Daimon. Revista Internacional de Filosofía, nº 84, 2021 pp. 63-80

ISSN: 1130-0507 (papel) y 1989-4651 (electrónico)

http://dx.doi.org/10.6018/daimon.481891

Licencia Creative Commons Reconocimiento-NoComercial-SinObraDerivada 3.0 España (texto legal). Se pueden copiar, usar, difundir, transmitir y exponer públicamente, siempre que: i) se cite la autoría y la fuente original de su publicación (revista, editorial y URL de la obra); ii) no se usen para fines comerciales; iii) se mencione la existencia y especificaciones de esta licencia de uso.

 

Provocative insinuations*

 

Insinuaciones provocativas

 

ÁLVARO DOMÍNGUEZ-ARMAS**

ANDRÉS SORIA-RUIZ***

 

 

Abstract: In this paper we analyse utterances that, without explicitly constituting hate speech, nevertheless convey a hateful message. For example, in the headline “Iraqi Refugee is convicted in Germany of raping and murdering teenage girl”, the presence of “Iraqi refugee” does not seem arbitrary. To the contrary, it is responsible for inviting a racist inference against Iraqi refugees. We defend that these inferences cannot be described as slurs, ethnic or social terms used as insults, dogwhistles or conversational implicatures. Rather, we propose that these inferences are insinuations, specifically provocative insinuations, as no conversational response seems effective at blocking them.

Keywords: Conversational implicatures, dogwhistles, inferences, insinuations, hate speech.

 

Resumen: En este artículo analizamos oraciones que, sin constituir explícitamente discurso de odio, no obstante transmiten un mensaje de odio. Por ejemplo, en el titular “Refugiado iraquí es condenado en Alemania por violar y asesinar una chica adolescente”, la presencia de “refugiado iraquí” no parece arbitraria. Al contrario, invita una inferencia racista en contra de los refugiados iraquíes. Argumentamos que estas inferencias no pueden ser descritas como slurs, términos neutros utilizados como insultos, dogwhistles o implicaturas conversacionales. En cambio, proponemos caracterizar estas inferencias como insinuaciones, concretamente insinuaciones provocativas, debido a que ninguna respuesta parece efectiva a la hora de bloquearlas.

Palabras clave: Implicatura conversacional, dogwhistles, inferencias, insinuaciones, discurso de odio.

 

---

Reccibido: 31/05/2021. Aceptado: 02/07/2021.

* Thanks to Pedro Abreu, Marcin Lewiński, Mora Maldonado, Dima Mohammed, Erich Rast, José Ramón Torices , and audiences at IFILNOVA (Portugal), UNICAMP (Brazil), and the University of Granada (Spain). This work has been funded by Portuguese national funds through FCT-Fundação para a Ciência e a Tecnologia, I.P., within PhD scholarship 2020.09683.BD, and project UIDB/00183/2020, as well as by COST Action CA17132, funded by the Horizon 2020 Framework Programme of the European Union.

** PhD student at IFILNOVA, Lisbon, Portugal (a.dgueza@campus.fcsh.unl.pt). Lines of research: philosophy of language, argumentation theory, political philosophy.

*** Postdoctoral researcher at IFILNOVA, Lisbon, Portugal (aruiz@fcsh.unl.pt). Lines of research: philosophy of language, formal linguistics, metaethics. Recent publications: Soria-Ruiz, A., & Faroldi, F. L. (2020), “Moral adjectives, judge-dependency and holistic multidimensionality”, Inquiry, DOI: 10.1080/0020174X.2020.1855241, and Soria-Ruiz, A., & Stojanovic, I. (2019), “On linguistic evidence for expressivism”, in M. J. Frápolli (ed.), Expressivism, Knowledge and Truth, Cambridge: Cambridge University Press, 86, pp. 155–180.

 

 

1 Introduction

 

Hate speech has the potential to undermine the space we all inhabit by damaging the dignity of members of vulnerable groups through the association of hateful ideas with their identity (McGowan, 2012, 2019; Waldron, 2012). In this paper we are concerned with utterances that, without explicitly constituting hate speech, nevertheless convey a hateful message. Our paradigmatic examples come from headlines published, e.g., in the New York Times (NYT) (2019) or the Spanish tabloid MediterráneoDigital (2021):

 

(1) Iraqi Refugee is convicted in Germany of raping and murdering teenage girl.

Iraqi refugees are despicable/dangerous/malicious.

(2) Un marroquí, detenido por esconder 187 kilos de hachís bajo la cuna de su hijo.

A Moroccan, arrested for hiding 187kg of hashish under his son’s cradle.

Moroccan people are despicable/dangerous/malicious.

 

Although these newspapers have little in common,1 both headlines have arguably a similar effect: they invite an inference (marked with “”) against Iraqi refugees and Moroccan people, respectively. For lack of a better term, we call this a racist inference (RI). The general phenomenon is that the appearance of social, ethnic or national group terms together with actions or events that pose a risk for society suggests that such actions are typically produced by members of such groups, thereby promoting the view that those groups are despicable, dangerous or malicious.

Let us start by pointing out two aspects of this phenomenon. First, RIs do not arise with any predicate in subject position. This can be seen by noting that, if we substitute “marroquí” for another predicate, e.g., “cartero” (“postman”) in the subject position of a sentence like (2), the hateful inference disappears:

 

(3) Un cartero, detenido por esconder 187 kilos de hachís bajo la cuna de su hijo.

A postman, arrested for hiding 187kg of hashish under his son’s cradle.

