The “All Lives Matter” response: QUD-shifting as epistemic injustice

Drawing on recent work in formal pragmatic theory, this paper shows that the manipulation of discourse structure—in particular, by way of shifting the Question Under Discussion mid-discourse—can constitute an act of epistemic injustice. I argue that the “All Lives Matter” response to the “Black Lives Matter” slogan is one such case; this response shifts the Question Under Discussion governing the overarching discourse from Do Black lives matter? to Which lives matter? This manipulation of the discourse structure systematically obscures the intended meaning of “Black lives matter” and disincentivizes future utterances of it.


Introduction
Linguistic exchanges are not haphazard affairs; rather, they are structured social interactions. One element of this structure can be modeled by what formal pragmatic theorists call Questions Under Discussion (hereafter QUDs). According to Roberts (2012)-building on the work of Lewis (1969), Stalnaker (1978), and Carlson (1982)-each discourse is structured around a set of interlocking questions arranged in a hierarchy, which represent the goals of the conversation and strategies for achieving those goals; at any given moment in the conversation, the QUD is the question which represents the most immediate aim of the discourse. Conversational moves are actions performed by discourse participants in service of achieving the goals represented by the set of discourse questions. In addition to structuring discourse, QUDs constrain these conversational moves; Schoubye and Stokke (2016), for instance, argue that what is said by a particular utterance is a function of the QUD, and van Kuppevelt (1996), van Rooij and Schulz (2004), Zondervan et al. 1 3 (2008), Benz and Salfner (2013) have shown that the QUD affects the interpretation of implicatures.
While linguistic theory has long been guided by the assumption that linguistic interaction is a fundamentally cooperative affair, recent work in the philosophy of language has highlighted myriad ways in which language may be used as a tool for the perpetuation of unjust social systems rather than the realization of shared goals. 1 Following Kidd et al. (2017), I'll use the umbrella term epistemic injustice to refer to "those forms of unfair treatment that relate to issues of knowledge, understanding, and participation in communicative practices." 2 Drawing on recent work in formal pragmatic theory, I show that the manipulation of discourse structure-in particular, by way of shifting the QUD-can constitute an act of epistemic injustice. I argue that the "All Lives Matter" response to the "Black Lives Matter" slogan is one such case. Though it began as a tone-deaf expression of support for the BLM protest movement, the "All lives matter" response quickly morphed into an expression of opposition and proposed corrective to the slogan, "Black lives matter". 3 Here are just a few examples which illustrate the corrective/oppositive use of "All lives matter": Protesters have made signs with the word "Black" in "Black lives matter" conspicuously crossed out and replaced with "all", 4 politicians such as Mike Pence have answered "All lives matter" in direct response to the question "Do Black lives matter?", 5 women taking down "Black lives matter" signs in Visalia, CA claimed that they had a problem with these signs "because all lives matter", 6 and a wedding photographer's contract was cancelled after she posted a "Black lives matter" sign on her social media page. The explanation given by the couple was that they could not support someone "who does not believe that ALL lives matter". 7 As many have pointed out, "All lives matter" cannot coherently serve as a corrective to what is said by "Black lives matter", given that the former entails the latter. 8 Rather, it serves as a corrective to a perceived implicature of "Black lives matter"; namely, that only Black lives matter (what Anderson 2017 calls the exclusive reading). I argue below that this implicature only arises given the assumption that the QUD is Which lives matter? and that "All lives matter" is proposed as an alternative answer to that particular QUD. However, the QUD governing the context in which protesters assert "Black lives matter" is rather, Do Black lives matter? Thus, ALM activists are shifting the QUD midstream, a move that results in the misinterpretation of "Black lives matter" (what Anderson, 2017 calls illocutionary flipping) and disincentivizes future utterances of that phrase (what Dotson, 2011 calls testimonial smothering). This manipulation of the discourse structure by way of shifting the QUD constitutes epistemic injustice insofar as serves to perpetuate an oppressive status quo.
Before I begin, let me say a word about what I hope to contribute to the discussion surrounding the "All lives matter" slogan, and to the literature on epistemic justice. I aim to offer a complete interpretive analysis neither of "Black lives matter" nor "All lives matter", but rather to focus on one particular aspect of a nuanced and complex discourse. Nor do I offer social/political insights that are not present-in one form or another-in various academic papers, news articles, and social media posts which engage with this topic. The way in which I propose to contribute to this discussion is to draw on these insights and situate them within the framework of formal pragmatic theory. My hope is that this will serve to more sharply delineate one particular mechanism behind the epistemic injustice of the "All lives matter" response-namely, manipulation of the discourse structure by shifting the QUDand provide the theoretical resources to identify other instances of epistemic injustice through the same mechanism.

