Reacting to homophobia in a French online discussion: The fuzzy boundaries between heteronormativity and homophobia

This article analyzes how participants of an online discussion thread related to a YouTube video on homophobia expressed their opposition to homophobia. Both the video and the 403 posts in the discussion thread are in French. On the surface, the data are characterized by strong antagonism between the stances that support and those that are critical of LGBTQ persons. However, a closer look at the posts expressing a pro-LGBTQ stance reveals considerable variation among them: they range from an open deconstruction of homophobia to more ambivalent positions that draw on ideologies circulating within the heteronormative order and are naturalized in the everyday discourse of spontaneous online interactions. We analyze five categories of posts expressing different forms of pro-LGBTQ stances to highlight their fuzzy boundaries with homophobic stances. The analysis draws on argumentative discourse analysis, focusing on process types used to construct arguments and topoi, as well as deictic elements through which the authors of these posts express their distance vis-à-vis homophobia and LGBTQ persons.


Introduction
Homophobia could be succinctly defined as hatred, fear, mistrust, and/or discrimination against non-heterosexual persons (see e.g. Griffin, 2017, s.v. homophobia), rooted not only in individual thought processes but also in culture and social interactions (Herek, 2004: 11). France has taken several steps to ensure the rights and protection of LGBTQ persons: same-sex domestic partnerships have been legal since 1999, and same-sex marriage since 2013. Homophobic comments and different forms of discrimination against LGBTQ persons have been illegal since 2004, and homophobic motivation is an aggravating factor in the estimation of legal sanctions. However, a significant increase in homophobic incidents was registered in 2018, coinciding with a general increase in identity-based violence and the rejection of historically excluded individuals. One reason for the increase in reported cases of homophobia was that the victims had become less hesitant to lodge a complaint (DILCRAH, 2019). The online discussion analyzed in this article is part of the societal debate surrounding the increase in homophobic aggression.
Previous work on homophobia in France includes Provencher's (2010) analysis of media reactions to the stabbing of the Mayor of Paris in 2002, and to the action of the Mayor of Bègles in presiding over a gay marriage in 2004. According to Provencher, the French media framed the stabbing of the Paris mayor as an attack on democracy and the French Republic rather than a homophobic attack. The sending of thousands of letters to the Mayor of Bègles, in turn, illustrates the complex intersections between homophobia, parenthood, gender roles, nationalism, and anti-Semitism. Russell (2019aRussell ( , 2019b, who analyzed the links between populism and homophobia in France, argues that the populist Hommen group, which is opposed to same-sex marriage, only obliquely drew on homophobia; their main position was that of victimization, inspired by the perceived erosion of masculine supremacy (Russell, 2019a(Russell, : 56, 68, 2019b. Hugonnier (2021: 274-275) analyzed the discourses of protest movements opposing the access of lesbian couples to medically assisted reproduction. She also concludes that these movements adopted a defensive position to protect the heterosexual model of society that they thought was being threatened. At the same time, these groups tended to anticipate potential accusations of homophobia by putting forward arguments that could be used to refute such allegations.
As part of the ongoing societal debate on homophobia, the French television network France 2 broadcast a talk show entitled Homophobie -Stop à la vague de la haine ('Homophobia -Stop the wave of hate'), on 11 January 2019. The show was part of the series Ça commence aujourd'hui ('It starts today'). It featured a host interviewing two men and one woman who had been victims of homophobic aggression, a live audience, and two experts -a psychologist and a lawyer. Covering some of the most heavily publicized homophobic attacks in France in recent years, the YouTube version of the show triggered a multifaceted online discussion, which we therefore chose to analyze.
At the beginning of the show, the host announced that the topic was 'important' and 'urgent' because acts of homophobic aggression took place every 33 hours in France. The first interviewee, Bruno, was kidnapped by four men in the Parisian gay neighborhood of Le Marais, raped, and almost killed. The trial of his aggressors in 2011 was widely mediatized in France, and the homophobic motive of the crime was taken into account as an aggravating circumstance. Bruno suffered memory loss caused by a severe brain injury, hence most of the details he gave of the assault were based on the perpetrators' depositions. Brahim, the second interviewee, told the interviewer that he had been repeatedly raped by different young men in his own age group over the course of almost 10 years in his neighborhood north of Paris. According to him, the aggressors were closeted gay men; this series of acts did not lead to a lawsuit. Sandra, the third interviewee, was assaulted by three men in Strasbourg city center. She said that she looked stereotypically lesbian, she was wearing a rainbow belt, and she was just talking to her girlfriend. Although the physical assault was preceded by homophobic speech, the court did not consider the assault a homophobic crime. The victims in all these stories framed themselves as innocent parties, caught up in a series of events that they could not control. The location of the acts played an important role, and all the victims had sought a safer place in which to live after these incidents (cf. Leap, 2010: 188-189).
