THEORY AND DESCRIPTION IN AUTONOMIST AND COGNITIVE LINGUISTICS: Chapter Six

CHAPTER SIX: ELLIPSIS AND ASSOCIATED PHENOMENA

1.Introduction

This chapter starts from the notion that there must be something common to all forms of expression that are less than fully explicit. Naturally, there are differences as well -- otherwise they would not go by such disparate labels as Anaphora, Ellipsis, Right-Node Raising, Gapping, Conjunction-Reduction, Definite Descriptions, and so on. Nevertheless, it seems to me that Autonomist approaches to these phenomena have emphasized the differences at the expense of the similarities.

Functional Discourse approaches, seeing the sentences in a broader context, might be seen as more likely to explore the similarities. Likewise, a Cognitive approach might tend to view the underlying cognitive processes as probably very similar -- and thus to look for grammatical similarities.

One example where the two sorts of approaches mentioned above might lead to different conclusions occurs in Kuno (1975), which is a hybrid work with both Functionalist and Generativist aspects. He cites the following constraint:

(1) The Constraint on Verb Phrase Deletion: Verb Phrase Deletion can apply only to the VP that is VP-final.

However, compare the following examples:

(2) John wants to go to Paris, and Sally wants to, too.

(3) John wants to go to Paris, and Sally to Spain.

If one were to say that (3) was a counterexample to Kuno's constraint, the obvious retort would be that (3) is an instance of Gapping, not of Verb Phrase Deletion. There the discussion usually seems to stop, as if that were the last word on the matter.

But it is still possible to ask whether or not there is a higher-order generalisation to be found here. For example:

(4) The Constraint on Deletion in English.

Any type of constituent is deletable anywhere in English, apart from unaccompanied Noun Phrases and clause-final material which is not negated.

I suggest that this covers (2) and (3) with a single rule, and takes account of the fact that some item such as "too" is necessary in final position after an instance of Verb Phrase Deletion. The reference to negation is included because of examples such as:

(5) John wants to go to Paris, but not Sally.

The literature does not contain (as far as I am aware) any conclusive demonstrations that any alleged case of Ellipsis (or Deletion) is anything distinct from, or more than a case of something being left unspecified by the speaker/writer. Little attempt has been made in the literature to sharply delineate the supposed difference between these two types of case, as far as I am aware.

The Competence-Performance Distinction (see Chapter One) is used to argue that this question can be decided on grammarinternal, formal grounds -- yet, in practice, such grounds are never conclusive. There is an infinity of parameters relevant to the situations described in language which are (from time to time) left unexpressed, but it is not clear to me that one could say that all of them have been "elided" or "deleted", which would imply that they had all originally been present in some way.

For example, in (5) there are many relevant factors left unexpressed, such as what means of transport John wants to use, when he wants to go, for how long, etc.. Is there a clear way of defining Ellipsis so that (5) is not deemed to be "elliptical" as to those particular factors ? I doubt it.

Thus I propose the following rephrasing of (4):

(4') The Constraint on Inexplicitness in English.

Any type of information or constituent may be left totally unspecified in any part of any syntactic structure of English, apart from bare Noun Phrases (where special provisions apply), and clause-final material which is not negated.

I suggest a basic division of Ellipsis into Forwards Ellipsis and Backwards Ellipsis -- on the model of Backwards Anaphora (Cataphora) and Forwards Anaphora (Anaphora proper). The Antecedent precedes the point of ellipsis in Conjunction Reduction and Gapping, whereas the reverse ordering applies to Right Node Raising. Hence, I categorise Conjunction Reduction and Gapping as Forwards Ellipsis, and Right Node Raising as Backwards Ellipsis. I discuss what differences there are between these two types, and whether they amount to natural classes.

Many of the apparent differences between Conjunction Reduction (CR) and Gapping disappear if a non-standard approach is taken to constituent structure . If we assume a tripartite, instead of a binary branching of S ( NP-VP[-NP] -- i.e. with the object of a transitive verb being directly dominated by S -- instead of NP-VP), then both Gapping and CR can be seen to apply just to daughters of the conjunct concerned. This also explains the necessity for non-peripheral deletion to involve the first verb of the conjunct.

