Stephen C. Levinson – Pragmatics (part III)

 

Title: Pragmatics
Author: Stephen C. Levinson
Publisher: Cambridge University Press
Year: 1983
ISBN: 0-521-29414-2
Pages: 423

read the first part here, second part here

 

 

 

 

It might be held that a careful definition of the term speech act would virtually equate pragmatics with speech act theory. Some complication might arise at sub-speech act level – and indeed do, consider the tense of a sentence – but the point to be made is this: if one manages to understand the basics of speech act theory, then he is halfway through understanding at least the scope of pragmatics as a domain.

History

There are some interesting similarities between the birth of speech-act theory and the birth of argumentation theory. First, both were “unearthed” in response to the same philosophical excesses preached by logical positivism, namely the ‘unverifiability = meaningless’ doctrine[1]. This happened in one unusually prosper period at end of the fifties: Perelman’s & Olbrechts-Tyteca’s Treatise (1958), Toulmin’s The Uses of Argument (1958), Austin’s How To Do Things With Words (delivered in 1955), and Wittgenstein’s Philosophical Investigations (published in 1958). But a third interesting similarity is that there was little, if any, intellectual influence between these authors. Just as Toulmin had no idea of Perelman’s & Olbrechts-Tyteca’s work at that time – although in spirit the two works are very much alike, Austin had no idea of Wittgenstein’s work at that time – although, again, the two works are driven by the same philosophical conceptions. As Levinson notes, with a touch of wit, “Austin set about demolishing, in his mild and urbane way, the view of language that would place truth conditions as central to language understanding” (228), which, we recall, is exactly what Wittgenstein was doing against… his previous self.

Basics

I believe we pretty much covered the basics of speech act theory here on this blog. Obviously, reading our coverage does not amount to reading the sources themselves, but should rather be seen as a way to ease along with their main views. So here’s a list:

Austin’s How To Do Things With Words (1962)
Searle’s What is a speech act? (1965)
Searle’s Speech acts: An Essay in the Philosophy of Language (1969)
Searle’s A Taxonomy of Illocutionary Acts (1975)
Searle’s Indirect speech acts (1979)

These posts wrap up the theoretical input from Levinson’s pp. 226-243 (give or take). I will therefore not go into details any further. However, Levinson re-organizes[2] things so as to make clear some complications which speech act theory gives rise to; he, as it were, “abstracts” the common principles from the Austin & Searle approach to form what he names “irreducibility thesis” (Thesis, for short) and then he compares it to the “reducibility thesis” (Antithesis, for short). Solely by reading the title of the textbook, one can predict that the Antithesis, the attempt to reduce acts to semantics and syntax, must have failed. Let us then see why.

The Thesis could be pinned down by listing its main postulates: (1) uttering is not only expressing propositions but also performing actions, (2) among the actions performed the one called illocutionary act is being given as a privileged one for pragmatic inquiry[3], (3) the action in question is associated with the form by convention[4], (4) each type has its identifiable “classic” form, (5) a set of felicity conditions jointly define and constitute the nature of any identifiable act. Now, if we join all these, like Searle did, it’s almost impossible not to end up trampling on the domain of semantics, for what the Thesis ultimately amounts to is saying: “illocutionary force is an aspect of meaning, broadly construed, that is quite irreducible to matters of truth and falsity” (246). Acts are felicitous vs. infelicitous and this cannot be reduced to true vs. false. In other words, illocutionary force is an aspect of meaning which, since it belongs to the realm of actions, has as its “semantic value” the value of appropriateness.

The Anthithesis, as you might expect, states just the opposite: “there is no need for a special theory of illocutionary force because the phenomena that taxed Austin are assimilable to standard theories of syntax and truth-conditional semantics. The “dissenters,” as Levinson calls them, rejected Austin’s idea that a promise or a bet is assessable solely in terms of felicity and proposed the exact opposite conception: by uttering them, the speaker makes them true or false. So, when you say I hereby warn you not to behave improperly, what you have said is true. To be more precise, a certain combination of clauses (syntax) and a certain semantic value (truth-conditions) might be and have been designed to treat utterances as events. Events, we know, are either true or false. Levinson discusses in some detail the difficulties of such an approach. It must be frankly said that, although such theories may sound absurd, especially to those accustomed to the type of reasoning usually going on within pragmatics, such semantic theories are endowed with certain elegance. Nonetheless, they might miss the entire point of the notion of speech act: “the action-like properties of utterances” (259) Also, there is no way to relate (1) with (2) below:

(1) Why don’t you become an astronaut? (2) I ask you why don’t you become an astronaut, and if you can think of no good reasons why not, I suggest that you do

 

Indirectness

Levinson characterizes indirectness (see Searle’s Indirect Speech Acts above) as “a major problem for both Thesis and Antithesis” (p. 263) That is because both Thesis and Antithesis theorists are inevitably committed to the Literal Force Hypothesis (LFH) which could be summarized rather bluntly as: illocutionary force is built into sentence form. According to this hypothesis, the explicit performative, which could be boiled down to a form of the type I (hereby) V_p you (that) S, is either conventionally or truth-conditionally tied to a certain illocutionary force by virtue of the way the speaker “assigns” values to the “variables” V_p and S.

The next step is to view indirectness as a special case where the LFH is still standing and thus the LF must be, somehow, by some means, worked out. I’ll quote in full Levinson’s exposition of this conception:

Given the LFH, any sentence that fails to have the force associated with it […] is a problematic exception, and the standard line is to claim that, contrary to first intuitions, the sentence does in fact have the rule-associated force as its literal force, but simply has in addition an inferred indirect force. Thus, any usages other than those in accordance with [LFH] are indirect speech acts. (264)

But then (1) most usages are indirect, (2) what people do with sentences seems quite unrestricted by the surface form (i.e. sentence-type) of the sentences uttered and (3) some indirect speech acts have syntactic features of their own. How can the LFH still claim that force and form are strictly (and predictably!) associated? The two typical getaways are called idiom theory and inference theory. The former is the rather unsophisticated response of claiming that the primary (intended) act is achieved by an idiomatic expression (i.e. expressions which are not semantically equivalent to their parts). Thus, “Can you VP?” is an idiom for “I request you to VP”. Levinson notices that this explanation is not feasible (though it has been proposed by some theorists). First, both readings are simultaneously available and appropriate and responses to these utterances can be, within the same “speech event”, both available and appropriate. Moreover, consider this post. An idiomatic approach will eventually have to claim that all semantically non-equivalent instances of the e.g. indirect requests are idioms. Levinson scholarly loses his nerve and calls this an “embarrassment” (269)

Inference theory is what Searle illustrates in Indirect Speech Acts. In addition to the literal force in question, then, indirect speech acts manage to convey a second force which, by virtue of some Gricean-like inferences, is understood as primary. The entire procedure is explicated in the post on Searle’s famous article.