The show was posted on YouTube on 18 January 2019, generating a lively online discussion. 1 In this article, we analyze the argumentative and grammatical devices through which the participants of the discussion reacted to homophobia by deconstructing or either implicitly or explicitly aligning with the heteronormative order 2 . Although the discussion thread that forms our data consists of posts reacting to a video depicting homophobia in the form of physical violence, most of the participants also regard expressions of verbal violence as homophobic. Our specific goal is to shed light on the fuzzy boundaries and interlinkages between heteronormativity and homophobia by showing that anti-homophobic statements may implicitly convey heteronormative ideologies. In extreme cases, superficial anti-homophobia may even be used to perpetrate homophobia.
We define the heteronormative order as a structured constellation of discourses that convey and naturalize heteronormative ideologies in which heterosexuality is represented as a norm and a preferred form of affective and sexual behavior. The concept is related to heterosexism, namely a discriminatory attitude toward non-heterosexuals (Griffin, 2017, s.v. heterosexism), and the heterosexual matrix, namely a 'grid of cultural intelligibility' whereby a stable gender expresses a stable sex in stable bodies, the outcome of this process being the conception of desires as stable and heterosexual (Butler, 1990: 151).
Our choice of the term heteronormative order reflects recent debates in studies on gender and sexuality and in queer theory. Thus, as Bryant and Vidal-Ortiz (2008: 387-391) note in their discussion on the relations between the concepts of homophobia, heterosexism, and heteronormativity, the coinage of homophobia as a concept marked a significant transformation by gradually shifting the focus from a 'problem' caused by LGTBQ people themselves to negative reactions to LGTBQ people and sexualities. However, many scholars have criticized the concept for its overly individualistic and psychological nature, and for a lack of critical attention to the possible effects it may have. Homophobia is frequently associated with extreme violence, whereas heteronormativity and heterosexism account for 'structured anti-homosexuality' and its expression as homophobia. Thus, the concepts of heterosexism and heteronormativity 'have acted as correctives to some of homophobia's limitations' (Bryant and Vidal-Ortiz, 2008: 387-391).
Online interactions such as the discussion thread analyzed in this article pertain to everyday discourse (Maingueneau, 1991: 21) or mundane discourse (Pascale, 2019) consisting of ordinary interactions with no ties to specific institutions or practices. Everyday discourse has an important role in naturalizing ideologies, namely the shared representations produced and expressed in discourse (van Dijk, 2006) and derived from value systems, beliefs, and assumptions that are often taken for granted (Simpson, 1993: 5).
Everyday discourse is rarely performative: it does not have a direct impact on the world, unlike authoritative discourses such as legal discourse. However, the descriptive, constative utterances it contains create a reality for the felicitous enactment of performative utterances in normative discourses, such as those of law or religion (see Fish, 1980: 198-199). Consequently, it is important to analyze everyday discourse because it may naturalize hateful ideologies and ideologies opposing hate, for example.
We argue that everyday discourse is located in the interdiscursive space where the peripheries of different discursive formations meet and where their boundaries are constantly transformed (Foucault, 1972: 32, 97-99, 158-159, 172). Although seemingly neutral, everyday discourse is a central component of the interdiscursive space, which is also a melting pot for specific political, religious, or legal discourses, for example, as well as for the ideologies attached to them. Hence, the interdiscursive space is not neutral: it is a space in which ideologies and discourses are processed and transformed. In particular, the everyday, mundane discourse located mainly within the interdiscursive space has the potential to naturalize and rationalize dominant ideologies, and to give them the appearance of common sense (Simpson, 1993: 5-6) such as by normalizing hate speech (Pascale, 2019: 908-910). This, in turn, has an impact on the ways in which specific discourses 'systematically form the objects of which they speak' (Foucault, 1972: 56).
Our article is structured as follows. We present our data and the concepts operationalized as methodological tools in the analysis in the next section. Subsequently, we analyze 14 data examples, divided into five categories. Finally, in the discussion we reflect on the ideological roots of anti-homophobic arguments and the fuzzy boundaries of homophobia and heteronormativity.

