Chapter 3 that this could also happen with substitution. If a housewife on

Chapter 3 that this could also happen with substitution. If a housewife on seeing the milkman approach calls out Two ple.uc! she is using exophoric ellipsis; it is the context of situation that provides the information needed to interpret this. But exophoric ellipsis has no place in cohesion, w we shall not explore it a.ny further here. Let us summarize here the general features of reference, substitution and ellipsis, harking hack to what was said in tfle final paragraph of Chap- ter 3· All three are forms of presupposition, devices for identifying some- thing by referring it to something that is already there- known ro, or at least recoverable by, the hearer. Since this • something • that is presupposed

4.1 ELLIPSIS, SUBSTITUTION AND REFERENCE 145 may be an element in a preceding sentence, these devices have a cohesive effect; they contribute very largely to cohesion within the text. Reference is presupposition at the semantic ievel. A reference item sig- nals that the meaning is recoverable. though not necessarily in the form of the actual word or words required. For this reason a reference item cannot necessarily be replaced by what it presupposes; even if the presupposed item is present in the text, the reference to it may require an item. of a dif- ferent function in structure. At its simplest, reference is a form of situa- tional (exophoric) presupposition; but it is regularly used :in textual {end.ophoric) presupposition, pointing backwards (anaphoric) or some- times forward (cataphoric). In many styles of discourse, including almost all written language, reference is always textual rather than situational. Substitution, and here we include etli.psis as a special case of substitution, is presupposition at the level of words. and structures. When a substitute is used, it signals that the actual item required, the particular word -or group or clause, is recoverable from the environment; and the substitute pre- serves the class of the presupposed item., which may therefore be replaced in the • slot • created by it. The difference between substitution and ellipsis is that in the former a substitution counter occurs in the slot~ and this must therefore be deleted if the presupposed item is replaced. whereas in the latter the slot is empty - there has been substitution by zero. Unlike reference, substitution is essentially a textual relation; it exists primarily as. an anaphoric (-or occasionally cataphoric) device, and in its rare exo- phoric we it tends to give an effect of • putting the words in the other person's mouth~. ln tabular form; Substitution and Reference ellipsis Let~/ cf abstraction semantic lex:icogrammatical Primary source of situation text pTesupposition What is presupposed? meamngs items (i~ words, groups, clauses) Is class preserved? not necessarily ye• Is replacement possible? not necessarily ye. Use as a cohesive yes; anaphoric and yes; anaphoric (occa.s- tkvice cataphoric sionally cataphoric)

146 ExampJe:s: [4:5] a. This is a fine hall you have here. rm proud to be lecturing in it. b. This is a fine hall you have here. I've never lectured in a ftncr one. c. This is a ftne hall you have here. I've never lectured in 2 finer. Example (4:5a] is reference. [t would be possible to replace it by some expression containing the word hall; but it would have to be altered from the original (eg: in this .fine httll), a.nd it still sounds somewhat awkward. Examples (b) and (c) are substitution and ellipsis, and it would. be quite natural to add half after finer (deleting ()ne in (b)). In what follows, we shaH discuss ellipsis under three headings: Nominal ellipsis (4.2) Verbal dlipsis (4-3) Clausal ellipsis (4.4) There is one further general point to be made first. We noted ~hove, in Chapters 2 and 3. that a reference item, or a substitute, may relate to some- thing in the same sentence, such that the presupposition takes place within the confines of a single structure. This is no different in principle from any other instance of reference or mbstitution, though it may lu.ve certain specia1 features, such as the cataphoric reference of the to a <)_ualifier in the nominal group in example [2:61 J. But in the analysis. of texts. relations within the sentence arc &.irly adequately expressed already in structural terms., so that there is no need to involve the additional notion of co- hesion to account for how the parts of a sentence hang together. Between sentences, however. there are no strUCtural relations., and this is where the study of cohesion becomes important. For this reason in both these chapters we concentrated on reference and substitution as relations between sentences,. largely ignoring intra-sentence presupposition. We shall do the same here. Ellipsis. or something closely related to it, also occurs within sentences, as in [4:2-] above; and there arc certain special stntctural possibilities, types. of BRANCHING <.>tructute, which do not occur when the presupposition is between sentences. fu general we shall not be concerned with dJipsis within the sentence., for the same reason as already given; it can be explained in terms of sentence structure and does not constitute an ind.ependent agency of cohesion in the text. What we are interested in is ellipsis as a form of relation between sen- tences, where it is an aspect of the essential texture. The relevance of ellipsis in the present context is its role in grammatical cohesion.

1:47 4.2 Nominal ellipsis 4.2.1 Ellipsis within the nominal group By NOMINAL ELLIPSIS. we mean ellipsis within the nomina} group. The structure of the nominal group was outlined in 2.1. On the logical dimen- sion the structure is that of a Head with optional modification; the modi- fying elements include some which precede the Head and some which foUow it, referred to here as Premodifier and Postm.odifier respectively. Thus in those two fast electric tTains with panWgraphs the Head is trairu, the Premodifier is formed by those two fast electric and the Postm.odifier by with pantographs. The Modifier is combined with another structure. on the experiential dimension, which consists of the elements Deictic (d), Numer:a.tive {n), Epithet (e), Clas~oo (c), and Qnalifier (q), represented here by those, two, fast, electric and with pantographs respectively. The Deictic is normally a determiner, the Numerative a numeral or other quantifier, the Epithet an adjective and the Classifier a noun; hut these correspondences are by no means exact. There may be Suhmodifiers at various places; these are usually adverbs like so. very and. too. The Qualifier is normally a relative clause or prepositional phrase. The noun in this structure has the function referred to as the Thing. Most elements may occur more than once, .and the tendency for this to happen increases as one moves towards the later elements of the suucture. The function of Head, which is always filled., is normally served by the common noun., proper noWl or pronoun expressing the Thing. Personal pronouns are reference items and were described in Chapter 2; they will not he discussed. further. Proper nouns designate individuals, and are therefore not capable of further specification; they may sometimes be accompanied by descriptive modifiers, but these are not subject to ellipsis. Common nouns. on the other hand, designate classes; they are often fur- ther specified. and this: is the function of the elements Deictic. Numerative. Epithet and Classifier. Now under certain circumstances the common noun may be omitted and the function of Head taken on by one of these other elements. This is what is meant by nominal dlipsis. In a non-elliptical nominal group, the Head is the Thing. the noun designating the individual or class referred to. This m:.ay he a phenomenon of any kind: person, animate or inanimate object, abstraction, institution. process, quality. Sbte or relation. In an elliptical nominal gmup. this ele- ment is not expressed, and one of the other elements (Deictic, Numerative, Epithet or Classifier) functions as Head. This is very frequently a Deictic

148 or Numerative. much less frequently an Epithet. It is very rarely a Classi- fier; since the Clasgffier is usually a noun~ if i.t functioned as Head it would be liable itself to be interpreted as the Thing (so, for examplt; we cannot replace a tall brick chimney by.: taU brick; see 4.2.3 below}. In general, with exceptions to be noted below (4.2.3 and 4-2.].5). any nominal group having the function ofHead filled by a word that normally functions within the Modifier is an elliptical one. Nominal ellipsis therefore involves the upgrading of a word :limctioning as Deictic, Numerative, Epithet or Classifier from the status of Modifier to the status of Head. For example; (4:6] Four other Oysters followed them, and yet another four. In the seCond line four, wbich is a Numerative and therefore normally acts as Modifier. is upgraded to function as Head. Simibdy in [4:7] Which last longer, the curved rods or the straight rods?- The str.rigbt are less likdy to break. straight is an Epithet, functioning as Modifier in the question but as Head in the response. Both another four and the straight are elliptical nominal groups.. An dliptical nominal group dearly requires that there should be available &om rome source or other the information necessary for 611ing it out. Faced with anot:he., four. we need to know • another four what?' Normally~ the source of information is a preceding nominal group. A nominal group that is elliptical presupposes a previous one that is not, and it is therefore cohesive. If we "'h-ant to fill out .an elliptical nominal group. for text analysis purposes, there are two ways of doing so. One way is simply to ~push down' the element functioning as Head, making it a Modifier. and add the 'missing' Head in its place. (The question whether any other • missing' dements would have to be supplied is discussed in 4.2.3 below.) By this process another four in [4:6l would become another fcur oysters. The other way of doing it ls to keep the elliptical group as it is and add a partitive Qnalifier; this would give another four of the oysters. The partitive is pos-- sible only under certain conditions: generally, when the elliptical group designate! some aggregate - a subset, fraction, quantity or collective - that is different from that designated by the presupposed group. Hence in (4:7] the partitive form is not possible. The head noun in the partitive expression will be singular or non-singular (plural or ma:~s) according to

4-.2 NOMINAL ELLIPSIS .149 the type of aggregate: singular if the elliptical group is partitive in the narrower sense (ie designating a fraction), and non-singular otherwise. The former type arc less readily elliptical (hut see 4.2..34 below on inde- finite quantifiers): [4:8] a. How did you enjoy the exhibition?- A )or (of the exhibition) was vety good. though not all. b. How did you enjoy the paintings? - A lot (of the paintings) were very good, though not all. So an elliptical nominal group may always be replaced by its full, non- elliptical equivalent, either in simple form or in expanded. partitive form. In either case. the presupposed items arc restored. The two possibilities arise becawe the partitive type is in any case a regular form of the English nominal group~ obligatory in some instances, such as where there is quanti£ication within the deixis as in (4:9a], and optional in certain others such as[4:9h]: f4:9] a. Two of my rosebushes were uprooted. b. That was his most popular £1mjthe most popular ofhis films. The partitive Qualifier may itself contain an elliptical nominal group, as in one afthe three, any ofFred's. We may now modify the earlier statement that a nomina1 group having Deictic, Numerative. Epithet or Classifier as Head is always elliptical. If it contains a partitive Qualifier, it i-s not ellip- tical- nnless the partitive Qualifier is itself elliptical. Some further examples: a. the best. h. the best hat. [4: IO] Which hat will you wear? This is c. the hest of the hats. d. the be" of the three. e. the best you have. In .all cases the is Deictic, thrff is Numerative, hest is Epithet and hat is the common noun representing the Thing. Then: (a) is elliptical; the is Modifier, best is Head. (b) is non-elliptical; the best is Modifier, hat is Had. (c) is non-elliptical; the is Modifier, b~st is Head, of the hats is partitive Qualifier-, non-elliptical. (d) is elliptical; structure as (c), except that the partitive Qualifier <if the three is itself elliptical. (e} is dliptical; structure as (c). except that the Qualifier you have is not partitive.

IjO 4.2.2 Presupposition of nomimd elements An elliptical nominal group is cohesive; it points anaphoric.ally to another nominal group which is presupposed by it. But how much of the presu~ posed group is in fact included within the presupposition? So far we have merely indicated that the Thing designated by the com- mon noun is presupposed. But there may he other elements in the pre- supposed group which likewise do not occur in the eUiptical one~ for [4-: u] Here are my two white silk scarves. 1 can lend you one if you like. Here rme presumably presupposes not only scarves but also the garnishings white and silk; it could be filled out as one white silk scaif, or one cJ my white silk scarves. This makes it possible to state what CAN he presupposed. by reference to the structure of the nominal group in terms of the elements Deictic, Numerative, Epithet and Classifier; note that these elements occur in the order stated, followed by the Thing. In d)ipsis. the Thing is always presupposed. (We have already pointed out that in ellipsis the Thing is al- ways a common noun. since proper nouns and pronouns do not take defining Modifiers.) In addition, any element following the one that is up- graded in the elliptical nominal group may he presupposed. Thus If Head is filled by Thls must be presupposed: These maybe presupposed: Deictic Numerative Epithet Classifier Numerative, Epithet, dassifier Epithet. Classifier Classifier As already noted above. it is rare for the Classifier to occur as Head. These patterns are exemplified in [4:u]. In [4:ua], yours (Deictic) is Head, and the presupposed elements include not only scarves (Thing) but aho silk (Classifier), whiu (Epithet) and po,;bly twc (Numerative): [+: r2] Here .are my two white silk scarves. a. Where are yours? b. I used to have three. c. Can you see any b1ack? d. Or would you prefer the cotton?

4-2 NOMINAL ELLIPSIS l5I [.;: 12h] ;, J;ke [.;: II]; three presupposes scarf, silk and white. In (c) the dliptical group is any black; this presupposes scarf and silk. but there may be any number and they may not be mine. Fina1ly, in (d) only sa:if is P'empposed. In general, then, the range of possible presuppositions is dependent on the structure of the nominal group. It extends only over: that pan of the presupposed group which could follow the element acting as Head in the elliptical group. Those pans which would precede or be con- current with it are excluded fro-m the presupposition; and this restriction apparently applies even to subcategories within the Deictic and Numera- tive (see 4.2.4.1-4 below}, for example~ {4: r3J a. They haven't got my usual morning paper. Can I borrow yours! b. The first three buds all fell off. We'll have to watch the next. where yours excludes usual (even though yt:tUr usual could occur) because both your and usual are Deicric elements, and next excludes thr-ee (even though next three could occur) because both are Numerative elements. But there is considerable indeterminacy at this point. What can be presupposed, therefore, is anything having .a function in the series d-n-e- c that ls LA~ than that occupied by the Head of the elliptical group. Whatever has the same or a. preceding function is repu- diated. To exemplify once motey if the presupposed group is those two tal{ brick chimneys, the following table shows what is repudiated and what is not repudiated {and therefore may be taken over by presupposition) by the various elliptical groups; note that x stands for the function of Thing: If elliptical group These are These are not JS: repudiated: repudiated: which? (d) d ne c x=two tall brick chimneys three (n) dn e c x=ta11 brick chimneys two not so taU (ne) dne c x= brick chimneys some stone (d c) dnec x=chimneys The fu.rt:her 'to the right' the fmal dement of the presupposing group, the more usual it :is to presuppose by substitution rather than by ellipsis. We would expect some stone ones rather than .rome stone (stone= Classifier); and perhaps also two not so tall ones (tail=Epithet).

Ij2 However, it is not necessarily the case that everything that could be presupposed actua1ly is presupposed. We can take it as a general guiding principle that it will be-, but this wi1l certainly need to be modified to some extent. Consider an example such as [4: 14] Don't you like those three little white eighteenth-century stone cottages? - I prefer mine, My three little white eighteenth-century stone cottages? Or just my -cot- tage? The aruwer is possible even if 'mine • is one btge red Elizabethan brick and timbered one. We would accept any interpretation that made sense and was consistent with what we already knew. It is worth noting, at the same time, that a form such as mine in this context is in the strict sense of the term ambiguous: it could stand as the realization of a number of different selections. We do fmd a rough scale of probability, extending from right to left in the nominal group. Of the elements that MAY be _presupposed in any given instance, namely those that follow the element that is explicitly repudiated in the elliptical group, we have seen that the Thing, that which is designated by the Head in a non~lliprical structure, always is presupposed. Going 'from right to left', the Classifier, if present, is very likdy to be: the Epitbet somewhat less likely, and the Numerative less likely stilL The Deictic, being the first dement, rwrmally cannot he pcesupposed, by the principle illustrated in [4: r2] and [4-: 13] above. And there is an overriding principle that the ptesupposed items must be continuous.: it would not be possihle for mine in [4: I4] to presuppose three little- white but not eighte-enth- century sturu!. It is slighdy odd, therefore, to fmd all the elements in .a long nominal group. including the Classifier, repudiated in an instance where sttuc- tur.illy they could be pre.upposed. [4: 15] is only doubtfully acceptable, (4015]! thinkl"ll get one of those gorgeous big ted china dogs. Mine harks too much. In spoken English, there is often a phonological indication of the extent of presupposition. It is characteristic of an elliptical nominal group that its Head carries tonic prominence in the tone group. This is natural. since tonic prominence is the realization of MW or contrastive information, .and an elliptical nominal group (like one with substitution) is inherendy new. in the sense that it dHfecs in some respect from the one it presupposes; not necessarily having a different referent (it could refer contrastively to the same thing). but the function of an elliptical item is to start afresh. taking

I)J the presupposed item as a reference point. Thus the occurrence of an elliptica1 nominal such as mine allows us to predict that somewhere in the environment is an item expressing a Thing~ such as hat. which either was not 'mine' or. if it was. demanded some explicit reaffirmation of the fact, as it would for example in answer to the question Whose is this hat? - Il s mine. In this sense an elliptical group always embodies some new informa- tion. Now it ofien happens that the presupposed nominal group signals a particular point of repudiation- an element with which the presupposing elliptical group is specifically in contrast - by the device of tonic promi- nence. For exampJe, if your hat is to be followed by mine, the tonic wiU &11 on yow. This is especially likdy to happen in a question-answer sequence, or if the two nominal groups :are part of an utterance by the same speaker, -..yho may have planned the information structure as a whole: That's not YOUR hat. It's Mll'o"E. A MARKED tonic (tonic prominence falling on some dement other than the last) signals contrastive information; eg [4.: I6] The two WHITE silk scarves were beautifully made. Here the word white is tonic, and this gives an expectation that if an ellip- tical (or substitute) nominal group follows it will he one that repudiates white, such" why diJ ycu buy the pink(one)? 4.2.3 Types of nominal ellipsis We now consider in more detail some exa:mples of the most frequently occurring types of nominal ellipsis, with conunents on the words or word classes that function as Head in the elliptical group. We have already noted that the Classifier is very rarely Iefi: to function as Head. In {4: 17a-d] we have four examples in which the presupposing group contains. a Classifier;. they are given in a substitute form. with one(s) as Head. and only in the last of the four wou1d it be possible to delete the substitute leaving an elliptical group ending in a Classifier: [4: 17] a. Don't you like babies?- Yes. hut I c:an't stand crying ones. b. I've never tried Mn Sugdcn~ s cherry cake. but I like her ginger one. c. Borrow my copy. The library one is out on loan. d. Did you win a first prize?- No, I only got a third one. The principle behind this restriction is very clear. The dassifier is typi- cally. though not always. realized by a word that could also .realize the