Both the idiom and the inference responses, however, fail to[5] (1) explain why does one choose to perform acts indirectly instead of simply performing the needed one, (2) give no real theoretical (more the less empirical) evidence that Literal Forces are indeed conventional. If we are prepared to stand opposed to (2) then there’s no LF and there’s no indirect speech act. There’s only the context and the conversational goal(s) within that context, and we are only left with “a general problem of mapping speech act force onto sentences in context” (274). It is only in this sense that illocutionary force is “entirely pragmatic” (idem.) Are we then left with no theory? No, we can go back to the three, centuries-old sentence types: interrogative, declarative & imperative. Starting from there, we can adopt a Sperber & Wilson type of context-change theory (Levinson, obviously, does not cite from Relevance Theory since at that time it hadn’t been born yet). The basic assumption behind these theories is quite intuitive and could be expressed in the form: speech acts are operations on contexts; when a speech act is uttered, more has taken place than the (mere?) expression of “a” meaning. A context has been changed. The contribution that utterance makes to this change in the context is in fact the act (the “force of the act”, the “potential of the act” etc.) Levinson ends this section: “We await the full-scale theories that would provide answers […]”

In the remainder, Levinson expands on the notion of “speech event” and how a serious theoretical inquiry into this subject would “dethrone” the basic speech act theory. The whole idea behind speech-event approaches is quite unpretentions – and the attentive reader will recognize the radical loop back to Wittgenstein’s “meaning is use” and the closeness between “speech event” and “language game”. Utterances do not have previously mapped functions. They acquire these functions within contexts, i.e. within conversation. The examples Levinson gives in this section are instructive, but I think one of the examples from the next section illustrates better this language-game dependency. If we see the following exchange in isolation:

A1: I have a fourteen year old son B1: Well, that’s all right A2: I also have a dog B2: Oh, I’m sorry

 

it will be perceived as at least hard to comprehend or even strange. What force do we assign to B1 and A2? Moreover, B2 is downright meaningless. However, if we embed the exchange in a context where A is trying to rent an apartment and raises a series of possible reasons for rejection (to B, which would be the landlord), everything starts to make sense. It is in this vein that “meaning is use”.

---

[1] Note firmly that this is not the same with the “meaning = truth-conditions” doctrine as it is (still) used in formal semantics. In a positive sense, the principle has no existential implications and works just fine as a ‘metaphysical’ hypothesis not about the world, but about/between two languages (an object and a meta- language).
[2] As far as I can tell, this “re-organization” belongs to him and is not a general view of speech act theory.
[3] This might need some elucidation. The point is that, when uttering something, we are performing a lot of acts, on different levels. For instance when saying “He is therefore not guilty”, a speaker (a) makes some noises out of his mouth, (b) produces some words with meaning, (c) produces a statement which is an assertion, and we could go on higher levels (d) defends his client, (e) making a living, (f) doing justice etc. Among these, within pragmatics, (b) is the one which renders itself neatly to the study of language from a, to put it rather bluntly, “speaker-context-speaker” point of view.
[4] Uttering an act and indicating its force (explicitly, as in I promise to come, or implicitly, as in I’ll come) is assimilated by Searle in the essential condition (“X counts as doing Y”) and this condition is conventional
[5] Levinson repeatedly uses this verb which, I believe, makes the whole enterprise something of a straw man. Of course they fail: they don’t even try. Searle – as well as Grice, implicitly – deal with institutionally unmarked, “normal”, co-operative situations. So, of course they fail. They don’t even have the theoretical apparatus to deal with such questions. It’s like saying to a mathematician that he will be stunned to find out that “x” in his theorem could be a monkey as well as a moped.

Technorati Tags: Levinson,Stephen C. Levinson,Pragmatics,Speech act theory,speech acts theory,indirect speech act,illocutionary act,Searle,Austin,Grice

Stephen C. Levinson – Pragmatics (part II)

 

 

Title: Pragmatics
Author: Stephen C. Levinson
Publisher: Cambridge University Press
Year: 1983
ISBN: 0-521-29414-2
Pages: 423

read the first part here

 

 

Presupposition as a topic, as we are rightly informed from the very beginning, is probably the “oldest” to be part of what is now called pragmatics. This is both favorable and risky: on the one hand we have a steady bulk of concepts designed to attack precisely the problems raised by presupposition, on the other hand “a great deal [of work on this subject] is also obsolete and sterile.” (p. 167) Before we begin, I should mention that some past entries on this blog (the series starting with this post) have indirectly skimmed over the notion of presupposition while talking about the related subject of reference.

As with much of the linguistic work carried on at the beginning of the last century, inquiry into the notion of presupposition starts with Gottlob Frege. His insight on presupposition is, from a logical point of view, particularly elegant. Frege notes that a proposition like (p) “John is my friend” presupposes[1] something like (q) “[The name] ‘John’ designates something” and that this latter proposition cannot be part of its meaning. Leaving aside problems of object- and meta-language (as I see Levinson does too, for didactical purposes surely), we can say along with Frege that (q) “John designates something” cannot be a part of the meaning of (p) “John is my friend” because otherwise we could write p as p & q. Under these circumstances, negating p would yield something like ~ (p & q), which is logically equivalent to ~p ∨ ~q. Re-translating this into English, denying p would read “John is not my friend or John does not designate.” Thereby the idea that presupposition survives under negation was born.

A couple of years later, Russell’s theory of description (see here) brought to surface a puzzle which was supposedly overlooked by Frege. No, not the big-time paradox, but a smaller puzzle: sentences that lack proper referents can be meaningful. As we know, Russell’s answer to this was to deny that definite descriptions are logical subjects and to interpret them as conjunctions of propositions. The x is y, according to this long-lived theory, means something like There is an x & There’s no one else who’s x & x is y. What was even more appealing about this theory is that it was able to explain meaning-anomalies in terms of scope ambiguity (it depended on the scope of the negation).

But then Strawson came in the game and notice that The King of France is bald is true in A.D. 1670 and false in A. D. 1770. This led him to the idea that sentences are not true or false; only statements (uttering of a sentence in a certain context) are true or false. In a way, Strawson was back to Frege’s view when he said that the relation between (p) “The King of France is wise” and (q) “There is a present king of France” is such that (q) must be true in order to judge (p) as true or false. He termed this relation presupposition.

This is, certainly, only the beginning of the whole story. First notice that if Strawson’s intuitions are to be formalized (altered[2] in the form p ∥- q & ~p ∥- q), then it can be easily shown that such a system would render (r) The King of France exists as tautological. This means dropping the assumption of bivalence (p ∨ ~p) from the system, and thus allowing a third value of the kind neither-true-nor-false, or modifying the system to allow truth-value gaps. While this may be logically weird (for instance, it renders the good ol’ modus tollens invalid) it is semantically feasible. However, there are – as Levinson puts it – presupposition-like phenomena that don’t seem to fit satisfactorily within this doctrine. Some presuppositions are not about truth or falsity in the first place, but about things from the context (about, one might say, appropriateness). Keenan’s article “Two kinds of presupposition in natural language” (1971) established for good that there is a distinct class of such pragmatic relations between sentences and that these could be analyzed under the heading of pragmatic presupposition.

Although the swerve from Frege to Keenan was rather significant, the ‘negation test’ is still a general procedure of distinguishing presupposition from entailment (negation alters a sentence’s entailments but leaves its presuppositions untouched). This initial distinction gives rise not only to a large class of pragmatic inferences which could be characterized as presuppositions (Levinson is at pains to compile a four-page list of them), but also to an indefinite set of presuppositions generated, as it were, by the same statement. Examining these, and the presupposition triggers involved, Levinson notes that “defeasibility turns out to be one of the crucial proprieties of presuppositional behaviour.” In other words, it is a primary feature of presuppositions that they might be turned off because of certain linguistic and extra-linguistic (background assumptions) context. Though morbid, Levinson’s example illustrates the later sensitivity to extra-linguistic features of discourse.

