The semantics of slurs: A refutation of coreferentialism

Coreferentialism refers to the common assumption in the literature that slurs (e.g. faggot) and descriptors (e.g.male homosexual) are coreferential expressions with precisely the same extension. For instance, Vallee (2014) recently writes that ‘‘If S is an ethnic slur in language L, then there is a non-derogatory expression G in L such that G and S have the same extension’’ (p. 79). The non-derogatory expression G is commonly considered the nonpejorative correlate (NPC) of the slur expression S (Hom, 2008) and it is widely thought that every S has a coreferring G that possesses precisely the same extension. Yet here I argue against this widespread assumption by first briefly introducing what slurs are and then considering four sources of supporting evidence showing that slurs and descriptors are in fact not coreferential expressions with precisely the same extension. I argue that since slurs and descriptors differ in their extension they thereby differ in theirmeaning or content also. This article additionally introduces the notion of a conceptual anchor in order to adequately account for the relationship between slurs and descriptors actually evidenced in the empirical data, and further considers the inadequacy of common dictionary definitions of slurs. This article therefore contributes to the literature on slurs by demonstrating that previous accounts operating on the assumption that slurs and descriptors are coreferential expressions with the same extension, and that they thereby have the same meaning or content, are inconsistent with empirical data and that an alternative account in accord with Croom (2011, 2013a, 2014b) better fits the facts concerning their actual meaning and use. © 2015 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The humanities and social sciences have recently witnessed an explosion of fascinating new research on slurs and derogatory language (see for instance Cepollaro, 2015;Croom, 2015a,b;Jay and Jay, 2015;Beaton and Washington, 2014;Blakemore, 2014;Cupkovic, 2014;Jackson, 2014;O'Dea et al., 2014;Saucier et al., 2014;Weissbrod, 2014;Embrick and Henricks, 2013) and one widely held assumption in the literature -call it coreferentialism -is that slurs (e.g. faggot) and descriptors (e.g. male homosexual) are coreferential expressions with precisely the same extension (see for instance Dummett, 1973, p. 454;Hornsby, 2001, p. 129; first briefly introducing what slurs are and then considering four sources of supporting evidence showing that slurs and descriptors are in fact not coreferential expressions with precisely the same extension. I argue that since slurs and descriptors differ in their extension they thereby differ in their meaning or content also. This article additionally introduces the notion of a conceptual anchor in order to adequately account for the relationship between slurs and descriptors actually evidenced in the empirical data, and further considers the inadequacy of common dictionary definitions of slurs. This article therefore contributes to the literature on slurs by demonstrating that previous accounts operating on the assumption that slurs and descriptors are coreferential expressions with the same extension, and that they thereby have the same meaning or content, are inconsistent with empirical data and that an alternative account in accord with Croom (2011Croom ( , 2013aCroom ( , 2014b better fits the facts concerning their actual meaning and use. 2

Some basic features of slurs and their use
Slurs such as nigger, cracker, kike, chink, and slut are linguistic expressions that are primarily used and understood to derogate certain group members on the basis of their descriptive attributes (such as their race or sex) and expressions of this kind have been considered by many to pack some of the nastiest punches natural language has to offer. In discussions concerning the history and use of various race-directed and sex-directed slurs, Lemon (2013) explains that the slur nigger is a ''dark, degrading hateful insult for African Americans'', Foreman (2013) explains that the slur cracker ''is a demeaning, bigoted term [. . . ] a sharp racial insult that resonates with white southerners [. . . ] offensive and evidence of ill intent'', Verna et al. (2007) explain that the slur kike is ''a term of abuse for Jews'' whereas the slur chink is ''a term of abuse for Asians'' (p. 468), and Blackwell (2004) explains that the slur ''''slut'', a charge easy to level and hard to disprove, is an ambivalent emblem of women's perception of their sexuality'' (p. 141, my emphasis). As Anderson and Lepore (2013a) write, there are a wide variety of slurs ''that target groups on the basis of race ('nigger'), nationality ('kraut'), religion ('kike'), gender ('bitch'), sexual orientation ('fag'), immigrant status ('wetback') and sundry other demographics' ' (p. 25). So one basic fact about slurs is that an application of a particular slur in context occurs based on considerations of its systematic differential application-conditions, which concern descriptive attributes of targets such as their racial or sexual identity.
Another basic fact about slurs is that they are often considered to be among the most offensive of all linguistic expressions. The racial slur nigger, for instance, is commonly identified as ''one of the most racially offensive words in the language'' (Oxford English Dictionary, 2014) and Christopher Darden has popularly characterized the slur as the ''filthiest, dirtiest, nastiest word in the English language'' (Kennedy, 2002, p. 23;Knowles, 2009). Fitten (1993) even proposes that slurs should be considered ''fighting words'' since they have often been used to initiate violence and carry out hate crimes, and Jeshion (2013a) further suggests that ''Slurring terms are used as weapons in those contexts in which they are used to derogate an individual or group of individuals to whom the slur is applied or the socially relevant group that the slur references' ' (p. 237, my emphasis). In discussions concerning the history and use of various slurs, Cole (2013) has also discussed how slurs have been used to initiate fights in hockey, McIntyre (2013) has discussed how slurs have been used to initiate fights in football, and Hoover (2007) has discussed how slurs have been used to initiate fights on 2 For proposed adequacy conditions for accounts of slurs see for instance Hom, 2008, p. 426;Croom, 2011, p. 355;Croom, 2013a, p. 200. college campuses. In yet other discussions concerning the history and use of various slurs, Islam (2011) discusses how a 7-year old boy killed one of his classmates for targeting him with a homophobic slur, Reifowitz (2013) discusses how the slur nigger was the last word thousands of African Americas heard before they were savagely lynched by white supremacists, and Kemp (2014) discusses how 4 perpetrators in Philadelphia were recently charged for kidnapping, assaulting, and forcibly tattooing racial slurs on the arms of their victim. Resultantly, prohibitions against slurs are often so strong now that Craver (1994) even reports a case where one man was fired from his job for merely listening to a radio station show that had used slurs in conversation. So another basic fact about slurs is that they are among the most potentially offensive expressions that natural language has to offer.