Data, analytical tools, and methods
Our data, which were collected in May 2019, consist of 403 posts (23,515 words) related to the television show described above. They represent all the posts reacting to this video at that time (excluding those that had been removed by the moderator), in a single thread divided into 34 sub-threads, organized in a manner that bears some resemblance to naturally occurring oral conversations (Benson, 2017: 83). The largest sub-thread in this discussion comprises 42 posts. The way in which each sub-thread appears to the YouTube user is governed by a complex algorithm taking into account the date of publication of the posts, the number of responses, the proportion of likes to dislikes, as well as the 'reputation' of the participants in accordance with the like-dislike ratio of their posts. All posts were written in French, and they could have been written by any French-speaking person in France or elsewhere. We did not verify whether the posts were written by real persons or smartbots.
Having identified the sub-threads, we classified the data. Many discussants gave information about their gender, ethnicity, and sexual orientation in their username, the username image, or the contents of their message, but we did not quantify these categories because the identity depicted online does not necessarily reflect the identity and opinions of the person who writes the post. Instead, we based our categorization on the stance, understood as the positioning and evaluation expressed by the participants' posts in relation to the contents of their and others' posts, as well as the most prominent discourse objects in them (Du Bois, 2007: 163;Jaffe, 2009: 3;Kiesling, 2009). Key discourse objects in this discussion include homophobia, LGBTQ people, and LGBTQ sexualities and sexual practices.
Of the 403 posts, 198 were discarded because they consisted of personal communications between participants or related to discourse objects that were irrelevant to the analysis. In addition, given the goal to analyze the ways in which users reacted to homophobia, we discarded the 30 posts that clearly expressed a homophobic stance. The 175 remaining posts differ considerably. Whereas some purported to reverse homophobic arguments, for example, others directly attacked another user perceived as homophobic, or combined the condemnation of homophobia with statements implying that heterosexuality was the preferred sexual orientation thereby accusing LGBTQ persons of homophobia or, paradoxically, expressing homophobia. In other words, although all the posts expressed opposition to homophobia, some of them combined this with adherence to the heteronormative order.
To account for these differences, we analyzed the predominant argumentative features of the posts. We based our understanding of argumentation on argumentative discourse analysis (e.g. Amossy, 2012), according to which language use always has an argumentative dimension because the aim is to have at least a minimal effect on particular listeners and readers, and on reality in general. We analyzed the arguments in each post, namely statements logically leading to a conclusion via reasoning based on certain premises, as well as the topoi, namely the general principles taken for granted in a community. Topoi link the argument to its conclusion by supporting the reasoning on which it is based rather than constituting reasoning per se (Anscombre, 1995: 190-192;Wodak, 2001: 73-74). Understood in this way, topoi are based on a strong (real or imagined) ideological consensus in a group. For example, the topos of normalcy (see example 1 below) is founded on the implicit assumption that something that is common has to be accepted. Thus, a statement such as 'homosexuality is normal' constitutes an argument leading to (at least) two conclusions: (a) homosexuality has to be accepted and (b) homophobia is irrelevant.
In our analysis of the processes through which the participants constructed their arguments we draw on systemic-functional grammar (Caffarel, 2006;Halliday and Matthiessen, 2004: 168-175), which identifies six process types materialized in a clause. Although no process types appear exclusively in discriminatory discourse or its counterdiscourse 3 (such as one opposing homophobia), they are essential components of arguments and topoi, bringing into focus a specific aspect of the discourse object. In other words, process types, and especially their central elements, namely verbs, provide information about the ontological status of the object of discourse (Blumenthal, 2003).
Three process types are of particular relevance in analyzing discriminatory discourse and its counter-discourse aimed at deconstructing discrimination. Material processes construe happening, creating, changing, doing, and acting; they enable the description of typical actions and behavior among members of the discriminated group and/or of those who discriminate against it. Mental processes construe inner experiences and processes of consciousness: they directly describe the feelings and thoughts of human actors -in our data they are typically used to express the thoughts and feelings of the author of a post. In other words, mental processes facilitate the drawing of attention to a participant's inner reactions while at the same time detaching these reactions from the essence or typical activity of the group that triggers