(84) Sue cried before she finished her thesis (85) Sue finished her thesis

While (85) is a presupposition of (84), it is not a presupposition of another, syntactically similar statement (86) Sue died before she finished her thesis.

A second hot spot is known as the projection problem. Quite early in the development of theories about presupposition, it had been suggested that, if Frege’s Principle of Compositionality is to hold, then it should hold about presuppositions to. That is, if the meaning of a complex sentence S is a sum of the meanings of sentences S1, S2, S3… Sn that form S, then the same must hold about the set of presuppositions holding for S[3]. But this is definitely not the case because in some complex sentences presuppositions of lower clauses “fail to be inherited by the whole complex sentence. In other words, presuppositions are sometimes defeasible by virtue of infra-sentential context” (p. 193) Thus, (1) presupposes (2), but (1’) does not, and both (1), (1’) entail (3) which cannot be negated by a subsequent clause as in (4):

(1) John didn’t manage to pass his exam (1’) John didn’t manage to pass his exam, in fact he didn’t even try. (2) John tried to pass his exam (3) John did not pass his exam (4) *John didn’t manage to pass his exam, in fact he did pass it.

Both defeasibility and the projection problem in their more elaborate form presented by Levinson are highly restrictive as to the capabilities of a semantic theory to comprise and explain such phenomena. Truth-conditional theories have basically no way of distinguishing the context in which (to take defeasibility) presupposition does or does not survive. The “semantic presuppositionalist” (!) can claim ambiguity, e.g. between two types of negation, a presupposition-preserving and a presupposition-discarding one, but this is and ad hoc solution and can easily be shown to fail in other contexts where presupposition “evaporates.”

Pragmatic theories of presupposition – although in a better position to account for the phenomena – are in no easy place. A definition of presupposition as envisaged by the pragmatic accounts stresses the involvement of two, seriously pragmatic concepts: appropriateness (or felicity) and mutual knowledge. Following Gazdar’s account in Pragmatics: Implicature, Presupposition and Logical Form (1979), Levinson cites the following definition:

An utterance A pragmatically presupposes a proposition B iff A is appropriate only if B is mutually known [4] by participants

Thought as such, presupposition seems to overlap with Grice’s conventional implicature: they are not semantic inferences even though the association between the discourse and the presupposition it “filters” is more or less conventional. Some authors even formulated a Grice-like principle for pragmatic presupposition: “organize your contribution in such a way that the conventional implicate of the sentence uttered are already part of the common ground” (p. 209) Sperber & Wilson are among those who offered a pragmatic account outside both semantics (and Montague semantics in particular) and Gricean-like pragmatics.

---

[1] Frege was not using an exact equivalent of “to presuppose.” The term he used, Voraussetzung, means something like “condition”. This fits within his theory because such a “condition” must be fulfilled in order for the proposition at issue to have a truth-value. So John must designate something in order for “John is my friend” to be either true or false. The concept of presupposition was not to be dubbed until half a century later.

[2] The symbol “∥-“ is used by Levinson for “presupposes.”

[3] Again, and for the same didactic purposes I presume, Levinson talks rather loosely about the principle. The “components” of a complex sentence S need not be sentences themselves, but can be formulas with free variables – sometimes called “open” formulas. It was, I believe, Alfred Tarski who first found a way of dealing with this by formulating value assignment functions that would, as it were, “pretend” in the process of interpretation that any sub-formula (be it open or bound) is a sentence. Maybe then, by “sentence” Levinson generally refers to “formulas.”

[4] This condition might prove to be too strong for some cases, so Gazdar suggested the replacement of mutually known by “consistent with”.

Stephen C. Levinson- Pragmatics (part I)

 

Title: Pragmatics
Author: Stephen C. Levinson
Publisher: Cambridge University Press
Year: 1983
ISBN: 0-521-29414-2
Pages: 423

 

 

 

 

As I see it, this is the appropriate follow-up to Yule’s very general introduction to pragmatics (see here). As with any field, it’s good to start with the sources, the origins. But in the case of pragmatics this might turn out to be more difficult than it seems at first blush. The roots of pragmatics are not in one place: if you would want to study the subject of, say, reference you’d have to investigate certain texts from certain fields while trying to do the same for implicature might lead you to a rather distinct area of thought. Also, since the field itself is now fairly ripe (and it had been partially even at the time Levinson wrote this textbook), reading just the sources could be something of a miss. Take the concept of speech act. Reading Austin’s How to do things with words (see here) will get you on track, but it’s doubtful whether it will get you far ahead. It is the job of textbooks to compile the different and often divergent strands from within the field, and it is the advantage of the reader that they do.

 

Pragmatics: history & scope

The first to make use of the term pragmatics was Charles Morris. He was not so much interested in establishing a new term (more the less a new field of inquiry), but in putting some order in what was, in his view, a generalized theory of signs: semiotics. Within this theory, according to the link they study, branches of inquiry could be divided into what later was to be called Morris’s Trichotomy: syntax (who studies the relations between linguistic entities), semantics (who studies the relations between words and the world) and pragmatics (who studies “the relation of signs to interpreters). Carnap seems to have been especially influenced by the trichotomy when he wrote his Introduction to Semantics (1938). Levinson notes, however, that if Carnap’s view of pragmatics had dominated, it would have generated a subject purely interested in the study of deixis (see below).

So what is pragmatics about? What’s within its reach and what falls outside? Levinson goes about answering these questions by way of tackling different possible definitions. Here are some of them:

1. pragmatics is the study of those principles that will account for why a certain sentences are appropriate/felicitous, while others are anomalous (e.g. “Fred’s children are hippies, and he has no children”; “I hereby sing” etc.) 2. pragmatics is the study of language from a functional perspective: that is, it attempts to “explain facets of linguistic structure by reference to non-linguistic pressures and causes” (p. 7) 3. pragmatics is concerned with (following Chomsky’s distinction between competence and performance) the performance principles of language use 4. pragmatics covers context-dependent aspects of language (the grammaticalized, that is, encoded in language, relations between context and language) 5. pragmatics is the study of non-semantic features of language (or, as Gazdar formulated it, “meaning minus truth-conditions”)

Levinson finally settles somewhere between number 4 and 5, with the very important caveat that adopting such a definition means seeing the realm of pragmatics as depending on the semantic theory adopted. However, this defect can be overturned by specifying that the specific aspects of meaning with which pragmatics is concerned are defeasible – in the same sense Grice spoke about his implicatures as cancellable (see below and here). Not very flattering, seeing pragmatics almost as a hodge-podge of what’s left from the formal semantic analyses, but the negative definition certainly “has some cogency”. We simply must assume that, in communication (in producing meaning_NN[1], see here), some inferences are drawn by way of decoding the grammar and adding a semantic interpretation, and some other inferences are not. The latter need this hazy thing called: the context. For the sake of a neat textbook organization and few abstract, meta-theoretical discussions, Levinson builds his exposition around definition number 6.

 

Deixis

The term “deixis” is borrowed from Greek and it means something like “pointing” or “indicating”. The proto-typical examples are: demonstratives (this, that, those), pronouns (I, me, you), time and place adverbs (now, here) and some other features of language tied to the circumstances of utterance (e.g. tense). Pragmatics is concerned with deixis because it is a clear-cut case of the way in which language encodes (grammaticalizes) aspects from the context. An apparent example of the way utterances are anchored in the context: the sentence (1) “Anwar Sadat was the president of Egipt” is true or false whether Anwar Sadat was, in fact, the president of Egipt[2], while the sentence (2) “I was the president of Egipt” also depends on who is uttering the sentence: in order to assess its truth or falsity, we must know who is being replaced by “I”. Needless to say, both sentences are ‘anchored’ in the context due to tense: both (1) and (2), if “I” stands for “Anwar Sadat” are false if they are uttered in 1920.