The traditional assumption that slurs and descriptors are coreferential expressions
Having now briefly reviewed some basic facts about slurs in the previous section, this section will proceed to critically assess the widespread assumption that slurs (e.g. faggot) and descriptors (e.g. male homosexual) are coreferential expressions with precisely the same extension (Dummett, 1973, p. 454;Hornsby, 2001, p. 129;Williamson, 2003, p. 261;Whiting, 2007, p. 192;Whiting, 2008, p. 385;McCready, 2010, p. 5, 9;Anderson and Lepore, 2013a, p. 26-27;Anderson and Lepore, 2013b, p. 351;Vallee, 2014, p. 79). In an early and influential discussion involving the slur boche, for instance, Dummett (1973) proposed that ''The condition for applying the term to someone is that he is of German nationality; the consequences of its application are that he is barbarous and more prone to cruelty than other Europeans'', and that ''We should envisage the connections in both directions as sufficiently tight as to be involved in the very meaning of the word: neither could be severed without altering its meaning'' (p. 454, my emphasis). Williamson (2003) similarly argued that the slur '''Boche' has the same reference as 'German''' (p. 261) and that therefore ''the differences between 'Boche' and 'German' apparently play no role in determining reference, and so make no difference to the way in which the terms contribute to the truth-conditions of sentences in which they occur'' (p. 261, my emphasis; see also Williamson, 2009Williamson, , 2010. Whiting (2008) also argues that ''the meaning of [the slur] 'Boche' is given by whatever inferential rules govern (and thereby determine the meaning of) 'German''' (p. 385) while Vallee (2014) further claims that ''the extension of ''German'' is the set of German people, as is the extension of ''boche'', and the extension of ''Chinese'' is the set of Chinese people, as is the extension of ''chink'' (p. 79). Luvell Anderson and Ernie Lepore have accordingly considered the expressions boche and German, and the expressions chink and Chinese, as examples of ''slurs and their neutral counterparts' ' (2013a, p. 26-27, my emphasis) which they consider to be ''co-referential expressions for the same group' ' (2013b, p. 351, my emphasis). 3 Other examples of ''co-referential expressions for the same group'' have also been suggested by Whiting (2007) and include ''the pairs 'faggot' and 'male homosexual','nigger' and 'black' and 'Kike' and 'Jew''' (p. 192). McCready (2010) similarly suggests that ''Kraut is a pejorative term for German people on its nominal use'' and that ''the expressed content of Kraut is roughly that German people are bad' ' (p. 5, 9).
Given that two expressions ↵ 1 and ↵ 2 are commonly considered coreferring expressions just in case referent(↵ 1 ) = referent(↵ 2 ) for the expressions ↵ 1 and ↵ 2 (van Deemter and Kibble, 2000, p. 629;Reimer, 2009) 4 the traditional assumption in the literature on slurs has been that the referent(boche) = referent(German) for the expressions boche and German, that the referent(chink) = referent(Chinese) for the expressions chink and Chinese, that the referent(faggot) = referent(male homosexual) for the expressions faggot and male homosexual, that the referent(kike) = referent(Jewish) for the expressions kike and Jewish, that the referent(kraut) = referent(German) for the expressions kraut and German, and that the referent(nigger) = referent(African American) for the expressions nigger and African American (see for instance Dummett, 1973, p. 454;Hornsby, 2001, p. 129;Williamson, 2003, p. 261;Whiting, 2007, p. 192;Whiting, 2008, p. 385;McCready, 2010, p. 5, 9;Anderson and Lepore, 2013a, p. 26-27;Anderson and Lepore, 2013b, p. 351;Vallee, 2014, p. 79). As Hom (2008) popularly proposed in ''The Semantics of Racial Epithets'', ''For any racial epithet [or slur], [we can] call its nonpejorative correlate (NPC) the expression that picks out the supposed extension of the epithet but without expressing derogation towards members of that extension. For example, the NPC of 'chink' is 'Chinese', the NPC of 'kike' is 'Jewish', the NPC of 'nigger' is 'African-American', and so on'' (p. 417, fn. 4; see also Hornsby, 2001, p. 129). Hom (2010) further proposed in ''Pejoratives'' that ''for any derogatory word, D, and its neutral counterpart, N [. . . ] to call someone a D is to say that they ought to be subject to discriminatory practices for having negative, stereotypical properties because of being an N' ' (2010, p. 174, 180, my emphasis). Finally, Whiting (2013) writes that ''slurs express the same semantic content when used as their neutral counterparts; that is, slurs and their neutral counterparts contribute the same thing to what is said by uses of sentences involving them'' (p. 364, my emphasis). Clearly a perusal of the extant literature on slurs shows that the assumption that slurs and descriptors are coreferential expressions with precisely the same extension is widespread. The purpose of this section was to make this point clear. In the next section my purpose is therefore to turn to consider four sources of evidence demonstrating the falsity of this widely held assumption.

