The concept of orientation (argumentative orientation, oriented statement or expression), combined with the correlative concept of argumentative scale (Ducrot 1972), is fundamental to the theory of Argumentation within Language (sometimes referred to as AwL theory) developed by Oswald Ducrot and Jean-Claude Anscombre since the 1970s (Anscombre, Ducrot 1983, Ducrot 1988, Anscombre 1995a, 1995b, etc.). In this entry, the word discourse will refer exclusively to (polyphonic) monologue, not to dialogue or interaction.

The following equivalences can be helpful to grasp the general concept of meaning as argument, that is orientation towards a following statement, having the status of conclusion:

He said E1. What does that mean?
He says E1 in the perspective of E2
The reason why E1 is said is E2
The meaning of E1 is E2
E1, that is to say E2

1. But and the grammar of orientation

The stimulating case of but, has played a pivotal role in the construction of a grammar for argumentation. The privileged construction chosen to analyze this conjunction is schematized by “E1 but E2”:

The restaurant is good, but expensive.

The basic observations are as follows: E1 and E2 are true (the restaurant is good and expensive); but refers to an opposition; this opposition is not between the predicates “to be good’ and “to be expensive”: one knows that “everything good is expensive”, and tends to think that all expensive restaurants are necessarily good. The opposition is between the conclusions drawn from E1 and E2, functioning as arguments: if the restaurant is good, then, let’s have dinner there; if it is expensive, let’s go to another place; and the final decision is the latter. But here articulates two statements oriented towards contradictory conclusions, and retains the conclusion derived from the second argument.

Under such analysis, the meaning of but is instructional: connectives provide guidance for the interpretation of the speeches they articulate. They give the receiver the instruction to infer, to reconstruct from the left context E1 a proposition C opposed to something, not-C, that can be inferred from the context to the right of E2 (following E2). It is up to the listener to rebuild an argumentative opposition.

In the context of dialogical argumentation, these “instructors” themselves fall into the scope of an argumentative question, conditioning the reconstruction of the conclusions derived from E1 and E2. The preceding but came under a question like “Why not try this restaurant?”. If the question was “Which restaurant should we buy to make the best investment?”, the interpretation would be totally different: “This restaurant is good (= “delivers outstanding financial performance”) but is expensive (to buy)” the inferred, implicit conclusion would be “so, let’s invest our money somewhere else”. The argumentative question structuring the text creates the field of relevance and provides the interpretive constraints. This question and the relevant conclusions – answers are said to be “implicit” only insofar as the data supporting the analysis of but are generally limited to a pair of statements, the analyst considering that intuition can supply the missing context.

2. Linguistic constraint on the {argument, conclusion} sequence

As classical approaches, this theory considers argumentation essentially as a combination of statements “argument + conclusion”. The crucial difference lies in the concept of the link authorizing the “step” from argument to conclusion, that is the “topic”. The coherence of discourse is attributed to a semantic principle called a topos which binds the predicate of the argument to the predicate of the conclusion.

Ducrot defines “the argumentative value of a word” as “the orientation that this word gives to discourse” (Ducrot 1988, p. 51). The linguistic meaning of the word clever must not be sought in its descriptive value of a capacity measured by the intellectual quotient of the person concerned, but in the orientation it gives to a statement, namely, the constraints it imposes on the subsequent discourse. For example:

Peter is clever, he will solve this problem

is opposed to the chain perceived as incoherent:

* Peter is clever, he will not be able to solve this problem.

Such an argumentation is convincing indeed, because its conclusion, “solving problems” belongs to the set of predicates semantically correlated with “to be clever”. A set of pre-established conclusions is already given in the semantic definition of the predicate of the statement used as argument.

The two argumentative morphemes, little / a little give opposed argumentative orientations to the statement they modify, S. Orienting words:

he has taken a little food, he is improving
he has taken little food, he is getting worse.


Building upon such intuitions, Ducrot defines the argumentative orientation (or argumentative value) of a statement as “the set of possibilities or impossibilities of discursive continuation determined by its use” (ibid.).

The argumentative orientation of a statement S1 is defined as the selection operated by this statement among the class of statements S2 likely to follow it in a grammatically well-formed discourse. Theoretically, a first statement S1 can be followed by any other statement S2, both being independent linguistic units. According to the Argumentation within Language theory, the use of the first statement S1 introduces restrictions imposing certain characteristics upon the second statement S2; that is, it excludes some continuations and favors others.

