
the if-clause in (48).
(48) If he stands on a chair, John can touch the

ceiling.
If (10b) and (48) both receive translations of the form of

(49) (where the propositional value of p is supplied by the
adjunct or if-clause), then the fact that (10b) fails to
entail the truth of its adjunct follows automatically from
the semantics of can.
(49) can'(D(cb)(p»(AJobn-touches-the-ceiling')
(10a), on the other hand, doesn't have the character of a

conditional sentence; it isn't felt to mean 'if he has
unusually long arms, John can touch the ceilingl. The role
of the adjunct is instead an inferred one; furthermore, the
entailment properties of the adjunct make it necessary to
regard it as outside the scope of the modal.
Similar observations are true of (11a,b) and (12a,b):
in both cases, the weak adjunct can be understood to
condition the interpretation of the modal in the manner of
an if-clause, a fact which immediately accounts for its
entailment properties, under the assumed analysis of modal
conditionals; the strong adjunct, on the other hand, always
assumes some other sort of role, and does not fall within
the scope of the modal.
Thus, the entailment differences between strong and
weak adjuncts are a reflection of the fact that only the
latter may condition the interpretation of a modal in the
manner of an if-clause. Because weak adjuncts have this
capability, the fact that (9b)-(12b) fail to entail the
truth of their adjuncts follows as a direct consequence of
the assumed semantics for modals. On the other hand,
because (9a)-(12a) uniformly entail the truth of their
strong adjuncts, i t is clear that such adjuncts do not
appear within the scope of a modal.
Let me emphasize here that I am not suggesting that
weak adjuncts in modal contexts al ways serve to condi tion
the interpretation of the accompanying modal; examples (9b)-
(12b) have been chosen to bring out this use of weak
adjuncts, but it is by no means their only use in modal
contexts. Sentence (50), for example, could be understood
as a narrative sentence describing John's abilities on a
particular occasion of standing on the chair;
(50) Standing on the chair, John can see over the

fence.
in this case, the actual truth of the adjunct is entailed,

even though it is a weak adjunct. Thus, my claim is that
weak adjuncts, but not strong ones, .!!!!!.r condition the
interpretation of a modal, and that this is the source of
their entailment differences in modal contexts.
56 CHAPTER II
2.2.1. Weak Adjuncts in Modal Contexts
To provide a formal account of the foregoing analysis of the

entailment properties of weak adjuncts, very little
apparatus is necessary other than what has already been
proposed. Indeed, the desired consequences follow straight-
forwardly from the simple assumption that the class of free
adjuncts includes a subclass of conditional adjuncts, which,
like overt if-clauses, belong to the category t' of
conditional clauses. Under this assumption, the rule (42)
generating modal conditionals applies indiscriminately to
if-clauses and free adjuncts, so that sentences (9b) and
(45) receive analyses which are entirely parallel, apart
from the effects of the quantification rule S55. 8
(51 ) wearing that new outfit, (Bill) [wouldJ fool

everyone, t,· 18
[wouldJ, wearing that new (heO) [foolsJ everyone, t

MOD outfit, t'
(52) if he wore that new outfit, (Bill) [wouldJ fool
----------=----
everyone, t, 55,0
Bill, T if heO wore that new outfit, (heO) [wouldJ

fool everyone, t, 18
[wouldJ, if heo wore that (heo) [foolsJ everyone, t

MOD new outfit, t'
The syntactic analyses which (42) assigns to sentences

(10b)-( 12b) are likewise utterly parallel.
For such an approach to work, of course, members of the
category t' of condi tional clauses--free adjuncts and if-
clauses alike--must be assigned translations of the same
logical type. In the analysis of modal conditionals
presented in section 2.1, if-clauses were tacitly assumed to
translate as expressions of type t--that is, as sentences of
intensional logic, denoting truth-values. It seems qui te
natural to assume that conditional adjuncts are similarly
translated; indeed, the central question addressed in this
chapter--what is responsible for the entailment differences
between weak and strong adjuncts in modal contexts?--would
scarcely make sense if it weren't possible to regard a free
adjunct as denoting truth or falsehood. Thus, although
conditional adjuncts (like free adjuncts generally) have the

form of predicative phrases (verbal, adjectival,
preposi tional, and nominal), they translate as expressions
having a free variable x in 'subject position' (and thus
constituting sentences of intensional logic). For purposes
of simplicity, it is assumed here that the implicit subject
variable of a free adjunct is obligatorily bound by the
subject of the matrix clause 9 --that is, that the free
adjunct in (53) is true iff John is lying on the beach.
(53) Lying on the beach, John would surely catch cold.
(Explicit statement of the rules for deriving and

translating conditional adjuncts will be deferred until
Chapter III, section 2.2; the details of their formulation
hinge on issues other than those relevant to the present
chapter. )
If this approach to the semantics of conditional
adjuncts is assumed, then the translations assigned to
sentences (9b) and (45) by rule (44) are entirely parallel:
(54) would' (D(cb)( .... x-wearing-that-new-outCit'))

( .... Bill-Cools-everyone')
(55) would'(D(cb) ( .... Bill-wears-that-new-outfit' ))
( .... Bill-fools-everyone')
In that reading of (9b) which fails to entail the

actual truth of its free adjunct, the logical role of the
adjunct is determined by the way it conditions the
interpretation of the accompanying modal; very little need
be inferred about the logical relation between the free
adjunct and the matrix clause, since this relation is, in
effect, specified by the semantics of the modal would. The
translation in (54) captures this fact; the semantic role
which it attributes to the free adjunct in (9b) is entirely
determinate. Analogous consequences are likewise obtained
for the adjuncts in (10b)-(12b), whose function is again
precisely that of an if-clause conditioning the
interpretation of a modal.
58 CHAPTER II
2.2.2. Strong Adjuncts in Modal Contexts
Under the assumed semantics for modals, the fact that each
of the a sentences in (9)-(12) entails the truth of its
strong adjunct must be taken as evidence that these adjuncts
are outside the scope of the accompanying modal verb; thus,
the gross syntactic analysis for a sentence such as (9a)
must be one in which the adjunct is 'added on' to the main
clause only after the modal has already been added, as in
the informal analysis in (56).
(56) being a master of disguise, Bill would fool

everyone~
being a master of disguise Bill would fool everyone

~
would
B111 fools everyone
The adjunct in (56) serves not as an argument of the

accompanying modal,10 but as a sentence-modifier of some
sort. Accordingly, I shall, for the moment, refer
informally to such adjuncts as adsentential adjuncts.
The most probable logical role for the adsentential
adjuncts in (9a)-(12a) is that of a because-clause--(9a),
for example, seems naturally paraphrased by (57).
(57) Because he is a master of disguise, Bill would

fool everyone.
It is essential to recognize, however, that whereas the

conditional relation borne by the weak adjuncts in (9b)-
(12b) derives directly from the semantics of modals, the
because-type relation borne by the adjuncts in (9a)-(12a)
does not follow from the principles of English semantics at
all.
To see this, consider the following example:
(58) ?Because he is a master of disguise, Bill would,

nevertheless, never fool anyone.
(58) is odd because the subordinating conjunction becau~e

specifies a relation of causation/explanation between the
subordinate clause and the matrix clause, whereas the
'connective adverb' nevertheless specifies an adversative
relation; because and nevertheless are at odds, since the
relations they specify are totally antithetical. Consider,

however, the corresponding free adjunct sentence:
(59) Being a master of disguise, Bill would,

(59) seems considerably less peculiar than (58); the adverb

nevertheless doesn't obviously conflict with the adjunct as
it does with the adverbial clause in (58). In fact, the
adjunct in (59) doesn't appear to function as a because-
adverbial at all; on the contrary, the most likely
paraphrase for (59) is one in which the adjunct is rendered
as a though-clause--
(60) Though he is a master of disguise, Bill would,

Clearly it is the adverb nevertheless (and one's knowledge

of the fact that masters of disguise are normally good at
fooling people) that makes it natural to construe the
relation between the adjunct and the matrix clause in (59)
as one of adversativeness. 11 The point here is that the
strong adjunct being a master of disguise can't be viewed as
receiving a because-type interpretation in (9a) owing to the
effect of some semantic rule; instead, the logical role
borne by this adjunct must be able to vary according to the
inferences evoked by the context in which the adjunct
occurs.
In order to account for this variability in the
relation between a strong adjunct and its matrix clause, the
following approach will be adopted and developed here.
Sentence (57) might be plausibly argued to translate as
in (61), where because' is an intensional logic constant (of
type «s,t>,«s,t>,t») specifying the relation of
causation/explanation holding between the subordinate and
matrix clauses in (57).
(61 ) because' ("Bill-is-a-master-of-disguise')

("Bill-would-fool-everyone')
Sentence (9a) is like sentence (57) to the extent that it

serves to relate the proposition that Bill would fool
everyone to some auxiliary proposition; unlike (57),
however, (9a) is indeterminate with respect to the precise
nature of this relation. ThUS, (9a) might be assigned a
60 CHAPTER II
translation something like (62), in which L is a free

variable (also of type «s,t>,«s,t>,t») ranging over the
various logical relations which may hold between a free
adjunct and its matrix clause; L is, in effect, a variable
subordinating conjunction.
(62) L(A x-being-a-master-of -disguise')

(ABill-vould-fool-everyone')
As a free variable, L can have as its value any kind of

propositional relation whatsoever, including the highly
unlikely relations denoted by the intensional logic
expressions in (63).
(63) ApAq[p = q v ,[p = q]]

ApAq[ "p .. -,"p]
ApAq[ [vp v vq] + [vq .. O[vp & John-owns-a-dog']]]
In the analysis developed here, however, the interpretation

of a sentence such as (9a) depends not merely on rules of
semantics, but on the inferences of language users;
specifically, the nature of the logical relation holding
between the adjunct and the matrix clause in (9a) must be
inferred. Thus, in a given context of use, the range of
pragmatically admissible values for L--those determined by
the inferences of language users--will always be drastically
constrained, often to uniqueness.
The broader theoretical consequences of such an
analysis cannot be overlooked. My claim is that the
semantic interpretation rules for English underdetermine the
interpretation of a sentence such as (9a) , in that they
fail to specify any value at all for the relation
represented by L. Accordingly, the inferences of language
users playa cri tical role in the analysis advocated here:
they aren't merely deductions drawn on the basis of the
interpretations assigned to the sentences in a discourse;
rather, they may actually determine the interpretations of
the sentences in a discourse by fixing the value of some
element (such as L) whose value has not been determined by
the semantic rules of the language--that is, l~nguage users'
inferences may actually constitute a precondi tion for the
assignment of an interpretation to certain sentences. What
I am proposing, then, is that inferences playa much more
central role in the interpretation of sentences than has
generally been granted in work on formal semantics. For a

general discussion of this proposal, see Chapter VI.
Before proceeding, it will be desirable to make one
adjustment in the use of L to represent the relation between
a strong adjunct and its matrix clause. Recall that the
original reason for distinguishing strong and weak adjuncts
was that in modal sentences such as (9)-(12), the truth of a
strong adjunct is uniformly entailed (while that of a weak
adjunct is not); note also that sentences with strong
adjuncts also entail the truth of their matrix clause.
Thus, it appears that all pragmatically admissible values
for L are such that L(p)(q) entails both -p and -q, where p
and q denote propositions; that is, all pragmatically
admissible values for L give rise to facti ve entailments. 12
There are at least two ways of interpreting this fact.
It may simply be an accidental fact about pragmatically
admissible values for L that they happen always to give rise
to factive entailments. On the other hand, it may be that
the semantic rules of English, though failing to determine
the precise nature of the relation holding between a strong
adjunct and its matrix clause, nevertheless require that the
inferred relation give rise to factive entailments. I know
of no conclusive evidence deciding the issue; the fact that
this property of sentences with strong adjuncts remains
invariant in all contexts, however, at least suggests that a
semantic account is appropriate.
A semantic account can be provided as follows. Let K
be a constant of intensional logic having the same
denotation as (64) (where L is, as above, a variable of type
«s,t>,«s,t>,t», and p,q,r,s are variables of type <s,t»:
(64) ALApAq[L(p)(q) & II.rll.s[L(r)(s) -to [-r & -s]]]
Observe that the result of applying K to L is an expression

K(L) whose interpretation is exactly like that of L (i.e.
its value may be anyone of a range of functions of type
«s,t>,«s,t>,t») except that its values must always be
relations which can be true of a pair <p,q> of propositirins
at some index only if p and q are each true at this index.
If K(L) rather than L is used as the implicit subordinator
in the translation of sentences with strong adjuncts, then
the desired facti ve entailments are guaranteed. Thus, the
rough translation (62) for sentence (9a) is here revised as
in (65) (the latter is still no more than a tentative
translation, however).
62 CHAPTER II
(65) K(L)("'x-being-a-master-of-disguise')
(ABill-uould-fool-everyone')
K(L) can be analogously employed in the translation of

sentences (1 Oa)-( 12a). Moreover, modal sentences with weak
adjuncts can also be assigned translations similar to (65).
Recall, for example, that sentence (50) has a 'narrative
present' interpretation which entails the truth of its weak
adjunct:
(50) Standing on the chair, John can see over the

fence.
On this interpretation, the adjunct in (50) must be regarded

as adsentential rather than conditional; accordingly, it is
to be translated roughly as in (66).
(66) K(L) ("'x-staDding-on-the-chair')

("'John-can-see-over-the-fence')
Given the logical properties of K(L), (66) entails that ~ (=

John) is standing on the chair.
The implicit subordinator K(L) can, likewise, be used
in the analysis of nonmodal sentences;
(13 ) a. Being a master of disgui3e, Bill i3 fooling

everyone.
b. Wearing that new outfit, Bill is fooling
everyone.
Here, the adjuncts must be regarded as adsentential: since

there is no modal present, neither can be regarded as a
condi tional adjunct. (This is not to imply, however, that
all adjuncts which aren't conditional (e: P t ,) are
necessarily adsentential; in Chapter III, sections 3 and 4,
it is shown that adjuncts may function as members of yet a
third category--the category TA of set-level time-adverbs.)
For the time being, this analysis of adsentential
adjuncts remains an informal one: adsentential adjuncts
have not yet been assigned to any formal category of
expressions, and formal rules have not yet been proposed for
generating and translating adsentential adjuncts, nor for
combining such adjuncts with their superordinate clause.
The correct statement of these rules depends upon matters
beyond those addressed in this chapter. In Chapter III,
section 2, however, I propose that adsentential adjuncts in
fact belong to the category MTA of main tense adverbs;
formal rules for the derivation and distribution of such
expressions are presented in connection with this proposal.
In section 2 of Chapter IV, a semantic argument in favor of
this proposal is advanced.
Let me sum mari ze what has been shown in this section.
First, I have shown that what distinguishes weak adjuncts
from strong adjuncts is that the former but not the latter
may condi tion the interpretation of a modal, precisely in
the manner of an if-clause; this, ultimately, is what is
responsible for their entailment differences. If it is
assumed that weak adjuncts conditioning the interpretation
of a modal are members of the category t' of conditional
clauses, then their syntactic and semantic comportment
follows automatically from rules which are needed
independently for the analysis of modal conditionals.
Adjuncts not conditioning the interpretation of a modal--
including strong adjuncts generally and weak, nonconditional
adj uncts--m ay be informall y re garded as adsententi al
modifiers of some sort. They are interpreted as having an
implicit, variable subordinator K(L), whose logical
properties guarantee that they receive a factive
interpretation.
In all, this analysis of the interpretation of free
adjuncts in modal contexts seems satisfying. It has,
however, turned up at least two new problems:
(1) What independent criteria are there for
distinguishing strong and weak adjuncts? In the absence of
such a criterion, the observation that strong adjuncts
cannot condition the interpretation of modals is no
different from the virtually uninformative claim that some
adjuncts cannot condition the interpretation of modals.
(ii) !!...!:!.1 can't stron~ adjuncts condition the
interpretation of ~ modal? That is, even if an independent
cri terion can be found for distinguishing strong and weak
adjuncts, why should this distinction be reflected in the
range of interpretations which adjuncts may receive in modal
contexts? As I have shown, the manifestations of the
strong/weak distinction in modal contexts can be
insightfully described if two subclasses of free adjuncts
are distinguished--conditional adjuncts and adsentential
adjuncts; but a description of these manifestations does
not, in itself, explain why they should exist in the first
64 CHAPTER II
place. Presumably, an answer to question (i) will go a long
way towards providing this sort of explanation.
In the next section, problem (i) is addressed; as I
show, there is a clear independent criterion for
distinguishing strong and weak adjuncts. The second problem
(ii) is considered in more general terms in Chapter VI,
section 2.
3. A Semantic Correlate of the Distinction between Strong

and Weak Adjuncts
In this section, I show that Carlson's (1977) distinction

between individual- and stage-level predicates is highly
reliable as a correlate of the distinction between strong
and weak adjuncts.
In searching for an independent semantic correlate of
the distinction between strong and weak adjuncts, a
reasonable first hypothesis is that the relevant correlate
is aspectual in nature. Careful consideration reveals that
strong adjuncts always derive from stative predicates, and
that weak adjuncts often derive from nonstative predicates,
as the examples in (67)-(68) suggest.
(67) strong adjuncts:
Being something of an insomniac, Mary might be

reading at this hour.
Jim, having a wooden leg, can't remain standing

for long periods.
Measuring less than an inch, Larry's pet goldfish

would easily fit into this jar.
(68) Weak adjuncts:
Walking to the store, Mary can stop by to talk to

Jane.
Noticing the snake, Bill might run off.
John would probably have an accident, driving this

sort of car.
Nevertheless, it appears that the stative/nonstative
distinction is at best an approximation of the property
distinguishing strong and weak adjuncts, for although all
free adjuncts deriving from nonstative predicates are weak,
there are at least four sorts of cases in which adjuncts
deriving from stative predicates are also weak. First, free
adjuncts derived from statives of the stand-sit-lie class
are weak, not strong; in the se"ntences in (69), the actual
truth of the free adjunct needn't be entailed. 13
(69) The stand-sit-lie class:
Lying on the beach, you can get a better tan than

at home.
Sitting in the garden, Al might spot the turkey.
Second, free adjuncts derived from stative perception

predicates are weak, not strong; the adjuncts in (70), like
those in (69), show the distinctive entailment properties of
weak adjuncts in modal contexts.
(70) Stative perception predicates:
Seeing this rainbow, Hank would flip.
Larry, hearing this song, can get very

sentimental.
Third, all stative adjective phrases that are acceptable as

free adjuncts are weak rather than strong, as the examples
in (71) show:
(71) Stative adjective phrases:
Alone, she might not seem so formidable.
Sick in bed, John wouldn't be of much help.
Finally, most predicative prepositional phrases give rise to

weak adjuncts, though they are stative.
66 CHAPTER II
(72) Prepositional phrases:
In his maroon suit and whi te shoes, he must look

like a real salesman.
On a stage, he can be very winning.
These four kinds of cases show quite conclusively that

stativity cannot be the property which determines whether an
adjunct will be weak or strong. Hore generally, the
existence of sentence pairs like that in (73) seems to dash
all hopes that any kind of aspectual correlate of the
distinction between strong and weak adjuncts can be found.
(73) Asleep, Rover might not seem so ferocious.

(weak)
Being asleep, Rover might not seem so ferocious.

(strong)
There is, however, a correlate of a completely

different sort. As I now show, the distinction which Greg
Carlson has drawn between individual-level and stage-level
predicates appears to be what determines whether a given
adjunct is weak or stronb. In section 3.1, an informal
characterization of the distinction between stage-level and
individual-level predicates is presented. In section 3.2, a
formal account of this distinction is developed, with
special attention to the semantics of be. In section 3.3,
it is shown that the stage/individual distinction among
predicates correlates exactly with the weak/strong
distinction among free adjuncts. In section 3~4, I show
that the assumed correlation between Carlson's ontology and
the weak/strong bifurcation affords an insightful analysis
of prepositional adjuncts headed by as and with.
3.1. CARLSON'S ONTOLOGY OF STAGES AND INDIVIDUALS
In his dissertation, Reference to Kinds in English, Carlson

shows that two very broad classes of predicates can be
distinguished in English on both distributional and semantic
grounds. Among the diagnostic environments which
distinguish these two varieties of predicates is that in
(74) (cf. p.208ff);
(74) Mary saw John
Generally, if a predicate is acceptable in this sort of

environment, it is what Carlson terms a stage-level
predicate; examples are listed in (75).
(75) Stage-level predicates:
drunk
alone
asleep
sick in bed
naked
walk(ing) to the store
ly(ing) on the beach
carry(ing) a load of over 10 lbs.
driv(ing) this sort of car
wear(ing) his new outfit
stand(ing) on a chair
notic(ing) the snake
sit(ting) in the garden
hear(ing) this song
in his maroon suit and white shoes
on a stage
be(ing) a bastard
be(ing) silly
be(ing) arrested
be(ing) awarded the Medal of Honor
Individual-level predicates, on the other hand, cannot occur

acceptably in environments of this sort; the predicates
listed in (76), for example, do not appear naturally in
environment (74).
(76) Individual-level predicates:
be(ing) a master of disguise

be(ing) a sailor
be(ing) something of an insomniac
weigh(ing) over 200 lbs.
hav(ing) unusually long arms
hav(ing) a wooden leg
stand(ing) over six feet tall
intelligent
be(ing) intelligent
68 CHAPTER II
(76) [continued]
American
be(ing) American
fat
be(ing) fat
As Carlson demonstrates, the distinction between stage-level

and individual-level predicates is a profound one in
English; in fact, he argues that stage-level and individual-
level predicates are predicates of onto logically distinct
sorts of anti ties.
Three sorts of entities are distinguished in the
ontology developed by Carlson: kinds~ objects, and stages.
Kinds are entities of the sort denoted 4 by bare plural noun
phrases: the noun phrase dogs, for example, names the kind
of thing that dogs are. Objects are entities of the sort
denoted 14 by proper names: Fido names the object Fido.
Anything that is a kind or an object is, in Carlson's
terminology, an individual. Stages are 'spatially and
temporally bounded manifestation[s]' (p.115) of individuals;
these are, intuitively, the spatiotemporal slices of an
individual that show up as part of some event or state of
affairs. (English doesn't have any expressions which
directly denote stages, but stages do, according to Carlson,
figure extensively in the interpretation of complex
expressions.) Objects of a certain kind are said to realize
that kind; stages of a certain individual, to realize that
individual.
According to Carlson, individual-level and stage-level
predicates are predicates of individuals and stages,
respectively, as their names imply. This claim may embody
some intuition-the predicates in (75) do, after all, seem
to express highly temporary, accidental properties, while
those in (76) seem to express more inherent, essential sorts
of properties; but the original motivation for Carlson's
claim that individual-level and stage-level predicates
express ontologically distinct sorts of properties was that
it afforded a highly satisfactory account of the semantics
of bare plural noun phrases in English.
It has traditionally been assumed that bare plurals
have (at least) two kinds of meanings--one existential, as
in (77a), the other generic, as in (77b).
a. Dogs chewed up that newspaper.
b. Dogs are intelligent.
This has led some formally-minded researchers to postulate

two sorts of logical representations for bare plurals--one
involving an existential quantifier, the other a 'generic
quantifier'. Carlson, however, shows that in opaque or
multiply-quantified contexts, the bare plural behaves very
unlike any kind of quantified expression, resisting any
reliable quantificational paraphrase and uniformly failing
to participate in ambiguities of scope; this and related
facts lead him to the conclusion that bare plurals are not
interpreted as quantified expressions, but are instead just
names. He shows, furthermore, that the subject of sentences
(77a,b) actually has a single, constant meaning--that the
existential character of (77a) and the genericity of (77b)
are in fact to be attributed to their respective predicates.
Chew up that newspaper is a stage-level predicate, as the
diagnostic environment (74) shows: as such, it can't be
directly applied to kinds, rather only to stages;
accordingly, (77a) is interpreted as a predication of some
stage of the kind denoted by dogs, as in (78) (where R
denotes the relation between a stage and the indi vidual(s)
it realizes, x S is a variable over stages, and tense is
ignored).
(78) Vxs[R(X s , dogs') & chew-up-that-newspaper' (x s )]
Be intelligent, on the other hand, is an individual-level

predicate, as (74) suggests: it can apply directly both to
objects and to kinds; thus, (77b) is interpreted as a direct
predication of the kind denoted by dogs, as in (79).
(79) be-intelligent'(dogs')
According to these interpretations, the meaning of dogs

plays a constant role in (77a,b); it is clearly the
predicates whose meaning determines the existential or
generic flavor of these sentences. Carlson points out that
sentences which have proper names as their subjects also
vary in character depending on whether their predicate is
individual-level or stage-level; sentence (80a), for
example, refers to a specific event, while (aOb) expresses a
'generic' fact about Fido.
70 CHAPTER II
(80) a. Fido chewed up that newspaper.

b. Fido is intelligent.
This difference is explicitly accounted for in the way

Carlson iriterprets these sentences; their translations are
(81a,b).
(81) a. Vxs[R(xs,Fido') &
chew-up-that-newspaper'(x s )]
b. be-intelligent'(Fido')
Carlson's system thus provides a unitary explanation of

the intuitive variance in meaning in (77) and (80); this
variance is shown to be strictly a function of whether the
predicate involved is stage-level or individual-level.
Carlson's distinction between individual-level and
stage-level predicates is also useful in accounting for
other sorts of facts. He shows that certain constructions
admit only stage-predicates: existential there sentences
(pp.211ff) of the form of (82) allow stage-level but not
individual-level predicates, as (83) suggests;
(82) there Copula Noun phrase Predicate
(83) a. There was a dog chewing up· that newspaper.

b. *There was a dog being intelligent.
similarly, as (84) shows, progressive be combines only with

stage-level predicates (cf. pp.183ff).15
(84) a. Fido is chewing up that newspaper.

b. Fido is sitting on the new chair.
c. *Fido is being intelligent.
d. *Fido is weighing 30 Ibs.
Predicate nominals, on the other hand, are always 16

individual-level.
In the following subsection, a formal account of the
distinction between stage-level and individual-level
predicates is developed.
3.2. A CATEGORIAL DISTINCTION BETWEEN STAGE-LEVEL AND

INDIVIDUAL-LEVEL PREDICATES
Carlson (1977) has proposed a formal account of the

distinction between stage-level and individual-level
predicates; notwithstanding certain minor departures, the
analysis assumed here is essentially like Carlson's. First,
a categorial distinction is made between stage-level and
individual-level intransitive verb phrases, as well as
between stage- and individual-level predicative phrases:
the basic category of [walk], [dance],

stage-level intransi- [laugh],
tive verb phrases
the basic category of A
stage-level infinitive
phrases
PRPL s the basic category of
stage-level present
participial phrases
PSPL s the basic category of A
stage-level past parti-
cipial phrases
the basic category of asleep, ready,
stage-level adjective available,
phrases
stage-level pre~osi
tional phrases 1'(
individual-level intran-
sitive verb phrases
individual-level infini-
tive phrases
individual-level present
participial phrases
PSPL i the basic category of A
individual-level past
participial phrases
CHAPTER II
the basic category of intelligent, Cat,

individual-level adjec- Anerican, obnoxious,
tive phrases
the basic category of
individual-level ~repo
sitional phrases 1
NOM the basic category of
individual-level pregi-
cative noun phrases 1
PREDs is to be the supercategory of stage-level predicative

phrases; the membership of this supercategory is tQe union
of those of INF s , PRPLs, PSPL s, ADJ s , and Pps. PRED~ is the
supercategory of individual-level predicatiye phra~es, who~e
ruempersQip is the union of those of INF~, PRPL~, PSPL~,
ADJ 1, PP~, and NOM.
l1embers of ~REDs will be interpreted as sets of stages;
members of PRED~, as sets of individuals. So that this
difference will be reflected in the translations of
expressions of these categories, I will, following Carlson,
employ a sorted intensional logic. (The formal definition
of this logi9 appears in the Appendix.) The variables x~,
xo , xk, and xl are thus to be the nth variables over stages,
oBjecEs, kinas, and individuals, respectively, for any
nonnegative integer n. Similarly for variables of higher
types.
Noun phrases will always denote individual-level
property-sets; as a consequence, no stage-level verb phrase
will ever be interpreted as actually belonging to the
property-set denoted by its subject. For this reason,
stage-level verb phrases will have to be converted to
individual-level verb phrases by rule (85) before they can
join with a subject.
(85) 821. If a e: P1Vs, then F 21 (a) e: P1Vi, where F21 (a)

is ex.
The corresponding translation rule gives an expression

denoting the set of individuals having a stage in the
denotation of ex:
(86) T21. If ex e: P1Vs and ex translates as ex', then

F21 (a) translates as
AxiVxs[R(xs,X i ) & ex'(x s )].
Thus, for examp.le, F21 ([s.ee] Mary) = [see] Mary, which
translates as AX1VXs[R(xs,x 1 ) & see-Mary'(x s )].
The rule combining a noun phrase wi th an indi vidual-
level verb phrase is completely straightforward:
(87) S19. If cx PT and S E PIVi, then F 19 (a,S)

E E Pt ,
[-Tense], where F 19(cx,S) is
CON«a), AGR(cx,S».
Example: F 19 (John, [be] intelligent) = (John) [is]

intelligent
T19. If a E PI' S E P1Vi, and a, S translate as

a',S', then F 19 (a,S) translates as a'("'S').
Example: F 19(John, [be] intelligent) translates as

be-intelligent'(John')
The introduction of a distinction between stage-level

and individual-level verbal expressions affords a very
precise understanding of different uses of be in English. I
will briefly discuss these here.
As was seen above, stage-predicates with bare plural

subjects induce a kind of 'existential reading', as in
(77a); individual-level predicates like be intelligent
induce a 'generic reading'. This should lead one to suppose
that the sort of 'reading' produced by a predicate when it
has a bare plural subject will be yet another indicator of
whether it is stage-level or individual-level. This is in
fact true for most cases. There is, however, one well-
defined class of predicates which are individual-level, but
induce 'existential readings': these are predicates
consisting of be plus some stage-level predicative phrase.
According to the evidence of the diagnostic environment
(74), existential there sentences, and progressive
sentences,19 the predicates in (88) are individual-level
rather than stage-level (even thoug~ they contain stage-
level adjective or prepositional phrases); yet, they induce
74 CHAPTER II
an 'existential reading' in a bare plural sentence like Dogs

were on a stage.
(88) be(ing) drunk

be(ing) alone
be(ing) asleep
be(ing) sick in bed
be(ing) naked
be(ing) in his maroon suit and white shoes
be(ing) on a stage
Carlson thus translates the be 1 in (88) as (89) (p.410);
(89) APsAxivxs[R(xS,xi) & pS{xs}]
this guarantees that the result of applying be1 to a stage-

level predicati ve phrase will be an individual-level
predicate, but one which nevertheless yields the same kind
of 'existential reading' as a stage-level predicate (by the
rule in (86». This translation is of course impossible to
justify on the evidence of sentences like (90) and (91);
these would (correctly) receive the translations (92) and
(93) even if be1 were assumed to be a semantically empty
expression of category IVs/PRED s , owing to the availability
of rule (86) (and to the fact that asleep in that room and
on the table are stage-level).
(90) Dogs are asleep in that room.

(91) Sack lunches are on the table.
(92) Vxs[R(x s , dogs') & asleep-in-that-room'(x s )]

(93) Vxs[R(x s , sack-lunches') & on-the-table'(x s )]
But the absence of predicates like those in (88) from the

see them drunk, progressive, and existential there
constructions strongly suggests that they are not stage-
level; Carlson's translation for be, makes it clear how this
could be so •
. Be, is thus regarded as an expression of category
IV1/PRED s • By rule (94), it is combined with stage-level
predicative expressions of three different categories to
produce individual-level intransitive verb phrases.
(94) S25. If a E PIVi/PREDs and 8 E PpRPL s , PADJs, or

Ppps , then F25 (a,s) E PIVi, where F25 (a,s) is
CON(a,B).
Examples: F25 ([be], walking home) = [be] walking home
F 25 ([be], awake) = [be] awake
F25 ([be], in the barrel) = [be] in the barrel
T25. If a e; PIVi/PREDs, 6 e; PpRED s , and a,6

translate as a',6', then F25 (a,6) translates
as a'("8').
Examples: F25 ([ be], walking home) translates as

AxiVxs[R(x s ,xi) & walking-home'(x s )]
F25 ([be], awake) translates as
hiVxs[R(xs,x i ) & awake'(x s )]
F25 ([be], in the barrel) translates as
AxiVxs[R(x s ,xi) & in-the-barrel'(x s )]
In Chapter IV, I shall argue that be1 is, among other

things, the be of the progressive; there are, however, at
least three be's distinct from be1.
3.2.2. Be2 € Plyi/PREDi

Members of PRED i may appear in predicative constructions, as
in (95)-(98).
(95) John is intelligent. (intelligent € PADJi)

(96) John is a sailor. (a sailor E PNOM )
(97) John is from Wales. (from Wales e; Pppi)
(98) John is loved by Mary. (loved by Mary E PPSPLi)
The be 2 in (95)-(98) cannot be identified with be 1, because

the latter is of the wrong category to combine with
individual-level predicative expressions such as
intelligent, a sailor, from Wales, and. love4 by Mary.
Accordingly, be 4 is assigned to category Iy1/PRED1 • By rule
(99), be2 comblnes with four sorts of individual-level
76 CHAPTER II
predicative phrases to produce individual-level verb

phrases:
(99) 326. If a €: PIVi/PREDi and S

PADJi, PNOM ' Pppi,
€:
or PpSPLi, then F 26 Ca,S) E PIVi, where

F26 (a,S) is CONCa,S).
Examples: F 26 ([be], intelligent) = [be] intelligent

F26 ([be] , a sailor) = [be] a sailor
F 26 C[be], from Vales) = [be] from Vales
F 26 C[be], loved by Mary) = [be] loved by Mary
T26. If a €: PIVi/PREDi, S E PpREDi, and a,S
translate as a',S', then F 26 Ca,S) translates
as a,(A S ').
Be2 is probably genuinely without semantic content; that is,

it is apparently to be translated as (100).
(100)
If be2 translates in this way, then (99) assigns [be]

intelligent, [be] a sailor, [be] from Vales, and [be] loved
by Mary the translations intelligent', a-sailor', from-
Vales', and loved-by-Mary', respectively.
The careful reader will have noted that some
expressions consisting of be plus a predicati ve phrase do
occur as stage-level predicates--among them, be a bastard,
be silly, be arrested, and be awarded the Medal of Honor.
These instances of be are distinct from both bel and be~;
observe for example, that they may cooccur with bel' as ~n
(101) and (102).
(101) He iS 1 being polite.

(102) He iS 1 being forced to sign the contract.
3.2.3. Be 3 E PIVs/PREDi
The be3 in be a bastard, be a hero, be polite, be silly, be
obnox10us, and so on, is the 'active be' postulated by
Partee (1977:305-310).20 As she shows, expressions
resulting from the combination of be 3 with some predicative
expression are syntactically and semantically very different
from normal copulative phrases.
(i) They behave like active predicates in that they
may occur in the progressive:
(103) *John is being fat again.

*John is being alone again.
John is being a bastard again.
John is being obnoxious again.
(ii) They require animate subjects:
(104) John is noisy today.

The river is noisy today.
John is being noisy today.
*The river is being noisy today.
(iii) Their subjects are understood agentively;
(105 ) Be polite!
John is trying to be polite.
John is deliberately being polite.
We persuaded John to be polite.
that is, be3 has a meaning something like that of act

(like) :
(106 ) John is being/acting silly.

(107 ) John is being/acting like a jerk.
For this reason, be3 doesn't combine readily with

predicative phrases expressing properties that an individual
can't help having:
(108 ) *Be healthy!

*John is trying to be healthy.
*John is deliberately being healthy.
*We persuaded John to be healthy.
An additional property which distinguishes be 3 but which, to

my knowledge, hasn't been pOinted out before is that (iv)
be combines most readily and perhaps exclusively with
inaividual-level predicates (even though the resulting
78 CHAPTER II
combination is itself a stage-level predicate):
(109 ) a good citizen

a bastard
a hero
polite
silly
John is being foolish again.
obnoxious
paranoid
defensive
?*available
?*ready
This last fact suggests an interesting analysis of be 3 •

Consider sentence (110).
( 11 0) John is being a real pain in the neck.
This can, intuitively, be paraphrased with (111),
( 111) John is being the kind of person who is a real

pain in the neck.
or, in Carlson's idiom, with (112) (where the progressive is

for the moment ignored, and R' denotes the relation between
an object and the kind(s) it realizes).
(112 ) VXS[R(x S, John') & Vxk[Axo o [real-pain-in-the-

neck'(xo) ~ R'(Xo,X k )] & R(xs,x k )]]
(112) suggests a translation like (113) for be3:
( 113)
According to (113), be 3 doesn't shift stage-level properties

to the individual level (as be} does), but rather has just
the reverse property of shi ting an individual-level
property down to the stage level; that i.s, be j is an
expression of category IV s /PRED 1 , not of IV 1 /PRED. This
analysis embodies property (iv) above, and explains why the
result of combining be3 with a predicative phrase has all
the properties of stage-level predicates, e.g. (i). This
analysis doesn't account for properties (ii) and (iii);
these, however, are probably to be regarded as
conventionally implicated aspects of the meaning of be 3

rather than as part of its asserted meaning.
Be3 combines with an individual-level predicate in
accordance with the following rule:
(114) S24. If a € PIVs/PREDi and a € PNOM or PADJi, then

F24 (a,a) € PIVs, where F24 (a,a) is CON(a,a).
Example: F24 ([be], obnoxious) = [be] obnoxious
T24. If a € PIVs IPRED i , a € PpREDi, and a, a

translate as a', a', then F 24 (a, a) translates
as a'("'S').
Example: F24 ([be], obnoxious) translates as

AxsVxk[Axo O[obnoxious'(xo) .. R'(xo,x k )] &
R(Xs,x k )]
There is one final be which, like be 3 • joins with a

predicative phrase to produce a stage-level intransitive
verb phrase.
The be4 in be arrested, be awarded the Medal of Honor, be

forced to sign the contract, be thrown into the river. and
so on, is, of course, the be of the passive. To understand
why these passive be-phrases are stage-level predicates. it
is important to recognize that there are intuitively at
least two 21 sorts of transitive verb phrases (cf. Carlson.
pp.186ff): those like chase, which appear to denote
relations between stages (cf. the 'existential reading' of
dogs in Harry chased dogs all day), and those like love,
which seem to denote relations· between individuals (cf. the
fact that Harry loves dogs expresses a generic fact about
dogs). To capture this intuition. two categories of
transitive verb phrases are introduced:
so CHAPTER II
the basic category of [chase], [hit],

stage-level transitive [kiss], •••
verb phrases
the basic category of [love], [admire],
individual-level tran- [fear] ,
sitive verb phrases
Members of TVs are interpreted as functions from stages

to sets of stages. The rule (115) combining a member of TV s
with a noun phrase produces a stage-level intransitive verb
phrase:
( 115) 328. If a E: PTVs and B E: PT' then f'2S(a,B) e; Prvs,

where F28(a,B) is RWRAP(a, OBJ(B».
Example: F28 ([hit], John) = [hit] John
T28. If a € PTVS' S € PT, and a,B translate as

a',S', then F2S (a,B) translates as
'Api'AX s pi{"')..YiVys[R(yS,yi) &
a'(ys)(xs)]}("S').
Example: F28 ([hit], John) translates as

'AxSVys[R(yS, John') & hit'(ys)(x s )]
Members of TV i are interpreted as functions from

indi~iduals to sets of individuals. The rule (116)
combining members of Tyi with noun phrases yields an
expression of category IV~:
(116 ) 329. I f a € PTvi and B € PT, then F29 (a,B) € PIVi,

where F29 (a,S) is RWRAP(a, OBJ(S».
Example: F29 ([love], John) = [love] John

(116 ) [continued]
T29. If a E PTVi, 6 E PT , and a,6 translate as

a', B', then F 29 (a, B) translates as
A piA Xo pi{ "Ayi[ a'(yi)(xo)] H"B').
Example: F 29 ([love] John) translates as

Axo[love'(John')(xo)]
Following Bach (1980), I assume that passive rules

operate on transitive verb phrases to produce past
participial phrases with passive meaning. Given the
assumption that there are two distinct categories of
transitive verbs--TV 1 and TVs--it is clear that separate
passive rules must be formulated for these categories. The
passi ve rules for stage-level transi ti ve verb phrases are
(117) and (118): the former produces agentless passives
while the latter produces passives with overt by-phrases.
( 117) S41. If a E PTVs,

then F 41 (a) £ PPSPLs, where
F 41 (a) is DB(EN (a».
Example: F 41 ([knock] over) = knocked over
T41. If a e: PTVs and a translates as a', then

f41 (a) translate3 a3 t-xsVyS[ a'(xS)(yS)].
Example; f41([knock] over) tran31ate3 as

t- xSVys[knock-over' (x s ) (ys)]
(118) 842. I f a e: P TV san d B e: P T' the n F 42 (a, B)

e: PPSPLs, where F42 (a,B) is
CON(DB(EN(a», CON(by, OBJ(B»).
Example: F 42 ([knock] over, John) = knocked over by

John
T42. If a £ PTVs, B £ PT , and a, B translate as

a',B', then F 42 (a,B) translates as
h S [ B'(">.yiVys[R(yS,yi) & a'(xs)(ys)])].
82 CHAPTER II
( 118 ) [continued]
Example: F42 ([knock] over, John) translates as

AxSVys[R(yS, John') & knock-over'(xs)(ys)]
Similarly, the passive rules for individual-level t,ransitive

verb phrases are (119) and (120).
( 119 ) S43. I f a E PTvi, then F 43 (a) E PPSPL i, where

F 43 (a) is DB(EN(a».
Example: F43 ([love]) = loved

T43. I f a E PTvi and a translates as a', then
F 43 (Ct) translates as AxiVyo[a'(xi)(yo)J.
Example: F43 ([loveJ) translates as

".xiVyo[ love' (xi) (yo)]
(120 ) S44. If a E PTvi and B E PT ' then F 44 (a,B)

E PPSPLi, where F 44 (a,B) is
CON(DB(EN(Ct», CON(by, OBJ(B»).
Example: F 44 ([loveJ, John) = loved by John
T44. If Ct E PTVi, B £ PI' and Ct, B translate as

a',S', then F 44 (a,S) translates as
Axi[ B'("AYo[ a'(xi)(yO)])].
Example: F44 ([love] , John) translates as

Axi[love' (xi) (John')]
The stage-level past participial phrases produced by

(117) and (118) join with be4--passive be--to form stage-
level intransitive verb phrases. This fourth be, unlike be 1
and be 3 , doesn't change the 'level' of the expression it
joins with; it is of category IVs/PRED s • Like be 2 , it is
semantically empty.
( 121) S23. If a E PIVs/PREDs and B E PpSPLs, then
F23 (a,B) E P1Vs, where F23 (a,B) is CON(a,B).
T23. If a E PIVs/PREDs, BE PPREDs, and a,B

translate as a',B', then F23 (a,B) translates
as a'("B').
The individual-level participial phrases produced by

rules (119) and (120) join with be2 (e; PIVi /PREDi) to form
indi vidual-level intransi ti ve verb phrases. (See rule (99)
above. )
In this analysis, the phrase be arrested is stage-
level: because [arrest] is an expression of category TV s ,
F41 ([arrest]) and F23 ([be], arrested) are both stage-level.
Note, however, that, not all passive intransitive verb
phrases are stage-level in this account: be loved, for
eXqmple, is individual-level because [love] is a member of
TV1, and hence F 43 ([love]) and F~6([be], loved) are
indi vidual-level expressions. This 1S correct: be loved
satisfies none of the diagnostics for stage-predicates, as
(122)-(124) show.
(122 ) *Mary saw John being loved.

(123 ) *There was a man being loved.
(124 ) *Jane is being loved.
The object of this rather circuitous discussion of be

has been to show that the following four be's, which one
would like to be able to distinguish because of the varying
ways in which they interact with Carlson's stage/individual
distinction, can be distinguished on independent grounds.
be1 e: BIVi/PREDs
be2 e; BIV i /PREDi
be3 e; Brvs /PREDi
be4 e; B1VS /PREDs
Though the semantic details of the distinction between
stage-predicates and individual-level predicates are quite
complex, the relevance of this distinction to the
weak/strong bifurcation of free adjuncts is clear, as I
shall now show.
84 CHAPTER II
3.3. THE STAGE/INDIVIDUAL DISTINCTION AS A SEMANTIC

CORRELATE OF THE WEAK/STRONG DISTINCTION
The importance of the distinction between stage-level and

indi vidual-level predicates for the semantics of free
adjuncts is that it appears to coincide precisely with the
difference between weak and strong adjuncts. Observe that
if the predicates listed in (75) are used as free adjuncts,
they uniformly show the entailment properties of weak
adjuncts;
(75) Stage-level predicates:
drunk
alone
asleep
sick in bed
naked
walking to the store
lying on the beach
carrying a load of over 100 lbs.
driving this sort of car
wearing his new outfit
standing on a chair
noticing the snake
sitting in the garden
hearing this song
in his maroon suit and white shoes
on a stage
being a bastard
being silly
being arrested
being awarded the Medal of Honor
the predicates in (76) and (88), on the other hand, give

rise only to strong adjuncts.
(76) Individual-level predicates:
being a master of disguise

being a sailor
being something of an insomniac
weighing over 200 lbs.
having unusually long arms
having a wooden leg
(76) [continued]
standing over six feet tall

being fat
being intelligent
being American
(88) being drunk

being alone
being asleep
being sick in bed
being naked
being in his maroon suit and white shoes
being on a stage
A comparison of the distinction between stage-level and

individual-level predicates and the distinction between
statives and nonstatives reveals the precision with which
the former correlates with the difference between strong and
weak adjuncts: the fact that all adjuncts with nonstative
predicates are weak follows from the fact that nonstatives
are in general stage-level; and the fact that the stative
predicates exemplified in (69)-(72) give rise to weak rather
than strong adjuncts follows from the fact that these
predicates are, in Carlson's system, stage-level. Such
facts unequivocally point to the conclusion that strong
adjuncts are inherently distinguished from weak adjuncts in
that they ul timately deri ve from indi vidual-level rather
than stage-level predicates.
In the fragment developed here, this correlation
between Carlson's stage/individual distinction and the
weak/strong distinction is easily formalized. Recall that
two sorts of adjuncts have been proposed so far: (i)
conditional adjuncts, which belong to the category t' of
conditional clauses and may therefore condition the
interpretation of a modal (according to rule (42»; and (ii)
adsentential adjuncts, whose category membership remains to
be established. In section 2.2, the distinction between
weak and strong adjuncts was reduced to the following
generalization: whereas both weak and strong adjuncts may
serve as adsentential adjuncts, only weak adjuncts may serve
as conditional adjuncts. In view of the correlation between
weak adjuncts and stage-level predicates, this
generalization can now be formalized as a restriction on the
derivation of conditional adjuncts: all such adjuncts must
86 CHAPTER II
deri ve from stage-level predicates (i.e. members of PRED s ).

See Chapter III, section 2 for an explicit statement of the
rule generating conditional adjuncts, which incorporates
this restriction.
This analysis of free adjuncts in modal contexts
affords some extremely subtle insights into the semantics of
predicative expressions in English; most saliently, it
predicts that any predicative phrase exhibiting the
distributional characteristics typical of stage predicates
(appearing in the progressive, existential, and see them
drunk constructions) will also be able to serve in the
manner of an if-clause in modal contexts (but that
predicative phrases of other sorts will not be able to). In
the following 'case study', the accuracy of this prediction
for prepositional adjuncts headed by as and with is
demonstrated; moreover, I show that the proposed analysis
reveals a fine semantic connection between these two
prepositions on the one hand and the verbs be and have on
the other.
3.4. ADJUNCTS WITH AS AND WITH
As was seen in section 3.2.1, bel (E Prvi/PRI;:Ds) occupies a

special place in Carlson's system, since lt effectively
converts stage-level properties to properties of
individuals; thus, the adjective clean-shaven is a predicate
of stages, while the verb phrase be clean-shaven is
individual-level. This is reflected in the fact that, as
adjuncts, the former expression is weak, while the latter is
strong:
(125 ) Clean-shaven, Harold would look something like my

brother.
[on one interpretation, equivalent to (127)]
(126 ) Being clean-shaven, Harold would look something

like my brother.
[entails that Harold is clean-shaven; not
interpretable as (127)J
(127 ) If he were clean-shaven, Harold would look

something like my brother.
Thus, it is possible to have weak and strong 'alternants'

for many adjuncts: asleep/being asleep, drunk/being drunk,

in bed/being in bed, etc.
Superficially, however, certain strong adjuncts seem to
lack weak alternants. First, since predicative noun phrases
are inherently individual-level (and thus combine with be 2
(£ Prvi/PREDi) rather than be 1 ), adjuncts consisting of
predlcative noun phrases (with or without be 2 ) are uniformly
strong--
(128) (Being) a blonde, Mary might look something like

Jane.
[entails that Mary is a blonde; not
interpretable as (129)]
(129 ) If she were a blonde, Mary might look something

like Jane.
Second, strong adjuncts headed by verbs which are basically

indi vidual-level (e.g. have, weigh, own, etc.) apparently
lack weak alternants:
(130 ) Having green eyes, Mary might look something like

Jane.
[entails that Mary has green eyes; not
interpretable as (131)]
( 131) If she had green eyes, Mary might look something

like Jane.
Despite this apparent gap in the system of weak and

strong adjuncts in English, careful consideration reveals
that there are in fact expressions which function precisely
as weak al ternants for adjuncts such as those in (128) and
(130); these are predicative prepositional phrases headed by
as and with.
Consider, first, the preposition as. When it combines
with a noun phrase, the resulting expression may appear in
the usual diagnostic environments for stage-level
predicates, including the see them drunk and existential
constructions:
(132) Twenty years ago, we saw John as a young man.

(133 ) There were faculty members as chaperones.
Thus, according to the proposed analysis of free adjuncts in

88 CHAPTER II
modal contexts, as-phrases should be able to serve as weak

adjuncts, conditioning the interpretation of a modal in the
manner of an if-clause. This prediction is correct: (134),
on its most natural interpretation, is equivalent to (129).
(134) As a blonde, Mary might look something like Jane.
The equivalence of (134) and (129) reveals that as a

blonde serves precisely as a weak alternant for the strong
adjunct (being) a blonde (cf. (128». Thus, it can be seen
that as and bel have complementary functions: 22 whereas bel
shifts proper ies from the stage level to the individua
level, as does just the reverse, lowering individual level
properties (such as the intension of the predicative noun
phrase a blonde) to the stage level; as an expression of
category pps /NOM, as might therefore be assigned the
following translation.
(135 )
The complementarity of as and be1 is paralleled by that

of with and the individual-level verb have. Prepositional
phrases headed by with appear in the usual environments for
stage-level predicates:
(136) I saw him with a mustache.

(137 ) There were three girls at the party with green
eyes.
The proposed analysis thus predicts that with-phrases should

serve as weak adjuncts in modal contexts. This prediction
is borne out: the adjunct in (138) restricts the
interpretation of the modal might fully in the manner of the
conditional clause in (131).
(138 ) With green eyes, Mary might look something like

Jane.
Thus, with green eyes functions, in effect, as a weak

alternant for the strong adjunct having green eyes (cf.
(130». Accordingly, with in (138) might be translated as
in (139), where have' is the translation of the individual-
level verb have: 23
(139 )
As this brief discussion shows, the proposed analysis
of free adjuncts in modal contexts not only correctly
predicts that as- and with- phrases may serve as weak
adjuncts; it also reveals an intimate semantic connection
between as and be, and between with and have.
4. Chapter Summary
The foregoing discussion began with the observation

that free adjuncts may playa conditional role in modal
contexts. However, it was shown in section 1 that a
conditional interpretation is not possible for all adjuncts;
a distinction was made between weak adjuncts--those
admitting a conditional interpretation--and strong adjuncts,
for which such an interpretation is impossible.
In section 2, the difference between weak and strong
adjuncts was elucidated within the framework of Kratzer's
theory of conditional necessity and possibili ty: in this
framework, weak adjuncts playing a conditional role may be
viewed as restricting the interpretation of a modal in the
manner of an if-clause. It was proposed that such
conditional adjuncts and if-clauses belong to the same
category (t'); if this is assumed, then the conditional
character of an adjunct such as that in (140) follows
automatically from rules which are independently necessary
for the interpretation of modal sentences.
(140 ) Thus, written as a deletion rule, Gapping might

look like (28) •••
(Williams 1977:110)
Strong adjuncts (and weak adjuncts not serving to restrict

the interpretation of a modal) were informally analyzed as
adsentential modifiers headed by an implicit, variable
subordinator K(L), whose logical properties guarantee the
factive entailments associated with adjuncts of this kind.
In section 3, the distinction between weak and strong
adjuncts was shown to correlate precisely with Carlson's
independently motivated distinction between stage-level and
individual-level predicates: weak adjuncts have stage-level
predicates, and strong adjuncts have individual-level
predicates. Given this correlation, the semantic
differences between strong and weak adjuncts in modal
contexts may be viewed as the consequence of a restriction
go CH,4.PTER II
on the derivation of conditional adjuncts: all such

adjuncts must derive from stage-level predicates.
Certain issues raised here remain to be addressed. For
instance, is the requirement that conditional adjuncts
derive from stage-level predicates merely an arbitrary fact
about English grammar, or does it follow in some principled
way from the nature of Carlson's ontological distinction
between stages and individuals? Also, how do language users
go about resolving the semantic indeterminacy (represented
by 'K(L)') inherent in the interpretation of adsentential
adjuncts? Why, for example, are strong adjuncts such as
those in (9a)-(12a) so naturally inferred to play the role
of because-clauses? These are, ultimately, pragmatic rather
than semantic issues; I therefore defer discussion of them
until Chapter VI, where the pragmatic problems posed by
absolute constructions are centrally addressed.
Footnotes
1. Propositions are, as usual, taken to be sets of

possible worlds; a proposition p is thus true in a world w
iff w is a member of p. A proposition p follows logically
from a set A of propositions iff p is true in every world in
which every member of A is true. A set A of propositions is
consistent iff there is a world in which all members of A
are true. A proposition p is compatible with a set A of
propositions iff A U {p} is consistent.
2. She shows that Lewis' (1973) account of counterfactuals

is actually a special case of her own account of conditional
neceSSity! his systems of spheres can, without
modification, be regarded as simply a special sort of
conversational background, and her truthconditions for
would-conditionals, if relativized to a conversational
background of this sort, are provably equivalent to his
truthconditions for counterfactuals, if one ignores those
conditionals which are, in Lewis' system, vacuously true.
3. In more recent work, Kratzer (1981) has introduced a

number of subtle refinements into her treatment of modal
semantics. Since the factors motivating these refinements
do not bear directly on the issues addressed here, I shall
continue to assume the earlier, Simpler version of her
theory.
4. Here and throughout, a basic category is one which

cannot be defined in terms of other categories by means of a
slash. It isn't necessarily a lexical category.
5. See Chapter III, section 1.2.1 for discussion of the

rules of tense.
6. Throughout this study, I will employ the convention of

abbreviating the translation of a complex expression such as
the tenseless sentence John loses his license with a
hyphenated, primed expression: John-loses-his-license'.
(This is a departure from Montague's use of the prime: in
PTQ, primed expressions are always constants of intensional
logic.) The convention employed here is necessitated by the
length of many of the example sentences.
In many cases, the 'translation' which I provide for an
expression is not strictly a translation, but a logically
equivalent formula resulting from multiple applications of
lambda conversion, etc. This convention is likewise
assumed for expository convenience.
7. This analysis is based on the assumption that, in modal

conditionals, if-clauses have no meaning beyond the
proposi tion they express--essentially, that if' has no
meaning in such sentences; the admissibility of this
assumption is disputable, however. One might be willing to
treat if as a semantically empty expression in sentences
like (25 )-(28) on the grounds that they can be interpreted
perfectly adequately under this assumption. According to
this position, if serves to signal the occurrence of a
conditional clause without contributing in any way to its
meaning, much as a complementizer serves to signal the
occurrence of a sentential complement. In defense of this
position, one coUld cite subjunctive conditionals like (1),
whose antecedent clause is indistinguishable in meaning from
that of (ii) even though it lacks if;
(i) Had Mary seen the morning paper, she would have
known all about the fire.
(ii) If Mary had seen the morning paper, she would have
known all about the fire.
in (i), inversion takes the place of if as the signal of a

conditional clause. But the fact that (25)-(28) can be
interpreted satisfactorily under the assumption that if is
92 CHAPTER II
semantically empty would not be an argument in favor of this

assumption if there were some piece of meaning whose
presence correlated recurrently with that of if. M. L. Geis
(ms) has suggested that if correlates systematically wi th
the meaning' in the event that'I'in any event in which' on
the evidence of indicative conditionals; (iii), he argues,
is synonymous with both (iva,b).
( iii) I will leave if you leave.

(i v) a. I will leave in the event that you leave.
b. I will leave in any event in which you leave.
Notwithstanding Geis' extensive arguments for regarding if-

clauses as relative clauses like those in (iva,b), I admit I
have certain reservations about his claim. In particular:
(a) It's not clear how the equivalence of (i) with (ii)
can be accounted for if Geis' claim is accepted; without
independent motivation, an 'if-deletion' transformation
would beg the question.
(b) It's not clear that Geis' paraphrase for if is
valid for all kinds of condi tionals--in particular, modal
conditionals like those in (25)-(28): (va,b) seem odd as
paraphrases for (ii);
( v) a. ??In the event that Mary had seen the morning

paper, she would have known all about the
fire.
b. ??In any event in which Nary had seen the
morning paper, she would have known all about
the fire.
and, more importantly, it's not obvious that the meaning

attributed to if can be used to derive an acceptable model-
theoretic interpretation for conditionals like (ii).
(c) Geis' claim apparently precludes the identification
of condi tional if wi th 'whether if', an identification not
ruled out under the assumption that if is semantically empty
and merely serves as a signal of certain sorts of clauses.
Nevertheless, the intuitive appeal of Geis' analysis is not
negligible.
The question of how to analyze if raises what I believe
to be one of the most critical problems for compositional
semantics, namely the problem of redundancy. In the
analysis of modal-free conditionals like (vi), it seems hard
to avoid assigning some sort of semantic content to if--
perhaps the content suggested by Geis;
(vi) Mary knows Bill if Jane does.
in the analysis of (25)-(28), on the other hand, the modal

provides all the necessary information about the logical
role of the conditional clause, and thus makes if redundant
(and hence eliminable, as in (i)). How can such redundant
elements be incorporated into the semantic composition of an
expression? That is, how can two or more expressions make
the same contribution to the meaning of an expression of
which they are part--a contribution which anyone of them
could make independently? This problem is all the more
serious because it arises so commonly--as, for example, in
the double negative construction; thus, it's not obvious how
to provide an account of the synonymy of (viia-c) which is
compositional but in which not and no each have a constant
interpretation.
(vii) a. I don't have any money.

b. I have no money.
c. I don't have no money.
[= a,b in some dialects]
In the absence of a theory of redundancy, there is little

choice but to regard if as a semantically empty expression
in conditionals like (25 )-(28), given the assumed treatment
of modals.
8. See the Appendix for the formal statement of this rule.
9. This assumption would, of course, have to be suspended

in a systematic study of the control possibilities to which
free adjuncts are subject, but this is a concern that goes
beyond the scope of the present study.
10. The modal in (56) lacks an overt conditional clause, so

that its interpretation is relativized only to the (set of
all consistent subsets of the) conversational background
implici t in the context of its use; that is, rules (41) and
(43) rather than (42) and (44) are relevant for the analysis
of the main clause in (56).
11. See section 2 of Chapter VI for a discussion of

connective adverbs and their relevance to inferring the
94 CHAPTER II
logical role of a free adjunct or absolute.
12. Note that L(p) (q) can entail both ~p and ~ q at a

certain index even if the events (or states of affairs)
described by p and q do not actually take place at that
index, since p and q may both contain temporal operators.
Cases of just this kind will be discussed in Chapter III,
section 2.3.
13. Of course, the stand-si t-lie class of predicates isn't

unambiguously stati ve in the traditional syntactic sense--
cf. the acceptability of such predicates in the progressive.
See Dowty (1979: 173fO.
14. Of course, bare plurals and proper names denote

property-sets of entities, strictly speaking.
15. An exception is what Carlson (p.184) calls the

'comparative progressive':
(i) Kids are getting smarter.
16. An apparent exception are predicates like be a good

citizen in He's being a good citizen. These, however,
constitute a special construction. Vide infra.
17. Note here that I take prepositional phrases to be

predicative expressions, unlike Montague, who, in prQ, takes
them to be adverbial in nature. While I don't wish to argue
that all prepositional phrases are predicative, it is clear
that some are: John is in his white suit, Larry is under
the garbage can; it is this predicative use of prepositional
phrases that is of interest here. (I should note in passing
that the adverbial use of a prepositional phrase may be
reducible to the predicative use: for example, Mary is
playing in the yard might be analyzed as (i) rather than as
(ii) •
(i) playing'(Mary') & in-the-yard'(Mary')

(ii) in-the-yard,(Aplaying')(Mary')
Geis (1974) has argued extensively for just such an analysis

of locative prepositional phrases.)
18. There is no NOMs because predicative noun phrases are

always individual-level, as mentioned above.
19. An additional fact about the predicates in (88)

suggesting that they are individual-level predicates will be
discussed in Chapter III.
20. Cf. also Dowty's (1979:115,185) discussion.
21. Here I ignore verbs like seek, which are intensional

with respect to object position. See Carlson, pp.189ff.
22. In his analysis of with-absolutes, McCawley (1983:281f)

postulates a deri vational relationship between as and be,
with the former serving as a transformational replacement
for the latter. While it is clear that some sort of
connection exists between be and as (that they are, in some
sense, 'alternants' with respect to the weak/strong
distinction), they have distinct semantics, making a
derivational relationship dubious. Indeed, they contrast in
the environment under discussion:
(i) a. Being a blonde, Mary might look something

like Jane.
b. As a blonde, Mary might look something like
Jane.
The fact that as-phrases (unlike other members of

PRED s ) do not join with bel--that is, the fact that
expressions such as those in (ii) are unacceptable--can
perhaps be explained by appealing to some kind of principle
of conversational economy (cf. McCawley (1978»: since an
expression of the form 'bel as NOM' (e.g. be as a blonde)
will always be equivalent to the corresponding expression
'be NOM' (e.g. be a blonde), the latter, being
syntactically and semantically simpler, is systematically
preferred.
(ii) *Mary is as a blonde.

*Mary wants to be as a blonde.
23. One is reminded of Fillmore's (1969:37lff) arguments

for deriving have from be with. According to the analysis
developed here, such a derivation would be quite accurate,
from a semantic perspective: 'with T' has as its intension
a property of stages; bel' however, can shift this property
96 CHAPTER II
to the individual level, so that 'be1 with T' and 'have T'
will both have properties of individuals as their
intensions. (In the present fragment, however, the
intension of 'have T' (unlike that of 'be1 with T') is,
specifically, a property of objects.)
McCawley (1983) argues that with-phrases having a
possessional meaning are derived from augmented absolutes
through the deletion of the underlying subject and the main
verb have; thus, the adjunct with green eyes derives from
(something like) (i) under his analysis.
(i) with [ Mary having green eyes]
Ishihara (1982:63f) proposes a similar analysis, although

she assumes that the underlying absolute phrase has PRO
(rather than a lexical NP) as its subject:
(ii) with PRO having green eyes ]
But by virtue of their individual-level predicates,

absolutes such as those in (i) and (ii) (like the adjunct
having green eyes) are strong, as will be seen in Chapter V,
section 1; for this reason, a derivational relation between
(i)/(ii) and the weak adjunct with green eyes seems dubious.
See footnote 22 for analogous remarks concerning be and as.
See Ruwet (1978:194ff) for relevant arguments against a
similar rule for the deletion of avoir in French; cf. also
Riemsdijk (1978:66ff) for arguments involving Dutch.
C HAP T E R I I I
TENSE AND THE INTERPRETATION

OF FREE ADJUNCTS
In this chapter, the status of free adjuncts within the

system of English temporal semantics is investigated. Three
issues are centrally addressed here.
(i) The first of these is the temporal reference of
free adjuncts. While the time of the event (or state of
affairs) described by a free adjunct may coincide with that
of the event (state of affairs) described by its
superordinate clause--as in (1 )--i t is clear from examples
such as (2) that the time reference of a free adjunct is in
principle independent of that of the main clause.
(1) Lying on the beach, John smoked his pipe.

(2) Setting sail for the island in the fall of 1740,
he reached his destination in the spring of 1741.
Thus, the translation of a free adjunct must contain some

sort of temporal operator, analogous to that signalled by
tense in a finite clause.
In section 2 of this chapter, rules for the derivation
of free adjuncts are proposed which take account of this
temporal characteristic of free adjuncts; specifically, I
propose that all free adjuncts derive from temporal
abstracts. (Temporal abstracts are expressions having the
form of a sentence, free adjunct, or absolute, but denoting
a set of time intervals; as will be seen throughout this
chapter, they play a central role in the account of English
temporal semantics developed here.)
The remaining two issues addressed in this chapter
pertain to the distinction between weak and strong adjuncts.
Recall that in the preceding chapter, it was shown that two
varieties of free adjuncts can be distinguished in modal
contexts: weak adjuncts, which may restrict the
interpretation of a modal in the manner of an if-clause; and
strong adjuncts, which may not. In this chapter, I show
that weak and strong adjuncts are distinguished in two
additional, nonmodal environments: in sentences containing
97
98 CHAPTER III
adverbs of frequency (sometimes, often, never, always, ••• )

and in generic/habitual sentences.
(ii) Consider first the sentences in (3)-(6), each of
which contains a frequency adverb.
a. Being a sailor, John sometimes smokes a pipe.

b. Lying on the beach, John sometimes smokes a
pipe.
(4 ) a. Weighing four tons, our truck often makes the
bridge shake.
b. Carrying a load of over 1500 Ibs., our truck
often makes the bridge shake.
a. Having only ten minutes between her morning
and her afternoon classes, Anne never eats
lunch.
b. Studying for finals, Anne never eats lunch.
(6) a. A major stockholder of Texas Instruments,
Jane always uses a calculator.
b. Figuring her taxes, Jane always uses a
calculator.
Here, the ~ sentences have strong adjuncts, and the .!?

sentences, weak. Intuitively, the role played by the weak
adjuncts in the.!? sentences is very different from that
played by the strong adj uncts in the ~ sentences; in
particular, the frequency adverbs in these sentences seem to
interact in strikingly different ways with weak as opposed
to strong adjuncts. In sentence (3b), for example, the weak
adjunct lying on the beach seems to restrict the se.t of time
intervals qUantified by the frequency adverb sometimes, so
that sometimes specifies the frequency with which time
intervals at which the main clause is true (I.e. at which
John smokes a pipe) coincide with intervals at which the
adjunct is true (I.e. at which John is lying on the beach);
<3b) is thus naturally paraphrased by sentence (7), in which
the adjunct is replaced with an overt when-clause.
(7) When he is lying on the beach, John sometimes

smokes a pipe.
In sentence (3a), on the other hand, the strong adjunct does

not seem to restrict the set of intervals quantified by the
frequency adverb; (3a) certainly cannot be plausibly
p.araphrased with a sentence such as (8), in which the
TENSE AND FREE ADJUNCTS 99
adjunct being a sailor is replaced with a temporal adverbial

clause.
(8) When he is a sailor, he sometimes smokes a pipe.
On the contrary, the strong adjunct in (3a) seems most

naturally paraphrased with a because-clause. Similar
remarks hold for the sentences in (4)-(6).
In section 3 of this chapter, I shall argue that
frequency advebs denote binary relations between sets of
time intervals; accordingly, they require two arguments, of
which the first is typically a temporal adverbial clause (as
in (7)) and the second, a temporal abstract. If frequency
adverbs are analyzed in this way, then the intuitive
difference between weak adjuncts and strong adjuncts in
sentences such as (3)-(6) can be easily reconstructed in
formal terms: weak adjuncts may serve as the first argument
of a frequency adverb, while strong adjuncts may not.
(iii) Generic/habitual sentences are yet a third
environment in which weak adjuncts and strong adjuncts show
divergent behavior. Consider the sentences in (9)-(12).
(9) a. Being a businessman, John smokes cigars.

b. Lying on the beach, John smokes cigars.
( 10) a. Being a drunk, he drives very dangerously.
b. Drunk, he drives very dangerously.
( 11) a. Having a rather decrepit engine, the truck
makes funny noises.
b. In first gear, the truck makes funny noises.
(12 ) a. Weighing over 200 lbs., Mary looks fat.
b. Wearing her new outfit, Mary looks fat.
Here again, the ~ sentences contain strong adjuncts, and the

£ sentences, weak adjuncts. On their generic or habitual
interpretation, the ~ sentences are understood as expressing
generalizations about those time intervals at which their
adjuncts are true. Sentence (9b), for instance, is
interpreted as making a particular generalization about
intervals at which John is lying on the beach: the
generalization is that John smokes cigars at such intervals.
Thus, (9b) is naturally paraphrased by (13), in which the
weak adjunct lying on the beach is replaced with a temporal
adverbial clause.
(13) When he is lying on the beach, John smokes cigars.