154 .ELLIPSIS Thing: usually a noun, as in (b) and (c), or the -ing form of a verb as in (a). Hence a nomina1 group having such an item as its Head would norma11y be interpreted as non-elliptical (so I can't stand uying would he interpreted not as 'crying ones! ie crying babies' but simply as • the phenomenon of anyone crying'). There are some instances where a Classifier CAN function as Head, usually those where the elliptical interpretation of the resulting nominal group is in some way the most natural one: for example if the Classiflel' is not a noun. as in [4: 17d], or if the presupposing status of the nominal group is signalled by an anaphoric the as in [4: zzd} above. These. however, are a minority. But, as suggested in the previous section, we really have a gradation or 'dine • here, rather than a sharp distinction between the classifier and the rest. 'The structural formula Deictic - Numerative - Epithet - Classifier represents a gradual move, in the process of specifying the class of' things • that is expressed by the Head, from one type of specification to :mother: beginning, in the Delctic, with a kind of specification that is temporary. and related to the actual speech situation. and moving on to one that is increasingly permanent and inherent. Specification of the flrst k:ind is achieved by items in closed systems, such as this/thAt. or the pronominal possessives; that of the second kind by lexical items, which form "open' classes, Hence as one moves along this scale. the actual words used are more and more noun-like; they are words which themselves have the potential of expressing a class of' things~ such as is typically expressed by a noun functioning as Head, and so they are liable to he interpreted a'>: Head. This being the case, such words are LESS likely to function as Head when they are expressing something else. This does not mean that a nominal group having an Epithet or Classifier in it CANNOT be anaphoric and co- hesive; but it will tend to achieve this status by substitution rather than by ellipsis. The most characteristic instances of ellipsis,. therefore, are those with Deictic or Numerative as Head. Here the situation is the other vv-ay round: substitution is much Jess common~ and in some cases excluded altogether. So we have for example Deictic: Numerative: Ellipsis these, my. ••r the firs~ two, mon Substitution these ones, my ones, any ones the fost ones, two cnes. more ones

4.2- NOMINAL ELLIPSIS I$$ In principle any Deictic or Numerative clement can function as Head in nominal ellipsis, with some minor exceptions which will he noted. For the remainder of 4.2, therefore. we shall he concerned mainly with deictic and numerative elements, with a rdativdy short section on epithet ellipsis at the end. Deictics. are considered first. 4.2.3.1 SPE.CIFIC DEICTICS Following on from the earlier account of -deixis in the nominal group, we recognize a division of the Deictic element into two puts. one forming the Deictic properly so caUed and one which has been referred to as POST- DEICTIC. The words functioning as Deictic are mostly of the class of determiner; with the demonstrative, possessive and indefinite determiners forming a network of systemically related categories - one that indudes the articles, which are thus shown to be part of a wider system. Those functioning as Post-deictic are adjectives. In [.f.: 18} Here the other guinea-pig cheered, and was suppressed. the is Deictic and other is Post-deictic. Post-deictics are discussed in 4.2.3·3 hdow. Within the Deictic proper, the major distinction, and that which is most relevant to ellipsis, is into specific deictio {possessives, demonstra.tives and the} and non-specific (each, every. all, both. any, either, no, neither, some. and 4). Non-specific and specific deictics may be combined. only through the use of a partitive qualif1er, eg: eadt of my children, any of the answers, some tif that pudtling. The exceptions are all and fmth, which can be joined direccly to another determiner, in what is sometimes therefore referred to as PRE-DEICTIC position. as in all our yesterdays, lwth these gates. The words all and both very frequently function elliptically. They may refer back to a single nominal group; if so it will be plural, having the serue of' two • if presupposed by both and more than two if presupposed by all. There is no equivalent singular form; and curiously all is NOT used elliptically to refer to a mass noun. even though in non-elliptical nominal groups it is regularly • mass •, as in all the milk was sour. Forexampk: [4: 19} a. 'The men got back at midnight. Both were tired out, b. The men got back at midnight. All were tired. out. c. The milk couldn't be used. All was sour. Of these, {c) is impossible; there is in fi.ct no elliptical form here. just as there is no fOrm of substitution with mass nouns. At the same time,

Ij6 however, the item presupposed. by all or both may consist of separate nominal groups; so we would have to say rather that these words pre- suppose a certain number of entiti~ which may have been expressed -either in the form of one plural noun or :H- different nouns. singular or plural Furthermore these 'entities • are themselves- sets of any extent. We can summarize this by saying that both refers to two sets and all to three or more sets. These sets may be combined. in one nominal group, as in [4:19a and b) above, and in [4:20a] where the presupposed item is the parents, re '(i} the father- and (ii) the mother •. Or each one may be -a separate nominal group. not always in the same sentence but usually with some indication that they belong together, such as parallelism of structure; an example is [4:2ob] where Ewth presupposes [1) the parents and {ii) the childun. And in this case any one set may itself be complex and consist of a further coordination, as in [4:20e}, which has (i) paTents ami other respon- sible .uluh.! and (ii} children. [4:20) a. The parents could not be traced. Apparently both were abroad. b. The parents may enjoy it, but the children will be bored. You cannot please both. c. I£ parents and other responsibJe adults make no concessions. children will rebeL And both will be certain they are right. In the last type, there must be explicit linbge between the items that are being treated as a single set (parents and other responsible adults). We cannot have lwth presupposing (i) the boy's porents and his teachers. and {ii) the bey, in {4.:21], because there is nothing to show that the boy's parents and his teachers bdong together: [4: 21] The boy" s parents had no time for him . .M. school, his teachers could make little contact. Yet the boy had a lot of ability, if he'd tried.. [ suppose both were at fault, really. Sometimes it is not clear which items are being grouped together, and am biguit:y rerults, as in {4:22] The &therand the mother were so busy making money that the two chi1dren were left to rheir ovm devices. Nanrnally both were resentful. In addition to .ali3lld both, other Deictic elements regularly function as Head of an elliptical nominal gr-oup. The speciftc deictics are {i) demonstra- tives and the, and (ii) possessives. The demonstrarives (this, that, these, those.

4.2 NOMINAL BLLIPSIS 157 and which?) all occur elliptically, w-ith very great frequency. Since they are themsdves reference items (see 2.4 and 2.4.1), they are often anaphoric anyway; but wherever the nominal group could be 'fiUed out" with a noun Head, or by the substitute one(s), a demonstrative functioning as Head is in fact an example of ellipsis. One example will suffice~ tlwse is elliptical for those pills: [4:23) Take these pills three times daily. And you'd better h3ve some more of those too. The word the does not operate elliptically; since its ftmction is to signal that the • thing' designated is fully defined~ but by something other than the itself, it normally requires another item with it, as in the two, tire small (one)~ the one that got away. Where it could have occurred elliptica1ly it is replaced by its non-reduced cognate form tlutt. Possessives include both norninals (Smith's, my father's, etc) and pro- nominals (my, your, etc). The latter have a special form when functioning as Head; mine, ours, yours, his, has. theirs, whose, and (rarely) its. Whenever a possessive occurs as Head it is elliptical, and in the case of the third per- son pronominals this means, as noted in Chapter 2 above (see 2.3.4. [2:24]). that there is a double cohesive tie. An item such as hers presupposes both a person as possessor and a thing possessed, the former by reference (/re,), the latter by ellil"i' (the po,;essive)' [4: 24] Jwt ask Janet how to polish the bnssware. Hers sparKles. 4.2.J.2 NON--Sl'.BCIFIC DEICTICS The non-specific Dcictics are e.u-h, every, any, either, no, neither, a. and sl.ltnr. as well as all and both which have already been discussed. Of these, all occur as Head of an elliptical nominal except every, hut a and no have to be represented by the forms one and none respectively. Some examples: [4:25] a. I hope no bones are broken ?-None to speak o£ b. I won't be introduced to the pudding, please. May I give you some? c. Have some wine.- I don't see any \.Vine.- There isn't any. d. Write an essay on the Stuart kings. Two pages about each will do. e. His sons: went into business. Neither succeeded. Of these, either and neither are like both in presupposing two sets; and each presupposes two or more. Again, these may be expressed either as one

158 plural nominal group or separately; .and if they are separate, any one set may itself be a coordination. as in [4:26a and h]. Hence ambiguity may arise i.n the same way as with all and buth; for example if[4;26b} had three betlrooms. the each might presuppose just these. [4:26] a.. Smith andJones are on holiday. [wonder if either has left an addres~. b. The flat has a sitting-room, a dining-room and one bedroom. Each has a window overlooking the park. The non-dual equivalents of eitlter and neithcr are .1ny and no; they are like all, except that they can occur dliptically with singular and mass nouns. as in [4 :25a and c]. The two pairs are proportional: no is to any as neither is to either. No and neither are of course negative. hut are usually restricted to clauses of declarative mood where the verb is positive; while any and either occur in clauses which are interrogative or hypothetical, OR where the verb is negative, or is posi:tive and the sense is 'it doesn't matter which'. In the latter type any, when used elliptically, repudiates any car- dinal numeral in the presupposed group and is usually singular { = • any one') unless some numeral occurs with it. such .as any three. Hence [4~27) Here are thirteen cards. Take any. Now give me any three. In its interr-ogative, hypothetical or negative use (htu he any 1 if he has any; he ham•t any), the diiference between singular and plural is neutralized. In non-elliptical groups, the plural form is usually used (has he any friends?}. as it tends to be .also with no(hehas no friends); hut when any is Head of an dliptical group this may be filled out with either a singular or a plural noun Head irres.pective of the nunl~r of the presupposed group, and likewise if any is Subject it may occur 'ioV-i.th either singular or plural verbs: {4:28]a. IwantamapoftheLakeDistrict. } Ha > ant f the L ~ '- ~~tn· t ve you got any. w some maps o a&e L"Ul c . b. I'm expecting a letter. } {Has .any ccme? I~ m expecting some letters. Have any come? In the same w.ay the singular/plural distinction is neutralized with the negativerw. Its elliptical form n()fl€ ( =tw+one) shows that it was originally treated as singular, but usage is no longer consistent: [4:29] I've checked all the files. None were/was missing. The elliptical some was mentioned in Chapter 3 (3.2.3.3. [3:25] and

Ij9 Table 5}. This is the non-singular (mass or plural) form of the indefinite article, and when functioning elliptically. as Head, it is always in its non- reduced form, ie [sAm] not [sm]. The nominal group presupposed by it may be singubr or plural, and any numeral is repudiated: [4: 30J a. These apples are delicious. Let's buy some. b. I had a dozen tennis balls; where are they?- I've got some; you can borrow mine. In (b), some does not mean •some of the dozen'; cf: I haven't got any. Parallel to some in its non-reduced funn is one, which is the non-reduced form of the singular indefinite artide a. It is this from which is derived the nominal substitute one discussed in Chapter 3 (3.2 and 3.2..I). As pointed out there (3·3·3·3· [3: ISci]), it is difficult to distinguish elliptical one from one of the uses of the substitute one; but the difference a ppean in the plural. since the plural of the substitute one is ones whereas the plural of the deter- miner (indefinite article) one is some. In [4:31) the one is an elliptical in- definite article (if(J :27)): [4:31] But you make no remark?- I didn't know I had to make one, just then. The elliptical use of deictic dements is a major source of cohesion in English texts. The Deictic is the element in the nominal group that relates to the HERE. AND NOW. linking the thing referred to to its verbal and situational context. It is natural, therefore, that it should he typically used. as a means of harking back: to a thing that has already been mentioned, while at the same time recontextualiz:ing it by anaphoric or exophoric reference. 4.2.3.3 POS'l'-DEICTICS The words functioning as Post-deictic element in the nominal group are not determiners but adjectives. There are some thirty or forty adjectives used commonly in Deictic function, and a number of others used occa- sionally in this way; the fiequentones include r>ther. same. different, identical, usual, regular. certain, odd. famous, well-known, typical. obvious, They com- bine with the, a or other determiner (the combination a+ other being written and pronounced as one word tmf)ther); and they may be FOLLOWED BY a Numerative, wdike adjectives in their norm.al function as Epithet which must FOLLOW any numerative dement. The distinction of meaning between Deicric and Epithet (:and'!{ 2.5.I [2:8o] above) can be seen in

16o ll:I.LIPSlS Deictic Epithet the identical three questions the usual two comments a different three people the odd few ideas the obvious fint place to stop three identical questions two usual comments three different people a few odd ideas the first obvious place to stop Of the adjectives used in deictic function the ones which regularly oa.-ur dliptica1ly are same and ()!her. The elliptical use of the same was treated. as substitution(:;.z.s). since it has been extended to very general use including clause substitution, as in do the same; actually an examp1e such as [3 :st] above (rll htwe the same) is simply an elliptical nominal group with same :as Head. The Post-deictic other combines either with specific Deictic (the other, that other. etc) or with non-specific (any other-, another, etc), and when it is used as Head it has a special plural furm .,thus. The nominal group which it presupposes need not be of the same number. and any numeral in it is repudiated, a:s with any. Example: [4:32-] I've used up these three yellow folders you gave me. Can I use the other? which does not mean 'the other three'. With a specific Deictic, orher(s) refen to the last remaining member(s} of a set. and therefOre it presupposes that all others must have been specified. This explains the frequency with which it is preceded by another nomi.nal group. often also anaphoric (eg: one, some of them, the first, etc) which is both presupposed and presuppos- ing: presupposed by other, but itself also relating back to the ultimately presupposed item. For example [4: 33] A group of well-dressed young men suddenly appeared on the stage. One of them bowed to the audience; the others stood motionless. The original item must be semantically plural- it must refer to more than one set. as described in 4· Z.J. I above; and the elliptical otber{s) presupposes just as much of it as does the intermediate item. Here one of them and tht! others both presuppose well-Jressed young men. The elliptical use of oti:Pr illustrates very well the indeterminacy which may arise in the extent of presupposition. If we had just the example [4:34] I see you've wld those two large red china dogs. Have you any others?

4.2 NOMINAL BLl.IPSIS 161 in the absence of further evidence we could not tell whether to fill this out as china dogs. red china dogs or large red china Jogs. Similarly in [4: 3.3]: the other yellow ones or just the other one? As we pointed out earlier, the extent of the presupposition may he signalled in the spoken language by the lo- cation of the tonic nuderu.. So if in [4.: 34] the toni.c falls on red. others means 'of another colour'; if on large. it means • any small ones', and so on. Finally we may note that elliptical nomina[ groups with Deictic as Head may also he exophoric, either in the generalized sense or specifically to the context of situation: [4-;35] a. Some say one thing, others say another. b. All is lo-st. c. Ail go into the other room. d. Have you been to Mary's recently? e. I'll have the usual, plea~. 4.2.3.4 NUMERATIVES Of the elements occurring after the Deictic in the nominal group, only the Numerative and certain types of Epithet function at all regularly as the Head in ellipsis. l11e Numerative clement in the nominal group is expressed by numerals or other quantifying words, which form three subcategories: ordinals. cardinals, and indefinite quantifiers. The ordinals are first, next, last, second, third, fourth, etc; they are often used dliptically. generally with the or a possessive as Deictic: [4:36] Have another chocolate.- No thanks; that was my third. Like the superlative form of an adjective, which in many ways it resembles (ordinals are in a sense 4 superlative numerals'), an ordinal is itsdflikdy to he presupposing even if the nomina1 group in which it occurs is not dlip- t:ical; thus the serond questicm presupposes that there was a first question, and the fost questirm thn there is likely to be a next. Again like superlatives, ordinals are often cataphoric to a Qualifier which indicates the domain of the ordering, supplying the information • first. etc, in what respect ?'; for example, to leave in {4:37] Smith was the first person to leave. I was the second. Cardinal numerals are also frequent in ellipsis, and may be preceded by any Deictic that is appropriate in number, eg; the rhTee, these three, any three, all three, and a1so by post-deictic adjectives as in the usual three, the same three.

162. l!LLIPSIS [4:38) a. Have another-chocolate.- No thanks; rve had my three. h. 'The other messenger's called Hatta. I must have two, you know. One to come, and one to go: With both ordinals and cardinals the presupposed notm may be either singular or plural. but it cannot be a mass noun unless there is also some measure word present or presupposed- naturally, since • mass • = • t.mcount- .ahle •. For example if in (4: 36] and [4: 38a] we had· Have some more tea. the answer would still he possible in each case, but only because it could he interpreted as presupposing • cup(s) •. The indefinite quantifiers are itcrns such as much, many, more, most,fiw, several. a little • lets, a bit, hundreds, etc; they include numerous transient and mote or less slang expressions especially used by children. Like other items with a numerative function. they are very frequently used in ellipsis; being indefinite. they are usua11y not accompanied by a Deictic, except where a is demanded as in a lot, although the comparative forms more, Ji!Wer and less may be preceded by no or any. Someofthemare specific to either count or Dl.1SS nouns. Examples: [4: 39] a. Can all cat<> climb trees.?- They all can; and most do. h. 'You ought to have a wooden horse on wheels., that you ought!. - • ru get one: the Knight said thoughtfully to him- sel£ • One or two - several. • Many of the indefinite quantifiers derive from measure nouns; for example lot, amount, and the larger numbers such as hundred and thou:1anJ. Since these still require partitive Qualifiers {a lot of . .. ). they are not very clearly distinguished &om the general class of measure noun. which in- dudes quantitatives (eg: half, pi<u, J.zen), partitives (eg: part, side, end) and collectives (eg: group. se~ pack). For the purposes of cohesion. these also can be regarded as requiring to be • filled out' by a partitive Qualifier. and therefore as elliptical if functioning as Head. Hence Alice' s predicament in {4: 40] 'One side will make you grow taller, and the other side will make you grow shorter.' 'One side of wh.lt? The other side of what?' thought Alice to henelf. • Of tbe mushroom; said the Caterpillar,just ;as if she had asked it aloud. Some combinations of quantifiers are possible. namely ordinal numeral plus cardinal or, in a few cases, ordinal numeral p)us indefinite; usually in the order stated. The combination of cardinal plus indefinite occurs only

.j.2 NOMINAL BLLIPSIS 163 if the indefinite is comparative, as in three more. Such combinations are regularly elliptical, .eg: the last three, the next few. Whereas the specific Dei.ctics - the demorutratives and posscssives - tend to occur alone. being themselves reference items., the Numeratives, like the non-specific Deictics, tend to be filled out precisely by a reference item in the form of a partitive Qualifier with third person pronoun. So we often find any of them, the first of them, three cf them and so- on. These are of coune still cohesive. but the presupposition is of the reference type rather than eTiipsis. Like the Deictics, Numeratives in e1ilptical use may be exoplwric; eg in {4: 37] we might have had Smiilt was the first to leave, with person under- stood. The presupposed item will he assumed to be some general category of which the item referred to, here Smith, is a member. This can be demonstrated by [4:41] Her money will be the first to leave her. Her husband will be the next. whicll puts heT money and her husband into the same general category by presupposition. Note the special exophoric use of a possessive Deictic plus cardinal numeral to mean 'children'. as in the proud mother's remark (4:42] My three are absolute terrors. Indefinite quantifiers occur exophorically 1n expressions like (4;43] He expects a lot . .But you can't do much to help him. 4.2.J.j EPITHETS The function of Epithet is typically fulfiUed by an adjective. It is not com- mon to find adjectives occurring as Head in ellipsis- colour adjectives are perhaps the most usual - except in their comparative and, especially, superlative forms. This rdlects the fact that superlative and comparative -adjectives are really functioning in a way that is more like a Numerative; possibly instead of the function Numerative in the nominal group we should recognize a more general function Ordinative, which would in- clude superlative and perhaps also comparative adjectives, as ·\veil as the classes of word that function as Numeratives- proper (numerals and inde- finite quantifiers). The superlative adjective precedes other Epithets and, like ordinal numerals (if .f-2.3·4 above), is usually accompanied by the or a possessive Deictic. Note in this connection the difference between (a) and (b) in [4:44]:

