1976. Hanasaari - ONLY CONNECT ...

by C. George SANDULESCU, Stockholm.

(Paper given by C. George SANDULESCU in 1976, at the Third Scandinavian Conference of Linguistics, which took place at Hanasaari, near Helsinki, in Finland, between 1 and 3 October 1976; the Proceedings, edited by Fred KARLSSON were issued by the Text Linguistics Research Group of the Academy of Finland, Turku/Åbo, 1976, 404+16 pages.)

[Acknowledgements:    I should like to avail myself of this opportunity and thank the participants in the discussion of the present paper at the Helsinki meeting -- V. Sørensen (Århus),  C.Kock (Copenhagen), Einar Haugen (Harvard and Uppsala), R. Leite (Oslo)  and others as well as the chairman of the session  H. Basbøl (Odense) whose contributions have certainly improved the present version. A verbatim transcript of the discussions has already been issued by the Discourse Analysis Project, Department of General and Theoretical Linguistics, University of Stockholm, Sweden.]

                                                   [Full Text]

(Epigraph by Ludwig Wittgenstein, 1938 :)   I have often compared language to a tool chest, containing a hammer, chisel, matches, nails, screws, glue. It is not a chance that these things have been put together -- but there are important differences between the different tools -- they are used in a family of ways -- though nothing can be more different than glue and a chisel.

1.1    Outline of the problem. The present study is devoted to the discussion of the following relations --

(1)                        Cn            {    p ,  q }

(2)                        C [+s] / (n        ,     0)            ===>        K  ( 0   ,   c  )

where p and q are current hypermorphemes in sequentialization.  Cn is a standard or non-standard conjoiner (fn 1) actualized in the linear manifestation of discourse, whereas K is an abstract connector with its overt and covert matrices, as defined in Sandulescu (Texas, April 1976).  The main theoretical problem under discussion is that of the possible correlation between an actualized conjoiner C -- with its whole range of lexicalizations in a given natural language -- on the one hand, and the complex bundle of features going into the making of Ko and Kc   structures in the process of monitoring discourse in both production and perception. It is advanced that conjoiner status in discourse, particularly within a frame of reference of a consistent real-discourse model which should meet tight requirements of descriptive adequacy, is by no means identical with 'propositional connective' status in symbolic logic, or 'cojunction' status in conventional grammar, and should in no way be confused with them.

1.2    Connectedness. The itemizable category of conjoiner, as viewed in the present model of discourse, is highly dependent on the theoretical construct of connectedness, circulated in topology and algebraic linguistics (cf.Saloni et al. 1974). It is superordinated to the less satisfying notions of cohesion (Halliday & Hasan 1976), coherence (as differentiated from cohesion by H. Widdowson), and connexitivity as employed by language statisticians and quantitativists. Ongoing research into discourse structure shows that conjoiners tend to form open sets, analysable in terms of set theory, rather than closed  inventories, as hypothesized by conventional (including transformational) grammars.

1.3    The symbols p and q in symbolic logic. Taking p and q to stand for atomic propositions, more complex statements can be formed by connecting them in order to obtain molecular propositions. This is done via connectives: propositional connectives in modern logic derive from the conjunctions existing in natural language, but are defined explicitly by means of truth tables, logic having originally emerged as a purely normative discipline (fundamentally connected with truth-values in effective argumentation). The exact sense of logical connectives thus rests on their truth-functional definition:

(3)                                p                    C                    q

                                    0                      0                    0

                                    0                      0                    1

                                    1                      0                    0

                                    1                      1                    1

        If C is ascribed the lexicalization 'and', then the truth table is correct, and the conjuntion p & q is true provided p is true and q is true.  Assuming that --

(4)                                p        (  Aristotle is Greek  )

                                    q         (  the table is brown  )

the connection thus obtained --

(5)                        ( p & q )     ( Aristotle is Greek and the table is brown )

is acceptable in logic on the strength of its truth-value function exclusively. But this, it must be pointed out, is an excessively narrow theoretical foundation to base a discourse model on. Here is what an outstanding logician has to say regarding the goals of logic:

(CHURCH    1956 : 1)     Our subject is [ ... ] formal logic. Traditionally, formal logic is concerned with the analysis of sentences or of propositions and of proof with attention to the form in abstraction from the matter.

Imposing, however, even minimal requirements of descriptive adequacy on a real-discourse model, p & q as conjoined in ( 5 ) hardly meets connectedness conditions in any possible communicative situation. The same holds good if we delete the conjoiner and/or play with tense:

(6)                            Aristotle was Greek, the table is brown.

