A rich-lexicon theory of slurs and their uses

ABSTRACT In this paper, I present data involving the use of the Romanian slur ‘țigan’, consideration of which leads to the postulation of a sui-generis, irreducible type of use of slurs (‘identificatory’). This type of use is potentially problematic for extant theories of slurs. In addition, together with other well-established uses (derogatory, appropriated etc.), it shows that there is more variation in the use of slurs than previously acknowledged. I explain this variation by construing slurs as polysemous. To implement this idea, I appeal to a rich-lexicon account of polysemy. I show how such a theory can be applied to slurs and discuss several important issues that arise.

or resorting to the idea that a slur's meaning solely depends on the intentions, beliefs or attitudes of speakers. I hold, instead, that slurs are polysemous. I appeal to a certain type of theory from lexical semantics to account for polysemy (a 'rich-lexicon' theory) and show how it can be applied to the various uses of 'țigan' and to slurs in general.
I proceed as follows. In the first section, I very briefly present several characteristics of slurs and their main uses. Section 2 is dedicated to presenting the data and its interpretation. In section 3 I show how identificatory uses can be problematic for extant theories of slurs. The next two sections present the sketch of a rich-lexicon approach to polysemy, first as it applies to nouns in general, then in application to slurs. In section 6 I discuss several important issues that arise. I then summarize and conclude.

Slurs: what they are and how we use them
A common way to characterize slurs has been by way of the function they serve in discourse. The following quote from Robin Jeshion nicely summarizes what most authors agree the main purpose of using slurs is, namely 'to derogate or dehumanize (…), to signal that their targets are unworthy of equal standing or full respect as persons, that they are inferior as persons.' (Jeshion 2013, 232). It is also widely agreed that slurs do double duty: they manage to refer to or pick out a particular group (what I call in what follows 'the target group'), and derogate its members because of being members of that group. Slurs thus target both groups and individuals, the latter via group membership. The main task of a theory of slurs is to offer a mechanism by which to capture the derogatory aspect of slurs. 1 A lot of work has been put into discerning the characteristic traits of slurs. Among those that most researchers have agreed upon are nondisplaceability (Potts 2007): the derogatory character of slurs doesn't disappear when embedded under various linguistic constructions, such as speech or beliefs reports, conditionals or negation; derogatory variation (Hom 2008(Hom , 2010Popa-Wyatt and Wyatt 2018): the fact that slurs vary in their derogatoriness, both when they target the same target group and across target groups; independence (what Hom 2008 calls 'derogatory autonomy'; Anderson and Lepore 2013a): the fact that their derogatory character doesn't depend on the intentions, beliefs or attitudes of speakers. While accounting for these traits is a desideratum for any theory of slurs, I will be concerned in this paper only with the latter. For my purposes here, independence will be taken to be non-negotiable.
In addition to the characteristics just highlighted, slurs have been taken to have several types of uses. As we have seen, slurs' main use is the one by which they fulfill their function to derogate, namely their derogatory use. But they have other uses, too, including non-derogatory ones. 2 The most discussed such uses are appropriated 3 uses, mostly directed towards members of a target group to which the speaker belongs, but not with the purpose of derogating. One widely used example of a successfully appropriated slur is 'queer', which presumably has nowadays become devoid of derogatory content (many academic publications have 'Queer Studies' in their title, for example), or the n-word, which has been reserved for in-group use (Gloria Naylor's essay 'The Meanings of a Word' shows the great variety of forms its appropriated use can take). Essentially, what happens when slurs are used in such a way is that they cease to communicate a negative attitude towards the target group, and communicate a positive one instead. There are also what Hom (2008) has called nonderogatory, nonappropriated (NDNA) uses, whose purpose is to correct certain injustices stemming from the derogatory uses of a slur, as in 'Institutions that treat Chinese as chinks are morally depraved. ' (2008; 424). Anderson 2018, following previous work, points to what he calls referential uses; thus, he notes in connection to the n-word (more precisely, one of its derived forms) that it is often used by African Americans as a synonym for 'friend; buddy; man ' (2018, 9). The distinguishing mark of referential uses is that the speaker need not intend to communicate a positive attitude with the slur; a neutral attitude is the default. Metaphorical uses of slurs (or at least some of them), as in John Lennon and Yoko Ono's song 'Woman is the nigger of the world', attempt not to derogate, but to prove a point. 4 Finally, there are didactical or quotational uses such as those found in academic works or courts of law for purposes of illustration, where slurs are not used, but merely mentioned. Here, I won't pay attention to all these uses, but focus on appropriated and referential ones insteadmainly with the purpose of characterizing the novel type of use I postulate. I grant, of course, that a complete theory of slurs must account for all types of uses of slurs, although that is not my project here.
The existence of such a variety of uses points toward a certain variability of slurs: the same word, or even the same sentence, can be used derogatorily in one context and non-derogatorily in another. Even leaving aside metaphorical and didactical uses, this sensitivity to context is a remarkable feature of slurs, and any complete theory should account for it. What the elements in a specific context that allow for a felicitous use of a slur in a non-derogatory way are is a crucial question, as is the issue of whether the variation mentioned is semantic, pragmatic, or of yet another kind. My answer to some of these questions lies in construing slurs as polysemous. In sketching the rich-lexicon theory I think gives the best account of polysemy, I focus mostly on accounting for the main uses of slurs highlighted aboveincluding the one about to be presented. Although this not the end of the story (a compete theory of slurs having to account both for their various uses and for their characteristics), it is an important part of any theoretical endeavor. One advantage that I take such a view to have is that avoids ambiguity. However, it should be kept in mind that I see the view's main attraction not in accomplishing this feature, but rather in putting polysemy theories into sharper focus and thus contributing further to a fruitful line of research. Let's now have a look at the data.