As we noted above, there are various categories of deixis according to their reference: person deixis (first, second and third person), place deixis (proximal and distal), time deixis (which encodes points and spans of time relative to coding and receiving), discourse deixis (as when we say “This is what you call a squeaky voice”) and social deixis (also called honorifics). The way in which speakers point or indicate obviously cannot be arbitrary. The touchstone, or in fancy words, the “canonical situation of an utterance”, is taken to be the basic face-to-face conversation. Also, deixis is typically egocentric. Quite literally: the ‘ego’ of the communication is always in the ‘centre’: the central person is the speaker, the central time is the coding time, the central place is the speaker’s place at coding time etc.

It should be noted that not all uses of the typical deictic expressions are instances of deixis. Take the pronoun you below:

(deictic usage): What did you say? (gestural[3] deictic usage): You, you, but not you, are dismissed (non-deictic usage): You can never tell what sex they are nowadays

Also, there are some expressions which are not properly deictic but have a “built-in deictic component” (p. 83). These are English verbs like come vs. go. If one says “He’s coming,” he will have said something like “He is moving toward me” – where me must be understood as a deictic expression. However, if we say “I’m coming”, this means “the speaker is moving towards the location of the addressee.”

 

Conversational implicature

I talked about implicature here so many times that I feel going through the whole story again would be pointless. For those of you who are new both to the concept and to the blog, (Hello there!), you can take a look quickly here, then here (and maybe finally here).

Levinson writes what I believe are the most instructive 70-something pages on the subject of conversational implicature. He is, from the very beginning, anxious to point out not only the great many advantages which the concept brings but also the puzzles it gives rise to. “Implicature,” he writes, “stands as a paradigmatic example of the nature and power of pragmatic explanations of linguistic phenomena” (p. 97) The main output of such a theory is that – as Grice in fact pointed out in Prolegomena to Grice (1989) – that over and beyond the stable semantic core there is an unstable, contextual layer of meaning. This layer of meaning may be seen as “the set of possible implicatures”. As long as we are able to define and predict the usage of this set, the theory itself might turn out to be profitable. Before we go further, it might perhaps be handy to stress that this level-division is also present in the nature of the four maxims (Quality, Quantity, Relation & Manner): they are, indeed, maxims – not rules. They express maximally co-operative communication and thus function not as conventions but as putting into practice a rationale for co-operative exchange. As a matter of fact, one of Grice’s powerful arguments was the fact that the same maxims or some derivates of them seem to operate also within non-linguistic behavior.

I’m not sure what to make of Levinson introduction of the term standard implicatures. He is not quoting anybody when introducing the term, not giving it a precise definition. From what I can understand, he means the more comprehensive category, of which the generalized implicatures are only a part. In this case, I’m not sure if particularized implicatures – according to his terminology – are to be opposed to the generalized or the “standard” ones. Whatever is the case, one can still make the Gricean distinction between implicatures which arise from (merely) observing a maxim (call them generalized or standard, the point is they don’t need a special context), and those implicatures which arise from flouting (at least) one maxim. In principle, each maxim can generate each type of implicature (examples are taken from pp. 105-108, the sign “+>” is for “implicates”):

Maxim	Utterance	Implicature	Type
Quantity	John has two PhDs	+> I believe he has +> I have adequate evidence for it	Generalized
Quantity	Is your name John?	+> I don’t know your name +> I want to know if your name is John	Generalized
Quantity	John has 4 kids	+> He has only 4 kids +> He has kids	Generalized
Quantity	The flag is white	+> The flag has no other colours	Generalized
Relevance	Pass the salt	+> Pass the salt now	Generalized
Manner	John packed and left	+> John first packed and then left	Generalized
Quality	A: How’s the whether in Britain? B: Oh, you know Britain. All sunny and blue skies…	+> It’s an awful weather, as it generally is here in Britain	Flouting
Quality	That gimnast is made of plastic!	+> She’s not made of plastic, she’s as if she’s made of plastic, she’s flexible	Flouting
Quantity	War is war	+> “terrible things always happen in war, that’s its nature and it’s no good lamentig that particular disaster” (p. 111)	flouting
Quantity	Either he’ll come or he won’t	+> “calm down, there’s no point in worrying about whether he’s going to come because there’s nothing we can do about it” (p. 111)	Flouting
Relevance	Johnny: Hey Sally let’s play some tennis Mother: How’s your homework getting along, Johnny?	+> you may not be free to play yet	Flouting
Manner	Miss Singer produced a series of sounds corresponding closely to the score of an aria from Rigoletto	+> Miss Singer is a bad singer	Flouting

 

Now, while these seem clear-cut examples of some meaning being conveyed beyond what is being said, some theorists (e.g. J. Sadock in “On testing for conversational implicature”, 1978) readily pointed that some equally simple instances fail to meet all the traditional requirements. These “traditional requirements” refer to what Grice called the “characteristics” of implicatures: they are cancellable, non-detachable, calculable and non-conventional (for a clearer exposition of some of the difficulties, see here). Levinson points just some of the puzzles: some implicatures seem to be part of the semantic content (like “and” being read as “and2”= “and then”), some implicatures seem to be detachable (e.g. scalar implicatures) and some implicatures seem to be generated by non-cooperative (or loose) talk.

Probably the point where the theory of implicatures goes into deep water is the concept of conventional implicatures. Acknowledged as a concept, the conventional implicature would mean something like this: there is a type of meaning which is (a) non-truth-conditional but (b) not derived from the principles of conversation and (c) attached to certain linguistic expressions by convention. Conventional implicatures will be everything conversational implicatures are not: non-cancellable, detachable, not-caluculated etc. How are they, then, not part of the truth-conditional meaning? Grice’s example is pretty straightforward: the difference between two conjunctions (and and but) in the way they chain propositions. So, for instance, we can have:

(3) The restaurant is cheap and pretty near
(4) The restaurant is cheap but pretty far
(4’) The restaurant is cheap but fancy

In both (4) and (4’) there is an opposition implicated between the two predicates (“cheap” and “pretty far” in the first place, and “cheap” and “fancy” in the second). Note that, in a normal context, to say “The restaurant is cheap but pretty near” would be anomalous (unless uttered in a conversation where we want to decide what is the cheapest most farther restaurant we could go to). On the other hand, uttering (3) does not carry the same implication of some opposition being conveyed. Note that, in a normal context, saying “The restaurant is cheap and fancy” would, in a normal context, eventually implicate (4’).

There are other conjunctions which bring nothing to the truth-conditional meaning of the terms (Grice notes “therefore”, Levinson notes “however, besides, anyway, well, still” and others, p. 128). Also, social deixis (sir, madam, your honour etc.) could be explained in terms of conventional implicature. Another important aspect discussed (with which I will deal closely after the second part of this review) is the possibility of scalar (quantity) implicatures and that of causal (quantity) implicatures. The latter touches directly upon the subject of unexpressed premise. Bear in mind that in (Eemeren & Grootendors, 1984) unexpressed premises were treated as implicatures and it wasn’t until (Eemeren & Grootendorst, 1992) that the authors preferred the concept of indirect speech act.

In the next part I will review the next 3 chapters: Presupposition, Speech acts and Conversational structure.

