Printer Friendly


1. Expressives

Expressives are lexical items which encode attitudes. Mary says: "That bastard John stole my pen." According to many semantic theories for expressives (1) the way to understand such sentences is by isolating two distinct levels of information (2) encoded by the utterance, for instance:

(1) John stole Mary's pen.

(2) Mary has a negative attitude towards John.

It seems that (1) captures the truth-conditional, or propositional, content while (2) goes some way towards capturing whatever meaning is introduced by the expressive "bastard." On the face of things this is an attitude held towards John by the speaker. Similar locutions can express other attitudes. E.g. (among others): I. Expressive Interjections: for instance: "ouch!," "yippee!," "damn," and "yuck." II. Epithets: for instance: "my dear," and "your honour," derogatory epithets such as "idiot," and racial pejoratives.

It is tempting to suspect that this speaker dependence (SD) is a universal feature of expressives:

(SD) Expressive content is always evaluated from the perspective of the speaker. It is the speaker's attitudes that the expressive locution conveys. (3)

Hence, according to SD, whenever one comes across a term such as "bastard" it is a feature of the content of these expressions that they can only encode the attitudes of speakers. As such, (SD) predicts that expressives will typically resist embedding under propositional attitude reports, verbs of saying, etc. (cf. Kaplan, 1999: 8): that is to say, the expressive content of an expressive always occupies wide scope position with respect to propositional attitude operators. For example, John's utterance of the sentence "Jane believes that bastard James is a decent man" expresses John's disdain for James, not Jane's.

Upon further reflection, however, there seem to be counterexamples to SD. Some have already been mentioned in the literature. Additionally, in what follows we shall offer two more. From such data, Potts (2007: 166) replaces SD with PD (Perspective Dependence):

(PD) Expressive content is evaluated from a particular perspective. In general, the perspective is the speaker's, but there can be deviations if conditions are right.

PD is captured in Potts's semantic theory by the addition of a judge parameter, as first introduced by Lasersohn's (2005) semantics for predicates of personal taste. What PD facilitates is the use of an expressive to encode the attitude of the judge, rather than the speaker. The default setting for the judge parameter is that it takes the same value as the agent or speaker parameter but, under special circumstances, these two parametric settings can differ, allowing speakers to express the attitudes of perspectives distinct from their own. However, there is an important difference in the way that Potts makes use of the notion of a judge compared to the treatment offered by Lasersohn. Lasersohn appeals to a judge parameter in order to defend the thesis that the very same proposition can have different truth-values relative to distinct judges. Hence the judge parameter is not an element of the context of utterance in Lasersohn's theory, as this would make the content of an utterance dependent on a judge, rather than the truth-value that the proposition has. Thus judges are parameters of the circumstance of evaluation according to which propositions are evaluated as true or false. For Potts, however, judges are elements of context and the proposition expressed is dependent on the value of the judge parameter. The addition of the judge into the context therefore allows a shift in the perspective encoded by the expressive away from the speaker, hence the theoretical shift from SD to PD.

Despite the proposed counter-examples, we do not think SD should be rejected. In this paper we argue for two things:

(a) A rejection of Potts' proposed solution to the counterexamples. For while we accept that this solution may work for CE1-3, it will not work for CE4-5, and in fact makes the wrong predictions about expressive content in these cases.

(b) A denial that CE1-5 really are counterexamples to SD. We argue that the troublesome status of CE1-5 is in fact consistent with SD, and can be dealt with without the complications that Potts envisages.

We thus defend the status of SD as a universal semantic principle for expressives, offering an analysis of apparent deviations from the principle that is able to preserve a greater degree of simplicity than that found in its rivals. (4)

2. Counterexamples to SD

SD is a semantic claim about expressives. More precisely, it is a claim about their character, where "character" is understood, roughly, as a rule for the use or interpretation of a given locution in a context. (5) SD makes it impossible to force an expressive to encode an attitude that the speaker does not hold (or at least purport to hold by the way they speak) without recourse to metalinguistic operations such as quotation.

Understood in this light, SD predicts that the only cases where one who utters an expressive locution does not thereby express their own attitude are quotational contexts, or can plausibly be analysed as such. (6) SD, then, proposes a direct analogy with indexical expressions such as "I," at least as according to their behaviour in English. Kaplan (1977: 510-511) argues that the only environments where the referent of "I" shifts to someone other than the speaker are quotational contexts:

(3) (a) Otto said that I am a fool. ("I" denotes speaker)

(b) Otto said "I am a fool." ("I" denotes Otto)

The same also holds, if SD is accepted, for expressives:

(4) (a) Mary said that that bastard John stole my pen. ("bastard" encodes speaker attitude)

(b) Mary said "That bastard John stole my pen." ("bastard" encodes Mary's attitude)

Recent literature, however, has proposed a number of counterexamples to this view. We consider three such examples: (7)

First Counterexample: Kratzer (1999)

(CE1) My father screamed that he would never allow me to marry that bastard Webster.

