Hate speech mainstreaming in the Greek virtual public sphere: A quantitative and qualitative approach

: This study delves into the manifestation and characteristics of hate speech in the Greek online public sphere, specifically exploring its most prominent forms, namely racism, anti-immigrant sentiment, nationalism, sexism, and homo-phobia/transphobia. Combining quantitative and qualitative methods, the research analyzes popular Greek online news media. It aims to uncover the visibility and operational patterns of hate speech, addressing key questions about its prevalence and presentation on these platforms. Findings reveal the normalization of discriminatory speech, particularly sexism and nationalism, in the digital landscape. More-over, distinct themes and linguistic elements associated with different forms of hate speech highlight the intertwined nature of these discriminatory expressions. This research sheds light on the pervasive nature of hate speech in Greece’s online realm, emphasizing its implications for democratic principles.


Introduction
The increasing presence of racism, xenophobia and hate speech in European public spheres has been a phenomenon attracting great theoretical and empirical interest.During the last decade, hate speech has undergone a sharp increase in Greece, as documented by European reports (European Commission against Racism and Intolerance, 2015) and academic research (Assimakopoulos and Baider 2017;Dalakoglou, 2013;Triantafyllidou and Kouki, 2012).This development has been taking place against the backdrop of growing support for extreme right-wing political parties.
This phenomenon has been addressed by a voluminous strain of literature, devoted to the study of hate speech.In addition, many of these studies focus on the online environment which functions as a favorable locus for its articulation and mainstreaming.These 'words that wound' (Matsuda, 1993) do not merely describe the injury, or produce it as a result, but rather constitute the harm itself.Acting in a performative manner (Butler, 1997), they create what they name, thus constituting the subject in a subordinate position (Kuntsman, 2008).Therefore, hate speech entails significant repercussions for equality and democratic coexistence, since it legitimizes a reality of discriminations and exclusions, which appear as natural, given and fixed (Van Dijk, 2000).Moreover, the diffusion and normalization of hate speech in the public sphere entails tangible consequences for agenda-setting, political competition and policy making, influencing the way political discourse is articulated, both by political parties and the media (Vasilaki and Souvlis, 2021).
Within this context, this article attempts to delve into the visibility and particular characteristics of hate speech in the Greek online public sphere.More specifically, our research questions are: The paper is structured as follows: In the first part, the issue of hate speech is contextualized in the Greek case, by addressing both the role of political parties and dominant attitudes and media representations that could favor a rise in hate speech.The second part offers a theoretical discussion regarding the conceptualization of hate speech and the prominence of online media as the locus of its reproduction.To tackle the aforementioned research questions, in the third part, we will employ data from the program 'Hate Speech Monitor 2019 ' (HSM2019, Tsirbas, 2022).being a host of migrants (Ellinas, 2013).These sentiments later became even more profound after the refugee crisis fueled by the Syrian war (Dianeosis, 2020;European Commission, 2018).
Nevertheless, the pre-existence of an exclusive hegemonic identity is not confined to that of Greek national identity; attitudes towards gendered roles and equality should also be taken into consideration.As Tsimouris (2015) argues, intersections between the 'deep structures' of nationalism and masculinity are at play in the discourse of both political parties/institutions and the media, constructing a particular image of 'the Other' in Greek society.These underlying, 'banal' and, up to that point, muffled characteristics of Greek political culture re-emerged and intensified during the crisis, resulting in the dissemination of sexist discourses and practices, even actions of aggressive manhood such as the ones performed by Golden Dawn's political cadres, often in prime-time television programs (Avdela and Psara, 2012).Here, nationalist and sexist discourses are intertwined, since references to the Greek nation are rooted in a version of aggressive masculinity, which draws from traditional and stereotypical models of gendered relations to praise 'true' Greek men and target women who deviate (Avdela and Psara, 2012).Such affinities between nationalism and sexism have also been found in the context of other countries, like Romania and the US (Norocel, 2016;Shook et al., 2020).Turning now to the role of mass media, it has been attested that dominant representations of refugees/immigrants in the media were already negative before the advent of the migrant crisis (Karydis, 1996;Falekas et al., 2012).These representations often relied on a number of social myths and distorted notions, reproduced both by the media and extreme right political parties (Papastergiou and Takou, 2014), resulting in a rather adverse image in public opinion (Tsimouris, 2015) that could instigate or reinforce discriminatory social attitudes (Falekas et al., 2012).
