Gapping: A Functional Analysis Author(s): Susumu Kuno Reviewed work(s): Source: Linguistic Inquiry, Vol. 7, No. 2 (Spring, 1976), pp. 300-318 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/4177923 . Accessed: 17/01/2012 22:14 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry. http://www.jstor.org REMARKS 300 AND REPLIES Gapping: A Functional Analysis* SusumuKuno i. Gapping is a rule that converts structuresunderlying such sentences as (ia), (2a), and (3a) into those underlying (ib), (2b), and (3b), respectively.' a. b. (2) a. b. (3) a. (i) I ate fish, Bill ate rice, and Harry ate roast beef. I ate fish, Bill rice, and Harry roast beef. Tom has a pistol, and Dick has a sword. Tom has a pistol, and Dick a sword. I want to try to begin to write a novel, and Mary wants to try to begin to write a play. b. I want to try to begin to write a novel, and Mary to try to begin to write" to begin to write Ca play F to writeI Several attempts have been made to arrive at the conditions for the applicability or nonapplicability of Gapping, and to find explanationsfor the conditions. An example of an ungrammatical sentence produced by an illegal application of Gapping is (4b), which is due to Ross. I will follow Hankamer's (I973) practice of indicating a permitted application of Gapping by enclosing the deleted string in [], and a prohibited application of the rule by enclosing the deleted string in *[ ]. (4) a. I want Bob to shave himself and V [I want] Tom to wash himself.2 b. I want Bob to shave himself and Mary * [wants Bob] to wash himself. Jackendoff (I 97 I) contraststhe ungrammaticalityof (4b) with the grammaticality of (5): (5) a. John writes poetry in the garden, and Max V[writes poetry] in the bathroom. * Research represented in this article has been supported in part by the National Science Foundation's grants to Harvard University (GS-33263X and SOC-74I 2366). I received many valuable comments from the participants in my course "Functionalism in Grammar" at the I974 Summer Institute of the Linguistic Society of America, University of Massachusetts, Amherst, Massachusetts, at which I discussed some of the hypotheses that I present in this article. I am indebted to Stephen Anderson,Jorge Hankamer,Judith Aissen, Sandra Chung, Linda Shumaker, Sidney Greenbaum, D. T. Langendoen, and many others for their comments on earlier versions of the article. 1 These examples are from Ross (I 970), which was the first attempt to analyze the process in the framework of the generative theory of grammar. 2 There is some likelihood that the deletion process involved in the derivation of (4a) and many other sentences to be discussed in this article is not Gapping, but Conjunction Reduction of repeated leftmost strings. (See Stillings (I 975) for this position.) However, since Langendoen and Hankamer, whose hypotheses on Gapping this article is primarily concerned with, have treated these sentences as having been derived by the same process as (I)-(3), I will follow their practice and refer to these examples as gapped sentences. REMARKS AND REPLIES b. Paul Schachter has informed me that the basic order in Tagalog and related languages is VOS; Ives Goddard , [has informed me] that the unmarked order in Algonkian is OVS; and Guy Carden [has informed me] that the basic order in Aleut is OSV. (Ross (I970)) c. Charlie went into the bedroom at 5:30, and Lola ,[went into the bedroom] at 5:3I. He attributes the ungrammaticality of (4b) to the fact that the VP of the right conjunct contains an NP VP sequence (i.e. Bob and to washhimself),and he hypothesizes the following constraint: (6) Jackendoff's Condition on Gapping When the sequence NP VP appears in the VP of the right conjunct, Gapping cannot delete NP, leaving VP behind. This constraint correctly predicts the ungrammaticality of (7a) and (8a) with the interpretationsspecified in (7c) and (8c): (7) a. Bill is depending on Harry to find the way to the party, and Sue to find the way home. (Jackendoff (I97I)) b. __and V [Bill is depending on] Sue to find the way home.3 c. __and Sue *[is depending on Harry] to find the way home. (8) a. Max seemed to be trying to force Ted to leave the room, and Walt to stay a little longer. (Jackendoff (I97I)) b. and V [Max seemed to be trying to force] Walt to stay a little longer. c. and Walt * [seemed to be trying to force Ted] to stay a little longer.4 3 Some speakers do not allow deletion of prepositions in the course of gapping that would leave their objects behind. These speakers find (7b) ungrammatical. Similarly, observe the following example: (i) a. John studied with Mary, and Tom V [studied] with Jane. b. John studied with Mary, and Tom *[studied with] Jane. The same constraint applies to other deletion processes: (ii) a. The proposal appealed to Mary more than to Jane. b. *The proposal appealed to Mary more than Jane. It seems that the above constraint exists in order to facilitate the recovery of the syntactic and semantic functions played by the constituents left behind. 4 Jackendoff also notes that gapped elements may not necessarily be continuous: (i) a. Max seemed to be trying to force Ted to leave the room, and Walt (,) Ira. b. __ and Walt V [seemed to be trying to force] Ira [to leave the room]. He correctly observes that Gapping, thus, cannot be simply formulated as deletion of a variable. He suggests that the rule requires an unlike constituent somewhere in the VP. However, this suggestion is inadequate, as it would rule out sentences such as (iib): (ii) a. Yesterday, Tom came to see me, and today, Mary came to see me. b. Yesterday, Tom came to see me, and today, Mary V [came to see me]. (Gapping) c. Yesterday Tom 0, and today Mary, came to see me. (Right-Node Raising) Note that the VP of the second conjunct is identical to the VP of the first conjunct. It seems that the rule requires two unlike constituents somewhere in the sentence, not necessarily in the VP. REMARKS 302 AND REPLIES However, Hankamer (I973) observes thatJackendoff's constraint cannot prevent Gapping from applying in (9c), (i oc), and ( IIc) to yield (9a), (I oa), and ( IIa). (g) a. Max gave Sally a nickel, and Harvey a dime. and ' [Max gave] Harvey a dime. b. and Harvey * [gave Sally] a dime. c. (io) a. *Max gave Sally a nickel, and Harvey Susan. b. * and ; [Max gave] Harvey Susan. (* for semantic anomaly) c. and Harvey *[gave] Susan [a nickel]. (i I) a. Max wanted Ted to persuade Alex to get lost, and Walt, Ira. b. and ' [Max wanted] Walt [to persuade] Ira [to get lost]. c. and Walt *[wanted] Ira [to persuade Alex to get lost]. and Walt *[wanted Ted to persuade] Ira [to get lost]. Hankamer argues that, in