When did linguists begin to look at the presuppositions that lie behind the production of sentences? The first comments I heard about it were at the Linguistic Institute at Illinois in the summer of 1968, but by the second Institute at Illinois in 1969 the classrooms and lecture halls were full of presuppositions, and, as the reader will have noted, more than half of the contributions to the Linguistics section of this volume deal in one way or another with the same topic. I do not know just who introduced the term ‘presupposition’ to the linguistic literature, or exactly when, but Fillmore, Lakoff, McCawley or Ross would certainly be good candidates for the honor. Where did this new psycholinguistic notion come from? Fillmore (1969) traces his indebtedness to certain of the philosophers of ordinary language, particularly Austin (1962) and Alston (1964) with their concerns about ‘illocutionary acts’ and the ‘happiness or felicity conditions’ for the use of certain sentences, but also to Strawson’s paper ‘On Referring’ (1950) and most remotely to Frege (1892).

In recent linguistic papers the role of presuppositions in a wide variety of sentence phenomena has been described; as a small sample, in the choice among verbs of judging (accuse, criticize, blame) by Fillmore (1969), in the choice between some and any by Lakoff (1969), in the use of pronouns and reflexives (McCawley, 1968; Sampson, 1969). Also, the kinds of problems in grammatical description that led Chomsky (1965) to introduce the notion of indexing (e.g., choice among the teacher₁ hates himself₁ and the teacher₁ hates the teacher₂) come down to special instances of presupposing (cf. Sampson, 1969). Finally, we have those situations, like the demonstration with which I began this paper, where presuppositions are established by non-linguistic contexts. Olson (1969), moving along very similar lines, has described how varying comparison objects (black circle vs. white square) can influence the way a speaker will refer to a given object (white circle), saying. . . under the white one for the first contrast and. . .under the round one for the second, for example.

This preoccupation with presupposition has led to debate, within as well as between linguists, as to the nature of such phenomena. Are presuppositions linguistic in nature (prior, even if implicit, sentences) or are they psychological in nature (cognitive states or sets)? Morgan (1969) says that ‘many kinds of presuppositions “act like” previously uttered sentences. . .but to describe both with the same set of rules, it is necessary to represent presuppositions as trees to the left of the sentences they are associated with’ (pp. 66-7). Lakoff (1969), in explaining why there can’t be any some-any rule,¹ states that ‘ something that is present only at the most abstract levels of grammar must play a part in determining the form which surface structures must take. .. when some is used, the presupposition is necessarily positive; when any appears, it may be negative or neutral ’. But he also says that ‘ the beliefs and expectations of the speaker are reflected in his choice of some or any, and the meaning of the sentence is correspondingly changed ’. And Lakoff concludes that ‘whether [representation of presuppositions] would be done by assuming separate sentences (e.g., I hope this S is true, or 1 believe that the subject of this sentence is true), or in another way, is not clear. . .’ (pp. 612-13).

Other linguists seem to have accepted the non-linguistic, cognitive nature of presuppositions. McCawley (this volume), for example, in discussing indexing, states that ‘ indices exist in the mind of the speaker rather than in the real world. . . and. . . the noun phrases which speakers use fulfill a function comparable to that of postulates and definitions in mathematics: they state properties which the speaker assumes to be possessed by the conceptual entities involved in what he is saying’. Similarly Sampson (1969): ‘I suggest that the mind of a person entering into a speech situation should be regarded as containing a number of entities which I shall call “referents” [cf. McCawley’s term, “intended referent”] (p. 4). . .so we must consider the hearer’s stock of referents to be arranged in some kind of mental space (p. 7) [and finally]... I believe this is a genuine problem which would sooner or later have to be faced by any theory of language ’ (p. 9). Schlesinger (forthcoming) builds a general theory of language production in which a non-linguistic ‘intentional’ system is assumed to antedate strictly linguistic operations.

Psychologists, of course, have generally assumed that non-linguistic cognitions were the sources and destinations of sentences. James Deese (1969), for example, speaks of ‘understanding’ as occurring when a listener can make some external information, linguistic or perceptual, conform to some conceptual category, this category providing the abstract structure of an interpretation. And on the encoding side, according to Deese ‘ the primordial thought that gives rise to a linguistic interpretation is not itself linguistic’ (p. 519). He believes that these cognitive categories are derived from underlying perceptual relations, including such gestalt-like properties as grouping, contrast and similarity as well as classification in terms of superordination and subordination. Donald Campbell (1966), after making the astute observation that ‘you can’t teach a language by telephone’, asserts that ‘it is the child’s already available entifications of the physical environment which provide his trial meanings ’, and by ‘ entification ’ he is also referring to the gestalt-like perceptual tendencies toward grouping in terms of qualitative similarities and contrasts and the common fate (trans-positions as wholes) which characterize the things which ordinarily get named (cup and saucer, but probably not cupandsaucer and certainly not cupandlittlefingerandsoutheastcornerofroom). As Campbell points out, languages tend to follow Aristotle’s advice and ‘cut nature at her joints’.