This does not meet connectedness conditions for the reason that matter (i. e. the semantic interpretation) -- in the sense given it by Alonso Church -- becomes in a real-discourse model as important as form, and perhaps, even more so. Within such a model, it is the function of the twin K matrices to fuse matter and form in a unified descriptive procedure, not only by extending conjoiner range to cover all non-standard items, but also by providing a subtheory of discourse heads -- as sketchily outlined in Sandulescu (New York, March 1976) -- to take care of matter. An overt or covert sharing of identical or related discourse heads is a fundamental connectedness constraint, imposed on any p q sequence in order to build appropriate discourse.

(7)                                (Aristotle was Greek. His writing table was always brown.)

        In ( 7 ), the two items his and was are to be represented as 'sustained Person' and 'sustained Tense' respectively within the overt K matrix, whereas writing is only assigned a place as a connectednes marker within the covert matrix of the same K connector on the basis of the speech-act participants' mutual factual knowledge of a certain possible world (which incidentally also accounts in part for the use of Past Tense in the latter hypermorpheme:  the proper name Aristotle certainly functions as a covert tense marker). Lastly, always occupies an even more remotely covert position in the K hierarchy, fulfilling multiple functions (mainly 'diachronic' in implication) at the levels both of the given possible worlds and of a certain universe of discourse. The 'atmospheric' impact on discourse of such items is so far uninvestigated, having only been analysed impressionistically in stylistics. Worth an analysis in the above manner are also discourses such as --

(8)         Aristotle was Greek. He never had a writing table and it was always brown.

(9a)       Aristotle was Greek. His writing table will have been brown.                

(9b)       Aristotle was Greek. His gestures will have been Mediterranean.

        In these oversimple examples, which come nowhere near the complexity of an actual text, the obvious methodological restrictions imposed upon the data are that we have confined our remarks to two-hypermorpheme discourse only: the precedent (i. e. the first hypermorpheme) was kept constant, and only one central discourse head (viz. Aristotle) was resorted to. 

1.4    Connectives as the 'glue' of language. Turning again to fairly elementary symbolic logic for a statement of phenomena from the formal viewpoint, several types of propositional connectives can clearly be distinguished. Generally speaking, logical connectors are symbols which may be used together with one or more propositions to form or produce a new proposition. They in fact play such an outstandingly important role in the language, be it natural or formal, that it is next to inconceivable to have them eliminated from it. But connectors not only 'glue' propositions together, they 'control' propositions. It is on the basis of this control rather than 'conjoining' function that there are three distinct types of truth-functional connectives TFC in mathematical logic:

(a)    unary, or singulary, or one-place TFC's, controlling one single proposition;

(b)    binary, or two-place TFC's, controlling two propositions (the precedent and the subsequent in the terminology of our discourse model),

and finally,

(c)    n-ary, or n-place TFC's, controlling more than two propositions.

        In point of fact, there is no upper limit to their conjoining capabilities, as TFC's can logically be defined for any number of constituent propositions. However, natural language lexicalizations of propositional connectives only evince a two-place pattern of the standard formula --

                                        Cn    {    p    ,    q    }.

In consequence, conventional linguistics and all descriptive grammars deal with the middle type only in the above sub-categorization of logical connectives -- that of the standard conjunction.

        The problem we are faced with at this stage is expressible in terms of what should be given conjoiner status in a real-discourse model in order to meet connectedness constraints which are at least remotely analogous with those in symbolic logic. 

1.5    Objections to connective sub-categorization in logic. Contemplating the threefold conjoiner typology, however, the important remark must be made that modern symbolic logic has concentrated to such an overwhelming extent on two-place connectives -- conjunction, disjunction, implication, and equivalence -- primarily because they represent the outstandingly common type of connective that is lexicalized in natural language. The three-fold classification also points to the equally important fact that there can be no question of a 'zero-place connective' even remotely analogous to Montague's zero-place operation symbol (cf. 1974 : 99)(fn 3). And this simply because, in Wittgenstein's terms, a connective should be 'the glue of language', fundamentally presupposing at least one item to be organically tagged to at least one other item via 'this very thing', which may be variously called connective in mathematical logic, conjunction in conventional linguistics, and (standard vs. non-standard) conjoiner in our own model of discourse structure. But if a zero-place 'connecter' -- to use Lord Quirk's (1972 : 661ff) new-fangled spelling -- is a misnomer in all these disciplines, so is a one-place connecter in at least some of them.

(CHURCH 1956 : 36)    The chief singulary sentence connective we shall need is one for negation. In this role we shall use, in formalized languages, the single symbol  ~ , which when prefixed to a sentence, forms a new sentence that is the negation of the first one. The associated function of this connective is the function from truth-values to truth-values whose value for the argument falsehood is truth, and whose value for the argument truth is falsehood.