addition to Jackendoff's constraint given in (6), one would need the following constraint in order to block Gapping from applying to (9c) and (ioc): If the VP of the right conjunct is V NP NP, neither of the two NPs can be deleted by Gapping. (I 2) One would also need some totally new condition in order to block Gapping in (iI Ic). Hankamer correctly observes that if one proceeded in this fashion to add a new condition on Gapping for every new restriction one finds, one would end up with a mysterious collection of independent conditions, and that it would be tantamount to claiming that there is no connection among the fact that Gapping is blocked in sentences like (4b), (7c), and (8c), the fact that it is blocked in sentences like (9c) and (ioc), and the fact that it is blocked in sentences like ( I c). On the other hand, he argues, if transderivational constraints are permitted in linguistic theory, one could eliminate this collection of independent conditions in favor of a single, unified generalization: (I 3) Hankamer'sNo-AmbiguityCondition(NAC) Any application of Gapping is disallowed if the resulting output structure is identical to a structure derivable by Gapping from another source that has the "gap" at the left extremity.5 Incidentally, (iic) is the result of applying Right-Node Raising to the structure underlying (iia) and extracting the rightmost common constituent camefrom each conjunct. It represents the pattern that Ross would have attributed to "Backward Gapping", and is contradictory to his assertion that English, as an SVO language, displays only the Forward-Gapping Pattern. See Maling (I972) for the Right-Node Raising analysis of what Ross called Backward Gapping. 5 Hankamer relates his NAC to a more general principle that he refers to as the Structural Recoverability Hypothesis: (i) The StructuralRecoverability Hypothesis Deletion rules involving variables are universally subject to a transderivational condition which prevents them from applying in such a way as to introduce structural ambiguity. REMARKS AND REPLIES According to this hypothesis, (9c) and (i oc) are ungrammatical because, corresponding to each of these sentences, there is a derivation of the same output pattern (i.e. (9b) and (iob), respectively) that has the "gap" at the left extremity. The ungrammaticality of (i ic) is also due to the presence of (i Ib), which has the "gap" at the left-peripheral position. The NAC also predicts correctly that the (a) sentences below do not have the readings corresponding to (c) :6 4) a. Jack calls Joe Mike and Sam Harry. and I [Jack calls] Sam Harry. b. c. and Sam * [calls Joe] Harry. (I 5) a. Jack told Harry that Bill was a fairy, and Alex that Tom had warts. b. and I [Jack told] Alex that Tom had warts. c. and Alex *[told Harry] that Tom had warts. (i 6) a. Max wanted to put the eggplant on the table, and Harvey in the sink. b. and I [Max wanted to put] Harvey in the sink. c. and Harvey *[wanted to put the eggplant] in the sink. (I7) a. The court declared Edward insane, and his mother morally bankrupt. b. and I [the court declared] his mother morally bankrupt. and his mother * [declared Edward] morally bankrupt. c. (I - - Hankamer emphasizes the fact that it is identity of structure that is specified in the NAC. Gapping is blocked in (i 8c) even though there is no grammatical sentence with a left-peripheral gap having the same sequence of morphemes: (i8) a. *Jack asked Mike to wash himself, and Sue to shave himself. b. and , [Jack asked] Sue to shave himself. c. and Sue * [asked Mike] to shave himself. - According to Hankamer, the NAC blocks Gapping in cases where there could be structural ambiguity as to the location of the "gap". Other potential disambiguating factors, such as the reflexive pronoun in (i8) or the selectional anomaly in (i9), are irrelevant: (I9) a. *Max wanted the matrix inverted, and Harvey diagonalized. b. * and 2 [Max wanted] Harvey diagonalized. c. and Harvey * [wanted the matrix] diagonalized. Hankamer considers a few classes of apparent exceptions to the NAC. One class consists of sentences such as (20): This is an extremely interesting hypothesis, although I think it is too strong as it is stated. I believe that the tendency to avoid structural ambiguity, which Hankamer has attempted to capture in his hypothesis, does exist, and that one can find many instances of this tendency in operation in various languages. However, I do not think that Gapping observes this constraint, as I will show in the rest of this article. 6 Grammaticality markings are Hankamer's. There are speakers whose idiolects include the (c) interpretations for these sentences. The analysis of Gapping that I present in this article accounts for this phenomenon. 304 REMARKS (20) AND REPLIES ?Max writes plays in the bedroom, and Harvey in the basement.7 He notes that there seems to be a difference between cases like (I6), where the verb is strictly subcategorized for a PP, and cases like (20), where the PP is optional. He further notes that the grammaticality (or near grammaticality) of (2o) depends also on the fact that plays in writesplays is generic. (2 I), in which a play is specific, is somewhat worse than (20); and (22), in which theplay is definite, is even worse: (2I) (22) ??Max is writing a play in the bedroom, and Harvey in the basement. ?? ?Max is memorizing the play in the bedroom, and Harvey in the basement. The second class of exceptions to the NAC consists of sentences like (5b), in which a pronoun contiguous to the verb does not seem to count as an NP as far as the NAC is concerned. Jackendoff contrasts (5b) with (23) and hypothesizes that there is a cliticization rule in English whereby pronouns in immediate postverbal position become cliticized to the verb. (23) Paul Schachter has informed Haj Ross that the basic order in Tagalog and related languages is VOS; Ives Goddard *[has informed Haj Ross] that the unmarked order in Algonkian is OVS; and Guy Carden * [has informed Haj Ross] that the basic order in Aleut is OSV. Thus, he assumes that the object pronoun me in (5b) is not to be analyzed as an NP for purposes of limiting the applicability of Gapping. Langendoen (I974) questions Hankamer's characterization of the NAC as a transderivational constraint. What is at issue is the ungrammaticality of (i 8c). Langendoen argues that since there is no derivation of (i8b), due to the semantic anomaly of its underlying structure, the NAC that blocks (i8c) cannot be transderivational. Second, he examines sentences such as (24a) and (25a): (24) (25) a. b. c. Max sent Sally the messenger last week, and Susan yesterday. and , [Max sent Sally] Susan