100 CHAPTER III
Similar remarks hold for the other b sentences. The a

sentences, on the other hand, are not understood as
expressing generalizations about those intervals at which
their adjuncts are true. (9a), for example, is not a
generalization about intervals at which John is a
businessman; on the contrary, the most likely interpretation
for the strong adjunct being a businessman in (9a) is that
of a because-clause. The same is true of the strong
adjuncts in (10a)-(12a).
In section 4 of this chapter, I argue that
generic/habitual sentences such as (13) may be perspicuously
analyzed as possessing an implicit 'generalization operator'
G" denoting a special binary relation between sets of time
intervals. Like a frequency adverb, this operator requires
two arguments, the first, a temporal adverbial clause, and
the second, a temporal abstract. Under this analysis, the
intuitive difference between weak and strong adjuncts in
generic/habitual sentences such as (9)-(12) is easily
viewed in formal terms: weak adjuncts may function as the
first argument of G", while strong adjuncts may not.
The discussion in this chapter presupposes a refined
understanding of English temporal semantics; section 1 is
therefore devoted to a detailed exposition of the analysis
of tense and time adverbs assumed here. (This introductory
section is fairly long and intricate. I make no apologies
for this: in my view, the plausibility of the analyses
advanced in sections 2 through 4 can only be judged within
the framework of a carefully articulated theory of tense and
temporal reference; my object in section 1 is to develop
such a framework.)
In section 5 of this chapter, the proposed analysis of
the role of free adjuncts in the system of English temporal
semantics is summarized. The rules of syntax and
interpretation employed here appear in the formal fragment
of English presented in the Appendix.
1. Preliminaries
In this section I shall explain a number of assumptions

behind my account of the interaction of free adjuncts with
tense. In 1.1, I review some basics of interval semantics.
In 1.2, I discuss temporal abstracts: these playa central
role in the system of temporal semantics assumed here, since
they are necessary for the derivation of tensed sentences;
thus, 1.2 includes a discussion of the basic syntactic and

semantic rules for tense in the proposed fragment of
English. In 1.3, two categories of time adverbs are
distinguished, and the rules joining time adverbs with
temporal abstracts are set forth. In 1.4, a detailed
semantic analysis of temporal adverbial clauses is
presented. Finally, a short summary of the assumed system
of tense and time adverbs is presented in section 1.5.
1.1. INTERVAL SEMANTICS
In the fragment developed here for the syntax and semantics

of free adjuncts and absolutes, interpretations are assumed
to be relativized to intervals of time rather than to
moments or points in time. The reasons for taking intervals
rather than moments as basic are by now familiar; I refer
the reader to Bennett & Partee (1972) and to Dowty
(1979:138f) for arguments in favor of this position.
The definitions assumed for intervals and for various
temporal relations are essentially those of Dowty
(1979:139f,353f). Where T is the set of moments and < is
the standard dense linear ordering of T, an interval is
defined as a subset i of T such that if ml ,m3 e; i and ml ~
m? ~ mV then m2 ~ i. (A moment may thus be regarded as a
klnd oT interval if unit sets are identified with their
members.) Where i"i 2 are intervals, i, is a subinterval of
i2 iff il ~ i 2 •
The version of intensional logic employed in the
fragment developed here includes expressions referring
directly to intervals, as well as the following complex
expressions, where ~,~ denote intervals and ~ denotes a
truth value: now, moment, pastCn, presC~). f'ut(~), ~c ~,
~ < ~, and AT(~,~). The i r i nterpretat ion may be informally
characterized as follows (see the Appendix for a formal
statement of their semantics): at any interval i,
now denotes i;
moment denotes the characteristic function of the
set of unit sets of members of T;
pastC~) is true iff there is some nonempty
interval i' such that i" < i' < i, where i" is the
denotation of ~;'
pres(~) is true iff the denotation of 4 is a
subset of i;
102 CHAPTER III
fu t «(") is t rue iff i < i', wher e i' is the

denotation of ~;
C c ~ is true iff the denotation of C is a subset
of the denotation of ~;
~ < ~ is true iff for every moment m1 in the
denotation of ~ and every moment m2 in the denotation
of ~, mt < m2 ;
AT(~,~) is true iff ~ is true at the interval
denoted by ~.
The structure of an interpretation will be augmented by

an indication that some interval is to be the speaker's
interval--one which contains the moment of utterance, and
which is intuitively regarded as that span of time which the
speaker categorizes as 'present'. This enrichment of the
structure of models is necessary for the interpretation of
certain expressions--NOW, PRESa), FUT(~), and NONPAST(o!'):
at any interval i, relative to the speaker's interval i',
NOW denotes i';

PRES( n is true iff i" =- i', where i" is the
denotation of C;
FUT(o!') is true iff i' < i", where i" is the
denotation of C;
NONPASTa) is true iff there is no subil1terval i"
of the denotation of C such tha t i" < i'.
The speaker's interval is also necessary to

characterize the notion of an index of possible utterance:
a world-interval pair <w,i> is an index of possible
utterance iff i is a subinterval of the speaker's interval.
Hereafter, example sentences will always be assumed to be
interpreted at indices of possible utterance.
The introduction of expressions referring to time
intervals makes it convenient to introduce a new primitive
type into the system of types of expressions of intensional
logic: to the type e of entity-denoting expressions and the
type t of truth value-denoting expressions, the type i of
interval-denoting expressions is added. A new set of
complex types results from this addition by the familiar
recursive definition (i.e. given any types a,b, <a,b> and
<s,a> /are also types); thus, expressions of type <i,t>
denot~sets of intervals, those of type «s,<i,t»,t> denote
sets of properties of intervals, and so on. (See Dowty
1979:326.)
Throughout, the following variable conventions are to

be assumed:
t,t n are to be vO,i' vn,i' respectively, where n

is any positive integer;
I,I n are to be vO,(i,t>' vn,<i,t>' respectively;
pt,Q t are to be vO,<s,<i,t»' v 1 ,<s,<i,t»'
respectively;
pt is to be Vo , <s , «s , <i t t» , t»;
Mn is to be vn,<i,<i,t».
1.2. TEMPORAL ABSTRACTS
In the fragment developed here, tensed sentences derive from

a special category of expressions: the category of temporal
abstracts.
TAB the basic category of A

temporal abstracts
Expressions of this category may have the superficial form

of sentences, but they denote sets of time intervals. Thus,
the expression (14), as a member of category TAB, has the
denotation of (15);
(14) John walked home.
(15) At[past(t) & AT(t, John-walks-home')]
that is, it denotes the set of past intervals at which John

walked home.
The assumption of this category is motivated by two
important considerations. First, sentences may be modified
by indefinitely many time adverbs:
(16) Yesterday John saw Mary in the morning, after Bill

arrived.
To account for this fact, the rule joining a time adverb

with the expression which it modifies must be allowed to
104 CHAPTER III
iterate. The simplest way to accomplish this is to regard

time adverbs as joining with temporal abstracts to produce
temporal abstracts: for example, starting with the temporal
abstract (17a) (having the denotation of (17b»,
(17 ) a. John saw Mary.

b. At[past(t) & AT(t, John-sees-Mary')]
the addition of the time adverb in the morning yields the

temporal abstract in (18a) (whose denotation is as in
(18b»;
(18 ) a. John saw Mary in the morning.

b. At[in-the-morning'(t) & ex(t)] (where ex =
( 17b) )
the addition of after' Bill arrived yields yet a new temporal

abstract, as in (19);
(19 ) a. John saw Mary in the morning, after Bill

arrived.
b. At[ after-Bill-arri ved'(t) &: 6(t)] (where 6 =
(18b»
and so on. This sort of iteration is not possible in

systems in which time adverbs join directly with sentences
(expressions of category t) to produce sentences; in such
systems, iteration of the rule for time adverbs would
erroneously predict scope differences between the multiple
adverbs in sentences such as (16). (Cf. Dowty (1979:3280.)
See section 1.3 for discussion of the rule joining a time
adverb with a temporal abstract, and for a more formal
analysis of a sentence containing multiple time adverbs.
A second reason for postulating a category of temporal
abstracts pertains to the semantics of frequency adverbs
(discussed in section 3 of this chapter) and of the
'generalization operator' G" proposed in section 4. As I
show, these may be plausibly analyzed as binary operators
whose second argument has the form of a tensed sentence;
nevertheless, I argue that they denote binary relations
between sets of time intervals. Accordingly, it is
desirable to assume that the second argument of a frequency
adverb or of the operator G" is an expression with the form
of a tensed sentence, but which has a denotation of type
<i,t>--in short, that it is a temporal abstract. See

sections 3 and 4 for further discussion.
1.2.1. Rules of Tense
In the proposed fragment, temporal abstracts play an

important intermediate role in the production of tensed
sentences from tense less sentences. Specifically, tenseless
sentences are converted to tensed sentences in three steps:
first, they are converted into tense less temporal abstracts;
these are then converted to tensed temporal abstracts;
these, in turn, become tensed sentences.
The rule for changing a tenseless sentence into a
tenseless temporal abstract is (20):2
(20) Tenseless temporal abstract rule:
S56. I f <p e:: Pt , [-Tense], then F56 (<P)

PTAB , e::
[-Tense, -Adjunct, -Perfect], where fS6(¢) is

¢•
Example: f 56 «John) [is] happy) = (John) [is] happy

T56. I f <pPt and <p translates as <pI, then F56 (
£ <P)
translates as At[AT(t, <PI)].
Example: F 56 «John) [is] happy) translates as

At[AT(t, John-is-happyl)]
The three rules introducing tense into a temporal

abstract are (21)-(23):2
(21) Past tense rule:
S57. If a PTAB , [-Tense, -Adjunct, ~Perfect],

e::
then F57(cd e: P TAB , [+Tense, -Adjunct,

-perfect], where f S7 (a) is PRET(a).
Example: F57 «John) [is] happy) = (John) [was] happy

106 CHAPTER III
(21) [continued]
T57. If ex E: PTAS and ex translates as ex', then

F 57 (ex) translates as At[past(t) & ex'(t)].
Example: F 57 «John) [is] happy) translates as

At[past(t) & AT(t, John-is-happy')]
(22) Present tense rule:
S58. If ex E: P TAB , [-Tense, -Adjunct, -Perfect],

then F 58 (Ct) E: P TAB , [+Tense, -Adjunct,
-Perfect], where F58 (Ci) is Ct.
Example: F 58 «John) [is] happy) = (John) [is] happy
T58. If a E P TAB and Ci translates as a', then

F 58 (a) translates as At[PRES(t) & a'(t)].
Example: F 5S «John) [is] happy) translates as

At[PRES(t) & ATCt, John-is-happy')]
Many sentences deriving from abstracts produced by (22)

are acceptable only as narrative presents; specifically,
these will be sentences with 'interval predicates' (Dowty
1979: 184), such as John walks home today, Now Mary notices
the signal, and so on.
(23) Future tense rule:
S59. If a E P TAB , [-Tense, -Adjunct, -Perfect],

then F 59 (ex) E: P TAB , [+Tense, -Adjunct,
-Perfect], where F59 (a) is RB([will], a).
Example: F 59 «John) [is] happy) = (John) [will] be

happy
T59. If a E: P TAB and a translates as Ci', then

F 59 (ex) translates as At[FUT(t) & ex'(t)].
TENSE AND FREE ADJUNCTS 10'{
(23) [continued]
Example: F59 «.John) [is] happy) translates as

At[FUT(t) & AT(t, John-is-happy')]
The rule convertin~ a tensed temporal abstract into a

tensed sentence is (24):
(24) S60. I f a PTAB , [+Tense, -Adjunct, -Perfect],

£
then F 60 (a) £ Pt , [+Tense], where F 60 (a) is

a.
Example: F 60 «John) [will] be happy) = (John) [will]

be happy
T60. If ex PTAB and a translates as a',

£ then
F60 (a) translates as vt[a'(t)].
Example: F60 «John) [will] be happy) translates as

vt[FUT(t) & AT(t, John-is-happy')]
Thus, in this system, the sentence (25) derives as in

(26), and is assigned the translation (27).
(25) John was smart.
(26) (John) [was] smart, t, 60

I
(John) [was] s.art, TAB, 57
I
(John) [is] smart, TAB, 56
I
(John) [Is] saart, t, 19
~ .
JobO, T [be] smart, IV 1 , 26
~.
[be], IV1/PRED 1 smart, ADJ 1
(27) Vt[past(t) & AT(t, smart'(John'»]
A natural question which arises here is: why are three

steps necessary for converting a tense less sentence into a
108 CHAPTER III
tensed sentence? Wouldn't it be simplest to convert

tenseless sentences directly into tensed temporal abstracts,
as in (29), rather than to include the intermediate stage
(b) of tenseless temporal abstracts?
(28) (a) -+- (b) -+- (c) -+- (d)
tense less tenseless tensed tensed

sentence temporal temporal sentence
abstract abstract
(29) (a) -+- (c) -+- (d)
tenseless tensed tensed

sentence temporal sentence
abstract
As it turns out, the three steps in (28) are all necessary;

this is because there are two rules (S68 and S69 in the
Appendix) requiring the existence of tenseless temporal
abstracts; the motivation for these rules will be discussed
in section 1.4.1.
1.2.2. Shifting and NODshifting Tenses
The attentive reader may wonder why I have used 'PRES(O'

and 'FUT(O' in the translation rules for the present and
future tenses, but 'past(~)' in the translation rule for the
past tense. (Recall from 1.1 that 'PRES(O' and 'FUT(O'
are always interpreted relative to the speaker's interval,
but that the denotation of 'pastU)', 'preset)', and
'fut(O' depends on their scope relative to other tenses.)
This difference is motivated by the following facts: on the
one hand, present tense and future tense sentences can be
embedded under other tenses but still be interpreted
relative to the speaker's interval; on the other hand, there
is no clear evidence that past tense sentences must ever be
interpreted relative to the speaker's interval when they are
embedded under other tenses.
To see this, first consider sentence (30).
(30) John will claim that Mary is hitting Bill.
For most speakers, sentence (30) has an interpretation

according to which John will claim that Mary is hitting Bill

right now, not at the time of his claim; if analyzed as in
(31), (30) is assigned the translation (32), which induces
precisely this interpretation.
(31) (John) [will] claim that Mary is hitting Bill,

I t, 60
(John) [will] claim that Mary is hitting Bill,
TAB, 59
(John) [claims] that Mary is hitting Bill, TAB, 56

I
(John) [claims] that Mary is hitting Bill, t, 19
------------
John, T [claim] that Mary is hitting Bill, lVi, 21
I
[claia] that Mary is hitting Bill, IVS, 27
~
[claim], IVs/t (Mary) [is] hi tting Bill, t, 60
I
(Hary) [is] hitting Bill, TAB, 58
I
(Mary) [is] hitting Bill, TAB, 56
I
(Mary) [is] hitting Bill, t, 19
(32) Vt[FUT(t) & AT(t, Vxs[R(x s , John') &

clai.' (x s , ~Vtl[PRES(tl) &
AT(t 1 , Mary-is-hitting-Bill')])])]
If 'pres' were substituted for 'PRES' in (32), then this

interpretation could not be induced: since 'presa)', like
'past(~)', does shift in denotation when within the scope of
some other tense, the translation resulting from this
substitution would induce the distinct interpretation that
John will claim that Mary is hitting Bill at the time of his
claim. (This, of course, is also a possible interpretation
for (30); see below.) Thus, the present tense can have a
nonshifting interpretation within the scope of future tense.
(I should note, however, that I have encountered at least
one speaker who doesn't accept a nonshifting interpretation
for the subordinate clause of (30). In fact, some embedded
present tense clauses are universally acknowledged to lack a
110 CHAPTER III
nonshifting interpretation; sentence (33), for example, can

apparently only mean that John will claim that Mary is his
wife at the time of his claim--not at the speaker's
interval.
(33) Next year, John will claim that Mary is his wife.
It isn't clear to me why speakers should vary in their

interpretation of sentences like (30), nor why (30) should
have a nonshifting interpretation while (33) does not.)
Roughly similar facts hold for present tense sentences
within the scope of past tense. The factive sentence (34),
for example, has the entailment that Mary is walking home
right now;
(34) John discovered that Mary is walking home.
if (34) is derived as in (35), then i t is assigned the

translation (36), which captures this entailment.
(35) (John) [discovered] that Mary is walking home,

I t, 60
(John) [discovered] that Mary is walking home, TAB, 57
I
(John) [discovers] that Mary is walking home, TAB, 56
I
(John) [discovers] that Mary is walking home, t, 19
John, T
-----------
[discover] that Mary is walking home, lVi, 21
I
[discover] that Mary is walking home, IVs, 27
~
[discover], IVs/t (Mary) [is] walking home, t, 60
I
(Mary) [is] walking home, TAB, 58
I
(Mary) [is] walking home, TAB, 56
I
(Mary) [is] walking home, t, 19
(36) Vt[past(t) & AT(t, Vxs[R(x s , John') &

discover'(x s , Vt 1 [PRES(t 1 ) &
A
AT(t 1 , Kary-is-walking-home')])])]
If 'pres' were substituted for 'PRES' in (36), the resulting

translation would fail to capture the relevant entailment.
Thus, the present tense has a nonshifting interpretation
within the scope of past tense as well. «34) is,
nevertheless, different from (30): (34) seems to carry the
implication (implicature?) that Mary was already walking
home at the time of John's discovery; (30) has nothing
analogous to this component of (34)'s meaning.)
(Note here that without rule (22) to introduce
'PRES(~)' into the translation of a present tense sentence,
the embedded clauses in sentences (30) and (34) couldn't be
assigned a nonshifting interpretation. Thus, (22) is not as
expendable as i t may appear.)
Now consider sentence (37).
(37) John claimed that Mary will hit Bill.
This sentence has the interpretation that John claimed that

Mary will hit Bill sometime later than the speaker's
interval; it cannot mean that John claimed that Mary would
hit Bill sometime later than the time of his claim, but
possibly before the speaker's interval. If (37) is analyzed
as in (38), the rules discussed above assign it the proper
interpretation, namely that represented by translation (39).
(38) (John) [claimed] that Mary will hit Bill, t, 60

I
(John) [claised] that Mary will hit Bill, TAB, 57
I
(John) [claims] that Mary will hit Bill, TAB, 56
I
(John) [claias] that Mary will hit Bill, t, 19
~
John, T [claim] that Mary will hit Bill, lVi, 21
--------
I
[claim] that Mary will hit Bill, lVs, 27
[claim], IVs/t (Mary) [will] hit Bill, t, 60

I
(Mary) [will] hit Bill, TAB, 59
I
(Mary) [hits] Bill, TAB, 56
I
(Mary) [hits] Bill, t, 19
112 CHAPTER III
(39) vt[past(t) & AT(t, vxs[R(x s , John') &

cla~'(xs, A Vt1 [FUT(t 1 ) &
AT(t 1 , Mary-hits-Bill')])])]
If 'fut' were substituted for 'FUT' in (39), (37) would

recei ve the wrong interpretation, namely that of sentence
(40). Thus, future tense sentences do not shift in
denotation when embedded in a past tense sentence.
(40) John claimed that Mary would hit Bill.
Future tense sentences embedded within a future tense

sentence are problemati"c. Consider sentence (41).
(41) John will claim that Mary will hit Bill.
If the embedded future tense sentence has a nonshifting

interpretation, (41) should be able to mean that John will
claim that Mary will hit Bill sometime later than the
speaker's interval but possibly before the time of John's
claim. My intuitions about this sentence are not entirely
firm, but I don't believe that such an interpretation is
possible; if (41) has any acceptable interpretation, it is
one according to which John will claim that Mary will hit
Bill subsequently to his claim. It is unclear to me why
this should be so, especially in the light of the clearly
non shifting interpretation of the embedded future tense
sentence in (37). I leave this puzzle unresolved.
In spite of the complications connected with sentences
like (41), it is clear that both present and future tense
sentences are sometimes interpreted relative to the
speaker's interval when they are embedded under some other
tense. It is unclear, however, that past tense sentences
must ever be interpreted in this way. sentences (42) and
(43) will bring this out.
(42) John will claim that Mary hit Bill.

(43) John claimed that Mary hit Bill.
sentence (42) has the interpretation that John will claim

that Mary hit Bill sometime before his claim, possibly after
the speaker's interval; this is the interpretation induced
by the translation (45) assigned to (42) by the rules above
in connection with the derivation (44).
(44) (Joho) [viII] claim that Mary hit Bill, t, 60
I
(John) [will] cla~ that Mary hit Bill, TAB, 59
I
(Joho) [claims] that Mary hit Bill, TAB, 56
I
--------
(Joho) [clau.s] that Mary hit Bill, t, 19
John, T [claim] that Mary hit Bill, IV i , 21

I
--=----
[cla~] that Mary hit Bill, IVS, 27
[claim], IVs/t (Mary) [hit] Bill, t, 60

I
(Mary) [hit] Bill, TAB, 57
I
I
(Mary) [hits] Bill, t, 19

claim'(x s , AVt,[past(t,) &
AT(t" Mary-hits-Bill')])])]
If (42) had an additional interpretation on which the

embedded past tense clause were interpreted relative to the
speaker's interval, (42) would be able to mean that John
will claim that Mary hit Bill sometime before the speaker's
interval. Notice, however, that (42) would, under such an
interpretation, be true in a proper subset of the set of
situations in which (45) is true; it is thus impossible to
establish that (42) has an interpretation on which its
embedded clause has a nonshifting denotation.
Sentence (43) has the interpretation that John claimed
that Mary hit Bill sometime before his claim; this is the
interpretation represented by translation (47), which is
assigned to (43) on the analysis in (46). If (43) had an
interpretation on which the embedded clause were interpreted
relative to the speaker's interval, (43) could mean that
John claimed that Mary hit Bill sometime before the
speaker's interval but possibly later than the time of his
claim. (43) plainly lacks such an interpretation. ThUS,
there is no clear evidence that past tense sentences ever
have a nonshifting denotation.
114 CHAPTER III
(46) (John) [claimed] that Mary hit Bill, t, 60

I
(John) [claimed] that Mary hit Bill, TAB, 57
I
(John) [claims] that Mary hit Bill, TAB, 56
I
(John) [claims] that Mary hit Bill, t, 19
~
John, T [claim] that Mary hit Bill, lVi, 21
I
[claim] that Mary hit Bill, IV s , 27
~
[claim], IVS/t (Hary) [hit] Bill, t, 60
I
(Hary) [hit] Bill, TAB, 57
I
I
(Hary) [hits] Bill, t, 19

claim'(x s , ~vt1[past(t1) &
For these reasons, the tense rules in section 1.2.1

have been formulated so as to assign nonshifting
interpretations to the present and future tenses, but a
shifting interpretation to the past tense. I should
acknowledge, however, that there are certain interpretations
of tense which the proposed rules of tense cannot account
for; specifically, these are the so-called 'sequence of
tense' interpretations found in sentences such as the
following.
(48) John claimed that Mary was hitting Bill.

(30) John will claim that Mary is hitting Bill.
(40) John claimed that Mary would hit Bill.
(41) John will claim that Mary will hit Bill.
Sentence (48) is ambiguous: it can mean either (a)

that John claimed that Mary was hitting Bill sometime before
his claim, or (b) that John claimed that Mary was hitting
Bill at the time of his claim. The proposed tense rules
suffice to induce interpretation (a) for sentence (48), but
not to induce interpretation (b). Sentence (30) is

similarly ambiguous: it can mean either (a) that John will
claim that Mary is hitting Bill right now, or (b) that John
will claim that Mary is hitting Bill at the time of his
claim. As has already been seen, the rules proposed above
can induce interpretation (a); they are, however, unable to
induce interpretation (b). The rules are also incapable of
assigning (40) the nonmodal interpretation that John claimed
that Mary would hit Bill at some time after his claim.
Finally, the rules cannot assign (41) the interpretation
that John will claim that Mary will hit Bill at some time
after his claim.
According to the traditional view, the embedded clause
in (48) and (30) is, on the (b) interpretation, a present
tense sentence whose denotation shifts relative to the tense
in whose scope it appears; and the embedded clause in (40)
and (41) is, on the nonmodal interpretation, a shifting
future tense sentence. These 'sequence of tense'
interpretations for (48), (30), (40), and (41) can be
formally represented as in (49)-(52).
claim'(x s , "vt1 [pres(t 1) &

AT(t 1 , Hary-is-hitting-Bill')])])]
(50) Vt[FUT(t) & AT(t, VxS[R(x s , John') &

claim'(x s , "'vt1 [pres(t 1) &
AT(t 1 , Hary-is-hitting-Bill')])])]
(51) vt[past(t) & AT(t, Vxs[R(x s , John') &

claim'(x S , "vt 1 [f'ut(t 1 ) &
AT(t 1, Hary-hits-Bill')])])]
clai.'(x s , "'Vt1 [f'ut(t 1) &

In the fragment of English grammar developed here, no

provision is made for the assignment of translations such as
(49)-(52); but see Stump (1981b:166-176) for one possible
approach to the sequence of tense phenomenon.
116 CHAPTER III
1.3. TWO CATEGORIES OF TIME ADVERBS
Throughout this chapter, I assume the existence of two

categories of time adverbs. This category distinction
reflects a distinction between two very different sorts of
roles which a time adverb may play in the interpretation of
a sentence.
1.3.1. Time Adverbs of Category TA
In sections 3 and 4 of this chapter, it will be seen that a

time adverb may function as the argument of certain
expressions, and that the purpose of such a time adverb is
simply to specify a set of time intervals; for this reason,
a time adverb serving this purpose can be assumed to have a
set of time intervals as its denotation. Adverbs of this
sort will belong to a basic category TA of expressions with
denotations of type <i,t>:
TA the basic category of yesterday, today,

set-level time adverbs tomorrow, ...
The adverbs yesterday, today, and tomorrow, like most

time adverbs, do not shift in reference from one time to
another; that is, even if they occur within the scope of
some nonpresent tense, they always pick out a set of
intervals standing in a certain relation to the speaker's
interval. For example, in the interpretation of (53), today
can only pick out the set of subintervals of the day
containing the speaker's interval; (53) cannot be
interpreted to mean that John discovered that Mary was here
on the day of John's discovery.
(53) John discovered that Mary was here today.
Similarly, yesterday and tomorrow in (54) and (55) are

interpreted relative to the speaker's interval, not relative
to the time of John's discovery.
(54) John discovered that Mary was here yesterday.

(55) John discovered that Mary will be here tomorrow.
TENSE AND FREE ADJUNCTS "7
Thus, as members of category TA, today, yesterday, and
tomorrow denote the sets of intervals represented by (56a-
c); the nonshifting interpretation of 'NOW' guarantees a
nonshifting interpretation for each of these adverbs.
(56) a.
(= today')
b. AtVt,[[day'(t,) & At 2 [todaY'(t 2 ) + [t, < t2
& At 3 [[t, < t3 & t3 < t 2 ] + todaY'(t 3 )]]]] &
t ~ t,]
c. AtVt,[[day'(t,) & At 2 [today'(t 2 ) + [t 2 < t,
& At 3 [[t 2 < t3 & t3 < t 1 ] + todaY'(t 3 )]]] &
t ~ t,]
At a given interval, (56a) denotes the set of subintervals

of the day containing the speaker's interval; (56b) denotes
the set of subintervals of the day before the day containing
the speaker's interval; and (56c), the set of subintervals
of the day after the day containing the speaker's interval.
Some time adverbs, such as during the past summer, do
shift in reference from one time to another; for example,
sentence (57) can be understood to mean that John discovered
that Mary was here during the summer before John's
discovery.
(57) John discovered that Mary had been here during the
past summer.
Thus, as a member of category TA, during the past sua mer

translates as (58); the shifting interpretation of 'nov'
permits a shifting interpretation for this adverb.
(58) AtVt,[[[summer'(t,) & t, < nov] &

At 2 [[summer'(t 2 ) & t2 < now] +
[t 2 < t1 v t2 = t 1 ]]] & t ~ t 1]
1.3.2. Time Adverbs of Category MTA
I assume that, in addition to serving as set-denoting

expressions, time adverbs may also serve as functors,
118 CHAPTER III
joining with tense to characterize the interval at which

some sentence is true. In this role, time adverbs are
regarded as denoting functions from properties of time
intervals to sets of time intervals) Thus, I introduce a
category MTA of main tense adverbs with denotations of this
type:
MTA the category TAB/TAB at that timen ,

of main tense adverbs since noon
As members of MTA, the adverbs today, yesterday, and

tomorrow have the denotations of (59a-c), where today',
yesterday', and tomorrow' are (56a-c), respectively:
(59) a. AptAt[todaY'(t) & pt{t}]

b. AptAt[yesterday'(t) & pt{t}]
c. AptAt[tomorrow'(t) & pt{t}]
Similarly, as a member of MTA, during the past summer

translates as (60) (where during-the-past-summer' is (58»:
(60) APtAt[during-the-past-summer'(t) & pt{t}]
There appear to be members of MTA which do not have

counterparts in category TA (e.g. since-adverbials); on the
other hand, all members of category TA have counterparts of
category MTA. This relation can be captured by means of the
following category-changing rule. 4
(61) S63. If E: PTA' [-AdjunctJ, then F 63 (cx)

IX E:PMTA'
where F 63 (a) is a.
T63. If a £ PTA and a translates as a', then

F 63 (a) translates as APtAt[a'(t) & pt{t}J.
A set of pro forms of the form at that time n occur as

basic expressions of category MTA; such proforms translate
as in (62).
(62) at that tiaen translates as AptAt[t = tn & pt{t}]

These proforms play an important role in the derivation of

temporal adverbial clauses, as will be seen in section
1.4.2.
Main tense adverbs join with temporal abstracts to
produce temporal abstracts, as their category definition
TAB/TAB implies. The rule for joining a main tense adverb
with a temporal abstract is (63).
(63) S61. If a E: PMTA and

PTAB , [yTense, 6 Adjunct,
~ E:
~Weak,~Perfect], then F61a(a,~), F 61b (a,B)
E: PTAB , [YTense, 6 Adjunct, ~ Weak, ~Perfect],
where F 61a (a,B) is CONWIB(a,B) and F61b (a,B)

is CON(B,a).
Examples: F 61a (yesterday, (John) [was] happy) =

yesterday, (John) [was] happy
F 61b (tomorrow, (John) [will] be happy) =
(John) [will] be happy tomorrow
T61. If a E: PMTA ,
PTAB , and a,~ translate as
~ £
a',B', then F 61a (a,B), F 61b (a,B) translate as

a'(~B').
Example: F 61 b( tomorrow, (,John) [will] be happy)

translates as At[tomorrow'(t) & [FUT(t) &
AT(t, John-is-happy')]]
(63) feeds rule (24), so that sentence (64) can be derived

as in (66) (see next page) and assigned the translation
(65) •
(64) John was asleep yesterday.
(65) Vt[yesterday'(t) & [past(t) &

AT(t, Vxs[R(x s , John') & asleep'(x s )])]]
120 CHAPTER III
(66) (John) [was] asleep yesterday, t, 60
-----------
I
(John) [was] asleep yesterday, TAB, 61b
yesterday, MTA, 63 (John) [was] asleep, TAB, 57

I I
yesterday, TA (John) [is] asleep, TAB, 56
I
(John) [is] asleep, t, 19
~ .
John, T [be] asleep, IV l , 25
~
[be], IVl/PRED s asleep, ADJ s
One advantage of this analysis is that it allows

sentences with multiple main tense adverbs to be derived
easily, as was mentioned above. Thus, if in the morning is
an expression of category MTA translating as (67), sentence
(68) can be analyzed as in (69), and thereby assigned the
translation (70); (70) induces the desired interpretation
for (68).
(67) Ap t Atvt 1 [[morning ' (t 1 ) & t ~ t 1 ] & pt{t}]
(68) Yesterday, John saw Nary in the morning.
(69) yesterday, (John) [saw] Mary in the morning, t, 60

I
yesterday, (John) [saw] Mary in the morning, TAB, 61 a
yester~)
MTA, 63
[saw] Mary in the morning, TAB, 61b
______________
I in the morning, (John) [saw] Mary, TAB, 57
yesterday, MTA I
TA (John) [sees] Mary, TAB, 56
I
(John) [sees] Mary, t, 19
~
John, T [see] Mary, lvi, 21
I
[see] Mary, IV s , 28
~ Mary,
[see], TV s T
(70) vt[yesterday'(t) & Vt 1 [[morning'(t 1 ) & t ~ t 1] &

[past(t) & AT(t, Vxs[R(x s , John') &
Vys[R(yS, Mary') & see'(x s , yS)]])]]]
Thus, the possibility of iterating rule (63) is one

advantage of deriving tensed sentences from temporal
abstracts. Further advantages of this analysis will be seen
in sections 3 and 4.
In combination with the rules of tense introduced in
section 1.2.1, rule (63) allows sentences like (71 )-(73) to
be freely produced:
(71) John will leave yesterday.

(72) John left tomorrow.
(73) John is happy yesterday.
These kinds of sentences are not regarded as syntactically

ill-formed, but as contradictory;5 (71 )-(73) are assigned
the translations (74)-(76), which are provably
contradictory.
(74) vt[yesterday'(t) & [FUT(t) &

AT(t, Vxs[R(x s , John') & leave'(x s )])]]
(75) Vt[tomorrow'(t) & [past(t) &
AT(t, Vxs[R(x s , John') & leave'(x s )])]]
(76) Vt[yesterday'(t) & [PREset) &
AT(t, happy'(John'»]]
1.4. THE SEMANTICS OF TEMPORAL ADVERBIAL CLAUSES
Throughout this chapter, I shall stress similarities in the

semantics of free adjuncts and temporal adverbial clauses
(or what I shall simply call 'temporal adverbials').
Therefore, as a preliminary to the discussion of tense and
free adjuncts, I shall briefly outline the assumed semantics
for temporal adverbial clauses.
Temporal adverbial clauses are assumed to belong to
both of the categories of time adverbs discussed in section
1.3 above--the category TA of set-level time adverbs and the
category MTA of main tense adverbs. As members of the
former category, they in general translate into intensional
'22 CHAPTER III
logio expressions fitting sohema (77):
Here, (x' represents some expression denoting a two-place

relation between intervals. The identity of a' is
determined by the subordinating conjunction in the temporal
adverbial: in the translation of a before-adverbial, a'
denotes a relation of precedence; in the translation of an
after-adverbial, a relation of subsequence; in the
translation of a when-adverbial, the identity relation; and
so on. ~'represents a sentence of intensional logic; in
the translation of a given temporal adverbial, this serves
to restrict possible values for t,. Thus, the adverbial
clause when Mary sang, as a set-denoting expression,
translates as (an expression equivalent to) (78).
(78) AtVt,[t = t, & [past(t,) & AT(t" Mary-sings')]]
(78) denotes the set of intervals identical to some past

interval at which Mary sang. Similarly, the adverbial
clause while Mary sang translates as (79).
(79) AtVt,[[t ~ t, & ,aomentCt,)] & [past(t,) &

ATCt" Mary-sings')]]
(79) denotes the set of subintervals of a nonminimal

interval at whioh Mary sang. The adverbials before Mary
sang and after Mary sang translate as Cexpressions
equivalent to) (80) and (81).
(80) AtVt,[[t < t, & M(t,t,)] & [past(t 1 ) &

(8') AtVt,[[t, < t & MCt,t,)] & [past(t,) &

Here, M is a 'context variable' of type <i,<i,t»; the value

which it takes in a given context further restricts the
relation between the values of t and t,; in most cases, thig
is a relation implying some degree of temporal proximity.
Thus, (80) denotes the set of intervals preceding some other
interval i at which Mary sang and additionally bearing to i
the relation M; (81) denotes the set of intervals following
TENSE AND FREE ADJUNCTS '23
some interval i at which Mary sang and also bearing relation
M to i. (The exactness of these translations will be taken
up in section '.4.5.) As members of category TA, temporal
adverbials undergo rule S63 (see (6') in section '.3.2) to
become main tense adverbs; as members of category MTA, the
adverbials when Mary sang, while Mary sang, before Mary
sang, and after Mary sang thus translate as in (82a-d).
(82) a. AptAt[Vt,[t = t, & [past(t,) &

AT(t" Mary-sings')]] & pt{t}]
b. APtAt[Vt,[[t =- t, & ""1IIIOment(t,)] & [past(t,)
& AT(t" Mary-sings')]] & pt{t}]
c. AptAt[Vt,[[t < t, & M(t,t,)] & [past(t,) &
d. APtAt[Vt,[[t, <t & M(t,t,)] & [past(t,) &
In this way, temporal adverbials undergo the rule joining

main tense adverbs with temporal abstracts (rule (63) in
section '.3.2).
In the following two subsections, two significant
problems which arise in the analysis of temporal adverbial
clauses are dealt with.
'.4.'. The Analysis of Tense in Temporal Adverbial Clauses
The first problem ar~s~ng in the semantic analysis of

temporal adverbial clauses is that they characteristically
show a pattern of superficial tense-marking different from
the pattern found in independent clauses. Like independent
clauses, temporal adverbial clauses relating to a past time
interval show superficial past tense marking, as in (83);
(83) a. John left when Mary sang.

b. John left while Mary sang.
c. John left before Mary sang.
d. John left after Mary sang.
thus, the adverbials in (83) can be unproblematically

translated as in (78)-(81) above. Similarly, temporal
adverbial clauses relating to a present time interval are
'24 CHAPTER III
superficially present in tense, as in the sentences in (84)

(which are best read as 'narrative presents');
(84) a. John leaves when Mary sings.

b. John leaves while Mary sings.
c. John leaves before Mary sings.
d. John leaves after Mary sings.
the set-level translations in (85) might, accordingly, be

postulated for the adverbials in (84).
(85) a. AtVt, [t = t, & [PRESet,) &

AT(t" Mary-sings')]]
b. AtVt,[[t ~ t, & ,moment(t,)] & [PREset,) &
c. Atvt,[[t < t, & M(t,t,)] & [PRESet,) &
d. AtVt,[[t, < t & M(t,t,)] & [PREset,) &
But when a temporal adverbial clause relates to a future

time interval, it typically shows present tense verb
morphology:
(86) a. John will leave when Mary sings.

(Cf. ??John will leave when Mary will sing.)
b. John will leave while Mary sings.
(Cf. ??John will leave while Mary will sing.)
c. John will leave before Mary sings.
(Cf. ?John will leave before Mary will
sing. 7 )
d. John will leave after Mary sings.
(Cf. ??John will leave after Mary will sing.)
Clearly the adverbial clauses in (86a-d) cannot have the

set-level translations in (85a-d); that is, although they
show superficial present tense marking, they are not
interpreted as present relative to the speaker's interval.
How, then, are the adverbial clauses in (86a-d) to be
analyzed?
An initially plausible analysis of these adverbial
clauses would be that they have a shifting present tense
interpretation--that they translate as in (87), and are
understood to pick out a future set of intervals in (86)
because they are within the scope of the future tense
operator in the superordinate clause.
(87) a. AtVt1 [t = t1 & [pres(t 1 ) &
b. Atvt,[[t ~ tl & ~momentCtl)] & [preset,) &
c. At Vt, [ [t < t, & M( t , t, ) ] & [preset,) &
d. AtVt,[[t, <t & M(t,t,)] & [preset,) &
ATCt 1, Mary-sings')]]
Such a solution would not be acceptable, however. As Dowty

(1979:323) has shown, it is incorrect to assume that tense
and time adverbs have scope relative to each other, since
they work in unison to pick out a single interval; the
assumption that the subordinate clauses in (86) are within
the scope of future tense would clearly be inconsistent with
this fact.
An alternative approach would be to assume that the
subordinate clauses in (86) are semantically future in
tense, despite their superficial present tense marking; that
they translate as in (88).
(88) a. AtVt, [t = t, & [FUT(t,) &
b. AtVt,[[t ~ t, & ~ment(t1)] & [FUT(t,) &
c. AtVt,[[t < t, & MCt,t,)] & [FUT(t, ) &
d. AtVt 1 [[t 1 < t & MCt,t,)] & [FUT(t 1 ) &
ATCt 1 , Mary-sings')]]
This solution wouldn't require a scope relationship between

the tense and the temporal adverbial in each of (86a-d). It
would, however, force one to say that the adverbials in (89)
'26 CHAPTER III
are each ambiguous between a future tense interpretation

(that induced by (88» and a present tense interpretation
(that of (85» even though these interpretations are always
in complementary distribution: the future interpretation is
only found in the environment of a superordinate clause in
the future tense, while the present tense interpretation is
restricted to the environment of a superordinate clause in
the present tense.
(89) a. when Mary sings

b. while Mary sings
c. before Mary sings
d. after Mary sings
A preferable alternative would be to assume that the

adverbials in (89) are semantically indeterminate between a
present tense and a future tense interpretation--that they
are simply 'nonpast'. This is what I shall assume; thus,
the translations in (85), (87), and (88) are rejected, and
the adverbial clauses in (89) are each assigned a single
interpretation, namely that induced by the set-level
translation in (90).
(gO) a. HVt, [t = t, & [NONPAST(t 1 ) &

b. AtVt 1 [[t ~ t1 & ~DOment(t1)] & [NONPAST(t1)
& AT(t 1 , Mary-sings')]]
c. AtVt 1 [[t < t, & M(t,t,)] & [NONPAST(t 1 ) &
d. AtVt,[[t, < t & M(t,t 1 )] & [NONPAST(t 1 ) &
The temporal predicate 'NONPAST' does not, of course,

show up in the translations of independent clauses: the
time reference of John loves Mary, for example, isn't
broadly nonpast--it relates strictly to a present time
interval. Since the 'NONPAST' predicate is restricted in
this way to temporal adverbial clauses, I shall assume that
it is directly introduced when clauses of this sort are
generated; in the interests of uniformity, I make a similar
assumption regarding the 'past' predicate appearing in the
translations of past tense adverbial clauses such as those
in (83). Thus, the rules generating temporal adverbial
clauses join a temporal subordinating conjunction with a
tenseless temporal abstract, simultaneously introducing past
or nonpast tense.
For the statement of these rules, I assume that the
subordinating conjunctions before, after, when, and while
are basic expressions of the category TSC--
TSC the basic category of before, after,

temporal subordinating when, while
conjunctions
and that they translate as expressions of type

«s,<i,t»,<i,t», as in C9').
a. before translates as
AptAtvt,EEt < t, & M(t,t,)] & pt{t,}]
b. after translates as
AptAtVt,EEt, < t & M(t,t,)] & pt{t,}]
c. when translates as AptAtVt,Et = t, & pt{t,}]
d. while translates as
AptAtvt,EEt ~ t, & ~DOment(t,)] & pt{t,}]
By rules (92) and (93), before, after, when, and while

JOln with tenseless temporal abstracts to produce temporal
adverbial clauses of category TA:
(92) S68. If a e: PTSC and S e: PTAB' E-Tense, -Adjunct,

-Perfect], then F 68 (a,S) e: PTA' E-Adjunct],
where F68 Ca,a) is CONCa, DPCPRETCa»).
Example: F68C while. (Mary) Eis] singing) = while Mary

was singing
T68. If a E: PTSC' B E:
PTAB ' and a,B translate as
a',S', then F68(a,B) translates as
Ct'(~AtEpast(t) & S'(t)]).
128 CHAPTER III
(92) [continued]
Example: while Mary was singing translates as

AtVt 1 [[t =.
t1 & ...,moment(t 1 )] & [past(t 1) &
ATCt 1 , Mary-is-singing')]]
(93) S69. If a E: PTSC and B

PTAB ' [-Tense, -Adjunct,
E:
-Perfect], then F 69 (a,B) E: PTA' [-Adjunct],

where F69 (a,B) is CONCa, DP(S)).
Example: F 69 Cwhile, (Mary) [is] singing) = while Mary

is singing
T69. If ex E: Prsc' B PTAB ' and a,B translate as

E:
a', B', then F 69(01, B) translates as

a,(AH[NONPAST(t) & S'(t) ]).
Example: while Mary is singing translates as

Atvt 1[[t =- t1 & .,.oment(t 1)] & [NONPAST(t 1)
& AT(t 1 , Mary-is-singing')]]
With these rules, sentences such as (94) and (95) are
easily derived, as in (96) and (97).
(94) John will leave while Mary sings.

(95) John left after Mary sang.
(96) (John) [will] leave while Mary sings, t, 60

I
(John) [will] leave while Mary sings, TAB, 61b
~
while Mary Sings, MTA, 63 (John) [will] leave, TAB, 59
I I
while Mary sings, TA, 69 (John) [leaves], TAB, 56
~ I
while, (Mary) [sings], TAB, 56 (John) [leaves], t, 19
TSC I
(Mary) [sings], t, 19
(97) (John) [left] after Mary sang, t, 60
(John) [left] after Mary sang, TAB, 6'b
after Mary sang~(John) [left], TAB, 57

I I
after Mary sang, TA, 68 (John) [leaves], TAB, 56
~ I
after, ( Mary) [sings], TAB, 56 (John) [leaves], t, '9
TSC I
(Mary) [sings], t, '9
By the analyses in (96) and (97), sentences (94) and (95)

are assigned the interpretations induced by translations
(98) and (99).
(98) vt[Vt,[[t ~ t, & ,aoment(t,)] & [NONPAST(t,) &

AT(t" Mary-sings')]] & [FUT(t) &
AT(t, John-leaves')]]
(99) vt[vt, [[t, <t & M(t,t,)] & [past(t,) &

AT(t" Mary-sings')]] & [past(t) &
Because the while-adverbial in (94) is analyzed as 'nonpast'

rather than present in tense, (98) (like (99» represents a
noncontradictory interpretation, as desired.
Note that because rules (92) and (93) require tenseless
temporal abstracts as arguments, the decision (in section
'.2.') to convert tenseless sentences to tensed sentences in
three steps--as in (28)--is well motivated: it would not be
desirable to eliminate (b) as an intermediate stage in (28),
since this would undermine the analysis of temporal
adverbial clauses embodied in rules (92) and (93).
(28) (a) (b) (c) (d)
tenseless tense less tensed tensed

sentence temporal temporal sentence
abstract abstract
130 CHAPTER III
1.4.2. Unbounded Dependencies in Temporal Adverbial Clauses
A second significant problem which arises in the semantic

analysis of temporal adverbial clauses has been pointed out
by Mike Geis (1970:Ch.3). To appreciate this problem,
consider sentence (100).
(100 ) John left when Jane said that Mary was leaving.
Rule (92) above can only assign one translation to the when-
clause in (100), namely that in (101).
( '01) Atvt 1 [t = t1 & [past(t,) &

AT(t" Jane-says-that-Mary-vas-leaving')]]
But (100) is ambiguous: it can mean either that John left

at the time of Jane's assertion (this is the interpretation
associated with the translation in (101» or at the time at
which Mary was allegedly leaving. 8 Under the latter
interpretation, the when-clause in (100) would have to
receive a translation like (102); so far, however, no means
of assigning such a translation have been proposed.
(102) AtVt,[t = t, & Vt 2 [past(t 2 ) &

AT(t 2 , Vxs[R(x s , Jane') & say'(x s , "'[past(t 1 ) &
AT(t 1 , Mary-is-leaving')])])]]
The analogy of the when-adverbial in ('00) to the

relative construction in (103) is an instructive one.
(103 ) John left at the time at which Jane said that Mary
was leaving.
The ambiguity of (103) is clearly correlated with the

presence of two possible 'extraction sites' for the
prepositional phrase at which:
••• the time at which [Jane said that [Mary was

leaving _ _ ] _ _ ]
since (100) and (103) are ambiguous in exactly the same way,
i t is tempting to derive the when-adverbial in (100) by a
rule of unbounded extraction; this is in fact exactly what
Geis proposes. In the fragment developed here, however,
movement rules as such are not employed; unbounded
dependencies are instead captured by rules of variable
binding. Accordingly, a rule of variable binding is here
proposed to account for unbounded dependencies in temporal
adverbial clauses such as that in (100):
(104) 867. I f a £ PTSC and cp £ Pt , [+Tense], then

F 67 ,n(a, ¢) £ PTA' [-Adjunct], where
F 67 ,n(a,cp) is
CONCa, DEL(at-that-time n , DP(CP))).
T67. If a £ P TSC ' cP <: P t , and a,cp translate as a',

cp', then F 67 ,n(a,cp) translates as
a'(" At n [ cpl ]).
The variable bound by this rule is the intensional

logic variable tn of type i appearing in the translation of
the adverbial proform at-that-time n (<: BMTA ) introduced in
section 1.3.2 above:
(62) at-that-timen translates as APt At[t = tn & pt{t}]
According to the rules discussed so far, this proform can be

used to build up sentence (105) in two different ways--as in
(106) and (107) (see next page).
(105 ) Jane said that Mary was leaving at that time.
Under these analyses, sentence (105) is assigned the

translations (108) and (109), respectively; note the free
variable t4 of type i introduced by the proform at-that-
tiae4·
(106) (Jane) [said] that Mary was leaving at-that-time4'
I t, 60
(Jane) [said] that Mary was leaving at-that-time 4 , TAB, 61b
------------- I
at-that-time4'
MIA
(Jane) [said] that Mary was leaving,
TAB, 57
(Jane) [says] that Mary was leaving, TAB, 56
I
(Jane) [says] that Mary vas leaving, t, 19
132 CHAPTER III
(107 ) (Jane) [said] that Hary was leaving at-that-time4'

t, 60
(Jane) [saidJ that Hary was leaving at-that-time4' TAB, 57

I
(Jane) [saysJ that Mary was leaving at-that-time4' TAB, 56
I
(Jane) [saysJ that Hary was leaving at-that-time4' t, 19
~
Jane, [sayJ that Mary was leaving at-that-time4 , lVi, 21
T I
[sayJ that Mary was leaving at-that-time4 , IV s , 27
~
[sayJ, IVs/t (Hary) [was] leaving at-that-time4' t, 60
I
-----------
(Mary) [wasJ leaving at-that-ti.e 4 , TAB, 61b
at-that-time4' MTA (Hary) [was] leaving, TAB, 57

I
(Mary) [is] leaving, TAB, 56
I
(Hary) [is] leaving, t, 19
(108 ) vt[ t = t4 & [past(t) &
AT(t, Jane-says-that-Hary-was-leaving')JJ
Vt[past(t) & AT(t, Vxs[R(x s , Jane') &

say'(x s , ~Vtl[tl = t4 & [past(t 1 ) &
AT(t 1 , Hary-is-leaving')J])])]
When rule (104) joins the sentence derived in (106) or (107)

with a temporal subordinating conjunction a to form a
temporal adverbial clause (of category TA), the adverbial
proform at-that-time4 is deleted; concomitantly, the
variable t4 in the translation (108) or (109) is bound by a
lambda operator, and the intension of the resulting
expression serves as the argument of the function denoted by
a. Thus, the adverbial clause in (110) can be derived as in
(111) or (112); the translations induced by these
derivations are (113) and (114), respectively.
( 110) when Jane said that Mary was leaving

(111) when Jane said that Mary was leaving, TA, 67,4
~
when, (Jane) [said] that Mary was leaving at-that-time4'
TSC I t, 60
[as in (106)]
(112) when Jane said that Mary was leaving, TA, 67,4
~
when, (Jane) [said] that Mary was leaving at-that-ti.e4'
TSC I t, 60
[as in (107)]
( 113 ) AtVt 1 [t = tl & vt 2 [t 2 = tl & [past(t 2 ) &

AT(t 2 , Jane-says-that-Hary-vas-leaving')]]]
( 114) Atvt 1 [t = tl & vt 2 [past(t 2 ) &

AT(t 2 , Vxs[R(X S , Jane') & say' (x S , .... vt 3 [t 3 = tl &
[past(t 3 ) & AT(t 3 , Mary-is-leaving')]])])]]
Note that these are equivalent to the 'target translations'

in (101) and (102).
As an expression of category TA, when Jane said that
Mary was leaving (however it is derived) can be converted to
category MTA by rule (61) (see section 1.3.2) and thus enter
into the derivation of sentences like (100). (100) thus
receives the two analyses in (115) and (116).
( 115 ) (John) [left] when Jane said that Mary was

leaVing, t, 60
I
(John) [left] when Jane said that Mary was leaving, TAB, 61b
when Jane said that Mary was leavi~) [left],

MTA, 63 TAB, 57
I I
I
when Jane said that Mary was leaving, (John) [leaves],
TA, 67,4 TAB, 56
I
[as in (111)] (John) [leaves],
t, 19
134 CHAPTER III
( 116) (John) [left] when Jane said that Mary was

leaving, t, 60 I
(John) [left] when Jane said that Mary was leaving, TAB, 61b
~
when Jane said that Mary was leaving, (John) [left],
I MTA, 63 TAj' 57
I
when Jane said that Mary was leaving, (John) [leaves],
TA, 67,4 TAB, 56
I
[as in (112)] (John) [leaves],
t, 19
These distinct analyses correspond to the distinct

translations (117) and (118), respectively:
vt[vt 1[t = t1 & vt 2 [t 2 = t1 & [past(t 2 ) &

AT(t 2 , Jane-says-that-Mary~as-leavingl)]]] &
[past(t) & AT(t, John-Ieaves l )]]
(118 ) vt[vt 1 [t = tl & vt 2 [past(t 2 ) &

Vt3 [t 3 = tl &
AT(t 2 , Vxs[R(x s , Jane') & say'(x s , A
[past(t 3 ) & AT(t 3 , Mary-is-leaving')]])])]] &

[past(t) & AT(t,John-leaves')]]
These translations account for the observed ambiguity of

sentence (100).
One refinement in the statement of rule (104) is
probably necessary; this is because of a critical difference
between while-clauses and other sorts of temporal adverbial
clauses (first painted out by Geis (1970:Ch.3». To
appreciate this difference, compare (100) and (119):
(100) John left when Jane said that Mary was leaving.
( 119) John left while Jane said that Mary was leaving.
As was seen above, (100) is ambiguous: in (100), John's

departure may be understood to be simultaneous either with
Jane's assertion or with Mary's alleged departure. (119),
however, is not analogously ambiguous: it can only be
understood to mean that John left during Jane's assertion.
Thus, whereas the 'extraction site' in a when-clause may be
embedded indefinitely deeply, that in a while-clause of the
form 'while ~, cannot be embedded at all, but must be
associated with the highest clause~. Consequently, unlike
other sorts of temporal adverbial clauses, while-clauses
shouldn't be viewed as being generated by means of the rule
of variable binding in (104), since they don't show the
sorts of unbounded dependencies which originally motivated
this rule; on the contrary, it is sufficient to assume that
all while-clauses are generated by means of rules (92) and
(93).9
Under this assumption, sentence (119) receives only one
sort of analysis in the present fragment (unlike the
ambiguous sentence ('100»:
(120 ) (John) [leCt] while Jane said that Mary was

leaving, t, 60 I
(John) [lect] while Jane said that Mary was leaving,
~ TAB,61b
while Jane said that Mary was leaving, (John) [leCt],

MTA, 63 TAB, 57
I
while Jane said that Mary was leaving,
I
(John) [leaves],
~ TA,68 TAB, 56
I
while, (Jane) [says] that Mary (John) [leaves],
TSC was leaving, TAB, 56 t, 19
I
(Jane) [says] that Mary was leaving, t, 19
Note that the adverbial proform at-that-time n plays no role

in this derivation. The corresponding translation (121)
induces the desired interpretation for sentence (119),
according to which John's departure occurred during Jane's
assertion.
( 121) vt[vt 1 [[t ~ t1 & ,aoment(t 1 )] & [past(t 1 ) &

AT(t 1 , Jane-says-that-Hary-was-leaving')]] &
[past(t) & AT(t, John-leaves')]]
One might suppose that rule (104) eliminates the need

for rules (92) and (93) in the analysis of when-, beCore-,
and aCter-adverbials--that the latter rules are only
136 CHAPTER III
necessary for the analysis of while-clauses, as in (120).
This is not the case, however; to see why, consider sentence
(122) •
( 122) John will leave when Mary sings.
If rule (104) is employed in the analysis of its adverbial

clause, this sentence is assigned analysis ('23) and
translation (124);
(123 ) (John) [will] leave when Mary sings, t, 60

I
-------------
(John) [will] leave when Mary sings, TAB, 61b
when Mary sings, MTA, 63 (John) [will] leave, TAB, 59

I I
when Mary sings, TA, 67,4 (John) [leaves], TAB, 56
I
Wh~Y) [sings] (John) [leaves], t, 19
TSC at-that-time4' t, 60
I
(Mary) [sings] at-that-tiae4' TAB, 6,b
------------
at-that-time4' MTA (Mary) [sings], TAB, 58
I
(Mary) [sings], TAB, 56
I
(124 ) Vt[vt 1 [t = t1 & vt 2 [t 2 = t1 & [PRES(t 2 ) &

AT(t 2 , Mary-sings')]]] & [FUT(t) &
unfortunately, (124) is provably contradictory. since no

interval will ever satisfy both 'FUTa)' and 'PRESa)'. If
rule (93) is instead employed in the analysis of wben Mary
sings, as in (126) (see next page), then (122) is assigned
translation (125), which is noncontradictory.
(125 ) Vt[vt,[t = t,& [NONPAST(t 1 ) &

AT(t" Mary-sings')]] & [FUT(t) &
( 126) when Mary sings, MTA, 63
I
---------
when Mary sings, TA, 69
when, TSC (Mary) [sings] , TAB, 56

I
Since (122) clearly has a noncontradictory interpretation,

deri v ations like (126) are necessary; analogous derivations
are required for the analysis of the temporal adverbials in
sentences like (127) and (128).
(127 ) John will leave before Mary sings.

(128 ) John will leave after Mary sings.
Rules (92) and (93) can also be employed in the

analysis of when-, before-, and after-adverbials like those
in (129)-(134) (where (132)-(134) are understood as
narrati ve presents).
( 129) John left when Mary arrived.

( 130) John left before Mary arrived.
( 131) John left after Mary arrived.
( 132) John leaves when Mary arrives.
(133 ) John leaves before Mary arrives.
(134 ) John leaves after Mary arrives.
In these cases, however, (92) and (93) don't give rise to

any interpretations that can't be produced by means of
( 104) •
1.4.3. Temporal Adverbial Clauses with Main Tense Adverbs
In the two preceding subsections, the temporal adverbial

clauses which have been examined have been rather simple.
Somewhat more complicated are adverbial clauses like those
in (135)-(137); here, the subordinate clauses are themselves
modified by main tense adverbs.
( 135) Before Jane arrived yesterday, John left.

(136 ) While Mary was singing yesterday, John left.
(137 ) When Jane said that Mary arrived yesterday, John
left.
138 CHAPTER III
It is a virtue of the present analysis of tense and time

adverbs that it assigns the desired interpretations to such
sentences without any additional machinery.
Sentence (135), for example, may be derived as in
(138), and is assigned the translation (139).
( 138) before Jane arrived yesterday, (John) [left],

I t, 60
before Jane arrived yesterday, (John) [left], TAB, 61a
~
before Jane arrived yesterday, (John) [left], TAB, 57
MTA, 63 I
\ (John) [leaves], TAB, 56
before Jane arrived yesterday, I
~ TA,67,4 (John) [leaves], t, 19
before, (Jane) [arrived] at-that-time4 yesterday, t, 60

TSC I
(Jane) [arrived] at-that-time 4 yesterday, TAB, 61b
yesterday,
MT A, 63---------------
(Jane) [arrived] at-that-time4' TAB, 61b
____________
I at-that-tiae4 , (Jane) [arrived], TAB, 57
yesterday, MTA I
TA (Jane) [arrives], TAB, 56
I
(Jane) [arrives], t, 19
(139) Vt[vt 1 [[t < t1 & M(t,t 1 )] & vt 2 [yesterday'(t 2 ) &

[t 2 = t1 & [past(t 2 ) & AT(t 2 , Jane-arrives')]]]] &
Note that the iteration of rule (63) (by which the main
tense adverbs yesterday and at-that-time4 are added on
successively in the derivation of the before-clause in
(138» is critical for this derivation.
Since rule (104) never figures in the derivation of
while-clauses, sentence (136) receives an analysis somewhat
different from (138), namely (140). In (140), rule (92)
rather than rule (104) is used to generate the temporal
adverbial clause. By this analysis, (136) is assigned the
interpretation of (141).
(140) while Mary was singing yesterday, (John) [left],
I t, 60
while Mary was singing yesterday, (John) [left], TAB, 61a
while Mary was singing yesterday,
I
MTA, 63 ----------(John) [left], TAB, 57
I
(John) [leaves], TAB, 56
while Mary was singing yesterday, I
~ TA,68 (John) [leaves], t, 19
while,
TSC
---------
(Mary) [is] singing yesterday, TAB, 6'b
yesterday,
MTA, 63
I
yesterday, TA
(Mary) [is] singing, TAB, 56
I
(Mary) [is] singing, t, 19
( 141) Vt[vt 1 [[t ~ t, & ~ment(t,)] & [past(t,) &

[yesterday'(t,) & AT(t" Mary-is-singing')]]] .&
(Note that because rule (92) can be used to generate before-

clauses, it is possible to use (92) instead of rule (104)
for generating the adverbial clause in sentence ('35); in
this case, sentence (135) receives an analysis parallel to
(140), but is thereby assigned an interpretation equivalent
to (139).)
The analysis presented here predicts a four way
ambigui ty for sentence (137).
When Jane said that Mary arrived yesterday, John

left.
The predicted ambiguity corresponds to the four possible

analyses for the adverbial clause when Jane said that Mary
arrived yesterday:
140 CHAPTER III
(1) the analysis (142) corresponds to the
interpretation in which (a) yesterday indicates the time of
Mary's alleged arrival, and (b) the when-adverbial denotes
the set of intervals at which Jane spoke;
(142) when Jane said that Mary arrived yesterday,

~ TA,67,4
when, (Jane) [said] that Mary arrived yesterday
TSC at-that-ti~4' t, 60
I
(Jane) [said] that Mary arrived yesterday at-that-time4'
~ TAB,61b
at-that-time4'
MTA I
(Jane) [said] that Mary arrived yesterday,
TAB, 57
(Jane) [says] that Mary arrived yesterday, TAB, 56
I
(Jane) [says] that Hary arrived yesterday, t, 19
-------------
Jane, T [say] that Mary arrived yesterday, lVi, 21
I
[say] that Hary arrived yesterday, IVs , 27
~
---------
[say], IVs/t (Mary) [arrived] yesterday, t, 60
I
(Hary) [arrived] yesterday, TAB, 61b
yesterday, MTA, 63 (Hary) [arrived], TAB, 57

I I
yesterday, TA (Hary) [arrives], TAB, 56
I
(Hary) [arrives], t, 19
(ii) the analysis (143) corresponds to the

interpretation of sentence (137) in which (a) yesterday
indicates the time of Jane's speaking, and (b) the when-
adverbial denotes the set of intervals at which Mary
allegedly arrived;
( 143) when Jane said that Mary arrived yesterday,

TA, 67,4
~
when,
(Jane) [said] that Mary arrived at-that-time4
TSC yesterday, t, 60
I
(Jane) [said] that Mary arrived at-that-ti.e4 yesterday,
~ TAB,61b
yesterday, (Jane) [said] that Mary arrived at-that-time4'
M1 A, 63 I TAB, 57
yesterday, (Jane) [says] that Mary arrived at-tbat-time4'
TA I TAB, 56
---------
(Jane) [says] that Mary arrived at-tbat-time4' t, 19
Jane, T [say] that Mary arrived at-that-tiae4 , IV i , 21

I
-----=------
[say] that Mary arrived at-that-time4' IV s , 27
[say], IVs/t (Mary) [arrived] at-that-time4' t, 60

I
(Mary) [arrived] at-that-time4' TAB, 61b
~
at-that-ti.e4' MTA (Mary) [arrived], TAB, 57
I
(Mary) [arrives], TAB, 56
I
(Mary) [arrives], t, 19
142 CHAPTER III
(iii) the analysis (144) corresponds to the
interpretation of (137) in which (a) yesterday indicates the
time of Jane's speaking, and (b) the when-adverbial denotes
the set of intervals at which Jane spoke; and
(144) when Jane said that Mary arrived yesterday,

TA, 67,4
~
when,
(Jane) [said] that Mary arrived at-that-time4
I
(Jane) [said] that Mary arrived at-that-time4 yesterday,
~ TAB,61b
yesterday, (Jane) [said] that Mary arrived at-that-time4'
MTj' 63 ~ TAB, 61b
yesterday, at-that-time4 , (Jane) [said] that Mary
TA MTA arrived, TAB, 57
I
(Jane) [says] that Mary arrived, TAB, 56
I
(Jane) [says] that Mary arrived, t, 19
Ja~] that Mary arrived, lVi, 21

I
[say] that Mary arrived, IVs, 27
-----.:----
[say], IVs/t
(Mary) [arrived]. t. 60
I
(Mary) [arrived]. TAB. 57
I
(Mary) [arrives]. TAB, 56
I
(Mary) [arrives]. t. 19
(iv) the analysis (145) corresponds to the
interpretation in which (a) yesterday indicates the time of
Mary's alleged arrival, and (b) the when-adverbial denotes
the set of intervals at which Mary allegedly arrived;
(145 ) when Jane said that Mary arrived yesterday,

TA, 67,4
~
when,
(Jane) [said] tbat Mary arrived
I
(Jane) [said] that Mary arrived at-tbat-time4 yesterday,
TAB, 57
I
I
(Jane) [says] that Mary arrived at-that-time4 yesterday,
TAB, 56
(Jane) [says] that Mary arrived at-that-time4 yesterday,
~ t,19
I
Jane, T [say] tbat Mary arrived at-tbat-time4 yest~rday,
Iy1, 21
[say] that Mary arrived at-that-time4 yesterday, IYS, 27
[say], Iys/t (Mary) [arrived] at-that-t~4 yesterday,

I t, 60
-----------
(Mary) [arrived] at-tbat-time 4 yesterday, TAB, 61b
yesterday, (Mary) [arrived] at-tbat-time4' TAB, 61b

MTA, 63 _________
I at-that-time4 , (Mary) [arrived], TAB, 57
yesterday, MTA I
TA (Mary) [arrives], TAB, 56
I
(Mary) [arrives], t, 19
My intuition is that this prediction is correct--that

these are all possible ways of interpreting the adverbial
clause in (137).
144 CHAPTER III
1.4.4. Remarks on Some Unacceptable Temporal Adverbials
The rules proposed in 1.4.1,2 for the derivation of temporal

adverbial clauses allow a number of kinds of unacceptable
sentences to be generated; examples are (146)-(150).
( 146) *John will leave when Mary arrived.

("147) *John left after Mary is singing.
( 148) *John will leave after Mary arrived.
(149) *John left before Mary arrives.
( 150) *John will leave when Mary will arrive.
I shall briefly explain what I believe to be the reasons for

this unacceptability.10
The unacceptability of sentences like (146) and (147)
is, I believe, to be attributed to their semantics:
according to the rules presented above, (146) and (147)
should have only contradictory interpretations. (146), for
example, is translated as (151) by the present system of
rules;
( 151) Vt[vt 1 [t = t1 & vt 2 [t 2 = t1 & [past(t 2 ) &

AT(t 2 , Mary-arrives')]]] & [FUT(t) &
because no interval could ever satisfy both 'FUT(~)' and

'past(O' (at an index of possible utterance), (151) induces
a contradictory interpretation for (146). The
unacceptabili ty of a wide range of sentences can be
accounted for in this way.
The unacceptability of (148) and (149), however, cannot
obviously be explained in this manner. The translations
assigned to (148) and (149) by the rules presented here do
not induce contradictory interpretations:
(152) Vt[vt 1 [[t 1 < t & M(t,t 1 )] & vt 2 [t 2 = t1 &

[past(t 2 ) & AT(t 2 , Mary-arrives')]]] & [FUT(t) &
( 153) a. vt[Vt 1 [[t < t1 & M(t,t 1 )] & Vt 2 [t 2 = t1 &

[PRES(t 2 ) & AT(t 2 , Mary-arrives')]]] &
b. Vt[vt 1 [[t < t1 & M{t,t 1 )] & [NONPAST{t 1 ) &

AT{t" Hary-arriyea')]] & [paat(t) &
AT(t, John-Ieayes')]]
There are several approaches that could be taken in order to

account for the unacceptability of (1118) and (149).
Translations (152) and (153a,b) could be converted to
contradictions if '<I were replaced with '«~I, where the
latter is defined as follows: where ~,~ denote the
intervals i,i', ~«~ is true iff i < i' and either (a)
[past( ~) & paste ~)] is true, or (b) [NONPAST( ~) &
NONPAST(~)] is true. (Heinamaki (1974:72f) makes a
suggestion something like this.) If (152) and (153a,b) were
changed into contradictions in this way, the anomaly of
(1118) and (149) would be comparable to that of (146) and
(147). Promising though this approach might appear to be, I
don't believe it is correct. It implies that sentence (154)
should not be true in a situation in which the time
predicted by Mary for John's departure is already past, and
that (155) should not be true in a situation in which the
predicted time hasn't yet arrived; the judgement is
delicate, but I believe that this implication is false in
bot.b cases.
(1511) John will leave after Mary said he would.

(155 ) John left before Mary said he would.
Smith (1975:72) suggests that the anomaly of sentences

like (148) and (149) (and like (1116) and (147» is syntactic
rather than semantic in nature. She proposes 'a principle
that will predict the possible tenses of embedded time
adverbials': 'an embedded time adverbial may have any tense
that occurs in the main sentence.' (Here, 'tense' refers to
superficial tense-marking.) This principle does rule out
offensive cases like (146)-(149); unfortunately, it would
also rule out perfectly acceptable cases like (154) and
(156), and possibly also (155) and (157).
(156 ) John leaves when Mary said he would.

(157 ) John left when Jane says that Mary arrived.
I therefore doubt that the anomaly of (148) and (149) is

purely a superficial one.
My belief is that the unaccep~ability of (1118) and
146 CHAPTER III
(149) is pragmatic in nature. When a speaker uses a

sentence of the form cP after 1/J, cP before 1/J , and so on, slhe
normally pragmatically presupposes 11 the truth of 1/J, as
Heinamaki (1974:99ff) has observed; for example, when
uttering (158), a speaker presupposes the truth of (159).
(158) John will leave when Mary arrives.
(159) vt[NONPAST(t) & AT(t, Mary-arrives')]
This means that the use of (148) would indicate the

presupposition of (160).
(160) vt[past(t) & AT(t, Mary-arrives')]
In the presence of this presupposition, the assertion of

John will leave would suffice to establish the succession of
John's departure and Mary's arrival. Thus, the information
conveyed by an assertion of (148) could always be conveyed
by a much simpler assertion; for this reason, (148) may be
anomalous because it inherently violates a principle of
conversational economy (cf. Grice (1975:67), stalnaker
(1978:325)). Analogous remarks are possible for example
(149) and other, similar sentences (cf. Heinamaki 1974:50f).
Sentences like (150), whose adverbial clause has future
tense-marking, are especially problematic;12 they aren't
assigned contradictory interpretations, as the translation
(161) of (150) shows;
( 161) vt[vt 1[t = t1

& vt 2 [t 2 = t1 & [FUT(t 2 ) &
AT(t 2 , Mary-arrives')]]] & [FUT(t) &
and they aren't pragmatically anomalous in the way (148) and

(149) were shown to be. Furthermore, as Smith (1975;73)
points out, sentences like (150) are considerably better
when the adverbial shows verb phrase ellipsis;
(162 ) John will leave when Mary will.
And before-adverbials with future tense-marking are in

general quite acceptable:
?John will leave before Mary will arrive.

(164 ) John will leave before Mary will.
I find these facts quite puzzling, and have no fully

adequate explanation at present. Nevertheless, sentence
(150) may well be unacceptable for pragmatic reasons:
because of the meaning of when, sentence (150) is true iff
(165) is true, where (165) translates as (166).
(165 ) John will leave when Mary arrives.