Reconsidering the case that slurs and descriptors are coreferential expressions
Here I challenge prior scholarship on slurs by arguing for the claim that slurs and descriptors are not coreferential expressions with precisely the same extension, and accordingly, that the latter are not simply nonpejorative correlates (NPCs) of the former. The relationship between slurs and descriptors is more complex than that and I will clarify this relationship by introducing the notion of a conceptual anchor in section five. But first let us consider in this section four sources of supporting evidence showing that slurs and descriptors are in fact not coreferential expressions with precisely the same extension. The first source of supporting evidence suggesting that slurs and descriptors are not coreferential expressions with precisely the same extension comes from the discussion Szekely (2008) provides in ''Offensive Words''. Here Szekely (2008) discusses how slurs were used in his linguistic community and reports that the slur faggot was in fact used to apply to some but not all male homosexuals. As Szekely (2008) Croom, 2011, p. 352) It is evident from this discussion that Szekely (2008) provides that the slur faggot is primarily used by speakers to apply (if at all) to some male homosexuals -namely, just those with attributes that the speaker does not consider acceptable or equal but rather as unacceptable or unequal -but not necessarily to them all. This is the point that Szekely (2008) makes in claiming that ''I would never call a gay guy a faggot, unless he was being a faggot. But not because he's gay, you understand''. Further, the fact that the slur faggot is differentially used so that it is often applied to some but not all male homosexuals suggests that the slur faggot and the descriptor male homosexual are in fact not coreferential expressions with precisely the same extension at all. In other words this discussion by Szekely (2008) suggests that it is not the case that the referent(faggot) = referent(male homosexual) for the expressions faggot and male homosexual, as it is widely assumed in the literature.
The second source of supporting evidence suggesting that slurs and descriptors are not coreferential expressions with precisely the same extension comes from the discussion MacDonald (1999) provides in All Souls: A Family Story from Southie. Here MacDonald (1999) discusses how slurs were used in his linguistic community and reports that the slur nigger was in fact used to apply to some but not all African Americans (p. 61). As MacDonald (1999) explains in All Souls: I spent hours in our apartment in Old Colony trying to grasp this hierarchy of niggers that I'd discovered. I wanted to know exactly where I fit into the scheme. Of course, no one considered himself a nigger. It was always something you called someone who could be considered anything less than you. I soon found out that there were a few black families living in Old Colony. They'd lived there for years and everyone said that they were okay, that they weren't niggers but just black. It felt good to all of us to not be as bad as the hopeless people in D Street or, God forbid, the ones in Columbia point, who were both black and niggers. But now I was jealous of the kids in Old Harbour Project down the road, which seemed like a step up from Old Colony, having many families left over from when housing projects were for war veterans, and when some of the kids had fathers. Of course, we were all niggers if we went to City Point, so forget going there again to see the beautiful beaches and Castle Island. I wondered if the Point kids might be niggers to people who'd really made it, like out in tidy West Roxbury or the suburbs that everyone talked about moving to when they won the lottery. (p. 61, my emphasis) The ''few black families living in Old Colony [. . . ] weren't niggers but just black '', MacDonald (1999) explains, because ''They'd live there for years and everyone said they were okay'', whereas in contrast with the comparatively neutral descriptive expression black, the more affectively forceful slur expression nigger was understood to be especially reserved as ''always something you called someone who could be considered anything less than you'' (p. 61, my emphasis; for discussion on the use of slurs for social distancing see also Croom, 2011;Croom, 2013a;Camp, 2013). So similarly to the case of sex-oriented slurs like faggot that Szekely (2008) discussed in ''Offensive Words'', it is evident in this case that race-oriented slurs like nigger are also used (if at all) to apply to some African Americans -namely, just those with attributes that the speaker does not consider acceptable or equal but rather as unacceptable or unequal -but not necessarily to them all. This is the point that MacDonald (1999) makes when he says about some of the African American families living in Old Colony -that ''they weren't niggers but just black'' (p. 61, my emphasis). Further, the fact that the slur nigger is differentially used so that it is often applied to some but not all African Americans suggests that the slur nigger and the descriptor African American are in fact not coreferential expressions with precisely the same extension at all. In other words this discussion by MacDonald (1999) suggests that it is not the case that the referent(nigger) = referent(African American) for the expressions nigger and African American, as it is widely assumed in the literature.