yesterday. and *[Max sent] Susan [the messenger] yesterday. _ a. b. c. d. e. f. g. Max wanted Ted to persuade Alex to see Mary, and Walt, Ira. and Walt *[wanted] Ira [to persuade Alex to see Mary]. and Walt *[wanted Ted to persuade] Ira [to see Mary]. and Walt * [wanted Ted to persuade Alex to see] Ira. and *[Max wanted] Walt [to persuade] Ira [to see Mary]. and *[Max wanted] Walt [to persuade Alex to see] Ira. and I [Max wanted Ted to persuade] Walt [to see] Ira. - 7 The grammaticality judgment is Hankamer's. Many speakers accept (20) as perfectly grammatical. Note that Jackendoff regards (5c), which has exactly the same structure as (20), as grammatical. There are also are equally acceptable. speakers to whom (20)-(22) REMARKS AND REPLIES In both (24b) and (24c), the "gap" is at the left extremity. However, only the former sentence is grammatical. Similarly, in (25e,f,g), the gap appears at the left extremity, but only (25g) is grammatical. Langendoen notes that what distinguishes (24c) from (24b), and (25e,f) from (25g), is that the former contain NPs (i.e. the messengerof (24c), Mary of (25e), and Alex of (25f)) in the nonleft-peripheral material to be deleted, and the latter do not.8 On the basis of this observation, Langendoen proposes the following hypothesis: Langendoen'sNonleft-PeripheralNP Constraint(NLPNPC) Gapping cannot apply so as to delete nonleft-peripheral contain NP other than a clitic pronoun adjoined to its V. (26) strings that The above constraint is a syntactic one that can be stated in the standard theory of transformational grammar, which does not have the power of transderivational constraints. In the previous section, I have outlined past results in describing conditions on the English rule of Gapping. In this section, I will show that Gapping is a much more complex process than is implied by Hankamer's NAC and Langendoen's NLPNPC, 2. 8 There is clearly a way to state Hankamer's peripheral gap principle in such a way that it will prefer the gap in (24b) and (25g) as more left-peripheral than the gap in (24c) and (25e,f), respectively. What is more serious about Hankamer's principle is its claim that Gapping is blocked if the resulting output structure is identical to a structure that has the gap at the left extremity evenif the latteris ungrammatical. Let us examine what this claim entails in connection with (i8): (i8) a. *Jack asked Mike to wash himself, and Sue to shave himself. b. * c. and V [Jack asked] Sue to shave himself. and Sue *[asked Mike] to shave himself. According to Hankamer, (i8c) is ungrammatical because its structure is identical to that of (i8b), which has the gap (i.e. jack asked)at the left extremity in spite of the fact that it is ungrammatical due to the disagreement of gender of the reflexive. (i8b) is presumably "derived" from the deep structure shown in (i) via erroneous application of Reflexivization (or rather, the spelling rule for reflexives): (i) and Jack asked Sue [Sue shave Sue]s Now, once the principle allows the output of a gapped sentence to be transderivationally compared with another string that involves violation of a rule application (of the sort illustrated in (i8) and (i)), it produces an unmanageable situation. For example, observe the following sentence: (ii) Mary expects to win, and John V[expects] to lose. It becomes no longer possible to claim that (ii) is grammatical because there is no other derivation of the same structure that has the gap at the left extremity. Note the following derivation: (iii) a. b. c. Mary expects [Mary win]s and Mary expects [John lose]s. Mary expects to win and Mary expects John to lose. Mary expects to win, and *[Mary expects] John to lose. (iiic) is an ungrammatical sentence because Gapping has applied to nonparallel coordinate structures. But since Hankamer's principle allows (ii) to be checked against derivations of ungrammatical structures, it will decide that (ii) is ungrammatical because (iiic), which has the same output structure, has the gap at the left extremity. Unless Hankamer can propose a principled way to distinguish between cases such as (i8)-(I9) and ones like (iii), it seems that the NAC cannot be maintained. 306 REMARKS AND REPLIES and that it is conditioned by the interaction of syntax with at least three kinds of nonsyntactic factors: (a) perceptual, (b) discourse-based, and (c) semantic. Undoubtedly, there is a strong tendency for coupling the two constituents left behind by Gapping with the rightmost constituents of the first conjunct. It seems that this tendency is what Hankamer and Langendoen have tried to capture in their respective hypotheses. This tendency is probably due to the fact that the constituents in the first conjunct that were processed last of all (that is, the ones that appeared closest to the end of the conjunct) are easiest to recall for the purpose of coupling with the constituents left behind by Gapping. Following Langendoen and Grosu,9 I will refer to this tendency as the Minimal Distance Principle. I will restate it below: (27) The Minimal Distance Principle The two constituents left behind by Gapping can be most readily coupled with the constituents (of the same structures) in the first conjunct that were processed last of all. 2.I. In spite of the plausability of the Minimal Distance Principle, which is a perception-oriented restatement of Hankamer's NAC and Langendoen's NLPNPC, there are many instances in which this principle is not observed. First, note the following sentences. (28) a. My brother visited Japan in I960, and my sister - [visited Japan] in I96I.10 b. Harry told this story to his mother, and Tom I [told this story] to his father. There does not seem to be anything wrong with (28a). It is a clear counterexample to Langendoen's NLPNPC because the gapped string visitedJapan is a nonleft-peripheral string that contains an NP (i.e. Japan). It is also a counterexample to Hankamer's analysis because there is a derivation of the same sequence of morphemes with the ''gap" at the left extremity. (29) My brother visited Japan in I960, and ,'[my brother visited] my sister in I96I. Recall that Hankamer's NAC specifies structural ambiguity and that, according to him, other potential disambiguating factors involved in (29), such as anomalous contrast (i.e. contrast between a country and a person), are irrelevant. Similarly, (28b) is a clear counterexample to Langendoen's hypothesis because told this stogy is a nonleft-peripheral string that contains an NP. It is also a counter9 Discussed in their course "Interactions Among Systems of Verbal Ability" at the 1974 Summer Institute of the LSA. 10 Counterexamples similar to (28a) were noted by Sidney Greenbaum at the conference "Testing Linguistic Hypotheses", University of Wisconsin, Milwaukee, Wisconsin, May io-i I, 1974, at which Langendoen's paper was presented. REMARKS AND REPLIES example to Hankamer's NAC because there is a derivation of the same string (although it is semantically implausible) with the "gap" at the left extremity: (30) *Harry told this story to his mother, and I [Harry told] Tom to his father. (3Ia-c) are more examples of the same type: (3 i) a. b. c. Second, Hankamer's (32) a. b. c. Two days ago, John took Mary out to dinner, and this afternoon, ,'[John took Mary out] to the movies.11 John comes to see Mary every day, and Bill I [comes to see Mary] every other day. John reported this incident to the FBI, and Bill '[reported this incident] to the CIA.12 observe the following sentence, whose interpretation is consistent with NAC and Langendoen's NLPNPC: John believes Mary to be guilty, and Tom to be innocent. and I [John believes] Tom to be innocent. and Tom * [believes Mary] to be innocent. What is peculiar about this sentence and, as a matter of fact, about most crucial sentences that Ross, Jackendoff, Hankamer, and Langendoen have discussed, is the fact that the NPs that are involved are all proper names. Let us now examine sentences of the same pattern in which the subject NPs of the first conjuncts are clearly marked as NPs that represent "new information". (33) a. b. c. d. Of the people polled, 8o% believe the President to be guilty, and 20% I [believe the President] (to be) innocent. 50% of his constituents asked the Senator to vote for the bill, and 25% ' [asked the Senator] to vote against it. SomeRepublicanswant Ford to run for the Presidency in I976, and others ' [want] Reagan [to run for the Presidency in I976]. Of the ioo people contacted, 70 promised the fund raiser to donate $Ioo, and 30 '[promised the fund raiser to donate] $200. 11 (3 I a) is a case of gapping that has never been mentioned before in the literature on Gapping: instead of a subject remaining, a contrasting clause-initial adverbial does. The sentence is a counterexample to the NLPNPC, but it is not a counterexample to the NAC because there is no other derivation of (i): (i) Two days ago, John took Mary out to dinner, and this afternoonAdV,to the moviespp. (3Ib) has another reading, which derives from (i): (i) John comes to see Mary every day, and V [John comes to see] Bill every other day. (3Ic) has the second reading, which is semantically semiplausible: (ii) ?John reported this incident to the FBI, and V[John reported] Bill to the CIA. 12 (28a), (3 I b), and (3 I c) all involve adverbial constituents for which the main verbs are not strictly subcategorized, and therefore belong to the class of counterexamples to the NAC that Hankamer (1973) discusses in connection with (20). 308 REMARKS AND REPLIES e. Somepeoplego to Europe every year, and others "[go to Europe] every other year. hit Mary with a baseball bat, and another "[hit Mary] f. Oneof themuggers with a bicycle chain.13 g. Somepeoplelive in this city because they like living here, and others 2 [live in this city] because they don't have means to move to the suburbs. Note that all the above sentences involve violation of Hankamer'sNAC and Langendoen's NLPNPC.14 Third, in case the preceding context clearly shows that the nonleft-peripheralNP represents old information, Gapping can be applied in violation of Hankamer's and Langendoen's constraints. Observe the following question-answer pairs: (34) a. Q: With what didJohn and Bill hit Mary? b. A: John hit Mary with a stick, and Bill I [hit MYary]with a bdlt. (35) a. Q: Who persuadedwho to examineMary? b. A: John persuaded Dr. Thomas to examine Mlary, and Bill I [persuaded] Dr. Jones [to examine M-ary]. (36) a. Q: Who persuaded who to be examined by the doctor? b. A: John persuaded Mary to be examined by the doctor, and Bill "[persuaded]Jane [to be examined by the doctor].15 Fourth, observe the following question-answer pairs: (37) a. Q: Who did John persuade to examine who? b. A: John persuadedBill to examine Jane, and I [John persuaded] Tom [to examine] Martha. (38) a. Q: Who persuaded who to examine who? b. A: John persuaded Bill to examine Jane, and *[John persuaded] Tom [to examine] Martha. (33f) stands in marked contrast with (ia-c): (i) a. John hit Mary with a baseball bat, and Bill with a bicycle chain. b. and V[John hit] Bill with a bicycle chain. c. and Bill *[hit Mary] with a bicycle chain. Note that (33f), which is grammatical, has the same derivational history as (ic), which is ungrammatical. 14 There are speakers for whom (33e,f,g) are acceptable, though (33a,b,c,d) are not. They also reject sentences such as (i): 13 (i) Some wanted rice, and others beans. It seems that they reject applications of Gapping that would yield "an indefinite quantified NP + X", where X is a constituent for which the matrix or constituent verb is strictly subcategorized. I have no explanation for this phenomenon. 15 Although the (b) sentences are acceptable as answers to the (a) sentences, they do not seem to be acceptable in isolation even if the same stress pattern is imposed. Namely, to many speakers, (i) is not acceptable in isolation. (i) J6hn persuaded Dr. Thomas to examine Mary, and Bill [persuaded] Dr. Jones [to examine Mary]. It seems that one needs the context that would justify the specified stress pattern. REMARKS AND REPLIES (37b) and (38b) have exactly the same underlying structure, but the former allows the gapping of John persuadedand to examine,while the latter does not. The explanation for the above phenomena seems to lie in the facts (i) that only those elements that are contextually known can be gapped,16 and (ii) that only those elements that represent new, unpredictable information can be left behind after Gapping.17 In (37b), John is contextually known in that it appears in (37a); hence the grammaticality of the sentence. On the other hand, in (38b), Joohnis not contextually known; hence, it is not possible to delete this NP in spite of the fact that it appears in the first conjunct of the sentence. The fact that Gapping is a pattern that is used for presenting contrastive pairs of information segments, and that, because of this semantic function, the constituents left over after Gapping must represent new, unpredictable information can be seen in the following example: (39) Johni eats peas, and Johni/hei *[eats] rice.'8 Now, it is generally the case in English that the closer a given constituent is to sentence-final position, the "newer" the information it represents. For example, if (40) is uttered with neutral intonation, John is ordinarily taken to represent old information (referring to the discourse topic), and Jane, new information.19 (40) John persuaded Bill to examine Jane. This is, of course, along the lines of the general discourse principle of presenting old information first and new information last in a sentence.20 Now, given this functional 16 I am indebted to one of the anonymous readers of this article for LinguisticInquiryfor (37)-(38) and the generalization presented here. 17 Greenbaum (in Quirk, Leech, and Svartvik (i972, 537f.)) alludes to information focus as a reason for ellipsis, although he does not specifically discuss Gapping. It is natural that Gapping, which is an ellipsis phenomenon, should delete constituents that represent old information and leave behind constituents that represent new information. 