1975. Helsinki - Presupposition, Assertion, and Discourse Structure.
By C. George SANDULESCU, Stockholm.
(Paper given by C. George Sandulescu at the Helsinki Conference, and published in Reports on Text Linguistics: Approaches to Word Order, edited by Nils-Erik Enkvist and Viljo Kohonen. Meddelanden från Åbo Akademi Forskningsinstitut, No. 8, pages 197 - 214.)
1.0 Abstract. It is a matter of common knowledge that discourse units are wired together in a specific way to form run-on texts. 'A sentence might be compared,' Enkvist (1975b) says, 'to a piece of flex with plugs at either end, and these plugs have to connect with corresponding plugs at the end of neighbouring sentences'. The task of the present paper is to point out that in addition to "wired" linkage between hypermorphemes, there is also "wireless" communication [or transmission], obtained via discourse presuppositions -- so far not investigated at all in spite of the plethora of work done on the sentential variety. Both the discourse assertions and the discourse presuppositions are here interpreted as hypermorphemes, and represent overt and covert contributions respectively to discourse structure. It is being suggested that the covert structuring of discourse is at least as important as its corresponding overt counterpart precisely as a result of its far-reaching consequences on the actual linearization.
1.1 Linguistic units and structural span. The most outstanding question in any discussion of word order is that of the theoretical frame of reference: word order -- within what frame ? The answer is being provided by the complex binary opposition between the micro- and the macro- levels as outlined in (1):
MICRO-LEVEL Below the Sentence:
1.1 micro- Morpheme/Word.
1.2 mini- Phrase/Group.
1.3 macro- Clause/Sentence.
MINI-LEVEL The Sentence:
2.1 micro- Simplex Sentence.
2.2 mini- Complex Sentence.
2.3 macro- Compound Sentence.
MACRO-LEVEL Above the Sentence:
3.1 micro- Hypermorpheme.
3.2 mini- Partial Sequence.
3.3 macro- Total Sequence.
The issue of word order has bearing upon the upper ranges of the hierarchy and is the outcome of presentative movement transformations, as described by Hetzron (1975), and triggered by constraints characterizing overall discourse structure. As pointed out by Kiefer (1971), early transformational work was 'inclined to relegate the problem of word order in toto to performance'. It is within such an approach that constituent ordering was placed within a clear sentence perspective (cf. Danes ed 1974). Thus, Functional Sentence Perspective takes the SENTENCE as its fundamental structural span, and deals essentially with units placed immediately under it in the hierarchy. FSP has in consequence at least one point in common with early transformational approaches, and tends to concentrate excessively on phenomena of the type --
(1) Deutsch spricht man in Österreich.
[ASSERTS:] (a) German is spoken in Austria.
(b) The language spoken in Austria is German.
[PRESUPPOSES:] Really good German is only spoken in Austria.(echt Deutsch...)
We here consider that Topic, Comment, and even Focus are manifestations of the presentative function in the syntactic linearization. The Focus/Presupposition binary opposition of a sentence model is replaced by the Assertion/Presuposition dichotomy within a discourse model, the linearizing operations being taken care of by the Presentative Movement.
1.2 Cue overlap. It has already pointed out that 'cues from levels other than phonotactics or lexis may play a crucial role' in investigating word order. Furthermore, the author goes on to state --
[ENKVIST 1975a] Many of the forces affecting the word order of a clause or sentence do not reside within that clause or sentence. [...]Some of the parameters affecting word order must be sought outside the affected clause or sentence, in the textual or situational context, in the universe of discourse, in the culture and knowledge shared by the communicants, and in other domains outside the sentence itself.
The Assertion/Presupposition interplay at discourse level is one such domain. A discussion of morpheme sequentialization within a consistent discourse perspective would bring in useful cues for sentential description from the area of the hypermorpheme; such cues are not provided by information coming from the language units below sentence level. An investigation of constituent ordering (morpheme vs hypermorpheme) within the structural span of discourse poses quite different theoretical problems. As has already been pointed out by Ruth Kempson (1975 : 123), 'the structure of discourse is not predictable in the way the structure of a sentence is'. One argument in support of positing discourse as a separate level of linguistic investigation is that central categories of conventional linguistics -- e.g., Competence/Performance -- no longer seem to hold at the level of discourse, where ordinary theoretical constructs, such as style and context, are modified beyond recognition as a result of pressure coming from the area of pragmatics. After identifying pragmatics with performance, Deirdre Wilson goes on to exclude context altogether from the range of preoccupations of a competence model:
[WILSON 1975 : 60] ... wherever context plays a part in the interpretation of a sentence it is the job of a pragmatic, rather than a semantic theory to determine the part it plays.