(166) Vt[Vt 1 [t = t1 & [NONPAST(t 1 ) &
AT(t" Mary-arrives')]] & [FUT(t) &
Because (165) is structurall~~simpler than (150), perhaps it

is systematically preferred. j This seems like a reasonable
approach to (150). Unfortunately, it can't easily be
generalized to cover the anomaly of (167);
(167 ) *John will leave after Mary will arrive.
the meaning of after makes (167) nonequi valent to (168) in

the present system. I leave the problem open for the
moment.
(168) John will leave after Mary arrives.
1.4.5. Remarks on the Assumed Truthconditions for Temporal

Adverbial Clauses
In this subsection, I shall briefly discuss the

truthconditions which the above system assigns to sentences
with temporal adverbial clauses by virtue of the
translations (91a-d) of the temporal subordinators when,
while, before, and after. In particular, I shall contrast
these truthconditions with those proposed by Heinamaki
(1974).
(91) a. before translates as

AptAtVt1[[t < t1 & M(t,t 1 )] & pt{t1l]
b. after translates as
AptAtVt1[[t1 < t & M(t,t,)] & pt{t1}]
'48 CHAPTER III
c. when translates as AptAtvt,[t = t1 & pt{t1}]

d. while translates as
AptAtvt,[[t ~ t, & ,.oment(t,)] & pt{t,}]
Before I begin, let me point out two peculiarities of

Heinamaki's system. First, her truthconditions don't
account for the interpretations of wheD-, beCore-, and
after-adverbials in which the 'extraction site' is in an
embedded clause (see section '.4.2), as M. Geis has pointed
out to me; I shall ignore this rather severe shortcoming in
the ensuing discussion. Second, for reasons which are not
made clear, she regards (p.'7) a sentence ¢ as true at a
nonminimal interval i only if i is a maximal interval of¢'s
truth--one having no proper super intervals at which ¢ is
true. This assumption is contrary to usual conceptions of
the entailments of sentences with stative and activity
predicates; furthermore, it complicates her truthconditions,
and in fact makes them incorrect in certain circumstances.
This will be brought out in the discussion of the different
kinds of temporal adverbials.
, .4. 5. 1. When.
According to the fragment developed here, a sentence of the

form ¢ when 1jJ is true only if there is some interval at
which the tenseless forms of ¢ and 1jJ are both true. 14 This
is intuitively correct. Heinamaki, however, has argued for
somewhat more complicated truthconditions for when:
¢ when 1jJ is true if and only if

(i) ¢ is true at an interval i,
(ii) 1jJ is true at an interval i', and
(iii) a. i f ¢ or 1jJ (or both) is durati ve [i.e.
has a stative or activity predicate--
GTS], then there is an interval i" such
that i" ~ i and i" =- i'. (If ¢ is an
accomplishment, then i = i"; if 1jJ is an
accomplishment, then i' = i".)
b. if nei ther ¢ nor 1jJ is durati ve, then i'
< i. (Cf. p.27)
(iiia) is the set of truthconditions for ¢ when 1jJ in
those cases in which one (or both) of ¢ and 1jJ has a stati ve
or activity predicate. In (Hia), the truth of <p when 1/J is

equated with the intersection (rather than the identity) of
the intervals at which <P and 1/J are true. This is to account
for several kinds of facts:
(a) that <P when 1/J can be true if the intervals at
which ¢ and 1jJ are true partially overlap. This situation
can arise when both ¢ and 1/J are 'durati ve', i.e. have
stative or activity predicates. Thus, (170) is true if the
interval at which it was raining only partially overlaps
with the interval at which we were in New Orleans, as in
(171).
(170) It was raining in New Orleans when we were there.
( 171) i1
~
I I
-
~
"" ~
i3
--- .............
i2
It was raining in New Or leans at i 1•
We were in New Orleans at i 2 •
(Heinamaki assumes (p.17) that sentences with achievement

and 'fast accomplishment' predicates can only be true at
moments;15 this, plus the parenthesized restriction on
accomplishments in (Hia), guarantees that ¢ when 1jJ can't be
true if the intervals at which ¢ and 1/J are true only
partially overlap and ¢ or 1jJ has an accomplishment or
achievement predicateJ
(b) that ¢ when 1jJ can be true if the interval at which
<P is true properly contains that at which 1/J is true. This
situation can arise whenever ¢ is durative. Thus, (172) is
true if the interval at which we were crossing the street
properly contains that at which John noticed us, as in
(173) (see next page).
( 172) We were crossing the street when John noticed us.
(c) that ¢ when

can be true if the interval at which
1/J
¢ is true is properly contained by that at which 1/J is true.
This situation can arise whenever 1jJ is durative. Thus,
(174) is true if the interval at which the balloon broke is
properly contained in the interval at which Lydia was
150 CHAPTER III
( 173)
We were crossing the street at i 1 •

John noticed us at i 2 •
(174) The balloon broke when Lydia was playing with it.
(175 )
I I I
The balloon broke at i 1 •

Lydia was playing with it at i 2 .
playing with it, as in (175).

(d) that ¢ when t can be true if the intervals at
which ~ and ware true are identical. This situation can
arise when either ¢ or t is durative. Thus, sentence (170)
is true if the interval at which it was raining in New
Orleans is identical to the interval at which we were there,
as in (176).
( 176)
( )
It was raining in New Orleans at i 1 •

We were in New Orleans at i 2 •
Thus, according to Heinamaki, ~ when W can be true no

matter how the intervals at which ¢ and t are true intersect
(provided that ~ or 1/1 is durative).
There is, however, an important assumption behind
Heinamaki's claim that when ~ or <ji is durative, ~ when W is
true iff the intervals i,i' at which ¢ and t are true
intersect: this is the assumption (pointed out above) that
i and i' must be maximal intervals of the truth of ¢ and \).I.
This isn't a necessary assumption, however; in fact, if it
is set aside, the truthconditions for ¢ when 1/1 can be
greatly simplified. Let me explain why this is so.
Recent work on the aspectual categories of English

predicates (e.g. those of Taylor (1977) and Dowty
(1979:163ff)) have shed much light on the entailments of
sentences with stative or activity predicates. Generally,
if a sentence has a stati ve predicate and is true at
interval i, then it is true at all subintervals of i,
including the minimal subintervals consisting of a single
moment. As Dowty (1979: 173ff) shows, 'interval statives'
such as stand in the corner, lie under the bed probably
don't give rise to an entailment quite this strong;
nevertheless, if a sentence with an interval stati ve
predicate is true at i, then it is true at all nonminimal
subintervals of i. The entailments to which acti vi ty
predicates give rise are somewhat more complicated, because
there are at least two kinds of activity predicates (cf.
Taylor (1977), Dowty (1979:170ff)): homogeneous activities
such as sleep and Cloat, and heterogeneous activities, like
chuckle, walk with a limp, and average 55 mph. If a
sentence with a homogeneous activity predicate is true at
interval i, then it is true at all nonminimal subintervals
of i; but if a sentence with a heterogeneous acti vi ty
predicate is true at i, then the most that can be said is
that it may be true at some nonminimal subintervals of i--in
some cases, it won't be true at any proper subinterval of i.
These entailments of stative and activity predicates
have rather profound consequences regarding Heinamaki's
truthconditions. For example, if ¢ and 1jJ both have
momentary stative predicates and the intervals at which they
are true overlap in any way, then there will automatically
be some interval at which both q, and 1/J are true. For
instance, i f (177) is true at il in (171) and (178) is true
at i 2 , then both (177) and (178) are true at i 3 •
(177 ) It is raining in New Orleans.

(178 ) We are in New Orleans.
Similarly, if one of the two sentences ¢ and 1jJ has a

momentary stative predicate and the interval at which it is
true properly contains the interval at which the other
sentence is true (no matter what sort of predicate the
latter has), then, again, there will necessarily be an
interval at which q, and 1jJ are both true. Thus, if (179) is
true at i1 in (173) and (180) is true at i 2 , then both (179)
and (180) are true at i 2 ;
152 CHAPTER III
( 179) We are crossing the street.

(180) John notices us.
and if (181) and (182) are respectively true at il and i2 in

(175), then both (181) and (182) are true at i 1•
( 181) The balloon breaks.

(182 ) Lydia is playing with it.
The implication of these facts is clear: in cases in which

~ or 1jJ has a momentary stati ve predicate but neither cp nor
1jJ has an interval stative or activity predicate, my
truthconditions for ~ when If; give exactly the same results
as Heinamaki's.
In cases in which ~ or 1jJ has an interval stative or
acti vi ty predicate, my truthcondi tions actually yield
superior results. Recall that the truth at i of a sentence
S with an interval stative or activity predicate does not
entail the truth of S at every moment throughout i, and in
fact may not entail the truth of S at any proper subinterval
of i if S happens to have a heterogeneous activity
predicate. In the light of this fact, consider the
following situation: John drives from Kansas City, Kansas,
to San Francisco, averaging 60 mph; however, knowing the
reputation of the Kansas Highway Patrol, he is careful not
to exceed 55 mph while still in Kansas. This situation may
be represented as in (183).
(183 )
)
------------~-----------
John averages 60 mph at i 1 •
John is in Kansas throughout i 2 , averaging 55 mph.
Heinamaki's truthconditions predict that (184) should be

true of this situation, because the interval at which John
was in Kansas intersects the interval at which John averaged
60 mph, even though John averages 60 mph isn't i tsel f true
at the interval of intersection (Le. i2 in (183». This is
plainly wrong.
(184) When he was in Kansas, John averaged 60 mph.

(184) requires that some interval at which John averaged 60
mph coincide with an interval at which he was in Kansas;
this is exactly what my truthconditions predict.
Further examples could be adduced to confirm the point:
in cases in which ~ or ~ has an interval stative or activity
predicate, Heinamaki's truthconditions allow ¢ when 1/1 to be
true even if there is no interval at which ¢ and 1/1 are both
true; my truthconditions, on the other hand, do require that
<P and ~ be simultaneously true. In every case, the latter
truthconditions make the right prediction. Thus, when <P or
1/1 has a stative or activity predicate, Heinamaki's
truthconditions for ¢ when 1/1 are either equivalent to mine
or are demonstrably inferior.
Part (iiib) of the truthconditions in (169) is somewhat
problematic. Heinamaki observes that when ¢ and 1/1 both have
accomplishment or achievement predicates, <P when ~ seems to
have truthconditions very diffe~ent from those in (iiia).
Consider sentences (185) and (186).
(185 ) When John wrecked the car, Bill fixed it.

( 186) When John broke his leg, he made a pair of
crutches.
It's clear that (185) is understood to mean that John

wrecked the car before Bill fixed it, and that (186) is
understood to mean that John broke his leg before he made a
pair of crutches. For this reason, Heinamaki suggests that
when both ¢ and 1jJ have accomplishment or achievement
predicates, <P when ~ is true iff ~ is true at an interval
which precedes the interval at which ¢ is true.
There is one important problem with (iiib), namely that
when ¢ and 1/1 both have accomplishment/achievement
predicates, <P when 1/J can sometimes entail the simultaneous
truth of ¢ and 1(1. Consider, for example, sentences (187)
and (188).
( 187> When John wrecked the car, he somehow managed to

jump to safety.
( 188) When John broke his leg, he also hurt his elbow.
(187) can easily be understood to mean that John jumped to

safety as he wrecked the car, and (188), to mean that John
hurt his elbow just as he broke his leg. My truthconditions
predict this; Heinamaki's do not. Note also that when-
adverbials can be explicitly contrasted 1011 th aCter-
154 CHAPTER III
adverbials, even when the subordinate and superordinate

clauses both have accomplishment/achievement predicates:
(189) a. Q: Did Mary arrive when John did?

A: No, she arrived (just) after he did.
b. Q: Did Mary arrive after John did?
A: No, she arrived (right) when he did.
This shouldn't be possible with Heinamaki's truthconditions,

but it's predicted by mine. Finally, note that a when-
clause which is modified by the adverb exactly clearly has
the interpretation which I have proposed, even in sentences
in which both the when-clause and the main clause contain
accomplishment/achievement predicates:
(190 ) Mary arrived exactly when John did.
Sentence (190) is understood to mean that Mary's arrival

coincided precisely with John's. Given Heinamaki's
truthconditions for when, it is unclear how such an
interpretation could be possible; in particular, it is
unclear how the divergence of (190)'s interpretation from
Heinamaki's truthconditions might be attributed to the
semantics of exactly. The function of this adverb is simply
to emphasize the measurable exactness of some proposition:
in sentence (191), for example, exactly emphasizes the
exactness of the proposition expressed by (192).
( 191) John is exactly as tall as Bill.

(192) John is as tall as Bill.
In (190), exactly may be similarly viewed as emphasizing the

exactness of the proposition expressed by (193), provided
that when is understood as specifying a relation of
simultaneousness in (193), as I have proposed.
(193) Mary arrived when John did.
Still, something must be said about sentences like

(185) and (186). On the one hand, a second when 2 might be
postulated with truthconditions different from the when, of
simultaneousness; this solution would, however, be purely ad
hoc, and would provide no obvious way of explaining why
~ and Wmust both have accomplishment/achievement predicates
in a sentence of the form ¢ when 2 w.
A better alternative, I believe, is to assume that when

always entails simultaneousness and to develop a pragmatic
account of sentences like (185) and (186). Consider
sentences (191) and (192) again. Clearly, the only reason a
sentence like (191) is useful is because the corresponding
sentence (192) may be used with less than complete
exactitude; (192), for example, can be truthfully asserted
if John's height differs from Bill's, provided that the
difference is viewed by the speaker and hearer as
negligible--that is, provided there is no height
intermediate between John's height and Bill's height from
which either is saliently distinguished. In short, in the
actual use of sentences like (192), some measure of
imprecision is tolerated; beyond a certain point, it simply
isn't useful to make finer and finer distinctions in height
(indeed, in Gricean terms, it may be uncooperative to do
so). My point is that (193) is analogous to (192) in this
respect--that (190) is useful precisely because (193) may,
like (192), be used with some degree of imprecision. Given
two events e 1 , e 2 describable by sentences cjJ 1 and cjJ 2'
respecti vely, speakers may describe this pair of events by
means of a sentence cjJ1 when cjJ2 even if e 1 and e 2 are not
simultaneous, provided that the temporal difference between
them is negligible--that is, provided there is no salient
event intermediate between e 1 and e 2 which makes it
important to distinguish them temporally.
In an imprecise use of sentence (192), either one of
John or Bill may be negligibly taller than the other.
Correspondingly, the events e 1 , e 2 described by a sentence
cjJ1 when cjJ2 may be such that either e 1 precedes e 2 or vice
versa. Real world knowledge may, of course, favor one or
the other of these possible interpretations. Thus, in the
most natural interpretation of (185), the event described by
the when-clause precedes that described by the main clause;
this interpretation is most natural because of the way in
which car repairs normally succeed car accidents. (Similar
remarks hold for sentence (186).) In the most plausible
interpretation of (194), however, the event described by the
when-clause follows that described by the main clause;
(194 ) When Lindbergh crossed the Atlantic, he chose Long

Island as his starting point.
obviously, Lindbergh's choice of a starting point for his

156 CHAPTER III
flight across the Atlantic was a precondition for the flight

itself.
If this pragmatic analysis of the use of when-clauses
is valid, then it is not only possible but desirable to
assume that when specifies a relation of simultaneousness in
all cases.
1.4.5.2. While
According to the system developed here, a sentence of the

form ¢ while 1/J is true only if there is an interval i and a
nonminimal interval i' such that the tenseless form of ¢ is
true at i, the tenseless form of 1/J is true at i', and i is a
subinterval of i,.16 This is intuitively accurate. The
requirement that i' be nonminimal is meant to capture the
fact that while-clauses don't have instantaneous achievement
predicates:
(195) *Phil had an idea while he spotted the elephant

(through his binoculars).
Heinamaki's truthconditions are only slightly more

complicated:
(196) ¢ while 1/J is true i f and only if

(i) ¢ is true at an interval i,
(ii) 1/J is true at an interval i' which is not a
moment, and
(iii) there is an interval i" such that i" c i and
i" c i'. If </> is an accomplishment, 1 = i";
if 1/J is an accomplishment, i' = i". (Cf.
p. 35)
Here, the truth of </> while 1jJ is equated with the

i n t e r sec t ion 0 f the i n t e r val 0 f ¢' s t rut h wit h the
(nonminimal) interval of 1jJ's truth. If ¢ has an
accomplishment predicate, then the interval of intersection
must be i; so also if </> has an achievement predicate, given
Heinamaki's assumption that sentences with achievement
predicates are true only at moments; likewise, i f 1/J has an
accomplishment or achievement predicate, the interval of
intersection must be i'. Thus, Heinamaki's truthconditions
differ from mine in two ways: (a) they don't require i to
be a subinterval of i' in cases in which ¢ has a stati ve or
activity predicate; and (b) they require i and i' to be

identical in cases in which both ¢ and 1jJ have accomplishment
or achievement predicates. As I shall show, neither of
these differences is well-motivated.
Consider first the cases in (a), in which cP has a
stative or activity predicate. Heinimikils truthconditions
actually differ from mine in only a subset of these cases.
Recall that if a sentence with a momentary stative is true
at an interval i, then it is true at every subinterval of i;
this means that i f ¢ has a momentary stati ve predicate aind
the interval i at which ¢ is true intersects the interval il
at which 1/! is true, then there is necessarily a subinterval
i"' of i at which ¢ is true and which is a subinterval of
it. Thus, her truthconditions are equivalent to mine as
regards cases in which ¢ has a momentary stative.
On the other hand, her truthconditions make distinct
predictions in cases in which ¢ has an interval stati ve or
activity predicate. These distinct predictions are easily
shown to be wrong, however. For example, according to the
truthconditions in (196), (197) should be true of the
si tuation represented by (183) in 1.4.5.1;
(197 ) John averaged 60 mph while he was in Kansas.
this is surely wrong. For (197) to be true, John must have

averaged 60 mph at some subinterval of an interval at which
he was in Kansas; this is just what my truthconditions
predict.
Now consider those cases in (b), in which both cP and
1jJ have accomplishment/achievement predicates. In these
cases, Heinamaki's truthconditions require i and il to be
identical, where i and il are as in (196). According to
this requirement, (198) is false unless the interval at
which Mary sailed around the world is identical to the
interval at which John wrote a novel.
(198) While Mary sailed around the world, John wrote a

novel.
This requirement seems too strong; according to everyone I

consul ted, (198) requires at most that John wrote a novel at
a subinterval of the interval at which Mary sailed around
the world. This is, of course, exactly what is predicted by
the truthconditions I have proposed.
Thus, where Heinamaki's truthconditions for while
158 CHAPTER III
diverge from mine, their predictions are clearly less
satisfying.
1.4.5.3. Before
According to the present analysis, a sentence of the form

<P before 1/1 is true only if an interval at which the
tenseless form of <P is true precedes an interval at which
the tense less form of 1/1 is true. 17 Heinamaki's
truthconditions are the following: 18
(199 ) <P before 1/1 is true if and only if

(i) <P is true at some interval,
(ii) 1/1 is true at some interval, and
(iii) the reference point of the interval at which
<P is true precedes the initial point of the
interval at which 1/1 is true. (Cf. p.49)
Here, Heinamaki introduces the notion of a reference point:

if a sentence S has an accomplishment predicate, then the
reference point of the interval at which S is true is its
final point; if S has some other sort of predicate, then the
reference point of the interval at which S is true is its
initial point. Clearly, the truthconditions in (199) yield
the same results as mine in those cases in which <P has an
accomplishment predicate--or an achievement predicate, given
Heinamaki's assumption that sentences with achievement
predicates are true only at moments; furthermore, because
the truth at i of a sentence S with a momentary stative
predicate entails the truth of S at all moments in i, the
truthconditions in (199) are equivalent to mine whenever
<P has a momentary stative predicate. Thus, the
truthconditions in (199) give different results from mine
only when <P has an interval stative or activity predicate.
(199) can easily be shown to produce incorrect results in
some such cases; for example, unlike the truthconditions
which I assume, those in (199) predict that sentence (200)
should be true of the situation represented in (183) in
1.4.5.1.
(200) John averaged 60 mph before he left Kansas.
This is again incorrect. (200) clearly requires that there

be some interval at which John averaged 60 mph and which
preceded his leaving Kansas; this is just what my
truthconditions for <P before 1/1 predict.
1 .4.5. 4. After
In the fragment developed here, a sentence of the form

<P after 1/1 is true only if an interval at which the tense less
form of <P is true follows an interval at which the tense less
form of 1/1 is true. 19 Heinamaki's truthconditions for
<P after 1/1 are those in (201).
(201) <P after 1/1 is true if and only if

(i) <P is true at some interval i,
(ii) 1/1 is true at some interval, and
(iii) there is some i' such that i' c i and the
reference point of the interval at which 1/1 is
true precedes if. (For some speakers, if
<P has an accomplishment predicate, then i' =
L) (eL p.72)
In the parenthesized remark in (iii), Heinamaki implies

that speakers differ in their understanding of sentences
like (202), in which the main clause has an accomplishment
predicate:
(202) John built the house after he got more money.
the suggestion is that for some speakers, (202) can be true

in a situation in which John got more money after he had
already begun to build the house, while for other speakers
it cannot. My judgment follows those of the intolerant
speakers. Furthermore, I suspect that the tolerant speakers
can regard (202) as true in such a situation because of a
certain vagueness in what it means to build a house. For
example, it seems to me that erecting a structure on a
foundation that has already been laid might sometimes count
as 'building a house'; thus, tolerant speakers might regard
(202) as true in a situation in which John had poured the
foundation of the house before he got more money. On the
other hand, adding walls, doors, windows, and siding to a
previously built frame seems less like 'building a house'; I
would accordingly expect many fewer of the tolerant speakers
to regard (202) as true in a situation in which John had
built the frame of the house before he got more money. My
guess is that nobody would think of (202) as true in a
160 CHAPTER III
situation in which John had done everything but shingle the

roof before he got more money. Notice, too, that when the
main clause of a sentence like (202) has an accomplishment
predicate that isn't especially vague, the sentence probably
can't have an interpretation analogous to the tolerant
interpretation of (202); (203), for instance, would not be
true in a situation in which Bill had already drawn part of
the circle with a crayon before he found his pen.
(203 ) Bill drew a circle after he found his pen.
I therefore conclude that the parenthesized portion of (iii)

in (201) represents an essential component of the
truthconditions for ~ after ~ for all speakers.
If this is conceded, then Heinamaki's truthconditions
are in many cases equivalent to mine--specifically, when
neither ~ nor ~ has an interval stative or activity
predicate. When ¢ or ~ does have such a predicate, however,
the truthconditions in (201) make distinct predictions from
mine. In particular, they predict that ¢ a£ter ~ can be
true even if there is no interval at which 1jJ is true which
precedes an interval at which ¢ is true. For example, they
predict that (204) is true of the situation represented in
(183).
(204) John was in Kansas after he averaged 60 mph.
This is plainly wrong. (204) requires that an interval at

which John averaged 60 mph precede an interval at which he
was in Kansas; this requirement is embodied in the
truthconditions which I have assumed. ThUS, here, as in the
preceding cases, Heinamaki's truthconditions are in some
cases equivalent to mine, but give incorrect results in
other cases.
1.5. SUMMARY OF TENSE AND TIME ADVERB SYSTEM
In this preliminary section, the syntax and semantics of

fi ve categories of expressions have been discussed. These
categories are listed in Table 3 with the logical types to
which they correspond (see next page).
The syntactic composition of expressions of these
categories is exemplified in the derivation (205); the homo-
morphic translation tree corresponding to (205) is (206).
Table 3
Five Categories and Their Corresponding Logical Types
Category Type
t t
TAB <i,t>
TA <i,t>
MTA «s,<i,t»,<i,t»
TSC «s,<i,t»,<i,t»
(205) when John arrived, (Mary) [left], t, 60
-----------
I
when John arrived, (Mary) [left], TAB, 61a
when John arrived, MTA, 63 (Mary) [left], TAB, 57

I I
when John arrived, TA, 67,4 (Mary) [leaves], TAB, 56
I
W~[arriVed] (Mary) [leaves], t, 19
TSC at-that-time4' t, 60
(206) a (where a-i are as below)

I
b
c
~g
I I
d h
e0 i
a. vt[Vt 1 [t = t1 & Vt 2 [t 2 = t1 & [past(t 2 ) &

AT(t 2 , John.-arrives')]]] & [past(t) &
AT(t, Mary-leaves')]]
b. At [Vt 1 [t = t 1 & Vt 2 [t 2 :: t 1 & [past( t 2) &
AT(t 2 , John-arrives')]]] & [past(t) &
AT(t, Mary-leaves')]]
162 CHAPTER III
c. Apt)'t[Vt, [t = t, & vt 2 [t 2 = t1 & [past(t 2 ) &

AT(t 2 , John-arrives')] J] & pt{t}]
d. At[Vt,[t = t, & vt 2 [t 2 = t, & [past(t 2 ) &
AT(t 2 , John-arrives')]]]]
e. AptAtVt,[t = t1 & pt{t,}]
f. vt 2 [t 2 = t4 & [past(t 2 ) &
AT(t 2 , John-arrives')]]
g. At[past(t) & AT(t, Mary-leaves')]
h. At[AT(t, Mary-leaves')]
Mary-leaves'
With these preliminary assumptions about tense and time

adverbs spelled out, a consideration of the ways in which
free adjuncts may interact with tense can now be undertaken.
2. The Temporal Reference of Free Adjuncts
In the translation (208) of sentence (207), the temporal

operator 'AT' plays a critical role.
(207) John left Kansas City in the morning.
(208) Vt[Vt 1 [morning'(t 1 ) & t =- t,] & [past(t) &

AT(t, John-leaves-Kansas-City')]]
'AT' has the effect of displacing John's departure, from the

interval at which (207) is evaluated: (208) (hence (207»
is true at interval i only if John left Kansas City at some
time before i. In this example, the operator 'AT' might be
regarded as the semantic concomitant of the superficial
tense-marking in (207).
Now consider sentence (209), in which an adsentential
adjunct is employed.
(209) Leaving Kansas Ci ty in the morning, John reached

St. Louis at noon.
In this sentence, the time of John's departure is similarly

displaced from the interval at which (209) is evaluated:
(209), like (207), is true at i only if John left Kansas

City at some time before i. This temporal displacement
cannot, however, be regarded as the effect of the 'AT'
operator corresponding to the tense of the superordinate
clause, because the free adjunct and the main clause in
(209) relate to different time intervals (as their time
adverbs 20 show). Accordingly, I conclude that the
translation of a free adjunct includes an 'AT' operator
separate from that of the translation of the superordinate
clause; that is, I conclude that the translation of (209)
must be roughly as in (210) (where K(L) is the implicit,
variable subordinator associated with an adsentential
adjunct--Chapter II, section 2.2.2).
(210) K(L)("vt[ in-the-morning'(t) &

AT( t, x-Ieaving-l:ansas-City')])(" John-reached-
St.-LOuis-at-noon' )
Given this assumption. the rules for deriving free

adjuncts can finally be formalized.
2.1. A-ABSTRACTS
Free adjuncts of all categories are, like tensed sentences,

regarded as deriving from temporal abstracts in the proposed
system. Here, I will refer to a temporal abstract from
which an adjunct may derive as an A-abstract; formally,
abstracts of this sort constitute the subclass of temporal
abstracts represented by the feature [+Adjunct].
Recall from section 1.2.1 that the temporal abstracts
which give rise to tensed sentences themselves derive from
tense less sentences, via rule (20). A-abstracts, on the
other hand, derive directly from predicative phrases, in
accordance with the following two rules.
(211) S 1. If a e: PPREDs, then F 1,n(a) e: PTAB ' [-Tense,

+Adjunct, +Weak, -Perfect], where F1,n(a) is
a.
Example: F1,4(walking home) = walking home

164 CHAPTER III
(211) [continued]
T1. If (X £ PpREDs and (X translates as (X', then

F 1,n(a) translates as.
At[AT(t, VXs[R(x s , x~) & a'(X s )])].
Example: F1 , 4(walking home) translates

. as
At[AT(t, Vxs[R(x s , xJ) & walking-home'(x S )])]
(212) 82. If a £ PpRED i , ¢.. PADJi, then F 2,n(a) £ PTAB '

[-Tense, +Adjunct, -v/eak, -Perfect], where
F 2,n (a) is a.
Example: F2 ,4(being a sailor) = being a sailor
T2. If a E PpREDi and a translates as a', then

o
F2 n(a) translates as At[AT(t, a'(x n »],
, k .
At [ AT ( t , a' ( x n ) ) ], 0 rAt [ AT ( t , ex' ( x~ ) ) ] ,
according as a' is of sorted type (o,t'>,
<k,t'>, or <[o,k],t'>, respectively.
Example: F2 ,4(being a sailor) translates as

At[AT(t, being-a-3ailOr'(X~»]
Rule (211) converts a stage-level predicate into a

temporal abstract classified as [+Adjunct, +vJeak]; in
accordance with the observations of Chapter II, section 3
(where it was shown that the weak/strong distinction among
free adjuncts correlates exactly with Carlson's distinction
between stage-level and individual-level predicates), the A-
abstracts generated by (211) will be used in the derivation
of weak adjuncts. Rule (212), on the other hand, converts
an individual-level predicate into a temporal abstract
classified [+Adjunct, -Weak]; such abstracts will thus give
rise to strong adjuncts. 21 Note that in the intensional
logic translation of an A-abstract generated by (211) or
(212), a free variable 'xn' is supplied; this, of course, is
to represent the 'understood subject' of an adjunct deriving
from this abstract. 22 Moreover, the translation assigned to
an A-abstract al ways contains the temporal operator 'AT',
as required by the earlier observation that the temporal

reference of a free adjunct is in principle independent of
that of its superordinate clause.
2.2. DERIVING CONDITIONAL ADJUNCTS
Given these rules for the derivation of A-abstracts, it is

now possible to formalize the rule generating conditional
adjuncts (i.e. those of category t'). The syntactic portion
of this rule is trivial: conditional adjuncts are always
identical in form with the A-abstracts from which they
derive. As was concluded in section 3.3 of Chapter II,
conditional adjuncts derive ultimately from stage-level
predicates; thus, they derive most directly from A-abstracts
in the subclass [+Weak], generated by rule (211) above. 23
(213) S10. If a e: PTAB , [+Adjunct, +Weak, -Perfect],

then F 10 (a) e: Pt " where F 10 (a) is a.
Example: F 10 (lying on the beach) = lying on the beach
Recall from Chapter II that conditional adjuncts denote

truth-values--that is, their translations are of type t.
Accordingly, I assume that, given an A-abstract a denoting a
set I of time intervals, the conditional adjunct deriving
from a is interpreted as the result of quantifying
existentially over intervals in I:
T10. If a e: PTAB and a translates as a', then

F 10 (a) translates as Vt[a'(t)].
Example: F 10 (lying on the beach) translates as

vt[AT(t, VxS[R(x S, xJ) &
lying-on-the-beach'(x S)])]
Because conditional adjuncts derive from temporal

abstracts, no extra rules are necessary to account for the
existence of conditional adjuncts containing time adverbs,
such as that in (214).
(214) Wearing that new outfit today, Bill would fool

everyone.
166 CHAPTER III
The A-abstract wearing that new outfit can combine

unproblematically with the main tense adverb today, in
accordance with the very general rule (63) for combining
time adverbs with temporal abstracts; thus, with the rules
developed so far, (214) is assigned the analysis in (215),
which induces the translation (216).
(215) wearing that new outCit today, (Bill) [would] fool

everyone, t, 60
I
wearing that new outfit today, CBill) [would] fool everyone,
TAB, 58
I
wearing that new outfit today, (Bill) [would] fool everyone,
TAB, 56
I
wearing that new outfit today, (Bill) [would] fool everyone,
t, 18
[would] wearing that new CBill) [fools] everyone,

MOD outfit today, t', 10 t, 19
I
wearing that new outfit today, TAB, 61
I -------------
today, MTA, 63
today, TA
wearing that new outfit, TAB, 1,4
I
wearing that new outfit, PRPL s
(216) Vt[PRES(t) & ATCt, would'(DCcb)(A Vt1 [today'Ct 1) &

AT(t 1 , VxS[R(x s , x~) &
wearing-that-new-outfit'(xS)])]»(~Bill-fools
everyone'»]
(216), of course, represents the conditional interpretation

of (214): 'if he wore that new outfit today, Bill would
fool everyone.'
2.3. DERIVING ADSENTENTIAL ADJUNCTS
Recall from Chapter II, section 2.2.2 that free adjuncts

such as those in (217) and (218) can be informally
classified as adsentential: unlike conditional adjuncts,
they do not restrict the interpretation of a modal, but
somehow serve to modify their superordinate clause.
(217) Being a master of disguise, Bill would fool

everyone.
St. Louis at noon.
In Chapter II, it was proposed that such adjuncts have an

implicit, variable subordinating conjunction K(L), as in the
tentative translations (219) and (220).
(219) K(L)("'x-being-a-master-of-disguise') ("'Bill-vould-

fool-everyone' )
(220) K(L)( "'vt[ in-the-.orning'( t) &

ATCt, I-leaving-Kansas-City')])C"'John-reached-
St.-Louis-at-noon' )
No formal account of adsentential adjuncts was proposed,

however; their category membership was left open, and the
rules necessary for deriving them and introducing tbem into
a larger sentence were left unformulated.
Here. I propose that adsentential adjuncts are main
tense adverbs. That is, I propose that they belong to the
category MTA introduced in section 1.3.2 of the present
chapter, and that they therefore join with a temporal
abstract to produce a temporal abstract, in accordance with
rule (63). For the moment, I will simply elaborate the
formal details of this hypothesis; in section 2 of Chapter
IV, however, I shall offer a fairly compelling argument in
favor of the classification of adsentential adjuncts as main
tense adverbs.
The syntactic rule necessary for converting A-abstracts
into main tense adverbs is strai~~tforward: no change in
form results from this conversion.
(221) S11. If a £ PTAB , [+Adjunct, -Perfect], then

F 11 ,m,n(0I) £ PMTA , where F11,m,nCOI) is 01.
Example: . F 11 ,m,n(leaving Kansas City in the .arning) =
leaving Kansas City in the morning
Note that the A-abstracts to which (221) applies mayor may

168 CHAPTER III
not be [+Weak]; thus, adjuncts of category.MTA may have
either stage-level or individual-level predicates, in
accordance with the conclusions of Chapter II.
The translation rule corresponding to (221) is not so
easily arrived. at. As a preliminary hypothesis, consider
(222) as a possible formulation of this translation rule.
(222) T11. (tentative)

If II E PTAB and II translates as a', then
F11 ,m,n(a) translates as
AptAt[K(Lm)(Avt1[a'(t1)])(A[pt{t}])]
To appreciate the effects of (222), consider the translation

which it would cause to·be assigned to sentence (223).
(223) Leaving Kansas City, John reached St. Louis at

noon.
As an A-abstract, the expression (224) translates as in
(225) •
(224) leaving Kansas City
(225) At[AT(t, Vxs[R(x s , x~) &

leaving-Kansas-City'(x s )])]
If (224) were converted into a main tense adverb by (221),

rule (222) would assign it the translation (226).
(226) AptAt[K(L6)(Avt,[AT(t1' Vxs[R(X s , X~) &

leaving-Kansas-CitY'(xs)])])(A[pt{t}])]
By rule (63), (224) would, as a main tense adverb, join with

the tensed temporal abstract (227) with translation (228);
the resulting temporal abstract (229) would translate as in
(230) •
(227) (John) [reached] st. Louis at noon
(228) At[at-noon'(t) & [past(t) &

AT(t, John-reaches-St.-Louis')]]
(229) leaving Kanaaa City, (Joba) [reacbed] St. Louis at
noon
(230) At[K(L6)(AVt1[AT(t1' Vxs[R(x s , x~) &

leaYing-Kansas-City'(xs)])])(A[at-noon'(t) &
[past(t) & AT(t, Jobn-reacbes-St.-Louis')]])]
By rule (24), (229) could then be converted into a tensed

sentence translating as in (231).
(231) Vt[K(L 6 )(AVt,[AT(t" Vxs[R(x s , xt) &

leaYing-Kansas-City,(xs)])])(A[at-noon'(t) &
[past(t) & AT(t, JobD-reacbes-5t.-Louis')]])]
This sort of translation for sentence (223) is very

close to being satisfactory; there is one refinement to be
made, however, which necessitates a slight change in the
statement of the translation rule (222) corresponding to the
syntactic rule (221). Observe that in (231), the first
argument of the variable subordinator K(L6) is (232).
(232) A
Vt, [AT(t" Vx s[ R(s i
x, x4) &
leaYing~ansas-City'(xs)])]
(232) is an eternal proposition; that is, it is either true

at all time intervals or false at all time intervals
(depending on wheth~r or not there is some interval--any
interval--at which x: (= John) is leaving Kansas City). As
a consequence, the variable L6 in (232) caGoot be assigned a
value which specifies a particular temporal relation between
the event to which its first argument relates and that to
which its second argument relates; for example, if L6 were
assigned the value of (233), (232) could be true even if the
only time at which John left Kansas City followed the time
at which he reached St. Louis.
(233)
This is clearly a deficiency in (231) as a translation for
(223), since an important part of the logical relation
inferred to hold between the adjunct and the main clause in
(223) is temporal in nature: the event picked out by the
adjunct in (223) is almost inevitably inferred to precede
170 CHAPTER III
the event to which the main clause relates.

To account for the possibility of inferring a temporal
relation between an adjunct of category MTA and its main
clause, I propose the following revised translation rule for
the rule (221) creating adjuncts of category MTA:
(234) Tll. (final)
If a E PTAB and a translates as at, then

F 11 ,m,n(a) translates as
AptAt[K(Lm)(AVt,[Mn(t,t,) & a'(t,)])(A[pt{t}])].
By (234), the A-abstract (224) translating as (225) may be

converted into a main tense adverb translating as (235).
(235) Apt At[K(L 6 )( .... vt 1 [M 9 (t,t,) & AT(t 1, Vxs[R(x s , x~) &
leaving-Kansas-City'(xs)]»))( .... [pt{t}])]
Wi th (235) being used in place of (226), a new translation

can be built up for sentence (223):
(236) Vt[K(L 6 )( .... Vt,[M9 (t,t,) & AT(t" Vxs[R(x s , xt) &
leaving-Kansas-City'(x s )]) )( .... [at-noon'(t) &
[past(t) & AT(t, JobD-reaches-St.-Louis')]]»)
Note the free variable Mg of type <i,<i,t» introduced by

rule (234): M9 takes as its value a binary relation between
time intervals. As with L6 , the value of M9 isn't specified
by rules of semantic interpretation, but must instead be
inferred by language users. The inclusion of this
additional variable in the translation of an adjunct of
category MTA makes it possible to account for the inference
of a particular temporal relation between an adjunct of this
category and its superordinate clause: thus if M9 is
inferred to have the value of (237), then (236) (hence
(223» is true only if John's departure from Kansas City
preceded his arrival in St. Louis.
Pragmatically speaking, the variable Mg plays the same

kind of role in the proposed analysis as tne variable L6:
both represent aspects of an expression's denotation which
the semantic rules of the language fail to determine.
Moreover, only certain assignments of values to these
variables are pragmatically admissible, namely those
conforming to the inferences of language users. In short,
my claim is that a sentence such as (223) cannot be assigned
an interpretation unless the semantic indeterminacy
represented by the variables L6 and Mg is resolved
inferentially. See Chapter VI for a general discussion of
such inferences.
Rules (221) and (234) can be used in the analysis and
translation of free adjuncts more complicated than the one
in sentence (223). For example, the free adjunct in (238)
itself contains a main tense adverb:

St. Louis at noon.
No new rules are necessary for the analy~is and translation

of this sentence, however:
(239)
leaving Kansas City in the morning,
(John) [reached] St. Louis at noon, t, 60
I
leaving Kansas City in the morning, (John) [reached] St. Louis at"noon,
-------
TAB, 61a
~
leaving Kansas City in (John) [reached] st. Louis at noon, TAB, 61b
the morning, MIA, 11,6,9
I
leaving Kansas City in
the morning, TAB, 61b
at noon, ~lTA, 63
I
at noon, TA
I -
(John) [reached] St. LoU1S,
TAB, 57
I
~ (John) [reaches] St. Louis,
in the morning, leaving Kansas City, TAB, 56
i1TA, 6::; I TAB,1,4
(John) [reaches] st. Louts,
in the mornin!!., leaving Kansas City, t, 19
TA PRPL s
(240) [vt 2 [ morning'(t 2 )

Vt[K(L 6)(A Vt1 [M 9 (t,t 1) & &
t, ~ t 2 ] & AT(t"Vxs[R(x s , xt) &

leaving-Itansas-City'(x s )])] ])(A[noon'(t) &
[past( t) & AT(t, John-reaches-St.-Louis')]])]
In section 2 of Chapter IV t a compelling argument in

favor of analyzing ad sentential adjuncts as main tense
adverbs (as I have proposed here) is presented.
In this section, it has been shown that the temporal
172 CHAPTER III
reference of a free adjunct is independent of that of its
superordinate clause; accordingly, it is necessary to assume
that the translation of a free adjunct contains a temporal
operator ('AT') analogous to that signalled by the
superficial tense-marking of the superordinate clause. To
guarantee the appearance of such an operator in the
translations of free adjuncts, I have proposed that free
adjuncts of all categories derive from temporal abstracts
(specifically, from A-abstracts); under this analysis, the
appearance of main tense adverbs in free adjuncts follows
directly from rule (63), motivated independently in section
1.3.2 of this chapter. Rules have been proposed for
deriving A-abstracts from predicative phrases of various
kinds: A-abstracts deriving from stage-level predicates
(E PPREDs) are subclassified [+Weak], and thus give rise to
weak aojuncts of various categories, while A-abstracts
deriving from individual-level predicates (E PpREDi) are in
the subclass [-Weak], and therefore give rise to strong
adjuncts (which are al ways of category MTA in the present
system). In accordance with rule (213), conditional
adjuncts (i.e. of category t') derive from A-abstracts in
the subclass [+Weak]; (213) thus embodies the generalization
(uncovered in Chapter II) that conditional adjuncts must
ultimately derive from stage-level predicates. By rule
(221), adsentential adjuncts derive from A-abstracts of
ei ther subclass; I have proposed that such adjuncts are in
fact in the category MTA of main tense adverbs.
In the remaining two sections of this chapter, it is
proposed that free adjuncts may also serve as expressions of
one additional category--the category TA of set-level time
adverbs.
3. Frequency Adverbs and the Distinction between strong

and Weak Adjuncts
In Chapter II, it was shown that two varieties of free

adjuncts--strong and weak--can be distinguished on the basis
of their entailment possibilities in modal contexts. As I
shall show in this section, strong and weak adjuncts have
distinct interpretations in other sorts of contexts as well.
Consider sentences (241)-(244).
(241 ) B. Being B sailor, John sometimes smokes a pipe.


pipe.
(242) a. Weighing four tons, our truck often makes the
bridge shake.
b. Carrying a load of over 1500 lbs., our truck
often makes the bridg~ shake.
(243) a. Having only ten minutes between her morning
lunch.
b. Studying for finals, Anne never eats lunch.
calculator.
The b sentences, whose adjuncts are weak, have

interpretations on which the frequency adverb in the main
clause seems to quantify intervals at which the adjunct is
true; the a sentences, which have strong adjuncts, do not
allow such an interpretation. This difference between
strong and weak adjuncts is reminiscent of that observed in
modal contexts in Chapter II. There, the distinct
entailment possibilities of strong and weak adjuncts were
shown to be the result of a difference in scope relative to
an accompanying modal; weak adjuncts with narrow scope were
shown to have precisely the function of if-clauses. Here,
the observed difference between strong and weak adjuncts is
evidently linked to the presence of a frequency adverb, as
is the fact that in each of the ~ sentences, the adjunct
intuitively plays the role of a time adverbial. Elucidating
the nature of this difference therefore requires a good
understanding of the semantics of frequency adverbs; this is
discussed in the following subsection.
3.1. THE SEMANTICS OF RELATIVE FREQUENCY ADVERBS
In English, there are two varieties of frequency

adverbs. 26 One variety, consisting of what I shall call
fixed frequency adverbs, is exemplified in (245); the other,
consisting of relative frequency adverbs, is exemplified in
(246). When applied to a sentence S, both sorts of adverbs
can be understood to specify the frequency with which S is
true during a single, continuous time interval; thus, both
(247) and (248) are understood to mean that John is sick was
174 CHAPTER III
(245) Fixed frequency adverbs
yearly, every year, once a year, annually;

monthly; weekly; daily; hourly
(246) Relative frequency adverbs
always; usually; often, frequently; sometimes,

periodically; occasionally, sporadically,
infrequently, every so often, now and then, now
and again, Once in a while, from time to time;
rarely; never
(247) John was sick every month in 1980.

(248) John was often sick in 1980.
true with a certain frequency during 1980. Relative

frequency adverbs, however, can also be understood to
specify the frequency with which intervals at which S is
true appear in a sequence of discontinuous intervals. For
example, sentence (249) is not simply understood to mean
that something which happens often is that Jane uses a
calculator to figure her taxes;
(249) Jane often uses a calculator to figure her taxes.
(249) would in fact be consistent with Jane's using a

calculator only very infrequently, if she happened to figure
her taxes only infrequently (say, once a year). (249) is
instead felt to mean that intervals at which Jane uses a
calculator appear often in the sequence of intervals at
which she figures her taxes.
This way of interpreting (249) brings out the essential
difference between fixed and relative frequency adverbs.
Fixed frequency adverbs specify frequencies whose periods
are of a fixed length; for example, yearly specifies a
frequency the length of whose periods is a year. Relative
frequency adverbs, on the other hand, specify frequencies
the average length of whose periods is proportional to the
length of some other interval (e.g. the interval determined
by 1980 in (248» or to the 'length' (i.e. number of
members) of some sequence of discontinuous intervals (e.g.
the sequence of intervals at which Jane figures her taxes in
(249». Thus, the interpretation of a relative frequency
adverb is always relative to some interval or sequence of
intervals; or, since an interval can be regarded as a

sequence of continuous smaller intervals, relative frequency
adverbs can, in fact, always be viewed as being interpreted
relative to some sequence of intervals. In sentences like
(249), the relevant sequence is implicit. It can, however,
be explicitly designated by a temporal adverbial; in (250),
for example, the adverbial when she's figuring her taxes
picks out the sequence of intervals relative to which often
is interpreted.
(250) When she's figuring her taxes, Jane often uses a

calculator.
It is most convenient simply to regard the temporal

adverbial in a sentence like (250) as a set-level adverb
(i.e. of category TA) rather than as a main tense adverb;
such an adverbial will have an unordered set of intervals
rather than a sequence of intervals as its denotation, but
the inherent linear ordering of intervals can be used to
impose the ordering necessary for the interpretation of the
frequency adverb.
Adverbs like often are regarded as basic expressions of
the category RFA of relative frequency adverbs:
RFA the basic category of always, often,

relative frequency frequently, sometimes,
adverbs occasionally, never
To formalize the relativity of these adverbs, 1 regard them

as denoting two-place relations rather than functions; the
extra argument position is to be filled by the (often
contextually implicit) set of time intervals which
determines the sequence of intervals relative to which the
adverb is interpreted. 27
On first consideration, i t might seem that a relative
frequency adverb a should translate as an expression a' of
type «i,t>,«s,t>,t». Then one could say that, where 8'
denotes a set of time intervals and ~, denotes a
proposition, a'( 8')( CP') is true iff intervals at which the
proposition denoted by CP' has truth as its extension appear
with the required frequency in the sequence of intervals
determined by the denotation of 8' (and .~). Under this
approach to the interpretation of relative frequency
176 CHAPTER III
adverbs, sentence (250) would, for example, translate as

(251) :
(251) often'( AtVt 1 [t = t1 & [NONPAST(t 1) &

AT(t 1, Jane-is-figuring-ber-taxes')]])
(AJane-uses-a-calculator')
(251) would be true iff intervals at which the proposition

denoted by AJane-uses-a-calculator' has truth as its
extension show up with the right frequency in the sequence
of intervals determined by the denotation of (252).
(252) Atvt,[t = t, & [NONPAST(t,) &

AT(t" Jane-is-figuring-her-taxes')]]
This approach seems plausible enough until nonpresent

tense sentences such as (253) are considered.
(253) When she figured her taxes, Jane often used a

calculator.
In the analysis of this sentence, tense cannot be regarded

as having narrower scope than often; if it were, (253) would
translate as (254), which plainly misrepresents the
truthconditions of (253). (254) could, for example, be true
even if Jane never used a calculator when figuring her
taxes, since the second argument of often' in (254) denotes
an eternal proposition.
(254) of'ten'(Atvt, [t = t, & vt 2 [t 2 = t, & [past(t 2 ) &

AT(t 2 , Jane-figures-ber-taxes,)]]])(Avt[past(t) &
AT(t, Jane-uses-a-calculator')])
On the other hand, tense cannot be regarded as having

wider scope than often in sentences like (253). To see why,
consider sentence (255).
(255) When it rained during the past summer, our roof

often leaked.
In this sentence, the frequency adverb often is restricted

by a wben-clause which itself contains a main tense adverb,
namely during tbe past su.mer; thus, if tense had wider
scope than often in this sentence, it would necessarily also
contain within its scope the when-clause and its main tense
adverb, as in the translation (256).
(256) Vt[past(t) & AT(t, often'o.t, Vt 2 [t, = t2 &

vt 3 [t 3 = t2 & [during-the-past-su••er'(t 3 ) &
[past(t 3 ) &
AT(t 3 , it-rains')]]] ])("'our-roof-leaks'»]
As was seen in section '.3.' of this chapter, the denotation

of the adverb during the past SUmmer shifts from one index
to another-recall the proposed translation (257), in which
the semantic properties of 'now' guarantee a shifting
interpretation:
(257) Atvt,[[[su.mer'(t,) & t, < now] &
At 2 [[summer'(t 2 ) & t2 < nov] .... [t 2 < t, v

t2 = t,]]] &t ~ t,J
Accordingly, if tense had wide scope in (255), this adverb

would be interpreted relative to some past interval rather
than to the speaker's interval; that is, (255) would mean
that there is some past interval i such that our roof often
leaked during the summer preceding i, where i could in
principle be indefini tely remote (e.g. twenty years past).
It is clear, however, that (255) lacks this interpretation:
(255) is true when uttered at interval i' only if our roof
often leaked during the summer immediately preceding i'.
These problems can be avoided under a slightly
different analysis, which I shall adopt here. In this
analysis, relative frequency adverbs are taken to denote
binary relations between sets of time intervals (i.e. of
type «i,t>,«i,t>,t»). Where a' is the translation of
such an adverb and 8', y' denote sets of time intervals,
a'(a')(y') is true iff it is true that members of the
denotation of y' appear with the required frequency in the
sequence of intervals determined by the denotation of a'.
(The 'required frequency' associated with a given relative
frequency adverb is sometimes quite straightforward. For
example, always'(a')(y') is true iff all members of the
denotation of 8' are also members of the denotation of y';
never'(a')( y') is true iff none of the members of the
denotation of 8' are also members of ·the denotation of y'.
178 CHAPTER III
i"lore often than not, however, the 'required frequency'

associated with a given frequency adverb is vague and
context-dependent;2H as I have shown elsewhere (Stump
'981a), it's possible to formulate truthconditions for such
adverbs which account for this indeterminacy. I won't
propose complete truthconditions here, however; all that is
necessary for the present discussion is that for any
relati ve frequency adverb a, there is some range R of
frequencies such that the truth of a'(B')(Y') depends on the
existence of a frequency f within range R with which members
of the denotation of y' appear in the sequence of intervals
determined by the denotation of B,.29)
In this analysis, relative frequency adverbs are
assumed to join with temporal abstracts rather than full
sentences. Thus, two rules are employed to introduce
relative frequency adverbs into larger expressions: one
joins a frequency adverb a with a set-level adverb B and a
temporal abstract to produce a new temporal abstract--in
this case, B explicitly restricts the interpretation of a; a
second rule simply joins a frequency adverb a with a
temporal abstract to produce a new temporal abstract--a is,
in this instance, only implicitly restricted. The two rules
are the following.
(258) 364. If a e: PRFA , B e: PTA' and y e: PTAB , [6Tense,

-Adjunct, ~Perfect], then F64a(a, B, y),
F 64b (a, B,y) e: P TAB , [aTense, -Adjunct,
~Perfect], where F 64a (a, a, y) is
CONWIB(B, y'), F 64b (a,B,Y) is CON(y', B), and
y' is ROS(a,y).
Example: F 64a (never, after heo mowed the lawn, (he o )

[trimmed] the hedges) = after heO mowed the
lawn, (heO) never [trimmed] the hedges
T64. If a e: P RFA , B e: PTA' y e: PTAB , and a,B,

y translate as a',B', y', then F 64a (a,B,Y),
F 64b(a,B, y) translate as
At[a'(B')(At,[t 1 ~ t & y'(t,)])].
TENSE AND fREE ADJUNCTS '79
(258) [continued]
Example: after heO mowed the lawn, (heo) never

[trimmed] the hedges translates as
H[never'(H,Vt 2 [[t 2 < t, & M(t" t 2 )] &
vt 3 [t 3 = t2 & [past(t 3 ) &
AT(t 3 , heO-mows-the-lawn')]]])( At, [t, =- t &
[past(t,) & AT(t" heo-trims-the-hedges')]])]
S65. If a e: P RFA and B e: PTAB , [YTense, -Adjunct,

'Perfect], then F 65 ,n(a,B) e: PTAB , [yTense,
-Adjunct, 'Perfect], where F 65 ,n(a,B) is
ROS(a,B).
Example: F65 ,6(occasionally, (a sailor) [strolled] by)

= (a sailor) occasionally [strolled] by
T65. If a e: PRFA , B E PTAB , and a,B translate as
a',B', then F 65 ,n(a,B) translates as
At[a'CIn)(At,[t, =- t & B'(t,)])].
Example: F65 ,6 (occasionally, (a sailor) [strolled] by)

translates as
H[occasionally'(I 6 )(At,[t, c t & [past(t,) &
AT(t" a-sailor-strolls-by')]])]
Under this approach, rule (258) can be used to analyze

sentence (250) as in (26') (see next page), which induces
the translation (260).
(260) Vt[often'(At,Vt 2 [t, = t2 &

vt3 [t 3 = t2 & [PRES(t 3) &
AT(t 3 , Jane-is-figuring-her-taxes')]] ])(At, [t, =- t
& [PREset,) & AT(t" Jane-uses-a-calculator')]])]
'80 CHAPTER III
---------
(261) when she is figuring her taxes, (Jane) often
[uses] a calculator, t, 55,'
Jane, T when she, is figuring her, taxes, (she,) often

[uses] a calculator, t, 60
I
when she, is figuring her, taxes, (she,) often [uses]
a calculator, TAB, 64a
oft~-lrn-s-h-e-,-i-S-f-i-gU-r-l.-·n-g-- (she,) [uses]
RFA her, taxes, TA, 67,0 a calculator, TAB, 58
~ I
when, (she,) [is] figuring her, (she,) [uses] a
TSC taxes at-that-timeo' t, 60 calculator, TAB, 56
I I
(she,) [is] figuring her, taxes (she,) [uses] a
at-that-tiaeo' TAB, 61b calculator. t. 19
~
at-that-tiaeo' (she,) [is] figuring her, taxes. TAB. 58
MTA I
(she 1 ) [is] figuring her, taxes, TAB, 56
I
(she,) [is] figuring her, taxes, t, 19
Similarly, the past tense sentence (253) is analyzed as in

(263) and translates as (262).
(262) Vt[often' (At, Vt 2 [t, = t2 & Vt3 [t 3 = t2 & [past(t 3)

& AT(t 3 , Jane-figures-her-taxes')]]])(At,[t, ~ t &
[past(t,) & AT(t" Jane-uses-a-calculator')]])]
(262) induces the proper interpretation for (253); note,

specifically, that (262), unlike (254), requires that past
intervals at which Jane figured her taxes coincide
frequently with past intervals at which she used a
calculator.
TENSE AND FREE ADJUNCTS '8'
(263) wben abe Cigured ber taxea, (Jane) octen [uaed]
a calculator, t, 55.'
------------
Jane, T wben sbe, Cigured ber, taxes, (sbe,) octen
[used] a calculator, t, 60
I
wben sbe, Cigured ber, taxes, (abe,) octen [used]
a calculator, TAB, 64a
---\\------------
often, when she 1 Cigured (she1) [used]
RFA her, taxes, TA, 67,0 a calculator, TAB, 57
~ I
when, (she,) [figured] her, (she ) [uses]
TSC taxes at-that-timeO' t, 60 a calculator, TAB, 56
I I
(she,) [figured] her, taxes (she1 ) [uses]
at-that-timeO' TAB, 6,b a calculator, t. 19
~
at-that-ti.eo' (she,) [figured] her 1 taxes, TAB, 57
MTA I
[ figures] her1 taxes. TAB. 56
I
(she1) [figures] her 1 taxes, t. '9
Sentence (255) is analyzed as in (265) and translates
as in (264).
(264) Vt[often'(At,vt 2 [t, = t2 &

vt 3 [during-the-paat-summer'(t 3 ) & [t 3 = t2 &
[paat(t 3 ) & AT(t 3 • it-rains')]]]])(At,[t, c t &
[past(t,) & AT(t,. our-rouf-leaks')]])]
(264) represents the desired truthcondi tions for (255); in
particular, the shifting main tense adverb during the past
suamer is interpreted relative to the speaker's interval
rather than relative to some past interval.
182 CHAPTER III
(265) wben it rained during tbe past su.aer, (our roof)
often [leaked], t, 60
I
wben it rained during tbe past su..er, (our roof) oftea
[leaked], TAB, 64a
often, when it rained during the (our roof) [leaked],
-----------
RFA paat su.mer, TA, 67,0 TAB, 57
I
when, (it) [rained] at-tbat-tt.eO (our roof) [leaks],
TSC during the past summer, t, 60 TAB, 56
I I
(it) [rained] at-that-tt.eO (our roof) [leaks], t, 19
during tbe past su.mer, TAB, 61b
I ------------
during tbe past summer,
MTA, 63
(it) [rained] at-that-tt.eO'
~ 61b
at-that-tt.eo'
during the past su.aer,
TA MTA I
(it) [rained],
TAB, 57
(it) [rains], TAB, 56
I
(it) [rains], t, 19
Sentence (266), in which the adverb often isn't
explicitly restricted by a set-level time adverb, is derived
by means of rule (259) as in (267), and is thus assigned the
translation (268), in which 16 is a variable over sets of
time intervals (whose value is inferred).
(266) Jane often uses a calculator.
(267) (Jane) often [uses] a calculator, t, 60
----------
I
(Jane) often [uses] a calculator, TAB, 65,6
often, RFA (Jane) [uses] a calculator, TAB, 58

I
(Jane) [uses] a calculator, TAB, 56
I
(Jane) [uses] a calculator, t, 19
(268) Vt[often'(I 6)( At1 [t 1 ~ t IX [PRESet,) &

AT(t 1, Jane-uses-a-calculator')]])]
The reader may have wondered why the expressions

produced by rules (258) and (259) must be temporal abstracts
rather than sentences. The reason is that main tense
adverbs may combine with them. Consider, for example,
sentence (2.59).
(269) When he was in Kansas City, John always went for a

walk after he ate supper.
Suppose that John had taken several vacations in Kansas

City, and followed the routine of taking a walk after supper
only during his most recent stay. Under these
circumstances, (269) could apparently be true; that is, the
when-clause in (269) can be understood to relate to a single
interval at which John was in Kansas City, and not
necessarily to all such intervals. On such an
interpretation, the when-clause cannot be within the scope
of the frequency adverb always; if it were, it would relate
to all intervals at which John was in Kansas City, no matter
whether it served as the first argument of always or as a
main tense adverb modifying the second argument of always.
Accordingly, in the derivation of (269), the when-clause
must be able to be 'added on' after the introduction of the
frequency adverb always; but since main tense adverbs join
with temporal abstracts rather than sentences, this means
that the rules (258) and (259) introducing relative
frequency adverbs must themselves produce temporal abstracts
rather than sentences. Since rules (258) and (259) are so
formulated, the desired interpretation for sentence (269) is
easily induced under the proposed analysis: if (269) is
derived as in (271), it is assigned the interpretation of
(270).
(270) Vt[vt 1 [t ~ t1 & Vt 2 [t 2 ~ t1 & [past(t 2 ) &
AT(t 2 , Jobn-is-in-Kansas-City')]]] &

always'(>'t 1Vt 2 [[t 2 < t1 & M(t" t 2 )] &
vt 3 [t 3 = t2 & [past(t 3 ) &
AT(t 3 , John-eats-supper')]] ])()ct 1 [t 1 c t &
[past( t 1) & AT( t l' Jobn-goes-for-a-walk')]]) ]
184 CHAPTER III
(271 ) wben be was in Kansas City, (John) always [went]
---------
for a walk at"ter be ate supper, t, 55,0
John, T wben bee> was in )[ansas City, (bee» always [went]

for a walk after heo ate supper, t, 60
I
vhen heo was in Kansas City, (beo) alvays [went] for a walk
----------
after heO ate supper, TAB, 61a
wben ben was in (hea) always [went] for a walk
Kansas City, MTA, 63 after heO ate supper, TAB, 64b
~_---;7r- ____
alwayS;
RFA
after beO ate supper,
TA, 67,0 I
(heO) [went] for a walk,
TAB, 57
(heo) [goes] for a walk, TAB, 56
I
(he o) [goes] for a walk, t, 19
The feasibility of this analysis of relative frequency
adverbs provides further justification for the category TAB
of temporal abstracts; moreover, temporal abstracts will be
indispensably employed in the analysis of a certain variety
of generic sentence in section 4.
The foregoing analysis of the interpretation of
relati ve frequency adverbs affords a very clear
understanding of the difference between strong and weak
adjuncts observed in sentences (241 )-(244), as will be seen
in the following subsection.
3.2. ADJUNCTS RESTRICTING THE INTERPRETATION OF RELATIVE

FREQUENCY ADVERBS
In the preceding subsection, relative frequency adverbs were
analyzed as denoting binary relations between sets of time
intervals; this analysis permitted an explicit account of
the way in which a time adverb--specifically, an adverb of
category TA--may restrict the interpretation of a relative
frequency adverb. In the light of this analysis, the
problem of explaining the observed differences in sentences
(241 )-(244) can be considered again.
(241) a. Being a sailor, John sometimes smokes a pipe.


pipe.
(242) a. Weighing four tons, our truck often makes the
bridge shake.
b. Carrying a load of over 1500 Ibs., our truck
often makes the bridge shake.
(243) a. Having only ten minutes between her morning
lunch.
h. Studying for finals, Anne never eats lunch.
calculator.
Consider sentence (241b). In this sentence, the weak

adjunct lying on the beach is felt to supply the temporal
restriction on the frequency adverb sometiaes--it specifies
exactly which intervals are the ones at which it is
sometimes true that John smokes a pipe; it thus appears to
serve exactly as a time adverbial would--as the first
argument of the frequency adverb. If this is so. then weak
adjuncts must be able to belong to the category TA of
expressions denoting sets of time intervals. The following
rule converting A-abstracts in the subclass [+Weak] to
members of TA might therefore be postulated. 30
(272) S9. If a e: PTAB , [+Adjunct, +Weak, -Perfect],

then F9,n(a) e: PTA' [+Adjunct], where F9 ,n(a)
is a.
Example: F9 ,o(lying on the beach) = lying on the beach

As a first approximation for the corresponding translation
rule, (273) might seem promising.
(273) T9. (tentative)
If a e: PTAB and a translates as 01', then

F9.n(a) translates as A tvt, [t = t1 & a'(t,)].
According to (273), the free adjunct in (241b) designates
the set of intervals at which John is lying on the beach.
This seems reasonable for (24' b), but I believe that
'86 CHAPTER III
identity may in general be too restrictive a relation

between the intervals denoted by t and t, in (273): in
(274), for example, the adjunct seems to designate some set
of intervals each of which is preceded by an interval at
which Leroy gets down from his highchair.
(274) Getting down from his highchair, Leroy sometimes

walks all the way to the living room.
This suggests that the temporal relation between the

intervals denoted by t and t, in (273) is semantically
indeterminate--that it is inferred, just like the logical
relation between the adjunct and the superordinate clause in
a sentence like (275).
(275) Lying on the beach, John smoked his pipe.
Recall that on the nongeneric, 'event' reading of a

sentence like (275), the adjunct serves as a main tense
adverb: thus, according to the analysis proposed in section
2.3, (275) translates as (276), in which the variables L6
and Mg represent semantic indeterminacies which can only be
resolved by inference.
(276) Vt[K(L 6 )(A Vt ,[M g (t,t,) & AT(t" Vxs[R(x s , x~) &
lying-on-tbe-beacb'(x s ) ]» ])(A[past(t) &
AT ( t , J obn-sllOkes-his-pipe') ] ) ]
If the relation between the intervals denoted by t and t, in

(273) is likewise viewed as semantically indeterminate, then
the '=' sign in (273) must be replaced with a free· variable
Mn over binary relations between time intervals (i.e. of
type <i,<i,t»), as in (277).
(277) T9. (final)
If a £PTAB and a translates as aI, then

Fg,n(a) translates as
Atvt, [Mn(t,t,) & a'(t,)].
(277) [continued]
Example: F9 ,o(lying on the beach) translates. as

AtVt 1 [Mo(t,t,) & AT(t" Vx s [R(x s • x41) &
lying-on-the-beach'(x s )])]
If rules (272) and (277) are assumed, then sentence (241b)

can be derived as in (278) and thus assigned the translation
(279).
(278) lying on the beach, (John) sollletimes [saokes]

a pipe, t, 60
I
lying on the beach, (John) sometimes [s.okes]
a pipe, TAB, 64a
sometiae~ing on the beach,

I
(John) [saokes] a pipe,
RFA TA, 9,9 I TAB, 58
lying on the beach, (John) [smokes] a pipe, TAB, 56

TAB, 1,4 I
(John) [s.akes] a pipe, t, 19
(279) vt[sometimes'(At, V~2[M9(t" t 2 ) &

AT(t 2 , Vx s [R(x s , x4)
1
&
s
lying-oD-the-beach'(x )])])(At 1[t 1 c t &
[PRES(t 1) & AT(t 1, John-saokes-a-pipe')]])]
(279) captures the intuition that the role of the weak

adjunct in (241b) is analogous to that of the adverbial
clause in (280).
(280) While he is lying on the beach, John sometimes

smokes a pipe.
Now consider sentence (241a). In this sentence, the

strong adjunct is not felt to provide a temporal restriction
on sometimes--(241a) doesn't mean that it is sometimes true
at intervals at which John is a sailor that he smokes a
pipe. Instead, the restriction on sometimes is determined
contextually, and the free adjunct bears some relation to
the entire sentence which follows, its intuitive function
being that of an adverb of causation/explanation. ThUS,
188 CHAPTER III
(241a) might simply translate as (281), where 13 is again a

variable over sets of time intervals.
(281) Vt[K(L 6 )(A vt 1 [M 9 (t,t 1 ) & AT(t 1, being-a-

sailor'(x~»])("[someti.es'(I3)(Atl [t 1 =- t &
[PRES(t 1 ) & AT(t 1 , Jobn-smokes-a-pipe')]])])]
This translation implies a syntactic analysis like (282), in

which the strong adjunct serves as a main tense adverb.
(282) being a sailor, (John) someti.es [smokes] a pipe,

I
-------
t, 60
being a sailor, (John) sometimes [smokes] a pipe, TAB, 61a
being a sailor, (John) someti.es [smokes] a pipe,

MTA, 111,6,9 ~ TAB, 65,3
being a sailor, sometiaes, (John) [s.okes] a pipe,
TAB, 2,4 RFA I TAB, 58
(John) [s.okes] a pipe, TAB, 56
1
(John) [s.okes] a pipe, t, 19
Thus, what distinguishes the a and b sentences of (241)

is that the strong adjunct being a-sailor in (241a) does not
serve to restrict the interpretation of the frequency adverb
sometimes, while the weak adjunct lying on the beach in
(241b) does restrict the interpretation of this adverb.
Moreover, careful consideration reveals that the adjuncts in
the ~ and ~ sentences of (242)-(244) differ analogously.
Evidently, the differences between strong and weak adjuncts
in the context of a relative frequency adverb are akin to
the differences observed in modal contexts in Chapter II.
strong adjuncts uniformly fail to be interpretable as the
first argument of a modal or relative frequency adverb,
serving instead as main tense adverbs. Weak adjuncts, on
the other hand, may restrict the interpretation of both
modals and relative frequency adverbs; when they do, they
are expressions of categories t' and TA, respectively. When
a weak adjunct belonging to one of these categories does
restrict the interpretation of a binary operator (of
category MOD or RFA) in the main clause, very little needs
to be inferred about the logical role of the adjunct; the

semantic properties of the main clause operator determine
this role at least partially and perhaps totally in some
cases. 31
Let me emphasize that weak adjuncts don't have to
restrict the interpretation of a main clause frequency
adverb if one appears)2 (283), for example, could
conceivably be understood as a narrative sentence de~cribing
something which John often did during a single occasion of
lying on the beach.
(283) Lying on the beach, John often thinks of home.
In such read ings of sentences li.ke (283), the adj unct does
not restrict the interpretation of the frequency adverb,
even though it is weak. The important generalization is
that a weak adjunct ~ restrict a main clause frequency
adverb, while a strong adjunct may not.
Nonpresent tense sentences having frequency adverbs
restricted by weak adjuncts are straightforwardly derived in
the proposed system. Sentence (284), for example, is
assigned the analysis in (285) and the translation (286).
(284) Carrying over 1500 Ibs., our truck often made the
bridge shake.
(285) carrying over 1500 Ibs., (our truck) often [made]

the bridge shake, t, 60
I
carrying over 1500 Ibs., (our truck) often [made]
often,
______,T----------_
the bridge shake, TAB, 64a
carrying over 1500 Ibs., (our truck) [made] the

RFA I TA, 9,9 bridgj shake, TAB, 57
carrying over 1500 Ibs., (our truck) [makes] the bridge

TAB, ',4 shake, TAB, 56
I
(our truck) [makes] the bridge shake, t, '9
s s i
(286) Vt[often'( At, Vt 2 [M 9 Ct"t 2 ) & ATCt 2 , Vx [R(x ,x4) &
carrying-over-1500-1bs.'(x s )]) ])(At, [t, =. t &
[past(t,) &
ATCt" our-truck-makes-the-bridge-shake')]])]
190 CHAPTER III
(286) represents the desired interpretation for (284).33

Similarly, the future tense sentence (287) may be derived as
in (288) and translated as in (289).
(287) Hearing that song, John always will think of Mary.
(288) bearing that song, (John) always [will] think of

Mary, t, 60
I
hearing that song, (John) always [will] think of
Mary, TAB, 64a
-------hear~ng
\. tha t song,
I
a 1 ways, (John) [will] think of
RFA TA, 9,9 Mary, TAB, 59
I
hearing that song, (John) [thinks] of Mary, TAB, 56
TAB, 1,4 I
(John) [thinks] of Mary, t, 19
(289) vt[ al ways'(A t1 vt 2 [M 9 (t,.t 2 ) & AT(t 2 , VXs[R(X s , X4i)

& hearing-that-song l (x s )])])(At 1 [t 1 ~ t &
[FUT(t,) & AT(t" John-thinks-of-Maryl)]])]
In this section, it has been shown that the distinction

between strong and weak adjuncts, motivated in two
independent ways in Chapter II, is further supported by the
semantics of free adjuncts in the context of adverbs of
relati ve frequency. As will be seen in the following
section, there are even further grounds for distinguishing
the same two varieties of free adjuncts.
4. A Generalization Operator
So far, two environments have been examined which motivate a
distinction between two varieties of free adjuncts: modal
sentences and sentences with relative frequency adverbs. In
this section, a third environment motivating a distinction
between strong and weak adjuncts is discussed. This third
environment is exemplified in sentences (290)-(293).
(290) Lying on the beach, John smokes cigars.