Intuitively, it would be strange for a prospective spouse to denote their future husband by the expression "that bastard." Hence, the thought goes, whereas "me" is indeed indexed to the speaker, it must be that "that bastard" does not in fact encode the speaker's attitudes at all. Rather, it seems that "that bastard" encodes the attitude of the father.

Second Counterexample: Schlenker (2003)

(CE2a) *I am not prejudiced against Caucasians. But if I were, you would be the worst honky I know.

(CE2b) I am not prejudiced against Caucasians. But John, who is, thinks/claims that you are the worst honky he knows.

According to SD, racial pejoratives, qua expressives, should not embed under any operation short of quotation. Yet in (CE2b) this is precisely what appears to be the case. For, if "honky" scoped out from the propositional attitude verb "thinks/claims" the result should be a contradiction as in (CE2a), where the speaker simultaneously disavows a negative attitude towards Caucasians and uses an expression which is felicitously used if and only if they have just such an attitude. Thus, the thought goes, "honky" must connote John's attitude, and not the speaker's, in direct violation of SD.

Third Counterexample: Potts (2007)

(CE3) A CPJ report on Venezuela tells us how problems have "escalated" in Venezuela under Chavez, i.e. the physical attacks against journalists under previous presidents have "escalated" to Chavez calling the opposition, which includes the media, names. This is very, very serious, but I don't think another coup attempt is called for until Chavez resorts to dramatic irony or sarcasm. But if that vicious bastard uses litotes, then there's no other rational choice than an immediate invasion.

Pace SD, the context of (CE3) is such that it is incorrect to think of "that vicious bastard" as connoting the weblog writer's attitude. Rather it seems to be the attitude of someone else. Commenting on this passage, Potts (2007: 175) says that the weblog writer's: "... general level of sarcasm is sufficiently high to shift the content of that vicious bastard away from her and onto her opponents (the authors of the CPJ report, of which she is sceptical)."

The above counterexamples are already familiar from the literature on expressives. Nevertheless we shall argue in what follows that they are open to certain obvious replies, and in this respect are deficient. To compensate, we offer two additional counterexamples of our own, which we do not think are so easily diverted:

Fourth Counterexample: The following passage is taken from Terry Pratchett's Making Money (2007: 307):

(CE4) She had the slightly wistful, slightly hungry look that so many women of a certain age wore when they'd decided to trust in Gods because of the absolute impossibility of continuing to trust in men. [...]
'What is your name, shister?' he asked.
'Berenice,' she said 'Berenice, er, Houser.'
Ah, no longer using the bastard's name, very wise, thought Cribbins.

Cribbins does not know Berenice's ex-husband, but has deduced from her manner that she is a divorcee. In this case, one might have expected his use of "the bastard" to be infelicitous, or at least inappropriate: for, were it really the case that "Bastard" could only express speaker attitudes, Cribbins should not be able to express a negative attitude towards someone he knows nothing about. But, rather, it seems that his use is felicitious precisely because SD fails: the attitude expressed by his use of "the bastard" is attributed to Berenice.

Fifth Counterexample

The final counterexample envisages a conversation between two individuals. Call them A and B:

(CE5) A. I just had to spend [pounds sterling]500 on a new clutch for my car

B. Ouch!

Assume that B is not subject to any unmentioned source of pain, discomfort, etc., during this dialogue. Given SD, her utterance should be infelicitous. But such dialogues are commonplace. B seems, contrary to SD, to use "Ouch!" to convey a state or attitude of A's. (8)

For SD to be acceptable it must be able to provide a solution to all of the above counterexamples CE1-5. Of course, the same requirement applies to alternative to SD, such as PD put forward by Potts. In the next section we describe this latter approach in more detail. We shall try to show that this proposal, while perhaps accounting for CE1-3 (though with reservations), nevertheless fails for CE4-5. By contrast, we shall try to show that all five counterexamples can be explained while holding SD intact.

3. Responses

If these counterexamples are genuine then they need a solution. One influential proposal, offered by Potts, proceeds by weakening SD, thereby generating a new principle he dubs perspective dependence (PD). Recall the formulation from Potts (2007: 166) given above:
Perspective dependence: Expressive content is evaluated from a
particular perspective. In general, the perspective is the speaker's,
but there can be deviations if conditions are right.

To see why PD is weaker than SD, consider any of the counterexamples we have outlined above. Unlike SD, there is no longer any restriction that the expressed attitude need be the speakers. Therefore the usages introduced in these counterexamples, where the speaker appears to be conveying another person's attitude and not their own, is accounted for by saying that the perspective has shifted away from the speaker--the person that we would normally expect to be the holder of the attitude expressed--to some other individual recoverable from the context.