Identificatory uses of slurs: the data
In 2007, Romanian president Trăian Băsescu has been recorded calling a journalist 'stinky gypsy' ('țigancă împuțită') and was subsequently fined by the National Council against Discrimination. 5 The slur 'țigan/țigancă', 4 That is not to say that such attempts are always successful. In the case above, the title and the lyrics of the song have sparked controversy, leading to the refusal of many radio hosts to play it on their stations. 5 The president apologized after the story broke to the public, but was fined nevertheless. For part of the story, see http://www.nbcnews.com/id/18789903/ns/world_news-europe/t/romania-president-sorrystinky-gypsy-slur/#.Xbg3NehKjb0.
roughly translated into English with 'gypsy', comes from the Byzantine Greek word 'athinganos', which supposedly meant 'untouchable' or 'dirty', thus encapsulating a negative connotation from the very beginning. The journalist targeted by the president was not a Roma ethnic, yet 'țigan' and its cognates are currently and commonly used as a slur targeting Roma ethnics. The derogatory use of 'țigan' is widespread and permeates most aspects of their lives. 6 However, while its derogatory use is very hard to contest, a certain number of members of the Roma community use 'țigan' to identify. I base this claim on a Romanian Governmental Report 2009 (henceforth RGR) aimed at gathering information about the social and economic situation of the Roma community (officially, as large as 650.000 people) in Romania. Among many other questions, Roma ethnics have been asked the following, with the percentage of answers displayed below: How do you usually refer to yourself and other Roma ethnics?
-'țigan': 44%; -'rrom': 32%; -no Romani speaker: 22%; -no answer: 2%. 7 According to these results, a relevantly large number of respondents (speakers and non-speakers of Romani alike) use 'țigan' to both identify 6 In this, 'țigan' differs from the way in which Cepollaro and Stojanovic 2016 take 'gypsy' to function: not as a slur, but as a word that could be used to derogate in certain contexts. Castroviejo, Fraser, and Vicente 2020 take the Spanish 'gitano' to be an ESTI ('ethnic/social terms used as insults'), a type of expression that, according to them, can be used to derogate but differs from slurs. I should make it clear that the claims in this paper concern only 'țigan' and its uses, and might not straightforwardly apply to uses of similar terms in other languages. Whether such terms function as slurs and what other uses they have depends on complex historical, social and political issues. 7 The exact questions were 'What do you usually say?' ('Dumneavoastră cum spuneţi de obicei?', RGR: 46, question RD13) and 'What do you usually say when you speak Romani?' ('Când vorbiţi în limba romani dumneavoastră cum spuneţi de obicei?', RGR: 46, question RD13a)which, in themselves, are quite vague. Nevertheless, it is clear from the contexts of the questionnaire that those whom the answer should apply were the participants themselves and other Roma ethnics from their community. The answers to another question are also relevantnamely, 'Among the following expressions which best describes your identity? Do you first and foremost consider yourself … ?' ('Din expresiile următoare, care sunt acelea care descriu cel mai bine identitatea dvs.? Vă consideraţi în primul rând … ?', RGR: 17, question ID4a), to which 33% of the participants answered with 'Romanian', 30% with 'rrom' (in two variants), 16% with 'țigan' (in two variants) and 21% with other terms. For discussion of the data, see Matei 2012. themselves and others from their community as members of the same ethic group. 8 What are we to make of this data? It is obvious enough that the use of 'țigan' by the respondents is non-derogatory. 9 But, as we have seen, there are several ways of using slurs non-derogatorily that have been discussed in the literature. A natural thought, then, is that the data above exemplifies one of such ways. In the reminder of this section I attempt to show that this is not the case and claim that the best way to make sense of the data is to postulate a sui-generis, irreducible type of use of slurs. The implication for the current debate is that this type of use must be taken into consideration when giving a complete theory of slurs.
As already mentioned, the most widely discussed non-derogatory use of slurs is their appropriated use. What exactly is the mechanism of appropriation is a contentious issue, but one that won't retain us here. 10 However, as a reviewer notes, this data cannot be used as it stands because the question doesn't specify the source of the offense. The respondents could have interpreted the question asking either how offensive they find 'țigan' when used by members of the Roma community or when used by those outside it. The significance of the data might thus be affected by the interpretation of the question. Further data might come from cross-linguistic studies. While the scarce evidence I got from personal interactions with citizens of other countries in Eastern Europe points to the existence of a similar use of slurs synonymous with 'țigan' in languages like Serbian, Hungarian, Bulgarian, Czech, Slovak etc., a thorough empirical study of the corresponding slurs in those languages would be of tremendous value for the investigation of this phenomenon. 9 To be sure, one could take the data to suggest that the respondents identify, both themselves and others, as members of the Roma community, but also derogateboth themselves and the others. Grigore 2008, for example, takes uses of 'țigan' to always carry a derogatory force even when used by members of the Roma community and traces it down to an internalization of the negative attitude they have been subjected to for centuries. While I don't deny that such 'internalized' derogatory uses exist within the Roma community, I'm skeptical that they cover the high percentages exhibited in the data presented above. Constantly derogating oneself and others from one's community strikes me as both psychologically implausible and socially damaging. (For a different claim involving the n-word, though, see Alexander 2010, who takes internalized derogatory uses of the word to be part of a coping mechanism used by stigmatized groups, and thus socially useful.) In any case, as Grigore herself admits (p.c.), an identificatory element remains in such uses, which is enough to make the data interesting. 10 It is customary to distinguish here between the process of appropriation and its result. The former involves the actions of those putting the meaning change in the service of social and political What is more important in this context is the fact that there is variation in the way in which slurs are used even within appropriation (as witnessed, for example, by the discussion of the n-word in Naylor 1986 or Kennedy 2003 andof 'queer' in Brontsema 2004). However, philosopherswhile being fully aware that this is a simplificationhave focused on two broad types of appropriated uses (e.g. Bianchi 2014; Anderson 2018; Jeshion 2020 calls them 'pride reclamation' and 'insular reclamation', respectively). One pertains to social and political movements that fight for the right of certain subordinated groups, such as people with non-traditional genders/sexual preferences or African Americans. Within such practices, appropriation amounts to purposely changing the meaning of a slur in order to express a positive, instead of a negative, attitude towards the target group. The second sense of appropriation involves using a slur as a term of endearment or to express camaraderie, solidarity etc. Here, too, the meaning of the slur is changed so that to express a positive attitude, but not (or not necessarily) in connection to social and political goals. The contexts in which a slur is used in each of these two senses of appropriation can differ significantly, which in turn might influence the way in which such uses are dealt with by the theory.