---

[1] Levinson has an interesting way of explaining Grice’s concept of non-natural meaning: “Suppose that Moriarty says that his watch broke, and from this Sherlock Holmes infers that he perpetrated the crime: although the information may have been indirectly conveyed, we should not be loath to say that Moriarty communicated it” (p. 15). See also here.

[2] Note, in passing, that by saying this we (along with Levinson, I’d say) are implicitly subscribing to a truth-conditional semantic theory.

[3] Deictic usages can be split between gestural (those which must be accompanied by a gesture, e.g. This is genuine but this is fake) and symbolic (all the rest). Non-deictic usage can be split between anaphoric (when a term picks out as referent same entity previously picked out, e.g. Mary heard and she knows), and non-anaphoric (all the rest). Note also that, since some linguistic expressions are typically deictic expressions, even when they are not used to produce deixis, authors still refer to them as “non-deictic usage of deictic expressions”

Technorati Tags: Pragmatics,Stephen Levinson,pragmatic,conversational implicature,implicature,deixis,deictic expressions,Grice,Herbert Paul Grice

The open door. A compilation

I want you to close the door
I’d be much obliged if you’d close the door
Can you close the door?
Are you able by any chance to close the door?
Aren't you going to close the door?
Would you close the door?
Won’t you close the door?
Would you mind closing the door?
Would you be willing to close the door?
Would it be convenient/hard/too much trouble for you to close the door?
Must you leave the door open?
Should you leave the door like that? 
You ought to close the door
It might help to close the door
Hadn’t you better close the door?
May I ask you to close the door?
Why not close the door?
It would be a good idea to close the door
I would appreciate it if you’d close the door
Would you mind awfully if I was to ask you to close the door?
I am sorry to have to tell you to please close the door
Did you forget the door?
Do us a favour with the door, love
How about a bit less breeze?
Now Johnny, what do big people do when they come in?
Okay, Johnny, what am I going to say next?
Johnny, how many times have I told you to close the door?
Johnny? The door.
Johnny?
*Chm, chm* Johnny!
Do you like sudden deaths, Johnny?
Not agaaain…
NO way. No fu#@%ng way! This is just… JOHNNY!
[*straight face*]
…

(Searle, 1979, pp. 36-39) & (Levinson, 1983, pp. 264-265)

Can you think of others?

Studies in Pragma-Dialectics – part III

 

Title: Studies in Pragma-Dialectics
Editors: Frans H. van Eemeren & Rob Grootendorst
Publisher: Sic Sat, Amsterdam
Year: 1994
ISBN: 90-800777-5-5
Pages: 291

First part here, second part here

Where were we with this? A, yes. The empirical component. For the first two, click here and here (read a generalized conversational implicature of consecution in that “and”). For an introduction to pragma-dialectics, click here, and finally, for some technicalities regarding what is “pragmatic” about pragma-dialectics click here and (then) here.

 

---

 

In the introduction to this series I said that reviewing the empirical component will be fun. Of course, each has his own idea of fun, so let me elaborate. To me, the concept of “conventional validity” is still in deep haze. I think this is mainly because, while pragma-dialecticians are anxious to point the clear gap between the ideal (normative) model and the actual argumentative reality, testing for conventional validity seems like a “bridging” activity. In other words, while in actual practice pragma-dialecticians have a very clear (and thoroughly explicated) idea of how should one handle the two dimensions, it feels like the moment they start working with the model itself (as opposed to working with linguistic input from the argumentative reality), the clear-cut line starts fading. Since the properties of the model (its problem-solving and its conventional validity) might turn out to be contrary to the general practice, I don’t seem to find an answer to the question: “What is the relation between the, let’s call them, rebutters and the model?” As I see it, the few sections from (Eemeren & Grootendorst, 2004, pp. 15-17) and the passage below (Eemeren et al., 1993, p. 23) either avoid a definite answer or I’m simply unable to see it [all italics are mine]:

“Empirical research can provide an important basis for evaluating [1] the validity of normative models of argumentation. Normative models can be evaluated against the two criteria of problem-solving validity and conventional validity mentioned in chapter 1. Both concepts of validity have empirical implications [2]. Problem-solving validity depends on the adequacy of the model as a description of effective practice – its ability to discriminate good argumentation from poor. To the extent that actual argumentative practice follows [3] the standards of a normative model but results in intuitively questionable [4] procedures or to the extent that actual argumentative practice departs from the standard but results in intuitively acceptable [4’] procedures, we should be skeptical [5]of the model’s problem-solving validity”

So, for each italicized word, respectively: (1) really? empirical research provides basis for evaluating the normative model?, (2) don’t they mean it the other way around?, that empirical datum might have theoretical implications? (3) does it ever? wasn’t that the whole point – that it doesn’t?, (4) according to whom? who’s intuition are we then following? the analyst only has the model and the interactants only have the practice – whose (neutral) intuition are we going to use if the two don’t connect?, (5) what does “be skeptical” means? does it mean “be skeptical” and leave it at that, or “be skeptical and thus do something with/to the model”?

Okay. Now we can review some of the articles from the empirical component. You can imagine that, biased as I am by my inability to answer (1)-(5), I cannot but find these articles as evincing precisely the same categories of strangeness.

 

Frans H. van Eemeren, Kees de Glopper, Rob Grootendorst and Ron Oostdam - Student performance in identifying unexpressed premises and argumentation schemes

To all the methodological objections above (which, again, may simply be thrust by my misunderstandings), this article is even more fun – in the above sense – because of the subject: unexpressed premise. I’m not going into detail with the subject of unexpressed premises, but consider this. About the step from the logical analysis to the pragmatic analysis (i.e. from the logical minimum to the pragmatic optimum – if confused, click here), they say [my italics]:

It goes without saying that the generalization should fit in well with the rest of the discourse and that the validity of the argument is to be maintained and no unwarranted commitments may be ascribed to the speaker (90)

It most certainly goes without saying. The only problem is their example is something like this:

(the enthymeme) “Amos is pig-headed, because he is a teacher” (the logical minimum) “If Amos is a teacher, then he is pig-headed” (the pragmatic optimum) “Teachers are pig-headed”

In what logic is: “Amos is a teacher + Teachers are pig-headed = Amos is pig-headed” valid? Well, in some (in predicate logic, maybe, if we manage to disambiguate the bare plural in the second clause), but definitely not in the same deductivist modus-ponens-ish logic. It would have been if the pragmatic optimum had been “All teachers are pig-headed.” But, as we can see, it’s not. So what are we to make out of this? That the notion of validity changes from one analysis to another? It’s all very confusing…

(To touch upon yet another thorny issue, one of their other examples is “Daniel is an actor [and actors are essentially vain], so he is certainly vain” (91, my emphasis). I talked about it here, points 4-6. Also, does the “certainly” in the conclusion means that now we can switch to modal logic? It’s all very confusing…)

So the empirical part goes like this. Hypotheses are formulated about the students’ abilities at (1) correctly identifying unexpressed minor premises and (2) correctly identifying causal argumentation (in fact, identifying it easier than the other two schemes). The method is: paper-and-pencil, multiple choice, not-mistake = skill. The design: facets, according to e.g. indicators or other stuff that might influence the choice. But now comes the funny part:

“The concepts of ‘unexpressed premisses’ and ‘argumentation schemes’ were defined with the help of examples. Furthermore, some examples of items were presented to demonstrate the test task” (94)

Oh, this is quite convenient. So you tell the subjects that “2+2=4” and that “3+3=6” and so on, and then you ask “4+?=8”. Moreover, you don’t just let them replace the “?” with their own choice, you say: “A. 4, B. 5, C. 79, D. 0”. The subjects pick variant A. and you’ve gathered empirical evidence for the fact that arithmetic is conventionally valid.

