On the Descriptive and Expressive Function of Derogatory Group Labels: An Experimental Test

By using a pseudoword paradigm, we tested whether derogatory labels (e.g., pejorative labels addressing group members) differed from category labels and general slur in their descriptive (i.e., pointing to group membership) and expressive functions (i.e., perceived offensiveness and social acceptability). Results indicated that derogatory labels were similar to category labels in their descriptive function, and had higher expressive function than slurs. Participants’ prejudice toward the groups that were targets of derogatory label reduced their perceived offensiveness than in turn increased their social acceptability.

Research in philosophical and linguistic traditions suggests that DGLs fulfill both a descriptive function, as they convey information about the group membership of their targets, and an expressive function, as they imply evaluative meanings (Anderson & Lepore, 2013;Croom, 2013Croom, , 2014. Alternatively, other authors suggest that DGLs have a purely expressive function (Hedger, 2012(Hedger, , 2013. The current research intends to deepen the understanding of the potential differences and similarities among category labels (i.e., labels that point to a group in a neutral manner; for instance "gay"), DGLs and general slurs in terms of descriptive (e.g., pointing to group membership) and expressive (e.g., perceived offensiveness) functions, given that different and sometimes mixed findings have been reported by research on this issue. Also, the current research analyze whether participants' levels of prejudice toward social groups are associated with the perceived offensiveness and social acceptability of DGLs, category labels, and general slur, an issue that is still debated in the literature.

Descriptive Versus Expressive Functions
In his analysis on linguistic expressions, Kaplan (1999) proposed the distinction between descriptive (i.e., semantic of meaning) and expressive (i.e., displaying an attitude/evaluation) content of such expressions. For example, the expression "I feel pain" has a primarily descriptive content, while "ouch" has an expressive content. Applying this distinction to the analysis of derogatory labels and common slurs, Hedger (2012Hedger ( , 2013 argues that DGLs primarily express negative affects/attitudes toward the target they are directed to, as in the case of racial slurs, losing their descriptive contents. Other authors (Anderson & Lepore, 2013;Croom, 2013Croom, , 2014 suggest that DGLs do convey expressive contents (e.g., evaluations) and, at the same time, work as descriptors, that is, they target certain group members on the basis of descriptive features such as their ethnicity, gender, or sexual identity.
Regarding the descriptive function, Carnaghi and Maass (2007) showed that homophobic epithets and category labels pointing to a gay sexual orientation equally activated label-consistent contents (i.e., stereotypical concepts) and inhibited label-inconsistent contents (i.e., counterstereotypical concepts), thus suggesting that both labels are similar on the descriptive dimension. Moreover, as DGLs are primarily used to degrade individuals on the bases of their actual or assumed group membership , and common slurs degrade an individual identity, without making reference to any group membership , DGLs are highly likely to have a stronger descriptive function than common slurs.
As for the different expressive function of category and DGLs, research has contrasted DGLs (i.e., homophobic epithet) and the corresponding category labels (i.e., gay) within the same experimental design, and has reported that DGLs are explicitly and implicitly appraised more negatively than category labels (Bianchi, Piccoli, Zotti, Fasoli, & Carnaghi, 2017;Carnaghi & Maass, 2007;Carnaghi, Maass, & Fasoli, 2011;Fasoli, Maass, & Carnaghi, 2015;Hunt et al., 2016). However, these studies have not compared DGLs to slurs that are unrelated to social groups (e.g., asshole). In this respect, O'Dea et al. (2015, Study 2), compared the perceived offensiveness of two racial slurs and a slur unrelated to racial groups (and a neutral term, such as buddy) and found the former labels being perceived more offensive than the latter. In similar vein, Preston and Stanley (1987) showed that, and among a variety of insults, the "worst thing" a man can be called is a DGL referring to homosexuality (Preston & Stanley, 1987; see also Saucier, Till, Miller, O'Dea, & Andres, 2015, Studies 3 and 4). Also, additional experimental research has reported that homophobic DGLs are rated as more offensive and insulting than slurs and category labels (Carnaghi & Maass, 2006). By contrast, Zotti, Carnaghi, Piccoli, and Bianchi (2018) reported that school staff perceived DGLs addressing gay and lesbian individuals as offensive as slurs unrelated to sexual orientation, and more offensive than category labels (i.e., gay and lesbian).
Moreover, additional research has analyzed the moderating role of prejudice in the perceived expressive function of category, DGLs and slurs. Specifically, participants who endorsed racial prejudice to a greater extent were also those who appraised racial epithets as less offensive, while no association was found between racial prejudice and the perceived offensiveness of general slurs (O'Dea et al., 2015, Study 2;O'Dea & Saucier, 2017). Zotti et al. (2018) reported that school staff's sexual prejudice was negatively and significantly related to the perceived offensiveness of DGLs, and it was positive and significantly associated with the perceived offensiveness of category labels referring to gay and lesbian individuals, while no association was found between sexual prejudice and the perceived offensiveness of general slurs. Hunt et al. (2016) found that adolescents' sexual prejudice was positively and significantly related to the perceived offensiveness of category labels referring to gay and lesbian individuals, while no association was found between sexual prejudice and the perceived offensiveness of DGLs as well as general slurs.
In sum, research has confirmed that DGLs are explicitly appraised more negatively than category labels, while the fact that DGLs can (or cannot) be assimilated to general curse words, thus constituting a specific cluster of derogatory language, is still a matter of debate. Also, a few studies have addressed the descriptive function of DGLs in comparison with category labels, and have limited their investigations to homophobic epithets (see also O'Dea et al., 2015;O'Dea & Saucier, 2017), thus preventing generalization of these results to other DGLs. Although the difference between category labels and DGLs on the expressive function has been largely acknowledged, a very limited number of studies has addressed the expressive function of DGLs in comparison with slurs, and has produced mixed findings. Finally, the research on the relation between perceivers' prejudice and the perceived offensiveness of category labels and DGLs have produced contrasting results, while the lack of association between prejudice and the perceived offensiveness of general slurs was consistently acknowledged.