Nevertheless, non-peripheral gaps are harder to process than (in particular) sentence-initial gaps. This is because their processing requires a delay, while subsequent context is processed and used to help decide what the "understood" element was meant to be. This delay entails extra effort in terms of both memory and processing. The speaker thus has less freedom in the case of Gapping, especially as ambiguity or garden-path effects may more easily occur.

Gapping is distributed similarly to DO IT Anaphora, which is used especially if the predicate is too complex for Gapping. Sometimes the semantics of the Gapped verb are such that DO IT is semantically inappropriate as an Anaphor:

One problem with Gapping ( as Kuno 1976 points out) is that Autonomists do not seem to have defined it adequately. The issue has been confused by the indiscriminate inclusion, under the heading of "Gapping", of examples where the subject of the non-initial clause is absent, as well as the verb. I opt for maximal clarity, by distinguishing such examples from Gapping proper and calling the former Conjunction Ellipsis (CE).

The difference is that Conjunction Ellipsis involves eliding two or more phrases (where a verb on its own can count as a "phrase") at the left extremity of a clause, whereas Gapping involves elision of the verb (with or without other lexical items as well) with a residue remaining at both extremities of the clause.

On the methodological level, I argue that a Functionalist/Cognitivist is justified in asking how plausible and probable a particular example is, and whether there are more plausible/probable alternative ways of communicating the same message. As mentioned elsewhere in the present book, Cognitivists may be more likely to bear in mind the fact that there are degrees of comprehensibility, and that the hearer uses strategies designed to cope with the more comprehensible, rather than the less comprehensible ways of expressing a given sentiment.

Conversely, a speaker can normally be expected to bear the processing needs of the hearer in mind to a greater or lesser degree. Whereas an Autonomist is in principle interested in any sentence that could in theory be generated by his system of rules, a Functionalist/Cognitivist may tend to be interested in example-types to a degree proportionate to their frequency of occurrence. This is because the more frequent example-types may be regarded as more prototypical (Zadeh(1965)).

To explain the apparent distribution of various forms of Ellipsis and Anaphora, I propose a Stylistic Anaphora Hierarchy (SAH). Conjunction Ellipsis (CE), for example, being at the top of the SAH, is never incomprehensible to the degree that Gapping can be --other factors being equal.

2.Backwards and Forwards Ellipsis

As my point of entry into this area, I would like to discuss Hudson (1976), which has the merit (from my point of view) of comparing three such structures -- even though he happens to conclude that they are completely separate phenomena. These are the structures he discusses:

(6) a. Mary opened the window at 8:00 and looked out (Conjunction Reduction).

b. Mary cooked the first course and Jane the second (Gapping).

c. Mary likes, and Jane would go anywhere to find, antique horse-brasses from the workshop of that genius in metalwork, Sam Small (Right-Node Raising).

Hudson (ibid) groups them together (for the purposes of the article) because they alone share two features: a) they reduce the length of a coordinate structure by reducing all but one of the conjuncts; b) and they apply only to coordinate structures -- with some exceptions in the case of Right-Node Raising (RNR). The exceptions that Hudson gives relating to RNR are as follows:

(7) a. Of the people questioned, those who liked outnumbered by two to one those who disliked the way in which the devaluation of the pound had been handled.

b. I'd have said he was sitting on the edge of, rather than in the middle of the puddle.

c. It's interesting to compare the people who like with the people who dislike the power of the big unions.

Hudson (ibid, 536) places great stress on the fact that he is not

concerned here with any kind of reduction process that isn't restricted to coordinate structures. In particular, our rules are not intended to cover either NP's whose head is missing as a result of identity-of-sense anaphora (e.g. John's as in John's is bigger than mine), or clauses whose time or place adverbial is missing but understood from the context (e.g. because it was too warm [in his study] in He turned the fire off in his study because it was too warm). Not surprisingly, both of these types of reduction are possible in coordinate structures, since they are possible virtually anywhere; but since they aren't restricted to coordinate structures, it would be both unnecessary and misleading to demand that the rules of CR etc. should cover them.