To effect this move, Potts (2007) draws on earlier work in relativist semantics, and in particular the work of Peter Lasersohn.9 Potts's proposal is that expressive utterances should be modelled by an ordered pair <s,c> of the sentence uttered s with a context of utterance c, which is a series of parameters c = <[a.sub.c], [l.sub.c], [t.sub.c], [w.sub.c], [j.sub.c], [[epsilon].sub.c]>, such that [a.sub.c] is the agent of c, [l.sub.c] is the location of c, [t.sub.c] is the time of c, and [w.sub.c] is the world of c. So far, this is familiar to students of indexical semantics (e.g. Kaplan, 1977). (10) The innovation comes from the addition of a judge parameter [j.sub.c] and expressive index [[epsilon].sub.c]. It is [j.sub.c] which motivates the move from SD to PD. The addition of this parameter allows the content of elements within the utterance to be determined by the attitude of a judge distinct from the agent (speaker) of the utterance.

Originally Lasersohn (2005) proposed the introduction of a judge parameter into the sequence of parameters that propositions are semantically evaluated with respect to in order to explain the apparent relativity of truth-value that he argues are found in judgements of taste. This relativity is made evident, Lasersohn argues, by so-called faultless disagreement--a phrase used to describe our intuition that there is no objective matter of fact which can decide disagreements in personal taste. Following Kaplan (1989), it is common in formal semantics to distinguish two levels of indexing: the context of utterance is a sequence of parameters which must be determined before a sentence can express a content. For example, indexical expressions require these parameters. Once these parameters have been fixed, we can evaluate the resulting proposition relative a further sequence of parameters, the circumstance of evaluation. Minimally, this sequence will contain a world and a time, but there is no formal obstacle to extending the sequence. Lasersohn's proposal is that faultless disagreement is most naturally explained by adding a judge parameter to the circumstance of evaluation. Consider two conversations between A and B:

(5) A. Bill missed his seminar presentation this morning.

B. Bill didn't miss his seminar presentation this morning.

(6) A. Presenting at seminars is fun.

B. Presenting at seminars isn't fun.

It would appear that there is a key difference between conversation (6) and conversation (5). Assuming that both utterances are paired to contexts which take the same value for "this morning," it does not seem possible for both (5A) and (5B) to be true. Either A or B must be factually mistaken. The same thing does not go for (6) however; because it is a subjective matter whether presenting at seminars is fun, (6A) and (6B) can be both true. A and B are disagreeing but neither is at fault: (11) B is able to contradict A in (6) without us having to infer that either A's statement or B's statement is false. This is explained within Lasersohn's semantic theory, because the two assertions are each evaluated with respect to a different judge (namely each speaker is the judge of her own utterance in this scenario). Potts adopts the notion of a judge parameter to account for the failure of SD in the above kinds of cases by taking them to be cases were the expressives encode the attitude of a judge who is distinct from the agent of the utterance. Furthermore, he appeals to Lasersohn's analysis of judgements of taste (Potts, 2007: 175) to argue that the inclusion of the judge parameter is independently motivated, hence not an ad hoc fix to the counter-examples. As noted above, however, Potts can only achieve the desired result here by making an important departure from Lasersohn's theory. Judges do not influence the content of utterances on Lasersohn's theory, they simply mediate over the correct evaluation of the content. This is indeed crucial to the notion of faultless disagreement: if different judges result in A and B meaning different things in the above examples, then there is no genuine disagreement between them and, a fortiori, no faultless disagreement between them--they are simply "talking past one another," B negating a proposition which turns out to be distinct from that which A asserts. The claim that the addition of judge parameters within the context of utterance is independently motivated by Lasersohn's account, therefore, is incorrect. Lasersohn's theory may provide independent reason for the notion of a judge parameter as a semantic concept, but it gives no independent reason for thinking that the concept is required for fixing the content of an utterance, which is what it is being invoked to do on Potts's account.

We think that the appeal to a contextual judge to explain the counter-examples is problematic for a number of further reasons even if one is willing to overlook the distinction between context and circumstance. CE1-CE3 do seem to involve a shift from the agent perspective to the perspective of a salient other as the holder of the attitude relevant to the use of the expressive (although we will shortly argue that this appearance is deceptive). However, there seems to be more than this at play in CE4 and CE5. A notable feature of these two cases is the "empathetic" quality of the speaker's use of the expressive to show some form of emotional or evaluative affinity with another salient individual. We suggest that the use of the expressive in these cases is felicitous only if the agent possesses a certain attitude towards this salient third party. Thus, rather than contradicting SD, they both appear to make the attitude of the speaker fundamental to the interpretation of the utterance (albeit in a different way to the standard, non-empathetic, uses of the expressives). Taking as an example the case of CE5: just as an appropriate or genuine use of "ouch" in standard--i.e. non-empathetic--uses requires the agent to experience some kind of pain or discomfort (see Kaplan, 2005) so an empathetic use of "ouch" requires the agent to hold the correct attitude (that of empathy) towards the object of their empathy if it is to be genuine. But it is in this respect that we see trouble for an account which abandons SD in favour of PD: we do not see how PD will be able to properly account for this attitudinal component in empathetic uses of expressives.