164 [4-: 44] a. Apples are the cheapest in autumn. b. Apples are cheapest in autumn. In {a) we may fairly ask 'the cheapest what?~; the cheapest is an elliptical group presupposing some item such as fruit. Example (h) is however not elliptical; it is like apples are dreap, and the domai.n of the superlative is provided hy the time element within the clause, ie 'cheaper in autumn than at other times •. Even where the superlative is diiptical, the presupposed group may still be \\.>ithin the dause. This happens only in equative clauses of the identify- ing type (those which are reversible, eg: apples are the cheapest~ the cheapest are apples). which arc probably the most frequent environment for ellip- tical superlatives. So in {4:45-a and b) we get two quite different notions of the qualities of the clown: [4:45] a. That down .is the finest fve ever seen. b. They are fine actors. That down is the finest r ve ever seen. In [4:4.sa] we assume that the presupposed item is clown, so although the }mestis. elliptic:at the presupposition is within the clause. In [4: 45b], on the other hand, the finest presupposes- actor from the preceding sentence. More accurately. [4:45b} is ambiguous;_it may mean 'the finest actor' or just 'the finest down'. Like an ordinal, a superlative presupposes some item that is semantically plural {more than one sct~ which as usual may be expressed in one nominal group or by a coordination); with the diffecence that, in the case of the superlative, this may also take the form of a mass noun. with the interpretation 'the ... -est kind of" • as in [4-:46} 'I told you butter wouldn't suit the works.' - •It was the best buttet.' As long as the clause is equative and the Subject is a common noun, an elliptical superlative as Complement will alway~ be ambiguous in this way. Other'"'rise, there is no ambiguity. If it is not equative. the superlative must refer to a preceding clause, as in [4:47a]; and likewise if the Subject is a proper noun, as in [4:47h) which cannot be interpreted as 'th.e finest Smith': [4:47} a. They are fme actors. ]ones always gets hold of the finest. b. They are .fine acton. Smith is the finest I've ever seen. As would be expected, a superlative repudiates all Numerative~ includ- ing cardinals, in the presupposed group. It may itself be si.ngnlar or plural. and if plural may be preceded by its own cardinal as in the three youngest.

r6:5 Comparatives are rather different from superlatives. Comparative adjec- tives are inherently presupposing by reference; this has been discussed above in 2.5. There must be a standard of comparison: anything that is bigger is bigger 'than' something else (which may be than itself under other circumstances). There are two specified sets involved. whereas with the superlative there is only one. This presupposition is not, however, an instance of ellipsis. One use of the comparative FORM of the adjective, always with deictic the, is actually semantically superlative: this is that in which the sense is 'the .• , -est of two'. An equarive clause having this type of comparative in complement position, such as Smith is the better actor, is of the IDENTIFY- ING type; and if the comparative functions as Head, as it does in [4:48a), then it is elliptical, just as a superlative would be (if [4:4-4al). The true comparative, however, does not take the, and an equative clause such as [4:48b] is not identifying but attributive. [4:48] a. Maty is the cleverer. b. Mary is cleverer. [4:48b] is not an elliptical clause. It presupposes by reference, but not by dlipsis; it cannot be "filled out' by a noun Head or a noun substitute. The structure, in fact, is that of(4:44h]. In other words, the three clawes apples tn'£ t:hetp. apples are cheaper r than pears') and apples are cheapest ('in autumn') are all attributive clauses,. and the nominal groups which func- tion as Attribute, those consisting just of an adjective (cheap, cheaper. chupest), are not elliptical fo:rtru. This function- that of Attribute in the clause - is. the only one in which an Epithet occurring as Head is not ellipticaL A nominal group with a true comparative as Head is, however,elliptical under alJ other conditions; for example [4:49] I'll buy you some prettier. These are less common than elliptical superlatives. but more common than elliptical uses of the ;adjective in a non-compared form (see below). The presupposed nominal group ma.y he count singular. COWlt plural or mass. There is one use of the true comparative which is confusing bccawe it :is preceded by tk and looks: like a superlative. This is as in [4: soJ The smaller the dog, the louder the bark. Here the .is not a Dcictic but a Submodifier with the sense of • by how much~. 'by that much' - originally not the definite article hut the

166 lii..LIPSrS instrumental case of that. There may be ambiguity between this type .and [4:48a], as in [4:51] Mrs Jones always uses Bliss. Her clothes are the whiter. - ~the whiter fOr it', or • the whjter of the two •? We have exemplified superlatives and comparatives only in the inflected forms -est and -er, but they may also of course be expressed by more . .. and most •... Everything that has been said applies to there fonns also. Notice that they differ from more and most as indefinite quantifters (4.2.3 .4); ambiguity may arise between the two, but only in the fu]l form of the nominal group, not when they are elliptical. Superlatives and, less often, comparatives may presuppose exophori- cally, as in JCIU take the biggest ('of the things in front of you'). Examples of idiomatic uses are the latest, in the special sense of' news' or • fashion •, and survival of the fittest. With comparatives we find mostly the superlative type with the, in the sense of~those who are ... -er than other people'. eg: the weakeT. finally. other items functioning as Epithet -that is, adjectives that are neither :superlative nor comparative- do not very ofien occur as Head in ellipsis. although colour words. which are anomalous in various ways. form something of an exception. Of the following. only [4: sza] is eUip- ticaJ; in [4; s:zb) grem is a noun: [4:52] a. The green suits you very welL b. Green suits you very well. So in (a) we could have the green mw, with substitution instead of ellipsis. In fact, this is the more usual pattern; substitution tends to be preferred to ellipsis wherever the presupposing nominal group contains an Epithet, or Classifier. Ellipsis occurs in an example such as {4: 53) I like strong tea. I suppose weak is better for you. But, as already noted. where the Epithet is functioning as an Attribute in the clause, it always appears as Head of the nominal group; in apples m-e cheap, cheap is not elliptical. Items like the rich, the long tmd the short of it. on the other hand, are dliptical, hut they are exophoric, and so do not contri- bute to cohesion. In its cohesive fimction, nominal ellipsis is largely con- fmed to instances where the presupposing clement is a Deictic or N umera- rive; in other v.rords. where it is one of the closed. system elements in the nominal group. Lexical elements usually require to be .accompanied by the substitute tme(5).

1.67 4-3 Verbal ellipsis 4·3-1 Ellipsis within the verbal group By VERBAL ELLIPSIS WC mean ellipsis wlth.in the verbal group. For example ln r4:54) a. Have you been swimming?- Yes, I have. b. What have you been doing?- Swimming. the two verbal groups in the answers, luJve (in yes I hatle) in (a) and swim- ming in (b), are both instances of verbal eHipsis. Both can be said to •stand for' have been swimming, and there is no possibility of • filling out' with any other items. So, for example, swimming in (b) could not be interpreted as 1 will be swimming or they are swimming. It could be interpreted only as I 1uwe bel!ff swimming; and it could, furthermore, be REPLACED .BY I h<Jve been swimming, since :!S in all types of ellipsis.. the full form and the elliptica1 one are both possible. An elliptical verbal group presupposes one or more words from a pre- vious verbal group. Technically~ it is defined as a verbal group whose structure does not fully express its systemic features - a1l the choices that are being made within the verbal group systems. The elliptical form swimming in {4:54b] has the features JOOS:ITIVE(as opposed to negative). FINITE (as opposed to non-ftnite) and ACTIVE (as opposed to passive). as well aS those of a particular tense, PRESENT IN PAST IN PRESENT; but none of these selections is shown in its. own structure. They have to be recovered by presupposition. A verbal group whose structure fu11y repre- sents all .its systemic features is not elliptical. This definition shows how verbal dlipsis differs from nominal ellipsis. In the verbal group, there is only one lexical element~ and that is the verb itself: su1im in [4: 54] above. The whole of the rest of the verbal group expresses systemic selections, choices of an either-or type (though not al- ways restricted. to two possibilities) which must be made whenever aver- bal group is wed. The principal systems arc: {I) Finiteness: finite or non-finite if finite: indicative or imperative if indicative: modal or non-modal (2) Polarity: positive or negative, and marked or unmarked (3) Voice: active or passive (4) Tense: past or present or future (recursively) These selections are obligatoty for all verbal groups. There is one other system. that of' Contrast: contrastive or non-contrastive', which appears

1:68 in spoken English only. since it is expressed by intonation. It is sometimes given partial expression in the written language by means of italics or other forms of typographical prominence eg: You WOULD do that! We shall not deai separately w:ith it as a verbal system; but :m me reference is made to cohesion by intonation in 5·9· Taken all together, the words that go to make up any non-elliptical verbal group, such as have been swimming. express all the features that have been selected.. In this instance it is fmite, indicative, non-modal, positive, active, and 'present in past in present'. But there is no direct conespondence be- tween the words and the features. We cannot pick out one word express- ing voice. another for tense and so on. The selections are expressed as a whole by the wards that are used and by their arrangement in a particular structure. Ellipsis in the nominal group was not described in this way, because the nominal group is not made up, to anything like the same extent, of gram- matical systems. It contains many more open choice (lexical) items.. Actually there is no difference in principle; the same theoretical definition of ellipsis i:s. valid for the nominal group also. But it would be much more complex to describe nominal ellipsis in terms of systems; so it was pre- sented in structural terms instead. For the verbal group, on the other hand, it is the system that provides the simplest way of explaining the facts of elli.p;;is, .and so the systems listed above have been used as the basis for organizing the present section. Being able to give a theoretical deftnition in these terms does not mean, however, that for every instance of .a verbal group we can always recog- nize whether it is elliptical or not just by looking at it. This is because, as we have already pointed out, the structure of the verba1 group does not represent its meaning in a Jirect and obvious way. In the firn: place, although all verbal groups express tense. voice, etc, we cannot identify each of these with a particular word or other element in the structure of the verbal group. Consider for example the verbal group has been seen. This ls finite, indicative, non-modal. positive~ passive, past in present. The feamres 'finite: indicative' arc expressed by the fact that the first word have is in the finite form has; 'non-modal' by the absence of a modal ele- ment; • positive • by the absence of a negative element~ • passive • by the word be in next to last place plus the fact that the verb see is in the passive participle fonn seen; • past in ... ' by the word have plus the fact that the next word be is in the past participle form been; and· ... in present' by the fact that the first word have is in the present tense form has. The whole thing is quite stWghtforward, although the details appear complex.

4• 3 VERBAL ELLIPSIS 169 In the second place, however. the fonru themselves are often multi- valent and even the whole verbal group may be ambiguous. Thus has is always finite and present; but have may he EITHER finite present OR non- finite, and so have been seen is ambiguous - it might he a. non-finite verbal group. The form saw (past finite) is distinct from seen (past or passive participle}; but in most verbs these are the same, eg: heard, tttadR, and all regular weak verbs. And have, be, and do occur BOTH as. realizations of the grammatical features of tense, voice, etc, AND as lexical verbs in their own right; in {4: 55), has, is .md does are grammatical in (r) and lexical in (z): [4:55] (I) (>) a. John has caught a co1d. a. Jolm has a cold. b. Mary is looking pretty. b. Maryis pretty. c. D<;,es John work well? c. John does his work well. So a1though the verbal group in English is extremely regular it is also faidy complex. It embodies a large number of systemic choices, especially those of tense, and it expresses these in ways which are not readily acces- sible to any kind of automatic recognition procedure. ln general, we can- not say just by looking at a verbal group whether it is elliptical or not. as we usually can with a nominal group; it is often necessary to consult the • eo-text' in order to find out. For example. each of the forms taking, has been and may have might be elliptical. or they might not. In [4:56]. they are non-elliptical in (r) but elliptical in {2) (with non-elliptical equivalents in square brackets): (4=56] (I) a, Taking photographs is a waste of time. b. Jane was secretary once. but I don't think Mary ever has been. c. Has he a car ? He may have. (2) a. What is he doing? Taking photographs. [is taking] b. Jane should have been told, but I don • t think she has been. [has been told] c. Has he seen it? He may have. (may have seen] But this is merely another aspect of what we have been stressing all along: cohesion is a feature of texts, and the question whether a particular instance is a cohesive form or not can often be settled only by reference to its textual environment.* * The description of the verbal group on which this section is based will be fOund in M. A. K. Halliday. Thr English vfflu:l grtntp ( •915 5. mim.eog~). An aco:nmt of it can be fuund in Geoffrey J Turner and Bemard A. Mo~ A lirtgNim£ dmripWn :md romputn progr11m for d.ilhm's 5puch. London. Routledge &: Kega:n h.ul. WlO (Ciupter 6}.

I70 4-3-2 Lexical ellipsis Nevertheless it may be helpful to approach the discussion of verbal ellipsis through a consideration of those instances where we CAN recognize that a particular verbal group is elliptical simply by inspecting its form. If we hear only the following sentence in a conversation [4: 57] h may or it may not. we know that the verbal items may and may not must be ellipticaL At least one word must he added following either of them in order to 4 fill out' the verbal group. The word tnil}' is a VERBAL OP.ERATOlt expressing "finite: indicative: modal~. It has no other functions, and cannot be a LEXICAL VERB. Hence may and may not have no lexical verb in them, and this is -sufftcient evidence to show that they are elliptical Any verbal group not containing a le:xical verb is ellipticaL (Note that the term 'lexical verb' includes the verbal substitute do discussed in 3-3 ~hove.} This enables us to identify one of the two types of verbal ellipsis, the one which we shall refer to as LEXICAL ELLIPSIS. It is the type of dlipsis in which the lexical verb is missing .from the verbal group. The other type is OPERATOR ELLIPSIS, described below in 4·3·3· All the modal operatoc-s can, could, will, would, shall, should, may, might, must. ought tc, .and is to .are alike in tha.t none of them can function as a lexical verb. (Here is to stands for all the funns am to, is to, are to, was to, were to; since this is a modal operator. it has no non-finite forms and no further variation in tense. There are two other modal operators, netd and dttTe; but they can also he used as lexical verbs. We ignore the special case of will=' bring about by willpower'. as in to will one's own destruction - as well as, of coUNe, will in to will one's fortune and r:an in to can fmit!) So any verbal group consisting of a. modal operator only can immediately be recognized as diiptic.al. Examples: {4: 58] Is John going to come?- He might. He was to, but he may not. - He should. if he wants his name to be considered. Here might, was to. may not and should are aU eUlptical verbal groups con- sisting of modal operator only~ each one of them could be filled out by the lexical verb come, or by the verbal :substitute do. The modal operators are always finite, and hence always occur in first position in the verbal group. There are other verbal operators, expresslng not modality but teme, which may be finite or non-finite~ any verbal group which ends in one of these is also elliptical, but here the situation is

4-3 VERBAL ELLIPSIS I7I 1ess clear because some of the items fWlctioning as temporaJ operators can be lexical verbs as well. The fmite temporal forms will, would, shall, slwuld (all these are temporal a~ wdl as modal) .and used to are ambiguous, like the modals; and so .are the non-finite equivalents of will, namdy (be) going to and (be} about to. All these are oper:ator::s; a verbal group which ends in any one of them kcb a Jexicaf verb and is therefore elJiptical. But be and have, which occur as temporal operators in an their forms. both finite and non-finite. function ALSO as lexical verbs; so no simple rule can be given to say that a verbal group ending in a form of be or have is eihptical- it may or may not be. The same applies to the one remaining verbal operator (finite only) do, which is the carrier of negative and marked positive polarity in simple present and past tense; in tkrs see, Jiti see, etc, de is :m operator, but do can also be a lexical verb (see above, 3·3-3·1) as weU as being the verbal substitute. To give some further examples. the verbal groups may be, are going to have and diJ(ifJ.J.J.6, [3'89-91D are non-elliptical in [4'59 (r)] but elliptical in [4,59(>)]' [4'59] (r) (non-elliptical)' a. He seems quite intelligent.- He may be. I agree. b. fve decided to leave.- I hope you~ re going to have second thoughts. c. Did Jane know?- No, but M>rydid. (z) (elliptical), a. Is he complaining? - He may be~ I don•t care. b. I haven•t finished it yet.- [hope you're going to have by tomorrow. c. Did Jane know? - Yes, she did. Here the distinction between elliptical and non-elliptical forms has to be recovered from the presuppored clause. The lexical verbs be and hav.e al- "\'\.'a}'S require a Complement. With all other verbs, there is a ge-neral rnle whereby if a Complement is omitted (by clausal dlipsis) then the lexical verb must also be either omitted or substituted. But this does not apply to be and have; these verbs may OCGUC with ellipsis of the Complement. as in ks9 (I>)] and [4'56 (rh and c)], the verbal grouP' then=lves being non- dliptical. Hence all that can be said is that if there is no Complement fol- lowing be or have there must be SOME ellipsis; hut it may be EITHER verbal ellipsis. with be, havt: as. operator, oa clausal ellipsis (c/ 4·4 below), with be, havtt as lexical vecb, and in order to determine which. it is necessaty to refer

1']2 :BLLTPSIS to the presupposed clause. If there is a Complement, and the verbal group ends in be or hav~. then there may be BtlHER verbal ellipsis~ with be,luwe as operator., OR no ellipsis at all, and be, hav~ as lexical verb; the clause in question is often ~mhiguous hy itself. eg: he h4S some cf the paintings in k6oa] and sbe is the do<torin [4:6ob]: [4:6o] (1) (non-elliptic.!): a. Has he .all these items in his own collection? - He has some of the paintings; fm not sure a bout the rest. b. She ought to know what to do. She is the doctor. (2) (elliptical): a. Has he sold his collec- tion yet? - He has some of the paintings; I'm not sure about the rest. b. Is she suing the hospital? - She is the doctor. As far as do is concerned. the lexical verb do also usually requires a Com- plement~ except in the special sense of'hc saris:fa.ctory', eg: will it do? The substitute do, however. does not; and a verbal group such as did in [4: 59C] may be non-elliptical, with substitute tic, as in (r), or elliptical, with opera- tor®, as in (2). The difference is shown by the fact that the non-presup- posing form of (1) would be Mary knew, with did RE!'LACED by knew, whereas that of {2) would he she did know, with the dliptical form did FILLED OUT by the lexical verb know. But since the verb..l operator do occutS as a finite form only, and hence comes first in the verbal group~ whereas the substitute do is a substitute for the lexical verb, and hence comes last, such instances of overlap can occur only with a verbal group consisting of just the one word, de. does, or did. With Jo the negative forms are unambiguous, since only the operator do has don't. J.oem' t and mdn' t as its negative forms. This is because the operator do is in fact simply a • carrier' of the expression of polarity: nega- tive (eg: didn't Me) and marked positive (eg: did s-u). the latter being the fonn used in interrogative dauses (did ycu s-ee John?, not saw you John?). So any verbal group consisting only of t~on•t. d~sn't or didn't must he ·elliptical,. and likewise any group consisting soldy of J(), does or did pre- -ceding the subject in an interrogative clause., eg: diJ you? The lexical verb do fOrms its negative like other lexical verbs: so we say hed~'t Jo his work properly. not he doesn't his work properly. Lexical be and lutw. on the other hand, form their- negatives like verh41 operators; hence Ire im't, is he?, he ~rasn•t. has hc-J may be either elliptical (operator only) or non-elliptical (lexical verb). There is ,a rider to this; there are actually two