1.3 The semantics of discourse. Mainstream linguistics is definitely against incorporating parameters coming from supra-sentential levels into the standard theory: inter-sentence relations are often dogmatically excluded from a theory of semantics of the standard type:
[KEMPSON 1975 : 123] ... inter-sentence relations are excluded by fiat since they constitute part of a theory of discourse, not a theory of semantics. [...] The syntax of a simple sentence cannot be allowed to have a conjoint sentence as part of its underlying structure for two reasons: (i) there is no syntactic evidence to justify such an underlying structure ... (ii) general conditions on recoverability prohibit such an analysis. [...] Similar reasons militate against a semantic rule stating as part of the interpretation of a simple sentence that the preceding sentence has contained in it an indefinite noun phrase. To incorporate either type of rule into the grammar would be to transform the grammar predicting the sentences of some language into a grammar predicting the discourse of some language. [...] Thus any attempt to incorporate such a prediction into the grammar is in principle doomed to failure.
Restricting grammar to sentential description is indeed a sensible suggestion, all the more so as discourse does indeed evince its own specific patterning. One thing is certain: it might perhaps be wise to reconsider the significance of cues coming from areas of investigation situated above the sentence, for 'inter-sentence relations' cannot be 'excluded by fiat from a theory of semantics'. They are already part and parcel of a semantic component of a theory of discourse. And, to be sure, a theory of semantics can never be pitched against an overall theory of discourse.
2.0 SENTENTIAL PRESUPPOSITIONS VS DISCOURSE PRESUPPOSITIONS.
2.1 Sentential presuppositions. All discussions of presuppositions in both philosophy and linguistics have deliberately been conducted on decontextualized -- but easily contextualizable -- language data. By way of illustration, the reader is referred to the set of Russellian "King of France" arguments -- so very politically tinged --, which were invariably put forth at a time when the "present King of Spain" had not yet been politically reinstated. The whole discussion originated just before the turn of the century in philosophy and only happened to reach mainstream linguistics three quarters of a century later, after 1968. The issue has now reached a stage of acute crisis, as is excellently pointed out in Wilson (1975) and Kempson (1975). The deadlock seems to be so complete that a drastic suggestion is made -- the total eradication of the theoretical construct in question from all semantic theory. This is done with an arsenal of arguments mainly coming from the field of discourse.
2.2 Discourse presuppositions. Presupposition in philosophy goes at least as far back as Heidegger, who, according to Richard Palmer (1969 : 135) pointed to the 'impossibility of presuppositionless understanding'; in fact, Heidegger's 'interpretation is never a presuppositionless grasping of something given in advance' for the simple reason that 'the most presuppositionless interpreter of a text has preliminary assumptions'. I am therefore inclined to derive two distinct presuppositional traditions in philosophy: (a) one centred around Russell and Frege, which formed the essence of practically all theoretical discussion in the 20th Century, focused on Sentence/Presupposition/Statement, and developing very close ties with both formal logic and truth-conditional semantics; (b) the tradition of discourse presuppositions, derived directly from Heidegger, and reaching linguistics as a result of the systematic investigation of language units larger than the sentence via text linguistics and hermeneutics. It is in fact Heidegger's 'textual presuppositions', reinterpreted in a sociological frame of reference by Rommetveit (1974) that form the basis of the discourse presuppositions, first discussed by Muraki (1972). We thus add to the already exceedingly list of presupposition typology (cf. Allwood 1975, for conventional, non-conventional, and other presuppositions) by suggesting the dichotomy sentential vs discourse presuppositions. Allwood distinguishes between (a) lexical presuppositions (i.e., connected with lexical items), and (b) thematic presuppositions (i.e., connected with the information structure of the sentence). As part of a similar taxonomic-typologic approach to the same linguistic phenomenon a case is made in support of discourse presuppositions as inter-sentential semantic phenomena sustained through the discourse across sentential and participant boundaries as well as intra-textual boundaries of any other type. Allwood's 'presuppositional irony' is after all best manifested at discourse level and is ultimately a feature of discourse structure and constituent ordering (See Section 3.4 of this paper).