This passage makes it seem rather extraordinary that Hudson should nevertheless group RNR with CR and Gapping. In addition, one could make the (important) point, in passing, that it is not at all obvious that in his study is in fact omitted or elided from because it was too warm in Hudson's example. Assuming anything has actually been ellided at all (and this assumption may be just an artefact of the Transformational approach), then it may just as well have been something like where he was sitting, or even a non-temporal expression like for him to work comfortably. There is an infinity of parameters relevant to the situation described in Hudson's example which are left unexpressed, but it is not clear to me in what sense one could say that all of them have been "elided" or "deleted". This I regard as the most serious problem with Generative (and Hudson (ibid) works within a Generative framework) approaches to Ellipsis and Gaps in general.

Given that the examples in (7) set RNR apart from Conjunction-Reduction (CR) and Gapping, it is all the more telling that the Antecedent precedes the point of ellipsis in CR and Gapping, whereas the reverse ordering applies to RNR. The one common feature that is left -- that all three structures reduce the length of a coordinate structure by reducing all but one of the conjuncts -- applies to other structures (see the passage quoted above) which Hudson explicitly excludes from the scope of his study for the sole reason that they also apply to other types of structures.

Hudson (ibid) quotes Tai (1969)'s Identity Deletion Rule:

(8) IDENTITY DELETION. Delete one of the two highest identical constituents in the pair of conjuncts under consideration according to the following principle: If these two identical constituents are left branches, deletion operates forward; if they are right branches, it

operates backward.

The point at issue (proposed by Tai, and denied by Hudson) is whether Gapping and CR constitute one single rule.

Hudson (ibid) claims that this apparently simple generalisation would have to be made very complicated in order to cope adequately with the facts. He says that Gapping can delete any type of constituent, as long as it contains the first verb of the conjunct -- whereas CR is restricted to daughters of the conjunct concerned. He also points out that Gapping (but not CR) leaves behind just two "balanced" constituents. He also claims that, unlike CR, Gapping cannot apply when the conjuncts are joined by but.

Hudson (ibid, 544) makes similar criticisms of the following Deletion Rule from Harries (1973);

(9) DELETION RULE (optional). In a coordinate structure in which each conjunct contains a constituent which is identical to a corresponding constituent in all other conjuncts, delete all but the leftmost of these identical constituents.

In fact, these apparent differences between CR and Gapping disappear if a different approach is taken to constituent structure and to the data that Hudson (ibid) cites. If we assume a tripartite, instead of a binary branching of S (i.e. NP-VP[-NP] instead of NP-VP), then both Gapping and CR can be seen to apply just to daughters of the conjunct concerned. This also explains the necessity for non-peripheral deletion to involve the first verb of the conjunct.

The notion of "balancing", as Hudson (ibid) uses it in connection with Gapping, seems unclear and subjective. He compares

(10) John invited Mary and Bill, Sue. (Gapping)

with

(11) John worked hard and was awarded the first prize at the end of the year. (Conjunction Reduction)

What Hudson appears to mean is that (10) involves two pairs of nouns, whereas (11) involves an intransitive verb + adverb, paired with a transitive verb + object -- the latter combination being much longer than the first. However, compare:

(12) John invited Mary, and his brother an old flame from years back. (Gapping)

In (12), Mary and an old flame from years back are arguably just as "unbalanced" as worked hard and was awarded the first prize at the end of the year in (11).

Hudson is also incorrect in stating that Gapping cannot involve conjuncts linked by but. Consider:

(13) The President brought his wife, but the King his concubine.

Having said that, I must admit that non-peripheral gaps are harder to process than (in particular) sentence-initial gaps. This means that the speaker has less freedom in the case of Gapping, as ambiguity or garden-path effects may more easily occur, as in this example from Hudson (ibid):

(14) John sings in the bath and Mary (sings) arias at the top of her voice.

Hudson does not describe (14) as involving anything resembling garden-path effects (he just uses it to show that "balance" is required for successful Gapping). However, the very fact that he inserts the gapped verb sings (in brackets) is an implicit recognition of the fact that the reader might otherwise have taken arias to be a (neologised) verb. Compare:

(15) John reads in the bath and Mary feminist books on equity for conscripts' wives.