4·3 VERBAL ELLIPSJS 173 distinct lexical items have, one meaning 'possess' and expandahle mto have got, the other meaning 'take' and not expandable. Only the former has the negative hasn't (eg: he hasn't any money). and that not in all dialects; the latter forms its negative like other lexical verbs, by tneans of the opera- tor do, as in he Joesn' t have breakfast. Apart from this exception the nega- tive formr. isn't. hasn't, etc may be either operator or iexica1 verbs, and hence one cannot say that a verbal group consisting of one of these forms alone is definitely elliptical: in [4:;55] they could occur in either column {1) or column (2). Finally there is the form to. We have seen that this occurs as part of the operators going lo, about to. used!{}, is to and ought to; and a verbal group in which to occurs finally, not followed by a lexical item, is bound to be elliptical (c/ 3.3;.2, (3: 70]). This applies also to a vcrbal gmupconsisting only of the word to, as a marker of the infinitive {that is, of the perfective form of the non-finite verb, ;g see, to have seen, etc); for example f4:6rl I'd better see him. I don't really want to. In what we are calling LEXICAL ELUPSlS, it is the lexical verb that is always omitted. Other words in the verbal group may also be omitted, v.-:ith the exception of whatever word is in first position- the finite opr:-r- ator if finite, and to or an -ing form if non-finite. So we may have, in answer to John should luwe been coming n•ery Jay: [ 4: 62 J ( 1) non-elliptical a. finite: r don't think he- . - has been coming b. nM-.finite, peifedive: At least I under- to have been coming stand him ... c. mm-finitf', impt'ifective: (2) elliptical has been has to have been to have to 1 think he rather having been coming having been regrets . . . having Lexical ellipsis is ellipsis 'from the right': it always involves omission of the last word, which is the lexical ve-rb, and may extend 'leftward', to leave only the first word intact. So for has been ccming we may fmd has been or simply has. With a very long verbal group there would be more possibilities: could have been going to be consulwd might he reduced, by lexical ellipsis, to could have b«n gfJing to be, could have bet>n going to, could

I74 have been. could have or simply rould. Usually the • outer~ forms .are pre- ferred: that whic:h is minimally dliptical with ONLY the lexical verb omit- ted,. or that with everything omitted that can be presupposed from the context. So following wasn't John going to be consult~? we would most probably find either he could have been going W be or he could have hem. But intermediate fOrms also occur. Note that the extent of the presupposition is not affected by these variations. Thus in [4:62] the elliptical fo:rms pre- suppose all the tense selections as well as the lexical verb: the form has. in (:u)~ stands for has been coming and not has come. In general • .any sekctions that are not explicidy repudiated are automatically presupposed. A very clear example oflexical ellipsis is provided by question tags. Al1 question tags have maximum lexical ellipsis and presuppose all the features of rhe relevant verbal group; so [4:63) a. John couldn't have been going to be consulted, could he? b. Mary didn't know, did she 1 c. They'll have been working on it aU night, won't they? The presupposition of particular systemic features is discussed in more de- tail in 4-3-4-I-4 below. 4·3·3 Operator ellipsis There is another type of ve:rbal ellipsis. which is ellipsis • from the left •. We shall refer to this as 'OPERA TOR ELLIPSIS' • since it involves only the omission of operators: the lexical verb always remains intact. Example [4: 54] showed the difference between the two: [4: 543) is lexical ellipsis. [4: 54bJ is operator ellipsis. In operator eUipsis the Subject also is always omitted from the clause; it must therefore be presupposed. One type of operator ellipsis. which is very frequent. v.-ill not concern us here, since it does not contribute to cohesion: this is operatO£ ellipsis within the sentence, in the context of coordination. In this type it is pos- sible to introduce a new Subject, as in [4: 64-c). So for example in [4:64} a. They must have been both watching and being watched. b. Afier we've brought them out so far~ made them trot so quick. c. Some were laughing and others crying. the verbal forms must luwe lu>en (bmh watching mu! being watched). have (brought . .. vut . .. and made •.• ) and were (laughing and .. . crying) are • branched': the operators arc structurally related to both halves of the

4•3 VERBAL llLLIPSIS I75 coordination, as indicated by the bracketing. Note that this also happens in lexical ellipsis, with one lexical verb being related to two or more co- ordinate operators; the most usual form of this is the coordination of positive and negative operator with or~ as in {4:65] They might or might not have objected, Verbal coordination of this type, however, accounts for a relatively small proportion of the total incidence of lexical ellipsis, which is more often BETWlffiN sentences (and therefOre cohesive in our sense). Operator ellipsis, when it occurs across sentences, is fonnd mainly in very closely bonded sequences such as question and answer, in which the lexical verb either supplies the answer to 'do what?•, as in {4-:54b], or repudiates the verb in the question, as in (4:66] Ha, ;he been ctying?- No, langhing. In most instances of operator ellipsis,. everything is presupposed but the lexical verb- that is. the entire selection within the systems of tense, voice, polarity and so on; and all words except the last are omitted. Occasionally it is: the voice, the choice of active or passive, that is being repudiate<L in which case if the elliptical group is passive the be immediatdy preceding the lexical verb must also be present, since it is part of the realization of the selection of passive; for example [4:67] What have you been doing? -Being chared by a bull. Operator ellipsis is fairly easy to recognize, with the provisos made in 4--J.I above, since there is no finjre dement in the eUiptical group. There are two sources of uncertainty, and these have to be resolved by reference to the surrounding text. One is that in most verbs -the past tense and the past or passive participle have the same form, so that an item like made in [4:64 b] taken on its own could be a simple past tense instead of being elliptical for have maJe. The other prob1em is that a FINITB verbal group WITH operator ellipsis ls identical with a NON-FINITE verbal group that is NOT elliptical, eg: being watched, made, singing, being chased, so that one has to ask whether the context demands a verbal group that is non-finite or one that is finite. But thls is not usually much of a problem, and it is further simplified by the fact that the perfective form of the non-finite verbal group near1y always has to at the beginning. The only point to note is that a non-finite verbal group may itselfhave operator ellipsis, either hy simple omission of IQ or • if it is marked for tense, by omission of the tense opera- tor (or operators), eg

1']6 [4:68] (>) non-elliptical: a. What do you want to do? -To go out to the pictures. b. Why was he so angry when the game was stopped? Because of h-aving been losing?- No~ having been winning ! As a final example, in (2) elliptical: Go out to the pictures. No; winning! {4:69] What must I do next?- Play your highest card. it is not very dear whether play is an elliptical indicative. for you must pl4y, or a non-elliptical imper-ative; nor does it seem to matter very much. It is probably the former: the tag would be shouldn't you? ,and we would prob- ably find the same form foUowing he, where it could not be imperative: What must he do next? - Play his highest cruJ.. But the difference in meaning is so slight that it is difficult to sense the ambiguity between the two. 4.J.4 Presupposition cif verbol group .symms We will consider in turn the various systems of the verbal group. asking whether, and under what circumstances, they are liable to presupposition in cases of ellipsis: whether, that iJ.I, the meaning is carried over when no selection from the system is expressed in the structure. We shall refer to polarity, fmiteness, modality, voice, and tense; with a very brief mention of the system of conttas:tiveness that is found only in the spoken language. 4.3..(.1 POLAIUTY Polarity is normally expressed at the beginning of the verbal group. A negative verbal group, if it is finite, hasn't or fUJt attached to the first word, eg: didn't know. did twt know. If it is non-6nite, it has not, usually as the first word, eg: not ha.ving known. not to have known, although the not may some- times follow the first verbal operator, eg: having not known, to have not krwwn. Other negative adverbs such as never, hanlly, luudly ever, may oc;cur ln place of tWl. The category of negative is not very sharply defined~ hut it is revealed by the choice of tag. There is a semantic parallel between he-'s here. isn't he? and he isn't hm?, is he-?; and the fact that the corresponding form with hardly ever is he's h4rdly filer here, is he?, with positive tag, shows that lurrdly ever is really a negative form. In lexical ellipsis, wh-atever else is omitted the frrst operator is always present. This means that the polarity will always in fi.ct he expressed, and

4• 3 V:BR.BAL BLLlPSIS 177 the question of what happens if it is omitted does not arise. This is a result of the structure, although it is not simply a gnm1112tical accident. The characteristic function of ellipsis is that of cohesion by presupposition, and there is a large class of cohesive sequences in which the one thing that can- not be presupposed is polarity: namdy those where the response (eg: to a yes/no question) serves precisely to sUPPLY the polarity. all else being taken for granted. For example [,;: 70j a. Were you laughing?- No, I wasn't. b. Cats like cheese.- They don't, do they?- Yes. they do.- Well,. some do and some don't. This makes it easy to understand the general principle whereby. what-- evet else may be presupposed in verbal (lexieal) ellipsis, the polarity has ro be made explicit. The principle .applies to both finite and non-finite verbal groups; and in the non-finite {perfective} there is. a special elliptical form of the negative, namdy not to, as in r d hate not to. not to would be silly. whieb expr=es simply the uon-finiren= and the pobrity, and nothing else. There is an equivalent positive form to which is, however, much m.Qte restricted.: we say I'd love to but we do not say to would br silly (cf 3.3.2. [3 :70<, d and eD. We should distinguish here the special type of negation in which the negative is attached specifically to some other element in the verbal group .,m {4:7IJ a. fve kept on tdephnning, hut they've simply been. not answenng. b. He says be's been not being informed about these devdop- ments. Here the verbal group itself is positive. and. certain items or features in it are explicitly negated: the lexical verb answer in (a). the lexical verb itifcms and the passive voice selection in (h). In such instances there is uncertainty about whether the polarity is presupposed with lexical ellipsis. Con- ceivably the response H• has to [4:71b] might be used to express agree- me~ ie 'he has been not being informed •. But there is a strong tendency in English for the polarity of the verbal group itself- that is. as expressed in first position- to determine the polarity of the whole, so that even here the negative is unlikely to he presupposed: one would expect rather (Quite right.) He hasn't in the sense of •1 agree' and (Oh yes.) He h~ with con- trastive intonation, in the sense oCI disagree •.

I78 Tuming now to operator ellipsis, we might expect that here. where among the items omitted is always the one which carries polarity. the polarity would naturally be presupposed by the elliptical verbal group. In fact as a rule it is not. although the reasons are different. As we have seen, operator dlipsis is ch-aracteristic of responses which are closely tied to a preceding question or statement, and which have the specific fi.mction of supplying, confirming or repudiating a lexical verb. The foJiowing is a typical sequence, illustrating how the polarity is restated each time: [4:72] A: What are you doing? (positive) B: Thinking. (positive; ·rm .. .') N da dr ( • ?') A: or y eanung. neganve; arentyou ... s: No, thinking. (positive; 'I'm ... ') If there is a change in polarity. this may go in either direction, from posi- tive to negative or from negative to positive; note that the final occurrence of thinking in [4:72] does not take over the selection of negative from the presupposed group. One typical context for a verbal group with operator ellipsis is as a response to a WH-question with the interrogative on the verb, such as J¥hat an yGU tWing? This is a demand fur a lexical verb, and the normal response is simply to supply the verb, everything else being omitted. Here it might be said that the polarity is presupposed. But there is re.olly no way of testing tbis statement. since the verbal group in the question is bound to be positive; one does not ask What aren't yPU d&ing? (except in the special instance of an echo question, where the polarity clearly IS presupposed in the response. as in Smith im't cooperating. -What isn't he dtling? - Cooperating). The other most usual context is that -of a yes/no question, and this is precisdy a demand for the polarity to he supplied; the polarity cannot therefore be presupposed. The words yes and no arc purely iodicaton of polmty, and they ace regularly dliptical for the whole of the presupposed clause (sec below, 4-4·3). But the speaker may repeat the lexical verb. in order to deny it or explicitly to afftrm it; in this case the polarity is always restated and, interestingly. operator ellipsis is possible only if the polarity is explicitly expressed - that is, if the answer is: negative (since the negative requires to be stated by not, with or without a preceding no) or. if it is positive, provided. it is introduced by yes: [4:73] (1) Weren•t you complaining?- (No,) Not complaining. (2) Were you complaining?- Yes, complaining. In (2) the answer could not be simply Complaining. If however the answer

4• 3 V.R:R.BAL ELLIPSIS 1.79 is a contradiction. involving a change of polarity, the elliptical form is much less likelyo (b) would be preferred to (a) in both [4073 (3) and (4)]o [4o73] (3) Weren't you com- plaining?- (4) Were you com- plaining?- a. Yes, complaining. b. {Yes,) I was complaining. a. (No,) Not complaining. b. (No,) I wasn't complaining. Occasionally a yes/no question may he answered (or, more accurately, responded to; such a response is not an answer) with a different lexical verb, and here, predictably, no ellipsis is possible: [4'74] Ao Were you thinking? B: I wasn't daydreaming, if that's what you mean. B• s response could not take the form Not daydreaming. Hence in verbal ellipsis of any kind the elllptical verbal group makes a new sdecrion in the .system of polarity: polarity is not included in what is presupposed. In lexical ellipsis, this is because the one clement that must be present, whatever else is omitted, is the initial element, and this is the one that carries the expression of polarity. In the case of operator ellipsis. the reason is semantic rather than grammatical; the expression of polarity is not required by the structure. but operator dlipsis is largdy restricted to responses in which either the polarity can only be positive (and the ques- tion of presupposition does not arise) or else it is precisely the information 'yes or no?' that is being asked fur, in which case it cannot possibly be presupposed in the answer. A conslderation of marked polarity {if 4·3-I above} would take us into too much detail. but it needs a brief mention to conclude this section. What is meant by MARKED F-OLARITY is the assignment of special p~ minence to the selection of positive or negative in order to draw attention to it. In the finite verbal group this is realized by the use of non-reduced forms of the finite operator or (where Ielevant) the negative: is, had, was, can, slMU, siwuld. etc instead of the reduced forms's, 'd~ 'll, [w.n:] for was, [Jdj for shculd (not distinct in writing). etc, not instead of n't, and also does see, did see instead of sees, saw. {Note that the non-reduced forms are NOT necessarily TONIC ('primary stress'}, though they must be SALIENT ('secondary stress').) In a verbal group with operator ellipsis, therefore~ it is impossible to express nurked pclarity; even if the presupposed item has it, as in [4: 75]. where doing is tonic and is is salient, :it tends merely to express the questioner's attitude, impatience or something of the son:

I8o [4:75] What is he DOING all this time?- Reading, probably. {//I what/ is he j DOING I all this I time If! -in any case it cannot he carried over. A verbal group with lexical ellipsis. on the other hand. must have the polarity marked; so the finite operator cannot he reduced: [+:76] a. Who'll put down five pounds?- I wilL (twt I1l) b. John's arrived, has he?- Not yet; but Mary has. (not Mary's) This applies to all positive forms. The negative may or may not he re- duced; we could have I w""'tm [4:J6a], and Maty ham't in [4:76b]- this is no doubt because tbe negative is itself a kind of marked polarity. The preference for marked forms of polarity in this type of verbal ellipsis is probably to be explained by the &et that so often in sequences of thi:. kind it is the expression of polarity that is the whole point of the response. 4-3-4.2 RNITENESS AND MODAI.JTY The Systems of FINITBNBSS and MODALITY are also closely associated with first position in the verbal group. and this largely determines the possibilities of their presupposition by mearu of ellipsis. A verbal group which is finite always expresses its finiteness in the first word. Either the gronp consists just of a finite funn of thC lexical verb. present or past (walk. wtJiks j walked)~ or it begins with a finite verbal operator; the latter is either a tense operator: (1} am, is. are; was. were [i.e finite forms of be] (2) have, has; had [ie fmite formsofh.we] (l) do, does; did (4) shall, will (5) used (to) or a modal operator: (6) shall, will, should, would, can, could, may, migh~ must, ought (to) (7) am to, is to, are to; was to, were to (j.e finite forms of be, plus to'] (8) need, dare \m one use) Any verbal group which does not have a finite form as its first word is automatically non-finite. A verbal group consisting just of the base form of the verb. eg: walk, is therefore ambiguous: it may he finite {present tense. eg: I walk)~ or non-finite (perfective. eg: made me w.Jk). But the dis- tinction is always clear in the context; moreover the non-finite perfective

.f. 3 VBllB-AL I!LLIPSlS I8I nearly always has to before it (eg: wanted me to walk). The imperative form walk has something of the finite and something of the non-finite about it. but is best treated as a finite vcrbal furm. Verbal groups with operators are never ambiguous as to finiteness. It is true that have and d() are ambiguous by themselves; but do occurs: as opera- tor ONLY in finite vecbal groups, while hav<' in a non-finite group is ALWAYS preceded by/(). In lexical ellipsis. as we have seen, the ellipsis is ~from the right' and the one element that is never omitted is the finite operator. So, as with polarity, there is no question of what happens if the finiteness is not expressed; it always is. A verbal group that is lexically dli:ptical is always explicitly either finite or non-finite. It cannot simply take over the selection made by the verbal group which it presupposes. There is no restriction of the presupposition of a finite verbal group by a non-finite or vice versa. We may have aU possible sequences: [4:77] a. [finite presupposed by finite] The picture wasn't finished. If it had been, I would have brought it. b. [finite prempposed by non-finiee] He's always being teased about it. I don't think he likes being. c. [non-finite presupposed by finite] What was the point of having invited all those people? - [ didn't; they just came. d. [non-finite followed by non-finite} It was hard work parcelling all those hooks. - I' m sure it was; and r d much prefer you not to have. With operator ellipsis, the situation is exactly reversed; here the first word MUST be omitted, whatever else is or is not present. and so the ellip- tical verbal group cannot express the choice between finite and non-finite. As Js to he expected, therefore. it takes over the selection from the pre- supposed group: [4:78) a. [finite: 'they are finishing'] What are they doing now?- Finishing their essays. b. (non-fmite: 'to be finishing'] What would you like them to be doing while you' re away? - Finishing their essays. A1) that has been said with ref§trd to finiteness applies equally to m~ dality. Modality (ie the choice between modal and non-modal, and, if

182 RLLIPSIS modal, among the various modal categories) is a subcategory of'finite'~ and is expressed by the presence or absence of a modal operator. In aver- bal group with lexical ellipsis, therefore, the modality is· always explicit. and there is no restriction on what may be presupposed by what: [4: 79] a. [modal P'esuppo:red by non-modal] I could hdp them.- Why don't you? b. [non-modal P'esupposed by modal] Are you going to tell her?- I ought to. c. [modal presupposed by same modal) He must have destroyed them. - Someone must have. cer- tainly. d. [modal p=uppored by dttli:rent modal] He must have destroyed them. -He may have. I suppose. In a verbal group with operator ellipsis, the modality is never explicit and, like the finiteness, is always carried over from the presupposed group: [4:80] a. [non-modal: [4:78a] 'they are finishing'] What are they doing now?- Finishing their essays. b. [modal: 'they will be finishing'] What will they be doing now, do you think ? -Finishing their essays. probably. 4·3·4·3 VOICE When we come to the system of VOICE (the choice between active and passive) the position is somewhat different. Voice is expressed towards the end. of the verbal group, by the presence ~ive) or absence (active} of some form of be or get just before a lexical verb, with the lexical verb in the passive participle form. Any verbal group displaying both these features is passive, eg: was stolen, has been robbed, being taken, get a"ested; all others are active, Therefore it does not follow automatically that an elliptical verbal group either will or will not contain an overt expres~ion of voice. Nevertheless in lexical ellipsis. the rule is quite clear; the voice selection is always presupposed. So although the examples in [4-: 81] make perfectly good. sense, they are impossible, because the e1liptical form here repudiates the voice of the presupposed verbal group. [4:8!] a. [active followed by passive: 'if it had been finished'] They haven't finished the p.icture. If it had been. I would have brought it.