 

Secondly, RIs do not arise simply as a consequence of associating negative actions or events with a group. An inference arises even in cases where the overall message is positive:

 

(4) Cuatro vecinos gitanos evitan que una mujer sufra una violación en plena calle.

Four Romani neighbours save a woman from being raped in the middle of the
street.

Contrary to expectations, these four Romani people are not despicable/dange
rous/malicious.

In this case, the headline seems to highlight the surprising character of the situation, where the saviours, and not the attackers, are Romani. Thereby, a racist inference against the Romani community still goes through, even if it is weaker or qualified.

Our research question is: how can RIs be described? In this paper, we consider the possibility of describing them as slurs (§2), ethnic and social terms used as insults (ESTIs, §3), dogwhistles (§4) or conversational implicatures (§5). We argue that RIs differ in crucial ways from each of those phenomena. We propose that RIs are a type of insinuation (Camp, 2018), in particular, a provocative insinuation (§6), whose effects are achieved via the mechanism of conversational eliciture (Cohen & Kehler, 2021). In §7, we discuss ways of confronting provocative insinuations, aimed at blocking their hateful effects. §8 concludes.

 

2. RIs as slurs

 

The presence of “Iraqi refugee”/“Morrocan” in (1–2) contributes to conveying a derogatory conceptualization of these groups via the inference that they are despicable/dangerous/malicious. Slurs give rise to similar derogatory inferences:2

 

(5) Bill is a N*

black people are despicable/dangerous/malicious.

 

Slurs are lexical items that convey a derogatory conceptualization of a group. Slurs are co-extensional with non-pejorative terms—so-called neutral counterparts (Falbo, 2021; Hess, 2020a, 2020b; Nunberg, 2018).3 The slur “N*” is a conventionally derogatory way to refer to black people. “N*” can be substituted by another non-derogatory term without triggering the derogatory inference—e.g., “Bill is black”.

Given that (1–2) convey a derogatory conceptualization of certain groups, one might be inclined to characterize “Iraqi refugee”/“Morrocan” as slurs (Hess, 2020a, 2020b) or their particular use in sentences like (1–2) as a type of slurring4 (Camp, 2013; Kukla, 2018). Moreover, slurs acquire and lose their derogatory meaning over time, so it could be the case that “Iraqi refugee” or “marroquí” (in Spanish), which on their face are not as aggressively pejorative as other paradigmatic examples, might be in the process of becoming a slur.5

However, there are several differences between our phenomenon and slurs. First, the RIs triggered by (1–2) are not limited to a single lexical item. Substituting “Iraqi refugee”/“marroquí” for (i.e.) “Iraqi asylum seeker”/“magrebí” would produce similar inferences.

Secondly, the predicates “Iraqi refugee” or “marroquí” are not conventionally derogative, as shown by the fact that they can appear in sentences that invite no derogatory inferences:

 

(6) The first Iraqi refugees arrived in Jordan in the aftermath of the 1991 Gulf War.

(7) Haz cuscús como un marroquí.

Make couscous like a Moroccan.

 

Moreover, “marroquí” is the neutral counterpart of a Spanish slur (“m*ro/a”).

Thirdly, the derogatory content of slurs and RIs project in different ways, as illustrated by the contrast between (8) and (9). As it’s well established, the derogatory content of slurs projects out of non-assertive environments such as negation, questions or modals (Bolinger, 2017; Camp, 2013; Cepollaro and Stojanovic, 2016; Croom, 2011; DiFranco, 2014; Jeshion, 2013a, 2013b):

 

(8)

(a) Bill is not a N*.

(b) Is Bill a N*?

(c) Bill might be a N*.

black people are despicable/dangerous/malicious.

 

However, the RIs of (1–2) do not arise as easily in similar environments. As shown by (9), RIs are substantially weakened in non-assertive contexts:

 

(9)

(a) Iraqi refugee is not convicted of raping and murdering a teenage girl.

(b) Is an Iraqi refugee convicted of raping and murdering a teenage girl?

(c) Iraqi refugee might be convicted of raping and murdering a teenage girl.

/ Iraqi refugees are despicable/dangerous/malicious.

 

Finally, another point of contrast between slurs and RIs is that slurs can be used as vocatives, while the predicates that trigger RIs cannot. Falbo (2021, 15) describes these “weapon uses” as the most toxic and offensive way that a slur can be used. She exemplifies this with a communicative situation where, e.g., a gay couple is on the street holding hands when an unprovoked stranger yells out “Hey F*!”. The thing to note here is that, if the speaker had yelled “Hey gay men!”—the neutral counterpart of “F*”—such an utterance would be very unnatural. The same thing is true of the predicates responsible for the inferences in (1–2). If a speaker yells out “Hey Iraqi refugee!”, this is likely to sound strange and unnatural.

In sum, although RIs are similar to slurs insofar as they derogate certain groups, RIs and slurs do not produce this effect via the same linguistic mechanism.

 

3. RIs as ESTIs

 

A more promising hypothesis might be to assimilate RIs to terms that, in contrast with slurs, do not have a demeaning effect on their targets as a matter of their lexical semantics. Such is the case of ethnic and social terms used as insults (henceforth ESTIs, Castroviejo, Fraser, and Vicente, 2020). ESTIs are terms that are generally neutral, but which, in some environments, convey a pejorative meaning built on negative stereotypes associated with the relevant group. ESTIs include ethnic, professional or even age terms like the following:

 

(10)

(a) Juan es un gitano.