This depiction stands in congruence with research conducted in other European countries that has pinpointed the racist undertone used against refugees and immigrants, who are often vilified by the media, either by being identified with criminal activities (Khosravinik, 2010;Lueck et al., 2015), or depicted as threat to the European way of life (Amin-Khan, 2012;Kallis, 2013a).It has been evidenced that the frames of illegality, criminality and securitization have dominated the Greek public sphere as well (Karamanidou, 2016).During the financial crisis, such hostile representations were aggravated due to the instrumental use by media and political parties of migrants as scapegoats for any social disarray.(Karamanidou, 2016;Doukarelli and Braessa, 2021).
Another important factor in the rise of hate speech in the Greek public sphere has been the unprecedented surge of extreme right-wing parties (Triantafyllidou and Kouki, 2012;Papastergiou and Takou, 2014) as along with the mainstreaming of their agenda (Vasilaki and Souvlis, 2021).Xenophobia and anti-immigrant attitudes, already present in a latent form within the Greek citizenry, intensified during the economic (and, later, refugee) crisis, when such issues became politicized and entered the agenda not only of far-right, but of mainstream parties (Ellinas, 2013;Tsimouris, 2015).Already before the outbreak of the crisis, these attitudes had found political representation in the Popular Orthodox Rally (LAOS); a nationalist populist extreme right-wing party that won ten seats in the 2007 national elections.Soon after, the neo-Nazi party Golden Dawn (GD) was best positioned to benefit from the electoral realignment caused by the crisis (Ellinas, 2013).After GD's first electoral success in the municipal elections of 2010, when it received 5.3 % percent of the vote in Athens (Georgiadou, 2013), the Greek media showed increased interest in its xenophobic rhetoric and fascist activity (Ellinas, 2013), in alignment with what literature suggests (Ellinas, 2009;Feischmidt and Hervik, 2015).Although most media treated the party critically, its 'social activities' were often greeted more positively (Ellinas, 2013;European Commission against Racism and Intolerance, 2015).Meanwhile GD was able to acquire significant online presence in online news sites, 'infiltrating the digital mainstream' (Siapera and Veikou, 2016, p. 1).GD managed to become the third strongest political party in the Greek party system 2014 and 2015, but it failed to enter the parliament in the 2019 election.In October 2020, GD's leadership was convicted of running a criminal organization under the guise of a political party and was sentenced to prison.However, in the 2019 elections, a new far-right nationalist party, Greek Solution (GS), entered the Greek parliament.
This rise of far-right extremism has led to a more generalized right-wing shift among competing parties (Tsimouris, 2015;Vasilaki and Souvlis, 2021).As Triantafyllidou and Kouki (2012) already noted in 2009, there was evidence that a rhetoric of intolerance and discrimination had been adopted by domestic political actors of the greater conservative family Moreover, this 'contagion' of mainstream political parties' discourse has been observed in many other European countries (Kouki and Vidra, 2014).This process of mainstreaming has a dual dimension (Feischmidt and Hervik, 2015): far right-wing parties and movements are rephrasing their political discourse, moving away from an overt rhetoric of hate and towards forms of 'incendiary racism' (Amin-Khan, 2012), while, at the same time, 'previously taboo ideas, frames, and practices are becoming the new 'common sense' for growing sections of European politics and societies' (Kallis, 2013b, p. 221), crossing the boundaries between 'extremist' and 'mainstream' discourses and, thus, opening the way for the further normalization and legitimization of extreme views and practices, both by mainstream and far-right political actors (Kallis, 2013b;Van Dijk, 2000).
Therefore, it is evident that in the case of Greece the economic and the refugee/ immigration crisis have intersected in a way that accentuated intolerance and processes of othering, favoring the rise of extreme right (Triantafyllidou and Kouki 2012;Tsimouris, 2015;Vasilaki and Souvlis, 2021) and online hate speech.These crises, however, have not operated on an empty canvas, but rather they have reinforced preconceived notions, attitudes and representations firmly rooted in the Greek society, laying the ground for the employment of hate speech in the public discourse (Karamanidou, 2016).