To see why we think an appropriate agential attitude is needed, rather than a contextually supplied judge, suppose that the judge of CE5 is indeed distinct from the agent. But then on the suggested analysis B's utterance of "ouch" could be appropriate in a case where B is glad that A has suffered the financial misfortune in question, or in a case where B deems the expense in question to be quite minor and an appropriate outlay for the new clutch, so long as the contextually determined judge herself has the attitude that she has suffered a financial misfortune and is upset about it. We find this counter to our intuitions. Secondly, note that B's utterance of "ouch" can be appropriate in cases where A does not hold the attitude standardly encoded by the term. Consider the following exchange:

A: I have just had to pay [pounds sterling]500 for a new clutch.

B: Ouch!

A: It's OK, I won [pounds sterling]100,000 on the lottery last week.

B: I still think [pounds sterling]500 is extortionate.

Assuming once again that B is not in any pain of any kind, B's first utterance appears to be an empathetic use of the expressive "ouch." But then, if it is an empathetic use, by PD's lights, the appropriate judge here ought to be A. On the other hand, B's utterance seems to be appropriate here, not because of any attitude that A holds, but because B thinks that [pounds sterling]500 is a lot of money to spend on a clutch. Taking A to be the parametric setting for the judge part of the context will not secure this result.

It is worth noting that there are important restraints placed on situations in which empathetic uses are felicitous that demand the preservation of SD and therefore are not explained by the relativist account. Consider the case where A is torturing B: if A inflicts pain on B and then says "ouch," this would be a bizarre and, we think, infelicitous utterance on the part of A. (12) Similarly, if A deliberately trips B over and then says "oops," there are a number of possible interpretations of the situation, but none of them are really interpretations on which A is expressing B's misfortune. It might, for example, be the case that A is expressing their own clumsiness in which case this is not an empathetic use at all, and certainly not a violation of SD. Or it could be that A is attempting to disguise their malevolence towards B by pretending the deliberate action was accidental, in which case this is a mock empathetic use, expressing A's mock empathy towards B. Similarly, perhaps A's utterance could be sarcastic but, as we argued in relation to CE3 above, sarcasm is best construed as a pragmatically introduced phenomenon, not a semantically encoded one. The restriction on empathetic uses is thus very clear--an empathetic use of an expressive is appropriate when and only when the speaker wishes to express a certain attitude that they have towards a contextually salient individual's situation. This is why we think a correct account of empathetic uses of expressives should preserve SD, as the speaker's perspective plays an ineliminable semantic role in such uses.

Our claim then is that PD makes the wrong predictions in the case of CE4-5. Now consider CE1-3: while PD does not seemingly make false predictions here we remain unconvinced that they are genuine counter-examples to SD at all, for reasons we shall now elucidate.

Let us start with CE1. In our view, this is simply a case of a metalinguistic operation on the expressive. Although there is no explicit quotation, the verb of saying "screamed" serves to generate what we will call a quasi-quotational environment. A quasi-quotational environment is contextually situated expression where the context generates a natural reading of the expression as quotational despite the absence of explicit quotation. In CE1, such an environment is generated by "screamed" in which the expressive is (with some pragmatic assistance from the context--namely the contextual obstacles blocking an interpretation of the expressive as transparently used by the speaker as they clearly don't think Webster is a bastard) construed as quoted. This is evident from the fact that the truth of the utterance here is dependent on the exact form of words used by the speaker's father: if he called Webster an "idiot," then the utterance will be false. Further confirming evidence of this hypothesis can be found by considering cases where the quasiquotational environment is a cross-linguistic one. Consider Mary's report to her English-speaking friends of her French speaking Father's objection to her proposed marriage to Webster:

(7) My Father screamed that he would never let me marry that connard Webster.

The truth of this utterance again depends on the particular pejorative used by Mary's father. The fact that we can insert the French pejorative into the context of an English sentence is evidence of a quotational context, on a par with other cases of "mixed quotation" widely discussed in the literature (see e.g. Cappelen and Lepore, 1997; Maier, 2016). Mary cannot, for example, report her father's outburst by using the Chinese expression "hundan" in the same context, despite the fact that her English-speaking audience have no less an understanding of Chinese than French (the "?" is used to mark this kind of infelicity within a context):

(7a)? My Father screamed that he would never let me marry that hundan Brian.

CE2 seems to us to be the weakest objection of all. For one thing, it seems highly dubious that (5) is felicitous at all. We suspect that the apparent felicity of the example is a function of the relatively mild racial pejorative used in the example. Examples using more politically explosive terms are likely to provoke somewhat different reactions. Again, we suspect that the reading CE2 is intended to highlight is a quasi-quotational one.

Similar arguments to those we considered regarding CE1 seem to apply here, but not quite so straightforwardly: although John need not have used the expression in question to talk about the addressee of the utterance, it has to be the sort of expression that John would use. For example, it is not felicitous to substitute for "honkey" a term that John is unfamiliar with. The term has to be one that there is a good justification for thinking John would himself use. Again this is strongly suggestive of a metalinguistic operation of the sort evident in CE1. (13)

We think, therefore, that there are only two ways to respond to counterexample 2: either it is not felicitous at all; or, if it is felicitous, it is also a quasiquotational case.