If this is taken into consideration, then the actual research is almost irrelevant (though well conducted, as appears from the article). Note, in passing, that the generic reading of the pragmatic optimum, due to its bare plural, is kept in the correct answers of each structuple. They illustrate:

He is a bad singer, because he cannot keep time
A. One who cannot keep time is a bad singer
B. Bad singers cannot keep time
C. Singers sometimes cannot keep time
D. One who cannot sing is also unable to keep time (95)

In case you’re wondering, the results were OK. The guys were guessing unexpressed premisses like there was no tomorrow. Moreover, in the concluding section, the authors note how “obvious” the Teachers are pig-headed type of gap-filler seems to be, almost forgetting that they told them that that answer-type is correct.

 

Bart Garssen - Recognizing argumentation schemes

Bart Garssen begins thus: “My starting-point is the pragma-dialectical typology of argumentation schemes. Ultimately, my object is to assess the feasibility of this typology for ordinary language users.” (105) A very important question; and an interesting one. How does he set about testing this? By making a (clever, if you ask me) step from “characterize the scheme” – that is, the particular type in the typology which corresponds to a link from premises to conclusion – to “criticize the scheme”. This is an evidently correct move in the pragma-dialectical approach because there is a precise connection between each scheme and its critical questions. So instead of saying (this argument is an instance of) “symptomatic”, “causal” or “analogy”, the respondents were ask to criticize the arguments. Their criticism – in the form of critical questions – was then recognized within the typology as pertaining to this or this scheme. Again, this is a clever move and it definitely made the research much easier.

So all the respondents had to criticize, while half of them also had to point to the particular scheme: “quality” (replaced “concomitance”), “analogy”, “causal”, “don’t know.” Some facets were designed according to the variable “with/without indicator.” So far so good.

But again, the elephant presents himself at the dinner table. This time, it is recognized as an elephant and the author tells us that he tried to kick him out of the house – but he doesn’t tell us how that work out. Let me be more specific.

The subjects are again instructed by means of a “short explanation concerning the three argumentation scheme” (108) The possibility of a damaging E-bias was apparently too great, so the author tried to control this variable by splitting it in half and checking for variances. This, at least, is how I read the paragraph “I have used two different instructions. One instruction provides information about premisses and standpoints, the other instruction contained additional information concerning the three argumentation schemes” (109) Then, cross-assigning these two variables, we get the following table (which looks great, and the elephant seems all gone):

	without indicators	with indicators
without instruction in argumentation schemes	Yaay!
with instruction in argumentation schemes

So, I know what you’re thinking: “Finally someone gets rid of The Elephant of Prior Instruction”, right? I just realize that E-bias might actually come from Elephant-bias! and that “the elephant is at tha table might mean “in the table”, as in our case” :) The “without instruction - without indicators” cell? That’s the best, right?! That’s as non-biased as you get. Well, you’re up for a disappointment, because here’s how the two variables were in fact combined with the facets:

	without indicators	with indicators
without instruction in argumentation schemes	critical comments	critical comments
with instruction in argumentation schemes	critical comments + multiple choice	critical comments + multiple choice

Oh, snap…

From this point on, it takes our Elephant three steps to get back at the table (i.e. – let’s put it a bit more seriously – for the “prior instruction” variable to occur as an unchecked independent variable and for it, instead of “subjects’ ability”, to influence the results)

Step 1. No multiple choice for the uninstructed subjects. Are we then able to tell apart the effects of the instruction? No.

Step 2. First row of the table: buh-bye! In a footnote, the author assures us that subjects have been randomly assigned to (1)-(4) in order to avoid class-effect. Good. But pay attention to what follows (same footnote, 108): “Condition 1 and 2 refer to the first research question, condition 3 and 4 to the second. In this article I only report the results of condition 3 and 4.” Wait, what?! So after he went through that trouble formulating the first, elephant-banishing research question (remember? “Do ordinarily language users who interpret an utterance as an argument and are not formally instructed in argumentation schemes have any articulated idea of the nature of the link between the premisses and the standpoint? [my italics]); and after he went through all that trouble checking the variable “instruction” by forming what could be seen as a control group of unsophisticated newbies, we’re not seeing the results of it? Why do it in the first place?

Step 3. Don’t even include in statistics. Indeed, even if not concretely considered, the statistics of (1) and (2) might have been of some importance, right? For future research, if nothing else. The table with the results, unfortunately, looks something like this:

	without indicators	with indicators
	%                                 sd	%                               sd
concomitance	71                              19.5	73                             18.7
analogy	81                              13.7	83                             14.4
causality	87                              13.0	89                             13.0

The results are then candidly announced: “According to the mastery norm of 80% the respondents are in general quite capable of identifying the three types of argumentation” (110) and “The research results show that ordinary language users who are briefly instructed about the schemes are quite capable of identifying the three prama-dialectical argumentaiton schemes.”

 

---

To conclude: Even if all the previous objections are answered – which can be the case, I only said I couldn’t find a satisfying answer, not that it hasn’t been given – the empirical component of the pragma-dialectical approach was in 1994 at an incipient stage where, although seemingly trying to refute conventional validity, researchers were only proving – over and over, and with great care – the fact that the pragma-dialectical theory is easily appraisable, that once taught, the model is recognizable. This, however, leaves the occurences of adverbs such as “intuitively” (used when explaining the validity of the model) quite inexplicable. 

Technorati Tags: pragma-dialectics,Frans H. van Eemeren,Bart Garssen,Studies in Pragma-Dialectics,argumentation schemes,empirical research,unexpressed premises,logical minimum,pragmatic optimum

Brief review of a biref history of argumentation theory. Briefly.

 

Title: Histoire des théories de l’argumentation
Authors: Philppe Breton, Gilles Gauthier
Publisher: La Découverte
Year: 2000
ISBN: 2-7071-3175-X
Nr. Pages: 123

 

 

 

 

 

 

 

Small book, big subject, general introduction. Which is good. Or bad. Depends. It’s good to have an overview of the field one is interested in, but the temptation to borrow ready-made assessments is always a bug, isn't it? The authors are honest and state from the very beginning that the aim of this short study is to offer a historical introduction – which I guess the careful reader must understand as terribly uncritical and at times vague. I was actually hoping to get at least a minimal grasp of the Francophone tradition in argumentation study. Considering that both the authors, Philippe Breton and Gilles Gauthier belong to that tradition, I thought at least one of the main interests would be a careful out-sketch of it.