        It is very hard to conceive of connectedness constraints on discourse  structure allowing for a 'one-place' conjoiner, analogous to negation, the singulary connective of symbolic logic. This is indeed a major point of divergence as regards connecter theory between the two disciplines. Negation does not in any way have a connective function, worth  that name, in discourse. No possible interpretation of negation in discourse can assign it a higher status as a connectedness marker, be it syntactic, semantic, or pragmatic, than that which may be accidentally assumed by any other non-conjoiner (fn 4).     

1.6    More than two place connectives. In a very recent and interesting paper, Gazdar & Pullum (1976) point to the fact that natural languages lexicalize only an extremely small range of TFC's. And the authors go on to demonstrate by using truth-value tables and other arguments that "the number of logically definable TFC's turns out to be literally greater than infinity". The real-discourse model that we are operating in requires that TF constraints on conjoiners be dropped are replaced by the pragmatics-oriented category of connectedness constraints, which take into account not only form but also matter (cf. the subtheory of discourse heads) and the attitudes of speech-act participants. Within this entirely different frame of reference, it still remains a very interesting suggestion to hypothesize  the existence of -ary conjoiners (cf. Section 3.5 of the present paper). 

2.1    The zero conjoiner. There is then the issue, by no means clear in interpreting discourse structure, of the zero conjoiner, which is quite different from the non-existent zero-place connective mentioned above. Such a conjoiner either does not emerge very often in the linear manifestation of discourse (the restricted approach), or it does emerge, literally, all the time (the comprehensive approach). Adopting the narrow approach, we may say that whenever such a conjoiner does emerge, it disturbs both intonation patterns  and conventional punctuation (i. e. graphemic) systems; here it is, provisionally illustrated by two sets of data, one in Swedish, the other one in Rumanian:

(10)      Man ska inte bara bo -- man ska trivas. (current Stockholm housing advert)

(11)      Nu mergem acasa, (ci) mergem la cinema.

        In the Rumanian example, the insertion is possible of an optional conjoiner ci, but the fairly low frequency of occurrence of this standard conjoiner in most discourse types makes its reinsertion rather improbable. Something somewhat similar may perhaps be said of the Swedish example (cf. utan, perhaps correlated with också). But in English sentences of the type --

(12)        We are not going home, we are going to the cinema.

no standard conjoinder is conceivably insertable; in writing, the linear manifestation takes the shape of two hypermorphemes not separated by period, but united by a comma. Even the semi-colon would be inappropriate.

2.2    Insertability vs. deletability. In Sandulescu (Oslo, April 1975) the complicated issue was discussed of the following principles: (a) the conjoiner insertability principle, as suggested by Katz & Fodor in 1963, and by virtue of which any text could become a sentence by infinite and indefinite and insertion; this is nowadays widely rejected by most linguists. There is at the other end of the scale, (b) the conjoiner deletability principle, suggested for the first time by Sandulescu (in the same paper) as a primary means of discourse formation, and as a cardinal communicative operation in monitoring discourse in reception. Hypothesizing such a principle also throws new light on the theory of paraphrase. The question is, particularly in the light of the conjoiner deletability principle, whether all instances of deleted conjoiner should or should not be regarded as instantiations of the 'zero conjoiner'. In this way, however, any two sentences separated by period, but united by a common semantic interpretation should indeed be connected by means of a zero conjoiner ! And to facilitate it, a symbolic-logic approach even discards, as we have seen earlier in the present study, the requirement of an inter-related or mutually connected semantic interpretation, ascribing all truth-value to the conjunction p & q, provided the two hypermorphemes are true, when taken separately. As the issue of the zero conjoiner is an extremely complex one, the solution of which depends on the completion in greater detail of other areas of the discourse model that we have adopted, we would like to leave the question open for further discussion, restricting, however, the use of the zero conjoiner to the instances in which the language under investigation does have a fully lexicalized standard item, which is optionally insertable in order to fulfil a standard-conjoiner function in the linear manifestation of the discourse, as illustrated by the Swedish and Rumanian examples // DATA //  in (10) and (11). 