This leaves us with CE3. We find it puzzling that Potts himself recognizes this use of the expressive as coming about via sarcasm, yet still holds that accounting for such cases requires us to modify the semantics of expressives. By contrast, we maintain that CE3 is very straightforwardly a case of sarcasm, and moreover can be dealt with using purely metalinguistic operations. For, just as with CE2, there are constraints on the terms that may be used in cases of sarcasm, as they have to echo the sorts of terms that we envisage the target speaker employing. However, there are differences. In particular, sarcasm often involves an element of hyperbole which allows us to exaggerate the echoic quasi-quotation to the level of a kind of mock-quotation: (14) sarcasm allows us to mock the speech of others and this means that we are not limited by the same constrains as we found in the quasi-quotational operations evident in CE1 and CE2. For example it may be that the authors of the CJP report would not use the expression "vicious bastard" but it is nonetheless licensed in the sarcastic context. The bottom line here, though, is that it is uncontroversial that sarcasm is best explained by pragmatic processes, not semantic ones. (15)

We have argued in this section firstly that some of the counter-examples to SD cannot be fully explained by the appeal to a judge parameter, and secondly, that those which can need not be, because there is no real reason to think they are genuine counter-examples to SD after all. This leaves us requiring an explanation of what is happening in those cases (namely CE4-5) that we so far have provided no solution for. We turn to that issue in the next section.

4. Empathetic Expressives

In the previous section we have established two things: (1) that the original motivation for PD (specifically CE1-3) is not in fact a counterexample to SD at all; (2) further examples that do apparently challenge SD (namely CE4 and CE5) cannot be explained by appeal to PD. In this section we will argue that CE1-5 can all be explained without rejection of SD. To do this, we will now show that CE4 and CE5 do not refute SD.

Our strategy for dealing with CE1-CE3 was to interpret them as subject to pragmatically introduced metalinguistic operations. All of them were argued to be quasi-quotational (albeit with the added complication of sarcasm-induced effects in CE3). Our suggestion is that a similar approach will explain CE4-CE5 also. However, there are important differences between CE1-CE3 and CE4-CE5 that must be taken into account and which require a more sophisticated analysis. No verbs of saying are present in these examples nor can they be plausibly paraphrased in terms of such constructions. This much is evidenced by the fact that the properties highlighted in the other examples suggesting the presence of an implicit quotation operation are absent in these cases. Consider CE5: B's response to A's reported outlay for the new clutch in CE5 is not subject to the same kinds of restriction on the appropriate expressions employed that we saw with CE1 and CE2. Indeed, it is perfectly appropriate for B to use an expression unfamiliar to A, or even one wholly absent from As idiolect. If A is a monolingual English speaker, and B is a native Mandarin Chinese speaker who speaks English as a second language, then B may appropriately say "ei yo!" in response to a report of the financial transaction given by A in English. The fact that A does not speak Chinese, and therefore does not understand B's utterance, does not make B's interjection infelicitous.

What this difference reveals, however, is not the absence of a similar metalinguistic operator. It simply shows that the form of quotation differs. The key difference between CE1-CE3 and examples like out CE4 and CE5 is that the former, but not the latter, are all reports of another agent/speaker's attitude. CE4 and CE5 are what we call "empathetic" expressives. They do not report or encode the attitudes of others (as is the case in CE1-CE3). Rather they express the speaker's empathy with the situation of others.

Our suggestion is that all the cases, despite these differences, are best understood as resulting from a metalinguistic operation whereby the expressive is interpreted as implicitly quotational. However, there are two important aspects to this operation that must be recognized. Firstly, the type of quotation on display here is what is often referred to as open quotation. Open quotation is a mixed form of quotation which is a hybrid of use and mention. It mentions material from another context (standardly a reported context), but also endorses the content of the quotation, thus using it as well as mentioning it. Mixed forms of quotation have been noted often. A famous example from Davidon (1979) is:

(8) Quine said that quotation "has a certain anomalous feature."

Here, the speaker does not just report (mention) what Quine said, but also uses the very words Quine said to communicate the information that quotation has a certain anomalous feature.

As Recanati (2010) explores in some detail, an important, but often neglected, feature of open quotation is that it is used routinely to effect a form of linguistic mimicry. We quote the exact words of someone when reporting their speech in order to enter into a form of pretence where we perform an (often mocking) impression of the reportee:

(9) Now it seems that Donald Trump is the best of friends with Little Rocket Man.

In (9) there are enough contextual cues to make it clear that the expression "Little Rocket Man" is mimicking Trump (the main contextual cue that the interpretation relies on being the shared knowledge that "Little Rocket Man" is Trump's nickname for Kim Jong-Un. One who lacks this information cannot interpret the utterance as intended). Evidently, the expression is best interpreted as implicitly quotational. Making the mimicry of Trump explicit requires us to make the quotation explicit:

(9*) Now it seems that Donald Trump is the best of friends with "Little Rocket Man."