How questions structure discourse
According to a popular semantic/pragmatic framework developed by Roberts (2012)-building on the work of Lewis (1969), Stalnaker (1978), and Carlson (1982)-discourses are structured around a set of interlocking questions representing conversational goals. Following a long philosophical tradition, Roberts assumes that conversation is fundamentally a cooperative affair aimed at gaining information about the world-thus, she claims, the overarching goal can be characterized as that of answering the Big Question: What is the way things are? This conception of the goal of linguistic exchange is highly idealized; conversations are often characterized by conflict, and aimed not at gaining information about the world, but rather at changing it-or preserving the status quo, as the case may be. 9 Robert's framework can nonetheless be helpful in understanding the discourse structure surrounding the "All lives matter" response, and I will adopt it (while rejecting the idealized background assumption) in what follows.
Whether or not the overarching goal of the conversation is shared, and whether or not it is aimed at figuring out what the world is like, our conversations are nonetheless structured and organized around goals, which can be represented by questions. Roberts notes that in each conversation there is usually a set of questions organized in a hierarchy, with some questions ("subquestions") serving as strategies for answering others ("superquestions"). 10 For instance, addressing the question Is it raining? might be a strategy for answering the question Should we go to the park today?, The QUD is the question which represents the most immediate aim of the discourse. Roberts suggests that we think of questions as arranged in a stack: once we answer the immediate question of what the weather is like outside, it will be replaced by the next question in the stack: Should we go to the park today? Since we have answered the question about the weather, we will now be in a better position to answer the question about the park, and so on.
Building on Groenendijk and Stokhof (1984), Roberts suggests that questions structure discourse by setting up a set of alternatives which partition the common ground into cells. The notion of common ground comes from Stalnaker (1978), and refers to the information that is mutually taken for granted by conversational participants. The common ground is modelled as a set of possible worlds-the "live options" for ways the actual world is, as far as the conversational participants are concerned. Questions serve to partition the space of live possibilities into potential answers; e.g., Is it raining? partitions this space into a cell containing the set of worlds in which it is raining and a cell containing the set of worlds in which it is not. Assertions, or answers to questions, aim to eliminate live possibilities; if you assert, "It's not raining", then you've proposed to eliminate the cell containing worlds in which it is raining from the common ground, leaving only the set of words in which it is not raining as live possibilities.
There are two broad types of questions: polar questions (yes/no questions) and wh-questions ("which…", "who…" "what…", etc.). While polar questions like Is it raining? partition the set of lives possibilities into two cells (corresponding to yes/ no answers), wh-questions, such as Who wants to go to the park? may partition the set of live possibilities into various numbers of cells, depending on the context. On Roberts' framework, the cells of Who wants to go to the park? are given by replacements of "who" by appropriate elements of the domain of the context. (The elements of the domain appropriate to "who" will be people, those appropriate to "what" will be things, etc.) For instance, if you and I are the only people relevant to our conversational goals, then the QUD Who wants to go to the park? will partition the set of possible worlds into those in which I want to go to the park, those in which you want to go to the park, and those in which both/none of us do. But it does not partition the space into worlds in which Boris Johnson wants to go to the park, because he is not included in the domain of the context-he is not relevant to our conversational goals. Thus, the way in which the QUD partitions worlds into cells will depend upon features of the context, including which people are taken to be conversationally relevant-this will be taken up again in Sect. 4.
Answers to questions can be complete or partial. For instance. If the QUD is Who wants to go to the park? in a context which includes just you and me, a complete answer will need to give a verdict on each of the resulting cells; that is, it needs to indicate whether or not you want to go to the park and whether I to go to the park. A partial answer is one which gives a verdict on some but not all of the cells. For instance, if in this context I assert "I want to go to the park", I have given a partial answer to the QUD Who wants to go to the park? I've proposed to eliminate the cell in which I don't want to go to the park, but I've left open whether you want to go to the park. Thus, I've not settled the QUD, but I've allowed us to make some headway by answering one of its subquestions. Conversationalists may give partial rather than complete answers for any number of reasons. Sometimes this is because they don't know the full answer; in other cases, speakers may know the full answer but only offer a partial answer, thus implicating a full (or, at least, less partial) answer. We'll take up the issue of implicature in the following section.

