Dan Sperber & Deirdre Wilson (1998) Irony and relevance: A reply to Drs Seto, Hamamoto and Yamanashi. In R. Carston & S. Uchida (eds.), Relevance theory: Applications and implications (John Benjamins), 283-293.

“…In this brief reply, we will look at three main issues. First, is verbal irony necessarily echoic? Should a category of non-echoic irony be recognised, as Drs Seto and Hamamoto propose? Second, is there a clear-cut boundary between ironical and non-ironical utterances, or are there borderline cases, as Dr Yamanashi suggets? Third, can the relevance-theoretic account of irony shed light on a range of more complex cases, including those discussed by Dr Hamamoto? We will end with some more general reflections on whether irony is a natural kind…” [PDF version]

Irony and relevance: A reply to Drs Seto, Hamamoto and Yamanashi

Dan Sperber and Deirdre Wilson

1. Introduction

The papers by Drs Seto, Hamamoto and Yamanashi make a valuable contribution to the analysis of verbal irony, both practical and theoretical. We would like to thank them for their positive comments on the relevance-theoretic account, and for the interesting questions they raise. In particular, we are grateful to Dr Hamamoto for his analysis of the relations between verbal and situational irony (based on the writings of Professor Kawakami), to Dr Seto for his insightful discussion of a variety of markers of echoic use, and to Dr Yamanashi for drawing attention to many problems with standard approaches to irony (for example, the fact that the interaction between metaphor and irony is incompatible with standard approaches).

In this brief reply, we will look at three main issues. First, is verbal irony necessarily echoic? Should a category of non-echoic irony be recognised, as Drs Seto and Hamamoto propose? Second, is there a clear-cut boundary between ironical and non-ironical utterances, or are there borderline cases, as Dr Yamanashi suggets? Third, can the relevance-theoretic account of irony shed light on a range of more complex cases, including those discussed by Dr Hamamoto? We will end with some more general reflections on whether irony is a natural kind.

2. Must irony be echoic?

Hamamoto and Seto see what Hamamoto calls the “vagueness of the echoed source” in many examples as a problem for our account. Nonetheless, we would like to defend the view that verbal irony is necessarily echoic.

The notion of echo we used in analysing irony is a technical one; it is deliberately broad, and goes beyond what would generally be understood by the ordinary-language word ‘echo’. It covers not only cases of direct and immediate echoes, as in (1), where B sarcastically repeats what A just said (and even these are not literally ‘echoic’ in the ordinary sense of the word), but also echoes of (real or imaginary) attributed thoughts, as in (2), and echoes of norms or standard expectations, as in (3) and (4):

(1) A: I’ll be ready at five at the latest.
…..B: Sure, you’ll be ready at five.

(2) A: I’ll be ready at five at the latest.
…..B: You mean at five tomorrow?

(3) A: I’ll be ready at five at the latest.
…..B: You’re so punctual!

(4) A: I’ll be ready at five at the latest.
…..B: It’s a great virtue to be on time!

This notion of echo is broad, but it does have limits. In the first place, given the background knowledge of speaker and hearer and the mechanism of verbal understanding, most utterances cannot be understood as echoic: there is no accessible representation that they might be taken to echo. In the second place, within the framework of relevance theory, an echoic interpretation is acceptable only if it contributes to the relevance of the utterance for the hearer (or may seem to the speaker to do so). Thus, if Peter asks at dinner “Could you pass the salt?”, and Mary utters the same words a while later, her utterance will not be understood as echoic, even though there is an accessible representation that she might be taken to echo. In normal circumstances, her utterance will achieve the expected degree of relevance as a genuine indirect request, and other, costlier interpretations will be disallowed. The communicative principle of relevance, and the criterion of consistency with the principle of relevance, place substantial constraints not only on when an utterance can be interpreted as echoic, but on what the source of the echo may be taken to be.

Despite these restrictions, we do claim that it is always possible to echo general norms or universal desires, or their particular instantiations in individual cases. Moreover, as Seto shows, irony can combine with emphasis or hyperbole: the original representation is exaggerated, so that its inappropriateness in the context becomes even more manifest. Take Martin’s example (Martin 1992: 80), quoted by Hamamoto:

(5) Oh great. That’s nice

It is always highly desirable that things should turn out well. We claim that someone who says emphatically that things are great when they patently are not does achieve her ironical effect by echoing a representation of what is always desirable. Similarly, “I like that”, in Seto’s two examples (6) and (7), echoes the universal desire for things to be such that we can like them:

(6) A: Bob has just borrowed your car.
B: Well, I like that!

