Upload
hoangtram
View
217
Download
3
Embed Size (px)
Citation preview
THE SEMANTIC VARIABILITY OF PRESENT-PARTICIPIAL FREE ADJUNCTS
KATSUKO TOMOTSUGU
1. Introduction
A free adjunct is Ha nonfinite predicative phrase with the function of an
adverbial subordinate clause.") It is signaled by an intonational break in
speaking or by commas in writing. It has a verbal or a non-verbal head: a
verbal head can be a present participle, past participle, or infinitive, and a
non-verbal head can be a noun phrase, prepositional phrase, adjective
(phrase), or adverb (phrase). The adjunct construction poses an interpreta
tion problem, i.e. the identification of the semantic relations between an
adjunct and its matrix clause. The aim of this paper is to elucidate a
mechanism of interpretation involved in the present-participial adjunct
construction.
The interpretation problem is attributed to the fact that the semantic
relations between the adjunct and the matrix clause are not determined in
form, i.e. adjuncts lack such subordinating conjunctions as when, because, if,
though, etc. Also, in adjuncts neither the tense nor the subject is marked.
Thus, two events are represented in a sentence without specifying their
semantic relations.
(1) Crossing the street, John was hit by a car.
Because of the indeterminacy in form, context plays an important role in
interpretation. This does not mean, however, that no systematic explanation
is possible in the adjunct construction. Stump demonstrates that both
[143)
144
semantic and pragmatic factors determine the role of free adjuncts. 2 His
semantic factors are modal verbs, relative frequency adverbs, and generaliza
tion. His pragmatic factors are inferences by language users based on
linguistic forms and world knowledge. The roles of adjuncts range from a
temporal setting to a logical relation and to a speaker's attitude. Sweetser
offers a new approach to lexical polysemy and pragmatic ambiguity, showing
that the same linguistic form can be applied to a different domain. 3 Her
analyses of conjunctions and conditionals are applied to multiple uses of free
adjuncts that have been distinguished intuitively.
The next section summarizes the characteristics of the present-participial
adjunct construction. Section 3 deals with the relations of modal verbs in
matrix clauses to the adjunct interpretation based on Stump's work. In
section 4, factors that motivate logical relations rather than temporal
relations, and the three domains concerned with the adjunct interpretation
are discussed. Predicate types of adjuncts, and modification of speaker's
reasoning or speech act figure in the interpretation processes. Adjuncts with
overt subjects, called absolutes, such as Other things being equal, the difference
in cost was decisive, and adjuncts with subordinating conjunctions, such as
While crossing the street, John was hit by a car, are beyond the scope of the
paper.
2. The present-participial free adjunct construction
The adjunct denotes an event, and the semantic relations expressed by
present-participial adjuncts are time, cause or reason, condition, concession,
accompanying circumstances, and manner.
(2) Writing the final term paper, John gave a large sigh of relief. (3) Knowing the importance of his words, the witness kept silence.
The adjunct in (2) serves as temporal overlap (when he wrote) or cause
145
(because he wrote), whereas the adjunct in (3) serves as cause (because he
knew) or concession (although he knew). From the fact that adjuncts are
paraphrasable, traditional studies proposed reduction operations for free
adjuncts from their full adverbial clauses. Adjuncts expressing temporal
relations, cause, condition, and concession can be restated using subor
dinating conjunctions. However, not all adjuncts are paraphrased by means
of adverbial clauses:
(4) Some of the apartments are of the terrace type, being on the ground floor so that entrance is direct.
(5) The siren sounded, indicating that a fire had broken out. (6) Bursting the paper cartridges, he scattered powder over the floor. (7) Tossing a ball and catching it, John talked to his mother.
Adjuncts expressing specification, explanation, and manner have no counter
part adverbial clauses. Matrix clauses and adjuncts in (4)-(6) refer to one and
the same event or state, so that their relations are not expressed by
conjunctions connecting two events. The adjunct of accompanying circum
stance in (7) shares the time and place with the matrix event and is asserted
simultaneously, rejecting a paraphrase by an adverbial clause.