In a sense, there is a context-shifting operation at work here. However, there is no need to postulate any semantic operator to carry the burden of that operation. (16) Once we recognize that the best available interpretation of the utterance is implicitly quotational it's context-shifting properties are just what we ought to predict, as we saw in 3(b) above. The heavy context dependence of the interpretation, we suggest, is an indication that it is pragmatically licensed, rather than the product of a semantic operator. The same sort of contextual cues that shift the nickname in (9*) can shift expressives too. Consider the case of racial pejoratives. These take wide scope readings by default and there are no semantic operators that can force them into narrow scope. If a speaker offers an unquoted attitude report of another, using a racial pejorative, the racism expressed by the pejorative incriminates the speaker, not the person whose speech is reported. Reporting someone's speech using direct quotation, however, forces the content--including the expressive content of the pejorative--into narrow scope:

(10a) Mary believes that there are too many Honkies living around here.

(10b) Mary said "there are too many Honkies living around here."

(10c) Mary believes there are too many "Honkies" living around here.

In (10a) the expressive content of "Honkies" projects into wide scope as expected.

In (10b) the quotation marks force the expressive content into narrow scope. In (10c) we have a mimicking case of quotation (this might be accompanied by explicit contextual cues such as mimicking Mary's voice when producing the word "Honkies" in an utterance of (10c)). Other contextual cues can make a reading like (10c) the only available interpretation of an utterance which superficially seems like (10a). Suppose that the speaker of (10a) is white, and that Mary is Black. Now it becomes clear that the intended interpretation of (10a) is (10c) and, similarly to (9), the availability of this interpretation requires knowledge of these facts on the part of the interpreter. (17)

Once we recognise these mimicking uses of quotation, we can make the further recognition that mimicry is not limited to reporting. We can also mimic the behaviour that we take ourselves to be liable to perform in hypothesised scenarios. This is what we propose is happening in cases of empathetic uses of like CE4 and CE5. Uses of empathetic expressives are also acts of mimicry but, rather than mimicking the speech of someone other than themselves, they are cases where speakers mimic the speech that they would utter if they were in the situation of their interlocutor (or whichever individual is made salient by the context).

Understood this way, we do not have a genuine departure from SD. The empathetic use of the expressive is understood as a form of open quotation but the quotation is not a report, it is simply a quote of what the speaker takes herself to appropriately utter in the counterfactual situation where she experiences what the person she empathizes with is actually experiencing. Effectively, this means that empathetic uses are generated by a metalinguistic operation which shifts the context that the expressive must be interpreted with respect to away from the actual context to the counterfactual one. But no change is made to the speaker-dependent character of the expressive. The character of "ouch," for example, dictates that the expression is to be used in situations where the agent is experiencing pain or discomfort. In our scenario, B is not in any kind of pain or discomfort, yet apparently can felicitously use the expressive "ouch" in response to A's pain or discomfort. Rather than construing this as a case where B expresses A's attitude, our analysis takes the utterance to express A's attitude of empathy--A is mimicking exactly what she would say without pretense were she in the situation that B is in. Empathetic uses, in short, are context-shifting operators, where the speaker mimics the discomfort they would experience were they in a situation where they experience the misfortune in question. With such uses, speakers put themselves "in the shoes" of the targets of their empathy.

The analysis that we propose not only preserves SD, thereby simplifying the semantics of expressives by removing the need for a judge parameter within the context of utterance, (18) it also manages to do so in a way which gives a uniform analysis of the apparent counter-examples to SD which we have discussed in this paper. All of them, on our proposal, are heavily context-dependent and, hence, best explained as subject to a contextually cued quotational interpretation that shifts the relevant elements of the utterances. (19,20)

Author Contributions

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.


(1.) See e.g. Bach (1999), Gutzmann (2015), Kaplan (1999, 2005), Kratzer (1999), Potts (2005, 2007), Simons et al. (2010).

(2.) Following Potts (2005) it is common to describe the two levels as the "at-issue" and "not-at-issue" content. The at-issue content is the truth-conditional content (what is said); the not-at-issue content may include presuppositional content, conventional implicatures, and so forth.

(3.) Potts (2005: 156) adopts a version of SD, citing with approval Cruse's (1986: 272) claim that a "characteristic distinguishing expressive meaning... is that it is valid only for the utterer, at the time and place of utterance. This limitation it shares with, for instance, a smile, a frown, a gesture of impatience...." Note, however, that Potts later goes on in his (2007) to reject SD, for reasons that we will discuss presently. Kaplan (2005: 6) also endorses a version of SD when he says: "the sentence that damn Kaplan was promoted is going to be expressively correct just in case the speaker has a derogatory attitude toward Kaplan."

(4.) Harris and Potts (2009) present experimental data which they argue also supports the phenomenon of perspective-shifting to non-speaker-oriented readings for expressives. Interestingly, however, they take the heavy burden placed on contextual cues, and the apparent availability of such readings in non-attitude-ascribing contexts, to count against a semantic explanation and favour a pragmatic explanation of the phenomenon.