Juan is a roma.

(b) Antonio es una portera.

Antonio is a doorwoman.

(c) Javier es un viejo.

Javier is an old-man.

 

Roughly speaking, (10a) attributes a set of stereotyped—negative—properties of Romani people to Juan, e.g., being a thief. Similarly, (10b) says that Antonio is a gossipy person. And (10c) describes Javier as, e.g., grumpy. The sentences in (10) convey such negative stereotypes relying on arguably neutral terms (“gitano”, “portera”, “viejo”).

In our examples (1–2) (repeated here) we also find ethnic or social terms (“Iraqi refugee”, “Moroccan”) triggering negative stereotypes. Therefore, describing RIs as ESTIs could be intuitively promising.

 

(1) Iraqi Refugee is convicted in Germany of raping and murdering teenage girl.

(2) Un marroquí, detenido por esconder 187 kilos de hachís bajo la cuna de su hijo.

A Moroccan, arrested for hiding 187kg of hashish under his son’s cradle.

 

Nonetheless, there is a crucial contrast between RIs and ESTIs: while (1–2) entail that the person referred to is an Iraqi refugee/Moroccan, the sentences in (10) do not entail that the subject belongs to the relevant group. The demeaning effect of ESTIs is achieved through the ascription of negative stereotypes to an individual, but without requiring that the individual actually belongs to the group in question. Castroviejo, Fraser and Vicente highlight a contrast between (10a) and the following sentence (without the indefinite article):

 

(11) Juan es gitano.

Juan is roma.

 

While (10a) does not entail that Juan is roma, (11) entails that Juan belongs to this group. And while (10a) attributes a set of negative stereotypes to Juan, (11) does not; it simply states that Juan is a member of this group.

In this respect, (1–2) are more similar to (11) than to (10a). Even though (1–2) convey certain stereotyped beliefs about a group, (1–2) entail that the subject is an Iraqi refugee/Moroccan. So even though (1–2) suggest that Iraqi refugees/Moroccans typically cause these dangerous actions, this is not the result of mere stereotyping, since the people alluded to in (1–2) really do belong to the relevant groups.6

We take this contrast between ESTIs and RIs to show decisively that RIs are a different phenomenon.

 

4. RIs as dogwhistles

 

Another possibility is to describe RIs as a kind of dogwhistle. A dogwhistle is a type of message designed to be taken up in different ways depending on the audience (Saul, 2018; Torices, 2021). Dogwhistles are usually conveyed by the presence of a word or idiom that, in addition to its normal interpretation, carries a coded interpretation targeted for part of the audience (call it the “target” audience). Dogwhistles can shift the boundaries of permissible discourse about a group (Maitra and McGowan, 2012; McGowan, 2019) without resorting to explicitly racist or discriminatory language.

For example, George W. Bush’s State of the Union speech from 2003 contained the following sentence (Saul, 2018, 362):

 

(12) Yet there’s power, wonder-working power, in the goodness and idealism and faith of the American people.

 

The phrase “wonder-working power” is a Biblical reference, and thus (12) has been described as an overt dogwhistle: a message with two plausible interpretations, one for the general audience and one for the target audience, namely Christians7—who must recognize the intention of the speaker to convey the private message.

Another example is the expression “inner-city”, used in the United States as a coded way of meaning “black’ (Saul, 2018, 367):

(13) Some people want to increase spending for new prisons to lock up violent inner-city criminals.

(13) is a covert dogwhistle, because the target audience need not recognize that a coded message is sent to them.

 

(1–2) might be described as overt dogwhistles if they are conceived as coded messages targeted to a private audience. (1–2) could be inducing in the audience certain beliefs about the dangerousness of Iraqi refugees or Moroccan people which, in turn, might lead to the implementation of policies against those groups. In 2019 there was public concern in Germany regarding the integration of refugees.8 One might say that the RI associated with (1) is a coded message from the NYT, aimed against the integration of refugees in Germany. The general audience would conceive (1) as a non-remarkable piece of news—after all, there were 720 murders and 7.724 rapes in Germany in 2019 (Statista Research Department, 2021)—but xenophobic subjects would understand (1) as an example in support of their views. If this was the case, (1–2) might be considered as a type of overt dogwhistle. But even if a xenophobic subject failed to recognize that a message is being sent to them, a defender of this view might want to treat (1–2) as covert dogwhistles—much like “inner city”—so that target audience recognition is not required.

However, there are two crucial differences between RIs and dogwhistles: first, dogwhistles are intended to be perceived only by the “target” audience, since they rely on the presence of expressions whose coded meaning is only understood by some hearers (“wonder-working power”, “inner city”). The inferences associated with (1–2), even if triggered by the presence of certain predicates (“Iraqi refugee”, “Moroccan”), do not rely on those predicates having a coded or hidden meaning: these inferences can be perceived by anyone. Secondly, dogwhistles lose part of their power when they are “uncovered” (Saul, 2018, 373; see also Torices, 2021). In particular, overt dogwhistles are more powerful insofar as the general audience does not recognize them, and covert dogwhistles only succeed if the general audience does not recognize the manipulative intention that underlies them. By contrast, the effects of RIs persist even if the audience recognizes the hateful inference.

We will discuss the effects of RIs in Section 7. For now, we take these observations to suggest that RIs cannot be described as dogwhistles.9

 

5.RIs as conversational implicature

 

Might RIs be a kind of conversational implicature (CI, Grice, 1989)?10 In a CI, the hearer makes an inference based on an utterance that violates one or more conversational maxims,11 while assuming that the speaker is cooperative.