The present participle -ing form does not necessarily denote the progres
sive. The difference between the finite form and the progressive is
neutralized in adjuncts: crossing the street represents the clause when he crossed
the street or while he was crossing the street. Stative verbs that do not take the
progressive show the form in adjuncts such as knowing in (3). The sentence
He was knowing the importance of his words is not possible. The progressive
itself does not appear in adjuncts: • Being crossing the street, John was hit by a
car. The present participle does not refer to a specific temporal point to
which the matrix event is connected; its function is a linkage of the adjunct
predicate to the matrix predicate. Since the temporal relations are not
146
determined 1ll the form, predicate types assist In the interpretation.
The perfect participle appears in adjuncts denoting anterior events.
Anteriority allows causal interpretation because being anterior is the
prerequisite of a cause.
(8) Having failed the exam twice, he didn't want to try agalll.
The perfect-participial adjunct expresses an event that is true in the interval
whose upper boundary is the time of the matrix event. The time span between
the adjunct event and the matrix event is not restricted. The adjunct event
can be a day or even several years before the matrix event. Time adverbs that
co-occur with the past tense can be used in the perfect-participial adjunct.
(9) Having been on the bus yesterday, he knows about the accident.
The perfect participle neutralizes the distinction between the preterit and the
perfect. As for the tense and aspect of the adjunct construction, tense is
determined by the matrix clause or a time adverb, and aspect is not restricted
to the progressive.
Word order in the adjunct construction often reflects the iconic order of
event. Adjuncts of condition and earlier time tend to precede their matrix
clauses, and adjuncts of subsequent time, accompanying circumstances,
specification, and result tend to occur in final position. Based on a corpus
study, Kortmann has found that a high percentage of those present-participial
free adjuncts/absolutes expressing temporal succession obey the iconicity
constraint: more than 80% of those expressing anteriority and about 98% of
those expressing posteriority.4
3. Adjuncts in the modal environment
Lack of conjunctions to specify the relations between events induces
various interpretations. In his formal account of the semantic variability of
147
the adjunct construction, Stump concludes that semantic and pragmatic
factors work jointly to determine the interpretation. Stump rejects the
syntactic derivation of adjuncts from corresponding adverbial clauses and
also the exclusive context dependency that rules out any formal explanation.
He demonstrates that predicate types in adjuncts determine the semantic
relation (logical role) when the matrix clause has a modal verb.5
(lO) a Wearing that new outfit, Bill would fool everyone. b. Standing on a chair, John can touch the ceiling.
(ll) a Weighing only a few tons, the truck might reach the top of that hill.
b. Having unusually long arms, John can touch the ceiling.
The matrix clauses have modal verbs of would, can, and might. The adjuncts
in (10) serve as conditional: if he wore that new outfit and if he stands on a chair.
The adjuncts in (11) serve as causal: because it weighs only a few tons and
because he has unusually long arms. For (10) the actual truth of the adjuncts is
not entailed; for (11) the actual truth of the adjuncts is entailed. The adjuncts
in (10) are termed weak adjuncts, whose actual truth can fail to be entailed,
and the adjuncts in (11) are termed strong adjuncts, whose actual truth is
entailed. The weak and strong adjunct distinction coincides with the
distinction between stage-level and individual-level predicates developed
by Carlson.6 Stage-level predicates include all dynamic predicates and some
stative predicates such as drunk, asleep, and available, denoting properties of
spatio-temporal manifestation of individuals. Individual-level predicates
denote inherent, essential sorts of properties. To explain the conditional
interpretation of weak adjuncts in the environment of modal verbs, Stump
assumes that modals are binary operators which specify logical relations
between two arguments, and weak adjuncts occupy the first argument. The
rough translation for (lOa) can be shown as follows:
148
would' CBill-wears-that-new-outfit') CBill-fools-everyone').7
Weak adjuncts condition the interpretation of the modal would to give the
same function as an if-clause, and their truth is not entailed. On the contrary,
strong adjuncts are outside the scope of modals and function as main tense
adverbs. The translation of (lIb) can be shown roughly as,
LC J ohn-has-unusually-Iong-arms ') C J ohn-can-touch-the-ceiling'),
where L denotes a logical relation.
The role of weak adjuncts is directly determined by the semantics of
modals. Stump analyzes four modal verbs must, can, would, and might with a
notice that other modal verbs can be dealt with as well. Of the four modal
verbs the positive can has no epistemic sense. The examples Stump gives for
would and might reveal the epistemic sense. Must has some peculiar features
and its peculiarity is reflected in the adjunct interpretation. Although
referring to the content of each modal verb is beyond the scope of his
research, the following discussion about must and can will elaborate on
Stump's view that modal semantics affects the role of weak adjuncts.
