On the Rhetorical Effectiveness of Implicit Meaning—A Pragmatic Approach

Abstract

This paper explores the interface between pragmatics and argumentation by considering the impact of different types of implicit meaning on different types of rhetorical effects. On the rhetorical front and drawing on classical rhetoric but going beyond the Aristotelian rhetorical triangle (ethos, logos, pathos), the paper discusses an open list of rhetorical effects affecting speakers, audiences, messages and the conversational flow of interaction. On the pragmatic front, the paper accounts for how specific features of different types of implicit meaning (presupposition, implicature and back-door speech acts) are likely to trigger specific rhetorical effects. In so doing, it discusses and justifies the need for and the feasibility of an experimental investigation of the rhetorical effectiveness of implicit meaning.

1. Introduction

During the 2016 presidential campaign, then-candidate Donald Trump made a habit of conveying disparaging and even legally reprehensible public statements.1 Interestingly, these messages were all plausibly deniable—and Trump often did deny having meant them. For instance, he was accused of (i) insinuating, after the first Republican primary debate, that the moderator’s (Megyn Kelly) alleged aggressiveness towards him was due to the fact that she was menstruating, and (ii) inciting murder on his Democrat rival, Hillary Clinton, by pondering whether “2nd amendment people” could “do something” to prevent her from being elected. These messages shared two properties: they (i) fulfilled an argumentative function of support that was (ii) implicitly expressed. From a communicative perspective, these messages are problematic: while their sheer presence in discourse is controverted (since they are deniable), they distract attention from main issues and support sneaky forms of misinformation. In other words, implicit meaning (henceforth IM) seems to offer a range of advantages to arguers who can hope to trigger different types of rhetorical effects.

Argumentation theory, since its inception, has investigated a number of aspects of argumentative practices, covering, among others, its epistemic dimension (i.e., how argumentation contributes to establish truthful and justified conclusions), its dialectical dimension (i.e., how argumentative discussions are regulated), its rhetorical dimension (i.e., how argumentation influences others), its linguistic dimension (i.e., what kind of linguistic and verbal resources arguers draw on as they engage in argumentation), and its interpersonal dimension (i.e., how the practice of argumentation unfolds in social interaction). In seeking to account for these, argumentation scholarship has gradually constituted the field and allowed to develop complementary research directions meant to elucidate the phenomenon. Within this panorama, this paper narrows down the inquiry to the rhetorical dimension of argumentation, which is chiefly concerned with the effects arguments have on their audiences during argumentative encounters. Within the realm of rhetoric, it specifically focuses on the rhetorically significant possibilities different types of IM have to offer to speakers. That is, we will be concerned here with the way pragmatic research can illuminate rhetorical issues. These will furthermore be investigated along four dimensions: effects on the speaker, effects on the audience, effects on the message and effects on the conversational dynamics. In so doing, we pursue three goals: (i) we aim to show that IM is a rhetorically significant category of verbal meaning; (ii) we aim to provide an initial and principled compendium of types of rhetorical effects that are likely to be triggered by different types of IM; and (iii) we discuss the need for and the feasibility of an experimentally grounded study of the rhetorical effects of IM.

In Section 2, we start by considering rhetorical effectiveness through its classical Aristotelian roots and discuss their limitations while reflecting on additional types of rhetorical effects that classical Aristotelian categories do not seem to cover. Section 3 subsequently attempts to extend the notion of rhetorical effectiveness by making the case that pragmatic research is ideally suited to this purpose, first by recalling extant research in pragmatics that already tackles issues pertaining to the field of rhetoric and, second, through a theoretical justification of why rhetorical effects are to be considered, in the speech act-theoretic tradition, as perlocutionary effects. Section 4, in turn, explains why IM itself is inherently suited to be exploited rhetorically by examining evolutionary, psychological, and social strands of research which support this claim. With these justifications in mind, Section 5 then considers the features of three categories of IM (implicature, presupposition and back-door speech acts) and their associated rhetorical effects, while Section 6 draws their respective matrix of correspondences. Section 7 concludes the paper with thoughts on the experimental possibilities research at the interface between pragmatics and rhetoric is likely to generate.

2. The Classical Take on Rhetorical Effectiveness (and Associated Problems)

The first—and still probably only—systematic account of rhetorical effectiveness is Aristotle’s model of rhetoric, which treats persuasiveness as an effect of discourse at the confluence of three dimensions: logos, ethos and pathos.2 In the next subsections, we briefly recall these classical Aristotelian categories and try to frame their study in a way that brings forward the relevance of approaching them with pragmatic tools.

2.1. Aristotle’s Triangle: Ethos, Logos, Pathos

Aristotle famously held that persuasion should not be restricted to the content of argumentation alone, as other parameters are deemed to influence the persuasiveness of a message:

“Of the modes of persuasion furnished by the spoken word there are three kinds. The first kind depends on the personal character of the speaker; the second on putting the audience into a certain frame of mind; the third on the proof, or apparent proof, provided by the words of the speech itself”.

(Aristotle 1995, pp. 1356a1–1356a14)

Ethos is defined by the degree of institutional integration, credibility, reputation, and authority of speakers (Braet 1992; Gallez and Reynders 2015). The construction of ethos is not restricted to discursive features and can be extended to paraverbal and non-verbal cues (e.g., appearance and gestures; Gallez and Reynders 2015). The speaker’s history and reputation make up what Amossy (2001) calls the ‘prior ethos’:3 in the situation where their ethos has been tainted either personally (e.g., as a result of a court case) or institutionally (e.g., because of their affiliation with a political party with a bad reputation), speakers will aim to remove the prejudices the audience might hold against them throughout their speech to regain the trust of the audience and restore their trustworthiness. The displayed ethos during the debate should appeal to the demographic attributes of the audience (e.g., age and culture) as well as the ethos the audience expects or prefers (e.g., reputable, expert in her field and benevolent; Braet 1992). The audience will be more persuaded by a speaker’s discourse if the latter is recognised as expert, benevolent, and trustworthy.4 Modern accounts of ethos, such as Mayer et al.’s (1995) model of organisational trust, propose to precisely circumscribe what ethos consists of, how it is shaped in discourse, and what language resources are available for its management. Experimental work conducted by Hovland et al. (1953) has furthermore assessed the credibility of the trustee with factors of expertise and trustworthiness, and found that integrity, ability, and benevolence could be singled out as key dimensions of trustworthiness. This three-fold description of trustworthiness, which incidentally resonates with the identification, by Aristotle, of three features “which inspire confidence in the orator’s own character”, namely “good sense, excellence, and goodwill” (Aristotle 1995), constitutes the foundation of Mayer et al.’s model of trust (1995).

In its Aristotelian conception, logos can refer to speech (and its verbal material, e.g., words, claims), myth (in the sense of ‘plot’), or rationality (or logic). In our discussion, we refer to the first of these and define logos in terms of the content of the message, which has to be clear, accessible and acceptable in order to be rhetorically effective, that is, to offer sound evidence for a given claim (Aristotle 1995). Because logos is given to us through the speaker’s choice of words, this is the persuasive dimension that best showcases the importance of wording in any persuasive endeavour. The content of a message, along with the stylistic choices implemented in its articulation, are directly responsible for the representation of the message that the audience will form—which in turn will trigger further perlocutionary effects. This is a clear indication that the study of rhetorical effectiveness, when focusing on logos, should incorporate a rigorous account of meaning, which pragmatics is able to offer.

