OF CONCEPTUAL STRUCTURE
Kumamoto Women's University
This paper is concerned with the presentation of constraints in
terms of Jackendoff's (1990) conceptual structure to accommodate
the acceptability of English and Japanese gapped sentences. It is
argued that the semantic aspects of Gapping are well captured by
referring to the substitution mech- anism based on the conceptual
structure of the antecedent clause to get the conceptual structure
of the gapped clause. The constraints prescribe the substitution
possibilities and reflect perceptual strategies contributing to the
difficulty of processing gapped sentences. This in turn
accommodates some exceptional cases to the constraints and the
difference between the English and the Japanese sentences.*
1. Introduction
Gapping, which is one of the elliptical phenomena, is closely
related to the understanding of the sentence. In Conceptual
Semantics proposed by Jackendoff (especially 1983, 1990), to
understand a sentence is to make the correspondence to the
conceptual structure of it. In this paper, Gapping
phenomena in English and Japanese are studied in the framework of
Jackendoff (1990) and constraints to capture the acceptability are
pre- sented.
Let us make clear the scope of the discussion by defining what the
Gapping construction is. It is defined as in (1).
*This is a revised version of the paper read at the Ninth National
Conference of the English Linguistic Society of Japan on November
23, 1991. Earlier versions were also read at meetings of the
Fukuoka Linguistic Circle (FLC) and the Kumamoto Linguistic Circle
(KLC). I am grateful to those who participated in the meetings,
especially to Toshiaki Inada, Keiji Konomi, Shoichi Tanaka and
Tatsuhiko Toda for invaluable comments and to two anonymous EL
reviewers for their useful criticisms and sugges- tions on an
earlier draft. I would like to thank Kim Shroeder and Paul Beaufait
for act- ing as informants. I am also indebted to Mary Gawienowski
and Daniel Kirk for kindly correcting stylistic errors of the
paper. All remaining inadequacies, needless to say, are my
own.
English Linguistics 9 (1992) 176-195 (C) 1992 by the English
Linguistic Society of Japan
-176-
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 177
(1) The construction in which the tensed verbs of the coordinate
clauses except one at the first or end are missing, leaving at
least two major phrase categories of the clauses.1,2
Thus the sentences in (2), in which the tensed verbs of the clauses
are not missing, do not involve Gapping though they have medial
missing ele- ments. (2a) is an example of Verb Phrase Ellipsis
(VPE) and (2b) is that of Left Peripheral Reduction (LPR) in which
the verb is not missing. By the same token, the Japanese sentences
in (3) are not examples involving Gapping either, since no verbs
are missing there.
(2) a. John was robbed in Paris, but he wasn't o in London. (Kuno
(1978: 29))
b. On Tuesday Linda washed the car, and on Wednesday o cleaned the
stove. (Imanishi and Asano (1990: 375))
(3) a. John-wa coohi-o Mary-to nomi, Tom -wa o -Top coffee-Acc
-with drink -Top
Susan-to non-da. -with drink-Past
'John drank a coffee with Mary and Tom drank one with Susan.'
b. John-wa Mary-o eiga-ni sasoi, Tom -wa o -Top -Acc movies-to ask
out, -Top
tsuri-ni saso-tta. fishing-to ask out-Past 'John asked Mary out to
the movies and Tom asked her to
go fishing.' Our interest is placed on such sentences as in (4) and
(5). In English Gapping, like those in (4), the right
(non-leftmost) conjunct contains the missing elements (gap), while
in Japanese Gapping, like those in (5), the left (non-rightmost)
conjunct contains them.
1The missing constituents are represented by o in all examples
throughout the paper. 2Some grammarians treat the following as
examples of Gapping:
(i) a. John must clean the shed and Peter o read the book. b. John
will sing and Mary o dance.
c. Harry may leave and Fred o stay. (Koizumi (1987: 243))
However, the acceptability seems to undergo idiolectal variance as
Koizumi (1987: 243)
points out, and Jackendoff s conceptual structure is not made clear
on the treatment of auxiliary verbs. Therefore, we do not take up
these examples in this paper.
178 ENGLISH LINGUISTICS, VOLUME 9 (1992)
(4) a. John ate the apple and Tom o the hamburgers. b. Mary went
into the kitchen and John o into the bathroom.
(5) a. Tom-wa hanbagaa-o o, John-wa ringo-o tabe-ta -Top
hamburger-Acc -Top apple-Acc eat-Past
(same meaning as (4a))3 b. John-wa furoba-e o, Mary-wa
daidokoro-e
-Top bathroom-to -Top kitchen-to
go in-Past (same meaning as (4b))
How are the meanings of Gapping sentences obtained? In other words,
how the correspondence between the conceptual structures and the
syn- tatic structures of them can be obtained? The conceptual
structure of the non-gapped conjunct of a Gapping sentence
corresponds to the syntactic structure by the correspondence rules
which Jackendoff (1990) proposes.4 It can be supposed that based on
the conceptual structure of the non-
gapped conjunct (which we will refer to as the antecedent
conceptual struc- ture no matter what the linear order of the
gapped and non-gapped con-
juncts is) the conceptual structure of the gapped conjunct should
be obtained through the substitution of the readings of the
remnants for the corresponding parts of the antecedent conceptual
structure. The antecedent conceptual structures of (4) and (5)
might be represented as (6). The con- ceptual structures of the
gapped conjuncts are the same ones as (6) except that the readings
of the remnants substitute for the corresponding argu- ments.