Must expresses present modality both in root and epistemic senses. An
obligation is present at the time of speech. Shalllwill have to replaces must to
refer to an obligation which does not exist at the time of speech but may arise
in the future, often dependent on the future fulfilment of a condition, e.g. If
you forget to take your passport, you 7l have tol *must come back for it.s Epistemic
must implies that the speaker has confidence in the truth of his utterance so
that it is not normally used when the logical conclusion concerns a future
situation, e.g. The parcel shouldl?must reach her tomorrow.9 The events on the
basis of which obligation or logical necessity is presented are actual. The
truth of the adjunct is entailed when its matrix clause has the modal must.
Therefore, even the weak adjunct receives a causal interpretation.
149
(12) Boiling the pot over, the woman must clean the stove.
The adjunct in (12) expresses the cause because she boiled the pot over, and not
the condition if she boils the pot over.
Another example in which modal verbs influence the adjunct interpreta
tion is the sentence:
; (13) Standing on the chair, John can see the performance.
In the case of can referring to possibility, the truth of adjunct is not entailed:
if he stands on the chair. On the other hand, as a sub-group of can of ability, can
occurs with verbs of perception (see, hear, feel, etc.) only for the purpose of
stating the present aspect, because the progressive form of those verbs is not
acceptable: * he is seeing. 10 When can refers to the present state, the truth of the
adjunct is entailed. In that case, (13) entails that John is standing on the
chair.
Stump has found two other environments where the weak/strong adjunct
distinction is effective: matrix clauses with adverbs of relative frequency
such as sometimes, always, and generic or habitual sentencesY
(14) a Weighing four tons, our truck often makes the bridge shake. b. Being a businessman, John carries a briefcase.
(15) a Carrying a load of over 1500 Ibs., our truck often makes the bridge shake.
b. Riding the commuter train, John carries a briefcase.
Again the actual truth of the adjuncts in (14) is entailed with a causal
interpretation. The function of the weak adjuncts in (15) is to specify the
time intervals in which the proposition expressed by the matrix clause is
true. Frequency adverbs and the generalization operator coined by Stump for
generic or habitual sentences, denote two-term relations, the first argument
of which is occupied by the weak adjunct.
150
Stump shows three linguistic environments, I.e. modal verbs, relative
frequency adverbs, and generic or habitual sentences, where the adjunct
interpretation is explained formally. Semantic relations between adjuncts
and matrix clauses without modals, frequency adverbs, and generalization
are indeterminate and left to the inferences of language users. Factors that
constitute the language users' inferences are stage-levellindividual-level
predicate types, duration of events, iconic word order, world knowledge, and
the presence of such connective adverbs as meanwhile, thus, nevertheless, etc.
According to Stump, linguistic information or language users' inferences, or
a combination of the two, accounts for the adjunct interpretation. For him
both semantics and pragmatics are involved in the study. In the next section,
what Stump calls inferences of language users (henceforth, speakers) will be
discussed from a different approach.
4. Inferences of speakers in the adjunct construction
The semantic variability of a single construction involves two divergent
ways of explanation: one is to find linguistic information that corresponds to
a rigidly defined function, and the other is to characterize the speaker's
inferences that cause multiple form-to-function mappings. This section takes
the latter way of explanation and shows how diverse meanings are structured
around the speaker's conception. Two kinds of processing are proposed for
the approach: a choice between temporal and non-temporal relations, and the
speaker's involvement in· the matrix proposition which is realized as
application to different domains.
Two events are placed side by side in the adjunct construction with the
adjunct verbs being nonfinite. Temporal relations are fixed when the matrix
event is connected to a time interval in the adjunct. Dynamic predicates are
schematized as time intervals with the end points at the beginning and at the
end. The time interval in the adjunct to which the matrix event is connected
IS not determinate:
(l6) a Crossing the street, John was hit by a car. b. Crossing the street, John entered the bank.
151
The matrix event refers to a middle interval of the adjunct in (16a) and the
final interval of the adjunct in (16b). Duration of events, presence of
endpoints, and our knowledge on the plausibility of action determine the
temporal overlap or temporal succession in the adjunct construction.