How the QUD constrains discourse moves
There are two main types of conversational moves that can be made in service of achieving conversational goals: asking questions and making assertions (i.e., answering questions). Both serve as strategies for answering the QUD. Assertions do so by providing a verdict on at least one of the cells given by the QUD's partition on the common ground, while asking questions set up strategies for doing so. For instance, If the QUD is Should we go to the park today? I might ask Is it raining? In doing so, I am temporarily changing to QUD so that we will be better positioned to answer the original QUD: as we saw before, once we've answered what the weather is like, we can return to the question of whether we should go to the park and be better positioned to answer it.
The QUD governing a discourse will constrain the sorts of conversational moves available to participants. Schoubye and Stokke (2016), for instance, argue that what is said by a particular utterance is a function of the current QUD. Defining the minimal content of an utterance as the minimal content that can be compositionally derived from the meaning of its parts in a context, Schoubye and Stokke propose that what is said by an sentence in a context is the weakest answer to the QUD that entails the minimal content of that sentence. 11 The constraint that what is said must entail the minimal content of the utterance preserves the connection to the literal meaning of the sentence, while the constraint that what is said be an answer to the QUD delivers intuitive predictions in the following types of cases: With an utterance of "Tipper is ready", the speaker will have said that Tipper is ready for dinner in a context where the QUD is Is Tipper ready for dinner?, while the speaker will have said that Tipper is ready for the interview in a context where the QUD is Is Tipper ready for the interview? Similarly, with an utterance of "I wrote my paper", the speaker will have said she wrote her paper while waiting for her laundry yesterday in a context where the QUD is What did you do while waiting for your laundry yesterday?, while she would have said more simply that she wrote her paper yesterday in a context where the QUD is What did you do yesterday? While Grindrod and Borg (2019) have pushed back on the idea that the QUD framework can "provide a model of how we arrive at semantically relevant content," they concede its usefulness in determining discourse moves and communicated content. 12 For our purposes here, it is not important to evaluate the details of Schoubye and Stokke's (2016) proposal or issues pertaining to the semantics/pragmatics divide; but their results lend strong support to the claim that the interpretation of discourse moves-at the pragmatic, if not the semantic level-is constrained by the QUD.
Further support for this claim comes from recent work in experimental pragmatics. For instance, van Kuppevelt (1996), van Rooij and Schulz (2004), Zondervan et al. (2008), Benz and Salfner (2013), show that interpretations of implicatures depend on the QUD. In experiments performed by Zondervan et al. (2008), e.g. audiences evaluating an utterance of "some pizzas were delivered", derived the implicature that not all pizzas were delivered in a context where the QUD was Were all pizzas delivered? but not in a context in which the QUD was Were some pizzas delivered? Such results show that Schoubye and Stokke's constraint on what is said-namely, that it answers the QUD-applies to interpretation of implicatures as well. This can be explained by appeal to Grice's maxim of quantity, which states that speakers should make their contribution as informative as is required (no more, no less) for current purposes of the exchange. If, as Roberts suggests, purpose of the conversation is to answer the QUD, it will be natural for audiences to reach for an implicature which fills in the remaining information when a speaker is in a position to give a complete answer but provides only a partial one. That is, they will look for a conversational contribution that provides the information required for the purpose of the exchange. But if the literal content of her assertion provides a complete answer to the QUD, audiences will be less likely to derive an implicature because no more information is required-the speaker has already done her job. Thus, in a context where the QUD is Were all pizza's delievered? audiences are more likely to derive the implicature that not all pizzas were delivered because "some pizzas were delivered" has supplied only a partial answer. The implicature that not all pizzas were delivered would provide the missing information. In contrast, in a context where the QUD is Were some pizzas delivered? an utterance of "some pizzas were delivered" has supplied a complete answer. In this context, the implicature would violate the maxim of quantity by providing more information than is demanded by the conversational goals, and thus is not likely to be derived.