2.3 Towards a formula for discourse presuppositions. Starting from discourse provisionally defined here as "a non-conjoined chain of sentences", we tentatively define discourse presuppositions as "whatever has to be assumed across conventional sentence boundaries -- possibly subsequent to conjoiner-deletion processes -- in order for discourse to be completely meaningful". The only researchers so far to deal with discourse presuppositions have been Muraki (1972) and Landesman (1972). There are two fundamental definitional strategies that one could indeed adopt in defining the phenomenon: either (a) select the widest possible and most comprehensive definition, or (b) resort to the narrowest and most restricted possible one. We are going to adopt the former strategy and bracket a substantial amount of implicit meaning under the same heading. As all presuppositions are treated as entailments by Wilson (1975) and at least part of them as implicatures (cf. Gazdar 1975), the premises exist for a provisional methodological decision to bracket several logical constructs under the same blanket term. The definitional test of discourse presuppositions is not operated in terms of negation (e.g. 'statement which is kept stable under negation'), but rather in terms of discourse heads (cf. Sandulescu 1975) and their possible interrelationships.
In contradistinction to a sentence, a discourse contains (a) assertions, and (b) presuppositions; whereas most of the assertions can be adequately handled at sentential level, most of the presuppositions require a frame of reference wider than the sentence. One main reason is that they form an undercurrent which is never quite capturable at the level of the linear manifestation of the discourse. One common way of defining presuppositions is via a two-term relationship, according to the 'X PRESUPPOSES Y' formula (cf. Leech 1974 : 291). The 'X PRESUPPOSES Y' way of defining them in its turn presupposes that both X and Y are sentences or phrases (cf. Ducrot 1971 : 191 - 221). Secondly, there is the assumption that both X and Y are in a paradigmatic, rather than syntagmatic, relationship. But what happens in discourse ? A discourse approach to presuppositions foregrounds the syntagmatic axis in the first place, and the issue of co-occurence becomes more important than any other relationship. It must therefore be emphasized that the X / Y interrelationship may be either (a) paradigmatic, or (b) syntagmatic, or even (c) a complex conflation of both.
3.0 INTER-SENTENTIAL LINKAGE.
3.1 Overt inter-sentential linkage. It is a matter of common knowledge that discourse units are wired together in a specific way to form run-on text. 'A sentence might be compared,' as Enkvist (1975b) says, 'to a piece of flex with plugs at either end, and these plugs have to connect with corresponding plugs at the end of neighbouring sentences.' But this is merely one aspect of the problem: the overt aspect. Such overt connectedness may be achieved, as is known, by the following means: (a) definite noun phrases with the definiteness feature derived from the fact that there was prior mention of that particular noun phrase in the discourse; (b) similarly, extensively recurrent pronouns (as is quite the case in a tightly packed narrative discourse); (c) sustained use of tense throughout the discourse with one occurrence in close correlation to the other, particularly the issue of tense sequentiality and degree of indirection of the discourse; (d) the expression in various ways of the three types of deixis -- Time, Place, and Person -- in the linear manifestation; (e) conjoinability via conventional conjunctions.
The theoretical model we propose thus assumes the existence of an abstract connector K between sentences. The internal structure of this abstract unit maps out all features related to discourse connectedness. A systematic investigation of K structuring is of vital importance to the formulation of an explicit theory of discourse. The abstract connector evinces an overt structure covering all connectedness markers emerging in the linear manifestation of the discourse: semantic significance of NP recurrence, sustained tense and time deixis, sustained place deixis, explicit conjoining via (a) conventional, and (b) unconventional conjoiners. All this builds the overt matrix Ko of the abstract connector K, which has both psychological and communicative reality in the complex process of discourse mapping (easily triggering rephrase phenomena in spoken discourse). Given its reality in the communication act, K should in no way be viewed as a mere inventory of random characteristics at meta-level, very much like the chaotic jumble of plugs on the floor of any good recording studio of any modern radio station. (The structure and function of the abstract connector K will be discussed in far greater detail in a forthcoming paper, and its communicative significance will be then placed in the right perspective; cf. Sandulescu 1976).