(291) Drunk, he drives very dangerously.
(292) In first gear, the truck makes funny noises.
(293) Wearing her new outfit, Mary looks fat.
Each of these sentences can be understood as expressing a

generalization about intervals at which its (weak) adjunct
is true. But if strong adjuncts are substituted for the
weak adjuncts in (290)-(293), this kind of interpretation is
not found; none of (294)-(297) can be understood as
expressing a generalization about intervals at which its
adjunct is true.
(294) Being a businessman, John smokes cigars.

(295 ) Being a drunk, he drives very dangerously.
(296) Having a rather decrepit engine, the truck makes
funny noises.
(297 ) Weighing over 200 lbs., Mary looks fat.
Above, it was shown that the distinction between strong and

weak adjuncts in modal contexts is a function of the fact
that weak adjuncts, but not strong, can serve as the first
argument of a modal; similarly, it was seen that weak
adjuncts can function as the first argument of a relative
frequency adverb, while strong adjuncts cannot. In
sentences (290)-(297), however, there is no explicit element
(akin to a modal or relati ve frequency adverb) relative to
which the observed difference in interpretation between
strong and weak adjuncts might be explained. In the
following subsections, I shall nevertheless provide
independent evidence of the existence of a phonologically
null 'generalization operator'; I shall argue that weak
adjuncts like those in sentences (290)-(293) can serve to
restrict the interpretation of this operator, while strong
adjuncts like those in (294)-(297) cannot, and thus that the
weak adjuncts in sentences (290)-(293) playa role
essentially analogous to those played by the adjuncts in
(298) and (299).
(298) Lying on the beach, John might smoke a pipe.

(299) Lying on the beach, John sometimes smokes a pipe.
192 CHAPTER III
4.1. TWO SORTS OF INTERPRETATIONS FOR TEMPORAL ADVERBIAL

CLAUSES
The distinctiveness of the class of sentences represented by

(290)-(293) can be appreciated by considering two different
ways in which certain temporal adverbial clauses can be
interpreted. In many cases, temporal adverbials can be
understood either to pick out a single interval of time or
to pick out indefinitely many distinct intervals. For
example, the adverbial clause when he was drunk in (300) can
be felt to relate to a particular past interval (perhaps the
only one) at which Bill was drunk;
(300) When he was drunk, Bill drove very dangerously.
on the other hand, it can also pick out an indefinite number

of such intervals--on this interpretation, (300) is roughly
equivalent to (301) and (302).
(30n Whenever he was drunk, Bill drove very

dangerously.
(302) When he was drunk, Bill always drove very
dangerously.
This pair of interpretations is by no means restricted to

when-clauses; a wide variety of temporal adverbial clauses
show a precisely analogous pair of usages, as the ambiguity
of each of (303)-(305) suggests.
(303) After he finished dinner, he smoked a cigar.

(304) He took aspirin before going to the dentist.
(305) She sat reading while she waited for her mother.
When a temporal adverbial clause is itself present in tense,

the 'multiple-interval' intepretation is highly preferred;
the adverbial clauses in (306)-(309), for example, do not
simply pick out a single occasion of lying on the beach,
being drunk, etc., except under the rather unlikely
narrative present interpretation of these sentences.
(306 ) While he's lying on the beach, John smokes cigars.

(307) When he's drunk, Bill drives very dangerously.
(308) When it is in first gear, the truck makes funny
noises.
(309) When she's wearing her new outfit, Mary looks fat.
The relevant generalization about sentences like (290)-
(293) is that they are understood to express a
generalization about intervals at which their adjunct is
true precisely when the adjunct funotions as a 'mul tiple-
interval' adverbial clause. Thus, just as (310) is
ambiguous between a 'single-interval' and a 'multiple-
interval' interpretation, so is (311);
(310) While he lay on the beach, John smoked cigars.

(311) Lying on the beach, John smoked cigars.
and it is just the possibility of a 'multiple-interval'
interpretation that distinguishes the weak adjuncts in
(290)-(293) from the strong adjuncts in (294}-(297).
How exactly are the two interpretations of temporal
adverbial clauses to be accounted for? One could, of
course, postulate pairs of temporal subordinating
conjunctions--when 1 , when 2 ; after1' after2; etc.; the
members of each pair would De identioal in interpretation
except that one member would give rise to a 'multiple-
interval' interpretation while the other would not. Such an
approach could, I believe, be made to account for the facts;
but it would imply--wrongly, I think--that the fact that all
members of a well-defined class of adverbial clauses have
both a 'multiple-interval' and a 'single-interval'
interpretation is merely accidental. It would therefore be
preferable to derive one interpretation from the other, or
to derive both interpretations from some common element of
meaning. In the following subsection, I shall develop the
means for doing just this.
4.2. GENERALIZATION OPERATORS

In this subsection, I shall argue that sentences like (290)-
(293) have an implicit 'adverb of generalization', and that
their (weak) adjuncts restrict its interpretation in
precisely the manner in which weak adjuncts may restrict the
interpretation of relative frequency adverbs. A brief
excursus on Carlson's operators G and G' will provide the
necessary background for discussion of this covert adverb.
In his dissertation, Carlson (1977) observes a
superficially unexpected asymmetry in the verb system of
English: although certain predicates (aliye, ayailable,
drunk) are always interpreted as stage-predicates (they
194 CHAPTER III
uniformly induce the 'existential reading' of a bare plural)

and certain others (widespread, rare, extinct) as predicates
of kinds, there are seemingly no predicates applying
uniquely to objects; in general, if a predicate can be used
of an object, it can be used of a kind as well. Carlson
presents extensive arguments that this state of affairs is a
consequence of the fact that, given any predicate known to
hold of some object, this predicate can be 'generalized' to
apply to the corresponding kind--that given the observation
that Fido has four legs, and Rover has four legs, and spot
has four legs, and Rex has four legs, one can generalize and
say that dogs have four legs. Carlson represents this
generalization of a predicate holding basically of objects
to the generic level as a generalization operator G' taking
object-level verb phrase intensions into generic verb phrase
extensions. The inductive generalization just mentioned
might therefore be represented as in (312):
(312) have-four-legs'(Fido')
have-four-legs'(Rover')
have-four-legs'(Spot')
have-four-legs'(Rex')
:. G' (Ahave-four-legs' ) (dogs' )
In order for G'(P) to hold for some kind xk, it need not
always be the case that all objects realizing xk have
property P; Dogs have four legs, for example, is not
falsified by the existence of some three-legged dog. But
G'(P) 'cannot be true of xk unless enough of the objects
realizing xk have P. So how many are 'enough'? Carlson
demonstrates at length that 'statements of frequency or of
absolute number of occurrences cannot be stated generally
for all generic sentences' (1979:59): while Lions are
.ammals applies to all lions, Lions give milk to their young
applies only to mother lions; Lions attain a weight of
several hundred pounds applies only to lions which reach
physical maturity; Lions attack zookeepers applies only to
lions in captivity; and so on. Ultimately, the question of
when a property ceases to be an accidental property of
certain objects and becomes an essential property of the
kind they realize is epistemological in nature, and thus one
which the grammar of English should not be expected to
answer.
Carlson also finds that any predicate (headed by a
verb) interpretable as a stage-predicate may also be
interpreted as a predicate of objects and kinds. He
suggests that this too is the result of generalizing the
application of predicates from 'lower' to 'higher' entities;
observing several stages of Fido chasing cars, one can
generalize and say that Fido (the object) chases cars--or
observing stages of several dogs engaged in the same
pursui t, one can say that dogs (the kind) chase cars.
Accordingly, a second generalization operator G taking
intensions of stage-predicates into extensions of
individual-level predicates is posited:
R(sO' Fido') & chase-cars'(so)

R(s1' Fido') & chase-cars'(s1)
... G(Achase-cars') (Fido')
In the present fragment, the operator G is introduced

by the following rule:
(314) S22. If a E Prvs, then F22 (a) E Prvi, where F22 (a)
is a.
T22. If a e: Prvs and a translates as a', then

F22 (a) translates as G("a').
As Carlson points out (p.280), 'the two G functions

together in a sense serve the inverse purpose of the
realization relation... Realization makes "available"
entities of a lower level from those of a higher level;
generalization makes available predicates from a lower level
to those of a higher level.'
r wish to suggest that a generalization operator in
some ways like those posited by Carlson is responsible for
the 'multiple-interval' interpretation of the time
adverbials in sentences like (303)-(309) and of the weak
adjuncts in sentences like (290)-(293). First, however, I
shall briefly demonstrate why the operator in question
cannot be identified with either of those which Carlson has
postulated.
196 CHAPTER III
Carlson discusses (pp.418ff) a type of sentence in

which a temporal adverbial clause seems to be within the
scope of a generalization operator. For instance, (315) can
be understood to mean, roughly, that any dog barks when it
sees a mailman--that the kind designated by dogs has the
generalized property of any object that barks when it sees a
mailman, as in (316).
<315 ) Dogs bark when they see a mailman.
<316 )
Carlson doesn't address the matter, but it seems intuitively

plausible to suppose that in the interpretation of
'habitual' sentences like (317), the adverbial clause is
again within the scope of a generalization operator;
Fido barks when he sees a mailman.
here, however, it would be the stage-to-individual level

operator G, as in (318).
(318 )
In accordance with this analysis, additional internal

structure could be supplied for (316):
(319) G,(AAXO(G(AAXS(xs_barks-wben_xs_sees-a-
mailman'])(xo)])(dogs')
Such an analysis would apparently afford a very nice account

of the fact that temporal adverbial clauses have both
'single-interval' and 'multiple-interval' interpretations:
it would allow one to say that the 'single-interval'
interpretation is basic, and that the 'multiple-interval'
interpretation arises when the adverbial is within the scope
of the generalization operator G; that this operator, not
the adverbial clause itself, is the source of the
multiplicity. Thus, suppose that (320) has the internal
structure (321).
(320)
(321) vt[Vt,[t = t, & vt 2 [t 2 = t, & [PRES(t 2 ) &
AT(t 2, see-a-mailman'(x s »]]] & [PRESet) &
AT(t, bark'(x s »]]
Here, the when-clause is interpreted as picking out a single
interval. But when (32') occurs within the scope of a
generalization operator, as in (318) and (319), it picks out
a number of distinct intervals: for any object 0, the truth
of (322) implies that there are a certain number of distinct
stages satisfying (323);
(322) G(AAX SVt[Vt 1[t = t1 Vt 2 [t 2 = t1 & [PRES(t 2 )

& &
AT(t 2 , see-a-mailman'(x s »]]] & [PREset} &
AT(t, bark'(xs»]])(o)
(323) [R(xs,o) & vt[vt 1 [t = t1 & Vt 2 [t 2 = t1 &

[PRES(t2 ) & AT(t 2 , see-a-llailman'(x s »]]] &
[PREset) & AT(t, bark'(x s »]]]
and given the ontological status of stages--spatiotemporal

'slices' of an individual--this implies that there are a
certain number of distinct intervals satisfying (324).
(324) Vxs[R(xS,o) & [vt,[t = t, & Vt 2 [t 2 = t1 &

[PRES(t2 ) & AT(t 2 , see-a-mailllan'(x s »]]] &
[PRESet) & AT(t, bark'(x s »]]]
The latter implication constitutes the 'multiple-interval'

interpretation of the adverbial clause in sentence (317).
This approach is intuitively sound, and could easily be made
to account for 'multiple-interval' interpretations involving
other sorts of temporal adverbial clauses, and for
'single/multiple-interval' ambiguities. Thus, the 'single-
interval' interpretation of (303) might be represented as in
(325); the 'multiple-interval' interpretation of (303), on
the other hand, might be induced by translation (326), in
which the generalization operator G again appears.
(303) After heO finished dinner, heO smoked a cigar.

198 CHAPTER III
(325) vt[vt 1 [[t, < t & M(t,t 1 )] & vt 2 [t 2 = t1 &

[past(t 2 ) & AT(t 2 , VxS[R(X S , X~) &
finish-dinner'(x s )])]]] & [past(t) &
AT(t, Vxs[R(x s , xg) & smoke_a-cigar'(x s )])]]
(326) G(A't..X Svt[Vt 1 [[t 1 <t

M(t,t 1 )] & Vt 2 [t 2 = t1 &
&
[past(t 2 ) & AT(ti' finish-dinner'(x s »]]] &
[past(t) & AT(t, smoke-a-cigar'(xS»]])(xoC)
Thus, the 'multiple-interval' interpretation of a temporal

adverbial clause seemingly may be derived from the 'single-
interval' interpretation by means of Carlson's operator G.
Despite the initial plausibility of this approach,
there are certain kinds of 'multiple-interval' sentences
which militate against it. This approach predicts that if a
temporal adverbial clause has a 'multiple-interval'
interpretation, then either its superordinate clause or the
adverbial clause itself must have a generalizable stage-
level predicate. Unfortunately, there are perfectly good
examples of sentences whose temporal adverbial clause has a
'mul tiple-interval' interpretation but which lack
generalizable stage-predicat~s. In (327), for instance,
both the adverbial clause and the main clause have kind-
level predicates;
<327 ) When cats are widespread, dogs are rare.
here there is simply no stage-level property to generalize;

yet (327) may be understood to mean 'whenever cats are
widespread, dogs are rare.'
Also, Carlson shows that predicates consisting of be
plus a stage-level adjective phrase (e.g. be asleep, be
drunk, and be available) cannot be generalized: (328), for
example, can't receive a generic interpretation, but instead
induces an 'existential reading' for its bare plural
subject. 34
(328) Sailors were drunk.
Thus, 'multiple-interval' sentences like (329) are an

additional problem for the approach outlined above:
When he is drunk, John is unavailable.

in (329), neither the adverbial clause nor the main clause

has a generalizable stage-level predicate.
From these facts, I conclude that Carlson's operator G
is not to be employed in explicating the 'multiple-interval'
interpretation of temporal adverbial clauses. The problem
with Carlson's operator is not the notion of generalization
~ se; for all sentences involving an adverbial clause with
a 'multiple-interval' interpretation imply a generalization
of a certain kind, including (327) and (329). The problem
with G instead appears to be in the sort of generalization
it embodies: generalization from the stage level to the
individual level. It is not clear that any shifting of
levels is implied in 'multiple-interval' interpretations.
I shall therefore introduce a new operator G" which
doesn't shift its argument to a 'higher' level, but embodies
a much simpler kind of generalization. G" is a dyadic
operator; like a relative frequency adverb, it denotes a
two-place relation between sets of time intervals. Where Sl
and y' are expressions of type <i,t>, the truth of
G"(SI)(yl) implies that a certain number of time intervals
in the denotation of y' are in the denotation of 6'.
Exactly how many intervals are enough for G"( S')( y') to be
true varies with S' and y': for example, sentence (330)
seems to imply that Mary has a terrible limp whenever she
walks without her cane, while the truth of (331) would seem
to admit some instances in which John remembered to pour the
orange juice.
(330) When Mary walks without her cane. she has a

terrible limp.
(331 ) When John fixes breakfast, he forgets to pour the
orange juice.
Necessary and sufficient conditions for the truth of

G"(S')(Y')are therefore not specified by the grammar.
The rule which introduces Gil is parallel to the rule
(258) (see section 3.1) which introduces relati ve frequency
adverbs that are restricted by a set-level time adverb: it
joins a set-level time adverb wi th a temporal abstract to
produce a temporal abstract. The rule introducing Gil is
(332) •
200 CHAPTER III
(332) S66. If a e: PTA and tl e: PTAB , [yTense, -Adjunct,

6Perfect], then F66a (a,tl), F66b (a,tl) e; PTAB ,
[yTense, -Adjunct, 6Perfect], where F 66a (a,tl)
is CONWIB(a,tl) and F 66b (a,a) is CON(a,a).
T66. If a e: PTA' tl E PTAB , and a,a translate as

a',a', then F 66a (a,tl), F 66b (a,tl) translate as
>.t[G"(a')(>.t 1 [t 1 ~ t & a'(t 1 )])].
According to (332), sentence (317) may be analyzed as in

(333) and translated as (334) .•
(333)
---------
Fido, T
(Fido) [barks] when he sees a mailBan, t, 55,0
(heO) [barks] when heO sees a mailman, t, 60

I
(heO) [barks] when heO sees a mailman, TAB, 66b
~
when beo sees a mailman, TA, 67,3 (beO) [barks], TAB, 58
~ I
when, (beo ) [sees] a mailman (beo ) [barks], TAB, 56
TSC at-tbat-time 3 , t, 60 I
I (beo ) [barks], t, 19
(heO) [sees] a mailman
at-tbat-time 3 , TAB, 61b
~
at-that-ti.e3' (beO) [sees] a mailman, TAB, 58
MTA I
[sees] a mailman, TAB, 56
I
Cbeo) [sees] a mailman, t, 19
(334) Vt[G"( At1 Vt 2 [t 1 = t2 & vt 3 [t 3 = t2 & [PRES(t 3 ) &
ATCt 3 , be o-sees-a-mailman')]]])(>.t 1 [t 1 ~ t &
[PRES(t 1) & AT(tl' heO-barks')]])]
Similarly, the past tense sentence (303) can be deri ved in

two distinct ways--(335) and (336)--and thereby assigned the
two distinct interpretations represented in (337).
after heO finished dinner, (heO) [a.oked) a cigar,

I t, 60
after heO finished dinner, (beO) [a.oked] a cigar, TAB, 66a
~
after heO finished dinner, (heO) [smoked] a cigar,
~ TA, 67,3 I TAB, 57
after, (heO) [finished] dinner (heO) [s.okes) a cigar,
TSC at-thjt-time 3 , t, 60 I TAB, 56
(heO) [finished] dinner (heO) [a.okes) a cigar, t, 19
at-that-time 3 , TAB, 61b
~
at-that-time 3 , (heO) [finished] dinner, TAB, 57
MTA I
[finishes] dinner, TAB, 56
I
[finishes] dinner, t, 19
(336) after he O finished dinner, (heO) [s.aked] a cigar,

I t, 60
after heo finished dinner, (heo) [a.oked] a Cigar, TAB, 61a
~
after heo finished dinner, (heo) [s.aked] a Cigar,
I MTA, 63 I TAB, 57
I
after heO finished dinner, (heO) [a.okes] a cigar,
~A, 67,3 TAB, 56
after, (heO) [finished] dinner (heo) [smokes] a cigar,
TSC at-that-time 3 , t, 60 t, 19
I
---------
(heO) [finished] dinner at-that-time 3 , TAB, 61b
at-that-time 3 , MTA (heO) [finished] dinner, TAB, 57

I
(heO) [finishes] dinner, TAB, 56
I
(heO) [finishes] dinner, t, 19
202 CHAPTER III
a. 'Multiple-interval':
vt[G"Ot 1vt 2 [[t 2 < tl & MCt 1,t 2 )] &

vt 3 [t 3 = t2 & [pastCt 3 ) &
ATCt 3 , heo-finishes-dinner')]]])(At,[t, =- t &
[past( t 1) & AT( t l' heO-smokes-a-cigar')]])]
b. 'Single-interval':
vt[vt, [[t 1 < t & M(t,t 1 )] & vt 2 [t 2 = t1 &

[past(t 2 ) & AT(t 2 , heO-finishes-dinner')]]] &
[past(t) & AT(t, heO-smokes-a-cigar')]]
And the problematic sentences (327) and (329) can be derived

as in (338) and (340), and thereby assigned the translations
(339) and (341).
(338) when cats are widespread, (dogs) [are] rare, t, 60
---------
I
when cats are widespread, (dogs) [are] rare, TAB, 66a
when cats are widespread, (dogs) [are] rare,

~ TA,67,7 TAB, 58
I
when, (cats) [are] widespread (dogs) [are] rare,
TSC at-that-time7 , t, 60 TAB, 56
I I
(cats) [are] widespread (dogs) [are] rare, t, '9
at-that-time7 , TAB, 61b
at-that-t~ (cats) [are] widespread, TAB, 58

I
(cats) [are] widespread, TAB, 56
I
(cats) [are] widespread, t, 19
(339) vt[G"(~t1vt2[tl = t2 & vt 3[t 3 = t2 & [PRES(t 3 ) &

AT(t 3 , widespread'(cats'»]]])(~tl[tl c t &
[PRESet,) & AT(t" rare'(dog~'»]])]
---------
(340) when he is drunk, (John) [is] unavailable, t, 55,0
John, T when heO is drunk, (heO) [is] unavailable, t, 60

I
-------=-
when heO is drunk, (heO) [is] unavailable, TAB, 66a
when beo is drunk, (beO) [is] unavailable, TAB, 58

~ TA, 67,4 I
/ ~ (heO) [is] unavailable, TAB, 56
when, (beO) [is] drunk I
TSC at-that-tu.e4 , t, 60 (beO) [is] unavailable,
I t, 19
-------------
(beO) [is] drunk at-tbat-time4 , TAB, 61b
at-that-time4' MTA (heO) [is] drunk, TAB, 58

I
(beO) [is] drunk, TAB, 56
I
(heO) [is] drunk, t, 19
(341) vt[G"(At 1vt 2 [t 1 = t2 & vt 3 [t 3 = t2 &

[PRES(t 3 ) & AT(t 3 , Vxs[R(x s , John') &
drunk'(x s )])J]])(At 1 [t 1 ~ t & [PRES(t 1 ) &
AT(t 1, Vxs[R(x s , Jobn') & unavailable'(x s )])]])]
4.3. ADJUNCTS RESTRICTING THE GENERALIZATION OPERATOR Gil
Given tbe foregoing account of the generalization operator

Gil, tbe analysis of sentences like (290)-(293) can now be
reconsidered.

(291) Drunk, he drives very dangerously.
(292) In first gear, the truck makes funny noises.
(293) Wearing her new outfit, Mary looks fat.
The interpretations of (290)-(293) on which they are

understood to express a generalization about intervals at
which their (weak) adjunct is true are just those in which
the adjunct has a 'multiple-interval' interpretation. Since
free adjuncts have already been introduced into the category
TA of set-level time adverbs, derivations for sentences
(290)-(293) are automatically produced by rule (332);
204 CHAPTER III
(342) in first gear, (the truck) [makes] tunny noises,

I t, 60
in first gear, (the truck) [makes] funny noises, TAB, 66a
in first gear,
TA, 9,9
I
-----------(the truck) [makes] tunny noises, TAB, 58
I
(the truck) [makes] fUnny noises, TAB, 56
in Cirst gear, I
TAB, ',4 (the truck) [makes] fUnny nOises, t, '9
sentence (292), for example, may be derived as in (342).

This derivation determines (343) as the translation for
(292); (343) induces the desired interpretation, in which
the adjunct in first gear specifies a set of intervals
relative to which a certain generalization holds. Observe
that very little needs to be inferred about the logical
relation between the free adjunct and the main clause under
this interpretation; the nature of this relation is largely
. specified by the operator G".3 5
vt[G"(At,vt 2 [M 9 (t"t 2 ) & AT(t 2 , Vxs[R(x s , x~) &
in-Cirst-gear'(x s )]) ])(A t, [t, =- t & [PRES(t,) &

AT(t 1, the-truck'('" AxiVXS[R(XS,X i ) &
make-fUnny-noises'(x s )]»]])]
Now consider sentences (294)-(297).
(294) Being a businessman. John smokes cigars.

(295) Being a drunk, he drives very dangerously.
(296) Having a rather decrepit engine, the truck makes
funny noises.
(297) Weighing over 200 lbs., Mary looks fat.
Recall that the strong adjuncts in these sentences lack the

'multiple-interval' interpretation found in (290)-(293);
they cannot be understood to specify a set of intervals over
which some sort of generalization is made (i.e. as the first
argument of Gil). As in modal sentences and sentences with
relative frequency adverbs, they may always be understood as
main tense adverbs. Here, the sentences which they modify
may either be understood as narrative presents or, more
naturally, as generic sentences; in the latter case,
Carlson's G operator, not the dyadic Gil operator introduced
above, is responsible for the generalization. Thus,

sentence (294) may be analyzed as in (344), and assigned the
translation (345).
(344) being a businesa.an, (John) [s.okes] cigars, t, 60

I
being a businessman, (John) [a.okes] cigars, TAB, 61a
being a businessman,
~(John) [s.okes] cigars, TAB, 58
MTA, ",6,9 I
I (John) [s.okes] cigars, TAB, 56
being a businessman, I
TAB, 2,4 (John) [smokes] cigars, t, 19
---------
John, T [saoke] cigars, lVi, 22
I
[s.oke] cigars, IVs, 28
(345) Vt[K(L6 )("'vt, [M 9 (t,t,) &

AT(t" being-a-bUsiness.an'(x~» ])("'[PRES(t) &
AT(t, G("'s.oke-cigars')(John'»])]
Under the proposed analysis, nonpresent tense sentences

of the character of (290)-(293) are straightforwardly
derived. Sentence (346), for example, may be analyzed as in
(347) and assigned the translation (348).
(346) Lying on the beach, John smoked cigars.
lying on tbe beacb, (.lobo) [s.oked] cigars, t, 60
--------
I
lying on the beacb, (.lobo) [smoked] Cigars, TAB, 668
lying on tbe beacb, (.lobo) [s.oked] cigars, TAB, 57

TA, 9,9 I
I (Jobo) [s.okes] cigars, TAB, 56
lying on tbe beach, I
TAB, ',4 (.lobo) [s.okes] cigars, t, '9
(348) Vt[G"(At 1vt 2 [M 9(t"t 2 ) & AT(t 2 , Vxs[R(x s ,

lying-on-the-beacb'(xs)])])(At, [t, =. t &
[past(t,) & AT(t" Vxs[R(x s , Jobn') &
s.oke-cigars' (x s )])]])]
206 CHAPTER III
In this section, a third environment has been found--

generic sentences--in which strong and weak adjuncts are
interpreted in distinct ways. To capture these differences,
a binary generalization operator Gil has been postulated for
the analysis of generic sentences of a common sort. In view
of this analysis, a clear generalization emerges about the
distinction between weak and strong adjuncts: the former
but not the latter can serve to restrict the interpretation
of a variety of different binary operators--including
modals, relative frequency adverbs, and the operator Gil_-
appearing in their superordinate clause.
5. Chapter Summary
In this chapter, the role played by free adjuncts in the

system of English temporal semantics has been examined at
length. A number of critical assumptions about tense and
time adverbs were set forth in section 1. The most
significant of these preliminary assumptions were: (a) the
assumption that tensed sentences and free adjuncts deri ve
from temporal abstracts, expressions denoting sets of time
intervals; and (b) the assumption that there are two
categories of time adverbs in English: (i) the category TA
of set-level time adverbs, which denote sets of time
intervals; and (ii) the category MTA of main tense adverbs,
which denote functions from properties of time intervals to
sets of time intervals, and which may thus combine with a
temporal abstract denoting a certain set S of time intervals
to yield a new temporal abstract denoting a subset of S.
In section 2, rules were formulated for deriving
temporal abstracts from predicati ve phrases. The 'A-
abstracts' so derived belong to the subclass [+Adjunct],
which is itself partitioned into two subclasses: A-
abstracts deriving from stage-level predicative phrases are
in the subclass [+Weak], while those deriving from
individual-level predicative phrases are subclassified
[-Weak]. In addition, a rule for deriving conditional
adjuncts (i.e. adjuncts of category t') from A-abstracts was
proposed; in accordance with the conclusions of Chapter II,
this rule applies only to A-abstracts in the subclass
[+Weak]. Finally, it was proposed that the 'adsentential'
adjuncts discussed in Chapter II are in fact main tense
adverbs (i.e. members of category MTA); accordingly, a rule
deriving main tense adverbs from A-abstracts was formalized.
In section 3, it was observed that strong and weak

adjuncts may receive different sorts of interpretations when
their superordinate clause contains a relative frequency
adverb: weak adjuncts may restrict the set of time
intervals quantified by the frequency adverb, while strong
adjuncts may not. A formal analysis of the semantics of
relative frequency adverbs was developed as a basis for
elucidating this new difference between weak and strong
adjuncts: in this analysis, relative frequency adverbs are
regarded as denoting binary relations between sets of time
intervals; accordingly, a frequency adverb of this sort
requires two arguments--a set-level time adverb (whose
function is to specify the set of intervals quantified by
the frequency adverb) and a temporal abstract. This
analysis affords a very clear view of the difference between
weak and strong adjuncts in the context of a relative
frequency adverb: the former, but not the latter, may serve
as the first argument of a frequency adverb, assuming the
role of a set-level time adverb. To give formal substance
to this generalization, a rule was proposed which converts
A-abstracts in the subclass [+Weak] into set-level time
adverbs (i.e. expressions of category TA).
In section 4, another difference between weak and
strong adjuncts was uncovered in generic/habitual sentences:
when a sentence of this kind contains a free adjunct, the
sentence may be understood as expressing a generalization
about intervals at which the adjunct is true only if the
adjunct is weak. This special interpretation of weak
adjuncts in generic sentences is just like the 'multiple
interval' interpretation of time adverbs appearing in
generic sentences. To account for this interpretation of
time adverbs, a generalization operator Gn was postulated,
whose semantic characteristics are reminiscent of those of
relative frequency adverbs; Gn denotes a binary relation
between sets of time intervals, and thus requires as its
arguments a set-level time adverb and a temporal abstract.
Under this analYSiS, the special interpretation of weak
adjuncts in generic sentences can be viewed as a consequence
of the fact that, as set-level time adverbs, weak adjuncts
may serve as the first argument of Gn ; thus, the rule
converting A-abstracts into set-level time adverbs--
originally proposed to account for the behavior of weak
adjuncts in sentences with relative frequency adverbs--
receives additional motivation in the proposed analysis of
generic sentences. Since this rule only affects A-abstracts
208 CHAPTER III
in the subclass [+Weak], strong adjuncts are correctly

predicted never to serve as the first argument of Gil 36
When a weak adjunct of category TA serves as the first
argument of a binary operator in the main clause (i.e. of Gil
or of a relative frequency adverb) the semantic properties
of this operator largely determine the logical role which
the adjunct is felt to play. Recall from Chapter II that
this is also true of conditional adjuncts serving as the
first argument of a modal verb. Nevertheless, certain
components of the interpretation of a free adjunct may not
be determined by rules of semantics; accordingly, a central
characteristic of the analysis of free adjuncts laid out in
this chapter is the use of free variables to represent those
aspects of an adjunct's interpretation which are
semantically indeterminate. For instance, in the
translation of an adjunct of category MTA, the variable L
(of type «s,t),«s,t),t») stands for the indeterminate
logical connection between the propositions expressed by the
free adjunct and its superordinate clause; and the variable
M (of type <i,<i,t») specifically represents the
indeterminate temporal relation between the events (or
states of affairs) described by the adjunct and the main
clause. (M serves a similar function in the translation of
an adjunct of category TA.) My claim is that these
indeterminacies are resolved by the inferences of language
users. This claim entails that inferences playa central
role in the interpretation of linguistic expressions--that
they may actually determine the truthconditions of
individual sentences.
Several issues raised in this chapter remain to be
discussed. First, why is it that weak adjuncts may
condition the interpretation of a binary operator (i.e. a
modal, a frequency adverb, or the operator G~ appearing in
their superordinate clause, while strong adjuncts may not?
Can this be viewed as a necessary consequence of the fact
that weak adjuncts derive from stage-level predicates, while
strong adjuncts derive from individual-level predicates?
Also, how is it that people draw the necessary inferences to
resol ve the temporally indeterminate relation between an
adjunct and its superordinate clause? Why, for example, is
a while-type relation most naturally inferred for the
adjunct in sentence (349), but an after-type relation for
the one in section (350)?
Crossing the street, he was almost hit by a car.

(350) Getting down from her highchair, she crawled into

the living room.
These issues will be addressed in detail in Chapter VI as

part of a general discussion of the role of inference in the
interpretation of free adjuncts and absolutes.
Footnotes
1. The reason for defining the past tense predicate in
this way will be taken up in section 1 of Chapter IV.
2. In section 2 of this chapter. I shall propose that free

adjuncts. like sentences. derive from temporal abstracts.
The temporal abstracts from which free adjuncts derive
(which are subclassified [+Adjunct]) differ in certain ways
from those giving rise to tensed sentences: (i) they have
the form of predicative phrases rather than sentences; (ii)
they do not show explicit tense-marking; and (iii) they may
be directly converted into expressions of categories t'.
MTA. and TA (with no change in form). but never to the
category t of declarative sentences. Because of these
differences. rule S56 is relevant only for the derivation of
temporal abstracts in the subclass [-Adjunctl. and rules
S57-S60 only apply to adjuncts so classified.
Despite these differences between temporal abstracts
classified as [-Adjunct] and those classified [+Adjunct].
they do have some similarities; for example. temporal
abstracts of all kinds undergo the rule joining a main tense
adverb with a temporal abstract. It is the existence of
such similarities which makes it plausible to regard
temporal abstracts in the subclasses [-Adjunctl and
[+Adjunct] as forming a single category rather than two
separate categories.
Temporal abstracts are also subclassified with respect
to the feature [+Perfectl. All members of the subclass
[+Perfect] denote sets of 'perfect' time intervals, in a
sense to be made clear in section 1 of Chapter IV; there, I
will explain why rules S57-S60 cannot apply to members of
the subclass [+Perfect].
3. That is. main tense adverbs are regarded as denoting

functions of type «s.<i.t».<i.t». rather than functions
of type «s,<i.t».t> (as Dowty (1979:326f) has advocated);
210 CHAPTER III
as was pointed out above, this analysis is motivated by the

fact that the rule joining a main tense adverb with a
temporal abstract must be allowed to iterate in the
derivation of a sentence with multiple time adverbs. (For a
formal exam pie of such a derivation, vide infra.)
As Dowty has pointed out to me, multiple main tense
adverbs can be accounted for under the «s,<i,t»,t>
analysis if it is assumed that multiple main tense adverbs
form a constituent; this, however, is a questionable
assumption--cf. (i)--and is in no way necessitated by the
analysis adopted here.
(i) Yesterday John saw Mary while he ~ walking home.
4. Thus, S63 establishes a derivational relationship

between main tense adverbs and certain set-level adverbs,
but only when the latter are in the subclass [-Adjunct]. In
the system proposed here, free adjuncts are generated as
members of both categories MTA and TA; in the case of free
adjuncts, however, members of the former category do not
derive from members of the latter category--rather, adjuncts
belonging to these two categories are independently
generated (although they do derive from a single, common
source: temporal abstracts). The reasons for this
peculiarity of free adjuncts are noted in section 3.2.
5. Dowty (1979;325) suggests that a sentence like John

viII leave today conventionally implicates (rather than
entails) that i is in the future, where i is the interval of
John's alleged departure. Under this approach, a sentence
like John will leave yesterday is not contradictory, nor is
a sentence like John viII not leave yesterday tautologous;
but they are both inappropriate, because their entailments
conflict with their conventional implicatures. This
approach is founded on the assumption that an unasserted (or
unassertable) aspect of the meaning of a sentence does not
contribute to its truthcondi tions. As this is not an
uncontroversial assumption (it would, for example, force one
to regard John leCt on Thursday and John viII leave on
Thursday as truthconditionally equivalent), I prefer to
assume that John will leave yesterday is a contradiction,
and John viII not leave yesterday a tautology, and that
their pragmatic peculiarity is simply a consequence of these
facts.
6. Note that without such a context variable, aCter Mary

sang would just denote the whole set of intervals following
the first one at which Mary sang; similarly, beCore Mary
sang would denote the whole set of intervals preceding the
last interval at which Mary sang.
1. Unlike other sorts of temporal adverbial clauses,

beCore-clauses sound reasonably acceptable with periphrastic
future tense-marking; I am not sure why this should be so.
8. I am of course ignoring certain other interpretations

of (100): the interpretation in which the subordinate
clause Mary was leaving functions as a 'shifted present',
and the two interpretations (parallel to those represented
by (101) and (102» in which Mary was leaving serves as a
'shifted futurate progressive' (cf. Dowty 1919:154fO. The
fragment developed here is incapable of producing these
three additional interpretations.
9. Therefore, in the fragment formalized in the Appendix,

rule (104) is prevented from applying to while.
10. Sentences like these are sometimes acceptable when a

'cyclic' concept of time is assumed. For example, in a
context in which Mary arrived at 2:00 p.m. yesterday, (146)
can be acceptably used to assert that John will leave at
2:00 p.m. sometime in the future. Cyclic interpretations
like this appear to be possible for when-, beCore-, and
aCter-adverbials; interestingly, while-adverbials appear not
to have such an interpretation. Interpretations of this
sort are not induced by the fragment developed here; they
would appear to require a model structure incorporating a
cyclic (rather than linear) ordering of time intervals.
11. According to Stalnaker (1918:321), 'the presuppositions

of a speaker are the proposi tions whose truth he takes for
granted as part of the background of the conversation. A
proposition is presupposed if the speaker is disposed to act
as if he assumes or believes that the proposition is true,
and as if he assumes or believes that his audience assumes
or believes that it is true as well!
12. Instances of temporal adverbial clauses with marked

future tense do show up in archaic forms of English:
212 CHAPTER III
(i) Immediately after they shall be assembled in

consequence of the first election, they shall be
di vided as equally as may be into three classes.
(U. S. Consti tution 1.3.2)
Nevertheless, in modern-day standard English, this is not

done; (i), for example, would be 'Immediately after they
are assembled ••• '
13. Cf. McCawley (1978), where it is argued that given two

equivalent expressions, the use of the more complex one
implicates that the less complex one is somehow less
appropriate in the circumstances at hand. Perhaps (150) is
routinely avoided because there simply aren't any
circumstances which would make (165) less appropriate than
(150); indeed, it is quite difficult to imagine what such
circumstances would be like.
14. Actually. this isn't quite accurate. because the

'extraction site' in tP may be in an embedded clause. For
simplicity's sake. I will ignore when-adverbials of this
sort in this section; their truthconditions will, in any
event, be analogous to those for the simpler sort of when-
adverbial in which the 'extraction site' is in the highest
clause.
15. Dowty (1979:88ff) assumes that sentences with

achievement predicates are true at nonminimal intervals,
because they may occur in the progressive. Nevertheless, I
believe that sentences with 'instantaneous achievement
predicates' such as spot (as in He spotted an elephant
through his binoculars) may only be true at a moment; note
the oddness of ??He was spotting an elephant.
16. These are essentially the truthcondi tions implied by

Bennett & Partee (1972:31).
17. Here again I am ignoring the possibility that the

'extraction site' in tP may be in an embedded clause; I am
also ignoring the contribution made by the context variable
M to the semantics of before-adverbials in my analysis. The
truthconditions which I assume are essentially those of
Bennett & Partee (1972:32).
18. Heinamaki also discusses the truthconditions for

connective. Such sentences will be ignored here.
19. Once again I neglect cases in which the 'extraction

site' in 1jJ is embedded. The truthconditions assumed here
are basically those of Bennett & Partee (1972:32).
20. Williams (1975:267) asserts that free adjuncts

('adverbial participles') cannot have sentence adverbs,
citing the 'strangeness' of (i) as evidence.
Ci) ?Sam, predictably seeking Mary's favor, is a fool.
I have found no such restriction on the natural occurrences

of free adjuncts which I have collected:
(ii) As the mare labored up the ravine, he dismounted

and walked beside her, occasionally stroking her
muzzle.
(OSI, 326)
(iii) Predictably defensive, evangelical politicians

charge that such criticisms are themselves
political sour grapes from liberal opponents.
(Newsweek 9/15/80, 31)
21. (212) is stated so as to prevent individual-level

adjective phrases from being converted into A-abstracts;
this is because such phrases do not in fact occur as free
adjuncts. Sentences (ia)-( iva), in which individual-level
adjective phrases appear as the sole predicates of free
adjuncts, are unacceptable, while (ib)-Civb), in which
stage-level adjective phrases appear instead, are fine.
Note, however, that copulative predicates deriving from
individual-level adjective phrases are perfectly acceptable
in free adjuncts, as (ic)-(ivc) show and as (212) predicts.
( i) a. *Intelligent, John answered all the

questions.
b. As alert as ever, John answered all the
questions.
c. Being intelligent, John answered all the
questions.
(ii) a. ·Mary, fat, couldn't get into her new outfit.

214 CHAPTER III
b. Mary, stiff from broadjumping, couldn't get

into her new outfit.
c. Mary, being fat, couldn't get into her new
outfit.
( iii) a. *Rufus jumped up and down, a little bit

strange.
b. Rufus jumped up and down, a little bit
intoxicated.
c. Rufus jumped up and down, being a little bit
strange.
(iv) a. *Quite rare, the timber wolf will die off

quickly.
b. Captive, the timber wolf will die off
quickly.
c. Being quite rare, the timber wolf will die
off quickly.
22. The complexity of T2 in (212) is a consequence of the

decision to introduce sorted types into the intensional
logic employed here. If a translates as an expression
denoting a set of objects (i.e. of sorted type <o,t'», an
object-level variable must be supplied as its implicit
'subject' if the resulting translation is to be defined on
all assignments of values to variables. Analogously, a
variable of the appropriate sorted type is necessary if the
translation of a denotes a set of kinds (Le. is of sorted
type <k,t'» or a set of individuals (Le. is of type
< [o,k] ,t'». See the Appendix for a formal account of the
sorting of enti ties into kinds, objects, individuals, and
stages.
23. Temporal abstracts, including A-abstracts, are

subclassified with respect to the feature [+Perfect]; all
abstracts in the subclass [+Perfect] denote sets of
'perfect' time intervals, a notion which will be defined in
Chapter IV, section 1. For reasons which will be made clear
there, the rule (213) converting A-abstracts into
expressions of category t' can only apply to abstracts in
the subclass [-Perfect].
24. Rule (221) applies only to A-abstracts in the subclass

[-Perfect], for reasons which will be made explicit in
section 1.3 of Chapter IV. cr. footnote 23.
25. Of course, I am assuming the standard abbreviatory

convention according to which a(X,y) = a(y)(X).
26. See stump (1981 a) for a discussion of these two sorts

of adverbs.
27. The motive for interpreting relative frequency adverbs

as two-place relations rather than as functions is thus very
much like the reason for interpreting modals relationally--
to 'factor out' a recurrent context dependency. Cf. Chapter
II, section 2.1.
28. As has been seen, relative frequency adverbs specify

frequencies the average length of whose periods is
proportional to the length of some other interval. They
are, however, usually vague with respect to the proportion
which they determine in a given context: for example, even
if the length of the commercial in question is known,
sentence (i) is vague regarding the number of times the fly
landed on my foot.
(i) During the commercial, the fly often landed on my

foot.
Also, the individual periods into which a relative frequency

adverb divides a given interval (or sequence of intervals)
may vary in length; for example, if different sailors have
strolled by at 1:07, 1:20, 1:29, and 1:51, Occasionally a
sailor has strolled by can be truthfully uttered at 2;00,
even though there is no fixed frequency with which sailors
have gone by.
29. Actually, because the denotations of S' and y' may be

infinite, a'(8')(y') requires a slightly more complicated
kind of interpretation: for any relative frequency adverb
a, there is some range R of frequencies such that a'(8'Hy')
is true iff each finite subsequence of the (possibly
infinite) sequence S of intervals determined by the
denotation of 8' has a finite supersequence s in S such that
for some frequency f within range R, members of the
denotation of y' appear in s with frequency f.
30. In section 2.3, I proposed that adjuncts of category

MTA derive directly from A-abstracts (by means of rule S11).
The introduction of adjuncts of category TA, however, would
216 CHAPTER III
seem to open the way to a rather different analysis of

adjuncts of the former sort. Recall that rule S63 shifts
set-level time adverbs in the subclass [-Adjunct] to the
category MTA. If this rule were generalized so as to apply
to members of the subclass [+Adjunct] as well, then adjuncts
of category MTA could be derived by applying rule (272) and
then rule 363 to an A-abstract. This would seemingly allow
rule S11 to be dispensed with. For two reasons, however,
this approach to the derivation of adjuncts of category MTA
would not work. First, all adjuncts of category TA are
weak, while adjuncts of category MTA may be either weak or
strong; if the latter were assumed to derive from the
former, then the fact that strong adjuncts occur as members
of the category MTA but not as members of the category TA
would become highly problematic. Second, the translation of
an adjunct of category MTA must contain the variable
subordinator K(L), according to the discussion in Chapter
II, section 2.2.2; if such adjuncts derived from adjuncts of
category TA by means of rule 363, the resulting translations
would lack this element.
Rule (272) applies only to A-abstracts in the subclass
[-Perfect]. The reason for this characteristic of rule
(272) will be made clear in section 1.3 of Chapter IV. Cf.
footnote 23.
31. In the proposed analysis, the translation of an adjunct

of category TA contains a free variable M, evaluated by
inference; in this one respect, the logical role of an
adjunct of this category depends on language users'
inferences.
Under the proposed treatment of condi tional adjuncts
(i.e. adjuncts of category t'), their logical relation to
their superordinate clause is completely determined by the
semantics of the modal verb in the main clause. It may be,
however, that they too have an element of semantic
indeterminacy, comparable to that represented by the
variable M in the tran~lations of other sorts of adjuncts.
Recall that M is used to represent the indeterminate
temporal relation between the event(s) (state(s) of affairs)
described by the adjunct itself and the event(s) (state(s)
of affairs) described by the main clause in sentences such
as (i) and (ii), whose adjuncts are main tense adverbs.
(i) Getting down from his highchair, Leroy hurt

himself.
(ii) Getting down from his highchair, Leroy walked into
the living room.
In these sentences, the temporal relation between the event

described by the adjunct and that described by the main
clause varies; the adjunct in (i) is naturally paraphrased
with a while-clause. while that in (ii) is best paraphrased
with an after-clause. The fact that a similar temporal
indeterminacy shows up in sentences with conditional
adjuncts--cf. (iii) and (iv)--suggests that the variable M
should figure in the translations of such adjuncts as well.
( iii) Getting down from his highchair, Leroy might hurt

himself.
(iv) Getting down from his highchair, Leroy might walk
into the living room.
32. An analogous point has already been made with respect

to weak adjuncts in modal contexts--see the discussion
accompanying example (50) in Chapter II.
33. The translation assigned to sentence (242b) is (i).
(i) Vt[ oCten' (A t1 vt 2 [M 9 (tl' t 2 ) &

AT(t 2 • Vxs[R(x s , xJ") & carrying_-Ioad-oC-
over-1500-lbs.' (x s )])])(At 1 [t 1 ~ t & [PRES(t 1 ) &
AT(t 1 , our-truck-makes-the-bridge-shake')]])]
The value inferred for M9 in (i) will, of course, specify

the appropriate relation between the intervals denoted by t,
and t 2 ; notice, however, that it may additionally specify
that these intervals satisfy some tense predicate, e.g.
'PRES(O'. I f this were not the case, (1) might not induce
the right interpretation for (242b).
34. This is an additional reason,for suppos~ng that the be

in be asleep translates as WS AX1VXs[R(x s ,Xl) & pS{x s }] (as
in (89) in Chapter II): ~his transla~ion guarantees that be
asleep (translation: AX1VX s [R (x s , Xl) & asleep'( x s )]) will
not be a possible candidate for generalization.
35. Nevertheless, the exact temporal connection between the

states described by the adjunct and the events described by
the main clause is viewed as semantically indeterminate;
218 CHAPTER III
this indeterminacy is represented by the free variable M9 in

(343).
36. The analysis proposed in this chapter for relative

frequency adverbs and for the operator G" was first set
forth in stump (1981 b). Since that time, Farkas & Sugioka
(1983) have proposed a somewhat different analysis of these
matters. According to their analysis, a when-clause
restricting a frequency adverb in a sentence such as (i) is
interpreted as the antecedent of a condi tional expression
'p c q' within the scope of the adverb, which is itself
interpreted as an unselective quantifier ~ 1a Lewis (1975):
(i) Bears are always intelligent when they have blue

eyes.
(ii) al ways' (have-blue-eyes' (x o ) c intelligent' (xo)),

where X O ranges over bears.
(An expression 'p c q', according to Farkas & Sugioka

(p.240), 'is true if p = T and q = T, and it is false if p =
T and q = F. The cases in which p = F would be irrelevant
for assigning a truth value ••• ') Similarly, the 'multiple-
interval' interpretation of a when-clause in a sentence such
as (iii) is as in (iv), where G is an unse1ective, generic
quantifier:
(iii) John is grouchy when it rains.
(i v) G( it-rains' c grouchy'(x s )), where XS ranges over

stages of John.
Al though this analysis yields a number of interesting

results, I have certain reservations about it. For one
thing, it provides no explicit account of the semantic
connection between the single-interval and multiple-interval
interpretations for a when-clause such as that in (v);
(v) When John heard that song, he thought of Mary.
recall that under my analysis, the when-clause in (v)

derives from a single source under both interpretations,
namely the set-level time adverb when John heard that song.
In addition, Farkas & Sugioka's analysis provides no account
of the fact that all kinds of temporal adverbial clauses,
and not simply when-clauses. can restrict the interpretation
of a frequency adverb. or receive a multiple-interval
interpretation in a generic sentence:
( vi) Before he goes to the dentist, John always takes

two aspirin.
(vii) After he finishes supper, John smokes his pipe.

One of the purposes of their analysis, of course, is to
account for the so-called 'atemporal' uses of when-clauses,
such as that in (i). Although I have not focused on such
uses here, they do not pose a serious problem for the
analysis which I have proposed; as I have shown elsewhere
(Stump 1981a), it is possible to assimilate the atemporal
interpretation associated with a frequency adverb to the
temporal interpretation, and the same appears to be true of
the atemporal interpretation of a generic sentence such as
(viii), if this is analyzed as having a generic operator
analogous to .the frequency adverb in (i).
(viii) Bears are intelligent when they have blue eyes.

G HAP T E R I V
ASPECT AND THE INTERPRETATION OF FREE ADJUNCTS
This chapter is devoted to the perfect and the progressive,

and their relevance to the semantics of free adjuncts.
Specifically, three matters are investigated here.
(i) An important peculiarity of free adjuncts is that
when they are in the perfect, they may be modified by a
wider range of time adverbs than may modify a sentence in
the present perfect, as the examples in (1) and (2) show.
(1) a. Having been on the train since noon, John

knows exactly why i t derailed.
b. Having been on the train yesterday, John
knows exactly why it derailed.
(2 ) a. John has been on the train since nooo.
b. *John has been on the train yesterday.
The perfect adjuncts in (1), which are modified by the time

adverbs since noon and yesterday, are both acceptable; of
the corresponding present perfect sentences, however, only
(2a) is possible--in (2b), the cooccurrence of the present
perfect with the time adverb yesterday is unacceptable.
Previous accounts of the semantics of the perfect have
failed to provide any insight into this puzzling difference
between free adjuncts and tensed sentences; this is true
even of the so-called 'extended now' theory of the perfect,
distinguished by McGoard (1978) as the most satisfactory of
four alternatives. In section 1 of this chapter, I develop
a new analysis of the semantics of the perfect in English;
what distinguishes this analysis is that it provides a sound
basis for explaining the difference in acceptability
exemplified in (1) and (2).
(ii) In section 2.3 of Chapter III, it was proposed
that those adjuncts failing to restrict the interpretation
of a binary operator (such as a modal, a frequency adverb,
or the generalization operator G") be analyzed as main tense
adverbs. In section 2 of the present chapter. an argument
in favor of this proposal is presented. This argument
220
ASPECT AND FREE ADJUNCTS 221
hinges on the semantic analysis of the perfect laid out in

section 1. According to that analysis, the perfect tenses--
past perfect, present perfect, and future perfect--are
derived compositionally: semantically, each consists of a
perfect operator within the scope of a tense operator.
Moreover, just as main tense adverbs may join with tense to
specify the particular past, present, or future time
interval at which some proposition is true, they may
likewise join with the perfect to specify the 'perfect'
interval during which some proposition is true. In the
latter function a main tense adverb is within the scope of
tense, but not in the former function. Accordingly, if free
adjuncts not restricting the interpretation of some binary
operator are in fact main tense adverbs, then two
predictions follow: (a) when modifying a sentence in a
perfect tense, as in (3), an adjunct of category MTA should
be interpretable in two ways, according as it is understood
to join with tense or with the perfect to characterize a
particular interval;
(3) Working at the post office during the past summer,

Mary had decided to go back to school.
in one of these two interpretations (namely that in which

the adjunct is interpreted as jOining with the perfect), the
adjunct should be understood to be within the scope of the
tense of the main clause, but not under the other
interpretation; and (b) a parallel pair of interpretations
should not be possible when an adjunct of category MTA
modifies a sentence in a simple tense, as in (4);

Mary decided to go back to school.
in this case, the only interpretation should be one in which

the adjunct is not within the scope of tense. As I
demonstrate in section 2 of this chapter, bot~ of these
predictions are borne out, a result which strongly confirms
the view that free adjuncts may function as main tense
adverbs.
(iii) The third section in this chapter deals
specifically with the semantic peculiarities of present
participial adjuncts, such as that in sentence (5).
(5) Crossing the street, John was hit by a car.

222 CHAPTER IV
The nai ve intuition of most English speakers is that

adjuncts of this sort are comparable to sentences in the
progressive aspect:
(6) John was crossing the street.
There is a certain amount of prima facie evidence against

the view that present participial adjuncts are progressive
in aspect; nevertheless, I argue in section 3 that this
intuition is basically sound. In particular, I argue that
the distinctive semantic characteristics normally associated
with the progressive are in fact typical of present
participial phrases in a variety of constructions, including
the free adjunct construction; I propose a semantic analysis
of present participial phrases which accounts for these
characteristics, and which allows the progressive to be
simply regarded as a predicative participle construction.
In section 4, the conclusions of this chapter are
summarized. The rules proposed in the following discussion
appear in the formal fragment presented in the Appendix.
1. The Perfect Tense and the Interpretation of Free Adjuncts
As was seen in Chapter III, one of the most salient features

of any relation holding between the propositions expressed
by a free adjunct and its superordinate clause is its
temporal dimension. The two propositions may be felt to
stand in any of a range of different temporal relations--
precedence, simultaneity, subsequence, and so on. The
precise nature of the relation is normally inferred by
language users;' see Chapter VI for a discussion of inferred
temporal relations.
One special way of explicitly indicating a relation of
temporal succession in free adjunct sentences is to employ
the perfect in the adjunct:
(7) Having been on the train, John knows exactly why

it derailed.
The propositions expressed by the free adjunct and its

superordinate clause are felt to be simultaneously true in
such sentences; (7) is understood to mean that (8) and (9)
are simultaneously true.
(8) John has been on the train.

(9) John knows exactly why it derailed.
But the perfect establishes a relation of succession between

the proposition on which it operates and the proposition
expressed by the main clause; thus, if (8) and (9) are both
true at some interval i, then by virtue of the meaning of
the perfect, sentence (10) will normally be true at some
interval preceding or ending with i.
(10) John is on the train.
A peculiarity of perfect adjuncts is the fact that they

occur with a broader range of time adverbs than do finite
clauses in the present perfect. For example, although both
(11a,b) are acceptable, only (12a) is possible for most
speakers.
( 11) a. Having been on the train since noon, John

(12 ) a. John has been on the train since noon.
Superficially, this has the appearance of a completely

idiosyncratic fact. Yet, as I shall now argue, it has
profound implications for the analysis of the English
perfect. In particular, I shall show that the 'extended
now' theory of the perfect, distinguished by McGoard as the
most adequate of several alternative theories, is
insufficient to predict the difference in acceptability
between (11 b) and (12b); I shall present a revision of this
theory which embodies certain generalizations overlooked by
McCoard, and from which the difference between (11b) and
(12b) follows automatically.
1.1. THE SEMANTIC UNSPEGIFICITY OF THE PERFECT
In his recent monograph on the semantics of the English

perfect, Robert McCoard (1978) discusses four major analyses
of the perfect which appear recurrently--though in a variety
of guises--throughout the rather prodigious literature on
this subject. After reviewing a considerable body of
224 CHAPTER IV
evidence, McCoard concludes that three of these are

inadequate, either because they mistakenly equate the
meaning of the perfect with some pragmatic implication
associated with the perfect in certain contexts, or because
they depend on an erroneous identification of the perfect
with a complex configuration of simple tenses. The only
defensible analysis, he argues, is what he calls the
'extended now' theory.
This theory is based on a distinction between two sorts
of time intervals. At a given interval i, those intervals
which precede i and. are separated from i are termed past
intervals, as in (13). Those intervals which end with i, on
the other hand, are called extended nows, as in (14).
(13) i' is a past interval relative to i.
i' i
~ ~
0(
(14) i' is an extended now relative to i.
( ------
-------------~~---------------
it
i
)
The extended now theory holds that the function of the

perfect is to locate an event within an extended now, while
the function of the preteri t is to locate an event at some
past interval. According to this analysis, sentence (15) is
true at interval i iff sentence (16) is true sometime during
an extended now relative to i, as in (17).
(15) John has seen Mary.

(16) John sees Mary.
------- ------
(17) (15) is true at i iff (16) is true at i'.
i' i
I
---------------~---------------
in
[in is an extended now relative to i; i' ~ i".

NB: it may overlap i.]
The preterit sentence (18), on the other hand, is true at
interval i iff sentence (16) is true at some past interval
relative to i, as in (19).
(18 ) John saw Mary.

(19 ) ( 18) is true at i iff (16 ) is true at i'.
i' i
~ ~
( »
[i' is a past interval relative to i.]
These distinct truthcondi tions can be formally represented

by introducing an intensional logic predication 'ID(C)' such
that, where ~ denotes interval i', ID(C) is true at index
<w,i> iff i is a final subinterval for i'. Given this
addition, translations (20) and (21) embody the
truthconditions for (15) and (18) prescribed by the extended
now theory.2
(20) vt[xn(t) & vt 1[t 1 ~ t & AT(t 1, John-see~ry')]]
(21) vt[past(t) & AT(t, John-sees-Hary')]
In this theory, the perfect is rightly regarded, not as an

aspect, in the strict sense, but as a tense. (See McCoard
(1978: Chapter 1) for arguments to this effect.)
The special merit of the extended now theory is that it
affords a semantic account of which sorts of time adverbs
may occur with the present perfect and which may not. While
some time adverbs, such as those in (22), may occur with
either the present perfect or the preterit, others, such as
those listed in (23), occur only with the preterit, and
still others, like those listed in (24), only with the
present perfect. (Cf. McCoard (1978: 135).) This difference
is illustrated in examples (25 )-(27).
(22) today, in the past, in my life, for three hours,

recently, just now, this week, already
(23) long ago, five years ago, once [= formerly],

yesterday, the other day, last night, in 1900, at
3:00, after the war, no longer
226 CHAPTER IV
(24) at present, up till now, so far, as yet, not yet,

during these five years past, lately, before now,
since noon
(25 ) a. John has been on the train today.

b. John was on the train today.
(26 ) a. *John has been on the train yesterday.
b. John was on the train yesterday.
(27) a. John has been on the train since noon.
b. *John was on the train since noon.
To see how the extended now theory accounts for these facts,
one must consider the semantic distinction between the three
sorts of adverbs listed in (22)-(24), and the way in which
adverbs enter into the interpretation of perfect sentences.
The adverbs in (22)-(24) are regarded as expressions of
the category MTA of main tense adverbs, and thus as denoting
functions from properties of time intervals to sets of time
intervals. They do, however, have three distinct sorts of
denotations. At a given index <w,i>, the adverbs in (22) in
general denote functions from properties to sets containing
both past intervals and extended nows relative to i; for
example, the denotation of today is that of (28), as has
already been seen (see (59a) in Chapter III, section '.3.2).
The adverbs in (23), on the other hand, denote functions

from properties to sets of past intervals; yesterday, for
instance, has the denotation of (29) (see (59b) in Chapter
III, section '.3.2), where today' is (30).
(29) APt At[Vt 1 [[daY'(t 1 ) & At 2 [todaY'(t 2 ) -+ [t1 < t2 &

At 3 [[t, < t3 & t3 < t 2 ] -+ toda Y'(t 3 )]]]] &
t =- t,] & pt{t}]
(30) AtVt,[day'(t,) & [NOW =- t, & t =- t,]]

Finally, the adverbs in (24) all denote functions from
properties to sets of extended nows; since noon, for
instance, translates as (3').
(31) AptAt[[][n(t) & vt,[[noon'(t,) & Vt 2 [day'(t 2 ) &

[t, =- t2 & now =- t 2 ]]] & t, < t]] & pt{t}]
Now, recall from Chapter III that main tense adverbs

join with tense to characterize a single interval. Thus,
the preterit sentence (32) is true at interval i iff there
is some past interval relative to i at which ('0) is true
and which is in the denotation of (33) at i.
(32) John was on the train yesterday.

('0) John is on the train.
(33) Atvt, [[daY'(t,) & At 2 [today'(t 2 ) .... [t, < t2 &

At 3 [[t, < t3 & t3 < t 2 ] .... todaY'(t 3 )]]]] &
t =- t,]
These truthconditions are captured by the translation (34)
for (32) (where yesterday" is (29»:
(34) vt[yesterday" ("At, [past(t,) &
AT(t" John-is-on-the-train')])(t)]
The case of the perfect is analogous under the extended now

theory: main tense adverbs join with the present perfect
tense to characterize a single interval. Thus, the present
perfect sentence (12a) is true at interval i iff there is an
extended now relative to i which is in the denotation of
(35) at i and during which ('0) is true. (NB: I t is the
extended now, and not its subinterval, which must be in the
denotation of the main tense adverb.)
( '2a) John has been on the train since noon.
At[xn(t) & vt, [[noon'(t,) & Vt 2 [daY'(t 2 ) &
[t, =- t2 & now =- t 2 ]]] & t1 < t]]

These truthconditions are embodied in the translation (36)
for ('2a) (where since-noon" is (31».
(36) Vt[since-noon"("At, [xn(t,) & vt 2 [t 2 =- t, &

AT(t 2 , John-is-on-the-train')] ])(t)]
If perfects and preterits are interpreted in this way,

then the unacceptable sentences (37) and (38) turn out to be
contradictions--in both sentences, the tense and the time
adverb give rise to conflicting entailments.
228 CHAPTER IV
(37) *John was on the train since noon.

(38) *John has been on the train yesterday.
For (37) to be true at interval i, there would have to be

some past interval relative to i that is in the denotation
of (35) at i; but there can be no such interval, since every
member of this set is an extended now relative to i.
Likewise, if (38) were true at interval i, then there would
have to be some extended now relative to i that is a member
of the set denoted by (33) at i; but this is impossible,
since members of the latter set will always be separated
from i. Note that sentences (39) and (40) are, on the other
hand, both acceptable, because the denotation of (30) at
interval i contains both past intervals and extended nows
relati ve to i.
(39) John has been on the train today.

(40) John was on the train today.
As these few examples suggest, the extended now theory

of the perfect affords a subtle and intuitive account of how
time adverbs join with tenses in English. (For a much more
detailed account along these lines, see Dowty (1979:339fO.)
In this section, I shall argue that the central claim
of the extended now theory--that the function of the perfect
is to locate an event within an extended now--is wrong; an
alternative conception of the perfect will be proposed. The
latter can be most easily grasped if a new piece of
terminology is introduced: any interval which begins
earlier than some interval i and lasts no later than i is to
be a perfect interval relative to i, as in (41).
(41) i 1, i Z' i3 are perfect intervals relative to i.
i
~
I I I I I )
My claim is that the real function of the perfect is to

locate an event within a perfect interval; that is, I shall
argue that the semantics of the perfect is such that (15) is
true at interval i iff (16) is true sometime during a
perfect interval relative to i, as in any of the situations

represented in (42). (This analysis predicts exactly the
same truthconditions as the extended now analysis for
perfect sentences without main tense adverbs; when such
adverbs are present, however, they make crucially distinct
predictions, as I shall show immediately below.)

(42) (15) is true at i iff (16) is true at a

subinterval of i'.
i
~
I I I
i
~
I I
I I
~
i'
[i' is a perfect interval relative to i.J
This claim--that the function of the perfect is to locate an

event within a perfect interval--would appear to be
immediately contradicted by the anomaly of (38); but I shall
argue that this is a pragmatic anomaly, not a semantic one,
as the extended now theory implies.
Consider sentence (7) again.
(7) Having been on the train, John knows exactly why

it derailed.
The free adjunct in this sentence is headed by the have of

the perfect. The extended now theory predicts that adjuncts
of this sort will contain time adverbs like since noon, but
not those like yesterday, which, it is argued, are
semantically incompatible with the perfect. Yet, as was
230 CHAPTER IV
observed above, perfect adjuncts may contain adverbs of

either sort:

There is, however, an intuitive difference between (11a) and

(11b): if these are paraphrased with sentences (43) and
(44), which have full adverbial clauses instead of free
adjuncts, it can be seen that while the perfect construction
in sentence (11a) functions essentially like a present
perfect, that in (11b) has the function of a preterit.
(43) Because he has been on the train since noon, John

(Cf. *Because he was on the train since noon, John
knows exactly why it derailed.)
(44) Because he was on the train yesterday, John knows
exactly why it derailed.
(Cf. *Because he has been on the train yesterday,
John knows exactly why it derailed.)
The use of the perfect construction in a preterit role

is in fact very widespread. It shows up in every nonfinite
structure that admits the perfect; in the structures in
(45), for example, the underlined phrases can function as
present perfects or as preterits, depending on the
accompanying adverb.
(45) Marked infinitive complements:
Bill seems to have slept yesterday/since

noon.
Unmarked infinitive complements:
Mary may have played the piano yesterday/

since noon.
Adverbial infinitives:
To have done the entire job yesterday/ since

noon, John must have had help.
(45) [continued]
Gerunds:
In spite of having studied yesterday/since

noon, John isn't confident of passing.
His having been in New York yesterday/since

noon is hard to explain-.--
Participial absolutes:
His father having driven yesterday/since

noon, John wants to take the wheel.
'Reduced relative clauses':
Anyone having spoken with Anne yesterday/

since noon should contact the police.
Furthermore, the perfect construction may serve as a

preterit in the past and future perfects. These compound
tenses are normally analyzed as the preterit or future of a
present perfect; and this analysis would appear to be
correct in the case of sentences like (46) and (47).
(46) John had worked on the problem since noon.

(47) John will have worked on the problem since noon.
But the acceptability of sentences like (48) and (49)

compels one to admit that the past and future perfect
constructions may express the preterit or future of a
preterit;3 otherwise, sentences (48) and (49) would be fully
as anomalous as (50).
(48) John had worked on the problem the day before.

(49) John will have worked on the problem the day
before.
(50) *John has worked on the problem yesterday.
It appears, then, that the perfect construction can

assume a preterit function in any nonfinite structure in
which it may appear, as well as in the preteri t and future
tenses--in other words, everywhere but in the present tense
(when there is no modal present). But it is just the
232 CHAPTER IV
present perfect that is actually in competition with the

preterit--there is nothing comparable to the preterit with
which the perfect may contrast in nonfinite constructions or
in the preteri t or future tenses. This is a cri tical fact
in the analysis which I now propose for the perfect.
In my analysis, the preterit tense is assumed to locate
an event at some past interval--exactly as in the extended
now theory. The perfect, however, receives a much broader
interpretation: a perfect sentence ~ without a time adverb
is true at interval i iff the tenseless form ~ of ~ is true
sometime during a perfect interval relative to i; a perfect
sentence ~ with a time adverb a is true at i iff there is
some perfect interval which is in the set of intervals
determined by a and during which ~ is at some time true.
Thus, sentence (15) is true at i iff (16) is true sometime
during a perfect interval relative to i;

and sentence (12a) is true at i iff there is some perfect

interval relative to i which is a member of the denotation
of (35) at i and during which (10) is at some time true.
(12a) John has been on the train since noon.
This proposal can be formally elucidated by introducing a

new intensional logic predicate 'perf'a)': where ~ denotes
a time interval i', let perf(O be true at index <w,D iff
i' begins before i and lasts no later than i--that is, let
perf(t) have the denotation of (51).
(51 ) [ vt 1 [ t 1 c t &t1 < nov] & -, vt 1[t 1 c t &

nov < t 1 ]]
The proposed interpretation for (15) and (12a) can now be

induced by the respective translations (52) and (53).
(52) Vt[perf(t) & Vt 1 [t 1 ~ t &

AT(t 1 , John-sees-Mary')]]
(53) Vt[since-noon"("'A t1 [perr(t 1) & Vt 2 [t 2 ~ t, &

AT(t 2 , John-is-on-the-train')] ])(t)]
Because a perfect interval may be either a past interval or

an extended now, my analysis predicts that perfects will
occur acceptably both with adverbs like since noon and with
those like yesterday; this prediction is borne out in the
vast majority of cases, as has been seen--everywhere but in
the present tense. The present perfect is distinguished
because it directly competes with the preterit; this, I
claim, is the source of the unacceptability of sentences
like (38).
(38) *John has been on the train yesterday.
In his article 'Conversational Implicature and the

Lexicon', McCawley (1978) argues that 'what is
conversationally implicated by an utterance depends not only
on the utterance but on what other utterances the speaker
could have produced but did not' (p.245); he discusses
several cases 'in which an utterance conversationally
implicates something by virtue of its "taking more effort"
or "taking the speaker further out of his way" than some
alternative utterance.' For example, he points out that
though one can say something is pale green, pale blue, or
pale yellow, it is odd to say that something is pale red,
owing to the availability of the adjective pink;
consequently, when someone does say that something is pale
red, s/he conversationally implicates that pink for some
reason does not apply--that slhe is talking about some color
intermediate between red and pink, even though this may be
much less pale than colors like pale green or pale blue. My
claim here is that the case of the present perfect is in
some ways analogous to this and other examples discussed by
McCawley.
Someone using a present perfect sentence like (15)
conversationally implicates that s/he has a reason for not
using the corresponding preterit (18), namely that John's
seeing Mary happened during some extended now which is
especially salient.