The political climate when data collection took place was marked by growing discontent with the governing SYRIZA party, for its handling of the third bailout program, the refugee crisis and the deadly wildfires in Attica in 2018 (Tsirbas, 2021).The signing of the 'Prespa Agreement' between Greece and North Macedonia, resolving the long-lasting dispute over its official name, was also dominant in the public discussion.The signing of the agreement faced strong public reaction, was accompanied by massive demonstrations, and proved to be one of the most salient and divisive electoral issues (Tsatsanis et al., 2020).

Conceptualizing hate speech
Hate speech is inherently related to processes of othering, stereotyping and discriminating, ultimately founded upon the notion of racism (Katsourides et al., 2018, p. 152).Even though racism has undergone various re-conceptualizations across time and space and a thorough examination of the concept is out of the scope of this examination, we will briefly refer to some of its 'universal' traits relating to this inquiry.As Goffman (1963) has famously demonstrated, racism operates through stigma, by vilifying, excluding or even dehumanizing and denying agency (Howarth, 2006) to particular individuals and groups on the basis of their race, skin color, national identity and religious beliefs.According to Van Dijk (1991), racism can be defined as a system of group dominance, both structural and ideological, and is thus reproduced simultaneously through individual attitudes and institutionalized practices, placing a particular importance upon the role of the media (Van Dijk, 2000;Amin-Khan, 2012;Falekas et al., 2012;Löwstedt and Mboti, 2017) and political parties (Triantafyllidou and Kouki 2012;Dalakoglou 2013).
Here, we will focus on a particular verbal expression of racism in its wider sense, namely what is usually referred to as hate speech.The conceptual ambiguity of the term, as well as the difficulty in demarcating its content from its legal connotations, has resulted in a plethora of equivalent or metonymic terms describing the same phenomenon, especially its online dimension, including extreme speech (Pohjonen, 2019), dangerous speech, fear speech, hateful speech, excitable speech (Butler, 1997), and cyberhate (Gagliardone, 2019).For the purposes of this research, the term hate speech was preferred due to its wide use and applicability, as well as inclusiveness in the phenomena it may encompass (Feischmidt and Hervik, 2015;Parekh, 2006).
This paper aims at moving beyond the legal aspects of hate speech (which have been thoroughly analyzed elsewhere (see, inter alia, Parekh, 2006;Brown, 2017), and delve into the use of such language in the virtual public sphere.Within this framework, hate speech operates in a threefold way: it distinguishes individuals or groups on the basis of one or more particular traits; it stigmatizes them due to the qualities associated with those traits and then it casts them off the realm of normal and appropriate social relations, resulting in them being 'legitimately exterminated, expelled, discriminated against, or at best tolerated as an unavoidable evil surviving on the margins of society' (Parekh, 2006).Drawing on this description, hate speech shall be understood as derogatory or abusive speech targeting specific group characteristics (Warner and Hirschberg, 2012, p. 19), that 'is meant to offend, exclude, intimidate, or discriminate against members of a group based on their race, religion, sex, sexual orientation, nationality or ethnicity' (Schulzke, 2016, p. 225).1 Relevant literature categorizes hate speech according to the particular social group that is targeted each time (Iordache, 2015;ElSherief et al., 2018).These may include immigrants, national, ethnic, or religious minorities, underprivileged groups such as drug addicts and, more recently, women, sexual minorities (LGBTQIA+) and people with disabilities (Lilian, 2007;Vasvari, 2013;Iordache, 2015;Frenda et al., 2019).