them. Relational processes describe symbolizing, having an identity, and having an attribute, and include attributive and identifying processes (Halliday and Matthiessen, 2004: 210-248). Attributive processes indicate classmembership and permit the description of the qualities and characteristics (the attributes) of a person, social group, or thing (the carrier), whereas identifying processes narrow class-membership to a class of one and allow the classification of persons and things by indicating their membership in a category. The attribute in an attributive clause is indefinite (an adjective or a noun phrase starting with an indefinite article, for example), whereas the identified element of an identifying clause is definite (often a noun phrase starting with a definite article). Another major difference is that the two parts of an identifying clause are reversible (e.g. Sarah is the director; The director is Sarah), whereas this is not the case with attributive clauses in ordinary language use (e.g. Sarah is a lesbian; *A lesbian is Sarah). By expressing what people and things do and what and how they are, material and relational processes (particularly attributive processes) strongly foreground the construal of actions and characteristics as general knowledge, without specifying the source of the information. 4 Argumentative and process-type analysis is complemented with micro-level linguistic analysis drawing on the insight that alongside the verb, pronoun choice also participates in the framing and contextualization of all interaction (Gumperz, 1992). Through such choices, participants of an online discussion can express their distance vis-à-vis the discourse objects by presenting themselves as insiders or outsiders in different processes, and by situating themselves within or outside the thematic schemata that are used for building specific arguments. Pronouns play a key role in the negotiations between distance and proximity: in a typical scenario, the first-and second-person pronouns I and you indicate presence, in-group, and/or participation, whereas third-person pronouns such as they indicate distance and out-group.
Pronouns used in online discussions show the extent to which the author assumes personal responsibility for the post and whether the post is addressed to all the other participants or just one. First-person pronouns indicate the deictic center, and both firstand second-person pronouns are inclusionary in the sense that they indicate that both the speaker and the addressee are taking part in the exchange: intersubjectivity only prevails between the I and the you (Benveniste, 1971: 223-230;Schiffrin, 2006: 106). Only the I (in French, in the subject position, je) can index the stance-taker (Du Bois, 2007: 132,152).
French has two pronouns corresponding to the English pronoun you. One of them, tu, is used only in the singular and indicates an informal and/or non-distant relationship between the interlocutors. It is also used generically, referring to any person, but our data contain very few such occurrences, presumably because such usage would create ambiguity in an online discussion with multiple participants. The generic you is absent from the examples analyzed in this article. The visual hierarchy of the thread (indentation) and/or the @ sign, followed by the addressee's username, also indicates that the writer is directly addressing another participant. 5 The other second-person pronoun, vous, can be either singular or plural. In the singular it indicates distance and formality, whereas in the plural it may imply distance and/or formality, or the lack thereof.
Third-person pronouns are exclusionary and refer to people who are not involved in the situation (Chilton, 2004: 336;Russell, 2019aRussell, : 36, 2019b. They may replace not only a proper name but also a common noun, hence utterances containing a third-person subject enable the expression of distance from the deictic center -the French pronoun on ('one') being a special case, as explained in the analysis of examples (8), (9), and (10). This distance may be neutral, but it could also relate to affective stance (Goodwin, 2007). An additional specific property of third-person pronouns is that they allow the creation of referential chains that do not name the persons being talked about, the first anaphoric usage creating a contextual model for subsequent posts (see Heritage, 1984: 242).
In the following we analyze 14 short examples of our data, representing the most typical argumentative features in each category. We explain the arguments put forth and the topoi evoked in each example, including the process types used to mobilize them, as well as salient phenomena in relation to the expression of distance. Alongside the French original, we give an English translation in italics, but we do not attempt to translate idiosyncratic features of the French source text such as typos or abbreviated and truncated words -we only comment on them when necessary. Although our aim is to translate the examples as accurately as possible, it is impossible to translate all the indexical information (Blommaert, 2006) in the posts, given that the grammar, style, and spelling differ significantly from standard French usage. These features are explained in the analysis when necessary. The translation follows the typography, punctuation, letter size, and use of smileys in the French original.