The third source of supporting evidence suggesting that slurs and descriptors are not coreferential expressions with precisely the same extension comes from the discussion Troyani (2013) provides in ''Guido Culture: The Destabilization of Italian-American Identity on Jersey Shore''. Here Troyani (2013) offers an informative discussion of the differential application-conditions for the slur guido and the descriptor Italian-American, which highlights the important fact that the application-conditions for typical targets of the slur guido involve attributes that saliently differ from the attributes involved in the application-conditions for typical targets of the descriptor Italian-American. So given the fact that there are saliently different application-conditions for the slur guido and the descriptor Italian-American, it is clear that the slur guido has in fact been used to apply to some but not all Italian-Americans. As Troyani (2013) explains in ''Guido Culture'': Some Italian Americans may aspire to live as Guidos or Guidettes. However, Guidos and Guidettes are not necessarily Italian American. The dubious Italian heritage of cast members suggests that if non-Italian Americans can adopt seemingly Italian-American Guido and Guidette characteristics and behaviours, these characteristics may also be achieved rather than inherited by Italian Americans. (p. 4, my emphasis) It is evident from this discussion that Troyani (2013) provides that the slur guido is primarily used by speakers to apply (if at all) to some Italian Americans -namely, just those considered to exhibit the saliently prototypical characteristics and behaviours of ''guido culture''-but not necessarily to them all. This is the point that Troyani (2013) makes when she claims that ''Some Italian Americans may aspire to live as Guidos or Guidettes. However, Guidos and Guidettes are not necessarily Italian American'' (p. 4). Further, the fact that the slur guido is differentially used so that it applies to some but not all Italian-Americans suggests that the slur guido and the descriptor Italian-American are in fact not coreferential expressions with precisely the same extension at all. That is to say, this discussion by Troyani (2013) suggests that it is not the case that the referent(guido) = referent(Italian-American) for the expressions guido and Italian-American, as it is widely assumed in the literature.
The fourth and final source of supporting evidence suggesting that slurs and descriptors are not coreferential expressions with precisely the same extension comes from the discussion Chris Rock (1996Rock ( , 1997Rock ( , 2002 provides in ''Niggas vs. Black People'', perhaps one of the most popular and controversial performances in the history of standup comedy (Hartsell, 2012;Hoggard, 2006;Farley, 2004;Schruers, 1997;Jeshion, 2013b, p. 309, 314-315). Here Rock (1996Rock ( , 1997Rock ( , 2002 offers an informative discussion of the differential application-conditions for the slur nigger and the descriptor black, which highlights the important fact that the application-conditions for typical targets of the slur nigger involve attributes that saliently differ from the attributes involved in the application-conditions for typical targets of the descriptor black (see for instance track 12 on Rock, 1997; track 8 on Rock, 2002). 5 So given the fact that there are saliently different application-conditions for the slur nigger and the descriptor black, it is clear that the slur nigger has in fact been used to apply to some but not all African Americans (Farley, 2004;Hoggard, 2006). As Rock (1996)  Niggas are shaking in their boots! ''Oh they're goin' to take our shit!'' Shit, a black man that's got two jobs, going to work every day, hates a nigga on welfare. Nigga get a job! I got two, you can't get one? (my emphasis, see also Rock, 1997, track 12 Rock, 2002 In their commentary on this highly controversial yet influential performance, Farley (2004) pointed out how ''Rock contrasted the values of middle class blacks with lower-income blacks who had succumbed to a kind of gangsta despair'' (my emphasis) while Hoggard (2006) also pointed out how ''Rock claimed it was socially acceptable for black people to refer to segments of the black population that degraded the black community through laziness and stupidity as ''Niggas'' [rather than ''Black People'']'' (my emphasis; see also Jeshion, 2013b, p. 309, 314-315). So similarly to the case that Szekely (2008) considered concerning the slur faggot, that MacDonald (1999) considered concerning the slur nigger, and that Troyani (2013) considered concerning the slur guido, it is also evident from this discussion that Rock (1996) provides in ''Niggas vs. Black People'' that the slur nigger is again primarily used (if at all) to apply to some African Americansnamely, just those with attributes that the speaker does not consider acceptable or equal but rather as unacceptable or unequal -but not necessarily to them all. In other words this discussion by Rock (1996) suggests that it is not the case that the referent(nigger) = referent(African American) for the expressions nigger and African American, as it is widely assumed in the literature.
We have now considered four sources of supporting evidence demonstrating the falsity of the widely held assumption in the theoretical literature that slurs and descriptors are coreferential expressions with precisely the same extension. By carefully considering empirical data on slur use, I have argued here that slurs and descriptors differ in their application conditions and thereby differ in their extensions. Further, because slurs and descriptors differ in their extensions they also differ in their meaning or content. The purpose of this section was to make this point clear. In the next section my purpose is therefore to turn to reconsider the relationship between slurs and descriptors and introduce the notion of a conceptual anchor to help provide a clearer understanding of their semantic or conceptual content.

Reconsidering the relationship between slurs and descriptors
In the literature on concepts or categories, a distinction is commonly drawn between classical accounts and family resemblance accounts (Rosch and Mervis, 1975). According to classical accounts, ''categories are defined by formal rules and allow us to make inferences within idealized law governed systems'' (Pinker and Prince, 1996, p. 332, my emphasis). So the classical account of categories maintains that category membership is determined by the possession of some common, essential, and criterial attribute. However, a substantive challenge for maintaining a classical account for all categories of natural language is that scholars remain unable to articulate necessary and sufficient conditions for most that are actually found from natural language (Fodor et al., 1980;Pinker and Prince, 1996;Rosch and Mervis, 1975). It is further clear from the extant literature on expressive expressions that this point applies a fortiori to slurs and general pejoratives (see for instance Potts et al., 2009).