Main concepts
As is already evident, over the past years, hate speech has conceptually expanded, in order to include most kinds of discrimination and potential 'targets'.This research, however, focuses on the most common and prominent categories of hate speech, namely, racism, anti-immigrant speech, nationalism, sexism and homophobia/ transphobia.Racism operates as an umbrella-term that describes discrimination against individuals on the basis of their race, skin color, ethnicity, religion etc.Even though modern racism is inextricably linked with the notions of nationalism (Balibar & Wallerstein, 1991, p. 46;Van Dijk, 1991;Troumbetta, 2000) and anti-immigrant othering (Löwstedt and Mboti, 2017, p. 120), in the context of the present paper it is mainly used with regard to race.Regarding the separate category of anti-immigrant speech, it differentiates between 'indigenous' people and immigrants/refugees, referring to the latter in a demeaning way, because of their perceived status.Nationalism refers to the invoking of Greek national identity in a way that creates exclusions and discrimination with respect to participation in the nation; when it highlights the Greek nation's 'superiority' by resorting to particular characteristics attributed solely to white, orthodox Christian Greeks, and moments from Greek history to support this claim; and/or when it uses derogatory references to other populations' national identity.Sexism is defined as the bulk of prejudices, stereotypes and discrimination stemming from the subjects' gendered identity, which is very prominent in discourse and employs a language that ignores and/or discredits subjects according to that identity (women in particular), imposes a dichotomic conception of gender and ascribes specific, quasi 'natural' characteristics to both 'female' and 'male' gendered identity.In their influential contribution, Glick and Fiske (1997) have proposed a conceptualization of sexism as ambivalent, consisting of two main components: hostile sexism and benevolent sexism.The former, describes negative attitudes and behaviors towards women, including beliefs and practices that overtly demean, devalue, or discriminate against women.The latter appears seemingly more positive at first glance (and is also perceived as such from the part of the sexist subject), but also draws on and reproduces traditional roles and stereotypes like the idea that women should be protected and admired for their roles as providers of care and as nurturers.Benevolent sexism adores and idealizes women, while at the same time considering them inferior to men; less capable and in constant need of support.These depictions also result in a double bind in which women who do conform to the stereotypically 'female' characteristics are presented as incapable and those who don't are unpopular and not 'female' enough.Therefore, both categories serve the maintenance and justification of sexism and gender inequality.Finally, homophobia/ transphobia includes negative/derogatory comments and hostile language directed against LGBTQIA+ individuals.It is based on a (fictional) heterosexual normalcy and discriminates subjects due to lack of conformity to the dominant heteronormative standard in defining their gender and sexuality.

Online media as the locus of hate speech
Mass media play a pivotal role in distributing, circulating, and reproducing racist ideas among the members of society (Löwstedt and Mboti, 2017;Van Dijk, 1999), both due to their decisive functions of discourse and communication with respect to the propagating of racism (Van Dijk, 1989), and the role of mass media in shaping public opinion.In this process, they are influenced by a number of factors, including corporate interests, ownership, media regulation, professional ideologies, news values, cognitive strategies (Van Dijk, 1989;Löwstedt and Mboti, 2017), and ties with the political, economic and cultural elites, all of which result in a discourse that favors particular dominant identities (e.g., white, western, male, and middle-class subjects) (Löwstedt and Mboti, 2017).
Hence, the mass media do not act as passive mechanisms that reflect public opinions and attitudes, impartially describing current events, but rather actively interpret and reconstruct them (Van Dijk, 1991, p. 41;Fryberg et al., 2012).Within this context, it is evident that the media can play a crucial role in reproducing, normalizing and legitimizing racism and hate speech in the public sphere (Cammaerts, 2018;Van Dijk, 1999, 1991;Cerase et al., 2015;Feischmidt and Hervik, 2015;Kallis, 2013b;Löwstedt and Mboti, 2017), making their study not only fruitful, but imperative in order to understand the process of mainstreaming hate speech .
This function becomes all the more prominent in the new online environment, with multiple scholars attesting to a sharp growth of racism and hate speech in the digital sphere (Klein, 2012;Assimakopoulos et al., 2017).In addition, women are under and misrepresented in the media, both traditional and online (Ward and Grower 2020;Haraldsson and Wängnerud, 2019;Humprecht and Esser, 2017), where they are often treated as sexual objects, depicted as sexually promiscuous and illequipped or inadequate to undertake positions of power and leadership, be they political or corporate, thus, invoking and reproducing traditional gender stereotypes (Iordache, 2015;Campus, 2013).Sexism and misogyny are particularly prevalent in advertisements but have also permeated mainstream columnists, TV shows, sports coverage, and political news websites (Courtney et al., 2020;Blumell, 2020;Iordache, 2015).Although the arrival of the internet and new technologies has, in many aspects democratized the sphere of information, it has also contributed to the normalization of hate speech, constituting a favorable field for expressing bigotry and hate (Klein, 2012;Hirvonen 2013;Cerase et al., 2015).Therefore, apart from the study of hate speech in websites, blogs etc., a voluminous strain of literature is particularly focused on social media and online commentary.In this paper, however, our focus will be placed upon news media and political parties' websites, since they constitute actors that bear particular significance and influence in the diffusion and mainstreaming of hate speech online.