2.3    Deep structure status ? At a time when certain generativists (cf. Chomsky 1976, Reflections on Language, Fontana, as reviewed by J. Searle, TLS, September 1976) seem to be dropping the notion of deep structure in favour of a modified version of surface structure, it would be preposterous on our part to postulate and hypothesize abstract levels and even more abstract levels of investigation, the 'existence' may be disproved within the short span of only a few years. There has been far too little investigation of discourse to justify positing the issues of both 'deep structure' and 'generation' of discourse. However, hypothesizing a zero conjoiner remains an interesting suggestion, particularly for at least some of the cases that D. Wilson (1975 : 33, 78, 84) presents as single sentences, but which Sandulescu (Åbo/Turku, November 1975) interprets as discourse, such as --

(13)        I just knew I'd win -- I can't see how I lost.

(14a)      Mary didn't clean the room: it wasn't dirty.  

(14b)      I didn't clean the bathroom: I cleaned the kitchen.   

(15)        Harry didn't criticize Bill for being the last man out of the room: he criticized Charley.

        The postulation of a zero conjoiner is, we repeat, fraught with dangers, and it is only a correlation between the linear manifestation of discourse and its corresponding semantic interpretation (far more exact and accurate than any researcher can produce at the moment) that can give an answer to this question. We wish to suggest by way of conclusion that it is only by providing a rigorous binary (overt vs. covert) structuring of K matrices that many of the current impressionistic conclusions about discourse can become scientific data.

3.1    Role of C in text structure. In order to ensure an empirical basis for the discussion of standard and non-standard conjoiners,  let us start from a text which belongs to a discourse type by definition cancelling the pragmatic boundaries of written vs. spoken transmission. Any text can be broken down into (a) an ordered set of hypermorphemes --

(16)                        p    q    r    ...    x    y    z    ,

defined as minimal clauses functioning as independent communicative entities in discourse, particularly given their propositional content, and (b) a set of conjoiners C of various types and categories. One important step in providing an explicit description of a text to contribute to its assignation to a discourse type is reducing it to a structure which can in logical terms be described as --

(17)                        ((p    ---    q) & (q    ---    r))    ---    (p    ---    r)).

        Assuming that all conjoiners are C in the linear manifestation, any text is of the shape --

(18)                        p    C    q    C    r    C    ...    x    C    y    C    z    ,

where p is the initial hypermorpheme of any discourse, and z is the final hypermorpheme of the same discourse, no matter whether the given discourse is in pragmatic or syntactic terms defined as a partial sequence pQ or as a total sequence tQ; such a notation makes unnecessary and redundant the use of any end-of-discourse marker. It may happen, however, that no lexicalized conjoiner C occurs in the linear manifestation of discourse and the text may take the hypothetical shape --

(19)                        p    q    r    ...    x    y    z    ,

but the discourse model presupposes the existence of relations of connectedness between the hypermorphemes sequentialized in the linear manifestation of discourse.

3.2    Three fundamental types of hypermorpheme sequentialization. Such connectedness relations can be explicitly described within the matrices and with the  formal devices of K, which concurrently functions as an overall symbolic marker of propositional connectedness. One obtains in this way three possible sequences of hypermorphemes in discourse:

(20)                        (alpha)    (    p    C    q    C    r    ...    C    x    C    y    C    z    )

(21)                        (beta)      (    p    K    q    K    r    ...    K    x    K    y    k    z    )

        The (beta) formula shows that the relations between hypermorphemes are abstract relations of connectedness, not realized in the linear manifestation of the discourse by means of any lexicalized conjoiner. Finally, a formula of the type --

(22)                        (gamma)    {    x    ;    p    ;    y    ;    q    ;    z    ;    r    }

would simply denote a set of hypermorphemes with no relations of connectedness whatever established between them. Certain researchers may label a sequence of hypermorphemes patterned on the last formula a 'non-text', or even a 'pseudo-text' ! But as any 'pseudo-text' may become a text provided certain pragmatic constraints are in force (fn 5 !), we prefer to call it 'an unordered set of hypermorphemes'. Such an unordered set of hypermorphemes may accidentally meet appropriateness constraints: putting together two sentences which are seemingly unrelated, a new semantic relation may emerge from discourse head association.

(23a)                        I met Enkvist on the corridor a few minutes  ago. There's going to be majority rule in Rhodesia in two years' time.

The 'accidental' connectedness relationship  between the two discourse heads --

(23b)                        Enkvist        vs.        Rhodesia

conveys the presupposed meaning in the underlying structure that --

(23c)                        It was Enkvist who communicated to me that this was so.

        The unordered nature of (22) is marked by the bracketing {   }  rather than in the conjoining system alone, as there is more to connectedness than just conjoining. This means that neither C nor K are insertable between the hypermorphemes, though relations of connectedness may accidentally occur.