(18) John saw Nary.
This implicature in effect eliminates the overlap between

the truthconditions of the perfeot and those of the
preteri t, so that, for pragmatic reasons, the present
perfeot seems to have essentially the truthoonditions
234 CHAPTER IV
attributed to it by the extended now theory. This analysis

implies that the perfect is somehow more 'marked' or 'takes
the speaker further out of his way' than the preterit. This
seems to be true, whether one considers the matter from a
syntactic or a semantic point of view. The perfect, a
periphrastic rather than merely inflectional tense, is
structurally more complex than the preterit. And
semantically, the job of the perfect is to locate an event
somewhere within a given interval, while the preterit always
simply locates an event at some interval. The perfect is
clearly the 'marked case' in both respects.
In accordance with this analysis, sentence (54a) is not
regarded as semantically anomalous (note that its
translation (55a) is not contradictory), but rather as
pragmatically anomalous.
(54) a. *John has seen Mary yesterday.

b. John saw Mary yesterday.
The use of (54a) conversationally implicates that (54b) for

some reason does not apply (just as (15) implicates that
(18) does not apply); yet, (54a) and (54b) (unlike (15) and
(18» are logically equivalent, as their respective
translations show.
(55) a. vt[yesterday"("At 1 [perr(t 1 ) & vt 2 [t 2 ~ t1 &

AT(t 2 , John-sees-Hary')]])(t)]
b. vt[yesterday,,(AAt 1 [past(t 1 ) &
AT(t l' John-sees-Hary')] )(t)]
Because the adverb yesterday unequivocally locates the event

of John's seeing Mary at a past interval, there would
normally be no motive for choosing the perfect over the
preterit in (54) (but see below). Thus, (54a) is
pragmatically anomalous because it conversationally
implicates that (54 b) does not apply, but at the same time
entails (S4b).4
It might be objected that the unacceptabili ty of (54a)
is simply too blatant to be pragmatic. Sentence (54a),
however, is actually acceptable in some contexts, as, for
example, in (56).
(56) So far, John has seen Mary yesterday, Bill has
seen her this morning,
Here, the perfect is accompanied by two adverbs--yesterday

and so far. So far, like since noon, determines a set of
extended nows; for this reason, it licenses the appearance
of the perfect in this sentence. (Note that sentences like
(56) are explicit contradictions in the extended now
theory.)
Nevertheless, I concede that the distinction between
finite present perfects and all other kinds of perfects may
have become 'grammaticized'--that is, this distinction may
no longer be merely pragmatic in status, but may have come
to be encoded in the rules of English semantics. If so,
then the problem posed by the perfect for stricter versions
of the compositionality thesis is very real: under the
assumption that grammaticization has taken place, the
perfect must have both an 'extended now' interpretation and
an interpretation of the sort I have advocated, the former
in present perfects, the latter elsewhere; if such is the
case, then it cannot be maintained that the perfect tenses
resul t uniformly from the compositional union of a Simple
tense operator with the perfect operator. There are ways of
softening the metatheoretical blow delivered by this
prospect; none of them are very elegant, though. Here, I
shall leave the question of grammaticization open: this is a
matter independent of the point which I hope to have
established, namely that the distinction between the present
perfect and other perfects is pragmatic in origin.
In the extended now theory, the perfect is simply
equated with the present perfect: yet, as we have seen, it
is precisely the present perfect that, for pragmatic
reasons, provides the least inSight into the true nature of
the perfect. As a consequence, the extended now theory
provides no account whatsoever of the use of the perfect in
a preterit role in nonpresent tense constructions. (Under
the extended now theory, the perfect would apparently have
to have two distinct sets of truthconditions, one for
present perfects, the other for other sorts of perfects, as
in the 'grammaticized' version of the theory advocated
here.) My analysis, on the other hand, allows all perfects
to be uniformly interpreted, whether they are present, past,
future, or nonfinite; furthermore, it recognizes the
pragmatically exceptional nature of the present perfect.
These are its principal advantages.
236 CHAPTER IV
In the following two subsections, a formal semantics

for the perfect based on the theory advocated here will be
incorporated into the developing fragment for free adjuncts
and absolutes.
1.2. THE FORMAL SEMANTICS OF THE PERFECT IN FINITE CLAUSES
In the formal fragment for the semantics of free adjuncts

and absolutes, perfect have is regarded as the unique
expression of the category of perfect auxiliaries:
TAB//TAB the category of {[have] }

perfect auxiliaries
Its translation is (57), where the interpretation of

'perf(t)' is as in (51) above.
As its category definition implies, perfect have joins

with a temporal abstract to form a new temporal abstract.
Specifically, have takes a tenseless temporal abstract in
the subclass [-Perfect] to form a tenseless abstract in the
subclass [+Perfect]:
(58) S62. I f a e: PTAB / /TAB and S e: PTAB , [-Tense,

-Adjunct, -Perfect], then F 62 (ex, S) E: PTAB ,
[-Tense, -Adjunct, +Perfect], where F 62 (a,S)
is PRB(a,S).
Example: F 62 ([haveJ, (John) [is] on the train) =

(John) [has] been on the train
T62. If a E: PTAB//TAB' B E: PTAB , and a,B translate

as a',B', then F 62 (a,B) translates as
U,(A S ').
(58) [continued]
Example: F 62 ([have], (John) [is] on the train)

translates as At[perf(t) & Vt,[t, c t &
AT(t,. Vxs[R(x s , John') &
on-the-train' (x s )])]]
One might suppose that the past, present, and future
perfects result from the application of one of the rules of
tense to a temporal abstract created by rule (58). This is
incorrect, however; for example, if the future tense rule
(as well as the rule converting tensed temporal abstracts to
tensed sentences) were to apply to the tenseless temporal
abstract (John) [has] been on the train, the resulting
expression (59a) would be assigned the translation (59b).
a. (John) [will] have been on the train

b. Vt[FUT(t) & [perf(t) & Vt,[t, c t &
AT(t" Vxs[R(x s , John') &
on-the-train' (x s )])]]]
(59b) is clearly an undesirable interpretation for (59a):
since no interval can satisfy both 'FUT(O' and 'perf(O',
(59b) predicts that (59a) should be an outright
contradiction. What's wrong with (59b), obviously, is that
it fails to capture the scope relationship between the
future and the perfect in (59a); thus, a better translation
for (59a) would be (60), in which the future tense is
interpreted as containing the perfect within its scope.
(60) Vt[FUT(t) & AT(t, vt,[perf(t,) & Vt 2 [t 2 ~ t, &

AT(t 2 , Vxs[R(x s , John') & on-the-train'(x s )])]])]
To guarantee the assignment of translations such as

(60), rule (6') has been included in the fragment:
(61) S70. If a E PTAB , [-Tense, BAdjunct, yWeak,

+Perfect], then F 70 (a) e: PTAB , [-Tense,
BAdjunct, yWeak, -Perfect], where F70 (a) is
a.
238 CHAPTER IV
(6') [continued]
Example: F70 «John) [has] been on the train) = (John)

[has] been on the train
T70. If a. f PTAB and a. translates as a.', then

F70 (a.) translates as At[AT(t, vt,[a.'(t,)])].
Example: F 70 «John) [has] been on the train)

translates as At[AT(t, Vt,[perf(t,) &
Vt 2 [t 2 ~ t, & AT(t 2 , Vxs[R(x s , John') &
on-the-train' (x s )])]])]
By means of rule (61), the sentence John will have been

on the train can now be assigned the analysis (62), which
induces the desired translation (60).
(62) (John) [will] have been on the train, t, 60

I
(John) [will] have been on the train, TAB, 59
I
----------
(John) [has] been on the train, TAB, 70
I
[have], TABllTAB (John) [ is] on the train, TAB, 56

I
(John) [is] on the train, t, 19
Central to such derivations is the feature [+Perfect].

[+Perfect] is that subclass of temporal abstracts generated
by rule (58); all such abstracts denote sets of perfect
intervals, in the sense defined. Since the direct
application of a rule of tense to a temporal abstract in
this subclass leads to the difficulties noted above, the
rules of tense are stated so as to apply only to temporal
abstracts in the subclass [-Perfect]. Thus, in order for an
abstract in the [+Perfect] subclass to become tensed, it
must first undergo rule (61), which transfers abstracts from
the [+Perfect] subclass to the [-Perfect] subclass.
[+Perfect] should not be thought of as the subclass of
temporal abstracts headed by the perfect auxiliary have,
since members of [-Perfect] may likewise be headed by this
auxiliary, provided they have undergone rule (61); instead,

the subclassification associated with the feature [+Perfect]
is best viewed as simply a means of guaranteeing the proper
sequence of rule applications in the derivation of compound
tenses. 5
One might ask whether the feature [+Perfect] could be
dispensed with by collapsing (58) and (61) into a single
rule. The answer is no, because applications of the rule
S61 adding main tense adverbs to temporal abstracts must be
able to intervene between applications of (58) and (61).
Such intervening applications of rule S61 appear in the
derivations (63) and (64) assigned to the sentences (12a)
and (12b), respectively.
(12) a. John has been on the train since noon.

(63) (John) [has] been on the train since noon, t, 60

I
(John) [has] been on the train since noon, TAB, 58
I
(John) [has] been on the train since noon, TAB, 70
----------
I
(John) [has] been on the train since noon, TAB, 61b
.
s~nce noon, MTA (John) [has] been on the train, TAB, 62
[have], TAB/ITAB (John) [is] on the train, TAB, 56

I
(64 ) (John) [has] been on the train yesterday, t, 60

I
(John) [has] been on the train yesterday, TAB, 58
I
----------------
(John) [has] been on the train yesterday. TAB. 70
I
(John) [has] been on the train yesterday. TAB, 61b
yesterday, (John) [has] been on the tra1n,

. TAB, 62
MTA, 63
I [have], (John) [is] on the train, TAB, 56
yesterday. TAB/ITAB I
TA (John) [is] on the train, t, 19
240 CHAPTER IV
(63) and (64) induce (65) and (66) as translations for

(12a) and (12b), respectively.
(65) Vt[PRES(t) & AT(t, Vt1[since-noon"(A~t2[perf(t2) &

vt 3 [t 3 =- t2 & AT(t 3 , Vxs[R(x s , John') &
on-the-train' (x s )])]])(t 1 )])]
(66) Vt[PRES(t) & AT(t, vt 1 [yesterday"(AAt 2 [perf(t 2 ) &
Vt 3 [t 3 =- t2 & AT(t 3, Vxs[R(x s , John') &

on-the-train'(x s )])]])(t 1 )])]
These translations are worth considering in detail, since

they clearly embody the essential features of the proposed
analysis of the perfect. Note, for example, that
translation (66) doesn't induce a contradictory
interpretation for (12b), because the set of perfect
intervals and the denotation of the set-level adverb
yesterday will always intersect (at an index of possible
utterance); this is consistent with the claim that the
anomaly of (12b) is pragmatic rather than semantic in
nature. Observe, however, that sentence (37), analyzed as
in (67), is assigned a contradictory interpretation by
translation (68):
(37) *John was on the train since noon.
(John) [was] on the train since noon, t, 60
----------
I
(John) [was] on the train since noon, TAB, 61b
since noon, MTA (John) [was] on the train, TAB, 57

I
(John) [is] on the train, TAB, 56
I
(68) Vt[since-noon,,(A),. t, [past(t,) &

on-the-train'(xs)])])(t)]
Past perfect and future perfect sentences with main

tense adverbs of various sorts are easily derived in this
framework. For example, sentences (69) and (70) may be

deri ved and translated as in (71)-(74).
(69) John had been on the train since noon.

(70) John will have been on the train today.
(71) (John) [had] been on the train since noon, t, 60

I
(John) [had] been on the train since noon, TAB, 61b
sinCe~had] been on the train, TAB, 57

MTA I
---------
I
[have], TAB/ITAB (John) [is] on the train, TAB, 56

I
(72) (John) [will] have been on the train today, t, 60
----------
I
(John) [will] have been on the train today, TAB, 61b
today, (John) [will] have been on the train, TAB, 59

MTA, 63 I
----------
I (John) [has] been on the train, TAB, 70
today, I
TA (John) [has] been on the train, TAB, 62
[have], TAB//TAB (John) [is] on the train, TAB, 56

I
(73) vt[past(t) AT(t, vt 1 [since-noon"( . . At 2 [per:f(t 2 )

& &
Vt 3 [t 3 ~ t2 & AT(t 3 , VXs[R(x s , John') &

on-the-train'(x S )])]])(t 1 )])]
(74) Vt[FUT(t) AT(t, vt, [today"( . . At 2 [perf(t 2 )
& &
Vt 3 [t 3 =- t2 & AT(t3' Vxs[R(x s , John') &

on-the-train'(x s )])]])(t 1 )])]
In accordance with the analysis which I have proposed, the

sentences (75) and (76) are assigned noncontradictory
- 242 CHAPTER IV
interpretations, namely those induced by (77) and (78)

(where the-day-before" is (79)).
(75) John had been on the train the day before.

(76) John will have been on the train the day before.
(77) Vt[past(t) &
AT(t, Vt1 [the-day-before,,(AAt 2 [perf(t 2 ) &
Vt 3 [t 3 =- t2 & AT(t 3 , Vxs[R(x s , John') &

on-the-train'(x s )])]])(t 1 )])]
(78) Vt[FUT(t) &

AT(t, Vt1 [the-day-before,,(AAt 2 [perf(t 2 ) &
vt 3 [t 3 =- t2 & AT(t3' Vxs[R(x s , John') &

on-the-train'(x s )])]])(t 1 )])]
(79) Ap t At 1 [vt 2 [[daY'(t 2 ) & Vt 3 [[day'(t 3 ) & nov c t 3 ] &
At 5 [t 5 =- t3 + [t 2 < t5 & At 6 [[t 2 < t6 &

t6 < t S] + t6 =- t 3 ]]]]] & t1 =- t 2 ] & P {t 1 }]
t
Recall that the extended now analysis founders on examples

such as (75) and (76).
1.3. THE FORMAL SEMANTICS OF THE PERFECT IN FREE ADJUNCTS
The derivation of perfect adjuncts involves rule (61);

two additional rules are also necessary, however. The first
of these combines have with a stage-level intransitive verb
phrase to produce an A-abstract in the subclass [+Weak,
+Perfect]:
(80) S3. If U E PTABIITAB and 8 Prvs, then F3 ,n(u,8)

E
E PTAB , [-Tense, +Adjunct, +Weak, +Perfect],

where F3 ,n(cx,8) is
DB(ING(CON(cx, DB(EN(8»))).
Example: F 3 ,4([have], [walk] home) = having walked

home
(80) [continued]
T3. If a E PTABIITAB' S E PIVs, and a,6 translate

as a',/3',
then F 3 ,n(a,/3) ~ranslates as
a,(AAt[AT(t, Vxs[R(x s , x~) & S'(x s )])]).
Example: F 3 ,4([have], [walk] home) translates as

At[perf(t) & vt1 [t 1 =- t &
AT(t 1 , Vxs[R(x s , x~) & valk-home'(x s )])]]
The second rule for the formation of perfect adjuncts

combines have with an indi vidual-level intransi ti ve verb
phrase to produce an A-abstract in the subclass [-Weak,
+perfectJ: 6
(81) S4. If a E PTAB//TAB and 6 E PIVi, then F4 ,n(a,/3)

E PTAB , [-Tense, +Adjunct, -Weak, +PerfectJ,
where F 4 ,n(a,6) is
DB(ING(CON(a, DB(EN(6»»).
Example: F 4 ,4([have], [be] a sailor) = having been a

sailor
T4. If a E PTABIITAB' 6 E PIvi, and a,/3 translate

as a',/3',then F 4 ,n(a,/3) translates as
areA At[AT(t, /3'( xg» J),
a,(AAt[AT(t, /3'(X:»]), or
a,(A)..t[AT(t, 6'(X~»]), according as 6' is of
sorted type (o,t'>, (k,t'>, or ([o,k] ,t'>,
respectively.
Example: F 4 ,4([haveJ, [be] a sailor) translates as

At[perf(t) & vt 1 [t 1 =- t &
AT(tl' be-a-sai1or'(x~»]]
One might suppose that the A-abstracts generated by
rules (80) and (81) can immediately undergo the rules S9,
310, and 311 responsible for converting A-abstracts into
244 CHAPTER IV
free adjuncts of categories TA, t', and HTA, respectively.

This is not the case, however. To see why, consider
sentence (82).
(82) Having been on the train, John will know exactly

why it derailed.
As an A-abstract, the free adjunct in (82) is assigned the

translation (83).
i
At[perf(t) & vt,[t, ~ t & AT(t" Vxs[RCx s , x4) &
on-the-train' (x s »)))
If this A-abstract were to be directly converted into an

adjunct of category MTA (by rule S,,), sentence (82) would
be assigned the translation (84):
(84) the-train'("Ax~[vt[K(L6)("vt1 [M 9Ct,t 1 ) &

[perf(t 1) & vt 2 [t 2 .=
t1 &
ATCt 2 , Vxs[R(x s , x~) &
on - i t 2' ( x s) ) ) ] ) ) (A [F UT (t) &
AT(t, JOhn-knovs-exactly-vhy-it 2-derailed'»))))
This translation for (82) isn't adequate, however:

according to (84), the adjunct in (82) necessarily relates
to a nonfuture state of affairs, i.e. one occupying a time
interval which is perfect relative to the speaker's
interval. Yet, it seems perfectly possible to interpret
(82) as meaning: 'because he will have been on the train,
John will know exactly why it derailed.' Accordingly, a
translation such as (85) is preferable to (84);
(85) the-t r a in' ("A xg[vt [K(L6)("vt 1 [M 9 (t, t 1) &

AT(t" vt 2 [perf(t 2 ) & vt 3 [t 3 ~ t2 &
AT(t 3 , VxS[R(x S , x!) &
on-it2 'Cx s »))))))("[FUT(t) &
AT (t , J ohn-know~xactly-"hy-i t 2 -derailed') ] ) ] ])
(55) is perfectly consistent with an interpretation in which

both the free adjunct and the main clause in sentence (82)
relate to future states of affairs.
Translations such as (85) can be straightforwardly

assigned if it is required that the perfect A-abstracts
generated by rules (80) and (81) undergo rule (61) before
being converted into free adjuncts of categories TA, t l , and
MTA by rules S9, S10, and S11, respectively. Accordingly,
the latter rules have been stated so as to apply only to A-
abstracts in the subclass [-Perfect].
Given these assumptions, sentence (82) can be derived
as follows:
(86) having been on the train, (John) [will] know

exactly why it derailed, t, 55,2
the t~ been on it2' (John) [will] know

exactly why it2 derailed, t, 60
------
I
having been on it2' (John) [will] know
exactly whyit2 derailed, TAB, 61a
having been on it2' (John) [will] know exactly why it2

MTA, 11,6,9 derailed, TAB, 59
I I
having been on it2' (John) [knows] exactly why it2
TAB, 70 derailed, TAB, 56
--------
I I
having been on it2' (John) [knows] exactly
TAB, 4,4 derailed, t, 19
[have], TAB//TAB [be] on it2'
(86) induces (85) as a translation for (82), as desired.

Sentences (11a,b) are the data which originally
motivated the proposed reanalysis of the perfect:

Recall that the extended now analysis of the perfect wrongly

predicts that sentence (11b) should be contradictory. Under
the present analysis, (11a,b) are derived as in (87) and
(88), and are thereby aSSigned the respective translations
(89) and (90):
246 CHAPTER IV
---------
having been on the train since noon,
(John) [knows] exactly why it derailed, t, 55,2
the train, having been on it2 since noon, (John) [knows]

T exactly why it2 derailed, t, 60
I
having been on it2 since noon,
(John) [knows] exactly why it2 derailed, TAB, 61a
having been on it2 Si~hn) [knows] exactly

I
MTA, 11,6,9 why jt2 derailed, TAB, 58
having been on it2 since noon, (John) [knows] exactly

I TAB, 70 why i t 2 derailed, TAB, 56
1
--------
having been on it2 since noon, (John) [knows] exactly
~TAB, 61b why it2 derailed, t, 19
since noon, MTA having been on it2 , TAB, 4,4
[have], TABIITAB [be] on it 2 , IV~.
----------
(88) having been on the train yesterday,
(John) [knows] exactly why it derailed, t, 55,2
the train, having been on it2 yesterday, (John) [knows]

T exactly why it2 derailed, t, 60
I
having been on it2 yesterday,
(John) [knows] exactly why it2 derailed, TAB, 61a
having been on it2 y~n) [knows] exactly

I
MTA, 11,6,9 why it2 Iderailed, TAB, 58
having been on it z yesterday, (John) [knows] exactly
I TAB, 70 why it2 Iderailed, TAB, 56
having been on it2 yesterday, (John) [knows] exactly
~' 61b why it2 derailed, t, 19
yesterday, MTA, 63 having been on it 2 , TAB, 4,4
I _________ .
yesterday, TA [have], TABIITAB [be] on it2' IV~
(89) the-train'("Ax~[vt[K(L6)("vt1 [M 9 (t,t 1 ) &

AT(t" Vt 2 [since-noon"(">" t 3 [perf,(t 3 ) &
Vt 4 [ t4::' t3 & AT ( T4 , Vx s[ R(xs ,x4~) &

on-i t 2 '(x s )])] ])(t 2 )])])("[PRES(t) &
AT(t, John-knows-exactly-why-it 2-derailed')])]])
(90) the-train'("Ax~[Vt[K(L6)("Vt1[M9(t,t1) &

AT(t 1, vt2[yesterdaY"("H3[perf(~3) &
t4::' t3 & AT ( T 4 , Vx s[ R(xs ,x ~)

Vt 4 [ 4 &
on-it 2 '(x s )])]])(t 2 )])])("[PRES(t) &
AT (t, J ohn-knows-exactly-why-i t 2-derailed') ]) ] ])
As is desired, neither of these translations induces a

contradictory interpretation: the interpretation of
'perf(~)' allows interpretations for perfect adjuncts which
are like those of a (pragmatically acceptable) present
perfect, as in (11a), or like those of preterits, as in
(llb); the accompanying adverb is, in both cases, the factor
determining which sort of interpretation is assigned.
In this section, I have argued for a new semantic
analysis of the perfect, according to which the function of
the perfect is to locate an event within a perfect interval;
unlike the extended now theory, this theory is compatible
with the use of the perfect in free adjuncts, as I have
shown. In the following section, I shall demonstrate that
this theory provides a compelling argument in favor of the
view that free adjuncts may serve as main tense adverbs.
2. An Argument for Free Adjuncts as Main Tense Adverbs
In view of the foregoing analysis of the perfect, it is

clear that main tense adverbs (i.e. time adverbs of category
MTA) can be employed in two distinct ways. On the one hand,
they may join with the past, present, and future tenses to
characterize the interval occupied by some event or state of
affairs. For example, in sentence (91), both the main tense
adverb and the tense help identify the interval at which
John left: the adverb serves to locate John's departure
within the present day, and the tense further locates it
within the past.
248 CHAPTER IV
(91) John left today.
The fact that the time adverb and the tense in (91)
characterize the same interval is reflected in the
translation assigned to (91) by the proposed fragment:
(92) Vt[today'(t) & [past(t) & AT(t, John-leaves')]]
Main tense adverbs don't only join with the past,

present, and future tenses, however; they may also join with
the perfect. Consider, for example, sentence (93).
(93) John had left since noon.
Two intervals are relevant to the interpretation of this

sentence: a past interval i at which (94) is true, and a
perfect interval i' relative to i, during which (95) is at
some time true.
(94) John has left.

(95) John leaves.
The main tense adverb since noon in (93) joins with the
perfect to characterize the interval i'; in particular, it
specifies a set of extended nows (relative to i) having i'
as a member. (Because it does denote a set of extended
nows, this particular adverb cannot join with the past tense
to characterize i; if it did, a contradiction would result.)
The fact that the time adverb in (93) joins with the perfect
to characterize a particular interval is captured in the
translation assigned to (93) (where, again, since-noon" is
(31» :
(96) Vt[past(t) & AT(t, Vt 1[since-noon,,(AAt 2 [perf'(t 2 ) &

vt 3 [t 3 ~ t2 & AT(t 3 , VXs[R(X s , John') &
leave'(x s )])]])(t 1 )])]
(31) AptAt[[xn(t) &vt1 [[noon'(t 1 ) & vt 2 [day'(t 2 ) &
[t 1 ~ t2 & now ~ t 2 ]]] & t1 < t]] & pt{t}]
Note that when a time adverb joins with the perfect, it

is within the scope of tense. In (93), for example, the
denotation of since noon is not evaluated relative to the
speaker's interval; that is, this adverb is not understood
to specify the set of extended nows beginning after noon on

the day of (93)'s utterance. Rather, it is evaluated
relative to the past interval i picked out by the past tense
in (93). On the other hand, a time adverb joining with a
simple tense to pick out some interval is always interpreted
as being outside the scope of tense, as in (92).
Not surprisingly, there are sentences whose ambiguity
turns on the two alternative uses for time adverbs
exemplified in (91) and (93). Sentence (97), for example,
can mean either that Harry wrote three novels during the
year 1978 (the perfect + time adverb interpretation) or that
as of 1978, Harry had written three novels, though not
necessarily in that year (the past tense + time adverb
interpretation).
(97) Harry had written three novels in 1978.
It is clear that this ambiguity depends upon the possibility

that a main tense adverb may join with the perfect rather
than with a simple tense to characterize some interval;
note, for example, that the simple past tense sentence (98)
corresponding to (97) is not similarly ambiguous.
(98) Harry wrote three novels in 1978.
In view of these facts, the claim that free adjuncts

may serve as main tense adverbs can be evaluated according
to a rather subtle test: if free adjuncts are in fact main
tense adverbs, then (a) in sentences in the past perfect,
present perfect, or future perfect, a free adjunct should be
interpretable in two different ways, according as it is
understood to join with the perfect or with the past,
present, or future tense; moreover, in the former but not
the latter case, it should be interpreted as being within
the scope of tense; on the other hand, (b) in sentences in
the simple past, present, or future, a free adjunct should
not have a pair of interpretations comparable to those in
(a), but should have only a single possible interpretation--
specifically, one according to which it is outside the scope
of tense.
As I shall now demonstrate, free adjuncts clearly pass
both of these tests, making it highly plausible that they
may serve as main tense adverbs (when not restricting a
binary operator in the main clause, such as a modal, a
relative frequency adverb, or the generalization operator
Gil).
250 CHAPTER IV
Consider the past perfect sentence (99).

This sentence is at least two ways ambiguous: on the one
hand, (99) can be understood to entail that Mary worked at
the post office during the most recent summer relative to
the speaker's interval; on the other hand, it can be felt to
entail that she worked at the post office during the most
recent summer relative to some past interval. Under the
former interpretation, (99) would not be falsified if Mary
had made her decision before she ever worked at the post
office--this interpretation can be singled out by inserting
already after had in (99); under the latter of the two
interpretations, Mary must have made her decision while
employed at the post office. This is clearly an ambiguity
in the role of the adjunct in (99). On the former
interpretation, the adjunct is understood to join with past
tense to characterize a single past interval; in this case,
(99) is to be analyzed and translated as in (100) and (101).
On the latter interpretation, however, the adjunct is
understood to join with the perfect to characterize a single
perfect interval; in this instance, (99) is derived as in
(102), and thus receives the translation (103).
(100 )
working at the post office during the past summer,
------
(Mary) [had] decided to go back to 3chool, t, 60
I
working at the P03t office during the past summer,
(Mary) [had] decided to go back to school, TAB, 61a
working at the post office (Hary) [had] decided

during the pa3t summer, MTA, 11,6,9 to go back to school, TAB, 57
I I
working at the post office (Hary) [has] decided
during the past summer, TAB, 61b to go back to school, TAB, 70
~ I
during the past working at the post (Mary) [has] decided
summer. MTA. 63 office. TAB. 1.4 to go back to school, TAB, 62
I I ______________
during the working at the [have], (Mary) [decides] to go back
pa3t 3UDJDer, post office, TAB/IIAB to school, TAB, 56
TA PRPL S I
(Hary) [decides] to go back to school, t, 19
(101 ) Vt[K(L 6 )("Vt 1 [M g (t,t 1 ) &

[during-the-past-summer'(t 1 ) &
AT(t 1 , Vxs[R(x S , x~) &
vorking-at-the-post-office'(x S)])] ])(A[past(t) &
AT(t, Vt 1 [perf(t 1 ) & Vt 2 [t 2 ~ t1 &
AT(t 2 , Mary-decides-to-go-back-to-school')]])])]
(102 ) working at the post office during the past summer,

(Mary) [had] decided to go back to school, t, 60
I
(Mary) [had] decided to go back to school, TAB, 57
I
working at the post office during the past suamer,
(Mary) [has] decided to go back to school, TAB, 70
I
(Mary) [has] decided to go back to school, TAB, 61a
working at the post office (Mary) [has] decided to go

during the past summer, HTA, 11,6,9 back to school, TAB, 62
I
working at the post office
during the past su••er, TAB, 61b
~
---------
[have],
TABIITAB
(Mary) [decides] to go
back to school, TAB, 56
I
during the past working at the post (Mary) [decides] to go
summer, MTA, 63 office, TAB, 1,4 back to school, t, 1~
I I
during the past summer, TA working at the post office, PRPL s
(103) Vt[past(t) & AT(t, Vt1[K(L6)(AVt2[Mg(t1,t2) &

[during-the-past-summer'( t 2 ) &
AT(t 2 , VxS[R(x S , x~) &
vorking-at-the-post-office'(x S )])] ])(A[perf(t 1) &
vt 2 [t 2 =- t1 &
AT(t 2 , Mary-decides-to-go-back-to-school')]])])]
Note that under the interpretation represented by

(101), the adjunct is outside the scope of past tense: this
is clearly the case, since the time adverb modifying the
adjunct is interpreted relative to the speaker's interval--
that is, it denotes the set of subintervals of the summer
immediately preceding the utterance of (99). In the
interpretation represented by (103), on the other hand, the
252 CHAPTER IV
adjunct is within the scope of the past tense: in this

case, the time adverb during the past summer is interpreted
relative to some past interval i, denoting the set of
subintervals of the summer immediately preceding i.
Clearly, the free adjunct in sentence (99) gi ves rise
to precisely the pair of interpretations that one would
expect under the assumption that it functions as a main
tense adverb. This assumption additionally entails that
sentence (99) should cease to be ambiguous in the way
described if it is converted to a simple past tense
sentence--

Intuitively, this prediction is exactly right: (104) can

only be understood to mean that Mary worked at the post
office during the summer immediately preceding the speaker's
interval. This is the interpretation induced by the
analysis assigned to (104) by the proposed system:
( 105)
-----
(Hary) [decided] to go back to school, t, 60
I
(Hary) [decided] to go back to school, TAB, 61a
working at the post office during (Hary) [decided] to go

the past summer, HTA, 11,6,9 back to school, TAB, 57
I I
working at the post office (Hary) [decides] to go
during the past summer, TAB, 61b back to school, TAB, 56
~ I
during the past working at the post (Hary) [decides] to go
summer, MTA, 63 office, TAB, 1,4 back to school, t, 19
I I
during the past summer, TA working at the post office, PRPL s
(106) Vt[K(L6)(AVt1[M9(t,t1) &

[during-~he-pas~-summer'( t 1) &
AT(t1' Vxs[R(x s • xt) &

working-at-the-post-officel( x S )])]] )(A[ past( t) &
AT(t, Hary-decides-to-go-back-to-schooll)])]
Note, in particular, that the free adjunct is outside the

scope of tense in this interpretation, as is clear from the
way in which its time adverb during the past summer is
interpreted.
Thus, the hypothesis that free adjuncts may serve as
main tense adverbs--first advanced in section 2.3 of Chapter
III--is palpably confirmed by the range of possible
interpretations for sentences such as (99) and (104): the
free adjunct interacts with tense and with the perfect in
exactly the ways in which clear examples of main tense
adverbs do. Specifically, they may join either with the
simple tenses or with the perfect to characteri ze a
particular interval, and can only be interpreted as being
within the scope of tense in the latter case.
My conclusion, then, is that adjuncts not restricting a
main clause binary operator may serve as main tense adverbs.
Indeed, I think an even stronger conclusion is warranted:
in the absence of clear counterevidence, it is reasonable to
assume that the only possible function for an adjunct not
restricting a binary operator is that of a main tense
adverb.
3. The Progressive Aspect and the Interpretation

of Free Adjuncts
In this section, the relevance of the progressive aspect to

the semantics of free adjuncts is considered.
English speakers presented with present participial
adjuncts like that in (107) often assume that they are
progressive in aspect;
(107 ) Crossing the street, John was hit by a car.
compare Kruisinga's (1932: 276) observation that 'the most

frequent of the verbal forms in this function [of free
adjunct--G. T. S.] is the ing; this is natural, for the ing
expresses the durative aspect, and this is often what is
required when attendant circumstances are referred to.'
A modern syntactician would no doubt regard this
identification of present participial adjuncts with
progressi ve forms as naive. S/he would point out (1) that
although momentary predicates may occur as present
participial adjuncts, as in (108a)-(110a), the same
predicates do not occur as progressives, as the
254 CHAPTER IV
unacceptability of (108b)-(110b) shows;
(108 ) a. Being a sailor, John smokes a pipe.

b. *John is being a sailor.
a. Having barely enough money for bus fare, Mary
decided to skip lunch.
b. *Mary was having barely enough money for bus
fare.
(110) a. Weighing five tons, our truck made the bridge
shake.
b. *Our truck was weighing five tons.
(ii) that verb phrases in the perfect may occur as present

participial adjuncts, as in (111), even though the
progressive may never govern such verb phrases in finite
declarative sentences, as the unacceptability of (112)
shows;
( 111) Having finished his work, John went to bed.

(112 ) *John is having finished his work.
and (iii) that present participial adjuncts deriving from

verb phrases in the perfect may themselves contain
progressives, as in (113).
(113 ) Having been reading the book, John knew all the
answers.
Resorting to semantic considerations, slhe might also point

out that (iv) some present participial adjuncts do not
obviously have a progressive meaning, such as that in (114).
(114) John drove from Indianapolis to Lexington,

stopping for a break in Louisville.
On the basis of this evidence, slhe would conclude that

present participial adjuncts have no more than a superficial
similarity to progressive forms.
Nevertheless, the naive intuition that present
participial adjuncts have something in common with
progressives has more merit than is immediately evident; in
fact, a close semantic comparison of present participial
adjuncts and progressives leads to some rather startling
conclusions, as I shall now show.
3.1. THE SEMANTICS OF PRESENT PARTICIPIAL PHRASES
It has recently been argued that an adequate

truthconditional account of the progressive must incorporate
both temporal and modal conditions.
The temporal dimension of the truthconditions has been
assumed for several years now. In unpublished work dating
from 1972, Michael Bennett and Barbara Partee ar~ued that a
progressi ve sentence is true at interval i iff the
corresponding nonprogressive sentence is true at some
interval of time properly containing i (but not ending with
i): according to their truthconditions, sentence (115) is
true at some interval iff sentence (116) is true at some
more encompassing interval, as in (117).
(115 ) John is pushing a cart.

( 116) John pushes a cart.
(117) (115) is true at i iff (116) is true at i'.
i i"
I I I I
~
i'
[ i c i ' : i " i LJ
Dowty (1979:Ch.2) has argued that this temporal

condition is by itself inadequate, owing to what he calls
the 'imperfective paradox'. The paradox consists in the
fact that progressives with accomplishment or achievement
predicates do not entail the corresponding nonprogressive
sentences (even though the truthconditions of progressive
verb phrases must apparently be stated in terms of their
constituent nonprogressive verb phrases);8 for example, the
truth of sentence (118) would not allow one to infer the
truth of (119)--(118) could well be true even if someone
caught the glass in midair.
(118) The glass was falling to the floor.

(119 ) The glass fell to the floor.
The consequence of this paradox is that the progressive

cannot simply be regarded as a temporal operator, according
to which the actual truth of a progressive sentence at some
256 CHAPTER IV
interval depends upon the actual truth of the corresponding

nonprogressive sentence at some surrounding time interval.
Instead, a modal dimension in the truthconditions for the
progressive must be acknowledged. Consider sentence (120).
(120 ) The glass is falling to the floor.

( 121) The glass falls to the floor.
Intuitively, the truth of this sentence at some interval i

implies that sentence (121) would be true at some interval
properly containing i if nothing unanticipated were to
impede the glass' fall subsequently to i--that is, if the
'natural course of events' were somehow allowed to flow
inertly starting at i.
Dowty (1979:145-150) proposes truthconditions for the
progressive which embody this intuition model-theoretically.
Crucial to these truthconditions is a category of possible
worlds which he calls inertia worlds. Given any index
<w,i>, a set of inertia worlds is assigned to <w,i>; this
set is to be thought of as containing exactly those worlds
that are like w up to and including i, and 'in which the
future course of events after this time develops in ways
most comRatible with the past course of events'
(1979:148).9 Employing this novel addition to model
structure, Dowty proposes an intensional logic progressive
operator PROG with the following truthconditions (where Inr
is the function assigning a set of inertia worlds to each
index):
( 122) [PROG til] is true at index <w,i> iff for some

interval i' such that i c i' and i is not a final
subinterval for i', and for all w' e: Inr«w,i»,
¢ is true at <w',i'>. (Cf. Dowty 1979:149)
According to (122), sentence (120) is true at index <w,i>

iff (121) is true at some interval which properly contains
(but doesn't end with) i in every inertia world assigned to
<w,i>, as in (123) (see next page). As the reader can
appreciate, this account of the progressi ve captures both
its temporal and modal dimensions.
In this subsection, I shall argue that the progressive
aspect in fact has no independent semantic status in
English--that its truthconditions are entirely determined by
general semantic properties of present participles.
A good construction to examine in order to find out
(123 ) (120) is true at <w,i> iff (121) is true at each

of <wO,P>, ••• <wn,i'>.
F
the set of
inertia worlds I )
assigned to I
<w,i> I
wn I
i"
-
i
w
--
~
I I
"'-"""
i'
[i c i'; i" 't i.]
about the semantics of present participles is the free

adjunct construction: not only can the full ranfbe of
present participial phrases occur as free adjuncts, but
other sorts of predicative phrases can as well; this allows
one to determine exactly which parts of the meaning of a
present participial adjunct derive from the participial
phrase itself, and which are part of the constructional
meaning of free adjuncts generally.
One conclusion which can be drawn on the evidence of
free adjuncts is that the lack of an entailment of
perfectiveness is not a property peculiar to the
progressi ve, It doesn't necessarily follow from sentence
(124) that John actually crossed the street, any more than
it follows from sentence (125); (124), like (125), may be
felt to imply only that John would have crossed the street
if nothing unanticipated had happened.
(124 ) Crossing the street, John was hit by a car.

(125 ) John was crossing the street,
It's important to recognize that the fact that the

adjunct in (124) lacks an entailment of perfectiveness is a
consequence of the fact that the adjunct consists of a
present participial phrase. Free adjuncts of other kinds--
those consisting of past participial phrases, adjective
phrases, prepositional phrases. or predicative noun
phrases--are never imperfective in this way. Consider, for
example, sentences (126) and (127).
258 CHAPTER IV
(126 ) Losing, the Phillies left the field.

(127) Beaten, the Phillies left the field.
These sentences differ in meaning only because the adjunct

in (126) needn't be understood as perfective, while that in
(127) must be. This shows that the imperfectiveness of a
present participial adjunct is not simply inferred; if
language users' inferences were responsible for the
possibility of understanding the adjunct in sentence (126)
imperfectively, they would, in the same way, admit a similar
interpretation for the adjunct in (127). Furthermore,
present participial phrases in other sorts of constructions
may be understood imperfectively; in each of the examples in
(128), the underlined participial phrase may have such an
interpretation.
(128 ) Adnominal participles:
Mary found the dying man.

[doesn't entail that he died]
The man crossing the street was hit by a car.

[doesn't entail that he crossed the
street]
Temporally restrictive adjectives (Dowty 1973):
John sat reciting the Iliad.

[doesn't entail that he recited the
Iliad]
Jane found Rover running across ~ field.

[doesn't entail that he ran across the
field]
Augmented adjuncts:
While copying the sentence into his notebook,

he ran out of ink. ---- ---
[doesn't entail that he copied the
sentence into his notebook]
(128 ) [continued]
Augmented absolutes:
With the children setting ~ the tents, Jane

and Bill went to look for firewood.
[doesn't entail that the children set up
the tents]
Although neither progressive sentences with

accomplishment or achievement predicates nor present
participial adjuncts with such predicates have an entailment
of perfectiveness, it is nevertheless possible in many
contexts to draw an inference of perfectiveness. For
example, one might well infer the simple preterit sentence
(129) from the past progressive (130); similarly, one might
infer (129) from the free adjunct sentence (131).
(129 ) John crossed the street.

( 130) Before entering the building opposite his office,
John was crossing the street.
( 1 31 ) Crossing the street, John entered the building
opposite his office.
Present participles in other constructions also seem to

admi t an inference of perfectiveness: sentence (132), for
instance, can be taken as either (1~i) or (134), and
sentence (135), as either (136) or (137);
(132) In the fourth inning, John bet on the winning

team.
(133) In the fourth inning, John bet on the team that
was winning.
(134) In the fourth inning, John bet on the team that
won.
(135 ) Everyone climbing the mountain received a prize.

( 136) Everyone who was climbing the mountain received a
prize.
(137) Everyone who climbed the mountain received a
prize.
similarly, sentence (138) can be understood as (139), while

(140) can be interpreted as (141).
260 CHAPTER IV
(138 ) When playing the march, the ensemble rose.

(139) When they were playing the march, the ensemble
rose.
(140) When playing the march, the ensemble was standing.
( 141) When they played the march, the ensemble was
standing.
I shall now propose a semantics for present participial

phrases which accounts for these important similarities
between progressives and other sorts of present participial
constructions. To make my proposal explicit, I introduce an
intensional logic expression log of type «s,t>,t> with the
following interpretation:
(142 ) Where $ denotes a proposition p, Iog($) is true at

index <w,i> iff for some interval i' such that i c
i' and i is not a final subinterval for i', and
for all w' E: Ind<w,i», p«w',i'» = 1.
Given an intransitive verb phrase with translation a', my

proposal is to interpret the corresponding present
participial phrase as Ax[Iog("'[a'(x)])].
To appreciate the content of this proposal, consider the
present participial phrase crossing the street, which
receives the translation (143).
(143)
According to (142), John has a stage in the denotation of

(143) at index <w,i> iff there is some interval i' such that
i is a proper subinterval of i', i is not a final
subinterval for i', and for each inertia world w' assigned
to index <w,i>, John has a stage in the denotation of cross-
the-street' (i.e. John crossed the street) at <w',i'>.
Under this analysis of the semantics of present
participial phrases, sentence (144) translates as (145).
(144 ) Crossing the street, John ran into Mary.
(145) AT(t" Vxs[R(x s , xJ)

Vt[K(L 6 )("'vt 1 [M 9 (t,t 1) & &
Iog( . . [cross-the-street'(x s )])]) ])("'[past(t) &
AT (t, John-runs-into-Mary')])]
According to the interpretation induced by (145), (144)

doesn't necessarily ~ntail that John crossed the street

4
(even assuming that x has John as its value). On the other
hand, (145) reflects the possibility of inferring that the
adjunct in (144) is perfective: the value inferred for Mg
may require that the intervals denoted by t and t1 be
separated by some interval during which John could have
finished crossing the street.
If present participial phrases are interpreted in the
manner just described, the progressive may simply be
regarded as a predicative construction, consisting of a
copula and a present participial phrase. 12
3.2. THE FORMAL SEMANTICS OF PRESENT PARTICIPIAL ADJUNCTS
In this subsection, rules embodying the generalizations of

the preceding subsection are proposed for the derivation and
interpretation of present participial adjuncts.
Present participial pl;rases are generated in two
categories: PRPL s and PRPL 1 (see Chapter II, section 3.1
for arguments motivating this category distinction).
Expressions of these twq categories derive from verb phrases
of category IV s and IV 1 , respectively. The rules accomp-
lishing these transfers are (146) and (147): 13
(146) S46. I f a € PIVi, then F 46 (a) € PpRPLi, where

F 46 (a) is DB(ING(a».
Example: F46 ([be] a sailor) = being a sailor
T46. If a E PIvi and a translates as a', then

F 46 (a) translates as Axo[Ing("[a'(Xo)])],
Axk[Ing("[a'(x k )])], or Axi[Ing("[al(x i ) ])],
according as a l is of sorted type <o,t'>,
<k,t'>, or <[o,k] ,t'>, respectively.
Example: F46 ([be] a sailor) translates as

Axo[Ing("[sailorl(xO)])]
(147) 345. If a e:PIVs, then F 45 (0I.) e: PpRPLs, where

F 4S (a) is DB(ING(a».
262 CHAPTER IV
(147) [continued]
Example: F45 ([walk] home) = walking home
T45. If a e: PIVs and a translates as a', then

F 45 (a) translates as Axs[Ing("[a'(x s )])].
Example: F45 ([walk] home) translates as

Ax s [lng("[walk-home' (x s )])]
Given these rules, sentence (124) can be analyzed as in

(148) •
(148) crossing the street, (John) [was] hit by a car,

I
---------
t, 60
crossing the street, (John) [was] hit by a car, TAB, 61a
crossing the street, MTA, 11,6,9 (John) [was] hit

I by a car, TAB, 57
crossing the street, TAB, 1,4 I
I (John) [is] hit
crossing the street, PRPL s , 45 by a car, TAB, 56
I I
[cross] the street, IV s , 28 (John) [is] hit
by a car, t, 19
On this analysis, (124) is assigned the translation (149).
(149) AT(t 1, Vxs[R(x s , xi>

Vt[K(L 6 )("Vt 1 [M 9 (t,t 1) & &
Ing("[ cross-the-street'(x s )])])] )("[ past(t) &
AT(t, John-is-hit-by-a-car')])]
According to the interpretation induced by (149), the

adjunct in (124) need not be understood as perfective; but
certain possible values for M9 would permit an inference of
perfectiveness.
Sentence (150) may be derived as in (151) and assigned
the translation (152).
(150) Being asleep, John was unaware of Mary.

( 151) being asleep, (John) [was] unaware of Mary, t, 60
-----------
I
being asleep, (John) [was] unaware of Mary, TAB, 61a
being asleep, MTA, 11,6,9 (John) [was] unaware of Mary,

I TAB, 57
being asleep, TAB, 2,4 I
I (John) [is] unaware of Mary,
being asleep, PRPL i , 46 TAB, 56
I
[be] asleep, IVl., 25
. I
(John) tis] unaware of Mary, t, 19
(152) Vt[K(L6)("'Vt 1 [M 9 (t,t 1 ) ~

AT(t 1, Ing("'Vxs[R(x s , x~) &
asleep'(x s )]» ])("'[past(t) &
AT (t, John-is-unaware-of-Mary')])]
Here there is no question of perfectiveness: owing to the

semantic properties of stative predicates, (153) is valid.
(153) Axi 0 [Ing("'Vxs[R(xS,x i ) & asleep'(x s )]) -+

Vxs[R()!:s,x i ) & asleep'(x s )]]
The validity of (153) may be one reason for the

unacceptability of momentary predicates such as be asleep in
the progressive construction:
(154 ) *John is being asleep.
The use of (154) would conversationally implicate that the

simple present John is asleep for some reason doesn't hold,
in accordance with the principle of conversational economy
discussed by McCawley (cf. also Dowty (1979: 180»; yet,
(154) entai Is that John is asleep. Thus, (154) may perhaps
be regarded as conversationally anomalous in roughly the
same way as (12b).
( 12b) *John has been on the train yesterday.
Note, however, that because there is nothing comparable to

the Simple tenses with which present participial phrases
compete in free adjuncts, the use of present participial
264 CHAPTER IV
statives 4like that in (150) poses no conversational

problems. 1
The progressive is, again, regarded simply as a
predicati ve construction. Thus, sentence (125) is anal yzed
as in (155). (155) produces (156) as a translation for
( 125) •
(155) (John) [was] crossing the street, t, 60

I
(John) [was] crossing the street, TAB, 57
I
(John) [is] crossing the street, TAB, 56
I
(John) [is] crossing the street, t, 19
~.
John, T
[be] cross1ng the street, lvi, 25
~ s
[be], IV1/PRED crossing the street, PRPL s , 45
I
[cross] the street, IV s , 28

Ing(A[cross-the-street'(x s )])])]
_. Chapter Summary
In this chapter, the perfect and the progressi ve and their

relevance to the semantics of free adjuncts have been
investigated. Whereas the main focus in Chapters II and III
was on the role played by a free adjunct in the syntactic
composition and semantic interpretation of its matrix
sentence, the focus in the present chapter has been
predominantly on the internal structure and semantics of
free adjuncts--specifically, perfect and present participial
adjuncts. Such adjuncts have been found to provide critical
insights into the semantics of the perfect and the
progressive in English. In section 1, it was shown that the
use of the perfect in free adjuncts (in fact, in all
constructions other than the present perfect) cannot be
accounted for by the extended now theory of the perfect; an
alternative theory was therefore proposed which is fully
consistent with the observed range of uses for the perfect.
In section 2, an argument was developed favoring the
conclusion that free adjuncts may function as main tense
adverbs: under the theory of the perfect proposed in

section 1, main tense adverbs may serve in two distinct ways
in the past, present, and future perfects, although only one
of these uses is possible in the simple tenses; free
adjuncts were shown to follow an identical pattern of usage.
Finally, in section 3, I have argued for a unitary semantic
analysis for present participial phrases; the semantic
peculiarities of present participial adjuncts follow from
this analysis, which additionally allows the progressive to
be regarded as nothing other than a predicative participial
construction.
Footnotes
1. But in some cases, e.g. sentence (1), a precise

temporal relation between the adjunct and its superordinate
clause is in fact entailed, due to the presence of time
adverbials in both clauses that specify particular times.
(i) Setting sail for the island in the fall of 1740,

he reached his destination in the spring of 1741.
2. Recall in particular the truthcondi tions for 'pasta)'

gi ven in section 1.1 of Chapter III: where ~ denotes
interval i', pasta) is true at index <w,D iff there is
some nonempty interval i" such that i' < i" < 1.
3. McCawley (1971) has pointed this out. McCoard

(1978:152) acknowledges that 'the opposition [of the
perfect] with the preterit exists only in tensed forms' and
seems (p.179) to accept the notion that 'some embedded
nontensed perfects are actually derived from preterits'. He
appears not to have noticed that the perfect can serve as a
preterit in the past and future perfect constructions,
though he cites (p.184) one of McCawley's examples which
shows this.
4. I should point out that in the cases discussed by

McCawley (1978), the conversational implicature arising from
the use of the 'marked' member a of some pair a,~ of
competing expressions never has the effect of making a
anomalous; instead, it simply causes a to be assigned an
interpretation different from that of B. For example, the
use of pale red instead of pink conversationally implicates
that pink is somehow inappropriate, but this implicature
266 CHAPTER IV
doesn't make pale red anomalous; it simply causes it to be

assigned a denotation different from that of pink.
In this respect, the cases discussed by McCawley are
different from the proposed pragmatic account of the
perfect. In my analysis, the conversational implicature
arising from the use of sentence (54a) instead of sentence
(54b) has the effect of making (54a) anomalous.
The reason for this difference between the cases
discussed by McCawley and the case of the perfect is clear.
Under the present analysis, (54a) and (54b) are logically
equivalent. Consequently, the conversational implicature
arising from the use of (54a) cannot cause this sentence to
be assigned an interpretation different from (54b); on the
contrary, this implicature--namely, that (54b) does not
apply, for some reason--is simply nonsensical, since (54a)
entails (54b). In the cases discussed by McCawley, however,
the two competing expressions are never logically
equivalent; indeed, it is precisely because their
denotations may differ that the 'marked' member of such a
pair of expressions may be appropriately used under certain
circumstances. Accordingly, the conversational implicature
arising from the use of pale red instead of pink makes
perfect sense: the color the speaker has in mind can't be
classified as pink, but falls somewhere within the range of
colors classifiable as pale red.
The line of reasoning embodied in the proposed analysis
of the perfect is not unprecedented. For example, Dowty
(1979:180) proposes that the anomaly of sentences such as
(i)-in which a momentary stati ve predicate appears in the
progressive aspect--is a consequence of the fact that such
sentences always entail their simpler, nonprogressi ve
variants (e.g. (ii». owing to the semantic properties of
momentary predicates.
(i) John is being awake.

(ii) John is awake.
I shall return to this idea in section 3 below.
5. Recall that the feature [+Tense] is used to prevent

iteration of the rules of tense. [+Perfect] does not serve
a parallel function in rule (58), however: though an
abstract is assigned to the subclass [+Perfect] by rule
(58), it may then be reassigned to the subclass [-Perfect]
by the rule (61), where it could then seemingly undergo rule
(58) again. This iteration is prevented, however, under the

assumption that as a member of PTABIITAB' have lacks a past
participial form (cf. Gazdar, Pullum & Sag (1982:597f»:
under this assumption, F62([have], (,John) [has] been on the
train) is undefined.
6. The statement of T4, like that of T2, is complicated by

the decision to translate English expressions into an
intensional logic sorted into kinds, objects, and stages.
Cf. footnote 22 to Chapter III.
7. Actually, they relativize the interpretation of

progressive sentences to moments rather than intervals. See
Dowty (1979:188) for an argument that progressives should be
interpreted relative to intervals whose length is possibly
greater than a moment.
8. Progressi ve sentences with acti vi ty predicates (walk,

sing, listen for something) pragmatically entail the
corresponding simple present tense sentence. See Dowty
(1979:Ch.3).
9. One might object to Dowty's analysis on the grounds

that in some (maybe even most) situations it is very hard to
decide what the natural course of events would be. But
Dowty himself remarks (1979:149) that "while there are
severely subjective differences among individuals' beliefs
as to how the world would "turn out" if left uninterfered
with, agreement on the truth of progressive sentences, to
the extent that such agreement obtains at all, presupposes
that such beliefs are held in common." 'Natural course of
events' may be a somewhat nebulous notion, but the role it
plays in the interpretation of imperfective progressives is
quite clear.
Note that Dowty relativizes the interpretation of
progressives to sets of inertia worlds. This is meant to
capture the intuition that there may be several mutually
exclusi ve natural courses of events-for example, at least
two natural courses of events may ensue when a coin is
flipped. Notice, however, that his truthconditions require
that the simple present tense sentence corresponding to a
progressive be true irr every member of the relevant set of
inertia worlds; see Dowty (1979:147f) for justification.
Dowty has pointed out to me that his characterization
of inertia worlds as identical to the actual world up to a
268 CHAPTER IV
certain time is inadequate--that it predicts that John was

crossing the street when the truck hit him cannot be true,
since all the relevant inertia worlds would be ones in which
the truck was coming down the street towards John (and thus
hit him). Dowty himself has suggested a possible
alternative characterization of inertia worlds as those that
are 'locally similar' to the real world up to the relevant
time--that is, those that are similar in the immediate
vicinity of, e.g., John. This alternative may, however, run
into problems with sentences like He was copying the
sentence into his notebook when he ran out of ink or
Inflation is destroying the economy.
10. A broader range of present participial phrases may

appear as free adjuncts than may appear as 'reduced relative
clauses' or temporally restrictive adjectives:
Being an experienced cook, John knows all about

wine.
*Anyone being an experienced cook should know all
about wine.
*I saw him being an experienced cook.
11. I do not wish to suggest that (133), (134), (136), and

(137) represent the only possible readings for (132) and
(135); other interpretations are clearly possible.
12. In partl,cular, the progressive consists of the bel of

category IVl./PRED S (see Chapter II, section 3.2.1) in
combination with a stage-level present participial phrase.
13. None of the intransitive verb phrases generated by the

fragment is headed by the have of the perfect; perfect have
is instead introduced only at the temporal abstract 'level'.
For this reason, phrases such as having left are not among
the present participial phrases generated by rules (146) and
(147); such phrases are instead only generated as A-
abstracts by rules (80) and (81). Note that lng does not
figure in the translation of such a phrase, according to the
translation rules in (80) and (81); indeed, nothing in the
semantics of such expressions suggests that it should.
14. Note that my analysis of present participial adjuncts

affords a pragmatic explanation of the anomaly of adjuncts
like that in (i).
(i) *Being carrying a suitcase, John entered the room.
In accordance with the principle of conversational

economy discussed by McCawley (1978), the use of the adjunct
in (i) conversationally implicates that the structurally and
conceptually simpler adjunct in (ii) somehow does not apply.
(ii) Carrying a suitcase, John entered the room.
Yet, under the assumption that time is dense (and that for
all w1 e: Inr«w,D), Inr«w,D) = Inr«w1,D», the adjuncts
in (i) and (ii) have equivalent translations in my analysis:
( iii) VxS[R(Xs,x i ) & Ing(A[carry-a-suitcase' (x s )])]

(i v) Ing(AVxs[R(xs,x i ) &
Ing(A[carry-a-suitcase'(x s )])])
Thus, the adjunct in (i) seems to be anomalous because its

use conversationally implicates that the Simpler adjunct in
(ii) somehow does not apply, despite the fact that the two
adjuncts are equivalent. (This explanation for the anomaly
of (i) parallels the proposed explanations for the anomaly
of (v) and (vi):
(v) *John has seen Mary yesterday.

(vi) *John is being asleep.)
In view of this explanation, the Doubl-ing Constraint

(see Chapter I, section 1.1) needn't be assumed to be solely
responsible for the anomaly of (i).
C HAP T E R V
THE FORMAL SEMANTICS OF ABSOLUTES
In the preceding three chapters, the formal semantics

of free adjuncts has been discussed in detail. A number of
important generalizations have emerged in this discussion;
chief among these are the following.
(a) There are two varieties of free adjuncts--weak and
strong--distinguished in the following ways: (i) weak
adjuncts derive from stage-level predicates, while strong
adjuncts derive from individual-level predicates (where the
stage/individual distinction is that drawn by Carlson
(1977»; (11) as members of the category t' of conditional
clauses, weak adjuncts may restrict the interpretation of a
modal verb in the manner of an if-clause, but strong
adjuncts may not serve in this way; and (iii) as members of
the category TA of set-level time adverbs, weak adjuncts may
restrict the interpretation of a relative frequency adverb
or of the generalization operator G" in the manner of a
temporal adverbial clause, but strong adjuncts may not serve
in this way.
(b) Uhen an adjunct--weak or strong--fails to restrict
the interpretation of some binary operator in the main
clause (Le. some modal or relative frequency adverb, or the
operator G"), its function is that of a main tense adverb-
that is, it serves as a member of the category MTA.
(c) The logical role of a (weak) adjunct restricting a
binary operator may be at least partially determined by the
semantic properties of that operator. Certain components of
the logical relation between a free adjunct and its
superordinate clause may, however, be left indeterminate by
the semantic rules of English; this indeterminacy is
especially clear in the case of adjuncts of category MTA,
which aren't associated with any sort of binary operator in
the main clause. Indeterminacies of this sort, which are
represented by free variables in the translations induced by
the proposed fragment of English grammar, can only be
resolved by the inferences of language users.
In this chapter, I shall investigate the semantics of
270
THE SENANTICS OF ABSOLUTES 271
the nominative absolute and augmented absolute

constructions; I shall argue that generalizations analogous
to (a)-(c) above are valid for these constructions as well.
In section 1, the semantics of absolutes in modal
contexts is examined. I argue that just as free adjuncts
fall into two subclasses according to their behavior in such
contexts, so do absolutes; moreover, this bifurcation of
absolutes is shown to correlate with the distinction between
stage-level and individual-level predicates. This
correlation is complicated, however, by an important
peculiarity shown by absolutes in modal contexts--a
peculiari ty which is without parallel in the earlier
discussion of free adjuncts: for an absolute to restrict
the interpretation of a modal in the manner of an if-clause,
it must not only derive from a stage-level predicate--it
must also be 'augmented' by the preposition with (or
without). Thus, the strong/weak distinction among absolutes
is slightly more complex than the corresponding distinction
among free adjuncts: for a free adjunct to be weak, it is
necessary and sufficient that it derive from a stage-level
predicate; for an absolute to be weak, however, it is
necessary that it derive from a stage-level predicate, but
not sufficient that it do so--it must in addition be
augmented. Rules embodying this finding are proposed for
the derivation of weak and strong absolutes.
Section 2 is devoted to the semantics of absolutes in
the context of a relative frequency adverb or of the
generalization operator G". The findings of the preceding
section are confirmed: augmented absolutes deriving from
stage-level predicates may serve to restrict binary
operators of these sorts, while other kinds of absolutes may
not. Thus, here too, absolutes show a distinction analogous
to, though slightly more complex than, the distinction
between strong and weak free adjuncts. Rules capturing this
additional similarity are proposed.
Section 3 is devoted to a brief demonstration that
absolutes which do not serve to restrict the interpretation
of a binary operator are best regarded as main tense
adverbs. The argument is parallel to that developed for
free adjuncts in section 2 of Chapter IV.
The conclusions of this chapter are summarized in
section 4. The rules presented in this chapter appear in
the formal fragment presented in the Appendix.
272 CHAPTER V
1. Modality and the Interpretation of Absolutes
In Chapter II, the class of free adjuncts was found to be

bifurcated in modal contexts: weak adjuncts can be
interpreted as if-clauses (more exactly, they can serve as
the first argument of an accompanying modal) while strong
adjuncts cannot. Careful consideration reveals that the
class of absolutes shows a similar although slightly more
complicated bifurcation.
1.1. WEAK AND STRONG ABSOLUTES
(1) His mother being a doctor, John would know the way
to the Med Center.
(2) The water being a little cold, the children must
stay on the beach.
(3) His arm being in a cast, Bill might not be asked
to participate.
In none of these sentences can the absolute be understood as

an if-clau03e; for example, (1) i03 not felt to mean 'If his
mother were a doctor, John would know the way to the Med
Center'. In03tead, (1 )-(3) uniformly entail the truth of
their absolutes. In view of the findings in Chapter II,
thi03 i03 exactly what one would predict, O3ince each of these
absolutes has an individual-level predicative phrase; thus,
the absolutes in (1 )-0) might be regarded as strong
absolutes analogous to the strong adjuncts discussed in
Chapters II-IV. As such, they may be tentatively
categorized as main tense adverbs; I shall argue in section
3 of this chapter that such a categorization is well-
motivated.
The analogy of the absolutes in (1 )-0) to strong
adjuncts should lead one to expect that absolutes with
stage-level predicates will show the behavior of weak
adjuncts--that they will, for example, play the role of an
if-clause in modal contexts. This expectation is, at first
glance, disappointed. Consider sentences (4 )-(6), in which
the absolutes have stage-level predicates.
(4) The truck in first gear, we would coast gently

downhill.
THE SEMANTICS OF ABSOLUTES 273
(5) Her hair braided, Jane must resemble Mary.
(6) The children asleep, Mary might watch TV.
These sentences are of marginal acceptability, and are

accordingly rather difficult to judge. Nevertheless, I
sense that, to the extent that any kind of interpretation
can be assigned to (4)-(6), they lack an interpretation
wherein the absolute functions as an if-clause; 1 (4), for
me, cannot mean 'if the truck were in first gear, we would
coast gently downhill'. Nevertheless, it would be wrong to
conclude from this fact that the distinction between strong
and weak adjuncts is not reflected among absolutes:
sentences (4)-(6) can be rather remarkably transformed if
the absolutes are replaced with augmented absolutes, as in
(7)-(9) •
(7) Wi th the truck in first gear, we would coast

gently downhill.
(8) With her hair braided, Jane must resemble Mary.
(9) With the children asleep, Mary might watch TV.
In each of (7)-(9), the augmented absolute can easily be

understood as the first argument of the accompanying modal;
for example, (7) is easily interpreted as 'if the truck were
in first gear, we would coast gently downhill'. Thus, these
augmented absolutes may be considered as weak absolutes
analogous to the weak adjuncts discussed in Chapters 11-
IV.2
What distinguishes the absolutes in (7)-(9) is that
they contain stage-level predicates and are, in addition,
augmented by a preposition. Absolutes lacking either or
both of these characteristics fail to serve as conditional
clauses in modal sentences; this is true of the unaugmented
absolutes in (1)-(6), as well as of the absolutes in (10)-
(12), which, though augmented, contain individual-level
predicates.
(10) vii th his mother being a doctor, John would know

the way to the Med center.
(11) viith the water being a little cold, the children
must stay on the beach.
(12) With his arm being in a cast, Bill might not be
asked to participate.
Thus, whereas the distinction between weak and strong free

274 CHAPTER V
adjuncts was found to correlate simply with the distinction

between stage-level and individual-level predicates, the
distinction between weak and strong absolutes is more
intricate: weak absolutes must contain stage-level
predicates, but must in addition be prepositionally
augmented; ~bsolutes lacking one or both of these properties
are strong • .J
Why it is that weak absolutes must be prepositionally
augmented is a fascinating question, all the more so since
strong aUbmented absolutes (such as those in (10)-(12»,
thOU9!'l acceptable, seem stylistically less natural than weak
ones. The pattern reflected in these facts--weak absolutes
must be augmented, while strong absolutes tend not to be--is
not at all accidental,but is, I think one manifestation of
a rather general phenomenon: weak adjuncts and absolutes
are in general more likely than strong ones to be
prepositional constructions.
One other instance of this tendency has already been
observed in section 3.4 of Chapter II, where it was seen
that certain strong adjuncts headed by individual-level
verbs have prepositional adjuncts (headed by as or with) as
their weak 'alternants', as in the following contrasting
pairs.
( 13) a. Being a blonde, Mary might look something

like Jane.
b. As a blonde, Mary might look something like
Jane.
( 14) a. Having green eyes, Mary might look something
like Jane.
b. With green eyes, Mary might look something
like Jane.
More striking confirmation of this tendency, however, can be

found in languages other than English.
French, like English, shows a distinction between weak
and strong adjuncts/absolutes; for example, according to the
native speakers with whom I have consulted, the adjunct in
(15a) is strong (just like the English adjunct being Swiss)
while that in (16a) is weak (like English wearing this hat):
(15) a. Etant suisse, elle parlerait peut-etre

l'allemand.
'Being Swiss, perhaps she would speak
German.' [not equivalent to (15b)]
,
b. Si elle etait suisse, elle parlerait peut-
etre l'allemand.
'If she were SWiss, perhaps she would speak
German. '
(16) a. Portant ce chapeau, il ressemblerait un peu ~
Jean.
'Wearing this hat, he would look a little
like Jean.' [equivalent to (16b)]
b. S'il portait ce chapeau, il ressemblerait un

peu a Jean.
'If he wore this hat, he would look a little
like Jean.'
Now, French free adjuncts may appear in two forms: unmarked

(as in (16a», or headed by the preposition en 'in' (as in
(17)).
(17) En portant ce chapeau, i l ressemblerai t un peu a

Jean.
'Wearing this hat, he would look a little like
Jean.' [equivalent to (16b)]
For weak adjuncts, the en construction is prefe·rred in

conversation, the unmarked construction being regarded as
somewhat 'literary'. For strong adjuncts, however, only the
unmarked construction is possible for many speakers; that
is, while (15a) is acceptable, (18) is not.
(18) *En ~tant suisse, elle parlerait peut-etre

l'allemand.
A similar phenomenon is observed with absolutes in

French: like free adjuncts, absolutes appear in either an
unmarked or a marked form; absolutes of the latter sort are
headed by avec 'with'. In weak absolutes, the use of avec
is essentially obligatory; indeed, the relative
acceptability of the French sentences (19a) and (20) is
comparable to that of their English counterparts, (21a) and
(22).
(19) a. Avec Jean dirigeant nos travaux, nous aurions

reussi. [equivalent to (19b)]
276 CHAPTER V
b. Si Jean avait dirige nos travaux, nous

.
aurlons '.
reussl..
(20) ?Jean dirigeant nos travaux, nous aurions reussi.
(21) a. With John guiding our work, we would have

succeeded. [equivalent to (21b)]
b. If John had guided our work, we would have

succeeded.
(22) ?John guiding our work, we would have succeeded.
On the other hand, the use of avec with a strong absolute is

completely unacceptable (cf. Ruwet (1978: 180f):
a. Jean ~tant malade, la reunion serait remise ~

plus tard.
'Jean being sick, the reunion would be
postponed.' [not equivalent to (23b)]
b. Si Jean etait malade, la reunion serait

remise a plus tarde
'If Jean were sick, the reunion would be
postponed. '
(24) *Avec Jean etant malade, la r~union serait remise

~ plus tard.
Thus, in French, weak adjuncts and absolutes tend to be

prepositional constructions (headed by en or avec), while
strong adjuncts and absolutes are never prepositionally
augmented; my view is that, ultimately, this reflects the
same phenomenon as the fact that weak absolutes must be
augmented in English while strong absolutes tend not to be.
Why should the functional distinction between weak and
strong absolutes be paralleled in this way by a formal
distinction? Although I don't have an indisputable answer
to this question, I offer the following speculation. Weak
adjuncts and absolutes may serve as expressions of
categories t' and TA; as members of these categories, they
appear as arguments of various sorts of operators appearing
in the main clause. Strong adjuncts and absolutes, on the
other hand, never serve as arguments of a main clause
operator, but instead express a proposition logically
distinct from that expressed by the main clause (indeed,

recall that K(L) represents the inferred relation between
these two propositions). Thus, whereas weak adjuncts and
absolutes may serve in various clearly subordinate
functions, strong adjuncts and absolutes are always much
less clearly subordinate in character. Perhaps this
difference is what is responsible for the fact that the
former tend to be prepositionally augmented, while the
latter tend not to be: in accordance with their
functionally subordinate status, weak adjuncts and absolutes
tend to assume a hypotactic form, while strong adjuncts and
absolutes, being less obviously subordinate in function,
tend to remain paratactic in form. 5
1.2. FORMALIZING THE DISTINCTION BETWEEN STRONG AND WEAK

ABSOLUTES
Recall that in Chapter III, section 2.1, it was proposed

that free adjuncts derive from temporal abstracts; this
proposal was meant to account for the fact that free
adjuncts, like finite clauses, have independent temporal
reference. The same is true of absolutes, as is especially
clear in examples such as (25) and (26), in which the
absolutes are themselves modified by main tense a'dverbs:
(25) \Olith the train departing at noon, we'll be lucky

to make it to the station on time.
(26) His family having I i ved in Paris in 1940, John
sometimes has nightmares about the war.
Accordingly, I shall assume that absolutes also derive from

temporal abstracts.
In Chapter III, the difference between weak and strong
adjuncts was formally reconstructed as a distinction between
two subclasses of A-abstracts: those deriving from stage-
level predicates were classified [+Weak], while those
deriving from individual-level predicates were classified
[-Weak]. The distinction between weak and strong absolutes
can be analogously formalized by means of rules (27) and
(28) (see next page).6
These rules both generate temporal abstracts in the
subclass [+Adjunct] (i.e. A-abstracts), but these differ
from the A-abstracts discussed in Chapter III in that they
do not have the form of free adjuncts: instead, they have
278 CHAPTER V
(27) 85. If a e: P T and S e: PPREDs, then F 5a Ca,S),

F 5b Ca,S) e: P TAB , [-Tense, +Adjunct,
-Perfect], F 5a Ca,S) e: [-Weak], and F 5b Ca,8)
e: [+Weak], where F 5a Ca,8) is CONCa,S) and
F5b Ca,s) is CON Cwith, CONCOBJCa),S».
Examples: F 5a Cthe truck, in first gear) = the truck in

f'irst gear;
F5b CheO' wearing that new outfit) = with himO
wearing that new outfit
T5. If a e: PI' 8 e: PpREDs, and a,8 translate as

a',8', then F 5a Ca,8), F 5b (a,8) translate as
Xt[ATCt, a'C"XxiVxs[RCxS,x i ) & 8'Cx s )]»].
Example: F 5a Cthe truck, in first gear) translates as

Xt[ATCt, the-truck'C"XxiVxs[R(X s , xi) &
in-first-gear'(x s )]»]
(28) S6. I f a e: PT and 8 e: PpREDi, rI PADJi, then

F 6a Ca,8), F 6b (a,8) e: PTAB , [-Tense, +Adjunct.
-rleak, -Perfect], where F 6a(a,8) is CONCa,8)
and F 6b Ca,8) is CONCwith, CONCOBJ(a),S».
Examples: F 6a CJane, being a doctor) = Jane being a

doctor;
F6b (she" being a doctor) = with her, being a
doctor
T6. If a e: P1" S e: PpREDi, and a,8 translate as

a',8', then F 6a (a,8), F 6b (a,8) translate as
At[ATCt, a'C"S'»].
Example; F6aCJane, being a doctor) translates as

Xt[AT(t, Ing("[doctor'{Jane')]»]
the form of nominati ve or augmented absolutes. (27)

generates A-abstracts containing stage-level predicates,
while those generated by (28) contain individual-level
predicates. 7 As was seen in the preceding subsection, weak
absolutes are distinguished in two ways--they must contain
stage-level predicates, but must in addition be
prepositionally augmented; accordingly. of all the A-
abstracts generated by the above rules, only the
prepositionally augmegted ones generated by (27) belong to
the subclass [+Weak).
1.3. DERIVING CONDITIONAL ABSOLUTES
In order to account for the fact that the weak absolutes in

sentences (7)-(9) receive a conditional interpretation,
absolutes of this sort must be introduced into the category
t' of conditional clauses.
(7) With the truck in first gear, we would coast

gently downhill.
(8) With her hair braided, Jane must resemble Mary.
(9) With the children asleep, Nary might watch TV.
In fact, if absolutes are assumed to derive from A-

abstracts, as was proposed in the preceding subsection, the
appearance of weak absolutes of category t' is guaranteed by
independently needed rules: in Chapter III, section 2.2,
the rule S10 was proposed for converting A-abstracts in the
subclass [+Weak] into conditional adjuncts; this very same
rule applies without modification to the weak A-abstracts
generated by rule (27), converting them into expressions of
category t'. Accordingly, the conditional interpretation of
sentence (8) can be induced by means of the analysis in
(30), which corresponds to the intensional logic translation
in (29). Similar analyses can be provided for sentences (7)
and (9) and their ilk.
(29) Vt[PRES(t) & AT(t, must'(DCcb)(h vt1 [ATCt 1, Jane's-

hair'C h ).xiVxs[RCx s ,xi) &
braided'(xS)]»]»(AJane-reseables-Hary'»]
280 CHAPTER V
(30) with her hair braided, (Jane) [must]

resemble Mary, t, 55,1
Jane, T
-------------
with her1 hair braided, (she1) [must]
resemble Mary, t, 60
I
with her1 hair braided, (she 1 ) [must] resemble Mary, TAB, 58
I
with her 1 hair braided, (she 1 ) [must] resemble Mary, TAB, 56
I
with her 1 hair braided, (she 1 ) [must] resemble Mary, t, 18
l
[must] , with her 1 hair brt~~,·de10d, M(Sahrey1)t[re1SgembleS]
MOD , ,
~ir braided, TAB, 5b
her 1 hair, T braided, PSPL s
In this section, I have shown that in modal contexts,

there is a distinction between two varieties of absolutes
that is analogous to the difference between strong and weak
adjuncts: absolutes which have stage-level predicates and
which are prepositionally augmented may restrict the
interpretation of a modal in the manner of an if-Clause, and
may thus be classified as weak; but absolutes which have
individual-level predicates or are not prepositionally
augmented cannot restrict the interpretation of a modal, and
are therefore to be classified as strong. In the following
section, further evidence motivating this distinction
between two sorts of absolutes is discussed.
2. Tense and the Interpretation of Absolutes
In this section, it is shown that weak absolutes can serve

as set-level time adverbs restricting the interpretation of
a relative frequency adverb or of the generalization
operator G". This special usage of weak absolutes provides
further motivation for distinguishing between two varieties
of absolutes, and constitutes an additional similarity
between weak absolutes and weak adjuncts.
2.1. ABSOLUTES AND RELATIVE FREQUENCY ADVERBS
Recall that in section 3 of Chapter III, it was shown that

weak adjuncts can restrict the interpretation of a relative
frequency adverb, as in (31);
(31) Walking home, John often stops to see Mary.
strong adjuncts were seen not to function in this way.