Overview of the Studies and Hypotheses
In two studies, we addressed the perceived descriptive and expressive function of category, DGLs and slurs within the same experimental design, and we relied on a modified pseudoword paradigm (see Formanowicz, Roessel, Suitner, & Maass, 2017) that allowed for this investigation by controlling for participants' familiarity with and frequency of use of these labels. This paradigm allowed us to constrain the category labels to their descriptive function, and the slur labels to their expressive function. We presented participants with three distinct nonwords (i.e., labels) and told them that one label was used to define members of a specific group without being judgmental (i.e., category label), one label was used to offend members of a specific group (i.e., derogatory label), and the remaining label was used to offend anyone (i.e., general slur label). The derogatory label differed from the general slur label in terms of the target of such language, being a group member for the former an uncategorized individual for the latter; the derogatory label differed from the category label in terms of offensiveness but no mention was made of the potential descriptive function of the derogatory label.
Moreover, as this research aimed to ascertain the influence of prejudice on perceptions of the expressive functions of category and DGLs, real groups were chosen to be the targets of category and derogatory labels, namely immigrants, obese individuals, and homosexuals, given the fact that strong prejudice is expressed toward each of these groups.
To analyze the descriptive function of DGLs, the extent to which these labels were perceived to define group membership was assessed. It is worth noting that no mention of this descriptive aspect was made in the derogatory label experimental manipulation.

Hypothesis 1
If DGLs were processed as having a similar descriptive function as category labels, in line with the claims of Anderson and Lepore (2013) and Croom (2013Croom ( , 2014, but contrary to those of Hedger (2012Hedger ( , 2013, derogatory labels would be judged as describing group membership as category labels, and more so than general slurs, which we experimentally constrained to a purely expressive function.

Hypothesis 2
As far as group descriptiveness was concerned, an ancillary aim was to test the consequences in terms of social perception of individuals and groups defined by category, derogatory, and slur labels. Indeed, if DGLs worked as a categorization device as category labels (i.e., define group membership), the target of derogatory labels would be perceived as typical of that category, as similar to other category members as the target of category labels, and as essentializing as category labels, and more so than the target of slurs. To our knowledge, this will be the first time that the perceived strength, stability, and resilience (i.e., essentialism; Carnaghi et al., 2008;Walton & Banaji, 2004) of being labeled by category, derogatory, and slur labels is assessed.

Hypothesis 3
We put forward that DGL and slurs were processed as more offensive and less socially acceptable than category labels. As for the DGLs and slurs, either these two types of labels could be equally perceived as offensive (Zotti et al., 2018;Hypothesis 3a) or DGLs could be perceived as more offensive than slurs (Carnaghi & Maass, 2006;O'Dea et al., 2015; Hypothesis 3b).

Hypothesis 4
We also addressed the perceived social acceptability of the DGLs, general slurs and category labels. Given that a higher level of perceived offensiveness of a label is related with a lower perception of acceptability of this label (Hunt et al., 2016;O'Dea et al., 2015), we suggest that the perceived acceptability of the labels would mimic the pattern of the expected results concerning the perceived offensiveness. Indeed, we put forward that DGLs and general slurs would be rated as less socially acceptable than category labels and, more important, that DGLs, compared with general slurs, would be perceived as either equally low in terms of social acceptability (Hypothesis 4a) or less socially acceptable (Hypothesis 4b).
Hypothesis 5 Furthermore, we tested the relation between participants' prejudice and the perceived offensiveness of category, derogatory labels, and general slurs. Given the mixed findings with regard to this issue, prejudice can either be unrelated (Hunt et al., 2016) or negatively related to the perceived offensiveness of DGLs Zotti et al., 2018; Hypothesis 5a). Also, prejudice and the perceived offensiveness of general slurs should be unrelated, while prejudice could be positively related to the perceived offensiveness of category labels (Hunt et al., 2016;Zotti et al., 2018; Hypothesis 5b).

Hypothesis 6
Moreover, previous research has demonstrated a positive association between prejudice toward a group and the extent to which expressing prejudice toward that group was socially accepted (Crandall, Eshleman, & O'Brien, 2002). Also, Hunt et al. (2016) showed that higher levels of sexual prejudice were associated with higher levels of social acceptability of homophobic epithets. Hence, individuals with higher compared with lower levels of prejudice would be more prone to process DGLs as more socially acceptable.

Hypothesis 7
Finally, O'Dea et al. (2015) claimed that perceiving derogatory labels as less offensive enhanced the perceived acceptability of their use. Similarly, Hunt et al. (2016) reported that the higher the level of the perceived offensiveness of derogatory labels, the lower the social acceptability of these labels. We hypothesized that individuals with higher compared with lower levels of prejudice would be more prone to consider DGLs as acceptable, because they perceived these labels as less offensive.
Both studies were run with bachelor students. Gender distribution in both studies was compatible with the distribution in the same bachelor. Samples of both studies were homogeneous in terms of age, nationality, sexual orientation, and BMI.

Procedure
The procedure was identical in both studies, otherwise indicated. Participants were handed a questionnaire introduced as a study on glottology aiming to test the understanding of obsolete terms taken from different languages. Participants were presented three terms and their definitions. All terms were nonexistent words in the Italian language. The three terms (i.e., Ciltano, Gorcio, Tiltese) were randomly assigned to represent different labels (i.e., category, derogatory, and slur) in different versions of the questionnaire. In the category label condition, a randomly selected term was described as: "[adj. and n.] term used to refer to a group of people who are homosexual/immigrant/obese, i.e. to describe, without judgment, an individual or a group of people who are not heterosexual/Italian/normal weight." In the derogatory label condition, a randomly selected term was described as: "[adj. and n.] term used to offend the group of people who are homosexual/immigrant/obese, i.e., to denigrate, negatively judge an individual or a group of people who are not heterosexual/Italian/normal weight." The difference between Studies 1 and 2 concerned the slur condition. In the slur condition, a randomly selected term was described in Study 1 as: "[adj. and n.] term used to offend people in a nonspecific manner, that is to denigrate, negatively judge anyone, in any situation and for any reason," while in Study 2 as: "[adj. and n.] term used to offend people, that is to denigrate, negatively judge an individual." While in Study 1, slurs were defined as offending people in a nonspecific manner, in Study 2, we used a definition that was modeled to be consistent with the experimental definition of derogatory label. Moreover, in Study 1, the definition of slur stated that this class of labels could be used in any situation, and such information could have cued that these slurs were less offensive than DGLs. In Study 2, this information was discarded. The order of presentation of labels was randomized as well as the social group they referred to (i.e., homosexuals, immigrants, obese people). Next, participants reported the extent to which (a) that label was offensive (i.e., perceived offensiveness), on a scale ranging from 1 (not at all offensive) to 6 (completely offensive); (b) that label was socially acceptable if stated in public (i.e., social acceptability), on a scale ranging from 1 (not at all acceptable) to 6 (completely acceptable); (c) that label was used to define a specific group of people who have some characteristics in common (i.e., perceived descriptiveness), on a scale ranging from 1 (it defines anyone) to 6 (it defines a specific group); (d) that label was used to point to a typical group member (i.e., perceived typicality), on a scale ranging from 1 (completely atypical) to 7 (completely typical); (e) people who were defined by that label were similar to each other (i.e., perceived similarity), on a scale ranging from 1 (not at all similar) to 6 (completely similar). Participants then filled out the essentialism scale (i.e., essentialism; McDonald's ω: .71-.78 in Study 1 and .66-.69 in Study 2). They indicated the extent to which (a) a person who has been labeled in that manner would also be labeled as such in the future, on a scale ranging from 1 (not at all likely) to 6 (completely likely); (b) to be labeled in that manner profoundly defined who this person was, on a scale ranging from 1 (not at all defining) to 6 (completely defining); and (c) a person who was labeled in that manner could change who he or she was, on a scale ranging from 1 (not at all) to 6 (very much). Participants' prejudice toward the three social groups were assessed by means of two items, one pertaining to positive and one to negative feelings scales ranged from 1 (not at all positive/negative) to 7 (completely positive/negative). Participants reported their age, gender, sexual orientation, nationality, primary language, height, and weight. Participants were then thanked and debriefed.
Confirming Hypothesis 1, category and derogatory labels were perceived as equally descriptive of group membership, and more so than slur labels. In line with Hypothesis 2, individuals who were addressed by using category and derogatory labels were considered to be more typical group members and considered as more similar to each other than individuals who were called by a slur label. Also, category and derogatory labels were perceived as more essentializing than slur labels. No difference was found between category and derogatory labels on these variables.
Confirming Hypothesis 3b, results indicated that derogatory labels were rated as more offensive than slurs, and both labels were perceived as being more offensive than category labels. Derogatory labels were perceived as being as socially acceptable as slurs, and both derogatory and slur labels were perceived as less socially acceptable than category labels, thus confirming Hypothesis 4a.
As in Study 1 and confirming Hypothesis 1, category and derogatory labels were perceived as equally descriptive of group membership, and more so than slur labels. Regarding Hypothesis 2 and as in Study 1, individuals who were addressed using category and derogatory labels were considered to be more typical group members and considered as more similar to each other than individuals who were called by a slur label. Also, category and derogatory labels were perceived as more essentializing than the slur label. Different from Study 1, the results showed higher levels on these variables related to category versus derogatory labels.
As in Study 1 and confirming Hypothesis 3b, results indicated that derogatory labels were rated as more offensive than slurs, and both labels were perceived as being more offensive than the category label. Derogatory labels were perceived as being less socially acceptable than slurs, thus confirming Hypothesis 4b, and both derogatory and slur labels were perceived as less socially acceptable than category labels.

Studies 1 and 2: Small-Scale Meta-Analyses
We presented consistent evidence between studies concerning perception of offensiveness and descriptiveness of category, derogatory, and slur labels. However, we also found inconsistent results between studies on levels of acceptability of derogatory and slur labels, and on variables related to social perception (i.e., typicality, similarity, and essentialism) when category and derogatory labels were compared. To increase the precision of the parameter estimates, we relied on a small-scale meta-analysis that combines the results obtained from different studies investigating similar questions (see Cumming, 2012Cumming, , 2014. Therefore, and following the procedure outlined (Riva, Brambilla, & Vaes, 2016;Rusconi, Sacchi, Capellini, Brambilla, & Cherubini, 2017), we meta-analytically combined the results from the effect sizes reported in Studies 1 and 2 (N = 200). This analytic approach would allow us to both boost precision and integrate inconsistent findings across studies. The meta-analysis showed that the weighted combined z score for acceptability (i.e., difference between derogatory and slur labels) was statistically significant (z = 2.94, p = .026) with the effect size being small (r = .21, d = 0.43). As far as variables related to social perception were concerned, the weighted combined z scores for typicality, similarity, and essentialism (i.e., difference between category and derogatory labels) were also statistically significant (z = 3.12, p = .017; z = 3.16, p = .016; z = 2.98, p = .024). The effect sizes for typicality (r = .22, d = 0.45), similarity (r = .22, d = 0.46), and essentialism (r = .21, d = 0.43) were small.
In sum, acceptability of DGLs seems to be comparatively lower than the acceptability of slur (Hypothesis 4b). Also, category labels seem to slightly emphasize typicality, similarity, and essentialism in comparison with DGLs.

The Relation Between Prejudice and the Expressive Function
Cross-Experimental Analyses. The cross-experimental analyses would inform us about whether participants' level of prejudice was related to (a) the perception of offensiveness of DGLs, category labels and slurs, thus testing Hypotheses 5a and 5b; (b) the acceptability of DGLs, category labels and slurs, thus testing Hypothesis 6; and (c) whether these patterns of results were independent from the studies. Studies 1 and 2 samples came from the same pool of participants. Furthermore, the procedure and the stimuli were the same, except for the slur conditions. Data collection occurred in class at the beginning of the first year of a Bachelor's in Psychology. Hence, the two studies were homogenous. Statistical cross-examination of these studies could be theoretically reliable (for a similar rationale and procedure, see Cherubini, Rusconi, Russo, &Crippa, 2013, andShamloo, Carnaghi, Piccoli, Grassi, &. Data from Studies 1 and 2 were merged together and analyzed by using the type of study (i.e., Study 1 vs. Study 2) as a between-participants factor. To test Hypotheses 5a, 5b, and 6, we computed a series of regression with participants' level of prejudice as the predictor. Participants' prejudice toward the three target groups (McDonald's ω: .84 and .78 in Studies 1 and 2, respectively) were averaged to form a single index of group prejudice (M = 2.59, SD = 1.00 and M = 2.28, SD = 0.92 in Studies 1 and 2, respectively). 1 The offensiveness and the acceptability of category, derogatory, and slur labels were used as dependent variables. We regressed each dependent variable on participants' group prejudice, the type of study (Study 1 vs. Study 2), and their interaction term. Supporting Hypothesis 5a, results indicated that the higher the prejudice, the lower the perceived offensiveness of derogatory labels (Β = −.11, standard error [SE] = .04, t = −2.90, p = .004; Figure 1). Nor the type of study (Β = −.20, SE = .22, t = −0.92, p = .36) neither the interaction between participants' level of group prejudice and the type of study (Β = −.13, SE = .08, t = 1.65, p = .09) were significant predictors of the offensiveness of DGLs.
Nor participants' level of group prejudice neither the interaction between participants' level of group prejudice and the type of study were significant predictors of the offensiveness or the acceptability of category and slur labels (ts < 1.28, ps > .20).
Mediation Analysis. The lack of significant interaction between participants' level of group prejudice and the type of study when analyzing offensiveness as the outcome variable allowed us to test Hypothesis 7 by merging the two studies together. We relied on medmod 1.0.0 to test the mediation outlined in the hypothesis. Participants' prejudice was entered as the predictor, the DGLs' offensiveness was used as the mediator, and their acceptability was entered as the outcome variable. The predictor was significantly associated with the mediator, path a: Β = −.11, SE = .04, p = .003, the mediator was significantly associated with the outcome variable, path b: Β = −.77, SE = .10, p < .001, the predictor was associated with the outcome variable, direct effect: Β = .16, SE = .06, p = .006; the total effect was B = .25, SE = .06, p < .001. Importantly, the indirect effect was significant, Β = .09, SE = .03, p = .006, Sobel's z = 2.85, p = .004. Hence, increasing levels of prejudice were associated with lower levels of perceived offensiveness of DGLs, which in turn enhanced their social acceptability.  Note. DGLs = derogatory group labels.