(6) a. [CAUSE([JOHN]α, [GO([APPLE],
[TO([IN([MOUTH OF([α])])])])])]
b. [GO([MARY], [TO([IN([KITCHEN])])])] The example in (7) is a case
of LPR and should be kept distinct from
Gapping. However, the verb is missing and the same mechanism to
obtain the conceptual structure of the elliptical conjunct as that
of Gapping seems to work. Moreover, it shares with the Japanese
Gapping construction the
3Throughout the paper the aspect of meaning, such as information
structure, is
ignored when it is stated that the meaning of a Japanese sentence
is the same as that of an English sentence.
4Such rules as Argument Linking, Argument Fusion, Adjunct Rule, and
Restrictive
Modifier Rule are proposed in Jackendoff (1990).
ENGLISH LINGUISTICS, VOLUME 9 (1992) 179
property that the missing parts should be necessarily located at
the end of the clause, not the middle, as is seen in (8).
Therefore, this type of LPR is touched upon in the following
discussion.
(7) John hit Mary with a baseball bat, and o o Bill with a bicycle
chain. (Kuno (1978: 63))
(8) a. Tom-wa Mary-to o o, John-wa Sue-to kohii-o -Top -with -Top
-with coffee-Acc
non-da. drink-Past 'Tom drank a coffee with Mary and John o o with
Sue.'
b. *Tom-wa o Mary-to o, John-wa kohii-o Sue-to non-da. Based on the
supposition that the conceptual structures of the gapped
clauses can be obtained by the substitution mechanism above
mentioned, constraints on the substitution are presented in the
subsequent sections to explain the acceptability of Gapping
sentences.
2. Semantic Parallelism Constraint
The basic constraint which immediately comes to mind is that in
(9).
(9) The remnants of Gapping must represent conceptually compati-
ble elements with the corresponding parts of the antecedent
conceptual structure.
It is a semantic parallelism constraint and naturally proposed
under the supposition that the conceptual structure of a gapped
clause is obtained based on the antecedent conceptual structure.
"Conceptually compatible elements" in (9) would be defined based on
two different notions: (i) the compatability of the conceptual
categories,5 and (ii) the identicalness of the semantic (thematic)
roles, which are regarded as structural positions in conceptual
structure in Jackendoff (1990).
Let us see the effect of (9).6
(10) a. The boy is APvery clever and the girl o NPa genius. b. The
boy walked Advslowly and the girl o ppwith great care.
5[PROPERTY] and [THING TYPE], which appear in (12a), can be
regarded as com-
patible. See Jackendoff (1983: 194-195). As for other categories,
the compatibility is defined by the identicalness.
6In the following, we have italicized relevant words and have added
syntactic or
conceptual categories to the original examples if necessary for the
expository purpose.
180 ENGLISH LINGUISTICS, VOLUME 9 (1992)
(11) *John eats ppat noon and Mary o ppat home. (Matsui (1988: 58,
60))
(12) a. [State BEIdent([Thing], [ATIdent([Property/Thing Type])])]
b. [GO([Thing], [Path])
Event [Manner]]
[pathTO([Plac, IN([Thing MOUTH OF([α])])])])])
Event [Place ATTemp([Time])]]
b. […(same as(a))]
Event [Place AT(Thing])]]
The italicized syntactically different constituents in (10) play
the same role as conceptual constituents to make the same
conceptual structures, which are shown in (12) ((12a) for (10a) and
(12b) for (10b)), satisfying (9) in both notions of conceptual
compatibility. As a result, Gapping is accept- able in (10). In
contrast, the italicized syntactically identical constituents in
(11) are incompatible as conceptual constituents since they contain
dif- ferent conceptual categories which appear in different
semantic fields (ie. have different semantic roles in its strict
sense) as is seen in (13) ((13a) for the left and (13b) for the
right conjunct of (11)). It violates (9) and this is why (11) is
not acceptable. The same kind of examples can be seen in
Japanese.
(14) a. (otona-ni naru-to) Tom-wa APsuratto kakkoyoku o adult-Dat
become-when -Top slender nice-looking John-wa NPzungurisita
himantai-ni natta.
-Top stout overweight body-Dat became
'(When they became adults) Tom became slender and John o a stout
man.'
b.?*Tom-wa ppshoogo-ni o o, Mary -wa ppie-de -Top noon-at -Top
home-at
shokuji-o shita. meal-Acc did '*Tom ate at noon and Mary o at
home.'