Alternatively, and in contrast with the classical account, the family resemblance account of categories maintains that category membership consists of a relationship in which case ''each item has at least one, and probably several, elements in common with one or more other items, but no, or few, elements are common to all items'' (Rosch and Mervis, 1975, p. 575;Wittgenstein, 1953). Pinker and Prince (1996) usefully distinguish family resemblance from classical categories by pointing out several salient ways in which they differ. First, family resemblance categories differ from classical categories in that the former lack necessary and sufficient conditions for category membership whereas the latter do not. Second, family resemblance categories differ from classical categories in that the former have graded degrees of category membership whereas the latter do not. Third, family resemblance categories differ from classical categories in that the former can be summarized by an ideal category member or prototype whereas the latter cannot. Fourth, family resemblance categories differ from classical categories in that the former have category members that tend to have characteristic non-defining attributes whereas the latter do not. Importantly, the family resemblance account avoids the challenge faced by the classical account in that the former does not maintain as the latter does that concepts or categories are strictly definable in terms of necessary and sufficient conditions. Rather, the family resemblance account maintains that most concepts of natural language are characterizable in terms of their family resemblance relationship. Indeed, Pinker and Prince (1996) explain that family resemblance concepts are characterizable in terms of ''correlations among features in sets of similar memorized exemplars, and allow us to make inferences about the observable products of history'' (p. 353) and Rosch and Mervis (1975) similarly explain that family resemblance ''prototypes appear to be just those members of the category that most reflect the redundancy structure of the category as a whole. That is, categories form to maximize the information rich clusters of attributes in the environment'' (p. 602; Rosch et al., 1976). The family resemblance account of concepts or categories is therefore more realistic than the classical account insofar as it provides an account of concepts that is actually in accord with the real rather than ideal nature of human psychology.
For example, what makes x a member of the category MUSIC on the family-resemblance conception is not some essentially criterial attribute that each and every x must have in order to be categorized as music. For a paradigmatic or prototypical musical work may typically or for the most part have a chorus and acoustic instruments, but could still be felicitously and informatively categorized as music even if it did not have a chorus or acoustic instruments, provided that the category MUSIC is that which is still most strategically apt among other options (for example, RECIPE, KOALA, and so on) available to that speaker for their current cognitive or communicative purpose. So with regard to slurs, my proposal here follows Croom (2011Croom ( , 2013aCroom ( , 2014b in adopting a family resemblance conception of category membership to account for the fact that the felicitous application of the slur faggot need not be restricted to male homosexuals (Szekely, 2008), that the felicitous application of the slur guido need not be restricted to Italian Americans (Troyani, 2013), and that the felicitous application of the slur nigger need not be restricted to African Americans (MacDonald, 1999;Rock, 1996), or more generally, that the felicitous application of a slur S need not necessarily target someone instantiating an associated neutral descriptive attribute G (Croom, 2014c, fn. 20;Croom, 2011, p. 352-357;Croom, 2013a,b). More specifically, the alternative approach proposed by Croom (2011Croom ( , 2013aCroom ( , 2014b suggests that the expression nigger (identified as N below) can be usefully understood as a family resemblance (rather than classical) category consisting in a constellation of prototypical attributes (identified as a 1 -a 10 below) such as the following: a 1 x is African American. 6 a 2 x is lazy. 7 a 3 x is subservient. 8 a 4 x is commonly the recipient of poor treatment. 9 a 5 x is athletic or musical. 10 a 6 x is sexually liberal or licentious. 11 a 7 x is simple-minded. 12 a 8 x is emotionally shallow. 13 a 9 x is a survivor, tough, or aggressive. 14 a 10 x is loud or excessively noisy. 15 Note that I am not suggesting here that attributes a 1 -a 10 should be understood as fixed in the precise rank-order provided in the example above, or that all of a 1 -a 10 are always involved in a contextindependent manner. Indeed that suggestion would be more in accord with the classical approach to concepts or categories and contrary to the very family resemblance approach I advocate here. And as I briefly discussed earlier, the classical approach has been increasingly undermined by a growing body of recent empirical literature on the nature of human memory, concepts, and sensorimotor cognition (Barsalou, 1999;Barsalou et al., 2003;Borghi, 2004;Barsalou et al., 2007;Barsalou, 2008Barsalou, , 2009Borghi and Riggio, 2009;Borghi et al., 2013;Dove, 2010Dove, , 2014. The alternative then that I propose here is that attributes a 1 -a 10 should be considered as rank-ordered based on the relative degree in which their attribution to x is taken as a salient indicator of category membership, and importantly, that this rank-order is re-organizable in a contextdependent manner. Further, in practice more attributes (for example, a 1 -a 15 ) or less attributes (for example, a 1 -a 2 , or even a 1 alone) could be involved in a given cognitive or communicative context. 6 Fredrickson, 1971, p. 41 and Asim, 2007, p. 12. 7 Asim, 2007, p. 27. 8 Fredrickson, 1971, p. 41 and Asim, 2007, p. 12. 9 Fredrickson, 1971, p. 41 and Asim, 2007, p. 12. 10 Alim et al., 2010, p. 128. 11 Asim, 2007, p. 27. 12 Asim, 2007, p. 27. 13 Asim, 2007, p. 27. 14 Anderson, 1999, p. 50 and Rahman, 2012. 15 Anderson, 1999 I provide a list of 10 attributes here merely to be both optimally informative (for listing a 1 alone would fall short of this) and economical (for listing a 1 -a 20 would go beyond this) in our discussion.
My suggestion here is that a 1 (African American) would be ranked relatively higher than a 6 (sexually liberal or licentious) and accordingly a 1 would be considered a more salient indicator than a 6 that the x possessing it is a member of N (nigger). Note also that although speakers may typically apply the slur nigger to targets attributed the highest-ranking (a 1 ) as well as the greatest quantity (a 1 -a 10 ) of attributes in N, my family resemblance account maintains (contrary to classical accounts) that speakers may still informatively or effectively apply that slur to x even if that x fails to possess the highest-ranking (a 1 ) or even the most (a 1 -a 10 ) attributes in N insofar as that is the most relevant and apt lexical choice for their purpose in a particular cognitive or communicative context (for further discussion of relevance in communication and cognition see also Sperber and Wilson, 1986;Wilson and Sperber, 2004). Importantly, however, in order for the choice of a speaker to refer to x as a nigger to be considered a strategically apt choice for that speaker, it must be assumed (at least for the purpose of that particular situation) that x possesses a practically sufficient set of attributes such that N is the most appropriate or serviceable category for the speaker to subsume x for their purpose in a particular cognitive or communicative context. So for example, in the case that a speaker intends to communicate that some x that they dislike, consider inferior, or would like to be distanced from possesses some subset of a 1 -a 10 under N, that speaker may make the strategic choice to use N in communication as that which most efficiently and economically predicates the intended attributes of x as well as most forcefully expresses a negative attitude towards x, or most successfully establishes a distancing relation with x, at least to the extent that N is better for this than other categories that are afforded to that language-user.
In order to account for cases of linguistic appropriation (for further discussion of slurs and appropriation see also in Rahman, 2012;Bianchi, 2014;Croom, 2013aCroom, , 2014bCroom, , 2015b, I further suggest that if, for instance, an in-group African American speaker X is intending to communicate that they are sufficiently similar to some hearer Y insofar as X and Y are both attributed a 1 , a 4 , and possibly other attributes (such as a 5 and a 6 , or a 9 ), and if X and Y know each other well-enough or have established enough common ground to understand that X does not dislike or consider themselves superior to Y and does not intend to communicate that Y possesses most of the other (typically negative) attributes belonging to N, then X as an in-group speaker might strategically choose to employ N as the category that most efficiently and economically predicates the intended (shared) attributes of Y , such as a 1 (African American), a 4 (commonly the recipient of poor treatment), and possibly other (positive) attributes -such as a 5 (athletic or musical), a 6 (sexually liberal or licentious), or a 9 (a survivor, tough, or aggressive) -at least to the extent that N is better for this than other categories that are afforded to that language-user (Croom, 2010a,b). In contrast with other classical accounts of slurs that incorrectly assume that the possession of some criterial attribute is essential for x to be considered a member of N (see, for instance Camp, 2013, p. 338, 342 fn. 16), the alternative account of slurs proposed by Croom (2011Croom ( , 2013aCroom ( , 2014b and expanded upon here maintains (in accord with the family resemblance account) that, for example, although different individuals that are referred to by the slur nigger are very likely to share different subsets of attributes (for example, a 1 -a 10 ) with other individuals also referred to by this slur (due to common knowledge of how this expression is typically used) it need not necessarily be the case (for the sake of the felicitous application of that slur) that each and every slurred x must share some criterial attribute with every other slurred x (for further discussion of interesting cases see also Sweetland, 2002, p. 514;Croom, 2011, p. 356;Croom, 2013a, p. 199). Rather, what is of importance on this alternative family resemblance account is that the use of a slur by a speaker may be considered a strategically apt enough or optimally relevant lexical choice for their purposes in a particular cognitive or communicative context. Further, instead of considering some descriptive attribute (a 1 or African American) as necessary or criterial for the felicitous application of a slur S (nigger), the alternative account I propose here is that the descriptive attribute (a 1 or African American) instead serves as a conceptual anchor for S, which may be understood as the most relevantly salient (rather than necessary) default descriptor (e.g. a 1 or African American) that helps communicative agents ground the apt application of S (e.g. nigger) towards its prototypical (rather than essentially categorical) targets (Croom, 2014c, fn. 20). In other words, whereas classical accounts of slurs maintain that the descriptor African American is a nonpejorative correlate (NPC) of the slur nigger and that African American and nigger are coreferential expressions with precisely the same extension, here I instead argue that the descriptor African American is at most only the conceptual anchor for the slur nigger since African American and nigger are not coreferential expressions with precisely the same extension. As I have shown above, the application-conditions for typical targets of the slur nigger involve attributes that saliently differ from the attributes involved in the application-conditions for typical targets of the descriptor black (Rock, 1996(Rock, , 1997(Rock, , 2002 so there is good reason to maintain that these expressions differ in their semantic or conceptual content. Nonetheless, my account still makes sense of the common association between slurs and descriptors by suggesting that descriptors still serve as salient anchors for the semantic or conceptual content of slurs. That is to say, although my account denies the classical view that the descriptor African American is an essential attribute for all felicitous targets of the slur nigger to possess, it nonetheless maintains that the descriptor African American is still practically useful in grounding a conventional understanding of what is most salient of typical targets of the slur nigger.