The lack of parallelism between the two conjuncts of (15) certainly makes it stylistically odd -- but not actually incomprehensible. A comparison of (14) and (15) shows how skewed examples can lead to wrong conclusions in Generative Linguistics.

To see that Gapping is not really so much more restricted than CR, we can follow Hudson (ibid) by pointing out that this structure actually (Hudson says only marginally) allows three separate constituents per conjunct. Here again, the particular example chosen influences the judgement. In (16), from Jackendoff (1971), the large amount of processsing required by all the NP's makes it harder to sustain Gapping:

(16) Arizona elected Goldwater Senator, and Massachusetts McCormack Congressman.

Jackendoff (ibid) also provides a more digestible example:

(17) Willy put the flowers in a vase, and Charlie the book on the table.

The following example of my own, however, seems even more clearly unimpeachable:

(18) They both got it wrong: Sam put the chopsticks in his ears, and Tom the wontons in his teacup.

3.Gapping and Similar Phenomena

Kuno (1976) is my focal point in this section, principally because his approach is more or less Functionalist.

The reader may recall the example of Gapping cited above:

(10) John invited Mary and Bill, Sue.

(19) Robin will tickle Sandy with a pencil, and Chris will do it to Pat.

(20) Max seemed to be trying to force Ted to leave the room, and Walt [seemed to be trying to force] Ira [to leave the room].

(21) Max seemed to be trying to force Ted to leave the room, and Walt seemed to be doing it/trying to do it to Ira.

Example (20) comes from Jackendoff (op.cit.). Example (21) shows, I think, that DO IT is not a very comprehensible Anaphor for a verb like "trying".

The following example from Kuno (1976) shows that not only simplex verbs can be gapped:

(22) I want to try to begin to write a novel, and Mary

to try to begin to write

to begin to write a play.

to write

[ ]

One of the facts that previous discussions of Gapping have focused on is the contrast between the following two types of examples (from Jackendoff (1971)):

(23) I want Bob to shave himself and Mary [wants Bob] to wash himself.

(24) John writes poetry in the garden, and Max [writes poetry] in the bathroom.

Generativists seem to regard (23) as ungrammatical, and (24) as grammatical. I myself regard (23) (in isolation, and without regard to intonation, stress, or contextual features) as hard to understand in the required sense, while (24) I find easily comprehensible, independent of such factors. Apart from the fact that (as always in such cases) I reserve the right --indeed the obligation -- to test (23) in plausible contexts, and with regard to intonation and stress patterns, my position on grammaticality vs comprehensibility essentially agrees that there is some contrast here that must be explained in some way.

See what happens when we put (23) into a plausible context:

(23') It's Bob's birthday tomorrrow, and, rather than giving him presents, we've all decided to get him to do something that WE want him to do to make US happy. Some of these tasks are fairly exotic, e.g.: Sally wants Bob to hop round the block, and Paul wants him to climb a coconut palm. On the other hand, some of the tasks are fairly banal, e.g.: I want Bob to SHAVE himself, and Mary to WASH himself.

Although I have not tested this myself, I feel sure that a statistical survey would show that more native-speakers would agree that (23') had the intended meaning than would think that of (23). No matter what the actual figures or proportions, I would be happy if it were thus shown that providing a context affected Comprehensibility Intuitions in that direction.

Be that as it may, the essential difference between (23) and (24) is that the former invites, and the latter does not invite confusion in the hearer's mind as to the Agent-Patient relationships in the gapped conjunct.

A problem with Gapping (as mentioned above) is that Autonomists do not seem to have defined it adequately. It seems to me that the issue has been confused by various authors' inclusion, in their discussions of Gapping, of examples such as:

(25) I want Bob to shave himself and [I want] Tom to wash himself.