(14a) is an example in which (9) is satisfied though the syntactic
categories of the italicized parts are different. On the other
hand, (14b) is an example in which the syntactic categories of the
italicized parts are the same, yet they do not satisfy (9).7
7The unacceptability of (i), which is a case of LPR, is an evidence
to confirm that the
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 181
These are examples which show the priority of the semantic
parallelism constraint over the syntactic parallelism constraint to
determine the accept- ability of Gapping sentences. However, the
problem is not so simple as it initially seems. Let us examine
conceptual compatibility more in depth. We introduced two notions
to define it above. Are both really required to define it and
maintain the priority of (9) over the syntactic parallelism con-
straint? Let us first suppose the matter of conceptual category
alone is enough to define the compatibility of conceptual elements.
If so, however, the examples in (15) would show the priority of the
syntactic parallelism constraint over the semantic one, in which
the conceptual categories of the corresponding parts are different
between the antecedent and the gapped clauses and they still are
acceptable.
(15) a. John hates NPthe war and Bob o NPSaddam Hussein. [EVENT]
[THING]
b.(?)John is proud of Nphis youth and Bob o o o NPhis
children.
[PROPERTY] [THING] Now let us consider the compatibility of
conceptual elements as a matter of the semantic role. The
italicized parts in (16) are the same in the syntactic and
conceptual categories. However, they are different in the semantic
roles, which are shown as the two different italicized parts in
(17), one being Agent, the other Instrument. Thus the semantic role
seems to be the
principal factor in (9) to determine conceptual compatibility. (16)
*NPPhil opened the first door and NPthe key o the second one.
[THING] [THING] (17) [CAUSE([α],[INCH([BE([β], [OPEN])])])
AFF-([PHIL]α, [DOOR]β)
[BY[CAUSE([α], [AFF-([γ], [β])])]
AFF-([α], [KEY]γ)]]
Then, when we consider the semantic parallelism as a matter of the
seman-
tic role alone, is it always possible to state that the semantic
constraint overrides the syntactic constraint? The answer is
negative. Note the
example in (18).
same mechanism to interpret the elliptical clause as the case of
Gapping works in some LPRs and they are also under the constraint
of (9).
(i) *John hit Mary ppwith a baseball bat, and o o Bill ppin the
park.
182 ENGLISH LINGUISTICS, VOLUME 9 (1992)
(18) *Sam hates NPreptiles, and Sandy o S/VPto talk to Oh. [THING]
[EVENT]
(Sag (1977: 141)) The italicized parts have the same semantic
roles, which would be the arguments represented by β ([β] and[Y])
in the conceptual structure (19),
yet (18) is unacceptable.8 (19) [HATE([α], [β])
State REACT-([X]α, [Y]β)]
To get a clear picture of the interaction between the semantic and
syn-
tactic parallelism constraints, observe the eight possible
combinations
below as we consider the semantic parallelism as distinct matters
of the
category and the role.
acceptability OK * OK * OK * * *
example no. (4)*(16)(15)*(11)(10)*(21) *(18)*(22)
(21) a. *John saw NPthe parade and Bob o s'that his boss would
[EVENT] [EVENT]
meet John. b. *John looked APpale and Bob o NPhis sadness.
[PROPERTY] [PROPERTY] (22) *Beth ate NPyogurt, and Norma o peat
midnight.
[THING] [PLACE] (Sag (1977: 141))
In (20) the circle and the cross mark the observance and the
violation of the constraints, respectively. (21) and (22) are
complementary examples to complete the list. From (20) we can draw
the conclusion that the semantic constraint is more significant
than the syntactic constraint to determine the acceptability of
Gapping sentences. This is clearly seen in (10) and (11), though
the determinant of the semantic constraint is different depending
on the observance of the syntactic constraint. If the syntactic
constraint is
8AFF in (17) is a function which shows Actor-Patient (Beneficiary)
relation. REACT in (19) is a function which shows the reaction of
the experiencer to the stimu- lus. They constitute the action tier
in contrast to the thematic tier, which deals with motion and
location. The action tier is shown when necessary.
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 183
observed, the role alone is the determinant of the acceptability,
which is clear from the left half of the list, especially from (15)
and (16). If the syntactic constraint is not observed, the
observance of the role parallelism is not enough to make the
sentence acceptable, which is seen in (18), that is, the stronger
semantic parallelism is required.
In other words, (9) is made specified like (9') as a constraint
which pre- dicts the acceptability of Gapping in accordance with
the observance of the syntactic parallelism constraint.
(9') a. The remnants of Gapping must represent the identical
semantic roles with the corresponding parts of the antecedent
conceptual structure. (weak form)
b. The remnants of Gapping must represent the identical semantic
roles and compatible conceptual categories with the corresponding
parts of the antecedent conceptual struc- true. (strong form)
The existence of the two forms suggests that Gapping is a
semantically
governed perceptual phenomenon in which syntactic category may help
process the sentence.