In addition to "wired" linkage (i.e. of the overt kind) between hypermorphemes, there is also "wireless" communication. By the side of the overt connectedness markers discussed above, there is a complex set of covert connectedness markers, which in their turn build a matrix, quite symmetrical in shape, but not at all similar in content, to the overt matrix Ko. These covert discourse-continuity markers form a separate matrix: the covert matrix Kc of the abstract connector K.
Such abstract matrices of both types necessarily exist in both discourse participants -- the sender and the receiver -- at inter-sentential boundaries: an adequate description of their structure is an essential part of a theory of discourse. The present discussion does not deal with the structure of the overt matrix Ko, as many of the overt characteristics have already been much discussed (cf. Halliday & Hassan 1976). For purely methodological reasons the overt abstract connector Ko is kept as empty as possible of discourse connectedness, mainly at the level of the linear manifestation of two-hypermorpheme discourse. This particular procedure has been resorted to in a deliberate attempt to delete "wired" linkage, and restrict intercommunication to the "radio'd" variety; in other words, attention will be focused on the covert matrix to the total exclusion of the overt one (as much as this is humanly feasible).
3.2 Covert inter-sentential linkage. In his discussion of sentence grammar dating back to 1957, N. Chomsky starts with one initial symbol -- that of S. In consequence, any conventional sentence-based linguistics takes as its starting point the same initial single symbol. A discourse model, however, even if it is of the simplest possible variety, must necessarily start from the following three initial symbols, evincing a close correlation between them:
(3) Sx Sy Sz
These correlated initial symbols are hierarchically organized. The present discourse model does not begin with syntax, but foregrounds the semantic parameters from the start, assuming that though discourse evinces definite structuring, it does not evince syntactic parameters of the conventional kind, which are commonly handled by any generative-transformational grammar (for a similar view, cf. Ruth Kempson 1975 : 123). Not only do we approach the problem from the semantic angle of vision, rather than the purely formal and syntactic one, but we foreground the most recent and very controversial area of semnatics, namely the theory of presuppositions. As was already mentioned, the current presupposition formula is --
(4) X PRESUPPOSES Y
where both X and Y are assimilated to hypermorphemes. As a result we can rewrite the above formula as --
(5) Sx PRESUPPOSES Sy
where sentential hypermorpheme x presupposes sentential hypermorpheme y. In discourse, however, this formula is radically changed in that discourse presuppositions operate on the following far more complicated pattern:
(6) The interrelationship of syntagmatically co-occurring Sx and Sy can only be established on the basis of an underlying Sz, where all symbols -- Sx, Sy, and Sz -- are to be interpreted as sets.
Thus, a two-hypermorpheme discourse pQ of the type --
(7) pQ ==> Sx + Sy
would presuppose the existence of at least one background sentential assumption or underlying statement belonging to the set Sz; in extreme cases, the set Sz may be open-ended as regards the possible number of underlying statements. In this way, discourse presuppositions carry with them three-term relationships which generally replace or supplement the more primitive and more conventional 'Sx PRESUPPOSES Sy'. Here is an extreme example in point:
(8) "Where is the bicycle ? I want to shave this morning !"
This triangular hypermorphemic interaction implies that the first two of the three elements involved are linearized; they are, in other words, part and parcel of the linear manifestation of discourse, whereas the third element, operating as a semantic link between the two, is never part of the same linearization. In terms of more traditional linguistics, if we assume that in syntactics we are always faced with the two axes -- one syntagmatic and the other paradigmatic (whichever way Homski may wish to beautify them) -- it is Sx and Sy that are always in a relation of overt co-occurrence, with Sx in initial position (functioning as The Precedent), and Sy in 'final' position (functioning as The Subsequent). Sz is in a paradigmatic relation to both Sx and Sy. Such a relation is always and diagrammatically triangular:
(9) Sx -------->-------------->----------> Sy
In a syntagmatic relationship, the dotted arrow indicates the linear direction of discourse, whereas the obvious position of Sz quite outside the linearization here indicates a paradigmatic relation between the surface manifestation and the depth interpretation. For reasons of concision and simplicity, we disregard the assertion aspects of the discourse and concentrate on its presuppositional aspects; we also disregard the possible relations between different Sz's as they clearly emerging in multi-hypermorpheme discourse, and concentrate exclusively on two-hypermorpheme discourse, where there is only one single set of Sz apparent. A correct interpretation of the triangle relation given in (9) emphasizes the passage from a sentence model of description (with a conjoining vulnerability quotient) to a discourse model. Other interesting examples illustrating the triangle relationship are the following:
(10 a) "Where is the soap ? I'm hungry !"