Analysis
We start the analysis with four examples of posts that openly deconstruct homophobia. Second, we consider two examples devoted entirely to an attack ad personam. Third, we examine four examples of expressions of empathy that are ambivalent in the sense that the defense of LGBTQ persons is accompanied with elements of heteronormative ideology. Fourth, we analyze two examples of posts whose authors produce reductive interpretations of LGBTQ people, and even blame them for homophobic aggressions. Fifth and finally, we scrutinize two examples representing the disclaimer 'I am not homophobic, but. . .,' whose authors explicitly state that they do not approve of all aspects of LGBTQ persons or sexualities although they claim to condemn homophobia. Therefore, the analysis moves from clearly anti-homophobic posts towards posts in which the antihomophobic opinion is no longer clear, and heteronormative ideology becomes more and more dominant. This categorization is based on statements expressed in the posts, as well as on the argumentative and linguistic tools mobilized in them.

Deconstructing homophobia and heteronormativity
Posts founded on arguments deconstructing the traditional arguments found in homophobic discourse are very common in the data (N = 72). Example (1) could thus be read as a counter-discourse, stating the opposite of the classical homophobic argument that dates back at least to Ancient Greek philosophy, according to which homosexuality is abnormal (Revol, 2003). It thereby reads as contrary to the traditional definitions of 'queer' as 'whatever is at odds with the normal, the legitimate, the dominant' (Halperin, 1995: 62). The two statements in the post are construed as attributive processes, allowing the presentation of the naturalness and normality (attribute) of homosexuality (carrier) as a universal truth. The two occurrences of the presentative construction c'est (translated as 'is') highlight the fact that the author's own point of view is erased (Modena, 2012: 301). However, through the capital letters in POINT ('period') the author marks the statement as a personal, emotional reaction indicating that the discussion is useless.
homosexuality is natural, it is normal, PERIOD.
This example mobilizes the topos of normalcy. This topos is accompanied with the topos of modernity and secularism in example (2), which mobilizes a counter-discourse similar to example (1) by characterizing homosexuality as 'just a sexual orientation' in an attributive process: the author of this post implies that the authors of homophobic posts represent the Middle Ages. The French law of 1905 separating the Church and the State and establishing the principle of secularism is used as an authoritative argument, again in an attributive process. Hence, the author implies that religious motivations automatically disqualify the argumentation in homophobic posts. In terms of author presence, this post is similar to (1): although the author is not made present through pronouns (the firstperson plural pronoun nous ['we'] refers to people in general), the exclamation marks as well as the interjection ZUT ('dang') indicate an emotional reaction of frustration.

Homosexuality is just a sexual orientation that's all! We are no longer in the Middle Ages damn it! The separation of the Church and the State was not insignificant! Everyone does what they want DANG it.
Compared to the first two examples, the counter-discourse mobilized in example (3) aligns with the tenets of traditional queer theory: it reverses the customary 'repetitions of pre-existing historicized citations' and the 'range of possible expressions of gender or sexual identity' (Barrett, 2003: 555) by presenting heterosexuality rather than homosexuality as repulsive. The author is clearly present in the statement, as indicated by the firstperson pronouns moi (emphatic) and me (direct object), and the fact that the reaction to the sight of straight people kissing is construed as a mental process, focusing on the author's inner feelings. The activity whose perception triggers this reaction is construed as a material process of doing in this post that intertextually parodies another post whose author feels uncomfortable when seeing a gay couple kissing each other (see example 13). This post also mobilizes the topos of normalcy. (

3) moi quand deux hétérosexuels s'embrassent devant moi ça me met mal à l'aise for me, when two heterosexuals kiss in front of me, it makes me uncomfortable
Claiming the right to free speech is a commonly used argument in homophobic discourse (Vernet and Määttä, 2021: 45-47). Some authors mobilize the topos of law in their counter-discourse, as in (4), whose author reacts to a previous post interpreted as homophobic by several participants. Rather than explicitly deconstructing homophobia, the writer refers to the law to show that the authoritative discourse has already done so and made it illegal. The post starts with two attributive processes containing the presentative construction c'est (translated as 'means' and 'it is'), which highlights the fact that the author's own point of view is erased (Modena, 2012: 301). However, in the following sentences, the author directly addresses another user, as indicated by the secondperson pronouns, as well as the explicit presence of the 'I' in the mental process starting the last sentence.

Attacks ad personam
The data contain 44 posts criticizing other participants for posts interpreted as homophobic. This category is characterized by verbal violence, insults, disparagement, and mockery, whereas arguments per se are missing. The stance on homophobia is somewhat ambivalent in example 5: the noun phrase là [sic] même chose ('the same thing') could refer to either LGBTQ sexuality or homophobic attacks. The author of this post reacts to another post proclaiming that LGBTQ people should not show themselves if they wished to be left alone. The wish that 'the same thing' would happen to the other participant implies that same-sex attraction is fatal, comparable to an illness or an accident. Fatality therefore emerges as the main topos mobilized in the post, and this interpretation is supported by the process type used to construe the statement: the verb faire ('do', 'make', translated as 'serve') is prototypically material, describing a change in the material world, although the change depicted in the clause Sa [sic] te ferais [sic] du bien ('It would serve you right') could also be interpreted as unfolding inertly (cf. Halliday and Matthiessen, 2004: 214).
(5) mdrr 6 mais quesque t'es conne ou con ! Sa te ferais du bien qui t'arrive là même chose sa te remettrais les idées en place ! LOLL you are really stupid! It would serve you right if the same thing happened to you that would put your thinking in its place! This category of posts contains a large quantity of second-person references. The following pronouns are used in (5) to refer to the author of a homophobic post: tu, truncated as t, ('you' [subject]) and te, also truncated as t ('you' [direct and indirect object]). The other participant is characterized as 'stupid' (conne ou con) in an attributive process constituting a face-threatening act (Goffman, 1955). The adjective shows that the gender of the attacked participant is unclear: both the feminine (conne) and masculine (con) forms of the adjective are used to mobilize the topos of ignorance.
Posts that we have categorized as personal attacks take a variety of forms. Another typical subcategory is represented in example (6) below, which also reacts to the participant who thinks that LGBTQ people should not show themselves if they wish to be left alone. Here, the attack consists of a virtual shout, and argumentation is absent.
The derogatory expression ferme ta gueule ('shut [the fuck] up') is truncated here and abbreviated as tg (i.e. ta gueule).