        The text to be outlined below is, like any real text, a complex combination of an (alpha) with a (beta) formula of discourse structure: it is submitted to the fundamental formal constraint of discourse in accordance with which the emergence of a C in the linear manifestation of discourse is optional, whereas the existence of a K is compulsory between all hypermorphemes, be they adjacent or remote.  

3.3    A partial sequence. As we are less interested at the moment in the semantic substance of the hypermorphemes, let us simply replace them by bracketed lower-case Latin letters (for reasons of convenience, we start the listing with a; and the occurrence of z is no longer an end-of-discourse marker, as was the case in the hypothetical formalization). We thus obtain the following text structure, itemizing only the lexemes outside hypermorphemes. (A) and (B) mark participant boundaries: 


(A) ((but) (my dear fellow) (excuse me for interrupting you) (you seem to be X-ing (a)))  #  (for) (after all) (even you must admit (that (b) (than (c))))  #

(B)    (b') (than (c'))  #    (d)    #    (e)    #    (f) than (g))    #    (h)    #    (i)    #    (j) (k(l(that m))))    #    (it is (n) (or (o))    #    (p(q)    #    (indeed (r) (and when (s)))) (because (t) (which (u(v(who w))))))    #    (no) (Avoc))    (x)    #    (y(and z(of whose (aa)))))    #    (bb(because    (cc(and(dd(ee))))))    #     (ff)    #    (gg(who(hh (how(ii)))))

(A)    (Bvoc)(jj(as if(kk)))    #    (ll(and(mm)))    #    (nn(but(oo)))    #

(The actual semantic interpretation of this whole text is left to the reader's literary imagination. Have a try!)

3.4    Discussion. This way of representing a text, any text, singles out three major sets of conjoiners. First, there is the comprehensive group of standard conjoiners -- the 'conjunctions' of conventional linguistics. They materialize two distinct types of grammatical relations -- co-ordination vs. subordination. It is only the 'logical connecters' ( and, but, or, for; cf Lord Quirk et al. 1972 : 661f) that approximate the function of propositional connectives in symbolic logic. The subordinators (who, when, how, that, than), the next set of conjoiners, pose major problems in the methodological process of disengaging hypermorphemes. Disengagement procedures have yet to be studied in discourse. Clearly, it is the subordinators that distinguish between discourse types: written discourse evinces highly elaborate subordinator patterns, this being one of the formal features on the basis of which it can be defined. Finally, it is the subordinators that are to an overwhelming extent compulsory in the linearization of a given discourse; their co-ordinative counterparts are to a large extent deletable from the linearization without affecting sentential well-formedness, and quite often, without substantially modifying the overall semantic interpretation of discourse. Subordinators should be given great attention as part of a discourse theory of paraphrase. 

        Non-standard conjoiners -- the next most important set -- are not generally assigned conjoiner status in conventional linguistics. Traditional, structural, and even transformational grammars have treated them as adverbs or particles deprived of any considerable syntactic significance. They were never viewed as conjunctions for they are 'parenthetical' to various degrees, and it is the semantic interpretation alone that ascribes them a connecter function. More recently (Halliday & Hasan 1976 : 267ff) has led to the study of continuatives (after all, of course etc).  The set of attitudinal disjuncts (indeed, possibly, apparently, actually; cf. Lord Quirk et al 1972 : 511ff) is a controversial subcategorization balancing repudiated against sustained information in discourse structure. Response markers yes and no presuppose two distinct linguistic phenomena: first, the existence of a foregoing question in relation to which they function both as answer and as 'reduplicative dummy'; secondly, the existence of a participant boundary between the question and the answer (FOOTNOTE 7). Parentheticals are a subset of non-standard conjoiners with a difference: they are syntactically even more disengaged than the continuatives or the response markers for the very fact that they are often linearized at the level of the sentence, clause or phrase. This essentially marginal syntactic character has made them be closely associated with performance phenomena and dismissed as such. But parentheticals fulfil a definite connecter function: all discourse types (cf. partial sequence pQ above) evince the use of one type of parentheticals or another. They clearly affect discourse structure in a most immediate way, and considerably modify semantic interpretation over longer stretches of linearized language. It is only a discourse model that can begin to accommodate parentheticals; their occurrence in discourse is culture-specific.

        To conclude this section, here is a conjoiner chart for the standard and non-standard conjoiners so far discussed: 

1.0                        Standard Conjoiners:

1.1                        Co-ordinative (logical connecters):    and, but, or;    for.

1.2                        Subordinative: that, than, how, who, of whose, when, which, as if.

2.0                        Non-standard Conjoiners:

2.1                        Continuatives:    after all,     of course.

2.2                        Attitudinal Disjuncts:     indeed,    possibly.

2.3                        Response Markers:    yes,    no.

3.0                        Parentheticals:

3.1                        (sentence) excuse me for interrupting you

3.2                        (clause) you must admit

3.3                        (phrase) my dear fellow,     in my opinion

3.4                        (word) NAME (in vocative function)

3.5                        (morpheme) well.