4-3 VERBAL ELLIPSIS 183 b. [passive foUowed by active: 'if she does beat him'] Johnny hates being beaten at any game by his sister. If she does, he sulks. c. [active followed by passive: 'she has never been loved'] Mary could love very deeply. Unfortunately she never has been. d. [passive followed by active: • she has forgiven them"} She is forgiven, apparently. But I don"t think she has them. No doubt the reason these are unacceptable is that the second sentence in each case involves a change in the alignment of structural functions. Either the Subject changes. the Actor/Goal relationship remaining the same. as in (a) and (b); or the Actor.JGoal rdarionship changes, the Subject remaining the same, as in (c) and (d). In either instance, and even if one element is an unexpressed "someone" as in (c), we feel the proposition should be restated in full. The voice selection, in other words, cannot be repudiated by an dJiptical structure; and the mere fact that the lexical verb needing to be supplied is already in the right form, as in (a) finished and (d) forgiven, is not enough to override the rule that voice must be carried over. Presumably we fed little in common between has forgiven and is forgiven, even though the participle is formally the same. In operator ellipsis, .as we saw earlier, the Subject is always omitted; it must therefore be carried over by presupposition. This means that we cannot have a change of Subject fur the elliptical group; so in an example such as [4:82] Were Australia leading England at the time, then?- No. Eng- land were winning. we cannot replace the second sentence by the elliptical form No, England winning. In other words, here. as in lexical ellipsis, the voice selection must he presupposed if the presupposing group is elliptical; it cannot be repu- diated. But there is one condition under which the voice can be repudiated in operator ellipsis: namely if the Actor/Goal rdationship changes. leaving -the Sub~t unaltered. One example of this was given in [ 4: 67} above; others would he [4-:83) a. Will you be interviewing today?- No; being interviewed. b. John has loved Mary for a long time. - Or :at least been loved by her.

184 IILLIPSIS TBNSB Non-finite, and Fmite non- finite modal, modal tenses tenses ( I2) : (36): read as read as far as f1 far as a < « (none) rrt !! t' in{=ent !V in{=ent V in{~=nt VI {past V!! m pm vm in{:~ 2 3 4 5 6 7 8 9 10 II 12 13 14 lj 16 17 18 19 20 21 22 23 24

185 Finite non-modal tense Non-finite, and finite modal tenses: (perfective, imperfective; modal) ' took / did take 2 take(s) f do("') take to take, taking; can take J wiU take 4 had taken 5 has taken II to have, having; can have+taken 6 will have taken 7 was taking 8 is taking Ill to be. being; can be+ taking 9 will be taking ro was going to take n is going to take IV to be. being; can he+ going/about 12- will be going to take to take 1:3 was going to have taken Y4 is going to have taken V to- be, being; can be+ going to have I 5 will be going to have taken taken r6 had been taking 17 has been taking VI to have; having; can have+been 18 will have been taking 19 was going to be taking zo is going to be taking VII to be. being; can be+ going to be 21 will be going to he 22 had been going to take 23 has been going to take VIII to have, having; can have+ been 24 will have been going going to take to take

<86 4·3·4-4 TENSE IX X XI XII 25 26 27 29 30 3I m present 32 futme 33 34 m present 35 36 The English tense system is complex, though its complexity is more ap- parent than reaL It is based on two very simple principles: ( 1) that there is a choice of past, present and future. and (2) that this choice may be made repeatedly (within limits), each new choice taking the previous one as its point of departure. Both these principles apply whenever a verb is used, tmlcss that vcrb is in the imp er ativc, which has no choice of tense. So, for example, l may choose a future tense: I will play. But having thus shifted my standpoint into the future [ may then take this a5 a base for a funher point, say past; I then get the tense 'pas-t in future', which is I will lw.ve played. Supposing once again I take this as a base line and select, say. pr-esent: the tense is then 'present in past in future'. I will have been

25 had been going to have taken 26 has been going to have taken 27 VP-ill have been going to have taken 28 V.'aS going to have been 29 is going to have been 30 will be going to have 3 I had been going to be 32- has been going to be 3 3 will have been going to be taking 34 had been going to have 35 has been going to have 36 will have been going to have been taking IX to have. having; can have+ been going to have taken X to be. being; can be+go.ing to have been taking XI to have, having; can have+been going to be taking XII to have. having; can have+been going to have been taking playing. This can happen up to five times. subject to increasing restrictions which end up by precluding a sixth choice altogether. The most complex tense form in English is one like had been going to have been playing~ which is 'present in past in future In past in past'. It may be hdpfui here to list the fu11 set of finite and corresponding non-finlte tenses of the English verb. The column headed a is the PRIMARY TENSE {FIRST ORDER TENSE); it is always expressed by a finite form, and a verbal group with primary tense is always finite. The other colunms represent the SECONDARY TE.NSBS (SECOND ORDER, THIRD ORDE.II: and SO on); the LAST OllDBR TBNSB is always the one that appears earliest in the NAME of the tense. Thus •pre- sent in past in future' has primary (first order) tense future, and secondary tenses past and present, of which the last order tense is present.

I88 It will be seen that the non-finite forms. which are also those of the fmite verbal group if it is modalized, are equal to the finite (non-modal) forms minus the • alpha-• or primat:y tense choice. So, for example, non-finite having takm corresponds to all three of the finlte tenses took. has taken and had taken. A tense form embodying only one choice is a SIMPLE tense; hence '(simple) past' I tcok, '(simple) present' I t.ke, '(simple) future' I will take. All other tenses are COMPOUND. Like the tense system itself. the principles of the presupposition of tense selections in verbal ellipsis look rather complex at first sight; but actually they are fairly -simple. Let us consider the follo""A"ing instances of lexical ellipsis: [4: 841 a. I protest. - Do you? b. He usually talks all the time. He didn•t, yesterday. c. It doesn't tum. -It will if you press it in first. d. She won't agree. -She did last time. e. Is he arguing ? - Yes. he always does. f. Was he going to apologize? He won•t now. g. Has she heard aboqt it yet?- No, but she soon will. h. You have been forgetting every morning. Today you did agam. The choice of tenses in these examples is as follows: Elliptlcal group (a} present ~)~t pm (c) present (d) future {e) present in present (f) future in past (g) past m present (h) present in past in present In each case the ellipricai verbal gmup makes a simple tense choice, which is fully explicit in the operator (did, Jces. will); hence the only presup- position is of the lexical verb, which is to be supplied in its base form. No tense selection is carried over from the presupposed group. Now consider the case where the second verbal group, the one that is elliptical, is making a compound tense selection. Here are some acceptable examples:

..of-. 3 VERBAL ELLIPSIS r!J9 [4:85] a. At least Stan has tried. I don~t think Bob has. b. fm going home this weekend. I shall be every weekend now. c. Are you dieting ? - I have been for some time. d. He was going to build it hinuel£ He isn't any longer. e. She really has been working hard. - And she's going to be again before long. (a) past in present (h) present in present {c) present in present (d) future in past (e) present in past in present past in present present in future present in past in present future in present present in future in present The following however are much less acceptable: [4:86] a. Have you discussed it yet?- No, we are now. b. You've been forgetting every morning. Today you have agam. c. He was going to tell us. But he rtill hadn't, yesterday. (a) past in present (b) present in past in present (c) future in past present in present past in present past in past For these to become acceptable, the second verbal group would have to be filled out by the lexical verb or verbal substitute; (a) discussing it{Jcing (b) forgottetJ/Jcne (c) to/J usfdom: The principle seems dear. In compound tenses, the tense selection is not made clear by the finite verbal operator alone; other elements are needed. and the form of the lexical verb itself may change. If the tense in the ellip-- tical verbal group is a compound one, then it must be such that the lexical verb can be carried over m THB SAME FORM. So in [4: 8.5] the elliptical verbal group could in fact be filled out by the lexical verb with its form Wl- changeJ, (a) tried, (b) going, (c) dieting, (d) (gcing to) bui/J, (e) wCTking. Thi• means that the la~t-order tense, the one that is EXPRESSED last in the verbal group (though it appears first in the NAME of the tense), is carried over from the presupposed group. If this changes. then the form of the lexical verb changes, and the lexical verb must be repeated (or substituted). as in k86].

190 To summarize: a verbal group with lexical ellipsis must have either a tense that is fully explicit even in the elliptical form, or one in which the- lexica1 verb can be carried over tmchanged from the presupposed group. In other words. either it has simple part, present or future; or, if the tense is compound, it has the same last-order tense as the presupposed group. So, for example. if the presupposed group has •future in present·~ he was going to letwe, there can be lexical ellipsis in a foHowing verbal group pr~ vid.cd. that that verbal group has any simple tense, eg past but 1u: didn•t (if [4:84f]). or a compound tense which is also ·ru.rure in ... ·, eg future in present but he isn~t new (if[4:SsdD. It is not that all other instances are totally unacceptable; we might accept b:1t he ~uzm•t in this instance, and. the following also: [4:87} a. fm staying at home this weekend. I haven't for some time. b. It was going to snow. they said. "Why isn"t it? where (a} has past in present presupposing present in present. and (b) has: present in present presupposing future in past. But these are aU a litt1e awk- W<ml. and a more natural form is th.at with substitute do: I haverl t done for some time, why tsn~ t it doing? We have .illustrated tense in verbal ellipsis by reference to finite verbal groups; but the same principles apply to those which .are non-finite. including instances where. of the two verbal groups involved in the pre- supposition, one is £nite and the other non-finite. Here are some varied examples: [4:88} a. He shows no sign ofhaving been studying.- He hadn.'t,lham't/ wasn't. b. She intends to come.- She won~t. c. Will he give in to them?- He doesn•t seem to be going to. d. We seem to be being followed. - I remember having been when we were here before. Presuppo,.d group (a) non-finite: present in past (correspondmg to all three:) (h) non-finite : tenseless (c) finit., future (d) non-finite (pafective): present finite: present in past in past finite: present in past in present finite: present in past finite: future non-finite: future non-finite (imperfective): past Verb forms which include the selection of 'future • at any point will normally be Jonger, in terms of words. than the corresponding forms

.q.. 3 VERBAL l!LLIPSIS I9r with present or past; in the simple tense, past and present consist of one word only (took. takes), future of two (will take). The 'marked positive' forms {see 4-3.4- above) of past and present are did tak, does takej and the paradigm of simpie tenses in spolc_en English is actually quite symmetrical: P"'t positive tuunarked marked he took he did take he takes he does take he'll take he v.'ill take negative unmarked marked he didn't take he did not take he doesn't take he does not take he won· t take he will not take Hence as we have already seen the elliptical forms of the simple tenses are .all forms consisting of one word: he did, ke tbes, he will . .But the non-finjte form of the future is be going to or be alwut to~ this is the form in which it occurs anywhere other than as primary tense. This does not affect the principles stated above, but it makes it simpler to state them by reference to the tense SYSTEMS: that is~ in terms of the selection of tenses in the verbal group. rather than in terms of the w-orcb. that are used to express the tense selections. With operator dlipsis, which as we saw eadier is characteristicalJy asrociated with question-answer sequences where the question centres around the lexical verb, the elliptical group normally takes over the total tense selection of the group wftich is presupposed. So; [4o89) a. What is he going to do with all that paraphematia?- Catch fuh. [~He's going to catch fish] b. Have you been digging?- No, weeding. f =I've been weed- ing] c. What should she ha.vedone?- Told the police. [~She should have told the police] This type of ellipsis is very frequent. and the result looks like an ordinary non-finite verbal group. If the question is a simple present or past tense, there is no possibility of verbal ellipsis in the response. as the verbal group consists of only one word. It is possible for a verbal group in such contexts to repudiate some or all of the tense selection of the presupposed group. but this has to be done

I~ I!LLJ:PSIS explicitly - anything that is omitted through ellipsis will be earned over. So we undernand the response in [4: 90] [4:90) He must have mended it.- Or been going to mend it, :rather. as he must have been going to mend it. It is possible to construct ambiguous examples if one tries hard enough~ eg [4:91] [4:91] He could have been going to mend it-or be mending it. where the response might be either he could have been going W be mending it (at some particular time later, eg just when you arrived), or he could be mending it (now). But the general principle is the usual one that whatever is not specifically repudiated is presupposed by the eUiptical form. 4.3·5 s"""'""'l '!{verbal ellipsis We can now give a brief summary of lexical and operator ellipsis in the verbal group. Lexical eiHpsis, it will be remembered, is ellipsis 'from the right": the final element in the verbal group. the lexical verb~ is omitted, and preceding elements may be omitted, all except the initial operator. Operator ellipsis is ellipsis • from. the left': the initial element in the verbal group (finite verbal operator, if finite; otherwise first non-finite operator) is omitted, and following dements may be omitted, all except the lexical verb. An elliptical verbal group carries over certain systemic selections from the group that it presupposes. The general principles regarding this pre- supposition are as follows: Polarity Finiteness and modality Voice Teuse Lexical verb Lexical ellipm Operator ellipsis in:opplicable (always inoppHcable (always not presupposed (except last order selection in compound tense) not presupposed presupposed (ean be repudiated under cer- tain conditions) presupposed unless repudiated in:opplicable (always

4·3 VEllBAL ELLIPSIS r93 This: pattern is rdatable to the different contexts of the two types of ellipsis. Operator ellipsis involves omission of the MODAL BLOCK- the Subject and .finite verbal operator (see 4.3.6 and 4·4·I below)- in the clause; this is the element that expresses mood. Operator ellipsis is there- fore characteristic of those contexts in which the mood is taken over from the previous clause. Typically this happens within the sentence, but we are not considering presupposition relations within the sentence because they do not form part of the total picture of cohesion, which is an inter- sentence relation. Between sentences. the typical context in which there is presupposition of mood is that of question and response~ hence, as we have seen, we find operator ellipsis in answers to questions. particularly those where what is asked for is the identity or confirmation of the lexical verb. eg: what Me you doing?. are you thinking (or ... )? So in operator ellipsis the finiteness is always presupposed. whereas the polarity never is. Tense and voice may or may not be; that is, they are presupposed unless repudiated. Lexical dlipsi~ on the other hand, lea~ out nothing of the mtXhl block, so that the mood of the clause is fully explicit: in a verbJ.l group with lexical ellipsis the finiteness is always expressed. so the question of its presupposition from an earlier verbal group does not arise. Lexical ellipsis occurs in those contexts where the lexical verb is not in question; the lexi- cal verb itself is therefore always: presupposed. and so is the voice, since the lexical verb carries with it the implications of its transitivity - if the love from John loves Mary is taken over by presupposition into the next clause, then naturally this presupposition extends also to the fact that it was loves and no"t is ltwed by. If we want to override this and talk about Mary loving John~ we must restate it as a new proposition. in full. Polarity however is not presupposed. In fact it is impossible not to restate the polariry, because it is tied structurally to the initial operator, which is always present in lexical ellipsis. But behind. this is a more im- portant reason,. namely that the polarity may he precisely the question at issue. as in sequences like Did Jvhnrome?- No, he didn't.- Didn't he?; as also in question tags, which are not treated in detail here because they are with- in the sentence and therefore not cohesive. Similarly, tense is not carried over; the primary tense choice has to be restated, being embodied in the initial operator, and tense also may be up for consideration, as in John came, didn't he? -No. but he will. Thus the pattern of presupposition reflects the different functions of the two types of verbal ellipsis in bringing about cohesion wjthin a text. We have illustrated mainly with question-answ-er sequences, because these

194 ELLIPSiS allow us to display the cohesive relations more clearly in a short space. But verbal ellipsis is characteristic of all texts, spoken and written, and provides an extremely subtle and flexible means of creating varied and intricate discourse. 4.3.6 Verbal ellipsis and the clause This final subsection is designed to provide a link between the present section and the next. We have seen that verbal ellipsis often entails the omission of other elements in the clause besides verbal ones. Specifically, operator ellipsis involves eUipsis of the whole MODAL element in the dame. and lexical ellipsis involves ellipsis of the who]e of the residue, the PB.OPOSJTIONAL element in the clause. So, for example, the clause the cat won't catch mice in winter has as its structure (on the interpersonal dimen- sion of meaning); the cat Modal Subject nominal wont catch Predicator verbal g~oup mice in winter Propositional Complement Adjunct nominal prepositional rf this is followed by Or chase bir~ with operator ellipsis, then the Subject the C4t is omitted as weU as the verba1 operator won't. If it is followed by won't it?, with lexical ellipsis~ then the remainder of the proposirional cle- ment, consisting of the Complement mice and the Adjunct in winter. is omitted along with the lexical verb catch. Verbal ellipsis is always accompanied by the omission of the related clause elements, those that are in the same part of the clause as the relevant portion of the verbal group. So in operator ellipsis, where there is omis- sion of the finite part of the verbal group, the Subject is also omitted.~ in Jexical ellipsis, where there is omission of the non-finite p.art of the verbal group, all Complements and Adjuncts are also omitted. These dements are omitted., that is to ~y. unless they are explicitly repudiated. It is im- portant to note that they can be repudiated; we might have, with lexical dlips-is.