General Discussion
This research analyzed the relative descriptive and expressive function of DGLs in comparison with category labels and common slur.
For the descriptive function, results revealed that DGLs and category labels are equally effective in pointing to category members, and more so than slurs (Hypothesis 1). DGLs and category labels work as linguistic tags that turn individuals into group members to a similar extent. As far as Hypothesis 2 is concerned, we expected DGLs to act as categorizing devices and highlight the perceived typicality of the individual they address as category labels and more so than general slurs. We indeed found DGLs to point to perceived typicality more so than general slurs, albeit to a different degree than category labels. Also, applying DGL or category labels, but not slur labels, to several individuals caused them to be perceived as very similar, thus mimicking the intragroup-homogeneity effect provoked by categorization (Corneille, Klein, Lambert, & Judd, 2002). Moreover, DGLs and category labels also convey essentializing views, albeit to a different degree, and more so than common slurs. To our knowledge, this is the first research endeavor to highlight this important consequence of DGLs in their descriptive function. Finally, our results suggest that category labels in comparison with DGLs point to slightly higher levels of typicality, similarity, and essentialism. We reason that these findings might be due to DGLs being characterized by dual descriptive and expressive functions, while category labels being primarily conveyors of descriptive meanings. Indeed, one possibility is that due to the fact that they also convey expressive meanings, DLGs might soften their descriptive function in comparison with category labels, especially in artificial setting in which both are compared.
For the expressive function, DGLs were perceived as more offensive (Hypothesis 3b) and less acceptable (Hypothesis 4b) than slurs. This pattern of results corroborates previous findings that pointed to higher levels of offensives of DGLs in comparison with slurs  and extends results on acceptability of DGLs in comparison with slur. As we described earlier in the Method section, we partly changed the definition of general slurs in Study 2 to make it more comparable with the definition of DGLs. The slight difference in the results between studies on the perceived social acceptability may be well due to this change, thus bolstering the idea that, with an improved definition of general slurs, their use is perceived as more socially acceptable than DGLs.
Significantly, and corroborating evidence that points to a significant and negative correlation between prejudice and DGLs' offensiveness Zotti et al., 2018), participants with low levels of prejudice appraised DGLs as more offensive in comparison with participants with high levels of prejudice (Hypothesis 5a). Additionally, participants' prejudice significantly affected their social acceptability. Indeed, participants with low levels of prejudice appraised DGLs as less socially acceptable in comparison with participants with high levels of prejudice (Hypothesis 6). Interestingly and differently from previous findings (Hunt et al., 2016;Zotti et al., 2018); in our studies, participants' level of prejudice was unrelated to the offensiveness of category labels, thus not supporting Hypothesis 5b, and consequently to their social acceptability. Also and as expected, participants' level of prejudice was unrelated to the offensiveness and the acceptability of slurs. In sum, in our studies the effects of prejudice seem to be selectively related to DGLs and specifically related to their general expressive function (i.e., perceived offensiveness and social acceptability). It is worth considering that we computed an average index of participants' level of prejudice toward the three social groups taken into account in our studies (e.g., gay men, immigrants, and overweight people). Previous research has examined the relationship between a specific group and offensiveness or social acceptability of derogatory labels pointing to this social group. It may be that the inconsistencies between previous studies might be due to different levels of prejudice toward the target groups they took into account. Finally, the mediation analysis revealed that enhanced levels of prejudice led to higher levels of social acceptability of DGLs, because DGLs were appraised as less offensive (Hypothesis 7), thus corroborating the model that was put forward by O'Dea et al. (2015). Given the correlational nature of our meditational analysis, an alternative model could be put forward. Specifically, and in line with models suggesting that perceived social norms influence the expression of prejudice (Crandall et al., 2002;Monteith, Deneen, & Tooman, 1996), it may be plausible that the perceived social acceptability of DGLs affects the levels of reported prejudice, which in turn impact on the perceived offensiveness of such labels. In a complementary analysis, we did not find support for such a model. 2 Given the correlational nature of both models, an experimental test is needed to ascertain the causal relation of the variables under examination.
A limit of this study relies on the artificial setting. Future research should enhance the ecological validity of these findings by using existing labels. Moreover, our participants were bachelor students and not evenly distributed by gender, as the gender makeup of our sample overlapped the gender distribution in the same bachelor. Indeed, our sample had a majority of women (2/3) who are reportedly less prejudiced toward some of the social groups we used in our studies (e.g., gay men and overweight people; Glenn & Chow, 2002;Herek, 1988). Future research may address the expressive and descriptive function of distinct classes of labels by relying on different types of samples and taking into account participant gender as a potential moderating factor.