3. Complete Constituent Constraint
The examples in (23) have been explained syntactically by various
ver- sions of the Major Constituent Condition initially proposed in
Hankamer
(1973), which prescribes that the remnants of Gapping should be the
con- stituents which are either immediately dominated by S or
immediately dominated by VP, which is immediately dominated by
S.
(23) a. *John spoke to Fred, and Mark o o Peter (Chao (1988: 12))
b. *Beth left after the firs act, and Norma o o the second
act.
(Koizumi (1987: 249)) However, they can be more comprehensively
accommodated by a semantic constraint referring to the conceptual
structure:9
(24) The remnants of Gapping must represent the complete argu-
ments or modifiers of the functions represented by the verbs
missing. (Otherwise the substitution of the readings for the cor-
responding elements in the antecedent conceptual structure may not
be allowed.)
9This section is based on Nishioka (1991), a work which limits
consideration to
English data only, but we consider Japanese and English data
here.
184 ENGLISH LINGUISTICS, VOLUME 9 (1992)
The conceptual structures of the gapped clauses in (23) would be
roughly represented as (25).
(25) a. [CAUSE([MARK], [GO([TALK], [TO([PETER])])])] b.
[GO([NORMA], [FROM([ROOM])])
AFTER([SECOND ACT])]
The right remnants of (23) lack the elements which express the
underlined
parts in (25) and cannot be the complete argument or modifier of
the func- tions expressed by the italics, causing the violation of
(24).
(26) a. John put on the hat and Mary o o the scarf. b. Charlie
entered the bedroom and Vera o the kitchen.
(cf. ?? Charlie went into the bedroom and Vera o o the
kitchen.)
(27) a.[CAUSE([MARY]α, [GO([SCARF],
[TO([ON([BODY OF([α])])])])])]
b. [GO([VERA], [TO([IN([KITCHEN])])])] In contrast, the acceptable
examples in (26) satisfy (24) since the remnants of the gapped
clauses represent the complete arguments of the italicized
functions which correspond to the gapped verbs in (27).
Does Japanese have this constraint? The examples in (28) are the
Japa- nese counterparts of those in (23) and seem to have
idiolectal variance in the acceptability.
(28) a. (??) Mark-wa Peter o o, John-wa Fred-ni hanashikaketa. -Top
-Top -to spoke
b. (??) Norma-wa nimaku o o o, Beth-wa -Top second act -Top
ichimaku-no-atoni deteitta. first act-Gen-after left
The speakers who accept (28) are not constrained by (24). Then what
can we say about the speakers who do not accept(28)? My informants
who fall into that category also do not accept the sentences like
(29).
(29) (??) Peter-ni Mark o o, Fred-ni John -ga hanashikaketa. -to
-to -Nom spoke
'To Peter Mark spoke, and to Fred John spoke.' In (29) the
functional Case ga representing Nominative is missing. How- ever,
functional Cases cannot appear in the conceptual structure and the
unacceptability of (29) cannot be accommodated by (24). This leads
us to suppose that some general constraint prohibiting the drop of
Cases works in both (28) and (29) for those speakers and (24) need
not be responsible for the unacceptability of (28). That is to say,
the idiolectal variance of
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 185
(28) as well as (29) could be attributed to the applicability
difference of the general constraint among individuals.10
Therefore, we would think that Japanese Gapping in contrast to
English is not constrained by (24). (24) is a perceptual constraint
which prescribes the preservation of the semantic relation between
the missing verb and the remnants of Gapping. The rea- son why
Japanese does not observe (24) seems to be related to the place of
the gap. Japanese Gapping always has the gap at the (right) end of
the clause. By the same token, the English LPRs, which have the gap
at the
(left) end of the clause, do not observe (24). (30) a. My sister
spoke to Mrs. Wimble on Friday and o o o the
dean on Saturday. (Sag (1977: 150)) b. Beth left after the first
act yesterday, and o o o the second
act today. Now consider the different acceptability between the
sentences in (31)
and (32).
(31) a. *John persuaded Bill to donate $200, and Tom o o to donate
$400. (Kuno (1976: 312))
b. *John shouted at Bill to run, and Tom o o o to stop.
(32) a. John promised Bill to donate $200, and Tom o o to donate
$400. (Kuno (1976: 312))
b. John vowed to Bill to join the race, and Tom o o o to win the
race.
This difference can also be accommodated by (24). The semantic
subject of each complement sentence differs between (31) and (32):
it is the
(prepositional) object (Bill) in (31) while it is the subject
(John/Tom) in (32), which may be represented as the two distinct
types of the conceptual structure.11
(33) a. [CAUSE([α],[AFF([β],)])]
AFF([]α, []β)]
AFF([]α, []β)]
(promise-type)
The italicized second argument of CAUSE i.e. the Effect corresponds
to the complement sentences in both (a) and (b) and the first
argument of the AFF function within the Effect i.e. the (embedded)
Actor is bound by β,
10Exact formulation of the constraint and the reason why it
undergoes the idiolectal variance on its applicability are left for
further study.