18 What is crucial here is the fact that due to the coordinate structure, the subjecthood of John in the second conjunct represents the most normal and expected state of affairs. Compare (37) with the following: (i) Before their arguments, John had usually agreed with Mary and she with him. Both she and himin the second conjunct are anaphoric, but they represent new, unpredictable information in the sentence. Similarly, observe the following sentence: (ii) During this difficult period, John argued with Mary, and she with herself. The fact that Mary is the subject of the second conjunct and the fact that it is the object, as well, is unpredictable from the information contained in the first conjunct. Note that (iii) is ungrammatical: (iii) *During this difficult period, John argued with Mary, and he with himself For the distinction that should be drawn between the concept "anaphoric" and the concept "old, predictable" information,see Kuno (1972, 271-272). 19 Of course, in many cases, the surface order of constituents is constrained by deep structure constituent order and cannot be altered, despite the fact that it may not be the ideal order in terms of "newness of information". However, when such situations arise (and they arise very often), emphatic focus stress is used to mark a nonfinal constituent that represents new information. 20 See Garvin (I963) for the outline of V. Mathesius's and other Prague School linguists' work on functional sentence perspective. See Quirk, Leech, and Svartvik (1972, 937-947) for new information occurring last in English sentences. 3IO REMARKS AND REPLIES sentence perspective (FSP) regarding the normal word order and the intended function of Gapping, namely that of deleting contextually known information and presenting contrastive (and therefore, new, unpredictable) information, it is easy to see why (4I), with neutral intonation on the first half of the sentence, is unambiguous: (4I) John persuaded Bill to examine Jane, and Tom, Martha. Since Tom and Martha are the two constituents left behind by Gapping, they must represent new information in the sentence, and therefore, they must be paired with the two constituents in the first half of the sentence that represent new information. When the first half of the sentence is pronounced with no emphatic stress on any of its constituents, Jane represents the newest information and Bill the second newest information (or, both Jane and Bill represent new information). John is taken as an NP that represents old information. Hence, it is not possible to pair Tom with John, and thus, the only interpretation that one gets for the sentence is that of (42): (42) John persuaded Bill to examine Jane, and I[John persuaded] Tom [to examine] Martha.2' On the other hand, when there is a preceding context that makes it clear that the ordinary functional-sentence-perspective flow of old-new information does not hold in a given sentence, as in (34)-(36), or when lefthand constituents are clearly marked as constituents that represent new information, as in (33), the two constituents left behind after Gapping can be paired with nonrightmost constituents of the first conjunct. I will summarize the observations I have made above: (43) The FSP Principle of Gapping a. Constituents deleted by Gapping must be contextually known. On the other hand, the two constituents left behind by Gapping necessarily represent new information and, therefore, must be paired with constituents in the first conjunct that represent new information. b. It is generally the case that the closer a given constituent is to sentencefinal position, the newer the information it represents in the sentence. c. Constituents that are clearly marked for nonanaphoricity necessarily represent new information in violation of (b). Similarly, constituents 21 If (40) is used as an answer to, say, "What happened?" without any previous mention of John, Bill, and Jane, all three NPs represent new information. In such a context, Gapping cannot apply: (i) a. Q: What happened? b. A: John persuaded Bill to examine Jane, and *(John persuaded) Tom (to examine) Martha. (ib) as an answer to (ia) is ungrammatical for the same reason that (38b) is an unacceptable answer to (38a). What is important to note here is that (40), when pronounced without any emphatic stress, may have only Jane representing new, unpredictable information, or both Bill and Jane or all three NPs representing new information, but that it cannot have only John or only John and Bill as representing new information. In such a case, a focus stress is obligatorily placed on john, or on John and Bill. REMARKS AND REPLIES that appear closest to sentence-final position necessarily represent old information (in violation of (b) ) if coreferential constituents appear in the corresponding position in preceding discourse. The above principle explains why (5b), repeated here, is grammatical: (5b) Paul Schachter has informed me that the basic order in Tagalog and related languages is VOS; Ives Goddard ; [has informed me] that the unmarked order in Algonkian is OVS; and Guy Carden [has informed me] that the basic order in Aleut is OSV. Personal pronouns in general represent old, predictable information.22 Furthermore, among personal pronouns, I/me represents the most "presupposed" information in that it refers to the speaker.23 For this reason, when the first conjunct of (5b) is pronounced with neutral intonation, with no emphatic stress, Paul Schachteris understood to represent much "newer" information than me does. Hence, Ives Goddardand Guy Cardenare paired with Paul Schachter,and not with me.24 2.2. The FSP hypothesis does not yet fully explain why (i8c) is not acceptable: (i8) a. *Jack asked Mike to wash himself, and Sue to shave himself. (Hankamer b. * c. (I973) ) and I [Jack asked] Sue to shave himself. and Sue * [asked Mike] to shave himself. - Why is it that even though (i8c) is the only semantically plausible derivation of (i8a), it still is not a possible interpretation ? It seems that the following semantic condition is in operation: (44) The Tendencyfor Subject-PredicateInterpretation When Gapping leaves an NP and a VP behind, the two constituents are readily interpreted as constituting a sentential pattern, with the NP representing the subject of the VP. (44) is needed to account for the grammaticality of the following examples: (45) a. John entered Harvard to study linguistics, and Bill I [entered Harvard] to study psychology. 