        These were the subsets of conjoiners to be almost exclusively derived from the very short text in 3.3.        More extensive text is sure to reveal more complex conjoining systems. Some are language-specific, thus correlating with differentials, others have communicative value correlating with universals. Whichever the case is, all such subsets require unified treatment in discourse.

3.5    n-ary conjoiners. A case should perhaps be made that at least some conjoiners function on an n-place basis, i. e. more-than-two-place-structure: 

(25)                        C [-s]    {    p    q    r    s    ...    }

Certain parentheticals may be particularly well suited for this function. Such an n-place non-standard discourse conjoiner is analogous in function with the logical connective --

(26)                        It is not the case that    {    p    q    r    s    }

However, given their clearly nominal character, a discussion of n-place conjoiners in discourse goes far beyond the bounds of the present paper; we have therefore merely limited ourselves to stating the issue.


4.1        A text, any text, is made up of two distinct configurations superimposed one upon the other: a configuration of sequentially ordered hypermorphemes pqrs...xyz, coupled with a configuration of conjoiners Cn's, insertable between the hypermorphemes on the basis of optional constraints. This twin configuration gives the essence of textuality.

4.2        The set of C's plays a cardinal role in structuring discourse, quite analogical -- though by no means identical -- with that played by propositional connectives in modern logic. The fundamental distinction lies in the fact that lexicalizations of C form two distinct subsets, subjected to different internal constraints.

4.3        In addition to closed subsets of standard conjoiners C[+s], also belonging to the set C is the open subset of non-standard conjoiners C[-s], including parentheticals.

4.4        Non-standard conjoiners, as defined in the body of the paper, can only emerge from a consistent correlation of the semantic interpretation of the discourse with the corresponding linear manifestation, as mirrored in K structures at various levels of operationalization in the hierarchy. 

4.5        Hypothesizing the existence of a zero conjoiner in the discourse model to be adopted requires further investigation. Such a zero element, to be duly interpreted formally only within the Kc matrix could emerge in the linear manifestation as either   C[+s]        or as    C[-s]  ,depending on the alternative realizations to be adopted in the process of monitoring discourse in production. This issue is closely related to a pragmatic theory of paraphrase at hypermorpheme level.

4.6        How to operationalize  C[-s]  in a formal model of discourse without reducing it to the status of a mere logical connective, ranging over more than two hypermorphemes, could be considered n-place connecters, analysable in terms of Montague's concept of satisfaction. 


(1)    In the talk given at the Helsinki Meeting of Linguists, C[+s] and C[-s], were called 'conventional' and 'non-conventional' respectively, but labels have since been changed largely on account of a remark by Einar Haugen during the discussions, pertaining to conventionality in language. We wish to express gratitude for insistence on terminological accuracy.

(2)    For a plethora of similar examples, see any elementary textbook, e. g. Suppes 1964, First Course in Mathematical Logic; Suppes 1957, Introduction to Logic; Stebbing, passim, etc.    Furthermore, any descriptive analysis of any partial sequence p q is impeded by the very fact that it is 'partial'. This is not at all a question of disambiguation by context, but rather an issue of overall discourse structure, which is an altogether different matter.

(3)    Montague's (1974 : 99) zero-place operator symbol The American President is, in the last analysis, a one-place symbol. This culture-specific place assignment to the expression is perhaps most obvious in British English or in 'Swedish English' (The American President Gerald Ford, den amerikanske presidenten Jimmy Carter) than it is in a purely 'American' 'context of use'. One should perhaps mention that Carter had difficulty during the 1976 TV-debates in finding the right term of address in spoken discourse: he rejected both Mr President and Mr Ford and was only left with the rather questionable vocative President Ford. ( A similar problem surfaced years later in the TV-debates in French between President Mitterand and Mayor of Paris Jacques Chirac: it surfaced quite dramatically in the actual quite spontaneous exchanges in spoken discourse...) In other words, within the given discourse structure president clearly became a one-place symbol. 

(4)    Negation does have  a disturbing effect  on connectedness in data of the type -- I have no brown table but it is square, or the exchange (A) What time is it ? (B) Not yet which still, in very specific situations, may make very appropriate discourse. These and other data may point to the fact that negation could function as a very strong 'disconnecter'. It may increase connectedness in dialogue but that in itself does not justify an analogy with logic.