We've just observed that whether an utterance provides a partial or complete answer-and as a result, whether it is reasonable to interpret the speaker as having implicated additional information-depends on the QUD. We can apply this observation to utterances of the sentence "Black lives matter". Whether or not such utterances provide a partial answer-thus raising the interpretive issue of an implicature-depends on the QUD. The polar question Do Black lives matter? partitions the common ground into two cells; those in which Black lives matter, and those in which they don't. An answer of "Black lives matter" provides a complete answer to this question by giving a ruling on each of these cells; it proposes to make the proposition that Black lives matter part of the common ground and rule out the possibility that they don't, thus settling the question. In contrast, the wh-question Which lives matter? partitions the common ground into as many cells as there are types of lives in the domain. (I.e., cells corresponding to any relevant completion of "___ lives matter.") The potential for the implicature that only Black lives matter-which, following Anderson (2017), I'll call the exclusive reading-could reasonably arise only when the QUD is Which lives matter? but not when the QUD is Do Black lives matter? given that it serves as an answer to the former, but not the latter question. That is, only in a context in which Which lives matter? is the QUD-and the speaker is in a position to give a verdict on each of the cells but offers an affirmative ruling only on the cell pertaining to Black lives-could it be reasonable to interpret her as implicitly giving a negative ruling on the remaining cells. 13 Similarly, the statement "All lives matter" could only reasonably serve as a corrective to "Black lives matter" if both are offered as opposing answers to the question Which lives matter? When the QUD is understood as Do Black lives matter? an utterance of "All lives matter" fails to provide an answer. While it may entail an answer given further assumptions about the speaker's beliefs (i.e., that black lives are included in the domain of discourse), in this case it nonetheless violates the maxim of quantity by providing more information than is required for the purpose of the exchange. Moreover, in such a case, the entailment would affirm rather than contradict the statement "Black lives matter". Thus, in order to make sense of a speaker's utterance of "All lives matter" as a relevant contribution offered in opposition to, or as a corrective to, the statement "Black lives matter", we would need to read her as operating under the assumption that both statements were made under the QUD Which lives matter? In this case, she will have derived the exclusive reading from "Black lives matter" and will be offering an answer which contradicts it.
Thus, I suggest that in order to explain the oppositional and corrective nature of the "All lives matter" response, we must assume that it both proports to answer the QUD Which lives matter? and interprets utterances of "Black lives matter" as proffering a complete answer to that same QUD by implicating the exclusive reading. In the following section I'll offer linguistic considerations that further support the claim that "All lives matter" responses purport to answer the QUD Which lives matter? but that utterances of "Black lives matter" purport to answer a different question, namely Do Black lives matter? If this is correct, then ALM activists are manipulating the discourse structure by shifting the QUD mid-discourse, which-I will go on to argue in Sect. 5-constitutes an act of epistemic injustice.

Which QUD?
In the previous section I suggested that the slogans "Black lives matter" and "All lives matter" are proffered as answers to different QUDs. In this section I'll offer linguistic support for this claim. As Roberts notes, QUDs are usually raised implicitly rather than explicitly within discourse, which makes selecting the right QUD for a given assertion an interpretive issue. Linguistic research has shown that there are at least two reliable cues for determining which QUD an assertion aims to answer: focus and context. 14 In this section I'll appeal to both in order to support the claims that the slogan "Black lives matter" purports to answer the question Do Black lives matter? while the response of "All lives matter" purports to answer the question Which lives matter?
Research has established that focus (intonation on a particular word or phrase) strongly correlates with certain types of questions. In particular, answers to whquestions tend to focus on the topic; i.e., an utterance will be felicitious only in the case that it is an answer to a question which is determined by replacing the focused element with an appropriate wh-expression. Consider the following sentence: 1. Xavier ate Swedish fish.
An utterance of (1) could be an answer to any of the following three questions: 2. What did Xavier eat? 3. Who ate Swedish fish? 4. Did Xavier eat Swedish fish?
As we have seen, what is ultimately communicated by an utterance of (1) depends on which question it aims to answer. In particular, it may give rise to exclusive reading implicatures if the QUD is (2) or (3), for which it gives a partial answer-but not (4), for which it gives a complete answer. In the former cases, if it can be assumed that the speaker is in a position to give a ruling on each cell of the partition, she will likely be read as implicating that Xavier ate only Swedish fish as an answer to (2) and that only Xavier ate Swedish fish as an answer to (3). In contrast, if (4) is the QUD, an utterance of (1) will not generate a reading of an implicature because it provides a complete answer to that question. Given that the QUD is often implicit, audiences common rely on a speaker's prosodic focus to determining which QUD they aim to address. Consider an utterance of the following sentence:

Xavier ate [Swedish fish] f
Replacing the focused element "Swedish fish" with the appropriate wh-expressionnamely, "what"-gives us back (2), for which (5) is a felicitous answer. In contrast, it would be infelicitous as an answer to (3) or (4). The same pattern holds for: 6.
[Xavier] f ate Swedish fish Replacing the focused element "Xavier" with the appropriate wh-question-namely, "who"-delivers (3), the only QUD for which it is a felicitous answer. Finally, unfocused (or sometimes called "broadly focused") answers are felicitous as answers to polar questions, to which the entire statement corresponds:
This data on focal patterning suggests that if "Black lives matter" were offered as an answer to the question Which lives matter? we should expect to observe focus on the word "Black". This could take the form of vocal prosodic stress on that word, or-in written utterances such as protest signs, t-shirts, social media posts, etc.the word might be underlined, written in all caps, red lettering, or bold print. Protest chants, however, tend not to include prosodic stress on the word "Black". Nor, according to a google image search of protest signs, BLM merchandise, social media posts and memes, etc. does there tend to be focus on the word "Black" through the use of textual highlighting. 15 This evidence provides support for the conclusion that utterances of "Black lives matter" are proffered as answer to the polar question Do Black lives matter? rather than the wh-question Which lives matter?.