(5.) More formally, characters are functions from contexts to contents which, in the case of indexical expressions, will be non-constant functions (see Kaplan, 1977). Note that, as applied to expressives, this is a more specific use of the word "semantics" than is occasionally used in the philosophy of language, where "semantics" is thought to be (near-) equivalent to "truth-conditional content" (the latter, for instance, seems to be the sense of "semantics" as it is used in Predelli (2005), whereas his use of the term in his (2013) seems much closer to ours).

(6.) The claim that expressive content cannot be embedded in non-quotational environments has been widely held in the past: see e.g. Cruse (1986), Quang (1971), Soames (2002), Zimmerman (1991).

(7.) In what follows, all underlined content is our emphasis.

(8.) A possible response to CE5 might be to challenge the assumption that the correct semantics for "ouch" interprets it as a subjective expressive that expresses the speaker's distress, discomfort, etc. In this case the example would not be so much a counter-example to SD as it is to the proposed character of "ouch." Our own view, however, is that the subjective analysis for "ouch" is correct. The core uses of the term (e.g. speaker's immediate response to painful stimuli) seem clearly subjective in this sense. Accepting departures from this norm would therefore be, at best, an indication of an ambiguity in the term. Our proposal, below, accounts for the cases like CE5 without postulating such ambiguity hence we see no need to challenge the subjective analysis.

(9.) Thus see esp. Lasersohn (2005). John MacFarlane has also contributed extensively to this area: see esp. MacFarlane (2005a,b). Note that Lasersohn has developed his own approach to expressive content in his (2007), and has developed a far more comprehensive statement of his position in his (2017).

(10.) Other parameters for demonstrata, addressees, etc. may need to be added for some indexical utterances but we have kept things minimal for the sake of simplicity here. It seems uncontroversial to assume that the general semantic framework outlined here applies to expressives, and we will assume as much in what follows. However, as will become clear, we do not accept the controversial part of the Pottsian proposal, namely that a judge parameter jc is required in the model.

(11.) See Kolbel (2004) and Lasersohn (2005, 2017) for further elucidation of this intuition.

(12.) There are, of course, contexts where such an utterance may be pragmatically licensed, despite its semantic infelicity.

(13.) We are not alone in suspecting that something metalinguistic is at work here. See Predelli (2013: 106-107) and Anand (2007: 205) for similar responses to such cases.

(14.) For more discussion of these and related issues see Recanati (2009: 226-228).

(15.) Note that in this respect, we follow Grice (1989: 32) in thinking of irony and sarcasm as a case of conversational implicature, and hence as possessing a level of meaning distinct from the conventional meanings a sentence expresses (see also Bach, 1994: 140; Huang, 2007: 34). Potts (2005: 23) also seems to endorse this picture, whereby conversational implicatures (except, perhaps, for generalized conversational implicatures which in any case are not of the same ilk as irony or sarcasm--see Levinson, 2000) are not encoded in the character of a locution. Hence we see no reason to think that a case of sarcasm requires us to modify our account of the semantics of expressive content.

(16.) Maier (2016) provides a compelling case against treatments of mixed quotation as semantic operations.

(17.) Thanks to Andreas de Jong for pointing out these kinds of cases.

(18.) Of course, there are independent reasons for thinking that judge parameters may be required in the circumstance of evaluation if one agrees with relativists like Lasersohn that there are genuinely relative truths. Our arguments here remain neutral on that point: our claim is simply that judges are redundant in the context of utterance as determinants of content. See Lasersohn (2017) for discussion of the difference between locating judges in contexts of utterance versus circumstances of evaluation.

(19.) A fundamental feature of apparently shifted expressives that has motivated our analysis is that they always rely heavily on contextual cues. This is widely recognized by theories which build shiftability into the semantics of expressives by the addition of a judge. For example, interesting data regarding so-called "modal particles" in German (which are often assimilated to expressives) appears to show that there are differences between them with regard to their shiftability, with some being apparently shiftable but only in response to explicit contextual cues, while others can never be shifted. Theorists like Gutzmann (2017) take this as evidence that some are judge-dependent, while some are speaker-dependent. Our alternative proposal is that such heavy context-sensitivity is evidence of a non-semantic influence of the sort we have outlined here.

(20.) We would like to thank audiences in both the Philosophy and Linguistics departments at The University of Manchester for helpful comments on presentations of this paper, and Martina Faller and Justina Berskyte for helpful discussions of these topics, as well as two anonymous referees for this journal.


Amaral, P., Roberts, C., and Allyn Smith, E. (2007). "Review of The Logic of Conventional Implicature by Chris Potts," Linguistics and Philosophy 30(6): 707-749.

Anand, P. (2007). "Re-Expressing Judgement," Theoretical Linguistics 33: 199-208.

Bach, K. (1994). "Postscript," in Thought and Reference. Oxford: Oxford University Press.

Bach, K. (1999). "The Myth of Conventional Implicature," Linguistics and Philosophy 22: 367-421.

Cappelen, H., and Lepore, E. (1997). "Varieties of Quotation," Mind 106: 429-450.

Cruse, D. A. (1986). Lexical Semantics. Cambridge: Cambridge University Press.