When an adjunct fails to be related temporally to the matrix clause,
non-temporal relations are induced. Such cases are found in the adjuncts of
negation, and in the adjuncts with stative predicates that are not bounded
temporally. In both cases, it is hard to identify a time interval to connect
another event, and the adjuncts are interpreted causally.
(17) a Having crossed the border, John waited for the coming of the night.
b. Not having crossed the border, John waited for the coming of the night.
The adjunct in (17 a) is temporal and the adjunct in (17b) IS causal.
(18) a Being a drunk, he fell down the stairs. b. Being drunk, he fell down the stairs.
The sentence (18a) is odd, because the time he fell down the stairs is not
identified with a specific time interval of his drunkeness. 12 In contrast to the
adjunct in (18b) that is temporally bounded, the adjunct in (I8a) does not
serve to restrict the time of the matrix event but serves a reason for it.
The roles of present-participial adjunct range from temporal relations to
speech-act qualifiers such as putting it mildly,frankly speaking (style disjuncts)
or to viewpoint references such as looking at it objectively (viewpoint
subjuncts)Y For differences of usage found in perception verbs, modality,
152
conjunction, and conditionals, Sweetser observes the same conceptual
structure underlying these distinct semantic areas. She proposes three
domains: content domain for the real-world events and entities, epistemic
domain for the speaker's reasoning processes, and speech-act domain for
conversational interaction. For the ambiguity of modals between root and
epistemic senses, she posits the same image-schematic structure for content
and epistemic domains. For example, may is viewed as a potential barrier
which is not actually barring some potential path. This concept applies to
the content domain yielding permission and the epistemic domain yielding
possibility, i.e. there is no barrier to the speaker's reasoning process reaching
a conclusion. Sweets er attributes multiple possibilities for interpreting
conjoined or conditional sentences to pragmatic ambiguity among domains.
Her example shows the same expression can be interpreted differently
depending on a domain:
(l9) She went, because she left her book 1ll the mOVIe theater last night. 14
The because-clause gives a reason for the departure in the content domain,
implying intention to recover the lost book, or a premise for a logical
conclusion in the epistemic domain: I know she went, because I discovered
that she left her book in the movie theater.
The free adjunct construction shares the interpretation with conjunctions
and conditionals. Three domains can be assumed for the adjunct interpreta
tion. In the epistemic domain, the speaker's knowledge state is conjoined to
his/her conclusion.
(20) a Wearing swimming trunks, you aren't going to the restaurant. b. Not being in the office, John has (evidently) gone home.
Sentences in (20) express causal conjunctions in the epistemic domain as
153
can be restated in (21).15
(21) a Since you are wearing swimming trunks, you aren't going to the restaurant, I conclude! guess! suppose.
b. Since John isn't in the office, he has (evidently) gone home.
Compared with the full adverbial clauses, the use of the adjunct in the
epistemic domain is restricted. One of the reasons is that the adjunct
construction keeps the unity of subject between the adjunct and the matrix
clause, leaving less chance to insert the speaker's view.
(22) a If John went to the party, he was trying to infuriate Miriam. b. If they're looking for an apartment, they're planning a wedding
before long.
These epistemic conditionals represent a reasoning process: If I know that X
is true, then I conclude that Y. 16 With the free adjuncts, their semantic relations
are reduced to the real-world relations as expressed in (23):
(23) a Going to the party, John was trying to infuriate Miriam. b .#Looking for an apartment, they're planning a wedding before
long.
The adjunct in (23a) does not represent the premise of the speaker but John's
manner for his plan. Sentence (23b) is unacceptable, because it fails to
express the speaker's reasoning from effect to cause which is encoded in
(22b). Furthermore, the adjunct does not express counterfactual con
ditionality.
(24) *Having visited him in hospital, you would know that he was moved to a different room. 17
Some linguistic cues contribute to expressing the speaker's viewpoint.
One of the cues is the modal verbs in epistemic sense. Another is what Quirk
154
et al. call viewpoint subjuncts. The verbs in viewpoint subjuncts constitute a
restricted semantic set which includes verbs of visual or mental perception
(e.g. consider, look at, see, view), or verbs of inert cognition (e.g. assume,
presume, suppose)18 Diachronically, many of these participles have
developed into conjunctions (considering, seeing, supposing, etc.).