Ambivalent expressions of empathy
Several posts condemning homophobia draw on empathy. Nevertheless, rather than deconstructing homophobia, many of them express empathy in an ambivalent way, implicitly combining it with the othering of LGBTQ persons or arguments that are typically used in homophobic discourse or heteronormative ideology. This category is represented by nine posts in the data.
In (7), empathy is combined with the topos of fatality, a topos that was also present in example (5) above.
why do you say that they have not done it on purpose to be in love with the same sex.
The verbal phrase faire exprès ('to do something on purpose' or 'to mean to do something') carries the implication that this form of love is not the most desirable: the expression j'ai pas fait exprès ('I did not do it on purpose') is typically used when apologizing for a mistake. The choice of this verbal phrase may have been innocuous, or it may echo arguments that are found in religious discourses according to which LGBTQ sexuality is a twist of fate that can happen to anyone; hence, empathy here is close to compassion or even pity. The material process centered on the verb faire ('do') implying an active role on the part of the doer of the deed also supports this interpretation. As the second-person subject pronoun tu ('you') indicates, the author is directly addressing another participant who had posted a homophobic statement; the third-person plural pronoun ils ('they') is used with reference to LQBTQ persons, marking a distance from the deictic center of the author of the post.
In (8), the topos of empathy towards LGBTQ persons is combined with a rejection of the topos of hate and the user's emphasis on being straight: (8) Je suis choquer de voir ces commentaire haineux je suis pas lesbienne rien de tous sa mais comment on peut prendre des humains et les considèrer comme de la merde.mais quesque ces que ce monde .pourquoi tant de haine pourquoi

I'm shocked to see these hateful comments I'm not a lesbian none of all that but how can one take humans and consider them shit.what kind of world is this .why so much hate why
This post characterizes LGBTQ persons through a mental process (les considérer comme de la merde, 'consider them shit'). The author's persona is firmly expressed by the personal pronoun je ('I') in the first clause depicting her reaction in an attributive process (I'm shocked), as well as in the attributive process of the second clause expressing both her gender and her sexual orientation. The four rhetorical questions introduced by the interrogative adverbs comment ('how'), pourquoi ('why', twice), and quesque ([sic], 'what') reinforce the expression of empathy, as does the characterization of LGBTQ people as humans and the rejection of the topos of dehumanization (Oliver, 2011) in the expression like shit.
The explicit argumentation of a straight identity could be used strategically to show that the author feels empathy towards LGBTQ persons even though her own sexual orientation is straight. However, the adverbial phrase rien de tous [sic] sa [sic], ('none of all that') clearly distances the author from LGBTQ persons, emphasizes the fact that she is not a lesbian, and presents LGBTQ sexuality as undesirable.
Tolerance combined with indifference is another strong topos in this category. (9) On ne demande à personne de les aimer juste de leur foutre la paix c'est simple.
FOUTEZ LEUR JUSTE LA PAIX Nobody is asked to like/love them just leave them alone it's simple. JUST LEAVE THEM ALONE The first pronoun, on, is a peculiar feature of the French language. Its function here is generic (Fløttum et al., 2007), so that the statement is presented as the point of view of all participants ('everyone' or 'anyone'). The English translation is intended to render this meaning through the passive voice. The point of view of all or anyone is also expressed by the indefinite pronoun personne ('no-one'). The third-person plural pronouns les (direct object) and leur (indirect object) also refer to LGBTQ persons in this post. Alternative stances towards LGTBQ people are reflected in the mental process centered on the verb aimer ('like') and the ostensibly material process centered on the vulgar phrasal verb foutre la paix ('leave alone'). The upper-case letters probably index indignation, as the six sad-faced emoticons at the end of the post imply. These semiotic resources allow the author to highlight the topos of empathy, whereas the general tone of the post is governed by the topos of relevance: sexuality should not be regarded as a societal problem or a reason for discrimination (see also examples 1 and 2).
The topos of relevance is also prevalent in (10), as the user expresses bewilderment over the fact that the same standards are not used for LGBTQ and straight people.
(10) Ça me dépasse qu'on se justifie sur le fait d'être homosexuel. Personne ne le fait à un hétéro. Je vois pas pourquoi chacun ne respecte pas les choix des autres. [-] It's beyond me that one has to justify oneself for being homosexual. No one does this to a straight person. I don't see why everyone does not respect other people's choices [-] As in (7) and (8), this post is framed as a personal opinion, as the first-person pronouns indicate. Here, as in (8), the third-person pronoun on ('one') has a specific function in the sense of 'someone' or 'anyone' (Fløttum et al., 2007), and the pronoun is rendered by the English pronoun one in the translation. The comparison between straight and LGBTQ persons is highlighted by the opposite indefinite pronouns personne ('no-one') and chacun ('everyone'). The pronoun autres ('others') might appear to be used as a device to 'other' LGBTQ persons, but in fact it refers to other people in general. Nevertheless, this choice of pronoun reinforces the qualification of homosexuality as a choice (choix), like any other life choice. The desired reaction to these choices is construed through a mental process centered on the verb respecter ('respect'), and the topos mobilized by the post is similar to example (9) above (tolerance combined with indifference). As a statement, therefore, it contradicts the generally anti-homophobic stance of the post and echoes arguments according to which people decide to be attracted to persons of the same sex although they could as well 'remain' straight and comply with societal norms.