Can the data involving 'țigan' above be taken to exemplify appropriated uses? This is not an unreasonable claim, but one that is ultimately false. First, let's consider the sense of appropriation involving conscientious social and political activism. An important consideration in this respect, and one that I take to be decisive, is the fact that there is a strong consensus among activists for the rights of Roma ethnics that the way to proceed in order to block the negative effects of using 'țigan' for the members of the Roma community is not to resemantize it, but to replace it with a different word: namely, the Romani word 'rrom' (which is in fact recognized by the state as the politically correct way to address Roma ethnics). 11 Thus, 'țigan' is simply not a good example of an appropriated use of slurs in the political sense.
Consider next the use of 'țigan' as expressing camaraderie or endearment. While it's quite plausible that Roma ethnics sometimes use 'țigan' to express camaraderie or as a term of endearment, I find it doubtful that this interpretation can be extended to cover all the data. For one, this use only involves addressing other people, while the respondents had to answer a question about what word they use both for others ideals, while the latter refers to the changed meaning of a slur. I am here concerned only with the second, which demands explaining the change in meaning by using hypotheses and theoretical tools from linguistics/philosophy. This is not to say, however, that the linguistic hypotheses are not affected by the form the process of reclamation takes. For more on this, see Jeshion 2020. 11 This, as a reviewer notes, is a good argument to take 'țigan' to be a slur in the first place. and for themselves. While one can certainly use the same word for both occasions, this cannot be generalized. Moreover, interpreting the data in this way won't explain the use of 'țigan' in responding to the questionnaire: first, the respondents were asked how they refer to themselves and others in their community (and not, for example, how the word 'țigan' is used by non-Roma ethnics), which suggests that identification as being part of an ethnic group is an essential ingredient in this type of use; second, the (official) context of the questionnaire itself doesn't seem to be one in which considering how to express camaraderie or endearment would have been appropriate. This interpretation is, then, ruled out too. Summing up the remarks in the last two paragraphs, I conclude that interpreting the data as exemplifying appropriated uses, in either of the senses distinguished above, is not plausible. 12 I also don't think that the data can be interpreted as exemplifying referential uses. As before, while there is no need to deny that Roma ethnics sometimes use 'țigan' referentially, I doubt that this interpretation can be extended to cover all the data. Again, as with camaraderie uses of 'țigan', this use involves only addressing other people, while the question respondents had to answer was both about other people and about themselves. But, more importantly, as far as I can tell, the identificatory component is missing from Anderson's description of referential uses (even if those referred to in such uses do, in fact, belong to one particular group). Identification, both of one's self and of others form the community, as belonging to a specific ethnic group, regardless of personal relations, is precisely what is distinctive of the data I'm investigating (as obviously witnessed by the questionnaire). Thus, I conclude that it is not best interpreted as exemplifying referential uses either. 13,14 12 A reviewer asks whether the data cannot be interpreted as a type of appropriated use that is different from both the political and camaraderie sense. I suppose it can, if one so wishes, but the point I'm making is that the data doesn't seem to fall easily under the two main broad forms of appropriated uses acknowledged in the literature. Moreover, even if it were best construed as a distinct form of appropriated use, it would still be a surprising finding. On the other hand, there is a trivial sense in which the data could be interpreted as exemplifying appropriated uses: namely, as the propensity to use a slur as a way of identifying as belonging to a group in the face of adversity, as part of the process of politically appropriating it. This involves a certain sense of pride (as in slogans like 'we're queer and we're proud'). Although the data is silent regarding the existence of an element of pride, such an element is missing in the uses of 'țigan' I'm focusing on (as argued by Grigore 2008). Also, this clashes with the fact that activists have taken the route of replacing the word with its Romani correspondent (I'm assuming here that the activists are attuned to the needs of the Roma community). Thanks to Dan López de Sa for bringing up the latter issue. 13 I'm open, though, to assimilating the two types of uses, as a reviewer suggests. If so, the theory I put forward in later sections will equally account for both. 14 Other putative ways of interpreting the data are that Roma ethnics use 'țigan' as captatio benevolentiae (Grigore 2008), as a conversational shortcut or as a sign of being charitable or ironic towards the non-Roma ethnics they engage with. While this might be so in certain cases, I'm again doubtful that this Given the above, I think that construing the data as exemplifying either of the two types of appropriated use or a referential use is on the wrong track. Since an interpretation of the data as exemplifying NDNA, metaphorical or didactical uses seems even more far-fetched, I conclude that the best way to make sense of it is to postulate a sui-generis, irreducible type of use of 'țigan'. For lack of a better word, I call this type of use 'identificatory'. This is a non-derogatory type of use of a slur that is similar to appropriated and referential uses in certain respects, but ultimately different from them (e.g. due to the fact that the identificatory aspect is essential). Whether other slurs, targeting either other ethnic groups or other kinds of groups, have such uses is an empirical question that is very much open at the moment. What is clearer, though, is that identificatory uses are potentially problematic for extant theories of slurs. In the next section I will show why, using as proxy the problems raised by appropriated uses. 15

Identificatory uses of slurs: the problem
Many authors writing on slurs take appropriated uses to pose a problem (or at least a 'puzzle' or a 'worry') for most extant theories of slurs. 16 Given the very different commitments of the various theories of slurs on the market, agreement stops when it comes to the best way of accounting for such uses. It is not my purpose here to investigate how all extant theories of slurs deal with such uses; neither is it to discuss in detail how any particular theory does. Rather, the following general considerations serve to situate and partially motivate the view I want to propose.
According to 'semantic' theories (e.g. Potts 2007;Richard 2008;Hom 2008Hom , 2010Hom and May 2013;Predelli 2013;Camp 2013;Jeshion 2013), the derogatory element of slurs is encoded in their semantic explanation can be generalized. I grant, however, that more research is needed in this direction. I warmly thank Delia Grigore and Gelu Duminică, both Roma activists, for discussion on this point. 15 Postulating a sui-generis, irreducible type of use of 'țigan' also seems to explain the difference in the percentages of the 'not offensive at all' answers exhibited in the first two rows of the table in footnote 8. In the first row are the answers of Hungarian ethnics vis-à-vis the slur 'bozgor', while in the second row are the answers of Jewish ethnics vis-à-vis the slur 'jidan'. While for both slurs there is a non-negligible percentage of 'not offensive at all' responses (19.6 and 22.9, respectively), the percentage of Roma ethnics that didn't find 'țigan' offensive at all is markedly higher (38.3%). Importantly, members of neither of the first two ethnic groups (Hungarian ethnics and Jewish ethnics) regularly use these words to identify themselves or others from their communities as members of a specific ethnic group, which suggests that such identification does play a role indeed in explaining the higher percentage found with Roma ethnics. However, as noted above, the data is problematic. 16 I should make it clear that, like many authors in the debate, I'm interested in cases in which a slur can have both a derogatory and a appropriated use at the same timeso, in synchronic, not diachronic semantics. Thus, I'm not interested in cases of slurs that have been successfully appropriated (such as, according to many theorists, 'queer'). It is plausible that the term acquired a new meaning via the process of appropriation. On this analysis, queer became semantically ambiguous upon appropriation. Initially, it was non-ambiguous, its only linguistic standing as a slur, one whose derogating capacity can be accounted for semantically. Later, it came to have another conventional use, one that is non-pejorative. (Jeshion 2013, 250).
Appeal to ambiguity, however, is not a satisfactory solution in relation to appropriated uses. While some slurs seem to pass the usual tests for ambiguity (e.g. Sennet 2016) 18 , other considerations militate against the idea. For example, taking slurs to be ambiguous leaves an important part of the process of appropriation out. As Rappaport writes, '[i]t is important to recognize that the power of appropriation comes in part from the fact that it is the same word that is being appropriated (2019, 17), while Tirrell nicely expresses the same idea when she says that an appropriated slur 'is the same word, with the same history, but with a new future' (1999,60). Related, and perhaps more importantly, ambiguity doesn't work in the context of appropriation (in the political sense) because it simply defeats its purpose. Arguably, while at the beginning of the process the aim of appropriating a slur is to make the members of the target group use it with a positive meaning, ultimately the aim is to make everyone 17 As the following quotes illustrate, the authors involved don't clearly distinguish between ambiguity in the sense of homonymy and in the sense of polysemy. While in both cases there are several meanings attached to a word (or at least to a unique string of symbols), in the former those meanings are disconnected, while in the latter they are interrelated. (Compare 'bank' -financial institution/river area with 'bottle'container/content, for example.) I interpret the authors mentioned below as proposing that slurs are homonymous, while according to the view I offer slurs are polysemous. 18 However, given that the accuracy of those tests has been widely disputed, not much weight should be put on this.
use it with that meaningthat is, to eradicate one of the meanings altogether. But this aim couldn't be achieved if the slur is construed as ambiguous. For what would be the basis on which to claim that one meaning of the term is mistaken and urge everyone to replace it with another meaning? In the case of 'bank' or other clearly ambiguous terms, the reaction to one's apparently improper use of the term with a certain meaning can at most be a requirement to clarify that meaning, and not to eradicate it. The same should be true if appropriated slurs are taken to be similar to 'bank'. Yet, things are different in the case of appropriated slurs: what a target group aims at by slur-appropriation is precisely replacing one meaning with another. Importantly, I don't think that what is known as 'the reclamation worry', originating in remarks by Anderson and Lepore 2013a and further discussed by quite a few authors (Ritchie 2017;Cepollaro 2017;Anderson 2018;Burnett 2020;Jeshion 2020) is a real problem for semantic views that appeal to ambiguity. In a nutshell, the problem Anderson and Lepore see is that mere postulation of ambiguity leaves a host of questions unanswered, such as who is allowed to use a slur in an appropriated manner, who is excluded from doing so, and why is that so. While these are no doubt important questions, it is not clear that they need to fall under the purview of semanticsan explanation in extra-linguistic terms based, for example, on social, moral and political considerations would work just fine. Additionally, several answers have been given to the worry, some of them from the standpoint of a semantic theory. 19 So, while I take ambiguity to be problematic, I do so for different reasons that those invoked by the reclamation worry.
Appropriated uses create problems not only for semantic accounts but for others as wellfor example, for 'pragmatic' approaches to slurs: accounts that take the derogatory element of slurs to be contributed by various pragmatic mechanisms associated with a (use of a) slur. While they might have an easier time than semantic views, there is considerable footwork to be done in order to account for appropriated uses. Views like Williamson's 2009 or Whiting's 2013, for example, that take the derogatory 19 Thus, Cepollaro 2017 argues that a simple explanation involving speakers' intentions and context is available to proponents of the semantic view. Both Cepollaro 2017 and Jeshion 2020 argue that the worry doesn't arise with other cases (e.g., honorifics) in which restrictions on the use of ambiguous words are present. Burnett 2020 claims that a derogatory attitude is or is not inferred (thus, not part of the linguistic meaning) based on the ideologies one subscribes to. A response amounting to a purely semantic proposal is Ritchie 2017, who takes slurs to be ambiguous between a derogatory and a non-derogatory meaning, the latter having an indexical element built in (basically, a 'we' that assures that the speaker is part of the right group). Ingenuous as it is, I find Ritchie's view ad-hoc. See also Cepollaro 2017 and Jeshion 2020 for further, and more substantial, criticism.
content of a slur to be contributed via an implicature to the effect that the target group is worthy of derogation need to provide an explanation of why the relevant implicature is not generated in appropriated uses. Further, views like Cepollaro's 2015; Cepollaro and Stojanovic's 2016; Marques and García-Carpintero's 2020 that take the derogatory content of a slur to be contributed via a presupposition to the same effect need to explain how come such uses are felicitous without the relevant presupposition being in place. Other views that don't fall squarely within the semantic/ pragmatic distinction are not without problems either: thus, the echoic view proposed by Bianchi 2014, while devised precisely with the aim of accounting for appropriated uses, has been criticized for giving a merely partial explanation of the phenomenon (by, e.g. Jeshion 2020), while Anderson and Lepore's 2013a, 2013b prohibitionist view seems to provide a good account of the uses at stake, but it has been rejected by the majority of the authors involved in this debate for independent reasons. Accounting for appropriated uses thus seems to be at least prima facie problematic for many of the views on the market. Now, the authors involved in the debate over the right account of slurs have not considered identificatory uses, but it is fair to say that the same solutions they offer for appropriated uses would probably be applied to these too. For example, it seems natural for proponents of semantic views to appeal to ambiguity, as in the case of appropriated uses; and although this is speculative, it is not unreasonable, given that identificatory uses pose a very similar problem to that posed by appropriated ones. The same considerations apply to pragmatic and the other types of views mentioned earlier: accounting for identificatory uses is not trivial and requires significant work. I thus conclude that it is very likely that most views currently on the market have troubles with accounting for the whole range of data, including identificatory uses. This, I believe, mandates the search for a better theory.
Before going on to sketch the view I prefer, let me first situate it. At one extreme we have ambiguity. We have seen that many of the extant views on the market (especially semantic views) would appeal to it in accounting for identificatory uses. We have also seen several reasons for which postulating ambiguity doesn't amount to a very satisfactory solution in the case of appropriated uses. Since we aim for a theory that accounts for the whole range of the data, this won't do. At the other extreme, one could embrace a radical contextualist view 20 , according to which 20 For example, the view known as 'truth-conditional pragmatics', defended, among others, by Recanati 2004 (although, for Recanati, speaker intentions are not the only contextual factor determining whether a certain word carries derogatory force depends mostly on the intentions, beliefs and attitudes of the speaker instead of being encoded in language. While such a view would perhaps make the right predictions in the case of identificatory uses, I take it to be untenable in general for slurs because it runs afoul of one of their most well-established traits: independence (in the sense employed in section 1). My aim in the reminder of the paper is to present a view that steers between the two extremes mentioned and has the means to account for all the uses of slurs focused on here. The view is that slurs are polysemous.

A rich-lexicon theory of nouns
Polysemy has been studied preponderantly in lexical semantics, where two broad types of theories dealing with the phenomenon are the main contenders: rich-lexicon theories and thin-lexicon theories. 21 According to the former, words have as their linguistic meaning rich conceptual structures comprised of interrelated dimensions of meaning which get selected in a particular context; according to the latter, the linguistic meaning of words is very thin (at minimum, just a string of symbols), with more robust meanings being constructed with the help of contextual clues and encyclopedic knowledge. A rich-lexicon theory of linguistic meaning fits better my purposes in this paper, so I will adopt that framework without entering the debate of which of the two is superior. Now, there are numerous such rich-lexicon theories on the market. Which particular theory to adopt in accounting for polysemy is, to a certain extent, indifferent to my purposes here. I will, however, flesh out the view I favor by using a theory based on qualia and qualia-structures (e.g. Pustejovsky 1995): units of meaning that encapsulate certain types of information and which are interrelated in certain ways. While appeal to qualia-structures has been criticized as being insufficient to account for all the variability data (see Hogeweg and Vicente 2020 for an overview), such a framework remains widely used in lexical semantics and will fit the purpose of this paper. To illustrate, then, I will use the rich-lexicon framework developed by Del Pinal 2018. While the framework was designed to substantiate a meaning, his attitude being quite lax in this respect). In relation to slurs, Kennedy 2003 is taken to hold such a radical view. 21 Prominent examples of works in lexical semantics (although not with application to slurs) are Jackendoff 1990; Pustejovsky 1995; Asher 2011. For a review article of how polysemy is treated in lexical semantics, see Falkum and Vicente 2015; and Hogeweg and Vicente 2020 for a detailed discussion of arguments in favor of each type of theory. For a nuanced view that is pluralistic (that is, accepts both rich-and thin-lexicon theories as applying to different types of words), see Vicente 2018. truth-conditional pragmatic view, it can be easily used to show how a rich-lexicon theory can deal with polysemy. Del Pinal's main idea is that the meaning of words can be represented by a tuple consisting in an extension ('E-structure') and a conceptual structure ('C-structure'). The conceptual structure comprises the various quale associated with a word. In the case of nouns, the structure is taken to comprise at least perceptual information about the objects referred to with the noun (call that dimension PERCEPTUAL), information about what those objects are made of or their parts (CONSTITUTIVE), how they came to being or the purpose of their creation (AGENTIVE), their typical function (TELIC). To put a bit more flesh on the bones, let me give a well-trodden example. Take the noun 'book'. Applying the abstract framework laid out above, the lexical entry for 'book' under the rich-lexicon theory considered is something along the following lines: The precise components of the conceptual structure will of course differ across types of words: while the one above is customary for nouns, the one usually associated with a verb, for example, might comprise completely different meaning dimensions. While it is no doubt important for a lexical semantic theory to mark these differences and draw a complete map of the meaning dimensions postulated, this is not my aim here. The example of 'book' will be enough for my purposes. Now, while the full linguistic meaning of a word comprises the conceptual structure with all its dimensions, in a particular use of a word in a given context certain dimensions are highlighted in the detriment of others. This, in essence, is a mechanism of selection. Although clearly important, I won't commit here to any particular way of conceiving this mechanism. As with choosing a particular rich-lexicon theory, the main purpose of this paper can be achieved if I remain neutral on this score, too. The mechanisms appealed to by Pustejovsky 1995; Asher 2011; Del Pinal 2018 or those described in Frisson 2009or Schumacher 2013 are all plausible candidates although, of course, which of them is the most adequate, both generally and for the task at hand, is something that remains to be established. 22 22 What I take these various mechanisms to have in common is that they are cognitive in nature, and are activated in the course of the compositional process that yields meaning to complex sentences from the meanings of their parts in context. The selection of meaning dimensions happens at the lexical Thus, instead of choosing one of the various mechanisms found in the literature, I propose to use placeholders. I will use the generic term 'foregrounding' to designate the process by which a certain meaning dimension is selected or highlighted in a certain context, and the generic term 'backgrounding' to designate the process by which a certain meaning dimension is taken out of the spotlight. To come back to our example, in a sentence like 'This book is bound in leather', the CONSTITUTIVE dimension of the lexical meaning of 'book' is foregrounded and the other dimensions are backgrounded: what is important in this context is what the book, in the physical sense, is made of. In contrast, in a sentence like 'This book is interesting', the AGENTIVE dimension is foregrounded while the other dimensions are backgrounded: what is important in this context is the content of the book and how it affects the reader. Notably, foregrounding and backgrounding take place even in the absence of linguistic clues from the context (a role played by the predicates 'bound in leather' and 'interesting' in the sentences above): a sentence like 'This book is great' could mean that the physical copy of the book is great, but also that it is great content-wise. 23 These mechanisms explain how words can get different meanings in different contexts, while preserving their full linguistic meaning. Importantly, this avoids postulating ambiguity: we are not talking here about two different words that happen to consist of the same strings of symbols, but about one word, characterized by a rich structure of interrelated meanings out of which a particular one (or, in some cases, several) is selected in a given context.

Applying the theory to slurs 24
The first step in applying this rich-lexicon theory to 'țigan' in particular and to slurs in general consists in the introduction in the conceptual structure associated with the relevant words of an evaluative dimension (EVA-LUATIVE) responsible for the slurs' derogatory force. 25 What exactly level (thus, before the meaning of the word enters into the compositional process) and so is not a case of 'pragmatic intrusion', or a 'secondary pragmatic process' of the type appealed to by Recanati 2004, for example. I will not address the issue of how composition works in the framework proposedalthough I agree it is a crucial issuebut see Del Pinal 2018 for details. 23 'Great' is, most likely, polysemous too. 24 Other explicit polysemy accounts of slurs are rare: see Ritchie 2017 andJeshion 2020. However, Ritchie is ambivalent between an ambiguity and a polysemy account, while Jeshion prefers a thin-lexicon theory (p.c.). I briefly engage with two more developed rich-lexicon theories in footnote 28 below. 25 Del Pinal and Reuter 2016 postulate a normative dimension in the lexical entry of expressions like 'father', 'scientist' or 'woman'. This idea pertains to an independent research programme taking various types of expressions in natural language to encode 'dual character concepts', whose main trait is that they have both a referential aspect (they refer to a class of objects) and a normative one (they encapsulate certain normative properties that makes the objects within that class to be what should be packed into this meaning dimension is a contentious issue. Arguably, it cannot contain merely a subjective-expressive component that would not properly differentiate slurs from other types of expressions (e.g. expressives like 'jerk' or 'asshole'). I take the EVALUATIVE dimension to contain elements of a more social nature, such as stereotypes (Williamson 2009), perspectives (Camp 2013) or ideologies (Burnett 2020) that ground the negative evaluation of the target group by users of a slurs. The stereotype associated with 'țigan' for example, is something like 'lazy, prone to stealing, unwilling to respect social norms etc.', which encapsulates the perspective of a racist ideology propagated by the bigoted majority. While this is merely an approximation, I take it to be enough to give an idea of what the EVALUATIVE dimension could consists in. The second step is determining what the other meaning dimensions in the case of slurs are. While there are considerable differences between 'book' and 'țigan', they are both nouns, and so we can use (at least as a starting point) some of the meaning dimensions listed in the lexical entry provided above for 'book'. Not all meaning dimensions that are present in the lexical entry of 'book' are needed in the case of 'țigan': for example, since the latter picks up a social group, postulating a TELIC dimension is not warranted (what would be the purpose of a social group?). Also, since 'țigan' refers to people belonging to an ethnic group, with a certain provenance, history and social standing, the introduction of a meaning dimension that encapsulates this aspect (call it ORIGIN) is desirable. 26 The other meaning dimensions from the lexical entry of 'book' can be kept. Thus, I propose to conceive of the lexical entry for 'țigan', under the rich-lexicon theory considered, as something along the following lines: This duality resembles that of slurs, which suggests a possible treatment of the latter along these lines. In principle, I am open to the idea of treating slurs as encoding dual character concepts. Whether this ultimately works requires inquiring into the relation between the evaluative dimensions slurs encode and the normative one postulated in the dual character concepts framework. I will leave this inquiry for another occasion, but for a recent proposal in that direction (although of a different kind than the one pursued here), see Marques and García-Carpintero 2020. 26 Thanks to Max Kölbel for suggesting this label.
There is surely room for further amendments and additions in light of future findings in cognitive science, linguistics and the theory of concepts, but I take this entry to serve as a sufficiently clear example of how the richlexicon theory sketched in the previous section can be applied to slurs.
When it comes to accounting for the various uses tackled in this paper, the same mechanisms appealed to in the case of 'book' and exemplified in section 4 will be used here too. In other words, the meaning dimension that is relevant for a use of a slur in a specific context is selected via foregrounding and backgrounding. Let's start with derogatory uses. When someone uses 'țigan' in this way, the EVALUATIVE dimension is foregrounded, while the other meaning dimensions are backgrounded. This can be graphically represented as follows (with the 'c' signifying that the word is used in a context and the bolded material signifying the meaning dimension that is selected in c):  The difference between using 'țigan' in a derogatory way and using it in an identificatory way thus consists in different meaning dimensions being highlighted in the two contexts, and thus selected as the meaning of the word in each context. Importantly, as with 'book', this avoids postulating ambiguity: we are not talking here about two different words that happen to consist in the same strings of symbols, but about one word, characterized by a rich structure of interrelated meanings out of which a particular one (or, in some cases, several) is selected in a given context. What goes for 'țigan' goes, mutatis mutandis, for other slurs used in an identificatory way.
What about appropriated uses? In section 2, I have argued that the data involving the use of 'țigan' by Roma ethnics in Romania pertains to a sui-generis, irreducible type of use, and that that use is different from appropriated uses. But can the rich-lexicon theory sketched above be applied to the latter as well? A positive answer to this question will lead to an important gain in terms of economy and theoretical unity.
Indeed, the theory can be minimally modified to account for appropriated uses too. The point where the modification has to be made concerns the EVALUATIVE dimension. As it stands now, EVALUATIVE captures the speakers' negative evaluation of members of the target group; but, given that in appropriation a positive evaluation is communicated instead, what has to be done is to allow it to become part of the EVALUA-TIVE dimension. To do this, I appeal to the notion of 'valence' (the type of evaluation, positive or negative, of an object): instead of directly encoding a negative evaluation in EVALUATIVE, we allow for the evaluation to have a certain valence, whose value will be determined in context. Thus, when someone uses a slur in a derogatory way, not only is the EVALUATIVE dimension foregrounded, but the value of the valence is a negative evaluation by the speaker of members of the target group. When someone uses a slur in an appropriated way, the EVALUATIVE dimension is also foregrounded (and not backgrounded as in the case of identificatory uses), but with the valence inversed, its value being now a positive evaluation by the speaker of members of the target group. To illustrate 27 , an appropriated use of 'țigan' in a context will thus be represented as follows: Thus, all three types of uses of slurs focused on in this paper can be accounted for in the rich-lexicon theory proposed. To repeat, understanding slurs as polysemous and allowing different meaning dimensions to be selected in different contexts via foregrounding and backgrounding avoids the postulation of ambiguity, preserving a unique linguistic meaning for slurs (consisting in a rich structure of interrelated meanings), despite the many uses they can be put to. This is one of the advantages that a polysemy account has over 27 As mentioned before, members of the Roma community are not keen on appropriating 'țigan'; however, I'm using this example to keep with the previous discussion and to better highlight the differences between appropriated and identificatory uses. merely postulating ambiguity: it is the same word, with the same linguistic meaning, that is up both for appropriation and identification. 28

Further issues
Before closing, I want to briefly address three issues that, prima facie at least, arise for the view put forward. The first concerns the role of speaker intentions in determining the semantic content of slurs. As stressed, I take it to be crucial that a theory of slurs respects independence the fact that the derogatoriness of slurs doesn't depend solely on the intentions, attitudes and beliefs of speakers. But, on the other hand, it seems that the speakers' intentions do matter in the framework adopted: namely, in deciding what meaning dimensions from the lexical entry of a slur are foregrounded and which are backgrounded in a certain context. This leads to a tension. I certainly agree that intentions play a role in the way in which meaning dimensions are selected from the rich lexical entries that slurs encode. But I claim that this role is highly constrained. The constraints that I think are at work here are of a social, moral and political nature. In other words, what meaning dimensions are selected in a certain context is a matter of the speaker going through the corresponding cognitive process, but whether the selection of a particular meaning dimension is legitimate in 28 Many of the semantic views on the market claim that the meaning of slurs is rich, at least in the sense that they comprise a descriptive component and an expressive/evaluative one. Some authors add further 'levels' of information that play a role in their account of slurs (e.g., Jeshion 2013). The view I propose is not essentially different from those, except perhaps by making the meaning dimensions more explicit and more structured. However, several views have postulated lexical entries for slurs that are more fine-grained, and thus closer to the view sketched above. I briefly present two such views. According to Croom 2011Croom , 2013, slurs encode rich conceptual structures that have as their meaning dimensions properties, both positive and negative, that the prototypical members of the target groups are taken to possess. Croom holds a family resemblance conception of category membership, according to which there is no essential property that all members of the target group need to share for the slur to apply, but that different members share some of the properties associated with them. The properties are ranked; this ranking, however, is context-sensitive and helps in communicating with slurs: a speaker selects among the properties in question those which are suitable for a given communicative situation. When a slur is used derogatorily, for example, the negative properties prototypical members of the target group are taken to have are selected, while when used appropriately, the positive ones are (or the negative ones are evaluated positively).
Neufeld 2019 proposes a richly structured 'essentialist' theory of slurs, according to which they are 'failed kind terms'that is, with null extension. The novelty of the view is the claim that part of the meaning dimensions of a slur is an 'essence' that is given a causal-explanatory role for the negative traits that members of the target group are perceived to have. That is, what explains and causes the members of the target group to have the negative traits perceived is precisely this 'essence' ('blackness', 'queerness' etc.); but since no such real essences exist, the terms fail to refer. Neufeld offers a view that features several meaning dimensions (negative stereotypes, essence, causality) but, in addition to Croom, she takes these to be causally organized in the way described. A detailed comparison between these three members of the rich-lexicon family, as well as an assessment of which handles the data presented in this paper better, would no doubt be of interest, but due to reasons of space it will have to be left for another occasion. that context is a matter pertaining to social, moral and political considerations. In this way, the selection of a meaning dimension is something that goes beyond the intentions, attitudes and beliefs of the speaker. To be sure, there is an interconnection between these, but the latter don't, in themselves, determine the former. This way of understanding the constraints in play accounts for the intuition that, for example, the (derogatory) use of a slur by a member of a dominant group to derogate a member of a subordinated group is illegitimate, as well as for the intuition that the (appropriated) use of a slur by a member of a subordinated group to fight injustice, or to express camaraderie or endearment towards another member of the same group is legitimate. Similarly for the intuition that members of certain ethnic groups (like Roma ethnics) can legitimately identify using a word that is a slur (like 'tigan'). Looking from the other end, this doesn't mean that the role many authors (e.g. Anderson 2018; Popa-Wyatt and Wyatt 2018; Hess 2020) think social, moral and political considerations play in a slur's life is downplayed: just because they are given a constraining, rather than a fully semantic, role doesn't mean they are not important. 29 While such moral, social, and political considerations introduce a certain degree of relativity into the picture (which is more or less to be expected), independence is saved. 30 The second issue I want to address concerns the equivalence between the truth-conditions of sentences containing slurs vs. sentences containing their neutral counterparts. Their presumed equivalence has been used as an argument against many extant theories of slursmore recently by Neufeld 2019, but also by Croom 2015. One of the main reasons this equivalence has been taken to be problematic is that it blocks an account of what is knownas'G-extending'usesofslurs:basically,theuseofaslurinordertoderogate someone who is not part of the target group. Since such uses are widespread, not accounting for them in a suitable manner limits a theory's power.