11See Nishioka (1991) for the detailed analysis of the difference
of the two types.
186 ENGLISH LINGUISTICS, VOLUME 9 (1992)
which corresponds to the object, in(a)while it is bound by α, which
corre-
sponds to the subject, in (b). (31) causes the violation of (24)
since the object (β) is missing and the Effect is incomplete
argument. On the other
hand, (32) does not involve such violation of (24) even if the
ohlect (β) is
mussing since α is liven.
(34) 50% of his constituents asked the Senator to vote for the
bill, and 25% o o to vote against it. (Kuno (1976: 316))
The Actor of the Effect in (34) is the object (the Senator), so it
is repre- sented as the persuade-type conceptual structure (33a).
However, the sen- tence in (34) is acceptable and it may be pointed
out as the exception of the explanation given above. The reason why
it is acceptable seems to lie in the use of the perceptually
outstanding special morphology (%). By the same reason, the example
in (35), which is similar to (23a), is acceptable though it
violates (24).
(35) Support for Bill increased to 70%, and o for John o o 30%.12
The use of % makes self-evident the correspondence between the con-
stituents in the antecedent and the gapped clauses and also makes
easier the substitution of the conceptual constituents proposed
here.13
As we have argued above, a wider range of lingustic facts
concerning Gapping can be accounted for by (24) than by any
syntactic approach.
4. Consecutive Arguments Constraint
Gapping sentences which have the object missing are unacceptable in
some cases:
(36) a. *Max gave Sally a nickel, and Harvey o o a dime. b. *Jack
calls Joe Mike and Sam o o Harry.
(Hankamer (1973: 26, 31)) c. *Mary considers John happy and Bill o
o unhappy. d. *Mary believes John a genius and Tom o o an
idiot.
(Matsui (1988: 62)) (37) a. * Max put the eggplant on the table and
Harvey o o in the
sink.
12The first missing element for support is not generated by Gapping
but by LPR. 13The special noun (the one with %) itself constitutes
an incomplete argument in
(35), while the other remnant (complement sentence) is incomplete
in (34). It seems that because of the self-evident correspondence
between % nouns, (34) has the same per- ceptual effect as LPR
(30).
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 187
b. * Max gave a nickel to Sally and Harvey o o to Suzan.
(Terazu (1975: 46)) However, not in others.14
(38) a. My father visited my uncle last night and my sister o o
this morning. (Matsui (1988: 64))
b. My brother visited Japan in 1960, and my sister o o in
1961.
(Kuno (1976: 306)) c. Mary saw her sister in Kyoto, and Sue o o in
Osaka.
The difference between the unacceptable Gapping in (36) and (37)
and the acceptable one in (38) can be found in the conceptual
status of the rem- nants of Gapping. In (36) and (37) the remnants
all correspond to the argu- ments of the functions represented by
the missing verbs whereas in (38) one remnant corresponds to the
modifier of the verb in each sentence.15
Jackendoff (1990: 258) proposes the thematic hierarchy such as
(39)
(here the relevant constituents are marked with*, an optional
element is enclosed by angle brackets):
(39) a. [AFF(X*, <Y>)] (Actor) b. [AFF(<X>, Y*)]
(Patient (AFF-) or Beneficiary (AFF+)) c. [Event/State F(X*,
<Y>)](Theme) F≠CS
d. [Path/PlaceF(X*)] (Location, Source, Goal)
The conceptual arguments are ordered according to this hierarchy
and
when the readings of the remnants of Gapping substitute for the
antecedent conceptual constituents to obtain the conceptual
structure of the gapped
clause, the following perceptual constraint seems to work.
14The inadequacy of Hankamer's (1973) No-Ambiguity Condition (NAC),
which
prohibits Gapping when the interpretation of the resulting output
structure as LPR is possible, is clear from the examples in (38) in
which Gapping is acceptable though they can be structurally
interpreted as LPR.
(i) a. My father visited my uncle last night and o o my sister this
morning. b. My brother visited Japan in 1960, and o o my sister in
1961.
(Kuno (1976: 306)) c. Mary saw her sister in Kyoto, and o o Sue in
Osaka.
15The conceptual structure of a modifier is fused into the one of
the modifiee by Restrictive Modifier Rule proposed in Jackendoff
(1990: 56).
(i) Restrictive Modifier Rule If YP is daughter of X" in XP, and
the conceptual structure of YP is [Cy], the conceptual structure of
XP is of the form […
[CY]].
188 ENGLISH LINGUISTICS, VOLUME 9 (1992)
(40) The remnants of Gapping may not correspond to two non-
consecutive arguments of the functions represented by the missing
verb.