Facts about the discourse context provide additional support for this claim. A QUD will tend not to be raised if it is already taken as common ground; otherwise it will introduce a redundant discourse goal. That is, it will raise an issue that has already been mutually taken to be settled by discourse participants. The discourse context surrounding the BLM movement is one in which it is taken as common ground-indeed platudinous-that all lives matter. 16 Thus, the question Which lives matter? -at least insofar as the context of the domain is limited to human livesfails to introduce a live issue to be settled; the common ground already includes the answer to this question, namely that all lives matter. However, in spite of the mutual acceptance of this platitude, we are also located in a context in which Black people are systematically treated as if their lives are not valuable, even by those who willingly endorse the slogan "All lives matter". This feature of the context raises the issue of whether Black lives are seen as an exception to this phrase by those who would endorse it with their words but not their actions (more on this below).
Recall that common ground includes information that is mutually accepted by discourse participants and as such is not determined solely by linguistic contributions, but may also include mutually apparent facts. The discourse surrounding the BLM movement takes place in a context in which many discourse participants, as Butler (2015) notes, "embrace a "'norm'" that effectively says "Black lives do not matter," one that is built up over time, through daily practices, modes of address, through the organization of schools, work, prison, law, and media." This context is one in which it is distinctly not yet common ground-that is, not mutually accepted by discourse participants-that Black lives matter. It is a context which in which a very particular question about the value of Black lives, rather than lives in general, is treated as unsettled: Do Black lives matter? 17 Even companies like Apple and Google have picked up on the QUD governing the BLM discourse context. Both Siri and Google Assistant, for instance, have been programmed to answer the question "Do Black lives matter?" with "Black lives matter". 18 They direct their users to BlackLivesMatter.com for further information about the subject, indicating that they take the QUD behind the statement "Black lives matter" to be Do Black lives matter? Similarly, Tiktok users have developed a trend of asking the question "Do Black lives matter?" and listing the answers that they have been given. 19 BLM supporters seeking to narrow or broaden the scope of the conversation introduce variations on Do Black lives matter? rather than Which lives matter?, asking questions such as: Do Black lives matter in Iowa? Do Black lives matter to Democrats? Do Black lives matter in bilingual education? 20 In all of these cases, BLM supporters have implicitly or explicitly picked up on the fact that the QUD governing the assertion of "Black lives matter" by BLM protesters is the polar question Do Black lives matter? and that the answer it offers is a resounding affirmative.
The sticking point, however, is that the context remains the same; it is one in which discourse participants verbally adopt the platitude that all lives matter, while readily participating in behaviours, structures, and institutions which constantly reinforce the idea that Black lives do not matter. There is a glaring inconsistency between the values affirmed in their speech versus the values affirmed in their participation in racist social structures. As Roberts (2012) notes, when audiences encounter this type of apparent inconsistency, they tend to interpret speakers in a way to make them consistent. 25 In particular, they might do this by adjusting what they interpret to be the intended domain of the speaker's utterance in a way that would make her words consistent. As Butler (2015) (echoing Mills, 1997) points out, the only way to make sense of the claim that "all lives matter" in a context in which it is treated as acceptable for Black people to be murdered on the streets is to assume that those who use this phrase do not include Black people in the domain of the discourse. That is, they do not think of Black lives as lives at all, or at least not lives that are relevant enough to be included in the statement "all lives matter". 26 One purpose of the statement "Black lives matter" is to make this exclusion salient. That all lives matter, as Keeyanga-Yamahtta Taylor notes, "has always been an assumption. The entire point of Black lives matter is to illustrate the extent to which Black lives have not mattered in this country." 27 Moreover, as Butler (2015) notes, this exclusion must be brought to salience in order to demand and effect change: "To 25 See Roberts (2012, p. 29). 26 Rand Paul, for instance, suggest that the slogan "Black lives matter" should be replaced with "Innocent lives matter", implicating that Black lives do not belong in the domain of "All lives matter" insofar as they are not innocent. https:// www. thegu ardian. com/ us-news/ 2015/ aug/ 27/ rand-paul-Black-lives-matter-name-change-all-innoc ent 27 Quoted in Victor (2016).