Ignoring the danger of anachronism for a moment, we could say, the authors assure us, that argumentation theory was born around 450-440 BC, due to precise social factors, in a precise historical environment. As anything stemming from that era, the birth of rhetoric is something of a legend, tied to an old teacher named Corax and his pupil Tisias. According to Cicero (who cites a lost fragment from Aristotle), there is no knowledge of anybody else teaching rhétoriké (Gr.) before Corax. Again, according to the legend (supported, nevertheless, by Roland Barthes and some historians) the idea of persuading was especially connected to the processes going on in Syracuse at that time. The city, a tyranny for longtime, knows a period of (ignoring anachronism once more) what we would now call a “democratic revolution.” In these circumstances, people literally had to argue for their land, originally, and be citizens, in some later political developments. A third thing that happened was the rise of a new judicial system:

Ces procès étaient d’un type nouveau ; ils mobilisaient de grands jurys populaires, devant lesquels, pour convaincre, il fallait être éloquent. Cette éloquence […] se constitua rapidement en objet d’enseignement [1] (p. 12)

Consequently, anyone who was anyone in that era grabbed the now famous rhetoric manuals. The authors seem to suggest that one can only dig out minor, external differences between these courses (Corax’s manual, for instance) and today’s “public speaking” manuals. Both could be seen as compilations of either (a) language tips & tricks, that is, stylistic devices, or (b) commonplaces, both of which students learned by heart.

Still within what the authors call “the birth of rhetoric,” the work of sophists “inaugure la conscience durable d’un vertige, celui d’un monde qui serait entièrement relatif au langage, créé et contenu dans la seule parole humaine. C’est ce vertige que tenteront de stabiliser Socrate et Platon, inaugurant ainisi la tradition philosophique qui tente en permanence depuis de réduire la sophistique à l’ombre, nuisible, portée par la philosophie” (15) Depending on who you ask, the sophists’ skepticism was either the best thing that could happen to argumentation theory – they were, after all, the first to indeed theorize about language – or the worst thing that could happen to reasonableness – they were, after all, the ones to teach individual success based on speaking skills, a big no-no for Occidental thought.

Socrates and Plato, though not directly engaged in creating proto-argumentation theory (like sophists, before them, and Aristotle, after them), had an enormous interest in cautiously inserting one particular view of language – fr. langage – into their epistemological views about what’s up with all this world just… being around. I find it particularly interesting that we can describe the proto-theories as descriptive (they were based on the rhetors’ past experiences of what works and what doesn’t), and accordingly see the Platonic movement as a normative (evaluative, dialectic etc.) turn in the proto-theoretical process. The critique that Plato and Socrates advanced against rhetoric was not a general on, but one which was focused on its “not-tied-to-the-perfect-Idea-ness” (be it Truth or any other value), a feature which the sophists tended to propagate. The “dialectical tiers” Christopher Tindale talks about here stem, I believe, from this particular historical view.

In this scheme, Aristotle falls somewhere in the middle. Although he stands in opposition to the ad hoc, cynical, “techniques mercenaries” of sophists, he also takes a dim view of Plato’s elitism, whose ideal City, viz. The Republic, is a place where citizens have little freedom for actual speech – dialectically tied or not. Keep in mind that Aristotle’s Rhetoric starts with this phrase: “Rhetoric is an antistrophos to dialectic” (Aristotle, trans. Kennedy, p. 30). Also, it is not a collection of prefabricated commonplaces, but a functional account: it devises methods in view of a (practical) purpose, that of persuading. The famous three genres (forensic, deliberative and epideictic) can then be seen as proto-“activity types” in which certain goals correspond to certain methods (enthymeme, example and amplification, respectively), and the methods are to be evaluated within that specific context.

After this, and with the notable exception of Cicero, argumentation theory went to bed for about two millennia – again, assuming that until now we’ve actually been talking about argumentation theory and not something else. The authors duly note the strange setting in which this happened:

“L’importance de la théorie argumentative va décroître au sein de la rhétorique au fur et à mesure, paradoxalement, que celle-ci va voir son rôle s’accroître et devenir finalement le contenu de tout enseignement” [2] (31)

This allows us to fast forward to the 1958 Moment, the appearance of Perelman & Olbrechts-Tyteca’s Traité de l’argumentation and that of Stephen E. Toulmin’s The Uses of Argument. The two authors begin then the third part of the history, the rebirth, with a somewhat detailed account of these two major works. Two main ideas are stressed with regard to both books: the fight against formal logic’s appropriateness for the study of arguments (though, to be precise, this fight was fought both on different grounds and with different tools), and the quest for a firm ground to study argumentation. For Perelman & Olbrechts-Tyteca, this firm ground was the “universal audience” – with all its constraints upon the speaker’s rationality, for Toulmin, it was the famous Model. What follows is a brief, fairly instructive summary of P&O-T’s schemes as they appear in the treaties. What I appreciated indeed was Breton & Gauthier’s first paragraph from the section on Toulmin: “La théorie de l’argumentation de Toulmin est aujourd’hui connue et reconnue surtout pour son célèbre modèle de l’argument. Cette réduction est caricaturale dans la mesure ou elle occulte sa préoccupation épistémologique fondamentale. […] À ce propos, il faut souligner la très grande originalité de Toulmin : il est le seul, ou du moins le premier, à s’opposer explicitement au positivisme et à son avatar logiciste en développant une théorie de l’argumentation” (54) Also, they offer an extensive treatment of Toulmin’s categories of arguments (analytic vs. substantial, warrant using vs. warrant establishing, necessary vs. unnecessary [or probable] etc.) and his major arguments against the “logiciste” reduction of all these to the analytic one as a paragon of reasonableness. I’ve never seen Toulmin’s model in French, so let me just have a dish of copy pasta right below:

	données (D)		Donc, (Q), conclusion (C)
		Vu que (G) garanties	|       si non (R) restrictions
		En vertu de (F) fondement	qualification modale

 

In the last chapter, Les recherches contemporaines en argumentation et rhetorique, the notable features of each major theory – or theorist(s) – are reviewed. It was surprising to find that, finally, francophone theories decided to translate “fallacy” as fallacie (Plantin) or, the authors’ choice, fallace. Yaay! We can now talk about Hamblin’s Fallaces. Aside from that, the authors repeatedly not that the field as a whole looks disintegrated, with novel ideas scattered here, with stable research programs, each with its fundamental differences. This, of course, can be good or bad: good, because a variety of answers to (more or less) the same problems allows for criticism and improvement; bad, because unless basic assumptions are left aside, communication between branches is virtually impossible. I’ll finish with a who’s-who list of the major authors briefly considered:

Anglophone research: Charles Hamblin, Trudy Govier, John Woods-Douglas Walton, Charles Willard, Anthony Blair, Ralph Johnson, Frans H. van Eemeren & Rob Grootendorst

Francophone research: Jean-Blaise Grize, Georges Vignaux, Chrisitan Plantin, Michel Meyer, Olivier Reboul

---

[1] This new type of legal process mobilized large popular juries in front of which, in order to convince, one had to be eloquent. It is this eloquence that turned rapidly into an object of study.

[2] The importance of argumentation theory will decrease within the field of rhetoric, paradoxically, as the latter’s role will rise and become, eventually, the content of learning as a whole.

H. P. Grice & Communication

(figure adapted after Levinson, 1983, p. 131)

An introductory glimpse into Relevance Theory

This post reviews the following article: Wilson, D. (1994). Relevance and understanding. In G. Brown et al. (Eds.) Language and understanding (pp. 35-59). Oxford: University Press

Relevance theory is, before anything, a pragmatic theory. There is, or should be by now, a steady consensus about what the word “pragmatic” means; for the purpose of this review we should note tentatively that pragmatics is concerned with the way in which the context contributes to language understanding. Levinson (1983, Chapter 1) argued for the translation of the above definition (or some similar ones) into an algebra-like relation like ‘pragmatics = meaning – semantics’ which I believe is an appropriate first-base conception. Wilson’s introductory article to relevance theory certainly holds to such theoretical delineations. Let us now follow the train of reasoning from the article, which is of the “example(s), then explanation” sort.