See, for example, the antecedent conceptual structures of (36a, c)
and (37a), to which the order of the arguments is given in
subscript according to (39).16
(41) a. [CAUSE([α], [GOPoss(CNICKEL]3, [TO([β])])])
AFF+([MAX)α1, [SALLY]β2)]
REACT(IMARY]α1, [JOHN]β2)]
AFF([MAX]α1, [EGGPLANT]β2)]
The remnants of Gapping in (36) and (37) must correspond to the
argu- ments given the numbers of 1 and 3 when the substitution
occurs in order to get the intended meanings, causing the violation
of (40). In contrast, the examples in (38) involve the substitution
of only one argument of the func- tions represented by the verb,
causing no violation of (40). See, for exam-
ple, the antecedent conceptual structure of (38a). (42)[GO([α],
[TO([β])])
AFF ([MY FATHER]α1, [MY UNCLE]β2)
[LAST NIGHT]]
(40) does not prohibit the examples such as the following. (43) a.
Arizona elected Goldwater Senator, and Pensylvania o
Schweiker o. (Jackendoff (1971: 24)) b. Mary believes John a genius
and Tom o Bob o.
(44) a. Max gave Sally a nickel, and o o Harvey a dime. b. Jack
calls Joe Mike and o o Sam Harry.
(Hankamer (1973: 26, 31)) c. Mary considers John happy and o o Bill
unhappy. d. Mary believes John a genius and o o Tom an idiot.
(Matsui (1988: 62)) The remnants correspond to the arguments given
the numbers of 1 and 2 in
(43) and to the arguments given the numbers of 2 and 3 in (44),
which are examples of LPR.
(40) is another perceptual constraint which prescribes the possible
substi-
16REACT in (41b) is in a mirror relation with AFF. Therefore, it
occu pies the same
position as AFF in (39).
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 189
tution. Therefore, when the correspondence of the substitution is
marked morphologically or contextually (pragmatically), (40) may be
overridden.17
(45) a. Drinking gives you cirrhosis, and smoking o o cancer.
(Oirsouw (1987: 40))
b. The Jones christened their first child Marjorie Victoria and the
Greens, o o Elizabeth Elnora. (Jake (1977: 169))
c. After her parents learned that Sally wanted to go to college,
Mr. Jones gave her a savings bond and Mrs. Jones, o o a set of
encyclopedias. (Jake (1977: 165))
Japanese is not usually constrained by (40) since the
correspondence is marked morphologically by the Case-markers.
(46) a. Max-wa 5 cent-o o o, Harvey-wa 10 cent-o -Top -Acc -Top
-Acc
Sally-ni yatta. -Dat gave
b. Jack-wa Mike-to o o, Sam-wa Harry-to Joe-o yobu. -Top -Com -Top
-Com -Acc call
c. Max-wa teburu-no-ueni o o, Harvey-wa nagasi-no- -Top
table-Gen-on -Top sink-Gen-
nakani nasubi-o oita. in eggplant-Acc put
The sentences in (46) have the same meanings as (36a,b) and (37a),
respec- tively, i.e. the same conceptual structures. However, they
are acceptable unlike the English counterparts.
The following English examples also do not undergo the constraint
of
(40) but in a different way. (47) a. Mary wrote a letter to her
parents and Jack o o to his sister.
b. Tom found happiness in money and Jack o o in knowledge.
(Terazu (1975: 48)) The antecedent con ceptual structures of (47)
are as follows:
(48) a. [CAUSE([α], [GO([LETTER]2, [TO([PARENTS])]3)])
AFF([MARY]α1,)]
AFF([TOM]α1,)]
17The use of pronouns as the indirect objects in (45a, c) is also a
perceptual factor to
allow the Gapping.
190 ENGLISH LINGUISTICS, VOLUME 9 (1992)
The readings of the remnants of Gapping may substitute for
arguments 1 and 3 since other possibilities of the substitution are
excluded semantically, that is, the substitution is semantically
specified. The substitution of the reading of the right remnant and
arument 3 is specified by the preposition and the substitution of
the reading of the left remnant and argument 2 causes semantic
anomalies in each sentence.
The lexical conceptual structure of write in the sense of (47a)
would be the following.
(49) [CAUSE([α), [GO([WRITTEN PRODUCT], [TO([])])])
AFF([]α,)]
That is, it imposes the selectional restriction on the Theme
argument that it should be a written product which can be the
object of transportation, of which the prototypical one is a
letter. This is why pronoun it in (50a) is interpreted as something
like a letter and the implicit Theme of (50b) is a letter.
(50) a. Mary wrote it to her parents. b. Mary wrote to her
parents.
Therefore, the left remnant of Gapping in (47a), JACK, cannot
substitute for Argument 2 by the selectional restriction.