QUD shifting as epistemic injustice
While I have questioned the assumption that linguistic exchange can be accurately modeled as a cooperative endeavor to gain information about the world, it is certainly the case that successful communication requires some level of cooperation and reciprocity from conversational participants. As part of a long philosophical tradition associated with the work of Paul Grice (1989), theorists of language have focused primarily on the way in which audiences depend on speakers to make their communicative intentions transparent. However, more recent work by linguistic theorists including Langton (1993), Hornsby (1995), Fricker (2007, and Dotson (2011), building on the work of Austin (1962) and Black feminist theorists such as Lorde (2007) and Collins (2000), have highlighted ways in which speakers are also dependent on audiences for understanding and uptake. 29 In this section I focus, in particular, on ways in which speakers depend on audiences to recognize and take up the QUD governing their assertions. I'll show that failure to do can constitute epistemic injustice. 30 Following Kidd et al. (2017), I'll use the umbrella term epistemic injustice very broadly to refer to "those forms of unfair treatment that relate to issues of knowledge, understanding, and participation in communicative practices." 31 As Anderson (2017) points out, there are two prominent ways that epistemic injustice is 28 Of course, there are other reasons that this assertion makes many White people uncomfortable. As Halstead (2017) notes, for those who are used to mattering as a zero-sum game, the valuing of some lives involves the devaluing of others. In particular, under a White supremacist system, the value of White lives is predicated upon the devaluing of lives of people of color. Cf. Atkins (2019). 29 Following, Austin (1962), Dotson (2011), Kukla (2014), and Anderson (2017) also highlight ways in which the moving parts involved in successful communicative exchange include context in addition to speaker and audience; background circumstances must be such as to facilitate mutual understanding and communicative success. 30 According to Roberts (2012), implicit QUDs can be characterized as presuppositions. Thus, failure to recognize and take up a QUD that a speaker implicitly aims to answer with her assertion qualifies more generally as a failure of presupposition accommodation. Though the exploration of QUD-shifting through the particular framework of presupposition accommodation failure strikes me as an important and interesting project, I will not have the space to do explore it here. 31 Kidd et al. (2017, p. 1). manifested in linguistic exchange: the obscuring of what the speaker meant and the loss of ability of speakers to perform uttterances and be interpreted correctly. I will argue below that manipulation of the discourse context by way of QUD-shifting can result in both; in particular, by shifting the QUD from Do Black lives matter? to Which lives matter? ALM activists assign to BLM activists an implicature that they did not perform, and disincentivize further utterances of "Black lives matter". I will also argue that this manner of QUD-shifting stems from what Dotson (2011) calls pernicious ignorance; i.e., ignorance that "follows from a predictable epistemic gap in cognitive resources". As Dotson points out, the pernicious nature of such ignorance is not determined by the intentions of individuals, but rather the ways in which it contributes to harmful practices-in this case, the failure to recognize and change a racist social structure. Because Anderson (2017) has made similar observations about the "All lives matter" response, I want to note how my proposal below both aligns with, and diverges from his: we both argue that the "All lives matter" response constitutes an act of epistemic injustice [Anderson (2017) uses the term hermeneutical injustice] insofar as it obscures speakers' intended meaning and impacts their ability to successfully perform future utterances. However, while Anderson focuses on the causal explanation for this injustice by invoking the ideology of color-blindness, I focus on articulating one particular linguistic mechanism through which this injustice is enacted. I don't see these proposals as being in conflict, but rather bringing out different, but related aspects of a complex social issue. 32 (E.g., the ideology of colorblindless is likely a partial explanation for the linguistic phenomenon that I aim to describe.) As Anderson (2017) notes, one way in which audience members may fail to uphold their end of the cooperative enterprise involved in successful communication is to obscure the speaker's intended meaning by making their speech act out to be something other than was intended. This failure of conversational reciprocity, which he calls illocutionary flipping, has received much attention in relation to sexual refusal; for instance, a paradigm case of illocutionary flipping is that of a woman's utterance of "no" in response to a man's sexual advances being interpreted by him as an act of consent rather than refusal. 33 Similarly, when ALM activists assign the exclusive implicature Only Black lives matter to an utterance of "Black lives matter", they misrepresent the speech act performed in such an utterance; they make it out to be an answer to the question Which lives matter? which gives a negative ruling on non-Black lives, rather than an answer to Do Black lives matter? which gives 32 My proposal concerning what is communicated by "Black lives matter" also differs from the proposal offered by Anderson (2017), who claims that the correct interpretation is what he calls the Inclusive reading-namely, that Black lives matter, too. This reading is better understood as an answer to Which lives matter? and concerns a topic which includes all the types of lives in the domain. In contrast, I am claiming that "Black lives matter" should be understood as answering Do Black lives matter? and concerns only the topic of Black lives. In this sense I agree with Atkins (2019), who claims that Anderson's (2017) interpretation is better understood as a repositioning-rather than a clarification-of "Black lives matter". 33 See, e.g., Hornsby (1993), Langton (1993), Hornsby and Langton (1998), Maitra (2009), andMcGowan (2014).