Pathos refers to the affective ways in which the message will influence the beliefs, attitudes, and values of an audience. It aims to put hearers “into the right frame of mind” (Aristotle 1995): the judgement given by an audience to a speaker is assumed to be different when, for instance, the former are joyful and happy than when they are grieving and hostile. The most persuasive arguments are usually deemed to be those that resonate with the audience’s most deep-held values: the deeper the value raised by the discourse is held, the more the argument will be impactful on the audience in terms of pathos. With the rhetorical support of logos, pathos is not only limited to impacting an audience’s affects; its influence can extend to actual behaviour (e.g., votes, uprising, or discrimination), even in unconscious ways.5

2.2. Some Problems with the Classical Account

Illuminating as Aristotle’s rhetorical categories can be, they face two problems: (i) the problem of multiple rhetorical effects, and (ii) the problem of the irreducibility of some rhetorical effects to any of the three existing categories. (i) is a problem of descriptive adequacy, as it emerges when more than one appeal is found in the same argumentative move, which challenges not only descriptive models but also explanatory accounts of rhetorical effectiveness. (ii) is a problem of scope, in the sense that some dynamic rhetorical effects that arise in conversations do not seem to be about persuasion per se, but rather about ways of managing the interaction by implementing constraints on the conversational flow. Let us take two examples to illustrate these problems. The first is an alleged conversation between former boxer Muhammad Ali and a flight attendant:6

• Just before takeoff on an airplane flight, the stewardess reminded Ali to fasten his seat belt. “Superman don’t need no seat belt”, replied Ali. “Superman don’t need no plane either”, retorted the stewardess.

In (1), the rhetorical effect of the flight attendant’s alleged retort is two-fold: on the one hand, it consists in ridiculing Ali by mocking him,7 which can be counted as a pathotic move, as it is destined to trigger the audience’s amusement and enjoyment at the expense of Ali. However, we should not underestimate the fact that the flight attendant’s response actually builds on Ali’s failure to draw relevant entailments of his own utterance, which is a reason to consider that the logotic dimension is present in (1) as well—indeed, as Superman can fly, he does in principle not need to embark on a plane to go somewhere. Finally, the flight attendant appears as a competent conversationalist, due to her ability to tip the rhetorical scales in her favour—an ethotic move if there is one. We see, thus, that one and the same utterance can simultaneously draw on all of Aristotle’s three rhetorical dimensions, with the consequence that the classification of the example is problematic. Even if we could in principle consider that all three dimensions do co-exist and contribute to the rhetorical appeal of the flight attendant’s utterance, which of the three is prevalent? How can we determine whether one is more fundamental to the rhetorical effect than the others? These are questions that the traditional Aristotelian account cannot answer.

Our second example comes from Sergiy Kyslytsya’s address at the UN on 24 February 2022, right after Russia’s invasion of Ukraine. The Ukrainian ambassador at the UN concluded his call to stop the war by addressing the president of the UN Security Council, who happened to be the Russian ambassador, in the following way:

2.

There is no purgatory for war criminals. They go straight to hell, ambassador.8

Beyond the blatant pathotic appeal of the example, with this implicit personal attack against Russian leadership, the Ukrainian ambassador invites a reaction from his target—which comes under the form of a clarification (“(…) we are not being aggressive against the Ukrainian people but against the junta that is in power in Kiev”). Yet, what are the conversational options of his interlocutor? In principle, he could address the attack by trying to counter it or by acknowledging it (which is of course unlikely in such a context), or he could ignore it, for instance by opting out or evading. Coming back to Kyslytsya’s attack, would we say that he has been persuasive in convincing viewers and audience members that Russian leadership members can be identified as war criminals? This is unclear, to say the least, but, crucially, this does not impact the rhetorical effectiveness of (2): it is indeed intuitively very clear that the Russian ambassador’s choice is much less appealing, from a rhetorical perspective, than the Ukrainian ambassador’s implicit attack—and thus, that there seems to be something rhetorical that is neither ethotic, nor logotic or pathotic.

The second of these issues (described in (ii) above), therefore suggests that other rhetorical effects, which are perhaps not systematically reducible to the categories of ethos, logos, or pathos, might be relevant to grasp the full measure of what the relatively vague notion of rhetorical effectiveness might denote. Indeed, an argument may be said to be rhetorically effective in a number of different scenarios, which are intuitively easy to identify, but which have not so far been systematically collected.9 For instance, an argument may be effective when (and this list is not exhaustive):

• its content is accepted as true or likely to be true by its audience
• it distracts the addressee from its argumentative flaws
• it fools the addressee into accepting the flawed argument
• it distracts the addressee from the course of the conversation and allows the speaker to change topic
• it allows the speaker to (plausibly) deny what they meant
• it allows the speaker to avoid reputational sanctions
• it makes the audience perceive the speaker as a trustworthy individual
• it silences the opponent
• it is able to weaken the reliability of opposing arguments
• forces an opponent to defend themselves
• …

While it seems straightforward to find ethotic, logotic and pathetic effects in this open list, not all of its members can be reduced to one of these three dimensions—the fact that an opponent is left speechless is arguably neither an ethotic phenomenon, nor a logotic or a pathetic one (even though it could be the consequence of some emotional state). Additionally, there are reasons to consider that some of the items on the list are related to each other—for instance, avoiding reputational sanctions can follow from the possibility of plausibly deny problematic contents, on behalf of the speaker—which might suggest that a more systematic classification may be possible. In fact, what this cursory look at the items on this list suggests is that some effects have to do with the conversational dynamics (i.e., what moves are permissible or likely in reaction to given moves), while some others have to do with classical persuasive dimensions (ethos, logos, pathos), and others with the sheer management of information flow. It remains that all of the above seem to denote different rhetorical advantages a speaker can hope to gain through their contribution in an argumentative exchange. For the purposes of this paper, we will consider that the rhetorical effects of a given argumentative move can concern the speaker, the audience, the message and, crucially, the conversational dynamics (see Section 3.2 and Section 6).

3. Pragmatic Insights into Rhetorical Effects

3.1. Some Pragmatic Insights into Ethos, Logos and Pathos

As seen above, example (2) seems to suggest that rhetorical effectiveness should perhaps not be a notion exclusively focused on persuasion, as non-persuasive, yet rhetorical, effects seem to exist and have already been studied. This is why we will propose to construe rhetorical effectiveness along the lines of perlocutionary acts, as defined in speech act theory (Austin 1962; Searle 1969; Oswald 2023). Should rhetorical effectiveness indeed be irreducible to either ethos, logos or pathos, then perhaps this should be taken as an indication that the very notion stands to gain from being tackled through a different framework. Based on recent work at the interface between pragmatics and argumentation, we argue that pragmatics can provide such a framework.