Materials and methods
In order to explore the aforementioned research questions, we will employ empirical data collected in the context of the HSM2019 research program.The project was conducted under the aegis of the Institute of Greek Politics of the Department of Political Science and Public Administration of the National and Kapodistrian University of Athens.The data used in the present paper refer to the 11 most frequently visited news websites, as ranked by alexa.com(2019; see Table 1), from the April 1, 2019, until the day of the European elections on May 26.

Website
Description URL The sample includes the body of online articles, notwithstanding the section they fell under.The selection of online media over traditional newspapers is a popular methodological choice in this line of research, due to the prominence of the virtual sphere as a locus of hate speech reproduction and dissemination (Klein, 2012;Hirvonen, 2013).The unit of analysis was every independent excerpt demonstrating one or more types of hate speech, resulting in a total of 1,015 unique entries.These categories comprised nationalist, racist, anti-immigrant, sexist, and homophobic/transphobic speech, as we have already discussed in more detail above.Coders were given the aforementioned operational definitions of hate speech categories, along with some indicative keywords and examples, and were instructed to document each excerpt that fell under one or more of these types.Hate speech excerpts that were treated in a condemnatory or critical manner were not included in the documentation.All sources were simultaneously monitored by two researchers rotating between media every week, thus enabling the calculation of intercoder percent agreement for every variable, as a measure of reliability.For each unit of analysis, a number of variables were coded, including information on the source, type of text and subject of enunciation, as well as the kind(s) of hate speech it demonstrated.
It should be noted that the aim of this inquiry was not to draw a representative sample in order to find the analogy between hate and non-hate speech on each website, but rather to document the absolute number of hate speech excerpts in every medium we examined.After data collection was completed, the bulk of coded data was reviewed by two independent researchers, achieving an intercoder reliability of more than 80 %.Then, mostly descriptive statistics were employed to tackle the quantitative aspects of our research question.The data were then qualitatively analyzed according to the principles of thematic analysis.Thematic analysis can be defined as 'a method for identifying, analysing and reporting patterns (themes) within the data' (Braun and Clarke, 2006, p. 79), while a theme, is a 'pattern found in the information that at minimum describes and organises the possible observations and at maximum interprets aspects of the phenomenon' (Boyatzis 1996, p. 3).Here, a coding reliability approach was adopted, in which themes are rather identified than developed (Braun and Clarke, 2016).In this approach, coding is seen as a process for looking for evidence of previously identified themes (Braun and Clarke, 2016), according to the conceptualization presented in the previous section.In the next phase, initial codes were reviewed again and transformed into themes and subthemes, capturing important aspects of the data in relation to the research question (Braun and Clarke, 2006, p. 82).

Quantitative analysis: The scope of hate speech mainstreaming
The total number of unique hate speech entries identified in the news websites under investigation was 1,015.By 'unique', we mean that each hate speech excerpt could be located several times by the coders, possibly in more than one website, but it was inserted only once in the database, as belonging to the source which published it first.Given the fact that the coding period was eight weeks long, this means that 18 unique hate speech excerpts, without any critical commentary, were transmitted in the Greek virtual public sphere each day, being viewed by literally millions of news website visitors.
The most visible hate speech category was sexism, with 50.2 % of the excerpts being categorized as such (Table 2).The second most salient kind of hate speech was nationalism, with 33.9 %, followed by racism (19.8 %), anti-immigrant speech (11.4 %) and anti-LGBTQIA+ speech, with 6.2 % (Table 2).
The main enunciators of hate speech are the news websites themselves through unsigned articles, which amount to 64.8 % of all entries (Table 3).This finding underlines the role played by online media in the (re)production of hate speech, in accordance with what the literature suggests.Moreover, the articulation of this kind of speech is mostly carried out without the composers' personally signing the texts; that is without assuming individual responsibility for the hate speech that is being voiced.The connotation of this finding is that hate speech is a fully embedded aspect of mainstream media because by widely adopting unsigned texts they willingly forfeit the right to make the 'personal responsibility of the author' an excuse.Furthermore, given the great degree of interconnectedness between the media and other structures of power, for example, politics and businesses, it is safe to assert that the main argument of this paper concerning hate speech mainstreaming seems rather solid.Also, the above-mentioned notion that the media actively interpret and reconstruct public opinions is further supported by these findings.