Careful consideration reveals an analogous difference
between weak and strong absolutes.
Consider sentences (32)-(34), which have strong
absolutes.
(32) His wife being a doctor, John sometimes stops by

the Med Center.
(33) Her arm being broken, Mary never played volleyball
last summer.
(34) Their truck weighing seven tons, they are
occasionally denied access to smaller bridges.
In none of these sentences can the absolute be understood as

supplying a temporal restriction on the accompanying
relative frequency adverb; for example, sentence (34) cannot
be interpreted as 'when their truck weighs seven tons, they
are occasionally denied access to smaller bridges'. Thus,
the absolutes in sentences (32)-(34) are not to be analyzed
as set-level time adverbs serving as the first argument of
the frequency adverb in the main clause. (In section 3, I
shall argue that such absolutes are in fact main tense
adverbs.)
Now turn to sentences (35)-<37), which have weak
absolutes.
(35) With her hair braided, Jane sometimes reminds me

of Mary.
(36) With her children asleep, Mary often watches TV.
(37) With his work done, John never goes straight to
bed.
The absolutes in these sentences can be very naturally

interpreted as supplying the temporal restriction on the
accompanying frequency adverbs. Sentence (36), for example,
can be interpreted as 'when her children are asleep, Mary
often watches TV'. Thus, these weak absolutes are
282 CHAPTER V
apparently to be analyzed as set-level time adverbs serving

as the first argument of a relative frequency adverb.
No new rules are necessary in order to assign the weak
absolutes in (35)-(37) t;.he desired sort of interpretation.
In section 3.2 of Chapter III, the rule 89 was proposed for
converting A-abstracts in the subclass [+Weak] into set-
level time adverbs; this rule therefore applies directly to
the augmented A-abstracts generated by (27). Accordingly,
sentence (36) can, for example, be analyzed as in (38); (36)
is thereby assigned the interpretation induced by (39),
according to which the absolute with her children asleep
serves as the first argument for often.
---------
(38) with her children asleep,
(Mary) often [watches] TV, t, 55,'
Hary, T with her, children asleep,

(She,) often [watches] TV, t, 60
I
with her 1 children asleep,
(she,) often [watches] tv, TAB, 64a
I
often with her, children asleep, (she,) [watches]
RFA TA, 9,9 TV'I TAB, 58
with her, children asleep, (she,) [watches] TV, TAB, 56
. ~TAB, 5b I
~ ~ (she,) [watches] TV, t, '9
ber, children, T asleep, ADJ s
Vt[often'( At1 Vt 2 [M g (t, ,t 2 ) &

AT(t 2 , Hary's-cbildren,(A AxiVXs[R(x s ,xi) &
asleep'(x s )]»])( t,[t,!::. t & [PRES(t 1 ) &
AT(t" Hary-watches-TV')]])]
Thus, in the environment of a relative frequency

adverb, absolutes are again seen to participate in a
distinction parallel to the distinction between strong and
weak adjuncts: strong absolutes may not provide the
temporal restriction on a relative frequency adverb, while
weak absolutes may do so.9 In the following subsection, yet
another environment will be seen in which these two sorts of
absolutes are distinct in their semantic behavior.
2.2. ABSOLUTES AND THE GENERALIZATION OPERATOR G"
In Chapter III, section 4, a generalization operator G" was

postulated to account for the generic nature of sentences
like (40).
It was found that weak adjuncts can restrict the

interpretation of G" by specifying the set of intervals over
which generalization takes place, but that strong adjuncts
cannot function in this way. Here, it is shown that weak
absolutes, like weak adjuncts, may specify the set of
intervals for which some generalization holds, while strong
absolutes, like strong adjuncts, cannot.
(41) His left arm being in a cast, John doesn't like to

play volleyball.
(42) His father having been a wheat farmer, Johns works
at the Board of Trade.
(43) The water being a little cold, Bill stays on the
beach.
In none of these sentences can the absolute be inter-preted

as specifying a set of intervals over which the main clause
generalizesj (41), for instance, cannot mean 'when his left
arm is in a cast, John doesn't like to play volleyball'.
Instead, the strong absolutes in (41)-(43) are most
plausibly analyzed as expressions of category MTA, as I
shall show in section 3.
Now consider sentences (44)-(46), which have weak
absolutes.
(44) With her hair braided, Jane reminds us of Mary.

(45) With her children asleep, Mary watches TV.
(46) I-li th his work done, John goes straight to bed.
These absolutes are easily interpreted as specifying a set

of intervals over which the main clause generalizes. For
example, (44) can be understood as expressing a
generalization about time intervals at which Jane's hair is
braided; similar interpretations are possible for the
absolutes in (liS) and (lI6). Thus, weak absolutes can
apparently serve as set-level time adverbs restricting the
284 CHAPTER V
interpretation of the generalization operator G".

As was seen in the preceding subsection, weak absolutes
of category TA are automatically generated by means of
independently needed rules; accordingly, the generic
interpretation for sentence (44) is easily induced under the
analysis in (47), which yields the translation (48).
(47 ) with her hair braided, (Jane) [reminds]

us of Mary, t, 55,1
Jane,
T ------------
with her 1 hair braided, (she 1 ) [reminds]
us of Mary, t, 60
I
with her 1 bair braided, (she 1 ) [reminds]
us of Mary, TAB, 66a
~
with her, hair braided, (she,) [reminds] us of Mary,
TA, 9,9 TAB, 58
I
with her 1 hair braided,
I
[reminds] us of Mary,
~AB, 5b TAB, 56
her, hair, braided, PSPL s

I
(she,) [reminds] us of Mary,
T t, 19
(48) vt[G"CAt,Vt 2 [M 9 (t"t 2 ) &

AT(t 2 , Jane's-hair'(~AxiVxs[R(Xs ,xi) &
braided'(x S )]»])(At 1 [t 1 =t & [PRES(t 1 ) &

AT(t" Jane-reminds-us-of-Hary')]])]
Similar analyses for sentences such as (45) and (46) can be

straightforwardly provided.
Just as weak adjuncts were seen in Chapter III, section
4 to be able to restrict the interpretation of the
generalization operator G", it has been seen here that weak
absolutes can do so as well; on the other hand, strong
absolutes, like strong adjuncts, cannot function in this
way.'O Thus, there are at least three environments
motivating the distinction between weak absolutes and strong
absolutes; moreover, these are the same envLronments
motivating the parallel distinction between weak and strong
adjuncts.
3. Absolutes as Main Tense Adverbs
In the foregoing sections, I have shown that as members of

the category t' of conditional clauses, weak absolutes may
serve as the first argument of a modal verb, and that as
members of the category TA of set-level time adverbs, they
may additionally serve as the first argument of a relative
frequency adverb, or of the generalization operator G".
Here, I shall argue that absolutes may also function as main
tense adverbs--in fact, that all strong absolutes do so, as
do those weak absolutes not serving as arguments of the
above-mentioned binary operators.
The argument here is parallel to that presented in
section 2 of Chapter IV, where I showed that the free
adjunct in sentence (49) has exactly the same combinatory
possibilities as a main tense adverb; as I shall show, the
same can be said of the absolute in sentence (50).

(50) With John leading the project during the past
summer, we had made a lot of progress.
To see this, consider first the interpretation of

sentence (51).
(51) During the past summer, we had made a lot of

progress.
Clearly, (51) is ambiguous, because the adverb during the

past summer may be interpreted ei ther wi th respect to the
speaker's interval (denoting the set of subintervals of the
summer immediately preceding the utterance of (51» or with
respect to a past interval i (denoting the set of
subintervals of the summer immediately preceding i). These
two interpretations are possible because the main tense
adverb during the past summer may combine either with the
past tense or with the perfect to characterize some
interval, as in the translations (52) and (53):
(52) Vt[during-the-past-summer'(t) & [past(t) &

AT(t, vt1 [perf(t 1 ) & Vt 2 [t 2 =- t1 &
AT(t 2 , we-make-a-lot-of-progress')]])]]
286 CHAPTER V
(53) Vt[past(t) & AT(t, Vt, [during-the-past-summer'(t,)

& [perC(t,) & Vt 2 [t 2 ~ t, &
AT(t 2 , we-make-a-lot-oC-progress')]]])]
In the latter case, in which the time adverb joins with the
perfect, the adverb is within the scope of tense; this is
not so in the former case, however.
Now consider sentence (50) again. (50), like (51), is
ambiguous. On one interpretation, (50) entails that John
led the project during the most recent summer relati ve to
the speaker's interval; on the other interpretation, it
entails that John led the project during the most recent
summer relative to some past interval. These two
interpretations can be represented as in (54) and (55).
(54) Vt[K(L 6 )(AVt,[M g (t,t,) &

[during-the-past-summer'( t,) &
Ing(A[lead-the-project'(x s )])])]]) (A[past(t) &
AT(t, Vt,[perC(t,) & Vt 2 [t 2 ~ t, &
AT(t 2 , we-make-a-lot-of-progress')]])])]
(55) Vt[past(t) & AT(t, Vt,[K(L6)(AVt2[M9(t"t2) &

[during-the-past-summer'(t 2 ) &
AT(t 2 , Vxs[R(x s , John') &
Ing(A[lead-the-project'(x s )])])] ])(A[perf(t,) &
Vt 2 [t 2 ~ t, &
AT(t 2 , we-make-a-lot-oC-progress')]])])]
In the interpretation represented by (55), the absolute is

within the scope of tense, but not in the interpretation
represented by (54).
This striking similarity in interpretation between the
absolute in (50) and the time adverb in (51) is naturally
accounted for under the assumption that absolutes may serve
as main tense adverbs. Indeed, under the proposed analysis,
no additional rules are necessary in order to assign
sentence (50) the two interpretations in (54) and (55): by
rule S" (first proposed in section 2.3 of Chapter III), the
A-abstract underlying the absolute in (50) may be converted

into a main tense adverb, as in the derivations in (56) and
(57); as is easy to verify, these derivations induce (54)
and (55) as translations for (50).
(56) with John leading the project during the past

summer, (we) [had] made a lot of progress, t, 60
I
with John leading the project during the past
summer, (we) [had] made a lot of progress, TAB, 61a
with John leading the

-------------
project during the past
summer, HTA, 11,6,9
(we) [had] made a lot of progress, TAB, 57
I
(we) [have] made a lot of progress, TAB, 70
I I
with John leading the (we) [have] made a lot of progress, TAB, 62
project during the past ~
summer, TAB, 61b [have], (we) [make] a lot of progress, TAB, 56
~ TAB/ITAB I
during the past _______ (we) [make] a lot of progress,
summer, MTA, 63 with John leading t, 19
I the project, TAB, 5b
during the past ~
summer, TA John, leading the project,
T PRPL s , 45
with John leading the project during the past summer, (we) [had]
made a lot of progress, t, 60
I
with John leading the project during the past summer, (we) [had] made a
lot of progress, TAB, 57
I
with John leading the project during the past summer, (we) [have] made a
lot of progress, TAB, 70
I
with John leading the project during the past summer, (we) [have] made a
lot of progress, TAB, G1a
~
with John leading the (we) [have] made a lot of progress, TAB, 62
project during the past ~
summer, hTA, 11,6,9 [have], (we) [make] a lot of progress, TAB, 56
I TAb/ITAb I
with John leading the (we) [make] a lot of progress, t, 1~
project during the past summer,
~ with John leading the project,

during the past sUlIIlmer, IHA, 63 TAB, 5b
I ~
during the past summer, TA John, T leading the project, PRPL s , 45
The ambiguity in sentence (51) is a consequence of the
fact that the time adverb during the past summer may join
either with the perfect or with the past tense to
characterize a particular time interval; accordingly, a
288 CHAPTER V
similar ambiguity is not found in the simple past tense

sentence (58) corresponding to (51).
(58) During the past summer, we made a lot of progress.
In (58), the time adverb is only interpretable relative to

the speaker's interval, denoting the set of subintervals of
the summer immediately preceding (58)'s utterance; thus,
according to the translation (59) assigned to sentence (58),
the time adverb is outside the scope of tense.
(59) Vt[during-the-past-summer'(t) & [past(t) &

AT(t, we-make-a-lot-of-progress')]]
In view of these facts, one would predict that if the

absolute in (50) is indeed a main tense adverb, then the
ambiguity of (50) should disappear in the corresponding
simple past tense sentence (60).
(60) With John leading the project during the past

summer, we made a lot of progress.
This prediction is borne out: in (60), the absolute can

only be interpreted as being outside the scope of tense, as
in the translation (62) associated with the analysis in
(61) •
(61) with John leading the project during the past summer, (we) [made]
a lot of progress, t, 60
I
with John leading the project during the past summer, (we) [made] a lot of
progress, TAB, 61a ~
with John leading the project (we) [made] a lot oC progress, TAB, 57
--------
during the past summer, MTA, 11,6,9 I
I (we) [make] a lot of progress, TAB, 56
with John leading the project I
during the past su.mer, TAB, 61b (we) [make] a lot of progress, t, 19
during the past summer, tHA, 63 with John leading the project, TAB, 5b
I ~
during the past summer, TA John, T leading the project, PRPL s , 45
(62) Vt[K(L 6 )(A Vt ,[M 9 (t,t,) &
[during-the-past-summer'(t,) &
Ing(~[lead-the-project'(xS)])])]])(~[past(t) &
AT(t, ve-make-a-lot-of-progress')])]
Clearly, there is good reason to suppose that absolutes

may serve as main tense adverbs; in fact, I know of no
evidence against the stronger conclusion that any absolute
not serving as the first argument of a binary operator is a
main tense adverb.
4. Chapter Summary
In this chapter, the semantic similarities between free

adjuncts and absolutes have been elaborated. As was seen in
section 1, absolutes can be sorted into two distinct
subclasses according to their semantics in modal contexts:
as members of the category t' of conditional clauses, weak
absolutes may restrict the interpretation of a modal verb in
their superordinate clause; strong absolutes, on the other
hand, cannot function in this way. This distinction among
absolutes in modal contexts is strongly reminiscent of the
distinction between weak and strong free adjuncts,
originally motivated in Chapter II. It is, however,
slightly more complex: for an absolute to be weak, it must
not only have a stage-level predicate--it must in addition
be prepositionally augmented. .
In section 2, weak absolutes were found to share two
additional distributional peculiarities with weak adjuncts:
in sentences containing relative frequency adverbs, weak
absolutes may function as set-level time adverbs,
restricting the interpretation of the accompanying frequency
adverb; and in generic sentences, they may serve in an
analogous way to condition the interpretation of the
generalization operator G". Strong absolutes, like strong
adjuncts, are never employed in these ways.
In section 3, I argued that absolutes not serving as
the first argument of a binary operator are main tense
adverbs, since they can be shown to interact with tense in
precisely the same way as indubitable examples of main tense
adverbs. A similar argument for free adjuncts was developed
290 CHAPTER V
in Chapter IV.
Thus, the overall analysis of absolutes advocated in
this chapter closely parallels the analysis of free adjuncts
set forth in Chapters II-IV. In particular, inference is
assumed to play the same kind of role in the interpretation
of absolutes as in the interpretation of free adjuncts. For
example, absolutes of category MTA, like free adjuncts in
this category, are assigned intensional logic translations
containing the variables Land M, which represent semantic
indeterminacies that can only be resolved by means of
language users' inferences. Such inferences, I claim, are
what is responsible for' the relation of causation or
explanation that is felt to link the absolute with its
superordinate clause in (63), for the relation of temporal
succession in (64), and so forth.
(63) His father having been a sailor, John knows all

about boats.
(64) The three of them walked out in a huff, with Mary
closing the door behind them.
In the following chapter, the nature of such inferences will

be considered in depth.
Footnotes
1. There is one construction which may express a condition

and which, from a historical perspective, is an unaugmented
absolute; this is the provided that S construction:
(i) Provided that all is safe, you may go.

(Visser (1972:1295»
Synchronically, provided (that) is probably best analyzed as

a kind of subordinating conjunction--cf. Visser's
discussion, and footnote 5 in Chapter VI.
2. In recent years, a fair amount of attention has been

devoted to the syntax of augmented absolutes--absolute
expressions of the form 'with (out) T PRED'. For example,
much discussion has focused on the proper constituent
structure analysis of such expressions. Riemsdijk
(1978:62ff), on the basis of data from Dutch, presents
several arguments in favor of analyzing augmented absolutes
as consisting of three principal coconsti tuents (P - N'" -

X"', in his system); in general, his arguments apply
equally well to the corresponding English data. Ruwet
(1978) proposes a slightly different analysis for the
corresponding construction in French: in his analysis, the
third coconstituent isn't X"', but is, underlyingly, either
PP or S; in the latter case, a rule deleting the copula
allows the full range of predicative phrases to appear as
the third coconstituent in surface structure. On the other
hand, McCawley (1983) argues that augmented absolutes have
the underlying structure
[pp with [S NP VP]],
which may be subsequently altered by rules deleting the

sequence 'NP have' (see footnote 23, Chapter 11), deleting
be (see footnote 3 below), or replacing be with as (see
footnote 22, Chapter II). Likewise, Ishihara (1982:Ch.2)
proposes a 'with 5' analysis within the Government and
Binding framework, and follows McCawley in assuming the
existence of rules deleting be and have. (In Ishihara's
analysis, with is regarded as a 'prepositional
complementizer' comparable to for, so that unaugmented
absolutes derive drom augmented absolutes by means of
complementizer deletion.) Still other analyses have also
been proposed: Beukema (1980) argues for a
with ENP NP SIt ]
analysis; and Beukema (1982) argues for a 'with [S']'

analysis. In my view, if a sufficiently rich semantic
theory is assumed, then the arguments favoring analyses such
as those proposed McCawley, Ishihara, and Beukema cease to
be compelling; thus, the analysis presented here conforms to
Riemsdijk's treatment.
Another issue which has been widely discussed in the
recent literature on absolute constructions is the manner in
which such constructions are adjoined to their superordinate
clause; see Ishihara (1981), Sakakibara (1982), and Fukuchi
(ms) for representative discussion.
3. Sakakibara (1982), McCawley (1983), and Ishihara

(1982:62f) argue for the derivation of verbless with-
absolutes such as (1) by means of a be-deletion
transformation, so that (i) and (ii) are underlyingly alike.
292 CHAPTER V
(Ruwet (1978) argues for a similar rule in French.)
(i) With the children asleep, Mary might watch TV.

(ii) With the children being asleep, Mary might watch
TV.
The absolute in (i), however, is weak, while that in (ii) is

strong; given that they contrast in this way, a derivational
relation seems questionable. Cf. Riemsdijk (1978:66ff).)
A similar objection can be raised regarding Ishihara's
(1982:Ch.2) claim that nominative absolutes derive from
augmented absolutes through the deletion of with: a
derivational relationship between the absolute in (i) and
that in (iii) seems implausible, since the former is weak
while the latter is strong.
( iii) The children asleep, Mary might watch TV.
The same objection applies to Berent's (1975) claim that

augmented absolutes derive from nominative absolutes by
means of a with-attachment transformation.
4. Sakakibara (1982:86) in fact asserts that absolutes

augmented with with 'must describe a more or less temporary
state'--that is, in the terminology employed here, they must
have stage-level predicates. Cf. also Beukema (1979: 161).
5. This recalls Berent's (1975:20) claim that 'the less

clearly marked the subordinate status of [an] absolute
conjunct, the less acceptable [the] absolute sentence will
be! He argues that with, like a subordinating conjunction,
makes the subordinate status of an absolute more explicit,
thereby making the absolute easier to process. He doesn't
distinguish between weak and strong absolutes, however; as
was noted above, strong adjuncts actually seem to sound
better without with.
6. As was noted in section 1.2 of Chapter I, with isn't

the only preposition which may 'augment' an absolute;
without can do so as well, with negative force--
(i) Without her hair braided, Jane might resemble

Mary.
Absolutes augmented with without are not generated by the
proposed fragment.
In order to generate perfect absolutes such as those in
(ii) and (iii), rules (iv) and (v) are included in the
fragment presented in the Appendix.
(ii) His family having lived in Paris in '940, John

sometimes has nightmares about the war.
( iii) With Mary having resigned her position, a new
search committee had to be formed.
(iv) S7. If a E: PT , S E: PTABIITAB' and y E: PIVs, then

F 7a Ca,
S, y), F 7b Ca, S, y) E: P TAB , [-Tense,
+Adjunct, +Perfect], F 7a (a,S,Y) e: [-Weak],
and F 7b (a, S, y) E: [+vleakJ, where
F 7a (a, S, y) is
CONCa, DB(ING(CON(S, DB(EN( y»»» and

F 7b (u,S,Y) is CON(with,
CON(OBJ(cx), DBCINGCCON(S, DBCEN(y»»»).
Examples: F 7a (John, [have], [leave]) = John having

left;
F 7b (John, [have], [leave]) = with John having
left
T7. If a e: PT , S E: PTABIITAB' Y £ PIVs, and ex, S,

y translate as ex', S', y', then F 7a (ex, S, y),
F 7b (u, S, y) translate as
B'(~At[AT(t, u'(~AxiVxs[RCxs, xi) &
y'(x s )]»]).
Example: F 7a(John, [have], [leave]) translates as

At[perf(t) & vt,[t, ~ t &
AT(t" Vxs[R(x s , John') & leave'(x s )])]]
294 CHAPTER V
(v) S8. If a e; PT , ~ e; PTAB1ITAB' and y e; Prvi, then

F 8a (a, ~, y), F 8b (a, ~, y) e; PTAB , [-Tense,
+Adjunct, -Weak, +Perfect], where FSa(a,~,y)
is CONCa, DB(rNG(CON(~, DB(EN( y»»» and
FSb(a,~,y) is CON(with,
CON(OBJ(a), DBCING(CON( ~, DB(EN(y»»»).
Examples: FSa(John, [have], [be] a sailor) = John

having been a sailor;
F Sb(John, [have], [be] a sailor) = with John
having been a sailor;
TS. If a e; Pro ~ e; PTABIITAB' y e; Prvi, and a, ~,

y translate as a', ~', y', then FSa(a, 6, y),
FSb(a, ~, y) translate as
~'( AAt [AT ( t, a' (A y') ) ] ) •
Example; F8a (John, [have], [be] a sailor) translates

as A. t[perf(t) & vt, [t, ~ t &
AT(t" sailor'(John'»]]
7. (2S) does not apply to individual-level adjective

phrases, since the latter do not occur as predicates in the
nominative and augmented absolute constructions. Recall
that there is a similar restriction on the formation of free
adjuncts, as noted in Chapter III.
8. Note that the 'possessional with' in the prepositional

adjunct in (i) (discussed in Chapter II, section 3.4) is
distinct from the 'augment with' in the absolute in (ii).
(i) With green eyes, Mary might look something like

Jane.
(ii) With the weather so nice, it's hard to get much
work done.
In the present analysis, the latter sort of with is

semantically null; in fact, it does not even exist as an
independent lexical item, since it is introduced

syncategorematically by rules (27) and (28). Nevertheless,
as Ruwet (1978:191ff), Sakakibara (1982:84ff), and McCawley
(1983:277ff) observe, some expressions having the appearance
of augmented absolutes do seem to invite a possessional
analysis; thus, the boldface expression in (iii) receives an
interpretation something like that of (iv).
( iii) With a girl in every port, Harry feels pretty

contented.
(McCawley (1983:2T())
(iv) Harry has a girl in every port.
Under such an analysis, however, the boldface expression in

(iii) could not in fact be regarded as an absolute at all,
since it lacks any overt subject (or 'possessor'); instead,
it would have to be viewed as a free adjunct consisting
simply of a predicative prepositional phrase. The notion
that a prepositional phrase may have the form 'preposition T
PRED' is, of course, independently well-motivated by
examples such as (v) and (vi) (cf. Sakakibara (1982:86f):
(v) I saw John with Mary as his escort.

with his face painted blue.
with his hands in his pockets.
with his shoes all muddy.
(vi) The man with Mary as his escort whistled 'Manteca'.

with his face painted blue
with his hands in his pockets
with his shoes all muddy
To provide a formal account of the possessional with-phrases

in (iii), (v), and (vi), one might include with among the
expressions of category (Pps /T)/PRED s ; with would then
translate as in (vii) (where have" is the translation of the
have in (iv».
(vii) APsA piAxsVXo[R(XS,xo) & pi{"Ayivys[R(yS,yi) &

have" (ps)( yS) (x o )] } ]
Under this analYSis, the with-phrases in (iii), (v), and

(vi) are stage predicates, but phrases headed by have (such
as the predicate of (iv» are individual-level: this seems
296 CHAPTER V
correct, since the former but not the latter can appear in
the usual diagnostic environments for stage predicates;
furthermore, the former expressions serve as weak adjuncts,
the latter as strong adjuncts--
(viii ) With three-inch claws on each paw, a Bengal tiger

would be a match for any lion.
[on one interpretation, equivalent to (x)]
(ix) Having three-inch claws on each paw, a Bengal

tiger would be a match for any lion.
[not equivalent to (x)]
(x) If it had three-inch claws on each paw, a Bengal

tiger would be a match for any lion.
It might be thought that, just as the with in a 'with T

PRED' construction may be either empty or possessional, so
may the with in a 'with T' construction. Certainly the with
in (i) is possessional; and it might be plausibly suggested
that that in (xi) is empty, rather like that in (ii) (cf.
Riemsdijk (1978:73f), Ruwet (1978:173), McCawley
(1983:282».
(xi) With this nice weather, it's hard to get much work
done.
However, unlike adjuncts and absolutes generally, the with-

phrase in (xi) does not obviously express a proposition; for
this reason, it is unclear whether it should be regarded as
an adjunct (or absolute) at all, rather than simply as a
prepositional phrase serving as a sentence adverb (like that
in (xii».
(xii) In this nice weather, it's hard to get much work

done.
(The semantic relation designated by with in (xi) does,

however, seem rather indeterminate. In this respect, at
least, the with-phrase in (xi) is similar to adjuncts and
absolutes--all three constructions depend for their
interpretation on the inferential contributions of language
users.)
9. A potential problem with this generalization exists in

absolutes such as those in (i)-(iii).
Her hair braided, Jane sometimes reminds me of

Mary.
( ii) Her children asleep, Mary often watches TV.
(iii ) His work done, John never goes straight to bed.
These absolutes have stage-level predicates, and because

they are unaugmented, they are strong, according to the
findings in section 1. The question here is whether these
absolutes are more like those in (32)-(34) or like those in
(35)-(37) as regards their interpretation relative to the
accompanying frequency adverb. While it does seem to me to
be very difficult to interpret these absolutes as supplying
the temporal restriction on the cooccurring frequency
adverb, my judgement is that such an interpretation may
nevertheless be barely possible. If so, then the absolutes
in (i)-(iii) would appear to contradict the claim that only
weak absolutes may serve as the first argument of a
frequency adverb. Even if the problematic readings of (i)-
(iii) are granted, however, this claim can still be
maintained under a slightly different conception of the
distinction between weak and strong absolutes.
Suppose the disputed readings are granted. Then it is
clear that all absolutes which may restrict the
interpretation of a relative frequency adverb have one
characteristic in common: they all have stage-level
predicates. This suggests that prepositional augmentation
may not be a necessary feature of weak absolutes after all--
that the presence of a stage-level predicate may be both
necessary and sufficient as a characteristic of weak
absolutes, as is assumed in Stump (1981b). If this
assumption is made, then two desirable consequences follow:
(i) the distinction between weak and strong absolutes
becomes exactly like the distinction between weak and strong
adjuncts--in both cases, whether or not an expression is
weak depends purely and simply on whether its predicate is
stage-level; and (ii) the claim that only weak absolutes may
restrict the interpretation of a frequency adverb can be
upheld.
Still, to maintain that any absolute having a stage-
level predicate is weak, one would have to come up with some
reasonable explanation for the very obvious difference
bet ween augmented absolutes such as those in (35 )-(37) and
their unaugmented counterparts in (i)-(iii): in comparison
298 CHAPTER V
with the latter, the former are much more naturally

interpreted as supplying the temporal restriction on a
frequency adverb. This difference is even more pronounced
in modal sentences such as (4)-(9): in these sentences, it
is virtually impossible to interpret the absolute as the
first argument of the accompanying modal unless the absolute
is augmented; indeed, this fact is what originally motivated
the conclusion that weak absolutes must be augmented.
Clearly, if prepositional augmentation isn't a necessary
feature of weak absolutes, it is unquestionably the norm in
weak absolutes serving as arguments of modals or frequency
adverbs.
10. This generali zation might be disputed on the basis of

examples such as (i)-(iii), in which the absolutes have
stage-level predicates, but are unaugmented.
(i) Her hair braided, Jane reminds us of Mary.

( ii) Her children asleep, Nary watches TV.
(iii) His work done, John goes straight to bed.
These absolutes seem to be very marginally interpretable as

specifying the set of time intervals over which their
superordinate clause generalizes. If so, then like the
absolutes in (44)-(46), they can apparently serve to
restrict the interpretation of the operator G"--this despite
the fact that they are strong, according to earlier
conclusions.
On the basis of such evidence, one might attempt to
argue that any absolute having a stage-level predicate is
weak, regardless of whether or not it is augmented; cf.
footnote 9. It is clear, however, that the augmented
absolutes in (44)-(46) are much more naturally interpreted
as restricting the operator Gil than their unaugmented
counterparts in (i)-(Hi).
C HAP T E R VI
INFERENCE AND THE LOGICAL ROLE OF FREE ADJUNCTS

AND ABSOLUTES
In this chapter, the role of inference in the proposed

analysis of free adjuncts and absolutes is considered in
detail. As a preliminary to the ensuing discussion, the
formal semantic analysis of free adjuncts and absolutes
developed in Chapters II-V is briefly summarized in section
1; as I show, this analysis provides a 'mixed' account of
the semantic variability of adjuncts and absolutes,
attributing equal importance to rules of grammar and
language users' inferences as determinants of the logical
function of an adjunct or absolute.
In section 2, the status of language users' inferences
in the proposed system is clarified. Two sorts of
inferences are distinguished: those which may determine the
truthcondi tional meaninl;, of an expression, and tho'se which
cannot; inferences resolving the semantic indeterminacies
inherent in the free adjunct and absolute constructions are
claimed to be of the former type. The diverse factors which
apparently gi ve rise to such inferences are considered in
detail: among these factors are Carlson's ontology of
kinds, objects, and stages; the relative duration of the
events or states of affairs to which an adjunct or absolute
and its main clause relate; word order; knowledge of the
world; and the presence of 'connecti ve adverbs' of various
sorts.
In section 3, I discuss the possibili ty that the role
of inference in the interpretation of free adjuncts and
absolutes is less significant that I have claimed--that free
adjuncts and absolutes are not semantically indeterminate at
all, but instead derive from fully determinate adverbial
clauses; in section 4, on the other hand, I consider the
alternative possibility that I have attributed too little to
inference--that the logical role of an adjunct or absolute
is simply inferred in all cases. I argue that these extreme
positions, according to which the logical function of an
adjunct or absolute is either wholly dependent upon or
wholly independent of the inferences of language users,
299
300 CHAPTER VI
cannot be maintained.
Finally, section 5 is devoted to the theoretical
implications of the proposed analysis. In particular, I
shall address the followinb central issue: Is the proposed
analysis consistent with Montagovian compositionality? That
is, can the meaning of an absolute construction be regarded
as a function strictly of the meanings of its parts and of
the manner in which they are combined, or must language
users' inferences be reckoned as a third potential
determinant of the meaning of such a construction? My
discussion of this issue will necessarily be very general
and somewhat speculative; nevertheless, some significant
conclusions regarding this and other matters do seem to
follow from the account of free adjuncts and absolutes set
forth here.
1. Summary of the Proposed Semantic Analysis

As was shown in Chapter I, the semantic variability of free

adjuncts and absolutes has attracted the attention of
various traditional grammarians, who have developed
inventories of the logical roles which adjuncts and
absolutes may intuitively play and have, more tentatively,
attempted to explain why a given adjunct or absolute is felt
to play a particular sort of role. Two sorts of approaches
to this variability were observed: on the one hand,
Jespersen (1940) and Curme (1931) attempted to link the
specific logical role of a given absolute to some feature of
its internal or external syntax; on the other hand, Quirk,
et al. (1972) and Kruisinga (1932) argued that the
particular logical relation which a given adjunct or
absolute is felt to bear to its superordinate clause is
simpl y inferred from context. In the semantic account of
free adjuncts and absolutes proposed here, my thesis is that
the factors which determine the specific logical role of an
adjunct or absolute are of more than one kind--that the
nature of the understood relation between a free adjunct or
absolute and its superordinate clause is jointly determined
by the principles of English semantics and the inferences of
language users.
In Chapters II-V, a model-theoretic semantics was
developed in an attempt to elucidate some of the ways in
which the logical role of a given adjunct or absolute may be
INFERENCE AND LOGICAL ROLE 301
Table 4
The Formal Characteristics and Category Membership
of Weak and Strong Adjuncts and Absolutes
Free Adjuncts Absolutes
formal category formal category

characteristics membership characteristics membership
Weak stage-level t,l stage-level til

predicates TA2 predicates; TA2
prepositional
MIA3 augmentation MIA 3
Strong individual-level MIA 3 individual-level MTA3

predicates predicates and/or
lack of
prepositional
augmentation
1. t' is the category of conditional clauses; adjuncts and

absolutes of category t' are derived from A-abstracts by
means of rule S10.
2. TA is the category of set-level time adverbs; adjuncts and
absolutes of category TA are derived from A-abstracts by
means of rule S9.
3. MTA is the category of main tense adverbs; adjuncts and
absolutes of category MTA are derived from A-abstracts by
means of rule 311.
determined semantically, independently of the inferences of

language users. Throughout the discussion in these
chapters, an important distinction among adjuncts and
absolutes was reflected time and again: the weak/strong
distinction. The formal and functional correlates of this
distinction are summarized in Table 4.
The central conclusion drawn in Chapters II-V is that
certain sorts of expressions--modals, relative frequency
adverbs, and the invisible generalization operator G"--may
take weak adjuncts and absolutes as arguments, and that when
they do, they wholly or partially determine their logical
role in the sentence. Consider, for example, sentence (1).
With her hair braided, Mary would resemble Jane.

302 CHAPTER VI
On the most likely interpretation of this sentence, the

absolute serves as the first argument of the modal would (as
such, it functions as a member of the category t'); for this
reason, the conditional nature of this absolute is entirely
determined by the semantics of the sentence itself.
Similarly, the weak adjuncts in sentences (2) and (3) can be
interpreted as the first argument of the accompanying
relative frequency adverb or generalization operator;
(2) Driving home, John always listens to the radio.

(3) Driving home, John listens to the radio.
because frequency adverbs and the operator G" are

interpreted relative to sets of time intervals, the
semantics of sentences (2) and (3) determines the logical
role of their adjuncts insofar as it requires that they
denote sets of time intervals.
Nevertheless, it was found in Chapters II-V that the
logical relation between an adjunct or absolute and its main
clause is often determined inferentially. For example, even
though the semantics of sentence (2) requires that its
adjunct denote a set of time intervals (on the most likely
interpretation), it does not specify which set of intervals;
the adjunct in (2) can be inferred to have the denotation of
either of the set-level adverbs in (4) and (5) (among other
less likely denotations).
(4) Upon driving home, John always listens to the

radio.
(5) While driving home, John always listens to the
radio.
(This indeterminacy in the meaning of an adjunct of category

TA is represented as a free variable M over two-place
relations between time intervals in the translation assigned
to the adjunct in (2) by the fragment developed in Chapters
II-V:
(6) AtVt,[M(t,t,) & AT(t" Vxs[R(x s , xi) &

driving-home' (x s )])].)
More generally, any adjunct or absolute not serving as the

argument of a modal, frequency adverb, or generalization
operator (i.e. any adjunct or absolute, weak or strong,
serving as a main tense adverb) is semantically
indeterminate as to its logical role. Consider, for
example, sentence (7).
(7) Being thirty-five years old, John will be eligible

to run for president in 1988.
The relation between the strong adjunct and the main clause
of this sentence is semantically indeterminate in at least
two ways. First, the temporal relation between the adjunct
and the main clause is open: (7) can be inferred to mean
either that John is thirty-five now, or that he will be
thirty-five in 1988. Temporal considerations aside, the
relevance of the adjunct to the main clause in (7) is not
strictly specified by the semantics of English. One would
almost surely infer that the adjunct in (7) provides the
explanation for the truth of the main clause; note, however,
that in sentence (8), the very same adjunct would most
likely be inferred to bear an adversative relation to the
main clause.
(8) Being thirty-five years old, John will

nevertheless be ineligible to run for president in
1988.
(These two sorts of indeterminacy are represented by the
free variables Land M in the translation assigned to (7):
(9 ) Vt[K(L)(~vt1[H(t,t1) &
AT(t l' being-thirty-five-years-old l ( x O ))])

("[in-1988'(t) & [FUT(t) &
AT (t, John-is-eligible-to-run-for-president')]]) J.)
In the following section, the nature of the inferences

necessary for resolving the semantic indeterminacy inherent
in most free adjuncts and absolutes and the means by which
language users arrive at such inferences will be considered.
2. The Role of Inference in the InterpretatIon

In the analysis summarized in the preceding section, the

inferences of language users were claimed to play a central
role in the interpretation of free adjuncts and absolutes.
304 CHAPTER VI
In this section, this claim is developed. First, the nature

of the inferences relevant to the interpretation of absolute
constructions is scrutinized: it is argued that the logical
role inferred for an adjunct or absolute in some sentence
actually constitutes a part of the truthconditions for that
sentence; in this respect, such inferences play a much more
significant role in interpretation than has generally been
assumed. Second, a number of diverse factors which may give
rise to inferences of this kind are discussed in turn.
2.1. TWO KINDS OF INFERENCES

In the following discussion, the' li teral meaning' of an
expression a in some language L is to be that part of the
meaning of a which is assigned by the grammar of L.
Correspondingly, an 'inference' is to be understood as
anything which is not part of the literal meaning of some
expression but which language users judge to be part of the
intended meaning of that expression. 1 (Here' meaning' is
meant to cover nontruthconditional as well as
truthconditional meaning.) If inferences are so regarded,
then it is clear that two varieties of inferences can be
distinguished.
When a sentence ~ is uttered in a particular context,
an inference drawn from cP mayor may not constitute part of
the truthconditional meaning of ¢. That is, it may be that
the literal meaning of cP underdetermines its truthconditions
in the context in question; in this case, an inference must
be drawn if cP is to be assigned any sort of interpretation.
On the other hand, the literal meaning of ¢ may suffice to
determine the truthconditions of cP in the given context; in
this case, any inferences drawn from ¢ are inessential for
the assignment of an interpretation to~.
An example will help bring out this difference.
Consider sentence (10).
(10) I saw a Picasso today.
The noun phrase a Picasso in this sentence is an example of

what Clark (1978:313ff) calls 'shorthand expressions'. A
shorthand expression 4 picks out a coherent class of
referents bearing some salient relation to the explicit
content of 4: the expression a Picasso, for instance, can
denote (the property set of) some member of any of the
classes listed in (11) (among others).
(11) Picassos;
works of art by Picasso

portraits of Picasso
people named Picasso
people bearing a strong physical resemblance to
Picasso
people with a level of artistic aptitude
comparable to Picasso's
autographs by Picasso
forgeries intended to pass as works of art by
Picasso
postage stamps bearing the image of a work of
art by Picasso
To interpret a sentence containing a shorthand expression,

language users must infer which class of referents the
shorthand is intended to pick out. This is especially clear
in examples such as sentence (12).
(12) I saw two Picassos today.
(12) isn't necessarily true if the speaker merely saw two

members of the set {x: x is in some way connected with
Picasso}--indeed, (12) is intuitively false if what the
speaker saw was one budding lithographer and 'Les
demoiselles d'Avignon'. For (12) to be true, the speaker
must have seen two members of some coherent class of
referents comparable to those listed in (11); that is, in
order to assign an interpretation to sentence (12), one must
infer what sort of Picasso the speaker has in mind, for this
constitutes part of the truthconditional meaning of (12),
even though it is not part of its literal meaning.
The situation with sentence (13) is different.
(13) I saw a painting today.
Given the appropriate circumstances, one might infer that

the speaker of (13) has a specific sort of painting in
mind--a forgery (e.g. if the speaker is a forgery expert), a
portrait of Lincoln (if slhe is a Lincoln biographer), a
painting by Andy Warhol, etc. This inference would not,
however, count as part of the truthcon-di tional meaning of
306 CHAPTER VI
(13)--the truth or falsehood of (13) can be determined in

the absence of any sort of inference about the sort of
painting the speaker saw. In fact, the inferred meaning is
in this case nothing more than a conversational implicature.
In what follows, a sentence whose literal meaning
underdetermines its truthconditions will be said to be
inde~~~inate with respect to a certain range R of
interpretations; if the truth or falsehood of such a
sentence is to be ascertained, the intended interpretation
must be selected from R by inference. Thus, sentence (12)
is indeterminate with respect to a range of interpretations
corresponding to the various possible denotations for the
shorthand expression two Picassos. On the other hand, a
sentence whose literal meaning suffices to determine its
truthconditional meaning may be unspecific with respect to
some range of potential implicatures; some particular
implicature may be inferred to be part of the intended
meanin~ of a sentence which is unspecific in this way, but
the truth or falsehood of the sentence would never be
sensitive to such an inference. Thus, sentence (13) is
unspecific with respect to the various implicatures to which
the noun phrase a painting could potentially give rise.
The difference between inferences which resolve
indeterminacies and those which resolve unspecificities is
of great importance for understanding the nature of free
adjuncts. Recall from Chapter I, section 2 the claim of
Quirk et ale (1972) that a logical role is inferred for a
free adjunct in essentially the same way as for a
nonrestrictive relative clause or coordinate clause--that
the possibility of inferring a causal connection in (14) is
paralleled in (15) and (16).
(14) John, being an Englishman, is brave.

(15) John, who is an Englishman, is brave.
(16) John is an Englishman, and he is brave.
In Chapter I, I disputed this claim on the grounds that the

causal connection in (14)--unlike that in (15) or (16)--
intui ti vely counts as part of what the sentence is used to
assert. This difference between (14) and (15),(16) can now
be stated in a slightly different way: while (15) and (16)
are merely unspecific with respect to the logical connection
between John's nationality and his courage, (14) is in fact
indeterminate with respect to this connection--to determine
the truth or falsehood of (14), one must infer some sort of
logical connection. This, I claim, is a general fact about
sentences with free adjuncts and absolutes: when the
logical role of the adjunct/absolute is not entirely
determined by the semantics of English (i.e. when the
adjunct/absolute isn't the first argument of a mOdal), such
sentences are indeterminate, not merely unspecific, with
respect to the logical relation holding between their
adjunct/absolute and its superordinate clause. Thus, what I
have represented with the free variables Land M in the
translations produced by the fragment developed in Chapters
II-V are indeterminate rather than unspecific aspects of a
sentence's meaning.
\ihat I shall examine in the following subsection are
the means by which language users resolve the indeterminacy
of sentences with free adjuncts and absolutes--that is, by
which they determine admissible values for the variables L
and M.
2.2. INFERENCES, ADJUNCTS, AND ABSOLUTES
The inferences by which language users resolve the semantic

indeterminacy of free adjuncts and absolutes appear to be
motivated by a variety of factors; here, I will discuss what
I perceive to be the most imlJortant of these factors.
First, whether an adjunct or absolute has a stage-level or
an individual-level predicate bears importantly on the
logical role i t is inferred to have. Second, the relative
duration of the events or states of affairs to which an
adjunct or absolute and its main clause relate places
certain limitations on the temporal relation inferred to
hold between the adjunct/absolute and the main clause
(whether or not the principal function of the
adjunct/absolute is that of a time adverb). Third, the
superficial order of an adjunct or absolute and its main
clause can be understood as iconic. Fourth, language users'
general knowledge of the world powerfully constrains the
range of logical roles that can be inferred for an adjunct
or absolute. Finally, certain sorts of 'connective adverbs'
can be understood as explicitly indicating the relation
holding between an adjunct or absolute and its superordinate
clause. Each of these considerations will be examined in
turn in the following five subsections.
308 CHAPTER VI
2.2.1. Carlson's Ontology
In Chapters II-V, Carlson's ontology of stages, objects, and

kinds was seen to play a central role in the interpretation
of free adjuncts and absolutes. In particular, it was shown
to be the basis for the distinction between weak and strong
adjuncts and absolutes: weak adjuncts and absolutes always
have stage-level predicates, and strong adjuncts and
absolutes generally have individual-level predicates
(although strong unaugmented absolutes may contain stage-
level predicates, according to the proposed analysis).
Recall that weak adjuncts and absolutes may serve as the
first argument of a modal (in the manner of an if-clause) or
as the first argument of a frequency adverb or
generalization operator (in the manner of an explicitly
temporal adverbial clause); in addition, when functioning as
main ten.:se adverbs, weak adjuncts and absolutes are often
felt to play an essentially temporal role, as in (17).
( 17) a. Crossing the street, he was almost hi t by a

car.
b. With the rain just starting to fall, we
reached St. Louis.
Strong adjuncts and absolutes, on the other hand, are often

felt to express reasons or causes; this is particularly true
of those strong adjuncts and absolutes whose predicates are
individual-level, as in (18).
(18) a. Having blue eyes, Jane looks a lot like Mary.

b. His father having been a sailor, John knows
all about boats.
These facts strongly suggest that there is something about

stage-level predications which makes them inherently more
suitable than individual-level predications for the
expression of temporal or conditional relations, and that
there is something about individual-level predications which
makes them naturally more suitable for expressing reasons
and causes.
On first consideration, this hypothesis appears to be
falsified: it is, after all, perfectly possible to have an
explicit because-clause with a stage-level predicate;
Because John heard Mary, we'll have to scrap the
plan.
similarly, explicitly temporal or conditional adverbial

clauses sound fine with individual-level predicates.
When John was a sailor, he traveled allover the

world.
If Bill were intelligent, he'd be able to figure
this out.
Nevertheless, careful consideration reveals that there is in

fact a pragmatic correlation between individual-level
predications and the expression of reasons and causes, and a
similar correlation between stage-level predications and the
expression of temporal relations. (The correlation between
stage-level predications and the expression of
condi tionali ty remains mysterious to me; see below.)
Recall the nature of stage-level properties--the
'accidental' properties exhibited by the space-time slices
of an individual: the interval at which a stage has a given
stage-level property is characteristically comparatively
short and discrete; those intervals at which stages of some
individual have a given stage-level property are often
discontinuous. Stage-properties contrast in these respects
with basically individual-level properties ('essential'
properties): the interval at (every moment of) which an
individual possesses such a property is normally
comparati vely long and lacks definite boundaries; an
individual-level property most often applies to a particular
individual (at every moment) throughout a single, continuous
interval. (As Dowty (1979: 179) observes, individual-level
predicates express the dispositions or potentialities of an
individual.) These differences 2 between the properties of
stages and individuals make the former much more sui table
for locating relatively short and discrete events or states
of affairs in time. One may locate a given event or state
quite precisely in time by relating it to the interval at
which the stage of some individual has some (stage-level)
property or other; for example, in (19), the interval at
which John fell down the stairs is located fairly exactly by
being related to that at which some stage of John was drunk.
(19) When John was drunk, he fell down the stairs.

310 CHAPTER VI
If an event or state is located by reference to the interval

at which some individual has a certain (basically)
individual-level property, much less precision is achieved:
(20) When John was a drunk, he fell down the stairs.
Thus, reference to the properties of some individual's

stages is important for establishing relatively precise
temporal relations with adverbial subordinate clauses, free
adjuncts, or absolutes.
This can be seen especially clearly in the case of
adverbial clauses, adjuncts, or absolutes serving to
restrict a relative frequency adverb or generalization
operator. The first argument of a relative frequency adverb
specifies the normally discontinuous set of time-intervals
over which the adverb quantifies. To the extent that the
membership of this set is ill-defined or its members vague,
the resulting quantification lacks informativeness; likewise
this is so if this set consists of only a single, continuous
interval. Accordingly, adverbial clauses, free adjuncts,
and absolutes with individual-level predicates (like when
John is a drunk, being a drunk, and (with) John bein.g a
drunk) aren't useful as arguments of relative frequency
adverbs. (Note the oddity of When he isa drunk/Being a
drunk, he oCten falls down the stairs if the adverbial
expressions are construed as restricting the interpretation
of oCten.) Similarly, adverbial clauses, adjuncts, and
absolutes restricting the operator Gil specify the normally
discontinuous set of intervals over which some
generalization holds; to the extent that they are vague in
the set of intervals they characterize, the resulting
generalization is uninformative; this is likewise the case
if they characterize a Single, continuous interval. For
this reason, adverbial clauses, free adjuncts, and absolutes
with individual-level predicates aren't very useful as
arguments of Gil. (Note the oddity of When he is a
drunk/Being a drunk, he Calls down the stairs if the
adverbial expressions are construed as restricting Gil.)
Clearly, there is a pragmatic correlation between stage-
level properties and the expression of temporal relations in
those instances in which an adverbial clause, free adjunct,
or absolute restricts a frequency adverb or generalization
operator: in these instances, only those adverbial
expressions with stage-predicates are uniformly good for
picking out the desired sort of set of intervals.
An additional piece of evidence suggesting that stage-

level predications are better suited than individual-level
predications for the expression of temporal relations is the
fact that augmented adjuncts with temporal subordinators
often sound quite odd with individual-level predicates; that
is, though one can say when he was a sailor, it sounds
strange to say when (being) a sailor.
In general, then, stage-level predications tend to be
better suited than individual-level predications for
expressing temporal relations. This tendency is apparently
one reason for the fact that language users routinely infer
that the logical role of a weak adjunct/absolute of category
MTA is temporal in nature, and for the fact that weak
adjuncts and absolutes, but not stron6 ones, may (as members
of category TA) restrict the interpretation of a relative
frequency adverb or generalization operator.
It can also be appreciated why adjuncts and absolutes
with individual-level predicates are well-suited to the
expression of reasons and causes. As was pOinted out above,
individual-level predicates express an individual's
'essential' properties--its dispositions or potentialities.
The essential properties or disposi tions of an indi vidual
are, of course, naturally viewed as being among the reasons
or causes for that individual's behavior; that is, the cause
of or reason for some event or state of affairs is routinely
sought among the dispositions of its participants. For this
reason, individual-level predications are good for
specifying reasons and causes; it is, therefore, not
surprising that language users tend to infer an explanatory
or causative role for strong adjuncts and strong absolutes
whose predicates are individual-level.
Thus, the pragmatic correlation between stage-
predicates and the expression of temporal relations and that
between individual-level predicates and the expression of
reasons and causes are not especially difficult to account
for, if one keeps in mind the fundamental distinction
between stage-properties and properties of individuals.
Indeed, the fact that these correlations do appear to be so
well motivated raises an important question: should the
distinction between weak and strong adjuncts and absolutes
be regarded as a formal, categorial distinction, or merely
as a pragmatic distinction? Recall that in the analysis
proposed in Chapters II-V, the weak/strong distinction was
formalized as a two-way subclassification of A-abstracts by
the feature [.:tWeakJ. As a way of guaranteeing that strong
312 CHAPTER VI
adjuncts and absolutes would never serve to restrict the

interpretation of a modal or frequency adverb, or of the
generalization operator G", the rules S9 and S10 converting
A-abstracts into expressions of categories TA and t' were
stated so as to apply only to members of the subclass
[+Weak); thus, in the proposed system, the weak/strong
distinction has the status of a category distinction. But
because the pragmatic properties of stage- and individual-
level predicates are so different, one might suppose that a
purely pragmatic account of the weak/strong distinction is
possible--that the distinctive semantic characteristics of
weak and strong adjuncts and absolutes are nothing more than
a reflection of the fact that stage-level predications are
well suited for the expression of temporal relations, while
indi vidual-level predications are well sui ted for the
expression of reasons or causes.
In Stump (1981b), I assumed that a pragmatic account of
this kind was feasible. Indeed, it does have a certain
amount of plausibility. Consider, for example, sentence
(21) •
(21) Being a sailor, John sometimes smokes a pipe.
In the analysis proposed in Chapters II-V, the strong

adjunct in (21) cannot serve as the first argument of the
frequency adverb sometimes because the latter expression
requires an argument of category TA, and the only adjuncts
belonging to this category are weak. But if individual-
level predications really are inherently bad for expressing
temporal relations (as was claimed above), then perhaps no
other explanation is needed for the fact that the strong
adjunct in (21) cannot be understood as restricting the
meaning of the accompanying frequency adverb: perhaps this
interpretation is simply excluded as pragmatically
anomalous.
Despite the initial plausibility of the claim that the
weak/strong distinction is a pragmatic one, there are at
least three phenomena which clearly cast doubt on this
claim. First, the fact that unaugmented absolutes with
stage-level predicates are strong shows that the semantic
differences between weak and strong absolutes cannot be
purely a consequence of the pragmatic differences between
stage-level and individual-level predicates. Moreover, it
is not clear how the pragmatic approach to the weak/strong
distinction could be modified to account for the fact that
weak absolutes require prepositional augmentation.
Second, it is not clear that there is a good pragmatic
explanation for the fact that weak adjuncts and absolutes
may serve as the first argument of a modal, while strong
adjuncts and absolutes cannot: I can think of nothing about
stage-level predications which somehow makes them more
useful for expressing conditions than individual-level
predications. There are, admittedly, certain parallels
between temporal adverbial clauses and conditional clauses--
cf. the idioms if and when, unless and until, and the fact
(noted by Lewis (1975)) that if-clauses can sometimes
restrict the interpretation of a relative frequency adverb
('adverb of quantification'); but it is far from obvious
that the special usefulness of stage-level predications for
expressing temporal relations is somehow responsible for the
fact that only weak adjuncts and absolutes may play the role
of a conditional clause. (This is perhaps the strongest
argument for maintaining a categorial distinction between
weak and strong absolute constructions.)
A final difficulty with the pragmatic account of the
weak/strong distinction arises in connection with the
problematic class of individual-level predicates consisting
of be 1 (see Chapter II, section 3.2.1) plus some stage-level
predicative phrase:
(22) bel + PRED s
be asleep
be in the barrel
be ready to leave
etc.
Recall that a number of diagnostics (the ability to occur in

the progressive, the ability to occur postnominally in
existential there sentences, the ability to occur in the see
them drunk construction, generalizability) all suggest that
such expressions are individual-level predicates, as does
the fact that they give rise to strong rather than weak
adjuncts and absolutes. The problem is that such
expressions do not express essential properties or
dispositions; on the contrary, the properties they express
are no less accidental nor less temporally bound than those
expressed by the stage-level predicative phrases from which
they derive (compare, for exaP1ple, the den.otations of asleep
and be asleep: asleep', AX 1 VX s [R(X s ,x 1 ) & asleep'(x s )]).
314 CHAPTER VI
Given this fact, the pragmatic properties of the adjuncts in

(23) should be identical, as should those of the absolutes
in (24).
a. Asleep, Fido is harmless.

b. Being asleep, Fido is harmless.
a. Wi th her children asleep, Mary occasionally
watches TV.
b. With her children being asleep, Mary
occasionally watches TV.
The obvious contrast between the adjuncts in (23) and

between the absolutes in (24) casts real doubt on the
prospects of reducing the weak/strong distinction to the
pragmatic differences between stage-level and individual-
level predicates.
Adjuncts and absolutes having individual-level
predicates such as those in (22) pose problems not only for
the pragmatic account of the weak/strong distinction; they
also pose potential problems for the claim (advanced above)
that adjuncts and absolutes with individual-level predicates
are inferred to express causes or reasons because of their
dispositional interpretation. Clearly, the logical relation
inferred for the adjunct in (23b) is just as likely to be a
because-type relation as the relation inferred for the
adjunct in (25); yet, unlike the adjunct in (25), the one in
(23b) doesn't express an essential property or disposition.
(25) Being a chihuahua, Fido is harmless.
This problem arises generally with adjuncts and absolutes

whose predicates are as in (22): since they don't express
essential pr:'operties or dispositions, why are they so
readily inferred to play a causal or explanatory role?
I admit I have no really satisfactory solution to this
problem. It may be that some special property of the verb
bel (above and beyond that of 'shifting' a stage-predicate
to the individual level) is responsible for this peculiarity
of predicates like be asleep, but I haven't been able to
find convincing evidence of any such property. Another
possibility is that the logical role of an adjunct or
absolute with being isn't inferred at all in Modern
English--that such constructions are an idiom with the
approximate meaning of a because-adverbial. A tantalizing
piece of evidence in favor of this point of view are the
being as (bow) and being (tbat) constructions present in
earlier English and in some contemporary American dialects
(mine included):
(26) You loyter heere too long, being you are to take
souldiers up.
(Shakespeare, Henry IV, Part 2, 1597, cited
by Visser (1972:1218)-)- - - -
(27) Being as you are a giant, you move naturally in

seven league boots.
(Visser (1972:1218»
These are unquestionably idiomatic constructions used

exclusively to express reasons and causes; perhaps adjuncts
and absolutes with being are analogous. 3 An additional
piece of evidence favoring the view that some sort of
idiomatization has taken place is the fact that being-type
adjuncts can seemingly play a broader range of logical roles
in languages other than modern English, and could also
apparently do so in earlier stages of English. In French,
for example, etant-adjuncts may be inferred to play a purely
temporal role:
(28) Etant petit gar~on, j'ai habite a Geneve.

'When I was a little boy, I lived in Geneva.'
Compare the following example from 17th century English:
(29) being young, he was a pensioner in the house

of this Ammonius.
[= Modern English 'when he was young ••• ']
(Dryden, Life of Plutarch, 1683, cited by
Visser (1972:1136))
In modern English, the free adjuncts being a little boy or

being young would not be inferred to play the sort of role
exemplified in (28) and (29). Since the ways in which
people draw inferences are presumably uniform from one
population to another, this cross-linguistic difference in
the interpretation of being-type adjUncts 4 would seem to
suggest that such adjuncts (and the corresponding absolutes)
are developing into idiomatic expressions of causation or
explanation in modern English. A parallel sort of process
has evidently already taken place in the being as (bow) and
316 CHAPTER VI
being (that) constructions, as well as in the seeing as

(how), seeing that, and given (that) constructions; the
latter developments reflect the strong and very general
tendency of free adjunct and absolute constructions to
develop historically into prepositional and adverbial clause
constructions. 5
Before proceeding to a discussion of how language users
determine the nature of the temporal relation holding
between an adjunct or absolute and its superordinate clause,
I wish to point out a relevant difference between absolute
constructions with stage-level predicates and those whose
predicates are individual-level. Recall that in section
1.4.4 of Chapter III, it was seen that, for semantic
reasons, when-clauses and while-clauses must agree in tense
with their superordinate clause; furthermore, sentences like
(30) and (31), though noncontradictory, were seen to be
pragmatically anomalous.
(30) *John will leave after Mary arrived.

(31) *John left before Mary arrives.
The upshot of these facts is that temporal adverbial clauses

must in general relate to an interval satisfying the same
tense predicate ('past( ~)', 'PRES( n', or 'FUT( n') as the
interval to which their superordinate clause relates, if
they are to be semantically and pragmatically acceptable. 6
Now, because adjuncts and absolutes with stage-level
predicates are characteristically inferred to have the
function of temporal adverbials, it is not surprising that
they are normally inferred to relate to an interval of the
same 'tense' as the interval to which their superordinate
clause relates. Consider, for example, sentences (32) and
(33) •
(32) vforking at the post office during the past summer,

(33) *Working at the post office during the past
summer, Mary will go back to school this spring.
In both (32) and (33), the adjunct relates to a past

interval, because of the adverb during the past summer which
modifies it. For this reason, sentence (33), in which the
superordinate clause is future in tense, sounds qui te bad.
«33) can, of course, be repaired by converting its adjunct
to perfect form:
(34) Having worked at the post office during the past

summer, Mary will go back to school this spring.
In this sentence, the adjunct itself need not relate to a

past interval--rather, it need only relate to some interval
at which it is true that Mary worked at the post office
during the preceding summer. This can be a future interval,
so (34) is fine.)
Adjuncts and absolutes with individual-level predicates
present a different situation. Adverbs of causation and
explanation, unlike temporal adverbials, are temporally
independent of their superordinate clause; this is, of
course, because the truth of a proposition p at index i may
qualify as a reason or cause for the truth of some other
proposi tion q at index j even if i and j satisfy different
tense predicates. For example, (35) is perfectly
acceptable:
(35) Because she is thirty-five years old, Mary will be

eligible to run for president in 1988.
Because absolute constructions with individual-level

predicates are so well suited for expressing causes and
reasons, it should not be any surprise that, unlike weak
adjuncts and absolutes, they can be inferred to relate to
some interval distinct in 'tense' from the interval to which
their superordinate clause relates. Thus, the adjunct in
(36) can be understood to relate either to the speaker's
interval or to 1988.
(36) Being thirty-five years old, Mary will be eligible

to run for president in 1988.
Thus, the difference between stage-level and

individual-level predicates has consequences for the
inferred time reference of free adjuncts and absolutes. In
the following subsection, another sort of consideration will
be seen to enter into the inference of a temporal connection
between a free adjunct or absolute and its superordinate
clause.
318 CHAPTER VI
2.2.2. Instantaneousness
Free adjuncts and absolutes of category MTA are always felt

to bear some sort of temporal relation to their main clause,
whether or not their principal function is felt to be that
of a temporal adverbial. As has been seen, strong adjuncts
and absolutes may relate to the speaker's interval; in most
cases, however, adjuncts and absolutes of category MTA
relate to some interval satisfying the same tense predicate
as the interval to which their superordinate clause relates,
and additionally bearing some sort of relation to it. (This
relation is represented by the free variable M in an
expression such as (37), the translation assigne2 to the
sentence Walking home, John found a dollar--
(37) Vt[K(L)(A vt1 [M 9 (t,t 1) & AT(t 1, Vxs[R(xs,x i ) &

Ing(A[valk-home'(xs»)))))(A[past(t) &
AT(t, John-finds-a-dollar'»)).)
In some cases, this relation is highly constrained by time

adverbs: in (38), for instance, the adverbs at 5:01 and at
5;07 require that the value of Mg in (39) be a relation
holding between 5:01 and 5:07.
(38) Leaving the station at 5:01, he arrived at the

bank at 5:07.
(39) Vt[K(L)(A Vt1 [M 9 (t,t 1) & [5:01' = t1 &

AT(t 1, Vxs[R(xs,x i ) &
Itmg(A[leave-the-station'(x s »)))) ])(A[5:07' =t &
[past(t) & AT(t,he-arrives-at-the-bank'»)))]
In most cases, however, lantSuage users must infer the

precise nature of the temporal relation holding between an
adjunct or absolute of category MTA and its superordinate
clause. The relative duration of the event or state of
affairs to which an adjunct or absolute relates and that to
which the main clause relates can form the basis for such an
inference. There are three broad cases to be considered:
1) that in which the free adjunct/absolute and the main
clause both relate to instantaneous events; 2) that in which
the free adjunct/absolute relates to an instantaneous event
and the main clause to a state of affairs or
non instantaneous event, or vice versa; and 3) that in which

the free adjunct/absolute and the main clause both relate to
states of affairs or non instantaneous events. Each of these
cases will be discussed in turn.
1) The adjunct/absolute and the main clause both relate
to instantaneous events.
This case covers sentences in which both the
adjunct/absolute and the main clause have instantaneous
achievement predicates. In such instances, language users
infer that the events to which the adjunct/absolute and the
main clause relate are simultaneous or successive. In
sentence (40), for example, it may be inferred that John
realized Bill's house was on fire just as he noticed the
smoke or sometime afterwards.
(40) Noticing the smoke, John realized Bill's house was

on fire.
The nature of instantaneousness precludes the possibility

that a relation of proper containment should be inferred.
Note here that states aren't normally felt to be
instantaneous; for instance, it's very difficult to come up
with a convincing example of a sentence in which a state may
be inferred to be no more than simultaneous with an
instantaneous event--this despite the fact that a stati ve
predication true at interval i is true at every moment in i.
Perhaps the truth of a stative predication p at some moment
m entails the truth of p at some proper superinterval of m.
If this is the case, then whenever an inference of
Simultaneity can be drawn between a state and an
instantaneous event, the stronger inference can be drawn
that the interval of the state properly contains the instant
of the event; the ever-present possibility of this stronger
inference accounts for the apparent incompatibility of
stativity and instantaneousness.
2) The adjunct/absolute relates to an instantaneous
event and the main clause to a state or noninstantaneous
event, or vice versa.
In sentences covered by case 2, the free
adjunct/absolute has an instantaneous achievement predicate
while the main clause has a stative, activity,
accomplishment, or noninstantaneous achievement predicate;
or vice versa. In such instances, language users infer a
relation of proper containment or of succession. for
example, in (41), it mCiY be inferred that Mary spotted the
320 CHAPTER VI
heron while she was walking home or, somewhat less probably,
after she got home; in (42), it seems quite natural to infer
a relation of succession.
(41) Walking home, Mary spotted a heron.