In his own account of ethos, Herman (2022, this volume) challenges classical accounts and argues that whatever audiences end up putting into the category of ethos is inferentially arrived at, and that a unified framework to account for these mechanisms should be a cognitive pragmatic one like relevance theory (Sperber and Wilson 1986, 1995; Wilson and Sperber 2012). Through a systematic survey of existing ethotic accounts, Herman shows that classical approaches fail to achieve explanatory adequacy, as they are limited to the identification of ethotic types and do not allow us to understand how audiences construct them. Moreover, with a relevance-theoretic analysis of a case study, Herman then convincingly shows how an investigation framed in terms of considerations of relevance can provide a cognitively plausible account of the actual ethotic effects of verbal utterances in their contexts of occurrence. Thus, cognitive pragmatics seems to be an apt framework to study rhetorical effects related to the management of speaker ethos.10

Recent work on relevance and emotions (Wharton et al. 2021) also seems to highlight a connection between cognitive pragmatic frameworks and the rhetorical dimension of pathos. In this work, the authors try to bridge the gap between the affective sciences and cognitive pragmatics by suggesting that the notions of relevance each field operates under have much in common, each being a subtype of goal relevance (Wharton et al. 2021, pp. 264–67). The gist of the discussion, which is relevant to the purpose of this paper, lies in the idea that accounting for communicative relevance is no different from accounting for emotional relevance: in other words, the mechanisms involved in figuring out what is emotionally relevant to an individual are likely to be similar to the mechanisms involved in figuring out what is communicatively relevant to an individual. From this, it follows that a relevance-theoretic framework is in principle equipped to account for emotional responses to communicative stimuli—i.e., for pathotic effects.

Logos has also been approached, although perhaps not in those terms, within cognitive pragmatics, for instance by Assimakopoulos (2021), Maillat and Oswald (2009, 2011), or Oswald (2016), who are interested in the persuasiveness of public messages in terms of their contents and ideological ramifications. To this effect, these authors provide an account of how the cognitive architecture of the mind, as described by relevance theory, makes it possible for messages to enter an addressee’s cognitive environment—or, in other words, how the content of a message ends up being accepted by its audience. Oswald (2016), for instance, considers that a given argument is said to be persuasive when it has withstood critical testing or when no counter-evidence has been brought forward against it (assuming its content is compatible with the addressee’s cognitive environment). The advantage of this construal of rhetorical effectiveness in terms of logos lies in its identification of specific conditions under which argumentative evaluation mechanisms unfold—namely epistemic strength and accessibility of information, which resonate with relevance theory’s extent conditions of relevance (Oswald 2016, pp. 264–68). Assimakopoulos (2021) adopts a similar viewpoint in his account of ideological effects that arise during interpretation, by extending the framework of relevance theory in order to cover ideological effects that are not necessarily intended by the speaker. The idea underlying both accounts is that verbal messages are likely to privilege—and, crucially, to constrain—the derivation of specific (ideological) representations. A relevance-theoretic analysis can then explain why these (ideological) representations are likely to be drawn by audiences, which, in the end, amounts to explaining their rhetorical effectiveness in terms of logos.

What this cursory look at some recent developments in cognitive pragmatics shows is that this particular strand of pragmatic research has much to offer to rhetorical scholarship. This idea is far from new: more than 20 years ago, Dascal and Gross (1999) attempted to provide a cognitive reading of Aristotelian rhetoric by describing persuasive interactions in terms of Gricean communicative causality. Over the years, the rhetorical account of Chris Tindale (1992, 2015, 2022, this volume) has consistently incorporated some key notions of both Gricean and relevance theory pragmatics, such as the notions of cognitive environment and of relevance, with the purpose of illuminating rhetorical phenomena. This work has, among other things, shed light on how the relevance of premises might vary from one audience to the next, on how irrelevance can turn into relevance, and shown how a pragmatic account of relevance can help in argumentative reconstructions. The thrust of Herman and Oswald’s (2014) edited volume on rhetoric and cognition was precisely meant to illustrate that contemporary pragmatics has much to offer to both classical and modern rhetoric, in that their respective objects of study overlap to a great extent.

The overall conclusion, if there is one, of this brief overview is that the study of meaning comprehension, in verbal communicative exchanges, has much to offer to the study of persuasion—that is, pragmatics has been shown to have a great deal to offer to rhetoric.

3.2. Rhetorical Effects Are Perlocutionary Effects That Can Be Studied Experimentally

A closer look at our earlier, albeit superficial, list of possible rhetorical effects reveals that all of them can ultimately be thought of as effects on the addressee. Indeed, ethotic effects, even if they are about the speaker’s perceived characteristics (competence, benevolence, integrity, etc.), are perceptions or representations that emerge as the result of the audience’s inferential work (Herman 2022, this volume); the same goes for logotic effects, which arise based on the audience’s processing of the verbal material contained in the speaker’s utterance, and for pathotic effects, which hinge upon the audience’s affective responses in the communicative event. The same also goes for effects on the conversational dynamics of the communicative event: these are audience-driven, insofar as they denote the range of possible follow-ups to a speaker’s utterance, including argumentative moves like refutation, challenge, etc. All in all, these can be considered as consequences that the comprehension of a message has had on conversational participants and on the conversation itself; as such, it makes sense to construe them as perlocutions, consistently with the Austinian definition, according to which perlocutionary acts “produce certain consequential effects upon the feelings, thoughts, or actions of the audience, or of the speaker, or of other persons” (Austin 1962, p. 101). The category of perlocutionary speech acts can thus be said to make some room for the inclusion of rhetorical concerns in pragmatic models. Further evidence of this connection can be found in the way Grice formulated his Cooperative Principle and its associated maxims, assumed to regulate communication at large: “I have stated my maxims as if this purpose were a maximally effective exchange of information; this specification is, of course, too narrow, and the scheme needs to be generalized to allow for such general purposes as influencing of directing the actions of others” (Grice 1975, p. 47). What this shows is that effects that emerge in conversation alongside comprehension, but are linked to the latter, have been recognised as relevant in the characterisation of human communication, and that the field of pragmatics, specifically, already has some theoretical categories to account for this. Here we suggest that the rhetorical effects that we are interested in can be accounted for as perlocutions, consistently with classical pragmatic theory (see also Oswald 2023, Section 2.3, for a discussion).

While perlocutionary effects have tended to be neglected by classical pragmatic research, and to the extent that their study has been recognised as controversial, to say the least (Gu 1993; Marcu 2000), they nonetheless remain effects that are intimately linked to the comprehension of locutionary and illocutionary speech acts—i.e., there is a strong relationship between the illocutionary act and the perlocutionary act, whether the latter is loosely denoted as a way of “influencing or directing the actions of others” or, more specifically, as a persuasive act (Marcu 2000). On this last point, it seems to us opportune to evoke the potentially very fruitful input that the field of experimental pragmatics (Noveck 2018; Noveck and Spotorno 2004) could provide in the study of perlocution. Initially developed as an empirical endeavour to lend support to theoretical models of meaning, experimental pragmatics has already made incursions into the study of some perlocutionary effects, as mediated by linguistic resources—see for instance Mazzarella et al.’s (2018) study of effects on trust and reputation based on explicit vs. implicit modulation of speaker commitment, or Weissman and Terkourafi’s (2019) empirical study of whether false implicatures qualify as lies. While experimental pragmatics is primarily focused on the study of meaning in context, it can also be recruited to provide better descriptions and explanations of perlocutionary effects and of the circumstances in which they are likely to be triggered. In other terms, the methods of experimental pragmatics are relevant and ready to be implemented to study the rhetorical effects of linguistic resources in argumentation.

Based on what precedes, we therefore circumscribe the types of rhetorical effects of argumentation we are interested in here into four categories, which we identify based on whom or what the message affects: (i) effects on the speaker (linked to ethos), (ii) effects on the audience (linked to pathos), (iii) effects on the message (linked to logos), and (iv) effects on the conversational dynamics (see also Table 1 below). Our next step, based on the salient feature of examples (1) and (2) above, is to reflect on the rhetorical effectiveness of IM as a communicative resource for meaning-making. In what follows, we develop some arguments that plead for such a study.