The linguistic constraints imposed by the argument upon the conclusion are particularly visible on quasi-analytical sequences, such as “this proposition is absurd, so it must be rejected”. By the very meaning of words, to say that a proposition is absurd, is to say that it must be rejected. This apparent conclusion is a pseudo-conclusion, for it merely expresses the definiens of the word absurd, “which should not exist” as testified by the dictionary:

A.- [Speaking of a manifestation of human activity: speech, judgment, belief, behavior, action] Which is manifestly and immediately felt as contrary to reason in common sense. Sometimes almost synonymous with the impossible in the sense of “which cannot or should not exist”. (TLFi, [Absurd]).

In a formula as famous as it is objectionable, Roland Barthes wrote that “language is neither reactionary nor progressive; language is quite simply fascist; for fascism does not prevent speech, it compels speech” ([1977], p. 366). Barthes’ perspective is certainly different from Ducrot’s. Nonetheless, in Ducrot’s perspective, the argument literally compels the conclusion — playing with words, one might say that the inference is compulsive. This is common argumentative experience, in ordinary language, hearing the argument is enough to guess the conclusion.

Ducrot’s theory is constructed on the linguistic observation that, regardless of its informational content, any statement specifies its possible continuations and excludes others. Not just any statement can follow any other statement, not only for informational reasons, but also for semantical and grammatical reasons. There are semantic constraints on discourse construction.

At the sentence level, this idea is expressed in the purely syntactic language of the restriction of selection. In its non-metaphorical use, the statement “Pluto barks” assumes that Pluto is a dog. Literally taken, barking carries a restriction of selection determining the class of entities it admits as subject. Similarly, but at the discourse level, S1 operates a selection upon the class of the statements S2 that can succeed it. An argumentation is a pair of statements (S1, S2), such that S2, called the conclusion, respects the orientation conditions imposed by S1, called the argument.

3. Meaning as intention

The AwL theory rejects the conceptions of meaning as adequacy to reality, whether logical (theories of truth conditions) or analogical (theories of prototypes). It is built on a quasi-spatial conception of meaning as sense, direction: what the statement S1 (as well as the speaker publicly) means, in a specific context, is the conclusion S2 to which this statement is oriented. The art of arguing here is the art of managing discourse transition.

The relation “argument S1 – conclusion S2” is reinterpreted in a language production perspective (Fr. perspective énonciative) where the meaning of the argument statement is contained in and revealed by the next statement. The understanding of what is meant by the statement “nice weather today!” is not developing a corresponding mental image or cognitive scheme, but accessing the intentions displayed by the speaker, that is, “let’s go to the beach”. This is in perfect agreement with the Chinese proverb, “when the wise man points to the stars, the fool looks at the finger”.

The meaning of S1 is S2. The meaning here is defined as the final cause of the speech act. The AwL thus updates a terminology referring to the conclusion of a syllogism as its intention. This reflects the fact that a reformulation connector such as that is to say can introduce a conclusion:

L1:   — This restaurant is expensive.
L2:   — That is/ you mean / in other words/ you do not want us to go there?

The theory has developed in three main directions, argumentative expressions, or orienting words; connectives as argumentative indicators; and the concept of semantic topos.

4. Some consequences

4.1 Reason in discourse

Tarski maintains that it is not possible to develop a coherent concept of truth within ordinary language, S. Probable. In Ducrot’s vision of argumentation, the question of the validity of an argument is re-interpreted as grammatical validity. An argumentation is valid if the conclusion grammatically agrees with its argument (if it respects the restrictions imposed by the argument). It follows that the rationality and reasonableness attached to the argumentative derivation are no more than an insubstantial reflection of a routine discursive concatenation of meanings, or, as Ducrot says, a mere “illusion”, S. Demonstration. This is coherent with the structuralist project reducing the order of discourse to that of language (Saussurian langue). Ordinary discourse is seen as unsuited to expressing truth and reality. It follows that discourse is denied any rational or reasonable capacity.