Kuno (1976) makes the point that examples such as (25) differ from Gapping proper, in that the subject of the second clause is elided along with the verb, but he decides to fall into step with the terminology of the authors whose work he is discussing. I will opt for clarity by distinguishing Gapping proper from examples such as (25), which I will treat as involving what I will call Conjunction Ellipsis (CE). The difference is that the latter involves eliding two or more phrases (where a verb on its own can count as a "phrase") at the left extremity of a clause, whereas Gapping involves elision of the verb (with or without other lexical items as well) with a residue remaining at both extremities of the clause.

The following is Jackendoff (1971)'s proposed solution to what he sees as the contrast between (17) and (18):

(26) Jackendoff's Condition on Conjunction Ellipsis

When the sequence NP VP appears in the VP of the right conjunct, Gapping cannot delete NP, leaving VP behind.

Hankamer (1973) cites several counterexamples to (26), such as:

(27) Max gave Sally a nickel, and Harvey a dime.

Hankamer considers that the interpretation whereby Max gave Harvey a dime is grammatical, whereas that under which Harvey gave Sally a dime is ungrammatical. Again, providing a context which heavily topicalises the notion of giving Sally some money would make the latter reading much more comprehensible, but there is a contrast here to be explained.

As Hankamer (op.cit.) points out, (26) would have to be complicated and made (even) less natural in order to cope with all the counterexamples he cites. He himself provides a more natural explanation:

(28) Hankamer's No-Ambiguity Condition (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.

This has the additional merit of being related to a more general principle (though, on the debit side, it confuses the issue by lumping together Gapping and Conjunction Ellipsis, as we have seen):

(29) The Structural Recoverability 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.

Kuno, who cites (28) and (29), feels that the latter is too strong, and that (28) is violated by certain examples of gapping. Langendoen (1974) criticises (28) on various grounds, of which the least valid (from a Functionalist point of view) is that there are technical reasons for denying (28) the status of a transderivational constraint. Despite the fact that this notion is outdated now, as far as Generative theory is concerned, it is worth making the Functionalist point that Autonomist Linguistics tends to get itself into this sort of position, where what might otherwise seem to be valid and natural generalisations are outlawed on grounds relating to the technical nature of the particular types of rules then in fashion.

However, Langendoen (op.cit.) does provide counterexamples to (28) as well, relating to the emphasis in the latter on the "left extremity".

(30) Max sent Sally the messenger last week, and Susan yesterday.

Langendoen considers that the reading whereby Max sent Sally Susan is grammatical, and that whereby Max sent Susan the messenger is ungrammatical, although in both cases the "gap" is at the left extremity. I myself doubt that comprehensibility judgements, if surveyed, would concur with Langendoen's grammaticality judgements in the case of (30) -- provided the examples were properly contextualised and involved plausible lexical items, e.g.:

(31) The perfume salesman has been sending samples to people in our neighbourhood recently. He sent Sally a sample last week, and Susan yesterday.

Actually, (30) is not a very likely sentence to be uttered in the first place, as the ambiguity can (and usually would) be avoided by using the Anaphorised alternatives:

(32) Max sent Sally the messenger last week, and he sent him to Susan yesterday,

(33) Max sent Sally the messenger last week, and he sent Susan to her yesterday.

Whereas Gapping and CE delete the "old" information entirely, these examples merely pronominalise it. A Functionalist/Cognitivist may be justified in asking how plausible and probable a particular example is, and whether there are more plausible/probable alternative ways of communicating the same message. Unlike Autonomists, Cognitivists are likely to bear in mind the fact that there are degrees of comprehensibility, and the hearer uses strategies designed to cope with the more comprehensible, rather than the less comprehensible ways of expressing the same sentiment.

Dowty (1980) expresses rather similar views:

A neo-Gricean conversational principle

If a language has two (equally simple) types of syntactic structures A and B, such that A is ambiguous between meanings X and Y, while B has only meaning X, speakers of the language should reserve structure A for communicating meaning Y (since B would have been available for communicating X unambiguously and would have been chosen if X is what was intended).

Likewise, the speaker can normally be expected to bear the processing needs of the hearer in mind to a greater or lesser degree. Whereas an Autonomist is in principle interested in any sentence that could in theory be generated by his system of rules, a Functionalist/Cognitivist is interested in example-types to a degree proportionate to their frequency of occurrence. This is because the more frequent example-types are regarded as more prototypical (Zadeh(1965)).