I further propose that for formal or strict concepts, such as those involved in mathematics or the sciences, the conceptual anchor is made to play not merely a guiding or grounding role, as in the case of most family resemblance concepts of natural language, but a strictly determinative or criterial role. The claim here is that it is the especially strict formality with which the concept is made to apply that ensures the strict formality of its semantic or conceptual content. So for instance, concepts of chemistry that are primarily used among and so normatively regulated by a community of chemists will thereby have semantic or conceptual content that is more formal or strict than other concepts from natural language that are not so normatively regulated (for further discussion of concepts and normativity see also Croom, 2012Croom, , 2010a. In other words, the content of H 2 O is more strictly formal than the content of water because the application of H 2 O is more strictly regulated by a narrower and more specialized community than the application of water is. One benefit of the account that I propose here is that it is flexible enough to accommodate informal or looser concepts like water, the application of which is less strictly regulated by a wider and less specialized community, by allowing the actual variety of their conditions for felicitous application to contribute to the specification of prototypical attributes (which may be listed in the form a 1 a n ) to be subsumed under the concept water. Yet my account also readily accommodates formal or strict concepts like H 2 O, the application of which is more strictly regulated by a narrower and more specialized community, by allowing the actual restrictiveness of their conditions for felicitous application to contribute to the specification of essential or criterial attributes (which may be listed in the form a 1 alone) to be subsumed under the concept of H 2 O.
Another benefit of the account that I propose here is that it is flexible yet robust enough to further account for semantic evolution and linguistic appropriation (for further discussion of semantic evolution see also Croom, 2008Croom, , 2013b. Since my account allows the actual variety of application conditions for a concept or expression S to contribute to the prototypical attributes (a 1 -a 10 ) understood to be subsumed under S, my account thereby allows a change or evolution in the actual application of S to result in a corresponding change or evolution to the meaning or content of S. A consistent and salient change in the use of S can thereby change the meaning or content of S by (i) introducing or removing prototypical attributes (a 1 -a 10 ) from S, or (ii) reorganizing the internal structure, rank-order, or affect felt towards the attributes (a 1 -a 10 ) in S. Consequently, counter-culture communities often constitute an effective force by influencing the change of the actual application of expressions or concepts and thereby our understanding of what those expressions or concepts mean. Being able to account for the formality as well as flexibility of concepts or expressions, along with the possibility for their evolution and appropriation, is a major advantage of my view that remains missing from others.
The purpose of this section was to reconsider the relationship between slurs and descriptors and introduce the notion of a conceptual anchor to help provide a clearer understanding of their semantic or conceptual content. Having accomplished this, the purpose of the next and final section is to conclude by considering the inadequacy of common dictionary definitions of slurs.

Considering the inadequacy of dictionary definitions of slurs
It should be granted that some scholars have previously attempted to offer a plausible case in support of the view that slurs and descriptors are coreferential expressions with precisely the same extension. Yet such attempts have been almost entirely based on standard dictionary definitions of slurs (see for instance Williamson, 2003, p. 261). In ''Understanding and Inference'', for example, Williamson (2003) draws upon the Concise Oxford Dictionary to argue that the slur boche is defined as a coreferential expression (that expresses contempt) for Germans (p. 261). Williamson (2003) for instance argues that the slur expression '''Boche' has the same reference as [the neutrally descriptive expression] 'German'. That is certainly the dictionary view of the matter. Under 'Boche', the Concise Oxford Dictionary gives the definition ' (Contempt[uous] for) German''' (p. 261, my emphasis).
Despite such an appeal to dictionary definitions, Kenneth Himma (2002) has convincingly argued in ''On the Definition of Unconscionable Racial and Sexual Slurs'' that dictionaries that simply ''define the word ''nigger'' as ''black person'''' and ''the word ''faggot'' as ''male homosexual'''' inappropriately imply that certain offensive claims -such as the racist claim that all black people are niggers and the sexist claim that all male homosexuals are faggotsare not just true (which is dubious enough) but are (even worse) necessarily true (p. 512; see also Hom, 2008, p. 421-422;Croom, 2011, p. 352). For ''Just as the definition of ''bachelor'' as ''unmarried male'' implies that it is conceptually impossible for there to be an unmarried male who isn't also a bachelor, Merriam-Webster's definition of ''nigger'' as ''black person'' implies that it is conceptually impossible for there to be a black person who isn't also a nigger'', Himma (2002) argues, and ''Likewise, Merriam-Webster's definition of ''faggot'' as ''male homosexual'' implies that it is conceptually impossible for there to be a male homosexual who isn't also a faggot'' (p. 512, my emphasis). ''Since these reprehensible claims constitute the very foundation for racist and heterosexist views'', Himma (2002) concludes, dictionaries such as ''Merriam-Webster must revise its definitions of these terms to avoid committing itself to such views'' (p. 512, my emphasis). Given considerations of this kind, the argument from definition that Williamson (2003) has offered in support of the claim that slurs and descriptors are coreferential expressions with precisely the same extension is unappealing insofar as it presumably rests on ''definitions [that] implicitly endorse racist and heterosexist claims'' (Himma, 2002, p. 512, my emphasis). The argument from definition is also unappealing on more matterof-fact grounds, since it is incorrect in implying, for instance, ''that it is conceptually impossible for there to be a black person who isn't also a nigger'' and ''that it is conceptually impossible for there to be a male homosexual who isn't also a faggot'' (Himma, 2002, p. 512, my emphasis;Croom, 2011, p. 352). For we have now thoroughly reviewed how an empirical consideration of the slur faggot in ''Offensive Words'' by Szekely (2008) showed that this slur was in fact felicitously applied to some but not all male homosexuals, how an empirical consideration of the slur nigger in All Souls by MacDonald (1999) showed that this slur was in fact felicitously applied to some but not all African Americans, how an empirical consideration of the slur guido in ''Guido Culture'' by Troyani (2013) showed that this slur was in fact felicitously applied to some but not all Italian Americans, and how an empirical consideration of the slur nigger in ''Niggas vs. Black People'' by Rock (1996) showed that this slur was in fact felicitously applied to some but not all African Americans. A consideration of such empirical cases of slurs suggests that the coreferentialist assumption is untenable.