4.2 A re-definition of homonymy and synonymy

As the theory is based exclusively on the concept of orientation, and not on quantitative data or measures, it follows that if the same segment S is followed in a first occurrence of the segment Sa and in a second occurrence of the segment Sb that contradicts Sa, then S does not have the same meaning in these two occurrences. As we can say “it’s hot (S), let’s stay at home (Sa)” as well as “it’s hot (S), let’s take a walk (Sb)” we have to admit that the statements “[are] not about the same heat in both cases” (Ducrot 1988, p. 55). This is a new definition of homonymy. By analogous considerations, Anscombre concludes that there are two verbs to buy, corresponding to the senses of “the more expensive, the more I buy” and “the less expensive, the more I buy” (Anscombre 1995, p. 45).

Conversely, we can assume an equivalence between statements selecting the same conclusion: if the same segment S is preceded, in a first occurrence by the segment Sa, and in a second occurrence by a different segment Sb, then Sa and Sb have the same meaning, because they serve the same intention: “it’s hot (Sa), I’ll stay at home (S)” vs. “I have work (Sb), I’ll stay at home (S)”. It is a new definition of synonymy, in relation to the same conclusion.

Finally, “if segment S1 only makes sense from segment S2, then sequence S1 + S2 constitutes a single utterance”, a single linguistic unit (Ducrot 1988: 51). One could probably go a step further, and consider that they make up a single sign, S1 becoming a kind of signifier of S2. This conclusion reduces the proper “order of discourse” back to that of the statement, even of the sign.

5. Orientation and inferring license

Ducrot opposes his “semantic” point of view to what he calls the “traditional or naive” view of argumentation (Ducrot 1988, p. 72-76), without referring to specific authors. Let’s consider Toulmin’s layout of argument.

— Argumentation is basically a pair of statements (S1, S2), having respectively the status of argument and of conclusion.

— Each of these statements has an autonomous meaning, and refers to a distinct specific fact, each of these facts being independently assessable.

— There is a relation of implication, a physical or social extra-linguistic law between these two facts (Ducrot 1988, p. 75).

This concept of argumentation can be schematized as follows. Curved arrows, going from the discourse level to the reality level, enact the referring process.

This conception may be “naive” insofar as it postulates that language is a transparent and inert medium, a pure mirror of reality. This is not the case for natural language (Récanati 1979); such conditions are only met by controlled languages ​​like the languages ​​of the sciences, in relation to realities that they construct as much as they refer to them.

Contrary to this view, the AwL theory emphasizes the strength of purely linguistic constraints. The orientation of a statement is precisely its capacity to project its meaning not only on, but also as the following statement, so that what appears as “the conclusion” is only a re-formulation of the “argument”. For the AwL theory, discourse is an arguing machine, systematically committing the vicious circle fallacy.

To sum up, the AwL theory opposes ancient or neoclassical theories and practices of argumentation, as a semantic theory of language opposes theories and techniques of conscious discursive planning, operating according to referential data and principles. For classical theories, argumentative discourse is likely to be evaluated and declared valid or fallacious. For semantic theory, an argumentation can be evaluated only at the grammatical level, as a concatenation (E1, E2) that is acceptable or not, coherent or not. In this theory, the compelling character of an argument is entirely a matter of language. It is no different from the coherence of discourse. To reject an argument is to break the thread of the ideal discourse. This position redefines the notion of argumentation; Anscombre speaks thus of argument “in our sense” (1995b, p.16).

6. Natural reasoning combines the two kinds of inferences

The transition from argument to conclusion can be based on a natural or social law or on a semantic coupling of the argument with the conclusion. These two kinds of inferences are currently connected in ordinary discourse:

You talk about the birth of the gods (1). You say, then, that at one time the gods did not exist; so you deny the existence of the gods (2), which is a blasphemy and punished by the law. So you will be punished (3) a pari, according to the law punishing those who speak of the death of the gods.

First, a semantic law deduces (2) from (1), S. Inference; second, a social law, having nothing to do with language or discourse, goes on from (2) to (3), the punishment being finally determined by an a pari alignment. Social law can be naturalized by somehow integrating the meaning of the words:

You are an impious man, impiety is punished with death, so you must die.

It is difficult to tell to what extent the very meaning of the word impious has integrated the law “impiety is punished with death”. Nonetheless, the link with social reality is clear: if I wish to reform the legislation, my revolt is not a semantic revolt. S. Definition; Layout.