My explanation in terms of Functional confusion needs to be expanded to read as follows:

(34) Stylistic Anaphora Hierarchy

In English, there exists a stylistic hierachy of Anaphoric devices in (partially) Complementary Distribution, which (in descending order) is:

1) Conjunction Ellipsis;

2) Pro-Form Anaphoras;

3) Gapping.

Where confusion would be possible as to Agent/Patient roles in an utterance, the speaker is expected to go high enough up the hierarchy to avoid such confusion.

The Stylistic Anaphora Hierarchy (SAH) implies that in (23), for example, the level of Comprehensibility is low, because there exists an alternative structure (Pro-Form Anaphora) which is higher up the hierachy, viz.:

(35) I want Bob to shave himself and Mary wants him to wash himself.

(23) I want Bob to shave himself and Mary to wash himself.

Of course, because of the masculine reflexive pronoun himself, no CE reading is possible in (23).

In fact, of the 30 examples of CE which are cited in Kuno (1976) (though not under that name, of course), only 10 (i.e. 1/3) are stated to be ungrammatical. Of these, 6 are ungrammatical for reasons having to do with the Semantics of the lexical items involved. This leaves only 1/5 of the total cases of CE cited which are alleged to be ungrammatical for syntactic reasons. Of the remaining 4, 3 involve a combination of Gapping and CE, e.g.:

(36) Max sent Sally the messenger last week, and [Max sent] Susan [the messenger] yesterday.

The first gap in (36) is a case of CE, and the second gap is a case of Gapping proper (as I have defined it). The Comprehensibility problem is caused by the Gapping gap, rather than by the CE gap, as can be seen by comparing the following versions:

(37) Max sent Sally the messenger last week, and Max sent Susan [the messenger] yesterday.

(38) Max sent Sally the messenger last week, and [Max sent] Susan the same messenger yesterday.

The word "same" has to be inserted in the second clause of (38) in order to naturally maintain the same word-order, including repetition of "the messenger" (rather than pronominalising to "him").

The only remaining case of alleged ungrammaticality involving CE is one where, in a passive sentence, the Agent Prepositional Phrase has been split up unnaturally, with the "by" being elided, and the rest of its Phrase being left behind:

(39) Bill was persuaded by John to donate $200, and [Bill was persuaded by] Tom to donate $400.

If the "by" is not elided, then this instance of CE ceases to be problematic:

(40) Bill was persuaded by John to donate $200, and [Bill was persuaded] by Tom to donate $400.

Thus Kuno really gives no examples of CE which are syntactically malformed -- whether this malformation is expressed in terms of "grammaticality" or Comprehensibility. Is this a random characteristic of the examples that Kuno happened to choose to discuss, or does this reflect something about CE itself ? I will choose the latter alternative, and proceed on the assumption that CE, being at the top of the SAH, is never incomprehensible to the degree that Gapping can be --other factors being equal.

Langendoen (op.cit.)'s own proposed solution runs as follows:

(41) Langendoen's Nonleft-Peripheral NP Constraint (NLPNPC)

Gapping cannot apply so as to delete nonleftperipheral strings that contain NP other than a clitic pronoun adjoined to its V.

Kuno (op.cit.) provides some counterexamples to the NLPNPC, such as (for example):

(42) My brother visited Japan in 1960, and my sister [visited Japan] in 1961.

Kuno's proposed solution involves a comparatively complex agglomoration of four separate constraints:

(43) 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.

(44) 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 sentence-final position, the newer the information it represents in the sentence.

c. Constituents that are clearly marked for nonanphoricity necessarily represent new information in violation of (b). Similarly, constituents that appear closest to sentencefinal position necessarily represent old information (in violation of (b)) if coreferential constituents appear in corresponding position in preceding discourse.

(45) The Tendency for Subject-Predicate Interpretation

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.

(46) The Requirement for Simplex-Sentential Relationship

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.