22 Personal pronouns are always anaphoric, but they do not always represent old, predictable information. For example, in (i) the first person pronoun represents new, unpredictable information. (i) a. Q: Who won the first prize in the competition? b. A: I did. See also footnote i8. 23 See Kuno (I 972, 284-285) for a peculiar aspect of the behavior of the first person pronoun in Japanese that can be attributed to this fact. 24 (23), preceded by a context that would make Haj Ross a noun phrase that represents old information (with the resulting destressing of the noun phrase), would behave the same way as (5b). 312 REMARKS AND REPLIES b. John stopped frequenting the Pussycat Lounge to please his wife, and Bill ' [stopped frequenting the Pussycat Lounge] to save money. c. John struck Mary as being honest, and Bill ' [struck Mary] as being sincere. d. Bill was persuaded by John to donate $200, and Tom ' [was persuaded by John] to donate $400. e. John is expected by Mary to work on Problem A, and Bill ' [is expected by Mary] to tackle Problem B. Note that all the above sentences violate Langendoen's NLPNPC and Hankamer's NAC, as well as the Minimal Distance Principle and the FSP Principle that I have discussedthus far. What seems to make these sentencesgrammaticalis the fact that the two constituents left over by Gapping in each of the sentences constitute a subjectpredicate relationship. The grammaticality of (45d) and (45e), which have nonleftperipheralgaps, is particularlynoteworthy because the correspondingactive sentences do not allow nonleft-peripheralgapping: (46) a. Bill was persuaded by John to donate $200, and Tom to donate $400. b. and *[Bill was persuaded by] Tom to donate $400.25 and Tom " [was persuaded by John] to donate $400. (47) a. John persuaded Bill to donate $200, and Tom to donate $400. c. b. and ' [John persuaded] Tom to donate $400. c. and Tom *[persuaded Bill] to donate $400. It seems that this difference can be attributed to the fact that, in the active sentence, the object of persuadeis coreferentialwith the underlying subject of the infinitive, while in the passive sentence, the subject of bepersuaded is coreferentialwith the underlying subject of the infinitive. In order to examine the effectofthe Tendency forSubject-PredicateInterpretation further, let us compare (47) with the following example: (48) a. John promised Bill to donate $200, b. c. and Tom to donate $400. and '[John promised] Tom to donate $400. and Tom '[promised Bill] to donate $400. For all native speakerswith whom I have checked the data, (48a) in the (48c) interpretation is considerably better than (47a) in the (47c) interpretation.The only relevant difference between (47) and (48) is the fact that the former involves persuadeand the latter promise.But this is a critical difference because in the NP1 persuadeNP2 VP 25 One might argue that (45d) is not a counterexample to the NAC because the (46b) derivation, which has the gap at the left extremity, is ruled out by the ban on deleting grammatical formatives (prepositions and conjunctions) that mark functions of the constituents that follow them (see footnote 3). However, as I mentioned in footnote 8, since Hankamer's NAC requires that ungrammatical derivations such as (46b) be checked transderivationally, it will rule out (45d) as ungrammatical. The same is true for (45e). REMARKS AND REPLIES pattern, the NP2 is coreferentialwith the underlying subject of VP, while in the NP1 promiseNP2 VP pattern, the NP1 is. The fact that Tomin (48c) is coreferentialwith the underlying subject of to donate$400 makes up for the violation of the Minimal Distance Principle and the FSP Principle, and makes (48a) an acceptable sentence in this interpretation.Sixty percent of the approximately6o native speakerswith whom I have checked the data get (48b) as the firstinterpretationof (48a), and 40% (48c). This phenomenon can be explained by the three principlesthat I have discussedin this article. (49) (A) (B) Minimal Distance Principle (47b) (47c)* (48b) (48c) + + (C) FSP Principle Tendency for SubjectPredicate Interpretation + - + + - + (47b) is in conformity with all of the three principles, and (47c) in violation of all of them. This explains why (47a) is not interpretablein the (47c) reading. On the other hand, (48b) is in conformitywith (A) and (B) and in violation of (C), while (48c) is in conformity with (C) and in violation of (A) and (B). Depending upon the relative strengths of these principles in individual idiolects, (48b) is the first interpretationof (48a) for some speakers, while (48c) is the first interpretation of the sentence for others.26 Let us now returnto (i 8c). In the NP1 askNP2 VPpattern, NP1 is not coreferential with the underlying subject of the VP. Therefore,Sueof (i8c) does not form a subjectpredicate relationship with to shavehimself.Thus, (i8a) in the (i8c) interpretation violates the Tendency for Subject-Predicate Interpretation, as well as the Minimal Distance Principle and the FSP Principle; hence the ungrammaticalityof the sentence. When promiseis used instead of ask, the Gapping pattern of (i8c) yields grammatical sentences: (50) a. Jack promised Mike to shave himself, and Sue to wash herself. b. * and I [Jack promised] Sue to wash herself. and Sue V [promised Mike] to wash herself. c. Some speakersmay regard (5oa) in the (soc) interpretationas awkwardor marginal, but it seems that all speakersagree that it is considerablybetter than (i8a) in the (i8c) interpretation. 