Grice, H. P. (1989). Studies in the Way of Words. Cambridge, MA: Harvard University Press.

Davidson, D. (1979). "Quotation," in Essays on Truth and Interpretation. Oxford: Oxford University Press.

Gutzmann, D. (2013). "Expressives and beyond. An Introduction to Varieties of Use-Conditional Meaning," in Gutzmann, D. and Gartner H.-M. (eds.), Beyond Expressives: Explorations in Use-Conditional Meaning. Leiden: Brill, 1-58.

Gutzmann, D. (2015). Use-Conditional Meaning: Studies in Multidimensional Semantics. Oxford: Oxford University Press.

Harris, J., and Potts, C. (2009). "Perspective-Shifting with Appositives and Expressives," Linguistics and Philosophy 32(6): 523-552.

Huang, Y. (2007). Pragmatics. Oxford: Oxford University Press.

Kaplan, D. (1977). "Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstratives and Other Indexicals," in Almog, J., Perry, J., and Wettstein, H. (eds.), Themes from Kaplan. Oxford: Oxford University Press, 481-614.

Kaplan, D. (1999). "What Is Meaning? Explorations in the Theory of Meaning as Use." Brief Version--Draft 1 (ms, UCLA)

Kaplan, D. (2005). "The Meaning of 'Ouch' and 'Oops,'" ms, paper given at UC Berkeley for the Howison Lecture in Philosophy. Transcript available at

Kolbel, M. (2004). "Faultless Disagreement," Proceedings of the Aristotelians Society 104: 53-73.

Kratzer, A. (1999). "Beyond Ouch and Oops: How Descriptive and Expressive Meaning Interact." This was written as a comment on David Kaplan's paper at the Cornell Conference on Theories of Context Dependency.

Lasersohn, P. (2005). "Context Dependence, Disagreement, and Predicates of Personal Taste," Linguistics and Philosophy 28(6): 643-686.

Lasersohn, P. (2007). "Expressives, Perspective and Presupposition," Theoretical Linguistics 33(2): 223-230.

Lasersohn, P. (2017). Subjectivity and Perspective in Truth-Theoretic Semantics. Oxford: Oxford University Press.

Levinson, S. (2000). Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press.

MacFarlane, J. (2005a). "The Assessment Sensitivity of Knowledge Attributions," Oxford Studies in Epistemology 1: 197-233.

MacFarlane, J. (2005b). "Making Sense of Relative Truth," Proceedings of the Aristotelian Society 105: 321-339.

Maier, E. (2016). "A Plea Against Monsters," Grazer Philosophiche Studien 93: 363-395.

Potts, C. (2005). The Logic of Conventional Implicature. Oxford: Oxford University Press.

Potts, C. (2007). "The Expressive Dimension," Theoretical Linguistics 33(2): 165-198.

Pratchett, T. (2007). Making Money. London: Doubleday.

Predelli, S. (2005). Contexts. Oxford: Oxford University Press.

Predelli, S. (2013). Meaning without Truth. Oxford: Oxford University Press.

Quang, P. (1971). "English Sentences without Overt Grammatical Subject," in Zwicky, A., Salus, P., Binnick, R., and Vanek, A. L. (eds.), Studies out in Left Field: Defamatory Essays Presented to James D. McCawley on the Occasion of His 33rd or 34th Birthday. Amsterdam: John Benjamins, 3-10.

Recanati, F. (2010). Truth-Conditional Pragmatics. Oxford: Oxford University Press.

Schlenker, P. (2003). "A Plea for Monsters," Linguistics and Philosophy 26(1): 29-120.

Simons, M., Tonhauser, J., Beaver, D., and Roberts, C. (2010). "What Projects and Why," in Li, N., and Lutz, D. (eds.), Semantics and Linguistic Theory 20: 309-327.

Soames, S. (2002). Beyond Rigidity. Oxford: Oxford University Press.

Tonhauser, J. (2012). "Diagnosing (Not-)At-Issue Content," Proceedings of Semantics of Under-Represented Languages of the Americas (SULA) 6: 239-254. UMass, Amherst: GLSA.

Zimmermann, E. (1991). "Kontextabhangigkeit," in von Stechow, A., and Wunderlich, D. (eds.), Semantik/Semantics: Ein internationales Handbuch der zeitgenossischen Forschung. Berlin and New York: Walter de Gruyter, 156-229.


Philosophy/School of Social Sciences, The University of Manchester, England

(corresponding author)


Philosophy/School of Social Sciences, The University of Manchester, England

How to cite: Stevens, Graham, and Nathan Duckett (2019). "Expressive Content and Speaker-Dependence," Linguistic and Philosophical Investigations 18: 97-112.

Received 5 July 2017 * Received in revised form 16 July 2018

Accepted 17 July 2018 * Available online 10 August 2018

COPYRIGHT 2019 Addleton Academic Publishers
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Stevens, Graha; Duckett, Nathan
Publication:Linguistic and Philosophical Investigations
Geographic Code:4EUUK
Date:Jan 1, 2019

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |