GRAMMATICALIZING ASPECT AND AFFECTEDNESS
by
Carol Lee Tenny
B.A., Oberlin College
(1977)
Submitted to the Department of Linguistics and Philosophy
in Partial Fulfillment of the Requirements
for the Degree of
DOCTOR OF PHILOSOPHY
at the
MASSACHUSETTS INSTIT!'TE OF TECHNOLOGY
August 1987
Q Carol Lee Tenny !987
The author hereby grants to MIT permission to reproduce and to
distribute copies of this thesis document in whole or in part.
Signature of Author-
Department offinguistics and Philosophy
August 1987
Certified by_
Kenneth Hale
Thesis Supervisor
Accepted by
Morris HalleMA 4 SSFUTrsNs)airman, Departmental Committee
OCT 2 3 1987
LIBRARIES
rCalvd
IMLh ,-
MITLibraries
77 Massachusetts Avenue
Cambridge, MA 02139
http://Iibraries.mit.edu/ask
DISCLAIMER NOTICE
Due to the condition of the original material, there are unavoidable
flaws in this reproduction. We have made every effort possible to
provide you with the best copy available.
Thank you.
Some pages in the original document contain text
that runs off the edge of the page.
Missing Page Number 149
Grammaticalizing Aspect and Affectedness
by
Carol Lee Tenny
Submitted to the Department of Linguistics and Philosophy
on August 24, 1987 in partial fulfillment of the requirements
for the Degree of Doctor of Philosophy in Linguistics
ABSTRACT
This thesis is an investigation of the interaction of aspect
and syntax. More particularly, the syntactic repercussions of
the aspectual property of delimitedness are examined.
Delimitedness -- the temporal boundedess of an event -- is
shown to have an effect on a wide range of syntactic
phenomena, including resultative secondary predicates,
verb-particle constructions, and certain case phenomena.
Affectedness is also shown to depend on delimitedness. The
interaction between affectedness and syntax is proposed to
take place in the Case module of the grammar.
An analysis of the property of affectedness in aspectual
terms leads to a theory in which the direct argument of a verb
'measures out' the event described by the verb over time, as
if on a scale. Affected arguments are direct arguments that
delimit the event on that scale. Non-affected direct
arguments also 'measure out' the event, though they do not
delimit it.
This aspectual property of direct arguments is the first of
a set of aspectual principles of argument structure. Three
additional principles are proposed: An event may be delimited
only by its internal arguments -- arguments within the verb
phrase at deep structure. Indirect arguments may delimit the
event parasitically through the direct argument, while
external arguments may not delimit the event at all.
Secondly, there may be only one 'delimiting' to a verb
phrase. And finally, secondary objects are always delimiting
elements.
Two specifically syntactic issues are addressed. First, it
is proposed that aspect is a syntactic category, and several
possible instantiations of aspect in phrase structure are
discussed. Secondly, the aspectual principles of argument
structure are applied to verb-particle constructions,
resultative secondary predicates, and double object
constructions; and these principles are shown to shed some
light on the syntactic behavior and structure of these
constructions. The usefulness of aspect as a tool for
syntactic investigations is demonstrated.
-2-
The' aspectual principles.of argument structure place
constraints on the kind of event participants that can be
internal arguments. In this way these principles provide a
principled mapping between the 'meaning' of verbs and their
syntactic representations. The Aspectual Interface Hypothesis
is proposed, which maintains that the two systems
'communicate' only through a common aspectual vocabulary.
Under this view only the aspectual information in thematic
roles is visible to the syntax, and thematic hierarchies are
not necessary in the mapping between 'meaning' and syntax.
The Aspectual Interface Hypothesis is consonant with a highly
autonomous syntax.
Thesis Advisor: Dr. Kenneth Hale
Title: Ferrari P. Ward Professor in Linguistics
-3-
Acknowledgements
I wish to thank my committee: my thesis advisor Ken Hale,
whose deep appreciation and feel for language was a source of
guidance and inspiration; Noam Chomsky, who by challenging me
at every turn forced me to struggle for clarity and precision;
and Richard Larson, who kept me focused and disciplined, with
my feet on the ground. I would like to thank three others who
also made special contributions: Sylvain Bromberger, who
guided me through the pitfalls of linguistic philosophy (any
mistakes in that department are my own); Beth Levin, who read
numerous drafts with great care; and Morris Halle, who
followed this work with interest, and whose lectures to me
about Linguistics always turned out to be lectures on Life. I
also received valuable comments and criticisms from: Steven
Abney, Jim Higginbotham, John Lumsden, Betsy Ritter, Barry
Schein and Peggy Speas.
I am grateful to the MIT Lexicon Project and the Nippon
Electric Company (and to Jay Keyser -- the mover and shaker)
for support and funding of a project on Japanese verbs, in
which I participated in the summer of 1985. I am also grateful
to the MIT Japan Science and Technology Program and the Athena
Language Project for a grant which I was awarded in 1985 to
develop a set of parsing rules for Japanese.
The members of the MIT linguistics community, full-timers
and passers-through, who have contributed in one way or
another to my education and experience here are almost too
numerous to mention. Here are a few I must thank, with
apologies to those whose names slipped through: Steven Abney,
Mark Baker, Andy Bares, Maggie Browning, Maggie Carracino,
Hyon Sook Choe, Jennifer Cole, Viviene Duprez, David Feldman,
Naoki Fukui, Alicia Gorecka, Carol Goslant, Isabelle Haik, Jim
Harris, Jim Higginbotham, Ewa Higgins, Alana Johns, Kyle
Johnson, Myung Yoon Kang, Kate Kearns, Mary Laughren, Beth
Levin, Juliette Levin, John Lumsden, Dinette Massam, Kiyoko
Masunaga, Jim McCloskey, Janis Melvold, Shigeru Miyagawa, Shin
Oshima, Nancy Peters, Tova Rapoport, Betsy Ritter, Luigi
Rizzi, Marc Ryser, Doug Saddy, Mamoru Saito, Barry Schein, Ur
Schlonsky, Gabriel Segal, Concepta Siembab, Brian Sietsema,
Michelle Sigler, Kelly Sloan, Peggy Speas, Richard Sproat,
Donca Steriade, Lisa Travis, Loren Trigo, Robert Van Valin.
I also wish to thank the Department of Linguistics for
taking me on when, in my late twenties, with a confused
history of work and study' behind me, I proposed to do graduate
work in a field I had never studied. In spite of all
frustrations, I have been grateful for the opportunity to
study what fascinates me.
-4-
For moral support during the writing of this thesis I owe a
special thanks to Steve and Nina Abney, Keith Bupp, Jennifer
Cole, Kim Jonas, John Lumsden, Kiyoko Masunaga, Jim Page, Ann
Renner, Betsy Ritter and her mother, Mrs. Kkthe Ritter, Kelly
Sloan; Ruth and Fred Hall, and my parents, brother and
sister.
Finally, a thanks to the dinosaurs in the Royal Ontario
Museum, who gave me my best idea.
This thesis is dedicated to my parents, Francis Briggs Tenny
and Nancy Naramore Tenny, who taught me curiosity, and to do
things for the right reasons.
"Anybody knows that if you believe what we tell you here
it's your own fault."
-- Morris Halle
-6-
Table of Contents
Chapter 1. Introduction 9
1.1 Introduction 9
1.2 Aspect 12
1.3 Delimitedness 17
Appendix: Overview of literature on aspectual verb classes 28
Chapter 2. The Grammatical Reality of Affectedness
and 4elimitedness 36
2.1 English verb-particle combinations 36
2.2 English resultative secondary predicates 42
2.3 Finnish accusative and partitive case. 47
2.4 Morphological aspect 51
2.5 Affectedness 58
2.5.1 English middles and noun phrase passivization 59
2.5.2 Japanese numeral quantifiers 63
2.5.3 Total affectedness 66
2.5.4 Affectedness and transitivity 69
Chapter 3. An Aspectual Theory of Affectedness 75
3.1 Affectedness as an aspectual property 75
3.1.1 Verbs of consumption and creation 76
3.1.2 Verbs of change of physical state 87
3.1.3 Verbs of abstract change of state 99
3.1.4 Achievement verbs 100
3.1.5 Verbs of motion 102
3.1.6 Affectedness as an aspec+4al property 105
3.2 Lexical entries for affectedness verbs 107
3.3 Affectedness and the compositionality of
delimitedness 112
3.3.1 Translation of spatial delimitedness into
temporal delimitedness 112
3.3.2 Parallels between the count/mass
and delimited/non-delimited distinctions 120
3.3.3 Some semantic approaches to the compositionality
of delimitedness: Verkuyl, Hinrichs, Dowty 125
3.4 Total affectedness 137
3.5 Delimitedness and case 142
Chapter 4. Aspectual Principles of Argument Structure 147
4.1 Introduction 147
4.2 Direct arguments 149
4.2.1 Verbs of imparting motion 151
4.2.2 Unergatives, reflexives and cognate objects 153
4.2.3 Verbs describing events with no change in the
internal argument 155
4.2.4 Conative and antipassive alternations 158
4.3 External arguments 162
-7-
4.3.1 External arguments and thematic roles 162
4.3.2 External arguments and the translation of spatial
delimitedness into temporal delimitedness 174
4.3.3 Summary 179
4.4 Oblique arguments 179
4.4.1 Delimiting oblique arguments 180
4.4.2 A delimiting requirement on secondary arguments 185
4.5 Summary 189
Chapter 5. Syntax of a Theory of Aspect 191
5.1 Introduction 191
5.2 Aspect and phrase structure 193
5.2.1 Semantic independence of tense/modality
and aspect 194
5.2.2 Syntactic independence of tense/modality
and aspect 202
5.2.3 Against a purely featural approach to aspect 206
5.2.4 Possible instantiations of aspect
in phrase structure 209
5.3 Aspectual particles, resultatives, and double
object constructions 215
Chapter 6. Aspect and the Mapping of Cognitive
Structure into Syntax 245
6.1 Introduction. The Aspectual Interface Hypothesis 245
6.2 Unaccusatives and unergative verbs 250
6.2.1 Introduction 250
6.2.2 Review of the syntactic approach to unaccusatives
and unergatives 252
6.2.3 Semantic differences between unaccusatives and
unergatives 259
6.2.4 The aspectual distinction between unaccusatives
and unergatives 267
6.3 The locative alternation 277
6.4 The body part/possessor alternation 284
6.5 Psych verbs 286
6.6 UTAH and thematic roles 294
6.7 Thematic hierarchies and the mapping of cognitive
structure into syntax 297
6.8 The AIH and learnability 310
BIBLIOGRAPHY 313
-8-
Chapter 1.Introduction
1.1 Introduction
This thesis is an investigation of aspect and its
interaction with syntax. More specifically it concerns one
aspectual property, delimitedness, and the role it plays in
syntax. Delimitedness -- the property of an event's being
bounded in time -- is a semantic property, but it is
implicated in a wide range of syntactic phenomena. Although
there are other aspectual properties of events that are
grammaticalized in many languages, delimitedness is the one
that has the most repercussions in syntax. The
delimited/non-delimited distinction seems to be a central
aspectual property of natural languages.
Aspect has long been recognized as a topic worthy of
interest to linguists outside the generative tradition, and
particularly among students of Slavic languages, where it is
heavily represented in the morphology. But aspect has not
been treated in generative theory in any great depth, because
its interaction with general principles of grammar has not
been understood. The nature of that interaction will be
elucidated in this thesis, and it will be proposed that
delimitedness interacts with NP-movement, with Case, and with
the syntactic organization of argument structure.
'.Affectedness', a property of arguments which has been
-9 -
discussed in the literature recently, has been linked with
many seemingly unrelated syntactic phenomena. The semantic
property of 'affectedness' plays a significant role in
predicting some kinds of syntactic behavior; but why
'affectedness' should do this, or how to understand
'affectedness' in syntactic terms, has not been clear. It
will be demonstrated in this thesis that 'affectedness' is
properly defined in terms of delimitedness. This will not
answer all questions about the syntactic effects of
'affectedness' but it will provide a framework for
understanding and investigating the phenomenon of affectedness
further.
In a deeper sense, this thesis is about the connection
between syntax and semantics. Aspectual delimitedness and
affectedness are two semantic properties that have syntactic
effects. In a theory of language espousing the autonomy of
syntax, these are interesting properties, particularly worthy
of study. This investigation will lead to a view in which
aspectual properties mediate between lexical semantics and
syntax in a highly constrained way. This view allows a strong
version of the autonomy of syntax to be maintained, and
uncovers the tools with which an explicit theory of thematic
roles could be developed. The interface between syntax and
lexical semantics or cognitive structure is to be found in the
correlation between certain aspectual properties and deep
structure argument positions. In particular it will be
proposed that the direct argument 'measures out' the event
described by the linguistic expression. Much progress has
been made in generative grammar by searching for the maximum
connection between syntax and semantics. I follow that
tradition here.
English is the principle language investigated in this
thesis, although reference will be made to relevant data in
other languages. The work is couched, for the most part, in
the syntactic framework of Government and Binding Theory, but
I have no doubt that many of these insights could be captured
in other frameworks. The semantic discussions will be cast in
informal terms. Some words with technical usages will be used
informally; these include 'event', 'measure', and 'meaning'.
The reader will be alerted to the usages of these words in the
course of the discussions. Also, the symbols A and E will be
used in place of the standard symbols for the universal (V)
and existential (a) .Aantifiers.
Chapter 2 presents some of the range of syntactic phenomena
in which delimitedness is implicated. Chapter 3 discusses
'affectedness'; argues that it is an aspectual property based
on delimitedness; and makes some proposals about how
affectedness is hooked up to syntax. Chapter 4 investigates
the aspectual properties of internal, external and oblique
arguments, and proposes some general principles of argument
structure. Chapter 5 digresses slightly in posing and
- 11 -
/considering two syntactic questions upon which aspect has some
bearing: the place of aspectual markers in X-bar theory, and
the deep-structure constituency of double object
constructions. Finally, Chapter 6 advances the idea that
aspectual properties of argument structure mediate the
connection between syntax and lexical semantics or cognitive
structure.
1.2 Aspect
'Aspect' usually refers to the organization with respect to
time, of an event represented by some linguistic expression
such as a verb, verb phrase, clause or sentence. This
includes such things as whether the event is understood to
involve change over time, whether it has a definite endpoint
or is ongoing in time, whether or not it is repetitive, and so
on. Many languages have morphological markers indicating
this kind of information. English has relatively little
aspectual morphology, but one example it provides is the
progressive form, marked by 'be -ing'. The progressive
characterizes an event as ongoing, even though the same verb
without the progressive marker may describe an event that has
an endpoint in time. For example, the sentence,
1. For general discussion of the kinds of aspectual phenomena
found across languages, see Comrie (1976) and Timberlake and
Chung (1985).
- 12 -
1
Patricia climbed a tree.
describes an event as having an endpoint. The sentence
describes the event as lasting a certain length of time. We
can tell this because it is possible to say:
2
*It took Patricia an hour to climb the tree.
However, when the verb is in the progressive form the event is
described as if it had no particular duration or endpoint:
3
Patricia was climbing the tree.
It took Patricia an hour to be climbing the tree.
Morphological markers of aspect such as 'be -ing' manipulate
the temporal organization of the event described by the verb.
Aspect in the example above is signalled by a morphological
marker on the verb, but this is not always the case. There is
a large body of aspectual information that is inherent in the
verb itself, or conveyed by the interaction of the verb with
its arguments. For example, the expression 'climbing a tree'
describes an action that has an endpoint built into it, but
'pushing a cart' does not. 'Winning the race' and 'growing
up' are both predicates that describe actions with an end, but
'winning the race' describes something that occurs in an
instant, whereas 'growing up' names an event that takes
years. This kind of information must be included in the
lexical entries of these verbs. Distinctions such as these
- 13 -
have been discussed in a long tradition of linguistic and
philosophical literature dating back to Aristotle, and will
figure greatly in this thesis. These phenomena are referred
to as Aktionsarten in much of the linguistic literature. Many
authors have classified verbs and/or predicates according to
whether they describe events that happen instantaneously or
over a period of time (the momentary/durative distinction);
whether they involve achievement of an end or not (the
telic/atelic distinction); and whether the predicates are
stative or nonstative. Some authors classify verbs alone into
aspectual classes, others classify entire verb phrases. The
aspectual property of delimitedness investigated in this work
will be shown to be compositional -- a property of verb
phrases or sentences rather than of lexical items. It is a
property determined by the interaction of a verb and its
internal arguments. Delimitedness is the property of a
linguistically described event which is bounded in time. The
distinction between delimitedness and non-delimitedness is
close to the telic/atelic distinotion.
Aspect is generally distinguished from tense in that tense
makes reference to a moment in time determined by the context
in which the expression is used -- the 'present', for example,
or the time at which the linguistic expression is uttered;
'aspect' does not refer to such contextual information which
locates the event in time, but to the internal time of the
event. Events as they are expressed linguistically have
- 14 -
temporal structure independent of reference -- this is what
'aspect' refers to. Tense, on the other hand, is indexical;
i.e., it is interpreted partly through the context in which it
is used.2 The English markers of present, past and future
mark tenses.3 If we say "Patricia is tall", the state of
Patricia's being tall is related to a particular point in time
-- the time at which the sentence is uttered. If we say
"Patricia will climb a tree", or "Patricia climbed a tree" we
express the event as taking place at some moment before or
after the present moment. The working definition I adopt of
the difference between tense and aspect is as follows: tense
must be interpreted partly through extragrammatical
2. There are three general approaches to a theory of tense in
the literature. Reichenbach (1947) established a method of
capturing the possible range of interpretations of various
tenses by setting up three reference points of time. These
are the speech time, (or time at which the sentence is
uttered); the event time, (or time the event actually takes
place); and the reference time (an internal time relating
speech and event time in some tenses). This system is widely
used and referred to by subsequent authors, among them, Smith
(1978) and Hornstein (1977), who extend and adapt the
Reichenbachian approach to a larger corpus of data. McCawley
(1971) treats tenses as underlyingly verbs, successively
embedded in a clause. Each embedding of a tense predicate
alters the sentence's reference to time. Finally, Partee
(1973) treats tense as a binding phenomenon. Partee notes
striking similarities between the behavior of tenses and
pronouns, and proposes that tenses are bound variables. All
of these theories address the relation of the time-reference
of events directly or indirectly to a timk line, or to other
events.
3. The English future is indicated by a modal rather than by a
tense marker, but for present purposes the English future acts
as a tense.
- 15 -
information; aspect is determined through information
contained within the linguistic expression.4
Although tense and aspect are interdependent in certain
ways, I will abstract away from this interdependence and
investigate aspect as independent of tense.
The term 'aspect' in its traditional usage refers to a broad
range of effects, generated by various parts of the language
including tense, morphological aspectual markers, adverbial
expressions, lexical meanings of predicates, and the syntactic
or semantic nature of their arguments. Aspect as such may not
be a unified phenomenon, and perhaps should not be treated as
one system. I do not intend to address in this thesis the
wide range of aspectual phenomena discussed in the literature
on aspect. Instead, I focus on one aspectual property, which
I label 'delimitedness', and which I demonstrate is of
fundamental importance in syntax and in cognitive or thematic
4. The.distinction between tense and aspect is not always
clearly maintained in the literature on the subject. The
English perfect, for example, is sometimes treated as a tense
and sometimes as aspect. It has often been classed as
aspectual because it seems to impose a completion on the event
described. (In the sentence, "Patricia had climbed a tree",
the tree-climbing is related to the present moment by
mediation through some unspecified moment in the past, before
which the event occured.) Reichenbach (1947) treats the
perfect forms as tenses and subsumes them elegantly in his
system of tense semantics. Syntactically, at least there is
reason to class the perfect forms with aspect. I return to
the English perfect briefly in Chapter 5, but apart from that,
the perfect forms will not be discussed in this thesis.
- 16 -
structure. I hold delimitedness to be one of the crucial
parameters of aspect, perhaps even the central parameter. A
general investigation of a wider range of aspectual phenomena
is beyond the scope of this work, but it is partly my thesis
that investigation of delimitedness will shed light on other
kinds of aspectual phenomena as well. Some predictions about
aspect in general, made on the basis of this investigation
into delimitedness, may have to await further research to be
fully tested.
1.3 Delimitedness
In light of the general discussion of aspect above, I turn
now to discuss the particular property of delimitedness which
is the topic of investigation of this thesis. Delimitedness
refers to the boundedness over time of an event as described
by a linguistic expression. A linguistically described event
is delimited if the sentence describes an event as something
that must transpire over a fixed length of time. It does not
matter whether that length of time is indicated in the
sentence. The sentence or event is delimited if it is
understood to mean that there is some point in time after
which the event is no longer continuing. I will illustrate
this in a general way through four classes of examples.5 For
5. It is my purpose to use delimitedness as a tool for
linguistic investigation. To this end it suffices if I
demonstrate that the speaker's intuitions, supplemented or
- 17 -
example, the sentence below describes an event of some
duration having a definite temporal endpoint. (These are
referred to as accomplishments.)
4
Kim will climb the silo.
The event described, Kim's climbing of the silo, is one that
must come to an end when he reaches the top of the silo. That
may take three minutes or three days, but if he climbs the
silo it will take a finite definite length of time.6  Now
compare 4 above with a sentence that describes an event of
indefinite duration. (These are referred to as activities.)
confirmed by certain independent tests, are adequate to
determining whether a linguistic expression describes a
delimited or non-delimited event. It is not my goal to
provide a formal treatment of delimitedness, or define it
within any formal system.
6. Delimitedness may seem to depend, in 4, on world knowledge
as well as on purely linguistic information. The
delimitedness of the climbing depends on the silo having
finite length. A certain class of verbs, to which 'climb'
belongs, translates spatial delimitedness into temporal
delimitedness. Nevertheless, the delimited interpretation of
the sentence would seem to depend on grammatical factors,
rather than on contextual information. First of all, the fact
that this verb can do that is linguistic information,
contained in the lexical entry for the verb. Secondly, in
languages where delimitedness is marked morphologically the
delimited reading would necessitate a finite silo, of which
the entirety is climbed once. Although delimitedness is not
morphologically marked in English, the delimited reading is
available as a grammatical option, even when not marked by
morphology. I will leave further investigation of this
problem for future research.
- 18 -
5
Kim will sleep in the silo.
The event described, Kim's sleeping, will not have a definite
duration; in fact, it need not even be finite. Kim may sleep
three minutes or three days or forever -- the sentence
expresses no restrictions on how long Kim sleeps.
Certain sentences expressing events that have little or no
duration in time, may also be delimited. (These are known as
achievements.) The following sentence is an example:
6
Kim will spot that sheep by the river.
6 describes a short-lived event. But even if it only takes a
moment to spot the sheep, there will be some definite moment
in the course of such an event, when sheep-spotting has come
to an end.
Sentences constructed with stative predicates describe
situations that are not delimited:7
7
Kim likes sheep.
Sentences like 1 and 3 which describe events as having a
7. Statives, strictly speaking, are better characterized as
describing situations that are independent of temporal
duration. We return to this in Chapter 4. For present
purposes statives can be considered to express events having
indefinite temporal duration.
- 19 -
finite or definite duration over time I will refer to as
describing delimited events. Sentences such as 2 and 4 which
describe events without a definite temporal endpoint describe
non-delimited events. These four classes of linguistic
expressions or linguistically described events are summarized
in the chart below. The classes were devised by Vendler
(1967).
Accomplishment definite duration delimited
of some length
Achievement [taving a definite delimited
endpoint, but of very
brief duration
or no duration
Activity indefinite duration non-delimited
Stative indefinite duration non-delimited
Two points of clarification are in order. It must be made
clear that though I speak of events being delimited, it is not
actually the event, but the linguistic representation of the
event that is delimited. I do not intend to make any claims
about the nature of events in the world, only about the
representation of events in natural language. It is very
possible that the linguistic expression of events is related
to the cognitive perceptions of events by human beingp, but I
make no claims about this either. Referring to the delimited
or non-delimitedness of events is simply a convenient
shorthand I will employ for referring to the delimitedness of
events as they are structured, represented or expressed by
- 20 -
natural language. I am making a three-way distinction between
(i) the linguistic expression, (ii) the event as represented
by that linguistic expression, and (iii) some event in the
world. I do not intend to make any claims about the (iii), or
discuss the relation between (ii) and (iii). Secondly, for
brevity and convenience, the term 'delimited' may be applied
to sentences or verb phrases that describe an event as bounded
in time, even though it is the linguistically expressed event
itself that is delimited, rather than the linguistic
expression.
Delimited and non-delimited sentences can be distinguished
by certain tests; in particular, by the application of
durative adverbials. Non-delimited activity and stative
sentences are completely acceptable with the durative
adverbial, under the interpretation that the event happened
only once during the time span expressed by the adverbial.
(This will be referred to as the semelfactive reading.)
8
Kim slept for three hours.
9
Kim liked sheep for three years. (Then he got bitten by one.)
These sentences describe a single event of sleeping or
sheep-liking. They do not describe a succession of such
8. Tests for aspectual classes using adverbial expressions are
discussed by Dowty (1979) and Vendler (1967), among others.
- 21 -
events. When durative adverbials are applied to a sentences
describing delimited events, the sentence either must be, or
may be, interpreted to mean the event happened again and again
during the span of time indicated by the adverbial. (This
will be referred to as the iterative reading.)9
10
Kim spotted that sheep for three hours.
11
Kim climbed the silo for three hours.
The sentences above describe delimited events. 10 cannot mean
that Kim spotted the sheep only once over the course of the
three hours. It must be interpreted to mean that the sheep
was spotted many times, or else it makes no sense. Sentence
11 may be understood to mean that Kim climbed the silo again
and again over three hours' time.10
9. If the sentence is understood to express an event happening
over and over an indefinite number of times, the sentence
expresses something with an indefinite duration in time. The
property of delimitedness as defined here is based on the
semelfactive reading and not on the iterative reading.
Sentences 10 and 11 are delimited sentences, even when they
are given an iterative reading.
10. Some speakers find that some sentences describing events
having duration as well as delimitedness, may have another
reading besides the iterative one: i.e., 11 might be
understood to mean that it took Kim three hours to climb the
silo once. When this reading is available, the sentence is
not a true accomplishment, but has more of an activity sense
to it. An appropriate paraphrase would be, 'for three hours
Kim was engaged in the activity of climbing the silo'.
Another example suggested to me by Steven Abney is a sentence
used by a mother to address her child on a playground,
climbing up and down on the play equipment: "You've climbed
- 22 -
When an adverbial expression like 'in an hour' is applied to
a non-delimited expression, the sentence is odd:
12
?Kim slept in an hour.
13
?Kim liked sheep in three years.
However, an 'in' adverbial applied to a delimited expression
is perfectly natural:
14
Kim climbed the silo in three hours.
15
Kim spotted that sheep in three hours.
In short, certain durative adverbial expressions are
diagnostic of delimited or non-delimited events: 'for'
adverbials with a semelfactive reading indicate non-delimited
events; 'for' adverbials with an iterative reading indicate
delimited events; and 'in' adverbials indicate delimited
events.
To summarize,
16
An event is delimited iff there is some point of time
during which the event transpires, but after
Mt. Everest long enough. Let's go home." The conditions
under which a semelfactive reading of a delimited expression
is possible are not yet clear to me. For present purposes
this does not invalidate the test, since it depends on the
possibility rather than the necessity of an iterative
interpretation.
- 23 -
which the event is no longer transpiring. If the state p is a
state in which the event is transpiring, and the state -p is
a state in which the event is not transpiring, then for a
delimited event there is some point of time when p becomes -p.1 1
Choosing the property of delimitedness to investigate, out
of the wide range of phenomena included in the term 'aspect',
calls for some justification. First of all, delimitedness has
been recognized in the literature as a central property of
aspect. What I have called the delimited/non-delimited
distinction is comparable to what has been called in the
literature the event/process, non-durational/durational,
telic/atelic, and definite/indefinite-change-of-state
aspectual distinctions described in the literature. (For a
more extensive discussion of this point and of the literature
on the subject, see the Appendix to this chapter.)
Secondly, delimitedness has been recognized as one of two
salient properties whose definition is necessary for a
discussion of lexical aspect. The other property -- the
durational vs. instantaneous quality of a linguistically
described event -- is arguably of less grammatical importance
than the delimited/non-delimited distinction.
The difference between accomplishments and achievements is
11. The term 'event' is used loosely and intuitively here to
indicate a situation or happening described by a linguistic
expression.
- 24 -
.one of duration. Accomplishment events have duration;
achievement events last only for a short interval or a moment
(if it is appropriate to describe them as 'lasting' at all).
However, many achievement verbs are easily reanalysed as
accomplishment verbs if one has the imagination to construct a
situation in which the achievement would take time. Dowty
(1979) pointed this out with the achievement verb, 'die'.
Dying is something that generally takes only a moment, but one
may occasionally say such things as, "the old man finished
dying", (Dowty's example) which indicates that the dying had
some duration. 12 13 In other words, imagination and
pragmatic knowledge can affect whether a verb is interpretable
as an achievement or an accomplishment verb. We would not
expect a purely grammatical distinction to be accessible to
such extra-grammatical factors as these.
The delimited/non-delimited distinction is quite different.
12. The conditions on the use of the English progressive are
more complicated than implied here. These include other
requirements on the event described besides duration. In this
thesis I do not discuss these conditions in any detail, and I
do not use the English progressive as a final diagnostic for
any aspectual property.
13. M. Halle has pointed out the following Russian proverb:
ne trudno umeret'
it is not difficult to die-PERF
no trudno umirat'
it is difficult to die-IMPERF
"It is not difficult to die, but to be dying."
There 'are verbs that can belong to both the class of verbs
that are associated with delimited events, and the class of
verbs that are associated with non-delimited events. In the
author's dialect 'dry', as in 'dry the clothes', is such a
verb. Drying the clothes can mean something one effects
(delimited or accomplishment reading), or it can mean
something one does, that takes an indefinite length of time
(non-delimited or activity reading). The verb 'die' has a
salient reading as an achievement verb, but the application of
some imagination makes the accomplishment 'meaning' available,
in English as well as in Russian. Pragmatic knowledge about
dying contributes to the interpretation of the verb as an
achievement. In the case of 'dry' no pragmatic information
biases one towards a delimited or non-delimited reading of the
verb, nor does imagination suffice to make us choose the
non-salient reading. The delimited/non-delimited distinction
is freer of pragmatic influence than the durative/non-durative
distinction.
Finally, the durative/instantaneous distinction does not
correlate with the wide range of grammatical/syntactic
phenomena that delimitedness does. Nor, to the best of my
knowledge, is the durative/non-durative distinction
grammaticalized in the morphology of any language, in such a
way that there is a particular morpheme that converts
achievement verbs into accomplishment verbs or vice versa.
Such morphemes do exist in many languages to mark the
- 26 -
delimited/non-delimited distinction. The wide range of
grammatical and syntactic phenomena that must be stated in
terms of delimitedness will be demonstrated in Chapter 2.
APPENDIX: Overview of literature on aspectual verb classes
This appendix is a brief survey of literature on aspectual
verb classes. Although a variety of approaches to a taxonomy
of lexical aspect have been proposed, they all pick out the
same or nearly the same properties as primitives. Something
like delimitedness is an important feature in all these
systems.
The earliest taxonomy of verb aspect cited in the literature
is Aristotle's. Aristotle discussed the differences between
logical entailments of various verbs with regard to tenses.
He was the first to point out that verb classes can be
distinguished depending on whether the sentence 'we are
VERBing' implies 'we have VERBed'. Aristotle discusses a
distinction in logical entailments between language describing
events that come to an end and language describing events that
have no definite endpoint: (Aristotle, Metaphysics 1048b,
trans. H. Tredennick)
...at the same time we see and have seen, understand and have
understood, think and have thought; but we cannot at the same
time learn and have learnt, or become healthy and be healthy.
We are living well and have lived well, we are happy and
have been happy, at the same time; otherwise the process
would have had to cease at some time...; but it has not
ceased at the present moment: we both are living and have
lived.
The philosophical literature has taken up this problem more
recently. Kenny (1963) developed a trichotomy of verb types
based on ideas of Aristotle's in the De Anima, Metaphysics and
- 28 -
Nicomachean Ethics, distinguished by the entailments of verbs
in present and past tenses. Ryle (1949) distinguished verbs
describing events which happen at a particular instant and end
in a particular result, (which he called achievement verbs)
from those which do not (task verbs). His distinction opposes
achievement/task pairs like 'cure' (which describes an event
having a temporal endpoint) and 'treat' (which describes an
event ongoing in time).
By and large the recent generative linguistic literature
looks to Zeno Vendler for the first discussion of this
problem. Vendler (1967) divided verb phrases into four
classes according to the type-of events they describe:
activities, accomplishments, achievements and states.
Activities are ongoing events like 'running', 'swimming' or
'pushing', which call for an indefinite period of time, There
is no terminal point when the activity must come to an end.
One may ask about the duration of an activity: "For how long
did she push the cart?" Accomplishments have a definite
terminus, as in 'drawing a circle' or 'making a chair'. They
transpire over a unique, definite period of time, and must be
questioned accordingly: "How long did it take to draw. the
circle?" Achievements, like 'recognizing his face' or
-reaching the summit', have a terminus but happen
instantaneously and so are to be distinguished from
accomplishments, which have duration. One must ask about an
achievement, "At what time did you reach the summit?"
- 29 -
NFinally, statives such as 'knowing the answer' are ongoing in
time. The progressive form is a criterion used in Vendler's
classification. It distinguishes activities and
accomplishments from states and achievements:
16
I am running. (activity)
I am making a chair. (accomplishment)
*I am recognizing his face. (achievement)
*I am knowing the answer. (stative)
Vendler set up this four way classification scheme using the
presence or absence of an event terminus, the momentary or
durative quality of the event, and the applicability of the
progressive form.
Subsequent aspectual classification systems in the
linguistic literature are based on Vendler's work. Various
authors have adapted or redefined Vendler's scheme according
to their own views. They have generally selected the presence
or absence of an event terminus -- the telic/atelic
distinction -- as the important parameter of aspect.
Mourelatos (1981) set up a trichotomy in which accomplishments
and achievements are grouped together as 'events' in
opposition to activities, which he calls 'processes'. States
are distinguished from all of these. Verkuyl (1972) mentioned
the tripartition into activities, accomplishments and
achievements (which he called the durative, terminative and
momentaneous aspects, respectively) but he focused his
- 30-
research on the opposition between non-durative and durative
aspect -- again the telic/atelic distinction. Hinrichs (1985)
is also based on an analysis of the telic/atelic distinction.
Dowty (1979), in a thorough treatment of aspectual classes
of English verbs, worked out an aspect calculus involving
partial lexical decomposition of verbs. It provides a system
of representation for the Vendler classes. This was part of a
larger program to illustrate the value of lexical
decomposition, based on the generative semantics tradition,
but couched in a Montague-type model-theoretic semantics. In
Dowty's system, stative predicates are the primitives out of
which other predicates are formed through combination with
primitive logical operators of a natural logic. Achievement
verbs are represented by the logical operator BECOME applied
to a stative predicate. BECOME(p) is defined (loosely) as:
"first not p and then p" or: "the state p comes about". This
captures the intuition that achievement verbs describe an
instantaneous change of state. The achievement verb
"discover" in the sentence "John discovered the buried
treasure in his back yard", is roughly:
17
BECOME[John knows that...]
Accomplishments contain the logical connective (p)CAUSE(q)
joining two embedded sentences. (For arguments motivating the
bisentential analysis see Dowty (1979).) The accomplishment
- 31 -
sentence "John painted a picture" has a structure something
like:
18
[[John paints] CAUSE [BECOME[a picture exists]]]
Activity predicates contain an operator DO applied to a
stative predicate. This is intuitively satisfying since
agentivity or volition often distinguishes many activity verbs
from statives. (e.g. Compare stative and active forms of
"taste" as in "This orange tastes good" and "John tastes the
orange".) The verb "look" in "Mary looks" would be roughly:
19
DO (mary [see...]
(Dowty acknowledges, however, that the operator DO is not as
well-motivated as the operators BECOME and CAUSE.) Not every
stative predicate appearing in these semantic representations
need be lexicalized in the language. The logical operators
BECOME, CAUSE and DO may appear in the semantic
representations of other verb classes besides the ones they
are primarily associated with, but in general they take wide
scope in their respective classes.
Dowty later introduces an interval semantics, in which the
primitive unit of time is taken to be the interval rather than
the moment. Redefining BECOME(p) with truth conditions using
time intervals, Dowty united achievement and accomplishment
- 32 -
verbs by ascribing to them both the operator BECOME. He
refined Vendler's classification by setting up five
distinctions determined by syntactic tests. These yield the
classification system reproduced in the table below (from
Dowty (1979) p. 184).
Non-agentive Agentive
States Ia. be asleep, be in 2a. be polite, be a
the garden hero (these
love, know belong possibly
here or in 4)
lb. interval statives: 2b. interval statives
sit, stand, lie sit, stand lie(with human subject)
Activities 3. make noise, roll 4. walk, laugh
rain dance (cf.2a)
Single change 5. notice, realize; 6. kill, point out
of state ignite (something to
sor4eone)
Complex change 7. flow from x to y 8. build (a house)
of state dissolve walk from x to y,
walk a mile
The distinctions and syntactic tests used for determining
these classes are quoted below from Dowty (1979):
I. Momentary (la and 'habituals' in all classes) vs.
iervalpredicates (1b, 2b, 3-8). Syntactic test:
abilityto appear in the progressive. (Note: 6 and
especially 5 appear less readily in the progressive than
other interval predicates.)
II. Predicates entailing definite or indefinite change
(3-8) vs. those entailing no chane (1 and 2).
- 33 -
Syntactic test: ability to occur in do constructions
(pseudo-clefts, do so reduction, etc7
III. Definite change of state predicates (5-8) vs.
activit predicates or indefinite change of state
predcates (3 and 4). Syntactic test: Does x was
V-ing (pragmatically) entail x has V-ed?
IV. Singulary change predicates (5-6) vs. complex
chanje predicates (7-8). Syntactic test: Is x
tiniAhed V-ing" acceptable?
V. Agentive (2,4,6,8) vs non-agentive (1,3,5,7)
predicates. Syntactivc test: ability to occur in agentive
contexts like imperatives, persuade x to do V do V
deliberately, etc.
Note that in III above, accomplishment and achievement
predicates are united as "definite change of state predicates"
while activities are "indefinite change of state predicates".
This roughly corresponds to what Mourelatos called the
'event/process' distinction, Verkuyl called the
'non-durational/durational' distinction, and traditional
literature refers to as the 'telic/atelic' distinction.
Dowty's definite change of state predicates are divided into
four classes as opposed to Vendler's two classes of
accomplishment and achievement verbs. However, the
distinction between singulary change and complex change
predicates has smaller scope than the distinction between
definite and indefinite change predicates. Like Mourelatos,
Verkuyl and Hinrichs, Dowty selected this parameter as a
crucial parameter of aspect.
In acknowledging the work of previous authors I have
introduced much confusing terminology. I have bypassed this
- 34 -
confusion by abandoning all the terms used to describe these
phenomena, together with the question of whether these various
authors meant exactly or only approximately the same thing.
The aspectual distinction investigated in this work is
non-delimitedness vs. delimitedness. It is roughly
comparable to the event/process, non-durational/durational,
telic/atelic, and definite/indefinite-change-of-state
aspectual distinctions described in the literature.
-'35n~
Chapter 2. The Grammatical Reality of Affectedness and
Delimitedness
Delimitedness has been defined in Chapter 1 as a property of
events which have a definite duration over time. It is an
aspectual or semantic property rather than a syntactic one.
Nevertheless cer Jain syntactic behavior is determined by this
property. The same thing is true of affectedess, another
semantic property related to delimitedness, which has
substantial repercussions in syntax. These are two semantic
properties that must be defined in the grammars of natural
languages. They have grammatical reality because there are
syntactic phenomena for which-the correct generalizations
cannot be stated without reference to delimitedness. A few of
these are discussed in the following sections.
2.1 English verb-particle combinations
English has a large number of partly lexicalized
verb-preposition combinations. 'Think through',, 'shut up',
'look over', 'throw out', and 'rely on' are a few examples. A
sub-class of these constructions are known as verb particle
constructions. These have been discussed by Bolinger (1971)
and Fraser (1976) among others. These particles exhibit the
syntactic trait of being able to appear on either side of the
object noun in transitive expressions. Bolinger (1971) used
this characteristic to distinguish them from what he called
pure prepositions, which may only appear before the noun, and
- 36 -
pure adverbials, which only appear after it. These three
classes are illustrated below in la, 1b, and ic respectively.
(Examples are from Bolinger (1971). He refers to these
particles as adverbial particles.)
I
a.
run up the flag
run the flag up
He looked up his friends.
He looked his friends up.
They bought out their competitors.
They bought their competitors out.
b.
climb up the tree
*climb the tree up
I can cope with Jones
*I can cope Jones with.
c.i
*I left there the keys.
I left the keys there.
I saw John yesterday.
*I saw yesterday John.
*He sold regretfully the business.
He sold the business regretfully.
With pronominal objects, the particles appear in the
postnominal position, which makes the difference between
particles and pure prepositions more pronounced:
1. Heavy NP Shift may apply in the adverbial cases. Bolinger
presents the example, 'I would sell regretfully any business
in which I had been engaged for half a lifetime.'
- 37 -
I
2a.
*run up it
run it up
*He looked up them.
He looked them up.
*They bought out them.
They bought them out.
b.
climb up it
*climb it up
I can cope with him
*I can cope him with.
The ordering of particles and pronominal objects is readily
explained by a condition that requires the item with the
heaviest "weight' to be clause-final where possible. The
heaviest 'weight' is associated with the most or the newest
information, or with the longest utterance. When the
pronominal is stressed it can be ordered postnominally:
3
You looked up THEM?! Those were the wrong ones to look up!
In this way, the ordering of particles with respect to
pronominal objects is related to Heavy NP Shift, which is
known to be a stylistic rule. The syntactic generalization
that the particle may appear on either side of the object
holds true, where the effect of stylistic or discourse
conditions is controlled for.
The degree of idiomaticization of a verb-particle
-38-
combinatioi. influences the separability of the verb and the
particle. Highly idiomaticized combinations are less likely
to be separable from the verb: 4a may be interpreted in the
literal or the figurative se.nse, but 4b may only be
interpreted in the literal sense.
4
a. Don't throw up your lunch.
b. *Don't throw your lunch up.
The data above show that there is a class of particles which
may occur on either side of a direct object noun phrase, and
which preserve a somewhat literal sense. These particles also
have a particular semantic property; they impart a resultative
sense to the sentence or verb phrase. In Bolinger's words,
"the particle must contain two features, one of
motion-through-location, the other of terminus or result."2
Bolinger notes that manner and time adverbials are excluded by
this account, as are place and stance adverbials (which may
contain result without motion), and directional adverbials
(motion without result). These are illustrated in Sa, 5b and
5c respectively:
5
a.
*He built well the fire.
*She stitched carefully the rip.
*I bought yesterday the stocks.
2. Bolinger (1971) p. 85
- 39 -
b.
*We brought here the bags.
*I left home the money.
*She crooked akimbo her arms.
c.
*He tossed upward the ball.
*They pulled downward the blinds.
An event that achieves a result is an event with a temporal
endpoint -- a delimited event.3 The class of verb particles
have the semantic property of imposing delimitedness on the
event described by a verb phrase or sentence, as well as the
syntactic property of appearing on either side of the noun
phrase object:
6
look up a name in the phonebook
look a name up in the phonebook
look over an article
look an article over
think up an answer
think an answer up
think through a problem
think a problem through
sit out a game
sit a game out
eat up an apple
eat an apple up
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
(delimited)
3. Not all expressions that have to do with the results of an
event are delimiting; e.g. purpose clauses. The sentence
'John sang in order to surprise Bill' describes a
non-delimited event, even though it contains a purpose
clause. Purpose clauses do not represent a result as
achieved; hence their inability to delimit.
- 40
If I 'look over an article' or 'think up a result' my action
of looking or thinking has a clear result or termination. The
article has been perused; there is an answer that did not
exist before. Either the object itself has been changed or
the activity has progressed through the 'extent' of the object
itself during the course of the event.
Prepositions which do not require a delimited interpretation
do not demonstrate the syntactic behavior of particles:
7
look at a photograph
*look a photograph at
look out a window
*look a window out
think about a problem
*think a problem about
sit on an idea
*sit an idea on
sit on a chair
*sit a chair on
eat at an apple
*eat an apple at
put on an act
*put an act on
non-delimited
non-delimited
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
non-delimited
non-delimited
non-delimited
non-delimited
Prepositions introducing goal phrases generally require a
delimited interpretation, but they do not express a result in
the same way that particles do, and they do not exhibit the
syntactic behavior of particles:.
-41-
8
fly to the moon (delimited)
walk to the cliff (delimited)
Although the events expressed by these verb phrases are
delimited events, having a distinct temporal endpoint, the
endpoints are not reached by either changing or 'traveling
through' the objects, moon and cliff. After looking up a
name, the name is in some sense 'up', (facetiously speaking)
but the moon is not 'to'. Particles have a special semantic
relationship with the direct object, not shared by goal phrase
prepositions such as 'to'.