Reductive interpretations of the causes of homophobia
Many participants express outrage at the attackers and rapists portrayed in the video to which they react without expressing solidarity, but rather adopting a position of an angry spectator who tries to identify the culprit. In this section, we analyze two examples of the 46 posts in which a revolt against homophobia is combined with arguments putting the responsibility for it on LGBTQ people, or even accusing them of being homophobic. Example (11) introduces the topos of the closet, which is prevalent in several sub-threads wherein the participants wonder how people claiming to be homophobic can commit same-sex rape.

Doubly cowardly
The argument is as follows: to rape a man, a male rapist needs to be sexually attracted to men, so the rapist must be gay. The implicit presupposition is that there would be less homophobia if the aggressors assumed that they were gay. Although this argument does not call into question the seriousness of the facts, it nevertheless leads to consideration of the possibility that these aggressions take place between LGBTQ people, and that homophobia is mainly an in-group problem. As a result, the heteronormative order remains safe. The distance in relation to the author's deictic center is highlighted by pronoun usage: the author refers to the rapists discussed in the video through the noun phrase ces gens-là, expressing both deictic and affective distance. The demonstrative ce accompanied with the particle là typically adds a derogatory connotation to the reference, so that in this case this phrase indicates that the author completely rejects 'these people' and does not want to have anything to do with them. The pronoun ils ('they') refers to the same group of people anaphorically. Attributive processes are used to characterize the rapists in the first and the third clause, as well as in the last clause where the verb is omitted, whereas raping is depicted through a material process in the second clause. In both cases, the facts are presented as uncontested and universally accepted. As the author's persona is grammatically absent in this post, the final qualification of homophobic acts committed by closeted LGBTQ persons could also be attributed to anyone who agrees with the argument. The author of example (12) urges the 'gay community' to react and defend themselves. As in (11), the post foregrounds the topos of cowardice. The use of the second-person pronoun vous ('you') shows that the message is addressed to the alleged gay community. The first two sentences of the post use the imperative mode (défendez-vous 'defend yourself', soyez solidaires 'show solidarity'), which suggests a strong incentive followed by a piece of advice. As a whole, the post seeks to teach lessons to LGBTQ people: the two sentences in the imperative mode are followed by a sentence describing the mental reactions triggered by inaction. In the next sentence, the author explains what LGTBQ people should do by using nouns that refer to the material processes depicting these actions (insultes, 'insults', coups, 'blows'). In the last sentence, which sounds like a bitter statement, the lack of solidarity among LGBTQ people is represented through an existential process, although here as well, the verb is omitted. This post borders on condescension, obscures the responsibility of the aggressors, and accuses the victims of being too weak -if they defended themselves, there would be less homophobia.