Implicature calculation follows roughly these steps (Grice, 1989; Cohen and Kehler, 2021). First, the hearer recovers the literal semantic content p of the sentence uttered by the speaker. Secondly, the hearer notices that p violates a conversational maxim. Thirdly, the hearer—while assuming that the speaker is aware of such maxims—infers that the sentence must have the secondary purpose of conveying an extra-semantic content q. Finally, the hearer treats the utterance as conveying the extra-semantic content q in addition to the literal semantic content p.

Consider the following standard example, based on a violation of the Maxim of Relation (Be relevant!, Grice, 1989, 32): Imagine a subject a, whose car is immobilised, approached by another subject b. The following exchange takes place:

 

(14)

a: I am out of petrol.

b: There is a garage round the corner.

 

In that situation, a first recovers the literal semantic content expressed by b as the proposition that there is a garage nearby. Secondly, a notices that this proposition is irrelevant vis-à-vis a’s initial utterance. That is, b’s utterance violates the Maxim of Relation. Thirdly, a assumes that b is a cooperative speaker. Thus, there must be a secondary purpose in pointing out the nearby garage, (e.g.) that the garage has petrol to sell. Fourthly, a infers from b’s utterance that there is a garage round the corner, and additionally, that it has petrol to sell.

One of the essential characteristics of conversational implicatures is that they are cancellable: the speaker can deny the extra-semantic content without contradiction. In the previous example, speaker b can follow-up their utterance with “… but in fact they do not sell petrol there” without contradicting themself.

The RIs of (1–2) (repeated here) could be described as conversational implicatures—based on the idea that the literal content of (1–2) is a violation of a conversational maxim.

 

(1) Iraqi Refugee is convicted in Germany of raping and murdering teenage girl.

Iraqi refugees are despicable/dangerous/malicious.

(2) Un marroquí, detenido por esconder 187 kilos de hachís bajo la cuna de su hijo.

A Moroccan, arrested for hiding 187kg of hashish under his son’s cradle.

Moroccan people are despicable/dangerous/malicious.

 

Specifically, it could be argued that mentioning the person’s nationality or international status is a violation of the Maxim of Relation and that this generates the inference.12 This hypothesis is immediately supported by the fact that RIs are cancellable: the speaker of (1), for example, can follow up their utterance with “…but in fact, Iraqi refugees are great people” without contradiction.

The line of reasoning of the audience of (1) might be as follows: first, the audience recovers the literal semantic content of (1), namely that the person convicted in Germany is an Iraqi refugee. Secondly, the audience notes that the literal content of (1) is in violation of the Maxim of Relation, since the nationality or international status of the perpetrator is irrelevant. Thirdly, in order to reconcile the literal content of (1) with the assumption that the speaker is being cooperative, the audience infers that the nationality of the perpetrator must be necessary to make sense of the story. Fourthly, a possible interpretation reached by the audience would be that these actions (murder, rape) are typically caused by Iraqi refugees and therefore, that Iraqi refugees are despicable/dangerous/malicious (a similar line of reasoning might be followed by the hearer of (2)).

However, there are salient differences between RIs and CIs, which speaks against subsuming the former under the latter. First, it is not clear that the extra-semantic content of (1–2) arises as a result of a threat of communicative failure. Above, we described the RIs of (1–2) as inferences triggered by a violation of the Maxim of Relation. But this account makes a bad prediction, and is independently questionable. First, if violating Relation was determinant for this phenomenon, any irrelevant predicate should invite similar inferences to those triggered by (1–2). But this is not the case, as illustrated by (3) (“A postman, arrested for hiding 187kg of hashish under his son’s cradle”). After all, the perpetrator’s occupation is no more relevant than their nationality. A defender of the implicature view would be hard-pressed to account for the difference between equally irrelevant predicates. As we argue in Section 6, we think that these inferences rely largely on world knowledge, and not just on the status of these different predicates vis-à-vis the Gricean maxims.

Moreover, from the point of view of a racist, the idea that mentioning the nationality or origin of the perpetrators in (1–2) violates the Maxim of Relation is questionable. If you are a racist, you will likely think that the nationality or origin of the perpetrator is essential to understand what happened. Put differently, thinking that (1–2) violate the Maxim of Relation requires the antecedent (presumably anti-racist) belief that nationality/origin plays no explanatory role. Without such belief, the premise that the literal contents of (1–2) are conversationally devious is missing, and the implicature calculation cannot take off. Presumably however, the RIs triggered by (1–2) do not require that the audience is anti-racist; such inferences can be drawn by any audience, racist or anti-racist.13

Secondly, there is an important contrast between CIs and RIs when they are cancelled in a subsequent utterance. When CIs are cancelled, the speaker appears uncooperative. Let us return to Grice’s example (14). Imagine that a wants to confirm that the garage sells gasoline. She asks, “Do they sell petrol there?” and b answers “No, they don’t”. Clearly, b will appear uncooperative. The reason for this is that, without the implicature, b’s original contribution seems completely irrelevant to the conversation.