4. The Rhetorical Appeal of Implicit Meaning

Having established how rhetoric can benefit from pragmatic insights, we now narrow down our focus to discuss some rhetorical specificities of IM, which, we argue, constitutes a pragmatic resource of particular interest for rhetoric. In this sub-section, we thus develop four lines of thinking that highlight the appeal of IM in communication and how the latter is likely to be exploited rhetorically.

4.1. Evolutionary Bases for the Existence and Prevalence of IM in Human Communication

As Reboul (2017) suggests, implicit communication does not seem to be the most efficient means of linguistic communication, both at the production and at the reception end of the process. For speakers, choosing to convey a message implicitly might be riskier and more open to misunderstandings than choosing to do so explicitly, as part of the meaning is not linguistically given, but left to the hearer to infer. Moreover, although IM may be less cognitively costly for speakers, this is at the expense of hearers, who are required to perform additional inferential work. From the perspective of politeness, despite the claim, drawn from early accounts such as Brown and Levinson’s (1987) or Leech’s (1983, 2014), that indirectness (and so IM) is more polite than explicitness, it has also been suggested by Pinker (2007, p. 442) that indirect requests “can make the speaker sound devious and manipulative”, since they aim at protecting speakers rather than hearers from unpleasant contents. Thus, IM, at first glance, does not seem to be an advantageous option, in terms of both its processing when it comes to securing a successful exchange of meaning and its impact on the social relationship between speakers and hearers.

One of the criteria for successful manipulation—as for deception—is that speakers must prevent their targets from detecting their manipulative intention (Oswald 2010). If this condition is met, the intention is hidden and can bypass the vigilance of its audience. If, however, the manipulative intention becomes public, not only will the manipulative attempt presumably fail, but negative reputational consequences might ensue for speakers. Interestingly, IM can mitigate this risk: speakers can cancel or deny having meant an implicature without this generating a contradiction in content—even if specific features of different types of IM might merely make denials merely possible instead of plausible (for instance, generalised conversational implicatures might be less plausibly denied than weak implicatures). Moreover, speakers can shift the responsibility of an incorrect interpretation of their utterance to hearers. IM thus offers a ‘way out’ (Pinker 2007; Pinker et al. 2008; Reboul 2017) because its associated content is conveyed off the record. In Reboul’s terms, IM allows “speakers to hide their ultimate intentions and to deny having them, while communicating content that is aimed at changing the addressee’s behaviour in such a way that these ultimate intentions will be satisfied” (2017, p. 206). IM, which many times appears in off-record communication (Gibbs 1999) is thus rhetorically appealing: it may be used to dissimulate or deny a speaker’s manipulative intentions, to allow speakers to avoid taking responsibility for a message conveyed and potentially shift that responsibility. All those benefits are related to the preservation of speakers’ ethos.

From this contrasted view, and given the stability of IM in human communication, it would therefore seem that its benefits outweigh its disadvantages. Thus, probably partly because of (the rhetorical advantages involved in the dissimulation of a speaker’s ultimate intentions and its positive impact on ethos, IM represents in principle a rhetorically attractive resource for arguers who seek to influence others.

4.2. Processing Advantages of IM for Rhetorical Effectiveness

The rhetorical appeal of IM can also be grounded in arguments about the way IM is cognitively processed. In particular, two facts of human psychology seem to plead for this claim: the extensively-discussed confirmation bias and myside bias (see Mercier 2016 for an overview) and our so-called cognitive optimism (Sperber et al. 1995).

The confirmation bias, which denotes our natural propensity to privilege confirmatory over falsificatory reasoning, has been famously studied through the Wason Selection Task (Wason 1968). More specifically, the confirmation bias has been defined as “the seeking or interpreting of evidence in ways that are partial to existing beliefs, expectations, or a hypothesis in hand” (Nickerson 1998, p. 175). While the reality of the confirmation bias has been called into question quite recently by Mercier, who proposes instead that the selection task gives evidence of the myside bias, namely the “tendency to find arguments that support one’s own views” (Mercier 2016, pp. 99–100), there seems to be experimental evidence that we have a strong tendency to stick to what we believe or assume to be the case in the face of contradiction or in reasoning tasks. In their own selection task study, Sperber et al. (1995) found that subjects set their mind on a hypothetical rule according to its accessibility and that they maintain it when the latter satisfies the expectations of relevance generated by the instructions of the task. However, more crucially for our purposes, they also found that subjects are not likely to re-check the output of their own inferences and dubbed this property of human cognition cognitive optimism.

Given that the comprehension of IM is a heavily inferential process (Grice 1989; Sperber and Wilson 1995)—insofar as a large portion of the meaning conveyed is not linguistically encoded but needs to be contextually worked out—it stands to reason that its derivation is also affected by the confirmation bias, or the myside bias, and that it is a cognitively optimistic process. That is, hearers are not only likely to consider that their interpretations of IM are accurate, but they are also less likely to submit them to critical scrutiny. As noted by Lombardi Vallauri and Masia, “discussing something that you have created yourself is much lesser a temptation than discussing something stated by someone else” (2014, p. 166)11. Accordingly, IM is less likely to trigger critical reactions. This is particularly important to consider in the context of argumentation (Oswald 2020): as critical scrutiny seems to be lower, or less likely to be thorough, for IM, it stands to reason that arguers might take advantage of it to communicate problematic contents, and to use it for rhetorical gain. In other words, IM is also rhetorically advantageous because of the way in which it is cognitively processed.

4.3. Social Advantages of IM for Rhetorical Effectiveness

In social interactions, the use of IM offers many advantages, both in dealing with opponents in a discussion or debate and in addressing an audience. In a study focusing on dyadic communicative settings (consisting of two interlocutors only) such as parent-child interactions, Bova (2011) investigated how parents use implicitness as an argumentative strategy with their children to achieve their goals during dinner conversations. He found that implicitness serves two purposes: it reduces the effectiveness of the child’s standpoint by persuading them to retract it and it reinforces the strength of the parent’s standpoint by anticipating possible consequences of the child’s behaviour. Through implicitness, parents are also able to reduce the undesirable effects of explicit orders so as to facilitate acceptance. This last point, regarding acceptance, is incidentally echoed in the work of Lombardi Vallauri (2018), who claims that “[e]xplicit assertions reveal the intention of the source to convince the recipient, which triggers our willingness to meet it with critical judgment, rejecting any message that is not completely truthful.” (p. 1, our translation). Under this lens, IM would presumably fare better than explicit meaning in terms of its rhetorical effectiveness.

Beyond dyadic configurations, the impact of IM may extend to triadic and multi-party conversational dynamics involving speakers and an audience who is not directly involved in the argumentative exchange as a participant. An empirical study on the modulation of commitment through explicit and implicit meaning has found that speakers who implied information that turns out to be false risk lower reputational and trust sanctions than speakers who conveyed the same information explicitly (through assertion), the latter being perceived as being more committed to the false information than the former (Mazzarella et al. 2018). The authors of this study explain that the difference of perceived speaker commitment depends on the type of content used by the speaker: people attribute lower commitment when a statement is implicit (i.e., implied) than when a statement is partially (i.e., presupposed) or fully explicit (i.e., asserted). This suggests that the use of IM is somewhat safer, as speakers who turn out to misinform are perceived as less at fault and receive lower reputational sanctions for implying the problematic statement than those who conveyed it more explicitly. This resonates with a previous study which found that greater commitment to the expressed proposition incurs greater reputational costs (Vullioud et al. 2017).12

Incidentally, further empirical data for these claims are now available. A recent corpus study of political speeches, which found that challengeable content types such as attacks and self-praises are more likely to be encoded implicitly as presuppositions, while less challengeable ones are more likely to be asserted (Masia 2020), resonates with the claims made here about the preference speakers would have for IM when it comes to conveying sensitive content that could backfire. Studies of propaganda such as Stanley’s (2015) go in the same direction in their observation that IM is regularly used to distract hearers or readers from the ideologically problematic features of the message, which are embedded under implicit contents.13 In conjunction with the cognitive and evolutionary advantages previously mentioned, it therefore seems that the use of IM is also rhetorically appealing in terms of its social advantages, as it may be used to shield speakers from reputational sanctions and to lower their responsibility for conveying problematic or uncertain statements. Moreover, in addition to ethotic benefits, IM has logotic benefits: if IM is indeed likely to reduce the strength of the opposition’s arguments (or even to discourage their presence altogether), then strategic speakers have much to gain in using it in contexts of argumentative disagreement. These are features that may come in handy in argumentative contexts where stakes can be high for participants.