1 3 a positive answer to a polar question whose topic concerns all and only Black lives. The mechanism by which this illocutionary flipping occurs is manipulation of the discourse context through shifting the QUD. The "All lives matter" response shifts the QUD from Do Black lives matter? to Which lives matter?, while presenting the discourse as if it had been governed by the latter all along; this allows ALM activists to assign the exclusive reading to utterances of "Black lives matter"-an implicature which could not have been derived under the intended QUD of those utterances, namely Do Black lives matter? This manipulation of the discourse serves to perpetuate an oppressive status quo: it directs attention away from the topic of the value of Black lives, enabling discourse members to evade confrontation with ways in which they participate in behaviours, structures, and institutions which reinforce the idea that Black lives do not matter. It stirs up confusion and divisiveness by systematically distorting affirmations of the value of Black lives as expressions of threat to the value of the lives of others.
Moreover, this act of illocutionary flipping comes from a place of pernicious ignorance, and perpetuates the conditions of such ignorance; misidentification of the QUD governing utterances of "Black lives matter" are more likely to be made by those holding social positions in which they are unaffected by-or benefitted bythe systematic devaluing of Black lives. It is more likely to be made by those holding social positions according to which they cannot make coherent sense of questions about the value of lives, when those questions are not centered on White lives. As Dotson (2011) points out, the culpability of such ignorance is beside the point. Its pernicious nature is due to the ways in which it contributes to a harmful practice, in this case the harmful practice of giving distorting interpretations of utterances of "Black lives matter", which in turn serve to perpetuate an oppressive status quo.
Similarly, the "All lives matter" response disincentivizes future utterances of "Black lives matter" through what Dotson (2011) calls testimonial smothering. Dotson (2011 notes that testimonial smothering usually occurs under conditions of testimonial incompetence, and when the utterance is unsafe. The "All lives matter" response occurs under, and perpetuates both of these conditions: Testimonial incompetence, according to Dotson (2011), is "the failure of an audience to demonstrate to the speaker that they will find the proffered testimony accurately intelligible. 34 In manipulating the discourse context by shifting the QUD from Do Black lives matter? to Which lives matter?, ALM activists both demonstrate their own testimonial incompetence and breed testimonial incompetence in others. As noted above, they demonstrate their own inability to find the proffered testimony accurately intelligibly by misrepresenting past utterances of "Black lives matter" as though they had been proffered as answers to a different QUD, and as though they had carried an exclusive implicature. In doing so, they render it more likely that future utterances will be distorted in the same manner-both by the fact that others may take their interpretation as authoritative, and by the fact that in uttering "All lives matter" they establish Which lives matter? as the new QUD. As we saw above, this is the only QUD under which the exclusive implicature could reasonably derived from utterances of "Black lives matter". Thus, establishing Which lives matter? as the new QUD places an additional burden on speakers who wish to utter "Black lives matter" as, in order to avoid misinterpretation, they will need to switch the QUD back to the original question in a context in which many discourse participants have demonstrated an active and committed refusal to accommodate it. In addition to creating conditions of testimonial incompetence, the "All lives matter" response renders future utterances of "Black lives matter" unsafe. According to Dotson (2011), an utterance is unsafe in the case that it can easily lead to false beliefs about its intended message, causing social, political, and material harm. In the case that utterances of "Black lives matter" are misinterpreted according to the exclusive reading, such harm could be manifested in, e.g., disproportionate force against BLM protestors by governmental agencies, perpetuation of White silence in both in the public and private sphere, violence from White supremacist groups, etc. 35 Finally, the smothering of utterances of "Black lives matter" is driven by, and perpetuates, the same pernicious ignorance involved in their misinterpretation.
Thus, the "All lives matter" response constitutes an act of epistemic injustice both insofar as it flips the illocutionary act performed by utterances of "Black lives matter" and disincentivizes future such utterances. Moreover, these acts of epistemic injustice are not individual, one-off failures of communicative reciprocity; they are systematic. The "All lives matter" response has become a symbol for a movement aimed at undermining racial justice. It is a hashtag on social media, a slogan for counter protestors at BLM gatherings, a thin veil to throw over expressions of hypocrisy and hate. It keeps the topic of discourse on White lives, allowing those who benefit from White supremacy to continue to live in ignorance of the disconnect between their participation in this system and the superficial platitudes they accept.