This does not happen with (1–2). Imagine that, after publishing (1), the NYT is accused of racism—e.g., on Twitter. Such accusations of racism might rely on the RI triggered by (1) about Iraqi refugees. The editor of NYT could post on Twitter “The NYT didn’t say/mean that”/“Iraqi refugees are wonderful people”.14 Setting aside the awkwardness of such an apology, these responses would cancel the RI of (1). And crucially, they would not make the editors of NYT look uncooperative for having published (1). However, if (1) was to be described as a CI, one would expect that, when the implicature is cancelled, the speaker should appear uncooperative and the original contribution should seem conversationally defective. Importantly, that does not mean that, if the editors of NYT cancelled such an inference, they would not come across as hypocrites. They probably would (especially given the existence of style-books that discourage these practices, see n. 12). But being a hypocrite is not the same as being conversationally uncooperative.15

In sum: when their inference is subsequently cancelled, (1–2) do not become conversationally defective, and speakers do not seem uncooperative. This suggests that (1–2) do not violate any conversational maxim, and thus that the RIs are not rescuing an otherwise uncooperative contribution. These observations suggest that cases such as (1–2) cannot be described as conversational implicatures.16

 

6. RIs as insinuations

 

Up to now, we have discarded an account of the RIs of (1–2) as a type of slurring, ESTIs, dogwhistles and conversational implicature. In this section, we defend that RIs can be characterized as insinuations (Camp, 2018; Fricker, 2012), specifically triggered by the mechanism of conversational eliciture described by Cohen and Kehler (2021).

Insinuations are “off-record” pragmatic inferences triggered by certain sentences. For example, suppose that a real estate agent says the following to potential buyers from a different racial background than the local majority (Camp, 2018, 43):

(15) Perhaps you would feel more comfortable locating in a more… transitional neighbourhood, like Ashwood?

Where the buyers perceive the explicit suggestion to seek a house elsewhere, the implicit insinuation is that they should feel uncomfortable in this neighbourhood by virtue of not belonging to the majority group. Another example arises from a situation where a driver is stopped for speeding. They might utter (16) in order to suggest a bribe (Lee and Pinker, 2010; Pinker, Nowak and Lee, 2008, cited in Camp, 2018):

 

(16) I’m in a bit of a hurry. Is there any way we can settle this right now?

 

Again, whereas the explicit message is an innocent question, the implicit insinuation is an invitation to accept a bribe. For a final example, if someone utters

 

(17) You know that Obama’s middle name is Hussein. I’m just saying.

the hearer might infer that, e.g., that Obama supports radical Islamism (Camp, 2018, 43).

 

An insinuation therefore consists of two messages. On the one hand, a literal “on-record” content—e.g., that the buyers would be more comfortable elsewhere—that is unobjectionable. On the other hand, these sentences carry an “off-record” message, e.g., that the buyers would be uncomfortable because they belong to a different racial group. The “off-record” content is such that the speaker cannot be blamed for having conveyed it. As a result, the literal content of these sentences is apparently innocuous, but they manage to subserve a controversial conversational point.

According to Camp, the main property that characterizes insinuations is that speakers can convey them without being held accountable for it. More specifically, Fricker (2012) claims that a distinctive feature of insinuations is that they are cancellable (Fricker says “deniable”) and disavowable: the speaker can cancel their content as well as deny having had the intention to convey that content without appearing uncooperative.17 For instance, if the speaker of (17) is confronted with “Are you suggesting that Barack Obama is a radical Islamist?”, they can answer “Of course not, he’s not a radical Islamist!”—cancellation—or “No, I didn’t say/mean that!”—disavowal. In this respect, insinuations behave differently from conversational implicature.18

We submit that (1–2) can be described as a type of insinuation. On the one hand, (1–2) convey the “on-record” content that an individual of a certain background committed a crime. On the other hand, (1–2) convey the “off-record’ content that individuals from that background are despicable/dangerous/malicious.

An immediate advantage of this hypothesis is that it bears out the prediction that RIs can be cancelled and/or disavowed by the speaker without appearing uncooperative. Consider the following dialogues as continuations of (1–2):

 

(1)

a. (An) Iraqi refugee was convicted in Germany because of raping and murdering
teenage girl.

b. Are you suggesting that Iraqi refugees are despicable/dangerous/malicious?

a. No, Iraqi refugees are great people/ I didn’t say/mean that!

(2)

a. (A) Moroccan was arrested for hiding 187 kg of hashish under his son’s cradle.

b. Are you suggesting that Moroccans are despicable/dangerous/malicious?

a. No, Moroccans are great people/ I didn’t say/mean that!

 

Thus, we consider (1–2) as insinuations which trigger in the audience certain associations of hateful ideas about a group. However, as illustrated by the previous examples, insinuations are a varied lot. We still need to point out the specific mechanism by which these inferences arise. The mechanism that underlies such associations is, we think, a form of what Cohen and Kehler (2021) call conversational eliciture. A conversational eliciture arises when, by choosing a particular predicate (among others), a speaker elicits inferences on the part of the audience that would not otherwise be drawn. This phenomenon can be illustrated by the following examples (Cohen and Kehler, 2021, 2-3):

 

(18)

a. The company fired the manager who was embezzling money. (Rohde et al. 2011)

b. The drug-addled undergrad fell off the Torrey Pines cliffs.

c. A jogger was hit by a car in Fresno last night.

 

On top of conveying their literal content, these sentences invite the inference that the embezzlement was the reason for the firing in (18a), that drugs caused the undergrad to fall off the cliffs (18b) and that the victim of the accident was jogging when the accident happened (18c). Note that the inferences in (18a–c) are cancellable. For instance, (18a) can have the following continuation:

 

(18a)

a. The company fired the manager who was embezzling money.

b. So she was fired because she was embezzling money?

a. No, she was fired because she’s always late!