I95 [4:92) The cat won·t catch mice in '."\<inter. a. [operator dli.psis; Subject repudiated: • nor will the dog chase rabbits in winter•] -Nor the dog chase rabbits. b.{lexical ellipsis; Complement repudiated: 'it will catch birds in winte:r1 - It will birds. c. [lexical ellipsis~ Adjnnct repudiated: 'it will catch mice in summer") -It will in sununec. But if there is verbal ellipsis, then any sttucturally-related element in the clause that is not contrastive with that in the presupposed clause must he omitted also. You cannot say, foUo"'Wg (4:92] above: d. -Nor the cat chase birds (repeating the cat} e. -It will birds in winter (repeating in winter) £ -It will mice in summer {repeating mice) nor is it possible to use a reference item in this context, eg {d) Nor it chase biTds. Such e1ements can be repeated or referred to only provided there is no verbal ellipsis: nor will the cat fit cluue birds, it will chase ht'rds in winter/ then, it wtll chase mice/them in summer. Hence in an example such as [4:93] Have you checked this page? -I have (done) THIS page. the answer is possible only with this in a contrastive sense, meaning either a dilferent page. or this page in contrast to others. The principle here is that which is common to all forms of ellipsis: namely, that although the structural dements themselves are not present in the elliptical item. the features that are realized by these elements ARE present. So a clause in which there is operator dlipsis of the verbal group has no Subject; but if the clause presupposed by it is indicative (indicative being the feature realized by the presence of a Subject). then it also is indi- cative even though it has no Subject. Similarly a dause in which there is lexical dlips.is of the verbal group has no Complement or Adjunct. hut it takes over any of the features realized by these elements (type of transitivity; time, place. manner, etc) that are present in the presupposed dause. There- fore if the elliptical clause is making a DIFFERENT selecrion within these features - referring to a different time. a different goal, different location, etc - this MUST be expressed overtly, in order to repudiate the previous selection; and on the other hand if it is making the same selection- ie if there is no contrast hetween the two clauses with respect to a given selec-

I9(S tion - this CANNOT be expressed overtly. Anything else would conilict with the basic function of ellipsis. which is to create cohesion by leaving out. under definite rules:. what can be taken over from the preceding dis- course, making explicit only what contrasts with it. Finally, we should mention the limits of cohesion through verbal ellipsis, in terms of the function of the clause in question. An elliptical verbal group cannot in general presuppose a verbal group in an embedded clause{one that is embedded in the narrower sense of the wmd, k RANK- SHIFTED~ cf 3.4-2 above}. Consider for example: (4:94] a. The policeman paid no attention to the girl who was driving the car.- Was she? b. The policeman paid no attention to Mrs Jones. who was driv- ing the car. -Was she? {4:943] is impossible; here who was driving the cat is genuinely embedded (rankshifted). so that it becomes part of the nominal group the girl who was driving the ctJr. In [4:94b]. however, where who was driving the car is not rank.shi.&ed, but is related to the other clause by hypotaxis. the reiponse is quite acceptable. A clause which is rankshifted loses its functional identity as a clause: it does not operate as an element of the sentence. But a hypo- tactic clause does not lose its identity; it is still an dement of sentence structure, and so readily serves .as the target of presupposition from another sentence.. This is the basis of the distinction between hypotaxis and rank- shift. which tends to be obscured in the use of the term • embedding •; and cohesion provides evidence of the importance of this distinction. Similarly; [4:95] o. I shall stay in the city when I retire this year.- Do you? b. I shall stay in the city. even though I retire this year. - Do you? Here again. the clause when I retire this year in (a) is ranhhifted and there- fore cannot be presupposed; so [4~95a] is unacceptable. But the clause even though I retire this year in (b) is hypotolctic; this, therefore. is accessible by presupposition. and [4:95b) is a perfectly good example of mhesion by verbal ellipsis. 4-4 Clausal ellipsis 4·4·1 Modal and propositional We bave included under verbal ellipsis all instances of ellipsis in the verbal

4-4 CLAUSAL ELLlPSTS 197 group. However. both types of verbal ellipsis, both operator ellipsis and lexical ellipsis, also involve ellipsis that is externa1 to the verb itself. affect- ing other ekments in the structure of the clause. We can therefore look at these two types: of ellipsis from another angle, taking the clause as the po-int of depanure. The clause in English. con- sidered as the expression of the 1lariom speech functions, such as statement, question, response and so on, has a two-part structure consisting of MODAL BLEM.HNT plus PROPOSITIONAL ELEMENT (if 4.3.6 above). :fur {4:96) (x) The Duke was I going to plant a row of poplars in the park (Modal element) (Proposicional element) The MODAL clement, which embodies the speech function of the clause, consists in turn of the Subject plus the finite clement in the verbal group, Strictly, the part of the verbal group that goes in the mod.al block is simply the finiteness, which may not be realized in a separate element: it may he fused with the remainder of the verb, as in simple past and present tenses pftmWti, pJant(s). The PROPOSITION At ELEMENT consists of the residue: the remainder of the verbal group. and any Complements or Adjuncts that may he present. The difference between a Complement and an Adjunct is. briefly. that the Complement could become a Subject i.f the clause was turned round in some way. eg: a row of poplars was going to be planted by the late Dukej whereas the Adjunct could not. In the favourite clause type the Modal element precedes the Proposi- rional, though it need not do; we may have the Duke was ~going to plant a row of popW.s. Proposi Modal element tional clement o< !.;:96] (3) A mw of poplm Proposi the Duke wa!> ~going to plant in the puk. Modal clement tiona1 element The two types of vcrha1 eUipsis are derivable from these two major divisions of the clause. Under certain conditions there is ellipsis of the Modal dement: thus {4:97) What was th.e Duke going to do?- Plant a row of poplars in the park.

.198 In the answer • the Modal element is omitted: the Subject and. within the verbal group. the finite operator was. Hence there is operator dlipsis in the verbal group. In other circumstances there may be ellipsis of the Proposi- tioml element: [4:98} Who was going to plant a row of poplars in the park?- The Duke was. Here there is omission of the Complement and the Adjunct. a~ within the verbal group. of the lexical verb plant: so we have lexical ellipsis in the verbal group. The verbal element going to, which is neither finite operator nor 1exical verb -it is a non-finite tense operator - is omitted in both examp!es here: this is one of the features of verbal ellipsis which can- not be accoWited for simply by reference to the clause (it was dealt with in a preceding section, 4--34--4). There is no need to repeat here the details of what from the clause stand- point are modal ellipsis and prepositional. ellipsis, since those have already been discussed in cotmection with operator dlipsis and lexical dlipsis in the verbal group. In brief,. modal dlipsls is associated with a context where there is no choice of mood in the clause- mood,. the choice of declarative, interrogative, imperative and their subcategories. is the realization of speech function, and is expressed by the Modal element. Likewise. in modal ellipsis the polarity is determined., and the Subject can be presup- posed from what has. gone before. Typically~ in other words, modal ellipsis. occurs in response to a WH- question asking 'what {did, does., etc) ... do?' (c{[4:97] above): [4:99] What were they doing?- Holding hands. The usual type of non-finite dependent clause is. in fact, simply a clause with modal ellipsis; but it is one which presupposes another dame within the same sentence. this being what is meant by 'dependent • • and so it does not enter into cohesion; an example would be [4-: too] Holding hands they stole quietly out of the house. Prepositional ellipsis, on the other hand, is associated with those instances where the mood and the poJarity are the principal components of the mes- sage: typically, responses to statements and yes/no questions. where the subject is presupposed by a reference item. as in [4:101] a. The plane has landed.- Has it? b. Has the plane landed ? - Yes, it has.

199 It is also found in response to WH- questions where the unk.no"vn element happens to be the Subject (q[4:98]above): [4: I02] Who taught you to spdl?- Grandfather did. In general, in a finite clause with either o.f these two types of ellipsis the verbal group will aiso be elliptical: Clause {I} modal ellipsis (2) propositional ellipsis Verbal group operator ellipsis kxical ellipsis But there are certain circumstances under which this does not hold. (I) Modal/operator ellipsis. If the verb is in simple past or present tense, modal e11ipsis may not involve operator diipsis; moreover it is not always possible to say whether it does or not! [4: 103] a. What did he do?- Ran away. (Run away.) (but note: What did he do, run away?) b. What do they do?- Run away. c. What does he do?- Runs away. (Run away.) It wou1d be possible to have run away in (c) and also in (a); they would then be, appropriately, instances of operator ellipsis, since the full forms would be He runs (=does+run) away. He ran (=did+run) away. There is some uneasiness about run away in these contexts. perhaps because it APP~ to be finite (and therefore wrong, either in number or in tense); on the other hand the non-elliptical forms also seem wrong, because they are dearly finite and yet lack a Subject, which is contrary to normal pat- terns. So the preferred form is often that with pronoun Subject added: he ran away. he runs away. With [4: ro3b] the problem does not arise, since the non-finite run would in this case a1so be the appropriate finite form. {2) Propositional/lexical ellipsis. There are two occasions when prepo- sitional ellipsis does not involve lexical dhpsis. One is very general: the speaker may use the substitute do rather than the elliptical form of the ver~l group (see above, Chapter 3, especially 3.3.2). Here 'speaker'. ~ always, includes 'writer'; but in &et substituticnt is m.ore common in spoken than in written English. ellipsis being often preferred in writing. Moreover, as akeady noted, there is considerable variation among dif- ferent dialects; and there are individual differences also. But with some speakers, at least, the substitute form of the verbal group may beuseJ in all instances of prepositional eUipsis except those where the verb is passive, which do not substitute at aU. and. those where the verb is be or have, which

200 l!.LLlPSIS 'substitute' for themselves (the verb do does in fact substitute by do~ though the result dQeS not show}. But substitution is less usual in question- answer sequences, which have marked polarity and therefore are more often elliptical. Some examples: [4:!04] Presupposed clau.e Elliptical (a) Has the plane Yes it has. landed? (b) Keep out of It ha•. si.ght till the plane lmds. (c) Who was piano? (d) WuJohn piano? (e) Are the rest linislred? (f) Do« Jane (g) De« Jane (b) Has May done her homework? (j) Has the weather been cold? (k) I hear Smith is having an operation? Peter was. though. Yes, they are. Yes. she does. Yes, she has. Yes, it has. He has. Substitute Yes it has done. It has done. Peter was doing. doing, though. full form Yes it has lmded. lt has landed. Peter was piano. piano, though. Yes, they are linislred. Yes, she does do. Yes, she does does. y "'· .he has done. (Ye. it has been.) (He has had.) •mg. smgs. Yes, she has done her homework. Yes,. it has been cold. He has had an operation. Strictly speaking in [4: 104g] there is no elliptical form (if [3 :9rl in 3·3·3-3). although effectively the distinction between ellipsis and substi- tution is neutralized here. This is, in fact, the second of the two occasions where propositionalellipsis does notlead to lexical dlipsis, and it is not very interesting; namely, in simple past or present tense with unmarked posi- tive pobrity. where there simply is no distinct elliptical form. Hence the

4·4 CLAUSAL ELLlPSIS 2CH difference between (f) and (g) above: whereas in [4: ro4f'J there IS a dis- tinct form with lexical ellipsis,. because the verb has marked polarity (the full form is J~s sing)~ in [4: 104g} there is not- the polarity is unmarked and the full form is simply sings. And since the verbs be aud have (ie the have meaning • possess'. whi-ch is replaceable by have got, as in he had a yacht; not that meaning "take', ~undergo', etc as in he had an ope:Taticm. he had brealifast) do not t.ke the verbal operator do-their 'marked positive .. form is simply the non-reduced is, has. hy contrast with reduced • s, etc - these verbs Nl!VER have an elliptical form in simple past and present tense. So [4:104] (cont'd.) Presupposed Elliptical Substitute fOnn clause (f) Does Jane Y.,., she does. Yes, she does do. (g)DoesJane Q) Is be suspicious? (m) Is be suspicious? (n) lhsbe(got) a pn:judice against it? (o) lhs he (got) a pr~ndice against it? No, but Mary does. Yes, he is. No, bus Jobn is. y.,., be ha~ No, but Jobn has. Full form Yes, she does smg. smg.. Yes. he is suspicious. No, but John's SU5p1ClOUS.. Yes, be has (got) a prejudice against it. No, butJobn has Oobn's got) a prejudice against it. Some varieties of English treat this haw like the majority of otber verbs and we the operator tlo with it; for speakers of such varietie5, example5 (n) and (o) would not be valid. To sununarize the circmnstances under which clausal ellipsis, modal or propositiona1, may be found unaccompanied by ellipsis in the verbal group: operator ellipsis may be avoided in simple past and present tense~ and ru~titution may be used in most instances instead of lexical ellipsis, the two being indistinguishable from each other in simple past and pre- sent (unmarked positive form), and indistinguishable also from the full form in the case of the verb. b. aod have(~ 'posses>'). Otherwise, verbal ellipsis and clausal ellipsis go together.

BLLIPSlS The next section (4-4-2) deals with some instances of the omission of single dements in the clause. After that W'e go on to consider clausal ellipsis in its typical context of question-response and other types of re- joinder, lint in direct qx:ech (4.4-3) md then in indirect •peech (4-4-4)- A fina1 section refers to ellipsis in clause complexes (4-4·5)- 4-4-Z No ellipsis of single elements It is not possible in English to say [4:105] Has she taken her medicine?- She has taken. Either we must reply with a full, non-elliptical clause she has taken her medicine (or she has taken it. using reference to presuppose hu mdicine); or we must omit BOTH her media.'ne AND the lexical verb take and say she has (or she has don£. using the substitute do in its place). Let us tabulate these, from the point of view of ellipsis: [4: Io6J a. lfu she taken her medicine?- (i) No ellipsis: (1) no presupposition She has taken her medi- cme. (2) presuppo.ition of Complement by reference She has taken it. (ri) Clausal ellipsis: (I) with verbal ellipsis She has. (2) with verbal •ubstitunon She bas done. It may be hdpful to give equivalent sets for Jo. have and be: f-c.: to6] b. Has she done her c. Has she had d. Has she- been homework? her breakfast? unhappy? (i) (I) She has done her She has bad her She has been homework. breakfast. unhappy. {2) She has done it. She has had it. (u) (1) She has. She has. She bas. (>) She has done. She has bad. She has been. (We cannot S2Y .she has been ~ at least not in answer to lun she been un- happy?, although this would have been acceptable if the Complement had been a noun, eg: Isn't it time she was secretary?- Oh no. she's bun it already.) The notion that Jo, have (ail senses} and be 'substitute fo.r themselves' is

203 useful in explaining the forms given lUlder (iiz). If these were not sub- stitute forms, they would be impossible in the same way that She has taken is impossible: it is not possible to le.1vc out the Complement but retain the Prcdicator {verbal group} intact. This in turn is part of a very general restriction on eUipsis, whereby it is not possible to omit single elements &om the structure of the clause. If a single elemem of clause structure is to be presupposed, for purposes of co- hesion,. it must he expressed by a reference item; m (to vary example [4:96]) [4: I07] The Duke has planted poplars in the park. Presupposing: we have: (a) The Duke He h.s plantecl poplan in the park. (b) popJars The Duke has pJanted them in the park. (c) in the park The Duke has planted poplars there. We cannot omit he in (a) or them in (b); and although we could omit there in (c), in the sense that it would still leave an acceptable clause struc- ture, there would be no presupposition and therefore no feature of• place • iJJ-.the dame. There is no type of daus:al ellipsis which rakes the form of the omission of single elements of clause structme. It should be stressed. once again that we are confining our definition of ELLIPSIS to its non-structural, cohesive sense~ that is, as a form of pre- supposition between sentences. Within the sentence, we find internal BRANCHING which may involve the omission of single elements of dause structure {as v.-ell as structures of any other rauk). for example: !4: I oS] a. John loves Mo!ry but is loved by Jane. b. Either Peter will play his cello or Sally her guitar. c. Anne cut out and Sarah sewed a dress for every doll. Similarly for combinations of two elements: [4:Io8Jd. Sybil takes coffee very strong butJoan rather weak. e. We climbed. Great Gable Qfi Tuesday and Sea Fell two days later. But here in all instances the two parts are structurally related, by coordina- tion, and the patterns of occw:rence are quite different. The same explana- tion holds for [4-:ro8} £ The cat catches mice in summer. -And the dog rabbits. g. The cat won't catch mice in winter.- Nor the dog rabbits.

204 ELLIPSfS Even though these are written as separate sentences, they are in fact linked by coordination; this pattern would not be possible with hut, so or other conjunctive dements (see Chapter s). Aside from this structural •branching', there remains one other phenomenon which is to be distinguished clearly from eUipsis; this is not in fact an instance of omission, and involves no presupposition of any kind, but it is sometimes referred to, rather confusingly. as if it was a form of ellipsis in the eh. we. Examples would be: [4:H)9] a. Simon's: playing. Let's: not interrupt. b. Sandra deans. for me when I' m out. c. Run! These are sometimes described as elliptical forms of. eg, Simon's pla.ying the piano, Sandt-ll cle-ans the fiat, You run! Actually however they are systematic variants in which nothing is omitted. any more than an expression of time or place can be said to he • omitted • from a clause which does not contain one. They have no systemic features which are not expressed in their structure. It is misleading to call them • elliptical• because this suggests they h:otve some rohesive function similar to that of the elliptical forms we are discussing here. whereas in fact they have none. If there was ellipsis of the Complement, they wou1d presuppose the Complement, which they do not. They do not presuppose any preceding item; in general, they cannot occur in contexts where there is presupposition,. for example [4: r to] a. Does Sandra dean the windows? - She deans for me when !*m out. b. They a!ked Simon to play some Chopin. When he started playing. it was LiS2t. In (a) the one thing the response could not mean is' she deans the windows for me~, which would make it like [4:105] above. This is borne out by [4: nob]. where the response is quite acceptable and started playing dearly does NOT presuppose Chopin. We have emphasized at various points in the discUs.sion that the distinc- tions we are drawing, while they are useful and important for explaining the patterns that lie behind the construction of text, are not to be taken too rigidly. When we say that there is no type of clausal ellipsis consisting in the omission ofindividual elements of structure of the clause. we are stat- ing :a generalization. one which explains why certain theoretically possible clause types cannot occur independently - though they may occur in branching structures. This does not mean that a. single dement of clause

4·4- CLAUSAL ELLIPSIS 20,S structure can never he presupposed undeL any circumstances. We can have sequences like f4:111] a. We went on the river yesterday. We had dinner out too. b. Can you read the print without your glasses? -No, but I can look at the pictures. where the second sentence in (a) also refers to 'yesterday• and the response in (b) refers to 'v.ithout my glasses'. But these are not elliptical sentences. They merdy imply, in the particular context in whlch they occur, the particular time, manner. etc referred to in the preceding sentence. So in I 4: I I I b] the response implies 'I can look at the pictures without my glasses •. but it does not itsdf embody a feature of manner, nor is without my glasses in any sense omitted from it. At the same time, the line between what is elliptical and what is not elliptical is not a completely sharp one. Most instances are dear; there is no doubt that the omission of modal and propositi anal elements in the clause, as in 14:97] and [4:98]. is to be explained as dii.ptical. whereas the types represented in [4:109-III] are not. But there are some doubtful in- stances. For example it might reasonably be suggested that in [4: tu] the second sentence is actually benefactive, the Beneficiary him being omitted by ellipsis: (4: riz] Are you sending jack anything for his birthday?- I thought of sending a book token. However, there appear to be no examples of the omission of just one ele- ment from the structure of a clause WHERE THAT BLEMENT IS OTHERWISB OBUGATO:RY- of the Subject, for example, or a Complement following a verb which must have a Complement (</[4-: 105] above}. Hence instances which on other grounds could he interpreted either as dliptical oL as non- elliptical, but which if regarded as elliptical would take the form of the omission of a single element of clause structure, should perh2.ps for that very Leason be excluded from the category of ellip5is. This is a theoretical decision, and one which would allow us to formulate a very general principle about cohesion in the clause. This principle is as follow>. Other than in a question-answer environ- ment (to be discussed in the remainder of this chapter). the basis of both elljpsis and substitution- and these, as explained earlier, .are essentially the same phenomenon -is the two-part structure • Modal plus Proporitional·. One or other of these elements may be presupposed,. as a whole; but the s.maller elements which make them up- Subject, Complement. Predica-