Politicians/political parties come in second as enunciators, with 18.4 % of the total excerpts, followed by journalists/reporters (5.8 %), anonymous hate speakers whose statements are being reproduced (3.8 %), members of high visibility profes-  3).Makeleio comes first in all five types of hate speech under examination.Sexism is more equally distributed among the different websites, thus verifying it is the most salient type of hate speech across the virtual public sphere.Sexism is more frequent in Makeleio (34.5 %) and Proto Thema (17.3 %) and it is quite salient, with more than 10 % of excerpts in three more outlets.Nationalism is also rather equally distributed among the different websites.Apart from Makeleio (45.1 % of nationalist speech), the tabloid Zougla (17.2 %) and Proto Thema (9.9 %) contain significant amounts of nationalist speech.Anti-immigrant speech, racism and anti-LGBTQIA+ speech on the other hand, are quite unevenly distributed among different outlets.Makeleio has in every case more than 50 % of the excerpts.The second outlet for racism and anti-LGBTQIA+ is Proto Thema, and regarding anti-immigrant speech it is IEfimerida (Table 4).In sum, sexism and anti-LGBTQIA+ speech are more common in sports and lifestyle outlets while nationalism and anti-immigrant speech is more salient in political outlets.Racism on the other hand is a generic feature of the Greek digital public sphere, since it can be found literally in all kinds of online media.
In Table 5, hate speech categories by enunciator are presented.It is interesting that while unsigned articles have the lion's share in every category, the hate speech category that is mostly conveyed by this particular enunciator type is sexism (80.6 %), followed by racism (73.6 %).On the other hand, nationalism is the type of hate speech that depends less on the medium (41.6 %) than any other and more on politicians (41 %) to be transmitted.Politicians also have a large share regarding anti-immigrant speech (19.8 %).Anti-immigrant speech and racism are the types of discriminatory speech where representatives of the military and the police are more vocal (3.4 % and 3.5 % respectively).Lastly, anti-LGBTQIA+ speech, besides being transmitted mostly in unsigned pieces, is the type of speech that journalists enunciate eponymously more than any other category (14.3 %) and that high visibility professions, the church/clergy and various technocrats contribute to the most.A plausible explanation for these variations is that sexism and racism, although mainstreamed and embedded, are not socially acceptable as personal labels, thus they are conveyed anonymously.This is not so much the case regarding other categories, which openly assert a kind of group dominance, such as nationalism.Table 6 contains the correlation coefficients between the different types of discriminatory speech.Phi correlation coefficients are presented.Phi is suitable for testing the relationship between binary variables.Sexism appears to be the more autonomous type of discriminatory speech, demonstrating negative correlations with all four other types of speech.This is expected because sexism is the hate speech category that is overwhelmingly conveyed anonymously, thus limiting the possibilities of co-existing with other types of hate speech.At the same time, three of these negative correlations are the three strongest of the whole matrix, namely with nationalism, anti-immigrant speech and racism.Also, rather intuitively, racist and anti-immigrant speech have the strongest co-existence relation.Anti-LGBT-QIA+ speech also seems to be a rather stand-alone type of hate speech, demonstrating negative correlations with sexism, nationalism and anti-immigrant speech, albeit weak ones.

Qualitative content analysis: The mechanisms of hate speech mainstreaming
Qualitative analysis aims at unpacking the mechanisms of hate speech mainstreaming.Specifically, it aims to address the particular ways in which hate speech is operationalized, as well as the thematic vehicles of its construction and transmission, by employing thematic analysis.
Sexism.In the case of sexism, the most common theme is the 'Aestheticization of the female body'.The male gaze (Mulvey, 2014) is prevalent, leading to a sexualized or even dehumanized depiction of women, who are (re)presented as sexual objects, serving the pleasure of the heterosexual male reader.E.g., 'We are talking about the Minister's blonde, juicy secretary.'2This theme is in accordance with the above-mentioned benevolent sexism type.Furthermore, according to a subtheme, 'Political correctness equals hypocrisy'.Here, the objectification of women is presented as natural and normal; all political correctness does is to conceal this normality.For example, an article stresses that 'the mere difference is that now, the conflict among the contemporary, civilized man who claims "it is her right to pose as she wants, against the stereotypes" and the other humble self who confesses that he prefers the picture "with the high stockings" is taking place in public.And, depending on each one's level of hypocrisy, it becomes disclosed or not.'3This attack against political correctness appears as an attempt by the editors of such texts to a priori deny allegations of sexism, by normalizing and naturalizing the aestheticization of women's bodies.