4.4. Documented Impact of IM on Decision-Making

A fourth line of thought to support the idea that IM is rhetorically appealing has to do with its concrete, practical, impact on decision-making tasks, which has been empirically established in the study of different types of IM.

In a study on the influence of racial cues which do not mention race explicitly but can allude to it implicitly, in the vein of dogwhistles (Saul 2018; Witten 2014), Valentino et al. (2002) found that implicit racial cues embedded in political advertisements, especially those that reinforce negative racial stereotypes, can indeed prime racial attitudes, and thereby influence the outcome of political decision making. In a study on code words and dogwhistles, Hurwitz and Peffley (2005) showed that in comparison to ‘violent criminals’, the implicitly negative racist connotation of the noun phrase ‘violent inner-city criminals’ was sufficient to influence participants to opt for more repressive policies toward violent criminals. Although its implicit/explicit nature is still debated within pragmatics (see e.g., Carston and Hall 2012; Davis 2019; Tendahl 2009; Wilson 2011), metaphor also has a significant measure of implicitness in the contents it conveys; Thibodeau and Boroditsky (2015) have studied its effect on decision-making processes related to anti-crime measures. Interestingly, they found that a text describing crime as a beast ravaging the city was more likely to prompt participants to opt for repressive policies (e.g., an increase in street patrols) than for preventive policies. By contrast, a description of crime metaphorically framed as a virus was more likely to prompt preventive measures (e.g., reforms of educational practices, creation of after-school programmes) than repressive ones. These three studies thus converge towards the idea that IM is, in fact, a powerful resource to influence actual decision-making processes.

Perhaps some further elaboration on whom is meant to be convinced is relevant here; indeed, in political debates, for instance, debaters rarely aim to convince their opponents. Although proponents and opponents are the main parties that are active in the debate, persuasion is usually aimed at the audience, who is ultimately the participant with the most active role in persuasion: its members are responsible for inferring IM and for deciding on new policies through their vote. As evidenced by the studies described above, a linguistic subtlety—even an extra word—conveying IM can impact policymaking. Worse, even: audience members need not be aware of it at all for it to have an effect. This crucially calls for a better understanding of the problematic usages of IM. We argue, in what follows, that this is likely to happen once we better understand which rhetorical pole different types of IM are likely to affect. To cater for this need, in the next section we offer a characterisation of different types of IM that have been described in classical semantic and pragmatic scholarship in terms of their potential rhetorical significance in argumentative discourse.

5. Some Pragmatic Resources for IM and Their Associated Rhetorical Effects

We have seen that IM, owing to many of its features, is likely to be used for its rhetorical appeal. However, the story should be broken down more specifically, to the extent that not all types of IM share the same features. IM is, in a minimal sense, meaning that is conveyed but that is not encoded in the linguistic material that the speaker has used in their utterance. We will accordingly consider, quite loosely, that IM is meaning that is overtly conveyed, but not linguistically articulated, in the vein of Allott, who defines the central data for pragmatics as “cases in which a speaker, in making an utterance, conveys something more than, or different from, the meaning of the words she uses” (Allott 2010, p. 2). Crucially, there are different kinds of IM, depending on their features, and it is nowadays common knowledge that not all of them behave the same way in communication. Based on this, we hypothesise that not all of them will be conducive to the same perlocutionary effects. In what follows we first survey extant characterisations of some types of IM in order to highlight their rhetorical potential, which we then link to their likelihood to trigger specific rhetorical effects.

Ever since Grice’s distinction between what is said and what is meant, the interface between semantics and pragmatics has been explored with an eye to clarifying the boundaries between both domains of meaning. While we do not wish here to contribute to that particular debate,14 we highlight that this line of research has yielded a whole range of different categories of IM. For the purposes of singling out general correspondences between types of IM and types of rhetorical effects, we decided to focus both on classical categories of IM like implicature and presupposition, and on two covert speech acts which have been discussed in terms of their strategic affordances in discourse, namely insinuation and dogwhistles. Below, we describe each of them and, based on these descriptions, theoretically justify which type of rhetorical effect they could be assumed to fulfil in argumentative discussions.

5.1. Implicature

In Grice’s framework, the term implicature covers components of speaker meaning that are not said (i.e., they are not linguistically articulated) and that are left for the hearer to be inferred—while still being speaker-intended. Grice distinguished conventional from conversational implicatures based on the way they can be drawn: conventional implicatures are triggered by the use of certain expressions or words that carry them, and thus they remain context-independent. To take an example, in (3), the meaning of the word ‘therefore’ conventionally implies (4), which acts as the major premise of the inference (Grice 1975, p. 44):

3.

He is an Englishman; he is, therefore, brave.

4.

All Englishmen are brave.

To the extent that conventional implicatures seem intimately tied to linguistic forms, doubts about their truly implicit status have been raised (see e.g., Bach 1999), which is why will not examine them further here.

Next to conventional implicatures, Grice distinguished two kinds of conversational implicatures: generalised conversational implicatures (GCIs) and personalised conversational implicatures (PCIs). GCIs do not depend on particular features of the context, unlike PCIs. In the absence of any special circumstances, a GCI of saying an utterance P is one that is normally conveyed by saying that P. As for PCIs, they are not automatically triggered: they are context-dependent and their derivation, in the Gricean model, requires the operation of the conversational maxims, in the sense that they arise based on the speaker’s observance or ostensive flouting of a maxim (and on the addressee’s recognition thereof).15 For instance, as an answer to the question “Does Nina care for the environment?”, an utterance like (5) can convey the GCI in (6) and simultaneously the PCI in (7): for the latter, the speaker is manifestly flouting the maxim of relevance, as the answer is not a simple yes/no answer which the format of the question was calling for:

5.

Nina has two cars.

6.

Nina has exactly two cars.

7.

Nina does not care about the environment.

Beyond types of implicatures, some of their properties are important for our purposes here. As discussed by Grice, conversational implicatures are calculable and defeasible.16 These, we contend, are relevant to the success of strategic exchanges in rhetorical terms.

The calculability of implicatures relates to the fact that they are not linguistically given by the speaker: they require the addressee to draw on their background knowledge and on the context, meaning that the responsibility for their derivation heavily rests on the latter’s shoulders. Consequently, there is a part of subjectivity in their derivation, and this makes them prone to being misunderstood, since an addressee might draw an unintended interpretation by failing to mobilise the contextual assumptions that the speaker meant them to mobilise. In turn, this means that a speaker who implies something can in principle deny the addressee’s interpretation by alleging that the latter has followed an inferential path that was not the one the speaker intended the addressee to follow. Moreover, as defeasible (or cancellable) inferences, conversational implicatures are inferences to the best explanation (Allott 2010; Geurts 2010): they are non-demonstrative inferences the conclusion of which may have to be given up in light of new evidence. Defeasibility, therefore, is the key to deniability: it is because an implicature can be cancelled by adducing new evidence that a speaker can effectively deny the interpretation that the addressee has reached. If we now look at defeasibility from the perspective of the addressee, there is a connection with the notion of cognitive optimism discussed above (see Section 4.2): given that the derivation of an implicature is a pragmatic inference, its conclusion not only might be biased by the addressee’s own expectations, but it may also fail to be critically questioned by them. Strategic speakers might thus rhetorically take advantage of this property to shield their ethos, by circumventing the responsibility of getting some problematic meanings across and, thus, by benefitting from fewer reputational sanctions.17

As defined by Sperber and Wilson (1995), implicatures come in different degrees of strength. The strength of an implicature depends on the degree of evidence speakers provide for their conveyed intention. When the possibilities of interpretation of an utterance become wider because either the stimulus or the context for interpretation is ambiguous or vague, the implicature is said to be weak. This makes it possible for a speaker to deny weak implicatures more easily than strong implicatures, the derivation of which is constrained by the increased accessibility and univocity of the intended context of interpretation. Crucially, thus, speakers can not only deny a weak implicature but also suggest that they meant an alternative implicature with almost identical relevance. Unlike strong implicatures, which make one interpretation more salient and mutually manifest than others, weak implicatures offer a greater degree of deniability which is coupled with the possibility of suggesting plausible alternative interpretations (see also Mazzarella 2021 for a discussion of the link between deniability and strength of communication).

It is not difficult to see how the deniability affordances of IM can be rhetorically exploited: by using IM (and by taking advantage of its associated deniability if need be), speakers may escape the sanctions they would face for explicitly asserting problematic contents (e.g., offensive, dubious, or socially delicate contents), and being called out for it. They could also take advantage of deniability to shift the responsibility for an interpretation towards the addressee, who is then put in the position of someone who either misunderstood or intentionally misrepresented their words—which, from a rhetorical perspective, is an effect on the conversational dynamics of the exchange, as it might prompt for defence or counter-attack on behalf of the addressee.18 At the same time, because of the way they are processed, implicatures are rhetorically appealing in terms of logos, as they (i) decrease the chances of their content being subjected to critical evaluation (see Section 4.2), and (ii) are more likely to be persuasive than their explicit counterparts for their processing features.

5.2. Presupposition

A presupposition is a type of implicit content (i.e., not linguistically articulated by the speaker) “whose truth is taken for granted in the utterance of the sentence. Its main function is to act as a precondition of some sort for the appropriate use of that sentence.” (Huang 2007, p. 65). Presuppositions represent essential background information that is not at-issue; they are neither part of the speaker’s intention, nor of the notional content of the message (Lombardi Vallauri and Masia 2014). Throughout the years, a semantic and a pragmatic account of presupposition have co-existed (see Saeed 2009, chap. 4.5): the former sees presupposition as being tied to the information a given sentence packages, while the former sees it as having to do with the inferences that conversational participants may draw on the basis of what the speaker said. Semantic presuppositions have been discussed at length in classical accounts; as for pragmatic, discursive and persuasive presuppositions, they have received attention in more recent research.

From a semantic perspective, the fact that a presupposition needs to be true in order for the sentence that presupposes it to make sense is taken as an indication that presuppositions are used to express formal relations between statements (Saeed 2009). In a nutshell, presuppositions semantically point to information that is assumed to belong to the common ground—while new information is generally not presupposed. This in turn explains that presuppositions usually remain true when the presupposing sentence is negated (‘Laszlo’s brother is tall’ presupposes that Laszlo has a brother, just as the negated version of the sentence, ‘Laszlo’s brother is not tall’, does). From a pragmatic perspective, presuppositions have been assumed to help manage the shared background (or informational common ground) between conversational participants, so that information is organised efficiently and for maximum clarity (Stalnaker 1974). This view also puts forward the fact that presuppositions can be accommodated: that is, a speaker who presupposes a piece of information that the addressee does not know can reasonably count on the latter’s accepting it for the purposes of successful communication (Lewis 1979)—if we tell you that Laszlo’s brother ran away, and you did not now Laszlo had a brother, you will presumably accommodate the existential presupposition to understand the utterance and automatically assume that Laszlo, indeed, has a brother. Discursive presuppositions (de Saussure 2013) behave somewhat similarly, but instead of being truth-conditional elements of meaning, they are relevance-conditional elements of meaning: they are preconditions for the meaningful relevance of statements. For instance, a sign reading ‘No guns allowed’ discursively presupposes that someone could find it relevant to wonder whether guns are allowed in the vicinity (assuming, for example, that the sign is at the entrance of the visiting area of a prison), that guns are dangerous/undesirable if carried in the vicinity, that guns may be allowed in other areas, etc. (adapted from de Saussure 2013, p. 179).

From a rhetorical perspective, presuppositions are appealing strategies in that they allow to point out that some information is (or should be) commonly shared, even when said information has not been verbally or linguistically encoded. Thus, the fact that the information is shared is not up for discussion, which may also decrease the chances of critical evaluation of its content. Since we have accepted this piece of information, why should we bother re-checking it? This can be particularly advantageous for manipulative purposes, for instance when speakers wish to convey doubtful contents without risking too much exposure—and it seems presupposition is able to do just that (Lombardi Vallauri 2016, 2021; Masia 2021). Thus, in terms of logos, presupposition might be well-suited to trigger two possibly related types of rhetorical effects: on the one hand, it could force the acceptance of doubtful contents, and on the other, because it conveys contents that are not at-issue (that is, contents which are not focal in establishing the contextual relevance of the utterance, but rather preconditions for it), it could decrease the chances of critical evaluation of doubtful information.

5.3. Back-Door Speech Acts

The survey of rhetorical effects conducted so far around implicature and presupposition is quite rudimentary. With back-door speech acts (Langton 2018; Witek 2021), the picture might get slightly denser, mainly because of their covert nature and polyfunctionality. Incidentally, it should be noted that their characterisation oftentimes acknowledges, albeit tacitly, their rhetorical appeal; this can be seen in Langton’s brief description of back-door speech acts as “low profile speech acts, enabled by presuppositions and their ilk, that tend to win by default” (Langton 2018, p. 146, our emphasis). Here we discuss only two examples (insinuation and dogwhistles) for reasons of space, with the understanding that the line of argument defended would extend to other discursive phenomena falling into the category of back-door (or covert) speech acts.

5.3.1. Insinuation

Insinuations and innuendoes are classically defined as “non-overt intentional negative ascription[s], whether true or false, usually in the form of an implicature, which [are] understood as a charge or accusation against what is, for the most part, a non-present party” (Bell 1997, p. 36). While there is a classical and a modern take on insinuation, which differ in whether they consider that insinuated contents are meant to be recognised by addressees as meant to be recognised or not (i.e., whether they are R-intended, in Grice’s sense, see Bach and Harnish (1979, chap. I.6)), a stable feature of insinuations is that they are used to ascribe negative descriptions to their targets while allowing the speaker to plausibly deny having meant the negative ascription (Oswald 2022). In other words, the key feature of insinuation is that it is plausibly deniable, which means that any insinuated content is accompanied by an alternative plausible meaning that is both contextually relevant and socially innocuous. For instance, imagine a situation in which both Winston and Nina brought cake to Laszlo’s party; imagine then that Laszlo joins them and utters (8)

8.