ELLlPSIS tor. Adjunct- may not be presupposed in isolation. The facts on which this principle is based are often indeterminate. as the facts of1anguage always are~ we do not force them into a mould. but in the uncertain instances we choose that interpretation which brings more of them within the scope of a single generalization- provided it is one which makes good sense_ Here it does.. It is the Modal-Proposi.tional structure which expresses the func- tion of the clause in the discourse. so it is natural that this structure should provide the means for integrating any clause into a coherent text together with what has gone before. 4·4·3 Ellipsis in qu~stion-answer tmd other rejoinder sequences Not all questions have an answer; but no less significant is the fact that not all answers have a question. The 'question and .answer • sequence is a standard pattern in language, and not surprisingly the special type of co- hesive relation that subsists between an answer and its question has its own characteristic grammatical properties.. At the same time there are other sequences involving rejoinders of one kind and another. Let us first make some terminological distinctions. Any observation by one speaker. whether it is a question or not, may be followed by an obser- vation by another speaker that is related to it by some cohesive tie. We shall refer to this very general category of sequel as a REJOINDER. A re- joinder is any utterance which immediately follows an utterance by a different speaker and is cohesively rdated to it. A rejoinder that follows a question will be called a RESPONSE. Within the category of responses there is a further distinction between DIRECT .RESPONSE.S and INDIRECT RESPONSES. A direct response is one which answers the question; it is either a form of'yes' or 'no • ~if the question is of the yesino type, or a specification of the information .asked for by the WH- element. if the question is of the WH- type. An indirect response is either one which comments on the question (cOMMENTARY). or one which denies its relevance {DISCLAIMER), or one which gives supple-- :mrnt:ary information implying but not actually expressing an answer {SUPPLEMENTARY RESPONSE). A direct response v.-ill also be referred to as an ANSWER. But note that the category of answer, which is the supplying of the particular inforiD2- tion that is called for by the question. is not limited to responses, because one can answer ones own quesnom. Other rejoinders, not following a question. jnclude ASSENT and -cON- TllADICTlON, following a statement; CONSENT and REFUSAL, following

Table 9: Types of rtjoinder JU!JOINDBD. (any cohesive sequd by different speaker) (not following question) a question) lE s P l N sE [other rei' cinders] I' (answering) (not answering) (following DIRECT RESPONSE JNDJRBC'f R:ESPONSJI statement) Statement Of a coaun;md) .-----...------LI--, command) (attitude to (evading (implying QUESTION 1 REJOINDER ASSENT CONTRADICTION CONSENT REfUSAL COMMJm- DtSCLAIMBll SUPPLBMBN~ TARY TARY RBSPONSll .. .. n d M • :: • " M

BLLlPSIS a command; and yes/no or WH- question following either a statement or a command. Some examples of all these types: (4:II3] a. It's going to rain. -(i) It might. (ii) Itisn"t. (iii) Is it? b. Leave me alone.- (i) I won't. (u) All right, I will. (ui) Why? The sequel sentences are rejoinders. since they are cohesive utterances by another speaker;. but they are not respcmes, because the presupposed items are not questions. [4: 113a] is a statement, to which the rejoinder is (i) an assent. (ii) a contradiction, (iii) a yes/no question; (b) is-a command. to which the rejoinder is (i) a r~ (ii) a consent, (iii) a WH- question. [4: II 3 J c. His John arrived? -Yes, he has. d. Whendidjohnarrive?- Yesterday. Here the two sequels are rejoinders of the 'response • type. and both are direct responses.. or 211Swers; they give the information that is being sought. [4!II3] e. Howdidtheybreakin?-I'llshowyouhow. £ Why didn't you tel!Jobn? -I did. Here the sequels are still responses. but indirect; the first is a commentary~ the second a disclaimer. (4: 113] g. Did you tell John?- He wasn't there. In (g) the response is also indirect,. but here the answer is implied(' ... so I couldn't~); these we shall can supplementary responses. Finally [4: 113] h. Did !lock the door? Yes of course I did. Here there is only one speaker. so the sequel is not a response; but it is an answer, since it gives the information required. 4·4·3·1 DJRBCI RESPONSES (1): YFS/NO QUESTIONS Answers to yes/no questions,. or POLAlt. QUESTIONS as they have been called, are very simply dealt with, as the instruction 'Answer ya or no!' suggests: the appropriate answer is ye.s or no. The words yes and no express simply a feature of polarity. They do not mean (as do their dictionary equivalents in some other languages) •you are right' and • you are wrong'; they mean • the answer is positive' and 'the answer is negative'. Hence their meaning is unaffected by the polarity of the question~ contrast the forms of the positive in French:

[4: II4) (I) Question a. Arc you coming?- b. Aren't you comlng?- (2) Answer, positive Yes ('f am') [oui] Yes('! am') [si] (3) Answer, negative No ('r'm not') [non] No \I'm not') [non] In this connection it is interesting to note that both yes and no occur more often as rejoinders t-o statements than they do as aruwers to questions; here both of them signal' I agree', • [ understand •, ~I' m listening' -keeping the channel of communication open - and the choice of one or the other simply follows the polarity of the preceding statement: [4:u5] a. The soloist wasn't very inspiring. -No, he seemed rather tired. b. The car's running very well. - Yes, I had it serviced recently. h is possible to consider yes and no as clause substitutes. But they are not really substitutes; for one thing, they can be accompanied. by part or even the whole of the dause for which they would be said to he substituting, and that is precluded from substitution as. usually defined. For example in [4:II44] the answer(2) could be yes, yes I am, or yes! am coming. They are realizations of a single clause feature, that of polarity. which is being expressed on its own instead of in association with the verbal group; and the fact that it is expressed on its own means that the whole of the remain- der of the clause is presupposed; hence their cohesive effect. The words yes and no express simple polarity. There are also complex expressions,. some meaning • either yes or no', eg: maybe. perhaps, ;md some meaning •both yes and no~, eg: sometimes, WUlJily. The former ace often combined with some modality. the speaker's assessment of the relative probabilities of "yes • and •no ', eg: probably, possibly. All these are appro- priate answers to yes/no questions; and they are also cohesive, since they presuppose all the remaining features of the clause other than the polariry. If the answer yes, or other expression of polarity, is accomp<mied by jus.t a part of the clause, this will be the Modal dement: yes I am, no I'm not, sometimes lu- does~ perhaps she has, possihly they might.and s.o on. The Modal element is itself sufficient as an answer, since it also carries the polarity (and presupposes the Propositional element of the clause); so [4-: r J4a] could be answered simply by I am, rm not. If both occur, as in yes I am, they func... tion jointly as the realization of a direct answer; .as distinct from (4: II6) Did you see anyone?- Yes, Shirley.

2"10 where the second part of the answer, Shirley, is an indirect response. giv- ing supplementary information (see 4-4·3·3 below). 4·4·3·2 DIRECT RBSPONSJlS (2): WH- QUESTIONS At .first sight the answer to a WH- question, or NON-POLAR QUES- TION, seems very different from the answer to .a yes/no question, since the information that is being sought by the two types of question is very different. A WH- question requires the specification of a particular item whjch is as it were missing from the clause. The respondent knows what the function of this item is in the clause structure, since this has been sup- plied by the questioner; he knows the total structure of the clause, in fact, and also the actual items that occur in all the other functions. He merely has to fill in the blank. The WH- expression itself indicate5 whether the missing item is participant or circumstance, and various other things about it: if it is a circumstance, whether it is time, place, cause, manner, etc (whm, where, why, lww, and an open-ended set of forms such as what time, what with, like what, which way,for whose .sake); ifitis a participant, whether it is from a limited set (which) or not, whether human (wlw, which/what person. etc,. or possessive whose, etc) or non-human (what, which/what thing, etc), and whether the question is one of degree (hew much/many. how long, etc) or of kind (what kind, like what, etc). The simpkst form of answer, therefore, is -one which. does. mer-dy fill in the blank: which supplies the appropriate nominal, adverbial or prepo- sitional gcoup to act as Subject or Complement or Adjunct, and as Actor or Goal or Beneficiary or Temporal or Locative or whatever function is required. So for example: [4! n7]a. What did I hit?- A root. (Complement; Goal) b. Who killed Cock Robin 1- The sparrow. {Subject; Actor) c. How much does Jt cost? - Five pmmds. (Complement; Range) d. How's the patient? - Comfortable. (Complement; Attri- bute) e. Till what time are you staying?- Half past three. (Adjunct; Temporal) £ What did you draw it with?- A pencil. (Adjunct; Irutru- ment) g. Whose gloves are these?- Sally~ s. {Complement; Identifier) The principle underlying these answers is, however, exactly the same as that which governs the answers yes and no to a question of the yes/no type.

4•4- CLAUSAL ELLIPSIS 2II In each ase. when givmg a direct response in its simplest form the speaker makes explicit just one thing. the information that the question calls for, and leaves all the rest to he presupposed by ellipsis. With a yes/ no question, this information is the polarity, so the answer specifies the polarity and presupposes aH else. ln a WH- question. the infortn.3tion required is the item occupying a particula~ function (strictly, a particular complex of functions) in the structure; the answer specifics this and pre- supposes the remainder of the clause. Hence the principle of clausal ellipsis. in clauses which are answers to questions, is general to all types of question. Any clause fWlCtioning as answer, in the sense defined above (see [4: I I3C and d]). has an elliptical form consisting of just one element. Which cle- ment this is is explicit in the form of the question, and all remaining fea- tures of the question clause- excepting. of course. its interrogative mood- :ire presupposed .. Just as with a yes/no question we may aho have longer, partially elliptical {or entirely non-elliptiC41l) forms of answer, so too these may he found with \VH- questions. If the WH- item is Subject the answer may, like the answer to a yes/no question, have propositional ellipsis; this is because the Subject falls within the Modal element. So the answer to [4-: 117b] might be The spaTTow did. Whether or not the WH- item is Subject. the answer can be filled out with no ellipsis at all; we could have The sparrow killed Cock Robin in answer to [4: 1I7h] and You hit a root in [4: IIp]. If the WH- item does not form a complete nominal or prepositional group by itself (eg: whose gkves, what ... with. till wh.tt time), then the simplest answer is one in which there i.s not only clausal ellipsis hut also ellipsis within the group. either nominal (eg: Sally's in [4:n7g]. for 'Sally's gloves'; if 4.2.J.I, [4:24] above) or prepositional We have not dis- cussed the dlipsis of prepositions in the Adjunct as a separate topic. since it occurs only in this context; it is however illustrated by [4: II?e and f]. where the prepositions till and with arc presupposed in the answer. In such instances there is an intermediate form of answer in which the group is not elliptical but the clause is; so we could have [4: t17] e'. Till what time are you staying?- Tlll half past three. f' _ What did you draw it with?- With a pencil. g'. Whose gloves are these? - Sally's gloves. There is no WH- verb in English; we cannot ask ycu·re whatting the eggs?- or rather. since the WH- item comes first in the clause, whatting are you tile eggs? Instead we have the form what (Jf'l! )'QU doing to the eggs? This involves the use of the 'pro-verb' do, in the combination Jo what? (if

2.12 ELLII>SIS 3·3·3·3). The exp£"ession do what?, like do that, presupposes the whole of the prepositional element in the clause other than anything that is repudiated. so that the question what are you doing? is appropriately answered by a proposition (ie a clause with modal ellipsis) rather than jwt a lexical verb, although tbe proposition might CONSIST just of a lexical verb. Thus relax- ing. frying eggs,foeding the ducks in the park could all aruwer do what? questions. If the do what 1 is not intended to embrace the entire proposition the iterru that are not being asked for are made explicit, eg; what were you doing in the park? Such items however can never occur as nominals; they must take the form of a prepositional group. Following do what?~ a pre- position is required even with an item functioning as Goal in the clause structure. usually to or with: [4: n8J a. What are you doing with the eggs?- Poaching them. b. What have the children done to the wheelbarrow? -Broken it, This is in accord with the general principle whereby all.PARTICIPANTS in the structure of the clause, in Engli~ may be related to the verb either directly as nominals or through the medium of a preposition; the preposi- tional form is used to make the function explicit where this is necessary - as it is here because we do not know what the verb is.* The answer fol- lows the nonnal pattern except that if there is a Goal-Complement in the question it is usually presupposed by pronoun reference rather than by ellipsis; so we havt:' them and it in the answers in (4: u8J. 4·4·3·3 INDIRECT RESPONSES There is one kind of response which is not a.n answer in the defined sense. but is what we called an INDIRECT RESPONSE; this may be a COM- MENTARY, a DISCLAIMER or a SUPPL.EMRNTARY. Any question may be greeted by a COMMENTARY, which is really a statement about the speak- er's attitude to the answer: his ignorance of it, for example, or his consent or refusal to give it. These. since tbey are in fact reports, have the elliptical potentialities of 'reporting-reported • sequences as described in 4·4·-4- be- low. Exampl., are (and g[4: UJe)): {4: II9] a. Is it Tuesday today?- I don't know. b. Why are the lights turned off?- I'm not supposed to say why. '* CfM. A. K. fulliday, 'Langtnge structuR" and languagr function', iu]<>hn Lyons (ed), New HoriZ<mJ in Linguisti&s {1-brm.~~ Penguin Boob,. 1.970), e~y p 16.4.

4·4 CLAUSAL l!.LUPSIS 2.1} Likewise, any question may be followed by a DISCLAIMER, which side- steps the question by disputing Its relevance. TypicaUy, a disclaimer in- volves moving from a yes/no to a WH- context, Or vice versa, eg (and if (40 IIJf]) 0 [4: 120] a. When did they cancel the hooking ? - Did they? b. "What's your tdephone number?- We're not on the phone. c. Have you tested the battery?- How? Some questions are framed so as to be difficult to disdaim, such as the notorious Wlten did you stop b~ating your wife?, to which there is no ELLIPTICAL response meaning unambiguously •I never have beaten her'. Normally however a response of the disclaimer type is elliptical; either it is declarative, with propositionat dlipsis, or it is interrogative, in which case it is of the opposite type to the question, and has response-question ellipsis (see below). The third type of indirect response is a SUPPLEMENTARY RESPONSE, which gives information other than that which is asked for but answeo the question by implication, for example (and if{4: IIJg]}: [4: 121] a. Can you make it stand up ? - If you keep still. h. Have we a car here?- Not unless you came :in yours. c. Did you get the application form?- It's on my desk. d. Aie you co-ming back today? - This evening. Characteristically there supplementary responses presuppose the entice question; and they stand in a definite structural relation to it- or rather to the declarative clause which would serve as a direct answer to it. by which they would he 'fll!ed out•. In [4:12.ra], for example, the full form would be (Yes) I um make it stand up if you keep stiU: that is, the direct answer, with the supplementary response hypotactically related to it, as a condition, The answer is positive unless repudiated by not, as in [4: urh]; the supple- mentary is usually conditional or causal r yes if ... ' > • yes: because ... }~ al- though it may be simply coordinated as in [4: nrcJ {'yes and .. .') and [-t.: 121d] ('yes, more specifically .. .'}-or even with an adversative impli- cation. the presupposed answer then being negative ('no but .. .') as in [4: I2.z] Did you collect the subscriptions?- Smith did. Supplementary responses. are typically associated with yes:Jno questions; it is difficult to answer a WH- question hy implication, and the nearest equivalent form of response to a WH- question is really a type of dis- claimer, like [4: 12ob]. But the various types of indirect response are all fairly similar and cannot be kept strictly apart.

2!4 4-4·3·4 A NOTE ON ZEUGMA As a postscript to this discussion of ellipsis in responses. we might add a brief note on zeugma. Zeugm.a is based. on ellipsis. and both the WH- and the yes/no types of question afford excellent opportunities for zeugmatic answers. These involve a transfer from one element of dausc structure to .another; with a WH- question this can be achieved. by a direct resqonse (ie an answer). as in [ 4: IZ3a and b]. but with a yes/no question it requires an indirect. supplementary response of the 'no but .. .'type. as :in [4:123c and d]: [4:rz,3} a. What has he been malcing?-A big mistake. b. How did you travd? - fu considerable discomfOrt. c. Did she make him a good wife? - No. a good husband. d. Was he shot in the street?- No. in the shoulder. In (b). for example, iww is intended in the question as 'by what means?' but is interpreted in the answer as • in what condition •. In its classical form as a figure of rhetoric, zeugma is embodied in coordinate structures, where the pattern is one of (structural) branching and not cohesion {eg; we travelled in buses and great discomfort). But the principle is the same. 4·4-3·5 OTHEll. REJOJNDI!RS We have already mentioned the difference between a :response and a rejoinder. A response is one kind of rejoinder, one which presupposes a question. and which therefore- Ins special potentialities for ellipsis, as dis- cussed in 4·4·3·!-2 C direct responses') and 4-4·3·3 ("indirect responses}. A rejoinder is any utterance by a second speaker which presupposes that of the first speaker whether it was a question or not. (W'e referred above to the frequent use of yu and no in r~oinders to statemenrs.) From the point of view of cohesion, there is no distinct .category of rejoinder; i.t is simply an ordinary element in a dialogue. and is covered by what has been said about dlipsis in general. But there remain a few observations to he made about rejoinders that are not responses to a question. but cohesive sequels to a statement or a command. QUllSTION R.EJOlND.ERS have the fUnction of querying a preceding statement or command, or eliciting supplementary information about it. (.r) One type is that which presupposes the entire preceding dause and seeks confirmation of it as a whole; these are yes/no questions and nearly always have the form of interrogative clauses with proposirional ellipsis., like the question tag at the end of a declarative or imperative clause, except

4·4 CLAUSAL BLLIPSIS .2I$ that with the rejoinder there is a different speaker and the polarity is never reversed. For example [4: 124] a. Peter's here.- Is he? b. Open that parcel. - Shall I? c. •1 com't believe that!" said Alice. •ean'tyou?' the Queen said in a pitying tone. (.2) In another type the speaker identifies one item as requiring confirma- tion; the remainder of the clause is omitted but this item is queried explicidy~ a. John? (To dinner?) [4:125) John's coming to dinner. b. Tonight? c. And Mary? (Not Mary?) These .are three typical instances. In (a), an existing eiement is echoed; in (b) a new element is added; and in (c) an existing dement is expanded. here by coordination. But the fOrm of ellipsis is the same in all three; only one element in the clause is present in the structure, the remainder being presupposed by ellipsis. Finally the speaker may similarly focus on one item in the clause hut query it in the form of a WH- question. Corresponding to {4: usa] we have the 'echo question' represented in [4:I26a]. meaning 'please repeat that' ( 4 [ didn't bear'. 'I'm surprised', etc}; this is the only type ofWH- question w-hich must be spoken on a rising tone. Corresponding to [4:12sb] is [4: u6b]. where the WH- item represents a new element and the whole of the clause i~ presupposed; and to [4~ use]. [4-:IUic} where the WH- item who else? asks for expansion of an existing element by coordination. a. Who? (tone z) [4' 126] John's coming to dinner. b. When? c. And who else? In some ins-unces the rejoinder ash for more specific :infOrmation about an item that is already present: [4: 126] d. John~s coming to dinner.-John who? e. Alice heard the Rabbit say, • A barrowful will do, to begin with.' • A barrowful of what', thought Alice. Often the item requiring further specification is itself indefinite (if[ 4: I 29d] below). eg: Someone·s coming to dinner. - J.Vho? This type could a1so per-