26 There are speakers who accept only the (48b) interpretation for (48a), and others who accept only the (48c) interpretation for the sentence. These wide idiolectal variations can be attributed to the difference, among idiolects, of relative weights of the three principles involved. Note that there are no such idiolectal variations for the interpretation of (47a). For all speakers, the primary interpretation of the sentence is that of (47b). For some speakers, the interpretation (47c) is nonexistent, and for others, it is weakly present in the sentence. But there are no speakers who accept only the (47c) interpretation to the exclusion of (47b). REMARKS 3I4 AND REPLIES 2.3. Observe the following question-answer pairs: (5 i) a. b. (52) a. b. Who persuaded who to examine Mary? John persuaded Dr. Thomas to examine Mary, and Bill V [persuaded] Dr. Jones [to examine Mary]. Who persuaded Dr. Thomas to examine who? John persuaded Dr. Thomas to examine Jaine, and Bill *[persuaded Dr. Thomas to examine] Martha.27 (sIb) violates the Minimal Distance Principle, but it is in conformity with the FSP Principle because the context (i.e. (5 Ia) ) makes it clear that the subject and object of persuaderepresent new, unpredictable information; hence, the sentence is grammatical as an answer to the question. The same is true for (52b), but it is almost unintelligible as an answer to (52a). Now, compare (52) with the following example: (53) a. b. Who promised Bill to examine who? Dr. Thomas promised Bill to examine Jaine, and Dr. Jones Bill to examine] Martha. V [promised (53b), which has the same Gapping pattern as the ungrammatical (52b), is grammatical for many speakers. Even for those who do not regard (53b) as perfect in the specified context, it is clear that it is considerably better than (52b). It seems that a tendency similar to that of Subject-Predicate Interpretation is in operation here. What seems to make (5Ib) and (53b) acceptable is the fact that the relationship between the two constituents left behind can be defined in terms of a for (s Ib), and Dr. simplex-sentential relationship (i.e. Bill persuadedDr. Jones.... In contrast, what seems to make (52b) unprocessable Jones examinesMartha for (53b)). is the fact that there is no simplex-sentential relationship between the two constituents Bill and Martha left over after Gapping. The above observation leads to the following generalization: (54) The Requirement for Simplex-SententialRelationship The two constituents left over by Gapping are most readily interpretable as entering into a simplex-sentential relationship. The intelligibility of gapped sentences declines drastically if there is no such relationship between the two constituents. It seems that the above constraint is perceptually motivated: it is difficult to recover the understood syntactic and semantic link if the two constituents do not hold a simplex-sentential relationship because the processing of the second conjunct would require supplying more than one verb. The following sentences all violate the Requirement for Simplex-Sentential Relationship: 27 I am indebted to Jorge Hankamer (personal communication (I 974)) for calling to my attention the contrastbetween (5I) and (52). REMARKS AND REPLIES Dr. Jones was persuaded by Bill to examine Jane, and *[Dr. Jones was persuaded] by John [to examine] Martha. b. Dr. Thomas promised Bill to examine Jane, and ? ?[Dr. Thomas promised] John [to examine] Martha. c. John visited Boston to see Mary, and *[John visited] Hartford [to see] (55) a. Jane. d. John hugged his sister to please his mother, and *[John hugged] his wife [to please] his father. One needs two verbs (namely persuadeand examine)and, therefore, complex-sentential structure to relate John and Martha of (55a). The same is true for (ssb) because it is not the case that John was to examine Martha, or that John promised Martha. The grammaticality judgment on (55c) is particularly clear-cut: it is considerably worse than (56), in which the two constituents stand in a simplex-sentential relationship to one another (i.e. john saw his uncle): (56) Bill visited Boston to see his father, and John ?[visited Boston to see] his uncle. Note that if Gapping leaves the entire infinitival VP behind in (55), we find grammatical sentences: (57) a. Dr. Jones was persuaded by Bill to examine Jane, and I [Dr. Jones was persuaded] by John to examine Martha. b. Dr. Thomas promised Bill to examineJane, and v' [Dr. Thomas promised] John to examine Martha. c. John visited Boston to see Mary and I [John visited] Hartford to seeJane. d. John hugged his sister to please his mother, and ' [John hugged] his wife to please his father. Since the NP and the VP in each sentence are major constituents of the matrix clause, they form a simplex-sentential relationship and do not violate the constraint under discussion. The following table scores, with respect to the four nonsyntactic constraints on Gapping, the crucial sentences that I have discussed in this subsection: Minimal Distance Principle (58) FSP Principle + + + (5ib), (53b) (*52b) (*55a,b,c,d) (?56) _ + - - (57a,b,c,d) + + - Tendency for SubjectPredicate Interpretation (not (not (not (not applicable) applicable) applicable) applicable) Requirement for Simplex-S Relationship + + + 3I6 REMARKS AND REPLIES The above table shows that the Requirement for Simplex-Sentential Relationship is a very strong and nearly inviolable constraint.28 3. The analysis of Gapping presentedin the preceding section still leaves many things unaccounted for. It does not explain the contrast between (20), (2 I), and (22) that Hankamerobserved.Similarly,it does not account for the fact, also noted by Hankamer, that (i6b) is not as acceptable as (20): Max writes plays in the bedroom, and Harvey ?[writes plays] in the basement. (i 6) b. Max wanted to put the eggplant on the table, and Harvey *[wanted to put the eggplant] in the sink. (20) Both sentences violate the Minimal Distance Principle and the FSP Principle and, therefore, the proposed analysis erroneously predicts that they would both be ungrammatical. Undoubtedly, the fact that (i 6b) has another derivation that is semantically semi-plausible, while (20) does not, is partly responsiblefor the difference in their acceptability: (59) a. *Max writes plays in the bedroom and ,'[Max writes] Harvey in the basement. (* for semantic anomaly) b. Max wanted to put the eggplant on the table, and ' [Max wanted to put] Harvey in the sink. It may also be that it is easier to apply Gapping to a string that representsa unitary self-contained semantic concept than to a string that does not. In (20), the gapped string writesplays, as a complete VP, representsa self-containedsemantic concept. On the other hand, in (i 6b), the gapped string wantedto put the eggplantdoes not represent a self-containedsemantic concept. The above constraintseems to be responsiblefor the contrast between (6o) and (6I): 28 All the grammatical gapped sentences that I have discussed in this article are in conformity with the Requirement for Simplex-Sentential Relationship except for (33a), repeated here: (33a) Of the people polled, 8o7o