Nina, your cake was absolutely fabulous!

in Winston’s presence. It would be legitimate, for instance if we knew that Laszlo secretly dislikes Winston and never lets an opportunity to disparage him pass, to infer that Laszlo meant (9):

9.

Winston is a bad cook

In other words, (8) can be used to insinuate (9). Crucially, if Winston retorts by calling the insinuation into question (‘Do you mean I’m a bad cook?’), Laszlo can always fall back on an interpretation that makes the literal meaning of his compliment to Nina contextually relevant, for instance by saying ‘No, I didn’t mean that; I just told her that because I know she just started and felt she can use the encouragement to keep baking cakes.’ Additionally, insinuations have been characterised as leaving reputational stains which, interestingly, are believed to remain even after the negative ascription has been denied (Bell 1997; Fraser 2001).

Insinuation is rhetorically appealing, and in a quite obvious way, within at least two rhetorical dimensions: ethos and logos. In terms of ethos, an insinuation usually takes the form of a plausibly deniable implicature; that is, the insinuator can disparage their target and still claim that the disparaging comment was not meant to be inferred and/or that they do not even believe that the disparaging comment applies (as in ‘Of course I don’t believe you’re a bad cook, Winston’). This can be seen as a way of shielding the speaker’s reputation, by denying malevolent communicative intentions and appearing to be non-committed to the latter. The ethotic advantage of insinuation therefore has to do with its shielding potential: the speaker can both deny malevolent intent and shift responsibility to others for having drawn the negative ascription. In terms of logos, the fact that insinuation is most of the time conveyed under the form of an implicature (a PCI in particular), i.e., through an inference the addressee is responsible for, makes the content less likely to be questioned (see Section 4.2). A straightforward consequence of this is that the content of the insinuation is more likely to be accepted as reflecting a state of affairs. We thus expect insinuations to be more persuasive in terms of logos. However, there is more, on at least two counts.

First, in terms of conversational dynamics, insinuation is likely to impose constraints on the responses the attacked party might have at their disposal. Because insinuation can function in argumentation as an ad hominem attack, the target is likely to either counterattack or to defend themselves. These argumentative moves are thus foregrounded as preferred, and they might in turn trigger a host of further audience attitudes that are decisive for determining which party has the advantage. For instance, seeing someone who has been attacked through insinuation defend themselves might give the impression that there was some truth in the attack, otherwise they would not even bother defending themselves; alternatively, if the target counterattacks with an exaggerated face threat towards the insinuator, they run the risk of coming across negatively. Additionally, in case the content of insinuation is called into question and subsequently denied by the insinuator, the target who has drawn the negative ascription runs the risk of doubling down on it in the eyes of the audience by addressing it. Thus, instead of having to justify the negative ascription, the insinuator can ‘argumentatively’ lay back and hope for their target to fail in their defence. In the process, an additional effect could be that the course of the conversation derails: the insinuator has thus much to gain, rhetorically speaking, from insinuating, because prompting a defence would likely affect the conversational flow and facilitate topic switches.

Second, the fact that insinuations are used to convey negative ascriptions can be rhetorically significant in terms of pathos. Many times, insinuations convey humorous contents, mainly because they appear as clever and cheeky off-record comments on someone’s misfortunes or shortcomings. Consistently with the main claims of superiority theories of humour (see Attardo 1994; Ritchie 2014), which explain humorous effects as by-products of a situation in which those who experience humour do so because they feel superior to the target of the comment, insinuation may represent a pragmatic resource through which speakers and their addressees might laugh about others. Crucially, thus, humorous insinuations achieve their effect by relying on the audience’s derivation of an implicit representation that simultaneously disparages its target and triggers a feeling of superiority. It consequently stands to reason that IM may be suited to trigger a rhetorical effect of target de-legitimation that arises through an affective response to mockery.

5.3.2. Dogwhistles

Dogwhistles “occur [ ] when a person or group sends a message which contains a second, coded interpretation meant to be understood by a select target person or group” (Witten 2014, p. 2). In politics, this strategy consists in conveying a controversial message to one part of the audience, who gets the coded message but not to another part of the audience, who remains unaware of the intended content (Saul 2018). For instance, to gather votes for a political initiative, a politician might say (10), where family values is understood as meaning Christian values (11) without alienating non-Christians or atheists, who could consider that this sentence denotes commonplace values as in (12).

10.

Thanks to your votes, we will be able to restore the family values you hold dear.

11.

Thanks to your votes, we will be able to restore Christian values (e.g., pro-life views, heterosexual-only marriage) you hold dear.

12.

Thanks to your votes, we will be able to restore the importance of families in society (e.g., existence of family benefits, family holidays).

Dogwhistles are implicit because their meaning, which co-exists with an innocuous or neutral meaning, needs to be inferred, and the fact that they target selected subsets of audiences suggests that one of their main intended rhetorical effects has to do with the relationship between the speaker and the audience—which itself could be counted among pathotic effects. On the one hand, dogwhistles can foster inclusion and strengthen the bond between those who get the dogwhistle and the speaker, as the correct interpretation of the intended meaning is likely to translate into ideological alignment between the speaker and the subset of the audience that gets it, thereby allowing a feeling of belonging to the same community to emerge or to be reinforced. On the other hand, dogwhistles could foster exclusion by portraying the targeted community as irrelevant and undeserving of inclusion in the conversation in progress. Still on the pathotic dimension of discourse, one could claim that because dogwhistles activate stereotypes people have about given communities, they can also trigger emotional responses that are likely to affect the way the discourse is processed in a rhetorically relevant way. Think, for instance, of people who hold racist beliefs and attitudes, and who are exposed to a racial political dogwhistle targeting a specific community: their resentment, if not their anger or fear, may be assumed to drive their response to the message. Another possible rhetorical effect is similar to the effect insinuation can have on ethos: dogwhistles are plausibly deniable, just like insinuations, which is a conversational move that could be used to shield the speaker’s image. In any case, it seems that the main rhetorical appeal of dogwhistles lies in their pathotic and ethotic affordances.

6. Towards a Matrix of Correspondences between IM and Rhetorical Effects

We started this paper by reflecting on the vagueness of the notion of rhetorical effectiveness, notwithstanding Aristotle’s classical account of ethos, logos and pathos. We then tried to justify that the field of pragmatics could represent an interesting and profitable option to tackle the notion and attempt to elaborate more fine-grained descriptions of types of rhetorical effects. We then narrowed down our inquiry to IM, first by justifying its rhetorical appeal and then by discussing the features of different types of IM in terms of the types of rhetorical effects these different types can be taken to be conducive to. Through our analysis, we can now offer a non-exhaustive, yet principled, typology of types of rhetorical effects, as triggered by types of IM (see Table 1 below), that can be experimentally operationalised.

Table 1. Possible rhetorical effects for different types of pragmatic meanings.