2I6 ELUPSTS baps be matched in the ye$./no series illustrated in [4: us]: ]ohn,.s coming to dinner.-John Smith?; Someone's coming to dinnu.- John? The other types of rejoinder to a statement or command wually take the form of an elliptical clause consisting of the Modal constituent only- that is. one with prepositional ellipsis - with pronoun Subject, but in the declarative fonn. In [4: 12-7]. (a) is an assent, {h) a contradiction, (c) a con- "'nt and (d) a refusal: [4: 127] a. •Everything's just as it was!' ~Of course it is/ said the Queen, h. ~ ... being so many d.Uferent sizes in a day is very confusing. • 'It isn't." said the Caterpillar. c. 'Our family always h.teJ cats: ""''Y• low, vulgar thing.! Don~t let me hear the name again!" 'I won't indeed!' said Alice, in a great hurry to change the subject of conversation. d. 'Never mind what they all say~ my dear, but take a return- ticket every time the train stops.' 'Indeed I sha"n't!' Alice said rather impatiently. These forms often combine with yes and no, as in the following: [4:127] e. It's none of their business.- Yes it is. f. 'It must come sometimes to .. jam today"; Alice objected. "No, it can't," said the Queen. As was remarked ear:lier. yes and ~w also occur alone; they are the forms of expression of positive and negative polarity when everything else in the clause is omitted by ellipsis: [4:127] g. It's col.J.- Yes. h. We're not late. -No. These are both instances of assent. In contradiction, the Modal consti- tuent of the clause is osua.lly added. and it must be added if the contradic- tion is positive; so in (h) we could not have just yes as a rejoinder (and if( e) above). The same pattern is fonnd following a command; we may have yes or no alone. but if yes occurs following a negative command, and there- fore signals a refusal, it must be accompanied by the Modal dement of the clause: [4: 127] j. Don~t tell anyone what you saw!- Yes. ( wilL k. Don't lce5 go back!- Yes, let's. All these rejoinders could be 'filled out' in a non-elliptical form. But

4·4- CLAUSAL ELLIPSIS 217 the pattem which is most typicl of or-dinary dialogue is that described here, with just one element mad.e explicit. and the remainder presupposed. This use of ellipsis to express the cohesive relation between a re;joinder and the utterances which preceded it,. in sequences which are not structured as question-answer, is a very characteristic aspect of the textUre oflinguistic interaction. 4·4·4 Ellipsis in 'reporting-reptmed' sequences There is one further context for clausal ellipsis. that of reported. speech. The type of ellipsis found in this context is closely related to some of the instances that we have alre2dy met,. particuJarly the· commentary~ type of indirect response [4.: I rg]. and the elliptical WH- question as rejoinclec [ 4: 126]. Here therefore we shall not be introducing a new type of elliptical structure, but bringing together various instances already met with which have something in common. What they have in common is the feature R.BPORTED, which is present in indirect speech: that is, indirect statements. yes/no questions and WH- questions. These are exemplified. in their full form in the second clauses of [4: r28a, b arul c) respectively: a. (that) he was coming. [4: I28) John didn't tell me b. if]whether be was coming (or not). c. why he was coming. lt is perhaps important to point out that the speech function of the report (statement, yesfno question, "WH- question) is a feature of the whole com- plex. even if the same verb (e~: teU) occurs in all types. The reported clause itself makes no independent selection of mood,_ and examples such as I asked John why was he late or I asked MtltJI would she enjoy it are not true interrogatives, as shown by the fact that. in the following. (ar): (a2} and (br): (b2-) are not proportional: (r) {>) (a) you'D enjoy it will you enjoy it? (b) I asked Mary if she'd enjoy it. I asked. Mary would. she enjoy it. 4·4·4• I lNDnt:BCf WH- QUESTIONS If the reported clause is .an indirect WH- question. it can he elliptical in the same way as its equivalent direct question, the WH- type interroga- tive clause. For example:

218 BLL[l'SIS [4: 129) a. Who cou1d have broken those tiles?- I can~t think. who. b. I said you would mend it for him. -I hope you didn't say when. c. The jewels are missing. - I wonder what else. d. They still have some copies in stock. - Did you ask how many? The target of presupposition Is not, of course, the immediately preceding clause, which would be the reporting one~ but the precedlng sentence; eg in (a} who presupposes ... could have broken those tiles and not I can't think. This is possible becawe a reponed clause is not embedded; the reporting-reported structure is a hypotactic one (if 3.4.2 above; also 4-3.6, [4:94-] and [4-:95])~ and therefore the reported clause can reach out beyond the bounds of the sentence in which it occun:. The conditions Wldex which elliptical forms tend to occur in indirect questions are those we have already met with in explaining ellipsis in direct questions (if [4-: u6] above). The 'echo' type occurs. where the presupposed clause was itself a WH- question,. as in [ 4-; 1293]; here however the elliptical clause does not take a rising tone (tone 2), since it is not itself interrogative. In [4: I29 b] a new element is added which was absent from the clause that is presupposed. In (c) and (d) an element already present in the presupposed clause is offered for expansion, either by coordination (c) or by further specification (d). Characteristically in the last instance the item in the presupposed clause is a non-specific form, with .some- (someo>ne, something, etc)~ or the indefinite article or other non-specific dcictic (see 4-2.J.2 above). As in a direct question. all features and elements of the presupposed clause are carried over unless repudiated.. The presupposed elements may include ~rt of a nominal or prepositional group where the WH- item is a modifier such as which, whose, how many; in an example such as [4: 129d]. In addition to the clausal ellipsis there is also nominal eUipsis, since how 11Mny cou1d be fi11ed out as how many copies. The mood of the presupposed clause is always repudiated, by the WH- item itself. Other elements also are sometimes repudiated, particularly the finite operator; in such instances, however, the WH- clause nonnally has to be non-elliptical, unless the WH- element is the Subject (which allows fur prepositional ellipsis) as in Who's gcing to lead the way? - I can~ t think who could. 4-4·4.2 INDIRBCT YES/NO QUESTIONS If the reported clause is a yes/no question, the most w:ual dliptical form of it is simply zero:

4·4 CLAUSAL ELLIPS-IS 219 [4: 130] a. Was that an earthquake?- I don't know. b. I wonder whether England won the cup. Have you heard? All features of the presupposed clause are carried over. (All, that is, except the polarity. Polarity has a special meaning in a yes/no question. different from its meaning elsewhere; it expresses the questioner's attitude to the question. as in Don't you know? meaning • I' m surprised', 'you ought to know~. In an indirect question which is elliptical. the polarity is simply neutralized.) In these instances the presupposed clause is itself a question, direct or in- direct~ and we could conclude that this feature is alse presupposed in the elliptical clause. It is not necessary for the presupposed dause to he a question, however, in order for the ellipsis. to be interpreted in this way. If the verb in the reporting clause is one that introduces a question, such as ask, then the elliptical reported clause will he interpreted as a question whatever the mood of the presupposed clause. In [4: 131a] the indirect yes/no question whether he was or not is entirely omitted by ellipsis even though the presupposed clause is not a question of a.ny kind: [4:I3I] a. John was very disappointed by the response. You can ask:. him. A more usual type is that in which there is a modality in the presupposed clause; [4: 13r] b. She might be better living away from home. I'm not sure. - and these in turn are related, as was pointed out in the context of the dis- cussion of clausal substitution (3-4-I.r). to elliptical modalized clauses in which everything except the modality is presupposed: [4: 131] c. I wonder if it~Jl rain on the day of the picnic:.-Probably. 4·4·4-3 INDIRECT STATEMENTS For an indirect statement. there is no equivalent elliptical form containing just the marker of the feature 'statement'. Here the cohesive form of the reported clause is the substitute so, or its negative not, as in [4: 132) I thought Mary was leaving today.- She hasn't said so. This has been discussed in 3-4.1.1. The elliptical form is again simply zero, as m [4:133] a. This mango is ripe. I know fi:om its colour. b. England won the cup. - Who told you?

lll.LIPSJS Again the whole of the presupposed clause is carried over, including the polarity; hut an elliptical indirect statement of this kind does not neces- sarily follow immediately after the presupposed clause, and the exact domain of the presupposition is sometimes. rather uncertain. Consider fOr [4: t34-] a. John's new cabinet is beautiful. I've seen it being made. You can tell him. Does this meap. 'you can tell him it's beautiful', or 'you can tell him tve seen it being made~, or both? If we expand the second sentence to [4:134] b. John's new cabinet is beautiful. I've seen it being made; it"s nearly finished.. You can te11 him. the amount of unce:rtainty becomes even greater. Even where there is no doubt which clause is presupposed, there may still be some more room for uncertainty, umally centring around the expression of modality; for [4::r3s] a. He hasn't finished. -l should have known. Y h b. Didn't you know? ou can go ome. N will kn c. oone ow. Here (a) might mean either 'I should have known that he hadn"t" or •1 should have known that he wouldn't•. More dearly still, whereas (b) means • didn'tyou know that you can?', (c) me2ns 'no one will know that you have gone'. The pattern of determination here is not easy to sort out. 4-4--4-4 AMBIGUITY BEl WEEN IND!llECT STATBMENTS AND INDIRECT QUESTIONS Fundamentally this is the same kind of ambiguity as may arise between a statement and a yes/no question if the reporting verb is one that can intr~ duce either. such as tell, say. report. know. and the reported clause is omitted by ellipsis. For example, in [4: 136]a. I think the cheque is still valid. The Bank can tell them. it is not dear whether the meaning is ~the Bank can tell them that the cheque is still valid • or • the Bank can tell them whether the cheque is still valid or not •. Again. it is not easy to state exactly what the relevant factors are; but the following examples would prestm1ably not be ambiguous:

4·4 CIA USAL ELLlPSIS 221 [4:136) b. The cheque is still valid. The Bank can tdl them. (-'that it is} c. The Bank can tell them. ( = • whether it is) d. The Bank told me. The cheque may still be valid. (='thatitmayhe/ With say the' zero • form of the .reported dause nearly always presupposes a question: he didn't say is likdy to mean ·whether .. :, not 'that.,.~. But with other verbs, and also with some adjectives such as cleM, the pre- supposed clause may be either question or statement. There is perhaps a tendency here for a • zcro' reported clause to be interpreted as a question if the reporting clause is negative, :u it frequendy is with these ambiguous expressions (it isn't clear. he didn't say are more 1ike}y to occur on their own as reporting-reported structuces than it is ckar, ht said); on the other hand, whereas I dcm't know is probably 'whether .. .'. I didn't know is more likely to be 'that .. .'. As (4:136] shows, various factors both in the reporting clause and in the presupposed clause seem to he relevant to the interpretation. 4·4·4·5 REPORTS AND FACIS JN RELATION TO CLAUSAL EU.IPSlS We should distinguish here between reported clauses and 'fact • clauses (cf above, 3.4..r.r). A REPORT clause. as already noted. is related hypo- tactically to the clause that contains the reporting verb; a sentence such as John said ~\fary was leaving consists of two clauses, the second dependent on (ie hypotactically related to) the first. A FACT clause is embedded. in the strict sense of downgraded in rank. or ·rankshifi:ed •; the sentence John predicted that Mary was leaving CONSISTS OF only one clause, which has embedded within it another one that no longer functions as a clause but functions as a nominal. There are various differences- between report and fact. which were summarized earlier (3.4.I.I); note that some verbs can occur with either. Because of the structUral difference between the two. a reported clause can be expressed cohesively through substitution or ellipsis., whereas a fact clause cannot. We have seen that there is no ellipsis of single elements in the structure of the clause; .a fact clause, being em- bedde~ functions as a single element, and hence cannot be omitted on its own. This explains why we cannot say {4: IJ7] a. The opportunity has now been iost.- I sincerely regret. A fact clause can on the other hand be expressed cohesively, as all single

222 liLLIPSIS elenrents of the clause can, by means of reference; so [4: 137b} is quite acceptable: [4: 137] b. The opportunity has now been lost.- I sincerely regret it/the fact. 4-4.5 Clausal ellipsis and dause crJmplexes Two or more clauses that are directly related in structure (as distinct from being related indirectly through r.ank.sh.ift) are said to form a CLAUSE COMPLEX. A clause complex may he either PARATACTIC or HYPO- TACTIC. In a paratactic clause complex the clauses have equal status. The relevant paratactic relation is that of coordination, ie ·and~ and 'or'; there are two others, namdy apposition and quotation, but we can ignore them here. In a hypotactic complex the clauses have unequal status. There are three types ofhypotactic relation in the clause: CONDITION (expressed by clauses of condition, concession, cause. purpose, etc}, ADDITION (expressed by the non~efining relative clause) and REPORT. Paratactic and hypotactic structures may combine freely in a single clause comp1ex. We oondude this chapter with a few observations on clausal ellipsis in clause complexes. This is: a big topic. and we have not attempted to treat it in full. We confine ourselves to types of clausal ellipsis not covered by the discussion on verbal eJlipsis; essentially to question-ans:wer, and reporting- reported sequences. The general principle is clear: an elliptical clause of whatever type may presuppose any clause in a complex. and will then automatically presup- pose in addition all clauses that are contingent on it: that is, all that come alter it (if paratactic) and all that are dependent on it (ifhypotactic). So fur example: [4:138] a. Smith was going to take part. but somebody telephoned and asked to see him urgently so he had to withdraw.- Who? b. I kept quiet because Mary gets very embarrassed if anyone mentions John's name. I don't know why. In (a}, which is paratactic, who? coheres wi.th wmeboJy; the presupposed clause is somebody telephoned. and the remainder of the sentence also falls within the donuln of the presupposition: the meaning is "who telephoned and asked to see him urgently such that he had to withdraw ?•. Likewise in the hypotactic example (b), the meaning is 'I don•t know why Mary gets emb.arrusedifanyonementioruJohn'sname'. The first clause, in each case, is outside the domain of what is presupposed.

223 However, there are a number of restrictions and limitations on this principle, as well as possible sources of ambiguity within it. Ambiguity may arise because it is uncertain which is the clause that is being presup- posed, for example [4: 139] a. So you knew the lawyer was responsible. I hadn't realized. meaning 'that you knew .. .' or <that the lawyer was responsible •. Com- pare [4-: 139} b. I finished writing that story and it's going to be published.- When? meaning ·when did you fitrish ... ?'or 'when is it going to be published?'. A number of factors come in to detennine what is the likdy interpretation. Some of them are quite specific; for example, if I in [4: 139a] is stressed contrastively (tone 4), the second interpretation becomes overwhelmingly probable: •you knew that the lawyer was responsible, but I didn't'. But there are also rome general considerations. There is a tendency to presup- pose what is nearer; that is~ to presuppose an element that occurs later in the dause complex rather than an earlier one, especially if the later one is rather Jong. (This makes the contrast between a clause complex and a simple dawe with embedding (ranksbift) in it even greater. Embedded clauses tend to occur in later positions; and since they cannot be presup- posed~ the target of presupposition in such cases tends to be towards the beginning. Compare. [4-:94-) above; [4:94h] is a hypotu.."ti.c clause complex, whereas in [4.:94-<1], which is a simple clause with embedding, the appro- priate rejoinder would he Didn't he?, presupposing the polireman.) In particular it is unusual to presuppose a dependent clause that precedes the clause on which it is dependent; an example such as f4:14oa] is rather un- likely~ although there are instances, such as [4-:14-ob], which seem to pose no problem: [4: I40] a. Seeing that Mary's left something behind I really think we should turn ronnd and go back to the hotel r m not sure wha~ b. Unless he gives up one of his bishops be's going t-o be in trouble. It doem"t matter which. Often an indefinite form such as somebody is such a clear invitation to presupposition that its presence is sufficient to override any limiting ten- dencies of one kind or another. In a very complex structure there are many clauses which would be potential candidates for presupposilion. and

2a.f. lll.LIPSIS for precisely that reason presupposition by ellipsis tends to be avoided. since it would lead to too much ambiguity; moreover the clauses which occur later in the complex, and are therefore more accessible from the point of view of distance, are also. in the case of hypotaxis, at greater depth and therefore from another point of view less accessible- because :m elliptical clause which presupposed them could not be filled out. A sentence such as [4: 141 a J is intractable in this respect: [4; 141] a. I shall be cross if you bceak that vase, which was a present from my boy friend. -Which? -if Which? means 'which boy friend?' the 'filled out' fOrm would pre- sumably be Which boy friend wos that vase, if I break which you will i1e crass, a present Jrcm? Nevertheless, given a dear invitation to prempposition in the presupposed complex,. such instances become possible; if, for example, the first sentence of[4: 141a] ended which was a present frcm someone I love, the rejoinder Whc? would cause no difficulty; compare [4:141-b and c): [4: 141] b. I'd like you to look at the painting, which my wife picked up somewhere in the country. - Where? c. Smith said if he could afford it he was going to buy the next- door house and rent it to someone he knew so as to keep it from being pulled down to make way for a block of flats. - Who? Presupposition of a paratactic clause complex by clausal ellipsis is pos-- sible only if all clauses fo1lowing the one that is presupposed are within the domain of the presupposition. In practice this usually means that they must be branched; th:u is. they must share at leart one element in common, typicaUy though not necessarily the Subject. So we can accept [4: r423]. but hardly (4a.pb]o (4: I42] a. I )eft my hooks here and somebody came in and either borrowed them or put them back on the shelf hut didn't say a word to me. I wish I conld find out who. b. I lefr my hooks here and somebody complained and the librarian put them back on the shelf but didn•t say a word to me. I wish I could find out who. However, the conditions which determine acceptability are by no means clear. It does not always twn out as expected, and presupposition can extend over a considerable structural distance. We have often cited

4-4 CLAUSAL ELLIPSIS 22) exampks from A lire (>nd if example [3: 40] in this connection), so perhaps the last word should remain with her: 14:143] •And who are these?• said the Queen, pointing to the three gardeners who were lying round the rose-tree; for~ you see, as they were lying on their faces, and the pattern on their backs \\tas the same as the rest of the pack, she could not tell whether they were ganleners~ or soldiers, or courtiers, or three of her own chiidren. • How should I know?' said Alice.