(10 b) The writer told the story of his future novel to his audience. She immediately wrote a letter to her brother John.
(10 c) They were asked to work more quickly in order to finish very early. He fell ill all of a sudden and went to bed.
(10 d) It is clear that he made a mistake. The city is far bigger than I thought.
(10 e) They are learning English. He has been playing foot-ball.
To summarize: Sx and Sy co-occur linearly; they both presuppose an underlying assumption. This three-term relation can be rephrased as follows:
(11 a) Sx PRESUPPOSES (part of) Sz
b) Sy PRESUPPOSES (part of) Sz
where all three symbols are assigned fixed positions in the hierarchy. There is furthermore a relation of identity between the two Sz's, and an obvious relation of diversity between Sx and Sy. We repeat that each of the three symbol stands for stets of one or more than one items. This set prerequisite is included into the model in order to permit an easy switch from the investigation of two-hypermorpheme discourse to that of multi-hypermorpheme discourse. The set provision also facilitates reference to the notions of research focus in relation to structural span. For this very reason, in addition to subscripts of the first degree -- Ox, Oy, Oz (Precedent, Subsequent, and non-Emergent) -- we propose the use of subscripts of the second degree for descriptive purposes, thus --
(12) Ox1 Ox2 Ox3 Ox4 Ox5 Ox6 ...
A two-hypermorpheme discourse can in this way be turned into a multi-hypermorpheme discourse by a corresponding expansion of the structural span. The focus is kept fixed on the Sx/Sy interrelationship, which need not be symmetrical. It must be emphatically pointed out that the above-outlined notational system describes an exceedingly complex communicative reality. The question of underlying statements in multi-hypermorpheme discourse is of fundamental significance: there is open-endedness in two directions -- overt vs covert --, one of which is non-emergent in the linearization.
3.3 The ultimate nature of the Sz. There are two fundamental kinds of Sz: (a) one that is kept constant throughout the discourse, and summarizes the global constraints on discourse, which in their turn are of three fundamental types -- syntactic, semantic, and pragmatic; (b) the shifting part of the Sz (in contradistinction to (a), which is the constant part of it), containing discourse features and constraint specifications which do change with the passage from one abstract connector to the other. This dichotomy is ultimately a continuum, on which a further intermediate stage (c) of zonal constraints establishes a bridge between the clearly permanent features of the discourse, subsumed under (a) above, and those subsumed under (b), which are strictly local constraints. GLOBAL, ZONAL, and strictly LOCAL are essentially methodological cuts on a continuum. This is in agreement with the following:
(13) All global features and constraints remain constant throughout the total sequence tQ.
(14) All zonal features and constraints remain constant throughout the partial sequence pQ.
(15) All local features and constraints characterize solely the structure of one single matrix of the abstract connector K; they are not recurrent in any of the senses in which both global and zonal features are, for the simple reason that as soon as they characterize two abstract connectors instead of one, they become zonal, rather than strictly local, features.
Working towards a discourse model which should incorporate, rather than eliminate, conventional sentence linguistics and which is bound to make connectedness explicit has its advantages: it provides among others, via the overt and covert connectors, a homogeneous treatment of heterogeneous features (cf. Pike 1967 : passim; cf. Weinreich, Labov & Herzog 1968 : 3.21). This means placing the syntactic on apar with the pragmatic; the semantic features are so overwhelmingly important that they are clearly superordinated to the other two components. Devised expressly with a view to incorporating the various features of the communication act, the model can easily cope with the notion of Participant Boundary (cf.Sandulescu 1975), coming from the area of micro-sociolinguistics. Finally, the treatment of absolutely all discourse features is a synthesis at another linguistic level -- that of discourse -- of the two or several tree-diagrams of the kind we got so accustomed to in the wild, wild forest of gradually diverging generative approaches: it is simply an attempt to co-ordinate the structural information carried by the tree-diagram.