Particles impart a sense of result. They force a delimited
interpretation of the event described, and they may appear on
either side of the object NP. Particles constitute an
aspectual class of words with a particular syntactic
behavior. Whatever the proper analysis is of the syntax of
these particles, some reference must be made to their
aspectual nature, or an important generalization has been
missed.
2.2 English resultative secondary predicates
Certain syntactic differences between resultative and
depictive secondary predicates have been noted in the
literature. Authors who have written on this subject include
Halliday (1967), Rothstein (1979), Simpson (1983),
Carrier-Duncan and Randall (1987), and Hale and Keyser
(1987). The observations in this section are drawn from their
- 42 -
work.
Resultative predicates are similar to particles except that
they may be adjectives or prepositional phrases. Resultatives
are underlined in the sentences below. (Examples 9 through 12
are from Hale and Keyser (1987).)
9
I cut the bread into thin slices.
I crushed the limestone to pieces.
I ground the corn fine.
The resultative prepositional phrase or adjective describes
the effect on the object, of the action described by the
verb. Depictives, on the other hand, describe the subject or
object independently of the effect of the verb action.
Depictives are underlined in the following examples:
10
This horse can't run a furlong tired.
I cut the bread hot.
Resultative secondary predicates are delimiting expressions,
and depictive secondary predicates are not. They represent
two distinct aspectual classes of expressions. Correlated
with these aspectual classes are four syntactic differences.
(i) Depictives may be predicated of subjects or objects, as
illustrated in 10 above, but resultatives may be predicated
only of objects, In 11 below, the resultative adjective makes
the sentence ungrammatical because the adjectives cannot be
- 43 -
associated with the subjects:
11
*My brother ran ragged.
(cf. I ran my brother ragged.)
*He talked into a corner.
(cf. He talked himself into a corner.)
(ii) Resultatives may occur in the middle construction
before the adverbial 'easily', but depictives are
ungrammatical or very poor in the same construction.
12
This bread cuts into thin slices easily.
??This bread cuts hot easily.
Rothstein (1979) observes two more syntactic differences
between resultatives and depictives. (Examples 13 and 14 are
from Rothstein (1979).)
(iii) There may only be one resultative per sentence, but
several depictives are possible:
13
a. *John washed the clothes clean white.
b. They eat meat raw, tender. (Rothstein's source: Simpson 1982)
(iv) When resultatives and depictives occur together, the
resultative must come first:
14
a. We hammered the metal flat hot.
b. *We hammered the metal hot flat.
- 44 -
This last property holds as well for particles. The
delimiting particle 'up' must precede the non-delimiting
depictive 'hot' in the examples below:
15
a. We ate up the muffins hot.
b. We ate the muffins up hot.
c. *We ate hot up the muffins.
d. *We ate the muffins hot up.
e. *We ate hot the muffins up.
f. *We ate up hot the muffins.
Example 15f shows that the condition on the relative order of
particles and depictives is not simply that particles must
precede depictives, but that depictives must follow both the
particle and the object.
Rothstein (1979) also notes interesting semantic differences
between resultatives and depictives. Resultatives are
selected by the verb. They may be nouns, adjectives, or
prepositional phrases. Resultatives are underlined in the
examples below. (Examples 16 and 17 are from Rothstein
(1979).)
16
a. We elected John president.
b. He painted the car a brilliant red.
c. He drank himself into a stupor.
Depictives, on the other hand, are not selected by the verb.
Furthermore they have particular and subtle semantic
constraints not shared by the resultative. In Rothstein's
words, "the attribute described by the predicate must be at
- 45 -
the same time an intrinsic property of the subject, and a
transitory one".4
17
a. John ate the peanuts salted/*salty.
b. John ate the meat raw/burnt/*tasty.
c. I met Mary drunk/in high spirits/*tall/*stupid.
d. We eat carrots raw/*orange.
The sentences in 17 are grammatical with the adjectives that
describe intrinsic and transitory properties. 'Salted',
describes a property of peanuts that is independent of John or
any other peanut-taster, but 'salty' describes a property that
is in the tongue, so to speak; of the taster, rather than
intrinsic to the peanuts. 'Salted' also is a temporary
property of peanuts, since they do not grow that way. 'Drunk'
is a transient property of Mary (giving Mary the benefit of
the doubt), whereas 'tall' or 'stupid' is not likely to be
SO.5
Note also that depictives may not be nouns:
4. Rothstein (1979) p. 153.
5. Hale (p.c.) and Rapaport (p.c.) note that good depictives
describe a transitory 'stage'.. No adjective is inherently
blocked from being a depictive secondary predicate. It is
simply required to be interpreted as a stage'. "John ate the
peanuts salty" is acceptable if "salty" is understood as a
stage'. Also see Rapoport (1987).
- 46 -
18
*John ate the meat steak.
I will return to a discussion of the semantic and syntactic
properties distinguishing resultative and depictive secondary
predicates in Chapter 5. The point to be noted here is that
they are delimiting and non-delimiting expressions
respectively, and that these aspectual differences are
correlated with different kinds of syntactic behavior. A
correct account of the syntax of these expressions must make
reference to the aspectual property of delimitedness.
2.3 Finnish case
In Finnish, delimitednefs is expressed. by case. A verb
phrase or a sentence with an object noun in accusative case
describes a delimited event. If the object is in partitive
case the event is non-delimited. Hein'Amlki (1984) shows that
accusative case is used when the event has some temporal limit
that must be achieved. (Examples 18 through 23 are from
HeingmAi (1984).)
19
a. Maria kantoi kirjaa
M. carried book-PART
"Maria was carrying a book."
b. Maria kantoi kirjan
M. carried book-ACC
"Maria carried the book."
Only 19b, where the object 'book' is marked with accusative
case, carries the inference that Maria carried the book to
- 47 -
some destination.
The temporal bound of the event, required by the accusative
case, may also be provided by a change of state:
20
a. Metsastaja' ampui vahingossa lehm'n
hunter shot accident-in cow-ACC
"The hunter shot a cow by accident."
b. Metsastoja ampui vahingossa lehmla"
hunter shot accident-in cow-PART
"The hunter shot (at) a cow by accident."
Example 20a is generally understood to mean that the cow was
shot dead, whereas 20b does not carry that implication. The
death of the cow provides an endpoint to the event.
The temporal bound of the event may be inferred or stated:
21
a. Maija luki kirjan
M. read book-ACC
"Maija read (all) the book."
b. Maija luki kirjan loppuun
M. read book-ACC end-to
"Maija read the book to the end."
c. Maija luki kirjan puoliv'liin
M. read book-ACC halfway through
"Maija read the book halfway through."-
d. Maija luki kirjan hajalle
M. read book-ACC to pieces
"Maija read the book to pieces."
In 21a, where no event terminus is stated, the object 'book'
itself provides the terminus for the event because the event
is understood to continue until all the book is read. In 21b
- 48 -
through 21d a bound for the event is specified, and the
sentence is understood accordingly.
If a bound for the event is not easily inferrable through
the verb and no bound is explicity stated, the sentence is
odd. Examples 22a and 22c with the object in partitive case
have no stated bound, and one is not easily inferrable from
the verbs 'heitti' and 'nosti'. If the explicitly stated
bound were omitted in the examples 22b and 22d, which have the
objects in accusative case, the sentences would be odd.
22 as
a. Tiina heitti keihasta
T. threw javelin-PART
"Tiina threw the javelin."
b. Tiina heitti keik'i'n mets"AZn
T. threw javelin-ACC forest-into
"Tiina threw the javelin into the forest."
c. Yrj5 nosti hattua
Y. lifted his hat-PART
"YrjB raised his hat (when greeting)."
d. Yrj5 nosti hattun hyllylle
Y. lifted his hat-ACC
"YrjB raised his hat onto the shelf."
Some verbs which usually take partitive objects can take
accusative objects if an event terminus is provided as in 23b
and 23d below:
23
a. Manne kehui hevosta
M. praised horse-PART
"Manne was praising the horse."
b. Manne kehui hevosen maasta taivaaseen
M. praised -horse-ACC earth-from heaven-to
- 49 -
"Manne praised the horse from earth to heaven."
c. MinS pelkan sotaa
M. fear war-PART
"I am afraid of war."
d. Pelkdsin iteeni puolikuoliaaksi
I-feared myself-ACC half-dead-to
"I scared myself half dead."
Recall that in English the application of durative
adverbials distinguishes between delimited and non-delimited
expressions. Delimited expressions in English are
ungrammatical with durative adverbials, in a semelfactive
reading. Durative adverbials in Finnish also distinguish
between delimited and non-delimited expressions. Durative
'for' adverbials are grammatical with non-delimited
expressions and ungrammatical with delimited ones.
24
a. Maria kantoi kirjaa tunnin
M. carried book-PART hour-ACC
"Maria carried a book for an hour."
b. Maria kantoi kirjan *tunnin
M. carried book-ACC hour-ACC
"Maria carried a book (to some place) *for an hour."
The examples above illustrate that the distinction between
delimitedness and non-delimitedness is grammaticalized in the
accusative and partitive cases in Finnish.6
6. According to Heinimaki, there are some Finnish verbs which
normally take accusative objects, but which may on occasion
take partitive objects. These are usually verbs of cognition
and perception. Heinmflki suggests that the reason for their
strong accusative orientation is that they are frequently used
- 50-
2.4 Morphological aspect
Certain generalizations about syntactic behavior must make
reference to delimitedness in order to be correctly stated.
This is true in a language like English, where delimitedness
is not overtly marked, and in a language like Finnish, where
it is marked in the case system. In some languages,
delimitedness is marked directly in the verbal morphology.
The fact that delimitedness may be grammaticalized
demonstrates its importance in natural language.
Linguists recognize several kinds of aspectual distinctions
7
marked in the morphology of natural languages. Among these
are the perfective, the progressive, and the perfect. Of
these it is the perfective that is most relevant to
as achievement verbs. In any case, it is not clear what the
semantic entailments of the partitive case are for these
verbs. Sometimes it imparts a partial quality to the event:
(i) Tunnen (hiukan) sit miesta (Leino, 1982)
I know little that man-PART
"I know that man (a little bit)."
but this may also be found with accusative case:
(ii) Tunnen hnet hyvin epataydellisesti
I know her-ACC very incompletely
"I know that man (a little bit)."
7. See Comrie (1976), Timberlake and Chung (1985).
- 51-
delimitedness.8 The perfective/imperfective opposition is
generally described as distinguishing between completed
actions and actions in progress, a distinction which for
practical purposes amounts to a delimIted/non-delimited
distinction. However, complicating factors enter into the
semantic analysis of perfective and imperfective markers for
two reasons; (i) there is variability in exactly what is
indicated by perfective and imperfective markers
cross-linguistically; and (ii) they may interact with tense,
modality, discourse, and adverbial expressions. I will
discuss each of these in order.
There is much freedom acrosS languages as to how the
perfective and imperfective markers register definite or
indefinite temporal duration of events. Chamorro and Russian
represent two different strategies that languages may employ
for classifying events aspectually. In Russian the perfective
marker is used when the event has both definite duration (is
telic) and is closed with respect to the time frame it is
viewed in. That is, a time frame is provided (possibly
through discourse factors) against which the event is
measured, and if the event has an endpoint within that time
frame it is closed with respect to that frame. If it does
8. The progressive and perfect are not irrelevant for a
discussion of delimitedness, but their connection with
delimitedness is less straightforward than that of the
perfective. The progressive will be discussed in Chapter 5.
-52-
not, it is open. In 25 below the 'when' clauee defines the
time frame for the event. In 25a the temporal duration of the
event indicated by the main clause event is not included
within that of the 'when' clause, and the imperfective marker
applies to the main verb. In 25b, the situation is reversed,
the temporal duration of the main clause is included within
that of the "when" clause, and the perfective marker applies.
(Examples in 25 and 26 are from Timberlake and Chung (1985).)
25
a. Kogda tanki dosti li kanala, fasisty kak raz
when tanks reach PERF) canal fascists just then
vzryvali most
blow up(IMPERF) bridge
"When the tanks reached the canal, the fascists just then
were blowing up the bridge."
b. Kogda tanki dostigli kanala, fasisty vzorvali most
when tanks reach(PERF) canal fascists blow up(PERF) bridge
"When the tanks reached the ezrnal, the fascists blew up
the bridge."
In Chamorro (a Western Austronesian language) there is an
aspectual marker -- the neutral or non-progressive marker --
which- applies to events which have a duration within the time
frame from which they are viewed, whether or not the events
have a definite duration. The verbs meaning 'run' and 'chase'
in 26a and 26b do not describe events with a definite
endpoint. The neutral marker (NL) applies because the events
are temporally contained within the time frame from which they
are viewed.
- 53 -
26
a. Malagu yu' nigap
run(NL) I yesterday
"I ran yesterday."
b. Lao ti ha-dulalak yu' i ga'lagu
but not 3SG-chase(NL) me the dog
"But the dog did not chase me."
Russian has an aspectual marker (perfective) that indicates
events with an inherent temporal limit (telic events) and a
temporal limit within the time frame. ChamorrQ has an
aspectual marker (neutral) that indicates events with a
temporal limit within the time frame, whether or not those
events have inherent temporal limits' (whether or not they are
telic). In both cases the aspectual markers apply in the
description of events with temporal limits, but the type of
events grouped together differs slightly between the two
languages.
Russian and Chamorro represent two different ways in which a
language may flesh out semantically, the aspectual
distinctions related to an event's definite duration over
time. I will not discuss further the variation in semantic
interpretation of morphological aspectual markers such as
these. The point of this discussion is to establish that
their 'meaning' includes some reference to the property of
delimitedness.
The second complicating factor regarding morphological
aspect and delimitedness is that delimitedness may interact
- 54 -
with tense in particular ways. Many languages have
co-occurrence restrictions between tenses and morphological
aspects. Comrie (1976) notes that ChiBemba (a Bantu language)
has three tenses: past, present, and future. The perfective
and imperfective aspects may occur with the past or the future
tense, but only the imperfective may occur with the present
tense. In the same vein, some languages seem to have combined
tense and aspectual markers in their morphology. Comrie
(1976) also observes that in Arabic there is an opposition
which we will call here perfective/imperfective, although it
indicates more than perfective aspect. The perfective is used
for perfective aspect, past tense, and relative past tense.
(Examples 27 and 28 are from Comrie (1976).)
27
a. perfective and past:
Jalas(PERI) tall '1-babi.
"They-sat-down Rt the door."
b. relative past:
?ajt?u(IMP) -ka ?idK j'marra(PER ) '1- busru.
I-come to-you when it-ripen the unripe-date
"I shall come to you when the unripe date ripens (shall
ripen)."
(The ripening of the date is past with respect to my
coming.)
The imperfective form is used to indicate imperfective meaning
or (relative) non-past tense.
28
a. imperfective and present:
?allahu ya amu(IMP) bi- ma ta4maluna(IMP).
God he- now about what you-do
"God knows what you are doing."
- 55 -
b. relative future:
?arsala(PERF) yutLimu(IMP) -hu bi- tlika.
he-sent he-inform him about this
"He sent (someone) in order to inform him about this."
c. imperfective (although with relative past):
Wa ?attabatpwr (PERF) ma tatlU(IMP) '1 hayitinu jalg
and they-foIlow what they-recite the dem6ns in
mulki sulaymrna.
reign Solomon
"And they followed what the demons used to recite in
Solomon' s reign."
The nature of the interaction of morphological aspect with
tense and discourse (and modality as well, although this is
not illustrated above) depends on complex semantic, discoursal
and perhaps even pragmatic factors that are beyond the scope
of this thesis. It must be mentioned that in spite of the
obvious interactions between tense and aspect that are
revealed by co-occurrence restrictions of the kind discussed
above, these do r.cb necessarily argue against treating tense
and aspect as separate entities. Tense and aspect are not in
complementary distribution, as they would be if they belonged
to the same syntactic category. Co-occurrence restrictions
may be explained by a variety of extra-syntactic factors. The
apparent merging of tense and aspect into single morphemes, as
in the Arabic examples, may actually be an instantiation of
such co-occurrrence restrictions. In this thesis aspect and
tense are treated as independent entities in natural
language.
These'examples from Russian, Chamorro, ChiBemba, and Arabic
-56-
serve to illustrate that there is not necessarily a simple and
direct mapping from delimitedness to morphological aspect. I
do not intend to discuss the factors entering into how a
language may use morphological markers of perfectivity. I
intend merely to illustrate that definite or indefinite
temporal duration (delimitedness/non-delimitedness) is one of
the properties referred to by morphological aspect, and
consequently delimitedness must be represented in the grammars
of these languages.
The preceding discussion has focused on the complications
inherent in looking to morphological aspect for clearcut
expressions of the delimited/non-delimited distinction.
However, there are languages in which morphology directly
expresses that distinction. Russian has a set of verbal
prefixes, different from the Russian perfective and
imperfective morphology discussed above, which mark
delimitedness directly. The prefixation process is
productive, although there are some lexical constraints on
which prefixes may affix to which verbs. These prefixes
convert a verb describing non-delimited events into one
describing delimited events. They may add a straightforward
aspectual content to the verb without imparting new lexical
meaning, as in the examples below. (Examples in 29 and 30 are
from Brecht (1984).)
- 57 -
29
kurit' "smoke"
vykurit' "finish smoking"
dokurit'a "smoke to a climax"
iakurit' "begin to smoke, light up"
zakurit'_a "smoke too much"
najr i'4sJa "smoke one's fill"
pokurit' "smoke for a while"
Prefixes may also alter the meaning of the verb slightly, but
in such a way that the verb describes a delimited event.
30
Sitat' "read"
perecitat' "reread"
taitat' "read out"
doita "read up to"
vy~itat' "find (in a book)"
Perfectivizing prefixes in Russian represent a relatively
productive morphological process that depends on the property
of delimitedness.
In summary, although for a variety of reasons morphological
aspect does not always instantiate a clear and direct mapping
between delimitedness and morphology, at the very least it
makes reference to the definite or indefinite duration over
time of an event. And some aspectual morphology directly
expresses the property of delimitedness. These facts argue
for the importance of delimitedness in the grammars of natural
languages.
2.5 Affectedness
The term 'affectedness' as it is used in the literature
describes the semantic properties of a class of verbs and
their direct arguments which have certain syntactic
properties. In this and the following chapter, the semantic
properties associated with 'affectedness' will be demonstrated
to be aspectual in nature. In this section the class of
affectedness verbs (verbs with 'affected' arguments) and their
special properties are introduced.
2.5.1 English middles and noun phrase passivization
M. Anderson (1979) describes the class of arguments which
cannot undergo passivization in noun phrases as 'affected'
arguments.9 Noun phrase complements of nominal predicates may
be passivized if they are affected arguments, in the sense of
being "changed or moved by the action of the head nominal".10
The sentences in 31 describe events in which the objects ('the
city' and 'the natives') are affected by the actions of
destruction or conversion expressed by the verbs. These
objects can be passivized, as in the b) examples. The
sentences in 32 describe events in which the objects (-'Bill'
and "the cat') are not affected by the actions expressed by
9. For M. Anderson, NP passivization is 'NP Preposing'. Her
syntactic account of the phenomenon depends on a rule-based
grammar, which I do not adopt. 'NP passivization' and 'NP
Preposing' name the same phenomenon, although they imply
different assumptions about grammar and different views about
the syntax of the phenomenom.
10. M. Anderson (1979) p.44.
- 59-
t e verbs, and these objects consequently cannot be passivized
in the b) examples:
31
a. The Mongols' destruction of the city.
b. The city's destruction by the Mongols.
a. The missionaries' conversion of the natives
b. The natives' conversion by the missionaries
32(These examples are from M. Anderson (1979).)
a. John's avoidance of Bill.
b. *Bill's avoidance by John.
a. Sally's pursuit of the cat.
b. *The cat's pursuit by Sally.
M. Anderson identifies two other classes of derived nominals
which have preposable object arguments, and which require a
slightly extended definition of affectedness. These are nouns
that imply creation, such as "performance', 'definition',
'publication', 'translation' and 'portrait', and nouns of
concealment or exposure. (Examples in 33 are from M. Anderson
(1979).)
33
a. The company's performance of the play
b. The play's perlormance by the company
c. John's definition of that word
d. That words' definition b7 John
e. Mouton's publication of the book
f. The books publication by Mouton
g. John's translation of the poem
h. The poems' translation by John
i. Mary's portrait of the Senator
j. The Senator's portrait by Mary
k. John's concealment of the knife
- 60 -
1. The.knife's concealment by John
m. Alice's exposure of the corruption
n. The corruption's exposure by Alice
M. Anderson includes these preposable object arguments of the
derived nominals as being 'affected' arguments, by an
extension of the definition of affectedness. Even though they
are not physically altered in the event described by the
nominal, their relation to the observer is altered. Looked at
in.this way, they are included in the class of affected
arguments.
Verbs with affected arguments have a second distinguishing
syntactic property. They may form middles. The following
examples are from Hale and Keyser (1987):
34
a. This wood splits easily.
b. This door opens easily.
c. This cinch tightens easily.
d. This wheel spins easily.
In 34a through 34d the argument is affected. Examples of
verbs which do not have affected arguments and cannot form
middles are shown below. (These examples are from Roberts
(1985).)
35
Idioms:
11. The following minimal pairs (pointed out to me by Beth
Levin) demonstrate a verb which can form the middle only in
its non-idiomatic usage: 'Fine china breaks easily' vs.
- 61 -
*Advantage takes of John easily.
Perception verbs with small clause complements:
*John sees sing easily.
Exceptional -case marking verbs:
*John believes to be a fool easily.
Double object verbs:
*Orphans give presents easily at Christmas.
Psychological activity verbs:
*This theorem learns fast.
Perception verbs:
*The mountains see beautifully after rain.
The semantic property that allows verbs to form middles has
been discussed by Hale and Keyser (1987) and by Roberts
(1985). Hale and Keyser (1987) identify the Lexical
Conceptual Structure12 of verbs that can form middles ae:
36
[x cause [y "undergo change"], (by...)]
where x and y are arguments of the verb. This
characterization says that a verb with two arguments, one
causing the other to undergo a change, is a verb that can form
a middle.
Roberts (1985) discusses affected arguments and the role
they play in the formation of middles. He labels the affected
arguments themes, and argues that the presence of an internal
'*Good news breaks easily'.
12. For discussion and explanation of Lexical Conceptual
Structure Hale and Keyser (1986, 1987).
theme argument entails a change of state. (Roberts 1985,
p.393):
A clause containing the n-place predicate F(ng...nn)
where n1 is an internal Theme argument denotes a
true proposition iff the referent of the Theme argument
undergoes a change of state.
Roberts defines the expression 'change-of-state' as follows:
...some property of the Theme held before the time with respect
to which the proposition containing the predicate is evaluated
and fails to hold after that time, or vice versa.
Under Robert's definition, an affected argument (or an
internal Theme argument) is one that undergoes a change of
state, and a change of state provides a temporal -bound for the
event.
The property of affectedness which permits the formation of
middles and noun phrase passives in English correlates with
the existence of a temporal bound for the event described by
the verb. In other words, affectedness depends on
delimitedness. This will be demonstrated in greater detail in
Chapter 3. The point to be made here is that the class of
affectedness verbs and affected arguments in English is
distinguished syntactically. This is another example of the
syntactic consequences of the semantic property of
delimitedness.
2.5.2 Japanese nominal quantifiers
- 63 -
Miyagawa (1987) has observed a type of NP movement in
Japanese that depends on the property of affectedness for the
correct statement of the syntactic generalization. Japanese
has a set of nominal quantifiers that may occur following the
noun phrase. These consist of a numeral and a clitic marking
the noun class. 'San-nin' which follows the subject in 37a
below, and 'san-satu' which follows the object in 37b, are
examples: (Data in this section are from Miyagawa (1987).)
37
a. Gakusei ga san-nin ofisu ni kita.
students NOM 3-clitic(people) office to came
"Three students came to my office."
b. Hanako ga hon o san-satu katta.
NOM book ACC 3-clitic(volumes) bought
"Hanako bought three books."
Scrambling of the nominal quantifier from the postnominal
position within the verb phrase to the head of the sentence is
possible when the object is affected (as in 37), and
impossible when it is not (as in 38):13
38
a. Futatu, Taroo ga mado o aketa (koto).
two-olitic NOM window ACC opened
"Taroo opened two windows."
b. Mittu, Ziroo ga onigiri o tukutta (koto).
three-clitic NOM rice balls ACC made
"Jiro made three rice balls."
13. Scrambling of nominal quantifiers out of post-subject
position is impossible for independent reasons. See Miyagawa
1987) for details. For more discussion of scrambling in
Japanese, see Saito (1985).
- 64 -
c. Nimai, akanboo ga sara o watta (koto).
two-clitic baby NOM plates ACC broke
"The baby broke two plates."
39
a. ?*Hitori, Tanaka-sensei ga gakusei o hometa (kotod).
one-clitic Prof. Tanaka NOM student ACC praised
"Professor Tanaka praised one student."
b. ?*Futari, Taroo ga kodomo o donatta (koto).
two-clitic NOM children ACC shouted at
"Taroo shouted at two children."
c. ?*Hitori, Hanako ga kurasumeeto o kiratte iru (koto).
one-clitic NOM classmate ACC hate
"Hanako hates one of her classmates."
The distinction between affected and unaffected arguments is
also marked by a construction referred to as the
'intransitivizing resultative' (Martin, 1975). This consists
of the continuative form of the verb, marked by '-te',
followed by 'aru' (one of two Japanese verbs that can be
translated as 'be'). This construction changes a transitive
verb into an intransitive one, and gives it a resultative
sense. It applies to verbs with affected arguments (39), but
not to verbs with unaffected arguments (40).
40
a. Mado ga akete aru.
window NOM opened
"The window is opened."
b, Onigiri ga tukutte aru.
rice balls NOM made
"Rice balls are made."
c. Sara ga watte aru.
plate NOM broken
"The plate is broken."
41
- 65 -
a. *Ano gakusei ga homete aru.
'that student NOM praised
"That student is praised."
b. *sono kodomo ga donatte aru.
that child NOM shouted at
"That child is shouted at."
c. *kurasumeeto ga kiratte aru.
classmate NOM hated
"A classmate is hated."
The constraint on scrambling of Japanese nominal quantifiers
out of verb phrases, like the constraints on English middles
and noun phrase passives, must be stated in terms of
affectedness.
245.3 Total affectedness.
Roberts (1985) and M. Anderson (1979) use the term
'affectedness' to mean the property of an argument which
undergoes a change of state during the course of the event
described by the verb. The term 'affectednens' has been used
in another sense; to describe an argument that is holistically
or totally affected by the action of the verb. I will refer
to this as total affectedness. (The examples in 42 below are
from S. Anderson (1977).)
42
a. Bees swarmed in the garden.
b. The garden swarmed with bees.
42b implies that the garden was full of bees, that is, all of
the garden contained bees, or the garden was totally affected,
- 66 -
while 42a does not have the same implication.'4 Total
affectedness is closely related to affectedness. It is also
an aspectual property based on delimitedness, since an event
in which the argument is changed in all parts of itself must
be a delimited event.15 Like affectedness, total affectedness
is only associated with the direct argument of the verb.16
Total affectedness is therefore a semantic property that is
syntactically constrained. In 42a 'the garden' is not t-he
direct argument of the verb. In 42b (where swarm is used as
an unaccusative verb) 1 the garden' is the direct argument.
This distinction is even clearer in transitive verbs. The
examples below are from English, Dutch, and Japanese:
43
English:
a. Jeremiah sprayed paint on the w"1l.
b. Jeremiah sprayed the wall with paint.
a. Josiah cleared dishes from the table.
b. Josiah cleared the table of dishes.
14. Some speakers of English get optional total affectedness
readings. These speakers find that a totally affected
interpretation of the garden' is a grammatical option in 42b
but not in 42a.
15. For many types of verbs a delimited reading necessitates a
totally affected interpretation.
16. Direct argument-hood is syntactically defined; the direct
argument of the verb is the sister to the verb at deep
structure.
17. The theory of unacusative verbs will be addressed in
detail in Chapter 6.
- 67 -
Dutch: (De Groot (1984); his source, Dik (1980))
a. Jan plant bomen in de tuin.
"John plants trees in the garden."
b. Jan beplant de tuin met bomen.
"John be-plants the garden with trees."
Japanese: (Pukui, Miyagawa, Tenny (1985))
a. kabe ni penki o nuru
wall on paint 4C paint(VERB)
"smear paint oh the wall"
b. kabe o penkt de, nuru
wall ACC paint with paint(VERB)
"smear the wall with paint"
a. taru kara sake o akeru
barrel from sake ACC empty
"empty sake from the barrel"
b. taru o akeru
barrel ACC empty
"empty the barrel"
In the b) sentences of 45 the object is the verb's direct
argument, and the sentences all carry the implication of the
object's having been totally affoted. After the event
described by the b) sentences has transpired, the entire wall
is covered with paint, the table is empty of dishes, the
garden is full of trees, and the barrel is completely empty.
What are the direct arguments of the b) sentences are the
indirect or oblique arguments of the a) sentences. In the a)
sentences, they are not understood to be totally affected.
Affectedness and total affectedness are ser.antic properties,
but they are correlated with distinct and coasistent syntactic
behaviors or syntactic positions. Like delimitedness,
affectedness and total affectedness are semantic properties
- 68O-
that must be accessible to the syntax of natural language
grammars.
2.5.4 Affectedness and Transitivity
The property of affectedness which correlates with English
middles and noun phrase passives, and the property of total
affectedness discussed in the previous section are only
associated with the direct arguments of particular verbe --
not with the indirect or external arguments. The correlation
between affectedness and direct argument-hood has led to the
idea expressed in the literature that affectedness correlates
with transitivity.18  With transitive verbs the direct
argument, i.e. the object, may be an affected argument, while
the subject may not be affected. Kenny (1963) expressed this
elegantly:1 9
...if we ask, regarding any action: what changes as a result
of the action? Is it the agent or the patient: i.e. is it the
subject or the object of the verb reporting the action? With
an important exception, to be considered later, the answer is
always: the object. When A 's B, it is essential that after
this event B should have changed; it is not essential that
A should have changed. The stove, after it has boiled the
kettle, may look, feel, and behave exactly as it did before
doing so; but the kettle cannot have been boiled by the stove
unless it is warmer than it was when the process of boiling
started. To find out whether you have washed the dishes
it is of little use to inspect you; whereas an examination of
the dishes is always a help to settling the question, and may
indeed settle it definitely, if they are still dirty. To be
18. See Hopper and Thompson (1979).
19. Kenny (1963 ) pp. 180-181 .
- 69-
sure, there will very often be a change in the agent as a
result of an action: when I have chopped down the oak-tree, I
am usually hotter and stickier than when I started. But this
is not essential to the truth of "I have cut down the oak" as
it is essential to its truth that the oak should not be in the
same condition as it was. If I am so strong or so skillful
that I can chop down the oak without turning a hair, no
matter; but no amount of strength or skill could make it
true to say that I had cut down the oak-tree without denting
its bark.
"The important exception" mentioned by Kenny refers to the
fact that an agent may be said to have changed because it has
carried out an action, i.e., some proposition in the past
tense about the agent is now true, which was not true before.
This is so with both the agent and the patient. But this is a
qualitatively different kind of change from that which must be
true of the patient's present-state.
Affectedness correlates with direct argument-hood rather
than with subject- or object-hood. Whereas the direct
argument of a tranpitive verb (a verb taking two arguments)
will be the object, the direct argument of an unaccusative
verb will be its subject. The subjects of many unaccusative
verbe are also affected arguments (e.g.: 'The ice melted',
'The grass burned'.)
It.has been argued that affectedness of the object is the
semantic property crucial for transitivity, rather than the
agency or volitionality of the subject. Tsunoda (1985) points
this out with the English examples below:
- 70 -
44
a. I hit him.
b. I hit at him.
44A is transitive because it has a nominative-accusative case
frame. The patient2C'him' in 44a is the direct argument of
the verb, and in this sentence it may be understood to delimit
the event like an affected object.21 However, 44a may express
a situation with or without volitionality and agency on the
part of the subject. The conative alternant of 44a is the
intransitive 44b, in which the patient 'him' is not a direct
argument and is not affected, and the subject must have the
properties of volitionality and agency. These two sentences
demonstrate that morphosyntactic transitivity (in the sense of
having a nominative-accusative case frame) does not depend on
volitionality and agency of the external argument (the subject
'I' in these sentences). If anything, it is linked to the
affectedness of the internal or direct argument.
Affezted arguments are direct arguments. HLowever, the
converse of this statement has also been expressed in the
literature: that direct arguments are affected arguments, or
20. The term 'patient' refers to the thematic role the NP has
in the sentence. Thematic roles will be discussed at greater
length in Chapters 4 and 6. An argument with a 'patient'
thematic role represents a participant in the event described,
which is the recipient of the action expressed by the verb.
21. The reader is warned that verbs of contact ('hit', 'poke',
'tap') are rarely affectedness verbs, although 'hit' may be
ambiguous in this respect. Verbs of contact will be discussed
further in section 3.3.1.
- 71 -
at least more affected than indirect or external arguments.
This is not true of affectedness in the explicit sense defined
in this chapter -- in the sense of the argument undergoing a
change of state that makes the event in which it participates
a delimited event. This is the sense of affectedness which
correlates with middles and noun phrase passives. However, it
is correct to say that an argument of the verb seems more
"affected" by the action described by the verb when it is
expressed as the verb's direct argument, than when it is
expressed as an indirect argument. (To avoid confusion, I
will indicate this intuitive use of the term "affected" by
enlosing it in quotation marks. The only such use of the term
in the entire thesis is in this section, 2.5.4.) There are
cases in which an argument may be expressed either as the
direct or indirect object of the verb in semantic
paraphrases. In these cases the argument seems more
"affected" when it is the verb's direct argument. English
provides many examples of this. The following are from Givon
(1984).
45
a. She swam the Channel. (3 'swim across the Channel')
b. He rode the horse. (= 'ride on the horse')
The sentences under 45 may seem to be semantic correlates with
their paraphrases at the right. However, 'swim the Channel'
has a sense of mastering the Channel that 'swim across the
Channel' does not imply. 'Ride the horse' means to control
- 72 -
the horse, while 'ride on the horse' means little more than to
be perched upon and transported by the horse. In both cases,
even though the object seems not to be greatly "affected" by
the action of the agent, it is more "affected" when it is the
direct argument than when it is the indirect argument. This
is true even of 45b which expresses a non-delimited event.
Reanalysis of a verb and a preposition is sometimes
possible, a process which makes an indirect argument into a
direct argument. One test for a reanalyzed verb-preposition
construction is the possibility of forming pseudo-passives
such as:
46
The ball was run over by a truck,
in which 'run over' is a verbal constituent and 'ball' its
passivized direct argument. S. Anderson (1977) noted that
pseudo-passives correlate with the argument seeming "affected"
in the event. (47 is from S. Anderson (1977).)
47
a. This bed has been slept in (by George Washington).
b. *Cleveland seemed run amok in by John.
The reanalysis of 'sleep in' that makes 47a grammatical also
implies the bed has been "affected" by George Washington's
sleeping in it. The bed is somehow changed or made special by
this heving happened to it. The ungrammatical 47b, on the
other hand, has no reanalysis, and no impication that
- 73 -
Clevqland has been basically changed by John's running amok in
it.
The property of affectedness which correlates with English
middles and noun phrase passives is associated only with
direct arguments -m the objects of transitive verbs and the
subjects of unaccusative verbs. Dire.t argument-hood itself,
even when not associated with affectedness in the strict
sense, imposes some special interpretation in which the
argument seems more "affected" when it is expressed as a
direct argument than when it is expressed as an indirect
argument. The direct argument is semantically as well as
syntactically priviledged. Chapter 3 will address
affectedness in the technical sense -- the sense in which it
is based on delimitedness. Chapter 4 will consider the
special properties of direct arguments (as opposed to indirect
or external arguments) that lead to their priviledged
interpretations.
- 74 -
Chapter 3. An Aspectual Theory of Affectedness
3.1 Affectedness as an Aspectual Property
The term 'affectedness' characterizes a semantic property of
some verbs and their direct arguments which correlates with a
range of cross-linguistic syntactic behaviors, among them
middle formation and noun phrase passivization in English.
The property of affectedness has clear syntactic diagnostics
but its semantic definition has been somewhat amorphous.
Semantic definitions of affectedness in the literature are
based on the notion of the affected argument of the verb being
caused to undergo some change during the course of the event
described by the verb. A more explicit and enlightening
def.inition of affectedness is-provided by treating it as an
aspectual property dependent on delimitedness. Affectedness
may be defined as the property of a verb, such that it
describes a situation or happening that can be delimit 1d by
the direct argument of the verb. Affectedness verbs describe
events which are 'measured out' and delimited by their
direct arguments. Affectedness.defined in this way as an
aspectual property more adequately characterizes the verbs
that allow middles and noun phrase passives than the
definition of affectedness based on the notion of 'undergoing
change'. This aspectual definition of affectedness applies to
1. The term 1event' is used here to mean the situation or
happening described by a verb, in which the arguments of the
verb participate.
- 75 -
a range of verb classes. Verbs of consumption and creation,
verbs expressing a change of state, achievement verbs, and
verbs of motion are encompassed in the aspectual definition of
affectedness. Some verbs which allow middle formation and NP
passivization but do not fall under the definition of
affeccednese based on 'change', fall under the aspectual
definition based on delimitedness. The aspectual definition
of affectedness also accounts for the fact that verbs capable
of ambiguity between delimited and non-delimited readings
selict the delimited readings in constructions such as middles
and NP passives. In the following sections the aspectual
definition of affectedness is developed and applied through an
investigation of the lexical semantics of several verb
classes.
3.1.1 Verbs of consumption and creation.
Hale and Keyser (1987) characterized verbs that form middles
as having a lexical conceptual structure like the following:
1
[x cause (y "undergo change"]
The idea that these verbs express a 'meaning' in which one of
the verb's arguments causes some change in the other argument,
makes fairly accurate predictions about which verbs will show
the syntactic behavior associated with affectedness. The idea
of change is central to affectedness. Consider what change
is, and how it is related to delimitedness. In the simplest
-76-
and most concrete view, change is the re-arrangement of things
or properties in space, viewed over time. We may view the
situation described by a verb as the collection of changes
that occur in (or among) each of the verb's arguments during
the course of the event described by the verb. A clange of
state in one of the verb's arguments will provida a distinct
temporal marker - an endpoint -- for the event. If there is
a central change in the event, the argument that undergoes
that change is responsible for much of the temporal quality of
the event. If the argument achieves a final or absolute
change of state it is responsible for delimiting the situation
described by the verb. Furthermore, if that argument
undergoes the change gradually, it 'measures out' the progress
of the event towards that endpoint in increments of change
over time. Many verbs of consumption and creation can be
characterized in this way. The verb 'eat' is an example:
2
eat an apple
The apple referred to by the direct object in 2 'measures out'
gradually the event described by the verb phrase. Someone who
eats an apple progresses through the event in increments of
apple. 'Eat an apple' is a delimited accomplishment because
the action of eating an apple may result in the total
consumption of the apple. We may think of an event as a
series of snapshots of the obJects involved, at points along a
time line. The snapshots record the property that is changing
- 77 -a
in the object. (Since the apple is being consumed, the
property that is changing is that of size or spatial extent.)
In the case of apple-eating there will eventually be some
snapshot in which the apple is gone, and all subsequent
snapshots will have no apple. It is the existence of this
distinctive point of time, provided by some changing property
of the object, that makes a delimited accompliahment.
With a verb like "eat', that describes a gradual change, the
direct argument can be said to 'measure out' the event. The
changing property of the argument may be partitioned into
equal sub-parts so that it becomes a scale, and the event
measured out' in terms of that scale. In the case of a verb
phrase like %eat an apple', the scale applied to the event is
finite, and 'measuring out' the event yields a definite
value. If I eat an apple, I eat one bite of it at time t,
another bite at t2 and so on until at some tn the apple is
gone because the apple has finite spatial extent. This is
true of all verbs that describe changes that progress over
time through subparts of the object that is the verb's direct
argument. If a piece of ice melts, some of it has melted by
the end of,t and more of it has melted by the end of t2 , and
all of it Was melted by tn. If I 4raw a circle, I draw part
of it.daring t1 , part of it during t2 and so on until at tn
the entire circle is drawn. The delimited reading of such
expressions entails that I never eat the same part of the.-
apple twice, or draw the same stretch of circle twice, and no
portion of the ice melts and freezes and melts again. It
assumes that I eat all the chunks of apple, and eat each of
them only once. In such a case the apple itself imposes a
finite and definite duration on my eating of it. Thus the
object is a measure of the event, as if I had performed one
apple's worth of eating.2
Now compare 'eat an apple' with 'push a cart'. Pushing a
cart is a non-delimited activity instead of a delimited
accomplishment. A snapshot series of cart-pushing will record
changes in the cart's locatioi over time but there will be no
distinctive snapshot to mark the end of the event.
The preceding discussion illustrates how delimitedness may
be' imposed on the event through the direct argument of the
verb. The property of affectedness is redefined here as
follows:
3
A verb is an affectedness verb iff it describes an
event that can be delmiTd by the direct argument of
the verb.
Taking the term 'argument' to refer ambiguously to either a
2. The delimited reading is not always required where it is
possible; in fact this varies across languages and may be
morphologically marked.
- 79 -
syntactic or a semantic element, I will define 'affected
argument' in both domains:
4
A syntactic argument is an affected argment. iff it is
the direct argument of an afecfedness verb, and its referent
delimits the event described by the verb.
5
A semantic argument is an affected argument iff it
independently delimits the event in whchit
participates.3
This view of affectedness explains why verbs like 'perform',
tread' or 'record', which do not seem intuitively to describe
actions which affect or change their objects, should belong in
this class. 'Perform', 'read' and 'record' may form middles,
and may appear in passivized noun phrases:
6
a. This play performs easily.
the play's performance by the company (M. Anderson, 1979)
b. This paper reads easily. It is very well written.
The paper's reading at the chemical society created a lot
of excitement.
c. This magnetometer data recorded easily. The data from the
echo-sounder was more of a problem, because the scale
continually needed to be readjusted.
This data's recording was not the problem. It was the
analysis that caused us so much trouble.
The referents of the direct arguments, 'play', 'paper' and
3. The term 'independently' is included in this definiiion to
rule out Goal arguments represented in the syntax as indirect
arguments of the verb. This will be explained more fully in
Chapter 4.
- 80 -
6'data', fmeasure out' the events of performing, reading and
recording. When a play is performed or a paper read, the
event of performing or reading proceeds through the play or
paper incrementally -- during each subinterval of performance
or reading time a sub-part of the play is performed, and a
sub-part of the paper is read. When a set of data is
recorded, more and more data is recorded until finally all of
it is recorded. The direct argument delimits the event.
Nothing else need be said to explain why 'perform', 'read' and
record' have affected objects.4 Hale and Keyser's definition
4. A digression is in order here about the variability of
speaker judgements about middles. The judgements here are
those of the author, which are shared by other speakers
consulted. However, these judgements are fairly liberal, and
many speakers have much more constrained middle formation in
their grammars. For thoae speakers with somewhat less liberal
judgements, it is predicted that they will find the same
pattern of relative acceptability based on the object's
ability to delimit the event. It is also predicted that
speakers with conservative middle formation will have
conservative NP passivization as well. For those speakers who
accept virtually no middles at all, this pattern of relative
acceptability will not be apparent. Other factors intervene
in the grammars of such speakers to rule out middle
formation. (These factors may include discourse or stylistic
conditions requiring the sentence to be construable as being
'abou;' the subject. Van Oosten (1977) argues in this vein
that middle formation is constrained by the requirement that
some properties of the subject must be understood as partly
responsible for the action expressed by the verb.) The
variability in judgements also reflects a variability in the
lexical meanings of verbs. Speakers do not learn and use the
same verbs in exactly the same manner. The verb 'read' is an
example. Most speakers get both a delimited and a
non-delimited reading for Mread the book', but for some
speakers only one or the other is possible. Which readings
are porsible depends on the %meaning' assigned to the verb in
the speaker's lexicon. In any case, it is my thesis that, in
spite of the variability of judgements about mid.les, it is
- 81 -
of affectedness based simply on 'undergoing change' cannot
account for verbs like 'read'. The aspectual definition of
affectedness subsumes a wider range of verbs with the
appropriate syntactic behavior.
Verbs of motion whose direct arguments are delimiting paths
appear to be somewhat problematic for this view, but on closer
inspection they behave in the same way as 'perform',, 'read'
and 'record'. The direct objects of the verbs below represent
an area, distance or object which is traveled through in the
course of the event described:
7
a. cross the desert -
b. traverse the continent
c. climb the mountain
d. swim the Channel
e. run a lap
f. walk ten miles
g. circumnavigate the globe
h. infiltrate the enemy battalion
These may form middles in the author's judgement:
8
a. The desert crosses more easily than the prairie for
settlers with large wagons.
b. Today the continent traverses in only four hours compared
with the weeks or months it took a hundred years ago.
c. That mountain climbs easily from the west side but it has
never been attempted from the east.
d. ?Thq Channel swims in fifteen hours for a swimmer in top
condition.
e. ?The last lap runs the hardest.
f. ??Ten miles walks easily in good shoes.
the delimitability of the event through the the direct
argument that is required by the grammar for middle formation
to be possible.