(5)    Nobody in quest of a novel to read would take up a telephone directory, though the latter is subjected to a clear set of textual constraints too, which make it either appropriate or inappropriate according to the year of publication.

        A telephone directory is a discourse ranging over a set of individuals existing in a given possible world, analogous to a novel. All propositions, however, are invariably of the same shape: an implied existential operator, accompanied by the deictic markers of coded location and coded tele-channel. This is adducible  to rigorous, but trivial,  formalization. Such a description does not of course cover 'secret' telephone numbers, which require additional constraints related to the interpretation of silence.  

(6)    This particular partial sequence has been selected from Oscar Wilde, 'The Critic as Artist', in:  Complete Works of Oscar Wilde, edited, with an Introduction by Vyvyan Holland, Collins, London and Glasgow, 1948/1969, pages 1022-1023: 

Ernest: But, my dear fellow--excuse me for interrupting you--you seem to me to be allowing your passion for criticism to lead you a great deal  too far. For, after all, even you must admit that it is more difficult to do a thing than to talk about it.

Gilbert: More difficult to do a thing than to talk about it ? Not at all. That is a gross popular error. It is very much more difficult to talk about a thing than to do it. In the sphere of actual life that is of course obvious. Anybody can make history. Only a great man can write it. There is no mode of action, no form of emotion, that we do not share with the lower animals. It is only by language that we rise above them, or above each other--by language, which is the parent, and not the child, of thought. Action, indeed, is always easy and when presented to us in its most aggravated, because most continuous form, which I take to be that of real industry, becomes simply the refuge of people who have nothing whatsoever to do. No, Ernest, don't talk about action. It is a blind thing dependent on external influences, and moved by an impulse of whose nature it is unconscious. It is a thing incomplete in its essence, because limited by accident, and ignorant of its direction, being always at variance with its aim. Its basis is the lack of imagination. It is the last resource of those who know not how to dream.

Ernest: Gilbert, you treat the world as if it were a crystal ball. You hold it in your hand, and reverse it to please a wilful fancy. You do nothing but re-write history. 

(7)    It is true that in the case of rhetorical questions there is no participant boundary; this very phenomenon is marked by the fact that certain languages possess specific or quasi-specific lexicalized response markers, which are, among others, used in conjunction with such questions: jo in Swedish, si in French, ba da in Rumanian, depending, of course, on the structure of the question itself. 

REFERENCES        to Only Connect ...

Church, A.        1956.  Introduction to Mathematical Logic, Princeton U. P. 378 pp.

Fretheim, T.        1974.    'Exclusive and Non-exclusive Coordination', in: Working Papers in Linguistics  from the University of Oslo (Linguistiske inlegg),  6,  63 pp.

Gazdar, G. & G. K. Pullum        1976.   'Truth-functional Connectives in Natural Language', in: Papers from the Twelfth Regional Meeting, Chicago Linguistic Society, mimeo 15 pp. 

Grice, H. P.        1975.    'Logic and Conversation', in: Syntax and Semantics, vol. 3: Speech Acts, ed. by Peter Cole & Jerry L. Morgan, New York, Academic Press.

Grimes, J. E.        1975.    The Thread of Discourse, Mouton, Janua linguarum, Minor 207, 408 pp.

Hale, A.        1973.    Clause, Sentence and Discourse structure in Selected Languages of Nepal, Oklahoma Summer Institute 40, 462 + 249 + 434 + 314 pp. 

Halliday, M. A. K.  & Hasan        1976.   Cohesion in English, Longman, 374 pp. 

Longacre, R. E.        1972.     Hierarchy and Universality of Discourse Constituents in New Guinea Languages, 2 vols: (a) Texts, (b) Discussion. Georgetown University, 268 + 176 pp.

Pike, K. L.        1964.    'Discourse Analysis and Tagmeme Matrices', in: Oceanic Linguistics, III, 1, 5-25.

Pike, K. L.        1975.    'Multidimensional Feature Arrays from Affix to Discourse', mimeo, 6 October 1975, 10 pp.

Pike, K. L. & E. G. Pike        1972.     'Seven Substitution Exercises for Studying the Structure of Discourse', in: Linguistics 94, 43-52.

Proceedings of the Nobel Symposium 39, devoted to the Theory and Practice of Translation, Stockholm, 6-10 September 1976.

Reichenbach, H.        1947.    Elements of Symbolic Logic, Macmillan, sNew York. 444 pp.

Quirk, R. et al.        1972.    A Grammar of Contemporary English, Longman.