believe the President to be guilty, and 2o7%, v[believe the President] (to be) innocent. The sentence seems to be better with to be deleted. Since 2o70 and innocentdo not enter into a simplex-sentential relationship in this derivation, the constraint under discussion erroneously predicts that the sentence should be of a much lower degree of grammaticality than it actually is. I do not fully understand why (33a), with to be deleted, is acceptable to many speakers. It seems that it is related to the fact that (i) is a perfectly grammatical sentence. (i) 20%7 believe the President innocent. Note that in (i), 2o70 and innocentconstitute a simplex-sentential relationship. The above situation contrasts with the case of (33b): (33b) 50O. of his constituents asked the Senator to vote for the bill, and 25% V[asked the Senator] to vote against it. Note that if to voteis deleted, the resulting sentence is of a dubious degree of grammaticality. REMARKS AND REPLIES (6o) a. John went out to buy beer, and Bill ' [went out to buy] fried chicken. b. John was glad to see Jane, and Bill ' [was glad to see] Martha. c. John was kind enough to help Jane, and Bill I[was kind enough to help] Martha. d. John began singinga chanson, and Bill '[began singing] ajapanese song. e. John went out singing a chanson, and Bill '[went out singing] a Japanese song. (6i) a. John came home to find his wife sick, and Bill *[came home to find] his child [sick]. b. John must be a fool to have marriedJane, and Bill * [must be a fool to have married] Martha. c. Tom went to Florida to learn to play tennis and Bill *[went to Florida to learn to play] squash. d. John was upset having received A - for the course, and Bill ??[was upset having received] B - [for the course]. In a very vague sense, one feels that the gapped strings in (6o) represent unitary semantic concepts and that those in (6I) do not. However, it seems hopeless to come up with a set of formal proceduresto determine whether a given gapped string forms a unitary semantic concept or not.29 4. I have shown in this article that the observed degree of acceptability of gapped sentences is derived from the interaction of at least five nonsyntactic factors: the Minimal Distance Principle (perceptual; due to Langendoen and Grosu), the FSP Principle (discourse-based), the Tendency for Subject-Predicate Interpretation (semantic and perceptual), the Requirement for Simplex-Sentential Relationship (semantic and perceptual), and the requirement for semantic unity of gapped strings (semantic). I have no doubt that many other factors that are still unknown affect the grammaticality of gapped sentences.30These nonsyntactic factors leave, in the domain of pure syntax, a set of rather uninteresting constraintson Gapping such 29 It seems that there is a correlation between what can serve as a "bridge" for extraction and what can be deleted by Gapping. Observe the following sentences: (i) a. b. c. d. (ii) a. b. c. d. This is the fried chicken that Bill wentout to buy. There were those who Bill was glad to see, and those who he was not. The song that Bill begansingingwas a Japanese song. The song that Bill wentout singingwas a Japanese song. *My wife, who I camehometofind sick, was in a lousy mood. *The woman who John was a fool to havemarriedwas Jane. *What did Tom go to India to become? *What grade was John upsethavingreceivedfor the course? The semantic unity seems to be the common factor that controls Gapping and Relativization, Wh-Q Movement, etc. I am indebted to one of the anonymous readers of this article for LinguisticInquiryfor the above observation. 30 Greenbaum, for example, points out that semantic and pragmatic factors, including one's knowledge of what is being spoken about, affect the grammaticality judgment of sentences that involve Gapping. For example, observe the following sentence: 3I8 REMARKS AND REPLIES as: (i) Gapping can apply only to parallel coordinate structures; (ii) Gapping can leave two and only two constituents behind; (iii) gapped elements must include matrix verbs; and (iv) Gapping cannot delete grammatical formatives (prepositions, conjunctions, and complementizers) that play the role of marking the functions of the constituents that follow them.31 The fact that Gapping depends heavily upon nonsyntactic conditions is not surprising at all considering the fact that it is a deletion operation, and considering the fact that most deletion and reduction processes (Verb Phrase Deletion, Backward Equi NP Deletion, so-called Super Equi NP Deletion, Backward and Forward Pronominalization, Reflexivization, etc.) are very heavily constrained by semantic, perceptual, and discourse-based factors. Examining these processes in the "idealized" syntactic environments would simply trivialize the phenomena under investigation. I believe that I have clearly demonstrated in this article the need for a functional approach in linguistic analysis. References Garvin, P. L. (I963) "Linguistics in Eastern Europe: Czechoslovakia," in T. A. Sebeok, ed., Current Trends in Linguistics, Vol. I, 499-522. Hankamer, J. (1973) "Unacceptable Ambiguity," LinguisticInquiry5, I7-68. Jackendoff, R. S. (I97I) "Gapping and Related Rules," Linguistic Inquiry 2, 2I-35. "Functional Sentence Perspective: A Case Study from Japanese and Kuno, S. (I972) English," Linguistic Inquiry 3, 269-320. Langendoen, D. T. (I974) "Acceptable Conclusions from Unacceptable Ambiguity," presented at the conference "Testing Linguistic Hypotheses", University of Wisconsin, Milwaukee, Wisconsin, May io-i I, I974. "On 'Gapping and the Order of Constituents'," Linguistic Inquiry 3, Maling, J. M. (I972) ioi-io8. Quirk, R., G. Leech, and J. Svartvik (I972) A Grammar of ContemporaryEnglish, Seminar Press, New York. Ross, J. R. (1970) "Capping and the Order of Constituents," in M. Bierwisch and K. Heidolph, eds., Progress in Linguistics, Mouton, The Hague, 249-259. Stillings, J. T. (I975) "The Formulation of Gapping in English as Evidence for Variable Types in Syntactic Transformations," Linguistic Analysis I: 3, 247-2 73. Department of Linguistics Harvard University Cambridge, Massachusetts 02138 (i) Jackendoff teaches linguistic courses in the morning, and McCawley V[teaches linguistic courses] in the afternoon. Greenbaum notes that (i) also allows the reading JackendoffteachesMcCawleybut states that the lack of semantic parallelism between linguisticcoursesand McCawleymakes this interpretation unlikely, and that our knowledge of Jackendoff and McCawley makes the same interpretation highly improbable. 31 This constraint, too, is probably perceptually motivated. See footnote 3.
© Copyright 2026 Paperzz