 

We think that the inferences associated with (1–2) can be described as insinuations that rely upon this very same mechanism: by linking a group with dangerous actions, they strongly invite the hateful inference that Iraqi refugees/Moroccans typically cause these actions and, thereby, that they are despicable/dangerous/malicious.

This hypothesis is consistent with the properties of RIs that we have observed throughout this paper: first, we observed that the RIs of (1–2) do not project out of entailment-cancelling environments in the same way as the derogatory inferences of slurs do. Conversational elicitures do not have the projective profile of presuppositions, and therefore, they are not expected to project out of entailment-cancelling environments.

Secondly, in contrast to ESTIs, we observed that the sentences that trigger RIs entail that the predicate applies to the individual in subject position: (1) entails that the murderer is an Iraqi refugee; (2) entails that the subject is Moroccan. Conversational elicitures also have this property: (18a) entails that the manager was embezzling money; (18b) entails that the undergrad was on drugs; (18c) entails that the victim was a jogger.

Thirdly, we saw that, in contrast with dogwhistles, the RIs of (1–2) persist despite being recognized by the audience: if the speaker of (1) is accused of inviting the inference that Iraqi refugees are despicable/dangerous/malicious, the inference will not thereby disappear or become weaker. Conversational elicitures also persist after being recognized by the audience: if a hearer accuses the speaker of (16a) of suggesting that the embezzlement caused the firing, that inference would not vanish.

Fourthly, as we have just seen, the RIs triggered by (1–2) are cancellable and disavowable, and so are conversational elicitures. Moreover, conversational elicitures do not rely on the violation of any conversational maxim, nor are they meant to rescue an otherwise uncooperative contribution.

We still need to account for examples like (3) and (4) (repeated here).

 

(3) Un cartero, detenido por esconder 187 kilos de hachís bajo la cuna de su hijo.

A postman, arrested for hiding 187kg of hashish under his son’s cradle.

(4) Cuatro vecinos gitanos evitan que una mujer sufra una violación en plena calle.Four Romani neighbours saved a woman from being raped in the middle of the street.

 

As we mentioned in the introduction, (3) does not invite an inference against postmen, and (4) invites the hateful inference that these four people surprisingly prevented, rather than committed, a crime.

Cohen and Kehler (2021) observe that there are sentences with similar structure to (18) that do not trigger the same inferences. For example, consider the following two alternatives to (18a):

 

(19)

a. The company fired the manager who was hired in 2002.

b. The company fired the manager who has a long history of corporate awards.

 

(19a) does not offer an explanation for the firing, and therefore does not invite the conversational eliciture that being hired in 2002 caused the firing. By contrast, (19b) gives a positive description of the manager, inviting the inference that the firing is counter to expectations. The reasons for these effects are extralinguistic. Embezzling money is an illegal activity and, as such, it can of course cause someone to be fired from their job. No such thing is true of the property of being hired in 2002 (although of course we can cook up an appropriate context: suppose that the company is firing its most veteran employees). Similarly, having a long history of corporate awards is a good thing, and thus, it goes against expectations that a company should fire someone like that.

We think that a similar explanation can capture, respectively, examples (3) and (4). First, consider (3). The nationalities and backgrounds of individuals are connected to racist attitudes, prejudices and stereotypes. By contrast, their occupation does not, in general, have this property. Thereby, “postman” in (3) does not offer an explanation for the arrest. Importantly, this is an extra-linguistic consideration. We can of course imagine a context in which there exists a prejudice against postmen. In that context, (3) would invite a similar inference to (1–2). Thus, we submit that sentences like (3) are like (19a) in the following sense: given world knowledge, in most contexts they do not invite an inference that the choice of predicate is guided by an explanatory purpose.

Similarly, due to prevalent racist prejudice against the Romani community, the hateful inference of (4) is that, contrary to expectations, Romani people were the saviours and not the attackers. This inference is, again, due to world knowledge: racist attitudes against the Romani community abound, and in that context, for a speaker to report a situation in which members of this community prevented rather than committed a crime, invites the inference that this is surprising. The situation is parallel to (19b).

 

7. How to respond? The provocative dimension of RIs

 

We finish this paper by arguing that the insinuations triggered by (1–2) are provocative, because they place the audience in a frustrating position. (1–2) provoke a variety of audience responses, none of which are ultimately successful in blocking their effects. The audience is thereby forced to succumb to the provocative dimension of RIs. Audience responses19 to these inferences can be categorized as follows:

 

1) Ignore the RI. Faced with (1–2), the audience might simply ignore the hateful inference. This strategy, however, has a risk: if it is not confronted, the hateful inference gets incorporated into the common ground of the conversation (Lewis, 1979; Stalnaker, 2002). If this happens, the common ground will contain racist attitudes about Iraqi refugees/Moroccans.20

2) Confront the RI. Alternatively, the implicit content of (1–2) might be confronted or challenged21 explicitly by the audience. The audience could accuse the speaker of having insinuated the implicit content, for example saying “Are you suggesting that Iraqi refugees are despicable/dangerous/malicious?”.

 

However, the speaker can reply to such a confrontation by different means. Regarding (1), for example, we saw that the speaker can cancel or disavow the inference (“No, Iraqi refugees are great people!”/“I didn’t say that!”/“I’m just stating facts!”). But the speaker can also reply “You said it, not me! But now that you mention it…’ (Camp, 2018, 46). This last type of responses can make the RI turn into the topic of the conversation. This would shift the agenda of the conversation. Participants abandon the flow of the dialogue to discuss whether or not there is a hidden inference under the utterance, or whether or not the inference is true. Even further, the content of the inference might gain credibility simply by coming under discussion.