Sandulescu, C. G.            1975a.    'Displacement Constraints on Discourse', in: Papers from  the Second Scandinavian Conference of Linguistics, Lysebu / Oslo, 19 – 20 April 1975, edited by Even Hovdhaugen, Department of Linguistics, University of Oslo, Norway,  pp. 282 – 313.

Sandulescu, C. G.                    Åbo/Turku           1975b.    'Presupposition, Assertion, and Discourse Structure', in: Proceedings on the Symposium on Word Order, ed. by N.-E.Enkvist, Åbo Akademi.

Sandulescu, C. G.           1976a.    'Learner Readability: The Status of Pragmatic Factors', paper presented at the 1976 Tenth Annual TESOL (Teaching of English as a Second or Other Language) Convention, New York City, 1 –7 March, Americana Hotel, (Abstract on p. 171).

Sandulescu,  C.G.             1976b.        'Mapping Discourse Structure' in: The Nordic Languages and Modern Linguistics, Proceedings of the Third International Conference of Nordic and General Linguistics, 5 – 9 April 1976, edited by John Weinstock, published by the University of Texas at Austin,  pp. 497 – 503.

Sørensen, V.        1976.    'Textuality -- a Pragmatic Approach' in: Proceedings of the Third Scandinavian Meeting of Linguists, Oslo.

Saloni, Z. & A. Trybulec        1974.     'Coherence of a Text and its Topology', in: Semiotica, 11 : 2, 101-108.

Thomason, R. H. ed.        1974.    Formal Philosophy, Selected Papers of Richard Montague, Yale U. P., 369 pp.

Trail, R. L.        1973.    Patterns in Clause, Sentence, and Discourse in Selected Languages of India and Nepal, Oklahoma Summer Institute 41, 4 vols.: 343 + 457 + 362 + 209 pp. 

Vaina-Pusca, L.        1976.    'La théorie des mondes possibles dans l'étude des textes', in:    Revue roumaine de linguistique, 21, 1, 35-49.

Wilson, D.        1975.    Presuppositions and Non-Truth-Conditional Semantics, Academic Press, New York, 161 pp.

Wise, M. R. & I. Lowe        1972.    'Permutation Groups in Discourse', in:    Language and Linguistics: Working Papers, Georgetown University, No. 4, 1972, 12-34. 

Wittgenstein, L.        1938.   'Lectures on Aesthetics', I, 4, in: Lectures & Conversations on Aesthetics, Psychology and Religious Belief, Compiled from Notes by Yorick Smythies, Rush Rhees and James Taylor, edited by Cyril Barrrett, Basil Blackwell, Oxford, 1966 / 1967, page 1. 


Ernest: But, my dear fellow--excuse me for interrupting you--you seem to me to be allowing your passion for criticism to lead you a great deal  too far. For, after all, even you must admit that it is more difficult to do a thing than to talk about it.

Gilbert: More difficult to do a thing than to talk about it ? Not at all. That is a gross popular error. It is very much more difficult to talk about a thing than to do it. In the sphere of actual life that is of course obvious. Anybody can make history. Only a great man can write it. There is no mode of action, no form of emotion, that we do not share with the lower animals. It is only by language that we rise above them, or above each other--by language, which is the parent, and not the child, of thought. Action, indeed, is always easy and when presented to us in its most aggravated, because most continuous form, which I take to be that of real industry, becomes simply the refuge of people who have nothing whatsoever to do. No, Ernest, don't talk about action. It is a blind thing dependent on external influences, and moved by an impulse of whose nature it is unconscious. It is a thing incomplete in its essence, because limited by accident, and ignorant of its direction, being always at variance with its aim. Its basis is the lack of imagination. It is the last resource of those who know not how to dream.

Ernest: Gilbert, you treat the world as if it were a crystal ball. You hold it in your hand, and reverse it to please a wilful fancy. You do nothing but re-write history. 

Oscar Wilde Quotation ends


(A) ((but) (my dear fellow) (excuse me for interrupting you) (you seem to be X-ing (a)))  #  (for) (after all) (even you must admit (that (b) (than (c))))  #

(B)    (b') (than (c'))  #    (d)    #    (e)    #    (f) than (g))    #    (h)    #    (i)    #    (j) (k(l(that m))))    #    (it is (n) (or (o))    #    (p(q)    #    (indeed (r) (and when (s)))) (because (t) (which (u(v(who w))))))    #    (no) (Avoc))    (x)    #    (y(and z(of whose (aa)))))    #    (bb(because    (cc(and(dd(ee))))))    #     (ff)    #    (gg(who(hh (how(ii)))))

(A)    (Bvoc)(jj(as if(kk)))    #    (ll(and(mm)))    #    (nn(but(oo)))    #