(7) I like that. Bob smashes up my car and then expects me to pay for the repairs.

Seto’s examples (20)-(28), which so well illustrate the various linguistic and rhetorical resources that may be used to highlight the ironical nature of an utterance, also involve the apparent expression of highly positive judgements. Hamamoto’s example (8) echoes a more specific representation, shared by speaker and hearer, of what is standardly desirable in the matter of looks (and also involves a pretence that the hearer’s dishevelled appearance is a deliberate hairstyle):

(8) You look perfect in your new hairstyle.

In all these cases, we claim that the irony is best analysed as involving the dissociative echoing of (possibly exaggerated) general desires or norms.

Broad though it is, this notion of echo provides the basis for a sharp contrast between standard desires and norms, which can always be echoed, and expectations and desires that go against these general standards, which can be echoed only if it is manifest to the speaker and hearer that they can be attributed to specific individuals. Thus, suppose John is playing in a tennis tournament; Peter expects him to be knocked out in the first round, and Bill is hoping that this will happen. If John makes it to the semi-final, Mary might say ironically to Bill and Peter:

(9) John was certainly knocked out early.

Here, she might be echoing Peter’s expectation, Bill’s hope, or both. In a situation where no one could be taken to have either expected or wanted John to be knocked out early, the utterance would fall flat.

The echoic theory of irony thus explains the striking fact – often noted but never explained – that irony tends to be ‘moral’: that it involves blame by apparent praise much more often than praise by apparent blame. It also explains why irony can occasionally take the form of apparent blame or criticism, as in (9), where specific desires or expectations, attributed to specific individuals, must be involved. If the notion of echo were more restrictively defined, the theory would fail to explain these aspects of irony.

Seto draws attention to an interesting further range of examples whose echoic status is in doubt. These involve relatively fixed expressions such as ‘a precious lot’ or ‘fat chance’ in (10) and (11) (or equivalents such as ‘C’est du beau’ in French):

(10) A precious lot you care about my wallflowers.

(11) Fat chance there is of Arsenal winning the Cup.

(10) and (11) would be ironical by standard definitions: for example, the speaker of (10) is accusing the hearer of not caring about her wallflowers rather than thanking him for caring. However, the ‘ironical’ interpretations have become grammaticalised to such an extent that it is hard to imagine these utterances communicating more regular ‘literal’ meanings.

Idiomatic ironies of this type show parallels with dead metaphors such as ‘deep difficulties’ in (12) and ‘back out’ in (13):

(12) The conservatives are in deep difficulties.

(13) The government backed out of its agreement.

A dead metaphor is defined in Preminger et. al. (1975: 184) as

“A metaphor which has been used so often in common parlance that its force as a figure of speech is no longer felt and which, therefore, is used as a literal expression.”

In more cognitive terms, dead metaphors have become associated with automatic interpretive routines which yield standard, though impoverished, interpretations. When routinely interpreted, they lose their potential for metaphor. However, as long as the original motivation remains transparent, their metaphorical potential may be revived by placing them in an appropriate context or subjecting them to conscious analysis; the revival of dead metaphors is sometimes seen as the special task of the poet.

We would suggest a parallel treatment for the idiomatic ironies in (10) and (11). What starts out as a genuine irony becomes associated with an automatic interpretive routine which assigns it a standard, though impoverished, interpretation. As a result, it loses both its original echoic status and its ironical force. However, as long as the original motivation remains tranparent, both may be revived in an appropriate context, or by conscious analysis. On this account, idiomatic ‘ironies’ such as (10) and (11) are not necessarily perceived as echoic, but a genuinely ironical reading is achievable only when an echo is perceived.

3. Are there borderline cases of irony?

Another important aspect of the echoic account of irony is its rejection of clear-cut boundaries between ironical and non-ironical utterances. Irony is an attitude to an echoed content. Since attitudes come in indefinitely many shades, we should expect there to be borderline cases which are neither clearly ironical nor clearly not, and also cases where the attitude expressed combines irony with other attitudes. We have illustrated this aspect of the theory with the case of Mark Antony repeating “Brutus is an honourable man”, each time with a more ironical attitude (Sperber & Wilson 1981, Sperber & Wilson 1990: 153-4). We are not sure that there is any significant disagreement about this example between us and Yamanashi, who discusses it very perceptively.

Yamanashi gives several examples of utterances such as (14), which “cannot be taken to be a hundred per cent ironic”:

(14) “You seem very clever at explaining words, Sir,” said Alice. “Would you kindly tell me the meaning of the poem called ‘Jabberwocky’?”

We agree that there is a touch of irony in Alice’s remark “You seem very clever at explaining words,” and that it is not a hundred per cent ironical. In fact there may well be a double echo in this single line. A page or so earlier in Lewis Carroll’s text, we read:

(15) Humpty Dumpty took the book, and looked at it carefully. “That seems to be done right,” he began.
“You’re holding it upside down!” Alice interrupted.
“To be sure, I was!” Humpty Dumpty said gaily, as she turned it round for him. “I thought it looked a little queer. As I was saying, that seems to be done right” (Lewis Carroll’s italics).

In (15), Humpty Dumpty uses ‘seem’ in a weak sense, where hardly any evidence is needed for something to seem to be the case. In (14), Alice can be read as echoing his use of the term and suggesting that she has no more evidence for her statement than he had for his. She is also clearly echoing the opinion that Humpty Dumpty has of himself. However, she seems to be more puzzled than actually convinced of Humpty Dumpty’s ineptitude. Her attitude can be understood as combining awe, bafflement, and a touch of irony.

Such cases should be an embarrassment for classical and Gricean approaches to irony: instead of a nice ‘meaning reversal’, the intended meaning seems to be somewhere up in the air, fluttering halfway between the literal and the standardly ironical. From the point of view of an echoic account, such examples are to be expected, and present no problems.

4. Some more complex cases

Given the variety of material that can be echoed _ general norms and desires, particular applications of these to specific cases, attributed past, present, or even future thoughts, actual or imagined utterances _, given that irony is on a continuum of attitudes to echoed material, given that irony can combine with emphasis, metaphor, pretence, parody etc., we should expect an indefinite variety of complex cases. Several such cases are discussed by Hamamoto. Some of them have been seen as raising problems for our approach. We would like to show how the echoic account can help with their analysis.

Here is an interesting example of what may look at first like a standard case of ironical blame. Kyoko learns that her husband Jiro has fiddled his travelling expenses and bought her a nice present. She says:

(16) You’re so naughty.

This is not like previous cases we have looked at, where someone was expected to do something wrong and did not, so that the expectation can be ironically echoed in order to praise or commend. The fact is that Jiro has broken a regulation about the use of travelling expenses. However, he has broken it to the benefit of Kyoko, who finds his behaviour worthy of praise rather than blame. She echoes what might be seen as a justifiable public criticism, while clearly dissociating herself from it. Her use of ‘naughty’, a word of mild reproach normally addressed to children, emphasises the fact that she sees no serious reason to blame Jiro, and that she has tender feelings towards him.

Another interesting example: Taro and his wife Hanako are environmental activists who work very hard on environmental issues, spending all their time away from home. Their son Jiro says:

(17) Our home is an environment

Hamamoto comments:

“Jiro’s remark is ironical but he does not seem to dissociate himself from the echoed opinion. What he does dissociate himself from is the situation where his house has been neglected and left messy by parents who in contrast care about the global environment problem.”

We would argue that a full analysis of this example reveals a subtle echoic dimension, and that this is the source of the irony. Jiro’s parents should recognise that their home is an environment (and that therefore they should care about it); however, they don’t. Jiro is echoing approvingly a thought that his parents should have. He is thereby dissociating himself, with a mixture of irony, reproach and regret, not from the thought directly echoed, but from the assumption implicit in this echo that his parents do have that thought.

Gibbs & O’Brien (1991), quoted by Hamamoto, give the example of a mother who sees her child’s room in a total mess and says:

(18) I love children who keep their rooms clean.

Here presumably (as in the preceding case), the speaker agrees with the literal meaning of her utterance, and would not want to dissociate herself from it. So where does the irony come from? We would argue that (18), literally understood, is inappropriate not because it is false, but because of the circumstances of utterance. What is being ironically echoed is the higher-order explicature (Wilson & Sperber 1993; Ifantidou-Trouki 1993; Ifantidou 1994) that (18), literally understood, is relevant in the circumstances. The circumstances should be such that the mother could relevantly say (18) without irony: that is, the room should be clean, and she should be able to praise her child sincerely. (For further discussion of these and similar cases from a variety of perspectives, see Martin 1992; Kreuz & Glucksberg 1989; Kumon-Nakamura & Glucksberg 1995; Perrin 1996; Cur‡o 1997).

5. Is irony a natural kind?

Seto believes that not all verbal irony is echoic. He therefore proposes a disjunctive definition: verbal irony is a type of implicit criticism involving either echoing or semantic reversal. If he is right, verbal irony has two distinct sub-cases involving two distinct interpretive mechanisms. It is not a natural kind.

Hamamoto and Yamanashi believe that not all irony is intentionally echoic. According to Hamamoto, a case of accidental (or situational) irony such as (19), which involves no deliberate echoing, should be grouped together with ostensively echoic cases:

(19) (Taro’s favourite food is Tira-misu, which was popular among young people and is now out of fashion. Hanako, not knowing it is his favourite, says to him:) There are still people who eat Tira-misu. Unbelievable, isn’t it?

Building on the work of Professor Kawakami, he suggests that both ostensive and situational irony should be dealt with by a single cognitive mechanism. If he is right, what we have been calling verbal irony is merely a sub-case of a more general phenomenon which has a better right to be called a natural kind.

We have argued against both these proposals. In section 2 above, we have defended the view that verbal irony as necessarily echoic. Elsewhere (e.g. Sperber & Wilson 1990) we have suggested that the echoic nature of verbal irony divides it from a range of non-echoic cases (situational irony, dramatic irony, romantic irony, irony of fate) which it resembles in some respects. We would like to end with some more general reflections on the status of irony as a natural kind.

The idea that there are two separate mechanisms involved in the interpretation of verbal irony is quite implausible. However, it has surfaced in many forms in the last 25 years. In early work (e.g. Cutler 1974), the semantic reversal mechanism was generally seen as basic, and ironical quotation (narrowly defined) was invoked merely to explain a small residue of troublesome cases. With the spread of the echoic account, these priorities have been reversed: now ironical echoing is generally seen as basic and semantic reversal as the minor but necessary addition. This seems to us as a step in the right direction, but one that does not go far enough.

In our own writings on irony, we have rejected the classical account for two separate reasons. On the one hand, we have tried to show that the semantic reversal mechanism was explanatorily inadequate; on the other, we have argued that an echoic account could deal with the full range of cases. In sections 2-4 above, we have continued this line of argument by showing how the echoic account deals quite naturally with a range of cases that are sometimes still seen as problematic. Here we would like to make a more general point. If our original arguments about the explanatory inadequacy of the semantic reversal mechanism had any value, it would be a mistake to assign this mechanism any role, however small.

Consider a standard irony such as (20), said by Mary to Peter in a downpour:

(20) It’s lovely weather.

On the two-mechanism approach, this utterance should have not only the range of echoic interpretations discussed above, but also a standard, non-echoic interpretation obtained by semantic reversal, meaning ‘It’s awful weather’. On this approach, it would then have to be explained how the practice of saying one thing and meaning the opposite could have arisen spontaneously in culture after culture; how children acquire it; how it could have become associated with the range of ‘ironical’ attitudes; why it lends itself to blame-by-praise more easily than praise-by-blame; how it fails to generalise beyond a few standard cases where, if no echoic interpretation is available, the irony falls flat. The fact that the semantic reversal mechanism is assigned only a limited role does not alter its explanatory inadequacy, and only increases the arguments for eliminating it entirely.

It is instructive to compare the two-mecbanism analysis of (20) with an analysis involving the notion of idiomatic irony proposed in section 2 above. If (20) were an idiomatic irony, it would have a standard interpretation of a type that might be assigned by the semantic reversal mechanism. There is a difference, though. On our account, even idiomatic ironies are echoic in origin. We are not proposing two distinct mechanisms; merely a single mechanism which (in a few stereotypical cases) may be bypassed by an automatic interpretive routine. This account is more plausible and more parsimonious than the two-mechanism approach, and should be preferred on general theoretical grounds.

According to Yamanashi and Hamamoto, an adequate account of irony should be more general still. In Yamanashi’s view,

“The existence of unintentional ironies …. indicates that there are more things in natural language than have been dreamt of in our semantic and pragmatic theory of figures of speech.”

Hamamoto proposes a general cognitive mechanism, based on the recognition of logical discrepancies between representations, which would apply equally to ostensive and unintentional cases such as (19) above. We would resist these proposals on two grounds, one involving the distinction between overt (ostensive) communication and other forms of information transmission, and the second the distinction between verbal irony and other types of irony.

The distinction between overt communication and other forms of information transmission is central to the Gricean pragmatic tradition to which we belong. Within this tradition, overt communication is seen as creating expectations not carried by other forms of information transmission, and involving separate principles and mechanisms (see for example Relevance, chapter 1). If this is right, then the search for a unified pragmatic theory which would cover both intentional and unintentional communication is a mistake. The issue is quite general and extends beyond the analysis of irony. It implies that ostensive and non-ostensive forms of irony involve different cognitive mechanisms and should not be treated together.

We would agree with Hamamoto (and have argued in Sperber & Wilson 1990: 152-3) that there is a common theme to all types of irony, ostensive and non-ostensive (situational, dramatic, romantic, Socratic, etc.): they all involve the perception of a discrepancy between a representation and the state of affairs it purports to represent. However, we claim that this is not enough to explain how verbal irony works. If it were, the classical account of irony, with its semantic reversal mechanism, would not fail in the ways it does. Ostensively ironical utterances exploit an echoic interpretive mechanism which is not involved in non-ostensive forms of irony. Although verbal irony has no clear-cut boundaries (because ironical attitudes shade off into others), the echoic mechanism defines a class of utterances which is a natural kind.

Let us end with one further example which fits well with the echoic account. It was used by Stephen Levinson in a discussion of irony in his textbook on pragmatics (Levinson 1983: 109):

(21) A: What if the USSR blockades the Gulf and all the oil?
…..B: Oh come now, Britain rules the seas.

(21B) is clearly echoic: it echoes the first line of Rule Britannia and a thousand Churchillian speeches, and conveys how hollow the sentiments behind them have now become. Levinson does not mention this. He uses (21B) to introduce a standard Gricean account of irony, a pragmatic variant of the traditional semantic reversal account. Yet the type of echoing involved in (21B) is particularly blatant. The fact that it could have been overlooked by an analyst as subtle and sophisticated as Levinson encourages us to persist in our view that other, less obvious types of echoing are still being overlooked.


Curco, C. 1997 The pragmatics of humorous interpretations: A relevance-theoretic approach. University of London PhD thesis.

Cutler, A. 1974 On saying what you mean without meaning what you say. Chicago Linguistic Society 10: 117-27.

Gibbs, R. and J. O’Brien 1991 Psychological aspects of irony understanding. Journal of pragmatics 16: 523-30.

Hamamoto, H. 1997 Irony from a cognitive perspective. (This volume.)

Ifantidou, E. 1994 Evidentials and relevance. University of London PhD thesis.

Ifantidou-Trouki, E. 1993 Sentential adverbs and relevance. Lingua 90: 65-90.

Kreuz, R. & Glucksberg, S. 1989 How to be sarcastic: The echoic reminder theory of verbal irony. Journal of experimental psychology: General 118: 374-86.

Kumon-Nakamura, S., Glucksberg, S. & M. Brown 1995. How about another piece of the pie? The allusional pretense theory of discourse irony. Journal of experimental psychology: General 124: 3-21.

Levinson, S. 1983 Pragmatics. Cambridge University Press, Cambridge.

Martin, R. 1992 Irony and universe of belief. Lingua 87: 77-90.

Perrin, L. 1996 L’ironie mise en trope: Du sens littéral des énoncés hyperboliques et ironiques. Editions Kimé, Paris.

Preminger, A. 1975 Princeton encyclopedia of poetry and poetics. Macmillan, London.

Seto, K. 1997 On non-echoic irony. (This volume.)

Yamanashi, M. 1997 Some issues in the treatment of irony and related tropes. (This volume.)

Sperber, D. & D. Wilson 1981 Irony and the use-mention distinction. In P. Cole (ed.) Radical pragmatics. Academic Press, New York: 295-318. Reprinted in S. Davis (ed.) 1991 Pragmatics: A reader. Oxford University Press, Oxford: 550-63.

Sperber, D. & D. Wilson 1990 Rhetoric and relevance. In J. Bender & D. Wellbery (eds) The ends of rhetoric: History, theory, practice. Stanford University Press, Stanford CA: 140-56.

Wilson, D. & D. Sperber 1993 Linguistic form and relevance. Lingua 90: 1-25.