In sum, any response that targets the RI explicitly seems inefficient at blocking the inference and, moreover, risks turning the conversation into a racist one, thereby serving the purposes of the insinuator.

We want to propose, in closing, that a more effective response might be to “counter-attack” the RI with similar inferences. Consider the following replies to (1):

 

(20) a. And another Iraqi refugee was selected to play with the Portuguese national
team. So what?

b. And a German was convicted of the same crime a week ago. So what?

 

(20a–b) are counter-attacks to the provocative insinuation in (1). These responses also invite conversational elicitures: (20a) invites the inference that Iraqi refugees are good football players; (20b) invites the inference that Germans are despicable/dangerous/malicious. Put differently, (20a–b) are attempts to undermine the RI of (1) using the same mechanism. If the inference of (20a) is accepted, or even if it becomes the topic of conversation and speakers start to discuss whether Iraqi refugees are good at football, the hateful effect of the original RI will vanish, or at least it will be “diluted” in combination with other, similar (but not hateful) inferences. Alternatively, if the inference from (20b) is accepted, or if it becomes the topic of the conversation, the hateful effect of (1) would be “compensated” by the hateful effect of (20b). The former strategy is perhaps more desirable, since it does not involve more hate speech, but both strategies might seem effective at neutralizing the original inference. On the other hand, note that (20a) is structurally similar to (4) (repeated here), which we observed gives rise to a hateful inference as well.

 

(4) Cuatro vecinos gitanos evitan que una mujer sufra una violación en plena calle.

Four Romani neighbours save a woman from being raped in the middle of the street.

 

This seems a consideration against saying things like (20a). However, there is a crucial difference between uttering sentences like (20a) or (4) out of the blue and uttering them as responses to sentences like (1): in the former case, they may contribute to hate discourse “from scratch”, by triggering the “counter-to-expectations” inferences that we observed; but in the latter case they may succeed at diluting the effects of a previous hateful utterance.

Before wrapping up, one might object to our counter-attack strategy as follows: failing to make explicit the harmful inferences associated with (1–2) can have a normalizing effect, effectively reinforcing the racist prejudices and biased ideology at play in these examples. We think that this depends, to some extent, on whether the conversation involves only two (speaker and addressee) or more than two parties. In the former case, we think that the risks considered above overcome possible benefits of confronting the RIs, especially if the speaker is a racist. But we acknowledge that, in the latter case, making explicit the RIs of sentences like (1–2) to a general audience might indeed be more beneficial than harmful. This can help other members of the audience spot a type of hate speech that would otherwise remain hidden.22

This is a mere sketch of different strategies that a speaker who wants to counter hateful inferences of this kind can deploy, and which to our knowledge have not been explored in any great detail in the literature. But this merits more careful attention than we have space for here. We leave this discussion for future work.

 

Conclusion

 

In this paper we have defended that headlines like (1–2) do not feature “Iraqi refugee”/“Moroccan” arbitrarily. Rather, they invite racist inferences against the relevant groups. We considered and rejected various possible accounts of this phenomenon, namely the hypothesis that these inferences are a type of slurring, that these inferences fall under ethnic or social terms used as insults (ESTIs), that they are understood as dogwhistles, and that they are a kind of conversational implicature.

In contrast to those accounts, we defended that the inferences of (1–2) can be described as insinuations, because they appear to be “off-record”, and thus cancellable and disavowable. Moreover, we argued that the RIs of (1–2) are produced via the mechanism of conversational eliciture—by choosing a particular predicate over others, the speaker elicits the inference that the property denoted by the predicate has an explanatory role. By choosing to mention that the perpetrator is an Iraqi refugee/Moroccan, the speaker invites the inference that being an Iraqi refugee/being Moroccan explains why they committed those crimes.

Finally, we pointed out that these insinuations have a provocative character, because they place the hearer into a frustrating position: any type of response that targets the RI seems inefficient at blocking its hateful effects. We tentatively proposed that the audience may seek a response that counter-attacks the RI with similar inferences, so that the hateful effects of the original utterance are diluted or somehow compensated.

 

References

 

Bolinger, R. J. (2017), “The pragmatics of slurs”, Noûs, 51(3), 439–462.

Camp, E. (2013), “Slurring perspectives”. Analytic Philosophy, 54(3), 330–349.

Camp, E. (2018), “Insinuation, common ground, and the conversational record”, in D. Fogal, D. W. Harris & M. Moss (eds.), New Work on Speech Acts, Oxford: Oxford University Press, pp. 40–66.

Carrera, M. (2017), “Enfocando a los “digitales de odio””, https://www.eldiario.es/catalunya/opinions/enfocando-digitales-odio_132_3518066.html.

Castroviejo, E., Fraser, K. & Vicente, A. (2020), “More on pejorative language: Insults that go beyond their extension”, Synthese, DOI: 10.1007/s11229-020-02624-0.

Cepollaro, B. & Stojanovic, I. (2016), “Hybrid evaluatives: In defense of a presuppositional account”, Grazer Philosophische Studien, 93(3), 458–488.

Cepollaro, B., Sulpizio, S. & Bianchi, C. (2019), “How bad is it to report a slur? An empirical investigation”, Journal of Pragmatics, 146, 32–42.