3.4 Prolegomena to semantic distance. Actual Sz structuring is determined among others by the semantic distance existing between co-occurring Sx and Sy:
(16) If the semantic distance between Sx and Sy is exceedingly great, then the set Sz is substantially overloaded.
(17) It is excessive overloading of the covert matrix of the abstract connector K that actually leads to the rejection of co-occurrence potentialities between Sx and Sy.
There are basically two kinds of illustration material of this phenomenonin two-hypermorpheme discourse: (a) there is one kind of language data involving connectedness across participant boundaries; (b) there is another kind of material -- the so-called run-on text, to resort to a looser terminology. The abstract connector is structured differently in the two cases in that it is "world-homogeneous" in one case, and "world-heterogeneous" in the other (world is here used in the Hintikka 1969 sense of "possible worlds", and not far at all from "universe of discourse" which is also gaining increased currency ...). The present approach to discourse connectedness thus attempts to avoid the uneasy relation so far existing between the semantic, pragmatic, and syntactic components of the model. There are two fundamental problems related to semantic distance between the hypermorphemes; these problems are somewhat similar to the issue of semantic distance between (content) morphemes: (a) one is the adequate theoretical explanation of the phenomenon, which, via discourse-head interrelationship, also covers the whole theory of figures in poetics, e.g.
(18) She dropped a tear and her handkerchief.
(b) the other one is the actual measurement of such distance. Measurement operations are, however, too heavily dependent for the time being on (i) pragmatic factors, (ii) overall discourse structure, and (iii) discourse typology. This particular point can easily be illustrated with data related to morpheme sequentialization:
(19) //DISCOURSE// (she dropped a tear.)(she dropped her handkerchief.)
(20) //SENTENCE// (She dropped a tear and her handkerchief.)
It is interesting to note that the operation of presentative movement transformations on such sentences directly affects the issue of semantic distance, as a matter of course considerably increasing it:
(21) It is a tear that she dropped, and her handkerchief.
(22) A tear she dropped, and her handkerchief.
(23) Her handkerchief she dropped, and a tear.
(24) Her handkerchief and a tear she dropped.
Measuring semantic distance means in other words measuring such distance between discourse heads at sentence level, and then analysing in a preliminary stage the effects produced by the presentative sequentialization of morphemes; in a subsequent stage, the procedure is analogical, but not qualitatively different, at discourse level, where the question is the degree of match and mismatch between hypermorphemes similarly linearized. It is only on the basis of the premises provided by semantic distance in discourse structure that we can begin to discuss the implications for the overall patterning of text derived from surface object, often called comment, and surface subject, often called topic.
4.0 ULTIMATE NATURE OF WILSON'S AND KEMPSON'S DEMONSTRATION DATA.
4.1 Discourse in disguise. Most of Wilson's examples can be turned into discourse on the basis of a very few simple processual rules of disjoining. As soon as the sentential demonstration data is turned into discourse, strong evidence is obtained in support of discourse presuppositions. It is worth remembering that discourse can and does indeed accommodate contradictory presuppositions in its structure, whereas the sentence does not. A case in support of discourse presuppositions concurrently maintains the validity of both argument and demonstration data against sentential presuppositions. Wilson's discussion of contradictory presuppositions at sentence level indicates how necessary it is to pay due attention to the Conjoiner Deletability Principle (cf. Sandulescu 1975).
(25) (Wilson 1975 : 51) //SENTENCE// No-one has ever claimed that Hiawatha was a communist: the claim that Hiawatha was a communist, then, has not destroyed Bill's faith in Longfellow.
(26) //DISCOURSE// No-one has ever claimed that Hiawatha was a communist. The claim that Hiawatha was a communist, then, has not destroyed Bill's faith in Longfellow.
(27) (Wilson 1975 : 49) //SENTENCE// Bill Bloggs might be here -- I don't know whether such a person exists.
(28) //DISCOURSE// Bill Bloggs might be here. # I don't know whether such a person exists.
(29) (Wilson 1975 : 73) // SENTENCE // If your teacher was a bachelor you were lucky, but if your teacher was a spinster you were unlucky.
(30) //DISCOURSE// If your teacher was a bachelor you were lucky. # If your teacher was a spinster you were unlucky.
In (30) for instance, discourse formation involves the following distinct operations, provisionally called "processual rules of disjoining": (a) but deletion; (b) period insertion; (c) initial capitalization. The processual rules turning compound and even complex sentences into discourse are either paraphrase rules, or graphematic rules. Very simple paraphrase rules are provided through the operation of the Conjoiner Deletability Principle, leading to the deletion of most co-ordinating conjunctions, and some of the subordinating ones. Subordinate-conjunction deletion must be accompanied by corresponding paraphrase in order to maintain intact the overall semantic interpretation of the discourse. Thus in a sentence like --
(31) (Wilson 1975 : 77) //SENTENCE// Either Joanie accused her husband of reading her letters or she accused him of not reading them
the discontinuous conjoiner either ... or ... can be rewritten as an adverbial conjoiner, such as alternatively. In fact, sentence (31) passes through the following stages:
(32) Either X accused Y of doing Z, or X accused Y of not doing Z.
This, in turn, can be rewritten as --
(33) Either (Sx) or (Sy), where (Sx (X accused Y of doing Z)) (Sy (X accused Y of not doing Z))
Then, (33) can be rewritten as discourse through the operation of a paraphrase rule of reinsertion of the non-conventional conjoiner into the linear manifestation of the discourse in order to preserve the initial semantic interpretation intact:
(34) (D (Sx (X accused Y of doing Z)) (K (alternatively)) (Sy (X accused Y of not doing Z)))
where alternatively is an adverbial replacement for the initial discontinuous conjunction. Both assertive and presuppositional structuring stay unchanged in the sentence-to-discourse transition. Thus, according to Wilson (1975 : 77), sentence (31) --
(35) //ASSERTS// Either Joannie indicated that her husband read her letters or she indicated that he did not read them.
(36) //PRESUPPOSES// (a) Joanie judged that reading her letters was bad. (b) Joannie judged that reading her letters was not bad.
All this makes us reshuffle (31) in the shape of the following two-hypermorpheme discourse:
(37) //DISCOURSE// Joanie accused her husband of reading her letters. Alternatively, she accused him of not reading them.
Such two-hypermorpheme discourse, in addition to preserving semantic appropriateness in point of conjoining, keeps unchanged the corresponding set of contradictory presuppositions. Besides disjoining, the only change that we suggest here is that the sentential presuppositions are to be interpreted as discourse presuppositions, primarily by virtue of the fact that the language segment under research focus has undergone discourse formation. There are other, more complex, instances of discourse formation among the wealth of examples put forth by Wilson in her attempt to delete Presupposition from a semantic theory of sentence linguistics:
(38) (Wilson 1975 : 75) //SENTENCE// If Sartre knows that Chomsky is alive, I'll be surprised, but if he knows that Chomsky is dead, I'll be amazed.
//PRESUPPOSES// (a) Chomsky is alive. (b) Chomsky is dead.
(39) (Wilson 1975 : 134) //SENTENCE// If John had left I would have seen him go; but I did see him go, so he did leave.
(40) (Wilson 1975 : 134) //SENTENCE// IBM has admitted that their computers often turn into frogs, so their computers obviously turn into frogs.
4.2 Defining discourse presuppositions in relation to testability conditions. The present, very brief discussion has selected the problem of presupposition and has related it to sentence structure and to corresponding discourse in order to bring out in bold relief the flimsy nature of the boundary. In consequence, basing very strictly a discussion of word order on such a boundary may have its dangers.
In order to emphasize the overwhelming significance of the underlying stream of implicit meaning in any discussion of linearization aspects as well as constituent ordering, we propose to define discourse presuppositions in the following way:
(41) A discourse presupposition is any background assumption which enters into the structure of the abstract connector K -- necessarily underlying discourse connectedness -- and thereby establishing semantic links between hypermorphemes within the discourse. Testability in discourse lies in the mismatch between emergent and non-emergent hypermorphemes, measured in terms of semantic distance. Some such presuppositions are discourse-internal, others (e.g. the existentials) are most often discourse-external.
5.1 Discourse semantics cannot do without presuppositions, though it may be conceivable that sentence semantics could do without them, as is suggested by Wilson (1975) and Kempson (1975). Presupposition thus becomes one of the most fundamental theoretical constructs of an emerging theory of discourse.
5.2 Discourse normally accommodates contradictory presuppositions between linearized hypermorphemes (e.g. the "to be or not to be" literary tradition ...), though it operates at assertion level ...
5.3 If sentential boundaries are relative, and so heavily dependent on alternative paraphrase, the actual frame of reference in which we speak of constituent ordering must still be an open theoretical question. Parameters vary extensively if we switch from grammatical structure to discourse structure.
5.4 Postulating the existence of an abstract connector K is the only way of providing an explicit description of connectedness in discourse.
5.5 The abstract connector K evinces an overt matrix Ko, structuring emergent information across sentential boundaries.
5.6 The abstract connector K evinces a covert matrix Kc, structuring non-emergent information across sentential boundaries.
5.7 The abstract connector K and the linearized hypermorphemes Sx and Sy are likely to be the fundamental units of discourse.
5.8 Discourse mapping, as a complex process endowed with both psychological and communicative reality, deals both with what is emergent -- e.g. linearized constituent ordering -- and with what is non-emergent -- the discourse presuppositions -- as well as with the complex interplay between the two.
5.9 The issue of semantic distance between both discourse heads (as constituent morphemes) and hypermorphemes, and the investigation of the empirical possibilities of measuring it form the most likely area of fruitful research for the future.
5.10 The presentative function, as discussed by Hetzron (1975), places the whole issue of constituent ordering in a new light. The balance between assertion and presupposition and the presentative movement ultimately determine the linearized structure of discourse and represent the real triggers of constituent ordering.
REFERENCES to "Presupposition, Assertion, and Discourse Structure":
ALLWOOD, J. 1975. 'Conventional and Nonconventional Presuppositions', in: Papers from the Second Scandinavian Conference of Linguistics, edited by E. Hovdhaugen, pp. 1-15.
DANES, F., ed. 1974. Papers on Functional Sentence Perspective, Mouton.
DUCROT, O. 1972. Dire et ne pas dire, Paris.
ENKVIST, N.-E. 1975a. 'Prolegomena to the Symposium' (in this volume).
ENKVIST, N-E. 1975b. 'The relevance of Theme, Rheme and Focus to the Description of Adverbial Placement in English', (in this volume).
GAZDAR, G. 1975. 'Implicature & Presupposition', Paper presented at the SLE/LAGB Conference, Nottingham, April 1975.
HALLIDAY, M.A.K. & R. Hasan, 1976. Cohesion in English, Longman.
HASAN, R. 1968. Grammatical Cohesion in Spoken and Written English, Longman.
HETZRON, R. 1975. 'The Presentative Movement' in: Ch. N. LI, ed., Word Order and Word Order Change, 1975, pp. 346-88.
KEMPSON, R. 1975. Presupposition and the Delimitation of Semantics, Cambridge U. P.
KIEFER, F. 1971. 'On the Problem of Word Order', in: M. BIERWISCH & K. E. Heidolph, eds, Progress in Linguistics.
LANDESMAN, C. 1972. Discourse and Its Presuppositions, Yale U. P.
LEECH, G. 1974. Semantics, Penguin.
LI, Ch. N., ed. 1975. Word Order and Word Order Change, Proceedings of the 1974 Conference on Word Order, Santa Barbara, California.
MURAKI, M. 1972. 'Discourse Presupposition', in: Papers in Linguistics, vol. 5, No.2 (Summer 1972), pp. 300-320.
PALMER, R.E. 1969. Hermeneutics, Northwestern U.P.
PIKE, K. 1967. Language in Relation to a Unified Theory of the Structure of Human Behavior, Mouton.
ROMMETVEIT, R. 1974. On Message Structure, London.
SANDULESCU, C.-G. 1975. 'Displacement Constraints on Discourse', in: Papers from the Second Scandinavian Conference of Linguistics, edited by E. Hovdhaugen, pp. 282-313.
SANDULESCU, C.-G. 1976. 'The Structure of Discourse Connectors' (forthcoming).
SINCLAIR, J. 1975. 'Discourse in Relation to Language Structure and Semiotics', (mimeographed).
WEINREICH, U., W. Labov & M. I. Herzog 1968. 'Empirical Foundations for a Theory of Language Change', in: Directions for Historical Linguistics, edited by W. P. Lehmann & Y. Malkiel.
WILSON, D. 1975. Presupposition and Non-Truth-Conditional Semantics, Academic Press.