First, Wilson states clearly that she is concerned with overt communication. The author gives the example of a senator holding a speech: he or she intentionally communicates some information, whereas unintentionally he may communicate all sorts of things, e.g. that he or she is nervous, or rich or intelligent.[1] The question arises: how do we understand what (s)he says? Although at least a minimal amount of linguistic encoding-decoding is performed (indeed, necessarily), there is definitely a gap between what the sentence means in virtue of its semantic and syntactical features, and what the speaker intended it to mean in virtue of it (the sentence) being uttered in a specific context. As Wilson notes – maybe something of an overstatement – “virtually any utterance can be used to show that this hypothesis [meaning = encoding/decoding] is wrong.” (p. 38)

What, then, is in this gap? Three categories of meaning[2]: (a) What did the speaker intend to say? (b) What did the speaker intend to imply? (c) What was the speaker’s intended attitude?, and finally, after a minimal theoretical input, (d) What was the intended set of contextual assumptions? To show that categories (a)-(d) are not part of the syntactic-semantic structure is a rather tedious job most of you are familiar with (either intuitively or theory-driven). We have stressed some similar points here and maybe here. Let us, however, follow the train.

I think Wilson’s example of (a) is rather obscure, so let us take a more clear-cut instance: deixis. It is only after we look at the context that the deictic I in the utterance I am in Amsterdam now can be explained (by pointing out who’s the source of the utterance). In fact, semanticists have a really hard time describing the truth-conditions of this utterance solely by appealing to linguistic explanation: it seems that we can spell out the conditions which would make I am in Amsterdam true only after we use the context to disambiguate the deictic I.

Anyway. Back to Wilson. For (b), any instance of conversational implicature could be used. This time, her example is clear. In a BBC radio interview, Mrs. Tatcher used the following utterance: I always treat other people’s money as if it were my own. Not only that this is an instance where what is meant (Mrs. Tatcher treats public money carefully) is more than what is said, but it is also an instance where the hearer must activate some additional knowledge to cancel the (perfectly valid, though improbable) implicature: I waste my money, so I am wasting yours as well.

With (c), we turn to a rather different type of pragmatic inference (judging from the abstraction made from (a) & (b) to (c) and (d), we could say a higher level inference – which would not be very far to the idea of knowing/recognizing speech events). Regardless of the propositional content of the utterance, one extra-linguistic decision the hearer must make is this: “is [the speaker] endorsing these propositions or dissociating herself from them” (p. 41)

In order to illustrate (d), we must introduce (I say “we”, I always say “we”, but let it be clear, there’s hardly ever any Is in all the wes I’m using) some of Wilson’s notion. So far we’ve seen that almost every time we inquire about what’s in addition to the utterance literal meaning, we more or less arrive at the answer: the context. What is this context? As any pragmaticist will tell you: eeer! wrong question. At the beginning of pragmatics (namely, in the sixties, say scholars like Hymes, Frith) there have been several attempts to create (invent?) from an ethno-anthropological perspective a taxonomy of possibly relevant contextual information. While providing a resourceful list of heuristics of context usage, no real theoretical underlying principle was ever provided. They were, in other words, ad hoc generalizations of the sources of pragmatic inferences. This, however, is not necessarily a problem (although it is a thorny aspect of pragmatics as a domain which, remember, is defined by making reference to the concept of “context”). As Wilson proceeds, the context is simply: “the set of assumptions brought to bear in arriving at the intended interpretation” (p. 41). Quite circular, but since it does not touch upon the predictive potential of the theory as a whole we’ll accept.

So the context helps hearers “narrow down” the intended interpretation. As the attentive reader, again, observed, this is simply recasting the problem in different terms because the same question remains. Thus (d): how do we know which intended assumptions language users use to arrive at the intended interpretation. Though trivial, and this point is especially well argued by the author, many theorists somehow assumed that the intended set is the intended set (not sets). But more often than not, this is not the case. To take the author’s example, if one asks How is your new partner? and the other party responds says Has a lot in common with X, a whole range of properties of X might possibly turn out to be part of meaning (they are both human, they are both mad, they are both tall, they both have parents etc). So the question is (d) How do we know which set of (X’s properties) assumptions is an utterance intended to activate?

“We do look for some implications, of course; but what we appear to do is to choose the minimal set of implications that would make the utterance worth listening to” (p. 43) This is the starting point of the relevance theory. When the range of possible speech-triggered interpretations occur, some faster some slower, since the complexity differs[3], there must be some criterion-like principle which is powerful enough to exclude all but at most a single interpretation. The criterion is, you guessed it: relevance.

Since every utterance is a request for attention it, as it were, arouses expectations of relevance. The hearer will choose that set of contextual assumptions which will fit these expectations of relevance. But what is relevance? In Sperber & Wilson (1986), relevance is defined in terms of contextual effect x processing effort. An utterance achieves contextual effects if it scrambles in some way some previously existing assumptions (whether this means denying, eliminating, confirming or using them). But any contextual effect comes along with the corresponding cognitive processing effort for the hearer’s part (see note 3). Consequently, relevance is a function of these two variables: “the greater the contextual effect, the greater the relevance; but the greater the processing effort, the lower the relevance” (p. 46) But the authors go one step further by saying that, as long as an interpretation fits the (“aroused”) relevance expectations, we stop at that one. 

So we as communicators find an interpretation which yields the greatest effect with smallest efforts and stop. But we are definitely not looking for the most relevant interpretation (since we are not even comparing all the possible ones, which is practically impossible because every utterance entails an infinity of equally true utterances[4]). So what do we do, as rational interactants? We choose the optimal relevance. An utterance is optimally relevant if “(a) it achieves enough contextual effects to be worth the hearer’s attention, (b) it puts the hearer to no gratuitous processing effort in achieving those effects” (p. 47).

Part (a) means that the utterance must not be (1) already an assumption, (2) information which the hearer is already processing at the time. At this point, the author reminds us that we are speaking of overt communication (in our words, intended meaning), so we must presume that the speaker is not leaving an easier (and equally effective) information to chip in the interpretive process. Therefore, the hearer assumes that the speaker is being cooperative, and it is now that the hearer activates clause (b): the first easy and effective interpretation is the intended one. Of course, he can miss. But this is what he does.

So behind all this cognitive back and forth, there’s a consistency criterion of the kind: An utterance is consistent with the principle of relevance iff the speaker might rationally have expected it to be optimally relevant to the hearer on that interpreatation. (p. 49) In other words: Did you find an interpretation which satisfies your expectations of relevance, satisfaction which the speakers could have foreseen? That is the one. As the authors say, any utterance has an “in-built stopping place.”

In the remainder, the author deals with the differences between this theory and Grice’s Cooperative Principle. I won’t go into that because the glimpse, as far as I’m concerned, is complete without it.

---

[1] Although not explicitly stated, this right of the bat restriction is nothing more than an application of Grice’s distinction between natural meaning and non-natural meaning (often abbreviated meaning_NN). See (Grice, 1957)

[2] A really attentive phrasing would have replaced ‘meaning’ here with something like non-linguistic inferences which amount to additional understanding.

[3] Consider for instance, in the Has a lot in common with X example, the assumptions “Both are annoying” vs “Both have a brother which also plays tennis with his left hand”

[4] The usual p → (p v q)

Technorati Tags: Relevance Theory,pragmatics,Levinson,Meaning,Grice,Herbert Paul Grice,Understanding,Communication and Cognition