As for the speech-act domain, English has a rather stereotyped expression
related to the act of speaking: putting it mildly, frankly speaking, etc. When an
utterance is interrogative or imperative, the adjunct is conjoined at the
speech-act domain.
(25) a Having done the washing-up, can you do the ironing, toO?19 b. Wearing swimming trunks, you aren't going to the restaurant!
The meaning of (25a) can be restated I ask you to do the ironing since/after you
have done the washing-up. Sentence (25b), which is the same construction we
have examined in the epistemic domain, means that I won't allow you to enter
the restaurant since you wear swimming trunks. These adjuncts are related not to
the content of the matrix clauses but to the act of asking and ordering.
Although Sweetser does not pose an image-schematic mapping in her
analysis of conjunctions and conditionals, some expressions show metaphor
ical mapping among domains. The order of adjuncts often reflects the order
of events. Event sequences are followed easily when the events are arranged
on a time line as well as spatially on a path.
(26) Leaving the base camp at five, we reached the top at noon.
Adjuncts with verbs indicating direction are used as shifting devices III
discourse, making explicit the place for the statement: e.g. moving back to the
second step, returning to the old question, going a step further.
In this section we have seen two processes that characterize the speaker's
inferences. One is concerned with a transition between different semantic
155
areas: from temporal to causal. The other is the application of event relations
in different domains, encoding the speaker's reasoning process or his speech
act.
5. Conclusion
For the various interpretations of free adjuncts, studies based exclusively
on form or context are inadequate. This paper reinforces the view that both
semantics and pragmatics are necessary for the adjunct construction. The
presence of modal verbs, relative frequency adverbs, and generalization
reduces the semantic variability of free adjuncts in two groups. Modal
semantics and predicate distinction determine the adjunct interpretation
from temporal to causal or conditional in these environments. Inferences of
speakers in the adjunct interpretation concern the transition between
semantic areas on one hand, and the three domains in each semantic area on
the other. The three domains by means of which Sweetser analyzes full
adverbial clauses are applicable to the analysis of free adjuncts. Content,
epistemic, and speech-act domains are posed either for temporal, causal, or
conditional relations, so that the domain analysis gives an extensive
explanation for the semantic variability of free adjuncts, including the
speaker's reasoning or his attitude in speaking.
NOTES
1 Gregory T. Stump, The Semantic Variability of Absolute Constructions (Dordrecht:
Reidel, 1985) 4.
2 Stump (1985). Its theoretical framework-is the formal semantics developed by
Richard Montague. The theoretical adequacy is of little concern to the present
paper.
3 Eve E. Sweetser, From Etymology to Pragmatics (Cambridge: Cambridge
University Press, 1990).
156
4 Bernd Kortmann, Free Adjuncts and Absolutes in English (London: Routledge,
1991) 117.
5 Examples (10) and (11) are discussed in Stump (1985).
6 Gregory N. Carlson, Reference to Kinds in English, University of Massachusetts
doctoral dissertation, 1977.
7 The full translation is
would' (D(cbWBill-wears-that-new-outfit')) ("Bill-fools-everyone').
Here, cb stands for the conversational background and D is the operator that
combines the set of proposition representing the conversational background with
the proposition expressed by the adjunct. See Stump 54. The problem of
identifying the underlying subject of adjuncts is neglected in the present
discussion:
8 Renaat Declerck, A Comprehensive Descriptive Grammar of English (Tokyo:
Kaitakusha, 1991) 384.
9 Declerck 409.
10 Jennifer Coates, The Semantics of the Modal Auxiliaries (London: Croom Helm,
1983) 90.
11 Examples (14) and (15) are discussed in Stump (1985). Sentences (14b) and
(15b) are modified from his original examples to give a plausible situation.
12 See Stump 309-310 for the relevant discussion.
13 Style disjuncts and viewpoint subjuncts are terms used in Randolph Quirk et
aI., A Comprehensive Grammar of the English Language (London: Longman, 1985).
14 Example (19) is discussed in Sweetser 77.
15 Sentences (21a) and (21b) are modified examples which are originally used in
Sweetser 80 and 78, respectively.
16 Examples (22a) and (22b) are cited from Sweetser 116 and 128, respectively.
17 Kortmann 231.
18 Kortmann 159.
19 Kortmann 231.