(42) John climbed down the well, discovering a sealed
metal box at the bottom.
In no case is a relation of simultaneousness inferrable,

owing to the sorts 6f predicates involved.
3) The adjunct/absolute and the main clause both relate
to states or non instantaneous events.
Subsumed under this case are sentences in which both
the adjunct/absolute and the main clause have stative,
activity, accomplishment, or non instantaneous achievement
predicates. In such instances, language users may infer
that the intervals to which the adjunct/absolute and the
main clause relate stand in a relation of identity, proper
containment, or succession. In (43), one might infer that
John sang throughout or at some period during his walk
beside the ri ver.
(43) Walking beside the river, John sang.
In (44), it may be inferred that the lowering preceded the

lying.
(44) Lowering himself laboriously onto the cot, Bill

lay exhausted.
Clearly, sentences subsumed under case 3 are subject to the

broadest range of possible inferences: since both the
adjunct/absolute and the main clause relate to events or
states having duration, the possibility of inferring a
relation of proper containment isn't ruled out (as it is in
case 1), nor is the possibility of inferring a relation of
simultaneity (as it is in case 2). Observe that a relation
of succession is, in principle, possible for all cases.
Thus, the nature of the temporal relation holding
between an adjunct or absolute of category MTA and its
superordinate clause is inherently limited by the duration
of the events or states of affairs to which these relate.
Although I won't demonstrate this in detail here, it can be
analogously shown that the set of intervals denoted by an
adjunct or absolute (of category TA) restricting a relative
frequency adverb/generalization operator is inherently

limited in an analogous way.
2.2.3. Word Order
One purely syntactic characteristic of free adjuncts and

absolutes which seems to influence language users when they
infer a logical role for such expressions is word order.
The most important respect 7 in which word order appears to
affect language users' inferences is as an iconic indication
of temporal succession. In (45), for instance, the counting
may be understood to follow the shouting (or, on a less
likely interpretation, to overlap), but not to precede it.
(45) Hilary shouted his name, counting the number of

echoes.
This is surely the effect of a familiar perceptually-based

phenomenon whereby a succession of words is felt to mirror
the succession of the events they describe (the same
phenomenon guides our understanding of coordinate
structures--cf. so-called 'asymmetric conjunction'--and of
separate sentences in discourse). If the free adjunct and
the main clause of (45) are placed in the reverse order, as
in (46), the counting may be understood to precede the
shouting--this is in fact the most likely interpretation. 8
(46) Counting the number of echoes, Hilary shouted his

name.
2.2.4. Knowledge of the World
A language user's knowledge of the world can impose enormous

constraints on the sorts of logical roles slhe will infer
for a free adjunct or absolute. 9 Knowledge of the world
may, for instance, determine the nature of the temporal
relation inferred to hold between an adjunct/absolute and
its superordinate clause, as when a language user concludes
from (47) that the interval of Bill's driving from Cheyenne
to Salt Lake City properly contains that of his crossing the
Rockies, on the basis of his/her knowledge of geography and
of driving (note that a different inference is produced if
the Nevada border is substituted for the Rockies in (47».
322 CHAPTER VI
(47) Bill drove from Cheyenne to Salt Lake City,

crossing the Rockies while it was still light.
Similarly, a language user's expectations as to how certain

kinds of events normally succeed one another might cause
him/her to infer from (48) that John's gaining ownership
followed his payment of enormous sums.
(48) Gaining full ownership of the company only after

several years, John paid enormous sums to one
stockholder after another.
(Observe that the iconicity of word order is overruled

here.)
Knowledge of the world can also give rise to inferences
about nontemporal relations between adjuncts/absolutes and
their superordinate clauses. For example, one's
expectations regarding how a person's preferences influence
what s/he agrees to do would lead one to infer an
adversative role for the adjunct in (49).
(49) Preferring to go dancing, Mary eventually agreed

to go skydiving.
(Knowledge of the world also enables language users to

infer that a present participial adjunct or absolute with an
accomplishment or achievement predicate is perfective (or
imperfective), and thus to eliminate the unspecificity
inherent in such constructions. For instance, expectations
regarding where cars drive would lead to an inference that
the adjunct in (50) is imperfective.
(50) Crossing the street, John was hit by a car.)
Clearly, the role of language users' real world

knowledge as a basis for drawing inferences about the
logical connection between a free adjunct or absolute and
its superordinate clause cannot be overestimated.
2.2.5. Connective Adverbs
In certain cases, the logical relation borne by a free

adjunct or absolute to its superordinate clause is spelled
out explicitly by a specialized sort of adverb. Adverbs of
this sort have been widely noted for their connective

function (cf, Halliday & Hasan (1976:Ch.5), van Dijk
(1977:Ch.3), and Warner (1979»; examples are those listed
in (51).
(51) Connective adverbs:
meanwhile therefore
then hence
subsequently thus
afterwards consequently
beforehand thereby
furthermore still
moreover anyway
in addition nevertheless
besides nonetheless
also instead
As the following examples show, connective adverbs of this

sort can determine the relation holding between a free
adjunct or absolute and its superordinate clause.
(52) Very slowly, he began to draw out the wire,

meanwhile shaking it enough to preserve the
interest of whatever might be on the other end.
(CL, 21-22)
(53) Thought by some to be a part of an otherwise

vanished circle, the standing stones of Kilchattan
have also been described as possible remnants of a
Druid astronomical observatory.
(CL, 146)
(54) Goats were originally brought to Colonsay because

it was known that they would assert territorial
rights to the highest ground, thus keeping sheep
away from the crags and contributing to their
safety.
(CL, 31-32)
324 CHAPTER VI
(55) Tourists who gawk at skyscrapers, look at maps, or

window shop with all the awe of a child at
Christmas stick out like sore thumbs, thereby
becoming easy marks.
(Woodall's Trailer and RV Travel/West 12/77,
11)
(56) Bleeding so profusely, amnesic and distraught, the

doctor nonetheless backed his Land-Rover through
the hedge, and, as if nothing had happened, went
on to collect his milk.
(CL, 48)
(57) Being very uncomfortable, he still fell asleep.

(Halliday & Hasan (1976:229»
Connecti ve adverbs can be used to specify a temporal

relation (cf. meanwhile in (52», a more or less neutral,
conjunctive relation (cf. also in (53», a relatioll of
consequence (cf. thus and thereby in (54) and (55», or an
adversative relation (cf. nonetheless, still in (56) and
(57». Note that the adverb may appear either within the
adjunct/absolute (as in (52), (54), and (55» or within the
main clause (as in (53), (56), and (57».
Despite the fact that the connective adverbs in (52)-
(57) explicitly indicate the logical role of the
accompanying adjuncts, inference does enter into the
comprehension of these sentences. In my judgment, a
connective adverb should be thought of as indicating a
logical connection between the proposition expressed by the
clause (or adjunct/absolute) in which it occurs and some
contextually salient propOSition, which may be expressed by
some other expression in the same sentence, by the preceding
sentence or paragraph, etc., but whose identity must in any
event be inferred. For example, I believe that (56) is
indeterminate with respect to the identity of the
proposition whose adversativeness to the main clause of (56)
is indicated by the adverb nonetheless, and that inferring
that the adjunct of (56) expresses the relevant proposition
is one way (probably the most likely way) of resolving this
indeterminacy. 10
2.3. PRAGMATICALLY ADMISSIBLE VALUES FOR LAND M
In the foregoing discussion, a number of factors have been

identified by which a language user might infer the logical
role of a free adjunct or absolute. For reasons of
exposi tion, these factors have been discussed in five
discrete sections; in reality, however, some of them
overlap, and two or more of them may jointly determine what
sort of inference is drawn in a particular instance. The
implications of the above discussion are nevertheless clear:
in any given case of a sentence that is indeterminate with
respect to the relation holding between its adjunct/absolute
and its main clause, the number of ways of resolving this
indeterminacy will actually be fairly limited.
In Chapters II-V, the semantically indeterminate
aspects of the meaning of a free adjunct/absolute were
represented with the free variables Land M.
(58) Walking home, John saw Mary.
(59) Being a sailor, John smokes a pipe.
(60) Vt[K(L)(~Vt,[M(t,t,) & AT(t" VXs[R(Xs,X i ) &

Ing(~[walk-homel(xs)])])])(~[past(t) &
AT(t, John-sees-Hary')])]
(61) vt[K(L)(Avt ,[M(t,t 1) &

AT(t 1 , Ing(A[sailor'(x o ) ]» ])(A[PRES(t) &
AT(t, John-smokes-a-pipe l )])]
For example, in the intensional logic translations of

sentences (58) and (59), the variables Land M represent two
sorts of indeterminacy:
1) L represents the range of nontemporal relations
which a free adjunct/absolute may bear to its superordinate
clause. Among the possible values for L are a relation of
causation or explanation, a relation of adversativeness, and
a more or less neutral, conjunctive relation.
2) M represents the range of relations which may hold
between the interval to which an adjunct/absolute relates
and the interval to which its superordinate clause relates.
(In a set-level adjunct/absolute, M represents the range of
relations which may hold between some interval at which the
326 CHAPTER VI
proposition expressed by the adjunct/absolute is true and a

member of the (set-level) denotation of the
adjunct/absolute.) Possible values for M are the identity
relation, relations of more or less proximate precedence and
subsequence, and the relation holding between the speaker's
interval and all intervals.
The inferences drawn by language users drastically
limit the number of useful values for Land M in
translations like (60) and (61). Thus, a value for L or M
is pragmatically admissible just when it embodies one of the
possible inferences by which language users might resolve
the indeterminacy which it represents.
An important issue is whether the number of
pragmatically admissible values for Land M is in general
strictly limited, or whether the range of admissible values
for these variables is in principle infinitely large. That
is, one might on the one hand suppose that there is a small
inventory of very basic logical relations relevant for the
interpretation of absolute constructions, and that the
interpretation of any given adjunct or absolute simply
requires that one of these relations be picked out, by
inference; on the other hand, it is conceivable that the
logical relations inferred in the interpretation of absolute
constructions are potentially infinite in diversity. Which
of these two positions is best motivated?
Unfortunately, this is not a question that can be
answered with certainty. Consider, for example, sentence
(62).
(62) Leaving Kansas City, John reached st. Louis at

noon.
Suppose (62) is uttered on two separate occasions, and that

on both occaSions, the adjunct is inferred to bear a
relation of temporal precedence to its superordinate clause.
Is the inferred temporal relation the same on both
occasions? Possibly, but not necessarily. On the one hand,
the simple relation 'ia precedes i m' might be inferred in
both cases (where i a , im are the respective intervals to
which the adjunct and the main clause relate). But it is
conceivable that in a given context, the relation inferred
in the interpretation of (62) could be one which places more
specific restrictions on ia: 'ia precedes im by at least
ten hours', 'ia precedes im and ia is on a Tuesday', etc.
If this is the case, then the relation of temporal
precedence inferred for (62) on a given occasion could in

principle be anyone of an indefinitely large number of such
relations.
Although this possibility cannot be dismissed, it is
equally possible that the range of pragmatically admissible
values inferrable for Land M is in fact finite--that there
is a small stock of highly salient relations which are alone
sufficient for the interpretation of absolute constructions.
This possibility is not at all implausible, for three
reasons:
(i) When people are asked to paraphrase absolute
constructions, the logical relation which they 'fill in' is
nearly always an extremely simple one--indeed, it is nearly
al ways a relation named by an overt connective, e.g.
because, although, when, while, before, after, and, or by
(means of). (Recall in this connection that the traditional
grammarians surveyed in section 2 of Chapter I always
attributed a finite range of possible logical roles to
absolute constructions.) This suggests that there may be a
small number of such relations which people normally rely on
to structure their perceptions and experiences, and that
these alone constitute the pragmatically admissible values
for Land M.
(ii) Just as the relation of temporal precedence in
(62) can be inferred to be anyone of a number of more
specific relations, so too can the temporal relation between
successive sentences in a discourse:
(63) John left Kansas City. He reached st. Louis at

noon.
It is plausible, however, that any specific details inferred
about the relation of temporal succession in (63) do not
have truthconditional status--that they are simply
implicatures. Accordingly, the same could well be true for
whatever minor details are inferred about the temporal
relation in (62); that is, i t could well be that such
details do not figure at all in the value inferred for Mo in
the translation (64) of (62). ~
(64) Vt [ K ( L 6 ) ( Vt 1 [~19 ( t,. t 1) &

A
AT(t 1 , Vx s [R(x s , x4)

1
&
leaving-Kansas-City'(xs»)))(A[at-noon'(t) &
[past(t) & AT(t, John-reaches-St.-Louis')]])]
32!:l CHAPTER VI
(iii) The view that the range of pragmatically

admissible values for Land M is highly circumscribed would
appear to be most compatible with the notion of an
inference, as I have characterized it. Recall that an
inference is here regarded as anything which is not part of
the literal meaning of an expression but which language
users judge to be part of the intended meaning of that
expression. If the 'success' of an inference is viewed as
the extent to which it embodies a speaker's intentions, then
it is clearly more likely that a hearer will draw a
successful inference if s/he has fewer choices to make about
what to infer. And it is clear that not just any relation
can be inferred in the interpretation of an absolute
construction: the relation between the adjunct and the main
clause in (65a) would never be inferred to be the relation
overtly expressed in (65b), no matter what the speaker's
intentions were.
(65) a. Having grown up in Boston, John knows all

about boats.
b. John grew up in Boston, but that's not why he
knows all about boats.
Thus, the interpretation of an absolute construction

requires a cooperati ve effort between speaker and hearer:
in determining the logical role of a free adjunct or
absolute, the hearer must attempt to judge the speaker's
intentions; the speaker, however, must see that the intended
relation is one which can be inferred reasonably easily
under the circumstances at hand. It is quite plausible that
the latter goal is nearly always achieved precisely because
speaker and hearer both tacitly assume that the relation to
be inferred can be drawn from a limited inventory of very
basic relations.
In this section, I have argued that a sentence
containing a free adjunct or absolute may be semantically
indeterminate--that its truthconditions may actually depend
on the logical role inferred for its absolute construction.
In addition, a broad array of factors has been found to form
the basis for inferences regarding the logical role of a
free adjunct or absolute; the 'ontological level' of the
predication expressed by the adjunct/absolute; the relative
duration of the events or states of affairs to which the
adjunct/absolute and its superordinate clause relate; the
order of the adjunct/absolute with respect to its
superordinate clause; language users' knowledge of the
world; and the presence of a connective adverb in either the
adjunct/absolute or its superordinate clause. The
inferences which language users draw on the basis of such
factors greatly constrain the range of pragmatically
admissible values for the variables Land M in the
intensional logic translations induced by the proposed
Montague fragment for absolute constructions. As I have
suggested, it is quite possible that there is in fact a
finite inventory of logical relations constituting the set
of pragmatically admissible values for Land M.
]. On the Possibility of Deriving Absolute Constructions

from Adverbial Subordinate Clauses
One objection that might conceivably be raised regarding the

proposed analysis of absolute constructions is that it
attributes too much importance to inference. One might, in
fact, try to argue that inference plays no significant role
in the interpretation of such constructions--that free
adjuncts and absolutes derive from adverbial subordinate
clauses whose interpretations are not at all indeterminate.
Consider the consequences of this 'transformationalist'
position. If absolute constructions did indeed derive from
adverbial subordinate clauses, then the free adjunct in (66)
and the augmented absolute in (67) would apparently have
multiple underlying sources, as in (68) and (69),
respectively.
(66) John saw Mary.

Walking home, { John often watches for eagles.
John would have seen the new billboard.
(67) { Mary put on her makeup.

With her hair braided, Mary always resembles Jane.
Mary might have won the lookalike
contest.
(68) While he was walking home, John saw Mary.

When he is walking home, John often watches for
eagles.
If he had walked home, John would have seen the
new billboard.
330 CHAPTER VI
(69) Once her hair was braided, Mary put on her makeup.
When her hair is braided, Mary always resembles
Jane.
If her hair had been braided, Mary might have won
the lookalike contest.
That is, the rule converting adverbial clauses into absolute

constructions would be a many-to-one mapping introducing
massive ambiguity. Under such an approach, it would not be
necessary to infer the logical role of an adjunct or
absolute, since this role would always be strictly
determined by its underlying representation. As a matter of
language processing, inferences would, of course, be
necessary for determining which of several possible
underlying sources was the "intended one, but this is true of
any sort of ambigui ty resulting from the application of a
transformation.
Here, I will explain why I think such an approach to
the semantic variability of absolute constructions would be
misguided.
First, it is uneconomical to postulate rules where no
rules are needed; and as I hope to have shown in the
preceding section, it is possible to provide a perfectly
plausible account of the logical roles played by absolute
constructions by assuming that these roles are inferred.
Moreover, the latter assumption is very well motivated,
since the logical role of an absolute construction a is
demonstrably dependent on both the semantic content of a
itself and its linguistic and extralinguistic context. For
example, the free adjunct crossing the street is assigned
three different logical roles in the three contexts in
(70) ;
(70) a. Crossing the street, John was hit by a car.

(most likely interpretation: 'while he
was crossing the street, ••• ')
b. Crossing the street, John entered the bank.
(most likely interpretation: 'after he
crossed the street, ••• ')
c. Crossing the street, John entered a different
county.
(most likely interpretation: 'by
crossing the street, ••• ')
on the other hand, i f the adjunct carrying the box under his
arm is substituted into (70a-c), it is not assigned the same
three logical roles; instead, it is interpreted as
expressing an 'attendant circumstance' in all three cases.
If the subtle dependencies illustrated in these examples
were somehow built into the transformation converting
adverbial clauses into absolute constructions (so that
crossing the street would, for example, derive from a while-
clause in context (70a), from an after-clause in (70b), and
from a by-phrase in context (70c», the resulting rule would
be so highly conditioned by semantic and pragmatic factors
as to be radically unlike any transformation which has been
seriously entertained. But to the extent that the
transformationalist approach failed to account for these
dependencies, it would have failed to provide a credible
account of the semantic variability of absolute
constructions.
Another reason for doubting the feasibility of the
transformational approach to absolute constructions is the
fact that such constructions cannot always be plausibly
paraphrased by adverbial subordinate clauses; as was pointed
out in Chapter I, Kruisinga (1932:275) has observed that
'the use of free adjuncts in English is greatly promoted by
the almost complete absence of conjunctions that can
introduce a subordinate clause expressing attendant
circum stances.' For example, the adjuncts in (71)- C7 4)
cannot be paraphrased without rather elaborate
circumlocution; these adjuncts can't simply be regarded as
'reduced' versions of full adverbial clauses.
(71) Exhausted, John walked home.
(72) Bill drove along the crater's rim, steering

skillfully to avoid falling one way or the other.
(73) Mary awoke, noticing she was late.
(74) Carrying the box under his arm, John entered the
bank.
Indeed, free adjuncts and absolutes can serve a variety of

specialized functions for which no other sort of subordinate
expression is especially well suited:
(i) One special function of free adjuncts and absolutes
is to clarify some aspect of the event described by the
superordinate clause, without referring to any separate
332 CHAPTER VI
event. In each of (75)-(79), for example, the adjunct and
the main clause are felt to relate to one and the same
action.
(75) ••• a man identified as Jimmy Lee Jones attacked

another man who had been sleeping on a rock in the
park, beating him with a wire litter basket.
(OSU Lantern 8/26/80, 8)
(76) Kate takes off a glove, exposing the diamond on

her third finger •••
(NY 8/4/80, 68)
(77) Now Kirk, talking enthusiastically, like a

football coach, tells the press that the Kennedy
forces are picking up a great deal of support on
the vote on the rule.
(NY 9/8/80, 50)
(78) Mora, using a special drafting eraser, rubbed his

friend's name from the documents he left behind,
and substituted his own.
(OSU Lantern 8/26/80, 3)
(79) IvIy finger, going down in the muck just before we

quit for the day, happened to go right through the
center of the circle of a bronze penannular
brooch •••
(eL, 140-143)
In other cases, the adjunct/absolute may be felt to relate

to an action or state of affairs distinct from that to which
the main clause relates, but one which is nevertheless, in
some sense, part of the same event; recall (71)-(74). In
many such cases, a free adjunct or absolute is used to
describe some part of a person's body as that person
participates in the event described by the main clause:
(80) White beard flowing, he drove the final spike •••

(CL, 122-123)
(81) She clapped her hands like a child, her lucid eyes
sparkling.
(OSI, 246)
(82) ••• a female matador lies across the horse's back,
her breasts bared, an espada still clutched in her
lifeless hand.
(NY 6/30/80, 57)
(83) When he was first acknowledged as chief, he stood

on top of a cairn, holding in his hand a white rod
that symbolized his responsibilities, •••
(CL, 151)
(84) Then he leaves, and the Carters and the Mondales

stand there, waving to the crowd.
(NY 9/8/S0, 99)
(Cf. also the frozen absolute expressions hat in hand, head

first, etc.) In all of the examples (75)-(84), the absolute
constructions are quite difficult to paraphrase with
adverbial subordinate clauses, since such clauses are
typically understood to relate to events or states of
affairs distinct from that described by the clause they
modify.
(ii) A related use of absolutes is found in sentences
whose main clause describes an event or state of affairs
involving a set ~ of participants; in such a sentence, an
absolute may characterize some specific member or subset of
~. Examples of this sort are (85)-(89).
(85) ••• the Benbergs stood by, he clasping his hands

and watching her closely, she wiping a plate round
and round with a sodden cloth.
(Visser (1972:1152»
( 86 ) S eve r a 1 aid e s t 0 May 0 r B y r n e - - 0 n e a n

administrative assistant and another her patronage
chief--turn up and stroll around amid the
delegation. (NY 9/8/80, 77)
(87) Donald McNeill has two, one right next to the

other.
(CL, 17)
(SS) Two Fridays ago, about fifty House members, most

of them relati vely junior, gathered to try to
figure out how to extricate the Party from Carter.
(NY 9/8/80, 41-42)
334 CHAPTER VI
(89) The tombstones were spaced out on the floor in

long rows, each stone about six feet long and
covered with carving in relie~
(CL, 70)
Absolutes of this sort may be elliptical (cf. Grady

(1972:6»:
(90) Into the middle of this tide went many of the

original clansmen of Colonsay, some early, some
later on, some after long stays on the mainland,
others more directly from the island, some
settling in the Lowlands, notably in Renfrewshire,
others going to Australia, Canada, or the United
States. [The first four absolutes are
elliptical.]
(CL, 8)
(91) the orchestra played some blues, a gospel

piece, and "I Got Rhythm," sometimes in unison,
sometimes in harmony, and sometimes free.
(NY 7/16/79. 80)
Absolutes of this kind uniformly resist paraphrase by means

of any sort of adverbial clause.
(iii) In the preceding section, it was seen that
absolutes with being are commonly inferred to provide th€
reason or cause for the truth of their superordinate clause.
Absolutes of this sort may additionally be understood to
explain other things that are in some way relevant to the
proposition expressed by the main clause. Consider, for
example, sentences (92)-(94).
(92) To confront another person with one's head

uncovered was a grave insult, the only exception
being Taoist recluses and Buddhist priests.
(PM, 174)
(93) The loss to him was two hundred and fifty pounds,
the equivalent of the savings of four good years--
a good year being one in which the net income of
the croft, including the government subsidies he
gets for his sheep and cattle, is about a hundred
and fifty pounds.
(CL, 33)
(94) Zelinsky once tried to answer the question of
where the South ends by taking a horse-mule census
within the territory in doubt--the theory being
that farmers who used mules to pull their plows
obviously lived in the South.
(NY 6/16/80, 107)
The absolutes in (92)-(94) are not naturally inferred to

provide a reason or cause for the truth of their
superordinate clause; for instance, (92) is not felt to mean
that confronting someone without a hat was an insult because
it was not an insult for Taoist recluses and Buddhist
priests to do so. Instead, the absolutes in (92)-(94) are
understood as explaining some notion ancillary to the
meaning of the main clause. The main clause of (92)
expresses an ancient Chinese rule of social behavior; the
absolute is understood to explain when this rule could be
suspended. Similarly, the main clause of (93) indicates
that a crofter's loss was the equivalent of four good years'
savings; the absolute is understood to explain what a good
year is. Thus, although absolutes like those in (92)-(94)
are inferred to provide an explanation, they are not,
strictly speaking, comparable to explicit because-clauses.
(Absolutes of this sort are usually postposed.)
For these reasons, I conclude that the transformational
approach to the semantic variability of absolute
constructions cannot be maintained: it is uneconomical,
technically infeasible, and is based on the observationally
inaccurate premise that absolute constructions simply serve
as replacements for full adverbial clauses.
4. On the Possibility that the Logical Role

of an Absolute Construction is Always Inferred
In the analysis which I have proposed, the logical relation

between an absolute construction and its superordinate
clause is viewed as being determined by purely semantic
principles in some instances, but by language users'
inferences in others. Specifically, I have distinguished
two fundamental subcases: (i) the logical role of a weak
adjunct or absolute may be partially or wholly determined by
the semantic properties of an operator appearing in the main
clause (in all such cases, the absolute construction is
formally represented as an argument of the relevant
336 CHAPTER VI
operator); but (ii) the logical role of an absolute

construction (weak or strong) serving as a main tense adverb
must be inferred. This account of the semantic variability
of absolute constructions might therefore be characterized
as a 'mixed' analysis, incorporating both semantic and
pragmatic components. One could, however, imagine a more
radically pragmatic approach, according to which the logical
role of an absolute construction is simply always inferred.
Analyses of this sort have been proposed in various
guises. For instance, one recurring proposal is that the
logical relation between an absolute construction and its
superordinate clause is always simply a very neutral, and-
type relation, and that the semantic variability of absolute
constructions is simply the effect of implicatures which may
be present in one context or another. (Cf. Berent
(1975: 10), Quirk et a1. (1972: 759) for informal proposals of
this sort.) A similar account would be one in which
absolute constructions always function as main tense
adverbs; because of the semantic properties of the operator
K(L) employed in the translation of absolute constructions
of category MTA, this would be similar to an and-type
analysis. All such analyses are immediately ruled out,
however, because of examples such as (95) and (96), in which
the truth of the adjunct or absolute is not entailed.
(95) Wearing that new outfit, Bill would fool everyone.

(96) ~ith her hair braided, Jane might resemble Mary.
A more radically 'inferential' analysis of the semantic

variability of absolute constructions would be one in which
the logical relation between a free adjunct or absolute and
its superordinate clause is always left completely open--in
which this relation is in all cases simply representable as
a variable L over relations between propositions. Under
such an approach, L could be inferred to have any sort of
value whatsoever; owing to the absence of the operator K, it
could even take on a value for which 'L(p)(q) ~ ~p' would
not be valid. Thus, even examples like those in (95) and
(96) would seemingly bow to such an analysis.
I have serious reservations about the wisdom of
pursuing a purely inferential analysis of this sort. In
particular, I think such an analysis must be rejected
because its fails to capture an important generalization:
the only cases in which the logical relation between an
absolute construction and its superordinate clause does not
allow the truth of the absolute construction to be entailed
are those in which this entailm,ent is blocked by a binary
operator appearing in the main clause (such as the modals in
(95) and (96»." Under the purely inferential approach,
there is no reason a priori to expect the nature of the
relation between an absolute construction and its
superordinate clause to correlate so strictly with the
content of the latter expression. Under the analysis which
I have proposed, however, the aforementioned generalization
follows as a direct and automatic consequence: the semantic
characteristics of K(L) guarantee that the truth of an
absolute construction will be entailed if this construction
is not an argument of some binary operator appearing in the
main clause.
Even if it is granted that the logical role of an
absolute construction conditioning the interpretation of a
modal is not simply inferred,- one might still attempt to
pursue a more radically inferential account of sentences
such as (97).
(97) Lying on the beach, John sometimes smokes a pipe.
Recall that in the 'mixed' analysis which I have

proposed, the adjunct lying on the beach serves as the first
argument of the adverb sometimes:
(98) Vt[sometimes'(At,v~2[M9(t"t2) &

s s 1
AT(t 2 , Vx [R(x , x4) &
lying-on-the-beach'(xS)])])(At,[t, c t &
[PRESet,) & AT(t" John-smokes-a-pipe')]])]
Note, however, that under the proposed analysis, the adjunct

in (97) can also be analyzed as a main tense adverb outside
the scope of the frequency adverb, as in (99); in fact, this
would be the only possible analysis if the adjunct were
strong.
(99) Vt[K(L6)(~Vt,[M9(t,t,) & AT(t 1, Vxs[R(x s , x~) &

lYing-on-the-beach'(xs)])])(~[sometimes'(I3)
(At 1 [t, =. t & [PRES(t,) &
AT(t" JOhn-smokes-a-pipe')]])])]
338 CHAPTER VI
Here, 13 is a variable over sets of time intervals; in a

particular context, its value is the set of time intervals
restricting the temporal quantification inherent in the
interpretation of sometimes.
One might attempt to argue that (99) is the only sort
of translation that need be assigned to sentence (97), on
the following grounds. If the value of 13 in (99) varies
according to context, it is perfectly possible that in one
context, it could take on the value of (100):
(100 ) At,Vt 2 [M 8 (t"t 2 ) & AT(t 2 , Vxs[R(x s , xt) &

lying-on-the-beach'(x s )])]
Such a value would, however, make (99) equivalent to (98)

under certain choices for L6 and M9 ; thus, one m.i.ght argue
that (99) covers the interpretation represented by (98), and
that the latter sort of translation should therefore be
dispensed with.
In my view, this sort of argument would be rather
dubious. As was pointed out in section 2.3 above, the
values inferred for Land M in the interpretation of an
absolute construction are subject to clear limitations--such
variables cannot just take on any value in any context. The
same, I claim, is true of the variable 13 in (99). In
particular, I claim that (100) wouldn't ordinarily represent
a possible value for 13 in (99). To see why, consider first
sentence (101).
(101) Because he read the commission's report, John

often refused to drink tap water.
A plausible intensional logic translation for (101) would be

something like (102).
(102) because,(AVt[past(t) & AT(t, John-reads-the-

commission's-report,)])(Avt [octen'(I3)( At1 [t, c t
& [past(t,) &
AT(t 1 , John-reCuses-to-drink-tap-vaterl)]])])
Although many plausible values can be imagined for the

variable 13 in (102), it is also easy to think of highly
improbable values. Among the latter is the set denoted by
(103) •
:>..t[past(t) &
AT (t, John-reads-the-commission's-report')]
That is, it is not at all natural to assign (101) the same

interpretation as (104) (in which the boldface when-clause
is understood as restricting the interpretation of often).
( 104) Because he read the commission's report, John

often refused to drink tap water when he read the
commission's report.
Why is the value of (103) so unlikely as a value for 13 in

(102)? The explanation, I believe, is this: the value
inferred for 13 must in general be a contextually salient
set of time intervals; moreover, it must be a set containing
a sufficiently large number of members, owing to the
semantic properties of frequency adverbs. Now, in and of
itself, an utterance of (101) would not heighten the
contextual salience of the entire set denoted by (103); on
the contrary, the truth of the because-clause in (10 1)
depends on the existence of only a single interval at which
John read the commission's report. Thus, the denotation of
(103) would normally be unlikely as a value for 13 in (102).
(It is, of course, possible to contrive special contexts of
utterance for (101) in which the denotation of (103) would
be salient; but in just these contexts, the set denoted by
(103) would in fact constitute a possible value for I,. The
invention of such a context i::; left a::; an exerci::;e for the
reader.)
In view of the foregoing di::;cus::;ion, consider again
sentence (97). Recall that if the free adjunct in this
sentence is analyzed as a main tense adverb, then (97)
translates as in (99). Notice, however, that the
pragmatically admissible values for L6 are always such that
the truth of (99) depends on the existence of only a single
interval at which John is lying on the beach. Thus, if the
free adjunct in (97) is analyzed as a main tense adverb,
(97) does not by itself heighten the contextual salience of
the entire set of intervals denoted by (100); accordingly,
the denotation of (100) is not normally a possible value for
13 in (99). This being the case, the translation in (98)
cannot be viewed as redundant.
I therefore conclude that language users' inferences
are not the only means by which the logical role of an
absolute construction may be determined: the theory which
340 CHAPTER VI
makes the best predictions about the semantic variability of

free adjuncts and absolutes is one in which two subcases are
acknowledged--one in which the logical function of an
absolute construction is partially or totally determined by
the semantics of a binary operator appearing in the main
clause, and one in which it is determined by purely
inferential means.
5. Theoretical Implications
In conclusion, I shall briefly discuss some of the broader

theoretical issues raised by the proposed semantic analysis
of absolute constructions. Most centrally, I shall consider
the implications of this analysis for the strict
compositionality that is one of the defining characteristics
of Montague semantics; this issue was first addressed by
Partee (1983a: 18ff), whose insights have greatly influenced
the following discussion.
Stated in its broadest, most theory-neutral form, the
Compositionality Principle is the principle that 'the
meaning of an expression is a function of the meanings of
its parts and of the way they are syntactically combined'
(Partee (1983a:1». On first consideration, it is not clear
that the proposed analysis of absolute constructions is
compatible with this principle. I have argued that free
adjuncts and absolutes are in many cases semantically
indeterminate--that their truth-conditional meaning may
depend upon the inferential contributions of language users.
If so, then is this a case in which the meaning of an
expression is not merely 'a function of the meanings of its
parts and of the way they are syntactically combined'?
Doesn't the suggested analysis of absolute constructions
force one to view the meaning of an expression as being, at
least potentially, a function of other, additional factors--
specifically, of language users' inferences?
These questions cannot, of course, be satisfactorily
answered without a sharper conception of what a meaning is.
In Montague's theory of reference, the meaning of an
expression is regarded as a function from possible worlds
and contexts of use to denotations of the type appropriate
for that expression. If this conception of meaning is
assumed, then whether or not the proposed analysis of
absolute constructions is compatible with the
Compositionality Principle depends on the importance one
attributes to the context of use as a determinant of the

interpretation assigned to an expression. It is clear that
Montague intended the context of an expression's use to
include a specification of values for free variables
appearing in that expression (cf. Montague (1970b: 228;244»;
this was, for example, the means he used to account for the
values assigned to third person pronouns in nonanaphoric
uses. If it is assumed that the context of use does in fact
specify values for free variables, then the question of
whether the proposed analysis of absolute constructions is
compatible with the Compositionality Principle can be
restated in a slightly different way: When a language user
infers the logical role of an absolute construction
appearing in some sentence, does this inference determine
the meaning of the sentence (i.e. the function from possible
worlds and contexts of use to truth-values) or does it
merely allow the language user to ascertain which assignment
of values to the free variables Land M is intended by the
speaker--as it were, to ascertain which context of use slhe
is 'in'? If language users' inferences do in fact determine
meaning in Montague's narrow sense, then the proposed
analysis for absolute constructions is almost certainly at
odds with the strict composi tionali ty of Montagovian
semantic theory; but if the task of inferring the logical
role of an absolute construction is simply viewed as the
task of figuring out which values for the variables Land M
are specified by the context of use, then the proposed
analysis is fully compatible with the Compositionality
Principle.
A useful perspective on this issue can be gained by
comparing the relational variables Land M to nonanaphoric
pronouns, following Partee (1983a:19ff). In Montague's
theory, the pronoun she in (105) is analyzed as containing
an implicit free variable x; accordingly, the context in
which (105) is used specifies the value for x.
(105) She is smart.
This is a natural way of analyzing the pronoun in (105),

since its referent is clearly dependent upon the context in
which (105) is uttered: she will generally refer to an
indi vidual prominent either in prior discourse or in the
perceivable surroundings. Now, as Partee (1983a:19f) points
out, the values assigned to the variables Land M in the
interpretation of an absolute construction do not seem to be
342 CHAPTER VI
sensitive to context in the same way that she is in (105):
When we hear [(106a)], even in isolation, we will all

probably interpret the propositional relation variable
as because; the only context we need is the sentenoe
itself. (And having just had a because in prior
discourse probably would make it only marginally more
likely that a subsequent sentence with a free adjunct
would be understood to express a because relation.)
(106 ) a. Being a master of disguise, Bill would fool

anyone.
b. Vt[K(L 6 )(A vt1 [M g (t,t 1) & AT(t,. being-a-
master-oC-disguise'( xo »]) (A [PRES( t) &
AT( t, Bill-vould-fool-anyone')])]
That is, the value assigned to L or M in the interpretation

of the absolute construction in (106a) is not chosen because
of its particular salience in prior discourse or in the
perceivable surroundings; it is chosen largely on the basis
of the content of (106a) itself (cf. sections 2.2.1-2,
2.2.5), and perhaps also on the basis of real-world
knowledge (section 2.2~) and ezpectations. Given that this
is so, is it correct to say that the relations inferred for
Land M are specified by the context of use?
Though Land M may initially appear to be quite
different from nonanaphoric pronouns, Partee points out that
they are in fact not so different after all--that there are
cases in which the referent of a nonanaphoric pronoun can
only be ascertained by inference from the meaning of the
sentence in which it appears. Citing (107)-(110) as
examples, she observes (pp.20f) that under the right
circumstances, an appropriate referent could be inferred for
the pronoun she in each of these sentences even if this
referent were neither perceivable nor salient in preceding
discourse; indeed, such could even be the case if (107)-
(110) were used discourse-initially.
(107) Did she bet her report card today?

(108) She still won't give me a raise.
(1 Og) Did she call and say when she's arriving?
(110) Oh, I see she brought my banjo back.
The limits of the background knowledge shared by speaker and

hearer guarantee that the range of potential referents for a
nonanaphoric pronoun will be fini te in any gi ven context;
thus, assuming that 'shared knowledg& is part of what makes
up a context of use, the range of potential values for the
pronouns in (107)-(110) can be regarded as being specified
by the context of use. Accordingly, in order to determine
the intended ref~rent for the pronoun in sentences such as
(107)-(110), one need only choose the most likely referent
from a finite number of contextually specified
possibilities. As Partee notes (p.21), the choice of a
referent for the pronoun in sentences such as (107)-(110) is
no different from the inference of a value for L or M in the
interpretation of an absolute construction, provided that
the range of possible (i.e. pragmatically admissible) values
for the latter variables is finite. Although I believe
there is room for doubt on this matter, it is certainly
conceivable that language users share a limited inventory of
basic logical relations from which the values of Land Mare
always drawn (see section 2.3); if this is in fact the case,
then this inventory can be viewed as part of the contextual
specification of shared knowledge, so that the logical role
inferred for an absolute construction can simply be regarded
as the role chosen as most likely among a limited number of
contextually specified possibilities.
According to this reasoning, absolute constructions are
no less compatible with the Compositionality Principle than
nonanaphoric pronouns. This conclusion should not, however,
be taken to imply that free adjuncts and absolutes are
unremarkable constructions after all. Quite the opposite is
true. Absolute constructions are distinguished from most
other widely studied constructions in that they normally
require language users to infer the intended value for a
free variable. That is, whereas a third person pronoun can
function (a) as a bound variable; (b) as a free variable
referring to someone/something salient in either preceding
discourse or the perceivable surroundings; or (c) as a free
variable whose referent must be inferred on the basis of the
content of the sentence in which it appears (and on the
basis of shared knowledge), the relational variables Land M
always seem to function in manner (c). Unlike the
interpretation of a personal pronoun, the logical role of an
absolute construction is routinely underdetermined by its
semantics.
This, I think, is why absolute constructions are of
such great theoretical interest; nevertheless, they do not
344 CHAPTER VI
seem to be unique in this regard. As Partee (p.c.) has

pointed out, there may be a variety of constructions whose
semantics routinely under determines their interpretation in
a particular context--that is, which routinely require
language users to infer some relation or other; see, for
example, Partee's (1983a:28ff; 1983b:5-7) discussion of
genitives and prepositional phrases and Clark's (1978:313ff)
discussion of shorthand expressions (mentioned above in
section 2.1). Similar analyses might likewise be proposed
for compounds (cf. in particular Dowty's (1979:314ff)
discussion), for topicalized constructions which don't bind
anything in the main clause (Partee (p.c.», and for
individual lexical items such as the verb have; further
research will no doubt turn up other such cases as well.
Thus, although semantic indeterminacy of the sort
exemplified by absolute constructions has heretofore
received little attention in the formal semantic literature,
it is quite possible that it will have to be confronted as a
very widespread phenomenon in human language.
In this chapter, the inferences whereby language users

resol ve the inherent semantic indeterminacy of sentences
with free adjuncts and absolutes have been examined; a
variety of factors have been found to contribute to such
inferences. Most importantly, it has been seen that, just
as there are definite semantic determinants of the logical
role which an adjunct or absolute intuitively plays (as was
demonstrated in Chapters II-V), there are quite obvious
pragmatic determinants as well. This fact appears to
validate the thesis defended here--that the intuitive
logical role of a free adjunct or absolute is a product of
both the semantics of English and the inferences of language
users.
Footnotes
1. Inferences of this sort are central to the intentional

theory of comprehension developed by Clark (1978).
Formal linguists have from time to time argued that the
semantics of English must include a deductive logic of some
sort (cf. e.g. Lakoff (1971), Nash-i/ebber & Sag (1979»; as
evidence, various phenomena have been cited which
purportedly reflect the interaction of a grammatical rule
with a deductive inference. Here, I won't be concerned with
deductive inferences.
2. I do not mean to suggest that the relative 'permanence'

of a property is always a reliable criterion for determining
whether it is a property of stages or of indi viduals. It
isn't: be a baby is individual-level, while alive is stage-
level, even though the interval at which someone is a baby
is normally shorter than that throughout which s/he is
alive; and be a student is individual-level, even though it
is possible to be a student at each of several discontinuous
intervals, while one is normally alive only throughout a
single, continuous interval. See Carlson (1977: 122£).
3. Note, however, that for some speakers, adjuncts and

absolutes with being can, in special circumstances, be
inferred to bear an adversative relation, as in (i) (see
section 2.2.5 below for discussion of examples like this).
(i) Being a gardener, he nevertheless knows nothing

about greenhouses.
And, as was pointed out in Chapter II, be-phrases consisting

of be4 plus a passive participial phrase are in fact stage-
level; thus, in each of (ii)-(iv), the relation between the
adjunct and the main clause can easily be inferred to be
temporal or circumstantial rather than causal.
(ii) Being carried through the streets, crowded with

early marketgoers, he wondered how his colleague
had managed to keep the fact that the dancer was
dead from his host of servants.
(PM, 110)
( iii) Gerald Rafshoon, being interviewed on television,

says that for Jimmy Carter "that's a badge of
courage, to be unpopular~
(NY 9/8/80, 90)
(iv) At this moment she's languishing on a couch in my

First Lady's room, being pampered with cold towels
and what have you!
(PM, 102)
4. I should note that there is additional cross-linguistic

variation in the status of being-type adjuncts. Although
346 CHAPTER VI
such adjuncts are strong in English (subject to the

exceptions noted in footnote 3) and in French (cf. section 1
of Chapter V), Feryal Yava§ informs me that in Turkish,
olarak-adjuncts appear to be weak:
(i) Doktor olarak, ben boyle sey yap-ma-m.

doctor being I this thing do-neg-1sg
If I were a doctor, I wouldn't do such a thing.
Olarak 'being' is in fact often translated as 'as'--quite

appropriately, given that as-adjuncts are weak in English;
cf. section 3.4 of Chapter II.
5. It is a truism of historical linguistics that the

pragmatic properties of a linguistic expression can, over a
certain period of time, be reinterpreted as centrally
semantic. For example, the word cloud derives from an Old
English word meaning 'hill', whose recurrent nonli teral (or
metaphorical) use in describing 'hills' of water vapor in
the sky ultimately altered the word's literal meaning;
similarly, indirect speech acts routinely become
conventionalized, so that Would you please pass the salt? is
no more a request for confirmation than Pass the salt.
please. For this reason, it is not surprising that free
adjuncts and absolutes sometimes develop into explicitly
prepositional or adverbial subordinate constructions: in
such cases, a pragmatic property (a recurrently inferred
value for L or M) is reinterpreted as part of the semantics
of the construction; this reinterpretation normally centers
on the head of the predicate of the adjunct or absolute,
e f fee t i vel yeo n v e r tin g i t i n t 0 apr e p o·s i t ion 0 r
subordinating conjunction.
Consider, for example, the case of during. In Middle
English, the present participle of the verb dure 'to last'
(from ~Uddle French durer) was regularly used in an absolute
construction (mirroring the use of durant in French):
(i) I have the power, duryng al my lyf,

Vpon his propre body.
(Chaucer, Canterbury Tales, D158)
An accurate translation of the absolute in (i) might be 'all

my life lasting' or 'as my entire life lasts'. Such
absolutes were useful for situating an event or state of
affairs within a particular interval; thus, in Chaucer's
time, (i) was likely interpreted as in (ii),
(ii) Vt[K(L 6 )(A vt1 [M 9 (t,t 1 ) &

AT(t 1, duryng-al-my-lyf') ])("[PRES(t) &
AT(t, I-have-the-pover-vpon-his-propre-body')])]
where the intended value of Mg was a relation of temporal

containment, e.g.
(The inferred value of L6 , on the other hand, would seem to

have been extremely weak, merely a relation of conjunction:
The regularity with which such a value could be inferred for

M in the interpr~tation of a during-absolute, coupled with
the gradual obsolescence of dure as a finite verb in
English, eventually caused during to take on the value of a
preposition of temporal containment, translatable as in
(iii); its status in modern English is unquestionablY that
of a full-fledged preposition. (Compare French durant,
German v3hrend.)
( iii)
Analogous developments have led to the reanalysis of

absolute constructions with predicates headed by pending,
notWithstanding, and except (cf. French pendant, nonobstant,
except~). other sorts of adjuncts ahd absolutes seem at
present to be in the course of reanalysis as overtly
subordinate adverbial clauses, though the verbs heading
their predicates are still recognizably verbs: seeing (as
(how», assuming that, granted that, given that, provided
that.
For further discussion, see Onions (1971;78n,
Jespersen (1940:51f;409ff), and Visser (1972:1217ff;1292ff).
6. Actually, this isn't quite correct. In sentences like

I'll leave when I said I would, in which the 'extraction
site' is embedded within the adverbial clause, the tense of
the adverbial clause doesn't have to agree with that of the
main clause, These can be safely ignored here, because free
348 CHAPTER VI
adjuncts and absolutes are never interpreted analogously to

such adverbial clauses.
7. For other ways in which word order tends to influence

language users' inferences, see Jespersen (1940:60ff); cf.
the discussion in Chapter I, section 2.
8. Oddly, it still seems quite possible to infer that the

shouting preceded the counting in (46).
9. Properly speaking, knowledge of the world is ultimately

responsible for the sorts of inferences examined in
subsections 2.2.1 and 2.2.2 above; but the sorts of
knowledge involved in such inferences are so specialized as
to deserve separate consideration.
10. In this connection, note that free adjuncts may

themselves have relative frequency adverbs:
with his heavy sword he has performed

decapitations, sometimes retaining as trophies the
heads he has severed.
(CL, 153)
( 11) The little people frequently used to dance by the

Temple of the Glen in the night, always singing a
monotonous song of theirs •••
(CL, 138-139)
(iii) Families sat on hillsides, often in snow or rain,

and watched their homes burn out.
(eL, 7)
In none of (i)-(iii) is the frequency adverb explicitly

restricted by a set-level time adverb. Thus, the adverbs
sometimes, always, and often in these sentences may be
regarded as having a free variable over sets of time
intervals as their first argument (cf. Chapter III, section
3.1): this variable is evaluated by inference. As it
happens, the main clause in each of (i)-(11i) provides the
necessary information for inferring the relevant set of time
intervals.
11. Here I am referring to the absolute construction as a

producti ve structure of English syntax. There are idioms
having the form of free adjuncts or absolutes whose meaning
is inherently conditional: all things considered, present
company excepted, assuming that X, provided that X, etc.
Such frozen expressions lack the semantic variability
typical of the productive construction, and are therefore
excluded from consideration here.
A P PEN D I X
A FORMAL FRAGMENT FOR FREE ADJUNCTS

AND ABSOLUTES
In this appendix is presented a Montague fragment for the

syntax and interpretation of free adjuncts and absolutes in
English; the rules developed in Chapters II-V are given here
in their final form. The fragment is presented in two
parts: rules of syntax and model-theoretic interpretation
for an intensional logic are gi ven in section 1, and rules
of syntax and intensional logic translation for a fragment
of English are given in section 2.
1. Intensional Logic
The version of intensional 10e,ic assumed for the fragment

presented here is defined syntactically and interpreted in
much the same way as in Montague (1973). Notable departures
are: 1) the inclusion of the following among the letters of
intensional logic: past, pres, fut, c, <, AT, PRES, FUT,
NON PAST (Chapter III, section 1.1), xn and perf (Chapter IV,
section 1.1); 2) the inclusion of the followinf:, expressions
among the constants of intensional logic: cb, C, D (Chapter
II, section 2.1), now, moment, NOW (Chapter III, section
1.1), and Ing (Chapter IV, section 3.1); 3) the sorting of
entities into kinds, objects, and stages (Chapter II,
section 3.1): 4) ttle addition of a new primitive type i--the
type of time intervals--and of all complex types that can be
derived by the usual recursive means usinl:, i (Chapter III,
section 1.1); and 5) the inclusion of the following in the
structure of a model: a conversational background
coordinate (Chapter II, section 2.1), a speaker's interval
coordinate (Chapter III, section 1.1), and a function
specifying the set of inertia worlds corresponding to a
given index (Chapter IV, section 3.1). Certain of these
departures recei ve comment as they arise in the following
statement of the structure and interpretation of intensional
logic.
The set of types, Type, may be defined as follows. Let
350
A FRAGMENT FOR FREE ADJUNCTS AND ABSOLUTES 351
s, e, i, and t be distinct objects (none should be <:In

ordered pair): the set of types is then the smallest set Y
containing e, i, and t: <a,b>, whenever a,b £ Y: and <s,a>,
whenever a £ Y.
The set of sorts to be employed is {k, 0, st} (the set
containing the sort of kinds, of objects, and of stages,
respectively), where k, 0, and st are distinct from one
another and from s, e, i, and t (and aren't ordered pairs).
This forms the basis for the set of sorted types, SType,
which is defined as follows (the account of sorting employed
here is modeled on those of Cooper (1975) and Carlson
(1977)): where i' and t' are distinct from each other and
from each of s, e, i, t, k, 0, and st (and aren't ordered
pairs), the set of sorted types is the smallest set Y
containing k, 0, st, it, and t': <a,b>, whenever a,b £ Y:
<s,a·>, whenever a £ Y; and [a,b], whenever a,b £ Y and for
some c £ Type, a,b 0 c. 0 is a relation holding
between certain members of SType and Type (deriving
ul timately from a sim ilar relation defined by Cooper
(1975:92)); i t can be recursively characterized as follows:
(i) t'
0 t (read: t' is included in t)
(ii ) i'
0 i
( iii) k, 0, st
CD e
(iv) If a,b Type, a' ® a, and b' ® b, then

®
£
<a', b'> <a, b>
(v) If a £ Type and a' CD a, then <s,a'> ® <s,a>
(vi) If a E: Type and a' ,b' ® a, then [a' ,b'] ® a
(vii) For all a E: Type, there is no at such that at '0 a

except as provided for in (i)-(vi). ~
[So, for example, «s,<st,t'»,<[o,k],t'»

«s,<e,t»,<e,t».]
0
Where E, W, I a re any sets (to be taken as the set of
possible entities, the set of possible worlds, and the set
of intervals of time, respectively), the set Da,E,W,I of
possible denotations of type a relative to E, W, I is
352 APPENDIX
characterized by recursive definition:
De,E,W,I =E
Di,E,W,I = I
D
t,E,H,I = {A, {A}} (where the empty set is identified
with falsehood, the unit set with
truth)
D<a,b>,E,W,I = {x: x a total or partial function from

Da,E,H,I into Db,E,W,I}
D<s,a>,E,\I,I = {x: x a total or partial function from

vi X I into Da,E,W,I}
= Sa E W I' the set of senses of type a
, " relative to E, W, 1.
The set Da',E,N,I of possible denotations of sorted

type a' relative to E, W, I is similarly characterized:
Dk, E, H', I = k'(E)
Do,E,W,I = o'(E)
Dst,E,I-J,I = st'CE)
(k', 0', and st' are functions connected with the respective
sorts; k'(E), o'(E), and st'CE) are all subsets of E, and
are to be taken as the set of kinds, of objects, and of
stages, respectively. k', 0', and st' must be so structured
that k'(E) n o'(E) = k'(E) fl st'(E) = o'(E) n st'(E) = A,
and that k'(E) U o'(E) U st'(E) = E.)
Di',E,W,I = I
Dt',E,W,I = {A, {A}}
D<a',b'>,E,'\J,I = {x: x E D<a,b>,E,ii,I (where a,b

E Type, a' ®
a, and b ' b) ®
and x has a total subfunction from
Da',E,W,I into Db',E,W,I}
(The total subfunction x' of x is simply that total function

having as domain the set A of all and only arguments for
which x has a defined value, such that for all yEA, x(y) =
x'(y).)
D(s,a'>,E,W,I = {x: x E D(s , a> " E W, I (where a E Type

and a' ® a) and x has a total
subfunction from W X I into Da',E,W,I}
= Sa' ,E,W,!
D[a',b'],E,W,I = {x: x E Da',E,W,I or x E Db',E,W,I}
[It will have been noted that kinds, objects, and
stages are here treated as sorts of entities, while
intervals are treated as things of a distinct type from
e n tit i e s • I n Do wt y' s (1 9 7 9 : 326) w0 r d s , ' the p rim a r y
moti vat ion for sorting is to allow certain variables
and constants to range over the whole domain of
entities, as well as allowing other variables to range
over only a part of it.' Since it is important that
there be variables ranging over kipds, objects, and
stage:;;--for exall)ple, the variable Xl over individuals
in AX1VXs[R(xs,x 1 ) & asleep'(x s )] must be of the sorted
type [o,k]--these must be treated as different sorts of
entities; but no expression of the fragment requires
that there be variables ranging over intervals and
entities, so that the type of intervals can be treated
as distinct from that of kinds, objects, and stages.
Some denotations of type (a,b> and (s,a> (a,b £
Type) are allowed to be partial functions so that the
set of possible denotations of sorted types (a' ,b'> and
(s,a'> (a',b' E SType, a' ®
a, b' ®
b) can
simply be characterized as a certain subset of the set
of possible denotations of types (a,b> and (s,a>; as
will be seen below, this does entail some complications
in the definitions of the denotations of certain
expressions.]
The language La of intensional logic is assumed to have
denumerably many variables and an infinite number of
constants of each sorted type. Given any nonnegative
integer n and any a' E SType, vn a' is the nth variable of
sorted type a'; Con a , is the set 6f constants, Var a , the set
354 APPENDIX
of variables, of sorted type a l •

The set MEL of meaningful expressions of LO is
recursively charac~erized. Whenever a,b E Type,
(1) Every variable and constant of sorted type a ' is

in the set MEa of meaningful expressions of type
a, where a' ® a;
(2) now, NOW E HE i ; moment E HE<i,t>;
Ing E HE«s,t>,t>;
cb E ME<s,«s,t>,t»;
C E ME «s,«s,t>,t»,<s,«<s,t>,t>,t»>'•
D E ME«s,«s,t>,t»,«s,t>,<s,«<s,t>,t>,t»»;
(3) If a E MEa and u is a variable of type b, then Aua
E NE<b , a>;
(4) If a E M~<a,b> and S E MEa' then a(S) E HE b ;
(5) If a,S E MEa' then [a = S] E MEt;
(6) If ¢,1j; E HE t , then -.¢, D¢, [¢ & 1j;], [¢ v 1j;],
[cj>+1J!],[cj>-1j;] EME t ;
(7) If ¢ E MEt and u is a variable, then VU¢, flucj>
E:HE t ;
(8) If ¢ E MEt and u is a variable of type e, then 1u¢
E HEe;
(9) If u E MEa' then Aa E ME<s , a>;
(10) If u E ME<s,a>' then -u E MEa;
(11) If U E MEi and ¢ E MEt, then AT(u,¢) E MEt;
(12) If a E NE i , then past(a), pres(u), fut(a),
perfCa), xn(a), PRES(a), FUT(a), NONPAST(a) E MEt;
(13) I f a,S E: ~1Ei' then [a ~ S], [u < S] E: MEt:
(14) Nothing is in any set NEa except as required by
(1)-(13).
The following two abbreviatory conventions are used:

y{cx} is [-Y](cx).
An interpretation of LO is an octuple
<E, W, I,~, F, if, K, Inr>,
where E, W, I are as above; < is a linear ordering of I; F

is a function wi th the set of all constants as its domain,
such that for any sorted type a' and any B € Cona" F(B)
E D<s a'> E W I; i* is a member of I;
, ",
[i* is to be thought of as the speaker's interval
relative to which nonshifting temporal expressions are
interpreted.]
K is a function with W X I as its domain and the set of sets
of subsets of W X I as its range;
[K is to be thought of as the conversational background
relative to which modals are interpreted.]
Inr is a function with H X I as its domain and the set of
subsets of W as its range, such that for any index <w,i>,
Inr«w,i» is a subset of the set of worlds identical to w
up to interval i.
[For any index <w,i>, Inr«w,i» is the set of 'inertia
worlds' corresponding to <w,i>; 'these are to be
thought of as worlds which are exactly like the given
world up to the time in question and in which the
future course of events after this time develops in
ways most compatible with the past course of events'
(Dowty (1979: 148)).]
Assume an interpretation So? of the form
<E, W, I,~, F, if, K, Inr>
and an So?-assignment g (a function with the set of variables

as its domain such that for any variable u of sorted type
a', g(u) E Da , ,E, W,I). For any meaningful expression a. of
Lo' [a] .awig is the denotation of a with respect to the
interpretation Q, the index <w,i> (s W X 1), and the Q-
assignment g. In particular,
(1) If a is a constant, then [a] Dwig is F(a)«w,D)j

i f OL is a variable, then [OL] .a wig is g(a);
(2) [now] .awig is i; [NOW] .awig is if;
[moment] D wig is that subset S of I such that for
356 APPENDIX
all i' E: I, i" E: S, i' is not a proper

subinterval of i";
[Ing] !} wig is that function h such that for any
subset p of W X I, h(p) = {A} iff for some
interval i' such that i c i' and i is
not a final subinterval for i', and for all
w' E InrC<w,i», pC<w',i'» = {rd, and
otherwise A;
[cb] !} wig is K;
[c] !} wig is that function h such that for any
function f from W X I to sets of subsets of
W X I and any (w',i'> E W X I, hCf)C(w',i'»
is the set of all consistent subsets of
f«w' ,i'»;
[D] Dwig is that function h such that for any
function f from W X I to sets of subsets of
vJ X I, any subset p of W X I, and any (w',i'>
E.: W X I, hCf)Cp)C<w',i'» is the set of all
consistent subsets of fC<w',i'» U {p} which
contain p.
U) If a E: HEa and u is a variable of sorted type b',
then [A uaJ !} wig is that function h with domain
Db,E,W,I Cwhere b' b) whose total
subfunction has domain Db',E,W,I and is such that
®
. [ J!} wig'
for a 11 x E.: Db',E,H,l' h ( x ) 1S a , where g ,
is the .a-assignment like g except for the
possible difference that g'Cu) is x;
(4) If a ,
E.: fJlE<a b> and S E.: ~1Ea' then [aCS)] .Qwig is
[a] !} wigc [s] .Q wig);
(5) I f a,S E: MEa' then [[0: = S]] Dwig is
{A} iff [ex] .Qwig and [8] .Qwig are defined and
identical;
A iff [ex] .Qwig and [8] .Qwig are defined and
nonidentical;
and otherwise undefined:
(6) I f q"lj! E: MEV then [-'q,] .Qwig is
{A} iff [¢].Q wig is A;
A iff [¢].Qwig is {A} ;
and otherwise undefined;

[O¢].Qwig is
{A} iff for every <w',i'> E: W X I, [¢] .Qw'i'g
is {A}:
, .,
A iff for some <w',i'> E: W X I, [<PJ Dw 1. g is
A;
[[q, & lj!]] .Qwig is
{A} iff both [<P] .Qwig and [lj!] .Qwig are {A}:
A iff [¢] .Qwig or [w] .Qwig is A and both are
defined;
similarly for v, +, -;
(7) If cP E: MEt and u is a variable of sorted type a',
then [Vu<P] .Qwig is
{A} iff [¢] .Q wig' is defined for all g' and
there exists x E: Da',E,W,I such that
[CP] .awig' is {A}, where g' is the
.Q-assignment like g except for the
possible difference that g'(u) is x;
A iff [CP] .Qwig' is defined for all g' and for
all x E: Da' , E , W, I' [CP] .awig' is fl., where
358 APPENDIX
g' is as above;
similarly for Au~;
(8) If ~ E MEt and u is a variable of sorted type a'

(where a' ® e), then [1U~] Qwig is t~a,t unique
member x of Da',E,W,I such that [¢] QWlg is {A},
where g' is the Q-assignment like g except for
the possible difference that g'(u) is X; and
otherwise undefined;
(9) If a E MEa' then [ ACt] Q wig is that function h with
domain W X I such that whenever <w' ,i'> E: VI X I,
Q ,.,
h«w',i'» is [a] w 1 g;
(10) I f a E: ME<s,a>' then [va] Qwig is [a] Qwig«w,i»;
(11) I f ~ HE t and ex ~1Ei' then [AT(ex, ~)] .Q wig is
,
E: E
.
{A} iff [~] .oWl g is {A}, where i' is
[cd!} wib;
A iff [¢] !}wi'g is A, where i' is as above;

(12) I f a E t-1E i , then [past(a)] !}wig is
{A} iff there is some nonempty i' E: I such
that [a] !}wig < i' < i;
A iff [a] !} wig is defined but there is no
such nonempty i I E: I;
[pres(a)] !}wig is
{A} iff [a] .owig ~ i;
/I. iff [aJ !}wig is defined but isn't a
subinterval of i;
[fut(a)] !}wig is
A FRAGHENT FOR FREE ADJUNCTS AND ABSOLUTES 359
{A} iff i < [Ct.] .QWig;
A iff [Ct.] .Qwig is defined but i t is not the

case that i < [ex].Q wig;
[per:f( a) ] .Q wig is
{A} iff (i) there is some nonempty i' =-

[ex].Qwig such that i' <i and (ii) there
is no i' =- [ex] .Qwig such that i < i';
A iff [ex] .Q wig is defined but conditions (i)
and (ii) aren't both met;
[xn(a)] .Qwig is
{A} iff i is a final subinterval for
[ex] .Q wig;
A iff [ex] .Q wig is defined but i is not a

final subinterval for [Ct.].Q wig;
[PRES(a)] .Qwig is
{A} iff [ex].Q wig -c i*',
A iff [cx] D wig i s d e fin e d but i s not a
subinterval of i*j
[FUT(ex)] .Q wig is
{A} iff i* < [ex] D wig;
A iff [Ct.] D wig is defined but it is not the
case that i* < [a].Q wig;
and otherwise undefined:
[NONPAST(Ct.)] .Qwig is
{A} iff there is no i' c [ex].Q wig such that
i' < i*:

360 APPENDIX
A iff [cx] .Qwig is defined but there is some

i' =- [cd .Qwig such that i' < i*;
(13) I f a,B e: ME i , then [[a=- B]].Qwi g is
{A} iff [a] .Qwig is a subinterval of
[B] .Qwig;
II iff [a] .Q wig and [B].Q wig are defined but

the former isn't a subinterval of the
latter;
[[et < BJ] .Qwig is
{A} iff [a] .Qwig < [B] .Qwig;
A iff [a] .Qwig and [B] .Qwig are defined but

the former doesn't precede the latter;
and otherwise undefined.
If ¢ £ MEt, then ¢ is true with respect to !1 and <w,i>

iff [¢J .Qwig is {A} for every !1-assignment g. If <w,D £
W X I, then <w,D is an index of possible utterance iff i c
"* •
1
2. Syntax and Translation Rules

for a Fragment of English
In this section, a fragment of English is formally defined

and translated into the intensional logic presented in
section 1.
2.1. SYNTAX
The set of categories of English is the smallest set Cat

such that (1) t, t ' , PRED s , INf s , PRPL s , PSPL s , ADJ s , pps,
IV s , TV s , PRED i , INF i , PRPL i , PSPL i , ADJ i , ppi, NOM, lVi,
TV i , CN, ~10D, TA, RFA, TAB, Meas, TSC ~ Cat, and (2) if A,B
A FRAGMENT FOR FREE ADJUNCTS AND ABSOLUTES 36'
€ Cat, then AlB, AIIB, AIIIB € Cat. The categories employed
in the fragment are:
t the basic category {( it) [rains]}

of declarative
sentences
t' the basic category l\

of conditional
clauses
the basic super- (B1NFS U BpRPLs U BpSPLs

category of stage-
level predicative U BADJs U EppS)
phrases
the basic category A

of stage-level
infinitive phrases
PRPL s the basic category A

of stage-level
present participial
phrases
PSPL s the basic category

of stage-level past
participial phrases
the basic category {asleep, ready, awake,

of stage-level happy, available, drunk,
adjective phrases unavailable, a little
cold, unaware of her,}
pps the basic category {in first gear}
of stage-level
prepositional
phrases
the basic category {[laugh], [dance],

of stage-level [walk] , [ drive] , [lie] ,
intransitive verb [run] for pre3ident,
phrases [bark], [go] dancing,
362 APPENDIX
[continued] [go] skydiving, [go] for
a walk, [go] back to
school, [stand], [shake],
[snarl], [work], [watch]
TV, [walk] home, [sing],
[leave], [arrive],
[leak], [stroll] by,
[coast] gently downhill,
[make] a lot of progress,
[have] nightmares about
the war, [live] in it 2 ,
[stop] by it2-' [stay] on
it 2 , [thinkJ of her,.
[run] into her l' [get]
down from it2' [walk]
into it 2 }
the basic category {[chase], [hit], [kiss],

of stage-level [unravel], [see],
transitive verb [arrest], [mow], [trim],
phrases [figure], [use], [touch],
[reach], [wear], [cross],
[hear], [ smoke], [ carry] ,
[braid], [make],
[answer], [finish],
[leave], [knock] over,
[sing], [lose], [fool],
[lead], [set] up, [eat]}
the basic super- (B1NFi U BpRPLi U BpSPLi

category of
individual-level U BADJi U Bppi U BNOM )
predicative phrases
the basic category

of individual-level
infinitive phrases
PRPL i the basic category A

of individual-level
present participial
phrases
PSPL i the basic category II

of individual-level
past participial
phrases
the basic category {fat, intelligent,

of individual-level American, obnoxious,
adjective phrases smart, widespread, rare,
thirty-three years old,
thirty-five years old}
the basic category II

of individual-level
prepositional
phrases
NOM the basic category II

of individual-level
predicative noun
phrases
the basic category A

of individual-level
intransitive verb
phrases
the basic category {[love], [know],

of individual-level [admire], [resemble],
transitive verb [fear], [have], [remind]
phrases of her,}
eN the basic category {post office, room,

of common-noun beach, drunk,
phrases businessman, license,
hat, master of disguise,
street, tax, calculator,
new outfit, chair,
barrel, sailor, hill,
truck, highchair, living
room, ceiling, doctor,
neighbor, unusually long
arm, roof, lawn, cat,
dog, car, pipe, cigar,
load of over 1500 Ibs.,
over 1500 Ibs., hedge,
364 APPENDIX
CN [continued] question, train, hair,
mailman, song, bridge,
funny noise, horse,
.other, child, water,
wife, father, wheat
Carmer, family, project,
way to it2 , top of it2 }
MOD the basic category {[must], [can], [would],

of modal verbs [might]}
TA the basic category {yesterday, today,

of set-level time tomorrow, in the morning,
adverbs in 1988, in 1940, during
the past su.mer, the day
before, at noon}
RFA the basic category {always, often,

of relative fre- frequently, sometimes,
quency adverbs occaSionally, never}
TAB the basic category A

of temporal
abstracts
Meas the basic category {almost a half mile, only

of measure phrases a few tons, four tons}
TSC the basic category {before, after, when,

of temporal while}
subordinating
conjunctions
tit the category of {necessarily,

sentence nevertheless, actually,
adverbs obviously}
(t/t)/t the category of {because, although}

tit-forming subordi-
nating conjunctions
1Q the category tilt {exactly why it2

of indirect derailed}
questions
Pps/NOM the category of as {as}
Pps/T the category of {on, in, at, with}

stage-level
prepositions
IVs/PRED s the category of {[be]}

be4 (see Chapter
II, section 3.2.4)
IVs/PRED i the category of active {[be]}

be3 (see Chapter II,
section 3.2.3)
IVs/INF i the category of. {[agree], [decide]}

stage-level INF1-
taking verbs
the category of {[claim], [discover],

stage-level sen- [say], [think],
tence-complement [announce], [promise]}
verbs
IVs/IVs the category of {very dangerously}

stage-level verb
phrase adverbs
IVs/ADJ i the category of. {[look] }

stage-level ADJ 1-
taking verbs
Tys/IV i the category of {[make] }

stage-level Iyl_
taking transitive
verbs
ADJi/INF i the category of {eligible}

individual-level
INFi-taking
adjectives
ppi/T the category of {from}

individual-level
prepositions
366 APPENDIX
Iyi/PREDs the category of {[ be]}
be, (see Chapter
II, section 3.2.1)
Iyi/PREDi the category of {[be] }
be2 (see Chapter
II, section 3.2.2)
Iyi/IQ the category of {[know] }
individual-level
IQ-taking verbs
Iyi/Heas the category of {[measure], [weigh]}

individual-level
measure verbs
the category of {[prefer] }

individual-level
INFi-taking verbs
MTA the category {at-that-time n , since

TAB/TAB of main noon}
tense adverbs
TABIITAB the category of {[have] }

perfect auxiliaries
T the category t/IV i {John, Mary, Bil.l., Jane,

of terms Fido, we, he4n' she4n+"
it~n+2' theY4n+3' that,
thl.s, enough money,
everyone, the Med Center,
the Board of Trade,
Wales, Kansas City, St.
Louis, Paris, dinner,
supper}
T/CN the category of {that, this, a, the,

determiners hisl:/ n ' her4n+l' its 4n + Z'
thel.r4 n+ 3 ,our J
[A number of the basic expressions of the fragment are
intuitively complex; since their internal syntax is
irrelevant to the analysis of free adjuncts and absolutes, I
shall leave such expressions unanalyzed in the interests of
keeping the fragment as simple as possible. This practice

is, of course, not unknown in Montague's own work--cf.
especially Montague (1970a).]
For any category A e; Cat, BA is the set of basic
expressions of A; a basic expression of the present fragment
is a member of UA e; Cat BA•
The set PA of phrases of category A (for A e; Cat) is
defined recursively by the rules of syntax. Employed in
these rules are the following morphological functions (see
section 4 of Chapter I):
Where a is of the form [B],
PRPL(a) is the pr€sent participle of a;
PSPL(a) is the past participle of ex;
PRES1(cx) is the third singular present form of cx;
PRES2(cx) is the third plural present form of ex;
PAST1(ex) is the third singular preterit form of cx;
PAST2(a) is the third plural preterit form of cx;
BASE(a) is the base form of cx (if cx e; BMOD '
BASE(cx) is undefined);
where ex is a common noun,
PLUR(cx) is the plural form of a.
In addition, the following syntactic subroutines are
employed: for any expressions a,B,Y,
EN(ex) is the result of replacing every expression of the
form [oJ] in cx with PSPL([oJ]) (and is undefined if some
bracketed expression in ex lacks a past participial
form) ;
ING(ex) is the result of replacing every expression of the
form [oJ] in cx with PRPL([oJ]) (and is undefined if some
bracketed expression in ex lacks a present participial
form) ;
NONF(ex) is the result of replacing every expression of the
form [oJ] in a wi th BASE( [cJ]) (and is undefined i f some
bracketed expression in ex lacks a base form);
DB(a) is the result of deleting all brackets from cx;
DP(ex) is the result of deleting all brackets and parentheses
from a;
SSDELn(cx) is the result of deleting the subscript n from
every occurrence of hen' him n , his n , she n , hern' itn'
its n • they n • them n • their n in CX;
PRET(ex) is the result of replacing each expression of the
form [cJ] in cx with PAST1([oJ]) i f [oJ] is PRES1([cJ]) and
with PAST2([oJ]) i f [cJ] is PRES2([']) (and is undefined
if some expression of the form [oJ] in ex is such that
368 APPENDIX
PRES1([6]), PRES2([eI]), PAST1([eI]), or PAST2([eI]) is

undefined) ;
SUBJ(~) is the result of replacing each expression of the
form [6] in ~ with PAST2([6]) (and is undefined if some
expression of the form [6] in a is such that PAST2([6])
is undefined);
OBJ(a) is hia n , her n , them n , or us if a is hen' she n , they n ,
or we, respectively; is a if a is any other member of
Pr; and is otherwise undefined;
DEL(~,B) is the result of deleting every instance of ~ in B;
ROS(a,B) is the result of placing a immediately after the
first parenthesized expression in B (and is undefined
if B has no parenthesized expressions);
RB(~,B) is ROS(a, DB(NONF(B»);
PRB(a,B) is ROS(V(a), DB(EN(B»), where V is the
morphological function such that the first expression
of the form [eI] in B is V([eI]) (and is undefined if B
has no bracketed expressions);
RWRAP(a,B) is the result of placing B after the first word
of a;
AGR(a,B) is the result of replacing each expression of the
form [eI] in B with PRES1([eI]) if ~ £ [-Plural] and with
PRES2([6]) if a € [+Plural] (AGR is undefined if ~ ¢
[-Plural] U [+Plural] or if there is some expression
[6] in B such that PRES1([eI]) and PRES2([6]) are
undefined);
SUB(a,B,y) is the result of replacing the first occurrence
of B in y with a (and is undefined if y has no
occurrences of B);
CON(a,B) is the result of concatenating a and B;
CONWIB(a,B) is the result of concatenating a and B with an
intervening intonation break.
The set PA of phrases of category A (for A € Cat) is
defined as follows.
so. (i) BA PA for every category A:

~
PINFs U PpRPL s U PpSPL s U PADJs U Ppps =

PpRED s ;
P1NFi U PPRPL i U PpSPL i U PADJi U Pppi U PNOM
= PPREDi.
( ii) P TAB U PTA = [-Adjunct] U [+Adjunct];
[-Adjunct] fl [+Adjunct] = A;
BTA =- [-Adjunct].
( iii) PTAB n [+Adjunct] = [-Weak] U [+Weak] ;
[-Weak] n [+Weak] = A.
(iv) Pt U PTAB = [-Tense] U [+Tense];
[-Tense] [l [+Tense] = A·,
Bt =- [-Tense].
(v) PTAB = [-Perfect] U [+Perfect];
[-Perfect] n [+Perfect] = A.
(vi) PT = [-Plural] U [+Plural);
[-Plural] n [+Plural] = A;
theY4n+3' we s [+Plural];
all members of BT other than theY4n+3' we,
are members of [-Plural].
Rules creating A-abstracts
S1. If a € PpREDs, then F 1,n(a)

PTAB , [-Tense, +Adjunct,
€
+Weak, -Perfect], where F 1 ,n(a) is a. [p.163)
S2. If a PpRED i , d PADJi, then F 2 ,n(a) € PTAB , [-Tense,

€
+Adjunct, -Weak, -Perfect], where F 2 ,n(a) is a.