To indicate what I consider to be the absurdity of treating presuppositions as prior, implicit sentences, let me borrow an example from Fillmore (1969) and expand on it. For the imperative sentence (9)

(9) Please shut the door

the ‘happiness conditions’ (presuppositions) include at least the following: (a) speaker and addressee have a role relation which permits such requests; (b) the addressee is capable of shutting the door; (c) the speaker has in mind some particular door and assumes the addressee can identify this door (saying go shut a door would be like saying go fly a kite!); (d) the door in question is presently open; and (e) the speaker wants this door to become closed. Now let us phrase appropriate prior implicit sentences for this speaker.

(10) I am your father (a)

(11) You are able to shut doors (b)

(12) There is a door here as we both know (c)

(13) That door is open (d)

(14) I desire the door to become closed, (e)

I submit that it is ludicrous to postulate that a speaker and a listener (to interpret fully) must riffle through (generate) such a series of sentences in order to create or understand (9) - and this is a comparatively simple sentence. Not only is it intuitively much more satisfying to assume that situational (perceptual) signs are shared by speaker and hearer and create the presuppositions, but processing of the above sort would be incredibly inefficient. Fillmore points out that we can infer such presuppositions from certain structural and semantic characteristics of the sentence, but then adds: ‘ more importantly, however, it needs to be pointed out that some of these conditions are really preconditions [italics mine] for the use of the sentence ’ (p. 96). In other words, such non-linguistic cognitions are predictive of the form and content of the sentence.

In an attempt to bring such presuppositions and indexings into the realm of familiar linguistic analysis, it seems to me that many linguists are now carrying their methods to ridiculous lengths. For example, Sadock (1969) proposes that there are super-hypersentences that are required to handle sentences like (15)

(15) I promise that this is the end

with the highest node being a super-hypersentence of the form Speaker- V-Addressee - S (where V is 4- performative complement), with this S dominating an ordinary hyper-sentence of the form Speaker- V - Addressee - S (where V is + promise) and with this S dominating in turn the sentence Speaker promise Address that this is the end. When will we have Extra-superduper-super-hypersentences? In a similar, if not so extreme, vein, we find McCawley (as reported in Morgan, 1969) analysing the semantically rather simple verb kill as a derived constituent CAUSE (BECOME (NOT (ALIVE))) as part of his attempt to give semantic representation in the form of tree-like structures. By a rule of ‘ predicate-raising ’ a series of embedded sentences becomes x causes become not alive y which is considered to be, if not synonymous with, at least an adequate paraphrase of x kill y. The underlying motivation here, as I see it, is to apply familiar linguistic constructs and rules to the semantic component, but if it can be shown that purely perceptual signs and their interrelations utilize the same semantic system (as when John perceives PAUL KILLS TOM) then I think this attempt is doomed to failure.

Nevertheless, presuppositions do influence the structure of sentences and, as Fillmore puts it (1969, p, 98), ‘any complete account of the grammatical description of a language will need to bring in presuppositional facts at many points’. Yet, if presuppositions cannot be handled with strictly linguistic constructs and rules, several fundamental issues arise: (1) Is the study of presuppositions a linguistic or a psycho linguistic business? Or, put another way, can presuppositions be considered as part of the grammar of a language, as Fillmore claims? (2) Can syntax be formally separated from semantics? As Maclay presents the situation in his overview of the linguistic contributions to this volume, ‘the thrust of many of the arguments that the well-formedness of sentences cannot be determined solely on formal or syntactic grounds... [and] once the autonomy of syntax has been denied, the existence of a distinct level of syntactic deep structure also becomes unacceptable’. (3) Can any clear distinction between competence and performance be maintained under these conditions? According to Maclay, ‘Katz and Fodor expanded the empirical domain of linguistics to include semantics with the ultimate result that the autonomy of syntax is now seriously questioned by linguists. The autonomy of competence may well be the next victim’. Such autonomy will clearly be questioned by the data on paraphrasing ‘things’ which I will present in this paper. It would appear that Chomsky was building better than he knew when, in more or less of an aside on the first page of his Language and Mind (1968), he transformed linguists into psychologists: ‘I think there is more of a healthy ferment in cognitive psychology - and in the particular branch of psychology known as linguistics - than there has been for many years.’ Welcome to the fold!

¹ Among Lakoff’s compelling examples are Who wants some beans? (speaker expects some - one does want beans) vs. Who wants any beans? (speaker expects no one wants beans) and If you eat any candy, I'll whip you (speaker is threatening you) vs. If you eat some spinach, I'll give you ten dollars (speaker is promising you).