Argument 2 in (48b) is PROPERTY and the left remnant in (47b),
JACK, which is THING, cannot substitute for it. This is because
PROPERTY can exist in another PROPERTY (KNOWLEDGE) but THING cannot
exist in PROPERTY.
When the substitution is specified in some manner, perceptual
constraint
(40) need not be observed.
5. Two-Argument-Limit Constraint
The final constraint presented in the paper is concerned with the
follow- ing examples.
(51) a. *Alan gave Sandy a book, and Peter o Betsy a
magazine.
(Neijt (1979: 90)) b. *Jack calls Joe Mike and Betty o Sam
Harry.
(Matsui (1988: 57)) (52) a.?*Willy put the flowers in a vase, and
Charlie o the book on
the table. (Jackendoff (1971: 26)) b.??Max gave a nickel to Sally
and Tom o a dime to
Harry. (Terazu (1975: 40))
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 191
(53) a. Mary played the piano on the stage on Monday and Betty o
the violin in the pit on Wednesday. (Matsui (1988: 70))
b. The French drink wine at 6:00 and the Germans o beer at 8:00.
(Goodall (1987: 84))
c. Peter talked to his boss on Tuesday, and Betsy o to her
supervisor on Wednesday. (Neijt (1979: 90))
The unacceptable sentences in (51) and (52) involve the
substitution of three arguments of the functions represented by the
verb, while the accept- able sentences in (53) involve the
substitution of only two arguments and modifiers. The substitution
of more than two arguments seems to cause the difficulty of
processing the sentence. We may as well propose the follow- ing
constraint:
(54) The substitution for the antecedent conceptual constituents
may not involve more than two arguments of the functions repre-
sented by the missing verb.
However, Japanese is not constrained also by (54).18
(55) a. Alan-wa Sandy-ni hon-o o, Peter-wa Betty-ni -Top -Dat
book-Acc -Top -Dat
zasshi-o yatta. magazine-Acc gave
(same meaning as (51a)) b. Jack -wa Joe-o Mike-to o, Betty-wa Sam-o
Harry -to
-Top -Acc -Com -Top -Acc -Com
yobu. call
(same meaning as (51b))
18 The Japanese sentence corresponding to (52a) may not be
accepted. (i) ??Willy -wa hana-o kabin-ni o, Charlie-wa hon-o
-Top flower-Acc vase-Loc -Top book-Acc teebru-ni oita. table-Loc
put
This is not because Japanese observes (54) but because Japanese
uses different verbs for put depending on the co-occurring words of
location. Thus the same meaning as (52a) should be represented as
(ii).
(ii) Willy-wa hana-o kabin-ni ire/sashi, Chalie-wa hon-o -Top
book-Acc vase-Loc put in/stick in -Top book-Acc
teeburu-ni oita. table-Loc put
192 ENGLISH LINGUISTICS, VOLUME 9 (1992)
This also seems to be due to the use of the Case markers, which
makes the substitution in the conceptual constituents easier by
showing the clear cor- respondence between the antecedent and the
gapped clauses.
Note that a slight difference in acceptability can be found in
English examples between (51) and (52). This is due to the fact
that one of the remnants is PP in each sentence of (52).
Prepostions seem to fulfill the same role as the Japanese
Case-markers do and make the substitution easier. Moreover, the
following examples, in which two of the three remnants are PPs, are
perfectly acceptable.
(56) a. Mary talked to Harry about the movie and Nancy o to John
about the play.
b. I heard from Mary about Paul's failure and Tom o from Nancy
about Jack's success.
(Terazu (1975: 39)) In this case, the substitution is fully
specified as the Japanese is because if the substitution of the two
remnants are specified, that of the other one is necessarily
specified. That is to say, (54) should be taken as a perceptual
constraint working when the substitution is not fully specified as
well as constraint (40) proposed in the previous section.
6. Concluding Remarks
We have argued that Gapping is an elliptical phenomenon which is
necessarily connected with the understanding of the sentence and
its acceptability might be well captured by constraints on the
conceptual structure. Under the supposition that the conceptual
structure of a gapped clause is obtained through the substitution
of the readings of the remnants for the conceptual constituents in
the antecedent conceptual structure, we have proposed four
constraints. They are (i) Semantic Parallelism Con- straint, which
is more significant than the syntactic parallelism constraint,
though the interaction of the two constraints can be seen in the
strength of the semantic parallelism; (ii) Complete Constituent
Constraint, which pre- scribes the semantic relation between the
remnants and the missing verb of Gapping to interpret the sentence
adequately and easily; (iii) Consecutive Arguments Constraint,
which prescribes the substitution possibilities un- less other
possibilities are somehow specified; and (iv) Two-Argument- Limit
Constraint, which prescribes the number and the kind of the rem-
nants. These constraints play different roles in defining the
possible sub-
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 193
stitution, though it might be said that (i) is the most fundamental
in that it states the premise for performing the substitution. It
might also be said that (ii) is more fundamental than (iii) and
(iv) in that (ii) states the unit of the substitution while (iii)
and (iv) state the substitution pattern and num- ber of the unit.19
These constraints reflect perceptual strategies, and they are
concerned with the difficulty of processing gapped sentences.