The second theme of sexist speech glorifies the 'Proper man'.Manhood is praised as a value reaffirmed through female conquests, often presenting itself as de facto rational, in stark contrast with the inherent irrationality of women.'Papadopoulos is the only [true] male that I have come upon in my career […].The only one who, without swinging, without pretense, without hesitations, talks with such an honest straightness.The vagina for him is what religious paintings are for every truly devoted Christian',4 an interviewee says, while another interviewee highlights that 'men, more than before, think that their rationalism is under attack.Women have the advantage of talking more easily '.5 Another theme refers to 'Ethical and value judgements,' whereby women are judged based on traditional values and morals.For example, the former Prime Minister's partner is referred to as the 'unwed, who has stretched from grubbing.'6Another excerpt claims that 'few things are so funny and hypocritical as a woman with deep cleavage being annoyed because people are dealing with her breasts.'7The aforementioned themes are in agreement with similar research findings regarding the content of sexist speech.According to Iordache (2015), the objectification of women, the emphasis on their physical appearance, and their portrayal as sexually promiscuous, are the most common themes of sexist discourse.
Nationalism.The first theme of nationalist speech is entitled 'Unique … and anti-leftist'.This theme refers to the issue of the name of North Macedonia and the signing of the Prespa Agreement, which proved to be one of the most salient and divisive electoral issues (as mentioned above).The common denominator is the uniqueness of Macedonia and its sole identification with Greece, combined with aggressive references to the Left, which is portrayed as a national traitor, surrendering the country to its enemies: 'The gathered citizens were yelling "scumbags, traitors, left" and "Macedonia is Greek" to the two candidates for the European elections, as well as to everyone else entering the city hall of Eordaia, where SYRIZA's "Europe of solidarity" was planned to take place.'8 Another theme pertaining to nationalist speech is that current affairs are framed in relation to 'Conspiracy theories', which primarily entail a secret plan to subvert the Greek nation.For example, an article claims that 'it became clear that Soros' aim was to destroy the national borders and the identity of all member-states for the purposes of a single world, a new order.Hence, the obvious plan is the destruction of national sovereignty, along with the cultural and religious characteristics […].It is also evident that Soros is just the tip of the iceberg and that a shady elite, responsible for actualizing these special plans, exists.'9 Racism.Beginning our analysis with racist speech, a recurrent theme is 'The division of labor in criminal activity'.Particular (ethnic) groups of people are linked with delinquent/illegal behaviors.For example, according to an article '...the most dangerous (aspect) is that within these squats refuge thugs and undocumented foreigners find refuge, whose origin, past and identity are unknown to the authorities.From Algerian cutthroats, Afghan cigarette sellers and drug dealers to suspicious people moving at the margins of global terrorism, they all compose the weird mosaic of "new Exarcheia".'10Thus, qualitative analysis is in line with the quantitative findings, indicating that racist speech is mainly xenophobic.Immigrants are perceived as threats to social order and security, as well as public health, a recurring theme in the representations of migrants in the Greek public sphere (Papastergiou and Takou, 2014;Doukarelli and Braessa, 2021).
The second theme is 'Allegories and metaphors;' that is, employing specific terms that normally denote something else, to describe a situation in an implicitly derogatory manner.An indicative example is the reference to 'the Kabul of Acharnon Street'11, or the repetitive mentions of 'Elladistan'.12It has to be noted that the use of specific linguistic strategies, such as metaphors, is not a new phenomenon: Baider et al. (2017, p. 38) stress that they 'are often employed with a view to constructing the exclusion of the out-group and the cohesion of the in-group,' while Van Dijk (2000, p. 35), also highlights the use of such 'rhetorical devices' in the reproduction of racism.