Although these four constraints together seem somewhat cumbersome, simplicity must take second place to observational adequacy. In other words, if they account for the facts of CE and Gapping adequately, then how simple they seem is less important. Moreover, even if one were to (as I will, below) propose an alternative Functional explanation, some of the above constraints are so self-evident that they may well retain some role (albeit peripheral) in the overall explanation.

A closer look at the 28 examples of Gapping proper that Kuno (1976) considers "ungrammatical" reveals that . of them (i.e. 22) come under what I call the Real Estate Institute approach to Intuitionising.

Just as it seems fairly common for potential house-buyers to be shown a house they can be counted on to consider undesirable first, and secondly the house that the Real Estate Agent thinks they are likely to want to buy (so that the contrast with the first house makes it seem even more appealing), so these 22 examples all involve a contrast between cases of CE (which, as we have seen, are virtually all judged to be "grammatical") and paired cases of Gapping proper.

I would predict that such pairing has a systematic effect on intuitions of Descriptive Grammaticality, Normative Grammaticality, and Comprehensibility, which could be experimentally investigated along the lines pioneered by Carroll, Bever, and Pollack (1981). The effect is likely to be one of making the Gapping examples seem more ungrammatical or incomprehensible than they would in other contexts.

I will give one example of this kind of pairing that Kuno (1976) carries out: We cited example (48) above : --

(23) I want Bob to shave himself and Mary [wants Bob] to wash himself.

Kuno pairs this with:

(23a) I want Bob to shave himself and [I want] Tom to wash himself.

Not only does it seem inappropriate to pair an example of Gapping proper (23) with one of CE (23a), but it also seems to me that (23) has a built-in bias towards failing the Intuitiontest, in that (assuming a Phonological Identity Criterion for this kind of Ellipsis) there is a Phonological difference between the third-person and first-person forms of the (singular) verb want. This means that the antecedent is not 100% compatible with the intended interpretation .

However, we saw what happens when we put (23) into a plausible context:

(23') It's Bob's birthday tomorrrow, and, rather than giving him presents, we've all decided to get him to do something that WE want him to do to make US happy. Some of these tasks are fairly exotic, e.g.: Sally wants Bob to hop round the block, and Paul wants him to climb a coconut palm. On the other hand, some of the tasks are fairly banal, e.g.: Jane wants Bob to SHAVE himself, and Mary to WASH himself.

Of the remaining 6 examples, 4 involve examples that are unusually long, so that the lower degree of apparent Comprehensibility is probably due to the load on the memory that is involved in gapping over large intervals. Some of these 4 are not only lengthy, but they also involve more than one gap, e.g.:

(47) John came home to find his wife sick, and Bill [came home to find] his child [sick].

As far as Comprehensibility is concerned, I see no problem with (47) at all, but one can see why an Autonomist might (or might not) have qualms about its "grammaticality", as a Normative Grammarian might well want to outlaw such sentences from the English language .

One of the other two examples of "ungrammatical" Gapping has been supplied with a detrimental context:

(48) a. Q: Who persuaded who to examine who?

b. A: John persuaded Bill to examine Jane, and [John persuaded] Tom [to examine] Martha.

Kuno (ibid) himself uses this example, contrasted with the same answer in a more appropriate context, to claim (rightly, in my view) that only "old" information can be gapped, and that what remains after Gapping must be "new" information.

Kuno's only other example of "ungrammatical" Gapping serves to underline the same point:

(49) John1 eats peas, and John1/he1 [eats] rice.

Thus a Functionalist approach that incorporates the SAH, exercises care with the contextualising and grouping of examples, distinguishes Gapping proper from Conjunction Ellipsis, and utilises the notion of Comprehensibility, rather than Descriptive Grammaticality, can come to the relatively simple conclusion that all cases of Gapping and Conjunction Ellipsis are comprehensible (to a degree dependent on the clarity of the context and the length and number of segments involved) -- provided "pragmatic" conventions relating to "old" and "new" information are observed.

4.Conclusion

This chapter has attempted to emphasise what is common to all forms of expression that are less than fully explicit. With this aim in view, a strong claim was made in the form of the following constraint:

(4') The Constraint on Inexplicitness in English.

Back to Contents Page