Importantly, during Quine's (1951) famous critique of the analytic/synthetic distinction in ''Two Dogmas of Empiricism'', he importantly pointed out how ''The word ''definition'' has come to have a dangerously reassuring sound, due no doubt to its frequent occurrence in logical and mathematical writings'' (p. 26, my emphasis) and further argued for the importance of clarifying linguistic ''notion[s. . . ] in terms relating to [actual] linguistic behaviour'' (p. 24; see also Quine, 1968, p. 185;Quine, 1990, p. 37-43;Fisher, 2011, p. 57;Smith, 2014, p. 494;Hylton, 2014, p. 40). 16 For as Quine (1951) correctly pointed out, even ''The lexicographer is an empirical scientist, whose business is the recording of antecedent facts'' concerning how language is actually used during human communicative interaction (p. 24, my emphasis). Quine (1951) illustrates this nicely during his discussion of the synonymous expressions bachelor and unmarried man in ''Two Dogmas of Empiricism'': how do we find that 'bachelor' is defined as 'unmarried man'? Who defined it thus, and when? Are we to appeal to the nearest dictionary, and accept the lexicographer's formulation as law? Clearly this would be to put the cart before the horse. The lexicographer is an empirical scientist, whose business is the recording of antecedent facts; and if he glosses 'bachelor' as 'unmarried man' it is because of his belief that there is a relation of synonymy between these forms, implicit in general or preferred usage prior to his own work. The notion of synonymy presupposed here has still to be clarified, presumably in terms relating to linguistic behaviour. (Quine, 1951, p. 24, my emphasis;Quine, 1990, p. 43).
In ''Two Dogmas of Empiricism'' Quine (1951) pointed out the important fact that dictionary definitions are always still empirically sensitive reports concerning how natural language is actually used, and accordingly, Quine (1968) further argued in ''Ontological Relatively'' that linguistic meaning is ''to be studied in the same 16 For instance Hylton (2014) offers a nice discussion concerning how ''For Quine, the criterion of successful communication, whether or not translation is involved, is fluent interaction, verbal and nonverbal: ''Success in communication is judged by smoothness of conversation, by frequent predictability of verbal and nonverbal reactions, and by coherence and plausibility of native testimony'' (Quine, 1990, 43)'' (Hylton, 2014, p. 40, my emphasis). Fisher (2011) further explains how Quine held that ''the evidence which guides our learning of a language is the behaviour of other language-users plus the observable circumstances surrounding that behaviour'' (p. 57, my emphasis). empirical spirit that animates natural science'' since ''Language is a social art which we all acquire on the evidence solely of other people's overt behaviour under publicly recognizable circumstances' ' (p. 185). 17 The considerations that Quine (1951Quine ( , 1968Quine ( , 1990 has offered here are important for our purposes because they serve as a methodological reminder for language theorists that being empirically sensitive to how natural language is actually used can only help one to carefully constrain and guide theorizing about slurs more specifically and natural language more generally. This is important since the meanings of linguistic expressions are not rigidly fixated on the past and impervious to semantic change, but are instead open to semantic evolution and renegotiation as communicators continue to make strategic use and sense of their available variety of expressions during the communicative exchanges of their social life.

Conclusion
Coreferentialism refers to the common assumption in the literature that slurs (e.g. faggot) and descriptors (e.g. male homosexual) are coreferential expressions with precisely the same extension (see for instance Dummett, 1973, p. 454;Hornsby, 2001, p. 129;Williamson, 2003, p. 261;Whiting, 2007, p. 192;Whiting, 2008, p. 385;McCready, 2010, p. 5, 9;Anderson and Lepore, 2013a, p. 26-27;Anderson and Lepore, 2013b, p. 351;Vallee, 2014, p. 79). Yet here I argued against this widespread assumption by first briefly introducing what slurs are and then considering four sources of supporting evidence showing that slurs and descriptors are in fact not coreferential expressions with precisely the same extension. I argued that since slurs and descriptors differ in their extension they thereby differ in their meaning or content also. This article additionally introduced the notion of a conceptual anchor in order to adequately account for the relationship between slurs and descriptors actually evidenced in the empirical data, and further considered the inadequacy of common dictionary definitions of slurs. This article therefore contributes to the literature on slurs by demonstrating that previous accounts operating on the assumption that slurs and descriptors are coreferential expressions with the same extension, and that they thereby have the same meaning or content, are inconsistent with empirical data and that an alternative account in accord with Croom (2011Croom ( , 2013aCroom ( , 2014b better fits the facts concerning their actual meaning and use.