Anti-immigrant speech.Turning to anti-immigrant speech, one identifiable theme is 'State of siege'.Greece is being depicted as an 'open unguarded field' (Papastergiou and Takou, 2014, p. 21); a castle under constant siege, which needs to endure the recurring waves of foreigners' attacks, which are threatening to distort the demographic composition of the country (see also Doukarelli and Braessa, 2021).According to one article, 'considering the pace of our country's settlement by the uninvited guests, it will not be long before it won't matter for the system what native Greek's opinion for politics is.Numbers will have imposed their own 'justice' in the place that gave birth to democracy and will soon have sharia implemented.'13 Another interesting theme depicts migration as a 'Holiday trip'.The process of migration is being flouted as a convenient and opportunistic journey, while immigrants are supposed to receive preferential treatment.For example, immigrants' transfer from the islands, where they were stranded, to the mainland is described as a cruise that abuses Greek people's money, with the Greek immigration minister, Mr Vitsas, seen as the cruise leader: 'Revealing photographs!Immigrants' "cruise" using Greek people's money and Vitsas as their "tour guide"-Rage among all Greeks.'14These findings stand in line with research on the field, which, among other findings, pinpoints the emergence of a discourse about 'immigrants having it too good in Greece' (Papastergiou and Takou, 2014).
Anti-LGBTQIA+ speech.Finally, homophobic-transphobic speech (which here is also referred to as anti-LGBTQIA+ speech) mostly revolves around the theme of 'Belittlement, mockery and shaming'.This theme includes derogatory references to any non-heteronormative sexual and gender identity, while at the same time, refusing the LGBTQIA+ subjects' right to self-identification.The most indicative example of the latter comes from the most popular news portal in Greece, asserting that 'no 13 Makeleio, 26/05/2019 14 Makeleio, 14/04/2019

Conclusion
The paper addressed the increasingly important issue of the proliferation of discriminatory speech in the digital public sphere.It took Greece as a case study, examining the eleven most visited websites in the Spring of 2019.The quantitative and the qualitative dimensions of hate speech, as they were explored here, most likely contribute to the normalization of both symbolic and physical violence and to the further diffusion of extreme-right wing ideology.This bears significance and repercussions for the quality of democracy.This issue has become all the more topical after the COVID-19 pandemic, as fear of the 'other' has dominated public spheres across the globe, intersecting with existing preconceptions and prejudices.In particular, in Greece, but also in other countries where immigration is one of the most salient issues, the proliferation of hate speech, in the form of othering or even in the dehumanizing of certain groups is directly linked to social (in)tolerance and the public accountability of law enforcement authorities.Furthermore, in Greece, from the emergence of the crisis in 2010, until the parliamentary election of June 2023, the popularity of extreme right-wing parties in the Greek parliament remained high, as did their ideas, even after the incarceration of the infamous neo-Nazi Golden Dawn leadership.The new insights that this study brings to hate speech research concern both the quantitative aspects and the qualitative themes through which we have tried to reconstruct the everyday enunciation of hate speech and which may be generalized theoretically in future studies.
More specifically, sexism and nationalism are the two most salient categories of discriminatory speech in the Greek digital public sphere, followed by racism, anti-immigrant speech and anti-LGBTQIA+ speech.The specific character of online media outlets seems to be a differentiating factor of discriminatory speech usage.
15 In.gr, 22/05/2019 16 Markos Seferlis a comedian known for defying political correctness.17 Lifo, 23/04/2019 Anonymous online articles constitute the main conveyor belt of discriminatory speech, especially for sexism and racism, followed by politicians/political parties.This finding underlines the mainstream character of hate speech, which is fully embedded in online media.Politicians and political parties play a great role in the transmission of nationalism and anti-immigrant speech.Anti-LGBTQIA+ speech is also transmitted mostly through unnamed pieces, as well as by eponymous journalists, high visibility professions, the church/clergy and various professionals/ technocrats.
From a qualitative perspective, distinct themes that serve as meaning conveyors, as well as specific lingos have been developed for every discriminatory speech category.'serious' outlets also take part in their usage.In addition, there are certain elective affinities between different types of speech, as in the relationship between racist and anti-immigrant speech, while sexism, on the other hand, is a stand-alone type of hate speech.

Table 1 :
HSM2019, List of news websites covered.
sions, like artists and athletes (2.3 %), members or representatives of the military and the police (1.3 %), representatives of the church and members of the clergy (1.3 %) and technocrats (1.1 %) (Table

Table 7 :
Main themes per hate speech category.matterhow much a transvestite puts make-up on, no matter how hard she tries, she cannot become a woman.'15Another excerpt illustrates an attempt to normalize the mockery of trans people, even resorting to ancient Greek comedians for support: 'Aristofanes used to curse Euripides and Socrates.Can't Seferlis16 curse two transvestites?He is not cursing him.He is just describing him and other people -who are worse than what they curse -laugh at him.'17