Conclusion
I argued above that utterances of "Black lives matter" are proffered as answers to the QUD Do Black lives matter? which has been raised in a context where the value of Black lives is daily and systematically called into question. I argued that utterances of "All lives matter", in contrast, are proffered as answers to a different QUD-namely, Which lives matter? I've appealed to linguistic research concerning patterns of correspondence between intonational focus and discourse QUD, and well as facts about the context of discourse in order to support these claims. If the foregoing is correct, the "All lives matter" response to "Black lives matter" shifts the QUD mid-discourse-a move which, I've argued, constitutes an act of epistemic injustice insofar as it obscures the intended meaning of utterances of "Black lives matter" and disincentivizes future such utterances, serving to systematically perpetuate an oppressive status quo. In conclusion, let me address a few potential worries for the foregoing arguments: 1. But not all QUD shifting is unjust! Absolutely. Almost any kind of misinterpretation-including that involved in QUD-shifting, can be completely innocuous. The ALM response is unjust because it perpetuates an oppressive social system, in the ways outlined above.
2. But the QUD-shifting in the ALM response may not be intentional! I agree, and would even conjecture that it is likely to be unintentional in most cases. Nonetheless, it is unjust. Like Dotson (2011), I take the issue of culpability to be irrelevant; it is unjust because it perpetuates harm, and not because of any particular speaker's intentions.
3. But it's not really epistemic! Let me remind the reader that I am using the term epistemic injustice in a broad, stipulative sense which is not limited to the epistemic, but includes "forms of unfair treatment that relate to… participation in communicative practices". The objection, then, just amounts to the complaint that I am using the term as a misnomer-in which case the reader is free to substitute their preferred terminology.
Nonetheless, it is worth pointing out the ALM response is intimately related to the epistemic in the following ways. First, interpretation is an epistemic phenomenon in the sense that to interpret someone is to form a belief about her conversational goals. The systematic misinterpretation of utterances of "Black lives matter" is thus an epistemic phenomenon. Moreover, this phenomenon is explained by epistemic states in the sense that the exclusive reading can only be reasonably attributed given ignorance of the state of racial violence in the U.S. Finally, the ALM response to this misinterpretation may affect epistemic states in the sense that it encourages further misinterpretations of utterances of "Black lives matter". 36 4. But it's obvious (and so unimportant)! Many may find it obvious that the "All lives matter" response involves changing the subject somehow, and that this changing-of-the-subject is morally problematic. So why are the foregoing arguments important, if they just tediously articulate what many find intuitively obvious? U s i n g theoretical tools to carefully articulate an inchoate-even if intuitively obvious-source of injustice is important because it serves to fill what Fricker (2007) calls a "hermeneutical lacuna". It provides the resources to more deeply understand the phenomenon, to clearly explain it to others (especially those for whom it is not so obvious), to identify instances of the same general type of phenomenon occurring in other contexts, and to mitigate against it. This paper is both an instance of applied philosophy and of theoretical philosophy. It applies theoretical resources from linguistic and moral theory to analyze a particular social phenomenon, but it also identifies a broader mechanism of epistemic injustice-namely, QUD-shifting-which has applications outside this phenomenon. Further explorations of the way this mechanism is used for purposes of epistemic injustice could include, for instance, contexts of domestic abuse in which the abuser gaslights the victim by shifting the QUD mid-discourse, thus misrepresenting her former speech acts and disincentivizing further utterances. Others may include criminal justice contexts in which QUD-shifting distorts the discourse context in ways that misrepresent testimony and disincentivize victims/the unjustly accused from speaking out.
The social and political implications of the "All lives matter" response I've highlighted here are not new; it's topic-shifting function is transparent to many, and theorists and lay-people alike have outlined various ways in which this response serves as a tool for epistemic injustice. I hope to have contributed to this discussion by building upon on these insights and situating them within the framework of formal pragmatic theory. This serves both to more sharply delineating one particular mechanism behind the epistemic injustice of the "All lives matter" response-namely, manipulation of the discourse structure through QUD-shifting-and to provide the theoretical resources to identify other instances of epistemic injustice performed through the same mechanism. Finally, I hope to have shown the usefulness of formal pragmatic theory as a tool for making sense of political discourse, while also calling into question some of its idealized assumptions about the cooperative nature of linguistic discourse.