[P.164]
S3. If a € PTABIITAB and S E PIVs, then F3 ,n(a,S) E PTAB '

[-Tense, +Adjunct, +Weak, +Perfect), where F3 ,n(a,S) is
DB(ING(CON(a, DB(EN(S»»). [p.242]
S4. If a € PTAB1ITAB and S € PIVi, then F 4 ,n(a,S) E PTAB ,

[-Tense, +Adjunct, -Weak, +Perfect], where F4,n(a,s) is
DB(ING(CON(a, DB(EN(S»»). [p.243]
370 APPENDIX
S5. If a e: P r and a e: PPREDs, then F 5a (a,a), F 5b (a,a)

e: P rAB , [-Tense, +Adjunct, -Perfect], F 5a (a,a) e:
[-Weak], and F 5b (a,B) e: [+Weak], where F 5a (a,a) is
CON(a,B) and F 5b (a,a) is CON(with, CON(OBJ(a),B».
[p.278]
S6. If a e: PT and B e: PpRED i , i PADJi, then F 6a (a,B),
F 6b (a,B) E PTAB , [-Tense, +Adjunct, -Weak, -Perfect],

where F 6a (a,a) is CON(a,a) and F 6b (a,B) is CON(with,
CON(OBJ(a),a». [p.278]
S7. If a e: PI' B e: PTABIITAB' and y e: PIVs, then

F 7a (a, B,Y), F 7b (a, B, y) e: PTAB , [-Tense, +Adjunct,
+Perfect], F 7a (a,a,y) e: [-Weak], and F 7b (a, a, y)
e: [+Weak], where F7a (a, a, y) is
CONCa, DB(ING(CON(B, DB(EN(y»»»
and F 7b (a,B, y) is
CON(wlth, CON(OBJ(a), DB(ING(CON(B, DB(EN(y»»»).
[p.293]
S8. If a E PT ' B e: PTA B I I TAB' and y e: P I Vi, the n

F 8a (a, B, y), FSb(a, B, y) E PTAB , [-Tense, +Adjunct,
-Weak, +Perfect], where F8a (a, a, y) is
CONCa, DB(ING(CON(a, DB(EN(y»»» and F 8b (a,a,y) is
CON(with, CON(OBJ(a), DB(ING(CON(B, DB(EN(y»»»).
[p.294]
Rule creating adjuncts and absolutes of category TA
S9. If a e: PTAB , [+Adjunct, +Weak, -Perfect], then F 9,n(a)

e: PTA' [+Adjunct], where F 9 ,n(a) is a. [p.185]
Rule creating adjuncts and absolutes of category ~
S10. If a E PTAB , [+Adjunct, +Weak, -Perfect], then F 10 (a)

e: P t " where F1Q(a) is a. [p.165]
Rule creating adjuncts and absolutes of category MTA
311. If a E PTAB , [+Adjunct, -Perfect], then F 11 ,m,n(a)

E P MTA , where F 11 ,m,n(a) is a. [p.167]
Rules creating augmented adjuncts
S12. If a is when or while and S E PpREDs, then F 12(a, S)

e: PTA' [-Adjunct], where F 12 (a.,S) is CON(a.,S).
S13. If a is when or while and S e: PpREDi, then F 13 (a.,S)

e: PTA' [-Adjunct], where F 13(a, B) is CON (a, B).
Rules creating conditional clauses
S14. I f $ e: P t ' [+Tense], then F 14 ($) E Pt" where F 14 ($) is

CON(iC, DP($».
S15. I f $E P t , [-Tense], then F 1S ($) e: Pt" where F 1S ($) is

CON(iC, DP(SUBJ(</l»).
Rule creating sentence adverbs
S16. I f a e: P(t/t)/t and </l e: Pt ' [+Tense], then F 16 (a,$)

e: Pt/ t ' where F 16 (a,$) is CON(a, DP($».
Rules creating sentences
S17. I f a E P MOD and $ Pt , [-Tense], then F 17 (a,$)

E E Pt ,
[ - Ten s e ], wher e F 17 ( a, $) is RB( a,$ ) • [ p. 51 ]
S18. If a e: P MOD ' $ E Pt " and 1/J E Pt ' [-Tense], then

F 1 8 ( a, </l, 1/J) E Pt ' [ - Ten s e ], w her e F 1 8 ( a, </l, 1/J) i s
CONWIB(</l, RB(a,1/J». [p.S1]
S19. If a E PT and B E PIVi, then F 19(a,8) E P t , [-Tense],

where F 1g (a,S) is CON«a), AGR(a,S». [p.73]
372 APPENDIX
S20. If a E Ptlt and 1> E Pt ' [+TenseJ, then F 20a Ca,1»,

F 20b (a,1» E Pt ' [+TenseJ, where F 20a (a,1» is
CONWIB(a,1» and F 20b (a,1» is CON(<j>,n:).
Rules creating intransitive verb phrases
S21. If a E PIVs, then F 21 Ca) E PIVi, where F 21 (a) is a.

[p.72J
S22. I f a E PIVs, then F 22(a) E PIVi, where F 22Co:) is a.

[p.195]
S23. If a E PIVs/PREDs and B E PpSPLs, then F 23 (a,B) E PIVs,

where F 23 (a,S) is CONCa,S). [P.S3J
S24. If a E PIVS IPREDi and B E PNON or PADJi, then F24 (a, B)
E PIVs, where F 24 (cx,B) is CON(a,B). [p.79]
S25. If a E PIVi/PREDs and B E PpRPLs, PADJs, or Ppps, then

F 25 (a,B) E PIVi, where F 25 (a,B) is CONCa,B). [p.75]
S26. If a E PIvi/PREDi and B E PADJi, PNOtvl' Pppi, or PpSPLi,

then F 26 (a,B) E PIvi, where F 26 (a,B) is CON(a,B).
[p.76]
S27. I f (X E: PIVs It and 1> E Pt , [+Tense], then F 27(0i,¢ )

E Prvs, where F27Cad) is CONC(X,CONCthat,DPC1»».
[p.31]
S28. If a E PTVs and B E Pro then F 23 (adl) E PIVs, where

F 2S (a,B) is RWRAP(a, OBJCB». [p.SO]
S29. If a E PTVi and B E PTf then F 2g Ca,B) E PIVi, where

F 2g Ca,B) is RWRAPCa, OBJCB». [p.SO]
330. If a E PIVi/Meas and B E PMeas' then F 30 Ca,B) E PIvi,

where F 30 Ca,B) is CONCa, B).
S31. I f a e: PIVslINFi and 8 e: PINFi, then F31 (a,8) e: PIVs,

where F 31 (a,8) is CONCa,8).
S32. I f a e: PIVi IINFi and 8 e: PINFi, then F 32 Ca, 8) e: PIVi,

where F 32 (a,8) is CON(a,8).
S33. If a e PIVs/ADJi and 8 e PADJi, then F 33 (a,8) e PIVs,

where F 33 (a,8) is CON(a,8).
S34. I f a e PIVi lIQ and 8 e: PIQ , then F 34 (a,8) e PIVi, where

F 34 (a,S) is CONCa,S).
S35. If a e PIVslIVs and S e PIVs, then F 3S (a,8) e PIVs,

where F 35 (a,8) is CON(8,a). [p.31]
S36. If a e PIVs and 8 e Ppps, then F 36 (a,S) € PIVs, where

F 36 (a,8) is CONCa,S).
S37. If a,S e PIVs, then F 37 (a,S) e PIVs, where F 37 (a,S) is

CONCa, CON(and,S».
Rule creating transitive verb phrases
S38. If a e PTVslIVi and 8 e PIVi, then F 38 (a,8) e: PTVS'

where F 38 (a,8) is CONCa, DB(8».
Rules creating infinitive phrases
S39. I f a e PIVs, then F 39 (a) e PINFs, where F 39 (a) is

CON(to, DB(a».
S40. I f a e PIVi, then F 40 Ca) € PINFi, where F 40 (a) is

CON(to, DB(a».
Rules creating past participial phrases
341. If a e PTVs, then F 41 (a) e: PpSPLs, where F 41 (a) is

DB(EN(a». [p.81]
374 APPENDIX
342. If a £ PTVs and is £ PI' then F 42 (a,iS) £ PpSPLs, where

F 42 (a,iS) is CON(DB(EN(a», CON(by, OBJ(iS»). [p.81]
S43. I f a £ PTvi, then F 43 (a) £ PpSPLi, where F 43 (a) is

DB(EN(cx». [p.82]
S44. I f a £ PTVi and 8

PI' then F 44 (a,B) £ PpSPLi, where
£
F 44 (cx,g) is CON(DB(EN(a», CON(by, OBJ(8»). [p.82]
Rules creating present participial phrases
S45. If a £ PIVs, then F45(cd £ PpRPLs, where F 45 (a) is

DB(ING(a». [p.261J
S46. I f a PIVi, then F 46 (a)

£ £ PpRPLi, where F 46 (a) is
DB(INGCa». [p.261]
Rule creating adjective phrases
S47. I f a £ PADJi/INfi and B

PINFi, then F 47 (a,g)
£ E: PADJi,
where F 47 (a.,g) is CONCa,S).
Rules creating nominal phrases
S48. If a £ PCN' then F 4S (cx) £ P NOM ' where F'4S(a) is

COtHa, a).
S49. If cx £ PT/CN and 8 E: PCN' then F 49 (a,8) E: PT'

[-Plural], where F 49 (cx,B) is CONCa,g).
S50. If cx £ PCN' then F 5o (cx) e: PI' [+Plural], where F'50(c;.)

is PLURCcx).
Rule creating possessive determiners
S51. If cx e: P T and is John, Mary, Bill, Jane, or Fido, then

f S1 (u) E: PT/CN' where F S1 (a) is CONCa, ·s).
Rules creating prepositional phrases
S52. I f a E PPps/NOH and S E P NOM ' then F 52 (a,S) E Ppps,

where F 52 «(X,S) is CONCa,S).
353. If a E ppps/ T and B

P T , then F 53 «(X,B) E E Ppps, where
F 53 (a,B) is CONCa, OBJ(B».
354. If (X E Pppi/T and B

P T , then F 54 (a,B) E E Pppi, where
F 54 (a,S) is CONCa, OBJ(B».
Quantification rule
355. If cx E P T and <P E Pt , [STense], then F 55 ,n(cx,<P) E Pt ,

[BTense], where:
(i) if <P is of the form '~ 1/ ~', where 1/ is
(hen)' (she n ), (it n ), or (they n ), then
F 55 ,n(a,<P) is SSDELn(SUB«a), 1/, <P»;
(ii) i f <P is not as in (U, then F 55 ,n(a,<P) is

S3DEL n (SUB(cx, 1/,<P », where 1/ is hen' him n ,
she n , hern' it n , they n , or them n ;
(iii) there is some nonnegative integer m such that
n = 4m+3 iff a E [+Plural];
(iv) 0: = OBJ(a).
Rules of tense
S56. I f ¢ E Pt' [-Tense], then F56(¢) E PTAB' [-Tense,

-Adjunct, -Perfect], where F 56 (<P) is <P. [p.l05]
357. If a E P TAB , [-Tense, -Adjunct, -Perfect],

then F 57(a) PTAS ' [+Tense, -Adjunct, -Perfect], where
E
F 57 (a) is PRET(a). [p.l05]

376 APPENDIX
S58. If u PTAB ' [-Tense, -Adjunct, -Perfect], then F 5aCa)

E:
E: P TAB , [+Tense, -Adjunct, -Perfect], where F 58 Ca) is

a. [p.106]
S59. If a PTAB ' [-Tense, -Adjunct, -Perfect], then F59 (a)

E:
E: P TAB , [+Tense, -Adjunct, -perfect], where F 59 (a) is

RB([will], a). [p.106]
S60. If a £ PTAB , [+Tense, -Adjunct, -Perfect], then F60 Ca)

E: Pv [+TenseJ, where F 60 Ca) is a. [p.107J
Main tense adverb rule
S61. If a e: P MTA and S E: PTAB , [yTense, 6Adjunct, ~ Weak,

7lPerfectJ, then F 61a Ca,S), F 61b Ca,S) e: P TAB , [yTense,
6Adjunct, ~Weak, 7lPerfect], where F 61a Ca,S) is
CONWIBCa,8) and f 61b Ca,8) is CONC6,a). [p.119]
Rule creating perfect temporal abstracts
S62. If Ct E: PTABIITAB and B E: P TAB ' [-Tense, -Adjunct,

-PerfectJ, then F62 Cu,8) E: P TAB , [-Tense, -Adjunct,
+PerfectJ, where F 62 Ca,S) is PRBCa,S). [p.236J
Rule creating main tense adverbs
S63. If U E: PTA' [-Adjunct], then F 63 Cu) E P MTA , where

f 63 (Ct) is Ct. [p.118]
frequency adverb rules
S64. If U E: P RFA , SEPTA' and y E: PTAB , [Hense, -Adjunct,

~Perfect], then F 64a Ca, 8,Y), F 64b Cu, 8,Y) E P TAi3 ,
[tlTense, -Adjunct, ~Perfect], where F 64a CCt, S,y) is
CON"lIIB(B, y'), F 64b Cu,B,y) is CONCy', (3), and y' is

ROS(a,y). [p.178]
S65. If a E P RFA and a E P TAB , [yTense, -Adjunct,

6Perfect], then F 65 ,n(a,a) E PrAB , [YTense, -Adjunct,
6Perfect], where F 65 ,n(a,a) is ROS(a,a). [p.179]
Rule introducing the generalization operator Gil
S66. If a E PTA and a

PTAB , [yTense, -Adjunct, 6Perfect],
E
then F 66a (a,a), F 66b (a,B) E PTAB , [yTense, -Adjunct,

6Perfect], where F 66a (a,B) is CONWIB(a,B) and F 66b (a,B)
is CON(a,a). [p.200]
Rules creating set-level ~ adverbs
S67. If a E PTSC ' a f while, and cp e: Pt , [+Tense], then

F 67 ,n(a,<I» E PTA' [-Adjunct], where F67,n(a, $) is
CON(a, DEL(at-that-tiae n , DP(<P»). [p.131J
368. I f a E P TSC and B E P TAB , [-Tense, -Adj unct,

-Perfect], then F 68 (a,B) E PTA' [-Adjunct], where
F 6 8(a,B) is CON(a, DP(PRET(B»). [p.127J
369. If a E PTSC and B E P TAB , [-Tense, -Adjunct,

-perfect], then F 69(01,8) E PTA' [-Adjunct], where
F 69 (a,8) is CON(a, DP(S». [p.128]
Rule removing abstracts from the subclass [+Perfect]
S70. If a E PTAB , [-Tense, SAdjunct, yWeak, +Perfect], then

F70 (a) E PTAB , [-Tense, aAdjunct, yWeak, -Perfect],
where F70 (a) is a. [p.237]
2.2. TRANSLATION
The type assignment for the categories of English is a

function f with Cat as its domain and Type as its range,
such that:
(1) f(t) = f(t') = t;

378 APPENDIX
(2) f(PRED s ) = f(INF s ) = f(PRPL s ) = f(PSPL s ) = f(ADJ s )

= f(Pp s ) = f(IV s ) = f(PRED i ) = f(INF i ) = f(PRPL i )
= f(PSPL i ) = f(ADJ i ) = f(ppi) = f(NOl-I) = f(IV i ) =
f(eN) = f(Meas) = (e,t);
(3) f(TV s ) = r(TV i ) = <e,<e,t»:
(4) f(MOD) = «s,«<s,t>,t>,t»,«s,t>,t»;
(5) f(TA) = f(TAB) = <i,t>;
(6) f(RFA) = «i,t>,«i,t>,t»;
(7) f(TSC) = «s,<i,t»,<i,t»:
(8) f(AlB) = f(AIIB) = f(AIIIB) = «s,f(B»,f(A»
whenever A,B £ Cat.
For any A £ Cat, if at £ PA' then at translates as an

intensional logic expression of type f(A).
The sorted type assignment for basic expressions of
English is a function c with UA £ Cat BA as its domain and
SType as its range, such that the following hold. (Here and
throughout, in' abbreviates the sorted type
[[<s,<o,t'»,<s,<k,t'»J,<s,<[o,kJ,t'»J; 0' abbreviates the
sorted type [<s,<o,t'»,<s,<[o,kJ ,t'»J; k' abbreviates the
sorted type [<s,<k,t'»,<s,<[o,k],t'»J; and in" abbreviates
the sorted type [[<s,<o',t'»,<s,<k',t'»J,<s,<in',t'»J.)
(1) c«it) [rains]) = t':

(2) for all at £ BADJs U BppS U BIVS' c(a) = <st,t'>:
(3) fpr all at £ BTVs, c(at) = <st,<st,t'»:
(4) c(widespread) = c(rare) = <k,t'>: for all other a
£ BADJ i U BCN U BMeas ' C(OI,) = <o,t'>:
(5) for all a £ BTVi, C(a) = <[o,kJ,<o,t'»:
(6) BMOD ' c(a) =
for all at £
«s,«<s,t'>,t'>,t'»,«s,t'>,t'»;
(7) for all a £ BTA , c(a) = <i' ,t'>;
(8) for all at £ BRFA , c(at) = «i' ,t'>,«i' ,t'>,t'»:
(9) for all a € BrQ U Bt/t , C(a) = «s,t'>,t'>:
(10) for all a 8 B(t/t)/t' C(a) = «s,t'>,«s,t'>,t'»:

(11) for all a 8 BppS/NOM U BIVS/INFi U BIVS/ADJi U
BrVS/PREDi, C(a) = <in',<st,t'»:
(12) for all a E Bpps/T' C(a) = <in",<st,t'»:
(13) for all a E BIVs/PREDs U BIVs/IVs, c(a) =
«s,<st,t'»,<st,t'»:
(14) for all a E BIVS/ ' c(a) = «s,t'>,<st,t'»j
t
(15 ) for all a 8 BTVS !IV i , C(a) = <in',<st,<st,t'>>>:
( 1 6 ) for all a E BADJilINFi U BrvilINFi, C(a) =
<in' ,<o,t' »;
(17) for all a E Bppi/T' C(a) = <in",<o,t'»;
(18) for all a E Brvi/PREDs, C(a) =
«s,<st,t'»,<[o,kJ,t'»:
(19) for all a E Brvi/PREDi, c(cx) =
[[«s,<[o,kJ,t'»,<[o,kJ,t'»,
«s,<o,t'»,<o,t'»J,«s,<k,t'»,<k,t'»J;
(20) for all a 8 BrVi/rQ' C(a) =
«s,«s,t'>,t'»,<o,t'»:
(21) for all a E BrVi/Meas' c(a) = «s,<o,t'»,<o,t'»j
(22) for all a 8 BTSC U B~1TA U BTAB/ /TAB' c(a) =
«s,<i' ,t'»,<i' ,t'»:
(23) c(theY4n+3) = <in',t'>; for all other a 8 BT' C(a)
= <0' ,t'>;
(24) for all a 8 BTlcN , c(a) = <0' ,<o',t'».
For any A E Cat, if a E BA, then a translates as an
intensional logic expression of sorted type e(a).
The following variable conventions will be followed in
the translation rules:
xS, yS, ZS are of sorted type st
XO, yO, o
xk , yk, k
380 APPENDIX
xi, yi, zi [o,k]

pS , QS <s,<st,t'»
pO, QO 0'
Pk , Qk k'
pi, Qi in'
pi in"
p, q, r, s <s,t'>
t, tl' t 2 , i'
I, I1 ' 1 2 , <i' , t' >
pt , Qt <s,<i',t'»
pt <s,«s,<i' ,t'»,t'»
M, Ml' M2 , <i',<i',t'»
L, L1 ' L2 , «s,t'>,«s,t'>,t'»
In addition, the following constants of intensional logic

are assumed: unaware-oC', live-in', stop-by', stay-on',
think-oC', run-into', get-down-from', walk-into' are
constants of sorted type <[o,k],<st,t'»; remind-of' is a
constant of sorted type < [o,k] ,< [o,k] ,<o,t'»>; way-to',
top-of' are constants of sorted type <[o,k],<o,t'»; Poss is
a constant of sorted type <[o,k],<[o,k],t'»: morning',
summer', noon', day' are constants of sorted type <i',t'>;
John', Mary', Bill', Jane', Fido', we', this', that' are
constants of sorted type 0; K is a constant of sorted type
«<s,t'>,«s,t'>,t'»,«s,t'>,«s,t'>,t'»>; H is a constant
of sorted type <[o,k] ,<st,t'»; H' is a constant of sorted
type <k,<o,t'»; G is a constant of sorted type
«s,<st,t'»,<[o,k] ,t'»; and G" is a constant of sorted
type «i',t'>,«i',t'>,t'».
Translations for the fragment of English are produced
by the following rules:
TO. (i) unaware of her, translates as

unaware-of' (xf);
(ii ) [live] in it2 translates as live-in' (x~);
[stop] by it2 translates as stop-by' ( x~) ;
[stay] on it2 translates as stay-on' (x~);
A FRAGMENT FOR FREE ADJUNCTS AND ABSOLUTES 38'
[get] down from it2 translates as

get-down-from' (x~);
[walk] into it2 translates as walk-intol(x~);
[think] of her, translates as think-ofl(xf)j
[run] into her, translates as run-into'(xf);
(iii) [remind] of her, translates as
remind-of' ( x f) ;
(iv) way to it2 translates as way-to'(x~)j
top of it2 translates as top-of'(X~);
(v) today (e BTA ) translates as
Atvt,[day'(t 1 ) & [NOW~ t1 & t ~ t,]]
(= today');
yesterday (€ BrA) translates as

Atvt, [[day'(t,) & At 2 [today'(t 2 ) -+ [t, < t2
& At 3 [[t, < t3 & t3 < t 2 ] -+ todaY'(t 3 )]]]] &
t~t,]j
tomorrow BTA ) translates as
(€
AtVt 1 [[daY'(t 1) & At 2 [todaY'(t 2 ) ~ [t 2 ( t,
& At 3 [[t 2 ( t3 & t3 < t,] -+ todaY'Ct 3)]]]] &
t ~ t,];
during the paat aummer BTA ) translates as
(~
Atvt,[[[summer'(t 1) & tl < now] &

At 2 [[summer'(t 2 ) & t2 ( now] ~ [t 2 ( t, v
t2 = t,]]] & t ~ t,];
at noon (~ BTA ) translates as At[noon'(t)]j
the day before (g BTA ) translates as
At 1vt 2 [[day'(t 2 ) & Vt 3 [[daY'(t 3) &
nov~ t 3 ] & At 5 [t 5 ~ t3 -+ [t 2 < t5 &
At 6 [[t 2 < t6 & t6 < t 5 ] -+ t6 ~ t 3 ]]]]] &
t 1 =- t 2 ];
382 APPENDIX
in the morning (e BTA ) translates as

AtVt,[morning'(t,) & t ~ t,];
(vi) as translates as
).piAxsVxi[R(xs,xi) & pi{xi}];
with translates as A piAXSVxo[R(xs,xO) &
pi{AAyi[have'(xO,yi)]}];
(vii) [be] (e PIVs/PREDs) translates as APS[VPS];
[be] (e PIVs/PREDi) translates as
APiAXSVxk[ Axo 0 [pi{ xc} .. R' (x o ,xk» &
R(xs,x k )];
[be] (e PIVi/PREDs) translates as
APSAxiVxS[R(xS,xi) & pS{x s }];
[be] (e PIVi/PREDi) translates as Api[vpi];
(viii) beCore translates as
AptAtVt,[[t < t, & M(t,t,)] & pt{t,}];
after translates as
AptAtVt,[[t, < t & M(t,t,)] & pt{t,}];
when translates as AptAtVt,[t = t, & pt{t,}];
while translates as
AptAtVt,[[t ~ t, & ~ment(t,)] & pt{t,}];
(ix) at-that-timen translates as
AptAt[t = tn & pt{t}];
since noon translates as
AptAt([xn(t) & Vt, [[noon'(t,) & Vt 2 [day'(t 2 )
& [t, ~ t2 & nov~ t 2 ]]] & t, < t]] & pt{t}];
(x) [have] (c BTAB//TAB) translates as
AptAt[perC(t) & vt,[t, ~ t & pt{t,}]];
(xi) John, Mary, Bill, Jane, and Fido translate as
APopo{John'}, Apopo{Mary'}, APopo{Bill'l,
Apopo{Jane'}, and APopO{Fido'}, respectively;
we translates as Apopo{we'};
hen' she n , itn translate as APOpO{X~};
that and this translate as APopo{that'} and
APopo{this'}, respectively;
theYn translates as Apipi{x~};
(xii) his n , hern' its n translate as
APOAQO[Qo{1Xo[POSS(x~,xo) & pO{xo}]}]:
theirn translates as
APOAQO[Qo{ 1Xo[Poss(x~,xo) & pO{xo}]}]:
our translates as
APOAQO[Qo{ 1Xo[Poss(we',x o ) & pO{x o }]}]:
(xiii) where g is a biunique function whose domain
is the set S of members of UA e: Cat BA not
mentioned in (i)-(xi) above, g(a) e: Conc(a)
whenever a e: S.
T1. If a e: PPREDs and a translates as

. a', then F 1 , n (a)
translates as At[AT(t, Vxs[R(x , ~) & a'(x s )]) 1
s
T2. If a e: PPREDi and a translates as a', then F 2,n (a)

translates as H[AT(t, a'(x~»], At[AT(t, a'(x~»], or
At[AT(t, a'(x~»], according as a' is of sorted type
<o,t'>, <k,t'>, or <[o,k] ,t'>, respectively.
T3. If a e: PTABIITAB' B e: PIVs, and a,B translate as a',B',
,
then F 3 n (a, B) translates as
.
a,(AAt[AT(t, Vxs[R(x s , x~) & B'(x s )])]).
T4. If a e: PTABI/TAB' S e: PIVi, and a,B translate as a',B',

then F 4 ,n(a,B) translates as a,(AAt[AT(t, B'(x~»]),
a,(AAt[ATCt, B'(X~»]), or a'(~At[AT(t, s'(x~»]),
according as S' is of sorted type <o,t'>, <k,t'>, or
<[o,k] ,t'>, respectively.
384 APPENDIX
T5. If a £ PT' B £ PPREDs, and a,B translate as a',B', then

F 5a (a,B), F 5b (a,B) translate as
At [ AT ( t, a' CAAx i Vx s [R Cx s , xi) & B' Cx s) ] ) ) ] •
T6. If a £ PI' B £ PPREDi, and a,B translate as a',B', then

F 6a (a,S), F 6b Ca,B) translate as At[AT(t, a,(A B,»].
17. PT' B e: PTABI !TAB' Y £ PIVs, and a, B, Y

I f a e:
translate as a' , B' , y' , then F 7a (a, B , y) ,
F 7b Ca, S, y) translate as
B,(Aa[ATCt, a'CAAxiVXS[R(X s , xi) & y'(x s )]»]).
T8. If a £ PI' S £ PTABIITAB' Y £ PIvi, and a, B, y

translate as a', B', y', then F 8a Ca, S, y),
F 8b Ca, B, y) translate as B,(AAt[AT(t, a,(A y'»]).
T9. If a PTAB and a translates as a' ,

£ then F 9 ,n(a)
translates as AtVt 1 [M n (t,t 1 ) & a'(t 1 »)·
T10. I f a e: P TAB and a translates as a', then F,O(a)

translates as vt[ a'(t)].
T11. If a £ PTAB and a translates as a', then F",m,n(a)

translates as
AptAt[KCLm)CAvt1[MnCt,t1) & a'(t 1 »))C A[p t {t}])].
T12. If a e: PTSC' B e: PpREDs, and a, S translate as a', S',

then F12 ,n(a,B) translates
. as
a,(AAt[ATCt, Vxs[RCx s , x~) & B'Cx s )])]).
T13. If a £ PTSC ' S e: PpREDi, and a,S translate as a',8',

then F13,nCa,S) translates as a,(AAt[AT(t, B'(x~»]),
a ' CAAt [ AT Ct , 8' Cx~» ) ), 0 r a' CAAt [ AT Ct , B' Cx ~ ) ) ] ) ,
according as 8' is of sorted type <o,t'>, <k,t'>, or
<[o,k] ,t'>, respectively.
T14. If cp E: Pt and cp translates as cp', then F 14 (CP)

translates as CP'.
T15. If cP E: Pt and cP translates as CP', then F 15 (CP)

translates as CP'.
T16. If a E: P(t/t)/t' cP E: P t , and cx,cP translate as cx',cp',
then F 16 (a,CP) translates as a'(~CP').
T 17. I f a e: PM 0 D' cP E: P t' and a, cP t ran s 1 ate a sa' ,CP " the n
F 17 (a,CP) translates as a'(C(cb»)(~cp').
T18. If a E: P MOD ' cP E: Pt " 1jJ e: Pt' and a, cP, 1jJ translate as
a', cp', 1jJ', the n F 18 ( a, cP ,1jJ) t ran s 1 ate s as
a'(D(cb) (~cp'»)(~1jJ').
T19. If a £ PI' S £ PIVi, and a,S translate as a',S', then

F 19(a,S) translates as a,(A S').
T20. If a E: Pt / t , cP E: P t , and a,cp translate as a',cp', then

F 20a (a,CP) , F 20b (a,cp) translate as a'("CP').
T21. If a E: PIVs and a translates as a', then F 21 (a)

translates as AxiVxs[R(xS,x i ) & a'(x s )].
T22. If a 10: Prvs and a translates as a', then F 22 (a)

translates as G("a').
T23. If a PIVs/PREDs, B E: PpREDs, and (l,S translate as

E
a',S', then F 23 (a,S) translates as a'("S').
T24. If a e: Prvs IPRED i , S E: PPREDi, and a,S translate as

a',S', then F 24 (a,S) translates as a'("S').
T25. If a Prvi/PREDs, S 10: PpREDs, and a,S translate as

10:
a', S', then F 25 (a,S) translates as a,(A S').

386 APPENDIX
T26. If a E PIVi/PREDi, 13 E PpREDi, and a,B translate as

a',B', then F26 (a,B) translates as a,(A B,).
T27. If a E PIVs/ t , <p E Ptf and a,<p translate as a',q,', then
F27 (a,<jl) translates as a,(A<jl').
T28. If a E PTVs, 13 E PI' and a,B translate as a',B', then

F28 (a,B) translates as \pi\x s pi{A\yiVys[R(yS,yi) &
a'(ys)(xs)]}(A B,).
T29. If a E PTVi, 13 E PI' and a,S translate as a',S', then

F 29 (a, B) translates as
\pi\X o pi{A\yi[a'(yi)(xo)]}(A B,).
130. I f a E
PIVi/Meas' 13 E PMeas ' and a,S translate as
a',S', then F 30 (a.S) translates as a,(A S ')'.
T31. If a E PIVs IINF i • 13 E PINFi. and a,S translate as

a', B', then F 31 (a.B) translates as a,(A S ').
T32. I f a E PIVi IINF i • 13 E PINFi, and a,S translate as

a', 13', then F 32 (a,S) translates as a,(A B,).
T33. If a E PIVs/ADJi. 13 E PADJi, and a,B translate as

a',B', then F 33 (a.B) translates as a,(A S').
T34. If a E PIVi/ IQ • 13 E P IQ • and a,B translate as a',B',
then F 34 (a, 13) translates as a'( AB,).
T35. If a E PIVs/IVs, 13 E PIVs, and a.B translate as a'.S',

then F35 (a,B) translates as a,(A B,).
T36. If a E PIVs, 13
Ppps, and a,B translate as a',B', then
E
F 36 (a,S) translates as \Xs[a'(X s ) & B'(x s )].
T37. If a,B E PIVs and a,B translate as ex',B', then F 37 (ex,B)

translates as \xs[ a'(X s ) & S'(x s ) J.
T38. If a E PTVs/IVi. B E PIVi. and a.B translate as a'.B'.

then F 38 (a. B) translates as a'("B').
T39. If a E PIVs and a translates as a'. then F 39 (a)

translates as a'.
T40. If a E PIVi and a translates as Ct', then F 40 (a)

translates as a'.
T41. If a E PTVs and a translates as a', then F 41 (a)

translates as AxSVys[a'(Xs)(ys)].
T42. If a E PTVs. SEPT' and a.B translate as a'.B'. then

F 42 (a.B) translates as Axs[B'("AyiVys[R(yS,yi) &
a' (x s ) (ys) J)].
T43. If a E PTVi and a translates as a'. then F 43 (a)

translates as AXivyO[o;'(xi)(y0)J.
T44. If ex E PTVi. B E PI' and ex.B translate as a'.B'. then

F 44 (a.B) translates as Axi[B'("Ayo[a'(Xi)(yo)])].
T45. If a E PIVs and

translates as a',
Ct then F 45 (a)
translates as Axs[Ing(" [a'( x s )])].
T46. If a E PIVi and a translates as a'. then F 46 (a)

translates as Axo[Ing("[a'(Xo)])]. Axk[ Ing(" [a'(X k )])],
or Axi[Ing("[a'(xi)])J. according as a' is of sorted
type <o.t'>. <k.t'>. or <[o.k].t'>. respectively.
T47. If a E PADJi IINFi. B E PINFi. and a. B translate as

a'.B'. then F 47 (a,B) translates as a'("B').
T48. If a E PeN and a translates as a', then F 48 (a)

translates as a'.
T49. If a £ PT/CN' B £ PCN' and a,B translate as ct'.B'. then

F 49 (a.B) translates as ex'("S').
388 APPENDIX
T50. If a PeN and a translates as a', then F 50 (a)

£:
translates as Apkpk{ 1x k Axo[R'(xo,x k ) .. a'(X o )]}.
T51. If a £: P T and a translates as a', then F51 (a)

translates as
APoAQo[Qo{1Xo[a,(AAXf[POSS(X~, XO) & po{xo}])]}].
T52. If a £: PPps/NOM' S £: P NOM ' and a,S translate as at,S'.

then F52 (a,S) translates as a,(A S ').
T53. If a £: PPps/T' S £: PT' and a,S translate as at,S', then

F 53 Ca,S) translates as a'CAS').
T54. If a £: Pppi/T' S £: PT' and a,S translate as at,S', then

F 54 Ca,S) translates as a,(A S').
T55. If a E: PT' <P E: Pt , and a,<P translate as a',tJ>', then

F55,nCOt,<P) translates as a'CAAx~[CP']) if there is some
nonnegative integer m such that n = 4m+3, and otherwise
as a'(AAx~[CP']).
T56. If cP E: Pt and cP translates as CPt, then F 56 (CP)

translates as At[AT(t, cpt)].
T57. If a E: P TAB and a translates as a', then F 57 (a)

translates as At[past(t) & a'(t)].
T58. If ex P TAB and a translates as a', then F 58 (a)

£:
translates as At[PRES(t) & ex'(t)].
T59. If a e; P TAB and ex translates as a', then F 59 (a)

transla tes as At [FUT( t) & ex'( t) ].
T60. If ot £: P TAB and ot translates as a', then F 60Ca)

translates as vt[ex'(t)].
T61. If ex E: PMTA , e £:
PTAB , and Ot,B translate as Ot',B', then
f 61a (a,6}, F 61b (ex,a) translate as ex'C .... B').
T62. If a E: PTAB1ITAB' S E: P TAB ' and a,S translate as at,S',

then F 62 (a,S) translates as at(ASt).
T63. If a E: PTA and a translates as a', then F 63 (a)

translates as APtx,t[at(t) & pt{t}).
T64. If a E: P RFA ' S E: PTA' Y E: PTAB , and a,S,y translate as

a',S',Y', then F 64a Ca,S,Y), F 64b Ca,S,Y) translate as
x,t[a'CS')CAt,[t, ~ t & y'(t,)])].
T65. If a E: PRFA , S E: PTAB , and a,S translate as a',S', then

F 65,nCa,S) translates as
At[a'CIn)(At,[t, ~ t & s'Ct,)])].
T66. If a E: PTA' S E: PTAB , and a,S translate as at,S', then

F 66a (a,S), F 66b Ca,S) translate as
H[G"Ca')(At,[t, ~ t & stet,)])].
T67. If a E: PTSC ' ¢ Pt , and a, ¢ translate as a', ¢', then

E:
F67,nCa,</» translates as a,(AAtn[</>'J).
T68. If a E: PTSC ' S E: PTAB , and a,S translate as a',B', then
F 68Ca,S) translates as a'CAx,t[past(t) & stet)]).
T69. If a E: PTSC ' S £ PTAB ' and a,S translate as a',S', then
F 69(a,S) translates as at(AAt[NONPAST(t) & s'(t)]).
nO. If a E: P TAB and a translates as at, then F 70 (a)

translates as At[AT(t, vt,[a'(t,)])].
REFERENCES
Babby, L. H.: 1975, 'The Syntax of Gerunds in Russian'.

Distributed by the Indiana University Linguistics Club.
Bach, E.: 1980, 'In Defense of Passive', Linguistics and
Philosophy 3, 297-342. -
Bennett, M. and B. H. Partee: 1972, 'Toward the Logic of
Tense and Aspect in English'. Distributed by the
Indiana University Linguistics Club.
Berent, G. P.: 1975, 'English Absolutes in Functional
Perspecti ve', in R. E. Grossman et al. (eds.), Papers
from the Parasession on Functionalism, Chicago
Linguistic Society. --
Beukema, F.: 1979, 'Some Reflections on 'With' in English',
Handelingen van he! vijfendertigste Nederlands
Filologencongres, Holland Universiteits Pers,
Amsterdam, pp.155-162.
Beukema, F.: 1980, 'Prepositions and Tenseless Sentence
Adverbials in English', in S. Daalder and M. Gerri tsen
(eds.), Linguistics in the Netherlands 1980, North-
Holland, Amsterdam an~New York. ----
Beukema, F.: 1982, 'On the Internal Structure of Free
Adjuncts', in S. Daalder and M. Gerritsen (eds.),
Linguistics in the Netherlands 1982, North-Holland,
Amsterdam and New York.
Beukema, F. and R. Verheijen: 1982, 'The Equi-noc-tial
Quandar~, Studia Anglica Posnaniensia 1_, 121-136.
Cantrall, W. R.: 1974, Viewppint, Reflexives, and the Nature
of Noun Phrases, Mouton, The Hague.
Carlson, G. N.: 1977, Reference to Kinds in English.
University of Massachusetts doctoral dissertation. [To
be published by Garland Publishing, Inc.]
390
REFERENCES 391
Carlson, G. N.: 1979, 'Generics and Atemporal When',

Linguistics and Philosophy 3, 49-98.
Chomsky, A. N.: 1982, Lectures on Government and Binding:
The Pisa Lectures (2nd revised edition), Foris,
Dordrecht.
Clark, H. H.: 1978, 'Inferring What is Meant', in W. J. M.
Levelt and G. B. Flores d'Arcais (eds.), Studies in the
Perception of Language, John Wiley and Sons,
Chichester.
Cooper, R.: 1975, Montabue's Semantic Theory and
Transformational syn~universftyof Massachusetts
doctoral dissertation.
Curme, G. 0.: 1931, A Grammar of the English Language, Vol.
III (Syntax), D--:- C. Heath and Company, Boston. --
van Dijk, T. A.: 1977, Text and Context: Explorations in the
Semantics and Pragmatics of Discourse, Longman, London.
Dowty, D. R.: 1972, 'Temporally Restrictive Adjectives,' in
J. Kimball (ed.), Syntax ~E.i Semantics, vol. 1,
Academ ic Press, Ne w York.
Dowty, D. R.: 1979, liord t"ieaning and Montague Gra~.!!!~,
Reidel, Dordrecht.
Dowty, D. R.: 1982, 'Tenses, Time Adverbs, and Compositional
Semantic Theory', Linguistics and Philosopy 5, 23-55.
Dowty, D. R., R. E. Wall, and S. Peters: 1981, Introduction
to Montague Semantics, Reidel, Dordrecht.
Farkas, D. F. and Y. Sugioka: 1983, 'Restrictive If/When
Clauses', Linguistics and Philosophy 6, 225-258.
Fillmore, C.: 1969, 'Toward a Modern Theory of Case', in D.
A. Reibel and S. A. Schane (eds.), ~odern Studies in
English, Prentice-Hall, Englewood Cliffs, ~ J.
Fukuchi, S.: ms, 'A Note on Participial Constructions in V4
System' •
Gazdar, G., G. K. Pullum & 1. A. Sag: 1982, 'Auxiliaries and
Related Phenomena in a Restrictive Theory of Grammar',
Language 58, 591-638.
Geis, M. L.: 1970, Adverbial Subordinate Clauses in English.
MIT doctoral dissertation.
392 REFERENCES
Geis, H. L.: 1974, 'What Do Place Adverbials Modify?', Ohio

state University Working Papers in Linguistics 18. The
Ohio State University Department of Linguistics.
Geis, H. L.: ms, 'The Syntax and Semantics of English
Indicative Conditionals'.
Grady, M.: 1972, 'The English Absolute Construction',
Linguistics 90, 5-10.
Grice, H. P.: 1975, 'Logic and Conversation,' in D. Davidson
and G. Harman (eds.), The Logic of Grammar, Dickenson,
Encino. ----------
Halliday, H. A. K. and R. Hasan: 1976, Cohesion in English,

Longman, London.
Heinamaki, 0.: 1974, Se.!!!antics of Eng~ish Temporal
Connectives. Distributed by the Indiana University
Linguistics Club.
Ishihara, R.: 1981, 'On Certain Prepositional Phrase
Modifiers', Linguistic Notes from La Jolla 9, 43-63.
Is hihara, R.: 1982, A Study of Absol ute Phrases in E ngli sh
Within the GoVernmentB1nding Framework. University of
California, San Diego, doctoral dissertation.
Jespersen, 0.: 1977, ! Hodern English Grammar on Historical
Principles, part V, George Allen and Unwin, London.
Kratzer, A.: 1977, "What 'Nust' and 'Can' Must and Can
Mean," Linguistics and Philosophy 1, 337-355.
Kratzer, A.: 1978, Semantik der Rede, Scriptor, Konigstein.
Kratzer, A.: 1979, 'Conditional Necessity and Possibility',
in R. Bauerle et a1. (eds.), Semantics fro..!!! Different
Points of View, Springer, Berlin.
Kratzer, A.: 1981, 'The Notional Category of Modali ty', in
H.-J. Eikmeyer and H. Rieser (eds.), Words, Worlds, and
Contexts, Walter de Gruyter, Berlin and New York. --
Kruisinga, E.: 1931, ! Handbook of Present-Day English, part
II.l, 5th edition, P. Noordhoff, Groningen.
Kruisinga, E.: 1932, ! Handbook of Present-Day English, part
II.3, 5th edition, P. Noordhoff, Groningen.
REFERENCES 393
Lakoff, G.: 1971, 'The Role of Deduction in Grammar', in C.

Fillmore and D. T. Langendoen (eds.), Studies in
Linguistic Semantics, Holt, Rinehart and Winston, New
York.
Lewis, D. K.: 1973, Counterfactuals, Harvard University
Press, Cambridge.
Lewis, D. K.: 1975, 'Adverbs of Quantification', in E. L.
Keenan (ed.), Formal Semantics of Natural Language,
Cambridge University Press, Cambridge.
McCawley, J. D.: 1971, 'Tense and Time Reference in
English', in C. Fillmore and D. T. Langendoen (eds.),
Studies in Linguistic Semantics, Holt, Rinehart and
Winston, New York.
McCawley, J. D.: 1978, 'Conversational Implicature and the
Lexicon', in P. Cole (ed.), Syntax and Semantics, vol.
9. Academic Press, New York. ---
McCawley, J. D.: 1983. 'What's with With?', Language 59,
271-287.
McCoard, R. W.: 1978, The English Perfect: Tense-Choice and
Pragmatic Inferences, North-Holland, Amsterdam.
Montague, R.: 1970a, 'English as a Formal Language', in B.
Visentini et al. (eds.), Linguaggi nella Societa ~
nella Tecnica, Edizioni di Communita, Milan.
[Reprinted in Montague (1974)]
Hontague, R.: 1970b, 'Universal Grammar', Theoria 36, 373-
398. [Reprinted in Montague (1974)]
Montague, R.: 1973, 'The Proper Treatment of Quantification
in Ordinary English', in J. Hintikka et al. (eds.),
Approaches to Natural Language, Re i de l, Dordrecht.
[Reprinted in-Montague (1974)]
Montague, R.: 1974, Formal Philosophy: Selected Papers of
Richard Nontague, ed. by R. Thomason, Yale Uni versi ty
Press, New Haven.
Nash-Webber, B. L. and I. Sag: 1978, 'Under Whose Control?',
Linguistic Inquiry 9, 138-141.
Onions, C. T.: 1971, Modern English Syntax (revised edition
of An Advanced English Syntax), st. Martin's Press, New
York.
394 REFERENCES
Partee, B. H.: 1976, 'Some Transformational Extensions of

I"lontague Gram mar', in B. H. Partee (ed.), Montague
Grammar, Academic Press, New York.
Partee, B. H.: 1977, 'John is Easy to Please', in A.
Zampolli (ed.), Linguistic Structures Processing,
North-Holland, Amsterdam.
Partee, B. H.: 1979a, '14ontague Grammar and the VJell-
Formedness Constraint', in F. Heny and H. S. Schnelle
(eds.), Syntax and Semantics, vol. 10, Academic Press,
New York.
Partee, B. H.: 1979b, 'Semantics--Mathematics or
Psychology?', in R. Bauerle et a1. (eds.), Semantics
from Different Points of View, Springer, Berlin.
Partee, B. H.: 1983a, 'Compositionality', draft of paper
presented at the 1982 Amsterdam Colloquium on Formal
Linguistics.
Partee, B. H.: 1983b, 'Formal Semantics and Conceptual
Structure', lecture handout.
Postal, P.: 1970, 'On Coreferential Complement Subject
Deletion', Linguistic Inquiry 1, 439-500.
Pullum, G. K.: 1974, 'Restating Doubl-ing', Glossa 8, 109-
120.
Pullum, G. K. and D. Wilson: 1977, 'Autonomous Syntax and
the Analysis of Auxiliaries', Language 53, 741-788.
Putnam, H.: 1975, "The Heaning of 'Meaning''', in K.
Gun d e r son (e d .), Lan g u age, !i i n ~ ..§.~ Kn 0 ~~ d g e ,
University of Minnesota Press, Minneapolis.
Quirk, R., S. Greenbaum, G. Leech, and J. Svartvik: 1972, A
Grammar of Contemporary English, Seminar Press, New
York.
Heuland, E. J.: 1983, 'Governing -ing', Linguistic Inquiry
14, 101-136.
van Riemsdijk, H.: 1978, !l Case Study in Syntactic
l'larkedness, Foris, Dordrecht.
Ross, J. H.: 1972, 'Doubl-ing,' Linguistic Inquiry 3, 61-86.
Ross, J. R.: 1973, 'A Fake NP Squish', in C.-J. N. Bailey
& R. W. Shuy (eds.), New Ways of Analyzing Variation in
English, Georgetown University Press, Washington, D. C.
REFERENCES 395
Ruwet, N.: 1978, 'Une construction absolue en frangais',

Lingvisticae Investigationes 2, 165-210.
Sakakibara, H.: 1982, 'Hith-Constructions in English',
Studies in English Literature, 1982 Enblish Number, 79-
95. Tokyo: Enblish Literary Society of Japan.
Scheffer, J.: 1975, The Progressive in English, North-
Holland, Amsterdam.
Smith, C.: 1975, 'The Analysis of Tense in English', Texas
Linguistic Forum 1, 71-89.
Stalnaker, n.: 1978, 'Assertion', in P. Cole (ed.), Syntax
and Semantics, vol. 9, Academic Press, New York.
Stump, G. T.: 1980a, 'Two Varieties of Free Adjuncts in
English', Paper presented at the LSA Summer Meeting,
Albuquerque, New Mexico.
Stump, G. T.: 1980b, 'Aspect in Free Adjuncts,' in M. H. T.
Ii end e r son (e d .) , 1 9 8 Q l1.!.~=-~.!E e ric..§. Lin g u i s tic s
Conference Papers, Uni versi ty of Kansas Department of
Linguistics, Lawrence.
Stump, G. 1.: 1981a, 'The Interpretation of Frequency
Adjectives', Linguistics and Philosophy 4, 221-258.
Stump, G. T.: 1981b, The Formal Semantics and Pragmatics of
Free Adjuncts and Absolutes in English. Ohio State
University doctoral dissertation.
Taylor, B.: 1977, 'Tense and Continuity', Linguistics and
Philosophy 1, 199-220.
Vendler, Z.: 1967, Linguistics in Philosophy, Cornell
Uni versi ty Press, Ithaca.
Visser, F. Th.: 1972, An Historical Syntax of the English
Language, part II, E. J. Brill, Leiden.
vi arner,R. G.: 1979, Discourse Connecti ves in English. Ohio
State University doctoral dissertation.
Hilliams, E. S.: 1975, 'Small Clauses in English', in J.
Kimball (ed.), Syntax and Semantics, vol. 4, Academic
Press, New York.
Hilliams, E. S.: 1977, 'Discourse and Logical Form',
Linguistic Inquiry 8, 101-139.
INDEX OF NAMES
Babby, L. H. 39 Greenbaum, S. 14, 17

Bach, E. 35, 36, H1 Grice, H. P. 146, 155
Bennett, M. 101 , 212,
213, 255 Halliday, M. A. K. 323
Berent, G. P. 38, 39, Hasan, R. 323
292, 336 Heinamaki, O. 145, 146,
Beukema, F. 38, 39, 291, 147ff, 212
292
Ishihara, R. 39, 96,
Cantrall, W. R. 38 291, 292
Carlson, G. N. 24, 42,
64ff, 71ff, 86, 89, Jespersen, O. 12, 14-16,
94, 95, 193-196, 198, 18, 19, 38, 39, 300,
199, 270, 308ff, 351 347, 348
Chomsky, A. N. 38
Clark, H. H. 304, 344 Kratzer, A. 23, 42, 44,
Cooper, R. 351 45ff, 89, 90
Curme, G. O. 12, 14, 19, Kruisinga, E. 14, 15,
300 17, 20, 300, 331
van Dijk, T. A. 323 Lakoff, G. 344

Dowty, D. R. 40, 95, Leech, G. 14, 17
101, 102, 104, 106, Lewis, D. K. 90, 218,
125, 151, 209-212, 228, 313
255, 256, 263, 266-268,
309, 343, 353, 355 McCawley, J. D. 39, 95,
96, 212, 233, 263, 265,
Farkas, D. F. 218 266, 269, 291, 295, 296
Fillmore, C. 95 McCoard, R. W. 220, 223,
Fukuchi, s. 39, 291 225, 265
Fuller, S. D. 5 Montague, R. 3, 91, 94,
340, 341, 350, 367
Gazdar, G. 267
Geis, M. L. 92, 94, 130, Nash-Webber, B. L. 344
134, 148
Grady, M. 39, 334 Onions, C. T. 347
396
INDEX OF NAMES 397
Partee, B. H. 77, 101, 292, 295
212, 213, 255, 340ff Smith, C. 145, 146
Postal, P. 39 Stalnaker, R. 146, 211
Pullum, G. K. 30, 37, Sugioka, Y. 218
267 Svartvik, J. 14, 17
Quirk, R. 14, 17, 20, Taylor, B. 151

21, 39, 300, 306, 336
Verheijen, R. 38, 39
Reuland, E. J. 38, 39 Visser, F. Th. 14, 16,
van Riemsdijk, H. 39, 17, 37-39, 290, 347
96, 290, 292, 296
Ross, J. R. 6, 38 Warner, R. G. 323
Ruwet, N. 38, 39, 96, Williams, E. S. 39, 213
291, 292, 295, 296 Wilson, D. 30
Sag, I. A. 267, 344 Yava§, F. 346

Sakakibara, H. 39, 291,
GENERAL INDEX
A-abstracts 163f, 165f, variability 14ff; see

167ff, 185, 206, 213- also augmented abso-
215, 242-245, 268, 277, lutes, free adjuncts,
279, 282, 287 nominative absolutes
absolute constructions: accomplishment predicates
adversative function 149, 153, 154, 156-158,
59, 324, 325, 345; 255, 259, 319, 320, 322
causal function 14-18, achievement predicates
58, 306, 308ff, 313ff, 149, 153, 154, 156-158,
317, 324, 342, 345; 212, 255, 259, 319,
circumstantial function 320, 322
14-18, 331ff; conces- activity predicates 149,
sive function 14, 17; 151-153, 156-158, 160,
conditional function 267, 319, 320; honJo-
14, 16, 17, 23f, 272ff, geneous 151; heteroge-
279f, 302, 308ff, 313, neous 151, 152
336f; contrastive func- [+Adjunct] (binary fea-
tion 15; development -ture) 27, 163, 206,
as prepositional 209, 216, 277
phrases or adverbial ad sentential adjuncts
subordinate clauses 58, 62, 85, 89, 163,
19, 290, 316, 346f; 166ff,206: see also
semantic variability 3, free adjuncts
300ff; sensitivity to adverbial subordinate
context 15, 18, 20f, clauses: as underlying
330f, 341ff; special source for absolute
uses 331ff; syntactic constructions 329ff
correlates of semantic AGR (syntactic subrou-
variability 18ff; tem- tine) 35
poral function 14-18, as 86ff, 95, 96, 274,
24, 188 f , 204, 222, 346
265, 281f, 283f, 302, aspect 225, 253; see
308ff, 313, 316f, also progressive aspect
318ff, 324, 325, 326ff, AT----(temporal operator)
345: traditional 102, 162, 172
thouBhts on semantic augmented absolutes 96,
398
GENERAL INDEX 399
259, 271, 292, 294; as conditional adjuncts 56,
arguments of frequency 62, 85, 89, 165f, 206,
adverbs 281f, 308, 216, 217, 279; see also
310, 320f; as arguments free adjuncts
of G" 283f, 308, 310, conditionals 41ff, 47ff,
320f; as arguments of 50, 52f, 90, 91ff
modals 308, 313, 336f; connective adverbs 58,
as main tense adverbs 93, 322ff
285f, 318ff; condi- contexts of use 340ff
tional interpretation control 7f, 37, 93
273ff, 279f; formal conversational back-
characteristics 13, ground 45ff, 90
290f; p~esent partici- conversational economy
pial 322; strong 146, 212, 233ff, 263,
273ff, 277ff, 280, 265f, 269
281f, 283f, 300ff, 308; CONWIB (syntactic sub-
weak 273ff, 277ff, routine) 35
280ff, 283f, 300ff,
308; see also absolute D (intensional logic con-
constructions stant) 52
augmented adjuncts 13, dangling participle 7
17, 258, 3i1 DB (syntactic subroutine)
33
bare plurals 68ff, 73f, declarative sentences 50
94, 198 DEL (syntactic subrou-
BASE (morphological func- tine) 34
tion) 30 Doubl-ing Constraint 6,
basic category 91 269
be 73ff, 86ff, 95, 96, DP (syntactic subrou-
313, 345, 346; active tine) 33
be 76f; passive be 79 Dutch 96, 290
binary features 27f
ellipsis 334
C (intensional logic con- EN (syntactic subroutine)
stant) 52 33
Case Filter 38 Equi-NP Deletion 38
cb (intensional logic extended nows 224
constant) 52
complementizer deletion free adjuncts 272, 289;
291 as arguments of
compositionality 93, frequency adverbs
235, 300, 340ff 184ff, 302, 308, 310,
compounds 344 320f, 337ff; as argu-
CON (syntactic subrou- ments of G" 203ff,
tine) 35 302, 308, 310, 320f; as
400 GENERAL INDEX
arguments of modals 184ff, 199, 207, 215,
53ff, 56f, 86, 308, 270, 280,281,297,
313, 336f: as main 310; semantics 175ff
tense adverbs 167ff,
220f, 244, 247ff, G (generalization opera-
249ff, 253, 265, 285, tor) 193ff
302f, 318ff, 337ff: G' (generalization opera-
conditional interpreta- tor) 193ff
tion 41ff, 53ff, 56f: G" (generalization opera-
correlation with Carl- tor) 100, 104, 190ff,
son's ontology 64, 199ff, 203ff, 207, 218,
84ff: entailment pro- 270, 280, 283f, 298.
perties 43: formal 302, 308, 310
characteristics 4ff; generic sentences 68ff,
free variable in sub- 99f, 190ff, 193ff,
ject position 57, 164; 203ff, 207, 219, 283f
misrelated 7; perfect genitives 344
220, 222ff, 229ff, 242; gerunds 11-13, 39, 231
present participial Government and Binding
221f, 253ff, 257f, 38, 291
261ff, 265, 322; re- grammaticization 235
lated 7; strong 43ff,
53ff, 58ff, 63, 84ff, have 86ff, 95, 96, 295f,
89, 97ff, 164, 172ff, 344
187, 188, 190, 191, hypot.axis 277
204, 206,207,270,
272, 296, 300ff, 308, I (intensional logic
318; temporal interpre- variable) 182, 188f,
tation 98ff; temporal 338
reference 97, 162ff; idioms 315, 348
unrelated 7; weak imperfective paradox 255
43, 53ff, 56f, 62, 63, imperfectiveness 257-
84ff, 89, 97ff, 164, 259, 261-263, 322
172ff, 185, 188, 190, implicatures 327, 336;
191, 203, 206, 207, 217 conventional 79, 210;
270, 272, 280, 282, conversational 233,
284, 296, 300ff, 308; 234, 263, 265f, 269
see also absolute indeterminacy 60, 90,
constructions 306f
french 96, 274-276, 291, index of possible utter-
292, 315, 346, 347 ance 102
frequency adverbs 98ff, individual-level predi-
104, 172ff, 302, 308, cates 24, 26, 64ff,
337ff, 348; fixed 71ff, 84ff, 89, 194ff,
173ff; relative 173ff, 206, 213, 243, 270-274,
GENERAL INDEX 401
279, 294, 295, 308ff, 48, 51, 90

313ff, 345; diagnostic logical consistency 90
environments 70
individuals 42, 68, 214, M (intensional logic
309 variable) 122, 170,
inertia worlds 256, 267, 186, 208, 211, 216,
268 217, 261, 262, 290,
inferences 20ff, 26f, 302, 303, 307, 318,
60ff, 169ff, 186, 188f, 338, 341ff, 347; prag-
204, 208, 216, 258, matically admissible
259, 262, 270, 290, values 171, 325ff, 343
296, 300, 302f, 304ff, main tense adverbs 118,
325ff, 330, 335ff, 137ff, 160, 206, 209,
340ff, 348; deductive 210, 216, 220, 226,
344f; factors contribu- 227, 239, 240, 247,
ting to 307ff 248ff, 253, 265, 270,
ING (syntactic subrou- 285, 302f; see also
tine) 33 time adverbs
instantaneousness 318ff main verb bracketing 31f
interval 101ff Middle English 346
interval statives 151- meaning 340ff
153, 157, 158, 160 modals 40, 41ff, 50,
256, 270, 280, 302,
K (intensional logic con- 313, 337; semantics
stant) 61, 167, 169, 45ff, 51ff
216, 277, 336, 337 momentary predicates
kinds 68, 194, 195, 198, 253, 263, 266
214, 267, 308 morphological functions
knowledge of the world 29f
321f, 342f, 348
narrative present 124,
L (intensional logic 204
variable) 163, 167, nominative absolutes
169, 170, 186, 208, 271, 294; as arguments
216, 277, 290, 303, of frequency adverbs
307, 336-338, 341ff; 296ff, 310; as
factivity 61, 94; arguments of G" 298,
logical type 60; prag- 310; as main tense
matically admissible adverbs 285f, 318ff;
values 60, 171, 325ff, formal characteristics
339, 343 8ff; present partici-
Latin 38 pial 231, 322; strong
logical compatibility 272, 280, 281f, 283,
49, 51, 90 300ff, 308; see also
logical consequence 46, absolute constructions
402 GENERAL INDEX
NONF (syntactic subrou- progressive aspect 5,

tine) 33 10, 26, 70, 73ff, 86,
220ff, 253ff, 257, 263,
OBJ (syntactic subrou- 264, 266-268; semantics
tine) 34 255ff
objects 68, 194, 195, PRPL (morphological func-
197, 214, 267, 308 tion) 29
Old English 37, 38 PSPL (morphological func-
tion) 29
parataxis 277
passive voice 5, 10, RB (syntactic subroutine)
79ff, 81ff, 345 34
past interval 224 realization relation 68,
PAST 1 (morphological 195
function) 29 reference point 158
PAST2 (morphological HOS (syntactic subrou-
function) 29 tine) 34
[+Perfect] (binary fea- Russian 39
-ture) 28, 209, 214, RWRAP (syntactic subrou-
216, 236, 238f, 242, tine) 35
243, 245, 266
perfect interval 228 sequence of tense 114f
perfect tense 5, 10, 26, set-level time adverbs
220ff, 222ff, 236ff, 28, 116, 175, 184ff,
248ff, 254, 264-266, 199, 203, 206, 210,
268, 286, 287, 316; 216, 280, 281, 283; see
'extended now' theory also time adverbs
220, 224ff, 232, 234, shorthand expressions
235, 245; semantics 304f, 344
236ff; and time adverbs Since Subject Erasure 39
223, 225, 229ff, 239f, speaker's interval 102,
245ff, 248ff 108ff, 116f,' 177, 286,
PLUR (morphological func- 288, 317
tion) 30 SSDELn (syntactic subrou-
[+Plural] (binary fea- tine) 34
-ture) 28 stage-level predicates
PRB (syntactic subrou- 24, 26, 64ff, 71ff,
tine) 35 84ff, 89, 165, 193ff,
PRESl (morphological 198, 206,242,268,
function) 29 270-274, 279, 295, 297,
PRES2 (morphological 298, 308ff, 313ff , 345;
function) 29 diagnostic environments
presupposition 146, 211 67, 70, 73, 86ff, 199,
PRET (syntactic subrou- 313
tine) 34 stages 42, 68, 195, 197,
GENERAL INDEX 403
214, 267, 308, 309 205, 206, 221, 224,

stative predicates 65, 225, 227, 231, 237,
85, 94, 148, 151-153, 238, 247, 248, 267,
156-158, 263, 319, 320 280, 286, 287, 316,
subinterval 101 317: shifting interpre-
SUBJ (syntactic subrou- tation 108ff, 116f.
tine) 34, 35 124f.211
subject parenthesization [~Tense] (binary feature)
32 28, 266
syntactic subroutines 32 time adverbs 116ff. 206:
multiple 103f, 120f:
temporal abstracts 27, and perfect tense 223,
28. 31, 32, 39, 97ff, 225ff, 229ff, 239f,
100f, 103ff. 127. 129, 245f, 248ff: shifting
163ff, 178, 183, 206~ interpretation 177:
209, 210, 214, 236-239, see also main tense
268 adver"bS:-set-level time
temporal adverbial adverbs, temporal
clauses 121 ff , 137ff, adverbial clauses
211: 'multiple inter- topicalization 344
val' interpretation transitive verb phrases
192ff, 195, 196ff, 218; 79ff
'single interval' Turkish 346
interpretation 192ff,
196ff; tense of 123ff: unspecificity 306f
truthconditions 147ff;
unacceptable 144ff; variable binding 131
unbounded dependencies
in 130ff. 148, 212, [+Weak] (binary feature)
347; ~ also time 28, 165, 168, 172, 185,
adverbs 206, 242, 243, 277,
temporal subordinating 279, 282, 311, 312
conjunctions 122, 127, with 86ff, 95. 96, 274,
132 294ff
tense 105f1', 123ff, 145, word order 321, 348
160, 162, 177, 189.

(Synthese Language Library 25) Gregory T. Stump (Auth.) - The Semantic Variability of Absolute Constructions-Springer Netherlands (1985)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Synthese Language Library 25) Gregory T. Stump (Auth.) - The Semantic Variability of Absolute Constructions-Springer Netherlands (1985)

Uploaded by

Copyright:

Available Formats

THE SEMANTIC VARIABILITY OF ABSOLUTE CONSTRUCTIONS

SYNTHESE LANGUAGE LIBRARY

TEXTS AND STUDIES IN

D. REIDEL PUBLISHING COMPANY

(Synthese language library; v. 25)

Published by D. Reidel Publishing Company,

Sold and distributed in the U.S.A. and Canada

In all other countries, sold and distributed

All Rights Reserved

I. THE SEMANTIC VARIABILITY OF FREE ADJUNCTS AND

3. A Semantic Correlate of the Distinction

III. TENSE AND THE INTERPRETATION OF FREE ADJUNCTS 97

1.4.5.4. ACter 159

IV. ASPECT AND THE INTERPRETATION OF FREE ADJUNCTS 220

V. THE FORMAL SEMANTICS OF ABSOLUTES 270

3. Absolutes as Main Tense Adverbs 285

VI. INFERENCE AND THE LOGICAL ROLE OF FREE ADJUNCTS

APPENDIX - A FORMAL FRAGMENT FOR FREE ADJUNCTS AND

The goal of this book is to investigate the semantics of

comprehension are not without relevance for referential

Partee, Mehmet and Feryal Yava§, Arnold Zwicky, and an

Excerpts from Elizabeth Drew, Portrait of an Election:

Excerpt from Brendan Gill, 'The Horizontal City,' The New

Sentence excerpted from The Oregonian, 9/14/80, p.C4,

THE SEMANTIC VARIABILITY

My purpose in this book is to investigate the semantic

Walking home, he found a dollar.

b. The nominative absolute construction:

His Cather being a sailor, John knows all

c. The augmented absolute construction:

With the children asleep, Mary watched TV.

These constructions have the bross syntactic characteristics

(2) The school is determined to avoid a scandal. The

(3) Grabbing a newspaper from a guard, Tom went back

(4) Transposed to a trumpet or saxophone, her

(5) Dead, they have taught the living.

The logical roles played by the free adjuncts in these

(6) Picasso, to himself and to many others~ was a god.

As in the cases in (2)-(5), English speakers have no

1. Introduction to Free Adjuncts and Absolutes

Here, I briefly revi~w the defining characteristics of the

1.1. THE FREE ADJUNCT CONSTRUCTION

A free adjunct is a nonfinite predicative phrase with the

(8) A center for shoe factories and breweries early in

(9) Unable to meet his eyes, Kate looks down at her

(10) The Second World War began, and, still scarcely in

(11) The W. S. Q. imitated the Ellington reed section,

In the following chapters, attention will be concentrated

(12) Glancing up at the Arch over the downtown skyline

(13) Published here in 1972, Thomas Keneally's novel is

(14) To tell you the truth, I have never really thought

A verbal adjunct may be in the active voice, as in (12),

(15) ••• it was the home of a hardy race of Gaelic-

(16) Arrived at the spot, the party lost no time in

It may show perfect 'aspect' alone, as in (17); or the

(17) Having come to this conclusion, I settled down to

(18) He was then drunk, having been all night taking

The progressive may not occur alone, as (19) suggests: this

might be attributed to Ross' (1972) Doubl-ing Constraint;4

(19) *Being eating, John couldn't answer.

Free adjuncts may be dependent upon a subordinate

(21) But when the man, baving backtracked, picked up

(22) He was called by Strauss, by Jordan, and by the

They normally occur either immediately before or immediately