'I am not homophobe, but': A rhetorical precaution
Finally, we analyze two posts using the disclaimer 'I am not racist/homophobe/sexist etc., but. . .,' which is commonly used as a preface to a prejudiced statement while at the same time highlighting the speaker's positive self-presentation (van Dijk, 1993: 29-30). This disclaimer appears in 28 posts in our data, in a variety of configurations, and it usually precedes a more or less homophobic statement. In (13), it takes the form 'LGBTQ behavior triggers a negative feeling in me, but homophobic aggressors are sick'.
(13) un couple homo qui s'embrasse devant moi me met mal alaise, mais je ne comprend pas cette haine que certains malades ont envers les homos c'est surtout les agressions qui me choque a gay couple kissing each other makes me uncomfortable, but I do not understand this hatred that some sick people have against gays it's especially the assaults that shock me The focus in this post is on the author, as exemplified by the first-person pronouns (moi, je, me ('I', 'me'). In addition, as in example (3) (14) to advance homophobia: the slur pédé ('faggot') appears in the first part, whereas the second part reveals that the author is only opposed to gay couples' adoption rights. In other words, the presence of the slur in the first part of the disclaimer disqualifies the statement that the author has nothing against gays: the post clearly expresses homophobia, although it starts with the rhetorical precaution that the author is not homophobic. This contradiction is also highlighted by the qualification of gays' sexual 'direction' as perverted and animal, and something that they should assume -in other words, accept with all its consequences, meaning that they are perverted and inhuman. Hence, the reference to animals adds a dehumanizing tone to this post.
According to Perreau (2003: 33-34), views advocating that only two parents of a different sex have the right to adopt a child are product of uncritically accepted anthropological and psychoanalytic wisdom that is naturalized as presenting cultural and biological evidence. The emphasis on adoption reflects the general tone of societal debates on samesex unions in France, focusing more on filiation than marriage (Fassin, 2009), and revealing the heterosexism that often accompanies homophobia (Fassin, 2008: 76 . This antagonism is also present in the contrast between the author's opinions and the description of LGBTQ people's actual mental state as volitional (veulent, 'want'), and their desired mental state as deontic (the subjunctive form qu'ils assument, 'may they assume').

Discussion
We have examined the argumentative and grammatical devices through which participants in a discussion thread related to a video on homophobia reacted to and opposed this phenomenon. No linguistic or argumentative features were used exclusively in the different categories of arguments opposing homophobia. However, certain links between particular argumentative strategies and linguistic tools could be identified. The first category consisted of posts drawing on the topoi of normalcy, modernity, relevance, and law to construct counter-discourses and counterarguments aimed at reversing homophobia and heteronormativity. Most posts in this category contained only minimal indicators of the presence of the author, and the statements were construed as material and attributive processes, enabling their representation as universally accepted facts that do not need to be justified. Many posts in the second category, namely that of personal attacks, were devoid of explicit argumentation, and homophobic statements in another post were attacked by resorting to hate speech.
The remaining three categories were characterized by the representation of LGBTQ persons and sexualities as distant from the deictic center of the authors. Many arguments defending LGBTQ persons were anchored in the heteronormative order, and the boundary between advocacy and marginalization of LGBTQ persons was frequently blurred. Thus, the third category consisted of ambivalent expressions of empathy, and the authors, whose open-mindedness was represented through mental processes, resorted to the topoi of tolerance and relevance and rejected the topoi of hate and dehumanization. At the same time, the topoi of fatality and indifference as well as of implicit heteronormative arguments created a contradiction between the pro-LGBTQ stance and the implicit assumption that homosexuality is a deviation from the heteronormative order. A similar contradiction was present in the fourth category, in which homophobia was represented as deriving from phenomena that are prevalent among LGBTQ persons, namely a lack of solidarity and the closet. Implicitly, these posts obscured the possibility that homophobia may be related to the heteronormative order. The last category featured users who condemned homophobia while at the same time expressing alignment with the heteronormative order. In extreme cases (example 14), tolerance of LGBTQ persons was combined with the expression of homophobia.
Our data show how legal provisions sanctioning homophobia are also reflected in everyday discourse: only 30 of the 403 posts openly expressed homophobia, and the posts analyzed in this article exemplify the ways in which ordinary people condemn it. However, the analysis also shows that there is considerable tension among the different views on homophobia, and that opposition is frequently contingent upon the ideologies circulating within the heteronormative order. Homophobia and heteronormativity are clearly different phenomena, exemplified by the fact that one can oppose homophobia while at the same time endorsing the heteronormative order. However, the heteronormative order makes homophobia possible, justifies it, and fosters its circulation.
Due to space constraints, we were not able to analyze all the relevant phenomena in these data, such as the interaction patterns among the discussion participants. In addition, because we focused on the links between the micro-level phenomena in the data and the discursive and ideological dimension, we were not able to examine the links between homophobia and current societal debates about gender. However, we hope we have shown the pervasiveness of the heteronormative order in the everyday discourse of spontaneous online interactions -which constitute an important site of everyday discourse in contemporary society. As the ideologies circulating within the heteronormative order are naturalized, their ideological character remains largely hidden, and difference automatically emerges as a defining characteristic of LGTBQ persons and their affective and sexual lives. Furthermore, given that the heteronormative order is deeply internalized, it is difficult to construct arguments that are not limited by it, and to use them successfully in a discussion. In fact, the first category of arguments, consisting of categorical statements presenting the counter-ideology as a universal truth, constitutes the best example of attempts to deconstruct the heteronormative order.

Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.