There- fore syntactic parallelism affects the strength of (i) and
perceptual marked- ness, or some way to ease processing them such
as the specification of the substitution, can naturally cause the
counterexamples to (ii)-(iv). Japanese, which does not undergo
(ii)-(iv), exemplifies such cases.
We believe some advancement toward the formalized semantic
treatment of Gapping encompassing the perceptual aspects has been
attained through this paper.
REFERENCES
Goodall, Grant (1987) Parallel Structures in Syntax, Cambridge
University Press, Cambridge.
Hankamer, Jorge (1973)“Unacceptable Ambiguity,”Linguistic lnguiry,
5, 17-68.
Hudson, Richard (1982)“Incomplete Conjuncts,”Linguistic INguiry 13,
547-550.
Hudson, Richard (1989)“Gapping and Grammatical Relation,”Journal
of
Linguistics 25, 57-94.
Imanishi, Noriko and Ichiro Asano (1990) Shoooo to Sakujyo
(Anaphora and
Deletion), Taishukan, Tokyo.
Inada, Toshiaki (1991)“Ray Jackendoff: Leading Figures in Current
Linguistics 2,”
Gengo 20.2, 82-89.
Inoue, Kazuko (1978) Nihongo no Bunpoo Kisoku (Grammatical Rules
in
19We have not found examples which show the interaction of these
constraints
except for those in (47), which suggest the interaction between (i)
and (iii). It could be
pointed out that the examples in (32) are evidence of the
interaction between (ii) and (iii). Note, however, that the Effect
represented by the complement sentences in (32) is not ordered by
the hierarchy (39) and thus the examples in (32) are not
constrained by
(iii) under the present analysis. If (iii) is extended to cover the
Effect and prohibits the substitution possibility in (32), the
acceptability would reflect the fundamentalness of (ii) over
(iii).
194 ENGLISH LINGUISTICS, VOLUME 9 (1992)
Japanese), Taishukan, Tokyo. Jackendoff, Ray (1971)“Gapping and
Related Rules,”Linguistic Inquiry 2,
21-35.
Jackendoff, Ray (1983) Semantics and Cognition, MIT Press,
Cambridge, MA.
Jackendoff, Ray (1987)“The Status of Thematic Relations in
Linguistic Theory,”
Linguistic Inquiry 18, 369-411. Jackendoff, Ray (1990) Semantic
Structures, MIT Press, Cambridge, MA.
Jake, Janice (1977)“Gapping, Pragmatics, and Factivity,”CLS 13,
165-172.
Koizumi, Naoshi (1987)“Gapping and Peripheral Deletions,”English
Linguistics
4, 238-252.
Koizumi, Tamotsu (1990) Gengai no Gengogaku: Nihongo Goyooron
(Implica- tional Linguistics: Japanese Pragmatics), Sanseido,
Tokyo.
Kuno, Susumu (1976)“Gapping: A Functional Analysis,”Linguistic
Inquiry 7,
300-318.
Kuno, Susumu (1978) Danwa no Bunpoo (Grammar of Discourse),
Taishukan, Tokyo.
Makino, Seiichi (1980) Kurikaeshi no Bunpoo (Grammar of
Repetition),
Taishukan, Tokyo. Matsui, Chie (1988)“Constraints on
Gapping,”Studies in English Literature 65,
57-74.
Gengogaku Zyoohoo (Current Trends in Overseas Linguistics) 6,
137-152, ed.
by Masaru Kajita and Felix Lobo, Taishukan, Tokyo. Neijt, Anneke
(1979) Gapping: A Contribution to Sentence Grammar, Foris,
Dordrecht.
Nishioka, Nobuaki (1991)“Constraints on Gapping: An Approach from
the
Perspective of Conceptual Structure,”Kyushu Studies in English
Literature 8,
47-73, Kyushu Branch of the English Literary Society of
Japan.
Oirsouw, Rovert van (1987) The Syntax of Coordination, Croom Helm,
London.
Rooryck, Johan (1985)“Gapping-Zeugma in French and English: ANon
Deletion
Analysis,”Linguistic Analysis 15, 187-229.
Sag, Ivan A. (1977) Deletion and Logical Form, Indiana University
Linguistics
Club, Bloomington. [Doctoral dissertation, MIT, 1976]
Terazu, Noriko (1975)“Coordinate Deletion, Gapping, and Right Node
Raising,”
Studies in English Linguistics 3, 19-65.
ON GAPPING: AN ANALYSIS IN TERMS OF CONCEPTUAL STRUCTURE 195
Department of English
Kumamoto Women's University
2432-1 Mizuarai, Kengun-machi