Cohen, J. & Kehler, A. (2021), “Conversational eliciture”, Philosopher’s Imprint, 21(12), 1–26.

Colegio de Periodistas de Cataluña (2020), Manual de estilo sobre minorías étnicas, https://aulaintercultural.org/wp-content/uploads/2014/04/manualdeestilo.pdf.

Consejo Audiovisual de Andalucía (2020), “Recomendaciones del Consejo Audiovisual de Andalucía sobre el tratamiento de la Inmigración en los medios audiovisuales”, http://www.consejoaudiovisualdeandalucia.es/sites/default/files/recomendaciones/Recomendaciones_2006_02_Tratamiento inmigracion.pdf.

Corredor, C. (2014), “Pejoratives and social interaction”, in P. Stalmaszczyk (ed.), Issues in Philosophy of Language and Linguistics, Łódź: Wydawnictwa Uniwersytetu Łódzkiego, pp. 39–55.

Croom, A. M. (2011), “Slurs”, Language Sciences, 33(3), 343–358.

DiFranco, R. (2014), “Pejorative language”, Internet Encyclopedia of Philosophy.

Falbo, A. (2021), “Slurs, neutral counterparts, and what you could have said”, Analytic Philosophy, DOI: 10.1111/phib.12217.

Fricker, E. (2012), “Stating and insinuating”, Proceedings of the Aristotelian Society Supplementary Volume, 86(1), 61–94.

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

Hess, L. (2020a), “Practices of slur use”, Grazer Philosophische Studien, 97(1), 86–105.

Hess, L. (2020b), “Slurs and expressive commitments”, Acta Analytica, 36(2), 263–290.

Jeshion, R. (2013a), “Expressivism and the offensiveness of slurs”, Philosophical Perspectives, 27(1), 231–259.

Jeshion, R. (2013b), “Slurs and stereotypes”, Analytic Philosophy, 54(3), 314–329.

Kukla, R. (2018), “Slurs, interpellation, and ideology”, Southern Journal of Philosophy, 56, 7–32.

Lance, M. & Kukla, R. (2013), “Leave the gun; take the cannoli! The pragmatic topography of second-person calls”, Ethics, 123(3), 456–478.

Lee, J. J. & Pinker, S. (2010), “Rationales for indirect speech: The theory of the strategic speaker”, Psychological Review, 117(3), 785–807.

Lewiński, M. & Mohammed, D. (2016), “Argumentation theory”, The International Encyclopedia of Communication Theory and Philosophy, DOI: 10.1002/9781118766804.wbiect198.

Lewis, D. (1979), “Scorekeeping in a language game”, Journal of Philosophical Logic, 8(1), 339–359.

Maitra, I. & McGowan, M. K. (2012), “Introduction and overview”, in I. Maitra & M. K. McGowan (eds.), Speech and Harm: Controversies over Free Speech, Oxford: Oxford University Press, pp 1–23.

McGowan, M. K. (2012), ““Whites only” signs and racist hate speech: Verbal acts of racial discrimination”, in I. Maitra & M. K. McGowan (eds.), Speech and Harm: Controversies over Free Speech, Oxford: Oxford University Press, pp. 121–147.

McGowan, M. K. (2019), Just Words: On Speech and Hidden Harm, Oxford: Oxford University Press.

Mohammed, D. (2019a), “Managing argumentative potential in the networked public sphere: The anti-#MeToo manifesto as a case in point”, in B. Garssen, D. Godden, G. R. Mitchell & J. H. M. Wagemans (eds.), Proceedings of the 9th Conference of the International Society for the Study of Argumentation, Amsterdam: SIC SAT, pp. 813–822.

Mohammed, D. (2019b), “Standing standpoints and argumentative associates: What is at stake in a public political argument?”, Argumentation, 33(3), 307–322.

Nunberg, G. (2018), “The social life of slurs”, in D. Fogal, D. W. Harris & M. Moss (eds.), New Work on Speech Acts, Oxford: Oxford University Press, pp. 237–295.

Pellicer, M. (2015), “Periodismo e inmigración: cómo deben informar los medios”, https://miquelpellicer.com/2015/08/periodismo-inmigracion-como-deben-informar-los-medios/.

Pinker, S., Nowak, M. A. & Lee, J. J. (2008), “The logic of indirect speech”, Proceedings of the National Academy of Sciences of the United States of America, 105(3), 833–838.

Popa-Wyatt, M. & Wyatt, J. L. (2018), “Slurs, roles and power”, Philosophical Studies, 175(11), 2879–2906.

Saul, J. (2002), “Speaker meaning, what is said, and what is implicated”, Noûs, 36(2), 228–248.

Saul, J. (2018), “Dogwhistles, political manipulation, and philosophy of language”, in D. Fogal, D. W. Harris & M. Moss (eds.), New Work on Speech Acts, Oxford: Oxford University Press, pp. 360–384.

Stalnaker, R. (2002), “Common ground”, Linguistics and Philosophy, 25(5–6), 701–721.

Statista Research Department (2021), “Number of German male suspects of rape and sexual assault Germany, by age group”, https://www.statista.com/statistics/1101516/male-suspects-german-rape-sexual-assault-age-group-germany/.

Torices, J. R. (2021), “Understanding dogwhistle politics”, Theoria: An International Journal for Theory, History and Foundations of Science, DOI: 10.1387/theoria.22510.

Waldron, J. (2012), The Harm in Hate Speech, Cambridge, MA: Harvard University Press.

 

Notes