- 82 -
g. The globe circumnavigates in a day with Pan Am.
h. The enemy battalion infiltrated surprisingly easily for
the guerrilla soldiers.
The examples above are admittedly somewhat awkward, but
comparable sentences with non-delimiting verbs are much worse:
9
a. wander the desert
b. travel the globe
c. pursue the enemy battalion
10
a. The prophet wandered the desert for forty years.
b. John traveled the globe (over) until he ran out of money.
c. The guerrillas pursued the enemy battalion.
11
a. *The desert wanders more easily than the prairie for
settlers with large wagons.
b. *The globe travels easily with Pan Am.
c. *The enemy battalion pursued surprisingly easily for the
guerrilla soldiers.
The verbs in 9 do not describe delimited events. They may
take paths as direct arguments (as in 10) but the middles
formed with these verbs (in 11) are much worse than those
formed with the delimiting verbs (verbs describing delimited
events) in 8.
Noun phrase passives show the same correlation of
acceptability with delimitedness:
12
a. The desert's crossing was inevitable, once gold was
discovered on the other side.
b. The continent's traversal takes five days.
c. ?That mountain's climb will be attempted before its
mapping has been completed.
d. ??The Channel's swim is less of a problem than its traversal
- 83 -
m
by sailboat.5
e. ?One more lap's run is impossible.
f. Ten mile's walk was enough.
g. The globe's circumnavigation was impossible until sextants
were develored.
h. The enemy battalion's infiltration was carried out at
night without mishap.
?The enemy battalion's infiltrating was carried out at
night without mishap.6
Comparison of the noun phrase passives in 12 describing
delimited events, with noun phrase passives in 13 describing
non-delimited events shows a strong difference in
acceptability:
13
a. *The desert's wandering wa' inevitable once gold was
discovered on the other side.
b. *The globe's traveling-was impossible until railways were
developed.
c. *The enemy battalion's pursuit was carried out at night
without mishap.
*The enemy battalion's pursuing was carried out at night
without mishap.
The property of affectedness defined in this way through
5. I do not have an explanation for why these particular
zero-derived nominals are awkward in these constructions. It
is propbably related to the fact that these verbs have strong
non-delimited readings that interfere with the less salient
delimited reading required for noun phrase passives.
6. '-ing' nominals are somewhat more difficult to construe as
noun phrase passives than many other nominals; '-ment'
nominals, for example. This is possibly because %-ing'
preserves the sense of ongoing action or erocess, whereas
-ment' signifies the completed action. -ing' nominals may
describe delimited events, but their delimited reading is not
as salient as it is for '-ment' or '-tion' nominals. This may
be the reason why '-ing' nominals make worse NP passives than
'-ment' or '-tion' nominals.
- 84 -
delimitedness does not depend on the presence of an agent, and
affectedness is not limited to the objects of transitive
verbs. Crmpare 'Jefferson dies' with 'Jefferson swims'. A
anapshot series over time of Jefferson's dying will be certain
to have one distiActive snapshot in which Jefferson is
fundamentally changed; he is dead. Snapshots of
Jefferson's swimming will have no such distinctive shot.
Affectedness depends only on the direct argument's ability
to delimit the event. Verbs with unaffected arguments such as
'puh' or swim' describe events that may aquire terminal
points. Jefferson may tire of swimming or he may drown or the
lak( may freeze over. The cart may fall to pieces, or the
road become impassable, and cart-pushing come to an end.
Jefferoon may swim only to the end of the lake, or the cart be
pushed only to the end of the block. Prepositional phrases
expressing a goal may be added to the verb phrase to delimit
it;
14
Jefferson swam to the end of the lake. (delimited)
push the cart to the end of the block (delimited)
However, these termiril points are not inherent in the verb's
direct argument as it participates in the event -- they are
7. This is Independent of medical controversies over what
(onstitutes death. However death is defined, there will be
one point of time when it has arrived.
. imposed externally. Verbs with affected arguments describe
events that have their terminal points built into the verb and
its immediate argument. These terminal points are interpreted
with no additional qualifications. It is sufficient to
observe the single argument, 'Jefferson' in 'Jefferson dies',
or 'apple' in 'eat an apple' in order to tell when the event
has come to an end.8
Verbs such as 'eat' and 'draw' have affected arguments which
measure out' the event according to the objects' spatial
extent. That property can provide an Inherently finite scale
against which to 'maasure out' the event, and these events are
consequently delimited whenev-r the object is itself finite.
They naturally include verbs of consumption, creation,
destruction, concealment, and enclosure. Verbs having dirict
objects that are themselves events, such as verbs of
performance, will also be included in this group. These are a
few examples:
15
eat, paint, translate, transfer, transmit, transpose,
perform, disperse, publish, define, corral
These verbs form middles and noun phrase passives:
8. The reader is reminded that affectedness is a linguistic
property, and is therefore a property of events as they are
organized and expressed by language. It is not a property of
events in the world.
- 86 -
16
Apples eat easily, but I don't like pineapples because they're
too much trouble to eat,9
One apple's eating was all it took to break a tooth.
Homer translates easily.
That poem's translation was a difficult undertaking.
Messages don't transmit easily to Mars.
The message's transmission was delayed by mechanical
problems.
According to the police, crowds disperse easily when it is
raining.
The crowd's dispersal was effected by the police.
Goats don't corral easily.
That goat's corraling took three hours and a lot of rope.
This group also includes verbs that describe a change that
travels through the object from one end to the other:
17
Hot bread cuts easily.
We'd better leave the cake's cutting to Mary.
Dollar bills rip in half easily.
The sail's ripping in half by the storm made the crew
decide to turn back.
34.2 Verbs of change of physical state
The verbs of consumption and creation discussed in the
previous section describe situations which are delimited by
9. For speakers with more conservative judgements of middles,
this sentence may improve if "apples' is specified: 'One apple
eats easily enough with dentures, but I don't think I could
eat four of them without my own teeth.' Also consider the
Catpbell's soup commercial, noted by Van Valin (1987a): 'The
soup that eats like a meal'. Nevertheless the fact remains
that "eat' and 'drink' do not form middles as readily as other
verbs of consumption and creation. These verbs have peculiar
properties for which I have no explanation.
- 87 -
the size or spatial extent of their direct arguments. With
verbs such as these the 'change' in the direct argument is
final and absolute, and therefore delimiting. But 'change'
may also be relative change, in which case it is
non-delimiting. Some verbs describe changes which are either
absolute or relative, and so they may occur in verb phrases
that are ambiguous between a delimited and a non-delimited
reading. In the delimited reading, the verb names the state
which is the final result of the change taking place in the
direct argument. In the non-delimited reading, the verb
simply names the property that is changing. 'Burn' and 'fade'
are examples of such ambiguous verbs:
18
The fire last year burned the old oak tree.
(The oak tree may have burned for awhile, or it may have
been completely burned up.)
That old photograph faded in thu attic trunk.
(The photograph might have gone on fading indefinitely in the
trunk, or it may be said to have reached the state where it is
faded.)
Verb particles remove the ambiguity and select the delimited
reading:
19
delimited or non-delimited:
a. The fire last year burned the old oak tree.
b. These children will grow taller.
c. Your socks will dry on the radiator.
. delimited:
a. The fire last year burned down the old oak tree.
b. These children will grow a.
c. Your socks will dry out on the radiator.
- 88 -
Abusch (1986) discusses this type of ambiguity in inchoative
verbs. Certain inchoatives that are related to adjectives may
describe either delimited or non-delimited situations.
20 (Abusch 1986; her source, Partee p.c.)
a. The weather has finally cooled.
b. The weather has cooled considerably.
21
a. The Atlantic Ocean widened in three decades.
b. The Atlantic Ocean widened for three decades. (Abusch 1986)
20a is naturally understood as having an end-state or as being
a delimited situation, while 20b has the implication of a
relative change, or simply a non-delimited cooling process.
The modifiers 'finally' and *considerably' select for each of
these readings. The expression 'in three decades' in 21a
indicates a completed situation, or a delimited reading of the
sentence, while 'for three decades' in 21b selects a
non-delimited reading. Since the verb 'widen' may be used
with either one, it may have either interpretation. This kind
of ambiguity is described by Abusch as an ambiguity between a
"become-adjective" (delimited) and a "become-adjective-er"
(non-delimited) interpretation. Evidence for this view is
provided by data such as.-the following. (Abusch, 1986):
22
a. The Atlantic Ocean is wide and is widening.
b. *The Atlantic Ocean is wide and becoming wide.
c. The Atlantic Ocean is wide and becoming wider.
22a has an interpretation in which the verb 'widen' describes
- 89 -
an ongoing process of change rather than an absolute change of
state. 'Widen' in this reading describes a non-delimited
situation, and is paraphrased by 22c rather than 22b.
'Becoming wide' in 22b is an expression describing a delimited
situation, because the adjective 'wide' provides an absolute
endpoint to the change. Once something is wide it cannot
continue to achieve wide-ness, hence the ungrammaticality of
22b. This is the reason why, in 22a, 'widen' is forced to have
the non-delimited interpretation. However, 'becoming wider'
in 22c describes a non-delimited situation, because 'wider' is
not an absolute state. There is no absolute endpoint for the
event at which 'wider-ness' is achieved.
Furthermore, Abusch notes (her source, Partee, p.c.) that
certain modifiers can diet nguish between the
"become-adjective" and become-adjective-er" readings of
inchoative verbs. (Abusch 1986)
23
Adjectives Comparatives Inchoatives
a.*a lot cool a lot cooler has cooled a lot
*quite a bit cool quite a bit cooler has cooled quite a bit
*ten degreees cool ten degrees cooler has cooled ten degrees
b. completely cool *completely cooler has cooled completely
absolutely cool *absolutely cooler has cooled absolutely
All the modifiers in 23 occur with inchoatives. The modifiers
in 23a, which also occur with comparatives, select for the
"become-adjective-er" or non-delimited reading. The modifiers
- 90-
in 23b, which express some absolute degree and do not occur
with comparatives, select for the "become-adjective", or
delimited reading, of the inchoative.1 0
Recall that 'for' durative adverbials may be used as tests
for delimitedness. These durative adverbials take
semelfactive readings with non-delimited sentences, and
repetitive readings with delimited sentences. The data below
shows that 'quite a bit' selects a non-delimited reading, and
completely' selects a delimited reading:
24
a. The Mid-Atlantic Ridge cooled a lot/quite a bit for fifty
years.
b. *The Mid-Atlantic Ridge cooled completely for fifty years.
24a can be understood as one event of' cooling, while 24b must
be interpreted as many events of cooling. Since this reading
is incompatible with the pragmatics of the situation, the
sentence is unacceptable.
On this basis, these two classes of modifiers can be used as
tests for delimitedness. Consider the following four verbs:
'tighten',, 'split', 'open', and 'spin'. With an unaccusative
reading, both classes of modifiers may occur with each verb.
For this test, consider only the reading in which the modifier
10. Degree modifiers do not all fall into one of these two
classes. Abusch lists the following degree modifiers that may
not occur with either comparatives or inchoatives: 'very',
'quite', 'so', 'pretty', and 'rather'.
- 91 -
applies to a single event, rather than a generic repetition of
events.
25
a. The cinch tightens a lot/quite a bit/completely.
b. The woods splits ?*a :ot/?quite a bit/completely.
c. The door opens a lot/quite a bit/completely.
d. The wheel spins a lot/quite a bit/completely around.
Consider these familiar middles once more. (Example 26 is
from Hale and Keyser (1987)):
26
a. The cinch tightens easily.
b. The wood splits easily.
a. The door opens easily.
d. The wheel spins easily.
When the two kinds of modifiers are applied to the middles in
26 a difference emerges. Although all the examples below are
stylistically awkward and somewhat unnatural, there is a clear
difference in acceptatility between the sentences in 27 and
28. Those in 097 with modifiers that select for-
non-delimitednese are much worse than those in 28, with
modifiers that select for delimitedness.12
11. 25b is odd with 'a lot' and 'quite a bit' for independent
reasons. 'Splitting' has a strong tendency to be interpreted
as happening nearly instantaneously. It is consequently
difficult to get these modifiers to modify the single event of
splitting. In the author's judgement it is possible, though
awkward, to get this interpretation with 'quite a bit'. This
will be the adverbial expression that is used with 'split' in
the examples that follow.
12. () A modifier like 'ten degrees'. is omitted from 27 for
two reasons. Its effect on acceptability is inconsistent, and
- 92 -
27
a. ?*The cinch tightens a lot/quite a bit, easily.
b. ?*The wood splits quite a bit, easily.
c. ?*The door opens a lot/quite a bit, easily.
d. ?*The wheel spias a lot/quite a bit, easily.
28
a. The cinch tightens completely, easily.
b. The wood splits completely, easily.
c. The door opens completely, easily.
d. The wheel spine completely around, easily.
Now consider the following noun phrase passives. They
include both middle-forming verbs (a-d) and pure unaccusative
verbs (e-f).
29
a. The cinch's tightening was difficult.
b. The wood's splitting was difficult.
c. The door's opening was difficult.
d. The wheel's spinning was difficult.
e. The lake's evaporating was a surprise.
f. The ice's melting was a surprise.
The two unaccusative exemples above may be used with either
set of modifiers:
30
The lake evaporated a lot/quite a bit/completely.
The ice melted a lot/quite a bit/completely.
it seems to have the ability to convert a non-delimited
sentence into a delimited one: "*It cooled ten degrees for
three hours' versus 'It cooled ten degrees in three hours'.
The other modifiers in this group do not affect the
'non-delimitedness of the sentences they apply to. (ii)
'Absolutely' is omitted from 28 because it is unnatural with
these particular verbs. Insofar as it can be used with these
verbs, in the author's judgement it behaves the in the way
same as 'completely'.
- 93 -
When the two classes of modifiers are applied to these noun
phrase passivesthe same distinction emerges:13
31
a. ?The cinch's tightening a lot/quite a bit was difficult.
b. ?*The wood's splitting a lot/quite a bit was difficult.
c. ?*The door's opening a lot/quite a bit was difficult.
d. ?The wheel's spinning a lot/quite a bit was difficult.
e. ?*The lake's evaporating a lot/quite a bit was a surprise.
f. ?*The ice's melting a lot/quite a bit was a surprise.
32
a. The cinch's tightening completely was difficult.
b. The wood's splitting completely was difficult.
c. The door's opening completely was difficult.
d. The wheel's spinning completely around was difficult.
e. Vie lake's evaporating completely was a surprise.
f. The ice's melting completely was a surprise.
Further evidence for the delimitedness of noun phrase
passives is adduced by comparing '-ing' nominals having the
direct argument in specifier and complement positions:
33
a. tha cooking of the stew
b. the stew's cooking
33a, which has the direct argument "stew' in complement
position, may be understood to mean a delimited or
non-delimited event of cooking. 33b, a noun phrase passive
with the object in specifier positior, can only be interpreted
as delimited. When the adverbial expressions 'in an hour' and
13. The difference between 31 and 32 is even more pronounced
when a-d are unaccusative. Substituting 'w:s a surprise' for
was difficult' improves 31. Using unaccusative verbs
abstracts away from complicating factors in middle formation.
- 94
'for an hour', which distinguish between delimited and
non-delimited expressions, are applied to 33b above, different
structures result:
34
a. iNP~the stew's] ([cooking] [in an hour]]j
b. iN(the stew's] [[cooking] (for an hour]]H
c. iNP[the stew's] (cooking]} [in an hour]
d. *INP[the stew's] (cooking]} (for an hour]
The ungrammaticality of 34d is even more pronounced with the
determiner 'several' 14:
35
*tNP[several stewe'] (cooking]I (for an hour]
'In an hour' applies to a delimited expression to yield
another delimited expression. 'For an hour' applies to a
non-delimited expression, but changes it to a delimited one.
The structure in 34d and 35 is impossible because it forces a
non-delimited reading of the specified NP.
Verbs that are ambiguously interpretable as describing
delimited or non-delimited situations describe delimited
situations when they occur in noun phrase passives.15 'Widen'
14. This was suggested by a comment of Barry Schein's.
15. '-ing' is a'nominalizing morpheme that preserves the
delimited/non-delimited ambiguity. There are some
nominalizing morphemes that do not; '-tion' for example,
requires a delimited interpretation of the event represented
- 95.-
and 'redden' are two examples. In 36 and 37 below, the a) and
b) examples show no difference in acceptability when the
adverbial expressions 'in an hour' and 'for an hour' are
applied. The noun phrase passives in the a) and b) sentences,
on the other hand, do show such a difference. Applying the
adverbial 'for an hour' makes the sentence awkward. This is
because 'for an hour' forces a delimited expression to be
interpreted as repetitive, and a repetitive reading is not
easily imposed on the situatione described in these
sentences:16
36
a. The federal official said that the highway's widening in a
year would be impossible.
b.?The city official said that the highway's widening for a
year would be a nuisance.
c. The federal official said that the widening of the highway in
year would be impossible.
d. The city official said that the widening of the highway for a
year would be a nuisance.
37
a. The western sky's reddening by air pollution in an hour
never happened before.
b.?The western sky's reddening by air pollution for an hour
by the nominalized predicate.
16. Some speakers do not allow NP passives formed by '-ing'
nominalisations of verbs like 'widen'. For these speakers 36a
and 36b will be bad sentences. These speakers should get the
corresponding difference between: 'the bypass' extension in a
year' and 'the bypass' extension for a year'. *
- 96 -
never happened before.17
c. The reddening of the western sky by air pollution in an hour
never happened before.
d. The reddening of the western sky by air pollution for an nour
never happened before.
Nominalised '-ing' nouns with subjects marked by genitive case
do not have to form delimited expressions:
38
a. John's trusting of the landlord (for a month/*in a month)
got him into trouble.
b. Harry's smoking (for a month/*in a month) is making him
sick.
Change of state verbs like those discussed in this section,
which may. be ambiguous between a delimited and a non-delimited
reading, select the delimited reading when they occur in
middles and noun phrase passives. This fact is not captured
by the view of affectedness as depending solely upon some
change in the direct argument. The correct generalization
requires making the distinction between delimited and
non-delimited change.18
Verbs of consumption and creation like 'eat' or 'draw' and
change of &tate verbs such as those above describe situations
17. The biterpretation looked for here is that in which the
reddening process is continuing for an hour, not the result of.
the reddening process.
18. Roberts' (1985) definition of the change required by
affectedness does in fact express a delimited change.
- 97 -
that are delimited through the direct argument. However, the
way the direct argument delimits the event is slightly
different in the two types of verbs. Verbs of consumption or.
creation describe events that progress through the object
physically, so that the finite spatial extent of the object
translates directly into the delimitedness of the event. With
a change of state verb describing a delimited event, the
endpoint is not necessarily achieved by progressing through
the object. How the action progresses through the spatial
extent of the object may be pragmatically interpreted. All
parts of the object may achieve the change of state
incrementally; or the new state may be arrived at over the
entire object at once; or some combination of both
possibilities may be true. The verb 'ripen' is an example. A
piece of fruit may ripen gradually from one end to the other;
or all parts of the fruit may ripen at an equal rate; or there
may be stretches of time when different parts of the fruit are
all ripening at the same rate, and stretches of time when tLey
are not. "Tan' is another example. 'Tanning the skin'
usually means tanning different parts of the skin at equal
rates, but some parts may tan faster than others depending on
the orientation of the sun, and the changing shadows. With
consumption/creation verbs and with change of state verbs, it
is not the spatial'extent of the object which is 'measuring
out' the event but some property central to the verb's meaning
which is associated with the object.
- 98 -
The me.taphor of the measuring scale applies to these verbs
as it does to verbs of creation and consumption. How the
scale of ripeness or tanness is applied to the event is
pragmatically determined, but what scale is to be applied is
specified in the lexical entry of the verb. Some property
central to the meaning of the verb becomes the scale that
'measures out' the event. With the verb 'ripen' it is the
property of ripeness, and with the verb 'tan' it is the
property of tanness that 'measures out' the event.
3.1.3 Verbs of abstract changes of state
Some verbs describing abstract changes of state also belong
in the class of affectedness verbs. These verbs may be
characterized in the same way as verbs expressing physical
changes of state. Three examples are, 'persuade' , 'confirm'
and 'bribe'. They participate in the middle and noun phrase
passive constructions:
39
a. Robert persuades easily. I talked him into buying a widget.
b. The Congressman's persuasion by the trade representatives
turned the decision in their favor.
a. Candidates nominated by the President confirm easily
in this administration.
b. The applicant's confirmation by the panel was finalized.
a. City officials bribe easily.
b. The mayor's bribing by the tycoon made big news.
- 99 -
19 The direct argument of an abstract affectedness verb
delimits the event, because the event is complete once the
direct argument has arrived at the new state. Persuading
Robert or confirming the applicant lasts until Robert is
persuaded and the applicant confirmed. In the delimited
reading of 'bribe', the act of bribing an offical is over when
the official has accepted the bribe, not when the briber has
offered it. Abstract change of state verbs like these are
encompassed in the aspectual definition of affectedness.
3.1.4 Achievement verbs
Many verbs with affected arguments can be characterized as
describing situations that are measurable over a certain time
interval along a scale provided by some property of the object
argument. These are accomplishment verbs, in Vendler's
terminology. However, there are other verbs with affected
arguments that seem to describe events with little or no
duration over time. These are Vendler's achievement verbs.
40
Glass breaks easily.
The glass' breaking made a terrible noise.
Bombs explode easily.
The bomb's exploding closed down the state offices.
19. These verbs are not generally ambiguous in the way that
physical change of state verbs are, although in the judgement
of the author *bribe' is ambiguous between a delimited and a
non-delimited reading.
- 100 -
Recall the discussion in Chapter 1 in which it is argued that
the distinction between achievement and accomplishment verbs
is not grammaticallly significant. It is delimitedness that
is grammatically important -- not duration. If the event is
one in which the argument undergoes a change of state, that
event has some duration, however little, merely because there
can be no change of state without a before and after. Even
though an interpretation is strongly favored in which the
temporal duration of the event is minimalized, the durative
quality of such an event is never completely eliminated. The
verb 'explode' is an example. Like 'eat', 'explode' takes an
affected object. It appears to differ from 'eat' in that an
event of something exploding usually happens in an instant,
while an event of eating scething has duration. Although an
explosion normally takes an instant, the explosion of a
supernova may take millions of years. 'Crack' is another such
verb. Cracking a pane of glass takes only an instant but
cracking the bough of a tree can take a certain amount of
time.2 0
Achievement verbs, therefore, describe events as having some
duration, however little. The events they describe take some
20. The reader is again reminded that it is the linguistic
representation of events that is under discussion here; that
is, events as they are organized and represented by language
-- not events as they are in the world.
- 101 -
finite amount of time to transpire. The endpoint of that time
period is registered in the direct argument -- e.g., the
cracked or exploded object. The aspectual definition of
affectedness based on delimitedness unites achievement verbs
with accomplishment verbs. A verb's lexical entry specifies
whether the verb describes an event as having extended
duration or little or no duration. There may be considerable
lexical variation among speakers of the same language, and a
speaker's lexicon may be influenced by the speaker's knowledge
or perception of the world. The difference between
accomplishment and achievement verbs is a lexical difference
without deep grammatical significance.
3.1.5 Verbs of motion
Even though the direct argument of a verb of motion does not
seem to delimit the event described by the verb, certain verbs
expressing motion caused by an agent may form middles:
41
a. Lorenzo pushes the cart.
b. This cart pushes easily.
a. Olga gallops the horse.
b. This horse gallops easily.
The a) forms illustrate the verbs in their regular transitive
usage and the b) forms illustrate the verbs used as middles.
Many verbs of motion in the middle usage often have a peculiar
inceptive interpretation. For many speakers the b) sentences
above actually mean this cart starts to push easily' and
- 102 -
'this horse starts to gallop easily'. An inceptive
interpretation of these sentences is forced by the
delimitedness constraint on middle formation. Verbs of motion
used in middles are actually used as change of state verbs.
That is, the moved object undergoes a change of state from
being an unmoving object to being a moving object. In these
cases, the direct argument -- the moved object -- is what
delimits the situation.
Verbs of motion used in the middle construction must have a
delimited interpretation. If it is not an inceptive
interpretation as above, it will be a completive
interpretation.21 As tests for delimitedness we can compare a
goal phrase and the expression 'a little ways'. The goal
phrase in 42a, 'to the other side', selects a delimited
interpretation while the expression *a little ways' in 42b
selects a non-delimited reading. Since 'cross the desert'
describes a delimited event, only the goal phrase 'to the
other side' may co-occur with it. 'A little ways' makes the
sentence bad:
42
a. cross the desert to the other side
b. *cross the desert a little ways
When this test is applied to verbs describing events of
21. Many speakers only get one or the other reading (inceptive
or completive), but not both.
- 103 -
* imparting motion such as 'heave', 'herd', 'drive' or 'push',
it is apparent that they describe completively delimited
events when they are used as middles:
43
a. Mary heaved the boulder (into the ditch/a little ways).
b. Boulders heave more easily than logs (into the ditch/*a
little ways).
a. Sarah herded/drove the cattle (to the stockyards/a little
ways).
b. Cattle herd/drive more easily than goats (to the stockyards/
*a little ways).
a. John pushed the cart (to Texas/a little ways).
b. The cart pushed easily (to Texas/*a little ways).
Verb-of-motion noun phrases with genitive subject arguments
should not be confused with noun phrase passives. Verbs of
motion that have true intransitive uses may appear to form
noun phrase passives when in fact they do not. 44 is an
example:
44
a. The horse's galloping down the road woke up the farmer.
b. The horse gallops.
c. The soldier gallops the horse.
44a is acceptable as a sentence describing a non-delimited
event because it is derived from 44b rather than 44c. 'The
horse' in 44a is not a deep structure object (direct argument)
of 'gallop'. Non-delimited constructions like 44a are
acceptable when the arguments are external arguments (deep
structure subjects).
Verbs of motion in true noun phrase passives are variable in
- 104 -
acceptability. They improve somewhat if the situation
described by the verb is understood as delimited:
45
a. *The cart's pushing along the river slowed down the
infantry.
b. ??The cart's pushing over the hill slowed down the
infantry.
3.1.6 Affectedness as an aspectual property
Investigation of the lexical semantics of five different
verb classes demonstrating the syntactic behavior typical of
affectedness verbs has shown that the semantic characteristic
associated with this syntactic behavior is more adequately
defined in terms of aspectual delimitedness, than simply as
'affectedness'. These five verb classes share the
characteristic that they describe events in which the referent
of the direct object is capable of delimiting the event. Some
property of the direct object, central to the verb's meaning,
is designated as 'measuring out' the event. Whether the event
described has some extended durati-on or little or no duration
is irrelevant. The five verb classes are summarized in the
- 105 -
table below:
Verb class Example of verb : Property of direct
with direct object: object that delimits event
Verbs of creation eat an apple size or spatial extent
or consumption draw a circle
translate a pcem
Verbs with event perform a play spatial or temporal
objects play a sonata extent
Verbs of motion cross the desert spatial extent
with delimiting
paths
Verbs expressing ripen the fruit ripeness
physical change of tighten the cinch tightness
state
Verbs expressing bribe the offical bribed-ness
abstract change of
state
Achievement verbs explode the bomb exploded-ndss
Verbs of motion push the cart i. state of being in
motion
ii. finite distance
moved
- 106 -
3.2 Lexical entries for affectedness verbs
The special properties of affectedness verbs must be marked
in the lexicon. Aspectual information about the verb will be
one part of the verb's lexical entry. This aspectual
information takes the form of a function associated with the
verb, that maps the direct argument into a two part value: a
change on some scale derived from some property of the
object,1 and an interval of time. The function is
represented as follows:
46
Averb(direct argument) = (Ls, At)
The scale may be finite or non-finite. Whether the scale is
finite, non-finite, or may be either, depends on the verb's
aspectual function. The scale associated with affdctedness
verbs may be finite; those associated with non-affectedness
verbs may not be finite.2  The finite scale is selected when
1. How that scale is derived may require a disjunct statement,
depending on whether the spatial extent of an object is to be
considered a property of that object in the same sense that
ripeness or wide-ness is a property of an object.
2. Within the class of affectedness verbs there may be
variation as to whether the scales are always finite, or have
a finite/non-finite option. Affectedness verbs are
distinguished because the AS associated with the aspectual
function is finite. Change of state verbs like 'ripen' or
widen', whose AS is determined by the change in some
property of the object, may be associated with scales that are
inherently finite, because the property names a final state.
- 107 -
(i) the verb has that option and (ii) the direct argument is
spatially delimited. (This will be discussed further in
section 3.3.1.) The finiteness or non-finiteness of the scale
will be represented by the following annotations: Af inS for a
finite scale and AinS for a non-finite scale. If Ans is
finite, At will also be finite.
47 below represents the aspectual functions associated with
the verbs 'eat', 'ripen' and-'explode', in the verb phrases
'eat an apple', 'ripen a fruit', and %explode a bomb':
However, verbs whose LS is determined by a change in the
spatial extent of the object seem to have both delimited and
non-delimited readings. For example, in (a) below 'climb the
beanstalk' describes a delimited event, while in (b) it
describes a non-delimited evr t:
a. Jack will climb the beanstalk in a !ay..
b. Jack will climb the beanstalk forever.
However, in environments where delimitedness is required, such
as NP passivization, the reading in which the event goes on
forever is not available:
??Jack's climbing of the beanstalk forever will be an amazing
sight.
Jack's climbing of the beanstalk in a day will be an amazing
sight.
If these differences between verbs like 'ripen' and verbs like
'climb' are real, then the aspectual- function of a verb must
be represented with a notation determining whether the scale
must be finite, must be non-finite, or may be finite or
non-finite. Verbs of change of state belong in the first
class, verbs of change of location in the second class, and
verbs of creation and consumption in the third class. This is
an issue that requires further research.
- 108 -
47
Aeat(an apple) = (/fin spatial extent of-apple, A1 t)
Aripen (a fruit) = ( A fin ripeness of fruit, a t)
Aexplode(a bomb) = ( Aft n explodedness of bomb, A t).
The size of the interval Ait will vary depending on the verb.
An accomplishment verb will produce a value for At that is an
interval of some duration, but an achievement verb will have a
At that is a very small interval, or perhaps a single point
of time.3 If Ait is a single point of time, A S will be
collapsed into a point, rather than 'spread out' over some
interval. Consequently, evidence for LS will be harder to
find for these verbs.4
The property that is associated by a verb with change over
time also depends on the verb. The verb specifies some
property central to its meaning, by which its direct argument
1measures out' the event. In some cases the property depends
jointly on the verb and its direct argument. The verb 'watch'
with a direct argument that represents an event uses the
3. I am bypassing the question of whether time is represented
in natural language as a succession of points or of intervals.
4. One important feature about AS has been presented as being
the fact that it can be described as a gradient. I leave it
as an open question as to whether there are degenerate cases
of verbs that collapse As into a single point of change.
Verbs with sources or goals as external arguments are possible
representatives of such a class. (More on this in section
4.3.1.) Even in these cases another important feature of ASS
is maintained: the change is describable as a change in a
single parameter.
- 109 -
temporal extent of the argument, as in 'watch a movte'. If
the direct argument is not an 'event object' the argument is
unaffected and the event is non-delimited, as in "watch a
bird'.
This property has a very important characteristic; it is
measurable on a scale. This is clearly true of a property
such as spatial extent, which is indicated by verbs of
consumption or creation; and the property of distance moved,
which may be specified by verbs of motion. However it is also
true of the properties associated with verbs expressing
changes of state, and achievement verbs. If some property of
the direct argument of a verb-expressing a change of state is
-measuring out' the event like a scale, it must be a property
that may be measured in degrees. A scale is made up of
discrete unit parts that are equal in value to one another,
and which can be ordered with respect to each other. The
interpretation of some adverbs and comparatives depends on
these characteristics of a scale. Adverbial phrases with a
comparative may refer to increments of these properties:
48
This banana ripens more every day.
Open the door a little more.
Unfold the paper a little wider.
Don't tan your skin so much.
Let the papaya ripen a little more.
Bend the bow a little more.
- Tighten the cinch a little more.
Loosen your belt a little.
Lengthen the rope a little mcre.
Shorten the rope a little.
Redden the image a little more.
-110-
Close the door a little more.
Because the property is measurable on a scale, it may be
represented as not quite achieved:
49
This apple is not quite ripe.
This branch is not quite broken. (I can't pull it off yet.)
The barn is almost collapsed, but still standing.
A partially exploded bomb was found in the train station.
Time, in the way we perceive it, also has a natural scale. It
can be divided up into unit intervals, and its directionality
allows the intervals to be ordered with respect to one
another. The complex value yielded by the aspectual function
is therefore an association of two scales. I suggest that
this value associating the change in a property of the direct
argument with the interval of time over which that change
transpires, is the linguistic representation of an event.5
The unit event is defined linguistically through a verb and
its direct argument.
It is the association of the change in property with a
change in time that is modified by rate adverbials:6
50
Greenhouse managers in Wisconsin ripen bananas slowly.
A bomb will explode slowly on slow-motion film.
5. Kahn(1981) points out that Aristotle had a similar notion
of events as change over time.
6. Richard Larson brought my attention to the relevance of
rate adverbials for this theory.
- 111 -
The composition of the aspectual function of a verb mirrors
the syntactic composition of a a verb with its direct
argument. Just as the verb takes its direct argument and maps
it into a verb phrase in the syntax, the verb's aspectual
function takes the direct argument and maps it into an event.
The verb phrase is the locus of the linguistic representation
of an event.7
3.3 Affectedness and the compositionality of delimitedness
3.3.1 The translation of spatial delimitedness into temporal
delimitedness
Affectedness verbs have the property that they describe an
event in which the direct argument of the verb is capable of
delimiting the event. The delimitedness of an event described
by an affectedness verb and its direct argument depends not
only on the lexical semantics of the verb, but on certain
properties of the direct argument of the verb. In this sense
delimitedness is compositional. Affectedness verbs have the
special .property that they allow the count/mass distinction in
the direct argument to be translated into a
delimited/non-delimited distinction in the verb phrase or
7. This view suggests that if there is an event position in
the argument structure of a verb, as suggested in Davidson
(1966) and Higginbotham (1985, 1986), it is associated with
the VP rather than with the verb.
- 112 -
sentence. The d'flimitedness of the events described by the
sentences below depends on the direct objects:
51
a. Dustin ate an apple (?for an hour/in an hour). (delimited)
b. Dustin ate snow (for an hour/??in an hour). (non-delimited)
The difference between 51a and 51b is due to the object of the
verb. If Dustin eats an apple, he finishes eating when the
apple is gone, but if he eats snow, he continues eating for an
indefinite period of time. He may even eat it forever if he
is in a world that never runs out of snow. The property of
noun phrases that contributes to the delimitedness of a verb
phrase is that of being spatially delimited. The noun phrase
an apple' is a count term, and therefore refers to something
that is spatially delimited. The noun phrase 'snow' does
not. An object or material that is spatially delimited has
some fixed quantity. Count nouns ('apple') describe things
that have clear boundaries and can be counted. Mass nouns
('snow') describe things that are undefined in extent. They
must be measured rather than counted. Bare plurals are the
same as mass nouns in this respect:8
52
Dustin ate apples (for an hour/??in an hour). (non-delimited)
Affectedness verbs are capable of translating the spatial
8. Also see Carlson (1977) for discussion.
- 113 -
delimitedness associated with the noun phrase that is their
direct argument into the temporal delimitedness associated
with the verb phrase. Spatial and temporal delimitedness are
parallel in many respects. A temporally delimited event has a
fixed duration, even if that duration is unknown. A spatially
delimited object or material has some fixed extent in space,
even if that extent is unknown. In the sentence 'Dustin will
eat an apple' the spatial delimitedness of the single apple
translates into the temporal delimitedness of an event of
eating a single apple.
The fact that affectedness verbs are capable of translating
the spatial delimitedness of their direct arguments into the
temporal delimitedness of the verb phrase distinguishes them
from verbs that do not have affected arguments. The verbs in
53 below are affectedness verbs. The events they describe are
temporally delimited because the direct objects are spatially
delimited:
53
a. eat an apple once (delimited)
b. eat one apple (delimited)
a. draw a circle once (delimited)
. b. draw one circle (delimited)
a. ripen the fruit once (delimited)
b. ripen one fruit (delimited)
a. perform the play once (delimited)
b.. perform one play (delimited)
a. break the window once delimited)
b. break one window delimited)
- 114 -
a. bribe the official once (delimited)
b. bribe one official (delimited)
a. open the door once (delimited)
b. open one door (delimited)
In the a) examples the event described by the verb phrase is
quantified with the frequency adverbial 'once'. This makes
the verb phrase event clearly delimited, and provides a
standard against which to compare the b) examples. In the b)
examples, the numerical quantifier 'one' is responsible for
the spatial delimitedness of the direct object noun phrase.
The a) and b) examples together form minimal pairs in which
the quantification is either over the verb phrase event or
over the noun phrase object. Although the a) and b) examples
are not necessarily paraphrases of each other, they are
equally clearly delimited.
Investigation of three classes of verbs with unaffected
objects demonstrates that they do not share this property of
translating spatial delimitedness into temporal
delimitedness. These three classes fall out from considering
separately the properties of the verb's internal and external
arguments. The crucial parameter is whether there is change,
motion or activity in one or both of these arguments during
the course of the situation or happening described by the
verb. For clarity of exposition these classes are discussed
under I through III in the following paragraphs.
I. Consider first the stative verbs. They describe
- 115 -
aituations in which no change, motion or activity occurs in
either the internal or external argument, while the situation
described by the verb lasts. Stative verbs describe
-situations in which there is no motion, change or activity at
all. This is the reason why statives only occur in
non-delimited verb phrases:
54
like mathematics non-delimited)
resemble your cousin non-delimited)
expect an arms shipment non-delimited)
see the forest (non-delimited)
be red (non-delimited)
The situations described by the verbs in 54 are
non-delimited. Applying a frequency adverbial like 'once'
makes the expression odd, unless 'once' is understood in a
different sense from that in the preceding examples under 53.
The salient reading of 'once' in the examples below is 'at
some time in the past'. The expressions are odd if 'once' is
interpreted in the sense of 'one event of _ing'.
55
?like mathematics once
?resemble your cousin once
?expect an arms shipment once
Where 'once' can be interpreted to mean 'one event of _ing',
it does not force a delimited reading of the expression:9
9. This is even clearer with the frequency adverbial 'twice',
which avoids the ambiguity of 'once': Tsee the forest twice' ,
'be red twice'
- 116 -
56
see the forest once (non-delimited)
be red once (non-delimited)
Quantifying the noun phrase object with the specifier 'one'
does not make the expression delimited:
57
like one branch of mathematics (non-delimited)
resemble one cousin (non-delimited)
expect one arms shipment (non-delimited)
see one forest (non-delimited)
II. The second class of verbs with unaffected direct
arguments are verbs that describe situations in which there is
change, motion or activity on the part of the external
argument but not the internal argument; and the action
performed by the external argument does not progress through
the spatial extent of the internal argument as it does with
verbs such as 'eat' or 'perform'. These verbs, like the
statives, occur only in non-delimited expressions. (There are
exceptions of a very particalar nature, discussed under III
below.)
58
watch a peregrine falcon (non-delimited)
remember the past (non-delimited)
observe the magnetometer (non-delimited)
Neither quantifying over the verb phrase event with 'once' or
over the noun phrase object with 'one' makes the expression
delimited:
- 117 
-
59
watch a peregrine falcon once
remember the past once
observe the magnetometer once
60
watch one peregrine falcon
remember one thing in the past
observe one magnetometer
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
III. There are some verbs that have unaffected arguments
but can appear in verb phrases expressing delimited events.
The adverbial 'once' selects the delimited reading ofthe verb
phrase. However, the quantifier 'one' applied to the object
noun phrase does not select the delimited reading:
tap the wall once
tap one wall
hit the wall once
hit one wall
beat the wall once
beat one wall
wave your hand once
wave one hand
poke the cat once
poke one cat
kick the door once
kick one door
(delimited)
(delimited,
(delimited)
delimited,
(delimited)
(delimited,
(delimited)
(delimited,
(delimited)
(delimited,
(delimited)
(delimited,
non-delimited)
non-delimited)
non-delimited)
non-delimited)
non-delimited)
non-delimited)
For the verbs in 61 above, quantifying over the verb phrase
event results in a clearly delimited expression, but
quantifying over the noun phrase object does not. These
expressions are delimited by an inherent subdivision of the
action of the external argument into units. These verbs
readily form corresponding nominals describing events (a wave,
- 118 -
61
a.
b.
a.
b.
a.
b.
a.
b.
a.
b.
a.
b.
a kick, a tap, a hit, a jump, etc.). The events they describe
are not delimited by the internal argument, and they may
easily be understood as repetitive. This means the verbs are
inherently ambiguous between a delimited and non-delimited
interpretation. (This is particularly clear with 'tap',
'beat' and 'wave' in 61 above.) When the event described by
the verb is delimited by the internal argument, the repetitive
reading is not the salient one.
This is not to say that a verb with an affected argument may
not express a repetitive event.'0  A repetitive interpretation
of the event described is possible with the proper frequency
adverbials. All the expressidns in 62 and 63 below, which
have affected and unaffected objects respectively, express
non-delimited, repetitive events:
62
draw one circle again and again (non-delimited)
open one door many times (non-delimited)
63
hit one wall again and again (non-delimited)
wave one hand many times (non-delimited)
However, when there are no frequency adverbials in the verb
phrase to force a repetitive reading, only the verb phrases
with affected objects can translate the spatial delimitedness
of the direct object into the temporal delimitedness of the
10. The term 'event' is used here to apply to many repetitions
of the same event, as well as a single, non-repeated event.
- 119 -
event. 1
1 12
3.3.2 Parallels between the count/mass and
delimited/non-delimited distinctions
The parallelism between the count/mass and the
delimited/non-delimited distinctions is brought out by certain
11. The observant reader will have noticed a gap in the
paradigm of possible verb classes. There are no verbs in
which there is motion, change or activity in the internal
argument, but not in the external argument. This would be,
for example, a hypothetical verb 'see-run' that could express
what the two verbs together express in 'see John run', where
'see' is non-volitional and non-agentive. I do not believe
this is an accidental gap, but as yet I have no explanation
for it.
12. Affectedness verbs are distinguished from other verbs by
the fact that they are capable of translating the spatial
delimitedness of their direct arguments into the temporal
delimitedness of the event they describe. However, there is
an open question as to whether they must necessarily do so.
This is because of the possible ambiguity introduced by the
repetition of events. If the property designated by the
affectedness verb is that of spatial extent, the correlation
between count/mass and delimited/non-delimitedness will be
direct. 'Eat snow' can refer to eating a definite amount of
snow an indefinite number of times, or eating an indefinite
amount of snow at one sitting. Only in the second reading is
spatial extent actually translated directly into temporal
delimitedness, but both express non-delimited events. If the
property invoked is something else besides spatial extent,
only in the first of these interpretations must spatial
delimitedness be translated into temporal delimitedness..
'Redden snow' can refer to reddening a definite quantity of
snow an indefinite number of times, or reddening an indefinite
quantity of snow once. In the latter case, if the reddening
does not progress from one 'end' of the snow to the other
(assuming we have some snow with 'ends'), but affects the snow
equally all over, the event will be delimited. This would be
an instance of the combination of a mass direct argument with
an affectedness verb to yield a delimited event. I am not
clear on the validity of these judgments. This issue needs
further research.
- 120 -
tests. There are many tests for distinguishing count and mass
nouns. Hinrichs (1985) lists the following four:
i) Co-occurence restrictions on quantifiers: Different sets
of quantifiers are associated with count and mass nouns.
Quantifiers like *each', "every', and cardinal numbers may
occur only with count nouns, and words like "little' and
'much' with mass nouns:13
64
each apple *each snow
every apple *every snow
three apples *three snows
*little apple little snow
*much apple much snow
Likewise, the modifiers of degree discussed in section 3.1
co-occur either with sentences describing delimited events or
with sentences describing non-delimited events. The examples
below are repeated from section 3.1:
64
a. ?*The cinch tightens a lot/quite a bit, easily.
b. The cinch tightens completely, easily.14
13. This is not to say that mass nouns may not be
reinterpreted as count nouns and vice versa. If snow is
understood to mean 'a snowfall' one may say, "We've had three
snows in January". And if apple is reinterpreted to mean
essence of apple', it may be said, "There's too much apple in
this fruit flavor".
14. A clearer example of a middle was suggested by Ken Hale:
%The cinch won't tighten completly.'
- 121 -
The sentences in 64 above are middles, and so must have
delimited interpretations. The expressions 'a lot' and 'quite
a bit', which co-occur with sentences describing non-delimited
events, are unacceptable in 64a; while 'completely', which
co-occurs with sentences describing delimited events, is
acceptable in 64b.
ii) Singular/plural modification: Count nouns may be
singular or plural, whereas mass nouns only occur in the
singular:
64
apple snow
apples *snows
Recall that a durative adverbial forces an iterative reading
only on expressions describing delimited events, not
expressions describing non-delimited events. The iterative
interpretation is, in a sense, a plural for events.
The next two properties of mass terms having to do with
closure of reference have been applied to the
delimited/non-delimited distinction by several authors.
iii) Cumulative reference: The sum of two denotations of a
mass term is in the denotation of that term. This is not true
for count nouns. If I have snow in my left hand and snow in
my right hand, and I put it all in my left hand, I still have
snow in my left hand. But if I take an apple in each hand and
transfer them to one hand, I do not have an apple in that
- 122 -
hand; I have two apples.
iv) Distributive reference: Dividing a denotation of a mass
term into two parts yields two new denotations of that term.
It is not so with count nouns. If I have snow in my right
hand, and put half of it into my left hand, I still have snow
in each hand. But if I take an apple in one hand, divide it
in two and take one part of it in each hand, I have not an
apple but a half an apple in each hand.
The parallel between the count/mass distinction for nouns
and noun phrases and the aspectual delimited/non-delimited
distinction for verbs and verb phrases has been noted by
several authors, among them, Taylor (1977), Mourelatos (1981)
and Hinrichs (1985). Taylor's treatment of this parallel,
which makes use of the property zf distributive reference,
will be summarized here.
In addressing Aristotle's trichotomy of state verbs
(statives), energeia verbs (activities) and kinesis verbs
(accomplishments'and achievements), Taylor equates the
energeia verbs with 'stuffs' like gold, and the kinesis verbs
with 'substances' like tables. (Taylor (1977) pp. 210-211):
...in general no (three-dimensional or other) space
within a table-is occupied by a table, whereas ever
three-dimensional area within a homogenous lump of gold is
* occupied-by a lump of gold. For a form of words to
encapsulate the difference, we may say that a homogenous
stuff fills, whereas a substance delimits, the space
it occupies.
- 123 -
Now, taking 'falls' and 'stabs' as paradigms respectively
of E- (energeia) and K- (kinesis) verbs, the views of the
last section can be summed up as the theses that falling
fills time as a homogenous stuff fills space, whereas
.stabbiig~elimits time as a substance delimits space.
.0..just as in general no spatial area within a table is
itself an area occupied~by a table, so the last section holds
that no period within a period of stabbing is itself such a
period; and just as every three-dimensional spatial area
within a lump of homogenous gold is itself such a lump of
gold, so every period within a period of falling is itself a
period of falling.
Taylor goes on to further divide 'homogenous stuff' into
homogenous and heterogenous stuff. The heterogenous stuffs
contrast with the homogenous stuffs in having a lower limit to
the applicability of the principle of distributive reference.
Thus a fruitcake is a heterogenous stuff, because division of
a fruitcake produces more fruitcake, but continued division
will in the end produce a lump of walnut or cherry, instead of
fruitcake. The temporal analogues of stuff -- the energeia
verbs -- are refined by Taylor to include this bifurcation.
Thus an arbitrarily small event of falling is still an event
of falling, but an arbitrarily small event of walking -- say,
lifting the heel off the ground -- is not an event of
walking. A final refinement equates the kinesis verbs with
heterogenous stuffs.15
The parallel between count/mass and
delimited/non-delimted-ness is manifested most clearly by
affectedness verbs, which translate one into the other.
15. For further discussion see Taylor (1977).
- 124 -
3.3.3 Some semantic approaches to the compositionality of
delimitedness
The translation of count/mass into
delimited/non-delimitedness has attracted attention in the
formal semantics literature. Some of the pieces of the
picture assembled here have direct or indirect counterparts in
the systems employed by semanticists who have considered this
phenomenon. I will discuss in general terms the basic
appropaches of Verkuyl, Hinrichs, and Dowty to this
phenomenon, and assess them in light of the discussion in-
sections 3.1 and 3.2. I will not go into detail about their
systems. For more rigorous and detailed exposition of these
ideas see Verkuyl (1985, 1987), Hinrichs (1985) and Dowty
(1979).
(ij Verkuyl
In his work on aspect, Verkuyl has taken certain insights
into aspectual phenomena and formalized them in a variety of
systems.16 I will not comment here on Verkuyl's
formalizations of his ideas, but will confine myself to
examining the ideas themselves, as they are put forth in his
most recent works. (Verkuyl (1985) and (1987)).
16. Yerkuyl (1972) appeals to phrase structure rule systems;
Verkyul (1978) to ideas from generative semantics; Verkuyl
(1987a) and (1987b) to set-theoretic semantic systems.
- 125 a
Verkuyl identifies those verbs that can translate spatial
delimitedness into temporal delimitedness as having a feature
[+ADD TO]. These are verbs expressing 'progress in time' or
'change'. The feature [+ADD TO] is associated with a
successor functign' that, loosely speaking, adds intervals of
time (in a model) to the intervals over which the sentence is
already true. Thus if a verb has the feature [+ADD TO] and
the sentence containing the verb is true of an interval from
time (a) to time (b), then the sentence is also true over the
interval from time (a) to time (c) when (c) is a point of time
after.(b). Spatial delimitedness of the object is represented
in Verkuyl's system as the feature [+SQA] (specified quantity
of A]. The presence of the feature [+SQA] results in the the
successor function's output being abridged. That is, when the
NP has the feature [+SQA] the successor function generates a
finite rather than an infinite number of successive time
intervals. The combination of a verb having the [+ADD TO]
feature with a noun phrase having the [+SQA] feature results
in the sentence containing the VP being true over a finite
interval of time. In this way spatial delimitedness is
translated into temporal delimitedness in Verkuyl's system.
Verkuyl has correctly identified the relevant
characteristics of verbs for this phenomenon, as being
something like change or progress in time. He has also
observed, correctly, that the semantic/aspectual interaction
between a verb and its direct argument is a special one. His
- 126 -
successor function, based on adding equal intervals to a time
line, has the important features of a scale, on which the
event is, in a sense, "measured out'. However, since Verkuyl
does not take note of the importance of the property of the
argument by which the event is 'measured out', his approach is
limited to verbs that 'measure out' the event through the
spatial delimitedness of the object (or the Specified Quantity
of A). A verb such as "ripen' which depends on the property of
ripeness instead of on spatial delimitedness for measuring out
the event, will not be subsumed in his system. Lacking a
reference to such a property the [+ADD TO] feature of
affectedness verbs is simply a diacritic, and does not explain
how these verbs characterize events as progressing in time
through the change in their internal arguments.
(ii) Hinrichs (1985)
Hinrichs (1985) provides a formal semantic approach to
Aktionsarten in English, drawing on work by Carlson (1977) and
cast in Montague Grammar. I will address only his treatment
of verbs which translate the count/mass distinction of their
direct objects into a delimited/non-delimited distinction in
the verb phrase. I will sketch out the major points of this
treatment.
The delimitedness of accomplishments and the
non-delimitedness of activities is treated by Hinrichs as
deriving from a 'heterogenous reference property' and a
- 127 -
'homogenous reference property' associated with
accomplishments and activities respectively. The 'homogenous
reference property' of an activity predicate is parallel to
the cumulative reference property of mass nouns. It ensures
that the temporal sum of two events of that activity is also
an event of the same kind; if there is an event of John's
sleeping from 5:00 to 6:00 and an event of John's sleeping
from 6:00 to 7:00, then there is an event of John's sleeping
from 5:00 to 7:00. The heterogenous reference property of
accomplishment predicates parallels the lack of distributive
reference for count terms. If there is an event of a pill's
dissolving from 10:30 to 11:00, then there is no event of the
pill's dissolving that occupies a temporal interval properly
included within the interval 10:30 to 11:00. (e.g. There is
no event of a pill's dissolving from 10:45 to 10:50.) These
requircuents of heterogenous and homogenous reference are laid
out in meaning postulates, which must be satisfied by
predicates of the appropriate type.
Hinrichs employs Carlson's idea of 'stages' realizing an
individual at a particular time and place, and extends this to
the idea of 'event stages', realizing an event at a particular
spatio-temporal location. Meaning postulates address (among
other things) the relationship between event stages realizing
an event of, for instance, 'building a house' and individual
stages realizing the individual participants in the
house-building at particular spatio-temporal location.
- 128 a
Accomplishment sentences like,
66
John built a house.
John ate a cake.
are accomplishments in Hinrich's system because the meaning
postulates associated with 'build' and 'eat' result in the
imposition of a uniqueness condition on stages of the direct
object 'house' or 'cake'. There may be no stage realizing
"house' or 'cake' which is associated with any substage
(temporal subpart) of the event stage realizing 'building a
house' or "eating a cake'. Such a substage would actually be
an event of building part of a house, which is not the same
kind of event as 'building a house'. For Hinrichs (following
Carlson) mass terms and bare plurals are linked with 'kinds'
as opposed to objects, and this allows them to escape the
uniqueness condition. If John eats cake for an hour, there
will be a smaller interval contained within that hour in which
John also eats cake.
In brief, Hinrich's approach relies on the homogenous or
heterogenous reference properties of count or mass terms,
together with meaning postulates associated with individual
predicates, to capture how affectedness verbs translate the
count/mass distinction of their direct objects into the
delimited/non-delimited distinction of the verb phrase. The
idea of the argument 'measuring out' the event is incipient in
Hinrich's system, although he doeE not make it explicit.' The
- 129 -
association of stages of objects with substages of events
approaches this idea. However, like Verkuyl, Hinrichs does
not focus on the property by which the direct object measures
out the event. His approach handles 'effected objects' (the
objects of verbs of creation and consumption) smoothly, but by
his own admission, it falls short in the treatment of what he
calls 'affected objects' (the objects of verbs expressing a
change of state), for which the relevant property is something
other than spatial extent. It is necessary to itefer to this
property in order to integrate these two classes of verbs.
For this reason Hinrichs fails to unify these verbs within a
general aspectual theory of affectedness.
(iii) Dowty (1979)
Dowty (1979) provides a model-theoretic treatment of why a
'BECOME'-type verb with a plural indefinite (bare plural) or
mass noun argument is aspectually durative, or non-delimited.
('BECOME' type verbs in Dowty's classification are delimiting
verbs. See appendix to Chapter 1.) His treatment gives an
account of why such a sentence can co-occur with a 'for'
durative adverbial. The BECOME operator contained in the
verb, when it is applied to the proposition p, makes the
sentence express a change of state:
67
BECOME p is true at t iff p is true at t and false at t-1.
The sentence,
-130-
68
John discovered the buried treasure in his back yard for
six weeks.
contains a BECOME-type verb and a countable argument and is
ungrammatical according to Dowty. Dowty translates the
durative adverbial 'for six weeks' as a quantified temporal
expression with a two-place AT operator, i.e. 'for all times
t such that t is a member of the period six weeks, it was true
at t that p'. Sentence 68 is translated as:
69
(At:tf.six weeks) AT(t, BECOME[John knows that...])
A contradiction is derived in-applying this formula, because
the time t at which the BECOME clause is true can be any
arbitrary moment. If that moment is t1 , then t5 is the moment
of change -n the truth value of the proposition embedded under
BECOME. But that moment could just as easily be chosen to be
t2, in which case t2 would mark the instant of change in the
truth value. Clearly there can only be one time t at which
the truth of the proposition changes from false to true.
Sentence 68 is therefore ungrammatical because the
interpretation would require that John discover the same thing
in his back yard over and over again for six weeks.
However the same type of sentence can have that
interpretation if the object of the verb is a plural
indefinite or mass noun:
- 131 -
70 1
a. John discovered fleas on his dog for six weeks.
b. John discovered crabgrass in his yard for six weeks.
Sentence 70b may be paraphrased as in 71a but not 71b:
71
a. For six weeks John discovered there to be some x such that
x is crabgrass and is in his yard.
b. *There ia some x such that x is crabgrass and for six
weeks John discovered x to be in his yard.
Dowty (following Carlson(1977)) assumes that indefinite
plurals and mass nouns have existential quantifiers within the
scope of the temporal quantifier A logical representation for
71 would be:
72
(At:t4Esix weeks) (Vx) [AT(t,BECOME[John knows that...])]
Since the value for x in this formula may change with the
value for t, a contradiction is avoided. The embedded clause
under the BECOME operator may become true at many different
times t, because the object discovered need not be the same at
each time t.
Dowty's treatment identifies a class of verbs -- those verbs
specifying the BECOME operator in their lexical entries -- as
describing delimited events when they have a definite direct
object, and non-delimited events when they have bare plurals
or mass nouns as direct objects. Insofar as Dowty's BECOME
class of verbs identifies the affectedness verbs, he has
- 132 -
identified a significant class. But s pointed out by
Hinrichs (1985) some verb phrase having 'event' objects like
'play a sonata' are not included in Dotwy's BECOME class.
Dowty comes closer than Verkuyl or Hinrichs to recognizing
the importance of the property of the object 'measuring out'
the event. The proposition p embedded under the BECOME
operator approximdtes the idea of a property changing over
time (if p maintains that some particular state obtains of the
object). However, Dowty does not characterize the event
described by a BECOME verb as being capable of being measured
out on a scale. In other words, p is only true or false;
there are no degrees of 'p-ness'. -Without degrees of 'p-ness'
the function of the adverbials and comparatives discussed in
section 3.3.2 cannot be captured.
There is a further problem with the analyses of Verkuyl,
Hinrichs and Dowty. They do not distinguish between bare
plurals and mass terms. While it is true that these may be
grouped together semantically with respect to direct (or
internal) arguments and delimitedness, they diverge in their
semantic behavior with respect to external arguments and
delimitedness. (This will be discussed in -Chapter 4.)
Furthermore, bare plural direct arguments need not always
create a non-delimited verb phrase event. The sentence 'John
opened doors' may represent an event in which all the doors
John opened, he opened at once by pushing a button. Such an
- 133 -
event is delimited even though it involves an indefinite
number of doors. There is an important difference between the
non-delimitedness that results from the mass-ness of direct
arguments and that which results from bare plural direct
arguments, which must be captured in a viable theory of the
compositionality of delimitedness.
These three authors who have addressed the compositionality
of delimitedness from a formal semantic viewpoint have each
proposed some elements of the aspectual function put forth in
section 3.2. The idea of an event as measurable over
cumulative intervals of time (as on a time scale), has been
capitalized on by Verkuyl and Hinrichs. 1 This idea is
incipient in the recognition that the properties of cumulative
and distributive reference apply to the
delimited/non-delimited distinction. The contribution of the
direct argument to 'measuring out' the event is suggested by
Hinrichs' association of stages of objects with subparts of
events. The importance of the delimiting change in the object
that takes place during the event is recognized by Dowty.
The contributions of the formal semantics literature to this
problem, while important, are limited because they have
addressed the phenomenon without making reference to the
17. The idea that sentences are true at intervals rather than
at points of time is also relevant here. Bennett and Partee
(1978) present this view.
- 134 -
syntactically interesting class of affectednese verbs, or
taking note of the fact that the aspectual property of
delimitedness has syntactic ramifications. On the basis of
syntactic behavior, an important class of verbs is identified,
which unifies verbs with 'affected', 'effected' and 'event'
arguments, and whose direct arguments may be objects or
subjects (in the case of unaccusatives). The syntactic
evidence draws attention to the semantic properties these
verbs share. The definition of this verb class in formal
semantic terms should be tailored specifically to single out
the syntactically interesting class of-affectedness verbs.
Verb classes which have both syntactic and semantic reality
are grammatically significant because they provide insight
into the interface of the two systems. Furthermore, when the
class of affectedness verbs is defined as a semantic as well
as a syntactic class, it leads to a statempnt that the direct
argument 'measures out' the event through the change over time
in a particular property -- a view which turns out to
encompass more than just affectedness verbs. This chapter and
later chapters present several reasons for adopting this
idea. This idea is implicit in some of the work of these
authors, but nowhere, to the best of my knowledge, is it
explicitly stated. I believe it is an idea that could be
translated into formal semantic language, and which would
contribute to formal semantic theory (e.g. by providing a
means of constraining meaning postulates associated with
a 135 -
particular verbs'.)
In conclusion, some of the ideas developed in this chapter
are mirrored in the work of Verkuyl, Hinrichs and Dowty. The
contribution of these authors is limited by their failure to
connect semantic theory with syntactic theory. Some of the
tools developed by these authors could be used with efficacy
to address semantic phenomena as defined through syntactic
means.
- 136 -
3.4 Total affectedness
The delimited interpretation of verb phrases or sentences
containing affectedness verbs depends on an interpretation in
which the object is totally affected; i.e., entirely consumed
or created (or changed). This is the salient interpretation
for these expressions in English. However, some speakers find
these expressions ambiguous between a delimited and
non-delimited reading. For these speakers the expression 'eat
an apple' can be paraphrased as 'eat an apple up' (delimited
and totally affected) or 'eat at an apple' (non-delimited and
not totally affected). 'Read a book' can mean either 'read a
book through' or 'read in a book'. Other examples of this
kind of ambiguity are listed below. The b) and c) expressions
are delimited and non-delimited paraphrases, respectively, of
the ambiguous a) expressions:
73
a. climb the mountain (ambiguous)
b. climb (all the way up) the mountain (delimited)
c. climb (around on) the mountain (non-delimited)
a. perform the play (ambiguous)
1* What verbs will allow this ambiguity seems to be dependent
on a highly individual property of the particular verbs.
Verbs that allow this ambiguity seem to be verbs in which the
spatial extent of the direct argument measures out the event.
In the author's judgement some verbs that do not allow this
ambiguity, but only take the totally affected interpretation
are: record, cross, traverse, swim, run, walk, circumnavigate,
transpose, transfer, disperse, corral, break, explode,
persuade, confirm, bribe.
- 137 -
b.
c.
a.
b.
c.
a.
b.
a.
a.
b.
0.
a.
b.
0.
perform the play (from beginning to end)
perform (some of) the play
paint the wall
paint (all of) the wall
paint (some of) the wall
translate the treatise
translate (all of) the treatise
translate (some of) the treatise
cut the cake
cut the cake through
cut the cake a little bit
(make a cut in the cake)
rip the flag
rip the flag through
rip the flag a little bit
(make a rip in the flag)
For many speakers, spray/load verbs make the distinction
between totally affected and non-totally affected readings
especially clear. In the examples below, repeated from
section 2.5.3, the event may be understood as delimited
through a totally affected interpretation of the direct
argument, but not of the indirect argument. The salient
reading is that in which the direct argument is totally
affected and indirect arguments are not necessarily totally
affected*
74
a. Jeremiah sprayed the paint on the wall.
b. Jeremiah sprayed the wall with paint.
2. It is true that if one runs out of paint, while spraying
the wall with paint, or runs out of wall while spraying paint
on the wall (which is unlikely), the spraying-event must come
to an end, but this is not a fact about the lingusitic
representation of the event; it is a fact about the world.
- 138 -
(delimited)
(non-delimited)
(ambiguous)
(delimited)
(non-delimited)
ambiguous)
(delimited)
(non-delimited)
(ambiguous)
(delimited)
non-delimited)
(ambiguous)
(delimited)
(non-delimited)
a. Josiah cleared the dishes from the table.
b. Josiah cleared the table of dishes.
Under the view advanced in section 3.2, in which
affectedness verbs characterize events as measured out by some
property of the verb's direct argument, the phenomenon of
total Affectedness receives a natural explanation. Total
affectedness and affectedness (as it is defined here) are one
and the same thing. If a verb describes an event which
'travels through' the spatial extent of its argument, the
event is delimited by the argument's spatial extent; and all
subparts of the argument have participated in the event. A
delimited, or totally affected, interpretation results from
using the entire scale defined by means of the object's
spatial extension.3 (It is unclear what 'total affectedness'
3. This suggests a possible means of representing the
difference between the totally affected and the non-totally
affected interpretations in 74. If the totally affected
interpretation results from using the entire scale, and the
non-totally affected interpretation from using only some
indefinite part of the scale, the distinction is easily
represented as either universal or existential quantification
over the scale dependent on the object's spatial extent. If
this scale is represented by Xn (where the scale is graduated
into units: X1, X2...Xn), the two interpretations may be
represented as follows:
A(Xd) totally affected, delimited
E(X ) not totally affected, non-delimited
Delimitedness and total affectedness is morphologically marked
in many languages. This view suggests a line of research
investigating whether some or all of these delimitedness
- 139 -
means with respect to verbs whose direct arguments measure out
the event by some other means than spatial extension (e.g.
'ripeness').
The phenomenon of total affectedness in Hungarian provides
interesting indirect evidence for the idea that semantic
composition of a verb and its direct argument involves forming
a scale out of some property of the direct argument.
Hungarian is a language in which total affectedness is
morphologically marked by a system of optional preverbs or
perfective markers. The sentences below are Hungarian
equivalents of English spray-load verbs. There are no
aspectual preverbs, and the sentences do not have the sense
that the direct'arguments are totally affected. In 75a, the
fat is the direct argument, and in 75b thd bread is the direct
argument of the verb. Not all the fat is necessarily smeared
in 75a, or all the bread smeared in 75b.
75 (De Groot (1984))
a. Janos zsirt ken a kenyerre
John fat smears the bread-onto
"John smears fat on the bread"
b. Janos zsirral keni a kenyeret
John fat-with smears the bread
"John smears the bread with fat"
markers are quantifiers indicating total affectedness. The
Hungarian aspectual preverbs (discussed in the following
paragraphs), for instance, may be thought of as
grammaticalizations of the universal quantifier A which is
necessary for the interpretation of total affectedness, and
which is not overtly marked in English. I leave this as an
open question. That research will not be pursued here.
- 140 -
However, when the preverbs are applied, the direct arguments
are understood as totally affected:
76 , , ,
a. Janos ra-keni a zsirt a kenyerre
John onto-smears the fat the bread-onto
"John smears fat on the bread"
b. Janos zsirral meg-keni a kenyeret
John fat-with perf-smears the bread
"John smears the bread with fat"
Evidence for the object scale measuring out the event comes
from an interesting difference in meaning between 75a and 75b.
According to DeGroot (1984) (p. 144), 75a "means: John puts
some fat on the bread, but does not necessarily smear the fat
evenly over (a part of) the bread. 75b on the other hand, has
the assumption that John indeed smears the fat evenly over (a
part of) the bread." Even though total affectedness is not
required of these direct objects because there are no
aspectual preverbs, there is still something required of them
because they are direct objects. Where 'bread' is the direct
argument .in 75b, all areas of the bread that are smeared with
fat are equally smeared. That is, even though the direct
argument 'bread' is not entirely affected it must be evenly
affected.4 The direct argument status of 'bread' forces it to
4. This view predicts that, for those speakers of English who
accept (a) below there should be a difference in meaning
between (a) and ):
- 141 -
be interpreted as changing on a scale. Because a scale is
divisible into parts of equal value, the change described by
such a scale is divisible into equal changes. Therefore where
the bread is. the direct argument, its parts are equally
affected. This is predicted by a theory in which a scale
measuring out the event is constructed on a property of the
object.
3.5 Delimitedness and case
Affectedness, a semantic constraint on the formation of
middles and NP passives in English, is actually an aspectual
constraint dependent on delimitedness. We must examine the
question of what exactly is the connection between
delimitedness and syntax. Since middles and NP passives are
formed by NP movement, we must assume that the interesting
correlation is between delimitedness and NP movement.
However, the delimitedness constraint does not apply to all
instances of NP movement; it does not apply to verbal
passivization or raising, for example.5 In the case of
passivization and raising there is some syntactic element
a. John ate the bread (but didn't finish it.)
b. John ate at the bread.
This is a question for further research.
5. In Chapter 6 the possibility will be raised of a
delimitedness constraint on unaccusatives as well. This is an
open question.
- 142 -
present -- the passive morphology or a modal verb -- which may
account for the lack of any delimitedness constraint on these
operations.
In the Government and Binding framework, the case assigning
properties of the verb force NP movement. If a verb assigning
accusative case to its direct argument loses its case
assigning properties, that argument moves to the position of
specifier of IP, where it is assigned case by INFL/AGR. The
delimitedness constraint may be rephrased in terms of case
theory in the following way:
77
Delimitedness-Case Constraint
INFL/AGR cannot assign case to the direct argument of a
verb describing a non-delimited event, without the mediation
of some special element appearing in the syntax.
Note that the' Delimitedness-Case contraint requires that the
syntax, speciffcally case thery, be able to refer to a
semantic featural distinction of +/-delimitedness. The need
for syntax to refer to features is not surprising. Lumsden
(1987) argues for the importance of features in the syntax of
natural languages. The Delimitedness-Case Constraint is
proposed here as a tentative explanation of the link between
aspect and case. It requires further research, and further
- 143 -
conceptual integration into a theory of grammar.6
There is evidence of a connection between aspect and case.
Recall the discussion of Finnish in Chapter 2. In Finnish,
accusative and partitive case-marking directly encode the
delimitedness or non-delimitedness of the verb phrase.
Additional evidence comes from Hindi, in which a semantic
distinction between affected and unaffected agents is
reflected in case assignment. Saksena (1980) discusses
morphologically related causative/non-causative verb pairs in
Hindi. The subject/agent of one member of the pair is the
object/causee in the other member. The object causee position
receives different case marking, depending on whether it is an
affected or unaffected agent. (D/A = dative accusative case
marker; DC = direct causative suffix.)
78
a. raam-nee nahaa-yaa
Ram-AGT bathe-PAST
'Ram bathed.'
b. mai-nee raam-koo/*see nahal-aa-yaa.
I-AGT Ram-D/A bathe-DC-PAST
6. It has come to my attention that Rizzi (1986) argues that
in Italian, objects of affectedness verbs may not be null
objects with arbitrary interpretation. This is a phenomeonon
that would not be substtmed under the Delimitedness-Case
Constraint. Also, the movement of nominal quantifiers in
Japanese discussed in section 2.5.2 may be movement to topic
position, in which case the Delimitedness-Case Constraint does
not cover it. Fin4lly, there appear to be certain
delimitedness constraints on IF movement as well (Kelly Sloan
(in prep)). These cases may not be subsumable under the
Delimitedness-Case Constraint.
- 144 -
'I bathed Ram.'
79
a. raam-nee peer kaat-aa.
Ram-AGT tree cut-PAST
'Ram cut the tree.'
b. mal-nee raam-see/*koo peer kal-aa-yaa.
I-AGT Ram tree cut-DC-PAST(m.)
'I made Ram cut the tree.'
In the event described in 78 above Ram is an affected agent,
since he undergoes a change of state from being unbathed to
being bathed. In 79 'Ram'is an unaffected agent, since he
undergoes no change of state in the event of cutting described
by the verb 'peer'. In both of the a) sentences 'Ram' is
marked with the agent case marker Tnee'. In the b) sentences,
where Ram is the object causee, the case marking distinction
emerges. "Ram' is marked with the dative/accusative case
marker 'koo' where Ram is an affected agent (78), and with the
instrumental case marker 'see' where he is an unaffected agent
(79).
The case markers 'koo' and 'see' are in complementary
distribution. In the few cases where a verb may have its
argument marked with either case marker, a semantic
distinction accompanies the case marking difference. (IC =
indirect causative suffix.)
80
m&l-nee raam-see/koo kitaab parh-vaa-ii
I-AGT Ram book read-IC-PAST(f.)
'I had Ram read the book.'
If 'Ram' is marked with the dative/accusative casemarker
- 145 -
'koo', the sentence means that Ram himself was made to read
the book. If 'Ram' is marked with the instrumental case
marker 'see', it means that the book was made to be read, and
Ram was an instrument to that end.
The affectedness constraint is stated in 77 above in terms
of the interaction of aspectual delimitedness and case theory
-- both of which must be included in the grammar for
independent reasons. This is an improvement over stating it
in terms of the interaction of a semantic property somewhat
amorphously described as 'affectedness', with individual
constructions such as middles, NP passives, and so on. Stated
in this way, the affectedness constraint will contribute to
directing future research into the theory of case and aspect.
Nevertheless it leaves a lot to be desired in explanatory
power. In the opinion of this author, a deeper understanding
of this phenomenon must await a deeper understanding of case.
- 146 -
Chapter 4. Aspectual Principles of Argument Structure
4.1 Introduction
There are two facts discussed in the preceding chapters,
which taken together, point to a connection between
delimitedness and argument structure: the property of
delimitedness is the basis of the property of affectedness;
and affectedness is only associated with direct arguments.
This suggests a connection between delimitedness and direct
argument-hood. If we can understand this connection, then
delimitedness will become a useful tool for investigating
argument structure. This is the central undertaking of this
chapter.
A deeper investigation into the relationship between
delimitedness and internal, external and oblique arguments
reveals some basic principles of argument structure: Events
are linguistically described as delimited through the change
in the direct argument. Direct arguments, in a sense,
measure out' the event over time, in a way that can be
precisely characterized. If a direct argument undergoes
change during an event, that change can be characterized as
change on a scale or change in a single parameter. This is
not required of external arguments. Furthermore, it is the
scale provided by the direct argument through which events are
linguistically delimited. The endpoint of the event must be
marked on that scale by the change within the direct argument
- 147 -
(as in the case of affectedness verbs) or by an oblique
internal argument which indirectly contributes an endpoint on
the scale provided by the direct argument. Delimitedness is
encoded within the VP; the external argument is outside the
domain in which delimitedness is determined. Finally, there
may be only one 'delimiting' associated with a verb phrase.
These principles of argument structure will turn out to have
consequences for syntax, and for the mapping of cognitive
structure into syntactic structure.
The discussion in this chapter will proceed as follows.
Since affectedness verbs have been discussed in Chapter 3, I
will concentrate on non-affectedness verbs in Chapter 4. I
will discuss a range of verbs, and show that while change in a
direct argument is characterizable as 'measuring out' the
event on a scale (section 4.2), change in an external argument
does not have to be so characterizable (section 4.3). Section
4.4 will demonstrate that oblique arguments which
linguistically delimit an event, such as sources or goals, do
so parasitically by means of the scale provided by the direct
argument.
It is necessary to clarify how I am using certain
terminology. The distinction between external, internal
direct and internal oblique arguments I take to be an
essentially syntactic distinction; the direct internal
argument is governed by the verb at deep structure, the
- 148 -
MITLibraries
77 Massachusetts Avenue
Cambridge, MA 02139
http://Iibraries.mit.edu/ask
DISCLAIMER NOTICE
MISSING PAGE(S)
149
- verbs with unaffected direct arguments or no direct
arguments -- we find that the special semantic status of
direct arguments extends to these verbs as well. The change
undergone by the direct argument in the course of the event is
characterizable as the scale which 'measures out' the event.
It is that scale through which the event is delimited.
This section is organized as follows. First I will discuss
non-affectedness verbs which describe events in which the
direct argument undergoes change but does not delimit the
event. Just as with affectedness verbs, some property of the
direct argument of these verbs becomes a scale by which the
event is "measured out' (even-though the direct argument
itself does not delimit the event on that scale.) Next I will
discuss non-affectedness verbs that have no direct arguments
or have optional direct arguments, and show that the events
they describe are delimited only when they have- a direct
argument. Then I will discuss non-affectedness verbs which
describe events in which there is no change, motion or
activity on the part of the direct argument, and show that
they describe non-delimited events. I take these three groups
of verbs to be representative of non-affectedness verbs
generally. Finally, I will mention several cases of minimal
pairs in which the same event participant is either a direct
object or an indirect'object. Where it is the direct object
it 'measures out' and delimits the event. Where it is an
oblique object it does not.
- 150 -
4.2.1 Verbs of imparting motion
There is a group of verbs that describe events in which
there is change, motion or activity (progress in time) in the
internal argument during the course of the event, but that
change does not result in a change of state,. or a delimited
event. These are not affectedness verbs, but their internal
arguments do, in a sense, 'measure out' the event although
they do not delimit it. The main examples of this class are
verbs of imparting motion such as 'push', 'shove' or 'roll'. 2
They are usually not affectedness verbs, and when combined
with a direct object they do not describe delimited events:
I
James pushed the cart. (non-delimited)
James drove the truck. (non-delimited)
James rolled the log (down the hill). (non-delimited)
Verbs like these, in their salient readings, describe events
in which the direct argument undergoes a change in location.
Just as with the verbs of change of physical state described
in section 3.1.2, a property of the object is singled out by
the verb as the locus of change. That changing property of
the internal argument, location, is measurable on a scale.
This is what some rate adverbials refer to; they modify the
2. Recall from Chapter 3 that verbs of motion may be used
occasionally as change of state verbs. This is not the usage
discussed here. These verbs more commonly describe
non-delimited events, which is how they are used above.
- 151 -
change in location (or distance traveled) of the internal
argument over time:
2
push the oart slowly
drive the truck fast
roll the log down the hill slowly
That scale is also invoked by the comparative '-er', which
expresses the change in gradients or degrees:
3
push the cart further
drive the truck further
don't roll the log quite as far down the hill
Finally (anticipating section 4.4), additional delimiting
expressions like goal phrases may be added to the verb
phrase. These expressions delimit the event described by
marking a point on the scale determined by the changing
location of the direct argument:
4
push the cart to San Francisco (delimited)
drive the truck to town (delimited)
roll the log to the top of the hill (delimited)
These verbs exemplify a class of verbs in which some property
of the direct argument is specified as changing over time in a
scalar fashion, just as in the case of affectedness verbs.
However, that change is non-delimiting, since the property
that changes has no endpoint or final change of state. The
direct arguments of these verbs do not delimit the event, but
they provide the scale along which the events are
- 152 -
delimitable.3 These ;erbs would have an aspectual function
like the following: 4
5
A (verbdirect argument) = (An distance moved, At)
6
Apush(the cart) s (An distance traveled by cart, Lt)
4.2.2 Unergatives, reflexives and cognate objects
Intransitive unergative verbs, which have no internal
arguments at all, describe non-delimited events:3
7
Mary ran (for an hour/*in an hour).
Mary danced (for an hour/*in an hour).
Mary sans (for an hour/*in an hour).
Where they may be understood to describe delimited events,
3. Recall from the discussion in section 3.1.2 that verbs like
ripen', 'redden' and 'widen' may have:non-delimited readings
(which are not possible in middles or NP passives but are
possible in a regular transitive usage of the verb). If these
verbs in their nondelimited interpretations constitute a
class of verbs describing non-delimited events, they belong
with the verbs of imparting motion above, in which the direct
argument 'measures out' the event but'does not delimit it.
The proper treatment of this type of verb (or verb
interpretation) is left as an open question.
4. See section 3.2.
5. These verbs are acceptable with 'in an hour' if they are
used inceptively: 'She will run/dance/sing in a hour.' In
this case the verbs are used as change of state verbs. That
is not the reading discussed here.
" 153 -
they are understood as if they had a reflexive object, or a
cognate object. Expressions with reflexive verbs describe
delimited events because the action described by the verb
progresses through the reflexive object during the event. In
the examples below, the reflexive object delimits the event
and disambiguates the verb phrase:
8
a. bathe
b. bathe oneself
a. shave
b. shave himself
a. wash
b. wash yourself
(non-delimited, delimited)
.(delimited)
(non-delimited, delimited)
(delimited)
(non-delimited, delimited)
(delimited)
Cognate objects delimit the event described by a verb by
making reference to one event of that type. The presence of a
cognate object forces a delimited interpretation: 6
9
a. sneeze
b. sneeze a horrific sneeze
a. laugh
b. laugh a mirthless laugh
a. dance
b. dance a dance
non-delimited, delimited)
delimited)
(non-delimited, delimited)
(delimited)
(non-delimited)
(delimited)
6. Some speakers can get a non-delimited reading for the b)
sentences in 8 and 9 as well. These are also relatively
acceptable with durative adverbials. 'John shaved himself for
an hour', "Mary sang a song for an hour'. This interpretation
results when the action does not progress through the object
during the course of the event. In any case, the fact remains
that where a delimited reading is possible it depends on an
actual or an understood object "measuring out' the event.
- 154 -
a. sing (non-delimited)
b. sing a joyful song (delimited)
Object deletion verbs which describe non-delimited events
when they are without a direct object describe delimited
events when they have a spatially delimited direct object:
10
a. John smoked. (non-delimited, delimited)
b. John smoked a Canadian cigarrette. (delimited)
a. Mary drank. (non-delimited, delimited)
b. Mary drank a jug of apple wine. (delimited)
These events are delimited through their direct arguments in
the same way as events described by verbs of consumption and
creation; the action taken by-the external argument progresses
through the spatial (or temporal) extent of the internal
argument over time.
4.2.3 Verbs describng events with no change in the internal
argument
Chapter 3 and sections 4.2.1 and 4.2.2 have demonstrated
that events are linguistically described as delimited through
some change in the internal argument. What about verbs that
describe events in which there is no change in the internal
argument? These verbs describe non-delimited events or
situations.
7. I have not extended the idea of a scale dependent on the
direct argument, to verbs that describe situations or events
- 155 -
Verbs in this group may be divided into two classes,
depending on whether or not there is change, motion or
activity in the external argument during the course of the
event described. Verbs that describe events in which there is
change, motion or activity on the part of the external
argument but not the internal argument include perception
verbs ('watch', 'observe'), verbs of contact ('tap', 'hit',
'beat', 'wave', 'poke', 'kick'), and miscellaneous other verbs
('kiss', 'remember').
11
observe the magnetometer (for an hour/*in an hour)
watch a moose (for an hour/*in an hour)
remember a dream (for ankiour/?in an hour)8
Verbs of contact in their salient readings describe
non-delimited events:9
in..which there is no change, motion or activity on the part of
.the internal' argument. However, something like this may be
appropriate, because the direct arguments of these verbs also
seem.to have a priviledged semantic status. To test this idea
it would be necessary to find some properties of a scale that
can be manifested without progress in time. Something like
this may account for the notion of "affected" discussed in
section 2.5.4. I will not pursue this here.
8. 'Remember' occasionally has a delimited usage, as when one
attempts to remember something and then suddenly succeeds. In
this case the internal argument is 'measuring out' the event;
'remember a dream halfway' can mean 'remember half a dream'.
9. A delimited/non-delimited ambiguity in the case of verbs of
contact is discussed in section 3.3.1.
- 156 a
12
wave your hand (for an hour/*in an hour)
beat the drum (for an hour/*in.an hour)
tap the wall (for an hour/*in an hour)
Finally, statives, which describe situations10 in which
neither the external nor the internal argument progresses over
time, always describe non-delimited situations: 11
13
Bill knows the way to San Jose. non-delimited
The two walls touch at the corner. non-delimited
This cheese stinks. non.;-delimited)
10. Although I am using the term 'event' in a non-technical
sense, it seems inappropriate to refer to statives as
describing events. I will refer to them instead as describing
situations.
11. The fact that neither the external nor the internal
argument progresses in time during the situation described by
a stative verb, gives statives their distinctive
characteristics. Stative verbs differ in a crucial respect
from non-statives. A distinguishing feature of stative
predicates is that they describe situations that do not
require temporal duration in order to transpire, whereas
non-statives describe events that require some stretch of
time. An event of singing is a non-stative event that
transpires over time; the singer must move her mouth and
activate her vocal chords in a certain way, and a fragment of
song must be produced. As we observe smaller and smaller
intervals of such an event, the event eventually loses its
identity as an event of singing. If the singer opens her
mouth a fraction of a millimeter, and emits a squeak, that is
not what would generally be considered singing. However, no
matter how small a temporal interval we consider, of a
situation described by a stative verb such as "resemble', the
situation is still describable by the verb. If John resembles
his cousin for a minute, he resembles his cousin for a
millisecond included in that minute. Statives describe events
that are independent of'temporal duration. (For further
discussion and some complications, see Dowty (1979), Taylor
(1977).) That is not to say that stative situations cannot
have duration; John may resemble his cousin for a month or a
year, and then have his hair dyed blue and lose all
resemblance to his cousin.
- 157 
-
(non-delimited)
Sections 4.2.1 through 4.2.3 have shown that an internal
argument of a non-affectedness verb may be characterized by
the verb as undergoing a change ('push', 'drive', etc.), or it
may not be characterized by the verb as undergoing any change
at all ('watch', 'hit', 'know'). If there is change in the
internal argument it must be characterizable as a change in a
single parameter, or a change that is measurable on a scale.
The event is then linguistically delimitable on that scale.
4.2.4 Conative and antipassive alternations
Consideration of the semantic contributions of affected and
unaffected direct arguments reveals that the semantic
contributions of direct arguments are subject to certain
general constraints; direct arguments either undergo no change
during the event, or they undergo change that can 'measure
out' the event. Only direct arguments may be totally
affected, because total affectedness depends on the direct
argument *measuring out' the event with the entirety of the
scale it provides. Examples of minimal pairs can be found in
some languages, in which the same event participant is
represented as the direct and oblique argument of the verb.
In these cases, only the sentences with direct arguments have
totally affected or delimited readings. Recall the discussion
of spray/load verbs in Chapter 2; the totally affected reading
is available (and for many speakers, salient) for whichever
- 158 -
My arm itches.
event participant is the direct argument but not available for
the one that is Nhe indirect argument. The English conative
alternation is another good example:1 2
14
a. eat the apple
b. eat at the apple
The totally affected interpretation is available for 14a, in
which 'apple' is the verb's direct argument, because it
measures out the event. In 14b, where 'apple' is an indirect
argument, it does not measure out the event and the totally
affected reading is unavailable.
Antipassive alternations provide minimal pairs in which the
argument may be direct or oblique (indirect). Bittner (1987)
has noted some interesting data from West Greenlandic Eskimo.
West Greenlandic Eskimo has three antipassive suffixes which
have certain aspectual properties. When one of these
antipassive suffixes is applied to a transitive verb (as in
15), the argument marked with accusative case becomes marked
12. The observant reader will have noticed that the conative
alternation is possible with verbs of contact:
hit the fence
hit at the fence
The difference in meaning obtains even though 'hit' usually
has an unaffected direct argument. This probably reflects the
ambiguity many verbs of contact have regarding whether they
describe delimited or non-delimited events (section 3.3.1.). I
leave this issue open.
- 159 -
with instrumental case, and the transitive verb becomes
intranritive (as in 16). (West Greenlandic Eskimo has
absolutive/ergative case marking; i.e., the subjects of
transitive verbs are marked with ergative case and the
subjects of intransitive verbs are marked with absolutive
case.)
15
Jaaku-p ujarak tigu-a-a
Jacob-ERG stone-ABS take-transitive.indicator-3sgE/3sgA
"Jacob took stone."
16
Jaaku ujarak-mik
Jacob-ABS stone-INSTR
tigu-si-vu-q
take-antipassive-intransitive.indicator-3sgA
"Jacob took stone."
Intransitive antipassive sentences like 16 above are
imperfective. Bittner writes:
For example, with an accomplishment verb like "kill", the transitivQ
form entails that the object referent is dead, whereas the
-si, -as i, and -nnig anitipassives are compatible
wTTh the v ctim being almost but not quite dead yet.
Similarly, for the verb iqqut- "bring inside", the
transitive form entails sthe agent has come in with the
object; while for the - ss i antipassive two of my
informants envisaged a sAtuation with a double door to the
house (e.g., for better insulation), and that the agent has
come in through the outer door but not yet through the inner
one.
In 16 the transitive verb form'is associated with a delimited,
totally affected interpretation. The antipassive forms are
associated with non-delimited, incomplete events.
- 160 -
Warlpiri exhibits a conative alternation. (Example 17 is
from Guerssel, Hale, Laughren, Levin, WhiteEagle (1985).)
17
a. Marlu luwarnu Jakamarra-rlu.
kangaroo shot Jakamarra-ERG
"Jakamarra shot the kangaroo."
b. marlu-ku-rla-jinta luwarnu Jakamarra-rlu.
kangaroo-DAT-3s0-PURP shot Jakamarra-ERG
"Jakamarra shot at the kangaroo."
When the 'kangaroo' is the direct argument of the verb (as in
17a above) it is an affected argument. When (as in 17b) it is
an indirect argument marked with dative case, it is
unaffected.
In these examples from English, West Greenlandic Eskimo and
Warlpiri an argument is affected (or totally affected) when it
is a direct argument, and unaffected when it is an indirect
argument. This is consonant with a theory in which only
direct arguments 'measure out' the event.
Direct or internal arguments represent event participants
that do not undergo change, or else undergo change that
measures out' the event described by the verb. They are
accordingly associated with thematic roles that meet these
semantic/aspectual requirements. Some characterizations of
thematic structure in the literature focus on the idea of
change over time. Gruber's idea of a theme as "the entity
which is conceived as moving or undergoing transitions" is an
example (Gruber (1965) p.38). Jackendoff's (1987) "path"
- 161 
-
constituent of conceptual structures is another.13 These
authors have noted the importance of the idea of 'change' in
linguistically described events, but they have not explained
it in aspectual terms, or explained the special relationship
between the direct argument and 'change'. I will return to a
discussion of thematic structure in Chapter 6.
4.3 External arguments
4.3.1 External arguments and thematic roles
In comparison with internal arguments, a wide range of
thematic roles may appear in the external argument position.
This is indicative of the familiar fact that the thematic
constraints on external arguments are much looser. The
variety of thematic roles that can be associated with external
arguments is illustrated below:
18
AGENT:
Chester dug a hole.
IATURAL FORCE:
The sun dried the clothes. (Levin p.c.)
INSTRUMENT:
The hammer broke the window. (Levin 1986)
(David broke the window with a hammer.)
13. Gruber's definition of theme' is not based entirely on
aspectual properties, so his 'theme' is not limited to direct
argument positions. It may be associated with external
arguments. Jackendoff's path complement, however, to the best
of my understanding, does not turn up in external argument
position.
- 162 -
EXPERIENCER:
Bill fears wasps.
SOURCE:
Katherine sold a cord of wood.
GOAL:
John bought a tractor.
John received a package in the mail.
LOCATION: (Levin 1986; her source, Perlmutter and Postal (1984))
This room sleeps five people.
(We sleep five people in each room.)
TEMPORAL LOCATION:(Levin 1986; her source, Perlmutter and Postal
1492 saw the beginning of a new era. (Levin 1986) (1984)
(A new era began in 1492.)
MATERIAL:
Water splashed the wall.
(Laura splashed the wall with water.)
RAW MATERIAL: (Levin p.c.)
This flour bakes good cakes.
NON-DELIMITING PATH:
The shore channels the current into the bay.
(The engineers channeled the current along the shore into the
bay.)
OTHER:
The box weighs five pounds.
The circle surrounds the dot.
I continue in the approach that a verb describes an event by
describing the changes that take place in its internal and
external arguments during the event. As we have seen in the
previous section, the verb describes very precisely what
changes take place in the internal argument, but it is less
specific about what happens in the external argument during
the course of the event. If there is change, motion or
activity on the part of the external argument, this change,
motion or activity is not necessarily completely specified by
- 163 -
the verb, or characterizable as a change in a single property
that 'measures out' the event. To see that the external
argument need not provide a scale for the event described by
the verb we need only to consider the most common thematic
role to be found in external argument position -- that of
agent.
19
a. Robin picked tomatoes.
b. Frank cut wood.
Examples 19a and 19b above describe events in which the
activity carried out by the external- argument is not
completely specified. Robin may have picked tomatoes by hand,
or with shears; Frank may have cut wood with a saw, an axe or
even a penknife. The sentence does not require that in the
event it describes the agent external argument uses any
particular method or tool. The change in the internal
argument -- the tomatoes or the wood -- does not allow such
freedom of interpretation. The difference between cut wood
and uncut wood is more explicit than the difference between
someone who is engaged in the activity of cutting and someone
not engaged in cutting.
Even when the instrument used by the external argument is
specified in the verb, as in:
20
Laura mopped the floor.
- 164 -
the change, motion, or activity engaged in by the external
argument may vary in many ways and still be described by the
same sentence. Laura may drop the mop in a bucket of water
and then apply it to the floor, or she may dump the water onto
the floor and then slosh the mop around in it. What is of
crucial importance here is the agent's manner of participation
in an event. The agent thematic role represents an event
participant that does undergo change, motion or activity. But
this activity is not completely specifiable. Mopping, picking
tomatoes or cutting wood involve a variety of sub-actions,
such as lifting, pushing or wringing the mop; grasping
tomatoes and pulling them off the stems; holding, guiding and
positioning the saw. One may infer from these sentences that
some or all of these sub-activities took place in the events
they describe, but it is not necessary that they took place
for the sentences to bc true.
More importantly, even if the action performed by the agent
were completely and explicitly specified by the verb, it would
not be undifferentiated with respect to the event. By this I
mean that the action would not be characterizable as a change
in a single parameter, unlike the change in a direct argument
which can be characterized by a change in some salient
property of the argument. It is easy to see that in an event
of- Laura's mopping the floor, as the event progresses in time
all that happens to the floor is that it changes from an
unmopped to a mopped state, while Laura undergoes change and
- 165 -
motion by moving around with a mop and sloshing it in and out
of a bucket. The internal argument, the floor, undergoes
change characterizable on a single parameter and specified by
the verb. The external argument, Laura, does not.
The agent thematic role represents a type of event
participant that is inherently unsuited to characterization by
a scalar change. The agent thematic role is always assigned
to the external argument position for this reason. The
external argument, unlike the direct argument, is not required
to provide a scale for the event described by the verb, and so
it is compatible with thematic roles that describe event
participants. that could not supply such a scale.
The relative freedom allowed in the interpretation of the
way the external argument participates in an event makes it
possible for tbc external argument to seem to be part of a
separate event. A number of authors have suggested1 4
something along these lines; the external argument engages in
one action or event, which causes another event to take place
in the internal argument. This intuition reflects the
differing semantic constraints on internal and external
arguments. Internal arguments in concert with a verb describe
an event in a narrow and precise way. (Hale and Keyser (1987)
14. Hale and Keyser (1987) adopt this view. It is also
implicit in Dowty's (1979) bisentential analysis of CAUSE.
- 166 -
describe the participation of the internal argument as the
"central event" which a sentence describes.) External
arguments, associated loosely with the 'central event', may be
construed as engaging in activity separate enough from the
central event' to aeem to be an event unto itself.)5
The tests that were applied to direct arguments to show that
they provide a scale for the event described by the verb work
differently with agent external arguments. Rate adverbials
may apply to the external argument (or rather to the activity
engaged in by the external argument), but when they do so they
modify, not a scale provided by the argument and mapped onto
time, but an unspecified collection of changes or activities
associated with the external argument and mapped onto time..
Consequently, it is not clear which of the actions or whether
all of the actions engaged in by the external argument are
being modified. Does the sentence 'Robin picked tomatoes
slowly' only refer to an event in which every movement of
Robin's involved in tomato-picking was a slow movement? Robin
may have reached for the tomatoes slowly, pulled them off the
vine slowly, and then tossed them quickly into a bucket. That
would be an event describable by the sentence, even though not
every component of Robin's motion was slow. However, if
'slowly' is taken to apply to the direct argument, tomatoes,
15. The reader is reminded that the term 'event' is used
loosqly and intuitively here.
- 167 -
it is clear what it meana about the event described by the
verb; the rate at which tomatoes left the vine and went into
the tomato bucket was slow.
The contrast between rate adverbials modifying the external
and the internal argument is made clearer by the following
sentences. The verbs in the a) sentences below are
unergatives, meaning their arguments are external. The verbs
in the b) sentences are unaccusatives and their arguments are
internal.1 6 The adverbial expressions 'a little bit at a
time' and 'a lot at once' make explicit reference to measuring
out something on a scale, over time. Therefore they are
anomalous with the external arguments of unergative verbs, but
acceptable with the internal arguments of unaccusative verbs:
21
a. The dancer danced slowly /* a little bit at a time.
The anrauncer talked slowly /* a little bit at a time.
b. The candle melted slowly /a little bit at a time.
The lake froze slowly /a little bit at a time.
22
a. ?Martha danced quickly, a lot of her dancing at once.
?The announcer talked quickly, a lot of him talking at once.
b. The candle melted quickly, a lot of it melting at once.
The lake froze quickly, a lot of it freezing at once.
Since there is only one argument in each of the sentences
16. The theory of unaccusative and unergative verbs vill be
addressed in Chapter 6. For now we accept these sentences as
minimal pair examples of verbs with either a single external
or a single internal argument.
- 168 -
above, there is no ambiguity as to which argument's change or
motion is being modified by the rate adverbial. In the a)
sentences, where the adverbial phrase modifies an external
argument, it is not modifying a scale. In the b) sentences,
where it is modifying an internal argument, it does modify a
scale.
Certain adverbials yield an explicit value on a scale for
the event described by the verb, and so they can be used as
tests for the presence of a scale. The adverbial 'halfway' is
an example. It is anomalous with agent external arguments:
23
*Martha danced halfway. -
*Thomas ate halfway.
and acceptable with internal arguments:
24
The lake froze halfway.
The candle melted halfway.
When it occurs in a sentence with both internal and external
arguments, it modifies the change in the internal argument.
The sentence,
25
Jacinta ate an apple halfway.
may be paraphrased as:
- 169 -
26
Jacinta ate half an apple.
but not as:
27
a. Half of Jacinta ate an apple.
b. Jacinta half ate and half gulped an apple.
27a would be a paraphrase of a sentence in which the scale was
constructed directly out of the external argument, Jacinta;
and 27b would be a paraphrase of a sentence in which 'half'
was applied to the unspecified collection of actions that
constitute Jacinta's actions of eating. Neither is a possible
paraphrase, so the adverb 'halfway' (which explicitly means
covering half the distance on-some sort of a scale) applies
only to the change or motion of the internal argument.17
The adverbial expressions that are odd with agent external
arguments ate odd with other thematic roles occupying external
argument position as well. (These tests only apply to
non-stat ives.)
17. Some unergative verbs of motion are acceptable with the
adverbial 'halfway': 'Aaron walked halfway'. But in these
sentences *halfway' modifies a change in location, not the
collection of actions involved in the activity of walking.
The sentence it not paraphrased by either of the following:
'Half of Aaron walked', or 'Aaron half walked and half ran'.
What the sentence means is 'Aaron walked half of some
distance'. Again, the scale that 'halfway' depends upon for
its interpretation can only be directly expressed through an
internal argument; 'walk halfway' must be understood as 'walk
half the distance'.
- 170 -
28
a. *The sun dried the clothes quickly, a lot of it (the sun)
drying the clothes at once.
b. The sun dried the clothes halfway.
possible paraphrases:
*Half the sdn dried the clothes.
Half the clothes dried.
29
a. *The hammer broke the window quickly, a lot of it (the
hammer) breaking the window at once.
b. The hammer broke the window halfway.
possible paraphrases:
*Half the hammer broke the window.
Half the window broke.
30
a. *The shore channeled the current quickly, a lot of it (the
shore) channeling the current at once.
b. The shore channeled the current halfway.
possible paraphrases:
*Half the shore channeled the current.
The shore channeled the current half of some distance
(to be traveled by the current).
31
b. ?This flour baked a good cake halfway.18
possible inferences:
*Half the flour is consumed.
The cake is half-baked.
Half the cake is baked.
32
b. ?Water splashed the wall halfway.
32 is ambiguous between a reading in which the direct argument
is affected and measures out the event, and one in which it is
unaffected. The adverbial 'halfway' only makes sense in the
18. In the judgement of this author -quickly' and 'slowly' are
odd when used with the verbs 'bake' and 'splash', in a
sentence without an agent. The adverbial tests are omitted
for this reason in 31 and 32.
- 171 -
former reading, in which case the sentence means water
splashed half the wall, not half the water- splashed the wall.
This is clearer in the following example:
33
?A bucket of water splashed the wall halfway.
possible inferences:
*Half a bucket of water splashed the wall.
A bucket of water splashed half the wall.
These tests are more difficult to apply in the case of verbs
taking a source or a goal as external argument:
34
John bought a tractor quickly.
Mary sold a cord of wood quickly.
These verbs express events with little or no duration, so it
is difficult to construe 'quickly', 'slowly' or 'halfway' as
modifying the change in the direct argument during the course
of the event. It is possible to understand "quickly' as
modifying the time it took for the buying or selling event to
take place. If it modifies the participants of the external
argument, then it modifies the unspecified collection of
actions that are involved in John's buying or Mary's selling,
which cannot be represented as a scalar change. If it
modifies the participation of the internal argument, it
modifies the wood's or the tractor's having changed hands.
The direct arguments undergo only a simple change of ownership
a change of state. The external arguments, being sources
or goals, represent event participants that undergo a change
- 172 -
at the point in time marking the end of the event -- but that
terminus is already registered in the internal argument's
change of state. The external argument's marking of the event
terminus is redundant. The verb phrases:
35
buy a tractor
sell a cord of wood
represent deJimited events independently of the external
argument. Moreover, it is only the change in the internal
argument that can be characterized as a change in a single
paramenter.
Examples 27 to 35 show that -even though external arguments
with thematic roles other than agents are not quite as
clearcut in their interaction with the adverbial tests above,
they nevertheless support the generalization that internal
arguments can "measure out' and delimit a linguistically
described event, but external arguments cannot. It is through
the direct argument that the event is defined temporally. The
direct argument is identified with the event in a very
particular way.
Although a variety of thematic roles may occupy the external
argument position, some alternations involving verbs that can
alternate external with internal or oblique arguments are
impossible when -the external argument would clearly delimit
the event independently of any internal argument:
- 173 -
36
a. Five people slept in this room.
b. This room slept five people.
c. Five people slept to New York (during a train ride from
Boston to New York.) 19
d. *New York slept five people.
37
a. The engineers channeled the current along the shore into the
bay.
b. The shore channels the current into the bay.
c. The engineers channeled the sewage into the bay.
d. *The bay channeled the sewage.
The type of thematic role possible in external argument
position is a reflection of aspectual constraints on argument
structure.
4.3.2 External arguments and the translation of spatial
delimitedness into temporal delimitedness
External arguments do not participate in delimiting
linguistically described events in the same way that internal
arguments do. Recall that the count/mass distinction of the
internal argument translates into a delimited/non-delimited
distinction in the verb phrases of affectedness verbs:
38
a. Charlie drank beer. (non-delimited)
b. Charlie drank a beer. (delimited)
a. Laura shoveled snow. (non-delimited)
b. Laura shoveled the snow in the driveway. (delimited)
19. This is somewhat awkward, but colloquial in the author's
dialect.
- 174 -
In the a) sentences, the internal argument is a mass noun and
the sentence is non-delimited. In the b) sentences the noun
phrase represents something countable and the sentence is
delimited. Thts does not happen with external arguments:
39
a. The heater dried the shoe. (delimited)
b. Heat dried the shoe. (delimited)
40
a. Snow surrounds the house.20  (vion-delimited)
b. Seven trees surround the house. (non-delimited)
All the internal arguments in 39 and 40 are countable, In the
a) sentences the external argument is a mass noun; in the b)
sentences it is countable. 'Dry' is an affectedness verb, but
the countability of its external argument has no effect on the
delimitedness of the sentence. The sentence is delimited in
both 39a and 39b, as we would predict for an affectedness verb
with a countable direct argument. Whether the external
argument is countable ('the heater') or not ('heat') has no
effect on the delimitedness of the sentence. 'Surround' in 40
is a stative (in this usage), and the verb phrase is always
non-delimited, regardless of whether its external argument is
mass or count. A countable external argument ('seven trees')
or a mass external argument ('snow') both yield a
20. 'Surround' also has a non-stative usage: 'The cops
surrounded the house.' I am only considering the stative
usage here.
- 17W -
non-delimited sentence when used with a stative verb.
Delimitedness is a property determined exclusively by the verb
and its internal arguments.
A digression is necessary here. There is an important
difference between non-delimitedness associated with bare
plurals and non-delimitedness associated with mass nouns. The
delimitednesss of a verb phrase containing an affectedness
verb makes no distinction between a mass noun ((a) examples
below) and a bare plural ((b) examples) in direct argument
position. Either one has properties of mass-ness as opposed
to countability and either one makes a delimited verb phrase:
41
a. Charles drank beer.
b. Charles drank mugs of beer.
a. Laura shoveled snow.
b. Laura shoveled driveways.
(non-delimited)
(non-delimited)
(non-delimited)
(non-delimited)
In the case of external arguments, however, there is a
difference. While the mass-ness of a mass noun external
argument does not contribute to the delimitedness of the
sentence, that of a bare plural external argument can:
The heater/heat dried the clothes.
Heaters dried the clothes.
The man/perseverance painted the wall.
Men painted the wall.
The child drew a circle.
Children drew a circle (each).
(delimited)
(delimited, non-delimited
(delimited)
(delimited, non-delimited
delimited)
,delimited, non-delimited
- 176 -
42
a.
b.
a.
b.
a.
b.
All the verbs in 42 are affectedness verbs and all internal
arguments are countable. All the a) sentences are delimited,
as we would expect. These are the sentences in which the
external argument is a single count noun or a mass noun.
However, the b) sentences, with bare plural external
arguments, have a non-delimited reading available as well.
How is it that the massness of the bare plural external
argument, but not of the mass noun external argument, may
translate into non-delimitedness of the sentence?
The non-delimited readings above result from an
interpretation in which each external argument performs a
separate event. In other words, the event is iterated through
the iteration of the external argument. For example, the
sentence:
435
The linguists drank a beer.
may be understood to mean that all the linguists drank the
same beer, in which case the event described is delimited
within the verb phrase in the familiar way. The linguists are
consuming together, a certain fixed quantity of beer. But it
may also mean that the linguis;s drank separate beers, in
which case each linguist's drinking of a beer was a separate
event. The single event consisting of several beer-drinking
events put together does not necessarily have a definite
endpoint in time, since the linguiste may drink their beers at
- 177 -
different indefinite times. This is the kind of
non-delimitedness that arises out of the iteration of separate
events an unspecified number of times. Since the events need
not transpire at the same time, they may stretch out
indefinitely over time. If the number of times is specified
by having a count plural external argument, the sentence is
delimited:21
44
Five linguists drank a beer. (delimited)
We must distinguish this kind of non-delimitedness, which
can interact with the external argument, from the
delimitedness that depends on-a single event, and which is
only determined within the verb phrase. The delimitedness
investigated in this thesis is of the latter type. (I will
leave the nature of the former As a separate topic for
research.) The verb phrase may be thought of as representing
the unit event, which may or may not be iterated by the
21. Verkuyl (1972) observed that bare plural indirect objects
can also, under certain circumstances, impart
non-delimitedness in the same way: 'Den Uyl overhandidge een
uur lang het PVDA-speldje aan congresgangers' (translation
from the dutch: "For an hour Den Uyl handed out the Labour
Party badge to 3ongress-goers".) In this example, the direct
argument ('the Labour Party badge') is countable, but the
indirect argument ('congress-goers') is a bare plural. The
fact that the sentence is compatible with the adverbial
expression 'for an hour' shows that it can describe a
non-delimited event. The non-delimitedness is contributed by
the bare plural indirect argument.
- 178 -
external argument.22 It is the delimitedness of the unit
event that is syntactically important in the ways described in
this thesis.
4.3.3 Summary
When the interaction of external arguments and delimitedness
is examined closely, it is evident that external arguments do
not contribute to the determination of delimitedness in the
same way that internal arguments do. They cannot delimit the
event through a scale as do the internal arguments of
affectedness verbs. External arguments are independent of the
event as measured out by the internal argument. They do not
provide a scale for the event, so they are not required to be
undifferentiated over time with respect to the event. The
change that occurs in the external argument during the course
of the event is not necessarily characterizable as a change in
a single parameter.
4.4 Oblique arguments
22. This view of the verb phrase as representing the unit
event is compatible with Schein (1986) who regards the verb
phrase as a kind of aspectual domain. He represents the
logical form of 'sum of plurals' readings and 'event dependent
quantification' as (a) and (b) below, respectively:
a. [e; VP(e)] INFL(e, NP)
b. Ae: VP(e)J INFL(e, NP)
The verb phrase in Schein's event logic is a restriction on
quantification over events.
- 179 -
4.4.1 Delimiting oblique arguments
Chapter 3 illustrated how direct arguments of certain verbs
may delimit the event described by the verb. Indirect or
oblique arguments, when they are sources and gQals, are also
delimiting expressions, as shown by these familiar tests with
durative adverbials:
45
a. push the car (non-delimited)
b.
c.
46
a.
b.
c.
d.
47
a.
b.
c.
d.
push the car for an hour
*push the car in an hour23
push the car out of the garage (delimited)
*push the car out of the garage for an hour
push the car out of the garage in an hour
It took an hour to push the car out of the garage.
push the car to a gas station (delimited)
push the car to a gas station for an hour
push the car to a gas station in an hour
It took an hour to push the car to a gas station.
The verb phrases 46b and 47b can have iterative
interpretations in which they are understood to mean the car
was pushed out of the garage or to the gas station over and
over again during the hour., (46b has only the iterative
interpretation and so is pragmatically odd. 47b has the
semelfactive as well as the iterative interpretation.) 46z-d
23. The adverbial test employs a particular reading of 'in an
hour'. The relevant reading is that in which it takes an hour
to push the car -- not the reading in which the car is to be
pushed at some time beginning an hour from the present.
- 180 
-
and 47c-d are grammatical with the expressions 'in an hour'
and "it took an hour to...', which indicate the event3
described are delimited.
Goal (and source) expressions delimit an event by marking an
endpoint (or a beginning point) on the scale defined through
the verb and its direct argument. In the sentences above,
they mark an endpoint in the direct argument's change in
location over time.
If there is a direct argument, a goal expression delimits
the event in concert with that direct argument. If the direct
argument does not delimit the event (if it is not an affected
argument) a goal phrase delimits the event by marking an
endpoint on a scale provided by the direct argument -- as in
the sentences with 'push' above. If there is a delimiting
direct arguent (an affected argument) in the verb phrase, an
additional goal only serves to specify the endpoint of the
event whose existence is already established by the theme:
48
eat an apple to the core
perform a play to the fourth act
Heinlmbki (1983) discusses this phenomenon in Finnish:
49
a. Metsist'ija ampui lehmb'n
hunter shot cow-ACC
"The hunter shot a cow."
b. Meteestaji ampui lehman kuoliaaksi
- 181 -
hunter shot cow-ACC dead-to
"The hunter shot a cow dead."
c. Metsast'ija ampui lehman silmpuoleksi
hunter shot cow-ACC eye-half-to
"The hunter shot a cow and blinded the cow in one eye."
Recall that accusative case on the direct object in Finnish
signals a delimited event, i.e. an event with some terminus.
What terminus that is may be specified by a goal phrase, but
the necessity of there being some terminus is indicated by the
direct object Itself. Hein'amaki writes (Heinamaki (1983)
p.157):
We see that the accusative form of the object in itself does
not semantically entail any particular result, such as
death in case of shooting; it only entails the existence of
a bound. The bound can be given explicitly, as in
(49b,c). But how-is the bound inferred in
(49a)? The sentence denotes a telic situation, i.e. a
situation that has a clear end point beyond which the
situation cannot continue -- namely the death of the cow.
Thus, if a sentence that has an accusative object denotes a
telic situation, the telic endpoint is inferred to be the
actual end point if no explicit bound is given in the
sentence. But, as was said before, this infertace is
pragmatic and can be cancelled by explicitly giving some
other bound. We also see from the examples in (49)
that kL-oliaaksi/silmlpuoleksi in (49b,c) are
not independent bounds added on top of the accusative object,
but ratier specificatioL1s of the bound, the existence of
which is implied by the acc- .ative object.
A deliml+ing rxpress'on such as a goal cannot delimit the verb
phrase independently of the s1 ale established by the direct
- 162 -
argument.24
Source phrases also delimit an event:
50
a. push the car (non-delimited)
b. push the car out of the snowbank (delimited)
An event may be both inceptively and completively delimited:
51
push the car out of the snowbank into the river
However, within a verb phrase an evsnt may be delimited at
most once inceptively and once completively. Since completive
delimitedness is much more prominent than inceptive
delimitedness in natural language, I will discuss only
complctive delimitedness. In the remainder of this thesis
'delimitedness' will mean completive delimitedness.
A verb phrase may be delimited only once:
24. A verb phrase with no direct argument can be delimited by
a goal phrase. In this case the goal marks an endpoint on a
scale provided by the verb's meaning, that tould be manifested
(in pure form) as a direct argument. Verbs of motion
illustrate this point. 'Carmen- walked to San Frar isco' may
be paraphrased as 'Carmen walked the distance to San
Francisco'. Although the external argument of this sentence,
Carmen, is the object with which the changing property of
location is associated, Carmen's Participation in the event is
not represented as that changing property* armen is also an
agent, engaging in activity not entirely characterizable as a
simple change of location. The pure linguistic representation
of that change must be manifested in a direct argument such as
'the distance' in the preceding example.
- 183
52
a.*push the car.to New York to San Diego
b. push the car to New York and push the car to San Diego
a. push the car to New York and San Diego
53
a.*push the car out of the snowbank out of the river
b. push the car out of the snowbank and push the car out of the riVe
c. push the car out of the snowbank and the river
52a is ungrammatical even though it could pragmatically have
the same meaning as 52b and 52c. 52a imposes two delimitings
on the verb phrase. 52b, by repeating the verb, describes two
separate events, one for each goal; and 52c merges the two
cities into one goal phrase delimiting the one event expressed
by the verb phrase. The same discussion applies to 53a
through 53c.
Gruber (1965) investigated Source-Goal patterns in depth,
and showed that certain kinds of verbs describing a change of
location can have more than one goal phrase. (Gruber refers
to these verbs as expressing 'Positional' transitions.)
(Example'54 is from Gruber (1965).)
54
John sent the book to New York to Bill.
In these cases, Gruber notes, the second goal phrase serves to
further specify the first. The preferred ordering of the two
goal phrases is from general to specific:
55
*John sent the book to Bill to New York.
Successive goal phrases are impossible with verbs describing
- 184 -
what Gruber refers to as 'Possessional' transitions:
56
*John gave the letter to New York to Bill.
*John gave the letter to Bill to New York.
Gruber explains the possibility of having more than one goal
phrase as depending on the ability of the verb to have a goal
which can be further specified (Gruber (1965) p.82 ):
The fact that the Possessional and Identificational differ
from the Positional verbs in that they permit only one
Source-Goal pair may be thought to follow from the fact that
the specifications for possession and identity, once made,
cannot be refined or elaborated upon. They are automatically of
absolute specification. Position may be specified in greater
degrees of accuracy.
In othsr words, even when there is more than one goal phrase
present in a verb phrase, there is still only one terminus of
the event expressed by that verb phrase.25
The preceding discussion leads to the conclusion that
oblique arguments may be delimiting expressions but when they
are, they are parasitic on the scale provided by the direct
argument. Furthermore, a verb phrase may only be delimited
once; i.e., there may be only one endpoint to the event
represented by a verb phrase.
4.4.2 A delimiting requirement on secondary arguments
25. For most speakers, the second goal phrase is set off by a
special intonation indicating it is a parenthetical and not a
true goal phrase.
- 185 -
6
There is a class of verbs of change of location, including
such verbs as 'lay' , 'put' , and 'set' which are peculiar in
that they require some extra delimiting expresssion in the
verb phrase. They are awkward or ungrammatical when used with
only a direct object,..but they improve if some delimiting
expression is added:
57
a. *lay the book
b. lay the book on the .table
c. lay the book down
58
a. *set the saw26
b. set the saw on the sawhorse
c. set the saw on its side
d. set the saw down
59
a. *put the pot
b. put the pot on the floor
c. put the pot there
d. put the pot down
Like the verbs of imparting motion, these verbs describe
situations in which the direct argument undergoes a change of
louation. The expressions in the b), c), and d) examples
above, which delimit the event or act of laying, setting, or
putting; do so by marking the endoint on a scale 'measured
out' by the object's changing property of location. This is
26. There is an acceptable sentence like 59a in which 'set'
has an entirely different- meaning: to prepare something for
use, or to choose the settings or parameters on instruments.
This is not the relevant reading. The usage of 'set' I am
concerned with is parallel to 'put'.
- 186 -
what rate adverbials and degree expressions modify:
60
set the saw down slowly
put the pot down slowly
61
set the saw a little further away
put the pot a little more towards the door
These particular verbs require extra delimiting expressions in
the verb phrase, because they describe events which must be
linguistically delimited, but which are not delimitable
through the direct argument alone. The verb is incomplete, in
a sense, without the delimiting expression.27  This leads to
another hypothesis about delimiting xpressions in the verb
phrase. If the verb requires some other expression besides a
direct argument to be in the verb phrase, that expression will
be delimiting. If we take the distinction between arguments
and adjuncts to be the following: arguments are required by
the verb, in order to make a grammatical sentence, whereas
adjuncts are optional, then secondary arguments (NP's required
by a verb in addition to a direct argument NP) are always
delimiting expressions.
This is also true of verbs entering into double object
27. These verbs, in combination with the extra delimiting
element may be complex affectedness verbs. Middles are
awkward but do not seem beyond hope: 'This heavy saw sets down
like a dream with this new handle attachment.' The fact that
they are complex verbs may make the middles more difficult to
process than they would otherwise be.
- 187 -
constructions,
62
give John a book
build John a house
in which both 'John' and 'book' or 'house' seem to be objects
of the verb. The object 'John', which I will call the dative
object, may also occur in a delimiting expression rather like
a goal phrase:
63
give a book to John
With the verb 'build', the dative argument occurs with the
preposition 'for', where it may have a benefactive or a
recipient reading:
64
build a house for John
The bene.~~.etive reading does not delimit; the recipient
reading does. When the dative is in a double object
construction (as in 62 above) the recipient or delimited
reading is selected.
This is true of other double object verbs:2 8
28. 63a and 63d are admittedly a little awkward because 'give'
expresses an event with little or no duration, while in an
hour' assuames the event has some duration. 64b and 64e are
somewhat possible if *build' is not understood as an
affectednesse verb. In spite of these complicating factors,
there is a clear difference in grammaticality that indicates
- 188 -
65
a. give John a book in an hour
b. *give John a book for an hour
c. It took an hour to give John a book.
d. give a book to John in an hour
e. *give a book to John for an hour
f. It took an hour to give a book to John.
66
a. build John a house in a day
b. *build John a house for a day
c. It took an hour to build John a house.
d. build a house for John in a day
e. *build a house for John for a day
f. It took an hour to build a house for John.
There is a clear difference in grammaticality between the
verb phrases modified with 'for an hour' and those modified
with in an hour', indicating that these expressions with
dative arguments are delimited expressions. The delimitedness
is contributed by the lative argument -- in some cases jointly
with the accusative argument. Double object construsCtions
will be discussed in greater detail in Chapter 5.
4.5 Summary
Four principles of argument structure have been advanced in
this chapter:
(i) If an internal direct argument undergoes change during
the event described, that change is characterizable as a
change in a single parameter or change on a scale.
the sentences are delimited.
- 189 -
(ii) Events are described linguistically as delimited
through the scale provided by the direct argument, and they
are delimited within the verb phrase; i.e., an event may be
linguistically delimited by an internal direct or indirect
argument, but not by an external argument.
(iii) There may be at most one 'delimiting' associated with
a verb phrase.
(iv) Secondary arguments are delimiting expressions.
Chapter 5 will examine the consequences of these principles
for syntax. In Chapter 6 we will see how far these principles
go toward explaining the mapping of cognitive structure into
syntax.
- 190 s
Chapter 5. Syntax of a Theory of Aspect
5.1 Introduction
Chapters 1 through 4 have argued that the semantic property
of aspectual delimitedness is relevant and accessible to
syntax. Chapter 5 is somewhat of a digression, and will
tackle two specifically syntactic questions in which aspect
bears on syntax. The first concerns the place of aspectual
markers in X-bar theory, and the second concerns the
interaction of aspectual principles of phraee structure with
the syntax of double object constructions. I will not get
much beyond posing the questions and will leave much
unanswered. However, this will support one of my central
theses; that syntactic questions may be posed partly in
aspectual terms.
To begin with, a brief review of the history of aspect .n
government and binding theory is in order. The earliest
models of generative grammar employed context-free rewrite
rules that expressed the possibilities of recursion and linear
ordering in natural language. In Chomsky (1965), tense,
modality and aspect were introduced under the auxiliary node
(AUX) at the level of the verb phrase.
1
AUI -- > Tense (Nodals) Aspect
8 -- > NP AUX VP
It was evident that context-free phrase structure rules were
-4 191 -
inadequate for handling the distribution of tense, aspect,
modality and agreement morphemes. This was one reason for the
drive to eliminate such rules from the grammar. In Chomsky
(1955) and Chomsky (1982) a transformation was employed to
move the verbal inflection from its position in Deep Structure
onto the verb. Tense and agreement morphemes were generated
together under the inflection node (INFL) and moved to their
position on the verb by a movement rule. INFL was viewed as
the head of S in the X-bar phrace structure system. In
Chomsky (1986) the movement uniting a verb and its inflection
was subsumed under the general process of movement-to-head, by
the proposal that the verb moves into the INFL position where
it amalgamates with its inflection.
Steeleet al (1981) argued that AUX was an independent
syntactic category. They did not subdivide AUX into syntactic
categories, so Aspect was grouped with Tense, Modality etc.,
under their approach. By treating INFL as the head of the
clause, the current tradition of generative syntax expresses
the intuition that tense, modals, and agreement take scope
over the clause. The intuition that aspect takes scope over
the VP rather than the clause was expressed indirectly in
Emonds (1986) who placed English aspectual markers under the
VP node in his phrase structure rules. In section 5.2 I will
suggest that Aspect .is an independent syntactic category,
separate from INFL.
- 192 .
5.2 Aspect and phrase structure
The discussion in Chapters 1 through 4 has concentrated on
the semantic property of delmitedness and its.repercussions in
syntax. Apart from the discussion in section 2.4, very little
has been said about syntactic markers of aspect -- words and
morphemes that specifically affect, alter or represent the
property of delimitedness. In this section I will address the
question of how aspectual markers such as these are related to
phrase structrre. In particular, I will argue that aspect is
a syntactic category, and I will consider several possible
instantiations of aspect in syntactic phrase structure. This
last issue will be left partly unresolved.
There are a variety of types of words and morphemes that
have been generally considered 'aspectual', but I will only
address those that are 'aspectual' in the sense intended here
-- they have to do more or less directly with delimitedness.
I take this approach because the primary topic of
investigation in this thesis is delimitedness; and because I
have adopted the assumption that delimitedness is one of the
1. Aspectual markers that indicate properties other than
delimitedness may contribute 'manings' based indtrectly on
delimitedness. For example, some languages have markers of
iterativity in their verbal morphology. These indicate that a
linguistically described event is to be understood as
happening again and again. An event must be delimited in
order to be interpreted in such a way. It is possible that
all or most aspectual meanings are built in some way on
delimitedness, but I will leave this as an open question.
- 193 -
central, if not the central property - or feature --
expressed by the category of aspect. This assumption is
supported by the evidence in Chapters 1 and 2 for the
syntactic importance of delimitedness. I will assume that
aspectual markers of delimitedness are syntactically
representative of a larger category of aspectual markers.
(The detailed working out of this idea must be left for
further research.) There are many obstacles to this approach
-- not least of which is the overwhelming complexity and
variety of tense and aspect phenomena across languages,
certainly beyond the scope of this thesis. But in order to
make any headway in the investigation of such complex
phenomena as aspect, it is necessary to constrain the
investigation by narrowing the the line of approach to the
problem.
The argument that aspect is a syntactic category will
proceed along the following lines. I will show that aspectual
markers cannot be part of INFL, and they cannot be simply
features on the verb. They have a syntactic identity and
distribution of their own.
5.2.2 Semantic independence of tense/modality and aspect
In considering the differences between these various
syntactic representations I will begin, perversely, with a
semantic discussion. The grouping of tense, modality and
aspect together with agreement (AGR) under the category
- 194 a
inflection (INFL) runs counter to their semantic groupings.
In this section four points will be discussed, supporting the
view that aspect and tense/modality should be semantically
distinguished. I will not include a discussion of AGR, since
it is semantically orthogonal to both tense and aspect.
(1) First, tense and modality have quite different semantic
functions from aspect. Tense markers may locate an event
temporally with reference to some point in time that cannot be
fixed solely by the grammar or by the sentence. Usually this
is the present moment, or the moment a linguistic expression
is uttered. Modality also refers to contextual or
extra-grammatical information.-when it indicates such things as
the speaker's opinion about the supposed truth or falsity of a
proposition; or how the statement is used as communication
among individuals (e.g., as a statement, question or
command). These kinds of modal 'meanings' cannot be
interpreted in a linguistic expression without knowledge of
who the speaker is, or how the sentence is used. In this
respect, tense and modality are indexical. Aspectual
delimitedness, however, is not. It is a property of
linguistically described events, one which is grammatically
indicated and does not require reference to contextual
information in order to be interpreted.
(2) Secondly, the same inflectional forms ar6 often used for
tense and modality (e.g. 'will'), but morphemes that express
- 195 -
both aspect and tense or aspect and modality are rarer. They
do not exist in English, and I have seen no incontrovertable
evidence of such forms in other languages.2  Furthermore,
tense and modality often seem to merge semantically as well as
morphologically; future times may easily be expressed as
possible times, for example.
(3) Third, as discussed in Chapter 2, delimitedness may be
compositional, having to do with the interaction of a verb and
its internal arguments. Tense and modality are not. This
demonstrates that the tense/modality system and the aspect
system have very different mechanisms for semantic
interpretation.
(4)-Fourth, aspectual delimitedness is interpreted
independently of tense and modality. It is true that tense
and modality sometimes appear to affect the interpretation of
the temporal organization of a linguistically described
event. However, tense/modality and delimitedness are
independent. This is illustrated in the remainder of this
2. It is possible to find single morphemes that seem to
express meanings incorporating aspect and tense. But this is
often due to independent semantic or pragmatic restrictions on
what tenses and aspects may occur together. In these cases,
it is possible that the morpheme directly expresses either
tense or aspect, and independeht principles constrain the
tenses or aspects that may co-occur with that specifically
indicated by the morpheme. To contradict (2) above it would
be necessary to find morphemes clearly indicating aspect and
morphemes clearly indicating tense, that have the same
syntactic distribution.
- 196 -
section, where the interact.on of delimitedness with the
English past, present and future forms is examined.
The English past tense imposes a temporal end to a
linguistically described event in a sense, simply because the
event has happened in the past and is therefore easily
understood as not happening in the present. An event that is
past has reached a definite temporal endpoint, but that is not
the same as being delimited in the sense we are
investigating. Consider the sentences:
3
Dustin ate the apple. (delimited)
Dustin slept. (non-delimited)
In both of these sentences, the past tense imposes a finite
duration on the event described, but the delimitedness of the
first sentence and the non-delimitedness of the second are
preserved. Dustin may have slept a day or a year (an
indefinite stretch of time), but he only ate the apple until
the apple was finished (a definite stretch of time).
Durational adverbials show that the delimited/non-delimited
distinction is preserved:
4
Dustin ate the apple (?for an hour/in an hour). (delimited)
Dustin slept (for an hour/??in an hour). (non-delimited)
The English present tense has many peculiar properties.
Stative verbs in the present tense have a natural
interpretation as describing a situation that continues over
- 197 -
an indefinite stretch of time:
5
Chizuko loves her grandchildren.
but non-stative verbs are odd in such an interpretation:
6
?Keith climbs a tree.
Non-statives tend to be shifted into habitual or narrative
interpretations in the present tense:
7
Keith climbs a tree every morning. (habitual present)
This is what happened to Keith: This bear comes
up out of the woods and starts chasing him
down the trail, and Keith climbs a tree to get out of its
reach... (narrative present)
The oddity of the English present tense is related to the
question of whether present time is interpreted as an interval
or as a moment.3 It is not my purpose to discuss the English
present tense here, but to illustrate the point that
extraneous and poorly understood phenomena enter into the
interpretation of the present tense, which' may complicate
judgements about delimitedness. Nevertheless, delimitedness
is clearly preserved under the present tense. Consider these
sentences in the habitual present:
3. See Dowty (1979) for. more discussion.
- 198 -
8
Keith eats an apple every morning.
Keith-sleeps every morning.
In the events as they are described above, it takes Keith a
definite length of time to eat an apple (every morning),
whereas he sleeps for an indefinite length of time (every
morning). The durational adverbial tests support these
judgements:
9
Keith eats an apple (?for an hour/in an hour). (delimited)
Keith sleeps (for an hour/??in an hour). (non-delimited)
The English future form, marked by the modal 'will', may
have a temporal or modal interpretation. The sentence:
10
The house will burn.
may indicate that the burning of the house will take place in
the future, or it may imply the present or future possibility
of the house's burning (i.e., the house is something that
would burn if it were exposed to fire). In either reading the
aspectual delimitedness of the verb phrase is unaffected:
11
Keith will eat an apple (?for an hour/in an hour). (delimited)
Keith will sleep (for an hour/??in an hour). (non-delimitp4)
Tense may obscure the speakers's judgements about
delimitedness, but on further consideration it is clear that
delimitedness is independent of tense. Although both tense
- 199 -
and aspect deal with the temporal nature of events, tense is
orthogonal to aspectual delimitedness.
This is also true of the relationship between modality and
aspect. Markers of modality can introduce particular
aspectual properties of their own. 12a below clearly
describes a delimited event. Native speakers report an
intuition that 12b describes a non-delimited event:
12
a. Keith ate an apple.
b. Keith can eat an apple.
Nevertheless, the delimited/non-delimited distinction is
preserved under the modal 'can':
13
Keith can eat an apple in an hour/*for an hour.
Delimitedness is also independent of adverbials of time.
Durational adverbials can impose a definite length of time on
an event, but when they do, the distinction between delimited
and non-delimited sentences is unaltered:
14
Keith slept for three hours.
Keith climbed a tree for three hours.
In the sentences above, the events of sleeping and
tree-climbing are expressed as having a definite duration over
time, but tree-climbing may be understood to have happened
iteratively, whereas there is only one event of sleeping
- 200 -
expressed. The distribution of semelfactive and iterative
readings that is determined by the property of delimitedness
is preserved. Even though the adverbials impose a definite
duration on the events, they do not obliterate the
delimited/non-delimited distinction.4 Tense, modality and
temporal adverbials do not enter into the determination of
aspectual delimitedness.
These four semantic reasons for separating aspectual
delimitedness from tense and modality have been reviewed:
(i) Tense and modality refer to extra-grammatical, contextual
information. Delimitedness does not.
(ii) Tense and modality often merge morphologically or
semantically. This is less apparent with aspect and tense or
aspect and modality.
iii) Aspect is compositional. Tense/modality is not.
iv) Delimitedness is interpreted independently of tense.
The fact that tense/modality and aspect are semantically
distinct is not sufficient grounds for separating them
syntactically as well, but it is circumstantial evidence in
favor of it. The basic syntactic categories of nouns, verbs,
adverbs, adjectives, determiners, etc. are quite distinct
4. There is precedent, in the literature on tense, for
treating time adverbials semantically as part of the tense
system. In Reichenbach (1947) and Smith (1978), adverbials
help determine the reference points around which the
interpretations of tenses turn. In Hornstein (1977)
adverbials map directly onto some of these reference points.
Tense and adverbials of time may be treated naturally as part
of a system of time reference that is distinQt from aspect and
delimitedness.
- 201 -
semantically; and the correlation of semantic with syntactic
distinctions is well-attested in natural language. However,
the proof of the pudding lies in the syntax.
5.2.3 Syntactic independence of tense/modality and aspect
Consider the English progressive marker, 'be -ing'. It has
the function (among other things) of making a verb phrase
describe a non-delimited event:
15
Harry was eating an apple for an hour.
?Harry was eating an apple in an hour.
Harry was sleeping for an hour.
*Harry was sleeping in an hour.5
The semantic function of the English progressive cannot be
described as simply converting a delimited reading into a
non-delimited one -- it is more complicated than that. I will
not attempt to decribe its "meaning' here. However, since
part of its function is clearly to change delimitedness into
non-delimitedness, I will constder it, for syntactic purposes,
to be an aspectual marker of delimitedness.
'Be -ing', has a different syntactic distribution from
tense/modality. Syntactic constraints on the occurrence of
tense/modality markers are irrelevant for aspect. English
5. This is a good sentence if it is understood inceptively, to
mean: 'Harry fell asleep within an hour'. This is not the use
of 'sleep' that I am considering here.
- 202 -
tense/modality must co-occur with a lexical subject or trace.
It cannot co-occur with PRO:
16
*Steve wants PRO goes to Vermont. PRES
*Steve wants PRO went to Vermont. PAST
*Steve wants PRO will go to Vermont. MOD
'Be -ing' may occur where tense may not:
17
Steve wants PRO to be going to Vermont. be -ing
The English perfective marker 'have -en' patterns together
with 'be -ing' in this respect. It is reasonable to assume
that it also belongs to the syntactic category of aspect.
This is not surprising, because it expresses non-indexical
'meaning', although its connection with delimitedness is less
straightforward.6
The relative ordering of tense/modality and aspect morphemes
provides some indirect support for separating aspect from
tense/modality (although it is not particularly strong
evidence in favor of the idea). The ordering is partly
explained if aspect and tense/modality are two separate
syntactic categories. When more than one of these markers
occurs together with tense/modality in a string of morphemes,
6. English aspectual verbs such as 'begin -ing' and 'finish
-ing' have a distribution similar to 'be -ing', and also
affect delimitedness. They may also belong to the category
aspect, but I will not discuss them.
- 203 -
tense/modality occurs at the ed~e of the string, and aspect
occurs closer to the verb:
18
Sung-hee has been going PRES - have-en - be-ing - go
Sung-hee had been going PAST - have-en - be-ing - go
Sung-hee will have been going MOD - have-en - be-ing - go
If tense/modality markers and aspect markers belonged to the
same syntactic category it would be possible (though not
necessary) for them to be interspersed in a string of
morphemes:
19
*Sung-hee have is going. have-en - PRES - be-ing - go
*Sung-hee have be goes-ing. have-en - be-ing - PRES - go
A general pattern is apparent in the distri-bution of
tense/modality and aspect: tense/modality is farther from the
verb, and aspect is nearer to it. Assuming there are no
discontinuous constituents, all elements of verbal morphology
belonging to one syntactic category should be separated from
all elements of verbal morphology belonging to another
syntactic category by a syntactic interface or boundary. The
distribution of tense and aspect morphemes discussed above is
consistent with this condition. Within that general pattern,
we see that aspect markers are not freely ordered, but
constrained within a fixed ordering. The fact that they have
a fixed order does not mean they cannot belong to the same
syntactic category. This ordering can come from independent
constraints not related to phrase structure and constituency.
- 204 -
It is commmon for verbal morphology to have a somewhat
arbitrary fixed ordering that is particular to the language.
There is more reason to unite 'have -en' and Tbe -ing' in one
category than to separate them into two.
The relative ordering of tense/modality and aspect morphemes
does not require that they be separate categories, but if they
are separate categories, some of the facts about their
relative ordering will follow.7
The data in this section demonstrates that tense/modality
and aspect belong to separate syntactic categories.
Furthermore, it favors an analysis in which tense/modality
takes syntactic scope over aspect, or aspect and verb together
make up a constituent which does not include tense/modality:
20
Sung-hee has gone. (PRES (haveeen go) )
Sung-hee was going. (PRES (be-ing go) )
On the basis of considerations such as these, Emonds (1976)
excludes 'be -ing' and 'have -en' from being generated in deep
structure under the category AUX (auxiliary). (Instead, in
Emonds' analysis, they are generated as syntactic main verbs,
7. If semantic or morphological merger only occurs under
conditions including adjacency at surface strucure, then this
view predicts that if tense and aspect ever merge in single
morpheme, that morpheme should be at the syntactic boundary
between the two categories.
- 205 -
under the VP node.) Emonds (1976) provides a precedent for
syntactically separating aspect from tense and modality. On
the basis of the data discussed in this section, any phrase
structure in which tense, modality and aspect are all grouped
together under the INFL node must be rejected.
5.2.3 Against a strictly featural account of aspect
It is posesible that aspect could be syntactically distinct
from tense/modality and still not be an independent category,
if aspect markers merely represent feature bundles attached to
a main verb. In this case aspectual markers could not claim
an independent node in phrase structure as they would if they
were a true syntactic category. It is clear that Tbe -ing'
must occupy a separate node from the main verb of a clause,
simply because it has lexical identity of its own. Adverbs
may intervene betweer "be' and the main verb:
20
Lisa was slowly walking down the road.
BE -ING
However, it may also be argued that some markers of
delimitedness that are bound morphemes also have independent
syntactic status from the main verb. Eckert (1984) observed
that similar aspectual markers in Russian and Czech can be
characterized as differing minimally in relative scope; The
Russian imperfective marker must be used whenever a verb is
used in an iterative context, regardless of the aspectual
- 206 -
class of the verb, and regardless of the fact thj.t the
iterative context is introduced by adverbial expressions. The
imperfective marker relates to the entire verb phrase. In
Czech, on the other hand, the perfective must be used when a
verb describing a delimited event is used in a repetitive
context. Since verbs in these languages may sometimes
describe delimited or non-delimited events, depending on
whether they are marked with the perfective or the
imperfective, aspectual scope ambiguities are possible.8
Russian sentences demonstrate an ambiguity that Czech
sentences do not have. The following example illustrates
this. The Russian verbs are both in the imperfective, and the
sentence has three possible interpretations. Both verb
phrases may represent delimited events, both may represent
non-delimited events, or the first may be non-delimited and
the second delimited. The Russian sentences have three
possible translations into Czech, because the Czech must
disambiguate this threefold ambiguity. Verbs are underlined
in the example below: (The examples in 21 are from Eckert
(1984).)
21
Russian:
Ja iz plena tri raza befal. I tri raza lovili.
IMF~~Imp
8. These Russian aspectual morphemes are a different set of
morphemes from those discussed in section 2.4.
-4207 a
Czech:.
a. Trikrat jsem utekl a trikrat me chytli.
"Three times I ran away and three times they
caught up with me."
b,. Tfikrat jsem utikal a trikrat me honili.
I~MW TMP
"I tried to run away three times and three times they
were after me."
V -'V0
c. Trikra't jeem utikal a trikrat me ch tli.
IMyP PERF
"Three times I tried to run away and three times they
caught up with me."
Because the perfective and imperfective markers in Czech are
sensitive to the delimitedness or non-delimitedness of the
individual verb, even in repetitive contexts, Czech does not
allow the range of scope ambiguities that Russian does.
The Russian imperfective marker may be reepresented as
taking scope over the entire verb phrase or V' projection.
Scope is represented in syntax by hierarchical phrase
structural relations. Extending this approach to the
aspectual markers, the natural representation of the Russian
imperfective would be something like 22a or 22b:
22a
?
asp VP
verb adverbs
22b
VP
asp VI
verb adverbs
- 208-
The corresponding Czech imperfective may be rerpesented as
taking scope only over the verb:
24
V'
asp verb
If the difference in relative scopes of aspectual markers such
as these is to be represented syntactically, the aspectual
markers must have syntactic reality, and it must be possible
for them to occupy independent syntactic nodes. Abney (1987)
and Baker (1986) argued that phonological and syntactic
word-hood do not necessarily coincide, It is not therefore
surprising to. find that some aspectual morphemes may be
represented as syntactic words. Lumsden (1987) argues that
all grammatical features (features associated with functional
categories) are syntactic features and are heads of syntactic
phrases. If this is true it is further support for the
plausibility of a syntactic category of aspect.
5.2.4 Possible instantiations of aspect in phrase structure
Adopting two plausible assumptions; that aspect is a
syntactic category, and that it conforms to the principles of
X-bar theory; there are several ways in which syntactic
markers of aspecu could be manifested in phrase structure.
These are illustrated below. (AGR is omitted from these
- 209 -
structures because it is irrelevant to the discussion.)
25
a) IP
NP I
INFL \
tns VP
mod /\
asp V'
V NP
b)
IP
NP I'
INFL \
tns VP
mod /\
Vi
asp-V NP
c) IP
NP I'
INFL \
tns ksP
mod /\
As
Asp VP
VI
V NP
d) VP
asp V 1
V NP
- 210 -
In 25a above, tense, modality and aspect are subsumed under
the single syntactic category of INFL (together with AGR). In
25b, aspect is distinct from tense and modality, but is not
syntactic. It is a bundle of features affixed to the verb,
and contained within the V node. In 25c aspect is a syntactic
category and a head, which takes the VP as its complement. In
25d aspect is a syntactic category that is not a head, but a
specifier.
Sections 5.2.1 through 5.2.3 have argued against phrase
structures 25a and 25b. However, there is another way to look
at phrase structure 25b. Aspectual features on the verb will
be percolated up the projectidn9 This is correct insofar as
the verb does contribute some aspectual features of its own to
the verb phrase. Recall that the verb phrase is a kind of
aspectual domain. Some aspectual markers have been analysed
as embedded main verbs (e.g., Emonds (1976)). If the features
on the head of the VP are aspectual, and the VP is an
adpectual domain, and 'be -ing' and 'have -en' are indeed
embedded verbs, it could be the case that aspect and verb
should be subsumed under a single category that heads a
phrase:
9. Lieber (1980) and Selkirk (1982) discuss feature
percolation.
- 211 -
25e
IP
NP I'
INFL \
tns AsP
mod /\
As'
Asp NP
25e is quite conservative; it looks like little more than a
renaming of the verb phrase as an aspect phrase. It is not
far removed from Emond's analysis. I leave it as an open
question whether this structure makes different predictions
from a structure in which a verb projects to a verb phrase.
We must consider the remaining possibilities, 25c, 25d and
25e. 25c and 25e are two possible ways in which the syntactic
category of aspect might head a syntactic phrase. 25c
predicts that an aspect phrase may be distinguished from a
verb phrase. To investigate this possibility properly it
would be necessary to find a language with aspectual
morphology -that tells upon the question. (English is not such
a language.) It is possible that 25c is an option in
universal grammar not fully instantiated in many or even most
languages.
25d, in which aspect is a specifier of the verb phrase, is
attractive because, as the Russian and Czech data illustrate,
some aspectual markers seem to have scope-taking
- 212 -
10properties.
This question of whether aspect is a head or a specifier is
parallel to the discussion of whether DET (determiner) is a
specifier for N or whether it heads its own projection.11
There are members of both the categories of aspect and
determiner that seem to have scope-taking or quantifier like
properties. Both Aspect and DET contain purely functional
information that is also associated with theverb or noun.
DET determines in large part whether the NP or DP will be mass
or count, and Aspect determines whether the verb phrase or
aspect phrase will be delimited or non-delimited. This is so
even though there is information about the mass/count or
delimited/non-delimited distinction in the noun or verb
itself. Whether aspectual markers of delimitedness are heads
of phrases or specifiers is a question that is beyond the
10. Peggy Speas has suggested another structure in which
aspect is not a head or a complement. This is an adjunction
structure like the following:-
VP
VP
as p
V
V NP
It is not clear how this structure would differ from -25d in
the predictions it makes. I leave this as an open question.
11. See Abney (1987), Fukui and Speas (1986), and Speas (1986)
on functional categories.
- 213 -
scope of this thesis to answer, although it is appropriate to
pose it here. To answer this question it will be necessary to
investigate the interaction of particular aspectual.
morphologies with the theory of functional categories, binding
theory and bounding theory. This is a timely question,
coinciding as it does with the research that is developing on
the nature of functional categories.
- 214 -
5.3 Aspectual particles, resultative secondary predicates,
and double object constructions
Aspectual particles, resultative secondary predicates, and
the dative arguments of double object constructions have much
in common syntactically. In this section the common syntactic
behaviors of these elements are reviewed, and considered in
light of the aspectual theory of argument structure developed
in Chapter 4. I hope to accomplish two purposes in this
section: first, to show that aspectual princples unify these
three kinds of syntactic elements and account for some of
their behavior; and second, to illustrate how an aspectual
theory of argument structure may be brought to bear on a
syntactic question. For the second project I will concentrate
on double object constructions. I will not develop an
original analysis of these constructions, but will discuss and
choose between existing solutions in the literature on the
subject. The conclusions reached will be few and tentative,
but sufficient to illustrate how aspect may be employed as a
tool in syntactic investigation. The aspectual theory
advanced here allows interesting questions to be posed, most
of which I will have to leave for further research.
Aspectual particles and resultative secondary predicates
have been discussed in Chapter 2. Double object constructions
have- not yet been discussed, and will be introduced here.
Double object constructions are exemplified by 26a below.
- 215 -
They are distinguished from prepositional datives as
exemplified in 26b:
26
a. Margaret sent Mary a package.
b. Margaret sent a package to Mary.
These constructions have received a great deal of attention in
the literature. A variety of analyses have been suggested for
them, focusing on the questions of (i) what is the syntactic
structure of 26a, and (ii) how are sentences like 26a and 26b
related? 26a and 26b differ crucially in that the dative
object is governed by the verb in 26 a but not in 26b. There
are three general ways in which the two constructions may be
related:
(i) the dative object (but not the accusative object) is base
generated in a position where it is governed by the verb, and
moved to a position where it is not (26a is the basic
form, and 26b derived),
(ii) the accusative object (but not the dative object) is base
generated in a position where it is governed by the verb, and
moved to a position where it is not (26b is the.basic
form, and 26a derived),
(iii) each form is base generated, and no derivational movement
necessary (26a and 26b are both base
generated).
These are general sketches of three possible approaches.
Naturally there are many refinements and variations on these
approaches developed by the authors who have worked on these
constructions. These variations have to do with explaining
the presence or absence of the preposition, or postulating
- 216 -
intermediate derivational structures. Several types of rules
inserting or deleting prepositions, or converting NP's into
PP's and vice versa, have been employed. I I will not discuss
these.
All three of these general approaches are represented in the
literature. Emonds (1972) argued that 26b was the base
generated form and 26a was derived from it by a
transformational rule permuting the dative and accusative
arguments. Qerhle (1976) and Kayne (1981) proposed that 26a
and 26b were both base generated. Fillmore (1965) adopted an
analysis in which the prepositional dative is derived from the
double object construction.2 -Chomsky (1975) and Larson (1987)
adopt analyses in which the base generated form is closer to
26a than 26b, but the relationship between the two forms is
more complicated in their analyses.3
1. It has come to my attention that Stowell (1981) also
adopted an analysis in which the double object is derived from
the prepositional dative.
2. Fillmore (1965) proposes a deep structure like 26a for 'to'
datives and particles, but not for 'for' datives.
3. Chomsky (1975) proposes a deep structure in which the
dative object forms a constituent with the verb:
the teacher -- gave to him - several books
In Chomsky (1975) the prepositional dative (a below) is
derived from the sentence above, and the double object (b
below) derived in turn, from the prepositional dative.
a. the teacher gave several books to him
- 217 -
This quick overview to the literature on double object and
prepositional dative constructions has focused on the deep
structure configurations ascribed to these constructions. In
Chapter 4, a correlation between deep structure position and
aspectual properties was proposed, and it is this correlation
that is the topic of this chapter.
Six unifying characteristics of aspectual particles,
resultative secondary predicates, and dative arguments of
double object constructions are described below.
Generalizations 1 through 4 are more or less familiar to
linguists working on these issues. Generalizations 5 and 6
concern the aspectual properties of these constructions, and
they are contributed by the view of aspect developed in this
thesis.
1. Each one of these constructions requires a postverbal NP
or accusative object. The aspectual particles are underlined
b. the teacher gave him several books
Larson (187) proposes a deep structure like the following:
NP V'
a letter
V PP
send to Mary
Both the double object and the prepositional dative
constructions are derived by Larson from this structure by
o.perations such as verb raising amd adjunction.
- 218 -
in the sentences below:
27
a. John ate an apple up.
John pushed the table over.
b.*John ate .4
*John pushi4 over.
Sentences in which there is an aspectual particle but no
postverbal NP involve unaccusatives or middles, and these have
a NP trace in the postverbal NP position.5
28
a. Martha dried her socks out.
b. The socksi dried ti out.6
c. The socksi dry t i out-easily by hanging them up.
d. *Martha dried out. (where Martha is drying something out)
In 28a there is an overt postverbal NP, just as in the
preceding set of examples. IU the unaccusative (28b) and the
middle (28c), there is a NP trace in postverbal position. In
(28d) there is neither a NP trace or an overt NP, and the
sentence is bad.
4. This is good for some speakers if interpreted with a
deleted object, in which case there is still an understood
postverbal NP.
5. The movement analysis for unaccusatives and middles is
reviewed in Chapter 6.
6. I am not taking a stand on whether the NP trace is before
or after the particle. It is placed before the particle
merely for convenience.
- 219 -
The same .is true of resultative secondary predicates. They*
require postverbal NP's which may be overt;
29
John scrubbed the floor clean.
John wrung his towel dry.
or they may be traces associated with unaccusatives;
30
The lakei froze ti solid.
The river drained ti dry.
or with middles:
31
Johni scrubs ti clean easily if you can hold him down long
enough. (Said of a small child who hates to be scrubbed.)
But resultative secondary predicates without a postverbal NP
or trace are ungrammatical:
32
*John scrubbed clean. (where John scrubbed something clean)
Finally, the dative argument of a double object construction
reqgires the presence of the accusative argument. The dative
arguments are underlined in 33 below. The accusative
arguments are 'a letter' and *a package' .
7. Just as with the aspectual particles, there are acceptable
constructions in which the accusative argument is understood:
'Kazuko' read to Taroo' means that Kazuko read something to
Taroo.
a 220 -
33
Tifa sent a letter to Muhend.
*Tifa sent to Muhend.
Kazuko mailed a package to Taroo.
*Kazuko mailed to Taroo.-
2. These three elements, with some minor qualifications, may
occur on either side of the postverbal NP or accusative
object. This is clear in the case of aspectual particles:
34
Mario ate an apple up.
Mario ate ap an apple.
Murry pushed the table over.
Murry pushed over the table.
Marcus dried his socks out.
Marcus dried out his socks.
and it is clear for dative arguments:
35
Tifa sent a letter to Muhend.
Tifa sent Muhend a ItTer.
Kazuko mailed a package to Taroo.
Kazuko mailed Tar oo a package.
Resultatives require a minor qualification. They are
generally not as good before the postverbal NP (36b) as after
it (36a). (The examples in (36a) are from Carrier-Duncan and
Randall (1987).)
36
a. The gardener watered the tulips flat.
The baker beat the eggwhites int-iEiff peaks.
- 221 -
b. ??The gardener watered flat the tulips.
*The baker beat into stiff peaks the eggwhitee.
However, the sentences in 36b above improve greatly when the
postverbal NP is heavily weighted phonologically and
informationally:
37
The gardener watered flat the tulips that she had planted
the week before andT2 not expected would ever come up.
?The baker beat into stiff peaks the eggwhites that his
assistant had made a special trip to a neighborhood farm
to get fresh eggs for.
Assuming the resultatives in postverbal position are not
prohibited by strict principles of grammar, but by priniciples
of felicity in discourse or style such as those that govern
Heavy NP Shift, resultatives, like aspectual particles and
dative arguments, may appear before the postverbal NP or
accusative object.
3. Particles, resultatives, and dative objects must be
unique in the verb phrase. The grammatical (a) and (b)
sentences below contain one of these syntactic elements. In
the ungrammatical (c) examples there are two of these
elements.
38
a. John ate the apple up.
b. John ate the apple _ rough.
c. *John ate the apple _p through.
a. John pushed the cart over.
b. John pushed the cart away.
c. *John pushed the cart over away.
- 222 -
39
a. John scrubbed the floor clean.
b. John scrubbed the floor dfry.
c. *John scrubbed the floor clean dry.
a. The lake froze solid.
b. The lake froze hiard.
c. *The lake froze solid hard.
a. The gardener watered the tulips flat.
b. The gardener watered the tulips sopping wet.
c. *The gardener watered the tulips flat sopping wet.
40
a. Tifa sent a letter to Muhend.
b. Tifa sent a leter toTEF7r.
c. *Tifa sent a letter~toMii-end to Izir.
4. The verb and the particle, resultative or dative object
are more tightly related thematically than the verb and the
postverbal NP or accusative Object. This was observed by
Emonds (1976) with respect to a verb and its dative object.
(The examples in 41 are from Larson (1987).)
41
Mary took Felix to the cleaners.
to task.
into consideration.
Felix threw Oscar to the wolves.
In these examples, the verb and its dative object jointly
select the accusative object. The existence of many such
idioms consisting of a verb and dative object that together
select the accusative object is evidence that a verb and its
dative object together assign thematic roles to the accusative
object.
Aspectual particles are quite productive, but it is not the
-223 -
case that any particle can occur with any particle-taking
verb. There are selectional restrictions between the verb and
the particle:
42
eat an apple up
*eat an apple out
?eat an apple through
?dry the socks up
dry the socks out
*dry the socks through
*hear the story up
hear the story out
hear the story through
Moreover, the verb and particle together select the postverbal
NP, rather than the verb and postverbal NP selecting the
particle. This is a subtle effect to test for, because the
range of 'meanings' contributed by the aspectual particle to
the linguistic expression is quite limited. These aspectual
particles require that the postverbal NP argument is entirely
affected, but 'they may express slightly different ways of
being entirely affected. 'Up' in 'dry up' implies a complete
drying of the surface of the object, while, 'out' in 'dry out'
has more of a sense of drying the object completely from
inside to outside. 'Up' and 'out' in concert with 'dry' can
select for different accusative objects:
45
*dry the.socks up
dry the socks out
dry the floor up
??dry the floor out
- 224 -
This is much easier to see with resultatives. First of all
it is clear that there is a thematic interaction between the
verb and the resultative:
44
water the tulips flat
?water the tulips rough
*water the tulips dry
wring the towel dry
?wring the towel flat
*wring the towel cold
Secondly, a verb and resultative together select the
accusative object. If the resultative is not incompatible
with the verb, then it selects the accusative object in
concert with the verb:
45
water the tulips flat
?water the sidewalk flat
?water the tulips shiny
water the sidewalk shiny
5. Particles, resultatives and dative objects contribute to
delimiting the event described by the verb phrase. Recall the
discussion in Chapter 2 of aspectual particles and
resultatives:
46
John dried his socks. (delimited non-delimited)
John dried his socks out. (delimitedS
John ate an apple. (delimited non-delimited)
John ate an apple R. (delimited)
John pushed the cart. (non-delimited)
- 225 -
John pushed the cart over. (delimited)
John scrubbed the floor. (delimited, non-delimited),
John scrubbed the floor clean. (delimited)
John pressed the pages. (delimited, non-delimited)
John pressed the .pages flat. (delimited)
This is not quite as obvious for dative arguments of double
object constructions, but it can be demonstrated. Culicover
(1982) observed that 'for' datives which are ambiguous between
a benefactive and a recipient reading (47a) are only
interpretable as recipients when expressed in the double
object construction (47b). (Examples 47 and 48 are from
Carrier-Duncan and Randall (1987), p 43-44.)
47
a. Max baked a cake for Felicia.
William wrote a sonnet for Felicia.
Alphonse knitted that nosewarmer for Felicia.
b. Max baked Felicia a cake.
William wrote Felicia a sonnet.
Alphonse knitted Felicia that nosewarmer.
Furthermore, they show that benefactive 'for' datives may not
appear in the double object position:
48
Jeeves drove that car for Madame.
The butler swept a room for the chambermaid.
Jack ate some spinach for his mother.
*Jeeves drove Madame that car.
*The butler swept the chambermaid a room.
*Jack ate his mother some spinach.
A -recipient argument differs from a benefactive argument in
the way it is understood to participate in a linguistically
- 226 -
described event. In the benefactive reading of 'Max baked a
cake for Felicia', we understand that Max has Felicia in mind
when he bakes the cake, but she does not necessarily receive
it. The argument 'Pelicia' does not contribute to delimiting
the event. However, in the recipient reading, Felicia is
understood to receive the cake. Felicia's receipt of the cake
marks a point of time in the described event -- a point of
time that delimits the event.8
The constraint that the dative object in a double object
construction must delimit the event explains several of the
special semantic properties associated with these objects.
Many of these are noted by Grden (1974). She observes that
with double object constructions, the dative and accusative
objects are understood to exist at the same time as each
other, in some relevant sense. (Examples 49 through 51 are
from Green (1974).)
S. In the author's judgement there is a possible reading of
'Max baked Felicia a cake' in which Felicia does not actually
receive the cake. However, there is still a distinction
between a) and b) below with regard to this benefactive-like
reading:
a. Max baked Felicia a cake.
b. Max baked a cake for Felicia.
In a), more than in b), Max believes or intends Felicia to
receive the cake. That intent supplies a standard which must
be met in the baking of the cake, and contributes to
delimiting the baking event. .
- 227o-
49
a. The American ambassador baked a cake for James I.
b. *The American ambassador baked James I a cake.
49a can mean the ambassador baked a cake for James I long
after James I was dead and gone, so that the cake came into
existence after James I had departed. The cake might have
been baked in James I's honor. 49b, however, cannot mean
this, and is pragmatically impossible because James I never
co-existed with cakes baked by American ambassadors. 49b must
be understood td mean that James I and the cake were extant at
the same time (or were believed to be by the ambassador --
this is the 'relevant sense' mentioned above). Likewise with
50:
50
a. She's going to sing a song for her late lover.
b. *She's going to sing her late lover a song.
50b is odd because the late lover cannot be understood to
exist when the song is sung. The late lover therefore cannot
be a true recipient of the song. This requirement does not
hold of 50a, which is not a double object construction, so 50a
is pragmatically interpretable.9 Finally, 51a might be said
9. Green (1974) mentions a double object verb, 'owe', which is
at first glance aberrant in this respect. The question is,
does the expression,
owe Bill forty dollars
describe an event measured out by forty dollars and delimited
by Bill? Forty dollars is measurable on a scale (even though
that scale is not necessarily 'spread out' over time) because
a 228 -
by someone who did not believe in Santa Claus, but 51b would
not:
51
a. Did you really write a thank-you note to Santa Claus?
b. Did you really write Santa Claus a thank-you note?
The semantic requirements on the dative object in a double
object construction --- that it must be understood to exist,
to be a recipient, and to coexist with the accusative object
-- can be subsumed under the single requirement that the
dative object delimits the event. We see that a necessary
(though not sufficient) condition for double object
constructions is that the dative argument contributes to
delimiting the event described by the verb.
There is another important semantic requirement on dative
objects of double object constructions. They must be animate
or sentient:
52
a. send a telegram to Bill
b. send Bill a telegram
53
a. send a telegram to Tokyo
b. *send Tokyo a telegram
it is possible to owe Bill $37, $38, or $39 dollars. Bill
does delimit the event (in a relevant sense) if he is
undetstood to be a hypothetical recipient of a $40 which is to
change hands. Viewed in this way, 'owe' is similar to other
double object verbs of this class identified by Green:
'bequeath', 'leave', 'offer', 'allot'. This is an
illustration of how verb meanings are constructed around a
core meaning, within a certain range of variation.
- 229 -
I
53b is considered impossible by most speakers, because the
dative object is inanimate. There is an acceptable reading
for 53b, however, in which %Tokyo' represents someone or
something that is cognizant of the arrival of the telegram --
a bureacrat in an embassy or government office, perhaps. The
requisite property is not animacy per se, but the ability to
undergo a change cf state that delimits the event, while being
a bystander or indirect participant in thL event. The fact
that this ability is naturally associated with emotional or
mental states explains why these objects are usually animate,
but this fact is not a linguistic fact. The view I am
advocating is that only the aspectual requirements on the way
the dative object participates in the event are linguistic
facts.
6. The postverbal NP's or accusative objects associated with
particles, resultatives and dative objects are capable of
measuring out and possibly delimiting the event described by
the verb. In this way they act like canonical direct
arguments. As we have seen, with particles the postverbal NP
may receive a totally affected interpretation in the absence
of a particle. The particle merely serves to disambiguate the
sentence:
54
a. eat an apple (delimited non-delimited)
b. eat an apple up (delimitedS
In the delimited reading of 54a above it is the boundaries of
- 230 -
the accusative object, apple, that provide the boundary of the
event as described by the verb. The same discussion applies
to resultatives. A totally affected interpretation of the'
accusative object is possible in the absence of the
resultative:
55
a. water the tulips
b. water the tulips flat
a. scrub the floor
b. scrub the floor clean
In the totally affected or delimited readings of the examples
above the action is understood to progress through the object
argument in the course of the-event, until it has applied to
the entirety of it. 10
The accusative argument in double object constructions
either delimits the event, or provides the scale on which it
may be delimited. In this way the accusative argument has the
canonical semantic/aspectual properties of a direct argument.
10. Even when, in the normal use of the verb, the accusative
object does not delimit the event, it does do so when there is
a resultative. For example, a sign advertising lead-free gas
in White River Junction, NH, says:
Drive your engine clean.
- Although 'engine' in 'drive your engine' is unaffected, so
that the sentence describes a non-delimited event, with the
resultative 'clean' the engine is understood to be affected
and to measure out the event. Some change in the engine
. registers the end of this event.
- 231 -
The verbs in 56 are double object verbs, but they may also
appear with only an accusative object. When they do, the verb
phrases express delimited events:
56
sell (Jim) a car
award (Jim) a prize
tell (Jim) a story
wire Jim the news
knit Jim a sweater
The events of selling a car or awarding a prize are delimi.ted
by a change of state of the direct arguments, the car and the
prize. (The change of state is a change in ownership.) The
events of telling a story, wiring the news, and knitting a
sweater-are delimited by the -event's progression through the
spatial or temporal extension of the story, the news, or the
sweater.
The verbs in 57, also double object verbs that are
grammatical with a sole accusative argument, have accusative
arguments that measure out the event through their change of
location:
57
Push (me) the wheelbarrow.
Lower (me) the rope.
The events of pushing the wheelbarrow and lowering the rope
are non-delimited events. However, they become delimited
events when there is a recipient (dative object) that marks
the end of the change in location of the moving object
- 232 -
(accusative object).-
Some double object verbs allow an ambiguity between a
reading in which the accusative object delimits the event, and
one in which the accusative object measures out the event but
the event is delimited by the dative object:
58
a. mail Charles a letter
b. throw Charles a ball
58a describes a delimited event, but the event is ambiguously
delimited. The event may end when the letter has been taken
to the post office, stamped, and dropped into the mailbox on
its way to Charles. Or it may end when Charles actually
receives the letter. The first interpretation is selected by
59a, the second by 59b:
59
a. It took two hours to mail Charles a letter at the post office,
b. It took two weeks to mail Charles a letter in Alaska.
The same thing is true of 58b. One ball-throwing event (one
throw of one ball) is a delimited event, but it may be
delimited in-two ways. The'event may be understood to include
everything up to the launching of the ball from the thrower's
hand. winding up the pitching arm, building up momentum,
releasing the ball. Or it may be understood to include the
progress of the ball until it achieves some distance or
arrives at some place. The first reading is emphasized in
60a, and the second is emphasized in 60b:
- 233 -
60
a. throw the ball with a jerky motion
b. throw the ball seven miles
Although the verb phrases in 58 describe ambiguously
delimited events, they must be interpreted, on any one
reading, as delimited in only one way:
61
*mail Charles a letter in two hours in two weeks
??throw Charles the ball with a jerky motion seven miles
??throw Charles the ball seven miles with a jerky motion
This is consistent with the aspectual principle of argument
structure advanced in Chapter 4, that states there may only be
one *delimiting' associated with one verb phrase.
The following generalizations about constructions involving
particles, resultatives and datives of double object
constructions have been stated in 1 through 6:
1)-These constructions all require a postverbal NP or
accusative object.
2) Particles, resultatives and dative objects may occur on
either side of the postverbal NP or accusative object.
3) Particles, resultatives and dative objects must be unique
in the verb phrase.
4) Particles, resultatives and dative objects are more
tightly thematically related to the verb than to
the postverbal NP or accusative object.
5) Particles, resultatives and dative objects delimit the
event described by the verb phrase.
6) The postverbal NP or accusative object measures out (and
- 234 -
may delimit) the event in the canonical fashion of direct
arguments.
The first point to be made concering these generalizations
is a reiteration of the central point of Chapter 2. These
constructions having similar aspectual properties also have
similar syntactic behaviors. Aspectual properties have
syntactic reality. We may go even further than this by
incorporating the aspectual principles of argument structure
advanced in Chapter 4 into the discussion. The aspectual
properties of these constructions actually explain some of
these syntactic behaviors.
First, the facts in 1 and 3 are not surprising in view of
the facts in 5 and 6. These show that the postverbal NP or
accusative object plays the canonical direct argument role of
providing the scale on which the event is measured out; and
the particle, resultative or dative object delimits the event
on that scale. The particle, resultative or dative object
(the delimiting element) requires the scale provided by the
postverbal NP or accusative object (the direct argument) in
order to delimit the event. These elements cannot delimit the
event independently of that scale, and therefore a sentence
containing only a delimiting element and not a direct argument
is uninterpretable. This results in generalization (1).
Generalization 3 is explained by the aspectual constraint dn
argument structure that requires that there be no more than
one 'delimiting' associated with a verb phrase. A verb phrase
- 235 -
with more than one particle, resultative or dative object
would have more than one 'delimiting'.
Viewed in light of the aspectual principles of argument
structure, generalizations 1, 3, 5 and 6 fit together into a
coherent, integrated picture of the argument structures of
these three constructions. Generalizations 1 and 3 are not
surprising from the point of view of the syntax of double
object constructions, because we expect arguments to be unique
and obligatory. But the syntax of particle and resultative
constructions provides less reason to expect these
generalizations to hold, since these elements are not
generally thought of as arguments. With the aspectual theory
of argument structure, 1 and 3 are natural corollaries of the
aspectual properties of particles and resultatives; and
particles and resultatives are unified with double object
constructions in a simple way.
I have not provided an aspectual perspective on
generalization 2 -- the fact that the delimiting element
occurs on either side of the accusative object. This
generalization is at the root of one of the central questions
about double object constructions: how are they related to
prepositional dative constructions? There is a large body of
literature addressing this question from the syntactic point
of view, and many syntactic issues bear upon it which I shall
not go into due to limitations of time and space. However, I
- 236 -
think it is reasonable to expect that aspectual issues will be
found to have considerable bearing on this question. A proper
elucidation of the aspectual implications for 2 will have to
await further research.
The syntactic and aspectual properties of particle,
resultative and double object constructions underscore the
relevance of aspect to syntax, argued for in Chapter 2.
However, merely demonstrating the fact that there is a
connection between aspectual properties and syntax does not
make aspect a useful tool for syntactic research. Some
minimal theory of that connection is necessary, through which
productive questions may be posed. The aspectual principles
of argument structure introduced in Chapter 4 do just that, by
proposing an inherent correlation between the aspectual
properties of an argument and its structural position.
Internal and external arguments, which have distinct aspectual
properties, are structurally distinguished at deep structure.
Since the arguments of unaccusatives, which- are deep structure
objects, share the aspectual properties associated with
canonical object positions (government by the verb), that
correlation is proposed to hold at deep structure. This is in
the spirit of regarding deep structure as the pure
representation of thematic relations, and capitalizing on the
connections between aspectual structure and thematic structure
discussed in Chapter 4. Continuing this line of approach, the
interesting question about double-object constructions for an
- 237 -
aspectual theory such as this is: what are the deep structure
positions of the two arguments (and particularly the
accusative arguient) in double-object constructions? Since
the accusative object or postverbal NP has the canonical
aspectual properties of a direct argument, we are interested
in its structural position at deep structure.1 1
The double object construction in 62 has five possible deep
structures represented in 63:
11. One important issue must be noted. I have assumed that
direct argument-hood is defined at deep structure, and argued
that aspectual properties associated with certain structural
positions are likewise defined at deep structure. (The bulk
of this argument rests on unaccusatives, which are discussed
in Chapter 6.) This view will have serious consequences for
analyses.of double object constructions. The logical
consequence of this view is that prepositional datives and
datives of double object constructions must. be base generated,
since they differ aspectually. The Barse-Lasnick data would
be difficult to account for under this view. There are some
arguments in favor of defining direct argumenthood at
S-structure, although for the reasons mentioned above, I do
not adopt that approach here.
- 238 -
62
give Mary a b'ook
63
(a)
VP
V NP NP
give Mary a book
(b)
VP
V . S
give 7 \
NP NP
Mary a book
(c)
VP
V PP NP
give /N a book
P NP
e Mary
(d)
V
give
(e)
VP
V\ /NP
a book
NP
Mary
VP
V' PP
to Mary
NP
e a book
a 239 -
V
giv
63a is the base generated structure proposed by Qerhle
(1975). 63b has been proposed by Kayne (1983). In 63b, the
two arguments form a constituent in a small clause structure.
This has certain syntactic consequences exploited by Kayne
involving the government of traces in double-object
constructions, and the substitution of an 'unambiguous paths'
condition for the conventional statement of c-command. (For
details see Kayne (1983).) In an earlier paper Kayne (1981)
proposed a structure like 63c, in which-an empty preposition
precedes the dative object. Kayne derives certain differences
between French and English dative constructions by this
means. In 63d, the verb and the dative object form a
constituent at deep structure. This deep structure
constituency has been proposed in some form by a number of
authors, including Fillmore (1965), Chomsky (1975), Bach
(1979), Larson (1987), Jacobson (forthcoming).12 63e was
assumed or argued by Emonds (1976) to be the base generated
form from which the double object construction was derived by
transformation.
Note that NP-movement (or transformations in the older view)
12. There is variation among these authors as to whether they
include the preposition in the deep structure. The important
thing here is that they all support the idea that the verb and
its dative object is a more basic constituent at deep
structure than the verb and its accusative.object.
- 240 -
need not apply to 63a through 63d to derive the surface word
order. However, certain binding facts noted by Barss and
Lasnick (1986) seem to indicate that the accusative object is
in the binding domain of the dative object, while the converse
is not true. Some kind of movement would have to apply to 63a
through 63d if the c-command relations indicated by these
binding facts were to be derived at surface structure.
(Larson (1987) proposes one possible means of accomplishing
this, involving verb raising and NP-movement.) Some kind of
movement would also be necessary to derive the surface word
order in a double object construction from a deep structure
such as 63e. I will not address the question of how these
possible deep structures in 63 are to be related to the
surface structure of a double object construction, although a
plausible account of this would be necessary in order to
firmly adopt one of these structures. I restrict myself to
considering whether there is any aspectual evidence for any of
these deep structure constructions.
Assuming a correlation between deep structure position and
aspectual properties, the accurative object must be
structurally distinguished because it is aspectually
distinguished. It has the canonical aspectual properties of a
direct argument. This should rule out 63a and 63b, in which
there is no structural asymmetry between the accusative and
dative objects. In 63a, there is an asymmetry between the
objects insofar as the dative object is adjacent to the verb,
- 241 -
while the accusative object is not; but this distinguishes the
wrong argument, since it is the accusative NP that we expect
to be structurally distinguished.
We are left with a choice between 63c, 63d, and 63e. Recall
generalization 4:
Particles, resultatives and dative objects are more
tightly thematically related to the verb than to
the postverbal NP or accusative object.
This fact would seem to indicate that the verb and the dative
object form a constituent at deep structure. If deep
structure positions reflect thematic structure, then we should
expect the verb and the dative argument to be sisters at deep
structure. This provides some independent reason for
selecting 63d as the deep structure of double object
constructions. If this is true, then the canonical structural
position for direct arguments is not necessarily sister to the
verb, but highest NP object dominated by the VP node. There
is some support for this from aspectual considerations. A
deep structure proposed by Belletti and Rizzi (1986) for
certain psych verbs in Italian has both of the verb's
arguments under the VP, with the argument having the aspectual
properties of a direct argument occupying the higher position
of the two:
Belletti and Rizzi propose that sentences such as,
- 242 -
64
Questo preoccupa kianni.
"This worries Gianni."
have underlying structures like the following:
65
S
NP VP
V' NP
V 0NP
ec preoccupa questo Gianni
For reasons that will be discussed in greater depth in section
6.5, the experiencer argument-'Gianni' measures out and
delimits the event, whereas 'questo' does not do so. The
experiencer must therefore be the direct argument in sentences
such as these. Belletti and Rizzi propose a generalization in
the form of a thematic hierarchy, that experiencer arguments
like 'Gianni' must be 'higher' in the phrase structure than
arguments like 'questo' .13 As a thematic hierachy, this is
merely a stipulation, but as part of a larger generalization
including dative objects of double object constructions, it is
evidence for the following:
13. Belletti and Rizzi refer to 'questo' here as a theme, but
since I do not believe that is an appropriate name for the
thematic role of this argument, I leave it un-named.
- 243 -
66
The direct or internal argument, the argument that is capable of
measuring out and delimiting the event, is the highest NP object
under the VP node.
66 is proposed as a hypothesis for future research.14
14. The argument structure selected above also allows us to
speculate about the special semantic properties of datives of
double object constructions. Recall that they are usually
animate, because they must delimit the event without being
acted upon directly. In this structure the dative object is
sister to the verb:
VP
V1 NP
a book
V NP
give Mary
It may be that the position as sister to the verb requires the
dative object to be understood as undergoing a change of state
(and thus delimiting the event); but since the role of true
direct argument is taken by the higher accusative argument,
only the accusative and not the dative can 'measure out' the
event. The dative argument is in a position where it must
delimit but may not measure out the event by being acted upon
directly. It would seem that this hybrid object-like
aspectual position is accompanied by hybrid object-like
aspectual properties. I leave this as a promising line for
further -investigation.
- 244 -
Chapter 6. Asvect and the Mapving of Cognitive Structurg
into vntax
6.1 Introduction. The Aspectual Interface Hypothesis
The nature of the interface between semantics and syntax is
a central problem in linguistics, and has been the source of
lively discussion and disagreement among linguists working in
this area. Of particular importance and currency is the
nature of the interface between syntax and what has been
variously called cognitive, thematic, or lexical semantic
structure -- loosely speaking, the 'meaning' of words. I will
refer to it here by the relatively neutral term of 'cognitive
structure'. What I mean by cognitive structure is the
representation of the 'meanings' of words in some appropriate
system that captures a native speaker's knowledge of the
lexical semantics of his or her language. The relationship
between syntax and cognitive structure has been widely
discussed in the literature, but there is disagreement about
whether they must constitute separate levels of representation
in a model of natural language; or if so, whether only two
levels are adequate instead of three or more. Some authors
maintain that cognitive structure translates directly into
syntax. Under this view the lexical 'meaning' of predicates
1. Fillmore (1968) adopts this view. This is also the
approach taken by generative semantics. See Lakoff (1971),
McCawley (1971 ) and Postal (1969).
- 245 -
entirely determines the manifestation of predicates and their
arguments in syntactic structures. Others hold the view that
the two levels are not reducible to one, and there is a
mapping between these levels that is, if not entirely
arbitrary, governed at least partly by thematic hierarchies or
linking rules.2 These are rules that predict how various
thematic roles, such as agent, theme or goal, will map into
external, internal or oblique argument positions in the
syntax. I will adopt the second view, which holds that
cognitive and syntactic structures must be fundamentally
distinct, and the cognitive representation of a predicate (its
'meaning') does not absolutely determine how its arguments
will be arranged in the syntax. I will not argue for this
view here.
Notwithstanding the separateness of syntactic and cognitive
structure, there is undeniably enough correlation between
'meaning' and syntactic structure to call for some explanation
of the connection between the two levels of representation.
Thematic hiearchies and linking rules provide a mechanical
description of the correlation, but rules such as these do
little to explain why, for instance, an agent is always found
2. Gruber (1965), Jackendoff (1972), Kiparsky (1985), Hale and
Keyser (1986, 1987) take this view.
3. Although I believe thematic roles are an inadequate means
of expressing the cognitive structure of a predicate, I will
use them as a convenient shorthand for the time being.
- 246 -
in an external argument position and a theme always becomes
the direct argument of a predicate. I will argue in this
chapter that the aspectual properties associated with
external, internal and oblique arguments outlined in Chapter 4
are sufficient to account for the apparent correlation between
syntactic argument structure and cognitive structure. I will
take the strongest view and state it in the strongest way --
that these aspectual properties of syntactic argument
structure are all that needs to be said of the mapping from
cognitive structure into syntax.4 Chapter 3 has shown that
affectedness, a semantic or cognitive property with syntactic
consequences, is actually an aspectual property; and Chapter 5
has demonstrated that the syntax of particles, resultatives,
and double object constructions is in part explained by the
aspectual properties of these constructions. These ideas
carried to their logical extreme suggests the following
hypothesis: .
Aspectual Interface Hypothesis IH)
The mapping between cognitive structure and syntactic
argument structure is governed by aspectual properties. Only the
aspectual part of cognitive structure is visible to the syntax.
This means that of the variety of information contained in a
thematic role such as agent, patient, actor, goal, etc., only
4. Although I refer to a mapping from cognitive structure into
syntax I do not mean to imply any necessary directionality in
the mapping.
- 247 -
the apectual information is relevant to syntax. The
prediction is made, that thematic hierarchies instrumental in
the mapping between thematic or cognitive structure and syntax
depend on aspectual properties.5 Furthermore, the AIH may
provide the basis of some theoretical underpinnings for
statements such as the Universal Theta Assignment Hypothesis,
formulated by Baker (1985) in the framework of Government and
Binding; and the Universal Alignment Hypothesis, formulated by
Perlmutter and Postal (1984) in the framework of Relational
Grammar.
Universal Theta Assignment Hypothesis (UTAH)
Identical thematic relatidnships between items are
represented by identical structural relationships between
those items at the level of D-structure.
Universal Alignment Hypothesis (UAH)
There exist principles of universal grammar which predict
the initial relation borne by each nominal in a given clause
from the meaning of the clause.
The UTAH and the UAH propose a strict correlation between
meaning' and structure. The aspectual constraints on
argument structure suggest a mechanism by which such a
correlation may be mediated.
5. The Aspectual Interface Hypothesis does not say that all
semantic relations, processes or operations in syntax depend
on aspectual properties. Predication, modification,
complementation, quantification, etc. do not; except insofar
as they interact with aspectual constraints on argument
structure.
- 248 -
A thorough and definitive demonstration of the validity of
the Aspectual Interface Hypothesis is beyond the scope of this
thesis. In this chapter I will merely attempt to demonstrate
its plausibility. The AIH will stand as a directive for
future research.
I will not develop or argue for any particular view of or
mode of representation of cognitive structure. A variety of
systems are in the literature, including: Gruber's system of
pre-lexical categorial structures (Gruber (1965));
Jackendoff's functional structures based on primitive semantic
functions (Jackendoff (1972)); Dowty's system of lexical
decomposition of verbs (Dowty-(1979), and see Appendix to
Chapter 1);6 systems based on thematic roles, such as
employed in Chomsky (1981) and elsewhere; and Hale and
Keyser's Lexical Conceptual Structure (Hale and Keyser, (1986,
1987)). Each of these may be partly on the right track. I
assume that some sort of representation of cognitive structure
is needed, but I will only compare these options, or speculate
on what sort of representation is the correct one, insofar as
the aspectual correlates of syntactic structure place demands
on this representation.
Four principles of argument structure have been advanced in
6. Talmy also exlores the possibilities of lexical
decomposition. See Talmy 1985).
- 249 -
Chapter 4:
(i) If an internal direct argument undergoes change during
the event described, that change is characterizable as a
change in a single parameter or change on a scale.
(ii) Events are described linguistically as delimited
through the scale provided by the direct argument, and they
are delimited within the verb phrase; i.e., an event may be
linguistically delimited by an internal direct or indirect
argument, but not by an external argument.
(iii) There may be at most one 'delimiting' associated with
a verb phrase.
(iv) Secondary arguments are delimiting expressions.
Now we examine how far these properties go in explaining why
certain arguments of verbs with particular cognitive
structures appear in the syntax as internal, external or
oblique arguments. In the next section it will be
demonstrated that principles (i) and (ii) above explain the
mapping of verbs into the unaccusative and unergative verb
classes.
6.2 Unaccusative and unergative verbs
6.2.1 Introduction
Intransitive verbs may be divided into two classes: the
- 250 -
unergatives and the unaccusatives.7 The sole argument of an
unaccusative verb is an internal argument, while the sole
argument of an unergative verb is an external argument.8
Unaccusative and unergative verbs provide a testing ground for
theories of internal and external arguments, because they are,
in a sense, minimal pairs. Comparison of these two verb
classes, therefore, is a promising place to begin
investigating the nature of internal and external arguments.
The distinction is a syntactic one; internal and external
arguments have different syntactic representations. A variety
of syntactic distinctions, showing that unaccusative and
unergative verbs have different syntactic behaviors, have been
mustered in a number of languages. The two verb classes also
show a striking semantic coherence cross-linguistically, but
in general the diagnostics in the literature for membership in
these classes, have been syntactic diagnostics.
First I will review the syntactic arguments for the internal
or external argument-hood of the arguments associated with
unaccusative and unergative verbs. Next I will show that none
7. Unaccusative verbs have also been called ergative verbs. I
will avoid a confusing proliferation of terminology by
referring to them in this thesis simply as unaccusatives. I
use the term 'intransitive' to mean a verb with only one
argument.
8. Burzio (1986), who introduced the unaccusative/unergative
distinction into Government and Binding theory, defines
unaccusatives as verbs that take a 'direct object' but assign
no thematic role to the subject.
- 251 --
of the semantic factors that seem to correlate with internal
or external argument-hood have perfect predictive power.
Finally I will show that the arguments of unaccusatives and
unergatives may be distinguished aspectually in that the
arguments of unaccusative verbs measure out the event in a way
that the arguments of unergatives do not. The arguments of
unaccusatives are therefore undifferentiated with respect to
the event described by the verb, and the arguments of
unergatives are not. (What is meant by "undifferentiated'
will become clearer in the following section.)
The aspectual theory of internal and external arguments
predicts three cross-linguistic classes, with respect to
meaning' or cognitive structure, of intransitive verbs: those
verbs that are always unergative, those that are always
unaccusative, and those that do not map consistertly to one or
the other class. These classes depend on whether the
cognitive structure of the verb is compatible or incompatible
with the aspectual properties of internal arguments. This
prediction will be borne out. The existence of this third
class shows that cognitive and syntactic structure must be
distinct levels of representation, because languages do make
some arbitrary choices in the mapping of cognitive structure
into syntax.
6.2.2 Review of the syntactic approach to unaccusatives and
unergatives
- 252 -
The class of unaccusative verbs was first identified by
Perlmutter (1978), who described their properties within the
framework of Relational Grammar. The idea of unaccusativity
was later introduced into the Government and Binding framework
in Burzio (1986). In both frameworks, the syntactic
differences between unaccusatives and unergatives follow from
the proposal that the argument of an unergative verb is
underlyingly a subject; whereas the argument of an
unaccusative verb is underlyingly a direct object, even though
it appears on the surface as a subject. In both theories,
subject- and object-hood are syntactically (rather than
semantically) defined, and they are used as notions with
cross-linguistic applicability.9 Both the Relational Grammar
and Government and Binding frameworks rely on multiple levels
of representation. It is these features that the two
frameworks employ in representing the difference between
sentences with unaccusative and unergative verbs. Both
frameworks derive the difference through universal rules of
syntax relating separate levels of representation.
Unaccusatives and unergatives differ as to whether what is the
subject at the surface level of representation is also the
subject at a deeper (or an initial) level of representation.
Internal and external arguments are distinguished in this
way. Although most of the discussion that follows is couched
9. For an opposing view see Van Valin (1987)
- 253 -
within the framework of Government and Binding theory, for the
purposes of the discussion here, either approach will
suffice. What is necessary is the idea that at the initial or
deep structure level of representation, unaccusatives have an
object argument and unergatives have a subject argument.
Since at deep structure the internal argument of a verb maps
to the object position and the external argument to the
subject position, unaccusatives have an internal argument, and
unergatives have an external argument.
The languages in which unergatives and unaccusatives are
easy to distinguish syntactically include Italian, English,
German, Albanian, Choctaw, Dutch, Lakhota, Turkish, Japanese,
Hebrew, and Basque.10  Much of the syntactic evidence for the
existence of the two classes is based on showing that the
arguments of unaccusatives (and the objects of transitive
verbs) pattern with the moved arguments of passives, and the
arguments of unergatives pattern with the unmoved subjects of
transitives. Evidence from Italian and English is discussed
in the following paragraphs.
Burzio (1986) provides several syntactic arguments for a
Deep Structure representation of unaccusative verbs, in which
10. See Burzio (1986) for Italian; Perlmutter (1978) for
English and German; Miyagawa (1987) for Japanese. Dubinsky
and Rosen (1987) is a guide to further sources in the
Relational Grammar framework, on a variety of languages.
- 254 -
the surface subject is the underlying object of the verb. He
distinguishes unergatives and unaccusatives on the basis of
partitive clitics. Italian has a partitive clitic, ne',
which may only be understood to apply to direct objects.
Burzio refers to this process as Tne cliticization', whereby
the clitic 'ne' occurs in a position preceding the verb, but
is associated with a direct object position following the
verb. It may not be associated with any other position:
(Examples 3 through 8 are from Burzio (1986))
3
a. Giovanni ne invitera molti.
Giovanni of-them will invite many
"Giovanni will invite many of them."
b. *Giovanni ne parlera a due.
Giovanni of-them will talk to two
"Giovanni will talk to two of them."
c. *Molti ne arriveranno.
many of-them arrive
"Many of them will a:-rive."
d. *Molti ne telefoneranno.
many of-them telephone
"Many of them will telephone."
(direct object)
(indirect object)
(subject)
(subject)
3 shows that the process of Tne' cliticization can be used as
a diagnostic for direct objects. Italian also has subject
inversion, in which the subject appears after the verb.
'Ne-cliticization' may occur in this construction with only
three types of clauses. It may occur with passives:
4
a. Saranno invitati molti esperti.
will be invited many experts
- 255 -
"Many experts will be invited."
b. Ne saranno invitati molti.
of-them will be invited many
"Many of them will be invited."
'Ne-cliticization' may also occur in the "inverted-subject"
form of the impersonal 'si' construction. This construction
is illustrated in 5a below, where the plural agreement on the
verb indicates that the subject is the noun phrase 'alcuni
articoli' following the verb. 5b shows that the 'ne' may be
construed with these "inverted subjects".
5
a. Si leggeranno volentieri alcuni articoli.
one will read(pl.) willingly a few articles
"A few articles will be read eagerly."
b. Si ne leggeranno alcuni.
one of-them will read(pl.) a few
"A few of them will be read."
The third environment where 'ne-cliticizatic'n' may occur is in
clauses with a particular class of verbs. These are the
unaccusatives:
6
a. affondarono due navi
sank two ships
"Two ships sank."
b. Ne affondarono due.
of-them sank two
"Two of them sank."
- 256 -
The subjects in the passive (4) and the impersonal 'si'
constructions (5) are generally assumed to be objects at Deep
Structure. I will not reproduce the arguments for this view
here. The crucial point to be made is that clauses with
unaccusatives are united with other constructions in which
surface subjects are deep structure objects. Burzio proposes
that the post-verbal subjects in 4 through 6 have not been
moved from their deep structure position, and this is the
cause of the syntactic difference that 'ne-cliticization' is
sensitive to. 'Ne-cliticization' is impossible with
unergatives (7) and transitives (8):
7
a. Telefoneranno molti esperti
will telephone many experts
"Many experts will telephone."
b. *Ne telefoneranno molti.
of-them will telephone many
"Many cZ them will telephone."
8
a. Esamineranno il caso molti esperti
will examine the case many. experts
"Many experts will examine the case."
b. *Ne esamineranno il caso molti
of-them will examine the case many
"Many of them will examine the case."
The analysis of postverbal unergative subjects as having
undergone movement, and unaccusative subjects as unmoved,
provides a syntactic, structural distinction between the two
clause types. Whether 'ne' is base generated in its surface
position or moved-there by derivation from Deep Structure,
- 257 -
(and Burzio writes that the arguments for either view are not
conclusive) the process of 'ne-cliticization' responds to that
syntactic distinction.
Perlmutter and Postal (1984) show that English
pseudopassives distinguish between unaccusatives and
unergatives. Pseudopassives are formed by moving what seems
to be an oblique argument of the verb into subject position.
The unergatives below form grammatical pseudopassives.
(Examples in 9,and 10 are from Perlmutter and Postal (1984).)
9
a. The bed was slept in by the Shah.
(derived from "The Shah slept in this bed."
b. The package was stepped on by a camel.
(derived from "A camel stepped on this package."
The unaccusatives in 10 do not form pseudopassives:
10
a. *The package was accumulated on by dust.
(derived from "Dust accumulated on this package.")
b. *The oven was melted in by the ice cube.
(derived from "The ice cube melted in the oven.")
A law of Relational Grammar (The 1-advancement Exclusiveness
Law)12  prohibits two advancements- (movements, in Government
and Binding terminology) in one clause, from a non-subject
12. See Perlmutter and Postal (1984) for details. Also see
Marants (1984) and Baker, Johnson, and Roberts (1986) for an
alternative account.
- 258 -
position to a subject position. The unaccusative
pseudopassives in 10 above are ungrammatical precisely because
they violate this law. Since the surface subject of an
unaccusative has already advanced from object position, a
second advancement of an oblique object is impossible.
Unaccusatives and pseudopassives are united under this
analysis in having surface subjects that are moved there from
some other position.
Italian and English are two languages that provide evidence
for the unification of arguments moved by syntactic
passivization with arguments of unaccusative verbs. I will
adopt the syntactic theory of-the distinction between
unergative and unaccusative verbs, on the basis of the kinds
of evidence cited above for Italian and English. In the next
section the semantic differences between unergatives and
unaccusatives will be considered.
6.2.3 Semantic differences between unergatives and
unaccusatives
When we consider the kinds of intransitive verbs that fall
into the unaccusative and unergative classes, the semantic
coherence within the two classes, and the semantic differences
between them, are impressive. However, no semantic diagnostic
has been proposed that is not merely a strong generalization.
In thts section I will illustrate the fact that each semantic
generalization about unaccusative and unergative verbs fails
- 259 -
as an absolute predictor of membership in one or the other of
these two verb classes (although some are more accurate than
others). A list, compiled by Perlmutter (1984), of the verbs
that generally belong in each class is repeated below in its
entirety. The verbs are English verbs, but Perlmutter
maintains that verbs of similar meaning make up the two
classes cross-linguistically.
Predicates Determining Initally Unergative Clauses'3
11
a. Predicates describing willed or volitional acts:
work, play, speak, talk, smile, grin, frown, grimace,
think, meditate, cogitate, daydream, skate, ski, swim,
hunt, bicycle, walk, skip (voluntarily), jog, quarrel,
fight, wrestle, box, agree, disagree, knock, bang, hammer,
pray, weep, cry, kneel, bow, curtsey, genuflect, cheat,
lie (tell a falsehood), study, whistle (voluntary), laugh,
dance, crawl, walk, etc.
Manner-of-speaking verbs:
whisper, shout, mumble, grumble, growl, bellow, etc.
Predicates describing sounds made by animals:
bark, neigh, whinny, quack, roar (voluntary), chirp, oink,
meow, etc.
b. Certain involuntary bodily processes:
cough, sneeze, hiccough, belch, burp, vomit, defecate,
urinate, sleep, cry, breathe, etc.
12
13. The term 'initially' in this heading refers to the idea
that the distinction between the two classes is stated at the
initial stratum (deep structure), and is lost when the
unaccusative argument changes from an object to a subject.
Also, in the Relational Grammar framework it is possible to
refer to unaccusativity and unergativity at various levels of
representation.
- 260 -
Predicates Determining Initally Unacgusative Clauses 14
a. Predicates expressed by adjectives in English:
a very large class, including predicates describing
sizes, shapes, weights, colors, smells, states of mind,
etc.
b. Predicates whose initial nuclear term15 is semantically a Patient
burn, fall, drop, sink, float, slide, slip, glide, soar,
flow, ooze, seep, trickle, drip, gush, hang, dangle, sway,
wave, tremble, shake, languish, flourish, thrive,
drown, stumble, trip, roll, succumb, dry, blow away,
boil, seethe, lie (involuntary), sit (involuntary),
bend (involuntary), etc.
inchoatives:
melt, freeze, evaporate, vaporize, solidify,
crystallize, dim, brighten, redden, darken, yellow, rot,
decompose, germinate, sprout, bud, wilt, wither, increase,
decrease, reduce, grow, collapse, dissolve, disentegrate,
die, perish, choke, suffocate, blush, open, close, break,
shatter, crumble, crack, split, burst, explode, burn up,
burn down, dry up, dry-out, scatter, disperse, fill,
vanish, disappear, etc.
c. Predicates of existing or happening:
exist, happen, transpire, occur, take place, and various
inchoatives such as arise, ensue, result, show up, end up,
turn up, pop up, vanish, disappear, etc.
d. Involuntary emission of stimuli that impinge on the
senses (light, noise, smell, etc.):
shine, sparkle, glitter, glisten, glow, jingle, clink,
clang, snap (involuntary), crackle, pop, smell, stink,
etc.
e. Aspectual predicates:
begin, start, stop, cease, continue, end, etc.
14. It should be noted that there is not universal agreement
on this classification. According to B. Levin (p.c.) and
Zaenen (1986), verbs of involuntary emission of stimulus such
as those in d (and also 'blush', which is is similar to these
but is listed under b (B. Levin, p.c.)) are not
unaccusatives.
15. 'Initial nuclear term' here refers to the argument of the
intransitive form of the verb.
- 261 -
f. Duratives:
last, remain, stay, survive, etc.
Many semantic generalizations are statable over the two
classes. These have been widely noted in the literature, and
certain semantic properties have been ascribed to one or the
other of the classes. Among these are:
(i) agency. By agency is meant physical causation. That
is, the argument does something or moves in some way,
voluntarily or involuntarily, to cause the action to take
place. This seems to be a property of unergative verbs. Some
unspecified 'sub-actions' must take place for the action
described by the verb to occur. 'Skating', 'walking' or
'bicycling' involves moving the legs in a particular way.
'Wrestling' entails doing particular things with the arms.
'Whispering' or 'barking' requires activating the mouth and
vocal tract. In order to 'hammer' one must hold the hammer,
as well as wield it. Various parts of the anatomy do various
things when someone 'sneezes' or 'belches'. Even 'meditating'
or cogitating' require a kind of initial action, in a
metaphorica sense, to set the mina in place ana aunch it on
a particuiar aina o menlaL activity.
unaccusazives generaixy ao not have this property. No other
action is necessary on the part of the argument besides that
specifically denoted by the verb. If something 'evaporates',
or 'melts', or 'reddens', or 'scatters', or 'diappears', or
- 262 -
'remains'; it does not have to do anything else but that. An
event of 'reddening' or 'disappearing' can take place without
any other noticeable events or changes transpiring.
Agency in this sense, then, is an attractive possibility for
a semantic distinction between the two classes. However, its
coverage is not perfect. The unergative verbs 'smile' and
'grimace' can be argued not to entail other actions on the
part of the smiler or grimacer (e.g., such as putting one's
hands to one's face to rearrange the facial muscles). The
unaccusative verb 'stumble' does seem to involve independent
motions of the legs or feet. The point to be made here is
that agency in the sense of physical causation does not yield
an absolute distinction between the two classes.
(it) volition or protagonist control. These properties
refer to agency in the sense that the argument wills the
action denoted by the verb. All of the unergative verbs in
Ila have this property, and none of the unaccusative verbs in
12a, b, c, d, and f have it (although it is true that some of
the unaccusatives have unergative counterparts involving
volition). However, volition will not distinguish absolutely
between the two classes. The unergative verbs in 11b such as
sneeze' or 'hiccough' are not volitional, and some of the
events denoted by the unaccusatives in 12e (e.g. 'start' and
'stop') can be willed. Neither physical causation nor
volition can be exclusively associated with one or the other
-263 -
class.
(iii) animacy. This is a property commonly associated with
agency, and always associated with volition (depending on
one's metaphysical view of the world). Most if not all of the
unergatives in 11 can only have animate arguments, and most of
the unaccusatives in 12 take only -inanimate arguments. But
again, this is only an imperfect generalization; there is an
unaccusative verb 'die' that requires an animate argument.
(iv) inherent lexical aspect: duration. All the unergatives
in 11 express situations with some duration over time
(accomplishments or activities in Vendler's terminology). A
large subgroup of the unaccusatives in 12, the inchoatives,
express situations that happen almost instantaneously. But
the unaccusatives also include many verbs like *burn', 'drip'
and 'shake', which express events that happen over some
interval of time. Duration is not an accurate diagnostic for
unaccusativity and unergativity.
(v) inherent lexical aspect: delimitedness. Unergatives
tend to describe non-delimited events, while unaccusatives
tend to decribe delimited events. This is true of the
unergatives 'work', 'smile', 'walk', 'mumble', and 'sleep';
and the unaccusatives 'melt', 'collapse', and 'vanish'. Many
intransitive verbs that take affected arguments (arguments
capable of delimiting events) belong in the class of
unaccusatives. 12b lists a number of them. But delimitedness
- 264 -
is no better a diagnostic than agency, volition, animacy or
duration, because there are exceptions to this
generalization. The unergatives 'cough', 'sneeze' and 'bang'
make delimited clauses if they are understood in the sense of
coughing, sneezing, or banging once. The unaccusative list
(12) includes many stative verbs, which always describe
non-delimited events.16
Cross-linguistically these semantic properties are
inconsistent diagnostics of unergativity and unaccusativity.
According to Van Valin (i987b), in Italian and Georgian the
split between unergatives and unaccusatives depends on lexical
aspect, while in Acehnese and~Tsova-tush it depends on
agentivity.1 7 Van. Valin argues effectively that such lexical
16. Within a stricter view of membership in the unaccusative
and unergative verb classes, it may in fact be true that
unaccusatives always describe delimited events: If Zaenen
(198E) and Levin (p.c.) are correct these statives verbs in
12 do not belong in the class of unaccusatives. Then if we
remove the class of purely aspectual verbs (12e and 12f; and
perhaps 'exist' in 12d as well) the remaining verbs in list 12
describe events which are delimited by the arguments of the
verbe. If this is true, the class of unaccusative verbs are
distinguished by an aspectual property like middles. I leave
this as an open question and assume for bhe moment the more
general view of the membership in the class of unaccusatives.
17. Italian verbs of motion are a particularly interesting
example. They may shift class, depending on whether the event
is delimited or not. Auxiliary selection in Italian makes
reference to the distinction between unerga.tive and
unaccusative verbs. Unergative verbs take 'avere' ("have")
and unaccusative verbs take 'essere' ("be"). (The data below
is from Burzio (1986).)
- 265 -
semantic properties correlate with the unaccusative/unergative
distinction. However, he proposes no consistent mapping from
these properties to the two verb classes. Rosen (1984) writes
that in Choctaw, protagonist control is generally irrelevant
to the distinction between unergativity and unaccusativity,
but there are a handful of verbs in the language for which it
is relevant.
It appears that none of the semantic properties identified
a. Giovanni ha telephonato.
"Giovanni has ('avere') telephoned."
b. Giovanni e arrivato.
"Giovanni has ('essere') arrived."
c. L'artiglieria ha affondato due navi nemiche.
"The artillery has ('avere') sunk two enemy ships."
d. Due navi nemiche sono affondate.
"Two enemy ships have ('essere') sunk.
Some Italian verbs of motion vary between unergattWve or
unaccusative, depending on whether or not they denote a
delimited event: (The data below is from Van Valin (1987b).)
(W)
a. Luisa ha corso nel parco. (non-delimited)
Luisa has run in the park.
"Luisa ran in the park."
b. Luisa e corsa a casa. (delimited)
Luisa is run to home
"Luisa ran home."
(ii)
a. L'uccello ha volato solo per qualche minuto. (non-delimited)
the bird has flown only for some minutes
"The bird flew just for a few minutes."
b. L'uccello e volato via. (delimited)
The bird is flown away
"The bird flew away."
- 265 -
above are adequate as diagnostics for unaccusativity and
unergativity.18 Yet we are left with semantic generalizations
striking enough to require an explanation. The explanation
becomes apparent when we consider the special aspectual
properties of internal arguments.
6.2.4 The aspectual distinction between unaccusatives and
unergatives
The difference in the aspectual properties of internal and
external arguments explains the semantic coherence of the
unaccusative and unergative verb classes. Recall from Chapter
4 that an internal argument that undergoes change during an
event, undergoes change in such a way that it provides a scale
by which the event described by the verb may be measured out.
The external argument does not have to meet this condition. A
crucial property of a scale is that all units are the same,
and for all practical purposes indistinguishable from each
other. The change undergone by the internal argument is
consequently undifferentiated with respect to the. event.
Furthermore, all of the scale is potentially involved in the
event. Consider the unaccusatives in the following sentences:
13
a. The apple ripens. (delimited)
b. The leaves burn. (delimited)
c. The puddle evaporates. (delimited)
18. See Rosen (1984) for additional discussion of this point.
- 267 -
d. The lake sparkled. (non-delimited)
e. The populace thrived. non-delimited)
f. The city existed. non-delimited)
g. The ruined building will remain. non-delimited)
(The examples are in past, present and future forms to
illustrate that tense does not play a role in this
phenomenon. They include both delimited and non-delimited
sentences.) The first three examples, 13a, b and c, are most
easily (but not necessarily) understood as delimited
sentences. Consider first the delimited readings. In 13a,
all the apple is consumed, and the consuming of any one part
of it does not differ from the consuming of the next part.
Likewise with 13b and 13c. If the leaves and the puddle were
to be partitioned into subparts of leaves and puddle, all of
those subparts would be understood to burn or to evaporate in
the course of the event. Even in the non-delimited readings
of 13(a)-(c) the argument provides a cale for the event.
Even if all the leaves are not burned, the event of burning
progresses through the leaves over time, without
distinguishing any leaf or part of a leaf from any other. The
same is true of 13b and 13c. Now consider 13d, e, f, and g.
They do not have delimited readings. Nevertheless, just as
with 13a, b and c, the event involves all hypothetical
subparts of the arguments equally. In 13d and e, the lake and
the populace are an undifferentiated mass as far as sparkling
and thriving are concerned. If lake and populace are
partitioned into subsections, the total amount of sparkling
- 268 -
and thriving is the sum of the sparkling and thriving of all
the subsections. The verbs in 13f and g are statives, but
they follow the same pattern. If the city exists, then all
the city exists, and the city is not internally differentiated
with respect to existing.1 9  If the building remains, the wall
of the building 'remains' in the same way as the roof does.
Wall and roof do not 'remain' differently.
The verbs in 13 are unaccusative, and therefore their
arguments are underlyingly internal arguments. The internal
arguments measure out the event or situation described by the
verb. The external arguments of the unergative verbs in the
examples below do not have the aspectual property of measuring
out the event:
14
a. Bill grins.
b. The children wrestle.
c. Mary mumbled.
d. The passengers coughed.
e. The lion will roar at feeding time.
In 14 the arguments are not partitinnable in such a way that
they measure out the event on a scale. Various hypothetical
19. The aspectual difference between internal and external
arguments was stated in Chapter 4 for arguments which undergo
change during the course of the event. The idea of being
undifferentiated with respect to the event is a logical
extension of the idea of a scale, to a situation in which the
argument does not undergo change, as in 13(d)-(g). This is an
issue that requires further research.
- 269 -
subparts of the arguments may be involved in the action, but
the subparts are involved in individual ways. When Bill
grins, Mary mumbles, or the lion roars, only their mouths,
faces or lungs take action -- not the rest of them. If the
passengers cough, it is their throats that do the coughing.
And if children are wrestling, although they may use almost
all parts of their bodies, not all parts are used in the same
way. Wrestling entails various kinds of actions -- grabbing,
pulling, or pushing, for instance -- which means doing very
particular things with arms and legs. An event can only
measured out through its argument if the participation of the
argument is partitionable into similar actions; in other
words, if it is characterizable as change along a single
scale. This is the basis of the aspectual distinction between
internal and external arguments. The distinction is clear
with verbs that have both unaccusative and unergative forms.
When we hear a lion roar (unergative and volitional), we know
that the source of the roar is localized to -one part of the
lion; but when we hear a waterfall roar (unaccusative and
non-volitional), we can only think of the roar as coming from
all over the waterfall. The lion is internally differentiated
in the event. The waterfall is not. The aspectual
distinction between internal and external arguments, based on
the principles of argument strucxure discussed in Chapter 4,
provides a more effective means of characterizing the
differences between the unergative and unaccusative verb.
- 270 -
classes, than do the properties of agency, volition, animacy,
duration or delimitedness.
The aspectual properties of internal and external arguments
explain why the semantic properties discussed in section 6.2.3
characterize the two classes to the extent that they do.
There are certain semantic characteristics that are more
naturally associated with the aspectual properties of internal
arguments than others:
(i) Agency, volitionality and animacy are not generally
associated with event participants which participate in the
event in an undifferentiated way, or for which change is
expressible along a single scale. This has been discussed
above and in Chapter 4.
(ii) Many unergative verbs are inherently durative because
duration tends to be associated with agency. (This has been
noted by Dowty (1979), among others.) Agency and volition on
the part of the subject mean that the subject must do
something in order to bring about the event. This constitutes
a set of unspecified actions understood to be part of the
event, and these actions take time. If there is agency on the
part of the subject, and particularly if the event involves
physical motion, it is possible that the event described will
be one that has duration. A durational activity like 'swim'
involves many unspecified actions that require motion or
physical change -- rotating the arms, kicking the legs,
- 271 -
propelling oneself forward in the water. These actions take
time. Unaccusatives such as inchoatives do not involve such
unspecified actions since they describe change characterizable
on a scale. When the lake freezes we do not see it doing
anything else in order to freeze. When an apple reddens, that
is all it does -- rqdden. It does not need to take a brush
and paint itself red, and so it could conceivably redden
almost instantaneously. For this reason unaccusatives often
describe non-agentive and non-durative events.
(iii) Delimitedness is more often associrted with
unaccusatives than with unergatives because internal arguments
may measure out the event, while external arguments cannot do
so. If the argument measures out the event, it may also
delimit it, as do the affected arguments discussed in Chapter
3. Some unaccusative verbs have affected arguments; the events
they describe are delimitable through their internal
arguments.20
The preceding paragraphs (i)-(iii) have explained why
semantic and aspectual properties such as agency, volition,
duration and delimitedness figure so strongly in the
distinction between unergative and unaccusative verbs. The
20. Sentences with unergative verbs may describe delimited
events, but when they do, the events they describe are
delimited through a real or implied object, often a cognate or
reflexive object.
- 272 -
reasons why these semantic properties correlate so strongly
with unergativity and unaccusativity are also the reasons why
they fail where they do, as diagnostics of class membership.
Perlmutter's lists of unergative and unaccusative verbs (11
and 12) notwithstarIing, the same predicates do not always
turn up in the same classes cross-linguistically. Rosen
(1984) cItes numerous examples, among which are the following:
15
'die'
'sweat'
'bleed'
Isuffer'
'hungry'
'dream'
'Cparkle'
Unaccusative in:
Italian
Choctaw
Turkish
Choctaw
Choctaw
Eastern Pomo
Dutch
Unergative in:
Choctaw
Italian
Italian
Italian
Italian
Dutch, Italian, Albanian
Italian
It is significant that the predicate-menaings in 15 could be
imagined to describe situations in which the argument had
either aspectual role (internal or external). These
predicate-meanings constitute the third group mentioned in the
introduction to this chapter, which do not belong consistently
to either the unergative or the unaccusative class of verbs
cross-linguistically. Verbs whose %meaning' does not
absolutely commit them to one aspectual class or the other
will vary from language to language.
- 273 -
Merlan (1985) discusses the differences between'
'subjectively' and 'objectively' inflecting intransitive
verbs, which for practical purposes, are equivalent to
unergative and unaccusative verbs respectively.21 She
examines this question with respect to eight languages
(Dakota, Seneca, Arikara, Tunica, Eastern Pomo, Batsbi,
Georgian, Mangarayi), and concludes that semantic
characterizations of these two classes are possible, but are
not without exceptions. Merlan found that semantically
descriptive predicates (e.g., Tbe wise') tend to be
objectively inflecting, but exceptions are to be found in
Seneca and Dakota. Most objectively inflecting intransitives
seem to be associated with involuntariness, although the
voluntary/involuntary distinction is inadequate for perfectly
characterizing the subjective/objective inflectional
distinction. Merlan writes further that verbs having to do
with bodily processes and functions (e.g. 'sneeze') cannot be
characterized as associating strongly with either class.2 2
21. In Merlan (1985), -subjective' inflection refers to the
marking of pronominal arguments of intransitive verbs by the
same inflectional forms used for subjects of transitive
verbs. 'Objective' inflection refers to the marking of
pronominal arguments of intransitive verbs by the inflectional
forms used to mark objects of transitive verbs. In these
cases the inflection on pronominal arguments marks whether it
is a internal or an external argument.
22. Merlan observes that certain verbs of bodily processes
tend to be found in the smaller of the two verb classes,
whether they are subjectively or objectively inflecting.
Furthermore, this minority verb class tends to contain verbs
- 274l-
Subjectively inflecting classes are likely to contain verbs of
inherently directed motion (e.g., 'jump',, fly')., while
objectively inflecting classes generally contain verbs of
manner of motion (e.g., 'fall', 'slip'). There is a
correlation between objective inflection and verbs describing
mental processes or physical or mental conditions. There is
also a correlation between subjective inflection and voluntary
activity. However, in spite the existence of these pronounced
patterns, Merlan concludes that:
Any formulation of the relation between lexical meaning and
inflection form is evidently an implicational statement and
not an absolute categorization...(Merlan (1985), p. 349)
The unaccusative and unergative verb classes, determined by
syntactic criteria, do not always contain the same
predicate-meanings cross-linguistically. Therefore the
mapping from 'meaning' to aspectual and syntactic structure
cannot be governed by inviolable principles. Whether an
intransitive verb takes an internal or an external argument
must be marked in the lexicon, but how it is marked is
influenced by the necessity of fitting the 'meaning' into an
aspectual structure. The marking of a verb's argument as
internal or external may be influenced by 'meaning', world
knowledge or even cultural beliefs. Dying or dreaming, for
example, may be viewed differently in different cultures, and
requiring animate subjects. If this is a general tendency
across languages I have no explanation for it.
- 275 -
the way these events are viewed may contribute to whether they
are perceived to fit into the aspectual mold of an internal or
external argument. Such extra-linguistic effects as these may
influence whether a predicate-meaning is instantiated as an
unaccusative or an unergative verb in a particular language.
But once a verb is marked in the lexicon as taking an external
or an internal arg'Ament, its interpretation conforms to the
aspectual constraints on verbs with internal or external
arguments.
Some types of verbs of motion are particularly variable. On
the one hand, the moving argument does not seem to change
during the event, and may engage in unspecified sub-actions to
bring about the motion, so it does not 'measure out' the
event. On the other hand, the property of location, which is
a property of the argument, certainly 'measures out' the
event. This explains why verbs of motion may go either way.
Under the view advanced here, there is no direct mapping
from thematic or cognitive structure to syntactic structure;
rather, the mapping is from cognitive structure to aspectual
structure. The aspectual structure is inherent in the
syntax. This is the reason why, while semantic properties
seem to map into syntax, the precise mapping has always eluded
discovery. Aspectual structures impose general constraints on
the types of events in the world that may be mapped into them,
in a way that purely syntactic structures cannot. In other
- 276 -
words, the aspectual structure provided by language -- the
distinction between internal and external arguments, for
instance -- is the mold into which the linguistic expression
of events in the world must be fitted.
Furthermore, thematic roles such as 'agent' or 'theme' are
not visible to the principles of syntax, but only to the
aspectual structure inherent in the syntax. The view I am
advocating supports a strong version of the autonomy of syntax
and semantics; and furthermore, provides an explicit model for
the interface between syntax and lexical semantics.
6.3 The locative alternation
Recall the class of spray/load verbs discussed in section
2.5.3. This is a class of verbs that exhibits the locative
alternation (so named by Rappaport and Levin (1984))
illustrated in 16 below:
16
a. spray paint on the wall
spray the wall with paint
b. load hay on the wagon
load the wagon with hay
c. clear dishes from the table
clear the table of dishes
d. cram pencils into the jar
crat the jar with pencils
I will refer to these verbs as spray/load verbs. Verbs in
this class have two arguments, one representing a material,
-277 -
and the other an area, flat surface or container onto (or
into) which the material is applied (or from which it is
removed) in the event described by the verb. The interesting
thing about these Verbs is that either the argument
representing some material ('paint' in 16a), or the argument
representing the surface on which the material is found ('the
wall' in 16a), may appear as the direct argument of the verb.
Verbs with similar 'mearings' exhibit this same alternation in
many languages. Dutch and Japanese examples from Chapter 2
are repeated below:
17
Dutch: (De Groot(1984), from Dk(1980))
a. Jan plant bomen in de tuin.
John plants trees in the garden
b. Jan beplant de tuin met bomen.
John be-plants the garden with trees.
Japanese: (Fukui, Miyagawa, Tenny (1985))
a. kabe ni penki o nuru
wall on paint ACC paint(VERB)
smear paint on the wall
b. kabe o penki de nuru
wall ACC paint with paint(VERB)
smear the wall with paint
to
a. taru kara sake o akeru
barrel from sake ACC empty
empty sake from the barrel
b. taru o akeru
barrel ACC empty
empty the barrel
A similar alternation in which goals may be direct arguments
is not generally possible:
- 278 -
18 -
push a cart to San Francisco
*push San Francisco with a cart
walk your dog to the corner
*walk the cornet with your dog
The locative alternation seems to be associated with the verb
having a goal argument that is a flat surface or area or a
container. Wo must ask the question why should the syntactic
arrangement of the verb's arguments depend on such a seemingly
trivial semantic property? The answer may be found in the
aspectual properties of the verb's direct argument. The
direct argument (if it unde.Aoes change) must provide a scale
for the event; it must measure out the event over time. A
material is easily construed as measuring out the event, since
it is consumed a little at a time during the course of the
event, and when it is used up the event is finished. However,
a goal is not so easily construed in this way, unless the gopJ.
is a flat surface or area or container. Consider an event in
which a material is applied to a flat surface. A flat surface
makes a natural scale for such an event because th3 material,
having no form of Its own, takes on the form of the surface to
whi^h it is applied, ad spreads out evenly across the
8urface. As it sp-ea'Is out it covers more and more surface
area, and so the surface area may be thought of as measuring
out the event. Liketise with a container. As the material
e'ters the container, the container becomes more and more
ftll. The fullness of the contairer measurco out the eveat.
- 279 -
This is what the cognitive structure associated with the
lexical entry of the verb must capture. This is the property
that allows the locative alternation to take place.
Spray/load verbs allow the goal to be their direct argument in
the syntax exactly because spray/load verbs describe events in
which the goal may be construed as measuring out the event.
We need only a cognitive structure expressing this fact about
spray/load verbs, and the aspectual constraints on the
interpretation of the direct argument will predict that the
alternation to be possible.23 Now consider the sentences:
23. More precisely, the aspectual constraints on the direct
argument predict that the verb may be marked as allowing the
alternation. It must be noted that the fact that a verb
describes an event involving a material and a flat surface or
container is a necessary but not sufficient condition for the
locative alternation to be possible. The verb's arguments --
the event participants -- must be marked as to whether they
are possible direct arguments. Besides satisfying the
aspectual constraints, being a direct argument requires
selection by the verb. For example:
a. fill the glass with water
b. *fill water in the glass
a) is possible and b) is not, although the event desribed is
of the kind that should permit the locative alternation.
There is semantic selection of the container argument by the
verb 'fill', since 'fill' requires as its direct object
something that can be filled:
fill the glass
?fill the wall (?with bookshelves)
*fill the pole
There is no semantic selection of the material argument by
'fill', because 'fill' holds no information about what the
container may be filled with.
The converse is illustrated by 'pour':
- 280 -
19
a. spray the paint iA the bucket
b. spray the bucket with paint
The preposition 'in' in 19a explicitly relates the bucket to
the event described in a particular way; the paint is directed
into a constrained area of the bucket, rather than being
spread out over some indefinite extent of surface area.
Although 19b is a grammatical sentence, it is not a paraphrase
of 19a. When the bucket is the direct argument of the verb,
the "in' meaning is lost and only the interpretation in which
the bucket measures out the event of spraying is available.
*pour the glass with water
pour water into the glass
'Pour' constrains the material to be something 'pourable':
pour water
?pour trees (?down the mountainside)
*pour pastures
but "pour' says nothing about what the material is to be
poured into or onto.
Direct argument-hood is associated with the ability of the
verb to semantically select the argument, but the aspectual
properties associated with direct argument-hood are more
important. Verbs which fulfill the aspectual conditions for
the locative alternation can often be reanalyzed as
participants in that alternation, even when they do not
semantically select for both material and goal arguments. All
the following are good in the author's judgements:
pour concrete on the roads
pour the roads with concrete
fill the coffee up to the rim of the cup
fill the cup up to the rim with coffee
- 281 -
19b can only mean that the paint was spread around some
surface area of the bucket.
Consider the verb phrases below:
20
a. spray the wall with water
b. spray the wall with a hose
In both (a) and (b) above, 'wall' is the direct argument. The
indirect argument is a material, 'water', in 20a and an
instrument, 'hose', in 20b. 20a can participate in the
locative alternation while 20b cannot:
21
a. spray water on the wall
b. *spray a hose on the wall
Water, a material, is consumed during the spraying. The water
sprayed during each subinterval of time during which the
spraying takes place is understood to be different from the
water sprayed in the next subinterval of time. The hose, an
instrument, is not consumed in thq spraying. As this event is
described, all the hose participates in each subinterval of
the event; rather than different pieces of the hose
participating in different subintervals of the event. Because
a material is consumed during an event, it can 'measure out'
the event over time, and can participate in the locative
alternation, appearing as a direct argument Since an
instrument is not consumed over time, it cannot measure out
the event, and cannot participate in the locative alternation
- 282 -
in 21b.
Path arguments have much in common with the goal arguments
of spray/load verbs. Paths may be direct arguments for the
reason that they measure out the event. Consider these
sentences:
22
a. walk the bridge
b. walk on/across the bridge
c. walk to the bridge
'Bridge' is the direct argument of "walk' in 22a but not in
22b and 22c. When it is the direct argument of the verb, it
can only be understood as a path, because in this position it
is forced to measure out the event of walking. For this
reason only 22b can be a paraphrase of 22a. 22c does not
express an event in which the bridge itself meabdres out the
event of walking, and so it cannot paraphrase 22a.
The behavior of spray/load verbs and path objects provides
further evidence that the direct argument of the verb is
constrained to provide a scale for the event described by the
verb when it undergoes change.
- 283 -
6.4 Body part/possessor alternation
Inalienable possession is a semantic relation with
interesting grammatical consequences in many languages. It is
exemplified by the relationship between body parts and their
possessors. English verbs of contact permit an alternation
such as the following when they have, as arguments, body parts
and their possessors:
23
a. hit Bill's arm
b. hit Bill on the arm
a. beat a man's shoulders
b. beat a man on the shoulders
a. kiss her sister's cheek
b. kiss her sister on the cheek
a. tap Joe's shoulder
b. tap Joe on the shoulder
a. poke the cat's nose
b. poke the cat on/in the nose
The verbs above are verbs of contact which have unaffected
direct arguments. In the a) sentences the body part is the
direct argument. In the b) sentences it is not the direct
argument; rather, the direct argument is the possessor of the
body part, and the body part is a further specification of the
direct argument. The prepositions 'on' or 'in' express the
idea of contact. This alternation is not possible with
affectedness verbs, even when they involve contact:
- 284 -
24
a. break John's foot
b. *break John on the foot
a. wash Mary's face
b. *wash Mary on the face
a. redden the clown's nose
b. *redden the clown on the nose
a. paint Bess' toenails
b. *pairt Bess on the toenails
Since an affectedness verb requires its direct argument to
'measure out' and delimit the event, that direct argument may
not be further specified as in the b) examples above. The
scale through which the event is measured out is determined by
the verb and the direct argument. Consequently the direct
argument is not modifiable by-adverbials that would alter the
scale itself. Some point on that scale may be marked by
delimiting adverbials:
25
paint Bess' toenails halfway/up to the edges/all over.
but this kind of aemantic contribution is quite different from
a further specification of the direct argument. The
requirement that the direct argument provide a scale for the
event blocks such further specification.
The view that this requirement is due to syntactic argument
structure rather than to the semantics of individual verbs is
supported by the fact that further specification of the direct
argument of affectedness verbs is impossible even when it is
implicit in the verb:
- 285 -
26
*decapitate the chicken on the head
*peel the apple on the skin
Even though it is the head and the skin that is removed in the
events as described above, the syntactic direct arguments --
'the chicken' and 'the apple' -- are what measure out and
delimit the events, as those events are described in these
verb phrases. The constraint is also not due to the fact that
the object referred to in the further specification may not
itself 'measure out' the event. 'Skin' may also be the direct
argument of 'peel':
27
peel the skin (off) the apple
The aspectual constraint on direct arguments provides an
explanation why the body part/possessor alternation is not
possible for affectedness verbs.
6.5 Psych verbs
Certain verbs expressing psychological states or events
('psych verbs') take as one of their arguments what is
generally referred to as an experiencer. In the two sentences
below, 'John' is an experiencer:
28
John fears ghosts.
Ghosts frighten John.
Because experiencers can appear in either internal or external
- 286 -
argument position, it would seem that they are a type of event
participant that does not map consistently into syntactic
argument structure. If this is true, psych verbe pose a
problem for the idea that there is a principled mapping
between cognitive structure and syntactic argument structure.
But this is only true as long as thematic roles, represented
by terms such as 'experiencer', are what the mapping
principles refer to. However, if we consider that it is not
thematic roles but zspectual principles, that govern the
interface between cognitive structure and syntactic structure,
then psych verbs are not aberrant.
Belletti and Rizzi (1986), in a paper on psych verbs in
Italian, show that there are two classes of Italian psych
verbs. One group has the experiencer as an external argument
and the other has the experiencer as an internal argument. 1
will review two of the pieces of syntactic evidence presented
by Belletti and Rizzi in support of this.
Deep subjects (non-derived subjects) in Italian can bind
anaphoric clitics, as in 29a below.. Derived subjects cannot,
as in the passive example in 29b and the raising example in
29c. (Examples 29 through 34 are from Belletti and Rizzi
(1986).)
1. Belletti and Rizzi actually discuss three classes of psych
verbs, but in two of these classes the experiencer is an
internal argument.
- 287 -
29
a. Gianni si e fotografato
"Gianni himself photographed."
b. *Gianni si e stato affidato.
"Gianni to-himself was entrusted."
c. *Gianni si sembra simpatico.
"Gianni to-himself seems nice."
Verbs like 'temere' ("fear", in 30) and 'preoccupare'
("worry", in 31) pattern like deep-subject verbs and
derived-subject verbs respectively:
30
Gianni si teme
"Gianni himself fears."
Io mi conosco
"I myself know."
31
*Gianni si preoccupa
"Gianni himself worries."
*Io mi interesso
"I myself interest."
Secondly, the Italian causative construction is incompatible
with embedded derived subjects:
32
a. Gianni ha fatto telefonare (a) Mario
"Gianni made Mario call."
b. *Gianni ha fatto essere licenziato (a) Mario
"Gianni made Mario to be fired."
The embedded phrase in 32b is passive and therefore has a
derived subject. 32b is consequently ungrammatical. 32a,
- 288 -
which has a non-derived subject in the embedded phrase, is
grammatical. The 'temere' class of verbs may be embedded
under a causative while the 'preoccupare' class of verbs may
not:
33
Questo lo ha fatto temere/apprezzare/ammirare ancora
di piu a Mario.
"This made Mario fear/estimate/admire him even more."
34
*Questo lo ha fatto preoccupare/commuovere/attrarre ancora
di piu a Mario.
"This made Mario worry/move/attract him even more."
Psych verbs in Italian are thus shown to fall into distinct
classes with distinct syntactic behaviors, due to the
difference in their syntactic argument structures. Verbs with
external-argument experiencers behave like deep-subject verbs,
while verbs with internal-argament experiencers behave like
derived-subject verbs.
The question that must be considered is whether
internal-argument experiencers and external-argument
experiencers actually have the same semantic properties. If
they do they pose a problem for a view such as that formulated
in Baker's UTAH and the UAH. I will examine the English
equivalents of the following Italian verbs cited by Belletti
and Rizzi. The 'temere' class of. verbs includes:
- 289 -
35
temere
conoscere
accettare
ammirare
apprezzare
rispettare
The 'preoccupare'
"fear"
"know"
"accept"
"admire"
"like"
"respect"
class of verbs includes:
36
preoccupare
interessare
attirare
commuovere
entusiasmare
affascinare
appassionare
spaventare
disgustare
"worry"
"interest"
"attract"
"move"
"excite"
"fascinate"
"excite"
"frighten"
"disgust"
The English glosses of the Italian verbs in 35 and 36 are
themselves psych verbs that belong to the respective classes.
The experiencer argument of the 'fear' class (under 37) is the
deep-structure subject, while of the 'worry' class (under 38)
it is the deep-structure object: 2
37
John fears the truth.
John knows the truth.
John accepts the truth.
John admires the truth.
John likes the truth.
John respects the truth.
38
The truth worrieb John.
2. Independent principles of grammar prohibit any other
analysis.
- 290 -
The truth interests John.
The truth attracts John.
The truth moves John.
The truth excites John.
The truth fascinates John.
The truth frightens John.
The truth disgusts John.
Since the salient readings for the examples in 37 and 38 are
non-delimited readings it is difficult to use delimitedness as
a test. However, some of the sentences above can in some
situations be construed as describing delimited events. These
non-salient readings are emphasized in the following
examples. (The psych verbs and their internal arguments are
underlined in these examples.)
39
The prophet went into the desert to fast and pray, and in three dayS
he knew the truth.
It took Sebastian a year to accept the truth about his
origins.
40
The play moved the audience in the first ten minutes of its
performance.
The livestock were usually slow to arouse to fear, but the
truck's backfiring frightened the cattle in an instant.
When these psych verbs do yield a delimited reading, it is the
internal argument that delimits the event:
41
After two days the prophet knew the truth halfway; he
knew half the truth/??half of him knew the truth.
- 291 -
?After the first two minutes of the play the audience knew the
outcome halfway; the audience knew half the outcome/*half of
the audience knew the outcome.
Sebastian accepted the truth about his origins slowly,
accepting only a little bit of the truth at a time/*only a
little of him accepting the truth at a time.
42
The play moved the audience quickly, moving a lot of the
audience at once/*a lot of the play moving the audience at
once.
?The play moved the audience halfway; half the audience
were moved/*half the play moved the audience.3
The truck's backfiring frightened the cattle quickly;
frightening a lot of the cattle atonce/*a lot of the noise
frightening the cattle at once.4
?The truck's backfiring frightened the cattle halfway;
Half the cattle were frightened/*Half the noise frightened
the cattle.
The judgements for the examples in 41 and 42 are somewhat
muddy. The fact that the internal argument delimits the event
is a little clearer in 43 and 44 below. Additional delimiting
expressions in the verb phrase refer to the scale provided by
the internal argument:
43
The children feared the movie to the end.
*The movie frightened the children to the end.
44
*The children feared the movie to death.
3. 'Halfway' is not meant to be a depictive secondary
predicate on the subject here.
4. These sentences are awkward because the speaker or hearer
must not only imagine a delimited reading of 'frighten', but a
durational reading as well.
- 292 -
The movie frightened the children to death.5
There is a further aspectual difference between internal
argument experiencers and external argument experiencers.
When a psych verb with an external argument experiencer
describes a delimited event, that event has duration, because
it must be an event in which the action progresses through the
internal argument over time:
45
accept the truth gradually
When a psych verb with an internal argument experiencer
describes a delimited event, the event is generally an
instaneous one, because the experiencer is undergoing a change
of state:
46
?frighten children gradually
The difference between 45 and 46 results from the differing
ways an experiencer and a non-experiencer undergo change; and
therefore, the differing ways in which they are capable of
delimiting an event. These are non-linguistic facts with
linguistic consequences. The AIH is responsible for the
translation of one into the other.
Psych verbs with experiencer arguments conform to the
5. 'To death' is a delimiting expression in its literal
interpretation.
- 293 -
aspectual constraints on internal and external arguments.
They do not represent an unprincipled mapping of cognitive
structure into syntactic structure. Rather than saying that
an experiencer is sometimes mapped to external argument
position and sometimes to internal argument position, the
correct generalization is that an experiencer is a kind of
event participant that may be cast in linguistic terms as
measuring out or as not measuring out an event.
6.6 UTAH and thematic roles
The Universal Theta Assignment Hypothesis of Baker (1985)
and the Universal Alignment Hypothesis of Perlmutter and
Postal (1984) propose a strict correspondence between thematic
roles (or 'meaning') and structural positions (or grammatical
relations). These hypotheses as they stand are stipulative.
It would be desirable to put them on a more principled br.is.
The aspectual principles of argument structure advanced in
this thesis can do that. I will focus in the following
discussion on the UTAH, although the consequences of these
ideas apply to both hypotheses.
The characterizations of thematic roles such as agent,
patient, theme, goal, source, etc. encompass a variety of
information. Being an agent, for example, may imply animacy,
volition, and causation. But the crucial thing about an agent
participant in a linguistically described event is that it
represents change, motion or activity that is not
- 294 -
characterizable as change in a single parameter or property.
Animacy, volition or causation may be associated with such
change for reasons discussed in section 6.2.4 -- but these are
extra-linguistic reasons. Such an event participant must be
mapped to the external argument position because of that
position's lack of aspectual constraints on the argument. A
theme represents an event participant that can be
characterized by a change in a single property -- and
consequently, themes are mapped to internal argument
positions. Gruber's characterization of the theme as, "the
entity which is conceived as moving or undergoing transitions"
(Gruber (1965) p. 38) partly captures this fact.6  Sources
and goals represent arguments that delimit the event
described, by indirectly putting a temporal limit on the
change (usually in location) undergone by another argument -
the direct argument. For this reason sources and goals may be
mapped to positions as indirect arguments within the verb
phrase. Instruments and materials differ minimally in their
aspectual properties. Materials are consumed over time,
during the course of the event, while instruments are not.
The difference between a material thematic role and an
instrument thematic role can be minimally characterized in
these aspectual terms -- that is, in terms of how they map
6. Since Gruber includes the subjects of unergative verbs of
motion in the category of themes, 'theme' for Gruber does not
mean change describable as change in a single property.
- 295 -
into the event temporally. Section 6.3 demonstrated that the
aspectual properties of instruments and materials may
determine how they appear in argument structure. The
aspectual properties of event participants, imperfectly
expressed by 'thematic roles', are what govern the position of
a verb's arguments in the syntax. The aspectual principles of
argument structure put the correlation between thematic roles
and argument positions on a principled basis.
Thematic roles are often employed in linguistic theory and
in the literature without the backing of a proper theory of
thematic roles. The theory of thematic roles, such as it is,
seems to lack a rigorous and consistent set of diagnostics of
the various role types. For this reason thematic theory is
looser than other aspects of grammar which are better
understood. The aspectual approach to thematic roles proposed
here offers a means of setting up a rigorous and explicit
typology of thematic roles that would interact with syntax in
a principled way. The aspectual information in thematic roles
is linguistic information, relevant to syntactic argument
structure. An aspectual characterization of thematic roles
would extract this information from the extra-linguistic
information, simplifying our picture of the interface between
syntax and lexical semantics. It would maintain a view of a
precise and computationally explicit syntax, divorced from
dependence on anything so loosely defined as thematic roles
(in terms of agent, patient, etc.). It would account for
- 296 -
regularities in the mapping between cognitive structure and
syntax simply and explicitly. For theoretical and empirical
reasons, an aspectual theory of thematic roles is an
attractive option.
6.7 Thematic. hierarchies and the mapping of cognitive
structure into syntactic structure
The Aspectual Interface Hypothesis holds that the mappping
between cognitive structure and syntactic argument structure
is mediated by aspectual properties. This makes the
prediction thnt thematic hierarchies are not necessary as
mapping principles between cognitive structure and syntactic
structure. Where thematic hierarchies appear to be necessary
they should be supplantable by the AIH. With this end in view
I will examine two thematic hierarchies proposed in the
literature; one by Jackendoff (1972) which applies in
passivization,7 and'one by Belletti and Rizzi (1986), which
applies to psych verbs.
Jackendoff (1972) argues for the following thematic
hierarchy:
47
1. Agent
2. Location, Source, Goal
7. Passivization may, in the broad view, be considered as
related to syntactic argument structure, since passivization
rearranges the arguments of verbs in particular ways.
- 297 -
3. Theme
Jackendoff (1972) maintains that the following Thematic
Hierarchy Condition holds in passivization:
48
The passive 'by'-phrase must be higher on the Thematic
Hierarchy (in 47) than the derived subject.
He cites the following sentences in support:
49
a. The bookcase was being touched by John.
b. The car was hit by John (?with a crash).
c. *Five dollars are cost by the book.
d. *Two hundred pounds are weighed by Bill.
e. *Harry is struck/impressed by Bill as pompous.
f. Bill is regarded by Harry as pompous.
According to Jackendoff, the Thematic Hierarchy Condition
explains these examples in the following way: In 49a and 49b,
'the bookcase' and 'the car' are Locations or Goals.8 The
sentences are only acceptable when 'John' is an Agent, even
though the non-passive counterparts are acceptable with John
in a non-agentive role:
50
John touched the bookcase.
John hit the car (?with a crash).
When 'John' is an Agent, the thematic hierarchy is satisfied
and the sentences are grammatical. In 49c and 49d, 'five
5. 1 will capitalize the names of thematic roles when they are
used according to Jackendoff's theory.
- 298-
dollars' and 'two pounds' are Locations, while the object of
the 'by'-phrase is a Theme. The thematic hierarchy is
therefore violated, and the sentences are bad. In 49e 'Harry'
is a Goal and 'Bill' is a Theme; the thematic hierarchy is
violated, and the sentence is bad. In 49f, 'Bill' is a Theme
and 'Harry' is a Goal, the thematic hierarchy is not violated,
and the sentence is fine. (Jackendoff determines the nature
of the thematic role from the kind of prepositions that may be
employed with it.)
On a closer examination of the data, it is not so apparent
that a Thematic Hierarchy Condition is necessary for a
satisfactory explanation. Consider first the psych verb
cases, 49e and 49f. Following the discussion in section 6.5,
these two sentences differ minimally in that 49e has the
experiencer as underlyingly an internal argument and 49f has
the experiencer as an external argument. The pattern
illustrated by 49e and 49f is not a general pattern with psych
verbs:
51
a. The uncertainty of the future is feared by many people who
lack faith in themselves.
b. Many people who lack faith in themselves are frightened by
the uncertainty of the future.
a. The power of the military is respected by the opposition
party.
b. The opposition party is disgusted by the power of the
military.
- 299 -
a. Chocolate ice cream with candy cherries is liked by every
five-year-old.
b. Every five-year-old is pleased by chocolate ice cream with
candy cherries.
The a) examples in 51 above have internal argument
experiencers and the b) examples have external argument
experiencers. All the sentences are in the passive form.
Though the b) examples may seem marginally more natural than
the a) examples, there is not a strong enough difference
between them to judge the a) examples as unacceptable or
ungrammatical. The difference may be due to conditions of
felicity in discourse -- perhaps a condition that requires the
subject to be salient by containing a relatively high amount
of new information. The tendency for the a) sentences to be
awkward is counteracted by increasing the emphasis on and
information in the subject:
52
??Ice cream is liked by five-year-olds.
(?)Chocolate ice cream with candy cherries is liked by
five-year-olds.
Chocolate ice cream with candy cherries IS liked by
five-year-olds.
(in response to the statement that five year olds
only like strawberry ice cream with marshmallows.)
The effect noted by Jackendoff then, is not of the kind that
warrants a Thematic Hierarchy Condition that interacts with
syntactic processes.
- 300 -
Now consider 52c and 52d. These are not, strictly speaking,
ruled out by the Thematic Hierarchy Condition, because they
are unacceptable without any 'by-' phrase being present at
all:
57
*Five dollars are cost.
*Two hundred pounds are weighed.
A Thematic Hierarchy Condition cannot account for this data
unless it refers to implicit 'by-' phrases, which would be
problematic. The sentences are not ruled out on pragmatic
grounds either. One could imagine a situation in which it
would be pragmatically possible to say something like this.
For example, in response to the question, "How much does this
book cost?" The sentences above are not ruled out
pragmatically, nor are they ruled out by the presence of an
improper 'by-' phrase.
The ungrammaticality of 57 is an interesting fact in need of
explanation. There are a number of verbs that do not
passivize well, verbs of measurement among them. The
unacceptability of passives of verbs of measurement may be
related to the unacceptability of passives of the copula:
58
a. The barn is red.
b. *Red is been by the barn.
a. Maxwell is a terrorist.
b. *A terrorist is been by Maxwell.
-301 -
Verbs of measurement express 'meaning' close to that of the
copula, and are in fact replaceable by the copula:
59
a. Bill weighs 200 pounds.
b. Bill is 200 pounds.
a. The book costs five dollars.
b. The book is five dollars.
The impossibility of passivizing the copula is not
surprising. The passive involves a morphological and
syntactic operation. Verbal morphology is often particular
about what classes of verbs it may be applied to. This in
itself does not constitute a need for something like a
Thematic Hierarchy Condition, since we expect verbal
morphology to have lexical constraints.
A final problem with the Thematic Hierarchy Condition as
propoded by Jackendoff is that it is not clear that the
thematic roles of the arguments can be clearly and
unambiguously diagnosed. Jackendoff assumes 'Harry' to be a
Goal in 53e and 53f. However 'Harry' in 53f (repeated below as
60) has many agent-like properties.
60
Bill is regarded by Harry as pompous.
For example, one may ask about 'Harry',
61
What did Harry do after Bill came to the meeting in a tuxedo?
- 302 -
and receive as an answer,
62
Harry regarded Bill as pompous.
The following is also possible:
63
What Harry did when Bill vaulted over the table was to regard
Bill with a look of astonishment.
The applicability of 'do' is generally considered a good test
for agency. If one may be said to 'do something', one has
agent-like properties. While 'John' in 64a below may have an
agentive or a non-agentive interpretation, 'John' must be an
agent in 64b:
64
a. John touched the wall.
b. What John did was to touch the wall.
The subject of 'regard' is in many ways like an agent, yet it
is identified as a Goal by Jackendoff. I do not mean to
establish here that it is one or the other, merely to
demonstrate that the thematic roles of event participants are
not always as straightforwardly identifiable as they might
seem.9
To conclude the preceding discussion, the Thematic Hierarchy
9. See Dowty (1986) for an interesting discussion of this
point.
- 303 -
Condition of Jackendoff (1972) is unnecessary, as it is
applied to passivization, although an interesting question
remains as to why it is impossible to passivize the copula.
Furthermore, the Thematic Hierarchy Condition lacks the
rigorous set of diagnostics for thematic roles that are
necessary for it to be useful as a truly grammatical
principle.
Jackendoff (1972) also argues for the imposition of the
Thematic Hierarchy in reflexivization and control. These are
phenomena that involve coreference and binding rather than the
syntactic manifestation of argument structure so I will not
address them. It is clear that lexical semantics plays a
large part in control, and not impossible that it plays some
role in reflexivization. However, since these are processes
of a more interpretive, less structural nature, their
intersection with 'meaning', lexical semantics or cognitive
structure need not be mediated by aspectual properties and the
Aspectual Interface Hypothesis.
Belletti and Rizzi (1986) have also proposed a thematic
hierarchy in the mapping (or projection, in their terminology)
of the arguments of Italian psych verbs into syntax (from
Belletti and Rizzi (1986)):
65
Given a Th-grid (Experiencer, Theme) the Experiencer is
projected to a higher position than the Theme.
...where 'higher' means 'asymmetrically c-commanding'.
- 304 -
Belletti and Rizzi argue on syntactic grounds that sentences
such as,
66
Questo preoccupa Gianni.
"This worries Gianni."
have underlying structures like the following:
67
NP VP
V" NP
NP
ec preoccupa questo Gianni
In this structure, 'Gianni' c-commands the other internal
argument, 'questo'. Belletti and Rizzi regard 'Gianni' as
experiencer and 'questo' as theme. They conclude that a
general mapping principle that maps thematic roles into syntax
is required, to explain this structure. If this is true it
poses a problem for the Aspectual Interface Hypothesis.
An aspectual account of the mapping is available. Recall
from the discussion in section 6.5 that an experiencer in
internal argument position may be capable of measuring out and
delimiting the event (i.e., of being an affected object). In
the absence of any other delimiting argument, the experiencer
- 305 -
must be the direct argument. Adopt the following refinement
of the principles of argument structure laid out in Chapter 4:
68
The internal argument, the argument that is capable of measuring
out and delimiting the event, is the highest NP object under the VP
node.
With the refinement as stated above, the thematic hierarchy
proposed by Belletti and Rizzi is subsumed under general
principles of argument structure. The fact that the
experiencer must c-command the other internal argument follows
from general aspectual constraints on argument structure.
The preceding discussion has argued that thematic
hierarchies proposed by Jackendoff and Belletti and Rizzi to
account for the mapping (or projection) of cognitive structure
into syntactic structure are either not necessary, or can be
stated within a general aspectual theory of argument
structure. This is a small portion of the work that needs to
be done to test the Aspectual Interface Hypothesis, but it
does demonstrate its plausibility. The strongest evidence for
the kind of interaction between thematic roles and syntactic
argument structure that has fueled thematic hierarchy theories
is the simple fact that, to the best of my knowledge, agents
always show up in external argument positions and themes show
up in internal argument positions. But this is accounted for
simply and explicitly by the aspectual theory proposed in this
thesis, in which very particular aspectual/semantic
- 306 -
constraints obtain of internal arguments. Themes are the kind
of event participants that can meet those constraints; 10
agents are the kind of event participants that cannot meet
them.
The Aspectual Interface Hypothesis imposes certain
requirements on the lexical entries of predicates. Predicates
must be associated with a certain number and type of event
participants -- this is nothing new. The type of the event
participants may be expressed by something like the terms
1agent', 'patient', etc. -- this is also nothing new.
However, the Aspectual Interface Hypothesis says that the
crucial information about theifatic roles is certain aspectual
information. Each event participant must be marked as to:
(i) whether it undergoes change or not, and if so, if it
undergoes change in such a way that it can measure out the
event; i.e, whether the event participant can be an internal
argument or not. Most verbs will have only one; spray/load
verbs will have two of these.
(ii) if it is not capable of being a direct argument,
whether it has any inherent case.
(iii) whether it must delimit the event indirectly by marking
a point on the scale laid out by the direct argument. If so,
it must be mapped to an oblique argument position within
the VP.
10. I am not using the term 'theme' here in Belletti and
Rizzi's sense in which the causer of the effect on the
experiencer is called a 'theme' (as in 65). This is a further
illustration the problems resulting from the looseness of the
traditional nomenclature for thematic roles.
-5 07 -
The requirements listed above leave several questions open.
First, the question as to whether there are arguments that
must be marked as external is left open. Agents are the only
candidates for such marking. It is clear that agents cannot
be internal direct arguments, but it is not clear that they
cannot be internal oblique arguments. For example:
69
John built a house with Bill.
The salient interpretation is one in which 'Bill' is an agent,
yet 'Bill' is not an external argument. Agents are usually
external arguments, but they do not have to be. Recall
furthermore from Chapter 4 that external arguments do not have
to be agents. These facts argue that it is the internal
argument and not the external argument that is marked in the
lexical entry of the verb.1 1
A second question left open is that of optional arguments
and adjuncts. How are they to be represented in the lexical
entry of a predicate? I have bypassed this problem up to now
by taking the broadest view and considering every NP within
the verb phrase as a potential argument. (The phrase 'with
Bill' for example, in 69 may be better regarded as an adjunct
than an argument phrase.) Aspectual properties were shown in
11. Williams (1981) and Stowell (1981) argue for the marking
of external arguments in lexical entries.
-308 -
Chapter 4 to have something to say about the difference
between adjuncts and arguments, but I will not follow this up
here. The question of the proper treatment of adjuncts versus
arguments in the lexical entries of predicates will be put
aside for future research.
The formulation of the requirements on lexical entries of
predicates outlined above suggests that the mapping from
cognitive structure to syntactic structure proceeds as
follows:
(i) map an argument that is marked as a possible direct
argument into the direct argument position. It receives
structural accusative case.12
(ii) map other arguments marked as possible recipients of
inherent case into oblique argument positions within the VP
(or positions as objects of prepositions).
(iii) map an argument into external argument position. If
there is an argument incapable of being mapped in (i) and (ii)
(this will usually be an agent), it must become an external argument
Under this mapping procedure the external argument is not
distinguished before it is mapped into the external argument
position. The external argument is certainly distinguished
syntactically (and aspectually) once it is mapped into the
syntax, but under the view espoused here, it is not inherently
so distinguished in the cognitive structure associated with
the lexical entry. It is rather the internal direct argument
12. Or nominative case after passivization.
- 309 -
that is special, in the lexical entry. Anything can be an
external argument but only very particular kinds of things can
be direct internal arguments. If (ii) and (iii) are freely
ordered, the mapping procedure will allow non-agents to occupy
external argument position (which does happen). The mapping
also predicts that agents will not be blocked absolutely from
occupying oblique argument positions, if the appropriate
inherent case (or the preposition) is available in the
language.
6.8 The AIH and learnability
The Aspectual Interface Hypothesis has been proposed as the
connection between cognitive structure and .linguistic
structure. The AIH maintains that the aspectual properties of
events constitute a kind of 'shared language' between the two
systems. We must ask the question: how do children learn the
aspectual properties of verbs? Do they first know the
aspectual meanings and hook them up to the right syntax, or do
they hear the syntax and discover the aspectual meanings? It
is difficult to answer such a question knowing as little as we
do about cognitive structure, but some speculation is
possible.
Cognitive structure and linguistic structure are separable,
since higher mammals and very small children who do not have a
command of language must be credited with sophisticated
cognitive abilities. When the child has the cognitive ability
- 310 -
to characterize or classify events, she also knows what
aspectual properties the events may have. The. process of
learning the argument structure of verbs is then a process of
connecting the general event type of a verb with a possible
aspectual characterization of the event. Some kinds of events
may be characterized aspectually in only one way; for these
the child has essentially nothing to do. Others may be
characterized in several possible ways; for these it is a
matter of choosing one of several possible aspectual
instantiations of the event type. Positive evidence will be
availdble from the syntax to guide this choice.
Even though verbal argument-structures must be learned, the
shared vocabulary of aspectual properties, at the interface of
cognitive and linguistic structure, makes that task relatively
simple. The AIH helps to explain how children can learn
language as quickly as they do.
To conclude, the Aspectual Interface Hypothesis contributes
to an explanation of the ease with which children learn
language; provides a principled basis for the correlation of
cognitive structure and syntactic argument structure; suggests
the outlines of an explicit theory of thematic roles; provides
a plausible alternative to thematic hierarchies and linking
rules; accounts for certain diathesis alternations simply; and
forms the basis of a principled statement of a constraint such
as the UTAH or UAH. The Aspectual Interface Hypothesis is
- 311 -
theoretically and empirically worthy of serious consideration.
- 312 -
Abney, S. (1967), "The Ebglish Noun Phrase in its Sentential
Aspect", Ph.D. Dissertation, MIT.
Abusch, D. (1986) "Verbs of Change, Causation, and Time", Center
for the Study of eon-wae and Information at Stanford
University, Reprint 7 CSLI-86-50.
Anderson, M. (1979) "Noun Phrase Structure", Ph.D dissertation,
University of Connecticut.
Anderson, S. (1977), "Comments on the Paper by Wasow", in
Culicover, P., Wasow, T., and Akmajian, A. (eds.), Formal
Syntax, New York: Academic Press. pp. 361-378.
Aristotle, Metapyics, Books I-Il, translated by H.
Tredennick,7(193, Cambridge, Massachusetts: Harvard University
Press.
Armstrong, D. (1981), "The Ancient Greek Aorist as the Aspect of
Countable Action", in P. Tedeschi and A. Zaenen (eds.),
Syntax and Semantics Vol. 14, Tense and Aspect, New York:
Academic Press. pp. 1-12.
Bach, E. (1979), "Control in Montague Grammar", in Linguistic
Inquiry 10, Cambridge, MA: MIT Press, pp. 515-531.
Baker, M. (1986), "A Theory of Grammatical Function Changing",
Ph.D dissertation, MIT.
Baker, M., Johnson, K. and Roberts, I. (1967), "Passive Arguments
Raised", me.; McGill University, UC Irvine, and Universite de
Geneve.
Bares, A. and Lasnick, H. (1986), "A Note on Anaphora and Double
Objects", in Li 17, Cambridge, MA: MIT
fres, pp. 347354.:
Belletti, A. (1986), "Unaccusatives as Case Assigners",
Lexicon Project Working Papers 8, Cambridge, MA: Center for
Cognitive Science at MIT.
Belletti, A. and Rizzi, L. (1986), "Psych-Verbs and Th-Theory",
Lexicon Project Working Papers #13, Cambridge, MA: MIT Center
for Cognitive Science.
Bennett, M. (1981), "Of Tense and Aspect: One Analysis", in P. Tedeschi
. and A. Zaenen (eds.), Syntax and Semantics Vol. 14, Tense
and Aspect, New York: Academic rees. pp. 13-30.
- 313 -
Bennett, M. and Partee, B. (1978), "Toward the Logic of Tense and
Aspect in Thglish", ms., Reproduced by the Indiana University
Linguistics Club.
Bittner, M. (1986), "On the Semantics of the Greenlandic
Antipassive and Related Constructions", University of Texas
at Austin, ms.; to appear in (1987), International Journal
of American Linguistics.
Bolinger, D. (1971), The Phrasal Verb in English, Cambridge,
MA: Harvard University Press.
Brecht, R. (1984), "The Form and Function of Aspect in Russian",
in Flier, M. and Brecht, R. (eds.), Issues in Russian
Morphosyn- , (1985), Columbus, Ohio: Slavica Publishers,
Inc. Pp. 10-34.
Burzio, L. (1986), Italian Syntax, A Government and Binding
Approach, Dordrecht: Reidel.
Carlson, G. (1977), "A Unified Analysis of the English Bare
Plural", Li stics and Philos 1, Dordrecht:
Reidel. pp137.
Carlson, L. (1981), "Aspect and Quantification", in P. Tedeschi and
A. Zaenen (eds.), Syntax and Semantics Vol. 14 Tense
and Aspect, New York: Academic Press. pp. 31 -64.
Carrier-Duncan, J. and Randall, J. (1967), "Lexical Rules and Lexical
Representations: Evidence from Resultatives", Harvard
University and Northeastern University, ms.
Chomeky, N. (1975), TheicalStructure of Linguistic
Thepry, Chicago: University of Chicago Press.
Chomsky, N. (1957), 4yntacticStructures, The Hague: Mouton.
Chcmsky, N. (1965), Aspects of the Theory of Syntax,
Cambridge, MA: MIT Press.
Chemeky, N. (1982), Lectures on Government and Binding,
Dordrecht: Foris.
Chomsky, N. (1986), Barriers, Cambridge, MA: MIT Press.
Cowrie, B. (1976), p Cambridge: Cambridge University
Press.
Culicover, P. (1982), , New York: Academic Press.
Dahl, 0. (1961), "On the Definition of the Telic-Atelic
(Bounded-Nonbounded) Distinction", in P. Tedeschi and A. Zaenen
- 314 -
(eds.), Sntax and Semantics Vol. 14, Tense and Aspect, New York:
Academic Press. pp.79-90.
Davidson, D. (1966), "The Logical Form of Action Sentences", in
Davidson, D. Essays on Actions and Events, Clarendon
Press, Oxford. pp. 105-122.
De Groot, C. (1984), "Totally affected. Aspect and three-place
predicates in Hungarian", in C. de Groot and H. Tommola,
eds., Aspect Bound. A voyage into the realm of Germanic.
Slavonic and Finno-Ugrian aspectology, Dordrecg: Foris.
PP 133-151.
Dik C. (1980), Studies in Functional Grammar, London and New
York: Academic Press.
Dowty, D. (1979), Word Meaning and Montague Grammar,
Dordrecht: Reidel.
Dowty, D. (1986), "On the Semantic Content of the Notion
"Thematic Role", Ohio State University, ms.
Dubinsky, S. and Rosen, C. (1987), A bibliography on relational
granoar through May 1987, with selected titles on lexical
functional grammar, Bloomington, Indiana: Indiana University
Linguistics Club.
Eckert, E. (1984), "Aspect in Repetitive Contexts in Russian and
Czech", in flier, M. and Timberlake, A. (eds.), The Scope
of Slavic Aspect, (1985), Columbus, Ohio: Slavica
PbIhshers, In. pp. 169-180.
Eonds, J. (1972), "Evidence that Indirect Object Movement is a
Structure-Preserving Rule", Foundations of Language 8,
pp. 546-561. Reprinted in M. Gross, M. Halle and M.P.
Schutzenberger eds.), The Formal AnaLysis of Natural
I g The Hague: Mouton 1975.
Enonds, J. (1976), A Transformational Approach to English
n , New York: Academic Press.
Enc, M. Tense Without Scope: An Analysis of Nouns as
Indexicals, PhD. dissertation, University of
Wisconsin-Madison.
Fassi Fehri, A., (1987), "Anti-Causatives in Arabic, Causativity
and Affectedness", Lexicon Project Working Papers 15,
Cambridge, MA: Center for Cognitive Science at MIT.
Fillmore, C. (1965), Indirect object constructions in English
and the ordering of trsformations, The Hague: Mouton & Co.
Fillmore, C. (1968), "The Case for Case", in Bach, E. and Harms,
- 315 -
R., eds., Universals in Linguistic Theory, New York:
Holt, Rinehart and Winston, Inc., pp. 1-88.
flier, M. (1984), "The Scope of Prefixal Delimitation in
Russian", in Flier, M. and Timberlake, A. (eds.),
The Scope of Slavic Aspect, (1985), Columbus, Ohio:
Slavica Publishers, Inc. 41-58.
Fraser, B. (1976), The Verb-Particle Combination in English,
New York: Academic Press.
Fukui, N. (1986), "A Theory of Category Changing and its
Applications", PhD. dissertation, MIT.
Fukui N., Miyagawa S., and Tenny C. (1985), "Verb Classes in English
and Japanese: A Case Study in the Interaction of Syntax,
Morphology and Semantics", Lexicon Project Working Papers M5,
Cambridge, MA: MIT Center for Cognitive Science.
Fukui, N. and Speas, M., (1986), "Specifiers and Projection", MIT
Working Papers vol. 8.
Gee, J. (1975), "'Get Passive': On Some Constructions with 'Get'",
Indiana University Linguistics Club mimeo.
Givon, T. (1984), Syntax. A Functional-Typplogical
Introduction, Vol. I, Amsterdam/Philadelphia:
John Benjamins Publishing Company.
Green G. (1974), Semantics and Syntactic Regularity,
Bloomington and London: Indiana University Press.
Gruber, J.S. (1965) "Studies in Lexical Relations", Ph.D.
dissertation, MIT; in Gruber, J.S. (1976), Lexical Structures in
Syntax and Semantics, Amsterdam: North Holland.
Guerssel, M. Hale, K.; Laughren, M.; Levin, B. and White bgle,
J. (19855, "A Cross-Linguistic Study of Transitivity
Alternation", in W.E. Eilfort, P.D. Kroeber and K.L Peterson,
eds., Pawrs from the Parassession on Causatives and
ntivi at the Tirst Rion eeti , Chicago
14-giis CScit, Chcago, I, pp. 4 -
Hale, K. and Keyser, J. (19S7), "A View from the Middle",
Lexicon Project Working Papers 10, Cambridge, MA: Center for
Cogitive Science at MIT.
Hale, K. and Keyser, J. (1986), "Some Transitivity Alternations
in English, Lexicon Project Working Papers 7, Cambridge, MA:
Center for Cognitive Science at MIT.
Halle, M. and Vergnaud, J. (1987), "Stress and the Cycle"
Linguistic Inquiry, 18.1, Cambridge, MA: MIT Press,
-
316 -
p . 45-64.
Halliday, M. (1967-8), "Note on transitivity and theme", Journal
of Linguistics, 3, pp. 37-81, 199-244.
Heinamaki, 0. (1984), "Aspect in Finnish", in de Groot, C. and
Tommola, H. (eds.), Aspect Bound. A vpyage into the realm of
Germanic, Slavonic and Finno-Ugrian aspectology, Dordrecht: Foris
Publications. pp. 155-178.
Higginbotham, J. (1985), "On Semantics", in Linguistic Inquiry,
16, Cambridge, MA: MIT Press, pp. 547-593.
Higginbotham, J. (1966), "Elucidations of Meaning", ms. MIT.
Hinrichs, E. (1965), "A Compositional Semantics for
Aktionsarten and NP Reference in Tnglish", Ph.D dissertation,
Ohio State University.
Holisky, D. (1981), "Aspect Theory and Georgian Aspect", in P. Tedeschi and
A. Zaenen (eds.), Syntax and Semantics Vol. 14, Tense and Aspect,
New York: Academic Press. pp. 127-143.
Hopper, P. and Thompson, S. (1980), "Transitivity in Grammar and
Discourse", 4u , 56. pp. 251-299.
Hornstein, N. (1977), "Towards a Theory of Tense", Linguistic
InauirYL2 Cambridge, MA: MIT Press. pp. 521-557.
Jackendoff, R. (1972), Semantic Interpretation in Generative
Grammn&', Cambridge, MA: MITPress.
Jackendoff, R. (1985), Semantics and Cognition, Cambridge,
MA: MIT Press.
Jackendoff, R. (1986), "Parts and Boundaries", talk delivered at MIT
Lexicon Seminar.
Jackendoff, R. (197), "The Status of Thematic Relations in
Linguistic Theory", in Li stic In ui , 18-3,
Cambridge, MA: MIT Press, pp. .94117
Jacobson, P. (forthcoming), "Phrase Structure, Grammatical
Relations and Discontinuous Constituents", to appear in G. Huck
and A. Ojeda, eds., Syntax and Semantics 20:
Discontinuous Constituency, New York: Academic Press.
Kangm -inn, E. (1984), "Tense, Aspect and Aktionsarten in
Finno-Ugrian", in de Groot, C. and Tommola, H. (eds.),
Aspect Bound. A yge into the realm of Germanic,
Slavonic and Finno-Ugrian aspec , Dordrecht F oris
ublicaions. pp.T7-96.
- 317 -
Kayne, R. (1985), Connectedness and Binary Branching, Dordrecht:
Foris Publications.
Keyser, S. and Roeper, T. (1984), "On the Middle and Ergative
Constructions in English", in Lin4stichInuiry15-3,
Cambridge, MA: MIT Press, pp. .38 417
Kenny, A. (1965),
Routledge and
Action Botion and Will, London:
--egan kaui.
Kiparsky, P. (1985), "Morphology and Grammatical Relations", ms.
draft, Stanford.
Klipple, B. (1987), "Verb Semantics and the Aspectual Property of
Boundedness", m. MIT.
Kuhn, T. (1981), "What are Scientific Revolutions?", Occasional Paper
18, Cambridge, MA: MIT Center for Cognitive Science.
Lakoff, G. (1971), "On Generative Semantics", in Steinberg and
Jakobovits, eds., Semantics: An Interdisciplinary Reader
in Philosopy Li W stics Anthrop loa and Ps chol
Iarson, R. (1967), "On the Double Object Construction",
Lexicon Project Working Papers 16, Cambridge, MA: Center for
Cognitive Science at MIT.
Laughren, M., Levin, B., and Rappaport, M. (in preparation),
"What's Behind Theta-roles: What Syntax Tells us about Lexical
Representations", me., MIT.
Levin, B. (1986): List of verb classes for the MIT Lexicon Project,
revised, unpublished.
Lieber, R. (1980), "On the Organization of the Lexicon",
dissertation, MIT.
Lumden, J. (197), "Syntactic Features: Parametric variation in
the history of English", PhD dissertation, MIT.
Marantz, A. (1984), On the Nature of Grammatical Relations,
Cambridge, MA: MIT Press.
Martin, S. (1976) A Reference Grammar of Japanese, New
Haven: Yale University Press.
May, R. (1985), Iggical Form: Its Structural Derivation,
Cambridge, Mk: MIT Press.
McCawley, J. (1971), "Interpretive semantics meets Frankenstein",
in Foundations of language 7, pp. 285-296.
- 318 -
-PhD.
lb
McCawley, J. (1971 ), "Tense and Time Reference in Tglish", in
Studies in Linguistic Semantics, New York: Holt,
Rinehart and Winston. pp. 97-113.
Merlan, F. (1985), "Split intransitivity: functional oppositions
in intransitive inflection", in J. Nichols and A. Woodbury,
eds., Grammar inside and outside the clause Some
approathes to lhory from the field, Cambrige:
Cambridge University Press, pp. 324-362.
Merrill, P. (1984), "Universal Quantification and Aspect in
Russian", in Flier, M. and Brecht, R. (eds.), Issues in Russian
C(1985), olumbus, Ohio: Slavica Publishers,
Mittwoch, A. (1971), "Idioms and Unspecified NP Deletion",
LiMstic Inqniry, 2.2, Cambridge, MA: MIT Press,
squib,7pp.5-5.
Mittwoch, A. (1980), "The Grumnar of Duration", Studies in
Lang , 4.2, Amsterdam: John Benjamins BV. pp. 201-27.
Miyagawa, S. (1967), "On 'Theme' Subjects in Japanese", talk
delivered at MIT lexicon Seminar; ms. MIT.
Mourelatos, A. (1981), "Events, Processes and States", in P. Tedeschi and
A. Zaenen (eds.), .ytax and Semantics Vol. 14, Tense and Aapect,
New York: Academic Press. pp. 191-212.
Oehrle, R. (1975), The Grammatical Status of the Eglish
Dative Alternation, PhD. dissertation, MIT.
Partee, B. (1975) "Some Structural Analogies between Tenses and
Pronouns in Eglish", Journal of Philosophy, Vol. 70, No.
18, New York: The Journal of Philosophy, Inc. pp.601-609.
Perlmutter, D. (1978) "Impersonal Passives and the Unaccusative
Hypothesis", in Proceedin of the Fourth Annual Meeting
of the Berkeley linguistics Society, Berkeley, CA.
Perlmutter, D. and Postal, P. (1984), "The 1-Advancement
Exclusiveness Iaw", in Perlmutter, D. and Rosen, C. (eds.),
Studies in Relational Grammar 2, Chicago: University of
Chicago Press. pp. 81-125.
Postal, P.(1970), "On the Surface Verb Remind", Linguistic
In4uiry Vol. 1, no. 1.
Rapoport, T. (1987), "Gopular, Nominal and Small Clauses: A Study
of Israeli Hebrew", PhD. dissertation, MIT.
Rappaport, M., and Levin, B. (1984), "A Case Study in Lexical
- 319 -
Analysis: tae Spray/Load Alternation". Revised as Rappaport,
M. and B. Levin, (1985), A Case Study in Lexical Analysis:
The Locative Alternation", ms, MIT.
Rappaport, M. and Levin, B. (1986), "What to do with Theta-Roles",
lexicon Project Working Papers 11, Cambridge, MA: Center for
Cognitive Science at MIT.
Reichenbach, H. (1947), Elments of Symbolic logic,
New York: Maamillan.
Riemsdijk, H van, and Williams, E. (1986), Introduction to
the Theory of Grammar, Cambridge, MA: MIT Press.
Rizzi, L. (1986), "Null Objects in Italian and the theory of
"s, Cambridge, MA:
Roberts, I. (1985), "The Representation of Implicit and
Dethematized Subjects", Ph.D. dissertation, University of
Southern California.
Rosen, C. (1984), "The Interface between Semantic Roles and
Initial Gramatical Relations", in Perlmutter, D. and Rosen, C. (eds.),
Studies in Relational Grammar 2, Chicago: University of
Chicago Press. pp. 38-77.
Rothstein, S. (1979), "The Syntactic Forms of Predication",
PhD. dissertation, MIT.
Russell, P. (1984), "Aspectual Properties of the Russian Verbal
Prefix na-",. in Flier, M. and Timberlake, A. (eds.),
The Scoof Slavic Aspect, (1985), Columbus, Ohio: Slavica
Publishers, Inc. pp. 59-75.
Htyle, G. (1949), The Concept of Mind, London: Hutchinson's
University Library.
Saito, M. (1985), "Some Asymmetries in Japanese and their
Theoretical Implications", PhD. dissertation, MIT.
Schein, B. (1986), "Event logic and the Interpretation of
Plurals", Ph.D. dissertation, MIT.
Scwartz-Norman, L. (1976), "The Grammar of 'Content' and
'Container"', Journal of Linguistics, 12, pp 279-287.
Seljakin, M. (1984), "On the Essence of the Category of
Aspectuality and its Lexical-Semantic Level in Russian"
in de Groot, C. and Tommola, H. (eds.), Aspect Bound.
A vo into the realm of Germanic, Slavonic and
FN nan as l , Dordrecht: Foris Publications.
PP. 3-5.
-320-
Selkirk, E. (1982), The Syntax of Words, Cambridge, MA: MIT
Press.
Simpson, J. (1982), "Secondary Predicates in English, Icelandic
and Finnish", s. MIT.
Simpson, J. (1985), "Aspects of Waripiri Morphology and Syntax",
PhD. dissertation, MIT.
Sloan, K. (in prep), "Wh quantifier interaction", ms., MIT.
Sith, C. (1978), "The Syntax and Interpretation of Temporal
Expressions in Tglish", Linguistics and Philosophy 2,
Dordrecht: Reidel. pp.43-T
Speas, M. (1966), "Adjunction and Projections in Syntax", PhD.
dissertation, MIT.
Steele, S., with Akmajian, A., Demers, R., Jelinek, E., Kitagawa,
C., Oehrle, R., Wasow, T. (1981), AnEncyclopedia ofAUX;
A Study of Cross-Linguistic Equivalence, Cambridge, MA:
MITress.
Stowell, T. (1981), "Origins of Phrase Structure", PhD. dissertation,
MIT.
Taylor, B. (1977), "Tense and Continuity", Ligfstics and
Philosophy 1, Dordrecht: Reidel. pp. 199-220.
TalsW, L. (1985), "Lexicalization Patterns: semantic structure
in lexical forms", int lo and tactic
descrition III; Grammatic categoriesWandthe lexicon,
T. Shopen, ed., Cambridge: Cambridge University Press.
Timberlake, A. and Chung, S. (1985), "Tense, aspect, and mood",
in Shopen, T. (ed.), Iangag:ae typolog and syntactic
description, Vol III, Grammatical categories and the lexicon,
Cambridge: Cambridge University Press. pp. 202-258.
Tsunoda, T. (1985), "Remarks on Transitivity", Journal of
Linguistics 21, London: Cambridge UniversityPress. pp. 385-396.
Van Oosten, J. (1977), "Subjects and Agenthood in English", in
Chicago Linguistic Society 13, pp. 459-471.
Van Valin, R. (1997a), talk on verb classes, delivered at MIT
Lexicon Seminar.
Van Valin, R. (1987b), "The Unaccusative Rypothesis vs. Lexical
Semantics: Syntactic vs. Semantic Apporaches to Verb
Classification", Talk delivered at NEW 17, to be published
- 321 -
in Proceedino of NXES 17;'Cambridge, MA:
MIT Deparbment of Liaguistics.
Vendler, Z. (1967), "Verbs and Times", In.Z. Vendler,
!4,2tics In ,hil Ithaca, New York: Cornell
Untiversity Press.
Verkoyl, H.J. (1972) On the Coppositional Nature of the
Aspyects, Dordrecht: R1eidel.
Verkuyl, H.J. (1976), "Interpretive Rules and the Description of
the Aspects", Foundations of language 14, Dordrecht: Reidel.
pp. 471-505.
Verkuyl, H.J. (1978), "Thematic Relations and the Semantic
Representation of Verbs Expressing Change", in Studies in
29 ,pp. 199-233.
Verkuyl, H.J. (1987a), "Nondurative Closure of Events", in
J.A.G. Groenendijk, D. de Jongh, and M.J.B Stokhof (eds.),
Studies in Discourse Representation Theorr and the Theory
of Generalized Quantifiers Proceedings of the 5th
Amsterdam Colloquium (1945, Dordrecht: Foris Publications.
Verkuyl, H.J. (19S7b), "Aspectual Asymetry and Quantification",
to appear in V. TEhrich & H. Vater (eds.), T 4ralitat,
(Proceedings of the Heidelberg Workshop on Tense and pect,
Feb. 1986)
Williams, B. (1981), "Argument Structure and Morphology", in
Linguistic Review 1:81-114
Woisetschlager, E. (1977), "A Semantic Theory of the English
Auxiliary System", PhD. dissertation, MIT.
Zaenen, A. (1967), "Are there unaccusative verbs in Dutch?", Paper
presented at NES 17, to appear in Proceedif of NE 17,
Amherst: U Mass GLSA.
- 322 -