Does this criticism apply to the view put forward here? It doesn't, and the reason is simple: that the lexical entries that the view assigns to slurs and to their neutral counterparts are different. As we have seen, in the 29 Although, as we have seen in discussing the EVALUATIVE dimension in section 5, they permeate the semantics proper as well, via stereotypes, perspectives or ideologies, which are essentially social/ moral/political phenomena. 30 Two reviewers ask whether ambiguity views cannot help themselves with the same move. I can't see why notonce they answer the worries about ambiguity spelled out in section 3 above, that is. What is important to keep in mind is that, under the polysemy view that I offer, the derogatory character of slurs is always part of their lexical meaning, regardless of the way it is used in a certain context, while this is not so in ambiguity approaches. In any case, even if this is not taken to be an advantage, a polysemy account is interestingly different from an ambiguity one and independently worthy of investigation (see also my brief remark at the end of section 1).
case of 'tigan' the EVALUATIVE dimension is always present: foregrounded in derogatory/appropriated uses, backgrounded in identificatory uses. However, no such dimensions appears in the lexical entry of 'rrom' (the politically correct term used by Roma activists which we can take here to serve as the slurs' neutral counterpart); in fact, the latter has all the meaning dimensions 'tigan' has, except EVALAUTIVE. Since the lexical entries of the two words are different, the truth-conditions of the sentences they appear in are not equivalent. 31 As for G-extending uses, the view yields those as well: when, for example, 'tigan' is used in that way, the ORIGIN dimension is backgrounded, thus freeing the slur for application to people that are not members of the Roma community. The third issue tackled, also used as an argument against many current theories of slurs by Neufeld 2019, but also by Hom and May 2013, is whether the sentences containing slurs are, strictly speaking, true. This would be problematic especially when slurs are used derogatorily because it presupposes that there is something in the world that has all the negative traits that members of the target groups are perceived to have. In the case of the two works cited, the problem doesn't arise because the extension of the slurs is empty. I am not sure what the best response to this worry is, but what I want to point out is that the tables can be turned: an important question for views that postulate null extensions for slurs is how to account for appropriated uses in particular and for non-derogatory uses in general. This seems to be notoriously difficult for views like these, because sentences containing slurs that are used in an appropriated manner are intuitively true. Hom's preferred solution is appeal to ambiguity, while Neufeld doesn't address appropriation at all. While I concede that an answer to the worry should be given, the issue of which of the opposed families of views fares better is far from settled.

Summary and conclusions
This paper had two aims. The first was to bring to light certain data about the use of the Romanian slur 'țigan' by Roma ethnics from Romania. The data, 31 I haven't provided any precise description of how various meaning dimensions affect truth-conditions, but the short (and trivial) answer is: by giving a specific meaning of a certain word, which then enters into the compositional computation of the entire sentence. For example, ORIGIN specifies a meaning that applies to people having the corresponding provenance, PERCEPTUAL a meaning that applies to people having the corresponding shape, size, color, etc. EVALUATIVE specifies a meaning that applies to people that meet the stereotype the speaker associates with members of the target group. Obviously, a more detailed account is needed. For one that might be used in the present case, see Del Pinal 2018.
gathered from a Romanian Government report from 2009, shows that a relevantly large number of Roma ethnics use the word to identify both as individuals and as a group (hence, the name 'identificatory' for this type of use). Further, I argued that this data cannot be subsumed under types of uses of slurs previously discussed in the literature (in particular, appropriated and referential uses) and that a better option is to claim that they point to a sui-generis, irreducible type of use of slurs. The extent, underlying conditions, and further characteristics of this use are still to be inquired into, and one would expect cross-linguistic studies to be of help here. Identificatory uses of slurs show that there is even more variation in the use of slurs than previously acknowledged. The second aim of the paper was to provide an account that explains this variation. The main idea behind the account I sketch is that slurs are polysemous. While this has been claimed more or less explicitly by various authors in the literature, there are relatively few detailed accounts to be found. I have chosen to implement this idea by appealing to a rich-lexicon framework borrowed from lexical semantics, according to which the lexical entries of nouns are comprised of several meaning dimensions, among which the specific meaning of a noun in a context is selected. While I remained neutral about the exact mechanism by which this selection is achieved (using the placeholders 'foregrounding' and 'backgrounding'), I have shown how the framework can be applied to 'țigan' and its uses, and to slurs in general. Finally, I have briefly discussed three pressing issues. While many details remain to be ironed out, and various questions to be addressed, the paper brings both new data to the fore and polysemy views of slurs into sharper focus. seminars at various institutions for their comments, objections and suggestions. Two reviewers for this journal have provided essential feedback, as well as encouragement. I also thank Mircea Dumitru, Oana Ungureanu, Sorin Costreie, Simina Bădică, Cătălina Tesăr, Mihaela Vețan, Sergiu Florean, Adrian Briciu, Julia Zakkou, Sara Tătaru, Aladin Tătaru, Lidia Negoi, Agustin Vicente and Sorin Stan for various ways in which they helped. Special thanks go to Delia Grigore and Gelu Duminică, both Roma activists, who have spent their valuable time discussing various issues tackled in this paper with me. Research for this paper has been funded by a Mobility Grant for Researchers from Diaspora from the Romanian Ministry of Education and Research (UEFISCDI, project number PN-III-P1-1.1-MCT-2016-0019). I also acknowledge the financial help of a Lise Meitner grant from the Austrian Science Fund (FWF, project number M2226-G24).

Disclosure statement
No potential conflict of interest was reported by the author(s).