IM Types	Possible Rhetorical Effects
	On Speaker	On Audience	On Message	On Conversational Dynamics
Implicature	– Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos) – Speaker is more likely to be successful in denying commitment to an interpretation	– Audience can be held responsible for an interpretation – Audience is less likely to critically evaluate the message – Audience can be prompted to defend or counter-attack – Audience can be distracted from the course of the conversation	– Message is more likely to be compatible with more than one interpretation – Message is less likely to be critically evaluated – Message is more likely to be persuasive	– Shifts in the burden of proof are more likely to be successful in the exchange – If speaker denies the implicature, challenging moves following speaker denial are increasingly likely – Likelihood of change of topic is increased
Presupposition	– Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos)	– Audience is less likely to critically evaluate the message – Audience is more likely to accept dubious or problematic contents when they are presupposed	– Message is less likely to be critically evaluated – Dubious or problematic messages are more likely to be persuasive	– Continuity of the exchange is likely to be preserved – Likelihood of challenging moves is decreased
Insinuation	– Speaker can plausibly deny malevolent intent – Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos) – Speaker is more likely to be successful in denying commitment to disparaging ascription – Speaker is more likely to benefit from decreased argumentative load	– Audience can be distracted from the course of the conversation – Audience can be held responsible for a disparaging interpretation – If audience = target, the risk of reputational stain (delegitimation due to mockery) increases – If audience = target and audience engages in challenge, the risk of justifying the reputational stain increases – If audience = target, the risk of increasing argumentative load increases	– Message is more likely to be compatible with more than one interpretation – Message is less likely to be critically evaluated – Message is more likely to be persuasive	– Likelihood of change of topic increases – Likelihood of challenging moves (e.g., when used as ad hominem) increases
Dogwhistle	– Speaker can plausibly deny malevolent intent – Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos)	– Emotional response more likely to be triggered – Complicit audience feels included in the same community as speaker – Unaware audience is excluded from the community of the speaker and complicit audience	– Message is ambiguous (but only to part of the audience)

Table 1. Possible rhetorical effects for different types of pragmatic meanings.

IM Types	Possible Rhetorical Effects
	On Speaker	On Audience	On Message	On Conversational Dynamics
Implicature	– Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos) – Speaker is more likely to be successful in denying commitment to an interpretation	– Audience can be held responsible for an interpretation – Audience is less likely to critically evaluate the message – Audience can be prompted to defend or counter-attack – Audience can be distracted from the course of the conversation	– Message is more likely to be compatible with more than one interpretation – Message is less likely to be critically evaluated – Message is more likely to be persuasive	– Shifts in the burden of proof are more likely to be successful in the exchange – If speaker denies the implicature, challenging moves following speaker denial are increasingly likely – Likelihood of change of topic is increased
Presupposition	– Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos)	– Audience is less likely to critically evaluate the message – Audience is more likely to accept dubious or problematic contents when they are presupposed	– Message is less likely to be critically evaluated – Dubious or problematic messages are more likely to be persuasive	– Continuity of the exchange is likely to be preserved – Likelihood of challenging moves is decreased
Insinuation	– Speaker can plausibly deny malevolent intent – Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos) – Speaker is more likely to be successful in denying commitment to disparaging ascription – Speaker is more likely to benefit from decreased argumentative load	– Audience can be distracted from the course of the conversation – Audience can be held responsible for a disparaging interpretation – If audience = target, the risk of reputational stain (delegitimation due to mockery) increases – If audience = target and audience engages in challenge, the risk of justifying the reputational stain increases – If audience = target, the risk of increasing argumentative load increases	– Message is more likely to be compatible with more than one interpretation – Message is less likely to be critically evaluated – Message is more likely to be persuasive	– Likelihood of change of topic increases – Likelihood of challenging moves (e.g., when used as ad hominem) increases
Dogwhistle	– Speaker can plausibly deny malevolent intent – Speaker is more likely to successfully avoid reputational sanctions (=shielding ethos)	– Emotional response more likely to be triggered – Complicit audience feels included in the same community as speaker – Unaware audience is excluded from the community of the speaker and complicit audience	– Message is ambiguous (but only to part of the audience)

Table 1 summarises the discussion conducted in Section 5 by giving an overview of the different kinds of rhetorical effects we can expect different types of IM to be likely to generate. A first observation is obviously that there is some overlap among types of rhetorical effects across different types of IM. This is unsurprising, given the similarity of some features across types of IM—for instance, insinuation many times comes under the guise of implicature (Bell 1997; Fraser 2001; Oswald 2022). A second observation is that there seems to be some imbalance in the categories (for instance, no conversational effects are listed for dogwhistles, or more effects on audience than effects on the message are listed for insinuation). This follows partly from the fact that this is but a cursory exploration of rhetorical effects of IM—which calls for more scholarly work—and partly from the fact that different types of IM have different features, which in turn target different aspects of the communicative exchange. A third observation, which is perhaps more of a theoretical order, is that some of these effects seem to be closely related. For instance, for insinuation, the fact that the speaker can plausibly deny malevolent intent (or an intent to disparage) is a direct consequence of the fact that the message is likely to be compatible with more than one interpretation; in turn, the denial might license the attribution of the disparaging interpretation to the audience, who might then feel pressure to challenge this shift in the burden of proof at the conversational level. This means that establishing such a matrix of correspondences between types of IM and their associated rhetorical effects requires careful consideration of the dimension that is rhetorically significant. At the same time, it generates a form of dependency between rhetorical effects, which, even under the classical Aristotelian account, is in fact unsurprising, given the simultaneous operation of ethos, logos, and pathos in rhetorical enterprises. Be that as it may, we hope to have shown with this table that there are quite a few profitable directions to explore at the interface of pragmatics and rhetoric to assess the impact different types of IM can have on a speaker’s rhetorical plans.

7. Conclusions

In this paper, we have tried to justify the peculiar status of IM in terms of its rhetorical appeal and, simultaneously, to promote, at the scholarly level, the integration of pragmatic insights into rhetorical and argumentative inquiry. We have also tried to set the grounds for a systematic investigation of different types of rhetorical effects by defending the hypothesis that different types of IM are likely to be used for different rhetorical purposes. In so doing, we have been led to draw a tentative matrix of correspondences between types of IM and types of rhetorical effects. While we are far from claiming that this settles the issue, we do claim that this is a fruitful starting point for rhetorical inquiry, which takes us to our third, and probably most important, take-home message, which is to be found at the methodological level, namely the idea that all the rhetorical effects mentioned here can be experimentally investigated.

Rhetoric has always been about assessing how different aspects of a communicative encounter are likely to impact the way the speaker’s message is received. Here we have highlighted that this is most of the times mediated by the verbal choices speakers make to convey their meaning. In particular, as we have seen, IM seems to license a host of rhetorical advantages. Now, crucially, this makes research at the pragmatics and argumentation/rhetoric interface extremely prone to experimental research, simply because types of IM can be linguistically operationalised as variables for comparison (pretty much in the same vein as Bonalumi et al. 2020; or Mazzarella et al. 2018). IM types represent different ways of saying the same thing, which allows us to compare how different types of IM fare when put to use for different rhetorical purposes.19 This means that all the effects collected in Table 1 could, in principle, be experimentally investigated to determine whether the different types of IM postulated may trigger them or not.

The significance of this type of investigation, for argumentation scholarship, is high. First, the use of experimental designs to test rhetorical effects—even if limited, due to the necessarily constrained and reduced nature of the experimental material—may empirically document claims that were, so far, only made theoretically. Second, our understanding of verbal argumentation, and specifically our understanding of fine-grained linguistic subtleties that could appear to be inconsequential, can be improved by bringing to light evidence that linguistic manipulations can be impactful (see also the work of Schumann 2022, this volume; Schumann et al. 2019, 2021). Finally, the kind of methodological and theoretical integration advocated here represents an original, novel, and welcome empirical addition to the already fruitful pragmatics and argumentation interface.